Corium Processing

Nuclear Safety and Regulation
NEA/CSNI/R(2016)15
State-of-the-Art Report on Molten Corium

Concrete Interaction and Ex-Vessel Molten
Core Coolability
©OECD 2017
NEA No. 7392
NUCLEAR ENERGY AGENCY

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT
The OECD is a unique forum where the governments of 35 democracies work together to address the economic,
social and environmental challenges of globalisation. The OECD is also at the forefront of efforts to understand
and to help governments respond to new developments and concerns, such as corporate governance, the
information economy and the challenges of an ageing population. The Organisation provides a setting where
governments can compare policy experiences, seek answers to common problems, identify good practice and
work to co-ordinate domestic and international policies.
The OECD member countries are: Australia, Austria, Belgium, Canada, Chile, the Czech Republic,
Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, Korea,
Latvia, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia,
Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States. The European Commission
takes part in the work of the OECD.
OECD Publishing disseminates widely the results of the Organisation’s statistics gathering and research on
economic, social and environmental issues, as well as the conventions, guidelines and standards agreed by its
members.
NUCLEAR ENERGY AGENCY

The OECD Nuclear Energy Agency (NEA) was established on 1 February 1958. Current NEA membership
consists of 33 countries: Argentina, Australia, Austria, Belgium, Canada, the Czech Republic, Denmark, Finland,
France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, the
Netherlands, Norway, Poland, Portugal, Romania, Russia, the Slovak Republic, Slovenia, Spain, Sweden,
Switzerland, Turkey, the United Kingdom and the United States. The European Commission also takes part in
the work of the Agency.
The mission of the NEA is:
– to assist its member countries in maintaining and further developing, through international co-operation, the
scientific, technological and legal bases required for a safe, environmentally sound and economical use of
nuclear energy for peaceful purposes;
– to provide authoritative assessments and to forge common understandings on key issues as input to
government decisions on nuclear energy policy and to broader OECD analyses in areas such as energy and
the sustainable development of low-carbon economies.
Specific areas of competence of the NEA include the safety and regulation of nuclear activities, radioactive
waste management, radiological protection, nuclear science, economic and technical analyses of the nuclear fuel
cycle, nuclear law and liability, and public information. The NEA Data Bank provides nuclear data and computer
program services for participating countries.
This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the
delimitation of international frontiers and boundaries and to the name of any territory, city or area.
Corrigenda to OECD publications may be found online at: www.oecd.org/publishing/corrigenda.

© OECD 2017
You can copy, download or print OECD content for your own use, and you can include excerpts from OECD publications, databases and multimedia products in your
own documents, presentations, blogs, websites and teaching materials, provided that suitable acknowledgement of the OECD as source and copyright owner is given.
All requests for public or commercial use and translation rights should be submitted to neapub@oecd-nea.org. Requests for permission to photocopy portions of this
material for public or commercial use shall be addressed directly to the Copyright Clearance Center (CCC) at info@copyright.com or the Centre français d'exploitation
du droit de copie (CFC) contact@cfcopies.com.
COMMITTEE ON THE SAFETY OF NUCLEAR INSTALLATIONS
The Committee on the Safety of Nuclear Installations (CSNI) is responsible for NEA programmes and
activities that support maintaining and advancing the scientific and technical knowledge base of the safety
of nuclear installations.
The Committee constitutes a forum for the exchange of technical information and for collaboration
between organisations, which can contribute, from their respective backgrounds in research, development
and engineering, to its activities. It has regard to the exchange of information between member countries
and safety R&D programmes of various sizes in order to keep all member countries involved in and abreast
of developments in technical safety matters.
The Committee reviews the state of knowledge on important topics of nuclear safety science and
techniques and of safety assessments, and ensures that operating experience is appropriately accounted for
in its activities. It initiates and conducts programmes identified by these reviews and assessments in order
to confirm safety, overcome discrepancies, develop improvements and reach consensus on technical issues
of common interest. It promotes the co-ordination of work in different member countries that serve to
maintain and enhance competence in nuclear safety matters, including the establishment of joint
undertakings (e.g. joint research and data projects), and assists in the feedback of the results to participating
organisations. The Committee ensures that valuable end-products of the technical reviews and analyses are
provided to members in a timely manner, and made publicly available when appropriate, to support broader
nuclear safety.
The Committee focuses primarily on the safety aspects of existing power reactors, other nuclear
installations and new power reactors; it also considers the safety implications of scientific and technical
developments of future reactor technologies and designs. Further, the scope for the Committee includes
human and organisational research activities and technical developments that affect nuclear safety.
NEA/CSNI/R(2016)15
Foreword
Since 2002, the Nuclear Energy Agency (NEA) has sponsored the Melt Coolability and Concrete
Interaction (MCCI) co-operative project in two phases in order to investigate ex-vessel melt
coolability and concrete interaction by means of separate-effects tests and large-scale integral tests
carried out at the Argonne National Laboratory (ANL). The second phase of the experimental
programme was completed in March 2010. Key elements of this programme included the conduct of
experiments involving real reactor material and associated analyses with the objectives of resolving
the ex-vessel debris coolability issues and addressing remaining uncertainties related to long-term two-
dimensional molten core-concrete interactions under both wet and dry cavity conditions. It was
expected that the achievement of these two objectives would demonstrate the efficacy of severe
accident management guidelines for existing plants and provide the technical basis for better
containment designs for the future plants. During both phases of the programme, a total of 13 separate
effects tests were conducted to provide data on various core debris cooling mechanisms, including two
specific tests to study mechanisms connected to new design features to enhance coolability by bottom
water injection. In addition, six large-scale integral tests were conducted to provide data on long-term
two-dimensional core-concrete interaction under both wet and dry cavity conditions and one specific
one-dimensional large scale test was conducted to investigate the approach for augmenting ex-vessel
corium coolability based on an externally-cooled surface concept. These tests provided a broad
database to support the development and validation of models and codes with an aim to assess the
behaviour of the full size plants under given conditions.
The Committee on the Safety of Nuclear Installations (CSNI) held a seminar in November 2010
where the major outcomes of the MCCI Project and other complementary experimental activities were
presented and discussed.
One of the recommendations of this wrap-up seminar was: “The preparation of a state-of-the-art
report on melt coolability and core concrete interactions that captures the last thirty years of
international research results”. A consensus was reached to start with the preparation of the
recommended report. A proposal was discussed within the framework of the Working Group on Risk
Assessment (WGAMA) and endorsed by the CSNI in late 2011.
The mandate and the need for a comprehensive state-of-the-art report (SOAR) can be illustrated
by the number of questions raised and discussed during MCCI Project and WGAMA meetings. These
questions are:
• What can be learnt from the experimental data base including the last results of the MCCI
Project and complementary national or European projects?
• What progress has been made and what is the level of remaining uncertainties on the
modelling of corium concrete interaction and molten core coolability?
• What could be concluded about the capabilities of the codes to predict corium concrete
interaction and molten core coolability with respect to containment integrity assessment in
plant application?
STATE-OF-THE-ART REPORT ON MOLTEN CORIUM CONCRETE INTERACTION AND EX-VESSEL MOLTEN CORE COOLABILITY, NEA No. 7392, © OECD 2017 3
NEA/CSNI/R(2016)15
• What are the remaining issues and opportunities to define further experimental or analytical
activities?
To address these questions, a working group was set up and a kick-off meeting took place in
April 2012.
After five working meetings, the present state-of-the-art report is the result of the work of a group
which was co-ordinated by Jean-Michel Bonnet (IRSN, France) and included representatives from
Japan, Italy, France, Germany, Russia, Spain and the United States. The technical secretariat was
carried out by Martin Kissane (Nuclear Energy Agency).
The contributors to this report were:
Fernando Robledo CSN, Spain (Chapter 1 leader)
Mitchel T. Farmer ANL, United States (Chapter 2 leader)
Michel Cranga IRSN, France (Chapter 3 leader)
Claus Spengler GRS, Germany (Chapter 4 leader)
Didier Vola IRSN, France (Chapter 5 leader)
Sudhamay Basu NRC, United States (Chapter 6 leader)
Kresna Atkhen EDF, France
André Fargette Areva, Germany
Manfred Fisher Areva, Germany
Jerzi Foit KIT, Germany
Akitoshi Hotta NRA, Japan
Christophe Journeau CEA, France
Cathy Marchetto IRSN, France
Akinobu Morita NRA, Japan
Evgeny Moiseenko IBRAE, Russia
Franco Polidoro RSE, Italy
Quan Zhou FAI, United States
ACKNOWLEDGEMENTS
Within the realm of past work, several synthesis papers exist in the open literature but only one state-
of-the-art report (SOAR) was published in 1995 (Alsmeyer H. , et al., 1995). This 1995 SOAR was
prepared in the framework of the specific European programme, ”NUCLEAR Fission Safety 1990-
1994”, of the European Atomic Energy Community Reinforced Concerted Action on Reactor Safety.
The MCCI SOAR working group also acknowledges the benefit of a European activity conducted
in parallel that aimed to write a report on knowledge improvement based on the work performed on
corium concrete interaction between 2004 and 2012 under the framework of SARNET.
4 STATE-OF-THE-ART REPORT ON MOLTEN CORIUM CONCRETE INTERACTION AND EX-VESSEL MOLTEN CORE COOLABILITY, NEA No. 7392, © OECD 2017
NEA/CSNI/R(2016)15
Table of contents
List of abbreviations and acronyms....................................................................................................15

Executive summary ..............................................................................................................................21
Lessons learnt from experiments.....................................................................................................22
Capabilities and shortcomings in current codes and models...........................................................25
Remaining issues in the areas of MCCI and debris coolability ......................................................26
Perspectives .....................................................................................................................................32
1. Introduction to the main phenomena involved: Brief description
of accident phenomenology ............................................................................................................33
1.1. The concept of melt stabilisation and severe accident termination..........................................33
1.2. Corium discharge from the RPV into the containment ............................................................34
1.3. Corium concrete interaction .....................................................................................................36
1.4. Relevant containment failure modes due to ex-vessel corium behaviour ................................48
1.5. Severe accident management issues and back-fitting measures ..............................................56
1.6. Prioritisation of severe accident research issues ......................................................................62
2. Molten core-concrete interaction and debris coolability ...........................................................65
2.1. Overview ..................................................................................................................................65
2.2. Database summary ...................................................................................................................66
2.3. Dry cavity MCCI experiments .................................................................................................72
2.4. Flooded cavity MCCI experiments ........................................................................................108
2.5. The Chernobyl accident .........................................................................................................129
2.6. Lessons learnt from molten core-concrete interaction experiments.......................................134
3. Major MCCI simulation tools and models ...............................................................................142
3.1. Modelling approach ...............................................................................................................142
3.2. Overall description of codes ..................................................................................................143
3.3. Comparison and synthesis of code modelling........................................................................164
3.4. Code application limitations ..................................................................................................183
3.5. Summary of overall code qualities and limitations ................................................................185
4. Status of validation and model uncertainties ............................................................................187
4.1. General remarks on the use of data from integral experiments for code validation ..............187
4.2. Considerations on the energy balance in MCCI ....................................................................187
4.3. Identification of phenomenological models to be validated and useful experiments ............191
4.4. Code validation status ............................................................................................................195
4.5. Material properties validation status ......................................................................................204
NEA/CSNI/R(2016)15
4.6. Scaling to reactor case ...........................................................................................................205

4.7. Summary of the codes’ validation status ...............................................................................208
4.8. Conclusions ............................................................................................................................211
5. Plant applications ........................................................................................................................213
5.1. Introduction ............................................................................................................................213
5.2. Plant safety requirements .......................................................................................................213
5.3. Reactor containment designs .................................................................................................219
5.4. Severe accident management guidance (SAMG) ..................................................................231
5.5. Plant applications of MCCI ...................................................................................................236
6. Summary and recommendations ...............................................................................................261
6.1. Long term core-concrete interaction behaviour .....................................................................263
6.2. Realistic plant simulations .....................................................................................................267
6.3. Coolability enhancement under top flooding conditions .......................................................271
7. Appendices ...................................................................................................................................273
7.1. Appendix on detailed code descriptions ................................................................................273
7.2. Appendix on comparison of code validation status and uncertainties ...................................312
7.3. Appendix on SEABOR crust test ...........................................................................................334
8. References ....................................................................................................................................341
List of figures:
1.2-1: Exemplary paths of main severe accident phenomena in the cavity .................................. 35
1.3-1: Typical corium–concrete pseudo-binary phase diagram calculated with
GEMINI2 and NUCLEA07 (for the silica-rich concrete used in VULCANO) ................ 38
1.3-2: Traditional view of CCI with conduction-limited upper crust at melt water
interface; CCI with water ingression and melt eruption cooling mechanisms ................... 46
1.3-3: Extensive crack structure in the mid-plane sample of the SSWICS-3 crust ...................... 48
1.4-1: Example of drywell head flange connection details ........................................................... 50
1.4-2: Drywell head flange area versus pressure .......................................................................... 50
1.4-3: Comparison of drywell pressures for selected cases .......................................................... 55
1.5-1: Different basic measures for retention and coolability....................................................... 61
2.3-1: Elements of the BETA test facility..................................................................................... 73
2.3-2: Section through the test crucible for BETA test V 1.8 ....................................................... 75
2.3-3: Temperature of the melt pool for BETA test V 1.8 ............................................................ 75
2.3-4: COMET-L2 test apparatus ................................................................................................. 78
2.3-5: Oxide melt temperature evolution in COMET-L2 ............................................................. 79
2.3-6: Axial and radial concrete erosion by the metal melt phase in COMET-L2 ....................... 79
2.3-7: Post-test debris distribution for COMET-L2 ...................................................................... 80
2.3-8: Concrete erosion by steel and by U02-Zr02 slugs.............................................................. 82
2.3-9: SURC-4 experiment apparatus ........................................................................................... 83
2.3-10: SURC-4 Axial erosion front location and melt temperature .............................................. 84
2.3-11: SURC-1and SURC-2 axial erosion front location and melt temperature ........................... 85
NEA/CSNI/R(2016)15
2.3-12: Schematic and photograph of the ACE MCCI test apparatus ............................................ 86
2.3-13: ACE phase C experiments ablation front locations and melt temperatures ....................... 87
2.3-14: Mass fraction of major fission products and U released as aerosols
during the ACE tests........................................................................................................... 88
2.3-15: Post-test debris configuration for MOCKA test 1.3 ........................................................... 89
2.3-16: Cavity erosion behaviour and post-test debris for MOCKA 1.7 test.................................. 89
2.3-17: Bottom view of MOCKA test section prior to concrete, and views of
post-test debris for MOCKA 3.1 with rebar and MOCKA 3.3 ......................................... 90
2.3-18: Schematic of HECLA test facility; photograph of the HECLA-5 concrete crucible ......... 91
2.3-19: Concrete ablation in the HECLA experiments ................................................................... 92
3-20: Schematic and photograph of the SICOPS test facility ...................................................... 93
2.3-21: Post-test debris configurations for oxidic melt and a mixed oxide-metal melt tests .......... 94
2.3-22: Concrete erosion velocity versus power level for oxide and mixed oxide-metal
SICPOS tests ...................................................................................................................... 95
2.3-23: Concrete crucibles pre-test showing locations of embedded ZrO2 samples ..................... 95
2.3-24: Images of SICOPS test 4 with wo fully exposed ZrO2 bottom samples embedded
in EPR™ sacrificial concrete ............................................................................................. 96
2.3-25: Schematic of VULCANO test facility................................................................................ 96
2.3-26: Post-test cavity erosion profiles for the VB-U7 and VBES-U2 experiments..................... 99
2.3-27: Radial-axial power split for oxidic corium VULCANO experiments................................ 99
2.3-28: Composite SEM micrograph from test VB-ESU2 (lower part of the bulk melt) ............. 100
2.3-29: Photograph of large metallic structure recovered from test VBS-U4 .............................. 101
2.3-30: Key features of the CCI test facility ................................................................................. 102
2.3-31: Side and top views of the CCI test section ....................................................................... 103
2.3-32: Average melt temperatures for the CCI tests ................................................................... 104
2.3-33: Lateral and axial cavity erosion rates for CCI tests .......................................................... 105
2.3-34: Axial debris morphology for siliceous concrete tests....................................................... 108
2.4-1: Dimensions of the COMET-L3 test section ..................................................................... 110
2.4-2: Measured melt surface temperatures for COMET-L3...................................................... 111
2.4-3: Axial and radial concrete erosion data for COMET-L3 ................................................... 112
2.4-4: COMET-L3 post-test debris configuration ...................................................................... 112
2.4-5: ECOKATS-2 during spreading and after surface flooding the spreading
area is 2 m x 2 m .............................................................................................................. 113
2.4-6: COTELS lava M and lava B test facilities ....................................................................... 114
2.4-7: Concrete crucible dimensions and TC locations for test series D .................................... 114
2.4-8: Temperature histories in concrete sidewalls for test D-2 with water injection
and test D-8a with a dry cavity ......................................................................................... 116
2.4-9: Vertical cross section after test D-8a ................................................................................ 117
2.4-10: Vertical cross section after test D-9 ................................................................................. 117
2.4-11: Axial ablation for tests D-8a and D-9, illustrating the effect of a mortar layer................ 118
2.4-12. Mass of particulate formed before water addition vs. input energy for
the B/C test series ............................................................................................................. 118
NEA/CSNI/R(2016)15
2.4-13: Mass of particulate formed before water addition vs. concrete ablation depth
for the B/C test series ....................................................................................................... 119
2.4-14: MACE M0 Axial Ablation Data and Melt Temperature .................................................. 121
2.4-15: MACE M0 Debris-water heat flux and post-test debris configuration ............................ 121
2.4-16: MACE M3b axial ablation data and melt temperature ..................................................... 123
2.4-17: MACE M3b debris-water heat flux and post-test debris configuration ........................... 123
2.4-18: MACE M3b upper surface debris as-found and after removal of particle bed
and volcanic structures ..................................................................................................... 124
2.4-19: Debris-water heat fluxes for late-flooded tests CCI-1 through -3, and
for early flooded test CCI-6.............................................................................................. 126
2.4-20: Average melt temperature for CCI-1 through -3 and local melt temperatures
measured across the extent of the debris zone for CCI-6 ................................................. 127
2.4-21: Photographs of CCI-6 melt eruption particle bed and fractured crust debris regions
recovered from CCI-6 ...................................................................................................... 128
2.5-1: Top viewof under-reactor room 305/2 ............................................................................. 130
2.5-2: Repartition of corium on the floor of the reactor cavity ................................................... 130
2.5-3: Corium samples (TOP 1-4: brown ceramic from the large vertical flow;
BOTTOM 1-4: black ceramic from room 304/3) ............................................................. 131
2.5-4: Schematic illustrating vertical corium flows from reactor cavity .................................... 132
2.5-5: Cross-section illustrating the cavity ablation profile due to MCCI ................................. 133
2.5-6: Synthesis of ablation profiles along 11 section axes ........................................................ 133
2.5-7: Cross-sectional view at the K + 2 m axis location; Significant lateral ablation
of ~ 3 m is noted. Vertical ablation: ~ 0.5 m below the original cavity, ~1.1 m
in the lateral tongue .......................................................................................................... 134
3.1-1: Application of the lumped parameter approach to a MCCI situation............................... 143
3.2-1: Major phenomena modelled in COCO ............................................................................. 144
3.2-2: Assumptions for direction of heat conduction and ablation ............................................. 144
3.2-3: Example of melt-pool configuration in CORCON code .................................................. 145
3.2-4: Cavity surface computation in CORCON code................................................................ 145
3.2-5: Assumed core/concrete interface structure when slag continuously drains
through crust and when slag is retained beneath crust ..................................................... 148
3.2-6: Sketch of the COSACO geometry model......................................................................... 150
3.2-7: Node displacement at the ablation front ........................................................................... 152
3.2-8: Sketch of pool configuration and of heat transfer mechanisms
described in MEDICIS ..................................................................................................... 152
3.2-9: Schematic view of the liquid corium pool with convection due to gas bubbling
A crust exists at the interface ........................................................................................... 155
3.2-10: Sketch of the main models of the WECHSL code ........................................................... 158
3.3-1: Evolution of heat transfer coefficient versus superficial gas velocity Jg
in the reactor case: Comparison between Greene’s correlation and new correlation
versus corium viscosity .................................................................................................... 165
3.3-2: Mixing rate % versus gas superficial velocity in a BALISE test ..................................... 166
3.3-3: Flow map in a MCCI code ............................................................................................... 167
NEA/CSNI/R(2016)15
3.3-4: Pool/concrete interface assuming a corium crust in thermodynamic equilibrium

(‘equilibrium crust model’) .............................................................................................. 170
3.3-5: Pool/concrete interface structure with deviation from thermal equilibrium
in MEDICIS code (‘non-equilibrium crust model’) ......................................................... 171
3.3-6: Pool/concrete interface structure (‘no crust model’) ........................................................ 172
3.3-7: Liquefaction curves of some tested (and French PWR) limestone
common sand concrete mortars (NUCLEA calculations) ................................................ 181
3.3-8: Liquefaction curves of some French PWR siliceous concrete mortars
(NUCLEA calculations) ................................................................................................... 183
4.2-1: Heat and mass transfer processes at the outer boundary and internal interfaces
of the MCCI pool (Spengler, 2013).................................................................................. 188
4.2-2: Pool temperatures and energy balances calculated with MEDICIS for the
experiment CCI2 and a reactor case ................................................................................. 189
4.2-3: Transient effect in MEDICIS calculations for the experiment CCI-2 due to
variation of initial temperature ......................................................................................... 190
4.2-4: Slightly faster initial ablation velocity due to transient loss of melt superheat in the
calculation with elevated initial temperature.................................................................... 190
4.4-1: CCI-2 blind benchmark: Melt temperature versus time ................................................... 198
4.4-2: CCI-2 blind benchmark: Ablation depth versus time ....................................................... 199
4.4-3: CCI-2 blind benchmark: Final cavity contours ................................................................ 199
4.4-4: COMET-L2 benchmark: CEA and IRSN presented a base calculation and
calculations with a modified model for a better agreement with the
experimental results.......................................................................................................... 200
4.4-5: COMET-L3 benchmark ................................................................................................... 201
4.4-6: Crucible shapes at the end of calculation for VB-U5 ....................................................... 202
4.4-7: Pool temperatures versus time for VB-U5 ....................................................................... 203
4.4-8: Final cavity shape for VB-U6: Experimental and calculated profiles .............................. 203
4.6-1: Decay power in the oxide and metal phases after start of the MCCI
in the reference reactor case ............................................................................................. 205
4.6-2: Mapping of experimental data of oxide experiments to a selected BWR
reactor scale on the basis of an ASTEC/MEDICIS calculation ...................................... 206
5.3-1: Generic containment designs. PWR900, PWR1300 ........................................................ 220
5.3-2: Containment sectional views: PWR900-CP0 Fessenheim, PWR900-CP0 Bugey,
PWR900-CP, PWR1300MWe-P4 and PWR1450MWe-N4 ............................................ 220
5.3-3: Typical EDF PWR900 reactor cavity ............................................................................... 221
5.3-4: Modifications implemented for the Fessenheim plants .................................................... 221
5.3-5: BWR Mark I containment design (3D cutaway and containment simplified
schematics from NRC ...................................................................................................... 222
5.3-6: BWR Mark II design (3D cutaway and containment schematics from NRC ................... 223
5.3-7: Different pedestal region designs ..................................................................................... 223
5.3-8: 3D cutaway of BWR Mark III containment ..................................................................... 224
5.3-9: BWR Mark III design (containment schematics from (NRC, 2011) ................................ 224
5.3-10: Cross-sectional view (schematic) of the lower part of the Konvoi containment.............. 225
NEA/CSNI/R(2016)15
5.3-11: Main components of the EPR core melt stabilisation system .......................................... 227
5.3-12: ESBWR SA design features including the BIMAC device.............................................. 228
5.3-13: EU-ABWR ex-vessel retention device ............................................................................. 229
5.3-14: VVER-1000 and VVER-1200 core-catcher devices ........................................................ 230
5.3-15: Description of the Passive Ex-vessel corium retaining and cooling system
of the EU-APR1400 ......................................................................................................... 231
5.5-1: MCCI scenario with a localised corium pouring and accumulation in reactor pit,
without or with water ....................................................................................................... 238
5.5-2: Corium transfer from reactor pit to other compartments through a breach or a gate ....... 238
5.5-3: Ablation kinetics assuming a homogeneous pool, LCS and siliceous concretes ............. 239
5.5-4: Ablation kinetics with possible pool stratification: LCS and siliceous
concretes from Cranga, et al., 2010 .................................................................................. 239
5.5-5: Cavity shapes with possible stratified pool for LCS after 4 days of interaction
and siliceous concretes from Cranga, et al., 2010 ............................................................ 240
5.5-6: Evolution of cavity boundary profile versus time in INRNE calculation
(MEDICIS\ASTECv2) in dry conditions from Gencheva, et al., 2012 ............................ 241
5.5-7: Mass of the metal layer in dry conditions from Gencheva, et al., 2012 ........................... 241
5.5-8: Axial ablation in dry conditions from Gencheva, et al., 2012 .......................................... 242
5.5-9: Lower head, lower core support plate and RPV-bottom .................................................. 243
5.5-10: Erosion profiles for a typical LB-LOCA scenario with assumed layered
and mixed melt configurations ......................................................................................... 244
5.5-11: Ablation front progression and RPV-bottom failure for various
outflow sequences ............................................................................................................ 245
5.5-12: Calculated Oxide melt compositions (in mol%) at the end of the temporary
retention in the pit ............................................................................................................ 246
5.5-13: Calculated axial and radial ablation for an ESBWR transient.......................................... 248
5.5-14: Non-condensable and steam production from core-concrete interactions
for an ESBWR transient ................................................................................................... 248
5.5-15: Calculated axial and radial ablation for an ESBWR transient where GDCS
deluge is successful .......................................................................................................... 248
5.5-16: Non-condensable and steam production from core-concrete interactions
for an ESBWR transient where GDCS deluge is successful ............................................ 249
5.5-17: Calculated axial and radial ablation for an AP1000 transient case
of limestone concrete........................................................................................................ 250
5.5-18: Calculated axial and radial ablation for an AP1000 transient case
of basaltic concrete ........................................................................................................... 250
5.5-19: Calculated axial and radial ablation for an AP1000 transient for a rapid relocation
sensitivity case with partially flooded cavity ................................................................... 251
5.5-20: Containment basemat Discretisation approach for CORQUENCH analysis ................... 253
5.5-21: MAAP and MELCOR core debris distributions at the end of the spreading phase ........ 254
5.5-22: Cross sections of containment ablation ............................................................................ 254
5.5-23: MELCOR cavity profile for cross section A and B after 1 hour of CORQUENCH
simulation time and end of simulation ............................................................................. 255
NEA/CSNI/R(2016)15
5.5-24: MAAP cavity profile for cross section A and B after 60 min. of CORQUENCH
simulation time and end of simulation ............................................................................. 256
5.5-25: Principal set-up of the event tree (GRS, 2002)................................................................. 257
5.5-26: Maximum vertical erosion calculated with MELCOR ..................................................... 258
7.1-1: Model of melt and concrete interface ............................................................................... 273
7.1-2: An example of phase diagram in COCO code ................................................................. 274
7.1-3: Value of internal Rayleigh number versus pool Height ................................................... 284
7.1-4: Diagram of crust temperature profile ............................................................................... 293
7.1-5: Pool/concrete interface structure in case of model 1 ........................................................ 297
7.1-6: Pool/concrete interface structure in case of model 2 ........................................................ 298
7.1-7: Diagram of crust modelling.............................................................................................. 302
7.2-1: Stratification of molten oxide and metal after test D-6 of COTELS project .................... 312
7.2-2: Comparison of ablation depth in D-6 test of COTELS project ........................................ 313
7.2-3: Ablation profile after CCI-3 test in NEA MCCI Project .................................................. 313
7.2-4: COSACO-validation: Comparison of calculated and measured evolution
of the concrete ablation .................................................................................................... 314
7.2-5: MAAP-validation: Comparison with ablation depth data for ACE-L2 ............................ 315
7.2-6: MAAP-validation: Comparison of concrete ablation depths for CCI tests ...................... 316
7.2-7: Proposal for the usage of effective heat transfer coefficients in
MEDICIS by GRS ............................................................................................................ 317
7.2-8: Ratio of lateral to axial convective heat transfer coefficients versus
pool viscosity deduced from the small-scale CLARA tests ............................................. 318
7.2-9: 2D cavity profile in VULCANO–VBU5. Results obtained with TOLBIAC-ICB,
base case and with a ratio 3 between the radial heat transfer coefficient and
the axial one Comparison with the experiment ................................................................ 319
7.2-10: Melt temperature in CCI-3 Test ....................................................................................... 319
7.2-11: Phase diagram and calculated corium/concrete condition................................................ 320
7.2-12: COSACO-validation: Comparison of calculated and measured
evolution of the melt temperature .................................................................................... 320
7.2-13: MAAP-validation: Comparison with temperature data .................................................... 321
7.2-14: MAAP-Validation: Comparison of melt temperatures for CCI tests ............................... 321
7.2-15: MEDICIS calculation for CCI-2 by IRSN ...................................................................... 322
7.2-16: MEDICIS calculation for CCI-3 by IRSN ....................................................................... 322
7.2-17: TOLBIAC-ICB calculation for CCI-2 by CEA................................................................ 323
7.2-18: TOLBIAC-ICB calculation for CCI-3 by CEA................................................................ 323
7.2-19: MAAP-Validation: Comparison of heat fluxes from corium to water for CCI tests........ 325
7.3-1: Outline of facility ............................................................................................................. 335
7.3-2: Heat transfer surface and test section tank ....................................................................... 335
7.3-3: Gap pattern of crust block and setting in test section water tank ..................................... 337
7.3-4: Dependence of measured heat flux on gap geometry ....................................................... 339
NEA/CSNI/R(2016)15
List of tables:
1.3-1: Typical ex-vessel corium composition for BWR ............................................................... 37
1.3-2: Molar U/Zr ratio for some LWRs ..................................................................................... 37
1.3-3: Order of magnitudes of concrete constituent proportions ................................................. 39
1.3-4: Chemical compositions of typical concretes used in LWR (values in w %) .................... 40
1.3-5: Solidus and liquidus temperatures for some typical concretes ......................................... 41
1.3-6: Initial Conditions, reactor Pit Geometry and evolution of the decay power..................... 44
1.3-7: Summary of coolability mechanisms observed in MACE integral tests .......................... 47
1.4-1: Timing of key events for MELCOR calculations ............................................................. 54
1.5-1: Input parameters varied in the 2D initial and boundary condition sensitivity studies ...... 59
1.5-2: Input parameters varied in the 2D modelling phenomena sensitivity studies................... 59
2.2-1: Dry cavity MCCI and aerosol generation integral experiments and accidents ................. 66
2.2-2: Flooded cavity debris coolability integral experiments .................................................... 67
2.2-3: Simulant/separate effect experiments on MCCI behaviour, properties,
and debris coolability ........................................................................................................ 67
2.3-1: BETA Series I experiments .............................................................................................. 74
2.3-2: BETA Series II experiments ............................................................................................. 77
2.3-3: ACE Phase C test matrix .................................................................................................. 87
2.3-4: Main parameters and results for the HECLA experiments ............................................... 92
2.3-5: Concrete compositions for the VULCANO experiments ................................................. 97
2.3-6: VULCANO 2D oxidic corium test matrix ........................................................................ 98
2.3-7: Concentration of volatile species in VULCANO concretes ............................................. 99
2.3-8: VULCANO 2D metal-oxide corium test matrix ............................................................. 100
2.3-9: Specifications for CCI tests ............................................................................................ 103
2.3-10: Lateral/axial ablation rate and power split estimates for CCI tests................................. 107
2.4-1: Parameter ranges for the COTELS experiments ............................................................. 115
2.4-2: Conditions for tests D-2, D-8a and D-9 .......................................................................... 116
2.4-3: MACE test specifications (Farmer, Kilsdonk, & Aeschlimann, 2009)........................... 120
2.5-1: Average elemental composition (wt%) of corium samples. Light elements
(mostly oxygen) have been omitted ................................................................................ 131
2.5-2: Materials present in the reactor cavity at the beginning of corium formation and
incorporated during the MCCI ........................................................................................ 132
2.5-3: Summary of vertical and lateral ablation measurements ................................................ 134
3.3-1: Status on thermal-hydraulics modelling ......................................................................... 169
3.3-2: Status on pool/concrete interface modelling ................................................................... 174
3.3-3: Status on coolability models ........................................................................................... 176
3.3-4: Methods available for evaluating in MCCI codes thermochemistry data
during MCCI ................................................................................................................... 178
3.3-5: Methods available for evaluating in MCCI codes concrete ablation temperature
and enthalpy .................................................................................................................... 179
3.3-6: Features of LCS concretes .............................................................................................. 181
NEA/CSNI/R(2016)15
3.3-7: Features of siliceous concretes ......................................................................................... 181

3.5-1: Overall evaluation of main code qualities and limitations ............................................... 185
4.3-1: List of phenomenological models in MCCI codes (i.e. basic capabilities) that can be
validated by integral MCCI experiments ......................................................................... 192
4.3-2: Integral MCCI experiments proposed for validation of specific models ......................... 193
5.2-1: Refined description of levels of DiD from (WENRA, 2014) ........................................... 214
5.4-1: Severe accident challenges and mitigation strategies for a PWR..................................... 235
5.5-1: Comparison of melt-through delays calculated without and with top water injection
from (Gencheva, et al., 2012) ........................................................................................... 242
5.5-2: MELCOR plant applications ............................................................................................ 246
5.5-3: MCCI sensitivity calculations for AP1000....................................................................... 250
5.5-4: Probabilities for the erosion of the concrete foundation due to core melt ........................ 259
7.2-1: Inputs of downward and sideward heat transfer coefficients ........................................... 315
7.2-2: Comparison of heat transfer coefficients in the approaches of GRS and IRSN ............... 316
7.3-1: Specification of ceramics test specimens ......................................................................... 336
7.3-2: Major test conditions ........................................................................................................ 338
NEA/CSNI/R(2016)15
List of abbreviations and acronyms
ABI Automated ball indentation

ABWR Advanced boiling-water reactor
AC Alternating current
ACE Advanced containment experiments
ACIWA AC-independent water addition
ADS Automatic depressurisation system
AEG Atomic Energy Act
AES91 A specific VVER (or WWER) design
AFNOR Association française de normalisation

AFW Auxiliary feed water
ALWR Advanced light-water reactor
ANL Argonne National Laboratory (United States)
APWR Advanced pressurised water reactor
ASTEC Accident source term evaluation code
ASN Autorité de sûreté nucléaire (France)
AtG Atomic Energy Act
BETA A large scale test facility to study melt concrete interaction in a cylindrical
concrete crucible
BiMAC Basemat internal melt arrest and coolability
BIP Behaviour of Iodine Project
BSAF Benchmark Study of the Accident at the Fukushima Daiichi Nuclear Power
Plant
BWR Boiling water reactor
BWRS Boiling water reactor stability
CANDU Canadian deuterium uranium reactor
CAV Cavity package
CCI Core concrete interaction
CEA Commissariat à l’énergie atomique et aux énergies alternatives (France)
NEA/CSNI/R(2016)15
CFD Computational fluid dynamics

CFR Code of Federal Regulations
CHF Critical heat flux
CHRS Containment heat removal system
CIS Commonwealth of Independent States
CLN Coolant
CMSS Core melt stabilisation system
CNRA Committee on Nuclear Regulatory Activities
COCO 3D Core calculation code
COLIMA Corium liquid and materials
COMET NEA computer code program
COPS Containment overpressure protection system
CORCON NEA computer program (Mechanical and Thermal Stress in Fuel Element
Clad)
CORPRO Corium properties database
COSACO Areva computer code for melt coolability and concrete interaction studies
COSCHEM A real-solution chemical-substance database
COTELS Computer code
CPU Central processing unit
CRC CRC Press
CRD Control rod drive
CSE Complementary safety evaluations
CSNI Committee on the Safety of Nuclear Installations
CVH Control volume hydrodynamics
DEH Direct electrical heating
DF Decontamination factor
DiD Defence in depth
DTA Differential thermal analysis
DVI Direct vessel injection
EC European Commission
ECOKATS Name of an experimentation within the European Union project ECOSTAR
ECOSTAR European Union project
EDF Électricité de France
EFCVS Emergency filtered containment venting system
EMF Electric melt furnace
NEA/CSNI/R(2016)15
ENSREG European Nuclear Safety Regulators Group

EOC Effective orthotropic heat transfer coefficients
EOP Emergency operating procedures
EPR European pressurised reactor
EPRI Electric Power Research Institute
ERI EURSAFE research issues
ERMSAR European review meeting on severe accident research
ESBWR Economic Simplified Boiling Water Reactor
ESF Engineered safety features
EUR European utility requirements
EURSAFE European expert network for the reduction of uncertainties in severe accident
safety issues
FCI Fuel-coolant interaction
FCHF Corium-pool-to-water heat flux (Modular accident analysis program code)
FE Finite element
FEM Finite element method
FINCCI Computer code
FRAG Computer code
FZK Forschungszentrum Karlsrühe (Germany)
GAREC Computer code
GDCS Gravity-driven cooling system
GEMINI Gibbs Energy MINImiser (software for thermo-chemical equilibrium solver)
HECLA Computer code

HEFEST Computer code
HMX Heavy oxides and metals
HOX Oxide phase
HPCI High pressure core injection
HSS Heated solid slug of steel
INRNE Institute for Nuclear Research and Nuclear Energy (Bulgaria)
IRSN Institut de radioprotection et de sûreté nucléaire (France)
IAEA International Atomic Energy Agency
ICI In-core instrument
IRWST Internal refuelling water storage tank
KEPCO Korea Electric Power Industry
KIT Karlsrühe Institute of Technology (Germany)
NEA/CSNI/R(2016)15
LBLOCA Large break loss-of-coolant accident

LCS Limestone-common sand
LDF Lower drywell flooder
LOCA Loss-of-coolant accident
LOX Light oxide phase
LPCI Low-pressure core injection
LWR Light water reactor
MACE Melt attack and coolability experiments
MAAP Modular accident analysis program
MBDBE Mitigation of beyond-design-basis events
MCCI Melt coolability and concrete interaction
MELCOR Methods for estimation of leakages and consequences of releases
MOCKA German experimental facility for melt coolability and concrete interaction
studies
NEA Nuclear Energy Agency
NPP Nuclear power plant
NRA Nuclear Regulation Authority
NRC Nuclear Regulatory Commission
NUCLEA French (IRSN) thermodynamic database for nuclear applications
OECD Organisation for Economic Co-operation and Development
OSSA Operating strategies for severe accidents
PAR Passive autocatalytic recombiners
PIRT Phenomena identification and ranking table
PLE Plant lifetime extension
PLINIUS French (CEA) experimental platform for corium studies
PRECOS Russian experimental programme that characterised ceramics relevant to in-
vessel and ex-vessel corium
PRHR Passive residual-heat removal
PSA Probabilistic safety analysis
PSR Periodic safety reviews
PWR Pressurised water reactors
RCIC Reactor core isolation cooling
RHWG Working group on reactor harmonisation
RPV Reactor pressure vessel
RWSP Reactor water storage pit
NEA/CSNI/R(2016)15
SAHRS Severe accident Heat removal system

SAM Severe accident management
SAMG Severe accident management guideline
SAMIME Severe accident management implementation and expertise in the European
Union
SAREF Safety research opportunities post-Fukushima
SARP Severe accident research priority
SESAM Senior Group of Experts on Severe Accident Management
SNL Sandia National Laboratories (United States)
SOAR State-of-the-art report
SOARCA State-of-the-art reactor consequence analysis
SOCRAT Russian severe-accident computer code
SSC Safety systems and components
STCP Source term code package
SURC Sustained urania concrete
TOLBIAC French (CEA) omputor code for analysis of melt coolability and concrete
interaction
TOP Technical opinion paper
TSO Technical Safety Organisations
WENRA Western European Nuclear Regulators’ Association
WEX Geman developed computer code for MCCI analysis of melt coolability and
concrete interaction
WWER Vodo-Vodyanoi Energetichesky Reaktor, a Russian-designed pressurised
water reactor (also sometimes referred to as WWER)
WGAMA Working Group on Accident Management and Analysis
NEA/CSNI/R(2016)15
Executive summary
Since 2002, the Nuclear Energy Agency (NEA) has sponsored the Melt Coolability and Concrete
Interactions (MCCI) co-operative project in two phases in order to investigate ex-vessel melt
coolability and concrete interaction by means of separate-effects tests and large scale integral tests
carried out at the Argonne National Laboratory (ANL). Key elements of this project included the
conduct of experiments involving real reactor material and associated analyses with the objectives of
resolving the ex-vessel debris 1 coolability issue and addressing remaining uncertainties related to
long-term two-dimensional molten core-concrete interactions under both wet and dry cavity
conditions. It was expected that the achievement of these two objectives would demonstrate the
efficacy of severe accident management guidelines for existing plants and provide the technical basis
for better containment designs for future plants.
Six months after the second phase of the experimental programme was completed, the Committee
on the Safety of Nuclear Installations (CSNI) held a seminar in November 2010 where the major
outcomes of the MCCI Project and other complementary experimental activities were presented and
discussed. One of the recommendations from the seminar was: “The preparation of a state-of-the-art
report (SOAR) on melt coolability and core concrete interactions that captures the last thirty years of
international research results”. A working group was established in April 2012 within the framework
of the Working Group on Accident Management and Analysis (WGAMA) to address this
recommendation.
The SOAR provides a background discussion of safety issues relevant to core-concrete
interactions and melt coolability and related containment failure modes, an overview of various
experiment programmes that have been carried out in the areas of molten-corium-concrete interaction
and debris coolability, a description and assessment of various analytical tools (“codes”) that have
been developed to analyse MCCI behaviour, and finally, a summary of plant analysis activities that
have been carried out using these codes. These various activities, carried out over the last three
decades, have significantly increased our level of understanding regarding MCCI behaviour under
both wet and dry cavity conditions. Depending upon containment design, regulatory requirements, and
accident management considerations that are unique to each country and reactor type, the current level
of understanding in this area is sufficient for conservative reactor safety assessments. However, a few
areas have been identified (particularly based on lessons learnt from Fukushima Daiichi) that may
warrant further investigation to reduce residual uncertainties.
While existing data and experiments indicate that debris coolability can be achieved within an
envelope that is principally based on concrete type, melt depth, and timing of cavity flooding, this
envelope does not encompass the full range of accident conditions that can be encountered in certain
plant configurations. Neither does the envelope encompass various abstractions of melt progression in-
vessel which give rise to different initial and boundary conditions for ex-vessel melt progression and
also, wide variations in concrete constituents within the two major types investigated in the
experimental programme and consequent effects of such variations. Also, the report focuses on the
1. The term “debris” refers to the corium melt in general and not only to the solid particles.
NEA/CSNI/R(2016)15
progress made in the last two decades on the thermal-hydraulic aspects of MCCI and mentions in
passing some early research programmes dealing with fission products aspects of MCCI. Finally, the
report discusses general aspects of severe accident management strategies aimed at achieving
melt stabilisation in both generation II and generation III reactors.
Lessons learnt from experiments
The general goals of the MCCI experiments under both dry cavity (i.e. without mitigation measures
involving water addition) and wet cavity (i.e. with mitigation measures involving water addition) have
been to: i) identify and characterise important phenomenological processes in order to facilitate model
development, and ii) provide experimental data to support validation of models and codes that are used
in reactor safety assessments. For dry cavity conditions, the research focused on evaluating the nature
and extent of core-concrete interaction, basemat and sidewall erosion, and concurrent fission product
release. For wet cavity conditions, the research focused on evaluating the effectiveness of water in
terminating the MCCI by quenching the molten core material and rendering it permanently coolable,
i.e. to achieve the melt stabilisation condition.
The various accident sequences and the possibility of operator intervention result in a broad range
of possible initial conditions at time of vessel failure. Following the accident at Three Mile Island and
some studies of melt interactions with concrete, it was presumed that core degradation would be very
heterogeneous with central regions of the core melting while peripheral regions were barely degraded.
Additional core materials would cascade for protracted periods from the reactor vessel as core debris
attacked concrete. A certain fraction of the cladding would not be oxidised at the time of core debris
relocation to the lower head of the pressure vessel and upon vessel breach there would be a chemical
component to the heat generation in the core debris. Additionally, the state of knowledge about late in-
vessel melt progression is incomplete (particularly for boiling water reactors, BWRs). Thus, there is
considerable uncertainty regarding the MCCI initial conditions that includes the timing of reactor
pressure vessel (RPV) failure; the initial temperature, mass, and composition of the core debris; the
possibility of segregation of metal and oxide melt phases; the pour rate of the melt from the RPV that
is determined principally by the melting rate of residual core material, and to a lesser extent by the
opening in the RPV lower head; and finally, the timing of water injection (if any).
Many of these parameters (e.g. power level in the ex-vessel core debris, which is indicative of the
time of vessel failure, as well as melt mass and composition) have been addressed in various
experimental programmes that are described in Chapter 2. It is important to recognise that as the
understanding of core degradation evolved since the Three Mile Island accident and now continues to
evolve since the Fukushima accident, modelling of in-vessel melt progression likewise will continue to
evolve. As such, no attempt has been made in this report to encompass the full range of possible initial
and boundary conditions (some of which are known at present) and to conclude that the current
understanding of the MCCI phenomena is complete.
The various test series described in Chapter 2 investigated the effects of melt composition,
concrete type, input power to melt, and in experiments with cavity flooding (wet cavity experiments),
the timing of water addition on two-dimensional core-concrete interaction and melt coolability.
Principal variables measured during the experiments included melt temperature and local concrete
ablation rates. For flooded cavity experiments, the debris/water heat flux after cavity flooding was also
estimated based on the rate of steam production from the interaction. Key observations from these
tests are summarised below.
Under dry cavity conditions, all tests exhibit the overall trend of decreasing melt temperature as
concrete ablation progresses and increasing heat transfer surface area as the melt expands into the
concrete. This trend depends also on the decrease of the interface temperature between the melt and
the upper crust as the melt becomes enriched with low melting concrete decomposition products. The
NEA/CSNI/R(2016)15
results from several reactor material experiments indicate that the concrete ablation process for oxidic
core melts is influenced by concrete types. For limestone-common sand concrete, the radial to axial
erosion rate and ablation depth are approximately one to one whereas for siliceous concrete, it is
approximately three to one. To investigate whether the relatively small melt pool aspect ratio (i.e. test
section width/melt depth) used in the experiments has an influence on the radial/axial power split
observed in the dry cavity experiments, a dedicated large scale experiment was carried out. The results
indicate that an increase in aspect ratio from approximately 1 (typical of most reactor material tests) to
about 3.2 has no noticeable effect on the ablation characteristics for siliceous concrete. This
observation lends additional credibility to the measured erosion rates and ablation depths in various
experiments. It is noted that the forensic examinations of Chernobyl Unit 4 are consistent with the
experiment observations for siliceous concrete thus giving credibility to long term extrapolation of
experimental data even if the ablation asymmetry is not yet understood from a mechanistic point of
view.
Post-test examinations have shown that the nature of the core-concrete interface is noticeably
different for limestone tests in comparison to siliceous tests. The differences in interface
characteristics may influence the heat transfer at the interface, yielding different concrete ablation
behaviour for different concrete types. The overall trend in the ablation front progression that has been
observed under experiment as well as Chernobyl Unit 4 examinations cannot, however, be explained
fully on the basis of our current understanding of the ablation behaviour and modelling of such
behaviour. Thus, extrapolation of the results to plant conditions remains somewhat uncertain due to
the lack of a more robust phenomenological model that can rationalise the differences in the observed
cavity erosion behaviour of the two concrete types used in the experiments. It is worth mentioning that
while variations of the two major concrete types (e.g. representative concrete type in French plants
with variations of siliceous aggregates and “serpentine” concrete used in one of the Argonne Core-
Concrete Experiments [ACE]) were used in some experiments, the database of such variations is
somewhat limited.
Several experiments have provided evidence that initial corium crust formation on cold concrete
surfaces can influence the early (tens of minutes to an hour) core-concrete interaction behaviour.
During this phase, basemat ablation is minimal and melt temperature remains high due to the
insulating effect of the crusts. This effect has been observed in both transient as well as sustained
heating reactor material tests. However, once crusts at concrete interface fail, concrete ablation
proceeds vigorously and the melt temperature declines due to the above mentioned effects. Although
the data are not conclusive, there is evidence that gas evolution from concrete decomposition can act
to destabilise these interfacial crusts.
The effect of unoxidised Zr cladding on the thermal-hydraulics of the core-concrete interaction
was investigated in several experiment programmes. The oxidation reaction between Zr and sparging
concrete decomposition gases (CO2, H2O) caused exothermic transients during which the melt
temperatures increased over a period of tens of minutes in the experiments. This transient behaviour
was observed in both reactor material as well as simulant experimental tests. The data further indicate
that, after the cladding is fully oxidised, melt temperature drops to a value consistent with fully
oxidised melt conditions. Aside from cladding oxidation effects, a limited number of experiments have
been conducted to examine the effect of significant structural steel 2 content in the melt on core-
concrete interaction behaviour. One outcome from these tests is the extensive amount of iron oxidation
that occurs with limestone-common sand concrete. Steel oxidation also occurred in tests with siliceous
concrete, but to a lesser extent in comparison to the limestone case. Concrete temperatures showed
2. Steel comes from in-vessel structures or rebars embedded in the concrete for its reinforcement
NEA/CSNI/R(2016)15
that axial and radial ablations were more pronounced in the areas where metal was found. This is
consistent with other oxide-metal simulant tests that have shown enhanced heat transfer at the metal-
concrete interface relative to the oxide-concrete interface. Notwithstanding these findings, the metal-
oxide database is noted to be limited and there is no clear understanding of the phenomenological
behaviour for this case (i.e. mixed vs. stratified metal-oxide conditions, and/or bifurcation between
these two states during the interaction). Therefore, extrapolation of experimental results obtained thus
far to plant conditions is somewhat uncertain.
The fission product release during core/concrete interaction is only briefly mentioned in the
report as it has not been the main focus of the MCCI experiments in the last two decades. The aerosols
released during corium/concrete interaction contain mainly elements from the concrete. The release of
uranium or low-volatile fission product is enhanced by the presence of metal in the melt and by the
higher gas content of limestone common sand concrete but remains low. Interaction with silicon to
form silicates tends to lower the release of fission products of main interest like barium and strontium.
Regarding debris coolability, the test series provided evidence of several heat transfer
mechanisms that can contribute to long-term corium cooling. When the core debris is flooded from
above, the question of whether or not a significant amount of the thermal energy will initially be
removed depends upon whether a stable crust is able to form that inhibits heat transfer from the melt to
the water layer. For a stable crust to form, two conditions must be met: (i) a thermal condition, viz.,
the melt/water interfacial temperature must fall below the corium freezing temperature, and (ii) a
mechanical condition, viz., the incipient crust must be stable with respect to local mechanical loads
imposed by the agitated melt. In this bulk cooling regime, efficient melt/water heat transfer occurs
predominately by radiation heat transfer across the agitated (i.e. area enhanced) melt/water interface,
in addition to entrainment of melt droplets into the overlying water and conduction.
As bulk cooling heat transfer continues, the melt temperature gradually declines. As the
downward heat transfer rate decreases, then melt sparging arising from concrete decomposition also
decreases. Thus, a point is eventually reached at which the thermal and mechanical thresholds for
interfacial crust formation are both satisfied, and an insulating crust forms between the coherent melt
zone and the water layer. This crust is characterised by some degree of porosity, or cracks, owing to
the necessity of venting concrete decomposition gases.
After the crust forms, completion of the quench process can only be achieved if water is able to
penetrate into the debris by some mechanism to provide sufficient augmentation to the otherwise
conduction-limited heat transfer process to remove the decay heat. The tests have revealed three
mechanisms that provide pathways for water to penetrate the debris. The first is water ingression
through interconnected porosity or cracks. The second mechanism is particle bed formation through
“volcanic” eruptions. In this case, concrete decomposition gases entrain melt droplets into the
overlying coolant as they pass through the crust. The third mechanism is mechanical breach of a
suspended crust. In particular, the thick crusts that form from water ingression could bond to the
reactor cavity walls, eventually causing the melt to separate from the crust as the MCCI continues
downwards. However, this “anchored” or suspended configuration is not expected to be mechanically
stable at reactor scale due to the low mechanical strength of the crust. The suspended crust situation is
somewhat different from a supported crust situation as in many magma flow from volcanic eruptions.
Eventually the suspended crust is expected to fail, leading to rapid and massive ingression of water
beneath the crust. This sudden introduction of water provides a pathway for renewed debris cooling by
the bulk cooling, water ingression, and melt eruption cooling mechanisms.
Several findings related to debris coolability are directly applicable to evaluating plant accident
sequences. In terms of the water ingression mechanism, the test results indicate that the heat transfer
correlation based on previous one-dimensional (SSWICS) tests is conservative insofar as calculating
ingression-limited crust growth behaviour. In particular, the correlation tends to under-predict the heat
flux to overlying water during time intervals when water ingression is occurring. In terms of the melt
NEA/CSNI/R(2016)15
eruption mechanism, significant eruptions have been observed in the case of limestone common sand
concrete. Eruptions have also been observed under early cavity flooding conditions for siliceous
concrete. For tests conducted in two dimensional cavity configurations, the eruptions appear to have
occurred under a floating crust boundary condition, which is expected at plant scale. Finally, the crust
breach data indicate that the crusts that form at the melt/water interface after cavity flooding have a
low mechanical strength, and cannot be mechanically stable at plant scale. Rather, the crust is
expected to fail and, thereby, maintain a floating crust boundary condition that will allow the melt
eruption and water ingression cooling mechanisms to proceed.
It is noted that the debris coolability experimental database consists almost exclusively of tests
conducted with oxidic melts. When a significant metal fraction is present in melt, it may result in a
stratified pool configuration. This type of pool structure has not been evaluated from a coolability
standpoint. Thus, additional analysis and testing may be required with melts containing a significant
metal fraction to further reduce phenomenological uncertainties related to debris coolability, as well as
core-concrete interaction.
Capabilities and shortcomings in current codes and models
One goal of this SOAR was to summarise and assess capabilities of various simulation tools currently
used in the world, focusing on models used to describe the corium concrete interaction phenomena and
the coolability mechanisms induced by top flooding. Chapter 3 of this report provides a generic
description of MELCOR MCCI module (i.e. CORCON), CORQUENCH, COSACO, ASTEC MCCI
module MEDICIS, TOLBIAC-ICB, WECHSL, COCO, MAAP MCCI module, and the SOCRAT
MCCI module. More detailed models descriptions are also provided as in the appendices.
All codes can currently analyse the case in which the corium is assumed to be instantaneously
spread over the entire floor of the reactor pit under dry cavity conditions. This situation is most
consistent with rapid high temperature melt pour scenarios in dry cavities, for which efficient
spreading can be expected. However, not all the codes can adequately model the impact of top
quenching. In addition, the ability to analyse situations in which concrete ablation is limited to a part
of the reactor pit walls and/or floor due to localised corium accumulations, as well as scenarios
involving multiple pours separated in time, cannot be currently treated in an easy and systematic way
using available codes. Moreover, many current codes are not able to treat MCCI scenarios that involve
spreading into additional cavities adjacent to the reactor pit. Finally, the effect of concrete type on
concrete erosion pattern viz., the experimental results for siliceous or limestone common sand
concrete cannot be explained satisfactorily by existing phenomenological models.
Another aspect documented in Chapter 4 of the SOAR relates to validation and discussion of
modelling. Detailed and often different models of various phenomena are recommended for use in
different codes, as a result of individual code validation work that is often based on separate effect
experiments. However, MCCI is a complex interaction of several phenomena and it is important to
validate the codes against integral MCCI experiments to gain an appreciation of the predictive
capability of these tools. The validation status of individual codes is described from a general
viewpoint, with additional details regarding higher-level phenomenological models provided in the
appendices. Since all models require material property data, the validation status of property data and
supporting correlations are also reported, along with an assessment of how the experiment data can be
scaled to reactor conditions based on available models in the codes.
The validation work is commonly focused on comparing key calculated results (i.e. corium
temperature and local or maximum ablation depths) with the experimental data. Transient effects often
impact the course of individual experiments – principally at the start of the interaction between the
newly generated melt contacting cold test section structures, resulting in the formation of melt crust at
the interface with the concrete that prevents its ablation – for which the codes cannot be assessed and
NEA/CSNI/R(2016)15
adequately validated since most codes do not have the necessary modelling capabilities. However, in
the longer term, the experiments enter a quasi-steady regime in which code predictions of concrete
ablation progression and temperature history are reasonably well understood. Good agreement with
oxide melt experiments in terms of ablation and temperature history are noted when heat transfer
coefficients on the order of 300 W/(m2K) and concrete decomposition temperature close to ~1 600 K
are selected as boundary conditions at the melt-concrete interface.
Among other relevant findings from code assessment activities, it is noted that anisotropic
ablation observed for siliceous concretes can currently be captured in the codes only via empirical
application of heat transfer coefficients based on observations from two-dimensional experiments. The
assessment of top flooding conditions on the course of the MCCI is not yet clear since the crust
anchoring effect, observed in many experiments but not expected at reactor scale, is difficult to take
into account and is modelled in very few codes. Uncertainties were identified for melts consisting of
both oxides and metals. In particular, in a stratified pool configuration the metal melt thermal material
properties suggest elevated heat transfer at the metal/concrete interface, but the overall transfer of
decay power (which is predominantly released in the oxide melt) to the concrete via the metal layer is
governed by the heat transfer at the interface between the oxide and the metal layer. Direct model
validation for this interlayer heat transfer is not yet possible due to lack of appropriate data from
experiments under MCCI conditions. Additional uncertainty is introduced due to a lack of knowledge
regarding stratification and mixing processes under MCCI conditions. For situations in which concrete
reinforcement serves as a continuous source of metal, its impact on the concrete ablation mechanism is
still not properly understood.
Based on these underlying experiment and code assessments, an additional high level goal of this
SOAR is to review applications of MCCI phenomena, models, and data to safety analysis of nuclear
power plants under severe accident conditions, particularly in the context of reactor safety
requirements and containment designs to address such requirements. Safety requirements,
promulgated by various international bodies, are discussed as well as containment designs for a
number of generation II and generation III plants with particular emphasis placed on features relevant
to the MCCI issue. Three idealised plant (containment) configurations for reference plant calculations
are also discussed, and a few example plant calculations are presented. These examples illustrate the
approach of plant idealisation (simplification) that is quite common and reasonable in the field of
safety analysis, noting the inherent uncertainties in severe accident phenomena. As with virtually all
other severe accident phenomena, extrapolating the results of scaled experiments to MCCI in plant
scale involves some idealisation of plant geometry and configuration. For full-scale plant safety
assessment, the approach appears to be pragmatic whereby MCCI phenomena are analysed based on
conservative assumptions with respect to the weaknesses of the containment design.
Remaining issues in the areas of MCCI and debris coolability
Notwithstanding the progress made in the field of core-concrete interaction, the apparent simplicity of
the treatment of thermal-hydraulics of a well-mixed corium pool in the presence of the concrete
decomposition gases is contrasted by the complexity of the concrete ablation mechanism where the
heated concrete, a highly heterogeneous material, is gradually incorporated into the melt through an
evolving melt-concrete interface that is still difficult to observe experimentally and capture from a
modelling point of view. Because many of the models are not mechanistic, several parameters are
often empirically fit to reproduce as best as possible the scaled experimental results. Attempts to
model MCCI by a multi-scale computational approach to eliminate these tuned parameters with more
mechanistic models have been unsuccessful to date. This is mostly due to the difficulty of observing
and measuring local phenomena needed to validate multi-scale modelling approaches.
NEA/CSNI/R(2016)15
Based on the foregoing results, this SOAR focused on identifying remaining issues and residual
uncertainties in the areas of MCCI and debris coolability, and formulates some recommendations for
addressing these questions in the near future in order to increase the reliability of reactor simulations.
These remaining issues and recommendation can be classified into three general topics of interest
that are:long-term core-concrete interaction behaviour; realistic plant simulations; and coolability
enhancement under top flooding conditions.
Long term core-concrete interaction behaviour
The past MCCI experiments with prototypic materials were relatively short in duration, which is
partially due to the fact that constraints did not allow significantly longer duration experiments.
Admittedly, in many of these experiments the test duration was adequate to assure at least partial melt
cooling and slowing down of basemat ablation to a level that could be considered acceptable for
regulatory purposes (i.e. ablation limited to a specified amount by say 24 hours into the accident).
The Fukushima accidents suggest that a much longer transient is quite likely. A recent report on
“Safety Research Opportunities Post-Fukushima – Initial Report of the Senior Expert Group”
(NEA/CSNI/R(2016) 19) noted that MCCI very likely occurred in one or more Fukushima Daiichi
units for some time, but did not lead to a significant melt release outside the containment vessel to the
reactor building. The analyses performed to date in the NEA Benchmark Study of the Accident at the
Fukushima Daiichi Nuclear Power Station (BSAF) project phase 1 have not provided a consensus
view on the MCCI issue. In particular, the termination of the MCCI process was found to have been
impacted significantly by differences in melt pour conditions predicted by different codes at reactor
vessel failure. These findings put into question those analyses results which predict ongoing MCCI for
a long time, especially in the presence of water. Hence, there is a need to obtain longer duration
experimental data if the shorter duration experimental data cannot be extrapolated to the reactor
situation with a high degree of confidence.
Extrapolation of existing knowledge to long term MCCI processes
Longer duration experiments will provide data needed to: (1) confirm that intermittent phenomena like
melt eruptions are reproducible; and (2) investigate if the crust formed by water ingression is stable.
Long duration experiments will also provide data on long term behaviour dealing with the final phase
of the interaction, i.e. the time when the heat flux to concrete is low enough that it can be dissipated by
conduction into concrete without further ablation, or the heat flux that is applied in a specific coolant
circuit of a core catcher. Finally, long term behaviour also refers to situations wherein the concrete
fraction within the melt and the heat flux level are representative of the situation after many hours of
interaction. The subject of long-term behaviour vis-à-vis further research needs and recommendations
is discussed further in the following paragraphs.
It is noted elsewhere that more rapid radial ablation (relative to axial) was observed for siliceous
concrete, whereas limestone common sand concrete showed an isotropic (uniform axial and radial)
ablation profile. Currently, there is no generally accepted phenomenological explanation for this
behaviour, and the question remains if it is reliable to extrapolate this result to reactor scale for a
longer duration ablation process. This general scaling issue is equally important to concrete ablation in
a wet cavity situation.
Another complex phenomena related to this issue is associated with intermittent ablation bursts
that are observed in experiments. It is not clear if this is a result of crust instability or rather a result of
concrete spalling due to mechanical instability. Depending on the phenomena, the characteristic time
period can be several hours; e.g. crust dissolution processes with siliceous concrete. In this case the
test duration has to be long enough to observe at least two or three ablation bursts. Data from these
NEA/CSNI/R(2016)15
long duration experiments will reduce uncertainties in current melt eruption models and will provide
better confidence in extrapolating to reactor scale.
It is important to recognise that in every facility, the size of the test section, the heating technique,
and/or the operating procedure always induces some transient system effects. These transient effects
are not modelled in the codes and it is not required for most of plant applications. As a result when the
transient effects are dominant, the codes cannot reproduce accurately the final cavity shape which is
commonly used to estimate the ablation rate. It is recommended that an experiment objective should
be to reduce the duration of the initial transient, and experiment techniques that can contribute to
homogenous initial melting of the corium and limit initial crust formation should be encouraged.
Furthermore, future tests should be designed to run under steady state conditions for a longer duration.
Under long test operating conditions involving top flooding, one systematic drawback of the
experiments is the top crust anchoring phenomena. In tests performed with top flooding, the upper
crust eventually anchored to the side walls. The anchoring phenomenon unrealistically reduces the
efficiency of the melt ejection phenomena because a gap between the pool and the upper crust appears
and then increases due to concrete densification upon melting as well as loss of liquid corium as
eruptions occur. At the beginning of the process, crust anchoring could also create a pressure build-up
effect below the crust that experimentally distorts the eruption process. Crust strength measurements
made on crust samples obtained from experiments and supporting structural analyses indicate that a
floating crust boundary condition is likely in full-scale reactor geometry. In this spirit, experiment
techniques that can promote a floating crust boundary condition in reduced scale experiments should
be encouraged.
The final step of the MCCI process is characterised by a core-concrete heat exchange surface so
large that the heat can be transferred by conduction to the remaining concrete without further ablation.
This scenario would yield a very viscous melt with high concrete fraction. Under such conditions, the
heat transfer models at the core-concrete interface may not be valid. Some codes utilise a quasi-steady
modelling approach in which conduction into the concrete is not modelled. Thus, all heat transfer from
the core debris is dissipated by concrete ablation, and as a consequence, the ablation never stops.
Some of these deficiencies in analytical tools can be addressed with data from longer duration
experiments.
Realistic plant simulations
Improving the realism in plant simulations inherently introduces more complexity. As a result, the
associated efforts have to be balanced with approaches that rely on invoking additional levels of
conservatism to define a bounding set of hypotheses for safety-relevant issues. Three major related
issues are: the presence of metal within the melt or within the concrete; the initial conditions for MCCI
based on melt pour conditions into the reactor pit; and the presence of impurities in cooling water
(e.g. seawater or brackish water).
Presence of metal within the melt or within the concrete
The presence of metal within the melt or within the concrete influences the ablation profile as soon as
stratification occurs. The stratification process is governed by the higher density ratio between metal
and oxides as soon as the fuel oxides become diluted with concrete oxides. While several experiments
have been performed with iron-alumina thermite simulant, only a few tests have been performed with
a fully prototypic metal-oxide core melt composition. It was not possible to establish from the results
clear evidence of stratification, but ablation was observed to be increased in front of the metallic
masses. For prototypic metal-oxide core melt, the oxidation kinetics and the stratification thresholds
are important as they influence the time window when the melt is stratified at reactor scale and as a
result, the prediction of the basemat melt-through time.
NEA/CSNI/R(2016)15
The initial phase of the melt-concrete interaction involving unoxidised cladding (zirconium) in
the melt has been investigated in a few reactor material experiments. This stage can lead to highly
exothermic metal oxidation reactions. Zirconium-bearing concrete-metal inserts were used in some
Argonne experiments in which a relatively small amount of Zr was incorporated into the melt just
prior to melt contact with the concrete basemat. However, it is likely that a significant fraction of the
Zr in the inserts was oxidised before the test initiation, thereby limiting the impact on the actual MCCI
phase of the experiment. Another aspect not investigated in experiments is the presence of uranium
within the metallic phase. During the in-vessel stage of the accident, uranium is found in the metal
phase in scenarios that lead to a significant fraction of unoxidised cladding in the lower head. Under
these conditions it seems appropriate to implement an oxidation model for uranium in simulation tools
and to perform sensitivity analyses.
Reinforcing bars in concrete play a double role as they are a continuous source of metal which is
prone to oxidation during ablation and additionally, they change the ablation mechanism. In particular,
some recent test results indicate that the presence of reinforcing bars in siliceous concrete leads to a
homogeneous ablation profile, which contrasts the results from reactor material tests carried out with
non-reinforced siliceous concrete in which anisotropic ablation was observed.
The presence of metallic inclusions in an otherwise oxidic crust could change the properties and
thereby impact the water ingression cooling mechanism. Specifically, the presence of metal could
influence the critical heat flux associated with cracks that form in the crust due to thermal contraction
induced by top flooding. To address this issue, additional experiments could be performed with
different metal contents in the melt and a representative gas release to promote good mixing
conditions. For these tests, as well as large scale experiments with sustained heating, new thermite
compositions need to be developed that would produce a melt with adequate metal fraction.
The effect of metallic inclusions in melt on the melt ejection cooling mechanism is different as it
is occurs over a longer duration. It would be interesting to evaluate the entrainment rate of pure metal
melts and check the morphology of the particles formed during the quenching process to assess their
coolability as well as their influence on the coolability of the particulate debris bed in general. As soon
as the specific technological challenges of metal-oxide experiments are resolved, tests with high metal
fraction are recommended.
Thermal stresses on concrete structures brought on by core debris interactions with concrete have
not been investigated in MCCI Programs. These stresses are largely inconsequential for below grade
reactor cavities but can be quite important for free standing cavities such as sub-atmospheric
containments and especially for reactor pedestals in boiling water reactors. The core debris
interactions place the inner region in compression where concrete is strong but the outer region in
tension where concrete is weak and easily cracks. This has structural implications which again have
not been investigated in MCCI Programs.
MCCI initial conditions following melt relocation into the reactor pit
It is often assumed that the MCCI phase starts as soon as the vessel fails and the corium mass in the
lower head (which in bounding analyses includes the entire fuel and structural inventory in the reactor)
is relocated into the reactor pit. This approach offers a degree of conservatism in terms of axial-melt-
through delay if one assumes that the melt is spread instantly over the entire surface of a dry pit.
However, when the reactor cavity is flooded, spreading may be limited leading to corium
accumulation in one part of the reactor cavity. This scenario will result in higher heat fluxes to
concrete and reduce the basemat melt through time if this accumulation is stable and does not
eventually spread out uniformly. Among other things, local corium accumulation in the reactor cavity
mainly depends on the corium temperature, corium pour rate, reactor pressure vessel (RPV) failure
NEA/CSNI/R(2016)15
location, amount of water present in the reactor cavity at RPV failure, and finally on the ability to
provide water continuously on top of the corium accumulation.
Such configurations are quite complex to study because they involve the formation and spreading
of corium accumulations under water as well as the possibility of boiling off the water inventory,
drying out the core debris, re-melting, and onset of concrete ablation. An ancillary issue is that core
debris in a reactor cavity, if not covered by water, exposes a great deal of concrete surface area to
intense convective and radiative heat flux. The gas generation and concrete degradation from this
exposed concrete cannot be neglected in the analysis of core concrete interactions and containment
integrity.
Depending upon the melt pour conditions and with a relatively shallow water layer, melt jet
fragmentation is expected to be minor. For this type of scenario, existing MCCI models that treat the
corium as an initially intact melt pool interacting with concrete may be employed as a reasonable
approximation. However, for deeper water pools melt jet fragmentation may be significant, leading to
formation of a coherent particle bed, or a compact melt layer commonly referred to as a cake
surrounded by particle bed. Depending upon the bed depth, decay heat level, particle size, and
porosity, the configuration may be coolable. However, if the dry-out limit for the bed is too low then
gradual reheating, dry-out, melting, and onset of concrete ablation will occur.
These particular configurations have not been extensively investigated as part of MCCI research,
nor can existing MCCI models address all of these configurations. However, there has been a
significant amount of research done in this area (both experiments and modelling) that generically
addresses particle debris bed coolability for both in-vessel and ex-vessel applications. Conducting
experiments that involve dry-out and melting of particle beds composed of reactor materials is a
technical challenge given limitations with current core debris heating techniques. Thus, a possible first
step to address this issue is to utilise existing models to evaluate coolability of particle bed formations
predicted for plant applications. If these analyses indicate that the beds are not likely to be coolable,
then effort should be devoted to developing appropriate experiment techniques to address this type of
behaviour.
Another related issue is that of multiple pours and how that affects the coolability of debris in the
reactor pit. Again, in all experimental and analytical studies concerning MCCI, it is traditionally
assumed that at RPV failure, the molten material (whether the entire reactor inventory or partial
inventory) is poured all at once and instantly spreads on the whole reactor pit surface. It is likely that
in some accident scenarios, the melt pour would be periodic which has two consequences: non-
uniform melt accumulation and non-uniform spreading. Conducting experiments with this kind of melt
configuration may be quite challenging, and an analytical extrapolation of experimental data for
symmetric and uniform melt configuration may be more worthwhile based on simulant data.
Presence of impurities in cooling water
The impact of impurities in cooling water on severe accident behaviour resurfaced following the
Fukushima accident. In particular, the use of sea water brought into question the impact of salt
(sodium chloride) on coolability mechanisms, on fission products chemistry, and the performance
(i.e. potential for clogging) of coolant loops. Generally speaking, any impurity in cooling water
(whether it is salt in sea water or other forms of impurities in brackish inland fresh water) can impact
one or more of these areas.
For the ex-vessel corium cooling mechanisms identified under top flooding conditions, the
formation of precipitate in the cracks of the upper crust or in the overlying debris bed could reduce the
dry-out limits for these formations. As the composition of water present in the sumps at the bottom of
the reactor building is complex and may depend on the accident management strategy, it seems easier
to address the issue in separate effect tests than in semi-integral experiments. Ongoing experiments in
NEA/CSNI/R(2016)15
Japan are addressing some aspects of the water impurity issue. To parametrically investigate the effect
of water impurities on melt coolability by water ingression, SSWICS-like tests could be run to
evaluate the impact on the cracks formation and on the crust critical heat flux. If warranted, more
complex experiments (i.e. with sustained heating) could be conducted to assess the behaviour over the
long term. For the melt ejection mechanism, the influence of impurities on debris bed coolability could
be investigated in separate effects tests that utilise existing facilities for investigation of dry-out in
debris beds for in vessel conditions.
The water at the bottom of the containment building will be highly contaminated with fission
products. If this water is used to cool the melt, the chemistry of the fission products will likely be
modified by gas bubbling and more generally by particulate entrained in the water. While the fission
product behaviour under such conditions is an ancillary issue related to the consequences of clogging,
water samples could be collected quite easily at the end of MCCI experiments to perform chemical
analysis in order to characterise the chemical composition. If some impurities in the water plays a role
in trapping other species released during MCCI, it would be useful to carefully select the composition
of the water before running these tests.
Methods for improving melt coolability under top flooding conditions
This SOAR is focused on ex-vessel corium coolability under top flooding which is largely regarded as
a generic accident management strategy for ex-vessel melt stabilisation in existing plants. The
improvement of melt coolability under top flooding conditions can also be viewed as a potential back-
fitting strategy for operating reactors. Moreover, for new reactors, spreading and top flooding can be
incorporated in the design phase as a generic approach.
In terms of improvement, it is noted that a larger initial corium spreading area will reduce the
downward heat flux to the concrete and hence, reduce the likelihood of basemat melt-through. One
approach for increasing melt spreading area for plants with limited floor space is to allow radial melt-
through of a barrier with subsequent spreading of a portion of the melt into the reactor building. This
situation is more likely for siliceous concrete but remains limited only to the level of corium above the
breach elevation. This corium spreading strategy may be more effective in a dry cavity situation, one
that also provides the benefit of eliminating the risk of steam explosion.
The coolability of debris can be more efficient if corium spreading is combined with flooding.
Water ingression mechanism is most efficient at the early flooding stage with little concrete present in
the melt, and that the melt eruption mechanism is also most effective in the early phase of the corium-
concrete interaction due to the higher melt gas sparging rate. Ideally, it is desirable to have an initially
dry pit to maximise spreading and to avoid the risk of a steam explosion, followed by early flooding.
In this case, the time window to add water is narrow and a subsequent melt pour after top flooding
cannot be excluded.
Another consideration is the composition of the concrete that is used for the reactor basemat at
the plant. For new plants, a recommendation can be made to consider high carbonate and/or hydrate
contents for the concrete used for reactor basemat. For existing plants with a potentially too thin
siliceous concrete basemat, a possible back-fitting measure could be to consider pouring an additional
(sacrificial) layer of high carbonate and/or hydrate concrete. In this case, since the thickness of this
additional layer can be limited for a specific plant site, the key piece of information needed is the
efficiency of the melt ejection mechanism so as to ensure that the liquid melt is transformed into a
coolable particle debris bed before reaching the original siliceous concrete.
NEA/CSNI/R(2016)15
Perspectives
In the coming years the examination of the debris in the three damaged Fukushima reactors will likely
provide additional insights that will enhance the understanding of MCCI phenomena at large scale and
under fully prototypic conditions. The findings will undoubtedly provide opportunities to gain
additional confidence in the application of simulation tools to existing plants. They will also provide
data and information for optimising severe accident management strategies for existing as well as
future plants.
One of the top level recommendations in the NEA-SAREF report (in preparation) is to organise
an MCCI workshop to discuss current state of MCCI knowledge, identify knowledge gaps, and
identify data needs to bridge the gaps – the idea being that the Fukushima decommissioning effort can
be informed by the outcome of such a workshop while at the same time, data collected during the
decommissioning activities can be optimised to bridge the MCCI knowledge gaps. In two companion
studies (one on severe accident knowledge gaps post-Fukushima and the other on Fukushima forensic
data needs), MCCI knowledge and data gaps were identified as high priority topics. These findings
confirm that in order to perform experiments and additional analysis to address more realistic
situations, it is necessary to improve the capabilities of existing facilities and to perform needed
experiments to bridge the knowledge gaps and reduce residual uncertainties. Since experimental
MCCI research with prototypic reactor materials is an expensive undertaking, a collaborative effort
among various nuclear safety research organisations in different countries is highly recommended.
NEA/CSNI/R(2016)15
1. Introduction to the main phenomena involved: Brief description

of accident phenomenology
Chapter 1 provides an overall picture of the whole content of this SOAR. Although most of the topics
included in this section are thoroughly developed in further sections of this report, Chapter 1 provides
the basic definitions and descriptions of relevant safety issues in order to get a better comprehension of
every aspect of the severe accident issues addressed by this SOAR.
The outline of Chapter 1 is as follows. Section 1.1 provides two key definitions for this report:
the concepts of melt stabilisation and severe accident termination.
Section 1.2 shows the main paths followed by a severe accident after the failure of the reactor
pressure vessel (RPV) in a light water reactor, with a special focus on molten core concrete
interactions and ex-vessel corium coolability mechanisms.
Section 1.3 provides the basic understanding of the concrete ablation by the corium and it also
describes the corium coolability mechanisms by top flooding. Because of corium and concrete are the
two major components involved in the severe accident phenomena addressed by this SOAR, it was
deemed that a brief description of these two components should be provided, including an overview of
their most relevant physical and chemical properties. Additionally, two safety significant phenomena
are also discussed in this section: radial vs. axial concrete ablation and corium stratification.
Containment failure modes associated with molten core concrete interactions are described in
section 1.4. Safety topics unaddressed by this SOAR are also included in this section.
Section 1.5 deals with general aspects of severe accident management and severe accident
management strategies aimed at achieving the melt stabilisation. Both, Gen II and Gen III reactors are
considered in this section. A brief description of recent back-fitting measures implemented or under
consideration in some Gen II reactor is also included.
Finally, section 1.6 provides the safety prioritisation of the different phenomena addressed by this
SOAR within the framework of severe accident phenomenology. This prioritisation is based on a
number of expert judgement activities conducted by the European Union and the SARNET group.
1.1. The concept of melt stabilisation and severe accident termination
The terms “melt stabilisation” and “severe accident termination” are widely used along this report.
Thus, a definition of these two key concepts is required, previously to any other technical information.
The following definition for melt stabilisation and severe accident termination is taken from (Sehgal,
2006).
“A severe accident cannot be characterised as stabilised and terminated until core melt/debris has
been cooled and quenched and kept in the latter stage for a long time. Achieving and maintaining
coolability of the melt/debris is paramount, since fission product release and non-condensable-gas
generation stops as the melt/debris temperature drops below ~ 1 000 ºC and containment integrity is
not seriously challenged anymore, if containment cooling cycle has been established.”
NEA/CSNI/R(2016)15
1.2. Corium discharge from the RPV into the containment
Figure 1.2-1 gives the major paths of severe accident progression after a RPV failure for an existing
light water reactor type with inverted containment atmosphere (Allelein & Burger, 2006). In order to
focus on the relevant phenomena related to core-concrete interactions and corium coolability,
hydrogen combustion processes within the containment and the potential containment failure by an in-
vessel steam explosion (the alpha-mode failure) are excluded from Figure 1.2-1.
According to Figure 1.2-1, the first question deals with the RCS pressure at the moment of the
RPV failure. If the RPV fails with high pressure in the RCS, the corium would be widely dispersed
throughout the containment. These scenarios are considered in the Section 1.4.5.
Assuming that the RPV fails with low pressure in the RCS and that RPV does not fail
catastrophically, e.g. due to steam explosion or by melt-through with subsequent unzipping/tilting of
the whole lower head, but rather locally in a mode (“fish-mouth”-type failure), e.g. observed in
FOREVER experiments (Theerthan, Giri, Karbojian, & Sehgal, 2002) and in low-pressure LHF
experiments (Chu, Pilch, Bentz, & Behbahani, 1998), a substantial amount of the in-vessel corium
inventory would pour onto the cavity floor. Section 1.3.1 provides more details about the total amount
of corium discharged from the RPV and corium constituents.
Some cavities designs allow for the presence of water before the RPV failure. In addition, severe
accident managements strategies allows for water injection into the reactor cavity before vessel
breach. Under these circumstances, the following fuel-coolant interactions can occur (Park & al.,
1992) (CSN, 1997).
• If an ex-vessel steam explosion occurs, the melt that is involved in the explosion process
may be dispersed out of the cavity. However, the melt not directly involved in the explosion
can still form a particulate debris bed. As mentioned in Section 1.4.7 debris bed coolability
and ex-vessel steam explosion are beyond the scope of this report.
• If there is no energetic ex-vessel steam explosion, the melt jet can break up in the water pool
by hydrodynamic forces. As a result, a portion of the initial melt will form a particle bed,
whereas the remaining will form a cake. As it is shown in Section 1.4.7 melt jet break up and
the associated debris bed coolability issues are beyond the scope of this work.
Figure 1.2-1 also shows the phenomenon termed steam spike. This phenomenon is considered in
Section 1.5.2.
Under dry cavity conditions, the debris pours into the cavity and accumulates at its bottom. The
debris transfers heat to the atmosphere by convection and radiation, and attacks the concrete substrate,
possibly leading to the containment melt-through. Sections 1.3.3 and 1.4.1 provide more information
for concrete ablation and basemat melt-through, respectively.
NEA/CSNI/R(2016)15
Figure 1.2-1: Exemplary paths of main severe accident phenomena in the cavity
Recently, interest for another molten core concrete phenomenon has been renewed: the limited
corium spreading in the reactor cavity. This phenomenon consists of an accumulation of the corium
poured into the reactor cavity within a reduced part of the total floor area available for spreading.
Limited corium spreading could occur both under wet and under dry cavity condition and could result
in different concrete ablation profiles in comparison with the ones in total spreading case. Chapters
5 and 6 provide more detailed information about this recently considered phenomenon.
According to Figure 1.2-1, the branch without water in the cavity at the time of melt release
considers the possibility of cavity flooding after vessel breach. Section 1.5.3 discusses the possibilities
to reach a manageable situation when water is poured atop the corium located on the cavity floor.
NEA/CSNI/R(2016)15
1.3. Corium concrete interaction
This section is aimed at providing a basic understanding on the concrete thermal attack by the corium
and the ex-vessel corium cooling mechanisms. In order to get an appropriate understanding of the
present research and analytical activities carried out in the field of core-concrete interactions and
corium coolability, it is needed to obtain an appropriate background on the main characteristics,
e.g. chemical composition, material properties and so on, of the two main substances involved in the
topic addressed by this SOAR: corium and concrete.
The most relevant characteristics of the corium from the perspective of this report are described
in Section 1.3.1. The amount of corium poured onto the cavity floor when the RPV fails and the
corium constituents are also described in this section. It is well known how important is to obtain an
appropriate knowledge of a number of material properties in the field of the severe accidents. Thus
Section 1.3.1 also deals with the corium thermophysical properties within the field of core concrete
interactions.
The main characteristics of the other relevant substance for this report, i.e. concrete, are described
in Section 1.3.2. More specifically, concrete constituents, and chemical composition of the concretes
used in LWR, some insights about the thermal response of the concrete under high heat fluxes and the
thermophysical properties of concrete are the topics addressed in this section.
The concrete ablation phenomenon is described in Section 1.3.3. Because of their potential high
impact on the plant safety two specific topics are also included in this section. The first one is
concerned with the recent experimental evidence of the possible asymmetrical concrete erosion in
some cases, i.e. some experiments have shown that the radial erosion rate could be faster than the axial
erosion rate. The second topic included in this section deals with the impact in plant safety of the
possible melt stratification during the MCCI process. More specifically, the influence on the timing of
the basemat melt through of the stratified vs. homogeneous melt is briefly discussed.
Finally, Section 1.3.4 describes with some detail the main ex-vessel corium coolability
mechanisms, namely: bulk cooling, water ingression, melt eruptions and crust breach.
1.3.1. Corium characteristics
Corium is a molten mixture of fuel material, partially or totally oxidised cladding, non-volatised
fission products and various structural materials. The main constituents of in-vessel corium are UO2,
ZrO2, Zr and steel.
Section 1.3.3.2 provides an example of a typical corium composition for a 900 MWe PWR. In
this case, the total amount of corium discharged onto the cavity floor is ≈ 150 tonnes. Oxidic corium
amounts to ≈ 100 tonnes (roughly, 67% of the total corium inventory). The remaining ≈ 50 tonnes
(one-third of the total corium inventory) forms the metallic phase of the corium, being Fe the main
contributor to this corium phase: 35 tonnes, accounting for 70% of the total corium metallic mass. It
should be kept in mind that the amount of molten steel varies greatly, depending on scenarios. As an
example, it can exceed 100 tonnes (for up to 175 tonnes of oxides) in EPR, in which there is a heavy
reflector (Nie, 2005).
In Three-Mile-Island 2 accident, a corium with the following average composition has been
found in the vessel lower head (Akers & McCardell, 1989):
77 wt% UO2, 17% ZrO2, 2% Ag, 1% Fe, 1% Cr, 1% Ni, 1% In
Note that the oxidic phase accounts for the 94% of the total TMI-2 in-vessel corium mass.
As for BWR, the in-vessel corium is richer in metallic content than the corresponding one to
PWR. (Greene, Hodge, Hyman, & Tobias, 1991) analyses the performance of BWR Mark III
NEA/CSNI/R(2016)15
containment during a short-term SBO. According to this analysis, the RPV fails at 3h38 min and the
following data for the ex-vessel corium composition are provided in Table 1.3-1.
Table 1.3-1: Typical ex-vessel corium composition for BWR
Constituent Mass (tonnes)
Oxides
UO2 163
ZrO2 17
Others 0.2
TOTAL 180.2 47%
Metals
Fe 121
Zr 48
Cr 23
Ni 10
B4C 0.3
TOTAL 202.3 53%
TOTAL
382.5
OXIDES + METALS
(SNL, 2012) provides the following data for the debris bed composition prior to lower head
failure for an unmitigated long-term SBO. Lower head failure occurs at 19.7 hours. Debris bed
accumulated in the lower head of the RPV represents the mass of nearly the entire core plus structural
materials below the core. This debris bed is composed of a mixture of molten stainless steel (~32% by
mass), unoxidised zirconium (~12%) and particulate debris containing uranium dioxide and metallic
oxides (the remainder). Failure of the lower head results in the rapid ejection of over 300 metric tonnes
of core debris onto the floor of the reactor pedestal in the drywell.
Table 1.3.2 provides typical molar U/Zr ratio for some LWRs.
Table 1.3-2: Molar U/Zr ratio for some LWRs
U/Zr Reactor Type
0.65 US Boiling Water Reactor
1.64 US PWR (Farmer, Spencer, & Aesclimann, 2000)
1.52 French 900 MWe PWR (Baïchi, 2001)
1.37 TMI-2 PWR
1.24 EPR (Nie, 2005)
1.3.1.1. Corium chemical thermodynamics

The list of physical and thermodynamic properties which plays a role in corium progression during a
severe accident is quite impressive: liquidus and solidus temperatures, enthalpy, density, viscosity,
thermal conductivity, emissivity, surface tension and so on. Additionally, most of these properties vary
significantly with the composition and/or the temperature of the mixture. As a result, a prioritisation
should be made according to their importance for the safety in order to conduct research activities
aimed at obtaining appropriate data for these variables. As far as the ex-vessel corium behaviour and
NEA/CSNI/R(2016)15
cooling is concerned, the knowledge of phase diagrams and viscosities are generally considered as the
most relevant variables.
Chapters 3 and 4 of this report provide detailed information about the modelling of relevant
material thermophysical and thermodynamic properties within the field of corium concrete
interactions.
Dedicated thermodynamic databases (Fukasawa, Tamura, & Hasebe, 2005), (Guéneau, et al.,
2006), (Cheynet, NUCLEA, 2007), (Bakardjieva, et al., 2010) have been developed and are coupled
with Gibbs Energy minimisers in order to estimate the phases present in equilibrium for any corium
composition and temperature.
Now, data about some relevant material properties are provided. UO2 melting point is at
3 120 ± 30 K (Fink & Petri, 1997). This value can only be applied to non-irradiated fuel. Irradiation
tends to increase the oxygen stoichiometry to UO2+x and decrease by at most a few hundreds of K the
liquidus temperature. The minimum liquidus temperature in the UO2-ZrO2 system is of the order of
2 840 K (Cohen & Schaner, 1963). When concrete is mixed with corium, the solidus temperature is
lowered to about 1 400 K, as shown in Figure 1.3-1. The solidification range spans over more than
1 000 K for most of corium-concrete mixtures, which facilitates corium spreading.
Figure 1.3-1: Typical corium–concrete pseudo-binary phase diagram calculated with GEMINI2 and
NUCLEA07 (for the silica-rich concrete used in VULCANO) (Journeau & Piluso, 2012)
1.3.2. Concrete characteristics

Concrete is a complex mixture of mainly cement, water, aggregates and additives (in a small
amount). Cement, usually in powder form, acts as a binding agent when mixed with water and
aggregates. Water is needed to chemically react with the cement (hydration) and to provide
NEA/CSNI/R(2016)15
workability to the concrete. Typical concrete is made out of two different types of aggregate
substances: a coarse aggregate (usually gravel) and a fine aggregate or sand. Sand 1 or fine aggregate is
generally defined as aggregate with a size lower than 4.75 mm. Coarse aggregate are above 2 mm and
seldom beyond 25-40 mm, therefore there is an overlap. Although aggregates are generally thought of
as inert filler within a concrete mix, a closer look reveals the major role and influence that aggregate
plays in the properties of both fresh and hardened concrete. Changes in gradation, maximum size, unit
weight, and moisture content can all alter the character and performance of the concrete mix.
Chemical additives can improve concrete performance by modifying its characteristics and
enhancing workability. Chemical additives have to be added during the concrete manufacturing
process, under the appropriate conditions, at the convenient speciation and at the exact quantities. Air-
entraining agents are one example of chemical additives. Air entrainment is the process whereby many
small air bubbles are incorporated into concrete and become part of the matrix that binds the aggregate
together in the hardened concrete. Entrained air must be used in all concrete that will be exposed to
freezing and thawing and deicing chemicals. Additionally, air-entrained concrete is more workable
than non-entrained concrete. As other examples, the addition of gypsum 2 retards the curing and limits
the heat release associated with curing concrete in large placements such as reactor basemat
foundations or the addition of polypropylene fibres enhances concrete porosity during a thermal attack
in order to prevent concrete spalling.
Concrete is usually reinforced by steel rebars. Depending on the design, steel rebars can account
for 6-16% in mass of the structural concrete. Steel rebars are a supplementary source of metal in the
long-term molten core concrete interaction (Journeau & Piluso, 2012). As it is shown in Chapter 6,
MOCKA experiments are providing additional insights on the effect of the steel rebars on the concrete
ablation profile (Foit, Fischer, Journeau, & Langrock, 2014), (Foit J. J., 2015).
Table 1.3-3 summarises the proportion of the different constituents of concrete (Journeau &
Piluso, 2012).
Table 1.3-3: Order of magnitudes of concrete constituent proportions
Water Air Cement Aggregates
Constituents Chemically bond
Total Evaporable (a) Total Sand Gravel
(b)
Vol % 14 – 22 - - 1–6 7 – 14 60 -78 25 – 45 25 – 45
Wt % 5–9 2.5 – 7 1.5 – 4 0 9 – 18 65 – 85 30 – 45 30 – 45
• Evaporable water consists of free and physically bound (absorbed) water. It is defined as the
water released after drying in a furnace at 105 ºC.
• Chemically bound water is forming hydrates which are stable at 105 ºC, mainly tobermorite
gel ~ (CaO)1.62SiO2H2O)1.5 and calcium hydroxide Ca(OH)2.
In fact the water proportion can vary as it depends also on the operating temperature and level of
humidity.
1. Sand is usually associated with silica but in some regions fine aggregates are produced only from crushed
limestone.
2. The potential of sulfur release from the gypsum is of considerable interest for both source term
considerations and for the performance of passive autocatalytic hydrogen recombiners whose catalysts are
irreversibly poisoned by sulfur.
NEA/CSNI/R(2016)15
More detailed information about the chemical composition of typical concretes used in LWR is
shown in Table 1.3.4. Note that concrete for LWR is mainly comprised of silica (SiO2) and limestone
(CaCO3).
In addition to above-listed typical concretes used for the civil structure in LWR plants, other
types may be applied for specific purposes, like in Russian designs to shield the radiation in the lower
cavity. Such concretes are typically fabricated with serpentine aggregates. Serpentine is a mineral of
the composition (MgO)6_(SiO2)4_(H2O)4 with a substantial content of water in a crystalline structure.
Another specific application of concrete, is its use as sacrificial material for severe accident mitigation,
as in the EPR™ reactor (Nie, 2005), where, the aggregates used for the concrete in the reactor pit
contain a large fraction of hematite (Fe2O3-rich ore).
Table 1.3-4: Chemical compositions 3 of typical concretes used in LWR (values in w %)
Specie Siliceous Basaltic LCS
(Roche, Leibowitz, & Fink, (Roche, Leibowitz, & Fink, (Roche, Leibowitz, & Fink,
1993) 1993) 1993)
SiO2 68.99 52.80 28.27
CaO 13.47 12.51 26.02
Fe2O3 1.00 8.58 1.64
MgO 0.7 3.03 9.62
Na2O 0.69 3.06 1.09
Specie Siliceous Basaltic LCS
(Roche, Leibowitz, & Fink, (Roche, Leibowitz, & Fink, (Roche, Leibowitz, & Fink,
1993) 1993) 1993)
K2O 1.41 1.41 0.57
Al2O3 4.4 11.26 3.48
CO2 4.23 - 21.41
H2O 3.68 4.23 6.13
(ev +
chem.)
Depending on the region where aggregates are mined the composition of calcareous aggregates
can vary. If calcareous aggregates come often from calcite limestone it is possible to find also
dolomite MgCa(CO3)2 which decomposes at lower temperatures. Siliceous aggregates can span also a
large range of compositions and properties. Basaltic aggregate is, for example, a eutectic material that
is relatively fluid when melted whereas higher silica content aggregates like granite can be higher
melting and much less fluid when they melt.
The response of concrete to high heat fluxes is complex. Concrete is an inhomogeneous material
which undergoes changes in composition as it is heated. Several authors have described the reactions
that occur with an increase of temperature in concrete. Below, the most relevant chemical processes
during the concrete heat-up are shown (Journeau & Piluso, 2012).
• 30–105 ºC. The evaporable water and part of the bound water escape. It is generally
considered that the evaporable water is completely eliminated at 120 ºC.
• 180–300 ºC. The loss of bound water from the first stage of decomposition of the CaO-SiO2-
H2O and carboaluminate hydrates takes place.
3. Figures may vary according to the exact composition of sands, aggregates and cements.
NEA/CSNI/R(2016)15
• From 300 ºC. Increase in porosity and microcracking.

• 374 ºC. Critical point of water above which no free water is possible.
• The EUROCODE (AFNOR, 2005) proposes to consider for simplified calculations that
concrete above 500 ºC has no mechanical strength.
• 573 ºC. α → β transformation of quartz aggregates.
• 700 – 900 ºC. Decarbonation of calcium carbonate in cement paste and carbonate aggregates
which forms CaO powder.
• 720 ºC. Second stage of decomposition of CaO -SiO2-H2O.
• From 1 100 to 1 250 ºC. Start of concrete melting. More details about the concrete melting
temperature are provided below.
Concrete heating process also changes its thermo-physical properties, mainly due to drying and
decomposition (Journeau & Piluso, 2012). Concrete density is generally around 2 000 –
2 500 kg m 3. The specific heat of concrete typically lies between 900 J kg-1 K-1 (at room temperature)
and 1 100 J kg-1 K-1 above 400o C. These are averaged values that may be higher in certain
temperature intervals, e.g. when chemical decomposition or evaporation processes take place. The
thermal conductivity of concrete is generally low (a few W/(m.K) and significantly decreases with
temperature.
(Roche, Leibowitz, & Fink, 1993) have measured the solidus and liquidus temperatures for
several types of concretes. Table 1.3-5 shows the results.
Table 1.3-5: Solidus and liquidus temperatures for some typical concretes
Siliceous Lime-siliceous Calcareous
Solidus temperature (ºC) 1 130 1 120 1 220
Liquidus temperature (ºC) 1 250 1 295 2 305
As it was previously mentioned, concrete starts melting around 1 100 to 1 250ºC. If the formation
of a liquid concrete layer is considered as the reference transition, then the effective melting
temperature generally corresponds to the temperature at which 30 – 50 vol% is liquid.
The decomposition enthalpy to bring concrete from room temperature to total melting is
generally between 1.8 MJ kg-1 to 2.5 MJ kg-1. To provide an order of magnitude, it requires
2 MJ kg-1 to heat a typical siliceous concrete from room to melting temperature, while 2.5 MJ kg-1
are needed for a limestone-rich concrete. This higher value is both due to the decarbonation latent heat
and the higher effective melting temperature.
When applying corium concrete heat transfer models based on a global heat transfer coefficient,
the relevant temperature at which the concrete ablates, the so-called ablation temperature, needs to be
specified. Unfortunately, the ablation temperature is not a thermodynamic quantity, at least not a
known thermodynamic quantity (Epstein, 1997). In the Sandia MCCI test programme the concrete
ablation front was identified with the 1 600 K isotherm as determined by the thermocouples cast into
the concrete basemat. In the Argonne ACE test programme a thermocouple reading of 1 673 K was
assumed to indicate the passing of the concrete ablation front. For the theoretical calculations shown in
(Epstein, 1997), an ablation temperature of 1 600 K was chosen. As a general procedure that can be
applied for various concretes (Nie, 2005) proposed to use the temperature at which the volumetric
liquid fraction in the concrete slag has reached 50%. Correspondingly, the concrete decomposition
enthalpy would be given by the heat required to lift the concrete from room to the thus-calculated
NEA/CSNI/R(2016)15
ablation temperature. Due to the observation that silica gravel was not melted at ablation
(e.g. (Journeau, et al., 2012), it has also been proposed to consider ablation temperature as the
temperature at which a significant liquid fraction has been reached for the mortar (cement and sand).
It also should be noted that the heat required to ablate concrete is relatively insensitive to the
chosen temperature selected as the ablation temperature in (Epstein, 1997).
1.3.3. Concrete ablation
The attack of concrete by core debris is largely thermal in a LWR (Bradley, Gardner, Brockmann, &
Griffith, 1993). Energy is generated in the core debris from radionuclide decay and from chemical
reactions and may be lost either its top surface or to the adjacent concrete. In either case, as long as the
heat source is sufficiently large, the situation rapidly approaches a quasi-stationary state where the
losses from the core debris balance the internal sources.
It has been experimentally shown that thermal ablation was the major process governing core-
concrete interaction. Molten core concrete interactions have a highly complex phenomenology
coupling concrete high-temperature behaviour, molten pool thermal hydraulics, thermochemistry,
mechanics, etc. (Journeau & Piluso, 2012).
The corium typically contacts concrete at an initial temperature higher than 2 400ºC. The first
instant of core-concrete interactions are controlled by the melt overheat which heats the concrete,
leading to the liberation of steam and carbon dioxide and to the initial melting of concrete. Concrete
decomposition bubbles will enter the corium pool, agitate it and, when available, oxidise the metal
components, basically in the sequence Zr, Cr, Fe (Powers, Brockmann, & Shiver, 1986). Oxidation of
Zr and Cr is a highly exothermic process that contributes in the initial interaction phase to significant
energy release in the melt. Heats of reaction for Zr oxidation are
Zr + 2H2O → ZrO2 + 2H2 + 600 kJ mol-1 Zr
estimated at 2 500K with NUCLEA091 thermodynamic database and GEMINI2 from
THERMODATA (Cheynet, 2007).
Zr + 2CO2 → ZrO2 + 2CO + 600 kJ mol-1 Zr
In addition to the reactions of metals with gases from the concrete, the so-called condensed phase
reactions between oxides and metals should be also considered. These condensed phase reactions were
highlighted by the SURC-4 experiment (Bradley, Gardner, Brockmann, & Griffith, 1993), (Copus,
Blose, Brockmann, Gopmex, & Lucero, Core-Concrete Interactions using Molten Steel with
Zirconium on a basaltic Basemat: The SURC-4 Experiment, 1989). SURC-4 showed a vigorous
interaction between metallic zirconium and a siliceous concrete with low gas content. In order to
explain this interaction, condensed phase chemical reactions between the Zr and concrete molten
oxides should be taken into account. The driving chemical reaction was:
Zr + SiO2 → Si + ZrO2 + 2.1 MJ/kg Zr
At high temperature, this reaction can be accompanied by an incomplete oxidation reaction
leading to the formation of SiO which leaved the pools surface as gas.
Condensed phase reactions are particularly important for core debris interactions with higher
silica, lower gas concretes such as the one used in the SURC-4 experiment. They are less important for
calcareous concrete that have low silica content and high gas content. Chapter 2 provides more details
about the SURC-4 experiment.
After this initial transient, lasting from 10 min to a few hours depending on reactors and
scenarios, the main source of heat will be the decay heat, which is mainly (>90%) generated in the
oxide phase.
NEA/CSNI/R(2016)15
The so-called spalling phenomenon, e.g. the breaking off of the layers or pieces of concrete from
a concrete surface exposed to high heat fluxes, plays a minor role in MCCI (Journeau & Piluso, 2012),
(Sevón, et al., 2010).
The ablation rate is, at the first order, the ratio of the heat flux to the enthalpy needed to heat and
melt a unit volume of concrete:
φ
v=
ρ (H concrete melting − H 0 )
Equation 1.3-1
Where ν is the ablation rate, is the heat flux, ρ the solid concrete density, Hconcrete melting the
φ
enthalpy at the “concrete melting” temperature (see Section 1.3.2) and H0 is the concrete enthalpy at
room temperature. Note that Equation 1.3-1 highlights that core-concrete interaction is characterised
by an imposed power, and not by an imposed temperature (Sehgal, 2011). This approximate
expression for the concrete ablation rate is valid as conduction in the still solid concrete is so low that
the ablation front progresses as least as quickly as the conduction heat wave, thus, conductive heat
losses are negligible at the first order. Some rough data about concrete ablation rates are provided
below (Journeau & Piluso, 2012), (Sehgal, 2011). For a typical PWR of 1 300 MWe, about 120 t of
oxides (UO2, ZrO2, …) and 80 t of metal (Fe, Cr, Ni, Zr) could relocate into the reactor cavity. For a
6 m diameter cavity, this leads to a melt height of the order of 1 meter. For a 1-m thick corium pool
and an initial decay heat of 15 MW, considering an isotropic heat flux distribution of 200 kWm-2, a
simple calculation gives an average initial ablation rate of the order of 20 cm/hr. Typical long term
ablation rates are several centimetres per hour.
The concrete decomposition products are miscible with corium oxides, so, during its interaction
with concrete, the melt will be continuously enriched in concrete decomposition products (mainly CaO
and SiO2). If several metres of basemat are eroded, silica and/or calcia can become the major
constituents of the melt, as it occurred during Chernobyl accident (due to mixing of corium with
biological protection sand), with a typical melt composition of the so-called “brown-chocolate”
(Pazukhin, 1994) corium, in which concrete and mineral protective materials have been mixed:
57% wt% SiO2, 11% Al2O3, 8% UO2, 6% ZrO2, 6% CaO, 6% MgO, 5% Na2O, 1% Fe2O3
It is thus clear that there is not a unique corium-concrete mixture composition but a continuum of
possible compositions depending on the reactor, the accident scenario and the instant in this scenario.
Therefore, it is not possible to measure the corium-concrete mixture properties for every case and
models have to be derived in order to estimate both its thermodynamic and thermophsysical
properties.
Chapters 3 and 4 provide detailed information about modelling of material thermophsysical and
thermodynamic properties.
The heat transfer mechanisms are strongly dependent on the (varying) compositions of the melt
and concrete. The heat flux through a given interface is given by
∅𝒊 = 𝒉𝒊 �𝑻𝒑𝒐𝒐𝒍 − 𝑻𝒊 � Equation 1.3-2
where hi is the convective heat transfer coefficient (mixed convection due to the sparging gases)
and Ti is the interface temperature for the considered boundary. A number of heat transfer coefficient
correlations have been developed for the heat transfer to the pool walls as well as for heat transfer
between two immiscible liquid layers. More detailed information about these heat transfer coefficient
correlations can be found in Chapters 3 and 4.
NEA/CSNI/R(2016)15
1.3.3.1. Radial vs axial ablation

Recent experiments on MCCI addressed mainly two subjects, namely the 2D aspects of the ablation,
e.g. axial versus radial erosion, and the role of crust formation and melt segregation upon onset of
solidification. These subjects are especially important in the analysis of the long-term erosion and
cavity formation. This section discusses the topic of radial vs. axial concrete erosion.
One of the major findings of these tests with pure oxidic melts on 2D erosion is the observation
that dry ablation tests with silica-rich concretes tend to present at least after some transient phase an
anisotropic ablation pattern, namely, radial rate ablation is faster than the axial ablation rate, resulting
in a more efficient ablation of the sidewalls compared to downwards ablation. This is the major result
of the CCI and VULCANO VB-U tests series. The synthetic views of ablation profile along 11 section
axes of Chernobyl basaltic basemat, resulting from ISTC CHESS R&D shows that in most of the
cases, the same preferential lateral erosion is observed (factor : ~2-4) (Pazukhin, 1994).
At the opposite, the tests with limestone-rich concrete (e.g. CCI-2, CCI-4) showed a more
isotropic ablation.
Chapter 2 provides detailed descriptions for these experiments. Chapter 6 provides insights about
the extrapolation of these experimental results to reactor scale. As for Gen. II reactors, safety
implications of a potential radial melt-through and the subsequent corium spreading in rooms adjacent
to the reactor cavity are addressed in Chapter 5.
1.3.3.2. Homogeneous vs stratified pool melt
According to (Bradley, Gardner, Brockmann, & Griffith, 1993), experimental evidence shows that the
various oxidic species in the melt are highly miscible. The same holds true for the metallic species. In
absence of bubbling flow, it should be expected that the core debris will stratify into distinct layers
based on their relative densities.
The stratification of metal and oxide melts under gas bubbling is a complex phenomenon and it
remains a real challenge (Sehgal, 2011). This phenomenon is influenced by the density differences, by
the viscosities, by the size distribution of metal droplets and by the convection processes in the melt
induced by the gas released from concrete.
In order to illustrate the safety relevance of the corium melt stratification a number of ASTEC
calculations are described below (Sehgal, 2011). Table 1.3-6 shows the main data for the reactor
geometry and the initial conditions and corium composition considered in the ASTEC calculations.
Table 1.3-6: Initial Conditions, reactor Pit Geometry and evolution of the decay power
Initial corium inventory (tonnes)
UO2 82
ZrO2 19.5
Fe 35
Cr 6
Zr 4.8
Ni 4
Reactor cavity radius 3m
siliceous concrete
Concrete characteristics
with 6% Fe.
Basemat thickness 3 m to 4 m
Initial corium temperature 2, 673 K
NEA/CSNI/R(2016)15
Time after scram Decay power (W/Kg U)

1h 283
3h 227
7h 190
15 h 157
20 h 145
50 h 108
9 days 63
Note, that no corium quenching by flooding was taken into account in the present calculations.
Because of the present large uncertainties on the pool configuration and its possible evolution,
three different scenarios concerning the pool configuration were considered in the ASTEC
calculations.
• Scenario 1. This scenario assumes that a homogeneous oxide-dominated melt configuration
is maintained during the whole MCCI phase. This configuration leads to rather slow concrete
erosion.
• Basemat melt-through occurs around 9 days for a basemat thickness of 4m.
• Scenario 2. It is assumed in this unlikely scenario that a fixed stratified melt configuration is
maintained throughout the calculations what is a very conservative assumption. This
configuration leads to the fastest concrete erosion.
• Timing for basemat melt-through ranges between 14 hours only and 3.4 days, depending on
the choice of the solidification temperature and on the basemat thickness.
• Scenario 3. Finally, a more realistic scenario is assumed with four successive phases. In this
scenario, the pool configuration is initially stratified with the oxide layer below the metal
layer because of the higher initial oxide density, followed rapidly by a homogenous pool.
Then, the pool is again stratified but with the metal layer below the oxidic one, and finally
the pool becomes again homogeneous after disappearance of the metal layer due to
oxidation. This evolution of the pool was evaluated using pool configuration switch criteria
consistent with BALISE experiments. The consequence of modelling this sequence of pool
configurations is the large delay of the predicted melt-through time by more than 24 hours,
compared to that obtained in the case of a steady stratified metal/oxide configuration.
• Scenario 3 results are deemed as very conservative because pessimistic boundary conditions
were chosen and the BALISE stratification criterion is very conservative.
Similar scenarios were run assuming LCS concrete. Longer melt through times were obtained for
LCS concrete due to two peculiarities of the type of concrete. On the one hand, LCS concrete poses
higher gas content than siliceous concrete (see Table 1.3.4). Higher concrete gas content leads to
higher superficial gas velocities that, in turns, result in reducing the time duration of the stable
stratified configuration phase or even suppressing it. On the other hand, ablation enthalpy for LCS is
higher than the one for the siliceous concrete due to the high decomposition enthalpy of carbonates,
causing a slower ablation velocity (see Section 1.3.2).
It is important to keep in mind that these layered configurations represent an idealised modelling
assumption: so far these layered configurations have never been observed experimentally under
prototypical conditions. However, usually both stratified and mixed configurations are considered in
the analyses.
NEA/CSNI/R(2016)15
After the end of the MCCI, in the absence of gas induced mixing and heating, the oxide and metal
layer should stratify according to their representative densities. A corresponding behaviour was
observed in the BETA, LACOMERA, MOCKA and SICOPS experiments. Deviating from this, the
recent VULCANO VB-US tests showed vertical regions filled with steel tongue as well as large
floating metallic drops. This phenomenon may either be explained by peculiarities of the gas flow (and
probably not of the heating) or by the presence of convective plumes observed within the melt
(Journeau, et al., 2012).
1.3.4. Ex-vessel corium coolability mechanisms
Cooling of ex-vessel core debris using an overlying water pool was previously investigated in the Melt
Attack and Coolability Experiments (MACE) Program at Argonne National Laboratory (ANL), under
the sponsorship of EPRI (see Section 2.4.5). Large-scale integral experiments (with a test section
lateral span up to 1.20 m) were conducted with melt masses of reactor prototypic materials ranging up
to 2 metric tonnes. These experiments suggested various heat transfer mechanisms that could provide
long-term debris cooling. More recently, ANL launched the NEA-MCCI Program (under the auspices
of the NEA) to address two specific issues. The first issue relates to several debris cooling
mechanisms identified in the MACE Program (specifically, the relative roles of these mechanisms to
achieve overall coolability), while the second issue relates to long-term, two-dimensional concrete
ablation by ex-vessel core debris under both dry and flooded cavity conditions.
Several heat transfer mechanisms have been identified through experiments that can contribute to
long-term corium coolability. These mechanisms are summarised in Table 1.3-7, while a physical
illustration is provided in Figure 1.3-2. In general, when water is introduced atop an MCCI, the
question of whether or not a significant amount of the thermal energy is initially removed will depend
upon whether a stable crust is able to form that inhibits heat transfer from the melt to the water over
layer. For a stable crust to form, two necessary conditions must be met: i) a thermal condition, viz., the
melt/water interfacial temperature must fall below the corium freezing temperature, and ii) a
imposed by the agitated melt. If either of these two conditions is violated, then stable crust formation
is precluded. In this bulk cooling regime, efficient melt/water heat transfer occurs due to conduction
and, predominately, radiation heat transfer across the agitated (i.e. area enhanced) melt/water interface,
in addition to entrainment of melt droplets into the water over layer.
OVERLYING WATER
δCL
CONDUCTION LIMITED UPPER CRUST -
PERMEABLE WITH RESPECT TO GAS
FLOW, IMPERVIOUS WITH RESPECT TO
WATER INGRESSION
CORIUM MELT POOL
POSSIBLE LOWER CRUST
DECOMPOSING
CONCRETE
DRAWING: TRADITIONAL MCCI MODEL

WITH WATER PRESENT
DRAWING NO.: MCCI167
DRAWN BY: D. KILSDONK 2-4746
DATE: 10/18/02
FILE: Traditional_MCCI_Model_with_Water.DWG(AC28)
(a) (b)
Figure 1.3-2: Traditional view of (a) CCI with conduction-limited upper crust at melt water interface;
(b) CCI with water ingression and melt eruption cooling mechanisms
NEA/CSNI/R(2016)15
decreases. Thus, a point will eventually be reached at which the thermal and mechanical thresholds for
zone and water layer. The physical configuration at this point consists of an ongoing MCCI at reduced
temperature with a crust atop the melt (Figure 1.3-2(a)). The crust will be characterised by some
degree of porosity, or cracks, owing to the necessity of venting concrete decomposition gases.
After incipient crust formation, completion of the quench process can only be achieved if one of
two conditions is met. The first condition is that the melt depth lies below the minimum depth at
which decay heat can be removed via conduction heat transfer alone (~10 cm). This case is trivial, and
is not addressed in any further detail. The second condition is that water is able to penetrate into the
debris by some mechanism to provide sufficient augmentation to the otherwise conduction-limited
heat transfer process to remove the decay heat. Three potential mechanisms have been identified
through experiments which provide pathways for water to penetrate the debris. The first mechanism is
water ingression through interconnected porosity or cracks (Figure 1.3-2 (b)). This process relies on
crack propagation through the material and, as such, is highly dependent upon the mechanical
properties, since thermal stress is a key factor.
Table 1.3-7: Summary of coolability mechanisms observed in MACE integral tests
Mechanism Description Experimental Evidence
Bulk Cooling Melt sparging rate is initially high enough to preclude High heat transfer rates measured during early
stable crust formation at melt/water interface. As a phase of the melt-water interaction. Data
result, high heat transfer rates occur due to conduction indicates that a coherent crust cannot form;
and, predominately, radiation across the agitated (area rather, crust segments are broken up and
enhanced) interface. This phase is terminated when a mixed into melt.
stable interfacial crust forms.
Melt Eruptions Melt dispersal occurs by an entrainment mechanism in Eruptions observed in all tests conducted with
which concrete decomposition gases carry melt limestone-common sand concrete after
through defects in the crust to the overlying coolant. incipient crust formation and in test with
The dispersed material is quenched and forms siliceous concrete with early flooding (CCI6).
coolable particle beds and high surface area volcanic The particle beds are characterised by high
formations. porosity and large particle size.
Water Ingression Corium shrinkage from an initially molten to a fully Melt/water heat flux far exceeds that which
quenched state amounts to ~ 18 vol%. This causes could be transferred by conduction across the
voids/defects to appear in the frozen material. Water (up to 10 cm) thick crusts formed during the
penetrates down through the voids/defects, tests. Post-test measurements indicate that
augmenting the otherwise conduction-limited heat crusts are permeable to both gas and water
transfer process. flows.
Crust Breach Due to water ingression, thick crusts will form and Partial crust failure and relocation events
bond to the reactor cavity walls. These crusts will not observed in MACE integral effects tests.
be stable in the typical ~ 6 m span of most plants. Various structural – mechanical analyses have
Thus, they will periodically fail, leading to renewed shown that crusts will not be stable at reactor
cooling by the above three mechanisms. scale.
The second mechanism is particle bed formation through “volcanic” eruptions. In this case,
concrete decomposition gases entrain melt droplets into the overlying coolant as they pass through the
crust. The entrained droplets then solidify in the overlying coolant and accumulate as a porous particle
bed atop the crust. The third mechanism is mechanical breach of a suspended crust. In particular, the
thick crusts that form from water ingression could bond to the reactor cavity walls, eventually causing
the melt to separate from the crust as the MCCI continues downwards. This phenomenon is caused by
limestone volume shrinkage due to decarbonation (loss of more than 40% of mass and similar
densities of CaO and CaCO3). However, this configuration is not expected be mechanically stable due
to the poor mechanical strength of the crust in comparison to the applied loads (i.e. the crust weight
itself, plus the weights of the overlying water pool and the accumulating dispersed material minus the
NEA/CSNI/R(2016)15
pressure of concrete decomposition gases). Eventually the suspended crust will fail, leading to rapid
ingression of water beneath the crust. This sudden introduction of water will provide a pathway for
renewed debris cooling by the bulk cooling, water ingression, and melt eruption cooling mechanisms.
Debris cooling can be substantially enhanced by a large-scale breach of crust that is formed at the
melt water interface. It is stipulated that an anchored crust (i.e. a crust formed at the melt-water
interface that is attached to the sidewall and otherwise suspended along the span) is likely to fail at the
plant scale because of its low structural strength. The presence of extensive cracks in the crust
(see Figure 1.3-3 as an example) contributes, in large part, to such low strength. Such a crust is
expected to fail under the applied load (namely, the combined weight of the crust, overlying particle
bed, and water pool minus the pressure of concrete decomposition gases) at the plant scale, and to
provide pathways for significant water ingression through macroscopic breaches and, as a result, will
provide additional cooling. The extent to which these various mechanisms can be effective in terms of
thermally stabilising the molten core debris is principally a function of melt depth and the timing of
water addition into the cavity. Section 1.5.3 provides more information about this important safety
issue.
Figure 1.3-3: Extensive crack structure in the mid-plane sample of the SSWICS-3 crust
1.4. Relevant containment failure modes due to ex-vessel corium behaviour
Ex-vessel corium poses a number of threats to the containment integrity as long as the melt
stabilisation is not achieved. This section provides a simple description of the most relevant
containment failure modes arising from core concrete interactions.
1.4.1. Basemat melt-through
Basemat melt-through is a late containment failure mode. This late containment failure mode occurs
when the coolability of the corium accumulated in the reactor cavity floor cannot be achieved. Under
these conditions, the continuous axial concrete erosion by the corium eventually results in the basemat
perforation with possible ground and water contamination.
As mentioned in Section 1.3.3.1, in addition to the axial concrete ablation, radial concrete erosion
should be also kept in mind. Fast radial ablation could also perforate cavity walls, leading to the
spreading of corium into adjacent rooms, e.g. instrumentation room. Whereas corium spreading is
favourable to its stabilisation, the presence of plant specific features such as drains, sumps etc. could
result in additional containment weakness. Also the thermal weakening and possible long term
deformation of load-carrying inner structures should be considered.
NEA/CSNI/R(2016)15
In the previous description of the basemat melt-through as a late containment failure, the term
“basemat perforation” should be understood as the perforation of the containment liner located below
the concrete slab within the reactor cavity. However, some nuclear power plants are designed with
additional concrete slab below the liner. This additional concrete slab could provide an additional
barrier to prevent the ground or water contamination. The credit of this concrete thickness slab as
safety measure is a plant specific issue that should be studied very carefully.
Realistic timing for the basemat melt-through usually ranges from one to several days after the
onset of the severe accident. This timing depends on many factors: plant specific features, such as
configuration of the reactor cavity; concrete type; severe accident sequence characteristics, i.e. amount
of corium discharged from the RPV, level of residual power, metal content of the corium, the presence
of water in the reactor cavity (either before the RPV failure or by pouring water atop the corium after
RPV failure), etc.
A great amount of research and analytical activities have been devoted to studying this late
containment failure mode for many years. These activities have provided an appropriate understanding
of the phenomena involved in corium-concrete interactions and the necessary information to determine
the feasibility of a number of severe accident strategies aimed at preventing this late containment
failure mode. Moreover, the core content of this SOAR is heavily focused on these two severe
accident phenomena. Section 1.6 provides information about the safety importance of these
phenomena within the field of the severe accident.
1.4.2. Containment failure by over-temperature
Containment system consists of several elements. In addition to the primary structure, these include
various components such as the mechanical (hatches, piping) and electrical (instrumentation and
controls) penetrations. There are a large variety of designs for these components and not all can be
tested. Nevertheless, a series of tests of a variety of containment penetrations types has been
conducted in order to determine their leakage characteristics under severe accident conditions
(Hessheimer M.F., 2006).
As an example of this kind of containment failure mode, severe accident analyses conducted on
boiling water reactor (BWR), equipped with Mark I type of containment, have indicated very high
temperatures in the drywell region as a consequence of the molten core concrete interactions (Claus,
1989) (NRC, 1990), (Herranz & al., 2012). Drywell region is the location of the electrical penetration
assemblies and the drywell head.
On the basis of the aforementioned experimental series, it could be concluded the loss of the
containment integrity due to the high temperatures within drywell as a result of the core concrete
interactions.
The SOARCA study for a BWR equipped with a Mark I type of containment (SNL, 2012) and
(Gauntt, et al., 2012) describes a potential failure of a containment mechanical penetration, namely the
drywell head flange. A short description of this potential failure of the drywell head flange is provided
below. Note, that drywell head flanges for other BWR plants may be different, so that the conclusions
of these studies could not be applicable to other plants.
Figure 1.4-1 shows an example of a drywell head flange. Note that the drywell head is removed
during refuelling to gain access to the reactor vessel. According to Figure 1.4-1, the drywell head
flange is connected to the drywell shell with bolts. The flanged connection also has gaskets. Gaskets
are replaced during each reassembly of the reactor vessel head, because they are exposed to constant
temperature and radiation, which contribute to their early degradation.
The drywell head flange bolts are pre-tensioned during reassembly of the head. This pre-tension
also compresses the gaskets in the head flange.
NEA/CSNI/R(2016)15
Figure 1.4-1: Example of drywell head flange connection details
During accident conditions, the containment vessel may be pressurised internally. The internal
pressure would then begin to counteract the pre-stress in the bolts. When the internal pressure
produces counter-stress equivalent to the bolt pre-stress, elastic strain of the head bolts will occur,
allowing a small gap to open up between the mating surfaces of the flange. Further increase in the
internal pressure would result in leakage at the flanged connection. Because the bolt stress is in the
elastic range of the stress-strain curve, when the containment pressure decreases, the gap reduces as
the bolt contracts and the leakage area is thereby reduced. In reality, some permanent strain may take
place in the flange region and seal degradation may also take place such that the leakage area may not
reduce to zero as the containment pressure decreases.
Figure 1.4-2 shows the leak area of the drywell head flange vs. drywell pressure. Figure 1.4-2
also contains the results obtained by the parametric analyses performed for this analysis. The selected
uncertainty parameters were: the modulus of elasticity (E) of the drywell head bolts, the torque
coefficient associated with the pre-tensioning of the head bolts (K), and the rebound thickness of the
drywell head gasket. More detailed information for this uncertainty study is included in (SNL, 2013).
Figure 1.4-2: Drywell head flange area versus pressure
NEA/CSNI/R(2016)15
1.4.3. Failure of the reactor pressure vessel pedestal in BWR

The reactor pressure vessel (RPV) is supported by the reactor pedestal in BWR. Failure of the pedestal
will result in gross motion of the RPV that could lead to the containment failure under specific
circumstances. A number of phenomena could lead to the failure of the pedestal, e.g. direct
containment heating, the occurrence of an ex-vessel steam explosion, the reactor pressure blowdown
from high pressure or the concrete erosion by molten core. Only the last mechanism will be considered
in this section. (Payne & al., 1990); (Brown, y otros, 1990); (Breeding, y otros, 1991) and (Harper,
1991) address this safety issue. The following discussion is focused on BWR equipped with Mark III
type of containments.
Pedestal failure is defined as the loss of support of the RPV such that gross motion of the vessel
results. If the vessel fails, core debris is released into the reactor cavity. Assuming the debris is not
coolable, it will participate in molten core-concrete interaction. During molten core-concrete
interaction, both the concrete and the extensive mesh of rebar in the pedestal region are eroded. If the
erosion into the pedestal wall is extensive, the pedestal may collapse from the load imposed by the
RPV and the shield wall. As previously said, failure of the pedestal will result in gross motion of the
RPV. Several large pipes are attached to the RPV that penetrate the drywell (e.g. steam lines and
feedwater lines). The motion of the RPV, and hence the motion of these pipes, can damage the
penetrations and fail the drywell boundary. The integrity of the drywell boundary can also be impaired
by damage of the drywell liner that results from the RPV motion. The combination of these events can
establish pathways that bypass the suppression pool. Relevant events, variables and phenomena
involved in the potential failure of the pedestal are: the presence of water in the cavity floor, the
amount of corium discharged from the RPV and the flow rate, corium superheat and amount of
unoxidised metal in the molten corium and decay heat.
1.4.4. Mark I containment failure by melt-attack of the liner
This early containment failure mode is specific of BWR equipped with a Mark I type of containment.
In this type of containments, relocation of molten core debris into the drywell cavity pedestal region
occurs upon RPV failure in a low-pressure severe accident. The liner failure issue arises because the
tight containment floor geometry relative to the large core inventory can lead to molten core debris
flowing a short distance and contacting the drywell liner. Failure of the containment liner shell at this
location could lead to rapid blow down of the drywell atmosphere into the reactor building and
subsequently into the environment without the benefit of the suppression pool scrubbing.
(Theofanous, Yan, & Podowski, 1993) provide with the details of the resolution of this important
safety issue. This study shows that flooding the drywell with water significantly reduces the thermal
load on the drywell liner to the point at which liner failure can be highly unlikely (conditional
probability less than 10-3 given a core melt accident). Substantial amount of water in the drywell
could be achieved by maintaining operational the drywell sprays throughout the severe accident, what
in turn underscore the importance of the implementation of appropriate severe accident management
strategies.
1.4.5. Coolability of the widely dispersed debris into the containment
In some cases, severe accident progression might result in the widespread debris dispersal into the
containment. This situation could occur in case of high pressure melt ejection, steam explosions,
vessel rocketing etc. The widespread debris dispersal might form a film on the containment boundary
structures with potential to challenge the containment integrity. This safety issue was analysed in
(NRC, 1990), (NRC, 1993) and (Harper, 1994) concluding that very little was known at that time
about the expected corium distribution under the previous energetic events. Nevertheless, if corium
comes to rest in a thin uniform layer air cooling will suffice. On the other hand, it is possible that drifts
of corium particles might accumulate in corners, in the wall-floor angle, and so on, that would be
NEA/CSNI/R(2016)15
enough to reheat and start core concrete interaction. Debris coolability was considered as very
uncertain in such scenarios.
(Alsmeyer H. , et al., 1995) also highlighted the safety concerns raised by the corium dispersion
in the containment, including the possibility of a delay loss of containment integrity.
The GAREC working group was also concerned with the problem of corium dispersion in the
containment (Seiler, et al., 1999).
Taken advantage of two plant specific experiments conducted for two PWR designed by
Westinghouse, (Hammersley, Cirauquib, Faigc, & Henrya, 1996) provides experimental evidence for
the potential of liner ablation when it gets in contact with the debris dispersed throughout the
containment. Due to a plant specific design feature in one of these two PWRs, there exists a possibility
of direct contact between the liner and the dispersed debris in the containment annular compartment
that might lead to a loss of the containment integrity.
These two experiments were conducted with stimulant materials and the containment liner was
simulated with a liner plate made of stainless steel. Liner plate was directly impacted by molten debris
in both experiments and once removed the solidified deposits and inspection of the liner plate surface,
no ablation or degradation was observed. Additionally, these two experiments provided the basis for
the analytical work shown in (Robledo & Lantaron, 1999), aimed to determine the needed conditions
leading to the liner failure by deposits on the liner.
(Meyer, Albrecht, Caroli, & Ivanov, 2009) show the most relevant findings obtained in the
integral tests performed in DISCO-H facility employing an iron – alumina melt as corium stimulant,
steam in the RPV and a prototypic atmosphere in the containment.
As for the potential impact on the containment integrity of the widely particle debris dispersed
into the containment, (Meyer, Albrecht, Caroli, & Ivanov, 2009) provide the following information:
• Tests with EPR geometry.
• “Debris in the DISCO vessel was recovered from (….) the dome surface and structures
inside the containment. (…) The debris in the containment was generally collected as small
particles on the wall, the top cover and on the floor”.
• “The particles larger than 5 mm are generally flat, because they hit a wall when they are still
liquid, the particles smaller than 2.5 mm have a spherical shape with a cavity in the centre”.
• Test with VVER-1000 geometry.
• “Less than 2% of the initial melt mass reached the containment as fine dust, with a median
diameter of ≈ 0.15 mm”.
• Tests with Konvoi geometry.
• “The debris in the containment was generally collected as small particles on all horizontal
areas, such as the subcompartment covers and hoses. (…) Only very fine particles could
enter the containment dome, here the mean diameter is 0.08 mm. These particles probably
were already solid when they entered the containment dome”.
1.4.6. Containment over-pressurisation
When the RPV fails at low pressure, the debris pours into the cavity and accumulates at the bottom of
the cavity. In the absence of corium cooling by water, the interaction of the corium with the cavity
concrete basemat results in ablation of the concrete substrate and the subsequent generation of non-
condensable gases, namely, steam, H2, CO2 and CO. As shown in Table 1.3.4, the quantities of steam
NEA/CSNI/R(2016)15
and CO2 released vary with the concrete type. Note that basaltic and siliceous concrete contains much
less CO2 gas than LCS concrete.
In addition to the containment basemat melt through, concrete ablation and the subsequent non-
combustible gas generation could lead to two additional potential containment failure modes.
As explained in Section 1.3.3, H2O and CO2 released from the concrete could oxidise the metals
contained in the corium. These exothermic chemical reactions have a twofold effect: increase the melt
temperature and generate flammable gases: H2 and CO, in addition to the H2 generated during the in-
vessel phase. The potential combustion of H2 and CO generated throughout a severe accident could
lead to a significant pressure peak in the containment, threatening its integrity. In addition to the
implementation of severe accident management guidelines (see Section 1.5) a number of safety
systems (e.g. passive autocatalytic recombiners (PAR’s), igniters, inerting the containment
atmosphere, etc.) have been designed and implemented in many LWRs in order to prevent this
containment failure mode (OECD/NEA, 1997), (IAEA, 2001).
The second possible containment failure mode is the so-called containment slow over
pressurisation. When the containment heat removal systems are inoperable, containment would fail by
slow pressurisation due to the build-up of the steam and non-condensable gases generated during the
ex-vessel phase of a severe accident. Under these circumstances, containment failure by slow over-
pressurisation is a late containment failure mode, because it takes place several hours after the RPV
failure.
However, containment venting could be activated in Mark I containments before the RPV failure
in some severe accident scenarios. This safety issue is analysed below. In order to prevent this
containment failure mode, filtered containment venting systems have been developed and
implemented in many NPPs (IAEA, 2001), (OECD/NEA, 1988).
(NRC, 2012) provides examples of severe accident scenarios where the containment venting
system could be activated even before the RPV failure. These scenarios are applicable for BWR with a
Mark I containment. Taking advantage of the SOARCA study (SNL, 2012), NRC has modelled with
MELCOR a number of both long-term and short-term station blackout (SBO) for Peach Bottom
leading to one of three possible outcomes: containment overpressure failure, liner melt-through failure,
or maintaining the containment intact as a result of venting or other mitigation measures. In modelling
a SBO, it is assumed that low-pressure core injection (LPCI), high pressure core injection (HPCI),
drywell spray, core spray and other engineered safety features (ESF), normally designed to run by AC
power, become unavailable for an extended period of time. However the reactor core isolation cooling
(RCIC) system is assumed to be running for 16 hours, because the RCIC is designed to provide core
cooling, the core uncovery and the subsequent accident progression is delayed accordingly. The RCIC
operation is controlled by battery, which acts as a power source for control valves that run the RCIC
pump on and off. The rationale followed to consider the RCIC to be running for 16 hours are provided
below.
The SOARCA study assumed RCIC operation for 4 hours. Many, if not most, USBWR Mark I
plants are equipped with batteries that will allow RCIC to run for an extended period of as much as
8 hours. Moreover, in the post-9/11 development of accident management strategies, conceivably even
a longer battery life for RCIC operation may have been considered. In Fukushima Daiichi Unit 2,
RCIC operation in excess of 70 hours has been reported although the reason for such an extended
operation is yet unknown. Likewise, in Fukushima Daiichi Unit 3, RCIC operation on the order of
20 hours has been reported followed by another 16 hours of HPCI operation that kept the core cooled.
With these considerations in mind, RCIC operation of 16 hours has been assumed in the MELCOR
calculations considered in this section.
NEA/CSNI/R(2016)15
The BWR Emergency Procedure Guidelines (BWREPGs), which form the basis for plant specific
emergency operating procedures, contain provisions for containment venting through the wetwell and
drywell (section 1.5 contains more detail information about EPGs and SAMGs). The primary function
of venting is to prevent containment failure by overpressure from steam and other non-condensable
gases. The BWR Mark I plants were originally designed with wetwell vents that had a low pressure
capacity. As a result of post-TMI improvements, the wetwell vents in many of these plants have been
upgraded and “hardened” for a high pressure capacity. Note that containment venting through the
wetwell has the advantage of attenuating fission products through suppression pool scrubbing.
Wetwell venting is activated when the containment pressure exceeds 60 psig (4.1 bar gauge) in the
cases analysed here.
(NRC, 2012) describes the results obtained by running 30 SBO scenarios with MELCOR for
Peach Bottom. Three cases were selected for this section. It has been respected the original
nomenclature for the cases contained in (NRC, 2012).
The three selected cases are briefly explained below:
• Case 2. SBO with total loss of safety systems, except for the RCIC. RCIC keeps running for
16 hours.
• Case 6. Same as case 2, but core spray is activated after the RPV failure. The rationale to
consider the activation of the core spray is provided next. Upon termination of RCIC
operation, a portable device, namely a diesel generator driven fire water system, was
considered to feed the low-pressure core spray system but only after RPV depressurisation.
A 300 gpm flow rate for the core spray was used in this case 6.
• Case 14. Same as case 2, but drywell spray is activated at 24 hours. Another mitigation
feature considered in the current study is drywell spray with a nominal flow rate of 300 gpm.
As in the case of core spray, the drywell spray is assumed to be operated by a diesel-powered
portable device. The drywell spray is actuated at 24 hours which, in most cases, correspond
to the timing of RPV lower head failure. A nominal flow rate of 300 gpm for drywell spray
was considered.
MELCOR results for these three cases are shown in Table 1.4-1. In addition, Figure 1.4-3 shows
the containment pressure evolution for the three cases.
Table 1.4-1: Timing of key events for MELCOR calculations

Event timing hr Case 2 Case 6 Case 14
SBO 0.0 0.0 0.0
RCIC terminates 17.9 17.9 17.9
Active fuel uncovery 22.9 22.9 22.9
Drywell pressure > 60 psig (4.1 bar gauge) – vent
22.8 23.3 23.2
opens if applicable
First Hydrogen production 23.6 23.6 23.2
Relocation of core debris to lower plenum 25.9 25.9 25.7
RPV lower head fails grossly 37.3 36.7 36.6
NEA/CSNI/R(2016)15
Figure 1.4-3: Comparison of drywell pressures for selected cases
1.4.7. Safety issues unaddressed by this report

MCCI SOAR focuses heavily on molten core concrete interactions and the cooling mechanisms for the
corium discharged from the RPV. As a result, a number of severe accident phenomena associated with
the ex-vessel phase of a severe accident are beyond the scope of the present SOAR. Below are listed
the safety issues unaddressed by this report.
• Corium jet impingement against concrete.
• Fuel coolant interaction after RPV failure.
• Steam spike.
• Melt jet break up.
• Debris bed coolability.
• Melt spreading.
• Fission products release.
• Recriticallity.
• Thermal stresses on concrete structures brought on by core debris interactions.
• Consequence of basemat penetration in terms of core debris migration into ground and
fission products release by leaching process with ground water.
Although this report focuses on corium-concrete interaction issues, one has to acknowledge that
valuable information can be found also in the works published by the nuclear or fire community to
address the concrete degradation due to surface heating by radiative heat flux (Chu T., 1978) or all the
work performed on sodium-concrete interactions.
NEA/CSNI/R(2016)15
1.5. Severe accident management issues and back-fitting measures
Accident management is essential to ensure effective defence in depth. (IAEA, 2009) provides the
following definition of accident management:
“Accident management is the taking of a set of actions during the evolution of a beyond design
basis accident:
• To prevent the escalation of the event into a severe accident;
• To mitigate the consequences of a severe accident;
• To achieve a long term safe stable state.
The second aspect of accident management (to mitigate the consequences of a severe accident) is
also termed severe accident management. Accident management is essential to ensure effective
defence in depth at the fourth level. According to the (IAEA, 2009), “the objective of the fourth level
of defence in depth is to ensure that both the likelihood of an accident entailing significant core
damage (a severe accident) and the magnitude of a release of radioactive material following a severe
accident are kept as low as reasonably achievable and, thereby, to reduce risk”.
(OECD/NEA, 1995) provides the following definition of severe accident management:
“Severe Accident Management consists of those actions that are taken by the plant staff during
the course of an accident to:
• Prevent core damage.
• Terminate progress of core damage and retain the core within the vessel.
• Maintain containment integrity, and
• Minimise off-site releases.
Severe accident management also involves pre-planning and preparatory measures for:
• Severe accident management guidance and procedures.
• Equipment modifications to facilitate procedure implementation, and
• Severe accident training.
The overall objective is to further reduce the risks of large releases. It is the responsibility of the
licensees to develop and implement a severe accident management.”
This definition includes the concept that there is some overlap between what is referred to as
accident management and severe accident management.
Accident management guidance has different forms depending on the considered domain. The
guidance for the preventive domain, therefore, takes the form of procedures, usually called emergency
operating procedures (EOPs), and is prescriptive in nature. EOPs cover both design basis accidents
and beyond design basis (IAEA, 2009).
In the mitigatory domain, uncertainties may exist both in the plant status and in the outcome of
actions. Consequently, the guidance for the mitigatory domain should not be prescriptive in nature but
rather should propose a range of possible mitigatory actions and should allow for additional evaluation
and alternative actions. Such guidance is usually termed Severe Accident Management Guidelines
(SAMGs) (IAEA, 2009).
NEA/CSNI/R(2016)15
The guidance should contain a description of both the positive and negative potential
consequences of proposed actions. The guidance for the mitigatory domain should be presented in the
form of guidelines, manuals or handbooks. The term ‘guideline’ here is used to describe a fairly
detailed set of instructions that describe the tasks to be executed on the plant, but which are still less
strict and prescriptive than the procedures found in the EOPs; that is, used in the preventive domain.
Manuals or handbooks will contain a more general description of the tasks to be executed and their
background reasoning.
Over the past fifteen years many of the nuclear power plants worldwide have been equipped with
a capability for severe accident management. This has been driven partly by the SAMGs developed by
owners groups for plant specific applications.
The impact of SAMGs has also been included in a number of Level 2 PSA studies which resulted
in a revision of core damage frequency, large early release frequency, and source term estimates. A
review was performed to identify the new developments and insights derived from Level 2 PSA
activities worldwide in the past ten years. Much of this new information was presented at the two
CSNI Workshops – the first on Level 2 PSA and Severe Accident Management held in Köln in March
2004, and the second on the Evaluation of Uncertainties in Relation to Severe Accidents and
Level 2 PSA held in Aix-en-Provence in November 2005.
After this period, it was clearly recognise that Level 2 PSA approach is now a common practice
for most reactors as explained in a 2007 CSNI-WG-Risk Technical Opinion Paper (TOP) on L2PSA
established after these two Workshops.
“The main message of this TOP is that the Level 2 PSA methodology may now be seen as
mature. This is reflected by the large number of high quality analyses that have been performed in
recent years and used to identify the potential vulnerabilities to severe accidents and the accident
management measures that could be implemented.
The Level 2 PSA is now seen as an essential part of the safety analysis that is carried out for all
types of nuclear power plants worldwide. The information provided by the Level 2 PSA is being used
by plant operators and Regulatory Authorities as part of a risk informed decision making process on
plant operation and more specifically on issues related to severe accident management.”
Nevertheless, it was also recognised that some discrepancies may exist in L2 PSA practical
methodologies, for example to introduce uncertainties associated to the severe accident phenomena.
After having promoting exchanges in the framework of the EC-SARNET, the European Commission
has sponsored the ASAMPSA2 project (ASAMPSA2, 2013) which has established some detailed best-
practices guidelines for L2 PSA development and applications, taken into consideration recent results
from research area. These guidelines summarise available knowledge on MCCI issues in 2011. They
will be update by the ongoing EC-ASAMPSA_E project (www.asampsa.eu).
Finally, the European Commission (EC) sponsored two studies: SAMIME (Severe Accident
Management Implementation and Expertise in the European Union) and SAMOS (the viability of
using computerised aids to assist in severe accident management) to understand the severe accident
management practice in member states and to investigate the feasibility of computerised tools for
accident and emergency management.
A brief description of the severe accident management strategies aimed to maintain containment
integrity for present LWR after the corium is poured onto the reactor cavity floor are described below.
1.5.1. Dry cavity and no water addition
As it can be seen in Figure 1.2-1, when core concrete interaction proceeds under dry conditions,
late containment failure is certain for most of the cases.
NEA/CSNI/R(2016)15
1.5.2. Reactor cavity flooded before RPV failure at low pressure

When the cavity is flooded before the vessel breach, there exist chances of formation of coolable
debris, especially with deep water pools. Available space for sufficiently deep pools depends on plant
design and is only available in some present LWR. With limited space in other types of reactors only
partial cooling and melt retention after a RPV failure can be expected from cavity flooding. As
possible counter-effects, the occurrence of steam explosions or strong steam spikes is considered in
Figure 1.2-1. Note that water has to be continuously added in order to achieve long term stabilisation.
Otherwise, only some delay for the basemat melt-through could be gained.
Steam explosions are beyond the scope of this report (see Section 1.4.7). The term steam spike
refers to pressure generated due to rapid quenching of the corium melt. For this process to occur the
melt must come into intimate contact with substantial quantities of water (NRC, 1985). The steam
spike phenomenon occurs when the molten debris is ejected from the reactor vessel and it encounters
water in the cavity or in the containment floor after being swept out of the cavity under high pressure.
The immediate corium/water interaction would generate a large amount of steam and produce a
pressure spike in the containment.
(Yang, 1990) provides very detailed information about the main results obtained by the NRC
during its investigations on the steam spike issue for American large dry PWR containments. Based on
these analyses, (Yang, 1990) concludes that the failure of the PWR large dry containment by a steam
pressure spike at the time of vessel failure is an event of relatively low probability. Note, that this
conclusion is applicable only for those PWR within the scope of the aforementioned study.
1.5.3. Water injection into the reactor cavity after RPV failure at low pressure
According to Figure 1.2-1, the branch without water in the cavity at the time of melt release is
considered to yield a manageable situation (i.e. termination of basemat erosion before melt-through or
over-pressurisation of the containment), only in case that by flooding of the cavity with water finally a
coolable configuration is established. Section 1.3.4 describes the various corium cooling mechanisms,
when water is poured onto the corium. Section 1.3.4 also states that their efficiency in terms of
thermally stabilising the molten core debris is principally a function of melt depth and the timing of
water addition into the cavity. Analytical activities supporting this statement are shown below.
The computer code CORQUENCH was developed within the frame of the MACE, NEA-MCCI
and MCCI-2 experimental projects. CORQUENCH code contains phenomenological models for the
ex-vessel corium cooling mechanisms, along with other models such as material properties and
concrete ablation models. Chapters 2 and 3 provide, respectively, detailed information about the
experimental programmes and the models contained in CORQUENCH.
(Robb & Corradini, 2009) show the results obtained in a sensitivity study performed with
CORQUENCH, whose main objective was to investigate the importance of system initial and
boundary conditions and also cooling phenomena to the progression of MCCI.
This sensitivity study was carried out over a range of input conditions for two concrete
compositions: LCS and siliceous. A 2D right cylindrical geometry was used to simulate reactor cavity
scale accidents and allowed for 2D concrete ablation. Note that the melt pool can only be modelled as
homogeneous in CORQUENCH. The input parameters and the range over which they were varied are
noted in Table 1.5-1.
NEA/CSNI/R(2016)15
Table 1.5-1: Input parameters varied in the 2D initial and boundary condition sensitivity studies
Parameter 2D Analysis Range Units
Cavity Pressure 0.1-0.5 MPa
Water Add Time Delay 5-120 min.
Time Since Reactor Shutdown 1-8 hrs.
Melt Initial Temp 2400-2900 K
Melt Initial Mass 45-180 tonne
Zr Initial Oxidation 20-80 w%
Cavity Diameter 5-10 m
This sensitivity study shows that with respect to initial and boundary conditions, the initial melt
mass and the cavity diameter are consistently the most important independent variables with respect to
ablation. Further investigation into sensitivity studies show that the initial collapsed melt height,
dependant from the above mentioned independent variables, is the most important initial condition.
The time delay of water addition was second in importance with the other independent variables
having tertiary importance.
As for the 2D modelling phenomena sensitivity studies, Table 1.5-2 shows the input parameters
considered. The range of values is based on post-test experimental data from the MACE test program.
Constant values for the melt entrainment coefficient, crust permeability, and crust thermal
conductivity were used instead of the phenomenological models available in CORQUENCH.
The main outcomes of this sensitivity study are shown below.
As for LCS concrete, the melt entrainment coefficient has a dominant effect on the total radial
and axial ablation, H2 gas generation, as well as the ratio of axial to radial ablation. The crust
permeability was second in importance followed by the other parameters.
Table 1.5-2: Input parameters varied in the 2D modelling phenomena sensitivity studies
Parameter Range Units
Particle Bed Porosity 33-59 %

Particle Bed Diameter 1.4-4.2 mm
Parameter Range Units
Melt Entrainment Coeff. 0.00-0.0025 -
Crust Permeability -11 m2

2.5-50 x10
Crust Thermal Conductivity 1.25-1.75 W/m K
Crust Strength 0.2-10 MPa
As for siliceous concrete, the melt entrainment coefficient was also the most influential parameter
to melt coolability, whereas the crust strength was now second in importance.
NEA/CSNI/R(2016)15
Note that the above-described analytical results are dependent upon the models used in
CORQUENCH and may only apply within the ranges of the parameters investigated. As more
accurate phenomenological models are crafted and implemented, new sensitivity analysis should be
accomplished for.
1.5.4. Back-fitting measures in Gen. II reactors aimed to prevent the basemat melt-through
Recently, modifications have been implemented in the French Fessenheim Gen II reactor in order to
prevent early basemat melt-through. Chapters 5 and 6 provide more detailed information about these
back-fitting measures that are briefly enumerated below:
• To increase the corium spreading area. In order to achieve it, dedicated holes in the vertical
walls of the reactor cavity could be drilled allowing for corium spreading into rooms
adjacent to the reactor cavity.
• To increase the thickness of the concrete basemat slab.
1.5.5. Generation III reactors
New reactor designs are expected to have enhanced capabilities both for preventing and for mitigating
severe accidents. However, severe accident management strategies remain an important element of
defence-in-depth for these new designs.
The new designs include safety enhancements not present in existing plants aimed at achieving
melt stabilisation. It basically could be distinguished between two classes of concepts:
• Concepts based on enclosure of melt and external cooling: enclosure concepts.
• Concepts relying on formation of particulate or porous debris as a basis for rapid quenching
and cooling due to surface increase.
Figure 1.5-1 gives an overview of the different concepts and also indicates advantages and
disadvantages. Here, measures are indicated from the side of constellations resulting from measures
and scenarios (left side) as well as from major retention and cooling concept (right side) and cooling
concepts (right side) and their requirements.
There exist certainly possibilities of combinations of basic goals and respective measures, partly
already depicted in Figure 1.5-1 i.e. some surface increase of the melt is provided by measures of
spreading of melt on a dedicated large area, as in the EPR concept (Fischer, 1999), (Seiler, et al.,
2003). However, since in this concept the melt layer may remain too thick (≈ 40 cm.) for removal of
decay heat in solid state, molten parts can exist over weeks or months, until the decay heat is
sufficiently reduced. This is because, at high decay power and without crediting melt fragmentation
and water ingress, heat can only be effectively transported to the cooled outer surfaces when the melt
remains in a partly liquid state. This means that the cooling from bottom and top which ensures the
enclosure of the melt inside crusts must be maintained over several months.
NEA/CSNI/R(2016)15
Figure 1.5-1: Different basic measures for retention and coolability

(Allelein & Burger, 2006) 4
The Tian-Wan core catcher concept (Seiler, et al., 2003), (Sidorov, 2000) represents another
enclosure concept, with a hot melt pool enclosed by crusts within a large steel vessel beneath the RPV.
Retention in liquid state has to be guaranteed over an even longer time by external water cooling of the
catcher vessel. Heat removal is enabled by diluting the melt with sacrificial material in the catcher
vessel, thus reducing the power density and increasing the surface for heat removal. As for ESBWR,
the BiMAC device is intended to arrest core melt progression in the lower drywell by cooling the
debris both from above and below (Gavrilas & Fuller, 2010).
The major concepts based on fragmentation and porosity formation are the deep water pool
(Seiler, et al., 2003), (Chu, Sienicki, Spencer, Frid, & Löwenhielm, 1995), (Lindholm, Holmstroem,
Miettinen, Lestinen, & Hyvaerinen, 2006) in the cavity yielding melt break-up and particulate debris
formation as well as the COMET concept (Alsmeyer & Tromm, 1999), (Alsmeyer, et al., 2004)
providing a coolable porous layer by means of water injection from the bottom into a spread layer.
Rapid quenching occurs in both concepts due to a large melt-water surface increase. It is further
supported by co-current flow modes of water and steam, especially in the COMET concept, provided
by the bottom injection of water.
A combination with melt spreading as in the EPR concept is to some extent essential also for the
COMET concept. The presently favoured design is optimised with respect to a relatively small layer
thickness of about 40-50 cm (as in the EPR concept). Easier implementation of the device is envisaged
with the COMET PC(A) concept (Alsmeyer & Tromm, 1999), where the water is injected via porous
concrete instead of specific injection nozzles. The water filled porous concrete may also be considered
to yield an additional protection as an enclosure.
Partly, the functioning of such concepts may be affected by concrete erosion, e.g. concerning
conditioning of the melt by mixing with sacrificial material or metallic/ceramic layering during a
4. On the right side of the figure (FP retention and coolability), melt surface has to be understood as melt-water
exchange surface.
NEA/CSNI/R(2016)15
phase before termination of erosion by stable enclosures or water injection and quenching. Further a
failure of a retention concept might occur. The analyses of possible weaknesses of such concepts and
of continued erosion processes after a failure still requires reliable knowledge about concrete erosion,
i.e. not only for existing types of reactors but also for considerations on new concepts.
The core melt stabilisation system (CMSS) and the Severe Accident Heat Removal System
(SAHRS) in EPR are examples of features designed to ensure core debris coolability. Numerous
features are incorporated into the ABWR design to help mitigate the effects of core concrete
interaction. These are: a large lower drywell flooder (LDF) system, an AC-independent water addition
(ACIWA) system, use of sacrificial basaltic concrete for the lower drywell floor, and a Containment
Overpressure Protection System (COPS). Ex-vessel melt stabilisation is included for other new reactor
designs as well including VVER-1000, APWR, and CANDU. In some cases, different strategies from
those used for existing reactors must be adopted.
All the above mentioned new safety features and accident management strategies have used
insights derived from two decades of molten core-concrete interaction and ex-vessel debris coolability
research findings.
1.6. Prioritisation of severe accident research issues
Once obtained a global view of the severe accident phenomena addressed by this SOAR, basically,
corium-concrete interactions and ex-vessel corium coolability mechanisms by top flooding, it is time
to highlight the safety significance given to these severe accident issues by the scientific and technical
community.
(Magallon, et al., 2005), (Schwinges, y otros, 2008) and (Klein-Heßling, et al., 2012) contain
detailed information on the activities carried out by the European Community and SARNET to
prioritise all the severe accident issues. This activity was deemed necessary to guarantee that the
research conducted on severe accidents remains efficient and conveniently centred. Before describing
these activities in detail, the main conclusions obtained in these decision processes related with the
matters addressed by this SOAR will be explained.
Three safety prioritisations for severe accident issues have been conducted by the Severe
Accident Research Priority (SARP) group and all of them concluded that the two following severe
accident issues should have the top safety priority, e.g. high.
• MCCI: molten pool configuration and concrete ablation.
• Ex-Vessel corium coolability, top flooding.
More detailed explanations about this decision process are provided below.
On 2001, the European Union considered that a work methodology aimed at ensuring that the
research conducted on severe accidents is efficient and well-focused should be developed and
implemented. As a result, it was created the EURSAFE Project within the FP5 (Magallon, et al.,
2005). EURSAFE accomplished the first PIRT (Phenomena Identification and Ranking Table) for
severe accidents. It integrated all the severe accident issues from core degradation up to release of
fission products from the containment, taking into account any possible countermeasures and the
evolution of fuel management. Two evaluations were established:
1. The safety importance ratio.
2. The knowledge ratio.
Starting with 1 016 identified phenomena, the list was reduced to 239 items important for safety,
of which 106 were found with significant lack of knowledge. These items have been summarised to
21 items, the so-called EURSAFE Research Issues (ERI). Two of these ERI were:
NEA/CSNI/R(2016)15
1. MCCI: molten pool configuration and concrete ablation

The research within this ERI would be aimed at improving the predictability of axial vs. radial
ablation up to late phase MCCI to determine basement material failure time and loss of
containment integrity.
2. Ex-vessel corium coolability, top flooding
The research within this ERI would be aimed at increasing the knowledge of cooling
mechanisms by top flooding the corium pool to demonstrate termination of accident
progression and maintenance of containment integrity
The SARNET Severe Accident Research Priority (SARP) group, within SARNET FP6,
accomplished a thorough review of this first PIRT four years later (Schwinges, et al., 2008). The main
conclusions reached by the SARP group were:
• 6 out of these 21 ERI were still considered with high priority.
• 4 ERI were re-assessed with medium priority
• 5 ERI were re-assessed with low priority.
• 3 ERI were marked as “issue could be closed”.
The two ERI concerned with this report, i.e. MCCI: molten pool configuration and concrete
ablation and ex-vessel corium coolability, top flooding, held their previous safety prioritisation: high.
Four years later, a new reassessment of the severe accident issues prioritisation was conducted by
the SARNET Severe Accident Research Priority (SARP) group within SARNET FP7 (2009-2013).
(Klein-Heßling, et al., 2012) describe this activity in detail, whose more relevant conclusions are
explained below.
In order to correctly address this problem, it was deemed that the organisations involved in the
working group should be selected from researchers, members of Technical Safety Organisations
(TSO) and utilities. Thus, the SARP work package was comprised by the following 12 organisations:
IRSN, Areva, CEA, EDF, KIT, GRS, JSI, KTH, PSI, RUB, TUS and VTT. GRS was the leading
organisation.
Relevant sources of information for this decision process were:
• The results of the ongoing SARNET work-packages and the related research projects and
experimental programmes.
• The results of the finished ASAMPSA2 project.
• The results of NEA projects (such as MCCI, THAI and BIP).
Although the procedure for the decision making was similar to that used in the previous SARP
works, some minor differences were introduced. The vote of phenomena was considered first before
discussing the safety relevance. The scope of the “safety-oriented vote” was widened. In addition to
the radiological consequences of a severe accident, the consequences on accident management
measures and procedures were also taken into account. Moreover, this reassessment also considered a
possible change of regulators’ point of view, e.g. the higher priority of long-term consequences of a
source term.
All steps of the decision process were documented to allow a well-founded judgement for the
end-users of this activity.
NEA/CSNI/R(2016)15
As for the review of the EURSAFE Research Issues, the following conclusions were obtained:
• The number of ERI increased to 25 up from the 21 considered in the previous work. A new
ERI was added and three previous ERI were split.
• This new added ERI is also related with this report: Thermal Database and deals with the
improvement of the thermo-dynamic and thermo-physical database for corium and fission
products. The priority assigned to this ERI was “medium”.
• The ERIs “MCCI: molten pool configuration and concrete ablation” and “ex-vessel corium
coolability, top flooding” retain their previous safety classification: high.
• The available data related to Fukushima was quite limited during the elaboration of this
reassessment. Thus, only four issues derived from Fukushima were proposed for further
consideration. These four issues are:
• Pool scrubbing under boiling conditions.
• Effect of impurities in water.
• Spent fuel scenarios.
• MCCI aerosol effect on chemistry.
NEA/CSNI/R(2016)15
2. Molten core-concrete interaction and debris coolability
2.1. Overview
The overall objective of this section is to review and assess principal findings from experimental
programmes as well as a reactor accident that have provided information on molten corium concrete
interaction under both wet and dry cavity conditions. The general goals of the experiments have been
to: i) identify and characterise important phenomenological processes in order to facilitate model
development, and ii) provide experimental data to support validation of models and codes that are used
in reactor safety assessments. For dry cavity conditions, the research has focused on evaluating the
nature and extent of the MCCI and concurrent fission product release. Conversely, under wet cavity
conditions the studies have principally focused on evaluating the effectiveness of coolant in
terminating the MCCI by quenching the molten core material and rendering it permanently coolable.
For dry cavity conditions, decay heat is continually dissipated by erosion of underlying concrete
and can eventually lead to containment failure by axial erosion through the extent of the concrete
basemat. Conversely, for BWR containments, radial erosion can undermine essential support
structures such as the reactor support pedestal. Aside from exacerbating fission product release,
continued ablation can lead to containment pressurisation by production of non-condensable gases
arising from concrete decomposition. In addition, for scenarios in which the core debris contains
significant amounts of unoxidised cladding and/or structural steel, then generation of flammable gases
(H2 and CO) from the interaction of the decomposition gases (H2O and CO2) with metals present in
the melt can also present a containment challenge.
Cavity flooding offers the opportunity to quench the core debris and prevent basemat melt
through, greatly attenuate fission product release from core debris, and terminate containment
pressurisation by non-condensable gas production. If core material that has relocated ex-vessel can be
quenched and rendered permanently coolable by formation of sufficient porosity within the debris for
water to ingress, then one significant aspect of the accident progression could be successfully halted.
However, steam will continue to be generated as decay heat is removed from the debris, and so
complete termination of the ex-vessel accident progression will inherently hinge upon: i) maintaining
adequate containment cooling, and ii) ensuring that sufficient heat can be extracted from the lower part
of the melt to prevent further basemat penetration. Furthermore, the distribution of debris found within
the TMI-2 vessel indicates that in case of vessel failure fuel will be not always dispersed into the
cavity at the same time during an accident. The quenching of relocated fuel released from a failed
vessel does not preclude continued degradation and fission product release from fuel remaining in the
core. These practical matters need to be factored into any evaluation of ex-vessel accident progression
involving cavity flooding.
The various accident sequences and the possibility of operator intervention result in a broad range
of possible initial conditions at time of vessel failure. Additionally, the state of knowledge about late
in-vessel melt progression is incomplete (particularly for BWRs). Thus, there is considerable
uncertainty regarding the MCCI initial conditions that include the timing of RPV failure; the initial
temperature, mass, and composition of the core debris; the possibility of segregation of metal and
oxide melt phases; the pour rate of the melt from the RPV that is determined principally by the melting
rate of residual core material, and to a lesser extent by the opening in the RPV lower head; and finally
the timing of water injection (if any). Many of these parameters (e.g. power level in the core debris,
which is indicative of the time of vessel failure, as well as melt mass and composition) have been
addressed in various experimental programmes that are described below.
NEA/CSNI/R(2016)15
2.2. Database summary
This section provides a summary of the database that has been generated over the last 40 years in
the areas of MCCI and debris coolability. Additional technical details regarding various experimental
programmes as well as the Chernobyl 4 reactor accident are provided in subsequent sections. Key data
available from integral experiments and Chernobyl for code assessment under dry cavity conditions
are summarised in Table 2.2-1, while flooded cavity data are summarised in Table 2.2-2. A summary
of simulant material and separate effect experiments that focused on providing phenomenological
insights and model development data is provided in Table 2.2-3.
Table 2.2-1: Dry cavity MCCI and aerosol generation integral experiments and accidents
Program and test(s) Description Phenomena tested or characterised
SNL
BURN-1, Large Scale Early transient tests with molten stainless steel Identification of basic phenomenology associated
Trans. Tests, TURC-1T, in concrete crucibles with core-concrete interactions for different
1SS concrete types
BETA
Series I: V1.1-V1.9; V2.1- Iron and Al2O3 tests in 2D cavities with Influence of power level on cavity erosion
V2.3; V3.1-V3.3 induction heating to simulate decay heat behaviour and aerosol release with SIL and LCS
concretes
Series II: V5.1-V5.3; Steel, Zr, and concrete oxide tests in 2D Influence of Zr cladding and B4C on erosion
V6.1-V6.2 cavities with induction heating to simulate behaviour and aerosol release with SIL and
decay heat serpentine concretes
SURC
QT-D,E; SURC-3, 3A, 4 Stainless steel tests in 1D cavities with Influence of Zr cladding on cavity erosion
induction heating to simulate decay heat behaviour and aerosol release for SIL and LCS
concretes
FRAG
1, 2a, 3, 4 Sustained heating of steel spheres over Ablation rate of fragmented core debris interacting
concrete surfaces with different types of concrete
HSS
1,2 Sustained heating of solid slugs of steel and Ablation rate of solid debris interacting with
UO2-ZrO2 over concrete surfaces concrete
COMET
L2 Iron and Al2O3 in 2D concrete cavities with Cavity erosion and gas release for SIL concrete
induction heating to simulate decay heat
MOCKA
1.1-1.7 Iron plus Al2O3, CaO in 2D concrete cavities; Cavity erosion and gas release for SIL concrete
chemical heating tests to simulate decay heat
HECLA
1-5 Transient stainless steel tests in 2D cavities Cavity erosion for FeSi (hematite) and ordinary SIL
investigating initial transient cavity erosion concrete.
phase.
ACE/MCCI
L1-L2; L4-L8 Core oxide tests in 1D cavities with direct Cavity erosion and fission product release for PWR
electrical heating (DEH) to simulate decay and BWR melts with SIL, LCS, LL, and serpentine
heat concretes
TURC
2-3 Transient core oxide tests in 1D crucibles Initial transient erosion phase and aerosol release
SURC
1-2 Core oxide tests in 1D cavities with induction Effect of Zr cladding on cavity erosion and aerosol
heating of susceptor plates to simulate decay release from PWR melt composition interacting
heat with LCS and Basalt concretes
SICOPS
Core oxide tests in 2D cavities with induction Simultaneous interaction of corium with concrete
heating to simulate decay heat and refractory material
VULCANO
VB- U4-U7, VB-ES U1-U4, Core oxide-stainless steel tests in 2D cavities; Cavity erosion for siliceous, LCS, and hematite
VBS U1-U4 induction heating to simulate decay heat concretes; separate effect tests with special
NEA/CSNI/R(2016)15
Program and test(s) Description Phenomena tested or characterised

concretes; focus on high metal content in corium
NEA-MCCI
CCI 1-5 Large scale, core oxide, prototypic power Cavity erosion during initial test phase for LCS
density tests in 2D cavities; DEH to simulate and SIL concretes
decay heat
Chernobyl
Unit 4 RBMK-1000 reactor accident involving core Erosion of basaltic concrete within the reactor
melt formation within the reactor cavity and cavity over the course of days
subsequent core-concrete interaction
Table 2.2-2: Flooded cavity debris coolability integral experiments

Program and test(s) Description Phenomena tested
SWISS
1-2 Stainless steel tests in 1D concrete cavities with Cavity erosion, debris cooling rate, and aerosol
induction heating to simulate decay heat scrubbing with melt interacting with LCS concrete
WETCOR
1 Oxide simulant in 1D concrete cavity with induction Cavity erosion, debris cooling rate, and aerosol
heating to simulate decay heat scrubbing with melt interacting with LCS concrete
COMET
L3 Iron and Al2O3 in 2D concrete cavities with Cavity erosion and debris cooling rate for siliceous
induction heating to simulate decay heat concrete
ECOKATS
2 Large scale transient spreading test with oxide Cavity erosion and debris cooling rate for siliceous
simulant; flooded following spreading phase concrete
COTELS
B/C-2 to B/C-9 High power density core oxide tests in 2D cavities Cavity erosion and the extent of debris quenching by
with induction heating to simulate decay heat the mechanisms of particle bed formation and water
ingression into fragmented debris
MACE and NEA-MCCI
MACE M0, M1b, Large scale, core oxide, prototypic power density Cavity erosion and debris cooling rate for LCS and
M3b, M4; MCCI tests in 1D and 2D cavities; DEH to simulate decay siliceous concretes; quantification of extent of cooling
CCI 1-4, 6; heat by melt eruption and water ingression mechanisms
Table 2.2-3: Simulant/separate effect experiments on MCCI behaviour,

properties, and debris coolability
Program(s) Description Phenomena tested
ARTEMIS 1D and 2D dry experiments using LiCl-BaCl2 eutectic to Cavity erosion and crusting behaviour at melt-
simulate concrete and pure BaCl2 (as well as concrete concrete interface
slag for two tests) to simulate melt. Emersion heaters to
simulate decay heat; gas injection to simulate concrete
decomposition gases
ÉCLAIR, CLARA, Controlled sparging of heated homogeneous pools with Effect of sparging on heat flux distribution to
BALI varying viscosity bottom and side surfaces of pool
EPSTEIN, CASAS Controlled sparging of pools with two immiscible fluids Onset of mixing between stratified layers
and CORRADINI,
BALISE
WERLE, GREENE, Controlled sparging of heated pools with two immiscible Heat transfer coefficient between two stratified
ABI fluids layers
PERCOLA and Controlled sparging of pools (varying viscosity for Melt eruption cooling mechanism (qualitative
U. WISCONSIN PERCOLA, 1-300 mPa.sec; water for UW tests). Upper and quantitative information on melt
surface covered with a special plate with penetrations entrainment)
that mocks up a porous floating crust.
UCSB Glycerine – liquid N2 tests with controlled gas injection to Debris cooling characteristics and morphology
simulate concrete decomposition gases (15 cm to 120 as a function of test section size
cm test section spans)
NEA-MCCI SSWICS Transient reactor material corium quenching tests in 1D Water ingression cooling and the mechanical
inert crucibles strength of quenched corium ingots
ROCHE Differential thermal analysis (DTA) and viscometer Liquidus-solidus and viscosity of core oxide-
experiments with core-concrete mixtures containing concrete mixtures
NEA/CSNI/R(2016)15
Program(s) Description Phenomena tested

varying amounts and types of concrete.
PRECOS Melting point evaluations for mixtures of UO2 and Phase diagrams for UO2–SiO2, UO2–CaO,
concrete constituents (SiO2, CaO) UO2–FeO–SiO2, and UO2–FeO–CaO mixtures
COLIMA CA-U4 Aerosol generation over MCCI pools and retention in Fission product aerosol generation and
concrete cracks deposition
Early transient experiments with steel melts were initiated in the mid-1970s at Sandia National
Laboratories (SNL) ( (Powers, Dahlgren, Muir, & Murfin, 1978), (Powers & Arellano, 1982-1) and in
Germany (Alsmeyer, et al., 1977) to identify basic phenomenology associated with core-concrete
interactions. The SNL transient testing programme continued into the 1980s using better-instrumented
systems ( (Powers & Arellano, 1982-2), (Gronager, Suo-Antilla, Bradley, & Brockmann, 1986) as
well as the incorporation of fission product mockups in the melt to assess fission product release
during the early stages of core-concrete interaction. The experiment capability was expanded again as
part of the Sustained Urania Concrete (SURC) programme to incorporate sustained heating of steel
melts under dry cavity conditions to further characterise phenomenology associated with long-term
core-concrete interaction. In particular, the SURC-3 (Copus, Blose, Brockmann, Gomez, & Lucero,
1990) and SURC-4 (Copus, Blose, Brockmann, Gopmex, & Lucero, Core-Concrete Interactions using
Molten Steel with Zirconium on a basaltic Basemat: The SURC-4 Experiment, 1989) tests evaluated
the effect of unoxidised cladding on the progression of core-concrete interaction for several concrete
types.
The sustained heating programme at SNL also included experiments carried out with solid
material to examine situations in which core debris may be quenched during relocation through water
following vessel failure, and then reheat to begin core-concrete interaction. These tests were carried
out as part of the HSS (Copus & Bradley, 1986) and FRAG (Tarbell, Bradley, Blose, Ross, & Gilbert,
1987) programs that examined the penetration of steel shot and slugs into different types of concrete
under sustained heating conditions.
In parallel with these efforts, an extensive experimental programme involving sustained heating
of metallic melts was initiated at Karlsrühe Institute of Technology (KIT) in the BETA ( (Alsmeyer,
1987) (Alsmeyer, et al., 1995) and COMET (Miassoedov A. , et al., 2008) test facilities in Germany.
These experiments provided valuable data on core-concrete interaction at a variety of power levels
including gas and aerosol releases during core-concrete interaction. Recently, as part of the HECLA
Program (Sevón, et al., 2010), transient metal tests have been conducted at VTT in Finland that
focused on quantifying the ablation characteristics for hematite concrete that is used as the sacrificial
material in the reactor pit of the European Pressurised Reactor (EPR). In addition, the MOCKA test
series (Foit, Cron, Fluhrer, Miassoedov, & Wenz, 2012) was carried out to study the interaction of a
simulant oxide and metal melt in a stratified configuration. To allow for a longer-term interaction
without the use of an external power supply, additional energy was added to the system by alternating
additions of thermite and Zr metal. These tests provide data on the heat flux distributions in the lateral
and axial directions during core-concrete interaction.
In terms of reactor material testing, a series of 1D experiments (ACE/MCCI) addressing thermal-
hydraulic behaviour and fission product release were carried out at Argonne National Laboratory
(ANL) ( (Thompson D. H., Fink, Armstrong, Spencer, & Sehgal, 1992), (Fink, Thompson, Armstrong,
Spencer, & Sehgal, 1995). Large scale 1D core melt tests (SURC) were also completed at SNL under
transient (Gronager, Suo-Anttila, & Brockmann, 1986) as well as sustained heating ( (Copus, Blose,
Brockmann, Simpson, & Lucero, 1992-1), (Copus, Blose, Brockmann, Simpson, & Lucero, 1992-2)
conditions that looked at erosion rates as well as aerosol generation. Two-dimensional core-concrete
interaction experiments ( (Journeau C. , et al., 2009), (Journeau, et al., 2012) under dry cavity
conditions have also been performed at the VULCANO test facility at Commissariat à l'énergie
atomique et aux énergies alternatives (CEA) in France. These tests examined multi-dimensional cavity
erosion behaviour, including the important parametric effect of high metal fraction in the core melt.
NEA/CSNI/R(2016)15
Core oxide and oxide-metal material interaction experiments in 1D and 2D cavities made of concrete
as well as refractory zirconia have been carried out in the Areva SICOPS facility (Langrock &
Hellmann, 2010). Finally, a series of large scale 2D core-concrete interaction experiments were
conducted as part of the internationally-sponsored NEA-MCCI Program at ANL ( (Farmer, et al., A
Summary of Findings from the Melt Coolability and Concrete Interaction (MCCI) Program, 2007),
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, &
Aeschlimann, 2010). These tests examined dry cavity erosion characteristics for several hours before
the cavity was flooded to provide additional data on debris coolability.
Finally, during the April 1986 severe accident at Chernobyl unit 4, tens of tonnes of molten
corium interacted with the reactor room concrete. Thus, this accident provided unique data on the
nature and extent of core melt interaction with silica-rich (i.e. basaltic) concrete over several days
under dry cavity conditions ((Bogatov, Borovoi, et al., 2007), (Pazukhin 1994).
The dry cavity test results show that core-concrete interaction during the early phase is influenced
by the extent of unoxidised cladding that is initially present in the melt. However, the remaining
cladding is rapidly oxidised within approximately the first 30 minutes of the interaction. During the
long-term, the nature of the core-concrete interaction is found to be a strong function of concrete type.
In particular, for silica-rich siliceous concretes the radial/axial ablation is found to be strongly skewed
radially, at least in the initial portion of the interaction when crusting on the concrete surfaces can
influence the power distribution. Conversely, for limestone-rich concretes, the radial/axial ablation
rates appear to be similar. For melts containing a significant fraction of unoxidised steel, stratification
of the steel and oxide phases is assumed after the initial transient phase has passed. Under these
conditions, the axial/radial power split appears to increase relative to tests conducted with little metal
present, but the data are too sparse to draw quantitative conclusions at this point. Refractory zirconia
has also been found to be effectively inert when in contact with core melt and concrete.
Fission product release has also been investigated. A range of parameters was addressed in one
study (Fink, Thompson, Armstrong, Spencer, & Sehgal, 1995) by a series of tests that used four types
of concrete (siliceous, limestone/sand, serpentine, and limestone) and a range of cladding oxidations
for both BWR and PWR core debris. The released aerosols contained mainly constituents of the
concrete. In the tests with metal and limestone/sand or siliceous concrete, silicon compounds
comprised 50% or more of the aerosol mass. Releases of uranium and low-volatility fission-product
elements were small. Releases of tellurium and neutron absorber materials (i.e. silver, indium, and
boron from boron carbide) were high.
For flooded cavity conditions, research has principally focused on determining the effectiveness
of water in terminating an MCCI by flooding the interaction from above, thereby quenching the
molten core debris and rendering it permanently coolable (Table 2.2-2). As a part of this work both
simulant and reactor material integral experiments have been completed to provide a database to
support model development and code validation activities. High temperature steel and oxide simulant
experiments were conducted at SNL as part of the SWISS (Blose, Gronager, Suo-Anttila, &
Brockman, 1987) and WETCOR (Blose, et al., 1993) programs to investigate coolability with
concurrent concrete erosion. Aerosol scrubbing due to overlying water was also measured as part of
these tests. Large scale steel and oxide simulant core debris cooling tests have also been conducted at
KIT as part of the COMET (Miassoedov, Alsmeyer, Cron, & Foit, 2010) and ECOKATS (Alsmeyer,
et al., 2005) programs. In terms of reactor material testing, the COTELS ( (Nagasaka, et al., 1999),
(Nagasaka, et al., 1999), (Zhdanov, et al., 1999), MACE (Farmer, Kilsdonk, & Aeschlimann, 2009),
and NEA MCCI Programs ( (Farmer, et al., A Summary of Findings from the Melt Coolability and
Concrete Interaction (MCCI) Program, 2007), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006),
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010) have been completed to investigate the
mechanisms of coolability under prototypic MCCI conditions. Depending upon melt composition and
conditions, the MACE and NEA MCCI tests have revealed four mechanisms that can contribute to
NEA/CSNI/R(2016)15
core debris quenching: i) bulk cooling in which gas sparging is initially sufficient to preclude stable
crust formation at the melt/water interface (and therefore, efficient heat transfer is achieved); ii) water
ingression through fissures in the core material that augments what otherwise would be a conduction-
limited cooling process; iii) melt (or volcanic) eruptions that lead to a highly porous overlying particle
bed that is readily coolable, and iv) transient breach of crusts that form during the quench process,
leading to water infiltration below the crust with concurrent increase in the debris cooling rate. In
addition to information on the core debris quenching mechanisms, some tests have provided valuable
data on the beneficial impact of water in scrubbing aerosols generated during core-concrete
interaction.
Although it is beyond the purview of this work to perform a detailed review of non-integral
experiments, it is important to note that a variety of simulant as well as reactor material separate effect
tests have been conducted to identify basic phenomenology associated with core-concrete interaction
and debris coolability, as well as providing data to support model development activities. Significant
parts of the MCCI code models are correlations derived from these simulant experiments. These
experiments are summarised in Table 2.2-3. The ARTEMIS test programme (Veteau, 2006) was
carried out at CEA-Grenoble to provide data on core-concrete interaction behaviour in 1D and 2D
configurations using a LiCl-BaCl2 eutectic mixture to simulate concrete and pure BaCl2 to simulate
core melt. Emersion heaters were included to simulate decay heat; gas was injected over the various
ablating surfaces to mock up concrete decomposition gases. These tests provided valuable data on
cavity erosion behaviour including the effects of crust formation and failure at the melt-concrete
interface.
As noted earlier, anisotropic ablation behaviour has been observed in reactor material
experiments for different concrete types ( (Journeau, et al., 2012), (Farmer, et al., A Summary of
Findings from the Melt Coolability and Concrete Interaction (MCCI) Program, 2007). Relevant
information in this area has been provided by the ÉCLAIR (Journeau & Haquet, 2009), BALI (Bonnet,
2000) and CLARA (Bottin, et al., 2016) experiments that have focused on providing quantitative data
on the heat transfer rates to horizontal and vertical surfaces in heated pools of varying viscosity that
are sparged with non-condensable gas. Recent CLARA experiments (Bottin, et al., 2016) provided
additional data in this area using a large scale apparatus. The data ( (Journeau & Haquet, 2009),
(Bottin, et al., 2016) indicates that in the absence of gas sparging, natural convection heat transfer
leads to a large asymmetry in the pool heat flux distribution wherein the heat transfer to the sidewalls
is dominant. Conversely, for low-viscosity pools, even a modest gas sparging rate (i.e. <1 mm/sec
superficial gas velocity) yields an isotropic distribution. The results of the CLARA experiments thus
indicate that asymmetric gas sparging alone is not responsible for the physical differences in ablation
behaviour observed for different concrete types.
In most LWR severe accident scenarios the core melt is expected to consist of discrete metal and
oxide phases and these phases are not miscible. Due to the density difference between these two
phases, stratification between these two phases may occur, and this can influence the heat transfer
partitioning during core-concrete interaction. On this basis, experiments have been performed by
Epstein
(Epstein M. , Petrie, Linehan, Lambert, & Cho, 1981), Casas and Corradini (Casas & Corradini,
1992), and Tourniaire and Bonnet (Tourniaire & Bonnet, 2003) in the BALISE facility at CEA in
order to characterise the threshold pool sparging rate for onset of mixing between two immiscible,
isothermal fluid layers. Aside from sparging rate, the thickness, viscosity, and density of the layers
were varied in these experiments. One key finding is that the gas velocity required for onset of mixing
between the layers increases as the density ratio of the two fluids increases. Aside from these
isothermal tests, heated pool experiments were carried out by Werle (Werle, 1982), Greene (Greene,
1992), and in the ABI facility at CEA-Grenoble ( (Cranga, et al., 2013), (Cranga, et al., Towards an
European consensus on possible causes of MCCI ablation anisotropy in oxidic pool, 2014) in order to
characterise the interfacial heat transfer coefficient between the two layers in a stratified state. These
NEA/CSNI/R(2016)15
experiments have shown that the interfacial heat transfer coefficient between two layers is generally
quite large when the pool is sparged, and that the presence of a metal layer can change the heat
transfer characteristics relative to that observed for other simulant fluids (e.g. water, oil, Freon).
Aside from the above mentioned programmes that address generic phenomenological issues
related to core-concrete interactions, other simulant and separate effects tests have been conducted to
provide data in the area of debris coolability. The PERCOLA experimental programme was carried
out at CEA-Grenoble by Tourniaire et al. (Tourniaire B. , Seiler, Bonnet, & Amblard, 2000) to provide
fundamental insights into the melt eruption process under well-controlled experiment conditions.
These tests utilised fluids with different viscosities to simulate the enrichment of corium with silica
(SiO2) during core-concrete interaction. The experiments revealed liquid ejection by single-phase
extrusion (or ‘fountain’) and two phase jetting mechanisms through a floating solid layer that
represented the interfacial crust that forms during core-concrete interaction. Robb and Corradini (Robb
& Corradini, September 2011) developed a facility for melt eruption experiments that took into
account the experience gained during the PERCOLA programme (Tourniaire B. , Seiler, Bonnet, &
Amblard, 2000). These experiments expanded the PERCOLA database leading to the development of
several entrainment models (Tourniaire B. , Seiler, Bonnet, & Amblard, 2006) including a
phenomenological entrainment model (Robb & Corradini, September 2011).
To identify phenomena associated with melt coolability, (Theofanous, Liu, & Yuen, 1998)
conducted coolability tests using glycerine and liquid nitrogen to simulate the melt and overlying
coolant, respectively. Gas sparging from MCCI was simulated by sparging air through a porous plate
located at the bottom of the apparatus; decay heat was simulated by the sensible heat deposited in the
glycerine due to cooling of the sparged gas. For small-scale tests, quench was not achieved due to the
formation of wall-anchored crusts that inhibited the melt-coolant interaction. However, for the larger
tests, quench was generally achieved, and the time to quench increased with initial melt depth and
decreased with gas sparging rate. The principal mechanism leading to coolability in the quenched tests
was volcanic eruptions, leading to enhanced debris surface area available for contact with the
overlying coolant.
Aside from the simulant material experiments, the SSWICS reactor material separate effect
experiments were conducted as part of the NEA MCCI Program ( (Farmer, et al., A Summary of
Findings from the Melt Coolability and Concrete Interaction (MCCI) Program, 2007), (Farmer,
Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010)
to provide data on: i) the water ingression cooling mechanism, and ii) the mechanical strength of crust
material formed during the quench process. These experiments provided the technical basis for the
development of a correlation for the effective crust dry-out limit due to ingression that can be used in
debris coolability analyses ( (Lomperski, Farmer, & Basu, 2006), (Lomperski & Farmer, 2007). The
crust strength data is valuable in terms of assessing the potential for crust anchoring to occur in plant
accident scenarios (Lomperski & Farmer, 2009), since anchoring and subsequent melt-crust separation
due to ongoing concrete ablation can effectively terminate quenching processes.
In addition to separate effect experiments focused on debris coolability, several experiments have
been conducted to measure additional material and transport properties for ex-vessel conditions. In
particular, (Roche, Steidl, Leibowitz, Fink, & Sehgal, 1993) conducted early experiments to evaluate
liquidus-solidus temperatures as well as viscosity for mixtures of core oxides (UO2, ZrO2) and
concrete for various concrete types. The PRECOS experimental programme (Bottomley, et al., 2012)
was subsequently conducted to evaluate phase diagrams for mixtures of UO2–SiO2, UO2–CaO, UO2–
FeO–SiO2, and UO2–FeO–CaO. Recent tests in the COLIMA facility (Journeau, 2005) have been
carried out to investigate aerosol generation over MCCI pools, as well as aerosol retention in physical
structures such as cracks in concrete.
Finally, although the current database assessment is focused on core-concrete interaction and
debris coolability under top flooding conditions, it is further noteworthy that several experimental
NEA/CSNI/R(2016)15
programmes have been conducted to evaluate the effectiveness of proposed core-catcher concepts for
Gen 3+ LWR plants. For ex-vessel conditions, two general methods have been explored for achieving
long-term melt stabilisation:
A crucible-type technique in which sacrificial material is first introduced into the melt through
ablation at the melt/crucible boundary (with the additives acting to lower the melt pool freezing point),
and then cooling of the crucible exterior boundary with water.
Melt fragmentation, in which water is introduced at the bottom of the melt pool at a slight
overpressure, and the ensuing steam formation acts to cool and solidify the melt in a highly porous
configuration that is readily permeable by water.
An example of the crucible technique is provided by the EPRTM (Fischer & A. Henning, 2009)
that features a spreading room adjacent to the reactor pit in which the melt is cooled by water from the
bottom and top after spreading is completed. Various experiments have been conducted to demonstrate
the heat removal capabilities of this system (Fischer, Herbst, & Schmidt, 2005), including a
generically designed large-scale reactor material experiment that illustrated the capability of a water-
cooled plate in stabilising a core melt after ablating through a layer of sacrificial concrete (Farmer,
Lomperski, Kilsdonk, & Aeschlimann, 2010).
The melt fragmentation technique, denoted COMET, was pioneered by Forschungszentrum
Karlsrühe (FZK). Both simulant and reactor material tests have shown that there are very effective
means for quenching and stabilising core melt (Alsmeyer, Farmer, Ferderer, Spencer, & Tromm,
1998), (Journeau & Alsmeyer, 2006). However, this concept has not yet been deployed in any reactor
design.
2.3. Dry cavity MCCI experiments
The previous sections have outlined phenomenology and supporting experiments related to core-
concrete interaction under dry cavity conditions. This section provides additional technical details
from selected experimental programmes regarding core-concrete interaction under dry cavity
conditions. The Chernobyl Unit 4 accident also provided valuable post-mortem information in this
area; details are provided in Section 2.5.
2.3.1. BETA experiment series
The BETA facility was constructed and operated by KIT (formerly Kernforschungszentrum Karlsrühe,
or KfK) in Germany as a large scale test facility to study melt concrete interaction in a cylindrical
concrete crucible (Figure 2.3-1) ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995). Two test series were
conducted. In the BETA Series I experiments, the melt (1) typically consists of 300 kg steel (Fe,Cr,Ni)
and 150 kg of simulant oxide (initially Al2O3, CaO, SiO2). The melt was generated by a thermite
reaction in the reaction tank (5) and poured into the concrete crucible (2) with an initial cavity of ID
38 cm. The initial melt temperature was ~ 2000°C.
NEA/CSNI/R(2016)15
Figure 2.3-1: Elements of the BETA test facility: 1) melt, 2) concrete crucible, 3) induction coil,
4) off-gas system, 5) thermite reaction tank, and 6) tank for measurement probes
Sustained heating was applied through a cylindrical induction coil (3) surrounding the crucible; the
power was deposited in the metallic phase. Instrumentation was provided to detect the location of the
ablation front, temperature and material sampling of the melt, off-gas composition and flow rate,
aerosol density and composition in the off-gas, and net input power to the melt. Video footage of the
melt upper surface behaviour was also obtained. In the Series I experiments, fission product simulants
were not placed in the melt. Table 2.3-1 summarises the 19 experiments that were carried out in the
Phase 1 test series (1984 to 1986). Most of the experiments used siliceous concrete, but three used
limestone or limestone/common sand concrete. Variations in input power were intended to simulate
different phases of the interaction; i.e. high power densities simulated the early high heat flux and
temperatures shortly after RPV failure, while low power input corresponded to the long term decay
heat level and quasi-steady melt characteristics after the transient cool-down phase.
For these tests, the interaction was dominated by the metallic melt due to the induction heating
technique that deposited energy only in that phase. In reactor situation, only the residual power is
mainly dissipated in the oxidic phase. The consequence is that the overlying oxide melt was cooler and
possibly below the melting temperature. Thus, concrete erosion by the oxide was limited and probably
overestimated by the metal. However, an advantage of induction heating is that interstitial heating
elements (e.g. electrodes) were not required, thereby allowing a true cylindrical cavity configuration to
be simulated. Additionally as the molten concrete and the metallic melt are not miscible, the volume of
the metallic heated layer does not increase with the concrete ablation and so the heat flux does not
decrease contrary to reactor situations.
NEA/CSNI/R(2016)15
Table 2.3-1: BETA Series I experiments

Test Melt Composition Power (kW) 1 Comments
V0.l Iron 0
V0.2 Iron 400 Test of facility
V0.3 Iron + Oxide 1 700

V1.1 Iron Pulsed Power failed
V1.2 Iron + Oxide Pulsed Lorentz forces excluded
V1.3 Steel + Oxide 1 000
V1.4 Steel 0 Transient
V1.5 Steel + Oxide 450
V1.8 Steel + Oxide 1900 No dispersion (CaO)
V1.9 Steel + Oxide 400 – 200 CaO added
V2.1 Steel + Oxide 120- 150
V2.2 Steel + Oxide 50-90 CaO added
V2.3 Steel + Oxide 240 CaO added
V3.1 Steel + Oxide 1 700 – 2 500 US Limestone/Quartz sand, heating from 0 to 66 s only
V3.2 Steel + Oxide 400 – 1 000 US Limestone, 30 min heating
V3.3 Steel + Oxide 600 – 200 US Limestone/Quartz sand, 60 min heating
V4.1 Steel + Oxide 1 000 – 300 600 mm dia. crucible
The oxide phase differs from prototypic melt due to the lack of UO2-ZrO2. Thus, the oxide
initially has a lower density but in the long term both melts are expected to have more similar physical
properties due to the addition of slag from the decomposing concrete. The result is that the BETA
experiments favour segregation of metal and oxide phases with the metal at the bottom. In the BETA
tests the oxide temperature is expected to be lower than the prototype because of the lack of internal
heating in that phase.
The metallic melt composition used in the Series I experiments provides a good representation of
the accident composition for structural steel, but there is a lack of unoxidised Zircaloy cladding which
is important in the early interaction phase. However, the role of Zr was investigated in the Series II
tests described below.
The experimental results for siliceous concrete are summarised as follows. For melts with high
temperatures and high input power input, characterised by small influence of crusts at the
melt/concrete interface, propagation of the metallic melt was predominantly downward. This trend
was especially pronounced in Test VI.8; the post-test cavity profile is shown in Figure 2.3-2. This test
simulated the early high temperature interaction including exothermal chemical interaction processes
from with an overall internal heating rate equivalent to 4.5 kW/kg. This power level exceeds the decay
heat level for a typical PWR at 2 hours following scram by a factor of ~20. The interaction was
1. For a 300 kg mass of steel and a 150 kg of simulant oxide the power density range from 111 W/kg up to
5 555 W/kg, whereas a typical power density for a PWR at 2 hours following scram is about 200 W/kg.
NEA/CSNI/R(2016)15
dominated by violent gas release that agitated the melt, depositing material on the upper crucible
walls. Over the eight minute heating period the melt eroded ~50 cm of the crucible axially at a nearly
constant erosion rate of 1 mm/s. The downward heat transfer was so effective that the melt cooled
down, despite the high heating rate, close to the solidification temperature (Figure 2.3-3).
The evolved gases from concrete decomposition (predominately H2O for siliceous concrete)
resulted in a substantial release of H2 from the reduction of steam by the metal. The molar ratio of H2
-H2O to CO-CO2 was characteristic of silicate concrete.
Figure 2.3-2: Section through the test crucible for BETA test V 1.8
Figure 2.3-3: Temperature of the melt pool for BETA test V 1.8
The low power BETA experiments illustrated the role of solidification processes (viz. crust
formation) on cavity erosion behaviour. The bottom crust of the metal melt at the lower metal/concrete
NEA/CSNI/R(2016)15
interface reduced the extreme downward heat transfer and a more balanced downward-sideward
concrete erosion pattern was observed. The crusts of the partly solidified melt were permeable to gases
released from the concrete, and these gases continued to agitate the melt, thus maintaining convection
which in the long term reduced the melt temperature to near the solidification point. The typical long
term erosion velocity was of the order of a few centimetres per hour. During this phase, H2 continues
to be produced as long as any metal was present in the melt.
The Series I tests also illustrated the tendency for aerosol release in the absence of fission
products and with no core oxides present in the melt. Condensation aerosols were observed, generally
in the form of agglomerated particles of a few µm in diameter. The primary aerosol particles had a
submicron diameter typical for condensation aerosols. All experiments showed intense aerosol release
during the first 1 or 2 minutes of the interaction that correlated with the high gas release immediately
after the melt was poured into the crucible. The release rate then fell to ~ 0.1 g per mole of gas
released from the concrete for both low- and high-power experiments.
The overall cavity erosion behaviour for the limestone concretes was similar to tests with
siliceous concrete, with the exception that lateral erosion for the limestone tests was more pronounced
relative to siliceous. The large amount of CaCO3 in the concrete is calcined, releasing CO2 near
700°C. The resulting CaO (burnt lime) is a very soft material and was found in the sectioned crucibles
in a zone a few millimetres ahead of the melting interface. At low heat transfer rates, this process
could influence the concrete decomposition process; i.e. instead of melting, the CaO could disintegrate
by spallation and dissolve in the melt. This behaviour is different than siliceous concrete that contains
crystalline silica (quartz) aggregate; this material remains intact and does not dissolve quickly in the
melt. For these tests CO2 dominated the gas release, and part of this gas was reduced to CO by
reaction with the metallic melt. One primary difference between the two concrete types is copious
aerosol production for the limestone tests. In particular, a dense white aerosol was released with a
typical diameter of 1 µm. The aerosol production rate was estimated to exceed 1.2 g per mole of gas
from concrete decomposition. The particles were predominately composed of CaO with traces of Na
and K.
The BETA II Series was performed from 1990 to 1992 and included 6 experiments (Table 2.3-2).
Experiments V5.1 to V5.3 investigated the influence of Zr (i.e. Zircaloy-4) on concrete erosion
following the SURC-4 experiments that indicated the potential for Zr-SiO2 condensed phase reactions
(Copus, Blose, Brockmann, Gopmex, & Lucero, Core-Concrete Interactions using Molten Steel with
Zirconium on a basaltic Basemat: The SURC-4 Experiment, 1989). In addition, B4C was added to
represent BWR-specific melt composition. Experiments V6.1 and V6.2 investigated the thermal
stability of a vertical concrete cylinder cooled by water on the outside as melt eroded the inner
surfaces. Finally, experiment V7.1 studies the behaviour of Russian serpentine concrete. All Series II
experiments were performed in cylindrical crucibles similar to the first test series with the same
induction heating technique. Modifications were made to allow the delivery of metallic Zircaloy and
fission product mockups to the melt, and to improve aerosol measurements and sampling. Redundant
and diverse thermocouple techniques were also provided to optimise melt temperature measurements.
For the V5 experiment series, the results indicate that extremely rapid oxidation of the Zircaloy-4
inventory during the first 2 to 3 minutes of the experiments. Material investigations showed that
depletion of Zr had been initiated by 1 minute with a simultaneous increase of Si concentration in the
metallic melt, as is expected to occur by the condensed phase chemical reaction Zr+SiO2 → ZrO2+Si.
In spite of the high energy deposition from Zr oxidation and from electrical heating, the temperature of
the metal in the three experiments dropped to near the freezing temperature within 150 s. This trend
was combined with rapid cavity erosion, high gas release, and extreme voiding of the melt. Gas
release showed a pronounced spike during the first interaction period and then fell to a lower level in
the latter phase of the tests. The axial versus radial erosion behaviour depended on the long-term
power input to the melt; this observation is in agreement with the BETA I experiment findings. High
NEA/CSNI/R(2016)15
aerosol release rates were observed during Zr oxidation; these fell after the Zr was depleted. In terms
of fission product transport measurements, substantial Te was released, whereas the release of Ce and
La were much greater than Sr and Ba. The release of Mo was small. Some portion of B4C that was
added to the melt (to represent BWR absorber material) led to formation of borates in the oxide phase.
These borates contributed to aerosol production.
Table 2.3-2: BETA Series II experiments
BETA test Initial melt Planned heating Main objectives
(~2 200 K) power (kW) 2
V5.1 300 kg Fe + Cr + Ni 400 PWR: Zr chemistry, consequences on
80 kg Zry-4 erosion and aerosol release
50kg Al2O3/SiO2/CaO
V5.2 Same as V5.1 200 BWR with B4C absorber, low power: Zr
chemistry and related processes
V5.3 Same as V5.1 800 BWR with B4C absorber, high power: Zr
chemistry and related processes
V6.1 300 kg Fc, Ni 120 Failure of concrete shield and possibility of
50kg Al2O3/SiO2/CaO late melt flooding
+ Zr02 + Cr203
V6.2 Same as V6.1 120 Failure of concrete shield and possibility of
late melt flooding
V7.1 300 kg Fe, Cr, Ni; 400 Serpentine concrete
80 kg Zr behaviour
50kg Al2O3/SiO2/CaO
The formation of SiO was of minor importance in these experiments, and this may be explained
by the relatively low temperature of the melt in comparison to the ACE experiments where SiO
formation and release played an important role (Fink, Thompson, Armstrong, Spencer, & Sehgal,
1995). The low freezing temperature of the metal is explained by the addition of Zr, Si, B, Cr, and Ni
to the Fe; this effect needs to be considered in computer codes.
The BETA V5.1 experiment was utilised for NEA Standard Problem ISP30 (Alsmeyer, Farmer,
Ferderer, Spencer, & Tromm, 1998). The conclusion from this exercise is that modelling of condensed
phase chemistry is required to describe the rapid Zr oxidation.
Experiment V7.1 conducted with serpentine concrete yielded much higher gas release, which is
attributable to the high content of bound water in the serpentine mineral. Consequently, substantial H2
release was observed. The MgO content of the decomposing concrete lowered the viscosity and
solidus temperatures of the multicomponent oxide melt in comparison with pure silicate concrete.
MgO contributed to aerosol production, nearly doubling the total aerosol mass.
The two experiments V6.1 and V6.2 on radial concrete penetration simulated the expected decay
heat level at 8 hours after initiation of concrete erosion. The experiments clearly demonstrated that
outside water cooling is unable to stabilise the melt in the concrete. This is due to the fact that the
concrete thickness by which heat can be transferred through the walls by conduction is only 5 to
10 mm, and this concrete thickness is mechanically unstable. Hence, further spreading of the melt and
direct contact with water will occur.
2. Power density ranging from 342 W/kg up to 1 860 W/kg
NEA/CSNI/R(2016)15
2.3.2. COMET-L2 experiment

The COMET-L2 experiment (Miassoedov A. , et al., 2008) was carried out using base technology
developed as part of the BETA Program ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995) to investigate long-
term melt–concrete interaction of metallic corium with siliceous concrete in a 2D cylindrical crucible.
The test was carried out under dry conditions with simulation of decay heat at intermediate levels during
the first phase of the experiment, and at low levels during the second phase. The apparatus is shown in
Figure 2.3-4.
Figure 2.3-4: COMET-L2 test apparatus
The corium melt is simulated by a metallic phase (~ 430 kg Fe and Ni) initially overlaid by an
oxide phase (~35 kg Al2O3–CaO). The melt was produced by a thermite reaction at an initial
temperature of ∼1 750°C in a separate melt generator. Oxide additives were incorporated into the melt
to lower the solidification temperature and increase the solidus–liquidus temperature range to be more
reflective of actual core debris (e.g. see (Roche, Steidl, Leibowitz, Fink, & Sehgal, 1993). The melt
was then poured into the 60 cm diameter concrete test crucible and induction heating was applied from
below the test section to the metal phase to simulate decay heat levels in the range of 150–250 kW.
This power level is indicative of ex-vessel accident conditions a few hours after the start of the cavity
erosion. In the second phase of the test, input power was reduced to ∼200 kW to observe the onset of
possible crust formation and its influence on the erosion process. The concrete crucible was
instrumented with thermocouples to detect the location of the ablation front and to provide an
indication of the melt temperature. An optical pyrometer was used to provide debris upper surface
temperature measurements. The facility also included an instrumented off-gas system.
Upper and lower bound estimates on the oxide phase temperature during the experiment based on
optical pyrometer measurements are provided in Figure 2.3-5. These measurements indicate a rapid
initial reduction in the oxide phase temperature, in agreement with the BETA measurements
(Alsmeyer, 1987), (Alsmeyer, et al., 1995). The cavity erosion data (Figure 2.3-6) indicates that the
initial axial-radial erosion ratio is ∼1:1 until the melt superheat is effectively lost by heat transfer to
the concrete. This occurs at ~100 seconds into the test sequence. However, during the longer-term
interaction, the axial erosion rate is two to three times higher than the radial rate. Heating of the melt
was terminated at 1 015 seconds when the axial erosion limit for the facility was reached. As shown
NEA/CSNI/R(2016)15
inFigure 2.3-7, final axial ablation depths were in the range of 9 to 17 cm, while radial erosion was
limited to ~ 3 cm.
Figure 2.3-5: Oxide melt temperature evolution in COMET-L2
Figure 2.3-6: Axial (solid lines) and radial (dotted lines) concrete erosion
by the metal melt phase in COMET-L2
NEA/CSNI/R(2016)15
Figure 2.3-7: Post-test debris distribution for COMET-L2

During the long-term interaction, quiescent periods were interrupted by short but intense periods
of melt eruptions that may be attributable to formation and failure of metal crusts at the core-concrete
interface. Low heat fluxes to concrete result in thermal decomposition of the concrete matrix and the
release of un-melted silica aggregate into the melt. When compared to heat fluxes experienced early in
the sequence, this process would lower the effective concrete decomposition enthalpy and may alter
the heat flux and concrete erosion patterns in the crucible. In addition, incorporation of solid aggregate
will cause bulk cooling of the melt to occur.
2.3.3. SNL dry cavity experiments
Sandia National Laboratories carried out several experiment series on core-concrete interaction to
support model development and code validation activities (Table 2.2-1). Both transient as well as
sustained heating tests were carried out, and the various experiment series are described under these
two headings below since these topical areas roughly correlate with the chronological order under
which the tests were carried out. Both simulant as well as reactor materials were used in the test
programmes.
Transient experiments
Early transient experiments were carried out as part of the Large Scale test series in which molten
stainless steel at ~ 1 700°C was poured into limestone-common sand (LSL tests 1-4), basalt (LSB
tests 2-4), and limestone (LSCRBR tests 1-2) hemispherical crucibles to provide basic information on
melt-concrete interaction processes. Each test crucible was exposed to molten steel several times to
simulate prolonged periods of interaction.
These experiments indicated that gas generation from concrete decomposition leads to a strong
initial melt agitation phase for all three concrete types. As the material cools, agitation diminishes to
the point where the interaction resembles a boiling liquid. A viscous layer of slag eventually forms
over the metal; the layer is permeable to the sparging gases. As the material cooled, the slag eventually
hardened but vent holes for gases persisted. H2 and CO generated by reduction of water vapour and
carbon dioxide from metals in the melt were detected over the interaction. Aerosols from oxidised
metal and concrete decomposition products were also detected. In general, these tests provided
qualitative indications of phenomena that were later confirmed in sustained heating experiments.
NEA/CSNI/R(2016)15
These early experiments were followed by the TURC transient tests that featured improved
instrumentation to investigate limestone-common sand (LCS) concrete erosion behaviour in 1D test
crucibles. The main objective was to compare metal melt erosion behaviour in tests TURC-1T and
TURC-1SS (Gronager, Suo-Antilla, Bradley, & Brockmann, 1986) to core oxide (U02-Zr02)
behaviour in tests TURC-2 and TURC-3 (Gronager, Suo-Anttila, & Brockmann, 1986). All
experiments used 41 cm ID refractory MgO crucibles lined at the bottom with a slab of LCS concrete.
The melts were first produced (using a thermite reaction for the steel tests and by induction heating of
embedded tungsten susceptors for the core oxide experiments) and then poured into the test crucibles.
Initial melt temperatures ranged from 2 100 to 2 500°C. Concrete erosion was limited to ~ 7 cm; the
erosion rate ranged from ~ 0.5 mm/sec at the start of the transient down to ~0.03 mm/sec near the end.
Gas measurements confirmed the formation of CO and H2. Melt temperature measurements during the
interaction were not obtained, and so information on heat transfer coefficients could not be obtained.
Aerosol measurements in TURC-1 SS indicated concentrations in the range of 15 to 80 g/m3,
which are considerably higher than the levels of ~5 g/m3 measured in the BETA tests. The difference
may be explainable as follows. The melt in the TURC-1SS experiment was covered by a thin layer of
oxide whereas in BETA the layer depth was significant. Second, the major release of aerosols in
TURC-1SS was in the form of volatile fission product simulants (i.e. Te, I, and Cs), and these
simulants were not present in the BETA tests. Low levels of Ce and La were also detected. The mean
diameter of the aerosols was 1 µm or less.
The core oxide melt masses in the TURC-2 and TURC-3 experiments were 147 kg and 46 kg,
respectively, with a composition reflective of a PWR (i.e. 70 wt% UO2 and 30 wt % ZrO2). In TURC-
3 experiment, 9 wt % metallic Zr was added to the melt. Both experiments indicated rapid crusting of
the melt upon introduction into the crucibles with no detectable concrete ablation. The small amount
of superheat in these melts in conjunction with the lack of input power to simulate decay heat allowed
the melts to cool without significantly damaging the concrete. Even though the cavities were not
ablated, steam and CO2 were still released and these gases were partially reduced to H2 and CO.
Detected aerosols consisted mostly of Te, I, and Cs. The low volatile fission products had low
concentrations or were below the detection limit. Note that an initial rapid crusting phase leading to a
delay in the onset of ablation under sustained heating conditions was also observed in the CCI tests
carried out in the NEA MCCI Program ( (Farmer, et al., A Summary of Findings from the Melt
Coolability and Concrete Interaction (MCCI) Program, 2007), (Farmer, Lomperski, Kilsdonk, &
Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010).
Sustained heating experiments

Early sustained heating experiments were carried out as part of the HSS (Copus & Bradley, 1986) and
FRAG (Tarbell, Bradley, Blose, Ross, & Gilbert, 1987) programs to examine the penetration of solid
steel and core oxide debris into different types of concrete. These tests examined scenarios in which
core debris may be quenched during relocation through water following vessel failure, and then reheat
to begin core-concrete interaction. This type of interaction may also occur late in the sequence after
cool-down and partial freezing of the metallic phase which would reside at the bottom of the melt pool
after the core oxides are diluted with slag from concrete decomposition.
The FRAG experiments used 45 kg of steel spheres (3-4 mm diameter) interacting with both
basalt and LCS concrete. The steel was inductively heated for up to four hours. In two of the four tests,
water was added to the debris after the interaction was initiated to determine the effect of flooding on
the erosion behaviour and temperature of the debris.
In the absence of water, the results show that the solid material reaches a temperature between
1 200 and 1 400°C. Axial erosion rates were in the range of 3 to 4 cm/h. Slag from the erosion process
migrated through the debris and formed a crust covering the upper surface. Two-thirds of the steam
and CO2 from concrete decomposition were reduced to H2 and CO by oxidation of the hot steel
NEA/CSNI/R(2016)15
spheres. The addition of water showed that a crust layer may limit water ingression; complete
quenching of the debris did not occur.
The HSS-1 and HSS-3 (Copus & Bradley, 1986) experiments produced similar results with a
heated solid slug of steel (HSS-1) and a heated slug of UO2-ZrO2 (HSS-3). Erosion was maintained in
these tests over periods of 3 to 4 hours at a typical rate of 1.5 cm/h; the total erosion depth was ~6 cm
(Figure 2.3-8). The average steel temperature was 1 350°C, while the bulk temperature of the oxide
temperature was 1 650°C; both of these temperatures are well below the melting points of the
respective materials.
Figure 2.3-8: Concrete erosion by steel (HSS-1, left) and by U02-Zr02 slugs (HSS-3, right)
(Copus & Bradley, 1986)
The FRAG and HSS tests thus showed that even with low heat fluxes imposed by hot solid
material, the material is able to melt the concrete as long as heat conduction into the concrete is unable
to remove the heat.
Aside from the solid material tests, concrete erosion by liquid melt with sustained heating was
investigated in the SURC test series. The SURC-4 test (Copus, Blose, Brockmann, Gopmex, & Lucero,
Core-Concrete Interactions using Molten Steel with Zirconium on a basaltic Basemat: The SURC-4
Experiment, 1989) was the first experiment to illustrate the role of condensed phase chemical reactions
between Zr and SiO2 from concrete erosion. Major components of the SURC-4 apparatus included a
sealed and water-cooled containment vessel, interaction crucible, induction coil, and zirconium delivery
tube (Figure 2.3-9 SURC-4 experiment apparatus). As for the TURC experiment series, the test crucible
consisted of a 40 cm ID refractory MgO cylinder with a basalt concrete slab located at the base, thus
limiting erosion to the axial direction. Thermocouples were used to measure melt temperature, as well as
local erosion rates and temperatures within the concrete. Additional instrumentation was provided to
monitor the composition and flow rate of evolved gases. An aerosol detection system was also
incorporated to measure the composition and release rates as a function of time.
NEA/CSNI/R(2016)15
Figure 2.3-9: SURC-4 experiment apparatus

Source: Copus, Blose, Brockmann, Gopmex, & Lucero, Core-Concrete Interactions using Molten Steel with
Zirconium on a basaltic Basemat: The SURC-4 Experiment, 1989.
The experiment was carried out with a 200 kg stainless steel charge with an additional 6 kg of
simulated fission products (Mo, Te, La2O3, CeO2, BaO). The melt was produced by in-situ induction
heating. At 105 minutes in the experiment sequence concrete erosion began. At 119 minutes, 20 kg of
zirconium metal was dropped into the melt. The experiment proceeded until 162 minutes,
corresponding to 57 minutes of concrete erosion with at a net heating rate that ranged from 51 to
62 kW. At this point, the test section failed which effectively terminated the experiment. After the
Zr was added, the concrete erosion near the centre of the crucible increased over several minutes from
16 cm/hr to 30 cm/hr (Figure 2.3.10). In addition, the gas volumetric flow rate (mainly H2) and
aerosol production rate increased dramatically at this time, apparently driven by the Zr oxidation
process. Simultaneously, an increase in melt temperature of ~130 K occurred (Figure 2.3.10), driven
by the exothermic nature of the Zr condensed phase chemical reaction, in parallel with gas phase
reactions. Large volumes of foam slag with 17-18% of oxidised ZrO2 also formed during this
interaction. Aside from these observations, the experiment provided no indications of the coking
reaction (formation of elemental carbon), as CO was continuously released. Regarding aerosol
chemistry, Te was found to be the dominant species with lesser amounts of Fe, Mn, Na, and
K released. For the other fission product simulants, Ce was detected at a level below 1%, while Ba,
Mo, and La releases were below the detection limit (< 0.1%).
The SURC-3 and SURC-3A experiments (Copus, Blose, Brockmann, Gomez, & Lucero, 1990)
were carried out to determine whether CO2 from concrete decomposition would be reduced by Zr to
produce elemental carbon through the coking reaction. The crucibles in these tests had an ID of
21.6 cm, and limestone concrete was utilised which would optimise the chances for coking reactions if
they were to occur. The SURC-3 experiment utilised a 1D crucible (concrete at the base), while the
crucible for SURC-3A was made entirely from limestone concrete.
NEA/CSNI/R(2016)15
Figure 2.3-10: SURC-4 Axial erosion front location (left) and melt temperature (right).
The experimental procedure was similar to that used in SURC-4. After the 50 kg stainless steel
charge was melted and concrete erosion initiated, Zr was added to the melt, resulting in a
Zr concentration of 2% in SURC-3 and 5% in SURC-3A. In both tests, the concrete ablation rate, melt
temperature, off-gas rate, and aerosol concentration increased over a period of 10-15 minutes
following Zr addition. In both experiments, CO was the dominant gas species both prior to and after
Zr addition, which confirms that the coking reaction did not occur at a significant level. Rather, the
energy source driving increases in melt temperature and concrete ablation rate was the highly
exothermic reaction between Zr and H2O and CO2 from concrete decomposition, yielding H2 and CO.
The typical temperature of the melt pool was 1 600°C, and the pool temperatures were reported to
increase by 80 to 200°C during the oxidation phase. The aerosols were comprised primarily of Te, Na,
K and Fe oxides. The release of other low volatile fission products was generally below 0.02%. The
SURC-3A experiment resulted in radial concrete erosion to a depth of 9 cm versus 25 cm in the axial
direction. This ratio is comparable to the BETA V3.2 experiment (Alsmeyer, et al., 1995).
The SURC-1 (Copus, Blose, Brockmann, Simpson, & Lucero, 1992-1) and SURC-2 (Copus,
Blose, Brockmann, Simpson, & Lucero, 1992-2) experiments were reactor material tests that provided
1D core-concrete interaction data for limestone-common sand and basalt concrete types, respectively.
Both tests utilised ~200 kg of core oxide (UO2 – ZrO2 ) containing 18 kg of metallic Zr. The overall
facility layout and crucible dimensions were the same as that used for SURC-4
(see Figure 2.3-9 SURC-4 experiment apparatus). Decay heat was simulated by induction heating of
tungsten susceptors located directly in the core debris. The input power ranged from 50 to 100 kW,
which corresponded to specific power densities of 250 to 500 W/kg. These power densities are
indicative of a PWR a few hours after shutdown. In addition to the UO2/ZrO2/Zr core charge for both
tests, 3.4 kg of fission product simulants were added prior to heating. These additives consisted of
BaMo04, La203, Ce02, and Nb205.
The tests showed two characteristic phases after onset of concrete ablation. In the first phase
when metallic zirconium was still present, the erosion rate was as high as 15-30 cm/hr
(see Figure 2.3-10 and Figure 2.3-11). After the Zr was fully oxidised, the ablation rate decreased to
5-15 cm/h. Since the decomposition enthalpy of limestone concrete is higher compared to Basalt, the
erosion rate in SURC-1 was lower than that compared to SURC-2. The high early erosion rate for both
tests is attributable to exothermic oxidation of Zr by concrete decomposition gases. Due to the high
carbonate content, the gas composition for the limestone concrete test was dominated by CO and CO2,
with CO dominant during the Zr oxidation process. In contrast, the gas composition for the basalt test
was dominated by H2. However, the cumulative gas release for this test was lower.
NEA/CSNI/R(2016)15
The temperatures of the melt pool (see Figure 2.3-10 and Figure 2.3-11) were initially controlled
by the high melting point of the core oxide mixture, but as concrete decomposition products were
added, the pool temperature decreased throughout the test to 1 700-1 800°C. In general, the
temperatures of the oxidic melts are several hundred degrees higher in comparison to the metallic melt
tests. Aerosol release rates ranged from 1 to 10 g/min for both tests. The aerosol compositions were
dominated by concrete species (Si, Na, K, and Ca). The U and Ba concentrations were less than 0.3%.
Releases of Mo, La, Ce and Nb were smaller still. Releases of some of the refractory components were
smaller for the basalt test, which may be attributable to silicates from the basaltic concrete.
Figure 2.3-11: SURC-1 (top) and SURC-2 (bottom) axial erosion front location (left)
and melt temperature (right)
2.3.4. ACE phase C experiments

The ACE Phase C experiments were carried out at Argonne National Laboratory between 1988 and
1991, see (Thompson D. H., Fink, Armstrong, Spencer, & Sehgal, 1992), (Fink, Thompson,
Armstrong, Spencer, & Sehgal, 1995). The objectives of these tests were to: i) quantify fission product
release during core-concrete interaction, and ii) provide data on thermal-hydraulic behaviour to
support model development and code validation activities. All experiments used an initial charge mass
of ~300 kg. The melt consisted of U02-Zr02, fission product simulants, concrete decomposition
products, and some structural materials to obtain a typical corium mixture. Several tests included
NEA/CSNI/R(2016)15
control rod material to evaluate the effect of this material on fission product release. Decay heat was
simulated using direct electrical heating (DEH) of the core debris at power levels experienced a few
hours into the accident sequence.
The apparatus was 1D with an internal cross-section of 50 cm x 50 cm (see Figure 2.3-12). The
concrete basemat was situated at the base of the apparatus. In most tests zirconium-bearing
concrete/metal inserts were placed immediately over the basemat that contained the initial metal
content of the melt, as well as volatile fission product materials (e.g. Te). The core melt was produced
in situ over the basemat by melting the powders from the top down after establishing an electrical
conduction path near the top centre of the powder bed. The inserts were included in the design so that
metals and volatile fission products would be introduced into the melt at the last stage of the melt
production process in order to minimise early oxidation of clad material and fission product release
during the heat-up phase that typically took 2 to 3 hours. Onset of concrete attack was defined by the
signal from a thermocouple located at the top surface of the concrete near the axial centreline. The
maximum allowable axial ablation depth for the facility was 13 cm; the typical operating time to
achieve this ablation depth was 1 hour.
Figure 2.3-12: Schematic (left) and photograph (right) of the ACE MCCI test apparatus
The facility was heavily instrumented with a complete gas and aerosol diagnostics system,
temperature measurements in concrete layer and melt pool, electrical power input, and additional
support information for the water cooling systems. Given the calculated (water-cooled) heat losses to
the sidewall panels of the apparatus, the gross heating power was adjusted online to compensate for
sidewall losses so that the net heating rate was maintained at the target level.
As shown in Table 2.3-3, seven experiments were performed. The principal parameters varied in
the test matrix were the concrete type, the extent of cladding oxidation, the core type (i.e. PWR vs.
BWR with the corresponding control materials and U/Zr ratios), and the net internal heat generation
which was generally consistent with decay heat at 2 hours after scram. The corium compositions also
included the inactive fission products mockups BaO, La2O3, SrO, CeO2, MoO2, SnTe, ZrTe2, and
Ru, along with the absorber materials B4C or Ag/In.
Principal findings from the ACE MCCI tests are summarised as follows. The temperature of the
core oxide melts were higher than originally expected, exceeding 2 100°C (2 400 K) in the majority of
the tests (see Figure 2.3-13). This trend is in agreement with the SURC-1 and SURC-2 test data
(Copus, Blose, Brockmann, Simpson, & Lucero, 1992-1), (Copus, Blose, Brockmann, Simpson, &
Lucero, 1992-2) during the early stages of the interaction. Subsequent melting point measurements by
NEA/CSNI/R(2016)15
(Roche, Steidl, Leibowitz, Fink, & Sehgal, 1993) indicated that these temperatures are due to the
extended solidus-liquidus range for core oxide materials.
Table 2.3-3: ACE Phase C test matrix
Test Concrete Net Heat Generation (W/kg Reactor Initial Clad Oxidation Control Rod
Type UO2) Type (%) Material
L5 LCS 325 PWR 100
L2 SIL 450 PWR 70
LI LCS 350 PWR 70
L6 SIL 350 PWR 30 Ag, In
L4 Serpentine/SIL 250 BWR 50 B4C
L7 SIL 250 BWR 70 B4C
L8 LL 350/150 PWR 70 Ag, In
Figure 2.3-13: ACE phase C experiments ablation front locations (left) and melt temperatures (right)
Downward erosion rates (Figure 2.3-13) cluster in two different bands; the first around
~ 4 mm/min (24 cm/h) and the second around ~ 1 mm/min (6 cm/h). This is in the range of the erosion
rates measured in the SURC-1 and SURC-2 tests. For tests with the lower erosion velocity, downward
heat transfer may be limited by an insulating crust at the melt-concrete interface, or possibly a highly
viscous sublayer in the melt.
For the fully oxidised test L5, off-gas analysis showed that no reduction of H2O or CO2 concrete
decomposition gases occurred. However, H2 and CO formed in all other tests where metal was present.
Downward migration of 25% to 50% of all decompositions gases through the concrete and out the
bottom of the unsealed apparatus was reported (Fink, Thompson, Armstrong, Spencer, & Sehgal,
1995) for these tests. This behaviour is most likely unprototypic. Gas release for the various
experiments increased in direct proportion to the amount of gases present in the different concrete
types, as expected.
Because of the high temperatures of the melts, the formation of gaseous SiO through endothermic
Zr-SiO2 reactions was important (Fink, Thompson, Armstrong, Spencer, & Sehgal, 1995). This
reaction most likely occurred in parallel with exothermic gas phase reactions between Si metal and
NEA/CSNI/R(2016)15
concrete decomposition gases. In several experiments the evolved SiO(g) condensed in the cooler
upper plenum of the test section where it decomposed into SiO2 and Si, as well as forming silicates
with other aerosols. These condensed reaction products nearly plugged the diluter in two of the tests
(L2 and L6), both of which utilised siliceous concrete (Fink, Thompson, Armstrong, Spencer, &
Sehgal, 1995).
Si species also dominated the aerosols for the limestone concrete tests. The release was highest
during insert ablation when Zr concentrations in the melts were highest. High Si concentrations in the
aerosols were also observed in the SURC-1 and -2 tests. However, in the lower-temperature BETA-II
and SURC-4 tests, major amounts of SiO(g) were not produced.
Melt foaming in tests L1 and L5 also occurred; this was due to a combination of appreciable
SiO(g) release coupled to high gas release rates from carbonate decomposition. In a real accident
foaming may reduce aerosol release and heat transfer to overlying structure. Foam formation in oxidic
pool during MCCI has been studied by (Tourniaire, Dufour, & Spindler, 2009).
Figure 2.3-14: Mass fraction of major fission products and U released as aerosols during the ACE
tests (Fink, Thompson, Armstrong, Spencer, & Sehgal, 1995)
With the exception of Te the release of fission products was small (less than 1% of the aerosol
mass). This is consistent with data from the SURC and BETA tests. Figure 2.3-15 provides a
comparison of the fission product elements and U released for the different ACE tests.
2.3.5. MOCKA experiments
The MOCKA experimental programme (Foit, Cron, Fluhrer, Miassoedov, & Wenz, 2012) is currently
underway at KIT in Germany. The purpose of the tests is to provide a more in depth investigation of
2D cavity erosion behaviour using a simulant oxide-metal melt in a stratified configuration. These are
transient experiments in which additional energy is supplied to the melt by means of alternating
additions of thermite and Zr to extend the duration of the interaction. The heat generated by the
thermite reaction and exothermic oxidation of Zr was mainly deposited in the oxide phase.
The experiments were performed in using siliceous concrete crucibles with an ID of 25 cm. The
thermite reaction produced an initial melt mass of 110 kg, consisting of a 42 kg Fe metal phase and a
68 kg oxide phase (56 wt% Al2O3 and 44 wt% CaO) at a temperature of ~1 900°C. The collapsed
height of the metal melt was about 13 cm.
The first experiments, MOCKA 1.1-1.3, were carried out to scope out the behaviour of an oxide-
metal melt in a cylindrical concrete crucible without external energy addition. Tests 1.1 and
1.2 utilised 1D crucibles, while the balance of the tests (1.3 to 1.7) utilised 2D crucibles. The axial
NEA/CSNI/R(2016)15
erosion by the metal melt in test 1.1 reached 3 cm. However, in test 1.2 no axial erosion occurred, and
post-test examinations indicated that this was due to the fact that the metal was encased in a ~0.2 mm
thick layer of insulating oxide. In the first 2D test (i.e. MOCKA 1.3), axial erosion was more
pronounced than radial (see Figure 2.3-15). This observation is consistent with the previous metal-
oxide sustained heating tests carried out in the BETA ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995) and
COMET-L (Miassoedov A. , et al., 2008) tests.
Figure 2.3-15: Post-test debris configuration for MOCKA test 1.3

To avoid the unexpected outcome of the 1.2 test, 3 kg of Zr were deposited near the crucible wall
in tests 1.4 and 1.5. The additional energy released during the exothermal Zr oxidation reactions lead
to considerably more concrete erosion in both these experiments. The ratio of the axial to the lateral
erosion was found to be ~2 in both tests.
Figure 2.3-16: Cavity erosion behaviour (left) and post-test debris (right) for MOCKA 1.7 test
To extend the duration of the interaction with the concrete and allow for significant concrete
erosion by the oxide as well as the metal in MOCKA 1.6 and MOCKA 1.7, thermite and Zr were
successively added to the melt over the course of these experiments. These tests were also performed
with a slightly higher initial Zr mass of 4 kg (compared to 3 kg in tests 1.3 and 1.4) positioned along
the initial concrete surfaces. Following the initial Zr oxidisation phase, 63 kg of Fe-Al2O3 thermite
and 24 kg Zr metal were added to the melt at 11 minutes into the experiment sequence for test 1.6.
Similarly, 117.5 kg of thermite and 34 kg Zr metal were added after 18 minutes in test 1.7. Since the
NEA/CSNI/R(2016)15
energy from the Zr oxidation reactions was mainly deposited in the overlying oxide, significant
ablation of the crucible wall occurred. In particular, for test 1.6 the axial-radial erosion depths for the
metal phase reached 10 cm and 5 cm, respectively, while for the oxide phase radial erosion reached
4.5 cm. For test 1.7 (Figure 2.3-16), the maximum axial-radial ablation depths reached 15 cm and
6.5 cm, respectively. The average axial ablation rate of ~ 7.7 mm/min for this test was as high as that
observed in the COMET-L2 experiment (Miassoedov A. , et al., 2008).
In summary, the MOCKA tests have produced overall results similar to the former BETA and
COMET-L experiments. However, in contrast to these previous experiments, significant lateral
concrete erosion by the oxide melt was observed when chemical energy was added to mock up decay
heat in the oxide. The pronounced downward erosion by the metal phase seems to be inherent for
melts containing a significant fraction of structural steel.
The programme is currently focused on assessing the influence of concrete reinforcement (rebar)
on cavity erosion behaviour since data in this area are sparse. Two tests with reinforced siliceous
concrete have been conducted thus far (MOCKA 3.1 and 3.2) (Foit, Fischer, Journeau, & Langrock,
2014). The initial thermite mass in the experiments was ~100 kg; 90 to 150 kg of thermite and Zr were
added to the melt over the first 10-20 minutes to extend test duration. Both experiments produced
isotropic ablation behaviour (i.e. ~7 cm in MOCKA 3.1 and ~11-12 cm in MOCKA 3.2). Two
additional tests were performed under similar conditions but without rebar in the concrete to form a
technical basis for assessing the effect of the reinforcement. Unlike the tests with the reinforced
concrete, these two counterpart tests yielded anisotropic cavity erosion behaviour similar to that
observed in the BETA tests ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995) (i.e. axial ablation
approximately twice radial; see Figure 2.3-17). Additional LCS tests have been recently performed to
assess also the effect of the concrete composition (Foit J. J., MCCI on LCS concrete with and without
rebars, 2015).
(a) (b) (c)

Figure 2.3-17: (a) Bottom view of MOCKA test section prior to concrete, and views of post-test
debris for (b) MOCKA 3.1 with rebar and (c) MOCKA 3.3 without
2.3.6. HECLA experiments

The HECLA test series (Sevón, et al., 2010) was carried out at VTT Technical Research Centre from
2006 to 2009. These experiments investigated the interaction between metallic melt and concrete by
pouring molten stainless steel into cylindrical concrete crucibles. These were transient tests
(i.e. external heating to simulate decay heat was not provided). The main objective of the study was to
examine the behaviour of concretes containing both siliceous as well as hematite (Fe2O3) aggregates.
The latter concrete type is used as a sacrificial material in the EPR reactor pit (Fischer & A. Henning,
NEA/CSNI/R(2016)15
2009) in Finland. Data on the high temperature behaviour of hematite concrete during core-concrete
interaction is very sparse in the literature.
A schematic diagram illustrating the basic elements of the test facility is provided in Figure
2.3-18. The facility consisted of an induction heating furnace for melting the steel charge and a steel
chamber in which the test was carried out. The melt was poured from the furnace into a pre-heated
tundish above the steel chamber. The melt flowed from the tundish through a bottom hole into the
chamber and fell into the concrete crucible. To prevent hydrogen burns, the steel chamber was inerted
with a continuous nitrogen purge.
HECLA-1 was a scoping test to shake down the facility using a small (20.4 kg) steel charge mass
at relatively low temperature and limited instrumentation. The last four tests were larger and employed
50 kg of stainless steel melt. The HECLA-1 and -2 crucibles were made of ordinary siliceous concrete,
while the last three experiments used hematite concrete. A photograph of the HECLA-5 crucible is
provided in Figure 2.3-18. All tests were carried out with 2D crucibles with an ID of 28 cm. The melt
pool depth was ~15 cm for the 50 kg steel charge.
Induction furnace
N2 in
Tundish
Concrete
crucible
Gases out
Figure 2.3-18: Schematic of HECLA test facility (left); photograph of the HECLA-5
concrete crucible (right)
The concrete materials for the HECLA-4 and -5 experiments were obtained directly from the
Olkiluoto 3 EPR construction site. The air content of the HECLA-4 concrete crucible was higher than
the prototype, but the HECLA-5 concrete was within the technical specifications for the plant. This
concrete contains 44.4 wt% SiO2 and 32.1 wt% Fe2O3. The density of the concrete was 2 620 kg/m3.
This is heavier (by ~ 15%) than ordinary siliceous concrete; the difference is due to the Fe2O3.
GEMINI2 calculations with the NUCLEA-09_1 database predict that the solidus and liquidus
temperature of this concrete are 1 140°C and 1 430°C, respectively; at 1 160°C, 50% of the concrete is
molten. The calculated solidus and liquidus temperatures agree with those measured using the
differential thermal analysis (DTA) technique (Sevón, et al., 2010).
Measured axial and radial concrete ablation depths as a function of time for the HECLA tests are
shown in Figure 2.3-19, while the main parameters and results are summarised in Table 2.3-4. The
final ablation depths ranged from 10 to 25 mm, with the exception of HECLA-3, for which no
NEA/CSNI/R(2016)15
sidewall ablation took place. This may be a consequence of the slower melt pouring rate that occurred
in this test (Sevón, et al., 2010). Comparison of the HECLA-2 ablation results with the three
subsequent tests that were performed with the hematite concrete indicates no clear difference in
ablation behaviour for the two concrete types. The only minor difference is the sidewall ablation rate,
which was slightly higher in HECLA-2. However, due to the relatively small number of tests and
scatter in the results, no definitive conclusion can be drawn.
25 25
HECLA-5
Basemat ablation (mm)
Sidewall ablation (mm)

20 20
HECLA-2 HECLA-4 HECLA-4

15 15
HECLA-2
10 10
HECLA-3
HECLA-5
5 5
0 0
0 25 50 75 100 0 25 50 75 100
Time (s) Time (s)
Figure 2.3-19: Concrete ablation in the HECLA experiments
Another important conclusion from the tests is that significant cracking (e.g. from thermal shock)
and/or spallation of the concrete did not take place during the melt pour or subsequent metal-concrete
interaction. In HECLA-4 a pit was formed at the centre of the basemat where the melt jet impinged.
However, the depth of the pit was relatively small (i.e. < less than 15 mm).
To summarise, important information was gained from the HECLA experimental programme on
the ablation characteristics and properties of hematite concrete used at the Olkiluoto 3 power plant.
Although the number of tests was limited, no dramatic differences in ablation behaviour were found.
In addition, chemical composition and mechanical property characteristics for this concrete type were
measured; this data is documented in (Sevón, et al., 2010).
Table 2.3-4: Main parameters and results for the HECLA experiments
Test
Parameter HECLA-1 HECLA-2 HECLA-3 HECLA-4 HECLA-5
Concrete type Siliceous Siliceous FeSi FeSi FeSi
Melt mass (kg) 20.4 50 50 50 50
Melt temperature (°C) 1 560 1 700 1 650 1 770 ~1 780
Average basemat ablation rate (mm/s) 0 0.28 0.26 0.17 0.24
Average sidewall ablation rate (mm/s) 0 0.26 0 0.20 0.15
Total basemat ablation (mm) 0 25 25 15 25
Total sidewall ablation (mm) 0 15 0 20 10
2.3.7. MCCI experiments in the SICOPS facility

Areva´s SICOPS experimental programme ( (Langrock & Hellmann, 2010), (Hellmann & Fischer,
2007), (Hellmann, 2000) was conducted with the objective of resolving two open technical questions
NEA/CSNI/R(2016)15
related to core-concrete interaction; namely: i) the influence of high core melt metal content, and
ii) 2D (axial vs. radial) power split during core-concrete interaction. This work was conducted to
support development of the EPR core-catcher concept that utilises hematite concrete as a sacrificial
material in the reactor pit, as well as a zirconia-based refractory layer that backs up this material
(Fischer & A. Henning, 2009).
The 1D tests addressed the phenomenology and kinetics of melt – concrete interaction with
mixed metallic/oxidic melts under sustained heating conditions. Experiments with prototypic core
oxide melt under identical conditions were performed for comparative purposes. The conditions
selected in the experiments were consistent with both the early and late phases of the core-concrete
interaction. Specific experiment objectives were to: i) evaluate concrete erosion rate as a function of
input power for different concrete types, ii) determine the effect of high metal content on the concrete
erosion behaviour, and iii) investigate the effect of metal content on melt physicochemical behaviour
during the interaction. In addition, non-destructive techniques were used to obtain information on the
location and amount of metal in solidified debris samples.
The 2D tests focused on clarifying differences in radial-axial power splits during core-concrete
interaction and identifying parameters that influence this ratio. The tests further addressed the
following questions: i) does power increase lead to increase of temperature and/or erosion rate, ii) are
crusts formed at melt- concrete interface, and iii) does the melt temperature decrease in the long term
below the melting point of steel reinforcement.
The SICOPS facility used a high-frequency induction heating method to simulate decay heat,
coupled with a cold crucible technique for containment of the core melt material. This technology can
be used to produce a broad spectrum of prototypic (UO2-bearing) melts for experiment purposes. The
crucibles for the experiments are relatively small-scale; i.e. 10 cm ID. The water-cooled crucible
absorbs only a small amount of the electromagnetic energy; typically 80-85% of the induction power
is absorbed by the melt (Hellmann & Fischer, 2007). A facility schematic along with a photograph is
provided in Figure 2.3-20.
Figure 2.3-20: Schematic (left) and photograph (right) of the SICOPS test facility
Information on melt front progression into the concrete was obtained from embedded
thermocouples. Heat losses are calculated from flow-rate and temperature rise of the crucible coolant
water; the net heat input into the melt is calculated as the difference between total input power minus
crucible cooling losses.
In terms of test results, melt temperature measurements were found to be difficult. Optical
pyrometer data for a mixed metal-oxide melt with 21 wt% initial Fe content indicated that the melt
NEA/CSNI/R(2016)15
was subcooled by ~150 K (i.e. melt temperature was ~1 830°C versus 2015°C liquidus temperature
predicted by GEMINI2/NUCLEA09). However, surface observations indicated that melt convection
was quite active at this temperature. At this stage the oxidic melt contained 45 wt% concrete. The heat
flux into the concrete was ~100 kW/m2 and the concrete erosion velocity was ~1.4 mm/min. This
result confirms observations from earlier MCCI tests in which samples of zirconia-based refractory
material were placed inside a prototypic molten corium pool during ongoing interaction with concrete
(Hellmann & Fischer, 2007). In those tests, melt compositions covered the range of expected
conditions from vessel failure through several hours of ex-vessel interaction. None of the zirconia
samples showed any measurable dissolution at the surface. It was therefore concluded that the MCCI
pool was subcooled and saturated in zirconia which reduces the driving force to dissolve zirconia to
effectively zero.
Concerning thermochemistry aspects, the solubility of metal and oxide phases was below the
detection limit of SEM/EDX as well as chemical analysis. Iron oxidation in the nine mixed metal-
oxide tests was high (20-45%). These data are from tests using siliceous concrete and are for melt
compositions initially containing 10 to 26 wt% Fe; the duration of core-concrete interaction ranged
from 10 to 30 minutes. These observations are qualitatively in agreement with findings from the
VULCANO VBS U1 and U3 experiments (Journeau C. , et al., 2009) where a large fraction of metal
was oxidised.
Figure 2.3-21: Post-test debris configurations for oxidic melt (left) and a mixed oxide-metal melt
(right) tests
To compare oxide and mixed oxide-metal melt erosion behaviour, an average erosion depth was
evaluated based on post-test debris examinations for all experiments (Figure 2.3-21). These data are
shown in Figure 2.3-22 as a function of net input power available for concrete dissolution (Note that
the interface area was ~100 cm2). The average erosion rates for oxide and mixed metal-oxide melts
are the same within the error bands for the experiments at a given power level. A linear relationship
exists, which is consistent with a purely thermal concrete erosion process.
NEA/CSNI/R(2016)15
Figure 2.3-22: Concrete erosion velocity versus power level for oxide
and mixed oxide-metal SICPOS tests
Aside from general aspects of core-concrete interaction behaviour, another key element of the
SICOPS test programme was to experimentally confirm that refractory zirconia is effectively inert
when in contact with molten core debris over a wide range of ex-vessel conditions. To evaluate the
stability of ZrO2 in contact with corium, samples were embedded in the bottom and sides of concrete
crucibles made of EPR sacrificial (hematite) concrete (Figure 2.3-23). After these tests, samples were
extracted for analysis by SEM. These studies showed that zirconia resists attack by oxidic melt under
these MCCI conditions. Material compatibility with metal melt was demonstrated in separate
experiments. Figure 2.3-24 illustrates results for SICOPS Test #4 with two fully exposed ZrO2 bottom
samples inside EPR sacrificial concrete. This test showed that ZrO2 was neither dissolved nor etched
by the melt during the interaction. As noted earlier, these results confirm that the core-concrete
interaction establishes a subcooled melt pool that is saturated with respect to refractory components
ZrO2 and UO2.
ZrO2
samples
(behind 5
Figure 2.3-23: Concrete crucibles pre-test showing locations of embedded ZrO2 samples
NEA/CSNI/R(2016)15
Figure 2.3-24: Images of SICOPS test 4 with wo fully exposed ZrO2 bottom samples embedded
in EPR™ sacrificial concrete
2.3.8. VULCANO experiment series

The VULCANO core-concrete interaction programme has been underway at CEA since 2003. A total
of twelve reactor material experiments have been carried out in the facility with sustained heating
(Journeau C. , et al., 2009), (Journeau, et al., 2012). These experiments have focused on addressing
two unresolved issues related to long-term core concrete interaction behaviour; namely, radial-axial
power split, and the effect of high metal content on cavity erosion behaviour.
A schematic that illustrates principal features of the facility is shown in Figure 2.3-25. Key
elements consist of furnaces for melting the corium constituents and an instrumented concrete test
section. The oxide phase for a given experiment is produced in a rotating plasma furnace. The corium
constituents are loaded as powders into the furnace and held in place by centrifugal force. A plasma
arc is initiated between two graphite electrodes located along the axial centreline of the furnace, and
the powders are gradually melted by the heat produced from this arc. After melting is completed, the
furnace is tilted and the oxide flows down a trough and into the concrete test section. Induction
furnaces are used to produce molten metal constituents, and they are introduced into the test section in
the same manner.
Figure 2.3-25: Schematic of VULCANO test facility

The concrete test sections are cylindrical half-sections (180° sectors) with an ID of 30 cm and
depth of 25 cm. The test sections are heavily instrumented with more than 100 Type K-type
NEA/CSNI/R(2016)15
thermocouples that are primarily used to monitor cavity erosion. Several high temperature (Type C)
thermocouples are also installed to monitor the melt pool temperature. Four parallel induction coils
surround the test section to provide sustained heating that simulates decay heat (Figure 2.3-25). Free
(unconnected) coils are also installed beneath the test section to provide electromagnetic shielding for
the lower part of the melt where the metal phase eventually segregates. The concrete crucible is a half-
cylinder with the dividing wall made of refractory material (MgO in first tests and ZrO2 thereafter).
Part of the input power is lost by conduction through the refractory wall and is removed by coolant
water flow through the induction coil. This allows a heat balance to be performed and so the gross
input power can be adjusted to offset losses, thereby maintaining input power at the target level.
One of the primary parameters varied in the experiment series has been concrete type, with a total
of five different compositions tested (Table 2.3-5). Compositions F and G are reflective of silica and
limestone-rich plant structural concretes, respectively, while concrete E represents the sacrificial
hematite concrete used in the EPR reactor pit at Olkiluoto, Finland. Concrete C is a specific
formulation using calcined limestone (“clinker”) aggregate that is similar to Concrete G, but without
carbon dioxide. Mortar F is mortar with a composition similar to Concrete F but without the larger
aggregates (gravel).
Table 2.3-5: Concrete compositions for the VULCANO experiments
Designation wt% Constituent:
CaO CO2 SiO2 Al2O3 Fe2O3 MgO H2O

Concrete C 49.2 - 29.6 3.9 2.1 0.9 10.7
Concrete E 12.7 1.4 45.5 3.3 32.9 0.3 3.7
Concrete F 16 9 63 5 - - 3
Mortar F 18 7 58 5 1 0.5 7.5
Concrete G 42 25 26 2 - - 4
As shown in Table 2.3-6, six 2D core oxide experiments using four different concrete types have
been carried out thus far in the VULCANO facility. The first four experiments (i.e. VB-U4, U5, U6,
and U7) utilised traditional plant concretes, while the last two (VBES-U2, -U3) used specialty
concretes for separate effect testing. The tests used 28 to 55 kg of prototypic oxidic corium, yielding
collapsed pool depths of 16 to 25 cm. The pools were heated for several hours while cavity erosion
behaviour was observed.
Before the start of the VULCANO (Journeau C. , et al., 2009), (Journeau, et al., 2012) and NEA
MCCI CCI (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, &
Aeschlimann, 2010) experimental programmes, it was generally assumed that concrete ablation would
proceed at roughly the same rate in the radial and axial directions. However, both programmes have
shown that radial ablation is significantly more pronounced for siliceous concrete. This behaviour is
illustrated in Figure 2.3-26 which shows the final cavity profiles for tests VB-U7 and VBES-U2; these
tests utilised concretes E and C, respectively. The cavity shapes are quite similar. In tests with the F,
E and C concrete types, ablation was initially asymmetric (radial/axial ablation ratio between 2 and 3).
However, in the long-term the asymmetry decreased. On the other hand, tests with limestone concretes
(VB-U6 and CCI-2 ( (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2010) exhibit more isotropic ablation. The final radial and axial ablation
depths for all seven oxidic corium experiments are shown in Figure 2.3-27. Except for tests VB-U6
and VBES-U3, all tests experienced more radial ablation than axial. The fact that similar behaviour
has been observed in two experiment facilities (VULCANO and CCI) operating at two different scales
(50-1 000 kg), with different geometries and heating techniques (induction vs. direct electrical
NEA/CSNI/R(2016)15
heating) indicates that the trend is most likely not an experiment artefact, but rather a real
phenomenological affect.
Table 2.3-6: VULCANO 2D oxidic corium test matrix
Test VB-U4 VB-U5 VB-U6 VB-U7 VBES-U2 VBES-U3
Concrete Concrete F Concrete F Concrete G Concrete E Concrete C Mortar F
Initial mass 45 kg 28 kg 31 kg 54 kg 45 kg 29 kg
Initial ~2 200 K ~2 400 K ~2 400 K ~2 500 K ~2 500 K ~ 2 500 K
temperature
Average net 14 kW 12.5 kW 9 kW 22 kW 15 kW 9.4 kW
power
Heating 1 h 40 2 h 30 2h 2 h 40 3h 2 h 40
duration
Ablation anisotropic anisotropic ~isotropic anisotropic anisotropic Limited
Pattern ablation
An examination of the VULCANO database suggests that the superficial gas velocity during
core-concrete interaction is not the primary parameter controlling changes in ablation behaviour for
the two concrete types. For a given ablation rate, superficial velocity is only a function of the quantity
of gas generated per unit volume of ablated concrete (see Table 2.3-7). It is unlikely that there is a
sharp transition between concrete types C and G. Small scale simulant material ÉCLAIR experiments
(Journeau & Haquet, 2009) indicate that the transition between highly anisotropic natural convection
heat flux distributions to an isotropic distribution occurs at quite low superficial gas velocity3, and this
velocity is below that estimated for all VULCANO tests.
Aside from volumetric flow rate, there is a significant difference in gas mass flow rate 4 between
tests that yielded anisotropic ablation patterns and test VB-U6 with concrete G that yielded isotropic
ablation. The higher mass flux for concrete G is due to the differences in molar mass between carbon
dioxide and steam. Note that models used to describe convection induced by gas sparging (Bonnet,
2000), (Tourniaire, 2006) are currently based on gas volumetric flow rate (superficial velocity) and not
the volumetric mass flow rate. The data further indicate that the composition of molten concrete
(which affects the corium solidus-liquidus temperatures, viscosity, chemical diffusivity, and ability to
dissolve corium crusts (Carenini, Haquet, & Journeau, 2007) is not a major factor influencing ablation
behaviour. This is evidenced by the fact that tests VB-ESU2 and VB-U6 yielded very different erosion
patterns, yet the non-volatile constituents of the concretes used in these tests
(i.e. C and G; see Table 2.3-6) are very similar.
3. The threshold corresponds to Reynolds number (calculated with buble diameter and superficial gas velocity
as characteristics length and velocity) around unity.
4. The gas mass flux during core-concrete interaction is equal to the product of the volatile gas mass density
in the concrete and the ablation rate
NEA/CSNI/R(2016)15
Figure 2.3-26: Post-test cavity erosion profiles for the VB-U7 and VBES-U2 experiments
Figure 2.3-27: Radial-axial power split for oxidic corium VULCANO experiments
Table 2.3-7: Concentration of volatile species in VULCANO concretes

Test V V VB VB-
B-U5 B-U7 ES-U2 U6
Concrete Designation F E C G
Quantity of Volatile Gas (kmol/m3) 8 6 14 18
(H2O) (CO2)
Volatile mass fraction (wt%) 5 1 11 31
2
Post-test examinations indicate that silica gravel may remain in a solid state in the corium pool.
On the other hand, limestone (CaCO3) is destroyed by decarbonisation around 700ºC to form a fine
lime (CaO) powder that was not found in the solidified corium pool for test VB-U6. Moreover,
chemical analysis of samples collected from tests VB-U4 and VB-U5 indicate that the pool
composition was depleted in silica. This finding indicates that the larger aggregate had not mixed with
NEA/CSNI/R(2016)15
the corium while the finer mortar did. Another finding from the post-test analyses is the presence of
corium-rich and corium-lean zones in the pool, indicating that solutal convection was occurring
leading to large local variations in composition (see Figure 2.3.29) and thus in density.
Zone rich in corium but lean

in concrete elements.
Zone rich in concrete but

lean in corium elements
corium and lean in concrete
Figure 2.3-28: Composite SEM micrograph from test VB-ESU2 (lower part of the bulk melt)
In addition to purely oxidic tests, four metal – oxide (VBS series) experiments have been carried out
in the VULCANO programme; one with limestone concrete, and three with siliceous (see Table 2.3-8).
Experiment VBS-U2 was repeated as VBS-U3 due to a poor induction coupling in the first experiment.
The tests were conducted by first pouring oxide into the cavity, and then 2 to 3 litres of molten stainless
steel are added from several separate steel furnaces. At this point, induction heating is applied to simulate
decay heat in the oxidic phase.
One of the most surprising results from the VBS-U1 experiment with limestone concrete was the
fact 90% of the initial steel mass of 15 kg was apparently oxidised (i.e. only 1.5 kg steel was recovered
following the test). Pre-test calculations predicted an iron oxidation rate of 4% per hour. In order to
oxidise 90% of the steel, it would have been necessary for all concrete decomposition gases from
ablated concrete to react with the steel, as well as all the gases from the balance of the crucible that
reached a temperature of 100°C. The remaining metal was completely depleted in Cr and significantly
enriched in Ni (viz. depleted in Fe). Zones in the oxide phase were enriched in concrete oxides (CaO,
SiO2) as well as steel oxides. Steel oxidation mechanisms are still under investigation at CEA.
Table 2.3-8: VULCANO 2D metal-oxide corium test matrix

Test VBS-U1 VBS-U2 VBS-U3 VBS-U4
Concrete Concrete G Concrete F Concrete F Concrete F
Oxide Load1 Corium 2 Corium 2 Corium 2 Corium 1
Initial oxide mass 35 kg 18 kg 36 kg 35 kg
Initial stainless steel mass 15 kg 17 kg 15 kg 24 kg
Heating duration 4h Poor coupling 4h 2 h 25
1
Corium 1 consists of 45/19/20/15 wt% UO2/ZrO2/SiO2/Fe2O3, while Corium 2 consists of 69/17/6/7/1 wt%
UO2/ZrO2/SiO2/Fe2O3/CaO.
NEA/CSNI/R(2016)15
In the tests with siliceous concrete, steel oxidation also occurred but to a lesser extent in
comparison to the limestone concrete case. As for VBS-U1, the remaining steel was depleted in Cr and
enriched in Ni. The most surprising result from the three tests with siliceous concrete is that spatial
segregation occurred, but it was not totally driven by gravity. In particular, steel layers were found on
both horizontal as well as vertical surfaces (Figure 2.3-29). Solidified metal drops were also found in
the oxide pool; sizes ranged from 0.1 mm up to several centimetres. Oxidic melt and un-melted silica
aggregate were also trapped in some parts of the metal layer.
Vertical metal
layer
Horizontal layer of metal
Figure 2.3-29: Photograph of large metallic structure recovered from test VBS-U4
For test VBS-U3 the metal layer did not cover the entire horizontal and vertical concrete surfaces,
but only an angular sector of slightly less than 90°. Concrete temperatures showed that axial and radial
ablation was more pronounced in the areas where metal was found. This is consistent with other
oxide-metal simulant tests (e.g. BETA (Alsmeyer, 1987), (Alsmeyer, et al., 1995) that have shown
enhanced heat transfer at the metal-concrete interface relative to the oxide-concrete interface. Further
note that tests VBS-U3 and VBS-U4 were carried out with different corium compositions that had
significant density differences due to variations in initial concrete content (see Table 2.3-8). This
indicates that the unique metal phase topology is not strongly linked to the oxide phase density. Due to
ongoing solutal convection, plumes with various concrete fractions – and thus different densities – are
present below the metallic material. For instance, analysis of a sample from test VBS-U3 indicates
estimated densities (evaluated at the pool temperature) of 6.2 kg/l for the metal, 6.6 kg/l for the
concrete-lean oxide, and 4.6 kg/l for the concrete-rich plume. This analysis indicates that
phenomenological behaviour near the core-concrete interface is more complex that that observed in
simulant material stratification experiments where only two liquid compositions are present. During
core-concrete interaction, miscible but yet unmixed oxides are present and coexist with immiscible
molten metal.
The four oxide-metal tests performed in the VULCANO facility are the first significant scale
experiments with a prototypic oxide-metal composition with decay heat simulation in the oxide phase.
In other experiments such as COTELS ( (Nagasaka, et al., 1999), (Nagasaka, et al., 1999), (Zhdanov,
et al., 1999), (Farmer, Kilsdonk, & Aeschlimann, 2009), the heat was injected into the metal phase due
to characteristics of the induction system used in those tests. The VULCANO VBS tests have
identified unexpected phenomenology regarding metal oxidation during core-concrete interaction, as
well as the spatial distribution of the metal due to segregation. These phenomena need to be better
understood in order to determine how to scale the results to plant conditions.
NEA/CSNI/R(2016)15
2.3.9. NEA MCCI Project CCI experiments

The NEA MCCI Project was carried out at ANL from 2002 through 2010 ( (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010). As part of
this work, a total of six large-scale reactor material core-concrete interaction experiments were carried
out with sustained heating. One of the primary objectives of these tests was to provide data on cavity
erosion behaviour and melt temperature response in order to support code verification and validation
activities. To augment the amount of information obtained from these tests, five of the six experiments
were flooded from above after a pre-defined concrete ablation depth was reached in order to provide
information on the nature and extent of debris coolability processes. This section summarises
information gained from the tests related to dry cavity erosion for siliceous and limestone-common
sand concrete types; the coolability results are summarised in Section 2.4. Specifications for all six
tests in the experiment series are shown in Table 2.3-9, tests CCI-1 through CCI-5 are relevant to the
current discussion. Key facility features are illustrated in Figure 2.3-30, while details of the test section
are shown in Figure 2.3-31.
The facility consisted of a test apparatus, a power supply for direct electrical heating (DEH) of
the corium, a water supply system, two steam condensation (quench) tanks, a ventilation system to
filter and exhaust the reaction product gases, and a data acquisition system. The apparatus for
containment of the core material consisted of a test section that was ~ 3.4 m tall. The internal
dimensions of the bottom test crucible varied in the test series, but the nominal configuration was a
square internal cross-section measuring 50 x 50 cm (Figure 2.3-31). The lower section was basically a
notch-type configuration in which the basemat and two sidewalls experienced concrete erosion, while
the other two opposing walls were lined with tungsten electrodes that were used to provide DEH to the
melt to simulate decay heat. The two electrode walls were made from refractory MgO; the electrodes
were backed up with a layer of crushed UO2 pellets to provide additional protection against attack by
the corium. Melt generation in all experiments was through an exothermic chemical reaction yielding
the target initial melt mass over a timescale of ~ 30 seconds. After the reaction, DEH was supplied at a
level that provided a heat flux in the range of 150 (for CCI-2 through -5) to 200 (for CCI-1) kW/m2 to
all melt surfaces in contact with concrete (as well as upper atmosphere) at the start of the experiment.
This power level range is consistent with conditions at ~ 2 hours following scram in a PWR. The
concrete basemat and sidewalls of the apparatus were instrumented with Type K thermocouples to
measure ablation front progression and Type C units (in tungsten thermowells) to measure melt
temperature.
Figure 2.3-30: Key features of the CCI test facility
NEA/CSNI/R(2016)15
Initial melt compositions, temperatures, and corresponding power levels for the various tests are
summarised in Table 2.3-9. All tests were carried out with fully oxidised PWR core melt with the
exception of CCI-4. This test utilised a BWR composition in which 78% of the cladding was initially
oxidised; the melt also contained 7.7 wt% stainless steel. This steel fraction is less than half that used
in the VULCANO metal-oxide tests (Journeau C. , et al., 2009), (Journeau, et al., 2012)
(see Section 2.3.8), and so this test can be considered as an intermediate data point when viewed in
context with the VULCANO results.
19.75" [50.2cm]
EXISTING MACE 50 cm X B
50 cm MgO TEST SECTION
NORTH
LIFTING LUGS
TUNGSTEN ELECTRODES
MgO INSULATION (9.5 mm OD)
ALUMINUM TRANSITION CRUSHED PELLETS

PLATE (1" THICK)
TUNGSTEN BACKUP PLATES
INITIAL POWDER HEIGHT

BASEMAT
50 cm (50 cm X 50 cm)
CONCRETE SIDEWALLS A A
TUNGSTEN ELECTRODES 50 cm
68.50" [174.0cm]
ELECTRODE SUPPORT ROD
TYPE K TC'S (9.5 mm OD TUNGSTEN)
47.0"
TYPE C TC IN TUNGSTEN
THERMOWELL
TYPE K TC'S CONCRETE SIDEWALLS (2)

BASEMAT 56.2 cm THICK, MAXIMUM
ABLATIION DEPTH = 35 cm
55 cm
MgO REFRACTORY SIDEWALLS
1.00" 22.13" [56.2cm]
DRAWING: CCI1 BOTTOM SECTION DRAWING: CCI1 BOTTOM SECTION

(WEST VIEW) CONCEPTUAL DRAWING (TOP VIEW)
ELECTRODE CLAMP DRAWING NO.: MCCI197
DRAWING NO.: MCCI297
DRAWN BY: D. KILSDONK 2-4746
8.0" TOP VIEW DRAWN BY: D. KILSDONK 2-4746
DATE: 4/25/03
IBEAM FILE: CCI1_BSWVC.DWG(AC86) DATE: 1/12/04
B
FILE: CCI1_BSTV.DWG(AC93)
VIEW FORM WEST 45.25" [114.9cm]
NORTH AT CENTER LINE SOUTH
Figure 2.3-31: Side (left) and top (right) views of the CCI test section
Table 2.3-9: Specifications for CCI tests
Parameter Specifications for Test:
CCI-1 CCI-2 CCI-3 CCI-4 CCI-5 CCI-6
Corium 100% oxidised 100% oxidised 100% oxidised 78% oxidised 100% oxidised 100% oxidised
PWR + 8 wt% PWR + 8 wt% PWR + 15 wt% BWR with 7.7 PWR + 15 wt% PWR + 6 wt%
SIL LCS SIL wt% SS and SIL SIL
10 wt% LCSb
Concrete typea SIL (US-type) LCS SIL (EU-type) LCS SIL (EU-type) SIL (EU-type)
Basemat cross-section 50 cm x 50 cm 50 cm x 50 cm 50 cm x 50 cm 50 cm x 40 cm 50 cm x 79 cm 70 cm x 70 cm
Initial melt mass 400 kg (25 cm) 400 kg (25 cm) 375 kg (25 cm) 300 kg (25 cm) 590 kg (25 cm) 900 kg (28 cm)
(depth)
Test section sidewall Electrode walls: Electrode walls: Electrode walls: Electrode walls: Electrode walls: Electrode walls:
construction Inert Inert Inert Inert Inert Inert
Other walls: Other walls: Other walls: Other walls: Wall # 1: Other walls:
concrete concrete concrete concrete concrete concrete
Wall # 2: Inert
Lateral/Axial ablation 35/35 cm 35/35 cm 35/35 cm 45/42.5 cm 40/42.5 cm 24/32.5 cm
limit
Initial melt temperature 1 950 ºC 1 880 ºC 1 950 ºC 1 850 ºC 1 950 ºC 2 300 ºC
System pressure Atmospheric
Melt formation tech. Chem. reaction (~30 s)
Melt heating technique Direct Electrial Heating (DHEH)
Power supply Constant @ 150 Constant @ 120 Constant @ 120 Constant @ 95 Constant @ 145 Constant @ 210
operation prior to kW kW kW kW kW kW
water addition
NEA/CSNI/R(2016)15
Criteria for water 1) 5.5 hours of 1) 5.5 hours of 1) 5.5 hours of 1) 7.0 hours of 1) 6.0 hours of 2.5 cm concrete
addition operation 2) operation 2) operation 2) operation 2) operation 2) ablation
ablation → 5 cm ablation → 5 cm ablation → 5 cm ablation → 5 cm ablation → 5 cm
of limit of limit of limit of limit of limit
Timing of water 68 minutes 301 minutes 108 minutes 380 minutes N/A 0.6 minutes
addition (actual)
Parameter Specifications for Test:

CCI-1 CCI-2 CCI-3 CCI-4 CCI-5 CCI-6
Inlet water flowrate 2 lps/20 ºC
and temperature
Water depth over melt 50 ± 5 cm
Power supply Constant voltage
operation after water
addition
Test termination 1) Tmelt < Tcon,sol, 2) ablation arrested, or 3) ablation →
criteria limit
Operational Summary Pronounced Symmetrical Pronounced Symmetrical Pronounced Limited ablation
lateral erosion in cavity erosion; lateral erosion; cavity erosion; lateral erosion; due to effective
one sidewall; successful cavity successful cavity mantle crust cavity not debris cooling
successful cavity flooding flooding prevented melt- flooded due to
flooding water contact off-gas system
when cavity was plugging
flooded
a
SIL denotes siliceous concrete, LCS denotes Limestone/Common Sand concrete.
b
After erosion of concrete/metal inserts and at start of basemat ablation.
Comparisons of key test results for the CCI tests are provided in Figure 2.3-32 and Figure 2.3-33
that provide the average melt temperature and peak lateral and axial concrete ablation rates for each of
the tests. Initial melt temperatures were in the range of 1 880-1 950ºC; differences were due to
uncertainty/variability in the reaction temperatures for the different chemical mixtures used to generate
the melt pools. During dry cavity operations, all tests exhibited the overall trend of decreasing melt
temperature as ablation progressed, which was due to a heat sink effect as relatively cool concrete slag
was introduced into the melt, as well as the increasing heat transfer surface area as the melts expanded
into the concrete crucibles. The decline in melt temperature may further reflect the evolution of the
pool boundary freezing temperature that decreased as additional concrete was eroded into the melt
during the tests (Roche, Steidl, Leibowitz, Fink, & Sehgal, 1993).
2100
Test CCI-1 (SIL Concrete)
2000 Test CCI-2 (LCS Concrete)

CCI-1 Water Addition Test CCI-3 (SIL Concrete)
1900 Test CCI-4 (LCS Concrete)

Average melt temperature (˚C)

CCI-3 Water Addition
1800
CCI-2 Water Addition
1700
1600
1500
1400
1300
1200
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425
Elapsed time (minutes)
Figure 2.3-32: Average melt temperatures for the CCI tests
NEA/CSNI/R(2016)15
35 35
Test CCI-2 (LCS Concrete)
Test CCI-4 (LCS Concrete)
30 30 Test CCI-1 (SIL Concrete)
Lateral ablation depth (cm)
25 25
Axial ablation depth (cm)

20 20
15 15
Test CCI-2 (LCS Concrete): North Wall

10 10
Test CCI-4 (LCS Concrete): North Wall
Test CCI-1 (SIL Concrete): North Wall
5 Test CCI-1 (SIL Concrete): South Wall 5

Test CCI-3 (SIL Concrete): South Wall
Test CCI-5 (SIL Concrete): North Wall
0
0
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425
Figure 2.3-33: Lateral (left) and axial (right) cavity erosion rates for CCI tests
One objective of the test series was to investigate the effect of unoxidised Zr cladding on the
thermalhydraulics of the core-concrete interaction. As noted earlier, CCI-4 was initiated with 22% of
the Zr initially present in an unoxidised state. As is evident from Figure 2.3-32, the effect of the
oxidation reaction between Zr and sparging concrete decomposition gases (CO2, H2O) was to cause
an exothermic transient in which the melt temperature increased by ~ 100 ºC during the first
20 minutes of the test. This same type of transient was observed in previous metal tests conducted in
SNL (Copus, Blose, Brockmann, Gomez, & Lucero, 1990), (Copus, Blose, Brockmann, Gopmex, &
Lucero, Core-Concrete Interactions using Molten Steel with Zirconium on a basaltic Basemat: The
SURC-4 Experiment, 1989) and at KIT in the BETA facility ( (Alsmeyer, 1987), (Alsmeyer, et al.,
1995). In addition, the effect of Zr oxidation was also investigated as part of the ACE/MCCI core
oxide test series (Thompson D. H., Fink, Armstrong, Spencer, & Sehgal, 1992). Unfortunately, tests
conducted in those experiments with partially oxidised melt and LCS concrete were of very short
duration (i.e. several minutes), and so the long term effects of the oxidation reactions on thermal-
hydraulic behaviour were not clear. However, CCI-4 ran for several hours past the point at which all
the cladding had oxidised. Moreover, CCI-2 can be considered a counterpart experiment insofar as
cladding oxidation is concerned. Comparison of the results indicates that cladding oxidation reactions
cause an early exothermic temperature transient in the melt, and after the reaction is complete, the
temperature drops to that consistent with fully oxidised melt conditions.
Additional examination of Figure 2.3-32 indicates that CCI-1 exhibited slightly different melt
temperature behaviour compared to the other fully oxidised tests. In this test, the melt temperature was
relatively constant over the first ~40 minutes of the interaction. One possible contributor to this trend
was the fact that this test was run at a 25% higher power level in comparison to the other CCI tests.
However, the lack of a temperature decline may have also been caused by crust formation at the core-
concrete interfaces that acted to insulate the melt. Relatively low heat transfer rates to the concrete
boundaries were evidenced by the low ablation rates exhibited over the first 40 minutes. Note that this
type of behaviour is consistent with other transient core oxide tests carried out at Sandia (Gronager,
Suo-Anttila, & Brockmann, 1986), wherein sub-stoichiometric melt [(U,Zr)O2-x] was dropped into
basalt and limestone-common sand concrete test sections and allowed to cool with no further heating.
In these tests, no concrete ablation occurred and the conclusion was drawn that crusts acted to
thermally protect the concrete. However, in the current tests the melts were continuously heated. Thus,
once the surface crusts failed in CCI-1, ablation proceeded vigorously and the melt temperature fell
rapidly in comparison to the other tests. This initial stable crust behaviour may have been linked to the
exceptionally low gas content for this concrete type in comparison to others used in the test series
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006). For instance, gas sparging at the core-concrete
interface may exert mechanical force(s) on the crust thereby freeing the material and allowing ablation
to proceed. If this interpretation is correct, then the reduced gas production rate may have allowed the
NEA/CSNI/R(2016)15
crust to remain stable over an extended period of time in CCI-1, which in turn caused the melt
temperature to increase.
Aside from CCI-1, other CCI tests also provided evidence of early crust formation that influenced
ablation behaviour (Figure 2.3-33). For CCI-2, both axial and lateral ablation rates were quite low and
the melt temperature relatively constant until ~ 30 minutes, after which time a period of rapid erosion
occurred. However, unlike CCI-1, these erosion bursts were not sustained. Rather, after ~ 5 cm of
ablation, both the axial and lateral ablation rates slowed significantly and approached quasi-steady
states. The reduced crust stability period for CCI-2 is consistent with the idea that gas sparging can
disrupt surface crusts, since the gas content of the CCI-2 concrete was significantly greater in
comparison to CCI-1.
Aside from initial crusting effects, examination of Figure 2.3-33 indicates that the long-term
ablation process is influenced by concrete type. Estimates of average lateral and axial ablation rates for
the five tests are provided in Table 2.3-10; the Table also includes estimates of the corresponding
concrete heat fluxes that were formulated based on the average erosion rates (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2010). As shown in the table, long-term lateral and axial ablation rates for
Tests CCI-2 and CC-4, both of which were conducted with LCS concrete, were about the same. For
CCI-2, the concrete erosion rate averaged 4 cm/hr over several hours of interaction before gradually
decreasing; the corresponding surface heat flux was ~ 60 kW/m2. For CCI-4, the fluxes were slightly
lower (i.e. ~ 40 kW/m2), but this is due to a surface scaling effect as the initial cavity size and
therefore input power level were reduced to expand the test duration. Thus, the lateral/axial heat flux
ratios for these LCS tests are ~1.
The relatively uniform power splits for CCI-2 and CCI-4 can be contrasted with the results of the
three tests conducted with siliceous concrete. For test CCI-1, the ablation was highly non-uniform,
with most of the ablation concentrated in the north sidewall of the test apparatus. As described above,
the results for this experiment were felt to be strongly influenced by transient crust effects, and so a
power split estimate was not formulated for this test. However, tests CCI-3 and CCI-5 seemed to
exhibit repeatable, albeit non-isotropic, ablation behaviour. For CCI-3, fairly symmetrical ablation
occurred insofar as the progression of the ablation fronts into the two opposing sidewalls of the
apparatus is concerned. However, unlike Test CCI-2, the lateral ablation was highly pronounced in
comparison to axial for this particular test. A similar trend was noted for CCI-5 that was conducted
with a single siliceous concrete sidewall. As shown in Table 2.3-10, lateral ablation averaged 10 cm/hr
during the late phases of the CCI-3 and CCI-5 experiments, while the axial ablation rate was limited to
2.1 to 2.5 cm/hr over the same time frame for the two tests. On this basis, the lateral/axial surface heat
flux ratios for tests CCI-3 and CCI-5 were estimated as ~ 4 and ~ 4.7, respectively. These values are
significantly higher than the near-unity ratios deduced for tests CCI-2 and CCI-4 with LCS concrete.
Thus, the data suggests that there is an effect of concrete type on the spatial heat flux distribution at
the core-concrete interface during dry core-concrete interaction. This same overall trend was observed
and documented in counterpart experiments performed in the VULCANO test facility (Journeau, et al.,
2012), (Langrock & Hellmann, 2010) (see Section 2.3.8).
NEA/CSNI/R(2016)15
Table 2.3-10: Lateral/axial ablation rate and power split estimates for CCI tests
Lateral ablation Axial ablation Lateral-axial heat
flux ratio
Test Concrete Type
Ablation Rate Heat Flux Ablation Rate Heat Flux
(cm/hr) (kW/m2) (cm/hr) (kW/m2)
CCI-1 SIL (US) N: 39.1 395 26.1b 265 –a
S: 8.4 86
CCI-3 SIL (EU) 10.0 97 2.5 25 4.0
CCI-5 SIL (EU) 9.8 95 2.1 21 4.7
CCI-2 LCS 4.0 58 4.0 59 1.0
CCI-4 LCS 2.7 39 2.8 41 1.0

a
Heat flux ratio not evaluated for this test due to large asymmetry in lateral cavity erosion.
b
Ablation burst; rate appeared to slow significantly after this time interval; see Figure 2.3-33.
Another finding from the CCI tests was that the nature of the core-concrete interface was
noticeably different for limestone tests in comparison to siliceous; see Figure 2.3-34. The ablation
front for the siliceous concrete tests consists of a region where the core oxide had locally displaced the
cement that bonded the aggregate. Conversely, for Test CCI-2 the front consisted of a powdery
interface in which the core and concrete oxides were clearly separated. The interface characteristics
may have influenced the heat transfer at the interface, thereby resulting in different ablation behaviour
for the two concrete types. Differences in interfacial conditions for these two concretes were also
found in the VULCANO tests (Journeau, et al., 2012), (Langrock & Hellmann, 2010).
Finally, one of the objectives for test CCI-5 was to examine the influence of melt pool aspect
ratio on the radial/axial power split under dry cavity conditions. In this experiment the test section
aspect ratio (i.e. test section width/melt depth) was increased to the greatest extent possible to more
accurately mock up prototypic conditions. The approach was to modify the test section design to
include a single concrete sidewall that would undergo ablation, whereas the opposing sidewall was
made from insulating material (MgO lined with UO2). With this approach the aspect ratio was
increased from a value of ~ 1 for CCI-3 to ~ 3.2 for CCI-5. The relatively close agreement in long-
term ablation behaviour for tests CCI-3 and CCI-5 (Figure 2.3-33) indicates that aspect ratio has little
influence on ablation characteristics. This observation lends additional credibility to observation of
skewed power splits for siliceous concrete observed in both the CCI and VULCANO tests. Finally,
data from the Chernobyl accident, involving core melt interaction with silica-rich (basalt) concrete
over several days, also support this observation (see Section 2.5). Aside from overall cavity erosion
behaviour, video data indicated that a crust was present over the melt surface during a large fraction of
dry cavity ablation phase for all tests. The crust contained vent openings that allowed eruptions to
occur as the tests progressed. The frequency and intensity of the eruptions was directly correlated to
concrete gas content. In terms of the chemical analyses conducted as part of the test series, samples
were collected to: i) characterise the lateral and axial composition variations of the solidified debris,
and ii) characterise the composition of corium regions that played key roles in the coolability aspects
of the tests (e.g. porous crust zones formed at the melt/water interface, and the material erupted after
cavity flooding in CCI-2). Analysis of samples taken to characterise the lateral composition variation
indicate that for most tests, the corium in the central region of the debris had a higher concentration of
core oxides in comparison to samples collected near the two ablating concrete sidewalls. Conversely,
samples taken to characterise the axial composition variation indicated the general trend of slightly
increasing core oxide concentration as the concrete surface is approached. For all three tests conducted
with siliceous concrete, two zones were present: a heavy monolithic oxide phase immediately over the
NEA/CSNI/R(2016)15
basemat that was enriched in core oxides, with a second overlying porous, light oxide phase that was
enriched in concrete oxides (see Figure 2.3-34).
Figure 2.3-34: Axial debris morphology for siliceous concrete tests (a) CCI-1, (b) CCI-3,
and (c) CCI-5, and LCS Test (d) CCI-2 (All tests were flooded except CCI-5)
2.4. Flooded cavity MCCI experiments
The previous sections have outlined phenomenology and supporting experiments related to core-
concrete interaction under dry cavity conditions. This section summarises results of experimental
programmes that have focused on studying the nature and extent of core debris cooling that occurs
when the core debris is flooded from above; experimental programmes that have addressed this issue
are summarised in Table 2.2-2. Alternative cooling strategies and associated experimental
programmes for stabilising core debris that do not rely on top flooding (i.e. core-catchers) are briefly
summarised in Section 2.2.
2.4.1. SNL flooded cavity experiments
Although the SNL R&D programme on core-concrete interaction primarily focused on dry cavity
erosion and fission product release, three melt coolability experiments were also conducted with
sustained heating and concurrent concrete erosion. These experiments include the SWISS tests (Blose,
Gronager, Suo-Anttila, & Brockman, 1987), conducted with stainless steel melts interacting with
limestone/common sand concrete, and the WETCOR test (Blose, et al., 1993) conducted with an oxide
simulant that also interacted with LCS concrete.
The SWISS apparatus consisted of an interaction crucible, melt generator, and a water delivery
system. The system also contained instrumentation to measure aerosol and non-condensable gas
release both before and after water was added to the test section. The MgO interaction crucible was
cylindrical with an ID of 21.6 cm. The concrete basemat was located at the bottom of the crucible. The
melt was produced by induction heating of a 46 kg 304 stainless steel charge mass in a melt generator
located over the crucible. After the melt was formed, it was dropped into the crucible to initiate the
test. Fission product decay heat was simulated with induction heating of the metal charge. Water was
added over the melt at a preselected time, and the depth over the melt was maintained at ~ 30 cm. The
heat flux to the overlying water was measured by the mass flow rate of the coolant and the differential
temperature rise across the water inlet and outlet plenums. An aerosol detection system was also
provided to quantify the extent of fission product scrubbing once water was added over the melt.
NEA/CSNI/R(2016)15
For the SWISS-1 test, water was added very late in the sequence due to test occurrences. Thus,
this was basically a dry MCCI experiment with minimal coolability data obtained. For SWISS-2, a
total of 44.2 kg of melt was delivered to the test section. The initial temperature was between
1 530 and 1 686ºC at the time of delivery. Net input power was maintained at 60 to 70 kW throughout
the balance of the test. The corresponding power density was 1500-1700 W/kg of fuel, which is
~ 5 times the anticipated power level of ~ 300 W/kg for a typical PWR at 2 hours into the accident
sequence. Water was added to the test section at 1.6 minutes past melt delivery. The peak melt/water
heat flux was measured as 1.4 MW/m2. After steady state conditions were established, the basemat
erosion rate was ~ 27 cm/hour, while the melt-water heat flux was ~ 0.8 MW/m2. Comparison of the
SWISS-1 and SWISS-2 ablation results indicates that, for these experiments, water addition had little
influence on the basemat erosion rate. However, post-test examinations indicated that a bridge crust
ranging from 5.1 to 6.4 cm thick had formed. This crust spanned the width of the test section and was
anchored to the test section sidewalls. As a result, an intervening gap developed between the melt and
crust as the MCCI progressed downward. This non-prototypic gap may have adversely affected the
quench progression in this experiment, as indicated by the simulant coolability experiments of
Theofanous et. al (Lomperski, Farmer, & Basu, 2006), as well as the MACE reactor material
experiments (Farmer, Kilsdonk, & Aeschlimann, 2009). As an important aside, the aerosol data
indicated an effective decontamination factor (DF) of ~ 10 for the 30 cm deep saturated water pool
maintained over the debris during the test.
The WETCOR test (Blose, et al., 1993) utilised an apparatus similar to that used for the SWISS
tests. The MgO interaction crucible was slightly larger, with an ID of 33.0 cm. The basemat was also
made from LCS concrete. In contrast to the SWISS tests, the melt was generated by in-situ heating of
a 34.1 kg oxide charge (76.8/16.9/4.0/0.9/0.5 wt% Al2O3/CaO/SiO2/Fe2O3/MgO). The sidewall
heating elements consisted of tungsten susceptor rings which were inductively heated. This approach
was intended to eliminate crust anchoring to the test section sidewalls after water addition, as occurred
during the SWISS tests, by maintaining the sidewalls above the freezing temperature of the melt. After
the melt was generated and concrete ablation established, the cavity was flooded, thus initiating the
experiment. Fission product decay heat after water addition was simulated by maintaining input power
to the sidewall heating elements.
Key results for the WETCOR test are summarised as follows. The sidewall heating technique
resulted in a non-uniform ablation front at the time of water addition; i.e. the ablation depth had reached
~ 4.3 cm at the centre of the test section, 0 cm at the mid-radius, and ~ 2.6 cm along the outer periphery
near the tungsten susceptor coils. The cavity was flooded at 529 minutes into the test. The net input
power to the melt after water addition was estimated as 28 kW; the corresponding specific power density
was ~ 0.61 W/cm3. Addition of water resulted in an initial, intense bulk cooling phase which lasted
1-2 minutes. The thermocouple data suggests that the melt temperature was reduced by ~100ºC during
this time. Dispersion of melt droplets into the overlying coolant was visible during bulk cooling. After
the initial intense heat transfer period, a stable crust formed thereby reducing the heat transfer to the
overlying coolant. The melt/water heat flux was estimated as 0.52 ± 0.13 MW/m2. After crust formation,
the thermocouple data indicated that the temperature stabilised in a range of 1480-1580ºC. The sidewall
of the test section failed 26 minutes after water was added, leading to a pressurised melt run-out from the
test section. Post-test disassembly revealed the presence of a 3 to 5 cm wide gap between the top crust
and the underlying solidified debris. As for the SWISS tests, this finding indicates that the crust had
anchored to the test section sidewall after water was added, which most likely influenced the cooling
behaviour. The WETCOR aerosol data indicated an effective DF in the range of 3 to 15 for the ~30 cm
deep saturated water pool maintained over the debris during the test.
2.4.2. COMET-L3 experiment
The COMET-L3 experiment (Miassoedov, Alsmeyer, Cron, & Foit, 2010) was carried out at KIT to
investigate 2D concrete erosion and debris coolability in a cylindrical crucible made from siliceous
NEA/CSNI/R(2016)15
concrete. The facility utilised base hardware and technology that was developed as part of the
preceding BETA experimental programs ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995). The concrete
crucible was located at the base of the test rig and was situated over an induction coil that delivered
heat to the melt metal phase to simulate decay heat. The cylindrical sidewall above the coolant device
(Figure 2.4-1) was formed from a refractory MgO ceramic ring that prevented thermal attack of the
upper test rig. The core melt simulant was generated externally and then poured through into the test
crucible at an initial temperature of ~1 700°C. The initial melt composition consisted of 459 kg metal
(90 wt% Fe, 10 wt% Ni) and 159 kg oxide (56 wt% Al2O3, 44 wt% Ca). The material was poured into
the concrete cavity through an opening in the lid of the apparatus. This initial charge thus yielded a
25 cm deep metal layer overlaid by 20 cm of oxide in the 60 cm ID concrete crucible. The crucible
was heavily instrumented with Type K thermocouples to detect ablation front location; melt surface
temperature was measured with an infrared camera.
Figure 2.4-1: Dimensions of the COMET-L3 test section
In terms of the test operating procedure, after the melt was added to the crucible the debris was to
be heated at a target power level of 200 kW. The cavity was to be flooded after concrete ablation had
proceeded 9 cm either radially or axially. At the end of the dry cavity erosion phase, the melt was to
be flooded by water from the top at a rate of 375 l/s 5.
In terms of the test results, in the initial phase of the test (less than 100 s), the superheated melt
underwent an intense interaction with the concrete, leading to a rapid decrease in melt temperature
(Figure 2.4-2). During this phase the steel erosion rates in the radial and axial directions were similar
(∼0.15 mm/s); see Figure 2.4-3. The erosion rate by the overlying oxide was small in comparison (due
in part to the lack of internal heating) and stopped after < 1 cm erosion. After the initial transient, the
interaction rate slowed and by ∼800 seconds the axial erosion rate fell to ~ 0.07mm/s; the lateral
erosion rate was ~ 1/3 this value. These rates are consistent with those measured in the COMET-L1
(Alsmeyer, et al., 2005) and COMET-L2 (Miassoedov, Alsmeyer, Cron, & Foit, 2010) experiments, as
5. In terms of difference with reactor case, as the heat is delivered only in the metal phase a constant power
remains injected in the melt after flooding because the metal ejection is very unlikely due to the large
density difference between metal layer and simulant top oxide layer.
NEA/CSNI/R(2016)15
well as the former BETA experiments ( (Alsmeyer, 1987), (Alsmeyer, et al., 1995) at low power
density. During this period burnable gases were released as a result of chemical reaction of steam and
CO2 with iron. The dominant gas was hydrogen.
Surface flooding of the oxide melt was initiated at 800 seconds. The upper melt surface was
devoid of crust and moderately agitated by sparging decomposition gases at the time water was added.
Flooding did not lead to strong melt/water interaction or penetration of water into the melt. The melt
surface was almost completely covered by crust after 60 seconds, and the surface was quenched by
140 seconds. The crust eventually anchored to the cavity sidewalls and separated from the melt below.
Concrete erosion continued, although at a reduced rate of ~0.04mm/s. The melt eventually reached the
maximum erosion limit of the crucible, at which time input power was terminated.
Post-test analysis of the solidified melt showed that a large void had developed under the surface
crust (Figure 2.4-4). The crust was bulged ~20 cm in the central region, which may have been due to
pressure from below due to build-up of concrete decomposition gases. Water ingression into the bulk
of the oxide and underlying metal was prevented by the suspended bridge crust and no melt was
ejected trough this crust. No efficient fragmentation mechanism was identified that could breach the
crust and contribute to cooling. An energy balance indicated that only about 20% of the input power
was transferred through the upper melt surface, intervening gap, and overlying bridge crust into the
water. This confirmed that the influence of surface flooding is indeed small in this experiment and
unable to stop the downward erosion. However, formation of anchored crusts is not expected at plant
scale (Lomperski & Farmer, Corium Crust Strength Measurements, 2009), and results of other
experimental programmes (Farmer, Kilsdonk, & Aeschlimann, 2009), (Lomperski, Farmer, & Basu,
2006) have indicated that gap formation can effectively terminate any debris cooling mechanisms that
could lead to long-term coolability.
Figure 2.4-2: Measured melt surface temperatures for COMET-L3
NEA/CSNI/R(2016)15
Figure 2.4-3: Axial (left) and radial (right) concrete erosion data for COMET-L3
Figure 2.4-4: COMET-L3 post-test debris configuration

2.4.3. ECOKATS-2 experiment
The ECOSTAR Project was carried out in the European Union to investigate sequences that play a key
role during the ex-vessel phase of a postulated core-melt accident (Alsmeyer, et al., 2005). As part of
this work, the ECOKATS-2 demonstration experiment (Alsmeyer, et al., 2005) was carried out to
provide melt spreading data for code qualification purposes. After the material had spread, the
interaction was subsequently flooded to provide insights into debris coolability at large scale.
In this experiment, 3200 kg of a metal-oxide mixture was poured at a rate of ~ 20 l/s onto a 2 m
x 2 m siliceous concrete basemat. The oxide consisted of a mixture of Al2O3, SiO2, CaO and FeO; the
mixture was designed to have a broad freezing range of 1 000 to 1 550°C to emulate that of a
prototypic corium (Roche, Steidl, Leibowitz, Fink, & Sehgal, 1993). The metal phase was pure iron.
The melt evenly spread to an equilibrium depth of ~ 20 cm in less than 60 seconds. Strong gas
release from the concrete ensued, with hydrogen flames rising from the melt surface. Following the
spreading transient, the debris upper surface was flooded at a rate of ~ 4 l/s to examine the cooling
behaviour. In spite of ongoing core-concrete interaction with substantial gas release and an agitated
melt surface, the flooding process was mild and did not cause intense interactions of melt and water
(Figure 2.4-5). The melt surface crusted over and small volcanoes developed that ejected melt from
below in the form of lava flows, but particle ejection was not observed. The surface crust was firmly
NEA/CSNI/R(2016)15
anchored to the test section sidewalls. Only little concrete ablation has finally been evidenced as the
melt was not heated after pouring.
Figure 2.4-5: ECOKATS-2 during spreading (left) and after surface flooding (right)
The spreading area is 2 m x 2 m
Cool-down of the melt was slow due to the upper crust which limited the ingression of water.
Later in the experiment the central part of the crust was lifted by gases from the decomposing
concrete. This created a porous surface structure which allowed more efficient cooling of the upper
4 cm of the melt surface, but the balance of the melt material solidified in a dense form that excluded
water ingression and efficient cooling. Based on this simulant experiment one may conclude that top
flooding alone would be unlikely to stop concrete erosion by corium melt in a plant accident by the
water ingression cooling mechanism. However, it is important to note that subsequent water ingression
measurements made as part of the NEA MCCI Program ( (Lomperski, Farmer, & Basu, 2006),
(Lomperski & Farmer, 2007) indicate that the dry-out limit due to water ingression is quite high for in-
vessel core oxide material. Conversely, the dry-out limit is quite low for concrete slag materials such
as the silica-rich oxide simulant used in this spreading experiment.
2.4.4. COTELS experiments
Core concrete interaction and debris coolability were studied experimentally within the framework of
the COTELS project ( (Nagasaka, et al., 1999), (Nagasaka, et al., 1999), (Zhdanov, et al., 1999),
(Maruyama, Tahara, Nagasaka, & Vassiliev, 2002), (Maruyama, et al., 2006) from 1995 to 2002 as a
joint study between NUPEC (Japan) and NNC (Republic of Kazakhstan). The experiments were
conducted in the testing complex at NNC using the “LAVA-M” and “LAVA-B” facilities; schematics
are provided in Figure 2.4-6. Test series B/C was conducted at LAVA-M while test series D was
carried out at LAVA-B. The objectives of the B/C test series were to identify the mechanisms of
debris cooling and to understand the effects of various parameters such as debris composition, aspect
ratio (melt depth/diameter), and water injection flow rate on the cooling behaviour. Series D was
composed of separate effect tests to clarify findings from the B/C series. A total of 21 tests were
performed in the two test series.
NEA/CSNI/R(2016)15
Figure 2.4-6: COTELS lava M (left) and lava B (right) test facilities
Each facility was composed of an electric melt furnace (EMF), a concrete vessel, an induction
heater, and a water supply system. The corium composition was loaded into the EMF and melted by
indirectly heating a graphite crucible with a tantalum inner layer. When the temperature of the melt
reached a specified value, the melt was poured into the concrete vessel. Induction heating of the core
debris in the concrete vessel was then initiated to simulate decay heat.
The pressure vessel (MR) of LAVA-B was a horizontally oriented cylindrical tank with an ID of
1.8 m, length of 2.5 m, and a wall thickness of 30 mm. The vessel inner surface was thermally
insulated to minimise heat losses. The EMF and MR were inerted with argon before each test.
Figure 2.4-7: Concrete crucible dimensions and TC locations for test series D
The concrete test sections were cylindrical and made of basalt concrete. The structure and major
dimensions of the crucibles are shown in Figure 2.4-7, along with the thermocouple locations.
Thermocouples were embedded in the concrete to measure temperatures and to track the ablation front
location.
NEA/CSNI/R(2016)15
Major parameter ranges for the COTELS tests are summarised in Table 2.4-1Table 2.4-1. The
melt compositions were mixtures of UO2, ZrO2, Zr and stainless steel. The maximum melt mass was
59 kg. Prototypic core melt compositions for BWR and PWR reactor configurations were defined on
the basis of MAAP4 severe accident analyses. For a BWR the corresponding composition was
55/5/25/15 wt% UO2/ZrO2/Zr/stainless steel, while for a PWR the composition was 78/17/5 wt%
UO2/ZrO2/stainless. In test D-6, steel alone was used in order to study the effect of a lower melting
temperature. The initial melt temperature for the oxide tests was ~3 000 K. For test D-6 in which steel
melt was used, the temperature was 2 170 K.
Table 2.4-1: Parameter ranges for the COTELS experiments
Parameter Range
Melt Composition UO2, ZrO2, Zr, Stainless steel
Melt Mass 42-59 kg
Initial melt temperature 2 170-3 350 K
Concrete type Basaltic concrete (simulated by quartz concrete)
Inner diameter of cavity 0.2-0.36 m
Gross heat input 0-175 kW
Flooding water flow rate 0-0.2 kg/s
The concrete used in the experiments simulated Japanese LWR basaltic concrete 6 in terms of
aggregate size, distribution, composition, strength, and field curing. As noted earlier, cavity aspect
ratio was one of the experimental parameters. Thus, cavity diameters of 20, 26, and 36 cm were tested
in the series. The corresponding initial melt pool depths ranged from 7 cm to 22 cm. Gross heat input
to the core debris ranged from 50 to 175 kW; the heating efficiency was in the range of 10 to 47%.
Typical results for three experiments are presented here. Test conditions are summarised in Table
2.4-2. Test D-8a is a reference dry cavity case while test D-2 was flooded to examine coolability. Test
D-9 was a dry cavity test with the cavity inner wall lined with mortar to examine the effect of
aggregate on MCCI progression.
The temperature histories measured in the sidewalls of the concrete crucibles for tests D-2 and
D-8a are shown in Figure 2.4-8. When water was injected in test D-2, the side wall temperature
gradually decreased following the sharp temperature increase at the beginning of the experiment. As a
result, the concrete ablation depth in test D-2 was much smaller in comparison to test D-8a. A cross-
sectional view of the crucible for test D-8a is shown in Figure 2.4-9. Post-test examinations confirmed
that a layer of degraded concrete including once-molten material formed at the side and below the
solidified debris. The degraded concrete layer contained a large amount of un-melted coarse and fine
aggregate. It appeared as though the water had migrated into the thermally degraded concrete side wall
region for test D-2. The underlying reason for this occurrence is that concrete becomes porous when
heated due to the release of concrete decomposition gases, and the corresponding loss of material
causes interconnected porosity to form.
6. Siliceous aggregates were made nevertheless of crystalline quartz instead of vitreous basalt, increasing the
concrete melting temperature.
NEA/CSNI/R(2016)15
Table 2.4-2: Conditions for tests D-2, D-8a and D-9
Parameter D-2 D-8a D-9

Melt Composition (%) UO2 55
ZrO2 5
Zr 25
Stainless steel 15
Melt Mass (kg) 48 54 57
Initial melt temperature (K) 3 120 3 270 2 620
Concrete type Basaltic concrete Basaltic with Mortar layer
Inner diameter of cavity (m) 0.26
Gross heat input (kW) 90 75 50
Initial pressure of Pressure Vessel (MPa) 0.3
Water injection flow rate 0.03 0 0
Delay of water injection start (min) 9 - -
A cross-sectional view of the concrete vessel for test D-9 is shown in Figure 2.4-10. As opposed to
test D-8a, a layer of once-molten mortar with un-melted fine aggregate has accumulated at the upper
surface of the core debris. This indicates that the molten mortar was pushed out by the heavier, crusted
core debris, and this material accumulated as a slag layer on top the debris. This same type of behaviour
was observed in the FRAG test series at SNL (Tarbell, Bradley, Blose, Ross, & Gilbert, 1987).
A comparison of the axial ablation depth between D-8a and D-9 tests is shown in Figure 2.4-11.
Because of the difference in the net induction heat input to the debris in both tests, the integral heat
input to the initial sensible heat was used as the horizontal axis. A larger ablation depth was observed
in test D-9 where the cavity inner surface was covered with a mortar layer. The comparison between
tests D-8a and D-9 implies that the downward relocation of the solidifying core debris could be
dominated by melting of coarse aggregate when the aggregate is thermally more stable than cement.
To confirm the hypothesis that the cement and aggregate used in these tests had different melting
points, a separate small- scale test was performed in which a concrete sample was heated to 1 623 K.
The results indicated the cement had completely melted while the aggregate remained in a solid state.
(a) (b)
Figure 2.4-8: Temperature histories in concrete sidewalls for (a) test D-2 with water injection
and (b) test D-8a with a dry cavity
NEA/CSNI/R(2016)15
Figure 2.4-9: Vertical cross section after test D-8a (no water injection, no mortar layer
Figure 2.4-10: Vertical cross section after test D-9 (with mortar layer)
NEA/CSNI/R(2016)15
Figure 2.4-11: Axial ablation for tests D-8a and D-9, illustrating the effect of a mortar layer
The relationship between particle bed mass, input power, and ablation depth for the B/C test
series is illustrated in Figure 2.4-12 and Figure 2.4-13. There is a tendency for the bed mass to
increase with the ablation depth as well as the integrated input power. These trends imply that power
input to the melt promotes concrete ablation, and the decomposition gases from the ablated concrete
generate particle beds due to melt eruptions. This same phenomenology has been observed in the
MACE (Farmer, Kilsdonk, & Aeschlimann, 2009) and NEA MCCI CCI tests ( (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010).
Figure 2.4-12. Mass of particulate formed before water addition vs. input energy for
the B/C test series
NEA/CSNI/R(2016)15
Figure 2.4-13: Mass of particulate formed before water addition vs. concrete ablation depth
for the B/C test series
Major insights gained from the COTELS project are summarised as follows:
1. No steam explosions occurred in any test when subcooled water was injected onto molten
UO2 – ZrO2
2. The melt was cooled after water injection and the MCCI was suppressed. This cooling
was attributed to water migration into paths formed in the thermally degraded concrete
side wall.
3. The test results imply that concrete ablation processes depend on whether or not a stable
crust forms at the interface between core melt and concrete.
4. The extent of particle bed formation above the monolithic core debris increases as the
ablation depth increases and with the integrated power before water injection. This
relationship implies that particle bed formation is driven by decomposition gases released
from ablated concrete.
In addition to these findings, the following quantitative conclusions were drawn:
1. The size distribution in the particle beds fits a Rosin-Rammler distribution, and this
distribution is independent of the presence or absence of water.
2. Steady state heat removal rates from upper surface of the core debris to overlying water
ranged 0.2 to 0.7 MW/m2.
2.4.5. MACE experiments
As part of an international consortium, a series of coolability experiments were conducted at ANL
from 1989 to 2002 as part of the Melt Attack and Coolability Experiments (MACE) Program (Farmer,
Kilsdonk, & Aeschlimann, 2009). The principal programme objective was to explore the possible
benefits of water addition atop an MCCI insofar as: i) quenching and stabilising core melt, and
ii) arresting or even terminating basemat ablation. Four operationally successful integral experiments
were conducted; test specifications are provided in Table 2.4-3. Early tests M0 and M1b were
conducted with 70% oxidised PWR melt compositions, while the later tests M3b and M4 were
conducted with fully oxidised core melts. Three tests in the matrix were conducted with LCS concrete,
while the fourth test was conducted with siliceous concrete. A principal parameter in the matrix was
test section lateral span, which was varied from 30 cm x 30 cm up to 120 cm x 120 cm. Core melt
masses in the various scale tests ranged from 130 to 1950 kg.
The MACE system was very similar to that utilised for the core-concrete interaction experiments
performed as part of the NEA MCCI Project (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006),
NEA/CSNI/R(2016)15
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010); see Section 2.3.9 and Figure 2.3-30 for further
details. (In fact, the MACE facility formed the technical basis for the follow-on NEA project). There
were, however, several differences in the facilities and these are briefly described below.
Table 2.4-2: MACE test specifications (Farmer, Kilsdonk, & Aeschlimann, 2009)
Parameter Test
M0 M1b M3b M4
Corium type 70% oxidised PWR 70% oxidised PWR 100% Oxidised BWR 100% Oxidised BWR
Concrete type LCS LCS LCS Siliceous
Test section internal 30 cm x 30 cm 50 cm x 50 cm 120 cm x 120 cm 50 cm x 50 cm
dimensions
Sidewall material Concrete MgO MgO MgO
Initial melt mass (depth) 130 kg (15 cm) 480 kg (25 cm) 1 950 kg (20 cm) 480 kg (30 cm)
Initial Melt Temperature 1 730ºC (estimate) 1 800ºC 1 830ºC 1 930ºC
Melt formation technique DEH DEH Chemical reaction Chemical reaction
Specific power (actual) 700-1 400 W/kg 350 W/kg 300 W/kg 300 W/kg
Ablation depth at flooding 1.3 5.0 1.3 3.8
Water collapsed pool depth 50 cm 50 cm 50 cm 50 cm
Inlet water flow rate/temp. 10 lps/20 ºC 2 lps/20 ºC 2 lps/20 ºC 2 lps/20 ºC
DEH power operating mode Constant voltage Constant voltage Constant voltage Constant voltage
The first difference was in the test sections that contained the core debris. In all four MACE tests,
an instrumented concrete basemat was located at the base of the test section. In three of the four tests,
the sidewalls of the apparatus were constructed from inert MgO, and so the tests were 1D. However,
the M0 test was conducted with concrete walls also, and therefore provided data on 2D erosion
characteristics. In the CCI tests, two walls were concrete while the other two walls were MgO.
In the first two experiments M0 and M1b, the corium powder charge was a mixture of crushed
UO2 pellets and ZrO2 powders, plus a small amount (typically 8 wt%) of calcined concrete powders
(CaO and SiO2). The concrete was incorporated to account for erosion that is expected to occur during
the corium spreading phase following breach of the reactor pressure vessel. The melt pool was
produced by gradually heating the powders using DEH over a period of several hours. However, for
tests M3b and M4 the DEH technique was replaced by specially designed exothermic mixtures that,
upon ignition, produced the target core melt composition over a timescale of ~ 30 seconds. This
approach was used in all tests conducted in the NEA MCCI Program.
Once the melt was produced and a specified concrete erosion depth was achieved, the cavity was
flooded. Water was delivered at a sufficient rate to ensure that the quench process would not be water-
starved (i.e. sufficient to sustain a quench rate of at least 10 MW/m2). Once the cavity was flooded,
power supply operation was switched from constant power to a constant voltage mode. This was due
to the fact that core material that is quenched is not effectively heated using DEH. However, operation
in a constant voltage mode maintained the specific power density in the remaining melt zone at the
target level. Thus, if significant debris quenching occurred, the input power would decrease while
operating in this manner. Steam and non-condensable gases from the interaction were vented through
a series of quench tanks and finally an off-gas system. This part of the facility was instrumented to
quantify the corium quenching rate (based on the steam condensation rate), as well as the gas
composition and flow rate from core-concrete interaction.
A summary of principal experiment findings from the MACE Program is provided below. Data
from two of the experiments (i.e. M0 and M3b) are presented in order to illustrate the type of
information that was obtained.
NEA/CSNI/R(2016)15
The scoping test M0 was performed in order to provide an early indication of the mode and
extent of corium quenching, and to provide information that would aid in future experiment designs.
Water addition in M0 began when the basemat ablation depth reached 1.3 cm (i.e. at 4.0 minutes;
see Figure 2.4-14). At this point, the melt surface camera indicated an agitated pool with no evidence
of an upper surface crust. A direct measurement of melt temperature was not available at the time the
cavity was flooded due to limited instrumentation (Figure 2.4-14). However, it is believed that the
melt temperature was below the Zr melting point (1 857ºC) since an intense interaction was not
observed when the Zr rod located at the original core-concrete interface was contacted by melt. Later
investigations indicated that the low melt temperature was due to extensive concrete sidewall ablation
during the corium heat-up phase before the melt contacted the basemat.
The melt/water heat flux data for M0 is shown in Figure 2.4-15, along with DEH input power
normalised by the initial cavity area. The data indicated that during the first three minutes of the
interaction, a large cooling transient occurred with the heat flux peaking at nearly 4 MW/m2. In the
next three minutes, the heat flux fell rapidly and a quiescent period was observed. This occurrence is
attributed to the formation of a stable interfacial crust (i.e. the large initial heat removal caused the
melt temperature and sparging rate to decline to the point where a crust could form in the presence of
the sparging gas). Although perturbations occurred, the heat flux over the next 30 minutes averaged
~ 700 kW/m2, which amounts to ~ 70% of the DEH input heat flux of ~ 1 MW/m2. After this time the
flux diminished steadily to a level of ~ 150 kW/m2 at the end of the test.
12 2500
Centerline
10 NE Quadrant
2000
SE Quadrant
SW Quadrant
8
Ablation depth (cm)
Temperature (ºC)
NW Quadrant
1500
1000
4
Junction Elevation (cm)

500
2 0.0 -0.1 -2.0
0.0 -10.0
0 0
-20 -10 0 10 20 30 40 50 60 70 80 90 -20 -10 0 10 20 30 40 50 60 70 80 90
Time (minutes) Time (minutes)
Figure 2.4-14: MACE M0 Axial Ablation Data (left) and Melt Temperature (right)
4500
Onset of eruptions
Melt-Water
4000 Input Power
cm
3500 + 50
3000 + 40 LOOSE DEBRIS

Heat flux (kW/m )
2
SINTERED PARTICLE BED

2500 + 30
2000 + 20 SOLID CRUST
+ 10 DATA POINT
1500
VOID
1000 0
SOLIDIFIED MELT
-10
500
30 cm SQ. BASEMAT
-20
0
25 cm THICK CONCRETE WALL
-20 -10 0 10 20 30 40 50 60 70 80 90
Time (minutes) DRAWING: MACE SCOPING TEST
ACTUAL SIDEWALL ABLATION
NORTH SOUTH DRAWING NO.: M1038
Figure 2.4-15: MACE M0 Debris-water heat flux (left) and post-test debris configuration (right)
NEA/CSNI/R(2016)15
Post-test examinations indicated that the crust was anchored by the tungsten electrodes. The crust
remained at the elevation where it initially formed. As the MCCI progressed downward, an
intervening gap formed between the melt and crust7, which most likely terminated efficient heat
transfer processes between the debris and water. For this particular test, crust-melt separation probably
occurred at ~ 30 minutes into the test sequence, as evidenced by the rapid decline in the heat transfer
rate at this time. Despite this, ejections of molten corium through the bridge crust were observed
(Figure 2.4-15). These ejections were driven by concrete decomposition gases, which caused pool
swelling and entrained melt through the crust into overlying water in the form of dispersed droplets.
These droplets quenched as they settled through the water pool and collected on the crust upper
surface to form a particle bed (Figure 2.4-15). By the end of the test, the bed depth had reached
~ 10 cm. The top half was loosely packed while the bottom half was sintered agglomerate. The mass
of the bed was ~ 23 kg, which amounts to ~ 20% of the initial melt mass. These ejections occurred
every 10-15 minutes, and they resulted in transient augmentation in the upwards heat flux to levels in
the range of 1-3 MW/m2.
During disassembly, the concrete sidewalls were found to be eroded by as much as 10 cm.
Sidewall thermocouple data indicated that radial ablation was occurring as early as 50 minutes before
the melt reached the basemat. It is estimated that at onset of axial ablation, the total corium mass was
~ 130 kg, of which 23 wt% was concrete decomposition products from sidewall erosion. This large
influx of concrete from the sidewalls during the pre-heat probably depressed the melt temperature to
the extent that the Zr was not significantly melted. Thus, the actual M0 test conditions reflect a 100%
oxidised PWR melt composition ~ 23% diluted with concrete oxides run at 2-4 times prototypic decay
heat level at 2 hours into the sequence.
Following the scoping test, the facility design was changed to allow larger scale tests of up to
75 cm x 75 cm scale to be performed. The test section was constructed with refractory sidewalls
(MgO) rather than concrete to prevent early concrete dilution of the melt. The tungsten electrodes
were recessed into the sidewalls to reduce the chances of crust anchoring on these surfaces. A new
0.56 MW power supply was installed and an online gas mass spectrometer was added to the system.
Test M1b was subsequently conducted in a 50 cm x 50 cm test section with a collapsed pool depth of
25 cm. This test ran for six hours following onset of ablation. The experiment was terminated on the
basis that most basemat thermocouples showed that the concrete temperature had stabilised, indicating
that ablation had been arrested. As in the Scoping Test, Test M1b provided evidence of high initial
debris cooling rates as well as periodic ejections of molten corium. A large melt eruption was
observed early in the sequence that led to a significant transient increase in the upwards heat flux up to
~ 1.7 MW/m2. The heat flux data also indicates that two minor eruptions may have occurred late in
the test.
Following M1b, the overall experiment approach for MACE was re-examined. In particular, the
facility was redesigned to accommodate the largest possible test section in order to minimise the
chances of crust anchoring to the test section sidewalls, as occurred in both previous tests. As shown
in Table 2.4-3, this effort resulted in a test section featuring an internal cross section of 120 cm x
120 cm. The corresponding initial melt mass was specified as 1 950 kg, yielding an initial collapsed
melt depth of 20 cm. The test section design was also upgraded to include a UO2 pellet liner between
the corium charge and the MgO sidewalls. The liner was added to eliminate reactions between the
corium and the sidewalls, as occurred in M1b. Finally, extensive laboratory testing was carried out to
develop an exothermic chemical mixture capable of producing a prototypic core melt composition
over a short timescale (i.e. ~ 30 seconds).
7. As CaCO3 decarbonates, it loses about half of its volume.
NEA/CSNI/R(2016)15
These efforts culminated in Test M3b that ran for six hours following onset of ablation. The
experiment was terminated on the basis that virtually all melt temperature thermocouples showed that
the temperature had fallen below the concrete solidus of ~ 1 100ºC (Figure 2.4-16). Water addition
began when the ablation depth had reached the target depth of 1.3 cm (i.e. at 52 minutes). Other
measurements indicated that the melt temperature at the time water was added was ~ 1 870ºC, and the
melt sparging rate was ~ 12 cm/sec.
The melt/water heat flux (Figure 2.4-17) indicated that during the first 20 minutes of the
interaction, a large cooling transient occurred with the initial upwards heat flux peaking at nearly
1.8 MW/m2. This peak is less than the 4 MW/m2 level reached for the M0 and M1b tests. The reduced
initial cooling rate was most likely due to the fact the melt temperature for M3b was relatively lower,
and also that the corium contained less concrete as an initial constituent (i.e. ~ 8 wt%) in comparison
to M0 and M1b. The lower melt temperature, in conjunction with the reduced presence of concrete
oxides, resulted in a more viscous melt which underwent a less effective bulk transient compared to
the other tests.
35 2500
Junction Elevation (cm)
8.0 3.5 1.5 -1.0
30 5.5 -6.5 -14.5
2000
25
Ablation depth (cm)
Temperature (ºC)
1500
20
15 1000
Centerline
NE Quadrant
10 SE Quadrant
SW Quadrant 500
5 NW Quadrant
CO/CO2 Gas Release
0 0
0 50 100 150 200 250 300 350 400 450 500 -50 0 50 100 150 200 250 300 350 400 450 500
Time (minutes) Time (Minutes)
Figure 2.4-16: MACE M3b axial ablation data (left) and melt temperature (right)
2000
Melt-Water
Input Power
Onset of eruption
1500
Heat flux (kW/m 2)
Periodic eruptions
observed
1000
500
0
0 50 100 150 200 250 300 350 400 450 500
Time (minutes)
Figure 2.4-17: MACE M3b debris-water heat flux (left) and post-test debris configuration (right)
Although there are fluctuations, the data generally indicates that the melt temperature did not
change significantly following the initial 20 minute cooling transient that resulted in ~ 1.1 GJ of
upwards heat removal. The lack of a bulk temperature decrease indicates that the heat removal was
dominated by an interfacial freezing process, as opposed to convective cooling of the entire melt pool.
Based on a simple energy balance analysis (Farmer, Kilsdonk, & Aeschlimann, 2009), this amount of
NEA/CSNI/R(2016)15
heat removal would have resulted in complete quench of ~ 1 000 kg of core material, which amounts
to ~ 50% of the initial melt mass. The post-test debris configuration (Figure 2.4-17 and Figure 2.4-18)
revealed the formation of a bridge crust that weighed 1 050 kg, which agrees within 5% of the
quenched debris mass predicted on the basis of the energy deposition in the quench system. Other test
data (i.e. power supply response) is consistent with quench of ~ 1/2 the initial melt mass during the
first 20 minutes of the test. Post-test measurements also indicate that this thick bridge crust was
permeable to both gas and water flows. Finally, debris temperature measurements logged during the
post-test cool down indicate that water passed through the crust and quenched the underlying debris
over the time interval 550-680 minutes after onset of ablation (or 80-210 minutes after the test was
terminated). Thus, these various sources of information indicate that a significant cooling transient
occurred over the first 20 minutes of M3b which rendered ~ 1/2 the initial melt mass quenched and
coolable by virtue of water ingression into a growing crust.
The crust that formed during the initial interaction also anchored to the sidewalls, eventually
leading to the formation of an intervening gap (~ 23 cm) as the MCCI continued downwards.
Separation is believed to have occurred at ~ 72 minutes, as evidenced by the steep decline in
melt/water heat flux and the resurgence of basemat ablation after this time. Over the next two hours of
the experiment, the melt temperature and ablation rate approached steady state conditions, while the
melt/water heat flux declined steadily. The reduction in upwards heat transfer was attributed to
increasing radiation heat loss to the sidewalls as the melt receded from the suspended crust due to
concrete densification upon melting.
Figure 2.4-18: MACE M3b upper surface debris as-found (left) and after removal
of particle bed and volcanic structures (right)
As in the previous tests, M3b provided evidence of periodic ejections of molten corium. An
eruption was observed on video during the initial stage of the interaction which led to a significant
increase in the upwards heat flux to ~ 700 kW/m2. Over the time interval of 252-328 minutes, the
input power was increased back to the initial target level of 390 kW. Following the power increase,
periodic eruptions were observed until the power was again reduced at 328 minutes. These eruptions
resulted in increases in the upwards heat flux to levels as high 700 kW/m2. The post-test debris
examinations (Figure 2.4-18) indicated that the ejected material was quenched in the form of a large,
centrally located volcanic formation overlying a debris bed which was composed of spherical
particles. The mass of the ejected material amounted to 505 kg, of which 310 kg was located in the
particle bed, while the balance (i.e. 195 kg) was in lava formations. The 505 kg of ejected material
amounts to ~ 26% of the initial melt mass.
After the power reduction at 328 minutes, the melt temperature steadily declined to below the
concrete solidus of ~ 1 100ºC at 470 minutes (Figure 2.4-16). This trend is most likely attributable to
NEA/CSNI/R(2016)15
the fact that the upwards heat transfer rate approximately balanced the input power during this time
(Figure 2.4-17). The test was eventually terminated on the basis that the bulk melt temperature had
fallen below the concrete solidus.
Post-test examinations further revealed that ~ 1/3 of the bridge crust failed and relocated
downwards at least once. This conclusion is based on the fact that a large fracture ridge was found in
the crust which roughly spanned two opposing corners of the test section (Figure 2.4-18). The crust on
the side of the ridge that had relocated was highly cracked, and the surface elevations in this area were
below the initial collapsed melt height. Thus, although the important objective of maintaining a
floating crust was not met in this increased scale test, the M3b post-test debris configuration
nonetheless suggests that a floating crust boundary condition is the expected situation at plant scale.
This observation was further reinforced by crust strength measurements subsequently made as part of
the NEA MCCI Program (Lomperski & Farmer, 2009).
During disassembly, the MgO sidewalls were found to be in very good condition, indicating that
the UO2 liner had functioned properly insofar as protecting the test section sidewalls. Thus, during the
first 4 hours of M3b, the test conditions reflect a 100% oxidised core melt composition initially diluted
with ~ 8% concrete oxides run at prototypic decay heat level at two hours into the accident sequence.
During the elevated power period from 252-328 minutes, the power density varied from 2-5 times the
target level.
Due to the fact that an anchored crust formed in M3b, the decision was made to return to the
smaller 50 cm x 50 cm test section. Furthermore, all previous tests were conducted with LCS concrete,
and therefore the decision was also made to conduct the final test with a siliceous concrete basemat to
round out the test matrix. This test, denoted M4, was conducted with a system very similar to that used
for M1b, with the exception that an exothermic chemical mixture was used to generate the melt as
opposed to DEH. The mixture was reformulated to contain 8.5 wt% calcined siliceous concrete, as
opposed to calcined LCS concrete used for the M1b and M3b mixtures. Test M4 was conducted with a
480 kg corium mass, yielding an initial collapsed melt depth of 30 cm in the 50 cm x 50 cm test
section (Table 2.4-3). The test section design retained the use of a UO2 pellet liner to protect the MgO
sidewalls. Final important additions to the design were two insertable “crust busters” which were to be
used to dislodge a bridge crust should one develop during the test.
M4 ran for 4 hours following onset of ablation, and the test was terminated on the basis that the
maximum permissible axial ablation depth of 31 cm had been reached. Water addition began when the
ablation depth reached 3.8 cm (i.e. at 22.8 minutes). The melt temperature was ~ 1930 ºC at the time
of flooding, and the melt sparging rate was ~ 15 cm/sec. Although there was efficient initial cooling of
the debris, a bridge crust formed early in the experiment with subsequent gap formation, and this
effectively terminated cooling processes. No melt eruptions occurred in this experiment, which may be
attributable to the formation of the gap. During disassembly, the MgO sidewalls were found to be in
very good condition. Thus, the test conditions for M4 reflected a 100% oxidised core melt
composition initially diluted with ~ 8% concrete oxides run at prototypic decay heat level at two hours
into the accident sequence.
2.4.6. NEA MCCI Project CCI experiments
Aside from examining the thermal-hydraulic aspects of molten core-concrete interaction under dry
cavity conditions, a second and equally important aspect of the NEA MCCI test series ( (Farmer,
Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010)
was to investigate debris coolability under both early and late-phase flooding conditions. The
hardware and procedures for these experiments were described previously; see Section 2.3-9.
A total of six tests were conducted as part of the series; specifications and operation summaries
for the individual experiments are provided in Table 2.3-9. Of the six tests, four provided quantitative
information on the extent of cooling when water contacts molten core debris. In CCI-4 direct melt-
NEA/CSNI/R(2016)15
water contact was precluded by the presence of a large mantle crust that formed in the upper regions of
the test section due to extensive melt foaming that occurred over the course of the experiment (Farmer,
Lomperski, Kilsdonk, & Aeschlimann, 2006). For CCI-5, the cavity was not flooded due to plugging
of the main gas line for the apparatus (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010). Thus,
these two tests are omitted from the discussion below.
In terms of phenomenology, four previously identified (Farmer, Kilsdonk, & Aeschlimann, 2009)
cooling mechanisms were targeted for investigation at the onset of the MCCI Program: i) bulk cooling,
ii) water ingression through cracks/fissures in the solidifying material, iii) melt eruptions, and
iv) transient crust breach. As a whole, the test series provided data on all four of these mechanisms. In
addition, Test CCI-2 provided data on a fifth cooling mechanism; that is, water ingress at the interface
between the core material and concrete sidewalls. In Test CCI-2, thermocouple measurements
indicated that water was able to ingress at this interface and effectively cool the concrete sidewalls to
saturation, thereby terminating the lateral cavity erosion process. This mechanism had been previously
identified in the COTELS reactor material test series ( (Nagasaka, et al., 1999), (Nagasaka, et al.,
1999), (Zhdanov, et al., 1999); see Section 2.4.4). Although this is a beneficial cooling mechanism, the
results need to be carefully scaled to plant conditions as surface-to-volume effects will diminish the
overall effectiveness as scale is increased.
Figure 2.4-19: Debris-water heat fluxes for late-flooded tests CCI-1 through -3 (left),
and for early flooded test CCI-6
As shown in Figure 2.4-19, the heat flux during the initial flooding stage was generally high for all
tests. For the two late-flooded tests conducted with siliceous concrete (CCI-1 and CCI-3), the initial
cooling rates approached the Critical Heat Flux (CHF) limitation of ~ 1 MW/m2 under saturated boiling
conditions. Thus, these fluxes were indicative of quenching of the upper surface crust that was initially
present for both tests. Although the lance was used to puncture the crust for these tests before the cavity
was flooded, the crusts were floating and the openings were generally small compared to the remaining
crust surface area over the melt. However, for test CCI-2, the upper surface was essentially crust-free due
to the insulating effect of the overlying crust mantles that had formed over the previous five hours of dry
cavity operations. Furthermore, for CCI-6 only a thin skin crust was present and the melt was well sparged
due to the early nature of the interaction. Thus, for both of these tests bulk cooling transients resulted in
which the initial cooling rates were quite high (i.e. ~3 MW/m2 and ~5 MW/m2 for CCI-6). As is evident
from Figure 2.4-19, the heat fluxes eventually fell below 1 MW/m2 after ~ 5 minutes. At this time, stable
crusts most likely formed at the melt-water interface, thereby terminating the bulk cooling transient.
NEA/CSNI/R(2016)15
Figure 2.4-20: Average melt temperature for CCI-1 through -3 (left) and local melt temperatures
measured across the extent of the debris zone for CCI-6 (right)
In general, the tests did not exhibit a pronounced decrease in overall melt temperature after cavity
flooding (Figure 2.4-20). This is despite the fact that the heat flux and power supply responses
indicated significant debris cooling. This type of behaviour can be rationalised by a latent heat transfer
process in which a quench front develops at the melt/water interface, as opposed to a sensible heat
transfer process in which the entire melt mass is cooled by convective heat transfer where the heat is
dissipated to the overlying water by conduction across a thin crust at the melt/water interface. Indeed,
after incipient crust formation at the melt/water interface, the bulk melt temperature response, power
supply response in constant voltage operating mode, 8 and post-test debris morphology are consistent
with development of quenched debris zones as opposed to bulk cool down of the entire melt mass by
conduction-limited cooling across a thin crust at the melt/water interface. In fact, as shown in Figure
2.4-20, the average melt temperature in test CCI-2 actually increased for a period after water addition,
while the debris was undergoing extensive cooling. Simple energy balance analyses coupled with post-
test examination results (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010) indicate that in order to
extract the heat fluxes from the core debris shown in Figure 2.4-19, water-ingression cooling must
have occurred in order to rationalise the thicknesses of solidified crust material discovered during
post-test examinations. The potential for water to ingress into core debris was also measured as part of
the SSWICS counterpart tests conducted as part of the NEA MCCI Program ( (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010). A
photograph of fractured crust material recovered during post-test disassembly of CCI-6 is provided in
Figure 2.4-21.
8. After water addition, the power supply was run in a constant voltage mode to maintain the power density in
the melt constant; see Section 2.6 for additional discussion regarding power supply operations for these tests.
NEA/CSNI/R(2016)15
Figure 2.4-21: Photographs of CCI-6 melt eruption particle bed (left) and fractured crust (right) debris
regions recovered from CCI-6
Aside from the water ingression mechanism, the tests also provided data on the nature and extent
of the melt eruption cooling mechanism after cavity flooding. In particular, significant eruptions were
observed for test CCI-2 that was conducted with LCS concrete, and for test CCI-6 that featured early
cavity flooding. A photograph of erupted material recovered during disassembly of CCI-6 is provided
in Figure 2.4-21. In contrast, no spontaneous eruptions were observed after late flooding for the two
other tests conducted with siliceous concrete (CCI-1 and CCI-3). As discussed by Bonnet and Seiler
(Bonnet & Seiler, 1992), the gas sparging rate during core-concrete interaction is the key parameter
influencing the melt entrainment process during eruptions. Thus, the reduced gas content for this
concrete type coupled with the fact that the MCCI was flooded in the late stage may have been a key
contributor to the lack of eruptions for these two tests.
Sufficient information was gathered during the CCI-2 and CCI-6 tests to evaluate the melt
entrainment coefficient after cavity flooding. This information can be used in models for assessing the
extent that core debris can be rendered coolable by virtue of top flooding in accident analyses. For the
CCI-2 test conducted with LCS concrete the average entrainment coefficient, defined as the ratio of
the melt volumetric entrainment rate to the hot gas volumetric flow rate (Bonnet & Seiler, 1992), was
found to be ~ 0.11% for CCI-2. This entrainment rate is consistent with that observed in previous
MACE integral effect tests carried out with LCS concrete (Farmer, 2005). Conversely, for CCI-6
conducted with siliceous concrete, data analysis indicates that the melt entrainment coefficient
averaged 0.04% over the course of the test. Both of these entrainment estimates are within the range of
that required to stabilise a core-concrete interaction over a fairly significant range of melt depths
(Farmer, 2005).
The entrainment coefficient data for these two tests are significant since the eruptions occurred
under a floating crust boundary conditions (as evidenced by the post-test examinations that indicated
the absence of a continuous void region below the crust upper surface in both experiments), and while
the input power was decreasing, so that the melt zone was not over-powered during the eruption
process. Thus, the data upon which the entrainment coefficient is based are deemed to be prototypic.
On this basis, the entrainment coefficient may be used to evaluate the effects of the melt eruption
cooling mechanism on mitigating the core-concrete interaction under plant accident conditions for the
case of LCS concrete over a range of cavity flooding times, and for siliceous concrete under early
cavity flooding conditions.
NEA/CSNI/R(2016)15
2.5. The Chernobyl accident
During the April 1986 severe accident at Chernobyl unit 4, tens of tonnes of molten corium interacted
with the reactor room concrete. Thus, this accident offers the unique opportunity to obtain information
on MCCI behaviour under dry cavity conditions at prototypic scale that augments the data provided in
the preceding sections.
In an RBMK reactor there is no pressure vessel but rather 1 661 pressure channels, as in a
CANDU reactor. The reactor structures (Figure 2.5-1) include a large (14-metre diameter, 2-meter
high) lower biological ‘OP’ shield plate that has no structural role, but houses radiological and thermal
shielding materials, including serpentinite and Mg3Si2O5(OH)4 (a residue from asbestos waste
processing). There are about 300 tonnes of sand between the Л (L) lateral water shield and the cavity
walls.
The reactor cavity is rectilinear with rough dimensions of 18.2 m x 23.6 m. The cavity is divided
into quadrants by a reinforced-concrete cross (1 m high, 1.4 m wide) that is visible on Figure 2.5-2. At
the centre of this cross, there is a 5.3-m high metallic beam assembly connected to the ‘OP’. The
concrete floor 9 lies at an elevation of +9.7 m. Above the concrete there is a layer of insulation and then
a metallic liner at the +10 m elevation. The vertical walls are also protected by a liner and insulation.
The concrete is basaltic (i.e. silica-rich) with a thickness of 1.7 m. There are 8 steam bleeder valves
which are attached to penetrations (steam dump tubes) in the reactor cavity floor.
Evaluations ( (Bogatov, et al., 2007), (Pazukhin, 1994) indicate that debris from the core and the
reactor shielding structure collected in the reactor cavity following the destruction of a 105-110° sector
of the ‘OP’ in the southeast (SE) quadrant (i.e. between Л (L)47 and И (I)46 in Figure 2.5-2) and the
southwest (SW) quadrant (i.e. between Л (L)48 and И (I)47 in Figure 2.5-2). Due to insufficient decay
heat removal and heat produced from oxidation reactions, the debris remained molten for several days.
The corium flowed through the piping and ablated the walls and floor of the reactor cavity.
Samples were collected, mostly from the upper surface, since it was very difficult to drill within
the corium mass. The SE and SW quadrants were covered with corium (total surface area of ~ 170 m²,
equivalent to the EPR spreading chamber). The corium was predominately a brown ceramic in the SW
quadrant, and ii) predominantly a black ceramic in the SE quadrant; see Figure 2.5-3 ( (Pazukhin,
1994), (Burakov, 2013). The compositions of these samples (see) are representative of that determined
from samples taken throughout the reactor building.
9. The floor of room 305/2 is not a basemat since there are several levels of rooms below this level, as visible in
Figure 2.5.1
NEA/CSNI/R(2016)15
Figure 2.5-1: Top view of under-reactor room 305/2 (Bogatov, et al., 2007)
Figure 2.5-2: Repartition of corium on the floor of the reactor cavity
NEA/CSNI/R(2016)15
Figure 2.5-3: Corium samples (TOP 1-4: brown ceramic from the large vertical flow;
BOTTOM 1-4: black ceramic from room 304/3) (Burakov, 2013)
Table 2.5-1: Average elemental composition (wt%) of corium samples. Light elements (mostly
oxygen) have been omitted. (Pazukhin, 1994)
Type Al Mn Fe Pb Cr Mg Ni B Si Ca Zr Na U
Black 4.8 0.3 1.4 0.05 0.2 2.4 0.1 0.04 29.8 5.5 3.2 4.2 4.7
ceramic
Brown 3.3 0.5 0.8 - 0.2 5.0 0.3 0.07 32.7 4.5 4.3 - 8.3
ceramic
It should be noted that there is a very low concentration of uranium in both samples (<9 wt%);
this is explained by the large amounts of structural materials mixed with the fuel. Table 2.5-2
summarises the inventory of materials that were present in the cavity, as well as estimates of the
masses of these materials that were incorporated into the corium based on the chemical analysis
results1.
‘Vertical lava flows’ occurred through the steam dump tubes. The corium flow stopped without
any intervention at the ground floor and did not damage the basemat. More than 40 boreholes were
made in the concrete walls and floor to determine the extent of the corium-concrete interaction in the
NEA/CSNI/R(2016)15
reactor cavity. Note that the vertical wall between rooms 305/2 and 304/3 was melted through (Figure
2.5-4) leading to what is known as the ‘horizontal lava flow’.
Table 2.5-2: Materials present in the reactor cavity at the beginning of corium formation and
incorporated during the MCCI (Bogatov, et al., 2007)
Material Amount in Rooms Incorporated into
504/2 10 and 305/2 after corium (tonnes)
the accident (tonnes)
Fuel (U) 120 90

Steel 1 300 11 < 20 12
Serpentinite mixture 580 160
Concrete from under the reactor plate - 130
Building construction concrete dropped into the vault 950 480
from upper level marks
Sand from the vault filling material 300 280
Zirconium ? 45
Graphite 750 virtually none
Figure 2.5-4: Schematic illustrating vertical corium flows from reactor cavity (305/2)
10. within the boundaries of the reactor space

11. excludes materials from the ‘cross’ component and non-melted components of the reactor bottom
12. 330 t of melt spread within the under-reactor rooms
NEA/CSNI/R(2016)15
Figure 2.5-5: Cross-section (along the K axis in Figure 2.5.1) illustrating the cavity ablation profile
due to MCCI (dashed lines show the initial floor position)
Figure 2.5-5 presents the estimated shape of the cavity ablation profile due to MCCI along the
K axis (see Figure 2.5-1). The ablation is clearly anisotropic with a maximum axial ablation depth of
65 cm and a maximum lateral depth of 2.4 m (ratio >3). Figure 2.5-6 provides a synthesis of the other
ten (10) cross-sectional measurements that were made.
Figure 2.5-6: Synthesis of ablation profiles along 11 section axes
In some of the cuts (i.e. from section I to section K-2 m) the corium spread was not complete and
there was little or no contact between the corium pool and the lateral walls. Thus, no lateral ablation
occurred in these locations. In the next cuts (e.g. from K to K+2.4 m) significant lateral ablation
occurred, reaching a maximum ablation depth of 3.5 m, while the vertical ablation was limited to
1.3 m. The lateral/axial heat flux split was thus ~ 2.7 in these regions. Also note that significant lateral
ablation occurred in the direction perpendicular to where the south-eastern wall of room 305/2 was
melted through. Table 2.5-3 summarises the ablation depths measured at several different locations.
The lateral/axial ablation splits range from 2 to 4.
The lateral/axial heat flux splits determined as part of these examinations are noted to be similar
to that found in in the CCI and VULCANO 2D reactor material tests conducted with silica-rich
concretes (see Section 2.3). Moreover, the tendency for ablation to move along an oblique line has
NEA/CSNI/R(2016)15
also been observed (see e.g. Figure 2.5-7) in Chernobyl. This comparison thus indicates that the
ablation anisotropy observed in the tests is not an experiment artefact or scaling distortion, but rather a
genuine property of core-concrete interaction with silica-rich concretes. However, it should be noted
that the corium in this reactor accident formed in the reactor cavity and was not poured in a liquid state
as is expected in LWR accident scenarios.
Finally, note from Table 2.5-1 that by the end of the accident, the black ceramic corium had less
than 10 wt% of (U,Zr)O2. Such a low concentration would occur in a PWR only after a very large
mass of concrete has been ablated, typically after several days of interaction.
Figure 2.5-7: Cross-sectional view at the K + 2 m axis location; Significant lateral ablation of
~ 3 m is noted. Vertical ablation: ~ 0.5 m below the original cavity, ~1.1 m in the lateral tongue
Table 2.5-3: Summary of vertical and lateral ablation measurements

Cross section Vertical ablation Lateral ablation Ratio
I 40 cm >63 cm > 1.6 13
I+1 200 1m No -
I +1 650 80 cm No -
K-3 000 70 cm No -
K-2000 80 cm No -
K-1 000 80 cm 50 cm 0.6
K 65 cm 240 cm 3.7
K+2000 110 cm 300 cm 2.7
K+2 400 130 cm 350 cm 2.7
L-700 60 cm 40 cm 0.7
2.6. Lessons learnt from molten core-concrete interaction experiments
2.6.1. Operational issues

Molten core-concrete interaction experiments conducted as part of the ACE/MCCI
(Thompson D. H., Fink, Armstrong, Spencer, & Sehgal, 1992), (Fink, Thompson, Armstrong,
Spencer, & Sehgal, 1995), MACE (Farmer, Kilsdonk, & Aeschlimann, 2009), and NEA MCCI
13. As the lateral wall was melted through, it is not possible to estimate what would have been the ablation if the
wall had been thicker.
NEA/CSNI/R(2016)15
Programs (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, &
Aeschlimann, 2010) used the direct electrical heating (DEH) method to simulate decay heat in the core
debris at a power level typical of a PWR at ~2 hours after reactor scram. 14 After melt formation, the
power was held constant during the dry cavity phase of the experiments. In many tests, the cavity was
flooded after a predetermined time or ablation depth was reached to provide additional data on core
debris coolability. Following water addition, the power supply operating mode was usually switched
from constant power to constant voltage. This was done due to the fact the DEH does not appreciably
heat solidified material, and so if debris quenching occurred, then the power density in the remaining
melt would increase to non-prototypic levels if constant power operation was maintained. For a given
melt resistivity, constant voltage operation preserves power density since the overall melt resistance
increases as debris is quenched, and so the electrical current (and thereby gross power) naturally
adjusts to maintain the target density 15. However, in reality the melt resistance evolves over the course
of the experiment due to changes in composition (by incorporation of concrete slag due to concrete
ablation) and temperature (by virtue of debris cooling) 16. There are uncertainties associated with
estimating the amount of ablation and melt temperature during the experiments, as well as the overall
affect that these parameters have on corium resistivity. These uncertainties need to be factored in to
the analysis of the test data.
There are generally two criteria used for the termination of an MCCI experiment. The first is that
the ablation depth (axial and/or radial) reaches a specified maximum value. For wet cavity tests, the
second criterion is that a predetermined time after water addition is met. The maximum ablation depth
specification ensures that there is no gross failure of the interaction crucible. The choice of test
duration after water addition, on the other hand, is predicated on the notion of the debris cooling rate
being reduced to a sufficiently low level such that further continuation of the experiment is not
expected to yield additional useful data. Generally, the ablation limit is a more restrictive termination
criterion that controls the test duration after water addition. This duration varies from as little as
30 minutes (typically for siliceous concrete) to ~ 2 hours or more (typically for limestone concrete).
Extrapolation of limited data from shorter duration tests to much longer duration reactor scale core-
concrete interactions may be problematic. Therefore, some degree of conservatism may be prudent in
specifying the test termination criteria.
Precise measurement of melt temperature in extremely high temperature environments has
always been a challenge in experiments. Even with the significant advances made in thermocouple
technology, it is not unusual to witness a large number of thermocouples failing in a given experiment.
Moreover, the reliability of data from surviving thermocouples may be questionable, especially since
these thermocouples are operated in extremely corrosive environments and the large dimension of
their protective sheaths may act as thermal bridges perturbing the temperature field. Consequently,
uncertainties in thermocouple data need to be factored into any analysis.
Finally, crust anchoring has been found to have a profound (and detrimental) influence on core
debris cooling behaviour that is not expected at reactor scale. To avoid crust anchoring in future
14. Neglecting noble gases, the power density in the melt would normally correspond to ~300 W/kg fuel at this
time for a PWR. The corresponding average heat flux to the surfaces of the melt for a 30 cm collapsed melt
depth was in the range of 150-200 kW/m2 at this power density level.
15. See Appendix B of reference (Farmer, Kilsdonk, Aeschlimann, & Lomperski, 2010) for a description of the
theory underlying constant voltage operation during core debris coolability experiments using the DEH
technique.
16. The resistivity of molten core debris increases as concrete decomposition products from ablation are
introduced into the melt, and as the melt temperature deceases.
NEA/CSNI/R(2016)15
experiments will most likely require a new concept for a test design. For example, in the PERCOLA
simulant material test programme (Tourniaire B. , Seiler, Bonnet, & Amblard, 2000), a novel method
was developed in which a structure simulating an upper crust was counter-balanced in order to create a
floating crust condition in a liquid pool. Challenges with this method include developing a
technological solution for maintaining a liquid state in the gap around the upper structure, and to avoid
any bypass by gases arising from concrete decomposition. An intermediate step could be to design
such a new test facility and to verify the concept using a medium temperature simulant fluid before
attempting to adopt the method for use in reactor material test facilities.
2.6.2. Material issues
The MCCI experiments conducted over several decades presented many materials issues ranging from
the choice of melt composition to the stability of structural materials used in test rigs due to the high
temperatures and corrosive nature of core melts. At the same time, these experiments also provided
significant learning opportunities for designing and conducting successful large-scale experiments
involving prototypic core materials.
Foremost among the material issues surrounding the MCCI experiments is the recognition that
the vast majority of tests with prototypic core materials are limited to oxidic compositions. In contrast,
it is likely to have a mixed melt composition (i.e. oxide and metal) in the reactor case due to the
presence of structural steel incorporated into the melt during the in-vessel phase, as well as the
presence of steel reinforcing bar in the concrete basemats of all plants. Thus, when extrapolating test
results to reactor cases, there should be some consideration of uncertainties regarding melt
composition, particularly as it relates to the presence of structural steel in the melt. The VULCANO
programme has conducted a few experiments with mixed metal-oxide compositions ( (Journeau C. , et
al., 2009), (Journeau, et al., 2012), but data are limited and additional experiments are needed to fully
explore the effect of high metal content on core-concrete interaction behaviour.
From the experimental viewpoint, conducting reactor material experiments with a significant
metal phase presents several challenges, a few of which are outlined below.
1. The prototypic metal phase contains zirconium or even uranium (for specific in vessel
conditions). Oxidation reactions for these metals are very exothermic, which can significantly
reduce the test duration and make the test interpretation more difficult due to the fact that the
chemical reaction power is hard to characterise.
2. Under prototypic conditions the decay heat is mainly located in the oxide phase, and this is
difficult to reproduce given current limitations on heating technology. In the event that the
metal phase stratifies, then a short circuit is created that defeats the direct electrical heating
method. Moreover, for facilities that utilise tungsten as the electrode material (such as the
ANL test facility), the formation of a low-temperature Fe-W eutectic can lead to rapid erosion
of the electrodes. For facilities that utilise the induction heating method (such as the
VULCANO facility), the magnetic screening of the lower pool part where the metal is
expected to be may not be sufficient because the presence of metal droplets in suspension
within the oxide can create a screen that significantly disturbs the power injection in the oxide
phase. As a consequence, the power distribution is not known with a high degree of
confidence.
3. In high temperature experiments, the design of intrusive probes is always difficult but in
metal-oxide systems the probes have to resist both phases, particularly after stratification
occurs. Nevertheless, there are indications that the CEA team has developed appropriate
technology to overcome these difficulties by selecting specific shielding materials. The
thresholds between stratified and homogeneous pools are extrapolated from simulant
experiments, and it was not possible to make online measurements during the VULCANO
tests. As no appropriate dimensionless analysis has been conducted to justify the choice of the
simulant fluid or the form of the correlation to express those thresholds, the design of an
NEA/CSNI/R(2016)15
appropriate probe to make measurements in prototypic material experiments should be

encouraged.
4. In metal-oxide experiment, it is also difficult to characterise from existing instrumentation the
time of change in between stratified and mixed configurations and no specific device has been
implemented.
Material compatibility, i.e. potential adverse interactions between melt and other structural
components of the test facility, is another issue that caused failures in some early MCCI experiments
and provided considerable challenges in designing future experiments. For example, the sidewall
failed prematurely in the MACE test M2 (Farmer, Kilsdonk, & Aeschlimann, 2009) due to chemical
interaction between siliceous melt and magnesia liner on the sidewall causing a spill of molten
material outside the test section and termination of the test. Another MACE test (M3) was also
terminated prematurely due to test section over-pressurisation that was caused by excessive hydrogen
production from the melt mixture during the chemical reaction stage of the experiment used to
produce the core melt. This over-pressurisation was due to moisture ingress into the mixture during
pre-test operations. The experimentalists learnt from these failures and challenges and designed future
experiments (in both the MACE series as well as the NEA MCCI series (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, 2006), (Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010) with a
particular focus on material compatibility issues.
In two recent MCCI experiments, steam line failure has occurred. While these failures did not
warrant test interruption or termination, the functionality of the quench system was compromised as a
result. Upon flooding, the steam production eventually became excessive and masked the test section
because the quench system was not fully functional. Excessive steam also caused moisture ingress into
various instrumentation ports (thermocouple junctions, pressure transducers, etc.) and the data
acquisition system, resulting in temporary loss of data channels and/or spurious recording of data.
Post-test examination of the failed steam line in the first of the two identified experiments
indicated thermal ablation of the flange material (stainless steel). Thus, measures were taken in
designing the second experiment whereby the inside surface of the flange was protected with mullite –
a ceramic material composed of alumina and silicate. Despite the protective measure, the stainless
steel flange in the second experiment experienced thermal ablation – as intense as in the first
experiment. Moreover, the post-test examination of the flange revealed no trace of the mullite liner
indicating perhaps its dissolution due to chemical interaction. This recent experience suggests that
additional care should be exercised in designing the steam line, quench system, and other components
of the integral test rig so as not to produce adverse consequences during test operation.
2.6.3. Results of MCCI experiments and their significance
The various test series summarised earlier in this section principally investigated the effects of melt
composition, concrete type and input power on two-dimensional core-concrete interaction under both
wet and dry cavity conditions. Principal variables measured during the experiments included melt
temperature and local concrete ablation rates. For flooded cavity experiments, the debris/water heat
flux after cavity flooding was also estimated based on the rate of steam production from the
interaction. Key observations from the tests are summarised below.
Under dry cavity conditions, all tests exhibit the overall trend of decreasing melt temperature as
ablation progressed, which was due to a heat sink effect as relatively cool concrete slag was
introduced into the melt, as well as the increasing heat transfer surface area as the melts expanded into
the concrete crucibles. The decline in melt temperature may further reflect the evolution of the pool
NEA/CSNI/R(2016)15
boundary 17 freezing temperature that decreased as additional concrete was eroded into the melt during
the tests.
The effect of unoxidised Zr cladding on the thermalhydraulics of the core-concrete interaction
was investigated in several experimental programmes. The oxidation reaction between Zr and sparging
concrete decomposition gases (CO2, H2O) caused exothermic transients in which the melt
temperatures increased by roughly 100 ºC over a period of tens of minutes in the experiments. This
transient behaviour was observed in both reactor material as well as simulant experiment tests. The
data further indicate that after the cladding is fully oxidised melt temperature drops to that consistent
with fully oxidised melt conditions.
Aside from cladding oxidation effects, a limited number of experiments have been conducted to
examine the effect of significant structural steel content in the melt on core-concrete interaction
behaviour. One surprising outcome from these tests is the extensive amount of iron oxidation that
occurs with limestone-common sand concrete. Steel oxidation also occurred in tests with siliceous
concrete, but to a lesser extent in comparison to the limestone case. The most surprising result from
tests with siliceous concrete is that spatial segregation occurred, but it was not completely driven by
gravity. In particular, steel layers were found on both horizontal and vertical surfaces. Solidified steel
were also found in the oxide. Oxidic melt and unmelted concrete aggregate were also trapped in some
parts of the metal layer. Concrete temperatures showed that axial and radial ablation was more
pronounced in the areas where metal was found. This is consistent with other oxide-metal simulant
tests that have shown enhanced heat transfer at the metal-concrete interface relative to the oxide-
concrete interface.
Despite these findings, the metal-oxide database is noted to be quite limited and there is not a
clear understanding of the phenomenological behaviour for this case (i.e. mixed vs. stratified metal-
oxide conditions, and/or bifurcation between these two states during the interaction). Therefore,
extrapolation of results obtained thus far to plant conditions is uncertain. The importance of these
uncertainties depends on the reactor design. For example, sensitivity analyses in French PWR900
calculations show that the impact of these uncertainties on the overall melt-through time is
manageable. For 3-4 m basemat thickness, the duration of such stratified configurations are very
limited in time (less than 10 hours) compared to the duration of ablation (several days) (Atkhen &
Spindler, 2010) (Cranga, Michel, Mun, & Barrachin, 2010).
Several experimental programmes have also provided evidence that initial crust formation on
cold concrete surfaces can influence the early (tens of minutes to ~hour) core-concrete interaction
behaviour. During this phase, basemat ablation is minimal and melt temperature remains high due to
the insulating effect of the crusts. This effect has been observed in both transient as well as sustained
heating reactor material tests. However, once surface crusts fail ablation proceeds vigorously and the
melt temperature declines due to the above mentioned affects. Although the data are not conclusive,
there is evidence that gas evolution from concrete decomposition can act to destabilise these interfacial
crusts. This is due to the fact that the duration of this crust stability period has been found to be
inversely proportional to concrete gas content in several experiments.
Aside from initial crusting effects, the data from several reactor material experimental
programmes indicates that the long-term ablation process for oxidic core melts is influenced by
concrete type. The results indicate that for high gas content concretes such as limestone-common sand,
17. If the presence of a crust at melt concrete interface is still an open issue, the freezing temperature between
melt and upper crust boundary will decrease as a function of concrete fraction whatever the hypothesis is
considered between liquidus temperature and temperature corresponding to a given solid fraction close to
a mobility threshold usually around 50 to 60¨%.
NEA/CSNI/R(2016)15
the radial/axial power split is ~ 1. Conversely, for low-gas content concretes such as siliceous, this
ratio is ~ 3. Forensic examinations at Chernobyl Unit 4 are also consistent with the experiment
observations for siliceous concrete. This overall trend in the ablation front progression that has been
observed under experiment as well as prototypic conditions cannot be explained on the basis of our
current understanding of these types of processes. Thus, extrapolation of the results to plant conditions
is currently uncertain due to the lack of a phenomenological model that can rationalise the differences
in the observed cavity erosion behaviour.
Post-test examinations in the experimental programmes has shown that the nature of the core-
concrete interface is noticeably different for limestone tests in comparison to siliceous; i.e. the ablation
front for the siliceous concrete tests consists of a region where the core oxide had locally displaced the
cement that bonded the aggregate. Conversely, for tests with limestone concrete, a powdery interface
exists in which the core and concrete oxides are clearly separated. The interface characteristics may
influence the heat transfer at the interface, yielding different ablation behaviour for the two concrete
types.
One question that has been raised is whether the relatively small melt pool aspect ratio (viz., test
section width/melt depth) used in the experiments has an influence on the radial/axial power split
observed in the dry cavity experiments. A dedicated large scale experiment was carried out to
investigate this effect. The results indicated that an increase in aspect ratio from ~ 1 (typical of most
reactor material tests) to ~ 3.2 has no noticeable effect on ablation characteristics for siliceous
concrete. This observation lends additional credibility to the measured power splits in various
experimental programmes. Finally (and as noted previously), the forensics examinations regarding
power split from Chernobyl are consistent with the test data for low-gas content siliceous concrete.
Aside from overall cavity erosion behaviour, video data indicates that a crust is typically present
over the melt surface during a large fraction of dry cavity ablation. The crusts contain vent openings
that allowed eruptions to occur as the tests progressed. The frequency and intensity of the eruptions
was directly correlated to the concrete gas content.
In terms of the chemical analyses conducted as part of the test programs, samples were generally
collected to: i) characterise the lateral and axial composition variations of the solidified debris, and
ii) characterise the composition of corium regions that played key roles in the coolability aspects of the
tests (e.g. porous crust zones formed at the melt/water interface, and the material erupted after cavity
flooding in CCI-2). Analysis of samples taken to characterise the lateral composition variation indicate
that for most tests, the corium in the central region of the debris had a higher concentration of core
oxides in comparison to samples collected near the two ablating concrete sidewalls. Conversely,
samples taken to characterise the axial composition variation indicated the general trend of slightly
increasing core oxide concentration as the concrete surface is approached. For tests conducted with
siliceous concrete, two zones were typically present: a heavy monolithic oxide phase immediately over
the basemat that was enriched in core oxides, with a second overlying porous, light oxide phase that
was enriched in concrete oxides.
Regarding fission product release during MCCI, the composition of aerosols produced has been
found to be different from that of aerosols generated under in-vessel conditions, which is likely
attributable to the addition of compounds of sodium, potassium, magnesium and calcium arising from
concrete erosion. The phenomenology leading to aerosol generation is complex; the mechanisms
include evaporation into gas bubbles generated from concrete decomposition (evolved gas species
include H2O/H2 and CO2/CO) and at the surface of the corium-concrete mixture, as well as by
mechanical generation of droplets due to sparging.
Sustained heating tests conducted with reactor materials and non-radioactive mockups of various
fission product species (Fink, Thompson, Armstrong, Spencer, & Sehgal, 1995) indicate that aerosols
contain mainly constituents of the concrete. In the tests with metal and limestone/sand siliceous
NEA/CSNI/R(2016)15
concrete, silicon compounds comprised 50% or more of the aerosol mass. Releases of tellurium and
neutron-absorber materials – silver, indium and boron (from boron carbide) – were high. Releases of
uranium and low-volatility FP elements were small in all tests. During ablation of the concrete, aerosol
composition remained fairly stable and particles were compact but varied considerable in size, the
majority being typically micron-sized (geometric diameter) but with some considerably larger sizes.
Regarding debris coolability, the test series have provided evidence of several heat transfer
mechanisms that can contribute to long-term corium cooling. When the core debris is flooded from
above, the question of whether or not a significant amount of the thermal energy will initially be
removed depends upon whether a stable crust is able to form that inhibits heat transfer from the melt to
the water layer. For a stable crust to form, two conditions must be met: i) a thermal condition, viz., the
melt/water interfacial temperature must fall below the corium freezing temperature, and ii) a
imposed by the agitated melt. If either of these two conditions is violated, then stable crust formation
is precluded. In this bulk cooling regime, efficient melt/water heat transfer occurs due to conduction
and, predominately, radiation heat transfer across the agitated (i.e. area enhanced) melt/water interface,
in addition to entrainment of melt droplets into the overlying water.
decreases. Thus, a point will eventually be reached at which the thermal and mechanical thresholds for
zone and water layer. The crust will be characterised by some degree of porosity, or cracks, owing to
the necessity of venting concrete decomposition gases.
After the crust forms, completion of the quench process can only be achieved if one of two
conditions is met. The first condition is that the melt depth lies below the minimum depth at which
decay heat can be removed via conduction heat transfer alone (viz. ~10 cm based on a typical PWR
decay heat level two hours after scram). This case is trivial, and is not addressed further. The second
condition is that water is able to penetrate into the debris by some mechanism to provide sufficient
augmentation to the otherwise conduction-limited heat transfer process to remove the decay heat. The
tests have revealed three mechanisms that provide pathways for water to penetrate the debris. The first
is water ingression through interconnected porosity or cracks. This process relies on crack propagation
through the material and, as such, is highly dependent upon the mechanical properties, since thermal
stress is a key factor. The second mechanism is particle bed formation through “volcanic” eruptions.
In this case, concrete decomposition gases entrain melt droplets into the overlying coolant as they pass
through the crust. The entrained droplets then solidify in the overlying coolant and accumulate as a
porous particle bed atop the crust. The third mechanism is mechanical breach of a suspended crust. In
particular, the thick crusts that form from water ingression could bond to the reactor cavity walls,
eventually causing the melt to separate from the crust as the MCCI continues downwards. However,
this configuration is not expected to be mechanically stable due to the poor mechanical strength of the
crust in comparison to the applied loads (i.e. the crust weight itself, plus the weights of the overlying
water pool and the accumulating dispersed material). Eventually the suspended crust will fail, leading
to rapid ingression of water beneath the crust. This sudden introduction of water provides a pathway
for renewed debris cooling by the bulk cooling, water ingression, and melt eruption cooling
mechanisms.
Several findings related to debris coolability are directly applicable to evaluating plant accident
sequences. In terms of the water ingression mechanism, the test results indicate that the heat transfer
correlation based on previous one-dimensional (SSWICS) tests is conservative insofar as calculating
ingression-limited crust growth behaviour. In particular, the correlation tends to under-predict the heat
flux to overlying water during time intervals when water ingression is occurring. In terms of the melt
eruption mechanism, significant eruptions have been observed in the case of limestone common sand
NEA/CSNI/R(2016)15
concrete. Eruptions have also been observed under early cavity flooding conditions for siliceous
concrete. For tests conducted in 2D cavity configurations, the eruptions appear to have occurred under
a floating crust boundary condition, which is expected at plant scale. Finally, the crust breach data
indicate that the crusts that form at the melt/water interface after cavity flooding are quite weak, and
will not be mechanically stable at plant scale. Rather, the crust is expected to fail and, thereby,
maintain a floating crust boundary condition that will allow the melt eruption and water ingression
cooling mechanisms to proceed.
Note that the debris coolability experiment database consists almost exclusively of tests
conducted with oxidic melts. As noted earlier, a significant melt metal fraction may be present that
may result in a stratified pool configuration. This type of pool structure has not been evaluated from a
coolability standpoint. Thus, additional analysis and testing may be required with melts containing a
significant metal fraction to further reduce phenomenological uncertainties related to debris
coolability, as well as core-concrete interaction.
NEA/CSNI/R(2016)15
3. MAJOR MCCI SIMULATION TOOLS AND MODELS
Major simulation tools currently used in the world are described focusing on models currently used to
describe the main phenomena involved. Considered codes are by alphabetic order: CORCON,
CORQUENCH, COSACO, MEDICIS (GRS), MEDICIS (IRSN), TOLBIAC-ICB, WECHSL, COCO,
MAAP, and SOCRAT (MCCI).
3.1. Modelling approach
Why is the lumped parameter approach currently used instead of more recent and possibly more
powerful approaches such as CFD, DNS? Answers to this question are multiple and can be listed as
follows:
As far as heat convection is concerned, no adequate and validated modelling of multiphase
thermal-hydraulics of a pool submitted to gas injection including mass, momentum and energy balance
equations is available in a 2D configuration, on the main reasons is that no well-established and
validated constitutive laws, determining conditions along boundary pool interfaces in case of gas
injection imposed at these interfaces are not available in particular in case of a flow turbulent regime
not to speak of the configurations with two immiscible liquid phases; an attempt performed at IRSN to
recalculate CLARA simulant experiments on 2D convection driven by gas bubbling remained so far
unsuccessful;
The real structure of a pool/concrete interface is only known very approximately and likely to be
linked to instability of crusts and boundary layers and depending on the pool scale; a detailed
description of these pool interfaces coupled with the pool thermal-hydraulics would be a huge task
which was never reached even by very far;
The strong coupling for a corium/concrete mixture of thermal-hydraulics, local corium
composition, affecting transport properties and freezing temperatures (Seiler & Froment, 2000),
increases further the task complexity;
The outcome of the recently performed R&D (Cranga, et al., Towards an European consensus on
possible causes of MCCI ablation anisotropy in oxidic pool, 2014) showed that the impact of the
structure of pool/concrete interfaces on 2D ablation is prevailing on those of the detailed 2D
convection mechanisms within the corium pool;
Last but not least, it is not sure that such a complex modelling is really required to perform
reactor predictions, since after at most a few hours, the use of an energy balance equation in a quasi-
state regime is reasonable and permits to evaluate the concrete ablation kinetics, provided that the heat
flux profile along the pool interfaces can be evaluated using simplified assumptions and models
validated against the experimental database.
Therefore the use a lumped parameter approach (Figure 3.1-1) based on the averaging of physical
quantities over the whole pool or separated layers is preferred and seems to be both easier and more
reliable. Such an approach requires knowing convective heat transfer coefficients for describing heat
transfer from bulk pool to interfaces and also between layers and detailed heat transfer mechanisms or
at least effective heat transfer coefficients for describing heat transfer across the pool interfaces.
NEA/CSNI/R(2016)15
1 convection pool bulk => interfaces

heat transfers in series
2 heat transfer across corium/concrete interfaces
3 heat transfer to upper interface
radiation
upper oxide crust concrete
slag
slag layer
layer
upper hconv liquid/gas)
(liquid/gas)
oxide layer
lateral hconv in oxide hslag (θ) layer

boundary
h oxide/metal
possibly present
axial hconv
metal layer
Oxide or metal crust at
corium/concrete interface
4 convection between layers within the pool
Figure 3.1-1: Application of the lumped parameter approach to a MCCI situation
SOCRAT code is an intermediate case: this code uses a finite element approach for determining
2D heat transfer within the pool but does not calculate the fluid velocity and flow pattern, which is
much less ambitious than CFD simulation tools (see Section 3.2.8).
3.2. Overall description of codes
Here are indicated main features for each code. Description of basic assumptions of lumped parameter
MCCI codes is further detailed in appendix 7.1. The reader is warned that the description of each code
is not exhaustive and only major features and specific models are addressed below.
3.2.1. COCO
Major phenomena that are modelled in the COCO, code developed by NUPEC and JNES are
illustrated in Figure 3.2-1. In the COCO code, the behaviour of molten fuel is described by a lumped
parameter model (0D model) and the transient heat conduction in concrete wall and basemat is
expressed on axisymmetric two-dimensional coordinate system coupled with the release models of
free water, bound water and carbon dioxide.
Pool configurations treated
The melt pool is either a homogeneous mixture or stratified layers of oxide and metal. The melt pool
may be surrounded by a top crust and melt/concrete-interface crust. There is a slag layer between crust
and concrete surface. The existence and thickness of crust are decided by the temperature field and the
phase diagram.
NEA/CSNI/R(2016)15
Heat transfers
Energy equations for debris temperature transient and heat conduction within concrete structure are
solved by the fully implicit finite difference scheme.
Correlations and models have been incorporated for solidification of melt to form crust, water
ingression into crust, melt eruption, boiling and radiation heat transfer from debris upper surface.
Boiling & Radiation

Melt Eruption at Crust Surface
Water Ingression Water

into the Crust Concrete
Crust Heat Conduction

in Concrete
Solidification of Debris
to Form Crust
Melt
Convective Heat
Transfer from Melt
Concrete Ablation
& Gas Release
Figure 3.2-1: Major phenomena modelled in COCO

Erosion algorithm
The ablation velocity is calculated using the latent heat of concrete melting, the heat flux to concrete
surface from the melt and the conductive heat flux to concrete inward. The heat flux to concrete
surface from the melt is calculated from the convective heat transfer from melt to crust/melt interface
and the heat conduction in crust (if there exists) and slag.
Region I
Single Volume Radial
Direction
Melt
Region II Region III

Radial
Axial Direction
Direction
Figure 3.2-2: Assumptions for direction of heat conduction and ablation
NEA/CSNI/R(2016)15
The ablations are assumed in one-dimensional. The basemat is ablated vertically, and for the
sidewall, the direction is radial. The lower sidewall (Region III in Figure 3.2-2) ablation is started after
the basemat ablation front reaches its level.
3.2.2. CORCON
CORCON-Mod3 (Bradley, Gardner, Brockmann, & Griffith, 1993) is a mechanistic computer code
that describes the core-concrete interaction phenomena relevant to the assessment of containment
failure and radionuclide release. The principal components of the CORCON system are the core
debris, the concrete, and the atmosphere and surroundings above the debris. The code also treats an
overlying coolant pool if one is present. The physical system considered by the code consists of an
axisymmetric concrete cavity containing debris in one or more layers. The code calculates heat
transfer rates from the debris to the concrete and to the top surface of the debris, as well as the heat
transfer between layers. The code also calculates concrete ablation, contingent generation of non-
condensable gases from core-concrete interaction, and consequent production of fission products.
Figure 3.2-3: Example of melt-pool configuration Figure 3.2-4: Cavity surface computation in
in CORCON code CORCON code
CORCON-Mod3 is integrated within the cavity (CAV) package of the integrated system-level
code, MELCOR, which models progression of severe accidents (i.e. accidents resulting in severe core
damage, possibly melting of the core, leading to release of radioactivity) in nuclear power reactors. By
default, the CAV package considers all debris, metallic and oxidic, to be mixed into a single layer of
homogeneous melt pool. However, there is option to model multiple layers in a molten pool. Boundary
conditions for the top surface of the debris are obtained from an associated control volume in the
Control Volume Hydrodynamics (CVH) package, which also serves as a sink for heat and gases
released during the interaction. If there is a water pool in the associated volume, it is assumed to
overlie the debris. A schematic of the CORCON melt pool configuration is shown in Figure 3.2-3.
The shape of the cross-section of the cavity is defined for computational purposes by a series of
“body points.” These are the intersections of a fixed series of rays with the cavity surface as shown in
Figure 3.2-4. Given the cavity geometry at the start of a time step, the shape change procedure
provides a new cavity shape at the end of the time step. The normal recession rate is used at each body
point, as calculated by the concrete ablation model.
NEA/CSNI/R(2016)15
3.2.3. CORIUM-2D
CORIUM-2D [(Parrozi & Fontana, 2013), (Parozzi, 2010), (Polidoro, 2013)] is a fast running code,
developed at RSE, aimed at describing heat transfer phenomena among corium, confinement
structures, and coolant under given LWR severe accident conditions or prototypical experiments. The
code is based on mass and energy balances; momentum transfer is not taken into consideration, while
liquid corium convection is considered, even with some simplifications.
The simulation is performed dynamically, using a modified Euler's predictor-corrector integration
method. The model calculates the corium and structures temperature field accounting for solid-to-solid
conduction, liquid corium convection, coolant convection and boiling, radiative transfer among dry
corium and structures and phase changes in corium and structural materials.
The main key assumptions in the code are:
• Solution of energy equations only, which means to assume:
• heat exchange driven by solid crusts;
• corium molten pool almost isothermal because of convection;
• fluid-dynamic analysis of the liquid pool not reliable because of non-
homogeneous properties;
• 2D dynamic analysis (Cartesian or axisymmetric);

• Arbitrary number of cells, not equally-spaced nodalisation;
• Corium, structural material and coolant properties self-calculated as a function of
temperature;
• Corium cells fixed-corium mass allowed to expand or collapse (confinement assumed at the
bottom and side walls);
• Levelling of corium pool.
The heat transfer model of molten corium convection is based on Fieg's and Kulacki-Goldstein's
experimental observations (Fieg, 1978), (Kulacki & Goldstein, 1972)] on internally heated liquids and
it is described by the following relations between Nusselt (Nu) and Rayleigh (Ra) numbers:
For upward facing surfaces Nu = 0.414 Ra 0.216 Equation 3.2-1
For downward facing surfaces Nu = 1.12 Ra 0.103 Equation 3.2-2
For lateral surfaces Nu = 0.163 Ra 0.244 Equation 3.2-3

within the range 107< Ra < 4•109.
Heat transfer through liquid-solid interfaces is calculated accounting pure heat conduction
through the related boundary layers. Heat transfer within the liquid bulk is evaluated through a
characteristic fluid velocity moduli u’ (natural and forced convection velocity fields assumed to be
comparable); this characteristic velocity u’ in a molten corium having pool height H, viscosity µ and
density ρ, at steady state is:
Reµ µ
u∞ = = 111.65 Pr −0.375 Nu1.25 Equation 3.2-4
ρH ρH
Re and Pr being the Reynolds and Prandtl numbers.
NEA/CSNI/R(2016)15
As regards the liquid levelling, a reasonable estimate of the mean velocity of the transferred mass
may be obtained by imposing that all the potential energy is converted into kinetic energy; to account
for energy dissipation by viscosity in turbulent regime, a small correction is then introduced.
Corium can be described as a debris bed, i.e. as a pack of solid particles arranged in a volume
filled with a motionless gas. The heat transfer mechanisms considered in the code are:
• Thermal conduction through gas;
• Radiant heat transfer between adjoining voids;
• Thermal conduction through solid;
• Thermal conduction through the gas film near the contact surface of two particles, radiant
heat transfer between solid surfaces.
To calculate the effective thermal conductivity ke of the bulk discontinuous material, CORIUM-
2D takes into account the contribution of heat transfer through the gaps and the solid material, as:
∆T
− ke = qgas + qs Equation 3.2-5
lp
lp is the average distance between the centres of two adjoining solid particles, ΔT the temperature
difference between the centres of two neighbouring particles, qgas and qs thermal fluxes through the
void fraction and solid phase.
The radiative heat transfer is computed by Monte-Carlo method accounting real 3D geometry and
inclination of enclosures, and re-iterated to account relocation of molten materials.
A complete library of corium and structural materials (in the oxidic and metallic phases) is
available for severe accident analyses of LWR reactors. The code description of system including
concrete is limited by the lack of models for the chemistry of corium-concrete interaction (loss of
concrete water, decarbonatation, oxidation of metals, etc.). For this reason, only cases implying a
small concrete ablation, if compared to the whole corium mass, can be reasonably described
(e.g. simulation of reactor cavity walls, thin concrete protective layers, etc.). In this light, as it is likely
to have a low influence on the overall system behaviour, no temperature dependence is assumed for
the concrete properties.
3.2.4. CORQUENCH
CORQUENCH, (Farmer, 2001), (Farmer, 2010), was originally developed to provide a simple,
modular model of MCCI behaviour that could readily be adapted to investigate the adequacy of
melt/water heat correlations as they were developed. However, it is now being used more broadly for
reactor applications.
The MCCI model within CORQUENCH can perform either a 1D or simplified 2D ablation
calculation. The 2D geometry can be selected to be either cylindrical or rectilinear, with average axial
and radial ablation depths calculated. Thus, the code does not perform a detailed calculation of the
spatial erosion pattern, but rather simplified geometries are calculated. The melt composition can
range from fully metallic to fully oxidic; in all cases, the two phases are assumed to be well mixed
(i.e. phase stratification is not modelled).
Erosion algorithm
In terms of heat transfer at the melt/concrete interface, a transient concrete decomposition model based
on integral thermal boundary layer theory is incorporated (Corradini, 1983). This model was upgraded
to account for the effects of transient concrete heat-up with simultaneous crust growth following initial
NEA/CSNI/R(2016)15
melt contact with the concrete. Sketches of the physical situations modelled in CORQUENCH are
provided in Figure 3.2-5. The inclusion of a concrete dry-out model can be important in evaluating
both the early and late phases of a core-concrete interaction. In the early phase, transient crust
formation can affect the timing of onset of ablation, while in the late phase of experiments with a
limited basemat thickness, heat transfer to underlying concrete can fall to low levels as the decay heat
decreases, and so conduction into the concrete behind the ablation front becomes important in
determining the overall ablation depth.
Although CORQUENCH has fairly refined concrete decomposition models, they are all one-
dimensional; no 2D. The calculated concrete response is based on 1D into the concrete; no 2D heat
conduction effects parallel to the ablating concrete surface are considered.
Heat transfer
The heat transfer coefficient at the melt/concrete interface can be selected from a variety of user
options. The code is capable of calculating transient crust formation, growth and failure at the core
concrete interface; this effect is important in the early stages of the core-concrete interaction when
crusts can result in an initial incubation period where ablation rates are suppressed. At the melt upper
surface, radiant heat transfer to overlying structure is calculated when the cavity is dry. When water is
present, a detailed heat transfer analysis is performed involving bulk cooling and follow-on transient
crust growth phases; the crust can be treated as impervious to water or coolable by water ingression.
Melt eruptions are also treated.
Figure 3.2-5: Assumed core/concrete interface structure when slag continuously drains through
crust (left) and when slag is retained beneath crust (right)
Energy balance
The conservation of energy equation includes the following energy source/sink terms: i) decay heat,
ii) mass flux of melt from the failed reactor pressure vessel (RPV), iii) chemical reactions between
metallic melt constituents Zr, Cr, Fe (in sequence) and concrete decomposition gases H2O and CO2,
iv) condensed phase chemical reactions between Zr and SiO2, v) downward (and sideward for 2D
case) heat transfer to concrete, including slag heat sink, and vi) heat transfer to overlying atmosphere
(wet or dry).
NEA/CSNI/R(2016)15
Code peculiarities
CORQUENCH is a simple code to run and is flexible; all user options can easily be selected through
user index options. The code produces text as well as graphical output files so key output results can
easily be graphed for evaluation.
3.2.5. COSACO
The MCCI code COSACO (Nie, 2004) has been developed with respect to the MCCI processes
foreseen in the core melt stabilisation concept of the EPR™. This concept includes a temporary melt
retention phase realised by sacrificial concrete in the reactor pit. During this phase all the corium will
be collected, independently of the accident scenario. For its long term stabilisation, the melt is spread
on the large surface of a core catcher within a water cooled crucible. To protect this crucible during the
corium spreading inside the core catcher its surface is covered with sacrificial concrete that will be
eroded.
To model the MCCI process, the COSACO code follows an energy-based approach. The specific
modelling of the thermo-chemical phenomena in the melt pool addresses the prevailing behaviour of
the considered multicomponent melts. The real solution database COSCHEM is integrated into the
modelling logic of COSACO by employing the equilibrium solver ChemApp (CHEMAPP, 1999). The
data stored in COSCHEM essentially bases on the real solution database TDBCR99 (Chevalier, 1999),
which was developed and validated e.g. in the CIT project of the EU’s fourth framework programme.
The MCCI is therefore modelled by a coherent description of the thermo-chemical behaviour of the
melt pool and its chemical reactions.
The assumed pool configuration of COSACO is oriented towards the EPR™ reactor pit. An axial
symmetric 2D geometry as sketched in Figure 3.2-6(a) is assumed. Other geometric pool
configurations, as e.g. the EPR™ core catcher, can be approximately defined via an angular sector
with an inert inner disc, as sketched in Figure 3.2-6(b).
The melt pool is typically composed of oxidic and metallic corium. In COSACO, these
constituents may either be in a layered mode or fully mixed. In the mixed mode a homogenous melt
pool due to the intense mixing by the rising MCCI gases is assumed. On the other hand, due to the
mutual immiscibility of the metallic and oxidic melt components, layered melt configurations can also
be assumed. For this layered mode, depending on the density of the layers two configurations are
distinguished. As long as the amount of ablated concrete oxides in the oxide melt layer is low, the
density of the oxidic melt is higher that of the metallic melt. Thus the oxide layer is located at the
bottom and covered by the metallic layer. On top of the metal, a light slag layer consisting mainly of
light oxides is considered. Later, when the density of the oxidic layer has sufficiently decreased due to
concrete addition, the density of the oxidic layer falls below that of the metallic layer. This yields a
layer inversion after which the oxide melt is located above the metallic layer. The slag layer from the
top is then mixed into the oxide melt.
Due to the gases released during the concrete erosion a certain amount of void establishes in the
melt. This yields an increase of the melt volume that corresponds to an increase of the melt layer
thickness. The so time dependent upper melt surface is considered in COSACO by the axial
discretisation of the lateral wall.
NEA/CSNI/R(2016)15
Refractory Layer
RAOC
Concrete
HECR
CDR2
R1
RAIC
CDBA
CD
ANCR
a) b)
Figure 3.2-6: Sketch of the COSACO geometry model
Erosion algorithm
Concrete erosion is determined using a quasi-static approach without any heat conduction in the
concrete. To calculate cavity profiles, an axial discretisation of the lateral walls is used. The locally
eroded concrete thickness is directly determined by the heat flux from the melt and the decomposition
enthalpy. Due to phase segregation, crusts of refractory material can be formed at the boundaries of
the oxidic melt layer. These crusts add an additional thermal resistance for the heat transfer to the
concrete.
Heat transfers
For oxidic or mixed melt pools, the heat transfer coefficients from the melt bulk to the surface of
another layer or to the crust covering the concrete surface are obtained by a correlation based on the
BALI tests (Bonnet, 1999). For the radial heat transfer, which was not investigated in the BALI tests,
the same correlation is used. Furthermore, if a crust exists at the interface with the concrete or another
layer, the effective heat transfer coefficient is determined by a series connection of the individual
thermal resistances.
In layered mode, the Kutateladze-Malenkov correlation (Kutateladze & Malenkov, 1978) is used
for the heat transfer from the metallic melt layer.
The heat transfer between two melt layers is determined by adding the individual thermal
resistances between the layers (connection in series), including that of the crust, if any, see 7.1.5.2.
Energy balance
The energy balance for the corium pool considers the decay heat, the sensible heat of the eroded
concrete, the heat transferred to other layers, the radiation from the top surface as well as the enthalpy
flow associated to the addition of the decomposed concrete and the exchange of various constituents
between layers. In the layered melt configurations, the decay heat is distributed between oxidic and
metallic layers. The energy involved in chemical reactions and phase transitions is internally
considered by the real solution database.
Code peculiarities
COSACO includes a coherent description of the thermo-chemical behaviour of the melt. Any chemical
reactions and phase transitions that occur during the ongoing MCCI are taken into account via the use
NEA/CSNI/R(2016)15
of a real solution data base in combination with the free enthalpy minimiser ChemApp (CHEMAPP,
1999).
3.2.6. MAAP
The Modular Accident Analysis Program (MAAP) is a widely used severe accident code owned by the
Electric Power Research Institute (EPRI). The current released version is MAAP5.02 (MAAP, 2011),
and MAAP5.03 is expected to be released in the summer of 2014. MAAP can simulate most all major
phenomena that might occur during the course of a severe accident, including the ex-vessel
phenomenon of molten core concrete interaction.
The MCCI model in MAAP considers that the reactor cavity contains corium as a
homogeneously mixed pool. It assumes material stratification is negligible even when appreciable
concrete ablation occurs. The corium pool is modelled as having a molten centre surrounded by crusts
at the bottom, along the side, and at the top. The mass and energy of the upper crust are tracked
separately from the remaining corium pool. This is done so that the upper crust can be quenched to a
temperature significantly lower than the remaining pool.
The model uses a simplified pseudo-binary phase diagram to determine thermal properties and
solid fraction within the pool. All constituents in the pool are grouped into two major phases: metals
and oxides. In each major phase, the constituents are further divided into sub-groups. Phase diagrams
are constructed for pairs of the sub-groups. Chemical reactions during MCCI are modelled assuming
instantaneous chemical equilibrium among all the reactants (constituents) in the corium pool at each
time step.
The corium pool interacts with several one-dimensional concrete heat sinks. In MAAP5.02 and
earlier versions, these heat sinks are the floor, the sidewall, and the upper wall (above the corium
pool). In MAAP5.03, the floor and sidewall of a sump in the cavity floor can also be modelled.
Therefore, in MAAP5.03 there are potentially 5 heat sinks interacting with the corium pool. If the heat
fluxes from the corium pool exceed the amount that can be conducted away from the surfaces of these
concrete heat sinks, the concrete is ablated. Ablation rates are proportional to the net heat fluxes
deposited onto the surfaces. The ablated concrete releases its constituents in gas and liquid forms. Gas
may enter or escape from the corium pool. Liquid is assumed to enter the corium pool immediately.
The surface area of the corium pool increases as the sidewall ablation continues.
The top surface of the corium pool can radiate to the upper portion of the vertical concrete wall if
the cavity is dry or transfer heat to water if the cavity is flooded.
Erosion algorithm
The corium pool is represented as a cylinder with its bottom and side surfaces in contact with concrete.
Concrete can be ablated if the heat fluxes from the bottom and side of the corium pool exceed what
can be conducted away from the concrete surface. As the ablation of concrete continues, the bottom
contact area is expanded, and the side contact area is calculated based on the volume of the corium-
concrete slag mixture. In MAAP5.03, more corium pool geometry options are available to users
besides the default cylindrical shape. One option is a sophisticated ablation front model which allows
evolvement from the original cylindrical shape to an axisymmetric curved shape. The second option is
a rectangular corium pool with two or four side surfaces ablating. This feature allows more accurate
representation of the corium pool geometry in a BWR Mark-I drywell and certain PWR cavities. The
third option allows a sump to be modelled. The shape of the sump can be cylindrical or rectangular. In
this case, the corium can contact and ablate the floor and sidewall in the cavity as well as the floor and
sidewall in the sump. The contact areas of the ablating surfaces are calculated based on the volume of
the corium mixture and the covered fraction of each surface.
NEA/CSNI/R(2016)15
The corium coolability models in MAAP include a water ingression model, a melt eruption
model, and a three-region corium pool model. In the three-region corium pool model, the corium pool
is represented by a particle bed, a solidified upper crust, and a molten pool. The water ingression
model has two options. The first option is a parametric model, which relies on user input to account
for uncertainties in the heat flux from the upper crust to water. The second option is a mechanistic
model in which the heat flux is based on models of crack extension, boiling, and dry-out in the heat
generating upper crust. The mechanistic model is now the default option in MAAP5.03.
3.2.7. MEDICIS
A lumped parameter approach based on a layer-averaged description is retained in MEDICIS code
developed by IRSN and GRS since ten years to be used as MCCI module in ASTEC integral code and
COCOSYS code, (Cranga, Fabianelli, Jacq, Barrachin, & Duval, 2005), (Cranga, Mun, Michel, Duval,
& Barrachin, 2008), (Duval & Cranga, 2008), (Cranga, Mun, & Marchetto, 2010), (Cranga, Michel,
Mun, & Barrachin, 2010), (Guillard, et al., 2010). Moreover for each layer, mass balance equations are
written per chemical element and the energy balance equation uses enthalpies of element mixtures,
which allow to solve mass and energy balance equations independently of the detailed corium
chemistry evolution.

The melt pool may be either homogeneous or stratified and may evolve versus time. Four layer types
are possible in the present version: an oxide layer, a mixed oxide/metal layer in the case of a
homogeneous pool, a metallic layer and the upper crust built-up at the pool upper interface.
hrad
hrad concrete
crust
top(xA*)
hconv top
oxide hconv (xA*)
hconv(θ)
hconv (θ) in hslag
hslag (θ)
oxide
h ox/met)
hGreen
o/m xhconv
F) (θ)
metal hconv (θ) crust
in metal
Figure 3.2-7: Node displacement at the ablation Figure 3.2-8: Sketch of pool configuration
front and of heat transfer mechanisms described
in MEDICIS
Erosion algorithm
The cavity geometry is either axisymmetric or is slab-shaped with two non-ablatable walls. The cavity
erosion algorithm assumes that the cavity boundary is a succession of truncated cones in case of an
axisymmetric geometry or a succession of prisms in case of slab geometry. The 2D profile of the
cavity boundary versus time is determined using the local energy conservation at each boundary node
and the Stefan’s relation to evaluate the ablation velocity and assuming that the ablation front moves
along the bissectrix of the angle made of the current node and the two adjacent nodes (see Figure
3.2-7).
NEA/CSNI/R(2016)15
Heat transfers
Heat transfers described by the code (see Figure 3.2-8) are the 2D heat convection described by a
convective heat transfer coefficient correlation within each pool layer from bulk pool to the pool
interfaces (per default correlation fitted on BALI data (Bonnet, 2000), the convective heat transfer
between oxide and metal layers described by a convective heat transfer correlation (Greene’s
correlation (Greene & Irvine, Heat transfer between stratified immiscible liquid layers driven by gas
bubbling across the interface, 1988) taking into account the build-up of a crust at the oxide/metal layer
interface and the heat transfer close to the pool/concrete interface determined by the pool/concrete
interface structure described below
Energy balance is written as:
d(mh) dmabl
=P decay −Pupwards + .h conc (T0 ) Equation 3.2-6
dt dt
where: m the mass of the corium pool, h the corium enthalpy, P the decay power, P the
dm abl ∫ φ .dS decay upwards
upwards heat loss rate, = is the ablated mass rate, h ( T ) the enthalpy of ablated concrete
at temperature T, T the dt ∆habl
conc
ablation
abl
temperature and T the ambient temperature, φ is the local heat flux
0
at the corium/concrete interface, ∫ φ .dS is the integral of this heat flux along the concrete ablation
boundary and ∆h = (h (T ) − h (T )) is the concrete enthalpy variation at ablation. The energy
abl conc abl conc 0
balance equation including metal oxidation reactions is solved in MEDICIS using a consistent set of
enthalpy functions for all involved reactants and products.
The upward heat transfer from the upper interface to the surrounding walls is due to radiation. A
simplified model takes into account only the upper crust and wall emissivities. A more detailed model
available in case of coupling to the ASTEC/CPA containment code takes into account shape factors
for the different walls, the possible ablation of concrete walls above the pool and the absorption of the
radiative heat flux by aerosols present in the reactor pit using a user’s input absorption length.
A source mass term due to successive corium pourings can be described taking into account the
pouring rate of each species and the temperature of the poured corium.
Code peculiarities
The tool flexibility permits an easy addition of model variants and modifications of the dependence of
convective heat transfer coefficient and of the pool/concrete interface structure versus the interface
orientation.
3.2.8. SOCRAT/HEFEST
HEFEST−EVA code (Highly Efficient Finite Element Solution of Thermal problems for Ex−Vessel
stage Analysis) is a module of SOCRAT system code package (Bolshov & Strizhov, 2006), which is
used for modelling melt-structures interaction after core meltdown at both in−vessel and ex-vessel
stages (Bolshov & Strizhov, 2006), (Moiseenko, Filippov, Drobyshevsky, Strizhov, & Kisselev,
2011). HEFEST−EVA can also function in a stand-alone mode. The numerical procedure is based on
finite element method (FEM) solution of 2D energy equation along with material relocation
algorithms and chemistry models. The treated configurations are: debris in lower plenum, normally or
inversely stratified molten pool in VVER lower head with its heat erosion, normally or inversely
stratified molten pool in VVER core catcher, melt in VVER reactor pit causing MCCI.
NEA/CSNI/R(2016)15

The calculation domain may have either axisymmetric or planar geometry. Pool can consist of either
one or two layers with oxide layer atop or below metallic one. Crust build-up (both upper and
sidewall) is simulated by change of properties of particular finite element (FE) when its material has to
solidify. Solidified FE is detached from the pool and its composition cannot be changed until it melts
again.
Erosion algorithm
Ablation model: thermal erosion is modelled as melt propagation in the composite solid medium
(Stefan's problem). The heat conductivity in the melt is taken much greater than in solid (effective
conductivity of convection). FEs with solid content in which "melt conditions" (temperature, liquid
neighbours, etc.) are satisfied are attached to the pool and the composition of the FE and of the pool
changes. Their composition is calculated by averaging the previous composition of the pool and of
newly added FEs
Heat transfers
Nonlinear, transient 2D heat conduction equations are solved with coefficients calculated in
accordance with the physical model of the considered phenomena: melting, relocation, chemical
interaction etc.
Energy balance
Energy balance in homogeneous medium is kept automatically. The balance is estimated and
controlled during all time step calculation procedures: numerical solution, material relocation,
chemical reactions. The discrepancies, arising mainly due to different material properties of the
components during relocation and melt propagation in non-homogeneous medium are distributed in
the melt.
Code peculiarities
HEFEST-EVA is a FEM code for wide range of nonlinear 2D transient thermal problems with
changeable internal boundaries. In particular, the configuration of the internal pool boundary is
calculated during the pool evolution. The bilinear quadrilateral FEs are used. The results are dumped
in HEFEST internal format and in VTK format (readable by open PARAView postprocessor). The
MCCI specificities are reflected in the user's adjustment in input data, such as material properties,
initial composition, options of melt relocation and arrival procedures, chemistry model options etc. In
present the recommendations for choice of specific parameters and typical data sets are being
developed based on experimental data used in the code models and typical configurations in MCCI.
3.2.9. TOLBIAC-ICB
The main hypotheses of the phase segregation model (Seiler J. -M., Phase segregation model and
molten pool thermalhydraulics during molten core concrete interaction, 1996), (Seiler & Froment,
2000) for an oxide melt interacting with concrete are the following: due to the high liquidus
temperature of oxide melts and despite the melting of concrete and the presence of gas issued from
concrete decomposition, a solid crust is assumed to deposit at the concrete wall. The species that
encrust are assumed to be the most refractory species (mainly UO2 and ZrO2 or to correspond to the
pool composition at solidification time). A slow crust growth, a high liquid diffusivity and a small
diffusion boundary layer thickness are also assumed to maintain local thermodynamic equilibrium. It
must be noted that the decay power in the crust is taken into account; for a transient lasting several
days, the crust thickness becomes large and this phenomenon is not negligible.
NEA/CSNI/R(2016)15
Figure 3.2-9: Schematic view of the liquid corium pool with convection due to gas bubbling
A crust exists at the interface (which is indicated on the zoom)
With this view, the pool is only composed of liquid and consequently has a low viscosity (Figure
3.2-9). The interface temperature between the liquid pool and the solid crust is equal to the liquidus
temperature corresponding to the current composition of the remaining oxide liquid phase. This
interface temperature is used to calculate the heat flux between the pool and the concrete. The crust
thickness is calculated using a thermal model, with the hypothesis of the succession of steady state
regimes. A global thermodynamic equilibrium between the liquid melt and the whole solid crust is not
considered, since equilibrium is considered to occur locally, only at the interface (not for the liquid
and crusts together). Moreover a large temperature gradient occurs within the crust thickness, and a
global equilibrium between liquid and solid would suppose an equilibrium temperature which is
difficult to define.
TOLBIAC-ICB uses this approach, which is an ideal model. The main point in this model is that
no parametric fitting is used. The drawback is that it gives an overestimation of the melt temperatures
for some kinds of concrete; however the concrete ablation has generally a low sensitivity to the melt
temperature. Moreover the code user may select an option in the code which deactivates the phase
segregation model, for sensitivity studies, with a fitting of the interface temperature between melt and
crust.

When the thermodynamic equilibrium provides two immiscible liquids in case of pool containing
oxides and metals, a mixing criterion is used to predict the stratification or mixing of these two liquids.
Either the BALISE correlation (Tourniaire, Seiler, & Bonnet, 2003) or the Epstein criterion (Epstein
M. , Petrie, Linehan, Lambert, & Cho, 1981) can be used. The first correlation depends only on the
density of the liquids and on the gas superficial velocity. The Epstein criterion also depends on the
height of the layers and takes the gas superficial velocity into account through the pool void fraction.
Concerning the physico-chemistry of the stratified melt, two options corresponding to two
different situations may be chosen. Following the miscibility gap model proposed by (Seiler, Froment,
& Defoort, 2003), both metallic and oxidic liquids can be assumed to be at thermodynamic
equilibrium. In such case, the two layers are globally bounded by one refractory crust and have the
same boundary temperature. A single crust composition is considered and calculated with GEMINI2
since no crust exists at the interface between the two layers. On the other case, both liquids are
assumed to be out of thermodynamic equilibrium. In this latter case, both liquids may be bounded by a
different crust: an oxide refractory crust for the oxide pool and a metallic crust in front of the metallic
pool. Since the liquidus temperature of the oxide is higher than the liquidus temperature of the metallic
pool, an oxide crust is expected to form at the interface between both layers. Thermodynamic
calculations are then performed for both liquids in order to determine the crust compositions and the
NEA/CSNI/R(2016)15
liquidus temperature related to each layer. Moreover in this case debris of refractory oxides in the
metallic layer are considered. Evidence of one scenario or the other is not clear, since MCCI
experiments using prototypic material with clear stratification do not exist. However the scenario with
no equilibrium between the two layers seems to be the more likely. But practically the scenario with
equilibrium is usually used with TOLBIAC-ICB, because it is simpler (no interfacial crust, no debris
of refractory oxides in the metal layer).
Erosion algorithm
Due to the low conductivity of concrete, conduction is not considered in the concrete walls. The
concrete walls are divided into individual cells, the size of which is defined by the code user (typically
5 cm for reactor case). The heat flux received by a cell is converted into ablation, and the progressive
ablation of the cells defines the shape of the cavity. The concrete walls above the liquid level are also
defined with individual cells, and the same ablation algorithm applies, giving the cavity shape. Several
concrete layers may be simply defined by changing the physical properties of the cells. However, for
long terms simulation a specific model has been implemented, which predicts the end of ablation when
the heat flux becomes low, and conduction must be considered (Tourniaire, Spindler, & Guillaumé,
2010).
Heat transfers
Taking into account remaining uncertainties related to the power splitting on the pool boundaries, the
code user may select several different heat transfer correlations on each boundary (BALI (Bonnet,
2000), Blottner (Blottner, 1979), Kutateladze-Malenkov (Kutateladze & Malenkov, 1978), …). The
code user may also simply modify the ratio between the bottom and lateral heat transfers. The
recommendation is to use weighting factors, between 2 and 3 for silica-rich concretes, around 1 for
limestone-rich concretes, rather than very different correlations, written in terms of different variable
groups, since this last solution may lead to unexpected variations of the power split.
A crust generally appears at the upper surface of the pool because of radiative heat transfer
towards the reactor cavity or water aspersion. A crust thickness and crust surface temperature are
calculated, supposing a steady state regime. The melt to crust heat exchange is governed by the heat
transfer coefficient and the interfacial temperature (pool liquidus temperature), and does not depend on
the external conditions. The crust surface temperature and crust thickness depend on the external
conditions, which may be radiation or heat exchange with a water layer. When there is no water
aspersion, the crust density is larger than the melt density (crust composed of refractory materials),
and the crust continually forms and sinks: it is then only taken into account for the heat transfer, and
no crust increase is taken into account. In case of water aspersion, crust porosity is supposed, and the
crust increase is taken into account in the model.
For the radiation above the melt surface, two levels of modelling exist in TOLBIAC-ICB. The
most simple radiation model only considers a radiation between two infinite planes, the upper crust
and the surroundings, depending on the surroundings temperature and the emissivity of the pool
surface and of surroundings. A more complex model also exists. The cavity above the melt level is
described, shape factors are considered in the radiation model, and ablation of the concrete vertical
walls above the melt level is taken into account. The ablation of a horizontal steel wall representing
the vessel structure above the melt may also be modelled. Heat losses through gas or aerosols escaping
from the melt pool to the containment may be taken into account through a coefficient of power loss
defined by the code user.
NEA/CSNI/R(2016)15
Energy balance
The steady state energy balance equation reads in case of a homogeneous melt as follows:
bottom walls lateral walls inert walls upper surface source terms
Sf.hf.(Tmelt-Tint) + Sl.hl.(Tmelt-Tint) + Sinert.hinert.(Tmelt-Tinert) + Su.hu.(Tmelt-Tu) + qs.Cps.(Tmelt-Ts)
= Pres + Pini
Input Power Initial overheat
This equation is used to calculate the melt temperature Tmelt.
S stands for the heat transfer area with the bottom wall (index f), with the lateral concrete wall
(index l), with the lateral inert wall (index i), and with the upper surface (index u).
h stands for the heat transfer coefficient with the bottom wall (index f), with the lateral concrete
wall (index l), with the lateral inert wall (index inert), and with the upper surface (index u).
Tint is the interfacial temperature between the melt and the crust. Tinert is the interfacial
temperature between the melt and the inert wall. Tu is the interfacial temperature between the melt and
the upper surface.
At the upper surface, if there is a crust, the interfacial temperature is the same as for the
interfacial temperature between the melt and the crust for the heat transfer with the walls. If there is no
upper crust the surface temperature is calculated through a heat balance at the upper interface.
Concerning the source term (corium from the pressure vessel at the beginning, and then metals from
the pressure vessel melting), qs is the source mass flow rate added to the melt during time step dt, Cps
is the heat capacity of this mass, and Ts is the temperature of this mass. For short transients and
especially for the simulation of experiments with a short duration, a transient term is included, which
writes: Mmelt.Cpmelt (Tmelt-Tn-1mlet)/dt.
3.2.10. WECHSL
WECHSL-Mod 3 was released in 1995 (Foit, Reimann, Adroguer, Cenerino, & Stiefel, 1995) and
further improvements are under development as new experimental or theoretical insight becomes
available. The outline of the WECHSL modelling is described below as an example for the
requirements and capabilities of a MCCI code.
The WECHSL code aims at modelling the physical and chemical phenomena governing the
molten core-concrete interaction in a severe reactor accident, when the molten core has penetrated the
pressure vessel. WECHSL models one-dimensional as well as two-dimensional melt-concrete
interactions in axis-symmetric concrete cavities. Main implemented models are documented in (Foit,
Reimann, Adroguer, Cenerino, & Stiefel, 1995) and (Foit J. J., 1997): The left side of Figure 3.2-10
shows models for a homogeneous oxide melt into which the metal phase is homogeneously dispersed
in form of droplets, whereas the right side of the same figure shows the layered melt configuration
with the metal layer at the bottom, overlaid by the oxide corium layer in the initially cylindrical
concrete cavity.
NEA/CSNI/R(2016)15
Figure 3.2-10: Sketch of the main models of the WECHSL code

To calculate one-dimensional experiments, the horizontal concrete may contain an additional
layer composed of concrete and metals to take into account those configurations that were used in
large-scale ACE experiments (Thompson, Farmer, Fink, Armstrong, & Spencer, 1997). The metallic
melt may contain Zr, Cr, Fe, Ni and Si. The oxide layer is composed of UO2, ZrO2, CaO, SiO2,
Al2O3, Cr2O3 and FeO. Internal energy can be produced by decay heat or by exothermic chemical
reactions. The mass and composition of the melt change as liquid concrete is mixed into the oxide
corium melt, and by the chemical oxidation of some of the metals. Energy is lost to the melting
concrete and to the upper containment by thermal radiation or by evaporation of sump water, possibly
flooding the surface of the melt.
The code performs calculations from the time of initial contact of a hot molten pool until long
term basemat erosion over several days with the possibility of basemat penetration.
3.2.11. Synthesis on main features of the MCCI codes
Main features of the MCCI codes: COCO, CORCON (Gauntt, MELCOR Computer Code Manuals,
Version 1.8.6, Vol. 1: Primer and Users’ Guide, Vol. 1, Rev. 3, NUREG/CR-6119, Rev. 3, SAND
2005-5713, 2005),CORQUENCH (Farmer, 2010), CORIUM2D (Parrozi & Fontana, 2013), COSACO
(Nie, Fischer, & Lohnert, 2007), MAAP (MAAP, 2011), MEDICIS (Cranga, Fabianelli, et al.
2005),SOCRAT (Bolshov & Strizhov, 2006), TOLBIAC-ICB and WECHSL (Foit, Reimann,
Adroguer, Cenerino, & Stiefel, 1995) are summed-up in the following table. Some brief indications on
available fission product release models are added in last table line.
NEA/CSNI/R(2016)15
NEA/CSNI/R(2016)15
NEA/CSNI/R(2016)15
NEA/CSNI/R(2016)15
NEA/CSNI/R(2016)15
NEA/CSNI/R(2016)15
3.3. Comparison and synthesis of code modelling
The purpose of this section is to compare available models in different codes and to point out
limitations of these models
3.3.1. Thermal-hydraulics models
Main heat transfer models to be considered are the following:
• - 2D heat convection;
• - Heat transfer between oxide and metal layers;
• - Oxide/metal pool configuration evolution.
They are addressed hereafter.
3.3.1.1. 2D heat convection

The main convection mechanism is gas-driven convection; correlations obtained by Kutateladze
(Kutateladze & Malenkov, 1978) or derived from BALI experiments (Bonnet, 2000), or still from
Deckwer’s surface renewal model (Deckwer, 1980), (Tourniaire & Varo, 2008) are most frequently
used in codes; the value of heat transfer coefficient is typically 100 up to a few 1 000 W/m/K versus
viscosity and superficial gas velocity. Free convection and still much more solutal convection (which
should be more efficient than free convection) which are not taken into account in codes should
increase further the convective heat transfer coefficients. In a previous analysis (Cranga, et al.,
Towards an European consensus on possible causes of MCCI ablation anisotropy in oxidic pool,
2014), it was shown that the convective heat transfer coefficient due to gas bubbling is significantly
higher than the effective heat transfer coefficient in oxide melts deduced from experiments. In the
frame of thermal resistance approach at the pool/concrete interfaces, the description of 2D convection
in the bulk pool using only the gas driven convective heat transfer is a reasonable approximation, this
is the case of MEDICIS IRSN code if using a “no-crust “model because of the prevailing thermal
resistance of pool/concrete interfaces; however if a stable crust model is used with a pool/crust
temperature imposed by thermochemistry, the heat flux distribution in the oxidic pool depends on the
level of convective heat transfer coefficient and on its distribution along the pool interfaces.
Experimental data on 2D convection due to gas bubbling were obtained recently in the frame of
the CLARA simulant programme (Cranga, et al., Towards an European consensus on possible causes
of MCCI ablation anisotropy in oxidic pool, 2014). And two sets of correlation have been proposed
from these results ( (Michel, 2015) and (Bottin, et al., 2016).
3.3.1.2. Oxide/metal heat transfer

The oxide/metal interface in case of pool stratification has very likely a different structure and
involves different heat transfer mechanisms compared to those of the pool/concrete interface, since the
gas bubbling will promote entrainment of tiny metal droplets into the upper oxide layer and intense
heat transfer from the entrained metal to the oxidic phase.
Available correlations on oxide/metal convective heat transfer coefficient are discussed in detail
in (Cranga, et al., 2010). One of the first correlations established for the heat transfer coefficient at a
liquid/liquid interface was originally proposed by Werle and relied on simulant experiments (Werle,
1982) performed only at a low gas velocity (<0.7cm/s). The formulation is based on a free convection
heat transfer correlation combined with an enhancement factor depending on the superficial gas
velocity.
The most frequently used heat transfer correlation (Greene & Irvine, Heat transfer between
stratified immiscible liquid layers driven by gas bubbling across the interface, 1988) is a semi-
NEA/CSNI/R(2016)15
empirical one assuming that the interfacial heat transfer process is a transient renewal process on both
sides of the interface and that the two resistances to heat transfer can be considered in series. In
Greene’s correlation, the heat transfer coefficient does not depend anymore on viscosity whereas
experimental results obtained by Greene do. Nevertheless this correlation is often used in MCCI codes
in lack of a better one.
Due to limitations of existing correlations on oxide/metal convective heat transfer, a work was
launched at IRSN to build an improved correlation taking into account the existing database. This new
correlation was fitted to experimental data points reported by Greene and obtained in the ABI test
series (Cranga, et al., State Of the Art Report on MCCI in dry conditions: analysis of experiments and
modelling, 2013). The analysis of ABI results showed that an approach taking into account thermal
resistances in series on both sides of the oxide/metal interface is not appropriate. Therefore a new
correlation (Cranga, et al., State Of the Art Report on MCCI in dry conditions: analysis of experiments
and modelling, 2013), aimed at describing the global heat transfer across the interface, was finally
−1.64
 U Tlow  U Tup 
selected: h = 13.8 Pr 1.65 − 0.67
Pr Re 2.37
Re − 0.35    Equation 3.3-1
low up low up  J  J 
 g  g 
where the subscripts up and low refer respectively to the upper and lower layers and UT is the bubble
terminal velocity.
Figure 3.3-1: Evolution of heat transfer coefficient versus superficial gas velocity Jg in the reactor
case: Comparison between Greene’s correlation and new correlation versus corium viscosity
An application of such a correlation was made to the reactor case by IRSN (Cranga, et al., State
Of the Art Report on MCCI in dry conditions: analysis of experiments and modelling, 2013). Material
properties were chosen in order to be representative of the reactor case with a wide range of oxide
viscosity to take into account the high viscosity increase in the long term phase due to the high silica
fraction caused by the large concrete ablation. These results are compared to those obtained with
Greene's correlation and plotted in Figure 3.3-1 1.
1. It can be noticed that the heat transfer coefficient given by the new correlation decreases from
μup = 0.005 to 0.5 Pa.s but increases from μup = 0.5 to 1 Pa.s. It is due to the fact that the terminal
bubble velocity depends also on the viscosity: this velocity decreases with the viscosity and the
heat transfer coefficient decreases with the terminal bubble velocity of the upper layer so it has
the opposite effect of the explicit viscosity of the upper layer which intervenes in the Prandtl and
Reynolds numbers.
NEA/CSNI/R(2016)15
With the properties chosen for a reactor case allowing a possible stratified configuration, the heat
transfer coefficient is higher with the new proposed correlation by a factor 2 to 5 than with that
currently used Greene's correlation. This work confirms also that the heat transfer at the interface
oxide/metal is very high compared to the convective heat transfer from the bulk oxidic pool to other
interfaces (see 3.3.1.1), if these results can be scaled up to the materials in the reactor case. The
possible presence of an oxidic crust at the oxide/metal interface will reduce heat transfer through this
interface but in a moderate way.
3.3.1.3. Pool configuration evolution model

There are mainly three experiments with simulant material and associated models available to study
the stratification criteria: Epstein (Epstein M. , Petrie, Linehan, Lambert, & Cho, 1981), Casas and
Corradini (Casas & Corradini, 1992) and BALISE tests (Tourniaire, Seiler, & Bonnet, 2003).
All these studies show the existence of three main configurations which can be observed by
increasing (or decreasing) the gas flow rate. For low gas velocity, the pool is stratified. When the gas
flow rate is increased, a significant part of the heavier phase is entrained in the light phase. This
corresponds to the onset of entrainment (Jgoe) on Figure 3.3-2. When the gas superficial velocity
increases further, all the heavy phase is entrained in the lighter part of the pool; this corresponds to the
full mixing (Jgfm).
Stratified Full mixing

Configurati
Jgoe Jgfm
Figure 3.3-2: Mixing rate % versus gas superficial velocity in a BALISE test
All the authors proposed a criterion for the superficial gas velocity for the onset of entrainment
(Jgoe) and for the full mixing (Jgfm). Among these criteria, the BALISE criterion appears as an upper
bound for the threshold of the stratified configuration. This criterion is based on the ratio between the
density of the light phase (ρL) and the heavy liquid phase (ρH).
and Jgfm = 0.054(ρ H − ρ L )/ ρ L m/ s J goe = 0.5 J gfm Equation 3.3-2
Several limitations on this modelling appear:

1. Obviously, all variations in the physical properties observed in an assumed MCCI with
metal and oxide have not been investigated experimentally. Furthermore, the effect of
thickness of the heavy layer and the effect of a 2D injection have not been investigated
either. Furthermore, this correlation, based on gas injection from the bottom only, assumes
that the whole gas rate flows through the pool and requires an extension to the case of 2D
gas injection and a clarified definition of superficial gas velocity used: it is to be considered
that either the superficial gas velocity is defined as the ratio of the gas volume rate to the
area of the interface (bottom + lateral) or the superficial gas velocity is defined as the ratio of
the produced gas volume rate to the pool horizontal section.
NEA/CSNI/R(2016)15
The few MCCI experiments with prototypic corium and steel have shown that the configurations
may be more complex that the two academic cases previously considered. For instance, in the three
VULCANO tests for which a significant mass of metal has been found after the end of the experiment,
the metallic mass was not found as a horizontal layer, but had a more complex shape, with metallic
phases vertically facing the walls. This shows that phenomena might be far more complex that those
represented in simulant material stratification experiments (Tourniaire, Seiler, & Bonnet, 2003) where
only two liquid compositions are present. During MCCI, miscible but yet unmixed oxides are present
and coexist with immiscible molten metals.
Application to the reactor case
In the reactor case, when MCCI starts, the heat fluxes are high and thus the superficial gas velocity are
so high that the “full mixing” configuration may be assumed. One of the reactor case specificity is the
decrease of the heat flux on the vertical and horizontal wall during the MCCI. This decrease is due to
the increase of the area bounding the pool and to the decrease of the power in the pool as decay heat
falls with time. The decrease of the heat flux induces a decrease in the gas flow rate bubbling in the
pool. If the gas flow rate falls below the threshold for full mixing, the metal starts to slump and settles
at the bottom of the pool. But as the metal heat transfer coefficient is higher than the oxide one, the
heat flux to the bottom becomes higher and the gas flow rate increases and the stratification
disappears. But as soon the gas flow rate reduces, stratification of the metal starts again. The stratified
configuration is really stable only when the power transferred to the bottom wall corresponds to a gas
superficial velocity lower than the threshold for the onset of entrainment. Most MCCI codes use
specific stratification models or user’s input data. Only TOLBIAC-ICB and MEDICIS use a model
derived from simulant experimental data. These codes use two thresholds depending on the current
configuration (see Figure 3.3-3). When the current configuration is fully mixed, configuration is kept
mixed until the superficial gas velocity becomes lower than a transition criterion, called Jg_stableS on
Figure 3.3-3. When the configuration is a stratified configuration, the superficial gas velocity must be
higher than a transition criterion, called Jg_stableM on Figure 3.3-3, for a change to a “stable” fully
mixed configuration.
Configuration Stratified Partial mixing Full mixing
Full Mixing
Jg< Jg stableS Jg>Jg stableS
Stratified
Jg<Jg_stableM Jf> Jg_stableM
Figure 3.3-3: Flow map in a MCCI code

The criterion derived from BALISE experiments uses the entrainment onset threshold Jgoe for
Jg_stableS, below which the stratified configuration is stable and the full mixing threshold Jgfm for
Jg_stableM, above which the homogeneous configuration is stable. However, to avoid oscillations between
stratification and full mixing, MCCI codes raise the Jg_stableM to make the stratified configuration more
NEA/CSNI/R(2016)15
stable, which promotes the axial ablation. The occurrence of these oscillations shows that the pool
configuration model is probably too crude and some intermediate configuration types, which cannot be
described with the present model, need to be included. These oscillations may be physical as stratification
enhances heat transfer through the steel layer to the bottom concrete and this increases gas sparging,
promoting de-stratification. A minimum metal layer thickness is required to maintain oxide/metal
stratification. It is chosen usually between 1 cm and 5 cm: the lower boundary corresponds roughly to the
size of a bubble and the upper boundary to the maximum thickness of the possible solid crust built at the
interface between the metal layer and concrete. For a layer thickness above the lower threshold value the
metal layer has some chance to be separated from the oxidic layer in presence of bubbling and above the
upper threshold value the metal layer should be at least partially liquid thus permitting heat convection with
the upper oxide layer.
A parametrical reactor study was performed with the ASTEC/MEDICIS code on the impact of pool
configuration evolution criteria (Cranga, Mun, & Marchetto, 2010) It highlighted the strong increase of the
melt-through time with the decrease of the superficial gas velocity threshold below which stratification
appears and a strong dependence on the minimal metal thickness required for pool stratification.
Thus, BALISE simulant experiments have established a convenient criterion for the pool
stratification onset. However several uncertainties remain on this stratification criterion: lack of a
dimensionless formulation, validation in 2D gas injection conditions, too high a conservatism
particularly in the prototypical case of a low ratio of metal to oxide volumes. And, above all,
understanding of intermediate configurations (partially stratified and partially mixed) has to be
improved.
3.3.1.4. Synthesis on thermal-hydraulics models
Features and limitations of available models on pool thermal-hydraulics are summed-up for each type of
phenomena on Table 3.3-1.
Following remarks can be made of sub-items of Table 3.3-1:
2D heat convection: two sets of consistent and validated correlations for describing 2D heat
convection have been obtained from CLARA experimental results (Cranga, Spengler, et al. 2013) and
depending on corium transport properties and superficial gas velocity profile along the pool/concrete
interface ( (Michel, 2015) and (Bottin, et al., 2016). These sets will contribute to reduce the impact of
uncertainties on 2D convection.
Oxide/metal heat transfer: Greene’s correlation gives a heat transfer coefficient value comparable
to experimental data or slightly lower; since the obtained heat transfer coefficient is much higher than the
lateral convective heat transfer coefficient according to experimental data, the use of Greene’s
correlation can be considered as enough conservative because it leads to a focusing of decay power in the
oxide layer towards the lower interface with the metal layer, thus promoting the axial ablation kinetics.
Criteria of oxide/ metal pool configuration evolution: in spite of uncertainties, BALISE type
stratification criterion Jg < bHS 2.Δρ/ρmin combined with a minimum metal thickness thm is likely
conservative if bHS > 0.027 and thm equal to at most 3 cm because it will promote stratification even in
situations where a stable stratification may be doubtful.
2. bHS is the BALISE stratification coefficient for the switch from homogeneous to stratified configuration.
Another coefficient bSH has been experimentally determined for the switch from stratified to homogeneous
configuration (Cranga, et al., Towards an European consensus on possible causes of MCCI ablation
anisotropy in oxidic pool, 2014).
NEA/CSNI/R(2016)15
Table 3.3-1: Status on thermal-hydraulics modelling

Phenomenon Codes Established knowledge/retained assumptions Limitations Impact on MCCI
including its predictions
description
Gas-driven convection is likely to be prevailing in all

codes(excepted for SOCRAT using free convection), Low impact of
existing correlations (Bali (Bonnet, 2000), Kutateladze hconv level
(Kutateladze & Malenkov, 1978), (Blottner, 1979),
(Konsetov, 1966), Deckwer (Deckwer, 1980) give Solutal convection High impact of ratio
2D heat
same o.m.1 not described hlat/ hax if “crust
convection
All codes model”
In most cases (SOCRAT excepted) hconv >> hint,exp2 Lack of data on 2D
(hconv)
hconv profile Low impact of ratio
Range of lateral/axial convection ratio hlat/ hax: hlat/ hax if no crust
MEDICIS: =1; model because
hconv is >> hint
TOLBIAC-ICB : 1 to 3 versus concrete type
Greene’s correlation (Greene & Irvine, Heat transfer

between stratified immiscible liquid layers driven by
gas bubbling across the interface, 1988) is used in
MEDICIS, TOLBIAC-ICB; Werle’s correlation in
WECHSL (Werle, 1982), Blottner’s one (Blottner, High on axial
No correlation
1979) in SOCRAT ablation kinetics in
Oxide/ metal All codes validated for reactor
case of stratification
heat transfer CORQUENCH, conditions
hconv o/m >> hconv ox(excepted for (depends on reactor
MAAP (transport
COSACO,SOCRAT where hconv o/m ~ hconv ox) design and
((hconv o/m) excepted properties, bubble
concrete
size
Limited impact of crust at oxide/metal interface composition)
Heat conduction for CORIUM-2D
High on axial
No mechanistic ablation kinetics
BALISE based criterion (Tourniaire, Seiler, & Bonnet, model (no because it
2003):Jg > bHS.Δρ/ρmin + minimum metal thickness dependence on determines
Oxide/ metal All codes, thm (CORCON,MEDICIS, SOCRAT, TOLBIAC-ICB) transport stratification and
pool CORQUENCH,
properties) downwards
configuration MAAP Mechanistic settling/entrainment model (CORCON)
focusing of decay
evolution excepted Excepted for non-
COCO: user’s input, COSACO: stratification according power (depends on
to sign of Δρ validated CORCON reactor design and
model concrete
composition)
1: o.m.: order of magnitude 2: hint,exp: effective heat transfer coefficient at pool/concrete interface deduced from
experiments
3.3.2. Pool interface models

As mentioned in section 3.3.1.1, it was pointed-out that the convective heat transfer coefficient due to
gas bubbling exceeds largely the effective heat transfer coefficient in oxide melts deduced from
experiments.
Moreover data from CLARA simulant experiments on 2D convection (Cranga, et al., Towards an
European consensus on possible causes of MCCI ablation anisotropy in oxidic pool, 2014) cannot
account for the heat flux distribution along the pool/concrete interface in 2D ablation MCCI
experiments versus the concrete but have the opposite trends. Indeed the MCCI experiments show
clearly that the ratio of lateral to axial ablation and so the ratio of lateral to axial heat flux is larger
than 1 in case of siliceous concrete that induces a higher corium viscosity and is equal to around 1 in
NEA/CSNI/R(2016)15
case of LCS concrete that induces a lower corium viscosity. Another aspect of CLARA results is the
distribution of convective heat transfer coefficients obtained in case of a zero injected gas velocity at
the pool bottom: in this case, the hlat/hbot ratio increases up to 2.2 at high viscosity and to still higher
values at lower viscosity. This situation might correspond roughly to a possible scenario involving the
build-up of an intact crust at the bottom interface preventing any gas flow through this interface and
explaining partly the prevailing lateral ablation in case of a siliceous concrete. However this might be
valid only if the bottom crust stays really completely tight to gases, which should be not the case at
large scale in the reactor case. In case of LCS concrete, this scenario is clearly contradicted by the
rather isotropic 2D ablation.
Consequently the convection anisotropy caused by bubble agitation within the liquid is not, at the
present state of knowledge, the major reason of the dissymmetry of 2D ablation which is much more
likely to be explained by the dissymmetry of lateral and axial pool/concrete interfaces; the description
of heat transfer across the pool/concrete interface is therefore of prime importance.
Different approaches have been proposed:
with stable crust at the pool/concrete interface using either a equilibrium crust model or with non-
equilibrium crust model or
without any crust at the pool/concrete interface.
3.3.2.1. Approach with active crust at the pool/concrete interface
This approach is used in 2 different models. The first model assumes the build-up of a crust at the
pool/concrete interface in thermodynamic equilibrium with the pool (so-called “equilibrium crust
model”); the pool/crust interface temperature is equal to the liquidus temperature at the pool
composition (see Figure 3.3-4) and the composition of the crust increment is that of the solid in
equilibrium with the liquid pool composition. This approach proposed by CEA (Spindler & Seiler,
2006) was implemented in the TOLBIAC-ICB code (Spindler, Tourniaire, Seiler, & Atkhen, 2005).
Figure 3.3-4: Pool/concrete interface assuming a corium crust in thermodynamic equilibrium

(‘equilibrium crust model’)
The justification is double: 1) the pool temperature in ACE tests follows approximately the same
trend as the liquidus temperature; 2) in the MCCI 1D test MACE3b (Levy, 2002) the temperature was
little modified by a large power increase, thus showing that the pool bulk thermal resistance is low and
then the molten fraction is high even in the region close to the pool/concrete interface. However the
comparison of results obtained respectively in 1D and 2D MCCI tests (Spindler & Seiler, 2010) shows
that the bulk melt solid fraction is higher in 2D ablation tests in particular with siliceous concrete and
consequently demonstrates that the pool/crust interface temperature may become much lower than the
liquidus one in a more realistic 2D ablation configuration. This model assuming the thermodynamic
equilibrium between the crusts and the pool, might be valid only in the asymptotic case with a reduced
surface to volume ratio as in 1D tests and with a rather low corium viscosity as obtained with LCS
concrete (Spindler & Seiler, 2010).
A variant of this model (also implemented in TOLBIAC-ICB) is to assume that the crusts have
the composition of the melt at the time they are solidifying (non-equilibrium stable crusts), remain
stable and that the interface temperature is nevertheless at liquidus temperature.
NEA/CSNI/R(2016)15
The second model assumes a deviation from thermodynamic equilibrium (so-called “non-
equilibrium crust model”) and takes into account a slag layer and/or a gas film between the concrete
and the crust, see Figure 3.3-5: this type of model was implemented in several codes: in the previous
version of COCO, CORQUENCH (Farmer, 2010), CORCON (Bradley, Gardner, Brockmann, &
Griffith, 1993), MAAP (MAAP, 2011), MEDICIS code (Cranga, Fabianelli, Jacq, Barrachin, &
Duval, 2005), SOCRAT (Bolshov & Strizhov, 2006), WECHSL (Foit, Reimann, Adroguer, Cenerino,
& Stiefel, 1995); the pool/crust interface temperature is evaluated by linear interpolation between
solidus and liquidus temperatures as:
Tint = (1 − γ ) Tliq + γ Tsol Equation 3.3-3
or is evaluated from the threshold volumetric molten corium fraction (Cranga, Mun, Michel, Duval, &
Barrachin, 2008):
Figure 3.3-5: Pool/concrete interface structure with deviation from thermal equilibrium in MEDICIS
code (‘non-equilibrium crust model’)
The non-equilibrium assumption is supported by the pool temperature measurements in CCI tests
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010) showing that the pool temperature stays more
or less below the liquidus temperature at the pool composition, requiring γ >0.
A still different model assuming that the solid fraction of the oxidic pool is non zero with crust
formation along the pool/concrete interface is used by Areva in the COSACO code (Nie, Fischer, &
Lohnert, 2007). In this code, two thermal resistances in series separate the oxidic pool bulk from the
decomposing concrete: a convective resistance (based on the BALI correlation (Bonnet, 2000) and a
conductive resistance through a crust layer. The overall solid fraction of the pool is determined by the
in-built free-enthalpy minimiser on the basis of the oxidic pool composition and temperature. The
obtained solid mass is assumed i) to accumulate along the sides of the oxidic pool as a crust and ii) to
be suspended in the bulk of the oxidic pool (as floating crystals). The split between these locations is
proportional to the magnitude of the corresponding interfacial and volumetric heat sinks. On the basis
of the solid mass redistribution, the solid density and the surface area of the considered boundary, a
crust thickness is calculated. The heat flux density to the concrete is then calculated with the following
expression:
Tox − Tdec
ϕ= Equation 3.3-4
1 δ
+
h λ
where h is the BALI heat transfer coefficient, δ the crust thickness on the considered boundary, λ
its thermal conductivity, Tox the temperature of the oxidic melt and Tdec the decomposition
temperature of the concrete (assumed to be equal to the temperature at which its solid volume fraction
is equal to 50%).
A similarity of this approach with the previously described “non-equilibrium crust model” is that
the bulk pool temperature and hence the pool/crust interface temperature can be lower than the
liquidus temperature but the pool/crust interface temperature is not a priori imposed.
NEA/CSNI/R(2016)15
3.3.2.2. Approach without active crust at the pool/concrete interface

This approach assumes no stable active crust, either no crust at all, in particular along the lateral
pool/concrete interface, or an inert crust, built-up possibly on the horizontal interface, for example in
case of siliceous concrete, with an interface thermal resistance because of a slag layer build-up (so-
called “no crust model”, see Figure 3.3-6): this type of model is proposed with some differences by
IRSN, GRS and VTT. Models from IRSN (Cranga, Mun, & Marchetto, 2010) and GRS, (Spengler,
2007), (Spengler, 2012), are implemented in the MEDICIS code; the VTT model (Sevón, 2008) was
introduced in a MCCI code called FINCCI (Sevon, 2011) and developed recently.
Figure 3.3-6: Pool/concrete interface structure (‘no crust model’)
A first justification of this approach was that in the ARTEMIS 1D test series it was found with
ASTEC/MEDICIS recalculations using such an approach, that the pool temperature follows
approximately the same evolution trend as the liquidus temperature. But in all tests (except for the first
one because of different initial conditions) the temperature deviates significantly from the liquidus
value staying above or below depending on the boundary conditions (ablation rate and superficial gas
velocity); consequently crust formation at the pool/concrete interface does not take place in most cases
at thermodynamic equilibrium with the pool.
A more recent justification of this approach came from the significant decrease of the pool
temperature below the liquidus one in 2D MCCI tests (Farmer, Lomperski, Kilsdonk, & Aeschlimann,
2010) in particular with siliceous concrete while keeping a sustained lateral ablation, suggesting that
the pool/concrete interface temperature and even the pool/crust interface temperature (if any crust is
present) are not directly related to the pool corium thermo-chemistry data. Moreover no evidence of
crusts along the pool/concrete interfaces was found from post-test examinations.
The approach retained recently in the MEDICIS code by IRSN assumes no crust build-up along
the pool/concrete interface. In this approach, the effective heat transfer coefficient from the pool bulk
to the concrete interface is obtained from the equation:
−1 −1 −1
heff = hslag + hconv Equation 3.3-5
where h slag is the heat transfer coefficient of the slag layer or equivalently of the pool/concrete
interface and hconv is the convective heat transfer coefficient from the pool bulk to the slag layer, both
coefficients possibly depending on the interface orientation.
The approach proposed by GRS, (Spengler, 2007), (Spengler, 2012) is also based on the use of an
effective heat transfer coefficient depending on the interface orientation; this effective heat transfer
coefficient aims at describing the overall heat transfer from the pool bulk up to the pool/concrete
interface but without a distinction between the convective heat transfer within the pool and the heat
transfer across the slag layer. This model was implemented as an option into the MEDICIS code.
Although this model is still more simplified, its application to 2D MCCI tests confirms the approach
proposed by IRSN ignoring any solidifying stable crust along the interface.
NEA/CSNI/R(2016)15
3.3.2.3. Upper pool interface models

Although the upward heat transfer does not impact the anisotropy of 2D ablation, it influences the
fraction of power radiated upwards, the value of the average ablation depth and the level of the pool
temperature. Most models implemented in MCCI codes for describing the upwards heat transfer from
the pool bulk use the following assumptions:
• Upwards convective heat transfer evaluated from a convective heat transfer correlation;
• Conduction across the upper crust formed on the top pool interface assuming the
thermodynamic equilibrium is reached or not as in the case of the first two pool/concrete
interface models.
Another approach proposed by GRS (Spengler, 2012) evaluates the upwards heat flux from the
bulk pool to the upper crust using an overall heat transfer coefficient, which is constant and in most
cases equal to around 300 W/m2/K. A very similar approach is also used by VTT (Sevon, 2011). It has
to be pointed out that these assumptions are derived from an overall fitting against experiments and
ignore the possible impact of an upper crust built-up at some imposed interface temperature.
Areva in the COSACO code (Nie, Fischer, & Lohnert, 2007) describes the build-up of an upper
crust whose thickness is depending on the heat flux distribution along the pool interfaces but without
imposed pool/crust interface temperature, as explained above in section 3.3.2.1 at the opposite of other
approaches dealing with an upper crust.
3.3.2.4. Synthesis on pool interface models
Features and limitations of available models are displayed for each type of phenomena on Table 3.3-2.
Specific limitations of models implemented only in some codes are outlined in red. Overall limitations
are outlined in purple.
Heat conduction behind the ablation front is neglected in most codes: this approximation is
justified in a quasi-state regime and in particular in medium and long term MCCI phases in the reactor
case (beyond around 5 hours); this is not the case in experiments with a thinner concrete basemat
especially in the late phase.
Following remarks can be made on pool interface models (Cranga, et al., Towards an European
consensus on possible causes of MCCI ablation anisotropy in oxidic pool, 2014):
1. Pool/concrete interface model: two main model approaches are available, a first one assuming a
stable crust and a second one assuming no stable crust along the pool/concrete interface and
hence imposing no precise pool/crust interface temperature. The detailed analysis on one side of
available MCCI test results and on the other side of CLARA data on 2D heat convection driven
by gas bubbling leads to the following conclusions: the profile of the thermal resistance along the
pool/concrete interface, instead of 2D convection, is likely to be the main explanation of the 2D
ablation. This is in favour of the “no crust” model approach at least in the short and medium
MCCI phases which can be investigated in MCCI experiments. More precisely a possible
explanation of the anisotropy observed with siliceous/clinker concrete, might be based on the
increase of the thermal resistance at the bottom concrete interface due to the pile-up of solid
concrete aggregates below an unstable crust or possibly a viscous slag layer. However the later
evolution of the thermal resistance in particular at the bottom interface in the long term MCCI
phase remains unknown. In case of LCS concrete, which releases only tiny aggregates easily
entrained into the pool, the thermal resistance of pool/concrete interface stays rather uniform and
is also largely prevailing over the bulk pool resistance because of the high convective heat
transfer at lower corium viscosity. The pool/concrete interface model impacts also on the bulk
pool temperature evolution : on one side the models with a lower pool/crust interface temperature
or without any crust lead to a lower pool temperature are partially supported by experimental
NEA/CSNI/R(2016)15
data; on the other side models with a higher pool/crust interface temperature, closer to the pool
liquidus temperature are also in broad agreement with experiments if taking into account
segregation of refractory material into built-up crusts, although to a less extent in case of
experiments with siliceous concrete; however the latter models suppose the existence of
refractory crusts along the pool/concrete interfaces which was not demonstrated up to now by
metallographic PTE1.
Pool upper crust model: Due to uncertainties on experimental boundary conditions in particular at
the top of the pool, it is difficult to select the best assumptions for the upwards heat transfer. However
the modelling with a low pool/crust interface temperature or without any crust seems to be supported
by experiments showing in most cases a thin crust with a crust renewal process or even no upper crust
at all in dry situation. The use of the same value of convective heat transfer towards the pool upper
interface as towards pool/concrete interfaces latter model also keeps a better consistency with CLARA
2D heat convection data, which is not possible in case of a high pool/crust interface temperature which
requires an increase of upwards convective heat transfer compare to lateral one in order to get a correct
prediction of the upwards radiative heat losses in 2D MCCI experiments. Such a dissymmetry could
be explained if higher viscosity is considered only for pool/concrete heat transfer traducing the
presence of a slag layer.
Table 3.3-2: Status on pool/concrete interface modelling
Interfaces Model Codes Available models/retained assumptions Limitations Impact on MCCI
type including its predictions
description
Pool/ “with All codes at Crust Tsolidification3 No stable crust Medium impact
concrete crust” least as an Between Tsolidus (CORCON, SOCRAT, WECHSL,..) and observed in exp. on pool
interface model option Tliquidus (TOLBIAC-ICB) No mechanistic temperature and
(excepted Fm,vol,threshold ~0.5 (MEDICIS-IRSN: model 1) model for the ablation kinetics
MEDICIS- Meshed crust and mushy zone in SOCRAT influence of
GRS) concrete type
slag layer model between crust and concrete : hint2 ~ from 300 No exp. data Low impact
W/m2/K to infinite (TOLBIAC-ICB) excepted if crust
is absent
“no MEDICIS For MEDICIS-IRSN, MEDICIS-GRS: hint2,bottom: 80 to 300 No mechanistic High impact on
crust” IRSN, W/m2/K depending on concrete type; hint2,lateral: 200 to 300 model for the 2D ablation
model MEDICIS- W/m2/K depending on concrete type; MEDICIS-IRSN (model 2): influence of profile
GRS, hint2 combined with hconv1; MEDICIS-GRS: hint includes concrete type
CORCON, convection
CORQUENCH CORCON, CORQUENCH: gas film model, modified Bradley’s
(as an option) correlation, Sevón’s correlations
Upper “with All codes Crust Tsolidification between Tsolidus (CORCON, SOCRAT, Uncertainty on High impact on
pool crust” excepted WECHSL,..) and Tliquidus (TOLBIAC-ICB); Fm,vol,threshold Tsolidification3 ablation kinetics
interface model MEDICIS- ~0.5 (MEDICIS-IRSN); meshed crust in SOCRAT and on radiation because it
GRS Radiative heat transfer (Stefan’s law) absorption by determines the
aerosols upwards heat
“no MEDICIS- Effective heat transfer derived from exp. Predicts no losses
crust” GRS upper crust;
model unreliable
extrapolation to
reactor case
1: hconv: convective heat transfer coefficient within bulk pool; 2: hint: heat transfer coefficient at pool/concrete interface;
3: crust formation temperature
1. This point is still an open issue as in MACE or CCI tests, some post-test measurements of uranium distribution inside the
melt show higher uranium concentration near the bottom concrete interface. It is nevertheless difficult to conclude between
the existence of a solid crust during the transient or the possibility to have a segregation process during the final cool-down.
NEA/CSNI/R(2016)15
3.3.3. Coolability models

Features and limitations of available coolability models are displayed for all phenomena (bulk cooling
before crust build-up onset, film boiling heat transfer and radiation before crust quenching onset, water
ingression and melt eruption after crust quenching onset) are listed on Table 3.3-3.
Specific limitations of models implemented only in some codes are outlined in red. Overall
model limitations are outlined in purple.
Taking into account Table 3.3-3, following comments can be made on available coolability
models In case of water injection:
Only radiative heat transfer from the pool upper interface to water is described by all codes;
Film boiling heat transfer is not described by some codes; moreover the “classical” correlations
for the film boiling regime are not adequate for describing heat transfer during MCCI and the
transition from film boiling to nucleate boiling regime is very likely much faster than predicted by and
occurs at a much higher heat flux level than predicted by “classical” correlations.
Phenomena of water ingression and melt eruption are the prevailing ones. Generally speaking,
two types of models are used: either simplified models or more detailed models including a deeper
understanding of main physical mechanisms.
As far as water ingression is concerned, the simplified model based on Darcy’s law requires
knowing the permeability of the upper crust. The more detailed model (Lomperski & Farmer, 2007)is
able to evaluate this crust permeability in consistency with other physical variables determining the
crust cracking due to thermal gradient such as the corium crust thermal diffusivity and linear
expansion coefficient. However the values of some physical parameters such as the crust temperature
at onset of cracking and other mechanical parameters taking into account the impact of crust material
are required.
As far as melt eruption is concerned, the simplified model based on Ricou-Spalding’s correlation
(Ricou & Spalding, 1961) is easy to use and accounts for both the variation of the gas and corium pool
density. Main uncertainty is on the value of this correlation for a reactor situation with real material
properties and a large scale floating upper crust. The detailed model taking into account the detailed
hole geometry (Tourniaire B. , Seiler, Bonnet, & Amblard, 2006) (mainly hole diameter and surface
density) permits a mechanistic description of melt eruption hydrodynamics but in spite of numerous
models involving in particular energy balance considerations with corium non-freezing conditions,
non-flooding and non-gas-build-up conditions (Farmer, Phenomenological Modeling of the Melt
Eruption Cooling Mechanism during Molten Corium Concrete Interaction, 2006), (Robb & Corradini,
2010), no really mechanistic modelling is able to predict the hole geometry in the reactor case.
NEA/CSNI/R(2016)15
Table 3.3-3: Status on coolability models

Phenomenon Model Codes Available models/retained assumptions Limitations Impact on
type including its MCCI
description predictions
Heat transfer Bulk All codes Radiative heat transfer (Stefan’s law) in all codes Uncertainties on Medium
before crust cooling (excepted for MEDICIS-GRS where an effective radiation because of
build-up onset heat transfer coefficient is used) absorption by the short
aerosols time period
combined with bubbling enhancement (Farmer,
2010)in CORCON and CORQUENCH codes
Heat transfer Film CORCON, Transition film boiling curve film boiling regime Heat flux in film Medium
after crust boiling CORQUENCH, based on Berenson correlation (Berenson, 1961) boiling regime because of
build-up onset TOLBIAC-ICB, during quenching the short
and before MAAP, In SOCRAT film boiling is introduced by of a corium pool time period
quenching SOCRAT dependence of effective heat transfer coefficient must be increased
onset vs time compared to
literature
Film boiling heat flux in MAAP is imposed as the correlations
lower limit of the heat flux from the corium to
water.
Ignored in MEDICIS-IRSN, COCO, WECHSL
Radiative All codes Radiative heat transfer (Stefan’s law) in all codes Medium
heat (excepted for MEDICIS-GRS where an effective because of
transfer heat transfer coefficient is used) the short
time period
Water All codes, Only simplified model : Darcy’s law in MEDICIS, Uncertainty on Low
ingression CORCON, COCO, SOCRAT; Kutateladze model (Kutateladze crust permeability excepted in
Heat transfer COSACO, & Malenkov, 1978)in MAAP value in simplified the early
after crust WECHSL model; phase as
quenching excepted Detailed model: CORQUENCH, TOLBIAC-ICB, Uncertainties on long as
onset MAAP thermo-mechanical concrete
parameters of oxide
detailed models fraction is
low
Melt All codes, Only simplified model (Ricou-Spalding (Ricou & Proportionality Very high
eruption CORCON, Spalding, 1961): ASTEC-MEDICIS, COCO, factor in simplified
COSACO, MAAP, SOCRAT model is very
WECHSL uncertain and
excepted Detailed model PERCOLA type (Tourniaire B. , questionable; large
Seiler, Bonnet, & Amblard, 2006) with hole uncertainties in
geometry: CORQUENCH, TOLBIAC-ICB detailed models:
no real
mechanistic model
for hole geometry
NEA/CSNI/R(2016)15
3.3.4. Material thermophysical and thermodynamic properties

3.3.4.1. Thermophysical properties
Excepted for the viscosity, thermophysical properties of mixtures are generally evaluated by codes
from the linear combination of properties of individual species weighted by their mass, molar or
volume fractions 2.
The evaluation of viscosity is different due to the high impacts of respectively the silica fraction
on the liquid oxide phase viscosity and the solid mass or volume fraction on a partially molten
corium/concrete mixture.
From a general point of view the accuracy of the thermo-physical property models can be
considered to be sufficient, except for the viscosity of a solid liquid mixture because of large
discrepancies between models of the impact of the solid fraction. In particular results obtained using
Stedman’s model or Ramacciotti’s one can lead to a discrepancy between two models of one order of
magnitude on the viscosity taking into account the solid fraction.
3.3.4.2. Thermo-chemical properties
Obtaining accurate thermo-chemical data for a oxide/metal corium mixture (in particular enthalpy and
molten fraction) is of prime importance for evaluating the pool temperatures and material transport
properties, which strongly influence heat transfer distribution within the corium pool.
Thermo-chemical data can be generated using a thermochemistry computing tool or simplified
correlations fitted on available data.
The used methods for computing thermochemistry data are listed in following Table 3.3-4; three
methods exist in practice for generating thermochemistry data for the corium/concrete mixture during
MCCI:
• interface with a thermochemistry tool: determining molten fraction versus temperature for a
corium composition ranging from the pure initial corium up to a pure concrete using the
thermo-chemistry tool, generate a thermochemistry database for the whole required
temperature and corium composition ranges and interpolate from this database during the
MCCI calculation; this method was chosen by ASTEC/MEDICIS using the GEMINI2 solver
and a detailed thermo-chemical database such as NUCLEA as in the case of
ASTEC/MEDICIS code; this method is a good compromise both keeping a sufficient accuracy
in generated data and minimising the computing time;
• internal coupling with a thermochemistry tool: required thermochemistry data (mainly molten
fraction, enthalpy) are computed every time interval during the MCCI calculation as in the
case of COSACO, CORCON, and TOLBIAC-ICB codes; this method is obviously the most
precise but at the cost of an extra computing time due to the use of a Gibbs minimiser, which
can often generate convergence problems; for TOLBIAC-ICB codes, when fast calculations
are needed, this coupling can be shunted by using pseudo-binary diagrams;
• interpolation between data obtained from an internal thermochemistry database or from
empirical correlations (in particular for solidus and liquidus temperatures) as in the case of
COCO, MAAP, WECHSL, SOCRAT codes; this method is the most simple but also the less
precise and its use becomes difficult in case of corium with a specific composition such as in
2. It must be stressed that mass, molar and volume fractions are significantly different in MCCI melts due to the
large density differences of (U,Zr)O2 - around 8000 kg/m3 - and concrete -around 2000 kg/m3-.
NEA/CSNI/R(2016)15
some high temperature simulant experiments or with a concrete involving significant fractions
of some particular minor species, which are not taken into account in the available database.
Methods available for evaluation thermochemistry data in MCCI codes are displayed on the next
table:
Table 3.3-4: Methods available for evaluating in MCCI codes thermochemistry data during MCCI
Code Interface with Coupling with thermochemistry Obtained from thermochemistry
thermochemistry tool tool (Gibbs minimiser+ database) database or correlations
(outside MCCI code)
ASTEC/ MEDICIS Yes: GEMINI2+NUCLEA No No

database
COCO No No Yes
CORCON/
No Yes1 Yes1
MELCOR
CORIUM-2D
CORQUENCH No No Yes
COSACO No Yes No
MAAP No No Yes
SOCRAT No No Yes
Yes: GEMINI2+NUCLEA Yes: GEMINI2+NUCLEA database
TOLBIAC -ICB Yes (default option)
database (optional)
WECHSL No No Yes
1: CORCON combines uses of a thermochemistry tool and of correlations?
3.3.4.3. Concrete ablation

The ablation process is very likely mainly related to the melting process of concrete components,
which might be the prevailing mechanism leading to the concrete ablation. More details on the
concrete ablation process are given in section 1.3.2 (Concrete characteristics).
The postulated threshold temperature permitting the concrete mixing with corium is called the
ablation temperature.
Values of ablation enthalpy (variation of enthalpy from ambient temperature up to ablation
temperature) and ablation temperature corresponding to the basemat concrete are essential to evaluate
the ablation velocity and so are required to perform a MCCI calculation. These physical are either
user’s input values or evaluated by the code itself.
In the first option (user’s input), they are given as user’s input. They must be derived from
empirical data if available or evaluated using an external thermo-chemistry tool, which permits to
calculate the concrete molten fraction and the concrete enthalpy versus temperature. More precisely
the ablation temperature can be deduced from a volumetric solid fraction threshold for the mixture of
all concrete oxides or for a given component (e.g. mortar).
In the second option (tabulation in the code), empirical values of ablation temperature and
ablation enthalpy are tabulated for a given set of concretes.
In the third option (calculation by the code), values of ablation temperature and ablation enthalpy
are determined from a thermochemistry tool which is internal to the code or coupled with the code
during the MCCI calculation.
NEA/CSNI/R(2016)15
Main information on the evaluation of concrete ablation temperature and ablation enthalpy in
each code is summed-up in Table 3.3-5.
Concrete ablation is assumed to occur at a single temperature although concrete is a
multicomponent mixture melting over a large temperature interval. All MCCI codes use this type of
assumption. Nevertheless this ablation temperature is expected to be a function of the composition of
the different concrete components corresponding to the different size classes of sand, gravel and
cement, see table Table 1.3-3: Order of magnitudes of concrete constituent proportions in Section
1.3.2, in particular of their different gas contents, and of the melting behaviour of these components.
There is some experimental evidence that silica gravel and mortar (mixture of sand, cement and water)
have different behaviours during ablation.
Table 3.3-5: Methods available for evaluating in MCCI codes concrete
ablation temperature and enthalpy
Code User’s input Tabulation in the code Obtained from thermochemistry
tool in the code
Yes: evaluated from thermo-chemistry

ASTEC/ MEDICIS No No
outside the code
COCO Yes No No
CORCON /
Yes No No
MELCOR
CORIUM-2D
CORQUENCH
COSACO
MAAP Yes No No
Yes – for impact of water
evaporation and water/CO2
SOCRAT Yes No
chemical release in ablation
enthalpy
TOLBIAC -ICB Yes No No
WECHSL Yes? No No
Considering the way MCCI is modelled in severe accident codes, ablation temperature has a large
impact on MCCI calculations for following reasons:
• It influences the interface temperature of the different pool/concrete interfaces and then the
heat flux distribution since the pool interface temperature may be different for upper pool
interface (not depending on the ablation temperature), the lateral one (equal to the ablation
temperature in the absence of stable crust) and at the bottom oxide interface, possibly
depending on the ablation temperature in case of a pool stratification or in the absence of
stable crust.
• It influences also the bulk pool temperature because of the impact on the corium pool
interface temperature at least in case of no stable crust and again this impacts the energy
balance because of the energy required to heat the concrete to the bulk pool temperature.
We recall here the classical formulation of corium pool energy balance during MCCI. The energy
balance equation may be written as:
 dh  dmabl
m  = P decay − Pupwards − .[hconc (T ) − hconc (T0 )] Equation 3.3-6
 dt  dt
NEA/CSNI/R(2016)15
where:
- m the mass of the corium pool, h the corium enthalpy, P decay the decay power, Pupwards the
upwards heat loss rate,
dmabl
- .[hconc (T ) − hconc (T0 )] is the heat sink due to ablation and heating of the ablated concrete
dt and chemical reactions,
mass
dm
- abl is the ablated mass rate, hconc ( T ) the enthalpy of ablated concrete at the bulk pool
dt
temperature T, Tabl the ablation temperature and T0 the ambient temperature (298K),
dm ∫ φ .dS
with: abl
= is the mass ablation rate, φ is the local heat flux at the corium/concrete
interface, ∫ φ .dS∆his
dt abl the integral of this heat flux along the concrete ablation boundary and
∆habl = (hconc (Tabl ) − hconc (T0 )) is the concrete enthalpy variation from the ambient temperature
up the ablation temperature.
When examining the previous energy balance equation, it appears that the value of the ablation
temperature impacts the ablation enthalpy and also the local heat flux (by imposing the pool/concrete
interface temperature) and influences clearly the energy balance. The exception is the case where the
pool interface temperature is not dependent on the ablation temperature because of a stable crust
built-up at a solidification temperature on each concrete interface during the whole MCCI phase and
in the absence of oxide/metal stratification. Hence the need appeared since a long time to evaluate
precisely the concrete ablation temperature in order to achieve realistic MCCI calculations.
Available values of the concrete ablation temperature are very empirical and were obtained often
indirectly from first MCCI experiments (Peehs, Skokan, & Reimann, 1979), (Skokan, Holleck, &
Peehs, 1979), (Powers & Arellano, 1982-1) (Powers & Arellano, 1982-2) and then later from some
extended programs: (Alsmeyer H. , et al., 1995), (Thompson, Farmer, Fink, Armstrong, & Spencer,
1997). Most empirical values of concrete ablation temperatures were found within the range: 1 500 to
1 700K.
However due to the lack of precise data for each concrete type, a more reliable evaluation
procedure was needed. This was the purpose of a recent work at IRSN (Barrachin, 2011) towards
defining this evaluation procedure. In the frame of this work, a wide review of experimental data
deduced not only from large-scale MCCI tests but also from past thermodynamic studies and separate
effect experiments was performed in order to estimate the ablation temperature for the different types
of concretes defined by their mass ratio (SiO2/(SiO2+CaO) called “r” (calcareous with a r value below
0.3, limestone common sand with an r range between around 0.35 and 0.65, siliceous concrete with a
“r” value above around 0.75). Detailed compositions of cements, aggregates and mortars (mixture of
cement and sand made of finer aggregates of size below 4 to 6 mm) for a large set of real concretes of
French PWR reactors were taken into account. Some features for a first set of limestone-sand (LCS)
concretes including typical examples of French PWR reactors (named RLCS1 to RLCS4), LSL
concrete (LCS concrete used in ACE tests) and VULCANO VB-U6 concrete are documented in the
table Table 3.3-6.
NEA/CSNI/R(2016)15
Table 3.3-6: Features of LCS concretes

ratio of
Concrete name Type of concrete components ratio of mortar
concrete
RLCS1 calcareous aggregates / siliceous mortar 0.376 0.567
RLCS2 calcareous aggregates / siliceous mortar 0.443 0.616
RLCS3 silico-calcareous aggregates / silico-calcareous mortar 0.485 0.420
RLCS4 silico-calcareous aggregates / siliceous mortar 0.646 0.554
LSL calcareous aggregates / siliceous mortar 0.53 0.715
VB-U6 silico-calcareous aggregates / silico-calcareous mortar 0.379 0.37
Some features for a second set of siliceous concretes including typical examples of French PWR
reactors (named RS1 to RS4), LSL and VULCANO VB-U5 concrete are documented in Table 3.3-7.
In this table, an example of a hybrid concrete with siliceous aggregates and calcareous mortar (called
RH1) is also provided.
Table 3.3-7: Features of siliceous concretes

ratio of
Concrete name Type of concrete components ratio of mortar
concrete
RS1 siliceous aggregates/siliceous mortar 0.791 0.687
VB-U5 siliceous aggregates/siliceous mortar 0.797 0.69
RH1 siliceous aggregates/calcareous mortar 0.591 0.302
For the calcareous refractory concretes, this analysis showed that a specific concrete ablation
temperature cannot be really determined on the basis of the available data. Indeed if a criterion based
on the liquid fraction is retained, the thermodynamic calculations logically lead to high “ablation”
temperatures which are not really observed in MCCI tests. For these MCCI tests, the experimental
observations seem to enhance the important role played by some species of the corium (the iron oxides
in particular) on the concrete degradation. When the concrete is refractory, chemical interaction with
some specific species of the corium or preferential interaction between some species of the concrete
may contribute to lowering the ablation temperature which is intrinsically high due to the high CaO
content in the calcareous concrete.
Figure 3.3-7: Liquefaction curves of some tested (and French PWR) limestone
common sand concrete mortars (NUCLEA calculations)
NEA/CSNI/R(2016)15
For the limestone common sand (LCS) concretes, the analysis (see Figure 3.3-7) showed that a
specific concrete ablation temperature for this type of concrete defined from an absolute criterion does
not really exist. It really depends on the chemical composition of the aggregates. An LCS concrete
with a lower r ratio (e.g. RLCS1 reactor concrete) contains only siliceous sand of low granulometry
but a large amount of calcareous aggregates, which are dispersed into powder at decarbonation and
mixed with other concrete components. So separation between mortar and aggregates does not occur
during melting and the mortar melting alone cannot be a relevant indication for the ablation process.
The bulk concrete with a higher calcia content (see RLCS1) has to be considered and the ablation
temperature derived from this bulk concrete melting is much higher than that of RLCS2 concrete
(see Figure 3.3-7).
As for calcareous concretes, other factors such as the detailed corium composition could
eventually play a role in the reduction of the ablation temperature. In case of an LCS concrete with a
higher r ratio (e.g. RLCS3 and RLCS4 reactor concretes), the undissolved part of large aggregates at
the time of the ablation process could be more significant since both concretes contain a non-
negligible content of refractory siliceous gravels. For both of these latter concretes, defining the
ablation temperature from the mortar melting could be more relevant than defining it from the bulk
concrete composition.
Nevertheless for most LCS concretes, the ablation temperature can be determined as the
temperature corresponding to 50% of liquid in the concrete at least for LCS concretes with a high
enough silica content (e.g. RLCS2 reactor concrete).
For the siliceous concretes which are frequently encountered in particular in French PWRs, the
ablation temperature can be chosen in the temperature interval between 1 650 and 1 700 K on the basis
of the liquefaction of the mortar (see Figure 3.3-8), since the refractory siliceous aggregates remain
solid and separate from mortar during concrete melting. Moreover taking into account the uncertainty
of alumina content in the French PWR mortar could lead to a reduction up to 100 K, i.e. 1 550-
1 600 K.
No rule could be proposed in case of untypical “hybrid” concretes with for example a refractory
mortar rich in CaO but a high fraction of mostly siliceous aggregates (see RH1 in Table 3.3-7). For
this latter concrete type, a temperature of 2000 K is recommended for lack of a better model.
Efforts for improving the method for the ablation temperature assessment should be focused on
following aspects:
1. Confirmation of the obtained preliminary trends are required, especially on the impact of
mortar melting on the ablation process in case of silica aggregates.
2. Application to intermediate concretes (between siliceous and more calcareous LCS concrete
types), or different concretes such as CCI2 (significant MgO fraction).
3. Identification and quantification of other phenomena or factors determining the ablation,
e.g. mechanical behaviour or liquefaction curve of bigger aggregates (calcareous and
siliceous), investigation of the influence of a third major species such as iron oxide in addition
to other major species (silica, calcia).
4. Last but not the least study of the influence of iron bars on the ablation process, which remain
so far mostly unknown.
NEA/CSNI/R(2016)15
Figure 3.3-8: Liquefaction curves of some French PWR siliceous concrete mortars
(NUCLEA calculations)
3.4. Code application limitations
3.4.1. Model simplifications

Several simplifications appear in the models of several codes:
1. Heat conduction is neglected behind the ablation front in MEDICIS, COSACO, WECHSL
codes.
2. Corium masses stored in crusts along the pool/concrete interface are not taken into account in
several codes (e.g. MEDICIS).
3. Main transport properties of mixtures (viscosity excepted) are evaluated using a mass or
volume averaging on the individual species; note that mass and volume average yield to fairly
different results due to the large spectrum of densities within core-concrete mixture
constituents.
4. The cavity shape may be non-realistic if some simplified assumptions are used, such as a
cylindrical cavity as in CORQUENCH codes.
5. The attenuation of radiative heat flux above the corium pool by aerosols is described by a
user’s parameter or a very simple evaluation of an attenuation factor.
6. The structure of the pool/concrete interface is described in all codes by rather very simple
models (stable crust or absence of crust with an imposed effective heat transfer coefficient
depending on the interface orientation or a gas film model).
The simplifications 1 to 3 are reasonable and the impact is limited at least in the quasi-state
regime and at the reactor scale. Nevertheless, the assumption 1 can lead to a too fast ablation in
experiments with rather thin basemat such as VULCANO tests because conduction heat losses are
ignored.
The impact of simplification 4 might be significant but the uncertainties on the aerosol features
(concentration and size distribution) are too high to permit a gain with a more mechanistic model.
The impact of simplification 5 is questionable if the aim is to obtain precise predictions of
basemat melt-through in the long-term phase of a reactor case
NEA/CSNI/R(2016)15
The simplification 6 is questionable but the use of more complex models for the pool/concrete
interface would be difficult because the experimental database for validation is too scarce and it would
make the code complex because of the number of input additional physical parameters and possibly
numerical instabilities (e.g. in case of crust instability model). However a crust instability model is
available in CORQUENCH code (Farmer, 2010).
3.4.2. Types of scenarios which can be described
Outside model limitations given above, code limitations concern the types of scenarios which can be
described. Indeed the basis scenario of a corium pouring in a dry reactor pit assuming a fast corium
spreading over the whole available area can be described by any available MCCI code.
However scenarios involving a corium spreading phase and still more presence of water with a
more or less sustained water injection might change drastically the initial corium geometry and the
subsequent evolution of the ablated cavity:
• In case of a corium pouring on a reduced part of the total available area, the real extent of
corium spreading should be taken into account and might influence the ablation kinetics
because of the reduced contact area between the corium inventory and the bottom and lateral
concrete pit walls.
• In presence of water initially in the reactor pit, first this initial water amount will quench at
least a part and possibly the total corium inventory, thus delaying the corium melting and the
onset of concrete ablation but by approximately 1 hour only. Second the injected water might
also hinder the corium spreading and limit the corium/concrete interaction to a restricted area
thus leading to a local higher heat flux and faster concrete ablation at least before some re-
spreading occurs.
Most of MCCI codes are unable to model properly the corium spreading3. At least some of them
can describe the impact of water injection and more precisely of the top quenching, but keeping the
initial reactor pit geometry, excepted in case of MELCOR; in particular the possible restriction of
concrete ablation to a part of reactor pit walls due to a localised corium accumulation because of water
present in other places of the reactor pit cannot be treated or at the opposite the additional corium
spreading of corium outside the initial cavity for example in the case of a concept permitting the
spreading in additional cavities localised around the first one.
3.4.3. Possible reactor pit geometries
The choice of cavity or reactor pit geometries in MCCI codes is limited. Most codes can treat either
slab geometry (with two ablatable lateral walls) in case of some MCCI experiments (ANL-CCI
facility) or an axisymmetric cavity (cylinder or succession of truncated cones).
Nevertheless the treatment of non-axisymmetric cavities for example with a lateral corridor or
large sumps (as in Fukushima Daiichi drywells) is not possible; only adaptations keeping axisymmetry
and changing the surface over volume ratio can be performed but leading in fact to a rather unrealistic
shape. The main reason is that the lumped-parameter approach used in most MCCI codes and leading
to averaging physical quantities such as the corium temperature over the whole pool can be a
reasonable simplification only if assuming an axisymmetric pool.
3.4.4. Concrete features
A question arises: which concrete types can be addressed by MCCI codes?
3. In this case, specific spreading codes must be used
NEA/CSNI/R(2016)15
Respective impacts of siliceous and LCS concretes on MCCI behaviours are known even though
the ablation pattern of concrete of intermediate compositions between the studied silica-rich and
limestone-rich concretes is not clearly known. Moreover the impact of another concrete type
(e.g. serpentinite) on 2D ablation behaviour is unknown so far. In this case, in order to get
conservative results and the shortest melt-through delay, it is recommended at the present state of
knowledge to assume an isotropic heat flux distribution the mixed oxide/metal pool and in the oxide
layer during the stratified configuration phase. On the opposite, if there are concerns about lateral melt
through, anisotropic ablation should be considered.
3.5. Summary of overall code qualities and limitations
Taking into account the previous analysis and the experience from users, main code qualities and
limitations are tentatively synthesised in the following table. Only existing code versions are
considered; improvements planned in future versions are ignored.
Table 3.5-1: Overall evaluation of main code qualities and limitations
Code Level of model detail Model choice Validation state Flexibility of
flexibility boundary
conditions,
geometry, corium
pour scenarii
high for thermochemistry
and heat transfer models, almost completed for
high for dry MCCI high; limitation on initial
ASTEC/ no concrete conduction; dry MCCI models;
models, limited for cavity shape: no non-
MEDICIS simplified description of insufficient on
coolability models axisymmetric cavity
debris bed and melt coolability models
eruption
high for thermochemistry, various kinds of model extended for dry MCCI cavity: axisymmetric
COCO heat transfer models, low parameters are given models concrete (plus non-
for stratification model by input ablative wall)
high for thermochemistry,

high for pool config. high; limitation on initial
CORCON heat transfer and extended for dry MCCI
evolution, reduced for cavity shape: no non-
/MELCOR stratification models / no models
pool/ concrete interface axisymmetric cavity
coolability models
high for pool/concrete
interface, crust and extended for dry MCCI high; limitation on initial
high in particular for
CORQUENCH coolability models, no models; in progress on cavity shape: no non-
coolability models
oxide/metal models; coolability models axisymmetric cavity
simplified cavity shape
mass-energy very fast running mainly for in-vessel high; applicable to
conservation and ex-vessel axisymmetric cavities
CORIUM2D confinement structures;
extended for dry MCCI
models
high for thermochemistry in general limited; low
and heat transfer models, for pool/ concrete extended for dry MCCI
COSACO
low for pool config. interface and pool models
evolution and coolability config. evolution
parametric heat
high; Users can specify
high for pool/concrete transfer model (user-
pouring conditions such
interface, simplified heat provided heat transfer
extended for dry MCCI as compositions,
transfer distribution, no coefficients). Multiple
models; coolability duration and
MAAP oxide/metal pool models; cavity shape models to
models compared with temperature. limitation
multiple cavity shapes to choose from.
CCI-2 and CCI-3 tests on initial cavity shape:
choose from: high for Parametric or
non-axisymmetric
coolability models mechanistic coolablity
cavity is not allowed
models available
NEA/CSNI/R(2016)15
Code Level of model detail Model choice Validation state Flexibility of

flexibility boundary
conditions,
geometry, corium
pour scenarii
SOCRAT high for pool/concrete high – most model in progress high;
interface, lack of gas- parameters can be
driven convection; simple defined by user
coolability models
high; Users can specify
pouring conditions such
limited for pool/
high for thermochemistry as compositions,
concrete interface(crust
and heat transfer models, extended for dry MCCI duration and
stability and pool/crust
optional treatment of models; validated temperature; limitation
interface temperature)
TOLBIAC -ICB concrete conduction; against MACE and CCI on initial cavity shape;
and heat flux
simplified description of tests for coolability possible simultaneous
distribution in oxidic
debris bed and melt models MCCI at the same time
layer; high for
eruption in different rooms (with
coolability models
possible corium transfer
between rooms)
low flexibility in
high for heat transfer particular for pool
limitations on cavity
models)/ simplified interface models
extended for dry geometry and initial
WECHSL correlations for (crust stability and
MCCI models cavity shape, no
thermochemistry, no pool/crust interface
treatment of pouring?
coolability models temperature) and heat
flux distribution
NEA/CSNI/R(2016)15
4. Status of validation and model uncertainties
The preceding chapter gives a short description of a range of detailed models (e. g. for the pool
interface, the convection in the pool, the coolability in a flooded scenario etc.) which are currently
used in recent applications of the major MCCI simulation codes. These detailed models, which are
recommended for usage by the different code users as the result of the individual code validation
work, are often based on theoretical considerations for separate effects and validated for separate
effect experiments. However, MCCI is essentially a complex interaction of several phenomena and it
is important to know about the performance of the codes for integral MCCI experiments.
This chapter focusses not on detailed mechanistic models like those described in Chapter 3 but on
more higher level capabilities of the codes, which can be validated by comparing with available data
of MCCI experiments. Sections 4.1-4.3 aim at identifying these higher-level capabilities of MCCI
codes. Then Section 4.4 presents the validation status of the codes with view to these capabilities.
Models require the usage of material properties like the viscosity of the corium. Section 4.5 reports the
status of material properties validation. Section 4.6 discusses how the data of experiments can be
scaled to the reactor situation and how this is supported by available models in the codes.
4.1. General remarks on the use of data from integral experiments for code validation
As the interactions between corium and concrete are composed of a complex combination of several
physical phenomena a calculation performed by a code dedicated to MCCI usually involves a complex
interaction of several detailed models. MCCI experiments are integral experiments; because of that
they cannot serve as a validation for each detailed phenomenological model included in the codes.
Detailed models are usually validated in separate effect tests, but for the case of MCCI it is difficult to
define meaningful separate effect tests, since some phenomena can hardly be separated or isolated
from each other, like e. g. the heat transfer between melt and concrete interface and the melting of the
concrete. Only the combination/interaction of models used in the codes can finally be validated in
integral experiments. It is at first necessary to reflect how experimental data obtained from integral
MCCI experiments can be used for code validation.
Therefore it is necessary to understand the interaction of models in an MCCI code. The
development of MCCI codes should help to understand the real physics. Several dedicated MCCI
codes have evolved from long-term international R&D projects on MCCI in the past. It is interesting
to understand trends predicted by the models when applied to typical MCCI conditions. The following
section gives some insights for the exemplary case of the MCCI code MEDICIS (Cranga, Fabianelli,
Jacq, Barrachin, & Duval, 2005).
4.2. Considerations on the energy balance in MCCI
MCCI is characterised by a complex combination of heat and mass transfer processes. Neglecting
energy sources and sinks from chemical reactions, the energy balance for the corium pool under dry
conditions in MEDICIS is written in the form:
NEA/CSNI/R(2016)15
𝑑𝐻
= 𝑃𝑑𝑒𝑐𝑎𝑦 − 𝑃𝑟𝑎𝑑 − 𝑃𝑖𝑛𝑡,𝑐𝑜𝑛𝑐 + 𝑃𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥,𝑐𝑜𝑛𝑐
𝑑𝑡 Equation 4.2-1
− 𝑃𝑚𝑎𝑠𝑠𝑓𝑙𝑢𝑥,𝑔𝑎𝑠
Symbol (unit) Explanation Represented in Figure 4.2-1 as
H (J) Enthalpy of corium pool
Pdecay (W) Energy source due to internal decay power
Prad(W) Energy flux due to radiation from the free surface Red arrow at the free surface of the pool
Energy flux due to convection to the concrete

Pint,conc (W) Red arrows along the concrete interface
interface
Energy flux with the concrete mass flux released into

Pmassflux,conc (W) Grey arrows along the concrete interface
the pool at the interface
Energy flux with the gas mass flux released from the
Pmassflux,gas (W) Grey arrow at the free pool surface
pool at the free surface
Figure 4.2-1: Heat and mass transfer processes at the outer boundary
and internal interfaces of the MCCI pool (Spengler, 2013)
Transient calculations with the MCCI code MEDICIS for a homogeneous pool configuration (e.
g. the experiment CCI-2 from the NEA-MCCI Project) and using constant heat transfer coefficients
between corium and concrete (as found to be fairly consistent with several experiments (Cranga, et al.,
Towards an European consensus on possible causes of MCCI ablation anisotropy in oxidic pool, 2014)
have shown, that the calculated pool behaviour tends to a quasi-steady state and remains in this state
for a long term (if there were no transient boundary conditions), in which the internal decay power
Pdecay is nearly, except for a small quantity ∆P, balanced by convection to the concrete interface
(Pint,conc) and by release at the free surface (Prad) (Spengler, 2013), (Spengler, Fargette, Foit,
Agethen, & Cranga, 2013), see Figure 4.2-2:
∆𝑃 = 𝑃𝑑𝑒𝑐𝑎𝑦 − 𝑃𝑟𝑎𝑑 − 𝑃𝑖𝑛𝑡,𝑐𝑜𝑛𝑐 Equation 4.2-2
NEA/CSNI/R(2016)15
CCI2: Temperature CCI2: Energy balance
Reactor scenario: Temperature Reactor scenario: Energy balance

Figure 4.2-2: Pool temperatures and energy balances calculated with MEDICIS for the experiment
CCI2 (Spengler, Fargette, Foit, Agethen, & Cranga, 2013) and a reactor case (Gencheva, et al., 2012)
Since ∆P shows to be much smaller than the sum of Prad and Pint,conc in the MEDICIS calculation
during the steady state regime, Equation 4.2-2 can be simplified to
0 ≈ 𝑃𝑑𝑒𝑐𝑎𝑦 − 𝑃𝑟𝑎𝑑 − 𝑃𝑖𝑛𝑡,𝑐𝑜𝑛𝑐 Equation 4.2-3

for this quasi-steady state time period.
According to this finding the pool behaviour in MEDICIS can be approximated for a long term
period by a kind of quasi-steady state balance, using the balance Equation 4.2-3 (redistribution of
internal power to the interfaces) in a slowly varying geometry.
During the course of MCCI tests there are transient effects (e. g. at the early stage, when the melt
eventually loses its superheat or some limited zirconium or chromium oxidation takes place).
Nevertheless, it can be assumed that the overall process on the long term may be described by a
sequence of such quasi-steady states. Since the MCCI codes aim at predicting the long-term, the
sequence of quasi-steady states should be tracked by the codes. However, for specific questions like
the MCCI dynamics on the short term, e. g. for the MCCI in melt retention concepts including the use
of core catchers, the capturing of transient effects may be important, too.
NEA/CSNI/R(2016)15
Figure 4.2-3: Transient effect in MEDICIS calculations for the experiment CCI-2
due to variation of initial temperature
Figure 4.2-4: Slightly faster initial ablation velocity due to transient loss of melt
superheat in the calculation with elevated initial temperature
For illustration, the effect of a transient as calculated by an MCCI code (MEDICIS) is indicated
in Figure 4.2-3, where the pool temperature calculated with MEDICIS for the MCCI experiment
CCI-2 is plotted. The original calculation assumes T0 = 2 153 K as initial temperature of the melt
whereas in a variation T0 + 300 K is assumed. Both calculations join into a common quasi-steady state
after approx. 45 min. This quasi-steady state is determined by a rough equality of injected power and
heat fluxes to the surface according to Equation 4.2-3and the pool temperature adapts to the
effectiveness of heat transfer mechanisms (which may be a function of pool temperature) and the size
and shape of the growing interface area. The calculation with elevated initial temperature leads to
slightly faster ablation in the early stage, but the effect on final ablation depth is relatively small
(Figure 4.2-4).
From the observed trends of the MEDICIS code the following general statements can be derived:
• Steady state is given by an approximate compensation of the power injected in the melt with the
power transferred at the boundaries (interface to concrete, top surface) according to Equation
4.2-3 in a fixed geometry and with a constant power. If there were no concrete erosion the pool
temperature would stay constant in such steady state. However, in MCCI there is no steady state
since there is concrete erosion and the pool boundaries are moving.
• In quasi-steady state the geometry and the level of power injected (in experiment and reactor
case) vary slowly, so that the process can be described by a sequence of steady states due to
Equation 4.2-3. MCCI code calculations tend to such a quasi-steady state behaviour.
NEA/CSNI/R(2016)15
• However, the whole energy balance is respected only if Equation 4.2-1 is fulfilled, which results
under typical conditions in a temperature decrease over the whole time period considered.
• With typical heat transfer coefficients in the order of a few 100 W/(m2 K) it takes the order of
1 hour experimental time until a quasi-steady-state establishes. The preceding temperature
transient is affected by melt thermal properties like specific heat, density, latent heat evolution,
void fraction etc. Differences in material properties would lead to different courses of
temperature evolution (i. e. slower or faster) computed by the codes during such transient. This
makes the precise recalculation of the temperature transient in short-running experiments (~ order
of 1 h) very ambitious.
• Since the variation of geometry and decay power is slow during typical MCCI conditions, the
quasi-steady approach should be adequate for the long term.
• The validation of the complete energy balance, Equation 4.2-1, requires the calculation of correct
ablation volume (indicated by geometrical ablation depths) and correct pool temperature
behaviour.
• A validation of the different terms in the energy balance (including the relative magnitudes of
different heat fluxes) is in most cases not accurate, since neither the ablation volume (as a
measure for the integral of Pint,conc) nor the heat transferred from the top surface (integral of Prad)
are known in experiments with satisfying precision. However, assuming that the experiment is
running in a quasi-steady state regime (as the models in MEDICIS suggest for slowly varying
boundary conditions), the comparison of data for ablation depths and pool temperature may help
to give more differentiated statements about the model validation status:
1. If heat fluxes at interfaces (i. e., top surface or interface with concrete) as computed by the
codes are compared with experimental data (i. e. upward heat flux or erosion velocity) for a
quasi-steady state period (and not during transient periods) a validation may be obtained for
the basic capability of calculating the correct heat flux distribution in the codes resulting in a
correct prediction of concrete ablation (abl)
2. If – in addition to the previous item – the melt temperature evolution (mt) computed by the
codes is compared to experimental data during a quasi-steady state period a validation of the
codes may be obtained for the effective heat flux modelling q = heff (T- Tint), which is used
with potentially different model parameters heff and Tint for the various interfaces. If the
MCCI runs predominantly in a quasi-steady state regime, differences in code predictions for
pool temperature can be referred to different parameters heff and/or Tint for some or all
interfaces of the melt pool, presumed that the heat flux distribution is correct. No further
information on other models e. g. viscosity, solid fraction evolution, impact of superficial
gas velocity, interface morphology etc. can be deduced.
• The quasi-steady state obtained after a transient is finally governed by the level of internal power
release (which should be a common data for all codes), the computed power distribution to the
different interfaces (i. e. power split in sideward, downward and upward direction), the area of
the interfaces and the effective heat transfer coefficients approximated for the specific interface.
4.3. Identification of phenomenological models to be validated and useful experiments
Considering the findings in Section 4.2 a list of potential higher-level, phenomenological capabilities
(or models) that may be validated based on available experimental data are given in Table 4.3-1.
Experiments which are proposed for model validation based on the current understanding are listed in
Table 4.3-2. Not all phenomena are covered by each experiment. Available data for validating basic
code capabilities are mapped here to the individual experiments.
NEA/CSNI/R(2016)15
Table 4.3-1: List of phenomenological models in MCCI codes (i.e. basic capabilities) that can be
validated by integral MCCI experiments
Abbreviation Phenomenological Prerequisites and conditions for successful validation of specific model
model
abl Concrete ablation Progress of (selected) local ablation depths vs. time and the final 2D ablation contour is
(ablation depth and 2D well met by the simulation. However, pool temperatures predicted by the code may still
profile) deviate from the measured ones, this is subject of the following item (mt).
mt Melt temperature In addition to abl the pool temperatures are also well met in the simulation. This
requires that the assumptions for heat transfer (i. e. the product of calculated effective
heat transfer coefficients h with the computed driving temperature differences T) and
the evolution of geometry represent a good approach of the experimental data. The
calculated heat transfer coefficients h are very specific to the assumed interface
temperature. The assumption of higher interface temperatures would require larger h to
transfer the same heat flux to the boundary at smaller T.
model
gr Gas release from MCCI Gas release rates (CO, CO2, H2O, H2) and integral gas masses released during MCCI
are well simulated. It has to be checked that there is no bypass of gas release in the
experiments (i. e., that all the gas goes through the pool and participates in chemical
reactions) and that there is no release of water from the concrete before the MCCI in
the test preparation phase.
ox Oxidation reactions Oxidation reactions of metals are simulated in agreement with experimental
observations. This can be checked by comparing the species composition in the gas
mass fluxes released from the pool surface and/or by comparing the species
composition in the melt between calculation and experiments.
tf Top flooding In the situation of a corium pool flooded from the top there is a crust growing on top of
the surface of the pool and the phenomenology is different from the dry situation
(boiling condition at the exterior of the crust or water ingression into the crust). In past
research programs three special phenomena effecting the cooling performance were
investigated:
1. Bulk cooling: Initially there is an intense interaction between liquid melt without
significant crust on top of it and water which leads to peak heat fluxes.
2. Water ingression: In the flooded situation the heat flux between the growing top crust
of the corium and the water is assumed to be controlled by the dry-out heat flux limit
according to the theory of water ingression. This heat flux is elevated compared to
thermal conduction through the crust. The dry-out heat flux limit generally depends
on material properties of the crust.
3. Melt eruption: Since the growing top crust of corium is porous, liquid melt is driven
through the crust into the water pool above the corium, where it is fragmented and
efficiently cooled. The entrained mass flow depends on the geometry of holes in the
crust and on the superficial gas velocity. Resulting heat fluxes are elevated
compared to thermal conduction through the crust.
Available models for the overall effect of top flooding or for the detailed mechanisms
are validated on the basis of measured heat flux data at the melt/water or crust/water
interface. Since the time behaviour of phenomena 1, 2, 3 is different (1: initial peak
transient, 2: elevated heat flux plateau, 3: periodic peak transients) validation may be
obtained for the specific detailed models 1, 2, 3. For the melt eruption, model can be
validated on the determination of the mass of particles debris. There are limitations in
interpreting test results with top flooding, since in most experiments with top flooding
the top crust attached to the sidewalls and thus formed an unrealistic boundary
condition.
NEA/CSNI/R(2016)15
model
tc Top crust formation Due to cooling of the melt at the top (heat transfer by radiation or due to boiling of
water) a crust forms on top of the corium pool. A validation of the top crust
thickness requires that the growth of crust thickness is well approximated in the
codes, if compared with experimental data on final crust thicknesses. However,
the available experimental data on top crust thicknesses are only weak indications
for real crust thicknesses during the heating period.
st Stratification evolution Situation of MCCI with stratified metal layer/oxide layer established due to density
differences might have a strong impact on the ablation rates. Due to the
enrichment with lighter concrete decomposition products the oxide layer will be
settled in the long term on top of the metal layer. Mixing and stratification is highly
affected by the gas flow through the melt; for high gas superficial velocities, oxidic
and metallic phases will be fully mixed. The different potential configurations
(mixed or stratified) will have large impact on the resulting ablation behaviour. The
only validation can be obtained by comparison with the final experimental
configurations characterised by the post-mortem analyses.
re Effect of reinforcement Rebars have the potential to strongly influence the 2D course of the MCCI once
the metal at the bottom starts to freeze. A validation requires that the erosion of
concrete including iron rebars is well simulated in the code if compared with
available experimental data.
fp FP/aerosol release FP and aerosol release rates and integral masses released during MCCI are well
simulated.
inc Incubation Due to the thermal shock between hot melt and cold concrete at the time of corium
slumping from the RPV an instantaneous crust may be formed and may attach to
the concrete wall. In experiments this crust is not effectively heated in contrast to
the liquid melt. This initial crust insulates thermally the melt pool. Heat is
transferred from the melt through the crust into the concrete but the erosion
(liquefaction of concrete) is slow or non-existing. In quasi-steady state the initial
crusts will be no longer present (depending on local heat fluxes). In this case the
erosion velocity is elevated compared to early times. A successful validation for
the incubation phenomena should predict the transient formation and failure of
contact resistances as indicated by tc-measurements in experiments.
Table 4.3-2: Integral MCCI experiments proposed for validation of specific models
Provides data for

Initial melt Concrete
Test series No. phenomenological Geometry
composition type
models
SURC (Copus & Powers,
1 abl, mt, gr, ox, fp, inc UO2, ZrO2 + Zr LCS 1D cyl.
1992)
2 abl, mt, gr, ox, fp, inc UO2, ZrO2 + Zr Basaltic 1D cyl.
4 abl, mt, gr, ox, fp, inc Steel + Zr Basaltic 1D cyl.
SWISS
(Blose, Gronager, Suo- 1 abl Steel LCS 1D cyl.
Anttila, & Brockman, 1987)
2 abl, gr, ox, tf Steel LCS 1D cyl.
WETCOR (NRC, 1993) 1 abl, mt, tc, tf Alumina based oxide LCS 1D cyl.
BETA (Alsmeyer, et al., Steel + alumina based
V1.8 abl, mt, gr, Siliceous 2D cyl.
1992) oxide
Steel + alumina based
V2.3 abl, mt, gr, Siliceous 2D cyl.
oxide
NEA/CSNI/R(2016)15
Provides data for

composition type
models
V3.2 abl, mt, gr,cs LCS 2D cyl.
oxide
V3.3 abl, mt, gr LCS 2D cyl.
oxide
V5.1 abl, mt, gr, ox Siliceous 2D cyl.
oxide + Zr
V5.2 abl, mt, gr, ox, fp Siliceous 2D cyl.
oxide + Zr
V5.3 abl, mt, gr, ox, fp Siliceous 2D cyl.
oxide + Zr
ACE (Sehgal & Spencer,
ACE Program Phase C:
Fission Product Release
L2 abl, mt, gr, ox, fp Oxidic corium + Zr Siliceous 1D rect.
from Molten Corium
Concrete Interaction MCCI,,
1992)
L5 abl, mt, gr, ox, fp Oxidic corium LCS 1D rect.
L6 abl, mt, gr, ox, fp Oxidic corium + Zr Siliceous 1D rect.
L8 abl, mt, gr, ox, fp Oxidic corium + Zr Limestone 1D rect.
MACE (Farmer, Kilsdonk, &
Aeschlimann, 2009)
M3b abl, mt, gr, ox, tf, tc Oxidic corium + Cr LCS 1D rect.
(Farmer, Spencer, Binder, &
Hill, 2001)
M4 abl, mt, gr,ox, tf, tc Oxidic corium + Cr Siliceous 1D rect.
M1b abl, mt, gr,ox, tf, tc Oxidic corium + Zr LCS 1D rect.
1D rect.
MSET1 abl, mt, tf, tc Oxidic corium + Cr Inert
COTELS (Maruyama, et al.,

D-6 abl Steel Basaltic 2D cyl.
2006)
COMET (Sdouz, et al.,
2006) (Alsmeyer, et al., L2 abl Siliceous 2D cyl.
oxide
2007)
L3 abl, tf Siliceous 2D cyl
oxide
VULCANO (Journeau C. , et
al., 2012) (Journeau C. , et VB-U5 Abl, (mt) Oxidic corium Siliceous 2D,1/2 cyl.
al., 2010)
VB-U6 Abl, (mt) Oxidic corium Siliceous 2D,1/2 cyl.
Ferro-
VB-U7 Abl, (mt) Oxidic corium 2D,1/2 cyl.
siliceous
limestone-
VBS-U1 abl,(st), (ox) Oxidic corium + steel 2D,1/2 cyl.
rich
VBS-U3
abl, (st), (ox) Oxidic corium + steel silica-rich 2D,1/2 cyl.
VBS-U4
NEA/CSNI/R(2016)15
Provides data for

composition type
models
CCI (Farmer, Lomperski,
Kilsdonk, & Aeschlimann,
OECD MCCI Project: 2-D
Core Concrete Interaction
2 abl, mt, inc, tf Oxidic corium + Cr LCS 2D rect.
(CCI) Tests, Final Report,
2006) (Farmer, Lomperski,
Kilsdonk, & Aeschlimann,
2010)
3 abl, mt, tf, tc, inc Oxidic corium + Cr Siliceous 2D rect.
4 abl, mt, inc Oxidic corium + Cr LCS 2D rect.
6 abl, mt, bc, tf, inc Oxidic corium + Cr Siliceous 2D rect.
Siliceous
SICOPS (Hellmann & (85%),
Fischer, 2007) (Hellmann, 9e abl, mt ZrO2, CaO, Al2O3 limestone 1D cyl.
2000) cement
(15%)
V3 abl, mt, inc Oxidic corium siliceous 2D inverse
V6 abl, mt, inc Oxidic corium siliceous 2D inverse
MOCKA (Foit, Cron, Fluhrer,
Miassoedov, & Wenz, 2012) 1.6 abl Siliceous 2D cyl.
oxide + Zr
(Foit, T., & Fluhrer, 2013)
1.7 abl steel + alumina based Siliceous 2D cyl.
oxide + Zr
4.4. Code validation status
Having clarified in the previous section, which information on higher-level phenomenological models
can be derived from available experimental data, this section presents the status of code validation
with regard to available information.
4.4.1. Status of individual codes
The individual status of code validation is described from a general point of view in this section for
each code (in alphabetical order) separately. Additional details are provided with regard to the higher-
level phenomenological model list of Table 4.3-1 in Appendix 7.2.
4.4.1.1. COCO
The COCO code is validated through the comparison with tests in the COTELS project, which was
conducted by NUPEC (Maruyama, et al., 2006), and tests in the NEA MCCI Project. The tests D-6
and D-11 in the COTELS project and tests CCI-2 to -6 in the NEA MCCI Project were analysed by
the COCO code.
4.4.1.2. CORCON
The CORCON module of the MELCOR code, which models core-concrete interaction phenomena,
has been validated in the past against a considerably large set of experiments. A detailed description of
the MELCOR code can be found in reference (Gauntt, 2005) and a detailed description of the
CORCON module can be found in reference (Bradley, Gardner, Brockmann, & Griffith, 1993).
NEA/CSNI/R(2016)15
This CORCON validation set includes the SURC (Copus & Powers, 1992), SWISS (Blose,
Gronager, Suo-Anttila, & Brockman, 1987), and WETCOR (NRC, 1993) experiments performed at
the Sandia National Laboratories (SNL), BETA tests performed at the Karlsrühe Institute of
Technology (KIT) (Alsmeyer, et al., 1992), and ACE and MACE tests performed at the Argonne
National Laboratory (ANL) (Sehgal & Spencer, ACE Program Phase C: Fission Product Release from
Molten Corium Concrete Interaction MCCI,, 1992). These tests cover a broad range of input power
conditions, different types of concrete (basaltic, limestone, limestone-common sand, and siliceous),
and both metallic and oxidic melts. The chemical reactions of zirconium (and other metals) at high
temperatures were also investigated in some of these tests. Moreover, several MCCI experiments were
conducted with an overlying water (Spencer, et al., 1992), (Sdouz, et al., 2006) (Alsmeyer, et al.,
2007), (Journeau C. , et al., 2012), (Journeau C. , et al., 2010), (Farmer, Lomperski, Kilsdonk, &
Aeschlimann, OECD MCCI Project: 2-D Core Concrete Interaction (CCI) Tests, Final Report, 2006),
(Farmer, Lomperski, Kilsdonk, & Aeschlimann, 2010). These tests provide important information on
the thermal-hydraulic behaviour of concrete basemat and fission product releases in the presence of
overlying water.
Assessment of CORCON against the MCCI experimental data fall into three main categories:
• Heat transfer between melt and concrete, and between melt and overlying water, if present,
and basemat ablation from core concrete interactions.
• Chemical reactions and gas release (mainly flammable gas production) due to core concrete
interactions.
• Fission product release.
The CORCON module can be used in stand-alone mode with appropriate input and boundary
conditions specified. Alternatively, it can be used within the cavity (CAV) package of the integral
code MELCOR.
Recent applications of MELCOR to address ex-vessel debris coolability by an overlying water
pool treated heat flux partitioning between melt, concrete, and overlying water in a parametric manner.
Though such treatment yielded reasonably satisfactory results in the prediction of transient core debris
cooling rates in experiments, uncertainties in parametric modelling were recognised. That is, a stable
water-impervious interfacial crust is assumed to form between the core debris and overlying coolant,
so even with this modelling upgrade the long term core-concrete interaction will continue unabated
without the possibility of quenching and stabilising the core debris.
4.4.1.3. CORQUENCH
CORQUENCH validation has principally focused on reactor material experiments since the code does
not currently possess the capability to import or evaluate material properties of simulants. Referring to
Table 4.3-2, in terms of dry cavity experiments, the validation matrix includes tests L2, L5, L6, and L8
of the ACE/MCCI test series, the SURC-1 and -2 tests conducted as part of the SURC test series, and
finally the CCI-2, -3, and -4 tests conducted as part of the NEA MCCI-2 Program. In terms of wet
cavity tests, the matrix includes the M1b, M3b, and M4 tests conducted as part of the MACE Program;
the CCI-2, -3 (late-flooded) tests conducted as part of the MCCI-1 Program, and finally the single
large scale CCI-6 test featuring early cavity flooding that was carried out as part of the NEA MCCI-2
Program. A complete description of the validation activities is provided in (Farmer, 2010).
4.4.1.4. COSACO
The COSACO code has been validated against a large number of 1D and 2D MCCI tests including:
MACE-M3b, MACE-M4, ACE-L5, NEA-CCI-1, BETA 5.2, CORESA 2.1, SURC4, VULCANO
VB-U5 and VB-U6. The calculated concrete ablation behaviour and melt temperature show an
NEA/CSNI/R(2016)15
adequate consistency with the experimental results. Furthermore, the results of various benchmark
calculations show good consistency with other MCCI codes.
On this basis, it is concluded that the general treatment of thermochemistry in COSACO and the
thermal-hydraulic models specifically implemented for the oxidic and metallic melts are well-suited to
analyse the characteristics of the MCCI process.
4.4.1.5. MAAP
The MCCI model in MAAP has been validated against a number of dry cavity MCCI tests:
• ACE tests L2, L5, L6 and L7 performed by Argonne National Laboratory ( (Thompson &
Fink, 1988), (Thompson & Fink, 1989), (Thompson & Fink, 1991a), (Thompson & Fink,
1991b)
• SURC-4 test by Sandia National Laboratory (Copus, Blose, Brockmann, Gopmex, & Lucero,
Core-Concrete Interactions using Molten Steel with Zirconium on a basaltic Basemat: The
SURC-4 Experiment, 1989)
• BETA test series V5.1, V5.2 and V6.1 by Kernforschungszentrum Karlsrühe (see (Alsmeyer,
et al., 1992) and (Alsmeyer & Firnhaber, 1990)
• CCI test series CCI-2, CCI-3 by Argonne National Laboratory (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, OECD MCCI Project: 2-D Core Concrete Interaction (CCI)
Tests, Final Report, 2006)
4.4.1.6. MEDICIS
The validation work of IRSN and GRS has recently been focused on experimental data from 2D
MCCI oxidic tests (NEA CCI-2,3,4,6 (Cranga, Mun, Michel, Duval, & Barrachin, 2008), (Cranga,
Mun, & Marchetto, 2010); VULCANO VB-U5,-U6 (Journeau C. , et al., 2012), LACOMERA-
COMET L2, L3 (Spindler, et al., 2007) and also by GRS only (Spengler, 2012). The analyses of these
experiments led to the proposal of a parametric approach for the 2D heat transfer coefficient
distribution along the melt/concrete interface in contrast to available mechanistic models.
4.4.1.7. SOCRAT/HEFEST
The SOCRAT/HEFEST verification programme has started recently, so only the rational for the
selection of experiments and some particular features of the code related to the selected experiments
are discussed here. Since the scope of applicability of the code should include concretes used in
Russian NPPs, the list of experiments used for verification differs from Table 4.3-2, but a number of
experiments from Table 4.3-2 are included.
4.4.1.8. TOLBIAC-ICB
The TOLBIAC-ICB validation matrix contains experiments from the SURC-, ACE-, MACE-, NEA-
MCCI, BETA- and COMET-test series as well as on VULCANO tests. Despite the difficulties of
defining appropriate initial and boundary conditions (e. g. initial melt composition, which is especially
important for the phase segregation model) for MCCI tests, and despite the low accuracy of some
measurements, the results of the simulations of most experiments with TOLBIAC-ICB are
satisfactory, even if the melt temperature is in some cases apparently overestimated. Nevertheless the
melt temperature measured at the end of the experiments corresponds to the liquidus temperature that
is calculated on the basis of the measured melt composition. The major problem seems the calculations
of the correct melt composition rather than the model assumption according to which the melt
temperature follows the liquidus temperature. Post-test analyses show evidence that part of the initial
melt is splattered on the walls or concentrates in initial crusts and that in ANL tests magnesia could be
ablated or dissolved and mixed with the corium.
NEA/CSNI/R(2016)15
4.4.2. Code benchmarks on MCCI experiments

In the frame of the European SARNET network and the NEA-MCCI Project code benchmark actions
have been performed to evaluate the uncertainties of the codes for the prediction of MCCI experiments
(Cranga, et al., 2010), (Journeau C. , et al., 2012), (Spindler, et al., 2007). A synthesis of these actions
is given here. It concerns first the 2D NEA-CCI-2 test (performed at ANL (Farmer, Lomperski,
Kilsdonk, & Aeschlimann, OECD MCCI Project: 2-D Core Concrete Interaction (CCI) Tests, Final
Report, 2006) with a homogeneous pool and a limestone/common sand concrete, which was used for a
blind benchmark. Secondly, the COMET-L2 and COMET-L3 2D experiments performed at KIT in a
stratified configuration were used as a post-test (L2 (Sdouz, et al., 2006) and a blind-test
(L3 (Alsmeyer, et al., 2007) benchmark. A third MCCI benchmark was performed for the VULCANO
experiments VB-U5 and -U6 (Journeau C. , et al., 2012).
The major uncertainties that are pointed out in these analyses concern mainly the corium/concrete
interface model and the heat flux distribution along lateral and bottom interfaces in case of a fixed
homogeneous pool configuration. In case of stratification, major uncertainties are related with the
initial pool configuration assumptions and subsequent configuration evolution models as well as with
the interlayer heat transfer.
4.4.2.1. The CCI-2 benchmark
The benchmarking work concerning the CCI-2 test (Farmer, Lomperski, Kilsdonk, & Aeschlimann,
OECD MCCI Project: 2-D Core Concrete Interaction (CCI) Tests, Final Report, 2006) was performed
within the NEA-MCCI Project. The results were presented at the MCCI Project seminar (Spindler, et
al., 2007). The most interesting results concerning code comparisons are the blind calculations
performed before the experiment. They were performed with the same data for all participants. The
input data for post-test calculations of the experiment were significantly modified after the experiment
(in particular concerning the corium mass involved in the interaction and the concrete composition). It
was shown, that models taking a strong coupling of heat transfer models with thermodynamical
equilibrium conditions into account (i.e. coupling of interface temperature between corium and
concrete with the liquidus of the residual melt) depend strongly on the melt composition. However, a
comparison of the post-test calculation results was not organised.
temperature of melt
ANL CORCON
2800
CEA TOLBIAC
EDF TOLBIAC
temperature K
2400 FRA COSACO

GRS MEDICIS
GRS WEX
2000 IRSN MEDICIS
JNES COCO
1600 KAERI CORCON
0 60 120 180 240 300 360 Measurements
time mn
Figure 4.4-1: CCI-2 blind benchmark: Melt temperature versus time
Very large temperature differences (Figure 4.4-1) are observed for the different codes, with three
kinds of initial behaviour: sharp increase, gradual decrease and sharp decrease. This behaviour is
related to the condition that is used for the heat transfer between melt and concrete. An interfacial
temperature equal to the liquidus temperature gives an increase of the melt temperature because the
NEA/CSNI/R(2016)15
initial temperature is lower than the calculated liquidus temperature. On the opposite, an interfacial
temperature equal to the solidus temperature gives an initial decrease of the melt temperature. After
about one hour, a quasi-steady state evolution of the melt temperature is reached. At the time water is
poured on the melt, a rapid decrease of melt temperature is observed in some calculations
ablation depth ANL CORCON-ax

ANL CORCON-rad
CEA TOLBIAC
0.4 EDF TOLBIAC-ax
EDF TOLBIAC-rad
FRA COSACO
0.3 GRS MEDICIS
ablation m
GRS WEX-rad
GRS WEX-ax
0.2
IRSN MEDICIS
JNES COCO-ax
0.1 JNES COCO-rad
KAERI CORCON-ax
KAERI CORCON-rad
0.0 Measurements ax
0 60 120 180 240 300 360
Measurements rad South
time mn Measurements rad North
Figure 4.4-2: CCI-2 blind benchmark: Ablation depth versus time
final shape
ANL CORCON
1.0
CEA TOLBIAC
EDF TOLBIAC
0.8 FRA COSACO
GRS MEDICIS
height m
0.6 GRS WEX

IRSN MEDICIS
0.4 JNES COCO
KAERI CORCON
0.2 Experiment S
Experiment N
0.0
0.0 0.2 0.4 0.6 0.8 1.0
width m
Figure 4.4-3: CCI-2 blind benchmark: Final cavity contours

The initial ablation rate differs depending on the code, and this discrepancy is connected to the
melt temperature behaviour. After about one hour, the ablation rates are less dispersed (Figure 4.4-2).
The final shape of the cavity (Figure 4.4-3) mainly depends on the choice made by the code user:
either isotropic heat transfer (which corresponds to what was observed in the experiment) or radial
heat transfer higher than axial heat transfer.
NEA/CSNI/R(2016)15
4.4.2.2. The COMET-L2/L3 benchmark

The COMET-L2 test (Sdouz, et al., 2006) was performed at Karlsrühe Institute of Technology (KIT)
in the frame of the LACOMERA project of the 5th European Framework Programme. The melt is
composed of oxide (alumina and calcia) and metal (iron and nickel). A sustained heating power was
released in the bottom metal layer for about 17 min. The COMET-L3 test (Alsmeyer, et al., 2007)
differs from COMET-L2 mainly by a top flooding that was triggered at 13 min. After an initial
transient period of about 100 s, characterised by an isotropic and fast ablation, the overheat of the
metal is lost and a quasi-steady state regime is reached, with a faster axial ablation compared to the
lateral ablation (factor 2 to 3), which is in agreement with the results of the BETA experiments at a
low power density.
Metal temperature Oxide temperature
2050 2050
2000 2000
AREVA AREVA
1950 1950
CEA CEA
1900 1900
temperature, K
temperature, K
CEA mod CEA mod
1850 1850
EDF EDF
1800 1800
FZK FZK
1750 1750
GRS-MEDICIS GRS-MEDICIS
1700 1700
1650 GRS-WEX GRS-WEX
1650
1600 IRSN 1600 IRSN
1550 IRSN mod 1550 IRSN mod
0 300 600 900 1200 1500 1800 VTT 0 300 600 900 1200 1500 1800 VTT
time, s time, s
a) Calculated metal temperature versus time b) Calculated oxide temperature versus time
Axial ablation Radial ablation
0.18 COMET-L2 SW 0.10 COMET-L2 t6 SW

COMET-L2 CL COMET-L2 t5 SW
0.15 0.08
COMET-L2 NE COMET-L2 t6 SE
ablated depth, m
ablated depth, m
0.12 AREVA AREVA

CEA 0.06 CEA
0.09 CEA mod CEA mod
EDF 0.04 EDF
0.06 FZK
FZK
GRS-MEDICIS 0.02 GRS-MEDICIS
0.03
GRS-WEX GRS-WEX
0.00 IRSN 0.00 IRSN
0 300 600 900 1200 1500 1800 IRSNmod 0 300 600 900 1200 1500 1800 IRSN mod
time, s VTT time, s VTT
c) Calculated and measured axial ablation versus d) Calculated and measured lateral ablation
time versus time
Figure 4.4-4: COMET-L2 benchmark: CEA and IRSN presented a base calculation and calculations
with a modified model for a better agreement with the experimental results
The COMET-L2 test was used for a post-test benchmark. The participants were Areva with
COSACO, CEA with TOLBIAC-ICB, EDF with TOLBIAC-ICB, FZK with WECHSL, GRS with
ASTEC/MEDICIS and with WEX (which had been developed by GRS based on WECHSL), IRSN
with ASTEC/MEDICIS and VTT with MELCOR. The same input data were used by all the
participants. CEA and IRSN presented a base calculation and additional calculations with some
modifications of the models in order to get a better agreement with the experimental results.
The scatter between the calculated metal temperatures in Figure 4.4-4 a) is about 150 K, but six
results are between 1 750 and 1 780 K. The scatter between the oxide temperatures in Figure 4.4-4 b)
is larger: about 350 K at 1000 s. There are no bulk temperature measurements for comparison. There
NEA/CSNI/R(2016)15
is also a large scatter concerning the ablation depth (Figure 4.4-4 c) and Figure 4.4-4 d), but it can be
noticed that, after the first phase corresponding to the initial overheat, the ablation rate is similar for all
the codes. Finally, when compared to the experimental results, it is found that the maximum axial
ablation is underestimated.
Metal temperature Oxide temperature
1950 1950
AREVA 1900 AREVA
1900
CEA 1850 CEA
1850
EDF 1800 EDF
temperature, K
temperature, K
1800 FZK 1750 FZK

1750 GRS-MEDICIS 1700 GRS-MEDICIS
1700 GRS-WEX 1650 GRS-WEX
IRSN 1600 IRSN
1650
UPM 1550 UPM
1600 1500 VTT
VTT
1550 1450
0 300 600 900 1200 1500 1800 2100 0 300 600 900 1200 1500 1800 2100
time, s time, s
a) Calculated metal temperature versus time b) Calculated oxide temperature versus time
Axial ablation Radial ablation
0.25 0.14
AREVA
0.12 AREVA
CEA
0.20 CEA
EDF 0.10
ablated depth, m
ablated depth, m
EDF
FZK
0.15 0.08 FZK
GRS-MEDICIS
GRS-MEDICIS
GRS-WEX 0.06
0.10 GRS-WEX
IRSN
0.04 IRSN
UPM
0.05 UPM
VTT 0.02
VTT
COMET-L3 R2 NE
0.00 0.00 COMET-L3 Z5 SW
COMET-L3 R3 SE
0 300 600 900 1200 1500 1800 2100 0 300 600 900 1200 1500 1800 2100 COMET-L3 Z5 NE
COMET-L3 R1 CL
time, s time, s
c) Calculated and measured axial ablation versus d) Calculated and measured lateral ablation
time versus time
Figure 4.4-5: COMET-L3 benchmark
The COMET-L3 test was used for a blind test benchmark. The participants were Areva with
COSACO, CEA with TOLBIAC-ICB, EDF with TOLBIAC-ICB, FZK with WECHSL, GRS with
ASTEC/MEDICIS and with WEX (which is derived from WECHSL), IRSN with ASTEC/MEDICIS,
UPM with MELCOR and VTT with MELCOR. The experimental results were not known when the
calculations were performed, but for some code, the model modifications tested in order to get a better
agreement with COMET-L2 were used for the simulation of COMET-L3. The results are presented
Figure 4.4-5(a)-(d).
The top flooding at 800 s is not really sensitive to the calculated results. It can be observed that
the scatter of the calculated oxide and metal temperatures (Figure 4.4-5(a) and Figure 4.4-5(b)) is
reduced compared to COMET-L2, because of fitting of some parameters to the results of COMET-L2.
In the initial phase, some codes give a heat transfer from the oxide layer to the metal layer and the
others from the metal to the oxide. In the second phase, before flooding, all codes predict heat transfer
from the metal to the oxide. After flooding, some codes give back a heat transfer from the oxide to the
metal.
For production of gas through oxidation of the metal layer (H2 and CO) the scatter is large, with
about a factor 5 between the larger and the lower values. There is also a large scatter concerning the
ablation depth (Figure 4.4-5(c) and Figure 4.4-5(d)). Some codes give results that are similar to the
NEA/CSNI/R(2016)15
experimental results. Some others overestimate the axial ablation or the lateral ablation, and the
ablated volume.
4.4.2.3. The VULCANO-VBU5/VBU6 benchmark
The results of two oxidic tests VB-U5 (silica-rich concrete) and VB-U6 (limestone-rich concrete) were
used for a benchmark on MCCI codes as a part of the European Severe Accident Research Network of
Excellence (SARNET). Ten participants from different European countries (CEA Cadarache France,
CEA Grenoble France, IRSN France, GRS Germany, KIT Germany, VTT Finland, Areva Germany,
EI Bulgaria, INRNE Bulgaria, RSE Italy) took part in this work. Six computer codes (TOLBIAC-ICB,
ASTECv2/MEDICIS, COSACO, CORQUENCH, WECHSL and CORIUM-2D) were used to perform
two independent calculations, each one representing the main phenomena arising during the
interaction between prototypic oxidic corium and siliceous or limestone concretes.
An example of the obtained final cavity shapes in case of the VB-U5 test is displayed in Figure
4.4-6. Clearly, codes imposing the anisotropy (by fitting a multiplicative factor applied to the
convective heat transfer coefficients or imposing values of thermal resistances at pool/concrete
interfaces) or using heat transfer models leading intrinsically to anisotropy permit to get a better
agreement for the final cavity shape.
Crucible shapes at the end of calculations of VB-U5 test
0
VTT
0 5 10 15 20 25
GRS
-5
AREVA
IRSN
-10
EI
INRNE
-15
H (cm)
CEA_gre_base
-20 CEA_gre_case2
CEA_gre_case3
-25 KIT
CEA_cad_2336K
-30 CEA_cad_2625K
RSE
-35 EXPERIMENT
R (cm) INITIAL
Figure 4.4-6: Crucible shapes at the end of calculation for VB-U5
Pool temperatures calculated for VB-U5 are shown in Figure 4.4-7. Most curves indicate a trend
of monotonous decrease. In most of the calculations the pool temperature decreases faster during the
first transient phase and this decrease slows down at later times; the only exceptions are TOLBIAC-
ICB (CEA) and CORIUM-2D (RSE) calculations with the highest pool/crust interface temperature. It
must be noted that the pool temperature measurement (2 403 K), which has been taken only during a
small time period around 1200 s, is somehow overestimated as it has arisen from recent analysis of
subsequent VULCANO experiments (the real temperature can be 30 to 200K lower). Only the
TOLBIAC-ICB calculations which follow the liquidus temperature (Figure 4.4-7) give fairly good
estimates of the high corium temperature measured.
NEA/CSNI/R(2016)15
Pool temperatures versus time for VB-U5 test

3000 GRS
VTT
2500 AREVA
IRSN
2000 EI
Temperature (K)
INRNE
1500 CEA_gre_base
CEA_gre_case2
1000
CEA_gre_case3
KIT
500
CEA_cad_2625K
0 CEA_cad_2336K
0 1000 2000 3000 4000 5000 6000 RSE
Time (s) EXPERIMENT
Figure 4.4-7: Pool temperatures versus time for VB-U5
Concerning the VB-U6 experiment with limestone rich concrete, similar fits have been observed.
For instance, Figure 4.4-8 presents the experimental and computed cavity shapes.
Crucible shapes at the end of calculations of VB-U6 test

0
VTT
0 5 10 15 20 25
GRS
-5 AREVA
CEA_gre_base
-10 CEA_gre_case2
CEA_gre_case3
-15 IRSN
H (cm)
INRNE
-20 CEA_cad_9kW
CEA_cad_9kW-Pcond
KIT
-25
EI
RSE
-30
EXPERIMENT
INITIAL
-35
R (cm)
Figure 4.4-8: Final cavity shape for VB-U6: Experimental and calculated profiles
• Many similarities have been identified in the predicted trends. Nevertheless some major
differences between modelling approaches were observed.
• The ablated volume is controlled by the ablative energy, thus it is impacted by the amount of
energy radiated through the upper surface which depends on the code heat transfer models
(heat convection distribution and interface structure) and also on the interface temperature to
the upper crust. Most codes overestimate the ablated concrete volume (based on a
comparison of final melt composition data) if significant conduction heat losses through the
concrete are not taken into account especially in case of VB-U6. Impact of the ablation
occurring during the initial transient phase can also be important, in particular in
experiments.
• Cavity shapes are rather well predicted: the VB-U5 with siliceous concrete required taking
into account anisotropy, either explicitly (with ASTECv2/MEDICIS, TOLBIAC-ICB) or
implicitly (with CORQUENCH); axial ablation is generally overestimated.
NEA/CSNI/R(2016)15
• TOLBIAC-ICB calculations provide good estimates of the high pool temperature measured,
whereas the other models give some discrepancies of several hundred kelvins at least in the
initial MCCI phase of the VBU5 experiment. However, the calculated temperatures cannot
be compared with the experimental one in the longer term since the overall pool temperature
evolution was only measured at one unique time and the measured temperature may be
significant only for the transient period of the experiment, during which validation of codes
is a very ambitious task.
• All codes predict a small final void fraction (even with limestone test VBU6). The reason is
probably that the drift flux model used to calculate the void fraction is not well applicable to
2D configurations, in particular in small scale experiments in which side area is greater than
bottom area, contrary to the reactor case.
• Experimental results indicate that crusts, if they exist, are likely to have a composition close
to the current pool composition. Therefore it unfirms the assumptions that crust could keep
the same composition from the beginning (i.e. a time where there was less concrete in the
pool) and that crusts contain a higher fraction of most refractory species. Such assumptions
are in line with the TOLBIAC-ICB code hypotheses. Experimental observations show that
crust formation, if any, is renewed all along the experiment and without any significant
segregation.
4.5. Material properties validation status
It must be reminded that in this high temperature range material properties are known with significant
uncertainties (Journeau, Piluso, & Frolov, 2004). Moreover, the codes are assuming that the melt is
homogeneous or that there are two homogeneous melt phases (one oxide and one metal) whereas the
VULCANO post-test analyses revealed the presence of concrete-rich and corium-rich plumes having
different compositions and thus different properties, which cannot be taken into account with the
current level of MCCI modelling.
Among the thermophysical properties, one of the uncertainties that may have the most
consequences will be the density difference between oxidic and metallic phases. From the BALISE
correlation, a 10% difference in the densities of phases can be mixed by a 0.5 cm/s sparging gas flow.
The uncertainty on oxidic liquid density is clearly above 10% for complex melts: it depends on
• The knowledge of the phase composition.
• The knowledge of the partial molar volume of each constituents.
• The validity of the assumption of ideal mixing (no excess volume is considered).
Unfortunately, there are very few data available to validate these calculations.
Concerning the thermo-chemical properties (phase composition, liquidus temperature,…) ex-
vessel corium experiments as VULCANO, EVAN, COMETA have been used to validate and improve
the NUCLEA database (Bakardjieva, et al., 2010). Nevertheless, uncertainties of several wt% in
composition (1/10th of the value for heavy elements, 1/4th for oxygen) and of more than 50 K in
temperature are to be expected. A thorough analysis of the liquidus of EPR concretes showed that, in
the vicinity of steep variations of liquidus temperature with respect to composition (300 K in a few
wt%), the uncertainty on the liquidus temperature can be of the order of 300 K in this extreme case.
Outside from these peculiar compositions, much lower uncertainties are expected which are consistent
with the overall uncertainties of the codes.
For oxide-metal melts, uncertainties on the composition of phases at equilibrium may be
significant, as the validation matrix is relatively scarce in MCCI compositions.
NEA/CSNI/R(2016)15
Finally, it must be reminded that thermodynamic equilibrium calculations assume that all phases
are in equilibrium and do not account for any kinetic effects, which may not be totally the case. For
instance, Scheil-Gulliver solidification has been found to explain better some of the compositions
found in VULCANO samples better than equilibrium calculations (Bakardjieva, et al., 2010).
4.6. Scaling to reactor case
A discussion of remaining uncertainties resulting from code validation must be combined with
uncertainties coming from the scaling to reactor case.
MCCI codes are validated on the basis of a range of experiments which were performed in test
sections with a length scale < 1 compared to reactor scale. A very important question is in how far the
codes’ models are also valid for predicting a large scale reactor scenario.
In this chapter higher-level code capabilities which can be tested in calculations for available
integral experiments are identified. In the following it will be shown how far the available data of
experiments may be evaluated for the long-term and large-scale reactor case.
4.6.1. On the mapping of experiments to reactor scale
With regard to the two basic features concrete ablation (abl) and melt temperature (mt) prediction
during quasi-steady state, two parameters that may govern the physics for the MCCI are usually
considered: i) the ratio of injected power divided by available surface area for cooling (this is the
approximate heat flux density in quasi-steady-state) and ii) the ratio of injected power divided by the
melt volume (this is e. g. a governing parameter for natural convection heat transfer). During the
MCCI the variation of oxide melt properties is large. This is captured by tracking both the above
mentioned parameters along the concrete weight fraction in the oxide melt, whereas for the steel melt
the material properties will stay more or less the same.
Using a simplified procedure, which is described below, it is checked, how the design basis of
some representative experiments would map to a typical MCCI reactor scale (using the volumetric
internal power and/or the heat flux densities as scaling parameter) as function of concrete weight
fraction.
Reference case is a core melt scenario in a German BWR with 156 t of UO2 and 30 t of ZrO2 in
the oxide phase and 73 t of Fe, 50 t of Zr, 11 t of Cr and 6.4 t of Ni in the metal phase. This scenario
was basis for a MCCI code benchmark in the SARNET project (Cranga, et al., 2010). After start of the
MCCI (defined by t = 0 s) the decay power in the oxide and metal phases as function of time is shown
in Figure 4.6-1.
Figure 4.6-1: Decay power in the oxide and metal phases after start of the MCCI
in the reference reactor case (Cranga, et al., 2010)
NEA/CSNI/R(2016)15
The MCCI in a reactor may involve a sequence of different pool configurations due to
segregation phenomena. Due to the non-miscibility of the oxidic and metallic liquids of which the
corium is composed an establishment of different configurations can be expected depending on the
strength of agitating effects like gas release from the decomposition of the concrete.
For simplicity the time evolution of typical heat flux densities at the surface of the corium and of
the volumetric power are evaluated here assuming a static, homogeneous corium layer (metal
dispersed in oxide melt). The required time-dependent data for this evaluation (i. e., surface area for
cooling of the corium, volume of the corium, corium and concrete masses in the pool etc.) are taken
from the ASTEC/MEDICIS BWR reactor benchmark calculation for the static homogenous pool
configuration and the siliceous concrete as published in (Cranga, et al., 2010).
In Figure 4.6-2 the power and specific heat flux density calculated with ASTEC/MEDICIS for
the reference BWR reactor scenario (Cranga, et al., 2010) are compared with estimated data of some
representative experiments. The evaluation of volume and surface area of the pool obtained from
ASTEC/MEDICIS includes the effect of voiding, whereas the experimental data are approximated for
the collapsed corium. Focusing on specific heat flux densities (blue lines and blue symbols in Figure
4.6-2) the SURC experiments (e.g. SURC-1) represent a good approximation of the selected reactor
case for the very early phase at low concrete fraction while the data of the CCI experiments represent a
good approximation of the reactor case for the period up to 60 wt.-% concrete. Regarding volumetric
power (red lines and symbols in Figure 4.6-2) the MACE-M3b experiment represent a good
approximation of the period between 20 and 40 wt.-% concrete in the melt for this reactor scenario.
Figure 4.6-2: Mapping of experimental data of oxide experiments to a selected BWR reactor
scale on the basis of an ASTEC/MEDICIS calculation (homogeneous case)
For a PWR scenario the overall melt mass will be smaller and the power density will be larger,
thus a better agreement with experimental data for volumetric power is expected for a PWR.
Considering the processes within a stratified MCCI pool configuration including a segregated
metal layer the situation is different: The composition of the steel melt will not vary strongly during
the MCCI like in the case of the oxide melt. The volume will also not vary so much. Some amount of
metal is leaving the layer due to oxidation, whereas some amount of iron is introduced into the layer
by ablation of the reinforcement bars. Thus the volumetric power released in the metal and the specific
heat flux density across the surface of the steel melt will not decrease as much as in the case of the
oxide layer. Since there is no strong variation of material properties with concrete erosion which could
NEA/CSNI/R(2016)15
be applied to the case of the metal layer, the question remains of how to map experiments with steel
melts to the reactor scale. There is no clear way to evaluate a scale for the segregated case.
Here, the following method was used to evaluate such a mapping: The transient evolution of the
layer contour (surface area and volume of the metal layer) in the BWR reactor scenario is taken from
the ASTEC/MEDICIS calculation for the stratified calculation case (Cranga, et al., 2010). For the
estimation of power per volume and heat flux per area, however, the total decay power in oxide and
metal layer is considered. This assumption is plausible because a focusing effect may redistribute a
large part of decay heat of the oxide via the metal layer to the concrete.
Using this methodology it is found that, when regarding heat flux density at the corium/concrete
interface, the BETAV5.2 experiment maps to the reactor calculation at the beginning of the interaction
and the COMET-L3 experiment maps to the reactor calculation after some time of interaction (~ 0.5
days). However, the evaluation on the basis of the calculation with ASTEC/MEDICIS indicates
smaller volumetric power levels compared to the design of available experiments. COMET-L3 may
approximately represent an upper boundary of the volumetric power level in the metal layer at later
stages of the selected BWR scenario (t ~ 2 days).
Although the agreement between calculated data and experimental data depends strongly on the
selected reactor scenario and remaining uncertainties of code results (obtained here with
ASTEC/MEDICIS) may be quite significant, the findings suggest that for the BWR scenario selected
experimental data are principally available for the early phase of interaction, but data are missing for
the very late phase (i. e., at large concrete fraction > 60 wt.-% in the oxide layer). This is in principle
valid also for a PWR scenario. A further uncertainty for extrapolation of experiments to reactor scale
is that in experiments with a segregated metal layer, the power was injected predominantly into the
metal phase, which is not in agreement to the reactor situation (large fraction of power release in the
oxide phase), except for the MOCKA experiments of KIT Karlsrühe, which are not fully examined yet
and for the VULCANO experiments of CEA in which unexpected shape of the metallic pool is not yet
explained.
4.6.2. Simplified transposition of important parameters from small to large scale
The exemplary study in the preceding section tries to map representative experiments to a reactor
scenario considering the heat flux density (W/m2) to the concrete or the volumetric power (W/m3) in
the melt, respectively, and the concrete admixture to the corium as key parameters for scaling, which
showed that experimental data cover the reactor scale up to approx. 60 wt.-% concrete (~ order of 1
day of interaction). This procedure was proposed on the basis of MCCI calculations with the MCCI
code MEDICIS. Such MCCI codes tend to evolve into a sequence of quasi-steady states, in which the
decay heat released in the melt is roughly redistributed into concrete melting and heat losses from the
free surface. According to these quasi-steady states the level of the pool temperature is governed then
by the efficiencies of heat transfer mechanisms and boundary conditions at the pool’s interfaces
(expressed in terms of effective heat transfer coefficients and interface temperatures).
Because of that any scaling of the temperature history recorded during a small-scale MCCI to the
reactor must take special care of the heat transfer between the melt and its interfaces. If this heat
transfer explicitly depends on a geometrical dimension of the pool (e. g. on the characteristic length of
the pool) then the temperature level of the MCCI will be scale-dependent. This may be the case if
mechanisms such as natural or solutal convection govern heat transport.
The heat transfer is not affected by the geometrical dimension of the pool if bubble-induced
forced convection is the driving mechanism for heat transfer in the pool: here the characteristic pool
length itself does not appear in the heat transfer law, it is replaced by a characteristic bubble length
(for example in the BALI heat transfer correlation which uses the Laplace constant in the definition of
its Nusselt number).
NEA/CSNI/R(2016)15
Currently the mechanisms of heat transfer between melt and interfaces are not well understood
and identified. Corium viscosity (at least in the boundary layer) and the gas release at the interface are
suspected as parameters affecting the efficiency of heat transfer and/or the 2D heat flux distribution.
Rough estimates for heat transfer coefficients and 2D heat transfer coefficient distributions for MCCI
were obtained by empirical evaluations of MCCI experiments. A range of experiments can be
recalculated with good agreement to experimental results applying these heat transfer coefficient
estimates at the interfaces. It is presently not clear, whether these empirical data are invariant to
scaling or not.
By properly selecting the geometry, the initial conditions and the heating power of a small-scale
MCCI and under the assumption that heat transfer is invariant to scaling, the relevant variables as
temperature, concrete erosion, melt composition of a large-scale MCCI system can simply be obtained
by appropriate transformation from the small scale MCCI ( (Fargette, 2013), (Spengler, Fargette, Foit,
Agethen, & Cranga, 2013). However, some strong conditions for the corium pool geometry must be
fulfilled.
For example, considering that one is interested in the temperature-history of the following
(initially cylindrical) reactor-scale MCCI (basing on the BWR scenario considered in section 4.6.1): R
= 3 m, H = 1.8 m, P = 20 MW. A small-scale MCCI experiment with the same initial melt
temperature, composition and the following characteristics could be performed: r = 3 / 10 = 0.30 m, h
= 1.80 / 10 m = 0.18 m, p= 20×106 / 102= 200 kW (assuming the heat flux density as one parameter
which should be invariant to scaling). By stretching the initial temperature profile by the same factor
10, we then obtain the reactor-scale temperature profile, presumed that both systems evolve according
to the constraint of ideal cylindrical geometry. In the example given the temperature measured at the
test scale at t = 0.5 hr corresponds to the reactor-scale temperature at t = 0.5*10 = 5 hrs.
However, the condition of identical initial specific power density (W/m2) in both systems is not
the only possible constraint for scaling. Under different scaling conditions, e. g. if the relevant
variables are related by the ratio of the melt heat capacities, the time evolution of the melt temperature
will be identical in both systems (Foit J. J., 2012 ).
A potential reason for the applicability of constant effective heat transfer coefficients at the
interfaces to simulate MCCI experiments may be referred to the combined effect of cooling of the melt
(→ increase of viscosity via increase of solid fraction with decreasing temperature) and dilution with
concrete (→ decrease of viscosity via decrease of solid fraction with enrichment of concrete) which
could result in a rather constant plateau of effective pool viscosity, dependent of thermo-dynamical
data supplied (Spengler, 2013). More research efforts are required on this item: An improved
knowledge of the heat transfer mechanisms at the pool interfaces is urgently required (but hard to
gain) in order to confirm and/or replace the empirical effective heat transfer coefficients and cover
potential effects of evolving MCCI conditions in the long term.
For a further confirmation of these findings additional theoretical studies are proposed: Reactor
scale applications should be investigated thoroughly with reference to the time behaviour of the
different terms in the energy equation, especially with regard to the validity of quasi-steady state
assumptions (→ power redistribution to the interfaces, pool temperatures governed by effective heat
transfer coefficients), and to the evolution and effect of more realistic cavity contours in contrast to
simplified geometrical assumptions (cylinder, hemispherical section, etc.). This task could be fulfilled
by a reactor benchmark with the contribution of several different MCCI codes.
4.7. Summary of the codes’ validation status
From a general point of view a fully satisfying validation of the different terms in the complete energy
balance in MCCI codes is not possible (see Section 4.1), since the quantities required are mostly not
NEA/CSNI/R(2016)15
available or inaccurate based on the documentation of available experiments, e. g. total energy used
for ablation, total energy released from the free surface etc.
Because of that most codes have been validated against various MCCI tests primarily on the basis
of maximum or average erosion depths and temperature history. Models show, that when the
experiments temporarily run in a kind of quasi-steady state, i. e., at a smooth temperature evolution
when the injected energy is substantially redistributed to the interfaces, a differentiated information
could be obtained from that: The 2D heat flux distribution calculated in the codes is validated, if the
propagation of erosion depths (or more precisely speaking: average erosion velocities) in 2D are well
predicted for those quasi-steady time periods. If in addition the temperature of the melt is well
predicted in the codes, this serves as a validation for the heat flux modelling q = h Δ T in 2D for the
quasi-steady time periods. The relation between the heat transfer coefficients htop, hside, hbottom in
combination with the imposed interface temperatures (concrete decomposition temperature, crust
formation temperature for the top crust) governs the 2D heat flux distribution in the codes, and the
magnitude of these heat transfer coefficients is a key parameter determining the long-term pool
temperature.
Important progress in the understanding of 2D MCCI has been recently achieved in the frame of
the SARNET research on MCCI (Cranga, et al., Towards an European consensus on possible causes of
MCCI ablation anisotropy in oxidic pool, 2014): Good agreement between code calculations and
experiments was obtained using effective heat transfer coefficients between the bulk of the corium and
its interface in combination with the approved “melting model” for concrete erosion. Such effective
heat transfer coefficients seem to be fairly constant throughout the experiments and are representative
for the overall heat transfer from the bulk to the bottom and to the lateral pool interfaces, respectively.
The actual progress in the codes as outcome of the recent MCCI research may be summarised as
follows: Experiments with oxidic melts confirm that total heat transfer coefficients (htc) between the
liquid melt and the concrete ablation interface range from several 10 (in case of initial crusts) up to of
a few 100 W/(m2 K). Comparing this to empirical correlations for known convection mechanisms
(natural, solutal or bubbly convection) these htc are rather small. The reason for such reduced htc is
referred to thermal resistances developing at the interface between the melt and the concrete. Such
thermal resistances may be interpreted as temporary crusts or highly viscous boundary layers. The
detailed convection mechanism in the liquid bulk does not have large effect on the overall htc. The
conditions controlling the stability of such interface structures in the long term seem to depend on the
concrete type. In several experiments with siliceous concretes initial crusts at the bottom were of
higher stability than at the sidewall. Since the presented MCCI codes do not yet feature any model for
the simulation of crust stability under typical conditions, empirical htc have to be imposed by the code
user. The research has shown that a local crust may or may not be present at the lateral or the axial
interface. Consequently, the distribution of htc according to the interface angle and concrete type must
be selected under the specific assumption that highly viscous thermal interface structures are present
or not in the long term. The code user should be aware of this issue and should select a conservative
parameter setting with view to the most crucial consequence for the specific accident sequence.
Taking benefit of this rather empirical methodology the code results show for many experiments
an acceptable level of predictive capabilities for the estimation of concrete ablation (1D and 2D),
particularly for homogenous pools with oxidic character and a limestone rich concrete. For the case of
a homogeneous oxide melt and a siliceous concrete the anisotropic ablation observed in experiments is
still related with a significant uncertainty with regard to quantification of the effect and to the
mechanistic interpretation. Up to now the codes require empirical input parameters governing a non-
isotropic 2D distribution of heat transfer for siliceous concretes, which may also be dependent on the
local crust failure time.
It should however be stressed that there is still an uncertainty in the prediction of MCCI pool
temperature levels. In the short term this is caused by insufficient knowledge of the material properties
NEA/CSNI/R(2016)15
during the initial temperature transient (i.e. evolution of latent heat etc.). In the long term this is most
probably related to uncertainties in mechanistic modelling of the effective heat transfer coefficient
distribution (i. e. the relation between htop, hside, hbottom) and in the related interface temperatures,
since the complex mechanistic and probably thermo-chemical processes contributing to the heat
transfer are not known sufficiently. It is assumed that the corium viscosity in the boundary layer is one
of the dominating properties governing the efficiency of heat transfer and thus controlling the
temperature drop throughout the MCCI (see (Fargette, 2012), (Foit J. J., 1997), (Hackel & Gröll,
1969). A direct extrapolation of 2D heat flux distributions observed at a small scale to a larger scale
(compare section 4.6.2) is still questionable as long as the mechanisms for the driving heat transfer
through the interface structure are not well understood.
The transient processes during the incubation period in MCCI experiments are not sufficiently
modelled in the codes. Several codes assume a transient crust at the corium/concrete interface which is
a necessary feature to predict the behaviour of initial crusts. However, MCCI codes assume generally
a conversion of heat flux into progression of ablation contour. To predict the phenomena during the
incubation period correctly heat fluxes at the corium/concrete boundary should lead to heating of the
concrete but – depending on the interface conditions – not necessarily to concrete erosion. Deepened
theoretical efforts on this item would be of benefit for answering the question on the scaling issue with
the incubation period.
With view to the phenomenon of gas release from the MCCI to the containment and oxidation
reactions in the melt the validation status is lagging behind. This is one consequence of the intensified
discussion on 2D ablation behaviour, since the gas release coming from the decomposition of the
concrete is regarded only as subsequent effect of 2D concrete ablation and impacts directly the
oxidation of metals. As soon as the discussion of 2D MCCI converges on a common interpretation
(and modelling) of the 2D ablation, the validation of models should be continued with regard to the
oxidation behaviour which will influence the composition of the gases released into the containment.
The validation of top flooding models and top crust formation in integral MCCI codes is not
totally satisfactory in the different integral codes. The most comprehensive validation matrix for the
phenomenon of top flooding was obtained with the code CORQUENCH (Farmer, 2010), including
separate effect and integral tests. Some integral codes integrate a modelling similar to the
CORQUENCH original models. Satisfactory validation on the available experimental database has
been obtained with these models, see a.o. (Tourniaire, Boulin, & Haquet, 2015). In other codes than
CORQUENCH the focus was not put on the top flooding/top crust formation issue until now, due to
the priority of investigating the 2D dry MCCI issues first. Although these top flooding models are
available, their application to large-scale MCCIs is questionable. These models were indeed derived
from tests in which crust anchoring to the sides of the crucible affected water ingression and
quenching. At a large scale, this phenomenon would not be present and a “floating crust” is expected.
For the assessment of the top flooding/top crust effect on the course of the MCCI in experiments the
crust anchoring effect has to be taken into account.
For the phenomenon of FP and aerosol release the richest data base is still represented by the
large scale experiments of the international Advanced Containment Experiments (ACE) Program on
melt behaviour and aerosol release during MCCI (Fink, Thompson, Armstrong, Spencer, & Sehgal,
1995). Model approaches for the fission product release by vapourisation are typically based on the
application of thermodynamic equilibrium models of an ideal or non-ideal melt in contact with a
gaseous phase. To calculate the thermodynamic equilibrium the Gibbs energy of the total system as a
function of temperature and composition has to be evaluated. The vapourisation of fission products in
the ex-vessel situation depends much on the temperature of the core melt during the interaction and on
the gas flow rate coming from the concrete decomposition and since the uncertainties for the
prediction of these quantities are still large the precision of calculated releases is – taking into account
the inherent additional uncertainties for thermodynamical modelling – currently estimated to be at
NEA/CSNI/R(2016)15
least one or two orders of magnitude (Sehgal, Nuclear Safety in Light Water Reactors: Severe
Accident Phenomenology, 2011). Mechanical aerosol generation by bursting of bubbles is not always
considered in the codes. In that case an underestimation of aerosol release should be expected. A
simple model for aerosol generation by bursting of bubbles is available in the VANESA code (Powers,
Brockmann, & Shiver, 1986), which is part of CORCON and MELCOR, respectively.
Beside the clarification of 2D ablation phenomena further large uncertainties remaining are
related to melt stratification and the impact of reinforcements on concrete ablation. Melt stratification
criteria have only been derived from experiments based on simulants and an extrapolation to real melts
remains questionable. The impact of reinforcements was not been thoroughly investigated in the past
due to the inherent heating problems. These reinforcements might significantly impact the progression
of the ablation front. Some new experimental data are available from the MOCKA test series at KIT
Karlsrühe but they were not yet investigated thoroughly.
With regard to a potential transposition of the experimental data from laboratory scale to large
scale, simple scaling laws were derived for a transposition of important parameters (e. g. corium
temperature, maximum ablation depths, etc.) from small to large scale under ideal conditions, notably:
homogeneous pool, retention of geometrical shape, restrictions on the variation of the heat transfer
coefficient (e.g. scale-independency) and a simplified treatment of the power split. Simplified
numerical approaches based on a simplified geometry to estimate important parameters are proposed
which permit to check the overall consistency of results obtained with more detailed MCCI codes and
which are useful for testing different MCCI modelling hypotheses based on bulk properties (e. g.
potential impact of viscosity on heat transfer coefficient). A closer look at available MCCI
experiments shows that experimental data are principally available for the early and mid-term phase of
interaction, but data are missing for the very late phase (i. e., at large concrete fraction > 60 wt.-% in
the oxide layer).
4.8. Conclusions
The available MCCI codes are validated on the basis of a broad range of experiments
(see Table 4.3-2), featuring simulant and prototypic materials in different geometries and with
different experimental heating devices. The focus of the validation work is commonly put on
comparisons of the calculated results for corium temperature and local or maximum ablation depths
with the experimental data.
Transient effects may have impact on the course of individual experiments – mostly at the start of
the interaction between the newly generated melt with the structures – for which the codes cannot be
assessed as “validated”. Such initial transients are related to the formation and stability of interface
structures in the boundary layer of the melt. For the longer term however, experiments enter a quasi-
stationary regime for which the code predictions on ablation progress and temperature history are well
understood: The 2D power distribution in the codes is governed by the system of effective heat
transfer coefficients in combination with imposed temperature conditions at the different interfaces.
Good agreement with oxide experiments (as regards the ablation and temperature history) are obtained
if htc in the order of 300 W/(m2K) are effectively used at the different melt/concrete interfaces in
combination with the decomposition temperature of the concrete selected close to ~1 600 K as
boundary condition to the concrete. Anisotropic ablation as observed for siliceous concretes can
currently be captured in the codes only via imposition of anisotropic htc at the 2D interfaces.
Uncertainties are identified for melts consisting of oxides and metals: in a stratified configuration
the thermal material properties of metal melt suggest elevated htc at the metal/concrete interface but
the overall transfer of decay power (which is predominantly released in the oxide melt) to the concrete
via the metal layer is finally governed by the heat transfer at the interface between the oxide and the
metal layer. A direct model validation for this interlayer heat transfer is not yet currently possible due
NEA/CSNI/R(2016)15
to lack of appropriate data from experiments under MCCI conditions. Uncertainty is in addition
introduced by insufficient knowledge of stratification/mixing processes under MCCI conditions.
The assessment of top flooding conditions on the course of the MCCI is not yet clear since the
crust anchoring effect which is observed in experiments but not expected for the reactor scale is
difficult to be taken into account.
The impact of concrete reinforcement on the MCCI is currently under investigation in the
MOCKA experiments a KIT. In current reactor calculation, the chemical impact of iron rebars is
considered (the reinforcement can be taken into account in the concrete composition), but not their
structural impact.
NEA/CSNI/R(2016)15
5. Plant applications
5.1. Introduction
This chapter deals with the applications of MCCI phenomena, models, and data to safety analysis of
nuclear power plants under severe accident conditions, particularly in the context of reactor safety
requirements, and containment designs to address such requirements. Safety requirements are
discussed first at a high level, promulgated by international bodies such as International Atomic
Energy Agency (IAEA) and Western European Nuclear Regulators’ Association (WENRA), followed
by requirements of national nuclear regulatory bodies such as US Nuclear Regulatory Commission
(NRC), Japanese Nuclear Regulation Authority (NRA), French Nuclear Safety Authority (ASN) and
German Nuclear Regulatory Authority.
Containment designs are discussed next for a number of generation II and generation III plants,
specifically those design features that address the MCCI issue. Following the discussion of safety
requirements and plant design features to address such requirements as well as the severe accident
management strategies, three idealised plant (containment) configurations, used in the preparation of
input for plant calculations, are discussed. A few example plant calculations are then presented.
This approach of plant idealisation is quite common and reasonable in the field of safety analysis,
noting the inherent uncertainties in severe accident phenomena. As with virtually all other severe
accident phenomena, MCCI has been investigated experimentally at a reduced scale using idealised
geometries and, in some cases, using simulant reactor materials. Results from these experiments and
other analytical approaches were used to develop models of core-concrete interactions, melt spreading,
and debris coolability, both with and without water present in containment as a mitigation measure.
Extrapolating the results of scaled experiments to MCCI in plant scale involves some idealisation of
plant geometry and configuration.
5.2. Plant safety requirements
5.2.1. IAEA safety requirements

The International Atomic Energy Agency (IAEA), through its publication of IAEA Safety Standards
series, promulgates general and specific safety requirements with regard to nuclear power plant
design, operation, and maintenance. The specific publication, “Safety of Nuclear Power Plants:
Design” (IAEA, 2012) establishes design requirements for safe operation of nuclear power plants,
prevention of accidents that could compromise safety, and for mitigation of the consequences of such
accidents. One of these requirements on the control of radioactive releases from the containment states
that the design of the containment shall ensure that any release of radioactive material from the nuclear
power plant to the environment is as low as reasonably achievable, and below acceptable limits under
accident conditions.
The above requirement is pertinent to the MCCI issue as it means that consideration should be
given to incorporating certain provisions into the plant design to enhance coolability of molten core
debris, and mitigate the effects of core-concrete interactions. In particular, these provisions [discussed
in (IAEA, 2004)] in more details are as follows:
• A means of flooding the reactor cavity with water to assist in the cooling process;
• Protection for the containment liner and other structural members, if necessary;
NEA/CSNI/R(2016)15
• Sufficient floor space on the basemat to spread core debris and to increase the capability of
cooling the debris;
• A reinforced sump or cavity to retain molten core material debris; and
• Use of concrete type for the containment floor that minimises adverse effects due to molten
core- concrete interactions.
5.2.2. WENRA safety requirements
The Western Europe Nuclear Regulators’ Association (WENRA) gathers representatives of the
regulators from seventeen European countries and Switzerland with the main objective of developing
a common approach for obtaining continuous improvement of safety. Due to its nature, WENRA does
not aim to propose regulatory standards; instead, it proposes Safety Reference Levels with the purpose
of harmonising safety practices that are linked to the IAEA Safety Standards. These representatives
take part in working groups including a working group on reactor harmonisation (RHWG) that
proposed statements on safety reference levels for existing reactors, including long term operation and
safety objectives for new reactors, and when needed states positions on specific safety issues or on
best practices that are part of safety evaluation processes such as PSA or PSR. Whereas a report on the
harmonisation on safety at the European level was published in January 2011, the Fukushima
situations led the WENRA to review approaches for both existing (WENRA, 2014) and new reactors
(WENRA, 2013). These Safety Reference Levels have been reviewed by the members of the RHWG
accounting for the revision of the Safety Standards made by the IAEA, for the results of the European
Stress Tests and the European Nuclear Safety Regulators Group (ENSREG) peer review reports and
for Safety Standards or Regulation reviews within national frames.
Table 5.2-1: Refined description of levels of DiD from (WENRA, 2014)
For existing plants and specifically for LTO, WENRA recommends the safety objectives of new
reactors be used as a reference for the definition through PSR of reasonable improvements that can be
implemented in existing reactors. The report on Safety Standards for New Reactors (WENRA, 2013)
reaffirms the concept of defence in depth (DiD) as the key concept that should be applied to all safety
related activities and proposes to refine the description of the five levels of DiD to consider separately
the control of accidents due to multiple failure events and to reinforce the separation between the
consideration of accidents without and with significant core melt; see Table 5.2-1.
NEA/CSNI/R(2016)15
For new reactors, WENRA defined seven safety objectives. The third of these objectives is on
Accident with Core Melt and specifies that these types of accidents should be included in the reactor
design, and that the accidents that can lead to large or early releases should be practically eliminated,
and for those events that are not prevented, they should lead to only limited time and area protective
measures for the public. The fourth safety objective is on the independence of the levels of DiD; this
objective implies that each Safety Feature should be diverse, structurally separated, and functionally
isolated. The RHWG stated that to fulfil the third safety objective in level four situations, independant
safety systems should focus particularly on containment integrity as the main protective barrier for the
environment. These independent safety features should guarantee containment heat removal, control
pressure rise including the effect of non-condensable gases, limit the risk of energetic events such as
hydrogen combustion, and should avoid any uncontrolled leakage paths (through penetrations
including melt-through, access hatches, and failure of liners). In addition, to limit the time and area
consequences for the population, the inside containment source term must be lowered, containment
venting must be efficiently filtered, and dedicated instrumentation must be developed as part of
emergency preparedness. The RHWG recommends to analyse the adequacy of these accident
progression and consequences mitigation features using combined deterministic (best-estimate) and
probabilistic methods.
Adopting this fourth safety objective (specifically, separating safety features devoted to
mitigation and preventing the progression of a core-melt accident) as a reference for existing plants
has major implications for LTO (different cooling systems, electrical power supply, containment
safety systems, IC cables, et al.).
5.2.3. NRC safety requirements
Safety and regulatory requirements for operating reactors (i.e. reactors of the Generation II type) in the
United States are codified in Title 10 of the Code of Federal Regulations (CFR), in particular, Part 50
of the code and its various appendices. There are no particular regulatory requirements to address
severe accident challenges. However, in the design process, the concept of defence-in-depth is
incorporated which provides multiple physical barriers to the uncontrolled release of radioactive
materials to the environment. Furthermore, the design and operation of a nuclear power plant must
provide redundant means to ensure fundamental safety functions: reactivity control, heat removal, and
confinement of radioactivity.
The safety requirements stipulate that consideration be given to severe accident sequences, using
a combination of engineering judgement and probabilistic methods, to assess severe accident
challenges and identify preventive and/or mitigation measures. Acceptable measures need not involve
the application of conservative engineering practices used in setting and evaluating design basis
accidents, but rather should be based upon realistic or best estimate assumptions, methods and
analytical criteria.
The current design safety requirements are met by ensuring that structures, systems or
components perform the following safety functions in a reliable manner:
• Prevent uncontrolled reactivity transients;
• Maintain the reactor in a safe shutdown condition after all shutdown actions;
• Maintain sufficient reactor coolant inventory for core cooling in and after accident
conditions not involving the failure of the reactor coolant pressure boundary;
• Remove heat from the core after a failure of the reactor coolant pressure boundary in order to
limit fuel damage; and
• Transfer heat from safety systems to the ultimate heat sink.
NEA/CSNI/R(2016)15
Since TMI-2, the US Nuclear Regulatory Commission (NRC) issued a number of severe accident
policy statements aimed at ensuring that the severe accident challenges to the current fleet of reactors
are appropriately managed by the industry and that any new reactor design (generation III varieties or
advanced) is as safe as or safer than the Generation II varieties with regard to severe accident
challenges. In response, the industry initiated the development and implementation of severe accident
management (SAM) strategies. This is a voluntary initiative by the industry with no regulatory
oversight requirements. However, the industry is required to keep NRC informed of the status of SAM
activities on a regular basis.
In the unlikely event of a severe accident in which the core melts through the reactor vessel, it is
possible that containment integrity could be breached if the molten core is not sufficiently cooled.
SECY-90-016 (NRC, 1990) recommended criteria for evolutionary LWR designs to provide sufficient
reactor cavity floor space to enhance debris spreading, and provide for quenching debris in the reactor
cavity.
The Electric Power Research Institute (EPRI) requirements document provides floor sizing
criteria of 0.02m2/MWt, and provisions to flood the lower drywell or reactor cavity. While this is not a
regulatory requirement, the design should ensure that the containment will accommodate the
consequences of core-concrete interactions (i.e. basemat ablation not to exceed the limit that will
challenge containment integrity and gas production not to exceed the amount that will over-pressurise
the containment) for approximately 24 hours.
In SECY-93-087 (NRC, 1993), the above provisions were formalised as requirements for the
evolutionary and passive LWR designs to meet the following criteria:
• Provide reactor cavity floor space to enhance debris spreading;
• Provide a means to flood the reactor cavity to assist in the cooling process;
• Protect the containment liner and other structural members with concrete, if necessary;
• Ensure that the best estimate environmental conditions (pressure and temperature) resulting
from core-concrete interactions do not exceed Service Level C for steel containments or
Factored Load Category for concrete containments, for approximately 24 hours; and
• Ensure that the containment capability has margin to accommodate uncertainties in the
environmental conditions from core-concrete interactions.
Since the Fukushima accident in March 2011, there is a renewed debate on the need to provide a
regulatory framework and oversight for beyond design basis accidents. The Near Term Task Force,
put together in the aftermath of the Fukushima accident, recommended that the Commission direct the
staff to initiate actions to enhance the NRC regulatory framework to encompass beyond-design-basis
events (including severe accidents) and their oversight (NRC, 2013). These actions included issuance
of orders and initiation of rulemakings. Work is currently in progress on the mitigation of beyond-
design-basis events (MBDBE) rulemaking. The proposed MBDBE rulemaking has several provisions
one of which is to make generically-applicable previously imposed requirements for the mitigation of
beyond-design-basis external events by Order EA-12-049 (NRC, 2012).
The MBDBE rulemaking is not expected to alter the provisions of SECY-93-087 (NRC, 1993)
with regard to the MCCI issue. However, in light of the lessons learnt from the Fukushima experience,
it may be necessary to address much longer transients and hence, containment behaviour during longer
duration MCCI.
5.2.4. Japanese safety requirements
Through severe accidents at TEPCO’s Fukushima Daiichi Nuclear Power Station, triggered by the
devastating natural events of March 2011, many important lessons on nuclear safety issues were
NEA/CSNI/R(2016)15
learnt. As one of the post-Fukushima actions, improvement of nuclear safety management and
regulation has been promoted by a new nuclear regulatory body, the Nuclear Regulation Authority
(NRA). As an output of a complete review of regulatory requirements and safety guidelines with the
aim of formulating a set of new regulations to protect people and the environment, new regulatory
requirements went into effect on 8 July 2013 that are being back fitted to operating commercial power
reactors.
Based on a concept of “Defence-in-Depth”, essential importance was placed on the third and
fourth layers of defence and the prevention of simultaneous loss of all safety functions due to common
causes such as earthquakes, tsunamis and other external events such as volcanic eruptions, tornadoes
or forest fires. Furthermore, countermeasures are required against internal fires and internal flooding,
and to enhance the reliability of on-site and off-site power sources to deal with the possibility of
station blackout (SBOs). In addition to the above-described countermeasures established at design
basis, it is also required to establish countermeasures for severe accident response against core damage
and containment vessel failure.
In view of MCCI, it is required to prepare equipment and procedures for cooling molten core
material relocated from the reactor pressure vessel (RPV) into the containment vessel in order to
prevent containment vessel failure in the event of severe core damage. The purpose of cooling of the
molten core material relocated to the bottom of the containment vessel is to mitigate the molten core
and concrete interaction) (MCCI) and prevent the spread and contact of the molten core with the
containment vessel boundary.
“Equipment and procedures for cooling the molten core fallen to the bottom of the containment
vessel” should meet the following conditions in addition to other conditions that are commonly
required for SA measures. Those measures that have the same or improved effect can be adopted.
Equipment for injecting water into the bottom of the containment shall be prepared.
Equipment for injecting water into the bottom of the containment shall be redundant or
diversified, independent and dispersed in different locations.
Relevant equipment shall be connected to alternative power sources.
5.2.5. German safety requirements
The revised German “Safety Requirements for Nuclear Power Plants” (BFS, 2012) contain
fundamental and general safety-related requirements within the framework of the non-mandatory
safety standards and rules that should be considered to prevent any damage caused by the operation of
the plant in agreement with the regulations of the Atomic Energy Act (AtG). Regarding the nuclear
power plants operated in Germany, these requirements have to be considered for modification
licences. The German safety requirements are general in nature and do not provide any regulation
specific to the consequences of MCCI. In the following these regulations are outlined, which generally
concern the phenomena and consequences of MCCI.
The superordinate objective is to ensure the confinement of the radioactive materials present in
the nuclear power plant and the shielding of the radiation emanating from them. In order to achieve
this objective, a “defence-in-depth” safety concept shall be implemented in which measures and
equipment are allocated to different levels of defence. Measures and equipment for internal accident
management shall be provided and planned for supplement levels of defence 4b and 4c of the defence-
in-depth concept. These levels of defence are characterised by the following plant conditions:
• Level of defence 4b: events involving the multiple failure of safety equipment.
• Level of defence 4c: accidents involving severe damage to fuel assemblies.
NEA/CSNI/R(2016)15
For level 4c accidents, mitigative internal accident management measures shall be provided for
the purpose of maintaining – by using all available measures and equipment – the integrity of the
containment for as long as possible, excluding or limiting releases of radioactive materials into the
environment, and achieving a long-term controllable plant state. Furthermore, measures shall be
planned to support external accident management in order to assess the consequences of accidents
with potential or actual releases of nuclear materials into the environment, and to mitigate as far as
possible their effects on man and the environment.
Taking into account both internal and external accident management planning, then any release of
radioactive material into the environment caused by the early failure or bypass of the containment
shall be excluded, or their radiological consequences shall be limited so that external accident
management will only be required to a limited spatial and temporal extent.
In order to fulfil these high level requirements, it is mandatory that the MCCI does not cause or
contribute to large and/or early releases on a technical level. This means:
• the gaseous release from MCCI must not threaten the containment by pressure build-up; if
the containment load limit could be reached, adequate countermeasures have to be
implemented (e.g. filtered venting).
• the erosion of the basemat must not lead to early containment failure, taking into account
also local inhomogeneities (cavities or penetrations) in the sump area.
With a view towards potential MCCI scenarios, the safety requirements demand that
depressurisation of the primary circuit be effectively carried out so that there will be no core meltdown
under high pressure. This means that melt ejection from the RPV under high pressure will not be
probable, so that a redistribution of melt to other containment rooms other than the reactor cavity will
be unlikely.
For the design of mitigating measures for internal accident management on level of defence 4c, a
spectrum of events shall be postulated that takes into account all relevant phenomena for accidents that
involve core melting. In this context, special attention shall be paid to those phenomena that are a risk
for containment integrity. Furthermore, the phenomena that have an effect on the release of radioactive
materials and on possible release paths to the environment shall be considered. If the present accident
management measures implemented prove to be ineffective, additional severe accident management
guidelines shall be provided for the emergency response staff. The suitability of the severe accident
management guidelines for achieving the fundamental safety functions shall be demonstrated.
Potential measures to flood the ex-vessel core-melt with water are currently under focus as part of
the considerations for developing SAMGs for individual plants.
5.2.6. French safety requirements
France has only one nuclear utility, Électricité de France (EDF), that is operating a fleet of
58 standardised pressurised water reactors (PWRs) (3 series of 900, 1 300 and 1 450 MWe). Similar to
other Generation II reactors, severe accidents were not considered in their original design. However,
Periodic Safety Reviews (PSRs) conducted every 10 years have led to the design, assessment, and
implementation by EDF of significant plant modifications to include SAM equipment such that as
Passive autocatalytic hydrogen recombiners (PARs), Emergency Filtered Containment Venting
System (EFCVS), containment strengthening, dedicated instrumentation (Cénérino, et al., 2016), and
SAMG.
A PSR is carried out for the whole reactor series considered. Associated studies must be
completed early enough so that modifications to equipment and documents can be deployed on
reactors from the start of their ten-year outage period. The 3rd PSR for the 900 MWe PWRs (34
reactors) ended in 2008-2009 and their ten-year outage period is ongoing. Their 4th PSR started in
NEA/CSNI/R(2016)15
2014 as their 4th outage period is planned from 2019 to 2029. The 3rd PSR for the 1 300 MWe PWRs
(20 reactors) is currently being finalised with their 3rd outage period planned from 2015 to 2021.
From 2009, the French Safety Authority (ASN) with the support of its technical organisation, the
IRSN, started the evaluation of the strategy proposed by EDF for the Plant Lifetime Extension (PLE)
from 40 to 60 years within the specific context of the construction on the Flamanville site of a first
EPR of the generation III (or III+) type. This EPR is close to two Gen II 1 300MWe PWRs that have
been operating since the mid-eighties. Consistent with the position of WENRA, the ASN required
EDF to refer to the safety objectives of the generation III reactors for all the safety studies made for
the Generation II PLE. Concerning severe accidents, these safety objectives for generation III reactors
including the EPR should lead to “only very limited protective measures in area and time for the
public”. Thus, the proposition made by EDF does not only focus on the ageing management of Safety
Systems and Components (SSC), but also includes a safety enhancement programme in which EDF
intends to examine the possibility of implementing measures that includes the improvement of the
EFCVS efficiency, the improvement of the containment decay heat removal without opening the
EFCVS, and measures to avoid corium basemat melt-through in case of RPV failure. The French
regulatory framework does not limit the duration of operation for nuclear facilities and propositions
made by EDF for PLE are evaluated plant by plant, in the frame of their 4th PSR.
In addition, post-Fukushima complementary safety evaluations (CSE) in 2012 led to the
statements ECS-ND1 and ECS ND16 of resolution 2014-DC-0403 (ASN, 2014) by ASN that also
requires EDF to investigate the possibility of improving containment heat removal systems and
implementing measures to avoid basemat melt-through in case of MCCI as part of a set of measures
called ‘’hardened safety. Solutions proposed by EDF will be evaluated in the CSE context plant by
plant. In the framework of the 3rd PSR EDF has been granted with a ten-year extension for the
Fessenheim plants (two 900MWe PWR operated since the late-seventies) with a specific requirement
to deal with their thinner reactor basemats. In answer to this ASN requirement, EDF has implemented
in 2013 significant modifications that includes an ex-vessel corium spreading surface area extension
and thickening. It is described in the section presenting the French PWR containment characteristics to
highlight the differences that can exist within the implementation of concepts for a given reactor type.
Finally, presently, both PLE and CSE proposals by EDF are examined in the frame of 4th PSR of
900 MWe reactors.
5.3. Reactor containment designs
Containment designs not only differ between reactor concepts, but also between different
implementations for the same reactor type. These differences arise from the implementation of safety
and regulatory requirements of the country or region where the plant was built. Differences relate to
the presence of drains or sumps, specific seismic risks, or specific geologic configurations that lead to
constraints on foundations, flooding risk prevention, and from the type of concrete components
(mortar, cement, aggregates) available close to the plant location, among others. These differences
must be accounted for when evaluating basemat melt-through risks. It is not the purpose of this section
to extensively review different implementations, but to present some of these concepts and indicate
general features that may need to be addressed as part of plant-specific MCCI analyses.
5.3.1. Examples of generation II reactors
5.3.1.1. PWRs operated in France
In French 900MWe PWRs, the containment building is made of reinforced concrete with a steel liner.
The surface of the reactor pit is a keyhole geometry of about 30 m² surrounded by vertical concrete
walls, see Figure 5.3-3. An access corridor is located ~1 m above the surface of the basemat. The core
instrumentation room (~50 m²) is separated from the reactor cavity by a vertical concrete wall of
~ 1.5 m thickness. This room is at the same level as the reactor cavity.
NEA/CSNI/R(2016)15
The containment of the French 1300/1450 MWe is ensured by two concrete vessels. The outer
one, designed to withstand external aggressions, is made of reinforced concrete and the inner one is
made of pre-stressed concrete; see the right part of Figure 5.3-2. The access corridor of the reactor pit
is at the same level as the basemat, or slightly above this level (depending on the sites). The total
surface (corridor + reactor cavity) is roughly 42-47 m². Similarly to French 900MWe PWRs, the
reactor pit is located in the centre of the reactor building. Therefore, MCCI studies focus mainly on
axial basemat penetration.
Basemat design and thicknesses (see Figure 5.3-2) and basemat concrete type depends on the
plants site. Most of the 900MWe PWRs have a basemat thickness of around 4.7 m whereas the
basemats of the 1 300MWe and 1 450MWe PWRs range between 3.1 m and 3.6 m thick. 900MWe
CP0 plants located near Fessenheim were designed with a thinner basemat of 1.5 m. 900MWe CP0
plants located near Bugey were designed with two successive concrete slabs. Aside from the reactor
cavity, some other compartments exist. Most of the plant basemats are made using siliceous-rich
concrete whereas the plants in 5 locations have been built using LCS concrete.
Figure 5.3-1: Generic containment designs. Right: PWR900. Left: PWR1300 (credit IRSN)
Figure 5.3-2: Containment sectional views: from Left to Right PWR900-CP0 Fessenheim, PWR900-
CP0 Bugey, PWR900-CP, PWR1300MWe-P4 and PWR1450MWe-N4 (credit IRSN)
NEA/CSNI/R(2016)15
Figure 5.3-3: Typical EDF PWR900 reactor cavity (ref: DES180) Left: side view;
Right: top view (credit IRSN)
As mentioned in section 5.2.6, in the framework of the 3rd PSR EDF had been requested by the
ASN to significantly reduce the risk associated with basemat melt-through before being granted a
10 years lifetime extension of the Fessenheim plant. EDF proposed very significant modifications that
have been evaluated by the ASN with the support of IRSN and considered to be satisfactory. The
implemented modification is a thickening of the basemat of both the reactor cavity and an adjacent
room with a 0.5 m thick layer of self-levelling LCS concrete; see Figure 5.3-4. The reactor cavity has
been linked to this adjacent room using a transfer channel including a concrete fusible plug. In this
adjacent room, concrete vertical walls have been built to prevent spray and sump water to fill the dry
surface area devoted to corium spreading. These modifications allow increasing significantly the
basemat melt-through delay in dry condition and, contrary to the initially designed configuration, no
accident scenario has been found with the current state of knowledge on MCCI that can lead to a
basemat melt-through within the first 24 hours after the accident starts; the period needed in France to
implement the first population protection measures.
Figure 5.3-4: Modifications implemented for the Fessenheim plants (ASN)
NEA/CSNI/R(2016)15
One can mention that to avoid any basemat melt-trough, the corium should spread on the whole
dry surface areas before cooling water is injected on top; this illustrates the tight link between the
MCCI issue and the water management issue. As an example of the possibility to flood the reactor pit
before vessel failure, one can mention for PWRs, the fact that after the inner containment spray
activation, the reactor cavity could be filled with water up to the RCS pipe level within a short delay of
1.5 to 2.5 hours. In spite of a flooded pit, if the reactor vessel fails the corium spreading on the
basemat will be affected and the possibility of an intense steam explosion has to be considered.
Considering these scenarios, the recent IRSN stand is that the reactor cavity should be kept dry
until the vessel lower head fails. This eliminates the risk of a large early radioactive release due to FCI
and also increases the chances to maintain the functionality of the structure, system and components
(SSCs) needed for SAM after vessel failure (e.g. water injection systems), as well as the chances for
the corium to spread over a larger area and then be cooled by water injection from the top as soon as
the water injection systems are recovered. This issue is being considered for NPPs as part of the long-
term operation safety evaluation process. Consequently, EDF proposals are expected for the 900 MWe
series in the framework of their 4th PSR.
5.3.1.2. US BWR
Mark I primary containments consist of an inverted light bulb-shaped drywell vessel surrounded at the
base by a torus-shaped pressure suppression chamber, as shown in Figure 5.3-5. The drywell and
suppression chamber are connected by vent lines (typically 8 equally spaced lines around the base of
the drywell).
Figure 5.3-5: BWR Mark I containment design (3D cutaway and containment
simplified schematics from (NRC, 2011)
NEA/CSNI/R(2016)15
Figure 5.3-6: BWR Mark II design (3D cutaway and containment schematics from (NRC, 2011)
The Mark II containment design retains the basic pressure suppression function of the
Mark I containment, but rearranges the drywell and the suppression chamber into an "over/under"
single pressure vessel configuration as illustrated in Figure 5.3-6. The vessel is supported by a
concrete basemat in which the bottom of the vessel is embedded. The drywell and pressure
suppression chamber are separated by a diaphragm slab. Vertical downcomers connect the two
volumes. Steam released into the drywell during a LOCA is directed into the downcomers and is
discharged below water level in the suppression chamber. The Mark II containment design has been
implemented using three different construction techniques: (1) free-standing steel, (2) reinforced
concrete with steel liner, and (3) post-tensioned concrete with steel liner. Mark II plants also have
different pedestal region geometries. These are shown in Figure 5.3-7.
Figure 5.3-7: Different pedestal region designs
The Mark III containment, shown in Figure 5.3-8, is substantially larger than either the Mark I or
II vessels and houses nearly all the reactor building components. The Mark III containment consists of
NEA/CSNI/R(2016)15
a drywell and pressure suppression chamber inside a primary containment shell that is surrounded by
an enclosure or shield building and various equipment rooms that function as part of the secondary
containment boundary. The Mark III containment design, Figure 5.3-9, has been implemented using
two construction techniques: (1) free-standing steel, and (2) reinforced concrete with steel liner.
Figure 5.3-8: 3D cutaway of BWR Mark III containment
Figure 5.3-9: BWR Mark III design (containment schematics from (NRC, 2011)
General features of GE Mark I, II, and III containment designs that can impact MCCI behaviour
are summarised as follows. As noted above, the US designs vary considerably from plant to plant, but
many feature sumps of various depths and overall volumes that can harbour deeper accumulations of
melt that can be more difficult to cool relative to accumulations on the basemat floor. These
containments feature pedestals that hold up the RPV, with access within the pedestal region through a
doorway. Thus, the geometry for melt spreading consists of the pedestal region with one or more
sumps, followed by flow through a relatively narrow (typically 90-100 cm) doorway opening into the
drywell area. Depending upon melt pour conditions, the doorway flow restriction can result in deeper
melt accumulations in the pedestal region (an example of this behaviour is provided in the
NEA/CSNI/R(2016)15
Section 5.5). Also, there is extensive below-vessel structure in BWRs consisting of control rod drive
(CRD) and in-core instrument (ICI) tube penetrations. This structure can hold up core melt (by virtue
of the heat sink) as well as cause a distributed melt pour condition that would be more amendable to
quenching if water is present on the drywell floor. This occurrence can significantly alter the initial
conditions for the core-concrete interaction phase of the accident.
As noted previously, in the Mark I containment design the steel drywell shell that forms the
pressure boundary is also in relatively close proximity to the reactor vessel, and so direct melt contact
of this structure can pose a threat to containment. For the Mark II containments, the downcomers are
located on the drywell floor, and contact with these structures by melt during core-concrete interaction
may cause localised failures, thereby providing a pathway for melt to relocate to the suppression pool.
However, the suppression pool in the Mark II design is inside containment and so this behaviour
would not bypass containment.
GE BWR plants are deployed across the continental US. The basemats in these plants are thus
constructed from an array of concrete types ranging from Limestone/Common Sand, siliceous, to
basalt.
5.3.1.3. German PWR of Type “Konvoi” description
In the German PWR of type Konvoi the overall thickness of the concrete foundation is
approximately 6 m. This thickness will be eroded during MCCI after several days of interaction, if
MCCi is not terminated by the overlaying water pool (see Figure 5.3-10). However, there are two
specific details which require special consideration within an analysis of a severe accident with MCCI.
It is assumed that these details are representative of features which may be found in similar plant
designs.
Figure 5.3-10: Cross-sectional view (schematic) of the lower part of the Konvoi containment
(GRS, 2002)
At floor level within the cylindrical concrete support structure for the RPV (outer border of
reactor cavity), there are 8 pressure equalisation flaps (see Figure 5.3-10) which will open in case of
pressure differences developing during an accident. Consequently the water from the containment
sump will flow into the annulus between the RPV support structure and the biological shield. About
0.7 m below the bottom of the cavity there are ventilation ducts in a star-like pattern to remove heat
from the concrete structures. The ducts are connected to openings at the bottom of the annulus
between the RPV concrete support structure and the biological shield (primary or inner boundary of
reactor cavity), so that water will enter the ventilation ducts as well.
NEA/CSNI/R(2016)15
One possible scenario identified within a PSA level 2 study (GRS, 2002) is characterised as
follows: When the melt erodes the concrete and penetrates into this network of ducts, melt will flow
towards the containment sump. After a failure of the steel ducts (rising vertically at the inside of the
containment sump) the melt will reach the sump region where it can spread over an area of approx.
200 m². Another scenario, which is possible, is dominated by a first dry MCCI phase and a radial
erosion of the biological shield until its failure. Thereafter, water ingression is possible as well as melt
release into annulus and sump and MCCI may continue. It is important for the further event
progression whether the melt will reach the suction pipe stubs of the emergency core cooling systems.
When the protection tube of the suction pipe develops a leak, for example due to melt impact, there
will be a containment leak. Furthermore, small melt mass flows due to settling of melt particles
generated after melt/water contact could also accumulate significant amounts of core material in the
inside of the sump suction pipes with the potential consequence of a containment failure.
The second detail is that in the bottom of the sump of the containment there are cavities which
provide some amount of free volume at the deepest level of the sump to collect remaining water for
drainage by pumps. Below these cavities there is only ~ 1 m of concrete left until the melt would reach
the steel liner of the containment. If melt fills up these special cavities the local thickness of the heat
generating melt will be higher than what is generally considered as coolable under flooded conditions
and under the 1D assumption of an infinitely extended melt layer, so that the coolability under top
flooding conditions is challenged in these spots.
5.3.2. Example of generation III reactors
5.3.2.1. EPR™
For stabilising the melt in a severe accident, the EPR™ relies on an ex-vessel core catcher located in a
lateral compartment below the reactor pit (see Figure 5.3-11). Thanks to the spatial separation between
pit and spreading compartment, the core-catcher is safe from potentially critical loads related to the
failure of the RPV. Furthermore, an unintentional flooding of the core catcher during power operation
does not affect the safety of the plant. As a consequence of this uncoupling, power operation and
design-basis mitigation measures remain unaffected by the existence of the core catcher.
The melt relocation into the core catcher is preceded by a phase of temporary melt retention in
the pit, which addresses the prediction that the release of molten material from the RPV will likely
take place in several pours. Temporary retention is achieved by a sacrificial concrete layer. Its ablation
and incorporation leads to predictable melt characteristics that are independent of the preceding
accident scenario. Ablation of the sacrificial concrete ultimately exposes the melt gate in the centre of
the pit bottom. After the gate fails, the accumulated melt will spread on the large inner surface of the
core catcher and its depth is expected to be lowered to the point where it would be coolable by
conduction.
Both the reactor pit and the core catcher are initially dry to eliminate the risk of energetic steam
explosion that could cause early containment failure. But the melt arrival thermally destroys triggers
that open passive flooding valves and initiates the gravity-driven inflow of water from the Internal
Refuelling Water Storage Tank (IRWST). The water quickly submerges the outside of the core catcher
and eventually floods the melt’s free surface. Decay power is extracted by evaporation and steam
release into the containment. The condensate flows back into the IRWST from where it is resupplied
by passive overflow. As an option, the Containment Heat Removal System (CHRS) can be used to
actively provide water to the core catcher. This will completely submerge the spreading compartment
and the reactor pit and stop further steam discharge into the containment as a pre-condition to reach
atmospheric pressure conditions in the long-term, without the need of a venting system.
The provision of the core-melt stabilisation system (CMSS) avoids the interaction of the molten
core with the structural concrete and with it the risk of i) penetrating the embedded liner,
NEA/CSNI/R(2016)15
ii) weakening and mechanical deformation of load-bearing structures as well as the basemat itself, and
iii) sustained release of non-condensable gas into the containment atmosphere.
The melt stabilisation process involves the following stages (see Figure 5.3-11):
• RPV failure and initial melt release;
• Temporary melt retention and accumulation in the pit;
• Opening of the gate and melt spreading;
• Flooding and quenching of the spread melt;
• Long term cooling and heat removal to the water.
Sacrificial Material
Protective Layer
IRWST
Spreading Compartment
Sacrificial Material
Core Catcher Melt Discharge Channel Protective Layer Melt Plug
Figure 5.3-11: Main components of the EPR core melt stabilisation system (CMSS)
(Fischer, Long-term melt stabilization as part of the severe accident mitigation strategy of the
European Pressurized Reactor (EPR), 2005)
During this sequence, a transformation of the molten corium into a coolable and cooled
configuration is achieved on the basis of simple physics and without requiring further operator action
or the use of internal or external active systems and by using existing technology and materials that are
commonly available and applied also in other industrial areas.
NEA/CSNI/R(2016)15
5.3.2.2. ESBWR
Figure 5.3-12: ESBWR SA design features including the BIMAC device (NRC, 2010)
The ESBWR design includes a device called the Basemat Internal Melt Arrest and Coolability
(BIMAC) device that is intended to arrest core melt progression in the lower drywell by cooling the
debris both from above and below. The BIMAC device is a passively cooled barrier to core debris on
the lower drywell floor. The design features a series of side-by-side inclined pipes, forming a jacket,
which is passively cooled by natural circulation when subjected to thermal loading. Water from the
Gravity-Driven Cooling System (GDCS) pools enters the pipes via connecting downcomers. Once the
pipes fill up, the debris is also cooled from above from water that flows out of them. The timing and
flows are such that the cooling by natural circulation becomes functional immediately upon actuation,
and cooling by top flooding takes effect subsequently. The design procedure of not immediately
adding water greatly reduces the probability of an energetic steam explosion.
5.3.2.3. EU-ABWR
The EU-ABWR design includes mitigative features against MCCI, consistent with the guidance in
SECY-93-087. The most important features are: a large lower drywell floor area with minimal
obstructions to the spreading of core debris; a lower drywell flooder system (active or passive); and a
sacrificial basaltic concrete layer for the lower drywell floor. The optional design for installing the
core catcher on the lower drywell floor is proposed to exclude any MCCI effects. The core catcher
consists of a round basin to capture the ejected core melt, with lower inclined cooling channels
axisymmetrically arranged, an annulus riser, an annulus downcomer and a central water chamber, and
the passive flooder system. After core melt ejection, the fusible valves of the passive flooder open, and
the suppression pool water flows into the peripheral annulus downcomer of the core catcher. After
core catcher flooding, natural circulation is established within the cooling channels and downcomer,
and the core melt is cooled passively from the top by the flooding water and from the bottom by the
lower inclined cooling channels (Suzuki, Tahara, Kurita, Hamazaki, & Morooka, 2009).
NEA/CSNI/R(2016)15
Figure 5.3-13: EU-ABWR ex-vessel retention device (Hamazaki, et al., 2011)
5.3.2.4. AP1000
The AP1000 design employs the concept of in-vessel retention by external cooling of the lower
head with a cavity flooding system to prevent vessel failure and consequent relocation of core debris
in the cavity and any potential MCCI. Nevertheless, in case of vessel failure, the cavity design
includes a thick layer of concrete to protect the embedded containment shell with an additional thick
concrete layer below the liner elevation. The AP1000 design also relies on safety grade RCS
depressurisation and incorporates plant features to promote debris spreading, consistent with the
guidance in SECY-93-087 regarding debris coolability.
5.3.2.5. APWR
The US-APWR design includes a large area in the reactor cavity to provide floor space for debris
spreading and quenching capability to cool the debris. The design would provide retention and long-
term stabilisation of the molten core debris inside the containment. The melt is cooled by cavity
flooding using two independent means: (1) containment spray utilising water from the in-containment
reactor water storage pit (RWSP), and (2) fire water. Flooding of the reactor cavity occurs by manual
initiation of the system when core damage is detected, provided the water is below a certain level.
Since the US-APWR is designed to fill the reactor cavity with water when a severe accident occurs,
external reactor vessel cooling may be also possible. However, in-vessel retention is not credited for
the US-APWR in the Level 2 PRA study due to its inherent uncertainty.
5.3.2.6. VVER-1000
A crucible-type core-catcher has been designed for the reactor pit of the AES91 model (Khabensky,
Granosky, & Bechta, 2009) of the VVER-1000 series of reactors and for its VVER-1200 evolution
(Zvonarev, et al., 2011). This core-catcher device relies on a steel vessel located directly below the
vessel that is externally cooled by water; the lower head has a conical shape to take benefit from the
critical heat flux increase similar to the ESBWR and US-ABWR core catcher concepts (see for
example (Suzuki, Tahara, Kurita, Hamazaki, & Morooka, 2009). A sophisticated honeycomb
sacrificial structure fills the main part of the core-catcher vessel where each steel cell contains oxidic
materials (see Figure 5.3-14).
NEA/CSNI/R(2016)15
Figure 5.3-14: VVER-1000 (Left) and VVER-1200 (Right) core-catcher devices
The design of this complex sacrificial structure has been defined to meet specific requirements
related to durability during the course of normal plant operations, the limitation of the steam explosion
risk in case of unexpected early flooding, the stability of the solidified melt configuration after the
accident progression has been terminated regardless of the coolant conditions, and finally the ease and
safety of the decommissioning operation. The oxidic sacrificial material blocks contained inside the
steel cells have been selected mainly to guarantee a fast inversion of the metallic and oxidic layers,
thereby increasing the volume of oxidic materials (this lowers the heat flux due to residual power
dilution and risk of recriticality), to limit the hydrogen generation and, similar to the EPRTM, to limit
the uncertainties related to melt composition for all possible accident scenarios. For the AES91
concept, the oxidic sacrificial composition is approximately 70% Fe2O3, 30% Al2O3, 5% SiO2 and
0.1% Gd2O3. After a delay period evaluated to be sufficient for the two layer configuration with the
oxide above the metal to be achieved (see (Zvonarev, et al., 2011) for an illustration of this evolution
through simulation results), the melt is flooded by water.
5.3.2.7. EU-APR1400
A core-catcher has also been designed for the EU-APR1400 version of the APR-1400 PWR reactor
developed by KEPCO in South Korea. This version of the APR-1400 includes a set of SAM devices
and measures added to satisfy specific safety requirements promoted in European countries and
especially the European Utility Requirements (EUR Organisation, 2012). Among these devices, the
Passive Ex-vessel corium retaining and cooling system (PECS) described on Figure 5.3-15 is based on
a core-catcher of inclined walls including a sacrificial concrete layer, a cooling gap and downcomers
and on two trains of gravity-driven cooling water supply systems (Granovsky, 2012).
NEA/CSNI/R(2016)15
Figure 5.3-15: Description of the Passive Ex-vessel corium retaining

and cooling system of the EU-APR1400
5.4. Severe accident management guidance (SAMG)
5.4.1. Introduction
Severe Accident Management Guidance (SAMG) provides an organisational and procedural
framework for coping with beyond design basis accidents. The principal objective of severe accident
management (SAM) is to bring a nuclear power plant to a stable and controlled state following a
severe accident (otherwise referred to as beyond design basis accident involving significant core
damage with the potential for breach of RPV and containment) and to minimise on-site and off-site
radiological consequences. The SAMG framework includes necessary hardware, procedures to prevent
and/or mitigate severe accidents and consequences, operator training to implement such procedures,
and an organisational structure to oversee the implementation.
Early on in its evolution, the approach to SAMG was to provide a set of accident management
procedures, as an extension to emergency operating procedures (EOPs) that cover a finite set of
initiating events. Since the TMI-2 accident, the need to consider a broader range of accident conditions
and a more complete set of initiating events was identified. The events at Fukushima Daiichi have
reinforced this concept. The technical basis for SAMGs was derived largely from the severe accident
research activities that followed the TMI-2 accident. It was found that a generic symptom-based
approach to SAMG is more appropriate to deal with more complex situations, provided the symptoms
and plant parameters associated with them can be correctly identified. This concept facilitated the
development of two generic SAMGs for operating reactors (i.e. reactors of the Generation II type) in
the United States: Boiling Water Reactor Owners Group (BWROG) SAMG for BWRs, and
Pressurised Water Reactor Owners Group (PWROG) SAMG for PWRs. These guidelines are
discussed later in this section in more detail.
Generally, the goal of SAMGs is to identify and implement a limited set of strategies that are
capable of mitigating the consequences of a severe accident. The symptom-based strategies are intended
to minimise the degree to which severe accidents can challenge the integrity of multiple fission product
barriers – fuel or core, RPV, and containment structures – that serve to limit potential off-site
radiological releases. The following objectives should be accomplished to preserve these barriers:
• Heat removal from overheated core;
• Control of containment atmospheric conditions (temperature, pressure, flammable gas
inventory, etc.); and
NEA/CSNI/R(2016)15
• Limited release of fission products into the environment.

The SAMG strategies are formulated to perform a few generic actions to meet these objectives.
The actions are:
• Restore cooling to the fuel in the RPV, and provide a means to cool the debris in the
containment as well as containment atmosphere and spent fuel pool;
• Provide containment overpressure protection by initiating controlled venting;
• Provide means to reduce flammable gas concentration in primary containment and
reactor/auxiliary building; and
• Maximise retention of fission products in the containment.
In the aftermath of the accident at the Fukushima Daiichi nuclear power plant (BWR units with
Mark I type containment) in March 2011, the nuclear power industry and the regulatory organisations
worldwide felt a sense of urgency to revisit the current SAMG and its technical basis. The Fukushima
accident brought to the forefront the potential for extreme external events to extensively degrade and
even fail the means by which symptom-based accident management measures can be deployed. Work
is currently in progress internationally to assess the full extent of consequences of the Fukushima
accident and the significant challenges it posed to accident management measures during and
following the accident. Some modifications to current SAMG are anticipated in the future as a result
of post-Fukushima forensic activities.
The types of challenges to the integrity of fission product barriers that arise during a severe
accident vary as do the SAMG strategies to mitigate these challenges. Moreover, the SAMG strategies
vary between plant types (BWR vs. PWR for example). Also, the plant-specific SAMG and the means
of implementing it are influenced by the status of the plant and symptoms (for example, the extent of
fuel damage and challenge to containment integrity). The variation may not be as significant with
regard to fuel matrix and cladding as fission product barriers. However, the variation is increasingly
more pronounced for RPV or reactor coolant system and primary containment as other barriers. In
what follows, plant-specific SAMGs will be discussed in further detail for BWRS and PWRs with
particular focus on those aspects of SAMGs that are related to the phenomenological topic of MCCI.
5.4.2. BWR Severe Accident Management
Table 5.4-1 summarises the severe accident challenges in aBWR containment, and also the SAMG
strategies for mitigation of challenges.
Table 5.4-1: Challenges and SAM strategies for BWR

Barrier Type Challenges SAM strategies
Containment Overpressure failure Control containment atmosphere (temperature,
pressure), control non-condensable gas generation,
also deploy a controlled containment venting
strategy
Containment bypass/isolation Control containment atmosphere (temperature,
pressure), deploy a controlled venting strategy
Direct containment heating Depressurise RPV 1
1. If the automatic depressurisation system (ADS) is available during an extended accident, then direct
containment heating does not pose a challenge and no additional SAMG strategies are required.
NEA/CSNI/R(2016)15
Table 5.4-1: Challenges and SAM strategies for BWR (Cont.)

Ex-vessel steam explosion 2 Control containment atmosphere and drywell cavity
water level management
Flammable gas combustion in containment and Control containment atmosphere and provide
reactor building hydrogen management systems (igniters, PARs,
etc.)
Liner failure 3 Flood drywell cavity
Core-concrete interaction Flood drywell cavity
Thermal degradation of penetration seal Control containment temperature
causing leakage 4
Of the above challenges, the ones most pertinent to the phenomena of MCCI and debris
coolability are: core-concrete interaction, containment over-pressurisation from non-condensable gas
build-up, and liner failure. The common SAMG strategy to mitigate these challenges is to provide
water in the drywell cavity by means of flooding and/or spray. Another mitigation measure to
supplement the provision for water management is containment venting. These particular challenges
(for both BWRs and PWRs) and corresponding SAMG actions are discussed below in further detail.
Core–concrete interaction
In a postulated core melt accident, if the molten core is not retained in-vessel despite taking severe
accident mitigation actions, the core debris will relocate into the reactor cavity region. The resultant
interaction with structural concrete could potentially result in basemat failure (through erosion) and/or
containment over-pressurisation (by non-condensable gas build-up) and consequent fission product
release to the environment. Although this is a late release event, the potential radiological
consequences (in terms of land and groundwater contamination, as well as latent cancer risk), could be
substantial and warrant effective strategies to prevent or mitigate such a release. As one of several
strategies, the SAMG for many operating LWRs includes flooding the reactor cavity in the event of an
ex-vessel core melt release. The effectiveness of cavity flooding to cool ex-vessel core debris and
mitigate core-concrete interaction (CCI) depends, among other factors, on the mode and timing of
water addition, as well as the heat transfer characteristics of the melt-water interaction.
Flooding the containment at or near vessel breach is considered to be one severe accident
mitigation strategy, among others. Flooding provides a water pool in the containment to cool the core
debris. In the early phase of the cooling process, core debris in contact with an overlying water
undergoes efficient cooling that is dominated by convection, resulting in bulk cooling of the entire
melt mass. The melt sparging rate is also high in this initial phase which precludes stable crust
formation at the melt-water interface. However, the sparging rate gradually decreases due to efficient
heat transfer, and a crust eventually forms. Absent any other cooling mechanisms, the bulk cooling
mechanism is limited by a maximum thickness of solidified melt below the melt-water interface. The
underlying core debris will continue to interact with the surrounding concrete unless another cooling
mechanism is in place.
2. The ex-vessel steam explosion issue is still not fully resolved from a risk perspective and there residual
uncertainties remain regarding the explosion loads and containment structural integrity.
3. Liner failure issue is particular to the Mark I containment type and the issue is considered resolved from a
risk perspective (see NUREG/CR-5423 and NUREG/CR-6025) provided drywell is flooded at or prior to
vessel failure and adequate provision is made for liner submergence.
4.. One of several possibilities explored in the context of Fukushima Dai-ichi accident. Other possibilities are
drywell head flange leakage, gasket failure, etc. – all potentially caused by excessive temperature.
NEA/CSNI/R(2016)15
After stable crust formation occurs, cracks and fissures develop in the solidifying material. These
cracks and fissures serve as pathways for water ingression into molten core material, and provide
augmentation to what would otherwise be an inefficient conduction-limited cooling process. The
effectiveness of this mechanism decreases with increasing concrete content in the corium and has been
found to not be highly sensitive to concrete type.
Another mechanism of cooling the core debris is by eruption of the molten material underneath
the crust formed at the melt water interface. The volcanic-type eruption process is brought about by
gases generated as a result of core-concrete interaction. The eruption mechanism is particularly
effective at high gas flow rates and at an initial stage of melt solidification when crust formed at the
melt water interface is not sufficiently thick or stable. This eruption mechanism results in transport of
melt through the breached crust into the overlying water pool where the melt is quenched to form a
coolable particle bed.
A third cooling mechanism that can supplement the otherwise conduction-limited cooling is
macroscopic crust breach. As noted previously, cracks and fissures in the crust provide pathways for
water ingression into molten material underneath the solidified core debris. Formation of substantially
wider and possibly interconnected cracks (macroscopic crack network) in a long-span crust can further
enhance water ingress cooling. Also, a long-span crust (typical of reactor scale) can collapse from
hydrostatic loading due to standing water atop the crust. This type of crust failure (akin to macroscopic
crust breach) can result in substantial water ingress into molten core debris thus providing additional
cooling.
Containment over-pressurisation
The discharge of steam and also non-condensable gases such as hydrogen, carbon monoxide, and
carbon dioxide from core-concrete interaction (that includes metal-water reactions) will likely increase
the internal pressure of containment. Flooding the drywell cavity will have a beneficial effect of
cooling the debris and slowing down core-concrete interaction and hence, slowing down non-
condensable gas build-up. Moreover, flooding and spray actions will have a beneficial effect on fission
product scrubbing. However, if the water is present in the cavity prior to vessel breach, there is some
risk of an ex-vessel steam explosion depending, among other things, on the depth of water pool. An
appropriate SAMG action is predicated on all these considerations.
Despite water management in the cavity, if the static overpressure in the containment keeps
increasing and eventually exceeds the containment pressure limit, failure will occur. The failure could
be that of containment penetration(s), containment seal(s) (such as the drywell head flange in BWR
Mark I containment designs), or a gross failure of the containment structure. Over-pressurisation could
be reduced by controlled venting and condensation of steam by containment cooling. The role of
overpressure in secondary containments is different because these are not designed as leak-tight
structures. Rather, these structures are designed in some instances with blowout panels to relieve
overpressure from events such as main steam line breaks (for example, at the Fukushima Daiichi
units).
NEA/CSNI/R(2016)15
Liner failure (BWR Mark I)

In BWR Mark I containments, relocation of molten core debris into the drywell cavity pedestal region
occurs upon RPV failure. The liner failure issue arises because the tight containment floor geometry
relative to the large core inventory can lead to molten core debris flowing a short distance and
contacting the drywell liner. In the absence of water on the drywell floor, the thermal load imparted on
the liner by the debris will breach the liner leading to containment failure. This containment failure
mode, specific to Mark I containments, can be mitigated by providing water on the drywell. Past
studies (Theofanous, Amarasooriya, Yan, & and Ratnam, 1991) have shown that flooding the drywell
with water significantly reduces the thermal load on the drywell liner to the point at which liner failure
is highly unlikely (i.e. conditional probability less than 10-3 given a core melt accident).
5.4.3 PWR Severe accident management
Table 5.4-1 summarises severe accident challenges in a PWR containment, and also the SAMG
strategies for mitigation of those challenges.
Table 5.4-1: Severe accident challenges and mitigation strategies for a PWR
Containment Overpressure failure Control containment atmosphere (temperature,
pressure), control non-condensable gas
generation, also deploy a controlled
containment venting strategy
Containment bypass/isolation Control containment atmosphere (temperature,
pressure), deploy a controlled venting strategy
Direct containment heating Depressurise RPV

Containment Ex-vessel steam explosion Control containment atmosphere and reactor pit
cavity water level management
Flammable gas combustion in containment Control containment atmosphere and provide
and reactor building hydrogen management systems (igniters, PARs,
etc.)
Core-concrete interaction Flood reactor pit cavity
Thermal degradation of penetration seal Control containment temperature
causing leakage
Of the above challenges, the one most pertinent to the phenomena of MCCI and debris coolability
is containment over-pressurisation from non-condensable gas build-up. The common SAMG strategy
to mitigate this challenge is to provide water in the cavity by means of flooding and/or spray. Another
mitigation measure to supplement the provision of water management is containment venting.
The Operating Strategies for Severe Accidents (OSSA) framework (Sauvage, Prior, Coffey, &
Mazurkiewicz, 2006) was developed for Gen-3+ plants (like the EPR™) which have dedicated,
generally passive severe accident mitigation measures implemented into their design. These measures
are supported by instrumentation which is adequately qualified and can be credited under severe
accident conditions. The OSSA framework is intended to support the emergency team in managing the
severe accident by recommending various actions deduced from deviations between the course of
events and the intended mitigation path. OSSAs replace the former emergency operating procedures
(EOP), while the SAMGs typically extend them.
NEA/CSNI/R(2016)15
5.5. Plant applications of MCCI
5.5.1. MCCI under dry cavity conditions

Station blackout scenarios can potentially cause the reactor core injection cooling (RCIC) system for
BWRs or auxiliary feedwater (AFW) system for PWRs to be inoperable thus leading to core melting,
reactor vessel failure, and subsequent relocation of core debris onto a dry cavity floor (if the SAMG
for a particular plant or plant type has no provision for a pre-flooded cavity) and consequent dry
MCCI, i.e. MCCI under dry cavity conditions. The cavity may not be fully dry, strictly speaking, in
many, if not most, plants. This is because there is always some nominal leakage during normal
operation (e.g. seal leakage, penetration, etc.). However, this small amount of water will quickly boil
off upon relocation of core debris on the cavity floor.
During the MCCI, depending on the progression of the melt and the particulars of the
containment design, the corium mixture may relocate into other rooms when walls are penetrated
(e.g. see Section 2.5 summary of the Chernobyl MCCI progression). The ability of melt to spread into
an adjacent room once a wall is breached (provided that these rooms are also initially dry) is
determined by the temperature, composition (through the melt viscosity), breach flow area, and also
the gravity head.
In terms of safety criteria, the purpose of MCCI simulations is to establish the time corresponding
to the loss of containment integrity that prevents fission product leakage through the failure location.
Given a realistic containment design, it is important to focus the analysis on the weakest point in that
design, which could be sumps, instrumentation pathways, and/or collecting channels or pipes
embedded in the basemat in the reactor pit. Also, melt relocation into an adjacent room could provide
bypass or constitute the weakest point in the design.
In some reactor designs, early radial melt through could lead to more rapid containment failure
than axial melt through under circumstances that include:
• Bypass to a lower room or spreading in a sump or other structure where the axial concrete
thickness is smaller, or
• When a liner is present and failure of this liner is regulatory considered as a containment
failure, lateral spreading to a position where the liner could be contacted.
Given the details of the containment design, both MCCI and spreading simulation tools may be
needed to evaluate with confidence the potential for containment failure by core-concrete interaction.
Uncertainties regarding the 2D progression of the melt during MCCI as well as the potential for the
material to spread needs to be addressed using a bounding approach that factors in weak points in the
containment design.
5.5.2. MCCI under flooded cavity conditions
Depending on the details of the accident sequence, the availability of safety systems and operator
actions, and the cavity geometry, water may be present on the cavity floor prior to vessel failure. Pre-
existing water has several potential consequences regarding MCCI behaviour:
• Corium-water interaction could lead to the formation of a particle bed that may or may not
be coolable (independently of the steam explosion risk).
• The corium mixture may be partially or fully solidified, depending on the pour rate and
masses of material involved (both water and corium), with the possibility to form a mound or
a pile.
• If a debris bed locally re-melts and forms a corium pool, subsequent spreading under water
onto the surface of the cavity is expected to be dramatically slower in comparison to a dry
NEA/CSNI/R(2016)15
cavity situation (similar to the lava flows under water, with successive crust formation / crust
failure events leading to step-wise melt relocation behaviour).
Depending upon the extent of the debris porosity, the time for the corium debris to re-melt can be
estimated with varying degrees of accuracy. There are ongoing analytical and experimental
investigations in this area, primarily focused on multi-dimensional aspects of debris bed dry-out
behaviour that should reduce uncertainties in these types of evaluations in the future.
5.5.3. Theoretical basis for plant calculations
There are various levels of sophistication currently employed in plant analyses regarding MCCI
behaviour. The simplest approach is to assume that, following vessel failure, the corium spreads
instantaneously over the entire available area in the reactor pit. This scenario can be described by
virtually all available MCCI codes today. However, more realistic scenarios involve additional factors
that can include:
• Limited corium spreading in the reactor pit.
• The presence of water with possible sustained water injection.
• Transfer to adjacent compartment(s) with varying elevation(s), leading to different melt
depths and different interaction geometries (e.g. typical cylindrical, notch, rectilinear).
These phenomena, possibly combined and/or coupled, could change considerably the subsequent
evolution of the ablation pattern, with potential variations in melt-through times in different
compartments. The following realistic factors may be expected:
In the case of corium relocation into a localised area within a large containment, a limited fraction
of the available floor area may be involved due to limitations on spreading resulting from a high
corium solid fraction, or complete solidification during spreading (see Figure 5.5-1. The extent of
spreading will influence the ablation kinetics because of the initially reduced contact area between the
corium inventory and the reactor pit concrete basemat and walls. There is a potential for subsequent
evolution of the debris bed (i.e. re-melting of solid debris, spreading, and delayed onset of ablation)
With pre-existing water on the containment floor, the water may quench or partially solidify the
relocating corium, thus requiring the corium to reheat before concrete ablation can be established. The
presence of water may also hinder corium spreading, particularly if the corium solid fraction is high,
further reducing the extent of core-concrete interaction (due to augmented cooling) compared to a dry
case (see Figure 5.5-1). However, if the debris is not fully or partially cooled, then this scenario may
lead to locally higher heat fluxes and faster ablation for some length of time until re-spreading occurs.
In the case of a corium transfer to adjacent compartments by failure of walls, localised flow
restrictions, or gates that are designed to fail, crucial phenomena to be investigated are the possible
plugging within the hole, as well as limitations on the spreading rate due to the flow restriction
(see Figure 5.5-2).
NEA/CSNI/R(2016)15
localized corium pouring and

Reactor pit accumulation
with or without
water solid debris/crust
molten corium
basemat
fast ablation
Figure 5.5-1: MCCI scenario with a localised corium pouring and accumulation in reactor pit,
without or with water
corium pouring from vessel:
multiple pours of minimum
mass fraction frac
Reactor
adjacent compartment
pit
spread corium
basemat
Gate/door/breach
Figure 5.5-2: Corium transfer from reactor pit to other compartments through a breach or a gate
Most of the available codes (see Section 4) can describe the impact of water injection and top
quenching, but within the confines of the originally assumed reactor pit geometry. There are limited
capabilities for evaluating the restriction of concrete ablation to a part of reactor pit walls due to
localised corium accumulations (as in Figure 5.5-1); see Section 5.5.4.
There are limitations with the current set of codes in their ability to model additional spreading of
corium outside the initial cavity as shown in Figure 5.5-2 cannot be modelled at the current time.
Some codes are able to model spreading, but most only cover the initial spreading phase(s) from the
reactor vessel involving hydrodynamics and corium freezing processes. Subsequent corium re-melting
and refreezing, as well as spreading involving wall failure(s), is currently not treated.
This summary shows that additional model development to treat more realistic scenarios is
needed to reduce residual uncertainties and to gain more safety margin. In particular, no dedicated
models are available to treat scenarios where the core debris may re-melt, leading to re-spreading of
core debris, as well as situations where core debris can relocate through pathways after localised
failures by ablation. There has been progress made on the ability to treat scenarios involving corium
spreading into more complicated containment geometries, with localised ablation treated within these
structures, an example of which is provided in the next section.
NEA/CSNI/R(2016)15
5.5.4. Results and outcomes of some benchmarking activities made on generic plants
5.5.4.1. Benchmark on a generic LWR situation
The MCCI reactor benchmark performed in the frame of the European SARNET Network of
Excellence in 2008 (Cranga, et al., 2010) compared predictions for a generic light-water reactor case
obtained by most of the available MCCI codes for a basemat thickness of up to 6 m. The objective of
this benchmark was to compare the trends of different code predictions in terms of a homogeneous
pool configuration with an isotropic heat flux distribution, as well as a configuration permitting pool
stratification.
The inventory was typical of that of a large BWR which explains the fast ablation kinetics even
in the case of a homogeneous pool assuming an isotropic heat flux. Main results are displayed in
Figure 5.5-3 and Figure 5.5-4.
Figure 5.5-3: Ablation kinetics assuming a homogeneous pool, LCS (left) and siliceous (right)
concretes see (Cranga, et al., 2010)
Figure 5.5-4: Ablation kinetics with possible pool stratification: LCS (left) and siliceous (right)
concretes from (Cranga, et al., 2010)
The comparison of code results showed some areas of agreement; i.e. for the homogeneous
configuration case, results from all codes are near each other for limestone-common sand concrete
and, to a less extent, for siliceous concrete (differences in the axial melt-through for a basemat depth
of 4 m at most by a factor of 2), except for MELCOR (UPM) and CORQUENCH (VTT) codes. But
they also exhibited strong discrepancies:
NEA/CSNI/R(2016)15
• For the case of a homogeneous melt configuration with siliceous concrete, MELCOR and
CORQUENCH codes deviate substantially from other codes. Besides reasons due to model
differences and different thermo-chemistry data, the impact of the particular cavity erosion
algorithm in the MELCOR code and of very simplified assumptions for determining the
cavity shape in the CORQUENCH code contribute to the deviations of these two codes
compared to the others;
Figure 5.5-5: Cavity shapes with possible stratified pool for LCS after 4 days of interaction (left)
and siliceous (right) concretes from (Cranga, et al., 2010)
• For the case in which the pool can stratify, deviations between codes become very large
mainly due to differences in configuration evolution assumptions and the oxide/metal heat
transfer model. Code deviations are still greater in case of siliceous concrete compared to
limestone-sand concrete because the stratified configuration lasts a longer time.
These results pointed out the need to improve the reliability of the pool configuration model and
the heat transfer model at the metal/oxide interface in case of a stratified pool particularly for typical
BWR core inventory configurations. They also illustrate the substantial need to validate the codes
against available 2D MCCI tests. It is also important to perform additional reactor material
experiments using siliceous concrete to further validate MCCI models in the case of siliceous
concrete, where the code discrepancies are larger and melt-through of the basemat might occur earlier
than in the case of limestone-sand concrete.
5.5.4.2. Benchmark on a VVER 1 000 situation
As an example some results are given for a MCCI reactor benchmark performed in the early 2010s in
the frame of SARNET and organised by INRNE on a VVER1000 reactor type (Gencheva, et al.,
2012). This benchmark confirmed general trends obtained by previous studies, in particular the MCCI
reactor benchmark exercise of previous Section, on the influence of stratification criteria and of the
level of the oxide/metal convective heat transfer.
The results of this study indicate that for VVER1000 reactors with a high iron mass fraction
concrete basemat (around 16% instead of around 7% in PWR), the stability of the stratified
configuration with metal layer below might be enhanced (see Figure 5.5-7), resulting in a melt-through
delay in the range of 24 to 30 h after MCCI onset (see Figure 5.5-8) when using the same pool
stratification assumptions and models as in PWR applications.
NEA/CSNI/R(2016)15
Seven calculations without quenching have been done with different MCCI computer codes: six
with MEDICIS/ASTECv2 code, and one with the WECHSL code. Calculations of the cavity
boundary profile evolution with MEDICIS\ASTECv2 under dry conditions are presented in
Figure 5.5-6 (INRNE calculation). Starting from an initially stratified pool configuration with the
metal above the oxide, a pronounced radial ablation is first observed in the INRNE computational
results. After approximatively 3 hours, the configuration then switches to homogeneous. A secondary
stratification with metal below appears at 12.6h after MCCI onset which results in faster axial erosion
until basemat melt-through is reached at 27h. Similar results are obtained by most of the other
organisations using ASTEC-V2 (IRSN, NUBIKI, TU). In these calculations, the standard Greene’s
correlation for evaluating the oxide/metal heat transfer was used. Accelerated axial ablation after pool
stratification with metal below occurs; this is due to the impact of high oxide/metal convective heat
transfer coefficient and the focusing of decay energy downwards in the final stratified phase. In the
GRS ASTEC-V2 calculation, the pool is assumed to be stratified all the time with the metal layer
below the oxide but with a lower heat transfer coefficient between the oxide and metal. In the KIT
(WECHSL) calculation the metal phase is assumed to remain homogeneously dispersed in the form of
droplets. The evolution of the metal layer mass is displayed below in Figure 5.5-7. The thickness of
this layer that is set to zero in a homogeneous configuration is reduced in stratified conditions due to
oxidation that mainly occurs by oxidation from concrete decomposition gases (steam and carbon
dioxide). These different assumptions lead to significant differences in the axial melt through delays
displayed in Figure 5.5-8.
Figure 5.5-6: Evolution of cavity boundary profile Figure 5.5-7: Mass of the metal layer in dry
versus time in INRNE calculation conditions from (Gencheva, et al., 2012)
(MEDICIS\ASTECv2) in dry conditions from
(Gencheva, et al., 2012)
NEA/CSNI/R(2016)15
Dry MCCI With top water injection

conditions
INRNE 27.3h 34.2h
IRSN 25.5h 30.9h

Figure 5.5-8: Axial ablation in dry conditions from
(Gencheva, et al., 2012) Table 5.5-1: Comparison of melt-through delays
calculated without and with top water injection
from (Gencheva, et al., 2012)
Due to the high iron content in concrete and the rather low gas content, which both contribute to
maintain the pool stratification with metal below until the end of the calculation, comparable fast axial
ablation and early melt-through have been observed for both dry MCCI conditions and in the case
where water is injected on top by INRNE and IRSN that computed both situations using the simplified
coolability models available at that time in MEDICIS\ASTECv2, see Table 5.5-1. The limited impact
of top flooding on melt-through delay in this typical VVER configuration (around 16% iron mass
fraction compared to around 7% for a PWR) would have to be re-assessed in case of significant
advances in knowledge related to the effect of metal on both ablation and cooling derived from the
experiments and studies recommended later in this chapter.
5.5.4.3. Support to the assessment of EPRTM
To be able to demonstrate that, for all considered severe accident scenarios, the core debris will be
completely released from the RPV before the end of the molten core interaction with sacrificial
concrete in the pit and thus before the start of melt spreading, the MCCI code COSACO, see Chap.
3.2.6, was complemented by a simple physical model for the lower RPV. In this model, the RPV lower
steel structures (incl. core support plate and lower head) are lumped into a single mass denoted as
“RPV-bottom”. Starting with the first release of melt into the pit, the model calculates the heat–up of
this structure considering solely the radiant heat emitted from the surface of the MCCI pool
(see Figure 5.5-9).
NEA/CSNI/R(2016)15
Heavy Reflector
Reactor Pressure
Vessel
Metallic Melt
Core Region
Metallic Melt
Metallic Melt
Lower Core Support Plate Metallic Melt
Oxidic Melt
Oxidic Melt Metallic Melt
Metal Melt Metal Melt
Metal Melt
Oxidic Melt Metal Melt
Oxidic Melt
Oxide
Metallic Melt
Melt
Metallic
OxideMelt
Melt
QRad
Oxide
Metallic Melt
Melt
QRad RPV-Bottom
Lower Core Support Plate
Oxidic Melt
Lower Head QRad
QRad
Sacrificial Concrete
(a) (b) (c)

Figure 5.5-9: Lower head, lower core support plate and RPV-bottom
In the analyses the RPV-bottom is assumed to ultimately fail once its temperature exceeds a pre-
set value (1 300°C). At this instant, the steel mass associated with the RPV-bottom and all remaining
in-vessel core debris is added to the MCCI pool.
Selected modelling assumptions
The cylindrical geometry of the EPR™ pit is modelled according to the scheme given in Section 3.2-5.
Only the sacrificial concrete layer is taken into account. Its outer perimeter is defined by the protective
layer, which is considered to form a static, adiabatic boundary. This is justified by the very low
thermal conductivity of the protective material and its experimentally confirmed stability against both
the metallic and oxidic melt fraction under MCCI conditions (Hellmann & Fischer, 2007).
To account for the uncertainties related to the gas-induced mixing between the oxidic and
metallic phase of the melt, two enveloping cases are considered for the analysis: the layered case (L),
in which all constituents of the MCCI pool, namely metal, oxide and slag (concrete accumulating atop
the upper metal layer) are stratified according to their densities and the mixed case (M), in which all
constituents are mixed and in thermodynamic and chemical equilibrium. In the layered case, the core
oxide phase always has a higher initial density than the steel. This is because any in-vessel “dense
metal” phase does not remain under the oxidising conditions during MCCI. The subsequent addition
of concrete decomposition products steadily reduces the density of the oxidic melt ρ oxide up to the point
when it becomes less than the density of the metal phase ρ metal . Therefore, the layered case (L)
includes the following two situations:
ρoxide > ρ metal >> ρ slag (before layer inversion),
ρ metal > ρoxide + slag (after layer inversion)
During the layer inversion the core oxides take the top position. The related incorporation of the
accumulated concrete slag causes a drastic drop in oxide density.
For the heat transfer to the concrete, the COSACO model predicts a nearly isotropic heat flux
distribution inside the oxidic/mixed pool due to effective forced convection heat transfer (gas induced
mixing), consistent with the results of the BALI experiments (Bonnet, 1999). These tests only model
the pool of the MCCI but not the specifics of the concrete erosion process. Therefore, they cannot
explain the faster radial melt progression observed in some 2D VULCANO and CCI experiments with
(non-reinforced) siliceous concrete, including EPR™ sacrificial concrete (see Chapter 2). With respect
to the proof of the retention function the isotropic model is nevertheless conservative as it maximises
axial erosion and thus leads to earlier gate melt-through.
NEA/CSNI/R(2016)15
Melt release from the RPV

Predicting the melt masses in the lower head at the time of RPV melt-through as well as the fraction of
this mass which is initially released into the pit entails significant uncertainties. The approach chosen
to deal with these uncertainties is to perform a bounding parametric variation of the masses of oxidic
and metallic melt released during the first pour which envelops the range of predictions of integral
codes.
In combination with: (i) conservative assumptions for the progression of the melt towards the
gate and (ii) a large variation of release time / decay power, this procedure effectively covers a wide
spectrum of severe accident scenarios, including Large(LB) / Small (SB) Break LOCAs.
Example of results for the MCCI in the reactor pit
The evolution of cavity profiles for an example LBLOCA calculated with either the layered (L) or
mixed (M) modelling option, is shown in Figure 5.5-10 for an assumed single-pour melt release from
the RPV. With the layered (L) assumption (left figure), the metal layer quickly expands in radial
direction. During this early phase, all zirconium contained in either the metallic or oxidic phase is
oxidised. The early contact with the protective layer insulates the metal layer along its circumference
and thus keeps its temperature high. Consequently, the heat fluxes and ablation rates in the metal
region after the ultimate layer flip are high, as they are supplied by the transient cool-down of the
metal layer.
L_LB-LOCA M-LB-LOCA
4 4
geometry geometry
6420.0 s 6420.0 s
6822.0 s 6822.0 s
7422.0 s 7422.0 s
3 8022.0 s 3 8022.0 s
8622.0 s 8622.0 s
9222.0 s 9222.0 s
Height [m]
Height [m]
9822.0 s 9822.0 s
10422.0 s 10422.0 s
2 11022.0 2 11022.0 s
11622.0 s 11622.0 s
12224.0 s 12222.0 s
12822.0 s
13422.0 s
1 1
0 0
0 1 2 3 0 1 2 3
Radius [m] Radius [m]
Figure 5.5-10: Erosion profiles for a typical LB-LOCA scenario with assumed layered (L on left)
and mixed (M – right) melt configurations
When applying the mixed melt assumption, the oxide phase is initially subcooled, because of the
postulated global thermal equilibrium. The calculated melt progression is therefore more complex as it
is influenced by the interplay between oxide phase reheating, gas mixing and crusts melting.
Depending on oxide superheat, the MCCI can temporarily involve high rates of erosion and gas
generation, as well as correspondingly high void and pool levels.
NEA/CSNI/R(2016)15
Figure 5.5-11: Ablation front progression and RPV-bottom failure for various outflow sequences
When multiple pours are assumed the transient evolution of erosion profiles becomes more
complex. However, to compare the various transient cases it is sufficient to compare their axial
ablation curves. Such example curves are given in Figure 5.5-11. The convention used to denote these
cases is: the first number (e.g. 40%) refers to the initial corium release fraction, while the two letters
denote the scenario: early (E, straight lines) or late (L, dashed lines) and the melt state: layered (L) and
mixed (M), respectively.
Figure 5.5-11 illustrates that ablation progresses slower for: (i) late scenarios (less decay power),
(ii) lower initial melt release fractions, and (iii) mixed melts, when compared to their layered
counterparts. It demonstrates that – for all analysed scenarios and assumptions – the “Failure of RPV
bottom” event occurs significantly before the melt has penetrated even half of the sacrificial concrete
layer provided above the gate. This proves the self-adjusting characteristic of the temporary melt
retention function. It seems worth emphasising that this function is not based on guaranteeing a certain
minimum duration of the retention period (it can vary with decay power and melt release sequence),
but a maximum erosion depth at the time when all core debris is present in the MCCI pool!
Once this is achieved the MCCI continues and further concrete decomposition products are added
to the oxidic melt, which causes a steady cool-down in accordance with the related decrease in melt
liquidus temperature.
The next important issue is the assessment of the spectrum of melt states at the end of the pit
MCCI, as this is needed to demonstrate the ability of the melt to spread. Figure 5.5-12 gives the
corresponding predictions for the final melt composition showing that for all analysed parametric
cases, the melt composition becomes dominated by concrete decomposition products and iron and
chromium oxide originating from the oxidation of steel (in total >50mol%).
The only small differences in the final melt composition and in particular in the core oxide
content between the analysed cases translate into correspondingly small variations in melt temperature
and material properties. In addition, also the final masses of the melt stay within a narrow range. This
is due to the presence of the protective layer which prescribes the amount of (sacrificial) concrete to be
incorporated into the melt.
This remains true also if radial ablation would be considered to be more pronounced than the
axial one. This is because the melt now reaches the cylindrical part of the protective layer earlier,
leading to a transition from 2D MCCI into a 1D MCCI. As a consequence, melt temperatures will
increase and a comparably higher fraction of the decay power will be transported to and emitted from
the upper free surface. This in turn will result in an earlier failure of the RPV bottom, a later failure of
NEA/CSNI/R(2016)15
the gate and – because of the higher temperature – a lower viscosity of the melt. All these changes
have a positive impact on melt accumulation and conditioning.
100%
90%
80%
70%
60%
Molar fraction
Cr2O3
50%
Concrete Oxides
Iron Oxides
40%
SiO2
Core Oxides
30%
20%
10%
0%
-L
M
L
M
E-
E-
E-
L-
L-
L-
E-
L-
40
%
%
%
%
d-
40
60
80
40
60
80
80
80
r
da
an
St
Figure 5.5-12: Calculated Oxide melt compositions (in mol%)

at the end of the temporary retention in the pit
The results of the presented analyses attempt to demonstrate that the retention phase in the pit
decouples the later processes in the core catcher from the preceding events, including the accident
scenario, in-vessel melt progression, RPV failure mode, and melt release sequence.
5.5.4.4. Plant application made in the US
The severe accident source terms for selected accident sequences in a BWR Mark I plant were
analysed using MELCOR in a 1993 study (NRC, 1993) for the purpose of comparing the MELCOR
results with the NUREG-1150 (NRC, 1990) results calculated by the Source Term Code package
(STCP). The three severe accident sequences are: low pressure shut down, short term SBO, and design
basis LOCA, concurrent with complete failure of the emergency core cooling system. For the short
term SBO scenario, two cases were run: one with dry cavity and one with flooded cavity. A second
study (BNL, 1990) involved MELCOR calculations of the long term SBO in a BWR Mark I plant with
failure to depressurise the reactor vessel, and a comparison with the results calculated by the STCP
package. Table 5.5-2 lists a number of plant applications of MELCOR in the 90s.
Table 5.5-2: MELCOR plant applications
Plant Country Plant and containment type
TMI-2 United States B&W PWR
Surry United States 3-loop Westinghouse PWR
LaSalle United States BWR/5, Mark II
Peach Bottom United States BWR/4, Mark I
Grand Gulf United states BWR/6, Mark III
Oconee United States B&W PWR
Calvert Cliffs United States CE 3-loop PWR
Zion United States 4-loop PWR
Point Beach United States 2-loop PWR
Browns Ferry United States BWR/4, Mark I
TVO Finland ABB BWR
Loviisa Finland VVER-440
Muhleberg Switzerland BWR/4, Mark I
Beznau Switzerland 2-loop PWR
Gosgen Switzerland 3-loop PWR
Leibstadt Switzerland BWR/6, Mark III
NEA/CSNI/R(2016)15
Plant Country Plant and containment type

“José Cabrera” Spain 1-loop PWR
Garona Spain BWR/3, Mark I
Almaraz (I and II) Spain 3-loop PWR
Cofrentes Spain BWR/6, Mark III
Asco (I and II) Spain 3-loop PWR
Trillo Spain 3-loop KWU
Vandellos Spain 3-loop PWR
Many of these plant applications involved core-concrete interaction calculations using the
CORCON module in the CAV package of MELCOR. The Peach Bottom and Zion calculations were
performed as part of MELCOR/MAAP comparison, whereas Oconee calculations were performed as
part of MELCOR/STCP comparison. MELCOR calculations for the Surry plant were performed for
updating the source term for three accident sequences (AG, S2D and S3D). A TMLB' station blackout
analysis was also performed for Surry to compare two different updates of MELCOR. The Browns
Ferry (same plant and containment type as Peach Bottom) calculations were done to investigate the
relative effect of plant-specific features on accident progression, in particular, ex-vessel phenomena.
MELCOR calculations for the LaSalle plant were performed as part of an integrated risk assessment in
the phenomenology and risk uncertainty evaluation programme. The Grand Gulf calculations were
performed as part of the Level 3 PRA to assess severe accident risks under low power and shutdown
conditions.
Some of these plant calculations have been repeated in later years with updated code versions of
MELCOR for new regulatory applications. MELCOR calculations were also performed in the late
1990s and early 2000s as part of the confirmatory safety analysis of advanced reactor designs
(generation III varieties). Notable among these applications are those relating to the ABWR, ESBWR,
AP600/AP1000, and EPR plant designs.
The ABWR MELCOR calculations were performed for two low-pressure and three high-pressure
sequences to study the potential effects of core-concrete interactions, and to evaluate the source terms
including those generated by such interactions (Kmetyk, 1994).Sensitivity studies were done on the
impact of assuming limestone rather than basaltic concrete and on the effect of quenching core debris
in the cavity compared to having hot, unquenched debris present.
The accident scenario for ESBWR calculations involved a transient initiated by a loss of power.
GDCS injection to the reactor pressure vessel (RPV) and wetwell injection to the RPV through
equalisation lines are not available. The ADS system is actuated if and when the water level in the
RPV downcomer reaches a specified level. The PCCS and the pool makeup system are available, thus
allowing long-term containment heat removal. However, the GDCS deluge and the basemat internal
melt arrest and coolability (BiMAC) system are not available for debris bed cooling. In addition,
containment venting is not available. Because the GDCS deluge and the BiMAC system are not
credited in this scenario, MCCI is predicted to occur following relocation of core debris from the
breached reactor vessel. MCCI results in concrete basemat erosion and production of concrete
decomposition gases.
MCCI calculations were performed for the scenario for accident duration of 48 hours using
MELCOR code version 1.8.6. This version, when developed, effectively had no phenomenological
(i.e. physics-based and supported by experimental data) model for ex-vessel debris coolability with an
overlying water pool. Heat flux partitioning (i.e. heat flux from melt to overlying water pool and heat
flux from melt to basemat as well as sidewall) in the melt pool is performed in an empirical manner.
User guidance (best practice values) is provided for the partitioning ratios, based on calibration of
calculated thermal-hydraulic signatures with selected test data. Using this version, the erosion of the
NEA/CSNI/R(2016)15
basemat concrete at the end of the calculation was calculated to be about 1.5 m in the downward
direction and 1.7 m in the radial direction (Figure 5.5-13).
Figure 5.5-13: Calculated axial and radial ablation for an ESBWR transient
As a result of continued MCCI as well as continued ex-vessel oxidation of remaining core

zirconium, large quantities of hydrogen and carbon monoxide were generated (Figure 5.5-14). The
amount of carbon dioxide production was negligibly small because of the basaltic concrete
composition assumed for the drywell floor. MCCI calculations were also performed for a variation of
the accident scenario whereby the GDCS deluge is successful and thus credited.
Figure 5.5-14: Non-condensable and steam production from core-concrete

interactions for an ESBWR transient
The ablation profile for this case is plotted in Figure 5.5-15. The non-condensable gas generation
is plotted in Figure 5.5-16.
Figure 5.5-15: Calculated axial and radial ablation for an ESBWR transient
where GDCS deluge is successful
NEA/CSNI/R(2016)15
Figure 5.5-16: Non-condensable and steam production from core-concrete interactions

for an ESBWR transient where GDCS deluge is successful
Calculations were also performed for the AP1000 design using the MELCOR Code
Version 1.8.6 to confirm the amount of basemat ablation. The calculations showed a maximum
ablation depth of about 1.3 m for both limestone and basaltic concrete 2.5 days after accident
initiation, assuming a dry reactor cavity and uniform distribution of debris within the reactor cavity.
The dry reactor cavity assumption was merely a simplification reflecting a bounding scenario. The
ablation rates predicted by MELCOR were lower than those predicted by MAAP, partially as a result
of late RPV failure in the MELCOR calculation (8 hours in MELCOR versus 2 hours in MAAP).
The accident scenario for AP1000 calculations involved a nominal 17 cm break in the direct
vessel injection (DVI) line located in one of the two reactor coolant system loops, resulting in flow of
water from one accumulator, one core makeup tank and the in-containment refuelling water storage
tank into the lower containment region, including the cavity. The ADS is successfully actuated,
resulting in a low reactor coolant system pressure during core damage. Even though one of the
accumulators and a CMT are assumed to inject water into the RPV to replenish the RCS inventory,
these systems are not sufficient to make up for the loss of inventory through the DVI line break. The
PRHR system is assumed to be unavailable for decay heat removal during this accident scenario.
Another accident scenario involves a spurious actuation of the ADS stages 1, 2 and 3 resulting in
the loss of RCS inventory eventually leading to a “medium” pressure inside the reactor coolant
system. Following the reduction of the reactor coolant system pressure below the accumulator
injection set point, it is assumed that one of the two accumulators would be able to inject. The failure
of the CMT injection and stage 4 ADS is also assumed, thereby, preventing automatic IRWST
injection for core cooling. Furthermore, the PRHR system is assumed unavailable for decay heat
removal during this accident scenario.
Figure 5.5-17 shows the maximum concrete erosion in the radial and the axial (downward)
directions. After 60 hours into the accident, the basemat is calculated to erode over 1.0 m in the radial
direction and about 1.25 m in the downward (axial) direction. The concrete basemat integrity is not
expected to be challenged for many more hours in this case; nevertheless, the containment steel liner is
predicted to be penetrated in about 44 hours into the accident. Figure 5.5-18 shows the equivalent
erosion plots for the case of basaltic concrete (sensitivity case 2).
NEA/CSNI/R(2016)15
4.00
3.50
3.00 Radial (Sensitivity case 1)

Axial (Sensitivity case 1)
Erosion Depth (m)

Radial (3D with vessel breach)
2.50 Axial (3D with vessel breach)
Cavity Wall Thickness
2.00 Cavity Floor Thickness
1.50
1.00
0.50
0.00
0 10 20 30 40 50 60
Time (hr)
Figure 5.5-17: Calculated axial and radial ablation for an AP1000

transient case of limestone concrete
Figure 5.5-18: Calculated axial and radial ablation for an AP1000

transient case of basaltic concrete
MELCOR calculations were performed for the eight sensitivity cases listed in Table 5.5-3 in order
to assess the impact of MCCI, the containment spray actuation, and the containment shell exterior water
coverage on the containment response behaviour in AP1000. The RPV failure modes, cavity conditions,
concrete type, and the actuation of containment sprays are shown in the table. It is assumed in the
calculations that any relocated debris will immediately cover the entire cavity floor area.
Table 5.5-3: MCCI sensitivity calculations for AP1000
Sensitivity Debris relocation mode Cavity Concrete type Containment cooling
from RPV condition
1 Slow Dry Limestone Sprays off
2 Slow Dry Basaltic Sprays off
3 Slow Partially Limestone Sprays off
flooded
4 Rapid Partially Limestone Sprays off
flooded
5 Not applicable (IVR) Deeply Not applicable Sprays on for 2 hours
flooded
6 Slow Partially Limestone Sprays off
flooded aggregate/common sand
7 Not applicable (IVR) Deeply Not applicable Sprays off; partial wetted coverage on
flooded containment shell exterior
8 Not applicable (IVR) Deeply Not applicable Sprays off; 60% wetted coverage on
flooded containment shell exterior
NEA/CSNI/R(2016)15
The sensitivity case 1 is based on the second accident scenario (i.e. spurious actuation of the ADS
stages 1 through 3) resulting in vessel breach. The cavity concrete is limestone aggregate type. It is
assumed that the RPV fails on the side resulting in a slow debris relocation rate to the cavity over an
extended period of time.
The sensitivity case 4 is a variation of sensitivity case 1 except this case considers a rapid
relocation of the core debris from the RPV into a partially flooded cavity. Figure 5.5-19 shows the
maximum concrete erosion in the radial and axial (downward) directions. At 60 hours into the
accident, the axial erosion is 1.30 m while the radial erosion is less than a metre. Based on the
calculated concrete erosion rates in the radial and the axial directions, the penetration of the basemat,
in either the axial or the radial direction, is not expected to occur for many days.
Figure 5.5-19: Calculated axial and radial ablation for an AP1000 transient
for a rapid relocation sensitivity case with partially flooded cavity
Two most recent examples of MELCOR plant analysis are: State-of-the-Art Reactor Consequence
Analysis (SOARCA), and containment venting analysis. The SOARCA project was a comprehensive
study focused on providing a realistic evaluation of accident progression, source term, and off-site
consequences for select scenarios for the Peach Bottom plant (BWR Mark I) and the Surry Power
Station (PWR) by using the best available MELCOR modelling features. The analysis considered
mitigative measures (e.g. emergency operating procedures, SAMG, etc.), plant capabilities, as well as
the most current emergency preparedness practices, thus contributing to a more realistic evaluation of
consequences. The containment venting study, likewise, is an evaluation of accident progression,
source term, and off-site consequences for select scenarios in a BWR Mark I) and the Surry Power
Station (PWR) by using the best available MELCOR modelling features.
The MCCI modelling capability of MELCOR is based on CORCON-Mod3. The models assume
that a conduction-limited crust forms at the melt-water interface and that this crust is impervious to
any form of water ingression. Thus, MELCOR predicts continued melt-concrete interaction
underneath the crust at a higher rate even after cavity flooding and hence, eventual basemat
penetration by MCCI. Also, MELCOR calculates the melt spread area through use of a mass balance
and a basic empirical hydraulic relationship which spreads the melt based on the melt height and
temperature. The code does not currently take into account viscosity or leading-edge freezing effects,
and thereby introduces some uncertainties in the melt spread calculation.
Since the CORCON-Mod3 formulation of MCCI into MELCOR, a number of international
research programs and experiments have been conducted (notably, MACE, NEA-MCCI1, and NEA-
MCCI2). Insights and models from these research activities have been integrated into the stand-alone
CORQUENCH code (see Section 4). Phenomena accounted for in CORQUENCH include crust
NEA/CSNI/R(2016)15
anchoring and debris cooling by water ingression and melt eruption mechanisms, and these
phenomena can have potentially large impacts on MCCI progression. The CORQUENCH code is able
to evaluate melt debris coolability under top flooding as well as the axial and radial ablation of
concrete due to MCCI in a more realistic manner. Also, with regard to melt spreading physics, a
second stand-alone code, MELTSPREAD, was developed. MELTSPREAD calculates the area over
which the melt spreads, taking into account a number of physical phenomena including varying melt
viscosity, heat transfer, and solidification at the leading edge of the spreading melt. In addition to
predicting the lateral spread of the melt, MELTSPREAD has the ability to predict melt-liner attack and
failure of the liner if the melt is predicted to contact the drywell liner (near the floor).
Recently, simplified versions of water ingression and melt eruption models akin to those
deployed in CORQUENCH have been incorporated into the latest working version of MELCOR
(MELCOR 2.1) and the new models are being assessed against experimental data and being
benchmarked against other codes.
5.5.4.5. Study of the Fukushima Daiichi Unit 1 (1F1) ex-vessel situation
In a recent analysis of Fukushima Daiichi Unit 1 (1F1) accident, the MELCOR v2.1 code predictions
of melt spreading and MCCI were compared to the MELTSPREAD and CORQUENCH 3.03 codes
(Robb K. R., 2014). Regarding melt spreading, MELTSPREAD includes relatively detailed modelling
of core debris relocation including fluid-mechanics effects, heat transfer to overlying water and
underlying concrete, and finally mechanistically calculated solidification at the debris leading edge. In
contrast, the MELCOR code predicts the melt spread area through the use of a mass balance and a
user-specified hydraulic relationship that spreads the melt based on the melt height and temperature.
Despite these differences in modelling approaches, the overall differences in predicted spreading
behaviour between the two codes were found to be not that large. Although agreement for this
particular 1F1 sequence is reasonable, it is not clear from this study how well the two codes would
compare under a different set of conditions such as a hotter or colder melt release. Additional
comparison of the MELCOR MCCI module with CORQUENCH under the same set of modelling
assumptions (i.e. with the water ingression and melt eruption cooling mechanisms deactivated in
CORQUENCH, so that an impervious upper crust boundary condition was applied in both
simulations) indicated that the two codes yield similar results in terms of cavity ablation and
combustible gas production.
Aside from this comparison, additional analyses were carried out with MELTSPREAD and
CORQUENCH to analyse the ex-vessel accident sequence for 1F1 (Robb, Francis, & Farmer,
Enhanced Ex-Vessel Analysis for Fukushima Daiichi Unit 1: Melt Spreading and Core-Concrete
Interaction Analyses with MELTSPREAD and CORQUENCH, 2013). The high-level objective of this
study was to provide best estimate predictions of ex-vessel core melt accident progression and final
debris configuration for this reactor accident. The evaluation required the integration of data from
several sources. In particular, three sets of melt pour conditions were compiled based on best-estimate
MAAP (Luxat & Gabor, 2013) and MELCOR (Gauntt, et al., 2012) simulations. The time-dependent
melt pour conditions (composition, flowrate, temperature) were used as input to MELTSPREAD to
predict melt propagation, basemat attack, cladding oxidation (viz. H2 and CO production), debris
cooling, and drywell liner attack during the transient spreading phase. One of the principal outcomes
of this analysis was the extent to which the floor was covered by core debris during spreading, since
the depth of debris to be cooled in the long term is inversely proportional to the floor area covered by
spreading melt. The MELTSPREAD predictions of the local melt depth, composition, and temperature
at the end of the spreading transient were then used to define input for CORQUENCH. CORQUENCH
was then utilised to evaluate the long-term debris cooling behaviour in the containment, including the
amount of concrete ablation and non-condensable gas generation.
One unique aspect of the CORQUENCH analysis was the development of a cavity discretisation
model that allowed localised MCCI behaviour within the pedestal and drywell regions to be analysed
NEA/CSNI/R(2016)15
in an integral manner. To this end, the containment was divided into six regions: sumps (1), inner
pedestal (2), inside edge of the pedestal (3), doorway (4) between the drywell and pedestal, an area
extending from the doorway to the drywell liner (5), and the far-field drywell (6). These regions are
numbered and illustrated in Figure 5.5-20.
Drywell Liner
Pedestal Wall
Figure 5.5-20: 1f1 Containment basemat Discretisation approach for CORQUENCH analysis
The sumps (1) were modelled as 2D cylindrical cavities with walls that are higher than the melt.
In reality, the sumps are square and the accumulated melt depth above the height of the sumps will not
be in direct contact with the sump walls. The initial collapsed melt depth in the sumps for the
MELCOR cases ranged from 0.2-1.9 m while all the MAAP cases were 1.4 m deep. The sumps are
1.2 m deep, and so the majority of the melt was in contact with the sump walls for these cases. The
balance of the inner pedestal region (2) included the area between and around the sumps; this region
was modelled using the 1D geometry option available as part of the CORQUENCH code input. The
pedestal edge (3) region was modelled using a 2D cylindrical geometry similar to that for the sumps.
The diameter of the cavity was the same as that of the actual pedestal, 5.0 m. The amount of melt per
unit area in the edge region, as determined from the MELTSPREAD results, was extrapolated to fill
the inner portion of the 2D cylinder. This conserved the melt height at the pedestal walls and the
pedestal curvature. These parameters were conserved in order to conserve the heat and mass transfer at
the walls. Other important parameters, such as combustible gas generation during core–concrete
interaction, were scaled appropriately. The doorway region (4), the area outside the doorway (5), and
the far-field drywell (6) were modelled as 2D notch geometries (also available as a user-defined
modelling assumption).
Although this methodology allowed the spatial ablation variation to be evaluated, there were
several compromises and limitations associated with the technique. Each region was modelled
independently;
i.e. heat and mass transfer between regions was not modelled. However, the cross-sectional area of
melt in neighbouring regions is much lower than that in contact with concrete and water (or the
containment atmosphere), and the lateral heat transfer between zones is expected to be much lower
than the heat transfer to the water (or atmosphere) and concrete. The swelling of the melt by the gases
released during concrete ablation may cause regions of the melt to rise and spread to other regions.
Not allowing the melt to spread from one region to the next results in a less coolable melt
configuration. Finally, radial ablation may cause regions to expand into one another. This effect will
increase with the extent of radial ablation, but for most regions this influence is expected to be
negligible.
NEA/CSNI/R(2016)15
The MELCOR and MAAP simulation results for Unit 1 span a wide range of melt pour
conditions from the RPV ranging from a gradual pour of low-temperature core debris (MELCOR) to a
rapid pour of high-temperature melt (MAAP). The results of the MELTSPREAD analysis were highly
dependent upon the pour scenario (see Figure 5.5-21). For the low-temperature gradual MELCOR
pour, the code predicted lethargic spreading of a highly viscous melt over a period of approximately
4 000 seconds. Concrete ablation is minimal in these cases since the high viscosity limits convection
from the melt to the underlying concrete. Despite the low flow rate, in all cases the melt is predicted to
eventually spread out of the pedestal doorway and contact the liner. When water is present, the core
debris spreading is limited to a total of approximately 33 m2.
MAAP Melt Spread Distribution MELCOR Melt Spread Distribution

(120 seconds) (4 000 seconds)
Figure 5.5-21: MAAP (left) and MELCOR (right) core debris distributions
at the end of the spreading phase
In contrast, for the high-temperature MAAP pours, MELTSPREAD predicted the sump plates
rapidly ablated through in less than 5 seconds. Furthermore, the concrete erosion during spreading was
significant with ~ 2m3 ablated in the first 2 minutes; the maximum ablation depth of ~ 20 cm outside
the pedestal door. The melt was also predicted to fully cover the pedestal and drywell annulus floor
areas, which sum to approximately 111 m2.
Figure 5.5-22: Cross sections of containment ablation
NEA/CSNI/R(2016)15
The results of the long-term debris coolability analysis with CORQUENCH assumed sufficient
water was injected into containment to cover the debris starting 15 hours after shutdown. Under this
condition the simulations indicate that the melt was coolable over the long term. The predicted
concrete ablation was less than that necessary to reach the liner through downward melt progression.
The MELCOR case, which contained relatively cool melt, readily cooled within 2.5 hours after
relocation with limited concrete ablation in the sump regions (~18 cm) and less than 10 cm ablation
elsewhere; see Figure 5.5-22 and Figure 5.5-23. A total of 76 kg of H2 and 103 kg of CO were
predicted to be generated during core-concrete interactions. The MAAP cases, which contained
relatively hot melt, cooled approximately 22.5 hours after melt relocation and resulted in 65 cm of
concrete ablation in the sump region and less than 23 cm elsewhere; see Figure 5.5-24. Large amounts
of H2 (700 kg), CO (750 kg), and CO2 (490 kg) were predicted to be generated during concrete
ablation for the MAAP cases.
(A) 16.4 hours after SCRAM (B) 16.4 hours after SCRAM
Figure 5.5-23: MELCOR cavity profile for cross section A and B after 1 hour of CORQUENCH
simulation time and end of simulation
NEA/CSNI/R(2016)15
Figure 5.5-24: MAAP cavity profile for cross section A and B after 60 min. of CORQUENCH
simulation time and end of simulation
5.5.4.6. Example of PSA2 analysis of MCCI phenomena: German PWR

In 2001 GRS performed a comprehensive probabilistic safety analysis (PSA) with the objective to
evaluate the available PSA methods and to demonstrate their usability for practical applications for an
advanced German PWR (GRS, 2002). The NPP at Neckarwestheim, unit 2 (GKN-2) was selected as
the reference plant for this study. Part of this work consisted of the level 2 PSA for normal power
operation, which covered the events from the beginning of core melting up to the release of
radionuclides into the environment. For deterministic simulations of the accident progression and thus
for safety assessments the integral code MELCOR was used, and for relevant individual aspects
concerning the containment atmosphere the code RALOC (Klein-Heßling & Arndt, 1996) was
applied.
The principal structure of the event tree used for the PSA is shown in Figure 5.5-25. With regard
to MCCI the following branching points are of concern:
• Events between RPV failure and melt-sump water contact, and
• Events after melt-sump water contact.
These events cover the following phenomena related to core melt behaviour in the containment:
• Melt spreading in the lower part of the containment, taking into account possible damage to
components (sump suction lines).
NEA/CSNI/R(2016)15
• Core-concrete interaction before and after sump water contact, and

• Penetration of concrete foundation.
Figure 5.5-25: Principal set-up of the event tree (GRS, 2002)

The event tree analysis requires that even very complicated processes have to be represented by
branching probabilities. Beside the well-known “classical” containment threats related to MCCI – the
pressure increase inside the containment due to long-term core-concrete interaction and the potential
penetration of the concrete foundation – the reference plant has an additional vulnerability with regard
to a possible melt attack on the sump suction lines.
The following sections summarise how these threatening phenomena are considered in the event
tree analysis.
Melt spreading in the lower part of the containment
Section 5.3.1.3 identifies a scenario, in which the melt may enter the sump region in the PWR under
consideration. There it can spread over an area of approximately 200 m². It is important for the further
event progression whether the melt will reach the suction pipe stubs of the emergency core cooling
systems.
For the assessment of melt spreading in the sump region the dedicated GRS code LAVA
(Allelein, Breest, & Spengler, 1999) was applied. Although the LAVA code is currently not validated
for the boundary condition of spreading under water this situation was approximately considered with
a reduced force of gravity and an increased heat transfer at the melt surface compared to a dry
spreading scenario. Assuming conservative conditions for the initial state of the melt spreading (no
initial undercooling of the melt) the LAVA calculations predicted that the melt will cover the whole
sump area shortly after it enters the ventilation ducts. Under the conditions described above, the lower
part of the sump suction pipes will then be surrounded by core material, but its surface level was
estimated to be lower than what would be necessary for a direct flow of core material into the opening
of the pipes. It can however not be excluded that the protection tube of these pipes will fail due to the
thermal load and melt and/or water may then enter the annulus between protection tube and sump
suction pipe. Furthermore, small melt mass flows due to settling of melt particles generated after
melt/water contact could also accumulate significant amounts of core material in the inside of the
NEA/CSNI/R(2016)15
sump suction pipes. When the protection tube of the suction pipe develops a leak due to melt impact
inside the annulus, there will be a containment leak.
Taking these phenomena into account, the probability of a failure of at least one sump suction
pipe was estimated to be between 0% and 10% (homogeneous distribution) for sequences with core
melt in the reactor cavity.
Core-concrete interaction before and after sump water contact and penetration of concrete foundation
According to existing experimental evidence a stopping of the core concrete interaction before
penetration of the foundation was not implemented in the event tree. This means that gas from the
interaction keeps being released. A considerable part consists of non-condensable gases
(e. g. hydrogen, carbon monoxide), leading to a continuous pressure increase inside the containment
until the pressure relief has to be activated. At the same time this means that all sequences with failure
of the RPV lead finally to a penetration of the melt into the ground.
The velocity of the erosion primarily depends on the heat flux from the melt downward into the
concrete. This is dependent on the decay heat and thus on the time period since reactor shutdown.
Uncertainties are due to the fraction of radionuclides which will leave the melt and the partition of the
heat flux to the bottom and to the top.
Figure 5.5-26: Maximum vertical erosion calculated with MELCOR
A number of MCCI calculations were performed with MELCOR version 1.8.4. The calculations
considered four variations: 1) three layer pool configuration & gas film model for the heat transfer
between melt and concrete; 2) five-layer pool configuration & gas film model; 3) homogeneous
configuration & gas film model; and 4) homogeneous configuration and slag layer model for the heat
transfer between melt and concrete. Further model options to be selected by the user were selected in
agreement with the recommendations by the code developers (NRC, 1997).
Based on these calculations (Figure 5.5-26) and taking into account the results of international
research projects on MCCI, for example from the MACE project, probability distributions for axial
erosion velocities for different time periods after RPV failure were defined for the event tree analysis.
From that the probabilities for different progress levels of erosion within certain time frames were
obtained (Table 5.5-4).
NEA/CSNI/R(2016)15
Table 5.5-4: Probabilities for the erosion of the concrete foundation due to core melt
Probability (first lines) and times (second lines) of the
Phase of concrete erosion transition from core damage states to various depths
of concrete erosion
„Dry period“ after RPV failure before core melt reaches
0.06 0.32 0.30
the water
< 5h 5-20 h 20-100 h
filled ventilation ducts
0.32
Time between core damage state until melt reaches 0.20 0.46 0.02
the containment steel shell embedded in concrete RPV intact
< 1 day 1-5 days > 5 days
Time between core damage state until melt reaches 0.29 0.32 0.07
the ground < 5 days 5-15 days > 15 days
Table 5.5-4 contains information about times which pass until certain erosion depths (to the
ventilation ducts, to the containment steel shell, to the ground) are reached. The values in the second
line of the table are typical time periods; the first line (bold numbers) contains the pertinent
probabilities (mean over all core damage states). There is a pit at the bottom of the sump where the
concrete is thinner than elsewhere (see Figure 5.3-10). This has been taken into account for the time
values given.
Conclusions
For all core damage states a small but significant probability of the failure of the containment function
due to the melt-through of a sump suction line or due to a failure of the containment pressure relief
system was obtained. If the RPV does not remain intact, a core-concrete interaction will take place. A
coast-down of this reaction was not assumed. Therefore all sequences with failure of the RPV finally
were assumed to involve finally a penetration of the concrete foundation as well.
NEA/CSNI/R(2016)15
6. Summary and recommendations
Since the publication of the second CSNI specialist meeting on molten core-concrete interactions or
MCCI (Alsmeyer, et al., 1992) over twenty years ago – a seminal report on the then state-of-the-art,
and also the publication of a European Commission state-of-the-art report (SOAR) on the subject
matter shortly thereafter (Alsmeyer H. , et al., 1995), significant progress has been made in practically
all aspects of research and development related to MCCI. The list includes prototypic material
experiments at different geometric scales (some approaching quarter scale to full-size reactor cavity
geometry), making considerable improvement to the analytical tools (many codes now incorporate far
more mechanistic models of the phenomena of interest), and finally performing an increasing number
of validation, verification, and benchmarking exercises as well as a much wider plant applications
ranging from confirmatory safety analysis to new and advanced reactor licensing. The current SOAR
documents this large body of work highlighting the progress made in the last twenty years related to
the understanding of concrete ablation and corium coolability mechanisms. This report takes
advantage of the SOAR on corium concrete interaction recently published by the SARNET European
severe accident network (Cranga, et al., State Of the Art Report on MCCI in dry conditions: analysis
of experiments and modelling, 2013).
The state-of-the-art report (SOAR) provides a background discussion of safety issues relevant to
core-concrete interactions and melt coolability and related containment failure modes, an overview of
various experiment programs that have been carried in the areas of MCCI and debris coolability, a
description and assessment of various analytical tools (“codes”) that have been developed to analyse
MCCI behaviour and finally, a summary of plant analysis activities that have been carried out using
these codes. The report focuses primarily on the progress made in the last two decades on the thermal-
hydraulic aspects of MCCI and mentions in passing some early research programs dealing with fission
products aspects of MCCI. Finally, the report discusses general aspects of severe accident
management (SAM) strategies aimed at achieving melt stabilisation in both generation II and
generation III reactors.
It is important to note that in the early days of MCCI research, the focus was mostly on
understanding the basic physical phenomena such as heat and mass transfer between various
components involved in MCCI (i.e. molten core, coolant, and structural concrete or metal), chemical
processes involved in the interaction, transport behaviour of various product species, and the impact of
MCCI on containment integrity – all through a combination of experimental and analytical research
activities. Much of the experiments in the early days were conducted at small-scale and with simulant
materials, and were carefully designed to study individual physical phenomenon in sufficient detail so
as to be able to develop models that can analytically replicate the phenomenon at full reactor scale.
In contrast, the majority of more recent MCCI experimental research focused on integral
experiments, mostly with prototypic core materials to assess the overall progression and effect of
MCCI.
The experimental programs described in Chapter 2 played a major role in improving the
understanding of MCCI phenomena and provided a much needed data base for model development
NEA/CSNI/R(2016)15
and validation. Most of the tools, discussed in Chapter 3, are now able to simulate with somewhat
simplified assumptions the following behaviour:
• Concrete ablation for different concrete compositions;
• Evolution of corium pool configurations (homogeneous or stratified) as a function of
concrete decomposition gas release, density evolution of the oxide and metal phases, as well
as oxidation reactions; and
• Different cooling mechanisms when water is added from the top.
The treatment of thermal-hydraulics of a well-mixed corium pool in the presence of the concrete
decomposition gases is contrasted by the complexity of the ablation mechanism where the heated
concrete, a highly heterogeneous material, is gradually incorporated into the melt through an evolving
melt-concrete interface that is still difficult to observe experimentally and capture from a modelling
point of view. Because many of the models are not mechanistic, several parameters are empirically
fitted to reproduce as best as possible the scaled experimental results. Attempts to model MCCI by a
multi-scale computational approach to eliminate these tuned parameters with more mechanistic
models have been unsuccessful to date. This is mostly due to the great difficulty of observing and
measuring directly the local phenomena needed to validate multi-scale modelling approaches.
Nevertheless, in the near future it seems reasonable to reach an intermediate level of complexity to
improve existing models. As an example, the recent activities performed by Seiler and Combeau
(Seiler & Combeau, 2014) to develop a universal model of transient interface temperature in
multicomponent solid-liquid systems should be encouraged.
With regard to cooling mechanisms, top flooding appears to be a practical accident management
strategy to achieve enhanced cooling. A boiling heat transfer condition in the absence of other cooling
mechanisms at the surface of the corium crust is not enough to significantly slow down basemat
ablation. More complex mechanisms play an important role in enhancing heat transfer between the
corium and the overlying boiling water pool. Depending upon the melt composition and conditions,
four mechanisms have been identified in flooded cavity tests that can contribute to core debris
quenching. These mechanisms include: i) bulk cooling in which gas sparging is initially sufficient to
preclude stable crust formation at the melt/water interface (and therefore, efficient heat transfer is
achieved); ii) water ingression through fissures in the core material that augments what would
otherwise be a conduction-limited cooling process; iii) melt (or volcanic) eruptions that lead to a
highly porous overlying particle bed that is readily coolable, and iv) transient breach of crusts that
form during the quench process, leading to water infiltration below the crust with concurrent increase
in the debris cooling rate. Bulk cooling is important as it establishes the initial conditions for the long-
term cooling processes that involves crust formation and growth by water ingression as well as melt
eruptions. Transient crust breach is important as it allows water to re-establish contact with underlying
core melt, thereby allowing additional cooling to proceed by the water ingression and melt eruption
mechanisms.
While existing data and experiments indicate that debris coolability can be achieved within an
envelope that is principally based on concrete type, melt depth, and timing of cavity flooding, this
envelope does not encompass the full range of accident conditions that can be encountered in certain
plant configurations. Neither does the envelope encompass various abstractions of melt progression in-
vessel which give rise to different initial and boundary conditions for ex-vessel melt progression and
also, wide variations in concrete constituents within the two major types investigated in the
experimental programme and consequent effects of such variations.
The validation work, summarised in Chapter 4, underlines the scaling issues and illustrates the
fact that we have reached a reasonable level of confidence in extrapolating experimental conditions to
different reactor situations. For full-scale plant safety assessments that are discussed in Chapter 5,
NEA/CSNI/R(2016)15
approaches appear more pragmatic whereby MCCI phenomena are analysed based on conservative
assumptions with respect to the weaknesses of the containment design.
The fission product release during core/concrete interaction is only briefly mentioned in the
report as it has not been the main focus of the MCCI experiments in the last two decades. The aerosols
released during corium/concrete interaction contain mainly elements from the concrete. The release of
uranium or low-volatile fission product is enhanced by the presence of metal in the melt and by the
higher gas content of limestone common sand concrete but remains low. Interaction with silicon to
form silicates tends to lower the release of fission products of main interest like barium and strontium.
These various activities, carried out over the last three decades, have significantly increased our
level of understanding regarding MCCI behaviour under both wet and dry cavity conditions. Much has
been learnt from these experiments in terms of the overall MCCI behaviour at a plant scale. The data
gathered has also helped develop coolability models as well as improving existing MCCI models in
many stand-alone as well as integral codes. Depending upon containment design, regulatory
requirements, and accident management considerations that are unique to each country and reactor
type, the current level of understanding in this area is sufficient for conservative reactor safety
assessments. Some other important lessons learnt from these activities are highlighted in the next
paragraph.
The accident sequences and the possibility of operator intervention result in a broad range of
possible initial conditions at time of vessel failure. Following the accident at Three Mile Island and
some studies of melt interactions with concrete, it was presumed that core degradation would be very
heterogeneous with central regions of the core melting while peripheral regions were barely degraded.
Additional core materials would cascade for protracted periods from the reactor vessel as core debris
attacked concrete. A certain fraction of the cladding would not be oxidised at the time of core debris
relocation to the lower head of the pressure vessel and upon vessel breach there would be a chemical
component to the heat generation in the core debris. Additionally, the state of knowledge about late in-
vessel melt progression is incomplete (particularly for BWRs). Thus, there is considerable uncertainty
regarding the MCCI initial conditions that includes the timing of RPV failure; the initial temperature,
mass, and composition of the core debris; the possibility of segregation of metal and oxide melt
phases; the pour rate of the melt from the RPV that is determined principally by the melting rate of
residual core material, and to a lesser extent by the opening in the RPV lower head; and finally, the
timing of water injection (if any).
The remainder of this last chapter highlights the most important remaining issues and residual
uncertainties therein, and provides recommendations to address them in the future in order to increase
the reliability of reactor simulations. These issues are: (1) long-term core-concrete interaction
behaviour; (2) realistic plant simulations; and (3) coolability enhancement under top flooding
conditions.
6.1. Long term core-concrete interaction behaviour
The past MCCI experiments with prototypic materials were conducted for a relatively short duration,
in part, because the facility constraints did not allow significantly longer duration experiments.
Admittedly, in many of these experiments, the relatively short duration was adequate to assure melt
stabilisation and slowing down basemat ablation to a level that is considered acceptable for regulatory
purposes (i.e. ablation limited to a specified amount by say 24 hours into accident).
The Fukushima accidents suggest that a much longer transient is quite likely. A recent draft
report (OECD/NEA, Under preparation) on safety research post-Fukushima by an international group
of experts noted that MCCI very likely occurred in Unit 1, and probably as well in Unit 3 for some
time, but did not lead to a significant melt release outside the containment vessel to the reactor
building. It is not clear if MCCI occurred in Unit 2 but if it did, it would have been after many hours if
NEA/CSNI/R(2016)15
not several days following accident initiation. The analyses performed to date in the NEA BSAF
project phase 1 have not provided a consensus view on the MCCI issue. In particular, the termination
of the MCCI process in Units 1 and 3 as calculated by the different codes (OECD/NEA, 2015) is
impacted significantly by differences in melt pour conditions predicted by different codes at reactor
vessel failure. The latest results of plant inspections at Unit 1 have provided interesting results; i.e. no
significant damage inside the lower containment is visible (Yamanaka, 2015). These findings put into
question those analyses results which predict ongoing MCCI for a long time, especially in the presence
of water. Hence, there is a need to obtain longer duration experimental data if the shorter duration
experimental data cannot be extrapolated to reactor situation with a high degree of confidence.
Longer duration experiments will provide data needed to: (1) confirm that intermittent
phenomena like melt eruptions are reproducible; and (2) investigate if the crust formed by water
ingression is stable. Long duration experiments will also provide data on long term behaviour dealing
with the final phase of the interaction, i.e. the time when the heat flux to concrete is low enough that it
can be dissipated by conduction into concrete without further ablation, or the heat flux that is applied
in a specific coolant circuit of a core catcher. Finally, long term behaviour also refers to situations
wherein the concrete fraction within the melt and the heat flux level are representative of the situation
after many hours of interaction (typically, 12 hours or more). The subject of long-term behaviour vis-
à-vis further research needs and recommendations will be discussed in the following paragraphs, with
particular reference to dry cavity and wet cavity situations, respectively.
6.1.1. Ablation in dry situation
In the past, experiments under dry cavity conditions were run successfully over a period of a few
hours. The main insight from these tests was that the ablation profile depended largely on the concrete
composition. Based on different experiments and different heating techniques, more rapid radial
ablation (relative to axial) was observed for siliceous concrete, whereas LCS concrete showed an
isotropic (uniform axial and radial) ablation profile. Since there is currently not an accepted
phenomenological explanation for this behaviour, the question remains if it is reliable to extrapolate
this result to reactor scale for a longer duration ablation process. The general scaling issue, equally
important to ablation in a wet cavity situation, is discussed later in this chapter.
Another more complex issue is associated with intermittent ablation bursts that are observed in
experiments; it is not clear if this is a result of crust instability or rather a result of concrete spalling
due to mechanical stability. Depending on the phenomena, the characteristic time period can be
several hours; e.g. crust dissolution processes with siliceous concrete. In this case the test duration has
to be long enough to observe at least two or three ablation bursts. In terms of scaling to confidentially
reproduce the phenomena that occur at the corium concrete interface, the typical approach has been to
preserve the prototypic heat flux to the concrete resulting from fission product decay heat in the core
debris. Under these conditions, extending test duration means that the concrete thickness has to be
significantly increased.
Also, it is important to recognise that in every facility, the size of the test section, the heating
technique, and/or the operating procedure always induces some transient system effects. These
transient effects are not modelled in the simulation tools (codes). As a result, when the transient effects
are dominant at the experimental scale, the codes cannot reproduce accurately the final cavity shape
which is commonly used to estimate the ablation rate. As a recommendation for future test
interpretations, we note that the comparison of ablation rates for a given concrete composition cannot
be deduced from the final cavity shape, but only from the test periods where both radial and axial
ablation reach steady states. As most transient behaviour induced by experiment techniques are not
modelled in the simulation tools, reproduction of the final cavity shape is not a legitimate validation
goal.
NEA/CSNI/R(2016)15
Since the thickness of the concrete walls and the corium mass will always be limited in
comparison to the reactor case, to address long term behaviour in future tests it is important to increase
the initial concrete fraction in the melt and reduce in a consistent way the power injected to reproduce
prototypic long term heat fluxes values. Future tests should also allow be able to run under steady state
conditions for a longer duration. In addition, an experimentation objective should be to reduce the
duration of the initial transient period leading to the onset of ablation in both radial and axial
directions. In this spirit, experiment techniques that can contribute to homogenous initial melting of
the corium, as well as limiting initial crust formation during the phase in which the melt initially
contacts the concrete (either by pouring or by in-situ melt generation), should be encouraged.
6.1.2. Ablation in wet situation
Core-concrete interaction under wet cavity conditions (i.e. in the presence of an overlying water layer)
has been the subject of intense investigation over the last two decades. Testing in this area has focused
on systematic investigation of ex-vessel debris coolability. For top flooding, two particular coolability
mechanisms – melt eruption and water ingression – were investigated in a number of separate effect
tests, latter more so than the former. In the context of long-term CCI, these two mechanisms
(or phenomena) are discussed below in more detail with particular emphasis placed on the current
level of understanding, knowledge gaps, and recommendations for future research where applicable.
6.1.2.1. Melt eruption mechanism
Experimentally, the melt eruption mechanism appears to be a periodic phenomenon which is currently
modelled in a simplified manner in the codes in terms of an average melt entrainment coefficient. The
most recent CCI test resulted in melt eruptions even with low gas content siliceous concrete, but the
phenomenon occurred only two times during the test that ran for ~ 2 hours. Thus, the extrapolation of
an average melt entrainment coefficient deduced from a test of relatively short duration to long-term
CCI is inherently uncertain and, as such, should be used with some caution. It is important to conduct
longer duration experiments to see if multiple melt eruptions occur. Data from these experiments will
reduce uncertainties in current melt eruption models and will provide better confidence in
extrapolating to reactor scale.
Under long test operating conditions involving top flooding, one systematic drawback of the
experiments is the top crust anchoring phenomena. In some promising tests performed with top
flooding, even in some recent ones involving early cavity flooding, the upper crust eventually
anchored to the side walls. The anchoring phenomenon unrealistically reduces the efficiency of the
melt ejection phenomena because a gap between the pool and the upper crust appears and then
increases due to concrete densification upon melting as well as loss of liquid corium as eruptions
occur. At the beginning of the process, crust anchoring could also create a pressure build-up effect
below the crust that experimentally distorts the eruption process (e.g. promoting an extrusion-type
ejection process versus an entrainment process). Crust strength measurements made on samples
obtained from reactor material core debris coolability experiments and supporting structural analyses
indicate that a floating crust boundary condition is likely for reactor applications involving pit
diameters of typically 6 m.
In this spirit, experiment techniques that can promote a floating crust boundary condition in
reduced scale experiments should be encouraged. Several attempts have been made in past
experiments to reduce the risk of anchoring by relying on the ablating nature of the concrete walls, or
to break the crust using a dedicated lance. However, these attempts have been largely unsuccessful.
The use of pure mortar for lateral inclined walls with non-intrusive sensors or heating electrodes could
be a direction to follow in future tests. The community continues to study novel approaches for
achieving a floating crust boundary condition in reduced-scale tests, but these attempts have not been
successful to date.
NEA/CSNI/R(2016)15
6.1.2.2. Water ingression mechanism

The current approach for modelling enhanced crust cooling by water ingression is based on a critical
heat flux concept. The data underlying these models principally comes from SSWICS experiments
(see section 2.2) that were transient core debris cooling tests conducted in an inert crucible without
sustained heating of the core melt. The results indicate that the critical heat flux depends on crust
strength and thermal contraction (which drives crack formation and growth), and the effectiveness of
this mechanism rapidly degrades as concrete content in the melt increases. Thus, water ingression
cooling is more efficient under early cavity flooding conditions. In terms of extrapolating to reactor
scale, the behaviour of this thick cracked crust over a longer duration accident sequence involving
subsequent melt pours on top of this material is unknown.
To address this issue, it is important to first realise one experiment limitation; i.e. the solid crust
(as well as particulate on top of the crust formed by melt eruptions) is not heated, so that power
simulating decay heat is only injected in the liquid phase. This limitation raises the issue of power
control during the tests. Different methods are available for heating core debris, but none of them is
ideal insofar as mocking up heat produced in core debris by fission product decay. The community
continues to investigate heating methods and operational approaches aimed at reducing experiment
distortions related to mocking up fission product decay heat in core debris regardless of the debris
state (e.g. liquid/solid, continuous/fragmented, etc.).
6.1.3. Termination of the ablation phase
The final step of the MCCI process would occur after several days of ablation in which the core-
concrete heat exchange surface becomes so large that the heat can be transferred by conduction to the
remaining concrete without further ablation. This scenario would yield a very viscous melt with high
concrete fraction. Under such conditions, the heat transfer models at the core-concrete interface may
not be valid. Regarding modelling of this long-term behaviour, some simulation tools utilise a quasi-
steady modelling approach in which conduction into the concrete is not modelled. Thus, all heat
transfer from the core debris is dissipated by ablation, and as a consequence, the ablation never stops.
Some of these deficiencies in analytical tools can be addressed with data from longer duration
experiments.
The above situation can be contrasted with dedicated core catcher designs (e.g. EPR or BiMAC)
in which the ablation phase in the spreading room is very short and terminated when the heat load
from the core melt is balanced by the heat removal in the cooling circuit. In these cases, the melt can
remain partially liquid and no more decomposition gases are released from the concrete. Under this
condition, the heat transfer mechanism switches to natural convection. To address such situations,
specific models are needed to allow a continuous description between a bubbling pool, natural
convection, and conduction-limited heat transfer regimes.
Regarding debris formations resulting from top flooding (see Figure 1.3-2 b and Figure 2.4-21), it
is important to verify that they are coolable. To compensate the reduction in volumetric heating in the
experiment resulting from crust formation, the heat flux at the bottom surface of the upper crust may
be increased by increasing input power. However, this same approach would not be viable for
formations resulting from melt eruptions that generally consist of volcanic structures with surrounding
particle beds. The makeup of the particle beds has been found to consist of 5-10 mm diameter particles
that are generally considered to be coolable due to the relatively large particle size. However, the
coolability of volcanic structures has not been assessed in the open literature. Because volumetric
heating of solid material in MCCI experiments is currently not possible, an alternative approach for
assessing the coolability of these structures would need to be undertaken involving analysis possibly
supported by well-designed separate effects experiments.
NEA/CSNI/R(2016)15
6.1.4. Scaling issue between experiment and reactor situation

The initial mass of corium, shape of the test section, and the number and thickness of the concrete
wall(s) used in the experiment are all connected to the issue of scaling test results with respect to time.
When the ablation heat flux evolution is preserved in the experiment, corium enrichment with concrete
slag occurs more rapidly in 2D experiments than in the reactor situation because the ratio between
concrete surface and corium volume ratio is not preserved. For example, in the 2D CCI experiments if
a homogeneous power distribution at the pool interface is considered for a 900 MW PWR the corium,
then the corium is enriched with concrete about 4 times faster in the CCI experiments than in the
prototype (or about 8 times faster in the VULCANO experiments). Conversely, in 1D experiments the
scaling is easier as the corium pool is enriched at a prototypic rate if the initial height of the core melt
in the experiment is preserved.
One way to conduct a shorter duration test that mocks up longer term behaviour is to reduce to a
reasonable extent the height of the corium in 1D experiment, or to increase the core-concrete surface
to volume ratio in the experiment. However, in the latter case note that periodic phenomena such as
melt eruptions or ablation bursts at any given time will have a more significant impact on the
evolution of melt composition. As noted earlier, another way to address long term behaviour is to start
the experiment with a higher fraction of concrete in the corium pool and to adjust the heat fluxes to
prototypic values. This approach is reasonable for addressing questions related to late phase flooding
strategies or to observing the evolution of the melt entrainment rate in a composition range
corresponding to long term scenarios, despite the crust anchoring drawback.
6.2. Realistic plant simulations
Improving the realism in plant simulations inherently introduces more complexity. As a result, these
associated efforts have to be balanced with approaches that rely on invoking additional levels of
conservatism to define a bounding set of hypotheses for safety-relevant issues. Three major topics of
interest in this regard are:
• The presence of metal within the melt or within the concrete;
• The initial conditions for MCCI based on melt pour conditions into the reactor pit;
• The presence of impurities in cooling water (e.g. seawater or brackish water).
Additionally, it is noted in passing that thermal stresses on concrete structures brought on by core
debris interactions with concrete have not been investigated in MCCI Programs. These stresses are
largely inconsequential for below grade reactor cavities but can be quite important for free standing
cavities such as sub-atmospheric containments and especially for reactor pedestals in boiling water
reactors. The core debris interactions place the inner region in compression where concrete is strong
but the outer region in tension where concrete is weak and easily cracks. This has structural
implications which again have not been investigated in MCCI Programs.
6.2.1. The presence of metal within the melt or within the concrete
In the reactor configuration, some metallic elements are always present within the melt. Depending on
the reactor situation and ongoing scenario, they can come from different sources that include the initial
melt pour, as well as sources within the containment that arise through ablation; i.e.
• Metallic melt constituents (primarily U and Zr, but also control rod, blade, channel box, and
canister materials) transferred as part of the melt pour at vessel failure, or during successive
pours;
• Steel coming from the vessel or remaining structures that is heated and melted by the
radiative flux from the corium in the reactor pit before top flooding; and
NEA/CSNI/R(2016)15
• Steel coming from the reinforcement in the concrete basemat and/or sidewalls when the
concrete is ablated, or from the containment liner (where applicable) when contacted by core
melt.
6.2.1.1. Impact of metal on concrete ablation profile
The presence of an immiscible metallic phase within the melt influences the ablation profile as
soon as stratification occurs. This phase is mainly composed of steel as zirconium and uranium
elements are rapidly oxidised. During this short term period, zirconium and uranium affect the density
of the metallic layer and their oxidation produces a significant heat of reaction increasing the initial
ablation rate. Then, or after the last pour from the reactor vessel, the metal phase composition is
predominately steel. In the metal layer, the heat flux to the concrete is higher because: i) the layer is
not diluted by the melted concrete (the metal coming from the reinforcement in the concrete is usually
consumed by oxidation), ii) the heat transfer between oxidic and metallic layers is high due to gas
bubbling (interface dispersion and induced convection), and iii) the high thermal conductivity of the
metal phase increases the heat transfer coefficient to the concrete.
The stratification process is governed by the higher density ratio between metal and oxides as
soon as the fuel oxides become diluted with concrete oxides, as well as the decrease in melt gas
sparging rate that generally occurs during the course of the MCCI. Gas sparging initially promotes
mixing of the metal and oxide phases, whereas the oxidation reactions leads to metal disappearance
and thus a gradual return to a homogeneous configuration.
While several experiments have been performed with iron-alumina thermite simulant, only a few
VULCANO tests have been performed with a prototypic metal-oxide core melt composition. It was
not possible to establish from the VULCANO results clear evidence of stratification but ablation was
observed to be increased in front of the metallic masses 1. Moreover, the results indicate that metals
oxidation is not only driven by gas liberation from the ablated concrete, but also from gas liberated
from the bulk of the concrete material heated by conduction behind the ablation front location. For
prototypic metal-oxide core melt, more generally the oxidation kinetics and the stratification
thresholds are important as they influence the time window when the melt is stratified at reactor scale,
and as a result the prediction of the basemat melt through time.
The initial phase of the interaction involving unoxidised cladding (zirconium) in the melt has
been investigated in a few reactor material experiments. This stage can lead to highly exothermic
metal oxidation reactions. Zirconium-bearing concrete-metal inserts were used in some Argonne
experiments in which a relatively small amount of Zr was incorporated into the melt just prior to melt
contact with the concrete basemat. However, it is likely that a significant fraction of the Zr in the
inserts was oxidised before the test was initiated, thereby limiting the impact of this metal on the
actual MCCI phase of the experiment.
Another aspect not investigated in experiments is the presence of uranium within the metallic
phase. During the in-vessel stage of the accident, uranium is found in the metal phase in scenarios that
lead to a significant fraction of unoxidised cladding in the lower head. The presence of the cladding
results in the formation of a heavy U-Fe-Zr metallic phase at the bottom of the oxide melt in the lower
head. During MCCI, it is likely that this metal layer will stratify near the beginning of the interaction.
Under these conditions it seems appropriate to implement an oxidation model for uranium in
simulation tools and to perform sensitivity analyses.
1. Additionally the distribution of the spatial power in between oxidic and metallic phase remains uncertain in
the VULCANO experiments.
NEA/CSNI/R(2016)15
Reinforcing bars in concrete play a double role as they are a continuous source of metal during
ablation, in addition to changing the ablation mechanism. In particular, recent MOCKA test results
using simulant materials indicate that the presence of reinforcing bars in siliceous concrete leads to a
homogeneous ablation profile, which contradicts the results from reactor material tests carried out with
non-reinforced siliceous concrete in which anisotropic ablation is observed (Foit, Fischer, Journeau, &
Langrock, 2014). One hypothesis is that the rebar increases the mechanical stability of dehydrated
concrete, thereby preventing gross failure and ablation bursts that could cause ablation anisotropy. In
contrast, for limestone common sand concrete the ablation profile in MOCKA tests is anisotropic
(larger sidewall ablation) and not modified by the concrete reinforcement (Foit J. J., 2015). This trend
is also observed in the VBSU-1 reactor material test carried out with non-reinforced limestone
common sand concrete (Foit, Fischer, Journeau, & Langrock, 2014). Running experiments with
reinforcing bars embedded in the concrete means that we first have to avoid short circuiting electrical
heating methods or introducing non-prototypic direct heating issues. In terms of scaling it is important
to consider a test section large enough to respect the size of the aggregate and the spacing between the
rebar and conduct the test over longer duration to reach a steady state 2.
6.2.1.2. Effect of metal on cooling mechanisms
Since the water ingression cooling mechanism depends on crust mechanical properties, the presence of
metallic inclusions in an otherwise oxidic crust could change the properties and thereby impact this
mechanism. Specifically, the presence of metal could influence the critical heat flux associated with
cracks that form in the crust due to thermal contraction induced by top flooding. To address this issue,
additional SSWICCS-like experiments could be performed with different metal contents in the melt
and a representative gas release to promote good mixing conditions. Some tests of this type are already
scheduled in the frame of the MIT3BAR Program (Journeau & Teisseire, 2015). For these tests, as
well as large scale experiments with sustained heating, new thermite compositions need to be
developed that would produce a melt with adequate metal fraction. An alternative would be to add
metal after the initial melting phase of an otherwise oxidic melt. Metal-oxide mixing is important
since, without it, the situation will be very similar to pure oxidic melt conditions. It is not necessary to
address the behaviour of a purely metallic crust because the water ingression mechanism is closely
linked with early flooding and in that case the pool is expected to be well mixed.
The situation for the melt ejection cooling mechanism is different as it is occurs over a longer
duration. The melt could be mixed at the beginning with a priori no major differences, and then after
stratification occurs the ablation rate will increase with a corresponding increase in the gas release
rate. After most of the oxide melt is ejected, the ablation rate would decrease again and the remaining
liquid would tend towards a purely metallic layer with a low viscosity. For a floating crust, a part of
the metal layer could ingress in the upper crust and remain heated It could be interesting even if it is
not of a high priority to evaluate the entrainment rate of pure metal melts and check the morphology of
the particles formed during quenching to assess their coolability as well as their influence on the
coolability of the debris bed in general. As soon as the specific technological challenges of metal-
oxide experiments are resolved, tests with a high metal fraction could be performed in order to reach
conditions in which most of the oxide is included in the upper crust and particle bed regions.
6.2.2. The initial conditions for MCCI or the pouring phases from the vessel to the reactor pit
This section provides a discussion of the expected MCCI initial conditions for realistic reactor
scenarios. For the purposes of simplification, it is often assumed that the MCCI phase starts as soon as
2. This is a challenge for the MOCKA experiments where the decay heat is simulated by adding chemically
energetic mixtures to the melt.
NEA/CSNI/R(2016)15
the vessel fails and the corium mass in the lower head (which in bounding analyses includes the entire
fuel and structural inventory in the reactor) is relocated into the reactor pit. This approach offers a
degree of conservatism in terms of axial-melt-through delay if one assumes that the melt is spread
instantly over the entire surface of a dry pit. However, when the pit is flooded spreading may be
limited, leading to corium accumulation in one part of the pit. This will result in higher heat fluxes to
concrete and reduce the basemat melt through time if this accumulation is stable and does not
eventually spread out uniformly. Among other things, this situation depends on the corium
temperature, pour rate, failure location, amount of water present at vessel failure, and finally on the
ability to provide water continuously on top of the corium accumulation.
Such configurations are quite complex to study because they involve the formation and spreading
of corium accumulations under water as well as the possibility of boiling off the water inventory,
drying out the core debris, re-melting, and onset of concrete ablation. An ancillary issue is that core
debris in a reactor cavity, if not covered by water, exposes a great deal of concrete surface area to
intense convective and radiative heat flux. The gas generation and concrete degradation from this
exposed concrete cannot be neglected in the analysis of core concrete interactions and containment
integrity.
Depending upon the melt pour conditions and with a relatively shallow water layer, melt jet
fragmentation is expected to be minor. For this type of scenario, existing MCCI models that treat the
corium as an initially intact melt pool interacting with concrete may be employed as a reasonable
approximation. However, for deeper water pools melt jet fragmentation may be significant, leading to
formation of a coherent particle bed, or a compact melt layer commonly referred to as a cake
surrounded by particle bed (see FARO (Tromm, Foit, & Magallon, 2000). Depending upon the bed
depth, decay heat level, particle size, and porosity, the configuration may be coolable. However, if the
dry-out limit for the bed is too low then gradual reheating, dry-out, melting, and onset of concrete
ablation will occur. These particular configurations have not been extensively investigated as part of
MCCI research, nor can existing MCCI models address this type of behaviour. However, there has
been a significant amount of research done in the area (both experiments and modelling) that
generically addresses debris bed coolability for both in-vessel and ex-vessel applications. Obviously,
conducting experiments that involve dry-out and melting of particle beds composed of reactor
materials is a technical challenge given limitations with current core debris heating techniques (see
above discussions). Thus, a possible first step to address this issue is to utilise existing models to
evaluate coolability of particle bed formations predicted for plant applications. If these analyses
indicate that the beds are likely to be uncoolable, then effort should be devoted to developing
appropriate experiment techniques to address this type of behaviour.
Another related issue is that of multiple pours and how that affects the coolability of debris in the
reactor pit. Again, in all experimental and analytical studies concerning MCCI, it is traditionally
assumed that at vessel failure, the molten material (whether the entire reactor inventory or partial
inventory) is ejected all at once and spreads on the reactor pit surface. It is likely that in some accident
scenarios, the melt pour would be periodic which has two consequences: non-uniform melt
accumulation and non-uniform spreading. Conducting an experiment with this kind of melt
configuration may be quite challenging, and an analytical extrapolation of experimental data for
symmetric and uniform melt configuration may be more worthwhile based on simulant data.
6.2.3. Presence of impurities in cooling water
The impact of impurities in cooling water on severe accident behaviour resurfaced following the
Fukushima accident. In particular, the use of sea water brought into question the effect of salt (sodium
chloride) on coolability mechanisms, the impact on the chemistry of fission products, and the
performance (i.e. potential for clogging) of coolant loops. Generally speaking, any impurity in cooling
water (whether it is salt in sea water or other forms of impurities in brackish inland fresh water) can
impact one or more of these areas.
NEA/CSNI/R(2016)15
For the cooling mechanisms identified under top flooding conditions for a corium pool
interacting with concrete, the formation of precipitate in the cracks of the upper crust or in the
overlying debris bed could reduce the dry-out limits for these formations. As the composition of water
present in the sumps at the bottom of the reactor building is complex and can depend on the accident
management strategy, it seems easier to address the issue in separate effect tests than in semi-integral
experiments. Ongoing experiments in Japan are addressing some aspects of the water impurity issue
(see Appendix 7.3). To parametrically investigate the effect of water impurities on coolability by water
ingression, SSWICS-like tests could be run to evaluate the impact on the cracks formation and on the
crust critical heat flux. If warranted, more complex experiments (i.e. with sustained heating) could be
conducted to assess the behaviour over the long term.
For the melt ejection mechanism, the influence of impurities on debris bed coolability could be
investigated in separate effects tests that utilise existing facilities that are studying dry-out in debris
beds for in vessel conditions in order to assess the long term coolability.
The water at the bottom of the containment building will be highly contaminated with fission
products. If this water is used to cool the melt, the chemistry of the fission products will likely be
modified by gas bubbling and more generally by particulate entrained in the water. While the fission
product behaviour under such conditions is an ancillary issue related to the consequences of clogging,
water samples could be collected quite easily at the end of MCCI experiments to perform chemical
analysis in order to characterise the chemical composition. If some impurities in the water can play a
role in trapping other species released during MCCI, it would be useful to carefully select the initial
composition of the water before running these tests.
6.3. Coolability enhancement under top flooding conditions
The SOAR is focused on ex-vessel coolability under the top flooding which is largely regarded as a
generic accident management strategy for ex-vessel melt stabilisation in existing plants.
The improvement of melt coolability under top flooding conditions can also be viewed as a
potential back-fitting strategy for operating reactors. Moreover, for new reactor designs spreading and
top flooding can be incorporated in the design phase as a generic approach.
In this regard, the first point is that a larger initial corium spreading area will reduce the
downward heat flux to the concrete. This approach can be used for a new design if the basemat is thick
enough. One approach for increasing spreading area for plants with limited floor space is to allow
radial melt-through of a barrier with subsequent spreading of a portion of the melt into the reactor
building. This situation is more likely for siliceous concrete but remains limited only to the level of
corium above the breach elevation. Referring to the presentation in Chapter 5 about the Fessenheim
plant in France, modifications like an unplugged, dedicated hole in the wall of the reactor pit can
promote additional spreading in a dedicated dry room. This spreading will be easier and faster in a dry
cavity situation, one that also provides the benefit of eliminating the risk of steam explosion.
The coolability of debris can be more efficient if spreading is combined with water flooding. In
the EPR and BiMAC core catcher designs, water cooling circuits located at the bottom are provided to
achieve this objective. For existing reactors, coolability is achieved mainly by top flooding as part of
the accident management strategy. Two methods can be used to improve the coolability under these
conditions:
• Early flooding to capitalise on the fact that the water ingression mechanism is most efficient
at this stage with little concrete present in the melt, and that the melt eruption mechanism is
also most effective in the early phase of the corium-concrete interaction due to the higher
melt gas sparging rate;
NEA/CSNI/R(2016)15
• High gas content in the concrete to maximise the impact of the melt ejection cooling
mechanism over the long term and accelerate the oxidation of the metallic phase.
The first point is closely connected to the SAM strategy. Ideally, it is desirable to have an initially
dry pit to maximise spreading and to avoid the risk of a steam explosion, followed by early flooding.
In this case, the time window to add water is narrow and a subsequent melt pour after top flooding
cannot be excluded.
The second point depends on the composition of the concrete that is used for the reactor basemat
and so on the plant location. If a recommendation is made to consider high carbonate and/or hydrate
contents for the concrete of new reactor basemat, a back-fitting measure for plant having a potentially
too thin siliceous concrete basemat could be to consider pouring an additional (sacrificial) layer of
such a concrete. In this case, since the thickness of this additional layer is obviously limited, the key
piece of information needed is the efficiency of the melt ejection mechanism so as to ensure that all
the liquid portion of the melt is transformed into a coolable debris bed before reaching the original
siliceous concrete.
In the coming years the examination of the debris in the three damaged Fukushima reactors will
likely provide additional insights that will enhance the understanding of MCCI phenomena at large
scale and under fully prototypic conditions. The findings will undoubtedly provide additional
confidence in the application of simulation tools to existing plants. They will also aid in optimising
SAM strategies for existing as well as future plants. In the meantime, in order to perform experiments
and additional analysis to address more realistic situations, it will be necessary to improve the
capabilities of existing facilities to overcome the technological challenges that have been described in
this chapter.
One of the top level recommendations in the NEA-SAREF report (in preparation) is to organise
an MCCI workshop to discuss current state of MCCI knowledge, identify knowledge gaps, and
identify data needs to bridge the gaps – the idea being that the Fukushima decommissioning effort can
be informed by the outcome of such a workshop while at the same time, data collected during the
decommissioning activities can be optimised to bridge the MCCI knowledge gaps. In two companion
studies (one on severe accident knowledge gaps post-Fukushima and the other on Fukushima forensic
data needs), MCCI knowledge and data gaps were identified as high priority topics. These findings
confirm that in order to perform experiments and additional analysis to address more realistic
situations, it is necessary to improve the capabilities of existing facilities and to perform needed
experiments to bridge the knowledge gaps and reduce residual uncertainties. Since experimental
MCCI research with prototypic reactor materials is an expensive undertaking, a collaborative effort
among various nuclear safety research organisations in different countries is highly recommended.
NEA/CSNI/R(2016)15
7. Appendices
7.1. Appendix on detailed code descriptions
This appendix provided a detailed description of most of the codes presented in the Chapter 3.
7.1.1. COCO code
7.1.1.1. Pool interface models
7.1.1.1.1. Pool/concrete interface model

Crust and slag (molten concrete components) are considered between concrete and melt as shown in
Figure 7.1-1.
T Heat Conduction Heat Transfer by Convection
Tmelt
Tconcrete
melting Tinterface
Tinitial
Boundary
Concrete Slag Crust Melt
Layer
Figure 7.1-1: Model of melt and concrete interface

The crust is a mixture of corium and calcined concrete components and it includes not only solid
but also fluid that does not move due to high viscosity. The temperature of crust surface (interface
between melt and crust) is determined so that the solid volume fraction of the melt at the interface will
become a prescribed value, which is input data of COCO and usually given as 0.74 (density in close
packing of equal spheres).
The growth of the crust is calculated by the balance of heat transfer from the melt and heat
conduction to ablation front through crust and slag layer. The ablation rate is also calculated by the
balance of the heat flux from slag layer and heat conduction in concrete (the temperature at the
ablation front is given as the melting point of concrete).
7.1.1.1.2. Pool upper interface model

The interface model between top crust and melt pool is the same as that of the crust on concrete. The
temperature of crust surface is determined so that the solid volume fraction of the melt at the interface
is equal to a prescribed value.)
7.1.1.2. Thermal-hydraulics models
7.1.1.2.1. Heat convection within a pool layer

The convective heat transfer rates from melt to surrounding boundaries are calculated with the
following correlation:
NEA/CSNI/R(2016)15
 nt ( m ) nb( m )

m  g β ∆T  m  g α 
3 3
Nu = a c + at 
m m
Pr 
 + a b

 ν2 Pr 
  Equation 7.1-1
 ν
2
    
Where,
Nu : Nusselt number, a c , at , ab , nt , nb : constants, g : gravitational acceleration,  : characteristic

length, β : Thermal coefficient of volumetric expansion, Pr : Prandtl number, ν : Kinematic viscosity
of melt, α : Void fraction in melt, ∆T : Temperature difference (melt temperature – boundary
temperature)
Superscript m : direction (basemat, sidewall or top crust)
The first term comes from heat conduction effect in melt, the second term expresses the heat
transfer effect due to the thermal natural convection and the third term is the agitation effect by gas
bubbles. The constants a c , at , ab , nt , nb are input data to the COCO code and determined by the
calculation by CFD code including the natural convection model and gas-liquid two-phase flow
model. The heat transfer coefficient is not zero even in the limit of ΔT  0 or α  0.. Indeed the
correlation is expressed by the addition of three independent terms.
7.1.1.2.2. Heat convection between oxide and metal layers in a stratified pool
The heat transfer coefficient between oxide and metal layer is assumed to be so large that oxide and
metal temperatures are the same.
7.1.1.2.3. Stratification criteria

The choice of stratification model is decided by the user. When the ratio of oxide to metal density is
small, the stratification model will be used. For example, an experiment using molten stainless steel
pouring on to concrete cavity showed that the calculation with the stratification option agreed with the
experimental results, because the density of molten concrete is far less than molten stainless steel.
The viscosity of melt liquid phase is calculated by Kendell-Monroe (Kendell & Monroe, 1971), Shaw
correlation (Shaw, 1972) or a correlation derived for basaltic concrete
The melt, which is a mixture of corium and concrete, is a two phase fluid of liquid and solid. In the
COCO code, the solid fraction of the melt is determined according to a phase diagram
(see Figure 7.1-2).
Solidus Curve Liquidus Curve
3000
2800
0.075
2600
Temperature(K)
0.05
2400
Mass fraction of solid
2200 =0.025
0.95
2000
1800 0.975
1600
1400
1200
0.0 0.2 0.4 0.6 0.8 1.0
Concrete mass fraction(-)
Figure 7.1-2: An example of phase diagram in COCO code
NEA/CSNI/R(2016)15
The phase diagram data are calculated with thermodynamic database (Fukasawa, Tamura, &
Hasebe, 2005). One example of the phase diagram for corium and siliceous concrete system is shown
in Figure 7.1-2.
The calculated results of solid fraction and corium fraction in solid at various concrete contents
and temperatures are tabulated and COCO utilises the table data to determine the solid fraction in
melts and the effective viscosity of the melt. The effective viscosity of the liquid-solid mixture μ is
determined by Stedman’s correlation (Stedman, Evans, & Woodthorpe, 1990) with the volume
fraction of solid Φ:
1+ φ / 2
µ = µm
(1 − φ )4 Equation 7.1-2
where μm is the viscosity of the liquid.

7.1.1.5. Coolability models
When there is no water at the top of the melt, heat transport by radiation from surface of melt or crust
top surface is considered. The view factors from melt surface to upper sidewall and ceiling are
calculated and all surfaces are assumed to be grey body. The temperature of the ceiling is a boundary
condition and given by input.
When there is water above the top of the melt, cooling effects of water ingression and melt
eruption are considered. One-dimensional computation grid is created in the crust (number of mesh
increases when the crust grows) and the composition and temperature are evaluated at each mesh
point. The composition of the crust is decided depending on the melt composition when the crust is
formed. The dry-out heat flux at each mesh point is calculated by a correlation derived by SSWICS
test in OECD/MCCI Project (Lomperski & Farmer, 2007) and the maximum cooling rate is limited by
the dry-out heat flux at each mesh point.
The melt eruption rate is calculated by the following equation,
jm = K e j g
Equation 7.1-3
where jg is the superficial velocity, subscripts m and g denote the melt and gas, and K e denotes the
entrainment coefficient, which is calculated by the following correlation.
K e = E (ρ g / ρ m )2
1
Equation 7.1-4
where ρ is a density
The proportionality constant E is an input data.
The debris erupted above the crust is assumed to be sufficiently cooled by water.
7.1.1.6. Code simplifications and limitations
• In COCO, axisymmetric cavity is assumed. Appropriate modelling is required for the
analysis of non- axisymmetric cavity, e.g. PCV floor drain sump pit of BWR.
• Additional melt pouring is not considered.
• Calculation of melt spread is out of scope.
NEA/CSNI/R(2016)15
• The mass released by melt eruption is excluded from the calculation. Because, the erupted
melt is assumed to be cooled by covering water in the form of debris bed.
• The calculation is terminated when all the melt is solidified.
• When the melt is cooled and the solid fraction increases, the melt shows Bingham plastic
behaviour. In COCO, the melt is assumed to be Newtonian fluid; therefore the simulation is
not exact for crust formation near the ablation front and for the heat transfer rate in the high
solid fraction condition of melt.
• In COCO, the exact correlation of boiling curve in not used in the calculation of rewetting
timing after water flooding. Instead, rewetting is determined when the heat flux at the top of
the crust falls below the dry-out heat flux of the crust. On the other hand, the surface of the
crust will be uneven and there will be cracks and voids which reduce the heat conduction in
the crust, therefore the temperature and heat flux at the crust surface will not be uniform and
local quenching will occur at low temperature point and the rewetting will spread all over the
crust surface from this point. Therefore, the rewetting heat flux in COCO is assumed equal to
crust dry-out heat flux.
7.1.2. CORCON code
Heat is removed at the boundaries of the pool, which are its top surface and its interface with concrete.
The internal temperature of the pool adjusts quickly so that these heat losses balance the internal heat
generation, and the heat transfer approaches a steady state. The heat transfer model allows for several
possible configurations in each layer: the layer may be completely molten, it may have a solid crust on
one or more surfaces, or it may be completely solid.
7.1.2.1.1. Pool/concrete interface models

Heat loss from the top of the molten debris is dominated by radiation to containment structures or to
the overlying water. Because of the fourth-power temperature dependence of the radiative flux, this
loss is rather insensitive to containment temperatures (unless they are very high). The decrease in
radiative heat transfer from the pool surface to the surroundings, associated with atmospheric
attenuation by aerosols, is approximately accounted for. Pool/concrete interface model
Heat transfer between molten core debris and reactor cavity concrete is controlled by the
bubbling of concrete decomposition gases through the melt. This process is similar to nucleate boiling
or gas sparging except that at the interface between the core debris and the concrete, gas is being
released coincident with melting of the concrete surface. Coincident with gas bubbling and concrete
melting at the interface, the molten core debris may begin to solidify as a crust adjacent to the melting
concrete surface. This crust may be stable or unstable depending on its growth rate, its strength
properties, and the disruptive forces acting to destabilise it.
CORCON has two models for interfacial heat transfer: gas film model and slag film model. At
extremely high gas generation rates, it may be possible to form a stable gas film at the melt-concrete
interface. When a stable film is present, heat transfer across the film is by combined radiation and
convection. For a stable gas film on a nearly horizontal surface, heat transfer is computed from a
mechanistic model based on momentum balance in a Taylor-instability bubbling cell. The result may
be cast in the form of a Nusselt number based on film thickness. For an inclined surface, a flowing
film model is used and both laminar and turbulent films are considered. CORCON-Mod3 employs a
simple transition between the laminar and the turbulent flow regimes to ensure continuity of film
thickness and heat-transfer coefficient with the appropriate limits. In all cases, Nusselt-type
correlations are used.
NEA/CSNI/R(2016)15
The simplified model for the concrete response in CORCON-Mod3 is based on a steady-state,
one-dimensional energy balance. For a steady temperature profile in the concrete, a simple heat
balance at the concrete surface provides information on concrete erosion (ablation) in terms of net heat
flux to concrete and its ablation enthalpy. The steady-state model calculates the generation of
decomposition gases. The ablation enthalpy for concrete is calculated internally, and consists of both
sensible and chemical energies. The sensible energy includes the energy necessary to raise gaseous
decomposition products to the concrete ablation temperature. The chemical energy is included using
experimentally determined heats of decomposition for three reactions: evaporation of free water,
release of chemically bound water from hydroxides, and release of carbon dioxide from carbonates.
The ablation temperature of concrete is not precisely defined because ablated material may not be
completely molten. In CORCON-Mod3, a melting range is defined by the concrete liquidus and
solidus temperatures, with the ablation temperature ordinarily chosen by the user to lie between them.
The choice affects the calculated heat of ablation. If the concrete contains reinforcing steel, the energy
necessary to raise it to the concrete ablation temperatures is included in the concrete ablation enthalpy.

Heat loss from the pool surface includes convective heat transfer to the atmosphere and radiative heat
transfer to the surroundings. Thermal radiation is the dominant mechanism. If desired, the radiative
effects of aerosols in the atmosphere may be included in the calculation, with an atmospheric opacity
determined from calculated aerosol concentrations. Once the optical thickness of the atmosphere is
known, the one-dimensional diffusion equation is applied for infinite, parallel, and optically grey
plates. Convection produces additional heat transfer from the pool surface. Unless the atmosphere is
truly transparent, however, convection and radiation are strongly coupled; the radiation tends to
increase thermal stability and reduce convection. In most cases, convection is found to be small so
only a simple convection model is included in CORCON-Mod3.
If water is present, it will form an additional layer at the top of the pool. This is likely to cool the
top of the melt below the solidification temperature, resulting in a solid crust on the surface. The crust
will progressively fragment and allow water ingression until the core debris is completely quenched.
An overlying coolant pool will also trap aerosols generated during the core-concrete
interaction.CORCON-Mod3 treats heat transfer to the coolant using pool boiling correlations. It does
not allow for either steam explosions or the progressive quenching of a molten pool to a coolable
debris bed.
7.1.2.1.3. Crust formation and freezing

CORCON-Mod3 has a relatively simple quasi-steady-state model for crust formation and freezing.
The model is formulated in terms of the average temperature of the layer, which is known from its
mass and energy content. The average temperature and the boundary heat fluxes at this steady state are
determined by the internal heating and the boundary temperatures for the layer. As a further
simplification, the problem is reduced to two independent one-dimensional problems, one axial and
one radial, by performing radial and axial averages, respectively, of the full two-dimensional
problems.
Within a one-dimensional calculation, a layer may be entirely liquid, entirely solid, or liquid with
a solid crust. For the axial case, a crust may exist on the top, on the bottom, or both. In liquid regions,
heat transfer is by convection (natural or bubble-enhanced) with a conduction limit. In solid regions, it
is by conduction. In the case of a liquid with crusts, the liquid layer is solved first using assumed
values of its average temperature and thickness or radius. For any surface at which a crust exists, the
boundary temperature is assumed to be the solidification temperature.
NEA/CSNI/R(2016)15
7.1.2.2. Pool thermalhydraulics

The convective bulk pool heat transfer model in CORCON-Mod3 has evolved from those used in
CORCON-Mod1, and CORCON-Mod2. For the bottom interface of the melt pool, where gas bubbles
may be injected from the incoming concrete, the heat transfer coefficient for a liquid layer is
calculated using the Kutateladze correlation (Kutateladze & Malenkov, 1978). For many fluids, the
transition velocity calculated using the Kutateladze correlation is comparable to the velocity for
transition from bubbly to chum-turbulent flow. For the upper interface of the uppermost melt layer, the
Kutateladze correlation is simply multiplied by an area enhancement factor derived by Farmer.
For heat transfer between liquid layers within the melt pool, Greene’s correlation (Greene &
Irvine, Heat transfer between stratified immiscible liquid layers driven by gas bubbling across the
interface, 1988) is used to calculate the heat transfer coefficient in each layer.
At sufficiently low gas velocities, heat transfer in molten debris is dominated by natural
convection. This process is modelled in CORCON-Mod3 by conventional Nusselt-Rayleigh
correlations.
CORCON models the melt pool as consisting of a number of layers contained in a concrete cavity.
These layers are, from bottom up, a heavy (i.e. dense) oxide phase (HOX), a heterogeneous mixture of
heavy oxides and metals (HMX), a metallic phase (MET), a heterogeneous mixture of metals and light
oxides (LMX), a light oxide phase (LOX), a coolant (CLN), and the atmosphere (ATM). Mixing,
stratification, and density changes resulting from material addition can lead to changes in the layer
orientation during a calculation.
For a multi-layered pool, CORCON-Mod3 considers heat transfer one layer at a time. Given a
trial set of interfacial temperatures, a solution is found for each layer. Newton’s iteration is then used
to revise the interfacial temperatures to satisfy the requirement that the heat flux must be continuous at
all interfaces between layers. The heat transfer model allows for several possible configurations in
each layer: the layer may be completely molten, it may have a solid crust on one or more surfaces, or it
may be completely solid. For the bottom interface of the melt pool, where gas bubbles may be injected
from the incoming concrete, the heat transfer coefficient for a liquid layer is calculated using the
correlation devised by Kutateladze.
As time progresses, the debris pool grows; its surface area increases, and decay heating
decreases. Therefore, pool temperatures and heat fluxes decrease, and the possibility of refreezing
arises. Substantial freezing of the metallic phase may occur. However, the large internal heating and
small thermal conductivity of the oxidic phase prevent the existence of steady crusts more than a few
centimetres thick. The bulk of this phase will remain liquid, probably for weeks.

Experimental evidence (Bradley, Gardner, Brockmann, & Griffith, 1993) shows that the various oxidic
species in the melt are highly miscible, as are the metallic species, but that the two groups are mutually
immiscible. In the absence of gas bubbling, the core debris will stratify into distinct layers based on
their relative densities. Mixing of the immiscible layers can occur at high gas fluxes or when the
densities of the layers are close. During a core-concrete interaction, there may be times in which the
molten core debris is mixed and other times in which the debris is stratified.
In CORCON-Mod3, the user has the option of selecting whether to begin a calculation with a
stratified or a fully mixed debris pool, and the user can select whether the code will perform the
entrainment and de-entrainment calculations. This allows the user the flexibility to begin a calculation
NEA/CSNI/R(2016)15
in a fully mixed configuration and then allow the code to calculate entrainment and de-entrainment, or
the user can force the debris pool to remain mixed by bypassing the mixing calculation. Similarly, the
user can begin a calculation with a stratified debris pool and then allow the code to calculate
entrainment and de-entrainment, or the user can force the pool to remain stratified by bypassing the
mixing calculation. Mixture layers stratify into distinct metal and oxide layers if density differences
become greater or the gas flow through the melt decreases.
7.1.2.3. Thermophysical and thermo-chemical properties
Calculation of the physical processes described in the preceding sections requires values for a wide
range of thermodynamic and transport properties. These properties are treated as functions of
composition and of temperature. However, the dependence is not always explicitly included in the
CORCON-Mod3 model; for example, the values used for surface tensions are independent of
temperature. In most cases, properties are required for the mixtures of species which make up the
components of the CORCON system, although the calculation of chemical equilibrium requires the
chemical properties of the individual species. With a few exceptions, such as the viscosity of oxide
mixtures with large silica contents, mixture properties are calculated from those of the constituent
species. A user option for specifying the phase diagrams of the metallic and oxidic phases is provided
in CORCON-Mod3. Also, properties used by the CORCON-Mod3 implementation of the VANESA
(Powers, Brockmann, & Shiver, 1986) model are included.
7.1.2.3.1. Thermodynamic properties

The thermodynamic properties calculated are density, specific heat, enthalpy, and chemical potential.
Mixture densities are computed from the molar volumes of the individual species. The temperature
range of the data from which density-molar volume relationship is generated varies considerably for
different species. For many of the oxides, it is from 1 200°C to 1 800°C, while for others it covers the
entire range from melting to boiling. The melting range, defined by the liquidus and solidus
temperatures, is prescribed by external models for mixtures of condensed species.
The specific heat of any species, condensed or gaseous, is represented in the form of a
polynomial function of temperature. A single range of temperature is used for all gaseous species, and
the fits are valid from 25°C to approximately 5 700°C. The fits for condensed species include both the
liquid and one or more solid phases. Mixture specific heats are computed by mass averaging of
component specific heats. Specific enthalpies are computed from integrals of the corresponding
specific heats. As with the specific heat, the enthalpy of a mixture is computed by mass averaging.
Chemical potentials for the species are required in the calculation of chemical equilibrium and are
computed from the molar Gibbs function. The VANESA model tabulates and uses free energy data in
the form of free energy functions which are related to the Gibbs function and enthalpy.
7.1.2.3.2. Transport properties

The transport properties computed in CORCON-Mod3 are the dynamic viscosity, the thermal
conductivity, the surface tension, and the emissivity. Detailed models are included for condensed
phase species and mixtures only. Gas-phase viscosity and thermal conductivity (required for
calculation of heat-transfer coefficients at the melt/concrete interface) are treated as constants using
representative values. The viscosity of molten oxides is quite complex, particularly, when significant
amounts of silica are present. For low-silica mixtures, the viscosity is computed from the Kendell-
Monroe expression (Kendell & Monroe, 1971). The viscosities of the species are determined using an
Andrade equation. For mixtures with higher silica content, the viscosity can be greatly increased by
the formation of strongly bonded chains of SiO4 tetrahedral. The viscosity is calculated from a model
proposed by Shaw. This was originally generated as a fit to the correlation developed by Bottinga and
Weill (Bottinga & Weill, 1972). The viscosity of the metallic phase is assumed to be represented by
NEA/CSNI/R(2016)15
the viscosity of iron (the major constituent). The coolant viscosity is computed by the standard
formula. CORCON-Mod3 contains a model for the enhancement of viscosity by suspended solids
Values for thermal conductivity for condensed phase species do not include any temperature
dependence. Mixture values are computed from the species values by mole-fraction averaging.
CORCON uses the same thermal conductivity for both the solid and liquid. This is a fairly reasonable
assumption for the oxide phase. The code has the provision for specifying a multiplier to be applied in
the calculation of the metal phase thermal conductivity. This modified thermal conductivity applies to
both the solid and liquid phases. Like thermal conductivity, values of surface tension for condensed
phase species do not include any temperature dependence. Mixture values are computed from the
species values by mole-fraction averaging.
The calculation of radiative heat transfer requires emissivities for the radiating surfaces. In the
CORCON-Mod3 code, only the emissivity of water (coolant) is stored as internal data. Values are
input by the user for the ablating concrete surface, for the oxidic and metallic melt phases, and for the
surroundings above the pool. The first is specified as a constant, while the last three may be input as
functions of either surface temperature or time.
7.1.2.3.3. Chemical reactions

In CORCON-Mod3, both the reactions of metals with gases from the concrete and the condensed
phase reactions between oxides and metals are modelled. Condensed phase reactions are particularly
important for core debris interactions with high silica, low gas concretes. CORCON-Mod3 assumes
that chemical equilibrium is achieved between the reactants during each time step. The chemical
equilibrium solver minimises the Gibbs free energy function subject to constraints on mass
conservation and on non-negativity of concentrations. The condensed phase reactants and oxidic
products are treated as ideal solutions. It should be noted that non-ideal solution chemistry has been
included in the vapourisation release model in VANESA. CORCON-Mod3 contains coding to
calculate the reduction of oxides at the pool surface by the oxygen-poor atmosphere above the melt.
If a coolant layer is present, CORCON-Mod3 calculates heat transfer to coolant using standard pool
boiling correlations. Corrections are made for the effects of gas injection at the melt-coolant interface
and coolant subcooling. Nucleate boiling is treated using the Rohsenow (Rohsenow, 1952) correlation
for the temperature rise and the Zuber correlation (Zuber, 1958)(with Rohsenow’s coefficient) for the
critical heat flux. The effect of subcooling on nucleate boiling is included, using the expression
recommended by Ivey (Ivey, 1962). The film boiling regime is based on the Berenson correlation
(Berenson, 1961) for the heat transfer coefficient in film boiling and for the temperature difference at
the Leidenfrost temperature (minimum film boiling point). Above the Leidenfrost point, the total heat
flux includes a radiation heat flux component and a convective heat flux component. The radiative
contribution is given for infinite parallel grey walls.
CORCON-Mod3 includes the effects of gas sparging. Both gas sparging (i.e. non-condensable
gas injection at the interface) and coolant subcooling can greatly increase the film boiling heat flux,
while also increasing the temperature at which the vapour film collapses (the Leidenfrost point). Gas
sparging increases film boiling heat transfer by increasing agitation of the coolant and of the melt
surface. In CORCON-Mod3, the enhancement to the film boiling heat flux due to gas sparging is
included as a multiplicative factor. The factor used depends on whether the surface underlying the
coolant is solid or liquid. If the surface underlying the coolant is liquid, then the enhancement factor is
calculated using a correlation of experimental results advanced by Greene (Greene, 1991). If the
surface underlying the coolant is solid, then the enhancement factor is calculated using a correlation
advanced by Duignan (Duignan & Greene, 1989). When the temperature of the core debris is
NEA/CSNI/R(2016)15
calculated to lie between the solidus and liquidus temperatures of the debris mixture, the two
enhancement factors above are weighted by the surface solid fraction.
Subcooling of the overlying coolant pool can also enhance heat transfer in the film boiling
regime. When the overlying coolant pool is subcooled, energy is removed from the gas film by the
overlying subcooled coolant. The net effect of this cooling is a reduction in the thickness of the vapour
film. The reduced film thickness permits greater heat transfer by conduction. The enhancement to heat
transfer owing to coolant subcooling in the film boiling regime is given by a multiplicative factor,
which is calculated using an equation of the form proposed by Siviour and Ede (Siviour & Ede, 1970),
and Dhir and Purohit (Dhir & Purohit, 1978). The effect of coolant subcooling on the minimum film
boiling temperature is calculated using a simple linear correlation of experimental data.
CORCON-Mod3 currently has no water ingression or melt eruption models. It is planned to
incorporate these models from CORQUENCH (Farmer, 2010) in the future.
7.1.2.5. Melt spreading
Mechanistic modelling of the spreading process is not attempted CORCON-Mod3. Instead,
CORCON-Mod3 allows the user to specify a time-dependent radius of the melt that is less than the
dimension of the confining cavity. This approach allows considerable flexibility in mimicking the
spreading process, while also rendering the possibility of specifying physically unreasonable melt
configurations. The option, as currently implemented, will adjust the melt radius to keep the melt
thickness between the maximum and minimum thicknesses specified by the user. Coolant may be
present in the cavity when the time-dependent melt radius option is invoked.
The CORCON-Mod3 code has some limitations which are listed below.
1) For models in dry conditions:
• The calculation of radiative heat loss from the pool surface is based on a one-dimensional
model.
• The convective loss from the pool surface is calculated using a constant heat transfer
coefficient.
• The calculated concrete response is based on one-dimensional steady-state ablation, with no
consideration given to conduction into the concrete or to decomposition in advance of the
ablation front.
• The solidification model assumes that a crust forms on any surface whose temperature falls
below the solidification temperature. The mechanical stability of the crusts is not considered.
• The code assumes that the crust has the same properties as the bulk liquid phase. This may
not be true if the liquid phase composition is changing with time.
• The gas-film model is used for radial heat transfer even after the melt solidifies. As a result,
no radial gap develops around a layer of the melt which has completely solidified. The
model assumes no radial gap around a layer of solidified melt for the purpose of heat transfer
calculations though, in reality, a radial gap can form. These modelling assumptions affect the
calculated shape of the cavity.
• The code assumes ideal chemistry for bulk phase chemical reactions.
• The code uses the Fe-Cr-Ni phase diagram for the metal phase that neglects important
metallic components such as Zr, Si, or Al that may be present in the melt at various times
during core-concrete interactions.
NEA/CSNI/R(2016)15
• The time-dependent melt radius model allows the user to mimic the spreading of a melt
across a horizontal surface, but it is not a mechanistic model of spreading.
2) For models in wet conditions:
• The code uses flat plate pool boiling correlations to model heat transfer to an overlying
coolant pool. Generally, the code predicts lower early heat fluxes than in the experiments
and long-term heat transfer by film boiling.
• The code currently has no water ingression or melt eruption models.
7.1.3. CORIUM2D CODE

This section is intentionally left blank.


Model assumption
The code simulation of heat transport phenomena within a molten pool is mainly based on Fieg's
(Fieg, 1978) and Kulacki-Goldstein's (Kulacki & Goldstein, 1972) experimental observations on
internally heated liquids. The models proposed by these researchers give some correlations which
allow determining the heat transfer from the pool to the surroundings. The pool is schematised as
having three types of surfaces (top, side and bottom). A characteristic heat transfer correlation is
proposed for each surface.
The Rayleigh number is defined as:
Ra = Gr ⋅ Pr Equation 7.1-5
where Gr is the Grashof number:
H 3 g ∆T β ρ2
Gr = Equation 7.1-6
µ2
and Pr is the Prandtl number:
c pµ
Pr = Equation 7.1-7
k
where:
H the pool height [m],
g the gravity acceleration (= 9.81 m/s2),
∆T the temperature difference between the centre and the periphery of the pool [K],
β the coefficient of thermal expansion [1/K],
NEA/CSNI/R(2016)15
ρ the density of the molten material [kg/m3],
µ the dynamic viscosity [Pa·s],

cp the isobaric specific heat [J/kg/K] and
k the conductivity [W/m/K].
As a matter of fact, the modified Rayleigh number, Raint, introduced by Fieg and commonly used
for internally-heated liquid pools in most literature, represents a simpler way to calculate the Rayleigh
number, starting from the specific heat source instead of the actual ∆T:
gβ
Ra int = S H 5Equation 7.1-8
ηα k
where:
µ
η= is the kinematic viscosity [m²/s];
ρ
k
α= is the thermal diffusivity [m²/s].
ρ cp
However, this way to account the Rayleigh number implies two important assumptions:
• the bulk-wall ∆T is a function of the internally generated heat flux exiting the liquid pool at
thermal equilibrium;
• the bulk-wall ∆T is due to a purely conductive heat transfer in a uniform heat-generating
medium, with no account for any liquid convection.
With these restrictions, ∆T can be approximated as:
H 2S
∆T = Equation 7.1-9
k
While this way to account ∆T could be a reasonable approximation in the case of a lab-scale
enclosure at low specific heat sources, it evidently leads to overestimate the Rayleigh number if the
liquid convection is effective. In the case of corium pools at reactor scale, unreasonable liquid bulk
temperatures would result. Thus, the code uses a realistic way to calculate the Rayleigh number for
internally-heated liquid pools, taking into account the actual bulk-wall ∆T.
In a stationary regime, it is:
Q H
≅ S ⋅ Equation 7.1-10
A 2
where S is the volumetric heat source [W/m3]. On the other hand, the Nusselt number is defined by
means of the following relation:
Q k
= Nu ∆T Equation 7.1-11
A H
Merging the two last equations, we have:
NEA/CSNI/R(2016)15
H
S⋅ ⋅H
Q⋅ H 2 S ⋅ H2
∆T = = = Equation 7.1-12
A ⋅ k ⋅ Nu k ⋅ Nu 2 ⋅ k ⋅ Nu
Obviously, the value of ∆T is limited by the difference between the boiling and melting
temperatures of corium, i.e.
∆Tmax = Tboil − Tmelt Equation 7.1-13
The Rayleigh number can now be evaluated as:
 S H 5 g β ρ 2 Pr SH 2
 ⋅ if < Tboil − Tmelt
 2k µ
Ra =  3
2
Nu 2k Nu Equation 7.1-12
 H g β ρ ⋅ Pr ⋅ (Tboil − Tmelt ) if S H ≥ Tboil − Tmelt
2 2

 µ
2
2k Nu
18
10
17
10
16
10
15
10
14
10
Raileigh number
13
10
12
10
11
10
10
10 Fieg's correlation
9 CORIUM-2D correlation
10
8
10
7
10
6
10
0 1 2 3 4 5 6 7 8 9 10
Pool height [m]
Figure 7.1-3: Value of internal Rayleigh number versus pool Height
The code derives the Nusselt numbers for the boundary surfaces, according to the correlations
deduced by Fieg for a heated pool having aspect ratio of H/D = 0.25, and within the range
107 < Ra < 4·109:
0.414 ⋅ Ra 0.216 for upward facing surfaces

Nu = 1.120 ⋅ Ra 0.103 for downward facing surfaces Equation 7.1-13
 0.244
0.163 ⋅ Ra for lateral surfaces
CORIUM-2D does not manage the liquid pool as a unique volume, with a bulk temperature and a
pool depth. In fact, because of the need of tracking the corium solidification, the cell nodalisation is
maintained in the molten pool, and the following strategy is adopted.
Taking a geometric mean of correlations of the last equation with a double weight for the lateral
surface, we can write:
Nu = 0. 333 Ra 0.202 Equation 7.1-14
NEA/CSNI/R(2016)15
which, when substituted in Equation 7.1-14 allows estimating the Rayleigh number Ra. If S and,
hence, Ra happen to be zero, Nu=1, i.e. pure conduction, is assumed.
Then, for each cell of the molten pool, separate Nusselt numbers for the flows in the horizontal
and vertical directions are computed. While the corresponding term of Equation 7.1-15 is used in the
former case, an interpolation between the correlations Equation 7.1-15 for the upward and downward
facing surfaces is assumed in the latter case with the relative weights of the two terms depending on
the position of the cell in the molten pool. The above procedure, implemented in CORIUM-2D code at
RSE, obviously leads to Rayleigh numbers much lower than those predictable with the usual
correlation for internal Rayleigh (see Figure 7.1-3). Moreover, with the physical restriction imposed
by the maximum
∆Tmax = Tboil – Tmelt, it turns out that the maximum Rayleigh number reachable in a reactor
situation is of the order of:
Ra max ≈ 1012 ⋅ H 3 Equation 7.1-15
Equivalent conductivity at steady-state regime: While heat transfer through a liquid-solid
interface is traditionally calculated accounting pure heat conduction through the related boundary
layer, within the pool bulk also the contribution of the inter-cell mass transfer must be taken into
account, as follows:
Q k ∆T Γ c p ∆T
= + Equation 7.1-16
A ∆x A
where:
∆x = distance between two adjacent cells exchanging heat [m];
Γ = mass flow between cells [kg/s];

A = flow area [m2].
The mass flow Γ can be computed as

Γ = u′ ρ A Equation 7.1-17
u' being the velocity of the mass transferred between cells, assumed to be the turbulent
component of the flow regime.
Now, three simplifying fundamental hypothesis are adopted to account for the convective heat
transfer.
1. It is assumed that the characteristic fluid velocity distribution (and then the related Nusselt
numbers) generated by corium natural convection is equivalent to the distribution which could
be obtained under forced convection, so that:
Nu nat = Nu forced Equation 7.1-18

Nunat and Nuforced being respectively the Nusselt numbers of the natural and equivalent
forced convections.
This hypothesis allows estimating a characteristic velocity of the molten corium, associated to
each cell. In fact, a relation between this velocity and the Nusselt number can be obtained
through the Dittus-Boelter forced correlation for a cooled liquid (taking an average of Nusselt
numbers, depending on the cell location within the pool):
NEA/CSNI/R(2016)15
Re µ µ
u∞ = = 111. 65 Pr −0.375 Nu 1.25
ρH ρH Equation 7.1-19
where u∞ is the characteristic velocity [m/s] of the molten corium at steady state.
2. Within a generic cell, the fluid moves in all directions with its characteristic velocity
(distinguishing only between horizontal and vertical motions, but not between up and down,
and right and left).
3. The velocity of mass transfer between cells may be thought as proportional to the actual
velocity of the fluid within each cell which may be computed from the characteristic velocity
u∞ applying a correction factor fτ for mechanical inertia of the corium (see next paragraph),
i.e.
u ′ = χ u = χ f τ u∞ Equation 7.1-20
where χ is the proportionality constant which may be set by the user (a default value of 0.02 is
provided by the code).
Under these hypotheses, Equation 7.1-18 becomes:
Q k
= ∆T + χ u ρ c p ∆T Equation 7.1-21
A ∆x
which, substituting the value of u given by Equation 7.1-21 leads to:
Q k  µ 
= ∆T + 111.65 fτ χ Pr -0.375 Nu 1.25 c p  ⋅ ∆T Equation 7.1-22
A ∆x  H 
It is possible to define an equivalent conductivity taking into account both pure conductive and
convective heat transfer, i.e.
Q ∆T
= keq Equation 7.1-23
A ∆x
where:
∆x
keq = k + 111. 65 f τ χ µ Pr −0.375 Nu 1.25 c p Equation 7.1-24
H
Because of the uncertainties of this kind of arguments, the option of arbitrarily changing the
proportionality constant between the turbulent and the characteristic velocity of the fluid within each
cell of the molten pool is provided in CORIUM-2D code by means of an input assignment. This
allows evaluating the importance of the convective motions inside the pool for the time evolution of
the system.
A pure conduction heat transfer model is considered at oxide/metal interface.

Input defined
NEA/CSNI/R(2016)15

The heat transfer at pool/concrete interface is described by a Fieg’s modified model and Kulacki-
Goldstein’s correlation modified for non-stagnant pool. A stable crust is assumed for the pool/concrete
interface structure.

The heat transfer at pool upper interface is described by a Fieg’s modified model and Kulacki-
Goldstein’s correlation modified for non-stagnant pool.

CORIUM-2D materials may be thought as an aggregation of different components, which are uranium
and plutonium (both oxide and metal), zirconium oxide, Zircaloy, carbon steel, water, stainless steel,
concrete, sodium, lead and lead-bismuth eutectic. For each material, a library of constant properties
(melting and boiling temperature, latent heat of fusion, and emissivity) and of temperature-dependent
properties (density, specific heat, thermal conductivity, dynamic viscosity, isobaric expansion
coefficient) is included in the code. Dependence on pressure is also taken into account for water. The
properties of each component are computed by means of empirical correlations, while the properties of
the composite materials are evaluated weighting the properties of each component with the
corresponding molar or mass fraction. In particular, each component may undergo a phase change
independently of the others, since solidus and liquidus temperatures are defined for each component.
In any case, thermal properties are evaluated for every component by means of the corresponding
correlations or solidus-liquidus interpolations and finally averaged.
7.1.4. CORQUENCH code

The pool-concrete interface can be modelled in several ways, depending upon user input options. In
particular, the interface can be modelled as devoid of crust material, or transient crust growth,
stabilisation, re-melting, and subsequent failure can be modelled. For situations in which the concrete
surface is modelled as crust-free, the heat transfer coefficient from the melt pool to the melt/concrete
interface can be selected from i) a slag film model (Bradley, 1988) (viz. Bradley’s modification to the
bubble agitation heat transfer model of Kutateladze and Malenkov (Kutateladze & Malenkov, 1978),
ii) gas film models similar to those deployed in CORCON Mod3 (Bradley, Gardner, Brockmann, &
Griffith, 1993), or iii) the empirical correlations developed by Sevón (Sevón, 2008) on the basis of the
CCI test results. When crust growth is modelled, heat transfer from the pool to the crust is modelled
using the Kutateladze and Malenkov correlation (Kutateladze & Malenkov, 1978). Crust failure can be
specified as occurring at a minimum thickness, or failure can occur based on a simple mechanical
loading model that is orientation-dependent.
Depending upon the pool thermal-hydraulic conditions, the upper surface can be devoid of a stable
crust, or transient crust growth is calculated by solving a growth rate equation; the crust material
composition is treated separately from the melt material composition. When water is present, a
NEA/CSNI/R(2016)15
detailed heat transfer analysis is performed at the crust-water interface involving a variety of potential
core debris cooling mechanisms; see 7.1.4.5. On the crust-pool side of the interface, the convective
heat transfer coefficient is calculated using the correlation of Kutateladze and Malenkov (Kutateladze
& Malenkov, 1978).
7.1.4.2. Thermalhydraulics models

A detailed fluid mechanics calculation of convection within the pool is not calculated. Rather, the pool
is treated as well-mixed and at a uniform bulk temperature. Heat transfer at the pool boundaries is
modelled using convective heat transfer correlations. The code can treat fully oxidic to fully metallic
melt compositions, or both phases can be present. However, the code assumes that the oxide and metal
phases are well mixed at all times; i.e. phase stratification is not modelled.
As noted in section 3.2.4, CORQUENCH treats oxide and metal phases as well-mixed. Stratification is
not modelled.

Stratification is not modelled. This assumption is not valid for melt pools containing a significant
metal content under low gas sparging scenarios that can be encountered, for example, with siliceous
concrete.
The MCCI conservation of mass equations and thermophysical property subroutines consider most
core and concrete metals and their corresponding oxides. Melt viscosity is calculated using the
Andrade formula (see Nazare et al. (Nazare, Ondracek, & Shulz, 1977) with a correction for SiO2 as
developed by Shaw (Shaw, 1972). Viscosity enhancement due to build-up of solids within the melt
can be calculated using either the Ishii-Zuber (Ishii & Zuber, 1979) or Kunitz (Kunitz, 1926) models.
Debris specific enthalpy for the various zones is calculated by weighting the individual constituent
enthalpies on a molar basis. Melt void fraction can be evaluated from one of several different
correlations; i.e. those due to Brockmann et al. (Brockmann, Arellano, & Lucero, 1989), Wallis
(Wallis, 1979), or Kataoka and Ishii (Kataoka & Ishii, 1987).
CORQUENCH does not perform a detailed thermo-chemical properties evaluation; rather, the
particular species of key melt constituents (e.g. UO2) is assumed. Distinct metal and oxide phases are
treated as part of the evaluation; as noted previously, these phases are assumed to be mixed. The rate
of oxidation of metallic melt constituents is based on the rate at which oxidising gases from concrete
erosion (CO2 and H2O) are sparging through the melt. The oxide phase diagram is based on the
Lambertson-Mueller phase diagram (Lambertson & Mueller, 1953) for UO2-ZrO2 with empirical
corrections for concrete content based on the data obtained by (Roche, Leibowitz, & Fink, 1993). For
situations in which crust is forming, the crust-melt interface temperature can be based on the crust
composition or the melt composition (user-input modelling assumption).
There is no treatment of chemical reactions between the melt and the atmosphere, or of reactions
in the atmosphere. Metals contained in frozen material do not undergo additional chemical reaction
(unless that debris is re-melted). In reality, some additional oxidation reactions may be expected due to
debris permeability and diffusion within the solidified material.
CORQUENCH does not currently evaluate fission product source term which is an important
element in reactor safety evaluations.
NEA/CSNI/R(2016)15

Heat transfer at the pool upper surface under wet-cavity conditions has been the principal focus area
for CORQUENCH development. Bulk cooling and incipient crust formation are calculated using the
models developed by Farmer et al. (Farmer, Sienicki, & Spencer, 1990), (Farmer, Spencer, Schneider,
Bonomo, & Theofanous, 1992). Following incipient crust formation, crust growth is calculated by
solving a growth rate equation; the crust material composition is treated separately from the melt
material composition. The melt-side convective heat transfer coefficient is calculated using the
correlation of Kutateladze and Malenkov (Kutateladze & Malenkov, 1978). For the case in which the
crust is treated as permeable to water ingression, then the crust dry-out limit can be calculated using
either a user-specified crust permeability, or the dry-out heat flux can be calculated using the
Lomperski and Farmer model (Lomperski & Farmer, 2007).
For situations in which a particle bed develops over the crust, the heat flux from the crust upper
surface may be limited by the particle bed dry-out limit. For this case, the bed dry-out limit is
calculated with the Lipinski correlation (Lipinski, 1980). The heat flux from the crust upper surface is
checked during the calculation to ensure that it does not exceed the effective dry-out limit. Particle bed
formation by melt eruptions is evaluated using the approach of Bonnet and Seiler (Bonnet & Seiler,
1992); i.e. melt dispersal is calculated by assuming that the melt entrainment rate is proportional to the
gas volumetric flowrate times an entrainment coefficient. Several options are provided for evaluating
the melt entrainment coefficient: i) the user may specify the coefficient directly, ii) the entrainment
coefficient can be evaluated with the Ricou-Spalding model (Ricou & Spalding, 1961), or the
coefficient can be evaluated using the model due to Farmer (Farmer, Phenomenological Modeling of
the Melt Eruption Cooling Mechanism during Molten Corium Concrete Interaction, 2006). Consistent
with test observations, the dispersed melt is assumed to be rendered in the form of an accumulating
particle bed (with specified particle diameter and porosity) on top of the crust.
The model can also calculate crust anchoring to the cavity sidewalls, as well as the subsequent
melt/crust separation phase which arises due to concrete densification upon melting. For a given cavity
span, the minimum crust thickness required to be mechanically stable due to the combined weights of
the overlying water pool, particle bed, and the crust itself is evaluated using a first-order plate strength
equation from Roark and Young (Roark & Young, 1975). During the calculation, the upper crust
thickness is compared with that predicted from the Roark and Young equation. When the thickness
exceeds the minimum required to be mechanically stable in the given cavity configuration, the crust is
assumed to attach to cavity sidewalls with the upper surface elevation fixed at the location at the time
of anchoring. Thereafter, the voided melt upper surface location is tracked relative to the crust location
so that the onset of gap formation can be predicted. When a gap does form, debris quenching by the
mechanisms of crust water ingression and melt eruptions is terminated, and there is a corresponding
reduction in upwards heat transfer due to solidification (latent heat) processes. Moreover, a heat
transfer resistance across the gap is introduced into the heat balance, which causes a further reduction
in upwards heat transfer. This methodology allows the prediction of the crust anchoring time and
location for comparison with test results such as those obtained in the MACE Program.
The main simplifications that arise from utilisation of a lumped parameter approach, simple geometry
assumptions, and other CORQUENCH limitations are summarised below.
Dry cavity condition models
In terms of interaction with the cavity, the atmosphere and surroundings above the pool surface
serve only to provide boundary conditions for heat and mass transfer from the pool, as the code
does not include calculation procedures to update the temperature, pressure, or composition of the
atmosphere or the temperature of the surroundings. The calculation of radiation heat loss from the
NEA/CSNI/R(2016)15
pool surface is based on a one-dimensional model with the boundary temperature constant at a
user-specified value. No radiation attenuation by aerosols is treated.
Wet cavity condition models

CORQUENCH has fairly detailed and robust models for core debris cooling under a variety of
scenarios ranging from initial bulk cooling, incipient crust formation, and further debris cooling
with the possibility of treating melt eruption and/or water ingression cooling mechanisms.
Although discrete debris regions with unique compositions are treated, the compositions of these
zones are treated in a bulk fashion; i.e. local composition gradients are not evaluated. The particle
size and porosity of beds formed by melt eruptions are user-specified as opposed to calculated.
Reactor cavity geometry

CORQUENCH utilises a highly idealised cavity modelling approach in which the cavity is
envisioned to consist of either a right cylinder or a notch-type configuration. Although lateral and
axial ablation rates are calculated, one of these (user-specified) geometries is assumed and
maintained over the balance of the calculation. Thus, true cavity rounding effects that are
observed in experiments are not calculated with this code.
Scenario limitations
CORQUENCH does not have a dedicated spreading model; the initial cavity dimensions are
specified and that geometry is assumed through the balance of the calculation. The code can
accommodate ongoing melt pour conditions from the RPV. This material is assumed to relocate
to the melt zone of the core-concrete interaction; i.e. no interaction with crust and/or particle bed
atop the melt zone is treated. In reality, spreading over this previously solidified material would
most likely occur. In addition, melt fragmentation and cooling during relocation through water
over the pool is not addressed in the code.
7.1.5. COSACO code
In COSACO, the heat transfer coefficient from the oxidic melt pool is calculated using a correlation
based on the BALI tests (Bonnet, 1999). The boundary temperature used together with this heat
transfer coefficient is equal to the concrete decomposition temperature. This decomposition
temperature is defined for most COSACO analyses as the temperature for which a solid volume
fraction of the concrete of 50% is reached.
A quasi-steady concrete erosion without considering thermal conduction in the concrete is used.
The locally eroded concrete thickness increment is determined by dividing the heat transferred from
the melt during a time-step by the concrete decomposition enthalpy.
For oxide pool temperatures below the liquidus temperature, phase segregation with crusts at the
pool boundaries is considered. The total amount of solid material in the pool is predicted by the
thermo-chemical solver and is then split between the pool bulk (as floating crystals) and the pool
boundaries (as a crust). The proportion of the total solid settling on a given boundary is derived from
the relative importance of the heat sink associated to this boundary. On the basis of this locally
deposited amount and the surface area of the boundary, a crust thickness is derived. An effective heat
transfer from the melt pool to the concrete is determined by a series connection of the resistances from
the pool and the crust. Using the same assumptions, the crust and the heat transfer coefficient at the
oxide/metal interface for layered melt configurations is determined. The effective heat transfer
NEA/CSNI/R(2016)15
coefficient is determined by the individual thermal resistances in the layers (connection in series),
including that of the crust, if any.
Unlike oxidic corium, the metallic melt has a high thermal conductivity and is less prone to the
formation of insulating crusts at the concrete surface. Therefore, no presence of crust is taken into
account for the metallic melt layer. The interface temperature is chosen as the interface temperature
between two semi-infinite solid slabs brought into contact. Since, in addition to the heat conduction,
also convection occurs in the metallic pool, the effective conductivity used for the calculation of the
contact temperature is empirically determined to be around ten times the molecular value. The heat
transfer coefficient from the metallic pool to the concrete is determined using the Kutateladze-
Malenkov correlation (Kutateladze & Malenkov, 1978).

In COSACO, radiation heat transfer from the melt pool top surface is modelled. For configurations
where the top melt layer includes refractory oxides, a top crust is assumed and considered in the
effective heat transfer coefficient. This radiation yields also concrete erosion in the upper lateral cavity
walls. For situation where the melt pool is flooded with water, a simplified boiling heat transfer
correlation is included.

The convection in an oxidic or mixed melt layer is determined by using a correlation derived from the
BALI experiments (Bonnet, 1999). Due to the intense mixing of the melt pool by the MCCI gases, the
same heat transfer mechanism in radial direction as downwards is assumed. Therefore, the same heat
transfer correlations are also assumed.
For metal melt layers, the Kutateladze- Malenkov correlation (Kutateladze & Malenkov, 1978) is
used.
In layered melt configuration, the BALI and the Kutateladze correlations are used for the oxide and
metal pool layer, respectively. A crust of refractory solid oxides can form at the interface. The overall
heat transfer coefficient is determined by means of a series connection of the different heat resistances.

The melt pool configuration (mixed or layered) is preset by user input. In layered melt configurations,
initially three layers with oxidic melt, metallic melt and slag are considered (from bottom to top). With
ongoing concrete erosion and addition of light concrete oxides, the oxide melt density decreases. This
yields a configuration with the oxidic melt layer (including also the slag) above the metal melt.
During the calculation, the melt state is determined by means of a thermo-chemical equilibrium solver,
The solid and liquid fractions of the melt components are retrieved from the obtained equilibrium
state. The thermo-physical properties of the melt (in mixed and layered configurations) are then
determined by the use of mixture laws. The viscosity of the melt layers is calculated using the model
of Urbain (Urbain, 1987). The influence of solidified fractions on the overall melt viscosity is
considered using a correlation suggested by Ramaciotti (Ramacciotti, Journeau, Sudreau, & Cognet,
2001).
NEA/CSNI/R(2016)15

In COSACO, the thermo-chemical properties of the melt layers are based on the real solution database
COSCHEM, a subset of the TDBCR99 database (Chevalier, 1999). The thermo-chemical equilibrium
is calculated within each time step using the ChemApp solver (CHEMAPP, 1999).
In COSACO, the top melt surface temperature is calculated. In dry conditions, radiation from this
surface is considered. Furthermore, flooding of the melt pool can be used. A simplified boiling curve
is included for this case.
No dedicated modelling on water ingression, melt eruption and debris cooling is currently
included. It is foreseen to implement these features.
7.1.6. MAAP CODE

Crust is always assumed to be present at the interface between the corium and concrete. When the
corium is first relocated (discharged) from the failed reactor vessel, a minimum crust thickness of
1 mm is assumed. The crust is assumed to be quite permeable to the concrete slag and gases. The
molten slag and gas can enter the corium pool immediately, and the interface between the crust and the
concrete does not have an accumulated slag or gas layer. The contact (thermal) resistance between the
crust and the concrete is also ignored. Therefore, the temperature of the outer surface of the crust is
identical to the temperature at the surface of the concrete.
The temperature profile in the crust is assumed to follow a parabolic distribution, as shown in
Figure 7.1-4. The inner surface temperature of the crust is equal to the molten (solidus) temperature
TF,m of the melt. The outer surface temperature of the crust is equal to the concrete surface
temperature Ti. Therefore, the temperature profile T(x) is given by
(T(x ) − Ti ) / (TF,m − Ti ) = 1 − (x / x c )2 Equation 7.1-25
where x is the distance in the crust measured from the inner surface and xc is the thickness of the crust.
The growth (or shrinkage) rate of the crust is based on the overall energy balance in the crust. This can
be represented as
dx c / dt = (q conv "+ q v ' ' ' x c − q cond ") / (ρ c LH c ) Equation 7.1-26
where qconv” is the convective heat flux from the molten corium pool to the crust, qv’’’ is the
volumetric decay power generation rate in the crust, the q cond " = −k c dT(x ) / dx x = x term represents the
c
conduction heat flux from the crust to the concrete, ρc is the density of the crust, and LHc is the latent
heat for the crust.
NEA/CSNI/R(2016)15
Figure 7.1-4: Diagram of crust temperature profile
Once the heat flux from the crust to the concrete is determined, the response of the concrete to the
imposed heat flux is determined by balancing energy at the ablation front. The concrete is modelled as
a one-dimensional slab, nodalised in the direction of heat transfer. Rates of temperature changes in the
slab are solved through the one-dimensional heat conduction equation. For a given heat flux qcond”,
the ablation rate is calculated through the energy balance at the interface as
dx cn / dt = (q cond "−q cond ,cn ") / (ρ cn LH cn ) Equation 7.1-27
where xcn is the thickness of the concrete slab, q cond ,cn " = − k cn dTcn (x ) / dt x = x is the heat flux
0
conducted away from the front, ρcn is the concrete density. The latent heat term LHcn is a lumped latent
heat of concrete including the latent heat for both melting and decomposition processes.
Unlike the corium pool/concrete interface, crust may or may not be formed at the upper interface. If
the cavity is dry, the temperature at the upper interface may exceed the melting (solidus) temperature
of the corium pool, and no crust is formed for this case.
If the crust exists at the upper interface, the temperature profile in the crust is assumed to follow a
parabolic distribution similar to the bottom and side crusts. The temperature at the upper surface of the
crust is determined by the energy balance between the heat loss through radiation (if the cavity is dry)
or convection (if the cavity is flooded) and conduction heat transfer through the crust. If the crust does
not exist, the upper surface temperature is determined by the energy balance between the upward heat
loss and convective heat transfer from the molten pool centre to the upper surface. The energy loss
from the upper crust to water is determined by the corium coolability model in Section 7.1.6.5.
The convective heat transfer is modelled in MAAP5.01 via two options. The first option relies on
users to provide the convective heat transfer coefficients for downward, sideward, and upward
convective heat transfer from the molten pool centre to the crusts. The heat fluxes are given by
hd (1 − f s )n (TF − TF ,m ) for bottom crust
 Equation 7.1-28

 hs (1 − f s ) ( TF − TF ,m )
n
q" = for side crust

 hu (1 − f s ) ( TF − TF ,m )
n
for upper crust
where hd, hs and hu are the respective heat transfer coefficients, TF is the average temperature of the
pool, and TF,m is the crust melting temperature or surface temperature (if upper crust does not exist).
The centre of the pool is modelled as fully or partially molten. If the centre is partially molten, the
NEA/CSNI/R(2016)15
effect of viscosity is represented by the term (1-fs)n, where fs is the solid fraction in the molten centre.
Exponent n is a user provided parameter to account for this effect.
Guidance has been provided in the MAAP manual to help users choose the values of the heat
transfer coefficients and the exponent. Default values of these parameters are hd=3 500 W m-2K-1,
hs=hu=3 000 W m-2K-1 and n=2.75. These values are from comparisons between MAAP predictions
and dry cavity MCCI experiments. The second option is used if users decide not to provide the values
of the coefficients. MAAP code will calculate the coefficients based on the correlations suggested by
Mayinger et al. (Mayinger, Jahn, Reineke, & Steinbrenner, 1976). for this case:
Nu u = 0.36 Ra 0.23 Equation 7.1-29
Nu d = 0.54 Ra 0.18 (H / R )
0.26
Equation 7.1-30
Nu s = (Nu d + Nu u ) / 2 Equation 7.1-31
where Nuu, Nud, and Nus are respectively the Nusselt numbers for upward, downward, and sideward
heat transfer, Ra is the Rayleigh number, H is the depth of the pool and R is the radius of the upper
surface of the pool. This correlation was actually based on an experiment to simulate an in-vessel
corium pool. The pool is assumed to be surrounded by crusts and turbulent natural convection is
developed in the molten pool.
7.1.6.2.2. Heat convection within oxide and metal layers in a stratified pool
The MAAP model assumes that the corium pool is chemically mixed and stratification of metal and
oxide is negligible. This mainly reflects the assumption that gases generated during concrete ablation
are effective at mixing the pool. As long as significant concrete ablation occurs, stratification can be
ignored.
No stratification is currently considered in MAAP. If a stratification model is implemented in the
future, stratification criteria will be considered at that time.
7.1.6.3. Thermo-physical properties
The thermo-physical property model in MAAP is based on the phase diagram method. This method
assumes that all constituents in the corium can be partitioned into metal and oxide phases. For metal, a
pseudo-binary phase diagram is constructed for steel versus other metals. For oxide, a phase diagram
is constructed for U-Zr-O versus other oxides. The mass of the constituents and total energy are
tracked at each time step in MAAP. The phase diagrams are used to find the temperature, solid
fraction and other properties that balance the mass and energy.
MAAP uses a formulation analogous to the Ishii-Zuber (Ishii & Zuber, 1979) correlation to
account for the viscosity effect. As discussed in Section 7.1.6.2.1, a term (1-fs)n is applied to the
convection heat flux, but the exponent is user-specified.
Chemical reactions during MCCI are simulated in MAAP with a model called METOXA. This model
assumes chemical equilibria at each time step. Temperature-dependent equilibrium constants and
activity coefficients are used to find the equilibrium chemical compositions at a specific time.
Chemical reaction heat is deduced from two consecutive time step compositions. Volatile fission
product compounds formed in reactions are assumed to be carried by gas from concrete ablation, and
released to the containment gas space.
NEA/CSNI/R(2016)15
Two sets of data were used by the METOXA model: Gibbs free energy data and reaction
enthalpy data. The former was mainly from the NRC code VANESA (Powers, Brockmann, & Shiver,
1986). For those species and compounds not covered in VANESA, data from other sources were used
(Jackson, 1971), Pankrantz (Pankrantz, 1982), Lange’s Handbook (Deans, 1985), CRC Handbook
(Weast, 1983). The enthalpy data are mostly from Jackson and Pankrantz.
Corium coolability is modelled using two options. The first option, which is consistent with the model
in all versions prior to MAAP5.01, uses input parameters to account for uncertainties in debris
coolability. The second option is a mechanistic model added in MAAP5.01 to address the
phenomenon of water ingression into the upper crust.
The water ingression phenomenon is modelled in MAAP5.01 in the first option as in earlier
versions of MAAP, which use a parameter FCHF to represent the heat flux from the corium pool to
water. This option assumes the heat flux from the pool is controlled by hydrodynamic stability at the
top interface. The formulation of the heat flux is analogous to the critical heat flux for a flat plate
suggested by Kutateladze (Kutateladze S. , 1951):
(
q ' ' = FCHF ⋅ gσ(ρ l − ρ g ) / ρ g )
2 1/ 4
ρ g h fg Equation 7.1-32
where ρl and ρg are respectively the densities of water and steam, hfg is the latent heat of evaporation,
and σ is the surface tension. Parameter FCHF is provided through user input.
The second option is a mechanistic model which does not rely on a user-provided FCHF. This
model combines the studies by Lister (Listner, 1974) [on cracking and water penetration into hot solid
rock, and a one-dimensional heat transfer model by Epstein (Epstein, 2006). The crust is represented
by three regions in the model: at the top a quenched zone where the crust temperature is close to the
water saturation temperature; in the middle a dry-out zone where the cracks are filled by superheated
steam, at the bottom is a crack-free zone where the solidified crust is contiguous. A constant
quenching speed is assumed in the model which allows the three regions to move downward at the
same speed u. Decay heat generation in the crust is considered in this model.
One of uncertainties in water ingression modelling is the role of metallic materials in the upper
crust. All the experiments on heat transfer from corium to water so far were performed with oxidic
materials. It is uncertain whether water ingression will still occur and what will be the magnitude of
the heat flux to water if the corium has a significant amount of metallic materials in it. The
mechanistic model in MAAP has shown that permeability and heat flux to water are sensitive to
properties, especially the thermal conductivity in the upper crust, which is a function of the amount of
metallic material present. To allow sensitivity studies, an option has been added in MAAP5 to allow
users to specify whether the thermal conductivity in the crust is calculated with all the constituents
(oxidic and metallic) or with only the oxidic constituents. The default option is to calculate the thermal
conductivity with only the oxidic constituents.
MAAP also has a melt eruption model, which is based on the Ricou-Spalding entrainment
correlation (Ricou & Spalding, 1961), given by
jm = E 0 (ρ g / ρ m )
1/ 2
jg Equation 7.1-33
where E0 is a constant, jm and jg are respectively the superficial velocities of entrained melt and gas,
and ρm and ρg are the respective densities of the melt and gas. According to Epstein (Epstein, 2006),
the value of E0 should be 0.08 which is consistent with the entrainment data reported elsewhere
(Tourniaire B. , Seiler, Bonnet, & Amblard, 2006) for the thin crust case.
NEA/CSNI/R(2016)15
The corium structure model in MAAP5 is a three-layer model, including a particle bed, an upper
crust and the remaining continuous pool. Mass and energy in each region are integrated separately.
The particle bed can be generated at the time of corium relocation by means of corium-water
interaction. Users can control (through options and modelling parameters) the amount of particles
generated due to the interaction and the amount deposited on top of the upper crust as a particle bed.
The average size of the particles during the corium-water interaction can be either a user-input or
code-calculated value. The particle bed can also be generated at the time of melt eruption. In this case,
the erupted corium mass is deposited on top of the upper crust as a particle bed. The average particle
size generated by melt eruption is a user input. Heat transfer from the particle bed to water is based on
a correlation suggested by Lipinski (Lipinski, 1980) or the one suggested by Henry et al. (Henry,
Epstein, & Fauske, 1982). The user may select either of these two models. Heat transfer from a dry
particle bed to surroundings is based on the model developed by Epstein (Epstein, Cheung, Chawla, &
Hause, 1981).
Water ingression is assumed to occur in the upper crust. If the heat flux from the particle bed to
the overlying pool of water is less than the critical heat flux, water is available to the upper crust and
ingression heat flux is evaluated. Growth or shrinkage of the upper crust is based on a heat balance
from the upper crust to water and from the molten corium below to the upper crust.
The corium below the upper crust is modelled as a contiguous pool with a molten centre and
crusts separating the molten centre and concrete. It does not contact water directly. Therefore there is
no direct heat transfer from this region to the water pool above.
7.1.6.5.1. Pool shape model
The corium pool in MAAP5 can be modelled as different shapes. The default option is to model the
pool as a cylinder. As ablation occurs in the floor and sidewall, the cylinder floor and sidewall move
but the cylindrical shape is maintained. The second option is to model the pool as a rectangle or square
with vertical sidewalls. Similar to the default option, if ablation occurs, the floor and sidewall
boundaries move but the shape is maintained. For this option, users can also specify whether all four
sidewalls ablate or only two of the sidewalls ablate. The third option is a sophisticated ablation shape
evolvement model added in MAAP5.03. This model assumes the initial cavity shape is cylindrical. As
ablation starts, different points along the floor and sidewall can move at different speeds. Eventually,
the ablation front evolves from a cylinder to an axisymmetric curved shape. The last option is to model
both a cavity and a sump. The cavity and sump can be cylindrical or rectangular. As ablation starts, the
floors and sidewalls of the cavity and sump can move at potentially different speeds, but the
cylindrical or rectangular shape of the cavity and sump is maintained.
The simplifications and limitations in MAAP are summarised below:
Thermo-physical property model:
The thermo-physical property model in MAAP is based on an approach of partitioning
different constituents in the corium pool into metal and oxide phases. This approach has a
limitation in some cases when the partitioning between metal and oxide phases is not straight
forward. For example, if a significant amount of uranium and zirconium mass is in the corium,
some amount of the uranium and zirconium will be associated with the oxide U-Zr-O, while
the remaining mass will be associated with steel and B4C. The partitioning rule should ideally
be based on more sophisticated multicomponent phase diagram for this case.
Coolability model:
One of limitations of the MAAP coolability model is the lack of a bubbling heat transfer
enhancement during bulk cooling. The surface area for heat transfer to water is simply the top
NEA/CSNI/R(2016)15
area (at a single elevation) based on volume versus height table for the cavity. If a gas
bubbling effect was considered, the heat transfer area would be larger than the top area based
on the volume versus height table. Also, the criterion of incipient top crust formation is based
strictly on the top surface temperature in MAAP5. If the temperature is less than the melting
temperature of corium, a crust is assumed to form. Again, if a gas bubbling effect were to be
considered, the incipient crust formation would be delayed. This limitation tends to produce
conservative (perhaps overly conservative) results when corium interacting with high gas
content concrete is flooded by water.
Pool shape model:

The corium pool in MAAP is assumed to be an axisymmetric shape. Non-axisymmetric shapes
are not currently allowed. Also, the sophisticated ablation shape model in MAAP5.03 is
currently available only for the (initially) cylindrical cavity.
It is worth noting that the limitations 1) do not significantly affect MAAP calculations of corium
temperature and ablation depth. The reason for this is that the heat transfer from molten corium to the
surrounding crust is based on a parametric approach in MAAP. This approach relies on heat transfer
coefficients hd, hs and hu and an exponent n provided through inputs. The suggested values of these
coefficients and exponent are based on extensive comparisons with dry and wet cavity MCCI
experiments. Therefore, even if there is some error in the thermo-physical properties, the experiment-
based coefficients and exponent will compensate for the error.
7.1.7. MEDICIS CODE
Two types of pool/concrete interface models can be handled by MEDICIS
Model 1: the heat transfer towards the pool/concrete interface is described the concept of a
solidification temperature Tsolidif determining the interface temperature between the convective
bulk pool zone and a mushy crust as indicated in Figure 7.1-5. In the frame of this model, the
heat flux to the interface φ is equal to h conv .(Tbulk − Tsolidif ) and depending on the solidification
temperature as long as a crust is present at the interface.
Figure 7.1-5: Pool/concrete interface structure in case of model 1
Model 2: this very simple model permits to account for the 2D ablation anisotropy, assuming
that no stable crust can build- along the pool/concrete interface whatever the concrete type; the
heat transfer from the bulk pool to the concrete interface is determined only by convection in
the bulk pool and heat transfer at the interface; the interface thermal resistance depending on
the interface orientation is imposed using the hslag heat transfer coefficient, as indicated in
Figure 7.1-6; this permits to model a simple slag layer made of concrete oxides or a solid
accumulation or an inert crust without any link with a solidification temperature.
NEA/CSNI/R(2016)15
Figure 7.1-6: Pool/concrete interface structure in case of model 2
Model 2*: In case that the convective heat transfer coefficient (htc) turns out to be much
larger than hslag as imposed by the user (an observation made in validation calculations for
typical experiments) an alternative to model 2 is discussed, which introduces an effective heat
transfer coefficient in replacement of the combination of slag layer and convective htc.
The concept of solidification temperature is used to evaluate the interface temperature between the
bulk pool and the upper crust at the upper pool interface in case of both pool/concrete interface
models. The solidification temperature is evaluated either from liquidus and solidus temperature using
the formulas Tsolidif = γ.Tsolidus + (1 − γ ).Tliquidus – the drawback of this criterion is its lack of physical
basis. Therefore a new more physical criterion is used for evaluating the solidification temperature
from a volumetric liquid fraction obtained by an interface with the NUCLEA thermochemistry
database (Cheynet, Chaud, Chevalier, Mason, & Mignanelli, 2004) and corresponding to a corium
mobility threshold of 0.5 (Cranga, Mun, & Marchetto, 2010).
Recommendations:
• Use of the model 2 without crust at the pool/concrete interface with a profile of hslag value
in the range of a few tens to a few hundred of W/m2/K depending on the concrete type.
• Evaluation of the temperature for the upper crust build-up (solidification temperature) from a
corium volumetric molten fraction equal to 0.5.
• If a crust forms heat transfer to the top surface is governed by 2 parameters: the heat transfer
coefficient for melt convection and the solidification temperature.
• Other combinations are still under discussion (based on the pool/concrete interface model
2*).
The only mechanism described is convection due to gas bubbling. The correlation used is derived from
BALI data (Bonnet, 2000). Per default heat convection is assumed to be non-depending on the
interface orientation angle; as a user’s option a multiplicative factor function of angle can be used.
The recommended choice for heat convection correlation and 2D distribution (if using
pool/concrete interface models 1 or 2) is the BALI based correlation with no dependence on the angle.
The correlation used for convective heat transfer is Greene’s correlation (Greene & Irvine, Heat
transfer between stratified immiscible liquid layers driven by gas bubbling across the interface, 1988)
with a possible multiplicative factor. The impact of possible crusts built-up at the oxide/metal interface
on heat transfer is taken into account: heat transfer between two adjacent layers is described by a
thermal resistance model taking into account heat convection within each layer on both sides of the
NEA/CSNI/R(2016)15
interface and heat conduction in the crust possibly appearing in each layer at this interface (in
particular in more refractory oxide layer).
The recommendation on correlation choices and assumptions is standard Greene’s correlation
with a possible oxidic crust.
The stratification criterion depending on gas bubbling is derived from BALISE experiments
(Tourniaire, Seiler et Bonnet 2003) (Tourniaire, Seiler, & Bonnet, 2003). The superficial gas velocity
Jg threshold value below which pool stratification occurs is Jg < b_HS (ρbot – ρtop )/ ρbot m/s. The
superficial gas velocity Jg threshold value above which pool stratification disappears if Jg > b_SH
(ρbot – ρtop )/ ρbot m/s.
A minimum thickness of metal layer is required for stratification occurrence and combined with
the criterion on minimum Jg.
Recommendations on parameters of stratification criteria: b_HS = 0.027; b_SH > 0.1 (to avoid
oscillations between stratified and homogeneous configurations); the minimum metal layer thickness
required for stratification is 3 cm.
7.1.7.2.4. Thermophysical properties
Urbain’s model (Urbain, Bottinga, & Richet, 1982) is used for determining the liquid oxide mixture
viscosity in combination with several optional models for evaluating the increase of the viscosity due
to the solid fraction (per default Stedman’s model (Stedman, Evans, & Woodthorpe, 1990), as an
option either Thomas correlation (Thomas, 1965) or Ramacciotti’s correlation (Ramacciotti, Journeau,
Sudreau, & Cognet, 2001).
A consistent set of simplified enthalpy functions is used for all species either in condensed phase
or in gaseous phase permitting to take into account oxidation reactions. An additive law is used for
evaluating the heat capacity of mixtures. The enthalpy evaluation for a partially liquid mixture takes
into account the molten mass fraction deduced from the interface with NUCLEA described hereafter.
In MEDICIS, thermo-chemical data are obtained by using the thermo-chemistry database NUCLEA
(Cheynet, Chaud, Chevalier, Mason, & Mignanelli, 2004). The needed thermo-chemical data (mainly
molten fraction) in the range of temperature and corium composition encountered during MCCI are
generated in two steps, using an interface with the GEMINI2 thermo-chemistry equilibrium code
(Cheynet, Chevalier, & Fischer, 2002) and the NUCLEA database. First a set of compositions of
interest is defined: this is the line in the compositions space obtained by mixing the given initial
composition mass with increasing amounts of the considered concrete, taking into account oxidation
reactions between the gases released from the concrete and the corium. Then, in a pre-processing
phase, the interface is run for a set of compositions chosen to realise a composition meshing of the line
defined above. Second, during the MEDICIS calculation, thermodynamic data are obtained by
interpolation from the interface results. This two-step procedure avoids direct calls to the GEMINI2
code solver during the MCCI calculation itself and permits very fast calculations with MEDICIS code.
7.1.7.3.1. Coolability models
Top quenching models implemented or to be introduced soon in ASTEC/MEDICIS code are the
following.
Before quenching onset of upper crust (bulk cooling), a standard scenario describing the upwards
radiation without impact of gas bubbling at the upper interface temperature occurs towards the boiling
water interface at saturation temperature; no description of transition boiling regime is done. After
NEA/CSNI/R(2016)15
onset of upper crust formation, power extracted at the upper crust interface is the sum of radiated
power and of power extracted by water ingression.
As alternative to the above mentioned standard scenario there is a boiling model available that
calculates the heat flux through the top interface including a developing crust layer under
consideration of a Nukiyama boiling diagram (Spengler, 2012).
Water ingression treatment is based on the use of a Darcy’s law with a crust permeability value
given as a user’s input.
The simplified model of melt eruption evaluates the melt entrainment using Ricou-Spalding’s
correlation (Ricou & Spalding, 1961); a proportionality factor determining the ratio between volume
rate of entrained melt and gas volumetric rate is a user’s input.
A more mechanistic model (Cranga, Mun, & Bottin, 2012) to be introduced in later ASTEC V2.1
version will include following features:
• Melt flow hydrodynamics model based on the fountain model deduced from PERCOLA’s
model (Tourniaire B. , Seiler, Bonnet, & Amblard, Liquid Ejection Through Orifices by
Sparging Gas – The PERCOLA Program, 2000).
• Gas phase distribution: gas flow distribution between crust cracks and eruption holes is
evaluated assuming the same pressure drop through the crust cracks and eruption holes.
• Hole geometry determination: constraints on hole diameter and hole density derived from
following criteria based on energy balance, hydrodynamics and mechanics aspects: non-bulk
freezing at the top of hole, non-bulk freezing of melt along hole lateral wall, limited lateral
crust build-up along the hole, no gas build-up below the crust, non-flooding of holes by
water, these models will be derived partly from more recently proposed models by M.
Farmer (Farmer, Phenomenological Modeling of the Melt Eruption Cooling Mechanism
during Molten Corium Concrete Interaction, 2006) and K. Robb (Robb & Corradini, 2010).
• Debris bed description.
Equations for the mass and energy balance equations and height evolution of debris layer are
detailed hereafter:
d =m
m  erup − m
 d=>crust
Equation 7.1-34
dTsat dP
 vap .(hst,v − hst,l ) = −(mw .cp w + md.cpd ).
m .
dP dt
+ Pcrust,upwards + Pd,decay + m [ ]
 erup . hcor (Tpool ) − hd (Tsat ) + m [ ]
 g . hg (Tcrust ) − hg (Tsat )
Equation 7.1-35
where: m  vap is the flow-rate of vapour generated in the water pool determined from previous equation,
m w is the mass of water present in the debris zone, md is the mass of debris, Pd,decay is the decay
power released in the debris bed, Pcrust,upwards is the power transmitted by conduction from the upper
crust up to the debris bed,
m erup is the mass transfer rate from the pool upper layer to the debris bed due to melt eruption, the
ejected melt is supposed to be at the pool temperature, ignoring heat exchange with the upper crust,
m d=>crust is the mass transfer rate of debris from the debris bed down to the upper crust defined below.
hcor is the average enthalpy of the corium in the pool, hg is the enthalpy of gases.
In case of dry-out due to an insufficient water flow rate, this dry-out is supposed to occur at the
bottom of the debris-bed. The mass transfer rate of debris from the debris bed down to the upper crust
is obtained from the following equation:
NEA/CSNI/R(2016)15
dh d,dry
 d=>crust = ρ d .
m .A d
dt Equation 7.1-36
no dry-out heat flux limitation on debris cooling is taken into account, i.e. the debris bed remains
cooled at saturation temperature as long as it remains quenched by water.
Main code simplifications or limitations, excepted for the lumped parameter approach, are the
following ones in the MEDICIS code:
• Te simplified structure of the pool/interface model where the slag layer may include a
combination of real slag film, gas film and in case of the “no crust” model (model 2) crust
fragments.
• Heat conduction is neglected behind the ablation front.
• Corium masses stored in crusts along the pool/concrete interface are not taken into account.
• Thermal inertia of crusts along the pool/concrete interface is neglected.
• In case of the upper crust, the crust mass increment is mixed with the whole crust mass and a
composition averaged over the total crust thickness.
• Main transport properties of mixtures (viscosity excepted) are evaluated using a mass or
volume averaging on individual species; this might be wrong for some transport properties
such as heat conductivity and still more surface tension.
• No intermediate pool configuration between the homogeneous and stratified ones is possible;
this causes in practice an overestimation of the time period during which the pool remains
stratified.
• The attenuation of radiative heat flux above the corium pool by aerosols is ignored or, in
case of the ASTEC version coupled to CPA containment module, evaluated crudely by a
simple attenuation length given as a user’s input.
• A detailed description of heat transfer from the upper pool interface to surrounding walls is
possible only if using the MEDICIS and ASTEC containment module CPA in a couple
mode.
• In the present ASTEC version, the debris bed built-up during melt eruption is not explicitly
described since ejected corium debris, once quenched, are mixed with the upper crust; this
leads to an underestimation of the impact of top quenching on the decrease of ablation
kinetics.
• No film boiling curve is taken into account to describe the upwards heat transfer during the
top quenching phase.
3) For description of reactor pit geometry:
• Non-axisymmetric cavity shape (e.g. with a lateral corridor) cannot be described adequately.
4) Concerning the possible scenario:
NEA/CSNI/R(2016)15
• No lateral melt-through can be taken into account excepted manually with a reduction of
corium inventory and a code restart.
• Successive corium pours from the vessel to the reactor pit can be described on a simplified
way: corium poured from the reactor vessel is assumed to be spread over the whole pit area
instantaneously; poured corium is distributed between oxidic and metallic layers; no direct
heat transfer between the present water inventory and the poured corium falling through is
assumed to occur.
• No corium spreading model is coupled with MEDICIS: in case of a multiple cavity concept,
the pouring of corium from one cavity to another connected cavity can be described at least
in some geometries (e.g. in EPR core-catcher); however in the general case, the spreading
from a cavity to another connected cavity with a possible stop of the slug by freezing at the
front cannot be treated.
7.1.8. SOCRAT CODE
The pool/concrete interface is not separated as definite material, specific boundary or so. The heat
transfer through the interface is ruled by thermal conductivity, heat capacity and latent heat of the
composite materials. Present interface model relies on user-defined properties (thermally dependent
properties, Tsol & Tliq), composition of individual pool components (fuel, oxidised and non-oxidised
metals, concrete) and chemical interaction model. Based on the components properties thermal
conductivity of the compound in each FE is calculated. As a result the zones with different thermal
properties appear that can be treated as representing mushy zone, crust or their combination. Since 2D
temperature field is obtained on each step, FE state (solid or liquid) and properties are determined by
its temperature. Gas film that can hamper heat transfer from the pool to the concrete is not considered.
The image shows heat conductivity in the area near the piece of melt-concrete boundary with
white related to the values of 1-2 W/m*K in solid concrete and black related to the values of several
hundred W/m*K in the melt pool. Here areas with different heat conductivity can be attributed to solid
concrete, crust, mushy zone or melt.
Figure 7.1-7: Diagram of crust modelling

Top crust description: pool/crust temperature, crust energy balance
NEA/CSNI/R(2016)15
Heat transfer at the top surface and upper crust boundary temperature is governed by the applied
boundary conditions. Usually these are convective conditions
Fconv = H (T − Tb )
and heat radiation conditions
FRAD = εσ B (T 4 − TB4 )
for the boundary heat fluxes calculated from the "infinite" temperature or from the enclosure radiation
heat exchange (HEFEST−EVA model). These boundary conditions are applied to each FE on the
boundary thus allowing their spatial variation. If calculated temperature of some FE at the surface is
below the solidus (Tsol) of the compound, this FE becomes solid. When all the FEs at the boundary
become solid that is interpreted as crust formation. The crust thickness results from the calculated
temperature field and upward heat flux. The crust elements may be referred separately for changing
their properties in accordance with the definite physical model. A heat exchange with external coolant
is ruled by heat transfer coefficient of 3rd kind condition, which is varied in accordance with the used
coolant model. Heat flux from the crust (and other) boundaries is calculated during the stiffness matrix
and load vector formation. Energy balance in the crust FEs is calculated in the frame of general
procedure.

The main tasks for the convective heat transfer model are: a) estimation of heat transfer to the melt
boundaries; b) estimation of temperature across the pool.
Mechanisms: Spatially non−uniform effective orthotropic heat transfer coefficients (EOC) are
used for modelling of convective heat transfer (Ozrin, Tarasov, Strizhov, & Filippov, 2010) (Filippov,
E.V., & D.D., 2013). The EOCs are estimated on the base of averaged Nusselt number taken from
empirical or CFD based correlations Nu(Ra) or Nu(Rai). Basic configurations for the identification of
the pool volume shape are torispherical (hemispherical) molten pool and vertical cylinder (with
internal heat generation or input heat flux). For the stratified molten pool in the reactor lower head the
additional algorithms are used along with EOC model for profiling of sidewall heat flux.
Correlations used: Nu(Rai) correlations for torispherical (hemispherical) pools (BALI, COPO II
Lo (Bonnet & Seiler, 1999), (Helle, Kymalainen, & Tuomisto, 1998). For the cylindrical molten pool
at high Rai some CFD calculations results are applied ( (Filippov, 2011), (Filippov, 2011b).
Dependence on the interface orientation: Interface orientation is taken into account implicitly in
the form of the correlations Nu(Rai) used for estimation of heat transfer coefficients in the pool of
particular shape.
Recommendations on correlation choices and 2D distribution: For curvilinear walls BALI
correlations may be recommended as most reliable for high Rai. But the bubbling contribution in the
heat transfer seems to be very high, and the question about the correlation choice is open to discussion.
Correlation used for convective heat transfer. In oxide layer − BALI and COPO 2 Lo. Metallic layer
in MCCI is located under the oxidic one and its heat transfer is ruled by bubbling − the above
mentioned EOC is used with high axial conductivity and with the radial conductivity close to the
nominal value. The additional EOC terms are based on Blottner model (Blottner, 1979). The available
data on convective heat transfer in cylindrical pools are limited by the range Rai=<1 014. For the
NEA/CSNI/R(2016)15
larger values of Rai the tested CFD results on natural convection are used ( (Filippov, 2011),
(Filippov, 2011b).
Impact of possible oxide crust on heat transfer. The temperature of a finite element containing the
melt and located near the melt boundary can decrease below the melting point. Then this element is
treated as solid and its heat conductivity decreases. Thus the crust formation on the wall and at the
oxide/metal interface and its influence on the heat transfer is treated automatically. If the element
melts again, its heat conductivity increases effective and heat transfer coefficient is applied.
Recommendations on correlation choices and assumptions. The main assumption – the bubble
dynamics doesn't change significantly sidewall heat flux distribution in comparison with the
conditions of the single phase high Rai experiments (BALI, COPO). This assumption should be
checked on a wide range of experiments with large size pools (now most of them are not available for
validation of HEFEST−EVA).
In current version the criteria are user-defined time and liquid fracture (for in−vessel calculations).
Minimum metal layer required for stratification occurrence − the thickness is greater than 2 cells
of the mesh
Superficial gas velocity threshold is expected to be introduced in further versions for ex-vessel
applications.
No recommendation on parameters of stratification criteria.
Three main properties are used in stiffness matrix formation in FEM procedure: heat conductivity,
specific heat, and density. The heat conductivity is represented by the effective orthotropic heat
transfer coefficient mentioned above, which is calculated using material properties and Nu(Ra)
correlations with some additional corrections. The coefficients depend on the pool aspect ratio and are
different for horizontal and vertical EOCs. Here the viscosity is used only in calculations of Rayleigh
number and may be constant or temperature dependent. In a liquid pool for each concrete component
(SiO2, CaO, MgO, etc.) as well as for core melt components (Fe, Zr, U and their oxides) a dependence
of heat conductivity (which is based on conductivity for liquid phase) on temperature can be defined
by user. In this way low heat conductivity values defined in input file allow accounting for silica and
refractory materials in the melt.
Enthalpies of species are calculated from their heat capacities defined by user in input file.
Mixture heat capacities are obtained as sum of weighted heat capacities of individual species.
Thermo-chemical properties of melt components are originated from IVTANTHERMO database
(Gurvich, Iorish, Chekhovskoi, & Yungman, 1993), (Gurvich, et al., 1989). Chemical reactions are
accounted by an additional heat source or sink in the reaction area based on the standard heat of
formation stored in the internal database. The properties of homogeneous mixtures are calculated by
averaging. The thermochemistry solver is called every time step. The reactions are divided into
interface and bulk types. No reactions on the metal−oxide boundary are considered. The eutectics and
other nonlinear effects violating the averaging rules are taken into account separately. The effects of
bubbling and of the composition on the viscosity aren't yet accounted.
Top quenching models:
NEA/CSNI/R(2016)15
The interaction with the outer coolant is described by 3rd kind (Robin) boundary conditions as
mentioned above: λdT/dt=H(x,T,t)(T−Tout(x,T,t). The physics of heat exchange is described by
H(x,T,t) and Tout(x,T,t) values − they may be varied in accordance with the correlations built in
HEFEST−EVA or may be calculated in external thermalhydraulics code (SOCRAT/RATEG), which
may be coupled with HEFEST−EVA in the full SOCRAT calculation. In particular, for VVER core
catcher the correlation (Lopukh, Loginov, & al., 2000) for the heat transfer coefficient H(x,T,t) is used
for taking into account film boiling and radiation heat exchange through the film during the melt top
flooding.
The radiative heat transfer is ruled by radiation boundary condition
λdT/dt=ε(x,T,t)(T4−T4out(x,T,t) where Tout(x,T,t) may be user's defined or calculated in built−in
enclosure radiation procedure (including the calculation of view factors matrix for quasi−cylindrical
cavities). No effects of gas bubbling are considered. No transition regimes are considered in quenching
modelling − the heat exchange with the coolant starts at definite time and the boiling is always
assumed to be film boiling.
After onset of upper crust due to quenching
Water ingression and melt eruption can be accounted by user input as above mentioned heat
removal coefficient H(x,T,t). No internal evaluation of heat removal increase due to water ingression
and melt eruption is performed in current version.
Debris description
No porosity is supposed in arriving from the RPV debris (melt). The model of its arrival is of
“donor-receiver” type. The geometry of arriving debris is prescribed by the mesh partitioning (the
order in which different cavity areas are filled is defined in input data). The boundary between the
melt and the solid debris moves when the solid elements melt and finally disappears when whole
debris is molten. Mass and energy balance are monitored in the same way as in all material relocation
procedures.
Main code simplifications or limitations are those expected for the 2D approach without solution of
the flow equations for the melt. For SOCRAT/HEFEST-EVA code they are the following.
• Concrete reinforcement is modelled as uniformly distributed steel component, in the same

way as for other inhomogeneities. Mixture properties are used for multicomponent materials.
• Crust treatment requires relatively thin meshing of the significant part of a calculation
domain that increases total CPU time.
• Melt relocations and convective heat transfer are treated simplistically, without solution of
flow equation; that is implemented as:
a) Melt relocation (arrival, stratification, spreading) is modelled as quick (one time step)
change of melt configuration.
b) Heat transfer in the melt is modelled by effective orthotropic heat transfer coefficients,
which are estimated on the base of the known correlations obtained for homogenous pool
(without stratification and bubbling) of definite shape.
• The correlation based approach demands verification (CFD) against representative set of
configurations.
NEA/CSNI/R(2016)15
• Transition from the correlations used for cylindrical cavity or for torispherical
(hemispherical) cavity lacks experimental support. (The correlations, being developed on the
basis of pools with fixed shape and size, may be incorrect for MCCI pools with changing
configurations).
• The effect of bubbling on heat transfer is not considered in present version.
• Melt eruption and water ingression can be accounted for only by user defined boundary
conditions.
• No film boiling curve is taken into account to describe the upwards heat transfer during the
top quenching phase.
• 2D approach for reactor pit geometry: that allows only axisymmetric configurations –
corridors and other non- axisymmetric elements can be treated by introduction of space with
equivalent area of melt-concrete interface or by solution of coupled problem in which the
surrounding compartments are modelled separately/.
4) Concerning the possible scenario:
• Successive corium pours from the vessel to the reactor pit is described on a simplified way:
corium poured from the reactor vessel is assumed to be located at the definite places, which
are now prescribed by the code user.
• Poured corium is distributed between oxidic and metallic layers; no direct heat transfer
between the present water inventory and the poured corium falling through is assumed to
occur.
• No one-run model of corium spreading along the basemat.
5) Chemistry model
• The reactions are conservative in terms of H2 and CO generation – the only limiting factor is
reagents availability.
7.1.9. TOLBIAC-ICB CODE
A crust exists at the pool/concrete interface. In the frame of the hypothesis of thermodynamic
equilibrium, the phase segregation model is used (Seiler J. -M., 1996), (Seiler & Froment, 2000). The
interfacial temperature Tint between melt and crust is then the liquidus temperature calculated by
GEMINI2, using the evolving melt composition.
The GEMINI2 code is developed by THERMODATA (Cheynet, 2007), and gives
thermodynamic equilibrium between species, by mean of minimisation of Gibbs energy. The data base
that is used by GEMINI2 is NUCLEA. As an option for fast calculations, the code user may select the
use of pseudo-binary laws, fitted from GEMINI2 results, which gives formulas for the liquidus
temperature depending on the concrete composition. By default, the crust composition is calculated
through a coupling to GEMINI2, the code user can decide to consider a homogeneous solidification
(crust layer having the composition of pool at the time of its formation) or he can use pseudo-binary
laws to calculate its composition. The crust is calculated assuming a steady state conduction regime
(a transient regime can be used).
NEA/CSNI/R(2016)15

A crust generally appears at the upper surface of the pool because of radiative heat transfer towards
the reactor cavity or water aspersion. Crust thickness and crust surface temperature are calculated,
supposing either a steady state regime or a transient crust growth. The melt to crust heat exchange is
governed by the heat transfer coefficient and the interfacial temperature (pool liquidus temperature),
and does not depend on the external conditions. The crust surface temperature and crust thickness
depend on the external conditions, which may be radiation or heat exchange with a water layer. When
there is no water aspersion, the crust density is larger than the melt density (crust composed of
refractory materials), and the crust continually forms and sinks: it is then only taken into account for
the heat transfer, and no crust increase is taken into account. In case of water aspersion, a crust
porosity is supposed, and the crust increase is taken into account in the model.
For the radiation above the melt surface, two levels of modelling exist in TOLBIAC-ICB. The
most simple radiation model only considers a radiation between two infinite planes, the upper crust
and the surroundings, depending on the surroundings temperature and the emissivity of the pool
surface and of surroundings. A more complex model also exists. The cavity above the melt level is
described, form factors are considered in the radiation model, and ablation of the concrete vertical
walls above the melt level is taken into account. The ablation of a horizontal steel wall representing
the vessel structure above the melt may also be modelled. Heat losses through gas or aerosols escaping
from the melt pool to the containment may be taken into account through a coefficient of power loss
defined by the code user.
The code user may select several different heat transfer correlations on each boundary as quoted above
in section 3.2.9. The code user may also simply modify the ratio between the bottom and lateral heat
transfers. The recommendation is to use weighting factor rather than very different correlations,
written in terms of different variable groups, since this last solution may lead to unexpected variations
of the power split.
In case of stratified configuration, the Greene and Irvine’s correlation (Greene & Irvine, Heat transfer
between stratified immiscible liquid layers driven by gas bubbling across the interface, 1988) at the
interface between the two-melt layers is recommended.
The code user can choose between three criteria for stratification: that from BALISE test performed at
CEA (Tourniaire, Seiler, & Bonnet, 2003), Epstein (Epstein M. , Petrie, Linehan, Lambert, & Cho,
1981), or BALISE and Epstein together. Details of the criteria are given by Tourniaire and Bonnet
(Tourniaire, Seiler, & Bonnet, 2003). The BALISE criterion depends only on the density of the
liquids, whereas the Epstein criterion depends on the densities and also of the height of the layers.
For the simulant fluids used in the BALISE experiment, the BALISE and Epstein criteria give
results of the same order. But, when applied to reactor cases, the tendency is different with a delayed
stratification in case of the Epstein criterion, due to the low thickness of the metallic layer. The criteria
are as follows:
BALISE
Stratification occurs if the gas superficial velocity Jg is lower than a limit velocity Jg lim :
NEA/CSNI/R(2016)15
(ρ H − ρL )
Jg lim = F1 × 0.054 × m/ s
ρL
Equation 7.1-37
where index H corresponds to the heavy liquid and L to the light liquid.
It must be noticed that only the gas superficial velocity corresponding to the bottom concrete
ablation is used in the criterion. The gas generated by lateral concrete ablation is not supposed
to participate to the global motion in the pool, but only to flow along the lateral wall.
EPSTEIN
For a homogeneous melt, stratification occurs if the equivalent density ρ eq after heavy and
light oxide mixing is greater than the density of the light liquid ρ L , with:
ρ H .(1 − α )  ρ .V 
ρ eq = .1 + L L  Equation 7.1-38
1 + V L / V H  ρ H .V H 
where V correspond to the volume of the layers.

For a stratified melt, mixing occurs if the equivalent density ρ eq of heavy oxide with void is
lower than the density of the light liquid ρ L , with a different equivalent density:
ρ eq = ρ H .(1 − α )
Equation 7.1-39
BALISE and EPSTEIN:
In this case, both criteria Jg lower than Jg lim and ρ eq greater than ρ L must be satisfied.
Additional conditions for the stratification and the de-stratification of the corium pool:
Another condition for the de-stratification of the pool is still used in the present version of the
code: when the pool is stratified, if the thickness of the liquid metal layer becomes lower than
3 cm, then the de-stratification of the pool takes place.
Physical properties are not of minor interest, since they are used for determination of melt height and
heat transfer areas (density), and also for heat transfer coefficients and void fraction model. They are
also used in the stratification model to determine oxidic/metallic pool configuration. The physical
properties of individual species used in TOLBIAC-ICB are mainly those defined in the frame of the
ECOSTAR European project (Journeau, 2003) for the densities, and also in the CORPRO data base of
CEA (Piluso, 2003) for specific heat, conductivity and surface tension.
The density of the melt mixture is obtained by a weighting in volume, whereas the heat capacity,
conductivity and surface tension are obtained using weighting in mass. The dynamic viscosity of the
melt is calculated following the method proposed by (Seiler & Ganzhorn, Viscosities of corium
concrete mixtures, 1997), (Sudreau, Ramacciotti, Cognet, Seiler, & Journeau, 2000) for mixture with
silica above liquidus temperature and (Ramacciotti, Journeau, Sudreau, & Cognet, 2001). The Urbain
model is used depending on the mole concentration of the species in the mixture. The species are
divided into network former (SiO2), modifiers (CaO, FeO, MgO, UO2, ZrO2) and amphoterics
(Al2O3, Cr2O3, Fe2O3, NiO). In case of a metallic layer, the Andrade model is used.
NEA/CSNI/R(2016)15

At time step n, the data for the considered melt are the mass of each species, and the temperature of
the melt for a homogeneous melt, or the interfacial temperature for a stratified melt. The basic
hypothesis is that thermodynamic equilibrium between the species occurs. An instantaneous
equilibrium is supposed. The resulting distribution of masses between solid and liquid phases after
equilibrium is given using pseudo-binary diagrams (default) or by GEMINI2 (optional)
In case of water aspersion, both radiation and convection with water are taken into account. The
convective heat transfer coefficient is derived from the boiling curve of water.
Melt ejection models are implemented in the code. The PERCOLA model is derived from the
PERCOLA experiments performed at CEA-Grenoble (Tourniaire & Seiler, 2004). In this model, the
ejection rate mainly depends on the superficial velocity of the gas issued from the concrete ablation.
Due to the large span of the reactor pit, the upper crust is supposed to float on the corium pool (no
sticking to the sidewalls). However, the code user has to define the number of holes in the crust per
square metre, the size of the holes and the crust porosity. The simpler Ricou and Spalding (Ricou &
Spalding, 1961) model may also be used, with one single user’s input coefficient.
A limitation of the debris bed height is taken into account, considering the dry-out limit. If the
dry-out limit is reached, equilibrium is supposed between the particles that are ejected, and the
particles that re-melt into the pool. The dry-out heat flux ϕdo is calculated using the correlation of
Lipinski (Lipinski, 1980):
ϕ do = 0.245.H LS .
[ρ L .ρV .g .D.ε 3 /(1 − ε )]0.5
[1 + (ρ V / ρ L )0.25 ]
2
Equation 7.1-40
where, HLS is the latent heat of water, ρL and ρV the density of liquid water and of steam, D the mean
diameter of the particles in the debris bed, and ε the porosity of the debris bed.
The maximum height emax of the debris bed is obtained when the dry-out heat flux corresponds to
the heat flux out of the pool ϕb, plus the internal heat of the debris Qd :
ϕdo = ϕb + emax. Qd Equation 7.1-41
Recently the water ingression model of Lomperski and Farmer (Lomperski & Farmer, 2007) was
also implemented.
In case of water quenching, the mass flow rate of steam formed due to the heat transfer to water is
calculated. The heat source is the heat flux out of the crust on one hand, and the decay power in the
debris bed above the crust on the other hand. Sensible heat from the water temperature Twat to the
saturation temperature is considered, together with the latent heat of water HLS. Finally, the steam mass
flow rate is given by the following relation:
Qsteam = ( Ssurf. ϕsurf + Qd.vold ) / [HLS + Cpeau.(Tsat – Twat)] Equation 7.1-42
Main code simplifications or limitations are the following:
• The TOLBIAC-ICB code assumes simplified academic (steady state) configurations: phase
macrosegregation, interface exactly at the liquidus temperature, either fully mixed or totally
stratified configurations.
NEA/CSNI/R(2016)15
• Ablation anisotropy of silica-rich concretes by the oxidic phase is modelled by a simple

experimentally-fitted coefficient increasing the lateral heat transfer coefficient.
• The PERCOLA model can be used to calculate the melt ejection but the code user has to
define the number of holes in the crust per square metre, the size of the holes and the crust
porosity.
• A lateral corridor can be considered.

4) Concerning the possible scenario
• Lateral melt-through is taken into account when a corridor is modelled. It is also possible to
consider a reduction of corium inventory during the calculation.
• Successive corium pours from the vessel to the reactor pit can be described on a simplified
way: corium poured from the reactor vessel is assumed to be spread over the whole pit area
instantaneously; poured corium is distributed between oxidic and metallic layers; no direct
heat transfer between the present water inventory and the poured corium falling through is
assumed to occur.
7.1.10. WECHSL code
The pool/crust interface temperature is the freezing temperature Tfreez, Tsol ≤ Tfreez ≤ Tfimmob, of
the oxide melt at which a stable crust starts to form. Steady-state heat conduction is assumed for thin
crusts, whereas in case of thick crusts the one-dimensional transient heat conduction is modelled.
Both the crusts and the liquid melt have the same composition. Between the crust and the
concrete a gas film exists.
The heat transfer between the melt bulk and the inside of the crust is determined by a discrete bubble
type heat transfer mechanism, with the driving temperature difference determined by the bulk
temperature and the freezing temperature, Tfreez, of the melt layer at the crust interface.
The modelling of the heat transfer from the melt to the concrete is based on the following ideas.
The heat transfer from the melt bulk to the concrete is characterised by processes forming boundary
layers at the melt pool surface facing concrete. The most important process which governs the heat
transfer phenomena is the release of large volume fluxes of gases from the decomposing concrete. If
the superficial velocity of the gases being released from the concrete is sufficiently high, a stable gas
film is formed between the melt and concrete. If the superficial gas velocity drops below a limiting
value, the heat transfer will be governed by a nucleate boiling type of discrete bubble gas release. As
the melt is intensively stirred by the gases released during concrete ablation, the bulk of each layer of
the melt is assumed to be isothermal. Because of the high thermal conductivity and the low viscosity
of the metallic phase, the temperature drop across the boundary layer of the metal layer is very small.
For an oxide melt with a high Prandtl number, Pr > 1, the temperature drop across the boundary layer
NEA/CSNI/R(2016)15
is quite significant. Due to cooldown of the melt, the temperature in the melt boundary layer facing the
concrete may drop below the liquidus temperature, which characterises the onset of solidification. The
initially thin viscous layer will grow and the solid volume fraction will increase with a further
temperature decrease. The strong variation of the viscosity with temperature in the temperature
boundary layers at the interfaces of the melt considerable reduces the heat transfer from the pool. A
correlation proposed in ( (Foit, Reimann, Adroguer, Cenerino, & Stiefel, 1995), (Foit & Miassoedov,
1995) is used for correcting the appropriate non-dimensional heat transfer parameter, i.e.
 µ 
n

Nu = Nu0 0.645 i  + 0,356, n = − 2 Equation 7.1-43
  µb  
Nu is the corrected Nusselt number and Nu0 the constant property solution. The viscosity µi is
the viscosity at the interface temperature, while µb is evaluated at the bulk temperature. The most
pronounced rheological changes with temperature occur at temperatures for which the solid volume
fraction, fimmob, passes through the critical range of 50-80 vol.%. This temperature corresponds to
the freezing temperature, Tfreez defined above in Section 7.1.10.1.1.
7.1.10.2.2. Heat convection between oxide and metal layer in a stratified pool

The viscosities of the oxide melt below the liquidus temperature are assumed to follow the modified
Pinkerton-Stevenson correlation. The results obtained with the modified correlation are in good
agreement with the viscosity measurements performed for the melt used in the KATS tests.
7.1.10.4. Thermochemical properties

For the oxide phase and the dispersed melt the solidus and liquidus temperatures are determined either
from a quasi-binary phase diagram or from a user input table. The solidus and liquidus temperatures of
the metallic phase in WECHSL are calculated from the chromium-nickel-iron, zirconium-iron and
silicon-iron phase diagrams.

The only coolability model is used in WECHSL is the film boiling model.

• The current level of validation (Foit, Reimann, Adroguer, Cenerino, & Stiefel, 1995) against
experimental data ACE, SURC, BETA, COMET-L, VULCANO, MOCKA, CCI test series
(see Section 2 for references) would appear to be good enough to justify the use of the MCCI
codes such as WECHSL for risk assessment studies, provided the uncertainties in the
predicted results are taken into account.
• No water ingression or melt eruption models
• Non-axisymmetric cavity shape (e.g. with a lateral corridor) cannot be described adequately.
NEA/CSNI/R(2016)15
7.2. Appendix on comparison of code validation status and uncertainties
For the most important phenomena the individual validation status is detailed (incl. the conclusions
e. g. which phenomena are adequately validated and which require more validation) and individual
uncertainties are outlined.
7.2.1. Concrete ablation and 2D cavity profile
7.2.1.1. COCO
The results of the comparison between COCO code and COTELS D-6 test and NEA-CCI-3 are shown
here as examples.
A core debris simulant (stainless steels in case of the D-6) was melted in an electric heating
furnace and it dropped into the concrete vessel set up in the pressure vessel. The concrete vessel was
equipped with a high frequency induction heating coil surrounding it to heat up the melt. The
axisymmetric concrete vessel was made of basalt concrete. A number of thermocouples were installed
to detect the ablation front. The input to the high-frequency induction coil was about 80 kW and the
heating efficiency is 30%. Therefore, the heating rate of the melt was around 24 kW. The test
continued 130 minutes.
Figure 7.2-1 left shows the observed result of a concrete vessel after the test. There was an ingot
of stainless steel under the glass-like clinker. The concrete was ablated about 160 mm downward and
55 mm laterally. Figure 7.2-1 right shows the calculated ablation profile. In the calculation, a stratified
layer model of metal was used, as the difference in density is large for metal and concrete. Since the
heat transfer coefficient of metal layer in the calculation is very large, the downward ablation is larger
than side wall ablation. The calculated profile is similar to the observed one.
Figure 7.2-1: Stratification of molten oxide and metal after test D-6 of COTELS project
The position of ablation fronts are shown in Figure 7.2-2. The plots are test results and the lines
show the maximum of the calculated vertical and horizontal ablation depth. The boxes at 130 min are
the depth measured after teardown. The calculated ablation depth fits to the test results. (As the
ablation was not uniform, the test results are scattered).
NEA/CSNI/R(2016)15
Test （Down) Test （Side）
COCO (Down) COCO (Side)
200
150
Ablation Depth(mm)
100 Post-test
50
0
0 50 100 150
Time(min)
Figure 7.2-2: Comparison of ablation depth in D-6 test of COTELS project
0.6 COCO
Initial
North Side Wall
0.4 South Side Wall
0.2
Y(m)
-0.2
-0.4
-0.6
0 0.2 0.4 0.6 0.8 1
X(m)
Figure 7.2-3: Ablation profile after CCI-3 test in NEA MCCI Project
The CCI-3 experiment of the NEA MCCI Project is a two-dimensional corium concrete
interaction test. The cross-sectional area of the cavity was 50 cm (distance between the tungsten
electrodes) x 50 cm (distance between the concrete sidewalls). The basemat and sidewalls were made
of siliceous concrete. The initial melt contained 15 wt.-% concrete components. The input to the direct
Joule heating electrodes was about 120 kW and the test continued for 146 minutes.
The fundamental principle of deriving the scaling law for simulating the rectangular cavity by
axisymmetric code is that the following ratios are the same in the test and the analysis:
1. Concrete sidewall surface area/basemat surface area
2. Total surface area of concrete/melt mass
3. Input power/melt mass
NEA/CSNI/R(2016)15
To satisfy these relations, the scaling laws (values in the analysis/values in the test) are:
• Coordinate in vertical direction, ablation depth = 1.
• Intensive variables (temperature, volumetric heat rate) = 1.
• Extensive variables (mass, total power) =(r/y) π = π.
where r: cavity radius in calculations (r=distance between sidewalls for CCI-3), y: =distance
between electrodes.
The calculated ablation profile after the test is shown with the test results in Figure 7.2-3. The
north and south side wall ablations in the test differ a little in the depth. The calculated profile is close
to the test results.
7.2.1.2. CORQUENCH
In terms of the ability to capture 2D cavity erosion behaviour, the code has been compared to NEA
MCCI tests CCI-2, -3, -4, and -6. For tests conducted with high gas content concretes such as
limestone-common sand (i.e. CCI-2 and -4), the code does a reasonable job of predicting isotropic
ablation behaviour, but does not perform as favourably for siliceous concrete tests (that exhibit strong
lateral ablation behaviour) when default heat transfer correlations are used. However, the code does
seem to perform better when the gas film breakdown modelling option is employed. This assumption
leads to better agreement for siliceous concrete tests.
7.2.1.3. COSACO
The validation results show an adequate consistency with the experimental results regarding concrete
ablation behaviour and melt temperature (see Section 7.2.2.3). As an example, the ablation depth
history as calculated for the MACE M4 test and CORESA 2.1 test are given in Figure 7.2-4 in
comparison to the experimental data. Furthermore, the results of various benchmark calculations show
good consistency with other MCCI codes.
On this basis, it is concluded that the general treatment of thermochemistry in COSACO and the
thermal-hydraulic models specifically implemented for the oxidic and metallic melts are well-suited to
analyse the characteristics of the MCCI process.
0.1
0.35
Case '1' 0.09 Horizontal-COSACO

Case '2' Exp. Vertikal
0.3 B-Array Exp. Horizontal
C-Array 0.08
Center Line Array
D-Array
0.25 E-Array 0.07
Mass Spectrometer
Ablation Depth, m
Ablation Depth, m
0.06
0.2
0.05
0.15 0.04
0.03
0.1
0.02
0.05
0.01
0
0
0 50 100 150 200 250 300 350
-25 0 25 50 75 100 125 150 175 200 225
Time, s
Time, min
a) MACE M4 b) CORESA 2.1

Figure 7.2-4: COSACO-validation: comparison of calculated and measured evolution
of the concrete ablation
NEA/CSNI/R(2016)15
7.2.1.4. MAAP
ACE tests and SURC-4 test are 1D experiments in which ablation can occur only in the downward
direction, while BETA tests are 2D experiments in which the ablation can occur in both the sideward
and downward directions. In general, ablation rates predicted by MAAP model compared better for the
ACE and SURC-4 tests than the BETA tests. Figure 7.2-5 shows an example of the ablation depth
calculated by MAAP compared with the ACE-L2 test data:
Figure 7.2-5: MAAP-validation: Comparison with ablation depth data for ACE-L2
Ablation rates measured in the BETA tests were appreciably higher than those in ACE and
SURC-4 tests. Although the BETA tests allow the ablation to occur in both directions, post-test
examinations revealed the ablation in the downward direction is much larger than that in the sideward
direction. It was postulated that the high ablation rate in the downward direction was caused by
spalling in the concrete (Epstein, 2004). The relatively smaller sideward ablation may be caused by the
formation of a foamed layer between the corium and the sidewall which separated the corium from
contact with the wall. It was difficult for the MAAP model to simulate this type of behaviour, since a
spalling model was not implemented (such a phenomenon may not be prototypic in a nuclear plant set
up), and the foamed layer model was not included.
The MCCI and corium coolability models in the latest released version MAAP5.02 have been
compared with CCI tests, which are the latest 2D dry and wet cavity tests, performed by Argonne
National Laboratory under the sponsorship from NEA (Farmer, Lomperski, Kilsdonk, & Aeschlimann,
OECD MCCI Project: 2-D Core Concrete Interaction (CCI) Tests, Final Report, 2006). Two in the
total of six tests have been selected for comparisons: CCI-2 and CCI-3, where the concrete for the
CCI-2 test is limestone-common-sand (LCS) concrete, and the concrete for CCI-3 test is European
siliceous concrete. For each test, MAAP5 calculations were performed with two sets of inputs for the
convective heat transfer coefficients and exponent (see section 7.1.6), as shown in the following table:
Table 7.2-1: Inputs of downward and sideward heat transfer coefficients

Tests Set 1: Varying Heat Transfer Coefficients Set 2: Constant Heat Transfer Coefficients
hd W/(m2K) hs W/(m2K) n hd W/(m2K) hs W/(m2K) n
CCI-2 3 500 3 500 2.75 300 300 0
CCI-3 1 000 3 500 2.75 80 300 0
The inputs for the set 1 assumes the heat transfer coefficients from the molten corium to surround
crusts vary with the solid fraction in the corium, while the inputs for the set 2 assumes the heat transfer
coefficients are constant.
NEA/CSNI/R(2016)15
Figure 7.2-6 shows comparisons of ablation depths between MAAP calculations and the
experimental data. The calculated ablation depths are in reasonable agreement with the experimental
data.
(a) CCI-2 Comparisons (b) CCI-3 Comparisons

Figure 7.2-6: MAAP-validation: Comparison of concrete ablation depths for CCI tests
7.2.1.5. MEDICIS
In the parametric approaches for the heat transfer coefficients used by IRSN and GRS, the heat
transfer is referred to the impact of an interface structure between melt and concrete (either by
introducing a significant thermal resistance hslag between melt and concrete as function of interface
angle or by using effective heat transfer coefficients). Here the dependence of this thermal resistance
versus the interface orientation determines directly the ablation profile and the final cavity shape. For
the several experiments with an LCS type of concrete a uniform thermal resistance along the interface
seems the best estimate whereas for several experiments with siliceous concrete it is suggested to
account for the anisotropy of the 2D ablation observed in several experiments by using an increased
thermal resistance in the downward direction (see Figure 7.2-7 and Table 7.2-2).
Table 7.2-2: Comparison of heat transfer coefficients in the approaches of GRS and IRSN
Concrete type model type heat transfer at bottom heat transfer at lateral
interface (W/m2/K) interface (W/m2/K)
LCS concrete IRSN “no crust model” 300 300
GRS global heat transfer bulk to interface 200 200
siliceous type IRSN “no crust model” 80 300

concrete
GRS global heat transfer bulk to interface 80 300
NEA/CSNI/R(2016)15
Figure 7.2-7: Proposal for the usage of effective heat transfer coefficients in MEDICIS by GRS
This simplified approach does not track the existence of a crust at the interface, but rather gives
an approximation of the overall effective heat transfer coefficient. Detailed mechanisms of the heat
transfer are not identified. It is presently not clear how thermal resistances vary with time, i. e. with the
melt bulk conditions (viscosity, superficial gas velocity, decay power) in the pool. There are
indications that effective heat transfer coefficients may be nearly constant throughout the experiments
or slowly decreasing, except for the initial phase in the experiments. The 2D heat flux distribution is
then mainly controlled by the 2D relations of the heat transfer coefficients to the different interfaces.
Most experiments are overestimated with regard to the integral eroded volume but are met
satisfactorily with regard to maximum erosion depths. The validation of the 1D experiment MACE-
M3b performed by GRS shows that the 2D set-up of the model can also be applied for a 1D scenario
(Spengler, 2012).
In IRSN approach the heat transfer to the pool/concrete interface is thus evaluated as a
combination of heat convection within the pool and of a thermal resistance at the pool/concrete
interface, without any stable crust. The anisotropy in heat flux distribution observed in tests with
siliceous concrete causing a high corium viscosity is explained only by the non-uniform profile of
thermal resistance along the pool/concrete interface. The convective heat transfer coefficient is kept
constant and does not depend on the interface orientation. This view is supported by the first CLARA
results on the heat convective coefficient distribution in a bubbling pool. These results
(see Figure 7.2-8) show indeed that the ratio of lateral to axial convective heat transfer coefficients
stays near one at high viscosity and exceeds largely one only at low viscosity where the convective
heat transfer coefficient becomes high and cannot impact on the effective heat flux distribution along
the pool/concrete interface because of the much lower conductance of the interface itself.
NEA/CSNI/R(2016)15
Lateral to bottom heat transfer coefficient ratio

vs Lateral to bottom superficial gas velocity ratio
Hlat/Hbot
3.50 µ=1mPa/s
µ=9mPa/s
3.00
µ=25mPa/s
2.50
µ=100mPa/s
2.00
µ=1000mPa.s
1.50
µ=10000mPa/s
1.00
0.50
0.00
0.1 1 10
Jglat/Jgbot
Figure 7.2-8: Ratio of lateral to axial convective heat transfer coefficients versus
pool viscosity deduced from the small-scale CLARA tests
7.2.1.6. SOKRAT/HEFEST
Since SOCRAT/HEFEST if a fully 2D code, it should correctly predict concrete ablation both in
1D and in 2D geometry. 1D experiments of SURC (1-4), SWISS (1-2) and ACE (L2, L5, L6, L8)
series were chosen for the validation matrix. Among the 2D experiments COMET (L2, L3) and BETA
(V3.2, V3.3) are of particular interest due to the different behaviour of the concrete of different types.
It can be noted that even in 1D experiments the fully 2D approach can be useful due to the effects of
the crucible walls and of inhomogeneous heating.
The analyses of the results of the MCCI oxidic tests (NEA-CCI-2,-3,-4 and -6; VULCANO VB-U5,
-U6) show that there can be anisotropic ablation rates depending on the composition of the concrete.
For silico-calcareous and siliceous concrete, the radial ablation rate is larger than the radial one. As
this phenomenon is not well understood, a parameter is introduced in TOLBIAC-ICB in order to
impose a higher heat transfer coefficient in the radial direction than the axial one. The value of this
parameter is advised from 1 (for calcareous concretes) to 3 (for siliceous concretes). A result of a
calculation performed with TOLBIAC-ICB on VULCANO VB-U5 is presented on Figure 7.2-9.
NEA/CSNI/R(2016)15
0,5
Base
0,4
CO 3
initial cavity
0,3
height (m)
experimental
cavity 45°
0,2
experimental
cavity 90°
0,1 experimental
cavity 135°
0
0 0,1 0,2 0,3
width (m)
Figure 7.2-9: 2D cavity profile in VULCANO–VBU5. Results obtained with TOLBIAC-ICB, base
case (blue curve) and with a ratio 3 between the radial heat transfer coefficient and the axial one
(green curve).Comparison with the experiment
7.2.2. Melt temperature

7.2.2.1. COCO
The results for CCI-3 are shown here as example. Figure 7.2-10 shows the temperature transients of
the melt in the measurement and COCO calculation. The temperature by COCO agrees well with the
measurements.
2500
2000
Temperature(℃)
1500
1000 COCO
Test WCL-3
Test WNW-2
500
Test WSE-1
0
0 30 60 90 120 150 180
Time(min)
Figure 7.2-10: Melt temperature in CCI-3 Test
The calculated concrete concentration and melt temperature are plotted at every time step on the
phase diagram which is used for the COCO analysis as shown in Figure 7.2-11. The plotted point
horizontal value means the averaged concrete fraction for melt and crust. As shown in Figure 7.2-11,
the melt is in a two phase regime of solid and liquid through the test duration.
NEA/CSNI/R(2016)15
Calculated State Solidus Liquidus

3000
2800
2600
0 min.
Temperature[K]
2400
2200 115 min.
2000
180 min.
1800
1600
25 min.
1400
1200
0.0 0.2 0.4 0.6 0.8 1.0
Concrete fraction[-]
Figure 7.2-11: Phase diagram and calculated corium/concrete condition
7.2.2.2. CORQUENCH
Melt temperature has been measured in all experiments against which CORQUENCH has been
validated, and the temperature predictions by the code have been compared with the data. In general,
the code seems to reasonably capture the magnitude and trend of the test data.
7.2.2.3. COSACO
As for the concrete ablation, the melt temperature as calculated by COSACO has been verified against
a large number of experiments, as e. g. shown in Figure 7.2-12 for the MACE M4 and CORESA 2.1
tests.
2000
Metal-COSACO
1900 Experiment
1800
Temperature, °C
1700
1600
1500
1400
1300
0 50 100 150 200 250 300 350
Time, s
a) MACE M4 b) CORESA 2.1

Figure 7.2-12: COSACO-validation: Comparison of calculated and
measured evolution of the melt temperature
7.2.2.4. MAAP
Melt temperatures were compared between MAAP predictions and dry cavity tests (ACE, SURC and
BETA tests) mentioned in the previous section. The temperature calculated by MAAP is the average
temperature of the corium pool, while the temperature measured in the experiments is the temperature
in the molten pool. For dry cavity cases, the two temperatures do not significantly differ from each
other. Figure 7.2-13 shows an example of the comparisons between MAAP and the ACE-L2 test.
NEA/CSNI/R(2016)15
Figure 7.2-13: MAAP-validation: Comparison with temperature data
In general, the temperatures compared better than the ablation rates. Even the temperatures in the
BETA test series showed acceptable agreement with the experiment data.
Figure 7.2-14 shows the comparisons of the melt temperatures between MAAP5.02 calculations
and CCI-2 and CCI-3 data. As shown, the temperatures calculated with the constant heat transfer
coefficients have better agreement with the data than those calculated with the varying heat transfer
coefficients.

Figure 7.2-14: MAAP-Validation: Comparison of melt temperatures for CCI tests
7.2.2.5. MEDICIS
When using an effective thermal resistance at the melt/concrete interface as proposed by IRSN and
GRS it is observed that the long term melt pool temperature in the calculation is governed by the
effectiveness of heat transfer processes and conditions at the interfaces (including also the top
interface) and by the evolution of interface area. Validation calculations performed by IRSN consider
a detailed model for the evolution of a top crust on the corium pool, in which the interface temperature
between pool and upper crust is determined from a volumetric molten corium fraction equal to around
0.5. Typical calculation results are shown in Figure 7.2-15 and Figure 7.2-16. The agreement with
experimental data (slight underestimation of pool temperature in CCI-2 and overestimation at the
opposite in CCI-3) is rather good in spite of the simple and consistent assumptions used for the heat
transfer at the pool/concrete interface in both tests. This gives hints that the assumption of stable crust
at the melt/concrete interface with an imposed pool/crust interface temperature is not required to
reproduce experimental data. A main conclusion is that the pool temperature is only weakly linked to
NEA/CSNI/R(2016)15
the melt liquidus temperature; the link between the pool temperature and the liquidus temperature is
suspected only through the liquid fraction in the melt, which equilibrates to a large enough value to
permit convective heat transfer to the pool interfaces. A second conclusion is that the heat transfer
between melt and concrete may be explained without considering the evolution of a crust along the
pool/concrete interface.
Tliq
Tliq Fliq
calc. Tpool
Texp
Texp
calc. Tpool Fliq
Figure 7.2-15: MEDICIS calculation for Figure 7.2-16: MEDICIS calculation for CCI-3
CCI-2 by IRSN by IRSN
All experiments noted in Section 7.2.1.6 except for COMET (L2, L3) and SWISS are used in
verification of the SOCRAT/HEFEST ability to predict the melt temperature. Those experiments,
where abrupt temperature changes due to chemical reactions were observed, are of special interest
here. It concerns the SURC series and some of the ACE experiments.
More abrupt temperature rises during chemical reactions are expected in HEFEST simulations
due to the limitations of the chemistry model, especially in those experiments where Zr was added to a
silica-rich melt like in SURC-4.
Typical calculation results are shown in Figure 7.2-17 and Figure 7.2-18. The calculation of CCI-2 is
performed considering that all chromium is oxidised and the initial temperature is taken as the liquidus
temperature of this initial mixture. The results show that the temperature is well predicted by the code
(except at the early beginning of the experiment).
The calculation of CCI-3 shows a short decrease of the melt temperature in the experiment. This
is not reproduced by TOLBIAC-ICB calculation which over-predicts the melt temperature. This is due
to the assumption that the temperature at the interface between the melt and the crust is equal to
Tliquidus. In case of a silico-calcareous (or a siliceous) concrete, this hypothesis gives higher melt
temperature.
NEA/CSNI/R(2016)15
2800
TOLBIAC-ICB
2600
CCI2
temperature (K) 2400
2200
2000
1800
1600
0 60 120 180 240 300
time (min)
Figure 7.2-17: TOLBIAC-ICB calculation for CCI-2 by CEA
2600
TOLBIAC-ICB
2400 CCI3
2200
temperature (K)
2000
1800
1600
0 20 40 60 80 100 120 140 160
time (min)
Figure 7.2-18: TOLBIAC-ICB calculation for CCI-3 by CEA

7.2.3. Gas release
7.2.3.1. CORQUENCH
CORQUENCH predictions of melt superficial gas velocity have been compared with data obtained in
ACE/MCCI tests L2, L5, L6, and L8. The code seems to reasonably capture the magnitude and trend
of the gas release in the tests. However, the timing of peaks associated with ablation through different
zones in the basemat that contain varying amounts of gas (i.e. concrete-metal inserts vs. pure concrete)
do not always line up due to the fact that the code does not precisely predict the time at which a given
boundary is reached.
7.2.3.2. COSACO
The gas release is directly linked to the concrete erosion history. Thus the validation work has mainly
focused on concrete erosion.
NEA/CSNI/R(2016)15
7.2.3.3. MAAP
Gas release is calculated by MAAP models, but has not been compared with the experiments.
7.2.3.4. MEDICIS
The emphasis on validation work has recently been on the 2D ablation behaviour. The gas release
coming from the decomposition of the concrete is regarded only as subsequent effect of (2D) concrete
ablation. However, for MACE-M3b the phenomenological models for gas releases of the species CO,
CO2 and H2 are successfully validated by GRS. Based on this it is assumed that the models for metal
oxidation in MEDICIS give realistic predictions.
The same list of experiments as in Section 7.2.1.6 is used for the verification of gas release simulation.
Here, an influence of Zr-oxidation on the gas flow rate and the composition is of special interest, so
the focus is mainly on the SURC series. Faster and shorter H2 and CO releases due to Zr oxidation
than what was observed in the experiments are expected.
The gas release coming from the decomposition of the concrete is regarded only as subsequent effect
of (2D) concrete ablation. It depends on the models for metal oxidation.
7.2.4. Oxidation reactions
7.2.4.1. CORQUENCH
CORQUENCH is able to calculate gas phase chemical reactions between metallic melt constituents
and sparging concrete decomposition gases (CO2 and H2O), as well as condensed phase reactions
between metallic constituents and concrete decomposition products (i.e. Zr and SiO2). Tests for which
CORQUENCH has been applied that included metals oxidation as well as instrumentation to detect
gas speciation included ACE tests L2, L6, and L8; MACE test M1b, and SURC tests 1 and 2;
see Table 4.3-2. The melt temperature predictions that include the effect of exothermic metal oxidation
reactions have been compared with test data (see 7.2.2.2), but at this point gas composition has not
been compared with the results of these experiments. This shortcoming will be addressed in future
work.
7.2.4.2. COSACO
In COSACO, the thermo-chemical equilibrium in the melt pool layer based on the minimisation of the
Gibbs enthalpy is calculated during each time step. Therefore, all possible chemical reactions and
phase transitions are considered. This includes especially the oxidation reactions as a source of
burnable gases released into the containment.
7.2.4.3. MAAP
Oxidation reactions are calculated by the MAAP chemical reaction model. However, the compositions
calculated by MAAP have not been compared with experiments.
7.2.4.4. MEDICIS
The gas release which comes from the decomposition of the concrete and which impacts the oxidation
of metals is regarded as subsequent effect of (2D) concrete ablation. It has recently been of second
order priority in the validation work by GRS and IRSN, see Section 7.2.3.4.
Again those experiments, where Zr was present in the melt, are of main interest here. Those are SURC
(1-4) and ACE (L2, L6, L8). Later we expect to include tests with Cr in the melt (most likely MACE
series).
NEA/CSNI/R(2016)15
The melt composition is calculated by GEMINI considering the chemical reactions between the
different species.
7.2.5. Top flooding
7.2.5.1. CORQUENCH
CORQUENCH has been compared with both early flooded (MACE M1b, M3b, and M4; CCI-6) as
well as late-flooded (CCI-2 and -3) experiments. As part of these analyses, the code predicts crust
growth by water ingression, particle bed formation by melt eruptions, as well as the potential for crust
anchoring during the experiments. The code has been compared to the melt-water heat flux data
logged during these tests, as well as the debris morphology determined as part of the post-test
examinations. In general, the code seems to reasonably reproduce the initial crust formation phase as
well as the heat fluxes and debris configurations determined as part of the experiments. Currently, a
modelling shortcoming is the ability to predict the transient nature (i.e. on vs. off) nature of melt
eruptions.
7.2.5.2. COSACO
Currently, only a simplified heat transfer model to a water pool above the melt is included in
COSACO. No code validation for top flooding has been performed.
7.2.5.3. MAAP
Figure 7.2-19 shows the comparisons of the heat fluxes from corium pool to water between
MAAP5.02 calculations and CCI-2 and CCI-3 data:

Figure 7.2-19: MAAP-Validation: Comparison of heat fluxes from corium to water for CCI tests
For CCI-2 test, MAAP5 calculates much lower heat flux from corium to water in the first
30 minutes after water injection. This is mainly because MAAP5 lacks a sophisticated bulk cooling
model to calculate the intensive heat transfer between the corium and water in the short initial
transient, when water is added on top of a corium pool in an initially dry cavity. For CCI-3 test,
MAAP5 calculates a more reasonable heat flux compared to the experiment. It seems CCI-3 test does
not exhibit a strong bulk cooling effect like CCI-2 test, but the heat flux measurement in this test
suffers from an uncertainty of the upper surface area of the corium which is quenched by water.
Figure 7.2-19 shows the heat fluxes estimated with the upper and lower bounds of the top surface area.
MAAP calculation is between the two heat fluxes and closer to the one estimated with the upper
bound of the top surface area.
NEA/CSNI/R(2016)15
The MAAP water ingression model was developed by Epstein (Epstein, 2004) based on earlier
work by Lister (Listner, 1974) on cracking solid rock. A variant of the model has been validated by
Lomperski and Farmer (Lomperski & Farmer, 2007) to compare against measurements obtained from
the SSWICS tests. The model has successfully predicted the levelling-off of the heat flux from the
corium to overlying water when the concrete slag mass fraction in the corium pool exceeds a critical
fraction. With adjustment of a parameter in the model, the heat fluxes can agree with the experiment
measurements within the data error range. There is no consideration of possible effects which evolved
gas “ventilation” might have on the crust permeability in the MAAP water ingression model.
Therefore, comparisons regarding the effect of ventilation on water ingression cannot be made right
now.
The melt eruption model in MAAP5.02 is based on the Ricou-Spalding entrainment correlation
(Ricou & Spalding, 1961). Epstein (Epstein & Zhou, 2012) has compared the entrainment rate
predicted by the Ricou-Spalding correlation with the experiments performed by Tourniaire et al.
(Tourniaire B. , Seiler, Bonnet, & Amblard, 2006) using simulant material. According to Epstein, an
entrainment coefficient of 0.08 generates an entrainment rate comparable with the Tourniaire
experiment for thin crust cases. There has been no consideration in MAAP on the density (number
concentration) of melt eruption sites in the upper crust. Therefore comparisons regarding the density
of the eruption sites cannot be made right now.
7.2.5.4. MEDICIS
After onset of quenching, cooling mechanisms are described in the present version of ASTEC
V2.0/MEDICIS by very simple models of water ingression and melt eruption:
Water ingression is described using a simple model function of an imposed permeability value;
this model does not depend on the concrete oxide fraction in corium and thus does not reproduce the
higher heat flux extracted in the early MCCI phase at low concrete fraction as observed in SSWICS
tests (Farmer, et al., A Summary of Findings from the Melt Coolability and Concrete Interaction
(MCCI) Program, 2007); in the longer term phase, water ingression is much less efficient and becomes
almost negligible compared to the impact of melt eruption, so that the crudeness of the model has
limited consequences on quenching predictions;
Melt eruption is described using derived from that of an older CORQUENCH version (Farmer,
2001) based on Ricou’s correlation (Ricou & Spalding, 1961) for the evaluation of the corium
entrainment rate. A proportionality factor of Ricou’s correlation equal to around 0.08 permits to
reproduce roughly the averaged corium entrainment rate in MACE3B and CCI2 tests.
Recently, an additional model to account for a simplified treatment of the top flooding was added
to MEDICIS by GRS. This model is to be used in combination with the model options for effective
heat transfer coefficients at the boundaries of the melt pool and considers a simplified top crust
evolution and a boiling heat transfer at the interface between water and the top crust. The model was
successfully validated for MACE M3b and M4 as well as for CCI-2, -3 and -6. It was shown that
assuming solidus as interface temperature between melt and top crust the experimental data (heat
fluxes to the water) for these experiments can be closely approximated with a heat transfer coefficient
between melt and crust of about 200 W/(m2 K) or larger, which is of similar magnitude as the heat
transfer coefficients found and used between melt and concrete. This similarity is coherent with the
assumed interface temperatures between melt and top crust (interface temperature is defined as
solidus), as the decomposition temperature of the concrete of about ~ 1 500 to 1 600 K is not far from
typical solidus temperatures of the mixtures in the longer term, when some substantial amount of
concrete is mixed with the melt. However, the initial period after flooding showing intense interaction between melt
and water and corresponding peak heat fluxes cannot be resolved adequately with this simplified
model.
NEA/CSNI/R(2016)15
It is planned to use the SWISS-2 and COMET-L3 experiments to verify the effect of top flooding,
especially to check heat removal coefficients that are proposed to be used in simulations. Later it is
planned to consider also the experiments from the MACE series.
In case of water aspersion, the upper crust surface temperature becomes lower than the concrete fusion
temperature and the reactor vessel fusion temperature and then no ablation occurs above the melt
level. Both radiation and convection with water are taken into account. The convective heat transfer
coefficient is derived from the boiling curve of water.
Melt ejection models have been implemented in the code. The results of these models are in good
agreement with the results of MACE3B and CCI2 tests (Tourniaire, Boulin, & Haquet, 2015).
The water ingression phenomenon is described using a model derived from CROQUENCH
which depends on the concrete oxide fraction of corium. As this fraction increases, the water
ingression phenomenon becomes less efficient and its impact on the heat transfer calculation is
negligible compared to the melt ejection phenomenon.
7.2.6. Top crust formation
7.2.6.1. CORQUENCH
CORQUENCH contains a full set of models for calculating incipient crust formation and growth at the
melt-water interface. As part of the validation process, efforts have been made to predict end-of-test
crust thicknesses to that inferred during post-test examinations. In general, the code seems to
reasonably estimate these crust thicknesses, but there is lingering uncertainty as to whether the crust
thickness after the experiment is indicative of that present when the test was conducted.
7.2.6.2. COSACO
The temperature of the top of the melt surface is calculated by COSACO also including crust
formation in case of phase segregation. No specific validation on this item has been performed.
7.2.6.3. MAAP
The top crust thickness predicted MAAP has not been compared with experimental data.
7.2.6.4. MEDICIS
Some calculated data were evaluated by GRS for experiments involving top flooding (see Section
7.2.5.4) and using the simplified option in MEDICIS for top surface heat transfer under flooded
conditions. The final crust thicknesses calculated for the experiments are in the one-digit centimetre
regime. Generally, in the experiments slightly thicker final top crusts were observed. In view of the
uncertainties related with interface temperature, thermal conductivity of the crust, the ideal condition
of a plane crust in comparison to the real processes and the problem of how to evaluate an
experimental crust thickness at the time of power shutdown from post-test examinations the agreement
is satisfactory.
Since crust is treated automatically in the code, its thickness is derived from the upward heat flux, so
we can estimate the accuracy of top crust predictions by comparison of heat flux obtained in
simulations with one observed experimentally. Stable thin crust can transmit the same heat flux as
thicker unstable one, so we focus our attention on heat transmission rather than on actual crust
thickness, since not much experimental data about crust parameters is available anyway.
NEA/CSNI/R(2016)15
7.2.7. Effect of iron rebar

7.2.7.1. CORQUENCH
CORQUENCH is able to evaluate the effect of rebar on cavity ablation behaviour, either by
homogenising iron into the concrete composition, or by treating discrete zones in which the iron is
concentrated. However, the code has not yet been validated against this type of test as experiments of
this type are just now emerging.
7.2.7.2. COSACO
Iron rebar in the concrete has recently been implemented in the COSACO code. This includes the
addition of the steel components to the eroded concrete mass and an updated concrete decomposition
enthalpy. No other effects as e.g. heat conduction into the concrete are modelled.
7.2.7.3. MAAP
The effect of iron rebar on chemical reactions and gas releases is simulated by the METOXA chemical
reaction model in the code. However, since MAAP does not have a material stratification model,
frozen metal at the bottom or at the top cannot be mechanistically simulated by the current version of
the code. Therefore, no comparison on that aspect has been made.
7.2.7.4. MEDICIS
Iron rebar is considered in MEDICIS by providing specific data for the concrete in the input data set
(e. g. including a mass fraction of iron in the composition of the concrete and a different latent heat of
decomposition). No other impact of iron rebar is up to now considered in the models. The
phenomenon is not validated since there had not been any dedicated MCCI experiments on this
phenomenon in the past.
Iron rebar can be considered either by providing its exact 2D structure (which is extremely time
consuming since it requires very large meshes) or by declaring steel to be one of concrete component.
The first approach can be used for small-scale experiments, but has small practical value. The
verification of the second approach requires large-scale experiments, where the size of a single iron
element is much less than the size of the concrete block and there should be a large number of iron
elements, like in real NPP.
Iron rebar is considered in TOLBIAC-ICB by providing specific data for the concrete in the input data
set. This way, a mass fraction of iron in the composition of the concrete and a different latent heat of
decomposition is taken into account.
7.2.8. Stratification evolution and effect on heat flux distribution
7.2.8.1. CORQUENCH
As noted previously, CORQUENCH assumes oxide and metal phases are well mixed under all
conditions, and so no explicit attempt has been made to model stratification in experiments.
7.2.8.2. COSACO
No dedicated stratification criterion is implemented in COSACO. The usage of layered or mixed melt
configurations is predefined by user input.
In layered melt mode two configurations are considered. Initially the density of the core oxide
rich oxidic melt layer is higher than that of the metallic layer. Thus the oxidic layer will be located
below the metallic one. A layer of light oxides, mainly from the concrete eroded by the metallic melt,
is calculated on the top. With the ongoing addition of light concrete oxides into the oxidic melt layer
NEA/CSNI/R(2016)15
its density decreases. When the metal layer density is reached inversion of the layering pattern is
assumed. This yields a configuration where the oxide is located on top of the metallic melt layer. The
top slag layer of light oxides is then dissolved in the oxidic melt.
7.2.8.3. MAAP
Stratification of immiscible materials has not been considered in MAAP MCCI models. Therefore,
related comparisons cannot be made. The models may be subject to change in the future.
7.2.8.4. MEDICIS
Ablation of concrete by steel melts in a static stratified configuration is validated for several
experiments from the BETA and COMET test series with an alumina-iron based melt. Reliable
experimental data for a dynamic evolution of melt pool configuration (homogeneous <-> stratified)
from experiments with prototypical oxide and steel melts and prototypical power distribution in oxide
and metal are not available.
However the stratification criterion and the correlation used for heat transfer between oxide and
metal layers might be deduced from the results of simulant experiments:
1) The criterion for the switch from homogeneous to stratified configuration is deduced from the
onset of entrainment measured in BALISE experiments (Tourniaire, Seiler, & Bonnet, 2003).
2) The convective heat transfer between oxide and metal layers is evaluated from Greene’s
correlation (Greene & Irvine, Heat transfer between stratified immiscible liquid layers driven by gas
bubbling across the interface, 1988), which is in agreement within a factor of 5 with most
experimental data with simulant melts (Werle and ABI data).
Experimental data on dynamic behaviour of melt layers is unavailable now, and criteria used in
SOCRAT/HEFEST code (temperature, time and component mass ratio) seem to be of little use in melt
stratification during MCCI. Therefore we are not going to include this into present verification matrix.
When we introduce superficial gas velocity as another criterion for melt stratification, we will consider
this phenomenon in better detail.
Ablation of concrete by steel melts in a static stratified configuration is validated for several
experiments from the BETA and COMET test series with an alumina-iron based melt. Reliable
experimental data for a dynamic evolution of melt pool configuration from experiments with
prototypical oxide and steel melts and prototypical power distribution in oxide and metal are not
available.
7.2.9. FP/aerosol release
7.2.9.1. CORQUENCH
CORQUENCH does not possess the ability to model fission product release, and so validation against
fission product release data is not possible.
7.2.9.2. COSACO
No fission product or aerosol release model is implemented in COSACO.
7.2.9.3. MAAP
Fission product release by formation of volatile compounds is predicted by the MAAP chemical
reaction model. However, there is no comparison of the predicted fission product release with
experiments.
NEA/CSNI/R(2016)15
7.2.9.4. MEDICIS
The FP release model in MEDICIS is tested vs. some experiments of the ACE test series. The
uncertainties of the model for FP release are quite large. Depending on the chemical composition of
the corium melt the uncertainty is in the range of approx. one or two orders of magnitude, but may be
even much larger for some specific conditions (composition, temperature).
The experiments of SURC and ACE series will be used in FP/aerosol release model verification. Since
the FP/aerosol release rates strongly depend on the temperature of the melt which can differ
significantly in the middle and near the wall due to high viscosity of the melt with silica and
immiscible materials, we expect a high level of uncertainty here. Another source of uncertainty will be
an inhomogeneity of FP distribution inside the real melt, which cannot be accounted for in current
model.
No FP/aerosol release model is taken into account in TOLBIAC-ICB.
7.2.10. Incubation period
7.2.10.1. COSACO
No dedicated modelling is currently implemented in the COSACO code.
7.2.10.2. MAAP
The model in MAAP5.01 assumes crust will always be formed between the molten pool and the
adjacent concrete. Crust growth and shrinkage is based on an energy balance in the entire thickness of
the crust. When the corium is first relocated from the vessel, contact with the relatively cold concrete
will immediately start the growth of a crust. However the crust model in MAAP assumes there is
always decay heat generation in the crust. The incubation period observed in certain experiments is not
modelled by the current MAAP model, since the crusts in the experiments are free of heat generation.
7.2.10.3. MEDICIS
The phenomena observed during the incubation period in several 2D MCCI tests with oxidic melt are
referred to transient processes acting at the melt concrete interface (transient formation and re-melting
of crusts eventually accompanied by transient inhomogeneous heating conditions). For such transient
processes no adequate models are available in MEDICIS.
The crusts that appear when the melt touches cold concrete are treated automatically in
SOCRAT/HEFEST code if they are thicker than one cell of the mesh. Their dynamics can be
evaluated in experiments by a change in the heat flux, but it seems quite unreliable, therefore this
phenomenon is not of primary interest in SOCRAT/HEFEST code verification.
During the incubation period during several 2D MCCI tests with oxidic melts, several transient
processes take place. Such transient processes like inhomogeneous heating of the melt are not
considered in TOLBIAC-ICB.
7.2.11. Remaining uncertainties
7.2.11.1. COCO
In the COCO code, the melt is assumed to be a Newtonian fluid and the heat transfer rate is calculated
with this assumption. However, the mixture of corium and concrete shows the characteristic of
NEA/CSNI/R(2016)15
Bingham plastic on a certain condition. When the behaviour of Non-Newtonian fluid affects the heat
transfer between the melt and concrete, the ablation model of COCO would not be appropriate.
7.2.11.2. CORCON
MELCOR is considered a state-of-the-art code for severe accident modelling and analysis, and it has a
reached a reasonably high level of maturity over the years as evidenced by its wide acceptance and
utilisation over a broad range of applications in regulatory decision support. Nevertheless, it is
important to recognise the phenomenological uncertainties in MELCOR and their significance to
MELCOR results. Moreover, it is important to understand the compounding effect of various
uncertainties on the ultimate parameter of interest i. e., source term for all practical purposes. Some of
the more important uncertainties are briefly discussed in this section.
Generally, in the case of dry cavity and metallic melts, CORCON-Mod3 thermal-hydraulic
calculations were found to be reasonably in good agreement with the test data. For oxidic melts in a
dry cavity, uncertainties in heat transfer models played an important role for two melt configurations –
a stratified geometry with segregated metal and oxide layers, and a heterogeneous mixture. Some
discrepancies in the gas release estimates were noted in a few cases. These discrepancies were
attributed, in part, to condensed phase chemical reactions modelling and, in part, to experimental
uncertainties. In the case of wet cavity, good agreement was found between the experimental data and
code calculations except, again, for the gas release data.
The CORCON-Mod3 assessment against SWISS and MACE experiments indicate that the crust
model in the code may often lead to sudden appearance and disappearance of crusts in a single
calculation time step (very small time). This crust instability is an artefact of the model employed in
the code. Results of the code validation exercise made it evident that differences exist between the
experimental data and the code predictions. Gas release predictions for individual species show
improvements when three-dimensional temperature profiles are accounted for. The combined gas
release prediction is not affected as such. This suggests that for integral plant calculations, further
modifications of CORCON-Mod3 in this area are not warranted.
The validation exercise revealed that the CORCON-Mod3 model dealing with chemical reactions
in the metallic melt is important with regard to the code's thermal-hydraulic capability. Specifically,
oxidation reactions with two metallic components – zirconium and silicon – are important. Silicon
appears in the melt as a result of condensed phase reactions between zirconium and silica. The
chemical heat release resulting from these reactions is exothermic at low temperatures, and
endothermic at high temperatures when SiO(g) is formed. Using thermodynamic data bases, the
temperature at which the reaction changes from exothermic to endothermic is estimated to be about
2 350 K. The assessment provides information concerning a possible range of uncertainties in
calculations for tests where temperature exceeds the above value. While in plant calculations involving
siliceous concrete interacting with the core debris consideration of the SiO chemistry model will
improve the gas release prediction, overall improvement in thermal-hydraulic and fission product
prediction over the entire duration of core-concrete interactions is not likely to be significant.
7.2.11.3. CORQUENCH
CORQUENCH has been exercised against a variety of experiments under both wet and dry cavity
conditions. When assessing these results as a whole, the code seems to reasonably predict melt
temperature and overall cavity erosion behaviour for dry cavity cases, and also seems to reasonably
predict melt temperature trends observed in tests. However, as the model currently stands, it does not
do well insofar as predicting lateral/axial power splits for different concrete types (i.e. differences
between limestone and siliceous concretes). In addition, the code is not able to assess potential
stratification issues associated with core melts that contain a significant fraction of structural steel.
Finally, the code has not been validated against tests with concrete that contain a significant amount of
rebar. Regarding these last two areas (i.e. rebar and high metal content melts), there are currently
NEA/CSNI/R(2016)15
experimental programs that are providing data that will help reduce modelling uncertainties. In terms
of debris coolability, it is important to develop models that can rationalise the transient nature (i.e. on
vs. off) nature of melt eruptions.
7.2.11.4. COSACO
One central remaining uncertainty is the extrapolation of test results to reactor-scale. The thermo-
hydraulic phenomena driving pool convection and concrete ablation might exhibit a scale-dependency
which is not captured in the models since the investigation of large-scale MCCI is not practically
feasible.
Another uncertainty relates to the stratification of the oxidic and metallic melt components in the
pool. For large melt pools, as they would occur at plant scale, layered melt configurations can be
expected. Due to the different melt/concrete interface conditions (no crust expected for metallic melt)
the heat transfer to the concrete differs. Thus, to get a consistent description of the global lateral
concrete erosion the inter-layer heat and mass transfer processes are decisive, as they influence the
layer temperature and thickness.
7.2.11.5. MAAP
Efforts to validate the models in MAAP have been focused on ablation distance and average corium
temperature. Uncertainties discovered through the validations are:
The code does not have a good shape model to account for the evolution of the ablation front
observed in several 2D MCCI tests. This may not be a big issue when corium spreads on top of a large
flat area. It may be an issue when corium is confined in a small space such as the sump in the reactor
cavity.
A mechanistic convective heat transfer model might be needed for the transition between natural
circulation (low gas evolution rate) and the gas-agitated heat transfer rate, and to allow separate
prediction of downward and sideward heat transfer rates for input to the cavity shape model.
A sophisticated bulk cooling model is needed to calculate the intensive heat transfer between the
corium pool and water during the initial transient after water injection. Although the current model in
MAAP5 is conservative in terms of melt temperature and ablation rate without the sophisticated
bulking cooling model, it may be overly conservative when high gas content concrete is ablated.
A material stratification model might be needed in MAAP5 to calculate certain conditions with
high metallic content in the corium pool and small gas evolvement from concrete erosion.
The thermophysical property model in the code needs to be compared against detailed predictions
using the NUCLEA database (Cheynet, Chaud, Chevalier, Mason, & Mignanelli, 2004) to check for
any potential improvements.
7.2.11.6. MEDICIS
Regarding the stratified situation the validation of the inter-layer heat transfer is rather weak, since in
available experiments layer temperatures and inter-layer heat fluxes are not measured and the erosion
is dominated by the metal layer. There is still a large uncertainty on this parameter. However results
obtained from simulant experiments indicate that the convective heat transfer between oxide and metal
layers might be very high compared to the heat transfer from the bulk oxide layer to the lateral
concrete interface. So if the pool stratification really occurs, a focusing of decay power towards the
metal layer becomes very likely. Therefore the main uncertainty is the criterion of pool stratification in
a realistic situation with real material.
Regarding 2D ablation, the slower axial ablation observed in case of siliceous concrete might be
explained by the presence of a solid accumulation or crust built-up in the MCCI early phase increasing
NEA/CSNI/R(2016)15
the thermal resistance of the bottom interface. However the evolution of this solid accumulation in the
long term phase and in presence of decay power in the reactor case is unknown.
It has been further observed, that there may be transient effects (like an initially superheated melt
in contrast to quasi-steady state conditions) during which the constant heat transfer coefficients are not
a good approximation. This is particularly due for the initial phase in the experiments with metal melt.
The effect of such transients in the experiment on the final results is however not large.
Another important observation is that the heat transfer coefficients are tuned to the prediction of
maximum erosion depths, which may occur in experiments only at some singularity. Since in the
calculation homogeneous conditions are assumed, most experiments are overestimated with regard to
the integral eroded volume but are met satisfactorily with regard to maximum erosion depths.
Depending on the objective of the MEDICIS calculation the user should be aware of this “maximum
erosion depth”-“integral erosion volume”-mismatch.
Generally it is not clear, which physical mechanisms dominate the heat transfer in the melt and
why the efficiency of this mechanism seems constant for a long time of the experiments. Recent
investigations on the effect of viscosity suggest, that the heat transfer coefficients should decrease at
later times for concrete fractions in the melt > 60 wt.-%. For a verification of this hypothesis there are
not enough experimental data available. A need for additional experimental data for MCCI in the late
phase, with large concrete fractions and low specific heat fluxes is thus identified.
As far as corium quenching is concerned, very large uncertainties remain on the quenching
efficiency in case of top quenching, in particular on the efficiency of melt eruption: the model used in
MEDICIS is derived from that of an older CORQUENCH version (Farmer, 2001) and is based on a
very simple correlation assuming that the entrained corium mass rate is proportional to the gas rate
leaving the pool interface and is not depending on the hole geometry; these two points are clearly open
to critics. Moreover the intermittent nature of melt eruption is accounted for even by more recent
models as those introduced in the recent versions of CORQUENCH code (Farmer, 2001).
The validation results show a good estimation of the ablation rate but an over estimation of the melt
temperature (Spindler, Tourniaire, Vandroux, Seiler, & Gubaidullin, 2005). Generally speaking, large
uncertainties exist in the MCCI experiments due to the high temperature with prototypical materials.
Their simulation is not easy, since an interpretation of the test conditions is necessary. As illustrated
by ACE-L2 simulation, the melt temperature as calculated in TOLBIAC-ICB is very sensitive to the
melt composition. This latter depends on the way corium is heated and melted, and on peculiar
phenomena which can occur during the experiments (crust formation, splattering, ceramic dissolution,
etc.). A good prediction of the melt composition thus requires a precise simulation of theses
phenomena. The crust thickness on the concrete wall, for example, is probably underestimated in the
code since the cooling effects of the decomposition products of concrete which cross the crusts are not
taken into account. The thermodynamic data base, which is used by GEMINI2 to obtain the liquidus
temperature, may also induce uncertainty in the calculation. At last, it should be noted that the order of
magnitude of the experimental uncertainty on temperature is 50 K.
Regarding 2D ablation, in case of siliceous concrete, the erosion rates are anisotropic. This is not
a well-known phenomenon, CLARA experiments should provide models explaining it (more
representative than a ratio between heat transfer coefficients defined by the code user). In case of
siliceous concrete, the interface temperature between the crust and the melt seems to be lower than the
liquidus temperature. This phenomenon is not well represented by TOLBIAC-ICB. Further
investigations are needed.
Gas production has a great importance concerning the melt ejection phenomenon. In case of a
siliceous concrete, it is important to consider which gas has to be considered (H20 or H2 which
NEA/CSNI/R(2016)15
depends on metal oxidation). The models based on chemical reactions deduced from pseudo-binary
diagrams need to be assessed.
In case of water aspersion, the melt could freeze very quickly. The interaction between a frozen
melt and the concrete is not well-known. It seems that in TOLBIAC-ICB the erosion algorithm
implemented privileges the axial erosion.
The water-ingression modelling is not satisfactory in case of metal-oxide corium. No
experimental data are available in this case.
The primary source of uncertainties in 2D MCCI simulation is the applicability of BALI correlations
for melt pools typical for MCCI with a flat melt layer and a strong bubbling. Errors in evaluating
boundary heat flux distribution especially in a stratified melt lead to wrong prediction of concrete
erosion rate, which governs wide range of processes, including gas release, FP release, etc.
7.2.11.9. WECHSL
The current level of agreement with experimental data from the ACE, MACE, SURC, BETA,
COMET-L, VULCANO, MOCKA, CCI test series would appear to be good enough to justify the use
of WECHSL for risk assessment studies, provided the uncertainties in the predicted results are taken
into account. Nevertheless, plant calculations still require extrapolation beyond the existing
experimental database. In particular the treatment of long term radial ablation by the oxide corium
fraction layered over the metallic melt fraction is not yet well established, and is expected to affect the
prediction for axial ablation by the metal melt.
7.3. Appendix on SEABOR crust test
In the accident of TEPCO’s Fukushima Daiichi Nuclear Power station, seawater was injected into the
RPV for extended period due to loss of the freshwater source. In the present accident management in
Japanese commercial light water reactors, seawater is regarded as the alternative water source. Under
long-term injection of seawater, concentration of salt increases as coolant evaporates in the core and
precipitates will deposit in narrow gaps in fuel bundles. When the debris bed is formed in the lower
plenum or the cavity and overlaid with seawater, precipitates are produced in the particulate debris and
crust surrounding the molten debris. It is anticipated that localised precipitation clogs flow paths and
may degrade the heat removal significantly. Highly concentrated boric acid may co-exist under severe
accident conditions in order to prevent re-criticality. It is also desirable to gain knowledge of
thermophysical properties of precipitates produced from mixture of seawater and boric acid.
The molten core-concrete interaction (MCCI) of a piled debris bed on the cavity floor is
dominated based on a balance among the decay heat generated from radioactive nuclides, the heat
transfer to overlying pool water and the heat transfer across molten core-concrete interfaces. When
submerged in a pool, surfaces of molten debris are covered with crust that is mainly composed of
oxide materials. Although structure of crust is very complex in nature, its permeability and thermal
conductivity are relatively low in a base matrix. Based on past experimental knowledge on MCCI, it is
interpreted that the heat transfer across crust is enhanced by localised heat transfer mechanisms of
water ingression through cracks.
Salt precipitation decreases permeability in a crust base matrix and increases flow resistance of
cracks, which may degrade the heat removal to overlying pool water across the upper crust and
eventually enhance concrete ablation. In order to deepen understanding of this mechanism and
obtaining the experimental database to establish physical models, the Regulatory Standard and
Research Department, Secretariat of Nuclear Regulation Authority (S/NRA/R) planned the debris
crust heat transfer test under salt precipitation as a part of the Seawater – Boric Acid Injection Heat
NEA/CSNI/R(2016)15
Removal Test (SEABOR Test). The test consists of two categories, the mockup fuel bundle test
(SEABOR Bundle Test) and the debris crust test (SEABOR Crust Test).
7.3.1. Test Facility
Figure 7.3-1 shows a layout of the SEABOR Crust Test facility. Figure 7.3-2 shows a detailed plan of
the test section water tank and the heat transfer surface. A simulated crust block made of oxide porous
media is set with exposing its upper surface to be flush with the bottom surface of the tank.
Preconditioned solutions, deionised water, artificial seawater, or mixed solution of artificial seawater
and boric acid, are replenished so that the water level is maintained within a specified range. The crust
is heated uniformly at the bottom surface by the copper heat transfer surface consisting of the top
plate, the bottom block and ten sheath heaters. The copper heat transfer surface and the crust block are
thermally connected via the thin SUS304 board (100 microns in thickness). Ceramic insulators to
suppress an amount of heat loss in the lateral direction surround the crust block.
Figure 7.3-1: Outline of facility
ポリカーカーボネード
polycarbonate
水槽
water wall
level
meter water
surface
液面
external heater
SUS wall
ステンレス水槽
210x210
(320)
500
crust 試験体
support
block pin
ポリカーカーボネード
サポートピン polycarbonate
水槽
wall
Oリング＆
O-ring and copper plate
パッキン
packing 銅基板熱電対 thermal insulator
断熱板
T/C
50x50x30 （4本）ステンレス薄板
SUS thin board
250x250x100μm
Unit: mm
153
銅ブロック
copper thermal 〇○○○○ embedded
ヒーター
block heaters
（900Wx10本）
insulator
断熱材 (x10)
joint bolt
締結ボルト
Figure 7.3-2: Heat transfer surface and test section tank
NEA/CSNI/R(2016)15
The copper top plate and the thin SUS304 board are connected with diffusion welding. The crust
block is connected onto the SUS304 board with a bonding agent made of fire-proof ceramic and
inorganic polymer (Aron Ceramic W of Toa Gosei). This thin SUS304 board also works as a seal to
separate the water region from the copper top plate. This copper top plate is attached to the copper
block with a joint bolt. Four sheath type thermocouples of the diameter 0.25mm (Okazaki-Seisakusho
SUPER COUPLE K) are inserted into holes of the diameter 0.3mm drilled along the centreline of the
copper top plate at the pitch 2mm. Temperature and heat flux on the surface are evaluated based on
temperature signals and the linear auto-regression method. The maximum measurable temperature is
900 Kelvin.
The makeup water system consists of the water tank and the pump. Three kinds of working fluids
are used, deionised water, 2.5 times concentrated artificial water or mixed solution of artificial
seawater and boric acid at the same weight concentration as the artificial seawater. The water level is
maintained by controlling replenishment of solutions by positioning the ball valve at the tank inlet. In
making mixing fluids, artificial seawater and boric acid are stirred and prepared at planned
concentration levels in different tanks. Two solutions are mixed together in the makeup water tank and
heated at about 95 Celsius. Preconditioned solutions are supplied to the test section water tank by the
makeup pump. The bypass line is attached to re-circulate solutions back to the makeup water tank if
necessary.
7.3.2. Test Specimens
Figure 7.3-3 illustrates the simulated crust block made of oxide porous media employed in this test.
The entire crust block is enveloped by a cuboid of 50mm×50mm×10mm. A dimensional tolerance is
smaller than 0.5mm. The crust block is composed of small pieces that are arranged with a fixed gap
distance, 2mm, in order to realise three gap patterns, no gap, number-sign shaped gap and square
shaped gap. As shown in Table 7.3-1, a combination of two oxide materials, Silica and Alumina, and
four different particle diameters are employed as test parameters.
Table 7.3-1: Specification of ceramics test specimens
Material IDs Mat-A Mat-B Mat-C Mat-D Mat-E
Silica
100 100 100 60 -
Comp
ositio
(SiO2) %
n
Alumina (Al2O3) % - - - 40 100

Particle diameter mm 0.6 0.3 0.3 0.3 0.4
Porosity % 26 32 19 36 32
Medium heat Large heat
Comment Large particle Large porosity Small particle
conduction conduction
NEA/CSNI/R(2016)15
Unit: mm
(a) No-gap (b) Number sign shaped gap (c) Square shaped gap
(Thickness is 10mm for all cases)
(d) Connection of crust block (e) Crust block surrounded by (f) Setting of crust block
is onto thin SUS304 board ceramic insulator in test section water tank
Figure 7.3-3: Gap pattern of crust block and setting in test section water tank
7.3.3. Test procedures
The test is conducted according to the following steps.
Table 7.3-2 summarises major test conditions.
1. Solutions (artificial seawater and/or Boric acid) are prepared at the specified target
concentration in different tanks.
2. Solutions are moved to the makeup water tank and heated at the specified temperature.
3. Finished solution is moved to the test section water tank by the makeup pump.
4. Solution in the test section tank is heated up to the saturated condition by the external
heaters.
5. After reaching to the saturated condition in the test section water tank, the heaters embedded
in the copper block are put on to heat the test specimen.
6. By monitoring the water level in the test section tank, replenishment of solution is adjusted
to maintain the water level within the specified range.
7. When a rapid temperature increase is observed from any of the thermocouples embedded in
the copper plate, it is judged that the departure from nucleate boiling occurred and the heater
power is turned off.
8. The heater power is also turned off if the monitored test condition enters a risky zone, such
as measured temperature values approaching the melting point of copper.
7.3.4. Test results
Influences of gap pattern, porosity and solution composition are investigated in this test. Figure 7.3-4
compares dependence of the heat flux on the wall super heating (Figure 7.3-4 (a)) and the heating time
(Figure 7.3-4 (b)) among three different gap patterns for Material-E made of Alumina of the particle
diameter 0.4mm submerged in solution of 2.5 times concentrated artificial seawater.
NEA/CSNI/R(2016)15
Table 7.3-2: Major test conditions

No. Item Setpoint Comment
Water Deionised water
Seawater Artificial seawater 2.5 time concentrated
1 Solutions composition
Artificial seawater 1.25 times + Boric acid at
Mixed solution
the same weight concentration as seawater
Preheating by in the makeup water tank

2 Temperature of supplied solution 95 Celsius
(Max. 1kW)
Volume of water reserved in the test
3 14 L 210mm×210 mm×320 mm (height)
section tank
Saturation condition is maintained in the test
4 Temperature of reserved water 100 Celsius
section tank
0.9 kW × 10 leaders
Shown on the
5 Power of the underlying heaters Controlled at three levels:
right
2.6, 5.0, 7.0 kW
Target water level 300 mm Varies due to evaporation and supply
Criteria of solution Start when the water level becomes lower
6
replenishment Replenishment About 1.7 L at
than 280 mm
quantity one shot
Stop when the water level exceeds 320 mm
Occurrence of Sudden increase of the copper base
critical heat flux temperature
7 Criteria of test termination Heater failure or
Heater : 950 Celsius
the copper base
Copper base : 900 Celsius
melting
In a log-log plane that employs the heating time as the abscissa, the heat flux increases
approximately linearly with regard to the wall super heating in case of the no gap pattern. In this case,
the melting point of the copper block comes before the departure from nucleate boiling and the test is
terminated at this point. The two gap patterns show higher heat removal than does the no-gap pattern
and the number sign shaped pattern shows higher heat removal than does the square shaped pattern.
From three boiling curves, it is interpreted that the heat transfer is enhanced as the total gap area
becomes larger. The gap works as a flow path through which the coolant water is directly supplied
over the heated surface without passing through the base matrix that possesses the much larger flow
resistance. At the same time, generated bubbles can move smoothly through gaps and remove
generated heat from the heated surface without passing through the base matrix.
By increasing the wall superheating, a sudden decrease in the heat flux is observed at about 400K
and 500K for the square shaped pattern and the number sign shaped pattern, respectively. Beyond this
point, two gap cases approach the no-gap case. In these test conditions, salt precipitation occurs during
an evaporating concentration process. It is possible that precipitations deposited in gaps induce
clogging or higher flow resistance of gaps and results in noticeable degradation of the heat transfer.
The counter current flow limitation in gaps can also be attributed to these trends. Further
investigations are planned to clarify underlying mechanisms.
In the three gap patterns shown in Figure 7.3-4, heating time varies from 140 min to 210 min.
When employing heating time, which can be regarded as an index of accumulation of precipitations,
as the abscissa, similar trends can be observed.
NEA/CSNI/R(2016)15
Heat Flux
Heat Flux
Number Number
Square Square
No Gap No Gap
Wall Super Heating Heating Time (min)
(a) Heat Flux vs Wall Super Heating (b) Heat Flux vs Heating Time
Figure 7.3-4: Dependence of measured heat flux on gap geometry
NEA/CSNI/R(2016)15
8. References
AFNOR. (2005), Eurocode 2: Design of concrete structures. Part 1-2: General rules – Structural fire
design, European Standard NF EN 1992-1-2.
Akers, D. W., and McCardell, R. (1989), Core Materials inventory and behaviour. Nuclear
Technology(87), 214-223.
Allelein, H., Breest, A., and Spengler, C. (1999), Simulation of Core Melt Spreading with LAVA:
Theoretical Background and Status of Validation, OECD Workshop on Ex-Vessel Debris
Coolability.
Allelein, H.-J., and Burger, M. (2006), Considerations on ex-vessel corium behavior: Scenarios, MCCI
and coolability. Nuclear Engineering and Design, 236(19-21), 2210-2236.
Alsmeyer, H. (1987), BETA Experiments in Verification of the WECHSL Code: Experimental Results
on the Melt-Concrete Interaction. Nuclear Engineering and Design, 103, 115-125.
Alsmeyer, H., and Firnhaber, M. (1990), Specification of the International Standard Problem ISP-30:
BETA V5.1 Experiment on Melt_Concrete Interaction. ISP-30: BETA V5.1.
Alsmeyer, H., and Tromm, W. (1999), The COMET concept for cooling core melts: evaluation of the
experimental studies and use in the EPR. Report FZKA 6186, EXV-CSC(99)-D036,
Forschungzentrum Karlsrühe, Karlsrühe. Germany.
Alsmeyer, H., Adelhelm, C., Diilmann, H.-G., Foit, J., Heinle, M., Ratajczak, W., Tromm, W. (1995),
BETA Experiments on Melt-Concrete Interaction: the Role of Zirconium and the Potential
Sump Water Contact During Core Melt-Down Accidents. Nuclear Engineering and Design,
154, 61-68.
Alsmeyer, H., Adelhelm, C., Dillmann, H.-G., Foit, J., Heinle, M., Ratajczak, W., Tromm, W. (1992),
"BETA-Experiments on Zirconium Oxidation and Aerosol Release During Melt-Concrete
Interaction", Second OECD-CSNI Specialist Meeting on Molten Core Debris-Concrete
Interactions. Karlsrühe, Germany.
Alsmeyer, H., Albrecht, G., Meyer, L., Häfner, W., Journeau, C., Fischer, M., Bandini, G. (2005). Ex-
vessel Core Melt Stabilization Research (ECOSTAR). Nuclear Engineering and Design, 235,
271-284.
Alsmeyer, H., Barleon, L., Koster, J., Michael, J., Müller, U., & Reimann, J. (1977), Ein Modell zur
Beschreibung der Wechselwirkung einer Kernschmelze mit Beton. KfK-Report.
Alsmeyer, H., Cenerino, G., Cordfunke, E. H., Fioravanti, D., Fischer, M., Foit, J., Wegener, D.
(1995), Molten corium/concrete interaction and corium coolability. A State of the Art Report.
European Commission.
Alsmeyer, H., Cenerino, G., Cordfunke, E. P., Fioravanti, D., Fischer, M., Foit, J., Wegener, D.
(1995), Molten corium/concrete interaction and corium coolability—a state of the art report.
European Commission, Luxembourg.
NEA/CSNI/R(2016)15
Alsmeyer, H., Cron, T., Fluhrer, B., Messemer, G., Miassoedov, A., Schmidt-Stiefel, S., and Wenz, T.
(2007), The COMET-L3 Experiment on Long-Term Melt-Concrete Interaction and Cooling by
Surface Flooding, FZKA 7244.
Alsmeyer, H., Cron, T., Merkel, G., Schmidt-Stiefel, S., Tromm, W., Wenz, T. (2004), Melt cooling
by bottom flooding: the experiment CometPC-H4 and H-5. Forschungzentrum Karlsrühe,
Karlsrühe, Germany.
Alsmeyer, H., Farmer, M., Ferderer, F., Spencer, B., Tromm, W. (1998), The COMET-Concept for
Cooling of Ex-Vessel Corium Melts. ICONE-6. San Diego, CA USA.
Alsmeyer, H., Kuczera, B., Lecomte, C., Sehgal, B., Smith, B., Smith, P., Turland, B. (1992),
Specialist Meeting on Core debris/concrete interactions: summary and recomendations. NEA
Specialist Meeting on Core debris/concrete interactions.
Alsmeyer, H., Miassoedov, A., Cranga, M., Fabianelli, R., Ivanow, I., Doubleva, G. (2005), The
COMET-L1 Experiment on Long-Term Concrete Erosion and Surface Flooding. NURETH-
11. Avignon, France.
ASAMPSA2. (2013). ASAMPSA2 Best-practices guidelines for L2 PA development and applications
Volume 1 – General; Technical report, Volume 2 – Best practices for the Gen II PWR, Gen II
BWR L2PSAs. Extension to Gen III reactors. Volume 3 – Extension to Gen IV reactors.
European Commission.
ASN. (2014). Resolution 2014-DC-0403 of 21th January 2014, Retrieved from www.french-nuclear-
safety.fr/References/ASN-Resolutions/ASN-Resolution-2014-DC-0403-of-21th-January-2014.
Atkhen, K., Spindler, B. (2010), Use of MCCI- Projects and CCI French R&D program for reactor
applications and future needs to reduce remaining uncertainties. MCCI-OCDE seminar.
Cadarache.
Baïchi, M. (2001), Contribution à l'étude du corium d'un réacteur nucléaire accidenté; aspects
puissance résiduelle et thermodynamique des systèmes U-UO2 et UO2-ZrO2, PhD Thesis,
Joseph Fourier University. Grenoble, France: in French.
Bakardjieva, S., Barrachin, M., Bechta, S., Bottomley, D., Brissonneau, L., Cheynet, B., Wiss, T.
(2010), Improvement of the European thermodynamic database NUCLEA. Progress in
Nuclear Energy, 52, 84-96.
Barrachin, M. (2011). Assessment of Ablation Temperatures of Concretes from Past MCCI and
Thermodynamic Experiments, IRSN report, DPAM-SEMIC-2011-315, 2011 (restricted).
Berenson, P. J. (1961), Transition Boiling Heat Transfer from a Horizontal Surface. Journal of Heat
Transfer, 83, 351-358.
BFS. (2012), Safety Requirements for Nuclear Power Plants. Retrieved from
www.bfs.de/de/bfs/recht/rsh/volltext/A1_Englisch/A1_11_12_SiAnf_eng.pdf
Blose, R. E., Gronager, J. E., Suo-Anttila, A. J., & Brockman, J. E. (1987), SWISS: Sustained Heated
Metallic Melt/Concrete Interactions With Overlying Water Pools. NUREG/CR-4727.
Blose, R. E., Powers, D. A., Copus, E. R., Brockmann, J. E., Simpson, R. B., & Lucero, D. A. (1993),
Core-Concrete Interactions with Overlying Water Pools – The WETCOR-1 Test. NUREG/CR-
5907.
NEA/CSNI/R(2016)15
Blottner, F. G. (1979), Hydrodynamics and Heat Transfer Characteristics of Liquid Pools with Bubble
Agitation. SAND 79-1132.
BNL. (1990), MELCOR Simulation of Long-Term Station Blackout at Peach Bottom. Brookhaven
National Laboratories.
Bogatov, S., Borovoi, A., Gavrilov, S., Lagunenko, A., Pazukhin, E., & Khvoschinskiy, V. (2007),
Database on Location and Status of Nuclear fuel at Unit-4 of Chernobyl NPP before and after
the Accident. Kurchatov Institute Pre-print N° 130-11/2.
Bolshov, L., Strizhov, V. (2006), SOCRAT – The system of codes for realistic analysis of severe
accidents. ICAPP06. Reno, USA.
Bonnet, J.-M. (1999), Thermal hydraulic phenomena in corium pools for ex-vessel situations: The
BALI experiment. OECD workshop on ex-vessel coolability, FZKA 64753. Karlsrühe,
Germany.
Bonnet, J.-M. (2000), Thermal hydraulic phenomena in corium pools for ex-vessel situations: the
BALI experiment. ICONE8. Baltimore, Maryland, USA.
Bonnet, J.-M., & Seiler, J.-M. (1992), Coolability of Corium Spread Onto Concrete Under Water, the
PERCOLA Model. 2nd OECD (NEA) CSNI Specialist Meeting on Molten Core Debris-
Concrete Interaction. Karlsrühe, Germany.
Bonnet, J.-M., & Seiler, J.-M. (1999), Thermal hydraulic phenomena in corium pools: the Bali
Experiment. ICONE7. Tokyo, Japan.
Bottin, M., Amizic, M., Guyez, E., Seiler, J.-M., Ratel, G., Samaille, M., Bonnard, J.-C. (2016),
Molten Corium–Concrete Interaction: Investigation of convective heat transfer in a pool with
gas sparging. Annals of Nuclear Energy, 93, 35-42.
Bottinga, Y., & Weill, D. F. (1972), The Viscosity of Magmatic Silicate Liquids: A Model for
Calculation, American Journal of Science. 272, 438-475.
Bottomley, D., Stuckert, J., Hofmann, J., Tocheny, L., Hugon, M., Journeau, C., et~al. (2012), Severe
Accident Research in the Core Degradation Area: An Example of Effective International
Cooperation Between the European Union (EU) and the Commonwealth of Independent
States (CIS) by the International Science and Technology Center. Nuclear Engineering and
Design, 252, 226-241.
Bradley, D. R. (1988), Modeling of Heat Transfer Between Core Debris and Concrete. National Heat
Transfer Conf. HOUSTON, Texas, USA.
Bradley, D. R., Gardner, D., Brockmann, J., & Griffith, R. (1993), CORCON-MOD3: An Integrated
Computer Model for Analyse of Molten Core-Concrete Interactions. NUREG/CR-5843.
Breeding, R. J., Harper, F. T., Brown, T. D., Gregory, J. J., Payne, A. C., Gorham, E. D., Amos, C. N.
(1991), Evaluation of Severe Acciedent Risks: Quantification of Major Input Parameteres.
Expert´s Determination of Strucutural Response Issues. NUREG/CR-4551.Vol.2, Rev.1, Part
3.
Brockmann, J. E., Arellano, F. E., & Lucero, D. A. (1989), Validation of Models of Gas Holdup in the
CORCON Code.
NEA/CSNI/R(2016)15
Brown, T. D., Breeding, R., Jow, H., Higgins, S., Shiver, A., Helton, J., Amos, C. (1990), Evaluation
of Severe Accident Risks: Grand Gulf, Unit 1. Appendices. NUREG/CR-4551. Vol. 6, Rev.1,
Part 2.
Burakov, B. (2013), Material study of Chernobyl “lava” and “hot” particles. International Experts’
Meeting on Decommissioning and Remediation after a Nuclear Accident. Vienna, Austria.
Carenini, L., Haquet, J., Journeau, C. (2007), Crust Formation and Dissolution During Corium-
Concrete Interaction. ICAPP ’07. Nice, France.
Casas, J., Corradini, M. (1992), Study of void fraction and mixing in a pool configuration by an
upwards gas flow. Nuclear Technology, 99, 104-119.
Cénérino, G., Rahni, N., Chevrier, P., Dubreuil, M., Guigueno, Y., Raimond, E., & Bonnet, J.-M.
(2016), Severe accident mitigation strategy for the generation II PWRs in France – Some
outcomes of the ongoing periodic safety review of the French 1300 MWe PWR series.
Progress in Nuclear Energy, 87, 32-38.
Chatelard, P., Reineke, N., Arndt, S., S., B., Cantrel, L., Carenini, L., Piar, L. (2014), ASTEC V2
severe accident integral code main features, current V2.0 modelling status, perspectives.
Nuclear Engineering And Design, 272, 119-135.
CHEMAPP. (1999), CHEMAPP – The thermochemistry library for your software. Herzogenrath,
Germany: GTT-Technologies.
Chevalier, P. (1999), TDBIV – TDBCR - Thermodynamic Databases for Nuclear Chemistry –
Documentation. EU 4th Framework Program INV-CIT(99)-P039.
Cheynet, B. (2007), NUCLEA. Retrieved from http://hal.archives-ouvertes.fr/hal-00165418/en/
Cheynet, B., Chaud, P., Chevalier, P. F., Mason, P., Mignanelli, M. (2004), NUCLEA. Properties and
Phase Equilibria in the Nuclear Systems of Interest. Journal de Physique IV, 113, 61-64.
Cheynet, B., Chevalier, P., Fischer, E. (2002), Thermo suite. Calphad, 26(2), 167-174. Retrieved from
http://hal.archives-ouvertes.fr/hal-00165418/en/
Chu T., Y. (1978), Radiant heat evaluation of concrete- A study of the erosion of concrete due to
surface heating. SAND77-0922, Sandia Laboratories, Albuquerque.
Chu, C. C., Sienicki, J. J., Spencer, B. W., Frid, W., Löwenhielm, G. (1995), Ex-vessel melt coolant
interactions in deep water pool; studies and accident management for Swedish BWR´s. Nucl.
Eng. Des.(155), 159-213.
Chu, T. J., Pilch, M., Bentz, J. H., Behbahani, A. (1998), Experimental investigation of creep behavior
of reactor vessel lower head. Workshop on In-Vessel Core Debris Retention and Coolability.
Garching, Germany.
Claus, D. B. (1989). Severe Accident Testing of Electrical Penetration Assemblies. NUREG/CR-5354.
Cohen, I., & Schaner, B. E. (1963), A metallographic and X-Ray study of the UO2-ZrO2 system. J.
Nucl. Mater.(9), 18-52.
Copus, E. R., & Bradley, D. R. (1986), Interaction of Hot Solid Debris with Concrete, NUREG/CR-
4558.
NEA/CSNI/R(2016)15
Copus, E. R., Blose, R. E., Brockmann, J. E., Gomez, R. D., Lucero, D. A. (1990), Experimental
Results of Core-Concrete Interactions Using Molten Steel with Zirconium,. NUREG/CR-
4794; SAND86-2638.
Copus, E. R., Blose, R. E., Brockmann, J. E., Gopmex, R. D., Lucero, D. A. (1989), Core-Concrete
Interactions using Molten Steel with Zirconium on a basaltic Basemat: The SURC-4
Experiment, NUREG/CR-4994, Sandia National Laboratory.
Copus, E. R., Blose, R. E., Brockmann, J. E., Simpson, R. B., Lucero, D. A. (1992-1), Core-Concrete
Interactions Using Molten Urania With Zirconium on a Limestone Concrete Basemat, The
SURC-1 Experiment, NUREG/CR.
Copus, E. R., Blose, R. E., Brockmann, J. E., Simpson, R. B., Lucero, D. A. (1992-2), Core-Concrete
Interactions Using Molten UO2 With Zirconium on a Basaltic Basemat, The SURC-2
Experiment, NUREG/CR.
Copus, E., Powers, D. (1992), The SURC Test Series. Second CSNI Specialist Meeting on Molten
Core Debris-Concrete Interactions, Karlsrühe, Germany.
Corradini, M. L. (1983), A Transient Model for the Ablation and Decomposition of Concrete, Nucl.
Tech, 62, 263.
Cranga, M., Fabianelli, R., Jacq, F., Barrachin, M., Duval, F. (2005), The MEDICIS code, a versatile
tool for MCCI modelling. ICAPP05, Seoul Korea.
Cranga, M., Ferry, L., Haquet, J. -F., Journeau, C., Michel, B., Mun, C., Atkhen, K. (2010), MCCI in
an oxide-metal pool: lessons learnt from VULCANO, Greene and BALISE experiments.
ERMSAR-10, Bologne, Italy.
Cranga, M., Michel, B., Mun, C., Barrachin, M. (2010), Outcome of the available 2D MCCI database
for the reactor case in dry conditions. MCCI-OECD seminar, St-Paul lez Durance, France.
Cranga, M., Mun, C., Bottin, M. (2012), Analysis of top quenching models during MCCI: status of
knowledge and proposals of developments for the ASTEC code, SARNET2-MCCI-P12, IRSN
report, PSN-RES-SAG 2012-116.
Cranga, M., Mun, C., Marchetto, C. (2010), MCCI in an oxidic pool: lessons learnt from MCCI-
OECD and VULCANO real material experiments in dry conditions, MCCI-OECD seminar.
St-Paul lez Durance, France.
Cranga, M., Mun, C., Michel, B., Duval, F., Barrachin, M. (2008), Interpretation of real material 2D
MCCI experiments in homogeneous oxidic pool with the ASTEC/MEDICIS code, ICAPP08.
Anaheim, California.
Cranga, M., Spengler, C., Atkhen, K., Bechta, S., Bottomley, P., Dejardin, P., Fargette. (2013), State
Of the Art Report on MCCI in dry conditions: analysis of experiments and modelling,
SARNET.
Cranga, M., Spengler, C., Atkhen, K., Fargette, A., Fischer, M., Foit, J., Spindler, B. (2014), Towards
an European consensus on possible causes of MCCI ablation anisotropy in oxidic pool, Annals
of Nuclear Energy, 74, 72-88.
Cranga, M., Spindler, B., Dufour, E., Dimov, D., Atkhen, K., Foit, J., Schmidt, W. a. (2010),
Simulation of corium concrete interaction in 2D geometry, Progress in Nuclear Energy, 52,
76 - 83.
NEA/CSNI/R(2016)15
CSN. (1997). A Regulatory Evaluation of the Vandellós-II Probalistic Safety Analysis (Level-2),
ERI/CSN 97-801.
Deans, J. A. (1985), Lange’s Handbook of Chemistry, 23rd Edition. McGraw-Hill.
Deckwer, W. (1980), On the Mechanism of Heat Transfer in Bubble Column Reactors, Chem. Eng.
Science, 35, 1341-1346.
Dhir, V. K., Purohit, G. (1978), Subcooled Film Boiling Heat Transfer from Spheres, Nuclear
Engineering and Design, 47, 49-66.
Duignan, M. R., Greene, G. A. (1989), Enhanced Convective and Film Boiling Heat Transfer by
Surface Injection, Dissertation, State University of New York at Stony Brook, New York.
Duval, F., Cranga, M. (2008), ASTECV2 MEDICIS MCCI module Theoretical Manual, N.T.
DPAM/SEMIC 08/102.
Epstein, M. (1997), MCCI Thermohydraulics, MCCI Heat Transfer Models – A Comparison with
Prototypical Test Data and Some New Ideas, ACEX TR-C-17 Draft.
Epstein, M. (2004), A Model of Heat and Mass Transfer Beneath an Ablating Concrete Surface. J.
Heat Transfer, 126, 290-294.
Epstein, M. (2006), Dryout Heat Flux During Penetration of Water Into Solidifying Rock. J. Heat
Transfer, 128, 847-850.
Epstein, M., Zhou, Q. (2012), Ex-Vessel Corium Melt Coolability, FAI/12-0047.
Epstein, M., Cheung, F., Chawla, T., Hause, G. (1981), Effective Thermal Conductivity for Combined
Radiation and Free Convection in an Optically Thick Heated Fluid Layer, J. of Heat Transfer,
103, 114-120.
Epstein, M., Petrie, D., Linehan, J. H., Lambert, G. A., Cho, D. H. (1981), Incipient stratification and
mixing aerated liquid-liquid or liquid-solid mixtures, Chem. Eng. Sci., 36, 84-87.
Epstein, M., Petrie, D., Linehan, J., Lambert, G., Cho, D. (1981), Incipient stratification and mixing
aerated liquid-liquid or liquid-solid mixtures, Chem. Eng. Sci., 36, 84-87.
EUR Organisation. (2012), European Utility Requirements for LWR Nuclear Power Plants documents.
Retrieved from Volume 4:
www.europeanutilityrequirements.org/Documentation/EURdocument.aspx
Fargette, A. (2012), MCCI Pool Temperature and Viscosity: A Discussion of the Impact of Scale, 5th
European Review Meeting on Severe Accident Research (ERMSAR-2012), Cologne, Germany.
Fargette, A. (2013), A simple Procedure for Transposing Small-Scale MCCI Temperature Histories to
the Reactor-Scale. 6th European Review Meeting on Severe Accident Research (ERMSAR-
2013), Avignon (France).
Farmer, M. T. (2001), Modeling of Ex-Vessel Corium Coolability with the CORQUENCH Code,
ICONE-9. Nice France.
Farmer, M. T. (2005), OECD MCCI Analytical Support (Stand Alone Task): Melt Eruption Modeling
of the MCCI Tests, NRC/MCCI/2005-TR01.
NEA/CSNI/R(2016)15
Farmer, M. T. (2006). Phenomenological Modeling of the Melt Eruption Cooling Mechanism during
Molten Corium Concrete Interaction. ICAPP06. Reno, Nevada, USA.
Farmer, M. T. (2010). The CORQUENCH Code for Modeling of Ex-Vessel Corium Coolability Under
Top Flooding Conditions, Code Manual – Version 3.03. OECD/MCCI-2010-TR03.
Farmer, M. T., Kilsdonk, D. J., & Aeschlimann, R. W. (2009). Corium Coolability under Ex-Vessel
Accident Conditions for LWRs. Nuclear Eng. Technology, 41, 575-602.
Farmer, M. T., Kilsdonk, D. J., Aeschlimann, R. W., & Lomperski, S. (2010). Core-Concrete
Interaction Test Six (CCI-6) Final Report. MCCI-2010-TR04.
Farmer, M. T., Lomperski, S., Kilsdonk, D. J., & Aeschlimann, R. W. (2006). MCCI Project Final
Report. OECD/MCCI-2005-TR06.
Farmer, M. T., Lomperski, S., Kilsdonk, D. J., & Aeschlimann, R. W. (2006). MCCI Project: 2D Core
Concrete Interaction (CCI) Tests, Final Report. MCCI-2005-TR05.
Farmer, M. T., Lomperski, S., Kilsdonk, D. J., & Aeschlimann, R. W. (2010). OECD MCCI-2 Project
Final Report. OECD/MCCI-2010-TR07.
Farmer, M. T., Lomperski, S., Kilsdonk, D. J., Aeschlimann, R. W., Basu, S., Lomperski, S., Basu, V.
(2007). A Summary of Findings from the Melt Coolability and Concrete Interaction (MCCI)
Program. ICAPP ’07. Nice, France.
Farmer, M. T., Sienicki, J. J., & Spencer, B. W. (1990). CORQUENCH: A Model for Gas Sparging-
Enhanced, Melt-Water, Film Boiling Heat Transfer. ANS Winter Meeting on the Thermal
Hydraulics of Severe Accidents. Washington, D.C. USA.
Farmer, M. T., Spencer, B. W., & Aesclimann, R. (2000). Liquidus/solidus and Zr solubility
measurements for PWR and BWR core melt compositions. Wissentschaftlische Berichte
FZKA 6475, 380-393.
Farmer, M. T., Spencer, B. W., Binder, J. L., & Hill, D. J. (2001). Status and Future Direction of the
Melt Attack and Coolability Experiments (MACE) Program at Argonne National Laboratory.
ICONE-9. Nice, France.
Farmer, M. T., Spencer, B. W., Schneider, J. P., Bonomo, B., & Theofanous, G. (1992). Modeling and
Database for Melt-Water Interfacial Heat Transfer. 2nd CSNI Specialist Meeting on Core
Debris-Concrete Interactions. Karlsrühe, Germany.
Fieg, G. (1978). Heat Transfer Measurements of Internally Heated Liquids in Cylindrical Convection
Cells. 4th PAHR Information Echange Meeting. Varese, Itally.
Filippov, A. (2011). Numerical simulation of the experiments on turbulent natural convection at
cylindrical pool of heat generating liquid. J.Eng. Thermophys.
Filippov, A. (2011b). Numerical simulation of turbulent heat transfer in oxidic melt at corium catcher
of NPP with VVER-1200. J.Eng. Thermophys.
Filippov, A., E.V., M., & D.D., K. (2013). Verification of SOCRAT/HEFEST code solutions of the
problems of transient heat transfer in conjugated solids and in orthotropic medium. Bull. Russ.
Acad. Sci. Energetics.
Fink, J. K., & Petri, M. C. (1997). Thermophysical Properties of Uranium Oxide. Argonne Natiola
Laboratory Report ANL/RE-97/2.
NEA/CSNI/R(2016)15
Fink, J. K., Thompson, D. H., Armstrong, D. R., Spencer, B. W., & Sehgal, B. R. (1995). Aerosol and
Melt Chemistry in the ACE Molten Core-Concrete Interaction Experiments. High
Temperature and Materials Science, 33(1), 51-76.
Fischer, M. (1999). Main conceptual features of the EPR melt retention concept. Workshop on Ex-
vessel Debris Coolability. Karlsrühe, Germany.
Fischer, M. (2005). Long-term melt stabilization as part of the severe accident mitigation strategy of
the European Pressurised Reactor (EPR). ICAPP'05. Seoul, Korea.
Fischer, M., & A. Henning. (2009). EPR™ Engineered Features for Core Melt Mitigation in Severe
Accidents. ICAPP ’09. Tokyo, Japan.
Fischer, M., Herbst, O., & Schmidt, H. (2005). Demonstration of the Heat Removing Capabilities of
the EPR Core Catcher. Nuclear Engineering and Design, 235, 1189-1200.
Foit, J. J. (1997). Modeling of Oxidic Molten Core-Concrete Interaction in WECHSL, Nucl. Eng. and
Des., Vol. 170, pp 73-79 (1997). Nucl. Eng. and Des., 170, 73-79.
Foit, J. J. (2012 ). MCCI: From Experiment to Reactor. 6th Progress Meeting SARNET2-WP6-MCCI.
Budapest.
Foit, J. J. (2015). MCCI on LCS concrete with and without rebars. ERMSAR-2015. Marseille, France,
24-26 March 2015.
Foit, J. J., & Miassoedov, A. (1995). Modelling of Viscosity and Heat Transfer of Complex Oxidic
Melts in WECHSL. Wiss. Ber. FZKA, 5507.
Foit, J. J., Cron, T., Fluhrer, B., Miassoedov, A., & Wenz, T. (2012). MOCKA Experiments on
Concrete Erosion by a Metal and Oxide Melt. 5th European Review Meeting on Severe
Accident Research (ERMSAR). Cologne, Germany.
Foit, J. J., Fischer, M., Journeau, C., & Langrock. (2014). Experiments on MCCI with oxide and steel.
Annals of Nuclear Energy, 74, 100-109.
Foit, J. J., Reimann, M., Adroguer, B., Cenerino, G., and Stiefel, S. (1995). The WECHSL-Mod3
Code: a computer program for the interaction of a core melt with concrete including the long
term behavior; Model descriptions and User’s manual. FZKA 5416.
Foit, J., T., C., & Fluhrer, B. (2013). Interaction of a Metal and Oxide Melt with Reinforced Concrete
in MOCKA Experiments. The 15th International Topical Meeting on Nuclear Reactor Thermal
– Hydraulics, NURETH-15. Pisa, Italy.
Fukasawa, M., Tamura, S., & Hasebe, M. (2005). A SUPPRIMER. J. Nucl. Sci. Technol., 42[8], 706,
42.
Fukasawa, M., Tamura, S., and Hasebe, M. (2005). Development of thermodynamic database for U-
Zr-Fe-O-B-C-FPs System,. Journal of Nuclear Science and Technology(42), 706-716.
Gauntt, R. (2005). MELCOR Computer Code Manuals, Version 1.8.6, Vol. 1: Primer and Users’
Guide, Vol. 1, Rev. 3, NUREG/CR-6119, Rev. 3, SAND 2005-5713. Sandia National
Laboratories.
Gauntt, R., Kalinich, D., Cardoni, J., Philipps, J., Goldman, A., Pickering, S., Phelan, C. (2012).
Fukushima Daiichi Accident Study. SAND2012-6173.
NEA/CSNI/R(2016)15
Gavrilas, M., and Fuller, E. (2010). US NRC´s Use of MCCI-2 Data. OCDE MCCI seminar.
Cadarache, France.
Gencheva, R., Stefanova, A., Groudev, P., Cranga, M., D. Dimov, D., Spengler, C., Foit, J. a. (2012).
SARNET benchmark on VVER 1000: Molten Core Concrete Interaction Reactor Test cases.
ERMSAR-12. Cologne, Germany.
Granovsky, V. S. (2012). Modeling of melt retention in EU-APR1400 ex-vessel core catcher. ICAPP
2012. Chicago.
Greene, G. A. (1991). Heat, Mass, and Momentum Transfer in a Multi-Fluid Bubbling Pool. Advances
in Heat Transfer, 21, 270-345.
Greene, G. A. (1992). Experimental Modeling of Heat and Mass Transfer in a Two-fluid Bubbling
Pool with Application to Molten Core-Concrete Interactions. NUREG/CR-5875.
Greene, G. A., and Irvine, T. F. (1988). Heat transfer between stratified immiscible liquid layers
driven by gas bubbling across the interface. ANS National Heat Transfer Conference.
Houston, TX.
Greene, G. A., Hodge, S., Hyman, C. R., and Tobias, M. L. (1991). The response of BWR Mark III
Containments to Short-Term Station Blackout Severe Accident Sequences. NUREG/CR-5571.
Gronager, J. E., Suo-Antilla, A., Bradley, D., and Brockmann, J. (1986). TURC1: Large Scale
Metallic Melt-Concrete Interaction Experiments and Analysis. NUREG/CR-4420.
Gronager, J. E., Suo-Anttila, A. J., and Brockmann, J. E. (1986). TURC2 and 3: Large Scale
UO2/ZrO2/Zr Melt-Concrete Interaction Experiments and Analysis. NUREG/CR-4521.
GRS. (2002). Assessment of the Accidental Risk of Advanced Pressurised Water Reactors in Germany
(Draft for Comment), GRS-184. Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) mbH.
Guéneau, C., Chatain, S., Dumas, J. C., Lechelle, J., Rado, C., Defoort, F., Konings, R. (2006).
FUELBASE: a thermodynamic database for advanced nuclear fuels. HTR2006 : Third
International Topical Meeting on High Temperature. Johannesburg, South Africa.
Guillard, G., Seropian, C., Bentaib, A., Luther, W., Bosland, L., Reinke, N. (2010). User’s Guidelines
for the integral code ASTEC V2.0 Rev. 1. ASTEC-V2/DOC/10-05.
Gurvich, L. V., Bergman, G. A., Medvedev, V. A., Veitz, I.V., and Yungman, V. S. (1989).
Thermodynamic properties of individual substances. fourth edition, Hemisphere Publ. Co, NY,
L. NY.
Gurvich, L., Iorish, V., Chekhovskoi, D. V., and Yungman, V. S. (1993). IVTANTHERMO - A
Thermodynamic Database and Software System for Personal Computer. User's guide. Boca
Raton: CRC Press, Inc.
Hackel, A., and Gröll, W. (1969). Zum Wärmeübergangsverhalten zähflüssiger Öle.
Verfahrenstechnik 3, 4, 141-145.
Hamazaki, R., Nakagawa, T., Katagiri, N., Tahara, M., Komuro, M., Tobimatsu, T., Kurita, T. (2011).
Evaluation on core melt retention in core catcher of Toshiba's EU-ABWR. ICAPP. Nice,
France.
NEA/CSNI/R(2016)15
Hammersley, R. J., Cirauquib, C., Faigc, J., and Henrya, R. (1996). Direct Containment Heating
Experiments for Vandellos and Asco Nuclear Power Plants. Nuclear Engineering and
Design(164), 287-296.
Harper, F. T. (1991). Evaluation of Severe Accident Risks: Quantification of Major Input Parameters.
Expert´s Determination of Containment Loads and Molten Core Concrete Interaction Issues.
NUREG/CR-4551. Vol. 2, Rev.1, Part 2.
Harper, F. T. (1994). Evaluation of Severe Accident Risks. Volume 2: Quantification of Major Input
Parameters. Part 6: Determination of Parameter Values not Quantified Using Formal Expert
Elicitation. NUREG/CR-4551, Rev. 1., SAND86-1309.
Helle, M., Kymalainen, O., and Tuomisto, H. (1998). Experimental Data on Heat Flux Distribution
from a Volumetrically Heated Pool with Frozen Boundaries. Workshop on In-Vessel Core
Debris Retention and Coolability. Garching, Germany.
Hellmann, S. (2000). New Experiments on the Interaction of ZrO2 Material with Corium Melts.
Report EU-INV-CIT(99)-P037 (2000).
Hellmann, S., and Fischer, M. (2007). SICOPS experiments on simultaneous interaction of molten
corium with concrete and refractory material. OECD MCCI Seminar. Aix-en-Provence,
France.
Henry, R., Epstein, M., and Fauske, H. (1982). Cooling of Debris Beds – Methods of Analysis for
LWR Safety Assessments. International Meeting on Thermal Nuclear Reactor Safety (pp.
1421-1432). Chicago, USA: NUREG/CP-0027.
Herranz, L. E., and al., e. (2012). The Effect of Safety Systems Performance on A Fukushima-Like
Scenario: ICs, S-RVs and Suppression Pools. CSARP Meeting. Bethesda, USA.
Hessheimer M.F., and. a. (2006). Containment Integrity Research at Sandia National Laboratories. An
Overview. NUREG/CR-6906.
IAEA. (2001). Implementing of Hydrogen Mitigation Techniques and Filtered Containment Venting.
IAEA Technical Committee meeting held at GRS. Cologne, Germany.
IAEA. (2004). Design of Reactor Containment Systems for Nuclear Power Plants. IAEA Safety
Standards.
IAEA. (2009). Severe Accident Management Programs for Nuclear Power Plants. Vienna, Austria:
IAEA Safety Standards Series.
IAEA. (2012). Safety of Nuclear Power Plants: Design Specific Safety Requirements. IAEA Safety
Standards Series.
Ishii, M., and Zuber, N. (1979). Drag Coefficient and Relative Velocity in Bubbly, Droplet or
Particulate Flows. AlChE Journal, 25.
Ivey, H. J. (1962). Acceleration and the Critical Heat Flux in Pool Boiling. Chartered Mechanical
Engineering, 9, 413-427.
Jackson, D. D. (1971). Thermodynamics of Gaseous Hydroxides. Lawrence Livermore Laboratories.
Journeau, C. (2003). A review of some corium physical properties. European Commission.
NEA/CSNI/R(2016)15
Journeau, C. (2005). Severe Accident Research at the PLINIUS Prototypic Corium Platform. ICAPP
’05. Seoul, Korea.
Journeau, C., and Alsmeyer, H. (2006). Validation of the COMET Bottom-Flooding Core-Catcher
with Prototypic Corium. ICAPP ’06. Reno, Nevada USA.
Journeau, C., and Haquet, J.-F. (2009). Convection Heat Transfer Anisotropy in a Bubbling Viscous
Pool—Application to Molten Core–Concrete Interaction. Nuclear Engineering and Design,
239, 389-394.
Journeau, C., and Piluso, P. (2012). Molten Core Concrete Interation. In R. J. Konings,
Comprehensive Nuclear Materials (Vol. vol 2, pp. p. 635-654).
Journeau, C., and Teisseire, B. (2015). CEA post-Fukushima R&D programmes on PWR Severe
Accidents. International Experts’ Meeting on Strengthening Research and Development
Effectiveness in the Light of the Accident at the Fukushima Daiichi Nuclear Power Plant.
Vienna: IAEA.
Journeau, C., Haquet, J.-F., Letexier, B., Greco, A., Spindler, B., Gencheva, R., Polidoro, F. (2012). A
SARNET Benchmark on Two VULCANO Molten Core Concrete Interaction Tests. 5th
European Review meeting on Severe Accident Research (ERMSAR-2012). Cologne
(Germany).
Journeau, C., Piluso, P., and Frolov, K. (2004). Corium Physical Properties for Severe Accident R&D.
ICAPP’04. Pittsburgh.
Journeau, C., Piluso, P., Correggio, P., Ferry, L., Fritz, G., Haquet, J. F., Parga, C. (2012).
Contributions of the VULCANO experimental Programme to the Understanding of MCCI
Phenomena. Nucl. Eng. Technol.(44), 261-272.
Journeau, C., Piluso, P., Haquet, J. F., Boccaccio, E., Saldo, V., Bonnet, J. M., Brissonneau, L. (2009).
Two-Dimensional Interaction of Oxidic Corium with Concretes: The VULCANO VB Test
Series. Ann. Nucl. Energy, 36, 1597.
Journeau, C., Piluso, P., Haquet, J.-F., Boccaccio, E., Ferry, L., Monerris, J.,.Correggio, P. (2010).
Oxide-Metal Corium –Concrete Interaction Tests in the VULCANO Facility. OECD-NEA
MCCI Seminar. Cadarache, France.
Kataoka, I., and Ishii, M. (1987). Drift Flux Model for Large Diameter Pipe and New Correlation for
Pool Void Fraction. Int. J. Heat Mass Transfer.
Kendell, J., and Monroe, K. P. (1971). The Viscosity of Liquids, Ideal Solution of Solids in Liquids.
Journal of the American Chemical Society, 39(9), 1802.
Khabensky, V., Granosky, V., and Bechta, S. a. (2009). Severe Accident Management Concept of the
VVER-1000 and the Justification of Corium Retention in a Crucible-type Core Catcher.
Nuclear Engineering and Technology, 41(5), 561-574.
Klein-Heßling, W., and Arndt, S. (1996). RALOC MOD4 Cycle AE: Program Reference Manual. GRS
Köln.
Klein-Heßling, W., Sonnenkalb, M., Van Dorsselaere, J. -P., Chatelard, P., Raimond, E., Clément, B.,
Auvinen, A. (2012). Ranking of Severe Accident Research Priorities. 5th European Review
meeting on Severe Accident Research (ERMSAR-2012). Cologne, Germany.
NEA/CSNI/R(2016)15
Kmetyk, L. (1994). MELCOR 1.8.2 calculations of selected sequences for the ABWR. SANDIA report
SAND94-0938.
Konsetov, V. (1966). Heat Transfer During Bubbling of Gas Through Liquid. Int. J. Heat Mass
Transfer, 9, 1103-1108.
Kulacki, F. A., and Goldstein, R. J. (1972). Thermal convection in a horizontal fluid layer with
uniform volumetric energy source. J. Fluid Mech. [2] 55, 271-287.
Kunitz, M. (1926). An Empirical Formula for the Relationship between Viscosity of Solution and
Volume of Solute. J. General Physiology, 9.
Kutateladze, S. (1951). A Hydrodynamic Theory of Changes in the Boiling Process under Free
Convection1951. Izv. Akad. Nauk. SSSR Otd, Thekh. Nauk., 4.
Kutateladze, S. S., and Malenkov, I. G. (1978). Boiling and Bubbling Heat Transfer Under Conditions
of Free and Forced Convections, 6th International Heat Transfer Conference, Toronto, Canada.
Toronto Canada.
Lambertson, W. A., and Mueller, M. H. (1953). Uranium Oxide Phase Equilibrium Systems: III, UO2-
ZrO2. J. Am. Ceram. Soc.
Langrock, G., and Hellmann, S. (2010). An Overview of Areva MCCI Laboratory-Scale Tests with
Prototypic Corium (1D/2D, Including Mixed Metallic/Oxidic Melts). OECD/MCCI-2 Seminar
- OECD-NEA MCCI Seminar. Cadarache, France.
Levy, S. (2002). Summary of coolability studies undertaken by ACE (advanced Containment
experiments)/ MACE (Melt Attack and coolability Experiments)/ACEX (ACE/MACE
extension). Non-public reference of the MACE project.
Lindholm, I., Holmstroem, S., Miettinen, J., Lestinen, V., and Hyvaerinen, J. (2006). Dryout heat flux
experiments with deep heterogeneous particle bed. Nucl. Eng. Des.(236), October 2006 2049-
2060.
Lipinski, R. J. (1980). A Particle-Bed Dryout Model with Upward and Downward Boiling. Trans.
Amer. Nucl. Soc.
Listner, C. R. (1974). On the Penetration of Water into Hot Rock. Geophys. J. R. Astron Soc, 39, 465-
509.
Lomperski, S., and Farmer, M. T. (2007). Experimental evaluation of the water ingression mechanism
for corium cooling. Nucl. Eng. Des. 237, 237, 905-917.
Lomperski, S., and Farmer, M. T. (2009). Corium Crust Strength Measurements. Nuclear Eng.
Design, 239, 2551-2561.
Lomperski, S., Farmer, M. T., and Basu, S. (2006). Experimental Investigation of Corium Quenching
at Elevated Pressure. Nuclear Eng. Design, 236, 2271.
Lopukh, D. B., Loginov, I. A., and al., e. (2000). Experimental study of the processes during water
feed onto steel melt. Nauchno-prakticheskii seminar "Voprosy besopasnosti AES s VVER,
Scientific-Practical Seminar on Problems of Safety of Nuclear Plants with VVER (in
Russian).
St. Petersburg.
Luxat, D., and Gabor, J. (2013). Fukushima Technical Evaluation: Phase 1 – MAAP5 Analysis. EPRI.
NEA/CSNI/R(2016)15
MAAP. (2011). EPRI, Transmittal Document for MAAP5 Code Revision MAAP5.0.1, FAI/11-1161,
November, 2011.
Magallon, D., Maillat, A., Seiler, J. -M., Atkhen, K., Sjövall, H., Dickinson, S., Tinkler, C. (2005).
European expert network for the reduction of uncertainties in severe accident safety issues
(EURSAFE). Nuclear Engineering and Design(235), 309-346.
Maruyama, Y., Kojima, Y., Tahara, M., Nagasaka, H., Kato, M., Kolodeshnikov, A. A., Vassiliev, Y.
S. (2006). A Study on Concrete Degradation During Molten Core/Concrete Interactions.
Nuclear engineering and Design, 236, 2237-2244.
Maruyama, Y., Tahara, M., Nagasaka, H., and Vassiliev, Y. S. (2002). Recent Results of MCCI
Studies in COTELS Project. Third Korea-Japan Symposium on Nuclear Thermal hydraulics
and Safety (NTHAS3). Kyeongju, Korea.
Mayinger, F., Jahn, M., Reineke, H. H., and Steinbrenner, V. (1976). Examination of
Thermalhydraulic Processes and Heat Transfer in a Core Melt, BMFT RS 48/1, Institute für
Verfahrenstechnik der T.U. Hannover, 1976. Hannover Germany.
Meyer, L., Albrecht, G., Caroli, C., and Ivanov, I. (2009). Direct Containment Heating Integral Effects
tests in geometries of European nuclear power plants. Nuc. Eng. Des.(239), 2070-2084.
Miassoedov, A., Alsmeyer, H., Meyer, L., Steinbruck, M., Groudev, P., Ivanov, I., and Sdouz, G.
(2008). Results of the QUENCH-L2, DISCO-L2, and COMET-L2 Experiments Performed
wiithin the LACOMERA Project at the Forschungszentrum Karlsrühe. Nuclear Engineering
and Design, 238, 2017-2026.
Miassoedov, Alsmeyer, H., Cron, T., and Foit, J. (2010). The COMET-L3 Experiment on Long-Term
Melt–Concrete Interaction and Cooling by Surface Flooding. Nuclear Engineering and
Design, 240, 258-265.
Michel, B. (2015). New set of convective heat transfer coefficients established for pools and validated
against CLARA experiments for application to corium pools. Nuclear Engineering and
Design, 286.
Moiseenko, E. V., Filippov, A. S., Drobyshevsky, N. I., Strizhov, V. F., and Kisselev, A. E. (2011)..
Varia - an Application for Mass Computing and Statistical Analysis of the Simulation Results
in BEPU Safety Assessment. Proc. ICONE19 May 16-19, 2011, Chiba, Japan. ICONE19.
Chiba Japan.
Nagasaka, H., Kato, M. S., Cherepnin, Y., Vasilyev, Y., Kolodeshnikov, A., Zhdanov, V., and Zuev,
V. (1999). COTELS Project(1): Overview of Project to study FCI and MCCI during a Severe
Accident. OECD Workshop on Ex-vessel Debris Coolability, (pp. 285-292). Karlsrühe,
Germany.
Nagasaka, H., Sakaki, I., Kato, M., Vasilyev, Y., Kolodeshnikov, A., and Zhdanov, V. (1999).
COTELS Project(3): Ex-vessel Debris Cooling Tests. OECD Workshop on Ex-vessel Debris
Coolability, (pp. 302-308). Karlsrühe, Germany.
Nazare, G., Ondracek, G., and Shulz, G. (1977). “Properties of Light Water Reactor Core Melts,”
Nucl. Tech., 32, 239 (1977). 32.
Nie, M. (2004). ‘Temporary Melt Retention in the Reactor Pit of the European Pressurised Water
Reactor (EPR). PhD Thesis, IKE University of Stuttgart.
NEA/CSNI/R(2016)15
Nie, M. (2005). Temporary Melt retention in the reactor pit of the European Pressurised Water
Reactor (EPR). Dr.Ing. Thesis, Universität Stuttgart, Germany.
Nie, M., Fischer, M., and Lohnert, G. (2007). Advanced MCCi modelling based on stringent coupling
of thermalhydraulics and real solution thermochemistry in COSACO. ICONE-10. Arlington,
VA, USA.
NRC. (1985). Estimates of Early Containment Loads from Core Melt Accidents. NUREG/CR-1079.
NRC. (1990). Evaluation of Severe Accident Risks: Surry Unit 1. NUREG/CR-4551, Vol. 3, Rev. 1,
Part 2.
NRC. (1990). Evaluation of Severe Accidents Risks: Peach Bottom, Unit 2. NUREG/CR-4551 Vol. 4,
Rev. 1, Part 2.
NRC. (1990). Evolutionary Light Water Reactor (LWR) certification issues and their relationship to
current regulatory requirements. SECY-90-016.
NRC. (1990). Severe Accident Risks: An Assessment of Five US Nuclear Power Plants Accident
Analysis. NUREG-1150.
NRC. (1993). Core-Concrete Interactions with Overlying Water Pools: The WETCOR-1 Test.
NUREG/CR-5907; SAND92-1563.
NRC. (1993). Evaluation of Severe Accident Risks: Zion Unit 1. Appendix A. NUREG/CR-4551 Vol.7,
Rev.1, Part 2A.
NRC. (1993). Policy, Technical, and Licensing Issues Pertaining to Evolutionary and Advanced Light-
Water Reactor (ALWR) Designs, (Item II Q). SECY-93-087.
NRC. (1993). Severe Accident Source Term Characteristics for Selected Peach Bottom Sequences
Predicted by the MELCOR Code. NUREG/CR-5942..
NRC. (1997). MELCOR 1.8.4 - Reference Manual. USA, Washington.
NRC. (2010). ESBWR Design Certification Probabilistic Risk Assessment. NEDO-33201, Revision 6.
NRC. (2011). Chapter 6.5, Primary Containments. In Manual, General Electric Advanced Technology.
NRC ADAMS Accession Number ML11263A359.
NRC. (2012). Consideration of additional requirements for containment venting systems for Boiling
Water Reactors with MARK I and MARK II containments. Enclosure 5a: MELCOR Accident
Analysis. SECY-12-0157.
NRC. (2012). Issuance of order to modify licenses with regard to requirements for mitigation
strategies for beyond design-basis external events. ORDER EA-12-049.
NRC. (2013). US Nuclear regulatory framework commission staff recommendation for the disposition
of recommentation 1 of the near-term task force report. SECY 2013-0132.
OECD/NEA. (1988). Proceedings of the specialist meeting on filtered containment venting systems.
NEA/CSNI-148. Paris, France.
OECD/NEA. (1995). SESAM. Severe Accident Management Implementation. Senior Group of Eperts
on Severe Accident Management (SESAM).
NEA/CSNI/R(2016)15
OECD/NEA. (1997). OECD/NEA/CSNI Workshop on the implementation of hydrogen mitigation

techniques. Winnipeg, Canada.
OECD/NEA. (2015). CSNI BSAF project phase 1 final report.
OECD/NEA. (Under preparation). Safety Research Opportunities Post-Fukushima - SAREF.
Ozrin, V., Tarasov, O., Strizhov, V., and Filippov, A. (2010). A model for calculating composition and
density of the core melt in the water-moderated water-cooled reactor in case of severe
accident. Thermal Engineering.
Pankrantz, L. (1982). Thermodynamic Properties of Elements and Oxides. US Bureau of Mines
Bulletin, 672.
Park, C. K., and al. (1992). Evaluation of Severe Accident Risks: Zion Unit 1. NUREG/CR-4551 Vol.
7, Rev. 1, Part 2A.
Parozzi, F. (2010), Fast Running Models for Molten Corium Coolability for safety Analysis of
Nuclear Reactors, 65° Congresso Nazionale ATI, 13 september 2010. Congresso Nazionale
BALI. Sardinia.
Parrozi, F., and R. Fontana (2013), CORIUM2D Advanced Code, User’s Manual, RSE report.
Bologne.
Payne, A. C., and al., e. (1990), Evaluation of Severe Accidents Risks: Peach Bottom, Unit 2. Main
Report. NUREG/CR-4551. Vol. 4, Rev.1, Part 1.
Pazukhin, E. M. (1994), Fuel Containing Lavas of the Chernobyl NPP Fourth Block: Topography,
physicochemical properties and formation scenario. Radiochemistry, 36, pp. 109-154,.
Peehs, M., Skokan, A., and Reimann, M. (1979). The Behaviour of Concrete in Contact with Molten
Corium in the Case of a Hypothetical Core Melt Accident. Nuclear Eng. and Design, 46, 192–
198.
Piluso, P. (2003). CORPRO V1.5, Base de données propriétés physiques du corium. CEA.
Polidoro, F. (2013). Preliminary Assessment of the Thermal Interaction Among Corium and In-vessel
Core Catcher in a Large Sodium Fast Reactor, The 15th International Topical Meeting on
Nuclear Reactor Thermal – Hydraulics. NURETH-15. Pisa, Italy.
Powers, D. A., and Arellano, F. E. (1982-1). Large-Scale Transient Tests of the Interaction of Molten
Steel with Concrete. US Nuclear Regulatory Commission. Sandia National Laboratory.
Powers, D. A., and Arellano, F. E. (1982-2). Direct Observation of Melt Behavior during High
Temperature Melt/Concrete Interactions. NUREG/CR-2283.
Powers, D. A., Brockmann, J., and Shiver, A. W. (1986). VANESA: A mechanistic Model of
Radionuclide Release and Aerosol Generation during Core Debris Interactions with
Concrete. NUREG/CR-4308.
Powers, D. A., Dahlgren, D. A., Muir, J. F., and Murfin, W. D. (1978). Exploratory Study of Molten
Core Material/Concrete Interactions. SANDIA-77-2042, Sandia National Laboratories.
Ramacciotti, M., Journeau, C., Sudreau, F., and Cognet, G. (2001). Viscosity Models for Corium
Melts. Nuclear Engineering and Design, 204, 377–389.
NEA/CSNI/R(2016)15
Ricou, F., and Spalding, D. (1961). Measurements of entrainment of axisymmetrical turbulent jets. J.
Fluid Mechanics, 11, 21-32.
Roark, R. J., and Young, W. C. (1975). Formulas for Stress and Strain, 6th Ed. McGraw Hill.
Robb, K. R. (2014). Ex-vessel Core Melt Modeling Comparison Between MELTSPREAD-
CORQUENCH and MELCOR 2.1. ORNL/TM-2014/1.
Robb, K. R., and Corradini, M. L. (2009). Ex-Vessel Corium Coolability Sensitivity Study with the
CORQUENCH Code. NURETH-13. Kanazawa City, Ishikawa Prefecture, Japan.
Robb, K. R., and Corradini, M. L. (September 2011). Melt Eruption Modeling for MCCI Simulations.
NURETH-14. Toronto,Ontario,Canada.
Robb, K. R., Francis, M. W., and Farmer, M. T. (2013). Enhanced Ex-Vessel Analysis for Fukushima
Daiichi Unit 1: Melt Spreading and Core-Concrete Interaction Analyses with MELTSPREAD
and CORQUENCH. ORNL/TM-2012/455.
Robb, K., and Corradini, M. (2010). Towards understanding melt eruption phenomena during molten
corium concrete interactions. ICONE 18. Xi’an, China.
Robledo, F., and Lantaron, A. (1999). Spanish Regulatory Perspective on Ex-vessel Corium
Coolability Issues. OECD Workshop on Ex-vessel Debris Coolability. Karlsrühe, Germany.
Roche, M. F., Leibowitz, L., and Fink, J. K. (1993) Solidus and Liquidus Temperatures of Core-
Concrete Mixtures. NUREG/CR-6032.
Roche, M. F., Steidl, P. F., Leibowitz, L., Fink, J. K. and Sehgal, B. R. (1993). Viscosity of Corium
Concrete Mixtures at High Temperatures. ACE-TR-C37.
Rohsenow, W. M. (1952). A Method of Correlating Heat Transfer for Surface Boiling of Liquids,.
Trans. ASME, 74, 979-976.
Sauvage, E., Prior, R., Coffey, K., and Mazurkiewicz, S. M. (2006). OSSA – An optimized approach
to severe accident management: EPR application. ICAPP'06. Reno - Nevada (United States).
Schwinges, B., Journeau, C., Haste, t., Meyer, L., Tromm, W., and Trambauer, K. (2008). Final
Evaluation of Severe Accident Research Priorities and Proposal Research Activities.
SARNET.
Sdouz, G., Mayrhofer, R., Alsmeyer, H., Cron, T., Fluhrer, B., Foit, J., Wenz, T. (2006). The COMET-
L2 Experiment on Long-Term MCCI with Steel Melt. FZKA 7214 / SAM-LACOMERA-D15,
Forschungszentrum Karlsrühe.
Sehgal, B. R. (2006, October). Stabilization and Termination of Severe Accidents in LWRs. Nuclear
Engineering and Design, 236(19-21), 1941-1964.
Sehgal, B. R. (2011). Nuclear Safety in Light Water Reactors: Severe Accident Phenomenology.
Academic Press.
Sehgal, B. R., and Spencer, B. W. (1992). ACE Program Phase C: Fission Product Release from
Molten Corium Concrete Interaction MCCI,. Second OECD-CSNI Specialist Meeting on
Molten Core Debris-Concrete Interactions. Karlsrühe, Germany.
Seiler, J. -M. (1996). Phase segregation model and molten pool thermalhydraulics during molten core
concrete interaction. Nucl. Eng. Des., 166, 259-267.
NEA/CSNI/R(2016)15
Seiler, J. -M., and Froment, K. (2000). Material effects on Multiphase Phenomena in the Late Phases
of Severe Accident of Nuclear Reactors, Multiphase Sci. technol., 12, 117-257, 2000. 12.
Seiler, J. -M., and Ganzhorn, J. (1997). Viscosities of corium concrete mixtures. Nucl. Eng. Des., 178,
259-268.
Seiler, J. -M., Azarian, G., Gandrille, P., Dumontet, A., Dutheillet, Grange, J. L., Valin, F. (1999).
GAREC Analyses in Support of Ex-Vessel Retention Concept. OECD Workshop on Ex-vessel
Debris Coolability. Karlsrühe, Germany.
Seiler, J. -M., Froment, K., and Defoort, F. (2003). Theoretical analysis for corium pool with
miscibility gap. Nucl. Technol., 141.
Seiler, J.-M., Sehgal, B. R., Alsmeyer, H., Kymalainen, O., Turland, B., L., G. J., Cirauqui, C. J.
(2003). Analysis of corium recovery concepts by the EURO-CORE group. Nucl. Eng.
Des.(221), 119-136.
Sevón, T. (2008). A Heat Transfer Analysis of the CCI Experiments. Nucl. Eng. Design, 238(1-3),
2377.
Sevón, T. (2011). A heat transfer Analysis of the CCI and VULCANO experiments. Research report
VTT R-00483-11.
Sevón, T., Kinnunen, T., Virta, J., Holmström, S., Kekki, T., and Lindholm, I. (2010). HECLA
Experiments on Interaction Between Metallic Melt and Hematite-Containing Concrete.
Nuclear Engineering and Design, 240, 3586-3593.
Shaw, H. R. (1972). Viscosities of Magmatic Silicate Liquids: An Empirical Method of Prediction.
American J. of Science, 272, 870.
Sidorov, A. (2000). Core melt localising facility for Tianwan NPP. Design and operation. Applied
Research Workshop “Safety issues of NPP with WWER”. St. Petersburg, Russia.
Siviour, J. B., and Ede, A. J. (1970). Heat Transfer in Subcooled Film Boiling. Fourth International
Heat Transfer Conference. Paris, France.
Skokan, A., Holleck, H., and Peehs, M. (1979). Chemical Reactions Between Light Water Reactor
Core Melt and Concrete. Nuclear Technology, 46, 255–262.
SNL. (2012). State-of-the-Art Reactor Consequence Analyses Project Volume 1: Peach Bottom
Integrated Analysis. Sandia National Laboratories.
SNL. (2013). State-of-the-Art Reactor Consequence Analyses Project. Uncertainty Analysis of the
Unmitigated Long-Term Station Blackout of the Peach Bottom Atomic Power Station. Draft
Report.
Spencer, B., Farmer, M. T., Armstrong, D. R., Kilsdonk, D. J., Aeschlimann, R. W., and Fischer, M.
(1992). Results of MACE Tests M0 and M1. Second OECD-CSNI Specialist Meeting on
Molten Core Debris-Concrete Interactions. Karlsrühe, Germany.
Spengler, C. (2007). Theoretical Manual for GRS options in corium/concrete interface in MEDICIS.
Spengler, C. (2012). Analysis of Top Flooding During Molten Corium Concrete Interaction (MCCI)
with the Code MEDICIS using a Simplified Approach for the Combined Effect of Crust
Formation and Boiling. Annular Meeting of Nuclear Technology. Stuttgart, Germany.
NEA/CSNI/R(2016)15
Spengler, C. (2012). Validation of the MCCI-Module COCOSYS-MEDICIS, Technical Note. GRS-V-

RS1190-1/2012.
Spengler, C. (2013). A Simplified Quasi-Stationary Model for Investigating the Relation between
Corium Viscosity and Corium Temperature in Molten Corium Concrete Interaction (MCCI).
NURETH-15. Pisa, Italy.
Spengler, C., Fargette, A., Foit, J., Agethen, A., and Cranga, M. (2013). Transposition of 2D Molten
Corium-Concrete Interactions (MCCI) from Experiment to Reactor. ERMSAR-2013. Avignon
(France).
Spindler, B., and Seiler, J. M. (2006). Simulation of MCCI with the TOLBIAC-ICB code based on the
phase segregation model. Nuclear Engineering and Design, 236, 2264-2270.
Spindler, B., and Seiler, J. M. (2010). Lessons learnt from numerous molten core concrete interaction
tests: geometry effects, concrete type effects, physico-chemistry analysis. MCCI-OECD
seminar. Cadarache, St-Paul lez Durance, France.
Spindler, B., Atkhen, K., Cranga, M., Fischer, M., Kawabe, R., Kim, H., Spengler, C. (2007).
Simulation of Molten Corium Concrete Interaction: The CCI-2 Benchmark Organized in the
Frame of the OECD-MCCI Project. MCCI Project Seminar. Cadarache, France.
Spindler, B., Atkhen, K., Cranga, M., Foit, J., Garcia, M., Schmidt, W., Spengler, C. (2007).
Simulation of Molten Corium Concrete Interaction in a Stratified configuration: the COMET-
L2-L3 Benchmark,. ERMSAR-07.
Spindler, B., Tourniaire, B., Seiler, J. -M., and Atkhen, K. (2005). MCCI analysis and applications
with the TOLBIAC-ICB code based on phase segregation model. ICAPP’5.
Spindler, B., Tourniaire, B., Vandroux, S., Seiler, J., and Gubaidullin, A. (2005). MCCI Analysis
Based on the Phase Segregation Model with Application in the TOLBIAC-ICB Code.
NURETH-11. Avignon, France.
Stedman, J., Evans, J., and Woodthorpe, J. (1990). Rheology of composite ceramic injection moulding
suspensions. Journal of Material Science, 25, 1833–1841.
Sudreau, F., Ramacciotti, M., Cognet, G., Seiler, J. -M., and Journeau, C. (2000). Calculation
methodology for corium viscosity, Report. European Commission Euratom Research
Framework Program 1994-1998, Nuclear Fission Safety.
Suzuki, Y., Tahara, M., Kurita, T., Hamazaki, R., and Morooka, S. (2009). Heat Removal Capability
of Core-catcher with Inclined Cooling Channels. ICAPP'09. Tokyo, Japan.
Tarbell, W. W., Bradley, D. R., Blose, R. E., Ross, J. W., and Gilbert, R. W. (1987). Sustained
Concrete Attack by Low Temperature, Fragmented Core Debris. NUREG/CR-3024.
Theerthan, S., Giri, A., Karbojian, A., and Sehgal, B. R. (2002). EC-FOREVER experiments on
thermal and mechanical behavior of a reactor pressure vessel during a severe accident.
Stockholm, Sweden: Royal Institute of Technology.
Theofanous, T. G., Amarasooriya, W. H., Yan, H., and and Ratnam, U. (1991). The Probability of
Liner Failure in a Mark I Containment. NUREG/CR-5423.
Theofanous, T. G., Liu, C., and Yuen, W. W. (1998). Coolability and Quench of Corium-Concrete
Interactions by Top-Flooding. MACE-TR-D14.
NEA/CSNI/R(2016)15
Theofanous, T. G., Yan, H., and Podowski, M. Z. (1993). The probability of Mark-I Containment
Failure by Melt-Attack of the Liner. NUREG/CR-6025.
Thomas, D. (1965). Transport characteristics of suspensions. Journal of Colloid Science, 20, 267-277.
Thompson, D. H., and Fink, J. K. (1988). ACE MCCI Test L2: Test Data Report Volume I - Thermal
Hydraulics. Argonne National Laboratory.
Thompson, D. H., and Fink, J. K. (1989). ACE MCCI Test L5: Test Data Report Volume I - Thermal
Thompson, D. H., and Fink, J. K. (1991a). ACE MCCI Test L6: Test Data Report Volume I - Thermal
Thompson, D. H., Farmer, M. T., Fink, J. K., Armstrong, D. R., and Spencer, B. W. (1997).
Compilation, analysis and interaction of ACE Phase C and MACE experimental data.
Chicago, IL, USA: Argonne National Laboratory.
Thompson, D. H., Fink, J. K., Armstrong, D. R., Spencer, B. W., and Sehgal, B. R. (1992). Thermal-
Hydraulic Aspects of the Large-Scale MCCI Tests in the ACE Program. 2nd CSNI Specialist
Meeting on Core Debris-Concrete Interactions. Karlsrühe, Germany,.
Thompson, D., and Fink, J. (1991b). ACE MCCI Test L7: Test Data Report Volume I - Thermal
Tourniaire, B. (2006). A Heat Transfer Correlation Based on a Surface Renewal Model for Molten
Core Concrete Interaction Study. Nucl. Eng. Des, 236, 10-18.
Tourniaire, B., and Bonnet, J.-M. (2003). Study of the Mixing of Immiscible Liquids by Sparging Gas:
Results of the BALISE Experiments. NURETH 10. Seoul, Korea.
Tourniaire, B., and Seiler, J. -M. (2004). Modeling of viscous and inviscid fluid ejection through
orifices by sparging gas. ICAPP'04,. Pittsburgh, PA USA.
Tourniaire, B., and Varo, O. (2008). Assessment of two-phase flow heat transfer correlations for
molten core-concrete interaction study. Nucl.Tech., 164, 143-151.
Tourniaire, B., Boulin, A., and Haquet, J. -F. (2015). Simulation of MCCI with top flooding with the
TOLBIAC-ICB code. ICAPP'2015. Nice, France.
Tourniaire, B., Dufour, E., and Spindler, B. (2009). Foam formation in oxidic pool with application to
MCCI real materials experiments. Nuclear Engineering and Design(239).
Tourniaire, B., Seiler, J. -M., and Bonnet, J. -M. (2003). Study of the Mixing of Immiscible Liquids:
Results of the BALISE experiments. NURETH-10. Seoul Korea.
Tourniaire, B., Seiler, J. -M., Bonnet, J. -M., and Amblard, M. (2000). Liquid Ejection Through
Orifices by Sparging Gas – The PERCOLA Program. ICONE-10. Arlington, VA USA.
Tourniaire, B., Seiler, J., and Bonnet, J. M. (2003). Study of the Mixing of Immiscible Liquids by
Sparging Gas: Results of the BALISE Experiments. NURETH 10. Seoul, Korea.
Tourniaire, B., Seiler, J.-M., Bonnet, J.-M., and Amblard, M. (2006). Experimental Study and
Modeling of Liquid ejection through orifices by sparging gas. Nucl. Eng. And Design, 236,
2281-2295.
NEA/CSNI/R(2016)15
Tourniaire, B., Spindler, B., and Guillaumé, M. (2010). Transient conduction heat transfer modeling in
concrete for the simulation of long-term phase of molten core-concrete interaction. Nuclear
Technology, 170, 201-209.
Tromm, W., Foit, J., and Magallon, D. (2000). Dry and wet spreading experiments with prototypic
materials at the FARO facility and theoretical analysis. Wiss. Ber. FZKA.
Urbain, G. (1987). Viscosity Estimation of Slags. Steel Research, 58, 111-116.
Urbain, G., Bottinga, Y., and Richet, P. (1982). Viscosity of liquid silica, silicates and alumino-
silicates. Geochimica and Cosmochimica Acta, 46, 1061–1072.
Veteau, J. M. (2006). ARTEMIS Program: Investigation of MCCI by Means of Simulating Material
Experiments. ICCAP’06. Reno, Nevada, USA.
Wallis, G. B. (1979). One-Dimensional Two-Phase Flow, 2nd Ed. New York: McGraw Hill.
Weast, R. C. (1983). Handbook of Chemistry and Physics, 63rd Edition. CRC Press.
WENRA. (2013). RHWG Report on Safety of new NPP designs. Retrieved from
http://www.wenra.org/publications/.
WENRA. (2014). RHWG Safety Reference Levels for Existing Reactors. Retrieved from
http://www.wenra.org/publications/.
Werle, H. (1982). Enhancement of Heat Transfer between two Horizontal Liquid Layers by Gas
Injection at the Bottom. Nuclear Technology, 59, 160-164.
Yamanaka, Y. (2015). Fukushima information update. 4th SAREF meeting. Tokyo.
Yang, J. W. (1990). PWR Dry Containment Issue Characterization. NUREG/CR-5567.
Zhdanov, V., Vasilyev, Y., Kolodeshnikov, A., Cherepnin, Y., Sakaki, I., and Nagasaka, H. (1999).
COTELS Project(4): Structural Investigation of Solidified Debris in MCCI. OECD Workshop
on Ex-vessel Debris Coolability, (pp. 309-316). Karlsrühe, Germany.
Zuber, N. (1958). On Stability of Boiling Heat Transfer. Trans. ASME, 80, 711-720.
Zvonarev, Y., Tsurikov, D., Kobzar, V., Volchek, A., Kiselev, N., Strizhov, V., and Filippov, A. a.
(2011). Numerical Analysis of Core Catcher Efficiency for VVER-1200. Physics of Atomic
Nuclei, 74(13), 1845-1853.
Nuclear Safety
2017
NEA
2017
State-of-the-Art Report on Molten Corium

Concrete Interaction and Ex-Vessel Molten
Core Coolability
Activities carried out over the last three decades in relation to core-concrete interactions and melt
State-of-the-Art Report
coolability, as well as related containment failure modes, have significantly increased the level of
understanding in this area. In a severe accident with little or no cooling of the reactor core, the residual on Molten Corium Concrete
State-of-the-Art Report on Molten Corium Concrete Interaction and Ex-Vessel Molten Core Coolabilitys
decay heat in the fuel can cause the core materials to melt. One of the challenges in such cases is to
determine the consequences of molten core materials causing a failure of the reactor pressure vessel.
Molten corium will interact, for example, with structural concrete below the vessel. The reaction between
corium and concrete, commonly referred to as MCCI (molten core concrete interaction), can be extensive
Interaction and Ex-Vessel
and can release combustible gases. The cooling behaviour of ex-vessel melts through sprays or flooding
is also complex. This report summarises the current state of the art on MCCI and melt coolability, and
thus should be useful to specialists seeking to predict the consequences of severe accidents, to model
Molten Core Coolability
developers for severe-accident computer codes and to designers of mitigation measures.
Nuclear Energy Agency (NEA)

46, quai Alphonse Le Gallo
92100 Boulogne-Billancourt, France
Tel.: +33 (0)1 45 24 10 15 NEA
nea@oecd-nea.org www.oecd-nea.org NEA No. 7392

Corium Processing

Uploaded by

Copyright:

Available Formats

Corium Processing

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Corium Processing

Uploaded by

Copyright:

Available Formats

Nuclear Safety and Regulation

State-of-the-Art Report on Molten Corium

NUCLEAR ENERGY AGENCY

NUCLEAR ENERGY AGENCY

Corrigenda to OECD publications may be found online at: www.oecd.org/publishing/corrigenda.

List of abbreviations and acronyms....................................................................................................15

4.6. Scaling to reactor case ...........................................................................................................205

3.3-4: Pool/concrete interface assuming a corium crust in thermodynamic equilibrium

3.3-7: Features of siliceous concretes ......................................................................................... 181

List of abbreviations and acronyms

ABI Automated ball indentation

AFNOR Association française de normalisation

CFD Computational fluid dynamics

ENSREG European Nuclear Safety Regulators Group

HECLA Computer code

LBLOCA Large break loss-of-coolant accident

SAHRS Severe accident Heat removal system

Lessons learnt from experiments

Capabilities and shortcomings in current codes and models

Remaining issues in the areas of MCCI and debris coolability

Long term core-concrete interaction behaviour

Extrapolation of existing knowledge to long term MCCI processes

Realistic plant simulations

Presence of metal within the melt or within the concrete

Presence of impurities in cooling water

Methods for improving melt coolability under top flooding conditions

1. Introduction to the main phenomena involved: Brief description

1.1. The concept of melt stabilisation and severe accident termination

1.2. Corium discharge from the RPV into the containment

1.3. Corium concrete interaction

1.3.1.1. Corium chemical thermodynamics

1.3.2. Concrete characteristics

• From 300 ºC. Increase in porosity and microcracking.

1.3.3.1. Radial vs axial ablation

Time after scram Decay power (W/Kg U)

CORIUM MELT POOL

POSSIBLE LOWER CRUST

DRAWING: TRADITIONAL MCCI MODEL

1.4. Relevant containment failure modes due to ex-vessel corium behaviour

Figure 1.4-1: Example of drywell head flange connection details

Figure 1.4-2: Drywell head flange area versus pressure

1.4.3. Failure of the reactor pressure vessel pedestal in BWR

Table 1.4-1: Timing of key events for MELCOR calculations

Figure 1.4-3: Comparison of drywell pressures for selected cases

1.4.7. Safety issues unaddressed by this report

1.5. Severe accident management issues and back-fitting measures

1.5.2. Reactor cavity flooded before RPV failure at low pressure

Parameter 2D Analysis Range Units

Cavity Pressure 0.1-0.5 MPa

Water Add Time Delay 5-120 min.

Time Since Reactor Shutdown 1-8 hrs.

Melt Initial Temp 2400-2900 K

Melt Initial Mass 45-180 tonne

Zr Initial Oxidation 20-80 w%

Cavity Diameter 5-10 m

Particle Bed Porosity 33-59 %

Crust Permeability -11 m2

Figure 1.5-1: Different basic measures for retention and coolability

1. MCCI: molten pool configuration and concrete ablation

2. Molten core-concrete interaction and debris coolability