Proceedings of the 4th Ulcos seminar

1st & 2nd October 2008

New Direct Reduction (DR)

Iron ore Reduction with CO and H2 Gas Mixtures

Thermodynamic and Kinetic Modelling
Raymond Longbottom and Leiv Kolbeinsen1
1

Department of Materials Science and Engineering, Norwegian University of Science and Technology
(NTNU), Trondheim, Norway
leiv.kolbeinsen@material.ntnu.no

The reduction of iron ore pellets has been studied using different techniques. Thermodynamic studies, experimental investigations and mathematical modelling have all been undertaken to better understand the behaviour
of different pellet types in the new direct reduction process. The mathematical pellet model gives a good fit to
most of the experimental conditions used in this work. There are some discrepancies between the experimental
and calculated results under certain conditions, which are thought to be due to limitations in the experimental
set up rather than fundamental issues in the model. The micromodel indicates that the hematite within the
pellets is reduced to magnetite quickly, which in turn is reduced fairly quickly to wstite. The reduction of
wstite to metallic iron seems to be the limiting stage in the reduction of the pellets, which is in line with what
would be expected.
bon Capture (CC) as well as alternative process
schemes (e. g. Wiberg) can be investigated
through computer model simulations and verified
experimentally using the TGA.

Introduction
The reduction of iron ore pellets has been studied
using different techniques. Thermodynamic studies,
experimental investigations and mathematical modeling have all been undertaken to better understand
the behavior of different pellet types in the new
direct reduction process.

A robust and flexible model for reduction/oxidation

is, of course together with the TGA, the core of a
laboratory investigation of possible ways of integrating and verifying knowledge and competence on
both the micro and the macro scales of this type of
processes.

Experimentally, it can be seen that the pellets reduce

quickly under the proposed conditions. However, the
reaction is very endothermic and causes a large
temperature drop in the sample under the small
scale (250 g) conditions that have been studied. The
experimental equipment gives consistent results with
good repeatability. A single pellet (or particle) model
(the micro model) is integrated into a model for the
whole process as well as for the pellets in the retort
of a laboratory scale Thermo Gravimetric Analyser
(TGA), and this facilitates the following main applications:

Model overview
The most frequently used reduction gases are CO
and H2.
Reduction with CO: When the initial state is hematite, and the temperature is over 570C, reduction of
iron oxide will occur in three steps:

Fe 2 O3 Fe3 O 4 FeO Fe

1. Combining mathematical model of retort samples

of ~100 single pellet with TGA measurements allow for determination of parameters of the equations describing the resistances in the model

Reactions at 25C are:

o

3Fe 2 O 3 +CO R 2Fe3 O 4 +CO 2 H 298 =-7.8 kJ molFe

Fe3 O 4 +CO R 3FeO+CO 2

2. Combining mathematical model retort samples

with mathematical model of process makes it possible to calculate profiles of gas concentration and
temperature distributions in the process (e. g.
shaft).

FeO+CO R Fe+CO 2

H o298 =+11.2 kJ molFe

H o298 =-15.7 kJ molFe

Summing these reactions gives reduction of hematite

to iron:

3. Computed gas and temperature profiles can be

used to control input values to the TGA, and the
resulting recorded change of sample weight as
function of time and result of investigations of the
pellet sample after the experiment will give extra
verifications of the model and parameters

Since wstite is meta-stable below ~570C, reduction should occur in only two steps below this temperature. Magnetite reduces directly to metallic iron
without first being converted to wstite:

4. Effects of changing the gas & temperature profiles

by alternative methods for Syn Gas (SG) and Car-

Fe3 O4 +4CO R 3Fe+CO 2

Fe 2 O3 +3CO R 2Fe+3CO 2

H o298 =-13.3kJ molFe

H o298 =-4.5 kJ molFe

Reduction with H2: This reduction is similar to the

one with CO. It occurs in three steps when temperatures are over 570C and two steps when the temperature is below 570C:

will occur so that the Water Gas Shift Reaction

(WGS) equilibrium:

3Fe 2 O3 +H 2 R 2Fe3 O 4 +H 2 O (g) H o298 =-1.0 kJ molFe

is satisfied. This reaction moves excess oxygen between the c- and h- part of the gas mixture, without any changes in oxidation degree of the gas.

CO + H 2 O = CO2 + H 2

Fe3 O 4 +H 2 R 3FeO+H 2 O (g) H o298 =+24.9 kJ molFe

FeO+H 2 R Fe+H 2 O(g)

o
H 298
= 41.1 kJ mol

H o298 =+25.4 kJ molFe

Fe3 O 4 +4H 2 R Fe+4H 2 O (g) H o298 =+50.4 kJ molFe

Fe 2 O3 +3H 2 R 2Fe+3H 2 O (g) H o298 =+49.4 kJ molFe
If one compares the Baur-Glassner diagrams in Figure 1 below it is seen that from gas utilisation point
of view hydrogen is the best reduction gas dealing
with high temperatures, while CO-gas is the best at
low temperatures. From an enthalpy view point,
hydrogen reduction is generally endothermic, and
reduction with carbon monoxide is generally mildly
exothermic.

Figure 2: Water Gas Shift Reaction (WGSR) equilibrium

diagram in general; definition of variables. R denotes unoxidized gas species (CO + H2) and RO denotes the oxidised
gas species (CO2 + H2O). Temperature axis is normal to
the paper and T2 is lower than T1.

The dash-dot line in Figure 2 indicates the WGSR

equilibrium line for a gas mixture with given c/h-ratio
and a given degree of oxidation () in the gas. Using
R for the unoxidized gas species (CO + H2) and RO
for the oxidised gas species (CO2 + H2O) this line is
given by:

Figure 1: a) Stability of iron oxides as function of temperature and oxygen potential, b) Baur-Glassner diagrams
show realisation of oxygen potential in form of CO2- and
H2O-content in mixtures with CO and H2 respectively.

h = hc c +
Temperature change will shift the equilibrium position along this line depending on temperature, while
oxygen pick-up by the gas will move the line to the
right on the diagram without changing its slope. The
slope will change as a result of e.g. carburisation of
solids or removal of water or carbon dioxide from the
gas.

Reduction with CO-H2 mixture: Often a mixture between the two gases, CO and H2, are used as reduction gas. The ratio between C/H varies dependent on
source and production-route. The most interesting
source is the use of natural-gas, methane. It is not
possible to use methane as it is; it must be converted to CO and H2. This occurs by using partial
combustion or reforming with CO2 and/or H2O.

Model description

Partial combustion:

2CH 4 + O2 = 2CO + 4 H 2

o
H 298
= 35.7 kJ mol CH 4

Reforming:
o
CH 4 + CO2 = 2CO + 2 H 2 H 298
= 247.3 kJ mol CH 4
o
CH 4 + H 2 O = CO + 3H 2 H 298
= 206.2 kJ mol CH 4

When both CO and H2 are present all the reaction

R. Longbottom and L. Kolbeinsen

Reduction of a pellet can roughly be depicted by the

well known Shrinking Core Model (SCM), but this
needs to be modified to include the intermediate
products magnetite and wstite that exist between
hematite and metallic iron. A single pellet is represented as a small ball with hematite in the centre
followed by magnetite, wstite and then metallic iron
in concentric layers. The reducing gas in the model
should at least have two active components; hydrogen and carbon monoxide, and should include the

New Direct Reduction (DR) - October 2008

possible internal gas reactions like the water gas

shift reaction (WGSR).

can be split into two reactions, one involving the

solids, and the other involving the gases:

The classical SCM needs to be modified both with

respect to the solids, but also with respect to the
gas. Earlier work at NTNU [1] - [6] has shown that
the oxygen potentials in the gas and solids can replace gas compositions in this type of modelling.

FeOx (s) = FeOx-1 (s) + O2 (g)

(eq.2)

R (g) + O2 (g) = RO (g)

(eq.3)

The Gibbs free energy of equation 1 is given by

It should also be noted that the mathematical model

of course should be formulated as a set of time dependent differential equations in some way and it
should have a clearly defined interface with the surroundings; mainly input and output gas compositions
and amounts per unit volume. This is shown in Figure 3.

G1 = G D1 RT ln

pR
pRO

(eq.4)

while for equations 2 and 3

G D 2 = RT ln
D

(eq.5)

pO2( s ) 1/ 2

G 3 = RT ln

pR pO2 ( g ) 1/ 2
pRO

(eq.6)
Since equations 2 and 3 combine to give the whole
reaction, the standard Gibbs free energy can be
given by:

D
1

G = RT ln

Figure 3: Flow sheet for the mathematical model of pellet

reduction, showing inputs and outputs.

1
pO2( s ) 1/ 2

+ RT ln

pR pO2 ( g ) 1/ 2
pRO

p
RT
RT
ln pO2 ( s ) +
ln pO2 ( g ) + RT ln R (eq.7)
pRO
2
2

and substituting into equation 22, we get:

RT
p
RT
p
G1 =
ln pO2 ( s ) +
ln pO2 ( g ) + RT ln R RT ln R
2
pRO
pRO
2
=

in the

Based on the above the reduction and oxidation of

iron-oxygen compounds may be visualized as driven
by differences in oxygen potentials set up by gas
mixtures involved and the solids. This is shown in
Figure 4.

(eq.8)

ln pO2

of the CO/CO2 and H2/H2O fractions of

the gas. This allows us to calculate reaction rates for

both gases simultaneously, with the gas compositions linked through the water gas shift reaction.
In the shrinking core model, there are three processes that can govern the reaction rate: the diffusion
of the reducing and product gas through the gas
boundary layer surrounding the pellet; the diffusion
of the gas through the product layers formed during
reduction, within the pellet to the reaction interface;
and the rate of the reaction itself at the interface

The reaction:

R. Longbottom and L. Kolbeinsen

This difference gives the driving force for the reduction of the iron oxides by the reducing gas. It allows
the computation of the driving force of reduction by
both hydrogen-water vapour mixtures and carbon
monoxide-carbon dioxide mixtures concurrently. A
similar consideration can be used for the water gas
shift reaction, giving a driving force of the difference

Figure 4: Clouds indicate gas mixtures and reactions internally (homogeneous or catalytic heterogeneous) transferring oxygen according to the water gas shift equation
(WGSR). Oxygen transfer between gas and solid is also
indicated. Numerical values for the oxygen potentials depend on compositions as well as temperatures.

FeOx (s) + R (g) = FeOx-1 (s) + RO (g)

RT
ln pO2 ( g ) ln pO2 ( s )
2

(eq.1)

New Direct Reduction (DR) - October 2008

Finally, branches 8 and 9 are solid state diffusion through

the magnetite and wstite layers.

between the reduced and raw material. Which of

these is the most important during the reduction of
the pellet depend upon the temperature, the composition of the reducing gas and the physical properties
of the pellet.

In the above consideration, hydrogen refers to

the hydrogen-water vapour mixture, with the
same for carbon monoxide. The geometric and
resistance terms of the reaction rate equations
are combined in the G-matrix. The driving force
for the reaction is placed into the e-vector,
while the reactions rates are given in the calculated i-vector. The reaction rates are given by:

Since the gas atmospheres that are utilised in the

model and experiments has two pairs of reducing
gases, as well as inert gas, there is an adjustment in
the model to reflect this. This is done by calculating
the fractions of the total gas that contain CO+CO2
and H2+H2O, and using these as multipliers for the
rates of the appropriate equations.

i = G(u + e)

The derived kinetic rate equations for the different

processes each contain three main sections. There
are:

(eq.9)

where u describes the drop in the potential across

each branch in the circuit, which is also an unknown.
To eliminate u as an unknown in equation 9, another
matrix Q is defined. The Q-matrix is defined using
electrical circuit theory. The vector u can be given
by:

a geometric term, which is usually the surface area

of the interface at which the process is occurring;
a resistance term, which is the appropriate kinetic
rate- or diffusion-coefficient, which we have incorporated as an Arrhenius expression, which allows the
effect of temperature to be considered; and

u = QTv

(eq.10)

where v is voltage drop across the columns of the Qmatrix, which can be expressed in terms of the
known e, G and Q. Using Kirchhoffs Law,

a driving force, which in this case is the thermodynamic driving force, given by the difference in the
logarithm of the oxygen potentials for the process.

Qi = 0

(eq.11)

and multiplying the rate equation 9 by Q and substituting in equation 10, we get:

To allow for simpler calculation of the rates of reaction, an electrical equivalent was formed of the reduction. The circuit diagram for the reduction of a
pellet is show in Figure 5. The individual branches of
the circuit are named in the figure caption.

Qi = QG(u + e)
0 = QG(QTv + e)
v = (QGQT)-1QGe (eq.12)
Which, substituting back into equation 9, gives as
the reaction rates based on the known quantities e,
G and Q.
i = G(1 QT(QGQT)-1QG)e

(eq.13)

The e-vector and the G- and Q-matrices are defined

in the appendix.
The general mass balance equation is

C j
t

+ N j = R*j

(eq.14)

where the first term is the accumulation or depletion

with time of species j in the pellet; the second term
*
is the transport of j in and out of the pellet; and R j
is the total rate of formation or consumption of the
species j in the appropriate reactions.
In the simulation of the reduction of a single pellet,
there is no mass transport of solid phases in or out
of the unit volume. The mass transport of the gaseous phases is already included within the micro
model. In this case, we can neglect the transport
term in equation 14. R*j is calculated by the summa-

Figure 5: Electric analogue for reduction of hematite pellets. Batteries are representing the equilibrium oxygen
potentials for gas mixtures (7, 19) and solid solutions (10,
11, and 12). Reduction with hydrogen of wstite (1), magnetite (2), and hematite (3) will all produce water, and
oxygen (as water) is transported out of the interior parts of
the pellet through resistances in the product layers (4, 5,
and 6). A parallel set of branches illustrates the reduction
(13, 14, and 15) by carbon monoxide and oxygen transport
(16, 17, and 18). Branches 21, 22, and 23 are representing
WGSR on the three solid reactant/product interfaces, while
branch 20 is for WGSR in the bulk gas. Four possibilities for
this reaction are necessary due to varying catalytic properties of the solids compared to the bulk gas conditions.

R. Longbottom and L. Kolbeinsen

tion of the appropriate terms in the vector i. Because

i is in terms of moles, the sum of the terms of i is
divided by the unit volume to give a concentration. It
is also necessary to take into account the stoichiometry of the reactions. For the numerical simulation, we

New Direct Reduction (DR) - October 2008

Results of Reduction Experiments

discretise the differential. This gives a mass balance

equation of:

Several experiments have been performed to date on

the KPRS pellets supplied by LKAB and CVRD pellets
supplied by Arcelor Mittal. The sample size of pellets
in all experiments was a nominal 250g (~50 KPRS
pellets or ~30 CVRD pellets).

t
j 1
j
C j =
x j 1 (iCO
+ i Hj 21 ) + x j (iCO
+ i Hj 2 )
V

where xj-1 is the stoichiometry of the reaction of

formation of the species j; xj-1 is the stoichiometry of
the reaction of consumption of the species j; i kj 1 is

Main gas mixtures used in experiments are indicated

in Figure 7 and Table I below

the element of the i-vector for the formation of j with

the gas k; and similarly i kj is the element of the ivector for the consumption of j with the gas k.

Figure 7: Gas mixtures used in experimental investigations

Table I Gas Mixtures

Figure 6: Interaction between two (three) counter current
processes with different time constants

This single pellet (or particle) model (the micro

model) is integrated into a model for the whole process as well as for the pellets in the retort of a laboratory scale Thermo Gravimetric Analyser (TGA), and
this facilitates the following main applications:
1. Combining mathematical model of retort samples
of ~100 single pellet with TGA measurements allow for determination of parameters of the equations describing the resistances in the model

c/h

Temperature
range (C)

0.33

0.41

0.40

0.14

435 - 500

0.50

0.30

0.33

0.14

700 - 760

0.00

0.07

0.06

0.14

880 - 910

Gas A is only able to reduce hematite to magnetite

and will generally only apply to the upper parts of a
reduction shaft. Gas B can also reduce the magnetite to wstite, but will not be able to produce iron.
The last gas composition, C, is able to produce
iron, but is somewhat oxidized relative to the input
reduction gas as seen in Figure 7.

2. Combining mathematical model retort samples

with mathematical model of process makes it possible to calculate profiles of gas concentration and
temperature distributions in the process (e. g.
shaft).

Generally speaking laboratory experiments were run

in three different modes:

3. Computed gas and temperature profiles can be

used to control input values to the TGA, and the
resulting recorded change of sample weight as
function of time and result of investigations of the
pellet sample after the experiment will give extra
verifications of the model and parameters

1. Steady state conditions: Keeping gas and

temperature constant throughout the duration of
the experiment. This is of course a situation a pellet never will experience on the decent through a
DRI shaft, but such experiments are useful for
setting initial values for adjustable model parameters.

4. Effects of changing the gas & temperature profiles

by alternative methods for Syn Gas (SG) and Carbon Capture (CC) as well as alternative process
schemes (e. g. Wiberg) can be investigated
through computer model simulations and verified
experimentally using the TGA.

2. Step change conditions: Both temperature and

gas compositions are changed stepwise during the
experiment but kept constant between steps.
Again a situation not found in a real process, but
useful for fine tuning the model parameters and
investigate the robustness of the mathematical
model calculations.

A robust and flexible model for reduction/oxidation

is, of course together with the TGA, the core of a
laboratory investigation of possible ways of integrating and verifying knowledge and competence on
both the micro and the macro scales of this type of
processes.

R. Longbottom and L. Kolbeinsen

Gas

3. Continuously changing conditions: Both temperature and gas compositions are changed by
ramping between the step values used under 2.

New Direct Reduction (DR) - October 2008

above during the experiment. This is a situation

which closely resembles the change in temperature and gas composition a pellet will experience
as it travels down a real shaft.

effected much more by the change in the gas composition than the changes in temperature, which
were put out of sync with each other to see the
effect of each.

Below some examples of these three different experimental modes are presented.

a)

a)

b)
b)

c)
Figure 8: Results of reduction of pellets using the constant conditions for KPRS pellets. Experiment mode 1: a)
gas composition, b) crucible temperature, c) mass change.

Figure 9: Results of reduction of pellets using the step

change conditions for KPRS pellets. Experiment mode 2: a)
gas composition, b) crucible temperature, c) mass change.

The temperature shown is the measured temperature, and it is noted that the crucible temperature
drops significantly at the start of the experiment.
This is due to the endothermic reactions when hydrogen is the dominating reducing agent. Gas composition, though, is shown to be constant. This is
partly due to the fact that the gas composition
shown is that of the gas mixing, but the amounts of
gas supplied are at least 5 times higher than the
amounts consumed by the reaction at any time.

In this case the drop in temperature is not noticeable, partly because the rate is now determined by
the change in the ability of the gas composition to
reduce the various oxides, but also to the lowered
rate in the experiment due to the lower temperatures
used for most of time.

Figure 9 shows the reduction of the pellets under the

stepped change conditions, where the temperature
and gas composition are stepped between the three
gas compositions. The reduction rates tend to be

R. Longbottom and L. Kolbeinsen

New Direct Reduction (DR) - October 2008

The parameters that can be adjusted within the

model to give a better fit to the experimental results
include kinetic rate constants for all of the reduction
reactions as well as for the water gas shift reaction,
diffusion coefficients for both gases through the
product layers, and the activation energies for the
reactions, although the values of these have been
taken from literature and will hopefully not need to
be adjusted. As such there are many degrees of
freedom to adjust the model, so that fitting of the
model to the experimental results should be good,
but the fundamental nature of these parameters
may give some understanding in the reduction the
pellets.

a)

b)
a)

c)
Figure 10: Results of reduction of pellets using the ramp
change conditions for KPRS pellets. Experiment mode 3: a)
gas composition, b) crucible temperature, c) mass change.

b)

Figure 10 shows the reduction of the pellets under

the ramped change conditions, where the temperature and gas composition are ramped between the
three conditions A, B, and C. Again the changes
in temperature are put out of sync with the changes
in gas composition to see the effect of each.

Figure 11: Comparison between the experiment and the

results of the micro model for the KPRS pellets reduced
using the Gas C conditions (Mode 1). a) the experimental
fractional mass loss is compared to the model; b) the calculated concentrations of the different iron species is shown.
Inputs as in Figure 8

Figure 11 shows the results of the modelling of the

reduction of KPRS pellets using Gas C (as shown in
Figure 7) with the inputs to the model, taken from
the experimental log. Under these conditions, a close
fit between the model and the experimental results
can be achieved fairly easily.

Results of Mathematical Modelling

The modelling completed so far has been comparisons between the single pellet micro model and the
experimental results indicated in the previous section. For these tests, the inputs into the model have
included the gas compositions and temperature from
the experimental tests at each time step. The purpose of this is to calibrate the parameters within the
micro model to the experimental results, so that the
micro model can be used as a part of the greater
shaft model.

R. Longbottom and L. Kolbeinsen

From the model, it can be seen that the hematite is

reduced quickly, and disappears from the system
early in the reduction. Wstite and metallic iron are
produced from early on during the reduction, meaning that the magnetite concentration is limited in

New Direct Reduction (DR) - October 2008

extent, and disappears after around 20 minutes or

so. Wstite to iron seems to be the main limiting
reaction, but still occurs quickly. As the concentration
of wstite within the pellet decreases, the rate of
reaction also decreases, until the reaction virtually
stops after about an hour when the reaction is
largely complete.

rent fit, the kinetic parameters were significantly

increased from those in the previous simulations.

The results from modelling the reduction of KPRS

pellets under the step change conditions are shown
in Figure 12. Under these conditions, a reasonable
match between the experimental and the calculated
results can be achieved, with only a little tuning from
the kinetic parameters found from the previous three
experiments. The biggest deviations are at the beginning of the reduction, where the induction period is again noticed at low temperatures with Gas A
present. However, at higher temperatures, and with
increasingly reducing gases, the two curves match
much better.

a)

b)
a)

Figure 13: Comparison between the experiment and the

results of the micro model for the KPRS pellets reduced
using the ramp change conditions (Mode 3). a) the experimental fractional mass loss is compared to the model; b)
the calculated concentrations of the different iron species is
shown. Inputs as in Figure 10

Even now, it can be seen that the fit and shape of

the curves are somewhat different between the
experimental data and what is calculated. While the
three different curves that were mentioned in the
previous section are simulated, their positions are
different, with the flattening of the reduction curves
occurring at lower extents of reduction and lower
reaction times in the model than in the experiment.
The positions of these cannot be easily changed
without changing the thermodynamics of the model,
as they are caused by a delay in the onset of the
formation of wstite and metallic iron respectively.

b)
Figure 12: Comparison between the experiment and the
results of the micro model for the KPRS pellets reduced
using the step change conditions (Mode 2). a) the experimental fractional mass loss is compared to the model; b)
the calculated concentrations of the different iron species is
shown. Inputs as in Figure 9

It is thought that in this case, the differences in the

perceived thermodynamics of the model and the
experiment are caused by temperature gradients
within the sample, caused by the heating of the
sample during the experiment. Also, the effect of
slight differences between the actual and logged gas
atmosphere would also be exaggerated in this case
as opposed to having a constant gas atmosphere.

Figure 13 shows the results from the modelling of

the reduction of KPRS pellets under the continuously
changed (ramped) gas atmospheres and temperatures. In this case, it was much harder to find a
reasonable match between the experimental and
calculated fractional mass loss curves. For the cur-

R. Longbottom and L. Kolbeinsen

The effect of the temperature gradients can be reduced by repetition of the experiments, since the

New Direct Reduction (DR) - October 2008

vertical tube furnace used for these experiments has

been replaced with an exposed element muffle furnace, which has a larger isothermal zone. The new
furnace is also capable of faster heating and cooling
rates, which should allow better control of the crucible temperature in the future.

set up rather than fundamental issues in the model.

The micro model indicates that the hematite within
the pellets is reduced to magnetite quickly, which in
turn is reduced fairly quickly to wstite. The reduction of wstite to metallic iron seems to be the limiting stage in the reduction of the pellets, which is in
line with what would be expected.

Figures showing results for similar experiments and

simulations for the CVRD pellets are omitted here.
There is little difference between the two cases, with
similar kinetic parameters used for these pellets.
Since the model takes into account the different
pellet diameters, the main difference noticed in the
experimental results between the two pellet types
does not have such a large effect on the model.

The shaft model is working, but with known issues.

Some further work is required for this model to work
satisfactorily, but this is not envisioned to be an
extensive task for a simple shaft model. It should be
noted, however, that the formulation chosen for the
micro model described, as well as its interaction with
the surroundings is done in a manner intended to
facilitate any chosen complexity in the shaft model.

Also omitted are the results from simulations of a full

shaft model. In this model the micro model plays a
significant role in the mass balances for both gas and
solids, as well as for the energy balances for the
same [2], [7], [8] an important feature of the shaft
model is the fact that it is a dynamic description
where said balances are on partial differential form.
In fact we may say that the micro model presented
here is contained in a rudimentary macro model
(shaft model) containing only one cell and described
by equation 14 given earlier.

Acknowledgements
Grant number 158907 from the Norwegian Research
Council for the project DISvaDRI - Dispersoids In
Steel and value added Direct Reduced Iron in the
SIPNMATR program for allowing use of TGA equipment. Research scientist, Dr. Sean Gaal is thanked
for his efforts in continuously improving the performance of said TGA equipment.

References

A dynamic shaft model allows starting the simulation

from a known situation, e. g. a cold shaft filled with
hematite pellets, and run it until a steady state is
reached.

[1] Kolbeinsen, L.: Reduction of Iron Ore Mathematical

model for process control. PhD - thesis, Metallurgisk Institutt NTH May 1982 (In Norwegian)
[2] Kolbeinsen, L. & Onshus, T.: "Reduction of Iron Pellets
in a Shaft furnace - A Dynamic Approach". Proc. of the
6th Process Technology Conf. - Fifth International Iron
and Steel Congress, pp. 37-43, Washington April 1986.

Conclusion
The reduction of iron ore pellets has been studied
using different techniques. Thermodynamic studies,
experimental investigations and mathematical modelling have all been undertaken to better understand
the behaviour of different pellet types in the new
direct reduction process developed as part of the
ULCOS project [7], [8].

[3] Kolbeinsen, L. & Onshus, T.: "A generalized Shrinking

Core Model including the Water Gas Shift Reaction applied to Reduction of Hematite Pellets". Proc. of The
fourth Japan-Nordic Countries joint Symposium on Science and Technology of Process Metallurgy, The Iron
and Steel Institute of Japan, pp. 66-80, Tokyo, November 1986.

Experimentally, it can be seen that the pellets reduce

quickly under the proposed conditions. However, the
reaction is very endothermic and causes a large
temperature drop in the sample under the small
scale (250 g) conditions that have been studied. The
experimental equipment gives consistent results with
good repeatability. However, may be some problems
during non-isothermal experiments, which have
hopefully been rectified by the replacement of the
furnace. Some further experiments will be carried
out, mostly of partial reduction of the pellets, so that
these pellets can be characterised and used as a
further comparison and check for the mathematical
model.

[4] Leiv Kolbeinsen: "Metallurgical Conversion of Natural

Gas" Proc. of The Fifth Japan-Nordic Countries joint
Symposium on Science and Technology of Process Metallurgy, Sept. 1992, Helsinki, Finland, pp 191-207
[5] Kolbeinsen, L. & Bsen, T.: "FERROCARBON - Production and use of Iron Carbide" Proc. of the 2. Int. Symp.
on Metallurgical Processes for the Year 2000 and Beyond
and the 1994 TMS Extraction and Process Metallurgy
Meeting . San Diego, September 1994
[6] Tveit, H. & Kolbeinsen, L.: "Export Processed Gas!" (In
Norwegian) Teknisk Ukeblad Teknikk 142 (11) March
1995 pp 28-29
[7] Leiv Kolbeinsen: Summary of SP3 activities at Department of Materials Science and Engineering, NTNU ULCOS//SP3 RTD/WP3.2&3.3/ntnu 2006c

The mathematical pellet model is working, and can

give a good fit to most of the experimental conditions used in this work. There are however, some
discrepancies between the experimental and calculated results under certain conditions, which are
thought to be due to limitations in the experimental

R. Longbottom and L. Kolbeinsen

[8] Leiv Kolbeinsen & Ray Longbottom: Summary of SP12

activities at Department of Materials Science and Engineering, NTNU ULCOS//SP12 RTD/ ntnu 2008

New Direct Reduction (DR) - October 2008

Appendix Definition of Vectors and Matrices used in the Micromodel

Nomenclature
Ci

molar concentration of compound i

Cc

molar concentration of the carbon containing fraction of the gas (CO + CO2)

Ch

molar concentration of the hydrogen containing fraction of the gas (H2 + H2O)

DeffJ ,i

effective binary diffusion of gas fraction i through solid J

EaJ,i

activation energy for reduction to J by gas i

EaWGS
kf

activation energy for the water gas shift reaction

mass transfer coefficient through the gas boundary layer

koJ,i

frequency factor in Arrhenius equation for reduction to J by gas i

koJ

frequency factor in Arrhenius equation for the water gas shift reaction at the inner interface of solid J or bulk gas

Nj

moles of species j

gas constant

R*j

total rate of formation/consumption of species j in the relevant reactions

rJ

radius of reaction interface at the inner surface of solid J

ro

radius of pellet

Temperature

Time

GJD

Gibbs free energy of reduction to J in the solid phase

GiD

Gibbs free energy of the fraction in the gas phase containing i

Superscripts
g

bulk gas phase

Fe

metallic iron

Magnetite

wstite

Subscripts
c

carbon containing fraction of gas

hydrogen containing fraction of gas

Fe

metallic iron

Magnetite

wstite

R. Longbottom and L. Kolbeinsen

10

New Direct Reduction - October 2008

Table A-1: Diagonal terms of the G-matrix (dimension 2323), all others are zero.

Term
G(1,1)

G(2,2)

G(3,3)

G(4,4)

G(5,5)

G(6,6)

Equation

Remark

E
2rW 2 RT
Ch koFe, h exp
RT
EaW, h
W
2rM 2 RT
Ch ko , h exp
RT
EaM, h
M
2rH 2 RT
Ch ko ,h exp
RT
1

ro rFe
1

D Fe 2r r RT k h 2r 2 RT
eff ,h
f

Fe o
o
1
W

Deff
, h 2rW rFe RT
rFe rW
1

DeffM ,h 2rM rW RT
r r
Fe
a ,h

G(13,13)

G(14,14)

G(15,15)

G(16,16)

G(17,17)

G(18,18)

EaFe,c

2r 2 RT
C k exp
RT W
E W
Cc koW,c exp a ,c 2rM 2 RT
RT
E M
Cc koM,c exp a ,c 2rH 2 RT
RT
1

ro rFe
1

+
D Fe 2r r RT k c 2r 2 RT
eff ,c
f

Fe o
o
1
W

2
Deff
r
r
RT
W Fe
,c
rFe rW
1

DeffM ,c 2rM rW RT
rW rM

R. Longbottom and L. Kolbeinsen

interfacial reduction of Fe2O3 with H2

diffusion of H2/H2O through boundary layer

& Fe-layer

diffusion of H2/H2O through FexO-layer

diffusion of H2/H2O through Fe3O4-layer

potential of H2/H2O in gas phase
solids diffusion in Fe3O4
solids diffusion in FexO
potential of solids at FexO/Fe interface
potential of solids at Fe3O4/FexO interface
potential of solids at Fe2O3/Fe3O4 interface

Fe
c o ,c

G(19,19)

interfacial reduction of Fe3O4 with H2

0
0

G(7,7)
G(8,8)
G(9,9)
G(10,10)
G(11,11)
G(12,12)

interfacial reduction of FexO with H2

interfacial reduction of FexO with CO

interfacial reduction of Fe3O4 with CO

interfacial reduction of Fe2O3 with CO

diffusion of CO/CO2 through boundary

layer & Fe-layer

diffusion of CO/CO2 through FexO-layer

diffusion of CO/CO2 through Fe3O4-layer

potential of CO/CO2 in gas phase

11

New Direct Reduction - October 2008

Table A-1 (cont.)

Term

Equation

G(20,20)

G(21,21)

G(22,22)

G(23,23)

Remark

E
2ro 2 RT
kog exp
RT
EWGS
koFe exp a
2rFe 2 RT
RT
WGS
a

EWGS
koW exp a
2rW 2 RT
RT
EWGS
koM exp a
2rM 2 RT
RT

WGS reaction in gas phase

WGS reaction on pellet surface &
FexO/Fe interface
WGS reaction on Fe3O4/FexO interface
WGS reaction on Fe2O3/Fe3O4 interface

Table A-2: Terms of the e-vector (dimension 23), all other terms are zero.

Term

Equation

Remark

e(10)

G D
2 Fe
RT

potential of solids at FexO/Fe

interface

e(11)

GWD

2
RT

potential
of
solids
Fe3O4/FexO interface

at

potential
of
solids
Fe2O3/Fe3O4 interface

at

e(12)

e(7)

e(19)

R. Longbottom and L. Kolbeinsen

G D
2 M
RT
GHD / H O
2
2

2
RT
D
GCO

/ CO2

2
RT

Potential of H2/H2O in gas

phase
Potential of CO/CO2 in gas
phase

12

New Direct Reduction - October 2008

The Q-matrix (dimension 1123) is:

Q=

[ 1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

R. Longbottom and L. Kolbeinsen

13

New Direct Reduction - October 2008