Microstructure based creep constitutive model

for precipitation strengthened alloys: theory

and application
B. F. Dyson*
A kinetic creep equation for precipitation strengthened alloys has been derived using the climb/
glide particle bypass micromechanism. It is applicable only when stresses lie within the range
delineated by the dislocation network strength and the athermal yield strength – the lower of
particle shear strength or Orowan strength. In this ‘dispersion controlled’ creep regime, strain rate
is a hyperbolic sinh function of stress and interparticle spacing, making the state of dispersion an
important microstructural feature. Existing equations have been used to quantify the evolution of
dislocation and particulate microstructures and also stress redistribution between the two phases.
The resulting constitutive equation set has been used to predict proof stresses, minimum creep
rates and lifetimes for Nimonic 90. Comparison to industry data over a 600–900uC temperature
range concludes that proof stresses are creep controlled once temperatures are above 700–
725uC. Minimum creep rates and lifetimes over a similar temperature range are well predicted
between 600 and 700uC but become progressively less accurate at higher temperatures/lower
stresses, particularly lifetimes.
Keywords: Superalloys, Microstructure based kinetic equation, Creep, Proof stress

Dedicated to the memory of Professor Malcolm McLean

Introduction Similarly, tertiary creep is predominantly modelled using

a single empirical (damage) variable in spite of a large
In the power generation and propulsion industries, a metallurgical literature demonstrating that there are
small fraction of the many components is safety critical. many alloy specific microstructural features that evolve
Their mode of operation and/or geometry often generate during service and which could be accelerating creep rate
complex thermal/mechanical loadings either globally or (and therefore reducing load bearing capacity) either
locally within the component. Whether global or local, singly within different stress/temperature regimes or
excessive inelastic strain and/or fracture can cause simultaneously. The latter is more likely in complex
operational failure and its prediction requires a creep industrial alloys. This does not mean that empirical
constitutive equation set for the relevant alloy. approaches cannot model the shapes of creep curves (i.e.
Applying computer based modelling to the histori- plot the correct strain trajectory) under a variety of
cally protracted and expensive experimentally based steady conditions of stress (load) and temperature. They
development of high temperature alloys, such as super- clearly can, otherwise many papers would never have
alloys and 9–12%Cr steels, is a relatively new activity. It reached publication. The important point is that the
has a common problem with service failure modelling: number of arbitrary variables required increases drama-
few constitutive equation sets have foundations in tically as an increasing number of different stress,
physics, never mind microstructure and this deficiency temperature and stress states is imposed. Their limita-
has to be addressed if both applications are to progress. tions can easily be revealed by conducting a simple
The majority of equation sets have their origin in the diagnostic test: uniaxial testing under non-steady
applied mechanics field. The kinetic (i.e. creep rate) thermal/mechanical loadings. This type of test is
equations have usually been empirical and assumed to particularly fruitful for thermally unstable alloys, which
obey an arbitrary power law dependency on stress, even are manifest during service in both the power generation
when the power law exponent is subsequently found to and propulsion industries.
be a function of stress, temperature and inelastic strain – An early paper1 rationalising the microstructural
a sure sign that it is not the correct physical law. origins of tertiary creep with the empirical approach
(now called continuum damage mechanics)2–4 was
limited to those cases where each micromechanism acted
Department of Materials, Imperial College London, Exhibition Road, alone. The key point in the paper was identifying each
London SW7 2AZ, UK microstructural state variable whose evolution could
*Corresponding author, email drb.dyson@btinternet.com at least be approximately quantified – for example,

ß 2009 Institute of Materials, Minerals and Mining

Published by Maney on behalf of the Institute
Received 18 August 2008; accepted 11 September 2008
DOI 10.1179/174328408X369348 Materials Science and Technology 2009 VOL 25 NO 2 213
Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

dislocation density, particle volume fraction and size,

etc. and to use current knowledge to predict its evolution
and influence on creep rate. However, to progress
further and to quantify the effects on creep rate of what
could be two or three different microstructural variables
requires a kinetic (creep rate) equation that is an explicit
function of these variables. To derive this with any
degree of certainty requires a theoretical knowledge of
the creep mechanism for each specific class of alloy.
A generic kinetic creep equation for precipitation
strengthened alloys has been derived in the present
paper, using the climb/glide particle bypass microme-
chanism. All relevant microstructural parameters are
embedded within it and the consequential constitutive
model used to predict proof stresses, minimum creep
rates and lifetimes in the nickel base polycrystalline
alloy, Nimonic 90. The predictions are compared with a
published industrial database covering the temperature
range 600–900uC.

Derivation of kinetic creep equation

Microstructure and mechanism
Figure 1 is a schematic showing the essential cross-
sectional features of the dislocation and particulate
microstructure being modelled: a unimodal set of
randomly distributed spherical particles of radius rp
fully coherent and having a volume fraction wp with 1 Schematic illustration of particle strengthened alloy,
interparticle spacing lp loaded under uniaxial tension s. loaded under uniaxial tension, with interparticle spa-
The dislocation network has density rT with corre- cing, smaller than dislocation spacing (dislocations
{1=2
sponding spacing rT greater than the interparticle from two sources, S1 and S2, by pass spherical parti-
spacing lp. A fraction of dislocation links will be able to cles at particles ‘A’ and ‘C’ by mixed glide and climb
act as Frank Read sources (S1 and S2 in Fig. 1) once involving vacancy emission ‘e’ and at particle ‘B’, by
the local shear stress in the matrix tm is greater than the both vacancy absorption and vacancy emission)
1=2
shear ‘network’ strength ^tnet (~aGbrT ), where G is the
shear modulus and a a constant. S1 and S2 are both average, there will be an equal number of ‘e’ and ‘a’ type
under a shear stress tD (~tm {^tnet ) (Fig. 1). With climbing dislocations provided that the particulate
coherent particles, there are two possible micromechan- microstructure remains sensibly spherical or cubic.
isms that can give rise to a strain rate insensitive This simplification will be utilised below when formulat-
(athermal) shear yield strength ^tath : particle shearing and ing the kinetic creep law.
Orowan looping. The creep model will apply only when Ansell and Weertman6 wrote the seminal paper
the matrix shear stress is within the range delineated by quantifying the mechanism illustrated in Fig. 1. They
^tnet and ^tath , i.e. when ^tnet vtm v^tath . assumed climb was the rate controlling process and their
Dislocation sources become active once the shear resulting equation predicted that the steady state creep
stress exceeds ^tnet , assuming a negligible Peierls Nabarro rate was linear in stress with a magnitude that decreased
resistance. In the absence of diffusion, dislocations will with increasing particle size. This is in contrast to
glide until they reach the nearest particle producing a laboratory experience where creep resistance is invari-
negligibly small shear strain that will be ignored here. At ably highly non-linear in stress and where coarsening the
test temperatures greater than approximately half the particulate microstructure – thereby increasing particle
alloy’s homologous temperature, usual laboratory test- size – usually increases creep rate.
ing times enable dislocations to climb along the particle/ A different approach is now presented.
matrix interface. The slip plane/particle geometry
dictates the direction of climb5 and the need for an Kinetic creep equation
associated coupled glide. In Fig. 1, the local climb force The interaction of dislocations with particles (Fig. 1) can
is such that the dislocation from source S1 climbs by also be represented by a square lattice array of point
vacancy emission ‘e’ along the particle ‘A’/matrix forces (Fig. 2). Brown and Ham7 discussed their use in
interface (along with coupled glide) until it is in a assessing mechanical threshold or ‘strength’ of finite
position to glide within the matrix. After being stopped sized particle arrays – but in the absence of diffusion. At
again at particle ‘B’, it climbs by vacancy absorption ‘a’ stresses below this mechanical threshold and in the
and the above process is repeated. The dislocation from presence of diffusion, dislocations can continuously
source S2 , moves by vacancy emission along the rearrange themselves through climb and it is postulated
particle, ‘C’/matrix interface and then glides within the that there will always be a dynamic fraction of points
matrix to particle ‘B’, where climb/glide is again by where a dislocation will be able to overcome the local
vacancy emission. Under uniaxial compressive stressing, pinning force by a thermally activated process involving
these emission/absorption situations are reversed. On diffusion, thereby allowing localised slip to occur. This is

214 Materials Science and Technology 2009 VOL 25 NO 2

 Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

:
cm ~rbug (5)
(ii) ug&Ge(b/rp)wplp; i.e. the glide rate within the
matrix is fast and the shear strain rate is
controlled by the rate at which climbing disloca-
tions are ‘released’ from the particle dispersion
and will be termed the dispersion controlled shear
:
creep rate cd=c
:
cd=c ~rb(b=rp )wp lp Ce (6)

h
Taking the interparticle spacing i square lattice
lp as the
spacing (Fig. 2) lp ~1:6rp (p=4w)1=2 {1 . Inserting lp
into equation (6), gives
2 Point obstacle representation of Fig. 1 illustrating the
: h i
stochastic ‘escape’ of climbing dislocations to enable cd=c ~1:6rb2 wp (p=4w)1=2 {1 Ce (7)
localised glide within matrix
Ge is given by Friedel9 for climb of a vacancy emitting
represented in Fig. 2 by the dotted ‘point’. It is a similar jog ‘e’ (see Fig. 1)
process to thermally activated glide below the athermal
 
threshold in the absence of diffusion and which has been cj Ds blp
termed ‘jerky glide’.8 There, a dislocation spends most of Ce,e ~ exp bt D {1 (8)
b2 kT
its time stationary but in this case, the ‘stationary’
fraction is in fact climbing and thereby continually where Ds is the volume diffusion coefficient cj the
providing a fresh concentration of potentially ‘glide dislocation line jog density blp is the slip activation area
active’ sites of the type represented by the dotted ‘point’. shown in Fig. 2 and btD, the force on the dislocation,
Climbing and gliding dislocation densities rc and rg where tD is given by tD ~tm {^tnet .
respectively are therefore not independent in this model For a vacancy absorbing jog, the jump frequency Ge,a
and the mobile dislocation density r is given by the is given by
following  
cj D s blp
r~rc zrg (1) Ce,a ~ 2 1{exp{ btD (9)
b kT
The climbing dislocation density in contact with The number of vacancy emitting and absorbing jogs will
particles is rcwp and it is assumed that a fraction b/rp be approximately equal when the particulate micro-
will always be in a position to ‘escape’ from the set of structure is spherical or cubic, so the nett frequency Ge
particles after undergoing a single activated jump with will be given by the mean of equations (8) and (9), thus
frequency Ge, followed by localised glide within the
matrix. The dislocation glide density rg will therefore
cj D s tD b2 lp
increase at a rate rcwp(b/rp)Ge but will be immobilised Ce ~ 2
sinh (10)
temporarily at the next particle at a rate rgug/lp, where b kT
rgug is the gliding flux and lp the interparticle spacing. The dispersion controlled shear creep rate within the
: :
Their net rate of increase rg is therefore given by matrix cd=c is given by substituting equation (10) into
: ug equation (7)
rg ~rc wp (b=rp )Ce {rg (2)
lp h i  
: tD b2 lp
Equation (2) can be integrated after substituting for rc cd=c ~1:6rwp (p=4w)1=2 {1 cj Ds sinh (11)
kT
using equation (1), with rp and wp constant and setting
rg50 at t50 to give An important point to note that the dispersion term lp is
in the sinh function of equation (11) and therefore very
Ce (b=rp )wp lp sensitive to stress. Particles coarsen with time and so the
rg ~r
ug zCe (b=rp )wp lp rate of change of lp becomes an important factor in

 predicting long term performance. The pre sinh term
ug zCe (b=rp )wp lp depends only on volume fraction wp which remains
1{exp { t (3)
lp constant during standard isothermal creep tests, but
decreases with temperature. The change of wp with
The term in square brackets is a transient and will be
temperature requires quantification in order to predict
neglected, giving the following matrix dislocation glide
behaviour at different temperature levels and/or under
flux
variable temperature conditions.
Ce (b=rp )wp lp
rg ug ~ rug (4) Microstructure based equation set
ug zCe (b=rp )wp lp
Equation (11) needs to be expressed in terms of tensile
There are two limits to equation (4) stress and strain in order to analyse the uniaxial datasets
(i) Ce (b=rp )wp lp &ug , i.e. the glide rate within the given in the following section. Standard relationships
matrix is slow and rg increases until can be found in the literature for a polycrystal loaded
rg %r (rc %0) and the matrix shear creep rate under uniaxial tension, giving the dispersion controlled
: :
cm (~brg ug ) is given from equation (4) by tensile strain rate ed=c

Materials Science and Technology 2009 VOL 25 NO 2 215

Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

h i  
: 1:6 1=2 sD b2 lp
ed=c ~ rwp (p=4w) {1 cj Ds sinh (12)
M- MkT
-
where M - is the Taylor factor (%3) and sD ~sm {^snet .
To be of practical use, equation (12) has to be an
explicit function of applied stress s. The method used to
achieve this (i.e. going from the microstructural scale to
the macro or global scale) is developed in the section on
‘Appendix’, based on the work of Ion et al.10 There are
two issues to consider:
(i) the relationship between the local creep rate in
one of the phases and the global creep rate that is
measured in the laboratory
(ii) the necessary transfer of load from weaker to
stronger parts of the microstructure – in this case 3 Comparison of measured volume fraction of gamma
from creeping matrix to elastic particles – and its prime in laboratory prepared In 597 as function of tem-
associated kinetics, along with the mechanism perature with model fit using equations (18) and (19)
and mechanics dictating the magnitude of load
partition. 20 000 h over a range of temperatures.12 These data can
The global creep rate for the dispersion controlled be represented by the following empirical relationship
mechanism is given in equation (30) in the section on  
3 3 3|105
‘Appendix’ r ~ri zexp { (17)
RT
: 1:6 h i
eG ~ rwp (1{wp ) (p=4w)1=2 {1 As particle sizes increase, there will be a proportionate
M- increase in |lp| and thus an increasing creep rate,
" #
(s{si {^snet )b2 lp equation (13).
cj Ds sinh (13)
MkT
- Volume fraction of c9
The standard commercial heat treatment of Nimonic 90
where sD has now been replaced by the applied stress s
consists of solution treatment at 1080uC with an ageing
and an internal stress si which has the following
treatment at 700uC for 16 h. The c9 solvus temperature is
evolution equation, derived in equation (33) in the
approximately 1000uC and the mechanical data dis-
section on ‘Appendix’
! cussed below ranges from 600 to 900uC. Since the c9
: wp E si : volume fraction occurs in the pre sinh of equation (13)
si ~ 1{ 1 eG (14) and also indirectly through the |lp| in the sinh function,
1{wp si an analytical expression for the temperature dependence
1
The upper bound value for internal stress si is given by of wp becomes an important requirement. An indirect
method will be presented since no published data were
equation (32) in the section on ‘Appendix’
found in the literature.
1 2wp Figure 3 is from Gibbons and Hopkins13 and shows
si ~ s (15) c9 volume fraction as a function of temperature in a
1z2wp
laboratory alloy based upon commercial IN597. The
Two further evolution equations are required: one for following lever rule equation used here to fit the data
the dislocation density r and another for the inter-
particle spacing lp. The dislocation density appears C0 {Ce (T)
wp (T)~ : (18)
directly in the pre sinh term of equation (13) and 0 23{Ce (T)
indirectly through the network strength term in the sinh where C0 is the concentration of the main c9 forming
function. They act in opposition. Since the initial elements in the alloy and Ce(T), their summed equili-
magnitude of ^snet is of order unity in superalloys – brium matrix solute concentration, given by
because of low initial dislocation densities – and the  
7250
applied stresses are usually several hundred MPa, there Ce (T)~17exp { (19)
is invariably a net increase in creep rate due to increasing T
dislocation density, at least in the early stages of creep. At the solvus, wp(T)50 and so C05Ce(Ts) in equa-
There is still no universally accepted equation for the tion (18), where Ts is the solvus temperature.
evolution rate of dislocations in superalloys, so the Substituting C05Ce(Ts) into equation (19) gives the
empirical one suggested by Dyson and McLean11 will be following equation for Ts
used here
: : 7250
r~K e (16) Ts ~ (20)
ln(17=C0 )
where K5300ri The solid line in Fig. 4 is a plot of solvus temperature as
Equation (16) takes no account of the kinetics of a function of C0 given by equation (20). Experimental
trapping or annihilation (by climb processes). data from several superalloys are in good agreement
Coarsening occurs in c9 dispersions and average considering that only a three parameter model has been
particle size data for Nimonic 90 are available up to used. Nimonic 105 has been included to illustrate the

216 Materials Science and Technology 2009 VOL 25 NO 2

 Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

4 Comparison of predicted solvus temperature as function

of atomic percent of AlzTizNbzTa using equation (20) 5 Comparison of 0?1% proof stress data with model pre-
with solvus data from several nickel superalloys dictions: it is concluded that at temperatures lower
than ,700uC, yield is by particle cutting mechanism; at
well known consequence of an alloy having a liquidus higher temperatures, yield is controlled by dispersion
temperature that is below the c9 solvus. controlled creep mechanism
The predicted solvus for Nimonic 90 is sufficiently
close to the experimental value to justify using equa- performed under conditions of constant total strain rate
tions (18) and (19) to provide, in this instance at least, starting from zero strain and terminating when the
an analytical method of predicting c9 volume fraction as inelastic component had reached 0?1%, thus simulating
a function of temperature. the laboratory test data. The imposed total strain rate
was 1024 s21, which is the industry standard. The
Comparison of model predictions with predictions closely follow the data at temperatures
nimonic 90 experimental datasets greater than ,700uC. Below 700uC, experimental data
Proof stresses and predictions continue to diverge.
A comprehensive set of experimental 0?1% proof stress Minimum creep rates
data (INCO data sheets) for Nimonic 90 has been Figure 6 shows minimum creep rate as a function of
digitised, plotted in Fig. 5 and compared to predictions stress for Nimonic 90. The experimental data have been
using three mutually exclusive yield mechanisms. digitised from a paper by Harrison and Evans15 and are
The particle cutting model of Brown and Ham7 gives from a number of different sources and batches, so
the following equation for tensile yield strength accuracy cannot be guaranteed. However, it is extensive
(
 ) and unique in the open literature on nickel base
^ capb 4wp (T) 1=2 superalloys.
sc ~M - {wp (T) (21)
2b p The solid curves have been generated from the same
constitutive equation set used in the previous section but
where capb is the antiphase boundary energy. in addition, equations (16) and (17) were added. These
Equation (21) applies only when the condition rp.Gb2/ allowed dislocation generation and particle coarsening,
2capb is met. Putting capb50?35 J m22 gives rp.5?3 nm. which were not required for the short time computations
The initial c9 radius in Nimonic 90 is 12?5 nm,12 so for proof stress. In this case, numerical integration was
equation (21) should apply. The dashed line in Fig. 5 :
performed under constant load (s~s:e) to produce
shows the results. At temperatures ,700uC, the cutting
strength and proof stress data are virtually coincident
and remain so as the temperature decreases. Above
700uC, the experimental data and the cutting predictions
progressively diverge. The steep decline in cutting
strength .950uC is due to the rapid decrease in wp as
the solvus temperature of 1000uC is approached.
The Orowan bypass model as modified by Ashby,14 was
used to assess the potential upper bound yield condition
sor ~0:13 M(Gb=l
- p )ln(rp =b) (22)
The results are shown as the dot dashed line in Fig. 5. It
illustrates that under the fully heat treated condition,
Nimonic 90 will never yield by Orowan looping.
The solid line in Fig. 5 is the 0?1% proof stress
predicted using equations (13)–(15) with initial disloca-
tion density, ri51010 m22 and combined activation
energy for diffusion and jogs, Qj,s5320 kJ mol21 and 6 Comparison of industry generated minimum creep rate
Ds,051024 m2 s21. The variation of c9 volume fraction data in temperature range 700–900uC, with predictions
was given by equations (18) and (19). Integration was of dispersion controlled creep model

Materials Science and Technology 2009 VOL 25 NO 2 217

Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

inelastic strain as a function of time until the minimum

creep rate could be identified. Considering that the solid
curves are not ‘fits’, the agreement is good. However,
there is a systematic over prediction of creep rates at
lower stresses, which becomes greater the higher the
temperature.

Creep lifetimes
Figure 7 shows rupture lifetime as a function of stress
for Nimonic 90, again using the same digitised database
from the paper by Harrison and Evans.15 Predicted
creep curves are shown as solid lines and were generated
by numerically integrating the same constitutive equa-
tion set used in the previous section, but under constant
load. The lifetime database covers a larger temperature
range than was the case with minimum creep rates. The
two lowest temperatures – where no corresponding 7 Comparison of industry generated rupture lifetime data
minimum creep rate data were supplied – are remark- in temperature range 600–900uC, with predictions of
ably well predicted, particularly the database at 650uC, dispersion controlled creep model
which extends to 20 000 h duration.
Short term lifetimes are also well represented at all necessary to perform pretreatment and subsequent creep
temperatures by the model but not longer lifetimes at tests at the same temperature in order to avoid other
temperatures .700uC. The shortfalls in predicted life- potential changes in microstructure.
time clearly become systematically larger as tempera- The dispersion controlled creep equation (13) has
tures increase. successfully predicted the steep decline in 0?1% proof
stress data at temperatures .700uC. Below 700uC,
Discussion calculations using equation (20), demonstrate that
particle shear truncates the field of activity of creep
The microstructure based kinetic creep equation (13), controlled yielding. Nimonic 90 is therefore weaker at
has been termed ‘dispersion controlled’ because creep lower temperatures than it would have been had shear
rates are predicted to be highly sensitive to changes in not intervened. The value used for capb in computing ^sc
the dispersion parameter lp the interparticle spacing. is higher than the usual range quoted, 0?1–0?2 J m22,
Particle coarsening is an important feature of most, if but is similar to the maximum of 0?3 J m22 measured by
not all, alloys in service. To date, its influence on service Raynor and Silcock,16 who studied a number of low
lifetimes has never been quantitatively assessed, quite volume fraction nickel base alloys.
possibly because no creep law explicitly contains the If the rapid decline in proof stress of Nimonic 90
parameter lp; a necessary requirement. Testing the beyond 700uC is a consequence of dispersion controlled
predictions of equation (13) becomes, therefore, very creep, then proof stress data should be dependent upon
important for assessing its validity. To that extent, the the magnitude of the testing strain rate: the faster the
results presented in Fig. 6 are a disappointment. The imposed strain rate, the higher the Proof Stress.
progressive increase in the magnitude of shortfalls in
predicted lifetime as temperatures rise could be inter-
preted as being the consequence of gross overestimates Conclusions
of the effect of coarsening; thereby essentially condemn- 1. The derived ‘dispersion controlled’ creep rate
ing the dispersion controlled model. However, in (equation (13)) is characterised by its high sensitivity
mitigation, it is also observed that minimum creep rates to the magnitude of interparticle spacing. The creep data
in Fig. 7 progressively overestimate creep rates at higher available in this work was isothermal and has proved
temperatures, suggesting that the cause of progressive inadequate for testing this prediction unambiguously.
shortfalls in lifetime may be related to small errors in This is a consequence either of equation (13) being
initial parameter values within equation (13). Creep wrong or because of small uncertainties in the magni-
rates and lifetimes are inextricably linked. Additionally, tudes of initial parameters in equation (13) and/or in the
there may be errors in the parameter input into the evolution laws. Preaging tests are suggested as being the
empirical dislocation multiplication process. A more best way of establishing whether equation (13) ade-
systematic and rigorous computational effort than quately describes long term behaviour in Nimonic 90.
performed here would be needed to establish or deny 2. Proof stress calculations predict that the rapid
this unambiguously. decline in experimentally determined values at tempera-
A much better and conclusive method would be tures .700uC is a consequence of yield being due to
experimental. Equation (13) predicts that a thermal dispersion controlled creep. If true, then proof stress
treatment of Nimonic 90 under zero load would result data should be dependent upon testing speed. Below
in a subsequent increased creep rate when compared to 700uC, proof stress should be independent of testing
the as received alloy, tested under the same stress. speed, since the particle cutting mechanism is athermal.
Figure 6 suggests that a temperature of 750uC would be
sufficiently high to ensure the need for only relatively Acknowledgements
short pretreatment times. There is anecdotal evidence
that similar testing in other nickel base superalloys does Many thanks are due for many fruitful discussions with
indeed reduce creep resistance. In this case, it would be several colleagues over the years: in particular, Doctor

218 Materials Science and Technology 2009 VOL 25 NO 2

 Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

Stephen Osgerby (Alstom, UK) at NPL; the late : : :

Professor Malcolm McLean at Imperial College s~sp wp zsm (1{wp ) (26)
London and his then post doctoral students, Doctor Multiplying equation (23) by (12wp), equation (24) by
Hector Basoalto (QinetiQ, Farnborough) and Doctor wp, adding and substituting equation (26) and further
Sanjay Sondhi (G.E., Bangalore). noting that strain compatibility requires equality of
: : :
eT,m , eT,p and eT,p , (the global (measured) total strain
References rate), gives

: 
1. M. F. Ashby and B. F. Dyson: in ‘Advances in fracture research’, : Em s :
(ed. S. Valluri et al.), Vol. 1, 3–30; 1984, Oxford, Pergamon Press. eT,G ~ z(1{wp )em (27)
Em (1{wp )zEp wp Em
2. L. M. Kachanov: Izv. Ak. Nauk SSSR Otdel Tekh Nauk, 1958, 8,
26–31. When both moduli are equal and given by E, this
3. Y. N, Rabotnov: Proc. XII IUTAM Cong., (ed. H. Hetenyi and W. reduces to
G. Vincenti), 137–141; 1969, Stamford, Springer-Verlag.
4. J. Jason and J. Hult: J de Mécanique Appliquée, 1977, 1, 69–84.
: :
: s : s :
5. M. F. Ashby and D. R. H. Jones: ‘Engineering materials’, 173; eT,G ~ z(1{wp )em (sm ,T,Si ): zeG (28)
1980, Oxford, Pergamon Press. E E
6. G. S. Ansell and J. Weertman: Trans AIME, 1959, 215, 838–843. i.e. global (measured) creep rate is given by
7. L. M. Brown and R. K. Ham: in ‘Strengthening methods in : :
eG ~(1{wp )em . The measured creep rate for the disper-
crystals’, (ed. A. Kelly and R. B. Nicholson), 68; 1971, Amsterdam,
Elsevier. sion controlled model given by equation (12) will
8. U. F. Kocks, A. S. Argon and M. F. Ashby: ‘Thermodynamics & therefore be given by
kinetics of slip’, 93; 1975, New York, Pergamon Press. h i  
9. J. Friedel: ‘Dislocations’, 113; 1964, New York, Pergamon. : 1:6 1=2 sD b2 lp
eG ~ r(1{wp )wp (p=4w) {1 cj Ds sinh (29)
10. J. C. Ion, A. Barbosa, M. F. Ashby, B. F. Dyson and M. McLean: M- MkT
-
‘The modelling of creep for engineering design’, Report DMA
A115, NPL, Teddington, UK, 1986. Defining an internal stress, si5s2sm, equation (29) can
11. B. F. Dyson and M. McLean: Acta Metall, 1983, 31, 17–27. be rewritten
12. J. Heslop and W. Betteridge: ‘The nimonic alloys’, 2nd edn; 1974,
London, Edward Arnold. : 1:6 h i
13. T. B. Gibbons and B. E. Hopkins: ‘The creep behaviour and eG ~ r(1{wp )wp (p=4w)1=2 {1
M-
structure of Ni–Cr base alloys’, Report DMA 253, NPL, UK, 1976. " #
14. M. F. Ashby: Proc. 2nd Bolton Landing Conf. on ‘Oxide (s{si { s^net )b2 lp
dispersion strengthening’, 143; 1968, New York, Gordon and cj Ds sinh (30)
Breach Sci. Pub. MkT
-
15. G. F. Harrison and W. J. Evans: Proc. Int. Conf. on ‘Engineering
aspects of creep’, Sheffield, UK, September 1980, The Institute of This internal stress is now the variable requiring
Mechanical Engineers, 69–76. quantification. Using equation (25) and its above defini-
16. D. Raynor and J. M. Silcock: Met. Sci. J., 1970, 4, 121–130. : wp : : :
tion, it follows that si ~ 1{w (sp {s). Substituting for sp
17. T. M. Pollock and A. S. Argon: Acta Metall. Mater., 1992, 40, 1– p
30. using equation (24) and using equation (28) to eliminate
:
eT,p , gives
Appendix wp E :
:
The following is based upon the work of Ion et al.10 and si ~ eG (31)
1{wp
aims to quantify the stress redistribution that takes place
between matrix and particles as inelastic strain accumu- The load transfer from matrix to particles increases with
lates by creep and to quantify its effect on the measured increasing inelastic strain, with a corresponding decrease
creep rate. in matrix loading and a fall in global (measured) creep
:
Consider a volume fraction wp of particles with rate; its progress is monitored by si . A minimum in creep
Young’s modulus Ep under applied uniaxial stress s. rate will be reached when a stress relaxation mechanism
: becomes active. Several have been suggested in plasticity
The total strain rate, eT,m within the matrix m having
Young’s modulus Em is given by and creep: particle fracture; particle/matrix interface
: decohesion, and inelastic deformation within the particle,
: sm : controlled either by plasticity or creep. Clearly, all are
eT,m ~ zem (sm ,T,Si ) (23)
Em stress level/state controlled and therefore most likely to
The first term in equation (23) is the rate of change of activate at lower temperatures and/or in more creep
matrix elastic strain. The second is the matrix creep rate, resistant (stronger) alloys, where the operational stresses
are higher.
which is a function of the matrix stress sm temperature,
T, and the set of state variables Si, representing An alternative non stress controlled mechanism is the
upper bound collapse value of the internal stress. The
potentially evolving microstructural parameters. Since
stress transfer process has similarities to the one
the particles are assumed to behave elastically, the total
: associated with hardness testing and suggests a way of
strain rate eT,p within the particles is given by 1
getting an approximate value for si . In hardness testing,
:
: sp load is applied to the indenter causing flow in the matrix,
eT,p ~ (24) whereas, in the present case, flow in the matrix induces a
Ep
stress within the particle. The well known result for
Stress and stress rate equilibria at constant volume hardness testing is that the stress on the indenter
fraction are given by (induced stress in the particle) is approximately three
times the tensile yield stress (flow stress within the matrix
s~sp wp zsm (1{wp ) (25) 1
sm). The relationship between si and s can now be

Materials Science and Technology 2009 VOL 25 NO 2 219

Dyson Microstructure based creep constitutive model for precipitation strengthened alloys

easily found using the definition of si (;s2sm) and fields would ‘harden’ the matrix, as Pollock and
1 1
equation (25) with sp ~3sm Argon17 have pointed out for very high volume
fraction superalloys (wc’ %70%). Any of the above
1 2wp stress controlled mechanisms may intervene before si
si ~ s (32)  1
1z2wp reaches si . This will become more likely at
1 1 lower temperatures/stronger alloys because the larger
Thus si and therefore sm are proportional to s through stresses involved will favour a stress controlled inter-
a microstructural term f(wp). Equation (32) contrasts vention.
with the one given by Ion et al.,10 which assumed It is likely that there will be a spectrum of critical
diffusion creep to be the stress relaxation mechanism internal stresses within the microstructure – reflecting
and is therefore potentially the lower bound stress local differences in volume fraction for example – rather
controlled mechanism. 1
than a unique value si . A stochastic approach to si
The implicit assumption here is that the strain fields evolution seems therefore to be appropriate and a simple

between particles do not interact significantly. At best, 1{si
this would limit the use of equation (32) to low volume empirical probability function 1
si
has been used to
fraction alloys – a simple slip line argument suggests transform equation (31) into the following
less than 20–30% volume fraction. For larger volume !
1
fractions, si would be expected to be greater than : wp E si :
si ~ 1{ 1 eG (33)
given by equation (32) because the interaction of strain 1{wp si

220 Materials Science and Technology 2009 VOL 25 NO 2