S. Spigarelli, R. Sandström
www.elsevier.com/locate/msea
PII: S0921-5093(17)31514-9
DOI: https://doi.org/10.1016/j.msea.2017.11.053
Reference: MSA35768
To appear in: Materials Science & Engineering A
Received date: 27 September 2017
Revised date: 10 November 2017
Accepted date: 14 November 2017
Cite this article as: S. Spigarelli and R. Sandström, Basic creep modelling of
a l u m i n i u m , Materials Science & Engineering A,
https://doi.org/10.1016/j.msea.2017.11.053
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Basic creep modelling of aluminium
S.Spigarelli1, R. Sandström2
1. DIISM, Università Politecnica delle Marche, via Brecce Bianche 60131 Ancona, Italy
2. Materials Science and Engineering, KTH Royal Institute of Technology, Brinellvägen 23, S-
*
Corresponding author: Stefano Spigarelli. tel. +39 071 2204746. e-mail: s.spigarelli@univpm.it
Abstract.
In recent years a basic creep model that does not involve adjustable parameters has been developed.
The main feature of this model is that it is fully predictable and the assumptions at its basis can be
easily verified once the output is compared with experimental data. This model, initially developed
for pure copper, has been here applied to pure aluminium. A critical issue has been identified with
the controlling mechanisms during power-law breakdown. The increase in the creep rate at high
stresses and low temperatures can be quantitatively explained from the raised climb rate due to the
deformation-induced increase in concentration of vacancies. The model can also account for the
fairly wide range of stresses where aluminium follows power-law creep with a creep exponent of 4
to 5. At slightly lower stresses, the creep exponent increases somewhat due to the presence of an
internal stress. Since no adjustable parameters have been required, the model represents a notable
enhancement over the conventional approach, which is based on the use of the power-law equation
and requires fitting of experimental data to determine the values of the material parameters.
The creep response of pure aluminium has attracted a great deal of academic interest, leading to the
publication of many papers, if one considers their limited, to be fair, industrial relevance as creep-
resistant alloys. A non-exhaustive picture of the extensive coverage of this subject can be obtained
from the relevant chapters in [1-3]. The cause for such a debate can be traced back to the very
nature of this material, whose simple microstructure was an ideal case-study to identify the main
dislocation creep mechanisms and separate their relative contributions to creep strength.
In most cases, the secondary creep rate ( ) dependence on applied stress (σ) and temperature (T)
has been described by the conventional power-law and Arrhenius equations, in the form
n
D Gb Q
A 0sd exp sd 1)
kT G RT
where A is a material parameter, k is the Boltzmann constant, G is the shear modulus (G=30220-
16T MPa for Al), b is the length of the Burgers vector (b=2.86x10-10m for Al) and R is the gas
constant. D0sd in Eqn.1 is the pre-exponential factor in the Arrhenius equation for self-diffusion and
Qsd is the activation energy for self-diffusion in the metal. The stress exponent n in aluminium is
about 4 to 5 at temperatures above 250ºC [4,5]. Above a certain stress, power-law breakdown
occurs and the slope of the curve describing the strain rate dependence on applied stress in double-
log coordinates increases progressively with applied stress level. At room temperature, the stress
activation energy equivalent to that for self-diffusion identifies the so-called “class M” (Metal)
materials. The typical class-M response of pure coarse grained Al is shown in Figure 1, which plots
the steady state creep rate as a function of stress from the well-known Servi-Grant dataset for Al
99.995% (grain size 2mm) [5] and other results on Al 99.999% obtained by Mecking at al (grain
Several theories, reviewed in [7,8], have been later developed to justify the power law creep
behaviour of pure metals and the resulting values of stress exponent and activation energy.
Although Nabarro considered these theories “unconvincing” [8], they could nevertheless represent a
basis to develop new creep models for more complex materials. Yet, the creep community was
almost invariably satisfied with using the power law in its phenomenological form to describe the
response of innumerable materials, even with extremely complex microstructures (see [9-16] for
just few very recent examples). In these cases, n, Q and A are calculated by a best fitting procedure
of the experimental data and cease to have a direct correlation with microstructural phenomena.
Even in those cases where authors attempt to develop constitutive models based on the physics of
phenomena, a number of material parameters need to be calculated by fitting the experimental data
(see, for example, [17]). The fact is that the power law remains phenomenological in nature, and
for this reason, alternative approaches have been sought to introduce constitutive equations directly
derived from the operating mechanisms. In this line of thought, Sandström has proposed models for
plastic deformation of fcc metals based on physical mechanisms [18,19]. These models have been
successfully applied to copper and austenitic stainless steels, see for example [20,21], and can
handle not only power-law creep but also power-law breakdown. For example, the equations
proposed can correctly predict a creep exponent as high as 65 for copper at 75ºC [21]. These
models do not require any best-fitting procedures, being based only on a number of physical and
constitutive equations [22,23]. Since no adjustable parameters are involved, the models are fully
predictable. Thus, to predict the creep rate, the only data required are the experimental ones (stress
and temperature, in the case of a pure metal, where composition is not an issue). The
straightforward development of the model is its application to the other metal which, with Cu,
constitutes the classical case study of creep of class-M materials, i.e. pure Al [8]. Only once the
suitability of the basic model to describe the creep response of the high-purity metal is attested,
subsequent developments to describe the effects of the atoms in solid solution and of secondary-
The model presented in [18-23] is applicable to fcc metals with large grain size. This means that
grain size (d) should be larger than subgrain size (dsub). When grain and subgrain sizes become
equivalent, deformation mechanisms change and grain boundaries get a more dominant influence
[24]. Materials can be classified according to their grain size. For example, Mohamed gives 20 m
as the lower limit for large grain materials. Micrograined materials fall in the range 1–10 m,
ultrafine-grained (UFG) materials have grain sizes between 300 and 900 nm and nanocrystalline
(nanograined) materials between 20 and 200 nm [25]. A group of ultrafine nanocrystalline materials
with grain sizes below 10 nm can also be introduced. Although fine-grained materials will not be
discussed in this paper, it might be of interest to give some brief general comments, since
deformation mechanisms in fine grained materials are extensively covered in literature and in
creep rate. There are numerous models for describing how this can take place. Most authors agree
that the dominating mechanism is grain boundary sliding (GBS), i.e. deformation takes place by
neighbouring grains sliding against each other. GBS cannot occur without an accommodating
mechanism around the grain boundaries. A number of empirical models for GBS have been
formulated for fine-grained materials, but no basic model seems to be available (for a review, see
[26]). In agreement with experiments, these models suggest a stress exponent of 2 and a creep rate
that is inversely proportional to the square of the grain size for micro-grained materials [26,27].
This dependence is also frequently observed for superplastic alloys. For both micro-grained and
slip. It is therefore quite possible that the behaviour in the micro-grained range can be extended into
the ultrafine-grained range, but little experimental data is available to confirm it. Nanocrystalline
materials have a reduced dislocation density. In spite of this, accommodation takes place by
ordinary dislocation slip [28] and by emission and absorption of partial dislocations at the grain
boundaries [29]. The partial dislocations lead to the formation of deformation twins that can
improve both strength and ductility by increasing the dislocation and thereby work hardening. The
presence of deformation twins has been confirmed both by means of transmission electron
microscopy and molecular dynamics (MD) simulations. This is also observed for high stacking fault
materials such as aluminium [30]. For ultrafine nanocrystalline materials, the accommodation
mechanism is assumed to be based on diffusion creep. This result is primarily obtained through MD
simulations [31]. In summary, GBS is believed to be of importance at all grain sizes, but the
The brief analysis above clearly shows that very fine-grained Al alloys behave quite differently with
respect to conventional coarse-grained alloys, where creep is controlled by dislocation climb and
glide and class-M behaviour can be easily identified. For this reason, they will be not analysed in
this paper, which will rather deal with dislocation creep of pure coarse-grained aluminium. The
basic model introduced in [18-23] will be applied. In particular, the model results in an enhanced
creep rate at high stresses and low temperatures, which, as mentioned above, originates from a
questioned. Thus, a theoretical justification of the enhanced dislocation mobility in the region of
“power law breakdown” will be also presented, by introducing a new physical parameter, i.e. the
2. The model
The model, originally developed for Cu, is based on physically derived equations. It will now be
The total dislocation population is formed by dislocations stored in subgrain boundaries and free
dislocations forming a network inside the subgrains. In special cases, it is essential to distinguish
between dislocations in the sub-boundaries and those in the subgrain interiors. This applies for
example to cold worked materials [32]. However, this distinction is not important for pure
aluminium at high temperature, where free dislocation density plays a major role, and will therefore
be ignored here. The free dislocation density (ρ) is usually related to the stress by the well-known
i d i mGb 2)
where m is the Taylor factor (m=3.06 for fcc metals) and d mGb 1 2 is the dislocation
hardening term. The internal stress σi represents the strength of the pure annealed large-grained
metal, i.e. the stress required to move a dislocation in the absence of other dislocations, and is a
constant (=0.2-0.4; the intermediate value 0.3 will be considered in the following).
The evolution of the dislocation density during straining can be expressed as [18,32]
d m 2
M l 2 3)
d bL
where is a constant ( =14.7 in pure Cu), l is the dislocation line tension (l=0.5Gb2), M is the
dislocation mobility and L is the dislocation mean free path, i.e the distance travelled by a
CL
L 4)
CL being a strain-hardening constant. Equation 4 remains valid as long as it gives a value of the
dislocation mean free path smaller than the grain size. This implies that, since the dislocation mean
free path cannot exceed the grain size, for vey fine-grained materials, under very low applied
stresses, i.e. for very low dislocation densities, L should equal d. This is not obviously the case for
The first term on the right-hand side of Eqn.3 represents the strain hardening effect due to
dislocation multiplication, which is more rapid when L and, consequently, CL assume low values
and/or the dislocation density is high. The second term on the right-hand side of Eqn.3 describes the
effect of dynamic recovery, i.e. strain dependent recovery. The third term takes into account static
recovery, which is time dependent recovery. Unfortunately, the different types of recovery are not
named consistently in literature, which is something one should be aware of. In this context, for
consistency reasons, in this work the terminology used in [18,19,22,23] was adopted.
d b3 Q
exp exp sd
D0 sd b
Mc 5)
kT RT
kT
Creep results obtained in the low-temperature/high strain rate regime are characterised by higher
stress exponents and lower values of the activation energy for creep. It is often suggested that this
requires that glide is taken into account. Following [34], the glide of dislocations through an
p
q
d Q
f 2 exp 1 d 6)
dt RT Rmax
where f, Q, p, q and Rmax are unknown parameters. Rmax was specified as the maximum back stress
and was chosen as the true tensile strength, corrected to take into account the effect of necking in
very ductile metals. To find the values of f and Q, a unified model for climb and glide was
formulated according to a procedure proposed by Nes et al [35]. The resulting expression for the
p
q
d b3
M cg
D0 sd b
exp exp
Qsd
1 d
7)
kT Rmax
kT RT
Originally, the ranges 0<p1 and 1<q2 were suggested [34]. Other studies showed that Eqn.6
works very well for copper and austenitic stainless steels with p=2 and q=1 with Rmax equivalent to
the true stress corresponding to the ultimate tensile strength of the material [21]. This has also been
At steady state, with the climb glide mobility according to Eqn. 7, Eqn.3 gives
2M cg l bCL d
3
ss 8)
m CL d mG mGb
The term in the denominator that involves is a consequence of the dynamic recovery. It is
especially important when describing tertiary creep [37]. The model based on the combination of
Eqns.7 and 8 requires the determination of two constants (CL and ω) and of σi. Roters et al gave a
mdint 1
2 9)
b nslip
dint is the interaction distance where dislocations of opposite sign can annihilate each other and
nslip=12 is the number of slip systems; dint can be taken as dislocation core diameter. With ab initio
calculations, a core diameter for aluminium of 2.5 b was obtained [39], which gives = 15.
In Eqn.3 the dynamic and static recovery terms give a maximum dislocation density. For the
2
m
max
DR
10)
bCL
At low temperatures, the dynamic recovery and static recovery terms in Eqn. 3 have the same size
as for copper when the maximum dislocation densities are inserted [6]. By analysing creep data, it
can be shown that the same applies to aluminium at room temperature. This means that Eqn.10 can
m m 2G
CL 11)
b DR 0.5
max Rmax y
In the second equality, the Taylor Eqn.2 is used. The maximum dislocation stress is estimated as the
difference between the tensile strength Rmax and the yield strength y.
At higher temperatures, the static recovery term in Eqn.3 is larger than the dynamic recovery term.
bC L
0.5 13)
2M cg l m
where Eqn.12 has been inserted. As mentioned above, this term is close to unity at room
temperature, so the change in the dislocation density must be taken into account
m T 0.5 m m2G
CL 14)
bT 0.5 RT 0.5 b RT 0.5 Rmax y
where the suffixes T and RT indicate high-T and room-T respectively, i.e. the same result as in
Eqn.11. In the derivation, it was assumed that there is no temperature dependence of CL, an
assumption that is consistent with the result of Eqn.14. With Rmax 75 MPa (1.5RUTS) and a yield
strength close to 20 MPa, CL = 85 is obtained. For a dislocation density as low as 5x1011 m-2, this
value gives a dislocation mean free path close to 120 m, which is well below the grain size of the
The determination of the internal stress, which is temperature and strain rate dependent, is here
based on the assumption that the annealed dislocation density (a) and the σi values account for the
annealed yield strength of the pure metal. The yield stress is thus given by Eqn.2 [40], where a is
virtually nihil. The term 0 includes the dependence on grain size, that is
kd
i i0 15)
d
where i0, the internal stress for a material with an infinitely large grain size, depends on the
impurity content of the metal, while kd is the Hall-Petch constant. The variation of kd as a function
of temperature, reported by Blum et al. [4], was used to estimate the effect of grain size on the yield
stress of high-purity Al as a function of temperature (grain size d1=1 mm, strain rate 5x10-4 s-1 [40].
A way to describe the temperature dependence of the yield strength is to assume that it is
proportional to the creep strength in the creep range, whereas below the creep range the yield
strength is proportional to the shear modulus. Pure Al data cover all the interval between these two
extreme cases, and for this reason the following expression was used
i A y creep G 16)
where creep is the creep stress which corresponds to a given steady state creep rate. For a strain rate
of 5x10-4 s-1 as in [40], the Servi and Grant dataset for pure-Al with a 2mm grain size [5] was used
to obtain the relevant creep stress at different temperatures. The Ay constant was determined to
obtain a reliable estimate of the yield strength of high-purity coarse-grained Al. The results
presented in [40] for an 99.999 Al with d=1mm can be used to recalculate the yield stress (σyi) of a
kd kd
yi y1 17)
d di
which, combined with Eqn.2, gives, for di=2mm and a=1x1010 m-2, Ay=4.2x10-3.
Comparing Eqns.5 and 7 one finds that the dislocation mobility is enhanced by the following factor
Q d
2
fc lg l exp sd 18)
RT Rmax
assuming p =2 and q =1. This expression will now be justified. During plastic deformation the
number of vacancies increases. The non-conservative motion of dislocations generates and absorbs
vacancies. Mecking and Estrin formulated a model, which demonstrated that there is a net increase
in the vacancy concentration due to plastic deformation [41]. Their result can be expressed as
c 22 d
0.5 19)
c0 Dsd G
spacing between vacancy sinks. In [41], a factor 0.1 was used in Eqn.19. However, a detailed
derivation shows that it should be replaced by 0.5 and a factor 2 should be introduced. Following
[41], is taken as the subgrain size dsub. It is well known that the subgrain size can be related to the
applied stress
K subbG
d sub 20)
Ksub is a constant that is about 18 for aluminium [42]. Inserting Eqn.20 into Eqn.19 gives
c 2 K sub 2 b 2 d G
0.5 21)
c0 Dsd 2
This increase in the vacancy concentration raises the climb rate in proportion. Thus, the resulting
c
gc lim b 1 22)
c0
The expression 18) is compared to the vacancy concentration enhancement (Eqn.22) in Figure 2. In
the Figure, experimental values for the creep rate are used and a good general agreement between
Eqns.18 and 22 is obtained. The temperature, stress and strain rate dependencies are well covered.
Quite a surprising result was obtained. The enhancement of the creep rate in relation to the high
temperature climb model in Eqn.7 is well justified for copper and austenitic stainless steels [20]. It
was inspired by an empirical model for glide. Instead, it turns out that it can be fully explained by
the enhancement of the climb rate at low temperatures due to the increased vacancy concentration.
This should not be considered as the final proof that creep at ambient temperatures is fully
controlled by climb. However, it makes it much simpler to explain why full stationary creep is
observed in aluminium and copper at near ambient temperatures. Yet, taking glide and cross slip
into account results in a continuous build-up of a forest of edge dislocations, which would make the
creep rate slow down and eventually stop, as it is observed for logarithmic creep [43].
The two datasets shown in Figure 1 are substantially consistent with each other, except at
intermediate temperatures, where the Servi-Grant creep rates seem to be somewhat lower when
The first point to be addressed is the selection of the proper value of the self-diffusion coefficient.
As a matter of fact, this apparently trivial problem requires an analysis of the most recent findings
on this subject. The “traditional” values of the diffusion coefficient parameters for self-diffusion in
Al are D0sd=1.86x10-4 m2 s-1, Qsd=143.4 kJ mol-1 [44] or D0sd=1.72x10-4 m2 s-1 and Qsd=142.1 kJ
mol-1 [45,46]. The accuracy of these estimates was challenged in [47], which reported a wide
collection of literature results (Figure 3). The Figure shows how the value of 143 kJ/mol can
describe the high-temperature literature data very well, but strongly underestimates the value of the
diffusion coefficient at lower temperatures. An excellent description of the data collected in [47] is
Since all the parameters have now been quantified, the equations can be used to model the steady
state creep rate dependence on applied stress, by simply substituting σ and T into Eqns.7, 8 and 16,
with CL=85, =15, Rmax =75 MPa, Ay=4.2x10-3, Qsd=122 kJ mol-1 and D0sd=8.3x10-6 m2s-1 (Figure
4), without any data-fitting. The correlation between the theoretically and calculated curves and the
experimental data from the Mecking et al dataset is excellent; a larger deviation is indeed observed
in the case of the Servi-Grant results, at intermediate (533K) temperatures, although on a whole the
Figure 5 plots the fclgl and gclimb factors, now calculated from the model values of the strain rate. The
excellent agreement already observed in Figure 2 is confirmed, which indicates that a detailed
An interesting implication of the model used in this study is that it results in a marked curvature of
the strain rate vs stress plots for stresses below 1 MPa. This effect is simply caused by the presence
of an internal stress that, once calculated by Eqn.16, becomes non-negligible in this region when
compared to the stress applied. Although Eqn.16 can be supposed to overestimate the internal stress
in the very-low stress / very-high temperature region, the same experimental behaviour was indeed
Figure 6 plots a collection of dislocation density values as a function of stress for Al [40,49] and for
Al-Zn, which is thought to behave as the pure metal on this regard [40]. The same Figure 6 shows
the curves obtained by Eqn.2 with =0.3 and i=0, and with i given by Eqn.16. The description of
the experimental data is excellent, except in the low stress region, where, as mentioned above,
Eqn.16 is thought to overestimate the internal stress. Incidentally, it is interesting to observe that the
data reported in the Figure for pure-Al, with the exception of the value corresponding to the lowest
stress, were obtained by relating the RT yield stress after various levels of severe plastic
deformation by ECAP (Equal Channel Angular Pressing) and the corresponding dislocation
densities measured by X-ray analysis [49]. Data in Figure 6, which thus represent the total
dislocation content at room temperature, closely align on the same curve identified by steady state
flow stress at high-T and the corresponding free dislocation contents [40]. At high temperature,
torsion experiments in pure-Al demonstrated that the steady-state value of the flow stress is
unaffected by substantial changes in subgrain size and sub-boundary orientation [50]. On the other
hand, cell formation in Cu was observed to significantly affect the creep response at low
temperatures [32], an effect that was attributed to the high density of dislocation forming cell-walls.
A similar mechanism can be expected to be operative also in pure Al, and, as a matter of fact, the
model presented in Section 2, which considers only the free dislocations, somewhat overestimates
the steady state creep rate at 300 and 366 K (Figure 4). Gubicza et al. [49] reasoned that the fact that
the yield strength obeys the Taylor equation using the dislocation density values determined by X-
ray line profile analysis attests that the main strengthening mechanism is still the interaction
between dislocations, also in these ultrafine-grained material, irrespective whether they are stored in
5. Conclusions
A basic model for the description of creep in fcc metals has been applied to pure Aluminium.
The set of physically based constitutive equations proposed contains two important parameters: the
internal stress, representing the stress required to move a dislocation in the matrix and the strain
hardening constant CL, which, in combination with the free dislocation density, determines the
A comparison between model prediction and data suggests that power-law breakdown, i.e. the
increase in strain rate at low temperatures, can be quantitatively explained from the raised climb
rate due to the deformation-induced increase in concentration of vacancies. The model proposed,
which does not require any specific variation in constitutive equations to describe the experimental
range of steady state creep rate, can also account for the fairly wide range of stresses where
excellent description of the secondary creep rate with stress. In this sense, the model represents a
notable enhancement over the conventional approach, which is based on the use of the power-law
equation, thus constituting an excellent base for further development aimed at the description of
single phase solid solution alloys (class A metals) and age hardening aluminium alloys.
Table of most important fundamental symbols
[1] H.Oikawa, T.G.Langdon, The creep characteristics of pure metals and metallic solid solution
alloys, in Creep behaviour of crystalline solids, R.Wilshire and R.W.Evans Eds., Pineridge Press,
[3] M.E.Kassner, M.T.Pérez-Prado, Fundamentals of creep in metals and alloys, Elsevier, 2004, 11-
120.
[4] W. Blum, J. Hausselt, G. König, Transient creep and recovery after stress reduction during
[5] I.S. Servi, N.J. Grant, Creep and stress rupture behaviour of aluminium as a function of purity,
[6] H. Mecking, A. Styczynski, Y.Estrin, Steady state and transient plastic flow of Aluminum and
[7] F.R.N.Nabarro, Do we have an acceptable model for power-law creep?, Mater.Sci.Eng. A387-
[8] F.R.N. Nabarro, Creep in commercially pure metals, Acta Mater. 54 (2006) 263-295.
[9] A.A.Khamei, K.Dehghani, Effect of strain rate and temperature on the hot tensile deformation
of severe plastic deformed 6061 aluminum alloy, Mater.Sci.Eng. A627 (2015) 1-9.
[10] Q.Zhang, W. Zhang, Y. Liu, Evaluation and mathematical modeling of asymmetric tensile and
creep and uniaxial creep measurements for commercial purity aluminium, Acta Mater. 111 (2016)
31-38.
[12] Y.Li, Z.Shi, J.Lin, Y.-L.Yang, Q.Rong, Extended application of a unified creep-aging
constitutive model to multistep heat treatment of aluminium alloys, Mater.Des. 122 (2017) 422-432.
[14] Y.Xu, L.Zhan, L.Xu, M.Huang, Experimental research on creep aging behaviour of Al-Cu-Mg
alloy with tensile and compressive stresses, Mater.Sci.Eng. A682 (2017) 54-62.
[15] L.Zuo, B.Ye, J.Feng, X.Kong, H.Jiang, W.Ding, Effect of Q-Al5Cu2Mg8Si6 phase mechanical
[16] X.Jiang, Y.Zhang, D.Yi, H.Wang, X.Deng, B.Wang, Low-temperature creep behaviour nd
[17] C.Li, M.Wan, X.-D.Wu, L.Huang, Constitutive equations in creep of 7B04 aluminum alloys,
[18] R. Sandström, J. Hallgren, The role of creep in stress strain curves for copper, J. Nucl. Mater.
[19] R. Sandstrom, Basic model for primary and secondary creep in copper, Acta Mater. 60 (2012)
314-322.
[20] S. Vujic, R. Sandstrom, C. Sommitsch, Precipitation evolution and creep strength modelling of
[21] R. Sandström, Fundamental Models for Creep Properties of Steels and Copper, Trans. Indian
[23] R. Sandström, H.C.M. Andersson, Creep in phosphorus alloyed copper during power-law
[24] T.G. Langdon, A unified approach to grain boundary sliding in creep and superplasticity, Acta
[25] F.A. Mohamed, Deformation mechanism maps for micro-grained, ultrafine-grained, and nano-
[27] F.A. Mohamed, On the creep transition from superplastic behavior to nanocrystalline behavior,
[28] F.A. Mohamed, Correlation between the deformation of nanostructured materials and the
[29] Y.T. Zhu, T.G. Langdon, Influence of grain size on deformation mechanisms: An extension to
[30] S. Ni, Y.B. Wang, X.Z. Liao, R.B. Figueiredo, H.Q. Li, S.P. Ringer, T.G. Langdon, Y.T. Zhu,
The effect of dislocation density on the interactions between dislocations and twin boundaries in
[31] T.G. Desai, P. Millett, D. Wolf, Is diffusion creep the cause for the inverse Hall–Petch effect in
[32] R. Sandström, The role of cell structure during creep of cold worked copper, Materi.Sci.Eng.
(1979) 1
[35] E. Nes, K. Marthinsen, Modeling the evolution in microstructure and properties during plastic
[36] H.D. Chandler, Effect of unloading time on interrupted creep in copper, Acta Metall.Mater. 42
(1994) 2083-2087.
[37] R. Sandström, Formation of a dislocation back stress during creep of copper at low
approach based on three internal state variables, Acta Mater. 48 (2000) 4181-4189.
[39] R. Wang, S. Wang, X. Wu, Edge dislocation core structures in FCC metals determined from ab
initio calculations combined with the improved Peierls-Nabarro equation, Physica Scripta, 83
(2011).
[40] M.E.Kassner, Taylor hardening in five-power-law creep of metals and Class-M alloys, Acta
[41] H. Mecking, Y. Estrin, The effect of vacancy generation on plastic deformation, Scripta Metall.
14 (1980) 815-819.
[42] C.M. Young, S.L. Robinson, O.D. Sherby, Effect of subgrain size on the high temperature
strength of polycrystalline aluminium as determined by constant strain rate tests, Acta Metall. 23
(1975) 633-639.
[43] F.R.N. Nabarro, The time constant of logarithmic creep and relaxation, Mater.Sci.Eng. A309–
[45] T. S. Lundy and J. F. Murdock, Diffusion of Al26 and Mn54 in Aluminum, J. Appl. Phys., 33
(1962) 1671-1673.
[46] S. L. Robinson and O. D. Sherby, Activation energy for lattice self-diffusion in Aluminium,
[47] L. Zhang, Y. Du, Q. Chen, I. Steinbach, B.Huang, Atomic mobilities and diffusivities in the
fcc, L12 and B2 phases of the Ni-Al system, Int. J. of Mater. Res. 101 (2010) 1461-1475.
[48] S. Straub, W. Blum, Does the “natural” third power law of steady state creep hold for pure
[49] J. Gubicza, N.Q. Chinh, T.G. Langdon, T. Ungár. Microstructure and strength of metals
processed by severe plastic deformation, Proc. Ultrafine Grained Materials IV, Y.T.Zhu,
[50] M.E. Kassner, M.E. McMahon, The dislocation microstructure in Aluminum deformed to very-
large steady state creep strains, Metall. Trans. 18A (1987) 835-846.
List of Figure Captions
Figure 1. Experimental steady state creep rate as a function of stress for Al99.995% [5] and
Al99.999% [6]. In the former case, the data were obtained from constant stress creep experiments
Figure 2. Comparison of climb glide factor fclgl in Eqn.18 with increase in vacancy concentration
due to plastic deformation gclimb in Eqn.22 as a function of stress, using experimental values for
stress and creep rate from [6]. A detail of the high-T trends is given on the right.
Figure 3. Diffusion coefficient for self diffusion (see ref.[47] for the sources of the original
experimental data). The Figure also plots the curves describing the most widely used relationships
(Qsd=143) and the dependence of the coefficient of diffusion on temperature recalculated in this
Figure 4. Description of the steady state creep rate as a function of the applied stress described by
the p=2 model, and comparison with the experimental data presented in Figure 1. The curvature in
Figure 5. Comparison of climb glide factor fclgl in Eqn.18 with the increase in vacancy concentration
due to plastic deformation gclimb, Eqn.22 as a function of temperature. Results are shown for five
strain rates.
Figure 6. Free dislocation density as a function of modulus compensated stress for pure Al [39,48]
and Al-Zn (from [40]). The curves were obtained by Eqn.2 with =0.3, being i=0 (broken line) or
i given by Eqn.16 (solid line). Also in this case, the effect of a slight overestimation of the internal