You are on page 1of 11

KINETICS OF FLOW AND STRAIN-HARDENING7

H. MECKINGS and U. F. KOCKS


lnstitut fur Allgememe Metalikundc und Metallphysik. RWTH Aachen. W. Germany, and
Materials Science Division, Argonne National Laboralory, Argonne, IL 60439. U.S.A.

Abstract-The kinetics of glide at constant structure and the kinetics of structure evolution are corre-
lated on the basis of various experimental observations in pure f.c.c. mono- and polycrystals. Two
regimes of behavior are identified. In the initial regime, the Cottrell-Stokes law is satisfied, hardening is
athermal. and a single structure parameter is adequate. With increasmg importance of dynamic recovery,
be it at large strains or at high temperatures, all of these simple assumptions break down. However, the
proportionality between the flow stress and the square-root of the dtslocation density holds, to a good
approximation, over the entire regime; mild deviations are primarily ascribed to differences between the
various experimental techniques used. A phenomenological model is proposed, which incorporates the
rate of dynamic recovery into the flow kinetics. It has been successful in matching many experimental
data quantitatively.

R&urn&Nous avons relic la cinttique du glissement a structure constante et de I’tvolution de la


structure a partir de diverses observations sur des monocristaux et des polycristaux c.f.c. purs. Nous
avons distingut deux regimes. Au tours du premier regime. la loi de Cottrell et Stokes est virifiee, le
durcissement est athetmique et il suffit dun seul paramttre pour la structure. Lorsque Timportance de la
restauration dynamique augmente, que ce soit aux grandes deformations ou aux temperatures Clevees,
toutes ces hypotheses simples s’effondrent. Cependant, la contrainte d’ecoulement reste proportionnelle a
la racine car&e de la densitt de dislocations, a une bonne approximation, dans tout le regime; quelques
faibles deviations sont principalement attribuC;es aux diR&et&s entre Ies diverses techniques exptrimen-
tales utilis&s. Nous proposons un modtle phtnomenologique, qui permet d’introduire la vitesse de la
restauration dynamique dans la cinttique de l%coulement. I1 a permis de rendre compte quantitative-
ment d’un certain nombre de resultats experimentaux.

Zu.sammenfaasuag-Anhand verschiedener experimenteller Beobachtungen an reinen kfz. Ein- und


Vielkristallen wurden die Kinetik der Gleitung bei konstanter Struktur und die Kinetik der Struktur-
entwicklung miteinander korreliert. Zwei Bereiche unterschiedlichen Verhaltens wurden aufgefunden. Im
Anfangsbereich ist das Cottrell-Stokes-Gesetz erfdllt, die Verfestigung ist athermisch und ein einziger
Strukturparameter ist ausreichend. Mit zunehmendem EinfluB der dynamischen Erholung bei groger
Dehnung oder hoher Temperatur werden alle diese einfachen Annahmen ungtiltig. Der lineare Zusam-
menhang zwischen FlieDspannung und Wurzel aus der Versetzungsdichte dagegen bleibt in guter N;iher-
ung im gesamten Bereich bestehen. Kleinere Abweichungen werden im wesentlichen den Unterschieden
zwischen den verschiedenen experimentellen Techniken zugeschrieben. Es wird ein phlnomenologisches
Model1 vorgeschlagen, welches die Geschwindigkeit der dynamischen Erholung bei der Kinetik des
FlieDens beriicksichtigt. Dieses Model1 kann viele experimentelle Daten quantitativ beschreiben.

INTRODUCI’ION: SIXUCIURE deformation; but it is also true in a permanent sense


AND CHANGE when the material strain hardens.
A classical abstraction that divides plastic behavior
It is a fundamental tenet of materials science that the into two mechanistic steps has nevertheless proved
current properties of a piece of material depend en- useful: the flow stress a depends on the current struc-
tirely on its current metallurgical structure, and that ture
changes in properties result from changes in structure.
The structure may, however, change as the property is d = u(p;k, T) (la)
being tested, for example by aging. The dislocation and the structure evolves with strain l
structure occupies a specieal place in its relation to
mechanical properties, because it necessarily changes dp
- = E(p). (lb)
during the test. This is obviously true in a dynamic ck
sense, since it is dislocation motion that causes the In these simplified relations, it has been assumed that
the dislocation structure can be represented by the
t Work supported in part by the U.S. Department of
single parameter p, the ‘dislocation density’ (i.e., some
Energy.
$ Now at Technical University, Hamburg-Harburg, W. appropriate average of the dislocation distribution).
Germany. We have further assumed that, at given p, the flow
1866 MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING

stress depends on strain rate 6 and temperature T, but ‘plastic resistance’); however, a measurement of both
that the rate of evolution of p does not. Such a rate- the current flow stress and the current strain-harden-
independence of strain hardening is usually implied in ing rate might be sufficient. It is seen that the above-
phenomenological plasticity. More general cases, and mentioned possibility of a dependence of the ROW
more specific ones, will be discussed below. stress on the current rate of evolution could then be a
A measurement of the flow stress ‘at given struc- consequence of its dependence on two structure par-
ture’ [equation (la)] requires that a suitable back- ameters. We will find such a correlation borne out in
extrapolation method has been found which elimin- the experiments.
ates all efTectsof ongoing structure change [I, 23. This
is not necessarily possible, even in principle: the flow FLOW STRESS AND DISLOCATION
stress may also depend on the current rate of change DENSITY
of the dislocation structure [3-s], so that the separ-
ation of constant-structure kinetics from evolution is As a point of departure, we will discuss the specific
not as clear-cut as was implied in the relations (la) relation between flow stress and dislocation density
and (lb) [2]. It is the purpose of this paper to present that is in common usage. The flow stress ‘at zero
and review a number of experimental observations temperature’ is set equal to
that are most easily explained by such an assumption,
i = @bJp (3)
and are incompatible with relations (1). The same ex-
periments show, on the other hand, that the simple where b is the magnitude of the Burgers vector, P an
relations (1) are in fact well obeyed over a certain appropriate [7] shear modulus and i a constant of
range of external conditions: generally speaking, low order unity which depends, in part, on the strength of
temperatures and low strains. the dislocation/dislocation interaction. Thermal acti-
First, let us generalize the formal relations (1) in vation may lower this effective obstacle strength [7]
two respects: allowing that not only the flow stress, so that the flow stress at a finite temperature and
but also the rate of evolution of the substructure with strain rate becomes
strain may depend on strain rate and temperature;
c = ~(2,T)+b,/jp 1 ctpb& (4)
and that it requires two parameters (here called p1
and p2) to represent the macroscopic effects of the where s(G,7’) is a function that goes to 1 as T+ 0; it
substructure adequately: will be discussed later, on the basis of thermal acti-
vation kinetics. For now, it matters only that the flow
~=a(P,,Pz;&T) @a) stress is a product of a rate sensitivet term and a
Then, rwo evolution laws are required: structure sensitive term.
The flow stress as given by equation (3) and (4)
dp,
-=
de
~l(Pl.Pz;&n relates only to the impediment to dislocation motion
that is provided by other dislocations. In most
dp2 materials, there are other contributions to the plastic
- = E2hrPl;C n resistance. In some cases (e.g., lattice resistance [7],
de
solution hardening [7J, some grain size effects [IO]),
This set of relations will be sufficiently general for these are additive to the contributions discussed
all applications to be discussed here. An extension to above:
further parameters may be necessary when stress
0 = a@, T)*@& + a&, T). (5)
reversals are incorporated [6].
Comparison of relations (2) and (1) reveals a signifi- The rate dependence of co may be more important
cant difference relevant to phenomenological descrip- than that of a (or s), or it may be negligible; the less
tions of plasticity, in which the metallurgical structure rate sensitive term is often called an ‘internal stress’.
is of no concern, but instead the notion of a mechan- Some other contributions to the flow stress obey
ical stare of the material is introduced. When relation more complicated, nonlinear superposition rules [l 1-J.
(la) holds, a measurement of the flow stress at any For the purposes of this paper, we assume that the
standard strain rate and temperature identifies the superposition problem has been solved so that, from
current state of the material [7-g]: it is a macroscopic now on, u identifies only the dislocation/dislocation in-
equivalent to p. The state of a material obeying equa- teraction component of the flow stress, which changes
tion (2a). on the other hand, cannot be completely with strain (equation 4). The experimental results we
specified by one measurement [9]. such as of the flow will discuss relate only to pure f.c.c. materials, in
stress at standard conditions (the ‘reference stress’, the which all other contributions can be shown to be neg-
ligible (except at very small stresses). Furthermore, we
t All rate and temperature dependencies of interest here presuppose that self-consistent use of orientation
are due to thermal activation and are thus coupled. When factors is made in any application; we use u and e in
the influence of thermal activation is important, we say the
the general sense of ‘equivalent’ stresses and strains,
behavior is ‘rate sensitive’; when not, ‘athermal’. The tem-
perature dependence of p is assumed known and incorpor- and use r and y only when experimental data are
ated in the analysis. specifically reported in terms of resolved quantities.
MECKLNG AND KOCKS: KINETICS OF FLOW AND STRAlN-HARDENING 1867

TEM results

-
A (III> ”

PX Stoker, Holt
l PX Essmonn, et 01
-

Etch pit results

Fig. 1. Dislocation density vs flow stress for Cu at room temperature [12-211, normalized by
h = 0.256 nm, p = 42.1 GPa (decadic logarithms). Polycrystal (PX) tensile stresses were divided by 3.06
to convert to shear stress T. Vofume dislocation densities measured by TEM were divided by 2 to
convert to intersection density p. For single slip (541). etch-pit data refer to forest density, TEM data to
dislocations with nonprimary Burgers vectors. From the low stresse& 0.12 MPa was subtracted [20]. The
lines show the relation u = ati&, with a = 1 and 0.5.

The relation between flow stress and dislocation The data in Fig. 1 span three orders of magnitude
density has been investigated over the widest range of in stress, almost seven in dislocation density (from
the variables, and under the widest range of experi- about lo* to 10” ~m-~). Two observations must be
mental conditions, in pure copper. A large variety of made at first sight: each individual set of data obeys
these data [12-213 has been replotted in Fig. 1 in a the square-root relation quite well; and all data
way that would make departures from the square- together are compatible with it if a total variation in e
root relation [equations (4)] most clearly visible. All of a factor 2 is allowed, i.e. a variation in dislocation
of these data refer to room-temperature deformation,t density of e factor 2 either way from the mean. It
all are expressed in terms of tes~lved shear stress would seem that such a variation between data from
and no~i~ by the same shear modulus different investigators could easily be due to experi-
Ir Z
- JG4 *(c* I - dLz and all are reported as inzer- mental scatter or differences in evaluation techniques.
section (rather than volume) densities [12]. For the Beyond this first-order evaluation, Fig I may sug-
single-slip data, p is thefbresr density in the case of gest a correlation between lower values of a and
etch-pit studies [lS$, 201; in the one reported single- higher dislocation densities, or a systematic departure
slip investigation by transmission electron microscopy from the square-root relation [21-241. We find it sig-
(TEM) [14], p is the density of dislocations with non- nificant, however, that at least two other correlations
primary Burgers vectors, as measured after neutron can be made: TEM data (solid symbols) give consist-
irradiation. In the cases of polyslip deformation, p ently higher dislocation densities than etch-pit data
refers to the ‘total’ dislocation density [l;? 13,16-193. (open symbols); and polyslip deformation (in mono- or
The lowest stresses were corrected by a constant addi- polycrystals) gives consistently higher densities than
tive ~ntribution r. of O.l2MPa, presumably due to single slip [21]. For both of these correlations, plaus-
impurities and chosen such as to produce a fit with ible explanations are at hand. We find the difference
equation (5) [20]. in measurement technique the most convincing argu-
ment, as well as the one with the best experimental
correlation: all etch-pit data fit quation (5) with
t The single most extensive investigation[ZZ] was
undertaken at T = 4.2 K. Even if the mechanisms were a = 1, all TEM data with a 1: 0.5. Note, on the other
identical to those at R.T., a conversion of the data would hand, that if the data were fit by one straight line of
require a knowledge of the temperature dependence of the slope greater than one, the deviations are systematic,
impurity contribution. Furthermore, the published data namely of a sigmoidal nature. This is also evident in
contained an unknown, specimendependent yield stress
parameter.
the plot used by Basinski [22,24].
1:For this reason, Livingston’s [12) single-slip data are In conclusion, the proportionality between the flow
reported in the reevaluation by Jackson and Basinski [lS]. stress (due to d~~~ationfdis~ation tiermtions) and
1868 MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING

0
0 2 4 6 8 IO
u//l x IO’

10 20 30 40 Fig. 4. Change in flow stress (normalized by temperature


dependent shear modulus) upon a decrease in test tempera-
Fig. 2. The rate sensitivity of the flow stress vs the flow ture from the temperatures given lo 77 K, vs tensile flow
stress, both as a function of prestrain, for silver mono- and stress at temperature, in copper polycrystals, both as a
polycrystals [27]. The data points were determined from function of strain. Data points evaluated from Bullen and
instantaneous stress jumps upon a decrease in strain rate. Hutchison [29].
The solid lines here and in the next two figures are predic-
tions from a model lo be presented below, with one adjust-
able parameter listed in parenthesis.
Figure 2 shows such a Haasen plot for silver single
crystals in single slip and polyslip, and silver polycrys-
tals [27]. The measurements refer to the maximum
the square-root of the wrest) dislocation density is well
stress change in the initial transient after a decrease in
substantiated. Mild departures in the direction of less
strain rate by a factor 10. It is seen that all data points
rapidly rising stress at higher densities are compatible
fall well on a straight line through the origin for a
with the experiments, but not clearly derivable j?om
finite range of flow stresses, but deviate systematically
them.
above a fairly well defined value of the flow stress,
which we shall later identify with the beginning of
FLOW-STRESS KINETICS stage III work hardening. (The drawn lines in Figs 2,
3 and 4 represent curves calculated on the basis of a
The rate sensitivity of the flow stress is most con-
phenomenological model that will be proposed
veniently presented on a ‘Haasen plot’ [7,11.24,25],
further below.)
namely as ra!e sensitivity j3 vs flow stress a, both Figure 3 shows similar data for the single-slip
varying with strain. If equation (4) is valid, this plot orientation over a wide range of temperatures [273.
should be a straight line through the origin:
The same trend is evident up to about half the
melting temperature. For higher temperatures, the
lines appear curved from the beginning, although they
still go through the origin. We conclude that equation
The straight-line relation follows from the assumed (4) is valid in the limit of small strains at all tempera-
independence of s on p or, in other words, from the tures, and over a finite range of strains at low tem-
product nature of equation (4) [26]. If there were peratures.
other, constant contributions a0 to the flow stress Temperature changes can be analyzed in a similar
(equation 5). the line would remain straight but would way. Taking account of the temperature dependence
no longer go through the origin [ll]. of the elastic moduli, and of the fact that temperature
changes are usually far from infinitesimal, equation (4)
leads to

(7)

where u, and 11, refer to measurements at the refer-


ence temperature. The term s,/s is the ‘Cottrell-
Stokes ratio’ [28] for the strain dependent part of the
flow stress; when it is constant (‘Cottrell-Stokes law’),
a plot of A(a/il) vs u/p should give a straight line as
the strain is varied.
Bullen and Hutchison [29] have undertaken a
series of temperature change experiments on copper
Fig. 3. As Fig. 2. only for single crystals in single slip. but polycrystals. They measured the increase in flow
for a wide range of test lcmperatures. stress (back-extrapolated through the transient) upon
MECRING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDE~NG 1869

a change to 77 K after various prestrains at various


temperatures. Figure 4 shows our reevaluation of
their data according to equation (7). The lines are
curved, the more so the higher the temperature. This
again suggests systematic departures from the Cot-
treli-Stokes law at larger strains: the earlier and more
marked the higher the temperature.
The failure of equation (4) to hold at larger strains
can only mean that the factor s decreases with
increasing strain. If this were due to a decrease in i as
a result of the increasing stress [24], the departures
should be the same at a given stress for all tempera-
tures. Figures 3 and 4 are evidence against this inter-
20 60
pretation. We conclude that the increasing departure 1luPal 4o
,from the Cottrell-Stokes law at high strains must be Fig. 5. Strain hardening of silver single crystals in single
due to a decrease in the rate-dependent jiow stress slip, for a wide range of temperatures [32]. The ordinate is
&ctor s(G, T). In the following, we shall establish a proportional to the rate of evolution of the dislocation
correlation between the departures from the Cottrell- density, which is proportional to the flow stress I in the
athermal hardening regime.
Stokes law and the increasing importance of dynamic
recovery, i.e. the decreasing strain-hardening rate.
This interpretation unifies the strain and temperature 8 1 da/de. By analogy, the dynamic recovery term is
dependencies. described as a (negative) contribution to 8 (which is
still stress dependent).
ATIIERMAL HARDENING AND Figure 5 shows strain-hardening data for silver
DYNAMIC RECOVERY single crystals in single slip, over a wide range of tem-
peratures [32]. The evaluation was done according to
Strain hardening is due to the evolution of the dis-
equation (IO), using resolved shear stresses ‘I: and
location density with strain. According to equation
shears y. It is obvious that the initial behavior is we11
(4), it is appropriate to evaluate it as [30-321
described by a straight line ~rougb the origin,? for all
0 do (a@j2 dp temperatures, confirming the existence of an athermai
-=-- (8) hardening rate Bn. Up to about half the melting tem-
de 2 &
perature, the data follow a straight line for a finite
when c( is independent of strain. The evolution rate range of flow stresses: this is ‘stage II’ hardening. The
has been described as being composed of two higher temperature data exhibit an inirial slope close
parts [3, 33, 341: to the same @,, but are curved from the beginning.
This behavior can also be described by equation (10).
dp 1
- t,N,% For polycrystals and for single crystals in polyslip
-& = 3 c the behavior is simifar. Figure 6 compares these with
The first term describes the storage rate by a mean the single-slip case for silver at room tcmpcra-
free path A; the seoond a dynamic recovery rate that ture[35]. The (100) crystal displays a well-defined
renders dislocation segments of length L, ineffective stage II, i.e. a constant slope over a significant range
as N, of them per unit volume rearrange at a rate v,. of flow stresses; for the (111) orientation and the
The distinction between a hardening and a dynamic polycrystal, this range is smaller.
recovery component of the evolution rate becomes
meaningful by virtue of the hypothesis that the first is
of a strictly geometric or statistical nature,and thus
athermal, whereas the second is profoundly influenced
by thermal activation. In particular, the mean free
path A may be assumed to be pro~rtiona1 to the
dislocation spacing l/J;, and thus to t/a; then one
may write equations (8) and (9) to read
ado
- = a*[& - 8&, T, o)] (10)
h
where all the constants have been lumped into a ‘har-
dening rate’ 8,,: an athermal’ contribution to
120 aIMPa 1
20 bo TIMPal 60
t The small intercept at positive stresses corresponds to
an initiaf part of the stress strain curve that is concave Fig. 6. As Fig. 5, for silver single crystals of different orien-
upwards [31,32]. tations and polycrystals at room temperature [353.
1870 MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING

The tem~rature dependence of the shear modulus


has been taken account of in Fig. 7, that of s has not.
The reversible flow-stress change data reviewed in the
last section show that the change of s with tempera-
ture is small compared to that of 0,. {One is
approaching the flow stress plateau in a o--T dia-
gram.) In particular, the somewhat faster change of s
at larger strains cannot account for the large rate sen-
sitivity of strain-hardening at large strains.
Finally, we have noted in previous sections that a
may decrease slightly with increasing deformation; if
this is taken into account, equation (8) was not a
proper derivative of equation (4), but should be re-
placed by
I , I u dcr (apb)’ dp
__-+~@$*
Oo VfP I6 -=
& (12)
Fig. 7. As Fig. 5 but for copper polycrystals. Data evalu-
ated from Bullen and Hutchison [293. A decrease in a with strain could thus, through both
terms on the right-hand side of equation (12), conceiv-
Copper ~ly~st~s have been investigated over a ably explain sign~~nt deviations from linear strain
wide range of temperatures 129-J and are plotted hardening [36]. For an explanation of the entire stage
according to equation (10) in Fig. 7. The continuous III, however, this would require a to change by one or
curve at 77 K exhibits stage II behavior over a signifi- more orders of magnitude, depending on temperature,
cant range of stresses. The curves at higher tempera- for the data shown in Figs 5-7; this is incompatible
tures were available on a d~nt~uous basis only, with the direct data shown in Fig, 1, and indirectly
and not below about 3% strain. The points in Fig. 7 also with the relatively mild change with strain of the
are, nevertheless, most easily interpreted by ascribing rate and temperature sensitivity (Figs 2-4).
to them an initial behavior that is identical for all of We shall claim below that the cause-and-effect re-
them, and from which deviations occur the earlier the lation is precisely the reverse: the evolution law (rather
higher the temperature. Thus, all of them can be de- than the flow stress kinetics desc&ed by a) exhibits
scribed by the superposition of an athermal hardening evidence of a new mechanism (viz. dynamic recovery)
rate @, and a temperature sensitive dynamic recovery becoming more important at larger strains and higher
component 0,. The value of S, used to plot the temperatures, and this mechanism in turn influences
straight line that characterizes the low-strain asymp- flow stress kinetics; the feed-back through the vari-
tote in Fig. 7 corresponds to &ZOOin resolved shear ation of a on the evolution rate, according, to equa-
stress and strain (using [lo] a Taylor factor of 3.06) tion (12), is small.
Summarizing the strain hardening data, we assert
that an initial athermal hardening rate can be identified COBBELATION BETWEEN
for both single crystals and polycrystals at all tempera- STBAIN-HARDENING AND
tures, even when a ‘stage II’ in the sense ofan extended FLOWmESS KINETICS
straight portion of the stress strain curve cannot. lhe Comparison of the strain hardening behavior
existence of this athermai hardening rate 9,, which is shown in Figs 5, 6 and 7, with the reversible flow-
always within a fuctor 2 of the value l(lzOO on a stress kinetics displayed in Figs 2, 3 and 4, respect-
resolved basis, allows one to define a dynamic recovery ively, suggests that there is a correlation between the
rate t?, us rhe deviation of the actual net slope of the two: the deviations from Cottrell-Stokes behavior
stress strain curvefrom 0,. The temperature and strain- occur at about the same stress at which strain harden-
rate sensitivity of strain hardening reside in the dynamic ing deviates from being athermal, i.e. at the beginning
recovery term. of ‘stage III’. Two separate points are worth making
A number of small corrections must now be applied by way of abstracting this observation:
to the somewhat simplifi~ exposition given above.
Firstly, it bears repeating that the temperature depen- 1. There is indeed a ‘behavior’ that obeys all the
dence of the elastic modulus superimposes on any due simplest rules to a good approximation: hardening is
to dynamic recovery. Secondly, there is a trivial athermal [equation (lb)] or even constant (‘stage II’),
strain-rate and temperature dependence through the and the strain-dependent flow stress contribution con-
term s@, T) that enters into a in equation (10) [see sists of a product of a rate-sensitive term and a strain-
equation (4)]. Thus, one may write the strain-harden- dependent term. We will refer to this case as Cottrell-
ing rate as Stokes behavior’. It spans a finite regime of (low)
strains at low temperatures, but still describes asymp-
totic initial behavior at higher temperatures. This is
true for single crystals and polycrystals.
MECKING AND KOCKS: KINETlCS OF FLOW AND STRAiN-HARDENING 1871

The dependence of the initial hardening rate O* on


the new strain rate 4 is shown in Fig. 9 for aluminum
single crystals that were prestrained at T = 7J2 to a
flow stress of 10 MPa at a fixed g, [38,40]. it is seen
that decreases in strain rate lead to negative initial
hardening rates: the well-known phenomenon of
work softening [41]. The curve shown in Fig. 9 can be
described by
Q* = @&- r.i- tin (13)
where c is a constant that should depend on the state,
and n = 5.
The inset in Fig. 9 shows that the change in strain-
Fig. 8. Left: schematic of effect of strain-rate changes on hardening rate upon a change in strain rate is larger
stress-strain curve. Right: quantitative evaluation of than the d&Grence between the 8 values attained after
instantaneous flow stress T and strain-hardening rate 8 as continuous tests at the two strain rates from the
a function of strain rate +. Data on aluminium at half the
melting temperature [38). annealed state to the same flow stress. Had the strain
rate been decreased, the initial value 8’ can even be
negative, whereas the strain-hardening rate measured
in ~ntinuous tests is always positive.
2. Deviations from this simple behavior occur Thus, there is a transient in strain-hardening behav-
increasingly as dynamic recovery becomes more sig- ior upon a change in conditions. It lasts for typically
nificant: to the degree that strain hardening becomes about 3-5% strain. This transient constitutes evidence
rate sensitive, the Cottrell-Stokes law breaks down. for the necessity of introducing an additional struc-
Strain hardening is the more rate sensitive the lower it ture parameter [2,9]. If the flow stress characterized
is; by extrapolation, this regime includes steady-state the state completely, then 0 should be the same at the
flow [34,37-J. same u for the same current test conditions of 4 and T.
We will now demonstrate the correlation between This is not the case; however, after completion of the
strain-hardening and flow-stress kinetics on one transient, it becomes true. Thus, the additional state
further set of experiments, in which the effects of parameter must reach some steady-state value for the
strain-rate changes on both the flow stress and the given conditions, after a few percent strain.
hardening rate are measured. They also manifest the The ‘long transient’ discussed here is different in
need to introduce an additional parameter to charac- kind from the short ‘inverted’ transient that usually
terize the structure or state in this regime. extends for only a fraction of a percent strain[2].
The dependence of a property on strain rate and The long transient is observed only in stage III of
temperature is best tested at constant structure or work hardening, when dynamic recovery becomes
state. This concept was introduced by Cottmll[28] increasingly important. In the athermal hardening
for the flow stress (and used earlier in this paper); it regime, the strain-hardening rate is the same before
holds as well for the rate of evolution [9]. At a given and after a strain-rate change.
state (for example, after identical histories), the initial This completes our exposition of experimental
strain-hardening rate in new tests under various con- facts. We con&de that with increasing importance of
ditions may depend on the then current strain rate
and temperature just as the flow stress does. Unfortu-
nately, strain-hardening data have usually not been
reported upon abrupt changes in strain rate or tem-
perature, and they are not available for the two
materials discussed until now.
Figure 8 illustrates data for aluminum single crys-
tals deformed in single slip at half the melting tem-
perature [38] (similar data have been obtained for
~lycrys~lline aluminum alloys [39]). After changes
in strain rate of different sign and magnitude, both the
flow stress and the initial strain-hardening rate
change, and they appear to do so in a correlated fash-
ion. Increases in strain rate lead to increases in both
flow stress and strain hardening, at a decreasing rate 1/y Ilhci
as the magnitude of the change gets larger. Decreases
Fig. 9. The initial strain-hardening rate 8* after strain-rate
in strain rate lead to decreases-larger in magni-
changes away from jr. which was used in prcstraining to
tude-of both the flow stress and the initial hardening 10 MPa [38,40]. Inset: a transient of about 3% duration
rate. follows upon a strain-rate change in stage III.
1872 MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING

dynamic recovery, the vurious simple ussumptions that where the as yet arbitrary functionfmust go to 1 as 0,
were found to hold true during the initial ‘CottrelL goes to zero, and must decrease as 0, increases. We
Strokes behavior’ break down simultaneously; equa- shall use two easy heuristic representation of such a
tions (2a) to (2~) are needed as a general foundation. function [inserting equation (15) for the general glide
kinetics term .sA]:

A PHENOMENOLOGICAL MODEL
s = (i)“m-exp( -F- $) (17a)
Glide kinetics is generally described by an Arrhe-
nius equation [2,7]: and

+_)lh.(l-F*~) (17W
(14)

where AG(s,,) is meant to express the stress dependence where f3, has been introduced merely for normaliz-
of the activation free enthalpy. Inversion of equation ation, and F is an adjustable parameter (not necess-
(14) gives a special form, when equation (14) holds, for arily constant, but assumed so in the following); since
the stress factor s introduced in equation (4); namely it always comes out small compared to 1, the two
* 1/n expressions (17a) and (17b) are in fact equivalent, and
a
SE -=s,,(i,T)= ; . (13 will be used interchangeably, depending only on con-
b 0 venience.
In the last part of the equation, we have expressed sA Equation (17a) was originally derived on the basis
as a power-law in the strain rate, which is often used of a physical mechanism, which has not as yet been
for convenience in phenomenological expressions. It is published in detail, but was summarized in Ref. [ll].
truly equivalent to an inversion of equation (14) only It is based on the contribution made to glide kinetics
when by the ‘dislocation rearrangement strain rate’ during
dynamic recovery, and also provides a structural
interpretation for a second state parameter [ll, 421.
In the present paper, we shall merely use equations
(17) as empirical relations and prove that they de-
is independent of stress, as it often is to a sufficient scribe the data presented here quite well.
approximation. The temperature dependence of m at Equation (17a) leads to the following expression for
constant strain rate [which is more complicated than the rate sensitivity:
the explicit temperature dependence evident in equa-
tion (15’)] causes the temperature dependence of s and alns 1 F ae,
- =---_
thus of the flow stress in the power-law description aIn* T m 0, a In ( T
(15) c7l.
Equation (15) is a specific example of Cottrell- and using equation (13) for the instantaneous rate
Stokes behavior [equation (4)]: the flow stress a is a sensitivity of the hardening rate:
product [26] of 6, which depends only on strain, and
s, which depends only on strain rate and temperature.
This behavior would be violated if, for example, &
were not constant but depended on i. A mild depen- The equation was plotted as a solid line through
dence of this kind is actually likely [7]; however, it the data points in Figs. 2 and 3, using e,le, as evalu-
would go in the wrong direction to explain the ob- ated from Figs. 5 and 6, respectivelyt. We found that
served deviations from Cottrell-Stokes behavior at a value of m/n *F = 1 fit all the room-temperature
large flow stresses-and could not explain the corre- curves in Fig. 2; for higher temperatures, the value is
lation with dynamic recovery and the appearance of given in parenthesis on the respective curves in Fig. 3.
transients. (The temperature dependence is consistent with that
In order to describe and to test the postulated con- of m/n, and thus with F 2 constant.) The fit is
nection between deviations from Cottrell-Stokes kin- remarkable. Note in particular that the very different
etics and dynamic recovery, let us instead replace stress strain curves for the different single crystal
equation (15) by orientations and the polycrystal of silver give very
similar flow stress kinetics.
Temperature changes should similarly give, instead
of equation (7), a flow stress change according to
t For the strain-hardening rate, we have used the
‘steady-state’value at the base strain rate, 0; nevertheless.for
its rate sensitivity. we have used the instantaneous depen-
.$ (20)
dency given by equation (13). Moreover, we have assumed
this equation, which had been derived for AI single crystals
at half the melting temperature. to hold, in its functional where the linearized equation (17b) has been used,
form, for silver mono- and polycrystals at all temperatures. and S is the initial CottrelCStokes ratio at the two
MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING 1873

temperatures (sA,/sA).The copper data in Fig. 4 were proportional to the dislocation segment length [36]. If
fit using equation (20) with A@, = -Q,(T), and the inner cut-off radius is called rO, this gives
evaluating &/f1,, for each temperature from Fig. 7
(i = a&,:p In(l/rOv,‘p). (22)
(employing a computer smoothing technique). A
single value F = l/30 produced the good fit in Fig. 4. The elfective value of r = a0 In(l/r,,/p) then de-
Equations (17) thus describes these data very well. creases with increasing dislocation density. On the
Note that the correction factor for the _Howstress is other hand, it is assumed that the rate sensitivity is
always near 1 (F < 0.1, 0,/6’, < 1); nevertheless, the not proportional to the line tension, but
influence on !he rate and temperature seusitiuity is
fi = B-/lb J{) (23)
profound.
A further ramification of the new flow stress law with B independent of p. Then, the relationship
expressed by equation (16). as inserted into equation between flow stress and rate sensitivity can be
(4). is that an Arrhenius equation no longer describes expressed by
flow kinetics when the rate of dynamic recovery is
d = Cp*ln(D/Cp) (24)
substantial. The specific form given in equation (17) is
such that the observed flow stress should be lower where C z so/B and D = a&/r,,. Equation (24) de-
than that predicted by an Arrhenius equation [equa- scribes a curve, not a straight line, in a Haasen plot
tions (14) and (IS)] when the strain rate is low or the and is used to explain deviation from Cottrell-Stokes
temperature high. This is exactly what is observed in behavior.
the high-temperature flow-stress drop [2,5-J, and dur- The above description was based on a model [24]
ing long-term stress relaxation [43,44]. In all these in which the flow stress is composed of two parts: an
cases, the curvature in a plot of u vs T or In.5 has the athermal component described by equation (22), and
wrong sign to be consistent with thermally activated a thermally activated component which was assumed
glide over a physically reasonable obstacle in which to be so small compared to the athermal one that it
the activation area increases as the stress de- was not even included in equation (22) as an additive
creases [44]. Figure 8 of the present paper displayed term [45], even though it is solely responsible for
another case of such an unexpected curvature, but equation (23). The crucial assumption is that the two
here for the flow stress after a strain-rate change. The components have different p-dependencies. In our
temperature was there not quite in the flow-stress- opinion, this represents a superposition rule which is
drop regime, but the strain was large enough to pro- hard to justify theoretically and certainly refers to a
duce substantial dynamic recovery. The critical cri- very special obstacle distribution. In particular, it
terion for the departures from classical behavior is a appears to be in conflict with basic features of Cot-
low strain-hardening rate. trell-Stokes behavior.
The Cottrell-Stokes law is generally interpreted as
experimental evidence for the athermal and thermal
BASINSKIS MODEL stress components (if indeed they have a separate
existence) to be controlled by the same obstacles,
In the preceding we have presented a point of view namely forest dislocations [46]. The line tension con-
that emphasizes the correlation between the observed trols the area enclosed by dislocation segments at the
breakdown in Cottrell-Stokes behavior with the critical bow-out, and thus the statistics of the effective
breakdown in athermal strain hardening. The effect is obstacle spacing along the gliding dislocations.
mainly ascribed to an increase in the rate sensitivity Whether and in which form it appears in the macro-
of a in the u - ,/p relationship in equation (4) as the scopic flow stress relation depends on the obstacle
dynamic recovery component 0, of strain hardening strength and on the detailed averaging procedure, and
increases. If one expresses the rate sensitivity as
cannot be stated in a general way [7,47]. In any case,
however, it should enter both stress components in
precisely the same way when they are controlled by
the same obstacles.
then @a/a In i) is claimed to be a function of 0,. There If nevertheless the distinction between the p-depen-
is also a small decrease in the absolute magnitude of a dence of the two stress components were accepted on
which, however, is negligible in the proposed formal- a phenomenological basis, it could indeed explain the
ism. sign of the observed deviations from Cottrell-Stokes
Just the opposite point of view has been taken by behavior; to explain the observed magnitude, however,
Basinski [22,24] in a previous attempt to explain the the inner cut-off radius would have to have a value of
deviations from Cottrell-Stokes behavior at large rO z 15 b for the data reported by Basinski [24] (and
strains. It is based on the assumption that the absolute also for our evaluations at room-temperature, as
value of a in equation (4) decreases with increasing reported below), which is not easy to rationalize.
strain. This follows from the postulate that the flow Furthermore, the strong a@) dependence implied by
stress is proportional to the dislocation line tension, this value tends to overexplain the possible mild
and that the outer cut-off radius in the line tension is departure from the Q a fi proportionality shown in
1874 MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING

Fig. 10. Data from Fig. 3, but with curves derived according to Basinski’s model [23]. The cut-off radius
ror which determines the curvature, was matched to the room-temperature data, the other adjustable
parameter was matched to the low-stress regime for 7’ = 672,783 and 986 K, and is given in parenthesis.

Fig. 1. (See also the detailed discussion on the data in SUMMARY


Fig. 1.)
We assert that the experimental data for single
Finally, and for the present paper most signifi-
crystals of various orientations and for polycrystals
cantly, Basinski’s model is based on a modification
of pure f.c.c. metals at temperatures up to at least
of the athermal part of the flow stress in equation (22).
half the, melting point have the following cardinal
Therefore, in the framework of his approach, a de-
features:
scription of p(a) plots for different temperatures
should be possible according to equation (24) with a 1. An athermal hardening rate exists, which domin-
constant value of D (and thus of r,,), and an adjustable ates at low strains and low temperatures (‘stage II’),
value of C [and thus of B in equation (23)] for each but describes initial asymptotic behavior at all tem-
temperature. This procedure was applied to the data peratures. Departures from this behavior at larger
in Fig. 3, and the result is shown in Fig 10. The value strains (‘stage III’) are strongly rate and temperature
of D = 34OMPa was chosen such as to give the best sensitive.
fit for the room-temperature data; it is equivalent to 2 The Cottrell-Stokes law is obeyed (excluding the
Basinski’s value for r. = 15 b, when one assumes easy-glide regime from our considerations) over pre-
a0 = 0.2, which is consistent with the data in Fig. 1. cisely the same range, or in the same limit, in which
Now the values for C were determined such as to strain hardening is athermal.
match the low-stress (Cottrell-Stokes) regime for a 3. The Arrhenius equation, which gives an ad-
series of temperatures for which the /&r)-data were equate description of flow kinetics in the athermal
well documented (Fig. 3), namely 672,783 and 986 K; hardening regime, also breaks down when the rate of
these C-values are given in parenthesis on each curve. dynamic recovery becomes substantial.
As can be seen, the data points deviate considerably 4. The instantaneous response of the strain-harden-
from the calculated curves. ing rate to strain-rate changes is correlated with that
A good fit of the data by equation (24) could, in of the flow stress, in the dynamic recovery regime.
fact, be obtained-but only by allowing the quantity The classical decoupling of ‘reversible’ flow stress kin-
D. i.e. the inner cut-off radius, to vary with tempera- etics ‘at constant structure’ from the evolution kin-
ture. For the data at 783 K, for example, this required etics is no longer possible in this regime.
(for a0 = constant) an increase in r. by a factor of 20 5. A nontrivial dependence of flow stress kinetics
compared to room temperature, i.e. an increase from on evolution kinetics demands the introduction of a
r. = 1Sh to 3OOh. second structure/state parameter; this is in agreement
We conclude that Basinski’s model, even if it could with the appearance of an additional transient in just
he theoretically founded in a more satisfactory way, the same regime.
cannot explain the large-strain departures from Cot- 6. The proportionality between the flow stress and
treh-Stokes behavior at higher temperatures. In light the square-root of the dislocation density is well
of this rather limited success, it should not be used as documented for copper at room temperature. Mild
independent support for systematic departures from deviations are probably chiefly due to systematic dif-
the D % ,/{J proportionality either. ferences between etch-pit and TEM results.
MECKING AND KOCKS: KINETICS OF FLOW AND STRAIN-HARDENING 1875

These observations were quantitatively well de- 19. E. Giittlcr, Phil. Mug. 28, 1057 (1973).
scribed on the basis of a phenomenological model, in 20. H. Mecking and G. Bulian, Acru ntcrull. 24, 249 (1976).
which the classical equation for glide kmetics is modi- 21. P. Ambrosi, W. Homeier and Ch. Schwink, Scriptu
metull. 14, 325 (1980).
ficd by a factor that decreases from I as the rate of 22. Z. S. Basinski and S. J. Basinski, Phil. Muu. 9. 51
dynamic recovery increases. Thus, all the observed ( 1964): SW also in Dis/ocurbns irl So/ic/,s, (edited by
dcparturcs from classical behavior are linked in thei F. R. N. Nabarro) vol. 4. p. 261. North Holland,
Importance to the smallness of the strain-hardening Amsterdam (19791.
23. G. Schocck and R. Frydman, Phy.\icu SI~WS solidi 53,
rate whether this be at large strains or at high tem- 661 (1972).
peratures. 24. Z. S. Basinski. Srriptu meta// 8, 1301 (I 974).
25. P. Haasen. Phil. Mug. 3. 384 (1958).
A~~~rto~l~/c~/~/cn?c,r1rs-PProressor
K. Liicke contributed to the 26. R. A. Mulford and U. F. Kocks, Acru merull. 27, 1125
formulation of the concepts developed here and provided (1979).
unfaihng encouragement. We gratefully acknowledge the 27. H. Mecking. Huhiliturion rhrsis, RWTH Aachen (1973).
support of the U.S. Department of Energy, of the Deutsche 28. A. H. Cottrell and R.-J. Stokes, Proc. R. Sot. A 233, 17
Forschungsgemeinschaft (H.M.), and of the Alexander-von- (1955).
Humboldt-Stiftung (U.F.K.). 29. F. P. Bullen and M. M. Hutchison, Phil. Mog. 7, 557
( 1962).
30. F. R. N. Nabarro, Z. S. Basinski and D. B. Holt, Adr.
Phys. 13, 193 (I 964).
31. H. Mecking and K. Liicke, Z. Metal/k. 60, 185 (1969).
REFERENCES
32. H. Mecking. in Work Hardenina in Tension und Fati-
1. H. Meckmg and K. Liicke, Mater. Sci. Engng I, 349 gue (editedby A. W. Thompsonip. 67. A.I.M.E. (1977).
(1967). 33. K. Liicke and H. Mecking, in The Inhsmogeneity of
2. U. F. Kocks and H. Mecking, in Dislocation Modelling PIustic Deformation (edited by R. E. Reed-Hill) p. 223.
of Physical Processes (edited by C. S. Hartley et a/.) Am. Sot. Metals. (1973).
Pergamon, in press. 34. U. F. Kocks. J. Engng. Mater. Tech. (ASME) 98, 76
3. H. Mecking and K. Liicke, Acta metal/. 17, 279 (1969). (1976).
4. U. F. Kocks, in Fundamental Aspects of Dislocation 35. M. Biermann, Doctoral Thesis RWTH Aachen, W.
Theory, (edited by J. A. Simmons, R. deWit, R. Bul- Germany.
lough) (NBS Spec. Pub. 317). p. 1077 (1971). 36. D. Kuhlmann-Wilsdorf, in Work Hardening, (edited by
5. T. H. Alden, Phi/. Mag. 25, 785 (1972). J. P. Hirth and J. Weertman) p. 97. Gordon and
6. A. P. L. Turner, Metal/. Trans. A 10, 225 (1979). Breach (1968).
7. U. F. Kocks. A. S. Argon and M. F. Ashbv. _. Prop.
I 37. H. Mecking, in Dislocation Modelling of Physical Pro-
Mater. Sci. 19, (1975). - cesses, (edited by C. S. Hartley et al.) Pergamon,
8. U. F. Kocks. in Constitutiue Eauations in Plosticirv. Oxford (in press).
(edited by A.S. Argon) p. 81. MI? Press (1975). ” 38. H. Fischer, Doctoral Thesis RWTH Aachen, W. Ger-
9. U. F. Kocks, Chalmers Anniwrsary Volume, Prog. many.
Muter. Sci., p. 185 (1981). 39. H. Here, Proc. /Oth Biannial Gong. Ind. Deep Drawing
IO. H. Mecking, in Strength of Metals and A//oys, (edited Research Group, p. 179, Warwick, U.K. (1978).
by P. Haasen, V. Gerold and G. Kostorz) p. 1573. 40. H. Mecking, U. F. Kocks and H. Fischer, Proc. 4th Inc.
Pergamon (1979). Con/ on the Strength of Metals and Alloys Nancy,
11. U. F. Kocks, in Strength of Metals and Alloys, (edited France, p. 334 (1976).
bv P. Haasen. V. Gerald and G. Kostorz) I ,I). 1661. 41. R. J. Stokes and A. H. Cottrell, Acta metal/. 2, 341
Pergamon, Oxford (1979). (1954).
12. J. D. Livingston, Acta meta//. 10, 229 (1962). 42. U. F. Kocks and H. Mecking, in Strength ofMetals and
13. J. E. Bailey, Phi/. Mag. 8, 223 (1963). Alloys, (edited by P. Haasen, V. Gerold, and G. Kos-
14. U. Essmann. Phvsica status solidi 17. 725 (1966). torz) Pergamon. p. 345 (1979).
15. P. J. Jackson anh Z. S. Basinski, Can J. Whys. 45, 707 43. E. W. Hart and H. D. Solomon, Acta metal/. 21, 295
(1967). (1973).
16. U. Essmann. M. Rapp and M. Wilkens, Acta metal/. 44. C. C. Law and D. N. Beshers. Scrimu metal/. 6. 635
16, 1275 (1968). (1972).
17. G. van Drunen and S. Saimoto, Acta metal/. 19, 213 45. Z. S. Basinski, personal communication.
(1971). 46. Z. S. Basinski. Phil. Maa. 4. 393 11959).
18. M. R. Staker and D. L. Holt, Acta meta//. 20, 569 47. A. J. E. Foreman and I% J.‘Makin, PAi/. Mag. 14, 911
(1972). (1966).

You might also like