High strain rate behawior of metals

R J Clifton

Division of Engineering, Brown University, Providence RI 02912

Experimental results on the high strain rate response of polycrystalline

metals are reviewed, with emphasis on the behavior of pure metals. A
strong increase in flow stress with increasing strain rate is reported for
strain rates of approximately 10 5 s _ 1 and higher. This increase is
observed in pressure-shear plate impact experiments at nominally
constant strain rates from 10 5 s _ 1 to 10 6 s _ 1 . To improve understanding of
the increased rate sensitivity at high strain rates, pressure-shear,
strain-rate-change experiments have been conducted on O F H C copper
specimens. These experiments have been analyzed using a conventional
viscoplasticity formulation and an internal variable formulation in which
the hardening rate depends on the rate of deformation. Only the latter
formulation is successful in describing the observed response to the
change in strain rate. This observation is discussed in terms of its
implications for interpreting other dynamic plasticity experiments and
for improved understanding of the underlying dislocation mechanisms.
The enhanced rate sensitivity at high strain rates is concluded to be
related primarily to the rate sensitivity of strain hardening, not the rate
sensitivity of the flow stress at constant structure.

the loading and monitor the nominal stress and strain

1. INTRODUCTION
in the specimen. With this approach, dynamic stress-
strain curves, at constant strain rates, have been obtained
Plastic flow of metals at high strain rates, studied ex- for many metals at strain rates of 10 2 s" 1 to 10 3 s _ 1 (See
tensively for nearly fifty years, has recently attracted re- Lindholm (1964,1971) and Green ei al (1970)). These
newed interest because of the discovery of strong effects results show that, at a given strain, the flow stress in-
at very high strain rates and because of a growing accep- creases with increasing strain rates at the rate of a few
tance of the importance of these effects in dynamic failure percent per decade of strain rate. Larger increases are
mechanisms. Early interest in the subject was stimulated obtained for high purity metals; smaller increases are
by military applications, especially armor penetration. obtained for high strength structural alloys. The basic
While these applications continue to spur research, the concepts of the Kolsky approach have been extended to
need to understand dynamic plastic flow has been recog- torsional loading by Duffy, Campbell and Hawley (1971)
nized as being of critical importance in a broad range of and to tensile loading by Harding, Wood and Campbell
applications, including high-rate forming, high speed ma- (1960), Hauser (1966), Lindholm and Yeakley (1968), and
chining, explosive welding, and a variety of safety-related Nicholas (1980), with no fundamental change in the prin-
applications ranging from the crash-worthiness of vehicles cipal observations. Two modifications of the Kolsky ap-
to the retention of flying pieces of broken turbine blades. proach have resulted in important new insights. First,
From the perspective of failure mechanisms, plastic flow when the specimen dimensions are reduced, and the am-
at high strain rates plays an important role in dynamic plitude of the incident stress pulse is raised (bars of higher
ductile rupture and shear strain localization because of yield strength are used to prevent yielding) to achieve
the high strain rates that occur near the crack tip and in higher strain rates (10 4 s~ ! ), increased strain rate sensi-
the shear band, respectively. tivity is observed (See, e.g., Stelly and Dormeval (1977)).
Research on high rate deformation of materials has Second, when the specimen is subjected initially to pre-
been directed primarily towards the development of suit- straining at a lower strain rate and the strain-rate is in-
able constitutive models for describing the plastic behav- creased sharply in so-called "strain-rate jump" tests of
ior of materials in the high strain rate regime. Several the type introduced by Frantz and Duffy (1972), the
approaches have been followed. One, an experimental ap- stress-strain curve following the increase in strain rate
proach pioneered by Kolsky (1949), is to measure the re- lies below the stress-strain curve for a constant strain-
sponse of materials under nominally homogeneous states rate test at the higher strain rate. The latter tests have
of stress by placing a short specimen between two long been interpreted as indicating the lack of a "mechanical
elastic bars and using elastic waves in the bars to impose equation of state" relating stress, strain, and strain rate

&
Appl Mech Rev vol 43, no 5, Part 2, May 1990 S9 Copyright 1990 American Society of Mechanical Engineers

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S10 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

during plastic flow. T h e former tests were interpreted three-dimensional effects preclude a reliable interpreta-
initially as indicating t h a t at the higher strain rates the tion of the experiments within the framework of a one-
rate controlling mechanism of plastic flow changes from dimensional wave theory. These difficulties are overcome
the thermally activated motion of dislocations past obsta- in plate impact experiments in which plane waves are gen-
cles to the thermal-phonon resisted motion of dislocations erated by the impact of flat, parallel plates. Before wave
through the clear lattice. T h e experimental techniques reflections arrive from lateral boundaries, the wave prop-
and the principal experimental results for experiments at agation through the thickness is one-dimensional. Conse-
strain rates up to 1 0 4 s _ 1 have been reviewed by Hartley quently, one-dimensional wave theory can be used for nu-
and Duffy (1985) and by Nicholas (1982), who provides merical predictions of wave profiles to be compared with
an extensive list of references. those recorded in the experiments. Further simplifica-
tion of the analysis is obtained for cases in which the
Concurrently, experiments were developed for deter-
main platic wave propagates as a steady wave or only
mining quite directly the mobility of dislocations in crys-
the decay of the elastic precursor is required. Calcula-
tals at different stresses and temperatures. Johnston and
tion of steady wave profiles for an assumed constitutive
Gilman (1959) introduced the so-called "direct disloca-
model (See Johnson and Barker (1969)) requires only the
tion velocity measurement" approach in which specimens
solution of ordinary differential equations which, for el-
are subjected to a short-duration stress pulse and the dis-
ementary constitutive models, may be solved in closed
location velocity is estimated from the duration of the
form. Also, if the amplitude of the elastic precursor is
pulse and measurements of the dislocation positions be-
sufficiently small t h a t nonlinear elastic compressibility ef-
fore and after the pulse loading. This approach was ex-
fects are negligible, then the analysis of the elastic pre-
tended to torsional loading by Pope, Vreeland and Wood
cursor reduces to the solution of an ordinary differential
(1967) to obtain more well defined stress states and to
equation which, for elementary constitutive models (See
plate impact loading by Kumar and Clifton (1979) to ob-
Taylor (1965) and Johnson, Jones and Michaels (1970)),
tain higher dislocation velocities. Such experiments have
may also be solved in closed form. Interpretation of elas-
shown t h a t the intrinsic resistance of the lattice to the
tic precursor decay experiments has proved to be diffi-
motion of dislocations increases essentially proportionally
cult because steep gradients in the wave profile make it
with increasing dislocation velocity according to
difficult to measure the amplitude of the precursor accu-
rb = Bvd) (1) rately and because the amplitude is sensitive to disloca-
tion generation and multiplication at, and immediately
where r is the resolved shear stress, b is the Burgers vec- behind, the wavefront (See, e.g. Herrmann, Hicks and
tor of the dislocation, v^ is the average dislocation ve- Young (1970) and Meir and Clifton (1986)). Steady wave
locity, and B is the drag coefficient. Values of the drag profiles have proved to be more useful for examining the
coefficient B obtained from experiments on high purity plastic response of materials at very high strain rates,
crystals are found to be so small that dislocation veloc- say 1 0 7 s - 1 . An important conclusion from these exper-
ities approaching elastic wave speeds would be achieved iments is t h a t the flow stress increases strongly but less
at stresses which are less than, or comparable to, val- than proportionally, with strain rate at these high strain
ues obtained in other high strain rate experiments, such rates. Grady (1981) reports the shear stress increasing as
as Kolsky bar experiments. (For example, from the value the square root of the strain rate. One limitation of the
B = 1.7 x 10~ 5 Pa s~ reported for high purity copper by normal impact experiments is t h a t plastic strains of only
Jassby and Vreeland (1970), the shear stress r c at which a few percent are generated because the high strain rates
the dislocation velocity in (1) is equal to the shear wave are maintained for very short times.
speed is r c « 150MPa.) If the intrinsic resistance of the
clear lattice were the primary resistance to the motion of Larger plastic strains have been obtained by impacting
dislocations at the stress levels of high strain rate exper- plates which are parallel, but skewed relative to the di-
iments, then a sizable fraction of the dislocations would rection of approach to produce both pressure and shear
be moving at dislocation velocities of the order of elastic tractions on the impact plane. Such pressure-shear ex-
wave speeds and the resulting plastic strain rate would periments, introduced by Abou-Sayed and Clifton (1977),
be "much larger than observed in the experiments. Thus, have the attractive features that the resulting shear waves
it appears t h a t even at high strain rates the mobility of are both sensitive to the shearing resistance of the ma-
dislocations is strongly affected by the interaction of dis- terial and relatively easy to interpret because they are
locations with such obstacles as impurity atoms, second exactly one dimensional, as long as the impacting sur-
phase particles, other dislocations, and grain boundaries. faces are flat and parallel. Additionally, these experi-
T h e dynamic plastic response of materials has also ments show the effects of non-proportional loading paths
been investigated by stress wave experiments in which which arise because differences between the wave speeds
large amplitude waves are propagated in bars, tubes or for longitudinal and shear waves cause the normal stresses
plates. Comparisons of predicted and measured wave at any point to change before the shear stress. However,
profiles have been used to assess the validity of vari- unlike normal impact experiments in which steady waves
ous constitutive models and suggest improved models. are obtained when the wave-steepening effects of nonlin-
Experiments on longitudinal waves in bars, while the ear elastic compressibility are exactly balanced by the
easiest to conduct, have been difficult to interpret un- wave spreading effects of plastic flow, shear waves lack
ambiguously because the high strain rate region of pri- a wave-steepening effect and, therefore, tend to spread
mary interest is so near the impact end of the bar t h a t with distance of propagation (See Chhabildas, Sutherland

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S11

and Asay (1979), Kim and Clifton (1980), and Gilat and the experimental technique, as well as some of the early
Clifton (1985)). Correspondingly, the plastic shear strain results, have been given by Clifton and Klopp (1985).
rate decreases with distance of propagation. Li (1982), Without repeating the details, measurement of the skew
retaining the pressure-shear impact configuration, over- angle 9, the projectile velocity, V, and the transverse ve-
came the tendency for the shear strain rate to decrease locity, vjs(t), of the free surface of the target plate allows
by sandwiching a thin specimen between two hard elas- the nominal shear strain rate, j , and the nominal shear
tic plates in a manner t h a t is analogous to the method stress, r , to be computed from
introduced by Kolsky except for the replacement of bars
by plates. Because the wave propagation is one dimen-
sional, except for perturbations due to heterogeneity of
the specimen, there are essentially no end effects. There-
fore, large strain rates (10 5 s~ 1 to 1 0 7 s - 1 ) can be obtained T =
2^pC^sVfs' ^
by reducing the thickness of the specimen to thicknesses
of, say, 250//m to 2.5/zm. In this way, the flow stresses where t>o = VsinO is the transverse component of the pro-
of pure metals (Al, Cu, Fe) have been shown to increase jectile velocity, h is the specimen thickness, and (pc-^s is
strongly with increasing strain rate at strain rates of ap- the elastic shear wave impedance of the steel target. In-
proximately 1 0 5 s - 1 and higher (See Klopp, Clifton and tegration of the shear strain rate provides a measure of
Shawki (1985) and Huang and Clifton (1985)). the shearing deformation which, along with the computed
values for r , can be used to obtain a dynamic stress-strain
Discussion of this strong increase in flow stress with curve for the specimen material. After a few reverbera-
strain rate at very high strain rates is the principal thrust tions of waves through the thickness of the specimen, the
of this paper. Does this strong rate sensitivity of the flow stress in the specimen becomes nominally homogeneous
stress indicate t h a t the material response is much like and the "stress-strain curve" becomes representative of
that of a viscous fluid for which the flow stress depends the response for homogeneous deformation at approxi-
strongly on the current strain rate or, instead, does the mately constant strain rate. Such dynamic stress-strain
response remain like t h a t of a metal for which the flow curves are shown in Fig. 2 for annealed, commercially
stress at a given dislocation structure, depends weakly pure, aluminum. Steeply rising portions of the curves
on the current strain rate, while the structure depends correspond to intervals in which the stress state in the
strongly on the strain-rate history? First, results of con- specimen is not uniform because of insufficient time for
stant strain-rate tests will be summarized. Then, results wave traversals through the thickness of the specimen.
of strain-rate change tests will be presented. Interpreta- An indication of the variation of the rate sensitivity of the
tion of strain-rate change tests will be shown to be en- flow stress of aluminum over a wide range of strain rates is
hanced if the test is analyzed by considering the propa- shown in Fig. 3. D a t a points in Fig. 3 are obtained from
gation of the stress increment through the specimen. For experiments at nominally constant strain rate; at lower
strain-rate decrement tests on O F H C copper the anal- strain rates the curves correspond to fixed strain levels
ysis of this unloading wave will be shown to indicate a whereas at higher strain rates the flow stress becomes in-
weak dependence of the sensitivity of the flow stress on dependent of the accumulated strain (See Fig. 2) and
the current strain rate. This result will be examined fur- the d a t a points indicate the plateau stress reached in the
ther by considering the implications of this behavior for pressure-shear experiments. Such plots indicate t h a t the
the interpretation of other impact experiments used to increase in flow stress per decade increase in strain rate
characterize the plastic response of metals at high strain is much greater at strain rates of 10 5 s~ 1 and higher than
rates. at strain rates of 10 3 s~ x and lower. Similar behavior has
been reported by Klopp, Clifton and Shawki (1985) for
high purity iron and by Huang and Clifton (1985) for
2, EXPERIMENTS O F H C copper.

To gain better understanding of the origins of the strain

The pressure-shear plate impact configuration shown in
rate sensitivity at high strain rates, strain-rate change ex-
Fig. 1 has been used to study the plastic flow of metals at
periments have been conducted on O F H C copper in the
strain rates of 1 0 5 s _ 1 to 1 0 7 s - 1 . A detailed description of
pressure-shear plate impact configuration as a means for
probing the response of the material at different levels of
OfHC Copper Specimen stress and strain rate, but at nominally the same dislo-
Hard S t e e l / Hard Steel cation structure. T h e experimental configuration is the
same as that shown in Fig. 1 except that the hard steel
flyer plate is replaced by a two-plate assembly consisting
of a hard steel plate backed by a high strength aluminum
alloy (7075-T6). T h e t-X diagram for this experiment
is shown in Fig. 4. T h e specimen is subjected initially
to the high shearing rate imposed by the bounding steel
Fiberglass Tube Flyer Plate Target Plate
plates as shown in Fig. 5 by the points 2 and 3 repre-
senting the stress and particle velocity at the two faces
FIG 1. Schematic of pressure-shear plate impact configura- of the specimen. When shear waves reflected from the
tion. low impedance aluminum plate reach the specimen, the

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S12 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

150.0 y ° 2 . Q 5 x l 0 3 s " ' l S H 0 T 81-10

Second Window
100.0
w
w
CD
4—'

co
CO
<D 50.0 First Observed Signal
-C

co

OFHC Copper
0.0
0.00 0.05 0.10 0.15 FIG 4. Time-distance diagram for the pressure-shear,
Shear Strain
strain-rate-change experiment.
FIG 2. Dynamic stress-strain curves for 1100-0 aluminum (Li
(1982)). gratings. T h e o u t p u t of the T D I is monitored by a LeCroy
Waveform Digitizer which records 10,000 d a t a points at
shear stress is reduced sharply and, once a uniform state intervals of 0.742 ns. T h e recorded fringes are scaled to
of shear stress is established again in the specimen, the uniform amplitude and then reduced using a fast Fourier
shear strain rate is reduced strongly, as indicated by the transform, low-pass filtering, and multi-point differenti-
small separation between points 6 and 7 in Fig. 5. The ation to obtain the velocity-time profile. Details of the
shear strain rate can be obtained during this time (i.e. d a t a reduction procedure as well as the specimen prepa-
during the second window of Fig. 4) by replacing the ration procedures are given in a forthcoming paper by
numerator in (2) by (vjsi — vj^) as shown in Fig. 5. Tong, Clifton and Huang (1990). For the experiments de-
With this configuration, strain rate change tests can be scribed here, the specimens are O F H C copper foils with
carried out, starting at strain rates of 1 0 5 s _ 1 or higher a thickness of 25 /zm and a grain size which is initially 5-
for l p s , followed by substantially lower strain rates for 10 fim, but which becomes approximately 40-50 /-im after
approximately l//s. diffusion bonding to the target. Although the grain size is
For the strain-rate-change experiments as well as for greater than the thickness of the specimen the measured
recent constant strain rate experiments, the experimen- response is a measure of average behavior over differently
oriented grains because the responses of many neighbor-
tal technique has been upgraded to improve the resolu-
ing grains contribute to the measured motion at the rear
tion of the transverse particle velocity. The pitch of the
surface of the target.
diffraction grating placed on the rear surface of the tar-
get to enable the measurement of the transverse displace- Figure 6 shows results from one of the strain-rate
ment by means of the transverse displacement interferom- change tests. T h e strain-rate history is given by the down-
eter (TDI) introduced by Kim, Clifton and Kumar (1977) ward sloping curve, which begins when uniform stress
has been reduced from 200 lines/mm to as many as 1000 is established in the specimen. T h e stress-time profile
lines/mm by using a holographic technique to make the shows a decrease in shear stress when the sharp decrease
in strain rate occurs. Stress-time profiles are also shown
in Fig. 6 for a viscoplastic power law model (See Table

8. H»
140.0
0-- 26.6*
8" 33.7*
Shear Stres ; (MPa)

120.0

100.0

80.0

60.0

40.0

20.0
1.0 2.0 3.0 4.0 5.0
Shear Strain Rate (1/s, Log)
FIG 5. Stress-velocity diagram for the shear stress and trans-
FIG 3. Rate sensitivity of the flow stress in 1100-0 aluminum verse particle velocity in the pressure-shear, strain-rate-change
(Li(1982)). experiment.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S13

TABLE I. Model parameters for OFHC copper

Elastic: p c 2 2 = p = 44 G P a ,

pcx2 = 202 GPa, p = 8940 k g / m 3

£ 150 Viscoplastic Model:

th
7 = 7(7-, 7) = 7Jl +i 7g2-
where
60-

0 250 500 750 1000 1250 1500 1750

7i = To
Ta + {Ts - Ta){\ - 6 TO)
Time After Shear Wave Arrival (ns)

FIG 6. Stress-time profiles and strain-rate history for a 72 = imi")^ •

strain-rate-change test on OFHC copper (V = 0.112 mm//is,
8= 14.5°, h = 25n m).
Parameter values:
I) in which the material parameters are chosen to obtain
close agreement with the measured response in constant
strain rate experiments. This model does not provide a ra = 80 MPa, TS = 265 MPa,
satisfactory description of the response in the strain-rate
change test. Indeed, the measured and predicted pro- 7o = 0.2, m = 0 . 2 ,
files differ qualitatively at early times after the arrival of
the unloading wave: the former is continuous and slowly 7o = 5 x 10 5 s - \ m* = 0.75, j m - 3 x 10 6 s " 1 .
varying while the latter falls precipitously. These differ- I n t e r n a l V a r i a b l e M o d e l : Eqs. (21) - (25)
ences are discussed more fully in the section on Analysis.
At this point it is sufficient to note that the behavior is Parameter values:
contrary to the predictions of conventional viscoplasticity
and suggests the need for considering other constitutive 7o = l x l 0 5 s - 1 , m = 0.05,
models.

/?o = 8 x 10 5 s - 1 , mh = 0.515,
3. ANALYSIS

T 0 (0) = 60 MPa, v = 1-940,

Whenever the stress state in the high-strain-rate,
pressure-shear experiment is uniform through the thick-
ness of the specimen, the stress and strain rate in the spec- r s 0 = 73 MPa, j s l = 1 x 105 so - l
imen can be obtained directly from the measured particle
velocity by means of Eqns. (2), (3) and the analogous
equations m,i 0.05, 7»2 = 2 x 10 6 s-

urj - ufs
(4) ms2 = 0.2, 7 3 3 = 1.5 x 10 6 „-l

a
= 2^Cl)5t'^s' (5) ms3 = 0.525, A = 175, n = 0.75.

for the normal strain rate, e, and the normal stress a. In

Eqns. (4) and (5), u0 — Vcos# is the normal component of for determining the plastic response at fixed structure,
the projectile velocity, ujs is the normal component of the as in the strain rate change experiment, because that re-
velocity of the free surface of the target plate and {pc\)s sponse occurs before the stress becomes uniform through
is the elastic longitudinal wave impedance of the steel tar- the thickness of the specimen.
get. Even when the stress state is not uniform through Stress waves in the plastically deforming specimen sat-
the thickness of the specimen, the recorded motion of the isfy the m o m e n t u m conservation equations
free surface of the target plate is useful for interpreting
the behavior of the specimen at high shear strain rates. ex = Pout, (6)
However, relating the recorded motions to the material rx = Povt, (7)
response of the specimen requires a full analysis of the
wave propagation in the specimen for an assumed consti- and the compatibility equations
tutive model. Such an analysis is particularly important f* = u
x, (8)

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S14 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

It - vx, (9)
where subscripts denote partial differentiation and po is
the initial density of the material. T h e coordinate X cor-
responds to the coordinate, say Xi, in the direction of
wave propagation. T h e shearing deformation occurs in
the X\ — X2 plane. In addition, the stresses and strain
rates satisfy constitutive equations t h a t describe the elas-
tic/viscoplastic response of the material. Such equations
have been formulated by Ramesh and Clifton (1990) for
finite deformations in pressure-shear for isotropic materi-
als in which both the elastic and plastic deformations are
finite and by Klopp (1987) for face-centered-cubic metal
crystals in which the plastic deformation and rotation are 0 .05 .10 .15 0 .05 .10 .15 0 .05 0 .05 .10 ,15 .20
finite, but the elastic strains are infinitesimal. However, Shear Strain
here the complexities of finite deformations are put aside
to focus attention on the strain hardening issues that are FIG 7. Strain-rate jump experiments in torsion on thin-walled,
OFHC copper specimens (Senseny, Richman and Duffy
of primary interest in the strain-rate-change experiments.
(1975)).
Furthermore, the strains are small over the times of the
principal interest during, and following, the reduction in which ensure t h a t the total strain rates e22 and e 33 are
strain rate. zero as required by the geometry of pressure-shear im-
Governing equations for infinitesimal deformations in pact; E and v in (14) and (15) are Young's modulus and
pressure-shear have been presented and solved numeri- Poisson's ratio, respectively. For the calculations reported
cally by Gilat and Clifton (1985) and Gilat (1988) for here, isotropic hardening is assumed. With this assump-
interpreting waves generated by pressure-shear impact. tion the flow potential / is the Huber-Mises function
These equations consist of equations (6-9) and the con-
stitutive equations (16)
/ 0 « j ) = (2S>'is0") = -
1
*/- ^ f (10)
£« = T h e hardness measure To is assumed to evolve during
—-at
P0C1 + ${T,TQ^ da'
—
plastic deformation according to
It -r,+$(r,r0): , (11)
P0C2 or TO = H(TQ,JP)JP (17)

where c\, ci are, respectively, the elastic longitudinal and where H(TO,JP) is a hardening rate function which, along
shear wave speeds in the specimen, <&(r, r 0 ) is a viscoplas- with <£>(r, To), is to be determined from a series of exper-
tic strain rate function t h a t corresponds to the second iments. The form of the functions $ ( r , ro) and H(TQ,JP)
invariant of the plastic strain rate tensor, i.e. can be suggested by considering specific mechanisms of
plastic flow and strain hardening as done by Follans-
$ = 7P = (2c,- i e, 7 ) (12) bee and Kocks (1988), Zerilli and Armstrong (1987) and
Klepaczko (1988). However, in view of the lack of cer-
where the convention of summation over repeated indices tainty of the mechanisms at high strain rates, and the
is used and superposed dots denote the material time inherent difficulties of extending dislocation-based mod-
derivative which, for the infinitesimal strain description els to polycrystalline materials, the functions <3?(r, r 0 ) and
considered here, is equivalent to the partial derivative ( ) t . H(TO,JP) are represented here by simple, empirical ex-
The function $ depends on an equivalent shear stress r pressions. T h e stress TQ is identified as the flow stress,
defined by at the current dislocation structure, or "hardness", for de-
,1 formation at a reference plastic strain rate 70. T h a t is,
T=(nSijSijY' (13) the hardness or reference flow stress TQ is defined by an
idealized experiment in which the rate of plastic defor-
where s,-;- denotes the stress deviator ffij — go^iA'j, and mation is changed instantaneously to the reference strain
a reference stress TQ that represents the strain hardened rate 70; the corresponding change in flow stress is from r
state of the material. T h e function / is a flow poten- to TQ. For this definition of To it follows t h a t in a test at
tial /(cr, r, <722,<T33,) where cr22 and 0-33 are the normal a constant plastic strain rate equal to the reference strain
stresses on planes transverse to the direction of wave prop- rate, the stress TQ and the plastic strain rate 70 satisfy
agation. These normal stresses are evaluated by integra- 7 0 = $ ( r 0 , r 0 ) . Consequently, the function $ is required
tion of the equations to satisfy $7- -f <3>To = 0 along f = T0. This constraint is
satisfied by taking $ to depend only on the "overstress"
T—r0 or the "fractional overstress" T/TQ. From strain-rate
r -^22 - ^(0-33 + <?•) + $ ( r , r 0 )-^-— = 0, (14)
E ' (9(722 j u m p tests in which the strain rate is increased quickly
from 2 x 1 0 _ 4 s _ 1 to 3 x 1 0 2 s _ 1 it appears, as shown
033 ^(*22 + <r)+*(T,T0)^- = 0, (15) in Fig. 7, that a fractional overstress model provides a
Jb 0(733

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S15

better description. Furthermore, the quotient r/r0 is a the plastic strain rate function is modified to have the
suitable dimensionless variable for describing the driving form
force on dislocations when the primary resistance to their V Tn /
motion is provided by the stress fields of dislocation con- -) = (23)
TO V m m I
figurations which have been developed during the prior i + 0(£) h

deformation. T h u s , $ ( 7 , TO) is taken to have the form

where the exponent m/t (m/j > m) and the coefficient
/? (0 < P < 1) are introduced to model an enhanced rate
$(r,7- 0 ) = 7o</>(r/r 0 ), (18)
sensitivity of the flow stress at the highest strain rates en-
countered in the calculations. This equation agrees with
where <j> satisfies ^(1) = 1.
(20), except at large values of 7 / r 0 where the second term
To account for the strain rate dependence of strain in the denominator becomes dominant and reduces the
hardening, an elementary model is introduced to relate calculated strain rate. At sufficiently large strain rates
the hardening rate under any loading history to the hard- the strain rate function becomes
ening rate at a corresponding point of a constant strain
rate test at the current strain rate. The hardening rate To
function is chosen to be consistent with the observation $(r,r0) = (24)
that dynamic stress-strain curves at the essentially con-
stant shear strain rates of pressure-shear impact experi- To ensure t h a t the stress difference in (21) remains pos-
ments reach a plateau level at large strains and remain itive, the coefficient (3 and the parameter m;, are chosen
constant thereafter. To account for thjs_saturation of the such t h a t the curves representing the flow stress at con-
stress, the strain hardening rate H(TQ,JP) is taken to have stant structure do not intersect the curve TS(JP) of (22)
the form within the regime of stresses and strain rates t h a t are of
interest in the calculations. To this end, the coefficient /?
H(r0^) = h{^^-^)) (19) is taken to decrease with increasing TQ according to
To

where TS(JP) is the saturation stress at the plastic shear ro(0)

P = Po (25)
strain rate j p and 70 is a reference strain. The function h TO
is taken to satisfy h(0) = 0 to ensure t h a t the hardening
rate vanishes as the shear stress approaches the saturation where /?o,P are positive parameters and ro(0) is the ini-
stress for the current strain rate. tial value of ro. T h e additional parameters introduced by
these modifications are also given in Table I. T h e depen-
T h e choice of the reference strain rate can be made
dence of the saturation stress or steady-state stress on
arbitrary by requiring the functions <f> and h to be homo-
the strain rate is shown in Fig. 8, along with the rate
geneous functions of their arguments. Thus, the plastic
sensitivity at constant structure given by (18) and (23).
strain-rate function <f> and the strain hardening rate h are
T h e dashed curves indicate the rate sensitivity of the flow
taken to be, respectively,
stress without the modification (23) for high strain rates;
similarly, the limiting behavior (24) at high strain rates
v_ ) V ) '
(20) is given by the dotted curves. Although the modifica-
r
o r
o tion (23) was motivated by computational considerations,
the general features of the curves at high strain rates are
and believed to be representative of material behavior. Pa-
rs(yP) rameter values /?o, m /j>p, T o(0) for the modification are
h( n
-)=^(r,(T")-r) 7o . m
(21)
To constrained by the requirement t h a t the measured and
where m,n,A are constants and 70 has been replaced by
(^To") - "" to ensure t h a t the stress-time history r(t) for a
given strain-rate history jp(t) is independent of the refer-
ence strain rate. T h e function rs(jP)), based on results of
pressure-shear experiments on O F H C copper at constant
shear strain rate, is approximated by

^ ( T P ) ) = T30 + T (22)
• • • £ \7s
(!

Values for the exponents m„; and the dependent param-

eters T,i,j,i which provide the best fit of experimental
data for O F H C copper are given in Table I, along with
values for the parameters m,n,A. 0 2 4
Strain Rate (1/s, Log)
To avoid the use of unacceptably small time steps at
early times when large overstresses and, correspondingly, FIG 8. Modeled rate dependence of the flow stress at fixed
very large plastic strain rates are predicted by the model, structure and the steady state stress for OFHC copper.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S16 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

300'

600

;S 400-

IX.
a>

200-

100

2 4
Strain Rate (1/s, Log) 250 500 750 1000 1250 1500 1750
FIG 9. Modeled rate dependence of the hardening rate at Time After Shear Wave Arrival (ns)
fixed structure for OFHC copper.
FIG 11. Comparison of measured and computed stress-time
histories for a strain-rate-change experiment on an OFHC cop-
computed stress-time profiles match reasonably well at per foil (V = O.112mm//is,0 = 14.5°, h = 25/im).
early times before the stress becomes uniform through
the thickness of the specimen. pact experiment on O F H C copper at a relatively high
T h e hardening rate (21) is shown in Fig. 9. At fixed impact velocity is shown in Fig. 10. T h e calculations are
structure, the hardening rate increases with increasing made using the internal variable model of Table I and a
strain rate. T h e break in each curve occurs at the strain second order accurate difference method (see e.g. Ran-
rate for which the transition occurs in the strain rate ganath and Clifton (1972)). To eliminate numerical diffi-
sensitivity, according to the curves shown in Fig. 8. culties associated with shock wave fronts, a risetime of 1
Although there is insufficient experimental d a t a to con- ns is given to the loading at the impact face. T h e numer-
strain the hardening rate function for the highest strain ical solutions indicate that a homogeneous state of stress
rates shown, good agreement of computed and measured is established after approximately 100 ns. Before the cal-
stress-time profiles appears to require the hardening rate culated stress state becomes homogeneous, the agreement
to level off or even decrease at strain rates above, say, between the predicted and measured stress-time profiles
1 0 7 s - 1 . T h e preceding modifications affect the solution is not fully satisfactory. This lack of agreement indicates
significantly only during early times when the strain rate t h a t the model underestimates the plastic strain rate in
is much greater than the nominally constant strain rate the initial high stress regime. After the stress becomes
achieved at later times. T h u s , these modifications have uniform through the thickness, the main features of the
negligible effect on the comparisons in the regimes of pri- experimental curves are captured by the model.
mary interest in these experiments.
T h e same model for O F H C copper has also been used
A computed stress-time history for the transmitted to interpret the results of strain-rate-change experiments.
stress in a pressure-shear, constant-strain-rate plate im- In this case the calculations were modified to include the
waves reflected from the interface between the steel flyer
plate and the aluminum backup plate. Comparison of
6.0 9 the measured and computed stress-time profiles for one of
the strain-rate-change experiments is shown in Fig. 11.
The strain-rate history is given by the downward slop-
ing curve, which begins at the time when the calculated
stress in the specimen becomes uniform. The stress-time
3.0 to profile shows a small dip in the shear stress at the time
a: (t «s 500ns) when the normal stress in the specimen is re-
o> 200 c duced by the longitudinal unloading wave reflected from
'<t>

160-
2.0 W the front of the aluminum backup plate. Computation-
ally this dip corresponds to a non-proportional loading
effect (See Fig. 12). Sharp unloading of the normal stress
moves the stress state to a point on a neutral loading path
F 60- of constant r . Reloading along a similar p a t h , but with
slightly increasing r , restores the s t a t e of stress to nearly
250 500 750 1000 1250 1500 1750
that of simple shear.
Time After Shear Wave Arrival (ns)
T h e principal feature of interest in Fig. 11 is the stress-
FIG 10. Stress-time history for a pressure-shear experiments time profile at times beginning at t « 960ns when the
on an OFHC copper foil at nominally constant strain rates unloading shear wave arrives. At this time the mea-
{V = O.18O5mm/>s,0 = 22.0°, h = 25/im). sured nominal shear strain rate decreases sharply (again,

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S17

this shear strain rate should not be viewed as a uniform described above because the behavior at fixed structure
strain rate for homogeneous deformation of the specimen is described by a small value of the rate sensitivity pa-
until the stress becomes uniform through its thickness). rameter m.
The measured stress-time profile for the transmitted shear For conventional viscoplastic models the hardening
stress decreases gradually for approximately 600ns before is assumed to depend only on the accumulated plastic
the first arrival of unloading waves from lateral bound- strain, not the rate at which it is introduced. In this case
aries. Computed stress-time profiles based on the consti- h does not depend on the strain rate •yf and h = ft(ro)
tutive model described above are shown as dotted curves becomes simply the slope of the stress-strain curve at a
with the lighter and darker dots corresponding, respec- reference strain rate. For such models the hardness ro can
tively, to the stress at the front and rear faces of the be expressed as a function of the accumulated strain, re-
specimen. T h e model predicts correctly that the trans- sulting in the usual viscoplastic formulation in which the
mitted shear stress does not drop sharply when the un- plastic strain rate is a function of the current stress and
loading wave arrives. This prediction requires that the the accumulated plastic strain. When a model of this type
negative j u m p in stress at the front face of the specimen is used to describe the behavior of O F H C copper in the
be fully attenuated during the propagation of the unload- tests at constant strain rate the rate sensitivity parameter
ing wavefront through the specimen. The attenuation of must have the much larger value m = 0.2 in order to fit
this j u m p can be analyzed by considering the ordinary the observed increase in flow stress with increasing strain
differential equations rate. For this value of m the calculated stress-time pro-
files for the strain-rate change test, shown in Fig. 6, show
dr dv >MT,T0)T a sharp decrease in the transmitted stress at the arrival of
~P0C2 (26)
~dt 2f the shear wave of unloading. This drop in shear stress in-
dicates t h a t , according to (27), the discontinuity in shear
which hold along the characteristics X = c 2 at the lead- stress is not attenuated sufficiently because the change in
ing and trailing edge of the j u m p . Taking the difference plastic strain rate across the wavefront is too small. Sim-
between these two relations on the two sides of the j u m p ilar behavior has been observed in all strain-rate j u m p
and eliminating the j u m p in particle velocity [v] by using tests in O F H C copper. Thus, it appears that the strong
the j u m p condition [r] = — P0C2H o n e obtains rate sensitivity observed in constant strain rate tests on
O F H C copper must be attributed primarily to the rate
d[r] Poc2 $(r,r0)i sensitivity of the evolution of hardness, not to the rate
(27)
dt IT sensitivity of the flow stress at constant structure.

where square brackets denote the j u m p in the enclosed

quantities across the wavefront. Thus, the rate of atten- 4. DISCUSSION
uation of the stress j u m p [r] is proportional to the dif-
ference between the plastic shear strain rate on the two
sides of the wavefront. T h e hardness measure TO does not Comparison of the experimental and computational re-
change across the wavefront because the instantaneous sults for the strain-rate change experiment indicates that
response of the material is elastic. Therefore, the attenu- the large increase in flow stress in pressure-shear experi-
ation of [r] is a measure of the rate sensitivity of the plas- ments at very high strain rates is due largely to the strong
tic flow at fixed structure. From (27), strong attenuation rate sensitivity of the hardening rate. Although the com-
occurs when a sharp decrease in the shear stress at fixed parisons are made using simple empirical expressions for
structure causes a strong decrease in strain rate. Such be- the hardening rate and the flow rate at fixed structure,
havior occurs for the strain-rate history dependent model this overall conclusion does not depend sensitively on the
form of these functions. Quite generally, the lack of a
step-like decrease in the transmitted shear stress in the
strain-rate change experiments is a strong indication that
the flow stress decreases much less slowly with decreas-
260-
ing strain rate than indicated by the measured increase in
the saturation stress with increasing strain rate. Related
"•••..'•-., f = 236 MPa
indications of the influence of the rate of deformation on
200-
the hardening rate appear in other investigations of the
dynamic plastic flow of fee metals. Pioneering research
150-
related to the detection and modeling of such effects in
f = 221 M P a \ \ O F H C copper has been conducted by Follansbee, Kocks
100-
v W and co-workers (See, e.g., Follansbee and Kocks (1988)).

II \ \\
Their approach consists of conducting quasi-static and
50- split Hopkinson pressure bar experiments over a range of
strain rates and probing the hardness developed by mea-
0-
-300
-ii, -200
,
-100
r I
0 100
,U-
200 300
suring the so-called "mechanical threshold stress" of the
deformed samples at 0 K. T h e y report large increases in
(CV CT„)/V3 (MPa) the hardening rate at strain rates of 1 0 3 s - 1 and higher.
FIG 12. Stress trajectory at the rear surface of the specimen They model the behavior with a dislocation mechanics
in the strain-rate-change experiment of Fig. 11. based model in which the dislocation density is assumed

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S18 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

to be constant and the rate controlling process is the ther- experiment. Modifications of the functions used for the
mally activated motion of dislocations past obstacles. In flow rate and the hardening rate may be necessary at
their early work the hardening rate increased as strongly these lower strain rates; however, the principal elements
as linearly with increasing strain rate at high strain rates. of the strain-rate history dependent modeling presented
Recent evidence indicates that the linear increase may be here appears to provide an attractive framework for rein-
too strong, but t h a t a hardening rate which increases with terpreting plastic waves in bars and tubes.
the square root of strain rate at high strain rates may be
acceptable (See Tonks and Johnson (1989)). Unloading waves in plastic wave experiments offer fur-
ther possibities for examining the validity of various con-
Strain-rate j u m p experiments in the torsional Kolsky stitutive models for dynamic plastic flow. According to
bar configuration can also be interpreted as indicating a rate independent theory, unloading occurs elastically
that the hardening rate increases with'increasing strain and the unloading wave speeds are required to satisfy in-
rate. After the j u m p in strain rate the stress is less than equalities which, for example, require the speed of the
the stress in a constant strain rate test at the higher strain unloading wave marking the end of a plastic pulse propa-
rate. Furthermore, the initial strain hardening slope after gating along a bar to travel at a speed t h a t , as shown by
the j u m p tends to be larger than the slope, at the lower Lee (1953), is greater t h a n the plastic wave speed at the
strain rate, prior to the j u m p . Both of these features of state from which unloading is occurring and less than the
the curves are in qualitative agreement with predictions of elastic wave speed. According to a rate dependent theory
a strain-rate history dependent model with an enhanced of the conventional overstress type, there is no restriction
hardening rate at high strain rates. on the unloading wave speed and unloading waves are
relatively smooth because plastic flow continues during
Plastic waves in bars and tubes may also be interpreted unloading until the flow stress is reduced to that of the
as indicating strain rate history effects. Rate indepen- quasi-static stress- strain curve. Experiments by Bodner
dent analyses of the wave profile often provide quite good and Clifton (1967), for example, indicate t h a t the speeds
agreement with the experimental profiles, provided t h a t of unloading waves are consistent with the inequalities
the reference stress-strain curve is taken to be a "dynamic of the rate independent theory and t h a t unloading waves
stress-strain curve" with a higher hardening rate, espe- have relatively sharp wavefronts as predicted by such a
cially initially, than t h a t observed in quasi-static exper- theory. Viscoplastic models with weak rate sensitivity
iments. Plastic wave profiles for long bars and tubes, exhibit similar behavior. Strain-rate history dependent
subjected to constant velocity impact, consist of an elas- models of the type presented here would result in unload-
tic precursor, a main plastic wave, and a final constant ing wave behavior t h a t is similar to the predictions of a
state region. T h e main plastic wave propagates essen- viscoplastic theory with the exception t h a t the unload-
tially as a simple wave in which a given level of particle ing wave would be sharper because the rate sensitivity at
velocity or strain propagates at a constant velocity, as constant structure, which appears to be more appropriate
predicted by a rate independent theory. However, unlike for unloading, is less t h a n the total rate sensitivity during
the predictions of a rate independent theory, the main the evolution of hardening structures.
plastic wave makes a smooth transition to the final con-
stant state. These observations are qualitatively consis- Another means of probing the flow stress of a mate-
tent with the predictions of strain-rate history sensitive rial that has been subjected to very high strain rates is
models of the type presented here. T h e interpretation to examine the strength of the elastic release wave that
of the constant state region differs from t h a t of conven- can be propagated through the material in its shocked
tional viscoplastic models of the "overstress" type. Ac- state. Such experiments are conducted by normal impact
cording to such models a constant state is approached of a plate specimen by a somewhat thinner flyer plate.
as the overstress vanishes. The stress and strain in the T h e amplitude of the release wave, generated by reflec-
constant state region satisfy the quasi-static stress-strain tion of the incident compressive wave at the rear surface
curve which serves as the reference curve in determining of the flyer and propagating through the shocked mate-
the overstress. However, experiments indicate that the rial at the elastic longitudinal wave speed, is interpreted
locus of stress and strain states in the constant state re- as a measure of the yield strength of the material in the
gions for a series of experiments at increasing impact ve- shocked state. This experiment is similar to the strain-
locities is a curve which lies above the static stress-strain rate-change experiments in t h a t the material is subjected
curve. Similarly, if the strain-time profiles are used in a to high rate deformation for a short time, followed by
rate independent theory to predict the reference stress- rapid unloading. However, the strain rate has been re-
strain curve from the measured velocities of propagation duced to a small value before the unloading wave arrives
of different strain levels, the predicted curve lies above and the unloading wave is so strong t h a t the unloading
the static stress-strain curve. Such deviations from the causes plastic flow in the reverse direction. Thus, such ex-
static stress-strain curve, while not great at the nomi- periments provide primarily a probe of material response
nal strain rates of 10 2 s~ 1 t h a t characterize plastic wave under reverse loading, or an indication of the Bauschinger
experiments in bars and tubes, are of the type expected effect for a material with a dislocation structure that was
when the hardening rate increases with increasing strain generated at high strain rates. Such information is im-
rate. For the model described here, the nominally con- portant for an overall understanding of hardening at high
stant state region is reached as the reference stress in- strain rates, but it is not easily related to the unloading
creases and the flow stress decreases until the effective behavior in the pressure-shear, strain-rate-change exper-
strain rate becomes negligible on the time scale of the iment discussed here.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S19

Inclusion of the rate sensitivity of the hardening rate approximately 0.3D and (ii) the same equation relates T
in the description of rate dependent plastic flow was in- and D values over twelve decades of strain rate, ranging
troduced by Mecking and Kocks (1981). They introduced from creep rates to the very high rates of the pressure-
the rate sensitivity of strain hardening as a means for de- shear impact experiments; the corresponding range in the
scribing dynamic recovery of fcc metals. Subsequently, values of T and D is more than an order of magnitude.
Follansbee and Kocks (1988)and Follansbee (1988) ex- These two observations suggest that a deformation mech-
tended this model to describe the plastic flow of OFHC anism related to dislocations bowing out from cell walls
copper at high strain rates. They related the hardness of and sweeping across the cell is a primary mechanism of
the current dislocation structure to the so-called mechan- plastic flow in copper over a wide range of strain rates.
ical threshold stress, defined to be the extrapolated yield
To relate the rate of deformation for such a mechanism
stress at 0 K. The flow stress at constant structure was
to the applied stress one can express the rate of shearing
taken to depend weakly on the strain rate, in accord with
on a slip system due to the passage of a dislocation across
the rate controlling process being the thermally activated
a cell as
motion of dislocations post obstacles. The hardening rate . Ab
was taken to increase strongly with increasing strain rate. (29)
1= V.6.t'
This behavior was established experimentally; no physi-
cal mechanism was offerred as an explanation for the ob- where A is the area swept by the dislocation, per volume
served rate sensitivity of the hardening rate. V, per time .6.t. If the cell is assumed to be spherical and
the volume V is taken to be the volume in which only
Clifton (1989) attempted to explain the rate sensitivity one dislocation cell is active simultaneously, then V can
of OFHC copper at high strain rates by relating the final -3 r.;
flow stress to the dislocation cell structure of the recovered be represented as V = N D / V 2 where N is the number
specimen. This cell structure consists of nearly equi-axial of close-packed cells of diameter D which comprise the
cells with relatively thick walls as shown in Fig. 13. The volume V. If the area A is taken to be the area of a
final flow stress was found to agree reasonably well with diametral plane of the sphere, then
the value obtained from the relation
. rrV2b
_ 10.5pb (30)
T =-=--, (28) 1= 4N D.6.t)
D
where .6.t is the time for a dislocation to sweep across the
reported for polycrystalline copper by Staker and Holt
cell. If i' and D can be determined from experiments, then
(1972); the average dislocation cell diameter is repre- (30) can be used to estimate the product N .6.t. Now, the
sented in (28) by D. The agreement was remarkable in shear rate i' for a slip system is related to the macroscopic
two respects: (i) the constant in the equation is approxi- shear rate i'P by i' = i'P / K, where K, is a dimensionless pa-
mately the value obtained by considering the bowing of a rameter which can be estimated by making an assumption
dislocation between two obstacles spaced at a distance of regarding the orientation of the grain relative to the im-
posed shearing deformation. At large deformations it is
likely that the grain has rotated to a symmetry orien-
tation in which two or more slip systems are favorably
oriented for slip and have the same resolved shear stress.
With this assumption, the value of K, is approximately 2
and the value of N .6.t obtained from the measured values
of i'P and D is approximately 5 ns.
The time .6.t has been estimated (Clifton (1988)) by
considering the time required for a dislocation to sweep
across a cell when its motion is opposed by internal stress
fields due to other dislocations and the resistance of the
lattice. The resolved shear stress TR(X) acting on a central
segment of a dislocation as it sweeps across a cell can
be regarded as a superposition (motivated by Mughrabi
(1975))

TR(X) + TGW(X) + TID(X)

T
+ TLT(X) + TLR(X), (31)
where the quantities on the right side are the resolved
shear stress due to, respectively, the applied shear stress
T, the multipole dislocations in the cell wall, the imping-
ing dislocation from a neighboring cell, the line tension
of the bowed dislocation, and the long range stress field.
FIG 13. TEM micrograph of dislocation cdl structure for These terms and their superposition are shown schemat-
pressure-shear impact experiment on OFllC coppcr spccimcn ically in Fig. 14. The stress field TGW(X) due to mul-
(V == O.076mm//ls, 8 == 14.5°, h == 26Jlm, -yP ~ 1.1 X IO~S-I). tipoles is a short range field that decays quickly on the

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S20 MECHANICS USA 1990 Appl Mech Rev 1990 Supplement

dislocation density. By identifying the wavelength with

2.0
the cell diameter of the dislocation cell structure t h a t ul-
timately forms he concluded t h a t the cell diameter is in-
versely proportional to the square root of the dislocation
density. In this way the proportionality of the flow stress
to the inverse of the cell diameter, as given in (28), corre-
sponds to the flow stress being proportional to the square
root of the dislocation density. Such a model allows for
rate dependence of the hardening rate only if the dislo-
cation density in a sample that has been deformed to a
given strain at high strain rates is greater than the dislo-
cation density in a sample t h a t has been deformed to the
same strain at low strain rates. This possibility is allowed
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 in contemporary models of strain hardening by includ-
Normalized Distance ing rate dependent dynamic recovery through dislocation
climb and subboundary migration (See Kocks (1976) and
FIG 14. Schematic of the resolved shear stress at the center Gottstein and Argon (1986)). Neither dislocation climb
of a curved dislocation as it sweeps across the diameter of a
nor subboundary migration appear to be suitable recov-
dislocation cell.
ery mechanisms for explaining the rate sensitivity of the
scale of distances comparable to cell wall thicknesses. T h e hardening rate at the high strain rates considered here
remaining terms vary sufficiently slowly that the driving because insufficient time is available for the diffusion re-
stress on the dislocation becomes quite uniform once the quired for dislocation climb and because the mobility of
dislocation escapes from the region where the stress is the relatively wide dislocation cell walls is expected to
dominated by TCW(X)- T h e time At required for the re- be negligible for the temperatures and pulse durations of
mainder of the sweep can be estimated by replacing Uj the experiments. Thus, new mechanisms for the rate de-
in (1) by dx/dt and integrating along the motion of the pendence of the hardening rate appear to be required for
dislocation as a central segment sweeps across the diame- high strain rates; a mechanism based on an increased rate
ter D, driven by the stress TR(X); a relativistic correction of entanglement appears to be a possibility, b u t further
is used for B to limit the maximum dislocation velocity research is required to evaluate its viability.
to the shear wave speed C2. The resulting value for the
time At is approximately 0.25 ns and the corresponding
estimate for N is 20, which is comparable to the number 5. CONCLUDING REMARKS
N = 12 of cells in contact with a given cell in a close
packed array of spherical cells. Accounting for the escape
Based on the experiments and analysis considered herein
of the dislocation from the wall can be expected to in-
it appears t h a t some reasonably clear indications of es-
crease At and reduce N to a value that may be nearer
sential features of the modeling of the plastic flow of high
to 12. This observation suggests that, under nominally
steady conditions, each impinging dislocation would (on purity metals at high strain rates have emerged. In addi-
the average) cause one, and only one, dislocation to sweep tion, some attractive directions for further research can be
across an adjoining cell. If N were much less than 12, then identified. Such essential features and directions include
there would be a significant probability for more than one the following.
dislocation to be sweeping across a cell at the same time.
Intersection of these dislocations would lead to entangle- 1. T h e rate sensitivity of the hardening rate is an es-
ments and refinement of the cell structure, resulting in in- sential feature of the description of plastic flow over
creased hardness. T h e fraction of dislocations becoming a wide range of strain rates.
entangled would increase with increased strain rate, caus-
ing the hardening rate to increase with increased strain 2. At high strain rates the rate sensitivity of the flow
rate. stress at constant structure is considerably less than
the rate sensitivity of the hardening rate.
Whether or not the rate sensitivity of the hardening
rate is related to the kinetics of dislocation entanglements 3. In constant strain rate tests at high strain rates the
in the nominal 1 "' clear regions of dislocation cells has not flow stress saturates at modest strains. This behavior
been establish _ feed, the existence of a dislocation suggests t h a t a nominally steady dislocation struc-
cell structure at high strain rates is viewed as unlikely ture evolves in such experiments and that the charac-
by those who regard such structures as low energy con- terization of this structure and its relationship to the
figurations to which the dislocations relax after the ap- strain rate sensitivity of the saturation stress would
plied stress is removed. Holt (1970) provided support be a fruitful direction for further research.
for the idea that dislocation cell structures correspond
to low energy configurations by an analysis of the inter- 4. A successful model for the increased hardening rate
action of parallel screw dislocations. He showed t h a t a at high strain rates may result from an analysis of the
minimum energy configuration requires a modulated dis- kinetics of dislocation entanglements where the rate
location density and t h a t the wavelength of the modu- of formation of such entanglements increases with in-
lation is inversely proportional to the square root of the creasing strain rate.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 Appl Mech Rev vol 43, no 5, Part 2, May 1990 Clifton: High strain rate behavior of metals S21

5. Strain-rate change experiments are important in Inelastic Behavior of Solids (M F Kanninen et al Eds.) McGraw-
probing the plastic response of materials at nomi- Hill, New York, 521-542.
nally constant structure. When these experiments Harding. J , Wood, E D and Campbell, J D (1960). Tensile testing
are conducted in the pressure-shear plate impact con- of materials at impact rates of strain, J Mech Eng Sci 2, 88-96.
figuration for which one- dimensional wave theory Hartley, K A and Duffy, J (1985). The torsional Kolsky (split- Hop-
applies, the comparison of computed and measured kinson) bar, in Metals Handbook, Volume 8: Mechanical Testing,
wave profiles at early times after the arrival of the 9th Edition, American Society for Metals, Metals Park, 218-228.
wavefront of the strain-rate change is especially valu- Hauser, E F (1966). Techniques for measuring stress-strain relations
able in assessing the validity of proposed models. at high strain rates, Exp Mech 6, 395-402.
Herrmann, W , Hicks, D L and Young, E G (1970). Attenuation
Acknowledgements of elastic-plastic stress waves, Shock Waves and the Mechanical
Properties of Solids (J J Burke and V Weiss, Eds.) Syracuse
The author is especially grateful to Wei Tong who University Press, Syracuse, 23-63.
conducted the pressure-shear, strain-rate-change experi-
ments and did the associated calculations. A full account Holt, D L (1970). Dislocation cell formation in metals, J Appl Phys
4 1 , 3197-3201.
of this research will be presented as his PhD thesis. The
author is also grateful to John Jarrell for the TEM mi- Huang, S and Clifton (1985). Dynamic plastic response of OFHC
crographs of dislocation cell structures in the deformed copper at high strain rates, in Proceedings of IUTAM Symposium
on Macro- and Micro-Mechanics of High Velocity Deformation and
OFHC copper specimens and to the Office of Naval Re- Fracture, held in Tokyo, 63-75.
search for their support of this research.
Jassby, K M and Vreeland, T , Jr. (1970). An experimental study
of the mobility of edge dislocations in pure copper single crystals,
KEFERENCES Phil Mag 2 1 , 1147-1168.
Abou-Sayed, A S and Clifton, R J (1977). Pressure-shear waves in Johnson, J N and Barker, L M (1969). Dislocation dynamics and
6061-T6 aluminum due to oblique-plate-impact, J Appl Meek 9 9 , steady plastic wave profiles in 6061-T6 aluminum, J Appl Phys
79-84. 40, 4321-4334.
Bodner, S R and Clifton, R J (1967). An experimental investigation Johnson, J N , Jones, O E and Michaels, T E (1970). Disloca-
of elastic-plastic pulse propagation in aluminum rods, J Appl Mech tion dynamics and single-crystal constitutive relations: shock wave
89, 91-99. propagation and precursor decay, J Appl Phys 4 1 , 2330-2339.
Chhabildas, L C , Sutherland, H J and Asay, J R (1979). A velocity Johnston, W G and Gilman, J J (1959). Dislocation velocities,
interferometer technique to determine shear-wave particle velocity dislocation densities, and plastic flow in lithium fluoride crystals,
in shock-loaded solids, J Appl Phys 50, 5196-5201. J Appl Phys 30, 129-144.
Clifton, R J and Klopp, R W (1985). Pressure-shear plate impact Kim, K S , Clifton, R J and Kumar, P (1977). A combined normal
testing, in Metals Handbook, Volume 8: Mechanical Testing, 9th and transverse displacement interferometer with an application to
Edition, American Society for Metals, Metals Park, 230-239. impact of Y-cut quartz, J Appl Phys 4 8 , 4132-4139.
Clifton, R J (1988). Constitutive models for plastic flow at ultra- Kim, K S and Clifton, R J (1980). Pressure-shear impact of 6061-T6
high strain rates, in Dynamic Constitutive/Failure Models, Pro- aluminum, J Appl Mech 47, 11-16.
ceedings of Workshop held at University of Dayton, (A M Rajen-
dran and T Nicholas, Eds.) AFWAL-TR-88-4229, 28-55. Klepaczko, J (1988). Constitutive modeling in dynamic plasticity
based on physical state variables — a review, J de Physique, Col-
Duffy, J , Campbell, J D , and Hawley, R H (1971). On the use loque CS, suppl au n° 9, Tome 49, 553-560.
of a torsional split Hopkinson bar to study rate effects in 1100-0
aluminum, J Appl Mech 38, 83-91. Klopp, R W (1987). Plasticity of aluminum and iron at high shear
strain rate and high pressure, Ph D thesis, Brown University, Prov-
Follansbee, P S and Kocks, U F (1988). A constitutive description idence, RI.
of the deformation of copper based on the use of the mechanical
threshold stress as an internal state variable, Acta Met 36, 81-93. Klopp, R W , Clifton, R J and Shawki, T G (1985). Pressure-shear
impact and the dynamic viscoplastic response of metals, Mechs of
Follansbee, P S (1988). High-strain rate deformation mechanisms
Mails 4, 375-385.
in copper and applications for behavior during shock wave defor-
mation, in Shock Waves in Condensed Matter (S C Schmidt and Kocks, U F (1976). Laws for work-hardening and low-temperature
N C Holmes, Eds.) Elsevier, New York, 249-254. creep, J Engng Mater Tech, ASME, Ser H 98, 76-85.
Frantz, R A and Duffy, J (1972). The dynamic stress-strain behav- Kolsky, H (1949). An investigation of the mechanical properties of
ior in torsion of 1100-0 aluminum subjected to a sharp increase in materials at very high rates of loading, Proc Phys Soc London,
strain rate, J Appl Mech 39, 939-945. Series B 62, 676-700.
Gilat, A (1985). A viscoplastic theory with anisotropic hardening Kumar, P and Clifton, R J (1979). Dislocation motion and genera-
and its application to pressure-shear plate impact experiments, J tion in LiF single crystals subjected to plate impact, J Appl Phys
Appl Mech 52, 629-633. 50, 4747-4762.
Gilat, A and Clifton, R J (1985). Pressure-shear waves in 6061- T 6 Lee, E H (1953). A boundary value problem in the theory of plastic
aluminum and alpha-titanium, J Mechs Phys Solids 33, 263-284. wave propagation, Q Appl Math 10, 335-346.
Gottstein, G and Argon, A S (1987). Dislocation theory of steady Li, C H (1982). A pressure-shear experiment for studying the dy-
state deformation and its approach in creep and dynamic tests, namic plastic response of metals at shear strain rates of 1 0 5 s — ' ,
Acta Met 35, 1261-1271. P h D thesis, Brown University, Providence, RI.
Grady, D E (1981). Strain-rate dependence of the effective viscosity Lindholm, U S (1964). Some experiments with the split Hopkinson
under steady-wave shock compression, Appl Phys Lett 38, 825- pressure bar, J Mech Phys Solids 12, 317-335.
826.
Lindholm, U S (1971). High strain rate tests, in Techniques of Met-
Green, S J , Maiden, C J , Babcock, S G , and Schierloh, F L (1970). als Research, Vol 5, Part 1, Measurement of Mechanical Proper-
The high strain-rate behavior of face centered cubic metals, in ties (R F Bunshah, Ed.) Interscience, New York, 199-271.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms

 S22 MECHANICS USA 1990

Lindholm, U S and Yeakley, L M (1968). High strain-rate testing:

tension and compression, Exp Mechs 8, 1-9.
Mecking, H and Kocks, U F (1981). Kinetics of flow and strain-
hardening, Acta. Met 2 9 , 1865-1875.
Meir, G and Clifton, R J (1986). The effects of dislocation gener-
ation at surfaces and subgrain boundaries on precursor decay in
high purity LiF, J Appl Phys 59, 124-148.
Mughrabi, H (1975). Description of the dislocation structure after
uniaxial deformation at low temperatures, in Constitutive Equa-
tions in Plasticity (A S Argon, Ed.), MIT Press, Cambridge, 199-
250.
Nicholas, T (1980). Tensile testing of materials at high rates of
strain, Exp Mech 2 1 , 177-185.
Nicholas, T (1982). Material behavior at high strain rates, in Impact
Dynamics by J A Zukas, T Nicholas, H F Swift, L B Greszczuk,
and D R Curran, Wiley-Interscience, New York, 277-332.
Pope, D P , Vreeland, T Jr. and Wood, D S (1967). Mobility of
edge dislocations in the basal slip system of zinc, J Appl Phys 38
4011-4018.
Ramesh, K T and Clifton, R J (1990). Finite deformation analy-
sis of pressure-shear plate impact experiments on an elastohydro-
dynamic lubricant, manuscript submitted for publication, (to be
presented at 11th National Congress of Theoretical and Applied
Mechanics).
Ranganath, S and Clifton, R J (1972). A second-order accurate dif-
ference method for systems of hyperbolic partial differential equa-
tions, Comp Meth in Meek and Engr 1, 173-188.
Senseny, P E , Richman, M H and Duffy, J (1975). The influence
of annealing temperature on the strain rate sensitivity of copper
in torsion, in J Appl Mech 42 245-246.
Staker, M R and Holt, D L (1972). The dislocation cell size and
dislocation density in copper deformed at temperatures between
25 and 700 d e g C , Acta Met 20, 569-579.
Stelly, J and Dormeval, R (1977). Some results on the dynamic
deformation of copper, in High Velocity Deformation of Solids (K
Kawata and J Shioiri, Eds.) Springer-Verlag, Berlin, 82-97.
Taylor, J W (1965). Dislocation dynamics and dynamic yielding, J
Appl Phys 36, 3146-3150.
Tong, W , Clifton, R J and Huang, S (1990). Pressure-shear im-
pact investigation of strain-rate history effects in OFHC copper,
in preparation.
Tonks, D L and Johnson, J N (1989). Shock-wave evolution of the
mechanical threshold stress, in Shock Compression of Condensed
Matter Proc of the 1989 Am Phys Soc Topical Conference, Albu-
querque, N.M.
Zerilli, F J and Armstrong, R W (1987). Dislocation-mechanics-
based constitutive relations for material dynamics calculations, J
Appl Phys 6 1 , 1816-1825.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 09/19/2013 Terms of Use: http://asme.org/terms