ARTICLES
Shock deformation of face-centred-cubic
metals on subnanosecond timescales
E. M. BRINGA1 *, K. ROSOLANKOVA2 , R. E. RUDD1 , B. A. REMINGTON1 , J. S. WARK2 , M. DUCHAINEAU1 ,
D. H. KALANTAR1 , J. HAWRELIAK1 AND J. BELAK1
1
Lawrence Livermore National Laboratory, Livermore, California 94551, USA
2
Department of Physics, Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, UK
* e-mail: ebringa@llnl.gov
Published online: 17 September 2006; doi:10.1038/nmat1735
Despite its fundamental importance for a broad range
of applications, little is understood about the behaviour
of metals during the initial phase of shock compression.
Here, we present molecular dynamics (MD) simulations
of shock-wave propagation through a metal allowing
a detailed analysis of the dynamics of high strain-rate
plasticity. Previous MD simulations have not seen the
evolution of the strain from one- to three-dimensional
compression that is observed in diffraction experiments.
Our large-scale MD simulations of up to 352 million atoms
resolve this important discrepancy through a detailed
understanding of dislocation flow at high strain rates.
The stress relaxes to an approximately hydrostatic state
and the dislocation velocity drops to nearly zero. The
dislocation velocity drop leads to a steady state with no
further relaxation of the lattice, as revealed by simulated
X-ray diffraction.
nature materials VOL 5 OCTOBER 2006 www.nature.com/naturematerials
©2006 Nature Publishing Group
hock compression of condensed matter occurs in a wide
S variety of situations, including high-speed automobile and
aircraft collisions1 , explosive welding2 , armour penetration3,4 ,
meteor impacts5 , interstellar dust–dust collisions6 and inertial
confinement fusion7,8 . Despite considerable research during the
past half century9–11 , a detailed understanding of the three-
dimensional (3D) lattice relaxation processes during shock
compression beyond the elastic limit does not yet exist. Much of
what we know has been inferred from recovery experiments12,13
and analysis of wave profiles14 . The microscopic lattice response
and relaxation behind the shock front has been measured in situ
using time-resolved X-ray diffraction15–21 : shocked lithium fluoride
was observed to relax to the hydrostatic state within tens of
nanoseconds18 , whereas shocked single-crystal copper relaxed in
∼1 ns or less20 .
A detailed picture of the homogeneous nucleation of
dislocations as a function of shock strength emerges from large-
scale non-equilibrium MD simulations of shock-wave propagation
in perfect face-centred-cubic (f.c.c.) crystals22 . Surprisingly, when
X-ray diffraction patterns were calculated from these perfect crystal
simulations23 , the predicted lattice compression was essentially
uniaxial (1D), whereas the diffraction experiments unambiguously
showed a prompt plastic relaxation to a hydrostatic (3D) state20 .
Here, we resolve this dilemma by showing that MD simulations
of shock compression of copper single crystals with pre-existing
defects can achieve a nearly fully relaxed 3D hydrostatic state
provided that the simulations are large enough to access the spatial
and temporal scales of the experiment. This study represents the
first direct, one-to-one comparison and validation of strong-shock
MD simulations with dynamic experimental data at the same
spatial and temporal scales, both at the lattice level and globally.
The combination of the MD simulation detail with experimental
validation offers a powerful tool with which to explore new regimes
in the emerging field of materials science at extreme conditions.
We have carried out several very-large-scale MD simulations
to compare with experimental data from time-resolved diffraction
experiments of shocked single-crystal copper (2-μm-thick sample,
805
ARTICLES
~3D
Shocked,
relaxed
1
Intensity (a.u.)
az
ay
ax ax
a'z
Piston
Up 0.1
0 3
Lattice strain (%)
z
Dislocations: Motion
Figure 1 Dynamic lattice compression during shock loading. The upper red
curve shows the particle velocity versus time behind a shock front moving to the
right (z direction) at speed Us . The response of an initially cubic unit cell of side a0 ,
is illustrated schematically by the blue blocks. The initial strain is uniaxial
compression (1D), with a z < a0 and a x = a y = a0 and large internal shear stress.
The lattice responds by relaxing via volume-preserving dislocation flow to a more
3D-symmetric compression, az ∼ a x ∼ a y . The inset shows experimentally
measured diffraction data with a peak corresponding to the unshocked material at
zero strain and a peak corresponding to prompt (subnanosecond) 3D relaxation
(a z ∼ a x ) for shocked copper at 3%–4% strain20 .
∼0.4 ns shock transit time). The simulations covered up to
∼1-μm-thick copper samples, using up to 352 million atoms.
Our MD-simulated samples had pre-existing dislocation sources
in the form of prismatic dislocation loops. As illustrated in
Fig. 1, one surface of the system, the piston, was driven inwards
with a velocity Up , leading to a shock wave with velocity
Us > Up . Here, we describe two simulations in detail, both
with Up = 0.75 km s−1 and a peak pressure of ∼35 GPa, which
is above the homogeneous dislocation nucleation threshold for
copper. For the first simulation, Up was taken instantaneously from
zero to a constant value, launching a shock with essentially zero
rise time. Previous MD simulations of shocked samples almost
exclusively used a zero rise-time pulse because they spanned no
more than ∼10 ps in their entirety. We note that recent simulations
of shocks in nanocrystalline systems conducted by some of the
authors of this paper have a comparable number of atoms, but a
much shorter timescale24 . For the second simulation, the piston
velocity was increased in a linear ramp over 50 ps, and then held
constant at the same peak Up as the zero rise-time simulation.
This ramped velocity simulates experimental loading pulses which
have finite rise times, ranging from several picoseconds to several
nanoseconds. In the discussions to follow, compression is calculated
as (1 − a/a0 ), where a0 is the equilibrium lattice parameter and
a is the value determined from a diffraction analysis of both our
post-processed MD simulations and our in situ dynamic diffraction
experiments on copper (Fig. 1, inset).
A large density of dislocations, ρd , is created by the shock
compression in both simulations, as shown in Fig. 2a,b with
snapshots at ∼100 ps from both the fast-rise-time simulation and
the 50 ps rise-time simulation (see Supplementary Information,
Movies S1,S2). Whereas there is only homogeneous nucleation of
dislocations for the 0 ps rise-time case (Fig. 2a), there are three
regions of dislocation activity for the 50 ps rise-time case (Fig. 2b
and Supplementary Information, Fig S1), with Fig. 2c showing the
particle speed profile for Fig. 2b. The total dislocation density (ρd )
versus depth is plotted at several times in Fig. 2d, for the zero
rise-time case. ρd saturates at a steady-state value of ∼3 × 1013 cm−2
after ∼70 ps. The value of ρd at the homogeneous nucleation front,
ρmax , is very large compared with the density of pre-existing sources,
806
1D and there is no observable contribution to the final simulated
microstructure due to the initial defects. For the ramped case,
Shocked Unshocked shown in Fig. 2b, there is no dislocation production during the
initial 40 ps period of the ramp due to the large activation threshold
a0 a0 of the sources embedded in the simulation. Once that threshold
a0 a0
az a0 is reached, however, the pre-existing defects lead to dislocation
multiplication and partial stress relaxation. As a result, in the
‘mixed’ region of Fig. 2b, the homogeneous nucleation front is
6
not creating dislocations as prodigiously, and the final ρd , shown
Shock, Us in Fig. 2e, is roughly a third of that in the ‘pure’ homogeneous
nucleation region.
Nucleation Elastic
This sequence suggests that there are several relevant rates
inherent to the material: the rate of increase in dislocation
density (and the closely related hardening rate), the rate of
plastic relaxation and the ramp rate. How they compare with
each other determines the behaviour of the system. The rate of
plastic relaxation can be described by the Orowan equation25 ,
dεp /dt = ρm vd b, where dεp /dt , ρm , vd and b are the plastic strain
rate, the mobile dislocation density, average dislocation velocity
and the Burgers vector, respectively. During shock loading, d εp /dt
can be inferred from in situ X-ray diffraction and the Burgers vector
is known from the lattice. ρm and vd during shock loading are
largely unknown and no theory exists to predict how they evolve
under the rapid rise of shear stress in shock waves. These quantities,
however, can be obtained from the MD simulations. We find
that the strain relaxation is governed by the dislocation evolution
in two distinct regimes: initial unobstructed rapid dislocation
multiplication, followed by decreases in both the mobile dislocation
density and velocity as obstructions accumulate. The time history
of the relevant variables for a material element, calculated from the
MD simulation, is shown in Fig. 3. The 50 ps ramp and the sources
give an extra ∼1% lateral strain relaxation before homogeneous
nucleation of dislocations begins, compared with the lateral strain
in the zero rise-time simulation. The ramp drive with dislocation
sources achieves greater relaxation with a lower total dislocation
density. A longer ramp is expected to give a larger contribution to
the lateral relaxation.
A fully relaxed 3D state would be characterized by roughly zero
shear stress. The shear stress decreases behind the shock front22 ,
as shown in Fig. 4a for both simulations. More than half of the
relaxation occurs 10 ps after shock passage, and corresponds to the
dislocation transport regime (as opposed to the dislocation-loop
nucleation regime). This regime is captured in MD simulations
here for the first time. The shear stress evolves to an asymptotic
value of 0.43 GPa for the zero rise-time simulation and 0.34 GPa for
the ramped loading simulation. This relaxation is comparable to
that inferred from the diffraction experiments on copper13 . Using
a smaller cross-section and a sample ∼2 μm long we find that
the stress indeed saturates in less than 100 ps: nothing is gained
by continuing the simulations to longer times (see Supplementary
Information, Figs S2,S3). When the shear stress relaxes to its
asymptotic value, the dislocation density also reaches an asymptotic
value (Fig. 2d,e). It is clear that the sharpness of the front, which is
controlled by the ramp rate, determines the extent of relaxation. In
the homogeneous nucleation regime, the dislocation multiplication
rate is high, so the system quickly hardens to a high flow stress.
Below the homogeneous nucleation threshold, the ramped wave
can activate the sources and relax the shear at a much lower
hardening rate. The result of this change in the relative rates is that
the ultimate relaxation for the ramped wave is greater.
Given that significant stress relaxation has occurred, as shown
in Fig. 4a, the simulated X-ray diffraction would be expected to
give a nearly uniform 3D compression, meaning ∼5% in each
orthogonal direction for 15% volume compression. Some care is
needed in connecting the stress and strain relaxations, however,
nature materials VOL 5 OCTOBER 2006 www.nature.com/naturematerials
©2006 Nature Publishing Group
 ARTICLES

a d 100

ρd (1012 cm–2)
Time (ps)
10 48
98
138
158
b 1
0.1 μm
Shock
0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
z (μm)

Homogeneous Mixed Multiplication e 100

ρd (1012 cm–2)
10
Time (ps)
c 62
Vz (km s–1)

0.8 1 72
82
0.4 100
143
0 0.1
0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6
z (μm) z (μm)

Figure 2 Dislocation structure resulting from MD simulations of shocked [100] copper. a, Snapshot of a simulation with a shock drive of 0 ps rise time after t ∼ 100 ps,
showing only dislocation atoms. b, The same as a, except that the shock drive had a 50 ps linear-ramp rise time. In both cases, the copper crystal included pre-existing
dislocation sources. The colour in a,b is only to enhance the view of the dislocations. The three regions of dislocation activity—homogeneous, mixed and multiplication—are
marked. c, The particle velocity, V z , profile for the ramped shock, where the ramp extends over z = 0.29–0.43 μm. d,e, The resulting dislocation density profiles for the 0 ps
rise-time case (d) and the ramped case (e). The locations of the pre-existing sources are illustrated by red vertical arrows.

because the shear modulus is known to soften under uniaxial
compression (along the Bain path)26 . The lattice compression
inferred from X-ray diffraction (Fig. 4b) at times <10 ps in the
simulation indicates little lateral relaxation. It is only at times
much greater than 10 ps that the simulation shows significant
lateral relaxation, a result that was first observed in these very
long time simulations. For the zero rise-time case, the maximum
transverse compression is 4.3 ± 0.1%, for a compression of
7.1 ± 0.16% along the shock direction, where the error bars indicate
standard deviation. This has to be compared with the 5.2% needed
for isotropic compression. For the 50 ps rise-time shock with
pre-existing sources, the average final lattice compression along the
x and y directions was found to be 4.7 ± 0.1%, whereas the final
lattice compression along z , the direction of shock propagation,
was 6.25 ± 0.25%, indicating nearly complete plastic relaxation.
The remaining small difference with the experiment may be due to
the large source activation threshold in this simulation (∼10 GPa)
and the fact that most of the ramp lies below this threshold and
does not produce any dislocation activity, as indicated in Fig. 3. For
comparison, the shear stress corresponding to this compression is
[(2/3)1/2 C  {ε zz − (ε xx + ε yy )/2}], where C  = 24 GPa at ambient
pressure, giving 0.55 and 0.30 GPa for the final shear stress in the
zero rise time and ramped waves, respectively, in good agreement
with the values in Fig. 4a. In the early stages of the relaxation, the
softened shear modulus would need to be used, leading to a more
precipitous drop in the stress than in the strain.
On the basis of our simulations, we make two predictions for
experiments with shock rise times of only a few picoseconds. (1) We
predict that prompt 3D relaxation would not be observed in copper,
provided that diffraction measurements could be made at time
intervals of a few picoseconds behind the shock front, which is now
feasible experimentally27–29 . (2) For strong shocks, above the limit
for homogeneous nucleation of dislocations, with steep rises the
effect of pre-existing defects with a density ρ0 would be negligible,
provided that ρ0 ≤ ∼10l−2 , where l is the mean separation
between dislocations at the shock front. Therefore, shocking a
pristine single crystal (ρ0 ∼ 106−8 cm−2 ) in this manner would give
the same result (in terms of relaxation time, dislocation density
generated, VISAR (velocity interferometer system for any reflector)
traces, X-ray diffraction and so on) as shocking a pre-shocked or
radiation damaged crystal with ρ0 ≤ 10l−2 due to the dominance
of homogeneous nucleation for steep rise-time shocks. For a
35 GPa shock in copper, 10l−2 ∼ 1011 cm−2 . Finally, measuring
the time evolution of the crystal from uniaxial compression to
a 3D relaxed state would provide a measure of the product of
the mobile dislocation density and average dislocation velocity,
ρm vd , a determining quantity that has so far eluded measurement
techniques at ultrahigh strain rates. This rate of 3D relaxation is
important for understanding the dynamics of the lattice response
to high strain-rate deformation, in particular, due to a shock. The
large-scale MD simulations shown here comprise an important step
in elucidating these dynamics.
In conclusion, we have shown, with very large-scale MD
simulations at the correct spatial and temporal scales, the ability
to reproduce the experimentally observed global dynamics and
microscopic lattice response of shocked copper; we expect our
results to also apply to other f.c.c. metals. We show three
distinct regimes of lattice relaxation: dislocation nucleation (t1
in Fig. 3a), transport (t2 ) and immobilization due to tangling
and interactions (t > t2 ). The fraction of dislocations that remain
mobile, or more precisely, ρm vd , is the dominant factor in
determining the 3D relaxation rate. Given the detailed insights into
the lattice relaxation dynamics provided by our MD simulations,
it seems possible that experiments could be developed on future
facilities, with measurement timescales ∼10 ps, to actually observe
these distinct regimes of relaxation. Of particular importance
would be the ability to deduce both the total dislocation density and
the fraction that is mobile, as a function of time. Such experiments,
combined with very-large-scale MD simulations, such as those
described here, would provide an inroad into the new frontier of
time-resolved materials science at extreme pressures (>100 GPa)
and strain rates (>106 s−1 ).

nature materials VOL 5 OCTOBER 2006 www.nature.com/naturematerials 807

©2006 Nature Publishing Group

ARTICLES

a t2 t1 t0 a
1.0
2.0
0 ps
0.8
50 ps
Δ t2 Δ t1
ρm /ρmax
0.6 1.6
0.4

σ shear (GPa)
0.2 1.2

0
80 60 40 20 0 0.8
Time (ps)

b
0.3 0.4

0.2
vd /c 0

0
120 90 60 30 0
0.1 Time after shock (ps)

0 b
80 60 40 20 0 0 ps
Time (ps) 10 1D
Z; X–Y
c 0.04 50 ps
(dε /dt)/(ρmaxc0b)

Z; X–Y
8
0.02

Lattice compression (%)

0 6
80 60 40 20 0
Time (ps) 3D
d
6 4
ε (%)

2 2

0
80 60 40 20 0
Time (ps)
Figure 3 The dislocation dynamics of 3D relaxation in MD for the ramped shock
case. a, The evolution of mobile dislocation density, ρ m (t ), normalized by total
dislocation density, ρ max . The times indicated correspond to when dislocation
multiplication starts (t = 0 or t0 ), saturates (t 1 ), when the dislocations become
entangled and immobilized (t2 ), the dislocation nucleation period (t1 ) and the
dislocation transport period (t2 ). b, Average dislocation velocity, vd , normalized by
bulk sound speed, c0 . c, The plastic strain rate from the MD simulation, normalized
by ρ max c0 b. d, Accumulated plastic strain, namely, the time integral of the result
shown in panel c.
METHODS
MD SIMULATIONS
The MD code MDCASK was run using the EAM potential of Mishin et al.30 to
represent copper, with the longitudinal stress threshold for homogeneous
nucleation of dislocations σ zz ∼ 30 GPa. Pre-existing dislocation loops, with a
calculated threshold stress for source activation of σ zz ∼ 10 GPa, were 7 nm
long and separated by 145 nm along the shock direction, for a lateral sample
size of 72 nm. We generated shock waves along the [001] direction, using
200 × 200 × 2,200 and 200 × 200 × 1,600 f.c.c. unit cells for the simulation
shown in Fig. 2a and b respectively, with periodic boundary conditions
transverse to the shock-wave propagation. The source size is inversely
proportional to its activation threshold and such large lateral size is required to
avoid interaction among the periodic source images and to allow independent
source activation and evolution. The large length of the system is required for a
long-time shock propagation and to follow the long-term dislocation evolution.
The system was equilibrated at a temperature of 5 K before loading. The shock
front in the zero rise-time simulation develops a natural width of a few
0
0.6 0.5 0.4 0.3 0.2 0.1 0
z (μm)
Figure 4 Response of the MD lattice during 3D plastic relaxation. a,b, The
relaxation of the shear stress (a) and the corresponding lattice compression (b) from
simulated X-ray diffraction. Depth and time behind the shock front (bottom and top
scales) apply to both panels. The shock front is located at z = 0, with the piston
originally at z ∼ 0.55 μm. The black horizontal dashed line indicates the value for
full 3D compression. The black vertical dot–dashed line indicates the extent of our
previous simulations23 . In both panels, the red curves correspond to the 0 ps
rise-time shock case, and the blue curves to the ramped case. The curve for the
0 ps case includes diffraction for different ‘blocks’ of material at each z location. The
‘jitter’ of the curve indicates the magnitude of the error in our simulations, which is
∼0.2% well behind the shock front. We note that current experiments21 can
distinguish compression differences of 0.3%.
nanometres (subpicosecond rise time), with a volumetric compression of
∼15%. Shear stress is taken as [6−1/2 {σ zz − 0.5(σ xx + σ yy )}], as determined in
the region behind the shock front where it reached a spatially constant value
(except for a rise close to the ‘piston’).
Dislocations were identified using a centro-symmetry deviation (CSD)
filter31 . Atoms in a CSD range 0.075–0.25 were assumed to be part of
dislocation lines, allowing the calculation of dislocation densities. This method
only gives an upper limit of the dislocation density, and close comparison with
small volumes leads to dislocation densities that were generally lower by a
factor 2. ρd for a shock with a steep rise time can be estimated from the mean
separation between dislocation loops l ∼ a0 /ε v (ref. 22), where the stacking
fault spacing is equated with dislocation-loop separation, a0 is the lattice
constant and ε v = Up /Us is the volumetric strain. For our 35 GPa shock,

808 nature materials VOL 5 OCTOBER 2006 www.nature.com/naturematerials

©2006 Nature Publishing Group


ρmax ∼ l−2 ∼ 2 × 1013 cm−2 , which is consistent with our MD CSD analysis
(Fig. 2d). The dislocation density is also related to the flow stress in forest
1/2
hardening, τ = αGbρd , where the prefactor α = 0.3 for copper. G and b are
the shear modulus and Burgers vector for the compressed lattice. Using
ρd = 0.8 × 1013 cm−2 for the ramped case, and elastic moduli C11 = 168 GPa
and C12 = 121 GPa (the ambient values), we find τ = 0.39 GPa, in reasonable
agreement with the value from Fig. 4a. ρd is still 102 times larger than that
found in recovery experiments13 . This discrepancy may be attributed to several
causes not present in the MD shock simulation: (1) dislocation depletion from
tension waves due to the release at the free surfaces; (2) reduced dislocation
production due to rise-time, dislocation-source and strain-rate effects; and
(3) dislocation depletion due to long-time (μs–ms) thermal annealing of
recovered samples. For (1), we find in our simulations that ρd decreases by a
factor ∼3–5 within a few picoseconds of the shock hitting the back free surface,
and it is expected to decrease more at longer times after full release of the
compressed state.
Mobile dislocation densities and dislocation velocities were calculated as
follows. Given all atoms in dislocation lines in a certain region, a random atom
is chosen at time t , and the ‘persistence’ of that atom is checked in follow-up
snapshots of the dislocation lines. If the atom remains in the dislocation for a
time t , either the dislocation is fixed or it is moving at a velocity lower than
∼b/t . Repeating this procedure for a large number of dislocation atoms in a
given slice at different time steps, and using a sufficiently long t , a reasonable
estimate for the mobile dislocation density can be obtained. From the
distribution of ‘transit times’, an approximate velocity distribution for mobile
dislocations can also be inferred.
Simulated X-ray diffraction calculation allowed possible lattice rotations.
Corrections owing to large dislocation densities32 were evaluated using the
(200) and (400) peaks. Supplementary Information, Fig. S4 shows the
diffraction that would be recorded on film for the final lattice compression
in Fig. 4b.
Received 18 June 2006; accepted 10 August 2006; published 17 September 2006.
References
1. Itoh, M., Katayama, M. & Rainsberger, R. Computer simulation of a Boeing 747 passenger jet
crashing into a reinforced concrete wall. Mater. Sci. Forum 465/466, 73–78 (2004).
2. Acarer, M., Gulenc, B. & Findik, F. Investigation of explosive welding parameters and their effects on
microhardness and shear strength. Mater. Design 24, 659–664 (2003).
3. Cheng, W. L. & Itoh, S. High velocity impact of steel fragment on thick aluminum target. Mater. Sci.
Forum 465/466, 49–54 (2004).
4. Kumar, K. S., Singh, D. & Bhat, T. B. Studies on aluminum armour plates impacted by deformable
and non-deformable projectiles. Mater. Sci. Forum 465/466, 79–84 (2004).
5. Turtle, E. P. & Pierazzo, E. Thickness of a European ice shell from impact crater simulations. Science
294, 1326–1328 (2002).
6. Slavin, J. D., Jones, A. P. & Tielens, A. G. G. M. Shock processing of large grains in the interstellar
medium. Astrophys. J. 614, 796–806 (2004).
7. Haan, S. W. et al. Design and simulations of indirect drive ignition targets for NIF. Nuclear Fusion 44,
S171–S176 (2004).
8. Tokheim, R. E. et al. Hypervelocity shrapnel damage assessment in NIF target chamber. Int. J. Impact
Eng. 23, 933–944 (1999).
9. Smith, C. S. Metallographic studies of metals after explosive shock. Trans. Metal. Soc. AIME 212,
574–589 (1958).
nature materials VOL 5 OCTOBER 2006 www.nature.com/naturematerials
©2006 Nature Publishing Group
ARTICLES
10. Meyers, M. A. Mechanism for dislocation generation in shock-wave deformation. Scripta Met. 12,
21–26 (1978).
11. Meyers, M. A. et al. Constitutive description of dynamic deformation: physically-based mechanisms.
Mater. Sci. Eng. A 322, 194–216 (2002).
12. Gray, G. T., Hayes, D. B. & Hixson, R. S. Influence of the shock-induced alpha-epsilon transition in Fe
on its post-shock substructure evolution and mechanical behaviour. J. Phys. IV 10, 755–760 (2000).
13. Meyers, M. A. et al. Laser-induced shock compression of monocrystalline copper: characterization
and analysis. Acta Mater. 51, 1211–1228 (2003).
14. Vorthman, J. E. & Duvall, G. E. Dislocations in shocked and recovered LiF. J. Appl. Phys. 53,
3607–3615 (1982).
15. Johnson, Q., Mitchell, A. & Evans, L. X-ray diffraction evidence for crystalline order and isotropic
compression during the shock-wave process. Nature 231, 310–311 (1971).
16. Whitlock, R. R. & Wark, J. S. Orthogonal strains and onset of plasticity in shocked LiF crystals. Phys.
Rev. B 52, 8–11 (1995).
17. d’Almeida, T. & Gupta, Y. M. Real-time X-ray diffraction measurements of the phase transition in
KCl shocked along [100]. Phys. Rev. Lett. 85, 330–333 (2000).
18. Rigg, P. A. & Gupta, Y. M. Multiple X-ray diffraction to determine transverse and longitudinal lattice
deformation in shocked lithium flouride. Phys. Rev. B 63, 094112–094123 (2001).
19. Zaretsky, E. Multipeak pulse X-ray diffraction study of shocked single crystals. J. Appl. Phys. 93,
2496–2506 (2003).
20. Loveridge-Smith, A. et al. Anomalous elastic response of silicon to uniaxial shock compression on
nanosecond time scales. Phys. Rev. Lett. 86, 2349–2352 (2001).
21. Kalantar, D. H. et al. Direct observation of the α- transition in shock-compressed iron via
nanosecond X-ray diffraction. Phys. Rev. Lett. 95, 075502 (2005).
22. Holian, B. L. & Lomdahl, P. S. Plasticity induced by shock waves in nonequilibrium
molecular-dynamics simulations. Science 280, 2085–2088 (1998).
23. Rosolankova, K. et al. in Shock Compression of Condensed Matter-2003 (eds Furnish, M. D.,
Gupta, Y. M. & Forbes, J. W.) 1195–1198 (AIP, Melville, New York, 2004).
24. Bringa, E. M. et al. Ultrahigh strength in nanocrystalline materials under shock loading. Science 309,
1838–1841 (2005).
25. Hirth, J. P. & Lothe, J. Theory of Dislocations (Wiley, New York, 1982).
26. Swift, D. C. & Ackland, G. J. Quantum mechanical predictions of nonscalar equations of state and
nonmonotonic elastic stress-strain relations. Appl. Phys. Lett. 83, 1151–1153 (2003).
27. Gahagan, K. T. et al. Measurement of shock wave rise times in metal thin films. Phys. Rev. Lett. 85,
3205–3208 (2000).
28. Rose-Petruck, C. et al. Picosecond-milliångström lattice dynamics measured by ultrafast X-ray
diffraction. Nature 398, 310–312 (1999).
29. von der Linde, D & Sokolowski-Tinten, K. X-ray diffraction experiments with femtosecond time
resolution. J. Mod. Opt. 50, 683–694 (2003).
30. Mishin, Y. et al. Structural stability and lattice defects in copper: Ab initio, tight-binding, and
embedded-atom calculations. Phys. Rev. B 63, 224106–224121 (2001).
31. Kelchner, C. L., Plimpton, S. J. & Hamilton, J. C. Dislocation nucleation and defect structure during
surface indentation. Phys. Rev. B 58, 11085–11088 (1998).
32. Warren, B. E. X-ray Diffraction (Addison-Wesley, Menlo Park, California, 1969).
Acknowledgements
The authors would like to thank B. Sadigh for developing a faster version of MDCASK and for help
with the method to obtain mobile dislocation densities, G. Gilmer for help developing the code,
W. Cai and V. Bulatov and especially M. de Koning for help developing the initial dislocation sources,
and the MCR and Thunder computer teams for constant help and support. We would also like to
thank P. Erhart, M. Meyers, M. Schneider, J. McNaney, J. Colvin, J. Stölken, M. J. Caturla, B. D. Wirth
and M. Kumar for fruitful discussions and J. Aranibar, A. Caro, and W. J. Nellis for useful comments.
For the LLNL authors, this work was carried out under the auspices of the US Department of Energy
by the University of California, Lawrence Livermore National Laboratory under contract
No. W-7405-Eng-48.
Correspondence and requests for materials should be addressed to E.M.B.
Supplementary Information accompanies this paper on www.nature.com/naturematerials.
Competing financial interests
The authors declare that they have no competing financial interests.
Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/
809