You are on page 1of 13

metals

Article
Effect of Grain Boundary Misorientation on Spall Strength in
Ta via Shock-Free Simulations with Relatively Few Atoms
Jo Caulkins 1,2 , Carlisle Fauver 1 , Sara Adibi 3 and Justin Wilkerson 1, *

1 J. Mike Walker ’66 Department of Mechanical Engineering, Texas A & M University,


College Station, TX 77843, USA
2 Materials Science and Engineering Department, Carnegie Mellon University, Pittsburgh, PA 15213, USA
3 Mechanical Engineering Department, San Diego State University, San Diego, CA 92182, USA
* Correspondence: wilkerson@tamu.edu

Abstract: A suite of 37 molecular dynamics simulations is conducted at two system sizes to sys-
tematically characterize the role of grain boundary (GB) misorientation on spall strength in pure
BCC tantalum (Ta). The systems studied consist of bicrystals with a single [110] symmetric tilt grain
boundary. Two loading conditions are compared: (i) homogeneous extension under uniaxial strain
simulated in this study and (ii) piston/flyer impact of sample, which induces heterogeneous deforma-
tion via shockwave propagation along the length of the sample. The piston/flyer impact is taken from
the literature and run on the same set of GB misorientation angles using LAMMPS. The major finding
here is that both methods result in similar spall strength predictions, but the homogeneous extension
method generally requires two to three orders of magnitude fewer atoms and similar reductions in
computational costs. Spall strength results systematically overpredict using this method, by about
10% for the dataset three orders of magnitude smaller than piston/flyer simulations, and 5% for
the dataset two orders of magnitude smaller. Lastly, the effect of system size and pre-compression
Citation: Caulkins, J.; Fauver, C.; magnitude on spall strength is systematically characterized.
Adibi, S.; Wilkerson, J. Effect of Grain
Boundary Misorientation on Spall
Keywords: tantalum; molecular dynamics; spall strength
Strength in Ta via Shock-Free
Simulations with Relatively Few
Atoms. Metals 2022, 12, 1586.
https://doi.org/10.3390/
1. Introduction and Background
met12101586
Molecular dynamics (MD) is a powerful tool for understanding the role of microstruc-
Academic Editors: Alain Pasturel and
ture in the deformation and failure response of materials subject to extremely high strain-
Xing-Qiu Chen
rate loading. Such fundamental understanding is vital to many applications of national
Received: 5 July 2022 importance [1,2] such as armor, space shielding, and stockpile stewardship. Dynamic
Accepted: 19 September 2022 tensile strength under uniaxial strain, commonly termed spall strength [3], is particularly
Published: 23 September 2022 sensitive to microstructure and alloy content. These microstructure–spall strength relation-
ships are often non-intuitive and not necessarily simple extrapolations of the quasistatic
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
yield strength-microstructure relationships. For example, pure metals often have higher
published maps and institutional affil-
spall strength than alloys [4–6]. The spall strength of single crystals are generally a factor
iations.
of 2 or 3 higher than the spall strength of coarse-grained polycrystals [7]. Contradictory
to the well-known Hall-Petch relationship, most experimental grain size studies report
decreasing spall strength with decreasing grain size [8–12]. It has been argued that this
anomalous response is due to grain boundaries serving as particularly weak damage nucle-
Copyright: © 2022 by the authors. ation sites [8,13]. Generally, higher angle grain boundaries are more susceptible to damage
Licensee MDPI, Basel, Switzerland. nucleation than low-angle grain boundaries. That said, certain special high-angle grain
This article is an open access article boundaries, e.g., Σ3 and Σ9, are more resistant to void nucleation than some low-angle
distributed under the terms and grain boundaries [11,14]. The complexity of microstructure–spall strength relationships
conditions of the Creative Commons is further exacerbated in body center cubic (BCC) materials such as Ta, which exhibits
Attribution (CC BY) license (https:// two isochoric deformation modes: complex (e.g., non-Schmid effect [15]) slip systems and
creativecommons.org/licenses/by/ twinning [16].
4.0/).

Metals 2022, 12, 1586. https://doi.org/10.3390/met12101586 https://www.mdpi.com/journal/metals


Metals 2022, 12, 1586 2 of 13

The fundamental reason why grain boundaries are particularly detrimental to spall
strength is not yet fully understood. It is certainly possible that grain boundaries are inher-
ently weaker than grain interiors due to a greater degree of atomic disorder along grain
boundaries. Such arguments have been studied in the literature, but no clear correlations
between GB energy/volume and spall strength were observed in atomistic calculations of
pure Ta [17]. In alloys, heterogeneous segregation of impurity atoms to grain boundaries
may partially explain the apparent weakness of the grain boundaries. Alternatively, or
perhaps in combination, local stress concentrations are induced along grain boundaries,
which can drive premature failure at the grain boundaries. The stress concentrations are
governed by elastic and plastic incompatibilities of lattices on either side of the GB. These
effects have been systematically studied via crystal plasticity calculations, finding that
elastic and plastic stress concentrations are primarily affected by the tilt component of GB
misorientation with the twist component being relatively inconsequential [11,18–20]. MD
simulations have also found that both Cu and Ta systems fail more quickly on grain bound-
aries oriented normal to the load direction due to the higher resolved normal stress [21–23].
Strong interactions of dislocations and twins with grain boundaries, e.g., dislocation pile
ups and twin-GB intersections, may drive additional stress concentrations along grain
boundaries [21,24]. Some atomistic studies point to this mechanism being integral to the
propensity for intragranular failure in Ta polycrystals due to a high density of intersecting
slip systems and twins [25]. That said, the contribution of each of these mechanisms to GB
spall failure is not yet fully understood in pure or alloyed materials.
Normal plate impact tests are the most common configuration to experimentally
probe the spall behavior of materials [26]. Figure 1, reproduced from Chen et al., shows a
diagram of their flyer-plate simulations for reference [21]. Such tests consist of a flyer plate
being launched at high velocity into a target plate initially at rest. The impact generates
compressive shock waves that propagate through the thickness of the flyer and target
plates, shown in red dotted lines. These waves progress diagonally to show them moving
with time. A compressive wave (red) passes the GB before rarefaction at the free surface,
and later the rarefaction waves (blue) meet at the GB. By construction, the deformation is
one-dimensional with a macroscopically uniaxial strain state induced.
In this type of simulation, the aspect ratios, plate diameters normalized by plate
thicknesses, of both plates are typically greater than 5 to minimize contamination of the
one-dimensional uniaxial state by release waves emanating from the lateral surfaces. This
compressive shock typically induces plastic deformation and a temperature rise, both of
which can alter the material microstructure, e.g., dislocation density. Since spall strength
is sensitive to such shock-induced microstructural changes, spall strength unfortunately
becomes a poorly defined material property that is sensitive to the test conditions, e.g., shock
pressures, shock durations, shock rise times, and shock profiles. Often, simulations adopt
a similar test configuration by making use of flyer and target “plates” that are tens or
hundreds of nanometers in thickness. While this results in MD simulations that more
closely mimic the nature of the plate impact experiments, albeit with short length and time
scales [27], it also preserves the unfortunately complex dependence on test conditions that
cloud the interpretation of spall strength as a material property.
To overcome this shortcoming of plate impact MD simulations, here we advocate the
use of shock-free homogeneous deformation of MD samples. The homogeneous simu-
lations have two major advantages. Firstly, these simulations enable the de-coupling of
spall strength and shock-induced microstructural changes that depend on shock pressures,
shock durations, shock rise times, and shock profiles. This enables a more systematic
understanding of the role of microstructure on the fundamental dynamic tensile strength
of a material with known microstructure. Secondly, these simulations often require roughly
two to three orders of magnitude fewer atoms and similarly lower computational costs
than comparable plate impact MD simulations. The large system sizes required for plate
impact MD simulations are typically driven by the longitudinal lengths required to form
a quasi-steady shock profile. In addition, relatively long simulations times are required
may partially explain the apparent weakness of the grain boundaries. Alternatively, or
perhaps in combination, local stress concentrations are induced along grain boundaries,
which can drive premature failure at the grain boundaries. The stress concentrations are
governed by elastic and plastic incompatibilities of lattices on either side of the GB. These
Metals 2022, 12, 1586 effects have been systematically studied via crystal plasticity calculations, finding3 that of 13
elastic and plastic stress concentrations are primarily affected by the tilt component of GB
misorientation with the twist component being relatively inconsequential [11,18−20]. MD
simulations have also
for the shockwaves found thattoboth
to propagate the Cu
freeand Ta systems
surface fail more
of the target quickly
plate and on grain
reflections to
boundaries oriented normal to the load direction due to the higher
subsequently arrive at the spall plane. On the other hand, the system size for homoge- resolved normal stress
[21–23].
neous MD Strong interactions
calculations of dislocations
is driven entirely byand thetwins with grainwith
microstructure boundaries, e.g., possible
the smallest disloca-
tion pile ups and
representative twin-GB
volume intersections,
element may drive
that captures additional
all relevant stress concentrations
microstructural features along
being
grain boundaries
the lower bound on [21,24].
system Some atomistic
size. For many studies pointof
materials tointerests,
this mechanism being
e.g., single integral
crystals or
to the propensity
bicrystals of pureforor intragranular
solution strengthenedfailure inalloys,
Ta polycrystals
this lowerdue to a is
bound high density
quite small.ofItin-
is
tersecting slip systems and twins [25]. That said, the contribution
also reasonably small for some nanocrystalline materials [28,29]. Here, we will directly of each of these mecha-
nisms
compare to GB
ourspall failure
24,000 is not yet
and 200,000 atomfully understood MD
homogeneous in pure or alloyed
calculations materials.
with 20 million atom
plateNormal plate impact
impact simulations of tests
BCC Ta arepublished
the mostby common
Chen et configuration to experimentally
al. [21]. Both simulation methods
probe
result the spall behavior
in similar of materials
non-monotonic [26]. Figure
dependencies of GB1, reproduced
misorientation from onChen
spall et al., shows
strength. Thata
diagram
said, the of their flyer-plate
homogeneous MDsimulations
calculationsfor arereference [21]. less
significantly Suchcomputational
tests consist ofexpensive
a flyer plate
by
being
a factor of ∼ 10at
launched 3 for
high velocity
24,000 atoms intoand ∼ 102plate
a target initially However,
for 200,000. at rest. The theimpact
accuracygenerates
of the
compressive
results seemsshock to be waves
reduced that propagate
somewhat through
from the thickness
the smaller simulation of the flyer
size, and target
relative to the
shock simulation
plates, shown in red results.
dotted The datasets
lines. Theseare compared
waves progress in Table 1.
diagonally to show them moving
with The
time.structure of the paper
A compressive waveis(red)
as follows.
passesSection
the GB2beforedescribes the computational
rarefaction method-
at the free surface,
ology
and used
later thehere. Section waves
rarefaction 3 presents(blue)themeet
results
at and discusses
the GB. the comparison
By construction, between these
the deformation is
results and those with
one-dimensional fromaChen et al. [21]. Finally,
macroscopically uniaxial a summary
strain state is given
induced. in Section 4.

Figure 1.
Figure 1. Schematic
Schematic reproduced
reproducedfrom fromChen
Chenetetal.al.
showing their
showing simulation
their setup
simulation for reference
setup of flyer
for reference of
plateplate
flyer simulations [21]. This
simulations [21]. is notisthe
This notsetup of theofpresent
the setup simulations,
the present but the
simulations, butsetup of simulations
the setup of simu-
the present
lations simulations
the present are validated
simulations against. against.
are validated The dashedThe red linesred
dashed represent the compressive
lines represent wave,
the compres-
sive
and wave,
dashedand dashed
blue blue linesthe
lines represent represent the rarefaction
rarefaction wave. wave.

Table 1. Comparison of both present datasets with that of Chen et al. [21].

Homogenous Homogenous
Simulation Type Shock Flyer Plate
Deformation Deformation
Simulation Size 24,000 atoms 200,000 atoms 20,000,000 atoms
Computational time
30 min 4h 450 h
normalized to 1 core
Mean spall strength 19.62 GPa 18.58 GPa 17.33 GPa

2. Methods
Here, we simulate the spall failure of bicrystal BCC Ta containing a single tilt GB
aligned normal to the impacting projectile. Similar configurations have been adopted previ-
ously in crystal plasticity simulations, e.g., Nguyen et al. [18,19], as well as in experiments
on bicrystals and tri-crystals target plates, c.f. [30]. Our simulations show the behavior
Metals 2022, 12, 1586 4 of 13

of a single GB explicitly, without resolving the entire polycrystalline structure common


in spall experiments. Additionally, the GBs are modeled as perfectly bonded interfaces,
without the effect of local GB characteristics. All GBs in this work have a loading direction
parallel to the GB normal, which is most susceptible to failure [10,31,32]—up to an order
of magnitude more likely to nucleate a void than loading directions perpendicular to the
GB normal [17]. The orientation allows the probing of properties versus misorientation
Metals 2022, 12, x FOR PEER REVIEW 5 of 14
angle without needing to factor in the effects of the GB’s orientation with respect to the
loading direction, which was found to be sometimes significant [17,20]. A large benefit of
this method is allowing for reduced simulation size and computational time.
replicates the passage
We construct of a compressive
and study bicrystal BCCwave before spall
Ta specimens withand generates
a single defects
symmetric tilt in
GBthe in
otherwise perfect lattice, and (3) a tension stage that deforms
the middle of the specimen as shown in Figure 2. The GBs studied here are all coincident the material until failure to
measure its strength. The last two stages approximate
site lattice (CSL) boundaries. The GBs are generated using the MD simulator LAMMPS the loading conditions generated
by the passage
(Large-scale of a compressive shock
Atomistic/Molecular wave and
Massively subsequent
Parallel Simulator), tensile (spall) failure
specifically induced
by joining the
by release wave interactions, following the similar procedures
two crystals at the GB plane and deleting any overlapping atoms [33]. Grain boundaries are advocated by Nguyen et
al. [18,19].
visualized in OVITO. Two atoms are defined to be overlapping if they are within a specified
cutoffInradius
the minimization
of each other. stage,In initialtheruns,
unrelaxed
this radius GB was
formedvaried by from
joining0 totwo
2.1 Åcrystals
to test its in
LAMMPS is relaxed by giving each atom a random initial velocity
significance (for reference, Ta’s atomic radius is 2.2 Å). Cutoff radii that are too small leave corresponding to tem-
perature,
atoms 300 K.inside
partially Then,ofthe one simulation
another, whichprogresses
generatesat 300 K forhigh
a very 50 picoseconds (ps) using
initial compressive the
stress
NPT that
state ensemble (holding
perpetuates number (N)
throughout the of particles, giving
simulation, systemunrealistic
pressure (P) and temperature
results. Once the cutoff (T)
constant)
radius to relax
is large atoms
enough totoremove
equilibrium positions. Three
the overlapping atomsseeds
at thewere used toincreases
GB, further generate initial
make
velocities
no differenceto test stochasticity.
until the atomic The mean spall
separation strength
distance in thefrom the is
lattice three seedsatiswhich
reached, reported pointfor
each misorientation angle.
every atom would be deleted. The cutoff radius used here (1.64 Å) is large enough to avoid
After
artificial minimization,
initial compressive the stress
GB is compressed
states while in the z direction
remaining belowtothe 12.5 % uniaxial
atomic separation en-
gineeringThis
distance. strain over 25 ps
procedure to to approximate
generate GBs is the passagewith
consistent of athe~40methodology
GPa compressive adopted shock in
wave,etwhich
Chen al. [21].nucleates somewhat realistic crystallographic defects (e.g., dislocations) in
the grains. The compressive
Simulations for the set stage is carried were
of boundaries out atruna strain
at tworateboxof 5sizes
× 10to1/s while relax-
investigate the
ing with thesize
simulation NVE ensemble
effects. For the(holding
smaller number
set, the(N) of particles,
simulation boxvolume (V) andto
size is chosen energy
be 20 (E)re-
peated unit cells in
constant) at each timestep. the x and y directions, with 30 unit cells in the z direction which is the
direction
The of deformation.
final stage of theThat makes the
simulation total
is the size 6.6
tensile nm inwherein
portion, x and y,the andGB9.9isnm in z,along
pulled with
∼the24,z-axis
000 atoms (not exact
until failure using because
the same of edge
straineffects
rate as andinremoving
the compressiveoverlapping
stage.atoms) and
This repli-
periodic
cates theboundary conditions
tensile stresses in all three
experienced dimensions.
at the The other set
GB from rarefaction wavesof simulations
interactingwas run
in spall.
doubling
The pointthe of size along each
maximum stress dimension,
in this stage with is ∼ 200, 000 atoms.
indicative These
of the first simulation
void nucleation domain
since
sizes are substantially
the material will resolve smaller than the
the stress size adopted
through by Chen
deformation. Thus,et al.wewith
useits 15–20 million
ultimate tensile
atoms [21]. The interatomic potential used here is the Ta2
stress (under uniaxial strain conditions) as a proxy for cavitation strength. Data EAM (embedded atomfrommodel)the
potential
simulation byareRavelo
output et every
al. [34]. This is theThis
picosecond. same potential
limits used byofChen
the accuracy et al., and
the method agrees
relative to
well with
the rate ofmodel
tensilegeneralized
stress change, pseudopotential
since a void could theory calculations
nucleate between performed
outputs. The by Moriarty
accuracy
et
of al.
thefor pure
spall Ta systems
strength up to ~180in
measurements GPathis[21,34,35].
work are thus 0.5 GPa.

(a) (b) (c) (d)

Figure 2. Cont.
Metals 2022, 12, x1586
FOR PEER REVIEW 65 of
of 14
13

(e) (f) (g) (h)


Figure
Figure 2.
2. Progression
Progression of
of simulation
simulation on on an
an example
example GB
GB (misorientation
(misorientation angle
angle 80.63°)
80.63◦ ) to
to illustrate
illustrate the
the
process. Top left to top right (a–d): (a) Before relaxation, (b) after relaxation, (c) halfway compressed
process. Top left to top right (a–d): (a) Before relaxation, (b) after relaxation, (c) halfway compressed
(−6.25% strain), (d) fully compressed (−12.5% strain). Bottom left to bottom right (e–h): (e) After
(−6.25% strain), (d) fully compressed (−12.5% strain). Bottom left to bottom right (e–h): (e) After
some tension back to 0% strain for comparison with b, (f) after 9.17% strain in tension, (g) after
some tension
18.33% strainback to 0% strain
in tension, for comparison
(h) final with b,
state in tension (f) after
(27.5% 9.17%The
strain). strain
redinand
tension,
blue(g) after denote
atoms 18.33%
strain in grains
different tension,
in(h)
thefinal state in tension (27.5% strain). The red and blue atoms denote different
bicrystal.
grains in the bicrystal.
3. Results
Throughout deformation, a system level stress tensor is computed via a summation of
Analyzing how the stages of the simulations progress on a representative GB gives
the kinetic energy tensor and the virial tensor, i.e.,
insight about what exactly is happening in the simulations. From the 37 GBs explored in
this work, the GB that was selected
1 N for (this illustration
1 Nis the 80.63° misorientation angle
∑ ∑
(k) (k) (k) (k)
GB with 24,000 atoms shown m k)2vand
σI J =in Figures I v J3. + rI f J (1)
V k =1
V k =1

Here, N is the number of atoms, V is the system volume, and m(k) is the mass of
the k-th atom. The position, velocity, and force vectors associate with the k-th atom are,
(k) (k) (k)
respectively denoted as v I , r I and f I . Throughout this work the component of the
I = J = z component of the stress tensor, i.e., σzz , will be reported and analyzed. This
component of the stress tensor represents the axial normal stress aligned with the direction
of uniaxial deformation.
The simulation for each GB proceeds through three stages: (1) an initial minimization
stage that relaxes the atoms to minimum energy positions, (2) a compressive stage that repli-
cates the passage of a compressive wave before spall and generates defects in the otherwise
perfect lattice, and (3) a tension stage that deforms the material until failure to measure its
strength. The last two stages approximate the loading conditions generated by the passage
of a compressive shock wave and subsequent tensile (spall) failure induced by release wave
(a) (b)
interactions, following the similar procedures(c) (d)
advocated by Nguyen et al. [18,19].
In the minimization stage, the unrelaxed GB formed by joining two crystals in
LAMMPS is relaxed by giving each atom a random initial velocity corresponding to tem-
perature, 300 K. Then, the simulation progresses at 300 K for 50 picoseconds (ps) using the
NPT ensemble (holding number (N) of particles, system pressure (P) and temperature (T)
constant) to relax atoms to equilibrium positions. Three seeds were used to generate initial
velocities to test stochasticity. The mean spall strength from the three seeds is reported for
each misorientation angle.
After minimization, the GB is compressed in the z direction to −12.5 % uniaxial
engineering strain over 25 ps to approximate the passage of a ∼ 40 GPa compressive
shock wave, which nucleates somewhat realistic crystallographic defects (e.g., dislocations)
in the grains. The compressive stage is carried out at a strain rate of 5 × 109 1/s while
Metals 2022, 12, 1586 6 of 13

relaxing with the NVE ensemble (holding number (N) of particles, volume (V) and energy
(E) constant) at each timestep.
The final stage of the simulation is the tensile portion, wherein the GB is pulled
along the z-axis until failure using the same strain rate as in the compressive stage. This
replicates the tensile stresses experienced at the GB from rarefaction waves interacting in
spall. The point of maximum stress in this stage is indicative of the first void nucleation
since the material will resolve the stress through deformation. Thus, we use ultimate tensile
stress (under uniaxial strain conditions) as a proxy for cavitation strength. Data from the
simulation are output every picosecond. This limits the accuracy of the method relative to
the rate of tensile stress change, since a void could nucleate between outputs. The accuracy
of the spall strength measurements in this work are thus ± 0.5 GPa.

3. Results
Analyzing how the stages of the simulations progress on a representative GB gives
insight about what exactly is happening in the simulations. From the 37 GBs explored in
this work, the GB that was selected for this illustration is the 80.63◦ misorientation angle
GB with 24, 000 atoms shown in Figures 2 and 3.
The effects of the minimization stage can be seen by contrasting Figure 2a with Fig-
ures 2b and 3a with Figure 3b. Before minimization, there are pockets of empty space
between the grains, seen in white in Figures 2a and 3a. The coloring in Figure 3 denotes the
centrosymmetry parameter, which measures the degree of local lattice disorder, represent-
ing defects surrounding each atom. The blue regions are the closest to a perfect lattice, and
the red regions have the highest lattice disorder, representing defects. Centrosymmetry is
calculated using the following equation, with 8 nearest neighbors (N) for BCC [33].

{ N2 } → → 2
CS = ∑ R i + R i+ N
2
(2)
{ i =1}

Figures 2e and 3e show the system after the compressive stage has completed and
was additionally pulled in the tensile stage until the system returned to its initial size as
Figures 2b and 3b. Contrasting this with Figure 3b, colored regions of defects generated
from the deformation are visible that were not present previously in the system. This
shows the defects generated by the compressive stage, and why it is implemented in the
simulations. Because of this, these are plastic deformations that change the structure for the
tensile stage, which is the goal of the compressive stage. This can be seen in the hysteresis
between the compressive and tensile curves in the stress–strain diagram presented in
Figure 4, showing the energy lost to plastic deformation. Additionally, notable on the
stress–strain diagram in Figure 4 is a local minimum in tensile stress at about −10 % tensile
strain. The minimum in the curve represents the defect nucleation point.
The tensile portion of the simulation generates larger and larger tensile stresses until
failure, shown in Figure 4. This deformation grows the defects generated by the compres-
sive stage, with the colored portions in the bottom left and top right of Figure 3e turning
into the dislocations seen in Figure 3g. There is also more lattice disorder at the GBs in
Figure 3g, both in the center and the top and bottom due to periodic boundary conditions.
Eventually, this leads to the material failure shown in Figure 3h.
different grains in the bicrystal.

3. Results
Analyzing how the stages of the simulations progress on a representative GB gives
insight about what exactly is happening in the simulations. From the 37 GBs explored in
Metals 2022, 12, 1586 7 of 13
this work, the GB that was selected for this illustration is the 80.63° misorientation angle
GB with 24,000 atoms shown in Figures 2 and 3.

Metals 2022, 12, x FOR PEER REVIEW 7 of 14

(a) (b) (c) (d)

(e) (f) (g) (h)


Figure 3.
Figure 3. The
The same
same simulation
simulation steps
steps asas Figure
Figure 2,2, shown
shown color
color coded
coded byby centrosymmetry
centrosymmetry parameter,
parameter,
all with the same scale. Red has a higher centrosymmetry parameter, blue has lower. Top left to top
all with the same scale. Red has a higher centrosymmetry parameter, blue has lower. Top left to
right (a–d): (a) Before relaxation, (b) after relaxation, (c) halfway compressed (−6.25% strain), (d)
top right (a–d): (a) Before relaxation, (b) after relaxation, (c) halfway compressed (−6.25% strain),
fully compressed (−12.5% strain). Bottom left to bottom right (e–h): (e) After some tension back to
(d)
0% fully
straincompressed (−12.5%
for comparison withstrain). Bottom
b, (f) after 9.17%leftstrain
to bottom right (e–h):
in tension, (e) 18.33%
(g) after After some tension
strain back
in tension,
to 0% strain for comparison with b, (f) after 9.17% strain in tension, (g) after 18.33%
(h) final state in tension (27.5% strain). The line of high centrosymmetry atoms from the top leftstrain in tension,
(h) finaltostate
corner in tension
the GB (27.5%
in (e) shows strain). The
a dislocation line of high
generated centrosymmetry
during the compressive atoms from the top left
stage.
corner to the GB in (e) shows a dislocation generated during the compressive stage.
The effects of the minimization stage can be seen by contrasting Figure 2a with Figure
2b andThese steps3a
Figure were repeated
with Figurefor
3b.all 37 GBminimization,
Before misorientation angles examined
there are pocketsbyofChenemptyet al. [21].
space
The main advantage in the method used in this work is achieving a speedup
between the grains, seen in white in Figures 2a and 3a. The coloring in Figure 3 denotes in computa-
tional time for simulations.
the centrosymmetry The simulations
parameter, run in the
which measures the degree
presentof
work
localused about
lattice 24,000 rep-
disorder, and
200,000 atoms for 155 ps. Run on only one core, each smaller simulation took 30 min,
resenting defects surrounding each atom. The blue regions are the closest to a perfect lat- and the
larger size simulations took 3–4 h. Chen et al. used simulations with 15–20 million atoms
tice, and the red regions have the highest lattice disorder, representing defects. Centro-
for a total of 110 ps [21]. Run on 180 cores, they took 2.5 h per trial. The computational time
symmetry is calculated using the following equation, with 8 nearest neighbors (N) for BCC
decreases linearly with the number of cores up to 2000, so the speedup in this paper is on
[33].
the order of 103 and 102 for the two sets when compared on the same number of cores. The
present work aims to validate this model against the results from Chen et al. [21] by testing
cavitation strength vs. GB misorientation angle on the same GBs.
𝐶𝑆 = 𝑅⃗ + 𝑅⃗ (2)

Figures 2e and 3e show the system after the compressive stage has completed and
was additionally pulled in the tensile stage until the system returned to its initial size as
2b and 3b. Contrasting this with Figure 3b, colored regions of defects generated from the
deformation are visible that were not present previously in the system. This shows the
defects generated by the compressive stage, and why it is implemented in the simulations.
Because of this, these are plastic deformations that change the structure for the tensile
stage, which is the goal of the compressive stage. This can be seen in the hysteresis be-
tween the compressive and tensile curves in the stress–strain diagram presented in Figure
4, showing the energy lost to plastic deformation. Additionally, notable on the stress–
Metals 2022, 12, x FOR PEER REVIEW 8 of 14
Metals 2022, 12, 1586 8 of 13

Figure
Figure 4. Axial
4. Axial stress–strain
stress–strain curve
curve for for
the the example
example GB GB (misorientation
(misorientation angle
angle 80.63°)
80.63 during
◦ ) during com-
com-
pressive (dotted black) and tensile stages. The red line left of 0% strain is the tensile portion pulling
pressive (dotted black) and tensile stages. The red line left of 0% strain is the tensile portion pulling
the sample back to the initial length. This is only shown for the same boundary that compression is
the sample back to the initial length. This is only shown for the same boundary that compression is
shown with to be more legible. Three other randomly selected representative boundaries are shown
shown with
in the to bestage
tensile morefor
legible. Three other randomly selected representative boundaries are shown
comparison.
in the tensile stage for comparison.
The tensile portion of the simulation generates larger and larger tensile stresses until
Figure 5 shows the void nucleation stress vs. GB misorientation angle for the simula-
failure, shown in Figure 4. This deformation grows the defects generated by the compres-
tions carried out in this work and Chen et al., both with the same GBs [21]. Our results show
sive stage, with the colored portions in the bottom left and top right of Figure 3e turning
similar predictions for void nucleation stress through the misorientation space studied here,
into the dislocations seen in Figure 3g. There is also more lattice disorder at the GBs in
apart from a vertical shift in the data. Additionally, notable is a much better agreement
Figure 3g, both in the center and the top and bottom due to periodic boundary conditions.
between the two datasets in the 0◦ –125◦ region, with worse agreement at high angles,
Eventually, this leads to the material failure shown in Figure 3h.
125◦ –180◦ . We are not certain of the reason behind this, but a possibility is that at high
These steps were repeated for all 37 GB misorientation angles examined by Chen et
angles, there is more free volume at the GBs, and thus a more stochastic nature at the
al. [21]. The main advantage in the method used in this work is achieving a speedup in
sizes studied. The vertical shift in the results from our smaller dataset averaged 2.29 GPa
computational
higher than those of time
Chenfor et
simulations. The simulations
al., a difference of about 10%, run andin the
the results
presentfrom
work used
our about
larger
24,000 and 200,000 atoms for 155 ps. Run on only one core,
dataset averaged 1.25 GPa higher than the data of Chen et al., a difference of about 5% [21].each smaller simulation took
Two possible factors influencing this vertical shift were investigated and are discussed15–
30 min, and the larger size simulations took 3–4 h. Chen et al. used simulations with
20 million
shortly. atoms for
Additionally, thea dispersion
total of 110of psthe
[21]. Run
data on 180
points in cores,
this work theyistook 2.5 hthan
greater per that
trial.ofThe
computational time decreases linearly
the data from Chen et al. [21]. This is likely because with the number of cores up to 2000,
they varied local structure for each so the speedup
GB,
2
in this paper is on the order of 10
averaging two local structures per GB to obtain each data point. Our results came fromsame
and 10 for the two sets when compared on the a
number
single set ofof cores.
local The present
properties per GB, work aims
which to validatehas
intrinsically this model
more againstInthe
variance. results
future work,from
we Chen
hope to et al. [21] bythe
leverage testing
fastercavitation
simulation strength
time ofvs. thisGB misorientation
method to test spall angle on theresponse
strength same GBs.
to local GBFigure 5 shows the void nucleation stress vs. GB misorientation angle for the simula-
properties.
tions
Thecarried out in explanation
first possible this work and forChen et al., both
the vertical shiftwith
in voidthe same GBs [21].
nucleation stress Our results
is size
show similar predictions for void nucleation stress through
effects, because the two sets of simulations from this work were run with two and three the misorientation space stud-
ied here, apart from a vertical shift in the data. Additionally,
orders of magnitude fewer atoms than those of Chen et al. [21]. Finding the effects of notable is a much better
agreement
simulation sizebetween
was thethe two datasets
primary motivationin theof0°–125°
running region, with worse
simulations for allagreement
37 GBs at at two high
angles,sizes.
different 125°–180°. We are
Our results not24,000
with certain of theand
atoms reason behind
200,000 atoms this,
arebut a possibility
compared is that
in Figure 6. at
Thehigh angles,
larger thereagreed
dataset is moremorefree volume
closely withat theresults
GBs, and fromthusthe a more stochastic
flyer/plate nature at the
experiments
sizes studied.
performed by Chen Theetvertical
al., with shift in the
a mean results from
difference ourGPa
of 1.25 smaller datasettoaveraged
compared 2.29 GPa2.29 from GPa
ourhigher
smaller than those[21].
dataset of Chen
The et al., a with
trends difference of about 10%,
misorientation angleand the results
remained fromfor
similar our larger
both
sizedataset
datasets.averaged
To further 1.25 GPathe
study higher than
effects thesimulation
of the data of Chen size,etthe al.,size
a difference
of a single of aboutis5%
system
varied
[21].inTwo
orders of magnitude
possible factors by doubling the
influencing thissize in all shift
vertical three were
dimensions, whichand
investigated increases
are dis-
cussed shortly. Additionally, the dispersion of the data points in this work is greater than
Metals 2022, 12, x FOR PEER REVIEW 9 of 14

Metals 2022, 12, 1586 9 offor


that of the data from Chen et al. [21]. This is likely because they varied local structure 13

each GB, averaging two local structures per GB to obtain each data point. Our results came
from a single set of local properties per GB, which intrinsically has more variance. In fu-
the
turenumber
work, we of atoms
hope by a factor of
to leverage the8. faster
The results are shown
simulation time in
of Figure 7, which
this method contains
to test spall
more sizesresponse
strength than were studied
to local GBfor all 37 GBs.
properties.

Figure 5. Void nucleation stress vs. GB misorientation angle for BCC Ta bicrystal. Data shown from
Figure 5. Void nucleation stress vs. GB misorientation angle for BCC Ta bicrystal. Data shown from
Chen et al. [21] (red) and this study (blue). Error bars on our results show the standard error between
Chen et al. [21] (red) and this study (blue). Error bars on our results show the standard error between
2022, 12, x FOR PEER REVIEW 3 random seeds for initial velocity. The datapoints show the mean. 10 of 14
3 random seeds for initial velocity. The datapoints show the mean.
The first possible explanation for the vertical shift in void nucleation stress is size
effects, because the two sets of simulations from this work were run with two and three
orders of magnitude fewer atoms than those of Chen et al. [21]. Finding the effects of sim-
ulation size was the primary motivation of running simulations for all 37 GBs at two dif-
ferent sizes. Our results with 24,000 atoms and 200,000 atoms are compared in Figure 6.
The larger dataset agreed more closely with results from the flyer/plate experiments per-
formed by Chen et al., with a mean difference of 1.25 GPa compared to 2.29 GPa from our
smaller dataset [21]. The trends with misorientation angle remained similar for both size
datasets. To further study the effects of the simulation size, the size of a single system is
varied in orders of magnitude by doubling the size in all three dimensions, which in-
creases the number of atoms by a factor of 8. The results are shown in Figure 7, which
contains more sizes than were studied for all 37 GBs.

Figure 6. GraphFigure
comparing voidcomparing
6. Graph nucleationvoid
stress vs. GB misorientation
nucleation angle for BCC Ta
stress vs. GB misorientation bicrystal
angle for BCC Ta bicrystal
from our smaller (blue) and larger (yellow) datasets. Error bars on results from the 24,000
from our smaller (blue) and larger (yellow) datasets. Error bars on results from atom
the 24,000 atom
dataset show the standard error between 3 random seeds for initial velocity. The datapoints show
dataset show the standard error between 3 random seeds for initial velocity. The datapoints show
the mean.
the mean.
Figure 6. Graph comparing void nucleation stress vs. GB misorientation angle for BCC Ta bicrystal
Metals 2022, 12, 1586 from our smaller (blue) and larger (yellow) datasets. Error bars on results from the 24,000 atom 10 of 13
dataset show the standard error between 3 random seeds for initial velocity. The datapoints show
the mean.

Figure 7. PlotFigure
showing first showing
7. Plot void nucleation
first voidstress vs. number
nucleation stressofvs.
atoms in simulations
number of atoms in(simulation
simulations (simulation
size) for a BCC Ta bicrystal
size) for a BCCsystem with single
Ta bicrystal system GB (misorientation
with angle 80.63°). angle
single GB (misorientation left to◦ ).right:
From 80.63 From left to right:
3000 atoms; 24,000 atoms (smaller
3000 atoms; dataset
24,000 atoms size); 200,000
(smaller dataset atoms (larger dataset
size); 200,000 size); 1,600,000
atoms (larger atoms;
dataset size); 1,600,000 atoms;
~15,000,000 atoms (result from
~15,000,000 atomsChen et from
(result al. [21]).
Chen et al. [21]).

Figure 7 shows the 7void


Figure nucleation
shows the void strength
nucleationversus number
strength versus ofnumber
atoms in ofthe simula-
atoms in the simulation
tion for a BCC
forTaa bicrystal with misorientation
BCC Ta bicrystal angle of 80.63°.
with misorientation angleThe second
of 80.63 ◦ . The
datasecond
point fromdata point from
the left shows
theour
leftresult
showsfor this
our boundary
result for thiswith 24,000 with
boundary atoms, and the
24,000 atoms,thirdanddata
thepoint
third data point
shows our result
showswithour200,000. The200,000.
result with fifth data point
The fifthisdata
the result
point is from
the Chen
resultet al. for
from Chen theet al. for the
same GB [21]. The data point on the far left shows that running our simulation
same GB [21]. The data point on the far left shows that running our simulation with evenwith even
fewer atomsfewer
(3000)atoms
increases
(3000)the gap relative
increases the gaptorelative
the result from
to the Chen
result frometChen
al. [21].
et al.Con-
[21]. Conversely,
versely, as the sizesize
as the of the simulations
of the simulations increase, the the
increase, third andand
third fourth data
fourth point
data trend
point trend to- towards the
wards the result
resultfrom
fromChen
Chenetetal.al.[21].
[21].The
Themean
meandifference
differencebetween
betweenthe thevoid
voidnucleation
nucleation stress data
stress data presented
presentedhere with24,000
herewith 24,000 atoms
atoms and
and those
those of Chen et al. [21] was 2.29 GPa, GPa, whereas the
mean difference with 200,000 atoms was 1.25 GPa, so this could explain why there is a
vertical shift in our data. This suggests that our results might agree quite closely if the
simulations were performed on the same length scale.
Another size effect that was noted is that smaller system size increases the dependence
on initial conditions. Changing the seed in LAMMPS that determines initial random
atomic velocities introduces some stochasticity to smaller simulations. Void nucleation
stress results for the 24,000 atom dataset varied by up to 20% with changing seed for
one misorientation angle, though most differed by less than 5%. However, the average
void nucleation stress across the runs does not change, nor does the trend of dependence
with misorientation angle. To account for this, we report the mean spall strength for
each misorientation angle using three random seeds, and included standard error bars on
Figures 5 and 6, the figures that report on the 24,000 atom dataset.
To investigate another possible reason for the vertical shift in void nucleation stress,
the degree of compression during the second stage of the simulation was varied from 0 to
50 GPa in Figure 8, with linearly increasing compressive strain for each point. Because
each point is set at a fixed strain level, the stress levels are not evenly spaced. At 0 GPa of
compressive stress, the tensile portion of the simulation is trying to deform a perfect crystal,
which is harder to do than when defects are present. This is because deformation often
happens through the motion of dislocations, which requires dislocations to be formed and
to be moved. The stress to form new dislocations is higher than the stress required to move
them, so lattice that already contains defects will have lower strength. The 0–20 GPa runs
so they all had similar cavitation strengths.
The most interesting results come from the fourth and fifth datapoints, around 35
GPa in compressive stress. The slightly higher point has 25% less compressive strain, but
almost the same stress. That is because the compressive stress is partially resolved in the
Metals 2022, 12, 1586 material through deformation. Thus, we can see that defects are generated at around 35 11 of 13
GPa. After that point, the cavitation stress does not decrease further when compression is
increased to 50 GPa, reinforcing the importance of the 35 GPa range. This does not de-
scribe the difference
(the first between thepoints)
three data results, butnot
did it does
havegive some
enough insight
stress about what
to nucleate is hap-
defects in the material,
pening in thesocompressive stage.
they all had similar cavitation strengths.

Figure 8. FirstFigure
void nucleation stress
8. First void vs. compressive
nucleation stress vs.stress for a single
compressive GBfor
stress inaBCC TaGB
single bicrystal
in BCCwith
Ta bicrystal with
24,000 atoms (misorientation angle 80.63°).
24,000 atoms (misorientation angle 80.63 ).◦

The most interesting results come from the fourth and fifth datapoints, around 35 GPa
in compressive stress. The slightly higher point has 25% less compressive strain, but almost
the same stress. That is because the compressive stress is partially resolved in the material
through deformation. Thus, we can see that defects are generated at around 35 GPa. After
that point, the cavitation stress does not decrease further when compression is increased
to 50 GPa, reinforcing the importance of the 35 GPa range. This does not describe the
difference between the results, but it does give some insight about what is happening in
the compressive stage.

4. Summary and Conclusions


Molecular dynamics simulations were run in LAMMPS on a set of 37 BCC Ta bicrystal
grain boundaries to study the relationship of misorientation angle with spall strength. Each
bicrystal was relaxed to minimum energy positions using the NVE ensemble, and then
compressed to normal to the GB to 12.5% uniaxial engineering strain at a strain rate of
5 × 109 1 /s to generate somewhat realistic defects. The bicrystals were then pulled in uniax-
ial tension along the same axis to failure, using the same strain rate. The combination of the
compressive and tensile stages replicated the effects of a spall shockwave perpendicular
to the plane of the GB. Two datasets were analyzed with the present method, comprising
24,000 and 200,000 atoms. The results from both system size datasets were compared to sim-
ulations run by Chen et al. on the same set of GB misorientation angles using 15–20 million
atoms to investigate a potential speedup in MD spall simulations [21]. The comparison
was carried out in terms of spall strength, measured by first void nucleation stress.
Both present datasets give a reasonable representative value of spall strength when
compared to the results of Chen et al. across the misorientation space studied [21]. However,
there is a vertical shift in the present data compared to that of Chen et al., averaging a
spall strength 2.29 GPa higher for the 24,000 atom dataset and 1.25 GPa higher for the
200,000 atom dataset [21]. Comparison in runtime between the simulations shows that
the 24,000 atom dataset runs three orders of magnitude faster than the simulations by
Chen et al., and the 200,000 atom set runs two orders of magnitude faster [21].
There are several potential implications of a faster method. A faster simulation
method that has comparable accuracy at finding trends could scan larger parameter spaces
Metals 2022, 12, 1586 12 of 13

in sensitivity analyses, and the most interesting results of these analyses can then be
investigated for greater accuracy by methods such as the piston/flyer impact simulation
used by Chen et al. [21]. For example, with our dataset of 37 GBs, a three order of magnitude
speedup could add two additional dimensions to the study in the same amount of time.
Alternatively, the same feature space can be studied with a higher grid density. This could
be useful for highly nonlinear, jagged features.

Author Contributions: Conceptualization, J.W.; data curation, J.C.; formal analysis, J.C. and S.A.;
funding acquisition, J.W.; investigation, J.C.; methodology, J.C. and J.W.; project administration, J.W.;
resources, J.W.; software, J.C.; supervision, J.W.; validation, J.C., S.A. and J.W.; visualization, J.C.;
writing—original draft, J.C.; writing—review & editing, J.C., C.F., S.A. and J.W. All authors have read
and agreed to the published version of the manuscript.
Funding: Portions of this research were conducted with high performance research computing
resources provided by Texas A & M University (https://hprc.tamu.edu (accessed on 5 July 2022)).
This work was partially supported by the U.S. Department of Energy through the Los Alamos
National Laboratory. Los Alamos National Laboratory is operated by Triad National Security,
LLC, for the National Nuclear Security Administration of U.S. Department of Energy (Contract No.
89233218CNA000001). The authors are grateful for support from the Advanced Simulation and
Computing Program’s Physics and Engineering Models subprogram (Program Manager Manolo
Sherrill), in part under sub-contract 464745. The views and conclusions contained in this document
are those of the authors and should not be interpreted as representing the official policies, either
expressed or implied, of the U.S. Government. The U.S. Government is authorized to reproduce and
distribute reprints for Government purposes notwithstanding any copyright notation herein.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data from this study are openly available at: Caulkins, Jo (2022):
Working spreadsheet containing misorientation data. figshare. Dataset. https://doi.org/10.6084/m9
.figshare.21183649.v1. Caulkins, Jo (2022): Random seed analysis. figshare. Dataset. https://doi.org/10
.6084/m9.figshare.21183643.v1. Caulkins, Jo (2022): Varying Simulation Size, Compressive Stress, and
Overlap parameter. figshare. Dataset. https://doi.org/10.6084/m9.figshare.21183538.v1. Caulkins, Jo
(2022): Stress/Strain data. figshare. Dataset. https://doi.org/10.6084/m9.figshare.21183532.v1.
Acknowledgments: We would like to thank Jie Chen for sharing computational details, and for
helpful discussion about his simulations. We would like to thank all reviewers for taking the time
and effort necessary to review the manuscript. We sincerely appreciate all valuable comments and
suggestions, which helped us to improve the quality of the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Cardonne, S.M.; Kumar, P.; Mihcaluk, C.A.; Schwartz, H.D. Tantalum and its alloys. Int. J. Refract. Met. Hard Mater. 1995, 13,
187–191. [CrossRef]
2. Murr, L.; Shih, H.; Niou, C.-S. Dynamic recrystallization in detonating tantalum shaped charges: A mechanism for extreme plastic
deformation. Mater. Charact. 1994, 33, 65–74. [CrossRef]
3. Hahn, E.N.; Fensin, S.J.; Germann, T.C.; Gray, G.T. Orientation dependent spall strength of tantalum single crystals. Acta Mater.
2018, 159, 241–248. [CrossRef]
4. Nitol, M.S.; Adibi, S.; Barrett, C.D.; Wilkerson, J.W. Solid solution softening in dislocation-starved Mg–Al alloys. Mech. Mater.
2020, 150, 103588. [CrossRef]
5. Curran, D.R.; Seaman, L.; Shockey, D.A. Dynamic failure in solids. Phys. Today 1977, 30, 46–55. [CrossRef]
6. Pedrazas, N.A.; Worthington, D.L.; Dalton, D.A.; Sherek, P.A.; Steuck, S.P.; Quevedo, H.J.; Bernstein, A.C.; Taleff, E.M.; Ditmire, T.
Effects of microstructure and composition on spall fracture in aluminum. Mater. Sci. Eng. A 2012, 536, 117–123. [CrossRef]
7. Razorenov, S.V.; Kanel, G.I.; Herrmann, B.; Zaretsky, E.B.; Ivanchihina, G.E.; Elert, M.; Furnish, M.D.; Chau, R.; Holmes, N.;
Nguyen, J. Influence of nano-size inclusions on spall fracture of copper single crystals. AIP Conf. Proc. 2007, 955, 581. [CrossRef]
8. Minich, R.W.; Cazamias, J.U.; Kumar, M.; Schwartz, A.J. Effect of microstructural length scales on spall behavior of copper. Met.
Mater. Trans. A 2004, 35, 2663–2673. [CrossRef]
9. Peralta, P.; Digiacomo, S.; Hashemian, S.; Luo, S.-N.; Paisley, D.; Dickerson, R.; Loomis, E.; Byler, D.; McClellan, K.; D’Armas, H.
Characterization of Incipient Spall Damage in Shocked Copper Multicrystals. Int. J. Damage Mech. 2008, 18, 393–413. [CrossRef]
Metals 2022, 12, 1586 13 of 13

10. Wayne, L.; Krishnan, K.; DiGiacomo, S.; Kovvali, N.; Peralta, P.; Luo, S.; Greenfield, S.; Byler, D.; Paisley, D.; McClellan, K.
Statistics of weak grain boundaries for spall damage in polycrystalline copper. Scr. Mater. 2010, 63, 1065–1068. [CrossRef]
11. Escobedo, J.P.; Dennis-Koller, D.; Cerreta, E.K.; Patterson, B.M.; Bronkhorst, C.A.; Hansen, B.L.; Tonks, D.; Lebensohn, R.A. Effects
of grain size and boundary structure on the dynamic tensile response of copper. J. Appl. Phys. 2011, 110, 033513. [CrossRef]
12. Chen, T.; Jiang, Z.X.; Peng, H.; He, H.L.; Wang, L.L.; Wang, Y.G. Effect of Grain Size on the Spall Fracture Behaviour of Pure
Copper under Plate-Impact Loading. Strain 2015, 51, 190–197. [CrossRef]
13. Wilkerson, J.W.; Ramesh, K.T. Unraveling the Anomalous Grain Size Dependence of Cavitation. Phys. Rev. Lett. 2016, 117, 215503.
[CrossRef]
14. Mikhailovskij, I.M.; Mazilova, T.I.; Voyevodin, V.; Mazilov, A. Inherent strength of grain boundaries in tungsten. Phys. Rev. B
2011, 83, 134115. [CrossRef]
15. Cho, H.; Bronkhorst, C.A.; Mourad, H.M.; Mayeur, J.R.; Luscher, D. Anomalous plasticity of body-centered-cubic crystals with
non-Schmid effect. Int. J. Solids Struct. 2018, 139–140, 138–149. [CrossRef]
16. Chen, C.; Hu, G.; Florando, J.; Kumar, M.; Hemker, K.; Ramesh, K. Interplay of dislocation slip and deformation twinning in
tantalum at high strain rates. Scr. Mater. 2013, 69, 709–712. [CrossRef]
17. Fensin, S.J.; Valone, S.M.; Cerreta, E.K.; Escobedo-Diaz, J.P.; Gray, G.T.; Kang, K.; Wang, J. Effect of grain boundary structure
on plastic deformation during shock compression using molecular dynamics. Model. Simul. Mater. Sci. Eng. 2012, 21, 015011.
[CrossRef]
18. Nguyen, T.; Luscher, D.; Wilkerson, J. The role of elastic and plastic anisotropy in intergranular spall failure. Acta Mater. 2019, 168,
1–12. [CrossRef]
19. Nguyen, T.; Luscher, D.; Wilkerson, J. A dislocation-based crystal plasticity framework for dynamic ductile failure of single
crystals. J. Mech. Phys. Solids 2017, 108, 1–29. [CrossRef]
20. Krishnan, K.; Brown, A.; Wayne, L.; Vo, J.; Opie, S.; Lim, H.; Peralta, P.; Luo, S.-N.; Byler, D.; McClellan, K.J.; et al. Three-
Dimensional Characterization and Modeling of Microstructural Weak Links for Spall Damage in FCC Metals. Met. Mater. Trans.
A 2014, 46, 4527–4538. [CrossRef]
21. Chen, J.; Hahn, E.N.; Dongare, A.M.; Fensin, S.J. Understanding and predicting damage and failure at grain boundaries in BCC
Ta. J. Appl. Phys. 2019, 126, 165902. [CrossRef]
22. Hahn, E.N.; Fensin, S.J.; Germann, T.C. The role of grain boundary orientation on void nucleation in tantalum. AIP Conf. Proc.
2018, 1979, 050008. [CrossRef]
23. Fensin, S.J.; Escobedo-Diaz, J.P.; Brandl, C.; Cerreta, E.K.; Gray, G.T.; Germann, T.C.; Valone, S.M. Effect of loading direction on
grain boundary failure under shock loading. Acta Mater. 2014, 64, 113–122. [CrossRef]
24. Paul, S.K. Effect of twist boundary angle on deformation behavior of <100> FCC copper nanowires. Comput. Mater. Sci. 2018, 150,
24–32. [CrossRef]
25. Cheng, M.; Li, C.; Tang, M.X.; Lu, L.; Li, Z.; Luo, S.N. Intragranular void formation in shock-spalled tantalum: Mechanisms and
governing factors. Acta Mater. 2018, 148, 38–48. [CrossRef]
26. Antoun, T.; Seaman, L.; Curran, D.R.; Kanel, G.I.; Razorenov, S.V.; Utkin, A.V. Spall Fracture; Springer Science & Business Media:
Heidelberg, Germany, 2003.
27. Hahn, E.N.; Germann, T.C.; Ravelo, R.; Hammerberg, J.E.; Meyers, M.A. On the ultimate tensile strength of tantalum. Acta Mater.
2017, 126, 313–328. [CrossRef]
28. Bringa, E.M.; Traiviratana, S.; Meyers, M.A. Void initiation in fcc metals: Effect of loading orientation and nanocrystalline effects.
Acta Mater. 2010, 58, 4458–4477. [CrossRef]
29. Meyers, M.A.; Traiviratana, S.; Lubarda, V.A.; Benson, D.J.; Bringa, E.M. The role of dislocations in the growth of nanosized voids
in ductile failure of metals. JOM 2009, 61, 35–41. [CrossRef]
30. Perez-Bergquist, A.; Cerreta, E.K.; Trujillo, C.P.; Cao, F.; Gray, G., III. Orientation dependence of void formation and substructure
deformation in a spalled copper bicrystal. Scr. Mater. 2011, 65, 1069–1072. [CrossRef]
31. Weaver, J.S.; Jones, D.R.; Li, N.; Mara, N.; Fensin, S.; Gray, G.T. Quantifying heterogeneous deformation in grain boundary regions
on shock loaded tantalum using spherical and sharp tip nanoindentation. Mater. Sci. Eng. A 2018, 737, 373–382. [CrossRef]
32. Weaver, J.S.; Li, N.; Mara, N.A.; Jones, D.R.; Cho, H.; Bronkhorst, C.A.; Fensin, S.J.; Gray, G.T. Slip transmission of high angle
grain boundaries in body-centered cubic metals: Micropillar compression of pure Ta single and bi-crystals. Acta Mater. 2018, 156,
356–368. [CrossRef]
33. Plimpton, S. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J. Comput. Phys. 1995, 117, 1–19. [CrossRef]
34. Ravelo, R.; Germann, T.C.; Guerrero, O.; An, Q.; Holian, B.L. Shock-induced plasticity in tantalum single crystals: Interatomic
potentials and large-scale molecular-dynamics simulations. Phys. Rev. B 2013, 88, 134101. [CrossRef]
35. Moriarty, J.A.; Belak, J.F.; Rudd, R.E.; Söderlind, P.; Streitz, F.H.; Yanget, L.H. Quantum-based atomistic simulation of materials
properties in transition metals. J. Phys. Condens. Matter 2002, 14, 2825. [CrossRef]

You might also like