Download as pdf or txt
Download as pdf or txt
You are on page 1of 33

August 31, 2005 10:22 WSPC/IJCM-j050 00054

International Journal of Computational Methods


Vol. 2, No. 3 (2005) 341–373
c World Scientific Publishing Company

COMPUTATIONAL STUDY OF INTERFACE EFFECT ON


IMPACT LOAD SPREADING IN SiC MULTI-LAYERED TARGETS

S. K. DWIVEDI
Institute for Shock Physics, Washington State University
Pullman, WA 99164-2816, USA

J. L. DING∗
School of Mechanical and Materials Engineering
Washington State University
Pullman, WA 99164-2920, USA

Y. M. GUPTA
Institute for Shock Physics and Department of Physics
Washington State University, Pullman, WA 99164-2816, USA

Received 28 May 2004


Accepted 25 April 2005

The effect of finite strength interface with friction on lateral load spreading and phenom-
ena of interface crack initiation and propagation in bonded multi-layered targets under
high velocity impact are presented through axisymmetric finite element simulations. The
finite element code DYNA2D, developed at the Lawrence Livermore National Labora-
tory, was augmented with a phenomenological damage model for the silicon carbide
(SiC) ceramic and contact/cohesive interface model. Simulations were carried out for
four target configurations under varying interface strength (T m ), critical strain energy
release rate (Gc ), and inter-layer friction coefficient (µL ). It is shown that the high wave
speed SiC remains a potential material for enhanced load spreading if the confinement is
ensured to reduce/delay its damage. The load spreading also increases with the increase
in µL that comes into play after the interface failure. However, the µL of unity or more
needed to approach the upper bound of load spreading found in the perfectly bonded
target layers is not practical. It is shown that the resilience of the multi-layered targets
depends on T m as well as the Gc . But, the lateral load spreading depends dominantly on
T m and reaches the upper bound with its increase. It is further shown that the interface
cracks initiate and propagate in shear, mode II. The crack speed is invariably the maxi-
mum at initiation and is of the order of the longitudinal wave speed of materials on either
side. The maximum initiation speed of 11.29Cs and 7.79Cs are predicted at the SiC-
aluminum and steel-SiC interfaces for the frictionless case, where Cs is the shear wave
speed of the more compliant aluminum or steel. The crack speed reduces monotonously
after initiation, but it remains in the intersonic region for more than 150 ns. Whether
the propagating crack tip attains the steady state speed depends on the available driving

∗ Corresponding author. E-mail: ding@mme.wsu.edu, Fax: 509-335-4662.

341
August 31, 2005 10:22 WSPC/IJCM-j050 00054

342 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

energy. The steady state crack speed of 0.86Cs to 1.21Cs is predicted only at the steel-
SiC interface at 2 mm depth from the impact surface, lasts for more than 3 µs, is shown
to be independent of the interface strength upto 300 MPa, and is also independent of the
friction coefficient.

Keywords: Load spreading; SiC multi-layered targets; interface crack propagation; mul-
tiple crack initiation; intersonic crack speed.

1. Introduction
Layered structures have been a major contender for developing light-weight armor
systems. Readers interested in the past work on the response of multi-layered tar-
gets under high velocity impact can refer to Kinslow [1970], Zukas et al. [1982],
Ammann et al. [1987], Corbett et al. [1996] and references cited therein. In one of
the earliest studies, a light ceramic armor composite made by bonding a ceramic tile
to a metal backing plate was investigated by Wilkins [1978]. Since then, there have
been many studies to understand the relevant issues of interest: the deformation
behavior of ceramics to shock loading [e.g. Grady (1988, 1994); Raiser et al. (1994);
Chen and Ravichandran (1994); Feng et al. (1996a); Nesterenko et al. (1996); Grady
(1998); Feng et al. (1998); Sundaram and Clifton (1998); Yuan et al. (2001)], the
use of confined ceramics in multi-layered targets [e.g. Hauver et al. (1994); Brar
et al. (1997); Espinosa et al. (2000a)], the development of constitutive models to
describe deformation in the presence of damage during shock loading [e.g. Seamann
et al. (1976); Rajendran and Kroupa (1989); Rajendran (1994); Rajendran and
Grove (1996); Addesio and Johnson (1990); Curran et al. (1993); Holmquist et al.
(1995); Holmquist and Johnson (2002); Espinosa (1995); Espinosa et al. (1998a)],
finite element analyses to understand deformation behavior of ceramics [Camacho
and Ortiz (1996); Repetto et al. (2000)], and finite element analyses of multi-layered
ceramic targets [Espinosa et al. (1998b)]. Similar to ceramics, fiber reinforced com-
posite materials have been another potential material for weight efficient armor.
Extensive studies have evaluated their deformation characteristics and failure modes
under high velocity impact loading [e.g. Cantwell and Norton (1991); Abrate (1991,
1994); Potti and Sun (1997)]. However, multi-layered targets remain an active area
of research evident from Mahfuz et al. [2000] and Gama et al. [2001], all related to
the layered armors.
In a recent paper, Gupta and Ding [2002] proposed a potentially useful concept
to enhance the resilience of multi-layered structures to impulsive loading by using
a high wave speed layer to rapidly spread the impact load and energy in lateral
direction and reduce the offensive threat to the structure along the impact direction.
The feasibility of the load spreading concept was demonstrated through numerical
simulations with idealized material properties. No material damage was considered
and the layers were assumed to be perfectly bonded together. The authors also
proposed a quantitative measure based on normalized dissipative energy density
(NDE) to characterize the load spreading that provided the needed tool to examine
the features of the response of a composite target to impulsive loading. Using this
measure, Gupta and Ding [2002] studied the effects of wave speed, layer geometry,
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 343

and the mechanical properties of layer and substrate on load spreading in several
configurations of multi-layered target under normal impact loading.
With the feasibility of the load spreading concept demonstrated and a quantita-
tive measure developed, the effects of damage of the layering material and imperfect
interface were investigated in a subsequent study by Robbins et al. [2004]. The cor-
relation between load spreading and penetration resistance of the target was also
addressed. In this work, the interface was idealized as either a thin epoxy layer or
as sliding surfaces with friction (layers stacked together without bond) simulated
with slideline. In reality, the intact interface between bonded target layers has finite
strength in tension as well as in shear. If the load applied to interface due to impact
exceeds its strength, the interface fails and new free surfaces are generated between
the layers. The free surfaces subsequently interact as per their frictional behavior.
Hence, the interface between the target layers generally consists of regions of intact
interfaces and free surfaces during the target’s response to impact. The slideline
approximation of Robbins et al. [2004] represented the completely failed interface
from the instant of impact. It did not address the behavior of the interface prior to its
failure. Similarly, the epoxy layer approximated a weak interface with finite strength
between target layers, but did not address the failure initiation and the post-failure
behavior of the interface. The objectives of the current research were to study the
pre- (strength dominated) and post-failure (friction dominated) behavior of inter-
faces and their correlation with load spreading and to gain insight into the initiation
and propagation of interface failure in layered targets at high velocity impact.
Although the phenomenon of intersonic crack propagation at bi-material inter-
face and unidirectional fiber reinforced composite is well known [see for example
Rosakis et al. (1998); Huang et al. (1999); Yu et al. (2002); Dwivedi et al. (2003)
and references cited therein], the study of interface crack initiation and propagation
in multi-layered targets during high velocity impact has not been reported to the
best of our knowledge. Xu and Rosakis [2002] have shown that interface crack speed
in steel-homalite-steel sandwiched target is often intersonic under very low normal
impact velocity of 20–33 m/s. The propagating crack at the first interface kinks and
extends into the homalite and emerges on the second interface, thus suggesting inter-
laminar to intra-laminar and back to inter-laminar transformations. The mismatch
in materials elastic properties and stress wave speed responsible for intersonic crack
propagation exists naturally in the multi-layered targets. For example, the interface
between diamond or SiC ceramic and steel or aluminum simulated in the previous
studies [Gupta and Ding (2002) and Robbins et al. (2004)] present the same con-
ditions in which the metals are more compliant than the diamond or SiC. Wu and
Gupta [2000] experimentally found the crack speed on the Nb/Saphire interface to
be more than the dilational wave speed of the stiffer Saphire. One question we may
ask: Is the intersonic crack propagation possible along the interface in multi-layered
targets under naturally occurring material and impact loading conditions?
The above issues were addressed in the current study using numerical simula-
tions. The code used was DYNA2D [1992], developed at the Lawrence Livermore
National Laboratory. A phenomenological constitutive model for SiC and a contact/
August 31, 2005 10:22 WSPC/IJCM-j050 00054

344 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

cohesive interface model for analyzing finite strength interface between material lay-
ers were developed and implemented in DYNA2D. The models were validated with
the existing plate impact experimental data to establish their adequacy and param-
eter values. The responses of the targets with various configurations and interface
properties were studied in terms of the NDE measure, crack length, and crack speed.
The following sections give the description of the multi-layered targets and simula-
tion code DYNA2D. The next two sections briefly describe the material constitutive
models and analyses procedure used in this work. The results are discussed in Sec. 5
followed by the conclusions.

2. Multi-Layered Targets and Simulation Code DYNA2D


The schematic of the two-dimensional axi-symmetric finite element simulations car-
ried out in this work is shown in Fig. 1. The 26 mm thick 114 mm diameter target
plates were analyzed for the normal impact by a 12.5 mm diameter 4340 steel sphere
at 600 m/s impact velocity. Due to symmetry, only half diameters of the impactor
and target plate were simulated. All targets had a 6 mm thick skin layer (either sin-
gle or multi-layered) on top of the 20 mm thick substrate. The substrate material
was 6061 aluminum, and the skin materials were SiC ceramic as the high wave speed
material, 4340 steel, and 6061 aluminum. The four target configurations analyzed
are listed in Table 1. The nomenclature used in Table 1 follows that used in the
work by Gupta and Ding [2002] and Robbins et al. [2004].

Fig. 1. Schematic of the impactor and target configurations used in the simulations. All dimensions
are in mm.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 345

Table 1. Targets analyzed with different skin configurations.

HDP FDP Analysis


Target Layer 1 Layer 2 Layer 3 Time (µs) Time (µs) Time (µs)

0S 6.0 mm — — 2.5 4.0


aluminum
6S 6.0 mm SiC — — 2.0 3.6 7.5
2s-4S 2.0 mm steel 4.0 mm SiC — 2.2 3.8
2s-2S-2a 2.0 mm steel 2.0 mm SiC 2.0 mm 2.4 3.9
aluminum

DYNA2D [1992] used for simulations is a well known explicit transient dynamic
code based on the updated Lagrangian formulation. A brief description of the finite
element formulation of the code can be found in Goudreau and Hallquist [1982]. The
4-noded quadrilateral element of the code was used for the spatial discretization of
the multi-layered targets. The equation of motion solved in the explicit method of
shock propagation analysis in the absence of damping and external forces can be
written as

M t Ü t = −Fct − Fit , (1)

where M , Ü , and Fc are the globally assembled nodal lumped mass vector, nodal
acceleration vector, and internal force vector due to stresses in the continuum ele-
ments and the superscript ‘t’ refers to the time. Fi is the augmented nodal force
vector due to the cohesive interface tractions discussed later. The above equation
is integrated using mid-point algorithm with variable time step ∆t which gives the
deformed configuration at time t + ∆t and incremental deformation strains. The
Jaumann stress rate measure available in the code was used to calculate the stresses
in the updated configuration. DYNA2D has a number of built-in material models for
calculating the deviatoric and volumetric response of materials under dynamic load-
ing. Artificial viscosity for numerical stability under shock loading is also included
in the code. The internal nodal force vector is calculated by integrating the stresses
over the element using the one-point quadrature rule. The code provides different
options with variable coefficients for controlling the hourglass mode of deformation.
The optimized contact search algorithm together with different methods [Hallquist
et al. (1985)] for imposition of the contact impenetrability conditions are available
in the code needed for the impact simulations of the multi-layered targets.

3. Material Models
3.1. Aluminum and steel
The Johnson-Cook strength model [Johnson and Cook (1985)] and the equation of
state EOS1 [Goudreau and Hallquist (1982)] built-in into the DYNA2D were used
to simulate the deviatoric and volumetric responses respectively of aluminum and
August 31, 2005 10:22 WSPC/IJCM-j050 00054

346 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

steel. The Johnson-Cook strength model and its parameters for the two materi-
als [Johnson and Cook (1985); Espinosa et al. (1998b); Holmquist et al. (2001)]
are listed in Table 2. Due to their unavailability, the parameters of the EOS1
were obtained through approximation. The EOS1 is expressed as [Goudreau and
Hallquist (1982)]:

P = C0 + C1 µ + C2 µ̄2 + C3 µ3 + (C4 + C5 µ + C6 µ̄2 )Ev ; µ̄ = max(µ, 0), (2)

where µ = ρ/ρ0 − 1 with ρ0 and ρ being the initial and current mass density
respectively, C0 through C6 are the material constants, and Ev is the internal energy

Table 2. Material parameters for the Johnson/Cook strength model.


p
n m ε̄˙ T − Tr
σ̄ = [A + B ε̄p ][1 + C ln ε̇∗ ][1 − T ∗ ]; ε̇∗ = ; T∗ =
˙ε̄0 T m − Tr
X ∆ε̄p σm
f ∗ ∗ ∗
D= ; ε = [D 1 + D 2 exp D 3 σ ][1 + D 4 ln ε̇ ][1 + D 5 T ]; σ∗ = ≤ 1.5
εf σ̄
Properties Steel (s) Aluminum (a) Unit

Density ‘ρ0 ’ 7823.0 2704.0 kg/m3


Shear modulus ‘G’ 77.50 27.58 GPa
Initial yield stress ‘A’ 0.792 0.342 GPa
Strain hardening coefficient ‘B’ 0.51 0.1138 GPa
Strain hardening exponent ‘n’ 0.26 0.42 GPa
Strain rate dependence coefficient ‘C’ 0.014 0.002 —
Temperature dependence exponent ‘m’ 1.03 1.34 —
Melt temperature ‘Tm ’ 2066.0 1198.0 K
Room temperature ‘Tr ’ 303.0 303.0 K
Reference strain rate ‘ε̄˙ 0 ’ 1.0 1.0 1/s
Specific heat ‘Cv ’ 452.0 896.0 J/kg-K
Pressure cutoff ‘Pcut ’ 0.32 0.125 GPa
Failure parameters ‘D1 ’ −0.8 −0.77 —
‘D2 ’ 2.10 1.45 —
‘D3 ’ 0.50 0.47 —
‘D4 ’ 0.002 0.00 —
‘D5 ’ 0.61 1.60 —
Elastic longitudinal wave speed ‘Cl ’ 5881.0 6480.0 m/s
Elastic shear wave speed ‘Cs ’ 3147.0 3190.0 m/s
Elastic bulk wave speed ‘Cb ’ 4577.0 5327.0 m/s
Mie-Gruneisen equations of ‘K1 ’ 163.9 76.74 GPa
state parameters used to derive ‘K2 ’ 294.3 128.3 GPa
the EOS1 parameters ‘K3 ’ 500.0 125.1 GPa
‘γ0 ’ 1.16 2.0 —
EOS1 parameters ‘C1 ’ 163.9 76.74 GPa
‘C2 ’ 230.9 62.43 GPa
‘C3 ’ 288.9 27.07 GPa
‘C4 ’ 1.16 2.0 —
‘C5 ’ −0.99 −1.7 —
‘C6 ’ 0.43 0.74 —
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 347

per unit volume. The more commonly used Mie-Gruneisen equation of state [see for
example Rice et al. (1958)] is given as
 γµ  γ
P = (K1 µ + K2 µ2 + K3 µ3 ) 1 − + Em , (3)
2 V
where K1 through K3 are material constants, γ is the Gruneisen parameter assumed
as ρ0 γ0 = ργ with γ0 being the initial value of the Gruneisen parameter, V is the
specific volume, and Em is the internal energy per unit mass. The values of the
parameters K1 through K3 and γ0 were available for the two materials [Johnson
and Cook (1985); Espinosa et al. (1998b); Holmquist et al. (2001)] as listed in
Table 2. Neglecting the constant C0 , the constants C1 through C6 of Eq. (2) were
obtained by assuming one-to-one correspondence between Eq. (2) and Eq. (3), i.e.
 γµ 
C1 µ + C2 µ̄2 + C3 µ3 ≡ (K1 µ + K2 µ2 + K3 µ3 ) 1 − ,
2 (4)
C4 + C5 µ + C6 µ̄2 ≡ γ

with C1 set to be equal to K1 . The values of Ci ’s so obtained for aluminum and


steel are also listed in Table 2.

3.2. SiC ceramic


Development of material models generally follow two approaches, namely, microme-
chanical and phenomenological. The former provides more insights into the defor-
mation and fracture mechanisms and their correlation with macroscopic material
behavior, but is usually computationally intensive. For ceramics, efforts in this
aspect can be found in the works by Seaman et al. [1976], Rajendran and Kroupa
[1989], Rajendran [1994], Rajendran and Grove [1996], Addesio and Johnson [1990],
Curran et al. [1993], Espinosa et al. [1995; 1998a], Camacho and Ortiz [1996],
and Repetto et al. [2000] etc. The latter approach homogenizes the microstruc-
ture in terms of some continuum material parameters and is computationally effi-
cient. Efforts in this aspect can be found in the works by Ohtani and Chen [1989],
Holmquist et al. [1995], Holmquist and Johnson [2002] etc. The degradation in
strength and moduli, modeled in both approaches, affects the load transfer from
layer-to-layer and lateral load spreading and are more important for the present
study compared to the actual mechanisms causing them. Accordingly, computa-
tionally efficient phenomenological approach was considered appropriate for the
present research objectives.
Two different approaches were used to model the inelastic response of SiC under
compressive and tensile mean stress loadings. For the former, similar to Ohtani
and Chen [1988], Holmquist et al. [1995], and Holmquist and Johnson [2002], the
present model is based on the assumption that the equilibrium flow strength σ̄e

(with σ̄ e = 3J 2 = 3S ij Sij /2, Sij being the deviatoric stresses) of the material
under compressive mean stress reduces from the flow strength of the intact material
σ̄i to the flow strength of fully damaged material σ̄f as a linear function of the
August 31, 2005 10:22 WSPC/IJCM-j050 00054

348 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

accrued damage D, i.e. σ̄ e = (1 − D) σ̄i + Dσ̄f . D ranges from zero for the intact
material to unity for the fully damaged material. σ̄i and σ̄f are assumed to depend
on the compressive mean stress or confinement. The present model is based on the
overstress theory in which the stress state can lie outside the yield surface and
return to it within the characteristic relaxation time, assumed to be an inherent
material property. The in-elastic strain rate (ε̇pij ) is derived from the overstress flow
potential (Ω), i.e.
 i n+1
p ∂Ω σ̄ − σ̄ e
ε̇ij = ; Ω=A , (5)
∂Sij σ̄h
where, the superscript ‘i’ refers to the elastically predicted instantaneous value
and σ̄h , flow strength at the Hugoniot elastic limit (HEL), A and n are material
constants. The deviatoric stresses and effective plastic strain rate can be derived
from Eq. (5) as,
 
∆t p 2
i
Sij = Sij 1 − 2GBψ i ; ε̄˙ = Bψ, (6)
σ̄ 3
where, G is the shear modulus, B = [(σ̄ i − σ̄ e )/σ̄h ]n , and ψ = 1.5A(n + 1)/σ̄h. With
n = 1 and σ̄h /2Gψ = TR , Eq. (6) reduces to the overstress model of Feng et al.
[1996a] with TR being the relaxation time. Following Ohtani and Chen [1988], the
increase in damage under hydrostatic compression is a function of the mean stress
and inelastic strain energy (W), i.e.
 C
Ẇ σ̄
Ḋ = β ; β = β0 , (7)
Wf P + Pc
where, Ẇ = σij ε̇pij , β0 and C are material constants, P is the pressure (negative
of the mean stress), and P c = (1 − D) Pic is the cut-off pressure below which the
strength is zero with Pic being the cut-off pressure for the intact material. Wf in
Eq. (7) is the mean stress dependent inelastic energy needed to reach the failure
strength. The expression Wf = 0.01(σ̄i + σ̄f ) is used in the present work, which
assumes that at constant mean stress, the material strength reduces linearly from
intact to failure value incurring 2% of the inelastic strain. It is evident from Eq. (7)
that the rate of damage in the present model is directly proportional to the dissi-
pation rate and inversely proportional to the mean stress. The volumetric response
in compression is calculated as
∆P = −K∆εv + γ∆E v , (8)
where, ∆P is the pressure increment, ∆εv is the volumetric strain increment, and
K = (C 1 + 2C 2 µ)(1 + µ)2 is the instantaneous bulk modulus with C1 and C2 being
the experimentally determined material constants [see Feng et al. (1998) for more
details].
Under tensile mean stress loading, the material response is described by a micro-
cracking model following the work of Taylor et al. [1986], and Rajendran and Kroupa
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 349

[1989]. The tensile response is assumed to be elastic, but with the elastic moduli
degraded by damage. The damage introduced by tensile loading is defined in terms
of the crack density parameter Cd and degraded Poisson’s ratio ν, i.e.
 
16 1 − ν 2 16
Ḋ = Ċd ; ν = ν0 1 − Cd , (9)
9 1 − 2ν 9
where ν0 is the initial Poisson’s ratio. The crack density parameter is given as,
 2
5 KIc
Ċd = mk εm−1 ε̇v , (10)
2 ρC l ε̇max
v
v

where m and k are the material constants, KIc is the mode I fracture toughness,
Cl is the instantaneous longitudinal sound speed, and ε̇max
v is the maximum mean
strain rate applied to the material till the current time. The degraded bulk and
shear moduli in tension are:
3K (1 − 2ν)
K = (1 − D)C 1 ; G= . (11)
2 (1 + ν)

The incremental stresses ∆σij in tension are then calculated as ∆σ ij = 2G∆εij +


δij K∆εv , where ∆εij are the incremental deviatoric elastic strains and δij are the
Kronecker delta. Equations (7) and (9) were used to establish smooth transition
from the compressive to tensile damage state and vice-versa.
For determining model parameters, σ̄h was taken from Feng et al. [1998], the
damage exponent C was assumed unity, and damage parameter k was taken from
Rajendran and Kroupa [1989]. The remaining three parameters ψ, n, and β0 and
the parameter m were determined by analyzing three plate impact unloading exper-
iments [Yuan et al. (2003)] and the Grady’s spall experiment [Addesio and Johnson
(1990)]. The parameters so obtained are summarized in Table 3. The flow strengths
σ̄i and σ̄f of the intact and fully damaged material are shown in Fig. 2 as a
function of the mean stress. The equilibrium flow strength obtained experimen-
tally by Feng et al. [1998] is also shown in Fig. 2 for comparison. The compar-
ison between the simulated and in-situ particle velocity profiles measured with
the electromagnetic velocity (EMV) gauges at two locations in unloading exper-
iments for the case of 1076 m/s impact velocity is shown in Fig. 3. The two gauges
were at 3 mm and 8 mm from the impact surface in this normal symmetric (SiC
on SiC) plate impact experiments. Figure 4 shows the comparison of the simu-
lated back surface velocity profile with the Grady’s spall experiment at 1542 m/s
impact velocity (experimental data was reproduced from Addesio and Johnson
[1990]. Figures 3 and 4 show that the present model predictions are in reason-
able qualitative agreement with the experimental results and the model accuracy
was considered sufficient for our research objectives. The set of parameters listed in
Table 3 were used in the present study to simulate the continuum response of the
SiC layer.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

350 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Table 3. Materials parameters for the SiC phenomenological


damage model.

Density, ‘ρ0 ’ (kg/m3 ) 3200


Shear modulus, ‘G0 ’ (GPa) 192.3
Intact strength, ‘σ̄i ’ (GPa)
0.3904 + 1.561 P P ≤ 5.507
−2.834 + 2.732 P − 0.1063 P 2 5.507 < P ≤ 12.85
14.72 P > 12.85 GPa
Failure strength, ‘σ̄f ’ (GPa)
0.283 P P ≤ 5.507
−0.5841 + 0.495 P − 0.01927 P 2 5.507 < P ≤ 12.85
2.598 P > 12.85 GPa
Inelastic coefficient, ‘ψ’ 107
Inelastic exponent, ‘n’ 2.0
Damage coefficient, ‘β0 ’ 0.3
Damage exponent, ‘C’ 1.0
Poisons ratio, ‘ν0 ’ 0.16
Effective stress at HEL, ‘σ̄h ’ (GPa) 8.989
Mean stress at HEL, ‘Ph ’ (GPa) 5.507
Intact cut-off pressure, ‘Pic ’ (GPa) 0.25
Micro-cracking coefficient, ‘k’ 2.163*1022
Micro-cracking exponent, ‘m’ 2.0
Fracture toughness, ‘KIc ’ (MPa m1/2 ) 5.2
Equation of state constants, (GPa)
‘C1 ’ 220.0
‘C2 ’ 361.3
Gruneisen parameter, ‘γ0 ’ 1.50
Elastic longitudinal
wave speed, ‘Cl ’ (m/s) 12175
Elastic shear wave speed, ‘Cs ’ (m/s) 7735
Elastic bulk wave speed, ‘Cb ’ (m/s) 8292

Fig. 2. Effective stress of the intact and fully damaged material of the SiC phenomenological
damage model with equilibrium effective stress of Feng et al. [1998] shown for comparison.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 351

Fig. 3. Comparison of the simulated in-situ particle velocity profile with experimental data of
symmetric plate impact experiment at 1076 m/s impact velocity by Yuan et al. [2003].

Fig. 4. Comparison of the simulated back surface velocity with the Grady’s experimental data at
1542 m/s impact velocity taken from Addesio et al. [1990].

3.3. Cohesive interface element


The interfaces in the multi-layered targets were simulated using the contact/
cohesive methodology [Espinosa et al. (1996, 2000b); Camacho and Ortiz (1996),
Dwivedi et al. (2003)]. In this method, the zero-thickness interface elements are
embedded between layers in the initial unstressed state such that their nodes coin-
cide with the surface nodes of the continuum elements as shown in Fig. 5(a). The
cohesive law [Needleman (1987); Espinosa et al. (1996, 2000b); Camacho and Ortiz
(1996), Dwivedi et al. (2003)] describes the normal tensile and tangential tractions
August 31, 2005 10:22 WSPC/IJCM-j050 00054

352 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

(a) (b)

Fig. 5. Schematic of the (a) interface element in tension-shear at time t > 0 and (b) irreversible
interface cohesive law of the contact/cohesive methodology implemented in DYNA2D.

at the interface as a function of relative normal and tangential displacement jumps


given in terms of the normalized displacement jump, λ, defined as
   2
2
un ut
λ= + , (12)
δn δt
where un and ut are the instantaneous normal and tangential displacement jumps
shown in Fig. 5(a) and δn and δt are their maximum values at failure. The rate
independent irreversible cohesive law used in the present work is shown in Fig. 5(b).
The interface tractions are assumed to increase linearly from the unstressed state
and attain the maximum value T m at the critical value λ = λc . On further loading,
the interface tractions reduce linearly and vanish at λ = 1. At this state, the element
is deemed to have failed. The area under the triangular loading path gives the critical
strain energy release rate Gc of the interface. The unloading before attaining the
maximum traction takes place reversibly; i.e. the unloading path coincides with
the loading path. However, any unloading (say, from λ = λu ) after attaining the
maximum traction (λu > λc ) takes place irreversibly along the path joining the
current state (λu , T i ) to the initial unstressed state (0,0) as shown in Fig. 5(b).
The corresponding cohesive law can be written by setting λu = λc initially and
λu > λc in case of irreversible unloading:
 
 U  U 1−λ i
Tn = n Tni
 Tn = n
 T
λu λ 1 − λu n
λ ≤ λu λ > λu , (13)

 U 
 U 1−λ i
Tt = t Tti Tt = t T
λu λ 1 − λu t
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 353

where, Un = un /δn , Ut = ut /δt , Tni = Tnm (1 − λu )/(1 − λc ), and Tti = Ttm (1 − λu )/


(1 − λc ) with Tnm and Ttm being the maximum normal and tangential tractions.
The critical strain energy release rate GcI in tension (mode I) and GcII in shear
(mode II) can be written as
GcI = 0.5δn Tnm ; GcII = 0.5δt Ttm . (14)
The calculation of nodal forces from the interface tractions in the local coordinate
system (r, z) and transformation to global coordinate system (R, Z) to obtain Fi of
Eq. (1) follows the standard finite element procedure using the one-point integra-
tion rule.
The contact/cohesive methodology is summarized in Table 4. The computations
of the normal and tangential tractions depend on the loading and state of interface.
Before interface failure (λ < 1), the tangential traction is calculated by cohesive law
while the normal traction is calculated by cohesive law in tension and by contact
algorithm in compression. Both tractions are determined from the contact algorithm
after interface failure (λ ≥ 1). The implementation of the contact/cohesive method-
ology in DYNA2D was validated by simulating the compression-shear experimental
data of Connor [1988]. Figure 6 shows the comparison of the simulated voltage
record of the EMV gauges with experimental data for the case of 605 m/s impact
velocity. In this experiment, z-cut sapphire impacted on the fused silica target with
14◦ angle between the impact velocity and impact surface. The two gauges were
at 2.1 mm and 5.3 mm from the impact surface. The EMV gauges start record-
ing voltage at the arrival of the faster traveling compressive wave, attain the peak
value, remain constant until the arrival of the shear wave, and then start reducing
due to the opposite polarity of the longitudinal and lateral particle velocities. The
z-cut sapphire and fused silica were well characterized [Connor (1988); Feng and
Gupta (1996b)] and remained elastic at the impact velocity used for simulation. As
a result, the voltage components of EMV gauges due to the longitudinal and lat-
eral particle velocities become equal in magnitude. The voltage should go down to
zero due to the opposite polarity after the shear wave arrives at the interface. The
voltage not returning to zero is due to the interface failure and resulting decrease
of shear stress and associated lateral particle velocity. This is shown in Fig. 6. The
simulations were carried out at constant GcII = 1500 N-m/m2 and four different T m
values, viz., 100 MPa, 300 MPa, 500 MPa, and 1200 MPa. The reduction in the EMV

Table 4. Contact/cohesive methodology implemented in DYNA2D.

Interface traction calculation

Load Interface state Tn Tt

Tension-shear λ<1 Cohesive Cohesive


λ≥1 Contact Open

Compression-shear λ<1 Contact Cohesive


λ≥1 Contact Friction contact
August 31, 2005 10:22 WSPC/IJCM-j050 00054

354 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Fig. 6. EMV gauge voltage record in compression-shear experiment [Connor (1988)] compared
with the contact/cohesive simulation results at constant Gc = 1500 N-m/m2 and four T m values
(T 1 = 100 MPa, T 2 = 300 MPa, T 3 = 500 MPa, T 4 = 1200 MPa).

gauge voltage from the peak state increases as the shear wave magnitude transmit-
ted through the interface increases due to the increase in the interface strength.
The voltage reduces to zero at 1200 MPa interface strength as shown in Fig. 6. The
close agreement of the simulation result for T 2 with the experimental data shows
the interface strength T m = 300 MPa and mode II critical strain energy release
rate GcII = 1500 N-m/m2 in this experiment, and establishes the adequacy of the
contact/cohesive model implemented in DYNA2D.

4. Analyses Procedure
The augmented DYNA2D was used to study the effect of finite strength interface
and friction on the lateral load spreading in multi-layered targets and associated
phenomena of interface crack initiation and propagation. In the simulations, it was
assumed that (a) the value of maximum strength and critical strain energy release
rate in mode I and mode II were equal; i.e. Tnm = Ttm = T m and GcI = GcII = Gc ,
(b) all the interfaces in a target had the same T m , Gc , and friction coefficient irre-
spective of the participating materials on either side, (c) friction coefficient remains
constant irrespective of the relative sliding velocity, and (d) the interface crack
propagates along the interface without branching into the matrix. A number of
trial simulations were carried out first for the 6S target to select the set of parame-
ters in hourglass control, artificial viscosity, and contact algorithm. The final values
so determined and used in all the simulations are as follows: (a) The tolerance 10−6
relative to the mesh size, (b) the stiffness parameter k = 0.3 of the Flanagan and
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 355

Belytschko [1981] procedure of hourglass control, (c) linear artificial viscosity coef-
ficient Qb = 0.5 and quadratic artificial viscosity coefficient Qa = 4.0 [see Goudreau
and Hallquist (1982)], and (d) the scale factor fSI = 1.0 for the penalty approach
of imposing impenetrability condition [see Hallquist et al. (1985)].
The minimum values for the interface strength and critical strain energy release
rate were also determined from these trial simulations to obtain sufficient crack
length along the S-a interface. Care was taken so that the propagating interface
crack stopped before it could interact with the unloading wave from the boundary.
For example, based on the elastic bulk wave speed (Cb ), the unloading wave from
the bottom surface will arrive at the S-a interface after 8.23 µs, while the propagat-
ing interface crack stopped at 6.37 µs. The sufficient crack length was necessary to
obtain discernible effect of increasing the friction coefficient on the newly generated
sliding surfaces as discussed later. The friction coefficient between the impactor
and target surface µI and the friction coefficient between target layers µL were
assumed to be zero in the trial simulations. The minimum values of T m = 100 MPa
and Gc = 300 N-m/m2 were obtained from the trial simulations. The value of the
critical λc was chosen to avoid any artificial stress wave reflections at the interface
2
during reversible loading. λc = 6.0 ∗ 10−17T m /Gc so determined was used in all the
analyses.
Finally, detailed mesh convergence analyses were carried out for the same 6S
target using the minimum T m = 100 MPa and Gc = 300 N-m/m2 whose results are
not discussed. The finite element mesh chosen from these results and used in all the
analyses had an uniform square of 0.25 mm size in the first 36 mm radius, which
graded gradually to a rectangle of aspect ratio 2 at the boundary. A uniform mesh
size of 0.25 mm was used along the thickness, which gave 26 160 continuum elements,
26 802 to 27 284 nodes, and 240 to 720 interface elements. Consistent with the
Robbins et al. [2004], all analyses were carried out until the elapsed time of 7.5 µs.

4.1. Normalized dissipative energy


As mentioned earlier, NDE was a quantitative measure proposed by Gupta and
Ding [2002] to evaluate the load spreading capability of the layered targets and is
defined as
NDE = Ed /Es , (15)

where Ed is the dissipative energy per unit initial volume; i.e. Ed = (ρ/ρ0 ) σij dεpij ,
and Es is the total energy imparted to the substrate divided by the initial substrate
volume. The radial NDE distribution is calculated for the elements adjacent to
the top surface of the substrate, while the axial distribution is calculated for the
elements adjacent to the axis-of-symmetry. The combination of axial and radial
distributions gives an overall view of the directional partitioning of the energy in
the substrate and measures the effectiveness of the load spreading. It should be
mentioned that the dissipative energy refers to a sum of all the inelastic work
without considering the final form of the energy; e.g. heat or surface energy.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

356 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

The NDE was evaluated on consistent geometric and time bases. The geometric
consistency was maintained by keeping the skin layer and substrate thicknesses same
in all the simulations. The time consistency was maintained by calculating NDE
distribution when the shock front reached the half (HDP) and full (FDP) depth
of the substrate. The times of the HDP and FDP for the four targets are listed in
Table 1. Assuming V and H represent momentum transfer between the layer and the
substrate along the vertical (axial) and horizontal (radial) directions, respectively
(see Fig. 1), the load spreading concept is to enhance momentum transfer in the
horizontal direction (H) to permit a larger portion of the target to absorb the
impact energy. The span of radial and axial NDE distribution gives a measure
of H and V , respectively. However, H or V alone does not provide a complete
measure of load spreading. The gradient of the radial NDE is equally important
that indicates the relative dominance of V and H. A steeper gradient implies a
more localized dissipative energy distribution in the substrate (V is more dominant
than H). In general, effective load spreading leads to a lower peak value of NDE,
but the converse is not necessarily true. The peak value could be low for a target
with poor V regardless of its H. Because of the rapid decay of NDE along the axial
direction, the lower peak value at the layer–substrate interface, due to either more
effective load spreading or lower V , would typically result in a smoother axial NDE
distribution [see Gupta and Ding (2002); and Robbins et al. (2004) for detailed
description of NDE].

4.2. Crack length and crack speed


The total crack length and crack speed were used to study the phenomena of inter-
face crack initiation and propagation. The interface crack was considered to have
initiated at a site if the interface element at that position failed first and the ele-
ments on its either side failed consecutively in the increasing or reducing radial
directions. The crack propagating in the positive radial direction is termed as for-
ward propagating crack, and the crack propagating in the negative radial direction
is termed as the backward propagating crack. The single or multiple initiation site,
initiation time, and temporal variation of the crack length were obtained from the
simulation results. The speeds of the forward and backward propagating crack tips
were calculated from the time derivative of the temporal variation using the central
difference scheme.

5. Results and Discussion


5.1. Load spreading in the presence of damage to SiC
and cohesive interface
The load spreading in the targets for the minimum values of the interface param-
eters: T m = 100 MPa, Gc = 300 N-m/m2 and µI = µL = 0.0 are shown in Fig. 7.
The results for the 0S and 6S targets agreed well with the previous studies. The
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 357

Fig. 7. Increase in the load spreading by SiC high wave speed layer (0S versus 6S) and further
increase due to the confinement by the steel layer (6S versus 2s-4S and 6S versus 2s-2S-2a) shown
at HDP (left) and FDP (right). (T m = 100 MPa, Gc = 300 N-m/m2 , µI = µL = 0.0)

load spreading is the minimum for the 0S target shown by the maximum V /H ratio
due to the lower bulk wave speed of the aluminum skin compared to the other skin
materials. A substantial increase in the load spreading occurs for the 6S target com-
pared to the 0S target. However, 6S load spreading is lower compared to the case
simulated by Robbins et al. [2004] without damage to the SiC and perfectly bonded
interface. The SiC layer undergoes rapid damage in the radial direction due to the
release of confinement by the rarefaction wave. The horizontal momentum transfer
(H) is significantly reduced due to the resulting strength and moduli degradations
and also due to the interface failure. The vertical momentum transfer (V ) to the
substrate in the axial direction remains largely unaffected due to the attenuation
of the rarefaction wave through the comminuted material that helps in maintaining
sustained confinement of the SiC directly under the impactor. The resultant higher
August 31, 2005 10:22 WSPC/IJCM-j050 00054

358 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

V /H ratio gives a steeper gradient and a higher peak value of NDEs for the target.
Similar results were also reported by Robbins et al. [2004] who simulated the inter-
face as an epoxy layer. However, in spite of the damage and interface failure, the
V /H ratio of the 6S target is considerably higher than the ratio for the 0S target
showing the enhanced load spreading by the high wave speed SiC layer.
As also shown in Fig. 7, load spreading in the 2s-4S target is enhanced compared
to the 6S and 0S targets. The compressive axial stress in the top 2 mm thick high
impedance steel layer delays the effect of the rarefaction wave and ensures con-
finement of the SiC layer over larger radial span. The damage to the SiC reduces
in the radial direction increasing the horizontal momentum transfer (H) and the
radial load span on the substrate seen in the zoomed figure. The vertical momen-
tum transfer (V ) also increases as a result, but the V /H ratio reduces compared to
the 6S and 0S targets. The more uniform load spreading, as a result, is seen by the
reduced peak NDEs and reduced slope of the radial NDE. The load spreading in the
2s-2S-2a target is lower compared to the 2s-4S target due to the loss in confinement
from the unloading wave reflected off the SiC-aluminum (S-a) interface. However,
the peak NDEs values and gradient of the radial NDEs are unaffected due to the
better impedance match at the layer-substrate interface (a-a compared to S-a).
The above results show that the high wave speed SiC layer remains a potential
material for the load spreading if the confinement is ensured to reduce or delay its
damage. The 2s-4S target has more load spreading and 2s-2S-2a target has similar
load spreading compared to the 6S target, even though the thickness of the SiC
layer is reduced from 6 mm to 4 mm and to 2 mm respectively.

5.2. Cohesive interface failure process


The phenomena of interface crack initiation and propagation simulated in the above
analyses of targets with T m = 100 MPa, Gc = 300 N-m/m2 and µI = µL = 0.0 are
discussed. The spherical impactor generates a point-source spherical wave at the
instant of impact, which expands into the intact material layers along the radial
and axial directions. As it propagates, the wave is attenuated due to the wave
divergence and the energy dissipation due to the material’s inelastic deformation
and damage. The rarefaction wave generated at the impact surface trails behind the
primary loading wave. As the sphere comes gradually into contact with the target
surface, subsequent compressive stress waves and associated rarefaction waves are
generated at the contact surface. These secondary stress waves have diminishing
magnitudes and travel through the inelastically deformed or damaged materials.
The expanding stress wave loads the interface between material layers in com-
pression and shear. The shear load is generated at the interfaces by the oblique
impact of the stress wave and differences in the mechanical properties and the stress
wave speed of materials on either side of the interface. As a result, the interface crack
initiates in shear; i.e. pure mode-II, and propagates in the forward and backward
directions. This is shown in Fig. 8 by the loading on the forward propagating crack
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 359

Fig. 8. Interface crack tip loading shown along the s-S interface (elements 11, 21, 31 and 41)
and a-a interface (element 531) in the 2s-2S-2a target. (T m = 100 MPa, Gc = 300 N-m/m2 ,
µI = µL = 0.0.)

tip along the s-S and a-a interfaces in the 2s-2S-2a target. The interface elements
11, 21, 31, and 41 were on the s-S interface along which the crack initiated at ele-
ment 5 and stopped at element 44. The center of the four elements was at 1.5 mm,
4.00 mm, 6.51 mm, and 9.02 mm respectively from the center of element 5. A small
magnitude of tensile traction develops during shear reversal, but the failure of the
interface takes place in shear mode II. The maximum loading rate for element 11
next to the initiation site and the minimum for element 41 near the end of the crack
propagation show that the crack propagation stopped when the driving stress wave
magnitude could no longer provide sufficient energy to drive the propagating crack
tip. The interface element 531 in Fig. 8 was located at a-a interface along which the
crack initiated at element 492 and stopped at element 535. The center of element
531 was at 9.8 mm from the center of element 492. The crack initiates and is forced
to propagate along the interface under pure mode-II loading as in the case of s-S
interface. However, the appreciable normal traction and change in the loading from
the pure mode II to the mixed mode during the failure of the element 531 is due
to the unloading wave that was generated at the newly created free surface at S-a
interface. This release of the crack tip shear stress field by the unloading wave and
resulting crack arrest was observed only at the a-a interface of the 2s-2S-2a target.
Figure 9 shows the variation of crack length with time at different interfaces in
the 6S, 2s-4S, and 2s-2S-2a targets. The backward propagating crack tip stops before
reaching the center line in all the cases and is not discussed. The total forward crack
length along S-a interface in the 6S target is 8.8 mm. The increased confinement and
reduced energy dissipation in the SiC layer of the 2s-4S target increases the crack
August 31, 2005 10:22 WSPC/IJCM-j050 00054

360 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Fig. 9. The interface crack length increase with time shown along the interfaces in 6S, 2s-4S (left)
and 2s-2S-2a (right) targets. The arrows show initiation time and radial position of initiation site.
(T m = 100 MPa, Gc = 300 N-m/m2 , µI = µL = 0.0)

length marginally to 9.0 mm at the S-a interface at the same 6 mm depth from the
impact surface. However, the crack length along the s-S interface in the 2s-4S and
2s-2S-2a targets are considerably different. The time required by the unloading wave
from S-a interface to reach the s-S interface is half for the 2s-2S-2a target compared
to the 2s-4S target. As a result, the total forward crack length along the s-S interface
is reduced from 16.5 mm for the 2s-4S target to 9.8 mm for the 2s-2S-2a target.
Figure 9 also shows the multiple crack initiation and coalescence along the S-a
interface in the 2s-2S-2a target with the lengths of the forward cracks being 4.5 mm
and 5.8 mm from the first and second initiation sites respectively. The simulation
showed that the stress wave in the steel layer generates a precursor wave in the
SiC layer at the failed s-S interface. The precursor wave precedes the primary wave
in the radial direction. It loads and initiates the second crack at the S-a interface
ahead of the first crack tip. This is shown in Fig. 10 by the radial stress contours at
three time instances: (i) When the pre-cursor wave reached the S-a interface, (ii) at
one intermediate stage, and (iii) shortly after the initiation of the second crack. The
newly initiated crack tip also travels in the forward and backward directions. The
forward traveling first crack tip and the backward traveling second crack tip finally
coalesce just before stopping, manifested by the almost vanishing crack speed at
the time of coalescence.
Figure 11 shows the crack speeds along the interfaces shown in Fig. 9. The crack
speed is invariably the maximum at the time of initiation in all the simulations and
the initiation speed is comparable to the longitudinal wave speed of the materials
on either side of the interface. The initiation speed at the s-S interface in the 2s-4S
target was 24.5 km/s which is 7.79Cs , 5.35Cb , and 4.17Cl of the compliant material
steel. The minimum initiation speed was 11.85 km/s for the second crack along
S-a interface of the 2s-2S-2a target which is 3.71Cs , 2.22Cb , and 1.83Cl of the
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 361

Fig. 10. Contours of the radial stress (GPa) at three time instances showing the pre-cursor wave in
the SiC layer that initiates the second crack at S-a interface in the 2s-2S-2a target shortly before
3.45 µs. (T m = 100 MPa, Gc = 300 N-m/m2 , µI = µL = 0.0)

more compliant aluminum. The maximum initiation speed 36 km/s was obtained
for the S-a interface of the 2s-4S target which is 2.96Cl of the stiffer SiC and
11.29Cs , 6.76Cb , and 5.55Cl of the more compliant aluminum. The increase in
longitudinal wave speed on compression under confinement is one possible reason
for the initiation speed being more than the elastic longitudinal wave speed of the
August 31, 2005 10:22 WSPC/IJCM-j050 00054

362 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Fig. 11. Interface crack speed at the interfaces of Fig. 9 determined from the temporal crack length
variation. (T m = 100 MPa, Gc = 300 N-m/m2 , µI = µL = 0.0)

SiC layer. However, the crack speed reduces monotonously after initiation to a value
less than the shear wave speed of the more compliant material except for the s-S
interface in the 2s-4S target. The propagating crack tip does not attain steady speed
in most of the cases. The simulations predicted steady crack speed only along the
s-S interface. The crack propagates along this interface at a steady speed of 2.7 km/s
(0.86C s steel) for 1.05 µs in the 2s-2S-2a target and at 3.8 km/s (1.21C s steel) for
4.2 µs in the 2s-4S target.
The above results show that the interface crack initiates and propagates in
shear, mode II. The crack speed is invariably the maximum at initiation followed
by de-acceleration. This is in contrast to the phenomenon of crack initiation at
speed below the shear wave speed of the more compliant Homalite followed by
acceleration recorded experimentally by Xu and Rosakis [2002]. The crack speed
remain in the intersonic/transonic region for a measurable time period. But, the
steady state intersonic crack speed is predicted only at the s-S interface at 2 mm
depth from the impact surface.

5.3. Effect of interface failure on load spreading


under increasing friction
The 6S, 2s-4S, and 2s-2S-2a targets were simulated for two additional µL values of
0.1 and 1.0 keeping the interface parameters same as used above: T m = 100 MPa,
Gc = 300 N-m/m2 . Two different µI values of 0.0 and 0.3 between the impactor
and target surface were also considered. However, no discernible effect of µI on load
spreading was found. The results presented below are for the case of µI = 0.0.
Figure 12 shows the load spreading at FDP for the case of 2s-4S target with
increasing friction coefficient along the free surfaces generated after interface fail-
ure. The load spreading in the same target for the perfect bond condition used by
Gupta and Ding [2002] and for the slideline condition used by Robbins et al. [2004]
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 363

Fig. 12. Increase in the load spreading with increase in the friction coefficient shown for the 2s-4S
target simulated with T m = 100 MPa, Gc = 300 N-m/m2 , µI = 0.0. Results obtained from perfect
bond condition of Gupta and Ding [2002] and the frictional slideline condition used by Robins
et al. [2003] are also shown for comparison purpose.

are also shown in Fig. 12. The increase in friction coefficient results in a larger span,
higher peak value, and steeper gradient in radial NDE. The load transmission to the
substrate becomes more effective due to the enhancement of the tangential traction
with increase in the friction after interface failure. However, because the tangen-
tial traction is related to normal traction through the friction coefficient, the load
transmitted will be higher around the center where the normal traction is greater.
This normal traction dependence results in an increased gradient and peak value
with the increasing friction coefficient. In agreement with the previous studies for
other target configurations, the load spreading in the 2s-4S target is the maximum
August 31, 2005 10:22 WSPC/IJCM-j050 00054

364 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

with the perfect bond condition and shows the upper bound. The load spreading
increases with friction coefficient and reaches the upper bound for µL = 1.0.
The load spreading with the slideline interface is similar to those with the cohe-
sive interface for the minimum values of the interface parameters used in the sim-
ulations. The marginal difference in the load spreading between the cohesive and
slideline treatments is more appreciable around the area of impact. The difference
reduces as the friction coefficient increases. At low friction, the tangential traction
carried by the friction is smaller than that by the finite strength of the interface
resulting in better load spreading by the cohesive interface. As the frictional coef-
ficient increases, the traction carried by the friction is comparable to that by the
cohesive interface. As the friction coefficient reaches 1, the peak value of NDE from
the slideline interface becomes larger than that from the cohesive interface.
Figure 13 shows the effect of increase in friction coefficient on the interface
crack initiation and propagation. As shown on the top, the increase in the friction

Fig. 13. Effect of increase in the friction coefficient on the interface crack length and crack speed
shown for the s-S interface of the 2s-4S target (top) and S-a interface of the 6S target (bottom)
simulated with T m = 100 MPa, Gc = 300 N-m/m2 , µI = 0.0.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 365

coefficient had no appreciable effect on the initiation speed, total crack length, final
steady propagation speed and duration of the steady state crack propagation along
the s-S interface of the 2s-4S target. For µL = 1.0, the crack speed reduced due
to the increased energy dissipation during the shear reversal and then increased
to 2.18Cs of the more compliant steel before reducing back to the steady speed
of 3.8 km/s. Similar phenomenon was observed experimentally by Xu and Rosakis
[2002]. The most discernible effect of the friction coefficient was obtained along the
interfaces at 6 mm depth from the impact surface. This is shown at the bottom of
Fig. 13 for the S-a interface in the 6S target. The increase in the friction coefficient
reduces the crack tip loading due to the increases in energy dissipation by friction
and local damage to the SiC layer near the interface. The phenomenon did not
affect the initiation speed, but the total crack length and the overall crack speed
decreased along the S-a interface. The same phenomenon reduced the magnitude
of the precursor wave originating at the s-S interface in the 2s-2S-2a target (not
shown). As a result, the second crack initiation (of Fig. 9) was delayed and the
initiation speed was reduced with an increase of the friction coefficient from 0.0 to
0.1, while the second crack did not initiate at the friction value of 1.0. Similarly, the
initiation speed along the S-a interface in the 2s-4S target reduced from 36 km/s to
19.3 km/s due to the increased energy dissipation at the preceding s-S interface in
the target.
These results show that it is possible to increase the load spreading in bonded
multi-layered targets with low interface strength by increasing the friction properties
of surfaces that will come into contact after the interface failure. However, friction
coefficient of unity, or more at higher impact velocities, to attain the upper bound
of the load spreading is not practical. The friction characteristic of the interfaces
affect the interface crack initiation and propagation to varying degrees depending
on the depth of interface and materials on either side due to the stress wave and
energy attenuation in the vertical direction. However, the friction has no effect at
2 mm depth along the s-S interface which is the interface of the SiC with the most
compliant steel considered in the simulations.

5.4. Effect of interface parameters on load spreading


The effect of the interface parameters on the load spreading in the 6S, 2s-4S, and
2s-2S-2a targets was studied by varying the interface strength T m as 100 MPa,
300 MPa, and 500 MPa. At each T m value, the critical strain energy release rate Gc
was varied as 300 N-m/m2 , 900 N-m/m2 , and 1500 N-m/m2 . All the analyses were
carried out with the fixed friction coefficient µI = 0.3 and µL = 0.1.
Figure 14 shows the load spreading in the 2s-2S-2a target at the three T m
values keeping Gc constant at 300 N-m/m2 . The NDEs simulated with perfect bond
condition is also shown in Fig. 14. The increase in the interface strength increases the
horizontal momentum transfer H and reduces the vertical momentum transfer V .
The appreciable reduction in the V /H ratio leads to higher peak values and reduced
August 31, 2005 10:22 WSPC/IJCM-j050 00054

366 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Fig. 14. Increase in the load spreading with increase in the interface strength shown for the case
of 2s-2S-2a target at HDP (left) and FDP (right) simulated with Gc = 300 N-m/m2 , µI = 0.3,
µL = 0.1.

gradient of the radial NDEs. The load spreading increases with increase in the
interface strength and reaches the upper bound of the perfectly bonded condition.
At T m = 300 MPa, the load spreading is similar to the upper bound except for
slightly lower peak values of the radial NDEs. The increase in strength from 300
to 500 MPa does not discernibly alter the load spreading characteristics. The load
spreading also increases with the increase in Gc at constant T m . However, the
increase was not as appreciable as with the increase in T m , seen in Fig. 14, and is
not discussed.
On the contrary, T m and Gc have comparable effect on the interface crack
initiation and propagation in these targets. The effect of increasing T m at constant
Gc = 300 N-m/m2 is shown along the s-S interface in the 2s-4S target at the top
of Fig. 15. The increase in the interface strength from 100 MPa to 500 MPa delays
the crack initiation and shifts the initiation site away from the axis-of-symmetry. It
also reduces the crack initiation speed, total crack length, and overall crack speed.
However, the effect at 500 MPa interface strength are more pronounced compared
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 367

Fig. 15. Reduction in the crack length and overall crack speed with increase in T m at constant Gc
along the s-S interface in the 2s-4S target (top) and with increase in Gc at constant T m (bottom)
along the S-a interface in the 6S target (µI = 0.3, µL = 0.1).

to the 100 MPa and 300 MPa. For example, the crack initiation speed reduces from
24.5 km/s at 100 MPa to 22.5 km/s at 300 MPa and to 14.4 km/s at 500 MPa. At
300 MPa, the steady state crack speed reduces marginally from 3.8 km/s (1.21Cs
steel) to 3.3 km/s (1.05Cs steel) and lasts for 3 µs. The steady state crack speed is
not attained at 500 MPa interface strength. Similar phenomena were observed at
the s-S interface in 2s-2S-2a target, but the effects were less pronounced compared
to the 2s-4S target due to the early release of confinement. The s-S interface in the
two targets fail at all values of T m and Gc .
The similar effect of increase in the critical strain energy release rate Gc at
constant T m = 100 MPa is shown for the case of S-a interface in the 6S target at the
bottom of Fig. 15. The crack initiation is delayed and the initiation site shifts away
from the axis-of-symmetry as Gc is increased from 300 N-m/m2 to 1500 N-m/m2 .
With the increase in the energy needed to drive the crack, the crack initiation
speed, total crack length, and overall crack speed are reduced. The effects are more
August 31, 2005 10:22 WSPC/IJCM-j050 00054

368 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

pronounced at Gc = 1500 N-m/m2 compared to the lower two values. The interface
did not fail at the two higher values of the interface strengths. The same phenomena
were observed at the S-a interface in the 2s-4S and 2s-2S-2a targets.
These results show that the interface strength and critical strain energy release
rate play an equally important role for maintaining the integrity of the multi-layered
targets. However, the interface strength is more important for load spreading. The
overall failure of the interface reduces with increase in both the parameters. The
effect was found more pronounced for interfaces at larger depth from the impact
surface. At 2 mm depth along the s-S interface, the crack initiation and propagation
remains unaffected till the strength of 300 MPa, with the exception of reduction in
the duration of steady state crack propagation.

6. Conclusions
Detailed axisymmetric finite element analyses were carried out to study the effect of
finite strength interface with friction on the lateral load spreading in multi-layered
targets and associated phenomena of interface crack initiation and propagation.
Two configurations (0S and 6S) studied previously and two new configurations
(2s-4S and 2s-2S-2a) were simulated with varying interface strength, critical strain
energy release rate, and friction along the newly created sliding surfaces after fail-
ure of the interfaces between target layers. The targets were first simulated with
varying Coulomb friction coefficients keeping the interface parameters unchanged.
Subsequently, the same targets were analyzed with varying interface parameters at
constant friction coefficient. For comparison, the simulations were also carried out
with the perfect bond condition and slideline interfaces used in the previous study.
The load spreading was studied with the NDE parameter and phenomena of inter-
face crack initiation and propagation were studied in terms of the crack initiation
time, site, crack length, and crack speed along the interface.
The present study demonstrates that the high wave speed SiC ceramic remains
a potential material for enhancing the lateral load spreading if the confinement
is maintained by upper material layer to reduce/delay the materials damage. The
load spreading in targets with low interface strength can be further increased to
approach the upper bound of the perfectly bonded target layers by increasing the
friction coefficient along the newly created sliding surfaces after interface failure.
This is in accordance with the previous study where interfaces between target layers
were treated as free surfaces (stacked together without bond) with friction. However,
it is not practical to achieve the friction coefficient of unity, or more at higher impact
velocities, to attain the upper bound of the lateral load spreading.
Lateral load spreading in the bonded multi-layered targets depends more on
the maximum interface strength than on the critical strain energy release rate
of the interface. The load spreading increases with the increase in the interface
strength and reaches the upper bound of the perfectly bonded target layers. The
present method lets the computational procedure determine the maximum interface
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 369

strength to attain the maximum possible load spreading for the given multi-layered
target, impactor, and impact velocity. Moreover, the resilience of the multi-layered
targets to high velocity impact depends equally on the interface strength and crit-
ical strain energy release rate of the interface. The phenomena of interface crack
initiation and propagation in bonded multi-layered target under varying layer con-
figurations and interface properties at high velocity impact were studied for the first
time to the best of our knowledge.
The interface crack between target layers initiates and propagates in shear,
mode II. The propagating crack tip stops due do the attenuation of the stress
wave and driving energy due to the dissipation in inelastic deformation, damage,
and interface failure. The change in loading from mode II to mixed mode due to the
unloading wave from nearby interfaces also unloads and stops the propagating crack
tip. The interface crack speed in multi-layer targets at high impact velocity is invari-
ably the maximum on initiation and can be of the order of the longitudinal wave
speed of materials on either side of the interface. The maximum initiation speed of
11.29Cs , 6.76Cb , and 5.55Cl of the more compliant aluminum was obtained at the
SiC-aluminum interface at 4 mm depth from the impact surface for the frictionless
case. The steel-SiC ceramic interface at 2 mm depth had the initiation speed of
7.79Cs , 5.35Cb and 4.17Cl of the more compliant steel for the frictionless case. The
crack speed reduces monotonously after initiation, but remains in the intersonic
region for a measurable time period of more than 150 ns. Whether the propagating
crack tip will attain the steady state depends on the available driving energy. This
makes the interface crack initiation and propagation a function of the impact veloc-
ity, interface properties, depth of interface from the impact surface, materials on
either side, degree of damage to layer materials and resulting degradation in strength
and moduli. The steady state crack speed of 2.7 km/s to 3.8 km/s was obtained only
at the steel-SiC ceramic interface at 2 mm depth from the impact surface.
The present study also concludes that it is possible to arrive at a combination
of all the factors, viz., the materials on either side of interface, interface properties,
depth of interface, and impact conditions to observe the steady state interface crack
propagation independent of the friction coefficient. The steady state intersonic crack
speed of 1.05Cs to 1.21Cs of the more compliant steel was obtained at 2 mm depth
from impact surface along the steel-SiC interface in the 2s-4S target. The crack
speed jumped to intersonic speed 2.18Cs during the shear reversal. The phenomena
present the opportunity to further study the interface crack propagation at this
interface in more detail.
At the end, certain features of the present simulations need to be reiterated.
The results depend on the phenomenological model of SiC ceramic and the con-
tact/cohesive model developed and used in the simulations. The results from the
two models agreed well with the plate impact experiment data. The simulations
also predicted the load spreading in agreement with the effect of degradation in
strength and moduli and improved load spreading under confinement. The phenom-
ena of interface crack initiation and propagation will change considerably leading
August 31, 2005 10:22 WSPC/IJCM-j050 00054

370 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

to the intra-laminar crack if the micro-cracking models are used to simulate the
fragmentation of the SiC and metal layers. However, it will not change the load
spreading characteristics that depend on the layer-to-layer load transfer controlled
by the strength and moduli of the layer materials. One can also envisage that the
presented features of the crack initiation and propagation at interface closer to the
impact surface (e.g. s-S interface at 2 mm depth) will not change due to the avail-
ability of the impact energy without much dissipation along the thickness and weak
interface acting as preferred path. However, the inference remains to be verified.

Acknowledgments
This work was supported by the Army Research Office Grant Number DAAH04-96-
10053 and the considerable interest by Dr. Kailasam Iyer is sincerely acknowledged.
Lawrence Livermore National Laboratory is acknowledged for providing the research
code DYNA2D and its various components.

References
Abrate, S. (1991). Impact on laminated composite materials. American Society of Mechan-
ical Engineers. Applied Mechanics Review, 44(4): 155–190.
Abrate, S. (1994). Impact on laminated composites — Recent Advances. American Society
of Mechanical Engineers. Applied Mechanics Reviews, 47(11): 517–544.
Addesio, F. L. and Johnson, J. N. (1990). A constitutive model for the dynamic response
of brittle materials. Journal of Applied Physics, 67(7): 3275–3286.
Ammann, W. J., Liu, W. K., Studer, J. A. and Zimmermann, T. (1987). IMPACT: Effect
of Fast Transient Loading, (ed.) A. A. Balkema, Rotterdam, USA.
Brar, N. S., Yuan, G., Espinosa, H. D. and Zavattieri, P. D. (1997). Experimental study
on interface defeat in confined ceramic targets. In: Proceedings of 97 APS Topical
Conference of Shock Compression of Condensed Matter, July 27–August 1, Amherst,
MA.
Camacho, G. T. and Ortiz, M. (1996). Computational modeling of impact damage in brittle
materials. International Journal of Solids and Structures, 33(20–22): 2899–2938.
Cantwell, W. J. and Norton, J. (1991). The impact resistance of composite materials —
A review. Composites, 22(5): 347–362.
Chen, W. and Ravichandran, G. (1994). Dynamic compressive behavior of ceramics under
lateral confinement. Journal De Physique, III 4: C8–177.
Connor, M. P. (1988). Shear Wave Measurements to Determine the Nonlinear Elastic
Response of Fused Silica Under Shock Loading. MS Thesis, Department of Physics,
Washington State University, Pullman, WA 99164.
Corbett, G. G., Reid, S. R. and Johnson, W. (1996). Impact loading of plates and shells by
free-flying projectiles: A review. International Journal of Impact Engineering, 18(2):
141–230.
Curran, D. R., Seaman, L., Cooper, T. and Shockey, D. A. (1993). Micromechanical model
for comminution and granular flow of brittle material under high strain rate application
to penetration of ceramic targets. International Journal of Impact Engineering, 13(1):
53–83.
Dwivedi, S. K. and Espinosa, H. D. (2003). Modeling dynamics crack propagation in fiber
reinforced composites including frictional effects. Mechanics of Materials, 35(3–6):
481–509.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 371

DYNA2D (1992). A Nonlinear, Explicit, Two-Dimensional Finite Element Code for Solid
Mechanics. User Manual, UCRL-MA-110630, Lawrence Livermore National Labora-
tory, University of California at Berkeley, USA.
Espinosa, H. D. (1995). On the dynamic shear resistance of ceramic composites and its
dependence on applied multiaxial deformation. International Journal of Solids and
Structures, 32(21): 3105–3128.
Espinosa, H. D., Emore, G. L. and Zavattieri, P. D. (1996). Computational modeling of
geometric and material nonlinearities with an application to impact damage in brittle
materials. Advances in Failure Mechanisms in Brittle Materials, (eds.) R. J. Clifton
and H. D. Espinosa, MD-75/AMD-219: 119–161.
Espinosa, H. D., Zavattieri, P. D. and Dwivedi, S. K. (1998a). A finite deformation con-
tinuum/discrete model for the description of fragmentation and damage in brittle
materials. Journal of the Mechanics and Physics of Solids, 46(10): 1909–1942.
Espinosa, H. D., Dwivedi, S., Zavattieri, P. D. and Yuan, G. (1998b). Numerical investiga-
tions of penetration in multilayered material/structure systems. International Journal
of Solids and Structures, 35(2): 2975–3001.
Espinosa, H. D., Brar, N. S., Yuan, G., Xu Y. and Arrieta, V. (2000a). Enhanced bal-
listic performance of confined multi-layered ceramic targets against long rod penetra-
tors through interface defeat. International Journal of Solids and Structures, 37(36):
4893–4913.
Espinosa, H. D., Dwivedi S. and Lu H.-C. (2000b). Modeling impact induced delamination
of woven fiber reinforced composites with contact/cohesive laws. Computer Methods
in Applied Mechanics and Engineering, 183: 259–290.
Feng, R., Raiser, G. F. and Gupta, Y. M. (1996a). Shock response of polycrystalline
silicon carbide undergoing inelastic deformation. Journal of Applied Physics, 79(3):
1378–1387.
Feng, R. and Gupta, Y. M. (1996b). Materials Models for Saphire, α-Quartz, Lithium
Floride, and Fused Silica for Use in Shock Wave Experiments and Wave Code Cal-
culations. Internal Report, Institute for Shock Physics, Washington State University,
Pullman, WA 99164, USA.
Feng, R., Raiser, G. F. and Gupta, Y. M. (1998). Material strength and inelastic defor-
mation of silicon carbide under shock wave compression. Journal of Applied Physics,
83(1): 79–86.
Flanagan, D. P. and Belytschko, T. (1981). A uniform strain hexahedron and quadrilateral
with orthogonal hourglass control. International Journal for Numercial Methods in
Engineering, 17: 679–706.
Gama, B. A., Bogetti, T. A., Fink, B. K., Yu, C.-J., Claar, T. D., Eifert, H. H. and
Gillespie, J. W. Jr. (2001). Aluminum foam integral armor: A new dimension in armor
design. Composite Structures, 52: 381–395.
Goudreau, G. L. and Hallquist, J. O. (1982). Recent developments in large-scale finite
element lagrangian hydrocode technology. Computer Methods in Applied Mechanics
and Engineering, 33: 725–757.
Grady, D. E. (1988). The spall strength of condensed matters. Journal of the Mechanics
and Physics of Solids, 36(3): 353–384.
Grady, D. E. (1994). Shock-wave strength properties of boron carbide and silicon carbide.
Journal De Physique, IV4: C8–385.
Grady, D. E. (1998). Shock-wave compression of brittle solids. Mechanics of Materials, 29:
181–203.
Gupta, Y. M. and Ding J. L. (2002). Impact load spreading in layered materials and
structures. International Journal of Impact Engineering, 27(3): 277–291.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

372 S. K. Dwivedi, J. L. Ding & Y. M. Gupta

Hallquist, J. O., Goudreau, G. L. and Benson, D. J. (1985). Sliding interfaces with contact-
impact in large-scale lagrangian computations. Computer Methods in Applied Mechan-
ics and Engineering, 51: 107–137.
Hauver, G. E., Netherwood, P. H., Benck, R. F. and Kecskes, L. J. (1994). Enhanced
ballistic performance of ceramics. 19th Army Science Conference, Orlando, FL, USA.
Holmquist, T. J., Johnson, G. R., Grady, D. E., Lopatim, C. M. and Hertel E. S. Jr.
(1995). High strain rate properties and constitutive modeling of glass. 15th Interna-
tional Symposium on Ballistics, Jerusalem, Israel.
Holmquist, T. J., Templeton, D. W. and Bishnoi, K. D. (2001). Constitutive modeling
of aluminum nitride for large strain, high-strain rate, and high-pressure applications.
International Journal of Impact Engineering, 25: 211–231.
Holmquist, T. J. and Johnson, G. R. (2002). Response of silicon carbide to high velocity
impact. Journal of Applied Physics, 91(9): 5858–5866.
Huang, Y., Wang, W., Liu, C. and Rosakis, A. J. (1999). Analysis of intersonic crack
growth in unidirectional fiber-reinforced composites. Journal of the Mechanics and
Physics of Solids, 47(9): 1893–1916.
Johnson, G. R. and Cook, W. H. (1985). Fracture characteristics of three metals sub-
jected to various strains, strain rates, temperatures and pressure. Engineering Fracture
Mechanics, 21(1): 31–48.
Kinslow, R. ed. (1970). High-Velocity Impact Phenomena. Academic Press, New York,
USA.
Mahfuz, H., Zhu, Y., Haque, A., Abutalib, A., Vaidya, U., Jeelani, S., Gama, B.,
Gillespie, J. and Fink B. (2000). Investigation of high-velocity impact on integral armor
using finite element method. International Journal of Impact Engineering, 24: 203–217.
Needleman, A. (1987). Continuum model for void nucleation by inclusion debonding.
Trans. ASME, Journal of Applied Mechanics, 54(3): 525–531.
Nesterenko, V. F., Meyers, M. A. and Chen, H. C. (1996). Shear localization in high-strain-
rate deformation of granular alumina. Acta Materiallia, 44(5): 2017–2026.
Ohtani, Y. and Chen, W.-F. (1989). A plastic softening model for concrete materials.
Computers and Structures, 33(4): 1047–1055.
Potti, S. V. and Sun, C. T. (1997). Prediction of impact induced penetration and delami-
nation in thick composite laminates. International Journal of Impact Engineering, 19:
31–48.
Raiser, G. F., Wise, J. L., Clifton, R. J., Grady, D. E. and Cox, D. E. (1994). Plate impact
response of ceramics and glasses. Journal of Applied Physics, 75(8): 3862–3869.
Rajendran, A. M. and Kroupa, J. L. (1989). Impact damage model for ceramic materials.
Journal of Applied Physics, 66(8): 3560–3565.
Rajendran, A. M. (1994). Modeling the impact behavior of AD85 ceramic under multiaxial
loading. International Journal of Impact Engineering, 15(6): 749–768.
Rajendran, A. M. and Grove, D. J. (1996). Modeling the shock response of silicon carbide,
boron nitride and titanium diboride. International Journal of Impact Engineering,
18(6): 611–631.
Repetto, E. A., Radovitzky, R. and Ortiz, M. (2000). Finite element simulation of dynamic
fracture and fragmentation of glass rods. Computer Methods in Applied Mechanics and
Engineering, 183(1–2): 3–14.
Rice, M. H., McQueen, R. G. and Walsh, J. M. (1958). Compression of solids by strong
shock waves. Solids State Physics, Advances in Rsearch and Applications 6, (eds.)
F. Seitz and D. Turnbull, pp. 1–63.
Robbins, J., Ding, J. L. and Gupta, Y. M. (2004). Load spreading and penetration resis-
tance of layered structures — A numerical study. International Journal of Impact
Engineering, 30(6): 593–615.
August 31, 2005 10:22 WSPC/IJCM-j050 00054

Load Spreading in SiC Multi-Layered Targets 373

Rosakis, A. J., Samudrala, O., Singh, R. P. and Shukla, A. (1998). Intersonic crack propa-
gation in biomaterial systems. Journal of the Mechanics and Physics of Solids, 46(10):
1789–1813.
Seaman, L., Curran, D. R. and Shockey, D. A. (1976). Computational model for ductile
and brittle fracture. Journal of Applied Physics, 47(11): 4814–4826.
Sundaram, S. and Clifton, R. J. (1998). The influence of glassy phase on the high strain
rate response of a ceramic. Mechanics of Materials, 29: 233–251.
Taylor, L. M., Chen E. P. and Kuszmaul, J. S. (1986). Microcrack-induced damage accumu-
lation in brittle rock under dynamic loading. Computer Methods in Applied Mechanics
and Engineering, 55: 301–320.
Wilkins, M. L. (1978). Mechanics of penetration and perforation. International Journal of
Engineering Science, 16: 793–807.
Wu, J. and Gupta, V. (2000). Observations of transonic crack velocity at a metal/ceramic
interface. Journal of the mechanics and Physics of Solids, 48: 609–619.
Xu, L. R. and Rosakis, A. J. (2002). Impact failure characteristics in sandwich structures
Part I: Basic failure mode selection. International Journal of Solids and Structures,
39: 4215–4235.
Yu, C., Pandolfi, A., Ortiz, M., Cocker, D. and Rosakis, A. J. (2002). Three-dimensional
modeling of intersonic shear-crack growth in asymmetrically loaded unidirectional com-
posite plate. International Journal of Solids and Structures, 39(25): 6135–6157.
Yuan, G., Feng, R. and Gupta, Y. M. (2001). Compression and shear wave measurements
to characterize the shocked state in silicon carbide. Journal of Applied Physics, 89(10):
5372–5380.
Yuan, G., Feng, R. and Gupta, Y. M. (2003). Unpublished experimental results. Institute
for Shock Physics, Washington State University, Pullman, WA 99164-2816, USA.
Zukas, J. A., Nicholas, T., Swift, H. F., Greszczuk, L. B. and Curran, D. R. (1982). Impact
Dynamics. John Wiley & Sons, Inc., New York, USA.

You might also like