You are on page 1of 9

doi: 10.1111/j.1460-2695.2006.01054.

Aspects of rapid crack propagation in silicon


D. SHERMAN
Department of Materials Engineering, Technion, Israel Institute of Technology, Haifa 32000 Israel

Received in final form 7 June 2006

A B S T R A C T Fracture experiments with silicon specimens in recent years have shown the need for a
new approach to the analysis of rapidly propagating cracks in single crystals. Behaviour
and phenomena have been revealed that fracture in these materials is rather different from
the fracture of both amorphous and polycrystalline materials. We show that continuum
mechanics is insufficient for analyzing crack propagation in single crystals since it is unable
to consider atomistic-scale phenomena. Accordingly, we describe basic phenomena asso-
ciated with rapid crack propagation in silicon: (i) anisotropic velocity-dependent R-curve
behaviour, as a key phenomenon dictating atomistic scale behaviour, (ii) crack deflection
from one cleavage plane to another as a mesoscopic scale phenomenon in single-crystal
fracture, (iii) the Rayleigh surface wave speed as the limiting crack tip velocity is re-
examined, (vi) the lowest crack velocity in brittle crystals is examined, and finally (v) the
interaction between crack path and preferred cleavage planes in single crystals is depicted.

Keywords crack deflection; cleavage; dynamic fracture; energy dissipation; single crystal;
velocity-dependent.

The situation with single crystals is different. In the cur-


INTRODUCTION
rent contribution we show that crack initiation, propaga-
Polycrystalline and amorphous materials are considered tion and path in single-crystals does require precise knowl-
to be isotropic; they have a single property characterizing edge and understanding of the fracture events on the
their resistance to crack initiation, namely either K IC or atomistic scale. The importance of such knowledge rests
G IC , both isotropic and defining the fracture resistance in in the fact that crack initiation and propagation occur on
the entire volume and in any plane and direction. K IC stands the atomistic scale, being associated with the breaking
for the fracture toughness, G IC , for the fracture energy, of bonds between atoms,1 which is of profound impor-
both being interrelated.1–3 When the stresses (defined by tance in single crystals (SCs); therefore, understanding
K IC ) or the energy (defined by G IC ) are sufficiently high, the material’s atomistic structure, crystallographic planes
the crack will initiate and then propagate, stably or un- and orientations, and their energies is essential. We there-
stably, its velocity and path being dictated by the dynamic fore continue this introduction with the atomistic aspects
properties of the material and by the appropriate physical of fracture.
rules.4–6 Cracks may propagate either with no additional Crack propagation in brittle SCs is associated with two
crack resistance, as is assumed for brittle materials or with spatial parameters—the plane of propagation and the di-
increasing resistance, as in ductile, usually metallic, mate- rection of propagation—together they construct the crack
rials. In that case, the material is considered as exhibiting system. In contrast to the fracture properties of poly-
an R-curve behaviour, its resistance to fracture increasing crystalline and amorphous materials, brittle single crys-
as the crack length increases.1 tals possess varying fracture energies for various planes
Fracture mechanics usually refers to continuum, and directions of crack propagation.7,8 The atomistic ar-
isotropic, homogenous and dense materials. This being rangement of brittle SCs, and in particular the interatomic
so, it is noteworthy that fracture mechanics does not re- forces, dictate the fracture processes, initiation and prop-
quire the precise description and knowledge of the ma- agation. Therefore, when a brittle SC is loaded to frac-
terial’s microstructure and of the events on the atomistic ture, it tends to cleave on one of the family of favoured
scale near the crack tip in order to predict fracture initia- crystallographic planes, generally on that with the high-
tion and propagation. est in-plane atomic density and maximum inter-planar

32 
c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
R A P I D C R AC K P R O PAG AT I O I N S I L I C O N 33

separation, similar to the conditions dictating dislocation length and crack’s velocity, respectively, G=G0 g(V ) where
mobility in metallic single crystals. G0 = G(a, V = 0, t, geometry, load) and g(V ) is a univer-
The decision as to which cleavage system will be the one sal dynamical function expressing the kinetic energy and
preferred for crack initiation in a specific geometry, crys- approximated14 to be: g(V ) ≈ 1 − V /C R , which implies
tallographic structure and loading configuration, is made that C R is the upper limit of crack velocity. Stroh15 argued
possible with the aid of fracture theory.9,10 For this pur- that C R , the Rayleigh free surface wave speed, is the the-
pose, and assuming that the crystal contains a pre-existing oretical upper limit for crack velocity in solids. It should
crack or flaw, one calculates the quasi-static strain energy be noted that C R depends only on the elastic constants of
release rate (SERR) per unit area of crack advance, G0 , and the material, and no microstructural or dynamical effects
evaluates, experimentally, the quasi-static fracture energy, were considered.
 0 , for each possible cleavage system. The crack initi- Experimental observations and numerical calculations
ates on the first system that fulfils the Irwin criterion:10 show that the crack velocity, V , never attains C R , even
G0 =  0 . The lower limit to  0 is usually assumed to be at large values of SERR.16–18 Nevertheless, the crack ve-
2γ S , which is twice the bond-breaking energy.9 However, locity may reach a significant fraction of C R .
this value may be higher due to the lattice trapping ef- We investigated the fracture behaviour of single-crystal
fect,7,8,11–13 which depends on the atomistic arrangement, silicon, since it is the best man-made material, having a
namely the different distances between the atoms consti- perfect crystallographic structure with almost no defects.
tuting the plane.  0 , the quasi-static fracture energy, is It is also considered to be a nearly ideally brittle mate-
usually assumed to be isotropic for the specific cleavage rial. Silicon has been widely investigated over the years
plane, so that it does not depend on crystallographic ori- as a leading material in the microelectronics and MEMS
entation on that plane. This only involves stretching the industries,19 and its properties in almost every field, in-
bond until the forces between the atoms vanish or until cluding crack propagation, have been extensively studied,
the bond between the atoms breaks. However, it may be both experimentally and numerically, from both macro-
anisotropic within the plane, again due to the lattice trap- scopic and atomistic aspects. Silicon is commercially avail-
ping effect. Furthermore, this value varies from one crys- able and inexpensive, and as such it is most appropriate as a
tallographic plane to another due to the different atomic model material for investigating various features of crack
structures. propagation. Silicon has a diamond-like microstructure.
Molecular dynamics simulations of a cleavage in sili- The preferred cleavage planes in silicon are the {110}
con7 demonstrate the anisotropic nature of a fracture in and the {111} planes, and they are inclined 35.26◦ to-
a silicon crystal. The {111}<110>, the {111} <112>, wards each other.
and the {110}<110> systems [{hkl}-crack plane, <uvw> In the next sections, we describe the phenomenon of
crack direction] are the ‘continuous’ systems; the crys- velocity-dependent R-curve behaviour in silicon, which
tallographic structure of these systems being in such an is a highly anisotropic behaviour. The crack deflection
order that several bonds in the vicinity of the crack tip phenomenon and the deflection velocity as a function of
are stretching and contributing to the crack propagation crystallographic orientation are calculated based on a rig-
process. These atoms share the load at the crack tip, and orous analysis of the experimental results. The deflection
the crack propagation is therefore considered ‘continuous’ velocity defines the maximum crack tip velocity (along the
and requires less energy during rapid propagation. On the (110) plane) that a crack can acquire. Finally, it is shown
other hand, the fracture in the {110}<001> system ex- that the maximum crack velocity may be very far from that
hibits a completely different behaviour: the complex crys- predicted by Stroh. More critical issues are also discussed.
tallographic structure of this system prevents load sharing
between the atoms close to the crack tip, and a single atom
P H E N O M E N A A S S O C I AT E D W I T H R A P I D
at the crack tip is carrying most of the load. The bond-
C R A C K P R O PA G AT I O N
breaking events are therefore abrupt, which results in a
relatively high energy consumption. We postulate that, We fracture under three-point bending (3PB), thin and
analogous to crack initiation, crack propagation is also wide silicon specimens cut from silicon wafers. The exper-
controlled by the anisotropic atomistic structure of the imental set-up is schematically shown in Fig. 1. The ex-
crystal. perimental procedure has recently been widely described
As to the stage of propagation, Freund14 has defined the in the literature,20 and we therefore proceed without a
criterion for a straight and smooth crack moving at a con- detailed description of the specimens and of the loading
stant velocity, which is similar to Irwin’s criterion:11 G = procedure; they are advantageous in examining aspects
, where G and  are the temporal dynamic energy re- associated with rapid crack propagation in single crystals
lease rate and the temporal fracture energy, respectively. and with the influence of a variety of crack tip velocities
G=G(a, V , t, geometry, load), a and V are the temporal crack along variety of crystallographic orientations in a single


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
34 D. S H E R M A N

 I =  I (V ). Furthermore, the cleavage energy is also a


function of the crystallographic orientation on the cleav-
age plane, θ , thus:  I =  I (V ,θ), and the total cleavage
energy, more rigorously, is:
 = (V, θ) =  R (θ ) +  I (V, θ ). (1)

The dependence of  R on θ is due to the orientation-


dependent lattice trapping effect.7 The velocity and the
orientation-dependent cleavage energy may be described,
for simplicity, by a set of linear functions:
Fig. 1 The three point bending (3PB) specimen, the
crystallographic orientations, and the definition of the main axes. (V, θ) = 0 (θ ) + ∂(θ)/∂ V · V. (2)

specimen. We describe the phenomena rather than the With the aid of the deflection phenomenon detected dur-
experiments performed to obtain them. ing bending experiments, we were able to define a proce-
In this regard, we first refer the reader to two impor- dure to calculate the velocity and orientation-dependent
tant parameters in our investigations. One are the quasi- cleavage energy of the {110} cleavage plane,  {110} (V ,
static surface energies,  110
0 and  111
0 , for the {110} and θ), by estimating ∂ {110} (θ )/∂V .21 A basic assumption
the {111} preferred cleavage planes, respectively. De- in that estimation was that ∂ {111} (θ)/∂V is constant and
termination of these properties was attempted experi- independent of the crystallographic orientation, due to
mentally and numerically, they were found to be widely the nearly isotropic nature of the (111) plane. Fortu-
scattered, and their exact value are still debated. Nev- nately, an estimation of ∂ (111) /∂V is possible,17 it was
ertheless, we use the results lately obtained by full- shown there that a crack moving on the {111}<112̄> sys-
density functional molecular dynamic simulations,7 with tem at a velocity of 2000 m/s gains only 0.1 J/m2 above
which they were calculated to be 3.46 and 2.88 J/m2 the quasi-static value. From finite-difference calculations
for  1100 and  111
0 , respectively. However, in specimens we obtain ∂ (111) /∂ V ≈ 5 · 10−5 J s/m3 for any crystallo-
cut from [001] silicon wafers the {111} plane is inclined graphic orientation. Using the values mentioned above
35.26◦ to the {110} plane, and therefore the total en- for the quasi-static cleavage energies7 and substituting
ergy dissipated by a crack running along the {111} plane ¯ 0111 − 0110 = 0.06 J/m2 , as well as the deflection veloc-
is ¯ 0111 = 2.88/cos(35.26) = 3.52J/m2 . In this case, the ity, we get:21
{110} plane becomes the low-energy cleavage plane. In
the following discussion, G and  therefore refer to a ∂ (110) (θ ) 0.06
= 5 × 10−5 + . (3)
unit area of the specimen’s cross-section, not to the crack ∂V Vdefl (θ )
surface.
The gradient on the left side of Eq. (3) is readily found
if V defl (θ ) is known. Indeed, V defl (θ ) was evaluated us-
Velocity-dependent R-curve behaviour of a crack
ing somewhat lengthy analysis of the deflection phe-
running on the {110} plane
nomenon.21 The results are shown in Fig. 2a. The gra-
We start this Section with a description of the fundamental dients ∂ {110} (θ )/∂V of the [11̄0] and the [001] crystallo-
material behaviour that, we suggest, is the key reason for graphic orientations for θ equal 0◦ and 90◦ , respectively,
the phenomena associated with rapid crack propagation. are 7.08.10−5 and 3.1.10−3 Js/m3 , respectively (a ratio of
Although the results described here were obtained follow- 44). We suggest that this large difference is to be attributed
ing a thorough analysis of the deflection phenomenon,20 to the wide variations in the atomic arrangement of silicon
we have chosen to describe it first. The observation of on the (110) plane.
crack deflection from the {110} to the {111} cleavage We now rotate the silicon crystal to view the crystallo-
plane in SC silicon as a function of crack tip velocity graphic structure that the crack front ‘sees’ when changing
was interpreted by atomistic energy-dissipation mecha- its orientation from [11̄0] to [001] (θ=0◦ to 90◦ ). The plane
nisms,20 in which the basic form of the internal cleav- remains the (110), and the axis of rotation is the [110]. For
age energy is suggested by:  =  R +  I , where  R and the [11̄0] direction, the atomic structure, as seen in Fig. 2b,
 I are the reversible and the irreversible fracture ener- undergoes only minor changes until θ ∼54.7◦ , which is the
gies, respectively. We postulate that  I is a function of [11̄2] direction (Fig. 2c). At that angle, significant changes
crack tip velocity, V , which is the bond breaking velocity occur in the crystal structure and are retained until {the
along certain system (plane and orientation) and therefore, angle of} θ ∼90◦ , (the [001] direction) (shown in Fig. 2d)


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
R A P I D C R AC K P R O PAG AT I O I N S I L I C O N 35

Fig. 2 Energy and crystallographic structure: (a) the energy gradients,∂ (11̄0) (θ )/∂ V, versus the crystallographic orientation and the
crystallographic arrangements for a crack propagating along, (b) the (110) [11̄0], (c) the (110) [112̄] and (d) the (110) [001] systems.

that possesses a completely different microstructure. We altered for different values of the two. The basic behaviour,
therefore suggest that at that point, the energy dissipation however, will not change.
due to phonon radiation is high compared with other ori- We postulate that the reason for the velocity-dependent
entations. This results in a high dynamic-energy gradient R-curve behaviour in nearly ideal brittle single crystals
of this system. Note also that when the crystallographic is phonon radiation, which is dependent on the veloc-
orientation changes from the [11̄0] to the [11̄2] direction, ity and the crystallographic orientation. The anisotropic
a ‘global load sharing’ mechanism takes over, as more than phonon radiation is a source of energy dissipation mech-
two bonds are participating in the bond-breaking mech- anisms and is strongly dependent on the crystal structure.
anisms, so that these mechanisms are continuous, while As such, the phonon radiation is responsible for various
from the [11̄2] to the [001] direction, ‘local load shar- phenomena associated with rapid crack propagation; the
ing’ prevails, in which only one bond carries most of the deflection described in the following is one of them.
load and the bond breaking mechanism is abrupt, and the
phonon radiation there is high. Another important out-
Crack deflection from the {110} to the {111} plane
come of this analysis is that while the gradients ∂ (110)
(θ)/∂V show wide variations, the total dynamic cleavage A most significant finding in our 3PB experiments was
energy at deflection,  (110) (V defl ,θ), Eq. (2), exhibits only the discovery of the macroscopic change in the preferred
small variations: It ranges from 3.52 to 3.66 J/m2 for the cleavage plane from (110) to (111) as a function of crack
[001] and the [11̄0] orientation, respectively. velocity and crystallographic orientation.20,21 Specimens
We note that the results described above are sensitive with relatively long notches (over 1 mm), and therefore
to the exact values of  110 ¯ 111
0 and 0 , and therefore may be with relatively low energy and velocity at crack initiation,


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
36 D. S H E R M A N

Irwin’s criterion11 is met, i.e., when G0 = 0 (point A in


Fig. 3b). The propagation line is velocity-dependent due
to phonon radiation at the moving crack tip (see below). G0
increases during crack propagation, as does , the material
property that resists crack propagation, on the (110) plane
(line AB), where the velocity-dependent dynamic energy
is the lowest for the two cleavage planes. At point B, the
crack deflects to the (111) cleavage plane, as this plane be-
comes the one with the least cleavage energy and the rest
of the propagation takes place over that plane. This is the
case of a crack propagating in the [11̄0] or any other di-
rection on both planes. However, for a crack propagating
on the (110) plane but in the [001] direction, the atomic
arrangement implies much higher phonon radiation and
therefore much steeper ∂ (110) /∂V (line AD in Fig. 3b).
As a result, the crack advancing in the [001] direction will
deflect at a much lower velocity (point D) than in the pre-
ceding case (point B), and thereafter will propagate along
the (111) plane. It is postulated that, on the macroscopic
scale, the deflection phenomenon is independent of C R
but does depend on the anisotropic microstructure of the
crystal.
With respect to this occurrence we have suggested a
more general energetic criterion for the selection of the
preferred plane of propagation20 with orientation depen-
Fig. 3 Schematic presentation of the deflection phenomenon. (a) A dence:
crack in specimens with intermediate notch length propagated first
on the (110) plane, thereafter the crack is deflected to the (111) G(V, θ) − i (V, θ) = max, (4)
plane. The crack’s direction is from left to right. (b) Schematic
diagram showing the deflection phenomenon from the energy where i is the i’th plane at the vicinity of crack’s plane.
point of view: the dynamic cleavage energy, (V )= R + I (V ) of Equation (4) is similar to Irwin’s criterion for crack initi-
both planes, as a linear function of crack tip velocity, for two ation.10 That criterion states that propagation occurs on
crystallographic orientations (near the [110] and near the [001]
the lowest-energy plane, or alternatively, on the weakest
R = CR .
directions), assuming C 111 110
plane for which G- is the maximum. At point B, de-
were cleaved along the (110) cleavage plane. In specimens flection to the (111) plane takes place, since it is now the
with short notches (below 0.5 mm), and therefore with rel- lowest-energy plane.
atively high energy and velocity, the fracture was always
over the (111) cleavage plane. The crack in specimens with
Rayleigh surface wave speed and the limiting
a notch of intermediate length (and therefore intermediate
crack velocity
energy and velocity) propagated first over the (110) plane,
and thereafter deflected to the (111) plane, as schemati- Using our experimental results concerning the crack de-
cally shown in Fig. 3a. Similar cases of crack deflection flection phenomenon, we constructed the critical veloc-
have recently been reported in the literature, but without ity at which the crack deflects from (110) plane to (111)
further analysis.18,22 plane as a function of the crystallographic orientations, as
As a crack rapidly advances in the material, a basic ques- shown in Fig. 4a. That plot defines the maximum veloc-
tion arises with regard to every single, discrete bond break- ity at which a crack can run on the (110) cleavage plane
ing mechanism: which is the selected path along which the as a function of crystallographic orientation along that
energy consumed in separating atoms is the lowest. The plane. For example, a crack running on the (110) [11̄0]
energy consideration for the deflection phenomenon is system will deflect to the (111) [11̄0] system at a velocity of
schematically shown in Fig. 3b, where both planes are as- ∼2900 m/s. On the other hand, when the crack is prop-
sumed to have the same C R , the Rayleigh free surface wave agating on the (110) [001] system, deflection to the (111
speed, for simplicity21 , for any arbitrary crystallographic [112̄] system occurs at a very low velocity, as low as 30 m/s,
orientation. As G0 gradually increases with increasing ex- which is two orders of magnitude lower than the previ-
ternal load, crack initiation and propagation start when ous velocity, meaning that a crack in the {110}<001>


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
R A P I D C R AC K P R O PAG AT I O I N S I L I C O N 37

tallographic orientations.23,24 The results for the {110}


cleavage plane are shown in the insert of Fig. 4b. The
results obtained for the deflection velocity vs. the crystal-
lographic orientation were normalized by Rayleigh speeds
and are shown in Fig. 4b. It is evident that as the crystal-
lographic orientation approaches the [001] direction, that
ratio drops to a few percent. For the most extreme case,
when the crack is propagating on the (110) cleavage plane
and in the [001] direction, the normalized deflection ve-
locity was calculated to be ∼0.01.

On the lowest velocity of a crack


The question whether a crack can be initiated and prop-
agate at a low velocity still puzzles many researches. Nu-
merical simulations, and molecular dynamics models in
particular, indicate a minimal velocity that the crack can
initiate at.25–28 Our 3PB specimens have demonstrated
that a crack can propagate at a velocity much lower than
that indicated by the simulations. To elucidate this point
we briefly describe the properties of our specimens, from
the velocity point of view. A detailed description can be
found elsewhere.29
The typical deformation upon bending plays a significant
role in our specimens. The tension/compression stress
field enforces a unique crack front. For the sake of sim-
plicity we assume that the shape of the crack front is that
of a quarter-ellipse having a long and straight ‘tail’ with
angle of inclination of 1◦ , see Fig. 5a. When the crack on
the bottom surface approaches the far edge of the spec-
imen, the tension/compression stress field changes, and
the crack front is no longer at ‘steady state’.
Fig. 4 (a) The deflection velocity, i.e., that at which the crack The crack front was identified by scribing the bottom
deflects from the (110) plane to the (111) plane, as a function of the surface of the specimens perpendicularly to the crack di-
crystallographic orientation, θ , with respect to the [11̄0] direction.
rection and observing the ‘Wallner lines’ or crack front
(b) The deflection velocity normalized by the Rayleigh surface wave
speed for the (110) plane as a function of crystallographic waves29 resulting from the crack’s interaction with the
orientations. The Rayleigh surface wave speeds for the (110) plane scribed scratch. The interaction generated a disturbance
as a function of the crystallographic orientations with respect to that propagated concentrically from the point of interac-
[11̄0] direction is shown in the insert.23,24 tion, marking the crack front by jogging it out of the plane,
as shown by an optical photograph of the fracture surface
of a crack running on the (110) cleavage plane of a sili-
system cannot propagate at a velocity higher than about con specimen, Fig. 5b. Note the nearly parallel lines of the
1% of C R .20 We postulate that this behaviour is due to the crack front near the top surface of the fracture, generated
changes in the atomic arrangement as the crystallographic by another scratch located to the right of the scratch pre-
orientation changes on the (110) cleavage plane, resulting viously introduced. The observed nearly parallel lines are
in variations in the phonon radiation and hence, in varia- of profound importance, since it indicates the low veloc-
tions in the energy dissipation mechanisms of each crys- ity that the crack can attain. This crack front was found
tallographic orientation. This point cannot be treated by to match in glass and in silicon specimens on the (110)
the conventional continuum mechanics approach, which cleavage plane.29 In both materials the fracture surfaces
does not ‘see’ atoms. are atomistically smooth. Recent new experiments with
The Rayleigh surface wave speeds on the {110} and our 3PB specimens have shown that profile also matches
{111} planes were measured over the entire range of crys- the crack front in silicon specimens cut from [110] wafers,


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
38 D. S H E R M A N

Fig. 5 (a) Schematic representation of the quarter-elliptical and the long-‘tail’ profile, the ridge pattern and the velocities associated with
crack propagation under 3PB loading, (b) the ‘Wallner line’ as the way to predicting the crack profile, and (c) the direction, θ , of the normal
velocity vector along the crack front and the normalized normal velocity, V n /V x, both as a function of the normalized location, y, with
respect to the height, h, of the fracture surface, y/h, of a point on that profile.

where the crack is running on the perpendicular, (111) At steady state, which is maintained for most of the
cleavage plane. crack’s propagation, the crack front moves at a nearly
We postulate that the crack front is constructed of a set constant parallel velocity (V x ) in the x direction. The
of ridges initiating on the bottom surface, Point A, and physically significant velocity of the ridge, V n , coincides
advances in the material towards the upper portion of the with the parallel velocity, V x , in the lower portion of the
specimen, Point B, as schematically shown in Fig. 5a for cross-section (see Fig. 5a). In the upper portion of the
an individual ridge and fully described elsewhere.29 The specimen that velocity is very low and almost normal to
velocity of propagation, V n , of an individual ridge is al- V x , and it coincides with the [001] direction. It is again
ways normal to the local crack front, see Fig. 5a. This stressed that the specimen, with its unique crack front,
normal velocity is the actual rate at which the bonds be- covers a large variety of velocities and crystallographic ori-
tween atoms along the crack tip path are breaking and is entations, in contrast, for example, to tensile specimens,
therefore the physical velocity. which consist of only one cleavage system. The normal


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
R A P I D C R AC K P R O PAG AT I O I N S I L I C O N 39

velocity at any point of the crack front is given by V n = of silicon and of glass are dissimilar in nature. This point
V x sin θ , where θ is the angle between the local normal requires a more thorough experimental program and a
to the crack front and the x-axis, as shown in Fig. 5c. The new theoretical treatment.
resulting normal velocity, V n , is also shown in Fig. 5c.
The normal velocity decreases as the crack approaches
the top portion of the fracture surface and was esti- SUMMARY
mated20,21,29 to be of the order of several tens of m/s, We demonstrate the differences between several aspects
see Fig. 5c. This is in contrast to the minimum velocity of rapid crack propagation in amorphous materials and
that a crack can decrease to, as demonstrated by numerical in single crystals. While the continuum mechanics ap-
simulations. According to the simulations, the minimum proach well predicts fracture initiation and propagation in
velocity of crack propagation is about 0.35 C R, 25–28 which isotropic materials, such as amorphous and polycrystalline
in silicon on the (110) cleavage plane in the [110] and the ones, it cannot predict changes in the crack path or account
[001] directions, is nearly 1600 and 1900 m/s, respectively. for material behaviour on the atomic scale. This being so,
The maximum parallel velocity of a crack running on the the problem of understanding the fracture of brittle single
(110) plane without deflection was measured to be about crystals is challenging. We have pointed out on velocity
1500 m/s. Our quarter-elliptical crack front shape with and orientation dependent cleavage energy as the key phe-
long tail contradicts that suggested minimal crack veloc- nomenon responsible for the major phenomena occurring
ity supposition. during rapid crack propagation. The velocity-dependent
R-curve behaviour is attributed to the velocity dependent
Crack path and cleavage plane phonon radiation, which is a source of energy dissipation.
We have, in particular, demonstrated the phenomenon of
It is well accepted that cracks propagate on planes with crack deflection from one cleavage plane to another due to
the lowest cleavage energy, these planes therefore known energy considerations, and the different values of velocity
as cleavage planes. In silicon, the {111} and {110} family dependent energy dissipation on both preferred cleavage
of planes are the cleavage planes, of which the latter has planes in silicon. The discovery that the deflection ve-
the higher cleavage energy. A crack cannot propagate on locity varies as the crystallographic orientation varies has
any other plane. Nevertheless, it is evident that when frac- led to another finding; the Rayleigh surface wave speed,
turing silicon wafers, the crack paths are not necessarily a constant dictated by the elastic constant of the material
straight. only, is well accepted as the upper limit to crack velocity,
It has lately been shown that, in glass30 and even in sin- but we have shown that the maximum crack tip velocity
gle crystal silicon,31 cracks can oscillate and propagate in is dictated by the microstructure. Furthermore, we have
a sinusoidal pattern with atomistically smooth surfaces. shown that a crack can propagate at a low velocity, as low
We have recently shown that a fracture surface in crys- as tens of m/s, the lowest range of velocities that we can
talline silicon may be perturbed, but the perturbations measure. This is in contrast to predictions by molecular
are smooth. In silicon, for example, when a crack is in- dynamics simulations. Finally, the cleavage plane as the
teracting with stationary dislocations, the fracture surface selection of the crack path (as a recommendation only)
is smooth and flat but exhibits nanometric perturbations, was discussed.
constructed, presumably, from atomistic steps along the
(111) and (110) cleavage planes to form complicated crack
paths.32,33 It is therefore possible for a crack to propagate Acknowledgement
along paths that are not straight and yet atomistically to
The support of the Israel Science Foundation through
follow the known cleavage planes.
Grant No. 1110/04 is gratefully acknowledged.
Several theories, aimed at predicting the crack path selec-
tion, have been proposed4–6 , and numerical procedures,
REFERENCES
based on these theories, have been developed. These the-
ories are either based on overall energy considerations 1 Lawn, B. (1993) Fracture of Brittle Solids. Cambridge University
or are local, referring to the crack tip stress analysis. We Press, Cambridge, UK.
postulate that the existing laws of crack path selection are 2 Broberg, K. B. (1999) Cracks and Fracture. Academic Press. San
inapplicable to single crystals. The crack path is first dic- Diego, USA.
3 Broek D. (1981) Elementary Engineering Fracture Mechanics.
tated by the preferred plane of propagation, namely, by the
Martinus Nijhoff, The Netherlands.
cleavage plane, and in certain conditions the crack will de- 4 Goldstein, R. V. and Salganik, R. L. (1974) Brittle fracture of
flect and follow a new rule. Preliminary experiments in our solids with arbitrary cracks. Int. J. Fracture. 10 507–527.
lab have shown that crack propagation paths in specimens 5 Cotterell, B. and Rice, J. R. (1980) Slightly curved or kinked
having the same geometry and loading scheme but made cracks. Int. J. Fracture. 16 155–169.


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40
40 D. S H E R M A N

6 Erdogen, F. and Sih, G. S. (1972) A special theory of crack 21 Sherman, D. (2005) Macroscopic and microscopic examination
propagation, In: Mechanics of Fracture, Vol. I, Noordhofe, of the relationship between crack velocity and path and
Leiden. Rayleigh surface wave speed in single crystal silicon. J. Mech.
7 Pérez, R. and Gumbsch, P. (2000) An ab initio study of the Phys. Solids. 53 2742–2757.
cleavage anisotropy in silicon. Acta Mater. 48, 22 Muhlstein, C. L., Brown, S. B. and Ritchie, R. O. (2001)
4517. High-cycle fatigue of single-crystal silicon thin films. J.
8 Spence, J. C. H., Huang, Y. M., Sankey, O. (1993) Lattice Microelectromechanical Systems. 10, 593–600.
trapping and surface reconstruction for silicon cleavage on 23 Coufal, H., Meyer, K., Grygier, R. K., et al. (1994) Precision
(111). Ab-Initio uantum molecular dynamics calculation. Acta measurement of the surface acoustic wave velocity on silicon
Metall. Mater. 41, 2815–2824. single crystals using optical excitation and detection. J. Acoust.
9 Griffith, A. A. (1920) The phenomena of rupture and flow in Soc. Am. 95, 1158.
solids. Phil. Trans. Roy. Soc. Lond. A221 163. 24 Xu, Y., Aizawa, T. (1999) Leaky pseudo surface wave on the
10 Irwin, G. R. (1958) Handbuch der Physik. In Fracture. Vol. VI, water-Si(110) interface. Phys. Lett. A. 260, 512.
(Edited by S. Flugge), Springer, Germany, 551. 25 Hauch, J. A. and Marder, M. P. (1998) Energy balance in
11 Thomson, R., Hsieh, C., Rana, V. (1971) Lattice trapping of dynamic fracture, investigated by a potential drop technique.
fracture cracks. J. Appl. Phys. 42, 3154. Int. J. Fract. 90, 133–151.
12 Sinclair, J. E., Lawn, B. R. (1972) Atomistic study of cracks in 26 Fratini, S., Pla, O., Gonzalez, P., et al. (2002) Energy radiation
diamond-structure crystals. Proc. Royal Soc. A 329, 83. of moving cracks. Phy. Rev. B 66, No. 104104.
13 Riedle, J., Gumbsch, P., Fischmeister, H. F. (1996) Cleavage 27 Heizler, S. I., Kessler D. A. and Levine, H. (2002) Mode-I
anisotropy in tungsten single crystals’. Phys. Rev. Lett. 76, fracture in a nonlinear lattice with viscoelastic forces. Phys. Rev.
pp. 3594–3597. E 66 No. 016126.
14 Freund, L. B. (1990) Dynamic Fracture Mechanics. Cambridge 28 Martin, T., Espanol, P. and Rubio M. A. (2005) Mechanisms for
University Press, Cambridge. dynamic crack branching in brittle elastic solids: Strain field
15 Stroh, A. N. (1957) Advances in physics. Phil. Mag. Supp. 6, kinematics and reflected surface waves. Phys. Rev. E 71 No.
418–465. 036202.
16 Marder, M. and Gross, S. (1995) Origin of crack tip instability. 29 Sherman, D. and Be’ery, I. (2003) The shape and the energies
J. Mech. Phys. Solids. 43, 1–48. of a dynamically propagating crack under bending. J. Mat. Res.
17 Hauch, J. A., Holland, D., Marder, M., et al. (1999) Dynamic 18 2379–2386.
fracture in single crystal silicon. Phys. Rev. Lett. 82, 30 Yuse, A. and Sano, M. (1993) Transition between crack patterns
3823–3826. in quenched glass plates. Nature 362, 329–331.
18 Cramer, T., Wanner, A. and Gumbsch, P. (2000) Energy 31 Deegan, R. D., Chheda, S., Patel, L., et al. (2003) Wavy and
dissipation and path instabilities in dynamic fracture of silicon rough cracks in silicon. Phys. Rev. E 67, Art. No. 066209.
single crystals. Phys. Rev. Lett. 85, 788. 32 Shilo, D., Sherman, D., Be’ery, I., et al. (2002) Large local
19 Madou, M. J. Fundamentals of Microfabrication. CRC Press, deflections of a dynamic crack front induced by intrinsic
New-York.1997. dislocations in brittle single crystals. Phys. Rev. Let. 89 No.
20 Sherman, D. and Be’ery, I. (2004) From crack deflection to 235504.
lattice vibrations – macro to atomistic examination of dynamic 33 Sherman, D. and Be’ery, I. (2004) Dislocations deflects and
cleavage fracture. J. Mech. Phys. Solids. 52 (2004) perturb dynamically propagating cracks. Phys. Rev. Lett. 93 No.
1743–1761. 265501.


c 2006 The Author. Journal compilation 
c 2006 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 30, 32–40

You might also like