Extreme Mechanics Letters 6 (2016) 31–36

Contents lists available at ScienceDirect

Extreme Mechanics Letters

journal homepage: www.elsevier.com/locate/eml

How does surface tension affect energy release rate of cracks

loaded in Mode I?
Chung-Yuen Hui a,∗ , Tianshu Liu a , Marie-Emeline Schwaab b
a
Field of Theoretical and Applied Mechanics, Department of Mechanical and Aerospace Engineering, Cornell University, Ithaca,
NY 14853, USA
b
Institut de Recherche en Génie Civil et Mécanique (GeM), UMR CNRS 6183, Ecole Centrale de Nantes, BP 92101,
44321 Nantes cedex 3, France

article info abstract

Article history: The role of surface stress on fracture of elastic solids has been studied by two group of
Received 20 October 2015 researchers with different conclusions. One group concludes that surface stress has no
Accepted 13 November 2015 effect on energy release rate except that the singularity of the crack tip field is stronger
Available online 22 November 2015
than the usual inverse square root singularity dictated by linear elastic fracture mechanics.
The other group concludes that the singularity of the stress field reduces to a logarithmic
Keywords:
singularity, thus implying that the local energy release rate is zero. In this letter we resolve
Surface tension
Crack
this paradox by examining the solution of a special case where surface stress is isotropic
Energy release rate and independent of surface strain. We show that surface tension resists crack growth
Stress intensity factor by lowering the applied energy release rate while retaining the inversely square root
Laplace pressure singularity of the elastic crack tip field. We demonstrate this idea by solving a perturbation
problem where the capillary length is much smaller than the crack length. A closed form
expression for the local energy release rate is obtained in the limit of small surface tension.
© 2015 Published by Elsevier Ltd.

1. Introduction
The characteristic length scale that controls the defor-
mation of a solid due to its surface tension σ is given by
the ratio of the surface tension to its shear modulus G. For
hard materials such as metals and ceramics, this ‘‘capil-
lary’’ length σ /G is smaller than atomic dimensions, hence
surface tension effect can be ignored. However, the cap-
illary length of soft materials such as hydrogels and elas-
tomers range from tens of nm to hundreds of µm. For this
class of materials, surface tension can drive shape change,
for example, a sharp corner in a soft material cannot re-
main sharp because of surface tension [1,2]. Surface ten-
sion can also flatten surfaces of structures made by replica
∗ Corresponding author.
E-mail address: ch45@cornell.edu (C.-Y. Hui).
http://dx.doi.org/10.1016/j.eml.2015.11.002
2352-4316/© 2015 Published by Elsevier Ltd.
molding [3,4] and cause Plateau instability in thin gel fil-
aments [5]. However, surface tension can resist deforma-
tion, for example, both experiments and theory has shown
that the contact mechanics of spheres and cylinders on
soft elastic substrates can be affected by solid surface ten-
sion [6–11]. Liu et al. [12] have shown that surface tension
induced Laplace pressure can cause closure of a crack in-
flated by hydrostatic pressure.
There are very few experimental studies examining the
role of surface stress in fracture since actual measurements
of surface stresses are very difficult to make. The first
theoretical study on the influence of surface stress on the
fracture of elastic solids were carried out by Thomson,
Chuang and Lin [13] in 1986 (hence forth referred as
TCL). In their model, they ignored the curvature of the
deformed crack faces, and the effect of surface stresses is
modeled as a line force acting at the crack tip. Their analysis
showed that even though the singularities due to the line
force are much higher than the typical inverse square root
32 C.-Y. Hui et al. / Extreme Mechanics Letters 6 (2016) 31–36
Fig. 1. Figure on left is a schematic of the geometry, where an undeformed semi-infinite plane strain Mode I crack occupying the negative x axis is subjected
to a remote far field controlled by the applied stress intensity factor KA . Figure on right shows the deformed crack face, which is subjected to a Laplace
pressure induced by local curvature of the deformed crack surface.
singularity in linear elastic fracture mechanics, they have
no effect on the crack tip energy release rate [see remarks
after Eq. (42) in their paper].
It is interesting to compare TCL’s result with the more
recent studies of Kim, Schiavone and Ru [14,15] (hence
forth refer as KSR). Using the surface constitutive model
of Gurtin and Murdoch [16], KSR [14] in 2010 formulated
a small strain theory of fracture to study the effect of
surface tension on fracture of a plane strain crack loaded
in Mode I and Mode II. Their numerical results showed
that the stress field directly ahead of a Mode I crack is
bounded, irrespective of the magnitude of the surface stress
(as long as it is not exactly zero). In a later paper [14], KSR
reexamined their numerical solution in greater details and
concluded that the stress field directly ahead of the crack
tip has a weaker logarithmic singularity. In both cases their
numerical results imply that the stress intensity factor (or
the local stress intensity factor Klocal = KA + KST ) is exactly
zero. In other words, the applied stress intensity factor KA is
canceled by the negative stress intensity factor KST caused
by the curvature induced Laplace pressure acting to close
the crack faces. Since a zero local stress intensity factor
implies that local energy release rate is also zero, their
result is the exact opposite of TCL’s. Furthermore, because
surface tension is represented as a line force, the stress field
near the crack tip in TCL model is more singular than the
inverse square root singularity of the classical theory and
this result also contradicts the result of KSR.
The goal of this letter is to resolve this paradox. To
simplify the analysis, we focus on Mode I cracks and the
surface stress σαβ is assumed to be isotropic and is a
constant independent of surface strain, that is, σ = σ I
where σ is the surface tension (force/length). To focus
attention on the crack tip, we consider the ‘‘Small Scale
Surface tension’’ problem in which the effect of surface
tension is confined to a region that is small with respect to
typical specimen dimensions. As in Small Scale Yielding in
classical fracture theory [17], the crack is semi-infinite and
lies on the negative x axis, as shown in Fig. 1. The boundary
condition is that at distances far from the crack tip, the
stress and deformation field approaches the usual inverse
square root singularity of the classical elasticity solution.
More details will be given in the next section.
The curvature induced Laplace pressure pL resisting the
opening of the crack (see Fig. 1) is related to the surface
tension σ and the curvature of the deformed crack faces κ
by:
v ′′
pL = σ κ κ=  3/2 , (1)
1 + (v ′ )2
where v is the crack opening displacement and a prime
denotes differentiation with respect to x. In KSR, v ′ is
assumed to be small everywhere and κ is approximated by:
κ = v ′′ . (2)
Due to this approximation, the governing equation de-
scribing the crack tip field is linear and the full machin-
ery of analytic function theory can be used to formulate
the crack problem. In contrast, TCL avoided this issue all
together by placing a line force at the crack tip—the curva-
ture induced Laplace pressure appears as a delta function
in their formulation [13].
It is easy to understand why Klocal had to be zero in
KSR’s analysis, that is, the local stress field near the crack
tip cannot have an inverse square root singularity. Indeed,
assuming Klocal is positive, the crack opening for the upper
crack face near the crack tip is:
2(1 − υ)Klocal √
v= √ −x , x<0 (3)
G 2π
where υ is the Poisson’s ratio and G the shear modulus.
Substituting (3) into κ = v ′′ gives
κ ∝ −Klocal |x|−3/2 , x → 0− (4)
Eqs. (4) and (1) imply that the Laplace pressure has a non-
integrable singularity; such a singular pressure field will in-
duce an infinite negative stress intensity factor at the crack
tip, which means that Klocal goes to negative infinity. This
is a contradiction to the original assumption that Klocal is
positive so the only possibility is Klocal = 0, which implies
that the stress is bounded or has a weaker singularity.
The fact that Klocal = 0 brings up the possibility a line
force can exist at the crack tip. Indeed, in linear elasticity
fracture mechanics, the crack tip deforms into a cusp shape
if the stress is bounded at the tip—the Dugdale-Barenblatt
 C.-Y. Hui et al. / Extreme Mechanics Letters 6 (2016) 31–36 33

model being a prime example [17–19]; whereas a logarith- The Laplace pressure is related to the curvature of the
mic singularity in stress implies the crack tip opens into a crack opening displacement v by (1). Substituting this into
wedge [see Appendix]. In both cases the curvature has a (6)–(7), (8) becomes:
delta function behavior which corresponds to existence of
2(1 − v)KA  (1 − ν)σ
a line force acting at the crack tip. Thus, the crack tip stress v(x) = −x/2π +
field in the KSR model can still be singular since the singu- G πG
√ √ 
0
v ′′ (t )

larity in curvature can cause a singular field similar to those −x + −t
studied by TCL. Specifically, an implicit assumption of KSR × 3/2 ln √ √ dt
1 + (v ′ (t ))2 −x − −t

−∞
is that curvature is non-singular; otherwise a concentrated
line force can act at the crack tip. x < 0. (9)
The key idea in this letter is that these difficulties can
For a given applied stress intensity factor, (9) is a non-
be resolved using the full curvature expression. To see this,
linear integral equation for the unknown crack opening
let us assume that the local crack tip field has form given
displacement v . Eq. (9) can be reduced to a more standard
by standard theory (3), then (1) implies that
form by differentiating both sides by x, resulting in:
v ′′ −π G2
(1 − v)KA (1 − ν)σ

κ x → 0− → ′ 3 = .
 
(5) −1
|v | (1 − v)2 Klocal
2 v ( x) = −
′
+ √ PV
G 2π x π G −x
 3
The key feature is that v ′′ and v ′ has exactly the same  √ 
0
−t v ′′ (t )
singularity, so the curvature is bounded at the crack tip. This × 3/2 dt x < 0 (10)
−∞ t − x 1 + (v ′ (t ))2

result implies that the curvature induced Laplace pressure
is finite everywhere along the crack face including the crack
tip. Thus, for sufficiently small surface tensions, the local where PV denotes principal value. More insight can be ob-
stress intensity factor is well defined and positive (hence tained by normalizing x by the elasto-capillary length and
there is a non-zero energy release rate). In addition, the the opening displacement v by the opening displacement
existence of finite curvature eliminates the possibility of of a traction free crack evaluated at a distance equal to the
a singular line force acting at the crack tip. elasto-capillary length behind the crack tip, i.e.,

πG 2(1 − v)KA (1 − v) σ

2. Formulation and analysis η= x, v= v̄. (11)
(1 − v)σ G 2π 2 G
To quantify our idea, we employed the usual small scale Using these normalized variables, the governing equation
yielding formulation where the crack is modeled as semi- (10) becomes:
infinite and lies on the negative x axis with its tip at x = 0.  0
√
For our case, the plastic zone is replaced by a zone where −1 1 −ξ
v̄˙ (η) = √ +√
surface curvature is significant. This formulation is valid in 2 −η −η −∞ ξ −η
the limit where the capillary length σ /G is much smaller
v̄(ξ
¨ )
 3/2 dξ η < 0
than the crack length a. For example, the shear modulus × (12a)
of a soft hydrogel is on the order of 100 to 1000 Pa, using ˙ ) 2
1 + β v̄(ξ

a crack length of 1 cm, σ /Ga ranges from 0.07 to 0.007,
where we have taken the surface tension of the hydrogel where a dot denotes differentiation with respect to the
to be equal to the surface tension of water, which is about normalized distance η and
0.07 N/m. Thus, for most soft materials, the region of
dominance of surface tension is expected to be localized 2(1 − υ)KA2
β≡ = 4 (JA /σ ) (12b)
at the crack tip. Let KA denote the applied stress intensity Gσ
factor. It is the stress intensity factor in the absence of is a dimensionless parameter. It is proportional to the ratio
surface tension. The displacement of the upper crack face of the energy release rate without surface tension, JA , to the
vA due to the applied KA field is surface tension. It is particularly interesting to note that if
2(1 − v)KA  the approximation κ ≈ v ′′ is used, (12a) becomes
vA = −x/2π x < 0. (6) √
G −1 1
 0
−ξ ¨
Surface tension σ caused a restraining pressure pL (x) to v̄˙ (η) = √ +√ v̄(ξ )dξ η < 0. (13)
2 −η −η −∞ ξ −η
close the crack, see (1). The crack face displacement vp due
to an arbitrary pressure field p acting on the crack faces is Note that there is no free parameter in (13). The absence of
well known [20]; setting p = pL (x) gives: β means that the near tip fields is universal, that is, increas-
√ √  ing or decreasing the surface tension could only change
(1 − ν) 0

 |x| + |t |  the region of asymptotic validity of the near tip fields, but
vp = pL (t ) ln  √
 √  dt . (7)
Gπ −∞ |x| − |t |  not its form. Our asymptotic analysis of (13) supports the
numerical results of KSR [14,15]—the normal stress σyy di-
Note that since the Laplace pressure is negative, vp < 0.
rectly ahead of the crack tip has a logarithmic singularity.
The actual crack opening displacement is the sum of (6)
This analysis is given in the Appendix for completeness.
and (7)
The local stress intensity factor Klocal can be expressed
v = vA + vp . (8) by adding the applied stress intensity factor to the stress
34 C.-Y. Hui et al. / Extreme Mechanics Letters 6 (2016) 31–36

intensity factor induced by the Laplace pressure. This can that the separation of perspective fracture surfaces directly
be done by examining the behavior of v ′ (x → 0+ ) ahead of a crack tip is resisted by atomic cohesive forces.
in (10) or by using a standard expression in fracture Since its introduction cohesive force model has been
mechanics [20]: used to study a wide variety of fracture phenomenon. In
the standard theory, the cohesive traction is a nonlinear
v ′′ (t )
  0
2 dt function of crack opening displacement. For our case,
Klocal = KA + σ √ . (14)
π −∞ [1 + v ′ (t )]3/2 −t the cohesive traction depends only on the curvature of
the crack face (or the first and second derivatives of the
In normalized form, (14) is crack opening displacement). Our model is mathematically
Klocal equivalent to KSR, except we retain the nonlinear terms in
= 1 + χ (β) , the evaluation of curvature. Retaining this nonlinear term
KA
smooth over the curvature field and avoids the dominance
0
v̄(ξ
¨ ) dξ

of surface tension over elasticity near the crack tip. The
χ (β) = 2 3/2 √ . (15)

2 −ξ result is that for small surface tensions or large β , the
−∞
1 + β v̄˙ (ξ ) local stress field has the same square root singularity, but
with a reduced stress intensity factor. The applied energy
Note that χ (β) is in general negative (v̄(ξ
¨ ) < 0) there is
release rate is also reduced by surface tension. Of course,
a reduction in stress intensity factor. In this letter we limit if β is small, that is, if the surface tension is large or
our study to small surface tension where β ≫ 1. In this the applied stress intensity factor small, then the Laplace
limit, the integral term in (12a) is small compared with the pressure could close the crack and eliminate the inverse
first term so that square singularity at the crack tip. We did not study this
−1 scenario, but a recent paper by Liu et al. [12] showed that
v̄˙ (η) = √ + O(1/β). (16) the local energy release of an internally pressurized crack
2 −η
subjected to large deformation can vanish or even become
Therefore a first order estimate of Klocal can be determined negative for sufficiently large surface tensions or small
by substituting v̄˙ (η) ≈ 2√−1
−η
in (16) into (15) and evalu- internal pressures.
ating χ (β). The resulting integral can be evaluated exactly Our approach can be readily applied to finite cracks in
and we found elastic solids, as long as the Green’s function for line load
acting on the crack faces is known. Consider for example a
Klocal 4
=1− . β ≫ 1. (17) plane strain crack of length 2a in an infinite elastic block
KA β subjected to remote tension σA . For this case, the crack
Since the energy release rate is proportional to the square opening displacement due to the Laplace pressure vp is
of the stress intensity factor, (17) is equivalent to related to the actual crack opening displacement v by
(1 − ν)σ a v ′′ (t )

vp =
 2
Jlocal 4 3/2
= 1− . β≫1 (18) Gπ −a 1 + (v ′ (t ))2

JA β
√
 a2 − x2 + √a2 − t 2 

where Jlocal is the energy release rate at the crack tip.
× ln  √ √  dt (19)
 
Eq. (18) indicates that surface tension reduces applied en-  a2 − x2 − a2 − t 2 
ergy release rate gradually and smoothly.
where we have used a standard result [20]. The crack
3. Discussion opening displacement vA due to the applied remote tension
ΣA is
Our model gives a satisfactory resolution of the differ- (1 − υ)ΣA 
ences in result between the work of TCL and KSR. To sum- vA = a2 − x 2 . (20)
marize, in TCL, the curvature of the deformed crack faces G
are ignored, the surface tension acts on the un-deformed The integral equation governing the unknown v is obtained
flat crack faces. The line force boundary condition corre- by combining (19)–(20):
sponds to the sudden change of slope at the crack tip. (1 − υ)ΣA 
In contrast, KSR imposed a closing pressure on the un- v(x) = a2 − x 2
G
deformed crack faces which is proportional to the cur-
(1 − ν)σ a v ′′ (t )

vature of the deformed crack faces. These two different + 3/2
Gπ −a 1 + (v ′ (t ))2

approaches gave very different results: in TCL, surface ten-
sion has no effect on the applied energy release rate, but in- √
 a2 − x2 + √a2 − t 2 

creases the stress singularity at the crack tip; while for KSR,
× ln  √ √  dt |x| < a. (21)
 
surface tension eliminates the local energy release rate by  a2 − x2 − a2 − t 2 
reducing the stress singularity at the crack tip.
Our approach is to treat the surface tension induced There is obvious limitation to the small strain approach
Laplace pressure as a cohesive traction acting on the crack used in this work, especially for very soft solids where the
faces. Cohesive traction (force) model was first introduced deformation can be exceeding large near the crack tip. An
by Barenblatt [19] to study brittle fracture. His idea is example of a crack subjected to very large deformation
 C.-Y. Hui et al. / Extreme Mechanics Letters 6 (2016) 31–36 35

can be found in Sun et al. [21]. The fact that we need to where the superscripts ± denote the values of φ as
retain the nonlinear term in curvature suggests that large z approaches the cut (−∞, 0] from above or below
deformation matters. Therefore, a consistent approach is to respectively. Also, the integral in (A.3b) is interpreted as a
carry out a fully nonlinear large deformation analysis such principal value integral. The asymptotic analysis is easiest
as those recently performed by Liu et al. [12]. The stress to carry out if we formulate (A.1) as a functional equation,
free reference configuration for a cracked elastic solid as shown below. Eq. (A.3b) implies that
can be obtained by imagining turning off surface tension √
0
−ξ f˙ (ξ )

(e.g. treating the crack surfaces with a self-assembled
π i φ + (η) + φ − (η) = dξ .
 
(A.4)
monolayer) in the absence of external load. Starting −∞ ξ −η
from this stress free configuration, imagine turning on
Substituting (A.4) into (A.1), (A.1) becomes,
the surface tension without applying any external load
(removing the monolayer). This sudden engagement of −1 πi  +
f (η) = φ (η) + φ − (η)

surface tension will immediately give a very concentrated √ −√
2 −η −η
line force at the crack tip. This line force will pull the
crack tip inwards and resulting in the opening the crack η ∈ (−∞, 0) . (A.5)
faces. Since the surface stresses depend on the deformed Now differentiate (A.5) with respect to η, it becomes
configuration, the line force will be smooth over by the
deformation, the curvature of the crack faces is expected −1 πi
f˙ (η) = φ (η) + φ − (η)
 + 
3/2
− 3/2
to be continuous. This picture is consistent with Liu et al.’s 4 (−η) 2 (−η)
simulation of a penny-shape crack in a neo-Hookean
π i d φ (η) + φ − (η)
 + 
solid [12]. − . (A.6)
Finally, we mention that the reduction of energy release (−η)1/2 dη
rate in our analysis is consistent with the earlier analysis of Substituting (A.6) into the RHS of (A.3a), we have
Wu [23] where an artificial finite tip radius was introduced
to avoid the singularity of curvature. This finite tip radius φ + (η) − φ − (η)
approach was also used in later works by Fu et al. [24] and πi

√ −1
φ (η) + φ − (η)
 + 
Wang et al. [25]. = −η −
4 (−η)3/2 2 (−η)3/2

π i d φ + (η) + φ − (η)

Acknowledgments −
(−η)1/2 dη
C.Y. Hui and T. Liu acknowledge support from the US 1 πi  +
φ (η) + φ − (η)

Department of Energy, Office of Basic Energy Science, = +
Division of Material Sciences and Engineering under Award 4η 2η
d φ (η) + φ − (η)
 + 
(DE-FG02-07ER46463). Marie-Emeline Schwaab was a
student from Ecole Centrale de Nantes who participated − πi . (A.7)
dη
in this research during her visit to Hui’s group at Cornell.
Her visit and work was supported by the Région Pays de la The form of (A.7) suggests that as η → 0− , the RHS of (A.7)
Loire. C.Y. Hui enjoyed the many discussions with Anand should dominate the LHS, so that the first order asymptotic
Jagota and Rong Long. behavior of φ can be examined by studying the equation

πi  + d φ + (η) + φ − (η)
 
1
φ (η) + φ (η) − π i
−

Appendix. Asymptotic behavior of (13) +
4η 2η dη
Let f (η) = v̇ (η), then (13) can be rewritten as: = 0 η → 0− . (A.8)
√ Let g (η) ≡ φ (η) + φ − (η) , then (A.7) implies that
 + 
0
−ξ f˙ (ξ )

−1 1
f (η) = √ −√ dξ . (A.1)
2 −η −η −∞ ξ −η dg g 1
− = η → 0− . (A.9)
Define dη 2η 4π iη
√
0
−ξ f˙ (ξ ) The general solution of (A.8) is

1
φ(z ) ≡ dξ z ̸∈ (−∞, 0] (A.2)
2π i −∞ ξ −z 1
g (η) ≡ φ + (η) + φ − (η) = − + iA (−η)1/2
 
(A.10)
where z = η + iζ is a complex variable. The function φ is 2π i
analytic in the entire complex z plane with the exception where A is an arbitrary real constant (this is because g is
of the cut (−∞, 0]. Using the Plemelj formulae, we have, purely imaginary according to (A.3b)). Substituting (A.10)
for η ∈ (−∞, 0), into (A.5), we have
√
φ + (η) − φ − (η) = −ηf˙ (η) πi
 
(A.3a) −1 1 1/2
f (η) = √ −√ − + iA (−η)
 0 √ ˙ 2 −η −η 2π i
1 −ξ f (ξ )
φ + (η) + φ − (η) = dξ (A.3b) = π A.
π i −∞ ξ − η (A.11)
36 C.-Y. Hui et al. / Extreme Mechanics Letters 6 (2016) 31–36

Thus, the asymptotic behavior of f as η → 0− is a constant, respectively. The asymptotic behavior of σ̄yy as η ap-
and since proaches the crack tip (η = 0) can be found using a result
of Muskhelishvili [22] where he showed that for v̄(η
˙ →
f (η) = v̄(η)
˙ ⇒ v̄(η) = π Aη η → 0− . (A.12)
0− ) → π A,
For non-zero negative A, the local shape of crack tip is a
σ̄yy η → 0+ = π A ln η.
 
wedge. (A.12) implies that the σyy (x) has a logarithmic (A.19)
singularity. For consistency, we should check that the LHS This result is consistent with a numerical finding of
of (A.7) is indeed small in comparison with terms of the KSR [15].
RHS as η → 0− . To do this, we solve (A.10) and found

1 iA √ References
φ=− − zln(z ). (A.13)
4π i 2π [1] C.-Y. Hui, A. Jagota, Y.Y. Lin, E.J. Kramer, Langmuir 18 (2002) 1394.
To check that (A.13) indeed satisfies (A.6), we note that [2] S. Mora, Y. Pomeau, Condens. Matter 27 (2015) 194112.
[3] A. Jagota, D. Paretkar, A. Ghatak, Phys. Rev. E 85 (2012) 051602.
1 iA √ [4] D. Paretkar, X. Xu, C.Y. Hui, A. Jagota, Soft Matter 10 (2014) 4084.
φ+ = − − −ηi[ln(−η) + π i] [5] S. Mora, T. Phou, J. Fromental, L. Pismen, Y. Pomeau, Phys. Rev. Lett.
4π i 2π 105 (2010) 214301.
(A.14) [6] R.W. Style, C. Hyland, R. Boltyanskiy, J.S. Wettlaufer, E.R. Dufresne,
1 iA √
φ− = − − −η(−i)[ln(−η) − π i]. Nat. Commun. 4 (2013) 2728.
4π i 2π [7] J.-M.Y. Carrillo, A.V. Dobrynin, Langmuir 28 (2012) 10881.
√ [8] Z. Cao, M.J. Stevens, A.V. Dobrynin, Macromolecules 47 (2014) 6515.
Therefore, φ (η) + φ (η) = −
+ − 1
2π i
+ iA −η. From (A.3a) [9] T. Salez, M. Benzaquen, É. Raphaël, Soft Matter 9 (2013) 10699.
and (A.14), we found [10] X. Xu, A. Jagota, C.-Y. Hui, Soft Matter 10 (2014) 4625.
[11] C.-Y. Hui, T. Liu, T. Salez, E. Raphael, A. Jagota, Proc. R. Soc. A 471
A√ (2014) 20140727.
φ + (η) − φ − (η) = −η ln (−η) as η → 0− (A.15) [12] T. Liu, R. Long, C.Y. Hui, Soft Matter 10 (2014) 7723.
π [13] T. Thomson, T.-J. Chuang, I.-H. Lin, Acta Metall. 34 (1986) 1143.
[14] C.I. Kim, P. Schiavone, C.Q. Ru, J. Elasticity 104 (2010) 397.
which means the term on the LHS of (A.7) can be neglected
[15] C.I. Kim, C.Q. Ru, P. Schiavone, Math. Mech. Solids 18 (2012) 59.
compared with any term on the RHS of (A.7). [16] M.E. Gurtin, A.I. Murdoch, Arch. Ration. Mech. Anal. 54 (1975) 291.
Our result (A.12) shows that v̄(η)
˙ is a constant. We [17] J.R. Rice, Mathematical analysis in the mechanics of fracture,
now show that σ̄yy (η) as η approaches the crack tip at in: H. Liebowitz (Ed.), Fracture: An Advanced Treatise (Vol. 2, Math-
ematical Fundamentals), Academic Press, N.Y, 1968, pp. 191–311.
η = 0 has a logarithmic singularity. To see this, recall (Chapter 3).
that any Mode I crack (with arbitrary normal traction on its [18] D.C. Dugdale, J. Mech. Phys. Solids 8 (1960) 100.
faces) can be viewed as a continuous distribution of edge [19] G.I. Barenblatt, Adv. Appl. Mech. 7 (1962) 55.
[20] H. Tada, P.C. Paris, G.R. Irwin, The Stress Analysis of Cracks Hand-
dislocations [17]; in particular, for an semi-infinite crack book, third ed., ASME publication, 2000, p. 87, ISBN: 10:0791801535.
lying on the negative x axis, [21] J.Y. Sun, X. Zhao, W.R. Illeperuma, O. Chaudhuri, K.H. Oh, D.J. Mooney,
J.J. Vlassak, Z. Suo, Nature 489 (2012) 133.
µ x
0
 ′
[22] N.I. Muskhelishvili, Singular Integral Equations, second ed., Dover

G
σyy (x) = dx′ x ̸= 0 (A.16) Publications, INC, New York, 1992, (Chapter 4).
2π (1 − υ) −∞ x − x′ [23] C. Wu, J. Mech. Phys. Solids 47 (1999) 2469.
[24] X.L. Fu, G.F. Wang, X.Q. Feng, Internat. J. Fract. 151 (2008) 95–106.
where µ x = −2dv/dx′ is the dislocation density. Using
 ′
[25] G.-F. Wang, X.-Q. Feng, T.-J. Wang, W. Gao, J. Appl. Mech. 75 (2008)
011001.
(11), the normalized form of (A.16) is:
0
µ̄ (ξ )

1
σ̄yy (η) = dξ (A.17)
2π −∞ η−ξ
where σ̄yy and µ̄ are the normalized stress and dislocation
density defined by
√
G β
σyy = σ̄yy (A.18a)
(1 − υ)
µ
µ̄ (ξ ) = −2v̄˙ (ξ ) = √ (A.18b)
β