Int J Fract (2007) 145:1–8

DOI 10.1007/s10704-007-9094-1

ORIGINAL PAPER

Crack shielding and amplification due to multiple

microcracks interacting with a macrocrack
Stefan Loehnert · Ted Belytschko

Received: 27 October 2006 / Accepted: 11 June 2007 / Published online: 7 August 2007
© Springer Science+Business Media B.V. 2007

Abstract We investigate the effect of crack shielding 1 Introduction

and amplification of various arrangements of micro-
cracks on the stress intensity factors of a macrocrack, In many brittle materials, the nucleation of many micro-
including large numbers of arbitrarily aligned micro- cracks can be observed around the tip of a propagating
cracks. The extended finite element method is used crack (Hoagland et al. 1973; Claussen 1976). These
for these studies. In some cases the numerical XFEM microcracks can have a significant influence on the
simulation provides results that are more accurate than propagation of the main crack. They can either cause
currently available analytical approximations because crack shielding or crack amplification. Crack shielding
the assumptions are less restrictive than those made reduces the stress intensity factors of the main crack
in obtaining analytical approximations. Stress intensity while crack amplification increases those values. These
factors for the tip of a macrocrack under the influence effects have been investigated primarily with exact ana-
of nearby microcracks are calculated under far field lytical methods for the few cases where this is possi-
mode 1 boundary conditions. For a microcrack aligned ble (Rose 1986a; Rubinstein 1985; Gong 1994), and
with the macrocrack the numerical results agree quite with analytical approximations under certain, at times
well with the analytically exact stress intensity factors. restrictive, assumptions (see e.g. Rose 1986b; Kacha-
The influence of the distance to the macrocrack tip and nov 1987; Kachanov et al. 1990; Chudnovsky and Wu
the rotation angle is investigated for unaligned micro- 1990; Chudnovsky and Wu 1991). For example, Rose
cracks, and it is shown in several examples with many 1986b assumes that the length of a microcrack is much
randomly distributed microcracks that the influence of smaller than its distance to the tip of the macrocrack,
those microcracks which are not in close proximity to which is equivalent to a point representation of the
the macrocrack tip is on the order of 5%. crack and not always a reasonable assumption. Ana-
lytical methods based on assumptions in three dimen-
Keywords Crack shielding · Crack amplification · sions were investigated e.g. by Laures and Kachanov
Finite elements · Extended finite element method 1991. Furthermore, studying the effect of multiple mi-
crocracks by analytical methods is difficult, see e.g.
Kachanov 1993. A more recent review of micro/mac-
rocrack interaction problems is given in Tamuzs and
Petrova (2002).
S. Loehnert (B) · T. Belytschko
Department of Mechanical Engineering, Northwestern
In this paper, the extended finite element method
University, 2145 Sheridan Road, Evanston, IL, 60208 USA (Belytschko and Black 1999; Moës et al. 1999) will
e-mail: [email protected] be used to investigate micro/macrocrack interaction

123
2 S. Loehnert, T. Belytschko

numerically under less restrictive conditions. In con- , ,

trast to the standard finite element methods, XFEM
allows for computations with an arbitrary number of
cracks without any meshing difficulties. Additionally
the stress singularities which occur at each crack tip are
included in the approximation field, so that very accu-
rate computations can be made with coarse meshes. The , ,
XFEM solution is at times more accurate than the solu-
tion obtained from an analytical method under certain Fig. 1 Reference element
assumptions. The extended finite element method has
been applied to multiple crack growth e.g. in Budyn et
where σ is the stress tensor. The linear strains are
al. 2004, but the interaction between many microcracks
was not investigated. 1

ε= grad(u) + (grad(u))T (2)
Here we show that the available analytical solutions 2
do not capture the full extent of stress intensity fac- where u is the displacement field. We consider isotropic
tor amplification. In fact, for cracks at angles to the materials, so the constitutive equations are
macrocrack, significantly higher amplification factors
σ = 2µε +
tr(ε)1 (3)
recur. We also examine the effect of large numbers
of randomly distributed microcracks in the vicinity of where µ and
are the Lamé parameters. The bound-
the macrocrack tip. Surprisingly, the effects are quite ary ∂ = ∂u ∪ ∂t of the domain  consists of a
minor, with both reduction and increases in the stress displacement boundary where
intensity on the order of 5% or less. u|∂u = u0 (4)
The outline of this paper is as follows. In Sect. 2, a
brief summary of the method is given. The governing and a traction boundary where
equations as well as the numerical procedure to account t|∂t = σ · n . (5)
for the discontinuity of the displacement field are pre-
sented, and the method to calculate the interaction inte- Since we employ the finite element method, we use the
gral and the stress intensity factors is described. In weak form of the equilibrium equation: find a kinemat-
Sect. 3, examples for a microcrack aligned with a mac- ically admissible u(x) such that
rocrack for two symmetrically positioned microcracks  
in the vicinity of the crack tip of the main crack and for σ (u) : ε(v) d − t · v d∂ = 0 (6)
a random distribution of multiple microcracks around a  ∂t
macrocrack are presented. For the example with the col- for all kinematically admissible v with v = 0 on ∂u .
linear cracks, the results for the stress intensity factors
are compared to analytical solutions. Since for the other
examples no exact analytical solutions were available, 2.1 Nodal enrichments
the numerical results are compared to existing qualita-
tive results in literature. Section 4 briefly summarizes The computations were performed by the extended finite
the results which could be found and gives some con- element method (see Moës et al. 1999). In this method,
clusions. the enrichment of the approximation field is able to
reflect the discontinuities within those elements which
are intersected by a crack (see Fig. 1). In addition, nodes
2 Summary of the numerical method belonging to elements which are completely intersected
by a crack and nodes touching a crack are enriched with
We consider static, linear elasticity in two dimensions a Heaviside step function. Thus the displacements are
in plane strain. Body forces are neglected. The equilib- approximated by
rium equations are 
4 
4
uh = N I (ξ ) u I + N I (ξ ) H (ξ ) b I (7)
div(σ ) = 0 (1) I =1 I =1

123
Crack shielding and amplification 3

Fig. 3 Domain for the evaluation of the interaction integral

Fig. 2 Relationship of nodal enrichments to crack geometry

The coordinates of the crack tip are xtip and ytip .
where The domain  is discretized with quadrilateral ele-
1 ments. Only the elements cut by the crack are enriched,
N I (ξ ) = (1 + ξ1 ξ1I )(1 + ξ2 ξ2I ) (8) so this is a local partition of unity for the enrichment.
4
are the standard shape functions, ξ are the reference The adjacent elements sometimes require blending ele-
coordinates, ξ I are the nodal reference coordinates (see ments (see e.g. Chessa et al. 2003). However, the error
Fig. 2) and made by not introducing blending elements is relatively
 small, so this special treatment is neglected here.
1 : ψ(ξ ) ≥ 0
H (ξ ) = Ĥ (ψ(ξ )) = (9)
−1 : ψ(ξ ) < 0
is the jump function depending on the sign of a levelset
function ψ(ξ ) which defines the geometry of the crack, 2.2 Calculation of the stress intensity factors
see Stolarska et al. 2001 for the use of levelsets in the
extended finite element method. The stress intensity factors are computed using a domain
If the crack ends in the interior of an element, the form of the interaction integral (see Moran and Shih
nodes belonging to this element are enriched by the 1987 or in the context of XFEM Belytschko and Black
crack tip enrichment functions 1999; Moës et al. 1999).
⎛ ⎞
4 
4 
4  (2) (1)

uh = N I (ξ ) u I + ⎝N I (ξ ) f j (x) b j I⎠ (10) (1,2) (1) ∂u i (2) ∂u i (1) (2)
I = σi j − σi j − σkl εkl δ1 j
I =1 I =1 j=1 ∂ x1 ∂ x1
I
with ∂q
√
ϕ 
d (17)
f 1 (r, ϕ) = r sin (11) ∂x j
√
2ϕ 
f 2 (r, ϕ) = r cos (12) Here, all measures with the superscript (1) are the ana-
√
ϕ2 lytical solutions for either mode 1 or mode 2 cracking
f 3 (r, ϕ) = r sin sin(ϕ) (13) and the measures with superscript (2) are the values

2
√ ϕ obtained by the numerical method under the given
f 4 (r, ϕ) = r cos sin(ϕ) (14) boundary conditions. The domain in which this inte-
2
and gral has to be evaluated is determined by the function
q(x). It is set to 0 on the outer boundary of this domain
r = (x − xtip )2 + (y − ytip )2 (15) and outside of this ring, and it is set to 1 on the inner
y − ytip boundary of the domain and inside of this ring (see
ϕ = arctan −α. (16)
x − xtip Fig. 3). Within the domain q(x) is given by the nodal

123
4 S. Loehnert, T. Belytschko

values q I and the standard shape functions N I


n
q(x) = N I (ξ ) q I . (18)
I =1

The elements for which q(x) is given by (18) are the

elements intersected by a circle with a given radius
r around the crack tip for which the stress intensity
factors are calculated. Since outside of this domain the
derivative ∂q( x)
∂ x ≡ 0, for the evaluation of the interac-
tion integral only this domain is of interest. Using the
interaction integral (17) the stress intensity factors can
be calculated by
E ∗ (mode1,2) E ∗ (mode2,2)
K1 = I and K 2 = I (19)
2 2
with
Fig. 4 Sketch of the geometry for a microcrack of length
plane stress : E ∗ = E, ν∗ = ν l = b − a aligned with a macrocrack
E ν . (20)
plane strain : E ∗ = , ν∗ =
1 − ν2 1−ν
This domain form of the interaction integral has the 3.1 Aligned cracks
disadvantage that a relatively fine mesh resolution is
required for closely separated cracks, since only one The purpose of the studies with a single microcrack
crack tip should be enclosed by the domain and at least were primarily to assess the performance of the numer-
one layer of elements which are not intersected by other ical method in this setting. A single microcrack aligned
cracks has to surround each crack tip. However, alter- with a macrocrack, as depicted in Fig. 4, was consid-
native forms of the contour integral with a contour ring ered. The domain was chosen to be a square with H =
remaining in very close proximity of the crack tip reveal B = 2. The macrocrack is of length B/2. Since the ana-
much worse results for the stress intensity factors. lytically exact results for this problem (see Rose 1986b
show that the stress intensity factors for the three crack
tips depend only on the ratio a/b (see Fig. 4), for the
numerical studies, two cases were considered.
3 Numerical studies
(1) The position of one of the microcrack tips (point
We first studied a single microcrack aligned with a mac- 3) is held fixed, and the near microcrack tip (point
rocrack and two microcracks arranged symmetrically 2) is moved. This leads to a variable crack length.
around the crack tip of a macrocrack. For all examples (2) The crack length is kept constant and the distance
mode 1 displacement boundary conditions (K 1 = 1) of the microcrack to the macrocrack tip is changed.
are applied on the entire boundary of the specimen.
r K1
 ∗
ϕ  ∗
The numerical results in Fig. 5 show that for the two
ux = 2π E∗ (2 −2ν ) cos 2 + (1 + ν ) cases the difference between the K 1 stress intensity fac-
sin ϕ2 sin(ϕ) on ∂ (21) tors for the three crack tips is very small. The difference
r K1
 ϕ  ∗
ϕ  
uy = mainly comes from fluctuations which arise because
2π E ∗ 4 sin 2 − (1 + ν ) cos 2 sin(ϕ)
the microcracks are close to the boundary where mode
The domain is chosen large enough so that the fluctu- 1 boundary conditions were applied. However, the
ations of the mode 1 conditions in the vicinity of the results are in good agreement with the analytical solu-
boundary due to the microcracks are small, and the far tion even when the ratio a/b is small. The results also
field mode 1 conditions are justified. All tests are com- show, that as for the analytical solution, the microcrack
puted in plane strain. causes crack amplification but not shielding.

123
Crack shielding and amplification 5

Table 1 Parameters for the test of one microcrack aligned at an

angle with the macrocrack
a a/b b l
0.05 0.13 0.3846 0.3346
0.05 0.2 0.25 0.2
0.05 0.25 0.2 0.15

Fig. 5 Stress intensity factors at the microcrack and macrocrack

tips for two aligned cracks

Fig. 7 K 1 stress intensity factor at the macrocrack tip as a func-

tion of the angle α

ratios a/b (see Table 1). Because the crack geometry

is not symmetric, the near tip field around the macro-
crack tip is not purely mode 1 and the mode 2 stress
intensity factor K 2 is not zero even though we apply
pure mode 1 displacement boundary conditions on the
entire boundary of the domain.
The mode 1 stress intensity factor as a function of the
Fig. 6 Sketch of a tilted microcrack directing to the macrocrack angle α is shown in Fig. 7. One can see that for the three
tip sets of parameters K 1 does not change significantly as
long as the angle α is smaller than 60◦ . Depending
on the ratio a/b it reaches its maximum at approx-
3.2 One microcrack aligned at an angle with the imately 35◦ to 42◦ . For larger angles, K 1 decreases
macrocrack and reaches the nominal value only in the limiting case
of α = 180◦ . Thus for this micro / macrocrack con-
We next investigate the effect of microcracks that are figuration the mode 1 stress intensity factor is always
not aligned with the macrocrack. This example has amplified. Figure 8 shows the mode 2 stress intensity
been studied e.g. by Rose 1986b analytically using a factor as a function of the angle α. One can see that
point representation for the microcrack. We varied the even though the crack geometry is not symmetric, the
angle α (see Fig. 6) and kept the distance of the near mi- mode 2 stress intensity factor is almost two orders of
crocrack tip to the macrocrack tip as well as the length magnitude smaller than the mode 1 stress intensity fac-
of the microcrack constant. The length and width of tor and thus of only minor importance. Rose 1986b
the square domain are chosen to be H = B = 2, and could obtain accurate results for this example only if
the macrocrack has a length of B/2. We investigate the length of the microcrack is much smaller than its
the effect of the microcrack on the mode 1 and mode 2 distance to the main crack tip. This limitation does not
stress intensity factor of the macrocrack tip for different exist for the numerical simulation.

123
6 S. Loehnert, T. Belytschko

Fig. 8 K 2 stress intensity factor at the macrocrack tip as a func-

tion of the angle α Fig. 10 Stress intensity factors as a function of the centerpoint
of the microcrack xc

l α

xc yc
H 1

2
3

Fig. 11 Stress intensity factor at point 1 for microcracks aligned

at an angle α with the macrocrack
B

Fig. 9 Sketch of the symmetric microcrack arrangement

position xc of the centerpoint of the microcracks in
horizontal direction. Depending on the position of the
3.3 Symmetric micro crack arrangement microcracks, either crack shielding or crack amplifica-
tion can be observed. It can also be seen from Fig. 10
We next investigate the effect of two microcracks of that when the distance between the microcracks and
length l symmetrically arranged around the tip of a the macrocrack is small, the crack shielding effect can
macrocrack. This example is similar to the case stud- be so strong that the stress intensity factor K 1 of the
ied by Rubinstein 1986. We consider different distances near crack tips of the microcracks is significantly larger
between the microcracks and the macrocrack tip as well than the stress intensity factor of the macrocrack tip.
as various angles between the microcracks and the mac- This would imply that damage then progresses through
rocrack (see Fig. 9). The domain is chosen to be a square growth of the microcrack.
with H = B = 2. For horizontal microcracks (α = Figure 11 shows the mode 1 stress intensity fac-
0◦ ), microcracks at three distances in vertical direction tor for the macrocrack for different alignment angles
are investigated: y = 0.031, y = 0.062 and y = 0.093. of the microcracks. The distance of the centerpoints
The length of the microcracks is held constant at l = of the microcracks to the main crack is held constant
0.1. at y = 0.062 and the alignment angles vary between
Figure 10 shows the mode 1 stress intensity fac- α = −60◦ and α = +60◦ . The length of the micro-
tor K 1 of the three crack tips as a function of the cracks is fixed at l = 0.1.

123
Crack shielding and amplification 7

Fig. 13 Histogram for the mode 1 stress intensity factor of the

macrocrack tip for 50 random distributions of microcracks

placing a straight crack at a random angle in each circle

Fig. 12 Arrangement of multiple microcracks
such that the middle of the crack is the center of the cir-
cle and the crack has a length equal to the diameter of
From Fig. 11 one can clearly see the significant the circle. Only the two microcracks in the close prox-
dependence of the crack shielding and amplification on imity to the macrocrack tip were unchanged for all com-
the rotation angle. Obviously, if one of the crack tips of putations. Recall that with only the two microcracks,
the microcracks gets very close to the crack tip of the the numerical result for the mode 1 stress intensity fac-
main crack, the interaction between the cracks becomes tor of the main crack is K 1 = 0.5085. For multiple
even more significant. For certain angles, apparently it cracks this value changes to K 1 = 0.4901 in average
is not possible anymore to clearly distinguish a range which is only 3.6% less. A histogram for the K 1 stress
of xc (the horizontal position of the center of the micro- intensity factors for the macrocrack tip obtained for 50
cracks) for which crack shielding occurs and a range different microstructures is shown in Fig. 13. From the
for which crack amplification occurs. It is possible that histogram it can be seen that the range of results in the
there exist multiple areas for crack shielding and crack macrocrack stress intensity factors is about 10%, but
amplification. its mean value, 0.0491, agrees quite closely with the
result obtained by just considering the nearest micro-
cracks. Thus the effects of additional microcracks are
3.4 Random micro crack distribution not substantial but also not insignificant. An image of
the σ yy stresses for one of the given random distribu-
To determine the influence of more distant microcracks tions of microcracks around a main crack is shown in
on the stress intensity factors of the macrocrack, we Fig. 14.
study models with multiple randomly distributed
microcracks of equal length. Two of the microcracks
are positioned symmetrically and horizontally with their 4 Conclusions
centerpoints at x = 0 and y = ±0.031. This enables
us to quantitatively compare the results for only two mi- The effect of microcrack orientation on crack amplifi-
crocracks obtained previously with the multiple cation and shielding has been studied. The results show
microcracks model. A sketch of the crack configura- that substantially more amplification of the stress inten-
tion is shown in Fig. 12. sity factor on the macrocrack can occur for noncollinear
For this study we performed 50 tests with different cracks than is predicted by the previous analytical solu-
random distributions of the microcracks. The random tions for collinear cracks. Similarly, for certain posi-
distribution of cracks were generated by creating a ran- tions of an adjacent microcrack, significant crack
dom field of equally sized non-overlapping circles and shielding has been shown.

123
8 S. Loehnert, T. Belytschko

3. Chessa J, Wang H, Belytschko T (2003) On the construc-

tion of blending elements for local partition of unity
enriched finite elements. Int J Numer Methods Eng
57(7):1015–1038
4. Chudnovsky A, Wu S (1990) Effect of crack-microcracks
interaction on energy release rates. Int J Fract 44: 43–56
5. Chudnovsky A, Wu S (1991) Elastic interaction of a crack
with a random array of microcracks. Int J Fract 49:123–140
6. Claussen N (1976) Fracture toughness of Al2 O3 with an
unstabilized ZrO2 dispersed phase. J Am Ceram Soc
59:49–51
7. Gong SX (1994) Microcrack interaction with a finite main
crack—an exact formulation. Int J Fract 66:R51–R56
8. Hoagland RG, Hahn GT, Rosenfield AR (1973) Influence
of microstructure on the fracture propagation in rock. Roc
Mech 5:77–106
9. Kachanov M (1987) Elastic solids with many cracks: A
simple method of analysis. Int J Fract 23:43
10. Kachanov M (1993) Elastic solids with many cracks and
related problems. In: Hutchinson J, Wu T (eds) Advances
in Applied Mechanics vol 30, pp 259–445
11. Kachanov M, Montagut E, Laures J (1990) Mechanics of
Fig. 14 Stresses σ yy for multiple microcracks crack-microcrack interactions. Mech Mater 10: 59–71
12. Laures J, Kachanov M (1991) Three-dimensional interac-
tions of a crack front with arrays of penny-shaped micro-
We have also investigated the effect of dense arrays cracks. Int J Fract 48:255–279
13. Moës N, Dolbow J, Belytschko T (1999) A finite element
of microcracks around a macrocrack tip. Even large method for crack growth without remeshing. Int J Numer
arrays of microcracks have little additional effect Methods Eng 46(1):131–150
beyond those which are in the immediate proximity 14. Moran B, Shih CF (1987) Crack tip and associated domain
of the crack. The overall effect of the additional micro- integrals from momentum and energy balance. Eng Fract
Mech 27(6):615–642
cracks is usually less than 5%, so analyses that account 15. Rose LRF (1986a) Effective fracture toughness of micro-
for only the nearest microcracks should generally be cracked materials. J Am Ceram Soc 69:212–214
adequate for determining amplification and shielding 16. Rose LRF (1986b) Microcrack interaction with a main
of the stress intensity factor of the macrocrack. crack. Int J Fract 31:233–242
17. Rubinstein AA (1985) Macrocrack interaction with semi-
infinite microcrack array. Int J Fract 27:113–119
Acknowledgements The support of the Army Research Office
18. Rubinstein AA (1986) Macrocrack-microdefect interac-
under grant W911NF-05-1-0049 is gratefully acknowledged.
tion. J Appl Mech, Trans ASME 53:505–510
19. Stolarska M, Chopp DL, Moës N, Belytschko
T (2001) Modelling crack growth by level sets in the
References extended finite element method. Int J Numer Methods Eng
51:943–960
1. Belytschko T, Black T (1999) Elastic crack growth in finite 20. Tamuzs VP, Petrova VE (2002) On macrocrack-microde-
elements with minimal remeshing. Int J Numer Methods fect interaction. Int Appl Mech 38:1157–1177
Eng 45(5):601–620
2. Budyn E, Zi G, Moës N, Belytschko T (2004) A method
for multiple crack growth in brittle materials without reme-
shing. Int J Numer Methods Eng 61:1741–1770

123
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.