You are on page 1of 16

Fracture Mechanics

Lecture Notes
Colin Meyer
Contents
1 Lecture on 28 January 2014

2 Lecture on 30 January 2014


2.1 Griffiths Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3
3

3 Lecture on 4 February 2014

4 Lecture on 11 February 2014


4.1 Small-Scale Yielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4
4

5 Lecture on 13 February 2014


5.1 Energy Release Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4
4

6 Lecture on 18 February 2014


6.1 Double Cantilever Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4
4

7 Lecture on 20 February 2014


7.1 Thin Film and Channel Cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5
5

8 Lecture on 25 February 2014


8.1 Singular Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 G-K Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6
6
7

9 Lecture on 27 February 2014

10 Lecture on 4 March 2014


10.1 Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2 Resistance Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7
7
8

11 Lecture on 6 March 2014


11.1 Loading Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8
8

12 Lecture on 11 March 2014


12.1 Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9
9

13 Lecture on 13 March 2014


13.1 Stress Corrosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9
9

14 Lecture on 14 March 2014


10
14.1 Rubber Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
15 Lecture on 1 April 2014
10
15.1 The J Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
15.1.1 Path Independence of the J integral . . . . . . . . . . . . . . . . . . . . . . . . . . 11
16 Lecture on 3 April 2014

12
1

17 Lecture on 8 April 2014


12
17.1 Elastic-Plastic Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
18 Lecture on 10 April 2014
12
18.1 Crack Tip Blunting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
19 Lecture on 15 April 2014

12

20 Lecture on 17 April 2014


12
20.1 Crack Bridging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
21 Lecture on 22 April 2014
12
21.1 Small-Scale Bridging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
22 Lecture on 24 April 2014
13
22.1 Fiber Pull Out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
23 Lecture on 29 April 2014
13
23.1 Large-Scale Bridging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
24 Lecture on 1 May 2014
14
24.1 Mixed Mode Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
25 Lecture on 8 May 2014
14
25.1 Spalling of Brittle Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
25.2 Interfacial Fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Lecture on 28 January 2014

This class is taught by Professor Zhigang Suo and given by the School of Engineering and Applied Science
at Harvard University. In the first class Professor Suo described the toughest hydrogel ever measured, see
Sun et al. (2012). This was used as a jumping off point to describe the meaning of Toughness. In mechanics,
especially fracture mechanics, toughness is the ability for a material to resist the growth of a crack. Physically
this means that a tough material is able to deform plastically and absorb energy without fracturing. Toughness
can also be thought of as the integral under the stress-strain curve from the onset of loading to the fracture
strain f . Consider a rubber sheet of length L that is pulled to a new length l, giving a stretch of =l/L.
The maximum stretch is about max 7. If a small incision is made in the side of the rubber and the
experiment repeated, the new maximum stretch on the order of max 2. Thus, the incision can knock down
the rupture stretch considerably! What makes a material tough is a question for the course.
The first material we will consider is Silica Glass, which we know to be elastic (if we remove the load the
body recovers, which is intrinsically associated with a tough material) and the fracture strain is very small,
f '0.1%. The Youngs modulus (or stiffness) of silica glass is on the order of EG '70 GPa and the strength,
S, which is the rupture stress, f , is S '100 MPa. The order of magnitude of pressure are: 106 Pa = 1 MPa
and 109 Pa = 1 GPa. Now consider a body made of glass under plane stress and there is a stress concentration
site. We are interested in the failure load. Using linear elasticity, we have that max =3Applied (for a circular
hole). Thus, to be safe we should make sure that max <S, so in this case the applied load cannot exceed a
third of the total strength as Applied <S/3 or Applied <S/C for a general stress concentration factor C. This
is very problematic; the strength S is measured experimentally and C is calculated from linear elasticity. The
problem is that if the strength of glass is measured experimentally the results vary by orders of magnitude.
Also, the stress concentration factor is not correct as it varies down to the atomic scale and depends on the
shape of the concentrator. Consider the stress concentration factor for an ellipse, C =1+2 ab . If we send b0,
then C . Inglis (1913) modified the ellipsoidal
stress concentration to express C in terms of the radius of
p
curvature so that C =max/Applied '2 a/, where =b2/a for an ellipse. The size of the crack is given
by a. The hypothesis here is that only the bulk size and local tip curvature of the flaw are important. The
2

sharpest flaw is the atomic scale, using this as the radius of curvature we have 1010 m. A representative
size of a flaw is a 106 m, the variability in the strength of glass can be attributed to the variability in
flaw size. This gives max/Applied '200. The max stress is called the theoretical strength, max =T h; the
strength of atomic bonds is 10 GPa and a rule of thumb is E/10. The experimental strength Exp is
closely approximated by Exp 'T h/C '10 GPa/200'100 MPa, which is the value often found.
From linear elasticity, we have that the stress concentration factor is given as max =CApplied, where C
is dimensionless, independent of Applied, and dependent on only the shape and not the size of the flaw. The
reason that max is linearly dependent on Applied comes from the linearity of the equations. The equations
of linear elasticity are
1
ij = (ui,j +uj,i),
2
ui,tt = ij,j +fi,
1+

ij =
ij kk ij .
E
2
There is no possible way to construct a factor that is dimensionless and dependent on Applied. The stress
concentration factor is independent of size because there is no intrinsic lengthscale within the equations of
linear elasticity and thus we cannot compare the dimensions of the flaw to an intrinsic lengthscale to achieve
a nondimensional factor. Furthermore, the fact that there is no intrinsic lengthscale is a general feature of
continuum mechanics, even for a nonlinear rheology or nonlinear momentum equation. Since the strength of is
very sensitive to the size of the flaw, clearly the strength cannot be predicted from a theory without a lengthscale!
2

Lecture on 30 January 2014

Observational evidence shows that the experimental strength of glass varies greatly with samples. This is
strong evidence that the ultimate strength is not a material property and it is then clear that it must depend
on the flaw size. It is problematic however because flaws are too small to observe and therefore, the strength
cannot be predicted from that. A British scientist Alan A. Griffith proposed a different picture.1
2.1

Griffiths Experiments

Griffith (1920) introduced the idea of surface energy into fracture mechanics. His idea was that atoms on the
crack surface have fewer bonds than those in the interior and as the crack grows the surface energy increases.
Thus, surface energy resists crack growth and the elastic energy decreases. Surface energy is what one
would can surface tension in fluid dynamics. Griffith was able to experimentally measure by heating glass
up to a semiliquid state.
3

Lecture on 4 February 2014

In this lecture we analyze the Griffith (1920) theory. Consider a crack of length 2a cut in a body of thickness
b and total width 2`, such that a  `, that is in uniaxial tension, . The reference elastic energy in the
system scales as
2
U0 a2b.
E
The surface energy scales as
U ab.
Hence the difference in elastic energy U is
2
U =4ab a2b,
E
where the correct constants have been inserted. This expression reflects the increase in surface area with the
introduction of a crack less the elastic energy. This curve is a parabola with roots a=0 and a=2a, where
a is the location of the maximum. We can find the maximum of this curve by taking its derivative, i.e.
1 dU
2
=0=42 a.
b da
E
The maximum occurs at a, which is given as
2E
a = 2 .

An interesting aside is that Griffith (1920) was presented to the Royal Society by Sir GI Taylor.

For a small crack, thermodynamics says that the energy must decrease, so the length of the crack decreases.
This is crack healing. For larger cracks, however, reducing energy occurs by increasing crack length.
This simple relationship is a fundamental equation in fracture mechanics. By rearranging it is commonly
written in three ways
r
2E
2a
2E
a= 2 ,
and =
.
=

a
2E
How is this
checked experimentally? Since is a material constant all that needs to be determined is
whether a=constant.
4
4.1

Lecture on 11 February 2014


Small-Scale Yielding

Following papers like Irwin (1948) and Orowan (1949), Griffiths theory was extended beyond the limited
application of glass to metals. The thought was that energy goes into the creation of new surfaces, giving
rise to the surface energy , but for metals energy is also spent generating a plastic zone around a crack.
Thus, we can define a combination of these two effects and call it . It is defined as
=2+wp,
where wp is the work done per unit area to create a plastic zone. Usually, for metals, wp  2. All that
was done was to replace the 2 in Griffith (1920) theory
by . This gives
r
E
=
.
a
In the case for glass, wp  2, and therefore this expression reduces to the Griffith (1920) theory. An
important assertion at this point is that is actually a material constant.
A measure of the plastic zone size is called rp. To satisfy what is called small-scale yielding, rp a. Glass
will always satisfy this as rp is extremely small.
5
5.1

Lecture on 13 February 2014


Energy Release Rate

The energy release rate is given as


U
G= ,
A
where U is the stored elastic energy and A is the crack area. Fracture occurs when G=, the fracture energy,
which is a material constant.
6
6.1

Lecture on 18 February 2014


Double Cantilever Beam

Obreimoff (1930) did some experiments and theory about splitting Mica. He modeled the system as a crack
of length c driven into a block of thickness 2H by a wedge giving an opening of and load P . Effectively
splitting the sample into two cantilever beams. For a single cantilever beam the deflection load relation is
P c3
=
,
(1)
2 3EI
where I is the second moment of area, with dimensions of length to fourth power. For a cross section B wide
and H tall, we have that
H 3B
I=
.
12
The stored elastic energy in the two beams is given
 as


1
P = P.
U=
2 2
4
Solving equation (1) for P , we have
3EI H 3BE
P=
=
.
2c3
16c3
Inserting this into the stored elastic energy gives
H 3BE2
U=
.
16c3
4

Now defining the crack area as A=cB, we can insert it to find


H 3B4E2
U=
.
16A3
The energy release rate is defined as
U 3 EH 32
=
.
G=
A 16 c4
There is a potential factor of 2 error in Suos notes online. It is interesting to note that as the crack length c
increases the energy release rate decreases, this is stable crack growth.
If instead of controlling , the load P was imposed,
can use the beam deflection to write
 we 
2
12 P c
G=
.
EH 3 B
Here we can see that the energy release grows as the crack grows, at least until G=, this is unstable crack
growth.
7
7.1

Lecture on 20 February 2014


Thin Film and Channel Cracks

Thin films can develop residual stress, one example coming from a thermal mismatch. This gives a stress
f =Ef T ,
which is independent of the film thickness. These stress can be very high but often does not fracture the
film. However, increasing the film thickness will lead to fracture. Think of long thin, i.e. channel cracks,
that are seen in paint. In the region right around the crack, a length in either direction that scales with
the film thickness hf , is released but a little further away is still loaded with f . The length of the fracture
is L. As the crack grows, it does so in a self-similar manner and after a length that scales with hf , a steady
state is reached. The elastic energy is then scales as
2
U h2f L+hf L.
E
The condition at fracture is that
2hf
,
=Z
E
where Z is the constant of proportionality named by Zhigang Suo.
This can also be done by considering the integral
Z hf
2
1
G=
(x)(x)dx.
hf
2hf 2
Here (x) is the stress far ahead of the crack and (x) is the opening in the wake. For the stress (x)=
and the displacement
4(1 2) q 2 2
(x)=
hf x .
E
Inserting this into the integral we find that
Z hf 2
q
2 4 (1 2 )
1
h2f x2dx.
G=
2hf h2f
E
Nondimensionalizing so that x=hf , gives
Z
22(1 2) h2f 1 p
G=
12d.
hf E
4 1
As the integral represents the half of a circle of radius unity, the area is /2, thus, we arrive at the result
2(1 2)hf
.
G=
4E
The dimensionless factor Z thus takes the value of Z = /4, where the Youngs modulus in the previous

expression is taken to be equally applicable for plane strain and plane stress, i.e. think of it as E.
Following the exact same procedure, the energy release rate for the Griffiths crack can be found. That
is, consider a crack opening displacement given as
4 2 2
(x)=
a x .
E
5

The energy release rate can be calculated analytically


as
Z
1 a
G=
(x)(x) dx.
2a a
Inserting this in, and nondimensionalizing x by writing
x=a, we have that
2 Z 1p
2 a
12d,
G=
E 1
where the integral works out to be /2, thus,
2a
.
G=
E
8 Lecture on 25 February 2014
8.1

Singular Stress

Here we are interested in the near tip field. If we think of a perfect mathematical crack which is traction free,
we can model the body as being infinite and the crack as being semi-infinite. The crack front is a straight
line and the crack faces are two planes. This means that any boundary value problem set up in this context
has no intrinsic lengthscale. We can consider the dimensional quantities to be ij , G, E, and r. We also have
the two dimensionless quantities and . Considering the first four, we have 4 parameters and 3 dimensions,
thus, a single -group may be formed:
ij2 r
1 =
.
GE
Writing this as a function of the other dimensionless
r variables we have that
EG
fij (,).
ij =
r
There are other possible dimensionless arrangements for the stress nondimensionalization, but linear elasticity
restricts stress to be proportional to the energy release rate (though not necessarily linearly). This is a fantastic
result because it shows that the stress is square-root singular.
An important relabeling of this result is to write

K = EG.
Now we solve the boundary value problem in linear elasticity to find fij (). There are three modes for a crack,
there is mode I which is standard opening, mode II which is shear, and mode III which is antiplane shear.
Here we consider that along z the crack is much less singular and that mode I is dominant. Hence, we can reduce
the problem to a 2D plane strain problem for the Airy function (r,). This gives the biharmonic equation
 2
2

1 1 2
4
=
+
+
=0.
r2 r r r2 2
The stress, in terms of , is given as


1 1 2
1
2
rr =
+ 2 2,
r =
, and = 2 .
r r r
r r
r
Following Williams (1957) mode I separation of variables we choose
(r,)=r+1f(),
and inserting gives
 2
 2

d
d
2
2
+(1)
+(+1) f()=0.
d2
d2
b
The solutions to this ODE are of the
, which gives
the algebraic equation
 2form f()=e
2  2
2
b +(1) b +(+1) f()=0.
This gives b=(1)i, (1+)i. Thus, we have that
f()=Acos[(+1)]+Bsin[(+1)]+Ccos[(1)]+Dsin[(1)].
The boundary conditions are that = r = 0 on = . Also, the problem is symmetric so B = D = 0.
Thus, we have
f()=Acos[(+1)]+Ccos[(1)],
and can apply the two boundary conditions to give an ugly eigenvalue problem for . Unfortunately, it
turns out that there are infinitely many admissible solutions. From our dimensional considerations we choose
=3/2. This then gives
 
C
3
= cos
.
2
r
6

Noting that at =0, this is related to K, we can see that 

K
=
cos3
.
2
2r
8.2 G-K Relation
Following Irwin, we can show how G and K are related to each other by doing a simple integral. Starting
with a crack in a reference state, called (1), if we allow it to grow by an amount b, to a state (2), we can
write the energy release rate as
U (1) U (2)
.
G=
b
The stress in the reference state and displacement in the current state
r as given as
(1)
(2)
K
8K
br
(1) =
and (2) =
.
E
2
2r
Now doing something crazy and imposing the stress in the reference state onto the displacement in the
current state, we can write G as
r
Z
Z
U (1) U (2) 1 b (1) (2)
1 b K (1) 8K (2) br

G=
=
dr =
dr.
b
2b 0
2b 0 2r E
2
Nondimensionalizing so that r =b, we have that Z r
8K (1)K (2) 1 1
d.
G=
4E

0
This integral is a form of the Beta function,Z B(x,y), and given as
1
(x)(y)
Beta(x,y)= tx1(1t)y1 dt=
,
(x+y)
0
where () is the Gamma function. Our integral has


 2
12 32
1
3
1 1

x= ,
y=
and B(x,y)=
=
= ,
2
2
(2)
2 2
2
where we have used the fact that ( +1)=(). Inserting this into our solution we have that
8K (1)K (2)
G=
.
8E
Now writing the reference and current state as identical, we have that
K 2 =EG,
which is the relation we set out to show.
9

Lecture on 27 February 2014

The K-annulus is the region which the square-root singular solution is valid. From this, an estimate of the
plastic zone size is
 
1 Kc
rp
,
2

where Kc = E is the critical stress intensity factor and Y is the yield strength.
10
10.1

Lecture on 4 March 2014


Composites

Consider the problem of finding the stress intensity for the residual stress of a fiber of radius R inducing
an edge-crack in the matrix of length a. We can use linear superposition of the Greens function to find
K. For a pair of loads a distance away on a crack ofslength 2c, the Greens function for K is given as
P
c+
K=
.
c c
Thus, changing our coordinates to the center of the fiber, we choose r =+R+a/2, that the crack edges are
at r =R and r =R+a, this is allowed for by setting c=a/2. Insert this into the Greens function, choosing
P = dr, and integrating yields
r
Z R+a
R2
rR

K=
dr,
r2 2a a+Rr
R
7

where we have used the expression


R2
.
2r2
Nondimensionalizing by choosing r =R and =a/R, we
can pull of the factors out of the integral to have
Z 1+ r
1
R
1
d.
K=
2 1+
2a 1
To get Wolfram Alpha to solve this integral it is helpful to make the substitution = u+1. Then the
integral becomes
r Z 1
r
r
K
1

=
du=
.
2
2 0 (u+1)
1u
2 2(1+) 23
R
Thus, we have that
r r
a
1
K
=
.
8 R 1+ a  32
R
R
This is the expression given in equation (8) of Lu et al. (1991).
=

10.2

Resistance Curve

The resistance curve or R-curve is a plot of the energy release rate G against the crack length a, see figure 1A.
A long time after crack initiation, the fracture energy is given by G=ss, which is a material constant. But
right at fracture initiation, there is a potentially different fracture energy, 0, which is also a material constant.
In fact, the entire R-curve is material dependent. All of this is under the assumption of small scale yielding.
The crack length to grow until steady state is give by L
ss , which should scale with the plastic zone size as
2
1 Kc
.
Lss rp
2 Y
11
11.1

Lecture on 6 March 2014


Loading Curve

For glass, the plastic zone size is very small and therefore we expect that Lss, will be very small, i.e. Lss 1
nm. Hence, a constant fracture energy, ss, is good for glass and the R-curve will look like a rectangle. We
can then write the energy release rate G as a function of the load and the crack length, so that G=G(,a).
G

II

ss

ss

a0

Lss

a0

Figure 1: A. Resistance curve. B. Loading and R curves superimposed. Curve I is the smallest load to
initiate stable fracture and curve II is the smallest load to initiate unstable fracture.
For a Griffith crack we have

2 a
G=
.
E
Thus, we can plot G versus a for different values of . For a Griffith crack, these will be straight lines in
the G-a plane, see figure 1B. The smallest load to initiate fracture is the one that, for an initial crack size
of a0, gives an energy release rate of 0, i.e.,
r
0E
=
.
a0
8

This crack cannot grow because, any growth would require G to be larger than the value of the loading
curve, see curve I in figure 1B. For this reason, this crack is referred to as stable. That is to say if +,
then the crack will grow a little but then stop.
On the other hand, consider the load that results in curve II in figure 1. If stress were to increase as
+, then crack would run away, but if stress decreased as , then the fracture would stop
because at a certain point G would be less than the loading curve. The location of this point in terms of
the stress and crack length is an important parameter because it represents the largest load that can be
applied to give a stable crack growth. The method of determining this point is by finding the location where
the loading curve is tangent to the R-curve, i.e.

G

=
.
a a a a
One parametrization of (aa0) is given as


aa0
=(ss 0)tanh
+0.
Lss
Using a Griffith crack, the tangency condition gives


2
2 a a0
= ,
(ss 0)sech
Lss
E
so that the critical crack length a is given as
s
!
2

a =a0 +Lssarcsech
.
E(ss 0)
Notice the importance of the initial crack length a0 in this analysis and in figure 1B: if the initial crack length
is larger then the entire R-curve shifts to the right and a lower applied load may lead to unstable crack growth.
12
12.1

Lecture on 11 March 2014


Fatigue

The main observation is that cyclic loading of a sample can cause fracture. Thus, we can plot a curve of the
amplitude of the cyclic load S and the number of cycles it takes to fail N. This curve starts out at U , the ultimate strength, and decreases down either to an endurance limit for most substances or essentially zero for some
(like Aluminum). The problem with this curve is that there is large variation in the number of cycles required to
fracture a certain sample (also, it is very laborious to construct). An innovation came from Paris et al. (1961),
which was to consider the energy release rate G. Even though it is an elastic quantity and fatigue is certainly
a plastic phenomenon, we can consider a K-annulus and work under the assumptions of small-scale yielding.
We can now examine how the crack
length a grows with each cycle N, Paris et al. (1961) found that it
scales as a power law with Kmax =S a, i.e.,
n n
da
n
=AKmax
=AS n 2 a 2 .
dN
Solving this ODE for a(N) gives
n 
n
1 n
1 n
2
2
a0 a
= 1 AS n 2 N.
2
Near failure we have that af a0, so we can write

n
n 
1 n
a0 2 = 1 AS n 2 Nf .
2
We can solve for the critical load S as
! n1
1 n
a0 2
S= n  n
.
2N
1
A
f
2
A common exponent that is used is n=4.
13
13.1

Lecture on 13 March 2014


Stress Corrosion

Observational evidence gives that we hang a weight on a sample of glass, where the weight is much lower
than the load needed to break the glass, after a while the glass will break. At first though one might think
that (i) creep, (ii) diffusion, or (iii) inertia play a role but it turns out that they do not. The only logical
explanation is that there is a reaction taking place. Consider a small flaw at the edge of the sample. At
9

the surface of the glass SiO2, if one of the oxygen molecules is replaced by an oxygen from H2O, then the
surface weakens. This is the process of stress corrosion.
Orowan (1944) modified Griffith (1920) theory by replacing in Griffiths expression by s, which includes
the effect of the water. The method of analysis is to put a crack of length a0 into a sample and watch
the length of the crack a(t) as a function of time. At some time tf , there is an asymptote and a diverges.
Considering small-scale yielding we can think of the crack-tip process as governed by G. Consequently, an
intuitive plot is that of velocity da/dt versus G. At the low end there is a threshold energy release rate called
Gth, approaching this point from above, the velocity drops quickly to zero. At a critical energy release rate
Gcr , the velocity diverges. The curve that connects these points is a material curve and is given as
2a
da
=V (G), for G=
,
dt
E
where V is some arbitrary function of the energy release
 rate
n and is the applied load. An example of this is
da
K
=B
.
(2)
dt
Kth
Here K is the stress intensity factor and is given as K = EG, and G is the Griffith energy release rate.
Inserting this into equation (2), gives
! n2
!
2
2
n
n
n
da
B
2
a0
a
a 2 =a 2 and thus
=
=t.
n n
2
2
dt
Kth
2n a
a2
0
At t=tf , the crack length diverges and the second term is negligible as compared to the first. Thus, we have
that

 n2
2
2a0
Kth
tf =
.
2n B n2 2a0
14
14.1

Lecture on 14 March 2014


Rubber Fracture

Rivlin and Thomas (1953) applied Griffiths theory to rubber by experimentally measuring the stored elastic
energy U. Then by taking a derivative with respect to the reference state they were able to determine the
energy release rate G.
15
15.1

Lecture on 1 April 2014


The J Integral

Previously in the course we have defined the energy release rate G as


U
G= ,
A
where U is the stored energy and A is the crack area.
Rice (1968) discovered that he could calculate
the energy release rate by doing an integral,
Z
u
J = W (ij )nx T ds,
x

where
Z ij
ij dij ,
W (ij )=
0

(3)

is the strain-energy density, nx is the component of the normal vector in the x-direction, and the traction T
is given as
Ti =ij nj .
Thus, we can also write the J integral as
Z
u
J = W T ds.
x

Notice that the derivative of the displacement is with respect to the coordinate x, i.e. it is taken in the
horizontal direction.
This form of the J integral is in the limit Zof small deformation. Another, equivalent form of the J integral is
W xi
J = W N1 Nk
dL.
Fik X1

10

Here the notation of finite deformation is used: X is the location (coordinates) of a material particle in the
reference state. x is the position of the same material particle in the deformed state. F is the deformation
gradient and defined as
xi
Fik =
.
Xk
Force balance in terms of finite deformation is written as
sik
=0,
Xk
where sik is the nominal stress and therefore the traction is written as
Ti =sik Nk .
Now the material model is
W
sik =
,
Fik
so this allows us to write the J integral as Z
J = W N1 TiFi1dL.

The first derivation we will do is to show that the finite deformation and small strain forms are equivalent.
Writing
ui
xi
ui =xi Xi implies
=
i1.
X1 X1
Rearranging and inserting this into the Zfinite deformation form of the J integral yields
ui
Tii1dL,
J = W N1 Nk sik
X1

where we have invoked the nominal stress and traction relationships. Then we can see from a balance of
forces that
Z
T1 dL=0.

Thus, we have that


Z

W N1 Nk sik

J=

ui
dL,
X1

which is just different notation for equation (3).


15.1.1

Path Independence of the J integral

Starting with equation (3),Zfor a closed loop , we can


 theorem
 to write the integral in the form
Z apply Greens
W

ui
ui
ds=

J = W n1 ij nj
ij
dxdy.
(4)
x1
xj
x
A x

Now using the chain rule on the strain-energy density gives


W W ij
ij
=
=ij
.
x ij x
x
Now since strain can be written in terms of displacement
we have

 that
ij ij
ui
uj
ij
=
+
.
(5)
x
2 xj x xix
Since the stress tensor is symmetric, i.e. ij =ji, and relabeling enables ij Aij =jiAji, means that equation
(5) simplifies to


ij

ui
ij
=
ij
,
x xj
x
where we have also used the fact, from a balance of forces, that
ij
=0.
xj
Inserting
this into equation (4),





Z gives that 
Z
Z
ui
W

ui

ui

ui
J = W n1 ij nj
ds=

ij
dxdy =
ij

ij
dxdy =0.
x1
xj
x
x
xj
x

A x
A xj
Thus, the integral J around any closed curve is zero. It is now fair easily to see that the integral between any
two points is path independent. It is also important to note that the area enclosed must be free of singularities.
11

16

Lecture on 3 April 2014

The value of the J integral is equal to the energy release rate G.


17
17.1

Lecture on 8 April 2014


Elastic-Plastic Fracture Mechanics

Here we can use the same theory as developed for linear elastic fracture mechanics except use the stress-strain
distribution determined from experiments.
18
18.1

Lecture on 10 April 2014


Crack Tip Blunting

McMeeking (1977) focused on the opening and blunting of the crack tip. He measured a length b as the
opening at the intersection of two lines (separated by 90) and the crack face, where this is centered at the
crack tip. By dimensional considerations we have that 
E
GY bf
.
Y
This blunting is an inherently plastic phenomenon and it turns out that hydrostatic pressure does not affect
plastic deformation. At the crack tip, the surface is constrained to be traction free but just inside the material
along the =0 line there can be hydrostatic pressure. Thus, the stress can increase beyond that of the yield
strength. The peak is on the order of 22 4Y and occurs at about r 3b.
19

Lecture on 15 April 2014

The J-annulus is an analogy to the K for elastic-plastic fracture mechanics. This was developed by Hutchinson
(1968) and Rice and Rosengren (1968).
20
20.1

Lecture on 17 April 2014


Crack Bridging

In a Russian publication, Barenblatt (1959) describes how the crack tip field can be modeled as the separation
and breaking of springs (Barenblatt, 1962).this is referred to as a cohesive zone model.
21
21.1

Lecture on 22 April 2014


Small-Scale Bridging

At the crack tip, extending over a length L, there is a local stress 0 that scales with or is equal to the yield
strength Y . The crack is small compared to any external lengthscales and the bridging zone length L is
much smaller than the crack length a. This is a perfect scenario to invoke small-scale yielding, which means
that we can represent the loading as an applied stress intensity KA. Using linear superposition, as the plastic
region is very small, we can represent the rest as linear elastic, and then solving a boundary value problem.
To do this, we use the Greens function
r
2 P
,
K=

where P is the applied load and is the location. We then superimpose
many of these loads by doing the integral
r
Z L
0
2
d =2
K=
0 L.

0
Thus, we have that
r
2
Ktip =KA 2
0 L.

Since the tip region is taken care of by the bridgingrmodel, we can set Ktip =0, and find that
2
KA =2
0 L.

12

22
22.1

Lecture on 24 April 2014


Fiber Pull Out

One aspect of composites that makes them so strong is that as a crack propagates through the matrix, the
fiber is able slide with respect to the matrix. During this time the fiber stays intact and the friction between
the matrix and the fiber is extremely important in determining the toughness of the composite.
23
23.1

Lecture on 29 April 2014


Large-Scale Bridging

Bridging is a traction-separation model at the crack tip. Dugdale (1960) published a paper with a nice little
experiment and an appropriate theory. This paper turned out to be foundational. The experiment that he
did was to pull a thin ductile sheet with a crack of length 2a and thickness t with an applied load A. What
he observed was necking in the region around the crack tip and from this measured a bridging zone length
L at the crack tip. Based on the opening of the crack , one can specify a traction model, which is =().
Dugdale (1960) had the bright idea of approximating this curve as a constant, the yield strength Y , which also
is often written as 0. Then using linear elasticity, he solved a boundaryvalue 
problem for Ktip, which gives
p
p
20
a
Ktip =A (a+L)
(a+L) arccos
.

a+L
Since the bridging zone takes care of the dynamics at the tip, it makes sense that Ktip = 0. Thus, we we
can solve for applied load as a function of the bridging
 zonelength L, as
2
a
A = arccos
.

a+L
Defining the nondimensional variables,
A
L
=
and L = ,
0
a
we can write our expression as


2
1
= arccos
.

1+L
Now we can examine the two limits. For L a, we have that =0 and for L a, this gives =1. This makes
sense because if there is almost no bridging zone, then the applied load must be very small. However, if the bridging zone is massive, so to must be the load, but since it is constrained to be less than the yield strength, =1.
Cottrell (1963) took this analysis and extended it to consider the opening at the inner edge of the bridging
zone, called tail. This can now be formed into a well-posed
value problem,which
was solved to give
 boundary

p
4A p
8
a
a+L
0
tail =
(a+L)2 a2
arccos
(a+L)2 a2 aln
.
E
E
a+L
a
From before we also have that


p
20 p
a
Ktip = A (a+L)
(a+L) arccos
=0.

a+L
At fracture, tail =0, and we can nondimensionalize both
to give
 expressions

p
0E
p
1
(1+L )2 1+ln(1+L ),
0 =
=
(1+L )2 1arccos
8a0
2
1+L


2
1
=
arccos
.

1+L
Inserting the second expression into the first we see that
0 =ln(1+L ).
We can also insert this back into the expression for Ktip =0, to find


2
= arccos e0 .

Griffiths estimate is
s
r
r
E
E800 2 20
A =
which implies =
=
.
a
20E02 00
Now, defining
1
a

A=
= 0E and G =
,
0
00
80
13

we have the two functions

r
n 1o
2 2G
2
A
and G =
.
BZ = arccos e

A
For small A , the bridging zone model gives, BZ =1. For the same limit, the Griffith model diverges. In the
other limit, A , the two models give identical results. Thus, a common simplified model is to use =1
for small A and use Griffiths model for larger A .

24
24.1

Lecture on 1 May 2014


Mixed Mode Fracture

A simple experiment is to cut a crack at an angle from the vertical and pull it in vertical uniaxial tension.
If =/2, then this is the Griffith (1920) scenario and the only mode of fracture at the crack tip is mode
I, i.e. KI , which is opening. At any other angle , there is also a shearing mode of fracture, mode II, or
KII . This shearing mode will cause the crack to move horizontally, or kink. Erdogan and Sih (1963) asked
the simple question: what is the kink angle? They define the kink angle as the angle between the the
new direction of crack growth and the old direction of crack growth. In other words, the total angle from
the vertical to the new crack is +. Using a homogeneous, isotropic, elastic medium under small scale
yielding, we can see that the stress field must be linearly proportional to KI and KII , and given as
KI I
KII II
ij (r,)=
ij ()+
ij ().
2r
2r
Solving the boundary value problem 
forI () and II
(), we have
 that

 

II
2
II
2
and ()=3sin
cos
.
()=cos
2
2
2
Experimentally, Erdogan and Sih (1963) varied and measured . Their theoretical insight was that the
material should fail in the direction of maximal stress, , and that this will not necessarily be along .
Decomposing the applied load into its along crack and perpendicular to crack components we have
=Asin2() and =Asin()cos().
Then, KI and KII are given as

KII
=cot().
KI = a=Asin2() a and KII =Asin()cos() a thus
KI
To find the angle of maximum hoop stress, we take our expression for and take the derivative with respect
to and set this equal to zero. Then wefindthat
2cot()

= p
.
tan
2
1+ 1+8cot2()
So that
)
(
2cot()
p
.
=2 arctan
1+ 1+8cot2()
Now we can use this to check some limits. For =/2, we have the Griffith (1920) crack and there should
be no kink, and this is true: =0. Now, as 0, 
the cotangent
function diverges, hence we are left with

1
=2 arctan =70.53.
2
25 Lecture on 8 May 2014
25.1

Spalling of Brittle Plates

Thouless et al. (1987) discovered experimentally that if a thin film of thickness h, under residual tension , is
placed on top of a substrate, then a crack will begin along the interface and then dive into the substrate.
After a little while of crack propagation, it will be parallel to the interface and at a depth d. Ignoring the
start-up details, we construct a simplified picture: a thin beam of thickness d, bent by a moment M at
the end, but also with a force P applied at the end. Now, both M and P lead to shearing, but in opposite
directions and therefore it is possible that KII could be zero. Writing the energy release rate G in terms of P
and M, we have that
P 2 6M 2
G=
+
.
2Ed Ed3
14

As an aside this can be written in terms of the original parameters by writing P =h and M =h(dh)/2.
The energy release rate G is then partitioned into KI and KII as
2
K 2 +KII
G= I
.
E
Due to linear elasticity, KI and KII must be linearly proportional to P and M as
P
M
M
P
KI =a +b
and KII =m n ,
d
d
d3
d3
where a,b,c,d are pure, positive numbers and cannot contain any dimensionless parameters and the minus in
the mode II comes from opposing shear of P and M. Now we can see that G depends on the square of KI
and KII and the square of P and M but without any mixed terms, thus,we can insert these expressions for
the stress intensities into G to find 
2 
2
P
M
M
P
P 2 6M 2
2
2

GE =KI +KII = a
+b
n
+ m
= + 3 .
2d
d
d
d
d3
d3
Thus, we have that
1
a2 +m2 = ,
abnm=0 and b2 +n2 =6.
2
These are three equations for four unknowns, so there is one free constant. Hence, they can be written as

1
1
a= cos(),
m= sin(),
b= 6 sin(), and n= 6 cos().
2
2
To find we must solve a boundary value problem, which can be done by using integral equations. The
result is that =52. Now setting KII =0, we have that


tan()
dh
mP d=Mn which is hd =h
.
2
2 3
So that

3
d=h
which implies d3.83h.
3tan()
25.2

Interfacial Fracture

Williams (1959) extended his earlier analysis of the crack tip singularity (Williams, 1957) to a crack at the
weak interface of two dissimilar materials. He found that at =0
Kri
22 +i12 =
,
2r
where K =K R +iK I is complex, so that K R and K I are the real and imaginary parts of K. Note that the
1
dimensions of K are [stess(length) 2 i]. The dimensionless parameter  is a function of the ratio of the
elastic constants for the two materials, i.e.
!
341 + 12
1
= ln
.
2
(342) 12 +1
The next thing to do is to find the relationship between G and K. This was done by Malyshev and Salganik
(1965). They found


1
1
|K|2
G= +
E1 E2 cosh2()
Now examining the character of the stress relation we note that
ri =eiln(r) =cos(ln(r))+isin(ln(r)) and K =|K|ei.
Combining the expressions and separating into real and imaginary parts we have that
|K|
|K|
22 =
cos(ln(r)+) and 12 =
sin(ln(r)+).
2r
2r
In practice the value of epsilon is small and thus, at leading order, we can ignore the oscillatory behavior.
Then, when considering effects at the crack tip use the full expression.

15

References
G. I. Barenblatt. On equilibrium cracks, formed in brittle fracture. Dokl. Akad. Nauk, 127:4750, 1959.
G. I. Barenblatt. The mathematical theory of equilibrium cracks in brittle fracture. Adv. Appl. Mech.,
7:55129, 1962.
A. H. Cottrell. Mechanics of fracture. Tewksbury Symposium on Fracture, pages 127, 1963.
D. S. Dugdale. Yielding of steel sheets containing slits. J. Mech. Phys. Solids, 8:100104, 1960.
F. Erdogan and G. C. Sih. On the crack extension in plates under plane loading and transverse shear. J.
Fluids Eng., 85(4):519525, 1963.
A. A. Griffith. The phenomena of rupture and flow in solids. Proc. R. Soc. London, Ser. A, 221:163198, 1920.
J. W. Hutchinson. Singular behavior at the end of a tensile crack in a hardening material. J. Mech. Phys.
Solids, 16:1831, 1968.
C. E. Inglis. Stresses in a plate due to the presence of cracks and sharp corners. Trans. I.N.A, XLIV:115, 1913.
G. R. Irwin. Fracturing of metals. In The Principles of Fracture Mechanics, pages 147166. ASM, Cleveland,
1948.
T. C. Lu, J. Yang, Z. Suo, A. G. Evans, R. Hecht, and R. Mehrabian. Matrix cracking in intermetallic
composites caused by thermal expansion mismatch. Acta Metall. Mater., 39:1883 1890, 1991.
B. M. Malyshev and R. L. Salganik. The strength of adhesive joints using the theory of crack. Int. J. Fract.
Mech., 1:114128, 1965.
R. M. McMeeking. Finite deformation analysis of crack-tip opening in elastic-plastic materials and implications
for fracture. J. Mech. Phys. Solids, 25(5):357 381, 1977.
J. W. Obreimoff. The splitting strength of mica. Proc. R. Soc. London, Ser. A, 127:290297, 1930.
E Orowan. The fatigue of glass under stress. Nature, 154:341343, 1944.
E Orowan. Fracture and strength of solids. Rep. Prog. Phys., 12:185, 1949.
P. C. Paris, M. P. Gomez, and W. E. Anderson. A rational analytic theory of fatigue. Trend Eng., 13:914, 1961.
J. R. Rice. A path independent integral and the approximate analysis of strain concentration by notches
and cracks. J. App. Mech., 35:379386, 1968.
J. R. Rice and G. F. Rosengren. Plane strain deformation near a crack tip in a power-law hardening material.
J. Mech. Phys. Solids, 16:112, 1968.
R. S. Rivlin and A. G. Thomas. Rupture of rubber. i. characteristic energy for tearing. J. Polym. Sci.,
10(3):291318, 1953.
Jeong-Yun Sun, Xuanhe Zhao, Widusha R. K. Illeperuma, Ovijit Chaudhuri, Kyu Hwan Oh, David J. Mooney,
Joost J. Vlassak, and Zhigang Suo. Highly stretchable and tough hydrogels. Nature, 489:133136, 2012.
M. D. Thouless, A. G. Evans, M. F. Ashby, and J. W. Hutchinson. The edge cracking and spalling of brittle
plates. Acta Metall., 35:13331341, 1987.
M. L. Williams. On the the stress distribution at the base of a stationary crack. J. Appl. Mech., 24:109114,
1957.
M. L. Williams. The stress around a fault or crack in dissimilar media. Bull. Seismol. Soc. Am., 49:199204,
1959.
16

You might also like