ES 247 Fracture Mechanics http://imechanica.

org/node/7448 Zhigang Suo

Fatigue

Fatigue. Subject to monotonic bending, a metallic clip deforms, but may

not break. Subject to cyclic bending, however, the clip will break after some
cycles. Fracture caused by cyclic loads is known as fatigue. Examples of cyclic
loads include vibration, landing and taking off of an airplane, pressurization and
depressurization of a fuselage.
While all materials suffer fatigue, the effect is particularly pronounced for
ductile materials, such as metals and glassy polymers. For such a ductile
material, fracture may occur after a small number of cycles, when cyclic loads
repeatedly cause large-scale plastic deformation. In most engineering structures,
however, loads are kept small so that the structures deform elastically. Evan so, a
crack may initiate somewhere in such a structure, and extend by a small amount
during each cycle of the load. The structure remains elastic, except for a small
zone around the front of the crack. When the crack reaches a critical size, the
structure breaks by rapid fracture.
Fatigue may be divided into two phases:
• the initiation of a crack and
• the extension of the crack.

Approaches to the study of fatigue. Fatigue is of great practical

importance. Fatigue is also an extremely complex phenomenon. Its study has
gone on well over a century. Although much is known, most basic issues remain
unresolved. In particular, no approach exists to predict fatigue from first
principles. Qualitative pictures of various fatigue mechanisms are often too
vague to be of much use.
Fatigue has been studies at several scales:
• Atomic scale. Track accumulative activities such as dislocation generation
and organization of patterns.
• Microstructural scale. Fatigue striation. Persistent slip bands. The effect
of grains and inclusions.
• Component scale. In this lecture, we will focus on this scale. The Paris
approach is now used in engineering design.

The subject of fatigue itself can fill an entire course and more. In this course,
we will give one lecture on the subject, focusing on a single aspect: the
application of fracture mechanics to fatigue.

Frequency and amplitude of the cyclic load. When a metal is

subject to a cyclic load, the cycle-to-fracture is nearly independent of the
frequency of the load, but depends on the amplitudes of the load.

The S-N curve. This method had been in use before the work of Paris.
The stress fluctuates as a function of time. For the time being, consider the case

3/12/14 1 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

that the stress fluctuates within a constant range. Denote the maximum stress by
σ max , and the minimum stress by σ min . The range of the stress is
Δσ = σ max − σ min .
The mean stress is
σ + σ min
σ m = max .
2

S-N curve
σ x
TS
loading amplitude, S

x Experimental data
x ____ Curve through the data points to
x guide the eye.

x
x
σ x x
EL

2 4 6 8
1/4 10 10 10 10
cycle-to-fracture, N f

Experiment: Subject a bar to a cyclic stress. To be specific, cycles are

taken to be of equal tension and compression. Denote the amplitude of the stress
by S = σ max = −σ min . For a given amplitude S, cycle the bar until it fractures, and
record the cycle-to-fracture, N f . This experiment produces a data point (S, Nf).
Repeat the experiment for many amplitudes and bars, each producing a data
point (S, Nf). The final result is a curve, the S-N curve. Here are some
observations:
• Ultimate tensile strength. If the amplitude of the stress equals the
ultimate tensile strength, S = σ TS , the fracture occurs upon first loading,
N f = 1/4 .
• Endurance limit. For a given material (e.g., a steel), there may exist an
amplitude of the stress, called the endurance limit, σ EL . When the
amplitude is below the endurance limit, S < σ EL , the bar can sustain
infinite number of cycles without fracture, N f → ∞ . Aluminum does not
have an endurance limit: under cyclic load of even very small amplitude,

3/12/14 2 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

aluminum will fail after a finite number of cycles. Let me know if you find
an endurance limit for humans. I’d love to be immortal even if it means
doing sub-critical amount of work every day.
• Cycle-to-fracture. When the load amplitude falls between the ultimate
tensile strength and the endurance limit, σ EL < S < σ TS , the cycle-to-
fracture N f decreases as the load amplitude increases.

The use of the S-N curve in design. The S-N curve is specific to each
material. Once it is measured experimentally for a given material, one can use
the curve to answer design questions. Two types of questions are answered with
a S-N curve.
• Know the amplitude of stress, and predict the cycle-to-fracture.
• Prescribe a cycle-to-fracture, and determine the allowable amplitude of
stress.

Difficulties in using the S-N curve

• Large scatter in the S-N curve. For a prescribed amplitude of stress, the
experimentally measured cycle-to-fracture varies greatly from sample to
sample.
• To generate each data point on the S-N curve, one has to fracture a
sample. Many samples are needed to generate a single S-N curve, along
with statistics.
• Complexity in loads: multiaxial stress, time-history of stress, stress
concentration.

Use fracture mechanics to study the extension of a crack under

cyclic load. The Paris approach assumes that a crack pre-exists in a body.
Under a cyclic load, the crack grows a little bit each cycle. When the crack is long
enough, the body fractures.
The stress intensity factor is the loading parameter on a crack. For
example, for a crack of length 2a in an infinite panel subject to remote stress σ ,
the stress intensity factor is
K = σ πa .
When the applied stress σ cycles with time, so does
the stress intensity factor. The crack extension each
cycle is much smaller than the length of the crack, so
that
K max = σ max πa ,
and
K min = σ min πa .

Recall the slogan: the same K-history, the

same crack-tip process. The stress intensity factor is

3/12/14 3 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

the only messenger between the external load and the crack-tip process, however
complex this process is. Following Paris, we focus on a particular type of K-
history. While the stress intensity factor varies within each cycle, both Kmax and
Kmin vary slowly from cycle to cycle. Paris hypothesized that the crack extension
per cycle, da/ dN , is a function of the instantaneous Kmax and Kmin . For
example, the function may be written as
da
= f (ΔK , R ),
dN
where ΔK = K max − K min and R = K min / K max . This function is a material property.
For a given material, the function can be measured experimentally using a single
sample.
To prove this hypothesis, Paris et al. (1961) plotted the experimental data
of da/dN vs. ΔK for two materials from three independent studies. The data
were collected using specimens show in the figure. Included here is one of their
plots. They wrote:

It is worthy of special note that the data on these curves are from
three independent investigators, using many specimen sizes, i.e.,
widths from 1.8 to 12 in., thickness from 0.032 to 0.102 in., and
lengths from 5 to 35 in. The testing frequencies varied from 50 to
2000 cpm, and the maximum stresses on the gross area varied
from 6 to 30 ksi. On each graph, the materials are both clad
metals and bare metals. Therefore the correlation shown is surly
more than coincidental.

3/12/14 4 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

Paris also plotted the experimental data of da/dN vs. ΔK on a double

logarithmic basis. The figure shown here is taken from Paris’s PhD dissertation
(1962 Lehigh University). The straight line of such a plot suggests that the
experimental data fit a power law.
da m
= A(ΔK ) .
dN
A representative value for the exponent is m = 4. The growth also depends on R.
Fatigue crack growth rate vanishes if R =1, and increases as R reduces from 1.
The curve sketched in the figure is for a specific value of R, say, R = 0.1.
The above equation is now known as the Paris law. The equation is a
rather restrictive expression of Paris’s original hypothesis.

3/12/14 5 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

Using the Paris law in design. In a large component, a small crack of

initial size ai has been detected. The component is subject to a cyclic stress
between 0 and S. The component undergoes fast fracture when the crack size
reaches a f . How many cycles will the component last? Note that
ΔK = S πa .
The Paris law becomes
da
dN
(
= A S πa . ) m

This is an ordinary differential equation that governs the function a(N). For this
material, the exponent m = 4. Rearranging the above, we have
da
= dN .
Aπ 2 S 4 a 2

3/12/14 6 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

An integration from ai to a f gives the number of cycles to fast fracture:

1 ⎛ 1 1 ⎞⎟
Nf = ⎜ − .
Aπ 2 S 4 ⎜⎝ ai a f ⎟⎠
Main observations:
• The formula provides an explanation for the S-N curve.
• Typically, the initial crack size is much smaller than the final size, ai «a f , so
that the term involving a f is negligible.
• The S-N curve depends on the initial crack size. If the component has a large
initial flaw, the lifetime will be much reduced.
• Also, using the S-N curve without the knowledge of the initial flaw can be
dangerous

Threshold. Let us return to the experimental data of the crack extension

per cycle as a function of the stress intensity factor. The main features of the data
are indicted in the figure: threshold, fast fracture, and what in between.

threshold

fast fracture
log # %
" da $
dN

Paris law
da
= A(ΔK )m
dN

log (Δ K)

3/12/14 7 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

Lifetime prediction: division of labor. J.D. Achenbach,

Quantitative nondestructive evaluation. International Journal of Solids and
Structures 37, 13-37 (2000).

Crack size inspection Load history

σ (N )
ai

Elasticity Solution Fatigue test

da m
K = κσ a = A(ΔK )
dN

Remaining life
Nf

Historical Notes

The Griffith (1921) paper. A.A. Griffith, The phenomena of rupture

and flow in solids. Philosophical Transactions of the Royal Society of London, A
221, 163-198 (1921). In the Introduction of the paper, Griffith motivated his
study by describing fatigue of ductile metals. He wrote

The original object of the work, which was carried out at the Royal
Aircraft Establishment, was the discovery of the effect of surface
treatment—such as, for instance, filing, grinding or polishing—on the
strength of metallic machine parts subjected to alternating or repeated
loads. In the case of steel, and some other metals in common use, the
results of fatigue tests indicated that the range of alternating stress
which could be permanently sustained by the material was smaller
than the range within which it was sensibly elastic, after being
subjected to a great number of reversals.

The Thomas (1958) paper. A.G. Thomas, Rupture of rubber. V. Cut

growth in Natural rubber vulcanizates. This is the first paper in literature on the
application of the Griffith approach to the growth of a crack under cyclic load.

3/12/14 8 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

Thomas represented the cyclic applied load by a cyclic energy release rate. He
found that the extension of the length of a crack per cycle is proportional to the
amplitude of the energy release rate squared.

The Paris et al. (1961) paper. P.C. Paris, M.P. Gomez and W.E.
Anderson, A rational analytic theory of fatigue. The Trend in Engineering 13, 9-14
(1961). When I taught the course in 2010, I went online and found that the Trend
in Engineering is the alumni newsletter of the College of engineering, of the
University of Washington (http://www.engr.washington.edu/news/trend.html).
I could not find this paper online. John Hutchinson offered to write to Paul Paris
for a copy of the paper, which Paris sent by airmail. A scanned copy of the paper
is attached with this lecture.
The extension of a crack in a body under a cyclic load can be characterized
by using fracture mechanics, as described by Paris et al. (1961). You should begin
by reading this original paper. The paper is lucid and focused. I know of no
better description of the essential idea. Of the paper Paris (1998) would recount:

The initial reaction for this young researcher at that time was that we
were on to something big and important to the safety of aircraft and
other structures as well. It was time to tell the world about it, so a
paper was written clearly showing the correlation of data from three
independent sources on two materials. The tools of Fracture
Mechanics analysis needed to apply lab data to structural
configurations were also illustrated. Well, that paper was very
promptly rejected by three of the world’s leading journals. All of the
reviewers simply stated that “no elastic parameter, e.g. K, could
possibly correlate fatigue cracking rates because plasticity was a
dominant feature”. They proceeded to somehow totally disregard the
facts clearly demonstrated by the data!

Indeed, fatigue is inherently an inelastic phenomenon. It does sound

logical that an elastic parameter will have nothing to say about fatigue. At this
stage in our course I hope that you can see through the fallacy of this faulty logic.

References

P.C. Paris, Fracture mechanics and fatigue: a historical perspective. Fatigue &
Fracture of Engineering Materials & Structures 21, 535-540 (1998).

D.A. Lados and P.C. Paris, Parameters and key trends affecting fatigue crack
growth—A tribute to Professor Arthur J. McEvily’s contributions. Materials
Science and Engineering A468-470, 70-73 (2007).

M.F. Ashby and D.R.H. Jones, Engineering Materials 1, Pergamon Press, New
York. Chapter 15 provides a background of fatigue.

3/12/14 9 Fatigue
ES 247 Fracture Mechanics http://imechanica.org/node/7448 Zhigang Suo

R.I. Stephens, A. Fatemi, R.R. Stephens, H.O. Fuchs, Metal Fatigue in

Engineering, 2nd Edition, Wiley 2001.

S. Suresh, Fatigue of Materials, Cambridge University Press. 2nd edition (1998).

3/12/14 10 Fatigue