1

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Microstructure-Properties: II
Fatigue
27-302
Lecture 9
Fall, 2002
Prof. A. D. Rollett

Materials Tetrahedron
Processing
Performance
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Microstructure

Properties

Objective
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

The objective of this lecture is to explain the

phenomenon of fatigue and also to show how
resistance to fatigue failure depends on
microstructure.
For 27-302, Fall 2002: this slide set contains
more material than can be covered in the
time available. Slides that contain material
over and above that expected for this course
are marked *.

References
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Mechanical Behavior of Materials (2000), T. H.

Courtney, McGraw-Hill, Boston.
Phase transformations in metals and alloys, D.A.
Porter, & K.E. Easterling, Chapman & Hall.
Materials Principles & Practice, Butterworth
Heinemann, Edited by C. Newey & G. Weaver.
Mechanical Metallurgy, McGrawHill, G.E. Dieter, 3rd
Ed.
Light Alloys (1996), I.J. Polmear, Wiley, 3rd Ed.
Hull, D. and D. J. Bacon (1984). Introduction to
Dislocations. Oxford, UK, Pergamon.

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

sa :=
sm :=

Alternating stress
Mean stress
R := Stress ratio
e :=
strain
Nf :=
number of cycles to failure
A := Amplitude ratio
epl :=
Plastic strain amplitude
eel :=
Elastic strain amplitude
K := Proportionality constant, cyclic stress-strain
n := Exponent in cyclic stress-strain
c :=
Exponent in Coffin-Manson Eq.;
also, crack length
E :=
Youngs modulus
b :=
exponent in Basquin Eq.
m :=
exponent in Paris Law
K :=
Stress intensity
K :=
Stress intensity amplitude
a :=
crack length

Notation

Fatigue
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Fatigue is the name given to failure in response to

alternating loads (as opposed to monotonic
straining).
Instead of measuring the resistance to fatigue
failure through an upper limit to strain (as in
ductility), the typical measure of fatigue resistance
is expressed in terms of numbers of cycles to
failure. For a given number of cycles (required in
an application), sometimes the stress (that can be
safely endured by the material) is specified.

Fatigue: general characteristics

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Primary design criterion in rotating parts.

Fatigue as a name for the phenomenon based on the
notion of a material becoming tired, i.e. failing at
less than its nominal strength.
Cyclical strain (stress) leads to fatigue failure.
Occurs in metals and polymers but rarely in
ceramics.
Also an issue for static parts, e.g. bridges.
Cyclic loading stress limit<static stress capability.

Fatigue: general characteristics

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Most applications of structural materials involve cyclic

loading; any net tensile stress leads to fatigue.
Fatigue failure surfaces have three characteristic
features: [see next slide, also Courtney figs. 12.1, 12.2]
A (near-)surface defect as the origin of the crack
Striations corresponding to slow, intermittent crack growth
Dull, fibrous brittle fracture surface (rapid growth).

Life of structural components generally limited by

cyclic loading, not static strength.
Most environmental factors shorten life.

S-N Curves
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

S-N [stress-number of cycles to failure] curve defines

locus of cycles-to-failure for given cyclic stress.
Rotating-beam fatigue test is standard; also
alternating tension-compression.
[Hertzberg]
Plot stress versus the
log(number of cycles
to failure), log(Nf).
[see next slide,
also Courtney figs. 12.8, 12.9]

For frequencies < 200Hz,

metals are insensitive to
frequency; fatigue life in
polymers is frequency
dependent.

10

Fatigue testing, S-N curve

sa
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects

smean 3 > smean 2 > smean 1

smean 1
smean 2
smean 3

The greater the number of

cycles in the loading history,
the smaller the stress that
the material can withstand
without failure.

log Nf

Note the presence of a

fatigue limit in many
steels and its absence
in aluminum alloys.

Design
[Dieter]

11

Endurance Limits
Some materials exhibit endurance limits, i.e.
a stress below which the life is infinite: [fig. 12.8]
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Steels typically show an endurance limit, = 40% of

yield; this is typically associated with the presence
of a solute (carbon, nitrogen) that pines
dislocations and prevents dislocation motion at
small displacements or strains (which is apparent
in an upper yield point).
Aluminum alloys do not show endurance limits;
this is related to the absence of dislocation-pinning
solutes.
At large Nf, the lifetime is dominated by nucleation.
Therefore strengthening the surface (shot peening) is
beneficial to delay crack nucleation and extend life.

12

Fatigue fracture
surface
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

[Hertzberg]

13

Fatigue crack stages

Stage 1
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Stage 2

[Dieter]

14

Fatigue Crack Propagation

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Crack Nucleation stress intensification at crack tip.

Stress intensity crack propagation (growth);
- stage I growth on shear planes (45),
strong influence of microstructure [Courtney: fig.12.3a]
- stage II growth normal to tensile load (90)
weak influence of microstructure [Courtney: fig.12.3b].
Crack propagation catastrophic, or ductile failure
at crack length dependent on boundary conditions,
fracture toughness.

15

Fatigue Crack Nucleation

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Flaws, cracks, voids can all act as crack nucleation

sites, especially at the surface.
Therefore, smooth surfaces increase the time to
nucleation; notches, stress risers decrease fatigue
life.
Dislocation activity (slip) can also nucleate fatigue
cracks.

16

Dislocation Slip
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Crack Nucleation

Dislocation slip -> tendency to localize slip in

bands. [see slide 10, also Courtney fig. 12.3]
Persistent Slip Bands (PSBs) characteristic of
cyclic strains.
Slip Bands -> extrusion at free surface. [see next slide
for fig. from Murakami et al.]

Extrusions -> intrusions and crack nucleation.

17

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Slip steps
and the
stress-strain
loop

18

Design Philosophy: Damage Tolerant

Design

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

S-N (stress-cycles) curves = basic characterization.

Old Design Philosophy = Infinite Life design: accept
empirical information about fatigue life (S-N curves);
apply a (large!) safety factor; retire components or
assemblies at the pre-set life limit, e.g. Nf=107.
*Crack Growth Rate characterization ->
*Modern Design Philosophy (Air Force, not Navy
carriers!) = Damage Tolerant design: accept
presence of cracks in components. Determine life
based on prediction of crack growth rate.

19

Definitions: Stress Ratios

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Alternating Stress s a
Mean stress sm = (smax +smin)/2.
Pure sine wave Mean stress=0.
Stress ratio R = smax/smin.
For sm = 0, R=-1
Amplitude ratio A = (1-R)/(1+R).
Statistical approach shows significant
distribution in Nf for given stress.

See Courtney fig. 12.6; also following slide.

20

Alternating Stress Diagrams

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design
[Dieter]

21

Mean Stress
Alternating stress sa = (smax-smin)/2.
Raising the mean stress (sm) decreases Nf. [see slide 19,
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

also Courtney fig. 12.9]

Various relations between R = 0 limit and the ultimate

(or yield) stress are known as Soderberg (linear to
yield stress), Goodman (linear to ultimate) and
Gerber (parabolic to ultimate). [Courtney, fig. 12.10, problem
12.3]

sa

endurance limit at zero mean stress

s me an

s a s fat
1

tensile strength

tensile strength
smean

22

Cyclic strain vs. cyclic stress

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Cyclic strain control complements cyclic

stress characterization: applicable to thermal
fatigue, or fixed displacement conditions.
Cyclic stress-strain testing defined by a
controlled strain range, epl. [see next slide,
Courtney, figs. 12.24,12.25]

Soft, annealed metals tend to harden;

strengthened metals tend to soften.
Thus, many materials tend towards a fixed
cycle, i.e. constant stress, strain amplitudes.

23

Cyclic stress-strain curve

Objective
Crack
Initiation

[Courtney]

S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Large number of cycles typically needed to reach

asymptotic hysteresis loop (~100).
Softening or hardening possible. [fig. 12.26]

24

Cyclic stress-strain
Wavy-slip materials
generally reach asymptote
in cyclic stress-strain:
planar slip materials (e.g.
Objective
brass) exhibit history
Crack
dependence.
Initiation
Cyclic stress-strain curve
S-N
defined by the extrema,
curves
i.e. the tips of the
Cyclic
hysteresis loops. [Courtney
stress-strn
Crack
Propagate

Microstr.
effects
Design

fig. 12.27]

Cyclic stress-strain curves

tend to lie below those for
monotonic tensile tests.
Polymers tend to soften in
cyclic straining.

[Courtney]

25

Cyclic Strain Control

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Strain is a more logical independent variable

for characterization of fatigue. [fig. 12.11]
Define an elastic strain range as eel = s/E.
Define a plastic strain range, epl.
Typically observe a change in slope between
the elastic and plastic regimes. [fig. 12.12]
Low cycle fatigue (small Nf) dominated by
plastic strain: high cycle fatigue (large Nf)
dominated by elastic strain.

26

Strain control
of fatigue
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

[Courtney]

27

Cyclic Strain control: low cycle

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Constitutive relation s K e n
for cyclic stress-strain:
n 0.1-0.2
Fatigue life: Coffin Manson relation:
e p
2

ef 2N f

ef ~ true fracture strain; close to tensile

ductility
c -0.5 to -0.7
c = -1/(1+5n); large n longer life.

28

Cyclic Strain control: high cycle

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

For elastic-dominated strains

e e
at high cycles, adapt
sa E
sf 2N b
Basquins equation:
2
Intercept on strain axis of extrapolated
elastic line = sf/E.
High cycle = elastic strain control:
slope (in elastic regime) = b = n/(1+5n) [Courtney, fig. 12.13]
The high cycle fatigue strength, sf,
scales with the yield stress high
strength good in high-cycle

29

Strain amplitude - cycles

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

[Courtney]

30

Total strain (plastic+elastic) life

Low cycle = plastic control: slope = c
Add the elastic and plastic strains.
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

e el e pl sf

2N f
2
2
2
E

ef 2N f

Cross-over between elastic and plastic control is

typically at Nf = 103 cycles.
Ductility useful for low-cycle; strength for high cycle
Examples of Maraging steel for high cycle
endurance, annealed 4340 for low cycle fatigue
strength.

31

Fatigue Crack Propagation

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Crack Length := a.
Number of cycles := N
Crack Growth Rate := da/dN
Amplitude of Stress Intensity := K = sc.
Define three stages of crack growth, I, II and III,
in a plot of da/dN versus K.
Stage II crack growth: application of linear elastic fracture
mechanics.
Can consider the crack growth rate to be related to the applied
stress intensity.
Crack growth rate somewhat insensitive to R (if R<0) in Stage II
[fig. 12.16, 12.18b]

Environmental effects can be dramatic, e.g. H in Fe, in

increasing crack growth rates.

32

Fatigue Crack Propagation

Three stages of crack
growth, I, II and III.
Stage I: transition to a
Objective
finite crack growth rate
from no propagation
Crack
below a threshold value
Initiation
of K.
S-N
Stage II: power law
curves
dependence of crack
Cyclic
growth rate on K.
stress-strn
Stage III: acceleration
Crack
of growth rate with K,
Propagate
approaching
Microstr.
catastrophic fracture.

da/dN
I
II

Kc

III

effects

Design

Kth

33

*Paris Law
Paris Law:
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

dc
m
A(K)
dN

m ~ 3 (steel); m ~ 4 (aluminum).
Crack nucleation ignored!
Threshold ~ Stage I
The threshold represents an endurance
limit.
For ceramics, threshold is close to KIC.
Crack growth rate increases with R (for
R>0). [fig. 12.18a]

34

*Striations- mechanism
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Striations occur by development of slip bands

in each cycle, followed by tip blunting,
followed by closure.
Can integrate the growth rate to obtain cycles
as related to cyclic stress-strain behavior. [Eqs.
12.6-12.8]

cf

dc
NII
c0 dc / dN

cf

NII

c0

dc
m
m
A s c

35

*Striations, contd.
Provided that m>2 and is constant, can integrate.
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

A1s m 1m / 2 1m / 2
N II
c0
cf
(m / 2) 1

If the initial crack length is much less than the final

length, c0<cf, then approximate thus:
A1s m 1m / 2
N II
c0
(m / 2) 1

Can use this to predict fatigue life based on known

crack

36

*Damage Tolerant Design

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Calculate expected growth rates from dc/dN

data.
Perform NDE on all flight-critical components.
If crack is found, calculate the expected life of
the component.
Replace, rebuild if too close to life limit.
Endurance limits.

37

Geometrical effects
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Notches decrease fatigue life through stress

concentration.
Increasing specimen size lowers fatigue life.
Surface roughness lowers life, again through stress
concentration.
Moderate compressive stress at the surface
increases life (shot peening); it is harder to nucleate a
crack when the local stress state opposes crack
opening.
Corrosive environment lowers life; corrosion either
increases the rate at which material is removed from
the crack tip and/or it produces material on the crack
surfaces that forces the crack open (e.g. oxidation).
Failure mechanisms

38

Microstructure-Fatigue Relationships
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

What are the important issues in microstructurefatigue relationships?

Answer: three major factors.
1: geometry of the specimen (previous slide); anything on the
surface that is a site of stress concentration will promote
crack formation (shorten the time required for nucleation of
cracks).
2: defects in the material; anything inside the material that can
reduce the stress and/or strain required to nucleate a crack
(shorten the time required for nucleation of cracks).
3: dislocation slip characteristics; if dislocation glide is confined
to particular slip planes (called planar slip) then dislocations
can pile up at any grain boundary or phase boundary. The
head of the pile-up is a stress concentration which can
initiate a crack.

39

Microstructure affects Crack Nucleation

The main effect of
microstructure (defects,
surface treatment, etc.)
Objective
is almost all in the low
stress intensity regime,
Crack
i.e. Stage I. Defects,
Initiation
for example, make it
S-N
easier to nucleate a
curves
crack, which translates
into a lower threshold
Cyclic
for crack propagation
stress-strn
(Kth).
Crack
Propagate Microstructure also
affects fracture
Microstr.
toughness and
effects
therefore Stage III.

da/dN
I
II

Kc

III

Design

Kth

40

Defects in Materials
Descriptions of defects in materials at the sophomore level
focuses, appropriately on intrinsic defects (vacancies,
dislocations). For the materials engineer, however, defects
Objective
include extrinsic defects such as voids, inclusions, grain
boundary films, and other types of undesirable second phases.
Crack
Initiation Voids are introduced either by gas evolution in solidification or
by incomplete sintering in powder consolidation.
S-N
Inclusions are second phases entrained in a material during
curves
solidification. In metals, inclusions are generally oxides from the
Cyclic
surface of the metal melt, or a slag.
stress-strn
Grain boundary films are common in ceramics as glassy films
Crack
from impurities.
Propagate
In aluminum alloys, there is a hierachy of names for second
Microstr.
phase particles; inclusions are unwanted oxides (e.g. Al2O3);
effects
dispersoids are intermetallic particles that, once precipitated, are
thermodynamically stable (e.g. AlFeSi compounds); precipitates
Design
are intermetallic particles that can be dissolved or precipiated
depending on temperature (e.g. AlCu compounds).

41

Metallurgical Control: fine particles

Objective
Crack
Initiation
S-N
curves

Tendency to localization of flow is deleterious to the

initiation of fatigue cracks, e.g. Al-7050 with nonshearable vs. shearable precipitates (Stage I in a
da/dN plot). Also Al-Cu-Mg with shearable
precipitates but non-shearable dispersoids, vs. only
shearable ppts.

Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

graph courtesy of J.
Staley, Alcoa

42

Coarse particle effect on fatigue

Objective
Crack
Initiation

Inclusions nucleate cracks cleanliness (w.r.t.

coarse particles) improves fatigue life, e.g. 7475
improved by lower Fe+Si compared to 7075:
0.12Fe in 7475, compared to 0.5Fe in 7075;
0.1Si in 7475, compared to 0.4Si in 7075.

S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

graph courtesy of J.
Staley, Alcoa

43

Alloy steel heat treatment

Objective
Crack
Initiation

Increasing hardness tends to raise the endurance

limit for high cycle fatigue. This is largely a function
of the resistance to fatigue crack formation (Stage I in
a plot of da/dN).

S-N
curves

Mobile solutes that pin

dislocations fatigue
limit, e.g. carbon in steel

Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

[Dieter]

44

Casting porosity affects fatigue

Gravity cast
versus
squeeze cast
versus
wrought
Al-7010

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects

Design

[Polmear]

Casting tends to result in porosity. Pores are effective sites for

nucleation of fatigue cracks. Castings thus tend to have lower
fatigue resistance (as measured by S-N curves) than wrought
materials.
Casting technologies, such as squeeze casting, that reduce porosity
tend to eliminate this difference.

45

Titanium alloys
[Polmear]
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn

Crack
Propagate

Microstr.
effects
Design

For many Ti alloys, the proportion of hcp (alpha) and bcc (beta) phases
depends strongly on the heat treatment. Cooling from the two-phase region
results in a two-phase structure, as Polmears example, 6.7a. Rapid cooling
from above the transus in the single phase (beta) region results in a twophase microstructure with Widmansttten laths of (martensitic) alpha in a beta
matrix, 6.7b.
The fatigue properties of the two-phase structure are significantly better than
the Widmansttten structure (more resistance to fatigue crack formation).
The alloy in this example is IM834, Ti-5.5Al-4Sn-4Zr-0.3Mo-1Nb-0.35Si-0.6C.

46

*Design Considerations
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

If crack growth rates are normalized by the elastic

modulus, then material dependence is mostly
removed! [Courtney fig. 12.20]
Can distinguish between intrinsic fatigue [use Eq.
12.4 for combined elastic, plastic strain range] for
small crack sizes and extrinsic fatigue [use Eq. 12.6
for crack growth rate controlled] at longer crack
lengths. [fig. 12.21.]
Inspection of design charts, fig. 12.22, shows that
ceramics sensitive to crack propagation (high
endurance limit in relation to fatigue threshold).

47

*Design Considerations: 2
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Metals show a higher fatigue threshold in

relation to their endurance limit. PMMA and
Mg are at the lower end of the toughness
range in their class. [Courtney fig. 12.22]
Also interesting to compare fracture
toughness with fatigue threshold. [Courtney fig.
12.23]

Note that ceramics are almost on ratio=1 line,

whereas metals tend to lie well below, i.e.
fatigue is more significant criterion.

48

*Fatigue
property map

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design
[Courtney]

*Fatigue
property map

49

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design
[Courtney]

50

*Variable Stress/Strain Histories

Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

When the stress/strain history is

stochastically varying, a rule for combining
portions of fatigue life is needed.
Palmgren-Miner Rule is useful: ni is the
number of cycles at each stress level, and Nfi
is the failure point for that stress.
[Ex. Problem 12.2]

ni
1

i N fi
* Courtneys Eq. 12.9 is confusing; he has Nf in the numerator also

51

*Fatigue in Polymers
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Many differences from metals

Cyclic stress-strain behavior often exhibits
softening; also affected by visco-elastic
effects; crazing in the tensile portion produces
asymmetries, figs. 12.34, 12.25.
S-N curves exhibit three regions, with steeply
decreasing region II, fig. 12.31.
Nearness to Tg results in strong temperature
sensitivity, fig. 12.42

52

Fatigue: summary
Objective
Crack
Initiation
S-N
curves
Cyclic
stress-strn
Crack
Propagate

Microstr.
effects
Design

Critical to practical use of structural materials.

Fatigue affects most structural components,
even apparently statically loaded ones.
Well characterized empirically.
Connection between dislocation behavior and
fatigue life offers exciting research
opportunities, i.e. physically based models
are lacking!