A Survey of

Materials Science
IV. Mechanical Properties

J. W. Morris, Jr.
Department of Materials Science and Engineering
University of California, Berkeley
Fall, 2008
Materials Science Fall, 2008

PART IV: MECHANICAL PROPERTIES 655

CHAPTER 19: MECHANICAL BEHAVIOR 658

19.1 Introduction 658

19.2 Uniaxial Tension 659

19.2.1 The load-displacement curve 659
19.2.2 The engineering stress-strain curve 660
19.2.3 True stress and true strain 665
19.2.4 Varieties of tensile behavior 668

19.3 Creep 669

19.3.1 Creep deformation 669
19.3.2 The steady-state creep rate 670
19.3.3 Superplasticity 671
19.3.4 Stress rupture 671
19.3.5 Stress relaxation 671

19.4 Fracture 672

19.4.1 Fracture toughness 672
19.4.2 Plane strain fracture toughness 673
19.4.3 Sources of critical cracks 673

19.5 Fatigue 674

19.5.1 The s-n curve 674
19.5.2 The fatigue crack growth rate 675

19.5 Stress Corrosion Cracking 676

19.6 Hardness 678

CHAPTER 20: ELASTIC PROPERTIES 681

20.1 Introduction 681

20.2 Elastic Properties of Isotropic Materials 682

20.2.1 The basic load geometries 682
20.2.2 Uniaxial tension; Young's modulus and Poisson's ratio 683
20.2.3 Hydrostatic compression; the bulk modulus 685
20.2.4 Shear stress and the shear modulus 686

20.3 Physical source of the elastic moduli 689

20.3.1 The bulk modulus 689
20.3.2 The shear modulus 691

Page 653
Materials Science Fall, 2008

20.4 Specific Modulus; Stiffness 692

20.5 Engineering the Elastic Modulus 694

20.5.1 Composition 694
20.5.2 Microstructure 694
20.5.3 Composite materials 694

CHAPTER 21: PLASTIC DEFORMATION 698

21.1 Introduction 698

21.1.1 Engineering significance of plastic deformation 698
21.1.2 Mechanisms of plastic deformation 699

21.2 The Yield Strength 700

21.2.1 Deformation by dislocation motion 700
21.2.2 The critical resolved shear stress 702
21.2.3 The tensile yield strength of a single crystal 704
21.2.4 The yield strength of a polycrystal 705

21.3 Microstructural Control of the Yield Strength 707

21.3.1 Structural hardening; the Peierls-Nabarro stress 707
21.3.2 Grain refinement 709
21.3.3 Obstacle hardening 710
21.3.4 Solution hardening 712
21.3.5 Dislocation hardening 714
21.3.6 Precipitation hardening 717

21.4 The Influence of Temperature and Strain Rate 719

21.4.1 The variation of yield strength with temperature 719
21.4.2 The influence of strain rate on strength 721

21.5 Work Hardening 721

21.6 Plastic Instability, Necking and Failure 724

21.6.1 The Considere criterion 724
21.6.2 Tensile elongation 726

CHAPTER 22: FRACTURE 729

22.1 Introduction 729

22.1.1 Engineering importance 729
22.1.2 Why things fall apart 730

22.2 Fracture Mechanics 731

22.2.1 Crack propagation from a pre-existing flaw 731
22.2.2 The fracture toughness 733
22.2.3 Fracture-sensitive design 736

Page 654
Materials Science Fall, 2008

22.3 Microstructural Control of Fracture Toughness 738

22.3.1 The fracture mode 739
22.3.2 Choice of fracture mode; the ductile-brittle transition 741
22.3.3 Suppressing brittle fracture 743
22.3.4 Raising the toughness in the ductile mode 745
22.3.5 Raising the fracture stress in the brittle mode 746

22.4 Fatigue 747

22.4.1 Fatigue crack growth 748
22.4.2 Nucleation-limited fatigue 752
22.4.3 Fatigue-safe design 753

Part IV: Mechanical Properties

It was Eudoxas and Archytas who were the originators of the now
celebrated and highly prized art of mechanics. They used it with great
ingenuity to illustrate geometrical theorems, and to support by means of
mechanical demonstrations easily grasped by the senses propositions which
are too intricate for proof by word or diagram ... Plato was indignant at
these developments, and attacked both men for having corrupted and
destroyed the ideal purity of geometry. He complained that they had caused
her to forsake the realm of disembodied and abstract thought for that of
material objects, and to employ instruments which required much base and
manual labor ...
- Plutarch, The Life of Marcellus

The final class of material properties we shall consider are the mechanical properties
that govern the behavior of materials that are subject to mechanical loads. Almost all
engineering systems experience mechanical loads. The materials used in the individual
parts of the system must support those loads without deforming so that they lose their fit or
breaking so that they cease to function at all. Moreover, the manufacturing operations that
make engineering parts are most frequently mechanical processes such as machining,
forging, or mechanical forming. The mechanical properties of the materials that make these
parts must be compatible with the manufacturing procedures that are to be used.

Several qualitatively different kinds of mechanical behavior are pertinent to

engineering design, including the following types.

(1) Elasticity. A material that experiences a relatively small load deforms elastically,
that is, it undergoes a shape change that is, ordinarily, proportional to the load, and
disappears when the load is removed. Elastic deformation must be taken into account in
engineering design since it changes the shapes of parts and affects their fit to one another.
Elastic deformation may even cause a mechanical instability in which the part buckles or

Page 655
Materials Science Fall, 2008

breaks. Examples of elastic instability include buckling of the columns that support a
building or denting the fender of an automobile.

(2) Yielding and plasticity. Under higher loads the material deforms plastically, that
is, it undergoes a shape change that is retained when the load is removed. At normal tem-
peratures, plastic deformation begins when the applied load exceeds the yield strength of
the material. If the load on a material rises beyond its yield strength, its plastic deformation
increases monotonically with the load until it finally breaks at the ultimate tensile strength.
Plastic deformation cannot usually be tolerated in an operating device, since it alters the
geometry of the parts. Hence most devices must be designed so that the mechanical loads
they support do not reach the yield strength. On the other hand, plastic deformation does
provide a margin of safety that acts against catastrophic failure at loads near the yield
strength. Consequently, the materials used in most structures are required to have a certain
degree of plastic deformation before failure. On the other hand, plastic deformation is the
usual method by which parts are formed into shape. The materials that are used in these
parts must be capable of significant plastic deformation without failure.

(3) Creep. A material that is subjected to a static load deforms at a finite rate even
when the magnitude of the load is below the yield strength. This phenomenon is known as
creep. The creep rate is sensitive to the temperature as well as the load. It is very small
when the temperature is below about half the melting point. Creep deformation is a
potential problem when a material is used at high temperature, for example, in the hot
section of a gas turbine. Creep can alter the geometry of the part, and will eventually lead to
failure by a fracture mechanism that is known as creep rupture.

(4) Premature fracture. When a part fractures prematurely, at loads well below its
ultimate tensile strength, this is usually because it contains a flaw that concentrates the load
so that it exceeds the local strength of the material. The flaw generates a crack that
propagates to failure. The fracture toughness of a material is a measure of its ability to
tolerate internal flaws. A high toughness provides a margin of safety against premature
failure.

(5) Fatigue. A well-made part is ordinarily safe from premature fracture because it
contains no significant flaws. However, if the part is subject to a cyclic load, it may spon-
taneously develop a crack through a mechanism that is known as fatigue, even if the maxi-
mum load in the cycle is below the yield strength. Cyclic loads cause cracks to grow at an
accelerating rate until they reach critical size and propagate to failure. Fatigue failures are a
disturbingly common source of the premature failure of engineering devices. Fatigue must
be taken into account in the design of any structurally critical part that is subject to a cyclic
load.

(6) Stress corrosion. A second mechanism that can introduce flaws that lead to
fracture is stress corrosion cracking. When a part is loaded in a corrosive environment, it
may corrode in a pattern that forms sharp-tipped cracks that propagate into the material.
When these grow to critical size, the part breaks. Susceptibility to stress corrosion cracking

Page 656
Materials Science Fall, 2008

is an important consideration in the design of any part that is to be used in a corrosive

environment.

(7) Friction and wear. When materials that are in contact move with respect to one
another they generate frictional forces that oppose their relative motion. The magnitude of
the friction depends on the normal load that forces the parts together, and also depends on
the nature of the materials, including whatever lubricant may be used at the interface.
Where there is friction, there is inevitable wear; material is mechanically removed from one
or both surfaces. Friction is a problem because it consumes power, and also because it
generates heat that may soften or weaken the parts. Wear distorts the geometry of the parts,
and eventually leads to failure. Friction and wear are particular problems in the operation of
gears and bearings, and in the forming of parts with mechanical dies.

We shall discuss each of these mechanical phenomena below. However, to put

them into context, it is useful to consider a simple mechanical test, the uniaxial tensile test,
in which most of these phenomena are revealed.

Page 657
Materials Science Fall, 2008

Chapter 19: Mechanical Behavior

The Burgundian-French rivalry promptly precipitated an arms race

of the sort so familiar to us today. Metalworkers on both sides aimed at a
single goal: to make guns mobile without sacrificing their battering power.
Between 1465 and 1477, designers solved their problem brilliantly by
resorting to smaller, denser projectiles. They discovered that a
comparatively small iron cannonball could strike a more damaging blow
than a stone projectile (which fractured on impact) ever could, no matter
how large. This meant that guns could be made smaller, but had to be
stronger, too ...
As it happened the Burgundians, though they had taken the lead in
mobile siege artillery, were not destined to profit from it. Instead, Charles
the Bold, Duke of Burgundy, being too impatient to await the arrival of his
guns on the battlefield at Nancy, led a cavalry charge against a massed
formation of Swiss pikemen ... and met his death on the point of a pike.
Burgundian lands were then swiftly partitioned.
... If a single ruler had continued to control that region and been able
to monopolize the new guns for as along as a generation, he might have
been able to use his mobile wall destroyers to achieve hegemony wherever
the new siege guns could reach. Instead, the kings of France and the
Hapsburgs of Germany divided up the Burgundian guns and gunmakers ...
assuring a standoff, which became permanent, thanks to the swift
improvement in fortifications after 1500.
... For Japan as well as the other empires of Asia, the effect of
gunpowder weapons was to consolidate central power over comparatively
vast territories. Once a ruler had monopolized big guns or, in Japan,
disarmed the whole nation, there was no incentive to continue experiments
with improvements [while] among the rival states of Europe such
experimentation continued, and at a very rapid pace ... The amazing imperial
success of European nations in the eighteenth and nineteenth centuries
resulted - and all because in 1477 the Burgundian lands were divided be-
tween French and German rulers ...
Truly the course of history, even on a world scale sometimes turns
on single events, and on a skilled (or lucky) response in a moment of crisis.
- William McNeill, The Age of Gunpowder Empires, 1450-1800

19.1 INTRODUCTION

Let a material be made into a round bar and put in a device that grips its two ends
and pulls on them, as illustrated in Fig. 19.1. This is the normal geometry of a tensile test ,
which is the test used to generate much of the mechanical property data that is used in engi-
neering design.

The tensile testing machine can control either the load on the specimen, in which
case the test is said to be stress-controlled, or the specimen length, in which case the test is

Page 658
Materials Science Fall, 2008

strain-controlled. Most commonly, the tensile test is done in strain control; the specimen is
gradually elongated and the load recorded until the specimen fails. Other mechanical prop-
erties are measured by modifications of the tensile test. For example, creep properties are
tested by fixing the load and measuring the elongation as a function of time, or by fixing the
length and recording the load decay with time (a creep phenomenon known as stress
relaxation). The fracture toughness is measured by replacing the round bar with a plate that
is cracked on one side, and measuring the load necessary to propagate the crack to failure.
Fatigue resistance is measured by setting the machine to cycle the load in a programmed
pattern, and measuring the number of cycles required to fail the specimen, or, in a more
detailed test, measuring the rate of growth of a pre-existing crack. Stress corrosion
resistance is studied by immersing the specimen in a corrosive medium, usually under fixed
load, and measuring the rate of growth of a pre-existing crack.

P
...
Fig. 19.1: A cylindrical sample of length, L, and cross-sectional area, A,
in uniaxial tension.

A final variation measures the hardness of a material. In a hardness test one of the
specimen grips is replaced by a flat support, and the other by a small indenter that is im-
pressed into the material by loading it in compression. The size or depth of the indentation
made by a standard indenter under a standard load measures the hardness, which is the
property that controls the wear resistance of the material in most situations.

19.2 UNIAXIAL TENSION

19.2.1 The load-displacement curve

Let a round bar specimen be placed in a tensile machine and pulled under strain
control, so that the elongation is gradually increased and the load recorded. The relation
between the applied tensile load (P) and the change in length of the bar (ÎL) defines the
load-displacement curve for the sample. (In an actual test the specimen must be gripped at
its ends. To avoid end effects the specimen is shaped at its ends and the length is measured

Page 659
Materials Science Fall, 2008

between reference marks on the specimen well away from the grips, as shown in Fig.
19.2.)

P
P
L

ÎL
...
Fig. 19.2: A typical form for the load-deflection curve of a ductile metal
tested in tension.

While the form of the load-displacement curve is characteristic of the material, the
quantitative relation between the load and the displacement depends on the geometry of the
specimen as well. As you can easily verify by stretching rubber bands or bending paper
clips, the load required to achieve a given displacement increases with the cross-sectional
area of the bar (a thin rubber band is more easily stretched than a thick one) while the dis-
placement under a given load increases with the length of the specimen (a long rubber band
is more easily stretched by a given length than a short one). These phenomena reflect the
fact that the total elongation of the specimen is the sum of the elongations of each cross-
sectional element along its length, while the total load is the sum of the loads carried by
each element across a section through it.

19.2.2 The engineering stress-strain curve

Engineering stress and strain

To account for the specimen geometry in a simple way, we measure the load by the
engineering stress, s,

P
s=A 19.1
0

where P is the load, A0 is the cross-sectional area of the specimen, and the stress is force
per unit area. We measure the displacement by the engineering strain, e,

ÎL
e= L 19.2
0

Page 660
Materials Science Fall, 2008

where ÎL is the elongation, L0 is the specimen length, and the strain is dimensionless. The
variation of s with e is called the engineering stress-strain curve. It is simply the load-dis-
placement curve corrected for the dimensions of the specimen.

The engineering stress and strain are only approximate measures of the local loads
and displacements within the material. As the specimen elongates, its cross-sectional area
decreases, so the average stress across its section is always higher than the engineering
stress suggests. Moreover, the local stress (load per unit area) varies through the cross-
section; the material in the center of the specimen supports a higher load per unit area than
that near the periphery. Nonetheless, the engineering stress-strain curve is a sufficiently
good representation of the mechanical properties of the material that it is the one tabulated in
most compilations of mechanical properties, and provides the mechanical property data that
is used in all but the most sophisticated engineering designs.

su

sy

I II III

e
...
Fig. 19.3: The engineering stress-strain curve of a typical ductile metal.

The engineering stress-strain curve of a typical polygranular metal is illustrated in

Fig. 19.3. The curve can be divided into three regions that represent qualitatively different
kinds of mechanical behavior. The three regions are denoted by roman numerals in the fig-
ure.

Elastic deformation

Region I defines the range of elastic deformation. Elastic deformation ordinarily

has two characteristics.

(1) Elastic deformation is recoverable. If the stress is removed from the specimen,
the strain vanishes; the specimen returns to its original length.

(2) In most materials, the elastic stress is proportional to the strain,

s = Ee 19.3

Page 661
Materials Science Fall, 2008

where the constant of proportionality, E, is called Young's modulus. Equation 19.3 is

known as Hooke's Law, after its discoverer. Materials that obey Hooke's Law are called
linear elastic materials. Most structural materials are members of this class, including al-
most all metals and ceramics.

Stable plastic deformation

Region II defines the range of stable plastic deformation. Stable plastic deformation
ordinarily has three characteristics.

ep ee
e
...
Fig. 19.4: Relaxation of a specimen strained slightly into the plastic re-
gion. ep is the plastic strain, ee is the elastic strain.

(1) A portion of the strain is plastic. If the stress is decreased after the specimen is
loaded into the plastic range, the specimen relaxes along a line that is nearly parallel to the
linear elastic curve, as illustrated in Fig. 19.4. The specimen does not return to its original
length after the stress is removed. The remnant strain is called the plastic strain, and mea-
sures the permanent deformation of the specimen. The total strain of a specimen that is
loaded into region II is the sum of its elastic strain (ee) and its plastic strain (ep).

(2) The stress-strain curve is non-linear. The slope of the stress-strain curve,
ds/de, drops below the modulus, E, as the sample begins to deform plastically, and ordi-
narily decreases with the plastic strain as shown in Fig. 19.3. However, the slope remains
positive in region II. The engineering stress must be raised to increase plastic deformation.
This behavior is known as work hardening.

(3) The plastic deformation is uniformly distributed along the length of the speci-
men. This behavior is known as uniform elongation. The specimen increases in length, but
retains its shape, to within a uniform contraction.

Page 662
Materials Science Fall, 2008

The yield strength

The stress at which plastic deformation begins is called the yield strength of the
material, and denoted sy. The yield strength is the boundary between the elastic (I) and
plastic (II) ranges of the stress-strain curve. In some materials, the yield strength is well-
defined by a yield point at which there is an obvious discontinuity in the stress-strain curve.
An example is illustrated in Fig. 19.5a. However, most materials do not show a well-de-
fined yield point and there is some uncertainty in the location of the yield strength. To re-
move this uncertainty, it is conventional to define the yield strength as the stress that corre-
sponds to a plastic strain of 0.2%. To measure the yield strength according to this criterion,
a line is drawn parallel to the elastic portion of the stress-strain curve so that it intersects the
strain axis at 0.2%, as illustrated in Fig. 19.5b. The stress at which this line crosses the
stress-strain curve defines the 0.2% offset yield, which is the value of the yield strength that
is tabulated in most compilations of mechanical properties.

sy

s s

0.2%
e e
(a) (b)
...
Fig. 19.5: (a) The stress-strain curve of a material with a well-defined
yield point. (b) Illustration of the 0.2% offset criterion for the
yield strength (the elastic region has been exaggerated for clar-
ity).

Note that the conventional definition of the yield strength is such that there is some
measurable plastic deformation at stresses below yield. Engineers who design parts with
very tight tolerances must, hence, make sure that operating stresses are kept well below the
tabulated yield strength. If the part is a critical one, they should re-measure the yield
strength against the criterion that is pertinent to the particular application. (On a number of
occasions I have been contacted by a puzzled engineer who designed some metal part to op-
erate at stresses below the tabulated yield strength, and was shocked to discover that it
nonetheless experienced a measurable plastic deformation. He usually wants to know why
the published data is wrong, not why he is.)

Page 663
Materials Science Fall, 2008

Unstable plastic deformation

The engineering stress-strain curve of a ductile material usually passes through a

maximum, and enters the region marked III in Fig. 19.3. In region III the stress decreases
monotonically with the strain until the specimen finally fractures.

The behavior in region III is due to plastic instability. Rather than being uniformly
distributed along the specimen, the plastic deformation concentrates at a particular point,
which thins into a neck of the type illustrated in Fig. 19.6. As the sample necks down, its
cross-sectional area decreases. The actual stress on the material within the neck is P/A,
where A is its true cross-sectional area. As A decreases, the stress in the necked region in-
creases, even though the engineering stress, P/A0, decreases. The increase in stress causes
the sample to neck further, which causes the local stress to rise until the sample finally
breaks.

P
P

L L

P
P
...
Fig. 19.6: Necking of a tensile specimen due to unstable plastic deforma-
tion.

The ultimate tensile strength

The maximum value of the engineering stress is called the ultimate tensile strength,
and denoted su. When the test is done in strain control, su defines the boundary between
stable (region II) and unstable (region III) plastic deformation. It is the value of the engi-
neering stress at which plastic deformation first becomes unstable and forms a neck. If the
test is done in stress control the ultimate tensile strength defines the highest load a sample
with a given initial cross-section can support without breaking. The sample fractures when
the ultimate tensile strength is reached. Hence su is a critical parameter in engineering de-
sign.

The uniform and total elongation

In designing an engineering part it is often important to know how much the part
can be elongated before it fractures in tension. If the material is a ductile one with a stress-
strain curve like that shown in Fig. 19.3, the answer depends on how the test is controlled.
If the load is controlled, the sample fractures when the ultimate tensile strength is reached.
The plastic strain that corresponds to the ultimate tensile strength is called the uniform elon-

Page 664
Materials Science Fall, 2008

gation, eu, since the plastic strain prior to the ultimate tensile strength is uniformly dis-
tributed along the specimen.

If the strain is controlled, the sample necks before fracture, and has an additional
post-necking elongation. The total strain up to fracture at the neck is called the total elon-
gation, eT. Since the post-necking elongation is non-uniform and concentrated in the region
of the neck, the total elongation depends on the length of the specimen (Fig. 19.6). The
value of the total elongation that is usually tabulated is obtained by measuring the engi-
neering strain in a 2 in. section of the specimen (2 in. gage length) that includes the neck. If
some other gage length is used, it must be specified if the value of total elongation is to be
meaningful.

Reduction in area

The final tensile property that is often measured and tabulated is the reduction in
area prior to fracture at the thinnest part of the neck:

A0 - A
RA = A0 19.4

As we shall see below, the reduction in area is a measure of the true strain at fracture, and
is, hence, a measure of the useful ductility of the material. It is the tensile property that cor-
relates most closely to the fracture toughness of the material.

s
eT

eu

e
...
Fig. 19.7: The uniform and total elongation in a tensile test.

19.2.3 True stress and true strain

The engineering stress and strain generate useful numbers for mechanical design.
However, except in the limit of small strain, they are not very accurate measures of the
stress and strain actually borne by the material. The reason is that the sample thins as it

Page 665
Materials Science Fall, 2008

elongates and, being an inanimate object, has no reliable memory of what its original cross-
sectional area was.

Definition of the true stress and true strain

The actual value of the average stress on a cross-section of area, A, is called the true
stress, ß, and is

P
ß=A 19.5

Since

A 0 
ß = s  19.6
A

the true stress exceeds the engineering stress by a factor that increases as the sample be-
come longer and thinner.

To find a measure of the true strain we use the fact that an infinitesimal increase in
the stress on a uniform sample of length, L, produces an infinitesimal strain that is given by

dL
d‰ = L 19.7

If the stress is increased so that the length increases from L0 to Lf, the true strain is

Lf f L
dL  Lf 
‰= ⌠
⌡ d‰ = ⌠
⌡ = ln   19.8
L L0
L0 L0

When the strain is uniformly distributed along the length of the sample, as it is for
the elastic and stable plastic regions of the stress-strain curve, eq. 19.8 gives the true strain
that is associated with an increase in the overall length from L0 to Lf. However, when the
strain is non-uniform, as it is when the sample necks, eq. 19.8 can only be used by dividing
the samples into length increments that are so small that the strain can be assumed uniform
within each of them, and applying eq. 19.8 separately to each. One can do this, in practice,
by inscribing a fine grid on the surface of the specimen and separately calculating the strain
for each element of the grid by measuring the displacement of the grid lines.

However, there is an easier way to determine the local value of the true strain in a
tensile specimen. It is based on the fact that plastic strain conserves volume. As we shall
see later, the elastic strain does not ordinarily conserve volume. But if the total strain is
primarily plastic, as it ordinarily is by the time the sample begins to neck, then an element of

Page 666
Materials Science Fall, 2008

the sample that has length, L, and cross-sectional area, A, has a constant value of the
product, V = AL. Then

dL dA
d‰ = L = - A 19.9

and

A 0 
‰ = ln A  19.10
f

where Af is the final value of the cross-sectional area of a specimen whose original cross-
section was A0. Eq. 19.10 can be used to find the non-uniform strain in a round bar tensile
specimen by simply measuring the cross-sectional area along its length.

Relation to the engineering stress and strain

From the definition of the engineering strain (eq. 19.2), the relation between the
engineering strain and the true strain is

Lf - L0 A0 - Af
e= L0 = A0 = e‰ - 1 19.11

‰ = ln(1+e) 19.12

Using eq. 19.6, the relation between the engineering stress and the true stress is

 Af 
s = ßA  = ße-‰ 19.20
0

ß = s(1+e) 19.22

In the limit of small strain (‰ << 1), ‰ « e and ß « s. Most structural materials yield
at small values of the elastic strain. Hence we can usually ignore the difference between the
engineering strain and the true strain when the plastic strain is small.

The true stress-strain curve

The relation between the true stress and true strain in tension is called the true
stress-strain curve. The true stress-strain curve of a typical polygranular metal has a form
like that illustrated in Fig. 19.8. It is identical to the engineering stress-strain curve for
small strains, but differs qualitatively when the strain is large. In particular, there is no
obvious feature of the true stress-strain curve that reveals the value of the ultimate tensile
strength or the onset of plastic instability. This is because there is no change in the mechan-
ical behavior of the material at plastic instability.

Page 667
Materials Science Fall, 2008

As we shall discuss in more detail later, the plastic instability that determines the ul-
timate tensile strength happens because the rate of increase of stress with strain (the work
hardening rate, dß/d‰) becomes insufficient to preserve the geometry of the specimen. For
a cylindrical tensile specimen, this happens when

dß
d‰ = ß 19.23

As the specimen deforms, the stress, ß, rises while the work hardening rate, dß/d‰, drops.
The slope of the engineering stress-strain curve is

ds d(ße-‰ ) e-‰ dß - ße-‰ d‰ dß

-2‰ 


= = = e  d‰ - ß 19.24
de d(e‰ -1) e‰ d‰

which vanishes when eq. 19.23 is satisfied. Eq. 19.23 is called the Considere criterion.

ßf

ßy
ß

‰f
‰
...
Fig. 19.8: The true stress-strain curve of a typical ductile metal. The true
stress and strain at fracture are indicated.

While the true stress-strain curve conceals the ultimate tensile strength, it accurately
records the work hardening behavior of the material, correctly shows that hardening con-
tinues after the deformation has become unstable (in contrast to the engineering stress-strain
curve, which implies that the material softens after instability) and gives the values of the
stress and strain across the section at fracture. The design engineer is usually interested in
the engineering stress-strain curve, since the ultimate tensile strength is an important con-
sideration in design. The materials scientist is usually more interested in the true stress-
strain curve, since it is a more valid representation of the mechanical behavior of the mate-
rial.

19.2.4 Varieties of tensile behavior

Page 668
Materials Science Fall, 2008

The engineering stress-strain curve of a structural material usually has one of four
characteristic shapes. The stress-strain curves of polygranular metals and structural plastics
usually resemble that shown in Fig. 19.2, which we have already discussed in some detail.
Several important metals and alloys, including carbon steel, yield at a well-defined yield
point, and, hence, have stress-strain curves that resemble the one shown in Fig. 19.5a.
Brittle materials, including most structural ceramics, have stress-strain curves like that
shown in Fig. 19.9a; they do not yield (at least not to noticeable degree) but simply extend
elastically until they fracture. Elastomeric materials are capable of large elastic strains, and
have non-linear elastic regions like that shown in Fig. 19.9b. A pure elastomer, such a
vulcanized rubber, does not yield unless it is strained very slowly; it deforms elastically
until fracture.

Note that the "ultimate tensile strength" of a brittle or elastomeric material is not due
to plastic instability, but to simple fracture. As we shall see, the stress at which an elastic
material fractures is determined by the nature of the flaws the material contains rather than
by its inherent mechanical properties. For this reason, the ultimate tensile strength of a
brittle or elastomeric material is not a well-defined material property. It varies significantly
from sample to sample. While the "ultimate tensile strengths" of such materials are often
measured and tabulated, they do not constitute reliable design values. They give only a
rough indication of the level of tensile strength one might expect.

su su

s s

e e
(a) (b)
...
Fig. 19.9: (a) Stress-strain curve of a brittle solid. (b) Stress-strain curve
of an elastomer. In both cases the strain is elastic to fracture at
s u.

19.3 CREEP

19.3.1 Creep deformation

Consider a tensile specimen that is tested under stress control, and suppose that,
instead of elongating it to plasticity and failure, it is loaded to a stress below the yield
strength, and held at that stress for a long time. The specimen gradually increases in length.
This phenomenon is called creep. The creep rate (the strain rate at constant load) is,
ordinarily, almost vanishingly small unless the temperature of the specimen is greater than

Page 669
Materials Science Fall, 2008

about half of its melting point. However, when the temperature is a significant fraction of
the melting point the creep rate becomes appreciable, and leads to significant plastic
deformation in a relatively short time.

The normal behavior of a specimen that is tested in creep is illustrated in Fig. 19.10,
where we have plotted the strain as a function of time at constant stress. Creep occurs in
three stages. Stage I is primary creep. During primary creep the strain rate decreases
monotonically with time. Stage II is secondary or steady-state creep. This stage, which
may last for some time, is characterized by a constant creep rate. Eventually, the sample
enters tertiary creep. The creep increases monotonically, often because of progressive void
formation within the specimen, until the sample finally ruptures.

I II III

t
Fig. 19.10: The variation of strain with time in normal creep, showing I:
primary creep, II: steady-state creep, III: tertiary creep to rup-
ture.

19.3.2 The steady-state creep rate

The property that is most commonly used to characterize creep for design purposes
is the steady-state strain rate, the strain rate in the steady-state creep. The steady-state strain
rate governs the creep rate at the moderate strains that usually have the greatest engineering
significance. The steady state strain rate, •‰, increases with the applied stress and the tem-
perature. Most high-temperature creep data can be fit to relations of the form

- Q 
•‰ = Aßnexp kT  19.25

The stress exponent, n, has a value that indicates the dominant mechanism of creep. For
metals at moderate to high stress, the dominant mechanism is the climb-assisted motion of
dislocations through the bulk, and n falls in the range 4-7, depending on the material. At
very low stresses both metals and ceramics creep by diffusional processes with n = 1. The
activation energy, Q, is ordinarily the activation energy for the dominant diffusion step in
creep.

Page 670
Materials Science Fall, 2008

19.3.3 Superplasticity

Some materials adopt a creep mechanism at intermediate stresses that produces very
high tensile elongation prior to failure. Such materials are said to be superplastic. The total
strain of a superplastic material may exceed 1000% before failure, in contrast to the total
elongations of 30-40% that apply to exceptionally ductile materials in a normal tensile test.
Superplastic creep is characterized by a steady state strain rate that obeys eq. 19.25 with a
stress exponent, n « 2, and an activation energy, Q, near that for grain boundary diffusion.
Superplastic creep occurs in metals that are fine-grained, and maintain small grain size
during deformation. Their deformation is dominated by grain boundary sliding. Individual
grains are displaced with respect to one another by sliding along the boundary between
them. Superplastic creep is only observed at intermediate stresses; at both high and low
stresses it is superseded by bulk creep mechanisms with higher stress exponents.

19.3.4 Stress rupture

High temperature creep eventually terminates with the breakage of the specimen,
usually through the coalescence of voids that nucleate and grow in the material as it de-
forms. The time to creep failure is an important design parameter for devices that are in-
tended to operate at high temperature under load, such as turbine engines. The failure time
in creep depends on the stress and the temperature as well as on the material. The property
that is most often tabulated as a measure of creep failure is called the rupture strength. The
rupture strength is the stress that causes failure in some pre-selected time (often 1000 hr.
under load). The rupture strength is a sharply decreasing function of the temperature.

19.3.5 Stress relaxation

ß
decreasing
temperature

t
...
Fig. 19.11: Stress relaxation of a sample loaded and held at fixed strain.

Finally, suppose that a sample is given a fixed displacement that is maintained for a
long time. A common example is a bolt, which is tightened (strained elastically) and then
left alone. Creep occurs under the residual load in the sample. Since the strain is fixed, the
stress decreases with time, as shown in Fig. 19.11. This phenomenon is known as stress

Page 671
Materials Science Fall, 2008

relaxation. The stress relaxation rate decreases as the temperature decreases, as shown in
the figure.

Stress relaxation is important in engineering since it cause the disappearance of the

stresses that are used to bind a system together. Exposure to a sufficiently high temperature
for a sufficiently long time causes tightened bolts to loosen and shrunk-fit parts to come
free.

19.4 FRACTURE

19.4.1 Fracture toughness

If a specimen that contains a part-through crack is tested in tension, it usually breaks

at a load that is significantly below its yield strength. The crack becomes unstable at a
critical value of the stress and spontaneously propagates across the sample. The value of
the stress at which the crack propagates to failure depends on the size of the crack and on a
material property, called the fracture toughness, that measures its resistance to fracture.

The fracture toughness of a material is most conveniently studied by using a speci-

men that is a thick plate with a sharp-tipped crack of known length cut into one side, as il-
lustrated in Fig. 19.12. A tensile load is applied perpendicular to the crack, and increased
until the crack propagates. The critical stress, ßc, at which the sample breaks is inversely
proportional the square root of the crack length, a . The fracture toughness of the plate,
Kc, is defined as the product,

Kc = Qßc a 19.26

where Q is a geometric factor that depends on the ratio, a/W, where W is the width of the
specimen in the direction of crack propagation (Fig. 19.12). When W >> a, Q = 1.12 π
= 1.99.

ß
...
Fig. 19.12: A cracked plate loaded in tension used to measure fracture
toughness.

Page 672
Materials Science Fall, 2008

19.4.2 Plane strain fracture toughness

The fracture toughness, Kc, decreases as the plate thickness (T) increases. This is
because the specimen relaxes at its free surfaces, which changes the local distribution of the
stress at the tip of the crack. However, when the sample is very thick, the relaxation at its
lateral surfaces can be ignored. In this limit the sample is said to be in plane strain. When
the sample thickness is sufficient to achieve plane strain, the fracture toughness asymptotes
at the value, KIc, the plane strain fracture toughness. and becomes independent of thickness.

Kc

KIc
plane
strain

T
...
Fig. 19.13: The normal variation of fracture toughness (Kc) with the
thickness (T) of the test specimen. Kc asymptotes to the plane
strain fracture toughness, KIc, when the specimen is thick
enough to establish plane strain at the crack tip.

The plane strain fracture toughness, KIc, is the property that is ordinarily used to
characterize resistance to fracture. It is the preferred measure both because it is a material
property and because it provides a conservative design criterion. The critical fracture stress,
ßc, of a part that contains a crack of length, a, can be estimated from eq. 19.26 by setting Kc
= KIc. Since Kc ≥ KIc, the actual value of the critical stress is at least as high, and is
ordinarily higher, than the estimated value.

19.4.3 Sources of critical cracks

The critical crack length, ac, at which a crack propagates to fracture under a given
value of the applied stress, ß, is, from equation 19.26,

Kc2
ac = Q-2 ß  19.27

To ensure the reliability of an engineering structure one must guarantee that no critical crack
is present, and none will form during service. The cracks that cause the failure of engineer-

Page 673
Materials Science Fall, 2008

ing systems have three principal sources: manufacturing defects, fatigue crack growth and
stress corrosion cracking. One guards against manufacturing defects by controlling the
manufacturing process and inspecting the finished parts. Fatigue and stress corrosion
cracking require independent analysis, since they provide mechanisms that allow innocuous
flaws to grow to critical size.

19.5 FATIGUE

Suppose that a nominally flaw-free sample of a ductile material is placed in a tensile

machine and subjected to a load that oscillates with the time, as illustrated in Fig. 19.14. Let
the maximum value of the stress, ßmax, be below the yield strength of the material, but
above about 1/3 of the ultimate tensile strength. For many cycles of the load, nothing ob-
vious happens. The sample stretches slightly as the stress is increased to ßmax, relaxes to
its original length as the stress returns to zero, and compresses slightly as the load de-
creases to - ßmax. Then, often with no obvious warning, the sample breaks. This phe-
nomenon is called fatigue, an anthropomorphic image that suggests that the material grows
tired and weak.

A closer analysis reveals that cyclic stress provides a mechanism to nucleate cracks
and grow them until they reach critical size. Crack nucleation and growth occurs because
plastic deformation is never quite reversible. It eventually concentrates at local sites within
the specimen, and forms cracks. The irreversibility of plastic deformation at the tip of a
crack causes it to grow at an accelerating rate until it finally reaches critical size and propa-
gates to failure.

19.5.1 The s-n curve

The fatigue properties of a material are usually represented in one of two ways.
First, nominally flaw-free specimens are tested under reversed, uniaxial loading to produce
the s-n curve of the material. The s-n curve is a plot of the number of cycles to failure (N)
as a function of the maximum engineering stress, s, in a cycle between + s and - s. The
number of cycles increases dramatically as the stress cycle is decreased. Some materials,
including many common structural steels, have a fatigue limit, a value of the stress, sl, be-
low which fatigue cracks do not form. The fatigue limit, sl, is often about one-third of the
ultimate tensile strength, su. Most structural materials have no fatigue limit. However, the
stress that is associated with a life of n = 108 cycles is often used as a practical fatigue limit
since 108 cycles corresponds to a long operating time for most systems and since n in-
creases rapidly as s is lowered below this value.

When the material is likely to be stressed or strained into the plastic region, it is
common to represent its fatigue behavior in terms of a Coffin-Manson plot of n against the
cyclic plastic strain, Ήp. This plot can be easily generated by testing under strain control.

Page 674
Materials Science Fall, 2008

sm
su
s

-sm
s
t
sl

log(n)
...
Fig. 19.14: The s-n curve of a typical metal. The stress oscillates with
time as shown at right. The fatigue limit, sl, is either the
asymptote of the curve or the stress that causes failure in 108
cycles.

19.5.2 The fatigue crack growth rate

Most of the life of a flaw-free sample that is fatigued under cyclic load is spent in
the nucleation phase of fatigue crack growth. As a consequence, the s-n curve significantly
overestimates the life of a specimen that contains pre-existing cracks. This is a serious
shortcoming, since it is difficult to prevent small cracks or flaws in real parts that are cast,
formed, or machined into shape. For this reason modern engineering design often assumes
the presence of a small pre-existing crack, and attempts to ensure that it will not grow to
critical size during the life of the part. The data that is needed to support this design
methodology concerns the fatigue crack growth rate.

ln(da/dn)

ÎK th

ln(ÎK)
...
Fig. 19.15: The dependence of the fatigue crack growth rate (da/dn) on the
cyclic stress intensity (ÎK).

Page 675
Materials Science Fall, 2008

The stress cycle that drives the growth of a fatigue crack is the stress cycle at the
crack tip. As we shall discuss further below, this stress cycle is proportional to the cyclic
stress intensity, ÎK, where

ÎK = QÎß a 19.28

where Q is the geometric factor introduced in eq. 19.26, Îß is the nominal stress cycle in
the material well away from the crack, and a is the crack length. The crack growth per cy-
cle, da/dn, varies with the cyclic stress intensity, ÎK, as illustrated in Fig. 19.15. The be-
havior can be conveniently divided into three regions.

The first is the threshold region. There is a reasonably well-define threshold stress
intensity, ÎKth, below which a normal fatigue crack does not grow. One objective of fa-
tigue-safe design is to ensure that the largest crack is so small, or the cyclic stress is so low,
that ÎKth is never exceeded. When the cyclic stress intensity is near ÎKth, da/dn increases
rapidly with ÎK.

The second region is the power-law region, or Paris region (after the person who
first discovered it), in which

da
dn = A(ÎK)m 19.29

where A and m are constants that are characteristic of the material. Power-law behavior
pertains for most of the period of fatigue crack growth to failure, and, hence, this growth
law is often employed to estimate the service lives of parts that must be assumed to contain
pre-existing flaws.

The third region pertains when the crack length is very close to the critical crack
length, ac, for unstable fracture. The rate of crack growth increases rapidly as a approaches
ac.

We shall discuss later how this sort of information is used to estimate the life of an
engineering part or to set inspection intervals that guard against failure.

19.5 STRESS CORROSION CRACKING

Let a sample be loaded in tension while immersed in a electrolyte under conditions

where general corrosion is relatively slow. The sample may sit quite happily in this envi-
ronment for some time, and then suddenly fail by catastrophic fracture. This phenomenon
is known as stress corrosion cracking, and is due to the corrosion-assisted growth of
cracks to critical size. Stress corrosion cracking is a troublesome engineering problem
since it often leads to failure without warning under conditions in which general corrosion
is mild, and perhaps imperceptible.

Page 676
Materials Science Fall, 2008

Stress corrosion cracking is a crack growth mechanism that causes small flaws to
propagate until they reach critical size. The flaws that lead to failure may be either pre-ex-
isting manufacturing flaws or corrosion-induced flaws at the base of corrosion pits or
crevices. The enhanced stress at the tip of the flaw leads to the growth of a sharp-tipped
crack that propagates into the material. Depending on the material and the environmental
conditions, the crack may grow by anodic dissolution at a local anode, or by a hydrogen-
assisted crack growth process at a local cathode (in this case the phenomenon is often called
hydrogen embrittlement). In either case, stress corrosion cracking usually occurs under
conditions where general corrosion is mild; otherwise general corrosion would blunt or re-
move cracks that were growing in from the surface.

As in the case of fatigue, two types of test data are used to characterize stress cor-
rosion cracking. The simplest representation is a plot of the time-to-failure as a function of
applied stress for a nominally flaw-free specimen in a given corrosive environment. If the
environment is such that stress corrosion cracking occurs, these test yield curves of the
form shown in Fig. 19.16. The failure time increases dramatically as the load is decreased.
Failure is not observed at stresses below a threshold stress, which is almost always greater
than 0.5 sy, and is usually closer to 0.8 sy.

su

sc

t
...
Fig. 19.16: Typical variation of time-to-failure in stress corrosion cracking
with the applied stress. The threshold stress, sc, depends on
the precise environment.

Time-to-failure plots are not applicable to samples that contain pre-existing flaws.
Since these may be present, and, therefore, must be assumed to be present in most
manufactured parts, the more sophisticated analyses that are now used in engineering
design depend on direct measurements of the crack growth rate. The crack growth rate,
da/dt, in a given material in a given environment is determined by the stress intensity, K, at
the crack tip. By analogy to equation 19.26,

K = Qß a 19.30

Page 677
Materials Science Fall, 2008

The relation between da/dt and K usually has a form like that shown qualitatively in Fig.
19.17. As before, it is conveniently divided into three regions.

K Iscc

ln(da/dt)

KIc

ln(K)
...
Fig. 19.17: The stress corrosion crack growth rate as a function of the
crack tip stress intensity (K). KIscc is a strong function of the
environment.

Stress-corrosion cracks do not grow at stress intensities less than KIscc, the critical
stress intensity for crack growth. KIscc is not a material property; it depends on both the
material and the environment. Nonetheless, it is a useful design value if it is known, since
stress corrosion will not cause failure if K is held below KIscc, but will inevitably lead to
eventual failure if K > KIscc.

When K is raised above KIscc, the crack growth rate increases rapidly to a steady
state velocity (stage II). The steady-state velocity is maintained until the crack grows to the
extent that K approaches Kc, the fracture toughness of the material. When this happens
(stage III) the crack growth rate increases monotonically until the sample fails at K = Kc.

19.6 HARDNESS

The final simple mechanical property we shall mention here is the hardness. The
hardness of a material measures its resistance to indentation. It is tested with a device like
that shown in Fig. 19.20. An indenter is pressed into the surface of the material under a
pre-selected load. The hardness is associated with the degree of penetration. The hardness
is a qualitative measure of resistance to indentation that is a particularly good measure of
resistance to abrasive wear. It also provides a simple, qualitative measure of resistance to
plastic deformation that is widely used as a quality control tool to verify the strengths of
heat-treated metal parts.

While many different hardness tests have been used over the years, two are com-
monly used today. The first is the Rockwell hardness test. In the Rockwell test the inden-
ter is impressed into the surface to a pre-set load. The load is then relaxed, and the depth of
penetration is measured. The major advantage of this test is that the impression is made and

Page 678
Materials Science Fall, 2008

the depth of penetration measured in sequential operations on the same machine. The
Rockwell hardness test is quick and easily automated.

The result of the Rockwell hardness test is a unitless hardness number that is read
from a standard scale. The hardness number decreases with the depth of penetration, that
is, materials that are more resistant to penetration have higher hardness. Otherwise, the
Rockwell hardness number has no fundamental significance. There are, in fact, several
different variations of the Rockwell hardness test that vary in the load used and the shape of
the indenter. These are used for materials that have different hardness ranges and thick-
nesses. The Rockwell-C scale is the one that is commonly used for steel parts. This test
uses a conical, diamond indenter with a 120º included angle and a spherical cap with an 0.2
mm radius, and a 150 kg. major load. Rockwell A or B tests are used for softer metals.
Rockwell A uses a conical indenter with a smaller load, Rockwell B uses a spherical inden-
ter that makes a more shallow impression. Empirical relations have been proposed to relate
the various Rockwell scales to one another and to the yield or tensile strength of the mate-
rial. These are useful, but are only approximate, particularly when they are applied to ma-
terials that differ significantly from the materials for which they were developed.

...
Fig. 19.20: The hardness test.

The second common hardness test is the Knoop test, which is used to measure the
microhardness of small regions of the material. Microhardness readings are used, for ex-
ample, to determine the case depth of a part whose surface has been hardened by induction
heating or carburizing. The Knoop test uses an indenter that has the shape of a distorted
pyramid and makes an impression that has the elongated diamond geometry shown in Fig.
19.21. The indenter shape is such that the depth of the indentation is about 1/4 of the short
dimension of the diamond, and only about 1/30 of the long dimension. The Knoop hard-
ness is defined as

P P
K=A = 14.2 2 19.31
l

where P is the load, A is the projected area of the impression, l is the length of the impres-
sion, and the factor, 14.32, applies to the standard geometry of the Knoop indenter. In a
normal microhardness test the loading device and specimen are mounted in an optical mi-
croscope that is used to measure the length, l, of the impression. The impression need be
no more than about 10 µm in size, so very local values of the hardness can be measured.

Page 679
Materials Science Fall, 2008

Empirical relations have been developed that provide approximate conversions between the
Knoop and Rockwell scales.

The Knoop and Rockwell-C tests are often used together in the quality control of
metal parts. For example, the material specifications for a hardened gear, shaft or bearing
race often include a specification for the superficial hardness, given as minimum and
maximum acceptable values of RC , and a specification of the minimum case depth, the dis-
tance below the surface for which a specified value of RC must be maintained. The super-
ficial hardness specification can be verified non-destructively by performing hardness tests
on the surface of the part. To verify the case depth, however, it is necessary to section se-
lected parts and survey the hardness profile below the surface. This is ordinarily done with
a microhardness tester that uses the Knoop method. The Knoop values are subsequently
converted to RC to verify the specification that governs case depth.

...
Fig. 19.21: The shape of the impression made by a Knoop indenter.

Page 680
Materials Science Fall, 2008

Chapter 20: Elastic Properties

Robert Hooke ... gave the first rough law of proportionality between
the forces and displacements. Hooke published his law first in the form of
an anagram "ceiiinosssttuu" in 1676, and two years later gave the solution
of the anagram: "ut tensio sic vis", which can be translated freely as "the
extension is proportional to the force."
- I.S. Sokolnikoff, The Mathematical Theory of Elasticity

20.1 INTRODUCTION

Small mechanical loads cause elastic deformation. Elastic deformation is traceable

to displacements at the atomic level. The bonds between aotms are stretched by the applied
load until the restoring force of the stretched bonds is just sufficient to balance the
stretching force of the applied load. Since there is no long-range atom movement, elastic
deformation is recoverable; when the applied load is removed the atoms return to their
original, equilibrium positions.

The elastic properties of a material govern its response to small loads. The elastic
moduli relate small stresses and strains. The elastic moduli are the properties that are
normally used in the design of engineering structures and devices. Even large structures
must be constrained to very small strains. A 1% strain of a 50 story building changes its
height by almost 10 ft. If the building responded to wind gusts with even a fraction of this
strain, the occupants of its upper floors would be permanently seasick. The dimensional
constraints on mechanical devices are equally stringent. The allowable clearances between
moving parts in engines are measured in thousandths or ten-thousandths of an inch. Plastic
strains are intolerable, and uncompensated elastic strains of a small fraction of a percent
would cause catastrophic interference in many devices. Materials that are subjected to very
small strains are linear in their elastic response; the strain is simply proportional to the
applied stress, in accordance with Hooke's Law. Linear strains are additive; the response of
a material to a complex state of stress can be found by expressing the total stress as the sum
of individual, simple stresses, and summing the strains associated with each of these.

Since the configuration of atoms changes with direction in a crystal, the elastic re-
sponse of a crystal depends on the direction of the load; crystals are elastically anisotropic.
However, the crystalline materials that are ordinarily used in engineering structures are
polycrystalline metals or ceramics whose elastic properties average those of the individual
grains and are, hence, nearly isotropic. Glasses and amorphous polymers are automatically
isotropic. Isotropic materials require only two elastic moduli to specify their behavior. As

Page 681
Materials Science Fall, 2008

we shall see, these can be chosen to be Young's modulus, E, and Poisson's ratio, ˆ, though
other choices are also possible, and often convenient.

Since the most important structural materials are nearly isotropic, and since this case
is the simplest, we shall confine the following discussion to isotropic materials. Note,
however, that there are at least three common situations in which elastic anisotropy must be
taken into account. (1) Single crystals and strongly textured (oriented) crystalline films are
widely used in microelectronic devices. They are often subjected to thermal stresses, and
their elastic anisotropy must be considered in predicting their response. (2) Fiber compos-
ites are widely used in structures that support primarily uniaxial loads, since, as we shall
see, they can be engineered to have exceptional stiffness in the fiber direction. Fiber com-
posites have very different elastic properties in the directions parallel and perpendicular to
the fiber direction. (3) Even a fine-grained polycrystalline metal is often measurably
anisotropic. When the metal is rolled into a sheet or extruded into a bar or wire it develops
a crystallographic texture in which particular crystallographic directions align parallel and
perpendicular to the direction of the primary load. This texture produces an elastic
anisotropy that is sometimes large enough that it cannot be neglected. This is the reason
that tabulations of the elastic properties of metal sometimes list separate moduli for defor-
mation parallel and perpendicular to the rolling direction.

20.2 ELASTIC PROPERTIES OF ISOTROPIC MATERIALS

The elastic properties of an isotropic material are best understood by considering its
response to four basic stress states: tensile load, hydrostatic pressure, simple shear and
balanced shear. The basic stress states are illustrated in Fig. 20.1.

20.2.1 The basic load geometries

(a) (b) (c) (d)

Fig. 20.1: The four basic stress states: (a) tension, (b) hydrostatic com-
pression, (c) simple shear, (d) shear through balanced tensile
and compressive loads.

The simplest loading geometry is uniaxial tension or compression, as illustrated in

Fig. 20.1(a). Uniaxial tension is the load geometry that is ordinarily used to measure elastic

Page 682
Materials Science Fall, 2008

properties, not only because it is the most common loading geometry, but also because
complex stress states can be regarded as the superposition of uniaxial loads. Uniaxial
loading is complicated, however, by the fact that it produces both a shape change (the
sample gets longer) and a volume change (the sample expands slightly). At the atomic
level, shape and volume changes are qualitatively different. Shape changes alter the atom
configuration at constant atomic volume, while volume changes alter the volume at constant
configuration. It is, therefore, useful to consider elastic deformations that isolate these two
effects.

The volume changes at constant shape when the material is subjected to equal loads
along each of three perpendicular axes. This stress state is called hydrostatic compression,
and is illustrated in Fig. 20.1(b). A change in shape at constant volume is accomplished by
shear loads. There are two different ways in which a cube of isotropic material can be
deformed in pure shear The first is illustrated in Fig. 20.1(c); the cube is deformed by
equal and opposite shear stresses applied to opposite faces, and balanced so that there is no
net moment on the cube. The second, balance shear, is illustrated in Fig. 20.1(d); equal and
opposite uniaxial loads are applied in two perpendicular directions. Since the displacements
of the two perpendicular surfaces are equal and opposite, there is no volume change. As
illustrated in Fig. 20.2, a balanced shear is a simple shear of a square rotated 45º with
respect to the square cross-section of the parent cube.

Fig. 20.2: Illustration showing how equal and opposite displacements on

the surface of a square cause a simple shear of an inscribed
square that is rotated by 45º.

20.2.2 Uniaxial tension; Young's modulus and Poisson's ratio

When an isotropic material is loaded in uniaxial tension it elongates in the direction

of the load and contracts in the perpendicular directions, as illustrated in Fig. 20.3. For
small strains from an unstressed initial state, the tensile stress and tensile strain are related
by Hooke's Law:

ßx = E‰x 20.1

Page 683
Materials Science Fall, 2008

where ßx = P/A, the applied load divided by the cross-sectional area, ‰x = ÎLx/Lx is the
strain along the x-axis in the figure, and the constant of proportionality, E, is Young's
modulus.

y
x

Fig. 20.3: Uniaxial tension.

The lateral strain that is induced by a stretch in the x-direction is the same in the two
perpendicular directions, y and z, and is given by

‰y = ‰z = - ˆ‰x 20.2

where ˆ is Poisson's ratio. The fractional change in the volume of the body is, to within
terms of order ‰x2,

ÎV (Lx + ÎLx)(Ly + ÎLy)(Lz + ÎLz) - LxLyLz

V = LxLyLz

~ ‰x + ‰y + ‰z

= (1 - 2ˆ)‰x 20.3

When ˆ « 0.5, as it does, for example, for natural rubber and some other elastomers, ÎV «
0, and volume is conserved during tensile deformation. When ˆ < 0.5, as it is for structural
metals and ceramics, the volume increases during tensile strain. For many structural metals,
ˆ « 0.3.

Now suppose that stresses are applied in the y- and z-directions as well. So long as
the strains are small enough that terms of order ‰2 can be ignored, they can be found by
simply summing the strain increments due to each of the stresses. It follows that the total
strain in the x-direction is:

1 ˆ ˆ
‰x = E ßx - E ßy - E ßz

1
= E [ßx - ˆ(ßy + ßz)] 20.4

The strains in the y- and z-directions, ‰y and ‰z, are given by similar expressions. The three
equations for the strains, ‰x, ‰y and ‰z can be solved for the stresses. The result is

Page 684
Materials Science Fall, 2008

E
ßi = (1-2ˆ)(1+ˆ) [(1-ˆ)‰i + ˆ(‰j + ‰k)] 20.5

where the indices (i,j,k) take the values (x,y,z), (y,z,x) or (z,x,y).

It can be shown that any state of stress that can possibly be applied to a material can
be created by applying normal stresses, ßx, ßy and ßz, along three perpendicular axes, x, y
and z, which are the principal directions of stress for the particular stress state. Since
equations 20.4 and 20.5 hold for all possible choices of orthogonal x- y- and z-axes, it
follows that the two elastic constants, E and ˆ, are sufficient to characterize the elastic
behavior of an isotropic material under any state of stress.

Young's modulus and Poisson's ratio are not only a sufficient set of properties to
determine the elastic behavior of isotropic materials, they are a convenient choice for engi-
neering design. For this reason, E and ˆ are the elastic properties that are tabulated in most
compilations of material properties (often only the modulus is given since, as we shall see,
the values of Poisson's ratio are confined to a narrow range). On the other hand, E and ˆ
are not the most convenient choice of elastic properties for the purpose of understanding
elastic behavior, since the uniaxial tensile test that defines them mixes volume and shape
changes.

20.2.3 Hydrostatic compression; the bulk modulus

When a material sustains a hydrostatic pressure, P, its volume decreases by an

amount that is governed by its bulk modulus, ∫:

ÎV
P=-∫ V 20.6

The compressibility, ˚, is the inverse of the bulk modulus:

1 ∆V
˚ = - V  ∆P  = ∫ -1 20.7
T

To relate the compressibility to Young's modulus and Poisson's ratio, recognize that
a material under hydrostatic pressure, P, is subject to the triaxial stress

ßx = ßy = ßz = -P 20.8

Then, using equations 20.3 and 20.4.

ÎV
V = ‰x + ‰y + ‰z

(1-2ˆ) 3(1-2ˆ)
= E [ßx + ßy + ßz] = - E P 20.9

Page 685
Materials Science Fall, 2008

From which

E
∫ = ˚-1 = 3(1-2ˆ) 20.10

20.2.4 Shear stress and the shear modulus

Let a cubic element of a solid be subject to a shear stress, †, on opposite faces, as

illustrated in Fig. 20.2(c). The effect of this stress is to distort the square cross-section of
the cube into a parallelogram, as illustrated in Fig. 20.3. The shear strain is

ÎLx
©= L 20.11
y

When the shear strain is small, as we assume it is, it is also equal to the change in the angle
at the edge of the cube. The edge, which was originally a right angle, is distorted to the
angle

π ÎLx π
œ = 2 - tan-1 L  ~ 2 - © 20.12
y

ÎL x
†

†
†
Ly

π/2 - ©

Fig. 20.4: The shear strain, ©, caused by the balanced shear stress, †. The
strain, which is assumed small, has been exaggerated for clarity.

The shear stress and shear strain are related by the shear modulus, G:

† = G© 20.13

The shear modulus can be expressed in terms of E and ˆ by the following geometric argu-
ment which is, unfortunately, a bit tortuous.

Page 686
Materials Science Fall, 2008

Let a cube be sheared by the balanced stress illustrated in Fig. 20.1(d), and measure
the shear by the distortion of a square inscribed on the cross-section of the cube, as shown
in Fig. 20.2. As diagrammed in Fig. 20.5, the balanced stress (ß in the x-direction, -ß in
the y-direction) stretches the cube in the x-direction and compresses it in the y-direction, so
that

‰x = - ‰y = ‰ 20.14

Before straining the cube, imagine that a square is inscribed on its cross-section.
Let the edges of the square lie at 45º angles to the x- and y-axes of the cube, as shown in
the figure. After straining, the angle of the inscribed square at the left-hand side of the fig-
ure has changed from π/2 to π/2 - ©. The shear strain, ©, is related to the balanced tensile
strain, ‰, by the geometric relation

π ©  1+‰
tan 4 - 2 =  1-‰  20.15

Since both © and ‰ are small, this equation simplifies to

2ß[1 + ˆ]
© ~ 2‰ = E 20.16

where the final form of the equation follows from eq. 20.4.

L x(1+‰) -ß

L y(1-‰)

ß ß
π/2 - ©

-ß

Fig. 20.5: Cross-section of a cube that is subject to the balanced stress, ±ß,
which produces the balanced tensile strain ±‰. A square in-
scribed on the cross-section of the cube experiences the shear
strain, ©. (Lx = Ly = L.)

To find the shear stress that accomplishes the shear strain, ©, recognize that a tensile
stress on the surface of a body creates a shear stress on a plane that lies at an angle, œ, to the
normal. To find this stress, consider the situation shown in Fig. 20.6. The axial force is

Page 687
Materials Science Fall, 2008

F = ßA 20.17

where A is the cross-sectional area on which ß acts. The component of this force, Ft, that
is tangential to a plane whose normal makes an angle, œ, with the normal is

Ft = F sin(œ) 20.18

This force acts on an area

A(œ) = A/cos(œ) 20.19

so the shear stress is

Ft
†= = ßsin(œ)cos(œ) 20.20
A(œ)

For œ = 45º, as in Fig. 20.5,

† = ß/2 20.21

Fig. 20.6: Illustration showing the shear stress, †, induced in a plane whose
normal makes an angle, œ, with the plane on which the normal
force, ß, acts.

In the configuration shown in Fig. 20.5, the surface of the inscribed square is at 45º
to both the x- and y-faces of the cube. Both normal stresses contribute ß/2 to the shear
stress on that plane, so the balance shear stress, †, is

† = 2(ß/2) = ß 20.22

Substituting the expression for ß obtained from eq. 20.16 then gives

E
† = 2(1+ˆ) © 20.23

Page 688
Materials Science Fall, 2008

So the shear modulus is

E
G = 2(1+ˆ) 20.24

20.3 PHYSICAL SOURCE OF THE ELASTIC MODULI

As we discussed in Chapter 3, it is often convenient to divide the energy of a solid

into two parts:

E = E0(v) + E1({R}) 20.25

where E0(v), the dominant term, is determined by the atomic volume only, and E1({R}) is
determined by the configuration of the atoms for given atomic volume. This decomposition
of the cohesive energy provides a natural framework for understanding the elastic moduli.
The bulk modulus, ∫, governs the response to changes in volume at constant atomic con-
figuration, while (to the extent that we can ignore anisotropy) the shear modulus, G,
governs the response to configurational distortions at constant volume.

20.3.1 The bulk modulus

For a given atom configuration, the energy varies with volume as illustrated in Fig.
20.7. When the volume, and, hence, the interatomic separation is large, the atoms attract
and the energy and volume decrease together. When the interatomic separation is small, the
atoms repel and the energy decreases as the volume expands. Assuming low temperature,
the minimum of the energy determines the equilibrium volume. The pressure required to
maintain a given value of the volume is given by the differential

∆E
P = - ∆v 20.26

and is also plotted in the figure. The bulk modulus is

∆P ∆2E
∫ = - v ∆v = v 2 20.27
dv

that is, ∫ is given by the slope of the function P(V), or the curvature of the function E(V).
For small volume changes about the equilibrium volume, v0, the function P(v) is almost
linear, so the bulk modulus is nearly constant.

Note that the bulk modulus is determined by the curvature of the energy function at
its minimum value, not the minimum value itself. While there is a general tendency for the
bulk modulus to increase with increases in properties that are closely related to the cohesive

Page 689
Materials Science Fall, 2008

energy, such as the melting point and the sublimation energy, the correspondence is not
universal. For example, the diamond modification of carbon, which has the highest known
value of the bulk modulus, is not even the preferred phase of carbon at atmospheric pres-
sure.

P(V)

V0
E
V

E(V)

Fig. 20.7: The variation of energy and pressure with volume.

The bulk moduli of most non-transition metals can be understood on the basis of a
simple model. In these valence metals the radius of the ion core of filled electron shells is
very small compared to the effective radius of the atom in the solid. Most of the space
within the solid is filled by a "gas" of valence electrons, and its bulk modulus is, essen-
tially, the modulus of the electron gas. The isothermal modulus of an ideal gas is

∆P  kT
∫ = - v ∆v  = v = nkT 20.28
T

where v is the volume per particle, n = v-1 is the number of particles per unit volume, and k
is Boltzman's constant. In a metal the valence electron energy is essentially constant with T.
If we replace kT by the Fermi energy, which has a median value of about 4 eV for the
valence metals, then the compressibility is a linear function of the valence electron density
(number of valence electrons per unit volume):

∫ ~ 0.04n 20.29

where ∫ is the bulk modulus in units of 1012 dyne/cm2, and n is the density of valence elec-
trons in units of 1022 per cm3. Eq. 20.29 is a reasonable fit to the data for the valence
metals, which is plotted in Fig. 20.8.

Equation 20.29 does not work well for transition metals or covalent solids, both of
which tend to have much higher bulk moduli than predicted. The reason is relatively
straightforward. In transition metals, the partly filled d-shells in the inner core extend out-
ward to a significant fraction of the atomic radius. It follows that the free volume available
for the compression of the valence electron cloud is severely restricted, and its modulus is

Page 690
Materials Science Fall, 2008

higher. In covalent solids, the valence electron shells are essentially filled by shared elec-
trons. The Pauli exclusion principle restricts rearrangements of the distribution of valence
electrons, raising the modulus.

1.800

1.600 valence metals

transition metals
1.400
covalent solids
1.200
bulk modulus

1.000

0.800

0.600

0.400

0.200

0.000

0 5 10 15 20 25

electron density
Fig. 20.8: The bulk modulus plotted as a function of the valence electron
density for a number of valence metals, transition metals and co-
valent solids.

20.3.2 The shear modulus

The shear modulus, G, measures the resistance of the material to small changes in
atomic configuration at given volume. Those solids whose energies are relatively insensi-
tive to the precise atomic configuration, such as molecular solids like rubber, have relative
low shear moduli. Those solids whose configurations are relatively rigid, including par-
ticularly covalent solids with fixed bond angles, have relatively high values of G.

The elastic behavior of the solid is particularly sensitive to the relative values of ∫
and G. To study this, we use eqs. 20.10 and 20.21 to express the tensile elastic constants,
Young's modulus and Poisson's ratio, in terms of ∫ and G. the result is

9∫G
E = (G+3∫) = 2G(1+ˆ) = 3∫(1-2ˆ) 20.30

3∫ - 2G
ˆ = 2(G + 3∫) 20.31

Page 691
Materials Science Fall, 2008

Consider four cases:

(1) Let ∫ >> G, as it is for molecular solids, like rubber, whose energies are rela-
tively insensitive to the precise molecular configuration. In this case:

ˆ « 0.5, E « 3G 20.32

Since ˆ » 0.5, these materials are almost incompressible in tension, even if their bulk moduli
are relatively small.

(2) Let ∫ « 3G, as it does for many metals. Then

ˆ « 0.35, E « 2.7G 20.33

The value ˆ « 0.33 is a common and useful choice for structural metals.

(3) Let ∫ « G, as is the case when the atomic configuration is rigid, as in strongly
covalent solids like diamond. In that case

ˆ « 0.125, E « 2.25G 20.34

Such a material prefers a relatively large volume change when it is deformed in tension so
that its bond angles are not too severely distorted.

(4) Let G >> ∫, as would be the case in a material whose bond angles were much
more important than its bond lengths. In that case,

ˆ « -1, E « 9∫ 20.35

Since ˆ < 0, such a material would actually expand when strained in tension, since volu-
metric expansion is necessary to preserve bond angles. There is no natural material with ˆ
< 0; the lowest known value is about 0.1. However, researchers have recently made arti-
ficial, foam-like network solids that do have negative ˆ. These are interesting curiosities,
but have no current engineering applications.

20.4 SPECIFIC MODULUS; STIFFNESS

In the design of most engineering structures it is important to choose materials that

maximize stiffness, or resistance to elastic displacement. Put in its simplest terms, the
stiffness of an engineering structure governs the load, P, required to achieve a certain elon-
gation, ÎL. The smaller ÎL for given P, the higher the stiffness. Given Hooke's Law,

P ÎL
ß = A = E‰ = E L  20.36

Page 692
Materials Science Fall, 2008

or

P EA
ÎL = L 20.37

The right-hand side of this equation is a measure of the stiffness, where A is the cross-sec-
tional area of the structural member and L is its length.

Two situations are common in the engineering design of structural members of

given length, L. In the first, space-limited design, one would like the cross-sectional area,
A, to be as small as possible so the structure will not be excessively bulky. Buildings,
bridges and vehicles such as heavy trucks are designed under constraints like this. In this
case, the modulus, E, should be as large as possible so that the area, A, can be as small as
possible for given stiffness. Buildings, bridges, and transport vehicles are normally made
of steel or reinforced concrete, both high-modulus materials, rather than aluminum or plas-
tic, which have relative low elastic moduli.

In the second situation one would like the weight of the structural member to be as
small as possible for given stiffness. This is the case, for example, for aerospace vehicles,
camping equipment, and sporting equipment such as golf clubs, tennis racquets and fishing
poles, along with among many other weight-limited structures. The weight of a member of
length L is

W = ®AL 20.38

where ® is the density. Hence the stiffness is measured by the grouping

P E W
=   20.39
ÎL ® L2

To minimize weight (W) for given length and stiffness the quantity that should be maxi-
mized is the specific modulus,

E
E' = ® 20.40

The materials of choice for weight-limited structures are low-density metals such as
aluminum and titanium, and reinforced plastics such as fiberglass and graphite fiber com-
posites. While the moduli of these materials are typically well below those of structural
steels, their low densities have the consequence that their specific moduli are considerably
higher.

Page 693
Materials Science Fall, 2008

20.5 ENGINEERING THE ELASTIC MODULUS

One of the great limitations of materials engineering is its relative inability to control
the elastic properties of materials. Since these reflect behavior at the atomic level, they are
relatively insensitive to changes in the microstructure. However, there are instances in
which beneficial changes can be made, by adjusting the molecular structure, or by altering
the microstructure to introduce fibers or inclusions with exceptionally high moduli.

20.5.1 Composition

Small changes in the chemical composition of a solid do not ordinarily cause signif-
icant changes in its modulus. To a first approximation, the modulus of a solid solution is
the average modulus of its components, and is not significantly changed until the solute
content becomes so large that the nature of the material itself is changed. For example, the
moduli of the alloy steels that are used in engineering structures are almost the same, inde-
pendent of composition.

An important exception occurs in the alloying of aluminum with lithium. For rea-
sons that are still not well understood, the addition of about 2 weight percent Li to Al raises
its modulus by almost 10%. Since Li is a light element, the Li addition also lowers the
density of the alloy, so the improvement in the specific modulus is even greater. Because
of this behavior, Al-Li alloys have been under intensive development for a number of years
as candidate alloys for aerospace vehicles. The problem is to achieve improved elastic
properties without sacrificing other needed properties, such as strength, fracture toughness
and fabricability.

20.5.2 Microstructure

The elastic modulus is significantly affected by crystal structure, and can be signifi-
cantly altered by choosing the crystal structure when more than one structure is available.
For example, high alloy steels can be stabilized in the high-temperature, FCC structure, and
have lower moduli than similar steels with the normal BCC structure.

While crystal structure modification is not ordinarily a useful option for structural
metals, the structures of molecular solids often can be modified to improve elastic proper-
ties. The classic examples are the elastomeric solids, whose molecules are kinked and
coupled so that they behave like microscopic springs and permit large elastic deflections.
As we discussed in Chapter 6, both the internal structure of the elastomer molecule and the
cross-linking between adjacent molecules are controlled to adjust the elastic modulus.
Many engineering elastomers are complex copolymers that also contain inorganic fillers to
control the complex of mechanical properties.

20.5.3 Composite materials

The materials that offer the most attractive combinations of high modulus and low
density are ceramic materials, glasses, or fibers that are inherently brittle and, hence, un-

Page 694
Materials Science Fall, 2008

suitable for many structural applications. To achieve some of the benefit of these materials,
they are used as isolated fibers or particles that are embedded in a matrix of more ductile
material to inhibit brittle failure. The elastic modulus of a composite material is maximized
if the high-modulus component is used as a continuous fiber in the direction of primary
load. However, these fiber composites are highly anisotropic, and are best used in situa-
tions where the applied load is uniaxial. For structures that are subject to more complex
load geometries, it is still possible to obtain some of the benefit of high-modulus fillers by
cross-plying fibers in two or three dimensions, or using randomly-oriented, discrete rein-
forcement particles.

Fig. 20.9: Loading in an idealized fiber composite (load-bearing elements

in parallel) and particulate composite (load-bearing elements in
series).

The difference in behavior between a fiber composite and a particulate composite is

due to the manner in which the load is shared between the reinforcing particle and the ma-
trix. The difference is illustrated schematically in Fig. 20.9. To a first approximation, the
components of a fiber composite act in parallel, while those of a particulate composite act in
series. In a fiber composite, the elements have equal strain, while in a particulate composite
they carry equal stress.

Fiber composites

A fiber composite consists of continuous fibers of one material, embedded in a ma-

trix of a second material. In the usual case, for example, a graphite-epoxy composite, the
fiber is a high-modulus material that has the advantage of being very stiff and strong, but
the disadvantage of being brittle and, hence, liable to fracture. The matrix is a relatively
soft, ductile material that holds the fiber and shields it from fracture under normal loads.

The elastic modulus of a fiber composite is maximized when the fibers are aligned
parallel to one another, and the load is applied along the direction of the fibers. To estimate
the modulus of in this case, recognize that the fiber and matrix are stretched by almost
identical amounts by the applied loads (Fig. 20.9). Hence,

‰1 = ‰2 = ‰ 20.41

Page 695
Materials Science Fall, 2008

where ‰1 is the strain in the fibers and ‰2 is the strain in the matrix. The total load, P, re-
quired to achieve the strain, ‰, is

P = ß1A1 + ß2A2

= A(ß1f1 + ß2f2) 20.42

where ß1 and ß2 are the tensile stresses in the fiber and matrix, respectively, and f1 and f2
are their areal fractions in the cross-section. If E1 and E2 are the moduli of the fiber and
matrix, and ß is the overall average tensile stress,

P
ß = A = ‰(f1E1 + f2E2)

= E‰ 20.43

where E, the effective modulus of the composite, is the average of the moduli of the ele-
ments:

E = E1f1 + E2f2 20.44

It follows that the effective modulus of the composite in the fiber direction is most
strongly influenced by the high-modulus component, and increases with the areal fraction
of that component in the cross-section perpendicular to the applied load.

Equation 20.43 shows how the modulus of a material can be dramatically increased,
at least along the direction of the load, by introducing fibers of high-modulus material. In
particular, the specific modulus, E' = E/®, can be engineered to exceptionally high values by
inserting fibers of a low-density, high modulus material, such as the graphite fibers
described in Chapter 6, in a matrix of low-density material, such as epoxy. This is the
essential feature of the graphite composite materials that are used in many weight-critical
structures, such as aircraft parts and golf club shafts. These materials have the
disadvantage, however, that their properties are anisotropic; the exceptionally high modulus
is only achieved for loads that are applied parallel to the fiber axis. Hence they are most
useful in structures whose dominant load is in a single, predictable direction, as it is in the
shaft of a golf club or the blades of a rotating fan, among many other applications.

Particulate composites

To a first approximation, the elements of a particulate composite act in series (Fig.

20.9) so that they support the same stress (to be honest, this is a pretty crude approxi-
mation; computing accurate moduli for a particulate composite is a formidable problem in
mathematics). Assuming load sharing in series, the total strain of the composite is the sum
of the strains in the two elements:

Page 696
Materials Science Fall, 2008

‰ = ‰1f1 + ‰2f2

 f1 f2 
= ßE + E  20.45
1 2

so the modulus of the composite is obtained by averaging the reciprocals of the moduli of
the components

1 f1 f2
E = E1 + E2 20.46

It follows that the modulus of a particulate composite depends much less strongly
on the modulus of the more rigid component. Nonetheless, one can raise the modulus of a
material by filling it with particulates of a second, more rigid phase, and the product has the
advantage that it is more nearly isotropic in its elastic properties than a fiber composite of
the same materials would be.

A familiar example of a particulate composite is common fiberglass, which has

chopped fibers of glass distributed through an epoxy matrix. The modulus is less than it
would be if continuous glass fibers were used, but the material can be easily molded into
sheets with complex shapes, and can be made to be nearly isotropic in the plane of the
sheet, which makes it useful for structures, such as the hulls of ships and the underbodies
of aircraft, that sustain two-dimensional loads.

A second important class of a particulate composites includes the metal matrix com-
posites that are currently under development for many applications in aircraft. The most
widely used are aluminum alloys that contain dispersions of chopped fibers or crystalline
whiskers of silicon carbide. The high modulus of the SiC increases the modulus of the Al
alloy, while the lower density of SiC lowers the average density, adding a further increment
to the specific modulus.

Page 697
Materials Science Fall, 2008

Chapter 21: Plastic Deformation

Steel ... is capable of being made the hardest of all metals. The way
of making it so is this: only to heat it red-hot in the fire, and then throw it all
at once into cold water; and this matter of hardening is what is called
tempering it, and this makes it capable of cutting or at least breaking all sorts
of bodies without exception, even diamonds themselves: For it is certain
they will break in pieces with a small stroke with a hammer if it hits right.
- Jacques Rohault, Traite' de Physique (1671) [trans. John Clarke (1723)]

21.1 INTRODUCTION

An arbitrary deformation of a material can always be described as the sum of a

change in volume and a change in shape at constant volume (shear). Assuming constant
structure, the change in volume is recovered when the load is removed, since the atoms can
simply relax back to their equilibrium sizes. The change in shape, on the other hand, may
or may not be recovered, since the atoms can relax into new positions that are
configurationally identical to the original ones, but displaced from them. The part of the
shear that is recovered is elastic, the part that remains is plastic. Plastic deformation is a
permanent change in shape through shear.

In the tensile test, a material deforms plastically when it is loaded to a stress beyond
its yield strength. The extent of plastic deformation then increases with the applied stress
until plastic instability sets in at the ultimate tensile strength, the largest value of the
engineering stress that the material can withstand without failure. If the ultimate tensile
strength is exceeded, plastic instability leads to catastrophic necking and fracture.

A material also deforms plastically when it is subjected to a constant load.

Deformation under constant load is called creep. Creep deformation inevitably requires
some mass diffusion and, hence, occurs at a negligible rate unless the temperature is high
(T > 0.5 Tm is a rough criterion). When a material creeps under constant load, then, after
an initial transient, it deforms at a constant rate (steady-state strain rate) until accumulating
microstructural damage creates an instability that is characterized by a rapidly increasing
strain rate and eventual failure by creep rupture.

21.1.1 Engineering significance of plastic deformation

Engineers care about plastic deformation for at least four reasons. The first, and
most common concern, is to avoid it. Very few engineering structures are designed to
tolerate more than infinitesimal plastic distortions. They must, therefore, be designed to

Page 698
Materials Science Fall, 2008

operate at stresses well below the yield strengths of the structural materials that are used in
them. This requires that yield behavior be understood and controlled.

Second, plastic deformation is used to shape materials. The vast majority of

manufactured objects contain parts that are formed by plastic deformation. The successful
design and efficient manufacture of these products requires an understanding of the nature
and limitations of plastic deformation.

Third, plastic instability is a major concern in both manufacturing and product

safety. It restricts elongation and, hence, limits the possibility of manufacturing complex
shapes with simple forming operations. It also sets the ultimate tensile strength and, hence,
fixes the maximum load that can possibly be supported in a particular engineering structure.
A structure that is flawed or brittle may, of course, fail by fracture at loads that are well
below the ultimate tensile strength but, however well it is made, a structure can never
support a load that exceeds the ultimate tensile strength.

Fourth, plastic deformation is used as a diagnostic device to understand the

performance and interpret the failures of engineering structures. The pattern of deformation
in a structure often provides useful information on the magnitude and geometry of the
dominant loads it has experienced. This information may indicate service problems and
identify the cause of service failures.

21.1.2 Mechanisms of plastic deformation

There are four generic mechanisms of plastic deformation. While all of them are
important, we shall concentrate on the first, dislocation plasticity, since this is the dominant
mechanism of deformation of structural materials.

1. Dislocation plasticity. First, and most commonly, planes of atoms can slip over
one another like cards in a deck, leading to an overall shear that is localized within specific
atom planes. It is always energetically favorable to accomplish this slip a little at a time, as
one would move a large rug across a floor. And it is usually favorable to slip in increments
that correspond to a lattice displacement, so that the area of the plane that has slipped
maintains a perfect crystallographic match with the plane beneath it. In this case the
boundary of the slipped area is a linear defect, called a dislocation, that we defined and
discussed in Chapter 4. The motion, multiplication and interactions of dislocations are
responsible for the yield, work hardening and plastic instability that characterize the stress-
strain behavior of structural alloys.

2. Diffusion. Second, individual atoms can move so that the crystal becomes
longer in one or more of its dimensions and correspondingly shorter in the others. In a
crystal individual atoms move by diffusion, and this process is known as diffusional creep.
Diffusional creep is dominant at high temperature, where the rate of diffusion is
appreciable, and at low stresses, which are insufficient to produce a significant rate of
dislocation plasticity.

Page 699
Materials Science Fall, 2008

3. Structural transformations. Third, all of the atoms in the crystal, or some

subvolume of it, can move simultaneously to accomplish the shear. We have already
discussed spontaneous structural phase transformations, such as the martensitic
transformation in steel, that change the size and shape of the parent crystal.
Transformation-induced plasticity can be an important deformation mechanism in materials
that undergo martensitic transformations, particularly when the material is used at tempera-
tures where it is metastable with respect to a martensitic transformations that can be trig-
gered by the applied stress (as is the case with many austenitic (FCC) stainless steels), or
when it is subject to thermal cycling through the martensitic transformation (as is done to
achieve the shape memory effect in a number of modern alloys). In addition to those trans-
formations that change the basic crystal structure, it is also possible for the stress to drive
martensitic-like structural changes that reorient the crystal without changing the basic crys-
tal structure. The most familiar example is mechanical twinning, in which a region of the
crystal spontaneously shears into a configuration that has the same crystal structure as its
parent, but a different orientation in space. Mechanical twinning is an important low-
temperature deformation mechanism in metals that have non-cubic crystal structures, such
as the HCP metals, and is also foiund in cubic metals such as Cu and many steels, under
the appropriate combination of load and temperature.

4. Tumbling. Fourth, a material can be deformed by the cooperative, liquid-like re-

configuration of groups of atoms, molecules or small grains. This is an important mecha-
nism of deformation in amorphous metals, glasses or polymers, and is also important in the
high-temperature creep of fine-grained materials (where it is called grain-boundary sliding).
The scientific description of these deformation mechanisms is in its very early stages, and I
am coining a term by calling it tumbling, which is the best descriptive term I can think of
for what is going on. The deformation of very fine-grained or amorphous materials is an
active area of current research. The practical motivation for this research is the recent de-
velopment of bulk glassy metals and new, nanostructured alloys that have unusual mechan-
ical properties. A description of these deformation mechanisms is beyond the scope of this
course.

In this course we concentrate almost exclusively on dislocation plasticity.

21.2 THE YIELD STRENGTH

The yield strength of a typical crystalline material is related to the stress that is
needed to move dislocations to cause significant plastic deformation. To understand the
yield strength we must first understand the nature of the deformation caused by a disloca-
tion and the force that impels it to move.

21.2.1 Deformation by dislocation motion

As discussed in Chapter 4, the planar motion of a dislocation causes plastic defor-

mation by slip. The part of the crystal that is above the plane and behind the moving dislo-
cation is slipped by the vector, b, the Burgers vector of the dislocation, with respect to the

Page 700
Materials Science Fall, 2008

part below the plane (Fig. 21.1). When the dislocation passes through the crystal it dis-
places the material above the plane by the Burgers vector, b, with respect to that below it.

“ “
b † b

Fig. 21.1: Slip caused by the motion of an edge dislocation.

As shown in Fig. 21.1, the deformation that is accomplished by the dislocation is a

shear in the direction of b. It follows that the stress that causes dislocation motion is a
shear stress, †, resolved in the direction of b. It can be shown that the net force that acts on
the dislocation is

F = †b 21.1

where b is the magnitude of the Burgers vector and † is the shear stress that acts in the di-
rection of b.

For simplicity, Fig. 21.1 illustrates the shear caused by an edge dislocation. But our
conclusion is true whatever the type of the dislocation. For example, consider the dis-
location loop shown in Fig. 21.2. The crystallographic character of the dislocation changes
from edge (b perpendicular to the dislocation line) to mixed (b angled to the dislocation
line) to screw (b parallel to the dislocation line) around the circle of the loop. However, the
area enclosed by the loop has been slipped by the constant vector, b. If the loop is ex-
panded to sweep out the whole plane, the final configuration is identical to that shown in
Fig. 21.1; the material above the plane of the loop is slipped with respect to that below the
plane of the loop by the vector, b.

edge screw †

Fig. 21.2: A planar dislocation loop. The material above the area enclosed
by the loop is slipped with respect to that below the loop by the
Burgers vector, b.

Page 701
Materials Science Fall, 2008

Since an expansion of the dislocation loop causes a shear displacement in the direc-
tion of b, the stress that drives the expansion is the shear stress that acts in the direction of
b, as shown in the figure. Since an expansion of the loop is accomplished by displacing
the dislocation line normal to itself along the whole periphery of the loop, the force that acts
on the dislocation line, F = †b, is everywhere perpendicular to the dislocation line, and
oriented so that it tends to expand the loop (Fig. 21.3).

†b

b
†b †b

Fig. 21.3: The force that acts to expand a dislocation loop has magnitude,
†b, and is perpendicular to the dislocation at every point.

It follows that, in general, the force on a dislocation line has the magnitude given by
eq. 21.1, and is directed perpendicular to the dislocation line, whatever the crystallographic
character of the dislocation.

21.2.2 The critical resolved shear stress

Plastic deformation is driven by shear stress, though we ordinarily think of it in

terms of the tensile stress, ß, and quantify the yield behavior of a material by tabulating its
tensile yield strength, ßy. To see the connection between the tensile stresses and yielding
behavior, it is necessary to understand how a tensile stress can produce a resolved shear
stress that acts on a dislocation. We shall do this for the simple case of uniaxial tension.

Let a tensile stress, ß, be applied to a cylindrical specimen, as illustrated in Fig.

21.4. A dislocation with Burgers vector, b, lies in a plane whose normal makes the angle,
œ, with the tensile axis. The direction of b makes the angle, ƒ, with the tensile axis. The
force on the vertical face of the bar is

F = ßA 21.2

where A is the area of the face. The component of this force in the direction of b is

Ft = Fcos(ƒ) = ßAcos(ƒ) 21.3

Page 702
Materials Science Fall, 2008

The area, A', of the plane on which Ft acts is

A' = A/cos(œ) 21.4

Hence the resolved shear stress in the direction of b is

† = Ft/A' = ßcos(œ)cos(ƒ) 21.5

The resolved shear stress has its maximum value when both œ and ƒ lie at 45º to the tensile
axis, in which case

† = †max = ß/2 21.6

œ ƒ
ß

Fig. 21.4: A uniaxial tension, ß, produces a resolved shear stress, †, along

the Burgers vector, b, of a dislocation that lies in a plane whose
normal is tilted by œ from the tensile axis.

There is always a frictional resistance to the motion of a dislocation through a crys-

tal. In the simplest case, this is due to the need to break and restore bonds as the dislocation
is displaced through an elementary step. As we shall discuss below, several other mi-
crostructural mechanisms contribute to the frictional resistance, and these can be controlled
to adjust the yield strength of the material. The frictional resistance has the consequence
that, at least at low to moderate temperature, a dislocation is stationary until the resolved
shear stress on it exceeds a critical value, call the critical resolved shear stress, †c.

The critical resolved shear stress depends on the nature of the dislocation, which is
characterized by the Burgers vector, b, and the plane in which the dislocation lies. In
typical structural materials, †c has its least value when b lies in a close-packed direction (the
direction which provides the minimum value of |b| for a lattice dislocation), and when the
dislocation line lies in a close-packed plane. Since such dislocations are most easily moved,
they tend to dominate plastic deformation; the microscopic slip that happens during plastic
deformation tends to be in close-packed directions on close-packed planes. The combina-
tion of the slip plane and slip direction is called the slip system.

Page 703
Materials Science Fall, 2008

21.2.3 The tensile yield strength of a single crystal

To initiate plastic deformation in a crystal that is stressed in uniaxial tension, the

tensile stress, ß, must be large enough to produce a resolved shear stress, †, that exceeds
the critical resolved shear stress for a least one set of dislocations within the material. From
equation 21.5,

 †c 
ßy = mincos(œ)cos(ƒ) 21.6

signifying that the yield strength is the minimum value of the factor in brackets, where †c is
the critical resolved shear stress for a dislocation whose slip plane is tilted by the angle, œ,
from the tensile axis, and whose slip direction (the direction of b) lies at the angle, ƒ.

Two important results follow immediately from eq. 21.6. First, while the critical
resolved shear stress is a material property, the tensile yield strength is not. It depends on
the orientation of the crystal. Only its minimum value is a material property, and is realized
when the most favorable slip system (minimum †c) has œ = ƒ = 45º. In general,

ßy ≥ 2†c 21.7

Second, the preferred slip system may change with the orientation of the crystal.
This does not happen in FCC metals. In FCC the preferred slip system is {111}<110>, the
close-packed direction in the close-packed plane, and the angle between {111} planes is
sufficiently small that there is always a {111}<110> set available for glide at stresses not
too far above the minimum yield stress. In BCC metals, on the other hand, several slip
systems are used, including {110}<111>, {112}<111>, and {123}<111>. Each of these
systems has a <111> glide direction, which is the close-packed direction in BCC. They
differ in the choice of glide plane. The reason is that BCC is not a close-packed structure.
The geometry of the {110}, {112} and {123} planes is not all that different, and they have
comparable values of †c. Glide occurs on the plane that is most favorably oriented.

Fig. 21.5: The octahedron of {111} planes in the FCC structure.

Page 704
Materials Science Fall, 2008

Multiple slip systems are also used in HCP metals, but in this case the reason is the
low symmetry of the HCP crystal structure. The only close-packed plane in HCP is the
basal plane of the HCP cell. While dislocation glide is relatively difficult on the prismatic
planes of HCP (those angled to the basal plane), these become favored when the tensile
axis is either near (œ « π/2) or perpendicular (ƒ « π/2) to the basal plane. Similar consid-
erations apply to other crystals with non-cubic symmetry.

Finally, note that while dislocation slip produces a shear strain, shear at an angle to
the tensile axis causes a net elongation along that axis, as illustrated in Fig. 21.6. The
tensile stress that triggers slip is the yield strength that is measured in the tensile test.

elongation slip

ß ß

Fig. 21.6: Slip on a plane that is angled to the tensile axis causes elongation
in the direction of the tensile axis.

21.2.4 The yield strength of a polycrystal

When the material is a polycrystal, the definition of the yield strength is somewhat
ambiguous. Most polygranular materials exhibit two distinct kinds of yielding behavior
when they are stressed in uniaxial tension: local yielding, and general yielding.

ß ß

Fig. 21.7: Local yielding by slip on the most favorable slip system in a
polygranular body.

Local yielding occurs when the applied stress is sufficient to trigger dislocation mo-
tion in the weakest element of the microstructure. In the ideal case, this happens when the
applied stress is sufficient to move the most favorably oriented dislocations in the poly-
granular body, that is, when

ß = 2†c 21.8

Page 705
Materials Science Fall, 2008

Most real polygranular materials experience local yielding at even smaller stresses, since
they contain internal stresses left over from processing or prior service that add to the
applied stress, and heterogeneities that cause stress concentrations that magnify the applied
stress. Local yielding leads to a situation like that shown in Fig. 21.7; dislocation motion,
and, hence, plastic deformation, is confined to a few grains in the interior of the
polygranular body.

The earliest incidents of local yielding ordinarily do not propagate, since the grains
around the yielded grain are unlikely to have equally favorable slip systems. Nonetheless,
local yielding produces a net plastic elongation of the overall sample. If a volume, V', in a
sample of size, V, undergoes the strain, ‰', then it can be shown that the sample undergoes
the overall strain, ‰, where

V'
‰ =  V  ‰' 21.9

It follows that early yielding produces overall plastic strains that can be measured (at least
with instruments that are sensitive enough), and causes the stress-strain curve to deviate
from linearity.

As the stress is increased beyond that required to cause local yielding, an increasing
volume of the specimen is plastically deformed, and the stress-strain curve deviates more
noticeably from its initial, linear slope. Eventually, the stress becomes sufficient to cause
general yielding, in which the whole specimen behaves as an essentially plastic body.

The general yielding of a polygranular specimen requires that its grains be able to
deform simultaneously. This requires that the typical grain have enough active slip systems
to accomplish an arbitrary change of shape. It can be shown that at least five independent
slip systems are needed to accomplish an arbitrary change of shape. To activate five
independent slip systems, the yield stress must exceed the value given in eq. 21.8 by a
factor known as the Taylor factor. For a cubic crystal, the Taylor factor is approximately
1.5, so the stress required for general yielding, which is often used as the theoretical
definition of the yield strength, is

ßy ~ 3†c 21.10

Note two features of yielding in polycrystals. First, whether local or general

yielding is used as the criterion for the onset of plastic deformation, the material property
that is most important is the critical resolved shear stress, †c. The microstructural control of
yield strength is accomplished by manipulating the microstructure to adjust †c.

Second, the yield of a typical polycrystal is gradual rather than abrupt. The yield
strength is, therefore, largely a matter of definition. The usual practice is to define the yield
strength as the "0.2% offset load", that is, the stress required to accomplish a plastic strain
of 0.2%. The method of taking the measurement is illustrated in Fig. 21.8. A line is drawn
parallel to the linear, elastic portion of the stress-strain curve that intersects the strain axis at

Page 706
Materials Science Fall, 2008

a strain of 0.2%, and the yield stress is defined as the stress at which this line intersects the
stress-strain curve.

ßy

}
0.2% ‰

Fig. 21.8: The method of measuring the 0.2% offset yield strength.

The 0.2% offset yield has the dual advantages that it is relatively easy to measure in
practice, and, for most materials, corresponds fairly well to the stress required for general
yielding.

21.3 MICROSTRUCTURAL CONTROL OF THE YIELD STRENGTH

The yield strength is modified by controlling the critical resolved shear stress, †c.
The value of †c is controlled by placing obstacles in the plane of the dislocation that make it
difficult to move. The inherent value of †c is the Peierls-Nabarro stress that is due to the
crystal lattice itself; the atoms near the dislocation core must be moved and reconfigured for
the dislocation to move through an elementary step. Additional increments to †c come from
crystal defects that interact with dislocations. A dislocation is both an elastic and a
crystallographic defect. Its elastic field interacts with the fields of any other defects that
distort the crystal lattice, such as point defects and other dislocations. Its crystallographic
defect interacts with any other defects that disturb the crystallography of the lattice, such as
dislocations, grain boundaries and precipitate particles. The strength increment produced
by a particular kind of hardening defect depends on the intensity of its interaction with
dislocations and on its density within the material.

If several distinct, independent hardening mechanisms, such as solute atoms, dislo-

cations and grain boundaries operate in the same material, their strength increments are ap-
proximately additive. For this reason it is possible and useful to employ several different
mechanisms simultaneously to control the yield strength.

21.3.1 Structural hardening; the Peierls-Nabarro stress

The inherent resistance to dislocation motion is due to the fact that the atoms near
the core of the dislocation are displaced and reconfigured as the dislocation moves from one
stable position to another. Put simply, atomic bonds must be broken and re-established, as

Page 707
Materials Science Fall, 2008

illustrated in Fig. 21.9 for the case of an edge dislocation in a simple cubic solid. In order
for the dislocation to move, the force that acts on it (†b) must be sufficient to accomplish
this. The necessary value of the resolved shear stress, the Peierls-Nabarro stress, †p, has
been estimated for a simple dislocation, and is given by the expression

2G  2πd 
†p = 1-ˆ exp- b(1-ˆ) 21.11

where G is the shear modulus, ˆ is Poisson's ratio, b is the Burgers vector, and d is the
spacing between adjacent slip planes.

While eq. 21.11 is based on a simple model that does not account for many of the
important details that influence dislocation behavior in particular crystals, such as the pos-
sibility of splitting into partial dislocations that was discussed in Chapter 4, it is qualita-
tively accurate, and provides a good basis for discussing the relative inherent strengths of
various materials. Three of its features are particularly worth noting.

First, if eq. 21.11 is solved for values of the variables that are typical for simple
metals, it predicts a critical resolved shear stress of the order 10-4 to 10-3G, which is about
the right order of magnitude for typical metals in their purest forms.

(a) (b)

Fig. 21.9: The reconfiguration of atoms necessary for an edge dislocation

in a simple cubic crystal to glide by an elementary step.

Second, eq. 21.11 predicts that †p has its minimum value for slip on the atomic
planes that are most widely separated. These are the close-packed, {111} planes in FCC
metals, and the predicted value of †p for these planes is much less than for any of the non-
close-packed planes. This explains why slip on close-packed planes dominates in the de-
formation of FCC metals. On the other hand, BCC metals have several separate sets of
planes whose spacings are not all that different, which helps to explain why several differ-
ent slip systems are observed in BCC.

Third, eq. 21.11 predicts that the inherent strength of a material increases with its
shear modulus. In particular, materials that have very high shear moduli should be excep-
tionally strong. This is true. High-modulus oxides such as Al2O3 and SiO2, high-modulus
nitrides and carbides such as WC, TiC, TiN, SiC and Si3N4 are among the strongest

Page 708
Materials Science Fall, 2008

materials known. The ultimate case is diamond, which combines the highest known value
of G with the lowest value of ˆ, and has, as predicted, the highest inherent strength of any
known material.

However, despite their exceptional yield strengths, these super-strong materials are
rarely used as structural materials. For reasons we shall discuss in the following chapter,
their high yield strength has the consequence that they are exceptionally brittle, and fracture
easily under tensile loads. They can support compressive loads, and, for this reason, have
exceptionally high hardness. They are used as abrasives, in tool bits and as wear-resistant
coatings where their high hardnesses provide exceptional properties.

The inherent strengths of the common structural metals also follow eq. 21.11, and
scale with G. However, their inherent strengths (yield strength of a well-annealed, chemi-
cally pure material) are low. Microstructural hardening mechanisms are used to make them
more useful as structural materials.

21.3.2 Grain refinement

One of the simplest and most useful ways to strengthen a structural metal is by re-
fining its grain size. Grain boundaries are discontinuities in the crystal structure that act as
barriers to dislocations. Since the orientations of the active slip planes change abruptly at
grain boundaries, slip must be transmitted indirectly from grain to grain. When a dislo-
cation impinges on a grain boundary its stress field produces shear stresses on the poten-
tially active slip planes of the adjacent grain. These add to the applied load and help to
propagate plastic deformation by the motion of independent dislocations in the adjacent
grain.

Fig. 21.10: A dislocation pile-up in a grain that has yielded.

Large grains are particularly efficient at transmitting strain to their neighbors. When
a large grain slips, a number of dislocations glide along the preferred plane (or along
closely spaced, parallel planes) and pile up against the grain boundary, as illustrated in Fig.
21.20. The stress at the head of such a dislocation pile-up is magnified; if the resolved
shear stress is †, the stress at the head of a pile-up of n dislocations is n†. The larger the
grain the more easily extensive pile-ups develop, and the more easily strain is transmitted

Page 709
Materials Science Fall, 2008

across the boundary. The consequence is that the yield strength of a material decreases
with its grain size.

The yield strength of a typical metal varies with its grain size according to the Hall-
Petch relation:

K
ßy = ß0 + 21.12
d

where d is the mean grain size and K is a constant whose value depends on the material and
the characteristics of the microstructure. Refining the grain size can lead to a substantial in-
crease in strength.

It is relatively easy to control the grain size during the manufacture of structural
metals and alloys. These are normally processed by hot deformation (rolling or drawing)
into plates, sheets or bars. The hot deformation is done at temperatures above the
recrystallization temperature of the metal. Recrystallization produces fine grains that grow
on further exposure to elevated temperature, so the mean grain size is a function of the
temperature of hot deformation and the holding time after deformation. The grain size is
controlled by cooling the material at a pre-selected time after its final hot deformation.

21.3.3 Obstacle hardening

The other common hardening schemes employ microstructural obstacles that inhibit
dislocation slip through the grain interiors. These barriers may be solute atoms, forest
dislocations that thread through the slip plane, or small second-phase precipitates. In each
of these cases, the obstacle is a localized barrier that can be idealized as a point in the slip
plane.

†b †b
Ls R

Fig. 21.11: A dislocation, modeled as a flexible string, pressing against

obstacles that are separated by the distance, Ls.

Let a dislocation move over its slip plane under the action of a stress, †, that is sig-
nificantly larger than the Peierls-Nabarro stress, †p. In this case, the atomic structure of the
slip plane is relatively unimportant, and the dislocation behaves roughly like a flexible, ex-
tensible string. The dislocation has an energy per unit length, T, which is an effective line
tension, and has the approximate magnitude

Page 710
Materials Science Fall, 2008

1
T « 2 Gb2 21.13

The line tension opposes any increase in dislocation length.

Let the dislocation encounter a linear array of barriers that oppose its motion. The
dislocation presses against these barriers to create local configurations like that shown in
Fig. 21.11. To a reasonable approximation, the dislocation bows out between adjacent
obstacles to an equilibrium radius

T
R = †b 21.14

where †b is the force on the dislocation. Assuming that R > Ls/2, where Ls is the spacing
between adjacent obstacles, then the dislocation can only advance by penetrating through
the obstacles themselves.

The force that the dislocation exerts on the obstacles is due to its line tension, and
equal to

F = 2Tcos(¥/2) 21.15

where ¥ is the angle between the arms of the dislocation at the obstacle. If the obstacles are
equally spaced along the dislocation line,

Ls
cos(¥/2) = 2R 21.16

Let Fc be the force required for the dislocation to cut through the obstacle. As the stress, †,
is increased, the angle, ¥, decreases. The obstacle is passed when ¥ falls to ¥c, where

Fc
cos(¥c/2) = ∫ c = 2T 21.17

In practice, the force, Fc, tends to scale with the factor Gb2 just as T does, and the resistance
of the obstacle is characterized by the dimensionless strength, ∫c. Using equations 21.13,
21.14, 21.16 and 21.17, the critical resolved shear stress for dislocation motion through a
regular array of obstacles with spacing, Ls, is

Gb
†c « L ∫ c 21.18
s

In the typical real case, the obstacles are distributed over the plane in an approxi-
mately random pattern with mean spacing

Page 711
Materials Science Fall, 2008

1
Ls = 21.19
n

where n is the average number of obstacles per unit area. Research has shown that when
the obstacles are randomly distributed, equation 21.18 is changed to

†c « Gb∫ c3/2 n 21.20

Eq. 21.20 predicts that the critical resolved shear stress increases with the shear
modulus, with the 3/2 power of the obstacle strength, and with the square root of the
obstacle concentration. The equation holds reasonably well for hardening by solute atoms
(in the limit of small concentration), in which case ∫c is in the range 0.01-0.05, for
hardening by "forest" dislocations that thread through the glide plane, with ∫c in the range
0.1 to 0.3, and for hardening by small precipitate particles, with ∫c in the range 0.5-0.8.
Since the different types of microstructural obstacle have significantly different strengths,
they make essentially independent contributions to the critical resolved shear stress.

21.3.4 Solution hardening

Solute atoms never "fit" quite properly in the parent lattice, so there is always some
local distortion of the lattice in the vicinity of the solute (the misfit defect). Moreover, the
bonding around the solute is never quite the same as that in the parent lattice, so there is
also some difference in the local value of the elastic constants near the solute (this is often
called the modulus defect). The net result is that a solute atom acts as a barrier to
dislocation motion; a dilute distribution of solute atoms act as a distribution of obstacles of
the type considered in the previous section.

While the obstacle strength of solute atoms (∫c) is relatively small, the areal density
in the slip plane is relatively large, even when the solute concentration is much less than
1%. Solution hardening is an effective hardening mechanism that is widely used. When
the solution is dilute, the yield strength is given by an equation of the form

ß = ß0 + SGb c 21.21

where ß0 is the yield strength of a solute-free material with the same microstructure and S
is a constant. (When the solute concentration becomes appreciable, the strain fields of the
individual solute atoms overlap so they no longer behave like discrete obstacles. In this
regime the strength increases roughly as c2/3.)

The strength (∫c) of the solute defect is primarily due to its misfit in the parent lat-
tice. It follows that interstitial solutes strengthen an alloy much more effectively than sub-
stitutional ones, as illustrated in Fig. 21.12.

Page 712
Materials Science Fall, 2008

interstitial

ßy
substitutional

Ôc
Fig. 21.12: Solute hardening in the dilute solution limit.

Interstitial solution hardening is more effective in BCC crystals, where the

interstitial sites are small and asymmetric, than in FCC crystals, where they are larger and
equiaxed. The reason, as discussed in Chap. 4 for the particular case of the BCC structure,
is the relatively large lattice strain associated with an interstitial defect in the asymmetric
site. Carbon and nitrogen are the common interstitial solutes in structural steels, and they
harden conventional, BCC steels much more effectively than stainless steels that are
stabilized in the FCC structure.

On the other hand, while interstitials are less effective hardening species in FCC
crystals, their solubility is relatively high. High-strength, nitrided stainless steels are high
strength steels with many applications. Many commercial Al alloys are strengthened by
substitutional solutes. These are even less effective hardening species on an atom-by-atom
basis, but have high solubility in the FCC Al lattice.

The yield point

The diffusional mobility of solute atoms may also affect the strength, particularly
when the solute is a mobile interstitial or when the test temperature is relatively high. The
reason is that solute atoms are attracted to dislocations, and diffuse so that they accumulate
there, forming solute atmospheres.

compression

tension

Fig. 21.13: The formation of an impurity atmosphere by migration of

oversized solutes to the region of tensile strain near an edge
dislocation.

This phenomenon is illustrated in Fig. 21.13 for the case of a solute that is over-
sized for its position in the lattice, such as carbon in iron. The lattice just beneath an edge

Page 713
Materials Science Fall, 2008

dislocation is strained in tension, and an oversized atom has lower energy if it is sited there.
By the same argument, an undersized atom is attracted to the compressive region above the
dislocation line. If the solute is mobile, it will diffuse to the dislocation line to form an
impurity atmosphere.

The most important engineering consequence of impurity atmospheres is the yield

point observed in the room-temperature stress-strain curves of high-carbon steels.
Interstitial carbon atoms have moderate mobility even at temperatures near room tempera-
ture, and migrate to form atmospheres around dislocations. These are, essentially, rows of
carbon atoms just beneath the dislocation lines. Since there is a significant binding energy
between the carbon atoms and the dislocation, the dislocation cannot move until the
resolved shear stress is sufficient to literally rip it away from its atmosphere. As a conse-
quence, carbon steels that have formed dislocation atmospheres have high yield strengths
that are very well defined. Plastic strain initiates at a sharp yield point in the stress-strain
curve, as illustrated in Fig. 21.14.

Immediately following yielding a material that exhibits a yield point experiences

yield point elongation, a plastic elongation at a lower value of the stress. The yield point
elongation is due the fact that, once the dislocations are ripped free of their pinning atmo-
spheres, they are free to move at significantly lower stress until work hardening mecha-
nisms restore the strength. If the stress is controlled, and increased until the sample yields,
the yield point elongation may occur rapidly and dramatically, and appear in the form of
discrete bands of deformation across the sample.

yield
point
ßy
}

ß yield point
elongation

Fig. 21.14: The stress strain curve of a material that exhibits a yield point.

If a material is stressed beyond its yield point, unloaded, and re-loaded immediately
afterwards, the yield point disappears, since it takes some time for the carbon atmospheres
to reestablish themselves by diffusion. If the sample is heated slightly, or simply left at
room temperature for a time long enough for diffusion to occur, the yield point reappears.

21.3.5 Dislocation hardening

Dislocations interact with one another. Since they are crystallographic as well as
elastic defects, these interactions can be rather complicated, and a detailed discussion of
them will not be given here. The two simplest examples are illustrated in Fig. 21.15. If

Page 714
Materials Science Fall, 2008

two edge dislocations that have the same sign (same orientation of the extra half-plane) ap-
proach one another in the same slip plane, their elastic strain fields repel and drive them
apart. If two dislocations of opposite sign approach one another in the same slip plane,
their elastic fields attract, and they annihilate by joining their extra half-planes.

compression compression

tension tension

Fig. 21.15: Interaction of edge dislocations in the same plane. Like dislo-
cations repel, unlike dislocations attract and annihilate.

From the perspective of strength, however, the most important dislocation interac-
tions are the interactions between a gliding dislocation and the other dislocations that cut
through its glide plane (Fig. 21.16). The dislocations that intersect the plane are called
forest dislocations, and they provide obstacles to the motion of the gliding dislocation that,
to a reasonable approximation, can be treated as point obstacles in the glide plane.

Fig. 21.16: A mobile dislocation that is resisted by forest dislocations in

its glide plane.

The dislocation-dislocation interaction is much stronger than the dislocation-solute

interaction; the forest dislocations act as point barriers that have strengths (∫c) that typically
lie in the range 0.1-0.3. If the dislocations are randomly oriented, their density, n, the
number of dislocation that intersect a unit area of the glide plane, is one-half of the volu-
metric density of dislocations, ®, which is called the dislocation density, and is defined as
the total length of dislocation line per unit volume. The yield strength of a material in-
creases with its dislocation density according to the relation

ß = ß0 + åGb ® 21.22

Page 715
Materials Science Fall, 2008

Eq. 21.22 is in qualitative agreement with the obstacle model, eq. 21.20, and is reasonably
well obeyed by structural metals and alloys.

There are three common methods for controlling the dislocation density in structural
materials: heat treatment, mechanical deformation and martensitic phase transformations.

1. Heat treatment. A material is annealed at elevated temperature to remove dislo-

cations and lower its strength. Dislocations are non-equilibrium defects, and heat treatment
decreases their density by either of two mechanisms. The first is recovery. If a material
containing a high density of dislocations is annealed at a temperature high enough to permit
dislocation climb, dislocations migrate and interact, both with one another and with free
surfaces. Some of the dislocations are annihilated, others are gathered into stable, planar
configurations, such as low-angle grain boundaries (called subgrain boundaries). The net
effect is to leave the bulk of the volume relatively free of dislocations. The second mecha-
nism is recrystallization. If the dislocation density is high enough, and the material is
heated to a temperature above its recrystallization temperature, then, as discussed in Chap.
11, new, defect-free grains nucleate and grow at the expense of the old, producing a
microstructure that is relatively free of dislocations.

While heat treatments can decrease the dislocation density, they do not eliminate it
entirely. A typical structural alloy that has been recrystallized and annealed has a dis-
location density of 107 - 108/cm2.

2. Mechanical deformation. When a material is plastically deformed, dislocations

slip and interact strongly with one another. Their interactions produce a significant and
monotonic increase in the total dislocation line length. Hence the dislocation density, ®,
increases with strain, and the material work hardens according to eq. 21.22.

When the mechanical deformation is done at high temperature, as it is during the hot
deformation that is used to roll metal ingots into plates or sheets, work hardening is coun-
terbalanced by recrystallization and recovery. To achieve a high residual dislocation density
it is necessary to deform at relatively low temperature. Metal products that are strengthened
in this way are said to be cold-worked or cold-rolled. A severely cold-worked metal has a
dislocation density of 1010-1011, producing a dislocation hardening that can be two orders
of magnitude greater than that in the annealed condition.

3. Transformation strengthening. Transformation strengthening is possible in ma-

terials that undergo martensitic transformations on cooling. As discussed in Chap. 11, the
martensitic transformation changes the structure by shearing the parent lattice (FCC in the
case of structural steel) into the product (BCC in steel). This mechanical shear produces a
highly defective microstructure, and martensitic steels have very high strength in the as-
quenched condition. However, freshly created martensite is often very brittle. The reason
is that the martensitic shear is often accommodated crystallographically, by stacking faults
and lattice twins, producing high internal stresses that promote fracture. This is a particular
problem in high-carbon steels, since the trapped carbon produces a slightly tetragonal
structure that makes it difficult for adjacent martensitic grains to fit together without strain.

Page 716
Materials Science Fall, 2008

If the steel is properly alloyed, for example, by increasing the Ni content and de-
creasing the carbon, the mechanism of the martensitic transformation can be changed
slightly so that the transformation strains are accommodated by dislocations (the product is
called dislocated martensite). Martensitic steels of this kind combine very high strength
with reasonably good toughness is the as-quenched condition. (In ancient times the best
source of good, high-Ni iron was meteoritic material, which is often rich in Ni. The supe-
rior properties of steel that incorporated meteorite iron lends some credence to the legends
that say that the storied blades of antiquity were handed down from heaven. If the movie is
to be believed (fat chance) there was meteorite iron in the original Bowie knife.)

21.3.6 Precipitation hardening

The final type of hardening obstacle is a small precipitate in the interior of the grain.
As described in Chap. 11, such precipitates are normally introduced by aging a slightly
supersaturated material at relatively low temperature, so the precipitates nucleate primarily
in the grain interiors. The volume fraction of the precipitates is determined by the phase di-
agram, and is, hence, fixed by the composition and temperature. The size of the precipitates
then depends on the aging time. The precipitates form as very small particles, and coarsen
with time as the larger particles consume the smaller ones to decrease the total interface
area.

The yield strength of a precipitation-hardened material varies with the aging time
and temperature as illustrated in Fig. 21.17. The strength increases to a maximal value, the
peak hardness, then decreases on further aging, or overaging. For a given composition, the
strength increases more slowly, but rises to a higher value as the aging temperature is
lowered. Lowering the aging temperature increases the peak hardness because it increases
the volume fraction of precipitate phase (that is, it increases the obstacle density, n, in eq.
20.20). The time required to reach peak hardness increases because diffusion is slower,
which inhibits the particle coarsening that leads to peak strength.

lower
temperature
ßy

time

Fig. 21.17: The variation of yield strength with aging time for a precipita-
tion-hardened material.

The reason that there is a peak in the hardness is a bit subtle. In the most common
case, it is due to the fact that the hardening precipitates increase in strength with their size.
When the precipitates are very small, they are generally coherent, and dislocations can cut

Page 717
Materials Science Fall, 2008

through them. They behave as cuttable obstacles with strengths (∫c) of the order of 0.5.
As they grow, their strength increases, and, since their volume fraction remains constant,
their mean separation increases as well. However, in the early stages of coarsening the
increase in strength (∫c) generally outweighs the increase in separation (Ls = n-1/2), so the
material hardens.

The peak in the yield strength comes from the fact that there is an upper limit to the
strength of the obstacles, but no upper limit to their spacing. The maximum force that the
arms of a dislocation can exert on an obstacle (Fig. 21.11) is 2T, the force when the in-
cluded angle, ¥, is zero. When the force required to cut through the obstacle approaches
this value, the dislocation does not cut through the obstacle, but wraps around it, as illus-
trated in Fig. 21.18.

If the obstacle cannot be cut by the dislocation (an impenetrable obstacle) the dislo-
cation line is held back until the stress becomes so high that the diameter (2R) of the dislo-
cation loop is comparable to the obstacle spacing. When this is true the arms of the dislo-
cation extend well beyond the obstacles and approach one another (Fig. 21.18(a)). The two
arms of the bowed-out dislocation have opposite sign (note that the Burgers vector, b,
points out of the loops to the immediate left of the obstacles shown in Fig, 21.18(a), but
into the loops to their immediate right). Hence the arms attract one another. When they
meet on the far side of the obstacle, they annihilate to create a continuous dislocation that
lies beyond the obstacle row and leave behind loops of dislocation that surround each of the
obstacles that was by-passed, as illustrated in Fig. 21.18(b). A precipitate particle of maxi-
mum possible strength has a breaking angle, ¥c = 0, or a strength, ∫ c = 1. In fact, because
the arms of the dislocation attract one another, they help to pull the dislocation past, and an
impenetrable obstacle has an effective strength nearer ∫c « 0.8.

†b †b †b †b

b b

(a) (b)

Fig. 21.18: The interaction of a dislocation with impenetrable obstacles.

(a) The arms of the dislocation wrap around the obstacle and
attract one another. (b) The arms intersect and annihilate, pro-
ducing a propagating dislocation, and leaving dislocation loops
around the obstacles.

Once a precipitate has grown so large that it is impenetrable, its strength is fixed,
and is not changed by further growth. But the precipitates that have reached impenetrable
size continue to grow; expressed in the terms used in eq. 21.18, Ls increases while ∫c
remains the same. The result is that †c, and, hence, the yield strength, ßy, decreases with

Page 718
Materials Science Fall, 2008

further aging. The peak strength is reached when the largest obstacles just become impene-
trable. The material overages if any further coarsening is allowed.

Some materials form very hard precipitates, which are uncuttable even when their
size is very small. However, even in this case the hardness tends to increase to a maximum
as the alloy is aged. The reason is that such precipitates ordinarily have such high
interfacial tension with the alloy that nucleation is difficult and their volume fraction is
sharply restricted when their size is small. The increase in hardening with aging time is due
to an increase in the volume fraction of the precipitate phase.

21.4 THE INFLUENCE OF TEMPERATURE AND STRAIN RATE

21.4.1 The variation of yield strength with temperature

If the yield strength of a typical material is plotted as a function of its homologous

temperature (T/Tm, where Tm is the melting temperature), the result is a curve that resem-
bles that in Fig. 21.19. The yield strength is relatively insensitive to the temperature over a
range of intermediate values of the homologous temperature, but increases dramatically as
the temperature is lowered toward zero, and decreases dramatically as it is raised to near the
melting point. Ambient temperature is in the intermediate temperature regime for Al and its
alloys, is slightly into the low-temperature regime for typical structural steels, and is in the
high-temperature regime for low-melting metals like Pb.
thermal
activation

degradation
athermal
barriers

thermal

ßy

T/T m
Fig. 21.19: Typical variation of yield strength with homologous tempera-
ture.

Low-temperature strength

The rapid strength increase at low temperature is due to the strong temperature de-
pendence of hardening by obstacles that can be cut or passed by thermal activation. These
particularly include the weak, short-range obstacles created by isolated solute atoms. As
the temperature increases, the increased amplitude of atomic thermal vibrations produces an
effective vibration of the dislocation line, which permits it to cut through obstacles that
could not be bypassed by the stress alone.

Page 719
Materials Science Fall, 2008

As a rough rule of thumb, the low-temperature strengthening mechanisms are inef-

fective at temperatures above 1/4 to 1/3 of the melting point. The increase in strength at low
temperature is much more pronounced in the typical BCC metal than in the typical FCC
metal. A principal reason is the greater strength of interstitial solutes in BCC, which be-
come increasingly effective as thermal activation becomes more difficult and the full po-
tency of the interstitial solutes is revealed. Typical carbon steels have very high strengths at
cryogenic temperatures. Austenitic (FCC) stainless steels that are intended for service at
very low temperature are also strengthened by adding interstitials, with nitrogen preferred
to carbon because of its higher solubility in FCC alloys. While the low-temperature
strength increment per unit solute content is lower for FCC, the high nitrogen solubility has
the consequence that nitrided austenitic steels can achieve very high strengths at low tem-
perature.

Intermediate temperature strength

At intermediate temperature the yield strength is a relatively weak function of tem-

perature. The strength in this region tends to be controlled by dislocation-dislocation and
dislocation-precipitate interactions. These present obstacles that have relatively large effec-
tive sizes; they spread over at least several atom spacings in the slip plane, and are not eas-
ily passed by thermal vibrations of the dislocation. Room temperature is well within the
intermediate temperature regime for Al alloys, but is slightly below for typical structural
steels.

High temperature strength

At temperature above 0.5Tm the yield strength begins to drop dramatically. The
principal reason is the increasing rate of solid state diffusion, which affects both the dislo-
cations and the microstructural barriers. The microstructural effect is most important.
Dislocation configurations recover by climb and recombination, precipitates coarsen and
overage, and grains grow; virtually all of the available microstructural barriers become inef-
fective. At the same time, the high diffusivity makes it possible for dislocations to climb at
an appreciable rate, so the obstacles that remain are more easily passed.

The only practical mechanism for hardening metals that are to be used at a signifi-
cant fraction of their melting points is precipitation hardening by precipitates that are ther-
modynamically very stable, so that they do not coarsen at a rapid rate, and relatively large,
so that they are not easily passed by dislocation climb. The best high-temperature alloys are
members of a class of materials known as superalloys. These are based on Fe, or, even
better, on Ni, and are strengthened by large, blocky precipitates of Ni3X ordered inter-
metallics, where X may be Ti, Al, Nb, Ta or Mo. Such alloys are expensive, and they are
difficult to process, but they are the workhorse alloys of the high-temperature turbine in-
dustry.

Even better high-temperature strengths are exhibited by certain oxides (MgO and
Al2O3), carbides (SiC), and nitrides (Si3N4), which have high melting points and high
Peierls-Nabarro stresses to inhibit plastic flow. However, even at elevated temperature

Page 720
Materials Science Fall, 2008

these materials are relatively brittle, and are only rarely used in safety-critical structures
such as turbine engines.

In addition to the degradation of the tensile yield strength, high temperature also
makes the material liable to deform by creep if it is held under load. Creep deformation was
discussed briefly in Chap. 19.

In assessing the strengths of the common metals and alloys, it is useful to keep in
mind the principle that the homologous temperature (T/Tm), rather than the actual tempera-
ture, governs the strength. Materials with low melting points, like Pb and eutectic Pb-Sn
alloys, are very soft at room temperature. The reason is not so much their low inherent
strength (though they do have relatively low shear moduli) but their high homologous tem-
perature, which invalidates conventional deformation processes. Similarly, materials like
pure Mo have relatively high room temperature strength, largely because of their high
melting points.

21.4.2 The influence of strain rate on strength

The yield strength of a ductile metal or alloy always increases with the rate at which
the sample is strained. However, the magnitude of the rate effect varies dramatically with
the material and with the temperature at which the test is done. The reason is that the com-
mon rate effects have the same source as the temperature effects. If deformation is ther-
mally activated it becomes easier to accomplish as the rate of deformation is lower; more
time is provided for thermal activation. To a good first approximation, increasing the strain
rate is equivalent to decreasing the temperature.

It follows that materials exhibit pronounced strain rate effects when they are tested
in either the high-temperature or the low-temperature regimes, and are relatively insensitive
to strain rate when they are tested at intermediate temperature. As might be expected, the
strengths of structural steels are very sensitive to strain rate, while those of Al and its alloys
are almost constant, even when the strain rate is changed by several orders of magnitude.

21.5 WORK HARDENING

When a material deforms plastically it also hardens. The yield strength increases
with the strain, a phenomenon known as work hardening. The basic mechanism of work
hardening was described above, when we considered how dislocations harden materials.
As strain builds up inside the material, dislocations slip, intersect and interact with one
another, as illustrated schematically in Fig. 21.20. These interactions cause an increase in
the dislocation density (total dislocation length per unit volume) that is monotonic in the
strain.

Page 721
Materials Science Fall, 2008

Fig. 21.20: Dense dislocation network in a severely strained material.

The yield strength increases with the dislocation density according to eq. 21.22:

ß = ß0 + åGb ® 21.22

The associated work hardening rate, œ, is, then

dß åGb d®
œ = d‰ = - 21.23
2 ®  d‰ 

and depends on the rate at which dislocations multiply with the strain.

A major part of the fundamental research that has been done on work hardening
concerns the behavior that is peculiar to single crystals and is, therefore, of limited
engineering interest. There are two common representations of the work-hardening
behavior of the polycrystalline metals and alloys that are of interest in engineering. The
first comes from the parabolic stress-strain relation that is often assumed in continuum
mechanics:

ß = k‰n 21.24

where k is a constant, giving

œ = nk‰n-1 21.25

Equation 21.25 is analytically simple, but does not have any simple association with the
mechanisms of dislocation plasticity, and is not, in fact, very accurate.

A more useful relation comes from the Kocks-Mecking model of work hardening.
Assume that the material has undergone general yielding so that its dislocation distribution
is well developed and reasonably homogeneous on the macro scale. Then the instantaneous
value of the yield strength is given by eq. 21.22, and the work hardening rate is given by
eq. 21.23. To complete the model we require an expression for d®/d‰, the increase in
dislocation density with strain. A detailed analysis suggests that d®/d‰ is the sum of two
terms, one reflecting the rate of dislocation multiplication (hardening), and the second the

Page 722
Materials Science Fall, 2008

rate of dislocation annihilation (recovery). An analysis of the relation between ® and ‰

suggests that the multiplication rate is related to the frequency of intersection with forest
dislocations, and, hence, proportional to the spacing between forest dislocations in the glide
plane. It follows that

d®
  = C1 ® 21.26
 d‰ +

where C1 is a constant. The rate of recovery is the rate at which dislocations annihilate,
which happens predominantly through the intersection of dislocations that have a common
plane. An analysis of this process gives

d®
  = - C2® 21.27
 d‰ -

Summing 21.26 and 21.27, and substituting the result into 21.23 gives the Kocks-Mecking
relation for the work hardening rate

œ = œ0 - Cwß 21.28

where œ0 and Cw are constants.

ß
Fig. 21.21: The variation of the work hardening rate with the stress for
aluminum alloys.

The linear Kocks-Mecking relation is reasonably well obeyed by many engineering

materials, and is particularly applicable to FCC metals and alloys. However, it does not
apply to local yielding, nor to the transient behavior that may appear as local yielding de-
velops into general plasticity. The typical variation of the work hardening rate with the
stress for polygranular aluminum alloys is shown in Fig. 21.21. It contains three distin-
guishable regimes.

The first, to the left-hand side of the figure, is associated with local yielding. The
work hardening rate decreases rapidly as more and more local regions within the crystal
yield and contribute to plastic deformation. The second regime is a transient stage as re-

Page 723
Materials Science Fall, 2008

gions that have yielded locally link together and develop the microstructure an deformation
pattern that produces general yielding. The third regime is the linear work hardening be-
havior that is observed after general yielding.

21.6 PLASTIC INSTABILITY, NECKING AND FAILURE

The final subject we shall discuss here is plastic instability, which defines the ulti-
mate tensile strength of a tensile specimen, and triggers the necking deformation and failure
if the specimen is stressed or strained beyond its ultimate tensile strength.

21.6.1 The Considere criterion

To understand plastic instability, we must first understand how deformation actually

happens in a specimen, such as a tensile bar, that is stretched in tension. Plastic de-
formation is never strictly homogeneous. Just as a chain breaks at its weakest link, each
increment of strain happens at that volume within the specimen that is instantaneously
weakest. The macroscopic deformation appears uniform because each small element that is
strained is work-hardened. Because it is strengthened by work-hardening, the strained el-
ement ceases to be the weakest in the body, and the next increment of deformation happens
at some other place. In this way local deformation works its way back and forth through
the body. Necking happens when this process breaks down, and the strain remains con-
centrated at a particular site.

ß ß

Fig. 21.22: Incipient neck in a tensile bar.

To find the condition that produces plastic instability in a round tensile bar, consider
the tensile specimen shown in Fig. 21.22, and let the instantaneous strain thin a narrow
region, as illustrated, with considerable exaggeration, in the figure.

The cross-sectional area of the strained region of the specimen has decreased by
dA, so the stress it must bear under the applied load, P, is increased by the amount

P -dA 
dß = d(P/A) = - 2 dA = ß A  = ßd‰ 21.29
A

But the yield strength of the material has also increased, and is

dß
dßy =  d‰  d‰ 21.30

Page 724
Materials Science Fall, 2008

If dßy > dß, the material has hardened more than is necessary to support the increased load,
and the deformation is stable. The neck will not grow, and the next increment of de-
formation happens at some other point along the bar. If dßy < dß, on the other hand, the
incipient neck cannot support its increased load and the subsequent strain is localized there,
causing necking and failure. Since dß/d‰ is a decreasing function of strain (at least for the
common structural materials), the cross-over from stable to unstable deformation happens
when

dß
d‰ = ß 21.31

Equation 21.31 is known as the Considere criterion. It states that plastic instability occurs
when the work hardening rate falls below the true stress.

œ = dß/d‰
plastic
: ds/de = 0
instability
su
ß plastic
instability s
œ=ß

‰ e
(a) (b)

Fig. 21.23: True and engineering stress-strain curves for the same mate-
rial, showing the points at which plastic instability occurs.

The slope of the engineering stress-strain curve at plastic instability is, from eq.
19.24

ds d(ße-‰ ) e-‰ dß - ße-‰ d‰ dß

-2‰ 

 =0
= = = e  - ß 21.32
de d(e‰ -1) e‰ d‰ d‰

Hence plastic instability occurs at the maximum of the engineering stress-strain curve and
defines the ultimate tensile stress, su.

These relations are illustrated in Fig. 21.23. Fig. 21.23(a) shows an example true
stress-strain curve with the work hardening rate also plotted. Plastic instability intrudes
when these two curves intersect. Note that there is nothing in the behavior of the true
stress-strain curve itself to indicate that this has happened. Fig. 21.23(b) shows the engi-
neering stress-strain curve for the same material. In this case, the instability point is obvi-
ous. Plastic instability intrudes at the peak of the stress-strain curve, which defines su.

Page 725
Materials Science Fall, 2008

If the stress-strain curve is parabolic, as in eq. 21.24, then the uniform elongation is
easily found from equation 21.25, and is

‰u = n 21.33

More generally, the stress at which instability occurs can be found graphically from a plot
of the work hardening rate, œ, as a function of ß (Fig. 21.21). It is the intersection of the
work hardening curve with the line œ = ß, as illustrated in Fig. 21.24.

Finally, note that the Considere criterion matches a material property, the work
hardening coefficient, œ, with a geometric criterion, the increase in applied stress with local
strain. The latter is sensitive to the geometry of the specimen, and the simple Considere
criterion, eq. 21.31 only holds when the specimen is a cylindrical bar in uniaxial tension. If
the geometry of the specimen changes, the locus of plastic instability changes as well. For
example, a thin sheet specimen in uniaxial tension experiences local necking when

dß ß
d‰ = 2 21.34

Other criteria apply to other geometries.

plastic
instability
œ
œ=ß

Fig. 21.24: Location of the instability stress on a plot of œ vs. ß.

21.6.2 Tensile elongation

In addition to determining the ultimate tensile strength, plastic instability determines

the limit of the uniform elongation of a specimen. This is an important parameter in the
forming of materials, since a part that is formed to strains beyond the instability limit
develops non-uniformities in its cross-section, and is liable to fracture. The instabilities
that are of interest in forming are, of course, those that pertain to the material geometry
(usually a sheet) and the load configuration (usually biaxial) that is used. Specific forming
tests have been designed to define forming limit diagrams that govern particular forming
operations. Nonetheless, we can gain significant insight into the behavior of a material in
forming by understanding the factors that limit tensile elongation.

Page 726
Materials Science Fall, 2008

Fig. 21.23(a) illustrates how the work hardening behavior limits uniform elonga-
tion. The ultimate criterion is the intersection of the work hardening and stress-strain
curves, which suggests that the work hardening rate should be made as high as possible to
forestall this intersection. But this is deceptive. In fact, the tensile elongation is affected by
the yield strength, and by the work hardening behavior over the whole range of plastic
strain.

Fig. 21.26: Illustration of the deleterious influence of a high early work

hardening rate on the tensile elongation.

The simplest way to increase the elongation of a typical material is to decrease its
yield strength, particularly that part of the yield strength that is due to dislocations left over
from prior deformation. If a material can be softened without significantly changing its
work hardening behavior at higher strains, then there is an added, "free" elongation that oc-
curs before the stress reaches values that approach the work hardening rate. For this rea-
son, alloys that are intended for forming are usually supplied in a well-annealed, soft con-
dition.

Fig. 21.27: An increase in the work hardening rate at high strains increases
the total elongation.

Second, the elongation can be increased by decreasing the extent of early yielding,
where the work hardening rate is relatively high. As illustrated in Fig. 21.26, a high rate of

Page 727
Materials Science Fall, 2008

work hardening early in the deformation process raises the stress, and decreases the strain
at which the stress becomes equal to the work hardening rate. Since early yielding is asso-
ciated with heterogeneities in the microstructure, materials that are microstructurally homo-
geneous tend to have better tensile elongations.

Finally, one can increase the tensile elongation by increasing the work hardening
rate in the later stages of deformation, as illustrated in Fig. 21.27. Several metallurgical
tricks have been developed over the years to accomplish this. One of the more clever is in-
corporated in the so-called dual phase steels. These are alloy steels that are processed so
that their microstructures contain a fine mixture of relatively soft BCC ferrite, produced by
a nucleation-and-growth transformation from the FCC austenite, and relatively hard
martensite, produced by quenching a structure that is partly ferrite and partly residual
austenite. The ferrite yields at low stress, but its deformation does not significantly affect
the much harder martensite until work hardening has raised its yield strength to a value
comparable to that of the martensite. The martensite then yields fairly late in the deforma-
tion process. Its subsequent work hardening adds a significant increment to the large-strain
work hardening rate of the alloy so that plastic stability is maintained to high strains.

Page 728
Materials Science Fall, 2008

Chapter 22: Fracture

Stone-cutters fighting time with marble, you foredefeated

Challengers of oblivion
Eat cynical earnings, knowing rock splits, records fall down,
The square-limbed Roman letters
Scale in the thaws, wear in the rain. The poet as well
Builds his monument mockingly;
For man will be blotted out, the blithe earth die, the brave sun
Die blind and blacken to the heart:
Yet stones have stood for a thousand years, and pained thoughts found
The honey of peace in old poems.
- Robinson Jeffers, To the Stone-Cutters

22.1 INTRODUCTION

22.1.1 Engineering importance

The most threatening form of mechanical behavior is fracture, in which an intact

structure separates into two pieces. Sudden, unexpected fracture is the culprit in many of
the more famous engineering disasters, including such tragedies as the Liberty ship
crackups, the Ohio River and Kansas City Hyatt bridge collapses, the Tokyo and Sioux
City commercial airplane crashes, the in-flight destruction of the space shuttle Challenger,
the Great Boston Molasses Flood of nineteen-ought-whatever-it-was, the crack in the
Liberty Bell, and many others. In the background of these widely publicized disasters is a
steady stream of crashed vehicles, split pipes, exploded tanks, collapsed structures, broken
toys and malfunctioning electronic devices that failed because the materials inside them
broke.

While unexpected, catastrophic structural failures have plagued engineers for just
about as long as structural materials have been used, an accurate understanding of crack
propagation to failure has only come about in the last few decades. Fracture mechanics, the
now large subfield of engineering that specializes in the study of fracture, was not even
taught when I was a graduate student in Materials Science in the 1960's. The discipline
was born in the late 1950's, largely as a product of research initiated and funded by the
United States Navy.

The U.S. Navy had a strong, practical interest in developing a theory of fracture.
During the Second World War, the standard method of construction of sea-going hulls had
changed from riveted steel plates to welded ones. Welding was faster, cheaper and, in
theory, produced a more reliable product. But sailors do not float on theory, and the

Page 729
Materials Science Fall, 2008

welded tankers and transports that were turned out by the hundreds in yards like Henry J.
Kaiser's shipyard in Richmond, California, had a most annoying habit of splitting in two,
right through the center of the hull, with no prior warning. Since a vessel with half a hull is
not seaworthy, the Navy was anxious to know what had gone wrong with these "liberty
ships" and how the problem might be prevented in the future. The result was the field of
Fracture Mechanics.

The introduction to the subject that is given below here will necessarily be brief and
superficial. It is, nonetheless, sufficient to understand a good part of what happens in
common material failures.

22.1.2 Why things fall apart

Before we turn to the subject of fracture mechanics it is important to note that

structural materials fracture for three fundamentally different reasons: buckling, necking,
and cracking. The first two of these have virtually nothing to do with the material's resis-
tance to fracture itself.

The phenomenon of necking to failure was discussed in the previous chapter. If a

piece of material is loaded to a stress beyond its ultimate tensile strength, its plastic defor-
mation becomes unstable and concentrates, thinning the cross-section of the piece and
raising the true stress on that cross-section until it finally breaks. So long as a material has
sufficient toughness (resistance to fracture) that it does not break before plastic instability
sets in, its toughness does not matter very much. Once it begins to neck, it will inevitably
fracture.

Buckling is a similar phenomenon that was mentioned briefly in Chapter 20.

Buckling is an elastic instability. It is a common failure mode in structural columns that are
loaded in compression, and in sheets, such as automobile fenders, that are crushed or
dented. The simplest case is the column. If a column is loaded in compression it bows out
into an arc, whose deflection increases with the load. Eventually, the load reaches a value
where the bowing angle is so great that increasing the load actually decreases the elastic re-
sistance from the column, and it collapses. When a column buckles it forms a hinge at
some point along its length where deformation concentrates. Unless the material is very
ductile, it fractures at this hinge at some point during the buckling process. The toughness
of the material is almost irrelevant. The elastic instability has made fracture inevitable,
whatever the toughness may be.

The third mechanism of fracture, unstable crack propagation, is the subject of this
chapter. When a material that contains a crack or geometric flaw is loaded, the applied
stress concentrates at the tip of the flaw. When the local value of this stress reaches a
critical level the material separates, and the flaw propagates as an unstable crack that splits
the material apart. The material property that governs the stress at which the flaw
propagates is its fracture toughness. But a knowledge of the fracture toughness is not
sufficient to determine the load at which the material fractures. It is also necessary to know

Page 730
Materials Science Fall, 2008

the size and geometry of the flaws it contains, since these determine the stress concentration
factors that magnify the applied stresses and lead to fracture.

While buckling and necking are threatening failure modes, they are driven by the
macroscopic stress and are, hence, relatively easy to incorporate into engineering design.
Cracking is a much messier problem. Unless the material has very high fracture toughness,
the fracture stress is sensitive to the presence of flaws that are small, and difficult to detect,
and these may propagate to catastrophic failure at nominal stresses that are well below the
yield strength of the material. Compounding the problem, a material that is initially sound
may develop critical flaws during service, though fatigue or stress corrosion cracking, or
may lose its toughness during service via any one of several embrittling mechanisms.
Catastrophic crack propagation can strike without warning in structures that meet all normal
macroscopic design criteria, and has on a great many occasions.

22.2 FRACTURE MECHANICS

22.2.1 Crack propagation from a pre-existing flaw

The stress concentration at a flaw

Let a material specimen contain an internal crack of length, a, and let it support a
load that would produce a nominal tensile stress, ßa, as illustrated in Fig. 22.1. If the tip of
the crack is viewed under a sufficient magnification, it will appear rounded, as shown in
Fig. 22.2. Let its effective radius be ®.

ßa

ßT

ß
ßa
r
a

ßa

Fig. 22.1: A section from a material that contains a crack of length, a, and
is subjected to an applied stress, ßa. The stress concentration
at the crack tip results in the local stress, ßT.

Page 731
Materials Science Fall, 2008

The stress at the tip of the crack is magnified by the stress concentration factor of
the crack. If we assume (or approximate) elastic behavior up to the crack tip, the stress at
the crack tip, ßT, is given by the relation

a
ßT = Qßa ® 22.1

where Q is a geometric factor that depends on the shape of the crack and the geometry of
the specimen. A considerable part of the work of analytic fracture mechanics is concerned
with the evaluation of the geometric factor, Q, and the specific consequences of local plastic
deformation near the crack tip, but we do not need this sophistication to understand the
fundamentals of fracture. Moreover, we need not require that the flaw be a narrow crack,
as shown in the figure. Blunt cracks, voids and inclusions in the material also produce
stress concentrations that can be represented by equivalent cracks.

Fig. 22.2: Magnified view of the tip of the crack in Fig. 22.1, showing
the radius, ®.

Equation 22.1 has three important qualitative features. First, the crack tip stress, ßT
, increases linearly with the applied stress, ßa. Second, it increases with the root of the
crack length; the bigger the flaw, the higher the crack tip stress, and the greater the threat to
the integrity of the sample. Third, it increases with the inverse root of the crack tip radius,
®. This is a critical factor in determining the relative crack susceptibility of materials. If a
material is soft and yields easily, then plastic deformation near the crack tip causes the crack
to blunt, increasing the effective radius, ®, and decreasing the crack tip stress. In this sense,
flaws in ductile materials are self-healing. If the material is very strong, on the other hand,
the cracks that appear in it remain sharp, and produce intense magnifications of the applied
stress. This is the basic reason that strong materials, like silica glass, oxide ceramics,
carbides and nitrides, and diamond are very brittle. Incipient cracks in such materials
remain almost atomicly sharp.

The critical fracture stress

The simplest fracture criterion is the stress criterion suggested by Orowan. He pro-
posed that the material has a breaking stress, ßB , and spontaneously fractures if the local
value of the tensile stress exceeds this value. In this (somewhat oversimplified) model,
which we shall adopt here, the crack in Fig. 22.1 propagates when the crack tip stress

ßT = ßB 22.2

Since ßT increases with the crack length, a crack that has started to propagate will ordinarily
continue to do so. Eq. 22.2 is the criterion for unstable crack propagation to failure.

Page 732
Materials Science Fall, 2008

Eq. 22.2 refers to the local condition at the crack tip. The stress that is meaningful
to the engineer is the applied stress, ßa. Using eqs. 22.1 and 22.2, the sample breaks when
the applied stress, ßa, reaches the fracture stress, ßc, where

ßB ®
ßc = Q-1 22.3
a

If a material contains many flaws, as all real materials do, then the fracture stress is deter-
mined by the minimum value of the right-hand side of eq. 22.3. This defines the "weakest
link" in the material. As we anticipated above, the tensile stress at which a material breaks,
ßc, is not determined by its inherent breaking stress, ßB , but by a combination of ßB and
the size and geometry of the worst flaw it contains.

22.2.2 The fracture toughness

Fracture toughness

There are four independent parameters on the right-hand side of eq. 22.3. Two of
these, the geometric factor, Q, and the crack length, a, can be controlled or measured, and,
in particular, can be controlled in fracture experiments by fixing the geometry of the test
specimen.

The other two parameters, the inherent fracture stress, ßB , and the crack tip radius,
®, are extremely difficult to measure or predict. The reason is that both are influenced by
the details of processes that occur at the crack tip as it approaches the critical condition. An
inherent fracture stress, ßf, can be defined, for example, as the true stress at the final frac-
ture of a tensile specimen. But this stress is sensitive to the geometry of the specimen, and
averages behavior over a volume of material that is very large compared to that in which
fracture initiates at the tip of a crack. The effective crack tip radius, ®, is influenced by
plastic deformation or other relaxation processes that occur as the crack tip evolves into the
unstable configuration, and is virtually impossible to measure.

While ßB and ® cannot be measured, we can hope that they will have reproducible
values for near-critical sharp-tipped cracks in a given material. On this assumption, we de-
fine their product as the fracture toughness, Kc, where, from equation 22.3

ßc
Kc = Q 22.4
a

The fracture toughness, Kc, has the advantage that is readily measured in tests that employ
samples with cracks of known length and geometry. It has the second advantage that it is
an experimental number that is model-independent. There are theoretical alternatives to the
Orowan criterion for crack propagation that we used above. In fact, the energy-based
Griffith criterion, or modifications of it, is much more widely used. These alternative ap-

Page 733
Materials Science Fall, 2008

proaches lead to different expressions in the denominator of eq. 22.3. However, the dif-
ferent fracture criteria do not change the definition of the fracture toughness as given in eq.
22.4. (They are also unimportant in the qualitative discussion of fracture behavior that is
given below.)

Plane strain fracture toughness

There is, however, a problem with the fracture toughness, Kc, that is defined in eq.
22.4. It is not an unique material property. In particular, if Kc is measured for a number of
specimens that have different thicknesses, T, the value of Kc varies with T as shown in Fig.
22.3. After an initial transient for very thin samples, the toughness decreases with T to
asymptote at a constant value, called the plane strain fracture toughness, in the thick-sample
limit.

Kc

KIc
plane
strain

Fig. 22.3: The variation of fracture toughness (Kc) with the thickness (T)
of the test specimen. Kc asymptotes to the plane strain fracture
toughness, KIc.

The reason for this behavior can be explained from the basic relation of the Orowan
model:

Kc = ßB ® 22.5

While the inherent fracture stress, ßB , does not vary with T, the crack tip radius, ®, does.
The reason is that value of ® reflects the extent of crack tip blunting by plastic deformation
as the crack tip is loaded up to the critical stress. This deformation occurs in two
perpendicular planes (Fig. 22.4): deformation in the plane of the specimen tends to open the
crack, while deformation in the plane perpendicular to the specimen tends to thin the
specimen along the line of the crack tip (Fig. 22.4(b)).

The feature that changes with the thickness of the specimen is the thinning in the
thickness direction. The reason is illustrated in Fig. 22.5. Plastic deformation is driven by
shear, and concentrates along planes at 45º to the tensile axis, where shear is greatest. The
high stress at the crack tip is confined to the immediate vicinity of the crack tip, and extends
for only a short vertical distance along the sample, as shown schematically in the figure.
When the sample is thin, like the sample on the left in Fig. 22.5, it is possible to extend

Page 734
Materials Science Fall, 2008

shear along 45º lines across the whole width of the specimen while remaining within the
region of high stress. Hence the specimen thins significantly in the through-thickness
direction. This deformation contributes to crack blunting, increasing ® and, hence, Kc.
When the specimen is thick, like the right-hand sample, deformation can extend only a short
distance in from the lateral surfaces of the specimen before the shear planes outreach the
region of high stress. The result is that only the edge of the specimen thins. Over most of
the crack length through the specimen, only the deformation in the sample plane contributes
to crack blunting. The crack tip radius, ®, and, hence, Kc are smaller.

ß T1

Fig. 22.4: Side and front views of a thin fracture toughness specimen
showing plastic deformation in the sample plane and thinning
in the thickness direction.

When the through-thickness deformation can be ignored, only the in-plane defor-
mation contributes to crack blunting, and the sample is said to be in plane strain. The value
of the fracture toughness in the plane strain limit is called the plane strain fracture tough-
ness, and denoted KIc. The plane strain fracture toughness is the asymptotic value for large
sample thicknesses (Fig. 22.3). It is also the minimum value of the fracture toughness.

The plane strain fracture toughness, KIc, is the "fracture toughness" that is ordinar-
ily tabulated and used. It is the best measure of toughness for two reasons. First, it is a
material property that can be used with some confidence for design purposes. Second,
since it is the minimum value of the toughness, it provides a conservative estimate of the
fracture stress. It follows from eq. 22.4 and Fig. 22.3 that

KIc
ßc ≥ Q-1 22.5
a

A structure is definitely safe with respect to crack propagation so long as the peak stress
remains below the value of the right-hand side of eq. 22.5. (To ensure safety, we should

Page 735
Materials Science Fall, 2008

also take the geometric constant, Q, from the worst-case flaw geometry.) In the following
we shall use this conservative estimate for ßc.

T1 T2 ß

Fig. 22.5: Front view of thin (T1 ) and thick (T2) fracture specimens (the
side view of both would resemble Fig. 22.4), showing the
45ºC planes of maximum shear and the regions of significant
plastic deformation. The crack tip stress varies with height (d)
as shown at left.

22.2.3 Fracture-sensitive design

Eq. 22.5 shows that the critical fracture stress depends not only on the fracture
toughness, KIc, a property of the material, but also on the crack length, a, which is the
length of the worst flaw in the material. It follows that fracture-sensitive designs must in-
corporate some method for determining what the worst flaw in the material is.

The importance of flaw detection and characterization has given rise to a whole
subfield of engineering, which is called NDE, the non-destructive examination of materials.
Experts in NDE employ a variety of techniques to detect and measure structural flaws.
These range from fairly simple methods, such as visual inspection, often with the aid of
dyes or magnetic particles that makes surface cracks more visible, to complex methods that
employ x-rays or ultrasonic waves and elaborate computations to visualize results in three
dimensions. A detailed discussion is beyond the scope of this course. However, as is the
case in every field, the more elaborate the inspection, the longer the time and the higher the
cost. Engineers try to design structures so that critical flaws will be large and relatively
easy to find.

Fail-safe design

In fail-safe design one uses a material that is so tough that crack propagation is
highly unlikely at stresses below the ultimate tensile strength. If we conservatively estimate
the toughness by the plane strain fracture toughness, KIc, then the critical crack length, ac,
the length of a crack that would propagate to failure under the stress, ß , is

Page 736
Materials Science Fall, 2008

KIc2
ac = Q-2 ß  22.6

The critical flaw size increases with the square of KIc. If the fracture toughness is suffi-
ciently large ac is large and easily detectable (for example, a significant fraction of the cross-
section of the device) even if ß were to equal the ultimate tensile stress, ßu. A material that
satisfies this condition will fail, if at all, at its ultimate tensile strength, and the danger of
unstable crack propagation can be ignored.

A common approach to fail-safe design is to ensure that the critical crack length is
so large that the sample will undergo general yielding before fracture. Since common de-
sign rules require that the maximum operating stress be well below yield, this rule ensures a
good margin of safety against fracture. Under this criterion, fail-safe design suggests the
use of materials whose characteristic length

KIc2
a* =  ß  22.7
y

is a significant fraction of the section thickness of the specimen. The characteristic length,
a*, is a material property that is often used as a "figure of merit" for the quality of a struc-
tural material.

Flaw-sensitive design

The problem with a fail-safe design is that it virtually rules out the use of high-
strength materials. Since, as we shall discuss further below, fracture toughness tends to
decrease with yield strength, materials with very high values of KIc tend to have very low
strengths. While the use of low-strength materials in engineering structures is not limited
by fracture, it is limited by yield, since a structure must ordinarily be designed to operate at
no more than a fraction of its yield strength. This is a particular problem in weight-critical
structures such as aircraft, where low strength and high safety margins translate into high
weight and poor performance. Since aircraft structures are also safety-critical, a major part
of the effort to produce viable, flaw-sensitive designs has been in the aerospace industry.

Flaw-sensitive design is based on the symbiotic use of engineering design, materi-

als selection and non-destructive inspection. If ßD is the maximum value of the design
stress (with appropriate safety factors) then the critical flaw size for unstable crack propa-
gation is

KIc2
ac ≥ aD = Q-2 ß  22.8
D

Page 737
Materials Science Fall, 2008

where the geometric factor, Q, is chosen to represent a realistic "worst case". The structure
is safe if the non-destructive testing procedures that are specified can reliably detect flaws
that are significantly smaller than aD.

The difficulty with this procedure is that one must not only consider flaws that are
present in the structure as-manufactured, but must also consider the possibility of crack
growth during service by such mechanisms as fatigue and stress corrosion cracking. If
crack growth is a threat, as it invariably is in aircraft, then the structure must be periodically
re-inspected.

Despite these difficulties, flaw-sensitive design is widely and successfully used in

advanced structures.

22.3 MICROSTRUCTURAL CONTROL OF FRACTURE TOUGHNESS

The mechanical consideration that most often governs the selection of a structural
material for service in a fracture-sensitive structure is the combination of yield strength, ßy,
and plane strain fracture toughness, KIc. Both strength and toughness are critical properties
since failure may occur through either plastic deformation or fracture. The combination is
important since strength and toughness have an inverse relation to one another; an increase
in strength at given temperature almost invariably leads to a decrease in fracture toughness
(Fig. 22.6). In the design or selection of materials for cryogenic service it is desirable to
maximize the strength-toughness combination or, at least, to achieve values that lie within a
"design box" in a strength-toughness plot that is bounded by the minimum acceptable
strength and toughness values. Since both strength and toughness vary with the tempera-
ture the only strictly meaningful design box is one that is defined at the intended service
temperature.

K IC

ßy

Fig. 22.6: Typical relation between yield strength and fracture toughness
for a ductile material.

There is no reliable quantitative theory of the strength-toughness relation of struc-

tural alloys. However, research on the mechanisms of yield and fracture combined with
specific studies of the behavior of materials at temperatures has produced a qualitative
understanding of the strength-toughness combination that is useful for materials selection,
quality control and new alloy design.

Page 738
Materials Science Fall, 2008

22.3.1 The fracture mode

At the micromechanical level the fracture of a material is either ductile, in which

case the material is torn apart after considerable local plastic deformation, or brittle, in
which case the crack propagates with very little plastic deformation. In most cases there is
a first-order correspondence between the level of toughness and the fracture mode: a
change from a ductile to a brittle fracture mode causes a substantial drop in the fracture
toughness. It follows that the fracture mode is the first concern in interpreting the strength-
toughness characteristic of a material.

Fig. 22.7: Brittle crack propagation by intergranular fracture.

Since the simplest fracture modes are the brittle modes, we shall describe these first.
In a polygranular material, brittle fracture can occur in either of two modes. The first is
intergranular fracture, in which the crack propagates by separating grains along the grain
boundaries, as illustrated in Fig. 22.7.

Fig. 22.8: Brittle crack propagation by transgranular cleavage.

The second brittle fracture mode is transgranular cleavage, in which the crack prop-
agates by fracturing the individual grains, as illustrated in Fig. 22.8. Cleavage describes the
mechanism of brittle fracture of a crystal. When a crystal cleaves it separates along specific
atomic planes, the preferred cleavage planes whose separation is easiest. In metals and

Page 739
Materials Science Fall, 2008

covalent crystals, these tend to be the planes with the most dense atomic packing, since
fracture along close-packed planes breaks the fewest bonds. In ionic crystals, the preferred
cleavage are planes that are both relatively close-packed and electrically neutral, like the
{100} planes in the NaCl structure.

Because a propagating cleavage crack seeks out particular crystal planes, it must re-
orient itself when it passes across a grain boundary. This has the consequence that trans-
granular cleavage cracks follow a somewhat tortuous path through a polycrystal, and tend
to be arrested at grain boundaries.

Ductile fracture

There are, in fact, several fracture mechanisms that differ in micromechanical detail
that are properly called ductile. The mechanism that is most important is microvoid coales-
cence. The mechanism is illustrated in Fig. 22.9. The material does not so much fracture as
it rip itself apart; the process is often referred to as ductile rupture. It occurs in two steps.
Voids nucleate at inclusions, large precipitates or microstructural heterogeneities, then grow
to link up with one another. Inclusions, such as oxides and sulfides, are the dominant
nucleation sites for microvoids in most structural alloys. These create voids through
inclusion fracture or decohesion at the inclusion-matrix interface in the crack-tip strain field.
The voids grow by plastic deformation, and link up to propagate the crack.

Fig. 22.9: Crack propagation by ductile rupture. The fracture surface is

left covered with "ductile rupture dimples", as shown in the
scanning electron micrograph on the right.

Ductile fracture almost invariably produces a high fracture toughness, both because
the crack tip is blunted (high ®) and because ductile rupture requires local tensile stresses of
the order of the tensile strength (high ßB ). The ductility of the material at the crack tip may
not be apparent to the naked eye; since deformation is confined to the region of the crack tip
the macroscopic fracture surface may appear rather flat and brittle. But a closer examina-

Page 740
Materials Science Fall, 2008

tion, for example, at high magnification in a scanning electron microscope, reveals that the
fracture surface is rough, and densely decorated with small craters, called ductile rupture
dimples, that are left over from the voids that grew and joined to form the fracture surface.

22.3.2 Choice of fracture mode; the ductile-brittle transition

The choice of fracture mode depends, basically, on the yield strength of the material
in the region near the crack tip. If the material undergoes extensive yielding at crack tip
stresses that are below the critical stress for the easiest of the brittle fracture modes, then the
material will fracture in the ductile mode. If the material is so strong that critical stress for
brittle fracture is reached before the material has yielded over a sufficient volume to blunt
the crack tip, then the fracture mode will be brittle. This is the essential reason that soft
materials, like pure Al, Cu and Au, invariably fracture in a ductile mode while very hard
materials, such as silica, alumina, silicon carbide and diamond, are invariably brittle.

Most of the useful structural materials have intermediate strength, and can be either
ductile or brittle, depending on the microstructure and the temperature at which they are
used. Since strength is a function of temperature, these materials undergo a ductile-brittle
transition as the temperature is lowered (Fig. 22.10). At high temperature the material
fractures in a ductile manner by a microvoid coalescence mechanism and has a relatively
high fracture toughness. When the temperature falls below the ductile-brittle transition
temperature, TB , the mode of crack propagation changes to brittle fracture either by trans-
granular cleavage or intergranular separation, whichever is easiest.

Fig. 22.10: The ductile-brittle transition; the fracture toughness decreases

dramatically over a narrow range of temperatures near TB .

Page 741
Materials Science Fall, 2008

The qualitative source of the ductile-brittle transition and its relation to the yield
strength can be illustrated by the "Yoffee diagram" shown in Fig. 22.11, which represents
the relative likelihood of plastic deformation and fracture at the tip of a pre-existing crack.
As the applied stress is increased toward failure the stress at the crack tip reaches one of
two levels first: the "yield" stress, ßY, at which significant plastic deformation occurs, or
the brittle fracture stress, ßB , at which the crack propagates in a brittle mode by the most
favorable mechanism. Extensive plastic deformation at the crack tip limits the local stress
and inhibits brittle fracture. Hence the fracture mode is ductile and the toughness high if ßY
< ßB . The ductile-brittle transition temperature, TB , is that at which ßY rises above ßB .

ßII
ß
ßI

TB ßY

Fig. 22.11: The Yoffee diagram, illustrating the source of the ductile-brittle
transition. ßI and ßII are the critical stresses for the trans-
granular and intergranular fracture modes.

The "yield" stress in the Yoffee diagram is a qualitative concept that is not precisely
defined by any available theory. It is most closely related to the yield strength under plane
strain conditions, whose value is significantly above the tensile yield strength, ßy.
However, ßY varies with ßy, so the ductile-brittle transition is most pronounced in alloys
whose yield strengths increase rapidly at low temperature. The prominent example is car-
bon steel, in which carbon solutes in the interstices of the BCC structure cause a dramatic
rise in strength as temperature is lowered. (The liberty ships were made of welded carbon
steel, and most of those that failed broke up in cold water. The Titanic was also made of
carbon steel, and recent tests on recovered samples show that its hull would be brittle in any
ocean that was not almost boiling.) The ductile-brittle transition is less commonly observed
in FCC materials, such as austenitic steels, largely because of the lower increment to the
low-temperature yield strength by solute impurities.

As suggested by the Yoffee diagram, the fracture mode below TB is that which pro-
vides the smallest fracture stress, ßB . In BCC material this may be either transgranular
cleavage or intergranular separation. In FCC material the brittle mode is ordinarily inter-
granular. While there are isolated observations of transgranular cleavage in FCC alloys, the
cleavage stress is usually high enough that no brittle transition is observed unless an
intergranular fracture mode intrudes.

Page 742
Materials Science Fall, 2008

The ductile-brittle transition is significant to design for two reasons. The first is the
toughness itself. The materials that are brittle in service are those that are used at tempera-
tures below the ductile-brittle transition. It follows that fail-safe design must be done using
material properties at the lowest service temperature that matters. Many structural design
codes, which are guided by the knowledge that has been acquired since the liberty ship
fiasco, require that structures be designed to material properties measured 15º below the
lowest intended service temperature. (This provision can lead to fiascoes of its own. I was
once peripherally involved in the design of a high-field superconducting magnet. The
pertinent authorities reasoned that, since the magnet case was a container for liquid helium,
the cryogen for the superconductor, it qualified as a pressure vessel. Hence they required
that it be designed under the ASME Boiler and Pressure Vessel Code, which incorporates
the "service temperature minus 15º" rule in some of its pertinent sections. The problem
was that the temperature of liquid helium is 4.2K. We offered a Nobel Prize nomination to
anyone who designed a successful fracture test that met the code, but there were no takers.)

It also follows that the designer (or, more appropriately, the person immediately re-
sponsible for materials selection and quality control) must pay some attention to the metal-
lurgical factors that influence the ductile-brittle transition. A structural material that tests
superbly can be a disastrous choice if metallurgical variations that are permitted in the ma-
terial specifications cause a significant upward movement in TB .

22.3.3 Suppressing brittle fracture

The understanding of the ductile-brittle transition that is gathered in the Yoffee dia-
gram also suggests useful microstructural mechanisms to lower or eliminate the ductile-
brittle transition. One obvious method is to lower the yield strength (Fig. 22.12(a)). The
low-temperature strength increment can be specifically decreased by removing interstitial
solutes or by "gettering" them into relatively innocuous precipitates or second phases. For
example, ferritic steels that are intended for low-temperature service are often given
"intercritical" heat treatments that gather carbon into isolated pockets of retained austenite
phase or are alloyed with Ti to getter carbon into precipitates.

ß ßB ß ßB

ßy ßy

TB T TB T
(a) (b)

Fig. 22.12: Lowering TB by (a) lowering the yield stress; (b) raising the
brittle fracture stress.

Page 743
Materials Science Fall, 2008

The second obvious method is to raise the brittle fracture stress (Fig. 22.2(b)). The
best metallurgical method for doing this depends on the microstructural mechanism of the
brittle fracture that defines TB . If the fracture is intergranular its source is either a grain
boundary contaminant, such as the metalloid impurities S and P in steel, or an inherent
weakness of the grain boundary, as is apparently found in Fe-Mn alloys and in intermetallic
compounds such as Ni3Al. In the case of chemical embrittlement the alloy may be purified
of deleterious surfactants, alloyed to getter these into relatively innocuous precipitates, or
heat treated to avoid the intermediate temperature regime at which these impurities segregate
most strongly to the grain boundaries. When the grain boundaries are inherently weak the
metallurgical solution is the addition of beneficial grain boundary surfactants that serve to
glue them together. The most prominent of the beneficial surfactants is boron, which is ex-
tremely effective in suppressing intergranular fracture in Fe-Mn steels and in Ni3Al
intermetallics. Carbon is also an effective surfactant in Fe-Mn steels when it is present in
low concentration.

When the brittle fracture mode is transgranular, as it is in typical ferritic steels, the
most effective microstructural technique is to decrease the grain size of the alloy. This
raises the critical stress for transgranular cleavage by decreasing the mean free path of an
element of cleavage fracture. Of course, grain refinement also raises the yield strength,
which tends to increase the ductile-brittle transition temperature. But in many alloys, and
particularly in BCC structural steels, the influence of grain size on the fracture stress
significantly outweighs its influence on strength (Fig. 22.13). Grain refinement is a widely
used method for controlling TB in ferritic (BCC) steels and weldments.

ß
ßB

TB
ßy
T

Fig. 22.13: Lowering TB by grain refinement when brittleness is

transgranular.

Transgranular cleavage can also be suppressed by adding second-phase particles

that interrupt crack propagation and blunt cleavage cracks. Interestingly, relatively "dirty"
alloys that contain a high density of inclusions may have lower values of TB that alloys
with greater purity. The inclusions lower the toughness in the ductile mode, above TB , but
also lower the transition temperature by interrupting propagating cracks. The result may be
a surprising increase in toughness when the service temperature of the alloy is slightly
below the transition temperature of the "clean" material. (I once consulted with a
manufacturer of heavy vehicles who was shocked to encounter a brittle fracture problem

Page 744
Materials Science Fall, 2008

when his supplier "improved" the chemical quality of the steel he was using for heavy
structural members.)

22.3.4 Raising the toughness in the ductile mode

The fracture mode that is conducive to a favorable combination of strength and

toughness is ductile rupture with significant plastic deformation before fracture. The
characteristic variation of the fracture toughness of a ductile material with the yield strength
at constant temperature is shown in Fig. 22.6. Over the intermediate strength range of
greatest practical interest the toughness decreases monotonically as the strength is raised.

Since ductile fracture ordinarily occurs through the nucleation and growth of voids,
the fracture toughness can be improved by addressing either void nucleation or void
growth.

Void nucleation

The most effective way to increase the toughness of a high-strength structural alloy
is to eliminate void nucleation sites. Since these are, ordinarily, inclusions that result from
impurities in the alloy, alloy purification is an effective means for raising the toughness. A
change in the inclusion count at constant microstructure causes an increase in the plane
strain fracture toughness roughly according to the relation

ßy
KIc fi 22.9
Nv

where NV is the volume density of active inclusions. Interestingly, eq. 22.9 predicts that
the inclusion count has a much stronger influence on fracture toughness as the yield stress
rises, which suggests that the effect should be most apparent at the lowest temperatures and
in the highest-strength ductile steels. This prediction is in qualitative agreement with a
number of recent observations on the behavior of high-strength steels and weldments.

When it is impractical to eliminate inclusions, it is often beneficial to change their

shape or chemical composition to improve their resistance to void nucleation. For example,
the oxides and sulfides of iron that are the dominant inclusions in ordinary structural steels
tend to deform into long "stringers" during processing that are particularly efficient void
nucleation sites. A common method for improving toughness is "inclusion shape control",
in which minor chemical additions are used to gather O and S into stable precipitates that
are spherical in shape and less likely to nucleate voids. The common additives are Ca and
rare earth metals such as Ce and La.

Void growth

For a given inclusion distribution the best available ductile fracture theories lead to
predictions of the general form

Page 745
Materials Science Fall, 2008

KIc fi ‰f Eßy 22.10

where E is Young's modulus, ßy is the tensile yield strength, and ‰f is the strain to failure,
whose precise definition (and power) varies slightly from one model to another. The ex-
plicit dependence of the fracture toughness on the yield strength suggests that the two
should vary together, in with the experimental data that invariably shows a decrease in
toughness as the strength rises (Fig. 22.6). The resolution of this discrepancy lies in the
dependence of the failure strain on the yield strength; ‰f decreases strongly and mono-
tonically with ßy at constant temperature. Its influences dominates, so KIc decreases with
ßy as well.

Eq. 22.10 suggests that the best way to increase the strength-toughness relation,
KIc(ßy), is to improve the work-hardening characteristics of the alloy so that the failure
strain increases. This is difficult to do in a controllable way, particularly since the
microstructural theory of work hardening is not very well developed. While there has been
progress, the microstructural tricks that are used are beyond the scope of this course.

22.3.5 Raising the fracture stress in the brittle mode

There are many engineering structures that must be made out of high-strength ma-
terials that fracture in a brittle mode. Examples include optical devices, like windows and
optical fibers, semiconducting devices that employ brittle silicon chips, and very high-tem-
perature devices for which strong ceramics are the only viable option. In all of these cases
it is important to use materials that maximize toughness in the brittle mode.

There are three generic methods for increasing the toughness of brittle materials:
eliminating flaws, introducing compressive stresses and increasing crack growth resistance.

Eliminating flaws

Since a brittle material fractures at the worst flaw it contains, its fracture stress
(which is the apparent value of the "ultimate tensile strength" in a tensile test) is a very
sensitive function of its flaw distribution. Any successful method of eliminating the more
severe flaws has a very beneficial effect on the fracture stress.

The best method for controlling the maximum flaw size is, of course, care in manu-
facturing and handling. In addition, at least three methods have been used to eliminate
flaws after manufacture. The first is based on the fact that most of the severe flaws origi-
nate at the surface. If the surface is coated with a ductile material that penetrates flaws, such
as a transparent, ductile polymeric coating on glass, then the surface flaws can be sealed
and, effectively, eliminated.

The second method is based on the same approach: the mitigation of surface flaws.
The surface is lightly etched with a chemical that preferentially attacks surface cracks and
blunts them by removing material from the crack tips. Since flaws in brittle materials are

Page 746
Materials Science Fall, 2008

ordinarily very sharp, a chemical blunting of this type can significantly increase the fracture
stress.

A third method, that is imaginative, if it is not generally applicable, is useful for

thermoplastic materials like silica glass that can be rewelded after breaking. When a brittle
material is stressed to failure it breaks at its most severe flaw, so the remaining pieces of the
material necessarily have a higher fracture strength than the original. If the material can be
repaired after breaking, it is stronger. By repeating this process, the severe flaws in the
material can be eliminated and the strength (fracture resistance) raised.

Compressive stress

Cracks propagate in tension. If a compressive stress is added across the plane of

the crack, the fracture stress is increased by the amount of the compression, since this must
be overcome before the crack-tip stress turns tensile. A compressive stress can be added at
the surface of a brittle material by chemical doping or by heat treatments that cause the ma-
terial near the surface to expand relative to that in the interior. A familiar example is "tem-
pered glass", which is heat treated to produce a compressive surface layer.

Crack growth resistance

The crack growth resistance of brittle materials can be improved in two ways. The
first, and most universal, is the use of internal heterogeneities to inhibit crack growth, as is
done to inhibit transgranular cleavage in metals. The most effective microstructural hetero-
geneities are second-phase particles, particularly second-phase particles that are ductile, or
that delaminate at the particle-matrix interface to blunt propagating cracks.

A much more effective device is transformation toughening, which uses stress-in-

duced martensitic phase transformations to create plastic deformation and blunt propagating
cracks. This method is restricted to a few materials, the most important of which is ZrO2
(zirconia) and its alloys with yttrium. The reason is that the material must contain particles
that are metastable with respect to martensitic transformation, but so slightly metastable that
they transform under load. Moreover, it must be possible to distribute the transforming
particles densely through the microstructure, so that they are available when and where they
are needed. Despite these restrictions, a number of transformation-toughened ceramics
have been developed, and are used in such commercial products as ceramic nails that will
not corrode, and ceramic-headed golf clubs that are exceptionally light.

22.4 FATIGUE

Fatigue was reviewed briefly in Chapter 19. It is a mechanism by which materials

that are subject to cyclic stresses become liable to catastrophic fracture over time. It is a
serious problem in engineering, and a major source of unexpected failure.

Page 747
Materials Science Fall, 2008

The possibility that a nominally sound structure will deteriorate through fatigue has
been recognized for almost two centuries, or virtually as long as engineering structures
have been made of metal alloys that may experience fatigue. The word itself is an old one,
and expresses the point of view that the metal has somehow become "exhausted" and un-
able to maintain its integrity. Sometime in the middle of the nineteenth century the "ex-
haustion" theory was challenged by the "crystallization" theory. Fracture surfaces that are
produced by fatigue usually appear bright and shiny. Since it was not yet known that
metals are crystalline, and since the materials that were known to be crystalline were brittle
compounds with shiny surfaces, the notion was promulgated that fatigue resulted from the
crystallization and consequent embrittlement of metals that were amorphous in the normal,
ductile state. Unfortunately, both the "exhaustion" and the "recrystallization" models have
considerable resilience, and can still be found in what purports to be reliable literature in the
field of mechanics. The "exhaustion" theory is especially robust. Its modern form
suggests that cyclic deformation deposits energy in the material that accumulates until it
becomes so great that atomic bonds are disrupted and the material falls apart.

In fact, the fatigue process is not all that mysterious. Fatigue is, basically, a mech-
anism for growing cracks, which, if their growth is not stopped, will eventually reach criti-
cal size and propagate to failure. While we speak of "fatigue failures", fatigue is not,
strictly speaking, a failure mechanism. The failure mechanism is unstable crack growth of
the sort we have been discussing. Fatigue is a mechanism for producing cracks that are
large enough to propagate under stresses that would otherwise be safe. Since fatigue crack
growth requires a cyclic stress, fatigue only happens in structures that experience time-de-
pendent loads. Since fatigue crack growth takes time, a structure or device may operate
successfully for many years before fatigue becomes a problem. Since ultimate failure is
due to the unstable propagation of cracks that were stable until the last increment of growth
brought them to critical size, fatigue failures are sudden, usually unexpected and often
catastrophic.

22.4.1 Fatigue crack growth

The driving force for crack growth

Let a sample that contains a crack of length, a, be subject to a cyclic stress, as

shown in Fig. 22.14.

For a given value of the stress, ßa, the crack tip experiences the stress, ßT, where
ßT increases with the square root of the crack length, a (eq. 22.1).

 a 
ßT = Q ®  ßa 22.1

The cycle in the applied stress causes a cycle in the crack tip stress, ßT, that has the same
period. If the crack tip stress exceeds the yield strength, its cycle causes a local plastic de-
formation, which tends to open the crack as the load increases and close it again as the load
returns to zero.

Page 748
Materials Science Fall, 2008

There is still some uncertainty about the exact mechanism of crack growth under
cyclic stress. In fact, there are probably several mechanisms that differ in detail and vary
from one material to another. But it fairly easy to see that cyclic plastic deformation should
cause crack growth. Plastic deformation is not completely reversible, and, as follows ulti-
mately from the Second Law of thermodynamics, the consequence of its irreversibility will
be to deform the material in the direction that relaxes the applied load, in this case, the di-
rection that propagates the crack. Hence the crack grows, at a rate that should increase with
the cyclic crack tip stress, ÎßT.

ßa

ßT

ß
ßa ßa
r
a

ßa

Fig. 22.14: A material containing a crack is subject to a stress cycle like

that shown at right.

If we ignore the variation in the effective crack tip radius, ®, over the cycle, ÎßT is
determined by the applied stress cycle, Îßa,

 a 
ÎßT = Q ®  Îßa 22.11

with a constant of proportionality that varies with the length and sharpness of the crack.
While we cannot measure ÎßT directly, it is straightforward to measure the cyclic stress
intensity, ÎK, where

ÎK = ® ÎßT = ÎßaQ a 22.12

The crack growth rate

From eq. 22.12, the driving force for crack growth is proportional to ÎK. This sug-
gests that the fatigue crack growth rate, da/dn, the crack growth per cycle of the applied
load, should be an increasing function of the cyclic stress intensity. It invariably is. The
crack growth rate, da/dn, increases with ÎK in a pattern that is nearly the same for all duc-
tile materials, and is illustrated in the log-log plot in Fig. 22.15.

Page 749
Materials Science Fall, 2008

As illustrated in Fig. 22.15, fatigue crack growth behavior can be divided into three
regimes. The most important is the central, or power law regime, in which the crack
growth rate is given by the so-called Paris Law:

da m
dn = A(ÎK) 22.13

where the exponent, m, is a constant whose value is not too far from 2 for many materials.
The crack growth behavior deviates from this simple constitutive relation at both high and
low values of ÎK.

Kmax = K Ic
threshold

power law
da/dn

fracture:
ÎK t
ÎK

Fig. 22.15: The dependence of the crack growth rate on the cyclic stress
intensity (log-log plot).

To understand behavior outside the power law region, recognize that

ÎK = Kmax - Kmin = Q-1 a ßmax - Q-1 a ßmin 22.14

In the example shown in Fig. 22.14, which we shall use for simplicity, ßmin = 0, and

ÎK = Kmax = Q-1 a ßmax 22.15

On the low-stress side, the crack growth decreases dramatically with decreasing ÎK
until it vanishes at the threshold stress intensity, ÎKt. The usual reason for this behavior is
the phenomenon of crack closure. When the maximum value of K, Kmax, is small, the
magnitude of K is not sufficient to prevent the rough sides of the crack from contacting one
another over most of the range of ÎK. As soon as the crack sides contact, the crack is ef-
fectively closed, so the stress cycle is truncated and becomes ineffective. A certain mini-
mum value of Kmax is needed to force the crack open over a sufficient portion of its cycle
to drive growth.

On the high-stress side, Kmax increases until it becomes equal to the fracture tough-
ness, KIc. When Kmax = KIc the fatigue crack has reached critical size and propagates to

Page 750
Materials Science Fall, 2008

failure. For the last few cycles before failure, Kmax is sufficiently close to KIc that the crack
growth per cycle is unusually large.

The crack length

If we focus attention on crack growth in the power law regime, which ordinarily
dominates fatigue crack growth, and substitute eq. 22.11 into 22.13, the result is

da m m m/2
dn = A(ÎK) = A'[Îßa] [a] 22.16

When the applied stress cycle, Îßa, is a constant, eq. 22.16 can be integrated to find the
crack length, a, as a function of the number of cycles, n. When m = 2, which we take as a
representative value, the result is simple:

a(n) = a0e¬a 22.17

where the constant, ¬, is

¬ = AQÎßa 22.18

and a increases exponentially with n, as illustrated in Fig. 22.16.

fracture: a = a c
ac

a0

Fig. 22.16: Crack length as a function of cycle number.

The physical reason for the accelerating crack growth that appears in Fig. 22.16 is
the dependence of the driving force, ÎK, on the crack length, a. As the crack grows under
a constant cyclic stress, the driving force, ÎK, increases according to eq. 22.15, and the
crack grows faster.

Fig. 22.16 illustrates an important qualitative feature of fatigue that every engineer
(and, in fact, every person who uses equipment) should know. Fatigue crack growth accel-

Page 751
Materials Science Fall, 2008

erates with cycle number. A fatigue crack may remain small and, apparently, innocuous for
a very long time, before suddenly accelerating its growth and failing in a very short time.

The real situation is much worse than is suggested by eq. 22.16, which represents
fatigue crack growth under a constant cyclic load. Real devices experience a spectrum of
load cycles, most of which are relatively small. When the crack is small these minor loads
contribute almost nothing to fatigue crack growth. Most of them produce cyclic stress in-
tensities that are much less than the threshold value, ÎKt. But as the crack grows the cyclic
stress intensity produced by a stress cycle of given amplitude, Îßa, increases with a .
Stress cycles that are innocuous when the crack is small become damaging when it is large.
Fig. 2.16 illustrates crack growth as a function of the number of cycles. The user of a de-
vice is more interested in crack growth per unit time in service. The mounting significance
of minor cycles can increase the number of cycles per unit time by orders of magnitude,
dramatically affecting fatigue failure.

For example, I once did an analysis of a fatigue failure of a structural member in an

airplane that caused a fatal crash. Since the fracture surface was clean and very well
preserved, it was possible to trace out the propagation of the fatigue crack. The analysis
showed that the fatigue crack had grown to barely perceptible size over a very long period
of time, perhaps several years. It then grew from barely perceptible size to the several
inches that were required for instability in approximately 15 minutes of flight time. The
reason was that while the crack was microscopic only the most severe loads contributed to
its growth. However, once it reached that barely perceptible size, the normal operating
loads of the aircraft contributed to fatigue crack growth, with the consequence that the
number of damaging cycles per unit time changed from 1-2 per flight hour to 2-3x103 per
minute of flight time.

22.4.2 Nucleation-limited fatigue

The cracks that can grow to failure under fatigue conditions are those that have suf-
ficient length to generate cyclic stress intensities greater than the threshold value, ÎKt, un-
der the cyclic stresses that are experienced in service. If there are no such cracks, the ma-
terial is proof against fatigue unless such cracks can be nucleated. Unfortunately, they
often can be.

If the maximal value of the cyclic load on a material is sufficient to cause plastic de-
formation anywhere within it, for example, in favorably oriented grains or in the stress
concentration fields of inclusions, then the cyclic stress may set up a cyclic plastic defor-
mation that leads to crack nucleation. The mechanism is the formation of persistent slip
bands, which are microscopic bands of within which plastic deformation occurs with each
stress cycle. These bands can accumulate dislocations, and, eventually, decohere to nucle-
ate potential fatigue cracks.

Crack nucleation is responsible for the fatigue of nominally sound structural mem-
bers that are subject to relatively low stresses. The typical behavior of such a material un-
der a cyclic load that varies from sm to -sm is shown in Fig. 22.17. The sample fails after a

Page 752
Materials Science Fall, 2008

number of cycles that decreases with the stress amplitude, sm. The curve that represents the
fatigue life as a function of the log of the number of stress cycles is called an s-n curve.
The s-n curve terminates at the fatigue limit, sl. In steels and a number of other materials,
the fatigue limit is real; the material does not fatigue if sm < sl. In other materials, such as
Al and most of its alloys, there is no strict fatigue limit, but the s-n curve nearly asymptotes
by n = 108, and this is a good estimate of the practical fatigue limit.

sm
su
s

-sm
s
t
sl

log(n)
...
Fig. 22.17: The s-n curve of a typical metal. The stress oscillates with
time as shown at right. The fatigue limit, sl, is either the
asymptote of the curve or the stress that causes failure in 108
cycles.

The fatigue limit of a part in service does change if the part is subject to a mean
stress in addition to the cyclic stress, or if it contains notches or geometric irregularities that
introduce stress concentrations. The effect of there, and other pertinent variables, has been
studied, and design criteria are available that take them into account.

Studies of s-n fatigue have shown that the bulk of the fatigue life at a given value of
the plastic stress is associated with the nucleation of the fatigue cracks. The fatigue limit is
a stress so low that fatigue cracks are very unlikely to nucleate.

In many structures the fatigue cycle is a cyclic strain rather than a cyclic stress. A
simple example is a joint between dissimilar materials that expand by different amounts in
response to an elastic stress or a temperature excursion. In this case the number of cycles
to failure decreases with the cyclic plastic strain.

22.4.3 Fatigue-safe design

Fail-safe design

A part is safe against fatigue if the cyclic stress, sm, is below the fatigue limit (as
adjusted for mean stress and stress concentration effects) and if the material contains no
crack of size sufficient to generate a cyclic stress intensity that exceeds the threshold value.

Page 753
Materials Science Fall, 2008

When a structure is not weight-limited it is usually possible to design it to operate at low

values of the cyclic stress, and choose a material that has reasonable values of sl and ÎKt,
and can be fabricated and inspected to ensure a starting condition that is free of significant
flaws. This is the approach that is usually taken in design. Most of the parts that are cur-
rently used in operating engineering systems, from microelectronic devices to automobiles
to airplanes, are designed to fail-safe criteria.

Flaw-sensitive design

Fail-safe criteria do, however, impose a significant weight and performance penalty
since they require very low operating stresses. Where higher performance is essential and
cost is not limiting, as in critical parts in aerospace vehicles, it is possible to use an alterna-
tive, flaw-sensitive design procedure that is based on the combined use of non-destructive
and the fatigue crack growth curve.

To use a flaw-sensitive design, the fracture toughness and fatigue crack growth
curve of the material must be known. The part must be inspectable during service, and the
minimum capabilities of the non-destructive inspection procedure must be well-character-
ized.

Let the non-destructive testing procedure be definitively capable of identifying flaws

that are larger than some known minimum size. It follows that there is a "worst case" flaw
of effective length a* that cannot be detected. If a part is inspected without flaw indications,
it is conservative to assume that the "worst case" non-detectable flaw, a0, is present in it. If
this is the case then the initial rate of flaw growth is determined by the cyclic stress
intensity,

ÎK = QÎß a0 22.19

where Q is a constant (assuming the worst flaw geometry) and Îß is the cyclic applied
stress. This flaw size will grow with time at a rate that is predictable from a knowledge of
the cyclic operating stress and the fatigue crack growth curve. It is usually conservative to
assume that the flaw grows in the power law regime:

da m
dn = A(ÎK) 22.20

where da/dn is the crack growth per cycle and A and m are material constants. A conserva-
tive fracture criterion gives the crack size at fracture as

 KIc 2
ac = Q-1ß  22.21
max

where KIc is the plane strain fracture toughness and ßmax is the maximum of the cyclic ap-
plied stress. The structure is hence safe if the number of cycles is such that, say, a < 0.1ac.
If the material has sufficient fracture toughness for a crack of size 0.1ac to be found easily,

Page 754
Materials Science Fall, 2008

then the structure can be inspected after it has sustained enough cycles for the worst flaw,
a0, to grow to size 0.1ac. This criterion (or some other suitable criterion on the flaw size)
can be used to set an inspection interval. If a flaw is found on inspection the part is retired
or repaired. If no flaw is found, the largest flaw that can be present at that time is a0', the
largest flaw that cannot be reliably found by the inspection procedure. One assumes that
this flaw is present, and sets a new safe inspection period on this basis. In this approach,
periodic inspection is used to insure that the part will not fail in fatigue. Flaw-sensitive
design is now used for many critical parts.

It is, of course, a very good idea to inspect periodically for fatigue damage even
when fail-safe design has been used, since designs are imperfect, flaws may be introduced
during operation, and, for many reasons that are difficult to anticipate, parts may be subject
to cyclic stresses that are beyond those anticipated in the design. This is commonly done.
But it is also important to take appropriate action when a fatigue crack is actually found.
Unfortunately, this is not always done, largely because of ignorance of how fatigue cracks
behave. A fatigue crack that exists in a structure has grown to its present size, and will
continue to grow at a rapidly accelerating rate unless the part is taken out of service or the
part is repaired to remove the crack. It makes no sense to assume the crack is innocuous
because its present size is below the critical size for fracture. The crack is growing, and it is
its future size that one should worry about. A part that contains a visible fatigue crack must
be retired or repaired unless it is known that the crack will not grow to critical size before
its next inspection, or unless the failure of the part would not pose a threat to safety.

Page 755