2-1

Lecture 2

Mechanical Testing and Elementary Definitions of Stress and Strain

To determine mechanical properties, we must apply loads and measure the resulting deformation.
There are many ways to do this. Some common test configurations are described in this chapter.
In addition, we know from experience that the force required to produce a given deformation
depends on the object size—for example, suppose we have two springs, both made of the same
material and having the same shape but one larger than the other. The larger spring will be
stiffer. To remove the effects of object size so that we can focus on the properties of the material,
we introduce the very important concepts of stress and strain.

The Uniaxial Tension Test

In a uniaxial tension test, equal and opposite forces are applied at the ends of a right circular
cylindera subjecting it to mechanical tension as shown in Figure 2.1a. It is important that these
forces are distributed uniformly across the ends of the cylinder and parallel to the cylinder axis.
Phenomenologically, we see that the cylinder responds by stretching in the direction of the
applied forces and by shrinking in directions transverse to the applied forces as shown in Figure
2.1b.
F
before loading

A
L0 L

load applied
(a) F (b)

Figure 2.1: The uniaxial tension test.

We can normalize the force F applied to account for the diameter of the cylinder by dividing by
the cross-sectional area A (Fig. 2.1a). The resulting quantity is a normal stress (force applied
normal to the area) and is written
𝐹
𝜎= . (2.1)
𝐴

a
For now, we consider only materials that are both isotropic and homogeneous!

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-2

At the same stress, cylinders of different diameters will still experience the same force per
atomic bond on average, so we can be sure we are characterizing the material accurately, even if
we use cylinders of different diameters.
We can characterize the resulting elongation, normalizing for the length of the cylinder, by
integrating the infinitesimal elongations Dl normalized by the current length l as the sample is
stretched from its original length L0 to its final length L. The resulting quantity is the normal
strain and is written
!
𝑑𝑙 𝐿
𝜀=( = 𝑙𝑛 , . . (2.2)
!! 𝑙 𝐿"
At the same strain, cylinders of different lengths will still experience the same amount of stretch
per atomic bond on average, so we can be sure we are characterizing the material accurately,
even if we use cylinders of different lengths.
We immediately note that there are engineering simplifications for both stress and strain that
are convenient in some applications. First, we see that to determine the actual stress, we must
keep track of how much the cross-sectional area shrinks during the uniaxial tension test. If we do
this, then the stress given by Eq. 2.1 is the true normal stress. However, this requires constantly
measuring the sample diameter, so it is common in many cases to just report the force divided by
the initial (unloaded) cross sectional area, A0. The stress in this case is known as the engineering
normal stress and is given by
𝐹
𝜎#$% = . (2.3)
𝐴"
Although the engineering stress is more and more inaccurate the more the sample deforms, it has
an important application in structural materials as described in the next section.
Second, we define the engineering normal strain as
𝐿 − 𝐿" Δ𝐿
𝜀#$% = = . (2.4)
𝐿" 𝐿"
For small strains, L - L0 = DL is small and eENG ≈ e. The engineering strain was important before
the advent of electronic calculators when the natural log of a number had to be looked up in a
table or estimated using a slide rule. But it is still convenient to use when strains are small. For
example, it is often a reasonable approximation for the elastic strains in metals and ceramics. By
comparison, we refer to the strain given in Eq. 2.2 as the true normal strain.
Unless otherwise specified, we will use the true stress and strain values. The stress-strain
curve obtained in a uniaxial tension test provides information about elastic and plastic
deformation as well as fracture that constitute a “fingerprint” of the mechanical behavior of a
material. Some characteristic stress-strain types are shown below.

Stress-Strain Curves for Ductile Metals

A true-stress-true-strain curve that has characteristics of a uniaxial tension test of a typical
ductile metal is shown in Figure 2.2a. The curve begins with an elastic section that appears linear
but is actually very slightly curved. We know that it is elastic because if we unload from any
point in that region (say point A), the strain returns to zero. If we reload and go further, the curve
bends over and the strain increases much more rapidly with stress. If we unload from some point

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-3

(say point B) on this part of the curve, we find that we have some residual strain. To the extent
that the residual strain is permanent, we know that we have plastic deformation.

𝜎 B 𝜎 strain hardening
fracture
𝜎y perfectly
plastic
A

𝜀 0.02
𝜀
plastic elastic (a) (b)

𝜎ENG
ultimate tensile
. stress
𝜀 𝜎y
𝜎

𝜀 𝜀ENG (nominal)
(c) (d)

Figure 2.2: Typical stress-strain behavior in uniaxial tension tests of ductile metals.

For reasons that we will discuss later, the onset of plastic deformation is gradual and it is difficult
to identify the exact point at which plastic deformation starts. By convention, we define the yield
stress, sy, as the point where the residual plastic strain is 0.002b as shown in Figure 2.2b. At this
point plastic deformation is large enough to reliably detect, but not so large to be important in
most situations. The yield stress is thus treated as the point at which plastic deformation ‘begins.’
Note that sy depends strongly on the deformation history of the sample. Suppose a test is
conducted where the sample is loaded to point B and unloaded. The yield stress would just be the
point indicated in Fig. 2.2a. But now suppose the sample is then sent to another laboratory where
it is again tested in uniaxial tension. The second test would start where the first test ended. The
load would increase along a linear-appearing line directly to point B at which point plastic
deformation would begin. The researchers in the second laboratory would find that sy
corresponded approximately to point B. Thus, sy is not a materials property (although it could be
thought of as a property of a particular sample or class of identically prepared samples).
Because the stress needed to continue to deform the material increases with plastic strain it is
clear that the strength of the material is increasing. This is referred to as strain hardening. Strain
hardening can be quantified by comparing with the case where plastic strain would continue
indefinitely at sy—i.e. without strain hardening. That case is referred to as perfectly plastic and

b
This point is arbitrary, but is very well established.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-4

is shown in Fig. 2.2b. The difference between the perfectly plastic line and the actual curve is the
strain hardening. We also note that the stress level and the strain hardening in the plastic regime
increases with strain rate (𝜀̇ = 𝜕𝜀/𝜕𝑡) as illustrated in Figure 2.2c. The rate at which the stress at
a given strain increases with strain rate is known as the strain rate sensitivity. We will explore
all these concepts in detail later!
Another very important phenomenological observation is shown in Figure 2.3. In the initial
stages of plastic deformation, the cross-section of the sample shrinks at the same rate
everywhere. This is because if one part of the sample starts to deform faster than the others, it
becomes stronger due to strain hardening and stops deforming, allowing the other parts to catch
up. This is known as uniform plastic deformation. But the strain hardening rate decreases with
increasing strain. Eventually when one part starts to deform faster than others, the increase in
stress due to the local decrease in cross section is greater than the increase in strength due to
strain hardening and subsequent deformation will be concentrated in this region. This region
continues to stretch in the tensile direction and shrink in the transverse direction forming a neck.
The process of neck formation is called necking.

(a) initial (b) uniform (c) necking (d) failure

shape plastic
strain

Figure 2.3: Uniform plastic strain and necking.

Note that, to make the curves shown in Fig’s 2.2a, 2.2b, and 2.2c, we had to be able to obtain the
true stress and true strain at some point in the neck. This is very difficult to do! Consequently,
the data from uniaxial tension tests are often reported as engineering stress vs. nominal true strain
as shown in Figure 2.2d. The nominal true strain is simply determined from the overall extension
of a part of the gauge section that includes both the neck and surrounding uniformly strained
regions. Although the true stress in the neck is continuing to increase due to strain hardening, the
engineering stress bends over and starts to decrease as the force needed to generate the stress
required to deform the material decreases due to the decrease in the area of the neck.
Furthermore, the strain in the neck is much higher than the nominal true strain that is reported.
While such a curve is not very useful for studying the stress-strain behavior of a material,
especially at higher strains, it has great utility in structural engineering. Since sENG is just the
force applied to the sample scaled by a constant, the maximum value of the sENG–e curve

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-5

represents the maximum load that the sample can support. This point is referred to as the
ultimate tensile stress, or UTS. The scaling by A0 makes it possible to directly compare UTS
values from samples of different diameters.
Finally, once the stress and strain get to be high enough, the sample breaks into two (or
more!) pieces. This process can have many different geometric forms. If a crack propagates
through the sample, then this is a fracture process. If the neck in Fig. 2.3 simply shrinks to a
point, this would be referred to as ductile rupture. We will explore these geometries in detail
later.

Stress-Strain Curves for Brittle Ceramics

A true-stress-true-strain curve from a uniaxial tension test of a brittle ceramic is quite simple as
is shown in Figure 2.4. Because there is ‘no’ plastic deformation, the stress just increases linearly
with strain until the sample fractures.

fracture
𝜎

𝜀
Figure 2.4: Typical stress-strain behavior for a brittle ceramic.

Stress-Strain Curves for Polymers and Composites

There is no “typical” true-stress-true-strain curve that represents all, or even most, polymers and
the behavior of composites is even more diverse. But just to show how distinct such stress-strain
curves can be and to emphasize the time-dependent nature of polymer behavior, we show a
schematic true-stress-true-strain curve from the elastic regime of a typical elastomer (a category
that includes rubber bands) in Figure 2.5. The stress increases and decreases non-linearly with
strain. Although there is initially some residual strain when the sample is fully unloaded, we
know that the deformation is elastic because it recovers fully over time. We will learn about this
behavior later too!

𝜀
Figure 2.5: Stress-strain behavior in the elastic regime for an elastomer.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-6

Additional Test Details and Comments

Two important additional details regarding uniaxial tensile tests are worth noting. First, to
actually be uniaxial, there must be a way to ensure that the force is uniformly distributed over the
ends of the sample. This is done by producing samples in a “dogbone” shape with much thicker
grip sections at the end that taper to a much thinner gauge section in the middle, as shown in
Figure 2.6a. The grip sections are clamped or screwed into the testing machine and are subjected
to complex inhomogenous stress states. But because the grips are so much larger than the gauge
sections, the stresses in the grips are much lower than those in the gauge section such that the
vast majority of the deformation during the test occurs in the gauge section.

screws
grip
section crosshead

load cell

grips
gauge
sample
section
grips

hydraulic
actuator

base

(a) Tensile test sample (b) Screw-driven machine (c) Hydraulic machine

Figure 2.6: Tensile test sample and machines.

Second, there are two general types of testing machines. In one (Figure 2.6b), one end of the
sample is attached to a rigid base and the other is attached to a load cell that is in turn attached to
a rigid crosshead beam. The crosshead rides on massive screws that are attached to the base. By
turning the screws, the crosshead can be raised, stretching the sample. To the extent that the
machine is massive and rigid compared to the sample, the machine effectively imposes
displacements on the sample. Such a test is said to be conducted under displacement control
and the testing machine is said to be rigid. In the other type of test machine (Figure 2.6c), one
end of the sample is attached to a massive hydraulic cylinder that is attached to a rigid base and
the other is attached to a load cell that is in turn attached to a rigid crosshead beam that is directly
mounted to the base. The hydraulic cylinder is pressurized to apply force to the sample. Such a
test is said to be conducted under load control and the testing machine is said to be compliant.
Both machines have advantages and disadvantages and we will make note of some of them as we
go along.
Uniaxial tension tests are so common and so widely used that our understanding of
mechanical properties is dominated by (and distorted by) this kind of testing. To understand
mechanical behavior more accurately, we will have to learn to think more broadly. It is worth
noting that the reason that uniaxial tension tests are so widely used is because they are the easiest

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-7

tests to interpret, not because they are the easiest to conduct. In fact, conducting accurate uniaxial
tension tests can be quite challenging. We will note some of these difficulties and their effects in
relevant parts of the course.

Torsion Tests
In a torsion test, a right circular cylinder is subjected to equal and opposite torques applied
along the cylinder axis at the opposing ends of the cylinder as shown in Figure 2.7a. This causes
the cylinder to twist along its axis.
M A
Fs

Fs
(b)
x

h 𝜙

(c)
(a) M

Figure 2.7: (a) Torsion test. (b) Shear stress. (c) Simple shear.

Consider an infinitesimal cube at the cylinder surface with its top and bottom surfaces
perpendicular to the cylinder axis as shown in Figure 2.7b. The force transmitted to this cube by
the elements lying above and below it lie in the plane of the top and bottom faces as shown. We
denote this force as Fs. As before, we normalize the force by the area over which it acts to obtain
a stress. Since the force lies in the plane upon which it acts, this is a shear stress,
𝐹&
𝜏= . (2.5)
𝐴
Because the area does not change due to shear, there is no distinction between true and
engineering shear stresses as there was for normal stresses.
The conventional way to define the shear strain is shown in Figure 2.7c. We see that the top
of the cube is shifted by a distance x in the plane. This particular form of deformation is known
as simple shear. If the height of the cube is h then the engineering shear strain is defined as
𝑥
𝛾#$% = . (2.6)
ℎ
The angle between the deformed and undeformed vertical sides of the cube, f (in radians), is also
used as a definition of shear strain. We will refer to this quantity as the nominal shear strain, gn
= f. For small angles f ≈ tan(f) and gn ≈ gENG. We will not use gn much in this class.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-8

It is important to note that the shear strain as defined in Eq. 2.6 is not actually an accurate
measure of shear deformation. It is, however, the standard conventional representation of the
shear strain. We will learn about this distinction in the next lecture.
Torsion tests are not nearly as common as uniaxial tension tests but are convenient when
measurements of shear behavior are desired. Note that the shear strain, and thus shear stress,
increases with distance along the radius of the torqued cylinder so the stress state in a torsion test
is inhomogeneous. This makes interpretation of torsion tests much more complicated than
interpretation of uniaxial tension testsc.

Three- and Four-Point Bend Tests

Another type of test that is often used is a bend test, which is popular because sample preparation
is very easy; all that is needed is a rectangular bard of the sample material. Load is applied to the
sample along either 3 or 4 lines on the top and bottom faces of the sample as shown in Figures
2.8a and b, respectively. The sample bends in a complicated way in response (not shown).
Obviously, the stress distribution is very inhomogeneous. In the configurations shown, the
bottom of the bar is in tension, the top is in compression, and there are stress concentrations at
the contact points. However, for basic measurements, say finding sy or the fracture stress or a
rough measure of stiffness, such tests are very efficient.
F F/2 F/2

h L h L1 L2

F/2 F/2 F/2 F/2

(a) (b)

Figure 2.8: Three-point (a) and four point (b) bend tests.

In a 3-point bend test the highest tensile stresses are concentrated in a small region (roughly the
red region in Fig. 2.8a) opposite the single load line and the test thus samples effectively only
that small region. In a 4-point bend test the maximum tensile stress is constant along a region on
the opposite surface between the inner load lines (roughly the red region in Fig. 2.8b). The
maximum stress and strain can be calculated from the geometry of the sample and test and the
applied load and measured deflection.

Hardness Tests
By far the most widely performed mechanical property test is the hardness test (Figure. 2.9).

c
Sometimes people do tension tests of thin-walled tubes to minimize the range of radii involved, reducing the stress
variations within the sample.
d
Round rods or other shapes may be used.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-9

Indentation
Indenter
F
Indenter Indentation A
Sample Sample

cross section plan view

(a) (b) (c)

Figure 2.9: Hardness test.

In such tests a pointed or convex tip is pressed into a sample with a known load, F, for a fixed
time (Fig. 2.9a) and is removed (Fig. 2.9b). If there is a residual indentation (Fig’s 2.9b & c), a
measure of the area, A, of that indentation (usually either the surface area of the indentation or
the area of the indentation projected into the surface plane) is determined and the hardness is
defined as
𝐹
𝐻= . (2.7)
𝐴
The hardness is just the average pressure under the loaded indenter and is a measure of resistance
of the material to permanent deformation. One reason for the widespread use of hardness tests is
that, for ductile metals, H ≈ 3sy. Thus, sellers and purchasers of metals and devices made of
metals often use hardness tests to verify the identity and strength of the metals they are dealing
with. In addition, hardness tests can be scaled to very small dimensions (e.g. “nanoindentation”)
and have become very useful tools in micro- and nano-technologies.

An Elementary View of Stress States

As we will see, the stress state at a point in a loaded body is more complicated than a simple
scalar or vector quantity. In most cases, there are both normal and shear stresses and in all cases
we must keep track of the 3-D nature of the stress state. For example, in Figure 2.1 we looked at
a plane perpendicular to the applied force to find the normal stress in a uniaxial tension test.
However, there are many possible planes in the sample we could look at instead. If we were to
look at a plane lying at an oblique angle to the tensile axis, the force would no longer be
perpendicular to the plane nor would the area of the plane be the same as before. To see the more
general case, we consider a plane P lying at an angle q to the applied force direction in a uniaxial
test of a rectangular bar as shown in Figure 2.10. If we apply a stress sapp = F/A along the tensile
axis, what is the stress on plane P?

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-10

Figure 2.10: Oblique plane in a rectangular bar subjected to uniaxial tension.

We make an imaginary cut along plane P (Figure 2.11a) and determine the forces that must
act on the resulting surfaces to keep the pieces in static equilibrium (i.e. keep them from flying
apart).
F
F

F S

θ
F N
P
F
F
F
F F
N

θ
F S

F F
(a) (b)
Figure 2.11: Forces acting on an oblique plane

We resolve the force acting on each surface into two components, one normal to the plane
and one parallel to the plane as shown in Figure 2.11b. Let the normal component be FN and the
shear component be FS. If the applied uniaxial force is F, then we have FS = Fcos(q) and FN =
Fsin(q). If the cross-sectional area of the bar is A (Fig. 2.10), then the area of the oblique plane is
Aop = A/sin(q). Thus, the normal stress s acting on plane P is
s = FN/Aop = Fsin2(q)/A = sappsin2(q) , (2.8)

and the shear stress acting on plane P is

t = FS/Aop = Fsin(q)cos(q)/A = sappsin(2q)/2 . (2.9)

Clearly the values of both normal and shear stress depend on the orientation of the plane P
that we choose to look at! We can compute s and t for a plane at any angle q from Eq’s 2.8 and
2.9. The variation in s and t with q is shown in Figure 2.12.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker

 2-11

F/A o
τ
sin2 (θ )
σ F/2A o
2
sin (θ )

0 22.5 45 67.5 90 0 22.5 45 67.5 90

θ (Theta) θ (Theta)

Figure 2.12: Variation of normal (s) and shear (t) stress with angle q.

Importantly, we note that both normal and shear stresses are acting at every point in our
loaded uniaxial tensile test sample. From Fig. 2.12, we see that the normal stress is maximum on
planes oriented perpendicular to the applied stress axis and that the shear stress is maximum on
planes oriented 45˚ to the applied stress axis and has magnitude sapp/2. As it turns out, stress is
neither a scalar nor a vector, but a quantity known as a second rank tensor. We will explore this
further in Lecture 3.
Note that we could make very similar arguments for strain, which is also a second rank
tensor quantity that will be discussed in Lecture 3.

Summary
Uniaxial tension, torsion, bending, and hardness tests are the most common means of
determining general mechanical properties (stiffness and strength). There are many other tests
that are used to determine mechanical properties in particular geometries (e.g. thin layers) or to
study particular deformation mechanisms (e.g. anelastic, creep, or fracture). In all cases the test
geometry, conditions, and protocol are optimized to apply a known stress state that can be
analyzed to determine mechanical properties.

Cornell MS&E 5802 Lecture 2 Ó2024 S.P. Baker