LOW TEMPERATURE TESTING OF ELASTOMERS

Presented at the Spring Technical Meeting of the New York Rubber Group

Thursday, March 25, 1999

By,

Andrew C. Farinella
and
Daniel L. Hertz, Jr.
LOW TEMPERATURE DATA
Two basic questions (as suggested by John Ferry, 1986 [4]) that an engineer designing a rubber
seal might want to ask are:

1. After a sudden stress how long will it take for the elastomer to adjust its shape to maintain the seal?
2. If there are oscillating stresses, will there be sufficient time in the period of oscillation for the elastomer
to adjust its shape to maintain the seal?

Correct answers to these questions may have saved the lives of the astronauts aboard the Space Shuttle
“Challenger.” Thus it is imperative that we find the answers, prepare for the future and avert similar
tragedies.
A third question for a design or R&D engineer might want to ask is:
3. What tests can I use in the lab to analyze and verify my design?

In this paper we will discuss various historical attempts to answer these questions and propose
alternate methods.

PURPOSE OF ENGINEERING DATA

The key question we must ask here is why do we do these various tests for rubber properties? The
primary answers are 1) Quality Control and 2) to use the test results to predict performance under design
conditions. For example, if we want to design a rubber hanger to support 500 lbs., we can use the tensile
strength of the material as a design tool. From the equation:

PMax
Eq. 1 SU =
A
Strength = Max Load/Cross Sectional Area, we know the tensile strength of the material and the maximum
load corresponding to a given cross sectional area. For our design, we know the load to bear, and the
strength of the material. The unknown is the size (cross sectional area) of the hanger. Rearranging, we can
solve for the required area:

PDesign
Eq. 2 AMin =
Su
For our example, let’s assume a tensile strength of 2,500 psi. The load to bear is 500 lbs. Thus the cross
section must be at least 0.2 in2; for a square cross section, the length of a side is about 0.45 inches (the
square root of 0.2).
This example assumes that design conditions are (nearly) the same as test conditions. That is, the
loading is (nearly) static and tensile, and the temperature is the same as test conditions. Tensile strength is
a well-defined test, easily correlated for use in design*.

The whole purpose of engineering data is to classify and characterize the properties/behavior of
engineering materials in order to predict their behavior under other geometric, loading, or temperature
conditions. This reduces the amount of “trial and error” engineering and significantly reduces cost and
increases speed of design. Tests that cannot be correlated to other sets of conditions are of little use to the
design engineer. Mooney viscosity and many of the low temperature tests are just a few examples.

*
This is presented as a simple example for linear materials; other factors are involved for elastomeric
materials.
1
EXAMPLE OF POOR USE OF ENGINEERING DATA

Mooney Viscometer
As a quick example, let’s look at the typical industry use of the Mooney viscometer or cure
rheometer. Generally it is used as a quality control tool for various rubber compounds. Torque versus time
is plotted as the rotor either rotates (constant speed) or oscillates (constant frequency). Be aware that
viscosity (and consequently the measured torque) varies with shear rate. That is, the ratio of torque to rotor
speed is NOT a constant value. In fact, the viscosity of many polymer melts actually looks like this:

Viscosity vs. Shear Rate*

1.E+05

1.E+04

1.E+03

1.E+02

1.E+01

1.E+00
0 1 10 100
Shear Rate (s-1) *Bird, Armstrong, Hassager, p.173

Figure 1
It is worth noting that Mooney viscometers impose a shear rate of approximately 1 s-1, while typical shear
rates in injection molding operations are in the neighborhood of 1000 s-1. This information, combined with
the chart above, should raise a red flag to anyone using a Mooney viscometer as a check for injection
molding operations. Not only does the Mooney viscometer use a single (low) shear rate [6], but also its
data cannot be used to predict behavior in other geometric configurations. That is, there is no definable
“state of stress” for the sample in a Mooney viscometer.

INTRODUCTION TO MATERIAL BEHAVIOR

Perfectly Elastic Spring

Given a sinusoidal deflection (or similarly, strain) input, there will be a sinusoidal stress output
with no phase lag- i.e. tan delta is zero. No dissipation is occurring. The following two curves show a
sinusoidal strain input (black curve, y-axis is deflection in inches) and a perfectly elastic force response
(red curve, y-axis in units of pounds force) versus time. The curves are exactly in phase. That is, the only
difference between the two curves is in magnitude; both curves have the same argument, namely ω t. Thus
the input deflection is:

2
Eq. 3 y = A sin(ωt )
and the measured force response is:
Eq. 4 F = −kA sin(ωt )

0.1 0.2 0.3 0.4 0.5 0.6

-2

-4
A Purely Elastic System- A Spring

Figures 2a and 2b

As an example, think of a steel spring. The spring can be stretched between two fixed points and
years later, the spring will not lose the stored energy. We can get the energy back and perform work even
after a large amount of time. The further you stretch the spring, the more energy is stored, and it doesn’t
matter how fast you stretch the spring, only how far.

Purely Viscous Fluid

Given a sinusoidal deflection (or similarly, strain) input, there will be a sinusoidal stress output
with a 90º phase lag. Therefore, tan delta is undefined. This means that the input and output curves will
differ both in magnitude and in phase. The phase difference is exactly 90º. The fluid does not store any
energy- all the work is dissipated. Thus the input deflection is the same as before, and the output force is:
•
é ù
Eq. 5 F = ê− c y ú = [− cωA cos(ωt )] = [− cωA sin(ωt + π 2 )]
ë û
10

0.1 0.2 0.3 0.4 0.5 0.6

-5

-10
A Purely Viscous System- A Dashpot
Figures 3a and 3b

As an example, think of a damper on an office door. The faster you push the door, the harder it is
to close. Unlike the spring, if you push the door open, it will not close after you let go; the damper does not
store any energy. The force it takes to move the damper doesn’t depend on how far you move the damper,
only how fast.

3
Viscoelastic Materials
Sinusoidal Strain Input, Sinusoidal Stress Output- Phase Lag is somewhere between 0º and 90º;
there is some storage and some dissipation. The sinusoidal deflection input is the same as before, but the
output is a combination of the elastic and viscous responses:
Eq. 6 F = − kA sin(ωt ) − cωA cos(ωt )
This response can also be expressed as a single sine wave as follows:
Eq. 7 F = − f sin(ωt + δ ) , where the amplitude of the response is

Eq. 8 f = A 2 (k 2 + c 2ω 2 ) and the negative sign is simply an indicator of the direction of the
forces.

10

0.1 0.2 0.3 0.4 0.5 0.6

-5

-10
A Viscoelastic System- Polymers!

Figures 4a and 4b

As an example, think of a rubber band. There is some storage- you can stretch the rubber band
and launch it (i.e. recover the energy). There is also some dissipation- stretch a rubber band while holding
it over your upper lip and feel the heat buildup- this energy cannot be recovered!

Comparison of Responses:
We can plot all of the responses on the same axes to compare the results of a sinusoidal input.
They show respectively, the response of a purely elastic material, a purely viscous material and a
viscoelastic material.

10

0.1 0.2 0.3 0.4 0.5 0.6

-5

-10

Figure 5

4
Tan Delta
To define tan delta, we must first define delta. Delta (usually shown with the Greek letter, δ) is
the amount by which the sinusoidal force lags a sinusoidal displacement. Since the phase angle is
somewhere between 0º (for perfectly elastic) and 90º (for purely viscous), it is used as a measure of how
“viscous” a material is.

Force and Deflection vs. Time

10

5 Eq. 9 δ = ω (∆t )

0.1 0.2 0.3 0.4 0.5 0.6

-5

-10

Time Lag, ∆t

Figure 6

Loss Modulus
As we have seen, the response of a viscoelastic material to a sinusoidal input can be broken down
into a viscous component and an elastic component. For a static test, we can calculate a modulus as the
ratio of stress to strain. For a dynamic test, we can calculate a complex modulus as the ratio of stress and
strain amplitudes. Thus a complex shear modulus, G*, can be measured using a parallel plate rheometer. It
can be shown that the phase angle, delta, relates to the ratio of viscous to storage (elastic) modulus via the
following equation:

G' '
Eq. 10 tan δ =
G'
G*
G’’

G’
Figure 7

(which relates storage and loss modulus in terms of a right triangle and the Pythagorean Theorem) and that
G’ (storage modulus) and G’’ (loss modulus) are related by:

Eq. 11 G* = (G')2 + (G' ')2

Thus, we can measure G* and δ through simple experiments and calculate G’ and G’’.
5
COMMON ASTM TESTS

Brittle Point, ASTM D2137:

This test uses a hammer of a specific mass, moving at a specified speed toward five rubber
samples held by a grip. The samples and the grip are cooled to a low temperature. The hammer strikes the
samples at the specified speed and continues moving past. The temperature at which fracture occurs in
three of the five samples is said to be the brittle point of the material.

HAMMER

GRIP

SAMPLES
Figures 8a and 8b
This test has numerous flaws concerning its utility to the design engineer. The test samples are
undergoing a combination of bending and shear under dynamic conditions (impact). Thus the state of
stress in the material is not definable. Also, this is again a test that provides data at only one “shear rate”
(i.e. speed of hammer at impact). Polymer liquids have a viscosity that is exponentially dependent on
temperature as well as shear rate dependent at a given temperature. A design engineer cannot expect to use
brittle point data to design for low temperature applications.

TR Test, ASTM D1329:

The “TR,” or “time-to-return,” test uses samples clicked from a sheet (or an o-ring for method
D1414). The samples are stretched to a given extension in a sample rack. The samples and the rack are
cooled to a low temperature and allowed to reach equilibrium. The samples are released from the rack and
heated at a specified rate of 1ºC per minute. Stretch is recorded over time, and the time at which the
sample recovers 10% of the applied stretch is recorded as TR10.
Among the flaws of this test is the fact that time and temperature are coupled via a specified
heating rate of 1ºC per minute. Again, with the temperature and shear rate dependent viscosity of polymer
liquids, it is impossible to separate the effects of time and temperature using this test. Performing the test at
a different heating rate would produce different results. This test does not account for the variations in
thermal expansion among various compounds; i.e. the strain recovery is based on both the material’s
elasticity and the fact that it expands as temperature increases. The TR test does not provide the design
engineer with data that can be used for applications under conditions different than experimental.
A more logical test might be the time it takes a given sample to recover 10% of its elongation
performed at a fixed temperature (elastic recovery rate). However, even this test does not take into account
the fact that it may have taken much more energy to deform one compound vs. another. That is, a 90
durometer compound requires more work input to deform to 50% elongation than does a 50 durometer
compound. Thus, there is likely (depending on viscous losses) more energy stored in the stiffer compound
than in the softer one.
This is basically a measure of viscosity as a function of temperature (elastic recovery rate), and
can be done with a rheometer.

6
Stiffening at Low Temperature, ASTM D1053
This method employs a composite torsion rod. One section of the rod is made of a “torsion wire”
whose torque vs. angular deflection curve is known. The other section of the rod is the rubber sample. The
assembly is cooled to the test temperature. Then the top end of the composite rod is rotated 180º, while the
other end is held stationary. The rotation at the interface is measured. From the relative rotation of the
wire, the torque can be calculated. The shear strain in the sample can also be calculated from the measured
rotation. The reference to time scale in this test is the time the technician waits to take a data point after the
head has been set to 180º rotation (i.e. 10 seconds). The standard does allow for data to be taken over time
as the sample remains at fixed temperature.
There are still many drawbacks to this test. First of all, the test is neither strain controlled nor
stress controlled. It approximates a creep experiment, but the load changes with the displacement. Also,
there is no mention of time dependence. Also, the composite rod is not necessary for this test; using grips
on the sample combined with a torque transducer, we can perform this test easily- eliminating the “middle
man” of the torsion rod. Furthermore, the test says “turn the torsion head quickly but smoothly 180º”; the
interpretation of this statement can have a profound effect on the measured results if the material is highly
viscous.

Figures 9a and 9b

Glass Transition Temperature: DSC Method, ASTM E1356

“DSC” stands for Differential Scanning Calorimeter. This type of device measures heat flow to a
sample and the sample’s temperature. As the instrument displays heat flow versus temperature, we can see
endotherms (as the sample absorbs heat energy) and exotherms (as the sample releases heat energy). As a
polymer becomes crystalline or “glassy”, it undergoes a change that is similar to a change of phase.
Consider water as an example: as water reaches 32ºF, and we continue to remove heat, the temperature of
the water remains constant throughout the phase change. A similar process occurs at the glass transition
temperature for polymers. However, this process is strongly dependent on the rate of cooling. For
example, the following chart shows glass transition temperature data from a DSC instrument for a VF2
dipolymer fluoroelastomer.

7
 Tg vs. HEATING RATE (DSC)

20

15

Tg (°C)
10
Tg (°F)

0
0 2 4 6 8 10 12

-5

-10

-15

-20

HEATING RATE (°/MIN)

Figure 10, Data from Table 5

As the cooling rate increases, the glass transition temperature appears to decrease. Recall that the bosses of
most lab technicians want high productivity from their technicians; the quickest way for a tech to get more
tests done in a day is to increase the cooling rate! Thus, many materials might appear to have adequate low
temperature properties at high cooling rates, while in actual applications the same material performs less
than adequately. Think of your automobile as an example. How fast does it really cool down in winter? It
is actually a relatively slow process. Thus it is irresponsible to report the glass transition temperature (as
calculated from DSC data) as a single temperature given the strong dependence on cooling rate. Also, there
is no indicator of how the elastomer will perform at temperatures close to, but above, the reported Tg.
Should I not use the material for a seal that will see a minimum temperature of 10º above Tg, or will the
seal function properly at 1º above Tg? This test/data does not have the capability to make such a
determination.

8
Glass Transition Temperature: DMRT Methods, ASTM D5992
DMRT stands for Dynamic Mechanical and Rheological Testing. ASTM D5992 describes tests
for determining glass transition temperature via DMRT and tan delta data. A frequency of oscillation is
chosen, and the temperature is ramped up from a temperature well below Tg. (i.e. the test uses temperature
sweeps at a given frequency). The point at which tan delta reaches a maximum is said to be the material’s
glass transition temperature. Once again, time and temperature are directly coupled via a specified heating
rate, and again time and temperature effects are indistinguishable.
ASTM D5992 also describes a method of determining Tg using DMRT results for E’’, or loss
modulus. The procedure is very similar to that for the tan delta method and has the same deficiencies.
Sample data are given below for the DMRT methods described in ASTM to show the test’s
dependence on frequency. The data is for a VF2 dipolymer fluoroelastomer.

Table 1

Tg via DMRT and Tan Delta

Frequency (Hz) Tg (°C) Tg (°F)
1 -20 -4
10 6.5 43.7
50 11 51.8

Table 2

Tg via DMRT and Loss Modulus, G’’

Frequency (Hz) Tg (°C) Tg (°F)
1 -8 17.6
10 -4.5 23.9
50 -2 28.4

Proper Use of DMRT

Let’s now discuss a method that addresses the fundamental questions listed at the start of this
paper. Consider a constant stress (creep) experiment. We can measure the time it takes a given material to
reach its equilibrium configuration. In other words, let’s say we hang a weight on a rubber band, and
measure the time it takes to reach its full deflection. At room temperature, the response is nearly
instantaneous, thus making time measurement a difficult task. At a very cold temperature, the time could
be extremely long, thus making the experiment very time consuming. In any case, a quick response time is
what the designer is looking for in an elastomeric seal material. For now, to limit the time it takes to
perform a test we will consider (as proposed by Ferry) the time it takes the specimen to reach 90% of its
equilibrium deflection and call this time t90 (as shown in Figure 11 below). For elastomers with short
relaxation times (i.e. fast recovery rates), measurement of t90 will be nearly impossible; for non-glassy
elastomers, this time is only a fraction of a second.

9
 log J(t)

log (t90 )
log (t)

Figure 11

Let us now consider an application in which a seal undergoes oscillating stresses/loads. The
response of a seal under such conditions can vary significantly with temperature and frequency. A good
guideline for oscillating loads is to limit the vibrations to frequencies below that which cause a 10%
increase in storage modulus. The frequency at which storage modulus increases by 10% is called f10 (as
shown in Figure 12 below).
log G‘

log (f10 )

log (f)

Figure 12

At a given temperature, the frequency at which the storage modulus, G’, increases 10% can be correlated to
t90 as suggested by Ferry with the following equation:

0 .2
Eq. 12 t90 (sec) ≈
f10 ( Hz )
Using a simple geometry, such as circular parallel plates (as shown in Figure 13) or torsion, we
can easily induce a definable strain field into a sample and conduct a test to find G’ versus frequency at any
specified temperature. Using torque transducers we can easily measure the material’s response to various
inputs.

10
 SAMPLE

Figure 13
Examples of data from this type of test are shown below [4],[5]. The data compares two elastomers and
their relaxation times at -25ºC (-13ºF) and 25ºC (77ºF). Using the data from the dynamic test, we can
conclude that the polybutadiene material performs equally well over the entire temperature range, while the
butyl rubber experiences a significant change in relaxation time as temperature drops. By performing more
frequency sweeps at temperatures between -25ºC and 25ºC, we can determine the suitability of the
elastomer for a given application.
Table 3 (References [4] and [5])
Temperature, 1,4 Polybutadiene Butyl Rubber
°C f 10 (Hz) t90 (sec) f 10 (Hz) t90 (sec)
-25 13 0.015 0.005 40
25 250 0.0008 5 0.04

FREQUENCY TO INCREASE G' BY 10% VS. TEMPERATURE

300

250

200

150 1,4 Polybutadiene

Butyl Rubber

100

50

0
-30 -20 -10 0 10 20 30

TEMPERATURE (°C)

Figure 14, Data from Table 3

11
 ESTIMATED t90 VS. TEMPERATURE

45

40 Seconds
40

35

30

25

20 1,4 Polybutadiene
Butyl Rubber
15

10

0.04 Seconds
5
0.015 Seconds
0
-30 -20 -10 0 10 20 30

TEMPERATURE (°C) 0.0008 Seconds

Figure 15, Data from Table 3

SUMMARY
In short, the various methods proposed by ASTM for evaluation of low temperature properties
apply only to a limited set of conditions. In fact, the data produced by each of these methods does not even
produce a unique value for glass transition temperature. Even if the methods showed agreement, there
would be no scientific justification to enable prediction of material behavior in real-world applications. A
summary of the data is presented below. All tests mentioned above were performed for the same material,
and the calculated glass transition temperature (and brittle point) is plotted.

12
 Tg FOR VF2 DIPOLYMER ELASTOMER BY VARIOUS ASTM METHODS

15

10

-5

-10

-15

-20

-25

Figure 16, Data from Tables 4 through 8

Estimating elastic recovery rate as proposed by Ferry [4], [5] is a fast and useful method for estimating
material performance at low temperature.

13
 Low Temperature Properties for a Common VF2 Dipolymer Fluoroelastomer
Table 4
TR10
Tg (°C) Tg (°F)
-17 1.4
Table 5
Tg via DSC

Heating Rate Tg (°C) Tg (°F)

(°C/min)
10 -15 5
5 -12 10.4
1 -9 15.8

Table 6
Tg via DMRT Tan Delta
Frequency (Hz) Tg (°C) Tg (°F)
1 -20 -4
10 6.5 43.7
50 11 51.8

Table 7
Tg via DMRT Loss Modulus
Frequency (Hz) Tg (°C) Tg (°F)
1 -8 17.6
10 -4.5 23.9
50 -2 28.4

Table 8
Brittle Point
Heating Rate TB (°C) Tg (°F)
(°C/min)
1 -20 -4

Table 9

Example from J. Ferry

1,4 Polybutadiene Butyl Rubber

Temperature, °C f 1.1 t0.9 f 1.1 t0.9
-25 13 0.015 0.005 40
25 250 0.0008 5 0.04

14
SYMBOLS AND UNITS
Symbol Description Units
(English, SI)
A Cross sectional area of a specimen in2, m2
A Amplitude of deflection in, m
c Damping Coefficient lbf-s/in, N-s/m
f Cyclic Frequency cycles/sec
F Force lbf, N
f Amplitude of force response to a sinusoidal displacement input lbf, N
G* Complex Shear Modulus PSI, Pa
G’ Storage Shear Modulus PSI, Pa
G’’ Loss Modulus, Viscous Modulus PSI, Pa
log Base 10 logarithm
Pmax Maximum measured load in a tensile test lbf, N
Su Ultimate Strength PSI, Pa
Tg Glass transition Temperature ºF, ºC
y Deflection of a spring, damper or system in, m
δ Phase angle between sinusoidal output and input curves rad
π Ratio of circle circumference to diameter; angular measure Unitless
corresponding to 180º
ω Angular Frequency rad/sec

Unit Description
Pa SI Unit of Pressure (Pascal)
s or sec Seconds; unit of time measurement
Hz Unit of frequency (cycles per second)
N SI Unit of Force (Newton)
lbf English Unit of Force (pound force)
s-1 Units of shear rate, inverse seconds
Pa-s Unit of Viscosity (Pascal seconds)
t Time (seconds)
º Unit of Angular Measurement (degrees)
ºF Unit of Temperature Measurement (degree Fahrenheit)
ºC Unit of Temperature Measurement (degree Celsius)

15
REFERENCES

[1] 1998 Annual Book of ASTM Standards. Sections 9.01 and 9.02. The following standards were
referenced:

D1053 Standard Test Methods for Rubber Property- Stiffening at Low Temperatures: Flexible Polymers
and Coated Fabrics
D1329 Standard Test Method for Evaluating Rubber Property- Retraction at Lower Temperatures (TR
Test)
D1414 Standard Test Methods for Rubber O-Rings
D2137 Rubber Property- Brittleness Point of Flexible Polymers and Coated Fabrics
D5992 Standard Guide for Dynamic Testing of Vulcanized Rubber and Rubber-Like Materials Using
Vibratory Methods

[2] “Engineering Design with Natural Rubber.” NR Technical Bulletin, 4th Edition. Malaysian Rubber
Producers’ Research Association (MRPRA), 1974. ISSN 09560-3856.

[3] Bird, R.B., R.C. Armstrong, and O. Hassager. Dynamics of Polymer Liquids. 2nd Edition. New York:
John Wiley & Sons, 1987.

[4] Ferry, John D. Letter to Daniel L. Hertz, Jr. 20 November, 1986.

[5] Gent, A.N. Engineering with Rubber. New York: Hanser Publishers, 1992.

[6] Kalyon, D. Polymer Rheology. Lecture. Stevens Institute of Technology. Hoboken, NJ, 1999.

[7] Stevens, R.D. and Albert L. Moore. “A New, Unique Viton® Fluoroelastomer With Expanded Fluids
Resistance.” Paper #32. Rubber Division, American Chemical Society. Cleveland, OH, October
21-24, 1997.

[7] Tobolsky, Arthur V., and Herman F. Mark. Polymer Science and Materials. New York: Wiley-
Interscience, 1971.

16