You are on page 1of 8

Tribology International 34 (2001) 461468

www.elsevier.com/locate/triboint

The temperature, pressure and time dependence of lubricant


viscosity
Scott Bair *, Jacek Jarzynski, Ward O. Winer
School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0405, USA
Received 15 November 2000; received in revised form 6 March 2001; accepted 4 May 2001

Abstract
The general form of the pressure (Proc. Am. Acad. Arts Sci. 77 (1949) 117) and temperature (Physical properties of molecular
crystals, liquids and glasses (1968) 350) dependence of viscosity has been known for at least 50 years. Viscosity varies with
temperature in a greater than exponential manner and temperatureviscosity equations generally allow for an unbounded viscosity
at some characteristic temperature. At high-pressures the pressureviscosity response is likewise greater than exponential, often
following a less than exponential response at low-pressures. In spite of this known behavior, tribologists working in EHL have
generally assumed less than exponential pressure response as a means of applying the Eyring stress aided thermal activation theory
to the viscous regime of EHL traction. As justification, time dependence of the lubricant properties in the response to a pressure
transient has been advanced. We present acoustic, capillary and impact measurements for timescales less than EHL. While time
dependence of properties may be important in the viscoelastic regime of traction, this paper will show that for the timescale of
viscous response, a significant time dependence of viscosity is unlikely. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Viscosity; Temperature; Pressure; Time; Acoustic; Capillary

1. Background
The rheology of liquid lubricants in the Hertzian zone
of concentrated contacts has been of principal interest to
tribologists for more than thirty years. The shear
response of the liquid within the pressure generating
fluid inlet zone has been seen to be of less interest since
the Newtonian assumption results, with some exceptions, in accurate calculations of film thickness. This is
apparently because the magnitude of the shear stress
within the inlet zone is relatively low. The Newtonian
assumption has been shown to fail spectacularly for the
calculation of traction, however.
High-pressure rheology research related to full film
traction has followed two paths. The path chosen by this
laboratory has been to measure shear response outside
of the contact using rheometers and compare predictions
based upon the generated data and subsequent analysis
with measured traction. The second approach removes

* Corresponding author. Fax: +1-404-894-8336.


E-mail address: scott.bair@me.gatech.edu (S. Bair).

the rather difficult step of constructing rheometers that


operate at high-pressure and utilizes the contact itself as
a rheometer. The quantity, which can be measured
within the contact, is the traction force which is an average of the local shear stress over the entire contact area.
Because the nature of a given family of traction curves
is relatively simple and because the underlying rheology
is quite complex, many simplifying assumptions must be
made, such as the nature of the constitutive behavior, in
order to produce a viable solution. For a short review of
various assumed models see ref. [3]. These models predict traction when the relevant properties are obtained
from the same traction measurements. One of the earliest
assumed models was that of time-dependent viscosity.
In a thoughtful 1967 paper, Fein [4] considered that
the low shear viscosity, m, was an exponential function
of fluid density. Harrison and Trachman [5] later used a
free volume model to relate viscosity to volume in similar calculations. Changes in volume due to changes in
pressure were naturally retarded by volume viscoelasticity. The total volume change, or compression, was separated into an instantaneous or solid-like compression and
a time-dependent or relaxational compression. Thus, for

0301-679X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 0 1 - 6 7 9 X ( 0 1 ) 0 0 0 4 2 - 1

462

S. Bair et al. / Tribology International 34 (2001) 461468

Nomenclature
A1
A2
B1
B2
B
C1
C2
c
cp
cv
D
Ks
Ks
Kso
KT
KT
KTO
L
p
tc
ts
to
T
Tg
V
vocc
v
vo
z
g
m
mg
mo
r
t

Yasutomi parameter
Yasutomi parameter
Yasutomi parameter
Yasutomi parameter
Dolittle parameter
WLF constant
WLF constant
sound velocity
constant pressure specific heat
constant volume specific heat
capillary inside diameter
adiabatic bulk modulus
pressure rate of change of Ks
Ks at p=0
Isothermal bulk modulus
pressure rate of change of KT
KT at p=0
capillary length
pressure
compression time
time of shearing
observation time
temperature
glass transition temperature
average velocity
occupied volume
volume
volume for p=0
Roelands parameter
shear rate
limiting low shear viscosity
viscosity at the glass transition
viscosity for p=0
mass density
shear stress

short time applications of pressure, the volume change


would be less than the equilibrium value. Fein introduced the fictive pressure, that is, the pressure which
at equilibrium would produce the same compression. For
very slow compression, the fictive pressure becomes
essentially the glass transition pressure. Measurements
of viscosity by cooling across the glass transition by
Kobayashi et al. [6] and by compression across the glass
transition from this laboratory [3] indicate that changes
in viscosity with temperature and pressure are retarded
at the glass transition.
The idea that a time-dependent viscosity might be
important for viscous response was given support by
some in-contact measurements by Paul and Cameron [7]
in 1972. They measured leakage rates from a static contact entrapment. The viscosity deduced from these

measurements was seen to vary with time for minutes


even hours. These results, however, were quickly refuted
when Hutton and Phillips [8] conducted the viscosity
measurements out of contact with a pressurized Couette
viscometer. No compression time dependence was
observed. Wong et al. repeated the Paul and Cameron
measurements later [9] and found that the apparent time
dependence of viscosity was the result of time varying
shear rate in the flow from the contact and non-Newtonian viscous response.
Fein recognized that the ratio of volume (bulk) viscosity to the relaxational component of the bulk modulus, the structural relaxation time, is the same as the ratio
of shear viscosity to limiting shear modulus, the mechanical shear relaxation time. When the structural and shear
relaxation times are equal, then time dependence of vis-

S. Bair et al. / Tribology International 34 (2001) 461468

cosity should only occur for the time scales of shear viscoelastic response.
Time dependent properties are once again being
advanced to account for viscous response in contacts. By
1986, tribologists using the contact as a rheometer had
generally agreed upon the ReeEyring model as a
description of non-linear viscous response [10] as
opposed to the power-law behavior ordinarily observed
with rheometers [11:107]. This model is attractive
because it naturally embodies the sinh-law shape of
many traction curves. However, in order to retain this
shape, when integrated over the contact area with a
pressure-dependent viscosity, the viscosity must not rise
too rapidly with pressure. Also, if the only shear-thinning behavior admitted is Eyring, then the particular
horizontal positioning of individual traction curves for
different loads must be due to less-than-exponential
pressureviscosity response. This requirement is in contradiction with viscometer measurements of viscosity,
which have for at least fifty years shown greater-thanexponential pressureviscosity response at high-pressure
[1]. As justification for the use of viscous properties
which cannot be obtained with a viscometer, some
researchers applying the Eyring model to in-contact
measurements have once again invoked time dependent
behavior for viscous traction response [12,13]. These
researchers argue that steady-state measurements may
not yield the same properties as a short time contact
measurement and use the Evans and Johnson model [10]
for analysis, although Evans and Johnson did not
embrace the time-dependence argument.
With this paper we present measurements of the variation of viscosity with temperature and pressure for both
very short times of shear and compression and long
times of shear and compression.
2. Variation of viscosity with pressure
The most convenient instrument for the measurement
of the limiting low shear viscosity, hereafter simply viscosity, is the falling body viscometer with LVDT sinker
detection [14]. All of the falling body viscometers
presently used in this laboratory reset the sinker by
inverting the viscometer. An alternative approach has
been to lift the sinker with a solenoid immersed in the
liquid sample. By simply inverting the viscometer, the
electrical heating of the sample by the solenoid is avoided. We have demonstrated 1.6 GPa and 200C capabilities. Pressure has been calibrated against a Harwood
manganin cell to 1.3 GPa. The measurement of viscosity
under pressure is a well-developed field, with excellent
repeatability between laboratories. See Fig. 1 for
measurements of pressureviscosity response of a silicone oil by three laboratories [1,15]. Although the data
of King et al. [15] have significantly greater scatter, their
diamond anvil technique can achieve pressure to 10 GPa.

Fig. 1.
tories.

463

The viscosity of a silicone oil as measured by three labora-

Examples of measurements on formulated lubricants


are shown in Fig. 2 for a Jet II oil, L23699, and a traction
fluid used in a production toroidal CVT. For the jet lube,
less than exponential behavior persists to 0.5 to 1 GPa
depending upon temperature. For the traction fluid, the
less than exponential portion is confined to no more than
a wiggle below 0.2 GPa. Both liquids clearly display
greater-than-exponential behavior at all temperatures
except for the jet lube at 165C. We expect that at higher
pressures the greater than exponential response would
appear. Note that the local pressureviscosity coefficient
at high pressure may reach or exceed the local value at
ambient pressure.

3. Variation of viscosity with temperature


The viscosity of the liquids described above is plotted
against temperature for two elevated pressures each in
Fig. 3. The local temperatureviscosity coefficient
reaches very large values, as much as 0.25C1, when
the viscosity is high. Temperatureviscosity relations for
supercooled liquids must allow for the viscosity to be
unbounded at some temperature [2:350]. This singularity
accounts for the greater-than-exponential pressure vis-

464

S. Bair et al. / Tribology International 34 (2001) 461468

ation introduced by Yasutomi et al. [16]. The low-shear


viscosity is obtained from
mmgexp

2.3C1(TTg)F
C2+(TTg)F

(1)

where the glass transition temperature is


TgTgoA1ln(1A2p)

(2)

and the relative free volume expansivity is


F1B1ln(1B2p)

(3)

and A1, A2, B1, B2, C1, C2 and Tgo are parameters to be
evaluated. The parameters were regressed from the viscosity data and are listed in Table 1 for the liquids in
Figs. 2 and 3.
Note that these parameters are uniquely determined
only if there is data on each side of the pressure-log
viscosity inflection.

5. Time dependence of the limiting stress

Fig. 2. The variation of viscosity with pressure for a jet lube and a
traction fluid.

Fig. 3. The variation of viscosity with temperature at indicated pressures for the liquids of Fig. 1.

cosity response and the large values for temperature


viscosity coefficients.

Non-linear viscous response is of primary interest to


the study of EHL. In a series of experiments utilizing
the split Kolsky bar and pressure-shear impact, Ramesh
[17] has examined the influence of time on measurements of non-linear constitutive behavior. In early work
of this type the temperature increase associated with the
deformation made interpretation as constitutive behavior
impossible. However, for impact at moderate pressure
the temperature rise is small and comparison can be
made with isothermal rheometer measurements.
In Fig. 4, we compare measurements obtained with a
pressurized Couette viscometer [18] (circles) to the
pressure-shear impact measurement of ref. [17] (pluses).
For the Couette viscometer, the time required to attain
the given pressure, tc, is of the order of 100 seconds and
the stress and shear rate reaches steady values in about
ts=10 ms. For the impact test, compression is complete
in 1 microsecond and shearing occurs for even shorter
time. Clearly, these measurements encompass the time
scale of EHL traction rigs (105 to 103 seconds). However, the impact measurement fits the flow curve defined
by the longer time measurement.
We now are aware that limiting stress behavior is not
constitutive [19] as it results from cohesive failure or
slip within the liquid. But this behavior is clearly not
time dependent on the EHL time scale.

6. Viscosity for small time of shear


4. Free volume model
A viscosity model that can describe the temperature
and pressure response is the pressure modified WLF equ-

The effect of observation time upon viscosity can be


conveniently investigated with a capillary viscometer
[20,21]. In this instrument, pressure is generated with

S. Bair et al. / Tribology International 34 (2001) 461468

465

Table 1
Yasutomi parameters
Fluid

mg (Pas)

Tgo (C)

A1 (C)

A2 (GPa1)

B1

B2 (GPa1)

C1

C2 (C)

L23699
NissanCVT

1012
107

87.0
68.6

158
160.5

0.4476
1.816

0.194
0.3024

18.8
12.64

16.03
10.94

22.52
28.97

lary measurements asymptotically approach the falling


body results as wall shear rate, g N, is reduced. For isothermal Newtonian capillary flow, the average fluid
velocity is
Vg ND/8

(4)

for an inside diameter, D. If the corrected capillary


length is L, then the average shearing time is
toL/V
or
to
Fig. 4.
ments.

8 L
g N D

(5)

(6)

Flow curve comparing short time with long time measure-

two intensifiers placed in series. When the intensifier


pressure vessel is moved with respect to the two
intensifier pistons, liquid must flow from one intensifier
chamber to the other chamber. These chambers are connected by high-pressure tubing to a capillary. The pressures on both ends of the capillary are measured by
two transducers.
In Fig. 5, we present the viscosity of a polyalphaolefin
at two average pressures and one initial temperature.
Results for a falling body viscometer for an observation
(shearing) time of 1 to 3 s are plotted as circles. Also,
in Fig. 5 are results for a short capillary, plotted as
points, and a long capillary, plotted as pluses. The capil-

Referring to Fig. 5, the same curve has been drawn


through all four sets of capillary data. They are shifted
along the shear rate axis by about one order-of-magnitude, indicating that the non-linearity is not constitutive.
That is, the non-linearity is not related to shear rate (or
stress). Scales for observation time, to, are provided at
the top of Fig. 5. For each pressure, the curves are shifted
by two orders-of-magnitude of time, indicating that one
obtains the same response for differing observation times
down to 3 s.
The non-linearity may be explained by considering
that the viscous work done, on average on an element
of fluid, is proportional to mg 2Nto. The viscosity, m, is
initially the same for the two capillaries. The shift of
two orders-of-magnitude in time is compensated by the
shift of one order in shear rate so that corresponding
points on the pairs of curves represent the same amount
of viscous work. The non-linearity results from heating
and the viscous response is unchanged for observation
times of 3 s to 3 s.

7. Calculation of viscosity from very short time


compression

Fig. 5.

The viscosity of a PAO for various times and shear rates.

Measurements of the viscosity of a mineral oil, HVI


650 (Shell Research), are presented in Fig. 6. Four
measurements (xs) were made with a falling body viscometer at pressure less than 150 MPa [14]. These are
listed in Table 2 as well. Three measurements (pluses)
were performed with a Couette viscometer [8] at pressures above 300 MPa. From the low pressure, falling

466

S. Bair et al. / Tribology International 34 (2001) 461468

ments are listed in Table 2 as the ratio of volume at


elevated pressure, v, to the volume at atmospheric pressure, vo. A bellows type of piezometer was employed as
shown in Fig. 7. The change in length of a calibrated
bellows is measured with a potentiometer. To improve
the sensitivity of the measurement, the bellows is connected to a reservoir or bulb as shown in Fig. 7. To
demonstrate the accuracy of this instrument, we have
repeated the volume measurements of Bridgman
[23:128130] to within 0.3% for both water and 1-propanol over a range of temperature of 75C and pressure to
382 MPa.
The basic equation of Free Volume Theory is the Dolittle equation [16]:

mmoexp B

Fig. 6. The viscosity of a mineral oil, measured and calculated for


long and short times.
Table 2
Relative volume and viscosity of HVI 650 at 20C
p/MPa
0.1
34.5
69
146
241
310
345
393
441

v/vo
1
0.980
0.966
0.941
0.919
0.904

0.889
0.881

m/Pas
1.96
6.8
17.5
124

23800
91800
390000

body data we regressed the Roelands parameter,


z=0.592. Extrapolation with the Roelands equation [22]
is shown in Fig. 6 as the broken curve. The Roelands
equation predicts less than exponential response at all
pressures for this liquid. Less than exponential response
was obtained by Evans and Johnson [10], as well, from
traction measurements using the Eyring assumption (see
Fig. 6). However, the Evans and Johnson viscosity is
much greater than either the viscometer results or the
Roelands extrapolation.
We have also measured the volume of HVI650 as a
function of pressure. Some of these volume measure-

vocc
1
1

vocc
vo v Vocc

1
vo vo
vo

(7)

where vocc is occupied volume and B is a dimensionless


parameter. Utilizing the falling body viscosity data and
the relative volume data from Table 2 for p150 MPa,
we obtain B=4.260 and vocc=0.754 vo. The viscosity was
then calculated for p200 MPa from Eq. (7) and the
relative volumes obtained at the higher pressures with
the piezometer. See Table 2. The resulting calculated
viscosities are plotted in Fig. 6 as circles. Clearly, Free
Volume Theory correctly accounts for pressureviscosity behavior.
Ultrasonics can be used to probe the response of
liquids to very fast pressure changes. We have measured
the longitudinal sound velocity of HVI 650 using a highpressure acoustic interferometer (see Fig. 8). Two quartz
tranducing discs are separated by 3 cm of liquid under
pressure. A two period longitudinal wave is launched
from one disc and the time of flight to the opposite disc
is noted. The sound velocity is computed from the time
of flight. We have repeated the sound velocity measurements of Hagelberg [24] to within 1.5% for 1-propanol
at pressures to 400 MPa.
The measured sound velocity, c, for HVI650 is tabulated in Table 3 along with densities, r determined with
the piezometer. The adiabatic bulk modulus is calculated from
Ksrc2

(8)

and is listed in Table 3 and plotted against pressure in


Fig. 9.
Note that the plot of rc2 is roughly linear with pressure and can be represented by
KsKsoKsp

(9)

where Kso=1.89 GPa and Ks=11.3. Volume viscoelastic-

S. Bair et al. / Tribology International 34 (2001) 461468

467

Fig. 7. High-pressure piezometer.

Fig. 8.

High-pressure acoustic interferometer.

Table 3
Longitudinal sound velocity and density of HVI 650 at 20C
Frequency

p/MPa

c/m/s

r/kg/m3

Ks/GPa

1.4 MHz

0.1
34.5
362
393
428
0.1
34.5
69
103
138
172
207
241
310

1442
1596
2450
2500
2586
1451
1592
1701
1799
1889
2000
2089
2173
2351

894
912
999
1006
1011
894
912
926
937
947
956
965
973
989

1.86
2.32
6.00
6.29
6.76
1.88
2.31
2.68
3.03
3.38
3.82
4.21
4.59
5.47

0.5 MHz

Fig. 9. Ultrasonic measurement of the adiabatic bulk modulus of the


mineral oil of Fig. 5.

ity causes dispersion in longitudinal sound velocity


where c becomes a function of frequency and a plot of
rc2 turns upward with increasing pressure when structural relaxation is important. This upward curvature is
absent from Fig. 9. In fact this relatively low frequency is utilized in this laboratory to avoid dispersion

when measuring the adiabatic bulk modulus, Ks, at high


pressure. Comparing Figs. 6 and 9 shows that bulk viscoelastic response is absent to a shear viscosity of 105
Pas for times as short as 106 s.
We can calculate the viscosity of HVI 650 for very

468

S. Bair et al. / Tribology International 34 (2001) 461468

fast pressure transients using the volume obtained by


integration of Eq. (9) and the resulting volume, v,
inserted into Eq. (7). The ratio of compressibilities is
identical to the ratio of specific heats, cp/cv. Then the
isothermal bulk modulus is given by
KTKscv/cpKTOpKT

(10)

where KTo=Kso cv/cp and KT=Kscv/cp. Eq. (10) is the


Murnaghan equation of state [25] which in the integrated
form reads as

v/voexp

1
ln[(KTOpKT)/KTO]
KT

(11)

We have plotted viscosity calculated from Eq. (7) using


Eqs. (10) and (11) and the high frequency values of Kso
and Ks for two common values of the ratio of specific
heats. These are the two solid curves for cp/cv=1.10 and
1.15 in Fig. 6. From differentiation of a smooth curve
through the isothermal volumes obtained with the
piezometer we found on average that Ks/KT=1.1. The
curve in Fig. 6 for cp/cv=1.10 is an excellent representation of the viscosity measured with our viscometer
including the greater than exponential behavior. There
is no reasonable value of cp/cv which will produce either
the Roelands extrapolation or the Evans and Johnson
curve in Fig. 6.

8. Conclusion
For viscosity of the magnitude present in the Hertz
zone of EHL films for the viscous regime of traction,
the pressureviscosity behavior at pressures relevant to
traction generation is often greater than exponential.
Times of shear and times of compression characteristic
of the EHL transit time have no significant effect upon
this behavior.

Acknowledgements
This work was supported by a grant from the
Timken Company.

References
[1] Bridgman PW. Viscosities to 30,000 kg/cm2. Proc Am Acad Arts
Sci 1949;77:11728.

[2] Bondi A. Physical properties of molecular crystals, liquids, and


glasses. Wiley, 1968.
[3] Bair S, Winer WO. The high-pressure, high shear stress rheology
of liquid lubricants. ASME J of Trib 1992;114(1):12.
[4] Fein RS. Possible role of compressional viscoelasticity in concentrated contact lubrication. ASME J of Lubr Tech Ser F
1967;89(2):127.
[5] Harrison G, Trachman EG. The role of compressional viscoelasticity in the lubrication of rolling contacts. ASME J Lubr Tech
1972;94:30612.
[6] Kobayashi H, Takahashi H, Hiki Y. Viscosity of glass polymers
near and below the glass transition temperature. XIIIth Int. Congress on Rheology, 2000:491492.
[7] Paul GR, Cameron A. An absolute high-pressure microviscometer based on refractive index. Proc Royal Soc of London
1972;A331:17184.
[8] Hutton JF, Phillips MC. High pressure viscosity of a polyphenyl
ether measured with a new couette viscometer. Nature Physical
Sci 1973;245:16.
[9] Wong PL, Lingard S, Cameron A. The high pressure impact
microviscometer. STLE Trib Trans 1992;35(3):500.
[10] Evans CR, Johnson KL. The rheological properties of elastohydrodynamic lubricants. Proc Instn Mech Engrs Part C
1986;200(C5):305.
[11] Bird RB, Armstrong RC, Hassager O. Dynamics of polymeric
liquids, Vol. 1, Fluid mechanics. New York: Wiley, 1987.
[12] Fang N, Chang L, Johnston GJ, Webster MN, Jackson A. An
experimental/theoretical approach to modeling the viscous
behavior of liquid lubricants under EHL conditions. Proc. 27th
Leeds-Lyon Symp. Tribology, 2000.
[13] Olver AV, Spikes HA. Prediction of traction in elastohydrodynamic contacts. Proc Instn Mech Engrs 1998;212(J5):322.
[14] Bair S. An experimental verification of the significance of the
reciprocal asymptotic isoviscous pressure. ASLE Trib Trans
1992;36(2):1545.
[15] King HE, Herbolzheimer E, Cook RL. The diamond anvil cell
as a high-pressure viscometer. J of Applied Physics
1992;71(5):2078.
[16] Yasutomi S, Bair S, Winer WO. An application of a free volume
model to lubricant rheology. ASME J of Trib
1984;106(2):291303.
[17] Zhang Y, Ramesh KT. The behavior of an EHD lubricant at moderate pressures and high shear rates. ASME J of Trib
1995;117(4):6.
[18] Bair S, Winer WO. A new high-pressure, high-shear stress viscometer and results for lubricants. Trib Trans 1993;36:7215.
[19] Bair S, Winer WO. A rheological basis for concentrated contact
friction. Proc. LeedsLyon Symp., 1994:3744.
[20] Novak JD. An experimental investigation of the combined effects
of pressure, temperature, and shear stress upon viscosity, PhD
dissertation, The University of Michigan, 1968:14.
[21] Jakobsen J. Lubricant rheology at high shear stress, PhD thesis,
Georgia Institute of Technology, 1973:174175.
[22] Roelands C. Correlational aspects of the viscositytemperature
pressure relationship of lubricating oils. University Microfilms,
Ann Arbor, MI, 1966.
[23] Bridgman PW. The physics of high pressure. New York:
Dover, 1970.
[24] Hagelberg PM. Ultrasonic velocity measurements and B/A for
1-propanol at pressures to 10,000 kg/cm3. J Acoustical Soc of
Am 1970;47:158.
[25] Davis LA, Gordon RB. Compression of mercury at high pressure.
J of Chemical Physics 1967;46(7):2659.

You might also like