Вы находитесь на странице: 1из 10

367

Tribology Letters, Vol. 17, No. 3, October 2004 ( 2004)

Lubrication of silicon nitride and silicon carbide by water: Running


in, wear and operation of sliding bearings
Laurent Jordi, Christo Iliev, and Traugott E. Fischer*
Stevens Institute of Technology, Hoboken, NJ 07030, USA

Received 21 December 2003; accepted 25 February 2004

The purpose of this work was to establish the conditions for the operation and break-in of water-lubricated ceramic bearings. The experiments consisted of sliding 1 400 silicon nitride orcarbide balls against pre-polished disks of the same material
in water until tribochemical wear generates smooth conformal surfaces that allow hydrodynamic lubrication (l<0.002) by very
thin water lms. This running in was performed at various sliding speeds (0.014 m/s) and loads (0.520 N). The minimum
sliding speed for low friction were 0.04 m/s for silicon nitride and 0.5 m/s for silicon carbide, much lower than for conventional
bearings. The load carrying pressures were 6080 MPa, which is higher than the usually pressures of thrust bearings. The
hydrodynamic uid lm thickness was estimated with a standard integration of Reynolds equations modied for circular geometry, it was to be 515 nm for silicon nitride, 25 nm for silicon carbide. Operation over long distances (80 km) allowed us to
measure the wear rate during hydrodynamic lubrication; this was found to be <2 10)11 mm3/nm, a rate acceptable for industrial application. A novel method completed during this work allows the determination of the wear rate during run-in. It varies
with sliding velocity for silicon nitride, from 1 to 6 10)5 mm3/nm; it is constant at 4 10)6 mm3/nm for silicon carbide.
KEY WORDS: water lubrication, ceramics, running in, wear measurement, tribochemistry

1. Introduction
In 2001, Michael Gardos, at DARPA, set up a program to explore the feasibility of small, portable,
steam engines. Within this program, he decided to
investigate the possibility of lubricating bearings and
the piston-and-cylinder system with water. Earlier
work [1] had shown that silicon nitride and silicon carbide, when sliding on each other in water, obtain an
extremely smooth surface that allows hydrodynamic
lubrication at moderate velocities with a friction coecient smaller than 0.02. This phenomenon was later
studied with the aim of developing a polishing
technique [24]. Kato and coworkers have further
investigated the lubrication of ceramics by water
[510]; they have measured a friction coecient as low
as l = 0.001 [9] and found that texturing the surface
with shallow pits increases the load carrying ability of
the bearings [10].
After more than fteen years scientists are still
divided on the lubrication mechanisms responsible for
the low friction. Kato et al. [7] and Hsu [11] propose
that the combination of boundary lubrication by products of the tribochemical reaction and hydrodynamic
lubrication is responsible for the low friction. Fischer
et al. [1,12] state that the low friction is achieved by
hydrodynamic lubrication alone and that the water
*To whom correspondence should be addressed.
E-mail: tscher@stevens.edu

lm sustains the whole weight. This latter claim was


based on measurements of the Stribeck curve: when
the sliding velocity is decreased, the friction coecient
reverts to a high value of l=0.7 at very low speed. If
a lubricating layer were present, low friction would be
maintained at very low velocity. In situ FT-IR measurements [13] on the surface of silicon nitride sliding
in water and Auger spectroscopy studies of silicon
nitride surfaces polished in water [3] established the
absence of a reaction product layer on these surfaces.
Since hydrodynamic lubrication by the low-viscosity
water demands very thin uid lms and very smooth
surfaces, running in of the bearing is important. Kato
and coworkers have studied the dependence of the
run-in period on the roughness of the surfaces [6].

2. Objectives
Despite the previous work, a number of issues
remain outstanding. The rst is the question of wear
during hydrodynamic lubrication. This, in theory, is
zero but it must be veried experimentally because
some dissolution of the ceramic in water may occur
even in the absence of solid body contact. Another
important question is the range of sliding speeds and
loads at which wear (i.e., run-in) is purely tribochemical. It is conceivable that the chemical reaction is not
rapid enough at high sliding velocities or at high contact pressure and that mechanical wear mechanisms
1023-8883/04/10000367/0  2004 Springer Science+Business Media, Inc.

368

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

roughen the surface. Finally, one needs to establish the


highest load (or bearing pressure) such a bearing can
carry. The objectives of this investigation were therefore:
 to determine the operating limits of tribochemical
run-in of water lubricated bearings,
 to determine the wear rate of silicon nitride during
hydrodynamic lubrication,
 to determine the limits of velocity and load for
hydrodynamic lubrication by water,
 to rene the method for measurement of wear in
mixed lubrication.
We also studied the operation of the bearing, estimated the thickness of the hydrodynamic water lm
and veried that the expected hydrodynamic friction
coecient corresponds to the measured values.

the wear scar diameter on the ball at the end of the


run. Since we used a very simple tribometer, vibrations
at velocities above 1.5 m/s render the experimental less
reliable at these velocities.
Wear of the silicon nitride ball produces a at
round wear scar of radius q. The latter is measured
with an optical microscope. The wear volume is
approximated as
VW

pq4
;
4r

where q is the scar radius and r is the radius of the


ball (3175 mm). The wear of the disc is obtained from
a prolometer trace of the wear track. Wear volumes
as small as 2 10)6 mm3 can be measured on the ball
with reasonable precision (5%); the detection limit of
wear on the plate by prolometry is about 10)4 mm3
when the experiment is performed on a well-polished
at plate.

3. The experiments
The experiments were designed for the automatic
formation of a hydrodynamic bearing by running in.
A stationary silicon nitride or carbide ball slides
against a previously polished plate of the same material, at a chosen velocity and load. Initially, the contact is Hertzian; the contact area is too small and the
contact pressure is too high for hydrodynamic lubrication, therefore friction and wear rate are high. Wear is
tribochemical and creates a smooth, at, surface on
the ball so that a at-on-at geometry is quite rapidly
established. When the area of this wear scar is large
enough, hydrodynamic lubrication is established and
the friction coecient drops to very low values
(l < 0.02). This constitutes the running in of the system. The experiment thus consists of recording the
evolution of the friction coecient with time, from
boundary to mixed to hydrodynamic lubrication.
Observation of the wear scar on the ball and the wear
track on the disk allow us to measure the amount of
wear caused by run-in.

4. The tribometer
The tribometer used is ball on disc type where a stationary, non-rotating, ball is loaded with a dead
weight against a rotating disk. Deionized water is
poured on the ball by gravity from a reservoir and
ows into the contact. This insures good ooding
despite a rotational speed up to 4000 rpm. The friction
force is measured by a strain gauge and is recorded
continuously. The gauge is calibrated with dead
weights.
In each experiment the chosen variables are the load
and the velocity. The data obtained with this setup are
the evolution of the time-dependent friction force and

5. The samples
Silicon Nitride balls were commercial NBD200
bearing balls obtained from Cerbec, Saint-Gobain.
Their diameter was 6.35 mm (1/400 ). The ats were
rolls of silicon nitride sintered with isostatic pressure
and supplied by Ceradyne Inc; NBD200 discs obtained
from Saint-Gobain and TSN-03NH grade plates from
Toshiba.
Silicon carbide samples were polished balls, 8 mm
in diameter, manufactured by Koyo Corporation, and
graciously donated by Professor Kato. The plates were
hot pressed Hexoloy from Carborundum Corporation.
Just before the experiments, the balls were washed
with soap, rinsed rst with water, with acetone in an
ultrasonic bath during 10 min and rinsed again with
deionized water. One ball oers a fresh surface for several experiments when rotated on itself.
The plates were fastened on the sample holder with
thermoplastic glue; they were polished, rst with diamond abrasive and nished by tribochemical polishing
[2,3] to start experiments with a smooth surface.
Tribochemical polishing produced plate surfaces with
a roughness Ra < 5 nm for silicon nitride [3] and
25 nm for silicon carbide [14]. The plates were washed
with the same procedure as the balls.

6. Estimation of the water lm thickness


Hydrodynamic lubrication being essential to our
problem, we need to have information on the thickness
of the water lms that separate the two surfaces and
how it depends on the operating variables of the bearing, namely the sliding velocity and the load carried.
Measuring the thickness of water lms in the order of

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

10 nm would present a formidable challenge. The


thickness can be estimated by integration of Reynolds
equation for a thrust bearing formed of two surfaces
that make a very small angle with each other so that
their separation h1 at the entrance of the liquid is larger than the minimum separation h0 at the uid exit
end. Here we follow the treatment of Cameron [15]
who calculates the load P carried by a rectangular
thrust bearing of length B and width L, sliding at
velocity U, in a uid with viscosity g
P

gULB2
6Wk;
h2o

W is a dimensionless function of the attitude parameter k, dened as k = (h1)h0)/h0. The function 6W(k) is
tabulated in Cameron and has values between 0.14
and 0.16 when k varies from 0.6 to 2.5.
We do not know of an equation for a round thrust
bearing with diameter D = 2q. We use the following
approximation to calculate the load bearing force of
such a bearing. Using polar coordinates, we divide the
circle into a number of parallel strips of width DL =
q sin a and corresponding length B = 2q cos a. a = 0
is chosen along the sliding direction. For each strip, i.e.,
for each value of a, h0 and k are dierent from their values at a = 0. We calculate the load carried by each strip
using equation (2), inserting the proper values of DL (a),
B(a), h0 (a) and k (a). Finally, we add the loads carried
by all the strips and obtain equation (3) for the circular
thrust bearing. Since the diameter of the bearing D is
the equivalent of the length L and width B of (1):

369

The gure shows that the circular bearing carries


about 60% of the weight a square bearing the same
size (L = B = D) or 75% of the square bearing of
the same area. It is not possible to measure the attitude parameter k. Actual thrust bearings adopt their
own attitude k because their pads are articulated so
that the bearings nd their own angle. (The latter is
extremely small, around 10)5 radian or 2 of arc). In
our experiments, we believe that the angle is formed
by elastic deformation of the instrument: at the beginning of the experiment, the high friction force deforms
the sample holder very slightly; elastic recovery at low
friction straightens the holder and forms the angle.
Considering gure 2, we adopt the approximations
that 0.6<k<2 which gives us 6WC = 0.09 0.005.
With this value of 6WC, the estimation of the water
lm thickness becomes
r
gUD3
h0 0:3
:
P

The load P and the sliding velocity U are selected, the


viscosity of the water is known to be 1 cP = 10)3 Pas,
the diameter D of the wear scar is measured after run-in.
These water lm thicknesses will be shown in gure 10
when we present the results. We note that our arbitrary
range of k gives us an upper limit to the lm thickness.
Figure 2 shows that k values outside this range would
result in smaller estimated h0.
7. Results
7.1. Running in

gUD3
P
6WC k:
h20

Here 6WC is the factor equivalent to the one of equation (2), but recalculated for the circular bearing.
Figure 1 compares the factors 6W for rectangular and
round disk bearings.
0.18
0.16
0.14

As described above, running in starts with a round


ball sliding on the at with a Hertzian contact. Initially, friction is high (l = 0.7 for Si3N4 and 0.35 for
SiC). During run-in, wear of the ball forms a at wear
scar with an ultra smooth surface. The diameter D of
the at gradually increases and so does the load carrying force P (equation (3)) and friction decreases. This
is the regime of mixed lubrication. When the diameter is large enough that the whole load is carried
by the hydrodynamic forces and the minimum lm

6W*

0.12
0.1
0.08
0.06
0.04
0.02
0
0

2
K

Figure 1. Factors 6W* for a rectangular (equation 2, dotted line)


and a circular thrust bearing (equation 3 full line).

Figure 2. Evolution of the friction coecient during running-in of a


silicon nitride bearing. In this example, the run-in time is 8 s.

370

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

thickness h0 is larger than the combined roughness of


the two surfaces, we have hydrodynamic lubrication
with l<0.02. Figure 2 shows an example of the evolution of friction during run-in. Run-in time is 8 s in the
example of gure 2.
Figures 3 and 4 show the topography of the wear
scars for silicon nitride and silicon carbide after running in. At left the surface has undergone tribochemical wear through 55 km distance at 1 m/s velocity
under 4 N load. For comparison, the surface at right
was worn by mechanical mechanisms when sliding,
unlubricated, in air. The tribochemically run-in surface is very smooth. (The large particle is dust deposited after the test.) A high-resolution picture, bottom
left, shows that the surface is at with some hills
and valleys. The oblique electron beam incidence
(45) permits the estimation of the roughness, which
is about 10 nm. The holes in the surface do not
impede hydrodynamic lubrication; in fact, Kato et al.
[10] have shown that they increase the load carrying
ability. Figure 4 shows the wear scar of silicon carbide. This material is more porous than silicon

nitride and more brittle. In keeping with a previous


report by Zhu et al. [14] one can observe that some
grains stand slightly proud of the others; some grains
polish more slowly than others because of the
mechanical and chemical anisotropy of hexagonal
SiC. In addition, we observe that the inclusions
(carbon) do not wear tribochemically and therefore
contribute to the roughness of the surface, which is
about 25 nm.
Figure 5 shows run-in times for silicon nitride as a
function of velocity at 2 N load. In the case of silicon
carbide, the run-in times are somewhat longer because
its wear rate is lower and its surface is rougher and
demands a larger scar diameter. Figure 6 shows the
running-in distance, which as obtained from gure 5
by multiplying the time by the velocity. It rst shows a
rapid decrease with increasing velocity because a lower
bearing area is needed at high sliding speeds (equation
(4)). At velocities above 1.5 m/s, the running-in distance is essentially constant, most probably because
vibrations in the tribometer require a thicker uid
lms and consequently larger bearing diameters. The

Figure 3. Scanning electron micrographs of wear scars on silicon nitride balls. Top left: overview of a tribochemically worn-in scar. Top right,
a mechanically worn scar, for comparison. Bottom left: the tribochemically worn-in silicon nitride ball at 50,000X magnication, Bottom
right: mechanically worn silicon nitride ball at 2000X magnication.

371

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

100

Run-in time [s ]

Roll 1
Roll 2
Disc

10

2
Velocity [m/s]

Figure 5. Running-in time of a silicon nitride bearing as function of


the velocity at 2 N load.

35

Roll 1
Roll 2
Disc

R u n - i n d i s t a nc e [ m ]

30
25
20
15
10
5
Figure 4. High-resolution SEM micrographs of tribochemically
worn-in at on Sic ball. Top: 5000X magnication; upper half is the
wear scar, bottom half is a cut through the material. Bottom:
20,000X magnication of the wear scar.

gures show that run-in at 1 m/s velocity can be


obtained in 2 s at a distance of about 2 m when the
surfaces are polished. Previous investigations have
indicated quite long run-in times [1,6] because the
starting surfaces had a roughness around 2 lm. In the
manufacture of real machines, bearing surfaces are
polished; reliance on tribochemical polishing during
run-in would produce insucient precision.

7.2. Operating variables of water-lubricated ceramics


7.2.1. Hydrodynamic lubrication
We rst investigate the range of velocities and contact pressures at which low friction can be obtained. A
Stribeck curve, namely a measurement of friction from
very low to large velocities answers this question.
Figure 7 shows a Stribeck curve of silicon nitride sliding in water at 1 N load and gure 8 the same curve

0
0

Velocity [m/s]
Figure 6. Running-in distance of a silicon nitride bearing as function
of the velocity at 2 N load (obtained from gure 5).

for silicon carbide at 2 N. The upper curve is for


increasing velocities where running in still takes place.
The lower curve, for decreasing velocities, corresponds
to the nished bearing and is representative of the
lubrication. Friction stays low at velocities above
0.04 m/s for silicon nitride and 0.5 m/s for silicon carbide. As the sliding velocity is further reduced, the friction returns to high values, which establishes that
lubrication is purely hydrodynamic. A chemical form
of lubrication would result in low friction at all velocities. This nding has important and unfortunate implications for practical applications: water lubrication of
ceramics without additives is probably not possible for
reciprocating movement. The Stribeck curve for SiC
shows that higher velocities are necessary for hydrodynamic lubrication than for Si3N4. This is due to the
higher roughness of these surfaces (Ra=25 nm), which

372

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

140
Decreasing sliding velocity
Increasing sliding velocity

0.6

Scar radius roll 1


Scar radius roll 2
Scar radius disc

130
120

0.4
0.2

hd =

0.02

0.0
0

20

40

60

80 100 120 140 160

Scar radius [ m]

Friction coefficient

0.8

100
90
80

Sliding velocity,mm/s
Figure 7. Friction coecient as a function of sliding velocity (Stribeck curve) for silicon nitride sliding in water with 1 N load. Note
that the upper curve, with increasing velocity represents the runningin of the ball. The lower curve, with reducing velocity, represents the
operation of the nished bearing. We note that hydrodynamic lubrication obtains at velocities >0.04 m/s.

110

70
60
0

Velocity [m/s]
Figure 9. Radius of the at scar on silicon nitride the ball necessary
for hydrodynamic lubrication as a function of the velocity at 2 N
load.

Frict ion Coef ficient

0.4
0.3
0.2
0.1
0.0
0.0

0.5

1.0

1.5

2.0

Sliding Velocity (m/s)


Figure 8. Friction coecient as a function of sliding velocity (Stribeck curve) for silicon carbide sliding in water at 2 N load. The
upper curve was measure with increasing velocity and represents running-in of the ball, the lower curve, measured with decreasing velocity, represents the operation of the nished bearing. Note that
hydrodynamic lubrication obtains at v>0.5 m/s.

sponding minimum water lm thicknesses. These are


shown in gure 10. These lms are very thin, varying
from 5 to 15 nm, and correspond to the low roughness
of the silicon nitride surface (gure 3). At velocities
larger than 1.5 m/s, the lm thickness necessary for
hydrodynamic lubrication increases. We attribute this
to the vibrations in our rather crude tribometer, rather
than to an increase in roughness of surfaces.
An alternate technique of obtaining the minimum
water lm thickness is to run Stribeck curves of friction vs. velocity at several loads, as shown in gure 8.
One reduces the sliding velocity from maximum value,
which reduces the water lm thickness (equation (4)),
and notes the velocity Umin below which friction
30

7.2.2. Thickness of the hydrodynamic water lm


We now estimate the minimum thickness of the
water lm necessary to separate the two surfaces, prevent asperity contact and establish hydrodynamic
lubrication. Figure 9 shows the scar radii q obtained
by running in silicon nitride bearings at various sliding
velocities with the load P = 2 N. We recall that these
were obtained in runs similar to the one shown in gure 2: the diameter of the wear scar increases until friction reaches its minimum value, showing that
hydrodynamic lubrication has been achieved. Inserting
the scar diameters D = 2 q and velocities of gure 9
and P = 2 N into equation (4) gives us the corre-

25

Water film thickness [nm]

demands a thicker uid lm (gure 4). These measurements show that any lm, if present, does not lubricate; they do not prove the absence of a lm on the
surface, that absence has been established by Refs.
[3,13].

20

15

10

Roll 1
Roll 2
Disc

Velocity [m/s]
Figure 10. Estimated thickness of the hydrodynamic uid lm for a
silicon nitride bearing operating in water at 2 N load.

373

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

140

10

Bearing pressure [MPa]

Load at Transition (N)

12

8
6
4
2
0
0.000

0.200

0.400

0.600

0.800

1.000

Roll 1
Roll 2
Disc

120
100
80
60
40
20

Sliding Velocity (m/s)


Figure 11. Relation between load and velocity at which hydrodynamic lubrication starts for a silicon nitride bearing. Lower curve;
lowest velocity (abscissa) at which l <0.02 (such as 0.5 m/s in gure 13); it corresponds to a water lm thickness of 25 nm according
to eq. (7). Upper curve: velocity (abscissa) at which l 0.05; it corresponds to a water lm thickness of 19 nm.

increases for the load P. (This is 0.5 m/s at 2 N load


in gure 8.) Equation (4) predicts the load P and
velocity Umin are proportional to each other if the lm
thickness h0 is constant. Figure 11 shows this linear
relationship, from its slope on computes ho with the
help of equation (4). The gure shows that full hydrodynamic lubrication requires a water lm of 25 nm for
SiC. This corresponds to the rougher surfaces obtained
by tribochemical running in of silicon carbide (gure 4). A similar plot (upper line in gure 11) shows
that the friction coecient increases to l = 0.05
because of asperity contact when the water lm thickness is reduced to 15 nm.
7.2.3. Friction coecient
By using Newtons equation for viscous drag, it is
easy to compute the hydrodynamic friction coecient
of these bearings, using the water lm thicknesses and

0
0

Velocity [m/s]
Figure 13. Load bearing water pressure in silicon nitride bearings as
a function of sliding velocity for three dierent samples.

velocities just determined. This has been done for silicon nitride and is shown in gure 12. While there is
considerable scatter in these values, we see that they
are much lower than the experimental lHD @ 0.02 but
agrees with l = 0.001 obtained with the careful measurements of Kato et al. [9].
7.2.4. Load carrying uid pressure
The radii of the bearing scars of the balls (gure 9),
together with the loads carried, allow us the compute
the average pressure of the water lm p = P/pq2
where the symbols have the known meanings of load
and scar radius. The pressure is shown as a function
of sliding velocity at 2 N load in gure 13 for silicon
nitride. The pressure sustained by the water lm is
very large, reaching about 80 MPa at 1 m/s. It is much
larger than required for normal bearing design.

7.3. Wear rate during hydrodynamic lubrication


0.002
Roll 1
Roll 2
Roll 3

0.0015

0.001

0.0005

0
0

10

15

20

25

Load [N]
Figure 12. Friction coecients of silicon nitride lubricated by water.

It is important to determine the wear rate of the


ceramics during hydrodynamic lubrication. Theoretically, this is zero, so we expect it to be very small.
The measurement consists of running the bearing in
hydrodynamic lubrication for a long distance at sucient load and to measure the worn volume. The resolution of the measurement is limited by the smallest
volume change that can be measured and the total
sliding distance. Very small wear volumes can only be
measured on the ball via equation (1); wear volumes
on the disk by prolometry cannot be measured with
sucient sensitivity. In a representative experiment,
the run in ball had a at wear scar of 0.195 mm
diameter, which corresponds to 2.36 10)5 mm3
removed by run-in. We then ran the sample at hydrodynamic lubrication at 0.995 m/s velocity and 4 N
load for 16 h = 57.3 km. After the hydrodynamic

374

run, the wear scar diameter was 0.203 mm, corresponding to a total of 2.78 10)5 mm3 removed form
the ball. The wear during hydrodynamic lubrication
is obviously the increase in volume removed, namely
DV = 4.2 10)6 mm3.
This corresponds to a wear rate smaller than
2 10)11 mm3/Nm. This wear rate is in the range of
wear rates in well-lubricated machinery.

7.4. Wear rate during run-in


It is easy to measure the overall material removal
rate R during run in: one computes the volume of
material removed VW using equation (1) and the scar
radii gure 7, divides it by the load P and the run-in
distance s of gure 6 and obtains the overall rate

expressed in mm3/Nm. It is more useful to determine


the wear rate RC of the contacting asperities during
running in where the uid carries an increasing fraction f of the applied load P and the asperities carry
the remainder: PC = (1)f ) P. To compute RC, one
divides the wear volume VW by PC instead of P:
5

The task is to determine the factor (1)f ).


Our model of mixed lubrication [13,16] assumes that
the friction force F is a sum of hydrodynamic drag
FHD and friction at contacting asperities FC. Further,
it assumes that the friction coecient C of the contacting asperities is independent of velocity since it does
not have a hydrodynamic component.
The measured friction force F and friction coecient are then
F FHD FC

12
10

Not polished Toshiba plate


Polished Norton plate
Polished Toshiba plate

8
6
4
2
0
0.0

0.1

0.2 0.3 0.4 0.5


Friction coefficient

0.6

0.7

Figure 14. Measured wear rate as a function of measured friction


coecient in mixed lubrication of silicon nitride by water.

When the friction coecient l varies, as it does during run-in, the volume worn is

R VW =PC  s

RC VW =PC  s R=1  f:

Ball wear rate, (mm 3/Nm) x 10e - 6

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

VW PRC

l  lHD =lC : ds;

10

where s is the sliding distance. The integral is obtained


from the record of the friction coecient with time
(i.e., distance) as in gure 2. This allows us to determine the wear rate at contact RC as a function of sliding velocity and load. Figure 15 presents the wear rate
RC at asperities of silicon nitride vs. velocity with 2 N
load; it increases with sliding velocities and has an
average value about 1.5 10)5 mm3 /(Nm). At low
velocities, it has the much lower value of
5 10)6 mm3/Nm, in agreement with earlier results
[17]. Figure 16 shows that the wear rate RC of silicon
nitride at 1 m/s velocity is independent of the load. A
similar computation shows, in gure 17, that the wear
rate RC of silicon carbide, during run-in, is
RC 4 10)6 mm3/Nm, independent of the velocity.

with FC PC lC 1  fP  lC ; 6
4.00E-05

so that
7

lC is the friction coecient measured at very low velocities, where f=0.


We now obtain 1  f l  lHD =lC

3.00E-05

2.00E-05

1.00E-05

and equation (5) becomes


R 1  fRC RC l  lHD =lC :

Rc [mm3/Nm]

F=P l lHD 1  flC ;

Roll 1
Roll 2
Disc

Thus the measured wear rate R is a linear function of


the measured friction coecient l. This is illustrated in
gure 14 for the case of silicon nitride.

0.00E+00

Velocity [m/s]
Figure 15. The wear rate RC at contacting asperities as a function of
velocity at 2 N load for silicon carbide.

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water


5.00E-05
Roll 2
Roll 3

Rc [mm3 /Nm]

4.00E-05

3.00E-05

2.00E-05

1.00E-05

0.00E+00
5

10

15

Load [N]
Figure 16. Wear rate RC at contacting asperities of silicon nitride as
a function of load. Sliding velocity is 1 m/s.

8. Discussion and conclusions


The investigation has conrmed that silicon nitride
and silicon carbide form hydrodynamic bearings when
sliding in water even at quite low velocities. This
occurs because the low roughness of the surfaces, of
1025 nm, allows the formation of very thin water
lms. Running in of these bearings is possible at
velocities and contact pressures that occur in real
machinery. This constitutes thus a possible technology, which, however, has its limits. The fact that
lubrication in pure water is hydrodynamic raises the
questions of wear at start and stop, but we have
found that this is very low. More critical is the question of reciprocal sliding, as in pistons and cylinders
where the applicability of this technology still remains
to be investigated. We have chosen to study pure
water because this situation constitutes a large
domain of applications when oil-lubricated bearings
cannot be employed. The addition of boundary lubrication additives to water would be benecial in reciprocal sliding. The technology described here is

375

certainly not applicable to the lubrication of the piston and cylinder in steam engines where no liquid
water exists and the temperatures are high. There is
no guarantee that water vapor condensation would
occur in the pistonring interface, although such phenomena do occur as stiction in the operation of
readwrite heads of computers. The study of the
lubrication of steam engine cylinders would require
high-temperature and high-pressure equipment and
may best be studied in actual steam engines that are
instrumented for the purpose. In earlier work [18] we
have found reduced friction when silicon nitride slides
in humid ambient at elevated temperature. Whether
this phenomenon could be applied to the situation of
steam engines and would provide low enough wear
remains to be investigated. We also think that water
is not a suitable lubricant for rolling element ceramic
bearings because such geometries depend on elastohydrodynamic lubrication, which is based on a large
increase of viscosity with pressure (piezoviscosity).
The latter is known to exist in hydrocarbon lubricants but is absent in water. It would be of high
interest to investigate the sliding of silicon carbide
against graphite in water since this system is in wide
and successful use in rotating seals.

Acknowledgments
This work is dedicated to the memory of Michael
Gardos who passed away during its performance. We
are grateful to him, not only for initiating this work,
but also for the many stimulating contributions he
made to tribology and to our lives. We thank also
AFOSR for having administered the grant and Dr.
Koji Kato who graciously gave us silicon carbide balls
that he was using for his own research. Thanks are
also due Mr. Peter Krsko for the scanning electron
micrographs.

References

wear rates, (mm3/Nm) x10-6

6
5
4
3
2
ball wear rate
plate wear rate

1
0
0

50

100

150

200

250

300

sliding velocity, mm/s

Figure 17. Wear rate of silicon carbide at contacting asperities.

[1] H. Tomizawa and T.E. Fischer, ASLE Trans. 30 (1987) 41.


[2] S.R. Hah and T.E. Fischer, J. Eletcrochem. Soc. 145 (1998)
17081714.
[3] S.R. Hah and T.E. Fischer, J. Electrochem. Soc. 146 (1999)
15051509.
[4] V. Muratov and T.E. Fischer, Annu. Rev. Mater. Sci. 30 (2000)
2751.
[5] J. Xu, K. Kato and T. Hirayama, Wear 205 (1997) 5563.
[6] H.-C. Wong, N. Umehara and K. Kato, Wear 218 (1998) 237
243.
[7] J. Xu and K. Kato, Wear 245 (2000) 6175.
[8] M. Chen, K. Kato and K. Adachi, Wear 250 (2001) 246255.
[9] M. Chen, K. Kato and K. Adachi, Tribol. Int. 35 (2002) 129
135.
[10] X. Wang, K. Kato, K. Adachi and K. Aizawa, Tribol. Int. 36
(2003) 189197.
[11] S.M. Hsu, Presentation at DARPA Conference on Steam
Engines, Los Angeles, November 20, 2002.

376

L. Jordi et al./Lubrication of silicon nitride and silicon carbide by water

[12] V.A. Muratov, T. Luangvaranunt and T.E. Fischer, Tribol. Int.


31(1998) 601611.
[13] V.A. Muratov and T.E. Fischer, Tribochemical Reactions of
Silicon Nitride in Aqueous solutions, Finishing Adv. Ceram
Glasses p. 245257.
[14] Z. Zhu, V. Muratov and T.E. Fischer, Wear Mat. Intl. Conf.,
Wear 225229 (1999) 848856.

[15] A. Cameron, Basic Lubrication Theory (J. Wiley, New York,


1976).
[16] X.-Z. Zhao, J.-J. Liu and T.E. Fischer, Wear (1997).
[17] T.E. Fischer and H. Tomizawa, Wear 105 (1985) 29.
[18] H. Tomizawa and T.E. Fischer, ASLE Trans. 29 (1986) 481.

Вам также может понравиться