Elsevier Sequoia S.A., Lausanne - Printed in the Netherlands
IBM Research Laboratory, San Jose, CalijI (U.S.A.)
(Received May 15, 1972)
An experimental investigation into material transfer and wear has been performed for the case of a ferrite probe sliding at high speeds (between 25 cm/set and
1250 cm/set) and low loads (10 gm) on a steel disk, a nickel plated disk, and a polymercoated disk, the latter containing finely dispersed abrasive particles. Autoradiographic
methods together with microdensitometer analysis have been applied, resulting in
detailed information concerning the mechanism of transfer and its relationship with
wear. Large amounts of transfer are observed, both for sliding on the steel disk and
the nickel plated disk, and also for sliding on the polymer-coated abrasive disk.
The latter observation seems to suggest that large amounts of material transfer are
characteristic not only for sliding situations favoring adhesive wear, but also for the
general case of abrasive wear caused by hard particles dispersed in a soft matrix.
The understanding of material transfer and wear has been greatly enhanced
by the application of radioactive tracer techniques. The main advantage of this method,
it appears, is related to the high sensitivity and resolution with which radioactive
particles can be detected. This, together with the easy availability of radioactive
tracers by neutron activation, has made it possible to obtain detailed information
concerning the nature of the wear process as well as the relationship between wear
and material transfer.
In some of the earliest applications of radioactive tracers, Rabinowicz’ and
Rabinowicz and Tabor investigated friction and material transfer during the sliding
of lubricated and unlubricated metals. Both investigators found appreciable metallic
transfer, consisting, in general, of a relatively small number of large individual particles.
The results showed furthermore that transfer and wear were greatly reduced, but not
eliminated, if boundary lubricant films were applied to the sliding surfaces. Large
amounts of transfer were also found by Kerridge3, who studied the wear of annealed
tool steel against hardened steel using a pin on ring machine. Typical for the latter
experiment was a very high initial transfer rate which was seen to level off after a finite
time of sliding, reaching eventually an equilibrium value independent of further wear
on the pin. This, as well as other experimental evidence, led Kerridge to suggest that
Wear, 22 (1972)
transfer is not merely present as a result of sliding, but is an important first link in a
series of events resulting in the formation of loose wear particles.
The important role of transfer in the wear process was likewise observed by
Archard and Hirst4 in a study of mild wear of hardened tool steel ; by Kerridge and
Lancasters and by Steijn6 in investigations of the sliding of brass against hard metallic
surfaces; and in addition, by Hirst and Lancaster7 in a study of wear of brass on
hardened tool steel. Material transfer, however, has not only been observed during
the sliding of material combinations like soft/hard or hard/hard, but also during the
sliding of very hard material on a much softer surface as, for instance, tungsten carbide
on copper’.
Based on the above experimental evidence, it seems justifiable to conclude that
material transfer is indeed intimately linked with the wear process. It should be pointed
out, however, that most of the above results were obtained in experiments conducted
under conditions of high load (of the order of several kg) and low sliding speeds (of the
order of several cmjsec). Furthermore, most of the materials investigated were metals,
i.e. materials favoring adhesive wear during sliding. Although it seems likely that a
similar interdependence between material transfer and wear should exist for different
types of wear, or for sliding conditions of light load and high velocity, such a conclusion
may not be drawn a priori on account of the apparent changes in the physical situation
at the sliding interface.
The investigation reported in this paper is a study of material transfer and
wear under sliding conditions of light loads (of the order of 10 g) and high sliding
speeds (of the order of tenths of cm/set up to 1250 cm/set). In particular, material transfer and wear have been investigated in a pin on disk machine for sliding of a spherical
ferrite probe (hardness 750 kg/mm’) on a tool steel disk, a nickel plated disk of submicron plating thickness, and a polymer-coated disk containing finely dispersed
abrasive alumina particles of average size of approximately one micron. Autoradiographic methods have been applied together with optical microdensitometer analysis,
resulting in detailed information concerning material transfer and the mechanism of
wear. Although large differences exist in the wear behavior of the ferrite probe on the
metallic disks and the abrasive disk, appreciable amounts of transfer are observed for
the metallic disks as well as for the abrasive disk, thus indicating that material transfer
is typical not only for wear situations favoring adhesive wear, but also for certain cases
of abrasive wear.
The investigation was carried out in a set-up similar to a classical pin on disk
machine. Continuous sliding contact is established between the rotating disk, driven
by a variable speed motor, and a 3.1 mm diameter spherical ferrite probe (Fig. 1).
The probe is spring-loaded against the disk, which rotates in a closed environment of
filtered, slightly pressurized air. The cleanliness of the test chamber is monitored
continuously using a Royce particle counter capable of detecting airborne particles
larger than 0.3 microns in diameter. The load, applied to the probe by a cantilever
spring, is measured via high gain solid state strain gauges connected to a digital
readout. Relatively soft springs were used as loading springs in order to keep load
Fig. 1. View of spring-loaded
on disk.
variations due to vertical runout
Temperature and relative
dependence of the experimental
small range of temperature and
of the disk (typically 0.05 mm) below 1%.
humidity were monitored continuously ; however, no
results on these parameters was noticed within the
humidity fluctuations encountered.
Experimental procedure
The ferrite probes were irradiated prior to the experiment for 100 hours in the
Union Carbide nuclear reactor at Tuxedo, New York. Although a high neutron flux
density of 6.7 x 1013 neutrons/cm2/sec was used during irradiation, the level of radioactivity of the probe after irradiation was only in the low millicurie range on account
of a probe weight of only 89 mg. Thus, handling of the probe during the experiment
was possible without extreme health precautions or undue radiation exposure.
In a typical experiment, the ferrite probe of radius a was attached to the cantilever beam by means of a special mounting bracket ; thereafter sliding was allowed to
take place for prescribed conditions of load W, speed u, and time t. The total wear
volume V was then calculated from optical measurement of the wear scar diameter
d produced on the spherical probe, i.e.
j2a2 + %j
(a2 - %,‘I
or, since in all cases d < 2a,
64a *
Wear volumes as small as lo- lo cm3 were found to be resolvable with the
above technique, as long as the wear scar diameter d was measured under at least 200 x .
It should be pointed out that in the above calculation elastic deformation of the contact zone has been neglected, since only very light loads were applied to the probe.
Wear, 22 (1972)
Before each individual test, the probe surface was cleaned using isopropyl
alcohol; furthermore, a new unworn area of the probe was exposed for each new
test by rotating the probe in the holder. After all experiments were conducted, autoradiographs of the worn track areas were obtained by placing Kodak type AA X-ray
film in direct contact with the disk surface and developing the image formed on the film.
Exposure times of several weeks were generally employed, although it was found that
autoradiographs of sufficient resolution could be obtained with appreciably shorter
exposure times. To enable a ready comparison of different autoradiographs, without
the need for half-lifetime or exposure corrections, all autoradiographs, including
calibration autoradiographs, were exposed concurrently (25 days) and developed under
identical conditions.
Rapid decay of several short-lived isotopes occurs immediately following
irradiation so that the activity of the ferrite probe, after a slight cooling period, is
mainly due to the relatively stable isotopes Fes5, Fe”, Ni63, Ni6’, and Zn65. In
particular, the radiation emitted by these isotopes is (a)X-ray radiation of Fe55, Ni5’,
and Zn65 at the Mn-K-x, Co-K-x, and Cu-K-x line, respectively, (b) y-radiation of
Fe5’ and Zn6’ at 0.191 MeV, 1.10 MeV, 1.14 MeV, 0.82 MeV, and 0.32 MeV, and (c)
B-radiation of Fe5’ and Ni63 at 0.271 MeV, 0.462 MeV, 1.56 MeV, and 0.067 MeV.
Thus, exposure of the film is caused not only by X- and y-rays, but also by b-rays.
While the presence of several types of radiation is unimportant in qualitative
autoradiography, it becomes important in quantitative analysis. In fact, inaccurate
conclusions may result should self-absorption of P-rays be neglected in calibrating
autoradiographs using samples whose dimensions are appreciably larger than those
of average wear particles. That is to say, B-radiation effectively originates within a thin
surface layer of a large sample, and is thus essentially a surface phenomena, whereas
X-ray and y-radiation are essentially volume phenomena, relatively independent of
self-absorption. Hence, relative exposure of the film, based on volume ratios, decreases
for large calibration samples due to self-absorption, contrary to the case of small wear
particles for which self-absorption is negligible. It is thus apparent that calibration
results using large specimens are prone to errors, showing a tendency toward overestimating the amount of transferred material.
In order to avoid the difficulties outlined above, calibration of the autoradiographs was obtained as follows. The radioactive ferrite probe was dissolved at 140”C
in hydrochloric acid using a laboratory high pressure bomb. The resultant solution
was neutralized with sodium carbonate and, subsequently, was again made slightly
acidic with tartaric acid to keep the iron chloride in solution. Standard dilutions of
known amounts of radioactive material were then made and distributed over areas of
one square inch each. To limit the fluid from spreading further than the desired square
inch areas, low surface energy barriers were painted around the boundaries of the
squares prior to the application of the diluted sample. After drying, autoradiographs
were obtained of these square inch areas (Fig. 2) resulting in a relationship between
uncalibrated optical density and known amount of material per unit area.
Although the optical density across individual squares was reasonably uniform, optical integration was performed over the square areas using the experimental
set-up shown in Fig. 3. Diffuse white light is projected at an individual square after
Wear.22 (1972)
Fig. 2. Calibration autoradiographs.
Fig. 3. Schematic of opticai integration set-up.
Background Density
Optical Density
Fig. 4.
Calibration curve : radioactive material per unit area vs. optical density.
Wear, 22 (1972)
passing through a one inch square aperture. The light of intensity I is collected on a
photomultiplier tube, which was calibrated prior to the experiment using standard
density filters, thus establishing the desired relationship between calibrated optical
density d
.il I,dAij
and the known amount of radioactive material per unit area.
The calibration curve obtained with the above procedure is shown in Fig. 4.
It is clearly noticeable from this graph that optical density increases very slowly at low
area1 densities of radioactive materials, whereas it increases very rapidly at area1
densities above 0.001 mg/in2 although no saturation of the film occurs.
and wear
In Fig& typical autoradiographs
Fig 5. Typical
are shown for the steel disk, the electroplated
(a) Steel disk. (b) Nickel plated disk, (c)Abrasive
nickel disk, and the abrasive disk, respectively, all obtained under sliding conditions
of 10 g, and varying speeds and sliding times. From visual observation, it is apparent
that appreciable amounts of transferred material are present on all disks, and furthermore, that the distribution of transferred material is noticeably different from disk to
disk. In particular, transferred material in the case of the steel disk appears to be concentrated in relatively large discrete spots along the track (Fig. 6) while an almost
uniform distribution of transfer is present for the abrasive disk (Fig. 7). The distribution of transfer for the nickel plated disk, on the other hand, appears semi-uniform
with an appreciable number of small particles superimposed (Fig. 8).
It seems apparent that the differences in the autoradiographs are caused by
different mechanisms of wear prevailing in the different sliding situations. A similar
conclusion may also be derived from a qualitative examination of the wear scars
formed on the probe and the disk, or from a quantitative investigation of the probe wear
Wear, 22 (1972)
Fig. 6. Photomicrograph
of autoradiograph
Fig. 7. Photomicrograph
of autoradiograph
along a track
a track
on the steel disk (100 x ).
on the abrasive
disk (100x).
In Fig. 9(a), a typical wear scar is shown at 210 x for the ferrite probe after
sliding at 250 cm/set on the nickel plated disk. It is apparent from visual observation
that the wear scar is much like a metallographic etch, revealing clearly the grain structure ofthe ferrite. Microcracks and regions ofgrain pull-out are furthermore noticeable,
suggesting that the mechanism of wear is seemingly one of line scale abrasion as well as
one of large scale grain pull-out due to adhesion between the ferrite and the metallic
surface. A similar appearance of the wear scar is also found for sliding at low speeds on
the steel disk; however, a marked difference is noticeable for sliding on the abrasive
disk (Fig. 9(b)). Here, a large number of finely distributed grooves exist parallel to
the direction of sliding; no microcracks can be detected. The wear mechanism in this
Wear,22 (1972)
Fig. 8. Photomicrograph
of autoradiograph along a track on the nickel plated disk (100 x ).
case is seemingly one of pure abrasion without large scale grain pull-out. In the absence of grain pull-out, no large fragments should exist (as was already evidenced by
the autoradiographs for the abrasive disk (Fig. 7)) and it seems justifiable to conjecture
that the abrasive wear mechanism proceeds on a uniformly small scale, in distinction
to the above suggested two-level process in the case of the metallic disks.
The last statement should not be interpreted as if the wear rate of the ferrite
probe were lower for sliding on the abrasive disk than for sliding on the steel or nickel
plated disk. In fact, the opposite situation prevails at low speeds, as can be observed
from the plot of wear rates as a function of speed and time, shown in Fig. 10(a) and (b)
for sliding on the steel and abrasive disk, respectively. That is, at low speeds the wear
rates for sliding on the abrasive disk are typically higher by at least one order of magnitude than those on the steel disk; this behavior is changed only at high speeds
when appreciably higher wear rates are observed for the probe sliding on the steel
disk. From Fig. 10(a) and (b), it is also apparent that the probe wear rate increases
nonlinearly as a function of speed and sliding time for sliding on the steel disk,
while a linear increase in the wear rate as a function of time is observed in the case of
the abrasive disk.
Due to the generally large amount of wear occurring at high speeds. one may
expect large amounts of radioactive material along a track on the disk. While autoradiographic investigation confirms this conjecture, electron microscope investigations show transfer and wear particles present only in the case of metallic disks, as
shown in Fig. 11, for instance, for the steel disk. No definite signs of transferred particles are noticeable along tracks of the abrasive disk (Fig. 12) inasmuch as the abrasive disk appears unchanged during a wear test; i.e. the individual particles in the
coating in Fig. 12 are alumina particles already present at the beginning of sliding.
Since in the latter case a marked difference was observed in the distribution of transfer
particles along the autoradiographs, as compared to that on the metallic disks,
it is apparent that changing of material combinations can result in first order changes
in the size and distribution of transfer particles.
Wear, 22 (1972)
Fig. 9. (a) Typical wear scar on ferrite probe after sliding on the nickel plated disk (210 x ); (b) typical wear
scar on ferrite probe after sliding on the abrasive disk (210 x }.
The investigations reported in the previous section demonstrate clearly that
material transfer is intimately related with the wear process. To further enhance our
understanding of the situation at the sliding interface, quantitative information concerning the amount of transfer and its dependence on speed and sliding time is essential.
Due to the irregular distribution of transfer particles along wear tracks of the steel
disk, an estimate of transfer based on autoradiographs is iikely inaccurate, requiring
the analysis of a large number of individual particles. However, in the case of the abrasive disk, for which uniform distribution of transfer was observed, a situation much
r = 1250 cmlsec
v = 250 cmhc
2 0.2
. /
I /
Y = 250 cmlsec
Sliding Time [min.]
Sliding Time [min.]
Fig. 10. (a) Wear as a function of sliding time (ferrite probe on steel disk, 10 g load); (b) wear as a function
of sliding time (ferrite probe on abrasive disk, 10 g load).
Fig. 11. Electron
along wear track on the steel disk (2750 x ).
more amenable to analysis is encountered, permitting quantitative determination by
means of the following straightforward procedure.
First, the variation of optical density across the autoradiographic image of a
wear track is determined by microdensitometer analysis. The radial density profile
obtained in this way is then approximated by piecewise constant density regions of
varying width wiand density di (Fig. 13). Ifthe density gradients along the circumference
of the wear track are small enough to be negligible, the above approximation is clearly
equivalent to the assumption that the wear track of length li is made up of a number of
adjacent annular rings, each of density di, width wi, and length li, respectively. Since
the functional relationship between density and the amount of radioactive material
per in2 area, i.e. d=d(m), is known from the calibration curve obtained previously
Wear, 22 (1972)
Fig. 12. Electron photomicrograph
along wear track on the abrasive disk (1200 x , u= 250 cm/[email protected]
Fig. 13. Typical
’ min
profile across track on autoradiograph.
Fig. 14. Density profiles as a function of sliding time (u = 30 cmjsec, 10 g load).
(Fig. 4), it now follows that the amount of radioactive material Mi contained in each
annulus is given by
while the amount of radioactive material M contained in a wear track is given by
1M =
Figure 14 shows several density profiles for the abrasive disk corresponding to
wear tracks obtained as a function of sliding time under conditions of 30 cm/set and
10 g. It is apparent from visual observation that well defined density maxima exist for
all profiles which are nearly symmetric with respect to the center of the track. MoreWear, 22 (1972)
over, it can be observed that the maximum density of the profiles increases with increasing sliding time, indicating a monotonic increase in the amount of transferred
material with sliding time, or equivalently, with the number of repeated passes. Using
the above described procedure for calculating the amount of transferred material, we
find that the amount of radioactive material increases nearly linearly with sliding
time (see lower curve in Fig. 15). An even more interesting result is found, if one notes
that the total amount of transferred material is, to within 40 %, identical in value with
the total amount of probe wear, thus suggesting that most of the worn pin material is
deposited on the disk.
Fig. 15. Wear and transfer
as a function
of sliding time (ferrite probe on abrasive
disk, 10 g load).
Similar results regarding the linearity and the amount of transfer are also observed at sliding speeds of up to 250 cm/set (center curves of Fig. 15), as well as at
speeds appreciably less than 30 cm/set, although in the latter case the experimental
accuracy decreases on account of the low sensitivity of the calibration curve at low
optical densities. A much different situation is encountered as the sliding speed is
increased to approximately 1250 cm/set (upper curves in Fig. 15). There, an increase in
transfer is observed only during the initial stages of sliding; thereafter, an equilibrium
situation is seen to be reached with an approximately constant level of transfer independent of the total amount of probe wear.
It appears that the main cause for the observed dependence of transfer on speed
is related to the wear behavior of the abrasive medium itself. That is, at low speeds the
wear mechanism of the rider/disk interface seems to be mainly that of abrasive wear of
the probe due to the abrasive action of the finely dispersed alumina particles; very
little medium wear is present under these sliding conditions, as was evidenced by the
electron micrographs of Fig. 12. At high speeds, however, wear is observed on the
abrasive medium, even for the same number of traversals that resulted at low speeds
in unnoticeable medium damage. Hence, it appears justifiable to speculate that at
high speeds the rate at which abraded probe material is deposited in the medium
equals the rate at which transferred material is removed by medium wear. This mechanism apparently causes the experimentally observed constancy of transfer after an
initial sliding period. On the other hand, at low speeds, when the scale of medium wear
is much decreased, transfer of material is not counter-balanced by removal and,
therefore, a continuous increase of transferred material is observed.
Although the above considerations show the cause for the speed dependence of
transfer, no explanation has been given as to why, in general, transfer occurs in the
present wear situation. That is, one may expect large amounts of adhering particles in
an adhesive wear situation as, for instance, during the sliding of metals, but not in an
abrasive wear situation where one of the surfaces is a polymeric material containing
hard abrasives. Yet, it appears that the presence of large amounts of transferred material is related to the specific structure of the abrasive coating. That is, although the
hard abrasive particles in the coating are responsible for the abrasiveness of the medium, a large percentage of the medium consists of a soft polymeric material. Thus,
little resistance may be expected from the polymer, if abraded particles are pushed into
it during the repeated traversals of the probe over the same track. This then results in
the experimentally observed situation of large amounts of transfer.
It should be re-emphasized in support of the above suggested transfer mechanism that electron photomicrographs
of worn tracks did not clearly reveal the
presence of adhering particles, although autoradiographs indicated large amounts
of transferred material to be present. This, therefore, seems to indicate that abraded
particles are embedded in the coating rather than attached to the surface and, furthermore, that the average size of transferred particles is appreciably smaller in the
present abrasive wear situation than in a typical adhesive wear situation.
The experiments described in this paper have shown that material transfer
may be important not only during sliding at high loads and low speeds, but also during
sliding at low loads and high speeds. Moreover, it was found that changes in the
mechanism of wear, caused by introducing different sliding materials, may result in
substantial changes in the nature of transfer. As is evidenced from the autoradiographs
obtained, the most sensitive parameter in this regard is apparently related to the size
and distribution of transfer particles. Although a calculation of the individual particle
size is prone to error, a first approximation to the average particle size V,, may be
obtained as
where V is the total amount of transferred material, here assumed to be equal to the
total amount of worn probe material, and n is the number of individually resolvable
particles along a wear track on the autoradiograph. Clearly, this estimate for V,, is
likely to be larger than the actual average size, since the total amount of transferred
material could be smaller than the total amount of probe wear and, in addition, not all
transfer particles can be resolved in the autoradiographs as individual transfer particles. Applying the above analysis to the nickel plated disk, we estimate the average
particle size to be of the order of lo- I3 cm3. Since the smallest resolvable particles are
smaller than the average particle, it is apparent that the resolution of the autoradiographic method is beyond the order of lo-l3 cm3. Comparing this estimate with predictions of other investigators, for instance, Rabinowicz,’ who indicates the order of
lo- ’ 3 cm3 as the upper resolution limit for autoradiography, we see that the present
results suggest a slightly higher resolution limit, seemingly caused by the long exposure
times (25 days) and possibly higher radiation levels per unit volume.
Based on the previous considerations, it is apparent that the average size of
the transfer particles in the case of the abrasive disk is much smaller than that found
for the nickel plated disk or the steel disk. In fact, if the volume of the average wear
particle is smaller by only a factor of 10 to 100, say, we find that the linear dimension of
an average transfer particle, for the abrasive disk, would be as small as 1000 A.
This, then, would seem to be in general agreement with our previous considerations
concerning extremely small particle sizes in the case of the abrasive wear situation.
It should also be pointed out that throughout this paper only results were discussed for sliding under a constant load, with speed and sliding time as parameters.
Very similar results concerning the distribution of transfer particles and the appearance
of the wear scars are also found if the speed is kept constant and load and sliding time
are varied. Thus, it seems apparent that the basic mechanism of transfer in a given wear
situation is not likely to change as a function of experimental conditions such as load
and speed as long as the basic mechanism of wear remains unaltered.
One last point, emerging from the experimental results obtained in this investigation, should be emphasized. This point is that the mechanism of transfer and
embedding of material, suggested for the polymer-coated abrasive disk, seems to be
valid not only for the particular sliding situations investigated, but also for the general
case of abrasive wear caused by hard particles embedded in a soft matrix. From this
viewpoint, the results of this investigation may also be taken as an indication that
material transfer is typical both for adhesive wear and for abrasive wear caused by
hard particles dispersed in a soft matrix.
The author would like to express his appreciation to H. Werlich for very capable help in the microdensitometer work. The electron photomicrographs were obtained by H. R. MacQueen.
1 E. Rabinowicz,An
investigation ofsurface damageusingradioactive
metals, Et-it. J. Appl. Physics, 2( 1951)
2 E. Rabinowicz
and D. Tabor, Metallic transfer between sliding metals: an autoradiographic
Proc. Roy. Sot. (London), A 208 (1951) 455.
3 M. Kerridge, Metal transfer and the wear process, Proc. Phys. Sot. (London), B 68 (1955) 400.
4 J. F. Archard and W. Hirst, An examination
of a mild wear process, Proc. Roy. Sot. (London),
A 238 (I 957)
5 M. Kerridge and J. K. Lancaster, The stages in a process of severe metallic wear, Proc. Roy. Sot. (London),
A 236 (1956) 250.
6 R. P. Steijn, Sliding wear and metal transfer under unlubricated
conditions, Trans. ASME, J. Basic Eng.,
81 (1959) 67.
7 W. Hirst and J. K. Lancaster, The influence of speed on metallic wear, Proc. Roy. Sot. (London). A 259
(1960) 228.
8 J. Golden and G. W. Rowe, A simple apparatus
for studying the wear of tungsten carbide, Wear, 1
(1957/58) 491.
Wear, 22