Wear, 91 (1983)
SLIDING WEAR AND TRANSFER*
J. DON, T. C. SUN and D. A. RIGNEY
Metallurgical Engineering Department, Columbus, OH 43210 (U.S.A.)
The Ohio State University, 116 West 19th Avenue,
W. A. GLAESER
Structural Materials and Tribology Section, Avenue, Columbus, OH 43201 (U.S.A.) (Received
Summary It is well known that transfer of material from one component of a sliding pair to the other occurs in many tribological systems. In the present paper the authors describe their observations on transfer material and on debris particles. Detailed structural and chemical information has been obtained by using optical microscopy, scanning electron microscopy, transmission electron microscopy (TEM), scanning TEM and fluorescence analysis using energydispersive techniques (energydispersive analysis of X-rays) and wavelength analysis. The results show a clear connection between the transfer layer and the generation of loose wear debris for both unlubricated and lubricated sliding. Evidence of delamination of base material has not been observed in this work.
1. Introduction It is well known that transfer (and back transfer) of material from one component of a sliding pair to the other occurs in many tribological systems. In some systems, such as those involving the sliding of copper-based materials against various steels, the occurrence of transfer is obvious from the appearance of a coating of copper-colored material on the steel. In other systems, special techniques may be needed to detect transferred material. The importance of transfer is widely acknowledged. It is frequently included in discussions of the role of adhesion in friction and wear [l - 71. *Paper presented VA, U.S.A., 1983. ?Present address: hafen, F.R.G. 0043-1648/83/$3.00
in The Netherlands
However, most of these discussions have been rather qualitative because the tools used to study transfer have not provided enough structural or chemical information to indicate how transfer should be incorporated in friction and wear models. With the introduction and wide availability of modern electron beam instruments it is now possible to apply a variety of complementary techniques to characterize transfer material. For example, Sasada and coworkers [8 - 151 have published an excellent series of papers describing their careful experiments and analysis by scanning electron microscopy (SEM) combined with X-ray analysis. Other groups have used similar techniques and pointed out the importance of transfer for sliding wear phenomena [16 - 181 and for friction [la, 191. In the present paper the authors describe their observations on transfer material and on debris particles. Detailed structural and chemical information has been obtained by using optical microscopy, SEM, transmission electron microscopy (TEM), scanning TEM (STEM), and fluorescence analysis using energy-dispersive techniques (e.g. energy-dispersive analysis of X-rays (EDAX)) and wavelength analysis. The results provide new information on the formation of transfer layers and on the role of transfer material in sliding wear for both unlub~ca~d and lubricated systems.
The results described in this paper were obtained from several different series of experiments, some of which had different primary purposes. Therefore the details of experimental procedures vary somewhat from one set of experiments to another. In this section an outline of the general procedure is presented and variations are noted as necessary. Unlubricated (dry) and lubricated sliding wear tests were done using separate LFW-1 block-on-ring test machines , Each was modified by attaching a Plexiglas chamber to allow some control of the test atmosphere. The purpose here was not to exclude air totally but to provide a reproducible environment for a given series of tests. Most of the unlubricated tests were done in dry argon with a relative humidity of about 20%. A few of these tests were done in air with the same relative humidity. The lubricated tests were done in argon with a relative humidity of 30% - 40%. A typical load sequence for the unlubricated tests was 6’7 N (6.8 kgf) for 15 min, 133 N (13.6 kgf) for 15 min and then 200 N (20.4 kgf) for 140 min. The sliding speed was low (5 cm s-l) to avoid excessive heating. Other tests were done to study the behavior during early stages of running-in. These tests used a load of 67 N for 12,300,600 or 1200 s at a sliding speed of 1 cm s-l. The lubricated tests were done with a load of 133 N and a sliding speed of 6.5 cm s-l. Test rings were usually type 440C stainless steel, but M2 tool steel rings and carburized AISI-SAE 8619 steel rings plated with chromium or coated
with 1 pm of TiB2 were also used. In most cases the rings were used with their as-received surface finish, but they were cleaned in trichloroethylene and methanol. The rings for lubricated tests were carburized AISI-SAE 1018 steel heat-treated to 60 HRC and ground to a 0.25 I.crn center-line average surface finish. Test blocks included Cu-Ni, Cu-Al and Cu-A1203 alloys made from high purity metals as well as commercial materials (OFHC copper, Cu-Cu,O (tough pitch copper) and Cu-Be alloy 25). The blocks were shaped by fly cutting to avoid excessive damage in the near-surface region. Polishing steps included 240, 400 and 600 grit Sic abrasive paper (wet) followed by 3 and 1 I.crn diamond polishing compounds. Ultrasonic cleaning was carried out first in trichloroethylene and then in methanol. In addition, the blocks used for the shortest runs were electropolished to remove about 2 pm of material and then given a final light polish with 0.05 pm alumina. Lubricated tests were done with one of the following lubricants: (1) 1% solution of stearic acid in dodecane; (2) dibasic acid ester (MIL-L-7808); (3) DC 200 (dimethyl silicone). A hypodermic needle was used to apply lubricant to the test ring when the chamber atmosphere had stabilized. During the test, a lubricant drop remained suspended from the ring and could be observed through a low power microscope. Larger wear particles could be detected by flashes of reflected light as they tumbled about in the lubricant drop. Very small wear debris caused the drop to become cloudy. Debris samples could be withdrawn during each test or at the conclusion. They could then be observed by optical microscopy, or they could be prepared for STEM studies by filtering and incorporating the debris in a carbon film [ 211. For the dry sliding tests, wear was determined by measuring the weight changes of the block and ring and also the weight of the loose wear debris. Both the friction force and the near-surface temperature were monitored during sliding. Even with the highest load used, the temperature increase at about 1 mm below the interface was less than 50 ‘C; for the light load and slow sliding speed the corresponding increase was less than a few degrees Celsius. After a dry sliding test, the block, ring and debris were promptly examined by optical microscopy, SEM and EDAX. Next, the block was prepared for cross sectioning by applying a two-layer copper plating [22, 231. The first layer (20 pm thick) was formed in a copper cyanide bath to provide good adhesion. Then a thick plating (thicker than about 2 mm) was formed in a copper sulfate bath. Longitudinal sections (normal to the sliding interface and parallel to the sliding direction) and sometimes transverse sections (normal to interface and sliding direction) were prepared by first spark cutting to form slices about 0.8 mm thick and then electropolishing (1:l solution of H3P04:HZO) to reduce the thickness to about 0.2 mm. Next, a Struers Tenupol jet thinner was used with an electrolyte of H3P04-CrOs to prepare thin foils for TEM and STEM observations. Since the thinned area was not always located in the desired region near the interface, argon ion
beam thinning was used as a final step. Ion thinning was also useful for briefly cleaning any specimen foil which required additional observations at later times. Sections of debris particles were prepared by using similar techniques, as described by Sun [ 231. For normal TEM observations a Philips EM300 instrument was used. High resolution work and chemical analysis were done in a Philips EM400 instrument having a field emission gun, X-ray detector and convergent and microbeam diffraction capabilities.
3. Results 3.1. Scanning electron microscopy-energy-dispersive analysis of X-rays When a copper-based alloy rubs against steel a layer of copper-colored material can frequently be observed on the steel. Color changes may also develop on the copper alloy. Other materials systems may show similar direct evidence of changes in surface composition. In some cases the surface coating will be dark brown or black when viewed by eye or with the aid of an optical microscope. Such coatings are often thought to consist of oxide or some other product of reaction with the environment. Simple observations of surface color changes may be useful indicators of local changes in composition, but they should be interpreted very carefully. For example, a collection of very fine metallic debris will appear black even if no oxide is present, i.e. a color change can occur even without a composition change. Alternatively, a wear surface might have no color change to indicate the possibility of a composition change, yet it could still have a very thin coating or small patches of transfer material. Fortunately, modem SEM instruments are commonly equipped with X-ray detectors for chemical analysis. Thus it is not difficult to detect changes in composition on sliding wear samples, although quantitative work may be affected by the X-ray penetration depth and variable thickness of surface layers. Figure l(a) shows a wear surface of an Ni-40wt.%Cu block at low magnification. The scar is rough and irregular in shape, and the dark regions suggest that transfer material from the 440C stainless steel ring may be concentrated in local areas. Figure l(b) shows an EDAX iron map of this same wear scar. The same general outline of the scar is evident, but there is very little correlation of features in the SEM image with features in the EDAX iron map. The transfer material is spread much more uniformly than would be indicated by the SEM image. Counts obtained from the entire wear track provided an estimate that 10% - 20% of the material near the wear surface of this block is ring material. Similar results were found for tests with other Cu-Ni alloys, except that the amount of transfer depended on the alloy composition. Alloys which were high in nickel usually showed more transfer material (20% - 30%) whereas alloys high in copper showed no more than a few per cent of transfer material [ 171. In some cases the amount of transfer was very low,
(b) Fig. 1. (a) SEM micrograph of wear surface of Ni-40wt.%Cu block (type 440C stainless steel test ring; maximum load, 200 N; sliding distance, 540 m; sliding velocity, 5 cm s-’ ; air with 24% relative humidity): the arrow indicates the direction of motion of the ring surface. (b) EDAX iron map of the area shown in (a).
but some transfer material was present in all cases. Even when the wear scar remained very shiny, as for some copper samples, several per cent of transfer material could be detected. No correlation could be found between the surface appearance of these wear surfaces and the amount of transfer present. It is quite clear that it is difficult or impossible to estimate either the amount of transfer or even its presence by visual means only. Debris particles, mounted on carbon disks by using a suspension of colloidal graphite in isopropyl alcohol, were routinely examined using SEMEDAX techniques. As reported in the literature (e.g. ref. 14), two types of debris were observed. In some cases the particles were dark and included many that were very small; in other cases, especially for copper and some copper alloys, the particles were large shiny flakes. Figure 2 shows typical debris of the first kind (OFHC copper sliding against 440C stainless steel; 90 min). Although some are flakes, they seem to be composed of loosely compacted smaller particles. When touched by a needle they crumble easily. EDAX measurements indicate that transfer material is uniformly distributed throughout the debris. Figure 3 shows a higher magnification view of similar debris from a test in which Ni-40wt.%Cu alloy was slid against 440C stainless steel. The EDAX analysis of the debris cluster shows that it consists of about 70% of ring material. This ring material is very evenly distributed; it is not possible by using EDAX in the scanning electron microscope to detect separate particles of ring or block material. Since Debye-Scherrer X-ray patterns show that f.c.c. (Cu-Ni) and b.c.c. (iron) lines are both present in this kind of debris, it is clear that the pieces of block and ring material are smaller than the resolution of the SEM-EDAX systems and certainly smaller than the particles shown in Fig. 3 (less than about 0.5 pm in size). An example of the second type of debris particle is shown in Fig. 4, which is an edge view of a copper debris flake. This particle has a very
Fig. 2. SEM micrograph of debris particles from the sliding of an OFHC copper block against a 440C stainless steel ring (maximum load, 200 N; sliding distance, 540 m; argon atmosphere; relative humidity, 22%) : the particles include about 3% - 4% ring material. Fig. 3. SEM micrograph of one debris particle cluster from an Ni-40wt.%Cu conditions as for Fig. 1): this cluster contains about 70% ring material.
Fig. 4. Longitudinal (ring, 440C stainless 5 cm s-l):
view (SEM) steel; sliding
of a layered debris particle from an OFHC copper test distance, 540 m; maximum load, 200 N; sliding velocity, about
4% ring material.
pronounced layer structure. About 4% of the debris flake consists of 440C stainless steel ring material. Again, the mixture appears to be very homogeneous because the resolution of the SEM-EDAX equipment is limited. Analysis of sliding tests with many different block and ring materials has shown that transfer material was present on the block and/or ring for both mild and severe wear conditions. For most cases the debris particles, irrespective of their size or shape, consisted of intimate mixtures of block and ring material. In several cases (Cu-Ni blocks with high nickel content) severe damage to the ring occurred [ 17, 241. Large pieces of ring material were embedded in the block and they also appeared as loose debris particles. Both the block and the ring showed deep scratches. However, before this severe damage occurred, debris of the mixed type were generated and transfer layers appeared as in all other cases.
3.2. Tmnsmission electron microscopy and scanning transmission electron microscopy analysis The SEM analysis-EDAX has two limitations in this work. The first is the limited resolution, which is not sufficient to detect the smallest pieces of transfer material. The second is the fact that SEM techniques are for surface or near-surface analysis; they are not suitable for studies of internal structure and composition. It is natural then to turn to TEM analysis to obtain this complementary information. One of the first reports of the use of TEM for studying material below wear surfaces is described in ref. 25. Other work on TEM of materials affected by wear is reviewed in ref. 26. Figure 5 shows part of a typical longitudinal cross section through the wear surface of a copper block (sliding distance, 12 m). The sliding interface and the sliding direction are marked by the arrow. The gap was created by etching during the thinning of the foil for TEM analysis. A portion of the protective plating is visible above the gap. The lower part of the figure shows part of the highly deformed block material with its characteristic cell or subgrain structure; this region is described more fully in ref. 22. A well-defined boundary separates the cell region from the dark band, which has a very fine structure. Figure 6 includes a micrograph of a longitudinal section from another copper specimen (sliding distance, 540 m). In this case an EM400 instrument was used. The very fine-grained material has a structure similar to that shown in Fig. 5. Both iron and chromium were detected in the dark fine-structure region; neither was detected in the plating or in the deformed cell structure of the block itself. By focusing the electron beam to a diameter of about 100 nm, it was possible to obtain local composition results at various distances from the sliding interface. These data are plotted as per cent of ring material uersus distance below the surface in Fig. 6(b). Since the ring was made from 440C stainless steel, the sum of the percentages of iron and chromium was approximately equal to the percentage of ring material. The profile in Fig. 6 shows that the amount of transfer material varies considerably with depth. This is a real effect, with layers about 1 - 2 pm wide, probably related to the layering shown in Fig. 4. The data have been normalized for thickness variations arising from uneven etching during foil preparation. The imaging conditions used for Fig. 6 were sufficient to indicate the presence of composition layering, but they were still not good enough to determine the distribution of small pieces of transfer material. However, by focusing the electron beam to its minimum diameter of 10 - 20 nm, the results shown in Fig. 7 were obtained. In this micrograph, individual crystallites as small as 10 nm can be found. Therefore the electron beam is small enough to provide composition data from a small number of crystallites, and the local composition variations are quite large, as shown by the examples given at the left-hand side of Fig. 7. In general, dark areas in the figure seem to be lower in transfer material whereas light areas are high in transfer material These differences in appearance could be due to differences in the
(b) Fig. 5. TEM micrograph of a longitudinal section through the wear surface of an OFHC copper block (ring, 440C stainless steel; load, 67 N; sliding distance, 12 m; sliding velocity, 5 cm s-l). The arrow marks the position of the interface and the sliding direction. The dark band with fine structure is a transfer layer; the larger cell region is copper-base material. Fig. 6. (a) TEM micrograph of transfer layer on OFHC copper block, analyzed at marked positions by X-ray fluorescence (electron beam diameter, about 100 nm) (ring, 440C stainless steel; sliding velocity, 5 cm s- 1., load, 67 N; sliding distance, 540 m). The deformed bulk material, shown at the right-hand side, has a dislocation cell structure. (b) Results (normalized for thickness differences) of the X-ray fluorescence analysis of the positions marked in (a) shown as percentage ring material plotted against distance below surface.
electron transmission of the material locally or due to different etching behavior of the two materials. However, crystallite orientation could also influence contrast; therefore the lightness or darkness of a given crystallite is not a reliable indicator, by itself, of its composition. Still, the relative contrast may be useful for tentative identification of layers or bands of high or low transfer, whether they are the bands about 1 pm thick shown in Fig. 6 or the much thinner bands visible in Fig. 7. It should be noted that banding of this kind is not always present and the degree of banding varies from one region to another and from one test sample to another. It was not possible with copper samples to find areas of “pure” ring material by using these microanalysis techniques, although regions nearly free of ring material could be found. This indicates that the average size of the transferred ring particles is less than the beam size, i.e. less than about
10 - 20 nm. The average size of particles derived from the copper block was somewhat larger. Electron diffraction patterns of areas high and low in transfer material were also recorded. Regions low in transfer yielded relatively clear diffraction patterns, whereas areas high in transfer gave very poor patterns with little more visible than the central spot from the direct beam. This could be caused by the small size of the particles or by large plastic strains. Diffraction patterns obtained from a larger area (about 1 pm in diameter) of the fine-particle layer usually showed well-defined rings with very little preferred orientation. These patterns were similar to those described by many other investigators who examined parallel sections obtained by thinning from the back of wear samples. Our own ring patterns included rings of a-Fe from transfer material. These were weak because of the small proportion of transfer material present. Transfer layers with very fine microstructures were observed in all cross sections observed, but there were some differences between samples. For example, with Cu-A1203 the crystallite size was somewhat larger and the boundary between transfer layer and base material was not as obvious as in other cases. Also, for some OFHC copper samples, the transfer layer consisted of alternating layers of fine and very fine crystallites as shown in Fig. 8. In this case the sliding distance was only 3 m, but the transfer layer was well developed.
Fig. 7. High magnification view of the center part of Fig. 6. The arrows indicate the points analyzed and the corresponding amount of ring material at these locations; electron beam diameter, about 10 - 20 nm. Fig. 8. TEM micrograph of longitudinal section through an OFHC copper block (ring, 440C stainless steel; load, 67 N; sliding distance, 3 m). The arrow indicates the location of the interface and the sliding direction. The transfer layer consists of alternating layers of fine and very fine crystallites.
To study the development of the transfer layer, a very short test of one LFW ring rotation (sliding distance, 12 cm) was run. Even for such a short test, the near-surface microstructure consisted of well-defined elongated cells, and small patches of fine crystallites were present at the surface. SEM analysis-EDAX of the surface showed that some transfer had occurred. Except for the shortest test, which was too short to allow much accumulation of transfer material, the thickness of transfer material seems to range from several micrometers to tens of micrometers for the sections investigated. It is difficult to provide reliable values for average thickness because most of the work reported here has involved longitudinal sections. These have the advantage of showing the bending over of subsurface features caused by large plastic shear strains, but they do not provide a clear picture of variations related to surface topography [ 22, 271. A given longitudinal section may sample the structure through a ridge or a groove, but the thinning techniques do not alIow one to choose which. Somewhat better estimates of the average transfer layer thickness can be obtained by using transverse sections. Figure 9 shows an example for tough pitch copper. The boundary between the fine-particle layer and the deformed alloy substructure is well defined, but the thickness of the transfer layer varies strongly across the wear track. Debris particles were also examined by TEM and STEM techniques. Large particles from unlubricated tests could be mounted on a substrate and thinned in the jet thinner. Small powder-type debris and loosely bonded clusters were embedded in copper plating  and then processed in the same way as the sample cross sections. Figure 10 shows a micrograph of a large debris flake from a Cu-Be (Berylco 25) test. Its microstructure is very similar to that of the dark layer found in the Cu-Be block and on the copper block (Figs. 6 and 7).
Fig. 9. TEM micrograph of a transverse section through a Cu-CulO (tough-pitch copper) block (ring, M2 tool steel; load, 67 N; sliding velocity, 5 cm s-l ; sliding distance, 360 m): transfer material is pressed into base material (see ref. 27), and the thickness of the transfer layer depends on the position on the surface. Fig. 10. TEM micrograph of a large debris particle (block, Cu-Be alloy (Berylco 25); ring, M2 tool steel; sliding velocity, 5 cm s- r ; a load of 133 N for 180 m was followed by a load of 200 N for 180 m).
Very small debris particles showed a similar microstructure. Figure 11 is a micrograph of a debris particle from a Cu-Al,Os test block. The crystallite size is larger than in the previous case, but still it is very small. From SEM analysis-EDAX it was determined that this kind of debris contained transfer material from the ring. The small twins are like annealing twins in appearance, but the small crystallite size makes it difficult to use the presence of these twins as a means for estimating temperatures achieved during sliding.
Fig. 11. TEM micrograph of small debris particle (block, Cu-2.7%A1203; ring, M2 tool steel; sliding velocity, 5 cm s-l; a load of 67 N for 90 m was followed by a load of 133 N for 90 m and then a load of 200 N for 180 m): the linear features are microtwins.
Some of the debris obtained from lubricated tests were similar to debris resulting from dry sliding, e.g. a Cu-Al alloy block yielded copper-colored shiny flakes typically ranging in size between 10 and 30 pm. However, a second type of debris was rather different from any generated from the unlubricated tests [ 211. Such particles were dark and small (typically 1 - 5 E.crn in size) like one type of dry wear debris, but they were more translucent when viewed in a STEM instrument. Figures 12 and 13 show examples from a Cu-3.5wt.%Al sample lubricated with a solution of 1% stearic acid in dodecane. These micrographs show that the structure consists of small dark particles about 20 nm in diameter embedded in an amorphous gel-like material. EDAX indicated the presence of copper, aluminum and iron, and electron diffraction rings were identified for f.c.c. copper and CuZO. IR spectroscopy showed the presence of a soap and also some residual stearic acid. Thus these debris particles seem to consist of small crystallites of metal embedded in a metal soap. In other experiments, silicone oil was substituted for the stearic aciddodecane lubricant. The resulting debris particles looked very similar. High resolution EDAX showed that the small particles consisted mainly of copper with some aluminum and iron, and the gel was high in silicon from the lubricant. Thus, for both a chemically active lubricant and an “inert” lubricant, similar debris particles were generated. In both cases the metal particles were very small and they were embedded in a gel derived from the lubricant.
Fig. 12. TEM stereo pair of carburized AISI- SAE 1018 distance, 105 m; load, 133 metal particles embedded in
debris particle from a test with a Cu-3.5wt.%Al block and a steel ring lubricated by 1% stearic acid in dodecane (sliding N; sliding velocity, 6.5 cm s-- I): the particle consists of small a gel derived from the lubricant.
Fig. 13. TEM micrograph of wear debris particle from test described spots are metal particles; lighter areas are non-metallic and include diffraction pattern showed that both copper and iron are present.
in Fig. 12. The dark a stearate soap. The
All the results described to this point have been for tests involving two different materials sliding against each other. In these cases, the appearance of fine-grained material on the test samples is an important feature. One might ask whether similar layers develop during self-mated tests. To provide an answer to this question, a block of OFHC copper was tested with an OFHC copper ring. Both surfaces were highly polished (0.05 pm alumina) and thoroughly cleaned. The sliding distance was 12 m and the load was 67 N. Severe damage to both components occurred ~rned~~ly. The friction coefficient rose quickly to about 3 and then stabilized around 1.1 during the test.
Figure 14 shows part of a longitudinal section through the block surface. The grain size near the surface is very small, and a selected area diffraction pattern shows that the microstructure does not have a well-developed texture. The grains gradually increase in size below the surface, and they become elongated. The most striking feature of the fine-grained layer is its thickness of about 80 pm, which is much greater than the thickness values reported earlier. Figure 15 shows the same section much further from the interface; a well-defined boundary separates the fine-grained layer from the underlying cell structure of the copper block.
4. Discussion The development of a transfer layer has emerged from this work as an important phenomenon in sliding wear. Although the layer may be very thin and it may not cover the sample surface evenly, it can be detected even after
Fig. 14. TEM micrograph of longitudinal section through an OFHC copper block (selfmated test with an OFHC copper ring; load, 67 N; sliding distance, 12 m; sliding velocity, 5 cm s-l). Only a small portion of the fine crystallite region is shown; the full layer is about 80 pm thick. The arrow indicates the position of the wear surface and the sliding direction. Fig. 15. TEM micrograph of the deepest part of the layer shown in Fig. 14. The boundary between this layer (left-hand side) and the larger cells of the base material (right-hand side) is well defined. The arrow indicates the sliding direction.
very short sliding distances. Furthermore, the similarity of the fine crystallites which constitute both the transfer layer and various debris particles indicates that the transfer layer is closely connected with common wear processes. Before proceeding with a discussion of the results presented in Section 3, it is desirable to review briefly some related work already published. Several groups have used radioisotopes to detect transfer on wear surfaces [3 - 51. Of these, the work of Kerridge and Lancaster [4, 51 stands out for its originality and its sophisticated analysis. Kerridge and Lancaster presented several very interesting conclusions from pin and disk experiments: (1) equilibrium transfer rate is equal to pin wear rate; (2) there is no direct wear from the pin; all debris are generated from the transfer layer; (3) the original transfer particles are smaller than observed transfer patches and smaller than final debris particles. Sasada and coworkers [8 - 151 have come to similar conclusions by using SEM-EDAX studies. They suggested that transfer patches and layers develop by accumulation of small “transfer elements” and loose debris particles are derived from the transfer material. They concluded that this mechanism for sliding wear operated for both unlubricated and lubricated wear, and that this was consistent with available data on kinds of debris particles, size of debris particles, load dependence and effects due to different atmospheres. Rice et al. [16, 281 have discussed a surface layer (their zone 3) which develops quickly, has a very fine structure and has a mixed composition. They suggest that debris particles are generated at least in part from this layer and that the layer is replenished both from the counterface and from underlying zones. Kragelskii [ 291 suggested that “the wear fragments transferred from the slider mainly form a metallic film on the opposing surface, and this film is also removed as a result of wear by fatigue”. No information was presented concerning the structure or composition of the transfer film. Schell et al.  related the formation of debris particles to transfer processes and concluded that the debris consisted of an intimate mixture of small pieces of both rubbing materials, mechanically mixed to form a microcomposite. Also, when soft brushes were used to remove most of the debris particles as they formed, both the amount of transfer and the wear rate were reduced. Many others have assumed that adhesive transfer processes are responsible for sliding wear, but the details of the mechanisms have not been clear. Most commonly, it has been assumed that a piece of an asperity on one surface adheres to the mating surface, fractures so that it transfers to the mating surface, perhaps passes back and forth and then somehow yields a debris particle directly. The actual process seems to be somewhat more complex. The observations reported in refs. 4, 5, 8 - 18 and 28 and the results of the present work suggest that transfer layers form from small “transfer elements” [ll] which are smaller than typical debris particles. In addition, an
ultrafine gram structure is generated in the transfer layer, probably as a result of deformation and fracture processes. These very small grains may be smaller than the transfer elements themselves. Evidence from X-ray and electron diffraction and from the high resolution STEM work described in Section 3 indicates negligible in&diffusion, either between the very small grains in the transfer layer or between the transfer layer and the base material. This means that in these experiments the average near-surface temperature was not high enough to allow measurable diffusion. The transfer layer builds up very early during sliding; it is clearly visible after 2 or 3 m of sliding distance with a relatively light load of 67 N. This is long before loose wear debris can be detected. The TEM and STEM work revealed that some transfer material was present even after the shortest tests used. Apparently, very early asperity contacts give rise to small transfer elements which accumulate on the surface until some critical condition is reached and only then are loose debris particles generated from the transfer material. One possible critical condition is that the transfer layer reaches a critical thickness for the given sliding conditions (materials, load, environment etc.) . Such a condition would tend to give flake-like wear debris by a delamination process at or near the interface between transfer material and the base material. Delamination is not a new idea for wear. It was proposed by Styri in 1925  and by Fiichsel in 1929 . More recently Suh has emphasized delamination in his theory of sliding wear [33 - 353. His delamination work is based on the nucleation and growth of subsurface cracks in the base material. The theory acknowledges that large plastic strains are developed in the base material, but it does not incorporate any kind of transfer between the sliding materials. Evidence of simple delamination wear has not been observed in the present work. If delamination occurred in the base material, at least one side of the resulting debris flake should retain the cell structure of the deformed base material. Instead, wear flakes have the same ultrafine crystallite size as the transfer layer. Also, evidence of subsurface cracks in the base material has not been observed in cany of the sections studied in this work. Even for a commercial Cu-Be alloy con~in~g large cobalt-rich particles, no cracks appeared; instead, the deforming material flowed smoothly around the particles as if they were rocks in a stream. Suh et al.  have shown that cracks can nucleate around large inclusions in steel, but it is not clear that these cracks would grow and link up to produce delamination in the base material. In fact, recent stress calculations  and recent experiments on residual stress  indicate that the principal direction of propagation of cracks in the base material would not be parallel to the surface and would not cause delamination. It should be acknowledged, however, that some of Suh’s published photographs clearly show cracks running parallel to the sliding surface. This can be reconciled with the results of the present work if Suh’s delamination cracks propagated through transfer material. Without doing high resolution electron microscopy or local composition measurements, such as those
allowed by EDAX or other electron beam techniques, it is not possible to distinguish between delamination in the base material and delamination of transfer material. The distinction is important because the different mechanisms could require different methods for controlling friction and wear problems. Also, fracture calculations based on delamination in base material will be in error if actual delamination events occur in transfer material which has very different structure, composition and properties. One can only speculate about most of the mechanical properties of transfer material. Its hardness can be larger or smaller than that of the substrate material [ 271. The ultrafine grain size may allow super-plastic behavior. Both the deformation and the fracture of the transfer material could depend on environmental factors such as oxidation. Certainly, environmental factors could influence the adhesion events which are critical at several stages: (1) the formation of small transfer elements, (2) the adhesion of transfer elements to each other and (3) the adhesion of transfer material to the base material. If sliding wear occurs by removal of transfer material, and if steady state conditions are reached, then the rate of formation of transfer material must equal the wear rate. This raises the question of how the transfer layer can be replenished. One possibility is that base material is brought to the surface by plastic flow. Another possibility is that the base material is not uniformly covered with transfer material; then base material would be exposed locally to allow formation of additional transfer elements. No available theories predict the very small sizes of the crystallites in the transfer layer and in wear debris. Comminution limits have been proposed [ 38 - 401 but the predicted “ultimate” particle sizes are considerably greater than those of the particles observed in this work. It seems likely that surface energy is involved, and perhaps dislocation image force distances and changes in electronic behavior due to size effects. Obviously, there is room for more work on these very interesting small particles. If the ultrafine particles were simply part of a wide distribution of debris particle sizes, then one might argue that more attention should be given to the larger debris particles . However, the transfer layer, and therefore the larger debris particles, seems to be generated from smaller particles. Therefore they cannot be ignored. One of the important roles of a good lubricant is to reduce metal-tometal contact and thus to reduce adhesive transfer. Nevertheless, some direct contacts seem to occur, small metal particles are generated and transfer layers develop much as they do for unlubricated sliding. The structures of the transfer layer and of the debris particles are similar for lubricated and unlubricated wear. The lubricant slows the wear processes, but the mechanisms are basically the same with or without the lubricant. One way a lubricant can retard significant damage is simply by dispersing the small metal particles which seem to be precursors of the transfer layer. This is a simple dilution effect which makes it less probable that transfer elements will come close enough to adhere to each other. Larger volumes of fluid lubricant would
enhance this effect. Separation of the fine transfer elements is also achieved when a gel forms, as in some of the debris particles described in Section 3. Filtration could remove transfer elements before most of them could form a transfer layer; this would reduce the wear rate by reducing the rate of formation of the transfer layer, just as the brushes did in the tests reported in ref. 17. However, complete removal of transfer elements by filtration would probably be impractical because of the very small sizes of some of these particles. If it were possible to remove all transfer elements (or to keep them all apart), wear would still occur, but it might be a milder form of wear, and severe and sudden damage to the system would be unlikely. Adhesion theories of wear such as Archard’s  and delamination theories such as Suh’s [33 - 351 have not incorporated transfer layers and their role in wear. Therefore any wear equations based on such theories are inadequate for real sliding systems. If realistic wear equations can be developed for sliding, information on the properties of transfer layers will be needed. Also, it may be necessary to include information on local inhomogeneities, such as the layering evident in Figs. 4 and 6. Finally, it is tempting to suggest that the ultrafine crystallite transfer layer described in this paper is what some people still refer to as the Beilby layer . However, in all the cases we have observed, the layer is definitely crystalline, although some of the crystallites are as small as 3 nm. In brief, the transfer layer has such a different structure from that proposed by Beilby that it should not be called the Beilby layer .
5. Conclusions The work reported here has provided new information about the nature of transfer processes during sliding. The following list of conclusions is based largely on experiments with simple copper-base materials sliding against hard steels. Further experiments should be performed before these conclusions are generalized to other sliding systems. (1) Transfer patches or layers are common features of both unlubricated and lubricated sliding systems. (2) Transfer layers begin to develop very early, even before loose wear debris can be detected. (3) The component particles of the transfer layer are very small crystallites, 3 - 30 nm in diameter. They are derived from both sliding partners. (4) When unlike materials slide against each other, the transfer layer is a microcomposite consisting of an intimate mechanical mixture of crystallites derived from both sliding partners. (5) With self-mated tests, the crystallite sizes are a little larger and the transfer layer is much thicker than for tests with unlike materials. (6) Loose debris particles have the same structure and composition as the transfer layer.
(7) An important wear mechanism involves delamination of transfer material. (8) Simple delamination of base material has not been observed in this work, even for materials containing hard particles. However, similar work should be done on other systems (e.g. steels) to see whether similar results are found. (9) Fluid lubricants can reduce wear by dispersing and separating small transfer particles (transfer elements) before they can accumulate to form a transfer layer. (10) Environmental effects and various surface treatments should affect wear by affecting transfer processes and the structure and properties of transfer material. (11) Available wear equations are inadequate for real sliding systems because they do not incorporate the effects of transfer material. Realistic wear equations should depend on the properties of both sliding partners, on the transfer material which forms between them and on environmental effects.
Acknowledgments We are grateful to the National Science Foundation for research support under Grants DMR 7805719 and DMR 8120675. The work with lubricated systems was supported by the Army Research Office and the Office of Naval Research (ONR) under Grant DAAG 29-80-C-0145 and by the ONR under Grant N00014-82-C-0255. Professor W. A. T. Clark, Metallurgical Engineering, The Ohio State University, provided useful advice for electron microscope procedures and analysis. We also thank Dr. J. Bentley for his help and cooperation using the EM400 facility at the Oak Ridge National Laboratory as part of the U.S. Department of Energy SHaRE program.
References 1 F. P. Bowden and D. Tabor, The Friction and Lubrication of Solids, Part II, Clarendon, Oxford, 1964. 2 D. Tabor, Wear - a critical synoptic view, Proc. Znt. Conf. on Wear of Materials, St. Louis, MO, 1977, American Society of Mechanical Engineers, New York, 1977, pp. 1 - 11;J. Lubr. Technol., 99 (1977) 387 - 394. 3 E. Rabinowicz, Friction and Wear of Materials, Wiley, New York, 1965. 4 M. Kerridge, Metal transfer and the wear process, Proc. Phys. Sot., London, Sect. B, 68 (1955) 400 - 407. 5 M. Kerridge and J. K. Lancaster, The stages in a process of severe metallic wear, Proc. R. Sot. London, Ser. A, 236 (1956) 250 - 264. 6 M. Antler, Processes of metal transfer and wear, Wear, 7 (1964) 181 - 203. 7 E. M. Moore, P. F. Packman and J. J. Wert, A fracture mechanics approach to the problem of friction and wear, Tribal. Int., 7 (6) (1974) 242 - 244. 8 T. Sasada, H. Ohmura and S. Norose, The wear and mutual transfer in Cu/Fe rubbing,
I8 19 20 21 22 23 24 25
Proc. 15th Jpn. Congr. on Materials Research, Society of Materials Science, Kyoto, 1972, pp. 1 - 6. T. Sasada and H. Kando, Formation of wear particles by the mutual transfer and growth process, Proc. 16th Jpn. Congr. on Materials Research, Kyoto, 1973, Society of Materials Science, Kyoto, 1972, pp. 32 - 35. T. Sasada, S. Norose and Y. Shimura, Composition of wear particles produced under sliding friction of different metal combinations, Proc. 18th Jpn. Congr. on Materials Research, Kyoto, 1975, Society of Materials Science, Kyoto, 1975, pp. 77 - 81. T. Sasada and S. Norose, The formation and growth of wear particles through mutual material transfer. In T. Sakurai (ed.), Proc. Joint JSLE-ASLE Int. Lubrication Conf., Tokyo, 1975, Elsevier, Amsterdam, 1976. T. Sasada, S. Norose and J. Nagai, Wear in different metal combinations in vacuum, Proc. 21st Jpn. Congr. on Materials Research, Society of Materials Science, Kyoto, 1978. T. Sasada, S. Norose and H. Mishina, The behavior of adhered fragments interposed between sliding surfaces and the formation process of wear particles, Proc. Int. Conf. on Wear of Materials, Dearborn, MI, 1979, American Society of Mechanical Engineers, New York, 1979; J. Lubr. Technol., 103 (1981) 195 - 202. S. Norose and T. Sasada, Mutual transfer of rubbing materials and the mixing structures of wear particles formed in lubricating oil, J. Jpn. Sot. Lubr. Eng., Int. Edn., 1 (1980) 5 - 9. H. Mishina and T. Sasada, Observation of metallic structure of seized portion and mechanism of seizure, Proc. 24th Jpn. Congr. on Materials Research, Society of Materials Science, Kyoto, 1981. S. L. Rice, H. Nowotny and S. F. Wayne, Formation of subsurface zones in impact wear, Proc. 35th ASLE Meet., Anaheim, CA, 1980, in ASLE Trans., 24 (1981) 264 268. J. Schell, P. Heilmann and D. A. Rigney, Friction and wear of Cu-Ni alloys, Proc. Znt. Conf. on Wear of Materials, San Francisco, CA, 1981, American Society of Mechanical Engineers, New York, 1981, pp. 53 - 62; Wear, 75 (1982) 205 - 220. P. Heilmann and D. A. Rigney, Running-in processes affecting friction and wear, Proc. 8th Leeds-Lyons Symp. on Tribology, Lyons, 1981, IPC, Guildford, 1982. S. Mahdavian, Y. W. Mai and B. Cotterell, Friction, metallic transfer and debris analysis of sliding surfaces, Wear, 82 (1982) 221 - 232. Standard method for calibration and operation of the alpha model LFW-1 friction and wear testing machine, ASTM Stand. 02714-68, 1968 (ASTM, Philadelphia, PA). W. A. Glaeser, The nature of wear debris generated during lubricated wear-in, ASMEASLE Fall Meet., October 5 - 7, 1982. P. Heilmann, W. A. T. Clark and D. A. Rigney, Orientation determination of subsurface cells generated by sliding, Acta Metall., 31 (8) (1983) 1293 - 1305. T. C. Sun, Technique for preparation of wear debris particles for transmission electron microscopy, Wear, 79 (3) (1982) 385. P. Heilmann and D. A. Rigney, Experimental evidence for fatigue during sliding wear, Metall. Trans. A, 12 (1981) 686 - 688. K. R. Lawless, Critical appraisal and research opportunities - the materials research viewpoint, Interdisciplinary Approach to Friction and Wear, NASA Spec. Publ. SP181, 1967, pp. 433 - 478 (National Aeronautics and Space Administration, Cleveland, OH). A. W. Ruff, L. K. Ives and W. A. Glaeser, Characterization of wear surfaces and wear debris. In D. A. Rigney (ed.), Fundamentals of Friction and Wear of Materials, American Society for Metals, Metals Park, OH, 1981, pp. 235 - 289. J. Don, T. C. Sun and D. A. Rigney, Friction and wear of Cu-Be and dispersionhardened copper systems, Proc. Int. Conf. on Wear of Materials, Reston, VA, 1983, American Society of Mechanical Engineers, New York, 1983; Wear, 91 (1983) 191. S. L. Rice, H. Nowotny and S. F. Wayne, Formation of subsurface zones in impact wear, 35th ASLE Annu. Meet., Anaheim, CA, May 1980, Preprint 80-AM-2D-3.
29 I. V. Kragelskii, Friction and Wear, Butterworths, Washington, DC, 1965, pp. 126 131. 30 J. Don, Unlubricated friction and wear in the CupBe system, Ph.D. Dissertation. Metallurgical Engineering, The Ohio State University, Columbus, OH, 1982. 31 H. Styri, Mech. Eng., 4 7 (1925) 490. 32 M. Fiichsel, Organ Fortschr. Eisenbahnwes., 84 (1929) 413. 33 N. P. Suh, The delamination theory of wear, Wear, 25 (1973) 111 - 124. 34 N. P. Suh, An overview of the delamination theory of wear, Wear, 44 (1977) 1 - 16. 35 N. P. Suh, S. Jahanmir and E. P. Abrahamson II, The delamination theory of wear, Rep., September 1974 (Massachusetts Institute of Technology, Cambridge, MA) (Advanced Research Projects Agency, U.S. Department of Defense). 36 A. R. Rosenfield, Wear and fracture mechanics. In D. A. Rigney (ed.), Fundamentals of Friction and Wear of Materials, American Society for Metals, Metals Park, OH, 1981, pp. 221 - 234. 37 J. W. Ho, C. Noyan, J. B. Cohen, V. D. Khanna and Z. Eliezer, Residual stresses and sliding wear, Tech. Rep. 7, May 1982 (Department of Materials Science, Northwestern University, Evanston, IL). 38 K. Kendall, Complexities of compression failure, Proc. R. Sot. London, Ser. A, 361 (1978) 245. 39 K. Kendall, The impossibility of cornminuting small particles by compression, Nature (London), 272 (1978) 710. 40 J. T. Hagan, Impossibility of fragmenting small particles: brittle--ductile transition, Rep., 1981 (Cavendish Laboratory, Cambridge University, Cambridge, Cambs.). 41 E. Rabinowicz, discussion to T. Sasada, S. Norose and Y. Shimura, Composition of wear particles produced under sliding friction of different metal combinations, Proc. 18th Jpn. Congr. on Materials Research, Kyoto, 1975, Society of Materials Science, Kyoto, 1975. 42 J. F. Archard, J. Appl. Phys., 24 (1953) 981. 43 G. Beilby, Aggregation and Flow of Solids, Macmillan, London, 1921. 44 L. E. Samuels, Myths die hard - even in science, Physics of Materials, A Festschrift for Dr. Walter Boas, Griffin Press, Melbourne, 1979, pp. 238 - 243.