J. SPREADBOROUGH Battelle
July 28. 1961)
Experiments are reported which show that packets of graphite planes roll up on rubbing the material to form small roller bearings. Electron microscope studies of a natural graphite flake demonstrated that these rollers are formed only about a stylus path over the flake; they are not observed in undeformed regions. A mechanism for graphite lubrication is suggested which involves a surface orientation of crystallites and the subsequent formation of the rollers responsible for the low friction coefficients under normal conditions. ZUSAMMENFASSUNG
Es wird iibcr Experimente berichtet, welche ergaben, dass wahrend der Reibung Pakete van Graphitlagen aufgerollt werden und so auf dem Material eine Art Rollenlager bilden. Elektronenmikroskopische Urtersuchungen an natiirlichen Graphitflocken zeigten, dass solche Rollen nur in dem Bereich zn finden sind, wo ein reibender Stift iiber die Oberflache bewegt wurde; in unbcriihrten Bereichen wurden sie nicht beobachtet. Kin Mechanismus fiir die Graphitschmierung wird vorgeschlagen, welcher zuerst eine Orientierung der Kristallite annlhernd parallel zur Oberflache bewirkt und nachfolgend eine Bildung dieser Rollen, welche fur den niedrigen Reibungskoeffizienten, der miter Normalbedingungen beobachtet wird, verantwortlich sind.
The low friction coefficient of graphite at normal temperatures and pressures is of considerable importance. In a previous notei it was suggested that, apart from the usual explanation in terms of sliding graphitic layers, the formation of rollers made up of packets of layer planes might be responsible, in part at least, for the lubrication of graphite. The experiments described in the present communication were performed in the hope of showing that, under certain conditions, rollers are formed on rubbing a graphite surface, and to explore this idea further. EXPERIMENTAL
The experiments were of two kinds: the examination, by electron microscopy, of surfaces and debris, and the measurement of friction coefficients using a machine of the Bowden typez, in which the deflecting force was determined by the displacement of the anode of a valve forming part of an initially balanced bridge. The out-of-balance voltage was fed to an oscilloscope whose trace gave a direct indication of the stickslip behaviour and the friction coefficient. The graphite used included natural Madagascar flakes, fibrous natural Ceylon
FRICTIONAL BEHAVIOUR OF GRAPHITE
graphite, reactor graphite (grades A and B), and a graphite made from compacted natural flakes with no binder. For convenience in handling and identification samples were mounted on polished copper tables IO x 15 x z mm. Friction coefficients were measured using copper styli prepared from pure copper wire, diameter 1.5 mm, ground and polished to hemispherical tips. Samples were studied first with an optical microscope, and then prepared for electron microscope examination of the surface and debris in the following way: “Triafol”* discs of diameter I mm adhering to the small punch rod used to produce them were moistened with a drop of acetone and pressed on the appropriate region. After a few minutes the discs were removed gently, shadowed with a carbon + metal film, placed on a grid, and bathed in acetone vapour in a Soxhlet. This dissolved the Triafol leaving the replica + debris on the grid. Other deliberately abraded graphite samples were dusted directly on a microscope grid. Samples were examined in a Siemens Elmiskop I electron microscope at 80 or IOO kV. In order to examine the influence on the frictional properties of the intercalation of ferric chloride in graphite, natural flakes were heated with resublimed FeCls in an evacuated sealed tube for 24 hours at 280°C. The flakes were then washed with decinormal HCl and distilled water and dried on filter paper in air. RESULTS
Preliminary experiments on the variation of friction coefficient with applied load indicated that the measured coefficients tended to rise when the load was decreased to about I g. All subsequent experiments were performed with applied loads of at
Fig. 1. Oscilloscope
trace of friction coefficient for a polished showing stick-slip behaviour.
foil from Farbenfabriken
Bayer AG, Werke Dormagen,
least 1.5 g. The oscilloscope traces showed that considerable stick-slip occurred c\:cn for the polished polycrystalline samples ~~~a typical tract of one of thcsc is gilyen iI1 Fig. I. For the natural flakes, the stick-slip was even worse, and whilst the largest and best flakes were selected, the measurement of the friction coefficient \vas made morci difficult by their small size. The friction coefficients for such samples have a scatter OI about f 0.02; values for the \rarious kinds of graphite at room temperature in air are given in Table I. The differences between the various graphites are thought to bc significant. The ferric chloride intercalation seems to raise the friction coefficient of the natural
The values for the fibrous Ceylon samples the literature
for various commercial
with those reported
Photographs of abraded reactor graphite dust are shown in Figs. z and 3. The rolling-up of the flakes is clearly visible. There are two points which may be noted concerning these photographs: firstly, the rollers are not all parallel for the crystallites observed, and secondly they are not necessarily straight. At A in Fig. z, the rollers are kinked, suggesting that the rolling-up started in two places for the same layer packet and continued until the two rolls met. In most cases where one flake is observed the rollers are roughly parallel to each other as in B of Fig. 2. Examination of other graphites showed similar roller formations. A single flake used for friction measurements was studied. The replica pick-up technique was used to take seven samples, the arrangement of which is shown in Fig. 4. Two samples were of the tracks on the copper table, one was of the untouched region of the flake and the others were either wholly or partly on the stylus path on the flake. The results obtained are briefly described for each sample.
FRICTIONAL BEHAVIOUR OF GRAPHITE
;. 2. Transmission
of abraded reactor graphite.
Fig. 3. (As for Fig. 2) Wear, 5 (1962)
.Saq%e I showed some large and some small flakes of graphite (all pictures wt’~-t~ supplemented by diffraction patterns to verifythe observations). The large undisturbed flakes had dimensions of the order of IO p. Smaller folded-over flakes of size -5 ,u wcrt’
I Fig. 4, Sketch friction tracks,
showing the locations of the replica pick-up sample on the natural flake: I, L -debris on Vu table, 3 ~ partly on stylus path, 4 on stylus path, 5 on undisturbed flake. o nearly on top of 4, 7 on top of 0.
BEHAVIOUR OF GRAPHITE
Sample 2 had several flat flakes of graphite with rolled-up sheets clearly visible. Some rollers were bent and rollers of all axial orientations were observed. The diameter of the rollers was about 0.3 p (Fig. 6). Sample 3 contained a flat region of parallel lines on the replica, bounded on one side by an irregular surface with a few graphite flakes. The width of the parallel line
Fig. 6. Micrograph
from replica pick-up no.
showing various roller arrangements.
region, exemplified in Fig. 7, was of the order of 0.3 mm. Examination of the flakes by optical microscopy showed that the path track of the stylus was also of this width; thus the gouged parallel lines were identified with the track. To the side of the track some graphite flakes containing rollers were picked up, probably debris swept out of the path by the stylus. Sample 4 showed the parallel lines of the track, and pieces of deformed graphite from the flake surface. The flakes off the track looked similar to those of Fig. 5. Very little graphite was picked up; the replica of the surface of the flake is illustrated in Fig. 8. Sample 5 had large areas of graphite surface replica similar to Fig. 8 with some large flakes dotted about. These flakes seemed to be relatively undisturbed compared with those of the other replicas; this was brought out clearly by the diffraction patterns which showed the Kikuchi lines characteristic of highly perfect crystals. Wear, 5 (1962) 18~.30
of the surface
of the flake
off the stylus no. _,.
from a region
BEHAVIOUR OF GRAPHITE
Fig. 9. Large roller assemblies from replica pick-up no. 6.
Fig. 10. Large roller assemblies from replica pick-up no. 6. Wear, 5
Sam$e 6 showed large roller assemblies, typical of which are Figs. o and IO, superposed on the parallel lines of the track. Sample 7 contained the parallel line region of the path bounded on two sides by the blotchy region of which Fig. 8 was typical. The path region had no graphite pick-up on it, presumably because it had been scavenged by making sample 6. *Just off the path one or two pieces of graphite were observed, with prolific rollers which were bent and lay in various directions. Small pieces of the natural flakes which had been treated with ferric chloride were rubbed on a grid, which was examined. Diffraction photographs showed extra dots and rings. This sample changed with exposure to the beam; on increasing the intensity it jumped out of focus and large rings appeared and expanded. These were presumably due to the decomposition of the FeCl& intercalation compound. DISCUSSION
The pick-up experiments on the natural flake showed that rolled-up foils existed only in the regions where the stylus passed (care was used in mounting the sample to avoid deforming it or rubbing the surface.) The samples of debris swept off the flake at the end of the stylus path demonstrate that the shpping (and folding) of loose flakes also occurs. Such natural flakes are not perfect single crystals but rather aggregates of single crystal flakes which may be relatively easily detached from each other. However, the central region of the flake chosen (most of the stylus path) certainly consisted of few crystals and the observed Kikuchi lines from the pick-up sample 5 indicated a high local perfection. If shear within graphitic layers were responsible for lubrication the formation of rollers would be unnecessary; the natural flakes would seem to be the most favourable case for shear, and so the observation of rollers is taken as indicative that roller formation is easier than shear, under normal conditions. Before discussing the nature of the friction behaviour of graphite, a list is made of important experimental observations which a satisfactory lubrication mechanism must explain. Eq!wimentid
on the friction
bekcasiour of graphite
(a) The friction coefficient of graphite in the atmosphere was not affected by a reduction in pressure; on subjection to prolonged heating at 800°C i+z Dacao then cooling to room temperature the friction coefficient rose to 0.563. Introduction of water vapour or oxygen to a. pressure of I p gave a measurable decrease in friction and further admission to IOO ,IA pressure restored the friction to its atmospheric value of 0.15. Although a high friction coefficient was observed after outgassing, there was no seizure, unlike outgassed metals 4. It was suggested that the absence of seizure was due to the lamellar structure and that the recorded high friction coefficient implied a strong interface adhesion accompanied by slip and shearing of graphite planes below the surface. (b) SAVAGES found that the presence of water vapour to a pressure of 3 mm (or ammonia, benzene or other easily condensible vapours) gave good lubrication, but that hydrogen, nitrogen and carbon monoxide up to pressures of 600 mm gave wear as observed for outgassed graphite ilz vacua. Similar observations were made by ROWE”. SAVAGE concluded that the action of friction was to chop off those crystallites
which were not aligned in the plane of the surface and that the vapour formed a very thin surface layer responsible for the lubricating qualities. (c) A doubt was raised about SAVAGE’Smechanism by CARTERS,who showed that graphite had negligible wear in dry helium at 600°C; HOVERsuggested that the high temperature may allow enough intergrain plastic flow to permit the crystallites to rotate into the surface plane, presenting a smooth surface, and that the effect of vapour was to loosen the interparticle bonding and accomplish the same end. The friction coefficient of outgassed graphite decreases with temperature. For graphite on graphite it drops from 0.6 at room temperature to 0.2 at 15oo’C3. This behaviour occurs for different kinds of graphites. (d) When graphite is rubbed in one direction and the direction of rubbing is reversed, there is a short period during which high friction is observed and after which the friction coefficient falls to its normal, low value. It was suggested that graphite plates oriented with their slip planes on the surface and parallel to the direction of motion overlapped in the direction of stroke to form a “shingled” layer5. Stereoscopic electron microscope replica studies of graphite brushes13 showed that the graphite brush surfaces contained “a myriad of individually isolated graphite projections (tongues) stroked in the direction of motion so as to lie at varying obtuse angles to the base plane. These projections evidently provide the surfaces of the contact. They are undoubtedly flexible and elastic . . .“. (e) Electron diffraction observationsll~l3.13 indicated a high degree of surface orientation in rubbed graphite surfaces with the basal planes at a small angle ( -10’) to the surface, and the basal plane normals tilted against the direction of motion of the opposing body. Diffuse spots, corresponding only to orders of basal plane reflections, were observed. The “weak interlayer
The “classical” interpretation of the low friction coefficients of graphite in air in terms of the weak interlayer forces is unsatisfactory. FINCHES pointed out that cleavage by simultaneous rupture of bonds in a plane needs very high energy. This “ease of shear” idea is further discredited by the fact that graphite does not shear readily; mechanical tests show tensile failure unlessa highcompressionissuperimposed The roller mechanism involves the breaking of interlayer bonds piecemeal and so will require much less force than a shear process. Equally, the temperature dependence of the friction coefficient of outgassed graphite cannot be readily interpreted in terms of a more facile shear of the lamellae since the mechanical properties of graphite are not greatly influenced by temperature. This last fact is also contrary to the suggestion of DEACONAND GOODMAN~~, who interpreted the observed temperature decrease in friction of graphite irz vuczco in terms of a weakening of the binding between the crystallites. A mechanism fey the frictional
The imperfect nature of ordinary graphites cannot be too highly stressed; a typical commercial graphite consists of a number of crystallites, of only moderate perfection within themselves, in a framework of pores, partially ungraphitized binder and perhaps also ungraphitized carbon. The friction mechanism for polycrystalline graphites is visualized in terms of two Wear,
processes. Firstly, the surface crystallites will be rotated or displaced into a basal plane orientation nearly parallel to the surface. Rollers will form as soon the orientation approaches to within a few degrees of the basal plane orientation and will gives the low friction observed in air or suitable vapours. The process is illustrated schematztally in Fig. II. Figure
b) Fig. I I. A schematic illustration of the proposed friction mechanism. For simplicity, only a few crystallites are shown ; the graphite extends beyond laterally and below those sketched. The basal planeorientationsof the crystallites are indicated by the parallel lines within them. Loosecrystallites such as that marked L are easily displaced. Crystallites such as those marked R may rotate into a suitable orientation.
passage of the moving body. Figure rrb shows the moving body on the surface. Crystallites such as those marked R will be displaced to an orientation suitable for the production of rollers. Rollers will run off the crystallite and fill up crevices, etc. Loose crystallites will be displaced as that marked L and will either rub over other crystallites on rollers as shown in the figure or be swept away. The normal preparation of a graphite surface for general use or for friction studies will produce a surface orientation suitable for roller formation. In the case of single crystal flakes, it may be that the energy for the sliding of graphite planes is not much greater than that needed for roller formation. TSUZUKU~~ has reported the sliding of graphite planes in small single crystals, and the observations on the pick-up samples I and z indicated that initially some layers slid off the natural flake studied during the present work. However, as mentioned previously, natural flakes should be regarded as aggregates of single crystal flakes, some of which may have been easily detached during the first passes of the stylus.
and experimental observations
The friction process is visualized as being first an orientation of surface crystallites into an approximate (001) texture followed by a rolling up of layer planes. The observations of other workers listed previously may be explained satisfactorily in terms of this model. The effect of degassing is to raise the interlayer binding force and so make roller formation more difficult. It may be that the energy necessary to peel off rollers from a crystallite is then sufficiently high for another process - perhaps shear of layers or gross deformation of the surface - to occur. The influence of the various condensible vapours may be understood as a progressive loosening of the interlayer forces, first at the edges of crystallites and later inside them. This agrees with the observation that if water vapour is admitted to a previously outgassed sample the friction coefficient decreases over a considerable period of time. It is significant that the inert gases He, Nz, Hz (which are not expected to be chemisorbed) do not reduce the friction coefficient of outgassed graphite. The reduction of the friction coefficient at high temperatures in dry atmospheres or in vacua may be understood as a temperature-loosening of the interlayer binding forces, facilitating the formation of rollers, The electron diffraction observations on abraded graphite surfaces reported by many workers do not conflict with a roller mechanism. The only ~ffractions expected from a surface containing rolled-up layer packets will be the diffuse basal plane diffractions which are observed in practice. The reversal effect may be interpreted as follows: the “scales” of graphite lying almost in the basal plane orientation have to be reversed in inclination as proposed by SAVAGES; this involves a movement of the surface crystallites, the unrolling of previously formed rollers and the initiation of new rollers, all of which might be expected to give a temporarily high friction coefficient. The graphite-ferric
The results of other workers, quoted previously, concerning the chemisorption of vapours in graphite by reaction with the z electrons of the graphite all deal with substances which lower the cohesion between the lattice planes of graphitee. For example, graphite will take up bromine to form CaBr-a compound where the interlayer space distance is greater than that expected from the diameter of the Br atoms, implying that the interlayer graphite bonding is reducedie. ROWE states that bromine gives the lowest friction6. Pursuing the idea that the frictional behaviour of graphite is directly a function of its layer structure as evidenced by roller production, a compound was sought which would enhance the graphite interlayer bonding, thus making the formation of rollers more difficult. This is the reason for the preparation of the graphite-ferric chloride complex. The molecular compounds of graphite are more stable than the other ionic compounds and there is evidence that the bonds between the graphite and the ferric chioride are quite strongrr, and thus the interlayer forces may be stronger than for graphite. These compounds are also stable to solvents and mineral acids (except concentrated sulphuric and nitric acids) so it might be expected that if the friction behaviour were controlled by a roller mechanism the friction coefficient would be greater than would otherwise be the case without intercalation because of Wear, 5 (1962)
an increased interlayer force. (Steric effects, impeding the entry of vapours, do not seem likely since ammonia may enter a graphite-ferric chloride lattice and fully coordinate the ferric atoms.) The higher friction coefficient of the natural flake treated with ferric chloride compared to the untreated natural flakes (Table I) is in agreement with the above. Other layer compounds Deformed
flakes have also been examined
microscopy and rollers observed. There were very few rollers, however, compared to graphite, and MO& and similar compounds with Van der Waals interlayer bonding probably show low friction by planar slip. This is supported by the observation that such compounds show low friction when formed under vacuum conditions and higher friction in the presence of chemical vapoursc. Boron nitride, the classical analogue of graphite, shows an irreversible decrease in friction coefficient in the presence of organic vapours which lower the interlayer forces by disrupting the s electron pair. Boron nitride will also form intercalation compounds with ferric chloride which are quite stable to heat and solvents and which have the iridescent blue-black colour shown by the graphite-ferric chloride compounds. So probably the friction coefficient of BN will be increased by the intercalation of ferric chloride. It may be remarked in passing that Fig. 2 of DEACON AND GOODMAN’S paperis, which is a transmission electron micrograph of boron nitride, contains features which look suspiciously like rolled up sheets; nitride.
it is to be expected
that roller formation
will occur in rubbed boron
The author wishes to thank Dr. W. BOLLMANN for helpful and stimulating
Mrs. E. SHALA~~and Miss R. KOLLRACK for assistance with the photography and the preparation of the figures, and Battelle Memorial Institute, Geneva, for sponsoring the work.
W. BOLLMANN AND J. SPREADBOROUGH, Nature, 186 (1960) ~g. W. BOLLMANN AND H. TANNENBERGER, Proc. Intern. Congr. for Chronometry,
P. 75.5. F. I’. BOWDEN, J. E. YOLZNG AXI> G. ROWE, Proc. Roy. Soc.(London), A 21.z (rggz) 439. I’. I’. BOWDEN AND J. E. YOUNG, Proc. Roy. Soc.(London), A 208 (1951) 311. R. H. SAVAGE, J. Appl. Phys., 19 (1948) I. G. W. ROWE, Wear, 3 (1960) 274, R. L. CARTER, quoted by HOVE in ref. 8. J. E. HOVE, Trans. Met. Sot. AIME, 212 (1958) 7. F. I’. BOWDEN, Friction and Wear, (Proc. Symposium on Friction and Wear, Detroit Ig57), ElseVier Publ. Co., Amsterdam, Igsg, p, gr; D. M. KBNYON, Ph. D. Dissertatio+z, Univ. of Cambridge, 1956. E. F. FULLAM AND R. H. SAVAGE, J. ilppl. Phys.. 19 (1948) 654. J. 11’. MIDGELY AND D. G. TEER, Nature, 189 (1961) 735. P. v. K. PORGESS AND H. WILMAN. Proc. Phvs. Sot. flondon). 76 (1960) 513. R. F. DEACON AND J, F. GOODMAN; PVOC.Roj. Soc.(iondon),,A’zd> iIg;$ 464. G. I. FINCH, Proc. Phvs. Sot. (London). A 63 (1~50) 785. T. TSUZUKU, Proc. Third Coni. on Carbon, Buffalo, 1957. Pergamon Press, 1958, p. 433. W. R~~DORFF, Z. anorg. Chew, 245 (1941) 354. R. C. CROFT, Quart. Revs. (LoTzdon), 14 (1960) I. Weav,