GENERAL DISCUSSION Dr. H. Wilman (Imperial College, London)said : May I ask Briscoe and Tabor whether the approximate agreement of the shear strength of 3 x lo6 N m-2 for the calcium stearate of their fig. 5 with the approx. 3 x lo6 N m-2 at a contact pressure of about 10' N m-2 in fig. 4 means that the results in fig. 5 correspond to this contact pressure? If so, can they indicate to what they ascribe the rise in shear strength to more than 10' N m-' at the higher contact pressures in fig. 4? Do they consider that no regions of direct contact (of higher shear strength) of the two underlying substrates (in general, atomically uneven) are involved, presumably increasingly the higher the pressure ? Can the rise in shear strength correspond at least partly to a breaking-up of the molecular layers into progressively smaller crystalline regions and thus increase the extent of the grain boundaries and associated dislocations, and so increase resistance to shear? Their conclusion, that at high contact pressures the shear of long-chain materials such as stearic acid involves sliding of the chains lengthwise over each other, is interesting.I recall that Schoon pointed out that the angle of tilt of the chains to the ab planes in crystals of the a, B and y forms of stearic acid are consistent with the chains cohering along their length, but with different longitudinal shifts; and he suggested that such a sliding of the chains along each other by a multiple of the C-C spacing might be possible, and correspond to a series of other possible poly- morphic forms or metastable equilibria of molecule coherence.It seems not necessary that the molecule chains should lie quite parallel to the shear direction, i.e., the substrate surfaces. Electron diffraction results which I obtained on stearic acid in 1937 showed that rubbing a thick layer of stearic acid ( N 1 cm2 area at a few kg load) in one direction, caused a strong 2-degree preferred crystal orientation, as shown by a pattern of diffraction spots and arcs. The long axes of the molecules were then oriented not parallel to the substrate interface, but at about 5" to it (in a vertical plane through the rubbing direction), instead of at the normal angle of about 55" to the substrate when in (001) orientation. (Rubbing a very thin layer of stearic acid on to polished copper resulted, however, in a strong orientation with the molecule chains tilted in the rubbing direction, at the usual angle of about 55" to the substrate; but this could correspond to chemisorbed copper stearate.) Brummage found that in some cases rubbed n-paraffin layers were in part in an orientation with the chains at about 3" to the substrate surface, analogous to my observations above on stearic acid.Dr. B. J. Briscoe and Dr. D. Tabor (Cambridge University) said: We are grateful to Wilman for the points he has raised. His surmisal concerning fig. 5 is perfectly correct. The results in this figure for the polymers are based on direct friction experi- ments between polymers and glass where the contact pressure is of the order of several kg mm-2 (or several lo7 N m-2).These results have been compared in this figure with results obtained on calcium stearate films at a known contact pressure of lo8 N m-2. Since then we have transferred thin films of polymer to the glass substrate and studied the shear properties of the film by exactly the same procedure as that T. H. Schoon, 2. phys. Chern. B, 1937, 39, 385. see fig. 40 of G. I. Finch, J. Chern. Soc., 1938, 1137. K. G. Brummage, Proc. Roy. SOC. A, 1947, 188,414. 56GENERAL DISCUSSION 57 described for calcium stearate. The shear strength of the polymer films turns out to be somewhat greater than that quoted in fig. 5 but the effect of contact pressure closely resembles that shown in fig. 4 for calcium stearate etc. In these experiments, examination of the substrate before and after sliding shows no evidence for glass-glass contact even at the heaviest pressures employed.If such interaction was to occur the damage to the glass would be severe and detectable. Further, the results are very similar to those obtained in the bulk shearing of metallic soaps (see ref. (1 7), (1 8) of our paper) and the bulk shearing of polymers between diamond platens recently carried out in our laboratory by Dr. H. D. Flack. The increase in shear strength with pressure may or may not be due to grain boundary effects analogous to the increased hardness of metallic specimens as the grain size is reduced. It may not, however, be necessary to invoke such a mechanism. The contact pressures are very large compared with the modulus of these materials so that the molecular components are pressed close together.In these circumstances the interaction energy must be increased whether grain boundary effects are involved or not. Further many of the soaps lack any clearly defined crystalline structure. One of the un- certainties involved in the experiments quoted by Wilman is that electron diffraction can show the structure only after sliding has occurred. We have looked at the struc- ture of polymer films sheared between diamond platens using X-ray diffraction during the shearing process itself. Here the evidence supports the view that there is con- siderable, if not complete, orientation of the chains in the direction of shear. The evidence quoted for chain orientation is of great interest. Prof.M. W. Roberts (University of Bradford) said : Briscoe’s reference to a fraction of a monolayer of chemisorbed oxygen present on an iron surface is puzzling in the sense that oxygen rearranges very readily on iron to give an oxide. For example many “ monolayers ” form at a Po, N Torr and 77 K, i.e., with close to zero activation energy. Dr. B. J. Briscoe and Dr. D. Tabor (Cambridge University) said: We are grateful to Roberts for his comments. It may well be that the films formed in our experiments on iron were thin oxide films rather than a chemisorbed film. The main point we wished to emphasise was that far more oxygen was required to produce a comparable reduction of adhesion with copper surfaces. Dr. R. G. Linford (Berkeley Nuclear Lab.) said: The transition from retarded to non-retarded forces is shown in fig.2 of the paper by Briscoe and Tabor as occurring in the same region as the change-over from the dynamic to the jump method. In view of the extreme experimental difficulties of these elegant experiments, are the authors convinced that there is a definite transition and that it has been correctly located ? Dr. B. J. Briscoe and Dr. D. Tabor (Cambridge University) said: In reply to Linford, the detailed results given in our paper show that both methods give self--consistent results. The jump method covers separations from 2 to 20nm and shows that as the separation is increased the transition from non-retarded to retarded forces begins for separations greater than 12 nm. The resonance method covers separations from 120 nm down to 10 nm and shows that the transition from retarded to non-retarded forces begins at a separation of 50 nm and is complete when the separation is 12 nm.Thus, both techniques are in complete accord as far as the transition is concerned and in the region of overlap the absolute agreement in the magnitude of the force is58 GENERAL DISCUSSION extremely close. For both techniques the accuracy of the power law index n in the two regions is better than f O . l as indicated in fig. 3. Dr. J. R. Young (Duckhams Oils) said: The observations of Briscoe and Tabor relating to PTFE sliders on clean glass are puzzling. The results show that the wear mode of the polymer changes once smooth sliding has commenced. A thin orientated polymer layer is then transferred to the glass, and the surface of the slider becomes similarly orientated in the direction of sliding.If the slider is now turned through 90" and again moved over the previous track " lumpy " transfer with high friction is again observed in the initial stages. It seems unlikely that the transferred layer is well ordered and, in so far as the individual chain groups are concerned, may be considered a random two-dimensional array. If this is true one must ask how does the slider surface distinguish between the two directions of motion relative to the transferred layer ? Dr. B. J. Briscoe and Dr. D. Tabor (Cambridge University) said: Young has raised an interesting point but it is based on a slight misunderstanding of the statements in our paper. The only results of repeated sliding over the same friction track refer to the behaviour of an oriented slider over an oriented film.Here the van der Waals interaction involves a shear strength just about equal to or slightly less than the forces required to draw further material from the slider. The friction is low and there is practically no further transfer of PTFE. The effect of rotating the slider through 90" was limited to sliding over a clean glass surface.The interaction between PTFE and glass is strong : the orientation of the slider is unfavourable for easy drawing : a lump of PTFE is plucked out of the slider and the initial friction is high. The behaviour when the slider is rotated through 90" and slid over an oriented PTFE film is more complicated and is not discussed in the present paper though it is dealt with in the fuller paper by Pooley and Tabor.In effect, it depends on whether the transferred film completely covers the glass substrate. If it is a " good " continuous film, the friction and transfer are both low. If the film is tenuous so that a fair amount of glass is exposed through the film, the friction may be high. In general, if the interfacial adhesion is strong, the friction and transfer are large unless the sliding surface is favourably oriented. If the interfacial adhesion is low the friction and transfer are both small. Prof. D. D. Eley (Nottingham University) said: In relation to Weaver's paper, it is interesting to note that McAloon and Perkins,l using SCF CNDO theory have calculated an electron energy band structure for polyethylene with a band gap of 13-19 eV depending on the approximation used.Dr. J. R. Young (Duckhams Oils) said: With reference to the postulate of Allen et al. that migration is the process accounting for the formation of a filler-free layer at the adherend surface, the observations of Griffin are relevant. Griffin was seeking an explanation for the wide variations in friction and wear properties reported for apparently similar mixtures of solid lubricants (e.g., graphite, and molybdenum disulphide) in thermoplastic supports (e.g., polypropylene and polystyrene). He examined the distribution and orientation of particles close to the surface of injection- moulded test pieces. He found that (i) the surface of the test piece which had con- ' B.J. McAloon and P. G. Perkins, J.C.S. Faruduy Trans. 11, 1972, 68, 1121. G. J. L. Griffin, A.S.L.E. Trans., 1972, 15, 171.GENERAL DISCUSSION 59 tacted the mould was totally free from solid particles for a depth of 20-30 pm; (ii) the plate-like particles deeper in the solid were oriented parallel to the surface. arising from his theoretical study of the viscous flow of lamellar solid suspensions. It may, therefore, be unnecessary to invoke diffusion to account for the surface effect presently reported. These observations were judged consistent with predictions made by Jeffery Dr. H. Wilman (Imperial College, London) said : Can Sutcliffe and Cameron clarify what they assumed about the direction of motion of the (001) molecular layers over each other, relative to the crystallographic axes a and b ? Was the direction of relative motion taken to be a particular lattice direction, e.g., parallel to the x axis, or were all possible azimuthal directions considered and an average taken ? Was the relative motion in a given direction taken to be unidirectional, or in a zig-zag such as occurs in slip of close-packed sheets of atoms in metal slip planes over each other ((0001) hexagonal or (1 11) f.c.c.metals)? The long-chain molecules here are in a pseudo- hexagonal lateral packing, and analogous zig-zag slip may be expected. In our recent results we have shown that elastic hysteresis losses associated with periodic motion normal to the sliding direction (the units of one sheet moving in and out of a sequence of potential troughs as they slide over the opposing sheet of close- packed units) can contribute appreciably to the friction.Would not such a contribu- tion be expected similarly for sliding of sheets of hydrocarbon molecules over each other, and did their calculations include such periodic normal components or com- pressions along the molecule chains or their methyl end groups? With respect to the experimental values being much lower than those calculated, and this being attributable to the presence of dislocations, it is indeed probable with these layer structures that rotational slip would occur to an appreciable extent, and result in easier slip thereafter, giving rise to a series of relative azimuthal rotations such as those observed by Wilman.2 Dr. B. J. Briscoe and Dr.D. Tabor (Cambridge University) said: Cameron and Sutcliffe have embarked on a difficult problem. We have two comments. Both are concerned with the models they have chosen for the mode of shear in these materials. First, they consider that shear in a multilayer occurs between methyl groups midway between the two substrates. We wonder if this is a reasonable assumption. Recently, Dr. Israelachvili has carried out a series of experiments in our laboratory which seem to indicate that this type of shear is rarely realized. In his experiments he used mica crossed cylinders, the surfaces of which were covered with Langmuir- Blodgett layers of stearic acid. He was able to subject these layers to compressive and shear forces while observing the total film thickness to f0.2 nm using multiple beam interferometry. With a monolayer on each mica surface he found that the total film thickness was about 5.5 nm and the layers were stable both in compression, and shear. However, when three layers of acid were deposited on the surfaces the situation was rather different.The layers were stable to compressive forces, but the structure of the layers was destroyed immediately a shear force was applied. In these experiments the contact pressures were comparatively low (about lo7 N m-2) and the sliding speeds were of the order of mm s-l. In engineering systems, both the G. B. Jeffery, Proc. Roy. SOC. A , 1923, 102, 161. H. Wilman, Nature, 1950, 165, 321 ; Proc. Phys. SOC. A , 1951, 64, 329; A. D. Whapham J. Inst. Metals, 1956, 84, 109; A. D.Whapham and H. Wilman, Nature, 1955, 176, 460; Proc. Roy. SOC. A , 1956, 237, 513. ' P. V. K. Porgess and H. Wilman, Proc. Roy. SOC. A , 1959,252,35.60 GENERAL DISCUSSION pressure and the sliding velocity will be generally much higher. Hence we would expect that the structure of multilayer films would always be destroyed in these cases. Further, the fact that all surfaces are significantly rougher than mica would enhance this re-ordering. Secondly, apart from the experiment described above, we wonder if one ever has the situation where one Langmuir-Blodgett layer slides over another. We have found that a Langmuir-Blodgett layer deposited on glass has similar shear properties to the bulk material. There are perhaps two reasons for this similarity. (i) During sliding, additional material on the surface finds its way into the zone of contact and one is then sliding not on a monolayer, but on a wedge of material.Within the wedge (if it is formed), the molecules appear to be orientated with their chains parallel to the direction of sliding (see arguments in our paper). We also recall the observations of Flack which we described in reply to the question from Wilman on our paper. Flack was able to study the orientation of long molecular chains in polymers during shear using X-ray techniques. He found that there was considerable orientation of the chains in the direction of shear. Where this material comes from is uncertain when initially only one monolayer is present, but since the area of contact is small, very little material (less than 1 pg) would be required to provide a surface separation of 100 nm.(ii) The energy required to move a series of methyl groups over one another (Cameron and Sutcliffe model) is similar to that required to move methylene groups in the form of extended chains over each other. In their paper the authors mention that they can adjust p, the angle between the normal to the surface and the axis of the hydrocarbon chain. Could be set at n/2, or would a new model be required? We raise this point because Flack carried out a theoretical calculation of the forces involved in sliding long-chain molecules over one another in a direction parallel to the chain axis. His calculated shear strengths turn out to be about 10 times bigger than the observed values and are thus comparable with the values quoted by the authors for a very different molecular arrangement.Prof. W. C. Wake (City University) said: In reply to Young, I accept that a flow process can give rise to boundary zones free from suspended particles but the extent of flow during compression moulding is very small particularly at very low shear rates. For the adhesive supported on glass cloth there will be no flow in the centre of the annulus of the test piece. Both film and coated adhesive show filler particles in their surface as supplied even though a flow process would be involved in manufacture. I think the shear rates are too low to show the effect found by Griffin. The particle size of the aluminium in the adhesive was larger than 30pm and the carbon black very considerably smaller.Prof. D. D. Eley (Nottingham University) said: The presence of surface OH on rutile has been established by i.-r. spectrometry and I wonder whether they play a role in adhesion by hydrogen-bonding to the epoxy resin considered by Allen et aZ. ? Dr. H. Wilman (Imperial CoZZege, London) said : In the paper by Lin and myself, we had hoped to include a measurement of the coefficient of friction p of the Vickers diamond pyramid during ploughing along the surface of metals work-hardened by abrasion. We have now made this determination and find that on both freshly abraded Mo and Ni, ,u is about 0.47 when the load is half the hardness of the surface region in grams, and ploughing is parallel to the side of the square indentation. To make this measurement we held the front section of the microscope objective (on the front face of which the very small diamond for micro-indentation tests is mountedGENERAL DISCUSSION 61 centrally) in a collar having on two opposite sides extensions to about 4 cm length horizontally then bent down vertically to about 4 cm length so that a weighted scale pan could be hung from these extensions so as to load the diamond in a stable manner.A horizontal force to overcome the friction was applied by weights on a second scale pan on a thread passing over a low-friction pulley and attached low down close to the diamond. This observation makes it clear that the conditions are more complex than was assumed tentatively in our paper, which led to the estimate of s / p e 1.7 and p expected to be 3-4.There are four main possibilities to account for the observed p e 0.47 : (1) the real area of interface contact on the diamond is appreciably less than the apparent area (and is x times this) ; (2) the interface is mainly between the surface metal oxide and the diamond, and must have a low mean s/p - 0.4 or less ; (3) the flow pressure p of the piled-up metal just ahead of the ploughing indenter may be less than for the bulk metal, since it is projecting and thus under less constraint from any surrounding metal; (4) p under the dynamic conditions of ploughing is higher than for static indentation. Consider the first possibility. If the true contact area is x times the apparent area, eqn (l), (2), (3) and (5) must be replaced by : (1') w = ~ X A ; - ~ ~ A I , cot 81, F = sx(2A, +A,) + sxA,(sin 8' - 1) + pxA, cos 6', (5') but eqn (6) for p remains unchanged, thus fig.15 still applies. If, therefore, we use fig. 15 to conclude that the observed p of 0.47 must correspond to s/p-0.3 (for 8' = 68", i.e., 8 = 74", and A;/A; 1 0 . 4 7 as observed), then from eqn (3') we find x e 0.84 for Mo and 0.71 for Ni. If s/p = 0.4, we obtain only slightly different values, x = 0.86 for Mo and 0.72.for Ni. Thus, it appears that the observed p can indeed be accounted for mainly by (1) the true contact area being only 0.7-0.85 of the apparent area, and (2) a low s/p - 0.3-0.4 for the metal at the interface, this presum- ably being thus really metal-oxide surface film contacting the diamond (cf. (metal+ oxide) contacting (metal + oxide) with p == 0.28 in the experiments of Goddard, Harker and Wilman,l and Avient, Goddard and Wilman).2 We have already drawn attention in the paper to the presence of some furrows at the interface in the grooves ploughed on these rough work-hardened metal surfaces, hence this seems a reasonable explanation for the p value.The third possibility is difficult to assess, and it may be of only minor importance ; and the fourth possibility seems unlikely to have much effect because, although early work of Bowden and Tabor (see ref. (20), part I, 1950, p. 93) suggestedp for indium is increased considerably under dynamic conditions, other results such as those from shock-wave experiments on metals indicate only very small increases in yield stress as compared with that at lower loading rates. Dr. R. G. Linford (Berkeley Nuclear Lab.) said: Fig. 2 of the paper of Lin and Wilman shows a Talysurf trace of the groove. This is perhaps misleading as the magnifications in the horizontal and in the vertical directions differ by a factor of three for this instrument and the groove is therefore considerably less acute than shown. J. Goddard, H. J. Harker and H. Wilman, Nature, 1959, 184, 333. B. W. E. Avient, J. Goddard and H. Wilman, Proc. Roy. Soc. A , 1960, 258, 159.62 GENERAL DISCUSSION Interpretation of Talysurf traces often leads to confusion : for example, metal sur- faces prepared by normal engineering techniques yield profiles looking as jagged as the Alps whereas such surfaces more closely resemble the rolling hills of the Lake District. Dr. H. Wiiman (Imperial College, London) said: I thank Linford for reminding me that we had omitted to state the scales for the Talysurf trace in fig. 2. In fact, the vertical magnification is x 5000, while the horizontal magnification is x 200. The angle contained by the sides of the groove is 136", in reality, like the dihedral angle of the diamond pyramid (cf. fig. 4).