Wear, 118 (1987) 113 - 125
THE EFFECT OF SURFACE
J. V. REID Metallurgy Laboratory,
Reynolds Metals Company, Richmond,
J. A. SCHEY Department of Mechanical Engineering, N2L 3Gl (Canada) (Received January 21,1986;revised
University of Waterloo, Waterloo, Ontario
February 10, 1987; accepted March 31,1987)
summary The influence of bulk hardness, surface hardness and strain-hardening rate on the frictional behavior of Cu-Sn and Cu-Al alloys was investigated. Unlubricated specimens were subjected to severe relative sliding against tool steel (D2) and aluminum bronze (Ampco 25) anvils under constant interference conditions. A microindentation hardness survey, from the surface into the bulk, was carried out on sections of the specimens both before and after testing. The highest hardness was measured immediately below the wear surface; this hardness value was found to increase linearly with an increasing instantaneous strain-hardening rate. However, no correlation could be detected between the coefficient of friction and the bulk hardness, the strain-hardening rate or the surface hardness. There was a general trend for the coefficient of friction to increase with the greater increase in hardness produced during sliding against Ampco 25. Drawing on observations of metal transfer and surface damage made earlier, this increase could be attributed to the formation of junctions that were stronger than the parent material. There was no correlation with hardness increase when sliding against D2 and the coefficient of friction was often, but not always, lower. Thus a hardness increase during sliding led to increasing friction only with the metallurgicahy highly compatible pairs while differences in compatibility rather than hardness increase governed the magnitude of friction with less compatible pairs.
1. Introduction Adhesion between two materials is quantifiable by bringing two bodies together under a preset load and then separating them again; the ratio of the separating force to the applied force is the adhesion coefficient. The presence of even the thinnest surface films makes for a poor reproducibility for such experiments; therefore, in most studies of adhesion the specimens are 0043-1648/87/$3.50
0 Elsevier Sequoia/Printed in The Netherlands
twisted while in normal contact. This helps to break through any surface fihns that are present. From the measured normal separating force a coefficient of adhesion is again calculated, even though it is understood that the magnitude of the separating force is affected not only by adhesion but also by junction growth, surface roughening and the degree to which interface films are interrupted by the sliding (rotation) of the specimens. In most practical applications, separation of the contacting surfaces takes place not by normal separation but by further sliding. Thus for technological purposes it is preferable to utilize a test in which both the encounter and the ~p~ation of the contacting surfaces occur in sliding. This has the further advantage that deformation of the specimens can be chosen to assure the complete penetration of the surface films. A coefficient of adhesion cannot be determined; however, for practical purposes, it is more important to know the effects of adhesion on metal transfer, surface damage and the consequent increase in sliding force (or, if the sliding force is divided by the normal force, the increase in coefficient of friction). Therefore factors influencing the adhesion of copper alloys were studied in a sliding test apparatus. One of the specimens (specimen A) was short, ground with a small (2”) taper and was representative of tool materials. The other specimen (specimen B) was long and had a surface parallel to its base (Fig. 1). In the course of testing, specimen B slides over specimen A with sufficient deformation to assure penetration of the surface films. In work reported in a previous paper [l], alloys were selected to test current theories of adhesion, including the effects of stacking fault probability, hardness, strain-hardening rate, composition and microstructure.
Fig. 1. Configurations
are in millimeters).
115 TABLE 1 Composition
of specimen materials
Cu-4Al Cu-6.5A1 Cu-8Al Cu-5Sn Cu-9Sn Cu-13Sn Cu-30Zn 69.12 Cu-20Ni 76.7 D2 Ampco 25 79.25
4 6.5 8 5 9 13
Commercially pure copper, Cu-Sn, Cu-Al, Cu-Zn and Cu-Ni alloys (specimen B) were tested against D2 steel and Ampco 25, a multiph~ aluminum bronze (specimen A) (Table 1). All tests were performed unlubricated and at a relative velocity of approximately 1 cm s-i. The normal force, dependent on the predetermined interference and on the extent of metal transfer in the contact zone, varied from 20 - 40 kN and was always sufficient to cause plastic deformation of the surface of the softer B specimens. Adhesion was judged from the coefficient of friction (shear force divided by normal force), surface damage to specimen B and metal transfer to the harder specimen A. Although none of the current theories could explain all the results, the onset of metal transfer to specimen A could be correlated, for all combinations tested, with the ~ompatib~ty ratings of Rabinowicz  extended for alloys. The coefficient of friction and surface damage to specimen B decreased with increasing hardness, the ultimate tensile strength and the yield strength of untested specimens and with the stacking fault probability but only for the case of copper alloys sliding against Ampco 25; no clear trends emerged in sliding against D2. In the course of sliding, the surfaces undergo strain hardening; this must influence metal transfer and surface damage. The purpose of the present ~v~t~ation was to explore and, if possible, quantify the effects of surface hardening as opposed to bulk hardness_ 2. Exper~en~
2.1. Testing of specimens In preparation for exploring the hardness changes resulting from sliding, Cu-Al and Cu-Sn alloy specimens (specimen B, Table 2), in “before” and
Before polishing with alumina Standard polishing Repolished with alumina Before polishing with alumina Standard polishing Standard polishing Standard polishing Standard polishing Standard polishing Standard polishing Standard polishing Repolished with alumina Standard polishing Standard polishing Repolished with alumina Standard polishing Repolished with alumina Standard polishing
Cu-6.5Al Cu-8Al Cu-5Sn
D2 Ampco D2 Ampco D2 Ampco D2 Ampco
25 25 25
5*8 ?+a 3 + 10 85 + 2 22 + 13 IO + I.1 55 I! 11 37 + 8 22 + 7 34 f 6 120 95 39 146 128 116 112 74
f f f f f i: k f
9 12 5 18 5 6 5 7
“after” test conditions, were sectioned at a 45” angle to the wear surface and polished following the procedures outlined by Ahn et al. , A protective lacquer coating (‘“Microstop” by Michigan Chrome and Chemical Co.) was used to protect the wear surface during sectioning, using a diamond wheel of 102 mm diameter. Heating was limited by allowing 15 - 25 min to cut through the 3 mm thickness at a speed of 300 rev min-‘. Kerosene was used for lubrication and cooling. The specimens were placed so that the cutting wheel applied a compressive force across the wear surface of the specimen. Next, the specimens were plated, commercially, with a coating of nickel (0.127 mm} to aid edge retention during grinding and polishing. Plated samples were mounted in slow-curing epoxy at room temperature under an 83 kPa vacuum to remove bubbles. Wet grinding was performed manually, with light pressure and low speed, on 240, 320, 400 and 600 grit Sic papers. Specimens were ultrasonically cleaned in water for 5 min after each grit. The direction of motion was maintained approximately normal to the wear surface and in a direction from the plating towards the bulk material. The same direction of motion was used during rough polishing with diamond pastes (7 and 1 pm) and Struers DP lubricant (an oil emulsion) on DP cloth and during final polishing with alumina (0.3 and 0.06 pm) and water on microcloth. An attempt was made to follow the general rule of doubling the polishing time for each successive operation f4] in order to decrease the amount of strain hardening introduced by polishing. Specimens were ult~onic~ly cleaned in a 3% solution of an alkaline cleaner (Alconox) after each diamond paste step and
in water after each alumina powder step. Subsequently, the specimens were rinsed with warm water, then ethanol and dried with forced air. This sequence is referred to in the following as “standard polishing”. A microindentation hardness survey, using a Leitz hardness tester on antivibration mounts, was carried out on each polished section. The mounted specimens were located in a fixture so that the specimen surface was normal to the indenter within 0.06”. Using the Knoop indenter, a 25 gf load was applied for 5 s. The rate of descent was adjusted so that a minimum of 20 s was required to lower the indenter; prior to loading, the indenter was approximately 0.4 mm above the surface when focused with a 4O:l objective. Indentations were made at 12 locations ranging from 0.025 to 2.54 mm distance from the wear surface on the 45” angle; this corresponds to 18 3.796 pm actual depth below the wear surface. A minimum of seven readings were taken at a depth of 18 pm and a minimum of three readings at other locations. Readings were accepted only for microindentations that met the following criteria: a difference between semimajor diagonals of less than 10%; no excessive rounding or bulging of any of the apexes; no excessive rounding of the sides of the indentations; the existence of a sharp outline of all edges. 2.2. Testing of rolled specimen materials To calibrate the change in surface hardness in terms of true strain, four strips (of 10 mm width and 3 mm thickness) of each of the Cu-Al and Cu-Sn alloys were cold rolled on a laboratory rolling mill with a liberal quantity of a low viscosity lubricant applied to the rolls and both sides of the strips. Reductions ranged from 5% to 40%, corresponding to true strains of 0.04 - 0.5. Samples of “as-received” material and of the rolled strips were mounted in slow-curing epoxy, under vacuum, at room temperature. Grinding on Sic papers and polishing with diamond pastes were carried out as previously described; the direction of grinding and polishing was random. These samples were not polished with alumina since this was found to be unnecessary. A min~um of seven microindentations, using the Knoop indenter, were performed on each sample.
3. Results and discussion 3.1. Results of hardness tests on specimens To illustrate the results obtained on copper alloy specimens, plots of Knoop hardness number (KHN) us. depth below surface (to 540 pm) are given for Cu-4A1, before and after testing against D2 and against Ampco 25, in Fig. 2. Error bars show the interval of the mean calculated for each point at a 90% confidence level, assuming that the mean and standard deviation of the population are unknown. As expected, the highest hardness was measured immediately below the wear surface, i.e. at 18 pm depth.
Fig. 2. Change in KHN with depth below sliding against D2 (0) and Ampco 25 (0).
test (x) and after
The difference in the means of KHN was calculated, for a depth of 18 pm below the surface, for all copper alloy specimens tested against D2 and Ampco 25 (Table 2); again, a 90% confidence level was used. This number represents AKHN, the increase in surface hardness resulting from the sliding test. 3.2. Repeatability and accuracy of microhardness testing The microhardness test is very sensitive to the surface preparation of the specimen. Buckle  has demonstrated the importance of the effects of mechanical polishing; hardness values can vary by up to 50% with polishing procedure and time (the values decrease as the cold-worked layer is removed). This is demonstrated also by the data in Table 2; repolishing the copper alloy specimens (with alumina) decreased the changes in surface hardness. Electrolytic polishing is reported  to be the best method of surface preparation for hardness testing but longer etching times are required to remove the cold-worked layer. The only criterion for judging the completion of the operation is hardness. The cold-worked layer is completely removed when hardness values remain constant on continued electrolytic polishing. Electrolytic polishing was attempted in this study but excessive pitting made it difficult to obtain a surface satisfactory for hardness testing. Nevertheless, hardness values were measured on the electropolished surfaces of rolled Cu-4Al specimens and these gave a KHN us. strain curve that was below that of the mechanically polished specimens (Fig. 3). It is, therefore, quite likely that a work-hardened layer was present in all of the mechanically polished specimens and that the true surface hardness was lower than measured. However, only the differences in hardness (AKHN) are used for the evaluation, thus minimizing the influence of work hardening due to surface preparation. The anisotropy of the material also affects the microindentation hardness tests . This is illustrated by hardness values measured on Cu-13Sn (Fig. 4); a higher AKHN was measured when the major diagonal of the indenter was parallel to the rolling direction. All other measurements were made with the major diagonal of the indenter parallel to the rolling direction. A problem encountered in polishing the specimens was the hardness of the nickel plating relative to the copper-based substrate. The substrate
Fig. 3. Increase in surface hardness (KHN) as a function of strain produced by rolling Cu-4Al. Surfaces prepared by mechanical polishing (0) and electropolishing (0). Fig. 4. Increase in surface hardness (AKHN) as a function of strain for readings taken with the major diagonal of the indenter parallel (0) and perpendicular (0) to the rolling direction.
polished faster, leaving the nickel standing proud. A plating that is similar in hardness to the substrate would have been preferable. An attempt was also made to polish a mounted but unplated specimen; however, rounding of the edge occurred to an unacceptable degree. It is apparent that a great deal of time and patience are required to obtain accurate readings from microindentation tests. The values reported in this investigation may not be the lowest obtainable values; nevertheless, they represent true differences between bulk and surface hardness and are thus suitable for evaluating differences in material combinations. 3.3. Correlation of coefficient of friction with hardness There is no correlation at all between the actual hardness (KHN) at 18 E.trnbelow the wear surface and the coefficient of friction (Fig. 5(a)). Thus the general notion that the adhesion and, through the effects of adhesion on metal transfer and surface damage, coefficient of friction should be lower for harder materials could not be substantiated. However, there is a general trend, at least when sliding against Ampco 25, for the coefficient of friction to increase (rather than decrease) with greater strain hardening, as evidenced by the greater increase in hardness (AKHN) measured at 18 pm depth (Fig. 5(b)). This correlation implies, at least for metallurgically compatible pairs, that a large increase in surface hardness produced during sliding may cause the formation of junctions that are stronger than the parent material, resulting in a high coefficient of friction. An alternative explanation would attribute the increase in coefficient of friction to junction growth or ploughing, both of which would increase the shear force. However, they should also result in greater surface damage, yet no clear correlation between the increase in surface hardness (AKHN) and the extent of surface damage or metal transfer to the specimens, as reported earlier [l],could be found. The correlation for sliding against D2 was poor (Fig. 5(b)). Earlier work [l] has already shown that none of the usual measures of hardness could be correlated with the coefficient of friction. The present results show that there is no correlation with surface hardening either.
250 300 KHN
SPECIMEN B Cu-4AI Cu -4 Al IRE1 Cu-65Al CU-BAI cu- 5% Cu- 5% (RE) cu - 9% Cu-9SnlAEl cu - 13% Cu -13Sn (REI
SPECIMEN AWCU 25 cl
I Repohskd Specimen B
Fig. 5. Correlation of coefficient of friction with hardness at 18 pm depth below tested surface. (a) Coefficient of friction us. KHN, (b) coefficient of friction us. increase in KHN and (c) strain-hardening rate us. increase in KHN.
It should be noted that the compressive strength of Ampco 25 (1550 MPa, 39 HRC) was three times that of the hardest copper alloy, hence it should have given low friction if hardness rather than compatibility was the dominant criterion. The tool steel D2 was harder (3640 MPa, 59 HRC) and its compatibility was lower with the Cu-Sn alloys than with the Cu-Al alloys [l] and this difference in compatibility, rather than the surface hardening of the copper alloys, was reflected in the coefficient of friction. 3.4. Correlation of strain-hardening rate and surface hardness In an attempt to rationalize the differences in AKHN, a correlation was sought between the increase in surface hardness and the strain-hardening
-I 4 LOG
Fig. 6. Cu-4A1,
Logarithm of stress us. logarithm O; Cu-6.5A1,O; Cu-8Al,& Cu-5Sn,
of strain curves for specimen q; Cu-SSn, P; Cu-13Sn, ti.
121 TABLE 3 Strain-hardening rates for copper alloys Specimen Cu-4Al Cu-6.5Al Cu-8Al Cu-5Sn Cu-9Sn Cu-13Sn
0.43 0.36 0.34 0.42 0.48 0.41
0.29 0.36 0.31 0.36 0.40 0.41
4 (vs. Ampco
0.43 0.36 0.26 0.42 0.48 0.41
an equals strain-hardening rate from tension test for E > 0.1. bni (vs. D2) equals strain-hardening rate over the strain range corresponding to the strain induced in sliding against D2. ‘ni (us. Ampco 25), as above, but for sliding against Ampco 25.
rate. In a first attempt, the strain-hardening exponent n was calculated from the results of tension tests (Fig. 6) for strains greater than 0.1 (Table 3, n value). This gave a good correlation against AKHN for specimens slid against Ampco 25 but not against D2 (Fig. 5(c)). It was already known [ 11 that the lower adhesion against D2 resulted in less metal transfer and surface damage, even with alloys of high strain-hardening rate, and this could mean that specimens slid against D2 suffered less strain hardening. Since the copper alloys did not exhibit a constant strain-hardening rate (Fig. 6) it could then be argued that the correlation might be improved by taking the strain-hardening rates prevailing at the lower strains induced during the sliding test. To determine the level of strain, induced during the sliding test, the following procedure was adopted. (a) All Cu-Al and Cu-Sn alloys were rolled to different reductions to produce various strains and their KHN was measured, as in the example given in Fig. 3, where the interval of the mean is shown at a 90% confidence level. (b) The relationship between a change in surface hardness and strain was determined by subtracting the hardness at zero strain (before-rolling condition) from the hardnesses at specific strains produced by rolling. Thus the change in surface hardness (AKHN) is plotted against strain for each alloy in Figs. 7 and 8. (c) Changes in surface hardness resulting from the sliding tests (Table 2) are included in Figs. 7 and 8. Since, at this point in the argument, the actual surface strain is unknown, the symbols are placed at convenient but arbitrary strain values. In some cases (e.g. Fig. 8(a)), AKHN measured in the sliding tests was greater than the maximum AKHN calculated from the microindentation tests on the rolled specimens, implying a strain greater than 0.5 in the surface layer. (d) Figures 7 and 8 were used to determine if the specimen B materials were subjected to a strain greater or less than 0.1 during the sliding test; that
Fig. 7. Increase in surface hardness (fLKHN) as a function of strain induced by rolling: (a) Cu-4Al; (b) Cu-6.5Al; (c) Cu-8AI. The increase in surface hardness measured on specimens tested against D2 (A) and Ampco 25 (n) is also shown by symbols placed at arbitrary strains. Fig. 8. Increase in surface hardness (AKHN) as a function of strain induced by rolling: (a) Cu-5Sn; (b) Cu-9Sn; (c) Cu-13Sn. The increase in surface hardness measured on specimens tested against D2 (A) and Ampco 25 (0) is also shown by symbols placed at arbitrary strains.
is, if there was any merit in attempting to correlate the coefficient of friction with the strain-hardening rate calculated for a strain lower than 0.1. Based on steps (a) - (d), Figs. 7 and 8 reveal that surface strains in the sliding test were lower than 0.1 in the following cases: Cu-4Al us. D2 (Fig. 7(a)); Cu-8Al us. D2 and Ampco 25 (Fig. 7(c)); Cu-5Sn US.D2 (Fig. 8(a)); Cu-9Sn us. D2 (Fig. S(b)), i.e. in half the cases of sliding against D2. The ins~~eous str~-h~eni~ rates ni were determined, for the strains extracted from the above figures, using a least-squares regression, assuming a linear logarithm of the stress us. logarithm of the strain relationship over the appropriate strain range. The coefficients and “goodness of fit” parameters for these c~cu~tio~ are included in Table 4; stm~-h~dening rates obtained for strains greater than 0.1 and for the calculated lower strain ranges are summarized in Table 3. There was a fairly good linear correlation between the instantaneous s~~-h~den~g rate and AKHN (Fig. 9(a)), indicating that the procedure
cu-4Al Cu-8A1 Cu-8Al Cu-5Sn Cu-9Sn
0.290693 0.262884 0.314909 0.359900 0.404966
2.770597 2.958092 3.010445 2.828171 2.941112
0.04 0.06 0.09 0.05 0.09
- 0.13 - 0.12 - 0.28 - 0.22 - 0.17
Standard error of estimate 0.008613 0.002765 0.003096 0.012206 0.006536
Coefficient of correlation 0.975038 0.993076 0.998388 0.984254 0.992909
Coefficient of determination 0.950699 0.968200 0.996778 0.968755 0.985869
Coefficients of stress-strain relationship lg u = A + B lg E for indicated strain ranges (linear least-squares regression used to fit data)
AKHN (a) D
0254 11 026
I I ! I / 1 I I I I I 0 30
Fig. 9. (a) Instantaneous strain-hardening rate as a function of the increase in surface hardness at 18 pm depth below the tested surface. (b) Correlation of the coefficient of friction with the instantaneous strain-hardening rate.
was correct. However, there was still no correlation between the coefficient of friction and the instantaneous strain-hardening rate (Fig. 9(b)) for sliding against D2. 4. Conclusions The purpose of the present investigation was to study whether surface hardness or strain-hardening rate have a significant effect on the frictional behavior of Cu-Sn and Cu-Al alloys as reflected in the coefficient of friction measured in severe sliding against a copper-based die material or D2 die steel. The specimens, in “before” and “after” test conditions, were sectioned at a 45” angle to the wear surface. Subsequently, they were nickel plated and mounted in slow-curing epoxy prior to polishing. A microindentation hardness survey was carried out from the specimen surface into the bulk. Surface hardness values were also measured on strips of the alloys that had been rolled to produce various strains. The microindentation hardness test was very sensitive to the surface preparation of the specimen (i.e. polishing procedure and time) and the anisotropy of the material.
It has often been argued that adhesion and friction should be less for harder materials. Yet, in this work, no correlation between bulk hardness and coefficient of friction could be found. It can then be argued that surface hardness should be of relevance. Indeed, the highest hardness was measured immediately below the wear surface but there was still no correlation between this hardness value and the coefficient of friction. Thus the coefficient of friction is not necessarily lower for harder materials; results reported earlier [l] indicate that metal transfer and surface damage, indicative of adhesion, are not necessarily lower either. A second line of argument states that strain-hardening rates are of importance. A satisfactory correlation between surface hardness and stminhardening rate could be obtained by calculating the instantaneous strainhardening rate for strains induced by the sliding test. In general, a greater increase in surface hardness produced during sliding against Ampco 25 resulted in a higher coefficient of friction, suggesting that junctions stronger than the parent material must have formed. No such correlation was found for sliding against D2. The present results on the magnitude of friction, combined with observations made earlier [l] on metal transfer and surface damage, indicate that a greater increase in surface hardness or a higher strainhardening rate do not imply either lower adhesion or lower friction. Thus, reluctantly, we must conclude that hardness (or the hardness increase in the surface layers) cannot be used as a criterion for predicting the coefficient of friction in unlubricated sliding involving severe contact and penetration of surface films. Metalhugical compatibility had been found to determine metal transfer [ 11; it now appears that sliding of highly compatible pairs results in the coefficient of friction increasing with increasing strain-hardening rate of the softer material, whereas compatibility alone determines friction with less compatible pairs. Acknowledgments This work was carried out under an operating grant of the Natural Sciences and Engineering Council of Canada. We are deeply indebted to Mr. Stephen Woodall, who carried out the metallographic preparation and hardness testing for this work. References 1 J. V. Reid and J. A. Schey, Adhesion of copper alloys, Wear, 104 (1985) 1 - 20. 2 E. Rabinowicz, ASLE Trans., 14 (1971) 198 - 205. 3 T.-M. Ahn, P. J. Blau, K.-L. Hsu, D. A. Rigney and J. D. Schell, Wear, 56 (1979) 409 - 413. 4 Metallography, Structures, and Phase Diagrams, Metals Handbook, Vol. 8, American Society for Metals, Metals Park, OH, 1975, p. 3. 5 H. Buckle, Metal!. Rev., 4 (1959) 49 - 100. 6 M. A. Meyers and K. K. Chawla, Mechanical Metallurgy, Prentice-Hall, Englewood Cliffs, NJ, 1984, p. 617.