Adhesion of Metals to Polymers BY C. WEAVER Department of Applied Physics, University of Strathclyde, Glasgow Received 19th June, 1972 Methods of measuring the adhesion of thin films to bulk substrates are critically reviewed, with particular emphasis on the value of scratch testing methods. Results obtained by this method for a range of metals and polymers show that in many cases there is a considerable electrostatic component of adhesion, which is both time and temperature dependent. An examination of electrical conduction in polyethylene shows space-charge-limited currents, due apparently to positive hole injection at the anode. It is suggested that a similar mechanisms is the cause of the adhesion results. Some of the major difficulties associated with any work on adhesion are experi- mental.In the first place, there is the difficulty of obtaining perfect contact between two surfaces, particularly in the case of bulk materials. The most careful preparation of any surface leaves minute irregularities which are still large compared with atomic dimensions even if adsorbed layers can be eliminated. Consequently, when two surfaces are brought together, contact occurs only at the high spots and even if extreme pressure is applied to force the surfaces into intimate contact, the elastic stresses which are released when the pressure is removed cause a partial separation of the surfaces. These effects can be reduced by using a soft, ductile material such as lead or indium for one surface and a ball-bearing or similar hard indenter for the other, as shown by Bowden and Tabor.’ Plastic flow occurs and the force of separation becomes equal to the original applied force.Similar results may be obtained by annealing two metals in contact under pressure. The cases where these methods may be used are obviously restricted and it is still difficult to ensure uniform contact and freedom from localized stresses over the entire interface. An alternative approach is to try to produce a layer in intimate contact with a solid surface by freezing a liquid in contact with the surface, allowing solvent to evaporate from a solution as in the case of many adhesives, or condensing material on the surface, but this raises problems in detachment. It has frequently been pointed out that the strength of adhesive joints may be explained quite adequately in terms of van der Waals forces alone, e.g., de Bruyne,’ and the universal nature of these forces would suggest this type of bonding should represent the lower limit of adhesion when a perfect interface is achieved.There are several factors which indicate that this lower limit should be of the order of lo9 dyn cm-2 (lo8 N m-’). Theoretically, calculations of London dispersion forces 3-6 give values from about lo9 dyn cm-’ (lo8 N m-’) upwards while Tabor has estimated the force required to strip a pure hydrocarbon liquid from a surface leaving an absorbed layer as 6 x lo8 dyn cm-2 (6 x lo7 N m-’), considerably greater force being required to leave a clean solid surface. Experimentally, shearing forces up to 4 x lo7 dyn cm-2 (4 x lo6 N m-’) were required to remove ice frozen in situ from a metal surface and failure invariably occurred in the ice itself,8 suggesting that the interface was considerably stronger and 18C.WEAVER 19 probably above the value of 1.2 x 10' dyn cm-2 (1.8 x lo7 N m-2) obtained in friction experiments with ice. Admittedly, lower values of the order lo7 dyn cm-2 (lo6 N m-2) were obtained for surfaces which were not perfectly wetted, but this suggests imperfect contact. In experiments on friction between solid surfaces it is generally found that fragments of material are transferred from one surface to the other. Since ultimate strengths of most mtaerials are of the order of lo9 dyn cm-2 (10' N m-2) the adhesion bonding must be of at least the same order.A more direct measure of adhesion was obtained by Beams who tried to measure the centrifugal force required to detach an electroplated film from the steel rotor of an ultra-centrifuge. In most cases the rotor disintegrated before detachment and successful results were obtained only when bonding was reduced to the van der Waals level by putting a layer of albumin or oil between the film and the rotor. Bond strengths of the order 1.1 x lo9 dyn cm-2 (1.1 x 10' N m-2) were then obtained. These arguments and facts lead inevitably to the conclusion that any adhesion experiments giving an ultimate strength significantly below the van der Waals level should be examined most carefully for possible explanations. Imperfect contact can easily occur and it is difficult to avoid stress concentrations which lead to pro- gressive separation or peeling.Where results show a wide scatter which cannot be explained by the normal experimental errors, it suggests that these factors occur and particular significance should be attached to the highest values rather than average values, as representing a closer approach to perfect conditions, even though the average (or even the lowest) values may have greater practical significance. Various attempts have been made to measure the adhesion of thin vacuum- deposited films on different substrates, partly because of the technological importance and partly because the deposition process produces almost perfect interfacial contact. The difficulties arise in trying to remove the film from the substrate in a manner which gives a reliable indication of the adhesion.The most direct approach is to stick, solder or cement to the film surface a probe, rod or block of material which may then be used to apply a force to the film. An extensive series of measurements was made by Belser lo using flat-headed pins or short aluminium cylinders to apply a pull normal to the substrate. Results showed a wide scatter and were consistently an order of magnitude or more below the value obtained by Beams for a poorly adherent film, i.e., below the van der Waals level. The method naturally raises questions of adhesive (or solvent) penetrating the film and reaching the film-substrate interface, stresses produced during setting of the adhesive causing partial detachment, and non- uniform stress distribution over the contact area during the pulling process, all of which would affect the final result.More recently Butler l1 has used a short rod of square section which was then toppled sideways about one edge of the square contact area, producing an effective tensile pull at the other side, and Lin l2 has bonded a flat glass plate onto the top surface of his films so as to produce a lap joint which was then pulled to produce a shear stress. Accurate calculation was difficult but Butler estimated tensile stresses up to 4 x lo7 N m-2 before failure of the adhesive. Lin l 2 obtained results showing scatter over more than an order of magnitude with average values of 3 x lo5 N m-2 for gold on glass, 8 x lo5 N for copper on glass, and 1.6 x lo6 N m-2 for aluminium on glass, and maximum values about five times greater (e.g., > 5 x lo6 N m-2 for A1 on glass).Electron diffraction could not detect any adhesive at the film-substrate interface, but the figures all appear low compared with -lo8 N m-2 for van der Waals bonding. It has long been known, however, that the stress distribution in a lap joint is quite complicated 14* l 5 and photo-elastic models of lap joints have shown concentrations which depend, amongst other factors, on the shape of the glue fillets at either end of the joint. Many workers 16-19 on the20 ADHESION OF METALS TO POLYMERS adhesion of thin films have used scratch testing, in which a smoothly rounded point or stylus is drawn across the surface of the film under a load applied normal to the sur- face.As the load is increased, a point is reached at which the film under the moving point is detached and the critical load above which a clear scratch is produced then forms a measure of the film-substrate adhesion. Experimental measurements and calculations have shown that the horizontal force required to move the point across the surface does not play a significant part in the process and there is a fundamental difficulty in explaining how a compressive force normal to the substrate produces detachment of the film from the substrate, but it was observed that the process always involved plastic deformation of the substrate by the point. In 1960, a theory was presented which showed that as a result of this deformation a shearing force was produced at the film-substrate interface around the rim of the indentation produced by the point and a relationship was developed between the applied load and the shearing force.The most convincing evidence that the critical load is determined primarily by the adhesion between the film and the substrate is provided by measurements of two- layer metal films on glass. Fig. 1, shows the variation of critical load with time for s a o l IOSOA I .1 / I 2 5 0 i ISSOi 0 9 18 2 7 36 45 54 time/h FIG. 1.-Adhesion changes during annealing of Au-A1 thin-film diffusion couples with gold under- layers. different thicknesses of gold film deposited on glass and then overlaid with aluminium. Gold has a relatively poor adhesion to glass and the measurements start at a low value. Aluminium has a much higher adhesion to glass and, as the two metals interdiffuse, an abrupt change in adhesion occurs at exactly the stage where some aluminium (in the form of an intermetallic compound with gold) appears at the metal-glass interface.This can easily be detected by observation through the glass substrate. There is an exact parabolic relationship between the gold thickness and the time as measured up to the break in the adhesion curve, in accordance with diffusion theory ; diffusion coefficients may be calculated from the measurements. If the same metals are used but deposited in reverse order with the aluminium in contact with the glass substrate, the same diffusion occurs but the results are completely different, being determined mainly by the aluminium underlayer as shown in fig.2, which shows adhesion meas- urements for the two-layer film and for the aluminium and gold films separately. The detailed reasons for the different adhesions in this diagram have been explained elsewhere. O During recent years, the development of scanning electron microscopy has allowed examination of the scratches in more detail than was previously possible. As aC. WEAVER 21 result, the method has been criticized 21* 22 primarily because the micrographs show film becoming detached a short distance ahead of the moving stylus, and to a certain extent along the sides of the track. It is difficult to reconcile this with the remarkable success which the method has had both in measuring adhesion and in detecting alloying effects in bimetallic films.23 It certainly indicates that the process is more complex than originally envisaged, but the explanation probably lies in the fact that vacuum-deposited metallic films are almost invariably in a state of tensile stress and for very thick films, which are poorly adherent, spontaneous detachment can occur wherein the film curls away from the substrate and peels under the action of the internal stress.This means that if film detachment occms under the stylus and particularly around the perimeter of the contact area where the interfacial shearing lo0l 8 0 I 6ot P 0 16 2 0 3 0 4 0 time/h FIG. 2.-Changes in the measured adhesion of an Au-Al thin film diffusion couple on glass with an aluminium underlayer, together with corresponding changes in adhesion of individual Au and A1 films to the glass substrate.force is a maximum, the intrinsic tensile stresses in the film immediately around the stylus tend to lift the loosened edge and peel the film away from the substrate. This may be expressed somewhat differently by saying that a Griffith’s crack extends away from the stylus along the interface, but the length of such cracks is limited. According to the theory of crack p r ~ g a g a t i o n , ~ ~ ~ 25 the smallest stress 0 capable of extending the crack is of the order 0 N, s(3a/Z)+ where s is the ultimate strength (in our case the adhesion), a is the lattice spacing, I is the crack length and Q would of course be given by the intrinsic film stress. Movement of the stylus would tend to push down the loosened edge, creating a loop or hillock of detached film immediately ahead of the stylus, but for a thin film the bending moment in this loop would be insufficient to produce further crack propaga- tion and the loose film would be pushed aside until the stylus reached the end of the Griffith’s crack.The shear stress under the stylus would then repeat the loosening process. The net result is very reminiscent of a stick-slip process and such rhythmic effects may be seen in micrographs of the 22 This explanation modifies and expands the detail of how a film is removed but the initial stage is still dependent on the shearing force under the point. Other complications may arise. Some degree of peeling can occur with very poorly adherent films and it may be difficult to obtain completely clear scratches with highly adherent films because maximum shearing22 ADHESION OF METALS TO POLYMERS effects occur at the edge of the scratch.In practice, it has been found that the best results are obtained with steel points, frequently renewed, rather than diamond or sapphire styli, which have failed to give consistent results according to our experience. It should be obvious that the model is a simplification of a complex situation but an important advantage is that it consistently leads to adhesive strengths which compare favourably with the best theoretical and experimental data. An entirely different approach to the measurement of adhesion is by peeling a film from a substrate. In an ideal case, e.g., splitting of mica sheet,26 this is a revers- ible process and in most cases it leads to an energy of adhesion rather than an ultimate force.Theoretically, the energy would be given by the sum of the surface energies of the new surfaces produced, less the original interfacial energy. Derjaguin 2 7 9 28 was probably the first to notice the discrepancies between the peeling energy and the theoretical values, and attributed the high peeling energy to the separation of charges in an electrical double layer at the interface. Energies of the order of lo5 erg cm-2 (100 J m-2) were obtained for stripping of a polymer film from a metal surface and evidence was presented of electrical discharges in the gap between the separating FIG. 3.-Changes in adhesion with time for metal films on a polypropylene substrate.x , copper ; 0, gold ; + , silver ; 0, aluminium. surfaces. The work has been strongly criticized by Huntsberger and by Gardon 29 on the grounds that most of the experimental work of peeling was dissipated in plastic deformation. Similar effects may be observed in cohesive failure, where there is no reason for postulating an electrical double layer. Any electrical effects observed could easily be attributed to the heavy deformation of the peeled layer since there is evidence of piezo-electric effects in The major criticisms would of course apply to the measured work in any peeling process and it is doubtful whether the peeling energy has any theoretical significance in adhesion. Some of the lowest peeling energies have been obtained by Chapman 31 who peeled gold film off glass substrates using a backing of adhesive tape.Energies of the order of 2000 erg cm-2 (2J m-2) were obtained, and when thicker gold films were used so that the backing tape could be discarded, the measured energies fell to about 500 erg cm-2 (0.5 J m-2). However, a rate dependence and a pressure dependence were still observed, which suggests that energy is being dissipated and the true adhesionC . WEAVER 23 energy might be much lower. Estimated values of van der Waals bonding energy for various metals on alkali halide faces 32 are of the order of 100-300 erg cm-2 (0.1- 0.3 J m-2). ADHESION OF METAL FILMS TO POLYMERS Scratch testing has been used to measure the adhesion of metal films to various polymer surfaces. The films were deposited by vacuum evaporation on smooth plastic surfaces.In some cases, a glow discharge was used to clean the surfaces of adsorbed layers, but in other cases the films were deposited on the surfaces as placed in the vacuum chamber. Most metals showed a change in adhesion with time after deposition but in many cases the change was small. The largest effects were con- sistently observed with copper silver and gold; fig. 3 shows the changes in adhesion with time for a polypropylene substrate which had been subjected to a cleaning discharge. Similar effects may be observed for many polymers. In some cases such as polymethylmethacrylate and polycarbonate, the effects are much greater and in other cases such as P.T.F.E. the effects are smaller, but the general trends remain the same and the same group of metals shows the greatest adhesion. The fact that gold is amongst the metals showing maximum adhesion makes any form of chemical reaction most unlikely.Tabulated work functions show too much scatter to form a reliable guide but comparisons of work functions by contact potential measurements has shown that copper, silver and gold have work functions which are almost identical. 902 Application of d i IC har 9 e I I c, tk - + -+ + *'- 0 2 0 0 4 0 0 600 800 time/h x , Cu on Perspex ; 0, Cu on NaCl ; + , Cu on glass. FIG. 4.-Adhesion of evaporated copper films on Perspex showing elimination of charged layers by an ionizing discharge. These facts in conjunction suggest the possibility of an electrical component of ad- hesion. Fig. 4 shows some results obtained for copper films which were first aged for 200 h to develop increased adhesion and then replaced in a vacuum chamber and subjected to a glow discharge for a few minutes.The glow discharge produced no effects for copper on sodium chloride, which shows van der Waals bonding only, or for copper on glass, which shows an initial increase in adhesion due to oxide formation, For copper on Perspex, the high adhesion which had built up with ageing was reduced to van der Waals level and started to build up again. The process could be repeated. The manner in which the increased adhesion may be wiped out by an ionizing24 ADHESION OF METALS TO POLYMERS discharge provides almost perfect confirmation of the electrical nature of the in- creased adhesion, but does not explain the mechanism of charge transfer between the metal and the polymer.Derjaguin assumed electron transfer from the metal to the polymer so as to equalize the Fermi levels but this should be a rapid process compared with the ageing periods involved here, if it occurs at all. The conduction band in most insulators is no more than 1-2 eV below the vacuum level so that an electron at the Fermi level in a typical metal with a work function of the order of 4-5 eV would be faced with a potential barrier of the order of 3 eV. The chance of a metal electron having excess energy in this range is negligible. Measurements in our laboratories of conduction in polyethylene have shown space-charge-limited currents with gold and aluminium electrode~,~~ with a carrier mobility of the order of 3 x lo-* cm2 V-1 s-l (3 x m2 V-I s-l) at room temperature. When two different electrodes, having different conduction characteristics, were used on the same specimen, the observed conduction was always typical of the positive electrode.In conjunction with the other features this indicated positive hole injection at the gold and aluminium elec- trodes. Similar effects have been observed in less detail with polypropylene. This suggests that the charge transfer in adhesion is due to positive hole injection, i.e., electron transfer from the valence band of the polymer to the metal with the injected charge causing band bending until the Fermi levels equalize. This would of course imply that the effective Fermi level in the polymer normally lies above the Fermi level in the metal when these adhesion effects are observed. Any reasonable estimate for the Fermi level in the polymer would still leave a large difference between the valence band and the Fermi level in the metal so that, even allowing for band bending, there would be a substantial barrier to hole injection if the holes are injected directly into the valency band, but the space-charge-limited current characteristics show a total trap density 4 x lo1* ~ m - ~ (4 x m-3) expo- nentially distributed and the low mobility is typical of a hopping mechanism between traps.There seems little doubt that these traps play a part in charge transfer and adhesion as well as conduction. When a positive hole is trapped the trapping centre releases an electron, normally into the valence band or to another trap, but a trap sufficiently close to the polymer surface could release an electron directly to the metal electrode by tunnelling.The trapped positive hole would subsequently migrate away from the interface under the space-charge field, leaving the original trap free to repeat the process. This would allow the transfer of an electron to the metal at an energy level somewhat below the Fermi level where there is a reasonable chance of finding a vacant energy state, and the rate of charge transfer would be governed by the density of available vacant energy states in the metal as well as the trap density and energy. The problem of why copper, silver, gold, and to lesser extent aluminium, should show these charging effects to a much greater degree than other metals is a more difficult question which has not been completely resolved but these are all metals with odd numbers of electrons/atom and half-filled energy bands.Most other metals, apart from the alkali metals, conduct because of band overlap and have one band almost completely filled, with a spill-over into the next higher band. An elementary argument based on possible wave-vectors towards the top of the energy distribution suggests that there could be momentum restrictions on possible transfers, particularly if the metal shows some orientation effects such as fibre orientation about an axis perpendicular to the substrate. In presenting this paper I have drawn together the work of several research students, which I should like to acknowledge and I should like to thank Dr.G. A. P. Wyllie of Glasgow University for some helpful discussions.C. WEAVER 25 The continued support of the Ministry of Defence for this work is gratefully acknowledged. F. P. Bowden and D. Tabor, Proc. 2nd Znt. Congr. Surface Activity (Butterworth, London, 1957), 3, 386. N. A. de Bruyne, J. Sci. Znstr., 1947, 24, 29. P. Benjamin and C. Weaver, Proc. Roy. Soc. A, 1960, 254, 163. D. Taylor, Jr and J. E. Rutzler, Jr, Znd. Eng. Chem., 1958, 50, 928. J. R. Hunstsberger, Treatise on Adhesion and Adhesives, Vol. 1, ed. R. L. Patrick (Arnold, London, 1967), p. 21. B. N. Chapman, Ph.D. Thesis (Imperial College, London). D. Tabor, Rep. Prog. Appl. Chem., 1951,36, 621. L. E. Raraty, Ph.D. Diss. (Cambridge, 1955).J . W. Beams, 43rd Ann. Proc. Amer. Electroplaters SOC., 1956. 42453, 1954. R. B. Belser and W. Hicklin, Rev. Sci. Znstr., 1956, 27, 293. lo R. B. Belser, Interim Rep. No. 7, Project 163-176, U.S. Ordnance Contract DA-36-039-Sc- l2 D. W. Butler, J. Phys. E, 1970, 3, 979. l3 D. S. Lin, J. Phys. D, 1971,4, 1977. l4 C. Mylonas and N. A. de Bruyne, Adhesion and Adhesives, ed. N. A. de Bruyne and R. Houwink (Elsevier, Amsterdam, 1951), p. 91. L. Greenwood, T. R. Boag and A. S. McLaren, Adhesion, Fundamentals and Practice (Ministry of Technology, McLaren and Sons Ltd., London, 1969), p. 273. l6 0. S. Heavens, J. Phys. Rad., 1950, 11, 355. l7 P. Benjamin and C. Weaver, Proc. Roy. SOC. A, 1960,254,177 ; 1961,261,516 ; 1963,274,267. l 8 M. M. Karnowsky and W. B. Estill, Rev. Sci. Znstr., 1964, 35, 1324. l9 D. M. Mattox, J. Appl. Phys., 1966, 37, 3613. zo C. Weaver and D. T. Parkinson, Phil. Mag., 1970, 22, 377. 21 D. W. Butler, C. T. H. Stoddart and P. R. Stuart, J. Phys. D, 1970,3,877. z 2 D. W. Butler, C. T. H. Stoddart and P. R. Stuart, Aspects of Adhesion, Vol. 6, ed. D. J. Alner 23 C. Weaver and R. H. Hill, Adv. Phys., 1959,8,375. 24 A. H. Cottrell, The Mechanical Properties of Matter (Wiley, London, 1964). 2 5 E. Orowan, Reports Prog. Phys., 1949, 12, 185. 26 A. I. Bailey, Proc. 2nd Int. Congr. Surface Activity (Butterworth, London, 1957), 3, 406. 27 B. V. Derjaguin, Research, 1955, 8, 70. z8 B. V. Derjaguin and V. P. Smilga, Proc. 3rd Znt. Congr. Surface Activity (Butterworth, London, 29 J. L. Gardon, Treatise on Adhesion and Adhesives, Vol. 1, ed. R. L. Patrick (Arnold, London, 30 Yu. N. Novikov and F. T. Polovikov, Soviet Phys.-Solid State, 1966, 8, 1240. 31 B. N. Chapman, Aspects ofAdhesion, Vol. 6, ed. D. J. Alner (Univ. Lond. Press, 1971), p. 43. 32 P. Benjamin and C. Weaver, Proc. Roy. SOC. A, 1963, 274,267. 33 T. McGrenary, unpublished work, University of Strathclyde. 34 D. T. Morrison, Ph.D. Thesis (University of Strathclyde 1970). (Univ. Lond. Press, 1971), p. 55. 1960), 2B, 349. 1967), p. 320.