Strength and Failure Patterns of Metal-Metal Adhesives BY K. W. ALLEN, H. S. ALSALIM AND W. C. WAKE The City University, St John Street, London, EClV, 4PB Receiued 12th June, 1972 Torsional shear napkin-ring test-pieces have been used in adhesion studies with two titanium alloys, modified epoxy and epoxy-phenolic adhesives. The temperature profiles have been obtained and the failure surfaces studied. The metallography of the surfaces is recorded and the effect of a variety of preparative surface treatments. Most treatments lead to rutile, and in one case rutile needles bond to the adhesive more strongly than to the substrate. Surface roughness is shown to enhance bond strength. To achieve understanding about adhesives and to supply data for design purposes, stress rupture experiments have been used, most frequently based on a lap joint stressed in tensile shear.The stress distribution in this joint is complex, and torsional shear provides a simpler system and was advocated in 1951 by de Bruyne.l A torsional shear test associated with a cleavage test, such as peeling a flexible adherend, should enable adhesive design to be placed on a more rational basis. temperature/"C FIG. 1 .-(Strength, temperature) profile of Redux 775 adhesive. Two epoxy-based adhesives have been tested in torsional shear and their behaviour on two titanium alloys studied in detail. Napkin-ring test-pieces and apparatus for torsional shear testing are described el~ewhere,~-~ as also are the expressions used 38K . W. ALLEN, H. S . ALSALIM AND W . C .WAKE 39 for calculating the breaking stress from the observed t ~ r q u e . ~ In the present work, all calculations assumed elastic failures even though at elevated temperatures this is unlikely. The error is of no consequence as absolute stress values are not discussed. Previous work with torsional shear has shown that REDUX 775, a metal adhesive widely used in airframe manufacture and based on polyvinyl formal together with a phenol-formaldehyde resin, shows a breaking strength which varies in mechanism and magnitude with temperature. As shown in fig. 1, highest strengths are obtained at low temperatures where failure is by brittle fracture. Failure is wholly within the adhesive layer and, as would be expected from crack surface-energy considerations, strength does not vary much with temperature.Around room temperature, ductile failure is apparent in the adhesive and the temperature effect becomes more pronounced and typical of a flow process. Failure, although within the adhesive is, at this stage, characterized by clean areas on one of the adherends ; clean as far as can be seen by electron microscopy and ellipsometry. The (strength, temperature) curve was identical for aluminium, stainless steel and titanium alloy substrates, thus supporting the idea that failure is always within the adhesive. The surfaces of all three of metals were prepared by mechanical polishing and were not chemically etched. EXPERIMENTAL AND RESULTS BOND STRENGTH-TEMPERATURE PROFILES ADHESIVE BSL 308 ON TITANIUM ALLOY I M I 205 REDUX BSL 308, Ciba-Geigy Ltd.Duxford, is said to be a modified epoxy adhesive with a blend of resin based on bis-phenol A, together with a hardener, and is supplied as an unsupported film. Test pieces were ground flat and polished to 60 pm and were either used in this condition or were then subjected to one of several chemical etch treatments before bonding. Fig. 2 shows the temperature profiles of this ad- hesive with the various substrate treatments. At temperatures lower than 25"C, different levels of bond strength were obtained with different substrate treatments. Moreover, the (strength, temperature) curve was by no means level at these temper- atures although for some of the surface treatments there is less temperature dependence than for others. The highest strengths are given by treatment (.f), obtained with 2 h oxidation by 0.05 M hydrogen peroxide in 0.4 M sodium hydroxide at room temper- ature and the lowest strengths by treatment (e), aqueous hydrofluoric acid, 4 % w/v.Failure at these lower temperatures was not noticeably brittle in nature and apparently clean areas were seen on the adherend surface. At temperatures above 25°C there is no difference in bond strength between treatments. Neither does the broken surface suggest the ductile type failure expected at higher temperatures ; instead, the appearance of the broken test pieces remained unchanged over the entire temperature range and the adhesive appeared irregularly broken at the interface and across the thickness of the adhesive. However, careful examination of the apparently clean metal by phase contrast interference microscopy revealed the presence of a thin layer of transparent material.The position of the break is shown diagramatically in fig. 3 but the thin layer of adhesive on the substrate is not uniformly thin although, un- fortunately, it was not possible to measure its thickness. However, the ratio of areas of apparently bare metal and dark adhesive does vary. At the lower temperatures this ratio is greater than at higher temperatures when it becomes more nearly unity. BSL 308 is a dark material containing carbon black and aluminium powder, the Iatter being noticeably present in the surface as well as in the bulk of the film before bonding. After breaking the bond, the contact angles shown by water on the40 STRENGTH OF METAL ADHESIVES substrate and the matching adhesive surface were not significantly different.An attempt was also made to measure the refractive index of the thin transparent film mentioned above by flooding with liquids of different refractive indices. From this it I - 5 0 0 50 100 150 temperature/'% FIG. 2.-(Strength, temperature) profile of BSL 308 adhesive on titanium alloy IMI 205. (a) Plain wet polished surface, 0 ; (b) electrolytic oxidation in 0.1 N sulphuric acid, (first-order blue oxide) ; (c) electrolytic oxidation as (b), A (second-order yellow-oxide) ; (d) electrolytic oxidation as (b), 0 (second-order bronze oxide) ; (e) 4 % w/v aqueous hydrofluoric acid, x ; (f) sodium hydroxide+ hydrogen peroxide, + (black oxide region). appeared that the film was not entirely homogeneous as the reflection from some areas disappeared in a liquid of refractive index 1.598 and other areas in a liquid of 1.585.The refractive index of an epoxy resin is about 1.55 increasing with cross-linking and depending on the nature of the curing agent. The film was free from filler particles. I Adherznd FIG. 3.-Position of failure with BSL 308 adhesive on titanium alloy IMI 205. It appears therefore that some component of the adhesive migrates during curing to the interface and failure of the bond is close to, or at, the boundary of this migrated layer, although this should not be taken to imply that a sharp boundary exists.K . W . ALLEN, H . S . ALSALIM A N D W. C. WAKE 41 Material of low molecular weight would be expected to migrate preferentially from a broadly dispersed polymer.In some way the substrate surface is influencing the breaking of the bond at lower temperatures by initiating the failure process in different ways, although at higher temperatures this is not so and failure may be assumed to be wholly initiated by the adhesive. TITANIUM ALLOY IMI 205 This alloy, consisting of 85 % titanium and 15 % m~lybdenum,~ is classified as a p phase alloy and contains at the most about 5 % volume fraction of a phase. Molyb- denum stabilizes the /I phase and the electrolytic oxidation rate of the two phases is identical and hence with electrolytic oxidation there is no change in surface topography. In contrast, both hydrofluoric acid and sodium hydroxide preferentially attack one of the phases.With hydrofluoric acid, oxide formation occurs hydrolytically but with sodium hydroxide, hydrogen peroxide is included to bring about surface oxida- tion and its concentration limits the extent of surface dissolution. Alloy polished by a wet process has a surface consisting of a highly hydrated titanium oxide of unknown composition.6 Metal specimens prepared by polishing unfortunately acquired charge in the electron diffraction apparatus and hence the nature of the surface could not be identified in the present studies. The surface product from electrolytic oxidation is rutile, clearly identified by electron diffraction. Various oxide thicknesses, as judged by colour order, were produced by varying the voltage and the electrolyte. Attack by hydrofluoric acid produces a black oxide layer which was loosely adherent and patchy.When this alloy was dissolved in aqueous hydrofluoric acid, a molybdenum oxide (possibly Mo,O,) was precipitated. The black surface layer is due to molybdenum.’ The use of sodium hydroxide with hydrogen peroxide * produces a rough and porous surface due to preferential attack of one phase. The surface is a mixture of oxides with rutile predominating. Experiments, with a range of hydroxide/peroxide ratios and varying times of treatment, established that the strongest adhesive bonds were obtained with test pieces oxidized for short periods just long enough to form black oxide. If treatment is continued beyond this stage, one phase disappears from the surface leaving entrapped gases which appear as bubbles in the adhesive.The results used in fig. 2 were obtained with conditions to give optimum strength. Stereoscan and light microscopy of the adhesive surface after breaking showed evidence of substrate failure though whether oxide or metal was not ascertained but both a and B phases could be distinguished. ADHESIVE BSL 308 ON TITANIUM ALLOY I M I 318 Fig. 4 shows the temperature profiles of the same adhesive on this (Ti-Al-V) alloy, the results covering the same range as on the previous alloy. They show similar differences with the various surface treatments. The mode of failure, as shown by the appearance of the adhesive, was identical with that observed with IMI 205. TITANIUM ALLOY I M I 3 18 and is an a-P alloy which, in the samples obtained, had an alp ratio of about 70/30.As with IMI 205, the plain polished surface is presumed to be hydrated oxide,6 Electro- lytic treatment produces rutile with topography unchanged. Aqueous hydrofluoric acid attacks the a phase much faster than it attacks the p phase and TiO, (rutile) is formed on both phases but is precipitated as needles on the a phase. This, together This alloy consists of 90 % titanium, 6 % aluminium and 4 % vanadium- 5 0 0 5 0 100 I5 0 temperature/"C FIG. 4.-(Strength, temperature) profile of BSL 308 adhesive on titanium alloy IMI 318. (a) Plain wet polished surface, 0 ; (6) 4 % w/v aqueous hydrofluoric acid, + ; (c) electrolytic oxidation in 0.1 N sulphuric acid; 0, ( d ) sodium hydroxide+hydrogen peroxide, x (black oxide region).I . ' ' .- 5 0 100 150 2 0 0 250 300 - 5 0 0 temperature/"C FIG. 6.-(Strength, temperature) profile of Hidux 1197C adhesive on titanium alloy 318. (a) Plain wet polished, 0 ; (6) 4 % w/v aqueous hydrofluoric acid, x ; (c) as (6) but 10 % wlv nitric acid added, A.FIG. 5.-Titanium alloy IMI 318 after treatment with aqueous hydrofluoric acid. Electron micrograph ; carbon replica, platinum-shaded at 45". To face page 431K . W . ALLEN, H . S . ALSALIM A N D W. C. WAKE 43 with the shadow cast by the protruding, relatively unattacked p phase is the interpret- ation of the electronmicrograph shown as fig. 5. Sodium hydroxide preferentially dissolves material from the phase boundaries and the p phase, leaving a phase relatively untouched. A mixed oxide, containing rutile, is deposited on both phases but the surface is irregular and porous, much more so than with IMI 205.The conditions selected for surface treatment for the bonds included in fig. 4 were those giving optimum joint strength. ADHESIVE HIDUX 1197 c ON TITANIUM ALLOY IMI 318 This adhesive (Ciba-Geigy, Ltd., Duxford) is said to be a modified epoxy phenolic adhesive supplied as film on a glass cloth carrier. It also contains aluminium powder. Fig 6 shows the corresponding (strength, temperature) profiles for plain polished, and after etching with aqueous hydrofluoric acid, with and without the addition of nitric acid. The profiles are different in general form and, as might be expected from an adhesive reinforced by glass cloth, bond failure is apparently at the interface.The alloy appears superficially clean and the high finish of the metal is reflected in the surface appearance of the separated adhesive as if no true wetting had occurred. Electron micrography of a surface replica from broken joints made with polished adherends showed most of the adherend surface to be bare metal and this applied to joints over the whole temperature range at least to 250°C, after which decomposition of the adhesive is apparent. The apparently smooth surface of adhesive and adherend from joints made with etched surfaces were seen to be quite rough on microscopic examination at magnifications above 100, too rough for phase contrast or electron microscopy. A diagramatic view of the surface is shown in fig. 7 . The clear resin, completely free from aluminium has migrated to the interface and is more apparent at low temperatures whilst at higher temperature bare metal, or rather oxide, occupies more area.Where oxide or metal is apparent, it is as rough surface with bright spots of metal showed where protruding j? phase has been broken off. From areas of adhesive opposite to these bare patches, minute amounts of a black powder were recovered by layer I Ad he rend FIG. 7.-Position of failure with Hidux 1197C adhesive on titanium alloy IMI 138. treatment with dimethyl formamide and De-Solv.lo Microscopy showed there to be needle crystals together with small mauve irregular shapes typical of p phase. The needles were identified by X-ray analysis as ivtile. The addition of nitric acid inhibits the dissolution of the alloy by hydrofluoric acid by the formation of oxide and a smoother surface finish is produced with fewer but thicker needles on the a phase but the appearance after breaking the joint is similar.The proportions shown in fig. 7 are arbitrary and the damaged areas of the oxide substrate are, in fact, micro- scopic in area. DISCUSSION there seems to be a different mechanism for initiating failure at higher temperatures than at lower temperatures, though with the adhesives studied As with earlier44 STRENGTH OF METAL ADHESIVES it is not simply brittle fracture of the adhesive at low and ductile failure at higher temperatures. The appearance of the broken adhesive does not differ over the temperature range except insofar as with REDUX BSL308 there are more breaks through the adhesive from one metal interface to the other at lower temperatures, but there is no sign of drawn ductile edges at high temperatures as was reported4 for REDUX 775. Breaking through the adhesive suggests a form of brittle fracture which is prevented by the glass reinforcement of HIDUX 1197 C which throws the process to the interface with the metal, in fact, to the very thin adhesive layer between the glass cloth and the metal.This entails a higher rate of straining of adhesive near the interface than for unsupported adhesive where the shear is distributed across the total thickness, constant rate of loading being assumed. A further differentiation in the properties of the adhesive near to the metal is that caused by migration away from the filler present in both adhesives.Whether the migrated material is intrin- sically different in constitution or molecular weight or not, makes little difference to the argument since freedom from filler will alone suffice to change both molecular and fracture properties. There exists therefore an extreme case of the situation postulated by Good in which variation in these two properties is a function of distance from the interface and leads to failure close to, but not at, the interface. Where fracture occurs more easily through the bulk of the adhesive, there exists the tendency for the temperature profile to level off at lower temperatures (fig. 2 and 4) as was clearly demonstrated earlier (fig. l), but where reinforcement interferes with bulk fracture this feature of the (temperature, strength) profile is absent (fig.6). However, if fracture of the adhesive in one way or another were the sole determinant of strength, surface treatment of the adherend would not influence joint strength and it apparently does so at the lower temperatures. Polishing of the adherend with water-lubricated carborundum has in each series of experiments produced a greater scatter of results than most of the chemical treatments (fig. 2 and 4) and, for the supported adhesive has given the lowest bond strength over most of the temperature range. This may be the result of contamination as well as of a Bikerman weak boundary layer in the form of the hydrated oxide which has not been identified in this work for reasons stated. Electrolytically-produced oxide layers on IMI205 are very coherent rutile layers firmly secured to the substrate.The few data available suggest the thickness corresponding to a second-order yellow colour gives a significantly stronger bond than the thinner layer corresponding to first order blue. The loose molybdenum oxide surface produced by the action of hydrofluoric acid on IMI205 possibly accounts for the fact that this surface treatment gave the lowest results on this alloy, lower than simple wet polishing. On the other alloy, hydrofluoric acid gives a rough surface with the deposition of rutile needles and this gives bond strength in excess of polished metal (fig. 4 and 6). The roughest surfaces were produced by sodium hydroxide and hydrogen peroxide and these gave the strongest joints with REDUX BSL308.A feature of these joints was substrate failure. Previous work on the embrittlement of titanium alloys has always stressed the importance of hydrogen absorption as a cause, and with high concentrations, as might occur in acid etching, hydride formation would be expected. Examination of freshly etched specimens by optical and electron microscopy failed to reveal any hydride phase but there has been no attempt in the present work to investigate the effect of hydrogen. The indications so far in the present work are that a coherent, rough curface of rutile provides the best substrate for adhesives and this is best obtained by sodium hydroxide and hydrogen peroxide. This, coupled with information on the natureK . W. ALLEN, H .S . ALSALIM AND W. C . WAKE 45 of failure in the adherend suggests that at low temperatures there is a mechanical interaction which modifies adhesive behaviour. The authors thank their colleagues F. B. Elliot and M. Phillips for assistance with, and advice on, metallography, phase contrast interference and electron microscopy ; to D. F. Neale of Imperial Metal Industries (Kynoch) Ltd. for help with metallography and phase identification. This work has been made possible by a research agreement with the University by the Ministry of Defence Procurement Executive. Adhesion and Adhesives, N. A. de Bruyne and R. Houwink (Elsevier, Amsterdam, 1951), pp. 92 476. R. T. Humpidge and B. J. Taylor, J. Sci. Instr., 1967,44,457. Torsional Shear Adhesive Test Apparatus (H. W. Wallace and Co. Ltd., Croydon, 1969). H. Foulkes, J. Shields and W. C. Wake, J. Adh., 1970, 2, 254. IMI Titanium 205 (Imperial Metal Industries (Kynoch) Ltd., Birmingham, 1965), p. 3. S. H. Weiman, Corrosion, 1966, 22, 98. D. F. Neal, Imperial Metal Industries (Kynoch) Ltd., Birmingham, private communication. G. Bianchi, F. Mama and S. Trasatti, Proc. 2nd Int. Cong. Metallic Corrosion (National Assoc. Corrosion Eng., Houston, 1963), p. 905. ZMI Titanium 318 (Imperial Metal Industries (Kynoch) Ltd., Birmingham, 1969), p. 3. l o DE SOLV 8090 obtained from Oxley Developments Co Ltd., Ulverston. l 1 R. J. Good, (a) Aspects of Adhesion, 7, ed. D. J. Alner (Univ. of London Press., Ltd.), to be published ; (b) Amer. Chem. SOC. Div. of Organic Coatings and Plastics Chem. (Washington Meeting, 1971), 31, (2), 169.