Return to David Packham's home page

Key words: adhesion theory, anodised alumimium, anodised titanium, FPL etch, mechanical theory, macroroughness, microfibrous surfaces, polyimides, surface roughness, work of adhesion

The phenomenon of adhesion must have been known to our ancestors for millions of years and no doubt excited the curiosity of lively minds. We know that Newton was interested in it, and challenged ‘experimental philosophers’ to find out why things stuck [1]. It is only relatively recently, however, that the phenomenon has been the subject of sustained scientific study.

A key stage in the development of modern adhesion theories is usually considered to be the classic paper [2] 'On adhesives and adhesive action' published by McBain and Hopkins in 1925. This claimed to be the first systematic study of the action of adhesives and popularised the concept of ‘mechanical adhesion’. It is appropriate seventy years later to mark this publication by a review of the mechanical theory of adhesion and of its current status. This present paper sets out to do this. The intention is not to present a comprehensive review of 70 years of work. Such a review would be of inordinate length and of doubtful value as there have been many reviews over the years, some of mechanical adhesion in general, others of particular aspects of this field. Papers of Rinke[3] in 1945, of Reinhart[4] in 1954 and of Wake[5] in 1976, together with the references they cite, give good surveys of the situation at the times they were written. In the 1980s important reviews, mostly directed at polymer-metal adhesion, include those published by Venables[6], by Arslanov and Ogarev [7] and by Brockmann, Hennemann, Kollek and Matz[8]. More recent work by Davis[9] and by Critchlow and Brewis[10] should also be noted.

What this present paper does aims to do is to give (in section 2) an historical overview of the main stages of development of the mechanical theory up to the mid-1980s and then (section 3) to survey in more detail work of the last ten years in order to assess the present status of the theory (section 4). These last two sections can be seen as an updating of my own 1983 review, 'The adhesion of polymers to metals: the rôle of surface topography' in the important collection edited by Kash Mittal [11]. In section 2, I have also sought to minimise the overlap with my 1992 paper[12], written to mark the contribution of John Venables to the development of adhesion theory. A fuller assessment of his work on adhesion to porous surfaces on aluminium and titanium can be found there.

McBain and Hopkins considered that there were two kinds of adhesion, mechanical and specific. Specific adhesion involved ‘some species of interaction’ between the smooth non-porous surface and the adhesive. This interaction might be ‘chemical or adsorption or mere wetting’. Specific adhesion has developed into the model we today describe in terms of the adsorption theory.

In contrast, mechanical adhesion was only considered possible with porous materials. It occurred ‘whenever any liquid material solidifies in situ to form a solid film in the pores’. They cite as examples adhesion to wood, unglazed porcelain, pumice and charcoal. For McBain and Hopkins mechanical adhesion was very much a common sense concept, ‘It is obvious that a good joint must result when a strong continuous film of partially embedded adhesive is formed in situ’.

When they consider whether mechanical adhesion can act as a mechanism of adhesion on its own, in the absence of adsorption effects, their position was somewhat ambiguous. On one hand they wrote that ‘A liquid which wets a surface and is then solidified possibly always makes a joint’. Neglecting the equivocality of the ‘possibly always’, this seems to mean that ‘whenever wetting occurs, some form of adhesion must result’. This would be accepted today. By contrast, elsewhere in the paper McBain and Hopkins talk of mechanical adhesion as being ‘independent of adhesive action’. They found no evidence of adsorption of adhesive from solution onto filter paper, and concluded that the adhesive adsorption did not occur during the bonding of wood.

McBain and Hopkins distinguish adhesion to porous surfaces from adhesion to surfaces which are merely rough. Roughness, they report, has little influence on adhesion, ‘provided that the roughening has been carried out without pitting the surfaces appreciably’ If metal surfaces are ‘rather deeply furrowed’ the joints are decidedly weaker than those obtained with smooth surfaces. This reduction may be a result of either an increase in average adhesive film thickness, or there being insufficient adhesive to bridge the gap between the surfaces.

It is interesting to note how many of the themes of contemporary discussions of adhesion mechanisms (section 3) were broached in this classic paper in 1925.

Despite its intuitive appeal, the mechanical theory of adhesion was subjected to increasing hostile criticism. Browne and Truax [13] published a paper on wood adhesion soon after McBain and Hopkins. While stating that they had ‘no serious disagreement with McBain and Hopkins’ point of view’, their main emphasis was that specific adhesion was more important than mechanical. For joints with wood, they obtained high adhesion to smooth surfaces, and were unable to observe a correlation between adhesion and penetration of the glue into the wood. McBain and Lee replied by arguing that the penetration was on an ultra-microscopic scale and so could not be observed [14].

A survey of adhesion by Rinker and Kline[15] in 1945 advised that ‘in order to obtain a strong joint, a smooth surface is more desirable than a roughened one’, and glossed Browne and Truax's work as 'this mechanical interpenetration cannot account for more than a small fraction of the joint strength' [13].

In an influential review published in 1954 Reinhart considered the question of ‘specific versus mechanical adhesion’ [4]. On the basis of high adhesion to smooth surfaces and lack of microscopic evidence supporting the mechanical theory, he concluded that ‘mechanical adhesion occurs seldom, if at all’. The poorer adhesion often observed with rough surfaces was explained in terms of poor wetting leading to void formation, and the development of stress concentrations around projections on the rough surface. This conclusion was endorsed by Wake writing in 1965, ‘theories that mechanical interlocking .... added to the strength of a joint have been largely discredited’ [16].

Thus by the 1950’s the mechanical theory was all but abandoned, yet in the 1970’s it was again being taken seriously. The extent of the change can be judged again by quoting Wake, writing this time in 1976 in his book 'Adhesion and the Formulation of Adhesives'. Wake concluded [5] that 'adhesive joints frequently possess an important mechanical component essential to the performance of the joint but this type of component cannot suffice as the sole mechanism whereby surfaces are joined. It must be enhanced by, just as it enhances, specific adhesion'.

Most of the new work from the 1960's cited by Wake falls into one of two categories. The first is associated with the electroless deposition of metals onto plastics such as ABS and polypropylene. In the process the plastics must be etched in a way which produces pits on a micrometre scale. Such a topography had been shown to be a necessary, but not sufficient condition for adequate adhesion. The second category was concerned with adhesion to porous or microfibrous surfaces on metals. In 1969 Packham[17] had demonstrated the importance of pore structure in the adhesion of polyethylene to anodised aluminium, showing that the polymer entered the pores on the anodised surface. Arrowsmith [18] worked with electroformed copper and nickel, and argued that mechanical adhesion was the main mechanism of adhesion of these surfaces to epoxide laminates. There were further reports of the importance of topography in adhesion, for example, to copper[19] and titanium[20] with needle-like oxides.

It was developments at this time in electron microscopy (scanning electron microscopy and scanning transmission electron microscopy) and in electron spectroscopy (Auger and X-ray photoelectron spectroscopies) that enabled the physical structure and chemical composition of surface layers to be established in detail previously impossible. This, together with the increasing need of the aerospace industry for strong, consistent and durable adhesive bonds was among the factors that stimulated considerable work on pore-forming surface treatments for aluminium and titanium published during the late seventies and into the early years of the eighties.

It is not within the scope of this paper to present a comprehensive critique of the work of this period. Some such reviews have already been mentioned [7,11], others, written from a European industrial perspective, can be found in references 8, 21. Significant research was carried out in the United States, for example, by Boeing[22, 23, 24, 25] which led that company to adopt a standard phosphoric acid anodising pretreatment for structural bonding of aluminium[26]. Another American example is provided by Venables and his colleagues at the Martin Marietta Laboratories who, from the mid-1970's, undertook a study of the influence of pretreatment variables on the structure and adhesion properties of the films formed [12, 27, 28, 29, 30, 31].

The broad consensus that comes from most of this work is that strong bonds, and more particularly bonds of high durability, tend to be associated with a highly porous surface oxide, providing, of course, that the values of viscosity and surface tension of the adhesive are such as to allow it to penetrate the pores[11]. Thus for aluminium, a sulphuric acid/ dichromate etch, known as the FPL etch, produces a surface of low porosity and bond durability, compared to phosphoric acid anodisation[6, 17, 22, 24]. Figure 1(a) and (b) [33] gives very much simplified schematic representations which emphasises the contract between the surfaces mentioned. It must be emphasised that the actual structures depend upon the various pretreatment parameters and usually show features omitted here[6, 8, 30, 32,33].

The importance of porosity was brought out strongly in a 1984 review by Venables[6]. He concluded that for aluminium and titanium 'certain etching or anodization pretreatment processes produce oxide films on the metal surfaces, which because of their porosity and microscopic roughness, mechanically interlock with the polymer forming much stronger bonds than if the surface were smooth' This is as unequivocal a statement of the mechanical theory of adhesion as can be found in the original work of McBain and Hopkins.

Since this time, the acceptance of a 'mechanical theory' has not been seriously challenged, and it now has a generally accepted place within the canon of adhesion theories[9,10, 34, 35, 36]. It is, nevertheless, of value to examine work of the past decade or so to see what advances have been made. What kind of adhesion phenomena are now being rationalised in terms of the theory? Has the theory itself undergone significant development? If so, has the tendency been to reinforce or to undermine the orthodoxy of the eighties?

McBain and Hopkins considered that wood, unglazed porcelain and charcoal were substrates to which mechanical bonding occurred. What types of system are regarded today as ones where mechanical factors play a significant rôle in adhesion?

References to wood and ceramics still appear in the literature. For example, Bogner has reported correlations between bond strength, and increased penetration of the adhesive into the wood [37]. Correlations between ceramic porosity and adhesion have been published by Packham and Johnston [38] for unglazed porcelain and by Kanai [39] for alumina prepared by a sol-gel process.

The scope of the mechanical theory has moved well beyond these materials. The main features of recent work may conveniently be surveyed under three headings dealing, in turn, with metal, polymer and fibre substrates.

Discoveries in this area were principally responsible for the re-establishment of the mechanical theory in the 1970’s [5]. Microporous and microfibrous surface topographies can be produced to form a pre-treatment on many metals including magnesium, copper, steel, titanium and aluminium.

As was emphasised in the previous section, many of these pretreatments are important industrially, and give increased durability to the joint in humid conditions well as giving adequate initial adhesion. Treatments for titanium and aluminium have been reviewed within the past decade, for example by Davis [9] and by Critchlow and Brewis[10]. Bishopp et al.[33] have specifically discussed the effect of pretreatment of aluminium for bond durability. Broadly speaking, their results confirm the consensus discussed in the previous section. For unprimed surfaces they report considerably better durability for phosphoric acid anodising than for FPL etching. They found poorer durability for a chromic acid anodised surface which they attribute to inability of the adhesive to penetrate the narrow entry to the pores, see figure 1(c). It should be noted that they were using a toughened epoxy film adhesive with a 120 °C cure temperature which might have had a relatively high viscosity[11]. As their electron microscopy showed little damage to adhesive or substrate even after a fall in adhesion, they suggested subtle undermining of alumina film or disruption of physico-chemical bonds across interface as possible mechanisms of degradation. This is in contrast to the view of Venables[6] who argues that conversion of the aluminium oxide to hydroxide is a significant factor in bond degradation in a humid environment.

This is still an active research area where the concepts of the mechanical theory of adhesion are commonly invoked in order to rationalise the results obtained. Three recent papers of Wightman’s on bonding, respectively, to aluminium, steel and titanium illustrate current work.

Ko and Wightman [40] demonstrated the effectiveness of anodising aluminium-lithium alloys in either sulphuric acid or phosphoric acid. A porous surface oxide was produced into which a polysulphone adhesive penetrated giving a 'mechanical means of adhesion'. Bonds were obtained which were much more durable in high humidity than comparable ones made to surfaces without this high porosity.

Microfibrous surfaces on steel produced by a hydrothermal oxidation (cf. Figure 2 [41]) were used by Hollenhead and Wightman in a study of bonding with a thermoplastic polysulphone [42]. Lap shear strengths for the grit blasted steel, made for comparison, were 40% lower than those for microfibrous surfaces pretreatments: by increasing the strength of the interfacial region, failure is directed into the polymer where dissipation of fracture energy leads to a higher fracture stress. Although bonds with both pretreatments deteriorated with exposure to humid environments, those with the microfibrous surfaces were the more durable.

Filbey and Wightman[43] have clearly demonstrated the applicability of analogous principles in the bonding of epoxies to titanium alloys. They found that anodising either in chromic acid, based on a method outlined in Boeing patents[44, 45], or in sodium hydroxide was capable of producing a surface with pores of 40 to 50 nm diameter. Surface analysis showed the penetration of resin into these pores and wedge test results demonstrated the superior durability of these anodic pretreatments, compared with others which did not give a microporous surface. These results both confirmed and extended findings of Boeing[44, 45] and of Martin Marietta Laboratories[6]. In their recent review of 35 types of surface treatment for titanium, Critchlow and Brewis [10] conclude that microporous surfaces are ‘widely recognised ..... to produce the best durability results’, an effect ascribed by many authors to an increased likelihood of mechanical keying.

The generally good adhesion with a range of adhesives, usually combined with good bond durability in a humid environment [46], is generally associated with entry of the adhesive into the pores on the substrate surface.

The pores on anodised aluminium may be as small as tens of nanometres in diameter, and at one time scepticism was expressed as to whether any polymeric adhesive could enter them. There is now a wealth of evidence from electron microscopy and surface analysis [6, 33, 43, 47,48] backed up by theoretical calculations on idealised models [7, 11, 43, 49] which shows that polymeric adhesives and primers will often penetrate the microporous surface layers. The ease of ingress depends strongly on the polymer viscosity and time for which it is in a fluid state [11, 21, 49, 50], so a change of adhesive or of bonding conditions may significantly alter the extent of penetration.

Recent work has further confirmed the importance of adhesive penetration for good bond performance. Davies and Ritchie[51] in their work on bonding metal matrix composites to aluminium emphasise the importance of the ‘micro-composite’ interfacial region formed by penetration of the adhesive into the porous oxide on chromic or phosphoric acid anodised aluminium for the durability of the joints under stress and in humid environments. The 'micro-composite' region decreases the sharpness of the discontinuity in modulus between adhesive and substrate.

Sargent has recently published an investigation of bonding chromic acid anodised aluminium with epoxy and phenolic adhesives [48]. By variation of anodising conditions [in a way not revealed] a range of peel strengths were obtained. As the fine structure of the surface oxide increased, so did the peel strength [figure 3]. Depth profiling using argon ion bombardment and XPS showed evidence of a greater proportion of primer within the oxide layer, to a depth of at least 40 nm, for high peel strength specimens than for low peel strength ones. This result should be contrasted with that of Bishopp et al.[33] discussed above.

With increasing realisation of the environmental damage caused by many synthetic chemicals, there have been moves to develop pretreatments for aluminium which avoid the use of chromates. The widely used phosphoric acid anodisation is commonly preceeded by a chromate-containing FPL etch. As long ago as 1977 Russell [52] published results which suggested that an etch designated 'P' etch might be able to replace the FPL process, both as a treatment in its own right and as a preliminary to phosphoric acid anodising. The 'P' etch consisted of sodium sulphate and ferric sulphate dissolved in a sulphuric acid - nitric acid mixture.

A furtherdevelopment of this treatment is known as the 'P2' etch. It avoids using nitric acid and consists of ferric sulphate and sulphuric acid alone. The P2 etch produces an oxide topography comprable to that of FPL, and has shown some promise as a substitute for the FPL treatment[27, 53, 54]. It was even suggested as a replacement for the phosphoric acid anodising process itself[27]. More recent findings, such as the wedge test results shown in Figure 4, confirm its effectiveness when followed by the anodising, but do not support its use as a substitute[54].

Although some polymers are able to enter the pores produced on aluminium by sulphuric acid anodising [41], they are much finer than those from anodising in phosphoric acid [32, 55, 56]. There has been an interest in opening up the structure of the sulphuric acid-formed films by a post anodising dip. Arrowsmith and Clifford [57] have shown that a controlled etch in phosphoric acid after anodising in either sulphuric acid or chromic acid results in improved bond strength and durability of adhesively bonded aluminium compared with standard etching and anodising treatments. Digby and Packham [58] found that the durability of bonds to phosphoric acid anodised aluminium could also be improved by controlled exposure to phosphoric acid after anodising (Figure 5). Figure 6 shows the more open structure of the surface after such a treatment.

Plasma spraying [59] provides a method of producing a porous surface layer without the use of chemical solutions. The porosity is much coarser than that associated with anodising, being of the order of micrometres rather than tens of nanometres. Clearfield et al. [60] at Martin Marietta Laboratories compared anodising treatments for titanium alloy [Ti-6Al-4V] with plasma sprayed coating of the alloy. Durability results showed comprable performance. Alumina powder has been used on aluminium, titanium and steel[61]. More recent work from Martin Marietta by Davis et al. has reported on the potential of plasma sprayed coating as pretreatments for aluminium and titanium [62, 63]. The best performance on aluminium which they report is for a 40 µm thick coating of aluminium-silicon alloy mixed with polyester : crack growth in a wedge test was comparable to that for P2 and FPL

There has been a renewed interest in this coarser, micron sized, porosity in recent years. Indeed Sargent [48] reported influences of micron scale roughness as well as those discussed above which were associated with a far finer scale of roughness. Another example is the work of Morris and Shanahan on sintered steel substrates bonded with a polyurethane adhesive [64] They ascribed the much higher fracture energy for joints with sintered steel compared with those with fully dense steel to the mechanical interlocking of polymer within the pores. Extra energy was required to extend and break these polymer fibrils.

A picturesque example of providing a mechanical ‘key’ has recently been described by van der Putten who was concerned to bond copper directly to silicon in the context of integrated circuit technology [65].

Copper sticks poorly to silica but titanium tungstide sticks well. Using conventional lithographic techniques islands of TiW 0.1 µm thick were sputtered onto the silica and the photoresist was removed [figure 7a].

Palladium acts as a nucleating agent for the electroless deposition of copper. By treating the surface with palladium [II] chloride in hydrochloric acid a monolayer or so of palladium is deposited on the TiW surface. The palladium chloride solution also contains 1% of hydrofluoric acid which attacks the silica, undercutting the TiW islands [figure 7b]. Copper is now deposited, nucleating on the palladium covered TiW and growing from it. It is thus mechanically anchored to the silicon surface [figure 7c and d].

It is important to point out that this work on macroroughness qualifies, rather than supersedes, the earlier work[3, 4, 16] discussed above, where such roughness was generally reported to have an adverse effect on adhesion. Critchlow and Brewis [66] have recently reported that use of different grit sizes in the grit blasting of aluminium has only a small (albeit significant) influence on the durability of epoxy bonds.

The effect of roughness on this scale is a subtle one, depending very much on the detailed characteristics of the surface[11]. This point has recently been eloborated by Keisler and Lataillade[67] for steel joints made with epoxy resin and tested at high strain rates. Differences in the surface topography affected the deformation mechanism of the joint (cf. [68, 69]). Excessive roughness combined with sharp asperities and narrowly spaced valleys gave rise to poor wettability and to the creation of microcracks near the interface. Stronger joints, especially at very high test rates, were observed for some ‘laser roughened’ steel where wide valleys and asperities with a high radius of curvature were produced.

It is clear, then, that porous surfaces and the concepts of the mechanical theory of adhesion play a large part in the surface treatment of metals for structural adhesive bonding. Davis [9]has recently summarised the situation with respect to aluminium and titanium 'all high performance adhesive bonds require clean, microscopically rough substrate surfaces allowing wetting of polymer and ‘complete filling of the pores of nooks and crannies ....... the resulting mechanical bonds supplement or supplant the chemical bonds present'

In contrast with metals, polymers have low surface energies and are consequently wetted less readily by adhesives. Both classes of material may have cohesively weak surface layers, with polymers they result from migration of additives or of low molecular weight polymer fractions to the surface[70]. A wide range of treatments have been described which enhance adhesion to polymers. Some involve aggressive chemical treatment. Strong oxidising solutions, such as chromic acid or permanganate have been employed with hydrocarbon polymers, and highly reactive combinations such as sodium in liquid ammonia have been used for fluorocarbons. Alternative pretreatments to ‘wet’ chemical treatment include surface flaming, the action of gas plasma or reactive ion etching[71, 72, 73]. (It is important to appreciate that the plasma treatment referred to here and described in references 71, 72, 73 is very different from the plasma spraying of a porous layer referred to above, and described in reference 59.)

Surface analysis has established in some detail the various functional groups that different treatments introduce into the surfaces of different polymers. These are often polar oxygen-containing groups. It is commonly reported that surface roughening accompanies these chemical changes, and a ‘mechanical’ effect is sometimes claimed or mentioned as a possible contribution to the adhesion. The importance of the roughening of rubber reinforced plastics, such as high impact polystyrene and ABS is well established (see [5, 11]). Oxidising agents etch away the spherical rubber-rich phase where it comes close to the surface[74]. It may be easy to speculate about mechanical effects in other cases, but less easy to prove that they apply.

O’Kell et al. have recently reported the effects of treating polyethylene in a low energy (<1W) oxygen or nitrogen plasma[75, 76]. The changes induced were particularly well-characterised as the plasma cell was attached directly to the preparation chamber of an X-ray photoelectron spectrometer, so that spectra could be obtained immediately after treatment, as well as after exposure to air. Topographical changes were observed using atomic force microscopy. The plasmas introduced both chemical and topographical changes into the surfaces within the first few seconds of treatment. The authors consider the amount of oxygen functionality introduced to be the dominant parameter governing the level of adhesion of the polymer to aluminium. The importance of this conclusion would seem to be a little reduced, as bonds were made by melting the polyethylene against the aluminium. This might alter the surface chemistry and would almost certainly alter the topography.

Similar work treating polypropylene in oxygen and nitrogen plasmas has been published by Harth and Hisbt [77]. Again rapid formation of functional groups was observed, but roughening was slower. Adhesion specimens were made by bonding the treated polymer to aluminium using a polychloroprene adhesive. As the peel strength increased before the roughening became significant, the authors again considered that the introduction of chemical groups was the main cause of enhanced adhesion. They speculated that the 'filigrees' of polymer formed after longer treatment times were weak, but unfortunately gave no information about the failure modes which might have enabled a less ambiguous conclusion to be reached.

Recent publications on treatment of styrene-butadiene rubber (SBR) poly-p-phenylene-vinylene and PTFE-polyperfluoroalkylvinylether (PFA) copolymer report evidence for mechanical interlocking which together with chemical interaction contributes to the adhesion. These are now discussed.

Pastor, Sánchez and Martín chlorinated vulcanised S.B.R. with trichloroisocyanuric acid solution[78]. The treatment produced microcracks in rubber surface which, along with the formation of carbonyl, hydroxyl and ether functional groups, were considered to aid adhesion with a polyurethane adhesive .

Nguyen et al. studied the vapour deposition of aluminium on thin poly-p-phenylene-vinylene films [79]. Satisfactory adhesion to the aluminium was obtained when the polymer films were treated with oxygen by radio frequency sputtering. The roughness of the surface produced was considered important as it allowed deeper penetration of the deposited aluminium which subsequently formed an ‘Al-O-C complex’.

Surface roughening was found by Marchesi, Keith and Garton [80] to be a necessary stage in obtaining adequate adhesion of PFA to copper. The roughening was followed by chemical modification of the surface with sodium naphthalenide.

Bonding to polyimides has received attention in recent years because of their applications in the microelectronics industry and in high performance aerospace composites.

Buchwalter and Saraf[81] have treated polyimide surfaces prior to deposition of tantalum or chromium by argon sputtering, both on its own and in combination with reactive ion etching with carbon tetrafluoride. They have systematically characterised the chemistry and topography of the treated surfaces. The peel strength measured on peeling the metals from the optimally treated surfaces was much higher than that of the virgin polyimide (ca. 1000J/m² cf. 100J/ m²). Buchwalter carefully examined the failure modes and concluded that the increased adhesion was not a consequence of the increased surface roughness per se, nor of chemical change but of weak boundary layer removal with the increased energy dissipation perhaps being associated with cracking of underlying polyimide, toughened by the pretreatment [82].

Inagaki et al. treated polyimide films in a range of gas plasmas, deposited a copper layer and measured the adhesion[83]. They obtained infra-red spectroscopic evidence for co-ordinate bond formation between copper and the treated polymer surface. The argon plasma, which gave the highest peel strength, produced a rough surface with fine, high protuberances. The authors are undecided as to whether to ascribe the enhanced adhesion to mechanical keying, to weak boundary layer removal or to the co-ordinate bond formation. Unfortunately, they did not publish micrographs of the failure surfaces which could have been compared with Buchwalter’s to judge whether a similar explanation to hers applied here.

A particularly elegant method of producing polyimide with a rough surface and enhanced adhesion has been described by Saraf, Roldan and Derderian [84]. It is achieved by controlling the initial stages of formation of the polyimide from its dianhydride and diamine precursors, so no post-polymerisation treatment is required for adhesion. An example of the nanoscale roughness that can be produced is shown in Figure 8.

In order to understand how the roughness is achieved, it is necessary to remember the basic chemistry of formation of condensation polyimides[85]. An appropriate acid derivative, such as a dianhydride or diacyl chloride, and a diamine are dissolved in a solvent and a ‘soft bake’, typically at 70°C, is employed to remove solvent. Condensation to form a polyamic acid [PAA] occurs, Figure 9. A subsequent ‘hard bake’ at around 400°C eliminates water, and leads to the formation of the polyimide itself.

The method of Saraf et al.[84, 86, 87] relies on the fact that exactly the same polyimide would be obtained on hard baking of a polyamic ester, rather than polyamic acid: i.e. the carboxyl groups in the acid shown in figure 9 would be esterified and the corresponding alcohol, rather than water, would be eliminated during the hard bake. The polyamic ester can be obtained by simple modification of the original anhydride or acyl chloride.

Saraf et al.[84] produced poly (pyromellitic dianhydride-oxydianiline) [PMDA-ODA] from both the polyamic acid [PAA] and the ethyl ester PAETE. Precursors of both acid and ester were dissolved in solvent [N-methyl pyrrolidone, NMP], the solution spun on a suitable substrate and ‘soft baked’ at 70°C to remove the solvent. Depending on the proportions of the two precursors, polyimide films with different roughnesses on the scale of nanometres could be produced.

The rough surface is the result of two phenomena. As the NMP solvent evaporates during the soft bake phase separation by spinodal decomposition occurs. The PAA-rich phase swells more than the PAETE-rich phase: this produces surface roughness on a fine scale.

Imidisation occurs during the hard bake, but this is some fifteen times faster for PAA than for PAETE. Thus at an intermediate stage the former PAA phase has formed rigid polyimide while the PAETE is still essentially the softer precursor. As the PAETE is converted to polyimide further modification of the topography is spontaneously achieved.

Saraf et al.[84] ascribe the poor adhesion of conventional, untreated polyimide films to an ordered surface layer and the alignment of the polymer chains parallel to the film surface, as well as to the absence of nanoscale roughness. They claim that the technique described overcomes all these problems. They quote peel strengths for vapour deposited copper which are nine times greater than those to untreated conventionally formed polyimide.

A third area in which mechanical effects in adhesion are often reported deals with fibre-matrix adhesion in composites. Whereas adhesion to glass fibres is satisfactorily effected by use of silane coupling agents, carbon, Kevlar and high strength polyethylene fibres usually require chemical etching or gas plasma treatment. These modify the surfaces chemically, but many authors have reported a roughening which they see as contributing to the enhanced adhesion.

Treatments of carbon fibre in various plasmas have been reported to improve adhesion. For example, the formation of rough fibre surfaces and consequent improvement in adhesion to the matrix have been mentioned by Moyer and Wightman [88] for an oxygen plasma and by Zang[89] for oxygen and argon plasmas. In contrast, Yuan et al. [90] ascribed improved interfacial shear strengths in oxygen and argon plasma-treated fibres to the formation of hydroxyl, ether or aromatic groups, not to surface roughening.

Anodising has been shown by King et al. to produce surface roughness on pitch based carbon fibre[91, 92]. They used XPS and chemical probes in order better to understand the reasons for the improved properties of an epoxy composite made from the treated fibres. They concluded that the increased interfacial shear strength resulted primarily from improvements in mechanical interlocking rather than chemical interactions between fibre and matrix.

An interesting example of mechanical interlocking was reported by Marshall and Price in composites made from treated carbon fibre[93]. They talk of graphite sheet edges which 'increase the number of potential bonding sites', and also present a 'jagged surface profile [which may] increase mechanical keying'. It is curious in a paper in a primary research journal that they give no detail whatever of the surface treatment they used!

Ultra-high molecular weight polyethylene is now of interest as fibre reinforcement in composites. Bonding problems apply, similar to those addressed in their work on polyethylene sheet by O’Kell et al.[75, 76 ] (v. supra). Similar treatment methods have been applied.

Ward and Ladizeski [94] have used chromic acid treatment and plasma etching to achieve a range of values of both monofilament pull-out adhesion and inter-laminar shear strength, Table 1. This work provides an example of an adverse effect of high fibre-matrix adhesion. Enhanced adhesion resulting from oxygen plasma treatment prevented delamination during impact, and so reduced the impact strength [95].

Ward et al.[96] have also studied the effect of the oxygen plasma treatment time. At short treatment times, general oxidation of the polyethylene fibre surface occurred. With intermediate times, surface cross-linking predominated. Long treatments produced a rough surface 'which could give rise to a mechanical keying effect' .

Silverstein et al. have used chromic acid, potassium permanganate and hydrogen peroxide to treat polyethylene fibres[97, 98]. Of these three, they only found chromic acid effective in increasing the adhesion. They suggest the removal of weak boundary layers and surface roughening as well as surface oxididation as explanations for the better adhesion.

It can be seen then that the last decade has been one of steady progress and consolidation, during which the main features of the mechanical theory as it was re-established in the 1970's and 80's have been confirmed in a wide range of experimental situations. The theory has proved a 'useful' one in the sense that it has stimulated the development of new surface treatments for metals, polymers and fibres and has assisted in giving an understanding of their efficacy. There has perhaps been a tendency, now that the theory is again 'respectable', to invoke 'mechanical effects' somewhat uncritically wherever an increase in surface roughness has been observed.

It is now seventy years since McBain and Hopkins' classic paper proposing the mechanical theory of adhesion was published. In the intervening years, it has suffered from the vicissitudes of scientific orthodoxy[12], but is now well regarded and is often invoked to rationalise results of adhesion experiments. In the light of contemporary models, how might we regard mechanical theory of adhesion and its relationship to the adsorption theory?

Before addressing this question directly, let us consider the relationship between the forces acting at the interface between the bonded phases and, and the ‘adhesion’ measured by some destructive test performed on the adhesive joint. This practical adhesion, as Mittal[99] points out, reflects the properties of the interfacial region and, indeed, of the joint as a whole. The adhesive joint is usefully regarded as a ‘composite’. The interfacial forces are, of course, important, but the performance of the joint is a consequence of the interface, the bulk phases and the interactions between interface and bulk phases joined. The manner in which the components interact will depend upon the geometry of the joint, its physical form, and upon the way in which it is loaded.

This interdependent relationship can be expressed in simple mathematical form. Good did this when he applied the Griffith-Irwin theory of fracture to a joint comprising a bond between two phases[100]. The fracture stress, sf, is given by

sf = k(EG/l)½ (1)

where k is a constant, l is the length of the critical crack which leads to fracture and E is the modulus and G the fracture energy.

The modulus arises in the Griffith theory as its value determines the amount of elastically stored energy at a given strain. It is this stored energy which is released, providing the fracture energy, G when fracture occurs. Within the adhesive joint, E and G are semi-local properties. Fracture will occur where the term EG/l is lowest, whether at or near the interface or within one of the bulk phases. Factors which alter E or G or l locally within the joint may alter its strength and other properties.

The fracture energy G will involve a term associated with the energy required for bond breaking at the interface when the joint fails. These may be primary or secondary bonds. This surface energy term will be the thermodynamic work of adhesion Wa or of cohesion Wc , depending on whether fracture occurs at the interface or cohesively within one phase[101]. To this surface energy term must be added a term y representing other energy absorbing processes - for example plastic deformation - which occur during fracture:

G = Wa (or Wc) + y (2)

Usually y is very much larger than Wa (or Wc). This is why practical fracture energies for adhesive joints are almost always orders of magnitude greater than works of adhesion or cohesion.

A similarly approach may be adopted towards the peel test, where the ‘peel strength’ is essentially a peel energy, the fracture energy per unit area. The peel energy, P, then comprises a surface work term and (according to circumstances) terms for plastic, viscoelastic and other losses which occur during fracture[102]:

P = Wa (or Wc) + yplast. + yv/e + ...... (3)

Factors which enhance either W or y will increase the fracture energy - the adhesion - of the of the bond. Although again it is usual that

y >> Wa (or Wc)

a modest increase in W may result in a large increase in adhesion as y and W may be coupled. For some mechanically simple systems where y is largely associated with viscoelastic loss, a multiplicative relation has been found:

P = W × f(c,T) (4)

where f(c,T) is a temperature and rate dependent viscoelastic term[103, 104]. In simple terms, stronger bonds (increased W) may lead to much larger increases in fracture energy because they allow much more bulk energy dissipation (increased y ) during fracture.

Theories of adhesion can be understood in terms of the effects they predict for local modulus (E), surface energy (W), energy dissipating capacity (y) or defect size l. Let us use these ideas first to consider the mechanical theory, and then to compare the mechanical with the adsorption theory.

Adhesion to a rough or porous surface may be enhanced. How might this work?

An obvious way in which a rougher surface might lead to higher adhesion than a smooth one results from the increased area of contact. A surface with a higher 'roughness factor', defined as the ratio of the true area to the nominal area, would be expected to give a higher effective work of adhesion. This point has often been made, and in careful experiments with rubber adhesion Gent et al. have convincingly demonstrated the effect [105, 106]. With grit blasted steel they observed increases in adhesion by factors of 2 to 3 times which they ascribed to an increase in surface area.

Such a traditional roughness factor approach overcomes the complication of surface roughness, by treating the surfaces as if they were intrinsically the same as smooth ones, and applying a correction factor for the roughness.

But what if the surface is very much rougher? What about the porous surface of anodised aluminium shown in Figure 3 or a microfibrous oxide on steel, Figure 2? As the scale of roughness becomes finer, use of a simple scaling factor becomes increasingly unrealistic and unconvincing. It becomes unconvincing not just because of increasing practical difficulty in measuring the ‘true’ area of such surfaces, it is unconvincing because the roughness itself is an essential characteristic of the surfaces.

These surfaces are not just the same as smooth surfaces that happen to be rough. As we approach molecular scale roughness, indeed long before we get there, the molecular environment of the surface molecules, and therefore their surface energy, is quite different from that of similar molecules on a nominally plane surface. Their energy is a consequence of the topological configurations they take up. Moreover, it has been argued that many engineering surfaces are fractal, possessing a similar structure over a range of scales[107]. The area ascribed to such surfaces is arbitary, depending entirely on the scale chosen.

Further, roughness may develop as a result of the reorganisation of two phases once brought into contact. It would then be an intrinsic property of the interface concerned. Using the surface forces apparatus[101], Israelachvili and his colleagues have demonstrated the rearrangement of a solid-solid interface after formation by reptation, reorientation and exchange of molecular species[108, 109]. Roughness is produced where it was not previously present.

Even if it were practicable, it would make no scientific sense to apply some sort of ‘roughness factor’ to allow for the roughness of such surfaces. The concept of work of adhesion is applied in a macroscopic context: the appropriate area to be used in calculating it in these cases should therefore be the macroscopic area (i.e. ideal, geometric area) of the interface.

We see then that the effect of producing many of the more intricate forms of surface topography, described in this paper, is to enhance the surface energy and work of adhesion. The result is not fundamentally different from that obtained by increasing the surface energy by chemical modification. Both have the potential to enhance the measured adhesion: see equations 2, 3 and 4.

Modification of the topography of a surface may alter the way that the stress is distributed when the joint is loaded. This can increase the energy dissipation, y, which occurs during fracture, and provide an important increase in adhesion.

A conceptually simple example of this was the adhesion of silica to copper discussed above[65]. Here the stress is directed away from the low Wa interface (silica/copper) towards the stronger silica / palladium interface by the topography produced, figure 7. This is in many ways a simply 'key' of the sort that McBain and Hopkins envisaged. The surface topography protects weak regions from a high stress field.

The suggestion has been made that the occlusion of cohesively weak material in the pores or fibres of microfibrous surfaces might contribute to the good adhesion often associated with such surfaces[110, 111]. This can be understood as another example of the direction of the applied stress away from weak material towards the stronger bulk polymer. Arslanov and Ogarev[7] argue that the diffusion coefficient is too low for this to occur with polyethylene on anodised aluminium. Even if this be true in the particular example, there would seem to be no reason for dismissing the mechanism altogether.

Another way in which microfibrous and microporous surfaces enhance adhesion is by causing stress discontinuities at the interface. Fibre tips, for example, act as stress concentrators causing plastic yielding, at first locally, and then further into the bulk of the polymer. Evidence for this effect has been found, not only with high ductility polymers, such as low density polyethylene and ethylene vinyl acetate copolymers, but also with epoxy resins[69, 112, 113]. By altering the interfacial topography, the energy absorbing capabilities of the bulk materials are exploited to a greater extent. In terms of equation 3, the energy losses, yplast. , are increased.

Gent and Lin have shown that large amounts of energy can also be involved in peeling an elastic material from a rough surface[105]. The energy is essentially used for the elastic deformation of embedded filaments: this energy is lost because when the filaments become free, they immediately relax.

Gent and Lin experimented with rubber bonded to aluminium plates with regular arrays of cylindrical holes. The peel energy was low for the plates in the absence of holes. An energy balance analysis gives the ratio of fracture energy for peeling from the material with cylindrical pores G_a' to that from a smooth substrate G_a as

G_a' / G_a = 1 + 4 f l /a (5)

where l is the pore length, a its radius and f the ratio of pore area to total area of the plate[105]. Their experimental results demonstrated the essential validity of this relationship. Where pull-out alone occurred the work of detachment for their system increased by up to 20 times.

They further considered the situation where fracture of strands occurred. The extra work is proportional to depth of pores and for their system could be several hundred times the work of detachment from a smooth surface. This energy term dominates for deep pores.

There are obvious similarities between the polymer which has solidified within the pores of a microfibrous surface and fibres embedded in the matrix of a composite material. Standard treatments of fibre composites [e.g. 114] draw attention to the significance of the critical length of fibre. When short fibres are stressed axially, shear failure at the fibre/matrix adhesion is considered to occur, and the fibres may be pulled out of the matrix. Fibres greater than the critical length, with a consequently larger fibre/matrix surface area, fail in tension, and only the broken ends are pulled out. This, of course, is one of the points that Gent and Lin were demonstrating. The fracture toughness of the composite may be enhanced by energy terms associated with fibre fracture, with fibre/matrix adhesion and with fibre pull-out. By assuming that the fibre is linearly elastic and equating the interfacial shear force to the tensile force for a fibre of critical length l, it immediately follows that

2l / a = s/t (6)

where a is the fibre radius, s its tensile strength and t the interfacial shear strength. As in equation 5, the l / a ratio is significant.

Arslanov and Ogarev[7] use equation 6 to argue that the critical length of an adhesive in a microporous anodic film is very small, so the adhesive filaments will fail in tension and the most of the pore length is irrelevant to adhesion. Application of the simple model of equation 5 to this situation shows that even with a short length of elastic adhesive a useful increase in peel strength might be expected. For polyethylene embedded in a film formed by anodising in phosphoric acid, a ratio G_a' / G_a of three to four times is obtained.

In a realistic situation the adhesive filament will not act as a perfect elastic body uniformly stressed up to fracture. Uneven stress distributions and plastic yielding would be expected to increase the energy dissipation observed beyond that calculated for the ideal elastic model.

While calculations like those discussed involve serious simplifications and idealisations, they do serve to show that surface roughness per se is capable of increasing the fracture energy of an adhesive joint by a significant amount.

The mechanical theory of adhesion is associated with adhesion to rough and porous surfaces. Adhesion to such surfaces may be effective because their intrinsically high surface energy gives rise to a high work of adhesion in the bond. When stressed these surfaces may be able to redistribute the stress so as to increase energy dissipation during failure of the joint. These mechanisms are consistent with the theory discussed above, expressed in terms of equations 1 to 4.

McBain and Hopkins, like most commentators since them, tended to treat mechanical adhesion and specific adhesion (adhesion by way of the adsorption theory) as occurring by two entirely distinct mechanisms. Is this distinction really useful? Is there a danger of an unnecessary polarisation, where an integration might be preferable? Let us briefly examine adhesion as seen by the adsorption theory, considering adhesion to a 'smooth' surface.

If wetting is poor because of an inappropriate thermodynamic relationship between surface energies of adhesive and substrate, or because there is insufficient time for equilibrium to be reached before the adhesive sets, poor adhesion is likely to result because of a reduced area of contact and because of stress concentrations associated with interfacial voids. In terms of equations 1 and 2, Wa is low (part of the interface is air) and crack length l is increased. These considerations apply in the same way in the context of the mechanical theory. For thermodynamic or kinetic reasons poor penetration of adhesive into the interstices of the surface may result[11].

If wetting is complete, adsorption of the adhesive is considered to occur whether the surface is smooth or rough, but poor adhesion may result from low work of adhesion, a result of low surface energy. Adhesion may be enhanced by surface treatments some of which increase the surface energy and thence Wa. The treatment which raises Wa may be one which essentially alters the chemical groups in a smooth surface (although some topographical change will usually accompany it), or it may drastically alter the topography of the surface (although some chemical change will usually accompany it). Although Wa is increased the measured adhesion, G or P above, may be increased by a far greater extent (equation 4). Adhesion fracture energies are commonly far higher than the energies that can be ascribed to interfacial bonds, even to strong primary bonds. The treatment which increases surface energy is effective because it enables much more energy to be absorbed during fracture: it increases y (equations 2 and 3). This energy may be absorbed in regions near the interface or within the bulk phases present. This accounts for the widespread observation that high adhesion is associated with cohesive failure of the joint. Both adsorption and mechanical model work in a similar way, i.e. with an increase of work of adhesion leading to failure away from the interface.

Sometimes despite relatively strong interfacial forces, adhesion is low because stresses are strongly concentrated at the interface when the joint is loaded. This may be a result of a sharp discontinuity of modulus between adhesive and substrate. This is equivalent to introducing a critical defect, size l, close to the interface. Some surface treatments interpose a layer of intermediate modulus, lowering the stress concentrations and enabling larger volumes of material to be involved in energy dissipating viscoelastic or plastic deformation. The action of silane coupling has been described in such terms. The effect of an interfacial 'composite' region [51] where substrate and adhesive merge over a distance of nanometres or micrometres will similarly smooth out a sharp modulus change between the two phases.

It is important to recognise that the bias towards analysis, which is a characteristic of the scientific method, is not usually an end in itself, but a means of approaching the objective of achieving an overall rationalisation of phenomena[115]:

Discussion of adhesion phenomena separately in the terms of adsorption and mechanical theories may sometimes help to a clarify our ideas, but ‘adsorption’ and ‘mechanical’ effects can rarely be completely isolated from one another. Each theory emphasises a different aspect of a more comprehensive model which, in principle, relates molecular dispositions in the region of the interface to macroscopic properties of an adhesive joint. It might be a mistake today to lay too much emphasis on the distinction between the two classical theories, although this has been valuable at various times during the last seventy years in stimulating the development of new concepts and in suggesting fruitful experiments.

1. I. Newton, 'Opticks' 4th edn, book 3, part 1, query 31, 1730.

2. J.W. McBain and D.G. Hopkins, J. Phys. Chem. 29, 188 (1925).

3. R.C. Rinker and G.M. Kline, U.S. National Advisory Committee on Aeronautics, Tech. Note 989, August 1945.

4. F.W. Reinhart in Adhesion and Adhesives: Fundamentals and Practice, p.9 Society of Chemical Industry, London, 1954.

5. W.C. Wake, Adhesion and the Formulation of Adhesives, pp. 65-71, Applied Science, London, 1976.

6 J.D. Venables, J. Mater. Sci.19, 2431(1984).

7. V.V. Arslanov and V.A. Ogarev, Prog. Organic Coatings 15, 1(1987).

8 W. Brockmann, O.D. Hennemann, H. Kollek and C. Matz, Intll. J. Adhesion and Adhesives, 6, 115(1986).

9. G.D. Davis, Surf. Interf. Anal. 17, 439(1991).

10. G.W. Critchlow and D.M. Brewis, Intl. J. Adhesion Adhesives 15, 161 (1995).

11. D.E. Packham in Adhesion Aspects of Polymeric Coatings, (K.L. Mittal, ed.),

Plenum, New York, 1983, p. 19.

12. D.E. Packham, J. Adhesion 39, 137(1992).

13. F.L. Browne and T.R. Truax, Colloid Symposium Monograph 4, 258 (1926).

14 J.W. McBain and W.B. Lee, Ind. Eng. Chem. 19,1005(1927).

15. R.C. Rinker and G.M. Kline, Modern Plastics, p.168 (Nov. 1945).

16. W.C. Wake in Adhesion and Adhesives (R. Houwink and G. Salomon eds.),

vol 1, p. 405, 2nd edn. Elsevier, Amsterdam, 1965.

17. K. Bright, B.W. Malpass and D.E. Packham, Nature 223, 1360(1969).

18. D.J. Arrowsmith, Trans. Inst. Metal Finishing 48, 88(1970).

19. K. Bright, B.W. Malpass and D.E. Packham, Brit. Polym. J. 3, 205(1971).

20. K.W. Allen, H.S. Alsalim and W.C. Wake, Faraday Special Disc. Chem. Soc. No. 2 Solid/Solid Interfaces, Chemical Society, London, 1972, p.38.

21. O.D. Hennemann and W. Brockmann, J. Adhesion 12, 297(1981).

22. G. Farkas, Surfaces 14(93), 37(1975).

23. A.W. Bethune, SAMPE J. 11(3), 4(1975).

24. J.C. McMillan, J.T. Quinlivan and R.A. Davis, SAMPE Q. 7(3), 13(1976).

25. J.A. Marceau, SAMPE Quarterly 9(4), 1(1978).

26. Boeing Process Specification, Phosphoric Acid Anodising of Aluminum for Structural Bonding, BAC 5555, revision A, 1975.

27. J.D. Venables, J. Adhesion 39, 79(1992).

28. J.M. Chen, T.S. Sun, J.D. Venables and R. Hopping, Proc. 22nd SAMPE Symposium April 1997, p. 25.

29. T.S. Sun, J.M. Chen, and J.D. Venables, Appl. Surface Sci. 1, 202(1978).

30. J.D. Venables, D.K. McNamara, J.M. Chen, T.S. Sun and R.L. Hopping, Appl. Surface Sci. 3, 88(1979).

31. T.S. Sun, D.K. McNamara, J.S. Ahearn, J.M. Chen, D. Ditchek and J.D. Venables, Appl. Surface Sci. 5, 406(1980).

32. J.P. O'Sullivan and G.C. Wood, Proc. Roy. Soc. A 317, 511(1970).

33. J.A. Bishopp, E.K. Sim, G.E. Thompson and G.C. Wood, J. Adhesion 26,

237(1988).

34. A. Roche, Vide, Couches Minces, no. 257, 197 (1991).

35 K.W. Allen, J. Physique IV 3, no.C7 pt. 2, 1511(1993).

36 K.W. Allen, Intl. J. Adhesion Adhesives 13, 67(1993).

37. A. Bogner, Holz als Roh-und Werkstoff, 49, 271(1991)

38 D.E. Packham and C. Johnston, Intl. J. Adhesion Adhesives 14, 131 (1994).

39 H. Kanai J. Mater Sci. 26, 1815(1991).

40 C.U. Ko and J.P. Wightman, J. Adhesion 24, 93(1987).

41. D.E. Packham, Intl. J. Adhesion Adhesives 6, 225(1986).

42. J.B. Hollenhead and J.P. Wightman, J. Adhesion 37, 121 (1992).

43. J.A. Filbey and J.P. Wightman, J. Adhesion 28, 1(1989).

44. Y.Moji and J.A.Marceau, Anodising titanium to promote adhesion, U.S. Patent 3959091 (1976) {Chem. Abs. 85, 70057(1976)}.

45. Y.Moji and J.A.Marceau, Anodising titanium to promote adhesion, U.S. Patent 3989876 (1976) {via Chem. Abs. 85, 200102(1976)}.

46. R.G. Schmidt and J.P. Bell, Adv. Polym. Sci. 75, 33(1986).

47. B.W. Malpass, D.E. Packham and K. Bright, J. Appl. Polym. Sci. 18,

3249(1974).

48. J.P.Sargent, Intl. J. Adhesion Adhesives 14, 21(1994).

50. R.P. Digby, Ph.D. Thesis, University of Bath, 1996

51. R.J. Davies and M.D. Ritchie, J. Adhesion 38, 243(1992).

52. W.J. Russell, J. Appl. Polym. Sci., Appl. Polym. Symp. no 32, 105(1977).

53. N.L. Rogers, Proc. 13th National SAMPE, 1981, p. 640.

54. R.P. Digby and D.E. Packham, Vide, Couches Minces no.272 SS, p.586 (1994).

55. F. Keller, M.S. Hunter and D.L. Robinson, J. Electrochem. Soc. 100, 411(1953).

56. G.C. Wood and J.P. O'Sullivan, Acta Electrochimica 15, 1865(1970).

57. D.J. Arrowsmith and A.W. Clifford, Intl. J. Adhesion Adhesives, 5, 40(1985).

58. R.P. Digby and D.E. Packham Intl. J. Adhesion Adhesives, 15, 61(1995).

59. H. Reiter in Handbook of Adhesion (D.E. Packham, ed.) p. 192, Longman, 1992,.

60. H.M. Clearfield, D.K. Shaffer, S.L. Vandoren and J.S. Ahearn, J. Adhesion 29,

81(1989).

61. R.A. Pike, V.H. Patarini, R. Zatoski and F.P. Lamm, Intl. J. Adhesion Adhesives, 12, 227(1992).

62. G.D. Davis, D.K. Shaffer, P.L. Whisnant, G.B. Groff and J.D. Venables, 3rd Intl.

Conf. Adhesion and Surface Anal., Loughborough, April 1994, paper L13.

63. G.D. Davis, P.L. Whisnant, D.K. Shaffer G.B. Groff and J.D. Venables, J.Adhesion Sci Technol. 9, 527 (1995).

64. A.E.P. Morris and M.E.R. Shanahan, Intl. J. Adhesion Adhesives 14, 145 (1994).

65. A.M.T. van der Putten, J. Electrochem. Soc. 140, 2376 (1993).

66. G.W. Critchlow and D.M. Brewis, Intl.J. Adhesion Adhesives 15, 173 (1995).

67. C. Keisler and J.L. Lataillade, J. Adhesion Sci. Technol. 9, 395 (1995).

68. J.R.G. Evans and D.E. Packham, J. Adhesion 10, 177(1979).

69. P.J. Hine, S. El Muddaris and D.E. Packham, J. Adhesion 17, 207(1984).

70. L.A. Reeves, D. Kiroski and D.E. Packham, Surf. Interf. Anal. 23, 299(1995).

71. D.E. Packham (Ed.) Handbook of Adhesion, Longman, 1992, passim.

72. R.A. Ryntz, Prog. Organic Coatings 25, 73 (1994).

73. D.M. Brewis (Ed.) Surface Analysis and Pretreatment of Plastics and Metals, Applied Science, London, 1982.

74. D. Hace, V. Kovacevic and D Pajc-Liplin, 3rd Intl. Conf. Polymer Characterisation, Denton, Texas, Jan 1995.

75. S. O’Kell, M. Aindow and C. Jones, Vide, Couches Minces no. 272, 420 (1994).

76. S. O’Kell, T. Henshaw, G. Farrow, M. Aindow and C. Jones, Surf. Interf. Anal 23, 319 (1995).

77. K. Harth and H. Hisbt, Surface Coatings Technol. 59, 350(1993).

78. M.M. Pastor Blas, M.S. Sánchez Adsuar and J.M. Martín Martínez, J. Adhesion

Sci. Technol. 8, 1093 (1994).

79. T.P. Nguyen, K. Amgaad, M. Cailler and V.H. Tran, J. Adhesion Sci. Technol. 8, 821 (1994).

80. J.T. Marchesi, H.D. Keith and A. Garton, J. Adhesion 39, 185 (1992).

81. L.P. Buchwalter and R. Saraf, J. Adhesion Sci. Technol. 7, 925 (1993).

82. L.P. Buchwalter, J. Adhesion Sci. Technol. 9, 941 (1993).

83. N. Inagaki, S. Tasaka and K. Hibi, J. Adhesion Sci. Technol. 8, 395 (1994).

84 R.F. Saraf, J.M. Roldan and T. Derderian, IBM J. Res. Develop. 38, 441 (1994).

85. S.J. Shaw in Handbook of Adhesion (. D.E. Packham, ed) , p. 319, Longman, 1992.

86 R.F. Saraf, Macromolecules 26, 3623(1993).

87 N.C. Stoffel, E.J. Kramer W. Volksen and T.P. Russell, Polymer 34, 4524(1993).

88. D.J.D. Moyer and J.P. Wightman, Surf. Interf. Anal. 14, 496-504(1989).

89. B.Z. Jang, Composites Sci. Technol. 44, 333 (1992).

90. Y.L. Yuan, S.S. Shuy and J.Y. Lai, J. Appl. Polym. Sci. 42, 2525(1991).

91. T.R. King, D.F. Adams and D.A. Buttry, Composites 22, 380(1991).

92. T.R. King, D.F. Adams and D.A. Buttry, Composites Sci Technol. 44, 351(1992).

93. P. Marshall and J. Price, Composites 22, 388(1991).

94. I.M. Ward and N.H. Ladizeski, in, Composite Interfaces, (H. Ishida and J.L.

Koenig eds.), p. 37, North Holland, New York, 1986.

95. B. Tissington, G. Pollard and I.M. Ward, Composites Sci Technol. 44, 197(1992).

96. B. Tissington, G. Pollard and I.M. Ward, Composites Sci Technol. 44, 185(1992).

97. M.S. Silverstein and O. Breuer, Composites Sci. Technol. 48, 151 (1993).

98. M.S. Silverstein, O. Breuer and H. Dodiuk, J. Appl. Polym. Sci. 52, 1785 (1994).

99. K.L. Mittal, IBM J. Res. Develop. 38, 346(1994).

100. R.J. Good, J. Adhesion 4, 133(1972).

101. D.E. Packham, , Intl. J. Adhesion Adhesives 16, 121-128(1996).

102. T.A. Hatzinikolaou and D.E. Packham, 3rd International Adhesion Conference,

York, September 1987, paper 8

103. A.N. Gent and A.J. Kinloch, J. Polym. Sci. A2 9, 659(1971).

104. E.H. Andrews and A.J. Kinloch Proc. Roy. Soc. A 332, 385 and 401(1973).

105. A.N. Gent and C.W. Lin, J.Adhesion 32, 113(1990).

106. A.N. Gent and S.M. Lai, Rubber Chem. Technol. 68, 13(1995).

107. B. Bhushan, J.N. Israelachvili and U. Landman, Nature 374, 607(1995).

108. Y.L. Chen, C.A. Helm and J.N. Israelachvili, J. Phys. Chem. 95, 10736(1991).

109. H. Yoshizawa, Y.L. Chen, and J.N. Israelachvili, J. Phys. Chem. 97, 4128(1993). 110. D.E. Packham, K. Bright and B.W. Malpass, J. Appl. Polym. Sci. 18,

3237(1974).

111. A. Marceau, SAMPE Q. p. 1 (July 1978).

112. D.E. Packham in Developments in Adhesives - 2, (A.J. Kinloch, ed.) Applied Science Publishers, 1981.

113. T A Hatzinikolaou and D.E. Packham in Adhesion 9 (K.W. Allen,ed.) p.33 Elsevier Applied Science, London, 1984.

114. D. Hull, An Introduction to Composite Materials, Cambridge University Press,

1981, p. 142 et seq.

115. W. Wordsworth, The Prelude, II, 217-219. Text of 1805, Oxford University Press, 1970.

Epoxy resin reinforced with polyethylene fibres:

effect of fibre treatment on inter-laminar shear strength (ILSS)

and impact energy (after ref. 94).

Simplified schematic representation of the porous surface produced

on aluminium by the treatments (a) FPL etch, (b) anodising in phosphoric acid,

(c) anodising in chromic acid. After [33].

Microfibrous oxide formed on steel by oxidation in moist nitrogen at 400°C

Scanning electron micrograph, base +5.5mm[41].

Adhesion of epoxy resin to aluminium anodised in chromic acid to give different oxide porosities. Increasing fine structure of the oxide corresponds with increasing peel strength [48].

Crack growth on exposure to 96% relative humidity at 56°C for Boeing wedge test specimens of aluminium bonded with epoxy resin. The P2 pre-treatment is compared with the FPL etch used both alone and prior to phosphoric acid anodising (PAA) [54].

Change in strain energy release rate (G_I) on exposure to 96% relative humidity at 56°C for Boeing wedge test specimens of phosphoric acid anodised (PAA) aluminium

bonded with epoxy resin. The effect of a 10 sec phosphoric acid

post-anodising dip (PAD) is shown [58].

Scanning electron micrographs showing the change in surface structure of phosphoric acid anodised aluminium after a phosphoric acid post-anodising dip (PAD)

for (a) 10 sec, (b) 20 sec, (c) 30 sec [58].

Adhesion of copper to a silica layer achieved via deposition of

titanium tungstide islands[65].

Example of the nanoscale roughness on polyimide produced by phase separation [84].

Formation of a condensation polyimide [after 85].

4. Discussion: current status of the theory

5. Conclusions

Griffith-Irwin theory of fracture to a joint comprising a bond between two phases The fracture stress, sf, is given by

sf = k(EG/l)½

G = Wa (or Wc) + y

Usually y >> Wa (or Wc)

P = Wa (or Wc) + yplast. + yv/e + ......

y >> Wa (or Wc)

P = W × f

where f is a temperature and rate dependent viscoelastic term.