Interfacial chemistry of adhesive joint failure: an investigation by small area XPS, imaging XPS and TOF-SIMS Stephen J. Davis and John F. Watts* Department of Materials Science and Engineering, University of Surrey, Guildford, Surrey, UK GU2 5XH The strength and mode of failure of adhesively bonded iron substrates has been investigated. Joints were assembled from polished iron and from iron substrates that had been pretreated in a novel manner by the cathodic deposition of yttrium. Joints were exposed to an aqueous environment and a reduction in joint strength with increasing exposure time was noted. Failure surfaces of joints exposed to water for 1200 and 7500 h were examined by small-area XPS, imaging XPS and TOF-SIMS. An interfacial failure was observed, and although all surfaces were rich in nitrogenous species from the curing agent, no evidence of epoxy residues on the interfacial metal failure surfaces was recorded by TOF-SIMS.The metal failure surface at extended exposure times shows that there are two delamination fronts advancing from the exposed edge of the adhesive joint: one is associated with bond cleavage and gives rise to a 'zero-volume debond' and proceeds faster than the other, which is associated with gross separation of the adhesive and substrate, oxide growth on the exposed metal adherend and mass transport within the disbondment crevice. Adhesive bonding of metal components has been practised in the aerospace industry for many years. During this time a variety of pretreatment methods have been devised that provide excellent durability for joints specified for strategic appli- cations, which invariably involve the adhesive bonding of aluminium, titanium and their alloys.Recent years have, however, seen a widespread increase in the use of adhesive bonding for automotive and other high-volume applications which also require very good durability characteristics. Consequently, recent research has concentrated, not on pro- ducing an exceptionally strong bond, but on understanding the manner in which joint strength is reduced on exposure to aggressive environments, and developing pretreatments for steel substrates that will afford improved durability. A key step in understanding the mechanism of adhesive bond failure is the definition of the exact locus of failure.Surface analysis techniques such as X-ray photoelectron spec- troscopy (XPS) have been used for a number of years by several authors to study the exact locus of failure of adhesive bonds and organic coatings,',2 more recently, imaging spectro- scopies have been utilised to investigate the spatial distribution of failure sites.3 The ability to detect very low levels of adhesive, associated with a thin overlayer, has enabled very precise failure mechanisms to be suggested. It has been shown by several authors, for example ref. 4, that the composition of such overlayers, possibly affected by the segregation of minor components, can greatly affect the durability of adhesive bonds. Water ingress in adhesive bonds is known to greatly reduce their performance and has been highlighted as playing the major role in failure of environmentally exposed adhesive as surface energy considerations lead to a preference for a metal-water interface over a metal-polymer interface.* The nature of the weakening effect is, however, unclear.Cathodic disbondment has been identified as a route to failure in the case of steel substrates since 1929,9 however, there is still much debate as to the exact mechanism. Several workers have reported the major methods of failure due to cathodic delamination of organic Surface pre-treatments also play an important role in the performance of adhesive bonds; in the study described in this paper a rare-earth-metal pre-treatment described pre-vi~usly,~~-'~which involves forming a thin layer of yttrium hydroxide on a mild steel or iron adherend prior to bonding, has been considered. For this approach to work successfully it is necessary to form a tenacious coating of cationic yttrium which is not disrupted by environmental attack.Analysis of failed lap shear joints has enabled the exact locus of failure to be determined when such a pre-treatment is employed. In this paper we report the use of XPS (in small area and imaging modes) and time of flight secondary ion mass spec- trometry (TOF-SIMS) to study the failure surfaces of lap shear joints exposed to an aggressive environment (water at 30°C). The aim of the work is to use these surface analysis methods to provide analytical data, from which it is possible to deduce the interfacial chemistry of failure and provide a comprehensive model of the chemistry and electrochemistry that contributes to the observed reduction in joint strength.Methods and Materials Adhesive bonds In this study, single lap shear joints have been used as a method of comparing adhesive joint durability for joints with and without an inorganic adhesion promoter. Although not an ideal geometry offering fully quantifiable results in terms of fracture mechanics, it is suitable, as the failure load can be used in a relative manner to compare sets of samples. The single lap joint has several advantages over more complex joint geometries. To enable good surface analysis of the failed joint surface to be undertaken, it is necessary that the specimen is analysed in the 'as-failed' state.This limits the use of systems such as the double cantilever beam or Boeing wedge tests, as the failed surface is exposed to the testing environment; this results in post-failure attack which will dominate the results and possibly lead to erroneous conclusions being drawn about the actual failure mode. Adherends were made from a pure iron (Goodfellows, 99.5%) 1.02 mm thick sheet which was cut to size using a laboratory guillotine. A schematic of the lap shear joint used is given in Fig. 1. XPS of the as-received iron sheet showed the presence 1.02 mm IRON ADHEREND Fig. 1 Schematic of a lap shear joint J. Muter.Chem., 1996, 6(3), 479-493 479 of manganese at the surface. This is likely to be a result of segregation during the rolling or subsequent closed-coil annealing processes used to produce the thin sheet. The iron adherends were abraded with fine alumina powder prior to bonding to remove the surface-enriched manganese (confirmed using XPS). After abrasion, the iron was ultrasonically cleaned using acetone. A fully formulated two-part amine-cured epoxy adhesive (Ciba Geigy Araldite 2013) was used in this work. This was cured at 40°C for 16 h to ensure a high crosslink density. The glue line thickness was set at 150 pm; this was controlled by using two nylon wires as spacers in the adhesive joint. Nylon was selected as opposed to the more commonly used copper to minimise galvanic corrosion. Two series of lap shear joints were produced, one with the yttrium pretreatment and one control.A set of eight (four of each) were pulled to failure after cure to provide a control. Mechanical testing was performed on a J. J. Lloyd Tensile Tester, operated at 0.5mm min-' cross-head speed at room temperature. The remaining joints were immersed in milli-Q water (resistivity= lo1*Cl cm-') in sealed glass jars at 30 "C. Joints were removed periodically from the test solution, dabbed dry using a tissue and pulled to failure within as brief a time as possible (1-2 min) after being removed from the solution. The load to failure was recorded for each failed joint. Surface Analysis XPS Small-area and imaging XPS were used to study the failed lap shear joints.Standard XPS is an area integrating technique and has a relatively large analysis area (10 mm2). In this work it was essential to be able to analyse different areas within the failure region, thus a higher spatial resolution was necessary. Analysis of the failure regions was carried out using a VG Scientific ESCALAB 220i (BP Research, Sunbury) and a Scienta ESCA 300 spectrometer located at the RUST1 facility at Daresbury Laboratory. Both instruments offer high spatial resolution; however, the spectral resolution of the Scienta instrument was greater than that of the VG Scientific ESCALAB 220i (housed at BP Research, Sunbury). The XPS acquisition parameters used are given in Table 1.The Scienta ESCA 300 instrument was used to produce both point analyses and also energy-distance (E-X) plots. E-X plots enable both chemical and spatial information to be gained from one analysis. They are produced by plotting binding energy us. position, with the spectral intensity being represented as a heat scale on the z-axis (normal to the page). These plots are described in the literat~re'~.'~ and Plate 1 indicates how they are related to a complete spectrum and position on the specimen; the single spectrum of Plate 1 is effectively a hori- zontal line scan across the plane of the detector, the heat scale of the image now being transformed into the intensity scale of the spectrum. A montage of spectra can also be re-created from an E-X plot, as shown in Fig.2. Such analyses were carried out on different regions within the failure surface of 41 Intensity (Slit length/magnification) +aberrations ie 35/10 mm +50pm =3.55 mm 4 ( * ) !(Slit widWmagnification) +aberrationsSCAN AREA ie 80pm +50pm =130pm Plate 1 Creation of an E-X plot Distance 12 3 4 5 6 7 8 9 1C Split the E-X scan up mto a series of spectra Montage of spectra Fig. 2 Creation of a montage from an E-X plot the adhesive joint, thus enabling chemical state changes to be detected as a function of position. The analysis area for the Scienta ESCA 300 instrument is dependent on the mode of operation. For high-transmission analyses, the analysis area is defined by the X-ray footprint (6.0 x0.5 mm).However, for E-X scans an additional lens is added within the electron optics of the analyser chamber which has the effect of increasing the magnification of the transfer lens by a factor of ten. In this high spatial resolution mode the analysis area perpendicular to the hemispherical analyser acceptance aperture slit is given as (slit width/ magnification +spatial resolution), and parallel to the slit as (slit length/magnification +spatial resolution). It has been shownI7 that spherical abberations in the lens are small, and spatial resolution is typically 25 pm. Thus, for an E-X image with a 0.8 mm slit width, the analysis area displayed in the image produced is [3.5 mm +(2 x25 pm)] x [80 pm + (2 x25 pm)], i.e. 3.55 mm x 130 pm. In this work the VG Scientific ESCALAB 220i instrument acquisition parameter X-ray source X-ray power monochromated pass energy/eV survey high resolution take-off angle (to the sample surface)/degrees analysis area 480 J.Muter. Chem., 1996, 6(3),479-493 Table 1 XPS analysis conditions VG ESCLAB 2201 Al-Kr 15 kV, 20 mA, 300 W no 100 40 90 4x lo5 pm2 Scienta ESCA 300 Al-Ka 14 kV, 200 mA, 2.8 kW Yes 300 150 90 main lens =6.0 x0.5 mm E-X scan= 3.55 mm x 130 pm Fig. 3 Analysis regions across the failed lap shear joint after 7500 h in water at 30°C (a) XPS, (b)TOF-SIMS was used to study the failure surfaces ofjoints after environmen- tal exposure for 1200 h. After this time, two different failure regions could be visually identified.Initial results from these failure surfaces suggested that failure was truly interfacial between the adhesive and the adherend. This led to the Scienta ESCA 300 being used in future analyses of specimens exposed for 7500 h in water, as this spectrometer has better spectral resolution and has been used previously to precisely pinpoint the locus of failure in a complex system.'* Using the ESCA 300, XP survey spectra were recorded along a line from the specimen edge towards the central cohesively failed region. Starting from the overlap edge, spectra were recorded every 1 mm until the thick cohesively failed region was reached, this is shown schematically in Fig. 3(u) for the failure surface of a joint exposed to water for 7500 h.These survey spectra were then quantified using the appropriate sensitivity factors. The atomic percentage of each element present was then plotted against position to give elemental compositions across the surface. When displaying these data, the results were duplicated (assuming symmetry) to represent the entire joint overlap. This was undertaken to enable the results to be compared with those from TOF-SIMS in which analyses were acquired across the entire width of the overlap. These montages allow differences in spatial distribution of elements to be observed readily. Results from the VG Scientific ESCALAB 220i instrument are presented with the binding energy increasing from left to right; however, results from the Scienta ESCA 300 have the binding energy decreasing from left to right (increasing kinetic energy).This form of discrepancy arises as there is no univer- sally accepted format for plotting ionisation potentials and different manufacturers have adopted, for essentially historic reasons, either right-handed or left-handed formats. TOF-SIMS TOF-SIMS was utilised to provide further details regarding the organic material remaining on the metal side of the failed Table 2 Experimental conditions for TOF-SIMS analysis primary beam 26 kV Ga69+ ions specimen current/nA 2 pulse width/ns 20 pulse per pixel 32767 number of frames 50 magnification x 500 electron-flood gun/mA 0.1 3.0total ion dose (analysis)/lO" ions cm-2 joint.The high sensitivity combined with the ability to detect fragments of the adhesive system enabled SIMS to be used to fingerprint the adhesive. Analyses were made using a VG Ionex Type 23 mass spectrometer fitted with a Poschenreider analyser and a 30 kV gallium liquid-metal gun. Data were acquired using VG Ionex VGS 5100 software operating under RTll on a PDPll computer. This was then transferred to a PC via a conversion package to enable further processing using spreadsheet software (Borland Quattro Pro 4). Typical analysis conditions are given in Table 2. These parameters result in a static SIMS analysis, i.e. the total ion dose per analysis is well below the suggested threshold value for damage, of lOI3 ions cmP2.l9 All the SIMS data were normalised by dividing the intensity of the mass peak of interest by that of the total counts in the m/z 0-200 region minus gallium counts, i.e.[counts in peak]/ {[total counts m/z 0-2001 -[counts of (Ga6'+ +Ga7'+)]}. This corrects for the reduced total intensity which is observed during analysis and also for the increased signal associated with implanted gallium. Normalising also reduces any effect due to a variation in the specimen current which may occur between different analyses. To perform such normalisation it is necessary to process the raw SIMS data; this is achieved by transferring the intensities of each mass peak to a spreadsheet to facilitate inspection and subsequent normalisation. TOF-SIMS enables any fragments associated with the epoxy which may be present on the apparently metal failure surfaces to be detected.TOF-SIMS point spectra were recorded across both regions within the failure surface (i.e. thick cohesively failed epoxy and apparent metal surfaces) as shown sche- matically in Fig. 3(b).The analysis area for each spectrum was a 400 pm square. This enables plots of peak intensity us. position to be produced, thus allowing spatial differences to be detected easily. Results and Discussion Mechanical testing and observations on failure surfaces Results from the durability investigation are shown in Fig. 4, which shows that the yttrium treatment alone has had little affect on the failure load for these test conditions. The failure Fig.4 Durability results for lap shear joints (2013 adhesive) immersed in water at 30 "C for 1200 and 7500 h: yttrium treated sample us. control J. Muter. Chem., 1996, 6(3), 479-493 481 surfaces of the as-cured joints showed cohesive failure within the epoxy. This was evident to the naked eye as the thick black epoxy could clearly be seen on both surfaces. After 1200 h in water at 30°C the failure surface has two clearly defined regions. There is a thick cohesively failed region in the centre of the overlap surrounded by a shiny metal/smooth epoxy region. This outer region appears to be a result of interfacial failure. After 7500 h in water there are still two main regions of failure; however, as expected, the centre cohesively failed region is much smaller owing to prolonged exposure.Within the outer metal region there appear to be two different areas present as a halo around the centre region. The outer ring (zone 2) is slightly duller than the inner (zone 1). This is likely to be caused by oxidation of the metal surface on exposure to water. The failure surface can be defined by two failure fronts; front 1 is the fastest at 18.5 pm per day while front 2 is slower at 10.2 pm per day. These failure fronts are shown schematically in Fig. 5(u) and (b). To understand the mechanisms involved in the failure process it is necessary to consider the electrochemical activity present within the joint. The outer edges consist of exposed iron and will be anodic, whereas the inner regions (covered with the epoxy) will be relatively cathodic. As the epoxy is displaced (allowing access to the metal surface), this newly exposed region will become anodic and hence the corrosion path will move towards the joint centre.The following reactions will occur at the anode and cathode, respectively. anodic reaction: FetFe’’ +2e- (oxidation) cathodic reaction: 0,+2H,O +4e-+40H-(reduction) There are a number of possible cathodic reactions; however, the above equation is the most likely one in a non-acidic media and has been suggested as the most likely reaction to occur under an organic film permeable to oxygen.” Such reactions will result in environs of the crack tip becoming cathodic and the build up of OH-ions will result in a high pH.If an applied potential was present, this high pH could result in oxide reduction; however, at the free corrosion potential for iron, oxide reduction cannot occur.” This means that the inner shiny region is not cathodically reduced to give a thinner oxide, but rather that the outer region has experienced oxide growth during immersion in water. The extent of delamination (distance from the metal edge to the cohesively failed region) prior to catastrophic failure brought about by mechanical test has been investigated using Fig. 5 Schematics of (a) failed lap shear joint surfaces after 1200 and 7500 h in water at 30 “C;and (b)the disbondment fronts and crack tip 482 J. Muter. Chem., 1996, 6(3), 479-493 Fickian diffusion theory. The diffusion coefficient was deter- mined for the 2013 epoxy in water at 30°C by performing mass-gain experiments.Results indicated that if a critical water concentration is necessary for delaminatioq2, then failure is not controlled by water diffusing through the bulk of the adhesive. This was concluded as the water concentrations at the failure distances were not similar for the two exposure times when determined from modelled water penetration assuming Fickian diff~sion.’~It is likely that failure has occurred owing to enhanced interfacial diffusion of reactive species. Thus, the visual observations indicate that exposure to water brings about apparent interfacial failure between adhesive and substrate; the load-bearing ability of the joint following such exposure is a function of the area of the adhesive-substrate interface that has not been displaced by such electrochemical activity. This IS in agreement with the work of Gledhill and Kinl~ch,,~although they reported extensive corrosion of the steel substrate at the crevice mouth. This is presumably related to the test time and the ease with which the crevice mouth can open.In the present work a mere thickening of the oxide is observed rather than gross material degradation. A Cambridge Instruments Stereoscan 250 scanning electron microscope (SEM) was used to study the failure surface in more detail. Prior to analysis, the failure surfaces were first coated with a thin gold overlayer to prevent excessive sample charging. A 5 kV electron beam was also used during this investigation to minimise charging effects.During analysis, energy dispersive X-ray analysis (EDX) was performed in selected areas to determine the chemical composition of certain features. Despite the gold overlayer, meaningful results could be obtained using a windowed LINK EDX detector. As SEM is a bulk technique compared to more surface-specific analyses such as XPS and TOF-SIMS, care must be taken when assessing SEM micrographs. Results from SEM revealed the presence of many plate-like structures in the thick cohesively failed region; these are shown in Fig. 6(a). EDX was performed to determine their chemical composition. Typical plates have a maximum internal diameter (feret) of about 20 pm and are approximately circular in shape (they are not elliptical).EDX results indicate that the plates are iron- and potassium-rich compared to both the rough cohesively failed epoxy and the smooth epoxy on the epoxy side of the interfacial failed outer region [Fig. 6(b),(c) and (d)]. The smooth nature of these platelets in addition to their high potassium and iron content indicates that they come from a mica filler which has sheared along its potassium-containing cleavage planes, which are only weakly bonded by van der Waals’ forces. This cleavage process accounts for the smooth surface morphology detected by SEM. Mica is added to the adhesive as a filler and is therefore not unexpected at the fracture surface.Analysis with surface specific techniques (XPS) does not indicate the presence of these platelets in the bulk adhesive. This is because they are normally wetted by epoxy adhesive to a thickness greater than the analysis depth for techniques such as XPS (about 5 nm). Small-area XPS of the failed lap shear surfaces Small-area XPS (700 pm) was undertaken (VG Scientific ESCALAB 220i) on the failure surfaces of the 1200 h immersed sample in the outer failed region as well as the central cohesively failed region. The survey spectrum from the central region is presented in Fig. 7 and shows the expected peaks associated with carbon, oxygen and nitrogen (from the amine curing agent). There are also peaks associated with aluminium and silicon; these are likely to arise from additives used in the adhesive, e.g.the filler. Results from the outer region indicate that it has failed interfacially, and this is reinforced by studying the survey spectra and high resolution spectra for the carbon 1s peak. The metal side of the failed samples shows very low tion at 30°C27rather than immersed in water as in the earlier work, it is therefore unlikely that nitrogen was introduced from the test environment. It is probable that nitrogen present on the failure surfaces for the lap shear joints is either from a nitrogen-rich layer within the epoxy adhesive, or from re-adsorbed nitrogen from the adhesive or from nitrogen from the atmosphere which has adsorbed on the high-energy metal surfaces during failure.The nitrogen on the metal failure surface must be present as a discontinuous layer, because if it was present as a uniform overlayer, as an amine or similar adduct, the associated carbon signal detected on the metal failure surfaces would be much higher. The low iron signal on the epoxy side is likely to result from either the transfer of a small amount of iron from the metal substrate on failure or from the mica platelets in the epoxy. For the latter case to be true, the platelets must be at the surface and not covered with too thick an epoxy overlayer. If iron on the failed epoxy surface originates from the adherend oxide, then for the yttrium-treated sample such iron must result from the regions of back-deposited iron formed over the yttrium layer.This back deposition has been reported pre-vi~usly.~’The lack of any yttrium signal on the epoxy side indicated that failure has not occurred within the iron oxide beneath the yttrium layer. In view of the results associated with yttrium it is likely that the iron originates from the mica in the adhesive as the back-deposited regions only occupy a small fraction of the surface. XPS was also carried out on the failure surfaces of bonds after 7500 h exposure. These analyses were performed on the Scienta ESCA 300 spectrometer. XP survey spectra were acquired using the high-transmission mode (6 mm x 0.5 mm) from both the inner (shiny) region on the metal surface and also the outer (dull) region (Fig. 8). The spectrum from the inner region is comparable to that from a clean iron surface with its thin native oxide film.However, this inner region has a much higher chlorine signal than the outer region. The outer region has a reduced iron signal and higher oxygen, carbon and silicon signals. This can be seen in more detail by analysing the line scans presented in the next section. High-resolution core line spectra were also recorded for certain elements to enable chemical state information to be determined. The carbon 1s line acquired on a spectrometer with such high spectral resolution can often be used to determine the exact locus of failure, i.e. determine the presence of any organic overlayer by fitting the C 1s envelope. Such a spectrum is shown in Fig.9 for the ‘metal side’ of a failed joint. Although it is possible to produce an acceptable peak fit, as shown in Fig. 9, the result is still somewhat ambiguous as binding energies of individual components do not match well with those of adhesive and adventitious contamination expected on high energy surfaces. To resolve this uncertainty Plate 1 Ti Rough Cohesively Failed Epoxy Smooth Interfacially Failed Epoxy Si /Ads T’ I (4 0 2 4 6 8 10 12 14 16 18 20 Fig. 6 (a) SEM image of plate structures within the central cohesively failed epoxy region of a joint failed after 7500 h in water at 30°C. EDX results from: (b) plate structure in central cohesively failed region, (c) cohesively failed epoxy, and (d) smooth epoxy in the outer interfacial region.levels of carbon and a very clear iron signal, thus supporting the concept of interfacial failure. For the yttrium treated surface, there is a high yttrium signal and associated attenu-ation of the iron signal by the yttrium layer. The epoxy side, however, shows no yttrium and this confirms that a thin tenacious yttrium film has been deposited as failure has not occurred within the yttrium layer. The background slope following a peak can also offer much information about the hierarchy of species present on a specimen.25 Following the yttrium peaks at 301 and 313eV (Y 3p3I2 and Y 3~”~, respectively) the background has a negative post-peak slope (PPS), and this is indicative of a surface phase.25The nitrogen which is present on both metal surfaces could have originated from a number of sources.It may have arisen from water during the durability test” (no attempt was made to de-aerate the water), or from remnants of the amine curing agent which have segregated to the surface, from leaching of nitrogen-containing species from the adhesive which have re-adsorbed onto the surface26 or from the subsequent exposure of the failed surfaces to air on failure. Nitrogen was not present prior to bonding and this is confirmed by XPS analysis of an alumina-polished iron adherend. However, analysis of failed Boeing wedge test samples using this adhesive system show nitrogen to be present on the ‘metal surface’ even when the samples have been suspended in a controlled environment (75% relative humidity) created using a saturated NaCl solu-TOF-SIMS was used to fingerprint the epoxy adhesive and hence determine the presence of any residue on the ‘metal surface’.XPS line scans from the failed lap shear surfaces The results from quantified small area XPS analyses have been plotted us. position to give a representation of the sample surface across the entire joint overlap, i.e. an XPS line scan. These plots are given in Fig. 10 and 11 for the metal side and the epoxy side of the failed joint (7500 h in water), respectively. The centre of each plot corresponds to the thick cohesively failed epoxy, the data points either side of this central region correspond to the crack tip with the outer regions being the disbonded zone.Results for the carbon and oxygen signals indicate that the thick cohesively failed epoxy is relatively high in carbon and low in oxygen compared to the metal surface. This is to be expected as the clean iron oxide is high in oxygen and relatively free from carbon contamination. The epoxy side J. Muter. Chem., 1996, 6(3), 479-493 483 1 I I 1 1 I 1 1METAL: COWROL EPOXY :CONTROL40 1W-1 1 1 I 0-10-EPOXY :YTTRIUM --s-c 1s I binding energy/eV Fig. 7 Small area XP survey spectra from both the metal and epoxy side of a joint failed after 1200 h in water at 30 "C: yttrium treated sample us. control (VG ESCALAB 220i) shows the inverse of the metal side with a relatively lower carbon signal and higher oxygen signal compared to the smooth failed adhesive, although the differences are much less pronounced.The carbon signal is lower from the central region owing to higher signals from both oxygen and silicon; this indicates that silicon and oxygen are present in the cohesively failed central region. For the metal side, the iron signal is much higher at the crack tip; this is in keeping with the earlier result of Fig. 8 where survey spectra are compared from the inner and outer regions for the sample exposed to water for 7500 h. This indicates that the iron signal in the outer region is somewhat attenuated, and this is probably caused by a thicker oxide as well as other deposits which have formed during exposure to water over a long time.The iron signal on the epoxy side, although overall of much lower intensity, indicates that there is a reduced intensity for the cohesively failed region. The aluminium signal (for the metal side) is highest at the crack tip, with its intensity reducing towards the edges of the overlap. This is expected as the aluminium cations (from the adhesive) have segregated towards the cathode. The chlorine signal is highest at a point slightly away from the cathode. To understand the failure mechanism in more detail it is essential to determine the chemical nature of the elements present on the failure surfaces. To this end, the exact binding energies of both the chlorine and the silicon have been determined. High resolution spectra were recorded for both the C 1s and the C1 2p photoelectron lines.This enabled the 484 J. Mater. Chem., 1996, 6(3), 479-493 exact binding energy of the chlorine species to be determined. Spectra were recorded from both the metal and the epoxy side of the failed joint. The C 1s line was then fitted to determine the position of the C-C/C-H component. This was then given the value of 285.0 eV as a method of charge referencing. The C1 2p doublet was also fitted to determine the position of the C1 2p3/2 component. Its binding energy was then corrected using the previously determined level of charging. Results from analyses of the metal and epoxy side give a C1 2p3,, binding energy of 198.2 eV. This is in keeping with an alkali-metal chloride between an electropositive cation and an electronega- tive anion, i.e.Cl- from KC1 or NaCl which has predominately ionic bonding.28 A similar approach was adopted for silicon found in the outer region on the metal side and the binding energy of the Si 2p peak was determined as 102.2 eV. This is in keeping with a silicone or a silicate and suggests that the silicon in the outer region may have arisen from siloxanes present in the fully formulated adhesive, from contamination due to silanised glassware used during the durability testing or from the silicon (as a silicate) in the laboratory glassware itself. For the chloride ions, segregation away from the cathode is expected as anions present (i.e. chloride ions) are likely to be attracted to the anode (outer surface).This suggests that chloride ions from the adhesive segregate to the interface where they move towards the anode. Their inability to reach the anode is due to the very low volume debond (perhaps a 'zero volume debond') which occurs in the inner region (zone 1). 40' 3!j20. 0 I50i ::f 20 moo aoo 600 400 0 binding energy/eV Fig.8 Small area survey spectra from the inner (a) and outer (b) regions of a failed joint after 7500 h in water at 30 "C: control (Scienta ESCA300) 14t r 2 t 200 204 202 280 270 276 binding energy/eV Fig.9 XP high resolution C 1s spectrum from the metal side of a failed lap shear joint after 7500 h in water at 30 "C This makes movement of ions relatively slow.Their absence in the outer region (zone 2) suggests that transport is the controlling factor and that chloride ions in the outer regions are more mobile than those in the central (shiny) region. Increased mobility is caused by the bond in the outer region 'opening up'; this results in the chloride ions at the edge being displaced into the bulk solution by the movement of water within this more open region. The silicon signal for the metal side shows a slight increase in the central region but a large increase at the joint edge. The central signal is associated with the fillers, etc. in the adhesive; however, the very high signal at the edges is more likely to be due to post-failure contami- nation from the glassware used during the durability test.The far higher level of this contamination on the metal surface is expected as this surface will have a higher surface energy than the epoxy surface. The nitrogen signal from both the metal and the epoxy sides proves to be very interesting. Nitrogen is expected in the adhesive from the amine curing agent (aliphatic polyamine adduct hardener); however, the nitrogen signal is lowest for the cohesively failed epoxy. Both the metal surface and the smooth failed epoxy have a much higher nitrogen signal at the crack tip with the signal dropping (but still higher than for the cohesively failed epoxy) towards the edge of the joint. This suggests that either failure has occurred within a nitrogen-rich layer which has segregated towards the surface and has resulted in a discontinuous overlayer on the iron surface, or that nitrogen has been adsorbed onto the high-energy metal surface on failure.However, with the evidence available here it is not possible to determine the singular source of nitrogen on the metal failure surface. E-X images of the failed lap shear joint surfaces The images suffer from slight abberations across their width resulting from a reduction in counting efficiency in the centre of the channel plate detector. This is demonstrated in Plate 2(a), where a uniform gold band has been analysed.,' If the detector were perfect, the two spectral bands (from the 4f,,, and the 4f,,, photo-peaks) would have uniform intensity across the whole area of the detector. However, despite this slight inhomo- geneity, the results obtained can be used to study the transition between the two regions.E-X plots for C Is, 0 1s and Fe 2p are given in Plate 2 (b)-(d) for a region crossing between zone 1 (shiny) and zone 2 (dull) of a control sample. The C 1s and 0 Is images are not very clear and they remain fairly constant across the scan. The Fe 2p image of Plate2(d) is, however, far more informative. Both the oxide and metal components can be distinguished with a peak separation of about 5 eV. The metallic component reduces in intensity away from the crack tip. The fact that the metallic component does not show on the right of the E-X plot (away from the crack tip) indicates that the reduced intensity is due to oxide growth and is not merely a function of the analyser inhomogeneity.If this were the case, the metallic component would increase on the right-hand side also. To enhance this data, a montage of spectra has been recreated from the E-X plot (Fig. 12). Although quite noisy as a result of dissecting an image, this clearly shows the reduction of the metallic component across the sample surface. This indicates that the iron has a thinner oxide at the crack tip compared to that in the outer region. This result corroborates the visual inspection which clearly distinguishes the two regions as shiny and dull. The outer region (zone 2) is likely to experience oxide growth as it is relatively anodic compared with conditions at the crack tip, and is exposed (by the displacement of the epoxy) to a hostile environment.The as-received iron adherends will have an air- formed oxide typically 2 nm thick. TOF-SIMS spectra from 2013 epoxy adhesive Positive and negative SIMS of a failed polymer region from an adhesive bond was performed to determine the fragmen- tation pattern for the epoxy used in this work. Results from such analyses are given for cohesively failed epoxy [Fig. 13(a)]. Results indicate the presence of many silicon-containing frag- ments (m/z 28+, 43+, 73'); these are associated with siloxane contamination from the adhesive. The high m/z 39+ signal confirms the presence of potassium on the failure surface which has originated from fracture of the mica filler (SEM-EDX confirms the presence of platelet mica structures).The peak at m/z27' may be caused by the presence of aluminium at the cohesive fracture surface of the adhesive; however, this is unlikely as XPS results show a marked reduction of aluminium in the central epoxy region (Fig. 10). The m/z 27+ component is more likely to be caused by C,H,+ from the adhesive. Negative SIMS from the cohesively failed epoxy [Fig. 13( b)] reveals the presence of m/z 16- from oxygen, m/z 17- (hydrox- ide), m/z 12- (carbon), m/z 13- (CH-). There is also a small component due to nitrogen (m/z 14-) and CN- (m/z 26-). Such peaks are all associated with the adhesive system. The J. Mater. Chem., 1996, 6(3), 479-493 485 25.00 0 2 4 6 8 10 12 1 28.004 0 2 4 6 8 10 12 I 14 0.w7 7.00.6 00- 5.00- 4.00- s g 3.00-* 2! L O 8 2.007 2 4 6 8 10 12 ' 3.00 4 0 2 4 6 8 10 12 . 4.OO 0.004 0 2 4 6 8 10 12 2.004 0 2 4 6 0 10 12 ' 4 distance from joint edge/mm Fig. 10 XPS line scans (metal side) of a failed lap shear joint after 7500 h in water at 30 "C x-axis=distance from joint edge (in mm), y-axis = range of intensities for species present (concentration in atom%) (a) Carbon (Is), (b) oxygen (Is), (c) iron (2p),(d) nitrogen (Is), (e) silicon (2p), (f ) chlorine (2p), (g) aluminium (2p) 486 J. Mater. Chem., 1996, 6(3), 479-493 71.001 0 2 4 6 8 10 12 14 6.50 6.oO 5s 6.oo 4.50 1 I0.254 . ib 12 14 4.oo 0 2 4 6 6 0 2 4 6 10 12 ' 1.OO.c 0 2 4 6 8 I0 12 14 distance from joint edge/rnm Fig.11 XPS line scans (epoxy side) of a failed lap shear joint after 7500 h in water at 30 "C: x-axis=distance from joint edge (in mm); y-axis= range of intensities for species present (concentration in atom%). (a) Carbon (Is), (b) oxygen (Is), (c) iron (2p), (d) nitrogen (Is), (e) silicon (2p), (f) chlorine (2p). low intensity of nitrogen-containing species detected by SIMS precludes the nature of the surface nitrogen phase being determined. Spectra recorded for the 2013 adhesive are different to those from bisphenol A (which is the known base epoxy compound) and typical epoxy-related groups (m/z 135+, 191+, 252+ and 269+; 133-, 211- and 283-) are absent from the spectra recorded from the 2013 adhesive.Peaks in this fully formulated adhesive are associated with additives and indicate that these phases are dominant at the surfaces examined and are perhaps associated with a weakened zone within the adhesive which has failed preferentially. SIMS was also performed on the apparently metallic part of the failure region 2mm from the joint edge in zone 2 (see Fig. 3). These spectra are given for positive and negative ions in Fig. 14(u) and (b),respectively. The positive ion spectrum has dominant peaks at m/z 28' and 56+ from silicon and iron, respectively. However, the absence of peaks at m/z 43 and+ 73' shows that silicon is not present as a silicone or a siloxane, as peaks from Si(CH3) and Si(CH,), would be detected.XPS results confirm the presence of silicon at the joint edge and exact binding-energy measurements (Si 2p =102.2 eV) indicate that it is present either as a silicone or a silicate. The certainty of the XPS results (silicon is clearly present as both the Si 2p and Si 2s signals are very clear) confirm the presence of silicon in this region, therefore SIMS peak overlaps or doubly charged iron (mass 56/2) cannot explain the high m/z 28' component. From both the XPS and the SIMS results it is therefore clear that silicon in this outer region is present as an inorganic silicate rather than an organic silicon compound and that it has originated from the glassware used for durability testing. The negative ion spectrum is similar to that from the bulk adhesive; however, for the metal surface the m/z 12- (C-) and m/z 13- (CH-) peaks are smaller indicating a clean surface. The m/z 35- and m/z 37-(35Cl-and 37Cl-) peaks are com- paratively large for the metal region, this is an observation that is considered in more detail below.TOF-SIMSline scans of the failed lap shear surfaces TOF-SIMS was used to determine the exact nature of adhesive failure as the high sensitivity of SIMS allows very low contami- J. Mater. Chem., 1996, 6(3), 479-493 487 Plate 2 XPS E-X plots from the transition between the inner (shiny) and outer (dull) regions in a joint failed after immersion in water at 30°C for 7500 h. (a) Gold test band on silicon wafer, (b) C Is, (c) 0 Is, (d) Fe 2p. nant levels of species to be determined.Results from a series of point analyses across the joint overlap are displayed for mass fragments given in Table 3 as line scans (Fig. 15 and 16) in the same way as the XPS results. These plots allow the surface chemistry across the entire failure region to be examined. The m/z 56+ (56Fe) plot does not correlate with the m/z 28' (28Si) plot. This confirms that the m/z 56' peak is associated with iron and that it is not a double m/z28' (28Si) cluster peak. The iron signal varies across the failed region in a similar manner to the earlier XPS results, and this indicates that both techniques are suitable for analysing failure surfaces. The m/z 28' peak also shows a similar trend to the XPS results; however, the relative intensities are not the same.This is expected as the XPS quantification includes all silicon present (in different chemical states), whereas the SIMS data refers only to the m/z28+ fragment which is either present or is formed from fragmentation of larger fragments. The m/z 23' (Na+) component shows a clear trend with sodium present at the crack tip only. The level of sodium is very low, this is clear as it is not detected with XPS and the SIMS signal is relatively low considering the very high probability of sodium for ionisation. Sodium has segregated to the cathodic crack tip under the influence of the corrosion potential present in the system. The speed of ionic migration has been shown to depend upon the size of the solvation sheath which surrounds the ion in solution.The size of the solvation sheath is controlled by the charge/ionic radius.30 The detection of sodium at the crack 488 J. Muter. Chem., 1996, 6(3), 479-493 tip is expected as it is a small cation (charge/ionic radius= 9.80)31 which can easily migrate to the tip of the cathodic crevice. The m/z 40' fragment (calcium cation) is also intense around the cathodic crack tip. The speed of ion migration is about the same for calcium as for sodium as the charge/ionic radius for calcium ions is 10.15. This demonstrates that the movement of ions within the failure region is dictated by electrochemical considerations. The m/z 17- peak (OH-) confirms the high concentration of hydroxy ions at the crack tip.This is expected as they will be formed by the cathodic reduction of water at the cathode. The m/z35- and 37- data (35Cl- and 37Cl-) both show identical trends. The plots confirm that chlorine is present in the inner region near the crack tip and that it is absent from the outer surface of the joint overlap. This confirms the XPS result and once again raises the question of chloride ion mobility within the failed region. The chlorine profiles and the sodium profiles are not the same. It is likely that chlorine has reached the bond line (from the adhesive or water) and that it then starts to segregate towards the anode (outer region). This explains the displacement (compared to sodium) shown in the position of the chlorine-rich region towards the outer anodic region.The localised chemistry and migration of species within the adhesive bond are shown schematically in Fig. 17. The main aim of the TOF-SIMS investigation was to determine the exact locus of failure, i.e. was there a thin epoxy overlayer? Analysis of the results show that there are many mass peaks associated purely with the central adhesive region, 712 710 708 706 704 7 binding energy/eV Fig. 12 Montage of the Fe 2p peak showing the transition between shiny and dull regions within the failure surface (created from the E-X plot) 8000 'Si/CZH, cohesivelyfailed epoxy 7000 6000 K 5000 4000 3000 2000 -ul C3 1000 ;T& \8 x 0 0 10 20 3 40 50 60 70 80 90 100 al .-E 9000- 7 cohesively failed epoxy 8000- 7000- 6000- 5000-4000]300012000 CH; I IIP ,1111 0 10 20 30 40 50 60 70 80 90 100 masskharge Fig.13 TOF-SIMS spectra from the cohesively failed epoxy region of a failed lap shear joint after 7500 h in water at 30 "C:(a)positive ions; (b)negative ions 3500 (a "SiGH, =Fe metal surface 250030001 1000i 500-C 3 9 O HAr:111 -5 12000 .-(b) 0 metal surfaceC 10000-8000-6000-40001 c SCI 1112ooou0 0 10 20 30 40 50 60 70 80 90 100 masdcharge Fig. 14 TOF-SIMS spectra from the metal region (2 mm from the joint edge) of a failed lap shear joint after 7500 h in water at 30°C: (a)positive ions; (b)negative ions Fig. 16. Peaks at rn/z73+, 77+, 98+, 105+, 107' and 207' amongst others are all attributable to organic material.These mass fragments arise either from the bulk adhesive formulation (amine-cured epoxy) or from minor additions such as silicones/ siloxanes and fillers which are often added to fully formulated adhesives. Unfortunately, the exact content of the adhesive is unknown as it is a fully formulated commercial system. The clear discrimination between the central and outer regions for all of these mass fragments combined with the very high sensitivity of TOF-SIMS confirms that epoxy is not present on the apparently metal surface. This indicates that under the test conditions used here, true interfacial failure has occurred at the iron oxide+poxy interface; however, the presence of nitrogen on both sides of the failed joint (easily detectable by XPS) suggests either that failure has occurred through a surface nitrogen-rich phase or that nitrogen has back-deposited in the failed regions.General discussion The analysis of a failed lap shear joint has been investigated using a multi-technique approach. Failure of joints (both control and yttrium-treated) immersed in water is seen to involve a two-stage process and this can be observed visibly by inspecting the failure surfaces. The yttrium layer, although not enhancing adhesion in this case, has proven to be tenacious to environmental attack by water. Although bulk diffusion is thought to be a controlling parameter in the failure of adhesive joints,22 it has been shown that bulk diffusion theory cannot be used to model the failure of the system reported here.This confirms many previous beliefs that surface/interface diffusion rates are very different to those for bulk materials. This is likely as the surface (or very near surface) will have very different properties to the bulk in terms of crosslink density and porosity. Two clear regions develop on samples tested for 1200h in water, these are associated with bulk cohesively failed epoxy and an apparent interfacially failed region. XPS J. Muter. Chem., 1996, 6(3), 479-493 489 Table 3 Mass fragments analysed using line scans mass/charge ion/fragment 56+ 56Fe+ 28' 28s1+ 23' 'jNa+ 39 39K + + Y + +27' 27Al /CzH3 /CHN + H-Y=C-H 40' 40Ca+ 17 OH 35 c1 26 CN 73 28SiC3H9'+ 77 + 98' + 105' 0 0 II II107' H-SI-0--8-H + y3 207 + of this apparently metal region indicates that failure, if not uniformly interfacial, is pnmarily interfacial with possible small overlayer regions present This failure region appears identical to that identified as zone 1 on the longer exposed test sample After 7500 h in water, three regions appear on samples These are the central cohesively failed epoxy and a ringed outer 'metallic' surface which shows a bright metal surface at the joint centre (zone 1) and a duller region towards the joint edge which is a result of oxide thickening on exposure to water (zone 2), this is confirmed by XPS Analysis of the halo using small-spot analyses (Scienta ESCA 300) showed that chlonne was present in the inner region in fairly high quantities, whereas the outer (dull) region contained no chlorine Exact binding- energy measurements have enabled the chlorine to be dis- tinguished as a chloride ion from a highly ionic salt such as KCl or NaCl The distribution of species across the failure region has been further analysed using spatially resolved techniques and macro line scans have been produced both by XPS and TOF-SIMS The mobility of species can be assessed from these line scans and they allow the structure suggested in Fig 17 to be proposed On exposure, water permeates into the adhesive and diffuses towards the centre of the bond This process is controlled by surface diffusion or diffusion through a modified outer region within the epoxy The actual cause of delamination is still a matter of discussion with several routes being suggested Cathodic disbondment due to the presence of electrochemical potentials and a high pH at the crack tip will result in bond breaking and subsequent failure of the adhesive, this is ident- 490 J Muter Chem , 1996, 6(3), 479-493 ified as zone 1 Once disbonded, the epoxy displaces from the metal at a slower rate to the disbondment (bond breaking) to produce a crevice (zone 2) The disbondment front (just breaks the bonds) occurs at 18 5 pm per day to yield zone 1 whilst the slower delamination front (where the epoxy is physically moved from the metal surface allowing a crevice to form) proceeds at 10 2 pm per day in water at 30 "C which gives the visually identifiable zone 2 The presence of cations at the interface (detected by XPS and SIMS) is known to result in an increased failure rate due to solvated ions moving through the polymer allowing water ready access to the region3' The rate of failure is also dependent upon the cation present with delamination rates being higher for solutions containing potassium ions than lithium, sodium or barium electrolytes of equivalent concen- trations in the order of K, Na, Li and Ba A rapidly solvated ion will allow water to move quickly behind it, whereas a much slower ion (Mg2', charge/radius =27 78) will hinder the movement of water, resulting in slower failure lo In this case small mobile ions are present, therefore water will be able to move through the adhesive at a higher rate This will result in increased electrochemical attack and subsequently faster bond breaking Before separation, the delaminated surfaces merely remain in contact without displacing from one another This results in a very low volume debond (zero volume debond') At a critical point, the delaminated faces separate, allowing bulk solution to fill the crevice thus formed This increased volume allows post-failure oxide growth as well as ionic mobility Oxide growth is evident from both the survey spectra taken from the outer and inner regions and is demonstrated in the Fe 2p,,, E-X scan which shows the metallic component reducing in intensity towards the outside of the joint Increased ionic mobility is apparent from the chlorine signal which shows chloride ions being removed from the crevice region towards the more anodic outer surfaces Towards the centre of the bond (zero volume debond) the surfaces have not separated, thus ions move very slowly in this region The chlorine signal is much higher for this central region indicating that chloride ions are prevented (physically) from reaching the outer anodic surfaces Conclusions A thin yttrium hydroxide layer proves tenacious during expo- sure to water for 7500 h at 30°C However, this pretreatment alone does not improve durability After 1200 h, failure can be attributed to two clearly defined regions A central cohesively failed region and an outer appar- ently interfacial region There is no evidence of epoxy fragments on the metal surface, however, nitrogen is apparent (by XPS) on both failure surfaces The source of the nitrogen is not clear and three possible scenarios exist to account for its presence on the failed metal surfaces It is either present as a discontinu- ous layer resulting from failure within an enriched amine-rich surface phase within the adhesive, from leaching and sub- sequent re-deposition of nitrogen-rich phases from the adhesive or from adsorption of atmospheric nitrogen on failure After 7500 h in water at 30°C there is a ringed structure within the outer metal region This appears dull towards the joint edge and shiny towards the centre of the joint (crack tip) XPS confirms that a thicker oxide has formed in the outer region The inner zone corresponds to the interfacial failure (I e cathodic delamination) of the 1200 h joint Small-area XPS shows the presence of chlorine (as chloride ions) within the inner shiny region No chlorine is detected in the outer region It is proposed that although the entire failure surface has disbonded up to the crack tip, the delamination front (when a physical crack of a given volume is formed) has only opened up to the transition between the two regions It 50.00t I 20.::10.00 0.0040 2 4 6 8 101214 2 4 6 8 10 12 4 4.00d i 4 6 8 ioiirl 0.00 2 4 6 8 io i2 114 I distance from joint edge/mm Fig.15 TOF-SIMS line scans (metal side) of a failed lap shear joint after 7500 h in water at 30 "C: x-axis=distance from joint edge (in mm); y-axis=range of intensities for species present (in counts).m/z=(a) 56' (iron-56), (b) 28+ (silicon), (c) 23+ (sodium), (d) 39+ (potassium), (e) 27+ (aluminium/C,H,+/CHN+), (f) 40' (calcium), (8) 17-(OH), (h) 35-(chlorine-35), (i) 26- (CN-). J. Muter. Chem., 1996, 6(3), 479-493 491 5.00. 4.00. 3.00. 2.00. 1.00. 4 distance from joint edgelmm Fig. 16 TOF-SIMS line scans (metal side) of a failed lap shear joint after 7500 h in water at 30 "C: x-axis =distance from joint edge (in mm); y-axis=range of intensities for species present (in counts). m/z=(a) 73+, (b) 77+, (c) 98+, (d) 105+, (e) 107+, (f) 207'.Fig. 17 Schematic of failure within the lap shear joint is therefore proposed that two failure fronts exist in this system. The faster disbondment front operating at 18.5 pm day-' and the slower at 10.2 pm day-'. The data from XPS and SIMS analyses have been processed to produce line scans across the joint overlap. These plots 492 J. Mater. Chem., 1996,6(3), 479-493 show the spatial distribution of species present within the failure regimes. The presence of cations (sodium and calcium) at the crack tip indicate electrochemical activity whilst the absence of chloride ions in the outer region suggests that this region has better mobility, thus allowing anions to migrate towards the anode. The presence of chloride ions in the cathodic region near the crack tip confirms that this region is of much lower volume, thus preventing easy ionic movement. There is a high silicon signal in the outer region; both XPS binding energy measurements and SIMS analyses of this outer region enable this silicon to be identified not as a silicone but as silicate from the glassware.It is likely that contamination has occurred post-failure after the outer region has opened up, allowing movement of species from the bulk solution to occur. Electrochemical activity detectable from the movement of various ions indicates that the adhesive joint is under attack from cathodic delamination. However, without an applied potential, the pH achieved at the crack tip will not be sufficient to cause oxide reduction.A schematic of the failure region as a result of electrochemical activity prior to mechanical separation is proposed (Fig. 17). This work has shown the benefits of using a range of small- area and imaging spectroscopies including SAXPS, iXPS and TOF-SIMS in a complementary manner to study a complex system. S.J.D. gratefully acknowledges the financial support of both BP Research and the EPSRC for funding this CASE award. Special thanks go to Dr. Graham Beamson (RUSTI), Dr. Len Hazell (BP Research), David Bond and to Andy Brown for their invaluable assistance. References 1 J. F. Watts and J. E. Castle, J. Mater. Sci., 1983, 18,2987. 2 J. F. Watts, Su$. Interface Anal., 1988,12,497. 3 L. P. Haak, M. A. DeBolt, S. L.Kaberline, J. E. de Vries and R. A. Dickie, Surf. Interface Anal., 1993,20, 115. 4 A. M. Taylor, J. F. Watts, J. Bromley-Barratt and G. Beamson, Surf Interface Anal., 1994,21,697. 5 C. Kerr, N. C. Macdonald and S. Orman, Br. Polym. J., 1970,2,67. 6 A. J. Kinloch, Adhesion and Adhesives: Science and Technology, Chapman and Hall, London, 1987, pp. 363-366. 7 J. F. Watts, J. E. Castle and T. J. Hall, J. Mater. Sci. Lett., 1988, 7,176. 8 E. L. Koehler, Corrosion, 1984,40, 5. 9 U. R. Evans, Trans. Electrochem. SOC., 1929,55,243. 10 J. F. Watts, J. Adhesion, 1989, 31, 73. 11 J. M. Atkinson, R. D. Granata, H. Leidheiser Jr and D. G. McBride, IBM J. Res. Dev., 1985,29, 27. 12 J. E. Castle, 53rd International Meeting on Physical Chemistry, Paris, 1995, American Physical Society, in press.13 R. A. Cayless and D. L. Perry, J. Adhesion, 1988,26, 113. 14 R. A. Cayless and L. B. Hazell, Eur. Pat., EP 0331 284 Al, 1989. 15 S. J. Davis, J. F. Watts and L. B. Hazell, Surf Interface Anal., 1994, 21,460. 16 U. Gelius, C. G. Johansson, J. Larsson, P. Munger and G. Vegerfors,J. Electron Spectr., 1990,52, 747. 17 G. Beamson, D. Briggs, S. F. Davies, I. W. Fletcher, D. T. Clark, J. Howard, U. Gelius, B. Wannberg and P. Balzer, Surf. Interface Anal., 1990,15, 541. 18 A. M. Taylor, PhD Thesis, University of Surrey, 1994. 19 D. Briggs and M. J. Hearn, Vacuum, 1986,36,105. 20 R. A. Dickie and A. G. Smith, Chemtech, 1980,10,31. 21 M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, Pergamon Press, Oxford, 1966, pp. 177-182. 22 R. A. Gledhill, A. J. Kinloch and S. J. Shaw, J. Adhesion, 1980,11,3. 23 S. J. Davis and J. F. Watts, unpublished results. 24 R. A. Gledhill and A. J. Kinloch, J. Adhesion, 1974,6, 3 15. 25 J. E. Castle, R. Ke and J. F. Watts, Corrosion Sci., 1990, 30, 771. 26 J. F. Watts, R. A. Blunden and T. J. Hall, SurJ Interface Anal., 1990, 16, 227. 27 C. C. Telford, S. J. Davis and J. F. Watts, unpublished results. 28 Handbook of X-ray Photoelectron Spectroscopy, ed. Jill Chastain, Perkin Elmer Corporation, Minnesota, USA, 1992, pp. 62-63. 29 G. Beamson, personal communication, RUSTI facility, Daresbury Laboratory, 1995. 30 J. F. Watts, J. E. Castle, P. J. Mills and S. A. Heinrich, in Corrosion Protection by Organic Coatings, ed. M. W. Kendig and H. Leidheiser Jr, The Electrochemical Society, 1987, pp. 68-83. 31 R. D. Shannon and C. T. Prewitt, Acta Crystallogr., Sect. B, 1970, 26, 1046. Paper 5/05897J; Received 6th September, 1995 J. Mater. Chem., 1996, 6(3), 479-493 493