Fracture of epoxy bonded dynamic peel specimens containing interfacial layers

Interfacial layers have been detected in bonds formed between hot dip galvanised steel substrates and heat cured epoxy adhesive. The layers have a number of features which affect the impact performance of the bonds formed compared with the equivalent bonds with mild steel substrates. The mechanical properties of the interfacial layers have been found to be significantly different to that of the bulk adhesive. Analysis of the layers showed an absence of silicate filler particles within the layers. Bubbles were also found at the surface of the hot dip galvanised substrates as a byproduct of a chemical reaction, which may occur between the hardener in the adhesive and the substrate surface. Pyrolysis experiments showed that a gas was produced when zinc–dicyandiamide mixtures are heated at the adhesive cure temperature. The production of interfacial layers has also been found to be strongly dependent on subtle changes in the surface chemistry or structure of the substrates.


Introduction
Emission reductions and energy conservation issues are currently creating a need for lighter vehicle structures, while also maintaining strength.! Weight reductions can be partly accomplished using thinner gauge zinc coated steel sheet materials in the fabrication of body shells.Adhesive bonding is one area where significant improvements can be made by reducing the stress concentrations associated with traditional spot welds, thus allowing the use of thinner gauge sheets.Although some questions remain as to the suitability of adhesive bonding with some adhesive/substrate combinations, the epoxy-steel bond has now been shown to be as strong and as impact resistant as its spot welded counterpart.
Previous work by the authors examining fracture surfaces and sections through epoxy adhesive steel to steel bonds, containing metallic wire spaces to maintain the adhesive bond thickness, has shown that brittle interfaces may exist between the bulk adhesive and the spacers.A fracture surface from the location of such a wire spacer is shown in Fig. 1 and can be seen to be generally ductile, but has a brittle layer immediately adjacent to the track where the wire was located.This type of interface was found to occur predominantly around copper, zinc, and galvanised wires.The possibility of similar interfaces forming between zinc coated substrates and epoxy adhesives was therefore investigated.
Crompton 2 has previously observed the potential for layer formation in bonds between epoxy adhesive and aluminium substrates.He found a transition region between the substrate and the bulk adhesive in which the physical properties differed from the bulk adhesive.Knollman and Hartog 3 have found similar effects.More recent work by the authors of the present paper 4 has demonstrated the existence of interfacial layers between some structural epoxy adhesives and certain zinc coated steel substrates.These layers have been primarily found in some silicate filled epoxy adhesives in combination with hot dip galvanised substrates.These material combinations form a complex layered structure in which defects are fOJmed which have been shown to produce bonds with significantly reduced strength. 4The use of these materials in automotive structures is however desirable because of the enhanced corrosion resistance of zinc coated steel and the generally enhanced mechanical properties of epoxy bonds.The use of dicyandiamide (dicy) as a latent hardener is common in single part heat-cured epoxy adhesives.A number of studies 5 ,6 have shown that the dicy will react with metallic zinc when heated to a temperature similar to those used to cure the adhesives.Kinzler et al. 7 suggested that the products of such a reaction may migrate towards the bulk adhesive to create an interfacial layer adjacent to the zinc substrate.Gaillard et al. 8 ,9 have used grazing angle Fourier transform infrared (FTIR) microscopy to show that a reaction occurs between dicy and galvanised steel at i80 a C.
Pao et al. 10 have shown that in single lap shear joints the stress intensity factors and fracture parameters of interfacial cracks are strongly influenced by the ratio of the Young's modulus of the interfacial region to that of the bulk adhesive.Beevers and Bowditch ll have shown that the long term performance of bonds is strongly influenced by the rate at which water absorption occurs in epoxy adhesives.This has been shown to be most significant in the interfacial region of a bond and will be strongly influenced by the presence of defects in this region.A similar conclusion was reached by Dickie et al.,12 again highlighting the importance of the interfacial regions in an adhesive bond.

EXAMINATION OF CROSS-SECTIONS
Multiple bonds were prepared for sectioning and microscopy using various commercial grades of hot dipped zinc coated steel sheet.These materials have different spangle sizes (the grain size of the zinc), yield strengths, and thicknesses, some with and some without a chromate finish.The adhesive used was a high strength, single part, heat curing, calcium silicate filled epoxy.The adhesive used throughout this work is known to exhibit the formation of interfacial layers when used in conjunction with hot dip galvanised substrates.All the substrate surfaces were prepared by abrading with fine emery paper and de greasing with alcohol prior to assembly.The bonds were then cured according to the adhesive manufacturers' instructions at i80 a C. In order to determine whether interfacial regions existed between the substrate and the adhesive, sections were cut across bonds.After sectioning the specimens were mounted in a cold curing resin before being ground and polished according to a proprietary method. 13The polished specimens were examined using optical and scanning electron microscopy (SEM) and energy dispersive X-ray (EDX) analysis.Once areas of interfacial layer had been 1 Wire track with brittle sheath adjacent to wire identified, the Vickers hardness number for layers and bulk adhesive were determined using a Reichert microhardness tester.
tube furnace control thermocouple Argon gas 2 Pyrolysis oven used to study gas evolution from zincdicyandiamide mixtures of dicyandiamide and zinc were placed in a boat crucible and put in the middle of an air condenser tube which was surrounded by a muffle furnace set at a temperature of 180 a C for half an hour (see Fig. 2).The whole system was under an argon atmosphere throughout the experiment.
Heating was switched off after half an hour but argon gas was allowed to flow continuously for another hour in order to collect all the gas produced by the reaction.Any gas produced was bubbled through distilled water containing alcoholic phenolphthalein, which was used later for titration.

Results and discussion
Table 1 Indicence of filler barren layers in bonds with various finish galvanised steel sheet Table 1 shows some of the results from the optical microscopy that indicates the extent of the filler barren layers in the bonds for each of the different substrates.The phenomenon was found to be most prominent in bonds formed with the minimised spangle, chromate finish hot dip galvanised substrate and no layers were observed in any of the mild steel substrates.At its most distinct the layer extended at various thicknesses up to 30 Jlm and in some cases was observed along almost the whole length of a bond.A typical example of a well defined layer as shown in Fig. 3.This shows a layer adjacent to the zinc coating which is different in appearance to that of the bulk adhesive.The bulk adhesive can be seen to contain filler particles of various sizes, which are not apparent in the interface.The microhardness of the bulk adhesive was found to be 1•8SHV(S g) ± 0•11 compared with 4•S9HV(S g) ± 0'4S for the interfacial layer.This showed that the interfacial layer is significantly harder than the bulk adhesive despite the presence of the much harder silicate fillers in the bulk adhesive.The absence of calcium silicate filler in the interfacial layer was confirmed by carrying out elemental X-ray analysis shown in Fig. 4 and also by producing elemental distribution maps for silicon and calcium.These clearly showed the absence of filler in the interfacial layer compared to the high levels of both elements found in the bulk adhesive.The samples were coated with gold-palladium to prevent excessive charging in the SEM while still allowing EDX analysis.
Load-time curves from typical mild steel and hot dip galvanised impact peel test results are shown in Fig.

IMPACT PEEL TESTS
The dynamic performance of adhesively bonded joints and structures is obviously of primary importance in automotive applications.In order to assess the effect of the interfacial layers on the strength of metal to metal bonds, impact peel tests were conducted on the galvanised steel substrates and also on uncoated mild steel samples.In order to ensure that the results from the two sets of samples would be directly comparable, the mild steel substrates were prepared by removing the zinc surface layer from the respective hot dipped galvanised sheet using SO%HC!.All samples were again prepared in parallel by abrading with fine emery paper and degreasing with alcohol before assembly and cure.The impact peel tests were conducted using an instrumented falling weight drop tower, forces being measured using an instrumented tup travelling with a velocity of 2 m s-1, and stationary wedge in accordance with ISO 11343.
Force-time measurements obtained during the tests were converted into force-displacement data by numerical integration in order to allow the calculation of total energy consumed for each specimen.The test results from the impact peel specimens were compared for the samples made from both uncoated mild steel and hot dip galvanised substrates in combination with a high strength heat curing epoxy adhesive.The energies for each substrate type (hot dip galvanised or mild steel) were averaged and used to compare the performance of each hot dip galvanised substrate with its mild steel counterpart.The energies from the different hot dip galvanised substrates could not be directly compared owing to the nature of the impact peel test and its dependence upon the substrate thickness and yield strength.A T-test was subsequently performed on each hot dip galvanisedjmild steel pair of impact energies, to ensure that the difference observed was significant.The resulting fracture surfaces were examined using SEM.The same range of metal substrates and adhesive were used to make both mechanical test specimens and the samples which were cross-sectioned for optical microscopy and SEM.

CHEMICAL REACTION AND PYROLYSIS OVEN
A chemical reaction pyrolysis oven (shown in Fig. 2) was used to study gas evolution from reactions between dicyandiamide and zinc.One to one molar ratio samples propagate rapidly at an almost constant load.The total fracture energy in this case was found to be approximately 3 J.The substrates from the mild steel impact peel specimens suffered from considerable bending.The galvanised steel specimen substrates showed much less deformation, indicating that less energy has been absorbed than in the case of their mild steel counterparts.
Table 2 shows the average energies absorbed to fracture from both mild steel and hot dip galvanised specimens in each of the four series examined.These generally show increased energies in the mild steel specimens in all cases examined.
The fracture surfaces from the peel tests were examined using an SEM.The initial results showed a generally flatter fracture surface in the mild steel specimens with more crack branching in the hot dip galvanised specimens.There is also more evidence of brittle fracture surfaces in the hot dip galvanised specimens.All fractures are within the adhesive and many contain bubbles or cavities, which in the mild steel specimens, are larger and fewer in number.In the hot dip galvanised specimens the bubbles are smaller in size and greater in number.
Typical fracture surfaces resulting from specimens of the impact peel tests are shown in Figs. 6 and  mild steel specimen (Fig. 5a) rises sharply, and then drops and stays approximately constant for some time followed by a sequence of less sharp rises in load until the specimen fractures completely.The load peaks occur due to initiation of a fracture, which then arrests after propagating rapidly at a lower load in each case.The fracture surface exhibits stress whitening and more severe bending in the substrate at each point of crack initiation, corresponding to an increase in load.The total fracture energy of this specimen was found to be approximately 9 J.
Figure 5b shows a sample load trace for a galvanised specimen, which indicates a reduced fracture load with increasing propagation.
In this case there is little increase in load during propagation and the crack appears to  steel and hot dip galvanised substrates respectively.The fractures were found to be within the adhesive in all cases, being approximately in the centre of the adhesive layer in the mild steel bonds and close to the interface with the substrate in the galvanised steel specimens.The fracture surface in the mild steel specimens was found to be generally flat at the scale seen in Fig. 6, with a roughness evident at higher magnifications indicative of ductile tearing.Fine, almost spherical bubbles dominate the fracture surfaces from the galvanised substrates.These bubbles lie in the plane of the fracture, which lies close to the interface with the substrates and can be seen in Fig. 7. Similar bubbles are absent in the specimens with a mild steel substrate, although larger, irregularly shaped bubbles were often found on both types of substrate which were attributed to moisture.The bubbles in the hot dip galvanised samples suggest a reaction occurring between the adhesive as it cures and the zinc coating of the galvanised steel.
A more detailed examination of the fracture surfaces from the hot dip galvanised substrates revealed small brittle patches close to the substrate surface but within the adhesive.These brittle patches were always located in close proximity to a spherical bubble as shown in Fig. 8. EDX analysis of these brittle surfaces showed that there was no silicate filler contained in the adhesive at these locations.
The pyrolysis tests, in which specimens of the dicyandiamide hardener were heated with zinc metal powders in an inert atmosphere, showed an exothermic reaction when the temperature reached approximately 180 a C accompanied by a rapid gas evolution.The gas produced was shown to contain ammonia and provides a mechanism for the 7 Fracture surface from impact peel specimen with hot dip galvanised substrate 8 Brittle fracture region at edge of a bubble in fracture surface formation of gas bubbles observed in the hot dip galvanised fracture surfaces.Bubble formation was suppressed in one hot dip galvanised impact peel sample, by curing the bond at a pressure of 2 bar.The energy absorbed in this case was also reduced but only to 6 J, which although higher than that without pressure, is still much lower than the respective mild steel substrate.Examination of the fracture surface confirmed that there were no bubbles present in this case, but unlike the mild steel substrates, the fracture surface was heavily branched with cracks running off the main fracture locus towards the interface with the substrate.Small regions of brittle fracture surface were observed very close to the substrate surface, but within the adhesive.An example of a brittle region is shown in Fig. 9, which appears to indicate that the fracture is strongly influenced by brittle crack formation within the adhesive close to the interface with the metal substrate.An EDX analysis of the brittle layer shown in Fig. 9 indicated an absence of silicate filler particles in the adhesive within the layer.

Conclusions
1. Filler barren layers have been found to exist at the interface between hot dipped galvanised steel substrates and a high performance, calcium silicate filled epoxy adhesive.
9 Fracture surface of pressure cured epoxy on hot dip galvanised substrate Materials Science and Technology 2. The filler barren layers were found to be significantly harder than the bulk adhesive.
3. Impact peel tests on the galvanised steel specimens showed a significant reduction in energy to fracture.
4. Fracture surfaces resulting from the impact peel tests indicated a more brittle failure in the case of the galvanised steel substrates.
5. Examination of the fracture surfaces from the galvanised steel specimens revealed brittle regions in which the adhesive was barren of silicate fillers.
6.The increased formation of bubbles at the interface of the galvanised substrates did not completely explain the reduction in energy seen in the tests.It was concluded that the reduction in energy was in part due to the formation of interfacial layers.
S.From Fig.Sit can be seen that the fracture load for the

Finish5
on galvanised steel substrates Normal spangle, chromate (970) Minimised spangle, chromate (1000) Smooth finish, non-chromate (1225) Minimised spangle, non-chromate (1300) Layer appearance Some layers found A lot of layers found (up to 20 Jlm thick) A few layers found (up to 5 Jlm thick) A few layers found (about 1-3 Jlm thick) Load versus time curves for a mild steel and b galvanised steel adhesive substrates

3
Optical micrograph of interfacial layer between bulk adhesive (top) and galvanised steel

4
Energy dispersive X-ray analysis of a interfacial layer showing little silicon or calcium and b bulk adhesive showing presence of both silicon and calcium

Table 2
Total impact peel energies in Joules for the range of substrates tested