Phase Separation Influence on the Performance of CTBN-Toughened Epoxy Adhesives.
Epoxy adhesives are currently used in the airplane industry because of their overall performance characteristics. They can be formulated to perform adequately in extreme environments which include solvents, temperature extremes, and ultraviolet radiation and are relatively easy to process and apply. They also are easily reacted so that a wide range of curatives and catalysts can be used to initiate cure and the formation of a crosslinked three dimensional infinite molecular weight structure (1, 2). For higher end-use temperature epoxy systems, the crosslink density must be sufficiently high to prevent thermally initiated phenomena including creep, stress relaxation, or even degradation (1, 3-6). As a result, the cured epoxy matrices alone are somewhat brittle and prone to flaw or crack propagation. Toughening agents, including liquid carboxyl-terminated butadiene acrylonitrile (CTBN) elastomers, are frequently added to the uncured base epoxy formulations to improve fracture characteristics.
CTBN materials primarily toughen epoxy matrices through two separate mechanisms. Rubber-rich particles are created during the cure of the material and these particles work to moderate the stress field when loading occurs and induce localized yielding, crazing, and cavitation, three mechanisms which significantly increase the fracture energy properties of a material (7). In addition to these mechanisms, some CTBN material can remain in the epoxy-rich phase if phase separation is incomplete. This material acts in a flexibilizing role by reducing the crosslink density of the first phase. This, however, reduces the service temperature capabilities of the overall material system since the glass transition temperature decreases with increasing flexibilized rubber content. The role of flexibilized toughening versus dispersed particle size toughening for layered epoxy structures is not well understood from this material standpoint.
Limited information has been generated concerning the role of toughening additives which do not phase separate during the cure process. Several studies have been performed analyzing the morphology and phase compositions of epoxy systems toughened with either CTBN or thermoplastic materials (8-11). Hourston et al. investigated the polyetherimide (PEI) content in an epoxy matrix and observed that increasing PEI content led to more flexibilizing PEI in the continuous phase as determined by dynamic mechanical analysis. They stated this occurrence as a possible reason why mode I fracture energy results did not continually improve as toughener content was increased. Chen et al. observed the behavior of epoxy systems toughened with different molecular weight rubber tougheners. They found that an increased molecular weight led to more complete phase segregation, an increase in the size of the dispersed phases, and an overall decrease in the toughening properties. They indicated that a higher viscosity in the reacting medium produced smaller toughening particles. Lastly, Bucknall and Partridge performed tests on systems containing polyethersulphone, triglycidyl p-aminophenol, tetraglycidyl diaminodiphenylmethane, and diaminodiphenylsulphone. They observed that phase separation was suppressed for the polyethersulphone at tetrafunctional contents greater than 80 parts per hundred resin (phr). Most of the previously mentioned work has indicated that the phase separation process can be influenced by relative changes in chemistry, viscosity, etc. However, the work has not specifically focused on this influence on end-use properties, rather investigating morphological characteristics.
The purpose of this study is to investigate the bonding characteristics of CTBN toughened epoxy systems when used with polymeric composite materials. Specifically, this study will investigate the role of flexibilized rubber on the fracture energy and bonding behavior with both cocured and bonded composite materials to understand the relative importance of the toughening mechanisms. Collectively, this work will help provide a better understanding of the importance of the rubber phase separation process for liquid CTBN toughened epoxy adhesives when used with polymeric composite systems.
The epoxy materials utilized in the study consisted of a combination of difunctional DGEBA type resins and a tetrafunctional TGMDA resin. Epon[R] 836 (Epoxy Equivalent Weight = 315), 1001F (EEW = 535), and 828 (EEW = 190) were utilized for the difunctional systems with Araldite[R] MY9512 (EEW = 107) as the tetrafunctional system. Dicyandiamide was employed as the curing agent with diuron as a cure accelerator. Hyar[R] CTBN 1300 x 13 and CTBN 1300 x 8, both liquid rubber modifiers, were employed as tougheners. Finally, hydrophobic fumed silica (Aerosil[R] R202) was incorporated to increase the viscosity of certain systems and reduce the dispersed phase growth rates.
Table 1. Material Composition of Adhesive Resin Systems. Component Weight % Epon[R] 828 9 Epon[R] 836 13 Epon[R] 1001F 43 Araldite[R] MY9512 22 Dicyandiamide 6 Diuron 2.6 CTBN toughener 5/18 phr
A list of the composition for the various resins is shown in Table 1. For the toughened resins, the CTBN material was adducted with a difunctional epoxy for later incorporation into the resin. The adducted mixture was combined and mixed at 150 [degrees] C under a nitrogen atmosphere for one-half hour at which point triphenylphosphine was added to catalyze the reaction.
The completed resins were filmed on an industrial size prepregging machine using a hot plate temperature of 80 [degrees] C (12). Non-woven polyester scrim material, Reemay[R] 2250, was incorporated into the resin film for flow control and controlled bondline thickness. Final adhesive thicknesses of approximately 0.14 mm were obtained.
The different resin formulations were created by manipulating three variables including the silica content, CTBN toughener type, and toughener level. For clarity, the formulations will be referred to by a numbering system where the corresponding differences among resin systems are listed in Table 2.
Glass Transition Information
Glass transition information for the model adhesive systems was obtained using dynamic mechanical analysis (DMA). A TA Instruments 983 DMA module, interfaced to a Thermal Analyst 2000 Controller, was utilized. Samples were equilibrated at -90 [degrees] C for a period of five minutes after which time a 5 [degrees] C per minute heating ramp was applied. Two samples, obtained from bulk unfilmed material, were tested for each adhesive and the glass transition temperatures were recorded as the peak in the tan [Delta] response. Sample dimensions were approximately 8 x 3.5 x 1.5 mm and a frequency of 1.0 Hz was used with an oscillation amplitude of 0.2 mm.
Fracture specimens were prepared using an epoxy-carbon fiber 3k-70-PW material. Mode I and mode II fracture energies were tested using the double cantilever (DCB) and end-notch flexure (ENF) methods, respectively. Sample panels were constructed using eighteen plies of prepreg with one layer of adhesive in the midplane. A FEP (fluorinated ethylene propylene) film was placed in the middle to act as a crack starter. Both cocured and bonded specimens were tested for all adhesive systems. Bonded prepreg panels were cured at 177 [degrees] C and 585 kPa (85 psi) for two hours with one layer of epoxy-fiberglass peel-ply used to create a fresh surface prior to the bonding step. Before the bonding, the fiberglass layer was peeled off exposing the epoxy-carbon fiber surface. The surface was solvent wiped before film adhesive application. The panels, measuring 33 x 8.25 x 0.43 cm, were cured at 177 [degrees] C for two hours using heat-up and cooling rates of 2.8 [degrees] C per minute and a compaction pressure of 310 kPa (45 psi).
Table 2. Classification of Model Adhesive Resin Systems. Resin Toughener Toughener Silica System Type Level (phr) Addition 1 1300 x 8 5 N 2 1300 x 8 5 Y 3 1300 x 8 20 N 4 1300 x 8 20 Y 5 1300 x 13 5 N 6 1300 x 13 5 Y 7 1300 x 13 20 N 8 1300 x 13 220 Y 9 none n/a Y 10 none n/a N
DCB specimens were tested in a tensile loading frame with a crosshead speed of 2.54 cm per minute until a final displacement of 6.35 cm was encountered (13, 14). The energy required to propagate the resulting crack was recorded along with the sample width (approximately 1.27 cm) and crack length increase. Four samples were tested for each adhesive to give the reported value. All samples were precracked in the apparatus before actual testing to provide an initial sharp crack which was common to all specimens.
ENF specimens were tested in the same apparatus using a crosshead speed of 0.254 cm per minute (13, 14). The crack tip, after the precrack step, was positioned using a 2.5 x magnifying apparatus to be one inch from both the support and loading pins. The sample was loaded in a three point bending mode until crack propagation occurred. Specimens were then repositioned so that the crack tip was again at the required setting. One sample was tested for each adhesive from which five to six values were obtained and averaged.
Optical microscopy was used to investigate the morphology of the cured resins using both the DMA samples (bulk material) and fracture specimens. Both transmitted light (thin-section microscopy) and reflected light specimens were analyzed. When required, samples were placed in an Os[O.sub.4] solution (2.5 wt% with tert-butanol) for approximately eighteen hours to stain the CTBN unsaturated regions. An imaging analysis software program was used to calculate representative [TABULAR DATA FOR TABLE 3 OMITTED] rubber-rich dispersed phase volume fractions by using a binary imaging contrasting technique and recording characteristic pixel values (15). Scanning electron microscopy (SEM) was also utilized for morphological analysis with an accelerating voltage of 15 kV and a working distance of 20 mm.
Rheological properties including minimum viscosity and gel point temperatures were obtained to characterize the different adhesive systems. A Carri-med CSL 1000 Rheometer with an Extended Temperature Module (ETM) was used for the analysis with a parallel plate assembly, a 500 [[micro]meter] gap, a torque of 100 [Mu]N*m, an angular frequency of 6.28 rad/s, and a temperature ramp of 3.5 [degrees] C per minute.
Lap Shear Testing
Failure stress values were obtained using wide-area aluminum lap shear specimens. Panels were fabricated using 0.163 cm (0.064 in) thick aluminum which had been phosphoric acid anodized (PAA) and coated with a standard chromated primer. Adhesive layers were applied and the panels were cured at 177 [degrees] C for two hours using a compaction pressure of 310 kPa. The panels were machined and tested in a tensile loading frame using a crosshead speed of 0.127 cm per minute.
RESULTS AND DISCUSSION
Bulk adhesive viscosity profiles for the different formulations were obtained initially to determine the increase in flow viscosity resulting from the hydrophobic fumed silica and rubber toughening addition. Results for the different model systems appear in Table 3. Resin system 5 was not included in the rheological analysis.
The resin systems displayed expected characteristics for thermosetting polymeric materials. The resin sample from each pair (e.g. 1 and 2) that included fumed silica showed a viscosity increase on the order of 10-40 times due to the thixotropic behavior introduced by the colloidal silica particles. Resin 9, an untoughened formulation with silica, showed a unique reproducible behavior where the minimum viscosity temperature actually was significantly below the reaction initiation temperature. This most likely was a [TABULAR DATA FOR TABLE 4 OMITTED] consequence of the thixotropic behavior imparted by the silica particles/aggregates coupled with the increased chain mobility resulting from the thermal energy increase.
DMA specimens were tested for the various formulations using neat resin plaques cast at 177 [degrees] C using a 2.8 [degrees] C per minute heat-up rate. Most of the CTBN toughened epoxy subambient tests showed both low and high temperature transitions, characteristic of two distinct phases present in the materials. Resin systems were characterized for their transition temperature(s) and the results are displayed in Tab/e 4.
Photomicrographs of the toughened epoxy systems appear in Figure I and illustrate the different morphologies observed for the bulk resin systems. The changes in the toughener type, toughener amount, and silica addition had a strong influence on the observed morphologies. A wide range of microstructures was possible by manipulation of these variables. Resin system 6 did not show any phase separated material under transmitted or reflected visible light. This sample was also tested using SEM and a distinct second phase was not present. The photographic result for sample 8 was manipulated using computer software to more clearly illustrate the phase differentiation, however, the dispersed phase area or volume percentage was not affected by this procedure.
Image analysis of the photomicrographs was utilized with dynamic mechanical analysis results to calculate volume fractions of rubber and epoxy present in the different phases. The Gordon-Taylor relationship was used for the compositional analysis (16). This relationship equates the change in the glass transition temperature [Delta][T.sub.g,E] for the epoxy-rich phase with the mass fraction of rubber in this phase, [[W.sub.r].sup.c], such that
[Mathematical Expression Omitted] (1)
where [T.sub.g.E(r)] is the observed glass transition temperature for the epoxy-rich phase, [T.sub.g,E(0)] and [T.sub.g,R(0)] are the glass transition temperatures of the neat epoxy and rubber, and k can either be taken as a fitting parameter or a theoretical value related to the ratio of the thermal expansion jump of the materials at their respective glass transitions. For the present study, the k value was used as a fitting parameter since resin system 6 was completely flexibilized by the CTBN addition. This allowed a direct calculation of the k parameter which was then used for all subsequent systems containing this toughener. The k value for the 1300 x 8 CTBN systems was scaled from its theoretical value to give a result that was consistent with the ratio of the actual and theoretical values for the 1300 x 13 CTBN toughener. The volume fraction of rubber in the continuous phase was then calculated and with the aid of a mass balance the dispersed phase composition found. The results of this compositional analysis along with the average dispersed phase particle diameters appear in Table 5.
The untoughened systems' responses showed that the hydrophobic fumed silica addition did not affect the chemistry of the resin as far as curing behavior ([T.sub.g]) was observed. This was confirmed with differential scanning calorimetry (DSC) where identical heats of reaction were obtained for the two untoughened systems. The untoughened systems both possessed a subambient temperature transition, relatively faint and broad in nature, that has been attributed to a transition characteristic of the [-C[H.sub.2]CH(OH)C[H.sub.2]O-] segment in the crosslinked resin (1).
The behavior of the resin pairs (i.e. with and without silica) was quite consistent for all formulations. The continuous phase (epoxy-rich) transition temperature was slightly higher for the versions without silica except for the first two model systems showing that the dispersed phase nucleation or growth rate during the cure reaction was decreased by the silica addition. This seems reasonable assuming that the mechanism by which the dispersed phase was produced occurred by nucleation and growth rather than spinodal decomposition which possesses different kinetics. The nucleation and growth mechanism has been supported for similar epoxy-dicyandiamide-CTBN systems (17). Additionally, the characteristic bicontinuous phase formation seen during the spinodal decomposition process was not observed during or after cure for resin samples placed on a heated stage.
Assuming that the nucleation rate was entirely dependent on interfacial tension between the rubber and epoxy molecules, it would seem more likely that the growth rate of the rubber-rich phase was reduced by the viscosity increase from the silica addition. This resulted in more rubber remaining in the continuous phase after gelation and a lower observed [T.sub.g]. This is also supported by the fact that if the nucleation rate was increased by the silica addition a reverse in the trend of the epoxy-rich phase [T.sub.g] values would have been expected, and this was not observed. If the nucleation rate was increased due to fumed silica addition, and the same percentage of particles reached a critical radius size, one would have expected more complete phase separation and lower rubber percentages in the continuous phase, occurrences which were not seen experimentally. The growth rate dependence on viscosity has been part of several models used to predict phase separation behavior for toughened epoxy systems and has shown that the diffusional ability of the rubber molecules to join phase separated domains is a direct function of the viscosity (17, 18).
Table 5. Phase Composition Results for Model Adhesive Systems as Determined by Dynamic Mechanical Analysis and Optical Microscopy. Resin Volume Fraction Volume Fraction Average System Rubber in Rubber in Spherical Particle Continuous Phase Dispersed Phase Size ([[micro]meter]) 1 0.021 0.81 1.0-4.0 2 0.028 0.66 0.5-1.5 3 0.028 0.82 10-20 and 2-4 4 0.061 0.61 10.0 (non-spherical) 5 0.039 0.81 0.5-1.0 6 0.061 0 - 7 0.069 0.97 0.5-1.2 8 0.11 0.87 0.4-1.2 9 0 0 - 10 0 0 -
Rubber Type Dependence
The rubber type influence on the morphological characteristics showed expected behavior. The more polar toughener (1300 x 13) was more compatible with the epoxy mixture meaning that phase separation did not occur as early in the curing process as compared to the less compatible toughener. Consequently, assuming similar growth rates, the domain sizes were reduced for the 1300 x 13 systems.
Regarding compatibility, it would be expected that more CTBN rubber be present in the continuous phase of the 1300 x 13 toughened systems as was seen (assuming that phase separation was not pure). The composition of the dispersed phases, however, did not seem to be as strong a function of the toughener type.
Rubber Level Dependence
The rubber level dependence on the adhesives' morphologies was apparent for both the 1300 x 8 and 1300 x 13 systems. An increase in the rubber concentration resulted in a larger volume fraction of dispersed particles, but the concentration of these domains remained virtually constant for each rubber type irrespective of the rubber loading. The exception to this was the low and high rubber levels for the silica resin prepared with 1300 x 13 CTBN toughener where phase separation was not observed for the CTBN 5 phr situation.
It was also observed for model adhesive system 3 that the increased rubber loading level resulted in a somewhat bimodal distribution of dispersed particles. There were particles on the order of 10-20 [[micro]meter] in diameter and another grouping around 2-4 [[micro]meter]. Two particle size systems have been reported before but usually for systems incorporating bisphenol A as a chain extender (19, 20). This dual particle sized system suggests that the nucleation rate for this formulation may be coupled to the growth rate or that the growth process occurs by an Ostwald ripening process where the smaller particles were "frozen" in place when the molecular mobility was substantially decreased at gelation. However, this behavior would also be expected for other model formulations, provided that the growth mechanism did not change, and was not observed. Based on additional diffusion considerations, it also seems more probable that the former occurred.
In summary, the hydrophobic fumed silica addition did significantly affect the morphology and subsequent composition of the respective phases for the toughened epoxy formulations. There appeared to be a balance that was realized between the growth and phase separation processes. Growth rates may be decreased for the rubber domains as the viscosity is increased, but if the domains are not pure elastomer then the phase separation behavior becomes more complex. The rubber may still completely phase separate ff the incompatibility level is large enough and the curing reaction rate is not too fast, but the particle size will be decreased over a similar resin without silica. This particle size difference between toughened systems is very important since ultimate performance properties have been shown to be a strong function of domain size. (20-22). The mobility of the systems during and after cure is also important since an increased mobility can result in lower residual interfacial stresses generated after cool-down from the cure temperature for layered composite structures (19, 23).
It should be mentioned before continuing that bulk versus thin film material behavior has been shown to be important in regards to certain materials both from a morphological and intrinsic property point of view (24, 25). For this study, bulk materials used for morphological and glass transition information were prepared using thicknesses to closely approximate materials used in the mechanical property test specimens. Microscopy showed that the different resin morphologies were consistent when compared to the thin film materials contained in the composite applications.
Lap Shear Analysis
Lap shear tests with aluminum adherends were performed to characterize the relative adhesion properties. The results of the testing appear in Figure 2 and show that the manipulated variables did have an influence on the failure stress. The failure modes for the samples are also indicated with A signifying adhesive failure, C cohesive failure, and M representing "mostly" such that MC would imply mostly cohesive failure. The samples seemed to show a direct dependence on the rubber content in the continuous phase and the dispersed phase size. The samples with smaller particles and increased flexibilized rubber content showed the best performance for lap shear testing with aluminum adherends. The failure stress results seemed to parallel the failure strain values except for the systems which contained larger particles (systems 3 and 4), suggesting that the increased particle size created large enough stress concentrations to initiate early failure. In addition, samples 7 and 8 showed stress-whitened regions on the fracture surface after testing, indicative of plastic deformation and cavitation. The flexibilized rubber content and small particle size ([approximately] 1 [[micro]meter]) in these systems was sufficient to allow for significant energy dissipation during the testing.
Cocured Fracture Results
Figures 3 and 4 show the fracture results for the specimens cocured with the carbon fiber-epoxy prepreg material. Samples were void-free and adhesive thicknesses were approximately 0.08-0.10 mm. The mode I results for the cocured systems showed somewhat different trends than the mode II results. In mode I, the rubber type played a more significant role in the performance and the fumed silica addition did not as drastically influence the results. Model systems 1, 3, and 4 were also the only systems to show predominantly cohesive failure during the testing. It appeared that the morphological changes were not as important for resin systems in this testing. This issue will be discussed in detail later.
The systems toughened with the more compatible CTBN toughener (1300 x 13) did not show an increased toughness over the untoughened systems. All 1300 x 13 systems showed interfacial failure while most of the 1300 x 8 toughened formulations showed bulk cohesive failure. Adhesive or interfacial failure would cause morphological effects in the bulk resin to be masked since failure would not occur in the bulk where the toughening particles were contained. Therefore, for systems that failed interfacially, this test might not be a true indication of the value of these toughening domains. It does, however, illustrate the importance in adhesive applications in so far that failure modes can be manipulated with slight changes in chemistry and formulating procedure (19).
The particle size also played a role in the mode I results with larger particle systems showing improved results. Two possible reasons for this are stress relief during cool down and crack tip blunting in mode I loading. Larger rubber particles create larger tangential compressive stresses during the cooling period which would increase the load required to initiate failure during testing. However, larger particles do not toughen the bulk matrix to the same degree as smaller particles (0.5-10 [[micro]meter]) provided the failure is cohesive (20-22). Limited shear yielding and cavitation are observed with larger particles and crack tip blunting often is the main mechanism for toughening means.
The silica dependence itself did not significantly affect the mode I fracture results for the cocured systems. A very slight increase in the mode I toughness was seen with silica addition for the untoughened resins, possibly a result of a crack pinning mechanism and a slight increase in crack path length (26). The silica dependence on the fracture toughness by modification of the bulk morphology was evident. The change in the particle size between otherwise identical resin systems was found to slightly influence the mode I fracture toughness. Increased rubber levels were also found to increase the mode I toughness properties as expected, provided that the increased rubber content lent itself to phase domain formation and not complete flexibilization.
The mode II fracture cocured samples showed a strong dependence on particle size effects and flexibilized rubber content. The toughest samples in mode II were the untoughened resins followed by those with the smallest dispersed phase inclusions in the continuous phase. Larger dispersed phase diameter to adhesive thickness ratios result in larger stress values in the load-bearing crosslinked epoxy continuous phase for this failure mode and therefore a lower stress was required to initiate failure.
Therefore, the cocured fracture results showed that the manipulation of the nucleation and growth mechanism by fumed silica addition did influence fracture results. The main mechanism by which this happened was reduced particle size as a consequence of retarded disperse phase growth rates. Mode I results indicated that a switch to interfacial failure by residual stress build-up at the interface, along with crack blunting by large diameter particles, greatly inhibited mode I performance for the formulations prepared with the more epoxy compatible toughener. Mode II results showed that retarded growth rates for the dispersed phases were preferred because a smaller phase size led to increased shear fracture performance.
Bonded Fracture Results
The bonded fracture results for the model adhesive systems are shown in Figures 5 and 6. The specimens were void-free in the bulk and interfacial regions and adhesive thicknesses were on the order of 0.10-0.12 mm.
The effects of silica content, rubber type, and rubber concentration on fracture energy can be seen from [TABULAR DATA FOR TABLE 6 OMITTED] the graphs as well as the statistical results generated in Table 6 using an analysis of variance. The failure modes for the bonded systems also seemed to be a function of these variables. Samples 1-4, 6, and 10 showed primarily cohesive failure in mode I while the remainder showed an adhesive/interfacial or mixed failure mode. Table 6 also shows the interactions between variables and how they influenced the fracture energy values.
The rubber level had a large impact on the fracture energy while the rubber type showed to be not as important to the mode I fracture values for bonding situations. However, the rubber type seemed to have a profound effect on the type of failure that was observed. All systems prepared with the more incompatible rubber displayed cohesive failure while a majority of the samples toughened with the other more compatible CTBN material showed some adhesive failure. This failure dependence on toughener type was also seen for the cocured mode I samples and suggests that the systems with the larger dispersed phase failed in the bulk of the adhesive (i.e. the crack was deflected from the interface to bulk). However, this interfacial failure for the bonded specimens did not result in diminished mode I fracture values as is frequently observed. Reasons for this behavior will be discussed in the next section. The interfacial failure in mode I testing (specifically for systems 5 through 8) might actually have caused an increase in the perceived fracture energy so that actual mode I values would have been somewhat less. Often interfacial failure between two dissimilar materials is accompanied by mixed mode loading at the crack tip (27). This is a direct consequence of the crack being constrained to a certain path and often results in artificially high values for double cantilever beam tests since mode II contributions are added ([G.sub.IC] [less than] [G.sub.IIC] for most materials).
The modifying effect by the silica on mode I fracture was seen when looking at samples 5 and 6 which just differed in their silica content. The silica version had such diminished nucleation and growth kinetics that a dispersed phase was never formed, instead the CTBN remained flexibilizing the system. Thus, the bulk adhesive possessed a sufficiently low fracture toughness in comparison to a system with lower continuous phase rubber and dispersed rubber particles and the crack propagation never traveled to the interface. The other system, however, had some adhesive failure present but the bulk material was tougher on average and therefore an increased energy was observed.
The ENF results for the bonded composite samples implied that mode II was only affected for the samples highly loaded with the more compatible rubber. These two samples also displayed the most non-linearity in their load-displacement curves prior to catastrophic failure indicating more energy was absorbed by plastic deformation. Therefore, the continuous phase rubber content along with the dispersed phase size governed the bonded mode II fracture results. An increased continuous phase rubber content along with smaller dispersed phase particles allowed for the best mode II properties as far as energy absorption before failure was concerned (this combination was only achieved for resin formulations 7 and 8). System 6 had a relatively high continuous phase rubber concentration but did not possess a secondary distinct toughening phase for increased shear fracture performance, illustrating the relative importance.
In summary, the bonded fracture results displayed behavior indicating the importance of rubber type, rubber level, and silica content. Mode I toughness showed a strong dependence on rubber level while rubber type and dispersed phase growth retardation only manifested themselves in the failure mode and not the absolute fracture values. Mode II results for the bonded specimens indicated that a somewhat flexible continuous phase and more importantly a small domain size were desired and consequently the more compatible toughener was favored.
Cocured Versus Bonded
It is apparent from Figures 3 through 6 that the silica content, rubber type, and rubber content had a strong influence on the fracture properties of adhesively joined carbon fiber adherends. However, the large discrepancy between cocured and bonded results was unusual. The bonded specimens on average significantly outperformed the cocured samples. This intuitively was not expected and it should be mentioned that this trend may be specific to the type of high performance prepreg tested. Regardless of this fact, it brings up several issues which should not be overlooked or discounted when considering bonding and cocuring with adhesive materials.
The bonded versus cocured results indicated that the main mode of adhesion between the two resin systems was not a result of diffusion at the interface or interpenetrating networks. These issues have been commonly discussed as the main adhesion and/or toughness contribution mechanisms in certain similar polymeric systems, but in this study the diffusion model did not play a significant role (28-31). It appeared more that other adhesion modes such as mechanical interlocking were more important for this study. Mechanical interlocking appeared to have a reasonable influence on the precured panels. The panels which were sanded prior to adhesive application provided increased fracture results. The cocured resin systems, conversely, were smoother on a more microscopic level. Therefore the interfacially failed specimens would have slightly higher fracture values because of the larger asperities at the surface which would provide a larger frictional resistance. An example of these increased asperities or surface roughness is shown in Figure 7. This explains for mode I results why the bonded interfacially failed specimens showed similar fracture energy values to the cohesively failed bonded samples and why the same behavior was not seen for the cocured specimens. Because of the more saw-tooth pattern of the bonded adherends, the shear contribution to the DCB values was significant. However, this reason alone is not enough to fully explain the results since the untoughened bonded resins did not show increased mode II results as they would have if mechanical interlocking was the main factor. This implies that the toughening materials also had an influence on the cocured versus bonded results.
The [G.sub.IIC] results between the bonded and cocured specimens only showed significant differences for the highly rubberized systems (3, 4, 7, 8) with the smaller particle systems performing best. Stress relaxation tests were performed on bulk samples for all the adhesive formulations in order to try and simulate the mode II loading rate and other conditions. The stress decay ratio as a function of initial stress was obtained for the resin systems and the results were a direct function of the rubber remaining in the continuous phase. The increase of the stress decay values between the systems with limited flexibilizing rubber and those with more significant levels was only from 1.5 to 2.75% (stress percentage reduction over a period of sixty seconds). This indicates that the probability of increased fracture results through a stress relaxation mechanism was very small. It seemed more likely that the systems with mixed mode loading performed better by taking advantage of both the adherend roughness and dispersed phase particles.
Primary bonding between resin systems also did not appear to provide any significant adhesive force. If primary bonding were to occur, it would have been anticipated in the case of the cocured specimens and an increased interfacial strength realized. However, this was not observed experimentally, in fact, inferior numbers were seen in comparison to the bonded structures. The possibility of poisoning was considered for the cocuring specimens but this seems remote considering both systems were primarily epoxy based (DGEBA) resins. This was confirmed using differential scanning calorimetry where the different resin systems were put in contact with each other during cure and did not show a decrease in the overall heat of reaction. The reaction value was a weighted average of the individual components as would be expected if there were no interactions.
One difference between the sets of samples was that there were obvious signs of interlaminar or through-thickness resin flow in the cocured fracture samples. The flow usually was situated at prepreg-fiber overlap regions (tow interstitial areas) and commonly extended from one to three plies in distance from the interface. Despite this resin flow, adhesive layer thicknesses between the cocured and bonded samples were nearly identical, differing on the average by only 5-10%. The fracture path also never extended into these regions. Accompanied with the increased resin flow was a higher mobility between the resin systems for the cocured samples. This increased mobility along with other free energy reduction considerations is likely what caused some preferential location of the dispersed rubber-rich phases at the interface for the cocured samples. This phenomenon is illustrated in Figure 8 and might have contributed to the lower average values for cocured samples by creating a reduction in the interfacial toughness.
The effect of CTBN toughener content, type, and fumed silica addition on different properties for an epoxy adhesive were studied. Of special interest was analyzing the role that flexibilized rubber played in the final performance properties. Viscosity, morphology, mode I and mode II fracture energy analyses, and lap shear tests were conducted for different model resin systems with wide variations in the observed results.
The viscosity increased with silica addition and sufficiently slowed the dispersed phase (rubber-rich) growth rates such that smaller sizes were observed. The concentration of rubber in these phases also decreased with silica addition as determined using dynamic mechanical analysis and microscopy. Fracture samples were prepared with carbon fiber-epoxy prepreg systems under both cocured and bonded situations. Interfacial versus cohesive failure was found to have a strong effect on the fracture energy especially for the bonded specimens. The fracture results showed a definite dependence on the morphological differences that were created with the CTBN type, amount, and silica content. Mode I results favored systems with larger domains which helped to blunt the crack tip during loading while mode II results showed better behavior with small dispersed phase sizes. Aluminum lap shear results showed a strong dependence on flexibilized rubber content and dispersed phase size. Higher flexibilized rubber content along with small dispersed phase particles showed best performance in lap shear.
There existed significant differences between the bonded and cocured fracture sample behavior for the model adhesives with the bonded results generally outperforming the cocured. Mechanical interlocking for the bonded samples increased the fracture results provided that interfacial failure was observed. This was a result of the large asperities that created an increased frictional resistance during interfacial failure. Preferential migration of the dispersed phases to the carbon fiber composite host resin was also observed for the cocured specimens and diminished the fracture performance by reducing interfacial epoxy-epoxy interactions.
Collectively, this work provided information regarding toughened epoxy adhesive systems that helped show the property and morphological dependence on the manipulated flexibilizing and dispersed phase CTBN toughener content. Specifically the results relate to the joining of carbon-fiber polymer composites.
1. C. A. May, ed. Epoxy Resins, 2nd Ed., Marcel Dekker, Inc., New York (1988).
2. W. G. Potter, Epoxide Resins, Springer-Verlag, New York (1970).
3. B. S. Hayes, and J. C. Seferis, Journal of Applied Polymer Science, 61, 37 (1996).
4. F. Rodriguez, Principles of Polymer Systems, 3rd Ed., Hemisphere Publishing Corporation, New York (1989).
5. N. Rose, M. Le Bras, S. Bourbigot, and R. Debbel, Polymer Degradation and Stability, 45, 387 (1994).
6. S. V. Levchik, et al., Polymers for Advanced Technologies, 6, 53 (1995).
7. A. J. Kinloch, in Rubber-Toughened Plastics, C. K. Riew, ed., American Chemical Society, Washington, D.C. (1995).
8. D. Chen, J. P. Pascault, H. Sautereau, R. Ruseckaite, and R. J. J. Williams, Polymer International, 33, 253 (1994).
9. D. J. Hourston, J. M. Lane, and N. A. MacBeath, Polymer International, 26, 17 (1991).
10. C. B. Bucknall and I. K. Partridge, Polymer, 24, 639 (1983).
11. P. Bussi and H. Ishida, Polymer, 35, 956 (1994).
12. M. A. Hoisington, D. Thompson, and J. C. Seferis, in 37th International SAMPE Symposium, 37, 264 (1992).
13. N. J. Pagano, ed., Interlaminar Response of Composite Materials, Composite Materials Series, ed., R. B. Pipes. Vol. 4, Elsevier, New York (1989).
14. J. W. Putnam, Doctorate Thesis, Composite Prepreg Process Characteristics, University of Washington (1996).
15. W. Rasband, Void analysis was performed on a Model 7100/80 Power Macintosh using the public domain NIH Image program (written by Wayne Rasband at the U.S. National Institutes of Health and available from the Internet by anonymous ftp from zippy.nimh.nih.gov or on floppy disk from NTIS, 5285 Port Royal Rd., Springfield, VA 22161, part number PB93-504868).
16. M. Gordon and J. S. Taylor, Journal of Applied Chemistry, 2, 493 (1952).
17. J. P. Chen and Y. D. Lee, Polymer, 36, 55 (1995).
18. S.C. Kim, M. B. Ko, and W. H. Jo, Polymer, 36, 2189 (1995).
19. J. H. Klug and J. C. Seferis, Journal of Applied Polymer Science, 66, 1953 (1997).
20. C. K. Riew, E. H. Rowe, and A. R. Siebert, in Toughness and Brittleness of Plastics, R. D. Deanin and A. M. Crugnola, eds., American Chemical Society, Washington, D.C. (1976).
21. A. J. Kinloch, S. J. Shaw, D. A. Tod, and D. L. Hunston, Polymer, 24, 1341 (1983).
22. F. J. Sultan and F. J. McGarry, Polymer Engineering and Science, 13, 29 (1973).
23. S. W. Yurgartis, B. E. Mast, and M. D. Bush, Journal of Adhesion Science and Technology, 9, 199 (1995).
24. R. A. Gledhill, A. J. Kinloch, S. Yamani, and R. J. Young, Polymer, 19, 574 (1978).
25. W. D. Bascom, R. L. Cottington, R. L. Jones, and P. Peyser, Journal of Applied Polymer Science, 19, 2545 (1975).
26. Y. Nakamura, M. Yamaguchi, A. Kitayama, M. Okubo, and T. Matsumoto, Polymer, 32, 2221 (1991).
27. D. Hull and T. W. Clyne, An Introduction to Composite Materials, Cambridge, New York (1996).
28. M. S. High, P. C. Painter, and M. M. Coleman, Macromolecules, 25, 797 (1992).
29. Y. C. Chern, K. H. Hsieh, and J. S. Hsu, Journal of Materials Science, 32, 3503 (1997).
30. J. K. Kim and C. D. Hah, Polymer Engineering and Science, 31, 258 (1991).
31. S. S. Voyutskii and V. L. Vakula, Journal of Applied Polymer Science, 7, 475 (1963).
|Printer friendly Cite/link Email Feedback|
|Author:||Klug, Jeremy H.; Seferis, James C.|
|Publication:||Polymer Engineering and Science|
|Article Type:||Statistical Data Included|
|Date:||Oct 1, 1999|
|Previous Article:||Physicochemical characterization of alloy of polymide with varying degree of crosslinking through diisocyanates.|
|Next Article:||Monodisperse Poly(p-chloromethylstyrene) Microbeads by Dispersion Polymerization.|