The Effect of Thermooxidative Aging on the Durability of Glass Fiber-Reinforced Epoxy

1Department of Aerospace Engineering, Universiti Putra Malaysia, 43400 Serdang, Selangor, Malaysia 2Aerospace Manufacturing Research Center (AMRC), Level 7, Tower Block, Faculty of Engineering, UPM, 43400 Serdang, Selangor, Malaysia 3Medicontur Medical Engineering Ltd., Herceghalmi Road, Zsámbék 2072, Hungary 4Bay Zoltan Nonprofit Ltd. for Applied Research, Fehérvári Út 130, Budapest 1116, Hungary


Introduction
The importance of persistently improving civil aircraft performance and improving fuel consumption has resulted in novel improvement in the designs of aeronautical structures.This improvement has also led to a drastic growth in the application of advanced materials.Consequently, aeronautical industries have begun to use thin-skinned polymer matrix composite materials in aircraft construction.The Boeing 787 and Airbus A380 aircraft pioneered this innovative structural design.Such high speed civil aircraft offer considerable promise of being able to withstand long-term (60,000 h) exposure [1][2][3][4][5] to temperatures ranging from −40 to 177 ∘ C [5] and being sufficiently ductile to absorb low velocity impact energies [6][7][8][9][10].However, during thermal aging at moderate temperatures, for example, below the glass transition temperature, organic matrix composites deteriorate by matrix embrittlement resulting from thermooxidative degradation [11].Importantly, because oxidation is a diffusion-related phenomenon [12], using thin-skinned composites may accelerate the onset of observable changes in the mechanical, thermomechanical, and chemical properties that may result in lowered damage tolerance and structural integrity.As a result, investigating thin-skinned composite durability at the early stages of thermooxidative degradation is a crucial issue that needs to be understood to avoid premature in-service failures.
Thermal-oxidative aging may irreversibly change the chemical structure of polymer matrix composites [13,14].Alterations to the chemical structure during thermooxidative 2 Advances in Materials Science and Engineering degradation include postcuring [1,9,13,[15][16][17][18][19][20][21][22], loss of volatiles [2,[23][24][25][26][27], dehydration [13,28,29], chain scission [25,[30][31][32][33][34], additional cross-linking [1,13], and carbonyl growth [19,29].The initial chemical changes are accompanied by dehydration of secondary alcohols and the release of low molecular weight gaseous species due to random chain scission.However, at moderate heat exposures (24 ∘ C-177 ∘ C), chain cross-linking is the dominant chemical change in the matrix compared to chain scission [14].Indeed, the increase in the cross-link density of the cured matrix primarily occurs during the initial aging period and is caused by postcure reactions [13,17,35,36], which result in the excessive brittleness of the matrix [37].As aging proceeds in the presence of oxygen, susceptible chemical structures in the resin are oxidized to various carbonyl containing groups.Therefore, matrix embrittlement increases with the oxygen concentration and aging time [2,38].In addition to thermooxidative aging, the effects of physical aging [39][40][41] should be considered.Physical aging is universal, is independent of any chemical change, and is related to the gradual densification of nonequilibrium glassy structures.These properties result in a higher modulus and strength, lower toughness, and slower relaxation (creep and stress relaxation).The aging is also accompanied by enthalpy relaxation.Physical aging can be accelerated by annealing, such as a thermal treatment below but not too far from the glass transition temperature.These parameters are exactly the conditions used for mild oxidative aging.The effects of postcuring and chemical and physical aging are inextricably combined in epoxy resins.
The thermomechanical behavior in organic matrix composites can be changed as a result of postcure reactions and physical aging induced by thermooxidative degradation.More precisely, the cross-linking density is increased by a continued postcure reaction and physical aging processes restrict the molecular movement in the main chain of the matrix [17,35,41].The reduction in the molecular mobility results in an increase in the glass transition temperature [2,17,[39][40][41] and the storage modulus [20,[42][43][44][45].In fact, an evaluation of the viscoelastic behavior can detect changes in the state of the molecular motion as functions of time and temperature [39,46], which are responsible for relative stiffness in polymer matrix composites [39,47,48].Storage modulus and the peak, shape, and the area under tan  curve are typically used as indicators in studies on molecular crosslinking [49].
Matrix densification affects the mechanical properties of OMCs (organic matrix composites) in the early stages of oxidative aging and is induced by physical aging and/or chemical aging (postcuring).Research conducted by Tsotsis et al. suggests that the matrix dependent properties are enhanced during the postcure stage and matrix degradation resulted from densification or increased chain scission in the thermosetting matrix polymers, respectively [1].Other studies revealed that postcure reactions enhance the fiberdominant properties, such as the tensile strength and tensile modulus [15,41,50,51], and decrease the matrix dominated properties, such as toughness [3,52].Similarly, an increase in the tensile strength [15,41] and modulus [53][54][55][56] and a reduction in the toughness [41] caused by physical aging have been observed by several researchers.
The thermooxidative degradation of organic matrix composites has been observed [3,4,23,51,57,58] to be accompanied by the oxidation of the topmost surface of the matrix and by the formation of superficial microcracks.In fact, with continued aging, void formation occurred [59,60] and cracks developed from crazing by increasing extensional stress [61].A considerable amount of literature concluded that chemical shrinkage [23,25,26,[62][63][64][65] or notable variations in the matrix and fiber thermal expansion factors [36,66] may provide dilatational stresses (internal stresses) or synergistic effects of both.In addition, in morphological investigations, fiber-matrix interfacial adhesion also degraded and weakened as a result of moderate heat exposure [3,[67][68][69], resulting in fiber pullout failures as part of the fracture mode [70][71][72][73].
In this study, the influence of thermooxidative degradation on the durability of glass fiber-reinforced epoxy composite is evaluated from mechanical, chemical, and physical perspectives.This work seeks to relate the premature alterations in the thin-skinned composite panel to the deterioration of its durability.Examining thermal-oxidative degradation using a principles-based approach and by considering the original composite thickness aids in assessing the actual toughness behavior available for the use condition by mechanical testing.As a result, measuring the toughness value by calculating the area under the stress-strain curve method was used instead of the Charpy impact test due to the subsize thickness of the material available.Furthermore, the methodology introduced here is based on industrial interest because the thermooxidative aging of such thin-skinned composites has not received a great deal of attention.Therefore, this study is anticipated to contribute to the body of literature on the materials used for the manufacturing of aircraft and especially to the durability of the materials in terms of the pronounced short-term thermal-oxidative degradation.
This study uses aging of plain GFRPs (glass fiberreinforced plastics) laminates exposed to air for 800 h at a temperature of 82 ∘ C. Various tests are conducted to characterize the deterioration of the material, including tensile testing, dynamic mechanical analysis (DMA), Fouriertransformed infrared spectroscopy (FTIR), weight loss measurements, Scanning Electron Microscopy (SEM), and differential scanning calorimetry (DSC).

Material
The primary material used in the manufacturing of the plain laminates is the EHG250-68-37 prepreg.EHG250-68-37 prepreg is comprised of an E-glass fabric impregnated with the DGEBA based epoxy resin EH250 and manufactured using the US 7781 standard.The epoxy resin is a product of the Gurit Company.Based on the recommendation of the manufacturer, the laminates were cured in an autoclave [74].The length of the panel coincided with the plain weave on the weft direction.This coincidence stems from the proper stacking orientation of the four prepreg plies.Before cutting the 1 mm thick panels into appropriate test pieces using a Advances in Materials Science and Engineering 3 CNC machine, the material was inspected using an ultrasonic C-scan to identify any defects.

Isothermal Aging.
In accordance with the ASTM D5229/D5229M standard, the samples used in this study were dried to a constant weight under vacuum conditions at a temperature of 70 ∘ C before aging [75].This preliminary condition is essential because the matrix composite is hydrophilic in nature and variation in weight during aging serves as an oxidation index.To begin, the specimens were isothermally aged in an air circulating oven for a period of 800 h at 82 ∘ C. The chosen temperature is slightly above the standard use temperature (80 ∘ C) and is below   (110 ∘ C in TMA).For the sake of preventing damage to the specimen edges during pronounced thermooxidative aging, the edges were sealed using aluminum foil bonded with Dow Corning 3120 silicon rubber, which is stable at temperatures in excess of 300 ∘ C. In addition, the relative humidity in the oven was reduced using phosphorus pentoxide.This thermal condition acts as an artificial thermooxidative aging condition to reproduce the use condition for the short-term applications.

Tensile Test.
Using an Instron-5582 universal testing machine set at a 1 mm/min crosshead speed, the static tensile tests were performed in accordance with ASTM D3039/D3039M [76].Strain gauges were attached in the longitudinal direction (the weft direction axis of the specimen) to measure the tensile strain.The final values were taken from an average of the five specimens.

Dynamic Mechanical Analysis (DMA).
A dynamic thermal analyzer (TA instrument model Q800) was used to conduct the thermal mechanical tests on both the aged and the unaged samples according to the ASTM D7028-07e1 standard [77].A three-point bending geometry was used to apply deformation.The experiment was also conducted in scanning temperature mode, ranging from 25 ± 2 to 250 ∘ C. Other conditions include a 5 ∘ C/min heating rate and a 1.0 Hz oscillating frequency.

FTIR Spectroscopy.
From the aged and unaged samples, approximately 2 mg of the previously dried composite was used for the FTIR.The milled sample was pressed into a disc after being mixed with approximately 50 mg of anhydrous Potassium Bromide (KBr).Spectra were acquired in the transmission mode using a Thermo Nicolet Nexus 470 FTIR spectrometer.The disc analysis was performed by treating the KBr as the background reference.Acquisition of the spectra was performed over a range from 4000 to 400 cm −1 .Afterward, three separate analyses were conducted on each pallet at different locations to ensure the results were consistent.

Weight Loss Measurements.
To monitor the weight loss from the samples over the 800 h oxidative aging at 82 ∘ C, three specimens were cut from the original composite panel with a width of approximately 20 mm and a length of approximately 75 mm.In a periodic manner, the samples were removed from the oven, placed in a desiccator, and allowed to cool to room temperature.The samples were then weighed and returned to the oven.Weighing was performed using a microbalance accurate to within 10 g.Each data point reported is an average of three separate specimens weighed independently.

Scanning Electron Microscopy (SEM). A Hitachi SU1510
SEM was used to investigate the superficial matrix morphology.Additional SEM analyses were performed on the cryofractured specimens using liquid nitrogen to study the fracture behavior of the composites and possibly any matrixfiber adhesion.In both morphological investigations, the surfaces of the specimens were sputter coated with gold.Additionally, observations were conducted on at least three different locations on the samples to ensure that the reported microphotograph represented the typical morphology.

Differential Scanning Calorimetry.
The DSC measurements were performed using a DSC 823 system manufactured by Mettler Toledo with an air flow of 50 mL/min.A temperature scan was performed on 10 mg prepreg samples (EHG250-68-37) from 25 ∘ C to 250 ∘ C at a heating rate of 10 ∘ C/min.

Results and Discussion
4.1.Tensile Test.The comparison of both aged and unaged samples based on changes in the tensile strength and modulus is shown in Figures 1 and 2, respectively.The tensile strength and Young's modulus substantially increased under the pronounced aging conditions by 48% and 59%, respectively.These changes may be attributed to the increase in the crosslinking density of the epoxy matrix as a result of simultaneous postcuring and physical aging during thermooxidative degradation [15,41].To be more precise, by increasing cross-linking density, the cross-linked network within the epoxy matrix will bear the majority of the induced loads in the composite material [13].Consequently, the tensile strength and modulus of elasticity rose significantly over the prolonged aging period.
However, a 0.22% drop in strain-to-failure value was observed after pronounced oxidative aging process, as shown in Figure 3. Furthermore, the toughness measured by the area under the stress-strain curve decreased considerably by 49% after 800 h aging at 82 ∘ C compared to original sample, as shown in Figure 4.This remarkable loss in toughness and strain to failure may be explained by the embrittlement of the epoxy matrix induced by postcuring [37], physical aging [41], and the oxygen concentration [2,38].

Dynamic Mechanical Analysis (DMA).
The influence of thermooxidative aging over a period of 800 h at 82 ∘ C on properties, such as the glass transition temperature (  ), the onset temperature, mechanical damping (tan ), and the storage modulus (  ), was studied by conducting dynamic mechanical thermal analyses.The glass transition temperature (  ) is taken as the temperature at which the loss factor is maximized.Additionally, the onset temperature is determined from an inflexion point in the modulus curve.Therefore, this measurement is less accurate than when measured from the tan  maximum temperature.However, the onset temperature is used more by aircraft manufacturers compared to the glass transition temperature (  ) [78][79][80].
As shown in Figure 5, the glass transition temperature (  ) increases moderately over the period of induced thermooxidative aging.Additionally, tan  peak in the aged composite is relatively narrow compared to the original sample.Table 1 shows a quantitative comparison of the area under tan  peak between the aged and unaged samples.The results show a 19% reduction in the area under the damping peak after the induced thermal aging.Moreover, there appears to be doubling of the loss peaks in the unaged specimens at higher temperatures.This finding may be attributable to an inhomogeneous network structure or to the diffusion controlled oxidation process being stronger near the surfaces.
The behavior of the dynamic storage modulus (  ) during the pronounced thermooxidative degradation is shown in Figure 6.A minor increase is observed at the glassy state    compared with the original sample.Moreover, a trend in the increased rubbery state   with induced thermal aging is seen (Figure 6). Figure 7 shows the changes in the onset temperature after the 800 h thermal-oxidative process at 82 ∘ C compared to the original composite.The onset temperature drops from 83.58 ∘ C to 80.58 ∘ C, which is close to the maximum service temperature recommended by the manufacturer [74].
The observations above may be explained as follows.The alteration in   , tan  peak, and the storage modulus implies that the cross-link density of the matrix network's structure is being enhanced by the increasing temperature (as a result of postcuring and physical aging).Indeed, matrix densification results in the restriction of the molecular mobility, leading to a rise in the storage modulus and the glass transition temperature.However, the area under tan  peak decreases after the induced aging because of the increased cross-linking and steric hindrance in the matrix, which leads to a decrease in the damping capability (defined as the ability to absorb and dissipate energy or as the toughness) of the composite during pronounced aging conditions.This may be in accordance with the decrease in toughness observed in the tensile tests.Furthermore, the results from the onset temperature reveal that the thermal durability of the composite drops by approximately 3 ∘ C. This, in turn, leads to premature mechanical failures.

FTIR Spectroscopy.
In the hydroxyl regions (≈3800-2500 cm −1 ), we observe that for the pronounced aging condition the broad band of the spectra is centered at 3425 cm −1  (corresponding to O-H stretching of the hydroxyl group) and the peak at 2927 cm −1 (corresponding to C-H stretching of the methylene group) decreases in intensity, as shown in Figure 8.Additionally, a decrease in the absorption peak at 1095 cm −1 is noted as a result of the aging process (Figure 9).This decrease is responsible for the vibration of C-C-O associated with the secondary alcohol.From the changes in the peaks, we can deduce that the degradation of the epoxy resin originated in the -C-C(OH)-C-linkages.A theory was also formulated that the degradation of the epoxy resin secondary alcohol structure might be associated with the dehydration reaction that leads to the formation of C=C bands [81][82][83].A similar idea was also put forth where the degradation of the epoxy secondary alcohol structure may be attributed to the thermal-oxidative reaction that may lead to the formation of peroxides [84,85].From the FTIR observations, oxygen did not attack the material during the oxidative aging process to amine groups to form amide compounds.However, at approximately 830 cm −1 , moderate growth was observed in the intensity of the peroxide group.

Advances in Materials Science and Engineering
The IR spectra of the E250 epoxy system in the fingerprint region are shown in Figure 9.Although the absorption band at 915 cm −1 diminished or disappeared during the curing process, absorption of the N-H group in the curing agent at 1560 cm −1 and 1580 cm −1 was observed in both the aged and virgin samples.However, as the aging time increased, the absorption band at 1560 cm −1 and 1580 cm −1 decreased, as shown in Figure 10.The reduction in these bands may be attributed to the moderate temperatures associated with the thermooxidative aging advancing the mobility of the molecular chains, which increases the reaction probability between the resin and the curing agent (known as the probability of densification) [35].
In the carbonyl region (≈1900-1500 cm −1 ), we observe that the absorbance of the band at 1737 cm −1 , which corresponds to the characteristic stretching of C=O in phenyl, slightly increased in intensity (Figure 11).The formation of carbonyl groups may result from the oxidation of aromatic ether groups in the resin [19,29].This oxidation process is accompanied by a slight reduction in the intensity of the aromatic ether group, approximately 1035 cm −1 [29].The overall oxidation rate is very slow, significantly more so than under the high temperature oxidation conditions [86], as shown by the very moderate increase in the carbonyl band, which is a sensitive measure of the degree of oxidation in all organic polymers.

Weight Loss Measurements.
The weight loss percentage experienced by the composite specimens during the 800 h of thermooxidative aging at 82 ∘ C is depicted in Figure 12.The weight loss rises sharply during the initial 100 h of aging.This may be attributed to desorption of the residual moisture and volatiles and is not related to polymer degradation [27,87].A faster weight loss rate is evident over the 400 to 600 h range.Over this period, the mass loss rate reaches a peak of 0.13% at 600 h before staying constant over the remainder of the aging.This accelerated weight loss may occur due to the appearance of superficial cracks in the samples, which provide an increased surface area available to react with oxygen [88].

Scanning Electron Microscopy (SEM).
After 800 h of thermooxidative degradation at 82 ∘ C, notable changes occurred on the sample surface.First, a SEM micrograph showing the microstructure of an unpolished sample is shown in Figure 13 in the backscattered electron mode.Due to different phase contrasts in the scattered electron distribution, oxidized carbonaceous layers appear as white spots on the matrix surface in contrast to the original resin surface.Next, a significant increase in the size and concentration of voids (indicated by arrows) in the aged sample becomes evident (Figure 14(b)), which may be explained as a result of the thermooxidative degradation process.In addition, a superficial crack is observed in Figure 14(b), which is a result of crazing from local stress concentrations and/or regions of microstructural inhomogeneity.Indeed, crazing forms from interlocked voids with extremely tightly drawn fibrils [61].By increasing the extensional local stresses and matrix embrittlement induced by the elevated temperatures for a period of time, the fibrils are unable to absorb the dilatational thermal stress, which leads to the failure of fibrillated crazing structures and crack propagation.
The microscopic observations in the core cross-sectional area show the marginal effects of thermal degradation on the damage mechanisms that result in matrix shrinkage compared to the virgin sample (Figure 15).The resin shrinkage is related to the physical aging (densification) and/or volatilization of small molecular compounds during the postcure process.However, no evidence of fiber-matrix debonding caused by matrix shrinkage was observed.Therefore, the local stresses induced by resin shrinkage are insufficient to overcome fiber-matrix interfacial bonding but may be sufficient to weaken it.
Figure 16 illustrates the cryofractured sections of both the aged and unaged samples.An interesting observation from Figures 16(a) and 16(b) on the unaged sample is that the surfaces of the glass filaments have delicate coating of resin.This implies there is good bond strength between the matrix and the fiber.However, after temperature-induced degradation, the matrix completely detached from the filaments and fiber bundle, as shown in Figures 16(c) and 16(d), respectively.In addition, fiber pullout from the matrix was observed, as depicted in Figures 16(e) and 16(f).All the failure modes take the form of interfacial debonding, and the fiber pullout may be attributed to decreases in the interfacial fiber-matrix adhesion resulting from resin embrittlement as a result of thermal degradation.It is worth mentioning that the poor interfacial fiber-matrix bonding causes a reduction in interfacial dominant properties, such as the toughness.This reduction, in turn, leads to reduced damage tolerance and lack of long-term durability.
4.6.Differential Scanning Calorimetry.In addition to the DMA and FTIR analyses, DSC curves were also used to detect the possible postcuring and physical aging effects.
Figures 17 and 18 show the DSC curves for the unaged and aged specimens, respectively.
The differences are summarized as follows: (i)   calculated from the midpoint curved baseline (Δ  ) shifts from 79.78 to 81.29 ∘ C after aging, which may indicate postcuring.These   values should be compared to tan  maxima shown in Figure 5 (corresponding to 115.09 and 129.02 ∘ C).It is well known that DSC usually determines the glass transition to be at lower temperatures than the DMA.The increase in   is considerably larger in the DMA (approximately 14 ∘ C) than in the DSC (approximately 1.5 ∘ C).However, it should be noted that the DMA tan  curve of the aged specimen shown in Figure 5 is doubled, and there is a subcomponent close to the original one, which may correspond to the transition detected by the DSC.Advances in Materials Science and Engineering (ii) The post-  region of both DSC curves is complicated, and there are two broad, subsequent processes: one in the endothermal and one in the exothermal direction.Additionally, the unaged specimen exhibits a small but detectable melting peak, which disappears after aging.Under these conditions, the selection of the proper dividing line and the value of the integral obtained for the subprocesses are somewhat ambiguous.Nevertheless, we establish that the sharp melting disappears, the integral of the endotherm remains practically unchanged, and the integral of the exotherm decreases after aging.The interpretation (in the absence of the exact chemical composition) is tentative at best.The melting in the unaged specimen may be due to some unreacted component, which disappears after the thermal aging.The endotherm may be smeared overshoot attributable to physical aging (although it is usually sharper and increases with annealing) and the exotherm may be the residual reaction heat due to incomplete cross-linking, which diminishes after thermal aging (postcuring).

Conclusions
This study was designed based on industrial interests to evaluate the effect of thermooxidative aging on the durability of thin-skinned EHG 250-68-37 prepreg over a period of 800 h of isothermal aging at 82 ∘ C. The results obtained from the mechanical, chemical, and physical experiments lead to the following conclusions: (i) The tensile strength and modulus increase after aging, whereas the toughness measured by the area under the stress-strain curve decreases (due to lower elongation at break).These characteristics are probably attributable to the embrittlement of the matrix resin as a result of postcuring and oxygen concentration.
(ii) The glass temperature and the rubbery plateau modulus increase, which is clearly indicative of a postcuring process.
(iii) According to the FTIR observations, postcuring reactions occurred as the aging proceeded.Additionally, dehydration reactions cause the formation of secondary alcohol and aromatic ether groups in the resin, which are oxidized to carbonyl compounds.
(iv) From the electron microscopy observations, the presence of minor cracks resulting from postcuring is verified.However, no skin-core cracking is observed.In addition, matrix embrittlement induced by aging leads to fiber pullout and weakening of the fibermatrix adhesion.
(v) The overlap of the broad endothermal and exothermal process observed above   in the DSC make an exact interpretation difficult, but a diminishing exotherm could be identified and is compatible with postcuring.
These findings suggest that, in general, matrix embrittlement resulted from postcure reactions and physical aging plays a significant role in reducing the durability of thinskinned composite by causing premature failures.This study provides a better understanding of the nominal useful structural life for such composite materials.However, one source of weakness in this research which limited the thermal and FTIR analyses was that the chemical structure and compositional data of the matrix were not revealed by the manufacturer.Finally, it is recommended that further research investigate the effect that thermooxidative aging has on the chemical structure of the toughening agent in the resin.

Figure 13 :
Figure 13: SEM micrograph of an unpolished surface after aging treatment.

Figure 14 :
Figure 14: SEM observation on the polished matrix surface before (a) and after (b) 800 h thermooxidative aging at 82 ∘ C.

Figure 15 :
Figure 15: SEM micrographs of fiber ends at the cross-sectional area of unaged (a) and aged (b) sample.

Figure 16 :
Figure 16: SEM observation on the cryofractured surface of EHG250-68-37 composite specimens; (a) fiber-matrix adhesion of unaged sample; (b) zoom on fiber-matrix adhesion corresponding to (a); (c) fiber-matrix adhesion of aged sample; (d) zoom on fiber-matrix adhesion corresponding to (c); (e) matrix detached from fiber bundle of aged sample; (f) fiber pullout of aged sample.
a Area value indicates the area under tan  peak.