Polymer Degradation and Stability 93 (2008) 17861792

Contents lists available at ScienceDirect

Polymer Degradation and Stability

journal homepage: www.elsevier.com/locate/polydegstab

Photoageing behaviour of epoxy nanocomposites: Comparison between spherical and lamellar nanollers
` Benedicte Mailhot a, b, *, Sandrine Morlat-Therias a, Pierre-Olivier Bussiere a, Loic Le Pluart c, Jannick Duchet c, Henry Sautereau c, Jean-Franois Gerard c, Jean-Luc Gardette a
a b c

` Laboratoire de Photochimie Moleculaire et Macromoleculaire, UMR CNRS-UBP 6505, Universite Blaise Pascal, Ensemble Universitaire des Cezeaux, 63177 Aubiere Cedex, France ` Institut Universitaire de Technologie, Universite dAuvergne, Ensemble Universitaire des Cezeaux, B.P. 86, 63172 Aubiere Cedex, France Laboratoire des Materiaux Macromoleculaires, INSA de Lyon, UMR CNRS 5627, 20 Avenue Albert Einstein, 69621 Villeurbanne Cedex, France

a r t i c l e i n f o
Article history: Received 14 May 2008 Accepted 17 July 2008 Available online 25 July 2008 Keywords: Epoxy Nanocomposite Irradiation Oxidation AFM nanoindentation

a b s t r a c t
The photochemical behaviour of a nanolled epoxy resin has been studied. It has been shown that the ller content increases at the surface with irradiation time. Qualitative stiffness and adhesion measurements compare the surface properties of the lled and unlled samples upon ageing. Depth proling has been achieved by AFM nanoindentation and micro-FTIR. These two techniques allowed comparing stiffness and photooxidation of the aged samples. Both techniques showed an inuence of nanollers on thickness proles. The relationship between the oxidation process and its consequences on the physical properties is explained taking into account oxygen permeability and light diffusion. Additionally, it has been shown that, whatever the content (510 wt%), the nature (silica or different organomodied montmorillonites) or the shape of the ller (spherical or lamellar), the photoproducts were formed in comparable proportions and at similar rates as in the pristine matrix. 2008 Elsevier Ltd. All rights reserved.

1. Introduction Since a montmorillonite reinforced nylon nanocomposite was developed by Toyota group [13], much attention has been devoted to smectite as a reinforcement material for polymers. Because of the nanometer-size dispersed particles and their high aspect ratio characteristics, these nanocomposites exhibit improved mechanical, thermal and physicochemical properties in comparison to pristine polymers or conventional microcomposites. The most commonly used clay is the smectite group mineral such as montmorillonites (MT), which belongs to the general family of 2:1 layered silicates (phyllosilicate). The thickness of the single layer also called lamella is about 1 nm and the length varies from 100 to 200 nm. The lamellae form particles associated in aggregates. In order to optimize the polymerclay interactions, aggregates and particles should be delaminated. A practical problem in the synthesis of nanocomposites is to disperse the inorganic clay in an organic medium. This can be achieved by treating the clay so that it becomes organophilic. Usually, the cations located in the galleries between the clay

* Corresponding author. Laboratoire de Photochimie Moleculaire et Macro moleculaire, UMR CNRS-UBP 6505, Universite Blaise Pascal, Ensemble Universitaire ` des Cezeaux, 63177 Aubiere Cedex, France. Tel.: 33 473 177 166; fax: 33 473 407 700. E-mail address: benedicte.mailhot@univ-bpclermont.fr (B. Mailhot). 0141-3910/$ see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.polymdegradstab.2008.07.020

platelets are exchanged with organic cations such as alkylammonium ions. They lower the surface energy of the clay so that organic monomers with different polarities can diffuse between the clay layers. The degree of exfoliation and dispersion of the MT silicate layers in the nanocomposites can be investigated by using X-ray diffractometer and transmission electron microscopy (TEM). Diffraction peaks observed in the low-angle region indicate the basal spacing of ordered-intercalated and ordered-delaminated nanocomposites. However, if the nanocomposites are disordered, no peaks are observed in the XRD, and layered silicate can either be delaminated or intercalated. In such cases, TEM combined with XRD characterizes more accurately these materials. Two organo-montmorillonites (MMt) modied by dimethylbenzyl-tallowalkyl-ammonium ions and by C18 alkylammonium ions were tested and compared to conventional silica nanoller. The blends were prepared by an in situ polymerization of bisphenol-A diglycidyl ether (DGEBA) and polyoxypropylenediamine (Jeffamine D2000). Preparation and characterization of these nanocomposites have already been described [4]. Photoageing tests were carried out at l > 300 nm in the presence of oxygen. The aim of this work is to analyze the effect of nanollers on the mechanism and the kinetic of degradation of the polymer. The inuence of ller shape and concentration will also be discussed. In addition, the clay modies oxygen permeability and increases the light pathway through the sample. These phenomena

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792

1787

are expected to have an effect on the distribution of the oxidation products in the sample thickness. The chemical modications resulting from photooxidation are investigated by infrared and UV vis spectroscopies. Depth proling experiments were performed by AFM nanoindentation and by micro-FTIR. Those techniques give information on the stiffness, the adhesion and the concentration of oxidation products. The objective is then to correlate the chemical evolution to the changes of physical properties. 2. Literature Very few papers report on the photodegradation of polymer nanocomposites. It has been shown that the incorporation of layered silicates increases the rate of polycarbonate or polyethylene photodegradation [5,6]. Poly(p-phenylenevinylene)/layered silicate nanocomposites showed improved environmental stability against photodegradation in air [7]. For thermo- and/or photo-stabilized polypropylene formulations, the presence of organo-modied MT reduces the length of the induction period without changing the nature and the relative concentration of the photoproducts [810]. Concerning epoxies, little information concerns the degradation of bre/epoxy systems. Natural and articial ageing were shown to generate the formation of a thin photooxidation layer at the surface of epoxy resin [11]. The ablation of organic matrix was also observed in glass bre composites [12,13]. The comparison between natural and articial ageing proved that rainfall was a major parameter in amplifying the erosion phenomenon. The layer and the ablation of the organic matrix evolve spatially at a constant rate and despite an ablation of 1530 mm, photooxidation did not produce any mechanical weakening of the material [14]. Additionally, an NMR study revealed specic interactions between the glass bre and the epoxy resin. It is suggested that physisorbed or chemisorbed molecular water present at the surface of the glass bre assists in the opening of the epoxy ring by hydrogen bonding effect [15]. 3. Experimental 3.1. Materials The matrix was based on an epoxy resin (DGEBA from Vantico, n 0.15) and a Jeffamine D2000 hardener (Huntsman Chemical Co.). With stoichiometric conditions and optimized curing procedure, the glass transition temperature of this soft epoxy is around 50  C. The high molecular weight of the hardener allows achieving an elongation at break around 22% at room temperature. Synthesis of the organoclay and in situ polymerization conditions for the nanocomposite blends have already been described [16]. A commercial organophilic clay (Tixogel MP250) supplied from Societe Franaise des Bentonites et Derives was taken as reference organophilic clay. This clay is modied with dimethyl-benzyl-tallowalkyl-ammonium ions and its d-spacing is about 20 . All nanocomposite lling rates are expressed in grams of organoclay per 100 g of total resin (phr). The Tixogel contents in the polymer nanocomposite are 5 and 10 phr. The corresponding nanocomposites will be noted Tix5 and Tix10. Na-clay (Optigel EX0250) was obtained from Societe Franaise des Bentonites et Derives, based in Le Treport, France. Its cationic exchange capacity (CEC) is around 100 meq/100 g and its d-spacing is about 12.6 (hydrated clay). The amine used for the exchange was octadecylamine CH3(CH2)17NH2 from Aldrich. The content of the exchanged Optigel is 5 phr. The nanocomposite obtained will be called Opt5. Conventional silica ller was also used (Aerosil 200, from Degussa) with an average primary particle size of 12 nm. In the

blend, the Aerosil content is 5 phr. The nanocomposite will subsequently be named Sil5. 3.2. Irradiation Irradiations were carried out at wavelengths higher than 300 nm in the presence of oxygen. The device (SEPAP 12/24) is equipped with four medium pressure mercury lamps (400 W) and the samples are placed on a rotating carousel positioned in the centre. The temperature is regulated at 60  C and controlled by a Pt thermocouple. The device has been previously described [17]. 3.3. Characterization The chemical changes resulting from the oxidation reactions were followed by infrared spectrometry on Nicolet Magna-IR 860 equipped with an ATR accessory (4 cm1 and 64 scans, germanium crystal). The analyzed depth is wavelength dependent and is between 0.5 and 2.8 mm (3600700 cm1) [18]. Before proling, the oxidized samples (z3 mm thick) were microtomed in the irradiation direction into thin lms of z30 mm of thickness. Micro-FTIR experiments were performed on the Nicolet Magma-IR 760 spectrometer equipped with a Nicplan microscope. The spectra were recorded through a window of 10 100 mm, every 5 mm (4 cm1 and 512 scans). UVvis spectra were recorded on a Shimadzu UV2101 PC spectrometer equipped with an integrating sphere. 3.4. AFM analysis The measurement of nanomechanical properties involves acquisition of forcedistance curves with an atomic force microscope (AFM) [19,20]. The detection system measures the probe tip deection versus vertical displacement. The quantilever starts above the surface. The tip deection remains constant during the vertical displacement until the jump to contact with the surface (zero force, see Scheme 1). Just before tip-sample contact is made, the probe tip can be pulled down to the surface by attractive forces. After the contact, the deection signal increases due to repulsive forces between the tip and the sample: the tip indents the surface. During unloading, the cantilever deection decreases. Because the material response during unloading is elastic, parameters such as the stiffness, elastic modulus and hardness can be obtained from the elastic contact theory [21,22]. In the case of this study, we limited our experiments to qualitative measurements of stiffness deduced from the slope of the unloading curve. After unloading, the tip often adheres to the surface, causing a further decrease of the deection until the tip jumps out of contact (pull-off force required to cause detachment between the AFM tip and the sample surface). The difference between this minimal force and the zero force corresponds to the adhesive force Fad. AFM nanoindentation results were obtained with a Nanoscope IIIa from Veeco Instruments. Nanoindentations were performed

Scheme 1. Forcedistance curves during load/unload cycle.

1788

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792

with a tapping silicon tip RTESP-NCL from Veeco (given data: resonant frequency z150250 kHz, spring constant z2070 N/m and radius of curvature 510 nm). The same tip was used for all the indentations of this study. The surface and prole indentation procedures have been previously described [23]. After calibration of the tip on sapphire, for sample indentations, the maximum of deection was xed to 35 nm for a vertical displacement of 2000 nm. The reproducibility of the tests was checked by repeating the experiments several times with different samples. It is worthy to note that for MMt nanocomposites, the maximum analyzable stiffness has sometimes been reached when indentations were effected on lamellae. The corresponding slope values were then eliminated. After rejecting those values, the mean value on stiffness error (in a.u.) was estimated to be less than 5%.
Fig. 2. TEM picture of exfoliated regions in DGEBA/D2000 with (a) Tix5 and (b) Opt5.

4. Results and discussion 4.1. Morphological characterization In the rubbery epoxy/amine matrix, the states of dispersion of the two organoclays are completely different at the micrometer scale as can be seen in Fig. 1. In both cases, the dispersion is rather heterogeneous unlike the silica one. There are chemical interactions between organoclay and the reactive medium that govern morphologies since no shear work has been used. The Tixogel forms coarse aggregates up to 10 mm thick as can be seen in Fig. 1a whereas Opt5 (Fig. 1b) is dispersed more nely. Opt5 aggregates are smaller, more homogeneously dispersed in the matrix and one can as well distinguish very small aggregates barely visible at this scale level. However, these aggregates have roughly the same substructure. Whatever the organoclay used, the aggregates are made of diffuse light grey and darker zones, which look like some more or less dense parts of the aggregates. If a zoom is performed on the light grey zones, no kind of ordering at the nanometer scale is observed. This is veried with the two different organoclays as can be seen in Fig. 2. These zones look like exfoliated zones where the platelets are not parallel to each other any more. In contrast, the dark zone observables in the aggregates are intercalated zones. In these zones the platelets are clearly parallel to each other. The distances that can be measured on these TEM pictures correspond to the distances of 35 (for Tixogel) and 110 (for Optigel) measured by X-ray scattering for each one of the two organoclays. The conclusion is that morphologies obtained by chemical interactions (no additional shear) have to be considered in their whole and not only at the nanometer scale. Indeed the main differences one can observe in those nanocomposites are at the micrometer level with a better dispersion of the Opt5, whereas the inner structure of these aggregates is roughly the same with intercalated and exfoliated zones in both kinds of aggregates. Opt5 has, however, a better dispersion at the nanometer level as well since the d-spacing in its intercalated zones is much higher than the one of the Tixogel.

4.2. Stiffness analysis By using the tip of the AFM as an indenter, the load/unload response of the tip can easily be exploited to determine relative values of surface stiffness. Before irradiation, the surface stiffness shows that the gain due to the ller is rather low since the difference between the Sil5 and the DGEBA/Jeffamine samples is very small (Fig. 3). However, the increase of stiffness due to MMt is more signicant. Moreover, considering that in 5 phr of organoclay, 30% of the mass is represented by the organic modier, this increase is even more important compared to Sil5 that contains 5 phr of a 100% of inorganic ller. The difference between the two llers is principally due to their shapes and to their dispersion state in the network. The spherical silica does not increase the mechanical properties or the cohesion of the material [24]. In the case of MMt, beyond the intrinsic mechanical properties of the lamellae, this nanoller improves the nanocomposite cohesion, the lamellae playing a reinforcing role for the epoxy matrix [24,25].

4.3. Infrared analysis After initial characterization of the samples, photooxidation experiments were carried out. Fig. 4 shows the increase of carbonyl absorbance with irradiation time for the epoxy/amine alone (a) and for the sample containing 5 phr of MMt modied with dimethylbenzyl-tallowalkyl-ammonium ions (Tix5, b). The initial spectrum

Fig. 1. TEM picture of DGEBA/D2000 with (a) Tix5, (b) Opt5 and (c) Sil5.

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792

1789

Fig. 3. Initial stiffness of the DGEBA/Jeffamine and of the nanocomposites measured by AFM nanoindentation.

has been subtracted in order to directly visualize the formation of the photoproducts and to observe the polymer matrix changes. The comparison between lled and unlled samples shows that in terms of nature and stoichiometry, the photoproducts formed were identical. The photoproducts result from the degradation of DGEBA units and amine crosslinks. The photooxidation of the Jeffamine units represents a minor way of degradation. The maximum of absorbance at 1675 cm1 has been assigned to Ndisubstituted amides [18]. The only noticeable difference is that, in the nanocomposite, aromatic rings are only observed to change of substitution with a shift of the two bands at 1581 and 1607 cm1 to higher wavenumbers, in contrast with the unlled sample where a decrease of these bands was observed. For both samples, after 32 h of irradiation, the spectrum remains almost constant. The

spectra obtained for Opt5 and Tix10 were similar to the ones presented in Fig. 4(b). The infrared spectra of the Tix5 sample in the 13001000 cm1 region are represented in Fig. 5. The decrease at 1036 cm1 corresponds to the reaction of the phenoxy groups (leading to aromatic formates, 1739 cm1). The isopropylidene groups react during photooxidation (decrease at 1183 cm1) and mainly generate acetophenone (1685 cm1) and benzyl methyl ketone (1728 cm1). The isosbestic point at 1260 cm1 indicates a shift of the band at 1250 cm1 to higher wavenumbers. This shift is explained by a change of substitution of the aromatic ring. All these changes were previously observed in the non-lled sample [18]. However, interesting new phenomena were found. For long irradiation durations, increases at 1043 and 1078 cm1 were noted. In order to better visualize the formation of these new absorption bands, we have plotted in Fig. 6(a), the subtraction of the ATR spectra of sample Tix10 irradiated 32 h and the spectra of the same sample irradiated for longer durations (A). The transmission spectrum of a Tixogel pellet in KBr (B) was also added to this gure. The comparison between spectra A and B shows that the increase of absorbance after photooxidation of the Tix10 sample corresponds to an increase of the amount of ller in the analyzed thickness. Tixogel bands at 796, 884, 916, 1020, 1043, 1078, 1200 cm1 t with the ones of the sample. The increase of the nanoller content close to the surface with irradiation time has been also observed for Tix5 and Opt5 samples. In the Sil5 sample, an increase of a large band is observed around 1090 cm1 (Fig. 6(b)). This large band may be assigned to the SiO stretching vibration of the ller. This means that the concentration of silica increases even more signicantly close to the surface with the silica particles compared to clay lamellae. The surface enrichment with the nanoller during the rst 30 h of irradiation may result from an ablation of the organic matrix, with loss of small molecule oxidized products in the gas phase. The study of the degradation mechanism has shown that a lot of molecular products were formed [18,26]. An ablation of the organic phase has been found previously with epoxies containing glass bres [12] and on dimethacrylate networks synthesized by thermal free radical polymerization [27]. The increase of absorbance versus irradiation time has been measured at 1728 cm1, maximum attributed to a benzyl methyl ketone (Fig. 7) [18]. This ketone is one of the products resulting from the photooxidation of the DGEBA units. After 30 h of irradiation, the degradation rate slows down, essentially because the oxidation of the DGEBA units is maximal within the layer analyzed by ATRFTIR (inferior to 2 mm). The rate of oxidation is similar for all the samples. No signicant inuence of the ller could be detected in terms of nature of the photoproducts and of kinetics of

Fig. 4. ATR infrared spectra in the carbonyl region during photooxidation (after subtraction of the initial spectrum), (a) unlled sample, (b) Tix5 sample.

Fig. 5. ATR infrared spectra of a Tix5 sample in the 9701320 cm1 region during the photooxidation (after subtraction of the initial spectrum).

1790

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792

Fig. 8. Changes of the absorbance at 450 nm versus irradiation time.

amine sample is colourless whereas the composites are initially yellow. The higher is the initial discoloration, the lower is the increase of absorbance at 450 nm. After 30 h of irradiation, the same discoloration level is reached for all the samples as attested by the UVvis spectra. 4.5. Adhesion analysis Surface adhesion measurements were obtained by nanoindentation tests. After unloading, the tip adheres to the surface causing a negative deection of the tip until the force required to yield the detachment of the tip is reached. This negative deection characterizes the interactions between the tip and the sample. Before irradiation, the three samples presented similar tip/surface interactions (i.e. same area under the curve). The unloading curves shown in Fig. 9 were obtained by AFM nanoindentations on samples irradiated for 30 h. After irradiation, Tix5 and Sil5 nanocomposites present much lower interactions in comparison with DGEBA/Jeffamine. For the unlled sample, after 30 h of irradiation, the surface becomes tacky. The sticky products are highly oxidized products, resulting mainly from chain scissions [18]. The AFM adhesion tests showed that for the lled samples, the surface is less sticky. The reduced adhesion observed with the nanocomposites may be due to the fact that the sticky photoproducts adsorb at the surface of the polar nanollers, reducing the interaction parameters with the tip. The specic surface of lamellar MMt is higher than the one of silica due to its aspect ratio. Actually, MMt presents the fewest interactions with the tip. Another cause that should be invoked is that, in reason of a lower oxygen permeability due to the nanoller, the degradation involves more photolysis reactions (without oxygen), with formation of less photooxidised products. Photolysis experiments carried on the epoxy/amine sample generated a less sticky surface [28].

Fig. 6. (a) ATR infrared spectra of a Tix10 sample: A During the photooxidation (after subtraction of the spectrum of the lm irradiated 32 h); B Transmission spectrum of Tixogel in KBr, (b) ATR infrared spectra of an Sil5 sample during photooxidation (after subtraction of the initial spectrum).

degradation. No specic photoproduct resulting from photooxidation of the organic modier of MMt could be detected in the infrared spectra. 4.4. UVvis analysis Photooxidation generates an important yellowing of the epoxy/ amine sample, principally attributed to the degradation of the DGEBA units [18]. In absence or in presence of the ller, similar UV vis spectra were observed. The increase of absorbance measured at 450 nm versus irradiation time is rapid during the rst 30 h (Fig. 8). After this period, an invariant state is reached. The variation of absorbance is lower for the nanocomposite blends. However, the initial colour of the sample has to be taken into account. The epoxy/

Fig. 7. Changes of the absorbance at 1728 cm1 versus irradiation time.

Fig. 9. AFM nanoindentation curves of samples irradiated for 30 h (unloading part).

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792

1791

4.6. Photooxidation proles Depth proling experiments have been realized by micro-FTIR analysis after 30 h of irradiation. The results are reported in Fig. 10(a). Each point represents the absorbance at 1675 cm1 recorded through a 10 100 mm window and the spectra were recorded every 5 mm. For the three samples, an oxidation prole is observed in the rst 250300 mm. The unlled sample presents the highest concentration of N-disubstituted amide groups (1675 cm1). These photoproducts result from the oxidation of the amine crosslinks [18]. Generally speaking, in the sample thickness, the amount of photoproducts is lower for the samples containing silica (Sil5) and MMt (Tix5). This lower oxidation may be explained by the reduced oxygen permeability of the corresponding nanocomposites due to the presence of the nanollers [29,30]. This is especially true for the MMt nanocomposites in which oxygen pathway is considerably increased due to the aspect ratio of the lamellae. Additionally, during photooxidation, the oxygen barrier effect is amplied because of the accumulation of mineral matter at the surface as shown previously. In order to reveal the consequences of the evolution of the chemical structure on the mechanical properties of the material, we realized a second set of proles with the same samples irradiated 30 h and using the AFM nanoindentation technique. The photochemical mechanism of this crosslinked polymer is known to be governed by chain scission reactions and generates many oxidized products [18]. In a previous work, we showed that the photooxidation of the polymer matrix leads to a decrease of the stiffness from the core to the surface of the sample. After irradiation durations equal or higher than 30 h, the rst 25 mm of the surface correspond to a layer of maximal degradation [28]. To give evidence of an eventual role of the nanoller in the reinforcement of the polymer matrix, we measured the stiffness

prole after 30 h of photooxidation. The results are reported in Fig. 10(b). As previously, the curves indicate a stiffness decrease from the core to the surface. For the three samples, similar stiffness values were observed in the rst 25 mm. In that zone, the threedimensional structure of the polymer is lost and the ller does not provide any signicant reinforcement. A ller surface enrichment has been observed by ATRFTIR. However, the rst 5 mm could not be analyzed by nanoindentation, due to the excessive stickiness of the sample. The stiffness measured above 300 mm is constant and equal to the stiffness of the non-irradiated samples. It is also observed in this gure, that the samples present different depths of degradation, conversely to micro-FTIR analysis. The depth prole of the Sil5 sample is rather similar to that of the unlled resin; it is just slightly shifted to higher stiffness values. With MMt (Tix5), the stiffness is only affected until 80 mm in thickness, whereas with the sample containing the conventional ller (Sil5), the stiffness is modied until 150 mm. MMt appears to better preserve the samples mechanical properties in the chemically degraded zone. 5. Conclusions Polymer nanocomposites were prepared, characterized and submitted to ageing at l > 300 nm in the presence of oxygen. Based on a previous study of the photochemical behaviour of the DGEBA/ Jeffamine resin [18,28], the photodegradation analysis of the reinforced resins by silica or organo-modied montmorillonite has revealed several points: - the ller nature and content do not modify the chemical mechanism or the kinetics of degradation of the epoxy matrix; - the ller concentration increases close to the surface (a few microns) due to a loss of volatile photoproducts; - initially, the epoxy/amine sample is colourless and the composites are yellow. After 30 h of photooxidation the same discoloration level is reached; - owing to the llers, the surface of the nanocomposites is less sticky than that of the resin; - the presence of the llers reduces the depth of oxidation; - the improvement of the mechanical properties is more significant with MMt than with silica, initially and after photooxidation. The last three points are explained by the high aspect ratio of MMt, generating a larger contact area between the resin and the ller. References
[1] Kojima Y, Usuki A, Kawasumi M, Okada A, Kukushima Y, Karauchi T, et al. J Mater Res 1993;6:1185. [2] Kojima Y, Usuki A, Kawasumi M, Okada A, Kukushima Y, Karauchi T, et al. J Mater Res 1993;8:1179. [3] Kojima Y, Usuki A, Kawasumi M, Okada A, Kukushima Y, Karauchi T, et al. J Polym Sci Polym Chem 1993;31:983. [4] Le Pluart L, Duchet J, Sautereau H, Halley P, Gerard JF. Appl Clay Sci 2004;25:207. [5] Sloan JM, Patterson P, Hsieh A. Polym Mater Sci Eng 2003;88:354. [6] Qin H, Zhao C, Zhan S, Chen G, Yang M. Polym Degrad Stab 2003;81:497. [7] Lee HC, Lee TW, Lim YT, Park OO. Appl Clay Sci 2002;21:287. [8] Tidjani A, Wilkie CA. Polym Degrad Stab 2001;74:33. [9] Morlat S, Mailhot B, Gonzalez D, Gardette JL. Chem Mater 2004;16:377. [10] Morlat-Therias S, Mailhot B, Gonzalez D, Gardette JL. Chem Mater 2005;17:1072. [11] Monney L, Dubois C, Chambaudet A. Angew Makromol Chem 1999;273:6. [12] Monney L, Dubois C, Chambaudet A. Polym Degrad Stab 1997;56:357. [13] Monney L, Dubois C, Villa F, Chambaudet A. J Non-Cryst Solids 1999;26:208. [14] Monney L, Dubois C, Perreux D, Burtheret A, Chambaudet A. Polym Degrad Stab 1999;63:219. [15] Arvanitopoulos CD, Koenig JL. J Adhes 1995;53:15. [16] Le Pluart L, Duchet J, Sautereau H, Gerard JF. Macromol Symp 2003;1994:155. [17] Gauvin P, Lemaire J. Makromol Chem 1987;188:971.

Fig. 10. Photooxidation proles measured at 1675 cm1 by micro-FTIR (a) and by AFM nanoindentation (b) for an unlled sample (DGEBA/Jeffamine), a Sil5 sample and a Tix5 sample.

1792

B. Mailhot et al. / Polymer Degradation and Stability 93 (2008) 17861792 [24] [25] [26] [27] Du M, Zheng Q, Yang HM. J Soc Rheol Jap 2003;31(5):305. Yoon PJ, Hunter DL, Paul DR. Polymer 2003;44:5323. Monney L, Bole J, Dubois C, Chambaudet A. Polym Degrad Stab 1999;66:17. Rey L, Duchet J, Galy J, Sautereau H, Vouagnier D, Carrion L. Polymer 2002;43:4375. [28] Mailhot B, Morlat-Therias S, Ouahioune M, Gardette JL. Macromol Chem Phys 2005;206:585. [29] Gain O, Espuche E, Pollet E, Alexandre M, Dubois P. J Polym Sci Polym Phys 2005;43:205. [30] Jang-Kyo K, Chugang H, Ricky SCW, Man-Lung S. Comp Sci Techno 2005;65:805.

[18] Mailhot B, Morlat-Therias S, Ouahioune M, Gardette JL. Macromol Chem Phys 2005;206:575. [19] Tavares AC, Gulmine JV, Lepienski CM, Akcelrud L. Polym Degrad Stab 2003;81:367. [20] Blach JA, Watson GS, Buseld WK, Myhra S. Polym Int 2001;51:12. [21] Mareanukroh M, Eby RK, Scavuzzo RJ, Hamed GR. Rubber Chem Technol 2000;73:912. [22] Mailhot B, Komvopoulos K, Ward B, Tian Y, Somorjai GA. J Appl Phys 2001;89:5712. ` [23] Mailhot B, Bussiere PO, Rivaton A, Morlat-Therias S, Gardette JL. Macromol Rapid Commun 2004;25:436.