【正文】 trometer equipped with a DTGSKBr detector and a ZnSe IRE crystal. The spectra were measured at 4cm1 resolution and the wavenumber range was 650–4000cm1. The filmcovered IRE crystal with a filter paper above the sample film was mounted in an ATR cell and the spectra of the dry film was collected as background spectra, then without moving the sample, distilled water was injected into the filter paper while starting the data acquisition by a macro program. A typical sorption loop lasts about 15 min and the acquisition time interval was 27 s. The timeresolved ATRFTIR spectra of sorbed water in two wavenumber ranges were obtained from subtraction spectra of the background and the wet sample films (in Fig. 2).. Twodimensional correlation analysisSeveral spectra at equal time intervals in a certain wavenumber range were selected for 2D correlation analysis using the software ‘‘2D Pocha’’ written by Daisuke Adachi (KwanseiGakuin University). In the 2D correlation maps unshaded regions indicate positive correlation intensities, while shaded regions indicate negative correlation intensities.3. Results and discussion. Diffusion kinetics of glassy epoxyIt is now widely believed that the kinetics of diffusion in polymers is associated with a concentration gradientdriven diffusion process, as well as a relaxationcontrolled process. Using a semiempirical eqn. (1) expressing the initial shape of sorption curvesin which Mt and M1 are the mass of water sorbed at time t and at equilibrium. Three classes of diffusion behavior are distinguished, for Fickian diffusion n＝, for Case II diffusion n≥1 and for anomalous diffusion n1. Alfrey et proposed that the rate of diffusion is much slower, much faster or more parable than the rate of segmental relaxation respectively for these three diffusion types. The mass of sorbed water was measured by observing the variation of the absorbance intensity of OH stretching vibration band located at 3200–3800 cm1 (in Fig. 2A). The integrated area was plotted as a function of time to obtain the initial sorption curves (in Fig. 3) and the parameter n was calculated from fitting to eqn. (1). As shown in Table 1 quite consistent with the gravimetric results that among four epoxy samples the diffusion behavior of EPA and EPP were much approximate to Fickian diffusion which means a diffusion controlled process where the diffusion rate is much slower than the polymer relaxation rate. while EPB displays a typical Case II characteristic and EP follows anomalous diffusion. Fickian diffusion often occurs in rubbery polymers possessing sufficient segment mobility to allow solvent penetration. It also monly occurs when the activity of solvent is sufficient low or the diffusion occurs only in the free volume of the polymer. Diffusion of small molecules in glassy polymers often displays various deviations from Fickian diffusion and several different models have been proposed to describe the nonFickian behavior. Some examples are the stressdependent, the historydependent and dualphase diffusion models. 24～27In EP networks, the strong hydrogen bonding formed by hydroxyl groups makes the less mobility of the whole network which cumbers the segment relaxation process, as the result, epoxy segment relaxation rate is decreased to be parable with water diffusion rate. Furthermore considering the strong hydrogen bonding interaction between water and hydrophilic groups in EP epoxy, the widely used Fickian law is less appearing since its assumption does not involve this interaction, which maybe the main reason of observed anomalous diffusion behavior. More interesting that EPB has the lowest Tg and largest chain mobility among four epoxy samples, 22however, its diffusion behavior displays a typical Case II feature. This might be explained by the ‘‘ shielding effect ’’ of side group. The end group — is hydrophobic and pliable so the accessible free volume and hydrophilic sites of water diffusion is sheltered and reduced by its rapid movement. This ‘‘ shielding effect ’’ causes the segment relaxation controlling the diffusion process.Diffusion coefficient was calculated from ATRFTIR spectra by a nonlinear curve fit assuming Fickian diffusion, eqn. (2),12 in this expression, At and A∞ is the band absorbance of ATRFIIR spectra at time t and at equilibrium respectively, L is the film thickness, the parameter can be defined from eqn. (3)28.where λ is the wavelength of infrared beam in the ATR element, θ is the incidence angle of radiation at the polymer/element interface, n1 is the refractive index of IRE element and n2 is the polymer refractive index. As shown in Table 1, the calculated diffusion coefficients of EPA and EPP from ATRFTIR spectra are consistent with gravimetric results and the data fits very well to the Fickian eqn. (2) with little deviation, while the diffusion coefficient of EPB cannot be calculated by using Fickian diffusion equation due to it being characteristic of Case II diffusion. The data of EP display a higher deviation from the Fickian equation pared with that of EPA and EPP, additionally, the diffusion coefficient of EP is much lower than that of EPA and EPP. The reason for that is to retard the diffusion of water throughout the whole network owing to the stronger hydrogen bonding between water molecules and the epoxy networks.. Epoxy–water interaction by 2DIR spectroscopyThe synchronous correlation spectra of EPP in the range of 2800–3700 cm1 are shown in Fig. 4A. In the 1D reference spectra at the top and side of the 2D correlation maps two main bands in the range of 3200–3600cm1 and 2900–3000cm1 are assigned to the water OH stretching vibration and bulk CH vibration respectively. An asynchronous cross peak develops only if the intensities of two dynamic bands vary out of phase, delayed or accelerated, with respect to each other. 17In the corresponding asynchronous correlation spectr