Title: Impact of confinement on the dynamics and H-bonding pattern in low- molecular weight poly(propylene glycols)

Herein, we explored thermal properties, dynamics, wettability, and Hbonding pattern in various poly(propylene glycols) (PPG) of Mn = 400 g/mol confined into two types of nanoporous templates: silica (d = 4 nm) and alumina (d = 18 nm). Unexpectedly, it was found that the mobility of the interfacial layer and the depression of the glass transition temperature weakly depend on the pore size, surface functionalization, and wettability. However, interestingly, we have reported strengthening of the hydrogen bonds in samples confined in silica pores. Further, the unique annealing experiments on PPG-OH with the use of Fourier transform infrared spectroscopy revealed the reorganization of oligomers close to the interface and the formation of three distinct fractions, interfacial, intermediate, and bulk-like, in the infiltrated samples. These experiments might shed new light on the variation of the segmental/structural relaxation times due to annealing of materials of different molecular weights infiltrated into pores or deposited in the form of a thin layer.


INTRODUCTION
The behavior of soft materials under nanoscale spatial restriction conditions became an attractive field of research over the past decades. 1,2 A special attention was focused on understanding in detail the molecular mechanisms governing the variation in the physicochemical properties of confined compounds. 3,4 Interestingly, as shown by numerous theoretical and experimental studies, these fluctuations arise mostly because of the competition of three major effects: surface interactions, finite size, and free volume. 5−8 In fact, most investigations carried out to date demonstrated that the deviation in the molecular dynamics and phase/glass transition temperatures of confined materials is directly correlated with the size reduction realized by the decrease in both thicknesses of thin films and the pore diameter, d, of porous templates. 9 What is important, the positronium annihilation measurements carried out for liquids infiltrated into porous templates revealed that along with the change in the diameter of the nanocavities, a strong fluctuation in the free volume is noted. 10−12 At this point, it is worth quoting recent work by White and Lipson, who derived the cooperative free volume (CFV) model to describe general temperature and free volume-dependent structural relaxation behavior 10−12 of the bulk and confined materials. Briefly, one can stress that in this model, a group of molecules cooperate to obtain enough space for a rearrangement, and hence, the number of cooperating particles is inversely proportional to the free volume. This model was applied by Napolitano et al. 13 to explain enhanced dynamics of poly(4-chlorostyrene) and poly (2-vinylpiridine), prepared in a form of thin layers on the substrates differing in roughness. Moreover, recently, the CFV approach was used to describe segmental and chain dynamics of amino-terminated poly(propylene glycol) (PPG-NH 2 ) confined into native and silanized silica pores. 6 In this context, it is worthwhile to stress that a weak/negligible variation of the free volume below temperature connected to the vitrification of the interfacial molecules served as a base to formulate hypothesis that negative pressure controls dynamics of liquids confined in pores. 8, 14−16 Along with the development of the concepts relying on the variation in the finite size and free volume, an increasing number of papers highlighted the impact of the surface effects, as a dominant factor governing behavior of confined liquids. Especially, an increasing attention was paid lately to the processes occurring at the interface, 17−19 surface interactions, 20,21 and roughness. 13,22 One can recall that there were successful attempts to predict the direction and the magnitude of the confinement effect for various glass formers both deposited as thin films 20,23,24 and infiltrated within alumina [anodic aluminum oxide (AAO)] membranes 21,25 based on the interfacial energy, γ SL . As shown, the higher γ SL , the greater deviation from the bulk behavior is expected. 21,25 Nevertheless, it should be noted that although finite size and surface effects on the dynamics of nanospatially confined liquids are very often considered separately, recent works indicated some kind of entanglement between them. In this context, it is worthwhile to remind recent atomic force microscopy measurements performed for glycerol incorporated within alumina membranes of d = 10 nm, 26 demonstrating that surface tension and most likely interfacial tension γ SL varies with the surface curvature of applied nanochannels. 27−29 Moreover, a close correlation between both effects is also well illustrated in the case of the dielectric studies on PPG infiltrated within controlled porous glasses of various pore diameters (d = 2.5−20 nm). 30−32 As shown, the glass transition temperature, T g , of infiltrated PPG revealed a nonlinear (parabolic) dependence, where the rise of T g noted for smaller pore diameter (d < 3 nm) was discussed in terms of the adsorption effects overcoming the confinement one. On the other hand, in the case of various PPG derivatives within silica pores characterized by a little larger pore sizes (d = 4−8 nm), 3,5,6 a linear reduction of T g with increasing degree of confinement was observed for both native and silanized templates. This might indicate that the impact of the surface effects should be often discussed together with the finite size (especially the pore curvatures) which might induce changes in the surface interactions and also should be taken into account to predict the behavior of the confined liquids. 33−36 In this article, we probe the counterbalance between the confinement and surface effects on the chemically modified poly(propylene glycols) (PPG) derivatives of molecular weight M n = 400 g/mol incorporated into silica (native and silanized) and alumina (AAO) membranes of various pore diameters (d = 4 nm and d = 18 nm) measured using dielectric and infrared (IR) spectroscopy as well as differential scanning calorimetry (DSC). These investigations allowed to get insight into the variation in wettability, dynamics, and H-bonded pattern in the spatially restricted samples infiltrated into both kind of membranes.

MATERIALS AND METHODS
2.1. Materials. Poly(propylene glycol) (PPG-OH) and poly(propylene glycol) bis(2-aminopropyl ether) (PPG-NH 2 ) of M n = 400 g/mol with purity higher than 98% were supplied by Sigma-Aldrich. The nanoporous alumina oxide membranes used in this study (supplied from InRedox) are composed of uniaxial channels (open from both sides) with well-defined pore diameter, d ∼ 18 ± 2 nm, thickness ∼ 50 ± 2 μm, and porosity ∼ 12 ± 2%. Details concerning pore density, distribution, and so forth can be found on the Webpage of the producer. 37 The preparation of native and functionalized silica templates are presented in the Supporting Information file. Finally, it should be mentioned that density of the confined materials is assumed to be approximately the same as the bulk material at room temperature (RT).
2.2. Methods. 2.2.1. Broadband Dielectric Spectroscopy (BDS). Isobaric measurements of the complex dielectric permittivity ε*(ω) = ε′(ω) − iε″ (ω) were carried out using the Novocontrol Alpha dielectric spectrometer over the frequency range from 10 −1 Hz to 10 6 Hz at ambient pressure. The temperature uncertainty controlled by Quatro Cryosystem using the nitrogen gas cryostat was better than 0.1 K. Dielectric measurements of native silica and silanized membranes (of d = 4 nm) filled with PPG-OH, PPG-NH 2 , and PPG-OCH 3 were placed between stainless steel plates having 5 mm diameter. Moreover, dielectric measurements on empty membranes were also carried out to evaluate their contribution which turned out to be negligible, to the measured loss spectra (see Figure S1 in the Supporting Information file).
2.2.2. Differential Scanning Calorimetry. Calorimetric measurements were carried out using a Mettler-Toledo DSC apparatus (Mettler-Toledo International, Inc., Greifensee, Switzerland) equipped with a liquid nitrogen cooling accessory and an HSS8 ceramic sensor. Temperature and enthalpy calibrations were investigated using indium and zinc standards, and the heat capacity, C p , calibration was performed using a sapphire disc. After placing crushed templates in the aluminum crucibles, they were sealed and measured over a wide temperature range with cooling and heating rates equal to 10 K/min. Each experiment was repeated three times. In addition, measurements on empty membranes were carried out (please see Figure S1 in the Supporting Information file). It was found that in the range of the studied temperature, there is no change in heat flow, indicating no contribution from the silica and alumina to the heat capacity jumps detected for the membranes filled with studied PPGs.
2.2.3. Fourier Transform Infrared Spectroscopy (FTIR). The Nicolet iS50 FTIR spectrometer (Thermo Scientific) was used to measure the FTIR spectra of the bulk and the confined PPGs samples. FTIR spectra were recorded in the 4000−1300 cm −1 frequency region with a spectral resolution of 4 cm −1 . The limited spectral range resulted from the detector saturation in the region of Si−O and Al−O stretching vibrations. Each spectrum was obtained by averaging 32 scans. The measurements were carried at RT (293 K) and glass transition temperature (T g ) determined from the BDS measurements. The low-temperature IR spectra were obtained by using a liquid nitrogen-cooled Linkam THMS 600 stage (the temperature accuracy of ±0.1°C), which was adapted to the Nicolet spectrometer. The measurements were performed at a cooling rate of 10°C/min in a nitrogen atmosphere. The time-dependent IR spectra were measured at equal intervals, that is, every 1 min after the temperature stabilization at T = 183 K. The FTIR spectra of native, silanized silica, and alumina membranes were measured and are presented in Figure S1. The −OH stretching band decomposition was performed using the MagicPlot 2.9.3 software (MagicPlot Systems, LCC). The band occurring between 3050 and 3950 cm −1 was decomposed with the use of several Gaussian functions adjusting the intensity and the width of the fitting curves. The two wavelength intervals at ca. 2400−2500 and 3050− 3950 cm −1 (excluding the spectral region of the CH stretching band) were used for fitting. All spectral parameters were left free during the fitting procedure. The "best fit" was considered when the statistical parameter R was the lowest. One can stress that although the amount of −OH groups on silica surface can vary with the surface preparation method, 38 the peak position connected to the vibration of this moiety is located at ν Si−OH ∼3748 cm −1 . Interestingly, at this region, we did not observe any contribution from the PPG infiltrated into pores. Thus, there is no need to consider the influence of Si−OH groups from the silica membrane during the deconvolution process. In the case of alumina templates, the maximum of the OH band is found at 3640 cm −1 in the range of weak H-bonded bulk-like PPG-OH molecules. However, the analysis of the OH vibration in this frequency regime does not affect a discussion Tensiometer, GmbH Germany. The description of the instrument and procedures has been presented previously. 39,40 The measuring procedure at T = 298.2 K for all substances has been repeated dozen or more times. The temperature measurement uncertainty was ±0.1 K. The precision of contact angle measurements was 0.01°, and the estimated uncertainty was ±1.5°, whereas the uncertainty of surface tension was ±0.1 mN·m −1 . Density, ρ, required for the surface tension experiment was measured with an Anton Paar DMA 5000M densimeter with the uncertainty not worse than 0.0001 g·cm −3 . For the surface energy estimation of native and silanized silica, some of the following liquids were considered: water, ethylene glycol, diiodomethane, and glycerol. The dispersive and nondispersive part in the surface tension for these substances were taken from ref 21

RESULTS AND DISCUSSION
Dielectric loss spectra of studied poly(propylene) glycols (PPG) terminated by three different groups, −OH, −NH 2 , and −OCH 3 , incorporated into native and silanized silica templates (of d = 4 nm) are shown in Figure 1. Note that dielectric spectra for bulk substances, taken from ref 41, are presented in the Supporting Information file. In all cases, dielectric data revealed the presence of the dc conductivity related to the charge transport and the segmental (α) relaxation at higher frequencies reflecting the cooperative motions of the molecules and responsible for the liquid-to-glass transition. Herein, one can stress that because of low molecular weight, M n < 1000 g/ mol, of studied PPG, 16,25 the additional mobility related to the fluctuations of the end-to-end vector of the chain ends called usually as the normal mode process cannot be observed (see Figure 1).
One can also mention that often for the materials confined within silica membranes, the appearance of the interfacial process, reflecting reorientational motions of the polymers adsorbed at the surface of the pore walls, are widely reported. Interestingly, this specific relaxation is not observed in the case of studied PPG, independently to the applied porous matrix and functionalization (see Figure 1). This finding agrees with the data published by Arndt et al., 1 who studied salol and glycerol, differing in the number of hydroxyl units, infiltrated in silica pores of d = 4−8 nm. They found that while a reorientation of the adsorbed molecules can be detected for the former system (salol with one −OH group), it is not visible for the latter alcohol (glycerol with three −OH moieties). It was proposed to link this experimental observation to the balance between timescales of the exchange process between The Journal of Physical Chemistry C pubs.acs.org/JPCC Article core and interfacial molecules and experiments. 1 Note that if the exchange between both fractions is either fast or slow with respect to the time of the experiments, the interfacial process can be detected or not in the loss spectra respectively. However, we suppose that this phenomenon might be also somehow related to the number of hydroxyl units within the sample. Once the number of this particular moieties is equal to unity, the interfacial process is well visible as a separate loss peak, while in the case of materials with two or more −OH groups (i.e., glycerol or PPG-OH), this mode vanishes. This pattern of behavior is similar to what was previously found in the case of bulk alcohols, where for monohydroxy alcohols, an additional Debye relaxation (related to the formation of the hydrogen bonding supramolecular structure) is observed; while in polyacohols, there is no trace of this kind of mobility in the collected dielectric spectra. 26 Nevertheless to explain if there is any relationship between the appearance of the Debye process in monohydroxyalcohols and interfacial process in liquids confined in pores, further studies are required. In Figure 2, α-loss peaks of the bulk and confined samples were superimposed at the constant segmental relaxation times, τ α , for all investigated compounds. As shown, the shape of the α-loss peaks significantly broadens with increasing confinement because of the increased dynamical heterogeneity induced by the additional interactions with the pore walls. 15,42 However, it should be pointed out that the α-loss peak of PPGs within the functionalized (silanized) silica templates is, interestingly, narrower than the one recorded for PPG infiltrated into native silica. In general, this might suggest a change in the surface interactions between hydrophobic (silanized) and hydrophilic (native) templates. Note that the broader α-loss peaks within native templates might also be a result of the dynamical perturbation introduced by stronger interactions between the substrate and adsorbed molecules. It is worth mentioning that similar observations were made for other low and high molecular weight samples infiltrated in pores. 1,25,31,42 To explore in more detail, the molecular dynamics of confined materials and collected dielectric spectra were analyzed using Havriliak−Negami (HN) function with the conductivity term 43   where α HN and β HN are the shape parameters representing the symmetric and asymmetric broadening of given relaxation peaks, Δε is the dielectric relaxation strength, τ HN is the HN relaxation time, ε 0 is the vacuum permittivity, and ϖ is an angular frequency (ϖ = 2πf). Note that τ α was estimated from τ HN accordingly to the equation given in ref 44. Determined segmental relaxation times were plotted as a function of inverse temperature and shown in Figure 3. As illustrated, the τ α (T)dependences of all confined PPG is a bulk-like at a hightemperature region. However, below the temperatures denoted as T g,interfacial , they start to deviate from the bulk behavior irrespectively of the sample and porous template. As widely reported, this phenomenon is related to the vitrification of the materials adsorbed to the pore walls. 1,6,15,45 In this context, one can mention about "two-layer" or "core−shell" models often used to discuss/interpret results of molecular dynamics simulations or quasi-elastic neutron scattering investigations. 46 In view of these simplified approaches, liquids infiltrated in pores are considered as consisted of the molecules located in the center of the pores ("core") and adsorbed to the walls ("interfacial"). They are characterized by different densities and mobilities because of additional interactions with the solid substrate. 1,60 Analysis of the data presented in Figure 3 unexpectedly revealed that the bifurcation of τ α (T)-dependences of the confined PPGs occurs at similar τ α (log τ α ∼ −5.5) independently to the terminal groups, applied porous template, and functionalization (silanized or native). It is worth adding that the deviation of τ α (T)-dependences of PPGs infiltrated into alumina templates also occurs at comparable τ α , the same as in the case of material confined in silica pores (see insets of Figure 3a,b). Note that data for PPG derivatives within AAO membranes of d = 18 nm were taken from ref 41. In this context, it should be mentioned that the similar finding has been recently reported for a primary and secondary monohydroxy alcohols incorporated within alumina and silica pores. It was shown that the temperature dependences of the Debye relaxation times, τ D (reflecting the mobility of supramolecular self-assemblies 47,48 ), deviate from the bulklike behavior at approximately the same τ D irrespectively of the porous template, chemical structure, and architecture of the supramolecular structures formed in the studied systems. 49 Moreover, recently Tu et al. 50 found the same scenario in ionic liquids confined in the native and functionalized alumina pores. This indicates that the change in the porous template, hydrophobicity, or hydrophilicity of the pore surface does not affect segmental dynamics of the interfacial layer to much because the bifurcation of τ α (T)-dependences of the core PPGs occurs at similar τ α (log τ α ∼ −5.5). Interestingly, this agrees with the molecular dynamics simulations showing that although functionalization of the pore walls influences on the dynamics of the interfacial molecules, this effect is not significant. 51 To estimate the glass transition temperature, T g , obtained, data presented in Figure 3 were fitted using the Vogel− Fulcher−Tamman (VFT) equation 52 where τ ∞ is the relaxation time at finite temperature, D T is the fragility parameter, and T 0 is the temperature, where τ goes to infinity. It should be mentioned that the two VFT functions were applied for the confined systems because of the observed deviation in the slope of τ α (T)-dependencies. The first one (high temperature VFT) was used only for an accurate determination of a point (temperature), at which the slope changes (related to the vitrification of the interfacial layer and denoted as T g,interfacial ), while, the glass transition temperatures of the confined samples (in this case of the core polymers, T g,core ) were estimated from the second, low-temperature VFT fits. The values of all calculated glass transition temperatures are listed in Table 1. Note that T g,core is defined as a temperature at which τ α = 100 s. As observed, irrespective of the terminal group and applied template, estimated values of T g,interfacial are comparable. Moreover, the same scenario can also be observed in the case of T g,core (within the experimental uncertainty). This implies that the performed surface modification (silanization) along with the variation in the porous matrix and pore diameter have for some reason a marginal impact on the segmental dynamics of infiltrated PPG derivatives. This observation seems to be quite surprising taking into account that to date, a simple silanization of silica, leading to a prominent change of the polarity, specific interactions (H-bonds), hydrophilicity, and hydrophobicity, generally induced difference in the behavior of materials incorporated within native and functionalized templates. 30 To better understand reported findings, we further measured contact angles, θ, and surface tensions, γ L , and calculated the interfacial tension, γ SL , prior and after surface modification. 55,56 The estimated values of contact angles and surface tensions for all studied materials and surfaces are listed in Table 2. As observed, the contact angles of investigated samples change significantly depending on their chemical structure and type of matrices. Both PPG-OH and PPG-NH 2 are characterized by comparable contact angles on native silica surface (θ ∼ 25°), which indicates good wettability. However, their θ increases because of surface modification even up to θ ∼ 37°for PPG-OH on the silanized silica surface. The increase of θ suggests, in fact, the reduction of wettability on the modified interface with respect to the native one, more likely because of the strong change in the interactions between host and guest materials. In this context, one can also add that in the case of the alumina surface, the contact angle of both PPG-OH and PPG-NH 2 is extremely low (θ ∼ 6°indicates they wet this surface perfectly). On the other hand, PPG-OCH 3 surprisingly seems to wet all surface in a similar manner as its contact angle remains comparable for all surfaces (θ ∼ 6−7°).
Next, we estimated interfacial energy, γ SL , according to the Young equation (γ SL = γ S − γ L cos θ, where γ S is the surface energy). 55 Calculated values of γ SL are listed in Table 2. Interestingly, for the silanized matrices, we noted low values of interfacial energies for all examined compounds. This indicates a clear change in the interfacial interactions between materials and functionalized silica in the vicinity of the pore walls, where the dispersive interactions prevail. 56,57 In this context, one can recall studies on the thin films that revealed that the higher value of γ SL , the greater deviation from the bulk behavior because of the reduced mobility at the interface. 20,23,24 This approach was further applied to the porous materials by Alexandris et al. 21 for several polymers infiltrated within AAO membranes. 21,58 As shown, the increase of the difference between the glass transition temperature, T g , of the bulk and confined samples, ΔT g , enlarges with the rise of the interfacial tension. This relationship was well quantified by the systematic measurements of the wettability, allowing us to calculate the interfacial energy, γ SL , and T g of the spatially restricted polymers. 21,25 Further studies indicated that higher γ SL implies reduced mobility of the interfacial layers that consequently leads to greater depression of T g,core in pores. 25 In Figure 4, we have plotted the estimated values of γ SL versus ΔT g,core and ΔT g,interfacial calculated for PPGs infiltrated into alumina (d = 18 nm) and silica (d = 4 nm) templates. Note that ΔT g,core is the difference between T g,core and T g of bulk (ΔT g,core = T g,core − T g ), while ΔT g,interfacial is the discrepancy between T g,interfacial and T g of bulk (ΔT g,interfacial = T g,interfacial − T g ). As shown in Figure 4, both ΔT g,core and ΔT g,interfacial of all investigated PPG incorporated in porous matrices are similar (within experimental uncertainty). Surprisingly, despite a clear variation in wettability, interfacial energy and pore size (d = 18 nm vs d = 4 nm), no differences between studied systems infiltrated within alumina and silica templates can be observed. In this context, one can remind that recent studies on various PPG derivatives infiltrated into AAO membranes revealed that γ SL weakly depends on both their terminal groups and molecular weight, M n , whereas ΔT g,interfacial changes with both these factors. 59 The finding discussed above clearly indicates that although the interfacial tension is a very useful parameter to predict depression of the glass transition temperature of the polymers infiltrated into porous media, it is not sufficient to understand the complex dynamics of such heterogeneous systems. Therefore, the contribution of other factors, possibly related to the variation in the density packing and roughness of the pore walls must be considered as well.
In the next step, we have performed additional DSC measurements to support/confirm results of dielectric investigations. Thermograms recorded for all studied PPG derivatives incorporated within silica templates are presented in Figure 5. As illustrated, all samples exhibit the presence of the two endothermic processes, related to the vitrification of the interfacial (denoted as T g,interfacial ) and "core" (labeled as T g,core ) molecules located above and below T g of the bulk material, respectively (so-called the double glass-transition phenomenon). 42,60,61 It should be pointed out that even for PPGs infiltrated into silanized silica templates characterized by the extremely low value of the interfacial energy, (γ SL ∼ 2 mN· m −1 ), double glass transition was detected. The value of T g,interfacial and T g,core obtained from DSC measurements for confined materials and also T g of the bulk samples were added to Table 1. Although, there are some discrepancies between the value of T g determined from calorimetric and dielectric measurements, which can be due to the difference in the  The Journal of Physical Chemistry C pubs.acs.org/JPCC Article heating/cooling rate applied in both methods 41 they are similar. Furthermore, we also determined the length scale of the interfacial layer, ξ, which also can be obtained directly from DSC measurements 60 where d is the pore diameter; ΔC p,core and ΔC p,interfacial are the changes of the heat capacity at T g,core and T g,interfacial . Note that the application of eq 3 requires the following assumptions: (i) the volume of the material in the surface layer is proportional to the step change of its heat capacity, (ii) the density of the incorporated material does not change along the pore radius, and (iii) the shape of the pore is cylindrical. The values of the heat capacity and calculated thickness of the interfacial layer are listed in Table 1 and Figure 5. As observed, the estimated ξ reaches similar values for all PPGs when infiltrated into native silica membranes (within experimental uncertainty), which are comparable to those ones reported earlier for 2E1H 8 or for monohydroxy alcohols, 49 where the value of ξ oscillated around ∼1 nm. However, after the silanization, ξ decreases for PPG-OH and PPG-NH 2 because of change in the interfacial interaction (i.e., suppression of H-bonds). In contrast, in the case of PPG-OCH 3 , the length scale of the interfacial layer increases after the surface modification, most likely due to increased surface-dispersive interactions. 56,57 One can add that according to the literature, ξ varies dependently to the type and strength of interactions, including hydrogen bonds. 1 It is also worthwhile to add that ξ of low-molecular-weight PPG incorporated within the alumina templates of d = 18−150 nm increases with the pore size but was relatively independent to the terminal end-groups of studied oligomers. In addition, we compared and plotted the thickness of the interfacial layer estimated for the PPG oligomer infiltrated into pores, made of silica and alumina having different pore sizes; please see the inset in Figure 5. Data for PPG infiltrated in AAO were taken from ref 41. This graph clearly illustrates that there is a linear relationship between thickness of the interfacial layer and pore diameter, d, which indicates some entanglement between both parameters. Moreover, it is worthwhile to stress that the interfacial layer estimated from calorimetry for the samples infiltrated into larger pores barely agree with those calculated for the thin films. In this particular case, ξ is around ∼2−3 nm and weakly depends on the film thickness. 62 Hence, one can state that the interfacial layer determined from calorimetry for the infiltrated systems is related to the length scale obeying molecules/polymers of much slower dynamics with respect to  The Journal of Physical Chemistry C pubs.acs.org/JPCC Article the core material because of the perturbation introduced by interactions with the substrate. Taking advantage of the fact that both glass transition temperatures were well visible in the thermograms collected for the PPG confined in silica and alumina pores, we decided to find out whether this material will behave in a similar way as entangled cis-1,4-polyisoprene (PI) infiltrated in AAO membranes. 58 Just to mention that in this particular case, Politidis et al. have demonstrated that T g,interfacial is conditional and can be detected only for the samples cooled down below T g,core ; this observation allowed them to hypothesize that T g,core is a spinodal temperature. Therefore, as a subsequent point of our studies, we have performed additional calorimetric measurements using several temperature protocols to investigate the behavior of low-molecular-weight PPG derivatives infiltrated into silica templates of d = 4 nm to check how does various thermal histories influence the existence of the double glass-transition phenomenon. For this purpose, we carried out three cooling scans followed by heating in accordance to the following protocols: (1) cooling down to 184 K (deep below T g,core ), (2) cooling to 174 K (between both detected T g s), and (3) again cooling down to 184 K. Representative DSC thermograms obtained for infiltrated PPG-OCH 3 are shown in Figure 6a. Interestingly, in contrast to the data reported in ref 58, a prominent T g,interfacial appears in all registered thermograms questioning assignments of the low glass transition temperature as the spinodal temperature. To explain the discrepancy between results reported herein and the ones presented in ref 58, one should consider (i) different molecular weights of studied polymers (we focused only on PPG of M n = 400 g/mol), (ii) various porous templates (characterized by different finite size and surface interactions), and (iii) significantly different wettabilities and interfacial tension of PI and PPG on the alumina and silica surfaces.
Furthermore, we also carried out a series of DSC measurements with different heating rates, 5−20 K/min. Selected thermograms recorded for PPG-OH incorporated within native and silanized templates of d = 4 nm are presented in Figure 6b,c. As expected, both T g s shift toward higher temperatures with the increasing heating rate. Additionally, we also observed that the length scale of the interfacial layer, ξ, increases with lowering the heating rate in the case of PPG-OH infiltrated into native silica templates ( Figure 6b); whereas for PPG-OH within the silanized silica templates, ξ seems to remain constant independently to the applied heating rate (Figure 6c). This simple experiments indicated that although the dynamics of the interfacial layer is not so much different in the vicinity of the functionalized pore walls, the length scale of the molecules adsorbed to the pore walls is affected in a more significant way.
As a final point of our investigations, we have carried out additional FTIR measurements to gain information about the H-bonded pattern in the samples confined within silica and alumina nanopores. Figure 7 shows the comparison of FTIR spectra of bulk PPG-OH, PPG-NH 2 , and samples infiltrated in alumina and silica (native and functionalized) templates at RT It should be noted that the FTIR spectral data, especially obtained for alumina pores, were difficult to interpret because of additional strong contribution of the stretching vibrations of the −OH groups of the pore surface to the measured spectra. An informative band for the analysis of hydrogen bond interactions is that connected to the X−H stretching vibrations of the proton donor groups (the ν X−H ). This spectral feature occurs in the range of 3700−3000 cm −1 in the FTIR spectra of the studied systems. The band observed between 3000 and 2800 cm −1 is responsible for the stretching vibrations of the C−H groups of the carbon skeleton. The position and frequency of the ν X−H bands for the bulk samples differ between PPG-OH and PPG-NH 2 (see Figure 7). At 293 K (room temperature, RT) ν O−H band of bulk PPG-OH occurs as a single broad peak located at 3453 cm −1 , whereas the ν N−H band of PPG-NH 2 consists of three peaks at 3369, 3298, and 3201 cm −1 . Thus, the H-bonds in both PPG derivatives are of medium strength. The different profiles of the ν X−H band of PPG-OH and PPG-NH 2 correspond to the various types of Hbonded aggregates. Note that it seems that the H-bonded oligomeric structures dominate in PPG-OH, while probably a more complex H-bonding network (the ring-or chain-like structures) can be formed in PPG-NH 2 . The spectral modifications of the X−H stretching bands of PPG-OH and PPG-NH 2 accompanying the temperature drop are consistent with the trend reported in the literature and discussed in the Supporting Information file.
In the next step, the FTIR approach was applied to investigate the effects of the nanoconfinement of PPG samples after their incorporation into different nanopore templates (silica and alumina). Representative spectra measured at different temperatures for the sample infiltrated in porous matrices are presented in Figures S4 and S5 in the Supporting Information file. The interaction of PPGs molecules within silica or alumina membranes causes noticeable changes in their IR spectra that represent the change in the hydrogen-bonded pattern in the investigated systems. At RT, the IR spectra of the confined liquids exhibit a significant redshift of the ν X−H peak frequency values, relative to the bulk samples (Figure 7). This spectral effect is associated with the existence of stronger H-bonds in PPGs under nanoconfinement. Similar results are observed in IR spectra measured after the temperature drop. In detail, the larger redshift of the O−H stretching vibrations for PPG-OH molecules is observed for the native pores (14 cm −1 ) compared to the silanized pores (11 cm −1 ) or the alumina ones(12 cm −1 ) at T g . In the case of PPG-NH 2 , the most intense peak of the ν N−H band at 3359 cm −1 is shifted by 4 cm −1 in native pores and 3 cm −1 in silanized pores relative to the bulk at T g . On the other hand, the ν N−H peak in alumina membranes shows the same position as that in the bulk. Thus, the PPG molecules within silanized silica pores exhibit the smallest spectral changes (i.e., the redshift of the ν X−H peak frequency value) relative to other systems. This is because of the weaker interactions between the host and guest material. It is also observed that the ν X−H bands measured for the confined samples are much broader than those measured for the bulk materials. This indicates that PPG molecules also exhibit greater variability in the size of the H-bonded aggregates in a confined environment. In order to address this issue more carefully, we performed additional analysis relying on the deconvolution of the spectra measured in the 3000−3800 cm −1 region for PPG-OH; please see Figures 8 and S6. This oligomer was selected and described in detail because it interacts the most with the native pores. To evaluate the contribution of specific components in this complex spectral regime, data collected for bulk and confined PPG-OH were fitted to the combination of several Gauss functions. The procedure of the fitting of the −OH stretching bands was . For the sample confined in pores, an additional Gaussian component is required to describe the FTIR spectra in the 3198−3026 cm −1 region. This additional band is assigned to the ν OH in the molecules adsorbed (AM) at the interface. The comparison of the deconvoluted IR spectra in the −OH stretching vibration region for bulk and confined PPG-OH at RT and at T g is shown in Figures 8 and S6 in the Supporting Information file. Figure S7 illustrates the temperature variations of the spectral parameters such as peak position and integrated areas obtained for each component (fraction of molecule) from the fitting IR spectra of bulk and confined PPG-OH to the superposition of several Gauss functions. All these parameters are also listed in Table S1. On the other hand in Table S2, the percentage areas of the deconvoluted profiles are shown, which give information on the different populations of H-bonded structural arrangements in PPG-OH. From the analysis of the results obtained at T g (Table S1), one finds a redshift of the maximum of the all fractions of H-bonded oligomers in confined sample with respect to the bulk material. This indicates the enhancement of the H-bonding interactions under confinement. As shown in Figure 8, the largest peak area in bulk sample corresponds to the fully H-bonded PPG-OH molecules. On the other hand, quite a large variety of adsorption behavior of PPG-OH on the different pore walls is observed (Figures 8 and S6). In the native pores, the adsorption process is the most dominant because the AM component has the largest percentage in the OH band profile in selected temperatures. The silanized membranes exhibit similar behavior at higher temperatures; however, below 263 K, partially H-bonded PPG-OH structures predominate. In the case of alumina templates, the AM component also dominates at 293K−233 K, whereas at low temperatures, the interactions between FBM are the strongest.
Next, the change in contribution of various Gaussian components to the overall spectrum upon temperature drops was monitored. As shown in Table S1, the positions of FBM and IMBM substructures are shifted to lower wavenumbers with decreasing temperature, indicating the strengthening of the interactions between "fully" and "partially" H-bonded PPG-OH molecules. On the contrary, the blueshift of the WBM components is observed because of the growth of the degree of association of PPG-OH at T g . Simultaneously, the integrated area of the FBM and IMBM components increases and the WBM area decreases as the temperature is lowered indicating the growing organization of PPG-OH toward a fully H-bonded network. However, it should be pointed out that the FBM contribution in confined PPG-OH within the silanized silica templates exhibits opposite temperature effect, that is, it decreases with the lowering temperature. More detailed analysis based on the percentage contribution of the Gaussian component areas shows that in bulk PPG-OH, the FBM-type molecules are the dominating population, their proportion steadily increases from 48 to 62% as the temperature decreases from RT to T g . Interestingly, in all-confined PPG-OH, the AM sub-band has the largest percentage at RT. As the temperature is lowered, the adsorption process is rather reduced in favor of the fully (alumina templates) and partially (silica and alumina templates) H-bonded interactions. The greatest effect of temperature on the adsorption process is observed for the alumina membrane in which the AM contribution decreases from 51 to 23%, while the FBM and IMBM populations increases 20−42 and 24−30%, respectively. Thus, the variations of the respective populations illustrate that for confined materials, the adsorbed PPG-OH molecules are always the dominating population at RT. At T g , the interfacial H-bonded species are prevailing only in the native silica pores.
As a final point of our investigations, we have performed the time-dependent FTIR measurements on the PPG-OH confined into silica and alumina membranes to verify whether the annealing at T g,core will influence on the H-bonding network. The analysis of these spectra in the ν O−H band region shows both changes in the shape of the band profile and its broadening as a function of time ( Figure 9). The peak located around 3400 cm −1 , originating from the bulk-like core The Journal of Physical Chemistry C pubs.acs.org/JPCC Article molecules, essentially shows no variation over time (the light blue area in Figure 9). Simultaneously the strong intensity growth of the sub-band at the lower wavenumber (approximately 3250 cm −1 ) in both kind of pores (indicated as a light yellow area in Figure 9) was clearly detected. Interestingly, this band has been assigned herein to the vibration of the hydroxyl moiety in FBM. Moreover, IR spectra recorded during annealing also revealed appearance of the sub peak at ν OH ∼ 3120 cm −1 connected to the adsorbed molecules. This band is probably because of the formation of the strongest H-bonds between PPG molecules and hydroxyl units attached to the silica and alumina templates (highlighted as light pink area in Figure 9). Growing intensity of the band observed at 3250 cm −1 during annealing can be assigned to the formation of an intermediate layer of the molecules located between core and interfacial ones. However, because of overlapping of this band with that originating from the fraction of FBM of the bulk material, it was not possible to detect it in the FTIR spectra measured at different temperatures. Moreover, analysis of the position of the new band at 3120 cm −1 indicated stronger interactions between the PPG-OH molecules and the silica surface with respect to the alumina. It is worthy to mention that this oligomer wets the alumina surface much better with respect to the silica one. Hence, it is a clear indication that enhanced wettability does not have to necessarily mean stronger interactions between host and guest material at least in the case of associating H-bonded liquids. Moreover, various strengths of the H-bonds between PPG and hydroxyl moieties attached to the silica and alumina pores suggest different chemical characters of this functional group that affects their different tendencies in formation of these specific interactions in both kinds of materials. It must be also stressed that above discussed results obtained for confined PPGs differ from those obtained for water or alcohols infiltrated in nanoporous templates. For primary alcohols incorporated to the native and silanized silica pores, we have found that the strength of H-bonds in confined samples was weaker compared to those in bulk systems which was manifested in the IR spectra as the blueshift of the ν O−H peak frequency. 49 In the case of water confined in zeolites, the O−H bands were slightly redshifted with respect to bulk water, indicating, that under confinement, molecules are relatively strongly hydrogen-bonded. 64 Moreover, the IR spectra of water confined in controlled pore glasses proved that this fluid is perturbed on very large scales (more than 10 nm), even in pores of greater diameter (d = 55 nm). The position of the connectivity band (∼150 cm −1 ) increased when the pore size decreased, suggesting stronger H-bonding interactions between neighboring water molecules. Additionally, an important decrease of the FWHM of the connectivity band was found for the spatially restricted sample (70%) which was related to a different orientation dynamics of water (up to 55 nm) as compared to bulk liquid. 65 Besides, Baum et al. 66 highlighted the predominant effect of the pore size, the kosmotropic properties and the surface ions excess on the dynamics, and the structure of water molecules in the pore and within the interfacial layer. The molecular dynamics simulations of the IR spectrum of isotopically dilute HOD in D 2 O in 2.4 nm hydrophilic, amorphous silica pores 67 showed that −OH groups are involved in weaker H-bonds to the silica oxygen acceptors than to water, leading to blueshifts in their frequencies, although this spectral effect was not observed in the measured IR spectra. This fact was explained by the smaller transition dipole moments of these −OH groups. Similar results were reported for HDO in H 2 O in Aerosol-OT reverse micelles of varying sizes. 68 Although, in this case, the ν O−D bands exhibited a significant blueshift relative to the bulk liquid, indicating a weakening of hydrogen bonds between the OD groups because of the presence of the interface between the water pool and the surfactant headgroups. 67,69 Herein, it is also worth to mention about work by Zanotti et al. for water molecules in Vycor (a hydrophilic porous silica glass). 70 They have found the O−H stretching sub-band at 3230 cm −1 that was interpreted as a result of the existence of a monolayer of water molecules H-bonded to the silanol (Si−OH) groups of the Vycor surface. The position of this peak revealed that the H-bonds in this system are significantly stronger than in bulk liquid water (around 3400 cm −1 ). A further more detailed understanding of the molecular interactions of water confined in mesoporous silica was presented by Knight et al. 71 The authors fit the O−H stretch region using three Gaussian curves, representing unique water populations, described as network water (NW), intermediate water (IW), and multimer water (MW). However, these studies also showed a systematic blueshift in the IR peak locations of NW, IW, and MW in confined water. The authors suspected that water in pores is congregating around surface hydroxyl groups to form islands of highly coordinated localized regions. Hence, results of Knight et al. are similar to those reported herein. 71 However, it is worthwhile to point out that FTIR measurements on PPG confined in either silica or alumina membranes clearly revealed that although there are at least three fractions of molecules differing in the H-bond pattern in these conditions, the strength of these specific interactions is much stronger with respect to the bulk materials.
At the first sight, this experimental finding questions interpretation of the calorimetric data suggesting that there are two fractions of molecules differing in mobility and glass transition temperature discussed in terms of the "two-layer" (or "core−shell") model. However, it must be stressed that a change in the H-bonded pattern does not have to influence dynamics of confined systems to the extent allowing registration of the third glass transition temperature related to the vitrification of the intermediate layer. In this context, one can recall papers devoted to polymer thin films discussing the occurrence of the third (intermediate) layer. 23 As reported, T g of ultrathin poly(methyl methacrylate) (PMMA) films depends on the type of applied substrates, that is, increasing for silicon oxide (polar surface) because of the hydrogen bonding and decreasing for the nonpolar surface, together with thickness reduction. It was observed that the dynamics and T g of each layer differ from the corresponding bulk substances. This indicates that the nature of the interaction between materials and interface is one of the dominant factors in determining the glass-transition temperatures that strongly depends on the thickness of the film and the interfacial energy between the polymer and the substrate. Additionally, the formation of the third layer between the adsorbed layer and core volume was investigated for nanopores. 72 Using calorimetric measurements, it was revealed the existence of three T g s for PMMA confined into AAO nanopores of d = 300 nm, where (i) molecules near interfaces (of T g higher than for the bulk), (ii) molecules interacting at the center of nanopores (T g lower than for the bulk), and (iii) fraction located between abovementioned (characterized by intermediate T g in a nonequilibrium state). Interestingly for PMMA restricted The Journal of Physical Chemistry C pubs.acs.org/JPCC Article into d = 80 nm of AAO, only double T g s was noted. 61 Generally, the formation of the third layer was considered as being strictly connected with the weakening of the interfacial effect between polymer matrices together with increasing pore diameter. 23,72 Besides, it was indicated that the existence of the third layer and also the shift of T g strictly depends on the thermal history (heating/cooling rate, aging/annealing procedure) of the samples and might be strictly correlated with the exchange effects between the adsorbed layer, interlayer (two layers are trapped in a non-equilibrium state), and the core volume. 23,72 Note that recently, the presence of the additional interlayer under confinement was also reported for poly-(methylphenylsiloxane) infiltrated into AAO templates, where interestingly, the third T g was also detected for small pore size, d = 18 nm. 73 One can also add that the resolving of the intermediate layer within the confined PPGs with time might, in fact, help us to understand better the reported recent shift of the segmental/ structural relaxation process of various incorporated materials upon the annealing experiments performed at following temperature conditions, T g,interfacial > T anneal > T g,core . 62,74,75 Note that at the studied range of temperatures, the examined systems are highly heterogeneous (the interfacial fraction of molecules is vitrified and the core ones are not). Briefly, as the annealing experiment proceeds, the shift of the α-relaxation peak toward lower frequencies was observed 74,75 resulting in completely different τ α (T)-dependences of confined materials with respect to the measurements performed prior to annealing (see Figure S8 in the Supporting Information file). As assumed, the density packing of core and interfacial molecules was out of equilibrium, which was recovered upon sample annealing. Consequently, the confined system moves from the one isochoric condition to the other, characterized by different densities and dynamics. 15,41,76 Nevertheless, as indicated by complementary FTIR measurements, the observed variation of structural/segmental dynamics might be also related to the variation of the H-bonding pattern or processes ongoing at the interface. It looks that during the annealing, there is a strong rearrangement of the interfacial layer leading to the formation of strong H-bonds between PPG and either silica or alumina pore walls. Moreover, the enhancement of these specific interactions is accompanied by the formation of the intermediate layer being in some distance from the pore walls. 23,41,72 Thus, in the annealed samples, we can clearly distinguish into three fractions, namely, interfacial, intermediate, and bulk-like fractions of molecules in the liquids infiltrated into pores.

CONCLUSIONS
In this work, we investigated the behavior of three various poly(propylene glycol) derivatives of M n = 400 g/mol characterized by different abilities to form H-bonds and incorporated into alumina (d = 18 nm) and silica (native and silanized of d = 4 nm) templates. These studies enabled us to explore the impact of finite size and surface interactions on the overall behavior of substances in spatially restricted systems. The observed deviation of segmental relaxation times from the bulk-like behavior indicates that independently on the pore diameter and applied templates, they deviate at similar τ α . Interestingly, we observed that although previous reports suggested that the shift in glass transition temperatures correlates with the interfacial energy and wettability, such relationship does not hold for PPGs infiltrated within alumina (d = 18 nm) and native and functionalized silica (d = 4 nm) templates. It indicates that besides the surface interactions and finite size (curvature), also other factors, that is, specific interactions, and density packing, possibly surface roughness, should also be taken into account when discussing the behavior of the confined liquid. Additionally, for the investigated PPGs, we observed that T g,int occurs even if T g,core has not been reached before, which indicates that T g,core is not a spinodal temperature. Finally, using FTIR spectroscopy, it was also revealed that the strength of hydrogen bond interactions between the incorporated material and interface differs dependently to the applied templates. Moreover, the timedependent IR spectra recorded during annealing around the T g,core revealed rearrangement in the interfacial layer leading to the formation of very strong H-bonds between PPG and alumina or silica pore walls. Consequently, this simple experiment allowed us to visualize and distinguish an intermediate layer (rarely reported in literature) between interfacial and bulk-like fraction of molecules. This completely new finding might shed new light and allow us to better understand the shift of the structural/segmental relaxation upon annealing below T g,interfacial . It seems that variation in the H-bonds and density packing between molecules attached to the interface and the intermediate ones is responsible for this phenomenon.
Preparation of porous silica templates, DSC thermograms, dielectric loss, and FTIR spectra of empty silica templates, nitrogen adsorption/desorption isotherms measured for native silica templates, dielectric loss spectra of bulk PPGs, dielectric loss spectra collected upon annealing of confined samples, FTIR spectra measured for bulk and confined PPGs, and their decomposition as well as temperature evolution (PDF)