Pore structure and glass transition temperature of nanoporous poly(ether imide)


  • Tong Liu,

    1. Materials Science and Engineering Department and Rensselaer Nanotechnology Center,Rensselaer Polytechnic Institute, MRC-205, Troy, New York 12180
    Search for more papers by this author
  • Rahmi Ozisik,

    Corresponding author
    1. Materials Science and Engineering Department and Rensselaer Nanotechnology Center,Rensselaer Polytechnic Institute, MRC-205, Troy, New York 12180
    • Materials Science and Engineering Department and Rensselaer Nanotechnology Center,Rensselaer Polytechnic Institute, MRC-205, Troy, New York 12180
    Search for more papers by this author
  • Richard W. Siegel

    1. Materials Science and Engineering Department and Rensselaer Nanotechnology Center,Rensselaer Polytechnic Institute, MRC-205, Troy, New York 12180
    Search for more papers by this author


The effect of nanopores on the glass transition temperature (Tg) of poly(ether imide) was studied with differential scanning calorimetry. Nanoporous poly(ether imide) samples were obtained through the phase separation of immiscible blends of poly(ether imide) and polycaprolactone diol and by the removal of the dispersed minor phase domains with a selective solvent. Microscopy and statistical methods were used to characterize the pore structure and obtain the pore structure parameters. The pore size was found to depend on the processing time and the initial blend composition, mainly because of phase-coarsening kinetics. A decrease in Tg was observed in the nanoporous poly(ether imide) in comparison with the bulk samples. The change in Tg was strongly influenced by the pore structure and was explained by the percolation theory. © 2006 Wiley Periodicals, Inc. J Polym Sci Part B: Polym Phys 44: 3546–3552, 2006


As a unique branch of nanomaterials, nanoporous materials have received interest in recent years. Their specific properties lead to applications in a wide variety of fields, such as absorbents, catalysts, gas-separation materials, chemical/biological sensors, antireflection coatings, and low-dielectric-constant materials in microelectronic devices. There are different methods used to synthesize nanoporous materials, such as lithography and sol–gel chemistry. One of the simplest approaches is to create a two-phase mixture and remove the minor phase through physical or chemical means. Because of the high surface-area-to-volume ratio, nanoporous materials contain a large interfacial volume. This interfacial region has different properties than the bulk, such as density, crystallinity, and polymer chain mobility. The overall properties of the material reflect the combined effects of both the bulk and the interfacial regions, and depending on the amount of the interface present, the properties of the material can be altered drastically.

The glass transition temperature (Tg) is one of the most important properties of polymeric materials. It indicates the fundamental dynamics of polymer chains. Although glass transition behavior has been studied for a long time, the fundamental physics behind it still remains unsolved.1 Over the past 10 years, many researchers have focused on the effects of confinement on glass transition behavior, as deviations of Tg from the bulk value have been observed in various systems in confined geometries. A number of studies have concentrated on polymer thin films.2–11 It has been observed that Tg shifts occur when the film thickness decreases to ten to several hundred nanometers, and Tg depends strongly on both the film/substrate interface and the molecular weight of the polymer.3–8 Attempts have been made to interpret the confinement effects on Tg in polymer thin films from different aspects.2, 3, 7, 9–11 More recently, the glass transition behavior in polymer nanocomposites has been studied. The large filler/polymer interfacial area and surface-area-to-volume ratio in nanocomposites create a confined environment leading to changes in Tg. Similar Tg changes have been reported for different polymer nanocomposite systems.12–17 Both the sign and magnitude of the shift in Tg have been found to be influenced by the interaction between the filler and the polymer matrix.14–17 It has been suggested that the glass transition phenomenon observed in confined systems is due to the effects of the interfacial region. Many researchers have reported enhanced polymer chain mobility near a free surface.7, 9, 15–20 However, the understanding of the interface effects and Tg is still far from complete.21

This study focuses on the influence of the nanoporous structure on Tg. The article is arranged in the following order. The Experimental section includes the materials used, the spin-casting procedure, and the thermal and electron microscopy experiments performed. The Results and Discussion section includes the pore structure characterization and the Tg measurements. Finally, we present our conclusions.



Poly(ether imide) (PEI; Ultem 1000), with a weight-average molecular weight of 30,000 g/mol and a number-average molecular weight of 12,000 g/mol, was obtained from General Electric Plastics, Inc. Polycaprolactone diol (PCLD), with a number-average molecular weight of 503 g/mol, was purchased from Aldrich Corp. The densities of PEI and PCLD were 1.28 and 1.073 g/cm3, re spectively. The common solvent used in this study was dichloromethane (boiling temperature = 40 °C). Acetone was used to selectively remove PCLD from the PEI/PCLD blends to obtain the nanoporous structure.

Preparation of the Nanoporous PEI Films

PEI/PCLD blends were obtained through the mixing of PEI and PCLD with a magnetic stirrer in dichloromethane at room temperature until a transparent polymer blend solution was obtained. The total polymer content in the final solution was about 0.05 g/mL. PEI/PCLD blend films were obtained via spin coating (SCS P6700, Specialty Coating Systems, Inc.) at room temperature. The coating speed was kept at 500 rpm for all samples. Highly polished silicon wafers were used as substrates. The PEI/PCLD blend films were removed from the substrate with a water bath and were immersed in acetone for over 48 h to selectively remove PCLD to create a nanoporous structure. The nanoporous PEI films were dried at 40 °C in a vacuum oven for 48 h before further testing.

Pore Structure Characterization

The pore structure was characterized with a JEOL JSM-6330F field emission scanning electron microscope.22 Freestanding PEI films were embedded in an epoxy resin (Epon-812) via curing at 60 °C for 24 h. A Boeckeler PowerTome XL microtome was used to prepare cross sections of the embedding block. The sectioning direction was perpendicular to the film surface. Before scanning electron microscopy (SEM) analysis, the sample surfaces were sputter-coated with gold or platinum in an argon atmosphere. Image analysis methods were used to quantitatively determine the pore size and pore size distribution from 200 to 300 pores for each sample.23 Several SEM images (all at the same magnification) were analyzed for each sample. Because of the plastic deformation during the SEM sample preparation, the pores were elliptical in shape. The effective pore size was defined as the geometric average of the longest and shortest diameters of the elongated pores.

Statistical methods were applied to analyze the pore structure data because of the limited amount of data sampling available. A probability plot correlation coefficient hypothesis testing method was used to obtain the pore size distribution function.24, 25 The pore size data were first assumed to follow a theoretical distribution function. Then, the data were plotted against the theoretical distribution in such a way that the points formed approximately a straight line, which is the so-called probability plot. In this plot, the correlation coefficient associated with the linear fit to the data shows the quality of the fit. In practice, if the correlation coefficient is greater than 0.9, the distribution function is accepted to represent the data successfully.25 The distribution parameters can be estimated from the intercept and slope of the probability plot.

Thermal Characterization

Thermogravimetric analysis (TGA) was used to determine if the minor phase (PCLD) was completely removed after acetone immersion. TGA measurements were performed with a Mettler–Toledo TGA/SDTA 851e instrument in dried air. The operating temperature range was 25–1000 °C, and the heating rate was 10 °C/min.

The Tg measurements were carried out with a Mettler–Toledo DSC 822e differential scanning calorimeter. All measurements were performed between 25 and 250 °C at a heating rate of 10 °C/min. The Tg data were taken by straight lines being fitted to the differential scanning calorimetry curves before, during, and after the transition. The midpoint of the intersections was then defined as Tg and is reported in this study.



TGA was used to determine if the selective extraction of PCLD from the PEI/PCLD films was successful. TGA was performed on the pure components (PEI and PCLD) and on the PEI/PCLD films before and after acetone immersion. In Figure 1, the PCLD concentration in the PEI/PCLD blend is 20% (the blend compositions are reported as weight percentages). The onset-of-degradation temperatures (with 99% of the weight remaining) for pure PEI and PCLD were 430 and 170 °C, respectively. The PEI/PCLD blend showed both of these degradation temperatures: PCLD decomposed first, and PEI started to decompose at a temperature near the end of the PCLD decomposition. At the end of the PCLD decomposition, about 20% of the total weight was lost, which was the amount of PCLD present in the blend. The nanoporous PEI films did not show any weight loss within the 17–30 °C range, in which the PCLD decomposition was observed for the pure PCLD sample. The TGA thermograph of the nanoporous PEI was almost identical to that of the neat PEI. These results indicate that PCLD was successfully removed from the PEI/PCLD blend by acetone during the selective-solvent-extraction process.

Figure 1.

TGA plots of PCLD, PEI, a PCLD/PEI blend, and nanoporous PEI. The blend and nanoporous samples contained 20 wt % PCLD.

Pore Structure

The pore structure (average pore size and pore size distribution) was obtained through SEM imaging and statistical analysis. Figure 2 shows a typical SEM micrograph of the cross section of a nanoporous PEI film prepared from a PEI/PCLD blend with 10% PCLD. The pores were elongated along the microtome sectioning direction because of plastic deformation. For PEI (Ultem 1000), the tensile elongation at yield and the tensile elongation at fracture were 7 and 60%, respectively, and Poisson's ratio was 0.36. On the basis of these values, we could estimate that the maximum error of the effective pore size caused by the plastic deformation was about 11%. Pore sizes of up to 160 nm were observed with SEM. The arithmetic mean pore size was 64 nm for a PCLD concentration of 10%. The pore size distribution curve obtained by image analysis for this sample is shown in Figure 3, in which the number probability of pores is plotted as a function of the effective pore size. With statistical methods, the pore size data were found to follow a lognormal distribution function with a median pore size of 59 nm.26

Figure 2.

SEM micrograph of the cross section of a nanoporous PEI film obtained from a 10% PCLD/PEI blend.

Figure 3.

Pore size distribution of nanoporous PEI samples obtained from 10% PCLD/PEI blends.

Because the pore structure in nanoporous PEI films is developed through the phase separation of PEI/PCLD blends during spin coating, the size of the pores is directly related to the size of the dispersed PCLD domains in the immiscible polymer blend. Phase separation during spin coating is induced by solvent quenching. Because spin coating is a very rapid process, the slow dynamics of polymers make it hard to reach an equilibrium structure after solvent evaporation. In practice, the final domain size is the result of a finite time effect on molecular motion. It has been found that the late stage of domain growth during phase separation, which is called phase coarsening, is more relevant to the determination of the final domain size.27 During phase coarsening, the size of the domain increases with time, driven by the interfacial energy.28 The characteristic domain growth rate by phase coarsening has been found to follow a power law with time.29 This suggests that the formation of nanosized pores during spin coating is mainly due to the extremely short phase-separation time.

This result was further confirmed by a comparison of the final pore sizes created by two film processing methods: spin coating and casting. The initial PCLD concentration in the PEI/PCLD blend was kept at 10% for both methods. Compared with casting, spin coating is a much faster process because of the aggressive fluid expulsion by the rotational motion and the fast vapor-phase removal by the airflow field over the substrate.30, 31 It took approximately 5 s for the solvent to evaporate (and the film to form) during spin coating. However, film condensation took about 50 min in the casting method. It is obvious that the minor phase domains had more time to grow during solvent quenching in the casting process, and this is reflected in the median pore sizes obtained with these two processing methods. The median pore size of the final films was observed to be 628 nm in casting and 59 nm in spin coating with an initial PCLD concentration of 10%.

In addition to the processing time, we also investigated the influence of the polymer blend composition on the final pore size. Film samples with different initial PCLD concentrations were prepared via spin coating. Figure 4 shows the lognormal pore size distributions of nanoporous PEI films with initial PCLD concentrations of 3, 5, 8, 10, 13, 15, and 20%. The median pore sizes for these samples were 50, 54, 56, 59, 60, 69, and 75 nm, respectively. As discussed previously, the final pore size is directly related to the phase-separated minor domain size. The domain growth during phase coarsening can be viewed as a mass diffusion and coalescence process. As the concentration of the minor component (PCLD) in the polymer blend increases, a greater possibility occurs for the minor phase domains to coalesce and grow. Therefore, the average domain size increases as the PCLD content increases.

Figure 4.

Lognormal pore size distributions of nanoporous PEI samples with various initial PCLD contents.


To study the glass transition behavior of nanoporous PEI films, samples with different pore volume fractions were created by the variation of the initial PCLD content in the PEI/PCLD blend. The pore volume fraction was estimated from the PCLD weight content in the PEI/PCLD blend as follows:

equation image(1)

where φ is the pore volume fraction; mPCLD is the weight fraction of PCLD in the PEI/PCLD blend; and ρPCLD and ρPEI are the densities of PCLD and PEI, respectively. Figure 5 shows Tg of nanoporous PEI samples as a function of the pore volume fraction. A 22 °C reduction in Tg was observed with an increasing pore volume fraction. Changes in Tg have been observed in other confined systems, such as thin films and polymer nanocomposites with nonwetting interfaces, for which the area of the interfacial region is very large. Ash et al.14, 15 showed a dramatic decrease in Tg in alumina-filled poly(methyl methacrylate) nanocomposites. It was suggested that the nonwetting interfacial region surrounding the nanoparticles (which exhibits enhanced polymer chain mobility compared with the bulk polymer) leads to the change in Tg. Enhanced polymer chain mobility near a free surface has been observed in both experimental and computer simulation studies and has been suggested to be mainly due to the segregation of chain ends at the free surface.18–20, 32, 33 The impact of the interfacial region is not yet fully understood.21 The polymer affected by the interfacial region was first hypothesized to be on the order of several nanometers, however, it was found that the interfacial region may extend hundreds of nanometers into the bulk region.15 Recently, it has been observed that a significant depression of Tg near a polymer thin-film surface occurs as deep as several tens of nanometers, up to 100 nm, into the bulk region.34–36

Figure 5.

Tg of nanoporous PEI as a function of the pore volume fraction.

It is crucial to understand the mechanisms of why the existence of the high-mobility interfacial region might affect the overall polymer chain dynamics. The glass transition behavior was ex plained with a percolation mechanism by Hunt37–40 in the early 1990s. Long and Lequeux11 proposed a percolation mechanism, which explains the glass transition behavior in polymer thin films. More recently, Ash et al.15 extended Long and Lequeux's model to explain the behavior of Tg in polymer nanocomposites. The nanoporous PEI system studied here is very similar to a polymer nanocomposite system with a nonwetting interface. In the nanoporous system, an air–polymer interfacial region is generated between the pores and the polymer matrix, which can be considered a nonwetting interface. According to the percolation mechanism,11, 15 the polymer sample is composed of domains of fast and slow dynamics, which are each approximately 2 nm in size. The population of slow domains is controlled by the sample temperature. Upon cooling, the glass transition occurs because of the percolation of domains of slow dynamics by thermally induced density fluctuations. Even though domains of fast dynamics still exist in the system, they do not contribute to the overall viscosity, which is determined by the network of slow domains.11 In nanoporous polymers, a free-surface interfacial region that surrounds each nanopore results in high polymer chain mobility. These highly mobile regions are dispersed throughout the polymer matrix, and they contribute to the disruption of the percolation of the slow domains when the system is cooled from the polymer melt. In confined systems, such as nanoporous polymers and polymer thin films, the interfacial regions are sufficiently close together that the percolation of the slow domains is restrained by these interfacial regions, and this causes the glass transition to occur at a lower temperature compared with that of the bulk polymer.

The percolation mechanism provides an insight into the observed glass transition phenomenon in confined systems. In addition to the existence of the high-mobility interfacial region, the dynamic interaction between the interfacial regions is also necessary to cause a reduction in Tg. Recent work performed by Ellison and Torkelson9 and Bansal et al.17 provides more evidence to demonstrate the importance of the dynamic communication of interfacial regions in inducing the confinement effects on Tg. The study of the glass transition behavior in nanoporous polymers provides further understanding of the confinement effects on Tg and is useful for determining the nature of the glass transition and the impact of the interfacial region on the bulk properties of the polymer.


Nanoporous PEI was developed on the basis of the phase separation of PEI and PCLD blends during spin coating. After phase separation, the dispersed minor phase domains were removed by a selective solvent to form the nanoporous structure. SEM and statistical methods were used to characterize the pore structure and obtain the pore structure parameters. The pore size was directly related to the kinetics of the phase separation. The final pore size was found to be tens of nanometers in diameter and formed mainly because of the short-time phase-separation process during spin coating. The composition of the blend also affected the final pore size. With an increase in the PCLD content, the pore size was found to increase because of the greater probability of minor phase domain coalescence.

Tg of nanoporous PEI decreased from its bulk value with the addition of nanopores. This phenomenon is in agreement with glass transition observations in other confined systems. The presence of nanopores creates a large volume of the interfacial region that exhibits enhanced polymer chain mobility. These dispersed interfacial regions interrupt the percolation of domains with slow dynamics, and this leads to the observed decrease in Tg. In addition to the existence of the individual interfacial regions, the dynamic interaction between the interfacial regions is also important in determining the overall polymer chain dynamics in the system.


This work was supported by IBM and the Nanoscale Science and Engineering Initiative of the National Science Foundation under award number DMR-0117792.