Mechanical characterisation of polymer of intrinsic microporosity PIM-1 for hydrogen storage applications

For the PIM-1 material synthesised in this study, the number-average molar mass (M _n) was determined to be 76,261 g/mol, and the weight-average molar mass (M _w) was 193,074 g/mol, giving a polydispersity index (PDI, M _n/M _w ) of 2.5. By comparing (Table 1) the weight-average molar masses for PIM-1 obtained in other studies, Budd et al. reported M _w = 270,000 g/mol [18], and PIM-1 mechanically tested by the group of Song and Du had M _w = 57,000 g/mol [19] and M _w = 85,000 g/mol [20], respectively, indicating we had successfully synthesised PIM-1 suitable for film formation and characterisation.

After casting and evaporation, the PIM-1 film was transparent and had a bright yellow colour as shown in Fig. 2. These films could be deformed to a high curvature (Fig. 2a). It was found that films with a thickness over 60 μm were liable to shrinkage and corrugation during the evaporation process, and this prevented the formation of flat samples, free from stress concentrations which would influence the tensile testing data. However, when the film thickness was below 20 μm, it had insufficient mechanical strength and was easily damaged during removal from the glass petri dish and handling. Therefore, it was determined that optimisation of the film thickness parameter was crucial for the formation of planar and defect-free thin films to form appropriate mechanical test specimens. Five films with varying solution volumes were produced and measured in terms of their thickness as shown in Table 2. It was found that the optimal thickness for obtaining good quality films with sufficient mechanical strength for sample preparation and testing was between 25 and 45 μm. The films prepared with different amounts of solution resulting with different thicknesses are shown in Fig. 3. It can be clearly observed that three thinnest films (12, 18 and 22.8 μm in thickness) followed the shape of the watch glass and deformed under their own weight. The 12-μm film was very fragile, and could not be straightened due to damage caused by handling. The film with a thickness of approximately 40 μm maintained horizontal alignment, despite the watch glass shape, due to increased stiffness. The thickest sample, with average thickness of 65 μm, was distorted and deformed upwards on evaporation of the chloroform.

The nitrogen isotherm for a powder sample of PIM-1 is shown in Fig. 4. This isotherm shares many qualities with those previously reported, namely the combination of IUPAC Type I and Type IVa behaviour, displaying strong porosity in both the microporous (pore diameter <2 nm) and mesoporous (pore diameter 2–50 nm) ranges [18, 22, 27]. This combination of behaviour is seen in the sharp rise in uptake in the low P/P ₀ range, followed by a more linear rise through the middle of the isotherm. This isotherm displays very large hysteresis in the desorption curve, which is commonly seen for PIM-1 isotherms, and is likely indicative either of ‘throats’ in the porosity (narrow channels that limit the passage of gas molecules), or pore swelling caused by the presence of the adsorbate [18, 22].

The BET surface area calculated for this material is 804.0 ± 2.4 m²/g, taking a P/P ₀ range of 0.01–0.14. This range was determined based on the criteria of Rouquerol et al. [26]. As a check for validity, the authors suggest calculating the relative pressure at which monolayer coverage is assumed to complete, which is calculated using the formula [28]:where n _m is the monolayer capacity and C is the BET parameter. As the calculated value for C in this pressure range is 349.9, the relative pressure at monolayer coverage is calculated to be 0.0507, which is well within the range of pressures selected and so validating this choice. This BET surface area value compares well with those reported in the literature, as surface area values for PIM-1 powders are typically reported in the range of 750–860 m²/g [9, 18].

The pore size distribution for PIM-1 powder, calculated using the non-linear density functional theory (NL-DFT) model provided by the MicroActive software, is shown in Fig. 5. The distribution over the nanopore range (≤100 nm) is very broad, clearly demonstrating the wide range of porosity suggested by the initial isotherm. There are significant peaks at 0.92, 1.63, 2.13 and 4.95 nm within both the microporous and mesoporous ranges. The cumulative pore size, a measure of the total available pore volume in pores up to a certain width, gives a microporous pore volume of 0.25 cm³/g, and a total nanopore volume of 0.43 cm³/g. The total pore volume as estimated by total adsorption at P/P ₀ = 0.999 is 0.583 cm³/g. These values compare reasonably with previously demonstrated values for both microporous and total pore volume in PIM-1 powder samples (total pore volume of Budd et al.’s sample was 0.68 cm³/g at P/P ₀ = 0.98, with a “significant proportion of micropores with dimensions in the range of 0.4–0.8 nm” [18]).

Full N₂ isotherms were not successfully completed on PIM-1 film samples, as it took an impractically long time for the volumetric system to equilibrate in the low-pressure ranges. It is believed this is due to a low rate of mass transfer, due in part to the low temperatures at which the isotherm is being performed, and the distorted, tortuous nature of the porosity within the film which adds length and complexity to the diffusion pathway for the adsorptive molecules.

The high-pressure hydrogen isotherm can be seen in Fig. 6. The adsorption curve for PIM-1 powder, seen in black, follows a very typical excess isotherm shape for supercritical adsorption, reaching 1.49 wt% at 1 MPa, and a maximum of 1.66 wt% at 3.2 MPa. Previous literature has only tested hydrogen adsorption of PIM-1 up to 2 MPa, reaching a maximum uptake of 1.45 wt% at 1 MPa [10]. The excess uptake then declines as the pressure increases due to a greater rate of density increase in the bulk phase than in the adsorbed phase [29].

The adsorption uptake in the PIM-1 thin film samples initially performs identically to the powder samples, reaching an uptake of 1.59 wt% at 2.9 MPa before the excess decreases. This lower maximum uptake and subsequent higher drop in the excess adsorbed with pressure is likely to be due to the reduced available pore volume in the films, which is caused by the full relaxation of the polymer chains during the solvent casting process, eliminating any mesoporosity formed in the powder during the reprecipitation and leaving only the intrinsic microporosity [18, 30].

In addition, the desorption of hydrogen shows an unusual trend. Whilst it also generally follows an excess curve, it is irreversible compared to the adsorption curve. Initially, the desorption curve rises above the adsorption curve in the high-pressure region, before decreasing and otherwise not showing the peak in uptake demonstrated in the adsorption curve. This desorption curve continues to be lower than the adsorption all the way through to the low pressures (<0.2 MPa).

Whilst it is unknown what causes this irreversibility, it is feasible that a change in the polymer structure is the cause, perhaps due to polymer swelling. This idea is in part suggested because the effect is reversible by degassing; repeat runs on the sample following removal of the adsorbate show identical adsorption/desorption curves. This is true for multiple attempts, with degassing performed between each run. This phenomenon may be further tested by the running of adsorption/desorption cycles without degassing in between; this work is currently in progress.

To our knowledge, very little hydrogen desorption data on PIMs is available in the literature, despite this behaviour being important in understanding how a material may behave in a hydrogen storage system. Only Budd et al. [22] have described hydrogen desorption data (in the limited pressure range up to 1.5 MPa), in which they describe the hydrogen uptakes shown as having “no significant hysteresis”. Understanding the desorption response is an important step in understanding the response of the material to the cyclic loading and unloading of hydrogen within a working storage tank.

A total of 20 PIM-1 samples were tensile tested. The measured range of sample thickness was between 43 and 143 μm with average of 76 ± 25 μm. All samples were tested until rupture and most failed in the specimen centre, whilst a small number of samples broke at two positions simultaneously (Fig. 7a). This behaviour may suggest an equal distribution of the mechanical properties along the membrane, as two distinct pieces of the sample exhibited equal strength. The average ultimate tensile stress was 30.9 ± 5.4 MPa with strains to failure reaching 4.4 ± 2.0%; the maximum observed extension at break was 3.4 mm. These values are close to the mechanical properties of bulk polystyrene, which reaches a tensile stress of 40–48 MPa and a failure strain of 6–7% [31]. However, it should be taken into account that the mechanical properties of films do vary from those measured for bulk polymers, for example, polystyrene thin films exhibit a yield strength of 88 MPa [32]. For our PIM-1 samples, the yield strength had an average of 11.0 ± 1.9 MPa and failure strains of typically 1%. The Young’s modulus obtained for all samples had uniform values, on average exceeding 1.2 GPa, with standard deviation of 130 MPa, which shows high reproducibility of results.

The values of ultimate tensile strength and failure strain obtained in this study were lower than those reported in the work of Song et al. [19] and Du et al. [20], whilst our samples exhibited a higher Young’s modulus (Table 3). The ultimate tensile strength was over 30% lower (30.9 MPa compared to 47.8 and 47.1, respectively), strain at rupture around 66% lower (4.4% compared to 10% and 11.2%, respectively), whereas the average modulus of elasticity was 27% higher when compared with Young’s modulus estimated from stress–strain curve provided by Song (1.26 GPa compared to 1.15 GPa) [19]. Both studies [19, 20] used the same solution concentration as in this work and chloroform as the solvent, but the number-average molar mass and weight-average molar mass were lower (see Table 1). We believe that the most significant effect influencing the reported differences in ultimate stress and strain values may be associated with evaporation process and its rate. Rapid solvent evaporation may cause formation of macropores and cavities in bulk structure of polymer (Figs. 9, 10) thus introducing inhomogeneities, which are reflected in decreased mechanical properties. However, macrostructural features of PIM-1 films have not been investigated in previous studies which prevents direct comparison of internal film structures.

The mechanical properties measured here should satisfy the strain requirements for hydrogen storage tank liners. Taking into account a significant hydrogen adsorption of the polymer which would decrease the pressure and strains present in a 70 MPa hydrogen tank, the obtained level of flexibility and strain to failure is sufficient for a PIM-1 film to perform successfully as the tank liner. The component present in a Type IV hydrogen storage tank with the lowest strain to failure is the carbon fibres with a failure strain between 0.4 and 1.9% [33], which is significantly below the average 4.4% obtained for PIM-1 films. The lowest extreme failure strain reported in this study was not below 2.1%. However, further structural optimisation of PIM-1 has to be performed before applying it as a tank liner which will certainly influence the liner mechanical properties.

Dynamic mechanical thermal analysis of PIM-1 thin films was performed in tension configuration. Six samples were cooled to a temperature below −150 °C using liquid nitrogen and dynamically tested with increasing temperature until eventual failure of the sample at high temperature, as exhibited by a sharp decrease in the reaction force produced by the sample as the response to a constant applied displacement. This transition occurred on average at 354 °C, when the samples were subject to significant degradation and changed from an intense yellow to a black colour and failed in a brittle manner. Until the point of sample failure at high temperature, there was no other rapid transition observed which confirmed the results obtained by Budd et al. where decomposition occurred at around 350 °C [18]. Two small peaks were observed in storage modulus curves at around −50 and 80 °C which might indicate a mild transition (i.e. glass transition). The second peak was also visible in the curve presented in [18]. However, this transition was insignificant to mechanical behaviour of the sample that might influence its performance in intended applications. Table 4 summarises the DMTA data and shows some variation of the modulus values for all samples. Two values describing the modulus are reported: the average modulus over the temperature range and the modulus measured at 20 °C (Table 4). In general, tan δ, which is a measure of energy dissipation in a material, was relatively low (0.05 ± 0.004) which means that stress and deformation were nearly in phase. This in turn suggests that the complex modulus E* is almost equivalent to the storage modulus E′, whereas the loss modulus E′′ is low which indicates that our samples did not manifest significant viscoelastic behaviour in the temperature range examined and can be considered almost purely elastic. The averaged complex and storage moduli was over 960 MPa with a standard deviation not exceeding 240 MPa. Even though the obtained values of storage modulus are close to those reported in the literature (1 GPa), it should be noted that another solvent (tetrahydrofuran) was used for the preparation of samples in the work of Budd et al. [18]. The influence of the solvent on polymer films’ mechanical properties has been reported for such thermoplastics as PLA [34] and polystyrene [35]; therefore, this may also be the case for PIM-1. It is also necessary to consider the influence of polymer treatment on the results, as polymers are susceptible to processing history and every step, from polymer preparation in granular form to diluting it in different solvents, casting and drying in different conditions, may be significant to a final outcome. However, the shape of the curves and the tendency of the moduli to decrease with increasing temperature are consistent with those presented previously [18]. A typical example of such a DMTA curve is shown in Fig. 8.

The intrinsic microporosity of PIM-1 has been measured in terms of accessible surface area and gas adsorption but micropore sizes (less than 2 nm [12]) are below the resolution of scanning electron microscopy (SEM). Examination of the failure surfaces of PIM-1 after being subjected to tensile testing has shown high roughness and a layered structure with distinct ‘bumps’, and the ruptured surface clearly distinguishes itself from the defocused upper surface visible in the background (Fig. 9a) which was flat and uniform. At higher magnification, cavity structures with small round holes in the middle of the features were observed which appeared to be a cone-like structure (Fig. 9b). The roughness of the surface was similar along the edge, sometimes with larger bumps and level changes as shown in Fig. 9a.

In order to see if SEM observations of the failure surface of the tensile test specimen are tensile damage specific and related to the failure mechanism, we prepared control samples, where cross-sections were obtained by cutting and tearing the sample. One feature of the surface that was clearly observed in all samples was the presence of holes with diameters in range of 0.1–1 μm. All samples exhibited a porosity in the macro scale along the cross-section surfaces; however, these macropores are too large to introduce effective additional surface area in the presence of intrinsic micropores. Nevertheless, they may be an efficient access route for hydrogen to the micropores at the nanoscale, allowing the hydrogen molecules to penetrate inwards the film and improve the hydrogen adsorption. The distribution of the macropores was not highly regular, due to rather chaotic nature of the evaporation process, but the pores were evident along the cross-section. It was more common for the pores to appear in the centre part of the film, rather than at the film edges (Fig. 10).

We also observed that the film surface is homogenous and does not exhibit the macroporosity in contrast to the cross-sectional surface (Fig. 9a, top of the picture) of the tensile test specimens. This may suggest that there exists a denser, less porous outer layer formed during the relaxation of polymer chains during solvent evaporation that hinders gas adsorption process and this may explain the lower adsorption rates of PIM-1 films when compared to PIM-1 powder (1.59 wt% vs. 1.66 wt%). The observed cavities may result from the evaporation process during film formation. Evaporation of chloroform is relatively rapid and leads to the formation of secondary structures at the macro-scale level. In order to examine this hypothesis, we performed microscopic imaging of an evaporating solution in real time. When a drop of the PIM-1 and chloroform solution was observed under a transmission optical microscope, the polymer solidification was seen to clearly progress from the sample edges towards the centre. During the process, it could be observed that a rough/corrugated structure was formed on the surface of the dry PIM-1 resembling wrinkles and creases (Fig. 11). Due to a limited possible depth of the imaged sample, we performed the experiments with a small amount of solution. This mechanism of progressing, rapid solidification could be also observed at the macro scale when forming large surface films; therefore, the creased characteristics might be extrapolated to the larger scale sample case. This behaviour most possibly explains the appearance of holes and cavities inside the film observed with SEM imaging. The presented structural analysis might be useful in determining the optimal conditions to prepare films with possibly highest surface-to-volume ratio by adjusting solution concentration and evaporation surface.