Different polymeric membranes are candidates for polymer electrolyte fuel cells (PEFC) Page et al. (2012); Wang et al. (2011). High temperature polymer electrolyte fuel cells (HT-PEFC) operating at elevated temperatures between 100°C and 200°C (mainly in the range 140°C–180°C) are an interesting class of PEFCs with advantages for certain applications Kreuer (2013); Asensio et al. (2004). The most common material studied for low temperature polymer electrolyte fuel cells (LT-PEFC), operating at temperatures below 100°C is Nafion^® Page et al. (2012); Schmidt-Rohr and Chen (2008); Moilanen et al. (2008). Operating temperatures above 100°C in HT-PEFCs simplify significantly the water management during operation, since water is mainly present in the vapour phase. The CO tolerance is also better in HT-PEFCs compared to the LT-PEFC counterparts. The typical application range is at medium power (in the order of 5 kW) Liu et al. (2016). At the heart of a HT-PEFC is the proton conducting polyelectrolyte membrane between anode and cathode of the cell. The properties in terms of ion conduction and chemical stability are key for the performance of a HT-PEFC and are widely studied Kreuer (2013); Li et al. (2009); Wannek et al. (2009a). The operation temperatures of HT-PEFC above 100°C in acidic environment require polymerswith excellent thermal and chemical stability, one promising material being poly (2,5-benzimidazole)-based (PBI) membranes Wannek et al. (2009b). The aromatic backbone is the reason for its good thermal and chemical stability and its high glass transition temperature Kreuer (2013); Di et al. (2012). Due to the basic nature of PBI it can be doped with large amounts of phosphoric acid (PA) as an electrolyte, which has a high proton conductivity Vilciauskas et al. (2012) and provides in combination with the scaffolding PBI material a well-performing proton conducting membrane Maier et al. (2012). Conductivity data of phosphoric acid and PBI are presented in Ref. Korte et al. (2016). A recent review on PBI based membranes can be found in Ref. Aili et al. (2020).

The structure of the polymer membrane in PEFCs is an important parameter for the thermal stability and ageing of the membrane under the harsh operating conditions, and also has an influence on the conductivity due to the confinement of the electrolyte inside the porous membrane. Many studies of Nafion-based perfluorosulfonic acid (PFSA) membranes for LT-PEFCs use small angle scattering (SANS) for structural investigations Schmidt-Rohr and Chen (2008); Moilanen et al. (2008); Haubold et al. (2001); Elliot et al. (2000). Molecular dynamics (MD) simulations of a hydrated perfluoro sulfonyl imide (PFSI) model system address the question how nanostructure and proton transport are related Idupulapati et al. (2011). The relationship between structure and transport for similar systems was investigated recently in details Berrod et al. (2015). X-ray radiography Maier et al. (2012) and tomography Eberhardt et al. (2015) as well as neutron imaging Boillat et al. (2014); Arlt et al. (2015) served as tools to study HT-PEFC fuel cells or different part of its core component, the membrane electrode assembly (MEA). Structural length scales of a similar ABPBI membrane were investigated on a broad range of length scales with microscopic and scattering techniques Ivanova et al. (2017). The swelling of the PBI membrane upon doping with an alkali ion conductor has been investigated with SANS and prompt gamma activation analysis Babcock et al. (2019). Phosphoric acid diffusion in the swollen PBI membrane has been measured with quasielastic neutron scattering and analyzed by a fractal diffusion model Hopfenmüller et al. (2018). Aili et al., stated that the degree of crystallinity of PBI membranes increases under heating, which results in creating highly proton conducting pathways through the membrane cross-section by forcing phosphoric acid into the amorphous regions Aili et al. (2012). However, details of structural changes and its interpretation are lacking. This report aims of shedding light on the mesoscopic structural changes of the PBI membrane demonstrating irreversibility of these changes under thermal cycles from room temperature to typical operating temperatures of HT-PEFCs with neutron scattering and diffraction as well as with transmission electron microscopic techniques.

The temperature dependence of the structure of PBI membranes doped with D₃PO₄ has been studied with SANS on nanometer scale prior to heating to the typical operation temperature of HT-PEFCs and and at elevated temperatures in the range of operating temperatures. Figure 1 (left) shows two curves of the small angle scattering data for the initial membrane (top) measured at a temperature of 80°C (which is identical to the measurement at lower temperature), and the same membrane heated for the first time to the 140°C (bottom). Any further cooling and heating of the membrane resulted in no further change. The structures developed during the first heating cycle and discussed in more detail below remained also after cooling down and in all further heating cycles without additional changes.

The SANS intensity resulting from the measurement of PBI membrane doped with dPA at 80°C is fitted with a model combining the power law decay and the Teubner-Strey description for the broad shoulder at larger Q: I TS, LT Q = I TS Q + I 1 Q − p + b g r (3)

The resulting parameters for the low temperature phase were Q ₀ = 0.03 ± 0.001 Å⁻¹, d = 2π/Q ₀ = 20.9 ± 0.5 nm and ξ = 0.8 ± 0.1 nm.

The power law decay of the second term with I ₁ Q ^−p describes the large scale fractal structure of the membrane at small Q-values. It is then followed by the broad shoulder, described with Eq. 2, which finally results in a Q ⁻⁴ decay at large Q. The broad structure reflects the network of PBI strands in the swollen membrane.

As one can see from Figure 1 (left, bottom), for the measurement at 140°C the shoulder at large Q developed into a much more pronounced correlation peak. This has been accounted for by a second contribution of the form of Eq. 2. The high temperature SANS intensity I _TS,HT(Q) is fitted with the following function: I TS, HT Q = I TS Q + I TS,LT Q (4) which results in Q _0,HT = 0.073 ± 0.001 Å⁻¹, d _HT = 2π/Q _0,HT = 8.6 ± 0.1 nm and ξ _HT = 3.2 ± 0.3 nm.

The second Teubner-Strey contribution indicates that an additional structure with a characteristic average distance between the emerged objects, d _HT , has been created.

The scattering invariant gives a hint on the volume fraction occupied by one phase in a two phase system. If we take only the fitted component from I _TS(Q) and integrate it over Q, this leads to an indication of the volume fraction of the part of additionally emerged structure after the first heating cycle. Here, we find that the resulting volume fraction of the additional phase after heating is Φ = 0.04. We assumed here that the difference in scattering length density between the amorphous phase and the crystalline phase of PBI is approximately 0.5 × 10⁻⁶ Å⁻².

Crystalline polymers are typically denser than amorphous ones Kholodovych and Welsh (2007), thus we assume here an increase in density by a factor typical for polymers of about × 1.16 resulting in the estimate of the scattering length densities above. From the weight measurements during sample preparation the ratio between polymer and phosphoric acid is approximately 1:4, i.e., one may assume that only 20% of the sample consists of polymer. The part exhibiting a clear correlation peak from the Teubner-Strey analysis of the heated sample however can be attributed to 4% only, which we interprete as the formation of small crystalline PBI domains within the strands of amorphous PBI matrix. Following this estimation with the crystal volume fraction 0.04 and the polymer fraction 0.17, we have Φ_polymer − Φ_cryst = 0.13 left for amorphous polymer parts connecting and surrounding the crystals. The corresponding size qua diameter of the crystals is similar to the correlation length 8.6 nm from the Teubner-Strey analysis. Assuming spherical particles with r = ξ/2, this leads to a number of crystalline particles of N = 0.04/V _sphere = 2.3 × 10²¹ particles per cubic centimeter. With an average distance between the crystals of d = 8.6 nm, the maximum thickness of the connecting amorphous polymer strands can then be estimated by distributing the remaining 16% of polymer volume to thin cylinders, resulting in N V PBIcyl = N ( π R PBIcyl 2 d ) = 0.16 and therefore R _PBIcyl = (0.16/(Nπd))^1/2 = 1.6 nm with the radius of the polymer strand R _PBIcyl. The diameter of a polymer strand connecting the crystalline clusters is then approximately 2R _PBIcyl = 3.2 nm. Since the polymer strands obtained from the geometrical considerations above are smaller than the crystalline parts, a necklace structure where the crystal “beads” are connected with amorphous strands in between them is suggested.

The Porod part of the scattering (I ₁ Q ^−p in Eq. 3 with p = 3.2 and 3.6 for the initial and tempered membrane) can be connected to large scale inhomogeneities. It can be seen in a context of gels where large scale inhomogeneities are described by a Debye-Bueche scattering function (although we do not see the Guinier region here). It is characterized by a typical length scale of the inhomogeneity, R, and can be obtained from the slope at low Q of the SANS data. Since I (Q) of SANS is measured in absolute units it can provide the relevant length scale. Just from the Porod behaviour by taking the lowest-Q data point and assuming an ideal Q ⁻⁴ dependence, the following relation is used: I Q small = 2 π ϕ Δ ρ 2 3 R 1 Q small 4 (5)

For the inhomogeneities we assume a contrast obtained from crystalline vs. amorphous PBI. For the concentration of crystalline PBI we assume ϕ = 0.2 × 0.04 from considerations above. So we arrive at inhomogeneities of the size of either approx. 2 μm or 400 nm before and after the first heating to elevated temperatures. The latter value also seems to be motivated by the large scale brightness change of Figure 2.

The same effect of crystallite formation upon first heating of the membrane has been revealed at higher Q-values (sub-nm distances in real space) in the diffraction experiment. As one can see in Figure 1 (top right), the broad diffuse pattern at 1.6 Å⁻¹ corresponding to 3.9 Å transforms upon heating, thus leading to the formation of a number of relatively well-defined peaks at the following Q-values: 0.5, 1, 1.25 and 1.75 Å⁻¹. On length scales of characteristic molecular distances this indicates the higher degree of morphological ordering. This sharpening of the Bragg peaks in the sample after the first heating to 140°C reflects the formation of small PBI crystals, which are also responsible for the additional contrast in the SANS experiment with the pronounced additional correlation peak from the average distance between those semi-crystal particles. A previously heated sample doped with a higher amount of phosphoric acid is shown in Figure 1 bottom right. It demonstrates the irreversibility of the produced semi-crystalline structures. One may notice the different ratio of the coherent and incoherent contributions for these samples. However, the lower ratio in Figure 1 (top right) indicates the presence of high amount of hydrogen in the system, which is consistent with a moderate PA concentration of the sample. The Q-values of the peak positions indicated in the figures can be related to the following distances: 13.4 Å, 8.9 Å, 6.4 Å, 4.9 Å, 4.2 Å, and 3.6 Å. These distances correspond to certain inter- and intra-chain correlations, i.e., the inter-segmental distance between polymer chains as well as interatomic distances between phosphoric acid molecules appearing during dehydration of phosphoric acid (formation of pyro-phosphoric acid and water above 140°C) in accordance with the available literature information [see, e.g., References Trisno et al. (2022); Persson et al. (2018)]. The changes in the structure slow down dramatically after the first heating and main changes remains after multiple subsequent cooling and heating cycles, as shown in Figure 1 (bottom right) for a sample with an even higher content of PA, where we have observed sharper peaks at similar distances after several heating and cooling cycles. This can be considered as an evidence of irreversible thermally induced formation of a (semi-)crystalline structure during operation at elevated temperatures.

Transmission electron microscopy (TEM) has been used to highlight the structure of a swollen membrane. In real space, TEM provides a view onto the PA-swollen PBI membrane on similar length scales as SANS. However it lacks the possibility to perform thermal cycles during the measurements. Therefore only the initial structure of the PBI membrane filled with phosphoric acid could be investigated. Figure 2 shows cryo-TEM images on the very same location (although in two different magnifications) at few-milliseconds intervals. Initially there is no contrast between the PA and PBI domains and only a fully unstructured image is visible. The structural features of the membrane emerge then in subsequent images. It is obvious from this series of images that the phosphoric-acid-doped membrane inside the cryo-TEM is very sensitive to electron irradiation. At the low pressure (of 2.85 × 10⁻⁶ Pa), the vapor pressure of PA or remaining water contents might result in an evaporation at the outer parts of the membrane, making the remaining thermally more stable polymer structure appear. Pure PBI membranes show almost no structure in the TEM, as has been shown, e.g., in Ref. Holderer et al. (2015) for a similar (ABPBI) membrane.

Radiation induced damage can be exploited in cryo electron microscopy to highlight structures otherwise not visible. For example, in polymer science, Talmon’s group has used “radiation induced […] formation of a high contrast network in the specimen” due to the presence of ice layer on organic film, Talmon et al. (1986) and references therein. In the case of polymer blends, the different sensitivities of the polymeric components can result in a stronger mass loss in one component than in another, yielding a development of contrast at the start of irradiation in the electron microscope. See Michler (2008) and references therein. According to the pressure-temperature phase diagram of H₃PO₄, the boiling point of the phosphoric acid is at −124°C under a vacuum of 3 × 10⁻⁶ Pa which is the vacuum condition in the TEM column during a cryo-TEM experiment. Although the cryo-TEM experiment is performed at ca. −180°C, the area exposed to the electron beam may be warmer locally, inducing the beam damage in the H3PO4 sites in the PBI membrane. Thus, electron beam focussing reveals the structure of the film, presumably by PA or residual water evaporation due to local heating. The resulting larger areas where the structure surrounding the created holes emerges (see Figure 2) might be interpreted as the area where the crystalline regions can be formed in the first thermal cycle.

Figure 3 illustrates the schematic picture from these considerations for a swollen PBI membrane (left), initially in a fully amorphous state (left part of Figure 3), and after the first heating cycle with crystalline domains (right part of Figure 3). From our SANS results the microscopic picture of the PA-swollen PBI membrane can be represented by a continuous network of PBI strands with a diameter of about 3.2 nm, which is filled with PA. After the first heating cycles, about 1/5 of the PBI material forms crystalline clusters, which have an average distance from each other of about 8.6 nm and a diameter of roughly 3.2 nm. This is depicted by the sub-structures in Figure 3 (right). Neutron diffraction confirmed the increase of crystalline domains by a decrease of the width of the Bragg peaks. Also TEM images of swollen PBI membranes are consistent with this picture and the suggested structure of a swollen PBI network. The formation of crystalline domains seems to be an irreversible process, once the correlation peak is formed as shown by our neutron scattering data it stays for subsequent heating/cooling cycles.

Structural properties of phosphoric-acid-doped PBI membranes before the first use and after the first heating to operating temperatures have been investigated on length scales from atomic distances to micrometres with different neutron scattering techniques and electron microscopy. Scattering experiments allowed to observe the nano-structure of phosphoric-acid-doped PBI membranes under temperature conditions similar to operating conditions of the high-temperature fuel cells. Deuterated phosphoric acid has been used to achieve suitable contrast.

We found out that the first heating to the operating temperature of 140°C leads to a formation of a correlation peak significantly stronger than the shoulder-like broad slope of the initial sample. We interpret this as the formation of crystalline clusters in the thicker connecting points of the PBI network, which leads to an additional contrast with a characteristic repeat distance. Thermal cycles (not all curves are shown) indicate that only during the first heating of a PBI membrane, additional crystalline structures are created, most probably in small domains of the highly swollen PBI network. This higher crystalline ordering after heating the PBI structure for the first time is an irreversible process, and it can be speculated in which way such structures influence the mechanical stability of the membrane. Analysis of the scattering invariant together with peak position and correlation length of the additional peak allowed to deduce geometrical parameters of the PBI scaffold.

By combining small angle neutron scattering and neutron diffraction in reciprocal space with cryo-TEM images in real space a consistent picture of phosphoric-acid-swollen PBI membranes for fuel cell applications could be drawn. The overall picture obtained from cryo-TEM seems to be consistent with the initial stage of the phosphoric-acid-doped PBI membrane. Additionally, electron beam damage has been used to observe the structure of the chemically and thermally stable PBI membrane under cryogenic conditions. Further studies would be required to shed a light on the mechanical influence of the formation of the crystalline domains in phosphoric-acid-doped PBI membranes for high-temperature fuel cells.