(518d) Moisture-Resistant Graphene-Based Nanolaminate Membranes for Hydrogen Purification Enabled By Charge Neutralizing Nanofillers

More than 90% of 70 million tons of worldwide demand for hydrogen (H₂) is currently generated via steam reforming and the water‐gas‐shift (WGS) reaction of fossil fuel sources¹. This produces mixture of H₂ and carbon dioxide (CO₂) saturated with steam from which H₂ is subsequently purified by selectively adsorbing the CO₂ using aminated solvents. Unlike amine absorption processes, membrane separation requires no changes in phase behavior and is, in principle, the most energetically cost-effective route for primary purification of H₂ and CO₂ ². However, polymeric membranes do not have adequate separation capability of CO₂ from H₂ due to the high level of CO₂ solubility in organic materials ³. Costly metallic and ceramic membranes, operating on the principles of selective adsorption or molecular sieving, have extraordinary selectivity for H₂ against CO₂; ~1000 and ~300, respectively, and sufficient rates of permeance.

Ultra-thin graphene oxide (GO) has been proposed as a step-change membrane material to separate H₂ and CO₂. High selectivities (up to 1000) and triple-digit gas permeation unit (GPU) values (1 GPU = 3.35×10^−10 mol m^−2 s^−1 Pa ^−1) were reported ⁴. However, when the GO membranes are exposed to a humid environment, the hydrated GO sheets become negatively charged and will come apart due to the electrostatic repulsion that promotes the GO membrane's delamination. Such catastrophic swelling is an un-resolved obstacle to the practical implementation of this exciting technology⁵. In this work, we specifically target the stabilization of GO membranes using nanodiamonds (NDs) towards adverse humid conditions while attempting to maintain its overall high performance towards H₂/CO₂ separation⁶.

Results and discussion

GO membranes were prepared by vacuum filtration method of a suspension of single-layer GO sheets onto both ceramic and polymeric supports. 3 nm-sized NDs were controllably added to the GO solutions before vacuum filtration. The prepared membranes were designated as GOαND, where α (α= 5, 10, 20, 30, and 35) represents the weight concentration of ND particles. The atomic force microscope (AFM) images of the pure GO membrane revealed a smooth surface without visible defects. However, the surface roughness of the composite membrane increased with the addition of ND particles (Fig. 1a, b).

The cross-section view of the GO membrane displays a highly packed morphology with a uniform thickness of about 38±6 nm. The addition of ND particles (30 wt%) to the GO matrix increased the membrane thickness to 75±8 nm (Fig. 1c, d).

On ceramic supports, the produced native GO membranes were found to have performances comparable to or slightly better than those reported in the literature with initial H₂ permeance of around 1150 GPU and ideal gas selectivity of ∼284 over CO₂. However, when exposed to a water-saturated equimolar mixed gas feed, the GO membrane performances dramatically deteriorated over a 100 hr test at room temperature. Permeances and selectivity dropped by 55% and 70%, respectively. By contrast, membranes containing 30 wt% of NDs, GO30ND, were found to have permeances thrice the level of native GO membranes (to 3700 GPU) and with a relatively small reduction in ideal gas selectivity (to a_H2/CO2 ~ 210) when tested with dry gas. However, most significantly, there was only a ~5% and ~10% drop in permeance, and selectivity of GO30ND membranes, when tested extensively with a wet mixed gas feed (Fig. 1e, f).

The ND particles were positively charged (+45 mV), while the GO sheets carried a net negative charge (-48 mV) at pH=7. Thus, the proper assembling of NDs into the GO structure via strong electrostatic interactions can be presumed (Fig. 1g). Even though the NDs altered the stacking of the GO laminates, the composite membrane remained intact due to the electrostatic interaction and H₂ bonding between the GO sheets and ND particles. Hydrogen bonding in the GOαND membranes is verified by shifting prominent bonding peaks in FTIR spectra. C1s and N1s X-ray photoelectron spectroscopy (XPS) spectra confirmed the presence of ND in the GO mixtures. The X-ray analysis of GO membrane suggests highly ordered stacking of GO laminates (sharp peak at 2θ=6.15°) with the d-spacing of 0.93 nm (Fig. 1h). The intensity of the peak decreased and broadened by the introduction of NDs, which confirms the disruption of GO stacked ordering. Furthermore, the peak shifted slightly with the addition of NDs, and the equivalent d-spacing reached 0.89 nm in GO30ND membranes. The insertion of positively charged ND particles between the GO laminates diminished the negative charge effect, weakened the interlayer electrostatic repulsion, and reduced the d-spacing of GO sheets. Young’s modulus and hardness of the GOαND membranes both improved up to ∼25% compared to the pure GO membrane. The improved mechanical properties can be explained by the favorable interaction between GO and NDs, which is desired for long-term operation of the membranes.

The incremental addition of ND particles into the GO structure improved its gas permeation in a controlled and significant manner. At the same time, the separation factors were almost at the same level as GO membranes at low filler concentrations, i.e., up to 30 wt% (Fig. 1i). An H₂ permeance ~3741 GPU (α_H2/CO2=211.6) was observed in the GO30ND membrane; an exceptional ~300% enhancement in the permeance compared to the pure GO membrane. The decrease in the number of GO layers per laminates and the reduction in the crystallite width of the GO laminates are responsible for the permeance enhancement. N₂ adsorption test indicates a 500% enhancement in the pore volume from 0.036 cm³/g in pure GO to 0.17 cm³/g by adding 30 wt% ND particles, which facilitates the gas diffusivity in the composite membrane. Generally, compared across a wide spectrum of inorganic materials, the GO30ND membrane show both exceptional H₂ permeance (>3700 GPU) and H₂/CO₂ selectivity (>200).

The evaluation of gas separation properties of the membranes was investigated under mixed gas feeds. Using equimolar gas feeds similar to that from a WGS reaction, H₂ permeance and H₂/CO₂ selectivity of the GO30ND membrane decreased 6% and 13%, respectively. The reduction in the permeance and selectivity of the membranes under the mixed gas condition is, in general, due to the partial hindrance of H₂ molecules transport by highly adsorbed CO₂ molecules.

The stabilizing property of NDs against humidity ascertained from the immersion of GO30ND membranes in water. Here, thicker GO membranes (~200 nm) prepared over a larger area by vacuum filtration on polyethersulfone supports are seen to disintegrate upon immersion, whereas GO30ND membranes are stable over the same period. However, a more significant test of the ND stabilization effect is demonstrated by exposing the membranes cyclically to wet and dry feeds of equimolar H₂/CO₂ mixture (85% relative humidity) (Fig. 1j). Native GO membranes were not able to survive a single full cycle exposure, becoming fully permeable to both gases. Interestingly the membrane gets worse after being exposed to a second dry gas feed, suggesting that significant and irreversible structural reorganization takes place during both wetting and drying of the hygroscopic material. On the other hand, GOaND membranes are found to have a degree of reversibility in membrane properties when exposed to wet and dry gases, though there is a net reduction in both permeance and selectivity over a series of cycles. Although this reduction is minimized with higher ND content, even GO30ND membranes do not have perfect stability against the vigorous cyclic tests. This quasi-reversible variation of membrane performance under cycling or constant humid conditions suggests that the NDs stabilize the GO nanolaminates within the membrane.

Conclusion

In this study, we have demonstrated the benefit of the intentional addition of positively charged carbonaceous nanoparticles, here exemplified through NDs, to negatively charged GO membranes. We demonstrated that NDs with sp²/sp³ core-shell structure and positive surface charge could reduce the electrostatic repulsive forces between hydrated GO sheets, where their robust and GO compatible structures are intercalated between the GO laminates, strengthening the membrane structure. As a result, the delamination of GO nanofilms in the presence of humidity is suppressed. The NDs stabilize GO membranes against humidity while tremendously enhancing its intrinsic separation performance. The approach presented in this work will generate high stability 2D material composites, not only for membrane separation but also for other application areas where the field of 2D materials is gaining traction.

References

1. Voldsund, M., Jordal, K. & Anantharaman, R. H2 production with CO2 capture. Int. J. Hydrogen Energy 41, 4969–4992 (2016).
2. Merkel, T. C., Zhou, M. & Baker, R. W. Carbon dioxide capture with membranes at an IGCC power plant. J. Memb. Sci. 389, 441–450 (2012).
3. Zhang, M. et al. Electropolymerization of Molecular‐Sieving Polythiophene Membranes for Hydrogen Separation. Angew. Chemie 131, 8860–8864 (2019).
4. Kim, H. W. et al. Selective Gas Transport Through Few-Layered Graphene and Graphene Oxide Membranes. Science 342, 91–95 (2013).
5. Abraham, J. et al. Tunable sieving of ions using graphene oxide membranes. Nat. Nanotechnol. 12, 546–550 (2017).
6. Huang, G. et al. Overcoming humidity-induced swelling of graphene oxide-based hydrogen membranes using charge-compensating nanodiamonds. Nat. Energy 6, 1176–1187 (2021).