(383aj) In silico Synthesis of Carbon Molecular Sieves for High-Performance D2/H2 Separation Using Quantum Sieve Effect

Deuterium (D₂), a stable isotope of hydrogen, is an important gas with many applications, such as neutron modulators in heavy water reactors, deuterium arc lamps, non-radioactive tracers in chemical and biological reactions, H-NMR, long-lived large-scale integrated circuits (LSI), optical fibers with reduced optical attenuation, and deuterium-modified pharmaceuticals. To produce D₂, several methods such as cryogenic distillation (selectivity: 1.5 (20 K)), chemical exchange (selectivity: 2.3), centrifugal concentration, and laser have been used. However, their selectivity are low and their energy cost consumption for separation are high because the amount of D₂ is as low as 0.015% of H₂ and there is almost no difference in thermodynamic and chemical properties between H₂ and D₂. Therefore, it is desirable to develop a separation technology with higher efficiency.

The quantum fluctuations of the H₂ molecule in free space are negligible at room temperature, however the effect is not negligible when the molecule is placed in nanoscale space at low temperature. The diameter of D₂ which mass is twice that of H₂ is smaller than that of H₂. Therefore, D₂ is expected to have a faster uptake rate than H₂.

We have proposed a reasonable design guideline for carbon molecular sieve (CMS) with high O₂/N₂ separation performance [1] and synthesized CMS with high C₃H₆/C₃H₈ separation performance [2] by a combination of molecular simulation and experimental methods. In this study, using the in silico synthesis method of CMS obtained in these studies, we attempted to search for a synthesis guideline of CMS with high D₂/H₂ separation performance by applying transition state theory (TST) with Feynman-Hibbs effective potentials to the obtained CMS model and its actual preparation.

Simulation

To reproduce the non-equilibrium chemical vapor deposition (CVD) process onto an activated carbon AC model, we constructed a simulation cell with two gas density control cells at both sides of the AC model. Diffusion and reactions of CVD gas molecules were described by the molecular dynamics (MD) method with reaction state summation (RSS) potential for carbon-carbon interactions (RSS-MD), and the gas densities in the control cells were controlled by the grand canonical Monte Carlo (GCMC) method. The RSS-MD and GCMC simulations were performed alternately (RSS-GCMD). Temperature was fixed at 1000 K by Nose´–Hoover thermostat, and the equations of motion were integrated using the velocity-Verlet algorithm with a time-step of 0.05 fs. A slit pore model consisting of graphene layers was used as the AC model. Hypothetical aromatic carbon molecules (C₆, C₇, and C₉) were used as the CVD gas. The selectivity of D₂ over H₂ for the obtained CMS models were estimated by the TST with Feynman-Hibbs effective potentials.

Experimental Section

CMS samples were prepared by the CVD treatment with benzene under nitrogen atmosphere at 1000 K using pelletized AC KP-448 of Osaka Gas Chemicals Co., Ltd. D₂ and H₂ adsorption isotherms at 77 K were measured using a specific surface area and pore size distribution apparatus [Belsorp Mini, MicrotracBel Corp]. Uptake curves of D₂ and H₂ on those samples were measured by the volumetric method at 77 K using High-Precision Gas and Vapor Adsorption Analyzer [Belsorp Max, MicrotracBel Corp]. Moreover, the linear driving force (LDF) coefficients (k(D₂), k(H₂)) were determined from the uptake curves, and the kinetic selectivity S_LDF(D₂/H₂) was assessed as k(D₂) / k(H₂). Pore-mouth diameters of the AC and CMS samples were determined by the molecular probe method. The sample for the breakthrough curve measurement was crushed and sieved to 80/100 mesh, and packed into a SUS column (Shimadzu GLC Corporation, inner diameter 2.0 mm, length 0.5 m) weighing 2 g. D₂ and H₂ gas were distributed at 0.167 mL/s each through the column, and the gas composition at the column outlet was measured using a The breakthrough curve was obtained by measuring the gas composition at the column outlet using a quadrupole mass spectrometer [Thermo Fisher Scientific]. The total pressure was 101.3 kPa.

Results and Discussion

The relationship of the D₂ selectivity S_TST(D₂/H₂) to H₂ at 77 K calculated using the CMS model as a function of the pore mouth diameter d shows that the S_TST(D₂/H₂) increases with narrowing of d, and when d was smaller than 0.32 nm, the selectivity exceeded 1.5 by cryogenic distillation (temperature: 20 K) and 2.3 by chemical exchange method. The S_TST(O₂/N₂) calculated using a CMS model with S_TST(D₂/H₂) > 3 was 48.4. This suggests that a higher D₂/H₂ selectivity can be achieved by slightly advancing the pore narrowing than in the CMS for O₂/N₂ separation (S_LDF (O₂/N₂) = 30). In other words, the results suggest that it is possible to screen CMS expressing high D₂/H₂ selectivity by measuring and analyzing O₂ and N₂ adsorption rate curves at room temperature, without the difficult experiment of using D₂ at 77 K. Therefore, based on previous experimental results [1, 2], CMS was prepared by CVD of the precursor activated carbon KP-448 with benzene. The degree of pore narrowing was controlled by the CVD time so that S(O₂/N₂) was 28~50.

The selectivity at adsorption equilibrium of the selected CMS was evaluated by the ideal adsorbed solution theory (IAST) method and was found to be close to 2. In addition, the kinetic selectivity S_LDF(D₂/H₂) was 2.4, which is higher than the selectivity of 1.5 obtained by cryogenic distillation (20 K) and the selectivity of 2.3 obtained by chemical exchange. The selectivity of the CMS was also confirmed to exceed 3 when a D₂:H₂ = 1:1 gas mixture was distributed to the CMS.

[1] Yasuyuki Yamane, Hideki Tanaka, and Minoru T. Miyahara, “In silico synthesis of carbon molecular sieves for high-performance air separation”, Carbon, 141, 626-634 (2019).

[2] Yasuyuki Yamane, Minoru T. Miyahara, and Hideki Tanaka “High-performance carbon molecular sieves for the separation of propylene and propane”, ACS Appl. Mater. Interfaces, 14, 17878–17888 (2022)