Dataset

CSD-MOFDB data [Download]

The MOF database contained 72,618 non-disordered structures integrated within the Cambridge Structural Database (CSD, version 5.4) were used. The solvent molecules were cleaned from the structures by using in-house script based on the graph-labeling algorithm. The important structural properties including largest cavity diameter (LCD), pore-limiting diameter (PLD), and helium void fraction, etc., were calculated using open-source software Zeo++. We refined this dataset by removing structures based on the PLD less than the diameter of three adsorbed molecules so that all three adsorbents can access to the MOFs’ pores, and the diameters of N2, CO2 and CH4 are 3.64 Å, 3.3 Å and 3.8 Å, respectively. Some structures failed in GCMC calculation are also removed. We finally eliminated those structures and ended up with 7306, 6998 and 8562 MOFs for N2, CO2 and CH4, respectively. GCMC simulations were further carried out to calculate the adsorption data of MOFs for CO2, N2 and CH4 using RASPA software. We set 8 pressure points from the range of 0.2 kPa - 80 kPa, 5 kPa – 20,000 kPa and 100 kPa – 10,000kPa for N2, CO2 and CH4, respectively.

NIST-ISODB dataset [Download]

The NIST/ARPA-E database (https://adsorbents.nist.gov) contains experimental gas adsorption data of MOFs, while few of them have atomic coordinate files. We obtained all gas adsorption capacities (CO2 at 273 K, N2 at 77 K and CH4 at 298 K) through the application programming interface (API) provided by the official website. We further manually collected MOFs structures with crystallographic information file (CIF) that are not used in CSD-MOFDB dataset. Finally, we obtained 54 MOFs with 1876 pressure data points covering N2, CO2 and CH4 adsorbate molecules.