TransitionMatrix Monte Carlo (TMMC) simulations of binary mixtures of N_{2} and CO2 [110] were performed at T = 300 K and 350 K in the ZIF8 metalorganic framework (MOF). The main result of a GCTMMC simulation is the particle number probability distribution (PNPD), which is constructed by stitching together the particle number distributions from each window. The adsorption isotherm may be determined from the PNPD [10].
Simulations used different combinations of Monte Carlo moves depending on the number of particles, as described below. Low density windows used a conventional set of moves (i.e., no configurational bias) whereas high density windows used a configurational bias strategy known as Dualcut configurational bias (DCCB) Monte Carlo [13]. Additionally, lowdensity windows started in WangLandau biasing mode to quickly generate a guess of the PNPD, later switching to TMMC mode; data from the TMMC phase of the simulation was saved and reported here.
Other key simulation details common to all simulations are given below:
Fluid Model  TraPPE N_{2} and CO_{2}[11]  
LennardJones cutoff  12A (cut potential, no tail correction)  
Ewald Parameters  Set according to DL_POLY recipe [15], with relative tolerance 10^{5}  
Trial Move Types  ALL: Translation, Pivot rotation GrandCanonical: N < Nmax/2: Standard Insertion and Deletion N > Nmax/2: Dualcut Configurational Bias (DCCB) [13] Growth Isochoric Semigrand Canonical: Identitychange  
DCCB Details  Number of trials per atom: 4 Reference Potential: LennardJones + Ewald Realspace Cutoff Radius for Reference Potential: 4.5A  
Bias update freq  1.0e5  
Simulation Length  200 sweeps [FEASST]  
Physical Parameters  CODATA 2018 [12] 
Simulations were performed using the opensource FEASST Monte Carlo engine [11], using version 0.21.1 or 0.24.1.
The adsorbent MOF was reconstructed from publiclyavailable crystal structures and replicated to ensure that the simulation cell was at least twice the cutoff radius in all dimensions. Forcefields for each MOF were taken from published literature. The MOF structure and forcefield are provided in FEASST particle files in the data repository associated with this page (see "Data Availability" below). Coulombic interactions were handled using the Ewald summation method [2,3] (parameters listed in the metadata files). LorentzBerthelot combining rules were used to set the unlikeatom LennardJones parameters.
Simulation details specific to each MOF
ZIF8  
N_{max}  N_{2}:350; CO_{2}: 300 
Unit Cell Replication (N_{x}, N_{y}, N_{z})  (2,2,2) 
Cubic Box Dimensions (A)  34.023240 
Simulation MOF Mass (amu)  21846.91 
MOF Forcefield, Reference  Snurr [16] 
FEASST MOF Particle  data.ZIF8_Snurr_rep222 
The result of each simulation is the PNPD and average potential energy (for each N state). The PNPD may be used to compute the adsorption isotherm by the histogramreweighting procedure described by Shen and Errington [8]. All systems were single phase and, hence, no phase decomposition of the PNPD was necessary.
The pressure for a particular chemical potential pair was determined from TMMC simulations of the bulk N_{2}/CO_{2} fluid at the same cutoff.
Credit:
NIST

Credit:
NIST

Credit:
NIST

Credit:
NIST

Various data files used to generate the reference isotherms are available in a Git Repository: https://github.com/dwsideriusNIST/NIST_SRSW_Data/tree/master/N2_CO2_REF_ISOTHERMS
Files in the repository include:
FEASST particle file for the MOF material [includes atomic coordinates and the forcefield parameters]
FEASST particle files for TraPPE N_{2} and CO_{2 }
Particle number probability distributions for the adsorbed fluid mixture
Isotherm data files, including the individual species adsorption isotherms and selectivity and estimated uncertainties, formatted as AIF files [17]