Reverse Monte Carlo modeling in confined systems

V. Sanchez-Gil,E.G. Noya,E. Lomba

DOI: https://doi.org/10.1063/1.4861042

2013-12-17

Abstract:An extension of the well established Reverse Monte Carlo (RMC) method for modeling systems under close confinement has been developed. The method overcomes limitations induced by close confinement in systems such as fluids adsorbed in microporous materials. As a test of the method, we investigate a model system of $^{36}$Ar adsorbed into two zeolites with significantly different pore sizes: Silicalite-I (a pure silica form of ZSM-5 zeolite, characterized by relatively narrow channels forming a 3D network) at partial and full loadings and siliceous Faujasite (which exhibits relatively wide channels and large cavities). The model systems are simulated using Grand Canonical Monte Carlo and, in each case, its structure factor is used as input for the proposed method, which shows a rapid convergence and yields an adsorbate microscopic structure in good agreement with that of the model system, even to the level of three body correlations, when these are induced by the confining media. The application to experimental systems is straightforward incorporating factors such as the experimental resolution and appropriate q-sampling, along the lines of previous experiences of RMC modeling of powder diffraction data including Bragg and diffuse scattering.

Disordered Systems and Neural Networks

What problem does this paper attempt to address?

### Problems the paper attempts to solve

This paper aims to solve the challenges encountered in structural modeling in constrained systems (such as fluids adsorbed in microporous materials). Specifically, the authors developed an extended Reverse Monte Carlo (RMC) method to deal with the limitations imposed on the traditional RMC method by the tightly - constrained environment.

#### Main problems include:

1. **Molecular motion in the constrained environment**:
- In a tightly - constrained environment, such as an adsorbed fluid in a microporous material, the traditional RMC method is inefficient because molecular movement is extremely restricted. Most translation and rotation operations will be rejected, resulting in extremely low sampling efficiency.
- To overcome this problem, the authors introduced particle creation and annihilation attempts to allow more efficient sampling.

2. **Accurately recovering the adsorbate microstructure**:
- The authors hope to accurately recover the microstructure of the adsorbate from powder diffraction data and adsorption volume experiments, especially in the case where three - body correlations are induced by the constrained medium.
- To this end, they proposed the N - RMC method, where the number of particles \( N \) is an optimization variable.

3. **Handling the uncertainty of experimental data**:
- Experimental data usually has statistical uncertainty, which will affect the accuracy of the model. By introducing the uncertainty parameters \( \sigma_S \) and \( \sigma_N \), the authors explored the impact of these uncertainties on the performance of the method.

4. **Verifying the effectiveness of the method**:
- The authors used two zeolites (Silicalite - I and siliceous Faujasite) with significantly different pore sizes as test systems to study the adsorption of argon inside them.
- By comparing with the target structure factor generated by Grand Canonical Monte Carlo (GCMC) simulations, the effectiveness of the N - RMC method was verified.

#### Summary of mathematical formulas:

- The expression for the structure factor \( S(q) \) is:
\[
S(q)=\frac{2\pi^2}{NV\langle b\rangle^2}\sum_{q'}\frac{|F(q')|^2}{q'^2}\delta(q - q')
\]
where \( N \) and \( V \) are the number of atoms and the volume of the system respectively, \(\langle b\rangle\) is the average of the coherent scattering length, and \( F(q) \) is the correlation function between scattering nuclei.

- The difference structure factor \( S_{\text{diff}}(q) \) is expressed as:
\[
S_{\text{diff}}(q)=S_{\text{total}}(q)-S_{\text{zeo - zeo}}(q)
\]

- The deviation amount \( \chi^2 \) is defined as:
\[
\chi^2=\sum_{i = 1}^{N_q}\frac{(S_{\text{calc}}(q_i)-S_{\text{exp}}(q_i))^2}{\sigma^2(q_i)}
\]

- The acceptance probability \( P_{\text{acc}} \) is expressed as:
\[
P_{\text{acc}}=\min\left(1,\exp\left(-\frac{\chi_{\text{new}}^2-\chi_{\text{old}}^2}{2}-\frac{\Delta N_{\text{new}}^2-\Delta N_{\text{old}}^2}{2}\right)\right)
\]
where \(\Delta N^2=\frac{(N - N_{\text{exp}})^2}{\sigma_N^2}\), \( N \) is the instantaneous number of adsorbed particles, \( N_{\text{exp}}\)