Optimizing information content in MOF sensor arrays for analyzing methane-air mixtures

Abstract

In this study, multiple arrays of sensors composed from all possible combinations of nine different metal-organic framework (MOF) materials were evaluated for their methane-in-air sensing performance using molecular simulations. We considered all of the gas mixture compositions of CH4, N2, and O2, varying from 0% to 100% of each component in 1% steps (5151 mixtures in total), in all MOFs. Assuming the mass of adsorbed gas in each MOF can be measured using a microelectromechanical system (MEMS) device such as a surface acoustic wave (SAW) sensor, the expected signal response can be predicted from molecular gas adsorption simulations. By combining predicted signal responses from each MOF sensor element in an array, we are able to determine which sets of MOFs provide the most information about the gas mixture they are exposed to, as measured by the Kullback-Liebler divergence (KLD). The KLD values are then used for ranking array performances. We report results for both binary mixtures of CH4 and N2 and ternary mixtures that include O2. As expected, we found that increasing the number of elements in an array improves overall sensor performance; however, for a given array size, there is a wide disparity in KLD values between the best and worst arrays. This disparity highlights the potential inefficiency of choosing sensing materials for an array by experimental trial-and-error. Instead, we advocate the use of theory to intelligently select the best performing sensor arrays.