A generative artificial intelligence framework based on a molecular diffusion model for the design of metal-organic frameworks for carbon capture

Abstract

Metal-organic frameworks (MOFs) exhibit great promise for CO2 capture. However, finding the best performing materials poses computational and experimental grand challenges in view of the vast chemical space of potential building blocks. Here, we introduce GHP-MOFassemble, a generative artificial intelligence (AI), high performance framework for the rational and accelerated design of MOFs with high CO2 adsorption capacity and synthesizable linkers. GHP-MOFassemble generates novel linkers, assembled with one of three pre-selected metal nodes (Cu paddlewheel, Zn paddlewheel, Zn tetramer) into MOFs in a primitive cubic topology. GHP-MOFassemble screens and validates AI-generated MOFs for uniqueness, synthesizability, structural validity, uses molecular dynamics simulations to study their stability and chemical consistency, and crystal graph neural networks and Grand Canonical Monte Carlo simulations to quantify their CO2 adsorption capacities. We present the top six AI-generated MOFs with CO2 capacities greater than 2 $m mol/g$, i.e., higher than 96.9% of structures in the hypothetical MOF dataset.

Figures (17)

• Figure 1: CO2 adsorption values at 0.1 bar and 300K. CO2 adsorption capacity values at a pressure and temperature of 0.1 bar and 300 K, respectively, for the top six AI-generated MOFs' according to grand canonical Monte Carlo (GCMC) simulations and our modified crystal graph convolutional neural network (CGCNN) model.
• Figure 2: Visualization of the crystal structure of AI-generated MOFs.a--f Crystal structure of the top six AI-generated MOFs. The color code used to represent atoms is: carbon in grey, nitrogen in dark blue, fluorine in cyan, zinc in purple, hydrogen in white, and lithium in green.
• Figure 3: Functional groups in high-performing MOFs. Comparison of the proportion of selected functional group that appear in high-performing AI-generated MOFs and MOFs in the hMOF dataset.
• Figure 4: Properties of the hMOF dataset.a Depictions of the most frequent node-topology pairs in hMOF structures. b Cumulative distribution functions of their 0.1 bar CO2 capacities.
• Figure 5: CO2 capacities of hMOF structures at different catenation levels. Empirical cumulative distribution functions of MOFs in the hMOF dataset at 0.1 bar at different catenation levels. The x axis is capped at 4 $\textrm{m\,mol}\,\textrm{g}^{-1}$ to preserve details.