Visualization and Quantification of Geometric Diversity in Metal–Organic Frameworks

Abstract

With ever-growing numbers of metal−organic framework (MOF) materials being reported, new computational approaches are required for a quantitative understanding of structure−property correlations in MOFs. Here, we show how structural coarse-graining and embedding ("unsupervised learning") schemes can together give new insights into the geometric diversity of MOF structures. Based on a curated data set of 1262 reported experimental structures, we automatically generate coarse-grained and rescaled representations which we couple to a kernel-based similarity metric and to widely used embedding schemes. This approach allows us to visualize the breadth of geometric diversity within individual topologies and to quantify the distributions of local and global similarities across the structural space of MOFs. The methodology is implemented in an openly available Python package and is expected to be useful in future high-throughput studies.

Mapping the Porous and Chemical Structure–Function Relationships of Trace CH3I Capture by Metal–Organic Frameworks using Machine Learning

“…SOAP descriptors, on the other hand, do encode the chemical nature of the MOF backbone in the form of atomic neighbor environments within a cut-off radius of, typically, 4–8 Å. They are effective at capturing local chemical information, such as chemical bonds, interatomic/intermolecular interactions, and, generally, densely packed systems. , However, it has been shown that SOAP descriptors fail to capture long-range order, such as molecular packing . Furthermore, we show below that neither SOAP descriptors nor the pore descriptors on their own are effective at encoding the pore space within a MOF for the prediction of its performance in trace CH 3 I capture.…”

Coarse-grained versus fully atomistic machine learning for zeolitic imidazolate frameworks

“…We have previously shown that cg-ML models enable unsupervised learning in this domain, by visualising structural relationships between ZIFs and inorganic AB 2 networks. 12 Our present work now shows that energetics in Zn(Im) 2 (where Im = imidazolate) can be described, to useful accuracy, by supervised cg-ML models. Our study complements wider-ranging activities on cg force-field development, 13 and at the same time it addresses general questions about the nature of hybrid framework materials.…”

Coarse-Grained Crystal Graph Neural Networks for Reticular Materials Design

“…The term refers to the coarse-grain modeling approach, which involves simplification of complex atomic systems, e.g., reducing the number of degrees of freedom by representing groups of atoms as a single pseudoatom. Similar techniques have been applied to quantify MOF diversity , (in an unsupervised manner) and to establish structure–property relationships in a specific class of hybrid materials (zeolitic imidazolate frameworks); the PolymerGNN architecture, which introduces monomer units as molecular graphs, is also worth mentioning. In a recent study, coarse-grained representation was used for generating novel MOF structures within diffusion models.…”