HP$^3$-NS: Hybrid Perovskite Property Prediction Using Nested Subgraph

Longhan Zhang, Temesgen Muruts Weldengus, Ruixuan Song, Zhaoliang Lei, Haowen Wang, Yiteng Zhai, Jing Jiang

Keywords: Hybrid organic-inorganic materials; graph representation; material designing;

TL;DR: We proposed a nested graph representation for hybrid organic-inorganic crystalline material representation learning to learn both molecular and compound representation when predicting the material's property

Abstract: Many machine learning techniques have demonstrated superiority in large-scale material screening, enabling rapid and accurate estimation of material properties. However, data representation on hybrid organic-inorganic (HOI) crystalline materials poses a distinct challenge due to their intricate nature. Current graph-based representations often struggle to effectively capture the nuanced interactions between organic and inorganic components. Furthermore, these methods typically rely on detailed structural information that hinders the applications of the methods for novel material discovery. To address these, we propose a nested graph representation HP$^3$-NS (Hybrid Perovskite Property Prediction Using Nested Subgraph) that hierarchically encodes the distinct interactions within hybrid crystals. Our encoding scheme incorporates both intra- and inter-molecular interactions and distinguishes between the organic and inorganic components. This hierarchical representation also removes the dependence on detailed structural data, enabling the model application to newly designed materials. We demonstrate the effectiveness and significance of the method on hybrid perovskite datasets, wherein the proposed HP$^3$-NS achieves significant accuracy improvement compared to current state-of-the-art techniques for hybrid material property prediction tasks. Our method shows promising potential to accelerate hybrid perovskite development by enabling effective computational screening and analysis of HOI crystals.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7077