Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RingFormer: A Ring-Enhanced Graph Transformer for Organic Solar Cell Property Prediction

Zhihao Ding, Ting Zhang, Yiran Li, Jieming Shi, Chen Jason Zhang

TL;DR

This work tackles the challenging problem of predicting OSC properties from molecular structures by explicitly modeling ring systems. It introduces RingFormer, a hierarchical OSC graph and a RingFormer layer that combines atom‑level message passing with ring‑level cross‑attention and inter‑level fusion to capture multi‑level structure. The approach yields state‑of‑the‑art results across five OSC datasets, notably achieving about a 22.8% relative improvement on CEPDB for PCE prediction, and demonstrates robustness to backbone choices and dataset complexity. By capturing ring interconnections and hierarchically fusing atom and ring information, RingFormer accelerates accurate OSC screening and design.

Abstract

Organic Solar Cells (OSCs) are a promising technology for sustainable energy production. However, the identification of molecules with desired OSC properties typically involves laborious experimental research. To accelerate progress in the field, it is crucial to develop machine learning models capable of accurately predicting the properties of OSC molecules. While graph representation learning has demonstrated success in molecular property prediction, it remains underexplored for OSC-specific tasks. Existing methods fail to capture the unique structural features of OSC molecules, particularly the intricate ring systems that critically influence OSC properties, leading to suboptimal performance. To fill the gap, we present RingFormer, a novel graph transformer framework specially designed to capture both atom and ring level structural patterns in OSC molecules. RingFormer constructs a hierarchical graph that integrates atomic and ring structures and employs a combination of local message passing and global attention mechanisms to generate expressive graph representations for accurate OSC property prediction. We evaluate RingFormer's effectiveness on five curated OSC molecule datasets through extensive experiments. The results demonstrate that RingFormer consistently outperforms existing methods, achieving a 22.77% relative improvement over the nearest competitor on the CEPDB dataset.

RingFormer: A Ring-Enhanced Graph Transformer for Organic Solar Cell Property Prediction

TL;DR

This work tackles the challenging problem of predicting OSC properties from molecular structures by explicitly modeling ring systems. It introduces RingFormer, a hierarchical OSC graph and a RingFormer layer that combines atom‑level message passing with ring‑level cross‑attention and inter‑level fusion to capture multi‑level structure. The approach yields state‑of‑the‑art results across five OSC datasets, notably achieving about a 22.8% relative improvement on CEPDB for PCE prediction, and demonstrates robustness to backbone choices and dataset complexity. By capturing ring interconnections and hierarchically fusing atom and ring information, RingFormer accelerates accurate OSC screening and design.

Abstract

Organic Solar Cells (OSCs) are a promising technology for sustainable energy production. However, the identification of molecules with desired OSC properties typically involves laborious experimental research. To accelerate progress in the field, it is crucial to develop machine learning models capable of accurately predicting the properties of OSC molecules. While graph representation learning has demonstrated success in molecular property prediction, it remains underexplored for OSC-specific tasks. Existing methods fail to capture the unique structural features of OSC molecules, particularly the intricate ring systems that critically influence OSC properties, leading to suboptimal performance. To fill the gap, we present RingFormer, a novel graph transformer framework specially designed to capture both atom and ring level structural patterns in OSC molecules. RingFormer constructs a hierarchical graph that integrates atomic and ring structures and employs a combination of local message passing and global attention mechanisms to generate expressive graph representations for accurate OSC property prediction. We evaluate RingFormer's effectiveness on five curated OSC molecule datasets through extensive experiments. The results demonstrate that RingFormer consistently outperforms existing methods, achieving a 22.77% relative improvement over the nearest competitor on the CEPDB dataset.

Paper Structure

This paper contains 19 sections, 10 equations, 4 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. The RingFormer Method
  5. Hierarchical OSC Graph Construction
  6. RingFormer Layer
  7. Prediction Layer
  8. Experiments
  9. Experimental Settings
  10. Datasets and Evaluation Metrics.
  11. Baselines and Implementation Details.
  12. Overall Performance
  13. Model Analysis
  14. Conclusion
  15. Dataset Details
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Example of OSC molecules.
  • Figure 2: The RingFormer framework. For clarity, we showcase the framework with $L=2$ RingFormer layers.
  • Figure 3: (a) Performance improvement on molecules with varying numbers of rings; (b) and (c) Comparison of embedding visualizations, colors representing the number of rings.
  • Figure 4: PCE (%) prediction performance of RingFormer by test MAE when the number of RingFormer layers $L$ varies.