Multimodal Language and Graph Learning of Adsorption Configuration in Catalysis

Janghoon Ock,Srivathsan Badrinarayanan,Rishikesh Magar,Akshay Antony,Amir Barati Farimani

2024-10-13

Abstract:Adsorption energy is a reactivity descriptor that must be accurately predicted for effective machine learning (ML) application in catalyst screening. This process involves determining the lowest energy across various adsorption configurations on a catalytic surface, which can exhibit very similar energy values. While graph neural networks (GNNs) have shown great success in computing the energy of catalyst systems, they rely heavily on atomic spatial coordinates. In contrast, transformer-based language models can directly use human-readable text inputs, potentially bypassing the need for detailed atomic positions. However, these language models often struggle with accurately predicting the energy of adsorption configurations. Our study addresses this limitation by introducing a self-supervised multi-modal learning approach called graph-assisted pretraining, which connects well-established GNNs with emerging language model applications. This method reduces the MAE of energy prediction for adsorption configurations by about 10%. Furthermore, our findings demonstrate that graph-assisted pretraining enhances fine-tuning with different datasets, indicating the transferability of this approach. This method also redirects the model's attention toward adsorption configuration, rather than individual adsorbate and catalyst information, similar to common domain knowledge. Building on this, we propose using generative large language models to create text inputs for the predictive model, based solely on chemical composition and surface orientation, without relying on exact atomic positions. This demonstrates a potential use case of language models in energy prediction without geometric information.

Computational Engineering, Finance, and Science

What problem does this paper attempt to address?

The problem this paper attempts to address is: how to improve the accuracy of energy prediction in catalytic adsorption configurations using language models (especially large language models) without relying on precise atomic position information.

Specifically, the paper focuses on the following aspects:

1. **Challenges in Adsorption Energy Prediction**: Adsorption energy is a key reactivity descriptor in catalyst screening and requires accurate prediction. Due to the very small energy differences between different adsorption configurations (usually between 0.1 and 0.3 eV), the model needs to be able to distinguish these subtle energy differences. Traditional Graph Neural Networks (GNNs), although performing well in calculating the energy of catalyst systems, are highly dependent on the spatial coordinates of atoms, which are difficult to obtain in experimental data.

2. **Potential and Limitations of Language Models**: Transformer-based language models can directly handle human-readable text inputs, potentially bypassing the need for detailed atomic positions. However, these language models have accuracy issues in predicting adsorption configuration energies.

3. **Multimodal Learning Approach**: The paper proposes a multimodal pretraining method—Graph-Assisted Pretraining (GAP), which transfers the knowledge of Graph Neural Networks to the latent space of language models through a self-supervised process, thereby improving the accuracy of language models in the task of adsorption configuration energy prediction.

4. **Application of Generative Models**: Furthermore, the paper explores the use of generative large language models (LLMs) to generate text inputs for energy prediction in the absence of precise atomic positions. This approach demonstrates the potential application of language models in energy prediction tasks, especially when detailed geometric information is lacking.

Through these methods, the paper aims to enhance the performance of language models in the task of catalytic adsorption configuration energy prediction, making them more effective and reliable in practical applications.