Towards a Generalizable AI for Materials Discovery: Validation through Immersion Coolant Screening

2025-10-27

AI Summary

Existing AI models for materials science are typically designed for single-property prediction, suffering from poor generalizability, high redundant training costs, and attribute fragmentation that often leads to false negatives. Method: This paper introduces the Generalized Geometric Alignment Transfer Encoder (GATE) framework—a novel architecture that jointly models 34 cross-domain physicochemical properties (e.g., thermal, electrical, mechanical, optical) within a unified geometric representation space, enabling cross-property relational learning and zero-shot transfer. Contribution/Results: GATE eliminates the need for task-specific data collection or retraining, supporting direct deployment across diverse materials discovery scenarios. In an immersed coolant screening task, it efficiently identified 92,861 candidate molecules from a billion-scale chemical library; four candidates were experimentally or literature-validated and demonstrated performance comparable to or exceeding commercial coolants. The framework significantly improves both accuracy and efficiency in multi-objective materials screening.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Artificial intelligence (AI) has emerged as a powerful accelerator of materials discovery, yet most existing models remain problem-specific, requiring additional data collection and retraining for each new property. Here we introduce and validate GATE (Geometrically Aligned Transfer Encoder) -- a generalizable AI framework that jointly learns 34 physicochemical properties spanning thermal, electrical, mechanical, and optical domains. By aligning these properties within a shared geometric space, GATE captures cross-property correlations that reduce disjoint-property bias -- a key factor causing false negatives in multi-criteria screening. To demonstrate its generalizability, GATE -- without any problem-specific reconfiguration -- was directly applied to the discovery of immersion cooling fluids for data centers, a stringent real-world challenge defined by the Open Compute Project (OCP). Screening billions of candidates, GATE identified 92,861 molecules as promising for practical deployment. Four were experimentally or literarily validated, showing strong agreement with wet-lab measurements and performance comparable to or exceeding a commercial coolant. These results establish GATE as a ready-to-use, generalizable AI platform readily applicable across diverse materials discovery tasks.

Problem

• Developing generalizable AI for multi-property materials discovery
• Reducing disjoint-property bias in multi-criteria screening
• Validating framework through immersion coolant discovery without retraining

Innovation

• GATE framework learns multiple material properties jointly
• Aligns properties in shared geometric space to reduce bias
• Directly applied to immersion coolant discovery without retraining