Surrogate-Gated Generation and Foundation-Model Embeddings for Bayesian Materials Design

2026-06-26

AI Summary

This work addresses the high computational cost of property evaluation in materials discovery, which hinders the efficiency of generative closed-loop design. To mitigate this bottleneck, the authors propose a Gaussian process (GP) surrogate gating mechanism that performs low-cost screening of candidate structures between the generative model and the evaluator, thereby avoiding expensive fine-tuning. Integrating a pretrained diffusion model (MatterGen/CrystalFlow/ADiT), ORB embeddings, a GP surrogate, and a reinforcement learning–guided workflow, the method achieves performance comparable to exhaustive search using only one-fifth of the evaluation budget. Experiments demonstrate that, under a fourfold evaluation budget, the gating strategy significantly outperforms fine-tuning without gating. Density functional theory (DFT) validation confirms an average property prediction error below 2.5% and a Spearman rank correlation coefficient of 0.94, underscoring the approach’s effectiveness and generalizability.

Abstract

Closed-loop materials discovery iterates between proposing candidate structures and evaluating their properties, and property evaluation dominates the cost. In the generative variant, a learned prior proposes candidate crystals and a property oracle scores them; we ask whether a cheap probabilistic surrogate can triage the generator's output, and what such a surrogate must do well. Across three architecturally distinct pretrained diffusion priors (MatterGen, CrystalFlow, ADiT) and two targets (room-temperature heat capacity and bulk modulus), we insert a Gaussian process acquisition gate between structure generation and the oracle in an RL-steered generative workflow. The gate matches or exceeds ungated fine-tuning of the generative model while capping oracle calls at a fixed per-cycle budget. Budget-matched ablations isolate the mechanism. At an identical four-call budget, ranking-based selection outperforms arbitrary selection, confirming that the gain comes from the surrogate's choice; the gate comes within $\sim$9\% of exhaustive oracle spending at roughly one-fifth of the calls. A density-functional-theory check of the bulk-modulus discoveries confirms the learned oracle to within 2.5\% on average and the surrogate's ranking of the generated structures at Spearman $\rho = 0.94$. A cross-factorial benchmark of surrogate performance spanning mechanical, electronic, and vibrational properties identifies pretrained ORB embeddings with a Gaussian process as the most reliable combination, which we adopt as the building blocks of the proposed workflow. The complete pipeline is released as open-source software.

Problem

• Bayesian materials design
• surrogate model
• property evaluation
• generative materials discovery
• oracle budget

Innovation

• Surrogate-Gated Generation
• Bayesian Materials Design
• Foundation-Model Embeddings
• Gaussian Process
• Closed-loop Discovery