Redirecting to original paper in 30 seconds...

Click below to go immediately or wait for automatic redirect

arxiv_ai 75% Match Research Paper Materials Scientists,Computational Chemists,AI Researchers,Chemical Engineers 1 week ago

Accelerating Materials Design via LLM-Guided Evolutionary Search

generative-ai › diffusion
📄 Abstract

Abstract: Materials discovery requires navigating vast chemical and structural spaces while satisfying multiple, often conflicting, objectives. We present LLM-guided Evolution for MAterials design (LLEMA), a unified framework that couples the scientific knowledge embedded in large language models with chemistry-informed evolutionary rules and memory-based refinement. At each iteration, an LLM proposes crystallographically specified candidates under explicit property constraints; a surrogate-augmented oracle estimates physicochemical properties; and a multi-objective scorer updates success/failure memories to guide subsequent generations. Evaluated on 14 realistic tasks spanning electronics, energy, coatings, optics, and aerospace, LLEMA discovers candidates that are chemically plausible, thermodynamically stable, and property-aligned, achieving higher hit-rates and stronger Pareto fronts than generative and LLM-only baselines. Ablation studies confirm the importance of rule-guided generation, memory-based refinement, and surrogate prediction. By enforcing synthesizability and multi-objective trade-offs, LLEMA delivers a principled pathway to accelerate practical materials discovery. Code: https://github.com/scientific-discovery/LLEMA
Authors (4)
Nikhil Abhyankar
Sanchit Kabra
Saaketh Desai
Chandan K. Reddy
Submitted
October 26, 2025
arXiv Category
cs.LG
arXiv PDF

Key Contributions

Introduces LLEMA, a unified framework that combines LLMs with chemistry-informed evolutionary rules for accelerated materials design. It proposes a novel iterative process where LLMs propose candidates, surrogate models estimate properties, and multi-objective scoring guides subsequent generations, outperforming existing baselines in discovering stable and property-aligned materials.

Business Value

Significantly accelerates the discovery and design of new materials with desired properties, leading to faster innovation in industries like electronics, energy, and aerospace. This can reduce R&D costs and time-to-market for new products.

Paper Metadata

Innovation Type

Framework/Algorithmic

Deployment Feasibility

Moderate to High, requires integration of LLMs, evolutionary algorithms, and property prediction models.

Limitations Addressed

Slow pace of traditional materials discovery,Difficulty in exploring large chemical spaces,Balancing multiple, often conflicting, material properties

Performance Gains

Achieves higher hit-rates and stronger Pareto fronts compared to generative and LLM-only baselines.

Technical Tags

LLM-guided searchevolutionary algorithmsmaterials discoverysurrogate modelsmulti-objective optimizationcrystallographyproperty predictiongenerative models

Research Topics

Materials ScienceAI for ScienceGenerative AIOptimizationMachine Learning

Methods & Architectures

LLM-guided Evolutionary SearchSurrogate-Augmented OracleMulti-objective ScoringMemory-based Refinement Large Language Models (LLMs)Evolutionary Algorithms

Applications & Tasks

Materials Discovery Chemical Design Navigating Vast Chemical SpacesSatisfying Multiple Conflicting ObjectivesAccelerating Materials Design Discovering Novel MaterialsPredicting Material PropertiesOptimizing Material Design

Datasets & Benchmarks

Benchmarks
14 realistic tasks spanning electronics, energy, coatings, optics, and aerospace
Hit-ratesPareto frontsChemical plausibilityThermodynamic stabilityProperty alignment

Related Fields

Materials ScienceComputational ChemistryMachine LearningArtificial IntelligenceOptimization

Keywords

materials discoveryLLMevolutionary searchmaterials designgenerative AImulti-objective optimizationsurrogate modelscrystallographyproperty predictionAI for sciencechemical spaceLLEMA

Academic Context

#Materials Science#AI for Science#Generative AI#Optimization#Machine Learning

Commercial Potential

Potential Products
AI-driven materials design softwareCustom material discovery services
Target Industries
Materials ScienceChemical IndustryAerospaceEnergyElectronicsCoatingsOptics
Use Case Examples
Designing new battery materialsDiscovering novel catalystsDeveloping advanced coatings
Competitive Edge
Integrates LLM knowledge with evolutionary search, offering a more guided and efficient approach than purely generative or traditional search methods.
Market Opportunity
Large, driven by the demand for novel materials across multiple high-value industries.
Revenue Models
Software licensingcloud-based design servicesIP licensing for discovered materials.

Resource Requirements

Compute Needs
High, due to LLM inference and iterative search process.
Data Requirements
Requires curated datasets for training property predictors and potentially for LLM fine-tuning.
Deployment Constraints
Integration complexity, computational cost.
Scalability
Scalable through parallelization of evolutionary search and efficient LLM inference.

Production Readiness

Maturity Level

Research

Time to Market

Medium-term, requires significant engineering to productize.

Patent Potential

Moderate, for the specific framework and its applications.

View Full Paper Back to Papers