Materials Informatics: Emergence to Autonomous Discovery in the Age of AI

Area of Science:

• Materials Science
• Computer Science
• Data Science

Background:

• Materials informatics evolved from physics and information theory.
• Machine learning and AI integration, particularly since 2014-2016, transformed materials discovery.
• The U.S. Materials Genome Initiative spurred significant advancements.

Purpose of the Study:

• To provide a perspective on the evolution of materials informatics.
• To trace the historical context, current state, and future directions of the field.
• To review key methodologies and discuss challenges and solutions.

Main Methods:

• Review of historical contributions and foundational ideas.
• Analysis of machine learning and artificial intelligence (AI) integration, including deep learning and large language models (LLMs).
• Discussion of sequential design approaches like Bayesian Optimization and Reinforcement Learning, and transformer models.

Main Results:

• Machine learning and AI, especially LLMs, are now integral to property prediction, synthesis planning, and inverse design.
• Autonomous self-driving laboratories are emerging, driven by AI.
• Common pitfalls and solutions for LLMs and Bayesian Optimization in materials science are identified.

Conclusions:

• AI is evolving materials informatics from a predictive tool to a collaborative partner.
• Advances in active learning and uncertainty quantification pave the way for autonomous materials science.
• The field is moving towards reduced human involvement in the research loop.