Navigating the Landscape of Next-Generation Nonlinear Optical Materials Discovery: Opportunities and Challenges Driven by Artificial Intelligence

ConspectusNonlinear optical (NLO) materials, as a core role in all-solid-state lasers, quantum information, and optical communication, have made irreplaceable contributions to the development of the contemporary optoelectronics industry. Historically, the advancement of NLO materials has relied on empirical knowledge of physical and chemical principles. However, in recent years, the fast development of computational materials science based on quantum mechanics methods has provided a theoretical foundation for the rational design of NLO materials. Despite this, the vast chemical space of NLO materials still poses challenges to theoretical design methods. Computational search strategies that rely on first-principles calculations are limited by the exponential growth in the demand for computing resources, which makes the exploration of candidate NLO materials rather complex. Machine learning (ML) methods have shown considerable promise in material design due to their highly efficient prediction in recent years, offering innovative and efficient solutions to these critical bottlenecks. The main advantage of the ML method lies in its ability to deeply reveal complex structure-performance relationships and predict material properties in a relatively short time by constructing and training complex predictive models, thereby significantly accelerating the exploration of NLO materials. Given the significant progress and encouraging prospects made by AI technology in addressing the complex design challenges related to NLO materials, it has become imperative to systematically organize and summarize these research developments.This account systematically reviews the significant research progress made by theoretical design methods in NLO materials to address the practical challenges in material design. This account delves deeply into our self-developed NLO theoretical models and the development of ML methods. By innovatively embedding ML methods into computational frameworks, AI-driven NLO materials workflows have demonstrated outstanding capabilities: they can efficiently explore and analyze the complex structure–property relationships. This ability provides a powerful tool for effectively solving the two-way core problems in NLO material design - forward prediction and structure design. Based on the above analysis, we further propose a FCKI pattern aimed at combining systematic materialized knowledge with data-driven methods to construct an AI-based and application-driven NLO material design paradigm.Finally, we concisely summarize several key challenges and potential directions faced in the AI for NLO material design, such as data scarcity, model interpretability, and model innovation. The main objective is to significantly enhance the recognition of these often overlooked methodological bottlenecks that have limited the development of AI for NLO material design. Furthermore, we hope that this account can effectively encourage innovative research on next-generation NLO materials, thereby actively promoting the fast development of NLO materials under the paradigm shift driven by the era of AI.