Interviews with Faculty Researchers

– Interview with Prof. Yin Jun
Assistant Professor, Department of Applied Physics

Prof. Yin Jun and his team specialize in the theoretical and computational design of optoelectronic materials, focusing on sustainable, high-performance systems such as lead-free perovskites. Their research integrates artificial intelligence (AI) into traditional computational workflows, significantly enhancing the efficiency and predictive accuracy of materials discovery.

Using high-precision simulations, Prof. Yin’s team generates comprehensive datasets detailing crystal, electronic and optical properties. These rigorously verified datasets train AI models that can rapidly screen millions of material candidates and predict their performance in photovoltaic and optoelectronic applications. A major breakthrough is the development of machine learning potentials. By training AI on small-scale systems, the team can simulate large, complex materials, such as hybrid interfaces or perovskite structures containing hundreds of thousands of atoms, through custom force fields. This enables large-scale molecular dynamics simulations that were previously beyond reach.

Prof. Yin and his team also pioneer the search for environmentally friendly perovskites by replacing toxic lead with tin or other transition metals. AI tools help identify dopants that preserve optical performance while improving stability, balancing efficiency, durability and environmental safety. While efficiency is often prioritized in academic research, Prof. Yin notes that industry also emphasizes cost-effectiveness and practical deployment, such as alternatives to silicon solar cells.

Despite AI’s acceleration of discovery, challenges remain in bridging the gap between theoretical predictions and experimental observations. Simplifications in computational models, such as approximations for kinetics, solvent interactions and temperature effects, mean simulations cannot yet fully replicate complex experimental conditions. Prof. Yin believes that advances in computational power will allow future models to incorporate these factors, bringing theory and experiment closer together.

Experimental validation remains essential throughout. Prof. Yin’s workflow tightly couples theory and experiment, ensuring predictions retain physical meaning. The team plans to release their datasets after publication to support the broader research community, further advancing the integration of AI, computation and experiment, and bringing theoretical materials design closer to real-world application.

融合人工智能與運算技術　推動可持續光電材料設計

– 殷駿教授專訪
應用物理學系助理教授

殷駿教授的研究團隊長期深耕於光電材料理論與計算領域，致力於設計並開發具可持續性的高性能無鉛鈣鈦礦材料體系。團隊將人工智能融入傳統計算流程，大幅提升了材料研發效率與預測準確度。

團隊運用高精度模擬技術，建構了涵蓋晶體結構、電子特性及光學性質的大型數據庫，並透過嚴謹的實驗進行驗證，確保數據的可靠性。這些數據作為人工智能模型的訓練基礎，使模型得以在短時間內從數百萬種候選材料中快速篩選，並預測其在光伏與光電元件中的性能表現。其中一項重要突破在於開發出機器學習勢能模型：通過訓練人工智能理解小尺度系統的物理規律，進而實現對複雜大尺度材料結構的模擬，包括表面結構複雜、含有數十萬個原子的鈣鈦礦體系，使以往難以實現的含重原子大尺度分子動力學模擬成為可能。

團隊積極投入環保型鈣鈦礦材料的研發，以錫或其他過渡金屬取代具毒性的鉛元素。藉助人工智能技術，團隊迅速篩選出既能維持光學性能、又可提升材料穩定性的摻雜原子，在效率、穩定性與在環境友好之間取得平衡。殷教授指出，學術界通常聚焦於提升效率指標，而產業界則更關注成本效益與實際應用，例如開發可替代矽基太陽能電池的可行方案。

儘管新材料的開發效率已顯著提升，理論與實驗之間仍存在差距。由於計算模型在反應動力學、溶劑效應及溫度條件等方面仍需簡化，現有模擬結果未必能完全反映真實實驗環境。殷教授相信，隨着運算能力與模型方法持續進步，未來將能納入更多複雜因素，進一步縮小理論預測與實驗觀測之間的差距。

實驗驗證始終是科研的關鍵一環。殷教授的研究範式強調理論與實驗的深度融合，使各項預測在物理上的準確性顯著提升。團隊亦計劃在相關論文發表後，將數據庫公開予學術界共享，持續為綠色能源與先進光電科技的發展貢獻力量。