学术讲座
报告题目:Machine Learned Electronic Structures and Optical Properties for Organic Semiconductors
报告时间:2024-11-04 10:00
报告人: 林舟 助理教授
美国马萨诸塞大学阿默斯特校区
报告地点:曾呈奎楼B311
报告摘要:
Computational material discovery based on density functional theory (DFT) has achieved tremendous success in recent decades. However, the power of DFT on organic semiconductors (OSC) as molecular electronic materials suffer significantly from its computational complexity and intrinsic errors. Here we introduced a new exchange–correlation (XC) functional developed by us, referred to as ML-ωPBE, which evaluates the molecule-specific range-separation parameter (ω) in a range-separated hybrid (RSH) functional using a stacked ensemble machine learning algorithm and a composite molecular descriptor. [1] Compared to first-principles OT-ωPBE, a well-trained ML-ωPBE reaches a mean absolute error (MAE) of 0.00504 a0–1 for optimal ω’s, reduces the computational cost by 2.66 orders of magnitude, and achieves comparable predictive power in optical properties. In addition, ML-ωPBE shows a strong domain adaptation from closed shell molecules to open shell radicals. [2] From this study we concluded the importance of descriptors from semi-empirical quantum chemical calculations. Our study will set the stage for developing physics-based, and data-driven computational models for high-throughput material and drug discovery.
报告人简介:
林舟博士,2009年获中国科学技术大学化学物理学士学位,2015年获美国俄亥俄州立大学化学物理博士学位和校长奖学金,2015年至2018年在麻省理工学院和加州大学伯克利分校从事博士后研究工作,期间曾获美国化学会物理化学青年科学家奖。2020年9月林舟博士入职麻省大学安默斯特分校化学系担任助理教授和博士生导师,主要从事量子力学和机器学习理论的开发以及在化学中的应用,期间曾获美国国家自然基金会ADVANCE合作种子基金,美国科学进步研究协会Scialog碳中和研究基金,美国化学会石油研究基金,贝索斯地球基金会碳中和研究基金和Sanibel量子化学青年科学家奖。迄今已在知名化学期刊和计算机顶级会议上发表论文32篇,其中18篇为通讯作者或第一(含共同第一)作者。
欢迎老师同学们积极参加!
beat365官方网站
固体表面物理化学国家重点实验室