CAREER: Navigating Thermodynamic Landscapes for Phase Equilibria Predictions using Molecular Modeling and Machine Learning

职业：利用分子建模和机器学习在热力学景观中进行相平衡预测

基本信息

批准号：
2143346
负责人：
Yamil Colon
金额：
$ 51.08万
依托单位：
University of Notre Dame
依托单位国家：
美国
项目类别：
Continuing Grant
财政年份：
2022
资助国家：
美国
起止时间：
2022-03-01 至 2027-02-28
项目状态：
未结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=2143346&HistoricalAwards=false
关键词：
CAREER Navigating Thermodynamic Landscapes Phase

项目摘要

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).This CAREER project will use advanced computer simulations and machine learning algorithms to advance fundamental understanding of adsorption of gases in porous materials. Adsorption involves the concentration or rejection of molecules interacting with a material surface. It is a ubiquitous phenomenon present in our everyday lives and in many industrial and biological settings. Important technological applications that depend on adsorption processes include drug delivery, power production and energy storage, water harvesting, and others that affect the overall societal well-being of humanity. This research project makes use of powerful computational modeling tools to uncover a comprehensive picture of the interactions between the gas species and materials onto which they adsorb. This research will lead to fundamental insights into the adsorption process and the identification of promising new adsorbents that are crucial for technological advancements in areas of national importance including health care, climate change, and water scarcity. Integrated outreach and education components within this project include increasing literacy of machine learning at the undergraduate and graduate levels through course design; hosting middle school teachers through the Notre Dame Senior STEM Teaching Fellows Residency program to create course materials for 6-8th graders centered on probability and statistics; and translation of the middle school course material into Spanish for dissemination to Hispanic communities to improve their representation in STEM fields.This research program will integrate advanced molecular modeling and machine learning methods to create a universal gas adsorption model. By specifying the absorbent material, an adsorbate gas species, and the adsorption conditions (temperature and pressure), the model will be able to accurately predict the amount of gas that is adsorbed within the material pores at equilibrium. An adsorption model with such predictive capabilities would constitute an important engineering design tool, eliminating the current bottleneck posed by the high computational cost of screening all potential materials with molecular simulations and fundamentally advancing drug delivery, power production and energy storage (e.g., hydrogen), and atmospheric water harvesting and carbon capture technologies. The development of models to predict the nature of gas physisorption in porous materials will be developed within an active learning (AL) framework to efficiently navigate the large chemical spaces of adsorbates and adsorbents. The properties of absorbent materials and gas molecules will be represented as ‘features’ alchemically to maximize the range of materials and molecules that can be studied in a computationally feasible manner. The AL algorithm will inform, in an automated fashion, which simulations to perform to achieve accurate predictions with a limited number of simulations, thus allowing for an exhaustive yet efficient exploration of the feature space. The research plan is based on three objectives: (1) implement and validate an active learning framework capable of navigating adsorption landscapes, (2) navigate the feature landscapes of simple gas adsorbates, and (3) simultaneously navigate the feature landscapes of molecules and porous materials for gas adsorption. Because the proposed AL framework will be readily adaptable to other adsorption/material design scenarios, phase equilibrium studies beyond gas adsorption will benefit. These research efforts will be complemented by outreach efforts to middle schools and the public through bilingual curriculum development and middle school teacher training in probability and statistics, and dissemination of the course materials in Spanish to the local Hispanic community and in Puerto Rico.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该奖项的全部或部分资金来源于《2021 年美国救援计划法案》（公法 117-2）。该职业项目将使用先进的计算机模拟和机器学习算法来增进对多孔材料中气体吸附的基本了解。涉及分子与材料表面相互作用的排斥或排斥，这是我们日常生活以及许多工业和生物环境中普遍存在的现象，依赖于吸附过程，包括药物输送、发电和吸附。该研究项目利用强大的计算建模工具来全面揭示气体种类与其吸附的材料之间的相互作用。对吸附过程的基本见解和对有前景的新型吸附剂的识别，这些吸附剂对于医疗保健、气候变化和水资源短缺等国家重要领域的技术进步至关重要。该项目的综合和教育部分包括提高机器学习的素养。通过课程在本科和研究生阶段设计；通过圣母大学高级 STEM 教学研究员驻场计划接待中学教师，为 6 至 8 年级学生制作以概率和统计为中心的课程材料，并将中学课程材料翻译成西班牙语，以便向西班牙裔社区传播，以提高他们的代表性；该研究项目将整合先进的分子建模和机器学习方法，通过指定吸收材料、吸附气体种类和吸附条件（温度和压力）来创建通用气体吸附模型。能够准确预测平衡时材料孔隙内吸附的气体量，具有这种预测能力的吸附模型将构成一个重要的工程设计工具，消除目前筛选所有潜在材料的高计算成本所带来的瓶颈。分子模拟和从根本上推进药物输送、电力生产和能量存储（例如氢气）以及大气水收集和碳捕获技术将在主动学习中开发预测多孔材料中气体物理吸附性质的模型。 AL) 框架有效地导航吸附剂和吸附剂的大化学空间吸收剂材料和气体分子的特性将以炼金术的方式表示为“特征”，以最大化可以以计算可行的方式研究的材料和分子的范围。以自动化的方式告知要执行哪些模拟以通过有限数量的模拟实现准确的预测，从而允许对特征空间进行详尽而探索该研究计划基于三个有效的目标：（1）实施和验证。主动学习框架能够导航吸附景观，（2）导航简单气体吸附物的特征景观，以及（3）同时导航气体吸附的分子和多孔材料的特征景观，因为所提出的 AL 框架将很容易适应其他吸附/材料。这些研究工作将通过双语课程开发和中学教师概率与统计培训以及向中学和公众传播西班牙语课程材料来补充。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

期刊论文数量（5）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Using adsorbents to help society

使用吸附剂帮助社会

DOI：
10.33424/futurum431
发表时间：
2023-10
期刊：
Futurum Careers
影响因子：
0
作者：
Colón; Yamil
通讯作者：
Yamil

Metal–organic framework clustering through the lens of transfer learning

通过迁移学习的视角进行金属有机框架聚类

DOI：
10.1039/d3me00016h
发表时间：
2023-07
期刊：
Molecular Systems Design & Engineering
影响因子：
3.6
作者：
Cooper, Gregory M.;Colón, Yamil J.
通讯作者：
Colón, Yamil J.

Transfer Learning Facilitates the Prediction of Polymer–Surface Adhesion Strength

迁移学习有助于预测聚合物表面粘合强度

DOI：
10.1021/acs.jctc.2c01314
发表时间：
2023-04
期刊：
Journal of Chemical Theory and Computation
影响因子：
5.5
作者：
Shi, Jiale;Albreiki, Fahed;Colón, Yamil J.;Srivastava, Samanvaya;Whitmer, Jonathan K.
通讯作者：
Whitmer, Jonathan K.

Active Learning for Adsorption Simulations: Evaluation, Criteria Analysis, and Recommendations for Metal–Organic Frameworks

吸附模拟的主动学习：金属有机框架的评估、标准分析和建议