U.S.-Turkey Cooperative Research: Global Optimization of Transferable Molecular Step Potential Functions

项目描述：该项目支持俄亥俄州阿克伦大学化学工程系Richard Elliott博士和土耳其伊斯坦布尔Bogazici大学化学工程系Mehmet Camurdan博士以及土耳其伊斯坦布尔Koc大学工业工程系Metin Turkay博士之间的合作研究。本研究的目的是开发一种基于蒸汽压、密度、扩散率、气液平衡和液液平衡等物理性质数据推断分子相互作用参数的通用工具。像这样的物理性质表征在化学过程分析和模拟以及化学产品设计中起着重要作用。从可转移分子相互作用的角度来描述这些性质将允许对新化合物的性质和结构-活性关系进行预测。分子动力学模拟详细的分子结构，包括分支、环和键角，被用作区分同分异构体和空间效应的基础。基于这些预测，工程师们将能够在实验合成努力之前，在“计算机”中识别出有利的潜在化合物。智力优势：研究方法基于不连续分子动力学和热力学摄动理论（DMD/TPT）。DMD/TPT将分子势分解为离散的步骤，类似于方形阱势，但具有多个吸引阱。通过TPT，将吸引势的表征简化为一个可逼近的全局优化问题。步骤的深度是简单的连续参数，预测的物理性质的偏差构成目标函数。最优步宽的选择需要一定程度的整数规划。场地直径的优化在需要新的模拟时带来了重大的损失。制定系统全局优化策略对优化方法提出了新的挑战。虽然DMD/TPT此时特别适合这种优化，但也可以通过导数和直方图重加权方法来探测连续电位中物理性质对参数的敏感性。随着使用连续电位的研究人员取得进展，他们将看到需要一种可靠而有效的方法来优化可转移的电位参数。更广泛的影响：这些将包括综合研究、教育和国际交流。PI是化学工程热力学文本的合著者，并且已经整合了相关的先前NSF支持的结果，他的教学应该发展到包括需要广泛分子洞察力的产品设计。预计化学产品设计将在课程中占有越来越重要的地位。这一合作将展示一个国际研究的范例，充分利用双方的兴趣和能力。

0421849 ElliottDescription: This project supports a collaborative research between Dr. Richard Elliott, Department of Chemical Engineering, Akron University, Akron, Ohio and Dr. Mehmet Camurdan, Chemical Engineering Department, Bogazici University, Istanbul, Turkey and Dr. Metin Turkay, Industrial Engineering Department, Koc University, in Istanbul, Turkey. The objective of the research is to develop a general tool for inferring molecular interaction parameters based on physical property data for vapor pressure, density, diffusivity, vapor-liquid equilibria, and liquid-liquid equilibria. Characterizations of physical properties like these play a major role in chemical process analysis and simulation and chemical product design. Characterizing these properties in terms of transferable molecular interactions will permit predictions of properties and structure-activity-relations for new compounds. Molecular dynamics simulation of the detailed molecular structure, including branching, rings, and bond-angles, is used as the basis for distinguishing isomeric and steric effects. Based on these predictions, engineers will be able to identify favorable prospective compounds "in silico," prior to experimental synthetic efforts. Intellectual Merits: The methodology of the research is based on Discontinuous Molecular Dynamics and Thermodynamic Perturbation Theory (DMD/TPT). DMD/TPT breaks the molecular potential down into discrete steps, similar to a square-well potential but with more than one attractive well. Through TPT, the characterization of the attractive potential is reduced to an approachable global optimization problem. The depths of the steps are simply continuous parameters and the deviations in the predicted physical properties comprise the objective function. The selection of optimal step widths necessitates a degree of integer programming. The optimization of site diameter carries a significant penalty in requiring a fresh simulation. Developing a strategy for global optimization of this system presents a novel challenge for optimization methodology. Although DMD/TPT is particularly well suited to this kind of optimization at this time, it is also possible through derivative and histogram reweighting methods to probe the sensitivity of physical properties to parameters in continuous potentials. As researchers using continuous potentials make progress, they will come to see the need for a reliable and efficient methodology for optimizing transferable potential parameters.Broader Impacts: These will include integrated research and education and international exchange. The PI is the co-author of a text on Chemical Engineering Thermodynamics, and has already integrated results of related previous NSF support, and his teaching should evolve to include product design requiring extensive molecular insight. It is anticipated that there will be progressively greater importance for chemical product design in the curriculum. This collaboration will demonstrate a paradigm for international research that substantially leverages the interests and capabilities in both sides.