Mesophase Engineering through Coarse-to-fine Grained Modeling

在化学系化学理论，模型和计算方法项目的支持下，康奈尔大学的Fernando Escobedo旨在创建计算工具，以设计有机材料，在分子尺度上组织成复杂的结构。简单的物理模型在绘制软材料的一般性质和分子组织方面非常有效，但它们缺乏为分子分配特定化学物质所需的细节，并指导材料合成以实现所需的性质。Escobedo将开发方法，通过添加缺失的化学细节来实现广泛使用的物理模型的预测。Escobedo将联合收割机结合先进的分子模拟方法和机器学习策略，系统地识别最有可能实现对形成复杂结构的材料的简单模型预测的化学物质。虽然开发的方法预计将适用于许多类别的有机材料，但Escobedo将通过涉及能够自组装成三维网络的大型多功能分子的基准示例来展示它们的用途。因此，这项工作的结果可以通过指导研究人员为分离膜和光催化剂以及催化剂和吸附剂的多孔网络制定稳定的复合材料来影响先进材料行业。该项目将使博士生和本科生在计算材料研究的培训。在推广和教育方面，PI将协调在康奈尔大学组织的一系列新的研讨会，以庆祝学生在研究和包容性方面的成就。在这个项目中，Escobedo博士正在开发和应用分子模拟策略，以识别能够形成具有部分结构顺序的复杂相（称为中间相）的多亲性低聚物和功能化纳米颗粒。特别是，从一个计算效率高，但化学不可知（CA），粗粒度（CG）模型，形成一个目标中间相，Escobedo将开发一个方案，能够找到化学特定（CS）模型，是细粒度（FG）或原子详细的，保留中间相形成能力。这种方法将被应用到两个不同类别的CG模型，已显示出显着的承诺，在生成复杂的中间相：（I）非线性的多亲性低聚物，其独特的化学块带来纳米相分离，和（II）二元共混物的纳米粒子表现出非添加剂的混合行为。在这两种情况下，已经预测了多个复杂的3D网络阶段，并且可能可以访问许多其他阶段。将给定的CA CG模型映射到表现出相同的中间相行为的CS FG模型将需要迭代使用机器学习模型来搜索预定义的化学空间。重要的是，Escobedo不会直接搜索映射到目标CA CG阶段的CS FG模型，而是在快速、计算高效的CG级别使用选择“过滤器”。具体来说，候选化学物质将首先被映射到简单的CS CG模型中，以容易地识别那些能够形成目标相的化学物质;然后将这些化学物质映射到候选CS FG模型中，以进行进一步验证。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

With support from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry, Fernando Escobedo of Cornell University aims to create computational tools to engineer organic materials that organize into intricate structures at molecular scales. Simple physical models are very efficient in charting the generic properties and molecular organization of soft materials, but they lack the detail necessary to assign specific chemistries to the molecules, and to guide the materials synthesis to realize the sought-after properties. Escobedo will develop methods to allow the predictions of widely used physical models to be realizable by adding the missing chemical details. Escobedo will combine advanced molecular simulation methods and machine learning strategies to systematically identify chemistries most likely to fulfill the simple-model predictions of materials that form complex structures of interest. While the methods developed are expected to be applicable to many classes of organic materials, Escobedo will demonstrate their use with benchmark examples involving large multi-functional molecules capable of self-assembling into three-dimensional networks. Accordingly, results from this work could impact the advanced materials industry by guiding researchers to formulate stable composites for separation membranes and photovoltaics, and porous networks for catalysts and adsorbents. The project will enable the training of doctoral and undergraduate students in computational materials research. For outreach and education, the PI will coordinate a new workshop series organized at Cornell to celebrate the student accomplishments in research and in inclusivity.In this project, Dr. Escobedo is developing and applying molecular simulation strategies to identify polyphilic oligomers and functionalized nanoparticles capable of forming complex phases with partial structural order, called mesophases. In particular, starting from a computationally efficient but chemistry-agnostic (CA), coarse-grained (CG) model that forms a target mesophase, Escobedo will develop a scheme able to find chemistry-specific (CS) models which are fine-grained (FG) or atomistically-detailed that preserve the mesophase-formation ability. This approach will be applied to two distinct classes of CG models that have shown significant promise in generating complex mesophases: (I) Non-linear polyphilic oligomers whose distinct chemical blocks bring about nano-phase segregation, and (II) binary blends of nanoparticles exhibiting non-additive mixing behavior. In both cases, multiple complex 3D network phases have already been predicted and many others are potentially accessible. The mapping a given CA CG model into a CS FG model exhibiting the same sought-after mesophase behavior will entail the iterative use of a machine learning model to search through a predefined chemical space. Importantly, instead of directly searching for a CS FG model that maps into a target CA CG phase, Escobedo will use a selection “filter” at the fast, computationally efficient CG level. Specifically, the candidate chemistries will be first mapped into simple CS CG models to readily identify those able to form the target phase; these will then be mapped onto candidate CS FG models for further validation.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.