Collaborative Research: Designing Functional Materials with Optimal Learning

新产品和材料加工方法通常需要识别更强、更轻、更便宜或在某些方面更好的新材料。 用试错法寻找新材料可能是昂贵的，而且往往是无效的。 有了这个奖项，新的数学方法和计算机软件将被开发出来，以加速材料的发现。 计划中的方法将把可用的选择范围缩小到最有可能成功的选择范围，使新材料和新工艺的发现更加可靠，成本更低。该方法的示范将用于柔性有机太阳能电池中使用的材料，但该方法也可以适用于药物或食品添加剂中使用的材料。计划采用一种新的材料设计最佳学习方法，该方法使用贝叶斯实验设计和机器学习的进步来预测以前数据和领域专业知识的材料特性，并明智地建议物理和计算实验，这些实验将提供最支持发现的信息。 这些新的数学技术有望大大加快材料设计，提供更可靠的更好的材料，并减少实验工作。 该方法将在一组现有候选人，溶剂选择和加工条件下搜索有机半导体材料，并在此搜索中整合物理和计算实验。测试用例是沉积在碳纳米管（CNT）上的扭曲六苯并冠烯（c-HBC）的全有机太阳能电池系统。 该复杂体系涉及的问题包括在不同加工条件下c-HBC和CNT之间的络合等，其提供了对最佳学习和计算机模拟方法的严格测试，以预测加工-结构-功能三元组。这种方法广泛适用于各种材料设计问题。

New products and material processing methods often require the identification of novel materials that are stronger, lighter, cheaper, or better in some way. Searching for new materials with a trial-and-error approach can be expensive and often ineffective. With this award, new mathematical methods and computer software will be developed to accelerate materials discovery. The planned approach will narrow the available options to those that are most likely to succeed, making discovery of new materials and processes more reliable and less costly. Demonstration of the approach will be made for materials to be used in flexible organic solar cells, but the methods could also be amenable to materials for use in pharmaceuticals or to food additives.A new optimal learning approach to materials design is planned that uses advances in Bayesian experimental design and machine learning to predict material properties from previous data and domain expertise, and to intelligently suggest physical and computational experiments that will provide information that is most supportive of discovery. These new mathematical techniques promise to greatly accelerate materials design, providing better materials more reliably and with less experimental effort. The approach will be demonstrated in the search for organic semiconductor materials over a set of existing candidates, solvent choices, and processing conditions, and integrate both physical and computational experiments in this search. The test case is an all-organic solar cell system of contorted hexabenzocoronenes (c-HBC), deposited on carbon nanotubes (CNT). This complex system involves issues including complexation between c-HBC and CNT at different processing conditions, etc., which provide a stringent test of optimal learning and computer simulation methods to predict the processing-structure-function triad. This approach is broadly applicable to a diverse set of materials design problems.