Collaborative Research: Designing Polymer Grafted-Nanoparticle Melts through a Hierarchical Computational Approach

材料研究部和化学、生物工程、环境和运输系统司为该奖项提供资金，该奖项支持开发准确的分层建模方法的研究，旨在了解仅由化学接枝高聚物链的无机纳米颗粒组成的材料的结构和机械性能。聚合物接枝纳米颗粒代表着我们制造、使用和回收塑料的方式的范式转变，因为这些复合材料在强度和韧性方面有潜力超过传统的杂化材料(即聚合物和纳米颗粒的物理混合物)。这是因为聚合物直接附着在纳米颗粒上，从而确保这些材料构成聚合物和纳米颗粒的化学固定混合物--另一方面，物理混合物往往会分离。然而，纳米颗粒及其接枝物的大小不同，也就是所谓的分散性，导致了不寻常的流变性和断裂特性，但人们对此仍知之甚少。这项拟议的研究将利用分子模拟、理论和机器学习来帮助解释分散性是如何导致这些特性的，以及如何进一步利用它来发现超过当前纳米复合材料能力的材料。这些研究想法与教育和外展目标紧密结合，以丰富为STEM学生提供的K-12和本科课程。为此，该团队将创建模拟、聚合物物理、机器学习和材料设计的在线学习模块。其他活动包括通过西北大学预测科学和工程设计集群对研究生进行跨学科培训，分享力场和软件开发，软件商业化，以及为代表性不足的学生群体提供新的研究机会。材料研究司和化学、生物工程、环境和运输系统司为该奖项提供资金。该项目解决了一项关键的知识空白，即了解多组分聚合物接枝纳米颗粒(PGN)中纳米尺度的界面化学、限制和结构变化(接枝密度、接枝尺寸和核尺寸分散)如何控制干层延伸和熔融层交错，从而导致PGN新出现的中尺度有序性、动力学和失效机制。理解潜在的分子机制的一个障碍是，现有的计算范式，如分子模拟，要么缺乏预测这些系统的结构、动力学以及因此的性质的保真度或效率。这个问题将通过一种新的尺度桥接方法来解决，该方法将使用特定于化学的放大方法将粗晶PGN进一步提高到纳米颗粒水平，然后使用来自两个层次上的精细模型的互补数据来通知潜在变量高斯过程机器学习模型。自适应机器学习模型将使在未知参数空间中快速查询PGN属性成为可能。该项目包括三个研究方向：1)推导多组分体系的PGN对势能；2)研究颗粒尺寸和接枝双分散性对结构和流变学的影响；3)开发用于多目标PGN设计的自适应机器学习模型。最终，这种新的框架将产生多组分的PGN，它将打破阻碍传统纳米复合材料的强度和韧性之间的权衡。这些研究活动还将解决材料信息学中的悬而未决的问题，如在机器学习中纳入代表化学基团的分类变量，寻找最佳模拟批次大小，以及量化因噪声测量而产生的不确定性。发现由于胶体转变而表现出力学性能增强的PGN对于它们成为可伸缩、可加工和广泛用于纳米复合材料至关重要。为了丰富K-12和STEM学生的本科课程，该团队将创建新的活动，包括模拟、聚合物、机器学习和材料设计的在线学习模块。其他活动包括通过预测科学和工程设计集群对研究生进行跨学科培训，分享力场和机器学习软件开发，软件商业化，以及为代表不足的群体提供新的研究机会。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The Division of Materials Research and the Division of Chemical, Bioengineering, Environmental and Transport Systems contribute funds to this award, which supports research to develop an accurate hierarchical modeling approach aimed at understanding the structural and mechanical properties of materials comprised only of inorganic nanoparticles chemically grafted with polymer chains. Polymer grafted nanoparticles represent a paradigm shift in how we fabricate, use, and recycle plastics because these composites have the potential to outperform traditional hybrids (i.e., physical mixtures of polymers and nanoparticles) in terms of strength and toughness. This is because the polymers are directly attached onto nanoparticles, thus ensuring that these materials constitute a chemically fixed mixture of polymers and nanoparticles – physical mixtures, on the other hand, tend to demix. However, variability in the size of the nanoparticles and their grafts, which is known as dispersity, results in unusual rheological and fracture properties that remain poorly understood. The proposed research will employ molecular simulations, theory, and machine learning to help explain how dispersity causes these properties and how it could be further exploited for discovering materials that surpass the current capabilities of nanocomposites. These research ideas are closely coupled to educational and outreach objectives to enrich the K-12 and undergraduate pipeline for STEM students. For this purpose, the team will create online learning modules for simulations, polymer physics, machine learning, and materials design. Other activities include interdisciplinary training of graduate students through the Predictive Science and Engineering Design Cluster at Northwestern University, sharing of force field and software developments, software commercialization, and new research opportunities for underrepresented student groups.The Division of Materials Research and the Division of Chemical, Bioengineering, Environmental and Transport Systems contribute funds to this award. The project addresses a critical knowledge gap in understanding how nanoscale interfacial chemistry, confinement and structural variations (grafting density, graft size and core size dispersities) in multi-component polymer-grafted nanoparticles (PGNs) govern dry layer extension and melt layer interdigitation, and consequently the emergent mesoscale ordering, dynamics, and failure mechanisms of PGNs. A barrier towards understanding the underlying molecular mechanisms is that existing computational paradigms, such as molecular simulations, either lack the fidelity or efficiency for predicting the structure, dynamics and hence properties of these systems. This issue will be addressed through a new scale-bridging method that will further coarse-grain PGNs to a nanoparticle level using chemistry-specific upscaling methods, and then use complementary data from finer models at two levels of hierarchy to inform Latent Variable Gaussian Process machine learning models. The adaptive machine learning model will enable rapid query of PGN properties in unexplored parametric spaces. The project encompasses three research thrusts: 1) Derivation of PGN pair potentials for multi-component systems; 2) Investigation of particle size and graft bidispersity effects on structure and rheology; 3) Development of an adaptive machine learning model for multi-objective design of PGNs. Ultimately, this novel framework will result in multicomponent PGNs that will break strength-toughness tradeoffs that hamper traditional nanocomposites. These research activities will also address open questions in materials informatics such as incorporation of categorical variables representing chemical groups in machine learning, finding optimal simulation batch sizes, and quantifying uncertainty due to noisy measurements. Discovering PGNs that exhibit mechanical property enhancements due to colloidal transitions is critical for them to become scalable, processible, and broadly useful as nanocomposites. To enrich the K-12 and undergraduate pipeline for STEM students, the team will create new activities including online learning modules for simulations, polymers, machine learning, and materials design. Other activities include the interdisciplinary training of graduate students through the Predictive Science and Engineering Design Cluster, sharing of force field and machine learning software developments, software commercialization, and new research opportunities for underrepresented groups.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.