CDS&E: Toward a Pattern Recognition Framework to Identify Reaction Coordinates for Order-Disorder Transitions: Application to Block Copolymers

非技术性摘要该奖项支持计算研究和教育，以开发一种新的方法来确定原子或分子排列中包含的组织程度，因为它从无序或无序排列变为更有组织或有序的排列，特别是在材料的计算机模拟中。在许多材料的加工过程中，从长链状分子或聚合物制成的软材料到金属合金等无机固体，经常发生转变，涉及原子或分子自组织方式的性质变化，从更无序的排列到更有序的排列。这种内部秩序的变化通常会转化为材料性质的实质性变化。在许多有前途的构建块可用于产生所需的有序结构，嵌段共聚物，一类聚合物，脱颖而出，因为各种材料可以由它们制成具有高度规则的图案和纳米尺度的孔，在半导体薄膜，过滤器和导电膜的制造中的潜在应用。然而，实现这种高度规则的结构是困难的，往往涉及更多的艺术而不是科学。理解和控制微观有序的最重要的是要有良好的“有序参数”，研究人员可以使用这些参数来监控朝着所需的良好组织或有序状态的进展。这些序参量将大量微观和原子尺度的数据提炼成几个关键的量;目前，它们是以个案的方式开发的，通常涉及大量的计算工作和试错。在这个项目中，PI的目标是推进一个框架，以帮助识别有用的顺序参数的基础上，从模式识别领域的想法。就像有效的计算机程序已经被开发和部署用于人脸识别一样，例如，PI旨在开发描述伴随微观秩序的开始和传播的关键几何签名的技术。PI与实验组的密切合作将确保计算和实验工作之间的协同作用。该项目提供了一个环境，以培训和支持一名博士生，并为两名本科生提供研究经验，其中一人将开发推广教育模块。这项研究的结果也将用于PI教授的高级化学工程建模课程。非技术摘要该奖项支持以应用和开发模式识别（PR）为目标的计算研究和教育算法，其将允许用户从包含系统的配置文件的大量快照的输入轨迹数据中筛选出并识别转变动力学的良好描述符，从一种秩序状态到另一种秩序状态特别是，该提案寻求基于模式识别的反应坐标来跟踪嵌段共聚物系统中涉及的有序-无序相变的动力学，强调形成双连续相的不太清楚的过程，如gyroid，double diamond和plumbers nightmare，所有这些都由周期性交织的3D网络组成。PI设想，通过识别以这种方式发现的适当反应坐标，将获得对有序过程的更完整和直观的理解，以及如何将系统引导到导致不同稳定和亚稳态的替代途径。PI计划执行以下活动：（a）确定所寻求的双连续相形成的条件，需要自由能模拟以区分稳定相和亚稳相并产生感兴趣的参考有序结构的步骤。(b)模拟不同过冷条件下的过渡路径集合数据。(c)从步骤（a）中模拟的阶段开发候选的基于模式识别的描述符，并基于（B）中收集的数据评估其质量。(d)应用最佳反应坐标，从而获得物理洞察的动力学过程和相分叉。这项调查将提供模式识别算法和模式识别为基础的描述符被证明是有助于描述几何复杂的有序相的形成动力学。该方法可以找到许多新的应用，包括热，激光和溶剂退火的动力学研究，并在图案化基板上的嵌段共聚物薄膜的定向组装。

NONTECHNICAL ABSTRACTThis award supports computational research and education to develop a new method to identify the degree of organization contained in an arrangement of atoms or molecules as it changes from a disorganized or disordered arrangement to a more organized or ordered arrangement, particularly in computer simulations of materials. In the processing of many materials, from soft materials made of long chain-like molecules, or polymers, to inorganic solids like metal alloys, transformations often occur that involve changes in the nature of the way atoms or molecules self-organize from a more disordered arrangement to a more ordered arrangement. Such changes in internal order typically translate into substantial changes in the properties of the material. Among the many promising building blocks available to generate desirable ordered structures, block copolymers, a class of polymers, stand out because a variety of materials can be made from them with highly regular patterns and pores of nanoscale dimensions, for potential applications in the making of semiconductor films, filters, and conducting membranes. However, the realization of such highly regular structures is difficult and often involves more art than science. Paramount to understanding and controlling microscopic ordering is to have good "order parameters," quantities that researchers can use to monitor the progress towards the desired well organized or ordered state. Such order parameters distill a huge amount of microscopic and atomic-scale data into just a few crucial quantities; currently, they are developed in a case-by-case fashion, often involving significant computational effort and trial and error. In this project, the PI aims to advance a framework to help identify useful order parameters based on ideas from the area of pattern recognition. Just like effective computer programs have been developed and deployed for face recognition, for example, the PI aims to develop techniques for describing the key geometrical signatures that accompany the onset and propagation of microscopic order. The close collaboration of the PI with experimental groups will ensure a synergy between computational and experimental efforts. The project provides an environment to train and support a doctoral student and to provide research experience for two undergraduate students, one of whom will develop outreach educational modules. Results from this investigation will also be used in advanced chemical engineering modeling classes taught by the PI.NONTECHNICAL ABSTRACT This award supports computational research and education with the goal to apply and develop pattern recognition (PR) algorithms that would allow users to screen out and identify good descriptors of the transition kinetics from input trajectory data containing a large number of snapshots of configuration files of the system as it goes from one state of order to another. In particular, this proposal seeks pattern-recognition-based reaction coordinates to track the kinetics of order-disorder phase transitions involved in block copolymer systems, emphasizing the not-well understood processes that form bicontinuous phases like gyroid, double diamond, and plumbers nightmare, all made of periodic interweaving 3D networks. The PI envisions that by identifying appropriate reaction coordinates discovered in this way, a more complete and intuitive understanding will be attained of the ordering process and how the system could be steered into alternative pathways that lead to different stable and metastable states.The PI plans to execute the following activities: (a) Determine the conditions at which sought after bicontinuous phases form, a step entailing free energy simulations to discriminate stable from metastable phases and generation of the reference ordered structures of interest. (b) Simulate transition-path ensemble data for varying supercooling conditions. (c) Develop candidate pattern-recognition-based descriptors from the phases simulated in step (a), and assess their quality based on data collected in (b). (d) Apply the best reaction coordinate thus found to gain physical insight into the kinetics of the process and phase bifurcation. This investigation will provide pattern-recognition algorithms and pattern-recognition-based descriptors proven to be helpful in describing the formation kinetics of geometrically complex ordered phases. The method can find numerous new applications, including kinetic studies of thermal, laser and solvent annealing, and directed assembly of block copolymer thin films on patterned substrates.