Search for: All records

We present a new coarse-grained model and molecular simulation study of polysulfamide, a new class of polymer that could be a sustainable alternative to commodity polymers like polyurea. 

We present a computational method to analyze 2D small-angle scattering data from three-phase soft materials systems with structural anisotropy and output representative real-space structures of the three phases. 

Structural characterization of polymer materials is a major step in the process of creating materials' design-structural-property relationships. With growing interests in artificial intelligence (AI)-driven materials design and high-throughput synthesis and measurements, there is now a critical need for development of complementary data-driven approaches (e.g., machine learning models and workflows) to enable fast and automated interpretation of the characterization results. This review sets out with a description of the needs for machine learning specifically in the context of three commonly used structural characterization techniques for polymer materials: microscopy, scattering, and spectroscopy. Subsequently, a review of notable work done on development and application of machine learning models / workflows for these three types of measurements is provided. Definitions are provided for common machine learning terms to help readers who may be less familiar with the terminologies used in the context of machine learning. Finally, a perspective on the current challenges and potential opportunities to successfully integrate such data-driven methods in parallel/sequentially with the measurements is provided. The need for innovative interdisciplinary training programs for researchers regardless of their career path/employment in academia, national laboratories, or research and development in industry is highlighted as a strategy to overcome the challenge associated with the sharing and curation of data and unifying metadata. 

Collagen-like peptide heterotrimers are computationally designed to create percolated networks as a function of solvent quality and multifunctional materials of interest to the biomaterials community. 

In the field of materials science, microscopy is the first and often only accessible method for structural characterization. There is a growing interest in the development of machine learning methods that can automate the analysis and interpretation of microscopy images. Typically training of machine learning models requires large numbers of images with associated structural labels, however, manual labeling of images requires domain knowledge and is prone to human error and subjectivity. To overcome these limitations, we present a semi-supervised transfer learning approach that uses a small number of labeled microscopy images for training and performs as effectively as methods trained on significantly larger image datasets. Specifically, we train an image encoder with unlabeled images using self-supervised learning methods and use that encoder for transfer learning of different downstream image tasks (classification and segmentation) with a minimal number of labeled images for training. We test the transfer learning ability of two self-supervised learning methods: SimCLR and Barlow-Twins on transmission electron microscopy (TEM) images. We demonstrate in detail how this machine learning workflow applied to TEM images of protein nanowires enables automated classification of nanowire morphologies ( e.g. , single nanowires, nanowire bundles, phase separated) as well as segmentation tasks that can serve as groundwork for quantification of nanowire domain sizes and shape analysis. We also extend the application of the machine learning workflow to classification of nanoparticle morphologies and identification of different type of viruses from TEM images.