Bulletin of the American Physical Society
APS March Meeting 2019
Volume 64, Number 2
Monday–Friday, March 4–8, 2019; Boston, Massachusetts
Session B55: Big Data in Polymer and Soft Matter PhysicsFocus
|
Hide Abstracts |
Sponsoring Units: DPOLY GSOFT DCOMP Chair: Jonathan Whitmer, University of Notre Dame Room: BCEC 254B |
Monday, March 4, 2019 11:15AM - 11:27AM |
B55.00001: Descriptor of both domain structures and bridge chain network in a thermoplastic elastomer using graph Hiroshi Morita, Ayano Miyamoto The ABA triblock copolymer take micro-phase separated structure, and ABA chains take bridge or loop chains. In the elongation process, the coalescences and breaks of domains occurs and the conformations of ABA chains are also changed. To describe the structures of both domains and chain conformations, the simple descriptor for thermoplastic elastomer using graph theory is proposed. We performed the coarse-grained molecular dynamics simulation based on the TPE model by Aoyagi et al, and analyzed these structures using graph. |
Monday, March 4, 2019 11:27AM - 11:39AM |
B55.00002: Machine Learning for Modelling Microstructure Evolution in Polymer Mixtures Nigel Clarke Our aim is to enhance our modelling capabilities for microstructure evolution with machine learning. In particular, we focus on Gaussian processes (GPs), a popular non-parametric class of models used extensively in ML and uncertainty quantification, which have well documented predictive abilities. In our preliminary studies, we apply existing GP methodology to microstructure evolution to determine the feasibility of generating an emulator to supplement more traditional, computationally intense, approaches. As an exemplar, we focus on the non-linear Cahn-Hilliard equation for describing phase separation in blends. Spatio-temporal problems are particularly challenging for ML due to their high dimensionaility, hence we use a method recently proposed for using machine learning to predict video images, based on the idea of light cones, in which the present is only dependent on the past in the immediate spatial neighbourhood, analogous to real-space time-stepping numerical schemes for PDEs. We will present results which highlight the both the strengths and challenges of using ML for modelling microstructure. |
Monday, March 4, 2019 11:39AM - 11:51AM |
B55.00003: Impact of Dataset Uncertainties on Machine Learning Model predictions: The Example of Polymer Glass Transition Temperatures Anurag Jha, Anand Chandrasekaran, Chiho Kim, Ramamurthy Ramprasad Data-driven methods are seeing a revival and are deeply influencing multiple aspects of materials research. Materials property data from computations or experiments, are being utilized to create surrogate models using machine learning (ML) techniques. These models can be utilized to provide rapid predictions of the properties of new materials at a fraction of the cost compared to actual experimentation or computation. Moreover, a variety of techniques are being explored to “invert” the property prediction pipeline to allow for designing materials with desired target set of property values. The quality of the developed surrogate model, depends on the quality (and quantity) of the dataset used in the model training. Often, different experimental studies may report different values for the same property of the same material. This may be due to variations in measurement techniques, conditions, and sample quality among others. How should one treat these variances and what is their impact ? This question needs to be answered specifically, since it is paramount to the development of a good prediction model and helps understand its limitations. |
Monday, March 4, 2019 11:51AM - 12:03PM |
B55.00004: Molecular dynamics and machine learning assisted design of conjugated polymers for improved ionic conductivity Christian Nowak, Mayank Misra, Fernando A Escobedo Ionic transport in conjugated polymers is an area of increasing interest for applications involving such devices as sensors, batteries, and electronic ion pumps. In a previous work we used computational models to investigate the effect on ionic conductivity of specific chemistries of side chains attached to a polythiophene (PT) backbone, and made predictions that have been confirmed by experiments. Here we extend our materials design approach such that a chemical group “library” is considered from which polymers can be constructed. We employ two differing machine learning approaches to converge on a general set of high performing chemistries: a genetic algorithm and a neural network. We only consider PT-like molecules and take advantage of the fact that the crystalline arrangements simplify the task of microstructure modeling/prediction. We find some general trends for designing materials with improved ionic conductivity; namely, biasing the ion solvating groups towards the chain ends while still retaining good percolation of the solvation sites. Our ongoing efforts are focused on extending this scheme for optimization ionic conductivity to molecules with more complex morphologies such as Bolaamphiphiles. |
Monday, March 4, 2019 12:03PM - 12:15PM |
B55.00005: A charge density prediction model for organic molecules using Deep Neural Networks Deepak Kamal, Anand Chandrashekaran, Ramamurthy Ramprasad In an attempt to accelerate the pace of materials discovery, the community is increasingly using machine learning (ML) based techniques to rapidly map structure property relationships in materials. The Machine Learning methods has the advantage that their computational costs are negligible compared to those of the customary methods like density functional theory (DFT). The simulation properties of polymers is one such example where this limitation is most apparent . Here we propose a method to accurately predict charge densities of large organic systems by learning from pre-calculated examples of smaller systems. We present a novel fingerprint scheme which can numerically represent the local atomic environment. We then use a Neural Networks based models to learn the functional map between the local environments and the charge densities. Further, we introduce a recursive approach to improve these models by selectively incorporating poorly predicted examples. These charge densities are then used to calculate properties which depend on the charge density distribution like the dipole moments, dielectric properties for a host of unseen molecules thus eliminating the need to explicitly solve the laborious Kohn-Sham equations. |
Monday, March 4, 2019 12:15PM - 12:27PM |
B55.00006: ABSTRACT WITHDRAWN
|
Monday, March 4, 2019 12:27PM - 1:03PM |
B55.00007: Effectively Exploring Parameter Space: Design of Experiments and Machine Learning-assisted Organic Solar Cell Efficiency Optimization Invited Speaker: Erik Luber Organic solar cells (OSCs) are a potential cost-effective way to transform solar energy into electricity due to their potential for low-cost and high-throughput roll-to-roll production.[1] Improving the power conversion efficiency (PCE) and stability of OSCs are two of the most important tasks on the way toward commercialization. While much effort has been focused on developing new materials, optimization of processing conditions is equally important, where optimization is typically done in a haphazard manner using the experimenter's "intuition" or through one-variable-at-a-time (Edisonian) manipulation. However, such methods can fail to find the maximum PCE due to the high dimensionality parameter space of processing conditions and correlations between parameters. Moreover, laboratory-scale OSC fabrication is often low-throughput, time-consuming and expensive. Herein, we report an approach that uses Design of Experiments (DOE) along with machine learning (ML) to optimize solar cell efficiency. DoE is used to systematically explore the parameter space of processing conditions and ML is then utilized to estimate the PCE landscape as a function of the processing parameters. This process is then applied recursively to successively smaller regions of parameters space in regions of interest. Utilizing this process allows experimentalists to explore a larger parameter space with fewer experimental trials while obtaining valid and objective conclusions. Specific examples of concrete improvement of the power conversion efficiency of OSCs will be described. |
Monday, March 4, 2019 1:03PM - 1:15PM |
B55.00008: Using Particle Swarm Optimization and SCFT to agnostically identify the stable and low-lying metastable competitive morphologies of block copolymers. Carol Tsai, Kris T Delaney, Glenn Fredrickson The unguided search for the stable phase of a block copolymer of a given composition and architecture is a problem of global optimization with important ramifications from a materials discovery perspective. We discuss the development of a reciprocal-space Particle Swarm Optimization (PSO)-SCFT method in which Fourier components of SCFT fields near the principal shell are manipulated. Effectively, PSO-SCFT facilitates the search through a space of reciprocal-space SCFT seeds which yield a variety of morphologies. Using intensive free energy as a fitness metric by which to compare these morphologies, the PSO-SCFT methodology allows us to agnostically identify low-lying competitive and stable morphologies. In this talk, we present results for applying PSO-SCFT to conformationally symmetric diblock copolymers and miktoarm star polymers, and discuss the successes and challenges of the method. |
Monday, March 4, 2019 1:15PM - 1:27PM |
B55.00009: Prediction of suitable solvents and non-solvents for polymers using machine learning techniques Shruti Venkatram, Chiho Kim, Anand Chandrasekaran, Ramamurthy Ramprasad Solvent selection is essential for formulations in industrial and research processes like paints, cosmetics and pharmaceuticals. Identifying appropriate solvents for a polymer formulation is usually done by trial-and-error, and therefore, is time-consuming. To mitigate this problem, quantitative measures of solvent-polymer miscibility known as solubility parameters have been developed in the past. In the present study, we first assessed the performance of the Hildebrand solubility parameter to predict solvents for a set of benchmark polymers. Machine learning techniques, trained on a dataset of known polymer Hildebrand solubility parameters, were then used to predict the solubility parameter of a queried polymer. Matching the predicted value with known solvent solubility parameters was then utilized to identify suitable solvents and non-solvents for the queried polymer. This capability has been implemented at www.polymergenome.org. |
Monday, March 4, 2019 1:27PM - 1:39PM |
B55.00010: Applying Machine Learning to Structural Analysis using Pythia Matthew Spellings, Julia Dshemuchadse, Sharon Glotzer The recent explosion of interest and progress in machine learning (ML) methods has driven a proliferation of their application to soft matter systems. ML promises to deliver novel, automatic characterization techniques to solve previously insurmountable problems and it has already been successfully applied in several key areas for both disordered and ordered materials. However, researchers attempting to utilize ML methods often encounter challenges in finding the most appropriate representation of their data. To help alleviate this problem and foster reproducibility in these applications, we present Pythia, an open-source Python library for generating numerical descriptions of particle configurations. Pythia provides a palette of descriptors for users to select from, ranging from the simple to sophisticated. We demonstrate how Pythia can be combined with standard ML methods to quickly identify structures, analyze crystal grains, and study nucleation and growth of complex colloidal crystalline phases—all in a high-throughput manner. |
Monday, March 4, 2019 1:39PM - 1:51PM |
B55.00011: End-to-End Characterization of Colloidal Particles through
Holographic Microscopy and Deep Convolutional Neural Networks Lauren Altman, David Grier, Mark D Hannel II Analyzing holograms of colloidal particles with Lorenz-Mie |
Monday, March 4, 2019 1:51PM - 2:03PM |
B55.00012: An Anisotropic Langevin Equation for Protein Dynamics Eric Beyerle, Marina Giuseppina Guenza The analysis and description of protein motions is greatly facilitated by reducing the effective dimensionality of the system through coarse-graining and transforming to decoupled normal-mode coordinates. We have developed a coarse-grained, diffusive, Langevin equation to model protein dynamics, the Langevin Equation for Protein Dynamics (LE4PD), which accounts for hydrodynamic effects and mode-dependent free-energy barriers. Here, we extend the LE4PD to describe anisotropic, directional fluctuations of a protein’s residues, projected along the alpha-carbons. We compare the dynamics predicted by the LE4PD to a conventional method to model protein dynamics, principal component analysis (PCA), which does not account for free-energy barriers or possess an associated equation of motion. Testing the formalism on a molecular dynamics simulation of ubiquitin, coarse-grained at the alpha-carbon level, when both free-energy barriers and hydrodynamic effects are neglected, the normal modes predicted by both methods are identical. However, we find that including the barriers and hydrodynamic effects in the mode-dependent description can alter significantly the predicted kinetic and dynamic properties of the protein. |
Monday, March 4, 2019 2:03PM - 2:15PM |
B55.00013: ABSTRACT WITHDRAWN
|
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700