Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session S17: Chemical Physics at the Middle Scales of Soft Matter IIFocus
|
Hide Abstracts |
Sponsoring Units: DCP Room: Room 209 |
Thursday, March 9, 2023 8:00AM - 8:36AM |
S17.00001: Sparse, active learning of stochastic differential equations from data Invited Speaker: Benedikt Sabass Automatic machine learning of empirical models from experimental data is becoming feasible as a result of the increased availability of computational power and dedicated algorithms. I will discuss different approaches for the inference of governing equations from data. A robust method is proposed for the sparse solution of the inverse problem related to the inference of differential equations governing deterministic and stochastic systems. Next, I present a method that we call active learning of stochastic differential equations. In active learning, an inference of the stochastic dynamics is combined with perturbations to the measured system in a feedback loop. This procedure can significantly improve the inference of global models for systems with multiple energetic minima. If time remains, I will also discuss the use of deep convolutional networks and deep attention models for the automated analysis of biological cells. |
Thursday, March 9, 2023 8:36AM - 8:48AM |
S17.00002: Effects of many-body dispersion interactions on phase transitions of coarse-grained polymers Benedikt Ames, Mario Galante, Matteo Gori, Alexandre Tkatchenko Non-covalent interactions are crucial to the structure formation of soft matter. An important role is played by the van der Waals dispersion interaction, which is often modelled by efficient, but incomplete pairwise (PW) approximations. Substantially more accurate interaction energies are provided by the more recent many-body dispersion model (MBD) [1], but its relevance for (thermo)dynamic properties remains largely unexplored. In particular, since MBD predicts an increased effective interaction range which depends on the global configuration of the system, we anticipate enhanced cooperativity in conformational changes of polymers. |
Thursday, March 9, 2023 8:48AM - 9:00AM |
S17.00003: Machine learning enhanced computational reverse-engineering analysis for scattering experiments (CREASE) of soft materials to establish structure-property relationships Arthi Jayaraman, Christian M Heil Structural characterization is a critical step to establishing material design-structure-property relationships. To understand how the molecules assemble into structures and the morphologies they adopt at equilibrium and upon processing, one needs structural characterization at multiple length scales. Small angle scattering (SAS) enables 3D structural characterization across a wide range of length scales. The analysis of SAS results typically relies on fits using analytical models that are applicable for conventional shapes (e.g., spheres, cylinders, vesicles, etc.). With access to new soft materials chemistries and processing techniques, however, researchers observe unconventional structures for which existing analytical models are either too approximate or not applicable. Thus, there is a need for SAS analysis methods that do not rely on fitting the scattering profiles with analytical models. We have addressed this need with a new computational approach – Computational Reverse-Engineering Analysis for Scattering Experiments, CREASE. I will present how we use machine-learning enhanced CREASE to analyze SAS results from binary mixtures of spherical nanoparticles with varying composition, nanoparticle size distribution, and extent of mixing/aggregation. I will also show how we validate and then use the outputs from CREASE, in particular the 3D real-space configurations, as input to other computational methods that then predict macroscopic properties (e.g., optical response). |
Thursday, March 9, 2023 9:00AM - 9:12AM |
S17.00004: Stefan-Maxwell diffusivities in concentrated multi-species transport, and the Onsager's regression hypothesis. Maxim Zyskin, Charles W Monroe Stefan-Maxwell diffusivities play an important role in continuum models of multi-species transport because they describe drag between different species, as well as the entropy production rate. We investigate a lesser-known computational method for determining Stefan-Maxwell diffusivities that is better suited to the concentrated-solution regime. The method is based on Onsager's regression hypothesis, which in Casimir's interpretation states that autocorrelation functions of fluctuating quantities satisfy, in the linearized approximation, corresponding continuum model equations. We test this conjecture by comparing analytic computations of Stefan-Maxwell diffusivities for mixtures of Lennard-Jones gases, based on high-order expansions of the Enskog type, with molecular dynamics simulations that extract the same diffusivities from autocorrelation functions of Fourier-transformed species densities. We also compare the results with molecular dynamics diffusivity measurements by the familiar Green-Kubo approach. Analytic and molecular dynamics simulations of diffusivities match reasonably well, despite the fact that not all the assumptions of the regression-based continuum theory can be reproduced–or easily reproduced–in molecular dynamics simulations. We extend our study by testing the diffusivity measurement technique on complex liquids, implementing molecular dynamics simulation of Stefan-Maxwell diffusivities of liquid electrolytic solutions, which play an important role in battery modeling, and where robust computational methods covering a broad variety of salts and solvents compositions is much desired. For the electrolytes, analytic predictions are no longer available, but for the LiPF6 salts and a variety of solvents they have been measured experimentally in our group. |
Thursday, March 9, 2023 9:12AM - 9:24AM |
S17.00005: Collective Dynamics of Glass-Forming Liquids in Two and Three Dimensions Yangyang Wang, Michael S Jacobs, Zhiqiang Shen, Jan-Michael Carrillo, Bobby G Sumpter We evaluate the collective density-density and hydrostatic pressure-pressure correlations in two- and three-dimensional single-component coarse-grained polymer melts and binary Lennard-Jones liquids over a wide range of time and length scales using molecular dynamics simulations. At temperatures above the dynamic crossover, the mesoscopic collective dynamics in three-dimensional simulations can be qualitatively understood with the classical hydrodynamic theory. At lower temperatures, wavenumber-independent non-hydrodynamic modes emerge at long times. Such a connection between dynamic crossover and mesoscopic collective dynamics in 3D simualtions is critically tested in 2D systems and the results will be discussed in this talk. |
Thursday, March 9, 2023 9:24AM - 9:36AM |
S17.00006: Stability Study on Anion exchanged Donor-Acceptor based co-polymers Meghna Jha Semiconducting Polymers (SPs) have received widespread attention due to their promising qualities like superior absorbance/emission, easy chemical tunability, low-temperature solution processing, lightweight and flexible substrates, and low environmental toxicity. Molecular doping of SPs has been one of the most popular research topics in organic electronics over the past few decades. FeCl3 is a well-known p-type molecular dopant for SPs. Recently a new technique was introduced that is known as anion exchange that involves replacing the counter ion in a doped polymer with a more stable one to improve the longevity of the SP film. We show a D/A type co-polymer immersed in a dopant solution containing FeCl3 and the ionic liquid, LiTFSI. TFSI− counter ion is hydrophobic, stable under oxidizing conditions and weakly interacting with most cations. Under favorable conditions, TFSI- replaces the less stable FeCl4- counter ion. The main aim of this work is to study the degradation of SP films that are doped with FeCl3 vs that are anion exchanged. We perform experiments that include UV-Vis-NIR spectroscopy, 4-proble sheet resistance and Gas Chromatography-Mass Spectroscopy and X-Ray Absorption Spectroscopy to contrast the degradation between the two doped films. |
Thursday, March 9, 2023 9:36AM - 9:48AM |
S17.00007: Bottom-up coarse-graining scheme for the preservation of relevant slow degrees of freedom in soft materials. David G Rosenberger, Frank Noe, Cecilia Clementi, Andreas Bittracher, Clark Templeton, Wangfei Yang, Feliks Nüske Coarse graining approaches aim to recover relevant physical phenomena of molecular systems with computer simulations at lower resolutions than an atomistic representation. Ideally, the lower resolution still accounts for the degrees of freedom necessary to recover the correct physical behavior of the high resolution reference system, but with reduced computational cost. We make the argument that in soft matter contexts correct physical behavior corresponds to reproducing the long-time dynamics of a system.Thus, the retained degrees of freedom should correctly capture the rare-event transitions relevant for the long-time dynamics. Here, we propose a bottom-up coarse-graining scheme that correctly preserves the relevant slow degrees of freedom, and we test this idea for systems of increasing complexity. We show that in contrast to existing coarse-graining schemes, such as those from information theory or structure-based approaches, only the novel scheme is able to recapitulate the slow timescales of the system of interest. |
Thursday, March 9, 2023 9:48AM - 10:00AM |
S17.00008: Harnessing GPU-Enhanced Simulations for Efficient DFTB Metadynamics of Biochemical Systems Anshuman Kumar, Pablo Arantes, Aakash Saha, Giulia Palermo, Bryan M Wong The use of metadynamics to probe the thermodynamics of large chemical systems is computationally prohibitive due to the extensive sampling required to simulate the large degrees of freedom in these systems. To address this computational bottleneck, we utilize a heterogeneous CPU+GPU-enhanced density functional tight binding (DFTB) approach on Microsoft's Azure cloud platform to efficiently calculate the thermodynamics and metadynamics of biochemical systems. To first validate our approach, we calculate free energy surfaces of alanine dipeptide and show that our GPU-enhanced DFTB calculations qualitatively agree with computationally intensive hybrid DFT benchmarks, whereas classical force fields give significant errors. Most importantly, our GPU-accelerated DFTB calculations are significantly faster by up to 2 orders of magnitude. To further extend our GPU-enhanced DFTB approach, we also carried out a 10-ns metadynamics simulation of remdesivir, which is prohibitively out of reach for routine DFT-based metadynamics calculations. We find that the free energy surfaces of remdesivir obtained from DFTB and classical force fields differ, where the latter overestimates the internal energy contribution of high free energy states. Taken together, our benchmark tests, analyses, and extensions to large biochemical systems highlight the use of DFTB for efficiently predicting the free energy surfaces/thermodynamics of large biochemical systems. |
Thursday, March 9, 2023 10:00AM - 10:12AM |
S17.00009: Protein and Coupled Solvent Dynamics of Oligomeric and Fibrillar α-Synulein Katie Whitcomb, Kurt Warncke α-Synuclein (αS) is a protein that has an unspecified functional role in neurotransmitter release in brain neurons, and dysfunction associated with Parkinson's disease (PD).1 Monomeric αS is an intrinsically-disordered protein (IDP). Toward understanding the role of protein and coupled solvent dynamics in αS function and dysfunction, our established electron paramagnetic resonance (EPR) spin probe (TEMPOL) methodology was applied.2,3,4 Controlled-temperature ice-boundary confinement in frozen aqueous solutions of induced oligomeric and pre-formed fibrillar αS was used to localize TEMPOL to probe solvent phases specifically associated with αS. TEMPOL mobility in the presence of αS shows two distinct components at temperatures from 220 – 265 K, as for soluble globular proteins,5 but with dramatically higher fluidity. The temperature-dependence of the spin probe rotational correlation times and component weights, and hysteresis in these parameters for directional temperature change, are interpreted in terms of a high-fluidity, confinement-resistant aqueous-C-terminal protein domain (residues 96-140) mesophase. The results are relevant to the function of αS under high confinement conditions in the neuron presynaptic terminal region, and dysfunction, that involves oligomer and fibril permeation of phospholipid bilayer membranes. |
Thursday, March 9, 2023 10:12AM - 10:24AM |
S17.00010: Methodological advances in training bottom-up neural network coarse-grained force-fields Aleksander Durumeric, Frank Noe, Andreas Kraemer, Cecilia Clementi Bottom-up coarse-graining has recently been applied to parameterize neural network force-fields. However, progress in modeling small proteins has been slow. We here discuss two new methodological advances related to creating bottom-up coarse-grained neural network force-fields from reference atomistic data. These methods directly relate to improving and adapting Multiscale Coarse-Graining and generalized Yvon–Born–Green theory to the nonlinear data-hungry regime of neural network force-fields. We show that these approaches reduce the amount of data needed to accurately parameterize a coarse-grained model, sometimes with minimal additional computational cost. |
Thursday, March 9, 2023 10:24AM - 10:36AM |
S17.00011: Construction of phase diagram for binary polymer blend with unsupervised machine learning Inhyuk Jang, Arun Yethiraj Obtaining the phase diagram of complex fluid mixtures such as molecular liquid mixtures, polymer coacervates, and polymer blends is a challenging computational task. This is because traditional computational methods require the insertion of molecules, which is difficult for large molecules. In this work, we obtain the phase diagram of complex fluids using unsupervised machine learning (ML) methods. We show that construction of a feature vector is a crucial aspect to the success of ML methods. Using the ``affinity-based" feature, which assigns a "spin-like" variable to each atom, and the autoencoder we calculate the phase diagram of several model compounds. We show that the method is accurate for the phase diagram of simple binary mixtures and polymer blends, when compared to conventional methods. We also use the method to obtain the phase diagram of polymers in ionic liquids and address recent controversies. UML and local affinity feature vector open the new way to study phase behaviors of complex fluids without performing special simulations. |
Thursday, March 9, 2023 10:36AM - 10:48AM |
S17.00012: Efficient Sampling of Equilibrium Distributions with Scalable Autoregressive Flow Models Sherry Li, Grant M Rotskoff Deep generative models are a powerful tool for sampling equilibrium states of physical systems and estimating observable averages. However, training these models to sample complex molecular systems can be challenging since the underlying distributions are high-dimensional in nature and likely contain nonlocal interactions. We evaluate the quality of autoregressive flow models in the context of free energy estimation in solid lattice systems, using results from the Frenkel-Ladd technique as a benchmark. Moreover, the autoregressive structure enables partial reconstruction of the system of interest, which can be exploited to improve the scalability of the method. The autoregressive model can be trained on only parts of the system and autoregressively generate configurations of the entire system. |
Thursday, March 9, 2023 10:48AM - 11:00AM |
S17.00013: Neural-Network Pauli Repulsion Potentials for Density-Functional Tight Binding for Large Molecular Systems Leonardo Medrano Sandonas, Mirela Puleva, Martin Stoehr, Alexandre Tkatchenko Machine learning (ML) has been proven to be an extremely valuable tool for simulations with ab-initio accuracy at the computational cost between classical interatomic potentials and density-functional approximations. Similar efficiency can only be achieved by semi-empirical methods such as density-functional tight-binding (DFTB). In our previous work [J. Phys. Chem. Lett. 11, 16 (2021)], we substantially improved the accuracy of DFTB method for the prediction of multiple properties of small molecules by developing ML repulsive potentials (NNrep). However, these potentials do not account for long-range interactions which are crucial to investigate large/more flexible molecules and molecular dimers. Hence, we now employ a physics-inspired neural network architecture such as SpookyNet [Nat. Commun. 12, 7273 (2021)] to develop repulsive potentials supplemented with a treatment of many-body dispersion (MBD) interactions. In doing so, we show that the local and nonlocal interaction blocks in SpookyNet enhance the reliability of the MBD-corrected DFTB+NNrep models in performing modeling techniques like (global) structure search or vibrational analysis of flexible molecules and molecular dimers. Our study thus provides valuable insights for the fast access to reliable property calculation of diverse molecular systems. |
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