-
Adaptive Computing for Scale-up Problems
Authors:
Kevin Patrick Griffin,
Hilary Egan,
Marc T. Henry de Frahan,
Juliane Mueller,
Deepthi Vaidhynatha,
Dylan Wald,
Rohit Chintala,
Olga A. Doronina,
Hariswaran Sitaraman,
Ethan Young,
Ryan King,
Jibonananda Sanyal,
Marc Day,
Ross E. Larsen
Abstract:
Adaptive Computing is an application-agnostic outer loop framework to strategically deploy simulations and experiments to guide decision making for scale-up analysis. Resources are allocated over successive batches, which makes the allocation adaptive to some objective such as optimization or model training. The framework enables the characterization and management of uncertainties associated with…
▽ More
Adaptive Computing is an application-agnostic outer loop framework to strategically deploy simulations and experiments to guide decision making for scale-up analysis. Resources are allocated over successive batches, which makes the allocation adaptive to some objective such as optimization or model training. The framework enables the characterization and management of uncertainties associated with predictive models of complex systems when scale-up questions lead to significant model extrapolation. A key advancement of this framework is its integration of multi-fidelity surrogate modeling, uncertainty management, and automated orchestration of various computing and experimentation resources into a single integrated software package. This enables efficient multi-fidelity modeling across multiple computing resources by incorporating real-world constraints such as relative queue times and throughput on individual machines into the multi-fidelity sampling decision. We discuss applications of this framework to problems in the renewable energy space, including biofuels production, material synthesis, perovskite crystal growth, and building electrical loads.
△ Less
Submitted 28 March, 2025; v1 submitted 25 March, 2024;
originally announced April 2024.
-
Message-passing neural networks for high-throughput polymer screening
Authors:
Peter C. St. John,
Caleb Phillips,
Travis W. Kemper,
A. Nolan Wilson,
Michael F. Crowley,
Mark R. Nimlos,
Ross E. Larsen
Abstract:
Machine learning methods have shown promise in predicting molecular properties, and given sufficient training data machine learning approaches can enable rapid high-throughput virtual screening of large libraries of compounds. Graph-based neural network architectures have emerged in recent years as the most successful approach for predictions based on molecular structure, and have consistently ach…
▽ More
Machine learning methods have shown promise in predicting molecular properties, and given sufficient training data machine learning approaches can enable rapid high-throughput virtual screening of large libraries of compounds. Graph-based neural network architectures have emerged in recent years as the most successful approach for predictions based on molecular structure, and have consistently achieved the best performance on benchmark quantum chemical datasets. However, these models have typically required optimized 3D structural information for the molecule to achieve the highest accuracy. These 3D geometries are costly to compute for high levels of theory, limiting the applicability and practicality of machine learning methods in high-throughput screening applications. In this study, we present a new database of candidate molecules for organic photovoltaic applications, comprising approximately 91,000 unique chemical structures.Compared to existing datasets, this dataset contains substantially larger molecules (up to 200 atoms) as well as extrapolated properties for long polymer chains. We show that message-passing neural networks trained with and without 3D structural information for these molecules achieve similar accuracy, comparable to state-of-the-art methods on existing benchmark datasets. These results therefore emphasize that for larger molecules with practical applications, near-optimal prediction results can be obtained without using optimized 3D geometry as an input. We further show that learned molecular representations can be leveraged to reduce the training data required to transfer predictions to a new DFT functional.
△ Less
Submitted 5 April, 2019; v1 submitted 26 July, 2018;
originally announced July 2018.
-
Pump-probe polarized transient hole burning (PTHB) dynamics of hydrated electron revisited
Authors:
Ilya A. Shkrob,
Ross E. Larsen,
William J. Glover,
Benjamin J. Schwartz
Abstract:
Femtosecond PTHB spectroscopy was expected to demonstrate the existence of distinct s-p absorption subbands originating from the three nondegenerate p-like excited states of hydrated electron in anisotropic solvation cavity. Yet no conclusive experimental evidence either for this subband structure or the reorientation of the cavity on the picosecond time scale has been obtained. We demonstrate t…
▽ More
Femtosecond PTHB spectroscopy was expected to demonstrate the existence of distinct s-p absorption subbands originating from the three nondegenerate p-like excited states of hydrated electron in anisotropic solvation cavity. Yet no conclusive experimental evidence either for this subband structure or the reorientation of the cavity on the picosecond time scale has been obtained. We demonstrate that rapid reorientation of s-p transition dipole moments in response to small scale motion of water molecules is the likely culprit. The polarized bleach is shown to be too small and too short lived to be observed reliably on the sub-picosecond time scale.
△ Less
Submitted 20 June, 2007;
originally announced June 2007.
-
The structure of the hydrated electron. Part 2. A mixed quantum classical molecular dynamics - embedded cluster density functional theory: single-excitation configuration interaction study
Authors:
Ilya A. Shkrob,
William J. Glover,
Ross E. Larsen,
Benjamin J. Schwartz
Abstract:
Adiabatic mixed quantum/classical molecular dynamics simulations were used to generate snapshots of the hydrated electron (e-) in liquid water at 300 K. Water cluster anions that include two complete solvation shells centered on the e- were extracted from these simulations and embedded in a matrix of fractional point charges designed to represent the rest of the solvent. Density functional theor…
▽ More
Adiabatic mixed quantum/classical molecular dynamics simulations were used to generate snapshots of the hydrated electron (e-) in liquid water at 300 K. Water cluster anions that include two complete solvation shells centered on the e- were extracted from these simulations and embedded in a matrix of fractional point charges designed to represent the rest of the solvent. Density functional theory and single-excitation configuration interaction methods were then applied to these embedded clusters. The salient feature of these hybrid calculations is significant transfer (ca. 0.18) of the excess electron's charge density into the O 2p orbitals in OH groups forming the solvation cavity. We used the results of these calculations to examine the structure of the molecular orbitals, the density of states, the absorption spectra in the visible and ultraviolet, the hyperfine coupling (hfc) tensors, and the IR and Raman spectra of the e-. The calculated hfc tensors were used to compute the EPR and ESEEM spectra for the e- that compared favorably to the experimental spectra of trapped e- in alkaline ice. The calculated vibrational spectra of the e- are consistent with the red-shifted bending and stretching frequencies observed in resonance Raman experiments. The model also accounts for the VIS and 190-nm absorption bands of the e-. Thus, our study suggests that to explain several important experimentally observed properties of the e-, many-electron effects must be accounted for.
△ Less
Submitted 2 December, 2006; v1 submitted 20 November, 2006;
originally announced November 2006.
-
Instantaneous Pair Theory for High-Frequency Vibrational Energy Relaxation in Fluids
Authors:
Ross E. Larsen,
Richard M. Stratt
Abstract:
Notwithstanding the long and distinguished history of studies of vibrational energy relaxation, exactly how it is that high frequency vibrations manage to relax in a liquid remains somewhat of a mystery. Both experimental and theoretical approaches seem to say that there is a natural frequency range associated with intermolecular motions in liquids, typically spanning no more than a few hundred…
▽ More
Notwithstanding the long and distinguished history of studies of vibrational energy relaxation, exactly how it is that high frequency vibrations manage to relax in a liquid remains somewhat of a mystery. Both experimental and theoretical approaches seem to say that there is a natural frequency range associated with intermolecular motions in liquids, typically spanning no more than a few hundred cm^{-1}. Landau-Teller-like theories explain how a solvent can absorb any vibrational energy within this "band", but how is it that molecules can rid themselves of superfluous vibrational energies significantly in excess of these values? We develop a theory for such processes based on the idea that the crucial liquid motions are those that most rapidly modulate the force on the vibrating coordinate -- and that by far the most important of these motions are those involving what we have called the mutual nearest neighbors of the vibrating solute. Specifically, we suggest that whenever there is a single solvent molecule sufficiently close to the solute that the solvent and solute are each other's nearest neighbors, then the instantaneous scattering dynamics of the solute-solvent pair alone suffices to explain the high frequency relaxation. The many-body features of the liquid only appear in the guise of a purely equilibrium problem, that of finding the likelihood of particularly effective solvent arrangements around the solute. These results are tested numerically on model diatomic solutes dissolved in atomic fluids (including the experimentally and theoretically interesting case of I_2 in Xe). The instantaneous pair theory leads to results in quantitative agreement with those obtained from far more laborious exact molecular dynamics simulations.
△ Less
Submitted 26 October, 1998;
originally announced October 1998.
