Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Cluster computing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,405 records · Page 78
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Machine learning and shallow groundwater chemistry to identify geothermal prospects in the Great Basin, USA

This study discovers various geothermal prospects in the Great Basin, USA based on shallow groundwater chemical (geochemical) data. The geochemical data are expected to include hidden (latent) information that is a proxy for geothermal prospectivity. We processed the sparse geochemical data in the Great Basin at 14,341 locations including 18 attributes. Next, a non-negative matrix factorization with customized k-means clustering is applied to the geochemical data matrix that automatically finds three hidden geothermal signatures representing modestly, moderately, and highly confident geothermal prospects. The algorithm also evaluated the probability of occurrence of these types of resources through the studied region. There is a consistency between regional geothermal prospectivity as estimated by our ML methodology and the traditional play fairway analysis conducted over a portion of the study area. We also identify the dominant data attributes associated with each signature. Finally, our ML analyses allow us to reconstruct attributes from sparse into continuous over the study domain. The predicted continuous attributes can be used for future detailed geothermal explorations in the Great Basin.

15 GEOTHERMAL ENERGY↗
DOE OSTI2021

Baryonic effects in the Effective Field Theory of Large-Scale Structure and an analytic recipe for lensing in CMB-S4

Upcoming Large-Scale Structure surveys will likely become the next leading sources of cosmological information, making it crucial to have a precise understanding of the influence of baryons on cosmological observables. The Effective Field Theory of Large-Scale Structure (EFTofLSS) provides a consistent way to predict the clustering of dark matter and baryons on large scales, where their leading corrections in perturbation theory are given by a simple and calculable functional form even after the onset of baryonic processes. In this paper, we extend the two-fluid-like system up to two-loop order in perturbation theory. Along the way, we show that a new linear counterterm proportional to the relative velocity of the fluids could generically be present, but we show that its effects are expected to be small in our universe. Regardless, we show how to consistently perform perturbation theory in the presence of this new term. We find that the EFTofLSS at two-loop order can accurately account for the details of baryonic processes on large scales. Additionally, we compare our results to a hydrodynamical N-body simulation at many redshifts and find that the counterterms associated with the leading corrections to dark matter and baryons start to differ between redshifts z ≈ 3 and z ≈ 2, signaling the onset of star-formation physics. We then use these fits to compute the lensing power spectrum, show that the understanding of baryonic processes will be important for analyzing CMB-S4 data, and show that the two-loop EFTofLSS accurately captures these effects for ℓ ≲ 2000. Our results are also potentially of interest for current and future weak lensing surveys.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

On the Robustness of Network Community Structure Under Addition of Edges

Communities represent important functional modules in networked systems. A key goal in preserving such communities is understanding their robustness under perturbations. Previous research has studied the impact of node removals and edge removals on the community structure. However, the impact of edge additions on the robustness of the community structure is relatively unknown. Edge additions or false positive edges may simulate measurement errors or external exceptional events that threaten the functionality of networked systems. Here, we study the impact of edge additions on the community structure using Lancichinetti-Fortunato-Radicchi (LFR) benchmark networks. We show that, for a fixed network size, the impact of edge additions is greater on networks with initially weak community structure than on networks with strongly clustered structures. In addition, we find that the perception of the impact is also dependent on the community detection algorithm used to uncover communities. In particular, we found that modularity-based methods such as Leiden and Louvain are less affected than information-theoretic and message passing-based methods such as Infomap and Label Propagation. Our results demonstrate that edge addition can (a) significantly impact the community structure of networks based on their initial conditions, and (b) the perception of the impact is dependent on the community detection algorithm used. We describe limitations, open challenges, and how this methodology can inform the design of resilient networked systems under edge additions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Displacement Analysis of Geothermal Field Based on PSInSAR And SOM Clustering Algorithms A Case Study of Brady Field, Nevada—USA

The availability of free and high temporal resolution satellite data and advanced SAR techniques allows us to analyze ground displacement cost-effectively. Our aim was to properly define subsidence and uplift areas to delineate a geothermal field and perform time-series analysis to identify temporal trends. A Persistent Scatterer Interferometry (PSI) algorithm was used to estimate vertical displacement in the Brady geothermal field located in Nevada by analyzing 70 Sentinel-1A Synthetic-Aperture Radar (SAR) images, between January 2017 and December 2019. To classify zones affected by displacement, an unsupervised Self-Organizing Map (SOM) algorithm was applied to classify points based on their behavior in time, and those clusters were used to determine subsidence, uplift, and stable regions automatically. Finally, time-series analysis was applied to the clustered data to understand the inflection dates. The maximum subsidence is –19 mm/yr with an average value of –6 mm/yr within the geothermal field. The maximum uplift is 14 mm/yr with an average value of 4 mm/yr within the geothermal field. The uplift occurred on the NE of the field, where the injection wells are located. On the other hand, subsidence is concentrated on the SW of the field where the production wells are located. The coupling of the PSInSAR and the SOM algorithms was shown to be effective in analyzing the direction and pattern of the displacements observed in the field.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Description of Sensor Assignment Optimization Method as Deployed on a Multi-Node Cluster

Data analytic methods are being developed to address the problem of how to assign a sensor set in a nuclear facility such that a requisite level of process monitoring capability is realized and that the sensor set is sufficiently rich to determine the status of the individual sensors with respect to need for calibration. There is an awareness in the nuclear industry that data analytics combined with rich sensor sets represent a means to improve operations and reduce costs. In the industry the calibration problem has been previously approached as an empirical data-driven problem with several methods having been developed. However, the experience of the utilities over the past ten years with these methods indicates that the absence of physics-based information renders the data-driven approach less reliable. Complicating factors such as the inherent variability of operation (both equipment alignment and operating condition) can confound a pure data-driven approach while there are no rigorous guidelines for determining what constitutes an adequate sensor set. The solution under development to overcome these shortcomings supplements the data analytic method with process information in a so-called process-constrained data-analytic approach. Simple balance equations are written for generic components (e.g., mechanical pump, valve, and heat exchanger). These do not require a priori knowledge of process parameters, such as heat transfer coefficients or friction factors. All that is needed on the part of the utility user is to identify the components and how they are connected. This report describes the development of a parallel computing capability for determining the optimal sensor set. The optimal sensor set problem suffers from the curse of dimensionality. Computation time increases exponentially as the size of the system grows. To overcome this difficulty a pre-conditioner algorithm is developed to find an approximate solution close the actual solution. This serves as a seed for the full-blown algorithm and acts to constrain the space that must searched. The optimization algorithms are described and the implementation on a parallel computing platform is described. The application of the method to a use case we are solving in collaboration with our utility partner served to illustrate how the default sensor set in a nuclear plant may not provide sufficient coverage to infer sensor calibration status.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2021

A catalog of the diversity and ubiquity of bacterial microcompartments

Abstract Bacterial microcompartments (BMCs) are organelles that segregate segments of metabolic pathways which are incompatible with surrounding metabolism. BMCs consist of a selectively permeable shell, composed of three types of structurally conserved proteins, together with sequestered enzymes that vary among functionally distinct BMCs. Genes encoding shell proteins are typically clustered with those for the encapsulated enzymes. Here, we report that the number of identifiable BMC loci has increased twenty-fold since the last comprehensive census of 2014, and the number of distinct BMC types has doubled. The new BMC types expand the range of compartmentalized catalysis and suggest that there is more BMC biochemistry yet to be discovered. Our comprehensive catalog of BMCs provides a framework for their identification, correlation with bacterial niche adaptation, experimental characterization, and development of BMC-based nanoarchitectures for biomedical and bioengineering applications.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Neutron-neutron scattering length from the 6 He(p,pα)nn reaction

We propose a novel method to measure the neutron-neutron scattering length using the 6 He(p,pα)nn reaction in inverse kinematics at high energies. The method is based on the final-state interaction (FSI) between the neutrons after the sudden knockout of the α particle. We show that the details of the neutron-neutron relative-energy distribution allow for a precise extraction of the s-wave scattering length. Here, we present the state of the art in regard to the theory of this distribution. The distribution is calculated in two steps. First, we calculate the ground-state wave function of 6 He as a αnn three-body system. For this purpose we use Halo effective field theory, which also provides uncertainty estimates for the results. We compare our results at this stage to model calculations done with the computer code face. In a second step we determine the effects of the nn FSI using the nn t-matrix. We compare these FSI results to approximate FSI approaches based on standard FSI enhancement factors. While the final distribution is sensitive to the nn scattering length, it depends only weakly on the effective range. Throughout we emphasize the impact of theoretical uncertainties on the neutron-neutron relative-energy distribution, and discuss the extent to which those uncertainties limit the extraction of the neutron-neutron scattering length from the reaction 6 He(p,pα)nn.

6 ≤ A ≤ 19↗
DOE OSTI2025

Reduced-basis method for few-body bound-state emulation

Recent advances in both theoretical and computational methods have enabled large-scale, precision calculations of the properties of atomic nuclei. With the growing complexity of modern nuclear theory, however, also comes the need for novel methods to perform systematic studies and quantify the uncertainties of models when confronted with experimental data. Here, this study presents an application of such an approach, the reduced basis method, to substantially lower computational costs by constructing a significantly smaller Hamiltonian subspace informed by previous solutions. Our method shows comparable efficiency and accuracy to other dimensionality reduction techniques on an artificial three-body bound system while providing a richer representation of physical information in its projection and training subspace. This methodological advancement can be applied in other contexts and has the potential to greatly improve our ability to systematically explore theoretical models and thus enhance our understanding of the fundamental properties of nuclear systems.

cluster models↗
DOE OSTI2022

Simulations of x-ray absorption spectra for CO desorbing from Ru(0001) with transition-potential and time-dependent density functional theory approaches

The desorption of a carbon monoxide molecule from a Ru(0001) surface was studied by means of X-ray Absorption Spectra (XAS) computed with Transition Potential (TP-DFT) and Time Dependent (TD-DFT) DFT methods. By unraveling the evolution of the CO electronic structure upon desorption, we observed that at 2.3 Å from the surface, the CO molecule has already predominantly gas-phase character. While C 1s XAS is quite insensitive to changes in the C–O bond length, the O 1s excitation is very sensitive with the π* coming down in energy upon CO bond stretching, which competes with the increase in orbital energy due to the repulsive interaction with the metallic surface. We show in a systematic way that the TP-DFT method can describe the XAS rather well at the endpoints (chemisorbed and gas phase) but is affected by artificial charge transfer and/or incorrect spin treatment in the transition region in cases like CO, where there are low-lying π* orbitals and large exchange interactions between the core 1s and valence-acceptor π* orbitals. As an alternative, we demonstrate by comparing with experimental data that a linear response approach using TD-DFT employing common exchange-correlation functionals and finite-size clusters can yield a good description of the spectral evolution of the 1s → π* transition with correct spin and gas-to-chemisorbed chemical shifts in good agreement with experiment.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Kraken2 Metagenomic Virus Database

The Database: Kraken2 [1] database built from a classification tree containing over 700k metagenomic viruses from JGI IMG/VR [2]. (1) Wood, D. E., Lu, J., and Langmead, B. (2019). Improved metagenomic analysis with Kraken 2. Genome Biol., 20(1), 1–13. doi: 10.1186/s13059-019-1891-0 (2) Paez-Espino D, Chen I-MA, Palaniappan K, Ratner A, Chu K, Szeto E, et al. IMG/VR: a database of cultured and uncultured DNA Viruses and retroviruses. Nucleic Acids Res. 2017;45:D457-65. For Paper: Title: A k-mer based approach for virus classification in metatranscriptomic and metagenomic samples identifies viral associations in the Populus phytobiome and autism brains Abstract Background Viruses are an underrepresented taxa in the study and identification of microbiome constituents; however, they play an important role in health, microbiome regulation, and transfer of genetic material. Only a few thousand viruses have been isolated, sequenced, and assigned a taxonomy, which further limits the ability to identify and quantify viruses in the microbiome. Additionally, the vast diversity of viruses represents a challenge for classification, not only in constructing a viral taxonomy, but also in identifying similarities between a virus' genotype and its phenotype. However, the diversity of viral sequences can be leveraged to classify their sequences in metagenomic and metatranscriptomic samples. Methods To identify viruses in transcriptomic and genomic samples, we developed a dynamic programming algorithm for creating a classification tree out of 715,672 metagenome viruses. To create the classification tree, we clustered proportional similarity scores generated from the k-mer profiles of each of the metagenome viruses. We then integrated the viral classification tree with the NCBI taxonomy for use with ParaKraken, a metagenomic/transcriptomic classifier. Results To illustrate the breadth of our utility for classifying viruses with ParaKraken, we analyzed data from a plant metagenome study identifying the differences between two Populus genotypes in three different compartments and on a human metatranscriptome study identifying the differences between Autism Spectrum Disorder patients and controls in post mortem brain biopsies. In the Populus study, we identified genotype and compartment specific viral signatures, while in the Autism study we identified a significant increased abundance of eight viral sequences in Autism brain biopsies. Conclusion Viruses represent an important aspect of the microbiome. The ability to classify viruses represents the first step in being able to better understand their role in the microbiome. The viral classification method presented here allows for more complete identification of viral sequences for use in identifying associations between viruses and the host and viruses and other microbiome members. Acknowledgements and Funding This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. This research was also supported by the Plant-Microbe Interfaces Scientific Focus Area in the Genomic Science Program, the Office of Biological and Environmental Research (BER) in the U.S. Department of Energy Office of Science, and by the Department of Energy, Laboratory Directed Research and Development funding (ProjectID 8321), at the Oak Ridge National Laboratory. Oak Ridge National Laboratory is managed by UT-Battelle, LLC, for the US DOE under contract DE-AC05-00OR22725. This research used resources of the Compute and Data Environment for Science (CADES).

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Helium Bubble Growth and Retention Characteristics in Aging Palladium Tritide

The helium-related characteristics of aged palladium tritide, including tritium loss, the formation of He bubbles from tritium decay-induced He atoms, and He thermal desorption behavior are computed using a continuum model of bubble evolution that incorporates the log-normal bubble spacing distribution deduced from 3 He nuclear magnetic resonance (NMR) measurements. This spacing distribution produces significant differences between mean (expectation) values of the bubble size and pressure and the average (median) values of these aging characteristics obtained by a previous calculation. The new calculations find much of the He is retained in a large quantity of smaller bubbles with higher bubble pressures, improving the overall stability of the bubbles and extending the age of He retention. By contrast, the integrated tritide material characteristics of swelling and tritium pressure are found relatively insensitive to details of the bubble spacing distribution. Inter-bubble fracture is predicted to begin with small, closely-spaced bubbles and progress to include larger bubbles near the onset of rapid He release. The critical concentration for this onset increases almost linearly with material tensile strength and decreases with shear modulus. Release of He from the aging solid is modeled as a growing network of fractured inter-bubble ligaments. For small particle or small grain material, the physical size of the fracture cluster, relative to the material dimensions, becomes important. In contrast with normal aging, inter-bubble fracture occurring during thermal desorption appears to begin with large, widely-spaced bubbles, indicating elevated temperature techniques may be of limited use for evaluating development of the bubble fracture network.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2025

Feature engineering descriptors, transforms, and machine learning for grain boundaries and variable-sized atom clusters

Abstract Obtaining microscopic structure-property relationships for grain boundaries is challenging due to their complex atomic structures. Recent efforts use machine learning to derive these relationships, but the way the atomic grain boundary structure is represented can have a significant impact on the predictions. Key steps for property prediction common to grain boundaries and other variable-sized atom clustered structures include: (1) describing the atomic structure as a feature matrix, (2) transforming the variable-sized feature matrix to a fixed length common to all structures, and (3) applying a machine learning algorithm to predict properties from the transformed matrices. We examine how these steps and different combinations of engineered features impact the accuracy of grain boundary energy predictions using a database of over 7000 grain boundaries. Additionally, we assess how different engineered features support interpretability, offering insights into the physics of the structure-property relationships.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Isoenergetic two-photon excitation enhances solvent-to-solute excited-state proton transfer

Two-photon excitation (TPE) is an attractive means for controlling chemistry in both space and time. Since isoenergetic one- and twophoton excitations (OPE and TPE) in non-centrosymmetric molecules are allowed to reach the same excited state, it is usually assumed that they produce similar excited-state reactivity. We compare the solvent-to-solute excited-state proton transfer of the super photobase FR0-SB following isoenergetic OPE and TPE. We find up to 62% increased reactivity following TPE compared to OPE. From steady-state spectroscopy, we rule out the involvement of different excited states and find that OPE and TPE spectra are identical in non-polar solvents but not in polar ones. We propose that differences in the matrix elements that contribute to the two-photon absorption cross sections lead to the observed enhanced isoenergetic reactivity, consistent with the predictions of our high-level coupledcluster-based computational protocol. We find that polar solvent configurations favor greater dipole moment change between ground and excited states, which enters the probability for TPE as the absolute value squared. This, in turn, causes a difference in the Franck– Condon region reached via TPE compared to OPE. We conclude that a new method has been found for controlling chemical reactivity via the matrix elements that affect two-photon cross sections, which may be of great utility for spatial and temporal precision chemistry.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Disorder, interactions, and their interplay in novel narrow-gap Dirac materials and Weyl semimetals

Progress of the modern day condensed matter physics is to a large extent driven by the synthesis of new materials, advances in their experimental characterization and theoretical description. Recent discoveries of novel gapless Weyl semimetals, such as NaBi,CdAs, and BiTe-based films, in which magnetic dopants essentially suppress the gap, have added to the family of graphene and topological insulators actively investigated over the past decade. With the field of novel semimetals rapidly maturing, its focus necessarily shifts from demonstrations of the feasibility of such materials to their quantitative characterization. While the transport and optical properties of graphene and topological insulators are well captured within the picture of free non-interacting electrons, gapless 3D Weyl semimetals and narrow-gap 2D semiconductors with Dirac spectrum are known to be extremely susceptible to disorder and electron-electron interactions. This susceptibility obscures the manifestations of nontrivial band structure -- like quantum anomalous Hall effect -- of the new topological materials. Among particular projects to be addressed are: 1) optical conductivity of 3D gapless Dirac fermions in the presence of smooth disorder, 2) interplay of disorder and Coulomb interactions in the spectral properties of such fermions, 3) formation and structure of the impurity band with Coulomb supercritical clusters, 4) Coulomb interaction-driven renormalization of the electron spectrum and of the transport response in the presence of strong magnetic field, 5) instanton approach to the disorder-induced fluctuation states in zero-gap 3D materials, and 6) the role of disorder in quantum anomalous Hall effect. The proposal relies upon the investigators' previous broad expertise in interacting and disordered electron systems. The methods to be employed include perturbative diagrammatic technique, non-perturbative instanton and self-consistent approximations, hydrodynamics of electron liquid. Both analytical as well as numerical approaches are to be employed. The anticipated broader outcome of the proposal includes gaining an in-depth understanding of the interplay of the disorder and interactions under the conditions when this interplay has the most dramatic impact on observables. Traditionally, interaction effects are among the most challenging and interesting problems of condensed matter physics. Similarly, disordered systems typically present very difficult but extremely rich problems in the description of various materials. Importantly, understanding the spectral and transport properties of such materials not only presents the fundamental objective, but is also of particular interest for many applications, such as computation, memory, optics, plasmonics. In particular realization of the quantum anomalous Hall effect may lead to the development of low-power-consumption electronics. Indeed, a major constraint for practical use of the quantum Hall effect is limited by the requirement of the quantizing magnetic field. At the same time, the quantum anomalous Hall effect samples exhibit non-dissipative edge quantum transport in a zero magnetic field.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Machine learning to identify geologic factors associated with production in geothermal fields: a case-study using 3D geologic data, Brady geothermal field, Nevada

Abstract In this paper, we present an analysis using unsupervised machine learning (ML) to identify the key geologic factors that contribute to the geothermal production in Brady geothermal field. Brady is a hydrothermal system in northwestern Nevada that supports both electricity production and direct use of hydrothermal fluids. Transmissive fluid-flow pathways are relatively rare in the subsurface, but are critical components of hydrothermal systems like Brady and many other types of fluid-flow systems in fractured rock. Here, we analyze geologic data with ML methods to unravel the local geologic controls on these pathways. The ML method, non-negative matrix factorization with k -means clustering (NMF k ), is applied to a library of 14 3D geologic characteristics hypothesized to control hydrothermal circulation in the Brady geothermal field. Our results indicate that macro-scale faults and a local step-over in the fault system preferentially occur along production wells when compared to injection wells and non-productive wells. We infer that these are the key geologic characteristics that control the through-going hydrothermal transmission pathways at Brady. Our results demonstrate: (1) the specific geologic controls on the Brady hydrothermal system and (2) the efficacy of pairing ML techniques with 3D geologic characterization to enhance the understanding of subsurface processes.

58 GEOSCIENCES↗
DOE OSTI2025

Global Archaeal Diversity Revealed Through Massive Data Integration: Uncovering Just Tip of Iceberg

The domain of Archaea has gathered significant interest for its ecological and biotechnological potential and its role in helping us to understand the evolutionary history of Eukaryotes. In comparison to the bacterial domain, the number of adequately described members in Archaea is relatively low, with less than 1000 species described. It is not clear whether this is solely due to the cultivation difficulty of its members or, indeed, the domain is characterized by evolutionary constraints that keep the number of species relatively low. Based on molecular evidence that bypasses the difficulties of formal cultivation and characterization, several novel clades have been proposed, enabling insights into their metabolism and physiology. Given the extent of global sampling and sequencing efforts, it is now possible and meaningful to question the magnitude of global archaeal diversity based on molecular evidence. To do so, we extracted all sequences classified as Archaea from 500 thousand amplicon samples available in public repositories. After processing through our highly conservative pipeline, we named this comprehensive resource the ‘Global Archaea Diversity’ (GAD), which encompassed nearly 3 million molecular species clusters at 97% similarity, and organized it into over 500 thousand genera and nearly 100 thousand families. Saline environments have contributed the most to the novel taxa of this previously unseen diversity. The majority of those 16S rRNA gene sequence fragments were verified by matches in metagenomic datasets from IMG/M. These findings reveal a vast and previously overlooked diversity within the Archaea, offering insights into their ecological roles and evolutionary importance while establishing a foundation for the future study and characterization of this intriguing domain of life.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

An interregional optimization approach for time series aggregation in continent-scale electricity system models

Modeling electric power systems with high shares of weather-dependent resources requires tradeoffs between temporal, spatial, and operational resolution. Many studies perform time series aggregation using clustering algorithms to reduce the temporal dimension, but when modeling continent-scale electricity systems that are large enough to contain multiple independent weather systems, this approach requires large numbers of representative periods to minimize errors in regional wind and solar capacity factors. Here, a new optimization-based approach for representative period selection and weighting is introduced that minimizes regional errors in average renewable capacity factors and electricity demand. The method delivers higher regional fidelity with fewer representative periods than alternative clustering methods when applied to wind, solar, and demand profiles for the contiguous United States. When representative periods are selected from multiple weather years, the optimized method reproduces regional averages with lower error than a complete 365-day time series from any single weather year. The method identifies only representative (as opposed to outlying) periods but can be combined with an iterative "stress period" identification approach to guide efficient decision-making considering both average and high-risk weather conditions.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2020

Thermochemical Data Fusion Using Graph Representation Learning

Large databases are required for “Big Data” applications in catalysis and materials science. Thermochemical databases can be created by combining data from various sources and by correcting low-fidelity datasets to higher accuracy with minimal computation. To achieve this “data fusion”, thermochemical quantities of interest, calculated at various levels of density functional theory (DFT), need to be mapped to the same, high levels of theory. In this work, a graph theoretical, statistical framework is proposed for such tasks. Subgraph frequencies are shown to provide a natural representation for learning these fusion maps. The maps are linear and are learnt with automated descriptor selection. Using a dataset of as few as ~1% from the QM9 database of 133,885 molecules, these models can predict multiple thermochemical quantities at a higher level of theory with an accuracy of 1 kcal/mol. Here, the method is explainable, generalizable, and provides a diagnostic tool for outlier identification

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗