Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Data Reasoning”

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 325 records · Page 18
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Visualization for Insight and Data Analysis in Energy Research

This talk explores how advanced visualization technologies are transforming analytical reasoning and knowledge discovery in energy research, drawing on recent work at the National Laboratory of the Rockies' Computational Science Center. Through a series of scientific case studies, we demonstrate how immersive and high-resolution visualization environments enable scientists and engineers to identify previously unseen patterns and features - insights that often remain hidden in traditional desktop-based analysis. By embedding richer information into interactive analytics tools, these approaches support the exploration of complex, multivariate parameter spaces, where interaction itself catalyzes understanding. Beyond capability, we emphasize the critical role of visualization design grounded in perception and cognition, showing how visual encodings directly influence analytical outcomes. Spanning applications from materials science to integrated energy systems, these visualization approaches accelerate innovation and improve decision-making by enabling deeper, more reliable insight into increasingly complex energy data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A reduced order sulfuric acid decomposition model for a nuclear-powered hybrid sulfur cycle

A reduced order model of sulfuric acid decomposition within a bayonet chemical reactor was developed to support the U. S. Department of Energy Integrated Energy System program, and address the lack of knowledge in scaling and integration for joint chemical and nuclear processes. Sulfuric acid decomposition within a bayonet reactor was modeled to provide chemical and thermodynamic data relevant to advanced nuclear reactor-driven integrated energy systems based on desired operational scale and operational conditions. The temperature range required for high-temperature advanced nuclear reactor integrated energy systems, 750-850 °C, was shown to produce reasonable agreement (within a few percent relative error) with past models and experimental data, and yielded good efficiency results for bayonet reactor operations. The results of the reduced order model agreed with previous work from Savannah River National Labs within a maximum of 3.4% error on the decomposition of sulfur trioxide, and on previous Hybrid Sulfur flowsheets from Gorensek and Summers that showed operational temperature, pressure, and composition ranges for efficiency which made the Hybrid Sulfur cycle competitive with water electrolysis. Here, the agreement with previous high-fidelity models provided a framework for future Integrated Energy System grid evaluations with an advanced nuclear reactor and large-scale hydrogen production using a mathematical model to represent chemical operations.

08 HYDROGEN↗
DOE OSTI2022

MetaText: Compositional Generalization in Deep Language Models

Compositional generalization, the ability to infer and reason using novel combinations of previously encountered entities and structures, is a trait of great utility across a variety of deep learning tasks; it implies a certain ability to reason using consistent and therefore interpretable rules, to understand the constituent parts of its input at multiple levels of granularity, to be robust to spurious differences between semantically equivalent inputs, and more. We seek to understand the degree to which existing natural language models achieve or fall short of compositional generalization across a variety of tasks, whether there are distinct types of compositional generalization in practice, and to identify avenues of intervention through which we can improve models’ compositional generalization ability. After exploring these questions using an array of models, tasks, and potential interventions we see that large, pretrained language models have encountered sufficient training data to account for a variety of fine-grained compositional behavior but still struggle to reason at the level of phrases or larger language structures. Non-intrusive, data-based interventions in the form of augmenting individual sequences with compressed versions of themselves or deriving new examples from induced grammars prove insufficient to encourage greater levels of compositional reasoning, indicating that future work might benefit most from focusing on changes to a model’s inductive bias at the architectural or loss level, or by integrating compositionality-boosting data interventions into the large-scale pretraining process itself.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Unified descriptor for enhanced critical heat flux during pool boiling of hemi-wicking surfaces

Boiling heat transfer is dictated by interfacial phenomena at the three-phase contact line where vapor bubbles form on the surface. Structured surfaces have shown significant enhancement in critical heat flux (CHF) during pool boiling by tailoring interfacial phenomena. This CHF enhancement has been primarily explained by two structural effects: roughness, which extends the contact line length at the bubble base, and wickability, the ability to imbibe liquid through surface structures by capillary pumping. Here, in this work, we show that CHF enhancement on structured surfaces cannot be described by roughness or wickability alone. This result was confirmed using systematically designed micropillar surfaces with controlled roughness and wickability. Further, we performed a scaling analysis and derived a unified descriptor, which represents the combined effects of thin film density and volumetric wicking rate. This unified descriptor shows a reasonable correlation with CHF values with our experiments and literature data. This work provides important insights in understanding the role of surface structures on CHF enhancement, thereby providing guidelines for the systematic design of surface structures for enhanced pool boiling heat transfer.

42 ENGINEERING↗
DOE OSTI2021

Deep glassy state dynamic data challenge glass models: Configurational entropy models

This work was the result of the student's (Dongjie Chen) qualifying exam response. In it the experimental observation from our labs that the relaxation times of ultrastable glasses (a 20 million year old amber and a vapor deposited amorphous teflon) do not follow the expected Vogel-Fulcher or WLF types of divergence as the temperatures fall below the glass transition temperature. In the work we evaluated several theories by considering them in different ways. We used them as generally used and show that they do not capture the non-diverging behavior. Because the models were all entropy-based models, we also considered the possibility that the temperature dependence of the entropy used in the models is incorrect and chose two ways of estimating the entropy that do not follow the classical theoretical models. First we used the actually measured calorimetric response for the two ultrastable glasses, and we used the Milchev model, which does not go to zero above 0 Kelvin. Both of these approaches led to improved agreement between the theories and the experimental data. Finally, we also asked what the temperature dependence of the entropy would have to be so that the models agreed with the observed data. In some cases the entropy so-calculated was reasonable, while in others it was unphysical. In sum, the important point of the work is that the observation of non-diverging time-scales that suggest that there is no ideal glass transition can be reasonably captured by current theories if they take more realistic models of the entropy than is usually the case.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Rapid Adaptation of Chemical Named Entity Recognition Using Few-Shot Learning and LLM Distillation

Named entity recognition (NER) has been widely used in chemical text mining for the automatic identification and extraction of chemical entities. However, existing chemical NER systems primarily focus on scenarios with abundant training data, requiring significant human effort on annotations. This poses challenges for applications in the chemical field, such as catalysis, where many advancements have traditionally relied on trial-and-error investigations and incremental adjustment of variables. This hinders catalysis science and technology progress in addressing emerging energy and environmental crises. In this work, we propose a few-shot NER model that can quickly adapt to extract new types of chemical entities by using only a limited number of annotated examples. Our model employs a metric-learning approach to transfer entity similarity knowledge from high-resource chemical domains (with abundant annotations) to enable effective entity recognition in low-resource specialized domains (limited annotation). We validate the effectiveness of our model on a few-shot chemical NER benchmark built based on six existing chemical NER data sets. Experiments show that the proposed few-shot NER model can achieve reasonable performance with only 5 examples per entity type and shows consistent improvement as the number of examples increases. Furthermore, we demonstrate how the proposed model can be trained with large language model (LLM) annotated data, opening a new pathway for rapid adaptation of NER systems. Furthermore, our approach leverages the knowledge broadness of large language models for chemistry while distilling this knowledge into a lightweight model suitable for efficient and in-house use.

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

Development of HELIOS/BIPR/PARCS/MCNP6 Computation Route for WWER RPV Neutron Fluence Analysis and Validation Against Ex-Vessel Detector Measurement Data

Here, a computational route was developed for precise calculation of fast neutron fluence on a WWER-type reactor pressure vessel (RPV). The method is based on the transfer of neutronics data from HELIOS-2 lattice calculations and nodal diffusion neutronics data (power, density, and temperature) from BIPR7.1 and PARCS 3.36/PATHS core calculations into a three-dimensional (pinwise axially distributed) fixed neutron source for modeling of transport of fast neutrons from the reactor core to the outer surface of the RPV using MCNP6.2. Validation of the proposed computational method was carried out based on comparative analysis of MCNP6.2-predicted and neutron dosimetry–measured reaction rates [ 54 Fe(n,p) 54 Мn, 93 Nb(n,n') 93 mNb, and 58 Ni(n,р) 58 Со] on the outer surface of the Armenian Nuclear Power Plant (ANPP) Unit 2 RPV. Validation revealed that the MCNP6.2-predicted fast neutron fluence results are very sensitive to the ENDF-B neutron data. Particularly, MCNP6.2 with ENDF/B-VIII.0 significantly underpredicts (20% to 30%) fast neutron fluence while using ENDF/B-VII.1 data overpredicts it. Adding revised beta-released evaluations of 54 Fe, 56 Fe, 57 Fe, and 16 O from the International Nuclear Data Evaluation Network (INDEN) to ENDF/B-VIII.0 allows one to obtain reasonable agreement with measurement results for all types of measured reaction rates.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

Fathoming the anisotropic magnetoelasticity and magnetocaloric effect in GdNi

Intermetallic GdNi adopts a CrB type of crystal structure (space group Cmcm), and it orders ferromagnetically via a second-order phase transition at 70 K, exhibiting unusually strong spontaneous striction along the three independent crystallographic axes in the ferromagnetically ordered state. Here we introduce a microscopic model to describe anisotropic changes of lattice parameters and elastic contribution to magnetocaloric effect of GdNi. In the model, results of density functional theory (DFT) calculations are used as inputs into a Hamiltonian that includes elastic energy of an anisotropic crystal lattice, exchange interactions, and Zeeman effect. The magnetic and elastic Hamiltonians are coupled through an anisotropic Bean-Rodbell model of magnetoelastic interactions. This coupling gives rise to anisotropic changes in the lattice parameters observed experimentally, and the model reveals good to reasonable agreements between the current theoretical results and earlier experimental data, thus validating the model within the limits of assumptions made. We also show that DFT calculations with 4f electrons of Gd treated as core electrons lead to a more adequate estimate of elastic constants of GdNi in comparison with the LDA + U method where 4f electrons are treated as valence electrons.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2020

Measurement of the 239 Pu$(n, f$) prompt fission neutron spectrum from 10 keV to 10 MeV induced by neutrons of energy 1–20 MeV

Although the prompt fission neutron spectrum (PFNS) is an essential component of neutron-driven systems that has been measured for decades, there are still multiple glaring unknowns regarding the PFNS of major actinides in the fission neutron incident energy range, specifically with regard to multichance fission and pre-equilibrium neutron emission processes. The only impactful experimental 239 Pu PFNS measurements included in recent nuclear data evaluations were measured over a limited outgoing neutron energy range at thermal and 1.5-MeV average incident neutron energy, while other potentially impactful measurements have been shown to contain errors that resulted in either large uncertainty increases or in complete exclusion from nuclear data evaluation. Here, we report a measurement of the 239 Pu PFNS over a wide range of incident neutron energy (1–20 MeV) and three orders of magnitude in outgoing neutron energy (0.01–10 MeV) resulting from the Chi-Nu experiment at the Los Alamos Neutron Science Center. These results are the combination of separate PFNS measurements in the same experimental area, one using a Li-glass and the other a liquid scintillator detector array. Covariances between all PFNS data points from each detector and within each incident energy range were generated between all other data in both detector arrays and within all other incident neutron energy bins, yielding a single covariance matrix for all 1300 PFNS data points reported here. These covariances are based on a thorough assessment of systematic bias and uncertainties associated with the measurement, PFNS extraction technique, combination of data from each detector type, and other aspects of the analysis. The existence of covariances between PFNS data points in different incident neutron energy ranges yielded covariances between average PFNS energy values at each incident energy to be reported here as well, which allowed for firm statements to be made regarding a shape of a purely experimental mean PFNS energy trend for the first time. Although minor PFNS shape differences exist between the results reported here and recent nuclear data evaluations, the ENDF/B-VIII.0 and JEFF-3.3 PFNS evaluations agree reasonably well with the present results from 1-to 10-MeV incident neutron energy, which spans the well-measured 1.5-MeV incident neutron energy PFNS from Lestone and Shores as well as the onset of second-chance fission. However, while the pre-equilibrium component of the PFNS above 12-MeV incident neutron energy roughly agrees in position and magnitude with ENDF/B-VIII.0 and JEFF-3.3, clear differences relating to the relative magnitude of third-chance fission PFNS features are present in the PFNS shape and in the mean PFNS energy trends.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

Persistent Current Simulation for CCT Testing Magnet Used in EIC

The Electron-Ion Collider (EIC), a powerful new facility to be built in the United States at the U.S. Department of Energy's Brookhaven National Laboratory in collaboration with Thomas Jefferson National Accelerator Facility, will explore the most fundamental building blocks of nearly all visible matter. Here, there are many different types of superconducting magnets near the interaction region (IR) of EIC. Due to space constraints and special lattice requirements, Tapered CCT (canted-cosine theta) magnets have been used for EIC. At beam injection, the magnetic field is only ~5.5% of the maximum operating field. Considerable field errors will be generated from persistent current in superconducting strands even using very fine filament for those superconductors. A tapered CCT demonstrator magnet has been built and tested successfully at BNL since July 2020 to evaluate the key technologies for future tapered CCT magnets. In October 2023, BNL team also measured the persistent current in this demonstrator magnet. To validate the persistent current simulation methods for CCT magnets in EIC, this paper used a full 3D Opera Model and measured magnetization data from superconducting strand for the simulation. Simulation results showed reasonable agreement with recent measurement results.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2023

Spark Channel Dynamics of Electrostatic Discharges (Rev. 1) [Slides]

To model spark discharges, this work implemented, verified, and validated: 1-D radial conservative Lagrangian hydrodynamic scheme (with heat conduction); Eddington/P1 radiative transfer approximation, time-implicit scheme; RLC circuit solver, time-implicit scheme (adaptable to other circuit types). Novel analytics EOS fit for air up to temperatures of ~150,000 K. Novel "seed" electron concept to initialize finite conductivity in spark without pre-heating the air in the channel, as typically used in spark modeling literature. Reasonable, if conservative (from an engineering viewpoint), agreement with experimental data for indirect (circuit) and direct (interferometry) measurements of spark. Many potential avenues of future work: Two-temperature dryodynamics (requires two-temperature EOS development); 2-D axisymmetric geometry for axial (z) variation, modeling delectric electrodes; Kinetic equations for time-dependent charged species populations.

42 ENGINEERING↗
DOE OSTI2020

Thermodynamics of Uranium Tri-Iodide from Density-Functional Theory

Density-functional theory (DFT) is employed to investigate the thermodynamic and ground-state properties of bulk uranium tri-iodide, UI3. The theory is fully relativistic and electron correlations, beyond the DFT and generalized gradient approximation, are addressed with orbital polarization. The electronic structure indicates anti-ferromagnetism, in agreement with neutron diffraction, with band gaps and a non-metallic system. Furthermore, the formation energy, atomic volume, crystal structure, and heat capacity are calculated in reasonable agreement with experiments, whereas for the elastic constants experimental data are unavailable for comparison. The thermodynamical properties are modeled within a quasi-harmonic approximation and the heat capacity and Gibbs free energy as functions of temperature agree with available calculation of phase diagram (CALPHAD) thermodynamic assessment of the experimental data.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2023

Physics-Informed Deep Learning for Reconstruction of Spatial Missing Climate Information in the Antarctic

Understanding the influence of the Antarctic on the global climate is crucial for the prediction of global warming. However, due to very few observation sites, it is difficult to reconstruct the rational spatial pattern by filling in the missing values from the limited site observations. To tackle this challenge, regional spatial gap-filling methods, such as Kriging and inverse distance weighted (IDW), are regularly used in geoscience. Nevertheless, the reconstructing credibility of these methods is undesirable when the spatial structure has massive missing pieces. Inspired by image inpainting, we propose a novel deep learning method that demonstrates a good effect by embedding the physics-aware initialization of deep learning methods for rapid learning and capturing the spatial dependence for the high-fidelity imputation of missing areas. We create the benchmark dataset that artificially masks the Antarctic region with ratios of 30%, 50% and 70%. The reconstructing monthly mean surface temperature using the deep learning image inpainting method RFR (Recurrent Feature Reasoning) exhibits an average of 63% and 71% improvement of accuracy over Kriging and IDW under different missing rates. With regard to wind speed, there are still 36% and 50% improvements. In particular, the achieved improvement is even better for the larger missing ratio, such as under the 70% missing rate, where the accuracy of RFR is 68% and 74% higher than Kriging and IDW for temperature and also 38% and 46% higher for wind speed. In addition, the PI-RFR (Physics-Informed Recurrent Feature Reasoning) method we proposed is initialized using the spatial pattern data simulated by the numerical climate model instead of the unified average. Compared with RFR, PI-RFR has an average accuracy improvement of 10% for temperature and 9% for wind speed. When applied to reconstruct the spatial pattern based on the Antarctic site observations, where the missing rate is over 90%, the proposed method exhibits more spatial characteristics than Kriging and IDW.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Quantitative Power System Resilience Metrics and Evaluation Approach: Preprint

Power system resilience is an emerging topic and plays an essential role in helping power industry understand and respond to the increasing threats of extreme weather events. The first step of power system resilience analysis is to introduce metrics to quantify the resilience reasonably. Existing resilience metrics are typically restrained by the limited data for extreme event modeling and fall short in terms of physical interpretation and comparability. This paper develops novel quantitative metrics to evaluate power system resilience in pre- and post-event contexts. The developed metrics illustrate clear physical meanings and can be effectively used to compare resilience across different systems under different extreme events. Moreover, the developed metrics can be applied to both transmission and distribution systems. Simulation on a distribution system is employed to validate the effectiveness of the proposed resilience metrics and resilience evaluation approach.

power system resilience↗
DOE OSTI2020

SRNL Analysis of ICCWR LCM and WAMS data for Corrosion and Cracking (FY2020 Progress Report)

The development of algorithms for machine learning and data analysis for the 3013 Surveillance Program is a collaborative effort by the Savannah River National Laboratory (SRNL) and the University of South Carolina (USC). For corrosion detection, Laser Confocal Microscope (LCM) or Wide Area 3D Measurement System (WAMS) data is extracted from large binary files, with software written to convert the data to physical attributes (e.g. height, color and grayscale values; all as functions of a location in a plane projection). It is the objective of this project to produce a user-friendly interface that incorporated all operations needed to perform surface examination. For this reason, a Matlab-based Graphical User Interface (GUI) was created to integrate data input with software developed for processing and evaluation. In summary, the GUI permits selective downloading of binary data, interrogation of attributes, data labeling, flagging of significant features, execution of Machine Learning (ML) algorithms, output of parameters for trained ML algorithms, reports of ML model accuracy with respect to labeled data, and generation of graphical representations of various analyses.

3013 corrosion↗
DOE OSTI2022

Improved Cross Section Generation Capability of Griffin in FY22

The Griffin code is a Multiphysics Object-Oriented Simulation Environment (MOOSE) based reactor multiphysics analysis application jointly developed by Idaho National Laboratory and Argonne National Laboratory. The code includes a variety of deterministic steady-state transport solvers for fixed source, k-eigenvalue, adjoint, and subcritical multiplication as well as transient solvers for spatial dynamics with the improved quasi-static method. Griffin uses cross section data in the ISOXML format generated from external deterministic or Monte Carlo cross section generation codes. In recent years, the MC 2 -3 modules have been added to Griffin for fast reactor cross section generation, and the self-shielding application programming interface (SSAPI) was implemented in the ISOXML module for thermal reactor cross section generation. The on-the-fly slowing down method and double-heterogeneity treatment have been implemented to SSAPI and verified against particulate fuel-bearing graphite-moderated thermal reactor problems with high accuracy. This year, work has been focused on improving the cross section generation capability of ISOXML and streamlining the cross section generation procedures. In addition, the form function data were added to ISOXML in order to support the pin power reconstruction capability that was newly implemented in Griffin in this fiscal year. To facilitate the cross section generation using MC 2 -3 and SSAPI in Griffin, the cross section generation workflows have been set up for both fast and thermal spectrum reactors. The MOOSE action system tool was devised for fast spectrum problems, and the MOOSE stochastic tool was adopted to the branch calculation procedure for thermal spectrum problems. Meanwhile, to ensure the accuracy of group-constants, the thermal up-scattering kernel calculator accounting for resonance scattering was implemented in ISOXML, demonstrating the accurate computation of a Doppler-broadened scattering kernel of any Legendre order within a reasonable timescale. Other aspects of ISOXML, such as deletion solver and data, documentation, ISOXML file management, and the interface for Mixture, were improved as well.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

High-temperature seals for supercritical carbon-dioxide (sCO 2 ) turbines (Final Report)

This is the final report for project DE-FE0031924 titled “High-temperature seals for supercritical carbon-dioxide (sCO 2 ) turbines.” The report provides a summary of the entire project efforts from October 2020 through December 2024 including the high-temperature commercial dry gas seal (DGS) tests and thermal modeling of Task 2, as well as the high-temperature, large-diameter seal design and high-temperature tests of large-diameter seals in Task 3. A key outcome of Task 2 was the testing completion of specially instrumented commercial DGS in the GE-SwRI Apollo sCO 2 compressor (27,000 rpm). Test data from the DGS showed elevated temperatures upwards of 350 o F, which are close to the higher operating temperature limit of the DGS. The temperature measurements provide insight into the expected thermal loads on DGS operating in high-speed sCO 2 compressor and provided test data for validation of an in-house thermal model of the compressor/seal. Under Task 2.0, this report also presents the development of a steady-state conjugate heat-transfer model of the DGS operating in the sCO 2 compressor – a first of its kind model for modeling heat transfer of sCO 2 in an actual operating compressor. The findings of the thermal model show a reasonable match between temperature predictions of the model and the measured temperature data, also pointing out the validity of the approach and assumptions made in modeling the flows, heat transfer coefficients and windage modeling in the rig. Under Task 3.0, this report presents the preliminary design of a large-diameter hybrid face seal (14 inch and 26-inch diameter) for field testing in a land-based GE turbine. The preliminary seal design effort presented in this report under project DE-FE0031924 builds on the development and successful laboratory testing for such large diameter hybrid face seal under the prior DE-FE0024007 project. Key aspects of seal fluid analyses with CFD, mechanical design considerations and assembly considerations in a land-based turbine are presented. Finally, under Task 3.0, this report also presents the continued high-temperature testing of the 14-inch diameter hybrid face seal developed previously under the DE-FE0024007 program. Specifically, test data demonstrating successful non-contact seal operation and seal effective leakage of 0.001-inch with seal inlet temperatures above 700 o F are presented in this report. Successful hybrid seal operation in a laboratory environment for a large diameter (14-inch) seal at temperatures above 700 o F is a major technological milestone for this technology.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2020

Active Learning A Neural Network Model For Gold Clusters & Bulk From Sparse First Principles Training Data

Small metal clusters are of fundamental scientific interest and of tremendous significance in catalysis. These nanoscale clusters display diverse geometries and structural motifs depending on the cluster size; a knowledge of this size-dependent structural motifs and their dynamical evolution has been of longstanding interest. Given the high computational cost of first-principles calculations, molecular modeling and atomistic simulations such as molecular dynamics (MD) has proven to be an important complementary tool to aid this understanding. Classical MD typically employ predefined functional forms which limits their ability to capture such complex size-dependent structural and dynamical transformation. Neural Network (NN) based potentials represent flexible alternatives and in principle, well-trained NN potentials can provide high level of flexibility, transferability and accuracy on-par with the reference model used for training. A major challenge, however, is that NN models are interpolative and requires large quantities (similar to 10 4 or greater) of training data to ensure that the model adequately samples the energy landscape both near and far-from-equilibrium. A highly desirable goal is minimize the number of training data, especially if the underlying reference model is first-principles based and hence expensive. In this work, we introduce an active learning (AL) scheme that trains a NN model on-the-fly with minimal amount of first-principles based training data. Our AL workflow is initiated with a sparse training dataset (similar to 1 to 5 data points) and is updated on-the-fly via a Nested Ensemble Monte Carlo scheme that iteratively queries the energy landscape in regions of failure and updates the training pool to improve the network performance. Using a representative system of gold clusters, we demonstrate that our AL workflow can train a NN with similar to 500 total reference calculations. Using an extensive DFT test set of similar to 1100 configurations, we show that our AL-NN is able to accurately predict both the DFT energies and the forces for clusters of a myriad of different sizes. Our NN predictions are within 30 meV/atom and 40 meV/angstrom of the reference DFT calculations. Moreover, our AL-NN model also adequately captures the various size-dependent structural and dynamical properties of gold clusters in excellent agreement with DFT calculations and available experiments. We finally show that our AL-NN model also captures bulk properties reasonably well, even though they were not included in the training data.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗