Search NASA⌕ Search

SEARCH · Search NASA

Results for “association prediction”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Uncertainty quantification of mass models using ensemble Bayesian model averaging

Developments in the description of the masses of atomic nuclei have led to various nuclear mass models that provide predictions for masses across the whole chart of nuclides. These mass models play an important role in understanding the synthesis of heavy elements in the rapid neutron capture ( r ) process. However, it is still a challenging task to estimate the size of uncertainty associated with the predictions of each mass model. In this work, a method called ensemble Bayesian model averaging (EBMA) is introduced to quantify the uncertainty of one-neutron separation energies (S 1 n ) which are directly relevant in the calculations of r -process observables. Here, this Bayesian method provides a natural way to perform model averaging, selection, and uncertainty quantification, by combining the mass models as a mixture of normal distributions whose parameters are optimized against the experimental data, employing the Markov chain Monte Carlo method using the no-u-turn sampler. The EBMA model optimized with all the experimental S 1 n from the AME2003 nuclides are shown to provide reliable uncertainty estimates when tested with the new data in the AME2020.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2021

Evaluation of Potential Impacts of Free Chlorine during Washing of Fresh-Cut Leafy Greens on Escherichia coli O157:H7 Cross-Contamination and Risk of Illness

Addition of chlorine-based antimicrobial substances to fresh-cut leafy green wash water is done to minimize microbial cross-contamination during processing. We developed the FDA Leafy Green Risk Assessment Model (FDA-LGRAM) to quantify the impact of free chlorine concentration in wash water during fresh-cut lettuce processing on the extent of water-mediated cross-contamination between shredded lettuce and the associated risk of illness due to exposure to Escherichia coli O157:H7. At different contamination prevalence and levels of E. coli O157:H7 on incoming lettuce heads, the model compared the predicted prevalence of contaminated fresh-cut lettuce packages and the risk of illness per serving between: (1) a scenario where fresh-cut lettuce was packaged without washing; and (2) scenarios involving washing fresh-cut lettuce with different levels of free chlorine (0 ppm, 5 ppm, 10 ppm, 15 ppm, and 20 ppm) prior to packaging. Our results indicate that the free chlorine level in wash water has a substantial impact on the predicted prevalence of contaminated fresh-cut lettuce packages and the risk of illness associated with E. coli O157:H7 in fresh-cut lettuce. Results showed that the required level of free chlorine that can minimize water-mediated cross-contamination and reduce the corresponding risk of illness depended on contamination prevalence and levels of E. coli O157:H7 on incoming lettuce heads. Our model also indicated that the pathogen inactivation rate in wash water via free chlorine was a key model parameter that had a significant impact on the extent of cross-contamination during washing and the predicted associated risk of illness.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Columbia River Surface Water and Pore Water Unaligned FTICR-MS Data Associated with “Using Community Assembly Metrics from Metacommunity Ecology to Understand Environmental Metabolomes”

Surface water and pore water samples were collected at five sites along a shoreline transect of the Columbia River, southeastern WA, USA in 2017. This dataset is comprised of (1) high resolution characterization of dissolved organic matter via 12 Tesla Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) analyzed through the Environmental Molecular Sciences Laboratory (EMSL; https://www.pnnl.gov/environmental-molecular-sciences-laboratory) and (2) a metadata csv file. The FTICR-MS files are .xml, and the data package includes instructions for using Formularity (https://omics.pnl.gov/software/formularity) and an R script to process the data based on the user’s specific needs. The data are associated with the publication “Using community assembly metrics from metacommunity ecology to understand environmental metabolomes” that will be published in Nature Communications (Danczak et al.). The study aimed to investigate gaps in understanding the spatiotemporal organization of environmental metabolomes to improve the ability to predict ecosystem function. Associated sequencing data will be published via NCBI. The access information for NCBI and the publication DOI will be added to this abstract when they are available. Please use the data package’s DOI to cite the data package. We ask that you acknowledge the U.S. Department of Energy (DOE) Biological and Environmental Research (BER) Subsurface Biogeochemical Research (SBR) program when using the data. All data are free to be used for any purpose, such as for manuscripts, presentations, and grant proposals. There is no obligation to include data package authors as co-authors.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Peripheral Blood Transcript Signatures after Internal 131 I-mIBG Therapy in Relapsed and Refractory Neuroblastoma Patients Identifies Early and Late Biomarkers of Internal 131 I Exposures

131 I-metaiodobenzylguanidine ( 131 I-mIBG) is a targeted radiation therapy developed for the treatment of advanced neuroblastoma. We have previously shown that this patient cohort can be used to predict absorbed dose associated with early 131 I exposure, 72 h after treatment. We now expand these studies to identify gene expression differences associated with 131 I-mIBG exposure 15 days after treatment. Total RNA from peripheral blood lymphocytes was isolated from 288 whole blood samples representing 59 relapsed or refractory neuroblastoma patients before and after 131 I-mIBG treatment. We found that several transcripts predictive of early exposure returned to baseline levels by day 15, however, selected transcripts did not return to baseline. At 72 h, all 17 selected pathway-specific transcripts were differentially expressed. Transcripts CDKN1A (P < 0.000001), FDXR (P < 0.000001), DDB2 (P < 0.000001), and BBC3 (P < 0.000001) showed the highest up-regulation at 72 h after 131 I-mIBG exposure, with mean log2 fold changes of 2.55, 2.93, 1.86 and 1.85, respectively. At day 15 after 131I-mIBG, 11 of the 17 selected transcripts were differentially expressed, with XPC, STAT5B, PRKDC, MDM2, POLH, IGF1R, and SGK1 displaying significant up-regulation at 72 h and significant down-regulation at day 15. Interestingly, transcripts FDXR (P = 0.01), DDB2 (P = 0.03), BCL2 (P = 0.003), and SESN1 (P < 0.0003) maintained differential expression 15 days after 131 I-mIBG treatment. These results suggest that transcript levels for DNA repair, apoptosis, and ionizing radiation-induced cellular stress are still changing by 15 days after 131 I-mIBG treatment. Our studies showcase the use of biodosimetry gene expression panels as predictive biomarkers following early (72 h) and late (15 days) internal 131 I exposure. Our findings also demonstrate the utility of our transcript panel to differentiate exposed from non-exposed individuals up to 15 days after exposure from internal 131 I.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗
DOE OSTI2023

Predicting discomfort from glare with pedestrian-scale lighting: A comparison of candidate models using four independent datasets

After dark, pedestrians may experience discomfort from glare caused by outdoor lighting. While several models for measuring discomfort have been proposed, there is no consensus as to which model should be used. The performances of different models were investigated using datasets from four independent studies, comparing the degree of association between model predictions and subjective ratings, and the ability of a model to distinguish between discomfort and non-discomfort situations. The models tested are those proposed by Petherbridge and Hopkinson in 1950, Schmidt-Clausen and Bindels in 1974, Bullough et al. in 2008 and Lin et al. in 2014 and 2015. They also include two quantities: direct illuminance at the eye from the glare source and average source luminance. Of the models tested, the best performance was found using either the model proposed by Bullough et al. in 2008 or by direct illuminance at the eye.

Abboushi, B.↗
DOE OSTI2023

Deep learning integrates histopathology and proteogenomics at a pan-cancer level

We introduce a pioneering approach that integrates pathology imaging with transcriptomics and proteomics to identify predictive histology features associated with critical clinical outcomes in cancer. We utilize 2,755 H&E-stained histopathological slides from 657 patients across 6 cancer types from CPTAC. Our models effectively recapitulate distinctions readily made by human pathologists: tumor vs. normal (AUROC = 0.995) and tissue-of-origin (AUROC = 0.979). We further investigate predictive power on tasks not normally performed from H&E alone, including TP53 prediction and pathologic stage. Importantly, we describe predictive morphologies not previously utilized in a clinical setting. The incorporation of transcriptomics and proteomics identifies pathway-level signatures and cellular processes driving predictive histology features. Model generalizability and interpretability is confirmed using TCGA. We propose a classification system for these tasks, and suggest potential clinical applications for this integrated human and machine learning approach. A publicly available web-based platform implements these models.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Uncertainty bounds for multivariate machine learning predictions on high-strain brittle fracture

Simulation of the crack network evolution on high strain rate impact experiments performed in brittle materials is very compute-intensive. The cost increases even more if multiple simulations are needed to account for the randomness in crack length, location, and orientation, which is inherently found in real-world materials. Constructing a machine learning emulator can make the process faster by orders of magnitude. There has been little work, however, on assessing the error associated with their predictions. Estimating these errors is imperative for meaningful overall uncertainty quantification. In this work, we extend the heteroscedastic uncertainty estimates to bound a multiple output machine learning emulator. Overall, we find that the response prediction is accurate within its predicted errors, but with a somewhat conservative estimate of uncertainty.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Quantum physics of stars

Stars are slowly developing objects; the lifetimes of the different burning phases are determined by the strength of nuclear reactions, which in turn are defined by the quantum structure of the associated nuclei at the threshold and the respective reaction mechanisms. Stars, from the nuclear physics perspective, are cold environments where only a few of the key nuclear reactions have been measured at the actual stellar plasma temperatures. This is also the case for more dynamic astrophysical phenomena from the big bang to stellar explosions. Most of the nuclear reaction rates are therefore based on theoretical extrapolations. A number of discrepancies between these predictions and the associated stellar signatures have been observed, and many may be due to low-energy or near-threshold quantum effects. These effects need to be understood in order to reliably model nuclear reaction processes, not only for stars but also for low-temperature plasma environments such as controlled magnetic or inertial confinement fusion systems, which operate in similar temperature regimes. This review summarizes the various theoretical techniques presently used for deriving reaction rates and discusses possible quantum effects that may impact the reaction cross section near the reaction threshold. These resemble enhanced single-particle and cluster structures near threshold and associated interference effects. New experimental techniques such as deep-underground accelerators or the study of transfer reactions to mimic the quantum-mechanical transition strength, the so-called Trojan horse method, provide ways to directly or indirectly probe the reaction features that determine the reaction rates at stellar energies. Furthermore, this is demonstrated on a number of key nuclear reactions for different nucleosynthesis environments. Finally, current inconsistencies between experimental predictions and observations are discussed.

Models & methods for nuclear reactions↗
DOE OSTI2022

Methodology for assessing the maximum potential impact of separations opportunities in industrial processes

Separation technologies currently used in U.S. manufacturing industries are estimated to account for more than 20% of plant energy consumption. However, accurately determining the impact of new separation technology solutions can sometimes be difficult, especially when evaluating a slate of new candidate separation technologies, each of which has its own separation performance, energy demand, and capital cost. In these cases, a typical approach is to assess each new separation technology by collecting performance and cost information and then using that information to develop a techno-economic analysis to identify overall benefits. While this approach is thorough, it can be time consuming and can hinder reaching a critical understanding of the potential of a given separation challenge, especially when there is no known solution. To address these issues, we developed an assessment methodology, using industrial screening processes, that can be used to better understand the potential impacts of addressing a given separation challenge. This paper presents an overview of our separation challenge stream assessment methodology. The methodology involves defining an “ideal” separator and deriving the associated minimum separation energy. The “ideal” separator represents the most optimistic outlook of a given opportunity so the maximum impact from existing and not-yet-developed solutions can be assessed. Using established biorefinery models, we applied the methodology to 10 different separation challenge streams from two different biomass conversion platforms to identify the type of information that can be obtained. Three of the ten challenge streams assessed had maximum possible cost savings predictions >20%, and associated reductions in process energy carbon intensity ranging from 0 to 54%. Two streams had cost and energy savings potential that were < 5%. Some of the opportunity drivers from the various assessments include higher product yields, reduction or elimination of downstream equipment, new co-products, and cost savings associated with raw materials and energy consumption. The information from these assessments can help guide the selection or development of new separation technology solutions based on the various potential factors that drive the projected benefits.

09 BIOMASS FUELS↗
DOE OSTI2020

Deep neural network improves the estimation of polygenic risk scores for breast cancer

Polygenic risk scores (PRS) estimate the genetic risk of an individual for a complex disease based on many genetic variants across the whole genome. Here, we compared a series of computational models for estimation of breast cancer PRS. A deep neural network (DNN) was found to outperform alternative machine learning techniques and established statistical algorithms, including BLUP, BayesA, and LDpred. In the test cohort with 50% prevalence, the Area Under the receiver operating characteristic Curve (AUC) were 67.4% for DNN, 64.2% for BLUP, 64.5% for BayesA, and 62.4% for LDpred. BLUP, BayesA, and LPpred all generated PRS that followed a normal distribution in the case population. However, the PRS generated by DNN in the case population followed a bimodal distribution composed of two normal distributions with distinctly different means. This suggests that DNN was able to separate the case population into a high-genetic-risk case subpopulation with an average PRS significantly higher than the control population and a normal-genetic-risk case subpopulation with an average PRS similar to the control population. This allowed DNN to achieve 18.8% recall at 90% precision in the test cohort with 50% prevalence, which can be extrapolated to 65.4% recall at 20% precision in a general population with 12% prevalence. Interpretation of the DNN model identified salient variants that were assigned insignificant p values by association studies, but were important for DNN prediction. These variants may be associated with the phenotype through nonlinear relationships.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Preliminary IHLW Formulation Algorithm Description

This report documents the initial algorithm that could be used by the Waste Treatment and Immobilization Plant (WTP) in batching high-level waste (HLW) and glass-forming chemicals (GFCs) in the HLW melter feed preparation vessel (MFPV) (HFP-VSL-00001 and -00005). Not all Hanford tank waste can be accommodated by the models developed for this report and significant expansion of the model boundaries could be achievable to reduce the WTP mission life and total canister production count. The immobilized HLW (IHLW) must meet a series of constraints to be acceptable for disposal in the Monitored Geologic Repository, which are contained in the Specification 1 of the Contract (DOE 2000), the Waste Acceptance Product Specifications (WAPS, DOE 1996), and the Waste Acceptance System Requirements Document (WASRD, DOE 2007). The IHLW Waste Form Compliance Plan (WCP, 24590-HLW-PL-RT-07-0001, Rev 3) specifies that the formulation algorithm will be developed and used to comply with the constraints associated with glass composition and properties. This report is not an engineering calculation, does not provide design input, and is not an engineering study. Algorithm inputs include the chemical analyses of the blended HLW in the HLW blend vessel (HBV) (HLP-VSL-00028, the volume and composition of the MFPV heel, the volume and composition of the MFPV after waste addition, the volume and composition of MFPV batch after GFC addition, the compositions of individual GFCs, and the mass of glass in each canister. In addition to these inputs, uncertainties in the HLW composition and processing parameters are included in the algorithm. Using the above inputs, the algorithm calculates the following outputs: 1) the volume of HLW to be transferred from the HBV to the MFPV, 2) the mass of each GFC for addition to the MFPV, 3) the composition of the glass that will be produced along with uncertainties, and 4) the predicted properties, with associated uncertainties, of the resulting IHLW. The algorithm uses the property-composition models to calculate properties with associated uncertainties and compares them with various constraints to ensure that a processable feed is formulated and a compliant IHLW is produced. The GFC additions are determined using an optimization approach to provide high confidence that the HLW glass will meet all product quality requirements and key processing constraints. For most HLW batches there are many possible glass compositions that meet all constraints. In these cases, the glass composition is optimized for a series of target component concentrations and target property values. The algorithm also incorporates process measurement and product quality uncertainties, based on the work of Piepel et al. (2005). Estimates of the various process and measurement uncertainties that affect glass compositions and predicted glass properties have been previously reported (Piepel et al. 2005, 2006) and the impacts of these estimated uncertainties on the IHLW composition envelope that meets product quality and processing-related properties with sufficient confidence were evaluated. The details of work performed to date to develop this initial GFC addition and batching algorithm are summarized in Sections 4 and 5. An example data set is used to illustrate the calculations of the algorithm summarized in Section 6. Finally, in Section 7, there is a statement of the required work to achieve a final operational IHLW formulation control algorithm. This report is not an engineering calculation, does not provide design input, and is not an engineering study.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Comparison of venous and pooled capillary hemoglobin levels for the detection of anemia among adolescent girls

Introduction: Blood source is a known preanalytical factor affecting hemoglobin (Hb) concentrations, and there is evidence that capillary and venous blood may yield disparate Hb levels and anemia prevalence. However, data from adolescents are scarce. Objective: To compare Hb and anemia prevalence measured by venous and individual pooled capillary blood among a sample of girls aged 10–19 years from 232 schools in four regions of Ghana in 2022. Methods: Among girls who had venous blood draws, a random subsample was selected for capillary blood. Hb was measured using HemoCue® Hb-301. We used Lin’s concordance correlation coefficient (CCC) to quantify the strength of the bivariate relationship between venous and capillary Hb and a paired t-test for difference in means. We used McNemar’s test for discordance in anemia cases by blood source and weighted Kappa to quantify agreement by anemia severity. A multivariate generalized estimating equation was used to quantify adjusted population anemia prevalence and assess the association between blood source and predicted anemia risk. Results: We found strong concordance between Hb measures (CCC = 0.86). The difference between mean venous Hb (12.8 g/dL, ± 1.1) and capillary Hb (12.9 g/dL, ± 1.2) was not significant (p = 0.26). Crude anemia prevalence by venous and capillary blood was 20.6% and 19.5%, respectively. Adjusted population anemia prevalence was 23.5% for venous blood and 22.5% for capillary ( = 0.45). Blood source was not associated with predicted anemia risk (risk ratio: 0.99, 95% CI: 0.96, 1.02). Discordance in anemia cases by blood source was not significant (McNemar p = 0.46). Weighted Kappa demonstrated moderate agreement by severity (k =0.67). Among those with anemia by either blood source (n = 111), 59% were identified by both sources. Conclusion: In Ghanaian adolescent girls, there was no difference in mean Hb, anemia prevalence, or predicted anemia risk by blood source. However, only 59% of girls with anemia by either blood source were identified as having anemia by both sources. These findings suggest that pooled capillary blood may be useful for estimating Hb and anemia at the population level, but that caution is needed when interpreting individual-level data.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Survival analysis of localized prostate cancer with deep learning

In recent years, data-driven, deep-learning-based models have shown great promise in medical risk prediction. By utilizing the large-scale Electronic Health Record data found in the U.S. Department of Veterans Affairs, the largest integrated healthcare system in the United States, we have developed an automated, personalized risk prediction model to support the clinical decision-making process for localized prostate cancer patients. This method combines the representative power of deep learning and the analytical interpretability of parametric regression models and can implement both time-dependent and static input data. To collect a comprehensive evaluation of model performances, we calculate time-dependent C-statistics C td over 2-, 5-, and 10-year time horizons using either a composite outcome or prostate cancer mortality as the target event. The composite outcome combines the Prostate-Specific Antigen (PSA) test, metastasis, and prostate cancer mortality. Our longitudinal model Recurrent Deep Survival Machine (RDSM) achieved C td 0.85 (0.83), 0.80 (0.83), and 0.76 (0.81), while the cross-sectional model Deep Survival Machine (DSM) attained C td 0.85 (0.82), 0.80 (0.82), and 0.76 (0.79) for the 2-, 5-, and 10-year composite (mortality) outcomes, respectively. In addition to estimating the survival probability, our method can quantify the uncertainty associated with the prediction. The uncertainty scores show a consistent correlation with the prediction accuracy. We find PSA and prostate cancer stage information are the most important indicators in risk prediction. Our work demonstrates the utility of the data-driven machine learning model in prostate cancer risk prediction, which can play a critical role in the clinical decision system.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2020

Nuclear modification factors for jet fragmentation

Using a recently-developed perturbative-QCD approach for jet evolution in a dense quark-gluon plasma, we study the nuclear modification factor for the jet fragmentation function. The qualitative behaviour that we find is in agreement with the respective experimental observations in Pb+Pb collisions at the LHC: a pronounced nuclear enhancement at both ends of the spectrum. Our Monte Carlo simulations are supplemented with analytic estimates which clarify the physical interpretation of the results. The main source of theoretical uncertainty is the sensitivity of our calculations to a low-momentum cutoff which mimics confinement. To reduce this sensitivity, we propose a new observable, which describes the jet fragmentation into subjets and is infrared-and-collinear safe by construction. We present Monte Carlo predictions for the associated nuclear modification factor together with their physical interpretation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2020

Performance degradation due to anodic failure mechanisms in lithium-ion batteries

Here, we report a mechano-chemical model for anodic degradation during fast-charging of nickel-manganese-cobalt (NMC)/graphite (C) cell due to SEI growth, lithium plating/stripping, dead lithium storage, and film fracture of composite SEI and plated lithium film. Degradation of the battery is analyzed for a range of charging rates from 1 to 6 C-rates, and the influence of plating mechanisms – lithium plating and dead lithium deposition and recovery during stripping – on the film resistance of the anode are accounted for in the model. Dynamic evolution of the interfacial properties is modeled using rule-of-mixture approach. Model predictions of plating associated stress fields are used to compute critical energy release rate for film cracking. The results indicate an increased tendency of fracture for thinner SEI film with lithium plating at higher charging rates. The process of reforming the cracked film absorbs a significant portion of the electrode current thereby reducing the cell capacity and plating efficiency. The mechano-chemical model provides an extensive analytical framework for understanding the synergistic coupling of anodic degradation mechanisms, prognosticating conditions of SEI failure, and evaluating the capacity fade and efficiency of lithium-ion battery.

25 ENERGY STORAGE↗
DOE OSTI2026

Ab Initio Modeling of Aqueous Methanol Mixtures at DFT-SCAN Level Using Machine Learning Interatomic Potentials

Abstract Methanol–water mixtures find use in many applications, particularly catalytic energy conversion processes. Their importance has motivated numerous computational studies, most of which employed molecular dynamics based on classical force fields. These enable simulations of large systems on long time scales but do not reliably describe reactive dynamics involving bond breaking and bond formation. In contrast, ab initio molecular dynamics (AIMD) based on density functional theory (DFT) is generally more reliable for such applications but has a high computational cost, which discourages systematic studies of alcohol-water mixtures. To remedy this, we trained a machine learning interatomic potential capable of probing the properties of aqueous methanol mixtures at the DFT level using the SCAN functional. Our results show that SCAN qualitatively reproduces multiple key experimental features arising from the amphiphilic nature of methanol, including density, diffusion coefficients, X-ray structure factors, and Kirkwood–Buff integrals. We also find that structural correlations between water molecules are somewhat overestimated, leading to a stronger preferential association than that predicted by experiments. However, increasing the temperature by 30 K mitigates this effect and also recovers the correct mobilities of both methanol and water. These results indicate that SCAN provides an accurate description of methanol–water mixtures, making it a reliable choice for investigating the reactive dynamics in such systems.

Park, Sanghyun J. [Princeton University , , , ,]↗
DOE OSTI2024

Plasmon-Driven Ammonia Decomposition on Pd(111): Hole Transfer’s Role in Changing Rate-Limiting Steps

Here, ammonia (NH 3 ) has the potential to be a hydrogen carrier because it can be transported and stored with ease, but only if it also can be decomposed easily when needed. Understanding how to control the frequently rate-limiting N–H bond breaking and N–N bond forming on catalytic surfaces may help design efficient means for NH 3 decomposition. Yuan et al. recently demonstrated photocatalytically selective N–H bond breaking in NH 3 on plasmon-driven aluminum–palladium (Al–Pd) antenna–reactor heterostructures. Using embedded correlated wavefunction (ECW) theory, we predict that the rate-determining step (RDS) for NH 3 decomposition on Pd(111) via thermocatalysis (dissociating the first N–H bond, *NH 3 → *NH 2 + *H, in the ground state, where * means adsorbed) differs from that via photocatalysis (dissociating the second N–H bond, *NH 2 → *NH + *H, in the excited state). This result is consistent with the measured catalytic efficiency and selectivity of NH 3 -deuterium (D 2 ) exchange reactions (an indirect way to measure N–H bond breaking) on Al–Pd heterodimers. We also determine the origin of the observed selectivity of thermocatalysis and photocatalysis on Pd(111) toward doubly deuterated (NHD 2 ) and monodeuterated (NH 2 D) products, respectively, and explore viability of the full NH 3 decomposition path, also via ECW theory. Additionally, we predict that the associative desorption of *N as N 2 from Pd(111) is extremely difficult in thermocatalysis at least at low surface coverages; metal-to-adsorbate hole transfer in photocatalysis stabilizes the transition state for the first N–H bond dissociation, shifting the RDS to the second N–H bond breaking. Furthermore, the redistribution of electrons around *N upon excitation reduces the electron density in the Pd–N bonds, which may lower the barrier for N 2 associative desorption in photocatalysis. Thus, light-induced, plasmon-mediated, excited-state hole transfer may provide an efficient mechanism to accelerate NH 3 decomposition.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2026

Machine learning approach for vibronically renormalized electronic band structures

Here, we present a machine learning (ML) method for efficient computation of vibrational thermal expectation values of physical properties from first principles. Our approach is based on the nonperturbative frozen phonon formulation in which stochastic Monte Carlo algorithm is employed to sample configurations of nuclei in a supercell at finite temperatures based on a first-principles phonon model. A deep-learning neural network is trained to accurately predict physical properties associated with sampled phonon configurations, thus bypassing the time-consuming ab initio calculations. To incorporate the point-group symmetry of the electronic system into the ML model, group-theoretical methods are used to develop a symmetry-invariant descriptor for phonon configurations in the supercell. We apply our ML approach to compute the temperature dependent electronic energy gap of silicon based on density functional theory (DFT). We show that, with less than a hundred DFT calculations for training the neural network model, an order of magnitude larger number of sampling can be achieved for the computation of the vibrational thermal expectation values. Our work highlights the promising potential of ML techniques for finite temperature first-principles electronic structure methods.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗