Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep direct use”

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 343 records · Page 19
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Deep Learning for In-Situ Layer Quality Monitoring during Laser-Based Directed Energy Deposition (LB-DED) Additive Manufacturing Process

Defects are a leading issue for the rejection of parts manufactured through the Directed Energy Deposition (DED) Additive Manufacturing (AM) process. In an attempt to illuminate and advance in situ quality monitoring and control of workpieces, we present an innovative data-driven method that synchronously collects sensing data and AM process parameters with a low sampling rate during the DED process. The proposed data-driven technique determines the important influences that individual printing parameters and sensing features have on prediction at the inter-layer qualification to perform feature selection. Three Machine Learning (ML) algorithms including Random Forest (RF), Support Vector Machine (SVM), and Convolutional Neural Network (CNN) are used. During post-production, a threshold is applied to detect low-density occurrences such as porosity sizes and quantities from CT scans that render individual layers acceptable or unacceptable. This information is fed to the ML models for training. Training/testing are completed offline on samples deemed “high-quality” and “low-quality”, utilizing only features recorded from the build process. CNN results show that the classification of acceptable/unacceptable layers can reach between 90% accuracy while training/testing on a “high-quality” sample and dip to 65% accuracy when trained/tested on “low-quality”/“high-quality” (respectively), indicating over-fitting but showing CNN as a promising inter-layer classifier.

36 MATERIALS SCIENCE↗
DOE OSTI2026

NH 3 -Mediated Reactive Capture and Conversion: Integrating CO 2 Absorption from Flue Gas with CO Production via NH 4 HCO 3 Electrolysis

Efficient carbon capture and utilization require strategies that minimize energy penalties of CO 2 regeneration and compression. Reactive capture and conversion (RCC) address this challenge by integrating capture with direct electrochemical conversion. Here, we show an NH3-mediated tandem RCC system that couples capture of CO 2 from simulated flue gas (10% v/v CO 2 in N 2 ) with electroreduction of NH 4 HCO 3 to CO over a Ni single-atom catalyst (Ni-SAC). Speciation modeling and capture experiments revealed that a deep CO 2 capture with C/N ratio of 0.65 was achieved using 2.5 M NH 3 from simulated flue gas. Electrolysis of the resulting NH 4 HCO 3 on the Ni- SAC delivered an 85% CO Faradaic efficiency at 100 mA/cm 2 with excellent tolerance to NH 3 /NH 4 + as confirmed by DFT calculations and ab initio molecular dynamics (AIMD) simulations. Further, the technoeconomic analysis established a levelized total cost of CO manufacturing of $25.43/kmol, gauging the practical viability. Overall, this study holds great potential to decarbonize the chemical manufacturing industry while reducing synthetic production costs.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

Physics-Informed Deep Neural Networks for Learning Parameters and Constitutive Relationships in Subsurface Flow Problems

In this paper, we present a physics informed deep neural network (DNN) method for estimating parameters and unknown physics (constitutive relationships) in partial differential equation (PDE) models. We use PDEs in addition to measurements to train DNNs to approximate unknown parameters and constitutive relationships as well as states. The proposed approach increases the accuracy of DNN approximations of partially known functions when a limited number of measurements is available and allows for training DNNs when no direct measurements of the functions of interest are available. We employ physics informed DNNs to estimate the unknown space-dependent diffusion coefficient in a linear diffusion equation and an unknown constitutive relationship in a non-linear diffusion equation. For the parameter estimation problem, we assume that partial measurements of the coefficient and states are available and demonstrate that under these conditions, the proposed method is more accurate than state-of-the-art methods. For the non-linear diffusion PDE model with a fully unknown constitutive relationship (i.e., no measurements of constitutive relationship are available), the physics informed DNN method can accurately estimate the non-linear constitutive relationship based on state measurements only. Finally, we demonstrate that the proposed method remains accurate in the presence of measurement noise.

42 ENGINEERING↗
DOE OSTI2023

Detecting deviations from Gaussianity in high-redshift CMB lensing maps

ABSTRACT While the probability density function of the cosmic microwave background (CMB) convergence field approximately follows a Gaussian distribution, primordial non-Gaussianities, and small contributions from structures at low redshifts make the overall distribution slightly non-Gaussian. Some of the late-time component can be modelled using the distribution of galaxies and subtracted off from the original CMB lensing map to produce a map of matter distribution at high redshifts. Using this high-redshift mass map, we are able to directly study the early phases of structure formation. In this work, we forecast the detectability of signatures of non-Gaussianity due to non-linear structure formation at z > 1.2. Assuming the optimal case of no systematics, we find that it is challenging to detect such signatures in current surveys, but future experiments such as the deep field of CMB-S4 will be able to make detections of ∼7σ.

Astronomy & Astrophysics↗
DOE OSTI2024

Neural network emulation of flow in heavy-ion collisions at intermediate energies

Applications of new techniques in machine learning are speeding up progress in research in various fields. In this work, we construct and evaluate a deep neural network (DNN) to be used within a Bayesian statistical framework as a faster and more reliable alternative to the Gaussian process (GP) emulator of an isospin-dependent Boltzmann-Uehling-Uhlenbeck (IBUU) transport model simulator of heavy-ion reactions at intermediate beam energies. We found strong evidence of the DNN being able to emulate the IBUU simulator's prediction on the strengths of protons' directed and elliptical flow very efficiently even with small training datasets and with accuracy about ten times higher than the GP. Here, limitations of our present work and future improvements are also discussed.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2023

Exploring Classification of Topological Priors With Machine Learning for Feature Extraction

In many scientific endeavors, increasingly abstract representations of data allow for new interpretive methodologies and conceptualization of phenomena. For example, moving from raw imaged pixels to segmented and reconstructed objects allows researchers new insights and means to direct their studies toward relevant areas. Thus, the development of new and improved methods for segmentation remains an active area of research. With advances in machine learning and neural networks, scientists have been focused on employing deep neural networks such as U-Net to obtain pixel-level segmentations, namely, defining associations between pixels and corresponding/referent objects and gathering those objects afterward. Topological analysis, such as the use of the Morse-Smale complex to encode regions of uniform gradient flow behavior, offers an alternative approach: first, create geometric priors, and then apply machine learning to classify. This approach is empirically motivated since phenomena of interest often appear as subsets of topological priors in many applications. Using topological elements not only reduces the learning space but also introduces the ability to use learnable geometries and connectivity to aid the classification of the segmentation target. Here, in this article, we describe an approach to creating learnable topological elements, explore the application of ML techniques to classification tasks in a number of areas, and demonstrate this approach as a viable alternative to pixel-level classification, with similar accuracy, improved execution time, and requiring marginal training data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

First Demonstration of a Liquid Argon Time Projection Chamber Based Search for Intranuclear Neutron-Antineutron Transitions and Annihilation in $^{40}$Ar using the MicroBooNE Detector

We present a novel methodology to search for intranuclear neutron-antineutron transition ($n\rightarrow\bar{n}$) followed by $\bar{n}$-nucleon annihilation within an $^{40}$Ar nucleus, using the MicroBooNE liquid argon time projection chamber (LArTPC) detector. A discovery of $n\rightarrow\bar{n}$ transition or a new best limit on the lifetime of this process would either constitute physics beyond the Standard Model or greatly constrain theories of baryogenesis, respectively. The approach presented in this paper makes use of deep learning methods to select $n\rightarrow\bar{n}$ events based on their unique features and differentiate them from cosmogenic backgrounds. The achieved signal and background efficiencies are (70.22$\pm$6.04)\% and (0.0020$\pm$0.0003)\%, respectively. A demonstration of a search is performed with a data set corresponding to an exposure of $3.32 \times10^{26}\,$neutron-years, and where the background rate is constrained through direct measurement, assuming the presence of a negligible signal. With this approach, no excess of events over the background prediction is observed, setting a demonstrative lower bound on the $n\rightarrow\bar{n}$ lifetime in $^{40}$Ar of $\tau_{\textrm{m}} \gtrsim 1.

43 PARTICLE ACCELERATORS↗
DOE OSTI2020

NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations

In the last 50 years there has been a tremendous progress in solving numerically the Navier-Stokes equations using finite differences, finite elements, spectral, and even meshless methods. Yet, in many real cases, we still cannot incorporate seamlessly (multi-fidelity) data into existing algorithms, and for industrial-complexity applications the mesh generation is time consuming and still an art. Moreover, solving ill-posed problems (e.g., lacking boundary conditions) or inverse problems is often prohibitively expensive and requires different formulations and new computer codes. Here, we employ physics-informed neural networks (PINNs), encoding the governing equations directly into the deep neural network via automatic differentiation, to overcome some of the aforementioned limitations for simulating incompressible laminar and turbulent flows. We develop the Navier-Stokes flow nets (NSFnets) by considering two different mathematical formulations of the Navier-Stokes equations: the velocity-pressure (VP) formulation and the vorticity-velocity (VV) formulation. Since this is a new approach, we first select some standard benchmark problems to assess the accuracy, convergence rate, computational cost and flexibility of NSFnets; analytical solutions and direct numerical simulation (DNS) databases provide proper initial and boundary conditions for the NSFnet simulations. The spatial and temporal coordinates are the inputs of the NSFnets, while the instantaneous velocity and pressure fields are the outputs for the VP-NSFnet, and the instantaneous velocity and vorticity fields are the outputs for the VV-NSFnet. This is unsupervised learning and, hence, no labeled data are required beyond boundary and initial conditions and the fluid properties. The residuals of the VP or VV governing equations, together with the initial and boundary conditions, are embedded into the loss function of the NSFnets. No data is provided for the pressure to the VP-NSFnet, which is a hidden state and is obtained via the incompressibility constraint without extra computational cost. Unlike the traditional numerical methods, NSFnets inherit the properties of neural networks (NNs), hence the total error is composed of the approximation, the optimization, and the generalization errors. Here, we empirically attempt to quantify these errors by varying the sampling (“residual”) points, the iterative solvers, and the size of the NN architecture. For the laminar flow solutions, we show that both the VP and the VV formulations are comparable in accuracy but their best performance corresponds to different NN architectures. The initial convergence rate is fast but the error eventually saturates to a plateau due to the dominance of the optimization error. For the turbulent channel flow, we show that NSFnets can sustain turbulence at , but due to expensive training we only consider part of the channel domain and enforce velocity boundary conditions on the subdomain boundaries provided by the DNS data base. We also perform a systematic study on the weights used in the loss function for balancing the data and physics components, and investigate a new way of computing the weights dynamically to accelerate training and enhance accuracy. In the last part, we demonstrate how NSFnets should be used in practice, namely for ill-posed problems with incomplete or noisy boundary conditions as well as for inverse problems. We obtain reasonably accurate solutions for such cases as well without the need to change the NSFnets and at the same computational cost as in the forward well-posed problems. As a result, we also present a simple example of transfer learning that will aid in accelerating the training of NSFnets for different parameter settings.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

A deep decarbonization framework for the United States economy – a sector, sub-sector, and end-use based approach

Achieving the United States' target of net-zero greenhouse gas emissions by 2050 will require technological transformations and energy sector mitigation. To understand the role of dynamically evolving technologies, identify synergies and dissonance and the effect of allocating limited low-carbon biomass resources in decarbonizing the U.S. economy, we developed the Decarbonization Scenario Analysis Model. A Life Cycle Assessment based approach is implemented considering the U.S. economy as the functional unit, to estimate greenhouse gas mitigation potential for projected energy demand based on several sector-level and cross-sectoral decarbonization pathways. Direct and supply-chain emissions are accounted, resulting from changes in patterns of energy generation and consumption, technology breakthroughs, and reductions in fugitive emissions over time at the granularity of economic sectors, sub-sectors, and end-use. Decarbonization strategies are implemented over a reference case developed using Energy Information Administration (EIA AEO) projection of economic activities for 2020–2050. Based on the considered scenarios, 80–90% economy-wide decarbonization relative to the 2020 reference case is projected. Electrification, low-carbon fuels, and reduction of fugitive emissions play the most significant role to decarbonization. The majority of the remaining emissions are accounted to the supply-chain and end-use emissions from natural gas and diesel fossil-based fuels in heavy duty transportation and heavy industries, highlighting the need for developing low-carbon and carbon-negative alternatives to mitigate those fossil-based carbon emissions.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
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 OSTI2024

Thermal and Well Flow Performance of Closed-Loop Geothermal in the Wattenberg Area: Preprint

Closed-loop geothermal systems provide an alternative to resource-constrained hydrothermal systems and stimulation-intensive enhanced geothermal systems. In this work, we apply the slender-body theory (SBT) model, to simulate the well flow and heat transfer performance of U-loop well designs drilled in the Wattenberg area of the Denver-Julesburg Basin. Three U-loop well patterns are investigated including a single, double, and multi-lateral design. The subsurface within area is characterized by deep, hot (> 200degreesC) igneous/metamorphic basement rock underlying multiple sedimentary formations. The lateral section(s) of the U-loop lies within a target depth of 6 km where temperatures are estimated to approach 300degreesC. As a base case, conduction-only heat transfer is investigated through simulations with the SBT model within U-loops with open-hole laterals that exchange heat directly with the hot dry rock using water as a working fluid. The utilization of supercritical CO2 as a heat transfer fluid is also considered. For each scenario, the system performance in terms of annual heat production and temperature profile over a 20-year project lifetime are assessed. Also, the levelized costs of heat and electricity (LCOH and LCOE) are determined using a top-down technoeconomic analysis model. The results show that the performance and cost optimized U-loop design is one having an injection-production well spacing of 1,000 meters with ten 50-meter-spaced laterals that traverse a subsurface system with a temperature gradient of 60degreesC/km. By injecting 20 degreesC-water at a rate of 60 kg/s through this loop, an average heat production of 19 MWth can be achieved, resulting in an LCOE and LCOH of $136/MWh and $1.53/GJ, respectively, over a 20-year project life.

closed-loop geothermal↗
DOE OSTI2025

Modeling MTS pyrolysis and SiC deposition kinetics using principal component analysis and neural networks

Accurate chemical kinetics modeling is crucial for improving the efficiency of chemical processing and synthesis of ceramic matrix composites. Detailed kinetic models are computationally expensive due to the large number of transported chemical species, while the simplified physics-based models, such as single-step global mechanisms, are efficient but often overlook key chemical intermediates and pathways. Recent deep learning approaches promise accurate and cost-effective models. Yet, they require additional closures for the transported nonlinear latent variables, complicating integration with existing solvers. In this work, we develop a hybrid linear—nonlinear reduced model for silicon carbide deposition from methyltrichlorosilane precursor by combining principal component analysis (PCA) and autoencoder (AE) neural network (NN) approaches. PCA is used to identify a smaller set of linear transport variables, enabling direct reuse of conventional transport solvers. NNs then reconstruct the full chemical state from these reduced variables. We demonstrate the method on a chemical vapor deposition reactor—comprising a gas-phase pyrolysis plug flow reactor and a heterogeneous surface reactor—over a wide range of temperatures, pressures, and residence times. Our PCA–AE model achieves high accuracy with only five transported scalars, achieving an eightfold cost reduction compared to detailed mechanisms, in both a priori (using data from the test set only) and a posteriori (coupled with a differential equation solver). In conclusion, notable errors arise primarily near training domain boundaries and for long residence times, indicating the need for domain shift indicators and better long-horizon predictions in future reduced chemistry model development.

autoencoder neural networks↗
DOE OSTI2020

Oak Ridge National Laboratory Technical Input for the Nuclear Regulatory Commission Review of the 2017 Edition of ASME Section III, Division 5, ‘High Temperature Reactors’

To assist the Nuclear Regulatory Commission in its decision making on endorsement of the American Society for Mechanical Engineers Boiler and Pressure Vessel Code Section III, Division 5 (2017 Edition) for development of advanced non-light water reactors, the following Division 5 portions were reviewed: Article HBB-2000 Material; Article HCB-2000 Material; Article HGB-2000 Material; Mandatory Appendix HBB-I-14 Tables and Figures; and, Nonmandatory Appendix HBB-U Guidelines for Restricted Material Specifications to Improve Performance in Certain Service Applications. In addition to the 2017 Edition, the same parts of the 2019 Edition have also been reviewed as indicated in various sections of the report. This review was conducted by a collaboration of national laboratory and private sector participants with significant industrial experience, including some heavy lifting and deep diving from Clarus Consulting, LLC., all intended to achieve an objective, independent, and practical perspective. The report provides recommendations, descriptions of the evaluation methods, and the source references for the data used. To build confidence required for endorsement of the Code, this review was conducted as a verification and validation of the above Code contents. The objective of verification is to ensure that the Code is free of error – direct or implied; contains the information needed for its use, including proper coverage of the Code-specified materials for the intended application, and completeness and adequacy of references to other portions of the Code. The objective of validation is to authenticate that the Code tabulations and graphs represent design inputs consistent with what are determined using rules and methods specified by the Code. The authentication process used data that were assembled and/or generated independent of Code development, while the methods of analysis followed Code-specified methods where appropriate. The designated portions for this review cover the five alloys codified for high temperature reactor applications in Division 5, i.e. 316 SS, 304 SS, 800H, 2¼Cr-1Mo, and 9Cr-1Mo-V, regarding their general requirements, permitted specifications and design stress intensity values for pressure-retaining applications, deterioration in service, fatigue acceptance test, permissible weld materials, tensile and yield strength, expected minimum stress-to-rupture values (including for Alloy 718), weld stress rupture factors, permissible materials for bolting use, and restricted specifications in certain service applications. Additionally, stress intensity values for bolting materials including 316 SS, 304 SS and alloy 718 were reviewed. Analysis and discussion are also provided on contents outside of these designated Code portions where it was deemed relevant and necessary to develop a technically sound understanding of issues relating to the designated portions. Due to unavailability of sufficient test data on welds during the review period, the weld stress rupture factors in Tables HBB-I-10.14A to E, which cover a total of ten tables for the five alloys welded with twenty-eight different weld metals (some with similar properties), have been deferred to a future review effort. The review identified mainly two types of issues. The first type includes instances where the Code is found factually incomplete or incorrect, such as obsolete materials specifications listings, missing tabulation of stresses for bolting. Changes to the Code are recommended in these cases. The second type of issue includes instances where the Code tabulations and graphs are found to be less conservative than the review analysis results. In these cases, recommendations are made for further review and consideration where the difference in conservatism exceeds 10%, which is our threshold for questioning technical adequacy, meriting a risk assessment by the Nuclear Regulatory Commission and/or reactor designers. It is noted that this effort has been executed using all available data and established methods of analysis, including methods and criteria specified and used by the Code. As such, the findings that are presented in quantitative detail, in a format for convenient comparison with the Code, and with identification of where further review is recommended, should provide a sound technical basis for decisions about quantifying the implications of the reduced design margins and technical adequacy/inadequacy to form a basis for conditioning specific Code tabulation values on endorsement. Recommendations for specific changes to the Code, however, entail design conservatism considerations beyond the scope of this review effort, and are not made in this report.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Stacked networks improve physics-informed training: Applications to neural networks and deep operator networks

Physics-informed neural networks and operator networks have shown promise for effectively solving equations modeling physical systems. However, these networks can happen to be difficult or impossible to train accurately. Here, we present a novel multifidelity framework for stacking physics-informed neural networks and operator networks that facilitates training. We successively build a chain of networks, where the output at one step can act as a low-fidelity input for training a longer chain, gradually increasing the expressivity of the learnt model. The equations imposed at each step of the iterative process can be the same or different (akin to simulated annealing). The iterative (stacking) nature of the proposed method allows us to learn progressively features of a solution which could have been hard to learn directly. Through benchmark problems including a nonlinear pendulum, the wave equation, and the viscous Burgers equation, we show how stacking can be used to improve the accuracy and reduce the required size of physics-informed neural networks and operator networks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Sequential Infiltration Synthesis of Multinary Metal Oxides: The Role of Precursor Diffusion and Reactivity

Understanding volatile metal–organic precursor transport in polymer matrices during sequential infiltration synthesis (SIS) is critical to the design of hybrid materials of tailored composition and uniform structure. Here, we investigate the diffusion and reaction behavior of diethylzinc (DEZ) in poly(methyl methacrylate) (PMMA) matrices with variable indium and zinc incorporation. Unlike conventional SIS where precursors adduct or react directly with the polymer backbone, DEZ shows minimal interaction with PMMA. However, DEZ may interact strongly with In–OH or Zn–OH moieties that are previously incorporated into a PMMA thick film. Using spectroscopic ellipsometry, X-ray photoelectron spectroscopy depth profiling, and EDX line scans, we quantify Zn incorporation as a function of exposure time, polymer thickness, and previous infiltration steps. The Zn distribution evolves from surface-localized to deep within the ∼1 μm thick film with increasing DEZ exposures up to 30 min. A reaction-diffusion model incorporating a hindering factor to capture diffusivity decay due to product accumulation reproduces the Zn depth profile and yields physically meaningful parameters to guide future growth. Furthermore, this study provides a framework for interpreting metal–organic precursor infiltration behavior and highlights the utility of model-guided SIS process design.

diethylzinc↗
DOE OSTI2026

N 2 O Formation Mechanism and Suppression Strategy on Pt Catalysts for NO x Removal from Hydrogen-Internal Combustion Engines

Pt-based catalysts exhibit excellent low-temperature activity in the selective catalytic reduction of NO x with H 2 (H 2 -SCR), but their tendency to form N 2 O poses a significant challenge for practical use. This issue is further complicated by unclear formation mechanisms, hindering the development of more efficient catalysts. This study explored the N 2 O generation mechanisms on Pt catalysts supported by MgO, Al 2 O 3 , SiO 2 , and TiO 2 , aiming to achieve a deep understanding that could advance the Pt catalysts with high NO x conversion and minimized N 2 O emissions. Through systematic kinetics and characterization analyses, the direct influence of the support acidity and reactant dynamics (O 2 and NO) on N 2 O formation was clearly revealed. Notably, the Pt catalysts with strong NO adsorption capacity showed reduced N 2 O generation, highlighting the critical role of NO adsorption sites in the H 2 -SCR process. By incorporation of NO adsorption sites (i.e., MgO, BaO, CeO 2 ) onto Pt/SiO 2 , both the H 2 -SCR efficiency and N 2 selectivity (reduced N 2 O selectivity) were significantly enhanced, effectively reducing the N 2 O emissions through optimized surface NO adsorption. These findings provide a design framework for more selective Pt-based catalysts, advancing H 2 -SCR systems for effective NO x abatement from hydrogen-internal combustion engines, which is a promising carbon-free transportation technology.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Platinum Group Metal Catalysts: Supply Chain Deep Dive Assessment

The report “America’s Strategy to Secure the Supply Chain for a Robust Clean Energy Transition” lays out the challenges and opportunities faced by the United States in the energy supply chain as well as the federal government plans to address these challenges and opportunities. It is accompanied by several issue-specific deep dive assessments, including this one, in response to Executive Order 14017 “America’s Supply Chains,” which directs the Secretary of Energy to submit a report on supply chains for the energy sector industrial base. The Executive Order is helping the federal government to build more secure and diverse U.S. supply chains, including energy supply chains. This report focuses on the supply chain for catalysts, specifically platinum group metal (PGM) catalysts, used for decarbonizing energy technologies. Catalysts are substances that increase the rate, conversion, and selectivity of chemical reactions and are used in a variety of applications such as chemical manufacturing, petroleum refining, and catalytic converters. Catalysts containing PGMs (“PGM catalysts”) are particularly useful in widespread industrial applications, including the production of high-volume chemicals such as ammonia, acetic acid, nitric acid, and the refining of crude oil into petroleum products. The PGM metals possess extraordinary properties such as being active oxidation and hydrogenation catalysts; excellent electrical conductors and electrodes; and outstanding adsorbers of oxygen and hydrogen. Within the energy industrial base, PGM catalysts improve the energy and materials efficiency of petroleum refining and chemical industry processes and reduce energy consumption in manufacturing. In addition to their use in catalytic converters, PGM catalysts are important to maximizing the efficiency of emerging decarbonization technologies, specifically in proton exchange membrane (PEM) electrolyzers for green hydrogen production from water and PEM fuel cells for transportation and stationary energy storage. Green hydrogen is expected to play a significant role in decarbonization scenarios.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2024

Bridging the Gap Between LLMs and LNS with Dynamic Data Format and Architecture Codesign

Deep neural networks (DNNs) have achieved tremendous success in the past few years. However, their training and inference demand exceptional computational and memory resources. Quantization has been shown as an effective approach to mitigate the cost, with the mainstream data types reduced from FP32 to FP16/BF16 and recently FP8 in the latest NVIDIA H100 GPUs. With increasingly aggressive quantization, however, the conventional floating-point formats suffer from limited precision in representing numbers around zero. Recently, NVIDIA demonstrated the potential of using a Logarithmic Number System (LNS) for the next generation of tensor cores. While LNS mitigates the hurdles in representing small numbers, in this work we observed a mismatch between LNS and the emerging Large Language Models (LLM), where LLM exhibits significant outliers when directly adopting the LNS format. In this paper, we present a data-format/architecture codesign to bright this gap. On the format side, we propose a dynamic LNS format to flexibly represent outliers at a higher precision, by exploiting asymmetry in the LNS representation and identifying outliers through a per-vector basis. On the architecture side, for demonstration, we realize the dynamic LNS format in a systolic array, which can handle the irregularity of the outliers at runtime. We implement our approach on an Alveo U280 FPGA as a prototype. Experimental results show that our design can effectively handle the outliers and resolve the mismatch between LNS and LLM, contributing to an accuracy improvement of 15.4% and 16% over the floating-point and the original LNS baselines, using four state-of-the-art LLM models. Our observation and design lay a solid foundation for the large-scale adoption of the LNS format in the next-generation deep learning hardware.

Haghi, Pouya↗