Search NASA⌕ Search

SEARCH · Search NASA

Results for “floating structures”

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 127 records · Page 7

Computationally-driven discovery of second harmonic generation in EuBa 3 (B 3 O 6 ) 3 through inversion symmetry breaking

Nonlinear optical (NLO) crystals with superior properties are significant for advancing laser technologies and applications. Introducing rare earth metals to borates is a promising and effective way to modify the electronic structure of a crystal to improve its optical properties in the visible and ultraviolet range. In this work, we computationally discover inversion symmetry breaking in EuBa 3 (B 3 O 6 ) 3 , which was previously identified as centric, and demonstrate noncentrosymmetry via synthesizing single crystals for the first time by the floating zone method. We determine the correct space group to be P 6¯. The material has a large direct bandgap of 5.56 eV and is transparent down to 250 nm. The complete anisotropic linear and nonlinear optical properties were also investigated with a d 11 of ∼0.52 pm/V for optical second harmonic generation. Further, it is Type I and Type II phase matchable. This work suggests that rare earth metal borates are an excellent crystal family for exploring future deep ultraviolet (DUV) NLO crystals. It also highlights how first principles computations combined with experiments can be used to identify noncentrosymmetric materials that have been wrongly assigned to be centrosymmetric.

He, Jingyang (ORCID:0000000238863659)↗

Comparison of loads from HAWC2 and OpenFAST for the IEA Wind 15 MW Reference Wind Turbine

Reference wind turbines (RWTs) that reflect the state-of-the-art of current wind energy technology are necessary in order to properly evaluate innovative methods in wind turbine design and evaluation. The International Energy Agency (IEA) Wind Technology Collaboration Platform (TCP) Task 37 has recently developed a new RWT geared towards offshore floating-foundation applications: the IEA Wind 15 MW. The model has been implemented in two aeroelastic codes, OpenFAST and HAWC2, based on an underlying common ontology. However, these toolchains result in slightly different structural parameters, and the two codes utilise different structural models. Thus, to increase the utility of the model, it is necessary to compare the aeroelastic responses. This paper compares aeroelastic loads calculated using different fidelities of the blade model in OpenFAST (ElastoDyn and BeamDyn) and HAWC2 (prismatic Timoshenko without torsion and Timoshenko with fully populated stiffness matrix), where both codes use the DTU Basic controller and the same turbulence boxes to reduce discrepancies. The aeroelastic responses to steady wind, step wind and turbulent wind (per IEC 61400-1 wind class IB) are considered. The results indicate a generally good agreement between the loads dominated by aerodynamic thrust and force, especially for the no-torsion blade models. Discrepancies were observed in other load channels, partially due to differences in the asymmetric loading of the rotor and partially due to differing closed-loop dynamics, and they will be the subject of future investigations.

17 WIND ENERGY↗

Going beyond BEM with BEM: an insight into dynamic inflow effects on floating wind turbines

Blade element momentum (BEM) theory is the backbone of many industry-standard wind turbine aerodynamic models. To be applied to a broader set of engineering problems, BEM models have been extended since their inception and now include several empirical corrections. These models have benefitted from decades of development and refinement and have been extensively used and validated, proving their adequacy in predicting aerodynamic forces of horizontal-axis wind turbine rotors in most scenarios. However, the analysis of floating offshore wind turbines (FOWTs) introduces new sets of challenges, especially if new-generation large and flexible machines are considered. In fact, due to the combined action of wind and waves and their interaction with the turbine structure and control system, these machines are subject to unsteady motion and thus unsteady inflow on the wind turbine's blades, which could put BEM models to the test. Consensus has not been reached on the accuracy limits of BEM in these conditions. This study contributes to the ongoing research on the topic by systematically comparing four different aerodynamic models, ranging from BEM to computational fluid dynamics, in an attempt to shed light on the unsteady aerodynamic phenomena that are at stake in FOWTs and whether BEM is able to model them appropriately. Simulations are performed on the UNAFLOW 1:75 scale rotor during imposed harmonic surge and pitch motion. Experimental results are available for these conditions and are used for baseline validation. The rotor is analyzed in both rated operating conditions and low wind speeds, where unsteady aerodynamic effects are expected to be more pronounced. Results show that BEM, despite its simplicity, can adequately model the aerodynamics of FOWTs in most conditions if augmented with a dynamic inflow model.

17 WIND ENERGY↗

Small Co-doping induced magnetic and electrical transitions in single crystal CaF⁢e 0.95 ⁢C⁢o 0.05 ⁢O 3

CaFe⁢O 3 and CaF⁢e 0.95 ⁢C⁢o 0.05 ⁢O 3 single crystals were grown by combing floating-zone method with high oxygen pressure treatment. Both crystals show charge disproportionation of Fe (F⁡e 4+ → F⁡e 3+ + F⁡e 5+ ) accompanied by a metal-to-insulator transition as well as a crystal structural transformation from Pbnm to 𝑃⁢2 1 /𝑛. However, the slight introduction of Co significantly suppresses the critical temperature of charge disproportionation from 290 K in CaFe⁢O 3 to 260 K in CaF⁢e 0.95 ⁢C⁢o 0.05 ⁢O 3 . Different from the single antiferromagnetic phase transition as observed in the polycrystalline CaFe⁢O 3 , two sequential antiferromagnetic transitions are found to occur in these two single crystals with the Néel temperatures around 119 and 112 K for CaFe⁢O 3 crystal and 109 and 98 K for CaF⁢e 0.95 ⁢C⁢o 0.05 ⁢O 3 crystal, respectively. Neutron diffraction illustrates the formation of a spiral antiferromagnetic structure. Moreover, as the temperature further decreases to 65 K, a third magnetic transition accompanied by a second electrical transition is observed in the slightly Co-doped CaF⁢e 0.95 ⁢C⁢o 0.05 ⁢O 3 crystal. In conclusion, first-principles calculations suggest that the introduction of a moderate amount of Co and the peculiar spiral spin texture play an important role in the presence of such new magnetic and electrical transitions.

Xia, Hailiang [Chinese Academy of Sciences (CAS), ↗

Specific Ion Effects in Lanthanide–Amphiphile Structures at the Air–Water Interface and Their Implications for Selective Separation

The use of surfactants to attract dissolved ions to water surfaces and interfaces is an essential step in both solvent-based and solvent-free separation processes. Here, we have studied the interactions of lanthanide ions in the aqueous subphase with monolayers of dihexadecyl phosphate at air–water interfaces. With heavier lanthanides (atomic number Z ≥ 65) in the subphase, the floating layer can be compressed to an area/molecule of about half the molecular cross section, indicating bilayer formation. X-ray fluorescence and reflectivity data support this conclusion. In the presence of lighter lanthanides (Z < 65), only monolayers are observed. Subphase-concentration-dependent studies using Er 3+ (heavier) and Nd 3+ (lighter) lanthanides show a stepwise progression, with ions attaching to the monolayer only when the solution concentration is >3 × 10 –7 M. Above ~10 –5 M, bilayers form but only in the presence of the heavier lanthanide. Grazing incidence X-ray diffraction shows evidence of lateral ion–ion correlations in the bilayer structure but not in monolayers. Explicit solvent all-atom molecular dynamics simulations confirm the elevated ion–ion correlation in the bilayer system. This bilayer structure isolates heavier lanthanides but not lighter lanthanides from an aqueous solution and is therefore a potential mechanism for the selective separation of heavier lanthanides.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Origin of Intrinsically Low Thermal Conductivity in a Garnet-Type Solid Electrolyte: Linking Lattice and Ionic Dynamics with Thermal Transport

Understanding thermal transport in solid electrolytes is essential for improving the performance, reliability, and safety of all-solid-state batteries. Garnet-type lithium-ion conductors are promising candidates for solid electrolytes, yet their thermal-transport mechanisms remain poorly understood. Here, we connect the lattice and ion dynamics of single-crystal garnet-type Li 6.5 La 3 Zr 1.5 Ta 0.5 O 12 to its intrinsically low thermal conductivity. Our study reveals that the single crystals grown by the floating-zone method exhibit remarkably low glasslike thermal conductivity. Using first-principles calculations and inelastic-neutron-scattering measurements, we identify both the acoustic and numerous optical phonon modes, which stem from the complex crystal structure of the material. Notably, a low-energy optical branch exhibits an avoided crossing with acoustic phonons near 7 meV. These optical modes can enhance the scattering of heat-carrying acoustic phonons and reduce thermal conductivity. Furthermore, the calculated Grüneisen parameters are large, especially for the vibrational modes around 6 meV, indicating strong anharmonicity, with a noticeable contribution from lithium-ion vibrations. A two-channel thermal-transport model is employed to describe the weak temperature dependence of the thermal conductivity, which can be attributed to the substantial contribution of diffuson transport facilitated by the abundance of optical phonons and intrinsic anharmonicity. These results offer valuable insights into the thermal transport in a broad class of ionic conductors of interest for energy conversion and storage applications.

Ab initio calculations↗

FOCAL Campaign IV: Integrated System Control: Turbine + Hull

Campaign IV of the Floating Offshore-wind Controls Advanced Laboratory Experimental Program (FOCAL) aimed to generate a dataset enabling the validation of a floating offshore wind turbine with turbine and hull controls. The turbine considered in this campaign is the scaled IEA 15MW Reference Wind Turbine similar to the setup considered in FOCAL Campaign I. This turbine is mounted on the VolturnUS platform with integrated hull controls similar to the hull considered in FOCAL Campaign II/III. This system is deployed for a fully-coupled wind and wave testing campaign at the University of Maine's Harold Alfond Wind and Wave (W2) facility. System dynamics and turbine characteristics are measured to determine global performance and assess the effect of the control strategies employed. NREL's reference open-source controller (ROSCO) is used for active blade pitch control and hull structural control is examined through the use of tuned mass dampers (TMD's) tuned to the systems pitch and tower-bending natural frequency. The naming convention of the load cases considered are described in the data details section. Detailed properties on the modeled system are found in the following reference: Lenfest E., Floating Offshore-wind Controls Advanced Laboratory (FOCAL) Experimental Program - Campaign IV: 1:70 Model-scale Testing of the IEA-Wind 15MW Reference Turbine and the VolturnUS-S Platform. UMaine ASCC Report Number 23-57-1183.

17 WIND ENERGY↗

Oxygen Inhomogeneity and Reversibility in Single Crystal LaNiO{sub 3-{delta}}.

LaNiO 3&minus; &delta; single crystals have been obtained via high pressure floating zone growth under 149 bar of oxygen pressure. We find a radial gradient in the magnetic properties of specimens extracted from the as-grown boule, which we correlate with the appearance of ordered oxygen vacancy structures. This radial oxygen inhomogeneity has been characterized systematically using a combination of magnetization and X-ray scattering measurements. We establish the presence of rhombohedral ( R 3 &macr; c ), oxygen stoichiometric specimens at the periphery of the boule and the presence of a dilute concentration of ordered oxygen-deficient orthorhombic La 2 Ni 2 O 5 in the center. Furthermore, we demonstrate that the as-grown, oxygen-deficient central regions of the crystal can be annealed under high oxygen pressure, without loss of crystallinity, into fully oxygenated LaNiO 3 , recovering magnetic properties that are characteristic of stoichiometric specimens from the exterior region of the crystal. Thus, single crystals of LaNiO 3&minus; &delta; possess oxygen content that can be reversibly modified under oxidizing and reducing conditions.

crystal growth↗

A Study of Cost-Saving Potential of Load Flexibility Measures in Grid-Interactive Multifamily Buildings

With recent advances in smart technologies, more and more smart devices are penetrating the residential and commercial buildings market. The introduction of these smart devices is also helping IoT companies emerge with load aggregator roles in the sector. With more utility companies on the track of supporting OpenADR protocols, the aggregators could play a significant role in providing load flexibilities by automatically responding to demand response (DR) events and coordinating load flexibility measures between customers. This would benefit utility companies by reducing stress on the grid during critical peak demand hours as well as customers by allowing them to utilize utility rate structures advantageous to those able to reduce electric usage during high-demand hours. This study evaluates cost and energy savings from adopting multiple load flexibility measures in multifamily buildings. Combinations of different load flexibility measures, including space temperature floating, light dimming, automatic window shading, and water heater temperature floating, are considered. The simulations are performed using OpenStudio®, an open-source U.S. Department of Energy (DOE) simulation platform. For the case study, we used a midrise apartment building with weather conditions from Denver, Colorado. To compare climate zone differences, simulations were also performed for Los Angeles, California, and Chicago, Illinois. Initial results indicate that the application of automated load flexibility measures without careful consideration of dispatching strategies and DR program enrollments could significantly affect the savings. To get meaningful cost savings, aggregators need to encourage tenant awareness to curtail energy usage through occupant behavior in addition to dispatching automatic load flexibility measures. The outcomes from this study are believed to help load aggregators understand the risks and benefits of load flexibility opportunities.

building energy modeling↗

Frustrated Magnetism and Ferroelectricity in a Dy 3+ -Based Triangular Lattice

Triangular lattice magnets have attracted extensive research interest because they are potential hosts for geometrically frustrated magnetism and strong quantum fluctuations. Here, utilizing a laser floating zone technique, we report the first-time successful growth of a DyInO 3 sizable crystal, which contains Dy3+-based triangular layers. The fine-tuning of Indium stoichiometry was found to be the key factor in the stabilization of the desired hexagonal phase. The X-ray diffraction study of the crystal structure reveals a non-centrosymmetric P6 3mc space group. Switchable polarization, i.e., ferroelectricity, and ferroelectric domain configuration are experimentally demonstrated at room temperature. Anisotropic magnetic and thermodynamic measurements unveil antiferromagnetic interactions, the absence of long-range ordering down to 0.1 K, and a possible doublet ground state, indicating a strongly frustrated magnetism. Our findings suggest that the DyInO 3 crystal is an excellent platform for studying emergent phenomena and their interplay with coherent topological defects in the quantum realm.

36 MATERIALS SCIENCE↗

Ligand structure and adsorption free energy of nanocrystals on solid substrates

In this paper, we present an investigation on the absorption of alkylthiolated nanocrystals on a solid substrate. We calculate adsorption free energies and report a number of effects induced by the substrate. Nearest neighbor distances and bonding free energies are significantly different than for a free floating case, there is a weakening of bonding free energies among nanocrystals, and the adsorption is manifestly anisotropic, i.e., stronger along certain directions of the nanocrystal core. We contend that this last result accounts for the Bain transition (fcc → bcc) observed in experimental results. We report the presence of vortices induced by the substrate, which explain the increased nearest neighbor distance among nanocrystals, which is in excellent quantitative agreement with experimental results and with the predictions of the Orbifold Topological Model. Implications for the assembly of nanostructures and future experiments are also discussed.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Pushing the efficiency limit of low-cost, industrially relevant Si solar cells to > 22.5% by advancing cell structures and technology innovations (Final Report)

The overall objective of this program is to achieve ~23% bifacial n-type cell efficiencies by developing and implementing optimized homogeneous or selective boron (B) emitter on front and tunnel oxide passivated contact (TOPCon) on rear side, in combination with advanced fine-line screen-printing metallization with floating busbars. During this research project, first we developed a technology roadmap to drive the 21% n-PERT cell efficiency from 21% to 23% by transforming the cell design to n-TOPCon and establishing the requirements for each layer, including B emitter, rear n-TOPCon, n-base Si and screen-printed contacts. Next, consistent with our roadmap, we developed advanced homogeneous implanted B emitter (150-180 Ω/⌫) passivated with ALD Al 2 O 3 layer capped with PECVD SiN x /SiO x double-layer antireflection coating This gave a very low recombination current density of 10-15 fA/cm 2 prior to metallization. In addition, we demonstrated metallized recombination current density of ~31 fA/cm 2 for this advanced homogeneous B emitter with industrial screen-printed, fire-through contacts with 40 μm wide grid lines, floating busbars and implementation of an advanced Ag-Al paste which resulted in local or reduced area metal-Si contact under the grid lines with virtually no emitter surface etching. This paste reduced the full area metallized J oe,metal from >1100 fA/cm 2 to ~700 fA/cm 2 . We also developed novel processes for the formation of p + /p ++ selective B emitter by a) single B diffusion with selective etch back and b) two-steps diffusion with implanted B in field region and APCVD B diffusion under the metal grid. We achieved very low un-metallized recombination current density (J 0 ) of ~18 fA/cm 2 for the selective p ++ p + emitters (30/150 Ω/⌫) and metallized J 0 of ~28 fA/cm 2 with ~3% screen-printed metal contact to p ++ regions. Next, we developed the technology for n-TOPCon by growing phosphorus-doped LPCVD and PECVD poly-Si on top of ~ 15Å chemically grown (NAO) tunnel oxide. After an optimized anneal at 875 °C for 30 min, passivated n-TOPCon have unmetallized J 0 of ~ 5 fA/cm 2 which went down further to ~ 1 fA/cm 2 after a 700 Å SiN x capping layer and simulated contact firing cycle at 770 °C. After screen-printed fire-through metallization on this n-TOPCon with ~13% metal coverage, metallized J 0 value increased to only ~5 fA/cm 2 which is among the lowest reported value to the best of our knowledge for screen-printed metallization. Finally, we integrated all the above technology innovations and enhancements and demonstrated low-cost manufacturable screen-printed n-TOPCon bifacial Si solar cell with ~23% efficiencies. Based on the experimental and theoretical understanding developed in this project, we have developed a new technology roadmap that shows that implementation of busbarless contacts, 10-20 ms bulk lifetime Si, and selective B emitter or selective TOPCon on the front can drive ~23% efficient cells achieved in this research to ~25% at low-cost.

14 SOLAR ENERGY↗

Synthesis and characterization of the novel breathing pyrochlore compound Ba 3 ⁢Tm 2⁢ Zn 5⁢ O 11

In this study, a novel material from the rare-earth based breathing pyrochlore family, Ba 3 ⁢Tm 2 ⁢Zn 5 ⁢O 11 , was successfully synthesized. Powder x-ray diffraction and high-resolution powder neutron diffraction confirmed phase purity and the $F\bar4$3⁢𝑚 breathing pyrochlore crystal structure, while thermogravimetric analysis revealed incongruent melting behavior compared to its counterpart, Ba 3 ⁢Yb 2⁢ Zn 5⁢ O 11 . High-quality single crystals of Ba 3 ⁢Tm 2 ⁢Zn 5 ⁢O 11 were grown using the traveling solvent floating zone technique and assessed using Laue x-ray diffraction and single crystal x-ray diffraction. Further, thermodynamic characterization indicated paramagnetic behavior down to 0.05 K, and inelastic neutron scattering measurements identified distinct dispersionless crystal electric field energy bands, with the fitted crystal electric field model predicting a single-ion singlet ground state and an energy gap of ∼9 meV separating it from the first excited (singlet) state. Additional low-energy excitation studies on single crystals revealed dispersionless bands at 0.8 and 1 meV. Computed phonon dispersions from first-principles calculations ruled out phonons as the origin of these modes, further illustrating the puzzling and unique properties of Ba 3 ⁢Tm⁢ 2 Zn 5 ⁢O 11 .

36 MATERIALS SCIENCE↗

Projected Urban Morphology of the Los Angeles Area by the Year 2100

This dataset provides projections of urban building morphologies for the Los Angeles urban area at 30-meter spatial resolution. It contains 192 raster files that detail two primary building attributes: building footprint fractions (ranging from 0 to 1) and average building heights (ranging from 0 to 75 meters). The projections account for a wide range of future pathways, covering two Shared Socioeconomic Pathway (SSP) scenarios (SSP3 and SSP5), two population scenarios, two developed land intensification scenarios, and four distinct levels of intensification. The dataset was created using dual Generative Adversarial Networks (GANs) trained on 2015 land cover and building properties from the National Land Cover Database (NLCD) and Model America datasets. Supporting information on the dataset has been described in the LAUrbanAreaMorphologyProjections2100_README.txt file.

Pandey, Bhartendu↗

An adaptive model-free robotic force control strategy for hydrodynamic real-time hybrid simulation of floating offshore wind turbines

Real-time hybrid simulation (RTHS) - a cyber-physical testing approach - promises to enhance the simulation fidelity of the model-scale experiments used to prototype floating offshore wind turbines (FOWTs). In hydrodynamic RTHS (hydro-RTHS), actuators emulate aerodynamic forces on model-scale FOWT specimens subjected to physical waves in a hydrodynamic laboratory. Robotic arms are promising candidates for actuation in hydro-RTHS due to their compact multi-degree-of-freedom (DOF) capabilities. Unlike classical RTHS for seismic applications, which typically relies on displacement control, hydro-RTHS requires 6-DOF force control on newly designed floating prototypes in a model-scale setting, which presents significant challenges, including modeling uncertainties, directional asymmetry, configuration drift, bandwidth limitations, and time-varying delays. To mitigate these constraints without extensive pre-test calibration, this study proposes an adaptive model-free robotic force control strategy that combines task-space explicit force control with a secondary joint-space pose-keeping task. The Adaptive Feedforward Compensator (AFC) is integrated into the force control loop to compensate for time-varying delay. Experimental testing was conducted using a Franka Emika Panda robotic arm with a 1:50 scale FOWT specimen under operational wind and wave conditions. Results demonstrate stable and consistent 6-DOF force tracking. Effective delay compensation was observed, with low-frequency delay reductions ranging from 71.4% to 91.8% and improvements in low-frequency surge force tracking of 25.0% to 52.1%. This study enhances robotic actuation performance in hydro-RTHS and introduces a force control strategy that supports reliable robotic operation in uncertain floating environments. Future work will explore disturbance-observer mechanisms to further enhance wave rejection capabilities under extreme wind and wave conditions.

17 WIND ENERGY↗

Data Fusion to Enhance Quality Control and Analysis with Instruments at the Marine and Coastal Research Laboratory

Deploying environmental monitoring instruments in the marine environment can be challenging, facing challenges around device survivability, biofouling and corrosion, and consistent data collection. This project explores the use of data fusion – the process of integrating multiple data sources to produce more consistent, accurate, and useful information – to build a consistent long-term monitoring system at the Marine and Coastal Research Laboratory (MCRL) in Sequim, Washington. Unused instruments that had been acquired from past projects were inventoried and deployments planned on the MCRL pier and floating dock. A total of 8 instruments were deployed including a tide gauge, hydrophone, acoustic Doppler current profiler (ADCP), photosynthetically active radiation (PAR) sensors, meteorological station, and three water quality sensors. Deployments were planned to be well-protected around the pier structure and a maintenance schedule was created for cleaning and recalibration. An automated data pipeline was created to aggregate data on edge computers that push data to Amazon Web Services (AWS) cloud storage every 15 minutes, performing automated quality control and data transformations using the Time Series Data Analytical Toolkit (TSDAT). Continued efforts are underway to maintain this system into the future, take a data-driven approach to maintenance scheduling, improve the reliability of the system, and share the data with a variety of end-users.

54 ENVIRONMENTAL SCIENCES↗

FOCAL Campaign I: Advanced Wind Turbine Control Strategies

Campaign I of the Floating Offshore-wind Controls Advanced Laboratory Experimental Program (FOCAL) aims to generate a dataset enabling the validation of aerodynamic performance of a scaled turbine mounted on a rigid tower in a fixed condition. The turbine considered in the FOCAL testing campaigns is the IEA-Wind 15MW Reference Wind Turbine. This scaled model is capable of simulating advanced blade-pitch control strategies in a high-quality wind field. The turbine is fully instrumented to record a variety of parameters in real time such as structural loads and dynamics. The test data considered was generated at the University of Maine's Harold Alfond Wind and Wave (W2) testing facility. The Load Cases (LC) considered in this testing campaign are as follows: LC 1.X - Constant wind and blade-pitch with varying rotor speeds LC 2.X - Constant wind and rotor speed with varying blade-pitch LC 3.X - Varying wind with active closed-loop control Detailed properties on the modeled system are found in the following reference: Lenfest E., Floating Offshore-wind Controls Advanced Laboratory (FOCAL) Experimental Program - Campaign I: 1:70 Model-scale Testing of the IEA-Wind 15MW Reference Turbine. UMaine ASCC Report Number 23-40-1183. Details on the results of the verification and validation are found in the following reference: Mendoza, Nicole et al., "Verification and Validation of Model-Scale Turbine Performance and Control for the IEA Wind 15 MW Reference Wind Turbine," Energies, vol. 15, no. 20, 2022, https://doi.org/10.3390/en15207649.

17 WIND ENERGY↗

Efficient Mixed-Precision Matrix Factorization of the Inverse Overlap Matrix in Electronic Structure Calculations with AI-Hardware and GPUs

In recent years, a new kind of accelerated hardware has gained popularity in the artificial intelligence (AI) community which enables extremely high-performance tensor contractions in reduced precision for deep neural network calculations. In this article, we exploit Nvidia Tensor cores, a prototypical example of such AI-hardware, to develop a mixed precision approach for computing a dense matrix factorization of the inverse overlap matrix in electronic structure theory, S –1 . This factorization of S –1 , written as ZZT = S –1 , is used to transform the general matrix eigenvalue problem into a standard matrix eigenvalue problem. Here we present a mixed precision iterative refinement algorithm where Z is given recursively using matrix–matrix multiplications and can be computed with high performance on Tensor cores. To understand the performance and accuracy of Tensor cores, comparisons are made to GPU-only implementations in single and double precision. Additionally, we propose a nonparametric stopping criteria which is robust in the face of lower precision floating point operations. The algorithm is particularly useful when we have a good initial guess to Z, for example, from previous time steps in quantum-mechanical molecular dynamics simulations or from a previous iteration in a geometry optimization.

36 MATERIALS SCIENCE↗