Search NASA⌕ Search

SEARCH · Search NASA

Results for “continuous processing”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

High Performance Solvent for NGCC Flue Gas CO 2 Capture (Final Technical Report)

Amine-based solvent absorption is the most mature and reliable technology for large scale CO 2 capture, dating back to the 1930s when monoethanolamine (MEA) was used to treat acid gases from oil refineries. However, while strategic advancements have optimized the CO2 capture process, the cost of capture remains high, where current estimates suggest that CO 2 capture costs are around $\$$72/tonne of CO 2 . To address this, solvent development and optimization have become a focus of current research. This project sought to develop a high-performance solvent to reduce the overall cost of CO 2 capture from NGCC flue gas. Here, solvent optimization focused on: (1) reducing the energy required for CO 2 desorption in a reboiler, (2) improving CO 2 absorption and desorption reaction kinetics, (3) improving solvent stability, and (4) reducing environmental impacts. Susteon has developed and evaluated a promoted solvent, Sustenol™, for NGCC flue gas CO 2 capture. The optimized Sustenol™ also shows a higher dynamic CO 2 absorption capacity of ~0.5 mol CO2 /mol amine compared to 0.25 mol CO2 /mol amine for 30 wt% MEA. Additionally, the solvent is oxidatively, thermally and hydrothermally stable, which leads to lower solvent loss and emissions. These advancements have resulted in a solvent regeneration energy of 2.16 GJ/tonne of CO 2 which is >30% lower than current state-of-the-art commercial and emerging solvents. Combined with empirical data from the bench and pilot scale testing, this preliminary TEA study indicated the cost of CO 2 capture by Sustenol™ for 97% CO 2 removal at $\$$54/tonne and for 90% removal at $\$$49/tonne, with a pathway to achieve $\$$45/tonne of CO 2 with continued process and solvent advancements. Susteon has developed a technology roadmap to reduce the cost of CO 2 capture to <$\$$45/tonne for NGCC flue gas. Susteon plans to derisk this technology for commercial deployment through comprehensive solvent degradation testing, long-term testing in a pilot plant at 5 tonne CO2 /day and demonstration scale testing at 100 tonne CO2 /day with NGCC flue gas and engineering design studies to qualify Sustenol™ as a drop-in replacement solvent.

03 NATURAL GAS↗
DOE OSTI2022

Melt-Spun PAN Precursor for Cost-Effective Carbon Fibers in High Pressure Compressed Gas Tankage

The objective of this program was to synthesize melt-spinnable polyacrylonitrile (PAN) precursor that will lead to carbon fiber 25% less expensive than carbon fiber produced from conventional PAN precursor when produced in large volumes. The carbon fiber produced from this precursor needs to achieve ≥ 700 ksi tensile strength with this cost reduction to replace the incumbent fiber. In this program, the technology for producing precursor and then converted carbon fiber (CF) capable of meeting the 700 ksi target was to be demonstrated at a small scale in tows, containing ≥ 100 filaments. If this program moved on to a “Phase 2”, the processes would be scaled up to produce and test Type IV compressed gas storage tanks capable of meeting the 700-bar ultimate targeted FOA performance level. Precursor fiber production typically contributes a large fraction (~50%) to finished CF manufactured cost. Presently, all high strength CFs are manufactured from solution spun PAN precursor, accomplished at low temperatures using large amounts of environmentally unfriendly solvents that impart major impact to line speed and costs involved with handling and reclaiming the solvents and associated “waste streams” required in spinning and extracting those solvents. PAN melt spinning was demonstrated by the company BASF in CF precursor production and used in production of some other acrylic fibers, but significant production challenges remain for meeting the established CF metrics. Melt spinning requires use of plasticizers to suppress the melt temperature of the predominantly PAN formulation to a point sufficiently below the temperature of significant cross-linking. Fortunately, BASF and this team’s earlier work had demonstrated acrylic formulations melt spun with combinations of plasticizing solvents at much lower solvent levels. This project team made significant progress in understanding technical issues and identifying formulation pathways in earlier work in this area where >250ksi strength was achieved. The goal of this program was to pursue a more comprehensive approach with a larger, more diverse team having greater equipment design and overall production capabilities to push rapidly beyond the science to resolve remaining “engineering” issues associated with establishing stable, continuous processes for precursor fiber leading to enhanced, cost-effective CF and composite tank production.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Active- and transfer-learning applied to microscale-macroscale coupling to simulate viscoelastic flows

Active- and transfer-learning are applied to microscale dynamics of polymer flows for the multiscale discovery of effective constitutive approximations required in viscoelastic flow simulation. The result is macroscopic rheology directly connected to a microstructural model. Micro and macroscale simulations are adaptively coupled by means of Gaussian process regression (GPR) to run the expensive microscale computations only as necessary. This multiscale method is demonstrated with flows of a polymer solution as a model system. At the microscale level dissipative particle dynamics (DPD) is employed to model the fluid as a suspension of bead-spring micro-structures subjected to steady shear flow. The results yield the non-Newtonian viscosity and the first normal stress difference at strain rates as training data used in a GPR model. DPD parameters are calibrated with respect to experimental data for a real polymer solution. Compliance with these data requires adjustment of the DPD model's cutoff radius, which then becomes a function of the second invariant of the strain rate tensor. The FENE-P model is chosen for the macroscale description using the spectral element method (SEM) to simulate channel flow and flow past a circular cylinder. The DPD results at the lowest possible shear strain rate yield an estimate of the zero-shear rate viscosity, which allows the initiation of the macroscale flow by SEM as a Newtonian fluid. The resulting strain-rate field is surveyed to determine additional shear strain rate sampling points for the DPD system. This new information allows an initial fitting of parameters of the constitutive equation followed by new SEM simulations at the macroscale. Additionally, guided by active-learning GPR to select new sampling points, this process continues until convergence is achieved. The effectiveness of this new simulation paradigm for viscoelastic flows is tested with different macroscale operating conditions. The effective closure learned in the channel simulation is then transferred directly to the flow past a circular cylinder at low Reynolds number, where the results show that only two additional DPD simulations are required to achieve a satisfactory constitutive model. With an increase of the Reynolds number, the active-learning scheme automatically detects the inaccuracy of the learned constitutive model, and initiates additional DPD simulations for the extra data needed to once again close the microscale-macroscale coupled system. This new paradigm of active- and transfer-learning for multiscale modeling is readily applicable to other microscale-macroscale coupled simulations of complex fluids and other materials. Furthermore, the coupling between microscale and macroscale solvers can be seamlessly implemented with our open source multiscale universal interface (MUI) library.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Aggregation of Leslie matrix models with application to ten diverse animal species

Some grouping is necessary when constructing a Leslie matrix model because it involves discretizing a continuous process of births and deaths. Here, the level of grouping is determined by the number of age classes and frequency of sampling. It is largely unknown what is lost or gained by using fewer age classes, and I address this question using aggregation theory. I derive an aggregator for a Leslie matrix model using weighted least squares, determine what properties an aggregated matrix inherits from the original matrix, evaluate aggregation error, and measure the influence of aggregation on asymptotic and transient behaviors. To gauge transient dynamics, I employ reactivity of the standardized Leslie matrix. I apply the aggregator to 10 Leslie models developed for animal populations drawn from a diverse set of species. Several properties are inherited by the aggregated matrix: (a) it is a Leslie matrix; (b) it is irreducible whenever the original matrix is irreducible; (c) it is primitive whenever the original matrix is primitive; and (d) its stable population growth rate and stable age distribution are consistent with those of the original matrix if the least squares weights are equal to the original stable age distribution. In the application, depending on the population modeled, when the least squares weights do not follow the stable age distribution, the stable population growth rate of the aggregated matrix may or may not be approximately consistent with that of the original matrix. Transient behavior is lost with high aggregation.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Improving the Recyclability of Polymer Composites With Cellulose Nanofibrils

Cellulose nanofibers (CNFs) have been widely studied for their reinforcing potential in high-performance composites. While there are numerous publications on CNF-reinforced composites in a variety of polymer matrices, few have considered the recyclability of such thermoplastic composites and whether the incorporation of CNFs deteriorates or improves their performance upon reprocessing. In this study, two thermoplastic resins, poly(lactic acid) (PLA), and glycol-modified polyethylene terephthalate (PETg), were prepared with CNF reinforcement and thermomechanically recycled to investigate the effect of CNF inclusion on the composite properties after reprocessing as well as their effect on the composites’ number of useful life cycles. Additionally, changes in mechanical, thermal, rheological, molecular, and microstructural properties of the composites and/or base resins were monitored as a function of cycle numbers. As is typical, the polymers’ molecular weight and mechanical performance deteriorated with continued processing. However, the addition of spray dried CNF was found to better maintain the mechanical performance of both polymers throughout multiple recycling steps as compared to neat samples. For example, the tensile strength of PETg with 20 wt% CNF after 6 processing cycles was found to exceed that of virgin neat PETg, and higher loadings of CNF were found to preserve a higher yield strength during multiple rounds of reprocessing compared to PETg composites with lower CNF loadings. Ultimately this study indicates that the addition of CNF to some thermoplastic materials can increase both their sustainability by offsetting the use of high-embodied energy resins and their circularity by enabling performance retention over more use cycles.

42 ENGINEERING↗
DOE OSTI2022

Review of Computer-Aided Manufacturing (CAM) strategies for hybrid directed energy deposition

Hybrid additive manufacturing intertwines both additive and subtractive manufacturing layer by layer to digitally fabricate parts with complex geometries, improved surface finish and tight dimensional accuracies, the sum of which is difficult to obtain with any single process. Computer-Aided Manufacturing (CAM) software is required to orchestrate the machine toolpathing for both the deposition as well as the machining processes and is crucial for the successful fabrication of high quality structures. Additionally, CAM requires substantial operator input to account for challenging aspects of each fabricated structure. For example, deciding at which layer of deposition will the machining process continue to maintain access to complex cavities for finishing - internal features that would otherwise be unfinishable due to reach limitations or obstructions. Moreover, of the many commercially-available hybrid systems, each has a unique kinematic environment which can benefit from specific optimization of toolpath planning and substantial research has directly correlated toolpathing with the microstructure evolution, mechanical properties, porosity and residual stress state of the final fabricated part. This review explores the available strategies for CAM in the context of hybrid direct energy deposition, discusses the advantages and disadvantages of each and considers future CAM trends for this transformational digital manufacturing technology.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Modular machine learning-based elastoplasticity: Generalization in the context of limited data

The development of highly accurate constitutive models for materials that undergo path-dependent processes continues to be a complex challenge in computational solid mechanics. Challenges arise both in considering the appropriate model assumptions and from the viewpoint of data availability, verification, and validation. Recently, data-driven modeling approaches have been proposed that aim to establish stress-evolution laws that avoid user-chosen functional forms by relying on machine learning representations and algorithms. However, these approaches not only require a significant amount of data but also need data that probes the full stress space with a variety of complex loading paths. Furthermore, they rarely enforce all necessary thermodynamic principles as hard constraints. Hence, they are in particular not suitable for low-data or limited-data regimes, where the first arises from the cost of obtaining the data and the latter from the experimental limitations of obtaining labeled data, which is commonly the case in engineering applications. In this work, we discuss a hybrid framework that can work on a variable amount of data by relying on the modularity of the elastoplasticity formulation where each component of the model can be chosen to be either a classical phenomenological or a data-driven model depending on the amount of available information and the complexity of the response. The method is tested on synthetic uniaxial data coming from simulations as well as cyclic experimental data for structural materials. The discovered material models are found to not only interpolate well but also allow for accurate extrapolation in a thermodynamically consistent manner far outside the domain of the training data. This ability to extrapolate from limited data was the main reason for the early and continued success of phenomenological models and the main shortcoming in machine learning-enabled constitutive modeling approaches. Training aspects and details of the implementation of these models into Finite Element simulations are discussed and analyzed.

42 ENGINEERING↗
DOE OSTI2023

High-gradient magnetic separation of colloidal uranium oxide particles from soil components in aqueous suspensions

The separation of uranium oxide (UO 2 ) particles from soil-surrogate particles in aqueous suspensions was achieved using filtration enhanced by a magnetic field. Enhanced attraction of paramagnetic UO 2 colloids to a ferromagnetic stainless-steel filter placed in a strong magnetic field arises because of the positive magnetic susceptibility of the particles and the high-gradient field generated near ferromagnetic fibers. Enhanced uptake of smaller particles over larger ones occurs through Brownian motion that promotes the collision of particles with the ferromagnetic fibers of the filter. Hence, this work focused on UO 2 particles in the colloidal size range. Experiments used a water-cooled electromagnet and an array of permanent magnets. Chemical analysis showed that the magnetic field increased the capture efficiency of uranium particles from a range of 27–53% with the magnet off up to 98% with the magnet on after a single pass of the suspension through the filter. Further, the recovery of the UO 2 particles from the filter, however, was more difficult to achieve. Small amounts of UO 2 , together with significant amounts of background SiO 2 particles, were removed from the filter during a first flush with the magnetic field on. A much larger recovery of UO 2 was not observed until a second out-of-field flush was performed, which also released some SiO 2 . The degree to which particle separation was enhanced through the use of multi-stage filtration compared to single pass-through filtration was also examined. A design was suggested that could be used to optimize the separation efficiency for a continuous process.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2024

Overcoming barriers to medium-chain fatty alcohol production

Medium-chain fatty alcohols ( mc FaOHs) are aliphatic primary alcohols containing six to twelve carbons that are widely used in materials, pharmaceuticals, and cosmetics. Microbial biosynthesis has been touted as a route to less-abundant chain-length molecules and as a sustainable alternative to current petrochemical processes. Several metabolic engineering strategies for producing mc FaOHs have been demonstrated in the literature, yet processes continue to suffer from poor selectivity and mc FaOH toxicity, leading to reduced titers, rates, and yields of the desired compounds. Here, this opinion examines the current state of microbial mc FaOH biosynthesis, summarizing engineering efforts to tailor selectivity and improve product tolerance by implementing engineering strategies that circumvent or overcome mc FaOH toxicity.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Lattice response to ambient self-irradiation of δ-phase 239 Pu-Ga alloys exposed to cryogenic temperatures

Self-irradiation in plutonium is a continuous process that introduces radiation damage ranging from point defects to nanoscale radiation damage as well as decay daughter product impurities that require to be accommodated by the host material. These changes are affecting the structure and properties of the material by lattice swelling, mechanical strengthening, and other alterations. X-ray diffraction and dilatometry studies have shown that the room temperature self-irradiation of δ-phase Pu-Ga alloys (2 and 7 at%Ga) results in swelling of the lattice that saturates after 0.1–0.2 displacement per atom (dpa) of accumulated lattice damage. Here, for a better understanding of these self- irradiation processes, defect accumulation and damage evolution, this time-of-flight neutron diffraction study investigates the impact of ambient self-irradiation effects on the structural evolution of δ-phase 239 Pu-Ga alloys exposed to cryogenic temperatures (<120 K), associated with a varied history of storage time and thermal treatments. Two key behaviors were observed in the lattice evolution after exposure to cryogenic temperatures with respect to room-temperature accumulated self-irradiation damage. Samples with no self- irradiation damage accumulation under ambient conditions showed swelling of the lattice at the end of the measurement, while samples that accumulated significant self-irradiation damage under ambient conditions showed only a partial recovery of the lattice parameter value upon completion of the measurements (cooling followed by heating back to room temperature). Possible causes of the behavior are proposed, and their associated effects are discussed.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Pulling thin single crystal silicon wafers from a melt: The new leading-edge solar substrate

The Floating Silicon Method (FSM) has been established as a viable, stable method for growing single crystal ribbons directly from a silicon melt. With intense helium jet cooling to drive the linear progress of a [111] facet, pulled in the <110> direction, ribbons in the 0.6 – 3.0 mm thickness range can be grown at linear growth rates from 0.3 mm/s to >6 mm/s as reported in the literature. We report on recent progress towards growing (100) oriented ribbons with a net thickness of less than 200 microns and a ribbon width up to 18 cm using a stable, continuous process in the Leading Edge prototype furnace. Here, the 3D details of the single crystal growth are explained by extending the mechanics of the Limit Cycle Theory, with novel Internal Side Effect morphology described by a proposed Facet Flow Theory. Grown-in crystalline defect distributions are described as well as values of critical impurities like oxygen, carbon, dopants, and metals that are relevant for use as wafers for solar cells.

14 SOLAR ENERGY↗
DOE OSTI2024

Gas-phase surface modification to control catalyst structure and yields in methane dehydroaromatization

Methane dehydroaromatization (MDA) is a promising approach for direct methane transformation to aromatics and hydrogen. The benchmark catalyst Mo/H-ZSM-5 struggles to find commercial adoption because of thermodynamically-limited yields and rapid coking on Brønsted acid and molybdenum carbide species, especially on zeolite external surfaces. Here, gas-phase atomic layer deposition (ALD) overcoats H-ZSM-5 external surfaces with SiO 2 or Al 2 O 3 . NH 3 -TPD, HRTEM, and textural properties show that these overcoats exclusively passivate zeolite external surfaces. Under MDA conditions, SiO 2 gives softer coke and increases cumulative benzene yields by 25%, while Al 2 O 3 strongly decreases yields. H 2 -TPR and UV-visible and Raman spectroscopy show how the overcoats redisperse the MoO x precatalysts, especially over multiple deactivation and isothermal oxidative regeneration cycles. Combined with 27 Al-MAS NMR, MoO x redistribution and dealumination are seen as the causes of long-term deactivation over multiple regeneration cycles, and this process continues to occur regardless of the overcoat. Altogether, the deposition of a small amount of silica on the outer surface of Mo/H-ZSM-5 reduces the formation of hard coke, which could be regenerated by milder methods such as hydrogen treatment.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

A multifunctional technology platform for sorbent construction using polyacrylonitrile scaffolds

Polyacrylonitrile (PAN) is a synthetic polymer that shows high potential for use in a wide range of environmental remediation applications. PAN can be implemented in various ways within a batch or continuous process stream for use as a passive scaffold holding active gettering materials in place or where the PAN scaffold (e.g., beads, fiber mats, membranes) is functionalized with active chelating groups (e.g., amine, hydrazide, amidoximes, carboxyl). Application spaces covered in this review include remediation of heavy metals (e.g., Ag, As, Cd, Cr6+, Cu, Pb, Sb, and Se), high-dose fission products (e.g., 90Sr, 137Cs), radioiodine (i.e., 129I), noble gases (i.e., Xe, 85Kr), rare earths (e.g., Ce, Y), and actinides (e.g., Am, Pu, U). Methods for producing PAN composite sorbents are discussed. Options are also discussed for removing the PAN matrix following chemisorption of an active contaminant to minimize waste volumes requiring disposal.

polyacrylonitrile, composite sorbents, sulfides, a↗
DOE OSTI2021

Reactive Species and Reaction Pathways for the Oxidative Cleavage of 4-Octene and Oleic Acid with H 2 O 2 over Tungsten Oxide Catalysts

The oxidative cleavage of carbon-carbon double bonds (C=C) in alkenes and fatty acids produces aldehydes and acids valued as chemical intermediates. Solid tungsten oxide catalysts are low cost, non-toxic, and selective for the oxidative cleavage of C=C bonds with hydrogen peroxide (H 2 O 2 ) and are, therefore, a promising option for continuous processes. Despite the relevance of these materials, the elementary steps involved and their sensitivity to the form of W sites present on surfaces have not been described. In this work, we combine in situ spectroscopy and rate measurements to identify significant steps in the reaction, the reactive species present on catalysts, and examine differences between the kinetics of this reaction on isolated W atoms grafted to alumina and on those exposed on crystalline WO 3 nanoparticles. Raman spectroscopy shows that W-peroxo complexes (W-(η 2 -O 2 ) formed from H 2 O 2 ) react in a kinetically relevant step to produce epoxides, which undergo hydrolysis at protic surface sites. Subsequently, the CH 3 CN solvent deprotonates diols to form alpha hydroxy ketones that react to form aldehydes and water following nucleophilic attack of H 2 O 2 . Turnover rates for oxidative cleavage, determined by in situ site titrations, on WO x -Al 2 O 3 are 75% greater than those on WO 3 at standard conditions. These differences reflect activation enthalpies (ΔH ‡ ) for oxidative cleavage of 4-octene that are much lower for the isolated WO x sites (36±3 and 60±6 kJ∙mol -1 for WO x -Al 2 O 3 and WO 3 , respectively) and correlate strongly with the difference between enthalpies of adsorption for epoxyoctane (ΔH ads,epox ), which resembles the transition state for epoxidation. The WO x -Al 2 O 3 catalysts mediate oxidative cleavage of oleic acid with H 2 O 2 following a mechanism comparable to that for 4-octene. The WO 3 materials, however, form only the epoxide and do not cleave the C-C bond or produce aldehydes and acids. These differences reflect the distinct site requirements for these reaction pathways and indicate that acid sites required for diol formation are strongly inhibited by oleic acids and epoxides on WO 3 whereas the Al 2 O 3 support provides sites competent for this reaction and increase the yield of the oxidative cleavage products.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Mild-to-wild plastic transition is governed by athermal screw dislocation slip in bcc Nb

Plastic deformation in crystals is mediated by the motion of line defects known as dislocations. For decades, dislocation activity has been treated as a homogeneous, smooth continuous process. However, it is now recognized that plasticity can be determined by long-range correlated and intermittent collective dislocation processes, known as avalanches. Here we demonstrate in body-centered cubic Nb how the long-range and scale-free dynamics at room temperature are progressively quenched out with decreasing temperature, eventually revealing intermittency with a characteristic length scale that approaches the Burgers vector itself. Plasticity is shown to be bimodal across the studied temperature regime, with conventional thermally-activated smooth plastic flow (‘mild’) coexisting with sporadic bursts (‘wild’) controlled by athermal screw dislocation activity, thereby violating the classical notion of temperature-dependent screw dislocation motion at low temperatures. An abrupt increase of the athermal avalanche component is identified at the critical temperature of the material. Our results indicate that plasticity at any scale can be understood in terms of the coexistence of these mild and wild modes of deformation, which could help design better alloys by suppressing one of the two modes in desired temperature windows.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Unusual Li 2 O sublimation promotes single-crystal growth and sintering

Li 2 O is rarely used for cathode material synthesis due to its high melting point (1,438 °C). Here we discover that Li 2 O can sublimate at 800-1,000 °C under ambient pressure, opening new possibilities for cathode synthesis. We propose a mechanism that enables synthesis of single crystals-such as LiNi 0.8 Mn 0.1 Co 0.1 O 2 (NMC811) or LiNi 0.9 Mn 0.05 Co 0.05 O 2 (NMC90)-without direct contact with Li 2 O salts. We show that Li 2 O vapour successfully converts spent polycrystalline NMC811 into segregated single crystals without milling or post-treatment. The Li 2 O vapour, derived from Li 2 O solids, diffuses rapidly and reacts with precursors, mimicking a molten-salt environment, which facilitates single-crystal growth. The chemical lithiation process continuously drives Li 2 O sublimation, sintering the crystals. Single crystals derived from Li 2 O and fresh precursors or spent polycrystals exhibit outstanding cycling after 1,000 cycles in full cells. The demonstrated Li 2 O sublimation and its universal role in promoting single-crystal growth provides an effective approach for single-crystal synthesis, scale-up and recycling.

25 ENERGY STORAGE↗
DOE OSTI2022

Nonequilibrium DMFT + CPA for correlated disordered systems

We present a solution for the nonequilibrium dynamics of an interacting disordered system. The approach adapts the combination of the equilibrium dynamical mean-field theory and the equilibrium coherent potential approximation methods to the nonequilibrium many-body formalism, using the Kadanoff-Baym-Keldysh complex time contour, for the dynamics of interacting disordered systems away from equilibrium. We use our time domain solution to obtain the equilibrium density of states of the disordered interacting system described by the Anderson-Hubbard model, bypassing the necessity for the cumbersome analytical continuation process. We further apply the nonequilibrium solution to the interaction quench problem for an isolated disordered system. Here, the interaction is abruptly changed from zero (noninteracting system) to another constant (finite) value at which it is subsequently kept. We observe via the time dependence of the potential, kinetic, and total energies the effect of disorder on the relaxation of the system as a function of final interaction strength. Finally, the real-time approach has the potential to shed light on the fundamental role of disorder in the nonequilibrium dynamics of interacting quantum systems.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2024

DS-GL: Advancing Graph Learning via Harnessing the Power of Nature within Dynamic Systems

With the rapid digitization of the world, an increasing number of real-world applications are turning to nonEuclidean data, modeled as graphs. Due to their intrinsic high complexity and irregularity, learning from graph data demands tremendous computational power. Recently, CMOS-compatible Ising machines, i.e., dynamic systems composed of CMOS components, have emerged as a new approach that harnesses the inherent power of natural annealing within dynamic systems to efficiently resolve binary optimization problems and have been adopted for traditional graph computation, such as max-cut. However, when performing complex Graph Learning (GL) tasks, Ising machines face significant hurdles: (i) they are inherently binary and thus ill-suited for real-valued problems; (ii) their expensive all-to-all coupling network that guarantees effective natural annealing poses daunting scalability concerns. To address these challenges, this paper proposes a nature-powered graph learning framework dubbed DS-GL, which is the first effort to transform the process of solving graph learning problems into the natural annealing process within a parameterized dynamic system embodied as a CMOS chip. To tackle the two major hurdles, DS-GL first augments the Ising machine architecture to modify the self-reaction term of its Hamiltonian function from linear to quadratic, effectively serving as an energy regulator. This adjustment maintains the system’s original physical interpretation while enabling it to process continuous, real-valued data. Second, to address the scaling issue, DS-GL further upgrades the real-valued dense Ising machine by decomposing it into a mesh-based multi-PE dynamic system that supports efficient distributed spatial-temporal co-annealing across different PEs through sparse interconnects. By exploiting the inherent sparsity and component structures in real-world graphs, DS-GL is able to map complex graph learning tasks onto the scalable dynamic system while maintaining high accuracy. Evaluations with three diverse GL applications across six real-world datasets, including traffic flow and COVID-19 prediction, show that DS-GL can deliver from 102× to 106× speedups and 500× energy reduction over Graph Neural Networks on GPUs, with 5% - 20% accuracy enhancement.

Song, Ruibing↗