Search NASA⌕ Search

SEARCH · Search NASA

Results for “Matrix diffusion”

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 1,009 records · Page 56

Post-Deployment Characterization of Glass Fiber-Reinforced Thermoset and Thermoplastic Composite Tidal Turbine Blades

In 2021, the National Renewable Energy Laboratory (NREL) supported Verdant Power with the most successful tidal energy deployment in U.S. history. Three of their Gen5d 5 m turbines were deployed as part of the Roosevelt Island Tidal Energy project. Initially, the three rotors initially deployed were manufactured from glass fiber-reinforced epoxy composites. Midway through the deployment, one rotor was replaced with one manufactured at NREL. The new rotor utilized a novel infusible thermoplastic resin system. Since the deployment, one epoxy rotor and one thermoplastic rotor were returned to NREL for continued materials and manufacturing research. The two rotors underwent full-scale structural testing before being sectioned and cut into specimens for a variety of manufacturing quality tests, thermomechanical characterization, and evaluation of material performance in marine environments to understand the key differences between the fiberglass-reinforced epoxy and Elium composites used for the respective rotors. Matrix burn-off tests showed that the Elium blades had a considerably higher fiber volume fraction compared to the epoxy blades (61% vs. 49%). Environmental aging of the specimens showed that the epoxy laminates absorbed more water over the conditioning period; however, it was determined that the Elium laminates had higher diffusion coefficients, so they initially absorbed water faster. Finally, one full epoxy blade and one full Elium blade were conditioned at ambient temperatures for up to 11 months, while periodic mass measurements were taken. The datasets were extrapolated to assume a full 20-year operational life span, and it was determined that the blades would not reach full saturation during that time span.

composite manufacturing↗

Matrix operator theory of radiative transfer. II - Scattering from maritime haze.

Matrix operator theory is used to calculate the reflected and transmitted radiance of photons that have interacted with plane-parallel maritime haze layers. The results are presented for three solar zenith angles, three values of the surface albedo, and a range of optical thicknesses from very thin to very thick. The diffuse flux at the lower boundary and the cloud albedo are tabulated. The forward peak and other features in the single-scattered phase function cause the radiance in many cases to be very different from that for Rayleigh scattering. In particular, the variation of the radiance with both the zenith or nadir angle and the azimuthal angle is more marked and the relative limb darkening under very thick layers is greater for haze M than for Rayleigh scattering. The downward diffuse flux at the lower boundary for A = 0 is always greater and the cloud albedo is always less for haze M than for Rayleigh layers.

Kattawar, G. W.↗

Application of program LAURA to three-dimensional AOTV flowfields

Program LAURA (Langley Aerothermodynamic Upwind Relaxation Algorithm) is a robust, finite volume, single-level storage, implicit upwind differencing algorithm which has been documented and tested on several three-dimensional blunt-body flows. The algorithm can run at unlimited Courant numbers (relaxing the steady-state equations) but requires the inversion of only a 5 x 5 matrix per computational cell. An alternating directional sweep Gauss-Seidel substitution strategy is used to relax the governing equations. At present, the Euler and thin-layer Navier-Stokes equations using Sutherland's law for viscosity have been modeled for a perfect gas, equilibrium air, and nonequilibrium air chemistry neglecting diffusion. The equilibrium and nonequilibrium air chemistry options have been described in a companion paper. Good comparisons with experimental data and another calculation method for pressure distributions, aerodynamic coefficients, and heat-transfer distributions have been demonstrated for three-dimensional blunt-body flows.

Gnoffo, P. A.↗

Overview of Base Model in Parametric Studies Specific to Performance of U-Mo Plates

This paper provides an overview of the base model specifically developed to perform parametric sensitivity studies on the U-10Mo monolithic fuel system. U-Mo monolithic fuels are being considered for the conversion of test reactors into high-performance research reactors that operate using proliferation-resistant, low-enriched uranium (LEU) fuels. These plate-type fuels contain a high-density, low-enrichment fuel sandwiched between zirconium diffusion barriers and encapsulated in aluminum claddings. All U.S. high-performance research reactors have released the designs of their LEU monolithic fuel reactor cores. These designs include nearly 50 distinct fuel plate geometries with different operational parameters. Consequently, a single generic plate geometry representing all the extreme points in this design matrix is unrealistic. To evaluate the performance for various parameters, a set of sensitivity studies was performed. These studies considered various input parameters (i.e., geometric, operational, and material property-related). The results revealed valuable information about plate performance and the sensitivity of this performance to various modeling inputs. To establish a reference state for comparing these result, base model featuring representative irradiation conditions was developed. To capture in-reactor behavior accurately, incorporation of representative constitutive models capable of evolving properties with respect to temperature, irradiation time, and burnup was needed. The behavioral models considered burnup-dependent properties, swelling, creep, and degradation. This paper introduces the base model created for the parametric sensitivity studies. The detailed description of the procedure includes the model geometry, model discretization, thermo-mechanical coupling, material properties and behavioral models. This paper also provides selected results and assesses the performance of the base model.

42 ENGINEERING↗

Post-Deployment Characterization of Glass Fiber-Reinforced Thermoset and Thermoplastic Composite Tidal Turbine Blades

In 2021, the National Renewable Energy Laboratory (NREL) supported Verdant Power with the most successful tidal energy deployment in U.S. history. Three of their Gen5d 5 m turbines were deployed as part of the Roosevelt Island Tidal Energy project. Initially, the three deployed rotors were manufactured from glass fiber-reinforced epoxy composites. Midway through the deployment, one rotor was replaced with one manufactured at NREL. The new rotor utilized a novel infusible thermoplastic resin system (Elium from Arkema). Since the deployment, one epoxy rotor and one thermoplastic rotor were returned to NREL for continued materials and manufacturing research. The two rotors underwent full-scale structural testing before being sectioned and cut into specimens for a variety of manufacturing quality tests, thermomechanical characterization, and evaluation of material performance in marine environments to understand the key differences between the fiberglass-reinforced epoxy and Elium composites used for the respective rotors. Matrix burn-off tests showed that the Elium blades had a considerably higher fiber volume fraction compared to the epoxy blades (61% vs. 49%). Environmental aging of the specimens showed that the epoxy laminates absorbed more water over the conditioning period; however, it was determined that the Elium laminates had higher diffusion coefficients, so they initially absorbed water faster. Finally, one full epoxy blade and one full Elium blade were conditioned at ambient temperatures for up to 11 months, while periodic mass measurements were taken. The datasets were extrapolated to assume a full 20-year operational life span, and it was determined that the blades would not reach full saturation during that time span.

composite manufacturing↗

Irradiation Effects on Stability of δ-UZr2 phase in U-50 wt% Zr Alloy

U-50wt%Zr is a candidate metallic nuclear fuel with potential application in light water reactors due to its excellent thermal properties and high radiation tolerance. The Zr-rich UZr fuels possess greater swelling resistance and fission gas release characteristics compared with U-rich UZr fuels. In this current study, the δ-phase U-50wt%Zr is proton irradiated at room temperature to 1 displacement per atom (dpa) to provide insights on phase stability under irradiation conditions. High resolution characterization of Transmission Electron Microscopy (TEM) and Atom Probe Tomography (APT) characterization techniques are used to elucidate microstructural changes due to irradiation. TEM and APT results show highly oriented bcc β-Zr-rich platelet precipitates nucleating adjacent to α-U phases inside the UZr 2 matrix. Formation of this platelet morphology is characteristic of Widmänstatten structure which can be attributed to a variety of factors such as differences in thermal expansion coefficient between the phases, grain size, alloy composition, and cooling rate. The phases present are distinctly different than those observed through in situ annealing, but irradiation accelerates diffusion and phase separation kinetics. These microstructural changes in U-50wt%Zr are different from those achieved by pure thermodynamic or high temperature heavy ion irradiation experiments. Finally, our work, together with previous ones, highlight the necessity to study the U-Zr phase diagram under non-equilibrium thermodynamics conditions to support this material's deployment as a viable nuclear fuel form.

36 MATERIALS SCIENCE↗

In Situ Oxidation Studies of High-Entropy Alloy Nanoparticles

Although high-entropy alloys (HEAs) have shown tremendous potential for elevated temperature, anticorrosion, and catalysis applications, little is known on how HEA materials behave under complex service environments. Herein, we studied the high-temperature oxidation behavior of Fe 0.28 Co 0.21 Ni 0.20 Cu 0.08 Pt 0.23 HEA nanoparticles (NPs) in an atmospheric pressure dry air environment by in situ gas-cell transmission electron microscopy. It is found that the oxidation of HEA NPs is governed by Kirkendall effects with logarithmic oxidation rates rather than parabolic as predicted by Wagner's theory. Further, the HEA NPs are found to oxidize at a significantly slower rate compared to monometallic NPs. The outward diffusion of transition metals and formation of disordered oxide layer are observed in real time and confirmed through analytical energy dispersive spectroscopy, and electron energy loss spectroscopy characterizations. Localized ordered lattices are identified in the oxide, suggesting the formation of Fe 2 O 3 , CoO, NiO, and CuO crystallites in an overall disordered matrix. Hybrid Monte Carlo and molecular dynamics simulations based on first-principles energies and forces support these findings and show that the oxidation drives surface segregation of Fe, Co, Ni, and Cu, while Pt stays in the core region. The present work offers key insights into how HEA NPs behave under high-temperature oxidizing environment and sheds light on future design of highly stable alloys under complex service conditions.

36 MATERIALS SCIENCE↗

Structure and magnetism in strained Ge{sub 1-x-y}Sn{sub x}Mn{sub y} films grown on Ge(001) by low temperature molecular beam epitaxy

In this letter, we study the structural and magnetic properties of Ge{sub 1-x-y}Sn{sub x}Mn{sub y} films grown on Ge(001) by low temperature molecular beam epitaxy using X-ray diffraction, high resolution transmission electron microscopy, and superconducting quantum interference device. Like in Mn doped Ge films, Mn atoms diffuse during the growth and aggregate into vertically aligned Mn-rich nanocolumns of a few nanometers in diameter. Transmission electron microscopy observations in plane view clearly indicate that the Sn incorporation is not uniform with concentration in Mn rich vertical nanocolumns lower than the detection limit of electron energy loss spectroscopy. The matrix exhibits a GeSn solid solution while there is a Sn-rich GeSn shell around GeMn nanocolumns. The magnetization in Ge{sub 1-x-y}Sn{sub x}Mn{sub y} layers is higher than in Ge{sub 1-x}Mn{sub x} films. This magnetic moment enhancement in Ge{sub 1-x-y}Sn{sub x}Mn{sub y} is probably related to the modification of the electronic structure of Mn atoms in the nanocolumns by the Sn-rich shell, which is formed around the nanocolumns.

36 MATERIALS SCIENCE↗

Niobium hydride formation in superconducting qubit thin films

The formation of nonsuperconducting hydrides in 160–170-nm-thick films of niobium is examined. We identify six elastically distinct orientation relationships ɛ−Nb 4 ⁢H 3 takes within the matrix, solid-solution 𝛼−NbH 𝑥 phase. We employ a phase field model to assess the impact of elastic energy induced by the strain of phase transformation on the morphology and transformation dynamics of ɛ precipitates within a thin film. We consider the dimensions of the thin film, crystallographic growth direction, and diffusion rates to predict the timescale of hydride evolution. Leveraging the finite element method, we predict two-dimensional and three-dimensional equilibrium shapes of ɛ-hydrides within a bulk sample and in a thin film that has a traction-free surface. Our results suggest that niobium hydrides migrate to the free surface of the film. Precipitates which reach the free surface coarsen, while precipitates within the film dissolve. Precipitates in both two dimensions and three dimensions experience a repulsive interaction force at the free surface, that is attractive in the bulk, shown in experiment and theory of previous studies.

36 - MATERIALS SCIENCE↗

Scientific Basis For Ultra-High Burn-Up Nuclear Fuels

A key issue on the path to nuclear energy becoming an essential component of the U.S. clean energy strategy is complete burn of the nuclear fuel. We propose research that will enable high-burn-up fuels by establishing the basic science for development and qualification of advanced nuclear fuel that will couple modern computational materials modeling, fabrication, and characterization capabilities with targeted performance-testing experiments using ion-beam facilities. This work will establish the scientific foundation and guide selection of the optimum fuel type for advanced reactor concepts. We expect to experimentally quantify the phase stability and kinetics of phase transformations, inter-diffusion, microstructural evolution, micromechanical properties, and the influence of severe radiation environments on fuel performance and, by the end of the project, develop a validated model for advanced nuclear energy materials under extreme conditions of radiation, temperature, and evolving chemistry. We will provide a science-based path forward to an optimized inert matrix fuel, while contributing to the development of a validated nuclear fuel database. We expect this project will further establish LLNL’s credibility in the nuclear energy community.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

A 2D/1D coupling neutron transport method based on the matrix MOC and NEM methods

A new 2D/1D coupling method based on the matrix MOC method (MMOC) and nodal expansion method (NEM) is proposed for solving the three-dimensional heterogeneous neutron transport problem. The MMOC method, used for radial two-dimensional calculation, constructs a response matrix between source and flux with only one sweep and then solves the linear system by using the restarted GMRES algorithm instead of the traditional trajectory sweeping process during within-group iteration for angular flux update. Long characteristics are generated by using the customization of commercial software AutoCAD. A one-dimensional diffusion calculation is carried out in the axial direction by employing the NEM method. The 2D and ID solutions are coupled through the transverse leakage items. The 3D CMFD method is used to ensure the global neutron balance and adjust the different convergence properties of the radial and axial solvers. A computational code is developed based on these theories. Two benchmarks are calculated to verify the coupling method and the code. It is observed that the corresponding numerical results agree well with references, which indicates that the new method is capable of solving the 3D heterogeneous neutron transport problem directly. (authors)

97 MATHEMATICS AND COMPUTING↗

Hierarchical morphologies in co-sputter deposited thin films

Co-depositions of immiscible alloy films at specific processing conditions have yielded hierarchical microstructures which consist of distinct features at multiple length scales, often agglomerates and concentration modulations on the sub-micrometer-scale and fine nanoprecipitates in a matrix on the nanoscale. The present work examined a series of immiscible alloy systems: Cu-Mo, Cu-Ag, Cu-Fe, Cu-Ta, Mo-Ag, Cu-Mo-Ag, to determine the kinetic conditions favorable for hierarchical organization and the formation mechanism of such structures. Thin films of six immiscible systems were sputter co-deposited over a range of deposition rates from 0.12 to 2 nm/s and various temperatures from 400 to 800°C. The resulting microstructures indicate that hierarchical structures form with sufficient disparity in kinetic energy between the constituent atoms, one species being highly mobile (A) and the other relatively immobile (B). This condition arises typically at elevated deposition temperatures and reduced deposition rates but is also alloy dependent. The hierarchical structures form during deposition via phase separation and self-organization processes across the multiple length scales. The adatoms diffuse on the film surface with the highly mobile species swiftly agglomerating into A-rich domains within which B-rich nanoprecipitates form, often self-organizing into periodic arrays. The smallest B-rich nanoprecipitates in the A-rich domains are found to be coherent and in a metastable crystal structure (B taking the structure of A), but coarser precipitates that exhibit the equilibrium structure of B element. The A-rich domains are surrounded by a B-rich matrix that phase-separates into a concentration modulated structure. In conclusion, the observations are interpreted via a model incorporating material properties and process parameters.

36 MATERIALS SCIENCE↗

COMBINE7.1 - A Portable ENDF/B-VII.0 Based Neutron Spectrum and Cross-Section Generation Program

COMBINE7.1 is a FORTRAN 90 computer code that generates multigroup neutron constants for use in the deterministic diffusion and transport theory neutronics analysis. The cross-section database used by COMBINE7.1 is derived from the Evaluated Nuclear Data Files (ENDF/B-VII.0). The neutron energy range covered is from 20 MeV to 1.0E-5 eV. The Los Alamos National Laboratory NJOY code is used as the processing code to generate a 167 fine-group cross-section library in MATXS format for Bondarenko self-shielding treatment. Resolved resonance parameters are extracted from ENDF/B-VII.0 File 2 for a separate library to be used in an alternate Nordheim self-shielding treatment in the resolved resonance energy range. The equations solved for energy dependent neutron spectrum in the 167 fine-group structure are the B3 or B1 zero-dimensional approximations to the transport equation. The fine group cross sections needed for the spectrum calculation are first prepared by Bondarenko self-shielding interpolation in terms of background cross section and temperature. The geometric lump effect, when present, is accounted for by augmenting the background cross section. Nordheim self-shielded fine group cross sections for a material having resolved resonance parameters overwrite correspondingly the existing self-shielded fine group cross sections when this option is used. COMBINE7.1 coalesces fine group cross sections into broad group macroscopic and microscopic constants. The coalescing is performed by utilizing fine-group fluxes and/or currents obtained by spectrum calculation as the weighting functions. The multigroup constants may be output in any of several standard formats including INL format, ANISN 14** free format, CCCC ISOTXS format, and AMPX working library format. ANISN-PC, a one-dimensional (1-D) discrete-ordinate transport code, is incorporated into COMBINE7.1. As an option, the 167 fine-group constants generated by zero-dimensional COMBINE portion in the program can be used to calculate regionwise spectra in the 1-D ANISN portion, all internally to reflect the 1-D transport correction. The regionwise spectra are then used to generate mutigroup regionwise neutron constants. The 1-D neutron transport can be performed up to three stages, e.g., from a TRISO fuel to PEBBLE to 1-D full core wedge. In addition, COMBINE7.1 has now the capability of adjoint flux calculation through the 1-D ANISN transport. Photon transport capability is also added. For this, a photon production and photo-atomic cross section library, MATNG.LIB, was generated in MATXS format through NJOY code. The photon production cross section matrix is of 167 neutron - 18 photon groups. Photo-atomic cross sections, including heating, are in 18 energy groups.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Sulfur vacancies in photorefractive Sn{sub 2}P{sub 2}S{sub 6} crystals

A photoinduced electron paramagnetic resonance (EPR) spectrum in single crystals of Sn{sub 2}P{sub 2}S{sub 6} (SPS) is assigned to an electron trapped at a sulfur vacancy. These vacancies are unintentionally present in undoped SPS crystals and are expected to play an important role in the photorefractive behavior of the material. Nonparamagnetic sulfur vacancies are formed during the initial growth of the crystal. Subsequent illumination below 100 K with 442 nm laser light easily converts these vacancies to EPR-active defects. The resulting S = 1/2 spectrum shows well-resolved and nearly isotropic hyperfine interactions with two P ions and two Sn ions. Partially resolved interactions with four additional neighboring Sn ions are also observed. Principal values of the g matrix are 1.9700, 1.8946, and 1.9006, with the corresponding principal axes along the a, b, and c directions in the crystal. The isotropic parts of the two primary {sup 31}P hyperfine interactions are 19.5 and 32.6 MHz and the isotropic parts of the two primary Sn hyperfine interactions are 860 and 1320 MHz (the latter values are each an average for {sup 117}Sn and {sup 119}Sn). These hyperfine results suggest that singly ionized sulfur vacancies have a diffuse wave function in SPS crystals, and thus are shallow donors. Before illumination, sulfur vacancies are in the doubly ionized charge state because of compensation by unidentified acceptors. They then trap an electron during illumination. The EPR spectrum from the sulfur vacancy is destroyed when a crystal is heated above 120 K in the dark and reappears when the crystal is illuminated again at low temperature.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Oxidation and Ignition Behaviors of Molten Mg-Nd-Zr Alloy in Resin-Sand Mold

High-temperature oxidation behavior of molten Mg-Nd-Zr in the pyrolysis atmosphere of resin-bind sand mold was investigated. The results showed that pouring temperature higher than 1023 K would sharply increase the ignition risk of melt in mold. The oxide formed on Mg-Nd-Zr melt below 973 K consists of two layers: mainly MgO for outer layer; mixture of MgO and Nd{sub 2}O{sub 3} for inner layer. The inner oxide mixture layer played an essential role in the protection of Mg-Nd-Zr matrix. Based on the experimental results and thermodynamic analysis, the oxidation process containing three stages was proposed. Nd- and Zn-enriched zone was found at the subsurface of oxide scale, without the enrichment of Zr. DICTRA calculation results indicated that this phenomenon was caused by the depletion of Mg and up-hill diffusion of Zr. Thus, Zr showed limited protection to the melt in long-term oxidation.

36 MATERIALS SCIENCE↗

Report on the Implementation of Homogeneous Nucleation Scheme in MARMOT-based Phase Field Simulation

In this report, we summarized our effort in developing mesoscale phase field models for predicting precipitation kinetics in alloys during thermal aging and/or under irradiation in nuclear reactors. The first part focused on developing a method to predict the thermodynamic properties of critical nuclei such as the sizes and concentration profiles of critical nuclei, and nucleation barrier. These properties are crucial for quantitative simulations of precipitate evolution kinetics with phase field models. Fe-Cr alloy was chosen as a model alloy because it has valid thermodynamic and kinetic data as well as it is an important structural material in nuclear reactors. A constrained shrinking dimer dynamics (CSDD) method was developed to search for the energy minimum path during nucleation. With the method we are able to predict the concentration profiles of the critical nuclei of Cr-rich precipitates and nucleation energy barriers. Simulations showed that Cr concentration distribution in the critical nucleus strongly depends on the overall Cr concentration as well as temperature. The Cr concentration inside the critical nucleus is much smaller than the equilibrium concentration calculated by the equilibrium phase diagram. This implies that a non-classical nucleation theory should be used to deal with the nucleation of Cr precipitates in Fe-Cr alloys. The growth kinetics of both classical and non-classical nuclei was investigated by the phase field approach. A number of interesting phenomena were observed from the simulations: 1) a critical classical nucleus first shrinks toward its non-classical nucleus and then grows; 2) a non-classical nucleus has much slower growth kinetics at its earlier growth stage compared to the diffusion-controlled growth kinetics. 3) a critical classical nucleus grows faster at the earlier growth stage than the non-classical nucleus. All of these results demonstrated that it is critical to introduce the correct critical nuclei into phase field modeling in order to correctly capture the kinetics of precipitation. In most alloys the matrix phase and precipitate phase have different concentrations as well as different crystal structures. For example, Cu precipitates in FeCu alloys have fcc crystal structure while the matrix Fe-Cu solid solution has bcc structure at low temperature. The WBM model and KimS model, where both concentrations and order parameters are chosen to describe the microstructures, are commonly used to model precipitations in such alloys. The WBM and KimS models have not been implemented into Marmot yet. In the second part of this report, we focused on implementing the WBM and KimS models into Marmot. The Fe-Cu alloys, which are important structure materials in nuclear reactors, was taken as the model alloys to test the models.

Fe Alloys↗

Quantitative Relationships between Photosynthetic, Nitrogen Fixing, and Fermentative H2 Metabolism in a Photosynthetic Microbial Mat

The ultimate potential of any microbial ecosystem to contribute chemically to its environment - and therefore, to impact planetary biogeochemistry or to generate recognizable biosignatures - depends not only on the individual metabolic capabilities of constituent organisms, but also on how those capabilities are expressed through interactions with neighboring organisms. This is particularly important for microbial mats, which compress an extremely broad range of metabolic potential into a small and dynamic system. H2 participates in many of these metabolic processes, including the major elemental cycling processes of photosynthesis, nitrogen fixation, sulfate reduction, and fermentation, and may therefore serve as a mediator of microbial interactions within the mat system. Collectively, the requirements of energy, electron transfer, and biomass element stoichiometry suggest quantitative relationships among the major element cycling processes, as regards H2 metabolism We determined experimentally the major contributions to 32 cycling in hypersaline microbial mats from Baja California, Mexico, and compared them to predicted relationships. Fermentation under dark, anoxic conditions is quantitatively the most important mechanism of H2 production, consistent with expectations for non-heterocystous mats such as those under study. Up to 16% of reducing equivalents fixed by photosynthesis during the day may be released by this mechanism. The direct contribution of nitrogen fixation to H2 production is small in comparison, but this process may indirectly stimulate substantial H2 generation, by requiring higher rates of fermentation. Sulfate reduction, aerobic consumption, diffusive and ebulitive loss, and possibly H2-based photoreduction of CO2 serve as the principal H2 sinks. Collectively, these processes interact to create an orders-of-magnitude daily variation in H2 concentrations and fluxes, and thereby in the oxidation-reduction potential that is imposed on microbial processes occuring within the mat matrix.

Hoehler, Tori M.↗

Fast alogorithms for Bayesian uncertainty quantification in large-scale linear inverse problems based on low-rank partial Hessian approximations

We consider the problem of estimating the uncertainty in large-scale linear statistical inverse problems with high-dimensional parameter spaces within the framework of Bayesian inference. When the noise and prior probability densities are Gaussian, the solution to the inverse problem is also Gaussian, and is thus characterized by the mean and covariance matrix of the posterior probability density. Unfortunately, explicitly computing the posterior covariance matrix requires as many forward solutions as there are parameters, and is thus prohibitive when the forward problem is expensive and the parameter dimension is large. However, for many ill-posed inverse problems, the Hessian matrix of the data misfit term has a spectrum that collapses rapidly to zero. We present a fast method for computation of an approximation to the posterior covariance that exploits the lowrank structure of the preconditioned (by the prior covariance) Hessian of the data misfit. Analysis of an infinite-dimensional model convection-diffusion problem, and numerical experiments on large-scale 3D convection-diffusion inverse problems with up to 1.5 million parameters, demonstrate that the number of forward PDE solves required for an accurate low-rank approximation is independent of the problem dimension. This permits scalable estimation of the uncertainty in large-scale ill-posed linear inverse problems at a small multiple (independent of the problem dimension) of the cost of solving the forward problem.

97 MATHEMATICS AND COMPUTING↗