SEARCH · Search NASA
Results for “gradient methods”
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.
High-performance modeling acoustic and elastic waves using the parallel Dichotomy Algorithm
A high-performance parallel algorithm is proposed for modeling the propagation of acoustic and elastic waves in inhomogeneous media. An initial boundary-value problem is replaced by a series of boundary-value problems for a constant elliptic operator and different right-hand sides via the integral Laguerre transform. It is proposed to solve difference equations by the conjugate gradient method for acoustic equations and by the GMRES(k) method for modeling elastic waves. A preconditioning operator was the Laplace operator that is inverted using the variable separation method. The novelty of the proposed algorithm is using the Dichotomy Algorithm , which was designed for solving a series of tridiagonal systems of linear equations, in the context of the preconditioning operator inversion. Via considering analytical solutions, it is shown that modeling wave processes for long instants of time requires high-resolution meshes. The proposed parallel fine-mesh algorithm enabled to solve real application seismic problems in acceptable time and with high accuracy. By solving model problems, it is demonstrated that the considered parallel algorithm possesses high performance and efficiency over a wide range of the number of processors (from 2 to 8192).
Augmented Lagrangian formulation of orbital-free density functional theory
We present an Augmented Lagrangian formulation and its real-space implementation for non-periodic Orbital-Free Density Functional Theory (OF-DFT) calculations. In particular, we rewrite the constrained minimization problem of OF-DFT as a sequence of minimization problems without any constraint, thereby making it amenable to powerful unconstrained optimization algorithms. Further, we develop a parallel implementation of this approach for the Thomas–Fermi–von Weizsacker (TFW) kinetic energy functional in the framework of higher-order finite-differences and the conjugate gradient method. With this implementation, we establish that the Augmented Lagrangian approach is highly competitive compared to the penalty and Lagrange multiplier methods. Additionally, we show that higher-order finite-differences represent a computationally efficient discretization for performing OF-DFT simulations. Overall, we demonstrate that the proposed formulation and implementation are both efficient and robust by studying selected examples, including systems consisting of thousands of atoms. We validate the accuracy of the computed energies and forces by comparing them with those obtained by existing plane-wave methods.
Stochasticity and robustness in spiking neural networks
Despite drawing inspiration from biological systems which are inherently noisy and variable, artificial neural networks have been shown to require precise weights to carry out the task which they are trained to accomplish. This creates a challenge when adapting these artificial networks to specialized execution platforms which may encode weights in a manner which restricts their accuracy and/or precision.Reflecting back on the non-idealities which are observed in biological systems, we investigated the effect these properties have on the robustness of spiking neural networks under perturbations to weights. First, we examined techniques extant in conventional neural networks which resemble noisy processes, and postulated they may produce similar beneficial effects in spiking neural networks. Second, we evolved a set of spiking neural networks utilizing biological non-idealities to solve a pole-balancing task, and estimated their robustness. We showed it is higher in networks using noisy neurons, and demonstrated that one of these networks can perform well under the variance expected when a hafnium-oxide based resistive memory is used to encode synaptic weights. Lastly, we trained a series of networks using a surrogate gradient method on the MNIST classification task. We confirmed that these networks demonstrate similar trends in robustness to the evolved networks. We discuss these results and argue that they display empirical evidence supporting the role of noise as a regularizer which can increase network robustness.
A machine learning framework for accurate and robust analysis of radiation detector pulses
The microscopic properties of atomic nuclei are used to study various scientific questions. They are essential for understanding the fundamental forces of nature and the chemical evolution of the universe. Detecting decay radiation from radioactive nuclei makes it possible to probe these fundamental nuclear properties. Detector waveform traces may contain additional information about the radiation. Generally, advanced signal processing techniques are needed to extract this additional information, often involving fitting the waveform with model response functions using non-linear least-squares optimization with second-order gradient methods. While this is a powerful technique, it is also computationally expensive, leading to slow processing time, which scales with the volume of data. To address this problem, we have developed a machine learning (ML) approach that infers the characteristics of traces from a model detector response function. In particular, we are interested in classifying whether a single recorded trace consists of one or two pulse constituents and estimating the pulse parameters. Furthermore, our proposed ML method can precisely extract the pulses’ parameters, such as energy and timing information, and accurately classify the pulse multiplicity of a trace. Unlike non-learning-based approaches, our ML approach uses neural networks that are significantly faster at inference, as they do not require any optimization during this stage.
Perspective on Many-Body Methods for Molecular Polaritonic Systems
Recent advances in strong light–matter interactions have revealed a wealth of new physical phenomena in molecules embedded in optical cavities, including modified chemical reactivity, altered excitation spectra, and novel quantum correlations. To describe these effects from first-principles, the field of ab initio quantum electrodynamics (QED) has emerged as a compelling extension of quantum chemistry that treats electronic and photonic degrees of freedom on equal footing. In this Perspective, we review the growing landscape of many-body QED methods, including Hartree–Fock, density functional theory (QEDFT), time-dependent DFT (QED-TDDFT), configuration interaction (QED-CI), complete active space (QED-CASSCF), coupled cluster (QED-CC), quantum Monte Carlo (QED-QMC), and density matrix renormalization group (QED-DMRG), highlighting recent developments and implementations. We further explore real-time methods, gradient and Hessian formalisms, and the integration of nonadiabatic nuclear dynamics. Applications range from benchmark simulations of polaritonic chemistry to quantum simulations on emerging quantum hardware. We conclude by outlining future directions for theory development and interdisciplinary efforts at the interface of quantum chemistry, condensed matter, and quantum optics.
High-Quality SnSe 2 Single Crystals: Electronic and Thermoelectric Properties
In this work, high-quality SnSe 2 single crystals were successfully synthesized using a temperature gradient method. N-type characteristics and strong anisotropic transport properties of SnSe 2 single crystals were exhibited between the ab plane and the c-axis. At 673 K, the power factor (PF) value along the ab plane is 3.43 μW cm -1 K -2 , while it is 0.92 μW cm -1 K -2 along the c-axis. The ratio between thermal conductivities along the ab plane (κ ab ) and c-axis (κ c ) is on the order of 7.6 at 300 K, while this value is about 5.6 at 673 K. The thermoelectric figure of merit (ZT) in the c-axis (0.15) is higher than that (0.1) along the ab plane, according to the ultralow out-of-plane thermal conductivity. The electronic band structure results, which were examined by angle-resolved photoemission spectroscopy (ARPES) predicted the potential of improving the thermoelectric performance of SnSe 2 single crystals by electron doping.
Learning in continuous action space for developing high dimensional potential energy models
Reinforcement learning (RL) approaches that combine a tree search with deep learning have found remarkable success in searching exorbitantly large, albeit discrete action spaces, as in chess, Shogi and Go. Many real-world materials discovery and design applications, however, involve multi-dimensional search problems and learning domains that have continuous action spaces. Exploring high-dimensional potential energy models of materials is an example. Traditionally, these searches are time consuming (often several years for a single bulk system) and driven by human intuition and/or expertise and more recently by global/local optimization searches that have issues with convergence and/or do not scale well with the search dimensionality. Here, in a departure from discrete action and other gradient-based approaches, we introduce a RL strategy based on decision trees that incorporates modified rewards for improved exploration, efficient sampling during playouts and a “window scaling scheme" for enhanced exploitation, to enable efficient and scalable search for continuous action space problems. Using high-dimensional artificial landscapes and control RL problems, we successfully benchmark our approach against popular global optimization schemes and state of the art policy gradient methods, respectively. We demonstrate its efficacy to parameterize potential models (physics based and high-dimensional neural networks) for 54 different elemental systems across the periodic table as well as alloys. We analyze error trends across different elements in the latent space and trace their origin to elemental structural diversity and the smoothness of the element energy surface. Broadly, our RL strategy will be applicable to many other physical science problems involving search over continuous action spaces.
Pearson's Functions to Describe FSW Weld Geometry
Friction stir welding (FSW) is a relatively new joining technique particularly for aluminium alloys that are difficult to fusion weld. In this study, the geometry of the weld has been investigated and modelled using Pearson's functions. It has been demonstrated that the Pearson's parameters (mean, standard deviation, skewness, kurtosis and geometric constant) can be used to characterize the weld geometry and the tensile strength of the weld assembly. Pearson's parameters and process parameters are strongly correlated allowing to define a control process procedure for FSW assemblies which make radiographic or ultrasonic controls unnecessary. Finally, an optimisation using a Generalized Gradient Method allows to determine the geometry of the weld which maximises the assembly tensile strength.
Deconvolving the temporal response of photoelectric x-ray detectors for the diagnosis of pulsed radiations
Based on the conjugate gradient method, a simple algorithm is presented for deconvolving the temporal response of photoelectric x-ray detectors (XRDs) to reconstruct the resolved time-dependent x-ray fluxes. With this algorithm, we have studied the impact of temporal response of XRD on the radiation diagnosis of hohlraum heated by a short intense laser pulse. It is found that the limiting temporal response of XRD not only postpones the rising edge and peak position of x-ray pulses but also smoothes the possible fluctuations of radiation fluxes. Without a proper consideration of the temporal response of XRD, the measured radiation flux can be largely misinterpreted for radiation pulses of a hohlraum heated by short or shaped laser pulses.
Computational insights of water droplet transport on graphene sheet with chemical density
Surface gradient has been emerging as an intriguing technique for nanoscale particle manipulation and transportation. Owing to its outstanding and stable chemical properties, graphene with covalently bonded chemical groups represents extraordinary potential for the investigation of nanoscale transport in the area of physics and biology. Here, we employ molecular dynamics simulations to investigate the fundamental mechanism of utilizing a chemical density on a graphene sheet to control water droplet motions on it. Simulation results have demonstrated that the binding energy difference among distinct segment of graphene in terms of interaction between the covalently bonded oxygen atoms on graphene and the water molecules provides a fundamental driving force to transport the water droplet across the graphene sheet. Also, the velocity of the water droplet has showed a strong dependence on the relative concentration of oxygen atoms between successive segments. Furthermore, a multi-direction channel provides insights to guide the transportation of objects towards a targeted position, separating the mixtures with a system of specific chemical functionalization. Our findings shed illuminating lights on the surface gradient method and therefore provide a feasible way to control nanoscale motion on the surface and mimic the channelless microfluidics.
Variational approach for static mirror structures
Anisotropic static plasma equilibria where the parallel and perpendicular pressures are only functions of the amplitude of the local magnetic field are shown to be amenable to a variational principle with a free energy density given by the parallel tension. This approach is used to demonstrate that two-dimensional small-amplitude static magnetic holes constructed from a Grad-Shafranov type equation slightly below the (subcritical) mirror instability threshold identify with lump solitons of KPII equation, but turn out to be unstable. Differently, large-amplitude magnetic structures, which are stable as they realize a minimum of the free energy, are computed using a gradient method within two-dimensional numerical simulations where the regularizing effect of finite Larmor radius corrections is retained. Interestingly, these structures transform from stripes to bubbles when the angle of the magnetic field with the coordinate plane is increased.
A generalized Poisson and Poisson-Boltzmann solver for electrostatic environments
The computational study of chemical reactions in complex, wet environments is critical for applications in many fields. It is often essential to study chemical reactions in the presence of applied electrochemical potentials, taking into account the non-trivial electrostatic screening coming from the solvent and the electrolytes. As a consequence, the electrostatic potential has to be found by solving the generalized Poisson and the Poisson-Boltzmann equations for neutral and ionic solutions, respectively. In the present work, solvers for both problems have been developed. A preconditioned conjugate gradient method has been implemented for the solution of the generalized Poisson equation and the linear regime of the Poisson-Boltzmann, allowing to solve iteratively the minimization problem with some ten iterations of the ordinary Poisson equation solver. In addition, a self-consistent procedure enables us to solve the non-linear Poisson-Boltzmann problem. Both solvers exhibit very high accuracy and parallel efficiency and allow for the treatment of periodic, free, and slab boundary conditions. The solver has been integrated into the BigDFT and Quantum-ESPRESSO electronic-structure packages and will be released as an independent program, suitable for integration in other codes.
THE ATACAMA COSMOLOGY TELESCOPE: DATA CHARACTERIZATION AND MAPMAKING
We present a description of the data reduction and mapmaking pipeline used for the 2008 observing season of the Atacama Cosmology Telescope (ACT). The data presented here at 148 GHz represent 12% of the 90 TB collected by ACT from 2007 to 2010. In 2008 we observed for 136 days, producing a total of 1423 hr of data (11 TB for the 148 GHz band only), with a daily average of 10.5 hr of observation. From these, 1085 hr were devoted to an 850 deg{sup 2} stripe (11.2 hr by 9. Degree-Sign 1) centered on a declination of -52. Degree-Sign 7, while 175 hr were devoted to a 280 deg{sup 2} stripe (4.5 hr by 4. Degree-Sign 8) centered at the celestial equator. The remaining 163 hr correspond to calibration runs. We discuss sources of statistical and systematic noise, calibration, telescope pointing, and data selection. For the 148 GHz band, out of 1260 survey hours and 1024 detectors in the array, 816 hr and 593 effective detectors remain after data selection, yielding a 38% survey efficiency. The total sensitivity in 2008, determined from the noise level between 5 Hz and 20 Hz in the time-ordered data stream (TOD), is 32 {mu}K{radical}s in cosmic microwave background units. Atmospheric brightness fluctuations constitute the main contaminant in the data and dominate the detector noise covariance at low frequencies in the TOD. The maps were made by solving the least-squares problem using the Preconditioned Conjugate Gradient method, incorporating the details of the detector and noise correlations. Simulations, as well as cross-correlations with Wilkinson Microwave Anisotropy Probe sky maps on large angular scales, reveal that our maps are unbiased at multipoles l > 300. This paper accompanies the public release of the 148 GHz southern stripe maps from 2008. The techniques described here will be applied to future maps and data releases.
DUST CONTINUUM AND POLARIZATION FROM ENVELOPE TO CORES IN STAR FORMATION: A CASE STUDY IN THE W51 NORTH REGION
We present the first high-angular resolution (up to 0.''7, {approx}5000 AU) polarization and thermal dust continuum images toward the massive star-forming region W51 North. The observations were carried out with the Submillimeter Array (SMA) in both the subcompact (SMA-SubC) and extended (SMA-Ext) configurations at a wavelength of 870 {mu}m. W51 North is resolved into four cores (SMA1 to SMA4) in the 870 {mu}m continuum image. The associated dust polarization exhibits more complex structures than seen at lower angular resolution. We analyze the inferred morphologies of the plane-of-sky magnetic field (B ) in the SMA1 to SMA4 cores and in the envelope using the SMA-Ext and SMA-SubC data. These results are compared with the B archive images obtained from the Caltech Submillimeter Observatory (CSO) and James Clerk Maxwell Telescope (JCMT). The polarization percentage is about 1% to 4%, and it is found to decrease with higher intensity in our SMA images, which is a similar trend to that previously reported in the CSO and JCMT data. A correlation between dust intensity gradient position angles ({phi}{sub {nabla}I}) and magnetic field position angles ({phi}{sub B}) is found in the CSO, JCMT, and both SMA data sets. This correlation is further analyzed quantitatively. A systematically tighter correlation between {phi}{sub {nabla}I} and {phi}{sub B} is found in the cores, whereas the correlation decreases in outside-core regions. Magnetic field-to-gravity force ratio ({Sigma}{sub B}) maps are derived using the newly developed polarization-intensity gradient method by Koch et al. We find that the force ratios tend to be small ({Sigma}{sub B} {approx}< 0.5) in the cores in all four data sets. In regions outside of the cores, the ratios increase or the field is even dominating gravity ({Sigma}{sub B} > 1). This possibly provides a physical explanation of the tightening correlation between {phi}{sub {nabla}I} and {phi}{sub B} in the cores: The more the B field lines are dragged and aligned by gravity, the tighter the correlation is. Finally, we propose a schematic scenario for the magnetic field in W51 North to interpret the four polarization observations at different physical scales.
THE IMPORTANCE OF THE MAGNETIC FIELD FROM AN SMA-CSO-COMBINED SAMPLE OF STAR-FORMING REGIONS
Submillimeter dust polarization measurements of a sample of 50 star-forming regions, observed with the Submillimeter Array (SMA) and the Caltech Submillimeter Observatory (CSO) covering parsec-scale clouds to milliparsec-scale cores, are analyzed in order to quantify the magnetic field importance. The magnetic field misalignment δ—the local angle between magnetic field and dust emission gradient—is found to be a prime observable, revealing distinct distributions for sources where the magnetic field is preferentially aligned with or perpendicular to the source minor axis. Source-averaged misalignment angles (|δ|) fall into systematically different ranges, reflecting the different source-magnetic field configurations. Possible bimodal (|δ|) distributions are found for the separate SMA and CSO samples. Combining both samples broadens the distribution with a wide maximum peak at small (|δ|) values. Assuming the 50 sources to be representative, the prevailing source-magnetic field configuration is one that statistically prefers small magnetic field misalignments |δ|. When interpreting |δ| together with a magnetohydrodynamics force equation, as developed in the framework of the polarization-intensity gradient method, a sample-based log-linear scaling fits the magnetic field tension-to-gravity force ratio (Σ {sub B}) versus (|δ|) with (Σ {sub B}) = 0.116 · exp (0.047 · (|δ|)) ± 0.20 (mean error), providing a way to estimate the relative importance of the magnetic field, only based on measurable field misalignments |δ|. The force ratio Σ {sub B} discriminates systems that are collapsible on average ((Σ {sub B}) < 1) from other molecular clouds where the magnetic field still provides enough resistance against gravitational collapse ((Σ {sub B}) > 1). The sample-wide trend shows a transition around (|δ|) ≈ 45°. Defining an effective gravitational force ∼1 – (Σ {sub B}), the average magnetic-field-reduced star formation efficiency is at least a factor of two smaller than the free-fall efficiency. For about one fourth of the sources the average efficiency drops to zero. The force ratio Σ {sub B} can further be linked to the normalized mass-to-flux ratio, yielding an estimate for the latter one without the need of field strength measurements. Across the sample, a transition from magnetically supercritical to subcritcal is observed with growing misalignment (|δ|)
Light in the dark forest. Part I. An efficient optimal estimator for 3D Lyman-alpha forest power spectrum
The highly anisotropic nature of the Lyman-alpha (Lyα) forest data introduces a complex survey window function that complicates the measurement of the three-dimensional power spectrum ( P 3D ). In this paper, we present the first fully optimal estimator for P 3D , which exactly deconvolves the survey window function and marginalizes contaminated modes that distort the power spectrum. Our approach adapts optimal estimator techniques developed for the 2D cosmic microwave background data to the 3D case. To achieve computational feasibility, we employ the conjugate gradient method and implement the P 3 M formalism to handle large-scale and small-scale operations separately and efficiently. We validate our estimator using Monte Carlo mocks and Gaussian simulations, demonstrating its accuracy and computational efficiency. We confirm that mode marginalization eliminates distortions arising from quasar continuum errors and delivers robust power spectrum estimation, though it also inflates errors at large scales. This first implementation works in the flat-sky case; we discuss the remaining steps needed to generalize it to the curved-sky case. This formalism offers a foundation for the Lyα forest P 3D measurements and a new path toward cosmological constraints from the Lyα forest data.