SEARCH · Engineering Papers
Results for “function approximation”
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.
Toward Chemical Accuracy for Chemi- and Physisorption with an Efficient Density Functional
Understanding molecular adsorption on surfaces underpins many problems in chemistry and materials science. Accurately and efficiently describing the adsorption has been a challenging task for first-principles methods as the process can involve both short-range chemical bond formations and long-range physical interactions, e.g., van der Waals (vdW) interaction. Density functional theory presents an appealing choice for modeling adsorption reactions, although calculations with many exchange-correlation density functional approximations struggle to accurately describe both chemical and physical molecular adsorptions. Here, we propose an efficient density functional approximation that is accurate for both chemical and physical adsorption by concurrently optimizing its semilocal component and the long-range vdW correction against the prototypical adsorption CO/Pt(111) and Ar 2 binding energy curve. The resulting function opens the door to accurate and efficient modeling of general molecular adsorption.
Dynamic Modeling, Parameter Estimation, and Data Reconciliation of a Supercritical Pulverized Coal-Fired Boiler
A high-fidelity, spatially distributed dynamic model of a supercritical pulverized coal-fired boiler was developed for calculating transient thermal profiles of the boiler tubes, which are often not measured or unmeasurable in industrial settings due to the harsh operating conditions. In this work, rigorous models of the subcritical and supercritical steam properties, which can be highly nonlinear, were used to capture the transition between two-phase and one-phase flow, respectively, as supercritical boilers frequently transition through the critical point during load-following operation. The transient boiler model was also used for data reconciliation and parameter estimation for the purpose of model validation against industrial data. An approach was developed to reduce the size of large-scale reconciliation problem by focusing on the implementation of bias terms for the reconciled variables, which can be estimated in place of the reconciled variables themselves. In addition, a functional approximation of the reconciliation problem was defined so that it can be more simply optimized. Combining this functional approximation optimization with reformulation of the reconciliation problem considering measurement biases can significantly improve the tractability of many such large-scale problems. The validated boiler model was then used to study transient responses under load-following operation with specific attention on the calculation of unmeasurable variables.
Comparative Density Functional Theory Study of Magnetic Exchange Couplings in Dinuclear Transition-Metal Complexes
Multicenter transition-metal complexes (MCTMs) with magnetically interacting ions have been proposed as components for information-processing devices and storage units. For any practical application of MCTMs as magnetic units, it is crucial to characterize their magnetic behavior, and in particular, the isotropic magnetic exchange coupling, J, between its magnetic centers. Due to the large size of typical MCTMs, density functional theory is the only practical electronic structure method for evaluating the J coupling. Here, we assess the accuracy of different density functional approximations for predicting the magnetic couplings of eight dinuclear transition-metal complexes, including five dimanganese, two dicopper, and one divanadium with known reliable experimental J couplings spanning from ferromagnetic to strong antiferromagnetic. The density functionals considered include global hybrid functionals which mix semilocal density functional approximations and exact exchange with a fixed admixing parameter, six local hybrid functionals where the admixing parameters are extended to be spatially dependent, the SCAN and r 2 SCAN meta-generalized gradient approximations (GGAs), and two widely used GGAs. We found that global hybrids tested in this work have a tendency to over-correct the error in magnetic coupling parameters from the Perdew–Burke–Ernzerhof (PBE) GGA as seen for manganese complexes. The performance of local hybrid density functionals shows no improvement in terms of bias and is scattered without a clear trend, suggesting that more efforts are needed for the extension from global to local hybrid density functionals for this particular property. The SCAN and r 2 SCAN meta-GGAs are found to perform as well as benchmark global hybrids on most tested complexes. We further analyze the charge density redistribution of meta-GGAs as well as global and local hybrid density functionals with respect to that of PBE, in connection to the self-interaction error or delocalization error.
Density Functionals for Hydrogen Storage: Defining the H2Bind275 Test Set with Ab Initio Benchmarks and Assessment of 55 Functionals
Efficient and high-capacity storage materials are indispensable for a hydrogen-based economy. In silico tools can accelerate the process of discovery of new adsorbent materials with optimal hydrogen adsorption enthalpies. Density functional theory is well-poised to become a very useful tool for enabling high-throughput screening of potential materials. In this work, we have identified density functional approximations that provide good performance for hydrogen binding applications following a two-pronged approach. First, we have compiled a data set (H2Bind275) that comprehensively represents the hydrogen binding problem capturing the chemical and mechanistic diversity in the binding sites encountered in hydrogen storage materials. We have also computed reference interaction energies for this data set using coupled-cluster theory. Second, we have assessed the performance of 55 density functional approximations for predicting H 2 interaction energies and have identified two hybrid density functionals (ωB97X-V and ωB97M-V), two double hybrid density functionals (DSD-PBEPBE-D3(BJ) and PBE0-DH), and one semilocal density functional (B97M-V) as the best performing ones. We have recommended the addition of empirical dispersion corrections to systematically underbinding density functionals such as revPBE, BLYP, and B3LYP for improvements in performance at negligible additional cost. We have also recommended the usage of the def2-TZVPP basis set as it represents a good compromise between accuracy and cost, limiting the finite basis set errors to less than 1 kJ/mol.
Orbital dependent complications for close vs well-separated electrons in diradicals
We investigate two limits in open-shell diradical systems: O3, in which the interesting orbitals are in close proximity to one another, and (C21H13)2, where there is a significant spatial separation between the two orbitals. In accord with earlier calculations, we find that standard density-functional approximations do not predict the open-shell character for the former case but uniformly predict the open-shell character for the latter case. We trace the qualitatively incorrect behavior in O3 predicted by these standard density functional approximations to self-interaction error and use the Fermi–Löwdin-orbital-self-interaction-corrected formalism to determine accurate triplet, closed-shell singlet, and open-shell broken-spin-symmetry electronic configurations. Analysis of the resulting many-electron overlap matrices allows us to unambiguously show that the broken-spin-symmetry configurations do not participate in the representation of the Ms = 0 triplet states and allows us to reliably extract the singlet–triplet splitting in O3 by analyzing the energy as a function of Fermi-orbital-descriptor permutations. The results of these analyses predict the percentage of open-shell character in O3, which agrees well with conventional wavefunction-based methods. While these techniques are expected to be required in cases near the Coulson–Fischer point, we find that they will be less necessary in diradical systems with well-separated electrons, such as (C21H13)2. Results based on energies from self-interaction-corrected generalized gradient, local density, and Hartree–Fock approximations and experimental results are in generally good agreement for O3. These results help form the basis for deriving extended Heisenberg-like Hamiltonians that are needed for descriptions of molecular magnets when there are competing low-energy electronic configurations.
Strongly correlated electrons: Ground states, excited states, and transport. Final report
The research supported by this grant focused on the quantum mechanical theory of the electrons in materials and molecules. Progress was made in dealing with electronic correlation effects in the ground state energy of molecular systems, and with topological concepts to classify the electronic state of molecules and materials, including excitation and transport properties. The physical and chemical properties of molecules and materials derive from their electronic structure, but the latter cannot be calculated exactly even with the most powerful computers because the computational cost of solving the exact equations of quantum mechanics increases exponentially with the number of electrons. The exponential cost originates from the correlations among the electrons that repel each other via Coulombic forces. In this project we have developed a new functional approximation for the ground state electronic energy that includes explicitly, and in a controllable way, the effects of the interelectronic correlations. In addition we have further developed topological concepts for classifying the electronic states of periodic ring molecules and solids. Topological concepts are very powerful because they allow us to predict subtle properties of materials and molecules using very general geometrical properties of the electron wavefunctions that do not depend on the quantitative details of the electronic interactions, which are very difficult to calculate with high accuracy. The development of a new class of controlled functional approximations for the ground state energy of molecules and materials was the main goal of the project. It has been fulfilled with the formulation of the occupation-probabilities natural orbital functional theory (OP-NOFT). This approach introduces new theoretical concepts but practical application has proved to be harder than anticipated. So far it has been utilized only at its lowest level of approximation in the context of relatively small molecules (with up to 16 atoms). The study of topological properties of the electron wavefunctions in materials was not proposed in the original proposal but was prompted during the funding period by our interaction with leading experimental groups in materials chemistry and physics at Princeton University.
Advanced stationary and nonstationary kernel designs for domain-aware Gaussian processes
Gaussian process regression is a widely-applied method for function approximation and uncertainty quantification. The technique has gained popularity recently in the machine learning community due to its robustness and interpretability. The mathematical methods we discuss in this paper are an extension of the Gaussian-process framework. We are proposing advanced kernel designs that only allow for functions with certain desirable characteristics to be elements of the reproducing kernel Hilbert space (RKHS) that underlies all kernel methods and serves as the sample space for Gaussian process regression. These desirable characteristics reflect the underlying physics; two obvious examples are symmetry and periodicity constraints. In addition, non-stationary kernel designs can be defined in the same framework to yield flexible multi-task Gaussian processes. We will show the impact of advanced kernel designs on Gaussian processes using several synthetic and two scientific data sets. The results of our research show that including domain knowledge, communicated through advanced kernel designs, has a significant impact on the accuracy and relevance of the function approximation.
Assessing the accuracy of compound formation energies with quantum Monte Carlo
Accurately predicting the formation energy of a compound, which describes its thermodynamic stability, is a key challenge in materials physics. Here, we employ a many-body quantum Monte Carlo (QMC) with single-reference trial functions to compute the formation energy of two electronically disparate compounds, the intermetallic VPt2 and the semiconductor CuI, for which standard density functional theory (DFT) predictions using both the Perdew-Burke-Ernzerhof (PBE) and the strongly constrained and appropriately normed (SCAN) density functional approximations deviate markedly from available experimental values. For VPt2, we find an agreement between QMC, SCAN, and PBE0 estimates, which therefore remain in disagreement with the much less exothermic experimental value. For CuI, the QMC result agrees with neither SCAN nor PBE pointing towards DFT exchange-correlation biases, likely related to the localized Cu 3d electrons. Compared to the behavior of some density functional approximations within DFT, spin-averaged QMC exhibits a smaller but still appreciable deviation when compared to experiment. The QMC result is slightly improved by incorporating spin-orbit corrections for CuI and solid I 2 , so that experiment and theory are brought into imperfect but reasonable agreement within about 120 meV/atom.
Reconnaissance with JWST of the J-region Asymptotic Giant Branch in Distance Ladder Galaxies: From Irregular Luminosity Functions to Approximation of the Hubble Constant
Abstract We study stars in the J-regions of the asymptotic giant branch (JAGB) of near-infrared color–magnitude diagrams in the maser host NGC 4258 and four hosts of six Type Ia supernovae (SNe Ia): NGC 1448, NGC 1559, NGC 5584, and NGC 5643. These clumps of stars are readily apparent near 1.0 < F150W − F277W < 1.5 andm F150W = 22–25 mag with James Webb Space Telescope NIRCam photometry. Various methods have been proposed to assign an apparent reference magnitude to this recently proposed standard candle, including the mode, median, sigma-clipped mean, or a modeled luminosity function parameter. We test the consistency of these by measuring intrahost variations, finding differences of up to ∼0.2 mag that significantly exceed statistical uncertainties. Brightness differences appear intrinsic, and are further amplified by the nonuniform shape of the JAGB luminosity function, also apparent in the LMC and SMC. We follow a “many methods” approach to measure consistently JAGB magnitudes and distance moduli to the SN Ia host sample calibrated by NGC 4258. We find broad agreement with distance moduli measured from Cepheids, tip of the red giant branch, and Miras. However, the SN host mean distance modulus estimated via the JAGB method necessary to estimateH 0 differs by ∼0.19 mag among the above definitions, the result of different levels of luminosity function asymmetry. The methods yield a full range of 71−78 km s −1 Mpc −1 , i.e., a fiducial result ofH 0 = 74.7 ± 2.1(stat) ± 2.3(sys, ±3.1 if combined in quadrature) km s −1 Mpc −1 , with systematic errors limited by the differences in methods. Future work may seek to standardize and refine this promising tool further, making it more competitive with established distance indicators.
Assessments of epistemic uncertainty using Gaussian stochastic weight averaging for fluid-flow regression
Here, we use Gaussian stochastic weight averaging (SWAG) to assess the epistemic uncertainty associated with neural-network-based function approximation relevant to fluid flows. SWAG approximates a posterior Gaussian distribution of each weight, given training data, and a constant learning rate. Having access to this distribution, it is able to create multiple models with various combinations of sampled weights, which can be used to obtain ensemble predictions. The average of such an ensemble can be regarded as the 'mean estimation', whereas its standard deviation can be used to construct 'confidence intervals', which enable us to perform uncertainty quantification (UQ) with regard to the training process of neural networks. We utilize representative neural-network-based function approximation tasks for the following cases: (i) a two-dimensional circular-cylinder wake; (ii) the DayMET dataset (maximum daily temperature in North America); (iii) a three-dimensional square-cylinder wake; and (iv) urban flow, to assess the generalizability of the present idea for a wide range of complex datasets. SWAG-based UQ can be applied regardless of the network architecture, and therefore, we demonstrate the applicability of the method for two types of neural networks: (i) global field reconstruction from sparse sensors by combining convolutional neural network (CNN) and multi-layer perceptron (MLP); and (ii) far-field state estimation from sectional data with two-dimensional CNN. We find that SWAG can obtain physically-interpretable confidence-interval estimates from the perspective of epistemic uncertainty. This capability supports its use for a wide range of problems in science and engineering.
Dynamic aspects of graphene deformation and fracture from approximate density functional theory
Graphene is one of the most intriguing two-dimensional carbon materials. Its mechanical strength and failure are key concerns for materials engineering and applications. Despite the success of fracture mechanics, the mechanism of how pristine materials fail remains an elusive problem. While many theoretical studies based on molecular dynamics using empirical forcefields have tried to address this question, atomic-scale mechanics are not clearly understood. Especially, a widely employed bond-breaking approach based on the critical bond length has not been rigorously tested. Here, utilizing molecular dynamics simulations with density functional based tight binding, we investigate how the failure of the pristine material initiates. The Wiberg bond order (W BO ) to estimate the change of chemical bonds shows a transition from sp 2 (W BO ~ 1.33) to sp 3 (WBO < 1.0) during the deformation. However, it reveals that a single threshold value for either the WBO or bond length is insufficient to decide failure of pristine material without free edges or defects. Instead, collective behaviors of the local atomic group govern the fracture initiation of pristine graphene. Overall, our study provides dynamic mechanical responses based on quantum mechanics, which have not been captured by empirical forcefields, opening opportunities to design properties by precisely coupling the mechanics and quantum mechanics.
Erratum: “DFTB+, a software package for efficient approximate density functional theory based atomistic simulations” [J. Chem. Phys. 152, 124101 (2020)]
The implementation of the GPU support in DFTB+, as described in Sec. III C of the original publication,1 was developed based on a previous unpublished implementation by Jacek Jakowski. In order to acknowledge his work on this first implementation, the authors of the original publication wish to include J. Jakowski as co-author. The scientific content of the original publication is not affected.
Mathematical nuances of Gaussian process-driven autonomous experimentation
Abstract The fields of machine learning (ML) and artificial intelligence (AI) have transformed almost every aspect of science and engineering. The excitement for AI/ML methods is in large part due to their perceived novelty, as compared to traditional methods of statistics, computation, and applied mathematics. But clearly, all methods in ML have their foundations in mathematical theories, such as function approximation, uncertainty quantification, and function optimization. Autonomous experimentation is no exception; it is often formulated as a chain of off-the-shelf tools, organized in a closed loop, without emphasis on the intricacies of each algorithm involved. The uncomfortable truth is that the success of any ML endeavor, and this includes autonomous experimentation, strongly depends on the sophistication of the underlying mathematical methods and software that have to allow for enough flexibility to consider functions that are in agreement with particular physical theories. We have observed that standard off-the-shelf tools, used by many in the applied ML community, often hide the underlying complexities and therefore perform poorly. In this paper, we want to give a perspective on the intricate connections between mathematics and ML, with a focus on Gaussian process-driven autonomous experimentation. Although the Gaussian process is a powerful mathematical concept, it has to be implemented and customized correctly for optimal performance. We present several simple toy problems to explore these nuances and highlight the importance of mathematical and statistical rigor in autonomous experimentation and ML. One key takeaway is that ML is not, as many had hoped, a set of agnostic plug-and-play solvers for everyday scientific problems, but instead needs expertise and mastery to be applied successfully. Graphical abstract
Coherency-Constrained Spectral Clustering for Power Network Reduction
This paper presents a methodology for reducing the complexity of large-scale power network models using spectral clustering, aggregation of electrical components, and cost function approximation. Two approaches are explored using unconstrained and constrained spectral clustering to determine areas for effective system reduction. Once the system areas are determined, both loads and generators by type are aggregated, and their new cost function is approximated through polynomial curve-fitting or statistical methods. The performance of reduced networks is evaluated in terms of their ability to follow the true daily cost of the original system over a 24-hour period considering a set of several days. Two test systems are taken as test beds. Application of the methodology to a modified version of the IEEE 39-bus system reduces it from 17 generators to a 4-bus system and 9 generators with about 93% of accuracy. Similarly, the IEEE 118-bus system is reduced from 19 generators to a 3-bus system with three aggregated units achieving over 99% of accuracy. These findings address scalability challenges and enhance accuracy for high and mid-loading level conditions, and by aggregating thermal units with similar cost functions.
DeepONet-grid-UQ: A trustworthy deep operator framework for predicting the power grid’s post-fault trajectories
This paper proposes a novel data-driven method for the reliable prediction of the power grid’s post-fault trajectories, i.e., the power grid’s dynamic response after a disturbance or fault. Here, the proposed method is based on the recently proposed concept of Deep Operator Networks (DeepONets). Unlike traditional neural networks that learn to approximate functions, DeepONets are designed to approximate nonlinear operators, i.e., mappings between infinite-dimensional spaces. Under this operator framework, we design a novel and efficient DeepONet that (i) takes as inputs the trajectories collected before and during the fault and (ii) outputs the predicted post-fault trajectories. In addition, we endow our method with the much-needed ability to balance efficiency with reliable/trustworthy predictions via uncertainty quantification. To this end, we propose and compare two novel methods that enable quantifying the predictive uncertainty. First, we propose a Bayesian DeepONet (B-DeepONet) that uses stochastic gradient Hamiltonian Monte-Carlo to sample from the posterior distribution of the DeepONet trainable parameters. Then, we design a Probabilistic DeepONet (Prob-DeepONet) that uses a probabilistic training strategy to enable quantifying uncertainty at virtually no extra computational cost. Finally, we validate the proposed methods’ predictive power and uncertainty quantification capability using the New York-New England power grid model.
Introductory lecture: when the density of the noninteracting reference system is not the density of the physical system in density functional theory
We develop expressions for electron density defined through the linear response for general density functional approximations, demonstrating results for orbital functionals and for many-body perturbation theory, and explore the connections to developments in DFT.