arXiv subjects

Andrew J. Medford

Publications and source records attributed to Andrew J. Medford.

35 records · Page 2Linked to original sources

arXiv2023

Ab-initio investigation of finite size effects in rutile titania nanoparticles with semilocal and nonlocal density functionals

In this work, we employ hybrid and generalized gradient approximation (GGA) level density functional theory (DFT) calculations to investigate the convergence of surface properties and band structure of rutile titania (TiO$_2$) nanoparticles with particle size. The surface energies and band structures are calculated for cuboidal particles with minimum dimension ranging from 3.7 Å (24 atoms) to 10.3 Å (384 atoms) using a highly-parallel real-space DFT code to enable hybrid level DFT calculations of larger nanoparticles than are typically practical. We deconvolute the geometric and electronic finite size effects in surface energy, and evaluate the influence of defects on band structure and density of states (DOS). The electronic finite size effects in surface energy vanish when the minimum length scale of the nanoparticles becomes greater than 10 Å. We show that this length scale is consistent with a computationally efficient numerical analysis of the characteristic length scale of electronic interactions. The surface energy of nanoparticles having minimum dimension beyond this characteristic length can be approximated using slab calculations that account for the geometric defects. In contrast, the finite size effects on the band structure is highly dependent on the shape and size of these particles. The DOS for cuboidal particles and more realistic particles constructed using the Wulff algorithm reveal that defect states within the bandgap play a key role in determining the band structure of nanoparticles and the bandgap does not converge to the bulk limit for the particle sizes investigated.

cond-mat.mtrl-sci↗

arXiv2023

Maximum-likelihood Estimators in Physics-Informed Neural Networks for High-dimensional Inverse Problems

Physics-informed neural networks (PINNs) have proven a suitable mathematical scaffold for solving inverse ordinary (ODE) and partial differential equations (PDE). Typical inverse PINNs are formulated as soft-constrained multi-objective optimization problems with several hyperparameters. In this work, we demonstrate that inverse PINNs can be framed in terms of maximum-likelihood estimators (MLE) to allow explicit error propagation from interpolation to the physical model space through Taylor expansion, without the need of hyperparameter tuning. We explore its application to high-dimensional coupled ODEs constrained by differential algebraic equations that are common in transient chemical and biological kinetics. Furthermore, we show that singular-value decomposition (SVD) of the ODE coupling matrices (reaction stoichiometry matrix) provides reduced uncorrelated subspaces in which PINNs solutions can be represented and over which residuals can be projected. Finally, SVD bases serve as preconditioners for the inversion of covariance matrices in this hyperparameter-free robust application of MLE to ``kinetics-informed neural networks''.

cs.LG↗

arXiv2022

Robust and scalable uncertainty estimation with conformal prediction for machine-learned interatomic potentials

Uncertainty quantification (UQ) is important to machine learning (ML) force fields to assess the level of confidence during prediction, as ML models are not inherently physical and can therefore yield catastrophically incorrect predictions. Established a-posteriori UQ methods, including ensemble methods, the dropout method, the delta method, and various heuristic distance metrics, have limitations such as being computationally challenging for large models due to model re-training. In addition, the uncertainty estimates are often not rigorously calibrated. In this work, we propose combining the distribution-free UQ method, known as conformal prediction (CP), with the distances in the neural network's latent space to estimate the uncertainty of energies predicted by neural network force fields. We evaluate this method (CP+latent) along with other UQ methods on two essential aspects, calibration, and sharpness, and find this method to be both calibrated and sharp under the assumption of independent and identically-distributed (i.i.d.) data. We show that the method is relatively insensitive to hyperparameters selected, and test the limitations of the method when the i.i.d. assumption is violated. Finally, we demonstrate that this method can be readily applied to trained neural network force fields with traditional and graph neural network architectures to obtain estimates of uncertainty with low computational costs on a training dataset of 1 million images to showcase its scalability and portability. Incorporating the CP method with latent distances offers a calibrated, sharp and efficient strategy to estimate the uncertainty of neural network force fields. In addition, the CP approach can also function as a promising strategy for calibrating uncertainty estimated by other approaches.

physics.chem-ph↗

arXiv2022

Soft and transferable pseudopotentials from multi-objective optimization

Ab initio pseudopotentials are a linchpin of modern molecular and condensed matter electronic structure calculations. In this work, we employ multi-objective optimization to maximize pseudopotential softness while maintaining high accuracy and transferability. To accomplish this, we develop a formulation in which softness and accuracy are simultaneously maximized, with accuracy determined by the ability to reproduce all-electron energy differences between Bravais lattice structures, whereupon the resulting Pareto frontier is scanned for the softest pseudopotential that provides the desired accuracy in established transferability tests. We employ an evolutionary algorithm to solve the multi-objective optimization problem and apply it to generate a comprehensive table of optimized norm-conserving Vanderbilt (ONCV) pseudopotentials (https://github.com/SPARC-X/SPMS-psps). We show that the resulting table is softer than existing tables of comparable accuracy, while more accurate than tables of comparable softness. The potentials thus afford the possibility to speed up calculations in a broad range of applications areas while maintaining high accuracy.

physics.chem-ph↗

arXiv2022

A Universal Framework for Featurization of Atomistic Systems

Molecular dynamics simulations are an invaluable tool in numerous scientific fields. However, the ubiquitous classical force fields cannot describe reactive systems, and quantum molecular dynamics are too computationally demanding to treat large systems or long timescales. Reactive force fields based on physics or machine learning can be used to bridge the gap in time and length scales, but these force fields require substantial effort to construct and are highly specific to a given chemical composition and application. A significant limitation of machine learning models is the use of element-specific features, leading to models that scale poorly with the number of elements. This work introduces the Gaussian multipole (GMP) featurization scheme that utilizes physically-relevant multipole expansions of the electron density around atoms to yield feature vectors that interpolate between element types and have a fixed dimension regardless of the number of elements present. We combine GMP with neural networks to directly compare it to the widely used Behler-Parinello symmetry functions for the MD17 dataset, revealing that it exhibits improved accuracy and computational efficiency. Further, we demonstrate that GMP-based models can achieve chemical accuracy for the QM9 dataset, and their accuracy remains reasonable even when extrapolating to new elements. Finally, we test GMP-based models for the Open Catalysis Project (OCP) dataset, revealing comparable performance to graph convolutional deep learning models. The results indicate that this featurization scheme fills a critical gap in the construction of efficient and transferable machine-learned force fields.

physics.chem-ph↗

arXiv2022

ElectroLens: Understanding Atomistic Simulations Through Spatially-resolved Visualization of High-dimensional Features

In recent years, machine learning (ML) has gained significant popularity in the field of chemical informatics and electronic structure theory. These techniques often require researchers to engineer abstract "features" that encode chemical concepts into a mathematical form compatible with the input to machine-learning models. However, there is no existing tool to connect these abstract features back to the actual chemical system, making it difficult to diagnose failures and to build intuition about the meaning of the features. We present ElectroLens, a new visualization tool for high-dimensional spatially-resolved features to tackle this problem. The tool visualizes high-dimensional data sets for atomistic and electron environment features by a series of linked 3D views and 2D plots. The tool is able to connect different derived features and their corresponding regions in 3D via interactive selection. It is built to be scalable, and integrate with existing infrastructure.

cs.HC↗

arXiv2021

Kinetics-Informed Neural Networks

Chemical kinetics and reaction engineering consists of the phenomenological framework for the disentanglement of reaction mechanisms, optimization of reaction performance and the rational design of chemical processes. Here, we utilize feed-forward artificial neural networks as basis functions to solve ordinary differential equations (ODEs) constrained by differential algebraic equations (DAEs) that describe microkinetic models (MKMs). We present an algebraic framework for the mathematical description and classification of reaction networks, types of elementary reaction, and chemical species. Under this framework, we demonstrate that the simultaneous training of neural nets and kinetic model parameters in a regularized multi-objective optimization setting leads to the solution of the inverse problem through the estimation of kinetic parameters from synthetic experimental data. We analyze a set of scenarios to establish the extent to which kinetic parameters can be retrieved from transient kinetic data, and assess the robustness of the methodology with respect to statistical noise. This approach to inverse kinetic ODEs can assist in the elucidation of reaction mechanisms based on transient data.

cs.LG↗

arXiv2021

A Priori Calibration of Transient Kinetics Data via Machine Learning

The temporal analysis of products reactor provides a vast amount of transient kinetic information that may be used to describe a variety of chemical features including the residence time distribution, kinetic coefficients, number of active sites, and the reaction mechanism. However, as with any measurement device, the TAP reactor signal is convoluted with noise. To reduce the uncertainty of the kinetic measurement and any derived parameters or mechanisms, proper preprocessing must be performed prior to any advanced analysis. This preprocessing consists of baseline correction, i.e., a shift in the voltage response, and calibration, i.e., a scaling of the flux response based on prior experiments. The current methodology of preprocessing requires significant user discretion and reliance on previous experiments that may drift over time. Herein we use machine learning techniques combined with physical constraints to convert the raw instrument signal to chemical information. As such, the proposed methodology demonstrates clear benefits over the traditional preprocessing in the calibration of the inert and feed mixture products without need of prior calibration experiments or heuristic input from the user.

cs.LG↗

arXiv2021

Application of DFTB and machine learning to evaluate the stability of biomass intermediates on the Rh(111) surface

Biomass compounds adsorbed on surfaces are challenging to study due to the large number of possible species and adsorption geometries. In this work, possible intermediates of erythrose, glyceraldehyde, glycerol and propionic acid are studied on the Rh(111) surface. The intermediates and elementary reactions are generated from first 2 recursions of a recursive bond-breaking algorithm. These structures are used as the input of an unsupervised Mol2Vec algorithm to generate vector descriptors. A data-driven scheme to classify the reactions is developed and adsorption energies are predicted. The lowest mean absolute error (MAE) of our prediction on adsorption energies is 0.39 eV, and the relative ordering of different surface adsorption geometries is relatively accurate. We show that combining geometries from density functional tight-binding (DFTB) calculations with energies from machine-learning predictions provides a novel workflow for rapidly assessing the stability of various molecular geometries on the Rh(111) surface.

cond-mat.mtrl-sci↗

arXiv2021

Data Driven Reaction Mechanism Estimation via Transient Kinetics and Machine Learning

Understanding the set of elementary steps and kinetics in each reaction is extremely valuable to make informed decisions about creating the next generation of catalytic materials. With physical and mechanistic complexity of industrial catalysts, it is critical to obtain kinetic information through experimental methods. As such, this work details a methodology based on the combination of transient rate/concentration dependencies and machine learning to measure the number of active sites, the individual rate constants, and gain insight into the mechanism under a complex set of elementary steps. This new methodology was applied to simulated transient responses to verify its ability to obtain correct estimates of the micro-kinetic coefficients. Furthermore, experimental CO oxidation data was analyzed to reveal the Langmuir-Hinshelwood mechanism driving the reaction. As oxygen accumulated on the catalyst, a transition in the mechanism was clearly defined in the machine learning analysis due to the large amount of kinetic information available from transient reaction techniques. This methodology is proposed as a new data driven approach to characterize how materials control complex reaction mechanisms relying exclusively on experimental data.

stat.AP↗

arXiv2020

Heterogeneity in susceptibility dictates the order of epidemiological models

The fundamental models of epidemiology describe the progression of an infectious disease through a population using compartmentalized differential equations, but do not incorporate population-level heterogeneity in infection susceptibility. We show that variation strongly influences the rate of infection, while the infection process simultaneously sculpts the susceptibility distribution. These joint dynamics influence the force of infection and are, in turn, influenced by the shape of the initial variability. Intriguingly, we find that certain susceptibility distributions (the exponential and the gamma) are unchanged through the course of the outbreak, and lead naturally to power-law behavior in the force of infection; other distributions often tend towards these "eigen-distributions" through the process of contagion. The power-law behavior fundamentally alters predictions of the long-term infection rate, and suggests that first-order epidemic models that are parameterized in the exponential-like phase may systematically and significantly over-estimate the final severity of the outbreak.

q-bio.PE↗

arXiv2020

Population Susceptibility Variation and Its Effect on Contagion Dynamics

Susceptibility governs the dynamics of contagion. The classical SIR model is one of the simplest compartmental models of contagion spread, assuming a single shared susceptibility level. However, variation in susceptibility over a population can fundamentally affect the dynamics of contagion and thus the ultimate outcome of a pandemic. We develop mathematical machinery which explicitly considers susceptibility variation, illuminates how the susceptibility distribution is sculpted by contagion, and thence how such variation affects the SIR differential questions that govern contagion. Our methods allow us to derive closed form expressions for herd immunity thresholds as a function of initial susceptibility distributions and suggests an intuitively satisfying approach to inoculation when only a fraction of the population is accessible to such intervention. Of particular interest, if we assume static susceptibility of individuals in the susceptible pool, ignoring susceptibility diversity {\em always} results in overestimation of the herd immunity threshold and that difference can be dramatic. Therefore, we should develop robust measures of susceptibility variation as part of public health strategies for handling pandemics.

q-bio.PE↗

arXiv2020

TAPsolver: A Python package for the simulation and analysis of TAP reactor experiments

An open-source, Python-based Temporal Analysis of Products (TAP) reactor simulation and processing program is introduced. TAPsolver utilizes algorithmic differentiation for the calculation of highly accurate derivatives, which are used to perform sensitivity analyses and PDE-constrained optimization. The tool supports constraints to ensure thermodynamic consistency, which can lead to more accurate parameters and assist in mechanism discrimination. The mathematical and structural details of TAPsolver are outlined, as well as validation of the forward and inverse problems against well-studied prototype problems. Benchmarks of the code are presented, and a case study for extracting thermodynamically-consistent kinetic parameters from experimental TAP measurements of CO oxidation on supported platinum particles is presented. TAPsolver will act as a foundation for future development and dissemination of TAP data processing techniques.

cs.CE↗

arXiv2020

SPARC: Simulation Package for Ab-initio Real-space Calculations

We present SPARC: Simulation Package for Ab-initio Real-space Calculations. SPARC can perform Kohn-Sham density functional theory calculations for isolated systems such as molecules as well as extended systems such as crystals and surfaces, in both static and dynamic settings. It is straightforward to install/use and highly competitive with state-of-the-art planewave codes, demonstrating comparable performance on a small number of processors and increasing advantages as the number of processors grows. Notably, SPARC brings solution times down to a few seconds for systems with $\mathcal{O}(100-500)$ atoms on large-scale parallel computers, outperforming planewave counterparts by an order of magnitude and more.

physics.comp-ph↗

arXiv2019

Computational Study of Transition-Metal Substitutions in Rutile TiO$_2$ (110) for Photoelectrocatalytic Ammonia Synthesis

Synthesis of ammonia through photo- and electrocatalysis is a rapidly growing field. Titania-based catalysts are widely reported for photocatalytic ammonia synthesis and have also been suggested as electrocatalysts. The addition of transition-metal dopants is one strategy for improving the performance of titania-based catalysts. In this work, we screen d-block transition-metal dopants for surface site stability and evaluate trends in their performance as the active site for the reduction of nitrogen to ammonia on TiO$_2$. We find a linear relationship between the d-band center and formation energy of the dopant site, while the binding energies of N$_2$, N$_2$H, and NH$_2$ all are strongly correlated with the cohesive energies of the dopant metals. The activity of the metal-doped systems shows a volcano type relationship with the NH$_2$ and N$_2$H energies as descriptors. Some metals such as Co, Mo, and V are predicted to slightly improve photo- and electrocatalytic performance, but most metals inhibit the ammonia synthesis reaction. The results provide insight into the role of transition-metal dopants for promoting ammonia synthesis, and the trends are based on unexpected electronic structure factors that may have broader implications for single-atom catalysis and doped oxides.

physics.chem-ph↗

arXiv2019

Design and Analysis of Machine Learning Exchange-Correlation Functionals via Rotationally Invariant Convolutional Descriptors

In this work we explore the potential of a new data-driven approach to the design of exchange-correlation (XC) functionals. The approach, inspired by convolutional filters in computer vision and surrogate functions from optimization, utilizes convolutions of the electron density to form a feature space to represent local electronic environments and neural networks to map the features to the exchange-correlation energy density. These features are orbital free, and provide a systematic route to including information at various length scales. This work shows that convolutional descriptors are theoretically capable of an exact representation of the electron density, and proposes Maxwell-Cartesian spherical harmonic kernels as a class of rotationally invariant descriptors for the construction of machine-learned functionals. The approach is demonstrated using data from the B3LYP functional on a number of small-molecules containing C, H, O, and N along with a neural network regression model. The machine-learned functionals are compared to standard physical approximations and the accuracy is assessed for the absolute energy of each molecular system as well as formation energies. The results indicate that it is possible to reproduce B3LYP formation energies to within chemical accuracy using orbital-free descriptors with a spatial extent of 0.2 A. The findings provide empirical insight into the spatial range of electron exchange, and suggest that the combination of convolutional descriptors and machine-learning regression models is a promising new framework for XC functional design, although challenges remain in obtaining training data and generating models consistent with pseudopotentials.

physics.chem-ph↗

arXiv2018

Analysis of Photocatalytic Nitrogen Fixation on Rutile TiO$_2$(110)

Photocatalytic nitrogen fixation provides a promising route to produce reactive nitrogen compounds at benign conditions. Titania has been reported as an active photocatalyst for reduction of dinitrogen to ammonia; however there is little fundamental understanding of how this process occurs. In this work the rutile (110) model surface is hypothesized to be the active site, and a computational model based on the Bayesian error estimation functional (BEEF-vdW) and computational hydrogen electrode is applied in order to analyze the expected dinitrogen coverage at the surface as well as the overpotentials for electrochemical reduction and oxidation. This is the first application of computational techniques to photocatalytic nitrogen fixation, and the results indicate that the thermodynamic limiting potential for nitrogen reduction on rutile (110) is considerably higher than the conduction band edge of rutile TiO$_2$, even at oxygen vacancies and iron substitutions. This work provides strong evidence against the most commonly reported experimental hypotheses, and indicates that rutile (110) is unlikely to be the relevant surface for nitrogen reduction. However, the limiting potential for nitrogen oxidation on rutile (110) is significantly lower, indicating that oxidative pathways may be relevant on rutile (110). These findings suggest that photocatalytic dinitrogen fixation may occur via a complex balance of oxidative and reductive processes.

physics.chem-ph↗