Contributions to Diffusion in Complex Materials Quantified with Machine Learning.

Using machine learning with a variational formula for diffusivity, we recast diffusion as a sum of individual contributions to diffusion-called "kinosons"-and compute their statistical distribution to model a complex multicomponent alloy. Calculating kinosons is orders of magnitude more efficient than computing whole trajectories, and it elucidates kinetic mechanisms for diffusion. The density of kinosons with temperature leads to new accurate analytic models for macroscale diffusivity. This combination of machine learning with diffusion theory promises insight into other complex materials.

Discovering melting temperature prediction models of inorganic solids by combining supervised and unsupervised learning.

The melting temperature is important for materials design because of its relationship with thermal stability, synthesis, and processing conditions. Current empirical and computational melting point estimation techniques are limited in scope, computational feasibility, or interpretability. We report the development of a machine learning methodology for predicting melting temperatures of binary ionic solid materials. We evaluated different machine-learning models trained on a dataset of the melting points of 476 non-metallic crystalline binary compounds using materials embeddings constructed from elemental properties and density-functional theory calculations as model inputs. A direct supervised-learning approach yields a mean absolute error of around 180 K but suffers from low interpretability. We find that the fidelity of predictions can further be improved by introducing an additional unsupervised-learning step that first classifies the materials before the melting-point regression. Not only does this two-step model exhibit improved accuracy, but the approach also provides a level of interpretability with insights into feature importance and different types of melting that depend on the specific atomic bonding inside a material. Motivated by this finding, we used a symbolic learning approach to find interpretable physical models for the melting temperature, which recovered the best-performing features from both prior models and provided additional interpretability.

Machine Learning Methods for Multiscale Physics and Urban Engineering Problems.

We present an overview of four challenging research areas in multiscale physics and engineering as well as four data science topics that may be developed for addressing these challenges. We focus on multiscale spatiotemporal problems in light of the importance of understanding the accompanying scientific processes and engineering ideas, where "multiscale" refers to concurrent, non-trivial and coupled models over scales separated by orders of magnitude in either space, time, energy, momenta, or any other relevant parameter. Specifically, we consider problems where the data may be obtained at various resolutions; analyzing such data and constructing coupled models led to open research questions in various applications of data science. Numeric studies are reported for one of the data science techniques discussed here for illustration, namely, on approximate Bayesian computations.

Variational Principle for Mass Transport.

A variation principle for mass transport in solids is derived that recasts transport coefficients as minima of local thermodynamic average quantities. The result is independent of diffusion mechanisms and applies to amorphous and crystalline systems. This unifies different computational approaches for diffusion and provides a framework for the creation of new approximation methods with error estimation. It gives a different physical interpretation of the Green function. Finally, the variational principle quantifies the accuracy of competing approaches for a nontrivial diffusion problem.

Exact Model of Vacancy-Mediated Solute Transport in Magnesium.

Most substitutional solutes in solids diffuse via vacancies. However, widely used analytic models for diffusivity make uncontrolled approximations in the relations between atomic jump rates that reduce accuracy. Symmetry analysis of the hexagonal close packed crystal identifies more distinct vacancy transitions than prior models, and a Green function approach computes diffusivity exactly for solutes in magnesium. We find large differences for the solute drag of Al, Zn, and rare earth solutes, and improved diffusion activation energies-highlighting the need for exact analytic transport models.

Direct measurement of hydrogen dislocation pipe diffusion in deformed polycrystalline Pd using quasielastic neutron scattering.

The temperature-dependent diffusivity D(T) of hydrogen solute atoms trapped at dislocations-dislocation pipe diffusion of hydrogen-in deformed polycrystalline PdH(x) (x∼10(-3) [H]/[Pd]) has been quantified with quasielastic neutron scattering between 150 and 400 K. We observe diffusion coefficients for trapped hydrogen elevated by one to two orders of magnitude above bulk diffusion. Arrhenius diffusion behavior has been observed for dislocation pipe diffusion and regular bulk diffusion, the latter in well-annealed polycrystalline Pd. For regular bulk diffusion of hydrogen in Pd we find D(T)=D(0)exp(-E(a)/kT)=0.005exp(-0.23 eV/kT) cm(2)/s, in agreement with the known diffusivity of hydrogen in Pd. For hydrogen dislocation pipe diffusion we find D(T)≃10(-5)exp(-E(a)/kT) cm(2)/s, where E(a)=0.042 and 0.083 eV for concentrations of 0.52×10(-3) and 1.13×10(-3)[H]/[Pd], respectively. Ab initio computations provide a physical basis for the pipe diffusion pathway and confirm the reduced barrier height.

Direct diffusion through interpenetrating networks: oxygen in titanium.

How impurity atoms move through a crystal is a fundamental and recurrent question in materials. The previous models of oxygen diffusion in titanium relied on interstitial lattice sites that were recently found to be unstable--leaving no consistent picture of the diffusion pathways. Using first-principles quantum-mechanical methods, we find three oxygen interstitial sites in titanium, and quantify the multiple interpenetrating networks for oxygen diffusion. Surprisingly, all transitions contribute to diffusion.

Accurate and efficient algorithm for Bader charge integration.

We propose an efficient, accurate method to integrate the basins of attraction of a smooth function defined on a general discrete grid and apply it to the Bader charge partitioning for the electron charge density. Starting with the evolution of trajectories in space following the gradient of charge density, we derive an expression for the fraction of space neighboring each grid point that flows to its neighbors. This serves as the basis to compute the fraction of each grid volume that belongs to a basin (Bader volume) and as a weight for the discrete integration of functions over the Bader volume. Compared with other grid-based algorithms, our approach is robust, more computationally efficient with linear computational effort, accurate, and has quadratic convergence. Moreover, it is straightforward to extend to nonuniform grids, such as from a mesh-refinement approach, and can be used to both identify basins of attraction of fixed points and integrate functions over the basins.

Mechanism of a prototypical synthetic membrane-active antimicrobial: Efficient hole-punching via interaction with negative intrinsic curvature lipids.

Phenylene ethynylenes comprise a prototypical class of synthetic antimicrobial compounds that mimic antimicrobial peptides produced by eukaryotes and have broad-spectrum antimicrobial activity. We show unambiguously that bacterial membrane permeation by these antimicrobials depends on the presence of negative intrinsic curvature lipids, such as phosphatidylethanolamine (PE) lipids, found in high concentrations within bacterial membranes. Plate-killing assays indicate that a PE-knockout mutant strain of Escherichia coli drastically out-survives the wild type against the membrane-active phenylene ethynylene antimicrobials, whereas the opposite is true when challenged with traditional metabolic antibiotics. That the PE deletion is a lethal mutation in normative environments suggests that resistant bacterial strains do not evolve because a lethal mutation is required to gain immunity. PE lipids allow efficient generation of negative curvature required for the circumferential barrel of an induced membrane pore; an inverted hexagonal H(II) phase, which consists of arrays of water channels, is induced by a small number of antimicrobial molecules. The estimated antimicrobial occupation in these water channels is nonlinear and jumps from approximately 1 to 3 per 4 nm of induced water channel length as the global antimicrobial concentration is increased. By comparing to exactly solvable 1D spin models for magnetic systems, we quantify the cooperativity of these antimicrobials.

Designing Optimal Perovskite Structure for High Ionic Conduction.

Solid-oxide fuel/electrolyzer cells are limited by a dearth of electrolyte materials with low ohmic loss and an incomplete understanding of the structure-property relationships that would enable the rational design of better materials. Here, using epitaxial thin-film growth, synchrotron radiation, impedance spectroscopy, and density-functional theory, the impact of structural parameters (i.e., unit-cell volume and octahedral rotations) on ionic conductivity is delineated in La0.9 Sr0.1 Ga0.95 Mg0.05 O3- δ . As compared to the zero-strain state, compressive strain reduces the unit-cell volume while maintaining large octahedral rotations, resulting in a strong reduction of ionic conductivity, while tensile strain increases the unit-cell volume while quenching octahedral rotations, resulting in a negligible effect on the ionic conductivity. Calculations reveal that larger unit-cell volumes and octahedral rotations decrease migration barriers and create low-energy migration pathways, respectively. The desired combination of large unit-cell volume and octahedral rotations is normally contraindicated, but through the creation of superlattice structures both expanded unit-cell volume and large octahedral rotations are experimentally realized, which result in an enhancement of the ionic conductivity. All told, the potential to tune ionic conductivity with structure alone by a factor of ≈2.5 at around 600 °C is observed, which sheds new light on the rational design of ion-conducting perovskite electrolytes.

Data files for ab initio calculations of the lattice parameter and elastic stiffness coefficients of bcc Fe with solutes.

We present computed datasets on changes in the lattice parameter and elastic stiffness coefficients of bcc Fe due to substitutional Al, B, Cu, Mn, and Si solutes, and octahedral interstitial C and N solutes. The data is calculated using the methodology based on density functional theory (DFT) presented in Ref. (M.R. Fellinger, L.G. Hector Jr., D.R. Trinkle, 2017) [1]. All the DFT calculations were performed using the Vienna Ab initio Simulations Package (VASP) (G. Kresse, J. Furthmüller, 1996) [2]. The data is stored in the NIST dSpace repository (http://hdl.handle.net/11256/671).

Kinetic Monte Carlo investigation of tetragonal strain on Onsager matrices.

We use three different methods to compute the derivatives of Onsager matrices with respect to strain for vacancy-mediated multicomponent diffusion from kinetic Monte Carlo simulations. We consider a finite difference method, a correlated finite difference method to reduce the relative statistical errors, and a perturbation theory approach to compute the derivatives. We investigate the statistical error behavior of the three methods for uncorrelated single vacancy diffusion in fcc Ni and for correlated vacancy-mediated diffusion of Si in Ni. While perturbation theory performs best for uncorrelated systems, the correlated finite difference method performs best for the vacancy-mediated Si diffusion in Ni, where longer trajectories are required.

Computation of the lattice Green function for a dislocation.

Modeling isolated dislocations is challenging due to their long-ranged strain fields. Flexible boundary condition methods capture the correct long-range strain field of a defect by coupling the defect core to an infinite harmonic bulk through the lattice Green function (LGF). To improve the accuracy and efficiency of flexible boundary condition methods, we develop a numerical method to compute the LGF specifically for a dislocation geometry; in contrast to previous methods, where the LGF was computed for the perfect bulk as an approximation for the dislocation. Our approach directly accounts for the topology of a dislocation, and the errors in the LGF computation converge rapidly for edge dislocations in a simple cubic model system as well as in BCC Fe with an empirical potential. When used within the flexible boundary condition approach, the dislocation LGF relaxes dislocation core geometries in fewer iterations than when the perfect bulk LGF is used as an approximation for the dislocation, making a flexible boundary condition approach more efficient.

Direct calculation of the lattice Green function with arbitrary interactions for general crystals.

Efficient computation of lattice defect geometries such as point defects, dislocations, disconnections, grain boundaries, interfaces, and free surfaces requires accurate coupling of displacements near the defect to the long-range elastic strain. Flexible boundary condition methods embed a defect in infinite harmonic bulk through the lattice Green function. We demonstrate an efficient and accurate calculation of the lattice Green function from the force-constant matrix for general crystals with an arbitrary basis by extending a method for Bravais lattices. New terms appear due to the presence of optical modes and the possible loss of inversion symmetry. By separately treating poles and discontinuities in reciprocal space, numerical accuracy is controlled at all distances. We compute the lattice Green function for a two-dimensional model with broken symmetry to elucidate the role of different coupling terms. The algorithm is generally applicable in two and three dimensions to crystals with arbitrary number of atoms in the unit cell, symmetry, and interactions.

The chemistry of deformation: how solutes soften pure metals.

Solutes have been added to strengthen elemental metals, generating usable materials for millennia; in the 1960s, solutes were found to also soften metals. Despite the empirical correlation between the "electron number" of the solute and the change in strength of the material to which it is added, the mechanism responsible for softening is poorly understood. Using state-of-the-art quantum-mechanical methods, we studied the direct interaction of transition-metal solutes with dislocations in molybdenum. The interaction increases dramatically with increasing electron number and strongly influences the mechanisms responsible for plasticity in these materials. Our quantitative model explains solution softening of metals by using changes in energy and stress scales of plasticity from solutes.

Impurities block the alpha to omega martensitic transformation in titanium.

Impurities control phase stability and phase transformations in natural and man-made materials, from shape-memory alloys to steel to planetary cores. Experiments and empirical databases are still central to tuning the impurity effects. What is missing is a broad theoretical underpinning. Consider, for example, the titanium martensitic transformations: diffusionless structural transformations proceeding near the speed of sound. Pure titanium transforms from ductile alpha to brittle omega at 9 GPa, creating serious technological problems for beta-stabilized titanium alloys. Impurities in the titanium alloys A-70 and Ti-6Al-4V (wt%) suppress the transformation up to at least 35 GPa, increasing their technological utility as lightweight materials in aerospace applications. These and other empirical discoveries in technological materials call for broad theoretical understanding. Impurities pose two theoretical challenges: the effect on the relative phase stability, and the energy barrier of the transformation. Ab initio methods calculate both changes due to impurities. We show that interstitial oxygen, nitrogen and carbon retard the transformation whereas substitutional aluminium and vanadium influence the transformation by changing the d-electron concentration. The resulting microscopic picture explains the suppression of the transformation in commercial A-70 and Ti-6Al-4V alloys. In general, the effect of impurities on relative energies and energy barriers is central to understanding structural phase transformations.