Integrated workflows and interfaces for data-driven semi-empirical electronic structure calculations.

Modern software engineering of electronic structure codes has seen a paradigm shift from monolithic workflows toward object-based modularity. Software objectivity allows for greater flexibility in the application of electronic structure calculations, with particular benefits when integrated with approaches for data-driven analysis. Here, we discuss different approaches to create deep modular interfaces that connect big-data workflows and electronic structure codes and explore the diversity of use cases that they can enable. We present two such interface approaches for the semi-empirical electronic structure package, DFTB+. In one case, DFTB+ is applied as a library and provides data to an external workflow; in another, DFTB+receives data via external bindings and processes the information subsequently within an internal workflow. We provide a general framework to enable data exchange workflows for embedding new machine-learning-based Hamiltonians within DFTB+ or enabling deep integration of DFTB+ in multiscale embedding workflows. These modular interfaces demonstrate opportunities in emergent software and workflows to accelerate scientific discovery by harnessing existing software capabilities.

Sensitivity and dimensionality of atomic environment representations used for machine learning interatomic potentials.

Faithfully representing chemical environments is essential for describing materials and molecules with machine learning approaches. Here, we present a systematic classification of these representations and then investigate (i) the sensitivity to perturbations and (ii) the effective dimensionality of a variety of atomic environment representations and over a range of material datasets. Representations investigated include atom centered symmetry functions, Chebyshev Polynomial Symmetry Functions (CHSF), smooth overlap of atomic positions, many-body tensor representation, and atomic cluster expansion. In area (i), we show that none of the atomic environment representations are linearly stable under tangential perturbations and that for CHSF, there are instabilities for particular choices of perturbation, which we show can be removed with a slight redefinition of the representation. In area (ii), we find that most representations can be compressed significantly without loss of precision and, further, that selecting optimal subsets of a representation method improves the accuracy of regression models built for a given dataset.

Insights into the Emerging Networks of Voids in Simulated Supercooled Water.

The structural evolution of supercooled liquid water as we approach the glass transition temperature continues to be an active area of research. Here, we use molecular dynamics simulations of TIP4P/ice water to study the changes in the connected regions of empty space within the liquid, which we investigate using the Voronoi-voids network. We observe two important features: supercooling enhances the fraction of nonspherical voids and different sizes of voids tend to cluster forming a percolating network. By examining order parameters such as the local structure index (LSI), tetrahedrality and topological defects, we show that water molecules near large void clusters tend to be slightly more tetrahedral than those near small voids, with a lower population of under- and overcoordinated defects. We show further that the distribution of closed rings of water molecules around small and large void clusters maintain a balance between 6 and 7 membered rings. Our results highlight the changes of the dual voids and water network as a structural hallmark of supercooling and provide insights into the molecular origins of cooperative effects underlying density fluctuations on the subnanometer and nanometer length scale. In addition, the percolation of the voids and the hydrogen bond network around the voids may serve as useful order parameters to investigate density fluctuations in supercooled water.

Representations in neural network based empirical potentials.

Many structural and mechanical properties of crystals, glasses, and biological macromolecules can be modeled from the local interactions between atoms. These interactions ultimately derive from the quantum nature of electrons, which can be prohibitively expensive to simulate. Machine learning has the potential to revolutionize materials modeling due to its ability to efficiently approximate complex functions. For example, neural networks can be trained to reproduce results of density functional theory calculations at a much lower cost. However, how neural networks reach their predictions is not well understood, which has led to them being used as a "black box" tool. This lack of understanding is not desirable especially for applications of neural networks in scientific inquiry. We argue that machine learning models trained on physical systems can be used as more than just approximations since they had to "learn" physical concepts in order to reproduce the labels they were trained on. We use dimensionality reduction techniques to study in detail the representation of silicon atoms at different stages in a neural network, which provides insight into how a neural network learns to model atomic interactions.

An optimized interatomic potential for Cu-Ni alloys with the embedded-atom method.

We have developed a semi-empirical and many-body type model potential using a modified charge density profile for Cu-Ni alloys based on the embedded-atom method (EAM) formalism with an improved optimization technique. The potential is determined by fitting to experimental and first-principles data for Cu, Ni and Cu-Ni binary compounds, such as lattice constants, cohesive energies, bulk modulus, elastic constants, diatomic bond lengths and bond energies. The generated potentials were tested by computing a variety of properties of pure elements and the alloy of Cu, Ni: the melting points, alloy mixing enthalpy, lattice specific heat, equilibrium lattice structures, vacancy formation and interstitial formation energies, and various diffusion barriers on the (100) and (111) surfaces of Cu and Ni.

Energetics and atomic relaxations of Cu nanowires: the effect of local strain and cross-sectional area.

We have calculated the activation energies for several single atom and vacancy diffusion processes on Cu nanowires with the axial orientation of [Formula: see text], using the nudged elastic band technique based on the interaction potential obtained from the embedded atom method. It is shown that the dimer-initiated local strain and its relief at the transition state have a significant effect on the characteristics of self-surface diffusion mechanisms on nanowires. Contrary to the case for cylindrical multishell-type Cu nanowires, the vacancy formation energy for rectangular nanowires is maximum in the core region and is nearly zero at the corner of the nanowire. In addition, the activation energy barriers for the vacancy diffusion processes taking place in the core region are found to be higher than those occurring near the corner of the nanowire. Our calculations further show that the vacancy diffusion processes taking place near the corner of the wire are dictated by the lower coordination of the surrounding atoms. From the structural investigation of nanowires, we have also established that multilayer relaxations for rectangular nanowires with smaller cross-sectional area cannot be defined.