Symposium TC2 : Design, Discovery, and Understanding of Materials Guided by Theory, Computation and Data Mining

The known ferroelectric (FE) materials LiNbO₃ and LiTaO₃ can be regarded as derived from the A₂O₃ corundum structure with cation ordering. Here we consider more general binary (ABO₃) and ternary (A₂BB'O₆) corundum derivatives as an extended class of potential FE materials, motivated by the fact that some members of this class have recently been synthesized using high-pressure growth techniques. We first identify structure types within this class that allow polarization reversal between equivalent structures. These structures are all strongly polar with large spontaneous polarizations, but a material cannot be considered ferroelectric unless the polarization is reversible by the application of an electric field. This in turn is determined by the height of the energy barrier for domain reversal. We therefore report first-principles calculations of the energy barriers for FE reversal in a range of representative materials from this class. These are computed first for the case of coherent bulk reversal, and then also for the more realistic case of reversal via motion of 180° domain walls. We find that the pathways for domain reversal are not ones that might have naively been expected, and we use our results to identify several potentially new FE materials. We propose empirical measures that can provide a rule of thumb for estimating the barrier energies. We also discuss the conditions under which ferroelectricity is compatible with magnetic ordering, and evaluate whether members of this class are "hyperferroelectrics." We expect our results to assist in the experimental search for new FE materials in the corundum derivative family.

Gauss’s law dictates that the net electric field inside a conductor in electrostatic equilibrium is zero by effective charge screening; free carriers within a metal eliminate internal dipoles that may arise owing to asymmetric charge distributions. Quantum physics supports this view, demonstrating that delocalized electrons make a static macroscopic polarization, an ill-defined quantity in metals—it is exceedingly unusual to find a polar metal that exhibits long-range ordered dipoles owing to cooperative atomic displacements aligned from dipolar interactions as in insulating phases. Here we describe the quantum mechanical design and experimental realization of room-temperature polar metals in thinfilm ANiO₃ perovskite nickelates using a strategy based on atomicscale control of inversion-preserving (centric) displacements. We predict with ab initio calculations that cooperative polar A cation displacements are geometrically stabilized with a nonequilibrium amplitude and tilt pattern of the corner-connected NiO6 octahedra—the structural signatures of perovskites—owing to geometric constraints imposed by the underlying substrate. Heteroepitaxial thin-films grown on LaAlO₃ (111) substrates. Heteroepitaxial thin-films grown on LaAlO₃ (111) substrates fulfil the design principles. We achieve both a conducting polar monoclinic oxide that is inaccessible in compositionally identical films grown on (001) substrates, and observe a hidden, previously unreported, non-equilibrium structure in thin-film geometries [1]. We expect that the geometric stabilization approach will provide novel avenues for realizing new multifunctional materials with unusual coexisting properties.

1. T. H. Kim, et al., Nature, 533, 68 (2016)

This work has been done in collaboration with T. H. Kim, D. Puggioni, Y. Yuan, L. Xie, H. Zhou, N. Campbell, P. J. Ryan, Y. Choi, J.-W. Kim, J. R. Patzner, S. Ryu, J. P. Podkaminer, J. Irwin, Y. Ma, C. J. Fennie, M. S. Rzchowski, X. Q. Pan, V. Gopalan, J. M. Rondinelli.

Symmetry is a fundamental tool in physical sciences. An introductory course in materials science, solid-state physics, or chemistry begins with an introduction to symmetry. Distortions are ubiquitous in materials, ranging from molecular and crystal lattice vibrations, mechanical deformations, phase transitions, protein reconfigurations and others. One set of atomic configuration can rearrange to another configuration under an external stimuli, such as temperature, fields, and stresses. Representation theory is extensively used to study distortions and symmetry of distortions in materials.

In this talk, we propose a new antisymmetry operation called distortion-reversal (DR) symmetry, 1*, that reverses a distortion field. With this additional symmetry, we find that we can formulate distortion space groups and point groups (similar to magnetic groups), and ascribe them to a whole range of distortion phenomena such as the vibration modes of molecules, phase transitions, diffusion, and domain wall motion in materials with multiple order parameters. If a material exhibits distortions as well as magnetic phenomena, then both 1* and 1' (time reversal) are relevant, and we can describe the structure using double antisymmetry space groups (DASG’s). We can also ascribe distortion symmetry groups to distortions of electronic wavefunctions.

An important consequence of ascribing distortion groups to a generalized distortion is that it can predict a change in any property tensor of a material system in response to a distortion. I will discuss how the application of distortion symmetry can extend vastly from crystals and biomolecules, to metamaterials, electronic structure calculations such as Berry phase, to even generalized motion in Euclidean and curvilinear spaces. We have developed computer codes for Nudged Elastic Band module in Quantun Espresso that incorporates this symmetry and leads to significant efficiency and improved accuracy in determining minimum energy pathways of distortions. Examples of applications of this module to various materials design problems will be presented.

Reference:
(1) Vanleeuwen and Gopalan, DOI: 10.1038/ncomms9818

Real materials typically exhibit structural defects. The properties of many functional materials are often decisively controlled by these structural defects. The role of defects in oxides of mixed ionic-covalence bonds is not well understood. Specifically, the situation becomes even more complicated for prototypical ferroelectrics such as BaTiO₃, in which the perovskite structure allows for a wide variety of defects, including dipolar defects due to aliovalent doping. Using BaTiO₃ as an example, we simulate the polarization-electric field (P–E) and strain-electric field (ε–E) hysteresis loops in the presence of defect dipoles with molecular dynamics and find that a small concentration ( < 0.8%) of dipolar defects can dramatically influence the electro-mechanical properties of ferroelectrics. We demonstrate that despite the complex coupling between local electrostatic/elastic fields of defects and intrinsic ferroelectric polarization, the macroscopic switching behavior of ferroelectrics, such as aging effect and reversible giant electro-strain coupling, can be understood by the interplay between the local electric field from the defect dipoles, the intrinsic bulk polarization and the orientation of external driving field.

The application of ferroelectric materials in solar cells has a long history, dating back to the earliest photovoltaic devices. Now there is a resurgence of interest in these materials owing to their central role in recent efficiency breakthroughs. In ferroelectrics, the internal electric fields are thought to be an important ingredient that allows for efficient separation of the photogenerated charges. Most notably, hybrid halid perovskites have demonstrated remarkable efficiency gains with current values surpassing 22%. However, this record-breaking class of material is not without its shortcomings: toxicity and device stability are two major obstacles to commercial adoption. To overcome this bottleneck, a promising research strategy is to look for new materials with similar desirable properties but none of the drawbacks. The V-VI-VII chalcohalides have this potential [1,2].

The project is a major collabarative effort and this talk will focus on the high-level electronic structure calculations using quasi-particle self-consistent GW theory (QSGW), within an all-electron framework. QSGW is very much state-of-the-art for describing semiconductor band structures, providing quantitative gaps and accurate dispersion in cases where DFT fails. The results of a screening study will be presented, highlighting promising candidates and interesting band structure features, such as suitable direct/indirect band gap combinations and more exotic band splitting resulting from the interplay of spin-orbit coupling and internal electric fields (Rashba and Dresselhaus effects). Such effects are particularly relevant for photovoltaics, as recent work has shown that a spin-split indirect gap can result in a radiative recombination rate that is hundreds of times lower [3].

References
[1] K. T. Butler, J. M. Frost, and A. Walsh, Energy Environ. Sci. 8, 838 (2015)
[2] K. T. Butler, S. McKechnie, P. Azarhoosh, M. van Schilfgaarde, D. O. Scanlon and A. Walsh, Appl. Phys. Lett. 108, 112103 (2016)
[3] P. Azarhoosh, J. M. Frost, S. McKechnie, A. Walsh and M. van Schilfgaarde, ArXiv: 1604.04500

ABO₃ perovskite-oxide ferroelectrics are well known for their useful functional properties. These materials, as well as their solid solutions, exhibit rich phase diagrams that can be exploited, e.g., to obtain large piezoelectric and dielectric responses. Mesoscale-level investigations of their behavior usually utilize Landau phenomenological theory, where the system energy functional is represented by a polynomial expansion in powers of polarization and strain that is parameterized from experimental data. In this project, we present an approach for fitting the Landau functionals for perovskite ferroelectrics directly from first principles simulations with the help of statistical and machine learning tools. Initial data sets are created by computing the energies for a wide range of possible structural configurations involving polar and elastic distortions with standard density-functional theory (DFT) codes. A small fraction of this data is then processed by supervised machine learning algorithms to train a Landau-style polynomial model that can predict the system energies to within 20 meV of the DFT results.
We have also employed the insights and models obtained from the supervised machine learning studies for “on-the-fly” energy evaluations in Wang-Landau¹ Monte Carlo simulations of perovskite ferroelectrics for the investigation of thermodynamic properties. Specifically, the trained models were used to reject unfavorable structural configurations, that otherwise would have been calculated by much more expensive DFT computations. This has the potential to reduce the number of DFT calculations, resulting in a substantial speed-up of Monte Carlo simulations.
KCP and SMN are thankful to the National Science Foundation (DMR 1309114) for partial funding of this project. KCP also acknowledges the support from the Advanced Short Term Research Opportunity (ASTRO) program at ORNL. SFY, ME, and VRC were supported by the U.S. DOE, Office of Science, BES, MSED and the Office of Science Early Career Research Program (VRC). YWL was sponsored by the U.S. DOE, Office of Advanced Scientific Computing Research. The authors gratefully acknowledge the computational resources provided by NERSC, and the Oak Ridge Leadership Computing Facility (OLCF), which is supported by the Office of Science of the U.S. DOE.

References:
1 F. Wang and D. Landau, Phys. Rev. Lett. 86, 2050 (2001).

Computational material science is beginning to amass consistent data sets large enough to analyze via big data machine learning algorithms. Unlike other fields, this effort is stymied by the lack of adequate descriptors that can map to a plethora of properties that material scientists are interested in. Dielectric and piezoelectric response are important classes of materials in an increasingly electronic society that could greatly benefit from an accelerated discovery process. We present our recent efforts in understanding the performance of well-behaved structural and composition descriptors as well as evaluating novel descriptors based on distortions and connectivity. We will explore promising classes of piezoelectrics and dielectrics such as ordered anion compounds and hybrid organic perovskites to elucidate the physics of strong polarization response. This generalized feature set will then be projected into unknown material space to identify potential materials for ab initio calculation and possible experimental review.

Rapid devlopment of new materials requires an ability to predict conditions of growth of compounds. Here, I will explain a two-part approach towards developing a framework that can greatly accelerate the development of new materials using the example of pulsed laser deposition (PLD). PLD has become pivotal in nanoscience research, allowing fabrication of epitaxial thin films that have been used to explore eg. superconductivity in the cuprates [1], colossal magnetoresistance in manganites [2], and enhanced magnetoelectric behavior in ferroics [3]. However, despite the considerable advances made in understanding film growth by PLD, the technique is limited by lack of reproducibility and inability to make deterministic predictions apriori on final film properties given a set of growth conditions, and therefore limits PLD to a (relatively) slow, purely research-oriented regime without ability to scale for industrial application. For the framework, the first step involves mining of extant literature on PLD of oxides, to semi-automatically determine the growth conditions and final film properties of a select group of commonly studied oxides such as La-based manganites and ferroelectric thin films. Crowd sourcing is then used to match and assign the automatically identified parameters to the particular films being grown, with results feeding into a centralized database. Subsequently, data mining and machine learning techniques can be applied to determine the relevant correlations between growth parameters and the final film properties, and therefore predict conditions for similar or related compounds. In parallel, optimization of growth parameters in-operando requires information on plume dynamics [4] as well as film thickness and roughness (in-situ reflection high-energy electron diffraction). I will discuss strategies for combining these channels of information into developing a complete framework for understanding, predicting and optimizing PLD depositions, through machine learning/computer vision approaches. These strategies require community-based input, open source collaborations and will firmly plant PLD onto the grounds of big data science.

Research sponsored by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U. S. Department of Energy (SEED #8203, RVK, SVK, SY, RP). This research was sponsored by the Division of Materials Sciences and Engineering, BES, DOE (SVK, RKV). A portion of this research was supported (APB) and conducted at the Center for Nanophase Materials Sciences, which is a DOE office of Science User Facility.
References
[1] T. Venkatesan et al., App. Phys. Lett. 52 (14), 1193 (1988).
[2] S. Jin et al., Science 264 (5157), 413 (1994).
[3] J. Wang et al., Science 299 (5613), 1719 (2003).
[4] R. Wood et al., Phys. Rev. Lett. 79 (8), 1571 (1997).

Thermodynamic forces such as stress and temperature can alter the free energy landscape of the electronic charge carriers in semiconductors. By design, these thermodynamic forces can be tuned to promote a desired form of charge carriers. For example, the electronic conductivity and chemical expansion of metal oxides depend on the extent of the wave function spread of the electronic charge carriers. Delocalized (free) electrons generally lead to fast electronic conduction, whereas self-trapped (small polaron) electrons are thought to cause large chemical expansion in metal oxides. Here we adopt SrTiO₃ as a model system and computationally predict regimes of trapping versus delocalization for electrons as a function of hydrostatic stress and temperature. SrTiO₃ is a model system for the versatile perovskite family of metal oxides, and itself is a widely used crystal for electronic structure studies. Starting from density functional theory, we parametrize on-site Hubbard terms (+U) for Ti 3d states and O 2p states that adhere to the generalized Koopman’s theorem. This adherence qualifies density functional theory (DFT) equipped with the selected U to accurately describe trapped electrons and holes. Next we compute a temperature-hydrostatic stress predominance diagram for electron defects in the dilute limit in SrTiO₃ by combining density functional theory and the quasiharmonic approximation. At zero stress, free electrons are dominant and show no transition from large to small polaron behavior at half Debye temperature, consistent with experiments. Importantly, we discovered two regimes where the small polarons dominate – under tensile stress at low temperature, and under compressive stress at high temperature. The dominance of small polarons in these two regimes is due to the competition between the relative relaxation volumes of the small polaron and of the free electron as a function of temperature. Upon finite isothermal stretching, intriguingly the free electron induces a larger relaxation volume at high temperatures due its ability to generate more vibrational entropy compared to the small polaron. This is counterintuitive to the common understanding that charge localization leads to larger relaxation volumes. The analyses presented here provide the underpinnings of designing optimal thermodynamic conditions that can promote the desired form of electronic defects in metal oxides.

This lecture will discuss possibilities to discover and design new permanent magnets by employing computational high-throughput-screening and multi-scale-modeling approaches in order to substitute established magnetic materials like Nd₂Fe₁₄B, which have outstanding functionalities but also constraining criticalities. To discover promising magnetic phases, a combinatorial high-throughput-screening approach based on density functional theory (DFT) is employed to search for crystal structures and chemical compositions of intermetallic phases of transition-metal and rare-earth (RE) elements, which have good intrinsic ferromagnetic properties but contain less amounts of critical RE elements than Nd₂Fe₁₄B.
The development of permanent magnets from promising magnetic phases requires efficient scale-bridging approaches in order to take into account how the microstructure influences the magnetic behavior. The scaling-up in size of atomistic models can be achieved by employing efficient tight-binding total-energy models and bond-order potentials. Such approaches are derived from DFT data for magnetic iron phases, pure or with interstitial hydrogen or carbon, and they are capable to describe extended defects like grain boundaries or lattice dislocations.
A further route to multi-domains and poly-crystals is provided by phenomenological approaches like micro-magnetic or phase-field models, which can be parameterized as well using data bases of materials properties obtained from the atomic scale. The lecture will close with an outlook towards this direction.

We show that combining state-of-the-art computer simulations in the framework of the electronic structure theory and advanced experimental techniques allow one to investigate materials properties at the most fundamental level, and to transfer the knowledge to the broad and interdisciplinary research community. Moreover, the successful development of novel tools for the basic research gives the theory sufficient predictive power for the knowledge-based materials design. With modern computational tools at hand, we investigate and identify novel materials and exciting phenomena with strategic potential for future technological applications. We illustrate the capability of theoretical simulations in studies of properties of matter at extreme compressions, up to several millions of atmospheres [1] and show their relevance for understanding of properties of the Earth’s core [2]. Then we demonstrate the use of theory for knowledge-based design of novel materials, superhard coatings [3], isotope enriched silicon carbide with enhanced thermal conductivity for energy applications [4], alloys for spintronics [5], etc. We realize that time it takes to discover advanced materials and to prove their usefulness to a commercial market is far too long at present. There is a need to reduce it significantly. We argue that computer simulations coupled to high-quality experiment is an important tool to achieving this goal.

[1] L. Dubrovinsky, N. Dubrovinskaia, E. Bykova, M. Bykov, V. Prakapenka, C. Prescher, K. Glazyrin, H.-P. Liermann, M. Hanfland, M. Ekholm, Q. Feng, L. V. Pourovskii, M. I. Katsnelson, J. M. Wills, and I. A. Abrikosov, Nature 525, 226–229 (2015).

[2] L. V. Pourovskii, J. Mravlje, A. Georges, S. I. Simak, I. A. Abrikosov, “Fermi-liquid behavior and thermal conductivity of ε-iron at Earth's core conditions”, arXiv:1603.02287 [cond-mat.str-el].

[3] H. Lind, R. Forsén, B. Alling, N. Ghafoor, F. Tasnadi, M.P. Johansson, I. A. Abrikosov, and M. Odén, Appl. Phys. Lett. 99, 091903 (2011); H. Lind et al., “Coated cutting tool insert”, Application No/Patent No 11784719.4 – 1353, Priority EP/23.11.10/EPA 10192235

[4] O. Kordina, E. Janzén, O. Hellman, and I. A. Abrikosov, Patent application SE1330084-3

[5] A. S. Ingason, A. Mockute, M. Dahlqvist, F. Magnus, S. Olafsson, U. B. Arnalds, B. Alling, I.A. Abrikosov, B. Hjörvarsson, P. O. Å. Persson, and J. Rosen, Phys. Rev. Lett. 110, 195502 (2013).

Abstract not available.

The last few decades have seen significant advancements in materials research tools, allowing researchers to rapidly synthesis and characterize large numbers of samples - a major step toward high-throughput materials discovery. Data analysis advancements originating in machine learning also allow for more rapid conversion of the large amounts of collected data into actionable knowledge. Active learning, a branch of machine learning, promises to push high throughput materials research to the next level - autonomous materials exploration. Active learning allows the researcher to take a step back, and gives the choice of the optimal next experiment to perform to the algorithm. This in turn can potentially reduce tedious labor hours, reduce equipment hours, and accelerate materials exploration. In this talk we demonstrate the use of active learning to autonomously control X-ray diffraction systems in the lab and at the beamline for phase diagram determination from composition spreads. Materials of interest include Fe-Ga-Pd, TiO2-SnO2-ZnO, and Mn-Ni-Ge.

High-temperature shape memory alloys have applications in the automotive, aviation and biomedical sectors. The martensitic transformation temperature (M_s) and mechanical properties of these materials often exhibit a strong dependence on the chemical composition. Insight into the underlying mechanisms and phase stability of the competing austenite and martensite phases can help to device new alloys with the desired functional properties at elevated temperatures.
The free energy is a key quantity for determining the composition dependent transformation temperature between the involved crystal phases. We approximate the different contributions to the free energy as a function of composition based on density functional theory calculations and compare the stability of competing phases at finite temperature. In addition to the thermodynamic phase stability we investigate the mobility of the different elements within the alloy which is key to analyze segregation and redistribution at elevated temperatures.
We find that the 0K energy difference and the Debye temperature difference between the involved phases are the critical parameters for predicting the composition dependence of M_s. From our calculations we identify a one dimensional descriptor for estimating M_s that may be used in high-throughput screening of ternary or multicomponent alloys for a computationally guided development of novel high-temperature shape memory alloys.

Increased fuel efficiencies in automobiles can be achieved by making structural components lighter using alloys containing magnesium. Additions of rare-earth elements (RE) such as neodymium, yttrium and gadolinium have been shown to improve the mechanical properties of these alloys due to the formation of high-strength precipitates during the aging heat treatment. However, the crystal structure and thermodynamic stability of these metastable precipitates has remained an open question. In this talk, we will describe how ab-initio calculations combined with inputs from experiments revealed the stability of a unique hierarchy of RE orderings (β''') that emerge from the disordered solid solution during aging heat treatments. The elevated temperature properties of this hierarchy of orderings was studied with cluster expansion methods and Monte Carlo simulations. High-throughput first-principles calculations were used to explore the stability of similar hierarchies in other Mg-RE alloy systems. These studies showed that the crystal structures and morphologies of metastable precipitates in all Mg-RE alloys are very sensitive to misfit strains. The first-principles results of this study enable the formulation of design principles with which the properties of Mg-RE alloys can be further improved through precipitation strengthening.

Impurity (dilute) diffusion coefficients are the foundation of diffusion (mobility) databases that are essential for kinetic modeling of materials processes such as solidification, precipitation and creep. Impurity diffusion coefficients are traditionally measured using tracer experiments which are laborious, time-consuming and expensive; and thus only a few laboratories in the world are still performing such measurements. A new forward-simulation method allows very reliable extraction of impurity diffusion coefficients from regular diffusion couple profiles. This method together with high-efficiency diffusion multiples as well as an elegant liquid-solid diffusion couple geometry provides a new effective way to gather reliable experimental impurity diffusion coefficients. Examples of many alloy (Ni, Ti, Mg) systems will be used to illustrate reliability and effectiveness of the new method.

While high-throughput methods are commonly used for the development of functional materials, structural materials are still being developed by conventional processes and predictive paradigms. The main difference of both material classes is that mechanical properties are mainly governed by the interaction of microstructural elements such as grains, grain boundaries or precipitations, which can be adjusted by mechanical, thermal or thermo-mechanical treatments.
The collaborative research center 'Farbige Zustände' is developing a novel high-throughput method for the exploration of evolutionary structural materials. This requires new processes for the synthesis of samples, since thin film processes commonly used in high throughput processes are not capable of representing microstructures without emphasizing boundary effects. Furthermore, downscaled processes for thermal, mechanical and thermo-mechanical treatment are used to develop different microstructures.
Conventional material characterizations such as tensile testing are replaced by high-throughput characterizations, which result in descriptors. Such descriptors can be a simple deformation ratio achieved by mechanical treatment, or of more complex nature such as specific peak shapes in an XRD analysis. These descriptors are combined mathematical approaches to find a predictor function which maps descriptors to conventional material properties. Also, machine learning techniques will be used to identify possibly new mechanisms.
The method is expected to produce large amounts of data which can also be used for the development and validation of computer models for the evolution of microstructures. The initiative will provide a database with validated experimental data to interested researchers.

In materials science, structure is typically described using atoms as fundamental units and the properties are inferred through the spatial arrangements of atoms relative to each other. The length scales involved may vary from short-range (near-neighbors) and medium range to long-range length scales depending on the structure and properties of interest. Even when the analysis of structure at larger length scales is necessary, the characterization of the relative arrangement of atoms in the first coordination shell has proven to be of great importance. However, the description (or quantification) of this local structure is non-trivial and the automated analysis of such descriptors for large sets of data obtained through atomistic simulations is lacking.

In this talk, I will present a direct point-pattern matching technique for quantifying local atomistic structures. Point pattern matching (PPM) is a fundamental problem in pattern recognition with applications in a broad range of fields such as computer vision, computational chemistry, astronomy and computational biology. Generally, PPM algorithms are utilized to obtain a mapping between elements of two point sets such that the two sets differ only through rigid body transformations (rotations and translations) or jitter. These attributes, and in particular the fact that PPM algorithms account for small perturbations that arise due to thermal vibrations, make it ideal for the analysis of local atomic structure in disordered material systems. Another important aspect of the PPM technique is the fact that a metric quantifying the similarity (or dissimilarity) between the atomic environments may be easily computed. This allows for a better understanding of how structure varies, for example, as temperature is increased. Most notably, a metric will help facilitate the utility of unsupervised machine learning algorithms (such as clustering) in analyzing the underlying geometries in various material systems.

Mass transport controls crucial materials processing, such as segregation and precipitation, and properties, such as ionic conductivity, in a wide variety of materials. First-principles methods can determine the activated state energies at the atomic level involved in mass-transport such as vacancies moving in a crystal. Upscaling from activation barriers to mesoscale mobilities requires the solution of the master equation for diffusivity. For all but the simplest cases of interstitial diffusivity, and particular approximations with vacancy-mediated diffusion on simple lattices, calculating diffusivity directly is a challenge. This leaves two choices: uncontrolled approximations to map the problem onto a simpler (solved) problem, or a stochastic method like kinetic Monte Carlo, which can be difficult to converge for cases of strong correlations. Moreover, without analytic or semi-analytic solutions, evaluating derivatives of transport coefficients is also difficult. We describe and demonstrate the development of direct and automated Green function solutions for transport that take full advantage of crystal symmetry. To showcase the new functionality, we focus on magnesium alloys containing Al, Zn, and rare earth elements (Gd, Y, Nd, Ce and La), where current theoretical models to predict diffusivity from atomic jump frequencies make uncontrolled approximations that affect their accuracy. Density-functional theory identifies nine different solute-vacancy configurations from which symmetry analysis determine 17 transitions states corresponding to a 27-frequency model. Our Green function approach computes diffusivity for 14 solutes using the density-functional theory data. We find significant differences for solute drag of Al, Zn, and rare earth solutes, and improved predictions of activation energies for diffusion. The differences with prior predictions can be directly attributed to missing jumps in the 8- and 13-frequency models. The underlying automation also makes the extension of first-principles transport databases significantly more practical and eliminate uncontrolled approximations in the transport model.

Numerical studies on the kinetic and topological features of grain growth have been carried out for decades. However, there has not been a conclusive consensus even for the simplest ideal grain growth, which is mainly due to limitations in the computational accuracy of the grain-growth models and computer resources that have been employed to date. To address these issues, in this study, we utilize the multi-phase-field (MPF) model [1–3] that can simulate the curvature-driven migration of grain boundaries with high accuracy. In addition, the MPF simulations are dramatically accelerated via parallel computing on graphics processing units (GPUs); this is done by implementing the message passing interface (MPI) in a GPU code developed in our previous study [4]. By means of the MPF model and the parallel GPU code, we perform a series of ultra-large-scale simulations on two-dimensional and three-dimensional ideal grain growth, through which the growth rate and topology of grains in the ideal growth processes are elucidated in detail.
[1] I. Steinbach, F. Pezzolla, Physica D 134 (1999) 385–393.
[2] E. Miyoshi, T. Takaki, Comput. Mater. Sci. 112 A (2016) 44–51.
[3] E. Miyoshi, T. Takaki, Comput. Mater. Sci. 120 (2016) 77–83.
[4] E. Miyoshi, T. Takaki, (2016) submitted.

We have performed large-scale molecular dynamics (MD) simulations [1] up to 12 million atoms to discuss atomic nature in solidification [2,3] and subsequent grain growth [4] of pure iron systems. We revealed that there exist some amount of atoms with icosahedral configuration in the undercooled iron melt and these atoms increase with decreasing temperature. It is expected that accumulation of atoms with icosahedral configuration in the initial beta-relaxation regime of nucleation is the key to initiate the formation of bcc phase. Moreover, the Avrami exponents during solidification are approximately estimated to be close to 3 and 4 in two- and three-dimensional grain growths, respectively [3]. We estimated the grain growth exponent during the microstructure evolution after the solidification to be about 0.3, which is smaller than the ideal values during normal grain growth, 0.5. It is due to the anisotropy in the grain boundary energy and the kinetic coefficient is inherent during the MD simulation without any artificial parameter. Moreover, our trial to multi-GPU computation will be also introduced with larger systems.
[1] Y. Shibuta, M. Ohno, T. Takaki, JOM, 67, 1793-1804 (2015).
[2] Y. Shibuta, K. Oguchi, T. Takaki, M. Ohno: Scientific Reports, 5, 13534 (2015).
[3] Y. Shibuta, S. Sakane, T. Takaki, M. Ohno: Acta Materialia, 105, 328-337 (2016).
[4] S. Okita, W. Verestek, S. Sakane, T. Takaki, M. Ohno, Y. Shibuta, submitted.

The chalcopyrite Cu(In,Ga)Se₂ alloy is today a rather well-established compound for thin film solar cells. Emerging Cu-based materials are explored to benefit from the energetically high-lying Cu d-state in combination with low effective mass of the minority carriers [1,2,3]. In the present study, we explore the details in the optical properties of emerging Cu-based compounds, like for instance Cu₂ZnSn(S,Se)₄, Cu₂SnS₃, Cu(Sb,Bi)(S,Se)₂ [4,5], and Cu₃(Sb,Bi)(S,Se)₃ [4,5], employing the Heyd-Scuseria-Ernzerhof (HSE06) hybrid functional. We calculate the electronic structure and we analyze the optical properties in terms of the dielectric functions and absorption coefficients. By modeling the quantum efficiency of the compounds, we further discuss the optical response. The results help to understand fundamental physics of then emerging Cu-based materials in order to design and optimize solar cell devices.

[1] R. Chen and C. Persson, J. Appl. Phys. 112, 103708 (2012).
[2] R. Chen and C. Persson Thin Solid Films 519, 7503 (2011).
[3] S.G. Choi, et al., Appl. Phys. Lett. 101, 261903 (2012).
[4] M. Kumar and C. Persson, Appl. Phys. Lett. 102, 062109 (2013).
[5] M. Kumar and C. Persson, Energy Procedia 44, 176 (2014).

Electrodeposition of metals into the long and narrow pores of porous templates such as anodized aluminum oxide is an efficient method for fabricating dense arrays of nanowires, which have applications in a wide variety of devices such as thermoelectrics, sensors, and high-density magnetic data storage devices. Uniform nanowire lengths are desirable for many applications but difficult to achieve because often, a small fraction of nanowires reach the end of the template first and grow radially outwards, blocking and preventing growth in neighboring pores, resulting in an overall non-uniform nanowire array with many short nanowires that do not span the template.

To gain a better understanding of template-assisted electrodeposition, we develop a simple sharp-interface model and corresponding phase-field model for electrodeposition from a well-supported electrolyte into a porous template. Our model accounts for the electrode-electrolyte interfacial energy, ion transport via diffusion, and Butler-Volmer reaction kinetics. We perform two- and three-dimensional simulations of nanowire growth, including subsequent overflowing of the template to form hemispherical caps, and observe deposit morphologies and current evolution profiles that resemble those found experimentally. We propose that variations in pore cross-sectional area over the length of the pore can have a significant effect on growth rates and can thus contribute to non-uniformity in nanowire lengths. Another mechanism of reduced nanowire length uniformity that we propose is the presence of side reactions. Finally, we perform phase-field simulations to study the fraction of pore width that is filled for various electrodeposition parameters.

Intense research in renewable energy is stimulated by global demand of electric energy. Electrochemical energy storage and conversion systems namely, supercapacitors and batteries, represent the most efficient and environmentally benign technologies. Moreover, controlled nanoscaled architectures and surface chemistry of electrochemical electrode materials is enabling emergent next-generation devices approaching theoretical limit of energy and power densities and deliver electrical energy rapidly and efficiently. In this work we develop graphene-inorganic hybrid assembly highlighting the impacts of nanoscale internal microstructure producing tailored interfaces for improved electrochemical and electroanalytical properties. Molecular electrodeposition and facile hydrothermal synthesis techniques followed by thermal treatment are demonstrated to be effective approaches for nanoengineered electrochemical electrodes. The electrode assembly consists of supercapacitive graphene nanosheets and pseudocapacitive nanostructured transition metal oxides (TMeOs) synthesized on two- and three-dimensional graphene nanosheets facilitate chemically bridged (covalently and electrostatically anchored) yet tunable graphene-TMeO interfaces. The intrinsic microstructure and surface of these hybrids were characterized by electron microscopy combined with elemental mapping, X-ray diffraction and Raman spectroscopy. The electrochemical properties are investigated as asymmetric hybrid supercapacitors. We demonstrated that hybrids show improved electrochemical performance as compared with constituents by themselves. We attribute the remarkable findings due to interplay of (a) open pore system beneficial to ion diffusion and transport kinetics owing to larger accessible geometric surface area, (b) three-dimensional topologically multiplexed and highly conductive pathways provided by multilayer graphene, electrochemically reduced graphene oxide and hydrothermal processed reduced graphene oxide nanoscaffold architectures to ensure rapid charge transfer and electron/ion conduction (< 10 ms), and (c) synergistic integration of functional nanomaterials devoid of graphene sheets agglomeration with optimal transition metal (oxides) nanoparticles loading. Computational simulations via periodic density functional theory (DFT) with and without transition metal adatoms on graphene and graphene oxide sheets are performed. These calculations determine the total and hybridized partial electronic density of states (DOS) in the vicinity of Fermi level thereby complementing and synergizing experimental work in terms of various contributions toward surface/interfacial charge transfer sites on heterogeneous electrodes, electron/ion transport, pseudocapacitive and electric double layers.

Simulating dendrite growth with natural convection is great challenging, because there are large scale differences between dendrite and liquid flow. Although phase-field method is the most powerful numerical tool to simulate dendrite growth during solidification, the phase-field simulation needs large computational cost. Recently, we have developed a parallel GPU computational scheme to accelerate a very-large-scale phase-field simulation of the directional solidification of a binary alloy [1-4]. In this study, we try to apply the parallel GPU computation to the dendrite growth with natural convection. Here, we use a coupling model of phase-field model and lattice Boltzmann model [5]. By performing the large-scale simulations using the GPU supercomputer TSUBAME2.5 at Tokyo Institute of Technology, we investigate the effects of natural convection on the dendrite morphology formed during directional solidification of a polycrystalline binary alloy.
[1] Y. Shibuta, M. Ohno, and T. Takaki: JOM, 67, 1793 (2015).
[2] T. Takaki, T. Shimokawabe, M. Ohno, A. Yamanaka, and T. Aoki: J. Cryst. Growth, 382, 21 (2013).
[3] T. Takaki, M. Ohno, T. Shimokawabe, and T. Aoki: Acta Mater., 81, 272 (2014).
[4] T. Takaki, M. Ohno, Y. Shibuta, S. Sakane, T. Shimokawabe, and T. Aoki: J. Cryst. Growth, 442, 14 (2016).
[5] R. Rojas, T. Takaki, and M. Ohno: J. Comput. Phys., 298, 29 (2015).

Quantitative phase-field model has been increasingly utilized for simulation of microstructural evolutions during alloy solidification processes [1,2] since it can be exactly mapped onto the free-boundary problem [3]. However, the quantitative model is a non-variational model in that the correction term called antitrapping current is phenomenologically introduced into the diffusion equation. Hence, its theoretical basis has not been well established as a method of nonequilibrium thermodynamics. In this study, we put forward a way of variational formulation of quantitative phase-field model for alloy solidification processes on the basis of two-phase approach. It is demonstrated that the antitrapping current naturally emerges in the present variational formulation [4]. In addition, we systematically investigated numerical efficiencies of the quantitative phase-field models constructed with different polynomials of phase-field, focusing on the dendritic growth in a binary alloy.

[1] Y. Shibuta, M. Ohno, and T. Takaki: JOM, 67, 1793 (2015).
[2] T. Takaki, M. Ohno, Y. Shibuta, S. Sakane, T. Shimokawabe, T. Aoki: J. Cryst. Growth, 442, 14 (2016).
[3] M. Ohno and K. Matsuura: Phys. Rev. E 79, 031603 (2009).
[4] M. Ohno, T. Takaki and Y. Shibuta: Phys. Rev. E 93, 012802 (2016).

Recently, the possibility to add the spin degree of freedom to phase change materials (PCMs) has been explored from the theoretical and experimental sides. In principle, such a ferromagnetic PCM could open the way to realize a new multi-value memory, where the magnetic state as well as the electronic resistivity can be controlled by switching between the two phases. From the point of view of practical applications, the most crucial requirement is that the ferromagnetic (T_C) is higher than room temperature. Li and Zhang et al. performed density functional theory (DFT) calculations and ab initio molecular dynamics simulations for crystalline and amorphous Ge₂Sb₂Te₅ (GST) doped with several 3d transition metal (TM) impurities, including Cr, Mn, Fe, Co and Ni. They focused on defect formation energies, electronic structures and magnitude of local magnetic moments in these systems. However, it has not yet been elucidated what the stable magnetic structures are, nor how they depend on the type of 3d dopant.
The purpose of this work is to address this crucial issue. For this purpose, we study the electronic structure and the magnetic properties of GST doped with V, Cr, Mn and Fe. We understand the exchange mechanisms responsible for their magnetic order by calculating the exchange coupling constants (J_ij) between the TM impurities. This enables us to determine the stable magnetic state for each of the 4 dopants. Moreover, we estimate T_C for the systems showing ferromagnetic behavior by Monte Carlo simulations. Both the supercell method and the coherent potential approximation (CPA) are employed to describe this complex substitutionally disordered system. As regards the first approach, we consider a large unit cell containing 1000 sites to model the random distribution of the cations and of the impurities in doped cubic GST. Such a large-scale electronic structure calculation is performed using the program KKRnano, where the full potential screened Korringa-Kohn-Rostoker Green's function method is optimized by a massively parallel linear scaling (order-N) all electron algorithm.
We find that ferromagnetic states are favorable in the cases of V and Cr doping, due to the double exchange mechanism, whereas antiferromagnetic superexchange interactions appear to be dominant for Fe- and Mn-doped GST. The ferromagnetic interaction is particularly strong in the case of Cr. As a result, high T_C close to room temperatures are obtained for large Cr concentrations of 15 %

Bismuth, besides being a semimetal, displays puzzling superconducting properties. In its crystalline equilibrium phase, it does not seem to display superconductivity at accessible low temperatures. However, in the amorphous phase it displays superconductivity at ~ 6 K. Under pressure bismuth has been found to superconduct at 2.55 GPa (Bi-II monoclinic crystalline phase), 2.7 GPa (Bi-III tetragonal phase) and 7.7 GPa (Bi-V body centered cubic phase), having superconducting transition temperatures of Tc = 3.9 K, 7.2 K and 8.3 K, respectively. So the question is: what electronic or vibrational changes occur that explains this radical transformation in the conducting behavior of this material? In a recent publication¹ we argue that changes in the density of electronic and vibrational states may account for the behavior in the amorphous phase. Now we have undertaken an ab initio computational study of the effects of structural changes when crystalline bismuth is subjected to pressure. In order to see the effect of pressure alone we maintain the original crystalline structure and compress our sample, a 64–atom supercell, between 0 and 10 GPa; we then calculate the density of electron and phonon states. We shall present a detailed discussion and infer how relevant the observed changes are to account for superconductivity.

1. Mata-Pinzón Z., Valladares A. A., Valladares R. M., Valladares A. (2016) Superconductivity in Bismuth. A New Look at an Old Problem. PLoS ONE 11(1): e0147645. doi:10.1371/journal.pone.0147645

Many novel materials used in clean energy applications such as lithium iron phosphate (Li_xFePO₄), hematite, and certain perovskites are known to exhibit polaronic behavior. Being transition metal oxides, the strongly correlated interaction of the d shell electrons opens a gap and localizes conduction electrons into polaronic states, leading to the hopping conduction observed in these materials. To model electronic conductivities from a theoretical point of view, it is therefore necessary to calculate activation energies of such polarons. We study the effects of the electronic relaxation of semicore levels on polaron activation energies and dynamics. Within the framework of adiabatic ab initio theory, we utilize both static transition state theory and molecular dynamics methods for an in-depth study of polaronic hopping in our model system of Li_xFePO₄. Our results show that electronic relaxation of semicore states is significant in Li_xFePO₄, resulting in a lower activation barrier and kinetics that are one to two orders faster compared with calculations that do not incorporate semi-core states. In general, the results suggest that the relaxation of states far below the Fermi energy could dramatically impact upon ab initio polaronic barrier estimates for many transition metal oxides and phosphates.

Compared with cocrystal coformers, an explosive cocrystal has distinctive packing arrangement and complex intermolecular interactions. Identifying the spectral signatures of an explosive cocrystal and understanding the molecular low-frequency modes by means of the spectrum in the terahertz (THz) range are of great worth to the explicit mechanism of cocrystal formation. In this work, on the basis of the joint molecular dynamics (MD) simulations and solid-state density functional theory (DFT) calculations, we have investigated the THz absorption spectra of the CL-20/TNT cocrystal and its different directions as well as cocrystal coformers and determined the systematic and all-sided assignments of corresponding THz vibration modes. The THz spectral comparison of the cocrystal with different directions and the cocrystal coformers indicates that the CL-20/TNT cocrystal has five fresh low-frequency absorption features as unique and discernible peaks for identification, in which 0.25, 0.73, and 0.87 THz are attributed to intensive crystalline vibrations; 0.87 THz is also caused by C−H…O hydrogen-bonding bending vibrations; 1.60 and 1.85 THz features originate from C−H…O hydrogenbond stretching vibrations. In addition, the THz spectrum of the (120) crystal face has been used to conclude that the 0.07 THz peak is a symbol of the CL-20−TNT π−π stacking interaction. Also, CL-20−CL-20 interactions and C−H…O hydrogen bonding in the CL-20/TNT cocrystal have been further verified by means of the THz spectral features of (001) and (010) crystal faces. The THz spectra comparison of ε - and β-CL-20 crystals as well as (001) crystal face indicates that the CL-20 molecule conformation in the CL-20/TNT cocrystal is the same as that in β-CL-20 other than in ε -CL-20. All these inspections embody the superiority of THz technology in the characterization of energetic cocrystals, especially the weak intermolecular interactions.

Poly(vinylidene fluoride) (PVDF) is one of the best piezoelectric polymers owing to its net monomeric (CH₂CF₂) dipole moments well aligned in its β-phase crystals. Various efforts such as cold drawing, stretching, poling, copolymerization and inclusion of additives have been applied to maximize its piezoelectric coefficient by transforming the inherently-amorphous PVDF into the β-phase crystals, but the improvement has been limited due to the lack of systematic molecular-level understanding of the effect of the shear stress on the crystallization of PVDF. In the current study we therefore carry out non-equilibrium molecular dynamics simulations to virtually mimic the crystallization of amorphous PVDF under various amounts of shear stress. It is interesting to find that the β-phase crystallinity, the polarization, and the net dipole are maximized at a critical value of shear velocity rather than increasing monotonically with shear velocity. The presence of the critical shear velocity and its predicted value are in a good agreement the experimental observation (Org. Electron. 28, 67-72, 2016). Our simulations also predict a shear-induced development of various domains showing parallel, anti-parallel, and random alignments of dipoles.

The interest in calcium carbonate, CaCO₃ - one of the most abundant biominerals on the Earth - is growing in many branches of science, from biomedicine to environmental applications [1]. The formation of calcite, the most stable polymorph of CaCO₃, is either the main cause of problems in the water treatment process and a mechanism favored to act as a pH neutralizer in water filter systems. The inhibition of calcite growth, in industrial processes, occurs by using chemical additives often severely impacting the environment.
Organic molecules in aqueous solutions are good candidates in the inhibition of some biogenic crystals growth. The formic acid HCOOH is considered to investigate at atomic level the interaction between the carboxyl functional group -COO^- and the {10.4} hydrated surface of calcium carbonate, CaCO₃, in the form of calcite. Ab initio simulations based on the density functional theory are performed to study the adsorption of undissociated and dehydrogenated HCOOH in presence of water. Relevant adsorption energies obtained for HCOO^-+H2O on calcite predict that water is essential in the stabilization of the carboxyl group in its deprotonated form. The interfacial properties and the trend of adsorption energies for different coverages are given in details. The dissociation barriers of HCOOH on hydrated calcite are evaluated with the climbing-image nudged elastic band (CI-NEB) method.

[1] Meldrum, F.; Cölfen, H. Controlling Mineral Morphologies and Structures in Biological and Synthetic Systems. Chem. Rev. 2008, 108, 4332–4432

Solid state electrolytes (SSEs) with high Li conductivity can significantly improve Li ion accumulators in terms of electrochemical efficiency, thermal and mechanical stability, and environmental compatibility, leading to an enhanced range of applications for these high energy density batteries.
Compounds crystallizing in the structure of NaZr₂(PO₄)₃ (NZP) are regarded as promising SSEs, mainly because of their three-dimensional diffusion network enabling fast transport of Li ions through well-defined channels.
Starting from LiTi₂(PO₄)₃, we analyzed a huge variety of NZP materials by systematically screening the relevant parts of the periodic table, replacing atoms on the Ti sublattice fully and partly by tri-, tetra-, and pentavalent atoms, as well as the phosphate by silicate, vanadate, and arsenate anions.
The influence of different elements on preferred Li sites, Li mobility, and possible diffusion paths were analyzed by means of a combined approach of multiple computational methods with different levels of accuracy, ranging from density functional theory to molecular dynamics simulations with ionic bond valence potentials. Minimum energy paths and diffusion barriers were identified by making use of the nudged elastic band method, as well as saddle point and energy landscape calculations.

Materials achieving their catalytic functionalities with the help of structural phase transitions have drawn great attention in recent years for their diverse applications including fuel cells, catalysts, electrolysis cells, and rechargeable batteries. Since the phase transitions are determined by stabilities of the phases, their thermodynamics is one of crucial materials properties to investigate.
Oxide materials with reversible exsolution and incorporation (a.k.a. self-regenerative catalysts), which are generally precious metal doped perovskite oxides, are excellent examples of the materials with structural phase transitions. The form of precious metals in these self-regenerative catalysts can be tuned by operating conditions. These metal particles act as catalysts when they have surface segregation under reducing environment; however, they incorporate into the bulk host oxide materials under oxidizing environment and avoid potential catalytic deterioration.
In this presentation, reaction mechanism of the materials with self-regeneration has been predicted using Pd-doped BaCeO₃ (Pd-BaCeO₃) as model system. All chemical reactions (up to 300 reactions) with on compounds and molecules of Pd, Ba, Ce, and O are systemically screened based on ab-initio Ellingham diagram. This approach efficiently determines the stable forms of chemical species and the direction of reactions under various operating conditions.
We observe that self-regeneration can be accomplished by multiple reactions rather than a single reaction, and that the variable oxygen stoichiometry of the Pd-BaCeO₃ promotes the self-regeneration. Also, it is observed that temperature range for the self-regeneration is limited. Our study will provide guidelines for further understanding self-regenerative catalysts, and therefore will establish their materials design principles.

The spontaneous process of self-assembly arranges chaotic building blocks into ordered structures under the influence of non-covalent interactions. Peptides are the simplest biological blocks with the ability to self-assemble into various nanostructures. These nanostructures have many possible applications due to their biocompatible, chemical, electronic, and magnetic, and optical properties. Understanding self-assembly is a fundamental step in knowing the complicated structure formation. Considering the importance and the promise that self-assembly holds for material science and nanotechnology, there is a great deal of interest in uncovering the finer details of the intramolecular and intermolecular forces involved. By changing the functional groups of units we can tune the non-covalent interactions and eventually control the nanostructure architecture.
We have studied the physicochemical properties of cyclic dipeptides, such as phenylalanine-tyrosine (Phe-Tyr) to understand their self-assembly in both gas and solvent phases. The initial hypothesis of the formation of a three-dimensional aromatic arrangement due to a possibility of pi-pi stacking between the aromatic phenyl rings of the dipeptides and hydrogen bonding between the hydrogen donor and an acceptor.
Quantum chemical computations of several possible conformers of cyclo(Phe-Tyr) dipeptide were performed Density Functional Theory (MO5 functional ) and Møller–Plesset perturbation theory in both gas and solvent phase. We used 6-31G* basis set in all calculations and the polarizable continuum model was used to study the influence of solvent. The structures, energetics, thermodynamics, and IR spectra of the lowest energy conformers were analyzed and studied further. According to our hypothesis, we believe that the hydrogen bonds between hydrogen donors and acceptors of two dipeptide units give rise to a dimer and hexamer formation and later the pi-pi interactions between the aromatic benzene rings of peptide units play a role in stacking these hexamers on top of each other to form a nanotube. We anticipate to further understand not just the intramolecular forces in a single cyclo(Phe-Tyr) dipeptide unit, but also the intermolecular forces between several units which help in nanotube formation (dimer à hexamer à nanotube).

Characterizing the static and dynamic mechanical response of compliant tissues and polymers is challenging, but of great interest for understanding traumatic injury and designing tissue simulant materials. Impact indentation is a mechanical characterization technique that may be used to measure the impact response of compliant polymers and tissues under quasistatic conditions, with impact speed up to 10’s of mm/s [1], but understanding of the complex material response requires theoretical and computational treatment. In such experiments, the compliant sample is impacted by a pendulum with a small (~mm radius) spherical or flat punch indenter, and the resulting displacement vs. time is measured. Until now, analysis of these data has been limited to empirical parameters: maximum penetration depth, quality factor Q, and energy dissipation capacity K (i.e, the energy dissipated in a half cycle of the impact oscillation). Here we present novel, easily implementable methods for (1) predicting quasistatic impact indentation response from dynamic moduli (e.g., G’ and G’’ from oscillatory shear rheology), and (2) direct measurement of glassy and relaxation moduli from impact indentation data. The method of analysis is based on a linear viscoelastic model and therefore ignores nonlinear, inelastic, and poroelastic effects. An equivalent finite element simulation was developed to validate the new data analysis procedures. These simulations confirm that (1) the output parameters (e.g., Q and K) may be predicted analytically without the need of finite element simulations, and (2) the viscoelastic moduli may be measured accurately from analyzing the raw output data.

There are growing interests on magnetic nanowires, due to their potential applications in magnetic sensors and recording devices. In particular, the tunable magnetic and chemical properties of nanowires make them an excellent vehicle for applying forces to cells. In this work, we report a comparative ab-initio study based on the Density Functional Theory (DFT) of Ni_xFe_1-x nanowire periodic arrays by using atomic-orbital and plane-wave basis respectively through DMol³ [1] and CASTEP [2] codes. The numerical calculations were carried out within the Materials Studio framework, with the generalized gradient approximation (GGA) of Perdew-Wang 1991 (PW91) for the exchange-correlation energy. After performing a geometry optimization with the spin-polarized option, we calculate the electronic band structure, density of states, magnetic moments and X-ray diffraction spectra. On the experimental side, Ni_xFe_1-x nanowire arrays can be fabricated by the electrodeposition technique into nano-porous alumina template. The magnetic moment of nanowires is measured by means of the vibrating sample magnetometry (VSM), as done for the Cobalt nanowires [3]. The obtained ab-initio DFT results are consistent with structural and magnetic experimental data, for example, the variation of magnetic moment in Ni_xFe_1-x nanowires with their diameter or interwire distance.

This work has been partially supported by UNAM-IN113714 and CONACyT-252943. Computations were performed at Miztli of DGTIC, UNAM.

[1] B. Delley, J. Chem. Phys. 113, 7756 (2000).
[2] M.D. Segall, P.J.D. Lindan, M.J. Probert, C.J. Pickard, P.J. Hasnip, S.J. Clark and M.C. Payne, J. Phys. Condens. Matter 14, 2717 (2002).
[3] K. Bouziane, Y. Roussigné, S.M. Chérif, A.A. Stashkevich, M. Vazquez, M.R. Britel and M. Cherkaoui, Sensor Lett. 11, 2282 (2013).

High dielectric constant materials are promising field for the development of high energy applications like capacitors, photovoltaics, photonics and transsitors.¹ Recently enhancement in dielectric constant by incorporating metal atoms in the polymer chain backbone has been found to be a promising field due to the improved electrical properties and morphology of the materials.² In this study, a series of zinc (Zn, 3d¹⁰) and cadmium (Cd, 4d¹⁰) aliphatic esters containing 1 CH₂ to 8 CH₂ unit(s) were synthesized in a two phase reaction system and characterized for their dielectric applications. Here, in this study rational design of materials are predicted by Density Functional Theory (DFT) based on their ionic, electronic and total dielectric constant contribution, along with the structural prediction. With this computational theory, organic polymers ³ and metal containing polymers such as, tin-polyesters⁴ correlate nicely with the experimental findings in the similar fashion as with here with the Zn and Cd-system. We found all of the Zn and Cd-esters shown to have stable dielectric constant ca. 4 and 5.5, respectively and low loss ca. 1 % over a range of frequencies. An even-odd trend of CH₂ unit(s) on dielectric properties are observed for both Zn and Cd-ester as have seen for Sn-esters⁴ . However, when more CH₂ unit(s) are incorporated in the backbone structure, the dielectric constant is lowered due to a decrease in metal atom concentration per unit area within the coordination environment. There is potential to further expansion of the chemical space, with the help of DFT calculations, for other Metal Organic Framework (MOF), in subsequent studies.

References

1. Baldwin, A. F. et al. Effect of incorporating aromatic and chiral groups on the dielectric properties of poly(dimethyltin esters). Macromol. Rapid Commun. 35, 2082–2088 (2014).
2. Mannodi-Kanakkithodi, A. et al. Rational Co-Design of Polymer Dielectrics for Energy Storage. Adv. Mater. Prog. Rep. (2016). doi:10.1002/ adma.201600377
3. Sharma, V. et al. Rational design of all organic polymer dielectrics. Nat. Commun. 5, 4845 (2014).
4. Baldwin, A. F. et al. Rational design of organotin polyesters. Macromolecules 48, 2422–2428 (2015).

Research has determined that nanostructures have distinct and measurable forms that can be modelled as nanopatterns. Said patterns accord with thoroughly-studied and established laws of chemistry and physics. One of the most essential patterns in this context is the lattice, which is a repeating arrangement of atoms in a crystalline solid. Of interest in this case is the hexagonal lattice seen in graphene, an allotrope of carbon with a distinct honeycomb lattice where each vertex is comprised of a single atom. Other allotropes of nanocarbon such as fullerenes can be viewed as iterations of this basic lattice--these include a range of structures ranging from buckyballs to nano-onions. Of particular interest is the carbon nanotube due to a wide range of useful properties and potential applications. Thus an analysis of carbon nanotubes is shown, extending the graph-theoretic treatment of honeycomb lattice to this important nanocarbon allotrope. The analysis is subsequently extended to the spherical carbon nanostructures known as buckyballs. Based on the highly symmetric structure of the buckyball, graph theoretic analysis can be helpful in determining important aspects of structure such as stability. Using this approach, some basic fullerene properties are predicted. Parameters for continued analysis of fullerene structures are shown, and analogues for modelling other allotropes are presented.

We propose an “inverse” microstructure optimization process, driven by the desired physical properties of the material, through the “design-analysis-manufacturing” cycle, where the optimization of microstructure is accomplished through an integrated computational framework combining phase-field simulations to generate the microstructures and finite element modeling to predict the properties of the microstructure, while additive manufacturing is used to produce the optimized microstructure. The proposed optimization process is applied to a functional composite composed of flexible polymer as matrix and ferroelectric ceramic (BaTiO₃) as energy harvesting source. Based on our computational framework, the composite with three-dimensional (3D) interconnected geometry of BaTiO₃ is verified to have superb stress and heat transfer efficiency, which guarantee the excellent piezoelectric and pyroelectric properties. The optimized 3D interconnected microstructure is then practically fabricated using additive manufacturing.

Candida antarctica lipase B (CALB) is an efficient biocatalyst for esterification and polymerization reactions. However, synthesis activity of CALB needs improvement for some non-specialty products using substrates that have limited interaction with enzyme. In order to elucidate the important factors that are responsible for enzyme activity and stability and consequently improve the productivity of CALB, we performed two different studies using molecular dynamics (MD) simulations in combination with experimental verifications: (1) the effect of solvent and (2) specific amino acid changes on enzyme activity. For the solvent effect, our results demonstrated that the conformational changes of the active site cavity are dependent on ions: in ionic liquids with weakly coordinating anion, such as [Bmim][TfO], [Hmim][TfO] and [Omim][TfO] or in organic solvent such as tert-butnaol, the catalytic cavity has open conformation with high enzyme activity; in ionic liquids with strongly coordinating anion such as [Emim][TfO] and [Bmim][Cl], the cavity becomes closed, thus leading to poor enzyme activity. These significant change observed in [Emim][TfO] or [Bmim][Cl] results from strong non-bonded interactions between strongly coordinating anions and secondary structure near the cavity and these interaction induces a conformational switch from an α-helix to a turn, which resulted in reduced cavity entrance size and its activity. From aforementioned study, we selected amino acids that affect structural changes of catalytic cavity and replaced them with other amino acids to obtain variants with larger cavity size and better enzyme activity. Through combined computational and experimental approach, we identified specific amino acid changes near the catalytic cavity in the wild-type CALB that can make cavity size bigger and increase the enzyme activity approximately 3 to 12 fold over the wild-type. Overall, our computational observations not only explained the structure-activity relationship of CALB in various conditions but also, successfully provided experimental designs for synthesis reactions using CALB, which include choosing optimal solvents and right mutation sites for better enzyme activity. This study also indicates the power of molecular modeling for the experimental designs and improvements of enzyme reactions.

Cortisol is an important glucocorticoid hormone whose biochemistry influences numerous physiological and pathological processes. Moreover, it is a biomarker of interest for a number of conditions, including post-traumatic stress disorder (PTSD), Cushing’s syndrome, Addison’s disease, and others. Due to both the diurnal cycle and the pulsatile release of cortisol, an implantable biosensor capable of real time monitoring of cortisol concentration in adipose tissue can potentially revolutionize the diagnosis and treatment of these disorders, as well as provide an invaluable research tool. Towards this end, we develop a mathematical model, informed by the physiological literature, to predict dynamic cortisol concentrations in adipose, muscle, and brain tissues, where a significant number of important processes with cortisol occur. The mathematical model was applied to both a healthy adult male, as well as a model Cushing’s disease patient. Values predicted by the model were verified against reported measurements from the physiological literature. Our model can be used to inform the design of an implantable sensor, by optimizing the sensor dissociation constant, apparent delay time, and magnitude of the sensor output versus system dynamics. Measurements from such a sensor would help to determine systemic cortisol levels, providing much needed insight for proper medical treatment for various cortisol-related conditions.

We demonstrate the feasibility of predicting experimental materials properties (e.g. density, heat of vaporization, etc.) using polarizable force field from ab initio quantum chemistry information. To acquire parameter sets using the genetic algorithm (GA) optimization, we compared the interaction energy from quantum mechanics (QM) calculations and molecular mechanics (MM) with the parameterized polarizable force field for various small organic homo-cluster systems, excellent agreement with the two sets of data is achieved. However, when we use the GA parameterized force field conducting the molecular dynamics (MD) simulations for properties of those organic materials, such as density and heat of vaporization, the discrepancy between the experimental results and the simulated results appears with trends. We adjusted the parameters according to the trend using guided genetic algorithm (GGA) and improved the MD simulations results for better properties of those organic materials. Throughout the overall parameterization procedure, the experimental data is not involved but only for validation. We present a case study of how ab initio quantum chemistry calculations provide guidance for classical MM force fields.

LiNbO₃ is one of the representative ferroelectric materials showing high Curie temperature (~1400K). Because of the small Li ion with respect to the tolerance factor, LiNbO₃ does not form a stable perovskite structure. Such low degree of freedom for the possible phase transition should be the origin of relatively lower piezoelectric constant of LiNbO₃. In order to control the tolerance factor, imposing mechanical constrain and/or composition tuning are thought to be effective. For example, we found that in-plane tensile strain along the a axis, which is specific to the thin film, can largely enhance a piezoelectric constant in ZnO by using density functional perturbation theory [1]. The piezoelectric strain constant d33 was predicted to reach ~200 pC/N for 2.8 at% V-substituted ZnO at 5.5% in-plane strain, just before the phase transition from wurtzite to h-BN-type structure [1]. On the other hand, we have elucidated the unknown crystal structure of high-pressure phase of LiNbO₃ by utilizing novel crystal structure prediction method [2]. However, predicted high-pressure phases of LiNbO₃ were revealed to show no piezoelectricity. Thus, we explored various combinations of cation elements based on the LiNbO₃-type structure, and their thermodynamical stability and piezoelectric properties were investigated by using density functional perturbation theory. Notable finding of the present study is that 7 types new LiNbO₃ type compounds, showing superior piezoelectric constant compared to LiNbO₃, were found within 439 types compounds. Within those 7 types newly discovered compounds, AlTlO₃ was predicted to show the largest piezoelectric strain constant d33, approximately 60 pC/N. Formation energies of various phases in AlTlO₃ were calculated as a function of pressure, and LiNbO₃ type structure was confirmed to be stable above 7 GPa. Ferroelectric R3c phase was predicted to be energetically stable and show no phonon instability from zero to 2 GPa whereas paraelectric R-3c phase was more stable above 2 GPa. Dielectric constant of LiNbO₃ type structured AlTlO₃ was found to diverge at 2 GPa, indicating the quenchable nature. For the paraelectric R-3c phase of AlTlO₃, imaginary phonon frequency derived from A_2u^T mode was observed at Gamma-point, the same as LiNbO₃. These characteristics of the ferroelectric phase transition in AlTlO₃ were revealed to originate in displacements of Tl and O ions. Energy gain of such ferroelectric phase transition was quite low, approximately ~1 meV/f.u.. Thus, structural fluctuation between paraelectric and ferroelectric phases should be the origin of the large piezoelectricity in LiNbO₃ type AlTlO₃. Predicted piezoelectric properties of other compounds will be shown at the conference.
[1] K. Nakamura, S. Higuchi and T. Ohnuma, J. Appl. Phys., 111, 033522 (2012)
[2] K. Nakamura, S. Higuchi and T. Ohnuma, J. Appl. Phys., 119, 114102 (2016)

Industrially utilized materials often have interfaces, in which the atomic structure is largely different from that in the bulk. The atomic structure of the interface properties of materials comprehensively, the determination of the interface structure and identification of the central structures to bring about the peculiar properties to the interface are highly required. Thus, theoretical calculation of the interface structure have been extensively performed.
On the other hand, very exhausting calculation is necessary to determine single interface structure because the interface has many degrees of freedom. Even one coincidence site lattice (CSL) grain boundary (GB), one has to consider tens of thousands of candidate atomic configurations with different three dimensional rigid body translations and search the most stable structure from them. Accelerating the interface structure searching has been thus highly required.
To more efficiently determine the interface structure, a genetic algorithm method and a random structure searching algorithm method have been proposed [1-4]. Although these approaches can efficiently determine unknown interface structures, many trial calculations, more than several hundred, are still necessary to determine a single grain boundary structure. If the structure and energy of an unknown interface could be determined more efficiently, the investigation of interfaces would be dramatically accelerated. Here, we developed the quite powerful method to determine the interface structure with an aid of the machine learning technique [5, 6]. In this presentation, we will present about the systematic determination of the symmetric tilt GB of bcc-Fe using virtual screening and kriging methods.
For the structure optimization and energy calculation, we used the General Utility Lattice Program (GULP) code with a Finnis-Sinclair type embedded atom method (EAM) potential was used [7]. In the virtual screening method, based on the training data using 5 GBs, we generated a “predictor”, and applied it to 29 GBs. In the kriging approach based on Gaussian process, we found that the kriging approach is also able to accelerate interface structure searching. The detail will be discussed in my presentation [8].

References
[1] A. L.-S. Chua et al., Nat. Mater., 9 (2010) 418
[2] E. Vitkovská and P. Ballo, Appl. Phys. Cond. Matter: Proc. 19th Int. Conf., (2013) 199
[3] G. Schusteritsch and C. J. Pickard, Phys. Rev. B, 90 (2014) 035424
[4] K. Inoue et al., Mater. Trans., 56 (2015) 281
[5] S. Kiyohara et al., in submitted.
[6] S. Kiyohara et al., JJAP, 55 (2016) 045502.
[7] M. W. Finnis and J. E. Sinclair, Philos. Mag. A, 50 (1984) 45.
[8] This study was supported by a Grant-in-Aids for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology-Japan (MEXT; Nos. 25106003, 26630302, 26249092).

An omnidirectional mirror is an optical device that works for a wide range of light frequencies and all incident angles, in contrast to a notable dissipative loss at infrared found on conventional metallic mirrors. In this work, we report a multiscale design of a dielectric multilayer device consisting of an ab-initio quantum-mechanical calculation of the dielectric function for each semiconductor layer with a specific atomic structure [1], followed by a study of wave scattering through the device using the transfer matrix method within the classical electromagnetic theory. The designed omnidirectional mirror consists of stacked Bragg multilayer reflectors, each one tuned to a portion of the reflection band. The validation of this design was carried out on a free-standing porous silicon multilayer film fabricated by electrochemical etching of a p⁺-type [100]-oriented crystalline Si wafer alternating two anodic current densities and finishing with a high current in order to separate the multilayer from the substrate [2]. The measured infrared transmittance spectra are compared with those predicted from the multiscale design, observing a good agreement. Finally, we think that multiscale studies based on ab-initio calculations could be a useful alternative for the design of photonic devices.
This work has been partially supported by UNAM-IN113714. Computations were performed at Miztli of DGTIC, UNAM.
[1] P. Alfaro, A. Palavicini and C. Wang, Thin Solid Films 571, 206 (2014).
[2] A. Palavicini and C. Wang, Optics and Photonics Journal 3, 20 (2013).

Despite progress and developments of quaternary Cu₂ZnSnS₄ and Cu₂ZnSnSe₄ (CZTS). today solar cells with initial efficiency of about 12.6%, further understanding of CZTS as well as ways for its improvement is necessary. The control of formation/stability of point defects is one of promising approaches for increase of efficiency and stability of CZTS solar cells, because defects are one of major factors, related to phase formation, electrical and optical properties. Because of this, we perform in this work hybrid functional calculations of not only CZTS but also its alloys with Ge or Si. In particular, we analyze trends in electronic properties of CZTS materials with Ge or Si doping at different dopant concentrations. We further analyze optical properties of the materials in term of dielectric function and absorption coefficient. Based on the predicted results, we expand understanding of Cu-based solar cell materials and ways for controlling CZTS properties by doping.
References:
C. Persson, R. Chen, H. Zhao, M. Kumar and D. Huang, in Copper Zinc Tin Sulfide-Based Thin-Film Solar Cells, John Wiley & Sons Ltd, 2014, DOI: 10.1002/9781118437865.ch4, pp. 75-105.
W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T. K. Todorov, Y. Zhu and D. B. Mitzi, Adv. Energy Mater., 2014, 4, 1301465

Computational materials design has had many successes in predicting new compounds, never previously synthesized. A condition for a prediction to be useful is that the novel material can also be synthesized. While in the early days of computational materials design only functionality was considered in proposing new compounds, now it is common to evaluate stability of a potential compound through the convex hull construction, thereby ensuring that the compound is a thermodynamic ground state. I will show several examples of such novel predicted ground states that were also synthesized, and which have applications in the area of energy technology.
A more novel challenge is how one can predict the existence of metastable compounds. While it is generally assumed that energy above the ground state line is a good criterion to determine whether a compound can exist in metastable form, I will demonstrate that this is not a sufficient condition, and that one may need to explicitly account for the synthesis conditions under which metastable compounds form. While this is challenging I will show examples of how one can predict pathways through which metastable Mn-oxides and metal-nitride compounds can be created.

This work was supported by the Center for Next Generation of Materials by Design, an Energy Frontier Research Center of the Department of Energy

The materials fidelity of classical interatomic potentials has increased significantly over the last few decades. It is thus now meaningful to assess the uncertainty in the predictions of specific potentials. Here briefly review some well-known ideas in the economic theory of investment portfolio management and suggest that similar approaches may prove fruitful in uncertainty quantification of interatomic potentials. In particular, we show that the analysis of a potential in terms of the Pareto surface allows the parameterization with high materials fidelity and with high robustness. The efficacy of this approach is illustrated for the simple example of a Buckingham potential for MgO. The analysis of the Pareto surface to compare the potential materials fidelity of various functional form for interatomic potentials is discussed.

Synthesis of materials that exhibit ultra-low wear rates is important for both efficient metal manufacturing processes and for engineering of durable components in applications that involve mechanical contacts (e.g., piston-cylinder contacts, engine injection sections, orthopedics, electrical contacts, and general bearings). Mechanical properties of metal alloys can often be significantly improved by introducing into the microstructure a large density of interfaces. For instance, it has been shown that so-called ultra-mild wear can be achieved in metals that during sliding develop a nanocrystalline tribolayer near the sliding interface. Interfaces can be also introduced artificially into the coating, for instance by synthesizing multi-layer structures. A critical factor in the rational design of materials that experience high contact loads is understanding of how properties of interfaces can be controlled by such factors as alloy composition and synthesis conditions, as well as how deformation-induced defect interact with the interfaces. In this talk we will present results of targeted simulations and experiments aimed at providing such understanding in Cu-Ag alloys with the overall goal of providing principles for materials design and of informing higher length-scale models of deformation in mechanical contacts. We have found that the wear resistance of nanocrystalline metal alloys depends on the coupling between the grain size and the contact size and there is a grain size that minimizes wear. Increasing Ag concentration in nanocrystalline Cu first increases the strength of the alloy and then decreases it, which phenomenon can be attributed to the evolution of grain boundary complexions with dopant concentration and associated change in deformation mechanisms. We have also investigated how the interface type (controlled by synthesis) affects mechanical response of Cu-Ag multi-layer systems, including the possibility of dislocation transfer across interfaces, presence of plastic strain recovery, and the Orowan strength in confined layers. Finally, we will discuss our progress in the development of a continuum model, which combines finite element analysis, crystal plasticity, and phase field simulations, to simulate microstructural evolution in sliding contacts.

Optical microscopy (OM) is often the first technique that is used to characterize the surface of many materials because of its simplicity and low-cost. However the microstructural information obtained by OM is, in general, only qualitative. In this work, we present an advanced OM technique to assess the microstructure of polycrystalline metals in a quantitative way. Our technique relies on the collection of a series of optical micrographs of polycrystalline samples taken under controlled illumination conditions. Digital processing and numerical analysis of the micrographs enable the characterization of several microstructural quantities—such as grain size distribution, grain boundary character distribution, surface roughness, and crystallographic information of the constituent grains.
Owing to the simplicity of the measurements, the inexpensive equipment used, and the high throughput, we expect our technique to be widely used throughout the community to generate materials data. The data will be useful for the development of novel, microstructure-centered, materials designs.

The cost of experiment continues to rise, while the cost of computation falls exponentially according to Moore’s Law. In the face of this, it behooves experimentalists to turn to computation to extract sophisticated information from simple, inexpensive measurements. We have implemented just such an approach to extract values of materials properties from temperature-dependent current-voltage (JVT) measurements on solar cells.

Because the properties of each device layer all have impact on the J-V curve, in principle, information about the former is embedded in the latter. In fact, there exist many software packages (SCAPS-1D, PC1D, Sentaurus Device, etc.) that can accurately simulate these impacts. However, traditionally, researchers do not attempt to extract this information, instead using JV curves only as broader indicators of device performance. In order to solve this inverse problem (i.e. moving from electrical measurements to materials properties rather than the reverse), we use a Bayesian parameter estimation technique in concert with device simulation. First, we simulate JV curves throughout the space of parameters in which we wish to infer values in order to create a prior probability distribution (values of other properties are treated as fixed). Then, data from electrical measurements are fed in and compared to these curves to generate a multidimensional posterior distribution. This posterior can be projected into one-dimensional distributions over each property of interest, providing a most probable value as well as an uncertainty arising from a combination of measurement noise and statistical uncertainty.

Using this approach, we have demonstrated the ability to accurately infer values of properties such as minority-carrier mobility, minority-carrier lifetime, interface band offset, and surface recombination velocity that are significantly more challenging, expensive, and time-consuming to measure using traditional techniques. The procedure has been validated on devices made from both traditional (e.g., GaAs) and experimental (e.g., SnS) absorber materials. This technique promises to significantly accelerate the cycle of learning in development of new PV materials or device architectures. To that end, we are working to make the code publicly available so others can adapt it to their tools and workflow. It could also have impact in a wider area than just photovoltaics – any physical system for which there is a reliable predictive model to simulate the results of a measurement that can be performed in a high-throughput way would be amenable to this technique.

Microscopic interactions in electrolyte solutions and electrolyte/solid interfaces play central role in many chemical and biological processes including nucleation, particle mediated crystal growth, self-assembly, recognition reactions and materials synthesis. The challenge for theoretical treatment of microscopic interactions in heterogeneous condensed phase systems is in accurate description of many-body interactions that is valid for a wide concentration range from dilute solutions to the solid state. Empirical potential force-fields that use the concept of pair-wise atom-atom interactions are usually fitted to reproduce either the properties of dilute solutions or the properties of solid state, but are not suitable for the intermediate concentration range or for the description of freezing transitions, except in some very rare cases. Mesoscopic theories of electrolyte solutions, on the other hand, treat many-body interactions from first principles, but lack the essential microscopic detail. In this talk a minimum parameter-free classical density functional theory model for electrolyte solutions, which treats short-range and many-body interactions from first principles, will be discussed through the prism of its application to revealing multistep synthesis pathways leading to highly ordered mesoporous materials and predictive synthesis of nanostructured materials via particle mediated crystal growth.

The phonon Boltzmann transport equation (BTE) is widely utilized to study non-diffusive thermal transport. However, the BTE is notoriously difficult to solve and analytical or efficient numerical solutions exist only for very special geometries and considerations. We find a solution of the BTE in the thin film transient thermal grating (TTG) experimental geometry by using a recently developed variational approach with a trial solution supplied by the Fourier heat conduction equation. We obtain an analytical expression for the thermal decay rate that shows excellent agreement with Monte Carlo simulations from literature. We also obtain a closed form expression for the effective thermal conductivity that demonstrates the full material property and heat transfer geometry dependence, and recovers the limits of the one-dimensional TTG expression for very thick films and the Fuchs-Sondheimer expression for very large grating spacings. The results demonstrate the utility of the variational technique for analyzing non-diffusive phonon-mediated heat transport for nanostructures in multi-dimensional transport geometries, and will assist the probing of the mean free path (MFP) distribution of materials via transient grating experiments. We provide an extension of the standard MFP reconstruction procedure in order to account for the full material property dependence of the effective thermal conductivity and allow for efficient study of a wide variety of materials.
This work is supported by Solid State Solar-Thermal Energy Conversion Center (S³TEC), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, under Award Number: DE-SC0001299/DE-FG02-09ER46577

Nanostructures play an important role in interfacial heat transport, which have wide application in designing thermoelectric materials, phonon filters, thermal insulation materials, etc. However, it is challenging to find the optimal structure from tremendous number of candidates. Material informatics has been considered as one promising way to accelerate the material discovery and design.
In this work, the typical alloy and superlattices nanostructures for interfacial phonon transport were designed and optimized by using Bayesian optimization. The atomistic Green’s function was employed to calculate the phonon transmission and thermal conductance of the candidate structures during the optimization.
The alloy structure optimization deals with an interfacial Si/Ge alloy region, and the problem to be solved is how to organize the alloy atoms to obtain the best and worst interfacial thermal conductance for a given volume fraction. Through Bayesian optimization, eight structures with highest thermal conductance and one unique structure with lowest thermal conductance were successfully found. The highest conductance is around 2.3 times of the lowest thermal conductance. The optimal structures were obtained by calculating only around 4% of the total candidates, saving the computational resources considerably.
The Si/Ge superlattices structures with fixed layer numbers ranging from 8 to 16 were also optimized by arranging the order of Si/Ge layer to obtain the minimum interfacial thermal conductance. Compared with traditional periodic superlattices, the designed aperiodic superlattices can further decrease the interfacial thermal conductance by 20~50% through optimization. With increasing superlattices layer thickness as well as the number of interfaces, the thermal conductance decreases. Therefore, for a given superlattice thickness, the layer thickness and interface numbers are two competitive parameters, which thus gives rise to the optimal structure with minimum interfacial thermal conductance. The interference effect in superlattices was also characterized by comparing the phonon transmission from direct atomistic Green’s function calculation and cascade transmission model.
In conclusion, the presented optimization of alloy and superlattices structures for interfacial phonon transport has shown the effectiveness and advantage of material informatics in designing nanostructures to control heat conduction.

The transition metal carbides are a class of refractory materials known for having very high melting temperatures, high electrical and thermal conductivity and exceptional hardness. The IVB and VB transition metal carbides and nitrides form a rocksalt structure near equal parts metal and non-metal atoms. Despite the similar bonding and the same crystal structures, the IV carbides slip on {110} planes and the V carbides slip on {111} planes at low temperatures. The differences in available slip system makes the IV carbides more brittle at low temperature and more susceptible to cracking. In this talk, we present detailed DFT and analytical studies regarding slip and bonding in these materials. Notably, we show that the presence of an ISF on the GSF surface stabilizes dislocations on the {111} planes, which is lacking in the IVB carbides where {110} slip is observed. The existence of the ISF and its depth is shown to correlate strongly with the number of valence electrons as well as the structural energy differences between the B1 and tungsten carbide structures and is related to the angular nature of the bonding. This simple model predicts {111} slip in all the nitrides regardless of the group of the transition metal. Experimental work on hafnium nitride confirms {111} slip at both low and high temperatures, supporting model predictions. These results suggest methods to tailor the hardness and enhance ductility in these materials.

The ever-increasing spectrum of functionalities required for developing and optimizing materials fundamental to modern civilization requires efficient paradigms for materials discovery and design, beyond current serendipitous discoveries and classical synthesis-characterization-theory approaches. To bridge these complex issues will require integrated and direct feedback from multi-scale functional measurements to theory and must allow real-time and archival experimental data to be incorporated effectively. In this presentation, I will illustrate the use of correlated structural and functional imaging to simultaneously probe local structure and functionalities on atomic level. In scanning tunneling microscopy, direct mining of structural and electronic data is demonstrated for graphene and high temperature superconductive materials. The use of linear correlation analysis and kernelized correlation methods for analysis of the structure of resultant data sets is demonstrated. For electron microscopy, we explore the use of advanced segmentation methods to automatically index the images and build libraries of local chemical and physical configurations that can further be correlated with electron energy loss spectroscopy. On the mesoscale, we demonstrate direct mining of the domain imaging and dynamics to establish the correlations between the local structure and domain dynamics. Beyond establishing the relevant correlations, we extend this approach towards deep data methods that enable merging this knowledge with physical models. A smart data approach will enable algorithms for data identification, expert assessment, and ultimately, control over matter via real time feedbacks. I will further discuss the likely changes in the preponderant research paradigms enabled by real-time collaboration infrastructure, open workflow processes, and emergent text mining and image recognition tools.

This research is supported by the by the U.S. Department of Energy, Basic Energy Sciences, Materials Sciences and Engineering Division, and was conducted at the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National Laboratory by the Scientific User Facilities Division, BES DOE.

Ferroelectric materials offer compelling fundamental physical challenges and outstanding promise for applications, only some of which has been realized. Recently, a new paradigm for research in this field has emerged along with a new perception of its relationship to other fields. In particular, new experimental techniques have illuminated the extent and the limitations of analogies between ferroelectrics and other functional materials including ferromagnets, glasses, liquids, and liquid crystals. A consensus view may be that these analogies are all helpful but incomplete, leaving ferroelectrics as a class with unique properties. Signature features include strong and competitive interactions on different length scales, as well as coexisting long-range order and strong disorder.

Theoretical researchers have adopted and modified analytical techniques that originated in related disciplines, synthesizing them to create refined tools well suited to describe and understand the complexity of ferroelectrics. This includes elements of critical phenomena and phase transition theory, solid-state physics, crystal chemistry, theory of liquids, and nucleation and growth theory. On the one hand, this state of affairs could be considered idiosyncratic and perhaps nearly incomprehensible to those entering the field. Instead, I will put forward the view that ferroelectrics serve as the crossroads of materials science, chemistry, and physics.

In this lecture, several areas of contemporary ferroelectric research will be highlighted, in order to illustrate the integrated research paradigm described above, with emphasis on the relationships between physical mechanism and materials design as well as between theory and experiment. The dynamics of ferroelectric domain walls has seen great progress in the last decade. I will discuss these developments, with focus on current perceptions of how walls move and how they influence material properties. The relaxor ferroelectrics are perhaps the epitome of complexity, offering multiple length and time scale dynamics. I will talk about recent work in this field, including new understanding of the formation and behavior of polar nanoregions. An exciting materials design challenge has been the search for ferroelectric photovoltaics, and I will provide a report of progress toward materials with strong visible light absorption and photocurrent generation. As an extension of this, I will comment on connections between ferroelectricity in oxides and hybrid perovskites. Beyond bulk phenomena, I will highlight interesting developments in the area of interfaces between ferroelectrics and other active systems, including graphene and molecular fluids. The commonalities in physical mechanism and applicable synthesis and characterization techniques in these functional systems offer enhanced prospects for integrated electronics and optoelectronics, in which ferroelectrics concurrently play many roles.

The Materials Project (www.materialsproject.org) – part of the broader Materials Genome Initiative - is an effort to compute the properties of all known inorganic materials and beyond, and offer that data to the community together with online analysis and design algorithms. The current release contains data derived from density functional theory (DFT) calculations for over 70,000 materials, each with searchable associated properties such as relaxed structure, electronic state, energy storage capability, aqueous and solid stability, and more. The software infrastructure enables thousands of calculations per week – enabling screening and predictions - for both novel solid as well as molecular species with target properties. Current application areas include photocatalysis, thermoelectrics, beyond-Li energy storage, and alloy design. To exemplify the approach of first-principles high-throughput materials design, we will make a deep dive into a few of the materials design efforts which are currently enabled by the Materials Project infrastructure and computing resources. Such projects are i) stability of inorganic materials in aqueous electrolytes for battery, fuel cell and catalysis applications, and design principles for ii) novel multivalent intercalation cathode discovery and iii) electrolytes.

Mesoscopic simulations and experiments are used to examine the dynamics of thin films of phase-separating polymer solutions that contain nanoparticles. When phase separation occurs, the nanoparticles segregate to the interfaces between polymer phases, which locks the configuration of the phases into place. To control the morphology of the thin film, an electric field is applied to align the phases and the particles perpendicular to the film. The computational approach is a novel Brownian Dynamics-Cahn Hilliard (BD-CH) hybrid model that combines particle- and grid-based simulation techniques, employed with highly parallel computing for large-scale 3D simulations. Corresponding experiments utilize a thin-film polymer casting apparatus that simultaneously applies an electric field and elevated temperatures. Once the desired morphologies are obtained, one of the polymer phases can be dissolved away, leaving a membrane-like structure containing pore-channels that are lined with the nanoparticles, thus imparting functionality to the pore surfaces. This talk will focus on the comparisons between simulations and experiments, and the derived relationships between the channel diameter, the channel areal density, and the interfacial nanoparticle arrangements. This project is supported by NSF grant #1511896.

Computational material design based on machine learning techniques is a rapidly growing area in materials science. A machine learning technique enhances the exploration of materials and enables us to extract meaningful information and pattern from existing data. In the application of machine learning techniques, some fundamental properties and geometric properties to represent elements and crystal structure play an important role in making a prediction model of a target physical property, which are called “descriptors”. The accuracy of the prediction model depends mainly on (1) how to represent elements in a compound, (2) how to represent crystal structure of a compound, (3) how to combine the elemental and crystal structure features and (4) how to use these features as descriptors.
In this study, we apply several procedures for preparing descriptors to a large data set and two small data sets in order to discuss about the accuracy of the descriptors. The large data set is composed of the cohesive energy and bulk modulus for about 18000 compounds, computed by the density functional theory (DFT) calculation. The two small data sets include a data set of lattice thermal conductivity (LTC) for 110 compounds computed by the DFT calculation and a data set of the experimental melting temperature for 248 compounds. We adopt fundamental properties to represent elements in a compound such as atomic number, atomic mass, Pauling electronegativity, ionization energy, and so on. To represent crystal structure of a compound, pairwise representations and bond-orientational order parameters (BOPs) are considered. Then we regard a compound as a distribution of these properties. Finally we use quantities to represent the distribution as descriptors, such as the mean, standard deviation, skewness and covariance of the distribution. To make prediction model, we use only the kernel ridge regression.
Using the dataset of cohesive energy, we obtained a prediction model with the prediction error of 0.042 eV/atom from the mean, standard deviation and covariance of 21 elemental properties, 20 BOPs and 20 pairwise functions to represent crystal structure. This is much smaller than that of a prediction model (0.16 eV/atom) constructed by simply using elemental properties and radial distribution function. In the cases of the bulk modulus, LTC and melting temperature, we also obtained good prediction models from the same descriptor set as used in predicting the cohesive energy.

Since its start in 2011, Materials Project (MP, [1]) has grown into a world-wide resource for a materials sciences community of more than 20,000 users who rely on the portal as a trusted source to accelerate their research. As a result, they wish to help with MP's efforts by contributing back, but also ask for support in sharing their experimental and computational datasets alongside MP's curated results. This provides the opportunity for researchers in both domains to validate calculations or measurements almost instantaneously and use the disseminated data for integrated materials studies.

With the public announcement of our general contribution framework, MPContribs [2,3], we will present a sustainable solution for well-curated data management, organization and dissemination in the context of MP. The framework serves the purpose of collectively maintaining contributions to local and MP community databases as annotations to existing MP materials. It subsequently disseminates them through a generic interactive gateway powered by Jupyter notebooks or through custom project web apps enabled by the webtzite app kit [4]. The MPComplete service allows users to suggest new compounds for calculation by MP, thereby involving the community in the process of growing the available materials data. MPCite [5] creates persistent citations and facilitates sharing amongst collaborators by assigning Digital Object Identifiers (DOIs) to the more than 70,000 MP compounds. This feature extends to contributed experimental and computational data as well as data/DOI collections, and hence allows the MP user to achieve a new level of reproducibility in her research.

In this talk, we will give an overview of the full software stack employed to move MP toward an integrated analysis and validation hub for experimental and computational materials data. As a real-world example, we will demonstrate a MPContribs-driven spectroscopy data processing pipeline directly from the Advanced Light Source's (ALS) beamline at LBNL to MP and back. The pipeline will include a tailored web app for dissemination of ALS data, its interactive comparison and validation vs theoretical calculations, and visual overlays of composition-based observables such as magnetic moments with MP phase diagrams.

[1] Materials Project, https://materialsproject.org
[2] MPContribs, https://github.com/materialsproject/MPContribs
[3] MPContribs, arXiv:1510.05024, arXiv:1510.05727, MRS Spring 2016
[4] webtzite, https://github.com/materialsproject/webtzite
[5] MPCite, https://github.com/materialsproject/MPCite, SciPy 2016

As a red component, nitride phosphors are widely used in achieving warm-light-emitting diodes (LEDs) with high color rendition. There is therefore a huge impetus for the development of new nitride phosphors to further improve the luminous efficacy and color rendition of phosphor-converted LEDs. However, the harsh synthesis conditions (high temperature and/or high pressure) of nitride phosphors have severely hindered this development. In this talk, we outline a computationally efficient multi-property screening framework for identifying promising narrow-band red-emitting phosphors utilizing the Eu²⁺ activator based on first principles calculations. This screening relies on the use of first principles descriptors that correlate with desired red phosphor properties, such as high thermal/chemical stability, high quantum efficiency, near-UV absorption wavelength and narrow red emission band. We will demonstrate the applicability of our approach through its ability to identify existing well-known narrow-band red phosphors. Finally, we will propose several promising novel red phosphor host materials using this approach. Our study presents a highly efficient way to discover and design novel phosphors for illumination-grade lighting.

La₂Mo₂O₉ (LAMOX) and its derivatives represent a new family of fast oxygen ion conductors that have potential applications in intermediate temperature solid oxide fuel-cells. The non-centrosymmetric crystal symmetry of LAMOX is one of the most complex inorganic crystal structures with 312 unique atoms. This structural complexity combined with the high oxygen mobility in this system could lead to promising novel functional properties. In this work, we systematically image the electromechanical response and its coupling with local crystal structure of LAMOX, at the nanoscale, using piezoresponse force microscopy (PFM) and synchrotron X-ray diffraction microscopy (XDM) as well as bulk ultrasonic measurements.

Through the use of semi-supervised machine learning such as spectral clustering we can identify the evolution of the crystal phases in LAMOX as a function of temperature by XDM. We find the presence of sub-grain features, that are piezoresponse active as confirmed by PFM. Furthermore, these sub-grains domains show pronounced evolution with moderate heating (up to ~600 K), that cannot be accounted for by the thermally induced atomic displacements; instead, reordering of the oxygen sublattice via collective low-energy motions appears to be a plausible explanation. This thermal activation and local tip bias-induced switching properties of the piezoresponse present in these domains is further confirmed by PFM. The reported results shed information on the nanoscale oxygen ion hopping dynamics for this system that are inaccessible by macroscopic approaches and demonstrate the indispensability of data-intensive techniques to explore the wealth of information present in multimodal imaging of complex materials for the discovery of novel properties.

This research was supported by a Eugene P. Wigner Fellowship at Oak Ridge National Laboratory (NL). The PFM experiments were performed at the Center for Nanophase Materials Sciences, a U.S Department of Energy (DOE), Office of Science User Facility at Oak Ridge National Laboratory, and X-ray imaging at the Advanced Photon Source (33-ID-D), a U.S. DOE Office of Science User Facility at Argonne National Laboratory.

Piezoelectrics are important materials for a variety of applications including energy scavenging, low power sensing, and band pass filters in wireless communications. Computational searches for new piezoelectric materials have generated libraries of candidate materials, including metastable structures which have not yet been experimentally observed. Many of these metastable materials, particularly mixed metal oxides, likely have large kinetic barriers to crystal transformations and hence could persist at ambient conditions. We report a high throughput process utilizing millisecond time scale annealing for discovering and optimizing conditions to quench these metastable structures. Samples, deposited as amorphous films by sputtering or pulsed laser deposition, were laser spike annealed (LSA) over a wide range of conditions in a combinatoric manner using a lateral gradient (lgLSA) method. Resultant structures were identified using spatially resolved x-ray diffraction. In the TiMnO₃, LiMnO₂, and Li₄WO₅ systems we observed many computationally predicted phases including the potentially highly piezoelectric 3m point group TiMnO₃ and 4 point group LiMnO₂. In addition, we observed several phases that cannot be matched to the predicted structures. Finally, a TiMnO₃ – TiFeO₃ – TiNiO₃ pseudo-ternary composition spread, with predicted high piezoelectric coefficient phases, was characterized to identify optimal composition and processing regimes. The use of the lgLSA technique with composition spreads promises to dramatically accelerate the experimental synthesis of new phases.

Polymer properties are often determined by structure and dynamics on multiple length and time scales, which present a unique challenge for computational studies of polymers. All-atom molecular dynamics (MD) simulations have been used extensively to study properties of polymers at the length scales of hundreds of angstroms, however, as the simulation size increases, computations become slow. Coarse-graining (CG) of polymers based on all-atom MD simulations enables study of large spatiotemporal scale properties of polymers while retaining essential atomistic details of their monomers. This type of modeling is being increasingly used to study properties of polymers at large length and time scales. Polystyrene (PS) is extensively studied both experimentally and computationally due to its low glass transition temperature and narrow polydispersity. Several coarse-graining models of PS exist which could be categorized by number of beads used to represent its monomer. Here, we will present a new coarse-grain methodology based on a single all-atom simulation of a PS melt that can capture stereochemistry of the polymer chains in the melt. The CG model that we have used incorporates 2 beads to represent one monomer and presents an immense improvement over previous studies since it captures the stereochemistry of the polystyrene. The nonbonded interactions are obtained using the iterative Boltzmann inversion (IBI) scheme. These CG models can be back-mapped to the atomistic structure.

We have further extended this methodology to coarse-graining analogues of PS such as randomly sulfonated polystyrene (SPS) ionomer melts in which one additional bead is used to describe the sulfonate group. All the bonded distributions are extracted from a single all-atom melt simulation and converted to their corresponding energy using Boltzmann inversion. The nonbonded potentials for the two PS beads are taken from our previous work. Interactions between charged and uncharged beads were developed using iterative Boltzmann inversion method. Due to strong electrostatic interactions, the mobility of these melts even with a very small fraction of sulfonate groups is very small, making it difficult to obtain equilibrium distribution functions for Boltzmann inversions. Therefore we reduced the strength of the electrostatic interaction by setting the dielectric constant ε = 5 to fit the bonded and non-bonded coarse-grained potentials. With these successful models, we will provide new insights into impacts of interfacial roughness on interdiffusion.

An adsorption model in conjunction with an open source software package, Surface Evolver, has been used to simulate grain growth arising from mean curvature and the effect on its kinetics due to the dynamic redistribution of solute between grain boundaries and bulk. During grain growth, adsorption effects lead to the reduction in surface energy and modify the growth kinetics. The analytical model has been built using Gibbs' adsorption isotherms, the Langmuir-Mclean relation between bulk solute and the grain boundary solute, and the relation between these two equations. An initial grain boundary structure is generated through a Voronoi tesselation about random points, and hence a random arrangement of grains of the required number and the required average grain size is generated. The computation proceeds by computing the total grain boundary length during grain growth and re-partitioning solute between the boundary and bulk to conserve mass. The material parameters involved in the simulation are the segregation energy of the solute and the specific area of the segregant. The deviation in the kinetics from the classical linear dependence of square of grain size on time is shown, and the dependence of the rate constant on the material parameters and initial solute mole fractions is studied. The utility of Surface Evolver in aiding the design process for a stable grain structure is demonstrated by simulating the path of grain growth for a given set of initial conditions, including grain size and solute mole fractions, and the range of value attained by the bulk mole fraction is used to identify whether the system will reach a thermodynamically favourable state for precipitation. As an example of its practical application, specific material systems like Ni-P are simulated in this manner and different regions are identified in grain size-solute mole fraction space that correspond to different kinetic behaviour.

The structural phase field crystal (XPFC) model is often used to study material phenomena at atomic length-scales and diffusive time-scales, bridging the gap between atomistic and mesoscale models. As such, the XPFC model can be used to simulate grain boundary behavior in hexagonal close packed (HCP) metals, such as Mg. However, the XPFC model is phenomenological in nature and needs to be parameterized using experimental data or other computational results. A key feature of the XPFC model is that the two-body density correlation functions (DCFs) used to drive the dynamics of the model are constructed using Gaussian functions. Previous work has shown that changing the parameters that control the height and width of the Gaussian functions will affect the grain boundary energy (GBE) and structure in 2D hexagonal systems. Building on these results, 3D HCP grain boundaries are created using the XPFC model, with the DCF parameterized so that the GBEs agree with those calculated by 3D molecular dynamics simulations available in the literature [1]. The parameterized XPFC model will be used to study equilibrium and non-equilibrium phenomena associated with Mg grain boundaries, and the results may be used as input for larger scale continuum models such as phase field simulations.

[1] C. Ni, H. Ding, M. Asta, X. Jin, Scripta Materialia, 109 (2015), 94-99.

Developing protocols for making thin sheet Galfenol with abnormally grown Goss or Cube grains is challenging because the mechanisms that regulate grain boundary mobility and texture development in these alloys are not well understood. Grain boundary energy models do not account for extraneous driving forces caused by the control of surface energy from atmospheric annealing conditions. By characterizing the surface energy of specific Galfenol grains at room temperature, we can develop a more accurate thermodynamic-based framework for modeling abnormal grain growth and texture development that will be used to understand why a high temperature atmospheric anneal transforms myriad grains into highly textured, single-crystal-like polycrystalline material. To experimentally measure surface energy, we are developing contact angle methods to directly probe highly-textured and single-crystal Galfenol. By applying the two-liquid-phase contact angle method to an ion beam-patterned Galfenol surface, the probe water droplet will not spread as it does on a flat metal surface. The Young contact angle (flat surface contact angle) can be extracted from the Cassie-Baxter equation for use in calculating the surface energy of targeted crystal orientations. Ultrasonic and plasma cleaning are used to maintain a Cassie-Baxter wetting state to further increase the metal/probe-liquid/bulk-liquid contact angle. This method has potential for surface energy characterization of any metal with grain sizes >2 mm to accommodate the minimum size of the probe water drop.

Avik Mahata, Mohsen Asle Zaeem and Michael I. Baskes
Abstract: Homogeneous nucleation from undercooled fcc aluminum (Al) and bcc iron (Fe) melts was investigated by million-atom molecular dynamics (MD) simulations utilizing the second nearest neighbor modified embedded atom method interatomic potentials. The natural spontaneous nucleation was reproduced over a very large timescale without any influence of impurity, pressure and surface effects. A major difference between homogenous nucleation of Al and Fe was detected: solidification and nucleation from Al exhibited stacking faults while Fe did not show any crystalline defects.
MD simulations of homogenous nucleation showed three thermodynamic regimes: (I) Nucleation from the liquid: fcc or bcc small particles (initial nuclei) formed, and as particles grew, most of them dissolved back to the liquid phase. (II) Some particles reached the size of a critical nucleus (which was found to be ~0.82 nm for Al and ~0.77 nm for Fe) and passed the barrier of critical radius. (III) After overcoming the critical radius, further growth of particles led to a favorable condition for continuous solidification, and this eventually led to formation of bulk-crystalline solid. This three stage observation in MD simulation can be also related to the classical theory of nucleation, where the Gibbs free energy increases initially which is not thermodynamically favorable, then after it reaches a critical value (at the critical nucleus size) it shows a tendency favorable for the continuous solidification. The nucleation rate reached a maximum value of 42 ns^-1 (54 ns^-1) at the critical temperature of ~500K (~1,000K) for fcc Al (bcc Fe) simulation cell of 25×25×25 nm³ (28.5×28.5×28.5 nm³), which is equivalent to 2.7×10³⁷ (2.33×10³⁶) m⁻³s⁻¹. The typical nucleation rate for homogeneous nucleation of a pure metal was previously estimated by experiments to be in the order of 10³⁰ to 10⁴⁰ m⁻³s⁻¹ near the critical temperature of nucleation. Another important factor in nucleation is the induction time, which is defined as how often nucleation events occur, which also represents the ability of a system to sustain small thermal fluctuations. The higher the nucleation rate is, the faster a system escapes the metastable undercooled state. At the critical temperature, the induction time was found to be ~1.0 ns in both Al and Fe. In general, the minimum average grain size occurs when annealing at the critical temperature of nucleation. Microstructures obtained after solidification showed a minimum grain size of ~10 nm for Al and ~8 nm for Fe. Previous experiments showed that annealing at a constant temperature reduces crystal defects and creates larger grains, and a similar trend was observed in MD simulations for both Al and Fe.

Harnessing the mineralization of natural materials into complex microarchitectures holds the potential for functional structures. One model system is the combined precipitation of barium carbonate nanocrystals with amorphous silica, which yields thin-walled aggregates amenable to sculptural control by rational modulations of ambient conditions in a dynamic reaction-diffusion environment. To explain the resulting morphospace, we developed a geometrical morphogenesis framework based on the kinetics of the growth front, a curve that lays down a surface in space as it evolves over time. Our theory explains the observed range of precipitate morphologies, including vases, petals, and helices. It also defines environmental protocols to engineer architectures which we verify by growing mineralized shapes with proven optical properties. Altogether, our approach allows for the bottom-up control of precipitation-based mineralization patterns.

The design and discovery of new semiconductors with targeted multi-property functionalities for applications in solar energy conversion is a current challenge in developing new (photo-)electrode materials. Traditionally, theoretical predictions of ground state structures and their thermodynamic stability have facilitated the discovery of new materials and the optimization of known compounds. However, the field of materials design is experiencing a paradigm shift from near-equilibrium materials towards a much larger design space that includes materials far from equilibrium. In this work, we use theoretical and computational screening methods to identify and design metastable metal pnictide and metal chalcogenide alloys with desirable semiconducting properties for solar energy conversion applications. The materials design principles utilize both structure-property and composition-structure relationships to enable tunable multi-property functionalities. We leverage the possibility to overcome thermodynamic solubility limits by nonequilibrium thin-film growth to realize these novel semiconducting alloy electrode materials. Our theory guided synthesis employs a high-throughput combinatorial approach using pulsed laser deposition and reactive magnetron sputtering to rapidly screen composition and growth parameters to synthesize metastable spinel (Sn_1-xTi_x)₃N₄ and polymorphic Mn_1-xZn_xO alloys. A proof of principle is provided by initial photo-electrocatalytic device measurements that corroborate the materials by design features of tunable band properties and transport mechanisms to improve solar hydrogen generation using the novel alloy electrodes. Our results demonstrate a broadly applicable approach for synthesis of metastable materials and establish the predictive design and discovery of far from equilibrium materials for energy conversion.

Computer vision and machine learning methods were applied to the challenge of automatic microstructure recognition. Here, a case study on dendritic morphologies was performed. Two classification tasks were completed,
and involved distinguishing between micrographs that depict dendritic morphologies from those that do not contain this particular microstructural
feature (Task 1), and from those micrographs identified as depicting dendrites, different cross-sectional views (longitudinal or transverse) were identified (Task 2). Data sets were comprised of images taken over a range of magnifications, from materials with different compositions and varying orientations of microstructural features. Feature extraction and dimensionality
reduction were performed prior to training machine learning algorithms to
classify microstructural image data. Visual bag of words, texture and shape
statistics, and pre-trained convolutional neural networks (deep learning algorithms) were used for feature extraction. Classification was then performed
using support vector machine, voting, nearest neighbors, and random forest models. For each model, classification was completed using full (original size) and reduced feature vectors for each feature extraction method tested.
Performance comparisons were done to evaluate all possible combinations of
feature extraction, selection, and classifiers for the task of micrograph classification. Results demonstrate that pre-trained neural networks represent
microstructure image data well, and when used for feature extraction yield
the highest classification accuracies for the majority of classifier and feature
selection methods tested. Classification accuracies of 91.85 ± 4.25% and
95.74± 3.73% for Tasks 1 and 2 respectively, were achieved using pre-trained
neural networks for feature extraction. Thus, deep learning algorithms can
successfully be applied to the task of micrograph recognition. This work
is a broad investigation of computer vision and machine learning methods
that acts as a step towards applying these established methods to more sophisticated materials recognition or characterization tasks. The approach
presented here could offer improvements over established stereological measurements by removing the requirement of expert knowledge (bias) for interpretation of image data prior to characterization.

Often in materials research, a significant amount of data is collected and then distilled down to a single parameter. Conventional approaches to materials research ignore the vast data that is available and forgo leveraging data-intensive scientific exploration. In this study, a materials informatics approach is used to elucidate the microstructural evolution of nickel superalloy, PWA 1484, coupling experimentation with statistical correlations. Semi-automated digital analysis of optical and scanning electron microscopy (SEM) images was employed to identify and characterize microstructural changes in PWA 1484 from various time and temperature exposures. An image analysis (IA) routine has been developed in MATLAB to analyze the digital images captured from both characterization techniques. Circular polarized-differential interference contrast (Cir-DIC) optical images were acquired (Zeiss Axio Observer.Z1m) and processed with the IA routine to threshold the two-phase microstructural constituents γ/γ’ and provide a measure of the volume fraction of γ’. SEM images were acquired using the circular backscatter detector and ultra-high resolution mode on a dual-beam microscope (FEI Helios NanoLab 600). Image analysis of the SEM images further quantifies the microstructure such that individual γ’ precipitates are measured for particle size, shape factor (deviation from cuboidal γ’) and distributions of γ’. The IA routine has drastically reduced the time of data analysis, obtaining measurements and statistical analysis within seconds. The two imaging methods are combined to leverage the large area sampled with the lower magnification optical approach and the detailed γ’ precipitate characteristics measureable with the higher magnification SEM images. The end result is a rich data set that can be mined to provide a predictive modeling methodology to determine the time-temperature exposure of nickel superalloy coupons. Overall, digital image analysis allows for the direct characterization of material properties and quantification of material property distributions.

Nano-sized particles (NPs) are synthesized and modeled for decades exploring their various useful applications such as catalysis, magnetic storage media, photonics, and drug delivery. A fact that the useful properties of NPs can be improved further and even new NP functionality achieved by not only controlling the NP size and shape but also interfacing chemically or structurally distinct entities into single composite NPs encourages the exploration of such NPs. A typical example is core-shell NPs wherein the synergy of distinct atoms at the core/shell interface endows the NPs with otherwise unachievable functionality. Using the core-shell interface of 2.5 nm in size Fe core- Pt shell NPs as an example, we demonstrate that precise knowledge of the 3D atomic arrangement at functional interfaces inside NPs can be obtained by resonant high-energy X-ray diffraction (XRD) coupled to element-specific atomic pair distribution function (PDF) analysis. On the basis of the unique structure knowledge obtained, we scrutinize the still-debatable influence of core\shell interface on the catalytic functionality of Fe core−Pt shell NPs, thus evidencing the usefulness of this nontraditional technique for practical applications.

In this presentation, we describe the methods we are developing for real-time control of grain growth in metals. We are integrating control and simulation methods with in-situ FIB-SEM observations of thermal grain growth in polycrystalline copper films, with the goal of achieving active control of microstructural evolution.

Ten-channel micro-heater arrays were designed using finite element analysis (FEA) and fabricated using micro-fabrication techniques to allow for controlled heating profiles across one square mm of surface films in the FIB-SEM. Real-time observation using secondary electron imaging allowed assessment of the evolving grain microstructure with resolution down to a few nanometers and imaging increments as small as a few seconds. Statistical grain size distributions are currently used as a measurement of microstructural evolution over time and temperature. Coupled with feed-back control of the individual heater line resistances, we are aiming to experimentally show the ability to control the evolving microstructure in-situ. We also target creating unique microstructures based on specific thermal gradients and real-time feedback from imaging.

Preliminary in-situ control of the temperature in each channel of the micro-heater arrays has been established through the characterization of the electrical resistances of the ten channels.Measured temperatures between the heater lines and the chip surface differ only by a few degrees Celsius from the FEM calculations of the temperature distribution. Real-time grain growth at uniform temperature has also been characterized in bulk polycrystalline films between 300 and 500 degrees Celsius and compared to Monte Carlo grain-growth simulation results. The simulated microstructure is initialized to that observed experimentally at room temperature and then annealed until the mean grain area is equivalent to the annealed experimental structures. Results show that the standard deviations of the simulated annealed grain structures are within 10-15% of the corresponding experimental standard deviations.

The initial results from these grain growth studies are used to develop a methodology for growth of controlled microstructure distributions in polycrystalline films deposited on heater arrays with designed temperature profiles. While the current model material is polycrystalline copper, we believe this technique has potential to application to other material systems.

This work is supported through the National Science Foundation DMREF program under CMMI-1334283. This work has been supported by facilities in the Center for Materials, Devices, and Integrated Systems at Rensselaer Polytechnic Institute.

Development of graphene-based electronic devices requires the investigation of the structure and properties of graphene-metal interfaces. Although several theoretical studies are revealing important structural and electronic properties of such an interface for different metals, few studies exist at size and time scales larger and longer than those that can be addressed by ab initio methods. Here, results of classical molecular dynamics of titanium thin films adsorbed on graphene, with and without defects, are presented in terms of the thickness of the titanium film. Also, the influence of different types of substrates on the adhesion of the titanium to the graphene was analyzed. The substrates considered in this study are additional layers of graphene, copper, copper oxide of various terminations, and titanium oxide in the anatase phase. The results are compared to the free standing titanium/graphene system. We also investigate the effects of thermal fluctuations on the graphene-titanium interface for several titanium thicknesses. The simulations are carried out with the third-generation charge optimized many body (COMB3) potential. The results reveal the electrostatic role played by induced charges on these substrates on the adhesion of the graphene to the titanium films. Deformations on titanium/graphene systems can be explained by the different thermal expansion behaviors between these two materials.

Thermal or laser based sintering of nanoparticles is important to emerging technologies, like additive manufacturing, or to established technologies involving material annealing. Since the final microstructure and the presence of atomic defects impacts the electrical, optical, and mechanical properties of a sintered material, it is important to optimize the processing parameters. We have studied the processing-structure-property relationship during the sintering of Au nanoparticles using in situ electron microscopy and atomic-level computational tools. Our results suggest that sintering of nanoparticles leads to the formation of a coherent particle without any residual dislocations or interfaces, both of which are detrimental for optoelectronic applications. In addition, using our simulated free energy landscapes and experimental results we are able to reconstruct a sintering mechanism map which illustrates that, in contrast to the classical notion of diffusion aided sintering, in nano-scale materials sintering is aided by extensive dislocation activity. Finally I will illustrate how this study helped us to prepare high quality ZnO think films from sintering of ZnO nanopillars.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.

A central theme in modern day materials science is the search for novel structure-response relations in materials with evermore increasing spatial and configurational complexity. This materials complexity trend is equally matched only by the immense increase in data, both in size and modality, that emanate from modern imaging instruments. In this talk, I will present a novel approach to extracting relationships between the spatial distributions of crystal structure and electromechanical responses of complex oxide thin-films that is rooted in machine learning of multimodal imaging. In particular, by i) fusing imaging data from multi-frequency piezoelectric response force microscopy and X-ray diffraction microscopy, and ii) embedding in a weighted similarity graph, I will show how to extract quantitative and statistically rigorous couplings between the structure and response from this network using semi-supervised machine learning and spectral graph theory. The structure-response relations that are mined from the network are found to reveal both expected materials properties such as diminished piezoelectric response near misfit dislocations but also the presence of a range of responses that are weakly related to structural variations, indicating the strong and direct influence of other materials degrees of freedom not present in the network. The statistical generality and experimental robustness of the construction of a structure-response network by machine learning empirical data will be discussed and its application to salient materials science questions will be discussed.

Developing novel, functional nanomaterials requires characterizing their structure on the atomic scale. Current state-of-the-art scanning transmission electron microscopes (STEM) provides sub-Angstrom level resolution of atomic structure, but full three dimensional structures cannot be determined except in exceptional cases and through enormous effort. Here we explore how the the effective integration of experimental and computational methods can greatly enhance our ability to determine the atomic scale structure and composition of materials relative to either tool alone.

We introduce a genetic algorithm (GA) optimization scheme for determining the structure and composition of real nanoparticles from a combination of STEM data and atomistic modeling. This scheme optimizes populations of candidate particles based on comparisons of forward modeled STEM images with experimental counterparts and their calculated thermodynamic stability. A number of methods were developed that allow for a sufficiently fast STEM image simulation and theoretical stability calculation of thousands of generations of nanoparticles, each containing thousands of atoms. We demonstrate the effectiveness of this approach with simulated STEM images and then apply it to determine the structure of both colloidal Au test particles and catalytically active Pt-Mo bimetallic nanoparticles.

High-throughput computational studies for materials design are often encumbered by the weight of analysis: modern supercomputers are able to generate enormous amounts of data very quickly, but if structure identification is not also high-throughput, researchers can become the bottleneck for scientific discovery. While analytic methods exist to distinguish some simple structures, automatic identification of more complex structures is still lacking. Here, we describe how to create order parameters for crystal structure detection from data using off-the-shelf machine learning methods. These methods can be applied using unsupervised learning to map out regions of interest in an unknown phase diagram or supervised learning to identify exact, known structures. These algorithms can be used as a basis to intelligently explore parameter space -- one of the important components in engineering materials by design.

The insertion of voids of nanometer size in semiconductors provides a highly effective and widely applicable approach to control their thermal conductivity [1,2,3]. By performing molecular dynamics simulations, we examine the influence of voids of varying size on the thermal properties of silicon metalattices, which consist of a periodic array of nanopores in the crystalline structure of silicon. We examine the effects of geometry and chemical factors on thermal conductivity such as isotopic substitution, pore symmetry, volume fraction and radius. We find that the thermal conductivity of these nanostructures is reduced by orders of magnitude compared to that of bulk silicon when the diameter of the pores is in the range of 1 to 5 nm. To explain this result, we provide a systematic assessment of structural effects on heat conduction using an autocorrelation analysis of the phonons [4,5,6,7].

[1] Lina Yang, Nuo Yang, and Baowen Li. Extreme low thermal conductiv- ity in nanoscale 3D Si phononic crystal with spherical pores. Nano Lett., 14(4):1734–8, 2014.
[2] Jin Fang and Laurent Pilon. Tuning thermal conductivity of nanoporous crystalline silicon by surface passivation: A molecular dynamics study. Appl. Phys. Lett., 101(1), 2012.
[3] J. H. Lee, J. C. Grossman, J. Reed, and G. Galli. Lattice thermal con- ductivity of nanoporous Si: Molecular dynamics study. Appl. Phys. Lett., 91(22):2007–2009, 2007.
[4] Anthony J. C. Ladd, Bill Moran, and William G. Hoover. Lattice thermal conductivity: A comparison of melocular dynamics and anharmonic lattice dynamics. Phys. Rev. B, 34(8):5058–5064, 1986.
[5] J. E. Turney, E. S. Landry, a. J H McGaughey, and C. H. Amon. Predict- ing phonon properties and thermal conductivity from anharmonic lattice dynamics calculations and molecular dynamics simulations. Phys. Rev. B, 79(6):1–12, 2009.
[6] A. J. H. McGaughey and M. Kaviany. Quantitative validation of the Boltz- mann transport equation phonon thermal conductivity model under the single-mode relaxation time approximation. Phys. Rev. B, 69(9):094303, 2004.
[7] Asegun S. Henry and Gang Chen. Spectral Phonon Transport Properties of Silicon Based on Molecular Dynamics Simulations and Lattice Dynamics. J. Comput. Theor. Nanosci., 5(7):1193–1204, 2008.

It is well known that the laser flash method is the standard, most reliable, popular method to measure thermal diffusivity of solid materials. Picosecond and nanosecond pulsed light heating thermoreflectance methods are natural evolution of the laser flash method applicable to thermal diffusivity measurement of thin films by introducing picosecond and nanosecond pulse lasers as a sources of impulse heating instead of a sub-millisecond pulsed laser. Reliable thermal diffusivity data have been systematically measured for variety of thin films including metals, alloys, oxides, nitrides, carbides, et al. with thickness from 10 nm to several micrometers. Exact analytical solution of heat diffusion equation is fitted to experimentally observed impulse temperature response in order to calculate thermophysical properties such as thermal diffusivity, thermal conductivity and boundary thermal resistance.
Two configurations can be chosen: One configuration is that the surface of the thin film is heated t by a laser pulse and the temperature response of the heated area is observed by reflected intensity of the probe laser. This configuration has an advantage in order to measure thin films deposited on any surface including opaque to the heating and probe beams.
The second configuration is that the thin film is heated through the substrate by a laser pulse and the temperature response of the surface opposite to the heated area is observed by reflected intensity of the probe laser. This configuration has an advantage to observe diffusion of heat across the thin film directly. This configuration can give direct and crucial information to investigate transfer of energy deviated from diffusion equation. This configuration cannot be applied if the substrate is opaque to both heating laser beam and probe laser beam.
The principle, instrumentation, and systematic analysis of data based on the exact solution of diffusion equation including boundary thermal resistance.
This method have been successfully applied to measure thermophysical properties of thin films used in electronics such as processors, RAM, optical memory, magnetic memory, power device, thermoelectric thin films, and thin films for energy application. Examples of these applications will be introduced.

Integrated computational materials engineering of third-generation steels requires a multi-scale approach that passes first principles data to meso-scale (e.g. microstructural) models. Here, density functional theory is used to compute Al, B, Cu, Mn, Si, C, and N solute effects on the lattice and elastic constants of BCC, BCT, and FCC Fe. We propose a solute strain misfit tensor to quantify the solute dependence of the lattice parameters, and the strain contributions to changes in elastic constants. We also compute the effects of changes in chemical bonding due to solutes on the elastic constants. Computing the two elastic constant contributions separately is computationally efficient, and their sum agrees with costlier direct calculations that encompass both effects. The computed data estimates solute-induced changes in mechanical properties like strength and ductility, and serves to increase the predictive capabilities of phase field and crystal plasticity simulations.

The electronic transport properties and electrical conductivity in conducting polymers are related to their structure and morphology. In spite of hundreds of experimental and theoretical studies in the recent years, fundamental questions concerning the role of solvents and doping counter-ions are not completely understood and interpretation of many experimental results remain controversial. In this study we report atomistic molecular dynamics (MD) simulations of the effect of doping anions (tosylate and PSS) and the solvent concentration on the morphology of conducting polymer poly(3,4-ethylenedioxythiophene) (PEDOT). A crystallization mechanism in this material was studied in both solution and solid phase. Crystal nucleation was carried out in solution and then the solid stage was reached by removing solvent in several steps. Our simulations clearly confirmed a formation of pi-pi stacking in moderately charged PEDOT chains as a result of a predominant Van der Waals interaction. We also found that an intercalating configuration with inserted tosylate rings in between PEDOT chains to be the most likely structure at higher charge carrier densities in PEDOT chains. The obtained MD results are discussed to shine the light at recent experimental data reported for these systems.

In the past several years, Materials Genome Initiative (MGI) efforts have produced several examples of computationally designed materials in the fields of energy storage, catalysis, thermoelectrics, and hydrogen storage, as well as large data resources that can be used to screen for potentially transformative compounds. In this work, a data-driven computational framework is devised to extend computational screening efforts to materials synthesis routes. A novel software package, developed in the Python programming language, is used to extract synthesis parameters of materials from journal articles at scale using state-of-the-art natural language processing techniques. In this paper, we describe the first results of our compiled database of materials synthesis pathways, which we present in a web-accessible format.

The search for high strength materials to harden and toughen surfaces and for wear-resistant coatings has propelled the need to understand structural and mechanical properties of novel transition metal nitrides. While rocksalt phase TiN has been well studied, some other transition metal nitrides are relatively unexplored. WN is a potential hard-coating candidate due to the high melting point and hardness of tungsten and the high valence electron density of states which promises a high toughness in comparison to TiN. First principle calculations are employed to compare structural and mechanical properties of two cubic phases, the experimentally reported rocksalt and theoretically predicted NbO phase of WN. The rocksalt phase is both mechanically and thermodynamically unstable with a formation enthalpy H_f = 0.637 eV per formula unit (f.u.) and a single crystal shear modulus c₄₄ of -86 GPa. In contrast, the NbO phase is both mechanically and thermodynamically stable with H_f = �0.825 eV/f.u. and c₄₄ = 175 GPa. To explore the reasons for the mechanical instability of rocksalt WN, charge distribution and electronic density of states analyses are performed, revealing that the application of shear strain along [100] in rocksalt WN results in an increased overlap of t_2g orbitals which cause electron migration from the expanded [110] to the shortened [1^–10] direction. This electron migration populates states near the Fermi level yielding an energy reduction and therefore a negative shear modulus. Corresponding shear strain in the NbO phase causes an increase in energy and a positive shear modulus. The mechanical transition from the unstable NaCl to the stable NbO phase is further explored using supercell calculations of the NaCl structure containing C_v = 0 to 25 % cation and anion vacancies while keeping the cation to anion ratio constant at unity. The structure is mechanically unstable for C_v <5 %. At this critical vacancy concentration, the isotropic elastic modulus is zero but increases steeply to 445 GPa for C_v = 10 % and to 561 GPa for the NbO structure with C_v = 25 %. Correspondingly, the Vicker’s hardness estimated using Chen’s model, varies from 0 to 14 to 26 GPa as C_v increases from 5% to 10% to 25%. These results show that the experimental reports of a cubic WN phase can be explained by the mechanical stabilization of the rocksalt phase by a minimum of 5% anion and cation vacancies.
Keywords: Tungsten nitride, vacancy induced stabilization, shear.

Thermodynamic properties of hight temperature phase of NiTi and PtTi are calculated, using our recently developed P^4 method. The austenite phase of NiTi and PtTi exhibit harmonic phonon instabilities, making the standard lattice dynamics approaches insufficient to calculate their free energy. In the P^4 method, we propose to explore the potential energy surface by discrete sampling of local minima, surrounding the high-symmetry time-averaged structure, via a lattice gas Monte Carlo approach and by a continuous sampling via a harmonic lattice dynamic approach in the vicinity of each local minima.
The simple extension of the proposed P^4 method to solid solutions makes the calculation of technologically important compounds like NiTi and PtTi computationally feasible.

Crystallization is an important procedure for the purification and isolation of specialty chemicals. For new molecular entities, understanding their inherent tendency to crystallize is a key step for developing efficient industrial processes. To date, understanding the crystallization propensity of molecular solids has been primarily driven by empirical approaches. These studies show that crystallization may be influenced by a wide variety of processing parameters, including those of a structural, thermal, chemical and kinetic nature. Unfortunately, detailed trial-and-error studies that evaluate crystallization propensities for a diverse range of compounds have been lacking because they are expensive, time-consuming and inefficient, but more importantly failed experiments are rarely reported in the literature. Computational data-driven approaches based on a complete set of historical data coupled with machine learning methods would be invaluable to a number of industries that depend on successful crystallizations of new molecular entities. This presentation details the work that has been carried out towards the compilation and use of this type of historical data.
To achieve these goals, we have sourced over a decade worth of crystallization experiments from Pfizer and built a mineable database. With this database in hand we have developed machine-learning algorithms to predict the success or failure of crystallization for compounds within the database. The machine learning approaches have been trained using both traditional molecular-based descriptors and process-based descriptors such as experiment type and solvent-system. We have assessed the ability of different classification and regression models such as Random Forest and Support Vector Machine algorithms to predict the experimental outcomes. The major factors significant to predict crystallizability will also be described in this presentation.

Carbon nanotube synthesis critically depends on the chemical and physical states of the catalyst particle from which they grow. In the typical temperature range (900-1300 K) of SWNT synthesis, pure isolated Ni nanoparticles are solid although atomic diffusion can lead to significant shape changes. Under growth conditions, these nanoparticles are exposed to reactive carbon. Depending on temperature, carbon chemical potential and nanoparticle size, carbon can either stay adsorbed on the surface, or diffuse to subsurface or in the core of the nanoparticle, thereby inducing a partial or complete melting.
On the basis of the tight binding model developed for the Ni-C system coupled with grand canonical Monte Carlo simulations [1], we extend our previous calculations [2, 3] and calculate phase diagrams for Ni-C nanoparticles for sizes ranging from 1 to 3 nm diameter and for face centered cubic and icosahedral structures. As compared to bulk phase diagram, the nanometric size of the nanoparticles used to catalyze SWNT growth induces significant differences. A large liquid shell / crystalline core domain appears instead of the liquid/solid coexistence characteristic of the bulk. Much deeper eutectic points are observed and, for a 3 nm diameter nanoparticle, as shown in the figure, carbon segregation from a mostly solid surface only takes place at temperatures below 850 K, at which SWNT growth is not really efficient [5].
References
[1] H. Amara et al., Phys. Rev. B 79, 014109 (2009).
[2] M. Diarra et al., Phys. Stat. Sol. B 249, 12, 2629 (2012) .
[3] M. Diarra et al., Phys. Rev. Lett. 109, 185501 (2012).
[4] P. Steinhardt et al., Phys. Rev. B, 28, 2, 784 (1983).
[5] Y. Magnin et al., Phys. Rev. Lett. 115 205502 (2015)

Surface segregation leads to chemical disordering in magnetic alloy nanostructures and thus could have profound impact upon the magnetic properties of these nanostructures. In this study, we used the first-principles density functional theory calculation method to determine how Pt surface segregation would affect the magnetic properties of L1₀ ordered CoPt and FePt nanoparticles. For both CoPt and FePt nanoparticles, it was found energetically favorable to exchange the surface Fe (or Co) atoms with the interior Pt atoms and to induce the Pt surface segregation. Comparing the magnetic properties of the bulk-terminated and surface-segregated nanoparticles, we predicted that the surface segregation processes in the CoPt and FePt nanoparticles could cause a decrease in their total magnetic moments, a change in their (easy and/or hard) magnetization axes, and a reduction in their magnetic anisotropy. Furthermore, we identify that the Fe (or Co) atom leaving away from their sublattice sites is most responsible for the changes in the magnetic properties of the surface-segregated nanoparticles. Comparing CoPt and FePt, our calculations reveal that the fashions of surface magnetism canting are different on the surface of L1₀ FePt and L1₀ CoPt nanoparticles even with the same shape and size. For computational design of CoPt and FePt nanoparticles with desired magnetic properties, we investigated how the Cu doping and the particle shape would affect surface segregation. By tailoring the extent of surface segregation, the magnetic properties of CoPt and FePt nanoparticles could therefore be controlled.

Nanostructured metallic multilayers (NMMs) can serve as a test stage for understanding coupling between various strengthening mechanisms due to the ability to accurately control composition by magnetron sputtering syntheses. Often conventional large-scale strengthening mechanisms have been considered to be additive. For instance if the strength enhancement in a material due to the Hall-Petch mechanism as a function of grain size, is known, and the strengthening from Orowan strengthening due to precipitates or solid solution strengthening is known, to first order these are often added. In NMMs there is an additional strengthening mechanism of confined layer slip (CLS). However, it is not clear that these should always be additive, particular at high strength levels. Even some of the advanced approximations (e.g. Clyne method) is found to overestimate the combine effect. As there is no universal description of these mechanisms, the accuracy of any prediction can be challenged when there exist multiple strengthening mechanisms affecting each other. To test these relationships, a NMM was fabricated using a 30 nm layer thickness as a bi-layer of Cr and Cu with 2%Cr multilayer by magnetron sputtering. By annealing the samples Cr is precipitated in the Cu layer, and decreases solid solution content while increase precipitate content. Nanoindentation has been used to measure hardness changes as a function of annealing condition; moderate temperature anneals of 30 minutes at 383K in an inert environment lead to an increase in strength from 7 to 8.5 GPa. We then investigate the consequences of the simultaneous presence of multiple mechanisms by using molecular dynamics to establish relationships between CLS and precipitation hardening in NMMs.

Heusler compounds are being widely studied for their potential usage in spintronics, shape-memory devices, superconductors, thermoelectrics, topological insulators, etc. The crystal structure and its variants are ubiquitous, with more than 1000 Heusler compounds being reported. However, the phase space for possible Heusler compounds is orders of magnitude larger, raising the real possibility that many new, stable Heusler compounds are still awaiting discovery. We demonstrate a high-throughput computational DFT screening approach for ~200,000 potential Heusler compounds, and use this method to predict hundreds of new stable and metastable Heusler compounds, which we further examine for interesting functional properties. We highlight three distinct examples of computational discovery of Heusler compounds, demonstrating the extraordinary diversity of properties possible with this single structure type:
1) Efficient Thermoelectrics - We report the computational discovery of a class of hitherto unknown stable semiconducting full-Heusler compounds with ten valence electrons (X2YZ, X=Ca, Sr, and Ba; Y= Au and Hg; Z=Sn, Pb, As, Sb, and Bi) through high-throughput ab−initio screening. These new compounds exhibit ultralow lattice thermal conductivity, while preserving high power factors, and are thus promising high-efficiency thermoelectrics.
2) Off-stoichiometric Semiconducting Heuslers - Fe_2-xTiSb compounds were explored via a first-principles DFT based binary cluster expansion of Fe and vacancies on the Fe sublattice of the Heusler structure. Using the cluster expansion, we predict a novel, stable, nonmagnetic semiconductor phase with a structure based on an ordered arrangement of Fe and vacancies with composition Fe_1.5TiSb, i.e., between the full- and half-Heusler compositions.
3) Strengthening Precipitates - We present a search strategy for coherent precipitate Heusler phases. We screen for precipitates that either have a stable two-phase equilibrium with the host matrix, or are likely to precipitate as metastable phases. Our search produces known high-strength precipitates and also a number of new, currently-unknown promising precipitate/alloy systems.

There is huge effort in searching for high-performance thermoelectric(TE) materials. For that, understanding the thermal and electrical transports in complex materials is one of the key scientific issues to be pursued. In China, there are quite a few government-supported projects aiming at the discovery of new TE materials through in-depth understanding of the relationship between structure and electron/phonon transports by an integrated theory-experiment approach. This talk will specially focus on understanding thermal transport especially the diverse lattice dynamics and thermal transport in compound materials with different level of structural complexity. For the compounds with relatively simple chemical bond, the traditional approach to describe the thermal transport is based on the classical concept of “phonon” and the nonlinear phonon-phonon interactions such as three-phonon scattering. A concept of part-crystalline part-liquid state (or liquid-like), and even part-crystalline part-glass state (or glass-like), was demonstrated in some materials such as Cu₃SbSe₃ and Cu₂(S,Se) with chemical-bond hierarchy, in which certain constituent species weakly bond to other part of the crystal. Such a type of materials could intrinsically manifest the coexistence of rigid crystalline sublattices and other fluctuating noncrystalline sublattices with thermally induced large amplitude vibrations and even flow of the subgroup of species atoms. The large-amplitude vibrations and liquid-like flow of atoms can generate unusual severe phonon scattering and thermal damping due to the collective low-frequency vibrations similar to the Boson peak in amorphous or liquid materials, leading to the phenomenon of “phonon” scattering beyond the traditional anharmonicity.

The design of new materials is a key challenge in materials science: e.g. new materials for the sustainable production/storage/conversion of energy carriers are necessary to improve existing and to enable future energy systems. Design and of materials includes strategies for efficient discovery and optimization of new materials. A combination of computational and experimental material science, both applying high-throughput methods, is promising. By implementing and optimizing the combinatorial materials science approach in our group during the last ten years, we are trying to contribute to this development. It comprises the fabrication and processing of thin film materials libraries by combinatorial sputter deposition processes (40 elements available) and optional post-deposition treatments (e.g. thermal oxidation, annealing, dealloying), followed by the high-throughput characterization of the different thin film samples contained in these libraries, and in making the next step by the up-scaling of findings from materials libraries to uniform, single-composition samples or to larger, bulk material dimensions. The importance of defining adequate screening parameters and the according design of different materials libraries suitable for one or more screening parameters will be addressed. Our high-throughput material characterization methods are automated, fast, and mostly non-destructive: examples are EDX and RBS for composition, XRD for crystal structure, temperature-dependent resistance for phase transformation, high-throughput test stands for optical properties (color, transmission) and mechanical properties (stress, hardness, elastic modulus), and scanning droplet cells for photoelectrochemical properties screening. The obtained results for ternary and quaternary systems are visualized in the form of composition-processing-structure-function diagrams, interlinking compositional data with structural and functional properties. The talk will cover and discuss examples of the combinatorial development of intermetallic materials for superalloys and thermoelectric applications (Ti-Ni-Sn) as well as the development of metal oxide thin film materials libraries for solar water splitting (Fe-W-Ti-O, Fe-Al-Cr-O). Examples of cooperation with (high-throughput) computational materials science groups will be given. Finally, the importance of developing new materials – not just for themselves but to be part of a system – will be highlighted.
Funding of the German Research Foundation (DFG) is acknowledged

Many thermoelectric materials are multi-component systems that exhibit chemical ordering, which can affect both thermodynamic and transport properties. Here, we address the coupling between order and thermoelectric performance in the case of a prototypical inorganic clathrate (Ba₈Ga₁₆Ge₃₀) using a combination of density functional and Boltzmann transport theory as well as alloy cluster expansions and Monte Carlo simulations. The calculations describe the experimentally observed site occupancy factors and reproduce experimental data for the transport coefficients. By inverting the cluster expansion we then identify chemical ordering patterns that increase the power factor by more than 60%. This enhancement is traced to specific features of the electronic band structure. The approach taken in the present work can be readily adapted to other materials and enables a very general form of band structure engineering. In this fashion it can guide the computational design of compounds with optimal transport properties.

There is growing momentum in the automotive industry to harvest energy from the exhaust using solid state power conversion processes, which can convert kilowatt levels of electrical power from the vehicle’s engine. Solid state power conversion devices, such as thermoelectrics, depend solely upon the temperature gradients for their operation. Aeronautic gas turbine engines have temperature gradients as well throughout due to the enthalpy processes of combustion, which offers the possibility of generating electrical power for use in primary and secondary electrical systems in the aircraft. However, currently available thermoelectric materials do not possess the environmental durability and performance levels necessary to realize these benefits. New materials must be developed that can meet the requirements to harvest waste enthalpy from gas-turbine engines. Computational methods offers an efficient and systematic manner to design new materials and guide there development. A computational -based material approach was used to determine the suitability of ruthenium-doped gadolinium orthotantalate (GdTa _(1-x) Ru_xO4) as a practical thermoelectric material. The calculations were carried out using a projector augmented wave (PAW) method using a commercial code (Materials Design Inc.) MedeA incorporating the Vienna Ab-initio Simulation Package (VASP) as the computational engine. The calculation were based on density functional theory using the GGA-PBE exchange-correlation functional using an optimized mesh. This study makes predictions and comparison between experimental and theoretical data of electrical, structural, and crystallographic properties.

Quasi low-dimensional structures are abundant among known thermoelectric materials, primarily because of their low lattice thermal conductivities. In this work, we focus on quasi-2D binary and ternary compounds reported in the Inorganic Crystal Structure Database (ICSD) and computationally assess their potential for thermoelectric performance by using an improved version of our previously-developed descriptor (β_SE) for thermoelectric performance. The descriptor can be determined from first-principles density functional theory (DFT) calculations. The quasi-2D structures are identified using an algorithm we have developed specifically for the purpose of identifying quasi low-dimensional structures. The improved version of the descriptor explicitly accounts for van der Waals (vdW) interactions, which are present in quasi-2D materials and are poorly described in standard DFT functionals. As a consequence, we are able to more accurately predict phonon and charge transport in quasi-2D materials. With our improved descriptor, we correctly identify known quasi-2D thermoelectric materials such as SnSe, SnS, Bi₂Te₃ and Sb₂Te₃. We identify potentially promising candidates, a number of which have not been previously considered for thermoelectric applications. These include GeAs₂ (space group 55), In₂Te₅ (s.g. 9), as well as PbS (s.g. 36), PbSe (s.g. 62) in their high-temperature/high-pressure phases that are promising candidates for both p- and n-type doping. The approach presented in this work can be extended to other low-dimensional materials as well as other chemistries.

SnTe, a crystal-symmetry-protected topological insulator, is of interested in thermoelectric applications due to its outstanding ZT value. In this work, we investigate the electron-phonon interactions, which dominates the intrinsic electron transport including carrier’s mobility and Seebeck coefficient, using fully ab initio calculations. The effects of spin-orbital coupling on the electrical transport properties, especially electron mean-free-path, are quantitatively studied. We discuss how the topological property of SnTe is linked to its electron transport, both from the perspective of band structure and electron scatterings. This research offers the fundamental knowledges in designing thermoelectric devices and spintronic components. This work is supported by DOE EFRC (Grant No. DE-SC0001299).

Over the past decade, research into materials for thermoelectric (TE) applications has shown remarkable progress. The goal is to increase the thermoelectric figure-of-merit, ZT, which depends on both the electronic and the thermal transport properties, and today we have a variety of good TE materials with ZT in the range of 1 to almost 3. While theoretical modeling has become more sophisticated and accurate, the majority of advancements stem from experimental trial-and-error. To enable theory-driven innovation, a challenge is to develop fully predictive and rigorous approaches for calculating thermoelectric parameters that will provide insights into the material physics and help guide experimental discovery.

In this talk, I will present our progress towards a first principles-based capability to predict the electro-thermal transport and energy conversion properties of materials. Our approach is based on density functional theory (DFT), which is used to extract detailed material characteristics including electron and phonon dispersions, and scattering physics. For electrons and phonons the most common scattering processes are electron-phonon, impurity and 3-phonon scatterings, which need to be carefully treated. The DFT-calculated material properties are then combined with the Boltzmann transport equation to compute the charge and heat transport properties, such as the electrical conductivity, Seebeck coefficient, electronic and lattice thermal conductivities, and finally ZT. This work will help advance thermoelectrics, and other applications related to electronics and heat management, by providing fundamental insight, guiding experiment efforts and accelerating innovation through materials design.

Over the last two decades, thermoelectric materials have seen increasing improvements in their figure of merit zT mostly through lowering of the phonon thermal conductivity by either nanostructuring techniques or material’s intrinsic anharmonicity. Continued improvements in zT are needed for thermoelectric technology to compete with other energy conversion technologies. In this work, we study the thermoelectric property of half-Heusler materials using first principles calculations. We seek to enhance the electrical transport property, namely the power factor, which is the product of the electrical conductivity and the square of Seebeck coefficient. Using first principles method, we study the intrinsic electron transport properties of several half-Heusler materials, governed by electron-phonon interaction. We uncover that, besides of the traditional wisdom in selecting materials based on effective mass (band structure engineering), the electron scattering magnitude is also a determining factor. By combining the study of their band structure and electron scatterings, we discuss how one can find better new materials. We believe the obtained physical insight is not limited to this material system and may potentially help to find materials with better thermoelectric performance. This work is supported by DOE EFRC (Grant No. DE-SC0001299).

Grain boundaries (GBs) form a complex defect network whose collective structure strongly influences the mechanical, physical, and transport properties of materials. Attempts to quantify the structure of GB networks have employed metrics such as cluster sizes, Betti numbers, and triple junction fractions. However, all of these methods rely on the binary classification of GBs as either “special” or “general” and represent assumptions about what constitute important structural features. Microstructure design in the context of GB networks has been severely inhibited by the absence of a general structural metric. We present a new approach for characterizing the structure of GB networks based on eigendecomposition of graph representations of polycrystals. This method is capable of handling a continuous spectrum of GB types and naturally indicates the important features of the microstructure that govern the effective response of the polycrystal. We demonstrate the correlation between the natural measures of structure that emerge and the effective properties of model polycrystals and discuss the implications and potential for designing GB networks with tailored properties.

Several computational and experimental analyses rely on relative atomic velocities to identify excitation of bonds. These are often interpreted through the aid of normal mode calculations. We address two potential issues with this approach: 1. Relative velocities may not reflect the total potential energy absorbed by the bond. 2. Normal mode calculations do not include interactions between neighboring molecules, effectively assuming a similarity between condensed- and gas-phase chemistry. Lattice dynamics (LD) calculations of phonon states can be used to address the second issue, as they include interactions between neighboring molecules. We present here a means to extend LD calculations to partially address the first issue as well.

In a recent paper, co-authored by the present author[1], we introduced a new technique for analyzing phonon modes by computing the potential energy fractions of an ensemble of phonon modes calculated using the LD method. Applying this analysis to the energetic molecular solid α-RDX, we focused on the excitation of the N-N bonds, which are believed to play a central part in the solid’s chemical decomposition. We found that the phonon modes with the largest N-N bond excitations were not those analogous to the high-frequency normal modes corresponding to N-N vibrations, but among the low-frequency, intermolecular phonon modes typically associated with relative molecular vibrations.

In the present work, we expand the technique by comparing kinetic and potential energy fractions. We find that while there is significant potential energy in the N-N bonds among low frequency phonon modes, a negligible portion of the kinetic energy of these same modes is found in the vibrations of the corresponding N-N pairs. We suggest that this apparent contradiction may be explained through analogy the human femur during running. Despite significant stress, each femur deforms very little, and its vibrational kinetic energy represents a small fraction of the total kinetic energy of the human body, in the body’s center-of-mass frame. Similarly, the N-N bonds in α-RDX are stiff relative to many interactions in α-RDX, and thus act as a means to absorb stress while deforming relatively slowly during interactions between neighboring molecules. This implies that chemistry in the solid, and likely liquid, phase my involve interactions that cannot be identified through analyses that rely on relative atomic velocities.

1. B Kraczek and P Chung, J Chem Phys 138, 074505 (2013).

The downscaling of electronic devices has reached the range where the number of atoms in critical dimensions is countable, geometries are formed in three dimensions and new materials are being introduced. Under these conditions one can argue that the overall geometry constitutes a new material that cannot be found as such in nature and the distinction between new device and new material are blurry. The interactions of electronic, photons, and lattice vibrations are now governed by these new material properties and longer-range interaction mechanisms such as strain and gate fields. The Nanoelectronic Modeling tool suite NEMO5 is aimed to comprehend the critical multi-scale, multi-physics phenomena and deliver results to engineers, scientists, and students through efficient computational approaches. NEMO5’s general software framework easily includes any kind of atomistic model and is, insofar, able to compute atomistic strain, electronics band structures, charge density, current and potential, Schrödinger eigenvalues and wave-functions, phonon spectra, and non-equilibrium Green functions (NEGF) transport for a large variety of semiconductor materials and the software is entirely parallelized. We believe that such modeling capability is not available in any other modeling tool at this time.
In general NEMO5 uses various forms of orthogonal or non-orthogonal tight binding to represent the valence electrons of different materials. These basis sets are either obtained from the iterature or mapped from density functional theories (DFT) through a low rank approximation. These basis sets are utilized to represent experimentally relevant devices including typically 100,000 to 50 million atoms. NEMO5 can also import Wannier 90 functions directly from DFT codes, to skip the explicit mapping process. Atom position are typically determined by position relaxation within a Valence Force Field (VFF) / Keating model. Atom positions can also be computed in external applications and read-in explicitly into NEMO5. Electron transport through multiple quantum mechanical domains is based on the Non-Equilibrium Green function (NEGF) method.
NEMO5 algorithm and physics developments in the group span a wide range of devices and concepts from work with the leading semiconductor industries on technologies for the sub 10nm transistors, over optical devices to improve lighting, to foundational device physics for quantum computing in Silicon.

First principles quantum mechanical modeling schemes, such as Density Functional Theory (DFT), have been applied to a wide spectrum of materials characterization and/or property prediction problems. However, owing to large computational costs, traditional ab initio methods are, in a practical sense, limited to small length (<10nm) and time scales (<1ns), restricting their capability to model realistic systems. In our recent work, we proposed a novel machine learning based methodology that ‘learns’ from reference quantum mechanical data to predict atomic forces directly, given only the atomic configuration [1,2]. This scheme was able to successfully reproduce the vacancy and adatom diffusion coefficient, the phonon band structure and the thermodynamic properties of Al with quantum mechanical accuracy, while reducing the computational cost by roughly six orders of magnitude. In the present work we go beyond these properties, validating our scheme for diverse materials problems such as the melting behavior and the stress-strain behavior of Al. The results were found to be in quantitative agreement with those of ab initio methods and/or the semi-empirical potentials. This machine-learning scheme was also successfully extended to several different elemental systems (W, C, Si, Ti) with varying classes of crystal structures (FCC, BCC, HCP) and showcases the generalizability of our scheme. This work demonstrates that our regression-based machine learning approach can be used to reproduce the thermodynamic, mechanical, and other important properties of materials at quantum mechanical accuracy without sacrificing the speed of interatomic potentials, and its generalizability makes it suitable to explore various different chemical environments.



[1] V. Botu, R. Ramprasad, Int. J. Quantum Chem. 2014, DOI: 10.1002/qua.24836.
[2] V. Botu, R. Ramprasad, Phys. Rev. B. 92, 094306S.
[3] F. Dinga, K. Bolton, and A. Rosen, Eur. Phys. J. D 2005, 34, 275-277.

The diffusion of hydrogen in metals has strong implications on the irreversible damages of the host matrix. Therefore, the diffusion coefficient of hydrogen in metals is a fundamental data to design new protections preventing hydrogen embrittlement and safety materials. However, the apparent diffusion coefficient can be influenced by the presence of crystalline defects such as point defects, dislocations or grain boundaries. These defects induce long-range elastic displacement fields and non-linear displacements close to the defect core, which may accelerate or slow down the hydrogen diffusion through a trapping phenomenon. In this study, we determine the kinetic rates for the trapping and detrapping of H in a vacancy core and in the elastic displacement field induced by the defect in nickel up to the melting point from first principles calculations.

First, we determine the diffusion coefficient of hydrogen in an unconstrained Ni perfect crystal. We extend the calculations to a crystal under an external stress including uniaxial tension and compression, pure and simple shears and hydrostatic pressure. Here, we use the transition state theory, similarly to the previous study of Wimmer et al. (Wimmer et al., 2008, Phys. Rev. B). The calculations are performed up to the melting point where the Gibbs free energy of the diffusion barrier is expressed as a sum of vibration and electronic contributions from the computation of the phonon dispersion curve and the electronic density of states. The diffusion coefficients in the unconstrained crystal perfect are in good agreement with previous experimental data (Katz et al., 1965, Z. Metall., Louthan et al., 1974, Acta Met., Robertson et al. ). In particular, we found that the electronic excitations tend to increase the H diffusion coefficient.

Then, we calculate the kinetic rates for the trapping and detrapping of hydrogen in the displacement field induced by a vacancy and around the defect core. A close agreement is found between the detrapping energy calculated in this study with previous thermal desorption spectroscopy measurements (Lee and Lee, 1986, Metall. Trans. A; Oudriss et al., 2012, Acta Mat.). The contribution of the stress field induced by the defect is discussed from the results obtained for the perfect crystal under stress and the thermodynamic of stressed solids (Larché and Cahn, 1985, Acta Metall.). Finally, we attempt to define the apparent diffusion coefficient in the crystal in presence of vacancies from the solution of a set of equations describing the evolution of mobile and trapped hydrogen atoms and from a lattice-Boltzmann method where the local diffusion coefficients are implemented on each node of a 3D grid. The results from both methods are discussed and compared with experiments associated with sorption and desorption regimes.

The combination of transparency to visible light (optical band gaps in excess of 3.1 eV) and high electrical conductivity (conductivities > 10³ S/cm) in a metal oxide material is quite unusual as these are normally mutually exclusive properties.[1] This combination has been realised, however, for a small subset of metal oxides, termed transparent conducting oxides (TCOs), such as ZnO, CdO, Ga₂O₃, SnO₂, In₂O₃, BaSnO₃ etc.[2-5] The cationic species in these materials are all “post transition metals” and so all possesses a (n–1)d¹⁰ns⁰np⁰ electronic structure. The hybridization between the cations s states and the O p states typically yield low lying conduction band minima (high electron affinities) with excellent conduction band dispersion (low effective masses – high electron mobility).[6] Another cation which possesses an (n–1)d¹⁰ns⁰np⁰ electronic structure is Sb(V), however, oxides featuring Sb(V) have not enjoyed a huge amount of study to date. Recently, however, ZnSb₂O₆ has been identified from a high throughput computational screening to have an electronic structure that could be ideal for TCO applications.[7] In this presentation, we will detail a computational screening of all existing Sb(V) containing oxides for TCO and power electronics applications. We will highlight the importance of the connectivity of the SbO₆ octahedra to band dispersion, and pinpoint a number of ternary Sb(V) oxides that warrant further experimental study.

[1] P. D. C. King and T. D. Veal, J. Phys.: Condens. Matter, 23, 334214 (2011)
[2] M. Burbano, D. O. Scanlon, and G. W. Watson, J. Am. Chem. Soc. 133, 15065 (2011).
[3] D. O. Scanlon and G.W.Watson, J. Mater. Chem. 22, 25236 (2012)
[4] A. Walsh et al., Phys. Rev. Lett. 100, 167402 (2008)
[5] Z. Leben-Higgins et al., Phys. Rev. Lett., 116, 027602 (2016)
[6] H. Mizoguchi and P. M. Woodward, Chem. Mater, 16, 5233 (2004)
[7] G. Hautier et al., Chem, Mater, 26, 5447 (2014)

Nowadays state-of-the art DFT codes and high-throughput frameworks allow us to compute materials properties for large data sets. The Material Project¹ (MP) is one of the biggest project that aims to compute and share structures and properties of materials. As recently made for elastic² and piezoelectric³ tensors, we will present a large and freely accessible data set of transport related properties (effective mass, Seebeck, electronic thermal conductivity, …). This transport data has been computed on top of energy band structures available in MP, using the well-known BoltzTraP⁴ code inserted in a HT framework. Given the importance of electronic transport properties, the whole community of material science researcher will benefit from this database. We will present the work flow to obtain the data and the data set. We will also study some of the correlation between transport properties and present applications in the field of transparent conducting oxides and thermoelectric materials.

References:
(1) Jain, A.; Ong, S. P.; Hautier, G.; Chen, W.; Richards, W. D.; Dacek, S.; Cholia, S.; Gunter, D.; Skinner, D.; Ceder, G.; Persson, K. A. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation, APL Mater., 2013, 1, 011002, doi:10.1063/1.4812323.

(2) M. de Jong, W. Chen, T. Angsten, A. Jain, R. Notestine, A. Gamst, M. Sluiter, C. K. Ande, S. van der Zwaag, J. J. Plata, C. Toher, S. Curtarolo, G. Ceder, K. A. Persson, M. Asta; Charting the complete elastic properties of inorganic crystalline compounds, Scientific Data 2: 150009 (2015), doi:10.1038/sdata.2015.9

(3) M. de Jong, W. Chen, H. Geerlings, M. Asta, K. A. Persson; A database to enable discovery and design of piezoelectric materials, Scientific Data 2: 150053 (2015), doi:10.1038/sdata.2015.53

(4) Madsen, G. K. H.; Singh, D. J. BoltzTraP. A code for calculating band-structure dependent quantities, Comput. Phys. Commun., 2006, 175, 67–71, doi:10.1016/j.cpc.2006.03.007.

Phonon processes are well studied and understood for bulk semiconductors, however with the increasing prevalence of bottom-up fabrication of semiconductors from nanomaterials, we must consider what happens to phonons and electron-phonon coupling when materials are nanosized and how this knowledge can be leveraged to make better devices. Here we present how computational methods can provide insights into phonons in nanoscale materials and the behavior of these nanosized materials in devices.
We perform Ab-Initio-Molecular-Dynamics (AIMD) on PbS semiconductor nanocrystals (NCs), the most widely used material in NC-solid solar cells. The phonon density of states can be calculated from the time-dependent coordinates of the atoms extracted via AIMD. We have shown that our AIMD results are consistent with and shed light on experimental inelastic neutron scattering (INS) data [1]. We find that soft surfaces of the NCs give rise to both low and high energy phonon modes possessing large thermal displacements not found in bulk PbS. Analysis on the time-dependent energies and wavefunctions indicate that these modes couple strongly to the electronic states of the NCs. We show results for PbS NCs constructed with various surface passivating ligands, and analyze the impact of ligands on the electronic and phononic properties of the NCs.

[1] Bozyigit, D., et al. (2016). Soft surfaces of nanomaterials enable strong phonon interactions. Nature, (7596), 618–622.

Currently the creation and quantification of interatomic potentials for classical molecular dynamics calculations can be tedious. We present a high-throughput reproducible framework with a publicly available RESTful interface and website to potentials and their predicted properties. We use our open-source Python package DFTFIT to create interatomic potentials for molecular dynamics (MD) from density functional theory (DFT). It implements a popular variation of the Force-Matching algorithm, which is based on least square optimization. Our software is novel because of how it tightly integrates with mainstream DFT and MD packages, such as VASP, Quantum Espresso, and LAMMPS. One key benefit of this approach is that DFTFIT improves as more potentials become available in MD packages. We demonstrate the software on two model systems MgO and SiC. The potentials for both systems show improvement, which validates the effectiveness of the software by fitting two-body and three-body potentials. Additionally we will demonstrate the web interface by evaluating the potentials available in the NIST Interatomic Potentials Repository.

Self-consistent field theory (SCFT) has proven to be a powerful tool for modeling equilibrium microstructures of soft materials, particularly for multiblock polymers. A very successful approach to numerically solving the SCFT set of equations is based on using a spectral approach. While widely successful, this approach has limitations especially in the context of current technologically relevant applications. These limitations include non-trivial approaches for modeling complex geometries, difficulties in extending to non-periodic domains, as well as non-trivial extensions for spatial adaptivity. As a viable alternative to spectral schemes, we develop a finite element formulation of the SCFT paradigm for calculating equilibrium polymer morphologies. This method was designed to support massively parallel simulations. Utilizing this and the computational power of the Blue Waters super computer, we show the ability to investigate semi-flexible polymer chains which are of great industrial and scientific interest. Until recently, the computational complexity of the class of polymers has prevented simulation in all but the most limited cases. This new approach has enabled us to study diblock copolymers of varying stiffness with no limitations on the geometry. We present the details of the framework and selected results for semi-flexible chains.

We present the predictive modeling and experimental confirmation of the microphase separation of ternary polymer brushes. While block copolymers are the traditional system to study microphase separation, mixed-polymer brushes can also self-assemble into distinct lateral domains. Mixed-polymer brushes offer several advantages over traditional block copolymer systems. First, since the polymers are directly attached to the surface, they are more mechanically and chemically robust. Also the surface does not need to be planar, as when the brushes grow from the surface, any surface geometry can be used. Finally, features such as right angles, which are difficult for block copolymer systems, should be readily accessible with a mixed-polymer brush system. We have synthesized ternary polymer brushes consisting of polystyrene, poly (methyl methacrylate), and poly (4-vinyl pyridine). By using self-consistent field theory (SCFT) and experimental results, we have been able to predict a complete phase diagram, identify seven unique phase behaviors, and model both lateral and vertical phase behavior of this system. All phases observed experimentally correlate with the theoretical models. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S.Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.

Fibers made from polymer blends have conventionally enjoyed wide use, particularly
in textiles. This wide applicability is primarily aided by the ease of manufacturing such
fibers. More recently, the ability to tailor the internal morphology of polymer blend fibers
by carefully designing processing conditions has enabled such fibers to be used in
technologically relevant applications. Some examples include anisotropic insulating
properties for heat and anisotropic wicking of moisture, coaxial morphologies for
optical applications as well as fibers with high internal surface area for filtration and
catalysis applications. However, identifying the appropriate processing conditions from
the large space of possibilities using conventional trial-and-error approaches is a tedious
and resource-intensive process. Here, we illustrate a high throughput computational
approach to rapidly explore and characterize how processing conditions (specifically
blend ratio and evaporation rates) affect the internal morphology of polymer blends
during solvent based fabrication. We focus on a PS : PMMA system and identify
two distinct classes of morphologies formed due to variations in the processing
conditions. We subsequently map the processing conditions to the morphology
class, thus constructing a “phase diagram” that enables rapid identification of processing
parameters for specific morphology class. We finally demonstrate the potential for time
dependant processing conditions to get desired features of the morphology. This opens
up the possibility of rational stage-wise design of processing pathways for tailored fiber
morphology using high throughput computing.

A primary bottleneck for heat transfer in bulk linear polymers is believed to be weak inter-chain bonds, which are typically formed by weak Coulombic or van der Waals interactions. Recent experimental studies have shown an increase in thermal conductivity for certain polymers upon strengthening these inter-chain interactions, while for others an increase in thermal conductivity was not observed. For example, it has been shown that hydrogen bonding has a significant impact on heat transfer in some polymer mixtures, while in others the effect of hydrogen bonds on heat transfer is negligible. Similar discrepancies have also been reported for crosslinked polymers, where strong covalent bonds (approximately two orders of magnitude stronger than vdW interactions) have been used to enhance inter-chain bonding.

We use molecular dynamics simulations to study heat transfer in polymers with various inter-chain bonding configurations in order to resolve these experimental discrepancies. In particular we examine the bulk polymer mixtures polyacrylic acid (PAA) / polyacryloyl piperidine (PAP) and polyvinyl alcohol (PVA) / poly(4,4'-phenylene-pyromellitimide). We analyze hydrogen bond formation, polymer backbone stiffness, dihedral angle dynamics, and changes in vibrational modes to determine the physical origins of thermal conductivity enhancements. We observe large increases in thermal conductivity correlated with changes in backbone dihedral angles that occur in certain instances upon hydrogen bond formation. These angles have a significant effect on the radius of gyration and hence the polymer chain stiffness, which in turn governs the transport of vibrational (thermal) energy through the backbone covalent bonds.

In further studies, we find that hydrogen bonding between polymer molecules and water residuals remaining from polymer synthesis or absorbed from ambient humidity only leads to slight increases in thermal conductivity, because water molecules do not significantly alter the polymer morphology (e.g., backbone dihedral angles). For the same reason, polymer chains that are already stiff before mixing do not gain in thermal conductivity upon hydrogen bond formation during mixing.

In this contribution, we will discuss our efforts at guiding experimentalists to discover polymeric materials that can catalyse the photocatalytic splitting of water. Traditionally, photocatalytic water splitting is the preserve of inorganic semiconductor combined with transition or noble metal co-catalysts. Already in the 1980s, however, it was demonstrated that oligomers and polymers of p-phenylene under illumination with UV light could drive the reduction of protons to hydrogen [1], and the discovery twenty years later that carbon nitride could both reduce protons and oxidise water [2], kick started the field of polymer photocatalysts in earnest. While for the moment still relatively underexplored relative to their inorganic counterparts, these polymers offer the potential of photocatalysts based on earth abundant elements, whose properties can be easily tuned through copolymerisation [3,4], exploiting the vastness of the chemical space of organic chemistry.

Together with collaborators in Liverpool we use a combination of experimental and computational high throughput screening to find (new) classes of polymers that drive photocatalytic proton reduction and/or water oxidation, an activity that has already resulted in the discovery of a range of polymers that evolve hydrogen [4,5]. We will review in our contribution our computational approach and focus on the methodological challenges of predicting the potential of amorphous polymeric material as water splitting catalysts [6,7]. Specifically, we will discuss how to calculate the thermodynamic driving force for the desired solution redox reactions, the difficulty of benchmarking this to experimental data, the issue of exciton dissociation, and how the different requirements for a successful photocatalyst can result in competing requirements for material properties.

[1] S. Yanagida, A. Kabumoto, K. Mizomoto, C. Pac, K. Yoshino, J. Chem. Soc. Chem. Commun. 474. 1985.
[2] X. Wang, K. Maeda, A. Thomas, K. Takanabe, G. Xin, J.M. Carlsson, K. Domen, M. Antonietti, Nature Mater. 8, 76, 2009.
[3] J.X. Jiang, A. Trewin, D.J. Adams, A.I. Cooper, Chem. Sci. 2, 1777, 2011.
[4] R.S. Sprick, J.X. Jiang, B. Bonillo, S. Ren, T. Ratvijitvech, P. Guiglion, M.A. Zwijnenburg, D.J. Adams, A.I. Cooper, J. Am. Chem. Soc. 137, 3265, 2015.
[5] R.S. Sprick, B. Bonillo, R. Clowes, P. Guiglion, N.J. Brownbill, B.J. Slater, F. Blanc, M.A. Zwijnenburg, D.J. Adams, A.I. Cooper, Angew. Chem. Int. Ed. 55, 1792, 2016.
[6] P. Guiglion, E. Berardo, C. Butchosa, M.C.C. Wobbe, M.A. Zwijnenburg, J. Phys. Condens. Matter. 28, 074001, 2016.
[7] P. Guiglion, C. Butchosa, M.A. Zwijnenburg, Macromol. Chem. Phys. 217, 344, 2016.

We present a physical model to investigate the formation of the nano-scale surface patterns in biological materials through the interaction of anisotropic interfacial tension, bending elasticity, and capillarity at free surfaces. Focusing on the cholesteric liquid crystal (CLC) material model, the generalized shape equation for anisotropic interfaces using the Cahn-Hoffman capillarity vector, the Rapini-Papoular anchoring energy, and Helfrich free energy are applied to understand the evolution of the biological surface patterns such as multi-length scale patterns in floral petals. The physical model incorporating liquid crystal anisotropy of biological materials and bending elasticity of surfactant-like biomolecules, exhibits an analogy with harmonic motions in a driven pendulum with two different frequencies in orthogonal directions. Depending on the bending elasticity, liquid crystal chirality and anchoring strength, we classify the surface patterns and present the corresponding phase diagram. In the absence of bending elasticity, the model predicts a regular pattern of sinusoidal wrinkles. While, in the presence of bending elasticity, the system can evolve complex spatial patterns induced by the nonlinear coupling between LC chirality and the elastic wavenumber. These new findings can not only establish a new paradigm for characterizing surface wrinkling in biological liquid crystals, but also inspire the design of functional surface structures.

Recently, the novel properties of nanocrystalline(NC) semiconductors have been used to create optical up- and down-conversion devices that can allow capturing more of the solar spectrum, through energy transfer to and from organic semiconductors(OSC)[1][2].
In this work, we produce design principles for the transfer rate between NCs and OSC from two perspectives: geometry and electronic structure. We here simulate nanocrystals using atomistic molecular dynamics, allowing for the resolution of novel structural details about ligand shell. We find that the ligands undergo a transition from being upright to laying flat. Therefore, NC-to-NC and NC-to-OSC distances differ from the expected length based on the length of the NC ligands. The end result is quantitative elucidation of the morphology of the ligand shell, and the impact of the morphology on the distances across which the energy transfer occurs, which we have corroborated using TEM. The geometries we gained from the MD allows us to calculate the electronic structure of a realistic dot. Starting from the MD geometries, we use configuration interaction with constrained density functional theory (CDFT-CI) to predict the transfer rate enchantment from the ligands.

[1] Wu, M., Congreve, D. N., Wilson, M. W., Jean, J., Geva, N., Welborn, M., Baldo, M. A. (2015). Solid-state infrared-to-visible upconversion sensitized by colloidal nanocrystals. Nature Photonics Nature Photon, 10(1), 31-34. doi:10.1038/nphoton.2015.226
[2] Thompson, N. J., Wilson, M. W., Congreve, D. N., Brown, P. R., Scherer, J. M., Bischof, T., Baldo, M. (2014). Energy harvesting of non-emissive triplet excitons in tetracene by emissive PbS nanocrystals. Nature Materials Nat Mater, 13(11), 1039-1043. doi:10.1038/nmat4097

We show in this work that the mechanical properties of molecular crystals are strongly affected by the flexibility of the constituent molecules. To this end, we explore several kinematically restrained models of the molecular crystal cyclotrimethylene trinitramine (RDX) in the α phase. We evaluate the effect of gradually removing the flexibility of the molecule on various crystal-scale parameters such as the elastic constants, the lattice parameters, the thermal expansion coefficients, the stacking fault energy and the critical stress for the motion of a dislocation (the Peierls-Nabarro stress). The values of these parameters evaluated with the fully refined, fully flexible atomistic model of the crystal are taken as reference. It is observed that the elastic constants, the lattice parameters and their dependence on pressure, and the thermal expansion coefficient can be accurately predicted with models that consider the NO2 and CH2 groups rigid, and the N-N bonds and the bonds of the triazine ring inextensible. Eliminating the dihedral flexibility of the ring leads to larger errors. The model in which the entire molecule is considered rigid or is mapped to a blob leads to even larger errors. Only the fully flexible, reference model provides accurate values for the stacking fault energy and the Peierls-Nabarro critical stress. Removing any component of the molecular flexibility leads to large errors in these parameters. These results also provide guidance for the development of coarse grained models of molecular crystals.

Solidification of ionic liquids (ILs) is used in the synthesis of optically-active and magnetic nanomaterials based on ILs (termed ‘GUMBOS’ for Group of Uniform Materials Based on Organic Salts [1-3]). Possible applications of these nanomaterials are envisioned in optoelectronics, photovoltaics, separations, analytical chemistry and biomedicine. However, very little is understood about the solidification and nucleation of solid phases of ILs, which is important as ultimately the nanostructure of the ILs determines the properties of the IL-based nanomaterials. The solidification process from a supercooled, metastable liquid phase involves first the formation of a critical nucleus, which then grows irreversibly to form a macroscopic crystal phase. The critical nucleus is an activated, short-lived species that involves only a few molecules, and thus its direct detection in experiments is very challenging. Nucleation is also an extremely challenging problem for molecular simulations, as it involves timescales beyond those accessible through classical molecular dynamics (MD)

We studied the nucleation of crystals of a common IL, 1,3-dimethylimidazolium chloride [dmim+][Cl-], from its supercooled liquid phases in the bulk and in contact with graphitic surfaces. The string method in collective variables [4,5] was used in combination with Markovian milestoning with Voronoi tessellations [5,6] and order parameters for molecular crystals [7] to sketch a minimum free energy path connecting the supercooled liquid and the crystal phases, and to determine the rates involved in the nucleation process. This particular combination of methodologies is well suited to study nucleation of ILs, as these systems exhibit very slow dynamics and very large simulation box sizes are required in order to avoid finite-size effects. We will present and discuss results [8,9] obtained for the free energy profiles, mechanisms and rates involved in the nucleation processes of these systems, aiming at understanding the nucleation and growth of crystals of organic salts in the bulk, near surfaces and inside nanopores.

[1] A. Tesfai, B. El-Zahab, A. T. Kelley, M. Li, J. C. Garno, G. A. Baker and I. M. Warner, ACS Nano 2009, 3, 3244
[2] S. Das, D. Bwambok, B. El-Zahab, J. Monk, S. L. de Rooy, S. Challa, M. Li, F. R. Hung, G. A. Baker and I. M. Warner, Langmuir 2010, 26, 12867
[3] I. M. Warner, B. El-Zahab and N. Siraj, Anal. Chem. 2014, 86, 7184
[4] L. Maragliano, A. Fischer, E. Vanden-Eijnden and G. Ciccotti, J. Chem. Phys. 2006, 125, 024106
[5] V. Ovchinnikov, M. Karplus and E. Vanden-Eijnden, J. Chem. Phys. 2011, 134, 085103
[6] L. Maragliano, E. Vanden-Eijnden and B. Roux, J. Chem. Theory Comput. 2009, 5, 2589
[7] E. E. Santiso and B. L. Trout, J. Chem. Phys. 2011, 134, 064109
[8] X. He, Y. Shen, F. R. Hung and E. E. Santiso, J. Chem. Phys. 2015, 143, 124506
[9] X. He, Y. Shen, F. R. Hung and E. E. Santiso, J. Chem. Phys. 2016 (submitted)

This study targeted four configurational parameters and investigated their contribution to the strength of Carbon Nanotube (CNT) fibers by Coarse-Grained simulation. The goal is helping to make better multifunctional fibers through fusion process. The parameters have been chosen based on observations from the fusion experiment that elucidated their relative importance during this process. Parameters include the length distribution of CNTs inside the fiber, the average degree of misalignment of individual CNTs from the fiber’s axis, the extent of the CNTs that are bundled together (bundling extent), and the fiber density. Experiments shows increase in bundling extent, local density and length of CNTs and decrease in degree of misalignment. Studying these parameters can optimize the fusion parameters.
These molecular dynamic (MD) simulations are done with Coarse-Grained model that captures tension, torsion and bending. A periodic box with randomly distributed CNTs inside is created. Stretching the simulation box along fiber’s axis is started after local equilibration and during this process the stress has been measured (tensile test). For changing density the number of CNTs inside the box and for varying length distribution the max length have been changed. Furthermore, the average degree of misalignment and bundling extent are quantified based on the system’s evolution.
From analyzing the results of multiple studies, statistically meaningful trends for improving mechanical characteristics of CNT fibers have been found. These results will be compared with experimental results to better inform the fusion process for stronger CNT fibers.

The last few decades have seen significant advancements in materials research tools, allowing researchers to rapidly synthesis and characterize large numbers of samples - a major step toward high-throughput materials discovery. Data analysis advancements originating in machine learning also allow for more rapid conversion of the large amounts of collected data into actionable knowledge. Active learning, a branch of machine learning, promises to push high throughput materials research to the next level - autonomous materials exploration. Active learning allows the researcher to take a step back, and gives the choice of the optimal next experiment to perform to the algorithm. This in turn can potentially reduce tedious labor hours, reduce equipment hours, and accelerate materials exploration. In this talk we demonstrate the use of active learning to autonomously control X-ray diffraction systems in the lab and at the beamline for phase diagram determination from composition spreads. Materials of interest include Fe-Ga-Pd, TiO2-SnO2-ZnO, and Mn-Ni-Ge.

Because of steady increase in the availability of computing power, ab initio methods of computational materials science become everyday investigation tools in various research fields. This popularity of the first-principle-based atomistic modeling is in large part due to the performance of density functional theory (DFT), which could be used for simulations of even chemically and structurally complex materials, including minerals, fluids and melts [1]. However, because of intrinsic approximations, DFT is not always able to deliver reliable predictions. This is especially pronounced for f-elements bearing materials such as nuclear materials considered in nuclear waste management. For such materials standard DFT often fails also on qualitative level, predicting metallic states for even simple actinide-dioxides that are known to be insulators. Other properties, including the reaction enthalpies, are also often badly predicted [2] . In this contribution we will discuss our experience with different computational methods, including parameter free DFT+U method in which the Hubbard U parameters is derived ab initio, for prediction of various properties of f-electrons bearing materials [2-3]. We will show significant improvement obtained for structural and thermochemical parameters of various Ln-bearing ceramic materials [3-7] and actinide-bearing molecular and solid compounds [2] when the f-electrons correlations are explicitly accounted for. The necessity of the accurate ab initio data for designing of force fields and subsequent large scale atomistic simulations will be also discussed. Last, but not least, we will demonstrate that complementary experimental and atomistic modeling studies, supported by data mining techniques, result in superior and more complete characterization of challenging materials considered in nuclear waste management.
[1] Jahn & Kowalski, Reviews in Mineralogy and Geochemistry, 78, 691-743 (2014).
[2] Beridze & Kowalski, The Journal of Physical Chemistry A, 118, 11797 (2014).
[3] Blanca-Romero et al., The Journal of Computational Chemistry, 35, 1339 (2014).
[4] Li et al., The Journal of Solid State Chemistry, 220, 137 (2014).
[5] Kowalski et al., Journal of Nuclear Materials, 464, 147–154 (2015).
[6] Kowalski & Li, Journal of the European Ceramic Society, 36, 2093-2096 (2016).
[7] Li et al., Scripta Materialia, 107, 18-21 (2015).

Recently density functional theory (DFT) calculations with predictive performance can be routinely made by an ordinary PC cluster to construct database for further studies especially for the discovery of novel functional materials. Such DFT database is already available for public uses. However, it is still challenging to construct a database for a system when existence of stable or metastable compounds is not known, because construction of the convex hull is generally quite demanding.
The aim of the present study is the exploration of compounds on the convex hull and to search possible new phases in the ternary Sn-M-O (M = Mo or W) systems through DFT calculations and synthesis experiments. Although tin molybdate and tungstate have attracted interests in the fields of photocatalysts and controllable thermal expansion materials such as SnWO₄ and SnMo₂O₈, respectively, little have been known about the phase stability of these systems. After systematic search of stable compounds, we have actually succeeded in synthesis of some compounds by experiments for validation.
Prototype structures defined in the Inorganic Crystal Structure Database (ICSD) are adopted as structural models for the calculation, which cover a wide variety of compositions and crystal structure. This enables the extensive search for hypothetical phases. DFT calculations are carried out for the whole range of SnO_x-MO_y (M = Mo and W, x = 1 and 2, y = 2 and 3) pseudo-binary compounds made by ionic exchanges in the prototype structure. The VASP code is used for DFT calculations. All prototypical structures, namely 604 structures for (x, y) = (2, 4) system, 364 structures for (x, y) = (2, 6) system, 178 structures for (x, y) = (4, 4) system, 113 structures for (x, y) = (4, 6) system are chosen. Formation energies against constituent phases, dynamical stabilities of 2,518 compounds in total are calculated. After stability evaluation of compounds by the DFT calculations at the ordinary GGA-PBE level, additional calculations with the HSE06 functional are made for the candidates that are on or nearby the convex hull of the formation energy.
Based on the results of the computational screening, we have tried to synthesize the candidate compounds using various starting materials, conditions, and synthesis techniques, such as an ordinary solid-state reaction route, hydrothermal synthesis, co-precipitation and flux methods. As a result, we have successfully synthesized some compounds that are as-yet-unknown before the present study.

The analysis of phase transitions between solid, liquid and gas phases, and the corresponding atomistic patterns during various colloidal self-assembly techniques is crucial for understanding the dynamics of none-equilibrium and far from equilibrium processes [1]. Further, patterns formed in equilibrium conditions also require similar analyses, in particular in soft matter engineering. Most commonly, the pair-correlation function is utilized to identify a phase or pattern in a specific area [2], but this approach is not always enough for recognizing competing patterns. It is also not the most direct presentation of the phases and patterns.

Here, we demonstrate an alternative method which can be used for recognizing hexagonal, square, and amorphous solid nano-pattern symmetries. The method relies on “Lindemann histograms” to reliably detect patterns in the solid phase. In addition, it can be used to recognize and quantify solid, liquid, and gas phases in colloidal self-assembled nanopatterns. The demonstrated “Lindemann histogram” approach is based on the Lindemann parameter [3] distribution calculated per particle.

This numerical approach is applied to pattern recognition and symmetry classification of patterns formed by far-from-equilibrium self-assembly colloidal particles (500 nm diameter, polystyrene beads). These experiments were performed with an ultrafast laser starting strong drag currents (i.e, marangoni flow), to dissipatively organize the colloidal particles in various symmetries, and create controlled phases and pattern transitions at the nanoscale. [4]

The advanced detection algorithm developed was used to (i) detect the particles from the experimental data, (ii) then perform multiple parameter extractions, including Lindemann histograms. Both detection and parameter extraction methods are presented in this work. We argue that the Lindemann histograms provide the best visual and numerical finger-prints of the aforementioned patterns and phases for both mono- and multi-stable (pattern competition) patterns.

[1] Peng, et al, “Two-step nucleation mechanism in solid–solid phase transitions,” Nature Naterials, 14, 101, 2014.
[2] .Veatch, et al, “Correlation functions quantify super-resolution images and estimate apparent clustering due to over-counting,” PLoS ONE , 7, 2, 2012.
[3] Chakravartya, et al, “Lindemann measures from the solid-liquid phase transition”, The Journal of Chemical Physics, 126, 2007
[4] Ilday, et al, “Control of dynamical self-assembly of strongly brownian nanoparticles through convective forces induced by ultrafast laser”, MRS spring meeting, 2016.

New, functional materials can be created based on our growing knowledge of structure-property relationships of ordered structures. On the atomic scale the design space contains a finite number of element species, while soft matter or nanoscale systems allow a continuum of building blocks. To target a specific crystal structure, we have to explore the interactions between the constituting particles and how these factors determine their assembly. We aim to understand how different structures emerge from similar systems and conditions. In particular, we are seeking to understand when and how complex structures form – such as crystals with large unit cells [1] or quasicrystals [2]. For this purpose, we study the self-assembly and phase behavior of particles with coarse-grained, isotropic pair potentials. Using the highly parallel molecular dynamics code HOOMD-blue [3], we simulate a wide range of one-component systems and observe the resulting phases. We report a rich variety of crystal structures, ranging from the well-known sphere packings and other simple structure types, to giant-unit cell structures and quasicrystals. By exploring the structures, crystal growth, and phase transition behavior of these tunable model systems, we aim at describing and understanding diverse experimental systems on the atom and soft matter length scales under the same terms.

[1] J. Dshemuchadse, D. Y. Jung, W. Steurer, Acta Crystallogr. B 67, 269-292 (2011).
[2] M. Engel, P. F. Damasceno, C. L. Phillips, S. C. Glotzer, Nature Mater. 14, 109-116 (2015).
[3] J. A. Anderson, S. C. Glotzer, http://arXiv.org/abs/1308.5587 (2013). http://glotzerlab.engin.umich.edu/hoomd-blue/

We have developed a new single-particle dissipative particle dynamics (DPD) model for anisotropic particles with different shapes, e.g., prolate or oblate spheroids. In particular, the conservative and dissipative interactions between anisotropic single DPD particles are formulated using a linear mapping from the isotropic model of spherical particles. The proper mapping operator is constructed between each interacting pair of particles at every time step. Correspondingly, the random forces are properly formulated to satisfy the fluctuation-dissipation theorem (FDT). Notably, the model exactly conserves both linear and angular momentum. We demonstrate the proposed model's accuracy and efficiency by applying it for modeling colloidal ellipsoids. Specifically, we show it efficiently captures the static properties of suspensions of colloidal ellipsoids. The isotropic-nematic transition in an ellipsoidal suspension is reproduced by increasing its volume fraction or the aspect ratio of ellipsoid particles. Moreover, the hydrodynamics and diffusion of a single colloidal ellipsoid (prolate or oblate with moderate aspect ratios) are accurately captured. The calculated drag force on the ellipsoid and its diffusion coefficients (both translational and rotational) agree quantitatively with the theoretical predictions in the Stokes limit.

Understanding the early stages of electrochemical nucleation and growth is the cornerstone for nanoscale electrodeposition. Although studied since decades, the process is not yet fully understood. In this work, we introduce a new modelling approach to study the growth of a single hemispherical nucleus: Multi- Ion Transport and Reaction Model (MITReM). This approach takes into account the transport driven by diffusion and migration of all species in the electrolyte together with the electrochemical reactions at the electrode boundary. A Finite Element Method (FEM) is used to solve the balance equations for the concentration of all the active species and the electrolyte potential. In contrast to analytical models or discrete scale modelling techniques, the strength of this approach is that no assumptions on the dif- fusional or kinetic limitations have to be made. In addition, this novel platform allows to add further levels of complexity, such as multiple nuclei, adatom surface diffusion, aggregation, particle detachment, etc. The simulation results prove that, the initial growth stage of a 10 nm single hemispherical silver nucleus always starts under kinetic control, regardless of concentration and electrode potential. Later on, a transition from kinetic to diffusion control takes place. The time of transition depends on the imposed concentration and electrode potential. Moreover, the simulations clearly show that the growth rate is strongly affected by the imposed concentration and electrode potential, as it has been proven experi- mentally in countless occasions. Numerical simulation by MITReM proves to be of great interest to gain knowledge towards unravelling the early stages of electrochemical nucleation and growth processes.

In many quantum materials, the presence of atomic vacancies and impurity atoms creates a perturbation that strongly disrupts the surrounding electronic environment and can influence the overall macroscopic properties of the system. Scanning tunneling microscopy and spectroscopy (STM/S) offers a perfect tool to probe such perturbation by measuring a spatial dependence of quasipariticle local density of states (LDOS). Here, the interference between incident and scattered quasiparticles results in a periodic modulation of LDOS in the STS differential conductance maps, which can be observed as sets of peaks within the Fast Fourier Transform (FFT) of the data. This technique is known as quasiparticle scattering interference (QPI) method. Our aim is to analyze QPI patterns in the systems with spatially inhomogeneous structural and/or electronic order parameter fields, in which we expect a presence of more than one characteristic QPI patterns. We apply machine learning techniques to the sliding window FFT data extracted from the STS maps of graphene and FeAs-based compounds to get, for the first time, the insight into quasiparticle scattering in samples with spatially inhomogeneous physical properties. Experimental data contains both the 2D STM topographic images of the surface and 3D STS measurements, which can be viewed as a stack of 2D conductance maps on the same surface along the chosen energy dimension. For each of the STS maps sliding window FFT generates additional 3D array, where each 2D FFT pattern is also assigned spatial origin on the surface. Not only each of such patterns can contain multiple scattering source signatures, but also the selected size of the window can significantly affect the produced results. Furthermore, given the size and dimensionality of the data that could be generated using STM/S, we need to apply Big Data tools in order to create a reliable, robust method, which can further be applied to different types of materials. This necessitates both the introduction of physical constraints and learning from the data for each particular material. We propose a deep learning framework using unsupervised learning in order to extract features from the local FFT patterns, which will allow classification, and further projection of the features onto the conductance maps. We then use correlative learning in order to establish the connection between topographical and electronic structures of the materials studied. We further study whether the application of 3D deep learning to the STS data can provide clues towards understanding the connection between complex electronic phenomena and topography, as well as statistical trends in conductance maps, their FFT, energy dimension, and their interactions and correlations.

This work was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES). Fellowship support from the UT/ORNL Bredesen Center for Interdisciplinary Research and Graduate Education.

Due to the increasing demand for the miniaturization of magnetic devices, there has been a growing interest in the modeling of magnetic phenomena found in materials that present exchange bias. In particular, ferro-antiferromagnetic core-shell nanoparticles are an interesting case in which the magnetic properties of the nanostructure can be altered by adjusting their size, shape, and com- position. By employing Monte Carlo methods, we develop a computational scheme that efficiently models the magnetic behavior of these bi-magnetic core-shell nanostructures. Using a Heisenberg- Dirac-van Vleck Hamiltonian in combination with a continuous spin model, we simulate a wide range of hysteresis diagrams displaying exchange bias. Furthermore, we demonstrate our efforts towards improving the efficiency of the simulation algorithms, aiming to afford magnetic atomistic simulations of large nanostructures by using a discretized spin model based on the tesselation of a unit sphere upon GPUs. Our results allow for further semi-quantitative comparisons with existing experimental data, provide a means to discover new phenomena associated with these core-shell nanoparticles and other nanostructures, and supply a tool for the efficient design of new magnetic nanomaterials.

Cathode design frequently employs nano-sizing as one avenue to explore improved performance, particular to enhance ionic diffusion. In this aspect, understanding the reactivity of the surface, specifically towards oxygen loss and subsequent passivation of the surface is extremely important. In this work, we explore the surface stability and vulnerability towards oxygen loss of a representative cathode for one of the most exciting Li-ion cathode family to date – the Li excess materials. This class of material can be modeled with the end member of both solid solution and composite structure aspects, which is Li₂MnO₃ the most Li excess, highest capacity but also the worst degradation. In detail, we systematically investigated all possible low Miller index surfaces of the pristine Li₂MnO₃ with various cation ordering on each surface. We found surface reconstruction rules by means of local cation environments, which are utilized to optimize the equilibrium Wulff shape. For the resulting equilibrium particle shape, we established the threshold Li concentration for the spontaneous oxygen release, consistent with recent experimental observations. Finally, we identified which facets exhibit improved stability as compared to others, which provides a design recommendation for more stable particle morphologies and enhanced surface oxygen retention. Hence, our work provides a possibility for improvement and guidance for new directions of cathode design.

Unlike conventional engineering materials, biological materials present a unique challenge from a finite element modeling perspective, as it is difficult or impossible to obtain experimental data on the region of interest in the environment of interest. Therefore, certain simplifications and statistical methods must be utilized to obtain an accurate solution to the problem. A method for determining the effective viscoelastic properties of the transversely isotropic white matter is presented.

Magnetic resonance elastography (MRE) is an imaging modality used to determine the mechanical properties of biological tissues through magnetic resonance imaging, by imposing a harmonic force externally to the body, and monitoring the displacements. The resulting displacements can then be run through an inversion process to determine the estimated viscoelastic moduli. This inversion process involves certain assumptions about the material properties, including density and degree of anisotropy. The brain in particular is a difficult situation, as it is near impossible to obtain mechanical properties via other modalities, and certain areas of the brain display a relatively large amount of anisotropic behavior.

The objective of this research is to blend together various in vivo and ex vivo testing modalities to improve the inversion process. Diffusion tensor imaging can provide indications of the local geometry in vivo, and ex vivo experimental data can provide estimated micromechanical properties. A representative volume element (RVE) is then developed using a finite element solver, with appropriate geometry and material properties, and a direct steady state solver is then used to solve the equations of motion for the RVE with shear boundary loads applied. The computational result is then utilized to determine the effective properties of the RVE. While this is a relatively simple and straightforward technique, the uncertainties present in the system resulting from the imaging modalities and the ex vivo nature of the mechanical testing present an issue. Therefore, a statistical approach is utilized to obtain the estimated properties from several hundred or thousand RVEs to provide coverage of the potential solution space. This information can then be used to provide an estimate of the degree of anisotropy and other properties to improve the existing MRE inversion techniques.

With recent advances in computational hardware and experimental instruments, unprecedented levels of high quality data are being generated at an ever-increasing rate. Sophisticated data mining and machine learning methods are required to realizing the hidden knowledge within these large and complex data sets and to advance our understanding of materials science. In this talk I will discuss results from two machine learning projects, on solute diffusion and irradiation-enhanced embrittlement. The first topic, impurity transport in solid solutions, is a crucial aspect of many kinetic processes in materials science. With recent high-throughput density functional theory calculations, we have obtained a large database of more than 300 dilute solute diffusion systems. We apply several data mining algorithms to learn and to predict diffusion coefficients of new systems without requiring further calculations. The second topic, understanding and predicting the effect of radiation on reactor pressure vessel steels, is essential for the safe planning of light water reactor plant life extension. We use machine learning methods to investigate the Irradiation VARiable (IVAR) database, consisting of hardening data at more than 1500 composition and irradiation conditions. We find excellent interpolation capability and explore extrapolation to real world reactor conditions at low-flux and high-fluence. This work was a collaborative effort with the Informatics Skunkworks group (skunkworks.engr.wisc.edu) at UW-Madison.

Current solid oxide fuel cells (SOFCs) require temperatures in the region of 1000°C to operate. This is primarily due to the fact that current generation SOFC electrolytes, such as ytrria stabilized zirconia (YSZ), require high temperatures in order for sufficient ionic diffusion of the O^2- ions to occur.¹ It has been suggested that replacing YSZ with samarium doped ceria (SDC) or gadolinium doped ceria (GDC)² would reduce the operating temperature of SOFCs into the intermediate temperature (IT) range of 600-800°C, thus greatly reducing operating costs and increasing efficiency.

Classical molecular dynamics can be used to investigate ionic conductivity and its limitations in these systems. Here we compare the performance of two interatomic potentials derived for a range of trivalent dopants in ceria from ab initio data, a dipole polarizable ionic model (DIPPIM) and a rigid ion model (RIM).³ The DIPPIM allows for polarization effects resulting from induced dipoles whereas the RIM does not. In this study we aim to elucidate whether or not this system can be modelled successfully using a RIM or if a DIPPIM is necessary due to the large polarization effects caused by the presence of O^2- ions.

The effect of surfaces and interfaces, such as grain boundaries, on the oxide ion conductivity in these materials is essential to their performance as SOFC electrolytes. It has been suggested that the interfaces in these materials can result in reduced oxide ion conductivity; however, the majority of studies consider only the average effect of the interface and not the possible effects of different specifically defined interfaces. Here we discuss the effect of surfaces and specific tilt grain boundaries on the performance of SDC and GDC as oxide ion conductors. Segregation of the oxygen vacancies to the boundary is observed along with enhanced diffusion of the oxide ions parallel to the grain boundary.

[1] Boudghene et al, Renew. Sust. Energ. Rev., 6, 433-455 (2002).
[2] Zha et al, J. Power Sources, 115, 44-48 (2003).
[3] Castiglione et al, J. Phys: Condens. Matter, 11, 9009-9024 (1999).

Fundamentally understanding the charge transport phenomena in dielectrics is important for predicting the life-time of dielectric capacitors and prolonging it by mitigating or slowing down the voltage-induced resistance degradation and breakdown process. Experimentally measured current, impedance spectral, and electrocoloration evolutions during degradation for Fe-doped SrTiO₃ single crystal indicate that the effective resistance was decreased by two orders of magnitude and two regions with different conductivity were formed inside the material after long-time degradation. In order to understand the corresponding resistance degradation mechanism, a charge transport model was constructed incorporating the defect chemistry, charge transport, and interfacial charge injection. It is found that the electric field-induced migration of oxygen vacancies and the subsequent instantaneous reestablishment of the local defect equilibria that lead to the resistance degradation. At the beginning of the degradation, the interfacial injected electrons and holes play a more important role in increasing the conductivity and modifing the electric field profile. While with degradation time increasing, the intrinsic electrons and holes generated due to the defect equilibrium becomes more imporant for determing the conductivity and electric-field profiles. The simulated electrocoloration, effective conductivity, impedance modulus spectra agree with the experimental measurements, validating the developed program for modeling the resistance degradation for dielectric capacitors [1,2].

[1] J. J. Wang, H. B. Huang, T. J.M. Bayer, A. Moballegh, Y. Cao, A. Klein, E. C. Dickey, D. L. Irving, C. A. Randall, L. Q. Chen, Defect chemistry and resistance degradation in Fe-doped SrTiO3 single crystal, Acta Materialia 108,229(2016).
[2] J. J. Wang, T. J.M. Bayer, R. Wang, J. Baker, D. L. Irving, C. A. Randall, L. Q. Chen, Modeling Dielectric Degradation: Role of Interfacial Charge Injection, (in preparation).

A fully automated computational procedure to predict activation energies would be immensely beneficial to high-throughput studies of diffusion, domain wall motion, and phase transitions in materials. The nudged elastic band method is a widely used algorithm to determine these energy barriers along the minimum energy path connecting the initial and final state of a kinetic process. Yet, the calculated energy barriers depend critically on the choice of the initial energy path, so that multiple trials with different starting points are necessary to obtain accurate energy barrier predictions. The complexity of this problem can be drastically reduced by considering all of the distortion symmetries that map a possible energy pathway onto itself [1]. These symmetries, which consist of the composition of conventional spatial symmetry operations with an operation called distortion reversal, have recently been fully enumerated [2], allowing for their implementation into nudged elastic band calculations. By applying this symmetry-adapted approach, it is possible in many cases to reduce the calculations to half of the path, as well as explore additional pathways that are otherwise inaccessible without applying specific perturbations to the initial path. An implementation of the symmetry-adapted nudged elastic band algorithm has been completed in the open-source Quantum-ESPRESSO software package [3], permitting the automatic identification of the distortion symmetries that exist for a given distortion pathway. Testing has been completed for a variety of cases that contain distortion reversal symmetry, and for which the calculation of activation barriers with the nudged elastic band method is problematic.

[1] VanLeeuwen, B. K. & Gopalan, V. Nat. Commun. 6, 8818 (2015).
[2] VanLeeuwen, B. K., Gopalan, V. & Litvin, D. B. Acta Cryst. A 70, 24–38 (2014).
[3] Paolo Giannozzi et al. J. Phys.: Condens. Matter 21, 395502 (2009).

A novel mesoscale state comprising of ordered polar vortex-antivortex lattice has been demonstrated in ferroelectric superlattices of PbTiO₃/SrTiO₃. Here, we employ phase-field simulations, analytical theory and experimental observations to evaluate thermodynamic conditions and geometric length scales that are critical for the formation of such exotic vortex states. We show that the stability of these vortex lattices involve an intimate competition between long-range electrostatic, long-range elastic, and short-range polarization gradient-related interactions leading to both, an upper- and a lower- bound to the length scale at which these states can be observed. We first propose an analytical analysis, which provides a design rule to the search for vortex lattices, where the appropriate “characteristic length” related to the bulk domain wall width is required. The role of each individual energy contribution in the formation of a vortex state is further studied by calculating a size dependent phase diagram. The competing interactions of elastic, electric, Landau and gradient energies lead to a transition from a₁/a₂ twin polar states, to vortex lattice, and eventually to flux-closure lattices with increasing superlattice periodicity. Whereas the role of STO layers are further explored within the (PTO)₁₀/(STO)_n phase diagram, which indicate that the tunable depoling strength could help engineer multiple phases, while the existence of a weak ferroelectricity in the STO layer could facilitate the formation of the vortex lattice. Thus, our work not only contributes to the further understanding of domain formation mechanism in current (PTO)_n/(STO)_n superlattice system, but also stimulates future studies on developing superlattice-based novel material structures.

Bismuth ferrite (BiFeO₃ or BFO) is a multiferroic material that exhibits ferroelectricity at room temperature. A fundamental understanding of domain switching is critical to enabling technological applications of this functionality. In this work, the mechanism of ferroelectric switching and domain wall motion at an atomic scale is investigated using the nudged elastic band (NEB) method [1], implemented with first-principles electronic structure calculations. Different minimum energy pathways for switching are explored, employing a novel approach that makes use of a recently discovered symmetry of motion, namely 'distortion reversal symmetry' [2]. This theory expands upon the concept of spatial symmetry to include distortion and motion, finding hidden symmetries in physical processes. In the context of domain switching in ferroelectrics, it provides an elegant and robust approach to finding the minimum energy path, by breaking the symmetries of the switching pathway in a methodical manner. Applying this method to BFO, it is found that breaking certain symmetries allows for lower energy switching pathways. In particular, breaking the threefold symmetry by allowing the polarization vector to have a component perpendicular to its original orientation drastically reduces the activation energy for switching. Systematically breaking other distortion symmetries in this manner will help accurately determine the minimum energy pathway for switching, and also provide an insight into the closely related, and previously unexplored problem of minimum energy pathways for domain wall motion in BFO. The influence of the distortion symmetries of ferroelectric switching on the magnetic ordering will also be studied.

[1] H. Jónsson, G. Mills, and K. W. Jacobsen, “Nudged elastic band method for finding minimum energy paths of transitions,” Class. Quantum Dyn. Condens. Phase Simulations, pp. 385–404, 1998.
[2] B. K. VanLeeuwen and V. Gopalan, “The antisymmetry of distortions,” Nat. Commun., vol. 6, p. 8818, 2015.

Phase separation of a homogeneous state into a mixture of two or more phases is the manifestation of a common mode of materials instability. The formation of mixed phases attracts enormous attention from scientists since the mixed phases may give rise to enhanced responses under external stimuli due to the transition between the two phases. In a well-known phase de-composition process, the presence and local compositions of mixed phases are illustrated and calculated using the geometrical common tangent construction on free energy versus composition curves. In analogy to the phase de-composition process, we propose a new “phase de-strain” concept describing the coexistence of ferroelastic domains. The common tangent, level rule, equilibrium condition of chemical potential, and phase rule are revisited based on the phase de-strain model. Strain-temperature and strain-strain phase diagrams of PbTiO₃ are calculated based on the common tangent construction. Also, we construct a Landau-theory-based potential to describe the rhombohedral-like and tetragonal-like phases in compressively strained BiFeO₃ films, and the determined domain morphology from phase-field simulations exhibits excellent agreement with piezoresponse force microscopy (PFM) measurements.

Ferromagnetic manganese antimonide (MnSb) is a promising material for spintronic applications. Multilayered structures of ferromagnetic (FM) materials and semiconductors have attracted considerable interest for the purpose of spin-injection into the semiconductor. Even though the growth of MnSb on InP is challenging, the niccolite (n)-MnSb polymorph closely lattice matches InP which makes this an ideal system to study MnSb / III-V semiconductor interfaces without the complication of high strain. The MnSb/GaAs system is similar but with 3.2% lattice mismatch. Here we use spin-polarized density functional theory, via the CASTEP plane wave pseudopotential method, to investigate the structural, electronic and magnetic properties at the interface of n-MnSb/InP(111) and n-MnSb(0001)/GaAs(111) heterostructures. Our results show that the Mn-to-P termination of the n-MnSb/InP(111) and Mn-to-As termination of the n-MnSb(0001)/GaAs(111) superlatices have 63.9% and 61.2% spin-polarization, which is far higher than the bulk polarization of ~18% and should be very favourable for spin transport applications. Normally interfaces are seen as problematic in half-metallic FM structures, reducing or even reversing spin polarization, so this behavior is unusual. This high interface spin polarization is mainly due to the presence of Mn d orbital with contribution from the p orbital of P or As. The work of separation of the Mn-to-P and Mn-to-As terminations are 2.80 J/m² and 2.02 J/m² respectively, with interfacial bond lengths of 2.54 Å and 2.63 Å. These interfaces become less energetically unfavourable than in the bulk, while the others are more unfavourable and maintain the ferromagnetic nature of bulk n-MnSb. We discuss the prospects for experimental realization of these ferromagnetic/semiconductor superlattices.

Understanding the behavior of dielectric insulators experiencing extreme electric fields is critical to the operation of present and emerging electrical and electronic devices. Despite its importance, the development of a predictive theory of dielectric breakdown has remained a challenge, owing to the complex multiscale nature of this process. Here, we focus on the intrinsic dielectric breakdown field of insulators - the theoretical limit of breakdown determined purely by the chemistry of the material, i.e., the elements the material is composed of, the atomic-level structure, and the bonding. Starting from a benchmark dataset (generated from laborious first principles computations) of the intrinsic dielectric breakdown field of a variety of model insulators, simple predictive phenomenological models of dielectric breakdown are distilled using advanced statistical or machine learning schemes, revealing key correlations and analytical relationships between the breakdown field and easily accessible material properties[1]. Next we apply these design rules to a vast chemical space of perovskite materials in order to demonstrate the efficacy, generalizability and true predictive power of our model. Starting from several thousands of compounds, we systematically downselect 209 insultors which are dynamically stable in a perovskite crystal structure. After making predictions on these compounds using our machine learning model, the intrinsic dielectric breakdown strength was further cross-validated explicitly using first principles computations. Our analysis reveals that several boron-containing compounds are of particular interest, some of which exhibit remarkable intrinsic breakdown strength of about 2 GV/m[2].

References:
[1] C. Kim, G. Pilania, R. Ramprasad, Chem. Mater., 28, 1304 (2016)
[2] C. Kim, G. Pilania, R. Ramprasad, J. Chem. Phys. C., accepted (2016)

The link between changes in the material crystal structure and its mechanical, electronic, optical, and magnetic functionalities known as the structure-function relationship is at the heart of the contemporary materials science research. Of particular interest is a role of certain types of crystalline defects such as vacancies and substitutional dopants as well as the associated local strain in the physical properties of the technologically relevant materials. For example, chemical doping can turn a Mott insulator into an unconventional superconductor whose properties, such as a magnitude of the superconducting gap, are sensitive to variations of the inter-atomic bond lengths and bond angles in the crystal lattice.
The arrival of high-resolution scanning probe microscopic and spectroscopic (SPM/S) methods has recently allowed researchers performing simultaneous measurements of materials structure (e.g. lattice distortion around the defect) and functional properties (e.g. modulation of electron charge density) in a real space with a sub-nanometer precision. However, methods to cross-correlate information obtained in the structure ‘channel’ and function ‘channel’ and to describe the obtained correlation in terms of certain (linear or non-linear) physical models are very limited and yet are increasingly necessary due to the ever-growing volumes of relevant experimental data. Here, we have designed an automated approach for correlative analysis between structural and functional features from SPM datasets obtained on graphene with hydrogenated and oxidized defects. Specifically, by applying a combination of the Pearson correlation matrix, linear and kernel canonical correlation analysis, we were able to extract the detailed information on the correlation between the nanoscale strain and the parameters associated with the quasiparticle scattering. We found that the expansion of the lattice constant results in the enhanced scattering intensity in the stretched regions of graphene. Our analysis also revealed that the strength of coupling to strain is altered between different scattering channels which can explain an emergence of more than one superperiodic electronic patterns in the SPM images (the so-called fine structure of the superlattice). Finally, using a kernelized version of the canonical correlation analysis we uncovered a presence of non-linear associations between the strain components and the intensity of electron scattering in graphene. The physics implications of these findings will be discussed. In future, the approach outlined in our work can be applied to analysis of multi-modal datasets in electron microscopy, for example, for finding a correlation between position of atomic columns and features in electron-energy loss spectroscopy data.

This work was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), Materials Science and Engineering Division.

The double-layered ruthenate Sr3Ru2O7 exhibits a plethora of exotic phases. Here we use first principles theory to study the properties of this material at the surface. Our results demonstrate that the electronic properties are different from the bulk and intimately coupled with the RuO6 octahedra tilts and rotations.

The phonovoltaic mimics the photovoltaic in order to harvest hot optical phonons in a nanoscale p-n junction. That is, a non-equilibrium (hot) population of energy carriers (the optical phonon rather than the photon) drives electron-hole generation as it relaxes. Generated electrons are then separated by the intrinsic field in the p-n junction to generate power. However, phonons, unlike photons, tend to down-convert and generate low-energy acoustic phonons unable to generate power. Thus, the phonovoltaic requires both a band gap which matches the optical phonon energy and a strong electron-phonon (e-p) coupling relative to the phonon-phonon (p-p) coupling [C. Melnick and M. Kaviany, “Phonovoltaic. I. Harvesting hot optical phonons in a nanoscale p-n junction”, Phys. Rev. B, 93, 094302 (2016)]. No current semiconductors exist which meet the phonovoltaic requirements. However, if such a material is designed, the efficiency of the phonovoltaic cell can greatly exceed the thermoelectric efficiency.

Here, material candidates are surveyed and tuned graphene (any graphene-based material with a tuned band gap) is shown to be uniquely suitable. Using ab initio simulation and tight-binding theory, the e-p coupling in three tuned graphene materials is discussed: hydrogenated graphene, carbon boron-nitride composites, and bilayer-graphene under a field. The latter two materials achieve significant phonovoltaic figure of merit and enable a phonovoltaic efficiency exceeding thermoelectric devices, while the first does not. Maintaining the sp² hybridization of graphene and moving the band-edge away from the K points is shown to preserve or even enhance the e-p coupling of graphene.

The formation of oxygen vacancies and their effect on the n-type conductivity of transparent conducting oxides (TCOs) has remained a controversial topic. They have been determined to be deep donors in SnO₂ and ZnO, while it has been proposed that their formation on the surface of In₂O₃ would lead to intrinsic n-type conduction. Numerous computational studies on this subject have correlated the calculated defect properties with available experimental spectroscopic data (luminescence etc.), but, notably, contradict each other in the defect assignment. We use state-of-the-art hybrid quantum mechanical/molecular mechanical solid state embedding to determine the formation energy, electronic and optical properties of the oxygen vacancy in the three archetypal TCOs In₂O₃, ZnO and SnO₂. To ensure a high level of accuracy in our simulations, we employ three hybrid exchange and correlation density functionals (PBE0, B97-2 and BB1k) and compare their predictions with previous plane-wave pseudopotential based calculations and experiment. For ZnO and SnO₂, our results using the PBE0 functional are in excellent agreement with calculations done using a plane-wave basis set (which predict the oxygen vacancy to be a deep donor in all three systems). In bulk In₂O₃ we find, in contrast to previous studies, that the oxygen vacancy is a resonant shallow donor. Using BB1k, which has been developed specifically to treat both reaction energies and kinetic barriers accurately, for the cases of SnO₂ and ZnO we show that the vacancy is indeed a deep donor, but shallower than previously proposed. Our results are in excellent agreement with available deep level transient spectroscopy measurements and other relevant experimental data.

Transparent conducting materials are widely used as electrodes in devices such as solar cells, display screens and light-emitting diodes. The required properties of low optical absorption and low resistivity only coexist in a limited range of materials. The majority of transparent conducting oxides (TCOs) under active development are n-type with conduction bands in which cation s-orbitals hybridise with oxygen p-orbitals. The cation selection is generally limited to groups 12–14 of the periodic table.
Recently, zinc antimonate (ZnSb₂O₆) has been recommended as a possible TCO material in a high-throughput computational screening study.^[1] It has a wide bandgap (>3eV) and is reported to possess conduction band effective masses of ~0.2, which are ideal for high-mobility TCO applications. There is some experimental support for this as crystalline samples of ZnSb₂O₆ formed from sintered powder were reported to have promising conductivity in 2005.^[2] The best TCOs (e.g. BaSnO₃) only show high performance when doped with a suitable electron donor. In this presentation we will use hybrid density functional theory to examine the defect chemistry of ZnSb₂O₆, including intrinsic defects and extrinsic donors, to assess its suitability as a potential n-type TCO. Crucially, we will also utilize ab initio lattice dynamics calculations to gain insight into the temperature dependent stability window for the system, and additionally the formation energies of defects at realistic growth conditions.

[1] Hautier et al. (2014). Chem. Mater., 26 (19), 5447–5458.
[2] Kikuchi, N. et al. J. Am. Ceram. Soc., 88(10), 2793–2797.

A central trait of the perovskite oxides are their strong structure-property coupling. From a thin film point-of-view, this opens the possibility to rely on epitaxy to tune and modify the properties by epitaxial strain. Strain in the (001)-plane has been extensively studied, and has proved a valuable tool to control octahedral rotations and a plethora of functional properties. However, recent development in crystallographic growth has developed high quality thin films along other crystallographic directions such as the [111]. The (111)-interface is interesting because it has a buckled honeycomb lattice, and the oxygen octahedral share faces instead of corners as for (001)-oriented systems, possibly increasing the coupling across the interface. Here we use density functional theory (DFT) to compare the effects of biaxial strain in (001)- and (111)-plane, using LaAlO₃ as model system. LaAlO₃ is rhombohedral in bulk and (001)-strain is known to favor tetragonal phases. We find that tensile (111)-strain preserves the rhombohedral symmetry. On the other hand, compressive (111)-strain gives rise to three degenerate (△E< 0.1 meV/f.u.) monoclinic phases with different octahedral rotation patterns, not observed in bulk or for (001)-strain. The implications of having neither of the oxygen octahedra rotation axes parallel or perpendicular to the strain-plane when compensating for the applied strain will be discussed. Furthermore, the effect of trigonal d-orbital splitting from (111)-strain, compared to the cubic splitting from (001)-strain, is investigated. For LaAlO₃ this different orbital splitting causes a change in the sign of the derivative of the band gap as a function of strain, opening new pathways to tune the electronic properties of oxide systems. Finally, the effect of (111)-oriented strain on ferroic properties of similar systems will be discussed.

LaMnO₃-based perovskites have been widely studied as the oxygen electrode for high temperature solid oxide fuel cells (SOFCs). Sr-doped cubic LaMnO₃ is the conventional cathode material for high temperature operation due to its high electrical conductivity, good activity for the oxygen reduction reaction and compatibility with electrolyte materials.[1] Further research is required to improve the efficiency and reduce the operating temperature of SOFCs, due to the high costs and accelerated performance degradation associated with high temperature operation. At intermediate temperature (600 – 1000 K), the oxygen reduction reaction is often limited to the interface between the cathode, electrolyte, and the air (the three-phase boundary), due to the poor ionic conductivity of the cathode. In order to increase the region in which the reaction can take place, mixed ionic and electronic conductors can be used as cathode materials. Introduction of lower valence dopant cations at both the La and Mn sites can improve the ionic and electronic conductivity of orthorhombic LaMnO₃.[2] Here, we investigate doping with the alkaline earth metals Mg, Ca, Sr and Ba.

The formation energies of defects in orthorhombic LaMnO₃ have been calculated using PBEsol + U.[3] Oxygen vacancy formation, and its dependence on temperature, oxygen partial pressure and chemical potential, in bulk LaMnO₃ and at its low index surfaces has been examined. The energies of formation, under intermediate temperature SOFC operating conditions, of isolated defects and clustered pairs have been investigated, by placing defects at either the La or Mn site, to establish the most probable site at which they will be introduced. The charge compensation mechanism for the introduction of alkaline earth dopants has been examined by considering both ionic (formation of an oxygen vacancy for every 2 alkaline earth dopants introduced) and electronic compensation (a hole localised at the Mn site for each dopant introduced).

[1] J. A. Kilner and M. Burriel, Annu. Rev. Mater. Res., 2014, 44, 365 – 393
[2] L. Gan et al., J. Alloys Compd., 2016, 655, 99 – 105
[3] J. P. Perdew et al., Phys. Rev. Lett., 2008, 100, 136406

Materials that exhibit simultaneous order in their electric and magnetic ground states hold tremendous promise for use in next-generation memory devices in which electric fields control magnetism. Such materials are exceedingly rare, however, due to competing requirements for displacive ferroelectricity and magnetism. Despite the identification of a number of novel multiferroic materials and magnetoelectric mechanisms recently, known single-phase multiferroics remain limited by antiferromagnetic or weak ferromagnetic alignments, lack of coupling between the order parameters or have properties that only emerge well below room-temperature, stymieing device applications. We present a new methodology for the construction of single-phase multiferroic materials where ferroelectricity and strong magnetic ordering are coupled near room-temperature.¹ Starting with hexagonal LuFeO₃, a geometric ferroelectric with the greatest known planar rumpling, we introduce individual extra monolayers of FeO during growth to construct formula-unit-thick syntactic layers of ferrimagnetic LuFe₂O₄ within the LuFeO₃ matrix, i.e., (LuFeO₃)_m/(LuFe₂O₄)₁ superlattices.¹ The severe rumpling imposed by the neighbouring LuFeO₃ drives the ferrimagnetic LuFe₂O₄ into a simultaneously ferroelectric state, while also reducing the LuFe₂O₄ spin frustration. This increases the magnetic transition temperature significantly—from 240 K for LuFe₂O₄ to 281 K for (LuFeO₃)₉/(LuFe₂O₄)₁. Moreover, the ferroelectric order couples to the ferrimagnetism, enabling direct electric field control of magnetism at 200 K. Our results demonstrate a design methodology for creating higher-temperature magnetoelectric multiferroics by exploiting a combination of geometric frustration, lattice distortions and epitaxial engineering.¹

1. Julia A. Mundy, Charles M. Brooks, Megan E. Holtz, Jarrett A. Moyer, Hena Das, Alejandro F. Rébola, John T. Heron, James D. Clarkson, Steven M. Disseler, Zhiqi Liu, Alan Farhan, Rainer Held, Robert Hovden, Elliot Padgett, Qingyun Mao, Hanjong Paik, Rajiv Misra, Lena F. Kourkoutis, Elke Arenholz, Andreas Scholl, Julie A. Borchers, William D. Ratcliff, Ramamoorthy Ramesh, Craig J. Fennie, Peter Schiffer, David A. Muller, and Darrell G. Schlom (unpublished).

Hybrid improper ferroelectricity, where an electrical polarization can be induced via a trilinear coupling to two non-polar structural distortions of different symmetry, has recently been experimentally demonstrated for the first time in the n=2 Ruddlesden-Popper compound Ca3Ti2O7. In this talk I will first give an overview of recent experiments and theory. I will then present our new work in which we use group theoretic methods and first-principles calculations to identify possible domains and ferroelectric switching pathways. We identify low-energy paths that reverse the polarization direction by switching via an orthorhombic twin domain, or via an antipolar structure. We also introduce a chemically intuitive set of local order parameters to give insight into how these paths are relevant to switching nucleated at domain walls. Our findings suggest that switching may proceed via more than one mechanism in this material.

In accordance with Richard Feynman's 1959 statement, "there's plenty of room at the bottom," we explore the strain design space of low-dimensional materials for electronic and optoelectronic applications. Homogenous and inhomogeneous elastic strain [Nature Photonics 6 (2012) 866; Nature Communications 6 (2015) 7381], bending [ACS Nano 5 (2011) 3475], interlayer twist [Nano Letters 14 (2014) 5350] and slip [Nano Letters 15 (2015) 1302] lead to tunable, low-energy artificial atoms, artificial superlattices and pseudoheterostructures that can regulate quasiparticle motion [Adv. Mater. 26 (2014) 2572]. We also demonstrate production of kilogram-scale nanowires under large tensile elastic strain, which leads to improved superconductivity properties.

Epitaxial strain engineering is a powerful approach to generate and tailor collective phenomena and new functionalities in thin films of layered oxides, A_n+1B_nO_3n+1 (n=1-∞). Strong coupling between substrate-induced strain and polar lattice modes in thin films can stabilize new ferroelectric (FE) phases from nonpolar dielectrics or enhance electric polarizations and Curie temperatures. Recently, strain has also been exploited to induce novel metal-insulator transitions and magnetic reconstructions through its coupling to nonpolar lattice modes, including rotations of BO₆ transition-metal polyhedra. Although large epitaxial strains are believed to induce ferroelectricity, here we demonstrate using first-principles calculations and sophisticated structure-searching genetic algorithms that biaxial strain induces an unanticipated polar-to-nonpolar (P-NP) structural transition in (001) thin films of layered hybrid-improper ferroelectrics with the n=2 Ruddlesden-Popper structure at experimentally accessible strains. We show in detail for Ca₃Ti₂O₇ that the origin of the P-NP transition originates from the interplay of trilinear-related lattice mode interactions active in the layered oxides, and those interactions are directly strain tunable. We use this understanding to show that the P-NP transition also occurs in (001) Ca₃Mn₂O₇ and Sr₃Zr₂O₇ thin films and then establish a general design framework for realizing the P-NP transition using symmetry and crystal-chemistry principles. Our results call for a careful reconsideration of the role of strain-polarization coupling in FE films with nontrivial anharmonicities and offer a route to search for new functionalities in layered oxides through multimode coupling.
This work was supported by the National Science Foundation (NSF) through the Pennsylvania State University MRSEC under award number DMR-1420620.

Abstract: The original proposal to achieve superconductivity by starting from a quantum spin-liquid (QSL) and doping it with charge carriers, as proposed by Anderson in 1987, has yet to be realized. Here we propose an alternative strategy: use a QSL as a substrate for heterostructure growth of metallic films to design exotic superconductors. By spatially separating the two key ingredients of superconductivity, i.e., charge carriers (metal) and pairing interaction (QSL), the proposed setup naturally lands on the parameter regime conducive to a controlled theoretical prediction. Moreover, the proposed setup allows us to ``customize'' electron-electron interaction imprinted on the metallic layer. The QSL material of our choice is quantum spin ice well-known for its emergent gauge-field description of spin frustration. Assuming the metallic layer forms an isotropic single Fermi pocket, we predict that the coupling between the emergent gauge-field and the electrons of the metallic layer will drive topological odd-parity pairing. We further present guiding principles for materializing the suitable heterostructure using ab initio calculations and describe the band structure we predict for the case of Y$2$Sn$_{2-x}$Sb$_x$O$_7$ grown on the (111) surface of Pr$_2$Zr$_2$O$_7$. Using this microscopic information, we predict topological odd-parity superconductivity at a few Kelvin in this heterostructure, which is comparable to the $T_c$ of the only other confirmed odd-parity superconductor Sr$_2$RuO$_4$.

Layered oxides and sulfides form a large class of materials that includes many compounds used for electrochemical energy storage, including state-of-the-art Li-ion cathode materials (e.g., LiCoO₂) as well as novel Na-ion and Mg-ion electrode materials. While all of the layered oxides and sulfides share the same crystallographic building blocks, they differ in important ways, including the stacking of layers, ordering of intercalants, and intercalation thermodynamics. Understanding the trends in these properties will help to develop new battery materials. This work reports on high-throughput density-functional theory calculations of intercalated layered oxides and sulfides that elucidate patterns in structure and thermodynamics. Qualitative differences in properties between materials with different intercalants suggests that some of the challenges facing Na-ion and other beyond Li-ion chemistries may be unlike the challenges for Li-ion systems.

Owing to their low density, magnesium alloys have generated substantial interest as structural materials in automotive and aerospace applications. In particular, magnesium-rare earth alloys have demonstrated attractive qualities, including high yield strengths. For one magnesium-rare earth alloy, magnesium-neodymium, improved mechanical properties have been associated with the presence of metastable β’ and β₁ precipitates. However, recent experimental observations and first-principles modeling has indicated that the precipitates previously thought to have the β’ ordering in fact belong to a hierarchy of orderings, termed β’’’, with neodymium concentrations between 12.5 and 25 at. %. We present a series of phase field simulations to provide new insight into the formation of β’’’ precipitates in magnesium-neodymium alloys. We calculate the equilibrium morphologies of β’ and β’’’ precipitates for a variety of precipitate volumes and compare them with experimental observations. These simulations indicate that the formation of β’’’ is energetically preferable to β’ owing to decreased elastic energy. Precipitate morphology is shown to change as the precipitate volume increases as a result of the decreased role of the interfacial energy. Simulations examining the effects of precipitate-precipitate interactions on growth kinetics are also presented. Model parameters, including homogenous free energies, interfacial energies, stress-free transformation strains, and elastic constants, are determined from first principles calculations. The phase field simulations are performed using the newly released open-source PRISMS-PF phase field code.

Recently discovered carbon nanothreads, a one-dimensional form of sp3 carbon formed by high-pressure topochemical polymerization of crystalline benzene, complete the dimensionality/hybridization matrix of carbon compounds. Principles of materials design can be used to try to expand the family of possible nanothread-like materials to include a diverse range of materials properties ranging from piezoelectric to conductive and possibly superconductive, all within a package promising exceptional mechanical properties, in a system that is in some sense intermediate in character between a crystal and a polymer.

Defects play important roles in tuning the properties of bulk materials, but the effects in low-dimensional materials are more significant, because the large surface area in low-dimensional materials facilitates the formation of high density defects. The properties of low-dimensional materials can therefore be effectively modified by these defects. Understanding the roles of defects in tuning the magnetic and optoelectronic properties of two-dimensional (2D) materials is quite crucial for their promising applications in nanoscaled electronic devices.
Here, we focus on two typical semiconducting 2D materials, boron nitride (BN) and transition-metal dichalcogenides (MoS₂ and WS₂). We demonstrated that although perfect BN is nonmagetic with a large band gap (> 5.0 eV), stable ferromagnetic ordering can be achieved by fluorination or B-vacancy defects [1,2]. Using first-principles calculations, we proposed that the local magnetic moments arise mainly from the nitrogen atoms in the close vicinity of the defect. These magnetic moments interact in a ferromagnetic way via direct exchange mechanism, giving rise to stable ferromagnetism with Curie temperature higher than room temperature[1]. The co-exist of fluorination defects and B-vacancy defects can increase the ferromagnetism of 2D-BN [2]. For 2D transition-metal dichalcogenides, our first-principles calculations indicate that S-vacancy defects induce unfilled intermediate states (IS) within the band gap. The position and density of these IS can be tuned by controlling the density of S-vacancy defects. By introducing suitable S-vacancy defects, we achieved broadband laser saturable adsorption in few-layer MoS₂ [3] and WS₂ [4] films. The carrier density of these saturable adsorbers is 2-3 orders of magnitude higher than graphene. Using 2D-MoS₂ and 2D-WS₂ as saturable absorbers, we achieved high performance all-solid-state passively Q-switched laser operation with a pulse width of 60 ns [4], superior to those observed in other 2D materials.
[1] M. Du, X.L. Li, A. Z. Wang, Y.Z.Wu, X.P. Hao*, M.W. Zhao*, Angew. Chem. Int. Ed. 53, 3645-3649 (2014).
[2] H. B. Si, G. Lian*, A. Z. Wang, D. L. Cui*, M. W. Zhao*, Q. L. Wang, C.P. Wang, Nano Lett. 15, 8122-8128 (2015).
[3] S.X. Wang, H.H. Yu*, H.J. Zhang*, A. Z. Wang, M.W. Zhao*, Y.X. Chen, L.M. Mei, J.Y. Wang, Adv. Mater. 26, 3538-3544 (2014).
[4] G. Zhao, S. Han, A. Z. Wang, Y. Z. Wu, M. W. Zhao*, Z. P. Wang*, and X. P. Hao*, Adv. Func. Mater. 25, 5292-5299 (2015).

The coefficient of friction (COF) of graphene is reduced with increasing the relative humidity in the testing environment [1]. The role of water molecules in the atmosphere on friction reduction mechanisms of graphene was investigated by spin polarized density functional theory (DFT) calculations. Van der Waals (vdW) interactions were incorporated using the vdW-DF2 functional to accurately simulate the reaction between the water molecule and graphene surfaces with and without defects. The energy required for a water molecule to dissociate and be adsorbed on pristine graphene was larger (with an energy barrier of 3.53 eV/molecule) than the dissociative adsorption at mono vacancy site of graphene with an energy barrier of 1.27 eV/molecule. Dangling bonds near the vacancy became passivated by dissociated H and OH. Further dissociation of OH into O and H occurs on defected graphene. To understand the role of H and OH on the adhesion properties of graphene, bilayer AB graphene configurations with different H and OH facing each other were simulated. Results showed that the bilayer graphene interlayer separation can increase by 0.06-0.26 Å compared to the pristine AB graphene, whereas the interlayer adhesion energy can drop by 20%. The predicted increase in the interlayer separation distance agreed with experimental observation that the interlayer spacing between graphene layers increased from 3.4 Å to 3.5-3.8 Å during sliding [1]. The distortion of AB structure by the out of plane displacement (typically 0.4 A) of C atoms attached to OH molecules (akin to turbostratic structure) increased the structural disorder which in turn facilitated dissociation of water molecules on the graphene surface. Therefore, the friction reduction mechanisms of graphene under humidity can be interpreted as: (1) generation of dangling bonds at defect site; (2) passivation of graphene at defect site by dissociated water molecule; (3) graphene transfer due to increased interlayer separation and reduced interlayer adhesion energy; (4) low friction and wear are attained due to the formation of tribolayers.
[1] S. Bhowmick, A. Banerji, and A. T. Alpas, "Role of humidity in reducing sliding friction of multilayered graphene," Carbon, vol. 87, pp. 374-384, 2015.

Low-dimensional multiferroic materials with strongly coupled ferroic orders are particularly valuable owing to their great potentials for miniaturized device applications such as nanoscale transducers, actuators, sensors, photovoltaics, and nonvolatile memories. Nonetheless, perfect multiferroic materials especially in low dimensions are scarce, largely due to the stringent symmetry and chemistry requirements for practical applications at room temperature. Using first-principles theory, we predict that two-dimensional monolayer Group IV monochalcogenides including GeS, GeSe, SnS, and SnSe are a class of 2D semiconducting multiferroics with strongly coupled giant in-plane spontaneous ferroelectric polarization and spontaneous ferroelastic lattice strain, and are thermodynamically stable at room temperature. Their optical absorption spectra exhibit strong in-plane anisotropy with visible-spectrum excitonic gaps and sizable exciton binding energies. The low domain wall energy and small migration barrier together with the coupled ferroelastic-ferroelectric order and anisotropic electronic structures suggest their great potentials for tunable multiferroic functional devices by manipulating external electrical, mechanical, and optical field to control the internal responses. This may allow the realizations of conceptual devices such as 2D ferroelectric memory, 2D ferroelastic memory, 2D ferroelastoelectric nonvolatile photonic memory, and 2D ferroelectric excitonic photovoltaics. (Reference: http://arxiv.org/abs/1606.04522)

Graphite fluoride (GF) is comprised of stacked sheets of sp³-hybridized carbon atoms that are covalently bonded to fluorine atoms. Understanding the reductive defluorination of GF and related materials is important for controlling their functions as high energy density cathodes and synthetic precursors to graphitic nanostructures. While the reaction of GF with strong reductants is known to occur, recent experiments have shown that weaker reductants are capable of reductively defluorinating GF. After reduction, the defluorinated product becomes black and the Raman spectrum indicates the presence of conjugated carbon atoms. The ability of weaker reductants to defluorinate GF implies that defect sites provide lower-lying unoccupied orbitals that are capable of accepting electrons. DFT calculations have been carried out on a cluster model to identify a possible mechanism for the defluorination of GF via the reduction of various fluorine vacancies. The calculations reveal an “unzipping” of GF, where defect sites can regenerate and propagate upon reduction. Zig-zag and ring formation patterns are compared thermodynamically and kinetically to predict the dominant pattern of defect propagation.

The synthesis and applications of two-dimensional transition metal dichalcogenides such as molybdenum disulfide (MoS₂) in membranes, sensors, and microfluidic devices, involves an inevitable contact between these materials and various liquids. The existence of partially ionic bonds in MoS₂, as opposed to covalent bonds in graphene, suggests a plausible role for polar (electrostatic) interactions in determining the interfacial behavior on two-dimensional MoS₂ surfaces. Surprisingly, not only the equilibrium contact angle, which depends solely on the total interaction energy between the surface and the liquid, but also the friction coefficient and slip length, which depend on the spatial variations in the interaction energy, are independent of the inherent polarization in MoS₂. While the former is found to result from the negligible electrostatic interactions between MoS₂ and liquids, the latter results from the tri-layered sandwich structure of the MoS₂ monolayer. Specifically, the tri-layered sandwich structure results in spatial variations in the dispersion interactions that dominate over those in the electrostatic interactions. Furthermore, the wettability of MoS₂ is found to be controlled by a delicate balance between dispersion interactions and entropy. Our findings reveal that in spite of the existence of heteropolar bonds in MoS₂ compared to the existence of homopolar bonds in graphene, there is an unexpected similarity between the interfacial behaviors of these two surfaces. This finding, in turn, elucidates the similar physiosorption behavior of MoS₂ and graphene, including the use of similar liquids to exfoliate these 2D materials.

The influence of molecular conformation on electron relaxation and photophysical properties of a series of biphenyl derivatives have been investigated using density functional theory (DFT) and time-dependent DFT(TDDFT). The calculated absorption and emission properties of the series as well as phosphorescence quantum yield are in good agreement with the available experimental data. The spin orbit coupling values and the S → T intersystem-crossing matrix elements and crossing rate constants are explored as a function of the twist angle between the phenyl rings. The T → S₀ radiative and non-radiative rates are calculated and discussed for each member of the series. In addition, the T → S₀ radiative rate constants are calculated for twisted biphenyls then compared in planar ones. These results are discussed in the context of experiments performed on organic light-emitting diode (OLED) to develop the molecular design criteria of efficient OLEDs.

Two-dimensional (2D) materials, such as graphene and monolayers of transition metal-based compounds, exhibit a rich variety of electronic properties and novel 2D physics. In certain classes of layered transition metal-based compounds, recent work has indicated that it may be possible to exploit strain-induced structural transformations to rapidly switch between semiconducting and metallic in-plane crystal structures. To this end, we have developed an ab-initio informed, mesoscale continuum model based on the phase field microelasticity (PFM) approach to describe such transformations. The model incorporates the effects of transformation strains, domain boundaries and interfacial energies between the symmetry-related orientational variants of the transformed phase, long-range elastic interactions between domains, out-of-plane deformations, and coupling to applied stress or strain. Simulation results for both substrate-supported quasi-2D monolayers and suspended monolayers that can deform out of the 2D plane will be discussed, and issues of transformation reversibility and recoverability will be explored. This work has been supported as part of the Center for Computational Design of Functional Layered Materials (CCDM), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science under Award #DE-SC0012575.

2D forms of transition metal carbides and nitrides (MXenes) have shown interesting electrical, optical, and electrochemical properties. Similar to other 2D materials, their mechanical properties are believed to be high, rendering them attractive nanofillers for polymer matrix composites. High Young’s modulus, bending rigidity are necessary for developing MXene electrical and electrochemical sensors. However, measurements of the mechanical properties for single MXene flakes have been impeded by great experimental difficulties and have not been reported so far.
We investigate the mechanical properties and failure mechanisms of single-sheet titanium carbide MXenes of composition Ti_n+1C_n with n = 1 to 3, using large-scale classical molecular dynamics simulations with a three-body Axilrod-Teller potential to describe covalently bonded atoms and Lennard-Jones (6-12) two-body potential to describe non-covalent interatomic interactions. Young's modulus was calculated from the linear part of strain–stress curves obtained under tensile deformation of the samples. Strain-rate effects were observed for all Ti_n+1C_n samples. From the radial distribution function, it is found that the structure of the simulated samples is preserved during the deformation process. Failure mechanisms of Ti_n+1C_n were found to be sensitive to sheet thickness. Calculated values of the elastic constants are in good agreement with published DFT data. We also discuss in silico experiments to determine the bending rigidity of MXenes.

The atomic force microscope (AFM) is reaching a new level of maturity in the quantification of surface properties, beyond its original application as an imaging technique. Historically, most of this progress has been made using contact mechanics models with repulsive-regime force measurements to extract surface mechanical properties such as Young’s modulus or viscoelasticity. Additional useful information related to the chemistry of the surface is contained in the long-range, non-contact forces; however, reporting of forces beyond the region of mechanical contact (attractive-regime or long-range forces) has usually been limited to presenting the force-distance relation alone. Due to the complexity of the interaction between many forces (e.g. London dispersion, electrostatic, capillary and others), the approach of using physics models accounting for each of the relevant forces often fails to provide useful understanding of the resultant surface properties. Here we present an approach that avoids relying on complex physical models by suitably parameterizing force profiles to obtain experimentally meaningful quantities that do not depend on specific knowledge of the fundamental forces at work. We show based on measurements of several samples how parameters obtained directly from the force profile can be correlated to chemistry-related physical properties of the experimental system, without any dependence on fundamental physics-based models.

For more than 40 years, Plutonium and especially its phase diagram has been widely studied from both experimental and theoretical points of view. Nevertheless, it is still a challenging material to study and many questions remain opened in order to understand the phase transitions at low temperature and the dynamical stability of its structures.
It is experimentally observed that pure delta-Pu is not stable below 500K but can be stabilized with deltagen elements, such as Ga, Al, Am. However, numerical simulations based on first-principle calculations always found the delta-Pu phase stable even with the most sophisticated approximations.
In our study, we first reassess the dynamical stability of the delta phase by mean of Density Functional Theory combined with a Hubbard correction. We point out the critical role played by crystal symmetries in the stabilization of pure delta-Pu. Secondly we show the main influence of impurities – such as Ga, Fe, Al, U – on the stabilization of Pu in the low temperature range.

Energetic ions traveling in solids deposit energy in a variety of ways, being nuclear and electronic stopping the two avenues in which excitations are usually treated. This separation between electrons and ions relies on the adiabatic approximation in which both sub-systems remain disconnected. In a more detailed view, in which non-adiabatic effects are explicitly considered, electronic excitations weaken the atomic bonding, which translates into forces driving atoms apart. In this work, we use time dependent density functional theory (TD-DFT) and Ehrenfest dynamics to study first the non-adiabatic forces on ions. These forces translate into momentum transfer to ions well above the adiabatic prediction. At short time scales, most of the alteration of the adiabatic potential comes from the ionization of the core levels of the target ions, and it is only later that the usual picture of ’hot’ electrons becomes the dominant contribution. Second, we present a model for non-adiabatic classical molecular dynamics simulations that captures with high accuracy the wave-vector q dependence of the phonon lifetimes in agreement with quantum mechanics calculations. The model is parameter free, as its components are derived from ab inito-type calculations, and is readily extended to the case of alloys. We also show how this model removes some oversimplifications of the traditional ionic damped dynamics commonly used to describe this interaction.
Work performed at the Energy Dissipation to Defect Evolution Center, an Energy Frontier Research Center funded by the U.S. Department of Energy (Award Number 2014ORNL1026). This research used resources provided by the LANL Institutional Computing Program.

Squaraines are well suited to organic photovoltaics because of their high extinction coefficients over a broad wavelength range from visible to near infra-red (NIR). Moreover, their side groups can be changed with profound effects upon their ability to crystallize, leading to improvements in charge mobility and exciton diffusion. However, our group has shown the packing structure influences the rate of charge transfer and a complete and accurate reassessment of the excited states must be completed before the true charge transfer mechanism can be confirmed.
In this work we uncover the contribution of an intermolecular charge transfer state in a series of squaraines through essential states modeling validated by spectroscopic and X-Ray diffraction data. We demonstrate a connection between the presence of hydroxyl side groups, the size of anilinic alkyl groups, the packing structure and the relative population of this charge transfer state, which broadens the absorption spectrum into the NIR. We conclude by showing device data that describe the influence of the intermolecular charge transfer state on the power conversion efficiency.
We highlight that a complete understanding of the excited states in squaraine crystals will lead towards a prescription for derivatives that can be tailored for optimized exciton diffusion, charge transfer, higher mobilities and reduced recombination in small molecule OPV devices that promise reproducibility advantages over their polymeric counterparts.

Over the past decade, various high-throughput screening projects have exploited the power of modern computers in order to assess and compare known bulk materials using high-level electronic structure theory. The Materials Project, for example, has performed calculations on over 66,000 compounds[1,2] in order to help accelerate the process of materials discovery. But what fraction of chemical space does the number of known compounds analysed by this and similar enterprises represent? The task of exploring new combinations of the periodic table is a daunting one; forming a four component compound from the first 103 elements results in excess of 10¹² potential combinations. Such materials space is intractable to high-throughput experiment or first-principles computation and in order to tame this combinatorial explosion we can turn instead to an arsenal of rules that are the product of centuries of research.
We present the open-source SMACT (Semiconducting Materials by Analogy and Chemical Theory) package,[3] which implements such rules in the search through materials space in order to remove implausible compounds from the results. This includes the programmatic application of concepts such as charge neutrality and electronegativity. We go on to show how heuristic tools such as the Solid State Energy Scale as proposed by Pelatt et al.[4] can be employed to rapidly estimate key properties of chemical compositions such as band gaps and absolute electron energies. Crucially, these estimations can be made even before considering structure.
We illustrate this methodology with a search for new ternary chalco-halide materials for use as photoelectrodes for water splitting and show that the element compositions predicted as having desired properties are ideal candidates for structure predictor algorithms such as the probabilistic model proposed by Hauter et al.[5] We further demonstrate screening by structural analogy in assessing possible ternary combinations of elements that satisfy the radius ratio rule to form a perovskite lattice. We conclude that low-cost screening can be highly complementary to first-principles techniques as a means to navigate the combinatorial materials landscape.

Bibliography
[1] http://www.materialsproject.org (Accessed 01.04.16).
[2] P. Ong, W. D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V. L. Chevrier, K. A. Persson, and G. Ceder, Comput. Mater. Sci., 68, 314 (2013)
[3] http://www.github.com/WMD-group/SMACT
[4] B. D. Pelatt, R. Ravichandran, J. F. Wager, D. A. Keszler, J. Am. Chem. Soc., 133, 16852 (2011)
[5] G. Hautier, C. Fischer, V. Ehrlacher, A. Jain and G. Ceder, Inorg. Chem., 50, 656 (2011)

Experiments have shown that the oxygen storage capacity (OSC) of nano-ceria is size [1] and shape dependent [2]. For example, a dramatically increased OSC has been measured for very small nanoparticles (d < 5 nm) [1]. Based on previous theoretical calculations [3], this effect was suggested to originate from the adsorption of oxygen as superoxide ions at the corners and edges of octahedrally shaped ceria nanoparticles, thereby “supercharging” them [3]. As the magnitude of the adsorption energy is a key aspect of this mechanism (not too strong, not too weak), the theoretical method used in the mechanistic prediction needs to be able to calculate the oxygen adsorption accurately. Therefore, in the current study, we set out to refine our earlier predictions using a higher level of theory than in Ref. [3].

Here we have used the hybrid density functional HSE06 and made a modified version of it, where we tuned the amount of exact exchange to achieve better agreement with experimental data, especially with respect to the electronic structure and reduction energy of ceria, which are both closely connected to the computation of accurate oxygen adsorption energies. Using this modified functional, we find that the adsorption energy of O₂ molecules on ceria increases and brings the results into very good agreement with experimental observations. We conclude that the supercharge mechanism proposed earlier by us to explain the dramatically increased OSC seen in experiments stands firm. Moreover, we put forward a modified HSE06 functional for calculations of the interaction of oxygen with ceria.

[1] J. Xu et al., Chem. Commun. 46, 1887–1889 (2010)
[2] Z. Wu et al., Langmuir 26, 16595–16606 (2010)
[3] J. Kullgren, K. Hermansson, P. Broqvist, J. Phys. Chem. Lett. 4, 604–608 (2013)

There is a pressing need to systematically screen materials for surface related properties, such as photocatalytic and catalytic properties, through high-throughput calculations. The majority of high-throughput materials exploration efforts employ a DFT dataset of perfect crystals. However, construction of another DFT database by systematic surface calculations is essential to achieve better performance on surface-related property screening. There are a number of major issues that must be overcome to accomplish this task, and three examples are given below.

1) Automatic slab model generation. Nonpolar slabs are generally preferred over polar ones in high-throughput calculations as the latter require case-by-case treatment to avoid the "polar catastrophe”. In addition, case-by-case surface reconstruction becomes more relevant in the latter. Furthermore, a scheme is needed to automatically generate unique surface termination models given the crystal, orientation, and minimum slab and vacuum thicknesses [1].

2) Automatic nonpolar orientation determination. Checking all orientations to see whether a nonpolar slab can be obtained or not is inefficient. Moreover, logic must be employed to automatically choose specific orientations over others. In addition, the number of symmetry search and supercell generation should be reduced as much as possible for the sake of efficiency.

3) Converging calculations. Calculations are carried out using a slab-and-vacuum model under three-dimensional periodic boundary conditions, and relaxation of the surface is allowed in the calculations to obtain a reasonable model of the surface. However, excessive relaxation of the surface can destroy the bulk-like region at the center of the slab, especially when the bulk has high formation energy compared to the lowest energy polymorph.

We will show how these issues could be addressed and actually demonstrate generation of a high-throughput database of prototypical binary oxide surfaces. The surface energy is investigated first as an important quantity that determines which surface is likely to appear. Moreover, we focus on the ionization potential as a representative surface-dependent property.

[1] Y. Hinuma et al., Comp. Mater. Sci. 113 (2016) 221.

Phase field modeling has become significantly more popular in materials science and engineering given rapid improvements in computational power and numerical methods, and is now becoming a mainstream technique. As a result, the growing phase field community continues to develop a wide variety of codes, but it lacks benchmark problems to consistently evaluate, validate, and verify new implementations. The development of benchmark problems will enable the results of quantitative phase field models to be confidently incorporated into integrated computational materials science and engineering (ICME), an important goal of the Materials Genome Initiative. Following the example set by the micromagnetics community, the Center for Heirarchical Materials Design (CHiMaD) and the National Institute of Standards and Technology (NIST) are developing benchmark problems that test individual numerical or physical aspects of the codes. We discuss the benchmark problems developed to date, which focus on the diffusion of solute and the growth and coarsening of a second phase, linear elastic solid mechanics, and solidification. We demonstrate the utility of these problems by comparing the results of simulations performed with different numerical techniques. Finally, we discuss the needs of future benchmark problems and how the community can be involved.

This work was performed under financial assistance award 70NANB14H012 from U.S. Department of Commerce, National Institute of Standards and Technology as part of the Center for Hierarchical Material Design (CHiMaD). We gratefully acknowledge the computing resources provided on Blues and Fission, high-performance computing clusters operated by the Laboratory Computing Resource Center at Argonne National Laboratory and the High Performance Computing Center at Idaho National Laboratory, respectively.

CdTe is the most successful thin-film PV (TFPV) technology in the market to date. Although recent improvements in CdTe TFPV boosted module efficiency beyond 18%, the technology still has a strong opportunity for advancement in two main directions: efficiency increase by enhancing open-circuit voltage and reduction of metastabilities and degradation rates. Both directions rely heavily on achieving high and stable p-type absorber doping. In this work, we present results of experimental and theoretical studies that can explain defect chemistry behind Cu doping in CdTe and resulted in the demonstration of near 100% activation of Cu doping in polycrystalline CdTe absorber.

The proposed formation mechanism of Cu acceptors in CdTe absorbers involves two mobile species (copper and cadmium interstitial donors) as well as an immobile substitutional copper on a cadmium site. To become an acceptor, interstitial Cu donor replaces Cd on its site in crystalline lattice while forming an interstitial double-charged Cd donor. In this work, we demonstrate that removal of the created interstitial cadmium donors is critical to observe Cu activation and achieving unconditional stability of p-type doping.

Thermodynamic and kinetic simulations of Cu incorporation and activation in CdTe have revealed strong dependence of Cu solubility and Cu activation percentage on the boundary conditions for Cu and Cd species. As a result, the doping efficiency depends on the surface reactions and the properties of the Cu source. Facilitating surface desorption of cadmium products allowed us to achieve CV-measured hole density of 1e16 cm-3 for 2e17 cm-3 atomic Cu concentration in polycrystalline CdTe film. To our knowledge, this is the highest Cu doping efficiency ever obtained in CdTe. Taking into account the 0.23 eV ionization energy of substitutional Cu acceptor calculated using VASP HSE, achieved p-type doping corresponds to the theoretical maximum for the given atomic incorporation. Obtained doping shows no degradation over the course of high temperature accelerated-life-testing.

Technoeconomic modeling suggests that the current cost structure of photovoltaics (PV) manufacturing needs dramatic reductions in manufacturing costs to scale up deployment to impact climate change.[1] One way to achieve such reductions is to look to nontraditional materials that maintain their favorable optoelectronic properties even in the presence of the structural defects often introduced in less capital-intensive manufacturing techniques. We refer to such materials as “defect tolerant.”

In order for defects to be less detrimental to transport properties, they must be shallow rather than deep defects. Prior work has identified characteristics of the electronic structure needed to achieve such defect tolerance.[2] We have previously screened for defect-tolerant materials that exhibit these electronic structure characteristics in the valence band, typically found in materials containing partially oxidized heavy cations with significant s-orbital contribution to the top of the valence band.[3] While defect tolerance in the valence band allows shallow acceptor states, the screening relied on increased dispersion due to spin-orbit coupling to achieve shallow donor defects with energy levels close to conduction band edge.

In this work, we propose computational screening metrics to achieve an analogous type of defect tolerance in the conduction band to that previously demonstrated in the valence band. These materials have s-like bonding states comprising the bottom of the conduction band. We discuss DFT defect calculations predicting shallow defect levels as well as how a hybrid between these two materials classes could be achieved that would have both shallow acceptors and donors. Such a material could enable a revolutionary drop in the cost of PV and would likely have impact in other optoelectronic applications, as well.


[1] Needleman, D. B., Poindexter, J. R., Kurchin, R. C., Peters, I. M., Wilson, G., & Buonassisi, T. (2016). Economically Sustainable Scaling of Photovoltaics to Meet Climate Targets. Energy & Environmental Science, 9, 2122–2129. http://doi.org/10.1039/C6EE00484A

[2] Zakutayev, A., et al. (2014). Defect tolerant semiconductors for solar energy conversion. Journal of Physical Chemistry Letters, 5(7), 1117–1125. http://doi.org/10.1021/jz5001787

[3] Brandt, R. E., Stevanović, V., Ginley, D. S., & Buonassisi, T. (2015). Identifying defect-tolerant semiconductors with high minority-carrier lifetimes: beyond hybrid lead halide perovskites. MRS Communications, 5(2), 265–275. http://doi.org/10.1557/mrc.2015.26

In this work, a theory that exploits four observables in bimodal atomic force microscopy to produce maps of the Hamaker coefficient H is presented. The quantitative H maps can be employed by the broader community to directly interpret the high resolution of standard bimodal AFM images as chemical maps while simultaneously quantifying chemistry in the non-contact regime. In addition, We provide a simple methodology to optimize the range of operational parameters for which H is in closest agreement with the Lifshitz theory in order to 1) simplify data acquisition and 2) generalize the methodology to any set of cantilever-sample systems.

Understanding transport properties is the key to finding new materials and improving the performance of electronic devices. To gain such an understanding, first-principle computational methods have been used in the past as powerful tools to study transport phenomena; but due to the computational burdens involved and their limited applicability in the far-from-equilibrium realm, their application has been circumscribed. A novel first-principle, nonequilibrium thermodynamic-ensemble approach based on steepest-entropy-ascent quantum thermodynamics (SEAQT) is presented here as an alternative method for studying transport phenomena and predicting the thermal and electric transport properties of materials. Recent advances in SEAQT enable the modeling of far-from-equilibrium phenomena using thermodynamic principles (e.g., the Onsager relations and the Casimir condition), which until recently had been shown to only be applicable in the near-equilibrium realm. The exploitation of these principles in our model provides a significantly faster computational tool when compared to methods based on quantum/classical mechanics, e.g., molecular dynamics and the quantum and classical Boltzmann equations. With the help of Density Functional Theory (DFT), an ab-initio nonequilibrium thermodynamic model of the far-from-equilibrium behavior of charge carrier and point defect transport at an Al/SiO₂ interface is developed and presented here.

Theoretical study of atom diffusions in amorphous oxides is important for many applications such as non-violate memory devices and Li batteries. Reliable computational methods like density functional theory (DFT) can clarify the atomic diffusion behavior, but requires heavy computation, especially for the diffusion in an amorphous structure.

In this study, we examined (I) Single Cu atom diffusion in amorphous Ta₂O₅ (a-Ta₂O₅) and (II) Li atom diffusion inside amorphous Li₃PO₄ (a-Li₃PO₄) using neural network potentials which has both computational efficiency and sufficient accuracy. In doing so, we simplified the up-to-date high-dimensional neural network interatomic potential [1] as follows. In Case I, we used one neural network to learn the relation between local chemical environment of the Cu atom and energy variation along the diffusion path. In system II, the energy contribution of Li atoms were considered precisely with neural networks, while non-diffusive atoms (i.e. P and O) are considered with very small size neural networks. The number of parameters in the simplified neural network potentials (Case I ≈500; Case II ≈2000) are much smaller than that in a typical neural network potential (in a ternary system ≈8000). Thus, simplified neural network potential need much fewer training data.

In both Case I and Case II, the diffusion paths and corresponding activation energies obtained the neural network potentials and the nudged-elastic-band method agree well with those obtained by DFT, while the calculation speed was several orders of magnitude faster. We also constructed the diffusion path network, and extracted the effective diffusion activation energies.

[1] J. Behler and M. Parrinello, Phys. Rev. Lett. 98, 146401 (2007).

We present a combined first-principles simulation and experimental study of the effect of electron-phonon interaction on phonon transport at the single-mode level. There are growing interests in understanding mode-by-mode information of electron and phonon transport for improving energy conversion technologies, such as thermoelectrics and photovoltaics. Our first-principles simulation of electron-phonon interaction in silicon predicts that the phonon transport can be significantly suppressed by free charge carriers with their concentration above 10¹⁹ cm^-3, in contrast to previous beliefs. On the experimental side, although remarkable progress has been made in probing phonon-phonon interactions, it has been a challenge to directly measure electron-phonon interaction at the single-mode level. Here we use three-pulse photoacoustic spectroscopy to investigate the damping of sub-THz coherent longitudinal acoustic phonons by free charge carriers in silicon at room temperature. Building upon conventional pump-probe photoacoustic spectroscopy, we introduce an addition laser pulse to optically generate charge carriers, and carefully design temporal sequence of the three pulses to unambiguously quantify the scattering rate of a single phonon mode due to electron-phonon interaction. We find that at carrier concentrations around 10¹⁹ cm^-3, scattering by charge carriers provides a dominant contribution to longitudinal acoustic phonon decay at 250 GHz, exceeding the contributions of phonon-phonon and boundary scattering. Our results support predictions based on first-principles calculations and indicate the importance of the often-neglected effect of electron-phonon interaction on phonon transport in doped semiconductors. This work is supported by S³TEC, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Basic Energy Sciences, under Award No. DE-FG02-09ER46577.

The electromigration (EM) effect has been discovered in the early 20^th century. Huntington et. al. firstly proposed the “semi-ballistic model” based on momentum transfers between electrons and ions. The net force exerted on metal ions is the sum of electrostatic forces and electron wind forces. A characteristic index, effective charge z*, is proposed and have been extensively measured experimentally since then. At about the same time, Bosvieux and Friedel proposed the “charge polarization model”, suggesting that the force exerted on the metal ions results from the redistribution of its surrounding charges, namely the electron perturbation originated from the external field. Based on that, Sorbello systematically calculated theoretical effective charge, z*, with the aid of pseudo-potentials and obtained good agreements with experiments. However, none of the existing models has coupled the EM effect with lattice stability and thus the existing models cannot explain the lattice distortion and non-polarity effect in experiments. In this work, in situ synchrotron X-ray diffraction measurements for stripes under current stressing and corresponding ab initio calculations based on density functional theory (DFT) were performed for various pure elements and alloys. At early stage of current stressing, unequal lattice expansions were observed along the stripes; establishing a tensile strain gradient from the cathode towards the anode. Ab initio calculations show that more electron perturbation would cause more lattice expansion, which agrees closely with the in situ experiments. Therefore, the conclusion can be drawn that electrons at the cathode possesses more energy and would induce more electron perturbation. As the results, a strain gradient exists under electric current stressing. The electric current-induced strain gradients provide the driving force for atomic migration. The residual stresses/strains under electric current stressing also provide the driving force for the peculiar non-polarity effect. This combinatorial research with in situ synchrotron X-ray radiation and ab initio calculations reveals the thermodynamic foundation of the EM effect.

Previously, we have developed a collective molecular chemistry analysis approach that utilizes data mining techniques to extract molecular level chemistry information from the micro-scale Raman spectra of natural organic matter (OM) such as kerogen¹. One critical element of this approach is a molecular Raman spectra database containing a rich diversity of hydrocarbon molecules. Using ab initio calculations, we built a "kerogen genome database" (KGD) which contains DFT calculated Raman spectra of thousands of hydrocarbon molecules. However, the full potential of the KGD extends beyond being a part of our collective molecular chemistry analysis technique. We demonstrate one way of utilizing the KGD beyond the collective analysis with recovering the empirical relations between the aromatic cluster size and the Raman peak positions through purely statistical analysis over ensembles of molecular Raman spectra extracted from the database.

1 Liu, Y.*, Ferralis, N.*, Bryndzia, L. T. & Grossman, J. C. Genome-inspired molecular identification in organic matter via Raman spectroscopy. Carbon 101, 361-367, (2016).

The ionization ability of a compound in water, defined by its pK_a value, is a fundamental property particularly relevant to the pharmaceutical industry, where an estimated 63% of all drugs are ionizable[1]^,[2]. The pK_a parameter controls compound solubility and dissolution rate, and as a consequence has a great impact on its pharmacokinetic (PK) and pharmacodynamic (PD) profiles. In addition to these intrinsic properties, the pK_a value also determines whether or not a salt containing the drug can be made along with its compatibility with excipients in the drug product[3]. Early on in the development of a new compound, material may be scarce and expensive to synthesize, therefore computational predictions of physicochemical properties such as the pK_a can be invaluable when assessing the candidate’s likely success. These predictive methods are integral and vital parts of modern drug discovery[4].
Here we discuss the prediction of the pK_a using density functional theory (DFT) methods in combination with polarizable continuum calculations. First, in order to assess the ability of various contemporary DFT functionals for predicting the gas-phase deprotonation energy of a compound, high-level W1U calculations have been carried out on a subset of compounds for benchmarking purpose. Second, having retained the best levels of theory for the gas phase energy calculations, the proton exchange thermodynamic cycle was used in conjunction with several implicit solvent models (i.e. CPCM, SMD, and IPCM) to calculate the final pK_a value. Third, owing to the proton exchange methods sensitivity to the choice of reference acid, a fingerprint similarity scheme was developed to intuitively select the best reference. The predictive capability and accuracy of our approach in determining the pKa values of 198 acid-base pairs will be discussed in this presentation.



[1] J. I. Wells, Pharmaceutical preformulation: the physicochemical properties of drug substances, E. Horwood (1988).
[2] J. Comer, K. Tam, In Pharmacokinetic Optimization in Drug Research: Biological, Physicochemical, and Computational Strategies; B. Testa, H. van de Waterbeemd, G. Folkers, R. Guy, Eds. WileyVCH: Weinheim, New York (2001), pp. 275-304.
[3] G. Cruciani, F. Milletti, L. Storchi, G. Sforna, and L. Goracci, In silico pK_a prediction and ADME profiling, Chemistry & Biodiversity, 6, 1812–1821 (2009).
[4] C. Liao and M. C. Nicklaus, Comparison of nine programs predicting pK_a values of pharmaceutical substances, Journal of Chemical Information and Modeling, 49, 2801–2812 (2009).

Defects in materials can be regarded as long-lived excited states. As such, non-equilibrium defect populations created by neutron or ion bombardment, or laser irradiation, can store part of the energy employed to generate them. This work introduces the idea of storing energy in materials defects. We compute the amount of stored energy in a wide range of elemental materials, including tungsten, silicon, graphite, diamond, and graphene, and for several point defects such as vacancies, interstitials, and Frenkel pairs. Employing first principles calculations as well as previously measured and computed data, we demonstrate that a 1 at.% concentration of defects can store a significant amount of energy per volume and weight, 5 MJ/L and 1.5 MJ/kg respectively. For materials with covalent bonds, the stored energy densities are comparable to advanced solid-state storage technologies such as state-of-the-art batteries and supercapacitors. Practical ways to store and extract energy using defects are discussed.

The development of superhard materials has recently focused on systems containing transition metals with high valence electron density and light main group elements. The combination of these atom types provides the highly directional covalent bonding and incompressibility necessary to achieve the desired mechanical response. Rhenium sub-nitrides, Re₂N and Re₃N, are a perfect example of satisfying both of these criteria yielding high hardness (≈15 GPa to 20 GPa) and an extreme high bulk modulus (≈390 GPa). However, the high volatility of nitrogen at the high temperatures required to synthesize these phases makes them prone to the formation of high concentrations of nitrogen vacancies, which may alter the electronic and adversely affect the mechanical properties. Using first principle calculations, the vacancy formation energy in Re₂N and Re₃N is predicted at various pressures while their effect on the electronic structure and mechanical properties is investigated. The results indicate spontaneous vacancy formation in Re₂N at relatively low pressures (<10 GPa) up to about 12.5% nitrogen vacancy concentration. It is demonstrated that large concentrations of Re₃N nitrogen vacancies could be stabilized within the structure at high pressures. Including nitrogen vacancies in the crystal structures shows deterioration of bulk modulus, shear modulus, and the ultimate strength of these nitrides. Therefore, it is critical to control the reaction conditions in these materials to achieve the desired mechanical response. Nonetheless, despite the promising mechanical response of rhenium nitrides, Re is an extremely expensive and scarce starting material hindering its potential for large-scale application. Therefore, we have also initiated the development of new earth-abundant high hardness materials using high-information density plots based on a combination of first principle calculations and data-mining. This method is ideal to quantitatively balance mechanical response with sustainability allowing only viable compositions to be synthesized.

Peptide-based materials is an area that has seen a surge in interest recently due to its wide range of potential applications including targeted drug delivery, cancer treatment, tissue engineering, treatment of neurodegenerative diseases, nanoelectronics and antimicrobial surfaces. In this project, a coarse-grained model is employed in conjunction with Molecular Dynamics simulations to study the structure and dynamical processes underlying the aggregation behavior of a wide range of di- and tripeptides. The self-assembly of these molecules into various ordered nanostructures such as vesicles, nanotubes, nanorods and micelles, and disordered structures such as hydrogels is explored. In addition, mixed peptide systems are considered and the effect of their relative concentrations is investigated.

Initially developed by DuPont in 1941¹ for textile applications, polyacrylontrile (PAN) became more popular for producing carbon fibers with high tensile strength, high modulus and high carbon yield. In recent years, combining with carbon nanotubes (CNTs) to produce highly efficient PAN-CNT composite is a major interest for the carbon fiber community. With high internal tensile strength of individual tubes, CNT tends to template the graphitic interfaces during carbonization.² Even adding small amounts (<10 wt%) of CNT to the composite can dramatically change the mechanical properties of carbon fibers.² With this promise of improving mechanical properties of carbon fibers however, the interaction of PAN and CNT is still poorly reported. Specifically, a thorough understanding of early stage preparation of fibers such as solution processing of CNT-PAN composite is absolutely necessary to produce uniform and defectless precursor fibers. Dispersing agglomerated hexagonal lattice of SWNT bundle of diameters varying 5 to 100 nm with high binding energy between the tubes (~900 meV/nm)³ is a tremendous challenge to date. Debundling CNTs and thereby producing homogeneous mixture of PAN-CNT dispersion is one of the main goals for the processing stage of the composite mixture. To the best of our knowledge, here we represent the first atomistic study of PAN-CNT composite dispersion in three widely used solvents DMF, DMSO and DMAc including virtual electrons on the CNT backbone using modified PCFF-INTERFACE force field.⁴ The dispersion nature and specific molecular interactions were studied in dilute to concentrated (1 - 10 wt%) solutions both at lower and elevated temperatures (25 -75 °C). Although weak, some intermittent interactions of nitrile groups and hydrogen of polymers with the CNT surface were detected in the solvated systems. The surface interactions with CNT actually are found to be more prominent for the solvents (DMF and DMSO) than that for PAN(s). Comparing with experimental observations,^2,5 molecular dynamics study of polymer-polymer, polymer-solvent, CNT-polymer and CNT-solvent interactions in the multicomponent systems will be discussed in detail.

^[1]Engineering Materials: Applied Research and Evaluation Methods Ali Pourhashemi CRC Press 2015
^[2]Newcomb, B. A.; Giannuzzi, L. A.; Lyons, K. M.; Gulgunje, P. V.; Gupta, K. ; Liu, Y. ; Kamath, M. G.; McDonald, K.; Moon, J.; Feng, B.; Peterson, G. P.; Chae, H. G.; Kumar, S. Carbon 2015, 93, 502
^[3]Girifalco, L. A.; Hodak, M.; Lee, R. S., Phy. Rev. B 2000, 62, 13104
^[4]Heinz, H.; Lin, T.-J.; Mishra, R. K.; Emami, F. S. Langmuir, 2013, 29,1754
^[5]Newcomb, B. A.; Gulgunje, P.V.; Liu, Y., Gupta, K. K.; Kamath, M.G.; Pramanik, C.; Ghoshal, S.; Chae, H. G.; Kumar, S. Polym. Eng. Sci. 2015, 56, 361

Solar thermal fuels (STFs) is an unconventional paradigm for solar energy conversion and storage which is attracting renewed attention. A material absorbs sunlight and stores the energy chemically via an induced structural change, which can later be reversed to release the energy as heat. An example is the azobenzene molecule which has a cis-trans photoisomerization with these properties, and can be tuned by chemical substitution and attachment to templates such as carbon nanotubes, forming energy-harvesting nanostructures [A. M. Kolpak et al., Nano Lett. 11, 3156 (2011); T. Kucharski et al., Nat. Chem. 6, 441 (2014)]. By analogy to the Shockley-Queisser limit for photovoltaics (PV), we analyze the maximum attainable efficiency for STFs from thermodynamic considerations. We show a bound on the quantum yield of photoisomerization due to fluorescence, regardless of details of photochemistry. We emphasize the importance of analyzing the free energy, not just enthalpy, of the metastable molecules, and find an efficiency limit for conversion to stored chemical energy equal to the Shockley-Queisser limit for PV. We use these results to assess candidate molecules from high-throughput studies and determine key properties needed for practical applications.

Nanowires (NWs), being one of the exciting type of the nanomaterials, have been heavily studied owing to their remarkable mechanical, optical, electrical properties. These properties facilitate the building of new sensor systems, optoelectronics, nano-electro-mechanical systems (NEMS). Having such vast and charming applications of NWs, the study of their properties turns out to be exceptionally critical and essential. Unique properties are exhibited by nanoscale structures due to quantum size confinement and an extremely large surface-to-volume ratio compared to their bulk-counterparts. Hence, mechanical characterization of these novel nanostructures is crucial in order to fulfill reliability requirements of these devices. A significant work still lies ahead in the mechanical domain due to challenges encountered such as inconsistent nanowire fabrication, manipulation, alignment and attachment in a test setup. So far, many experimental studies have been conducted, however, although experimental methods provide realistic and meaningful information of the mechanical behavior of NWs, due to their extremely small dimensions, the application and manipulation of these experiments involve huge challenges in terms of sample fabrication, handling the small scale sample and also small-scale measurement setup with high resolution. In this study, we aim to provide a numerical approach in order to explore the mechanical properties of single crystal silicon nanowires. An-open-source-code of molecular dynamics simulator named LAMMPS, is used for computations. Specimens studied have a width (in-plane) of 3 nm, height (out-of-plane) of 10 nm, a length of 45 nm. In the computations, Si NWs are modeled in two ways: one is three-point bending which has a double-clamped beam configuration and displacement is forced to occur at the midpoint of the beam, the other one is uniaxial, in which the beam is fixed at one end and displacement happens in the axial direction. The simulations are conducted at the room temperature. In order to get more insight on the true mechanical behavior of silicon nanowires, a native oxide layer, having a thickness of around 15% of NW width, is modeled surrounding the Si NW. Embedded atomic potential model is implemented. The results indicate that at that scale, Si NWs have elastic modulus value of 143.6 GPa for the uniaxial case, and 146.6 GPa for the bending case, and yielding strength of 15.6 GPa and 16.2 GPa, respectively. Further in this study, different sizes of Si NWs will be investigated and results will shed a light on the scale effect in mechanical properties of SiNWs.

The development of dielectric materials has become increasingly desired as the world moves to be more electrified and alternative energy becomes more abundant. A high energy density is needed for parallel plate capacitor applications, leading to a search for high dielectric constant materials. This is driven by the linear relationship between dielectric constant and energy density. In a desire to quickly evaluate the vast field of organic dielectric materials, a co-design approach between density functional theory (DFT) computations and experimental synthesis was devised.^[1] Using band gap as a proxy for breakdown strength, materials with high dielectric constants and high band gaps were rapidly screened and several candidates were synthetically investigated.^[2] Among polyimides, two polymers stood out from the rest. One polyimide had a high dielectric constant ca. 7.8 and subsequent high energy density ca. 15 J/cc but a low glass transition temperature (Tg), while the other had a modest dielectric constant of 3.57 and energy density ca. 10 J/cc with relatively higher Tg. ^[3] Due to the complex size of the copolyimides that would be produced, a novel machine learning approach was used to quickly approximate the dielectric properties and justify the synthetic effort. ^[4,5] The copolyimides were able to maintain a high energy density despite a decrease in dielectric constant while improving their Tg. ^[6] Once experimental results were available, DFT calculations were run on the system to compliment the results and aid in validation of the machine learning technique.

References

[1] G. M. Treich, et al., Advanced Materials DOI: 10.1002/adma.201600377 (2016)
[2] V. Sharma et al., Nature Communications. 5, 4845 (2014).
[3] R. Ma et al., ACS Appl. Mater. Interfaces., 6, 10445 (2014).
[4] G. Pilania, C.C. Wang, X. Jiang, S. Rajasekaran, R. Ramprasad, Scientific Reports 3 (2013)
[5] A. Mannodi-Kanakkithodi et al., Scientific Reports 6 (2016)
[6] G. M. Treich et al., (manuscript under preparation)

Diffusivities of alloying element determine the direction and velocity of many processes in metals. Advancement of computational techniques make the determination of diffusivities of a large spread of different impurities in bcc-iron feasible. Ab-initio calculations have been performed, determining the diffusivities of impurity elements in bcc-iron. Two different exchange-correlation functionals, PW91 and PBEsol, were used to determine the vacancy formation and binding energies and migration barriers. Excess entropies and the attempt frequency for a jump were determined by calculating phonon frequencies in the harmonic approximation. The Le Claire 9-frequency model and transition state theory are used to determine the effective jump frequencies of each diffusing element. It is found that elements with low migration barriers, have a low correlation factor and this correlation is very temperature dependent, which acts as an effective activation energy for diffusion. PW91 and PBEsol give similar results, except for vacancy formation energies, which PBEsol overestimates consistently by approximately 0.5 eV. The influence of magnetism on activation energies for diffusion can be taken into account using either the Hillert-Jarl model or the Girifalco model. Both models, tough based on different principles lead to negligible differences. The diffusivity rates of all substitutionals in iron are controlled by the diffusivity of vacancies in iron. Each element except for Cobalt has a higher diffusivity than iron self-diffusion. Calculated results are compared with experimental data for 21 elements where such data was available. The calculated results of the functional PW91 show good agreement with experiments over most of the wide range of different elements. Both the migration barriers and the vacancy formation and binding energies are determined mostly by their column in the periodic table and much less by their size. The exchange-correlation functional PW91 gives a good agreement with experimental data.

The paucity of high performance p-type semiconductors has been a stumbling block for the electronics industry for decades, effectively hindering the route to efficient p-n junctions. Through the “chemical modulation of the valence band” (CMVB) pioneered by Hosono et al.,¹ copper based oxides and subsequently copper based chalcogenides have been a focal point for designing efficient p-type semiconductors, particularly transparent semiconductors. Our work extends this concept further into the pnictides with a group of ternary copper phosphides, M(II)CuP (where M(II) = Mg, Ca, Sr and Ba) and explores the relationship between coordination and band dispersion alongside explaining the coordination preference of Cu(I) based materials. Using hybrid density functional theory (DFT) we examine the structural and electronic properties of these four materials where we find hole effective masses matching the industry standard n-type TCOs: In₂O₃, ZnO and SnO₂ paving the way towards the design of high performance p-type materials.

1. H. Kawazoe et al. P-type Electrical Conduction in Transparent Thin Films of CuAlO2. Nature 389, 939–942 (1997).
2. B.A.D. Williamson; J. Buckeridge; R.G. Palgrave; D.O. Scanlon. Computational Prediction of Degenerate p-type Semiconductors. (2016), in submission.

The rapid rise in available computing power has made it possible to accelerate materials design by using ab-initio methods to calculate properties for tens of thousands crystalline materials. The calculation of many of these properties requires the evaluation of an integral over the Brillouin zone, which is commonly approximated by sampling a regular grid of points, known as k-points, in reciprocal space. However, finding an efficient k-point grid is not always straightforward. We have addressed this problem by creating a publicly available k-point grid server, backed by a database of tens of thousands of k-point grids, which is capable of rapidly generating highly efficient grids. We estimate that on average the use of grids generated by this server roughly doubles the speed well-converged calculations on crystalline materials. We present the underlying method behind this tool and benchmark results, along with additional features that we have implemented. We will also discuss how our tool can be integrated into an existing computing environment.

Neural network potentials (NNP) involve training an artificial neural network on density functional theory data to produce an interatomic potential to be used in molecular dynamics and other potentials-based techniques. Here a NNP for the alpha and beta phase of Ga₂O₃ is developed from thousands of DFT data points. The phonon band structure and density of states of bulk alpha and beta Ga₂O₃ has been computed using a NNP and compared to results from traditional DFT methods, serving as a test of the fidelity of NNP data. In addition, the NNP model enables us to extract any relevant anharmonic vibrational features. The vibrational characteristics of defects are also investigated using the NNP technique, enabling us to provide more accurate estimates of the vibrational entropy of oxygen and gallium vacancies, as well as dopant-defect pairs involving vacancies and tin/silicon dopants. Such contributions to the Gibbs free energy are important at the high temperatures involved in annealing a highly stable material like Ga₂O₃, and can be challenging to accurately compute using traditional DFT techniques, demonstrating the value of NNPs in this type of investigation. Finally, the details of NNP development for use in vibrational studies are discussed. Guidelines are presented for the production of the initial training data set, methods of analyzing the training output, and quantification of potential sources of error.

Virtual x-ray diffraction profiles using data from first-principles and classical atomistic simulations are used to quantify the uncertainty of minimum energy structures found in boron-based ceramics. Diffraction profiles highlight uniqueness stemming from subtle structural changes that occur when modeling complex materials using different first-principles techniques, basis-sets, pseudopotentials, and classical interatomic potentials. Diffraction profiles that are computed in reciprocal space are easily compared using over 45 distance and similarity metrics. These metrics highlight the structural uncertainty when using various computational approaches, and are used to quantify the fidelity of classical models. When possible, direct comparisons of simulated diffraction patterns are made to experimental data in order to provide additional insights.

Colloidal nanoplatelets are atomically flat, quasi-two-dimensional sheets of semiconductor exhibiting efficient, spectrally pure fluorescence. Their formation is not fully understood and especially surprising when the underlying crystal structure has cubic symmetry, for example in zincblende CdSe. In this case the crystal structure of the material appears to contradict its highly anisotropic shape. Here, we demonstrate that an intrinsic instability in the kinetics of crystal growth on small facets leads to highly anisotropic platelet crystallites even for materials with cubic crystal structure. This instability arises because the nucleation of a stable island–the rate-limiting step in crystal growth–is strongly reduced on narrow facets, as we confirm using density functional theory calculations for CdSe. Kinetic Monte Carlo simulations show, moreover, that the stochastic nature of crystal growth leads to nanoplatelets even from initial crystal seeds that are perfectly cubic in shape. Our model predicts enhanced growth rates only for a small range of narrow facet widths. Even within this range the growth rates rapidly decrease with facet size, in agreement with experimental results. Finally, we map our microscopic mechanism to standard concepts of volume, surface, and edge energy, allowing easy application of the growth instability criterion to many other crystalline materials.

Nanoparticle (NP) self-assembly that result in ordered crystalline arrangements are of interest as they give rise to materials possessing important emergent properties. However, proper control of such assembly processes can be challenging as NPs can easily phase separate out of solution. A popularized method to enhance molecular control of self-assembly is to use DNA to direct NPs assembly through hybridizing attractions between complimentary grafts. Despite initial successes, many efforts focused on the use of spherical NPs, which appear to have a limited range of accessible morphologies. Recent advances in the synthesis of non-spherical particles have directed attention towards using anisotropic particles for self-assembly. Motivated by these studies, we combine experiments and theory on a series of DNA-grafted nanocubes and propose a scaling theory for corona conformation for the resulting DNA-grafted NPs. Our results indicate a preferential partitioning of grafts of differing lengths to positions of maximum curvature on the nanoparticle. This coupling between graft length and surface curvature produces an uneven distribution of DNA on the particle, giving rise to non-uniform coronas that further emphasize the anisotropy of the grafting core. Furthermore, the resulting coronas pack into “lock-and-key” configurations that optimize the interplay between attractive and repulsive interactions between the individual building block. These results strongly suggest that corona-amplified shape complementarity is a powerful handle for precise control of self-assembly.

High Entropy Alloys are random solid solutions with many components. Recent experiments show that many of these materials have very high strength, high ductility, and high fracture toughness, especially at low temperatures. The precise mechanistic origins of these impressive properties have not been identified. Here, a model for the onset of plastic flow, i.e. the yield strength, of fcc HEAs as a function of composition, temperature, and strain rate is presented. The model adopts an effective-medium approximation wherein each element of the alloy is solute within the effective matrix of the alloy and then extends a recent predictive theory for dilute solutions [1,2] to arbitrary concentrations [3]. Direct molecular simulations on a 3-component model alloy, where the effective medium approximation can be executed and solutes introduced into the effective matrix material, show the predictive capability of the model as a function of composition. A simplified version of the theory using elasticity only is then developed for applications when detailed knowledge of solute/dislocation interactions is difficult to obtain. The elasticity theory elucidates the roles of many different solute and alloy properties in controlling the yield strength. The theory is then applied to the Fe-Ni-Cr-Co-Mn family of fcc HEAs and excellent agreement is obtained for the non-Hall-Petch (grain-size-independent) contributions to the strength. This agreement is achieved over a wide range of temperatures for a 4-component and 5-component alloy, and at room temperature for the available subset of binary, ternary, quaternary, and quinary alloys in this family [3]. Overall, this parameter-free model using first-principles input thus provides a basis for achieving the long-sought goal of computational design of alloys, within the context of solute-strengthening mechanisms.

G. P. Leyson, W. A. Curtin, L. G. Hector Jr., and C. Woodward, “Quantitative prediction of solute strengthening in aluminium alloys”, Nature Materials 9, 750-755 (2010).
G. Leyson, L. G. Hector, and W. A. Curtin, “Solute strengthening from first principles and application to aluminum alloys”, Acta Mater. 60, 3873-3884 (2012).
C. Varvenne, A. Luque, and W. A. Curtin, submitted to Acta Materialia (2016).

Within the last years computationally guided design strategies based on finite-temperature ab initio calculations have become a key pillar in materials design. Ab initio Gibbs energy computations have been successfully applied in the past to tackle various materials science problems ranging from phase stabilities of coating materials [1], new magnetocalorics for innovative cooling concepts [2] up to stacking fault energies for tailoring improved high-strength steels. All mentioned examples have one key challenge in common which is the ab initio modeling of magnetic fluctuations and their interplay with other degrees of freedom at elevated temperatures [3]. This talk gives a brief overview on the developed methods, which allow to include magnetic excitations and lattice vibrations on the same footing [4-6] and discusses the predictive power achievable by these new approaches.

[1] L. Zhou, F. Körmann, D. Holec, M. Bartosik, B. Grabowski, J. Neugebauer, and P.
Mayrhofer, Phys. Rev. B 90, 184102 (2014)
[2] B. Dutta, A. Cakir, C. Giacobbe, A. Al-Zubi, T. Hickel, M. Acet, and J. Neugebauer, Phys. Rev. Lett. 116, 025503 (2016)
[3] F. Körmann, T. Hickel, J. Neugebauer, Curr. Opin. Solid. St. M. 20, 77 (2016)
[4] F. Körmann, B. Grabowski, B. Dutta, T. Hickel, L. Mauger, B. Fultz, J. Neugebauer, Phys. Rev. Lett. 113, 165503 (2014)
[5] B. Alling, F. Körmann, B. Grabowski, A. Glensk, I. Abrikosov, and J. Neugebauer, Phys.
Rev. B 93, 224411 (2016)
[6] F. Körmann, P-.W. Ma, S.L. Dudarev, and J. Neugebauer, J. Phys. .Condens. Matter 28, 076002 (2016)

Traditional photovoltaics, comprised of materials like Si and GaAs, are treated as “free carrier” devices, in which the electron and hole are assumed to be independent, non-interacting particles. However, strong electron-hole interactions are common in many emerging semiconductors where they form bound states known as excitons. In this work we investigate the effect of excitons on the fundamental charge transport in novel photovoltaic materials with large exciton binding energies.

Cuprous oxide (Cu₂O) was selected for this work because it has been used as a model system in investigations of Bose-Einstein Condensation of excitons due to its long exciton lifetimes and high exciton binding energy. We develop a thermodynamic model and calculate the fraction of free electrons and holes in Cu₂O at quasi-equilibrium. Using this model, we find that over 20% of the generated population of carriers during photovoltaic operation could be excitons. These findings are corroborated by experimental investigations. Photoluminesence spectra of the exciton peak in Cu₂O at room temperature indicate the presence of large exciton densities under visible light illumination. We find that these large exciton densities contribute to current collection in Cu₂O/Zn(O,S) photovoltaics from spectral response measurements, which indicate that close to 10% of the short-circuit current density in these devices is due to exciton transport and dissociation.

Photovoltaic device performance was further studied using a device model which includes exciton transport. We find that excluding exciton effects in Cu₂O solar cells can underestimate the device performance by over 1.5 absolute percent. This work demonstrates that excitons can play a fundamental role in photovoltaic materials with large exciton binding energies and lays the foundation for further studies to optimize performance in such devices.

Condensed materials lacking crystalline symmetry feature in many naturally occurring processes (e.g. formation of passive films) or during material synthesis (e.g. thin film growth, alloy casting), and are relevant for a wide range of technological applications from electronics to catalysis to energy storage. Yet ab-initio, density functional theory (DFT) based modeling of amorphous materials requires simulations with relatively large number of atoms and exploration of the phase space, and therefore has often been considered a computationally demanding task. Nevertheless, recent advances in computational power allow us to scale up ab-initio molecular dynamics (AIMD) simulations of liquid and amorphous materials to many systems of technological interest. Here we present our recent efforts in building an automated, high-throughput framework to evaluate the properties of amorphous/liquid inorganic materials using AIMD. We will discuss the structural, thermodynamic and kinetic insights gathered from our existing calculations in metal oxide systems, alongside possible projections of these properties to wider chemical spaces using the local structural information. As more advanced technologies demand more complex materials, including metastable crystalline or amorphous materials, we expect the presented framework to find a growing number of applications in first-principles materials design problems involving non-crystalline materials over the next few years.

Thermal batteries based on reversible adsorption/desorption of certain molecules rather than conventional vapor compression is a promising alternative to exploit waste thermal energy for heat and cold reallocation. In this context, there is an increasing interest to search for novel porous solids able to achieve high energy density under low relative vapor pressure condition and with capability of regeneration (desorption) at low temperature, which are the major requirements for adsorption driven heat pumps and chillers. The porous crystalline hybrid materials named Metal–Organic Frameworks (MOFs), represent a great source of inspiration for sorption based-applications owing to their tunable chemical and topological features associated with a large variability of pore sizes [1]. With the wide variety of available choices for adsorbent–adsorbate pairs, the material selection is nontrivial. The good pair candidate depends on the amount of heat that can be extracted from the evaporator per adsorption cycle, which is proportional to the amount of vapor adsorbed and the evaporation enthalpy of the selected working fluid. The adsorbate could be water, methanol, ethanol or ammonia, while the MOFs should present large adsorption capacity and excellent hydrothermal stability. Concerning this last requirement, Al and Zr based MOFs have shown to be the most stable in aqueous environment [2]. Recently we have designed a new MOF named MIL-160, isostructural to CAU-10 [3], based on aluminum chains of corner-sharing octahedral AlO₄(OH)_2. The original organic linker, 1,4-benzenedicarboxylate was replaced to 2,5-furandicarboxylic acid [4]. This strategy proved to enhance both the hydrophilicity and the amount of water adsorbed at low p/p₀. The new solid was synthesized and its chemical stability/adsorption performances verified. We have extended this study by incorporating other polar heterocyclic linkers such as pyridine. In parallel, a series of other Al- and Zr-based MOFs have been selected and a comparative study has been performed. The adsorption isotherms were predicted by using force-field based Grand-Canonical Monte Carlo simulations and validated by experimental adsorption/desorption measurements. The chemical stability of the solids was also investigated by using first-principle calculation (e.g. DFT).
[1] S. K. Henninger, H. A. Habib , C. Janiak , J. Am. Chem. Soc. 2009 ,131 , 2776
[2] J. Canivet et. al. Chem. Soc. Rev., 2014,43, 5594-5617
[3] D. Fröhlich , et. al. Dalton Trans. 2014, 43 , 15300 ;
[4] A. Cadiau et. al. Advanced Material, 2015, 27(32), 4775-4780

The discovery of functional organic materials has established major technological advances in history. With the aid of high-performance computing, efficient in silico molecular design methods based on first-principles calculations have been adopted to expedite materials discovery compared with conventional trial-and-error experimentations. These enable high-throughput screening of potential candidates for target materials through a large-scale exploration prior to chemical synthesis. However, the screening approach would suffer from a low hit rate, while a large amount of computing resources is required.

In order to enhance the efficiency of candidate search, it is more favorable to selectively generate molecules that exhibit desired properties. The accumulated data of known molecules can be a fruitful source for new discovery. In this respect, inverse design methods have been developed to generate promising candidates based on data-driven knowledge. The main idea of these methods is to derive certain indices regarding the desired properties, which can be retrieved as molecular structures. However, they require some assumptions of designing rules or pre-defined molecular fragments to construct molecular structures from the indices, which involves direct incorporation of chemical knowledge. As a different knowledge is required for each application, the molecular design still depends on manual efforts by chemists.

Here we present a novel inverse design method to design new organic molecules with desired properties in a fully data-driven way. With the aid of emerging deep learning techniques, we built a model based on deep encoder-decoder architecture consisting of a deep neural network (DNN) and a recurrent neural network (RNN) learned from given molecular data. In this model, the DNN extracts hidden factors identifying the relationship between structural features and the corresponding properties of molecules in the data, and the RNN takes these factors to derive molecular structures. Deep learning enables to extract latent knowledge from the data, which is effectively used for the automated design of molecules without incorporating any external knowledge.

The effectiveness of the proposed method was demonstrated using PubChem. According to the results, a number of organic molecules that fulfill the target properties were generated with a considerable hit rate. In addition, the application results on the design of dopant materials for organic light-emitting diodes showed that the fully data-driven inverse design successfully works. We expect this method can serve as an enabling tool in searching promising chemical species and optimizing material structures for various applications.

Transformative outcomes for transportation and electrical grid require high performance, low cost energy storage systems beyond Li-ion, such as multivalent batteries, Li-S batteries and innovations in electrodes and electrolytes, alike. In this work, we present a coupled high throughput ab initio and MD simulation approach to understand complex behaviour of electrolytes, and predict their properties to obtain design metric for new improved electrolytes for multivalent and Li-S batteries. We uncover a novel effect between concentration dependent ion pair formation and anion stability at reducing potential, e.g., at the metal anode and used this knowledge to design novel electrolytes for multivalent batteries.^2,3 For Li-S, we study the effect of salt anion and the solvent on the solvation structure and dynamics of Li-polysulfide in the solution. We observe that counter anion and solvent interaction strength with Li⁺ is critical in controlling polysulfide solubility.⁴ This works shows that the combination of multi-scale modeling with experimental techniques provides unprecedented insight into the origin of the electrochemical, structural, and transport properties of electrolytes, which is crucial in designing electrolytes for beyond Li-ion batteries.

In the context of Materials-by-Design or Materials Genome Initiative (MGI) approaches to materials development, high-throughput experiments (HTE) are used to rapidly create structure-property relationship maps both to test the predictions of theory and to efficiently explore questions more easily addressed through experiment. However, to effectively use HTE in the spirit of MGI requires not only combinatorial materials synthesis and rapid property mapping but also automated data (and metadata) harvesting along with expression of the aggregated data in an accessible database to enable data analysis and data mining. At NREL, we have been using and continuously developing physical vapor deposition based composition gradient approaches to thin film combinatorial library synthesis and the associated spatially resolved property mapping since 2001. The data harvesting development began in 2008 and finally work on our purpose specific HTE analysis database began in 2014. Using the challenge of rapid analysis of large amounts of x-ray diffraction (XRD) and composition data to create structure phase maps as a unifying theme, we will present an overview of the complete NREL HTE ecosystem as it currently stands as well as a vision of how it might be integrated into multi-institute consortia such as envisioned in the recently created DOE Energy Materials Network. Our sputtering and pulsed laser deposition based combinatorial material synthesis capabilities will be demonstrated using examples of 1D and 2D composition gradients as well as 1D composition gradient with an orthogonal 1D temperature gradient covering oxides, nitrides, sulfides and mixed anion materials. The XRD patterns were measured using a commercial (Bruker D-8 Discover) diffractometer equipped with a 2D detector. The cation composition is routinely determined by x-ray fluorescence mapping. Anion composition as well as cation composition is also measured by Rutherford backscattering (RBS), albeit only on an as-needed basis. Our recently established HTE analysis database currently contains data from more than 850 combinatorial libraries representing roughly 23,000 effectively different samples. High-throughput analysis for x-ray diffraction will be demonstrated by application to the Co-Zn-Ni-O material system. In addition to composition and xrd data, the HTE analysis database also includes optical transmission and reflection spectra, electrical conductivity and deposition details information. The database can be accessed via an API or direct SQL queries. A web interface and graphing-program based user tools are being developed to facilitate routine use.