Modeling the behavior of materials composed of elements with different bonding and electronic structure character for large spatial and temporal scales and over a large compositional range is a challenging problem. Cases in point are amorphous alloys of Si, a prototypical covalent material, and Li, a prototypical metal, which are being considered as anodes for high-energy-density batteries. To address this challenge, we develop a methodology based on neural networks that extends the conventional training approach to incorporate pre-trained parts that capture the character of different components, into the overall network; we refer to this model as the ``implanted neural network'' method. We show that this approach works well for the Si-Li amorphous alloys for a wide range of compositions, giving good results for key quantities like the diffusion coefficients. The method is readily generalizable to more complicated situations that involve two or more different elements.

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

Inverse Compton scattering (ICS) is a unique mechanism for producing fast pulses - picosecond and below - of bright X- to gamma-rays. These nominally narrow spectral bandwidth electromagnetic radiation pulses are efficiently produced in the interaction between intense, well-focused electron and laser beams. The spectral characteristics of such sources are affected by many experimental parameters, such as the bandwidth of the laser, and the angles of both the electrons and laser photons at collision. The laser field amplitude induces harmonic generation and importantly, for the present work, nonlinear red shifting, both of which dilute the spectral brightness of the radiation. As the applications enabled by this source often depend sensitively on its spectra, it is critical to resolve the details of the wavelength and angular distribution obtained from ICS collisions. With this motivation, we present here an experimental study that greatly improves on previous spectral measurement methods based on X-ray K-edge filters, by implementing a multi-layer bent-crystal X-ray spectrometer. In tandem with a collimating slit, this method reveals a projection of the double-differential angular-wavelength spectrum of the ICS radiation in a single shot. The measurements enabled by this diagnostic illustrate the combined off-axis and nonlinear-field-induced red shifting in the ICS emission process. They reveal in detail the strength of the normalized laser vector potential, and provide a non-destructive measure of the temporal and spatial electron-laser beam overlap.

Inverse Compton scattering of laser photons by ultrarelativistic electron beam provides polarized x- to gamma-ray pulses due to the Doppler blueshifting. Nonlinear electrodynamics in the relativistically intense linearly polarized laser field changes the radiation kinetics established during the Compton interaction. These are due to the induced figure-8 motion, which introduces an overall redshift in the radiation spectrum, with the concomitant emission of higher order harmonics. To experimentally analyze the strong field physics associated with the nonlinear electron-laser interaction, clear modifications to the angular and wavelength distributions of x rays are observed. The relativistic photon wave field is provided by the ps CO2 laser of peak normalized vector potential of 0.5 < a(L) < 0.7, which due to the quadratic dependence of the strength of nonlinear phenomena on a(L) permits sufficient effects not observed in past 2nd harmonic study with a(L) approximate to 0.3 laser [M. Babzien et al., Phys. Rev. Lett. 96, 054802 (2006)]. The angular spectral characteristics are revealed using K-, L-edge, and high energy attenuation filters. The observation indicates existence of the electrons' longitudinal motion through frequency redshifting understood as the mass shift effect. Thus, the 3rd harmonic radiation has been observed containing on-axis x-ray component that is directly associated with the induced figure-8 motion. These are further supported by an initial evidence of off-axis 2nd harmonic radiation produced in a circularly polarized laser wave field. Total x-ray photon number per pulse, scattered by 65 MeV electron beam of 0.3 nC, at the interaction point is measured to be approximately 10(9).

We use machine-learning methods on local structure to identify flow defects-or particles susceptible to rearrangement-in jammed and glassy systems. We apply this method successfully to two very different systems: a two-dimensional experimental realization of a granular pillar under compression and a Lennard-Jones glass in both two and three dimensions above and below its glass transition temperature. We also identify characteristics of flow defects that differentiate them from the rest of the sample. Our results show it is possible to discern subtle structural features responsible for heterogeneous dynamics observed across a broad range of disordered materials.

Thin film photovoltaic cells are increasingly important for cost-effective solar energy harvesting. Layered SnS is a promising absorber material due to its high optical absorption in the visible and good doping characteristics. We use first-principles calculations based on density functional theory to study structures of low-index surfaces of SnS using stoichiometric and oxygen-containing structural models, in order to elucidate their possible effect on the efficiency of the photovoltaic device. We find that the surface energy is minimized for the surface with orientation parallel to the layer stacking direction. Compared to stoichiometric surfaces, the oxygen-containing surfaces exhibit fewer electronic states near the band gap. This reduction of near-gap surface states by oxygen should reduce recombination losses at grain boundaries and interfaces of the SnS absorber, and should be beneficial to the efficiency of the solar cell.

Photovoltaic cells based on SnS as the absorber layer show promise for efficient solar devices containing non-toxic materials that are abundant enough for large scale production. The efficiency of SnS cells has been increasing steadily, but various loss mechanisms in the device, related to the presence of defects in the material, have so far limited it far below its maximal theoretical value. In this work we perform first principles, density-functional-theory calculations to examine the behavior and nature of both intrinsic and extrinsic defects in the SnS absorber layer. We focus on the elements known to exist in the environment of SnS-based photovoltaic devices during growth. In what concerns intrinsic defects, our calculations support the current understanding of the role of the Sn vacancy (VSn) acceptor defect, namely that it is responsible for the p-type conductivity in SnS. We also present calculations for extrinsic defects and make extensive comparison to experimental expectations. Our detailed treatment of electrostatic correction terms for charged defects provides theoretical predictions on both the high-frequency and low-frequency dielectric tensors of SnS.

The nucleation of metastable germanium polymorphs on decompression is studied using in situ synchrotron x-ray diffraction. We show that the transition pathway is critically dependent on the hydrostaticity. Quasihydrostatic conditions result in the nucleation of the rhombohedral r8 phase, followed by the cubic bc8 and hexagonal diamond phases. In contrast, the presence of shear yields the tetragonal st12 phase. Thus, targeted nucleation of a metastable polymorph is now possible. This observation has implications for the technological exploitation of Ge, but also for other tetrahedral systems.

The formation of R8 germanium is reported. The β-Sn phase is first induced by the indentation of amorphous germanium (a-Ge) and the resultant phases on pressure release are characterized by Raman scattering. The expected Raman line frequencies for the various phases of Ge are determined from first-principles calculations using density functional perturbation theory of the zone-center phonons in the diamond, ST12, BC8, and R8 Ge phases. In addition to the R8 phase, traces of BC8 may also be present following pressure release.

We present first-principles calculations of the quasiparticle band structure of the group IV materials Si and Ge and the group III-V compound semiconductors AlP, AlAs, AlSb, InP, InAs, InSb, GaP, GaAs and GaSb. Calculations are performed using the plane wave pseudopotential method and the 'one-shot' GW method, i.e. G(0)W(0). Quasiparticle band structures, augmented with the effects of spin-orbit, are obtained via a Wannier interpolation of the obtained quasiparticle energies and calculated spin-orbit matrix. Our calculations explicitly treat the shallow semicore states of In and Ga, which are known to be important in the description of the electronic properties, as valence states in the quasiparticle calculation. Our calculated quasiparticle energies, combining both the ab initio evaluation of the electron self-energy and the vector part of the pseudopotential representing the spin-orbit effects, are in generally very good agreement with experimental values. These calculations illustrate the predictive power of the methodology as applied to group IV and III-V semiconductors.

SnS is a metal monochalcogenide suitable for use as absorber material in thin film photovoltaic cells. Its structure is an orthorhombic crystal of weakly coupled layers, each layer consisting of strongly bonded Sn-S units. We use first-principles calculations to study model single-layer, double-layer, and bulk structures of SnS in order to elucidate its electronic structure. We find that the optoelectronic properties of the material can vary significantly with respect to the number of layers and the separation between them: the calculated band gap is wider for fewer layers (2.72 eV, 1.57 eV, and 1.07 eV for single-layer, double-layer, and bulk SnS, respectively) and increases with tensile strain along the layer stacking direction (by ∼55 meV/1% strain).

We perform first-principles, density-functional-theory calculations of the electronic structure for the layered bulk materials SnS and GeS which are of interest for photovoltaic applications. Band gap corrections are computed within the GW approximation to the electron self-energy. The resulting quasiparticle gaps in both SnS and GeS are in excellent agreement with the majority of existing experimental measurements. In order to better understand the possible use of GeS layers as a carrier-confining barrier within a SnS-based photovoltaic device, we compute the band offsets for different orientations of a SnS/GeS heterojunction. We find the valence band offsets to be almost independent of interfacial direction while the conduction band offsets show a strong anisotropy as a result of the variation in the band gap caused by epitaxial strain along the different directions.

We have studied the stable end phase formed in amorphous germanium (a-Ge) films that have been subjected to a pressure-induced phase transformation under indentation loading using a large (20 mu m) spherical indenter. After indentation the samples have been annealed at room temperature to remove any residual unstable R8 and BC8 phases. Raman spectroscopy indicates a single broad peak centred around 292 cm1 and we have used first principles density functional perturbation theory calculations and simulated Raman spectra for nano-crystalline diamond cubic germanium (DC-Ge) to help identification of the final phase as hexagonal diamond germanium (HEX-Ge). Transmission electron microscopy and selected area diffraction analysis confirmed the presence of a dominant HEX-Ge end phase. These results help explain significant inconsistencies in the literature relating to indentation-induced phase transitions in DC- and a-Ge. ((c) 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)

The electronic structure and lattice dynamics for simple cubic (sc) P are calculated over the pressure range 0-70 GPa from first principles using the local-density approximation. The R phonon mode is found to be unstable below 20 GPa in the harmonic approximation, but may be stable down to a pressure less than 20 GPa when anharmonicity is considered. The electron-phonon coupling is calculated for pressures above 20 GPa, and the superconducting transition temperature T-c is found to decrease with increasing pressure throughout this pressure range. The result is in agreement with experimental results above 30 GPa. In contrast to experiment, no evidence for a decrease in T-c with decreasing pressure below 30 GPa is found. The structural transition from rhombohedral A7 to sc is also investigated. An interesting two-step transition is found to occur theoretically which may have relevance for the pressure dependence of T-c. Possible explanations for the discrepancy with experiment are discussed.