ænet-PyTorch: A GPU-supported implementation for machine learning atomic potentials training.

In this work, we present ænet-PyTorch, a PyTorch-based implementation for training artificial neural network-based machine learning interatomic potentials. Developed as an extension of the atomic energy network (ænet), ænet-PyTorch provides access to all the tools included in ænet for the application and usage of the potentials. The package has been designed as an alternative to the internal training capabilities of ænet, leveraging the power of graphic processing units to facilitate direct training on forces in addition to energies. This leads to a substantial reduction of the training time by one to two orders of magnitude compared to the central processing unit implementation, enabling direct training on forces for systems beyond small molecules. Here, we demonstrate the main features of ænet-PyTorch and show its performance on open databases. Our results show that training on all the force information within a dataset is not necessary, and including between 10% and 20% of the force information is sufficient to achieve optimally accurate interatomic potentials with the least computational resources.

Study of a charge transition-driven resistive switching mechanism in TiO2-based random access memory via density functional theory.

The nature of the conducting filament (CF) with a high concentration of oxygen vacancies (VOs) in oxide thin film-based resistive random access memory (RRAM) remains unclear. The VOs in the CF have been assumed to be positively charged (VO2+) to explain the field-driven switching of RRAM, but VO2+ clusters in high concentration encounter Coulomb repulsion, rendering the CF unstable. Therefore, this study examined the oxidation state of VOs in the CF and their effects on the switching behavior via density functional theory calculations using a Pt/TiO2/Ti model system. It was concluded that the VOs in the CF are in a low oxidation state but are transformed to VO2+ immediately after release from the CF. In addition, the short-range interactions between VOs were confirmed to facilitate the rupture and rejuvenation of the CF by reducing the required activation energy. Finally, an improved switching model was proposed by considering the charge transition of VOs, providing a plausible explanation for the reported coexistence of two opposite bipolar switching polarities: the eight-wise and the counter-eight-wise polarities.

An ab initio approach on the asymmetric stacking of GaAs 〈111〉 nanowires grown by a vapor-solid method.

This study provides an ab initio thermodynamics approach to take a step forward in the theoretical modeling on the growth of GaAs nanowires. In order to understand the effects of growth conditions on the involvement of stacking faults and polytypism, we investigated the vapor-phase growth kinetics under arbitrary temperature-pressure conditions by combining the atomic-scale calculation with the thermodynamic treatment of a vapor-solid system. Considering entropy contribution and electronic energy, the chemical potential and surface energies of various reconstructions were calculated as a function of temperature and pressure, leading to the prediction of the change in Gibbs free energy at each stage of nucleation and growth. This enabled us to predict the temperature-pressure-dependent variation in nucleation rate and formation probability of possible stacking sequences: zinc blende, stacking faults, twin, and wurtzite. As a result, the formation probabilities of stacking faults and polytypism were found to decrease with increasing temperature or decreasing pressure, which agreed well with available experiments. In addition, by showing that the formation probability of the stacking defects in GaAs nanowires grown along the 〈111〉B direction is about ten times higher than that along the 〈111〉A direction, the intriguing asymmetric stacking behavior during the growth along the polar direction and its dependence on growth conditions were fundamentally elucidated. The proposed ab initio approach bridges the gap between atomic-scale static calculation at zero-temperature and kinetic growth process under arbitrary vapor-phase conditions, and thus will contribute to the nanoscale growth not only for GaAs nanowires but also for other materials.

Equilibrium crystal shape of GaAs and InAs considering surface vibration and new (111)B reconstruction: ab-initio thermodynamics.

This work reports on the theoretical equilibrium crystal shapes of GaAs and InAs as a function of temperature and pressure, taking into account the contribution of the surface vibration, using ab-initio thermodynamic calculations. For this purpose, new (111)B reconstructions, which are energetically stable at a high temperature, are suggested. It was found that there was a feasible correspondence between the calculated equilibrium shapes and the experimental shapes, which implied that the previous experimental growth was performed under conditions that were close to equilibrium. In this study, GaAs and InAs were selected as prototype compound semiconductors, but the developed calculation methodology can also be applied to other III-V compound semiconductor materials.

Ferroelectric switching in bilayer 3R MoS2 via interlayer shear mode driven by nonlinear phononics.

We theoretically investigate the mechanism of ferroelectric switching via interlayer shear in 3R MoS2 using first principles and lattice dynamics calculations. First principle calculations show the prominent anharmonic coupling of the infrared inactive interlayer shear and the infrared active phonons. The nonlinear coupling terms generates an effective anharmonic force which drives the interlayer shear mode and lowers the ferroelectric switching barrier depending on the amplitude and polarization of infrared mode. Lattice dynamics simulations show that the interlayer shear mode can be coherently excited to the switching threshold by a train of infrared pulses polarized along the zigzag axis of MoS2. The results of this study indicate the possibility of ultrafast ferroelectricity in stacked two-dimensional materials from the control of stacking sequence.

Optical control of the layer degree of freedom through Wannier-Stark states in polar 3R MoS2.

Electrons in two-dimensional layered crystals gain a discrete positional degree of freedom over layers. We propose the two-dimensional transition metal dichalcogenide homostructure with polar symmetry as a prototypical platform where the degrees of freedom for the layers and valleys can be independently controlled through an optical method. In 3R MoS2, a model system, the presence of the spontaneous polarization and built-in electric field along the stacking axis is theoretically proven by the density functional theory. The K valley states under the electric field exhibit Wannier-Stark type localization with atomic-scale confinement driven by double group symmetry. The simple interlayer-dynamics-selection rule of the valley carriers in 3R homostructure enables a binary operation, upward or downward motion, using visible and infrared light sources. Together with the valley-index, a 2 [Formula: see text] 2 states/cell device using a dual-frequency polarized light source is suggested.

Surface reconstruction of InAs (001) depending on the pressure and temperature examined by density functional thermodynamics.

A detailed understanding of the atomic configuration of the compound semiconductor surface, especially after reconstruction, is very important for the device fabrication and performance. While there have been numerous experimental studies using the scanning probe techniques, further theoretical studies on surface reconstruction are necessary to promote the clear understanding of the origins and development of such subtle surface structures. In this work, therefore, a pressure-temperature surface reconstruction diagram was constructed for the model case of the InAs (001) surface considering both the vibrational entropy and configurational entropy based on the density functional theory. Notably, the equilibrium fraction of various reconstructions was determined as a function of the pressure and temperature, not as a function of the chemical potential, which largely facilitated the direct comparison with the experiments. By taking into account the entropy effects, the coexistence of the multiple reconstructions and the fractional change of each reconstruction by the thermodynamic condition were predicted and were in agreement with the previous experimental observations. This work provides the community with a useful framework for such type of theoretical studies.

Resistance switching behavior of atomic layer deposited SrTiO3 film through possible formation of Sr2Ti6O13 or Sr1Ti11O20 phases.

Identification of microstructural evolution of nanoscale conducting phase, such as conducting filament (CF), in many resistance switching (RS) devices is a crucial factor to unambiguously understand the electrical behaviours of the RS-based electronic devices. Among the diverse RS material systems, oxide-based redox system comprises the major category of these intriguing electronic devices, where the local, along both lateral and vertical directions of thin films, changes in oxygen chemistry has been suggested to be the main RS mechanism. However, there are systems which involve distinctive crystallographic phases as CF; the Magnéli phase in TiO2 is one of the very well-known examples. The current research reports the possible presence of distinctive local conducting phase in atomic layer deposited SrTiO3 RS thin film. The conducting phase was identified through extensive transmission electron microscopy studies, which indicated that oxygen-deficient Sr2Ti6O13 or Sr1Ti11O20 phase was presumably present mainly along the grain boundaries of SrTiO3 after the unipolar set switching in Pt/TiN/SrTiO3/Pt structure. A detailed electrical characterization revealed that the samples showed typical bipolar and complementary RS after the memory cell was unipolar reset.