Optimization and Validation of a Deep Learning CuZr Atomistic Potential: Robust Applications for Crystalline and Amorphous Phases with near-DFT Accuracy
Christopher M. Andolina, Philip Williamson, Wissam A. Saidi
We show that a deep-learning neural network potential (DP) based on density functional theory (DFT) calculations can well describe Cu-Zr materials, an example of a binary alloy system that can coexist in several ordered intermetallics and as an amorphous phase. The complex phase diagram for Cu-Zr makes it a challenging system for traditional atomistic force-fields that fail to describe well the different properties and phases. Instead, we show that a DP approach using a large database with ~300k configurations can render results generally on par with DFT. The training set includes configurations of pristine and bulk elementary metals and intermetallics in the liquid and solid phases in addition to slab and amorphous configurations. The DP model was validated by comparing bulk properties such as lattice constants, elastic constants, bulk moduli, phonon spectra, surface energies to DFT values for identical structures. Further, we contrast the DP results with values obtained using well-established two embedded atom method potentials. Overall, our DP potential provides near DFT accuracy for the different Cu-Zr phases but with a fraction of its computational cost, thus enabling accurate computations of realistic atomistic models especially for the amorphous phase.
In this work we use molecular dynamics simulations to study the stress overshoots of metallic glass Cu$_{50}$Zr$_{50}$ in three scenarios (unloading-reloading, slide-stop-slide, and stress serrations) that are associated with shear band relaxation. We found that, after the elastic recovery effect is factored out, atomic volume in the shear band barely changes during compressive relaxation but decreases during tensile relaxation, while local fivefold symmetry increases consistently for both cases. We propose that the atomistic mechanism for the related stress overshoots is due to the relaxation…
An approach is presented for the atomistic study of phonon transport in real dielectric nanowires via Green functions. The formalism is applied to investigate the phonon flow through nanowires coated by an amorphous material. Examples for a simple model system, and for real Si nanowires coated by silica are given. New physical results emerge for these systems, regarding the character of the transition from ballistic to diffusive transport, the low temperature thermal conductance, and the influence of the wire-coating interface on the thermal transport. An efficient treatment of phonon scatteri…
We have analyzed the atomic rearrangements underlying self-diffusion in amorphous Si during annealing using tight-binding molecular dynamics simulations. Two types of amorphous samples with different structural features were used to analyze the influence of coordination defects. We have identified several types of atomic rearrangement mechanisms, and we have obtained an effective migration energy of around 1 eV. We found similar migration energies for both types of samples, but higher diffusivities in the one with a higher initial percentage of coordination defects.
We employ ab initio molecular dynamics to simulate the response of hydrogenated amorphous silicon to light exposure (Staebler-Wronski effect). We obtain improved microscopic understanding of PV operation, compute the motion of H atoms, and modes of light-induced degradation of photovoltaics. We clarify existing models of light-induced change in aSi:H and show that the Hydrogen collision model of Branz3 is correct in essentials.
The atomic structure of amorphous materials is believed to be well described by the continuous random network model. We present an algorithm for the generation of large, high-quality continuous random networks. The algorithm is a variation of the "sillium" approach introduced by Wooten, Winer, and Weaire. By employing local relaxation techniques, local atomic rearrangements can be tried that scale almost independently of system size. This scaling property of the algorithm paves the way for the generation of realistic device-size atomic networks.
There is an increasing interest in the amorphous states of metal-organic frameworks (MOFs) and porous coordination polymers, which can be produced by pressure-induced amorphization, temperature-induced amorphization, melt-quenching, ball-milling, irradiation, etc. They can exhibit useful physical and chemical properties, distinct from those achievable in the crystalline states, along with greater ease of processing, and intrinsic advantages over crystals and powders, such as high transparency and mechanical robustness. However, these amorphous states are particularly challenging to characteriz…
One of the major theoretical riddles in shear banding instabilities is the angle that the shear band chooses spontaneously with respect to the principal stress axis. Here we employ our recent atomistic theory to compute analytically the angle in terms of the characteristics of the Eshelby inclusion that models faithfully the eigenfunction of the Hessian matrix that goes soft at the plastic instability. We show that loading protocols that do not conserve volume result in shear bands at angles different from 45$^o$ to the strain axis; only when the external strains preserve volume like in pure s…
Amorphous silicon (a-Si) is a widely studied non-crystalline material, and yet the subtle details of its atomistic structure are still unclear. Here, we show that accurate structural models of a-Si can be obtained by harnessing the power of machine-learning algorithms to create interatomic potentials. Our best a-Si network is obtained by cooling from the melt in molecular-dynamics simulations, at a rate of 10$^{11}$ K/s (that is, on the 10 ns timescale). This structure shows a defect concentration of below 2% and agrees with experiments regarding excess energies, diffraction data, as well as $…
Germanium is the base element in many phase-change materials, i.e. systems that can undergo reversible transformations between their crystalline and amorphous phases. They are widely used in current digital electronics and hold great promise for the next generation of non-volatile memory devices. However, the ultra fast phase transformations required for these applications can be exceedingly complex even for single component systems, and a full physical understanding of these phenomena is still lacking. In this paper we study nucleation and growth of crystalline Ge from amorphous thin films at…
Molecular dynamics simulations are performed to provide a detailed understanding of the functional degradation of shape memory alloys at small scale. The origin of the experimentally reported accumulation of plastic deformation and the anomalous sudden increase of the residual strain under cyclic mechanical loading are explained by detailed insights into the relevant atomic scale processes. Our work reveals that the mechanical response of shape-memory-alloy pillars under cyclic compression is significantly influenced by the presence of an amorphous-like surface region as experimentally induced…
Experiment-driven atomistic materials modeling: A case study combining X-ray photoelectron spectroscopy and machine learning potentials to infer the structure of oxygen-rich amorphous carbon
The annealing treatment in the advanced manufacturing process, e.g., laser-assisted manufacturing, determines the final state of glasses which is critical to its thermal, electrical and mechanical properties. Energy barriers analysis based on the potential energy surface offers an effective way to study the microscopic evolution of the inherent structures during the annealing process in a broadening timescale range, i.e., from atomistic timescale ( ~ ps) to experimental timescale (~ s). Here, we find the distribution activation energy barriers in the potential energy surface can be divided int…
Atomistic Insights into Cu/amorphous-Ta$_x$N Interfacial Adhesion via Machine Learning Interatomic Potentials: Effects of Stoichiometry and Interface Construction
Exponential band edges have been observed in a variety of materials, both crystalline and amorphous. In this paper, we infer the structural origins of these tails in amorphous and defective crystalline Si by direct calculation with current {\it ab initio} methods. We find that exponential tails appear in relaxed models of diamond with suitable point defects. In amorphous silicon (a-Si), we find that structural filaments of short bonds and long bonds exist in the network, and that the tail states near the extreme edges of both band tails are are also filamentary, with much localization on the s…
We present an information-based total-energy optimization method to produce nearly defect-free structural models of amorphous silicon. Using geometrical, structural and topological information from disordered tetrahedral networks, we have shown that it is possible to generate structural configurations of amorphous silicon, which are superior than the models obtained from conventional reverse Monte Carlo and molecular-dynamics simulations. The new data-driven hybrid approach presented here is capable of producing atomistic models with structural and electronic properties which are on a par with…
Molecular dynamics (MD) simulations enable the description of ma- terial properties and processes with atomistic detail by numerically solv- ing the time evolution of every atom in the system. We introduce Poly- merModeler, a general-purpose, online simulation tool to build atomistic structures of amorphous polymers and perform MD simulations on the re- sulting configurations to predict their thermo-mechanical properties. Poly- merModeler is available, free of charge, at nanoHUB.org via an interactive web interface, and the actual simulations are performed in the cloud us- ing nanoHUB.org reso…
We study the statistical properties of the yielding transition in model amorphous solids in the limit of slow, athermal deformation. Plastic flow occurs via alternating phases of elastic loading punctuated by rapid dissipative events in the form of collective avalanches. We investigate their characterization through energy vs. stress drops and at multiple stages of deformation, thus revealing a change of spatial extent of the avalanches and degree of stress correlations as deformation progresses. We show that the statistics of stress and energy drops only become comparable for large events in …
The mechanism determining the band alignment of the amorphous/crystalline Si heterostructures is addressed with direct atomistic simulations of the interface performed using a hierarchical combination of various computational schemes ranging from classical model-potential molecular dynamics to ab-initio methods. We found that in coordination defect-free samples the band alignment is almost vanishing and independent on interface details. In defect-rich samples, instead, the band alignment is sizeably different with respect to the defect-free case, but, remarkably, almost independent on the conc…
Localized basis ab initio molecular dynamics simulation within the density functional framework has been used to generate realistic configurations of amorphous silicon carbide (a-SiC). Our approach consists of constructing a set of smart initial configurations that conform essential geometrical and structural aspects of the materials obtained from experimental data, which is subsequently driven via first-principles force-field to obtain the best solution in a reduced solution space. A combination of a priori information (primarily structural and topological) along with the ab-initio optimizati…
Combined static and molecular dynamics first-principles calculations are used to identify a direct structural link between the metastable crystalline and amorphous phases of Ge2Sb2Te5. We find that the phase transition is driven by the displacement of Ge atoms along the rocksalt [111] direction from the stable-octahedron to high-energy-unstable tetrahedron sites close to the intrinsic vacancy regions, which give rise to the formation of local 4-fold coordinated motifs. Our analyses suggest that the high figures of merit of Ge2Sb2Te5 are achieved from the optimal combination of intrinsic vacanc…
We discuss an inverse approach for atomistic modeling of glassy materials. The focus is on structural modeling and electronic properties of hydrogenated amorphous silicon and glassy GeSe2 alloy. The work is based upon a new approach "experimentally constrained molecular relaxation (ECMR)". Unlike conventional approaches (such as molecular dynamics (MD) and Monte Carlo simulations(MC), where a potential function is specified and the system evolves either deterministically (MD) or stochastically (MC), we develop a novel scheme to model structural configurations using experimental data in associa…
We introduce a Gaussian approximation potential (GAP) for atomistic simulations of liquid and amorphous elemental carbon. Based on a machine-learning representation of the density-functional theory (DFT) potential-energy surface, such interatomic potentials enable materials simulations with close-to DFT accuracy but at much lower computational cost. We first determine the maximum accuracy that any finite-range potential can achieve in carbon structures; then, using a novel hierarchical set of two-, three-, and many-body structural descriptors, we construct a GAP model that can indeed reach the…
The size-dependent phase stability of gamma-Al2O3 was studied by large-scale molecular dynamics simulations over a wide temperature range from 300 to 900 K. For the gamma-Al2O3 crystal, a bulk transformation to alpha-Al2O3 by an FCC-to-HCP transition of the O sublattice is still kinetically hindered at 900 K. However, local distortions of the FCC O-sublattice by the formation of quasi-octahedral Al local coordination spheres become thermally activated, as driven by the partial covalency of the Al-O bond. On the contrary, spherical gamma-Al2O3 NPs (with sizes of 6 and 10 nm) undergo a crystalli…
We propose here a method to generate random networked amorphous structure using only readily available short-range properties like bond lengths, bond angles and connectivity of the constituents. This method is a variant of Monte-Carlo (MC) method wherein the basic constituents of an amorphous network i.e. rigid polyhedral units are connected randomly obeying certain steric constraints. The algorithm is designed to reproduce the medium-range order universally observed in glasses. The method somewhat resembles the reverse MC (RMC) method where a random move of an atom inside a box is accepted or…
Germanium is an extremely important material used for numerous functional applications in many fields of nanotechnology. In this paper, we study the crystallization of amorphous Ge using atomistic simulations of critical nano-metric nuclei at high temperatures. We find that crystallization occurs by the recurrent transfer of atoms via a diffusive process from the amorphous phase into suitably-oriented crystalline layers. We accompany our simulations with a comprehensive thermodynamic and kinetic analysis of the growth process, which explains the energy balance and the interfacial growth veloci…
We show that ZIF-4, a metal-organic framework (MOF) with a zeolitic structure, undergoes a crystal--amorphous transition on heating to 300 $^\circ$C. The amorphous form, which we term a-ZIF, is recoverable to ambient conditions or may be converted to a dense crystalline phase of the same composition by heating to 400 $^\circ$C. Neutron and X-ray total scattering data collected during the amorphization process are used as a basis for reverse Monte Carlo refinement of an atomistic model of the structure of a-ZIF. We show that the structure is best understood in terms of a continuous random netwo…
The atomistic mechanisms during lithiation and delithiation of amorphous Si nanowires ($a$-SiNW) have been investigated over cycles by molecular dynamics simulations. First, the Modified Embedded Atom Method (MEAM) potential from Cui et al. [J. Power Sources. 2012, (207) 150] has been further optimized on static (Li$_x$Si alloy phases and point defect energies) and on dynamic properties (Li diffusion) to reproduce the lithiation of small crystalline Si nanowires calculated at the {\it ab initio} level. The lithiation of $a$-SiNW reveals a two-phase process of lithiation with a larger diffusion…
Amorphous calcium carbonate (ACC) is an important precursor for biomineralisation in marine organisms. Among the key outstanding problems regarding ACC are how best to understand its structure and how to rationalise its metastability as an amorphous phase. Here, we report high-quality atomistic models of ACC generated by using state-of-the-art interatomic potentials to help guide fits to X-ray total scattering data. Exploiting a recently-developed inversion approach, we extract from these models the effective Ca$\boldsymbol\cdots$Ca interaction potential governing ACC formation. This potential…
Chemical segregation and structural transitions at interfaces are important nanoscale phenomena, making them natural targets for atomistic modeling, yet interatomic potentials must be fit to secondary physical properties. To isolate the important factors that interatomic potentials must capture in order to accurately model such behavior, the performance of four interatomic potentials was evaluated for the Cu-Zr system, with experimental observations used to provide validation. While experimental results show strong Zr segregation to grain boundary regions and the formation of nanoscale amorpho…
The band offsets between crystalline and hydrogenated amorphous silicon (a-Si:H) are key parameters governing the charge transport in modern silicon hetrojunction solar cells. They are an important input for macroscopic simulators that are used to further optimize the solar cell. Past experimental studies, using X-ray photoelectron spectroscopy (XPS) and capacitance-voltage measurements, have yielded conflicting results on the band offset. Here we present a computational study on the band offsets. It is based on atomistic models and density-functional theory (DFT). The amorphous part of the in…
Atomistic insights into the structural, thermal, and mechanical evolution of $Zr_{47.5}Cu_{47.5}Ag_{5}$ bulk metallic glass
Instituto Sabato, UNSAM, CNEA（萨巴托研究所，阿根廷国家原子能委员会）；Departamento de Materia Condensada, CNEA（凝聚态物质部门，阿根廷国家原子能委员会）；Department of Mathematical Sciences, UTEP（数学科学系，埃尔帕索大学）
1. Introduction 2. Molecular dynamics simulation of amorphous carbon structures 3. Atomistic simulation of the bombardment process during the BEN phase of diamond CVD 4. Growth of amorphous silicon 5. One-dimensional hopping in disordered organic solids
Chalcogenide phase-change materials (PCMs) show a significant contrast in optical reflectivity and electrical resistivity upon crystallization from the amorphous phase and are leading candidates for non-volatile photonic and electronic applications. In addition to the flagship Ge2Sb2Te5 phase-change alloy, doped Sb2Te alloys, in particular AgInSbTe used in rewritable optical discs, have been widely investigated for decades, and nevertheless the theoretical insights on the optical properties of this important family of PCMs are scarce. Here, we carry out thorough ab initio simulations to gain a…
Machine-learned potential for amorphous Indium-Tin-Oxide alloys
Center for Alloy Innovation and Design (CAID), State Key Laboratory for Mechanical Behavior of Materials, Xi’an Jiaotong University（西安交通大学合金创新与设计中心、材料力学行为国家重点实验室）
Structurally disordered materials continue to pose fundamental questions, including that of how different disordered phases ("polyamorphs") can coexist and transform from one to another. As a widely studied case, amorphous silicon (a-Si) forms a fourfold-coordinated, covalent random network at ambient conditions, but much higher-coordinated, metallic-like phases under pressure. However, a detailed mechanistic understanding of the liquid-amorphous and amorphous-amorphous transitions in silicon has been lacking, due to intrinsic limitations of even the most advanced experimental and computationa…
State Key Laboratory of Solid Lubrication, Lanzhou Institute of Chemical Physics, Chinese Academy of Sciences（中国科学院兰州化学物理研究所固体润滑国家重点实验室）；Department of Engineering Mechanics, School of Civil Engineering, Wuhan University（武汉大学土木工程学院工程力学系）；State Key Laboratory of Water Resources Engineering and Management, Wuhan University（武汉大学水资源与水电工程科学国家重点实验室）；Center of Materials Science and Optoelectronics Engineering, University of Chinese Academy of Sciences（中国科学院大学材料科学与光电技术研究中心）
The precise nature of chemical-bonding interactions in amorphous, and crystalline, chalcogenides is still unclear due to the complexity arising from the delocalization of bonding, and non-bonding, electrons. Although an increasing degree of electron delocalization for elements down a column of the periodic table is widely recognized, its influence on chemical-bonding interactions, and on consequent material properties, of chalcogenides has not previously been comprehensively understood from an atomistic point of view. Here, we provide a chemical-bonding framework for understanding the behaviou…
First-principles based modeling on phonon dynamics and transport using density functional theory and Boltzmann transport equation has proven powerful in predicting thermal conductivity of crystalline materials, but it remains unfeasible for modeling complex crystals and disordered solids due to the prohibitive computational cost to capture the disordered structure, especially when the quasiparticle "phonon" model breaks down. Recently, machine-learning regression algorithms show great promises for building high-accuracy potential fields for atomistic modeling with length and time scales far be…
Understanding various macroscopic pressure-volume-temperature properties of materials on the atomistic level has always been an ambition for physicists and material scientists. Particularly, some materials such as zirconium tungstate (ZrW2O8), exhibit multiple exotic properties including negative thermal expansion (NTE) and pressure-induced amorphization (PIA). Here, using machine-learning based deep potential, we trace both of the phenomena in ZrW2O8 back to a common atomistic origin, where the nonbridging O atoms play a critical role. We demonstrate that the nonbridging O atoms confer great …
We use atomistic simulations to study the resonant acoustic modes and compare different calculations of the acoustic mean-free path in amorphous systems with nanometric crystalline spherical inclusions. We show that the resonant acoustic properties are not a simple combination of the vibrations in the inclusions and in the amorphous matrix. The presence of the inclusion affects the transport properties mainly in the frequency range separating simple scattering from multiple scattering processes. However, propagation of acoustic wavepackets is spatially heterogeneous and shows that the amorphou…
Here, an approach to quantify the amorphous-to-disordered/crystalline transformation occurring in NaOH-activated ground granulated blast-furnace slag (GGBS) is outlined that combines atomistic modeling with in situ pair distribution function (PDF) analysis. Firstly, by using force-field molecular dynamics (MD) simulations, a detailed structural representation is generated for the amorphous GGBS that is in agreement with experimental X-ray scattering data. Use of this structural representation along with literature-derived structures for the reaction products allows for real space X-ray PDF ref…
