ZFZhao Fan

Zhao Fan

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Zhao Fan has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Convolutional neural networks enable high-fidelity prediction of rotation-invariant properties of amorphous and crystalline materials

    by Zhao Fan

    Over the past two decades, various machine learning (ML) methods have been broadly used to predict properties from structure of materials, which is valuable in accelerating the discovery and development of new materials for different applications. Convolutional neural networks (CNN), however, have not been widely used in materials science, although they have been demonstrated to possess the state-of-the-art learning capability in the field of computer vision and pattern recognition. Recently, it was shown that CNN are powerful in predicting orientation-dependent properties when using a new structure representation---spatial density map (SDM). However, it is unclear whether the deep learning (DL) framework combining CNN and SDM is useful for predicting rotation-invariant properties because of the lack of rotational invariance of SDMs. Here, we demonstrate that this DL framework can also enable high-fidelity prediction of rotation-invariant properties through two examples, i.e., predicting the flexibility volume of metallic glasses and the grain-boundary segregation energy of polycrystals. This is realized through training CNN models via data augmentation and defining a common protocol to establish a unique local coordinate system to compute SDMs when applying CNN models. The advantages of the DL framework arise from the completeness of the SDM and the excellent learning capability of CNN models. Thanks to the high fidelity of CNN models, additional physical insights into structure--property relationships are obtained. The DL method is expected to be useful for a broad range of topics in materials science.

  • Open Access

    Efficient machine learning interatomic potentials robust for liquid and multiple solid polymorphs of NaF and KF

    by Zhao Fan, Michael L. Whittaker, Mark Asta

    Achieving atomic-level understanding of crystallization of molten salts is of importance to a wide range of technological applications. Recent work [Fan , ] revealed that crystal nucleation in molten LiF salt is a multistage process according to the molecular-dynamics (MD) simulations based on an atomic cluster expansion (ACE) machine-learning interatomic potential (MLIP). In order to understand the influence of increasing cation size on nucleation pathways and nucleation rates of molten fluoride salts, here we develop two new ACE MLIPs for NaF and KF. The two ACE MLIPs feature DFT-SCAN-level accuracy for liquid and multiple solid polymorphs over a wide temperature (0–2000 K) and pressure (0–100 GPa) range, and also reproduce well a number of experimental data for solid and liquid equilibrium properties. The efficiency of the two ACE MLIPs enable million-atom-scale or microsecond-scale MD simulations. The two general-purpose ACE MLIPs are expected to be useful for atomistic simulations for different purposes, in addition to studying crystallization of molten NaF and KF salts.


  • Open Access

    Multistage nucleation pathway in LiF molten salt mirrors the crystal–melt interface structure

    by Zhao Fan, Deepak Rawat, Piotr Zarzycki, Michael L. Whittaker, Mark Asta

    Despite over a century of studies, fundamental questions remain about the processes governing crystal nucleation from melts or solutions. Research over the past three decades has presented mounting evidence for kinetic pathways of crystal nucleation that are more complex than envisioned by the simplest forms of classical theory. Such observations have been presented for colloidal and elemental systems with covalent and metallic bonding. Despite the technological and geochemical importance of molten salts, similar studies for these ionically bonded systems are currently lacking. Here we develop a machine learning interatomic potential for a model ionic system: LiF. The potential features quantum-level accuracy for both liquid and multiple solid polymorphs over wide temperature and pressure ranges and accurately reproduces experimentally measured properties. Thanks to the efficiency of the potential, which enables microsecond-scale molecular dynamics simulations, induction times for nucleation of LiF solids from their melts are computed over a range of undercoolings. With the aid of a set of robust local order parameters established here, the simulations reveal that homogeneous crystal nucleation in undercooled melts preferentially initiates from liquid regions showing slow dynamics and high bond orientational order simultaneously, and the second-shell order of both precritical nuclei and the surface of postcritical nuclei is dominated by hexagonal close packing and body-centered cubic local structure, even though the nucleus core is dominated by face-centered cubic structure corresponding to the stable rocksalt crystal structure. Finally, we establish a connection between the crystallization pathway and the equilibrium crystal–melt interface structure.


  • Structural origin of the anisotropic local mechanical response of amorphous silicon

    by Zhao Fan

    Deciphering atomic packing around “defects” in glasses remains a grand challenge in the glass community. Amorphous silicon (a-Si), as a prototypical network-forming glass, often exhibits opposite structure-property relations compared to metallic glasses (MGs), and the classic concept of free volume widely used for MGs is also invalid to a-Si. In this work, we surprisingly discovered that a-Si and MGs share the same structural origin responsible for their orientation-dependent local mechanical response. In addition, the favorite motif in a-Si, i.e., tetrahedra, centered on the atoms with the lowest plastic susceptibility have two dominant preferred orientations. This is in contrast to the case in MGs. These results will enhance our understanding of structure-property relationships in glasses and also provide some optimism for establishing a universal picture regarding “defects” for the two classes of glasses.


  • Open Access

    Microscopic mechanisms of pressure-induced amorphous-amorphous transitions and crystallisation in silicon

    by Zhao Fan, Hajime Tanaka

    Some low-coordination materials, including water, silica, and silicon, exhibit polyamorphism, having multiple amorphous forms. However, the microscopic mechanism and kinetic pathway of amorphous-amorphous transition (AAT) remain largely unknown. Here, we use a state-of-the-art machine-learning potential and local structural analysis to investigate the microscopic kinetics of AAT in silicon after a rapid pressure change. We find that the transition from low-density-amorphous (LDA) to high-density-amorphous (HDA) occurs through nucleation and growth, resulting in non-spherical interfaces that underscore the mechanical nature of AAT. In contrast, the reverse transition occurs through spinodal decomposition. Further pressurisation transforms LDA into very-high-density amorphous (VHDA), with HDA serving as an intermediate state. Notably, the final amorphous states are inherently unstable, transitioning into crystals. Our findings demonstrate that AAT and crystallisation are driven by joint thermodynamic and mechanical instabilities, assisted by preordering, occurring without diffusion. This unique mechanical and diffusion-less nature distinguishes AAT from liquid-liquid transitions.


  • Predicting the location of shear band initiation in a metallic glass

    by Z. Fan, E. Ma, M. L. Falk

    We report atomistic simulation results which indicate that the location of shear banding in a metallic glass (MG) can be ascertained with reasonably high accuracy solely from the undeformed static structure. Correlation is observed between the location of the initiation of shear bands in a simulated MG and the initial distribution of the density of fertile sites (DFS) for stress-driven shear transformations identified a priori based on a deep learning model devised in our recent work [Fan and Ma, Nat. Commun. 12, 1506 (2021)]. In addition, we demonstrated that one can judge whether a glass is brittle or ductile solely based upon its initial DFS distribution. These validate that shear bands in MG arise from nonlinear instabilities and that the as-quenched glass structure contains inhomogeneities that influence these instabilities. In this paper, we also reveal an important subtlety regarding the nondeterministic nature of athermal quasistatic shear simulations.


  • Open Access

    Convolutional neural networks enable high-fidelity prediction of path-dependent diffusion barrier spectra in multi-principal element alloys

    by Zhao Fan, Bin Xing, Penghui Cao

    The emergent multi-principal element alloys (MPEAs) provide a vast compositional space to search for novel materials for technological advances. However, how to efficiently identify optimal compositions from such a large design space for targeted properties is a grand challenge in material science. Here we developed a convolutional neural network (CNN) model that can accurately and efficiently predict path-dependent vacancy migration energy barriers, which are critical to diffusion behaviors and many high-temperature properties, of MPEAs at any compositions and with different chemical short-range orders within a given alloy system. The success of the CNN model makes it promising for developing a database of diffusion barriers for different MPEA systems, which would accelerate alloy screening for the discovery of new compositions with desirable properties. Besides, the length scale of local configurations relevant to migration energy barriers is uncovered, and the implications of this success to other aspects of materials science are discussed.

  • Atomic vibration as an indicator of the propensity for configurational rearrangements in metallic glasses

    by Huiyang Fan, Zhao Fan, Xiongjun Liu, Zhaoping Lu, En Ma

    In a metallic glass (MG), the propensity for atomic rearrangements varies spatially from location to location in the amorphous solid, making the prediction of their likelihood a major challenge. One can attack this problem from the “structure controls properties” standpoint. But all the current structure-centric parameters are mostly based on local atomic packing information limited to short-range order, hence falling short in reliably forecasting how the local region would respond to external stimuli (e.g., temperature and/or stress). Alternatively, one can use indicators informed by physical properties to bridge the static structure on the one hand, and the response of the local configuration on the other. A sub-group of such physics-informed quantities consists of atomic vibration parameters, which will be singled out as the focus of this article. Here we use the Cu64Zr36 alloy to systematically demonstrate the following two points, all using a single model MG. First, we show in a comprehensive manner the interrelation among common vibrational parameters characterizing the atomic vibrational amplitude and frequency, including the atomic mean square displacement, flexibility volume, participation fraction in the low-frequency vibrational modes and boson peak intensity. Second, we demonstrate that these vibrational parameters fare much better than purely static structural parameters based on local geometrical packing in providing correlation with the propensity for local configurational transitions. These vibrational parameters also share a correlation length similar to that in structural rearrangements induced by external stimuli. This success, however, also poses a challenge, as it remains to be elucidated as to why short-time dynamical (vibrational) behavior at the bottom of the energy basin can be exploited to project the height of the energy barrier for cross-basin activities and in turn the propensity for locally collective atomic rearrangements.


  • Tension–compression asymmetry in amorphous silicon

    by Yuecun Wang, Jun Ding, Zhao Fan, Lin Tian, Meng Li, Huanhuan Lu, Yongqiang Zhang, En Ma, Ju Li, Zhiwei Shan

    Hard and brittle materials usually exhibit a much lower strength when loaded in tension than in compression. However, this common-sense behaviour may not be intrinsic to these materials, but arises from their higher flaw sensitivity to tensile loading. Here, we demonstrate a reversed and unusually pronounced tension–compression asymmetry (tensile strength exceeds compressive strength by a large margin) in submicrometre-sized samples of isotropic amorphous silicon. The abnormal asymmetry in the yield strength and anelasticity originates from the reduction in shear modulus and the densification of the shear-activated configuration under compression, altering the magnitude of the activation energy barrier for elementary shear events in amorphous Si. In situ coupled electrical tests corroborate that compressive strains indeed cause increased atomic coordination (metallization) by transforming some local structures from sp3-bonded semiconducting motifs to more metallic-like sites, lending credence to the mechanism we propose. This finding opens up an unexplored regime of intrinsic tension–compression asymmetry in materials.