Atomic Size Research Articles

Exploring thermodynamic, physical and radiative interaction properties of quinary FeNiCoCr high entropy alloys (HEAs): a multi-directional characterization study

To qualify for nuclear applications, materials must meet specific criteria, including mechanical properties, high-temperature behavior, corrosion resistance, and high-temperature oxidation resistance. High Entropy Alloys (HEAs) are particularly suitable for these applications due to their unique properties. Consequently, we conducted a theoretical and simulation-based approaches to assess some critical properties including radiation shielding properties of some quinary FeNiCoCr HEAs. In this study, we focused on quinary FeNiCoCr HEAs, whose corrosion properties have been previously examined in the literature. We investigated the thermodynamic and radiation shielding properties of HEAs with sixteen different compositions. Our methodology included evaluating thermodynamic parameters such as Mixing Entropy (∆Smix) and Mixing Enthalpy (∆Hmix), as well as structural characteristics like Valence Electron Concentration (VEC) and Atomic Size Difference (δ). This allowed us to systematically deduce the phase behavior and stability of various HEAs. Through computational modeling, we assessed the radiation shielding capabilities of these alloys, particularly their effectiveness in attenuating gamma ray and fast neutrons. The results identified FeNiCoCrW as the alloy with the lowest fast neutron removal cross-section values, highlighting its potential for nuclear applications. Its high melting point and the synergistic interplay between its elemental composition and thermodynamic properties suggest broad applicability in extreme environments. Thus, FeNiCoCrW emerges as a promising HEA with multifunctional capabilities, warranting further exploration and potential integration into advanced engineering solutions.

Revisiting Stability Criteria in Ball‐Milled High‐Entropy Alloys: Do Hume–Rothery and Thermodynamic Rules Equally Apply?

Stability descriptors for the formation of solid solutions can be divided into two categories: inspired by Hume–Rothery rules (HRR) and derived from thermodynamic approaches. Herein, HRRs are extended from binary to high‐entropy alloys (HEAs) focusing on compositions prepared by ball milling. Parameters describing stability criteria are interrelated and implicitly account for the microstrains’ storage energy, more determinant than entropy increase in stabilization of HEAs and more effective in bcc structures than close‐packed ones (fcc and hcp). An effective temperature, Teff, is defined as the ratio between increase in metallic bonding energy of solid solutions with respect to segregated pure constituents and configurational entropy. This versatile parameter is used as a threshold for stabilization of HEAs at equilibrium and out of equilibrium. When Teff is below room temperature, HEA would be stable at equilibrium. When Teff is below melting temperature, HEA would be obtained by rapid quenching. Limitations related to electronegativity differences remain valid in mechanically alloyed solid solutions. However, ball milling broadens the allowed differences in atomic size to form HEA. Moreover, thermodynamic criteria can be surpassed in these systems, allowing the formation of single‐phase solid solutions beyond the compositional range predicted by those criteria.

Machine Learning‐Aided Discovery of Low‐Pt High Entropy Intermetallic Compounds for Electrochemical Oxygen Reduction Reaction

Advancing the design of cathode catalysts to significantly maximize platinum utilization and augment the longevity has emerged as a formidable challenge in the field of fuel cells. Herein, we rationally design a high entropy intermetallic compound (HEIC, Pt(FeCoNiCu)3) for catalyzing oxygen reduction reaction (ORR) by an efficient machine learning stategy, where crystal graph convolutional neural networks are employed to expedite the multicomponent design. Based on a dataset generated from first‐principles calculations, the model can achieve a high prediction accuracy with mean absolute errors of 0.003 for surface strain and 0.011 eV atom‐1 for formation energy. In addition, we identify two chemical features (atomic size difference and mixing enthalpy) as new descriptors to explore advanced ORR catalysts. The carbon supported Pt(FeCoNiCu)3 catalyst with small particle size is successfully synthesized by a freeze‐drying‐annealing technology, and exhibits ultrahigh mass activity (4.09 A mgPt‐1) and specific activity (7.92 mA cm‐2). Meanwhile, The catalyst also shows significantly enhanced electrochemical stability which can be ascribed to the sluggish difussion effect in the HEIC structure. Beyond offering a promising low‐Pt electrocatalysts for fuel cell cathode, this work offers a new paradigm to rationally design advanced catalysts for energy storage and conversion devices.

Machine Learning-Aided Discovery of Low-Pt High Entropy Intermetallic Compounds for Electrochemical Oxygen Reduction Reaction.

Advancing the design of cathode catalysts to significantly maximize platinum utilization and augment the longevity has emerged as a formidable challenge in the field of fuel cells. Herein, we rationally design a high entropy intermetallic compound (HEIC, Pt(FeCoNiCu)3) for catalyzing oxygen reduction reaction (ORR) by an efficient machine learning stategy, where crystal graph convolutional neural networks are employed to expedite the multicomponent design. Based on a dataset generated from first-principles calculations, the model can achieve a high prediction accuracy with mean absolute errors of 0.003 for surface strain and 0.011 eV atom-1 for formation energy. In addition, we identify two chemical features (atomic size difference and mixing enthalpy) as new descriptors to explore advanced ORR catalysts. The carbon supported Pt(FeCoNiCu)3 catalyst with small particle size is successfully synthesized by a freeze-drying-annealing technology, and exhibits ultrahigh mass activity (4.09 A mgPt-1) and specific activity (7.92 mA cm-2). Meanwhile, The catalyst also shows significantly enhanced electrochemical stability which can be ascribed to the sluggish difussion effect in the HEIC structure. Beyond offering a promising low-Pt electrocatalysts for fuel cell cathode, this work offers a new paradigm to rationally design advanced catalysts for energy storage and conversion devices.

First-Principles Calculations and Machine Learning of Hydrogen Evolution Reaction Activity of Nonmetallic Doped β-Mo2C Support Pt Single-Atom Catalysts.

The most widely used catalyst for the hydrogen evolution reaction (HER) is Pt, but the high cost and low abundance of Pt need to be urgently addressed. Single-atom catalysts (SACs) have been an effective means of improving the utilization of Pt atoms. In this work, we used a nonmetal (NM = B, N, O, F, Si, P, S, Cl, As, Se, Br, Te, and I) doped β-Mo2C (100) C-termination surface as the support, with Pt atoms dispersed on the support surface to construct Pt@NM-Mo2C. Using density functional theory (DFT) calculations, we selected catalysts with excellent HER activity. Among 117 candidate catalysts, 49 catalysts exhibited ideal catalytic performance with Gibbs free energy of hydrogen intermediate (H*) adsorption (ΔGH*) values less than 0.2 eV. The ΔGH* values of 16 catalysts were even lower than that of Pt (ΔGH* ≈ 0.09 eV), with PtI@N2/4-a-Mo2C demonstrating the best performance (ΔGH* = -0.01 eV). Combined with electronic structure analysis, we could understand the impact of charge transfer between Pt and the underlying NM atoms on the strength of the Pt-H bond, thereby promoting HER activity. Using machine learning (ML), we identified that the primary influencing factors of the HER catalytic activity in the Pt@NM-Mo2C system were the Bader charge transfer of Pt (NePt), the d-band center of Pt (εdPt), and the atomic radius of NM (RNM), with NePt having the greatest impact on the HER catalytic activity.

Enhancing mechanical properties in Ti-Containing FeMn40Co10Cr10C0.5 High-Entropy alloy through Chi (χ) phase dissolution and precipitation hardening

This research investigates the effects of adding 2 at.% Ti on the microstructure, phase transformation, and mechanical properties of FeMn40Co10Cr10C0.5 high-entropy alloy (HEA) in both as-cast and homogenized conditions. The incorporation of Ti increased the lattice parameter of the FCC matrix and caused greater lattice distortion (∼5.9 %) due to the larger atomic radius of Ti. The negative mixing enthalpy of Ti resulted in segregation at interdendritic regions, forming a Ti-rich Chi (χ) phase during solidification. The as-cast structure displayed a dendritic microstructure consisting of 77.9 % FCC, 20.6 % χ, and 1.5 % TiC phases. After homogenization at 1200 °C for 2 h, the χ phase was dissolved, increasing the volume fraction of TiC to 24 %. The homogenized sample exhibited a substantial improvement in tensile properties, i.e., with the ultimate tensile strength (UTS) rising from 631 MPa in the as-cast condition to 689 MPa, and yield strength (YS) increasing from 346 MPa to 419 MPa, while maintaining an elongation of ∼ 54 %. These enhancements are attributed to solid solution strengthening, increased carbide precipitation, and the dissolution of the χ phase into the FCC matrix.

Prediction of the optimal hydrogen storage in high entropy alloys

Hydrogen is the most abundant element in nature, characterized by its vast resources, recyclability, and environmentally friendly. Also, it can exist in various forms, such as gas, liquid, or solid, to meet different storage needs. In solid-state storage, the hydrogen storage alloys have received widespread attention due to hydrogen is stored in the form of hydrides. In recent years, high entropy alloys (HEAs), as a new type of novel alloy design concept have attracted a lot of attention in many fields including hydrogen storage due to their unique structures and excellent properties. Moreover, the freedom of element selection leads to various types of HEAs, further enhancing the potential for their application in hydrogen storage. This article focuses on the hydrogen storage performance of HEAs and provides a comprehensive overview of the hydrogen storage mechanism, thermodynamics, and kinetics. Additionally, we summarize and compare the hydrogen storage capacities of previously published HEAs and introduce two empirical parameters, valence electron concentration (VEC) and atomic size difference (δ), and their relationship with hydrogen storage capacity. By applying the K-means clustering algorithm, we conclude that the optimal hydrogen storage capacity of HEAs occurs when 4 <VEC < 6 and 5.3 % <δ < 7.5 %. This finding offers valuable guidance for the future design of efficient hydrogen storage HEAs.

Emerging ZnO Semiconductors for Photocatalytic CO2 Reduction to Methanol.

Carbon recycling is poised to emerge as a prominent trend for mitigating severe climate change and meeting the rising demand for energy. Converting carbon dioxide (CO2) into green energy and valuable feedstocks through photocatalytic CO2 reduction (PCCR) offers a promising solution to global warming and energy needs. Among all semiconductors, zinc oxide (ZnO) has garnered considerable interest due to its ecofriendly nature, biocompatibility, abundance, exceptional semiconducting and optical properties, cost-effectiveness, easy synthesis, and durability. This review thoroughly discusses recent advances in mechanistic insights, fundamental principles, experimental parameters, and modulation of ZnO catalysts for direct PCCR to C1 products (methanol). Various ZnO modification techniques are explored, including atomic size regulation, synthesis strategies, morphology manipulation, doping with cocatalysts, defect engineering, incorporation of plasmonic metals, and single atom modulation to boost its photocatalytic performance. Additionally, the review highlights the importance of photoreactor design, reactor types, geometries, operating modes, and phases. Future research endeavors should prioritize the development of cost-effective catalyst immobilization methods for solid-liquid separation and catalyst recycling, while emphasizing the use of abundant and non-toxic materials to ensure environmental sustainability and economic viability. Finally, the review outlines key challenges and proposes novel directions for further enhancing ZnO-based photocatalytic CO2 conversion processes.

Prediction of crystal structure, phase group, and stability of 2D materials through data science coupled with DFT

Two-dimensional materials offer unique mechanical strength, electrical conductivity, and tunable electronic properties, driving advancements in electronics, energy storage, and catalysis. The crystal structure of the materials is fundamental to several properties. In this study, we employ solely the elemental features to construct a random forest classification model achieving an 82.56% accuracy in predicting the crystal structure of the compounds. Additionally, one vs rest binary classifiers are developed for each crystal structure within our 2D materials dataset, with accuracies ranging from 86% to 99%, enhancing their practical utility. Features based on stoichiometric norms have been found to play a major role in the prediction because of the fact that certain crystal structures are associated with more balanced compositions, leading to lower norm values, while others with imbalances have higher norm values. It has also been observed that material compositions with an overall higher amount of p valence electrons tend to show monoclinic, orthorhombic and triclinic crystal structures. Moreover, in compounds with a lower total electronegativity of the constituent elements, the occurrence of trigonal and tetragonal systems has been found to be more probable. In a similar fashion, hexagonal crystal systems have been more often found in compounds where the sum of Van Der Waals radius of the constituent atoms is lower. To validate our model, Crystal structure AnaLYsis by Particle Swarm Optimization (CALYPSO) method, and Density Functional Theory (DFT) are utilized to predict the crystal structure, phase group, and stability of a 2D material.

Prediction of non-magnetic metal sphere radius using real part slope of inductance

The existing eddy current measurement thickness technology is mainly for one or more layers of metal plate. However, in real life, the radius of spherical metal is also a critical parameter, so using an eddy current sensor to measure it is also an important research topic. In this paper, real part slope of inductance is used as the characteristic value to characterize the radius of the non-magnetic metal sphere. Compared with the traditional real part of inductance, the addition of slope improves the accuracy and stability of detection data. The experimental results show that the maximum error rate of radius measurement is less than 2%, and the overall data stability is good, which can meet the measurement requirements.

The effect of Yb content on the dielectric properties of TlIn0,99Yb0,01S2 thin films

This paper investigated the impact of ytterbium content on the dielectric properties of TlInS2 and TlIn(0,999)Yb(0,001)S2 thin films with varying thicknesses. The dielectric properties of the thin films were analyzed using dielectric spectroscopy across a frequency range of 1 Hz to 1 MHz and a temperature range of 213 K to 373 K. The XRD analysis was used to obtain the structural information of TlInS2 and TlIn(0,999)Yb(0,001)S2 thin films. A comprehensive dielectric analysis of TlInS2 and TlIn(0,999)Yb(0,001)S2 thin films was presented. Using the Cole-Cole equation, It demonstrated that the thin films are electrically heterogeneous structures displaying a non-Debye behavior depending on temperature. The finding indicates that the free volume increases with increasing thicknesses due to the presence of Yb with a greater atomic radius than TlIn(0,999)Yb(0,001)S2 thin films. This study shows that even a small amount of Yb significantly impacts the dielectric properties of both films with the same thicknesses, thereby expanding its potential technological applications.

Structure and noncovalent interaction analysis of half-sandwich chiral-at-metal organoruthenium complexes with planar and η6-arene-metal axial chirality

The preparation of organometallic compounds has provided new molecular complexity in molecular science. The stereochemistry of organometallics has long been a great challenge due to the large atom radius and more valency. Herein, eight ruthenium piano-stool shaped Ru complexes 2a-d and 3a-d with η6-planar chirality and Ru-center chirality were successfully synthesized and resolved, and the absolute configurations were determined by X-ray diffraction analysis. Racemization experiment revealed the stability differences of the Ru complexes in various solvents. Furthermore, in asymmetric unit of complex 3b, there are two conformers bearing opposite η6-arene-Ru axial chirality. In the supramolecular organization, all of the Ru complexes are stabilized by hydrogen bonding, C-H···π and Cl/I···π interactions. π···π interaction was only observed in complex 3b. The Hirshfeld surface analysis and 2D fingerprint plot show that the eight complexes have a similar contribution composed of different noncovalent interactions to the stabilization of crystal packing. Interestingly, only complex 3b has 3.5% C···C interaction. The pairwise interaction energies were calculated using CE-B3LYP energy model and further visualized by 3D-energy framework, which show that the dominant interaction for all complexes is dispersion. Theoretical energies of each conformer of complex 2a-d and 3a-d with η6-arene rotated at different angles was calculated by DFT method, and the energy barrier between the two conformations was around 2-3 kcal/mol. Halide ligand exchange experiments on complex (SRu)-1a were performed and the plausible mechanism was proposed. These results have provided new evidence for the design of molecules with specific fine stereochemical information for potential applications in medicinal and material chemistry.

Mechanical and electrical properties of two-dimensional BN doped with carbon group elements: first-principles calculations

Abstract Research of low-dimensional nanomaterials provides a direction for solving the problems of energy and environmental pollution. In this work, the regulation mechanism of doping carbon group elements X (X = C, Si, Ge, Pb, Sn) on mechanical and electrical properties of 2D monolayer BN are investigated by first-principles calculations. Two doping sites were selected, replace B atoms (B15N16X) or N atoms (B16N15X). Lower relative enthalpies and the elastic constants, which conforming to the mechanical stability standard, fully prove the stability of the doping system. Compared with B15N16X, B16N15X has larger structural distortion, smaller elastic constants and modulus, and is more inclined to ductility. With the increase of atomic radius, the deformation degree increases and the elastic parameters decrease. C-doped by replacing B atoms improves the elastic mechanical properties of monolayer BN. Sn-doped and Pb-doped modulate the monolayer BN into ductility. More importantly, all doped configurations exhibit magnetism. The indirect band gap of the undoped system can also be modulated into a direct band gap, B15N16C, B16N15Si and B16N15Ge all have direct band gaps in the spin-down direction. Asymmetric impurity energy levels DOS further verify the magnetism of the reference system.

Microstructure evolution by thermomechanical processing and high-temperature mechanical properties of HCP-high entropy alloys (HEAs)

The solid solution strengthening and deformation mechanisms of the hcp-high entropy alloys (HEAs) have not been elucidated yet because of the complexity introduced by the martensite structure, particularly in alloys undergoing phase transformation from bcc to hcp structures. This study attempted to fabricate an equiaxed hcp microstructure for TiZrHf, TiZrHfAl, and TiZrHfAlSc using the hot rolling process and subsequent heat treatment. The equiaxed hcp phase was successfully formed in TiZrHfAl, but the martensite structure remained in TiZrHf and TiZrHfAlSc. The 0.2 % proof stress at 400 and 600°C increased with the increase of the average atomic size misfit δ and the mixing entropy . The low activation volume and high-stress exponent of TiZrHfAl with equiaxed hcp phase at room temperature suggest minor obstacles such as cluster or short-range order inhibit the movement of dislocation and leading to the significant solid-solution strengthening. However, the effect diminished at 600 °C.

Heteroatom‐induced tensile strain in copper lattice boosts CO2 electroreduction toward multi‐carbon products

AbstractStrain engineering on metal‐based catalysts has been utilized as an efficacious strategy to regulate the mechanism and pathways in various electrocatalytic reactions. However, controlling strain and establishing the strain‐activity relationship still remain significant challenges. Herein, three different and continuous tensile strains (CuPd‐1.90%, CuAu‐3.37%, and CuAg‐4.33%) are successfully induced by introducing heteroatoms with different atomic radius. The catalytic performances of CuPd‐1.90%, CuAu‐3.37%, and CuAg‐4.33% display a positive correlation against tensile strains in electrochemical CO2 reduction reaction (CO2RR). Specifically, CuAg‐4.33% exhibits superior catalytic performance with a 77.9% Faradaic efficiency of multi‐carbon products at −300 mA cm−2 current density, significantly higher than those of pristine Cu (Cu‐0%). Theoretical calculations and in situ spectroscopies verify that tensile strain can affect the d‐band center of Cu, thereby altering the binding energy of *CO intermediates and Gibbs free energies of the C–C coupling procedure. This work might highlight a new method for precisely regulating the lattice strain of metallic catalysts in different electrocatalytic reactions.

Microwave Remote Sensing with Hybrid Quantum Receiver.

Detecting a microwave signal that is emitted or reflected by distant targets is a powerful tool in fundamental science and industrial technology. Solid-state spins provide an opportunity to realize quantum-enhanced remote sensing under ambient conditions. However, the weak interaction between the free-space signal and atomic size sensor limits the sensitivity. This hinders the realization of practical quantum remote sensing. Here, we demonstrate active microwave remote sensing with a diamond-based hybrid quantum receiver by combining electromagnetic field localization at nanoscale with quantum spin manipulation. A method of differential spin refocusing (DSR) is developed to overcome the challenge of reducing the impact of inhomogeneities in spin-signal interaction, while the strength of interaction is enhanced by more than 3 orders with nanostructure. It improves the coherent interaction time of quantum receiver by 30-fold, substantially enhancing the sensitivity and stability. By detecting the reflected microwave with picotesla sensitivity, diamond remote sensing monitors the real-time status of a centimeter-sized target at 2 m distance. Our method is general to various solid-state spins. The results will expand the applications of solid-state spin quantum sensors in areas ranging from medical imaging to resource survey.

Tomography of entangling two-qubit logic operations in exchange-coupled donor electron spin qubits

Scalable quantum processors require high-fidelity universal quantum logic operations in a manufacturable physical platform. Donors in silicon provide atomic size, excellent quantum coherence and compatibility with standard semiconductor processing, but no entanglement between donor-bound electron spins has been demonstrated to date. Here we present the experimental demonstration and tomography of universal one- and two-qubit gates in a system of two weakly exchange-coupled electrons, bound to single phosphorus donors introduced in silicon by ion implantation. We observe that the exchange interaction has no effect on the qubit coherence. We quantify the fidelity of the quantum operations using gate set tomography (GST), and we use the universal gate set to create entangled Bell states of the electrons spins, with fidelity 91.3 ± 3.0%, and concurrence 0.87 ± 0.05. These results form the necessary basis for scaling up donor-based quantum computers.

Data-Driven Design of Single-Atom Electrocatalysts with Intrinsic Descriptors for Carbon Dioxide Reduction Reaction

The strategic manipulation of the interaction between a central metal atom and its coordinating environment in single-atom catalysts (SACs) is crucial for catalyzing the CO2 reduction reaction (CO2RR). However, it remains a major challenge. While density-functional theory calculations serve as a powerful tool for catalyst screening, their time-consuming nature poses limitations. This paper presents a machine learning (ML) model based on easily accessible intrinsic descriptors to enable rapid, cost-effective, and high-throughput screening of efficient SACs in complex systems. Our ML model comprehensively captures the influences of interactions between 3 and 5d metal centers and 8 C, N-based coordination environments on CO2RR activity and selectivity. We reveal the electronic origin of the different activity trends observed in early and late transition metals during coordination with N atoms. The extreme gradient boosting regression model shows optimal performance in predicting binding energy and limiting potential for both HCOOH and CO production. We confirm that the product of the electronegativity and the valence electron number of metals, the radius of metals, and the average electronegativity of neighboring coordination atoms are the critical intrinsic factors determining CO2RR activity. Our developed ML models successfully predict several high-performance SACs beyond the existing database, demonstrating their potential applicability to other systems. This work provides insights into the low-cost and rational design of high-performance SACs.

Surface morphology evolution, microstructural response and mechanical property variation of Au-ion irradiated CrNbZrMoV, TiCrZrMoV, TiNbCrMoV, TiNbZrCrV and TiNbZrMoCr high-entropy alloy coatings

In this work, the CrNbZrMoV, TiCrZrMoV, TiNbCrMoV, TiNbZrCrV and TiNbZrMoCr refractory high-entropy alloy (RHEA) coatings (1.67∼2.77 μm of thickness) were prepared by magnetron sputtering. Then, 6 MeV Au-ion irradiations with 2.5 × 1015 to 1.0 × 1016 ions/cm2 fluences were performed on these coatings at 473 K, and the surface morphology, microstructure and mechanical property were investigated. The peak damage of the CrNbZrMoV, TiCrZrMoV, TiNbCrMoV, TiNbZrCrV and TiNbZrMoCr coatings under 1.0 × 1016 ions/cm2 fluence are 48, 48, 44, 48 and 48 dpa, respectively. The peak Au concentration of the CrNbZrMoV, TiCrZrMoV, TiNbCrMoV, TiNbZrCrV and TiNbZrMoCr coatings under 1.0 × 1016 ions/cm2 fluence are 4.27 × 103, 4.12 × 103, 4.13 × 103, 4.16 × 103 and 4.37 × 103 appm, respectively. The surface morphology of the CrNbZrMoV, TiCrZrMoV, TiNbZrCrV and TiNbZrMoCr coatings were smoothed obviously. For BCC TiNbCrMoV coating, irradiation caused the growth of the nanocrystalline. For amorphous coatings, the crystallization occurred in the CrNbZrMoV, TiCrZrMoV and TiNbZrMoCr coatings after 2.5 × 1015 fluence irradiation (≥12 dpa), while the TiNbZrCrV coating remain mainly amorphous structure after all irradiation. It was found that irradiation induced continuous crystallization occurred not only at the surface but also in the peak damage zone of the coating, and then grew inside the irradiated region. Apparent irradiation hardening was observed in all the coatings except the TiNbZrCrV coating. The structural stability of these coatings under the current irradiation condition was discussed. Preliminary study shows that the great irradiation tolerance of amorphous TiNbZrCrV coating may be related to the lowest electronegativity difference (Δχ = 0.121) and large atomic size difference (δ = 9.042 %) that stabilize the structure and inhibit atomic diffusion, respectively. These findings provide the guidance for the development of high irradiation tolerance materials for future nuclear energy applications with great structural stability.

Exploring uncertainty in glass phase transitions through machine learning

Machine learning methods have shown significant potential and are widely used in modern physics research. However, the uncertainty linked to machine learning, arising from the opacity of its workflow, demands attention and consideration. This study investigates the application of machine learning models in analyzing the glass transition of Cu50Zr50 metallic glass. By employing supervised learning techniques with ResNet50, MobileNetV3, and GoogleNet image extraction models, the study reveals that while machine learning can capture variations in the disordered atomic structure during the transition process, different models may yield divergent results in determining the glass transition temperature. Moreover, variations in atomic sizes within the images can lead to fluctuations in the predicted transition temperatures. These findings highlight the inherent uncertainties associated with using machine learning to analyze continuous phase transitions and stress the importance of ensuring that the extracted structural features align with the physical characteristics of the transition process.