Search

The structure of evaporated amorphous Ge$_x$Sb$_x$Te$_{100-2x}$ ($x =$ 6, 9, 13) alloys was investigated by neutron diffraction, X-ray diffraction and extended X-ray absorption spectroscopy (EXAFS) at the Ge, Sb and Te K-edges. Large scale structural models were generated by fitting the experimental datasets (5 for each composition) simultaneously in the framework of the reverse Monte Carlo simulation technique. It was found that the alloys are chemically ordered (Ge and Sb have predominantly Te neighbors) and within the experimental uncertainty each component satisfies the $8-N$ rule. A comparison with the pair correlation functions of melt quenched Ge$_{20}$Te$_{80}$ revealed that the first minimum of $g_{\mathrm{TeTe}}(r)$ is shallower in the ternary alloys than in Ge$_{20}$Te$_{80}$. On the other hand, the separation of the first and second coordination environments of Ge atoms is stronger in the Ge-Sb-Te alloys investigated., Comment: 32 pages, submitted to Journal of Alloys and Compounds

Equivariant graph neural networks force fields (EGraFFs) have shown great promise in modelling complex interactions in atomic systems by exploiting the graphs' inherent symmetries. Recent works have led to a surge in the development of novel architectures that incorporate equivariance-based inductive biases alongside architectural innovations like graph transformers and message passing to model atomic interactions. However, thorough evaluations of these deploying EGraFFs for the downstream task of real-world atomistic simulations, is lacking. To this end, here we perform a systematic benchmarking of 6 EGraFF algorithms (NequIP, Allegro, BOTNet, MACE, Equiformer, TorchMDNet), with the aim of understanding their capabilities and limitations for realistic atomistic simulations. In addition to our thorough evaluation and analysis on eight existing datasets based on the benchmarking literature, we release two new benchmark datasets, propose four new metrics, and three challenging tasks. The new datasets and tasks evaluate the performance of EGraFF to out-of-distribution data, in terms of different crystal structures, temperatures, and new molecules. Interestingly, evaluation of the EGraFF models based on dynamic simulations reveals that having a lower error on energy or force does not guarantee stable or reliable simulation or faithful replication of the atomic structures. Moreover, we find that no model clearly outperforms other models on all datasets and tasks. Importantly, we show that the performance of all the models on out-of-distribution datasets is unreliable, pointing to the need for the development of a foundation model for force fields that can be used in real-world simulations. In summary, this work establishes a rigorous framework for evaluating machine learning force fields in the context of atomic simulations and points to open research challenges within this domain.

Pressure induced structural modifications in vitreous Ge$_{x}$Se$_{100-x}$ (where 10 $\leq$ x $\leq$ 25)are investigated using X-ray absorption spectroscopy (XAS) along with supplementary X-ray diffraction (XRD) experiments and ab initio molecular dynamics (AIMD) simulations. Universal changes in distances and angle distributions are observed when scaled to reduced densities. All compositions are observed to remain amorphous under pressure values up to 42 GPa. The Ge-Se interatomic distances extracted from XAS data show a two-step response to the applied pressure; a gradual decrease followed by an increase at around 15-20 GPa, depending on the composition. This increase is attributed to the metallization event that can be traced with the red shift in Ge K edge energy which is also identified by the principal peak position of the structure factor. The densification mechanisms are studied in details by means of AIMD simulations and compared to the experimental results. The evolution of bond angle distributions, interatomic distances and coordination numbers are examined and lead to similar pressure-induced structural changes for any composition.

Raman scattering, IR reflectance and modulated DSC measurements are performed on specifically prepared dry (AgI)x(AgPO3)1-x glasses over a wide range of compositions 0 < x < 0.60. A reversibility window is observed in the 0.095< x < 0.378 range, which fixes the elastically rigid but unstressed regime also known as the Intermediate Phase. Glass compositions at x < 0.095 are stressed-rigid, while those at x > 0.378 elastically flexible. Raman optical elasticity power-laws, trends in the nature of the glass transition endotherms corroborate the three elastic phase assignments. Ionic conductivities reveal a step-like increase when glasses become stress-free at x > xc(1) = 0.095, and a logarithmic increase in conductivity (sigma ~ (x-xc(2)t) once they become flexible at x > xc(2) = 0.378 with a power-law t = 1.78. The power-law is consistent with percolation of 3D filamentary conduction pathways. Traces of water doping lower Tg and narrow the reversibility window, and can also completely collapse it. Ideas on network flexibility promoting ion-conduction are in harmony with the unified approach of Ingram et al., who have emphasized the similarity of process compliance or elasticity relating to ion-transport and structural relaxation in decoupled systems. Boson mode frequency and scattering strength display thresholds that coincide with the two elastic phase boundaries. In particular, the scattering strength of the boson mode increases almost linearly with glass composition x, with a slope that tracks the floppy mode fraction as a function of mean coordination number r predicted by mean-field rigidity theory. These data suggest that the excess low frequency vibrations contributing to boson mode in flexible glasses come largely from floppy modes.

A low temperature Monte Carlo dynamics of a Keating like oscillator model is used to study the relationship between the nature of glasses from the viewpoint of rigidity, and the strong-fragile behaviour of glass-forming liquids. The model shows that a Phillips optimal glass formation with minimal enthalpic changes is obtained under a cooling/annealing cycle when the system is optimally constrained by the harmonic interactions, i.e. when it is isostatically rigid. For these peculiar systems, the computed fragility shows also a minimum, which demonstrates that isostatically rigid glasses are strong (Arrhenius-like) glass-forming liquids. Experiments on chalcogenide and oxide glass-forming liquids are discussed under this new perspective and confirm the theoretical prediction for chalcogenide network glasses., Comment: 4 pages, 3 figures

The rigid to floppy transitions and the associated intermediate phase in glasses are studied in the case where the local structure is not fully determined from the macroscopic concentration. The approach uses size increasing cluster approximations and constraint counting algorithms. It is shown that the location and the width of the intermediate phase and the corresponding structural, mechanical and energetical properties of the network depend crucially on the way local structures are selected at a given concentration. The broadening of the intermediate phase is obtained for networks combining a large amount of flexible local structural units and a high rate of medium range order., Comment: 4 pages, 4 figures

Classical molecular dynamics simulations are used to study the structure of densified germanium dioxide ($GeO_2$). It is found that the coordination number of germanium changes with increasing density (pressure) while pressure released systems do not exhibit a marked change in local structure as compared to the virgin system. The structural modification with pressure appears to be stepwise and gradually affects long range (through the reduction of the long-range correlations as seen from the shift of the first sharp diffraction peak), intermediate range (by angular reduction) and finally short range structure (by tetrahedron distorsion), Comment: 4 pages, 4 figures

Calorimetric experiments on network glasses provide information on the ergodicity (landscape) temperature of supercooled liquids and can be compared with a recent theory developed by Hall and Wolynes [PRL90, 085505 (2003)], Comment: 2 pages, 2 EPS figures RevTEX. to appear in Physical review Letters

There is growing evidence that electronic and molecular networks present some common universal properties, among which the existence of a self-organized intermediate phase. In glasses, the latter is revealed by the reversibility window obtained from complex calorimetric measurements at the glass transition. Here we focus on amorphous networks and we show how this intermediate phase can be understood from a rigidity percolation analysis on size increasing clusters. This provides benchmarks and guidance for an electromechanical analogy with high temperature super conductors, Comment: 15 pages, latex, 4 figures, to appear in J. of Superconductivity

Subj-class: Disordered Systems and Neural Networks, Comment: 12 pages, revtex, 3 figures

Simple statistical agglomeration models can provide a universal link between the local structure and the glass transition temperature in network glasses. We first stress the physical features of the models and the hypothesis made, and then show how to define the glass transition temperature. The models are applied to various types of binary, ternary and multicomponent chalcogenide glass networks and the predictions compared to experimental data., Comment: 11 pages, Revtex, 5 EPS figures, Proceedins of the Cargese CNRS school "Physics of glasses: structure and dynamics", Cargese May 1999

We give general topological rules which very accurately predict the chemical trends in glass transition temperature $T_g$ variation as a function of cross-linking. In multicomponent glasses, these chemical trends permit to distinguish homogeneous compositions (random network) from inhomogeneous ones (local phase separation). The stochastic origin of the Gibbs-Di Marzio equation is predicted at low connectivity and the analytical expression of its parameter emerges naturally from the calculation., Comment: 5 pages, Revtex, accepted for publication in Europhysics Letters, 2 figures not included, available at http://www.gcr.jussieu.fr/matthieu.htm#Research

We present in this letter a model of glass formation using energy barriers and a crystal nucleation process. We then analyze the corresponding dynamical equation in the vicinity of the stationary solutions. The occurence of a pure amorphous solution (i.e. glass) is due to the contribution of a ratio $\Lambda$ depending on the cooling rate and the crystal nucleation frequency. We also construct time-temperature-transformation (TTT) curves in order to relate the model with the kinetic treatment of glass formation., Comment: 4 pages, revtex, 3 figures postscript

In this article, we present a universal relationship between the glass transition temperature $T_g$ and the local glass structure. The derivation of the simplest expression of this relationship and some comparisons with experimental $T_g$ values have already been reported in a recent letter. We give here the analytical expression of the parameter $\beta$ of the Gibbs-Di Marzio equation and also new experimental probes for the validity of the relationship, especially in low modified binary glasses., Comment: 31 pages, LaTeX, 16 Postscript figures, mmi@ccr.jussieu.fr