Fast Crystallization of the Phase Change Compound GeTe by Large-Scale Molecular Dynamics Simulations

Sosso, Gabriele C.; Miceli, Giacomo; Caravati, S.; Giberti, Federico; Behler, Jörg; Bernasconi, Marco

doi:10.1021/jz402268v

Cited by 144 publications

(168 citation statements)

References 42 publications

(79 reference statements)

Supporting

Mentioning

161

Contrasting

Order By: Relevance

“…The presence of wrong bonds, particularly Te-Te, in our crystallized samples is striking, and Ge-Ge and Te-Te bonds are also present in simulations performed on GeTe [32]. The arrangement of Ge, Sb, and cavities in c-GST has often been discussed [12], and substantial displacements from the ideal rock salt positions may occur, particularly for Ge [6].…”

Section: Discussionmentioning

confidence: 61%

“…An interesting study of crystallization in GeTe has been performed by Sosso et al [32] using a force field generated from a neural network fitting to a large database of DF calculations of structures containing Ge and Te. They studied the temperature dependence in the range 500-700 K with a cubic 4096-atom cell, and performed test calculations with cells containing 512 and 1728 atoms.…”

Section: B Samples Without History Of Order: Run1run2run3mentioning

confidence: 99%

See 1 more Smart Citation

Simulation of crystallization inGe2Sb2Te5: A memory effect in the canonical phase-change material

2014

View full text Add to dashboard Cite

Crystallization of amorphous Ge 2 Sb 2 Te 5 (GST) has been studied using four extensive (460 atoms, up to 4 ns) density functional/molecular dynamics simulations at 600 K. This phase change material is a rare system where crystallization can be simulated without adjustable parameters over the physical time scale, and the results could provide insight into order-disorder processes in general. Crystallization is accompanied by an increase in the number of ABAB squares (A: Ge, Sb; B: Te), percolation, and the occurrence of low-frequency localized vibration modes. A sample with a history of order crystallizes completely in 1.2 ns, but ordering in others was less complete, even after 4 ns. The amorphous starting structures without memory display phases (>1 ns) with subcritical nuclei (10-50 atoms) ranging from nearly cubical blocks to stringlike configurations of ABAB squares and AB bonds extending across the cell. Percolation initiates the rapid phase of crystallization and is coupled to the directional p-type bonding in metastable GST. Cavities play a crucial role, and the final ordered structure is distorted rock salt with a face-centered cubic sublattice containing predominantly Te atoms. We comment on earlier models based on smaller and much shorter simulations.

show abstract

Section: Discussionmentioning

confidence: 61%

Section: B Samples Without History Of Order: Run1run2run3mentioning

confidence: 99%

Simulation of crystallization inGe2Sb2Te5: A memory effect in the canonical phase-change material

2014

View full text Add to dashboard Cite

show abstract

“…A depth-first search algorithm was then applied to identify clusters as connected components of the graph, similarly to the algorithm used in ref. 46 for the identification of crystal nuclei obtained from undercooled liquid GeTe. This clustering technique is used to compute variables n, n o , N I , and N II used to represent FESs reported in Figs.…”

Section: Methodsmentioning

confidence: 99%

Molecular-dynamics simulations of urea nucleation from aqueous solution

Salvalaglio

Perego

Giberti

et al. 2014

Proc. Natl. Acad. Sci. U.S.A.

Self Cite

161

231

View full text Add to dashboard Cite

Despite its ubiquitous character and relevance in many branches of science and engineering, nucleation from solution remains elusive. In this framework, molecular simulations represent a powerful tool to provide insight into nucleation at the molecular scale. In this work, we combine theory and molecular simulations to describe urea nucleation from aqueous solution. Taking advantage of well-tempered metadynamics, we compute the freeenergy change associated to the phase transition. We find that such a free-energy profile is characterized by significant finite-size effects that can, however, be accounted for. The description of the nucleation process emerging from our analysis differs from classical nucleation theory. Nucleation of crystal-like clusters is in fact preceded by large concentration fluctuations, indicating a predominant two-step process, whereby embryonic crystal nuclei emerge from dense, disordered urea clusters. Furthermore, in the early stages of nucleation, two different polymorphs are seen to compete.T he nucleation of crystals from solution plays an important role in a variety of chemical and engineering processes. However, the very small scale that characterizes the early stages of nucleation makes its experimental study rather challenging. In this regard, molecular simulations play an important role and much work has been devoted to the study of homogeneous nucleation in simple model systems like Lennard-Jones particles or hard spheres (1-3). More recently, growing attention has been paid to the computer simulation of nucleation from solution of organic and inorganic materials (4-11).However, these simulations have to face an important challenge. Nucleation is a typical example of a rare event occurring on a timescale that is much longer than what atomistic simulations can typically afford. The problem is most easily understood if we focus on molecular dynamics (MD), which is the sampling method used here but it plagues other methods as well. In MD, the shortest timescale to be used for a correct integration of the equation of motion is dictated by the fastest atomic motions such as vibrations or rotations. Due to this constraint in the integration step, present MD simulations can reach timescales up to microseconds unless specific hardware, accessible only to few, is used. Even in this case the scale of the accessible times can be stretched only to the milliseconds range, a far cry from the much longer scale of nucleation phenomena.These timescale limitations affect molecular simulations in several other research fields, such as ligand protein binding, protein folding, or slow chemical reactions, to name just a few. To overcome this limit, many enhanced sampling methods have been proposed (12-21). Several of them are based on the application of a suitable external bias potential (12-14) that speeds up configurational sampling and permits free energies to be evaluated and transition rates to be computed (22)(23)(24). Here, we shall use well-tempered (WT) metadynamics to enhance the nucleation...

show abstract

“…[68][69][70][71][72][73][74] In their approach, the system potential energy is expressed as a sum of environment-dependent energies of atoms (E i ) or their pairs (E ij ):…”

Section: Review Wwwq-chemorgmentioning

confidence: 99%

“…[69,71,75] Importantly for applications to materials modeling and design, this method has been shown to provide highaccuracy PESs for bulk materials as well as surfaces and nanostructures, [68,69,72,74,76,77] where other highly accurate approaches developed for molecular systems [22,67] are impractical.…”

Section: Review Wwwq-chemorgmentioning

confidence: 99%

Neural network‐based approaches for building high dimensional and quantum dynamics‐friendly potential energy surfaces

Manzhos

Dawes

Carrington

2014

Int J of Quantum Chemistry

196

148

View full text Add to dashboard Cite

Development and applications of neural network (NN)-based approaches for representing potential energy surfaces (PES) of bound and reactive molecular systems are reviewed. Specifically, it is shown that when the density of ab initio points is low, NNs-based potentials with multibody or multimode structure are advantageous for representing high-dimensional PESs. Importantly, with an appropriate choice of the neuron activation function, PESs in the sum-of-products form are naturally obtained, thus addressing a bottleneck problem in quantum dynamics. The use of NN committees is also analyzed and it is shown that while they are able to reduce the fitting error, the reduction is limited by the nonrandom nature of the fitting error. The approaches described here are expected to be directly applicable in other areas of science and engineering where a functional form needs to be constructed in an unbiased way from sparse data.

show abstract

Fast Crystallization of the Phase Change Compound GeTe by Large-Scale Molecular Dynamics Simulations

Cited by 144 publications

References 42 publications

Simulation of crystallization inGe2Sb2Te5: A memory effect in the canonical phase-change material

Simulation of crystallization inGe2Sb2Te5: A memory effect in the canonical phase-change material

Molecular-dynamics simulations of urea nucleation from aqueous solution

Neural network‐based approaches for building high dimensional and quantum dynamics‐friendly potential energy surfaces

Contact Info

Product

Resources

About