Device-size atomistic models of amorphous silicon

Vink, R. L. C.; Barkema, G. T.; Stijnman, M; Bisseling, Rob H.

doi:10.1103/physrevb.64.245214

Cited by 30 publications

(36 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An important speed-up is achieved by parallelization. We use an asynchronous master-worker parallelization paradigm, where the master proposes transpositions and workers report on their success, instead of a bulk synchronous parallelization proposed by Vink et al 10 In our procedure, the annealing temperature is slowly decreased from about 0.3 eV to about 0.15 eV per silicon atom. Following Barkema and Mousseau, we performed a zero-temperature quench every several thousand successful transpositions at the annealing temperature.…”

Section: Methodsmentioning

confidence: 99%

Nearly hyperuniform network models of amorphous silicon

2013

View full text Add to dashboard Cite

We introduce the concept of nearly hyperuniform network (NHN) structures as alternatives to the conventional continuous random network (CRN) models for amorphous tetrahedrally coordinated solids, such as amorphous silicon (a-Si). A hyperuniform solid has a structure factor S(k) that approaches zero as the wavenumber k → 0. We define a NHN as an amorphous network whose structure factor S(k → 0) is smaller than the liquid value at the melting temperature. Using a novel implementation of the Stillinger-Weber potential for the interatomic interactions, we show that the energy landscape for a spectrum of NHNs includes a sequence of local minima with an increasing degree of hyperuniformity [smaller S(k → 0)] that is significantly below the frozen-liquid value and that correlates with other measurable features in S(k) at intermediate and large k and with the width of the electronic band gap.

show abstract

Section: Methodsmentioning

confidence: 99%

Nearly hyperuniform network models of amorphous silicon

2013

View full text Add to dashboard Cite

show abstract

“…These optimizations are similar in spirit to the optimizations used in the scalable WWW algorithm. 8,9 The first optimization is aimed at reducing the CPU time spent on rejected bond transpositions. After a bond transposition in the original algorithm, the energy of the network is always completely minimized.…”

Section: Optimized Bond-switching Algorithm For Silicamentioning

confidence: 99%

Large well-relaxed models of vitreous silica, coordination numbers, and entropy

Vink

Barkema²

2003

Phys. Rev. B

Self Cite

View full text Add to dashboard Cite

A Monte Carlo method is presented for the simulation of vitreous silica. Well-relaxed networks of vitreous silica are generated containing up to 300 000 atoms. The resulting networks, quenched under the BKS potential, display smaller bond-angle variations and lower defect concentrations, as compared to networks generated with molecular dynamics. The total correlation functions T(r) of our networks are in excellent agreement with neutron scattering data, provided that thermal effects and the maximum inverse wavelength used in the experiment are included in the comparison. A procedure commonly used in experiments to obtain coordination numbers from scattering data is to fit peaks in rT(r) with a gaussian. We show that this procedure can easily produce incorrect results. Finally, we estimate the configurational entropy of vitreous silica.

show abstract

“…Indeed, the cooling rates that have previously been reported in DFT-MD simulations of a-Si (≈ 10 14 K/s) are several orders of magnitude faster than those in experiments (19)(20)(21). On the other hand, classical force fields require much less computational effort, giving access to nanometer-scale ("device-size") systems (9), but they are rarely accurate enough to fully correctly describe the structural variations present in the amorphous state.…”

mentioning

confidence: 99%

Realistic Atomistic Structure of Amorphous Silicon from Machine-Learning-Driven Molecular Dynamics

Deringer

Bernstein

Bartók

et al. 2018

J. Phys. Chem. Lett.

207

169

View full text Add to dashboard Cite

Amorphous silicon ( a-Si) is a widely studied noncrystalline 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 using a machine-learning-based interatomic potential. Our best a-Si network is obtained by simulated cooling from the melt at a rate of 10 K/s (that is, on the 10 ns time scale), contains less than 2% defects, and agrees with experiments regarding excess energies, diffraction data, and Si NMR chemical shifts. We show that this level of quality is impossible to achieve with faster quench simulations. We then generate a 4096-atom system that correctly reproduces the magnitude of the first sharp diffraction peak (FSDP) in the structure factor, achieving the closest agreement with experiments to date. Our study demonstrates the broader impact of machine-learning potentials for elucidating structures and properties of technologically important amorphous materials.

show abstract

Device-size atomistic models of amorphous silicon

Cited by 30 publications

References 23 publications

Nearly hyperuniform network models of amorphous silicon

Nearly hyperuniform network models of amorphous silicon

Large well-relaxed models of vitreous silica, coordination numbers, and entropy

Realistic Atomistic Structure of Amorphous Silicon from Machine-Learning-Driven Molecular Dynamics

Contact Info

Product

Resources

About