Novel method for realistically simulating the deposition of thin films from the gas phase and its application to study the growth of thin gold film on crystalline silicon

Abstract

We present a novel approach for simulating thin film (TF) deposition from the gas phase at the atomistic scale, combining molecular dynamics (MD) and time-stamped force-bias Monte Carlo (tfMC). In this approach, MD, with its fine temporal resolution, captures fast events, such as incident atom-substrate collisions, while tfMC simulates slow relaxation processes, enhancing temporal scale coverage. The proposed approach also adequately models deposition conditions, for example, by accounting for realistic energy and angle distributions in the description of the incident flux. To demonstrate its efficacy, we apply it to simulate the physical vapor deposition of a 3 nm Au TF on crystalline Si. We find that the entire deposition process consisted of four distinct stages: (i) the initial degradation of the Si substrate, (ii) formation of a mixed Au-Si interface layer, (iii) nucleation and growth of a polycrystalline Au layer, proceeding in a fashion close to the Frank-van der Merwe mode (layer-by-layer growth), and (iv) postdeposition relaxation of microstructure. The produced TF was comprehensively characterized, revealing that the deposited polycrystalline Au layer contained a considerable number of defects, including dislocations, stacking faults, grain boundaries, and Si impurities. The analysis also showed that in the simulated high-energy deposition the Si substrate was considerably degraded and that the disordered Au-Si layer which formed at the interface resembled the melt-quenched Au₈₂Si₁₈ eutectic. A comparison with an analogous MD simulation revealed that the MD + tfMC approach extended the accessible time scale 5-fold, allowing us to reach the microsecond scale, and yielding a TF with higher crystallinity and better-developed microstructure. The deposition rate used in the MD + tfMC simulation was two to 3 orders of magnitude lower than in other recent, but purely MD, simulations, being significantly closer to experiment.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata

Altmetrics