Predviđanje elektronskih i transportnih svojstava silicenskih nanovrpci uporabom neuronskih mreža

undergraduate thesis

• Title (english): Predicting electronic and transport properties of silicene nanoribbons using a neural network
• Author: Tin Župančić
• Mentor: Mirko Poljak (mentor)
• Committee member: Mirko Poljak (predsjednik povjerenstva)
• Committee member: Vladimir Čeperić (član povjerenstva)
• Committee member: Josip Žilak (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Računalne simulacije danas su standardni dio procesa istraživanja i razvoja nanomaterijala i nanokomponenti. Simulacije su ubrzane optimizacijama i masovnom paralelizacijom, ali usprkos tome, vrijeme izvođenja i dalje je dugotrajno. Neuronske mreže nedavno su pokazale potencijal za unapređenje tog procesa. Ovaj rad istražuje performanse neuronskih mreža na primjeru predikcije transportnih svojstava silicenskih nanovrpci (SiNR). Neuronska mreža s tri skrivena sloja trenirana je i evaluirana nad setom od ~10000 nanovrpci širine 2,1 nm dobivenih atomističkim simulacijama. Nakon toga performanse iste mreže (bez novog treniranja) testirane su za nanovrpce drugih širina. Mreža je pokazala dobru kvalitetu predikcije za SiNR širine 2,1 nm, a dodatna testiranja demonstrirala su da neuronska mreža ima potencijal za predviđanje svojstava nanovrpci i za veći raspon dimenzija.

Abstract (english)

Numerical simulations are nowadays a standard part of nanomaterial and nanodevice research and development. Various optimizations and massively parallel processing techniques are used to accelerate simulation process, but time to solution is still relatively high. Neural networks (NN) have recently shown potential for improvement of that process. This paper tests NN performance on silicence nanoribbon (SiNR) example, with the goal of predicting their transport properties. Atomistic simulations were used to generate a set of ~10000 SiNRs 2.1 nm wide, which was then used for training and evaluation. Additionally, the network was tested (without further training) on nanoribbons with different widths. NN has shown good prediction performance for 2.1 nm wide SiNRs and testing with other SiNR widths has demonstrated network’s potential for predicting the nanoribbon properties.

• Keywords: silicenska nanovrpca; pokretljivost elektrona; neuronska mreža; duboko učenje; TensorFlow; atomistička simulacija
• Keywords (english): silicene nanoribbon; mobility; neural network; deep learning; TensorFlow; atomistic simulation
• Language: croatian
• URN:NBN: urn:nbn:hr:168:121121
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 16:33:19