Aktivno učenje interatomskih potencijala za predikciju kristalne strukture

master's thesis

• Title (english): Active learning of interatomic potentials for crystal structure prediction
• Author: Janko Jurdana
• Mentor: Ivor Lončarić (mentor)
• Mentor: Danko Radić (komentor)
• Committee member: Danko Radić (predsjednik povjerenstva)
• Committee member: Ivor Lončarić (član povjerenstva)
• Committee member: Davor Horvatić (član povjerenstva)
• Committee member: Andrej Novak (član povjerenstva)
• Committee member: Matko Glunčić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2023-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Predikcija kristalne strukture predstavlja vrlo zahtjevan zadatak u modernoj fizici čvrstog stanja. Većina dosadašnjih metoda zasnovana je na teoriji funkcionala gustoće. Međutim, takve metode postaju računalno prezahtjevne kada su promatrani sustavi kristali organskih molekula. Razvojem metoda strojnog učenja, moguće je ubrzati izračune svojstava takvih sustava što otvara vrata proučavanju široke klase materijala. U ovom radu su takve tehnike iskorištene za predikciju energija osnovnog stanja i sila na atome treniranjem strojno naučenih potencijala. Svojstva izračunata tim tehnikama su potom uspoređena sa svojstvima dobivenim korištenjem metoda teorije funkcionala gustoće.

Abstract (english)

Crystal structure prediction poses a very demanding task in modern day solid state physics. Most of the existing methods are based on density functional theory. However, these methods become computationally too demanding when the studied systems are crystals of organic molecules. With the development of machine learning techniques, the calculations of the properties of such systems were drastically accelerated and thus opened the door to the study of a wide class of materials. In this paper, such methods were used for predicting ground state energies and forces on the atoms by training machine learned force fields. The properties calculated by these techniques were then compared with the properties obtained using density functional theory.

• Keywords: teorija funkcionala gustoće; strojno učenje; predikcija kristalne strukture; deskriptori
• Keywords (english): density functional theory; machine learning; crystal structure prediction; descriptors
• Language: croatian
• URN:NBN: urn:nbn:hr:217:041725
• Project: Number: UIP-2020-02-5675; Title: Povećanje prostorne i vremenske skale modeliranja materijala iz prvih principa pomoću strojnog učenja; Title: Extending length and time scales of first principles material modelling via machine learning; Acronym: ExtMatModelML; Leader: Ivor Lončarić; Jurisdiction: Croatia; Funder: HRZZ; Funding stream: UIP
• Study programme: Title: The University Integrated Undergraduate and Graduate Programme in Physics; Research Physics; specializations in: Research Physics; Course: Research Physics; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: sveučilišni magistar fizike (sveučilišni magistar fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-02-15 14:32:22