Please use this identifier to cite or link to this item: http://dx.doi.org/10.25673/33597
Full metadata record
DC FieldValueLanguage
dc.contributor.refereeMarques, Miguel-
dc.contributor.refereePaul, Wolfgang-
dc.contributor.refereeRinke, Patrick-
dc.contributor.authorGonçalves Marques, Mário Rui-
dc.date.accessioned2020-06-18T09:20:36Z-
dc.date.available2020-06-18T09:20:36Z-
dc.date.issued2020-
dc.identifier.urihttps://opendata.uni-halle.de//handle/1981185920/33794-
dc.identifier.urihttp://dx.doi.org/10.25673/33597-
dc.description.abstractDiese Arbeit leistet einen Beitrag zur Entwicklung und Charakterisierung neuer Materialien. Für viele Simulationen, z. B. Molekulardynamik-Simulationen, die zur Berechnung von Materialeigenschaften benutzt werden, ist es notwendig die Gesamtenergie und ihre Ableitungen tausende Male zu berechnen. Diese Zahl steigt für lange Simulationen oder große Systeme leicht in die Millionen an, was trotz effizienter Methoden wie Dichte-Funktional-Theorie extrem hohe Rechenkosten verursacht. Das Ziel dieser Arbeit ist die Entwicklung von Strategien diese Hindernisse mittels maschinellen Lernmethoden zu umgehen.ger
dc.description.abstractThis thesis provides a contribution to the problem of material discovery and characterization. Many simulations used to predict properties of materials, such as molecular dynamics and structural prediction, require thousands of total energy calculations (and its derivatives). This number can easily grow above millions for large systems or for long simulation times, which translates to high computational costs even for methods as efficient as density functional theory (which is the standard method to perform these calculations in material science). The aim of this thesis is to develop strategies to counter these obstacles using machine learning techniques.eng
dc.format.extent1 Online-Ressource (135 Seiten)-
dc.language.isoeng-
dc.rights.urihttp://rightsstatements.org/vocab/InC/1.0/-
dc.subject.ddc530-
dc.titleThe structure and dynamics of materials using machine learning-
dcterms.dateAccepted2020-05-05-
dcterms.typeHochschulschrift-
dc.typePhDThesis-
dc.identifier.urnurn:nbn:de:gbv:3:4-1981185920-337946-
local.versionTypepublishedVersion-
local.publisher.universityOrInstitutionMartin-Luther-Universität Halle-Wittenberg-
local.subject.keywordsThis thesis provides a contribution to the problem of material discovery and characterization. Many simulations used to predict properties of materials, such as molecular dynamics and structural prediction, require thousands of total energy calculations (and its derivatives). This number can easily grow above millions for large systems or for long simulation times, which translates to high computational costs even for methods as efficient as density functional theory (which is the standard method to perform these calculations in material science). The aim of this thesis is to develop strategies to counter these obstacles using machine learning techniques.-
local.subject.keywordsMaschinelles Lernen, Strukturvorhersage, Molekulardynamik, Kraftfelder neuronaler Netze, Defekte, Photovoltaik, Dichtefunktionaltheorie, Clustererweiterung, Cui, CZTS, Si, Cu, Au-
local.subject.keywordsMachine Learning, Structure prediction, Molecular dynamics, Neural Network force-fields, Defects, Photovoltaics, Density functional theory, Cluster expansion, Cui, CZTS, Si, Cu, Au-
local.openaccesstrue-
dc.identifier.ppn1701038692-
local.publication.countryXA-DE-
cbs.sru.importDate2020-06-18T09:18:09Z-
local.accessrights.dnbfree-
Appears in Collections:Interne-Einreichungen

Files in This Item:
File Description SizeFormat 
Mario_Marques_thesis.pdf11.78 MBAdobe PDFThumbnail
View/Open