Invito al seminario "Towards Machine-Learning based Modeling of Epitaxy and Nanostructures Evolution", 30 Gennaio 2024 (15.00-15.30)
Mer, 17/01/2024 - 11:49 — Anonimo
Cari colleghi,
Martedì 30 Gennaio alle 15.00 avremo ospite, in IMM Sede, il prof. Francesco Montalenti dell'Università di Milano-Bicocca, che terrà un seminario dal titolo:
Towards Machine-Learning based Modeling of Epitaxy and Nanostructures Evolution
Autori: Daniele Lanzoni, Roberto Bergamaschini, and Francesco Montalenti
Orario: 15.00-15.30 Aula Campisano, IMM Sede
Abstract
After having reviewed a few recent improvements leading to realistic continuum simulation of nanostructures epitaxial growth, I shall illustrate how Machine-Learning techniques can be exploited to speed-up conventional approaches. As a first example, I shall show that a convolutional, recurrent neural network (NN) approach can tackle the problem of predicting structural evolution by surface diffusion, after being trained on a phase-field generated dataset. Furthermore, I'll describe how it was possible to deep-learn the elastic chemical potential at the surface of a stressed
film, allowing us to accurately investigate Stranski-Krastanow island formation at a fraction of the computational cost required by the FEM-based numerical solution. Exciting perspectives, also in terms of extending machine-learning approaches beyond deterministic models, will be finally discussed.
film, allowing us to accurately investigate Stranski-Krastanow island formation at a fraction of the computational cost required by the FEM-based numerical solution. Exciting perspectives, also in terms of extending machine-learning approaches beyond deterministic models, will be finally discussed.
Short bio
Author of more than 150 publications on international journals, FM started using and developing computational methods for simulating diffusion processes and crystal growth already during his PhD in Physics (University of Genoa, 1996-1999), under the supervision of Riccardo Ferrando. He then moved to the Theoretical Division of the Los Alamos National Laboratory, where he worked for two years in the group of Art Voter, helping in developing accelerated molecular dynamics methods. In 2002 he joined the Materials Science Department of the University of Milano-Bicocca as a junior faculty researcher. There, he developed an interest for semiconductor films and heterostructures. Presently he still works in the same Department, but as full Professor in Theoretical Physics of Matter and Dean of the PhD program in Materials Science and Nanotechnology. In the last few years FM embraced the Machine-Learning revolution, epitaxy being the primary focus of application so far.
Author of more than 150 publications on international journals, FM started using and developing computational methods for simulating diffusion processes and crystal growth already during his PhD in Physics (University of Genoa, 1996-1999), under the supervision of Riccardo Ferrando. He then moved to the Theoretical Division of the Los Alamos National Laboratory, where he worked for two years in the group of Art Voter, helping in developing accelerated molecular dynamics methods. In 2002 he joined the Materials Science Department of the University of Milano-Bicocca as a junior faculty researcher. There, he developed an interest for semiconductor films and heterostructures. Presently he still works in the same Department, but as full Professor in Theoretical Physics of Matter and Dean of the PhD program in Materials Science and Nanotechnology. In the last few years FM embraced the Machine-Learning revolution, epitaxy being the primary focus of application so far.
Sentitevi liberi di veicolare l'informazione ai vostri collaboratori e studenti.
Vi aspettiamo. Grazie, Rosaria
-- Rosaria A. Puglisi, PhD Researcher at CNR-IMM Strada VIII n.5 Z.I. 95121 Catania Italy Off: +39 0955968237 https://hq.imm.cnr.it/users/rosariapuglisi https://hq.imm.cnr.it/womeninscience https://hq.imm.cnr.it/articles/il-sicilian-science-cluster ---
Istituto per la Microelettronica e Microsistemi del CNR - Webmaster Nicolò Parasole
