de | en

Hon.-Prof. Ph.D. Ignacio Cirac

Telefon
Raum
E-Mail
gi53rom@mytum.de
Links
Visitenkarte in TUMonline
Arbeitsgruppe
Max-Planck-Institut für Quantenoptik (MPQ)
Funktion
Honorarprofessor am Physik-Department

Lehrveranstaltungen und Termine

Titel und Modulzuordnung
ArtSWSDozent(en)Termine
Quantum Information Methods in Many-Body Physics
Zuordnung zu Modulen:
VO 2 Cirac, I. Schuch, N. Fr, 14:00–17:00
Fr, 14:15–16:15
sowie einzelne oder verschobene Termine

Ausgeschriebene Angebote für Abschlussarbeiten

Applying Tensor Network Methods to Large-Scale Mixed Bosonic/Fermionic Systems using Graphical Processing Units
Motivation: Systems with Bosonic (e.g. phonons) and Fermionic (e.g. electrons) degrees of freedom are generically hard, Fermions induce a sign problem and bosons increase the computational effort required by tensor network methods. At the same time, real materials exhibit coupling between phononic and electronic degrees of freedom which needs to be treated numerically. The Hubbard-Holstein Hamiltonian with strong coupling can serve as a model Hamiltonian. The Hubbard model itself is the subject of ongoing research and few results are available for the Hubbard-Holstein model with strong coupling (e.g. [1,2]). [1] http://link.aps.org/doi/10.1103/PhysRevB.92.241106 [2] https://journals.aps.org/prb/pdf/10.1103/PhysRevB.94.085115 Recently, GPUs have grown sufficiently that porting tensor network methods to run on GPUs appears feasible, but relatively little work has been done in this direction so far. The aim of the project is hence to extend the previous results by including more bosonic degrees of freedom and considering larger 1D and potentially 2D systems with a full tensor network method. Requisites: - second quantisation - basics of condensed matter physics - experience with C++11 programming - ideally also some experience using CUDA, MAGMA or similar technologies Approximate roadmap: - 2-3 months: learning MPS-DMRG, reviewing existing literature on Hubbard-Holstein model - 3-4 months: porting the SYTEN toolkit to make use of GPUs, e.g. by: + introducing a DenseGPUTensor class which resides solely on GPU memory + adapting the Tensor class to use either the DenseTensor or DenseGPUTensor + implementing functionality/wrappers required by MPS-DMRG, e.g.: - tensor reshaping - matrix-matrix multiplication - matrix decompositions: QR/SVD + preliminary benchmarks - 4-5 months: studying the Hubbard-Holstein model with more bosonic degrees of freedom and larger system sizes Supervision is provided jointly by computational experts from the Max Planck Computing and Data Facility and theory department members at MPQ.
geeignet als
  • Masterarbeit Physik der kondensierten Materie
  • Masterarbeit Theoretische und Mathematische Physik
Themensteller(in): Ignacio Cirac
Nach oben