helmut g. katzgraber
    Research

group

 
Helmut G. Katzgraber Assistant professor
Brigitte Surer Graduate student
Ruben Andrist Master thesis project
Francois Gaignat Undergraduate research project

Open positions in the group


research

 
Overview

The advent of fast and cost-effective computers as well as efficient algorithms has made computational physics into a powerful third way, besides experiment and theory, to do research. Disordered and complex systems, my main area of research, have many applications to other disciplines, such as biology, computer science, finance, etc. Due to their complexity, computational physics is the tool of choice to study these because, in general, analytic approaches are only fruitful for the simplest problems.

In the following, I provide a short background to my research in the physics of complex systems, and how I intend to improve our understanding of complex systems such as spin, electron, and structural glasses, as well as cold atomic gases and topologically-protected quantum computing. A key component of my research is developing innovative algorithms in order to achieve my research goals, as well as to study realistic model systems that are not accessible analytically.

Glassy systems

AT line Glassy systems are characterized by disorder and frustration. The most prominent representatives are spin glasses which exhibit high magnetic frustration, as well as structural glasses which are characterized by geometric frustration. Although spin glasses find few direct applications in experimental systems, they are paradigmatic models that deliver concepts relevant for a variety of systems such as optimization problems, disordered magnets and economics applications. However, despite ongoing research spanning several decades in the area of glassy systems, there remain many fundamental open questions.

One of my main research foci over the past few years has been the study of spin glasses where I have studied in detail the nature of the low-temperature phase [1,8], the existence of a spin-glass state in a field [2,3,9] (see figure for possible scenarios — for short-range spin glasses there is no spin glass state in a field), chaos in spin glasses [4], as well as if spin glasses obey universality [5]. In these studies, the analysis of a one-dimensional model with power-law interactions has proven to be very useful in elucidating the properties of higher-dimensional systems because the tuning of the power-law exponent corresponds to effectively changing the space dimension of the system. In addition, I have studied models for vortex glasses (gauge glass) [6] as well as hysteretic properties of disordered magnetic systems [7]. More recently I have branched out to the study of electron as well as structural glasses.

References:

Topologically-protected quantum computing

QDM Besides traditional implementations of quantum bits for quantum computers, there have recently been novel approaches using solid-state devices. In particular, Ioffe et al. have suggested the use of quantum Josephson junction arrays for the construction of a topologically-protected qubit (see figure). Based on the work of Kitaev, this implementation shows good prospects for large decoherence times and scalability; the latter being a crucial problem in common qubit implementations. These implementations exploit topological order, where the different states are characterized by topological properties of the system instead of broken symmetries. Such states have been observed in quantum Hall states, two-dimensional correlated Fermions, as well as Bosonic systems. The main microscopic Hamiltonian with a fractionalized quantum phase is the Rokhsar-Kivelson quantum hard-core dimer model on the triangular lattice. In two dimensions it forms a spin liquid with Z_2 symmetry relevant for quantum computing. Thus it is of paramount importance to better understand the properties of this model, study the robustness of the liquid phase towards perturbations of the Hamiltonian, and search for new models exhibiting these topologically-ordered phases.

Current research focuses on the feasibility of the implementation of a topologically-protected qubit [1] as well as the general study of exotic phases.

References:

Cold atomic gases

density My recent projects in the field of ultracold atomic gases are mainly driven by a productive collaboration with the experimental quantum optics group of T. Esslinger at ETH Zurich. Superimposing optical lattices to Bose-Einstein condensates has enabled the realization of complex quantum phases. Whereas theoretically and numerically exotic quantum phases are being predicted, experimentally they are far from being observed due to the complexity of the experimental setups. In particular, experiments require optical traps that can drastically change the critical behavior of the cold gas being studied. For example, while in a system with no trapping potential criticality has been observed numerically, recent simulations of systems in quadratic confining potentials have shown that the quantum criticality of the superfluid—Mott insulator transition observed in uniform systems is destroyed by the finite gradient of the confining potential. We have performed large-scale quantum Monte Carlo simulations of the two- and three-dimensional Bose-Hubbard model [1] where we show that quartic traps are better suited for experiments than quadratic ones. The figure shows the local density n for quadratic and quartic trap shapes. The central Mott plateau is considerably larger for the quartic trap.

In contrast to bosonic systems, the physics of strongly interacting fermions in 2D and 3D lattices is not yet properly understood. Indeed, accurate numerical simulations on interacting fermionic systems in 2D and 3D lattice models are NP (nondeterministic polynomial) hard, and no general numerical or analytical solution exists. Experiments on ultracold fermionic gases in optical lattices could thus be very useful in elucidating the properties of the fermionic Hubbard model and to probe exotic quantum phases. Recently, we have attempted to understand experimental results by the Esslinger group at ETH Zurich where ultracold fermionic atoms are ramped across a Feshbach resonance. Our results explain the molecule formation rates found and show that current experiments are performed at temperatures considerably higher than expected: lower temperatures are required for fermionic systems to be used as quantum simulators [2].

References:

Algorithms

energy distributions Algorithm development plays a key role in computational physics research. Even with ever-increasing computer power, the study of complex systems, such as glasses in particular, requires efficient algorithms.

My recent research has focused on improving the parallel tempering Monte Carlo method [1] using an optimized ensemble method, simple heuristic ground-state search algorithms for spin glasses [2,3,5], and computing energy distributions of disordered systems to high precision with a novel method using sampling in the disorder with a guiding function [4]. The figure shows a ground-state energy distribution for the mean-field Sherrington Kirkpatrick spin-glass model. While conventional approaches deliver 5 - 6 orders of magnitude, our algorithms extends this range to 16 - 18 orders of magnitude with the same numerical effort.

References:

Statistical mechanics data initiative

Hosted at www.stat-phys.org and initiated by myself, the goal of the statistical mechanics data initiative is to create a data repository along the lines of the arXiv where research groups can upload and publish their raw data sets. Furthermore, test instances for algorithm development will be shared. Stay tuned since the site is under construction!

last update: 06/2007


infrastructure

 
moo 3.0GHz dual quad-core Intel PowerMac (8Gb RAM, 2.25Tb storage) workstation
roo 3.0GHz dual quad-core Intel PowerMac (8Gb RAM, 1.0Tb storage) workstation
poo 1.83GHz dual core Intel Mac Mini (2Gb RAM, 1.1Tb storage) webserver
loo 1.8GHz dual-core Intel MacBook Air (2Gb RAM, 0.08Tb storage) laptop
brutus 128 cores of the ETH brutus cluster


[hardware financed by the Swiss National Science Foundation]