Many natural systems are far from equilibrium, either because having started from a non-typical initial state, they have not yet thermalized, or because they are restlessly driven out of equilibrium through exchanges of matter, energy or information with their surroundings resulting in currents that break time-reversal invariance. In such situations, the principles of equilibrium statistical mechanics do not apply: the theoretical understanding of the dynamical behaviour of many-body systems far from thermal equilibrium is one of the major challenges of contemporary statistical physics and a common trend of many studies carried out in our group.
Current works in this field follow two complementary strategies: one can explore universal features of non-equilibrium systems from a conceptual point of view; or, study some simplified models, classical or quantum, and derive some general understanding. In particular, rigorous studies of stochastic interacting particle processes are a long standing specialty of our group. This important part of our activity yields universal methods that can be applied to investigate the complex dynamical behaviour of biophysical matter (RNA, proteins and viruses) or of emergent artificial systems (communication networks, urban sprawl and cities).
Fluctuation relations, now currently used in experimental settings, establish structural properties of the probability distribution of some physical observable (say work) at a fixed time or in the long time limit.
Non-equilibrum systems can exhibit complex phase diagrams even in low dimensions and phase-coexistence that violate the classical laws of thermodynamics. For example, we have studied generic bistability in two-dimensional totally asymmetric kinetic Ising model, and obtained a complete classification of integrable open boundaries asymmetric simple exclusion processes with two species of particles.
Quantum non equilibrium dynamics exhibit qualitatively different properties from those of classical ones, that can be probed through the study of quantum walks. As a quantum walker propagates ballistically and its wave-function exhibits sharp ballistic fronts, it can easily avoid a static trap, so that it survives forever with non-zero probability (at variance with a classical random walker). Exact studies of model systems often rely on sophisticated theoretical techniques (algebraic integrability, field theory, random matrices). It is therefore natural that some of our works have a strong mathematical flavour, such as the study of record statistics of strongly correlated time series generated by a random walk or a Lévy flight on a line
Spatial networks that describe complex systems under the form of networks where nodes and edges are embedded in space. Transportation and mobility networks, Internet, mobile phone networks, power grids, social and contact networks, neural networks, are all examples where space is relevant and where topology alone does not contain all the information. Our main goal is to understand and model the structure, formation and evolution of these spatial networks with potential applications in different fields ranging from geography, urbanism, to epidemiology and the neurosciences.
Another class of problems that have challenging properties from the statistical physics point of view are disordered systems, in particular glasses and spin glasses. In IPhT we traditionally focus on understanding the theory and properties of super-cooled liquids and structural glasses. These topics were studied in statistical physics for many years, yet the last couple of years brought new enlightening results and development into which researcher in IPhT contributed significantly.
Many problems of interest in computer science related disciplines such as statistical inference, optimization or machine learning bear similarities with disordered systems and the methodology developed in statistical physics of disordered systems leads to remarkable results also in those disciplines. We study solvable models for high-dimensional inference and learning problems, analyse behaviour of associated algorithms, in particular message passing algorithms, study phase transitions in computational problem and their algorithmic consequences.
Our weekly seminar takes place every Monday at 14:00.
Postdoctoral positions are available each year in the Fall. Check this page or contact any staff member of the group.
Each member of the group can be contacted via email at firstname.lastname@example.org .
The full postal adress of IPhT is: Institut de Physique Théorique, CEA/Saclay, Bat 774 Orme des Merisiers, 91191 Gif-sur-Yvette Cedex, France.
Here are directions to the IPhT.