2025 IHES Summer School - Statistical Aspects of Nonlinear Physics

Europe/Paris
Le Bois-Marie

Le Bois-Marie

35, route de Chartres 91440 Bures-sur-Yvette
Description

The Summer School will be held at IHES from  June 23 to July 4, 2025. 

The summer school is a StatPhys29 satellite event

2025 IHES Summer School - Statistical Aspects of Nonlinear Physics

Statistical mechanics, non linear physics and mathematics have often progressed hand in hand, mutually enriching themselves by exchanging key questions and methods. This summer school will be organized around these interactions, with the aim of deepening them and extending them to current research questions. We have identified four active and quickly advancing topics : Random interfaces, Disordered landscapes and AI, Long-range interactions, and Active matter. Each one will be addressed simultaneously by two internationally renowned lecturers, a physicist and a mathematician. This original set-up for the summer school organization will provide two different points of view on the same topic and the tools to bridge them. We aim to bring together attendees from physics and mathematics, to provide them with opportunities to broaden their perspectives, from experimental physics to theoretical physics and mathematics, and to expand their scientific network.

Four courses given jointly by a physicist and a mathematician on the following themes :


This summer school is open to everybody. Priority will be given to PhD students and postdoctoral fellows but applications from more senior researchers are also welcome.

The courses are designed for a mixed audience of physicists and mathematicians. The aim is to provide an introduction to a wide range of topics, and to help students develop their skills and knowledge. 

Deadline for applications: February 23, 2025 


Contact: Cécile Gourgues
    • 1:00 PM 2:00 PM
      Welcome coffee & Registration 1h
    • 2:00 PM 3:30 PM
      Collective Behavior - Active Matter (1/3) 1h 30m
      Speaker: Julien Tailleur (MIT)

      Active materials are driven out of equilibrium at the microscopic scale, where individual units dissipate energy stored in the environment to inject momentum into the system. These energetic and mechanical drives out of thermal equilibrium endow active systems which a rich phenomenology, unmatched in passive systems. In this series of lectures, I will review standard models of active particles and discuss their modeling and collective behaviors. I will show how approximate theories and exact methods combine to shed light on a variety of collective phenomena encountered in active systems, focusing in particular on motility-induced phase separation and the transition to collective motion.

    • 3:30 PM 4:00 PM
      Coffee break 30m
    • 4:00 PM 5:30 PM
      Disordered Landscapes - Exploring the High-dimensional Random Landscapes of Data Science (1/3) 1h 30m
      Speaker: Gérard Ben Arous (New York University)

      Machine learning and Data science algorithms involve in their last stage the need for optimization of complex random functions in very high dimensions. I will briefly survey how these landscapes can be topologically complex, whether this is important or not, and how the usual simple optimization algorithms like Stochastic Gradient Descent (SGD) perform in these difficult contexts. I will mostly rely on joint works with Reza Gheissari (Northwestern), Aukosh Jagganath (Waterloo), Jiaoyang Huang (Wharton).

      I will first introduce the whole framework for non-experts, from the structure of the typical tasks to the natural structures of neural nets used in standard contexts. l will then cover briefly the classical and usual context of SGD in finite dimensions. We will then concentrate first on a simple example, which happens to be close to statistical physics, i.e. the so-called Tensor PCA model, and show how it is related to spherical spin glasses. We will survey the topological properties of this model, and how simple algorithms perform in the basic estimation task of a single spike. We will then move on to the so-called single index models, and then to the notion of "effective dynamics” and “summary statistics”. These effective dynamics, when they exist, run in much smaller dimension and rule the performance of the algorithm. The next step will be to understand how the system finds these “summary statistics”. We will show how this is based on a dynamical spectral transition: along the trajectory of the optimization path, the Gram matrix or the Hessian matrix develop outliers which carry these effective dynamics. I will naturally first come back to the Random Matrix tools needed here (the behavior of the edge of the spectrum and the BBP transition) in a much broader context. We will illustrate this with a few examples from ML: multilayer neural nets for classification (of Gaussian mixtures), and the XOR examples, for instance, or from Statistics like the multi-spike Tensor PCA (from a recent joint work with Cedric Gerbelot (Courant) and Vanessa Piccolo (ENS Lyon)).

    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Collective Behavior - Active Matter (2/3) 1h 30m
      Speaker: Julien Tailleur (MIT)

      Active materials are driven out of equilibrium at the microscopic scale, where individual units dissipate energy stored in the environment to inject momentum into the system. These energetic and mechanical drives out of thermal equilibrium endow active systems which a rich phenomenology, unmatched in passive systems. In this series of lectures, I will review standard models of active particles and discuss their modeling and collective behaviors. I will show how approximate theories and exact methods combine to shed light on a variety of collective phenomena encountered in active systems, focusing in particular on motility-induced phase separation and the transition to collective motion.

    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Disordered Landscapes - Exploring the High-dimensional Random Landscapes of Data Science (2/3) 1h 30m
      Speaker: Gérard Ben Arous (New York University)

      Machine learning and Data science algorithms involve in their last stage the need for optimization of complex random functions in very high dimensions. I will briefly survey how these landscapes can be topologically complex, whether this is important or not, and how the usual simple optimization algorithms like Stochastic Gradient Descent (SGD) perform in these difficult contexts. I will mostly rely on joint works with Reza Gheissari (Northwestern), Aukosh Jagganath (Waterloo), Jiaoyang Huang (Wharton).

      I will first introduce the whole framework for non-experts, from the structure of the typical tasks to the natural structures of neural nets used in standard contexts. l will then cover briefly the classical and usual context of SGD in finite dimensions. We will then concentrate first on a simple example, which happens to be close to statistical physics, i.e. the so-called Tensor PCA model, and show how it is related to spherical spin glasses. We will survey the topological properties of this model, and how simple algorithms perform in the basic estimation task of a single spike. We will then move on to the so-called single index models, and then to the notion of "effective dynamics” and “summary statistics”. These effective dynamics, when they exist, run in much smaller dimension and rule the performance of the algorithm. The next step will be to understand how the system finds these “summary statistics”. We will show how this is based on a dynamical spectral transition: along the trajectory of the optimization path, the Gram matrix or the Hessian matrix develop outliers which carry these effective dynamics. I will naturally first come back to the Random Matrix tools needed here (the behavior of the edge of the spectrum and the BBP transition) in a much broader context. We will illustrate this with a few examples from ML: multilayer neural nets for classification (of Gaussian mixtures), and the XOR examples, for instance, or from Statistics like the multi-spike Tensor PCA (from a recent joint work with Cedric Gerbelot (Courant) and Vanessa Piccolo (ENS Lyon)).

    • 1:00 PM 2:30 PM
      Lunch break (Buffet at IHES) 1h 30m
    • 2:30 PM 4:00 PM
      Collective Behavior - Active Matter (3/3) 1h 30m
      Speaker: Julien Tailleur (MIT)

      Active materials are driven out of equilibrium at the microscopic scale, where individual units dissipate energy stored in the environment to inject momentum into the system. These energetic and mechanical drives out of thermal equilibrium endow active systems which a rich phenomenology, unmatched in passive systems. In this series of lectures, I will review standard models of active particles and discuss their modeling and collective behaviors. I will show how approximate theories and exact methods combine to shed light on a variety of collective phenomena encountered in active systems, focusing in particular on motility-induced phase separation and the transition to collective motion.

    • 4:00 PM 4:30 PM
      Coffee break 30m
    • 4:30 PM 5:30 PM
      Posters 1h

      Marc Besse (Sorbonne Université): Memory Particles: Sensing and Adapting to the Environment
      Abstract: Living systems (e.g. brain, immune system, cell) must sense and respond to the environment to fulfill vital functions, thus displaying some kind of memory. Motivated by the description of cell trajectories, we examine Generalized Langevin Equation (GLE) as a simple yet effective model for agents endowed with memory. A key focus is on quantifying the memory embedded in a GLE and on investigating how such a memory-driven agent behaves both in static environments and in collective settings.

      Pierre Champagnac (Gulliver Lab, ESPCI): Non-linear Memory of a Probe in a Micellar Solution
      Abstract: Microrheology probes complex fluids with colloidal particles. However, the colloidal dynamics may be richer than what is expected from the macroscopic rheology of the fluid. For example, the recoil of a colloid after being dragged through a micellar solution that follows the Maxwell model unveiled a second relaxation time. At small driving velocity, the recoil is captured by a model where the colloid is coupled to two fictitious ``bath particles'' with harmonic springs. Each bath particle is characterized by its friction coefficient and the stiffness of its coupling to the colloid. This model being linear, it cannot capture the observed non-linear effects such as the saturation of the recoil at higher driving velocity, the confinement dependence of the effective memory kernel or the shear-thinning. We propose a model where the particle is coupled to ``bath modes''. Each mode is characterized by its friction coefficient, the stiffness of its coupling with the colloid, as a bath particle, and its wavenumber k. It reduces to a bath particle in the limit k -> 0. The dynamics of each mode can be integrated to obtain a non-linear non-Markovian dynamics for the colloid. Using analytical calculations and numerical simulations, we investigate whether a quantitative agreement with all the experiments can be obtained with a single set of parameters. Our results would guide the design of a more realistic microscopic model for this complex fluid. 

      Samantha Fournier (Institut de Physique Théorique, CEA Saclay): Learning with High Dimensional Chaotic Systems
      Abstract: Chaotic dynamics can naturally arise in high-dimensional heterogeneous systems of interacting variables. The simplest examples are random recurrent neural networks. I will discuss how to study simplified models of this kind through dynamical mean field theory (DMFT) and show that the corresponding chaotic dynamics can be tuned and shaped by synaptic connections to perform a set of interesting tasks. I will show how DMFT can be used to explore and describe the space of synaptic connections leading to good performances of the corresponding trained dynamical systems.

      Eirini Ioannou (Heriot-Watt University Edinburgh): Transfer Operators for Mean-field Stochastic Differential Equations
      Abstract: Mean-field models are interacting particle systems (IPS) in which all interactions to any one particle is replaced with an average or effective interaction. The goal is to develop a data-driven method for inferring global dynamics and meta-stable states of mean-field stochastic differential equations (SDEs) via transfer operators. More specifically, the algorithm used is the EDMD algorithm which estimates the Koopman and Perron–Frobenius operators. As these transfer operators are infinite dimensional, in order to be estimated, they need to be projected onto a finite-dimensional space using Galerkin projection. The evolution of observables of the system is governed by the Koopman operator, while the Perron–Frobenius operator describes the evolution of densities. Using the eigenfunctions of these operators, one can infer the meta-stable states of the underlying system.

      Ruan Miranda (Federal University of Bahia): Consensus Formation in the Multivariate Deffuant Model with Moderates and Extremists
      Abstract: Understanding the mechanisms that govern consensus formation is a central topic in probability theory and its applications to social, behavioral, and economic systems. The Deffuant model offers a probabilistic framework for studying how opinions evolve through pairwise interactions conditioned on a confidence bound. In this work, we investigate a multivariate extension of the Deffuant model defined on the complete graph, where agents' opinions are represented as points uniformly distributed in the unit square. We fix the convergence parameter $\mu=0.5$ and vary the interaction threshold $\tau$, which determines whether two agents interact based on their opinion distance. Our analysis assesses the role of moderates — agents with opinions inside the opinion spectrum — and extremists in achieving consensus. Through extensive numerical simulations, we identify a critical proportion of moderates required to achieve consensus under different tolerance regimes, suggesting that in populations with a high tolerance to interact, even a small number of moderates can significantly impact the speed of achieving agreement.

      Sakshi Pahujani (University of Cologne): Theory of Adaptation to a Moving Optimum
      Abstract: Phenotypic adaptation to long term persistent changes in the environment is typically studied using moving optimum models which consider a fitness function traversing the phenotypic space over time. Invoking the strong-selection-weak-mutation regime, we study adaptation as a walk towards this moving fitness optimum described by a continuous-state-discrete-time stochastic process. In this framework, at a critical speed of the optimum, we elucidate a transition from a regime where the phenotypic gap of the adapting population from the optimum attains a stationary mean value to one where it increases indefinitely. Through a special case of the problem, we provide an alternative description of this transition in terms of a force that drives the adaptive process and a corresponding potential that switches from being confining to non-confining at the critical speed. Further analysis of this case suggests that adaptation is rather successfully carried out, until the population becomes limited by the maximum rate at which it can adapt. Remarkably, despite the simplicity of this special case, its predictions align well with observations from the original, more complex model.

    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Disordered Landscapes - Generative AI and Diffusion Models: a Statistical Physics Analysis (1/3) 1h 30m
      Speaker: Giulio Biroli (LPENS)

      We will first present « diffusion models » which are nowadays the state of the art methods used to generate images, videos and sounds. They are very much related to ideas developed in stochastic thermodynamics, and based on time-reversing stochastic processes. The outcome of these methods is a Langevin process that generates from white noise (the initial condition) new images, videos and sounds. We will show that tools of statistical physics allow to characterise two main phenomena emerging during the Langevin generative diffusion process. The first one, that we call 'speciation’ transition, is where the gross structure of data is unraveled, through a mechanism similar to symmetry breaking in phase transitions. The second phenomenon is the generalisation-memorisation transition, which turns out to be related to the glass transition of Derrida’s random energy model. We will present analytical solutions for simple models and show numerical experiments on real datasets which validate the theoretical analysis.

    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Collective Behavior - Collective Motions with Congestion (1/3) 1h 30m
      Speaker: Bertrand Maury (LMO)

      Collective motions commonly refer to swarming phenomena, mimetism tendencies, collaboration behavior … We propose here to adopt a different standpoint, and to consider situations where interactions mainly stem from congestion principles. We shall in particular present some crowd motion models based on the assumption that entities have individual tendencies, which are counteracted by the presence of others, and investigate according to which principles « information » is propagated though the population. We will pay a special attention to models of the gradient flow type, which are built on the assumption that a given population tends to minimize some sort of global dissatisfaction (sum of individual dissatisfactions), together with an additional term which encodes the congestion contraints. We will present in details a microscopic version of this approach, together with macroscopic extensions, based on Optimal transportation principles. We shall also investigate how this framework, that is based on « classical mechanics » principles, and which carries in particular an Action-and-Reaction principle (symmetric interactions), can be distorted to elaborate more realistic interaction models, with asymmetric interactions between individuals.

      Microscopic model : 
      B. Maury, J. Venel, A discrete Contact Model for crowd Motion,  ESAIM: M2AN 45 1 (2011) 145-168.
       
      B. Maury, non smooth evolution models in crowd dynamics: mathematical and numerical issues, in Collective Dynamics from Bacteria to Crowds, An Excursion Through Modeling, Analysis and Simulation, Series: CISM International Centre for Mechanical Sciences, Vol. 553 (2014)
       
      S. Faure, B. Maury, Crowd  motion from the granular standpoint, Mathematical Models and Methods in Applied Sciences Vol. 25, No. 3 (2015) 463–493
       
      Macroscopic model (optimal transport setting)
      B. Maury, A. Roudneff-Chupin, F. Santambrogio,  A macroscopic Crowd Motion Model of the gradient-flow type, Mathematical Models and Methods in Applied Sciences Vol. 20, No. 10 (2010) 1787-1821
       
      Comparison micro-macro
      B. Maury, A. Roudneff-Chupin, F. Santambrogio, J. Venel,  Handling Congestion in Crowd Motion Modeling,  Networks and Heterogeneous Media, Volume 6, Number 3, September 2011, pp. 485–519
       
      B Maury, Grains de foules, Gazette des Mathématiciens No 152, avril 2017 (In French)
       
      Opinion propagation on a network
      H. Lavenant, B. Maury,  Opinion propagation on social networks : a mathematical standpoint, ESAIM: Proceedings and Surveys, 2020, Vol. 67, p. 285-335
       
    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:10 PM
      Short talks 1h 40m

      Amer Al-Hiyasat (MIT): A Cold Tracer in a Hot Bath: In and Out of Equilibrium
      Abstract: We study the dynamics of a zero-temperature overdamped tracer in a bath of Brownian particles. As the bath density is increased, the passive tracer transitions from an effectively active dynamics, characterized by boundary accumulation and ratchet currents, to a bona-fide equilibrium regime. To account for this, we eliminate the bath degrees of freedom under the assumption of linear coupling to the tracer and show convergence to equilibrium in the large density limit. We then develop a perturbation theory to characterize the tracer’s departure from equilibrium at large but finite bath densities, revealing an intermediate time-reversible yet non-Boltzmann regime, followed by a fully irreversible one. Finally, we show that when the bath particles are connected as a lattice, mimicking a gel, the cold tracer drives the entire bath out of equilibrium, leading to a long-ranged suppression of bath fluctuations.

      Anwar El Rhirhayi (Université d'Orléans): Collisional Kinetics and Large Deviation for Long-Range Interacting Systems
      Abstract: We study the slow collisional relaxation of systems with long-range interactions, where individual forces are weak but correlations accumulate over large timescales. At the kinetic level, this leads to the Landau equation, which describes the average evolution of the velocity distribution due to frequent, small-angle interactions. To resolve fluctuations around this mean-field behavior, we introduce a Langevin formulation that captures both the collective noise and the conservation constraints. We then apply large deviation theory to quantify rare events and derive an action functional whose minimizers correspond to the deterministic Landau dynamics. This approach connects microscopic stochastic dynamics with macroscopic transport and provides a framework for understanding fluctuations in long-range systems.

      Mathis Guéneau (Sorbonne Université - LPTHE): Large Deviations in Switching Diffusion: from Free Cumulants to Dynamical Transitions
      Abstract: We study the diffusion of a particle in a disordered medium. The disorder is effectively accounted for through the time-dependent diffusion coefficient $D(t)$, which follows a switching process. Specifically, $D(t)$ takes a constant value drawn from a distribution $W(D)$ for a random duration, then switches to another value for another random period. This process repeats at every subsequent renewal time. This model belongs to the class of “random diffusivity models”, introduced to describe diffusive processes where $\langle x^2(t) \rangle \sim 2Dt$, but with non-Gaussian distributions. Such models are widely observed in nature, including protein diffusion. We derive exact expressions for the time-dependent moments $\langle x^n(t) \rangle$ and demonstrate that, in the long-time limit, the cumulants $\langle x^n(t) \rangle_c$ are given by the free cumulants of a random variable distributed according to $W(D)$. This unexpected link to free probability theory opens new avenues for understanding the mathematical structure of such processes. Furthermore, for specific forms of $W(D)$, we compute the exact large deviations of the position distribution, uncovering rich behaviors and dynamical transitions.

      Jessica Metzger (MIT): Revisiting the Ratchet Principle: When Hidden Symmetries Prevent Steady Currents
      Abstract: The "ratchet principle", which states that non-equilibrium systems violating parity symmetry generically exhibit steady-state currents, is one of the few generic results outside thermal equilibrium. In this talk, I analyze exceptions to this principle observed in active and passive systems with spatially varying fluctuations sources. For dilute systems, I will show that a hidden time-reversal symmetry prevents the emergence of ratchet currents. At higher densities, pairwise interactions break this symmetry, but an emergent conservation law for the momentum may nevertheless prevent steady currents. I will show how the presence of this conservation law can be tested analytically and characterize the onset of ratchet currents in its absence. Our results reveal that the ratchet principle should be amended to preclude parity symmetry, time-reversal symmetry, and bulk momentum conservation.

      Charlotte Myin (Max Planck Institut for Dynamics and Self-Organization): The Statistical Physics of Nonreciprocal Flocking
      Abstract: Recently, nonreciprocal interactions have gained growing interest in active matter physics. To study the effect of nonreciprocal interactions on collective motion, we extend the Vicsek model, a paradigmatic model of flocking, to multiple species. Here, particles of the same species align, while interspecies interactions can be nonreciprocal. Through simulations of the microscopic model and linear stability analysis of coarse-grained hydrodynamic equations, we show that nonreciprocal interactions generically destabilize two-species flocks in a large part of the phase diagram. This instability leads to a new phase where one one the species condenses into a dense travelling band.

       

    • 4:10 PM 4:30 PM
      Coffee break 20m
    • 4:30 PM 5:30 PM
      Modelling Dense Crowds with Mean-Field Games 1h
      Speaker: Cécile Appert (IJCLab, Université Paris-Saclay)
      Joint work with Denis Ullmo (LPTMS)

      Game theory allows to model how agents can optimize their strategy in a competitive situation. In high density crowds, pedestrians compete for space. When the number of agents becomes large, the problem can be made tractable through a mean-field hypothesis [1]. Some experiments of crowd deformation by an intruder [2] have revealed that, even at densities as high as 6 ped/m2 , pedestrians anticipate the passage of the intruder in a way that none of the existing crowd models of that time could capture. In particular they behave very differently from granular matter. Crowd modeling based on Mean-Field Games (MFG) allowed for the first time to reproduce these experimental results [3], both for the density and displacement fields. In particular, the tuning of a single parameter - the anticipation horizon - allowed to account for several experimental conditions in which pedestrians have a variable level of information [4]. But it also raises numerical issues that will be discussed [5]. REFERENCES [1] J.-M. Lasry and P.-L. Lions. Mean field games. I – the stationary case. C. R. Acad. Sci. Paris, Ser. I 343 (2006) 619–625. [2] A. Nicolas, M.N. Kuperman, S. Ibáñez, S. Bouzat, and C. Appert-Rolland. Mechanical response of dense pedestrian crowds to the crossing of intruders. Sci. Rep. 9 (2019) 105. [3] T. Bonnemain, M. Butano, T. Bonnet, I. Echeverría-Huarte, A. Seguin, A. Nicolas, C. Appert-Rolland, and D. Ullmo. Pedestrians in static crowds are not grains, but game players. Phys. Rev. E 107 (2023) 024612. [4] M. Butano, C. Appert-Rolland, and D. Ullmo. Discounted mean-field game model of a dense static crowd with variable information crossed by an intruder. SciPost Phys. 16 (2024) 104. [5] C. Appert-Rolland, M. Butano, and D. Ullmo. Mean-field games for high-density crowds: the discount factor reproduces experiments but raises numerical issues. To appear in the Proceedings of TGF'24.
    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Collective Behavior - Collective Motions with Congestion (2/3) 1h 30m
      Speaker: Bertrand Maury (LMO)

      Collective motions commonly refer to swarming phenomena, mimetism tendencies, collaboration behavior … We propose here to adopt a different standpoint, and to consider situations where interactions mainly stem from congestion principles. We shall in particular present some crowd motion models based on the assumption that entities have individual tendencies, which are counteracted by the presence of others, and investigate according to which principles « information » is propagated though the population. We will pay a special attention to models of the gradient flow type, which are built on the assumption that a given population tends to minimize some sort of global dissatisfaction (sum of individual dissatisfactions), together with an additional term which encodes the congestion contraints. We will present in details a microscopic version of this approach, together with macroscopic extensions, based on Optimal transportation principles. We shall also investigate how this framework, that is based on « classical mechanics » principles, and which carries in particular an Action-and-Reaction principle (symmetric interactions), can be distorted to elaborate more realistic interaction models, with asymmetric interactions between individuals.

      Microscopic model : 
      B. Maury, J. Venel, A discrete Contact Model for crowd Motion,  ESAIM: M2AN 45 1 (2011) 145-168.
       
      B. Maury, non smooth evolution models in crowd dynamics: mathematical and numerical issues, in Collective Dynamics from Bacteria to Crowds, An Excursion Through Modeling, Analysis and Simulation, Series: CISM International Centre for Mechanical Sciences, Vol. 553 (2014)
       
      S. Faure, B. Maury, Crowd  motion from the granular standpoint, Mathematical Models and Methods in Applied Sciences Vol. 25, No. 3 (2015) 463–493
       
      Macroscopic model (optimal transport setting)
      B. Maury, A. Roudneff-Chupin, F. Santambrogio,  A macroscopic Crowd Motion Model of the gradient-flow type, Mathematical Models and Methods in Applied Sciences Vol. 20, No. 10 (2010) 1787-1821
       
      Comparison micro-macro
      B. Maury, A. Roudneff-Chupin, F. Santambrogio, J. Venel,  Handling Congestion in Crowd Motion Modeling,  Networks and Heterogeneous Media, Volume 6, Number 3, September 2011, pp. 485–519
       
      B Maury, Grains de foules, Gazette des Mathématiciens No 152, avril 2017 (In French)
       
      Opinion propagation on a network
      H. Lavenant, B. Maury,  Opinion propagation on social networks : a mathematical standpoint, ESAIM: Proceedings and Surveys, 2020, Vol. 67, p. 285-335
       
    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Disordered Landscapes - Generative AI and Diffusion Models: a Statistical Physics Analysis (2/3) 1h 30m
      Speaker: Giulio Biroli (LPENS)

      We will first present « diffusion models » which are nowadays the state of the art methods used to generate images, videos and sounds. They are very much related to ideas developed in stochastic thermodynamics, and based on time-reversing stochastic processes. The outcome of these methods is a Langevin process that generates from white noise (the initial condition) new images, videos and sounds. We will show that tools of statistical physics allow to characterise two main phenomena emerging during the Langevin generative diffusion process. The first one, that we call 'speciation’ transition, is where the gross structure of data is unraveled, through a mechanism similar to symmetry breaking in phase transitions. The second phenomenon is the generalisation-memorisation transition, which turns out to be related to the glass transition of Derrida’s random energy model. We will present analytical solutions for simple models and show numerical experiments on real datasets which validate the theoretical analysis.

    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:10 PM
      Short talks 1h 40m

      Hong-Bin Chen (IHES): The Convex Structure of the Parisi Formula for Multi-species Spin Glasses
      Abstract: We study the free energy of mean-field multi-species spin glass models. For such models wtih $D$ species, the Parisi formula is known to be valid, and expresses the limit free energy as a infimum over monotone probability measures on ${\mathbb R}_+^D$. We show here that one can transform this representation into a infimum over all probability measures on ${\mathbb R}_+^D$ of a convex functional. We then deduce that the Parisi formula admits a unique minimizer. Using convex-duality arguments, we also obtain a new representation of the free energy as an supremum over martingales in a Wiener space. Based on a joint work in preparation with Victor Issa and Jean-Christophe Mourrat.

      Arnaud Le Ny (LAMA UPEC & CNRS): Metastates for Long-range Ising Model with Random Boundary Conditions
      Abstract: Spin systems with random boundary conditions are among the simple examples of disordered systems, related to spin-glass models by Mattis transformation and disorder. We show that for polynomially decaying long range ising models the Metastates defined as distributionnal limit do exist and exhibit a transition from concentrated, dispersed or fully dispersed depending on the decay of the ising coupling. Joint work under revision with Eric Ossami Endo (NYU Shanghai) and Aernout van Enter (RU Groningen).

      Francesco Tosello (The University of British Columbia): Modelling Structured Data Learning with Restricted Boltzmann Machines in the Teacher-Student Setting
      Abstract: Restricted Boltzmann machines (RBM) are generative models capable of learning data with a rich underlying structure. We study the Teacher-Student setting, where a student RBM learns structured synthetic data generated by a teacher RBM. The amount of structure in the data is controlled by the number of hidden units of the teacher and the correlations between its weights (a.k.a. patterns).
      In the absence of correlations, we validate the conjecture that the performance is independent of the number of teacher patters and hidden units of the student RBMs. Beyond this regime, we find that the critical amount of data required to learn the teacher's patterns decreases with both their number and correlations. In both regimes, we show that, even with a relatively large dataset, it becomes impossible to learn the teacher's patterns if the inference temperature used for regularization is too low. In our framework, the student can learn teacher patterns one-to-one or many-to-one, generalizing previous findings about the Teacher-Student setting with two hidden units to any arbitrary finite number of hidden units.

      Léo Touzo (LPENS): Exact Results for Active Particles with Long-range Interactions
      Abstract: Active particles are objects or living organisms which possess some form of self-propulsion. When a large number of such particles interact together, surprising collective effects may emerge. However, the fact that the noise driving these systems is time-correlated makes exact analytical studies extremely challenging. I will present several exact results for such models in one dimension, focusing on long-range pairwise potential interactions. I will first explain how the non-equilibrium stationary density in models of run-and-tumble particles (RTPs) with long-range interactions can be studied using an extension of the Dean-Kawasaki equation. In the case of the 1D Coulomb interaction (attractive or repulsive), we obtained exact expressions for the stationary density for different types of confining potentials, which sheds lights on new non-equilibrium phase-transitions. Some results were also obtained for a repulsive 2D Coulomb interaction (log-gas), although the single-file constraint makes the study more difficult in this case. I will then focus on the fluctuations at the tagged particle level, for a gas of active particles on the circle with a generic Riesz (i.e. power law) interaction. In the limit of weak noise, we computed exactly and analyzed in different regimes a variety of correlation functions of the particle positions and interparticle distances, both for the Brownian Riesz gas and for its active counterpart, and showed that the activity plays an important role both at short times and at small distances.

      Thomas Tulinski (LPENS): Spherical Boltzmann Machines
      Abstract: We present the analytical solution to an idealized generative model, the spherical Boltzmann machine, trained on a fixed realization of the data. Depending on the level of noise during training, and on the intensity of the regularization on the weights, the machine can be found in a variety of condensed states after training, states which we characterize analytically. The solution, obtained with the replica method at a finite negative real number of replicas, involves various forms of replica-symmetry-breaking. This solution which can then be found back without replicas using spherical integrals.

    • 4:10 PM 4:30 PM
      Coffee break 20m
    • 4:30 PM 5:30 PM
      Cromosim : An Open Source Software for Crowd Motion Simulations 1h
      Speaker: Sylvain Faure (LMO)

      Cromosim (www.cromosim.fr)  is an open source software written in Python, it contains implementations of most microscopic crowd motion models : mainly Helbing’s Social Force Model, Cellular Automata, and granular models. This session  will give an overview of Cromosim’s main features, including the possibility to run evacuation scenarios  in non trivial buildings. Participant are encouraged to install the package ('pip install cromosim') in advance of the session.

      Cromosim  Software
       
      Important : For those who would like to run cromosim on a Windows environnement  : 
      you'll need a compiler to install a CroMoSim dependency(scikit-fmm). You have to install « Visual Studio Community» first (with itsC/C++ and Python environments).
       
    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Disordered Landscapes - Exploring the High-dimensional Random Landscapes of Data Science (3/3) 1h 30m
      Speaker: Gérard Ben Arous (New York University)

      Machine learning and Data science algorithms involve in their last stage the need for optimization of complex random functions in very high dimensions. I will briefly survey how these landscapes can be topologically complex, whether this is important or not, and how the usual simple optimization algorithms like Stochastic Gradient Descent (SGD) perform in these difficult contexts. I will mostly rely on joint works with Reza Gheissari (Northwestern), Aukosh Jagganath (Waterloo), Jiaoyang Huang (Wharton).

      I will first introduce the whole framework for non-experts, from the structure of the typical tasks to the natural structures of neural nets used in standard contexts. l will then cover briefly the classical and usual context of SGD in finite dimensions. We will then concentrate first on a simple example, which happens to be close to statistical physics, i.e. the so-called Tensor PCA model, and show how it is related to spherical spin glasses. We will survey the topological properties of this model, and how simple algorithms perform in the basic estimation task of a single spike. We will then move on to the so-called single index models, and then to the notion of "effective dynamics” and “summary statistics”. These effective dynamics, when they exist, run in much smaller dimension and rule the performance of the algorithm. The next step will be to understand how the system finds these “summary statistics”. We will show how this is based on a dynamical spectral transition: along the trajectory of the optimization path, the Gram matrix or the Hessian matrix develop outliers which carry these effective dynamics. I will naturally first come back to the Random Matrix tools needed here (the behavior of the edge of the spectrum and the BBP transition) in a much broader context. We will illustrate this with a few examples from ML: multilayer neural nets for classification (of Gaussian mixtures), and the XOR examples, for instance, or from Statistics like the multi-spike Tensor PCA (from a recent joint work with Cedric Gerbelot (Courant) and Vanessa Piccolo (ENS Lyon)).

    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Collective Behavior - Collective Motions with Congestion (3/3) 1h 30m
      Speaker: Bertrand Maury (LMO)

      Collective motions commonly refer to swarming phenomena, mimetism tendencies, collaboration behavior … We propose here to adopt a different standpoint, and to consider situations where interactions mainly stem from congestion principles. We shall in particular present some crowd motion models based on the assumption that entities have individual tendencies, which are counteracted by the presence of others, and investigate according to which principles « information » is propagated though the population. We will pay a special attention to models of the gradient flow type, which are built on the assumption that a given population tends to minimize some sort of global dissatisfaction (sum of individual dissatisfactions), together with an additional term which encodes the congestion contraints. We will present in details a microscopic version of this approach, together with macroscopic extensions, based on Optimal transportation principles. We shall also investigate how this framework, that is based on « classical mechanics » principles, and which carries in particular an Action-and-Reaction principle (symmetric interactions), can be distorted to elaborate more realistic interaction models, with asymmetric interactions between individuals.

      Microscopic model : 
      B. Maury, J. Venel, A discrete Contact Model for crowd Motion,  ESAIM: M2AN 45 1 (2011) 145-168.
       
      B. Maury, non smooth evolution models in crowd dynamics: mathematical and numerical issues, in Collective Dynamics from Bacteria to Crowds, An Excursion Through Modeling, Analysis and Simulation, Series: CISM International Centre for Mechanical Sciences, Vol. 553 (2014)
       
      S. Faure, B. Maury, Crowd  motion from the granular standpoint, Mathematical Models and Methods in Applied Sciences Vol. 25, No. 3 (2015) 463–493
       
      Macroscopic model (optimal transport setting)
      B. Maury, A. Roudneff-Chupin, F. Santambrogio,  A macroscopic Crowd Motion Model of the gradient-flow type, Mathematical Models and Methods in Applied Sciences Vol. 20, No. 10 (2010) 1787-1821
       
      Comparison micro-macro
      B. Maury, A. Roudneff-Chupin, F. Santambrogio, J. Venel,  Handling Congestion in Crowd Motion Modeling,  Networks and Heterogeneous Media, Volume 6, Number 3, September 2011, pp. 485–519
       
      B Maury, Grains de foules, Gazette des Mathématiciens No 152, avril 2017 (In French)
       
      Opinion propagation on a network
      H. Lavenant, B. Maury,  Opinion propagation on social networks : a mathematical standpoint, ESAIM: Proceedings and Surveys, 2020, Vol. 67, p. 285-335
       
    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:00 PM
      Disordered Landscapes - Generative AI and Diffusion Models: a Statistical Physics Analysis (3/3) 1h 30m
      Speaker: Giulio Biroli (LPENS)

      We will first present « diffusion models » which are nowadays the state of the art methods used to generate images, videos and sounds. They are very much related to ideas developed in stochastic thermodynamics, and based on time-reversing stochastic processes. The outcome of these methods is a Langevin process that generates from white noise (the initial condition) new images, videos and sounds. We will show that tools of statistical physics allow to characterise two main phenomena emerging during the Langevin generative diffusion process. The first one, that we call 'speciation’ transition, is where the gross structure of data is unraveled, through a mechanism similar to symmetry breaking in phase transitions. The second phenomenon is the generalisation-memorisation transition, which turns out to be related to the glass transition of Derrida’s random energy model. We will present analytical solutions for simple models and show numerical experiments on real datasets which validate the theoretical analysis.

    • 4:00 PM 4:30 PM
      Coffee break 30m
    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Long Range Interactions - Systems with Coulomb and Riesz Interactions (1/3) 1h 30m
      Speaker: Sylvia Serfaty (Sorbonne Université & Courant Institute of Mathematical Sciences)

      We will focus on the statistical mechanics of large systems of particles with Coulomb or Riesz long-range repulsion, analyzed via an electrostatic formulation. Topics covered include: equilibrium measure, expansion of the free energy, limit point processes, fluctuations.

      References:
      Lectures on Coulomb and Riesz gases”, https://arxiv.org/abs/2407.21194

    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Random Interfaces - Introduction to the Physics of the KPZ Universality Class (1/3) 1h 30m
      Speaker: Kazumasa Takeuchi (The University of Tokyo)

      The Kardar-Parisi-Zhang (KPZ) class is a prominent universality class known to describe a surprising variety of models and phenomena, including growing interfaces, directed polymers in random media, stochastic particle transport, and most recently even quantum systems such as the exciton-polariton condensate and integrable spin chains. Remarkably, the 1D KPZ class is exactly solvable in many aspects, opening a window to detailed statistical properties through its links to various mathematical problems. The lecture will include a pedagogical introduction to basic properties of the KPZ equation, an illustrative outline of characteristic distribution and correlation properties revealed by exact solutions, and a survey of experimental studies on growing interfaces.
      1. Introduction: why should we care this?
      2. Scaling exponents and universality classes
      3. Basic properties of the KPZ equation
      4. Distribution and correlation properties: stationary & non-stationary cases
      5. Experimental test of predictions from integrable models
      6. Distribution properties for general cases and variational formula

      Main reference: K. A. Takeuchi, "An appetizer to modern developments on the Kardar-Parisi-Zhang universality class", Physica A 504, 77-105 (2018).

      Slides of the lectures : https://bit.ly/ihes2025

    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:00 PM
      Long Range Interactions - Nonequilibrium Point Processes with Long-range Correlations Generated by Stochastic Resetting (1/3) 1h 30m
      Speaker: Satya N. Majumdar (LPTMS)

      In this course, I'll first discuss briefly the equilibrium systems with long-range correlations (such as the log-gas or the Riesz gas in coordination with the course by S. Serfaty). Then the main emphasis of my course would be on nonequilibrium stationary states of many-body systems with attractive long-range correlations. I will provide examples of how such states can be generated by stochastic resetting of N diffusing particles in one or higher dimensions. The joint distribution of the positions of the particles, though not factorizable due to strong correlations, has a certain conditionally independent structure that allows them to be exactly solvable for various physical obsevables. For example, in one dimension and in the large N limit, one can compute exactly, the average density, the extreme value statistics, the spacing distribution between particles as well as the full counting statistics. The presence of long-range correlations make these observables drastically different from the equilibrium ideal gas where the joint distribution in the stationary state factorizes.

      References:

      [1] M. Biroli, H. Larralde, S. N. Majumdar, G. Schehr, Extreme Statistics and Spacing Distribution in a Brownian Gas Correlated by Resetting, Phys. Rev. Lett., 130, 207101 (2023) [Editor’s suggestion].

      [2] M. Biroli, S. N. Majumdar, G. Schehr, Critical number of walkers for diffusive search processes with resetting, Phys. Rev. E, 107, 064141 (2023).

      [3] M. Biroli, H. Larralde,, S. N. Majumdar, G. Schehr, Exact extreme, order and sum statistics in a class of strongly correlated system , Phys. Rev. E 109, 014101 (2024).

      [4] M, Biroli, M. Kulkarni, S. N. Majumdar, G. Schehr, Dynamically emergent correlations between particles in a switching harmonic trap, Phys. Rev. E 109, L032106 (2024).

      [5] S. Sabhapandit, S. N. Majumdar, Noninteracting particles in a harmonic trap with a stochastically driven center, J. Phys. A: Math. Theor. 57, 335003 (2024).

      [6] M. Kulkarni, S. N. Majumdar, S. Sabhapandit, Dynamically emergent correlations in bosons via quantum resetting, J. Phys. A: Math. Theor. 58, 105003 (2025).

      [7] M. Biroli, S. N. Majumdar, G. Schehr, Resetting Dyson Brownian motion, https://arxiv.org/abs/2503.14733 (to appear in PRE, 2025).

      [8] M.R. Evans and S.N. Majumdar, Diffusion with stochastic resetting, Phys. Rev. Lett. , 106, 160601 (2011).

      [9] M. R. Evans, S. N. Majumdar, G. Schehr, Stochastic Resetting and Applications, J. Phys. A: Math. Theor. 53, 193001 (2020) (topical review).

      [10] S. N. Majumdar, G. Schehr, Statistics of extremes and records in random sequences (Oxford University Press) (2024).

    • 4:00 PM 4:30 PM
      Coffee break 30m
    • 4:30 PM 5:30 PM
      Posters 1h

      Juan Carlos Arroyave Blanco (IMPA): Universality of NESS for KPZ
      Abstract: We analyze the open $\textbf{WASEP}$($\alpha$), where each site can host up to $\alpha$ particles, coupled to boundary reservoirs fixing the densities $\rho_- = \beta$ and $\rho_+ = \gamma$. Under a suitable reference measure and hydrodynamic scaling, we show, via the relative entropy method, that the macroscopic density profile satisfies the inviscid Burgers-type equation: $\partial_x^2 u - v \partial_x u^2 = 0$. At the diffusive time scale, the associated density fluctuation field is tight, and its trajectories concentrate on energy solutions of the stochastic Burgers equation (SBE): $d\mathscr{Y}_t = \alpha \Delta_{\mathrm{Dir}} \mathscr{Y}_t \, dt + \frac{v\alpha}{2} \nabla_{\mathrm{Dir}}(u \mathscr{Y}_t) \, dt + v \nabla_{\mathrm{Dir}} (\mathscr{Y}_t^2) \, dt + \sqrt{2\chi(\rho)} \nabla_{\mathrm{Dir}} \, d\mathscr{W}_t$, where $\chi(\rho) = \rho(\alpha - \rho)$ and $\mathscr{W}_t$ denotes space-time white noise. This establishes WASEP($\alpha$) as a microscopic foundation for the universality of the SBE in non-equilibrium open systems.

      Ubaldo Cavazos Olivas (University of Warsaw): Ionic polaron and bipolaron in a Bose gas
      Abstract: Ultracold quantum many-body systems constitute an interesting research playground due to their wide range of applications, from precision measurements to transport phenomena in the field of condensed matter. One particular example are hybrid systems of atoms and ions, which are rapidly developing [1] and at ultralow temperature provide an ideal environment for the emergence of polarons. Namely, a quantum bath composed of bosonic atoms weakly coupled to an ion can be properly described by means of Bogoliubov theory. Nevertheless, this approach is no longer valid as soon as the strong coupling regime is taken into account, leading to an instability with an infinite number of bosons collapsing into the ion. Ion-atom systems feature long-range interactions which drive the system to form a many-body bound state with high density and large atom number [2]. In order to explore this physics and circumvent the bosons unstable behavior, based on [3], a variational approach is adopted. Employing a regularized potential that retains the correct long-distance behavior, we study the properties of interest in the formation of ionic Bose polaron and bipolaron, such as their energy, the number of bosons that takes part in the cloud formation, and the induced interactions which are tunable by the potential parameters.

      References:
      [1] Tomza, M. et al. Cold hybrid ion-atom systems. Rev. Mod. Phys. 91, 035001(2019).  
      [2] Astrakharchik, G.E., Ardila, L.A.P., Schmidt, R. et al. Ionic polaron in a Bose-Einstein condensate. Commun Phys 4, 94 (2021).  
      [3] Schmidt, R. and Enss, T. Self-stabilized Bose polarons. SciPost Phys. 13, 054 (2022).

      Adrien Escoubet (Université de Lille): Experimental Measurement of the Dynamical Structure Factor in a Normal Dispersion Fiber Recirculating Loop
      Abstract: The search for universality classes aims to group together many systems that exhibit identical asymptotic behaviors, regardless of their underlying microscopic mechanisms. In this context, Kardar, Parisi, and Zhang (KPZ) introduced their famous equation in 1986 to describe the growth of an interface between two phases [1]. This nonlinear stochastic equation was later found to be connected to the description of various systems within its universality class, such as the Eden model [2] and the free energy distribution in random polymers [3]. In 2004, Prähofer and Spohn developed a discrete growth model that allowed the theoretical determination of the dynamic exponent, which characterizes the evolution of the system’s characteristic length scale over time, yielding z=3/2 [4]. A recent study on a Bose-Einstein condensate with repulsive interatomic interactions hints at the defocusing nonlinear Schrödinger equation (dNLSE) belonging, in the regime of small modulation amplitudes, to the KPZ universality class [5]. The theoretical justifications are based in particular on the Madelung transformation, which maps the dNLSE to the stochastic Burgers equation, itself part of this universality class. The numerical results of the study recover the value of the dynamic exponent z through the computation of the dynamic structure factor, defined as the Fourier transform of spatio-temporal correlations. This is however only a numerical signature, and the conclusions are still up for debate. We develop a nonlinear optical experiment described by the KPZ universality class in a system governed by the dNLSE. The first numerical results show the reproducibility of simulations with parameters suitable for implementation in an optical system. A recirculating fiber loop allows full observation of spatio-temporal dynamics, enabling comparison with the model. We address the issues of spectral resolution and signal-to-noise ratio, which are the main obstacles to experimentally estimating the dynamic structure factor.

      References:
      [1] - M. Kardar, G. Parisi & Y. Zhang, Physical Review Letters, 56, 889-892 (1986).
      [2] - M. Eden, Dynamics of fractal surfaces, 4, 598 (1961).
      [3] - G. Amir, I. Corwin & J. Quastel, Communications on pure and applied mathematics, 64, 466-537 (2011).
      [4] - M. Prähofer & H. Spohn, Journal of Statistical Physics, 115, 255-279 (2004).
      [5] - M. Kulkarni & A. Lamacraft, Physical Review A, 88, 021603 (2013).

      Siddhant Mal (University of Michigan): Coherent Magneto-Conductance Oscillations in Amorphous Topological Insulator Nano-wires
      Abstract: Recent experiments on amorphous materials have established the existence of surface states similar to those of crystalline three-dimensional topological insulators (TIs). Amorphous topological insulators are also independently of interest for thermo-electric and other properties. To develop an understanding of transport in these systems, we carry out quantum transport calculations for a tight-binding model of an amorphous nano-wire pierced by an axial magnetic flux, then compare the results to known features in the case of crystalline models with disorder. Our calculations complement previous studies in the crystalline case that studied the surface or used a Green's function method. We find that the periodicity of the conductance signal with varying magnetic flux is comparable to the crystalline case, with maxima occurring at odd multiples of magnetic flux quanta. However, the expected amplitude of the oscillation decreases with increasing amorphousness, as defined and described in the main text. We characterize this deviation from the crystalline case by taking ensemble averages of the conductance signatures for various wires with measurements simulated at finite temperatures. This striking transport phenomenon offers a metric to characterize amorphous TIs and stimulate further experiments on this class of materials.

      Eduardo Pimenta (Federal University of Bahia): A functional Central Limit Theorem and weak Berry-Esseen Estimates for Non-homogeneous Random Walks
      Abstract: In this work, we establish a Trotter-Kato type theorem. More precisely, we characterize the convergence in distribution of Feller processes by examining the convergence of their generators. The main novelty lies in providing quantitative estimates in the vague topology at any fixed time. As important applications, we deduce functional central limit theorems for random walks on the positive integers with boundary conditions, which converge to Brownian motions on the positive half-line with boundary conditions at zero. 

      Enrique Rozas Garcia (University of Gothenburg): Universal Fragmentation in Annihilation Reactions with Constrained Kinetics
      Abstract: In reaction-diffusion models of annihilation relations, the late-time evolution to a final empty state is independent of the initial state of the system. This universal behaviour may be attributed to the diffusive dynamics allowing the complete exploration of the space of available states. In this poster, I discuss a reaction model where the exploration is hindered by constraining the dynamics to preserve the centre of mass.  With such constraints, the system does not evolve into an empty state but rather freezes into fragmented particle clusters. The late-time dynamics and final density are universal, and I discuss exact results for the final density in the large-reaction rate limit. This setup constitutes a minimal model for the fragmentation of a one-dimensional lattice into independent particle clusters.

    • 9:00 AM 9:30 AM
      Coffee break 30m
    • 9:30 AM 11:00 AM
      Random Interfaces - Lecture 1: Extreme Diffusion; or Was Einstein Wrong About Diffusion? 1h 30m
      Speaker: Ivan Corwin (Columbia University)

      In a system of many particles diffusing in a common environment, the first few particles often have outsized importance. How do they behave and what does that behavior tell us about the environment in which they have evolved? We will approach these problems by studying random walks in random environments and utilizing a connection with the Kardar-Parisi-Zhang stochastic PDE and universality class. I will describe recent work which uncovers new power-laws beyond those of Einstein's theory of diffusion and introduces the Extreme Diffusion Coefficient that captures new microscopic information about the environment.

      References and slides for the lectures

       
    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Long Range Interactions - Systems with Coulomb and Riesz Interactions (2/3) 1h 30m
      Speaker: Sylvia Serfaty (Sorbonne Université & Courant Institute of Mathematical Sciences)

      We will focus on the statistical mechanics of large systems of particles with Coulomb or Riesz long-range repulsion, analyzed via an electrostatic formulation. Topics covered include: equilibrium measure, expansion of the free energy, limit point processes, fluctuations.

      References:
      Lectures on Coulomb and Riesz gases”, https://arxiv.org/abs/2407.21194

    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:10 PM
      Short talks 1h 40m

      Jeferson Dias Da Silva (Federal University of Rio Grande do Sul): Spectral Properties, Localization Transition and Multifractal Eigenvectors of the Laplacian on Heterogeneous Networks
      Abstract: The spectral properties of matrices associated with heterogeneous networks provide critical insights into the dynamics of disordered systems. These heterogeneities arise primarily from two mechanisms: (i) disorder in the connections between nodes and (ii) disorder in the strength of these connections. The statistics of the Laplacian matrix's eigenvalues and eigenvectors play a key role in understanding diffusion processes, relaxation rates in trap models, consensus dynamics, and many other phenomena. While numerical solutions of the cavity equations have advanced our understanding of the Laplacian's spectral and localization properties, analytical solutions remain scarce beyond fully connected networks, where the topology is homogeneous. In this talk, I will discuss the spectral and localization properties of the Laplacian for highly connected heterogeneous networks. I will show that the spectral density diverges within the bulk of the spectrum when network topology fluctuations are sufficiently strong and the coupling strengths are random. This divergence signals a transition from non-ergodic delocalized to localized eigenvectors exhibiting strong multifractal scaling. For constant couplings, we will see that the bulk of the spectrum is characterized by a regular spectral density, and the corresponding eigenvectors are localized. These results highlight the profound impact of heterogeneous topology and random couplings on the spectral and localization properties of the Laplacian.

      Mingyi Hou (Uppsala University): Potential Theory and Metastability for Generic Non-reversible Langevin Processes
      Abstract: Metastability describes the behavior of a stochastic process evolving in a potential landscape with multiple local minima. The process typically remains near one minimum for a long time, fluctuating due to small noise, before making a rare and sudden transition to another minimum. The asymptotic estimate of the mean transition time in the low-noise limit is classically given by the Eyring–Kramers formula. A sharp version of this formula was rigorously established by Bovier et al. (2004), which has since inspired a large body of work on metastability. In this talk, I will briefly review the Eyring–Kramers formula and present recent developments in the potential theory approach to metastability for non-reversible Langevin processes. Emphasis will be placed on the analytical framework and new techniques for studying transition phenomena in systems beyond the reversible setting.

      Farzona Mukhamedova (King's College London): Physics Informed Neural Networks for Aggregation Kinetics
      Abstract: We introduce a novel physics-informed approach for accurately modeling aggregation kinetics which provides a comprehensive solution in a single run by outputting all model parameters simultaneously, a clear advancement over traditional single-output networks that require multiple executions. This method effectively captures the density distributions of both large and small clusters, showcasing a notable improvement in predicting small particles, which have historically posed challenges in computational models. This approach yields significant advancements in computational efficiency and accuracy for solving the Smoluchowski equations by minimizing the interval over which the physics-informed loss function operates, allowing for efficient computation over extended time-frames with minimal increase in computational cost. Due to the  independence of predefined shapes for bias or weight outputs, it removes the dependency on prior assumptions about output structures. Furthermore, our physics-informed framework exhibits high compatibility with the generalized Brownian kernel, maintaining robust accuracy for this previously unaddressed kernel type. The framework's notable novelty also lies in addressing four different kernels with one neural network architecture. Therefore with high computational efficiency, combined with low error margins it indicates significant potential for long-term predictions and integration into broader computational systems. (https://arxiv.org/abs/2410.06050).

      Yilin Ye (LPMC-CNRS – École polytechnique): First-passage Times to a Fractal Boundary: Local Persistence Exponent and its Log-periodic Oscillations
      Abstract: We investigate the statistics of the first-passage time (FPT) to a fractal self-similar boundary of the Koch snowflake. When the starting position is fixed near the absorbing boundary, the FPT distribution exhibits an apparent power-law decay over a broad range of timescales, culminated by an exponential cutoff. By extensive Monte Carlo simulations, we compute the local persistence exponent of the survival probability and reveal its log-periodic oscillations in time due to self-similarity of the boundary. The effect of the starting point on this behavior is analyzed in depth. Theoretical bounds on the survival probability are derived from the analysis of diffusion in a circular sector. Physical rationales for the refined structure of the survival probability are presented.

      Kohei Yoshimura (The University of Tokyo): Conservativeness in Quantum Thermodynamics
      Abstract: Equilibrium and nonequilibrium systems can be distinguished by whether the so-called thermodynamic force is conservative or not. For example, Brownian particles will exhibit the Boltzmann distribution when the external force is a potential force; a chemical system reaches detailed balance when the chemical affinity is closed regarding species within the reaction vessel. However, in open quantum systems, despite its similarity to classical Markovian processes, such a condition was not known. In this talk, by employing a fundamental structure in the quantum master equation, I will discuss how we can install such "conservativeness" to open quantum systems. 

      Ref: K. Yoshimura et al. Phys. Rev. Res. 7 (1), 013244 (2025).

    • 4:10 PM 4:30 PM
      Coffee break 20m
    • 4:30 PM 5:30 PM
      Short talks 1h

      Sofia Flores (Institut d'Astrophysique de Paris): Stellar Dynamics and Statistical Closure Theory
      Abstract: Stars orbiting a supermassive black hole in the center of galaxies undergo very efficient diffusion in their orbital orientations: this is "Vector Resonant Relaxation." Such a dynamics is intrinsically nonlinear, stochastic, and correlated, hence bearing deep similarities with turbulence in fluid mechanics or plasma physics. In that context, we show how generic methods stemming from statistical closure theory, namely, the "Martin-Siggia-Rose formalism”, can be used to characterize the correlations describing the redistribution of orbital orientations. In particular, we explicitly compare the associated prediction for the two-point and three-point correlation functions with measures from numerical simulations.

      Published reference : DOI https://doi.org/10.1103/PhysRevE.111.044111

      Ning Jiang (Wuhan University): On Self-organization  Models: Kinetic and Hydrodynamics
      Abstract: In this talk, we review some self-organization models arising in biology and physics, in particular, Vicsek model 1999 and its mathematical re-visit by Degond and Motsch in 2007. Starting from particle system, it can be derived the kinetic models (SOK) and more macroscopic hydrodynamic models (SOH). We made some rigorous justifications from SOK to SOH, using a Generalized Collision Invariants (GCI)-based Hilbert expansion. Moreover, our proof is employed some geometry of S^2 and SO(3). Furthermore, we introduce the relations of this problem to Landau damping, Taylor dispersion and enhanced dissipations.

      Eun Hee Ko (Institut d'Astrophysique de Paris): Orbit-averaged Fokker-Planck Equation in the Context of Cusp-core Transformation and its Application
      Abstract: Recent JWST observations have revealed superbubbles (SBs)—cavity-shell structures distributed across the galactic disk—formed by successive supernova explosions. The potential fluctuations generated by SBs can dynamically heat galactic systems. Using the orbit-averaged Fokker-Planck equation, we investigate the role of SB-driven stochastic heating in the context of cusp-core transformation. This formalism describes the cumulative impact of weak, local encounters induced by stochastic noise sources. By modeling the expansion and collapse of SBs, along with their inhomogeneous spatial distribution, we derive diffusion coefficients linked to the power spectrum of SB-induced fluctuations. Furthermore, we find simple analytic scaling relations that provide an intuitive understanding of how diffusion efficiency depends on noise source and system parameters.

    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Long Range Interactions - Nonequilibrium point processes with long-range correlations generated by stochastic resetting (2/3) 1h 30m
      Speaker: Satya N. Majumdar (LPTMS)

      In this course, I'll first discuss briefly the equilibrium systems with long-range correlations (such as the log-gas or the Riesz gas in coordination with the course by S. Serfaty). Then the main emphasis of my course would be on nonequilibrium stationary states of many-body systems with attractive long-range correlations. I will provide examples of how such states can be generated by stochastic resetting of N diffusing particles in one or higher dimensions. The joint distribution of the positions of the particles, though not factorizable due to strong correlations, has a certain conditionally independent structure that allows them to be exactly solvable for various physical obsevables. For example, in one dimension and in the large N limit, one can compute exactly, the average density, the extreme value statistics, the spacing distribution between particles as well as the full counting statistics. The presence of long-range correlations make these observables drastically different from the equilibrium ideal gas where the joint distribution in the stationary state factorizes.

      References:

      [1] M. Biroli, H. Larralde, S. N. Majumdar, G. Schehr, Extreme Statistics and Spacing Distribution in a Brownian Gas Correlated by Resetting, Phys. Rev. Lett., 130, 207101 (2023) [Editor’s suggestion].

      [2] M. Biroli, S. N. Majumdar, G. Schehr, Critical number of walkers for diffusive search processes with resetting, Phys. Rev. E, 107, 064141 (2023).

      [3] M. Biroli, H. Larralde,, S. N. Majumdar, G. Schehr, Exact extreme, order and sum statistics in a class of strongly correlated system , Phys. Rev. E 109, 014101 (2024).

      [4] M, Biroli, M. Kulkarni, S. N. Majumdar, G. Schehr, Dynamically emergent correlations between particles in a switching harmonic trap, Phys. Rev. E 109, L032106 (2024).

      [5] S. Sabhapandit, S. N. Majumdar, Noninteracting particles in a harmonic trap with a stochastically driven center, J. Phys. A: Math. Theor. 57, 335003 (2024).

      [6] M. Kulkarni, S. N. Majumdar, S. Sabhapandit, Dynamically emergent correlations in bosons via quantum resetting, J. Phys. A: Math. Theor. 58, 105003 (2025).

      [7] M. Biroli, S. N. Majumdar, G. Schehr, Resetting Dyson Brownian motion, https://arxiv.org/abs/2503.14733 (to appear in PRE, 2025).

      [8] M.R. Evans and S.N. Majumdar, Diffusion with stochastic resetting, Phys. Rev. Lett. , 106, 160601 (2011).

      [9] M. R. Evans, S. N. Majumdar, G. Schehr, Stochastic Resetting and Applications, J. Phys. A: Math. Theor. 53, 193001 (2020) (topical review).

      [10] S. N. Majumdar, G. Schehr, Statistics of extremes and records in random sequences (Oxford University Press) (2024).

    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Random Interfaces - Introduction to the Physics of the KPZ Universality Class (2/3) 1h 30m
      Speaker: Kazumasa Takeuchi (The University of Tokyo)

      The Kardar-Parisi-Zhang (KPZ) class is a prominent universality class known to describe a surprising variety of models and phenomena, including growing interfaces, directed polymers in random media, stochastic particle transport, and most recently even quantum systems such as the exciton-polariton condensate and integrable spin chains. Remarkably, the 1D KPZ class is exactly solvable in many aspects, opening a window to detailed statistical properties through its links to various mathematical problems. The lecture will include a pedagogical introduction to basic properties of the KPZ equation, an illustrative outline of characteristic distribution and correlation properties revealed by exact solutions, and a survey of experimental studies on growing interfaces.
      1. Introduction: why should we care this?
      2. Scaling exponents and universality classes
      3. Basic properties of the KPZ equation
      4. Distribution and correlation properties: stationary & non-stationary cases
      5. Experimental test of predictions from integrable models
      6. Distribution properties for general cases and variational formula

      Main reference: K. A. Takeuchi, "An appetizer to modern developments on the Kardar-Parisi-Zhang universality class", Physica A 504, 77-105 (2018).

      Slides of the lectures : https://bit.ly/ihes2025

    • 1:00 PM 2:30 PM
      Lunch break 1h 30m
    • 2:30 PM 4:10 PM
      Short talks 1h 40m

      Lorenzo Vito Dal Zovo (Politecnico di Torino): Combinatorial Mappings in the Study of the Inhomogeneous TASEP
      Abstract: In this presentation, we provide a combinatorial interpretation of the normalization function for the inhomogeneous TASEP, a TASEP in which at least one hopping rate differs from the others. Unlike the stochastic processes previously studied in the combinatorial literature, the inhomogeneous TASEP does not admit a steady state expressible in the matrix product form. Consequently, we work with the most general representation available, derived from the matrix-tree theorem. Our approach involves analysing the polynomial sequence given by the trace of the n-th power of the transition rate matrix, which we show to be connected to classical problems in enumerative combinatorics, specifically the enumeration of linear extensions of certain families of posets. This methodology is reminiscent of what has been done in the homogeneous case, where the matrices appearing in the matrix product formalism are interpreted as ladder operators and their products as lattice paths. Notably, our approach, even though it results in more challenging enumerations, does not rely on any a priori assumptions about the form of the steady state. We will then show instances where such enumeration results can already be achieved and the missing steps separating us from obtaining a closed-form expression of the normalization function.

      Carla Mariana Da Silva Pinheiro (University of São Paulo): Asymptotics for KdV Solutions in the Thinned Point Process for Unitary Random Matrix Ensembles
      Abstract: Eigenvalues in random matrix ensembles model several systems, such as Coulomb gas, quantum chaos, and even the distribution of trees in a forest. According to physics conjectures, the limit kernel arising in unitary ensembles with equilibrium measure vanishing as power (4k+1)/2 is built from solutions to the Painlevé I hierarchy. We studied a deformation of such kernels, which corresponds to the thinned point process, and proved that the multiplicative statistics are governed by the 2k+1 first equations of the KdV hierarchy. Finally, we derived the asymptotic behaviour of such solutions. This is based on a work in collaboration with Mattia Cafasso.

      João Miguel Machado (Lagrange Mathematical and Computational Research Center): A Statistical Physics Approach to the Mean Field Limit in Pairwise Interaction N-player Games
      Abstract: In this presentation, we shall discuss a statistical physics approach to show that a sequence of Nash equilibria to a family of N-player games with pair-wise interaction converge to a notion of equilibrium for a game with a continuum of players, being called a Cournot-Nash equilibrium. We identify a potential structure for our family of games, both in the discrete N-player formulation as in the continuous one, meaning that equilibria can be obtained via the minimization of a suitable functional. As a consequence, the convergence of equilibria can be obtained via a Gamma convergence result which takes inspiration in the works of S. Serfaty in the study of the mean field behaviour of Coulomb and Riesz gases. However, our results differ from these in the sense that our games have a dependence on the distribution of the players, being represented by an i.i.d. sample, describing the types of said players. Hence, we formulate our results in two situations, when players know their types beforehand, and when players only know the collective distribution before choosing their plays. This distinction gives rise to two Gamma convergence results, in the first situation being a convergence with full probability and the second being a Gamma convergence in the topology of random measures. This is a joint work with Guilherme Mazanti and Laurent Pfeiffer. 

      Parham Radpay (CEA-LIST, Université Paris-Saclay): The Injective Norm of Random Fermionic States and Skew-Symmetric Tensors
      Abstract: We study the injective norm of random antisymmetric tensors drawn from real and complex Gaussian ensembles. In quantum information theory, this problem corresponds to determining the geometric entanglement of random fermionic states. By applying the Kac–Rice formula on the Grassmannian manifold, we derive asymptotic upper bounds on the injective norm in two distinct regimes: for a fixed tensor order p and d→∞, and for fixed filling fraction p/d as both p,d→ ∞, where p is the rank of the tensor and d is the dimension of the underlying vector space. These correspond to lower bounds on the geometric entanglement. Furthermore, we explore the particle-hole duality in the fermionic case and the behavior of entanglement under this duality. This work, in collaboration with Stéphane Dartois, builds upon results from Dartois and McKenna (2024).

      Mathieu Yahiaoui (University of Melbourne): Random Winding Numbers for Determinantal Curves Associated with 2-matrix Models in the AIII Class
      Abstract: In this talk, I will present recent results on the random winding number of determinantal curves arising from the determinant of a two-matrix field evaluated along the unit circle, which are studied through average ratios of characteristic polynomials. I will provide exact formulas for the associated partition function and describe the asymptotic behaviour of the winding number in a broad class of random matrix ensembles known as Pólya ensembles. These include and generalize classical isotropic ensembles such as the Ginibre Unitary Ensemble, Wishart–Laguerre Ensembles, and Muttalib–Borodin Ensembles. This is based on joint work with Mario Kieburg.

    • 4:10 PM 4:30 PM
      Coffee break 20m
    • 4:30 PM 5:30 PM
      Competition at the Front of Expanding Populations 1h
      Speaker: Mehran Kardar (MIT)

      When competing species grow into new territory, the population is dominated by descendants of successful ancestors at the expansion front. Successful ancestry depends on the reproductive advantage (fitness), as well as ability and opportunity to colonize new domains. (1) Based on symmetry considerations, we present a model that  integrates both elements by coupling the classic description of one-dimensional competition (Fisher equation) to the minimal model of front shape (KPZ equation). Macroscopic manifestations of these equations on growth morphology are explored, providing a framework to study spatial competition, fixation, and differentiation, In particular, we find that ability to expand in space may overcome reproductive advantage in colonizing new territory. (2) Variations of fitness, as well as fixation time upon differentiation, are shown to belong to distinct universality classes depending on limits to gain of fitness. 

    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Random Interfaces - Lecture 2: How do Boundary Conditions Influence Random Interface Growth? 1h 30m
      Speaker: Ivan Corwin (Columbia University)

      In modeling real interface growth, there are several types of boundary conditions that can be imposed. In this lecture we will probe some of the predictions and results about how these boundary conditions influence the long-time behavior of interface growth. In particular, we will focus on the stationary measure for such growth models, as well as the time to approach stationarity and the current fluctuations in stationarity.

      References and slides for the lectures

       
    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Long Range Interactions - Nonequilibrium Point Processes with Long-range Correlations Generated by Stochastic Resetting (3/3) 1h 30m
      Speaker: Satya N. Majumdar (LPTMS)

      In this course, I'll first discuss briefly the equilibrium systems with long-range correlations (such as the log-gas or the Riesz gas in coordination with the course by S. Serfaty). Then the main emphasis of my course would be on nonequilibrium stationary states of many-body systems with attractive long-range correlations. I will provide examples of how such states can be generated by stochastic resetting of N diffusing particles in one or higher dimensions. The joint distribution of the positions of the particles, though not factorizable due to strong correlations, has a certain conditionally independent structure that allows them to be exactly solvable for various physical obsevables. For example, in one dimension and in the large N limit, one can compute exactly, the average density, the extreme value statistics, the spacing distribution between particles as well as the full counting statistics. The presence of long-range correlations make these observables drastically different from the equilibrium ideal gas where the joint distribution in the stationary state factorizes.

      References:

      [1] M. Biroli, H. Larralde, S. N. Majumdar, G. Schehr, Extreme Statistics and Spacing Distribution in a Brownian Gas Correlated by Resetting, Phys. Rev. Lett., 130, 207101 (2023) [Editor’s suggestion].

      [2] M. Biroli, S. N. Majumdar, G. Schehr, Critical number of walkers for diffusive search processes with resetting, Phys. Rev. E, 107, 064141 (2023).

      [3] M. Biroli, H. Larralde,, S. N. Majumdar, G. Schehr, Exact extreme, order and sum statistics in a class of strongly correlated system , Phys. Rev. E 109, 014101 (2024).

      [4] M, Biroli, M. Kulkarni, S. N. Majumdar, G. Schehr, Dynamically emergent correlations between particles in a switching harmonic trap, Phys. Rev. E 109, L032106 (2024).

      [5] S. Sabhapandit, S. N. Majumdar, Noninteracting particles in a harmonic trap with a stochastically driven center, J. Phys. A: Math. Theor. 57, 335003 (2024).

      [6] M. Kulkarni, S. N. Majumdar, S. Sabhapandit, Dynamically emergent correlations in bosons via quantum resetting, J. Phys. A: Math. Theor. 58, 105003 (2025).

      [7] M. Biroli, S. N. Majumdar, G. Schehr, Resetting Dyson Brownian motion, https://arxiv.org/abs/2503.14733 (to appear in PRE, 2025).

      [8] M.R. Evans and S.N. Majumdar, Diffusion with stochastic resetting, Phys. Rev. Lett. , 106, 160601 (2011).

      [9] M. R. Evans, S. N. Majumdar, G. Schehr, Stochastic Resetting and Applications, J. Phys. A: Math. Theor. 53, 193001 (2020) (topical review).

      [10] S. N. Majumdar, G. Schehr, Statistics of extremes and records in random sequences (Oxford University Press) (2024).

    • 1:00 PM 2:30 PM
      Lunch break (Buffet at IHES) 1h 30m
    • 2:30 PM 4:00 PM
      Random Interfaces - Introduction to the Physics of the KPZ Universality Class (3/3) 1h 30m
      Speaker: Kazumasa Takeuchi (The University of Tokyo)

      The Kardar-Parisi-Zhang (KPZ) class is a prominent universality class known to describe a surprising variety of models and phenomena, including growing interfaces, directed polymers in random media, stochastic particle transport, and most recently even quantum systems such as the exciton-polariton condensate and integrable spin chains. Remarkably, the 1D KPZ class is exactly solvable in many aspects, opening a window to detailed statistical properties through its links to various mathematical problems. The lecture will include a pedagogical introduction to basic properties of the KPZ equation, an illustrative outline of characteristic distribution and correlation properties revealed by exact solutions, and a survey of experimental studies on growing interfaces.
      1. Introduction: why should we care this?
      2. Scaling exponents and universality classes
      3. Basic properties of the KPZ equation
      4. Distribution and correlation properties: stationary & non-stationary cases
      5. Experimental test of predictions from integrable models
      6. Distribution properties for general cases and variational formula

      Main reference: K. A. Takeuchi, "An appetizer to modern developments on the Kardar-Parisi-Zhang universality class", Physica A 504, 77-105 (2018).

      Slides of the lectures : https://bit.ly/ihes2025

    • 4:00 PM 4:30 PM
      Coffee break 30m
    • 4:30 PM 5:30 PM
      Short talks 1h

      Florent Fougères (ENS, Paris): Deriving the Linear Rayleigh–Boltzmann Equation: Statistical and Long-time Results
      Abstract: For dilute gases, the derivation of the Boltzmann equation (1872) from the Newton microscopic equations has been mathematically conducted in 1975 by Oscar Lanford; and in the 2010s Isabelle Gallagher, Laure Saint-Raymond and Benjamin Texier gave precise quantitative estimates on the convergence of the empirical measure to the solution to this equation, yet only for very short times. In a linear context - close to equilibrium - this work has allowed long-time results, later on used to derive hydrodynamic limits - as suggested by the Hilbert’s sixth problem.
      In this presentation we show a more precise approach to prove far faster convergence rates in this linear case, based on a more physics-inspired time sampling, so as to give these results more physical meaning. We will also be able to tackle statistical aspects of this derivation in the framework of a gas mixture of tagged particles, which are the subject of an ongoing work.

      Chenjiayue Qi (IHES): Global Solution of a Functional Hamilton-Jacobi Equation associated with a Hard Sphere Gas
      Abstract: In recent years it has been shown for hard sphere gas that, by retaining the correlation information, dynamical fluctuation and large deviation of empirical measure around Boltzmann equation could be proved, in addition to the classical kinetic limit result by Lanford. After taking low-density limit, the correlation information can be encoded into a functional Hamilton-Jacobi equation. The results above are restricted to short time. In this talk, we establish global-in-time construction of a solution of the Hamilton-Jacobi equation, by analyzing a system of coupled Boltzmann equations. The global solution converges to a non-trivial stationary solution of the Hamilton-Jacobi equation in the long-time limit under proper assumptions. This talk is based on arxiv:2409.02805. For relevant slide and poster see https://sites.google.com/view/chenjiayueqi/

      Songbo Wang (École polytechnique): Size of Chaos for Gibbs Measures of Mean Field Interacting Diffusions
      Abstract: We investigate Gibbs measures for diffusive particles interacting through a two-body mean field energy. By uncovering a gradient structure for the conditional law, we derive sharp bounds on the size of chaos, providing a quantitative characterization of particle independence. To handle unbounded interaction forces, we study the concentration of measure phenomenon for Gibbs measures via a defective Talagrand inequality, which may hold independent interest. Our approach provides a unified framework for both the flat semi-convex and displacement convex cases. Additionally, we establish sharp chaos bounds for the quartic Curie-Weiss model in the sub-critical regime, demonstrating the generality of this method. This is a joint work with Zhenjie Ren.

    • 9:00 AM 9:30 AM
      Welcome coffee 30m
    • 9:30 AM 11:00 AM
      Random Interfaces - Lecture 3: What is the Universal Scaling Limit of Random Interface Growth, and What Does It Tell Us? 1h 30m
      Speaker: Ivan Corwin (Columbia University)

      KPZ scaling leads to a natural renormalization of the entire space-time trajectory of random interface growth models. What is the universal limit of this process? I will first introduce the KPZ fixed point and directed landscape which characterize this limit. Then, I will describe some implications of this full scaling limit, in particular to the theory of fluctuating hydrodynamics and multispecies interacting particle systems. Finally, I will sketch how the Yang-Baxter equation -- the fundamental relation from quantum integrable systems -- gives the key to unlocking this scaling limit.

      References and slides for the lectures

       
    • 11:00 AM 11:20 AM
      Coffee break 20m
    • 11:20 AM 12:50 PM
      Long Range Interactions - Systems with Coulomb and Riesz Interactions (3/3) 1h 30m
      Speaker: Sylvia Serfaty (Sorbonne Université & Courant Institute of Mathematical Sciences)

      We will focus on the statistical mechanics of large systems of particles with Coulomb or Riesz long-range repulsion, analyzed via an electrostatic formulation. Topics covered include: equilibrium measure, expansion of the free energy, limit point processes, fluctuations.

      References:
      Lectures on Coulomb and Riesz gases”, https://arxiv.org/abs/2407.21194