In a landmark paper, Scott Sheffield introduced a famous bijection, called the “hamburger-cheeseburger” bijection, to encode a certain model of random planar maps as a certain queue model in a kitchen selling hamburgers and cheeseburgers. Under this correspondence, natural geometric observables have a nice “burger” interpretation, for which Sheffield established scaling limit results. These scaling limits exhibit a phase transition in a special regime where the maps are believed to be “critical”. The goal of this talk is to present the analogue of these scaling limit results in the critical case, which can be thought of as the first convergence result of planar maps in the universality class of critical Liouville quantum gravity, in the so-called peanosphere sense. The talk is based on joint work with Xingjian Hu, Ellen Powell and Mo Dick Wong.

Many phenomena of interest in various applicative fields (epidemiology, neuroscience,…) can be idealized as interacting particle systems on random graphs. Various approaches have been proposed in recent years to develop tractable approximations of these dynamics that take the graph geometry and particle correlations into account. One of them, introduced by Lacker, Ramanan and Wu, focuses on dynamics on sparse graphs and their local limits. Analogously to mean-field models on complete and dense graphs, it is possible to establish so-called local-field equations on random trees that provide an autonomous description of the neighborhood of the root. In this talk, we will give a general overview of the local-field approach, as well as a recent result of quantitative propagation of chaos in this framework.

Exposé à Metz. Titre et résumé à venir.

Exposé à Metz. Titre et résumé à venir.

Date possible pour le colloquinte

La suite d’Oldenburger-Kolakoski est l’unique mot infini sur l’alphabet {1,2} qui commence par un “1” et est point fixe de l’opérateur de dérivation. En 1991, M.S. Keane conjecture que cette suite admet une fréquence d’1/2 pour la lettre “1”.

Les suites dites “auto-descriptives” sont une généralisation du mot d’Oldenburger-Kolakoski. Ces suites sont en bijection naturelle avec l’ensemble de toutes les suites sur l’alphabet {1,2} : une suite auto-descriptive est dite “dirigée” par son homologue naturelle sur {1,2}. Est-il possible d’inférer les fréquences de lettres de l’une à partir de l’autre ?

Je présenterai dans cet exposé deux approches à cette question : l’une probabiliste (Boisson, Jamet, Marcovici — 2024), l’autre analytique (Akiyama, Jamet, Marcovici, T.C. — 2024).

Many arguments in mathematics rely on the introduction of an auxiliary function that is compactly supported. Sometimes this compactly supported function is required to satisfy some additional regularity, but this cannot be pushed too far. It can for instance not be analytic as follows by the identity principle.

In this talk, we wish to present a technique that may bypass this obstruction. Namely, instead of considering a single function, we shall construct a sequence of function that are supported in the same compact, and satisfy some good uniform bounds on their derivatives.

This method is a powerful tool as it can, in some circumstances, turn a heuristic argument relying on a compactly supported analytic auxiliary function, into a rigorous proof. As an application, we shall discuss how this technique leads to improvements in some quantified Tauberian theorems.

One of the central topics in number theory is the study of $L$-functions and the distribution of their zeros. For example, the celebrated Prime Number Theorem is equivalent to the fact that the Riemann zeta function $\zeta(s)$ does not vanish on the line $\text{Re}(s)=1$. In this talk, we will focus on quadratic Dirichlet $L$-functions: in particular, the real zeros of the derivative of quadratic Dirichlet $L$-functions $L^\prime (s, \chi_d)$, where $d$ ranges over fundamental discriminants. Baker and Montgomery conjectured that there are $\asymp \log \log |d|$ real zeros of $L^\prime(s, \chi_d)$ in the interval $[1/2, 1]$ for almost all fundamental discriminants $d$. We will highlight some recent exciting progress that comes close to proving this conjecture and then outline the proof, which is based on ideas coming from analytic and probabilistic number theory. This is based on recent joint work with Youness Lamzouri.

In this talk, we will report on a joint work with Sanoli Gun,
where we study extreme values of $L$-functions attached to quadratic twists of Dirichlet characters.
We show that for any $\epsilon >0$ and Dirichlet character $F$ of odd conductor $q$, not necessarily a primitive form,
there exists at least $X^{1-\epsilon}$ fundamental discriminants $8d$ with $X< d \le 2X$ and $(d, 2q) =1$
such that $|L(1/2, F \otimes \chi_{8d})|$ takes large values.

La construction des nombres surréels de Conway donne également une nouvelle approche aux nombres réels, que je qualifierais de « quantique », opposée à  l’approche « usuelle », que je qualifierais de « classique ». J’essaierai d’expliquer cette opposition « quantique-classique » en la mettant dans le contexte de la théorie des ensembles (je résumerai en partie mon cours « Ensembles et nombres » que je donne actuellement dans le cadre de notre école doctorale). Cette séance se veut plus un forum de discussion qu’une présentation formelle de résultats : le sujet du « quantique » est pour ainsi dire la musique de fond du travail pour beaucoup d’entre nous, et il me semble intéressant de poser et de discuter des questions sur ce fond. 

Dans un travail récent, McNew, Pollack et Singha Roy obtiennent plusieurs résultats relatifs à la distribution du facteur premier médian des entiers lorsque ce dernier est défini en tenant compte de la multiplicité. En particulier, le comportement asymptotique des lois locales est étudié et fait apparaître une transition de phase qui n’est pas décrite. Dans cet exposé, nous présenterons une partie des améliorations et des résultats obtenus pour les lois locales et certaines applications.

In this talk, we will discuss diophantine approximations of irrational numbers by rational numbers, where the denominators are taken from certain interesting subsets of the positive integers. First, we will consider Diophantine approximations in which the denominators are drawn from the set of positive integers represented by a given positive definite integral binary quadratic form. Next, we will discuss Diophantine approximations where the denominators are restricted to the set of y-smooth (or friable) numbers for some given y > 0. Finally, we will outline some of the proofs.

Assume that $\omega_1, \ldots, \omega_n$ are i.i.d. uniform random points in $[0,1]$, which serve as the centers of shrinking intervals of given lengths $\ell_1 \ge \cdots \ge \ell_n$. The Dvoretzky covering problem asks for necessary and sufficient conditions on the sequence $(\ell_n)$ under which these random intervals cover $[0,1]$ infinitely often, almost surely. The problem was solved in 1972 by Shepp, and his work has since been generalized in several directions.

In this talk, I will discuss some deterministic analogues of Shepp’s result and their applications to the Littlewood Conjecture.

[Exposé en ligne diffusé dans la salle de séminaire]

A common question for geometric structures is that of rigidity. Infinitesimally rigidity is often controlled by the vanishing of a cohomology group. A common question then becomes when infinitesimal rigidity actually implies rigitidy. In this talk we discuss the case of regular foliations.

In the first part we define the terms involved and give an overview of the current state of the art. Moreover, we highlight a relation with the rigidity of group actions.

One encounters two issues in the literature. Many results require the foliation to have compact leaves and there is a general lack of examples.

In the second part of the talk, we take a step towards addressing those issues, by outlining a construction of infinitesimally rigid foliations with dense leaves. This is based on joint work with Stephane Geudens.

Je vais présenter un problème d’optimisation à frontière libre analogue au problème de Alt-Caffarelli pour les fonctions biharmoniques. Ce problème apparaît dans différentes questions d’optimisation de forme, dont la minimisation de la trainée d’un obstacle dans un fluide sous contrainte de mesure, la minimisation de la première valeur propre de l’opérateur de Stokes (ou de flambage) dans les domaines du plan, etc.. On s’attend à ce que la frontière libre obtenue soit généralement une union de courbes lisses, pouvant se rejoindre avec un angle d’environ 1.43pi, et je présenterai plusieurs résultats allant dans ce sens.

C’est un travail en collaboration avec Jimmy Lamboley.

Résumé à venir

Résumé à venir

Résumé à venir

Résumé à venir

Le succès de la chimiothérapie dépend à la fois de la stratégie d’administration du médicament et de sa capacité à éliminer les cellules cancéreuses tout en préservant autant que possible les tissus sains. Dans cette présentation, nous nous intéresserons à un problème de contrôle optimal avec des contraintes d’état appliqué à la chimiothérapie des tumeurs invasives, où la dose de médicament agit comme variable de contrôle. Étant donné que le traitement affecte à la fois les cellules tumorales et les tissus sains, l’objectif du
problème de contrôle est de réduire la densité tumorale en contrôlant la dose du médicament. Pour ce faire, nous modélisons l’action thérapeutique à l’aide d’une équation de réaction-diffusion non linéaire décrivant l’évolution d’une tumeur invasive sous traitement. Nous commençons par analyser mathématiquement le problème initial de valeur limite. Nous formulons ensuite le problème de contrôle optimal sous contraintes et en déduisons les conditions nécessaires à l’optimalité. Enfin, à l’aide de simulations numériques en 2D pour un cas de cancer du sein, nous illustrons l’importance des contraintes d’état dans les stratégies de traitement optimales, avant de conclure par quelques perspectives

The lecture addresses time‑dependent convection–diffusion problems with high Péclet number in thin 3D graph‑like networks of curvilinear cylinders connected by nodes of diameter $\mathcal{O}(\varepsilon).$ Inhomogeneous Robin boundary conditions with different intensity factors are imposed on the network boundary. As $\varepsilon \rightarrow 0,$ the network collapses to a graph and the diffusion terms vanish.

Such problems pose singular‑perturbation challenges that standard methods often cannot resolve. I present a systematic asymptotic framework for $\varepsilon \rightarrow 0,$ combining regular expansions on edges with node‑layer and boundary‑layer asymptotics to capture the multiscale flow structure. The analysis justifies reduced graph models, quantifies higher‑order corrections, and uncovers new phenomena in singular regimes.

In a striking contrast with the classical case of real-valued Sobolev functions, a Sobolev map with values into a given compact manifold N need not be approximable with smooth N-valued maps.
This observation, initially due to Schoen and Uhlenbeck (1983), gave rise to a whole area of research concerned with questions related to approximability properties of Sobolev mappings with values into a compact manifold. In this talk, I will give a broad overview of this research direction, its history, the main problems it is concerned with, important known results, as well as some recent contributions.

Bohr’s correspondence principle asserts that the predictions of classical and quantum mechanics coincide in the limit of large quantum numbers. This connection becomes especially striking when one studies Schrödinger-type equations for initial data minimising the uncertainty principle, known as “gaussian coherent states”, in the semiclassical limit. More precisely, in the context of quantum mechanics, one can derive a complete asymptotic expansion in the semiclassical parameter for solutions of such equations. The aim of this talk is to explain how to obtain a similar result in the framework of Quantum Field Theory for a certain class on analytic interactions, with a particular stress on spatially cutoff $P(\phi)_2$ interactions.

Dans ce travail, nous étudions un problème de contrôle optimal impliquant un modèle simplifié pour la protection d’un champ de culture. Plus précisément, nous considérons une protection sur un champ de culture et cherchons à placer des zones d’intervention, représentées par un contrôle, afin de maximiser la protection sur le champ pendant une période donnée. En utilisant une méthode de relaxation, nous prouvons qu’il existe un contrôle qui maximise la protection et, de plus, ce contrôle doit être de type bang-bang. Par ailleurs, avec des hypothèses supplémentaires sur la géométrie du champ de culture, certains résultats sur la forme de l’intervention optimale sont démontrés en utilisant des résultats de comparaison via les symétrisations de Schwarz et de Steiner. Enfin, des simulations numériques sont réalisées pour illustrer ces résultats.

A Euclidean Random Assignment Problem (ERAP) is the study of the random optimal cost of assigning blue points to red points, where these two sets form independent binomial point processes on the same metric space.

Originally introduced in physics more than three decades ago as toy models for spin glasses, ERAPs serve as discrete analogues of the celebrated Monge–Kantorovich problem in optimal transport and have found numerous applications, including in satellite-based Earth observation.

In this talk, I will first introduce ERAPs and review known results for cases where the underlying metric space has additional structure, such as Euclidean spaces of dimension $d \geq 1$.

Then I will show that, when blue and red points are uniformly distributed on the unit circle, the cost of the quadratic ERAP is asymptotically distributed as $\sum_{k=1}^\infty E_k$, where $(E_k)_{k \geq 1}$ is a family of independent $\mathrm{Exp}(k^2)$ random variables.

Talk mostly based on my PhD Thesis.

Dans cet exposé nous présenterons des résultats d’existence, d’unicité et de régularité de solution pour certains problèmes elliptiques du second ordre complètement non linéaires. Les opérateurs sont du modèle du p-Laplacien mais ne peuvent pas se mettre sous forme divergente. Les solutions le seront donc dans le sens des solutions de viscosité.  Dans la première partie, notre démarche sera d’abord de prouver l’existence de sous- et sur- solutions de viscosité. Puis, nous montrerons l’existence de solutions de viscosité à l’aide de la méthode de Perron. Nous prouverons l’unicité de la solution et discuterons sa régularité. Dans la deuxième partie, nous considérerons un opérateur symétriquement radiale (un des opérateurs  de Pucci) et prouverons l’existence et l’unicité de solution radiale dans un anneau. Enfin nous donnerons certains résultats récents et perspectives de recherche sur les propriétés de ces équations.

We consider linear waves on a bounded domain where one part of the boundary is governed by a coupled lower-dimensional wave equation (i.e., dynamic Ventcel/Wentzell boundary condition) and is subject to viscous damping. The other (possibly empty) part is left at rest. When the dynamic boundary geometrically controls the domain, we show that the total energy of classical solutions decays like 1/t. The proof relies on an analysis of high-frequency quasimodes, suitable boundary estimates obtained in different microlocal regimes, and a special decoupling argument. Optimality is assessed via an appropriate quasimode construction.

Ongoing work with Nicolas Vanspranghe (Inria team DISCO, L2S –CentraleSupélec).

More than 50 years ago, Moore predicted that the number of transistors on a microchip doubles every two years. This exponential growth is approaching its physical limit, highlighting the need for alternative computing paradigms. One promising avenue is neuromorphic computing, which aims to emulate the structure and function of the human brain. A key enabling technology is the memristor, a nonlinear resistor with memory. Memristors are capable of mimicking the dynamic conductance behavior of biological synapses, making them well-suited for implementing energy-efficient neural networks.

This talk focuses on the mathematical analysis of three-species drift-diffusion equations for memristors. We investigate the existence and boundedness of global-in-time weak solutions. The mathematical difficulties originate from the three-species situation and the different types of boundary conditions. These issues are addressed by combining free energy estimates with local and global compactness arguments. Additionally, we analyze memristor models coupled with electrical networks. One-dimensional numerical simulations capture the characteristic hysteresis behavior in the current-voltage curves, which are a fingerprint for memristive devices.

In this talk, I will present recent results on unique continuation for semilinear wave and Schrödinger equations with analytic nonlinearities. After recalling some motivation and classical results on the topic, I will describe a new method, introduced in a joint work with Camille Laurent (CNRS, LMR), that relies on analyticity-in-time regularization in finite time for solutions vanishing on a subset satisfying the Geometric Control Condition (GCC). The proof combines tools from control theory with ideas of Hale and Raugel on the regularity of attractors in dynamical systems. In a more recent work, we refined this approach and applied it to Schrödinger equations on compact manifolds, showing that the GCC suffices for unique continuation, thus answering in the affirmative an open problem posed by Dehman, Gérard, and Lebeau (2006) for the nonlinear case. The method is abstract and can also be applied to study similar questions for other PDEs.