arXiv subjects

Eric Bertin

Publications and source records attributed to Eric Bertin.

At least 19 recordsLinked to original sources

arXiv2026

Theory of collective learning in populations of adaptive agents

We investigate homogeneous populations of smart active agents that exchange information with their neighbors to perform a decentralized learning process aimed at achieving a prescribed macroscopic state. Such agents may, for example, represent simple microrobots. The exchanged information comprises tunable parameters governing the agent dynamics, referred to as the individual policy, together with an internal memory encoding previously visited states. This memory is used to evaluate a reward that quantifies the success of a policy to achieve the prescribed state. We extend the kinetic-theory description of collective learning in spatially homogeneous systems [Phys. Rev. Lett. 134, 248302 (2025)] and derive formal evolution equations for the distribution of policies across the population. A central outcome of our theory is the emergence of an effective reward function that fully determines the evolution of the policy distribution and encapsulates the microscopic details of the agents physical and memory dynamics. We obtain closed equations for the policy mean and variance which admit explicit time-dependent solutions under the assumption of Gaussian-distributed memories and polices. To illustrate the framework, we present a series of minimal microscopic models, considering both perfect and partial separation of physical, memory and policy exchange time scales, as well as models with one- and two-dimensional policies. The obtained theoretical results compare well with agent-based numerical simulations. The theory captures key aspects of collective learning, including the influence of population diversity and reward fluctuations on learning performance. Finally, we discuss potential applications to swarm robotics and machine learning, and highlight connections with classical models of biological evolution, including the Replicator equation and the Moran model.

cond-mat.stat-mech↗

arXiv2026

Yielding versus random organization: convex absorbing transitions in soft matter

We compare two different soft matter models, a generalized Random Organization Model (ROM) describing the stroboscopic dynamics of cyclically sheared suspensions, and an elastoplastic model describing the mesoscale dynamics of a yield-stress fluid under imposed stress. Both show absorbing phase transitions, sharing a peculiar mechanism: activity induces an internal noise which is transmitted over large distances by long-ranged mediated interactions, either hydrodynamic or elastic, which results in non-local creation of activity. They also both show convex transitions (i.e., the exponent $\beta >1$), in stark contrast with usual absorbing phase transitions, like (Conserved) Directed Percolation, which are concave ($\beta <1$). We further compare the dependence of the critical properties (activity mean value and fluctuations, avalanche statistics, low-wavenumber structure factor) on the decay exponent $\alpha$ of long-range interactions in both models, finding a qualitatively similar scenario. A smooth crossover is observed as a function of $\alpha$ between a concave transition regime for short-range interactions, with diverging fluctuations and compact avalanches, and a convex transition regime, with vanishing fluctuations and non-compact avalanches, for longer-range interactions. Although for a given range exponent $\alpha$, the values of critical exponents for both models differ, a good agreement between the models is found by parametrically plotting the different critical exponents as a function of the exponent $\beta$ of the mean activity. In this parametric representation, the concave regime is consistent with the behavior of the Long-Range Conserved Directed Percolation class, while the convex regime can be accounted for by a mean-field-type scenario with anomalous diffusion close to an absorbing boundary, inspired by the H\'ebraud-Lequeux model for the yielding transition.

cond-mat.soft↗

arXiv2025

Random organization criticality with long-range hydrodynamic interactions

Driven soft athermal systems may display a reversible-irreversible transition between an absorbing, arrested state and an active phase where a steady-state dynamics sets in. A paradigmatic example consists in cyclically sheared suspensions under stroboscopic observation, for which in absence of contacts during a shear cycle particle trajectories are reversible and the stroboscopic dynamics is frozen, while contacts lead to diffusive stroboscopic motion. The Random Organization Model (ROM), which is a minimal model of the transition, shows a transition which falls into the Conserved Directed Percolation (CDP) universality class. However, the ROM ignores hydrodynamic interactions between suspended particles, which make contacts a source of long-range mechanical noise that in turn can create new contacts. Here, we generalize the ROM to include long-range interactions decaying like inverse power laws of the distance. Critical properties continuously depend on the decay exponent when it is smaller than the space dimension. Upon increasing the interaction range, the transition turns convex (that is, with an order parameter exponent $\beta > 1$), fluctuations turn from diverging to vanishing, and hyperuniformity at the transition disappears. We rationalize this critical behavior using a local mean-field model describing how particle contacts are created via mechanical noise, showing that diffusive motion induced by long-range interactions becomes dominant for slowly-decaying interactions.

cond-mat.soft↗

arXiv2025

Kinetic theory of decentralized learning for smart active matter

Smart active matter has the ability to control its motion guided by individual policies to achieve collective goals. We introduce a theoretical framework to study a decentralized learning process in which agents can locally exchange policies to adapt their behavior and maximize a predefined reward function. We use our formalism to derive explicit hydrodynamic equations for the policy dynamics. We apply the theory to two different microscopic models where policies correspond either to fixed parameters similar to evolutionary dynamics, or to state-dependent controllers known from the field of robotics. We find good agreement between theoretical predictions and agent-based simulations. By deriving fundamental control parameters and uncertainty relations, our work lays the foundations for a statistical physics analysis of decentralized learning.

cond-mat.stat-mech↗

arXiv2024

R\'enyi complexity in mean-field disordered systems

Configurational entropy, or complexity, plays a critical role in characterizing disordered systems such as glasses, yet its measurement often requires significant computational resources. Recently, R\'enyi entropy, a one-parameter generalization of Shannon entropy, has gained attention across various fields of physics due to its simpler functional form, making it more practical for measurements. In this paper, we compute the R\'enyi version of complexity for prototypical mean-field disordered models, including the random energy model, its generalization, referred to as the random free energy model, and the $p$-spin spherical model. We first demonstrate that the R\'enyi complexity with index $m$ is related to the free energy difference for a generalized annealed Franz-Parisi potential with $m$ clones. Detailed calculations show that for models having one-step replica symmetry breaking (RSB), the R\'enyi complexity vanishes at the Kauzmann transition temperature $T_K$, irrespective of $m>1$, while RSB solutions are required even in the liquid phase. This study strengthens the link between R\'enyi entropy and the physics of disordered systems and provides theoretical insights for its practical measurements.

cond-mat.dis-nn↗

arXiv2024

Giant density fluctuations in locally hyperuniform states

Systems driven far from equilibrium may exhibit anomalous density fluctuations: active matter with orientational order display giant density fluctuations at large scale, while systems of interacting particles close to an absorbing phase transition may exhibit hyperuniformity, suppressing large-scale density fluctuations. We show that these seemingly incompatible phenomena can coexist in nematically ordered active systems, provided activity is conditioned to particle contacts. We characterize this unusual state of matter and unravel the underlying mechanisms simultaneously leading to spatially enhanced (on large length scales) and suppressed (on intermediate length scales) density fluctuations. Our work highlights the potential for a rich phenomenology in active matter systems in which particles' activity is triggered by their local environment, and calls for a more systematic exploration of absorbing phase transitions in orientationally-ordered particle systems.

cond-mat.stat-mech↗

arXiv2024

Protocol dependence for avalanches under constant stress in elastoplastic models

Close to the yielding transition, amorphous solids exhibit a jerky dynamics characterized by plastic avalanches. The statistics of these avalanches have been measured experimentally and numerically using a variety of different triggering protocols, assuming that all of them were equivalent for this purpose. In particular two main classes of protocols have been studied, deformation under controlled strain or under controlled stress. In this work, we investigate different protocols to generate plasticity avalanches and conduct twodimensional simulations of an elastoplastic model to examine the protocol dependence of avalanche statistics in yield-stress fluids. We demonstrate that when stress is controlled, the value and even the existence of the exponent governing the probability distribution function of avalanche sizes strongly depend on the protocol chosen to initiate avalanches. This confirms in finite dimensions a scenario presented in a previous mean-field analysis. We identify a consistent stress-controlled protocol whose associated avalanches differ from the quasi-static ones in their fractal dimension and dynamical exponent. Remarkably, this protocol also seems to verify the scaling relations among exponents previously proposed. Our results underscores the necessity for a cautious interpretation of avalanche universality within elastoplastic models, and more generally within systems where several control parameters exist.

cond-mat.soft↗

arXiv2024

A cascade model for the discontinuous absorbing phase transition between turbulent and laminar flows

We introduce a minimal model of energy transfer through scales to describe, at a qualitative level, the subcritical transition between laminar and turbulent flows, viewed in a statistical physics framework as a discontinuous absorbing phase transition. The main control parameter of the model is a Reynolds number that compares energy transfer to viscous dissipation on a large length scale. In spite of its simplicity, the model qualitatively reproduces a number of salient features of the subcritical laminar-turbulent transition, including the existence of an absorbing laminar state, the discontinuous onset of a metastable fluctuating turbulent state above a threshold Reynolds number, and a faster-than-exponential increase of the turbulence lifetime when increasing the Reynolds number. The behavior of the model is also consistent, at high Reynolds number, with the Kolmogorov K41 phenomenology of fully developed turbulence.

physics.flu-dyn↗

arXiv2024

Hidden collective oscillations in a disordered mean-field spin model with non-reciprocal interactions

We study the effect of introducing separable quenched disorder on a non-equilibrium mean-field spin model exhibiting a phase transition to an oscillating state in the absence of disorder, due to non-reciprocal interactions. In the disordered model, the magnetisation and its time derivative no longer carry the signature of the phase transition to an oscillating state. However, thanks to the separable (Mattis-type) form of the disorder, the presence of oscillations can be revealed by introducing a specific, disorder-dependent observable. We also introduce generalised linear and non-linear susceptibilities associated either with the magnetisation or with its time derivative. While linear susceptibilities show no sign of a phase transition, the third-order susceptibilities present a clear signature of the onset of an oscillating phase. In addition, we show that the overlap distribution also provides evidence for the presence of oscillations, without explicit knowledge of the disorder.

cond-mat.dis-nn↗

arXiv2024

Collective oscillations in a three-dimensional spin model with non-reciprocal interactions

We study the onset of collective oscillations at low temperature in a three-dimensional spin model with non-reciprocal short-range interactions. Performing numerical simulations of the model, the presence of a continuous phase transition to global oscillations is confirmed by a finite-size scaling analysis. By systematically varying the interaction range, we show that collective oscillations in this spin model actually result from two successive phase transitions: a mean-field phase transition over finite-size neighborhoods, which leads to the emergence of local noisy oscillators, and a synchronization transition of local noisy oscillators, which generates coherent macroscopic oscillations. Using a Fokker-Planck equation under a local mean-field approximation, we derive from the spin dynamics coupled Langevin equations for the complex amplitudes describing noisy oscillations on a mesoscopic scale. The phase diagram of these coupled equations is qualitatively obtained from a fully-connected (mean-field) approximation. This analytical approach allows us to clearly disentangle the onset of local and global oscillations, and to identify the two main control parameters, expressed as combinations of the microscopic parameters of the spin dynamics, that control the phase diagram of the model.

cond-mat.stat-mech↗

arXiv2024

Far-from-equilibrium complex landscapes

Systems with a complex dynamics like glasses or models of biological evolution are often pictured in terms of complex landscapes, with a large number of possible collective states. We show on the example of a stochastic spin model with non-reciprocal and heterogeneous interactions how the complex landscape notion can be generalized far from equilibrium, where collective states may exhibit spontaneous oscillations, often hidden by the presence of disorder. We identify relevant observables, like the density of entropy production, to unveil the presence of oscillations, and we characterize the complex landscape of our model in terms of a configurational entropy, that counts the number of nonequilibrium collective states with a given entropy production density.

cond-mat.dis-nn↗

arXiv2024

Behavioral transition of a fish school in a crowded environment

In open water, social fish gather to form schools, in which fish generally align with each other. In this work, we study how this social behavior evolves when perturbed by artificial obstacles. We measure the collective behavior of a group of zebrafish in the presence of a periodic array of pillars. When pillar density is low, the fish regroup with a typical inter-distance and a well-polarized state with parallel orientations, similar to their behavior in open water conditions. Above a critical density of pillars, their social interactions, which are mostly based on vision, are screened and the fish spread randomly through the aquarium, orienting themselves along the free axes of the pillar lattice. The abrupt transition from natural to artificial orientation happens when the pillar inter-distance is comparable to the social distance of the fish, i.e., their most probable inter-distance. We develop a stochastic model of the relative orientation between fish pairs, taking into account alignment, anti-alignment and tumbling, from a distribution biased by the environment. This model provides a good description of the experimental probability distribution of the relative orientation between the fish and captures the behavioral transition. Using the model to fit the experimental data provides qualitative information on the evolution of cognitive parameters, such as the alignment or the tumbling rates, as the pillar density increases. At high pillar density, we find that the artificial environment imposes its geometrical constraints to the fish school, drastically increasing the tumbling rate.

physics.bio-ph↗

arXiv2024

Yielding is an absorbing phase transition with vanishing critical fluctuations

The yielding transition in athermal complex fluids can be interpreted as an absorbing phase transition between an elastic, absorbing state with high mesoscopic degeneracy and a flowing, active state. We characterize quantitatively this phase transition in an elastoplastic model under fixed applied shear stress, using a finite-size scaling analysis. We find vanishing critical fluctuations of the order parameter (i.e., the shear rate), and relate this property to the convex character of the phase transition ($\beta >1$). We show explicitly that the CDP class is recovered when both properties are relaxed. We locate yielding within a family of models akin to fixed-energy sandpile (FES) models, only with long-range redistribution kernels with zero-modes that result from mechanical equilibrium. For redistribution kernels with sufficiently fast decay, this family of models belong to a short-range universality class distinct from the Conserved Directed Percolation class of usual FES, which is induced by zero modes.

cond-mat.stat-mech↗

arXiv2023

Biased motility-induced phase separation: from chemotaxis to traffic jams

We propose a one-dimensional model of active particles interpolating between quorum sensing models used in the study of motility-induced phase separation (MIPS) and models of congestion of traffic flow on a single-lane highway. Particles have a target velocity with a density-dependent magnitude and a direction that flips with a finite rate that is biased toward moving right. Two key parameters are the bias and the speed relaxation time. MIPS is known to occur in such models at zero bias and zero relaxation time (overdamped dynamics), while a fully biased motion with no velocity reversal models traffic flow on a highway. Using both numerical simulations and continuum equations derived from the microscopic dynamics, we show that a single phase-separated state extends from the usual MIPS to congested traffic flow in the phase diagram defined by the bias and the speed relaxation time. However, in the fully biased case, inertia is essential to observe phase separation, making MIPS and congested traffic flow seemingly different phenomena if not simultaneously considering inertia and tumbling. We characterize the velocity of the dense phase, which is static for usual MIPS and moves backwards in traffic congestion. We also find that in presence of bias, the phase diagram becomes richer, with an additional transition between phase separation and a microphase separation that is seen above a threshold bias or relaxation rate.

cond-mat.stat-mech↗

arXiv2023

Tailoring the overlap distribution in driven mean-field spin models

In a statistical physics context, inverse problems consist in determining microscopic interactions such that a system reaches a predefined collective state. A complex collective state may be prescribed by specifying the overlap distribution between microscopic configurations, a notion originally introduced in the context of disordered systems like spin-glasses. We show that in spite of the absence of disorder, nonequilibrium spin models exhibiting spontaneous magnetization oscillations provide a benchmark to prescribe a non-trivial overlap distribution with continuous support, qualitatively analogous to the ones found in disordered systems with full replica symmetry breaking. The overlap distribution can be explicitly tailored to take a broad range of predefined shapes by monitoring the spin dynamics. The presence of a non-trivial overlap distribution is traced back to an average over infinitely many pure states, a feature shared with spin-glasses, although the structure of pure states is here much simpler.

cond-mat.stat-mech↗

arXiv2023

Nonlinear dielectric response in glasses: restoring forces and avoided spin-glass criticality

Experimental measurements of nonlinear dielectric response in glassformers like supercooled glycerol or propylene carbonate have been interpreted as providing evidence for a growing thermodynamic length scale when lowering temperature. A heuristic picture based on coherently flipping `superdipoles' with disordered internal structure has been argued to capture the essence of the experimentally reported behavior, pointing to the key role of effectively disordered interactions in structural glasses. We test these ideas by devising an explicit one-dimensional model of interacting spins incorporating both the spin-glass spirit of the superdipole argument, and the necessary long-time decorrelation of structural disorder, encoded here in a slow dynamics of the coupling constants. The frequency-dependent third-order response of the model qualitatively reproduces the typical humped shape reported in experiments. The temperature dependence of the maximum value is also qualitatively reproduced. In contrast, the humped shape of the third-order response is not reproduced by a simple kinetically constrained spin model with non-interacting spins. To rationalize these results, we propose a two-length-scale scenario by distinguishing between the characteristic length of dynamical heterogeneities and a coherence length that monitors the effect of interactions. We show that both length scales are identical in the kinetically constrained spin model, while they have significantly different dynamics in the model of interacting spins.

cond-mat.dis-nn↗

arXiv2023

Discontinuous phase transition from ferromagnetic to oscillating states in a nonequilibrium mean-field spin model

We study a nonequilibrium ferromagnetic mean-field spin model exhibiting a phase with spontaneous temporal oscillations of the magnetization, on top of the usual paramagnetic and ferromagnetic phases. This behavior is obtained by introducing dynamic field variables coupled to the spins through non-reciprocal couplings. We determine a nonequilibrium generalization of the Landau free energy in terms of the large deviation function of the magnetization and of an appropriately defined smoothed stochastic time derivative of the magnetization. While the transition between paramagnetic and oscillating phase is continuous, the transition between ferromagnetic and oscillating phases is found to be discontinuous, with a coexistence of both phases, one being stable and the other one metastable. Depending on parameter values, the ferromagnetic points may either be inside or outside the limit cycle, leading to different transition scenarios. The stability of these steady states is determined from the large deviation function. We also show that in the coexistence region, the entropy production has a pronounced maximum as a function of system size.

cond-mat.stat-mech↗

arXiv2022

Nonequilibrium Phase Transition To Temporal Oscillations In Mean-Field Spin Models

We propose a mean-field theory for nonequilibrium phase transitions to a periodically oscillating state in spin models. A nonequilibrium generalization of the Landau free energy is obtained from the join distribution of the magnetization and its smoothed stochastic time derivative. The order parameter of the transition is a Hamiltonian, whose nonzero value signals the onset of oscillations. The Hamiltonian and the nonequilibrium Landau free energy are determined explicitly from the stochastic spin dynamics. The oscillating phase is also characterized by a non-trivial overlap distribution reminiscent of a continuous replica symmetry breaking, in spite of the absence of disorder. An illustration is given on an explicit kinetic mean-field spin model.

cond-mat.stat-mech↗