We study statistical properties of atmospheric particulate matter fluctuations using six years of daily ${\mathrm{PM}}_{2.5}$ concentration data from 54 Indian cities. Despite diverse urban settings and heterogeneous climatic conditions, we find that the fluctuations show strikingly universal behavior in both the distributional properties and temporal dynamics. After removing slow trends and seasonal components, the rescaled probability density functions of the residual fluctuations collapse onto a single curve and are well described by an exponentially modified Gaussian distribution. The rescaled residual time series for all the cities further exhibit certain robust dynamical features, with similar decay of autocorrelation functions, and power spectral densities displaying a $\sim 1/f$ decay at the tails. Finally, we propose a minimal stochastic model for the residual dynamics, which explains the observed universal features—the stationary distribution, temporal correlation, and spectral scaling.

Phase frustration is often used to represent an effective phase lag caused by signal transmission, sensing, processing, or response delays. In conventional phase-oscillator models, a fixed phase lag usually weakens the attractive component of the coupling and tends to hinder synchronization. Here we show that, when phase frustration is adaptively regulated by the collective state, it can instead provide a feedback mechanism for hysteretic synchronization transitions. In many oscillator populations, the effective phase lag need not remain fixed. When coherence is weak, the collective rhythm is difficult to predict and the effective phase mismatch can be large. When coherence grows, the rhythm becomes more predictable, so feedback or phase correction may reduce this mismatch. Motivated by this physical picture, we consider a population of globally coupled phase oscillators in which the frustration angle decreases with the global order parameter. This adaptive phase frustration creates a positive feedback between growing coherence and the effective attractive interaction. Numerical scans show that this feedback can change a continuous transition into a hysteretic jump with coexisting coherent states. The transition persists when the feedback exponent and frequency width are varied, remains visible in finite-size tests, and can also appear in networked systems. For the globally coupled Lorentzian case, the Ott-Antonsen reduction gives a low-dimensional explanation of the coherent branch and the forward and backward transition points. These results show that adaptive regulation of phase frustration provides a physically motivated phase-lag feedback mechanism for bistability and hysteretic synchronization transitions.

We study the percolation of strongly connected clusters (SCCs), in which sites are mutually reachable through directed paths, in systems with randomly oriented bonds by extensive simulations on hypercubic lattices from dimensions $d=2$ to 7 and on complete graphs. Below the upper critical dimension $d_u=6$, critical SCCs exhibit a nontrivial fractal dimension $d_{\mathrm{SCC}}$ with a nonmonotonic dependence on dimension: $d_{\mathrm{SCC}}$ reaches a maximum at $d=4$, in contrast to the fractal dimension of ordinary percolation clusters, which increases monotonically with $d$. The SCC size distribution follows $\sim s^{-{\tau }_{\mathrm{SCC}}}$, where the Fisher exponent is consistent with the hyperscaling relation ${\tau }_{\mathrm{SCC}}=1+d/d_{\mathrm{SCC}}$ and, unlike $d_{\mathrm{SCC}}$ itself, increases monotonically with $d$. For $d\ge d_u$, mean-field behavior is recovered with $d_{\mathrm{SCC}}/d=1/3$, consistent with complete-graph results. However, unlike finite-dimensional lattices, complete graphs exhibit a double-scaling structure in the SCC size distribution: Large SCCs are governed by the mean-field exponent ${\tau }_{\mathrm{SCC}}=4$, while small SCCs follow a distinct power law with exponent ${\tau }^{\prime }=1$. At $d_u=6$, logarithmic corrections are also observed. These results show that SCCs remain well-defined fractal objects across dimensions, exhibiting both a nontrivial dimensional dependence and a nontrivial crossover from finite-dimensional to mean-field behavior.

We study the run-and-tumble motion of a single autochemotactic agent that secretes and senses chemical cues using a reinforcement learning (RL) framework. In a chemical environment that changes dynamically via secretion, diffusion, and degradation of attractant chemicals, we study the effectiveness of RL for a successful navigation strategy. We show that RL enhances chemotactic performance. Starting from a random initial position, the probability of finding the agent in the favorable region (near the attractant peak) increases faster with time and reaches a higher value in the long-time limit when RL is used, compared to the case when the agent performs usual run-and-tumble chemotaxis without using RL. We also characterize the chemotactic performance of the agent by varying different parameters like rates of secretion, degradation, and diffusion. Importantly, we observe a nonmonotonic dependence of the chemotactic performance on the learning and exploration rates, indicating the existence of an optimal RL strategy for the most efficient navigation. We explain these observations as a result of the interplay between the preexisting attractant environment and the agent-generated attractant profile.

Molecular dynamics simulations are employed to investigate the tensile properties of water confined in carbon nanotubes (CNTs) with varying diameters, where CNTs serve as a prototype for xylem vessels critical to long-distance water transport in tall trees. Our results reveal that nanoconfined water exhibits unexpectedly ultrahigh tensile strength (∼126 MPa), which is significantly higher than that of glass (∼33 MPa). Meanwhile, the tensile strength of confined water is highly dependent on the nanotube diameter. The underlying mechanism for this unique tensile behavior is attributed to the synergistic effect of hydrogen bonding and nanoscale confinement. This work provides key insights into understanding the mechanism of water transport from roots to the canopy in tall trees.

We have developed explicit analytic expressions for the electron and ion electrical-conductivity tensors of a fully ionized two-component plasma. We assume the plasma is nondegenerate, nonrelativistic, and subject to electric and magnetic fields. The conductivity tensors give the Pedersen, Hall, and Birkeland currents of the electrons and ions as functions of the electric field, magnetic field, and plasma parameters. (By definition, the electron and ion Pedersen currents are those that flow in the direction of the component of E that is perpendicular to B. The electron and ion Hall currents flow in the $-\mathbf{E}\times \mathbf{B}$ and $\mathbf{E}\times \mathbf{B}$ directions, respectively. The electron and ion Birkeland currents flow in the direction of the component of E that is parallel to B.) We derive the conductivity tensors from Lorentz-force, kinetic, modified-two-fluid, and standard-two-fluid models of a plasma. The Lorentz-force model is based on the electron and ion Lorentz-force equations, and the probability distribution of random collisions. The kinetic model is based on the electron and ion Boltzmann equations in the relaxation-time approximation. The modified-two-fluid model is derived from the kinetic model, and differs from the well-known standard-two-fluid model. The Lorentz-force, kinetic, and modified-two-fluid models are in agreement and, in the steady state, predict the following: In the collisionless limit $({\omega }_{ce}\tau \to \infty ,$ ${\omega }_{ci}\tau \to \infty )$, the electron and ion Pedersen currents $\to 0$, whereas the electron and ion Hall currents dominate. In the collisional limit $({\omega }_{ce}\tau \to 0$, ${\omega }_{ci}\tau \to 0)$, the Pedersen currents dominate, whereas the Hall currents $\to 0$. (The quantities ${\omega }_{ce}$ and ${\omega }_{ci}$ are the electron- and ion-cyclotron frequencies, respectively; τ is the interparticle momentum-transfer collision time.) The results are consistent with elementary physical considerations. The standard-two-fluid model predicts instead that, in the steady state, the Pedersen currents are identically 0, and the Hall currents dominate, for all values of ${\omega }_{ce}\tau$ and ${\omega }_{ci}\tau$ (when $\mathbf{B}\ne 0$). The discrepancies with the Lorentz-force, kinetic, and modified-two-fluid models are caused by the collision terms of the standard-two-fluid model.

Granular clogging at narrow constrictions is commonly described as a memoryless Poisson process, capturing the stochastic onset of clogs but not the emergence of mechanically stable arches. Using discrete-element simulations of a vibrated two-dimensional hopper with repeated clogging events, we reveal a two-stage mechanism of clog formation: random Poissonian triggering followed by a finite structural-consolidation process that stabilizes the incipient arch. Tracking competing outlet-spanning chains further shows that the eventual final arch separates from transient alternatives only during this interval, through reduced geometric defects and strengthened weak contacts. Although consolidation durations vary strongly across events, their evolution is confined to a narrow directed trajectory in a state space defined by obstacle packing fraction, particle velocity, and effective coordination number. A graph neural network trained solely on particle-contact graphs independently recovers this ordering, showing that instantaneous configurations encode the consolidation state and clogging probability. Thus, clogging is random in time at initiation but predictable from configuration after structural consolidation begins.

We consider a two-dimensional athermal binary mixture of Lennard-Jones particles with persistent random active forces. The liquid phase of this system for active forces exceeding a threshold value exhibits self-organization with long-range spatial correlations of particle velocities and active forces. We study with simulations the development of these correlations from a random initial state. Several characteristics of the growth of correlations are measured and compared with those of phase-ordering kinetics of equilibrium systems after a quench from a disordered state. The motion of the particles in the long-time steady state is found to be dominated by two streams that flow in opposite directions.

Usually, seismic analyses are based on magnitudes which are obtained from models based on the energy released by each earthquake. The present approach supplements this description by including the intervals between consecutive events. This is justified by arguments coming from fundamental physics: energy and time are canonical variables, so they offer complementary and equally acceptable descriptions of the same phenomenon. We then apply both schemes to four seismic zones along the coast of the Pacific Ocean (part of the “circle of fire”) surrounding the following major earthquakes: $Mw8.2$ in Alaska (2021), $Mw7.1$ in California (2019), $Mw8.2$ in Iquique (2014), and $Mw8.8$ in Maule (2010). Following a basic statistical characterization, we applied mutability and diversity analyses—two information theory techniques related to Shannon entropy—to the magnitude and interval sequences. The analysis of the interval sequences provides an accurate description of the seismic activity. Magnitude and intervals present opposite variations in the dynamical analyses, thus offering complementary descriptions of the same phenomenon. However, the results for intervals render a more direct interpretation of the seismic sequence.

Bifurcation delay, the postponement of instability during slow parameter variation, is a generic feature of nonautonomous dynamical systems. We study how diffusive coupling, unequal time scales, and communication lags alter this delay in pairs of autonomous and nonautonomous oscillators, using an ecological canard model and the FitzHugh-Nagumo neuron model. The delay depends on the dynamical state of the autonomous oscillator: coupling to a stable autonomous oscillator increases it linearly with the coupling strength, while coupling to an oscillatory autonomous oscillator reduces it. An internal time-scale ratio controls how fast the nonautonomous unit grows beyond threshold and yields a closed-form asymptotic law for the delay. A finite communication lag produces a periodic modulation of the delay whose period is set solely by the Hopf frequency of the nonautonomous oscillator and is independent of the sweep rate. Perturbation analysis of the delayed characteristic equation gives a quadratic amplitude law, which we confirm numerically. These results extend the static amplitude-death condition of autonomous coupled oscillators to the slow-passage setting and give analytical expressions for delayed onset in coupled multiscale systems.

Many open problems across physical, biological, and engineered systems involve nonequilibrium transport processes where the governing conservation laws are known, but the underlying constitutive relations remain latent and time varying. Conventional data-driven approaches like deep neural networks capture statistical patterns but routinely violate fundamental mass conservation. Here I introduce a unified Bayesian inverse transport framework that resolves this by embedding Bayesian neural networks directly within exact partial differential equations in age-time space. By evaluating this framework on complex, real-world human cohort advection across China, Japan, and South Korea, I demonstrate that this physical constraint enables consistent uncertainty propagation and missing-data reconstruction from sparse observations. Beyond demography, this framework provides a generalizable foundation for observing and forecasting nonequilibrium boundary dynamics across various fields.

Real-world networked systems, such as power grids, operate inevitably under various types of noise originating from distinctive sources, which brings uncertainty and risks for systems' safe operation. As a major noise type, colored noise may enter the power system through multiple channels. For instance, through the large-scale integration of renewable energy sources and the demand-side power fluctuations, colored noise impinges on the stochastic frequency responses of power grids in a way that is not yet fully understood to date. In this work we systematically investigate the impact of noise color on the statistical properties of dynamic power grid frequency responses. Through abundant numerical simulations of a paradigmatic Kuramoto-like grid model subject to $1/f^b$ noise with a tunable color, we demonstrate that noise with a steeper power spectrum $(b\gtrsim 1)$ dominated by low-frequency components drives power grid responses into a regime of nonstationarity and non-Gaussianity, while the responses are Gaussian for $b\lesssim 1$ despite the non-Gaussianity of the noise. We further show that this transition at $b\approx 1$ occurs independent of the specific topological structure underlying the power grid systems. This suggests that the intermittent renewable power feed-in with Kolmogorov exponent $5/3$ may be a source of the observed non-Gaussian distribution of power grid frequency. Our results may thereby help better understand and control the collective frequency fluctuations in modern power grids.

Rumors now spread faster and wider than ever, driven by diverse communication channels. To capture this complexity, we develop a multilayer coupled model of rumor propagation that integrates individual behavioral differences, mass media influence, social reinforcement, and the interplay of cooperative and competitive dynamics across channels. Using mean-field theory, we analyze the system's dynamics and derive the critical threshold through the next-generation matrix method. Numerical simulations show that traditional mass media interventions are more effective in limiting single-channel spreaders, while new media interventions have stronger effects on dual-channel spreaders. However, when traditional mass media pressure is already high, intensifying new media interventions may fail to suppress offline spreaders and can instead raise costs or even cause negative outcomes. We also find that dual-channel spreaders are more responsive to changes in online rumor prevalence and play a stronger role in overall rumor dynamics than single-channel spreaders. Finally, validation with a Twitter dataset confirms the model's predictive power for offline rumor spread.

The replisome, a protein complex that drives DNA replication, comprises multiple DNA polymerases. During replication, one polymerase synthesizes the leading strand continuously, while another synthesizes the lagging strand discontinuously, producing short transient segments known as Okazaki fragments and gaps between them. After the synthesis of each Okazaki fragment, the lagging-strand polymerase dissociates either spontaneously or upon collision with the preceding fragment; however, the relative contributions of these mechanisms remain unclear. Here, we develop a biophysical model of semidiscontinuous replication by incorporating the stochastic dynamics of the lagging-strand polymerase. By computing the Okazaki fragment and gap-size distributions, we show the significance of polymerase dissociation in shaping these statistics. Applying the model to the T4 bacteriophage and B. subtilis replisomes, we find that collisions with preceding Okazaki fragments predominantly trigger polymerase dissociation.

Rendezvous is a critical task for multiagent systems, requiring agents to coordinate to meet at an unspecified location. However, achieving this in fluid environments presents a challenge, as it remains unclear how agents can exploit underlying fluid kinematics to facilitate convergence. In this study, we adopt a multiagent reinforcement learning (MARL) approach to develop rendezvous strategies in vortical flows. Compared to a naive strategy, where agents navigate toward their counterparts, MARL strategies significantly improve the rendezvous rate. MARL strategies also show transferability across varying vortex intensities, vortex scales, and swarm sizes. By breaking the symmetry of the state-action map, MARL strategy leverages a nonintuitive mechanism that prevents agents from becoming trapped in separate vortices, thereby enhancing rendezvous success. Additionally, a heuristic strategy is extracted from the learned strategy and also outperforms the naive strategy. Furthermore, a theoretical analysis demonstrates that large finite-time Lyapunov exponents identify where fluid effects separate adjacent agents and impedes rendezvous. Our findings reveal the important role that agent-fluid interactions play in multi-agent tasks and highlight the MARL capability to explore swarm intelligence in complex flow environments.

We propose a hybrid sampling method, tensor-network population annealing (TNPA), which combines tensor-network (TN) initialization with population annealing (PA). We apply this method to the two-dimensional $\pm J$ Edwards-Anderson Ising spin glass. The approach is motivated by the limitations of existing methods: TN-based samplers can become numerically unstable in frustrated spin systems at low temperatures, whereas conventional PA requires a long annealing schedule when started from the high-temperature limit. In TNPA, TN contractions are used only within a reliable temperature range to generate initial configurations that are close to the equilibrium distribution. The subsequent low-temperature equilibration is then carried out by PA. To stabilize the initialization process, we introduce a diagnostic based on the effective sample size that adaptively selects the initialization temperature. The proposed framework provides a practical and physically motivated route to low-temperature sampling by combining the complementary strengths of TN and PA.

Adhesive failure of soft polymer networks underpins the mechanical integrity of technologies ranging from packaging to electronics. Yet predictive descriptions of this behavior remain limited because the two main sources of energy dissipation at the peel front, friction and damage, are coupled in ways that are unknown. In prior work, we showed that the damage contribution to the adherence energy is intrinsically rate dependent [A. Arrowood et al., Proc. Natl. Acad. Sci. USA 122, e2509322122 (2025)], challenging viscoelastic fracture theories that treat it as constant. Here we provide further evidence for this picture by identifying a micro-stick-slip instability that emerges from this rate dependence. Using soft acrylic networks labeled with fluorogenic mechanophores, we correlate peel-front kinematics with molecular damage and show that, even under globally steady peeling, the crack front can alternate between a bulged, highly deformed, highly damaged state and a rounded, weakly deformed, weakly damaged state. The onset of this instability is governed by the network relaxation time, and thus by chemical composition and glass transition temperature, whereas the slip speed is $\approx 2.5\times$ the imposed peel speed and the crack-advance length remains nearly constant at 0.26 mm $\pm$ 0.04. Unlike classical stick-slip, which produces large force oscillations and centimeter-scale crack jumps, this micro-stick-slip instability generates only weak force fluctuations and instead reflects localized weakening within the damage zone. Overall, our results reveal a direct coupling between interfacial stress transfer and molecular damage at the peel front, motivating nested cohesive-zone models that incorporate rate-dependent molecular damage to predict adhesive fracture.

The interplay between activity and competing interaction length scales remains largely unexplored, despite its relevance to many soft and biological systems. Here, we study active Brownian particles interacting through a ramp-like core-softened potential that exhibits water-like anomalies in equilibrium. By varying the activity over a broad range of densities along two representative isotherms, one within the anomalous region and the other above it, we examine how self-propulsion modifies the structure and dynamics of the fluid. To gain local structural insight into how activity modifies local structure, we construct effective interactions from the steady-state pair correlations using iterative Boltzmann inversion. We find that activity progressively suppresses the anomalies of the passive fluid, although signatures of the underlying structural crossover remain visible in normalized quantities. The effective interactions reveal that self-propulsion lowers the distinction between the local environments and facilitates population transfer between the two characteristic length scales. These results indicate that activity primarily acts by facilitating population transfer between the two local environments, thereby reducing the structural competition responsible for the anomalous response.

We study collective ordering in confined trains of passive deformable particles driven by Poiseuille flow. An isolated particle adopts a slipperlike shape with two mirror-related orientations with respect to the flow direction, its tail tilting either upward or downward. We call a train polarized when all particles share the same inclination state. When oppositely inclined neighbors approach closely enough, the downstream particle can switch to match the upstream state, while the upstream particle remains unchanged. This directional switching arises from the fore-aft asymmetry of the slipper shape and its disturbance flow. Repeated along the train, this pairwise event generates a downstream-propagating switching front. A minimal bistable model with unequal upstream and downstream couplings captures both front propagation and arrest. In the propagating regime, interactions between fronts drive domain coarsening and can ultimately yield uniform polarization. When the interparticle coupling becomes too weak, fronts arrest, leaving persistent polarized domains. These results reveal a hydrodynamic route to collective polarization in externally driven suspensions of passive deformable particles, without self-propulsion or externally imposed asymmetric interactions.