Hysteresis is a major feature of the solid-liquid transition in granular materials. This property, by allowing metastable states, can potentially yield catastrophic phenomena such as earthquakes or aerial landslides. The origin of hysteresis in granular flows is still debated. However, most mechanisms put forward so far rely on the presence of inertia at the particle level. In this paper, we study the avalanche dynamics of non-Brownian suspensions in slowly rotating drums and reveal large hysteresis of the avalanche angle even in the absence of inertia. By using microsilica particles whose interparticle friction coefficient can be turned off, we show that microscopic friction, conversely to inertia, is key to triggering hysteresis in granular suspensions. To understand this link between friction and hysteresis, we use the rotating drum as a rheometer to extract the suspension rheology close to the flow onset for both frictional and frictionless suspensions. This analysis shows that the flow rule for frictionless particles is monotonous and follows a power law of exponent $\alpha =0.37\pm 0.05$, in close agreement with the previous theoretical prediction, $\alpha =0.35$. By contrast, the flow rule for frictional particles suggests a velocity-weakening behavior, thereby explaining the flow instability and the emergence of hysteresis. These findings show that hysteresis can also occur in particulate media without inertia, questioning the intimate nature of this phenomenon. By highlighting the role of microscopic friction, our results may be of interest in the geophysical context to understand the failure mechanism at the origin of undersea landslides.

Mutualistic interactions, those that are naturally beneficial for both interacting species, are recurrently found in ecosystems. Observations of natural systems show that if we draw mutualistic relationships as links between species, the resulting mutualistic network of interactions displays a widespread particular ordering called nestedness. In such an ordering, the mutualistic partners of a given species conform a subset of the partners of all species with larger degree, that is, of those species having more interactions. On the other hand, theoretical works show that a nested structure has a positive impact on a number of relevant features of mutualistic communities ranging from species coexistence to structural stability and biodiversity. However, how nestedness emerges and what are its determinants, are still open challenges that have led to multiple debates to date. Here we show, by applying a theoretical approach to the analysis of 167 real mutualistic networks, that nestedness is not an irreducibly macroscopic feature but an entropic consequence of the degree sequences (number of mutualistic interactions of each species). Remarkably, we find that an outstanding majority of the analyzed networks does not show statistically significant nestedness. These findings point to the need of revising previous claims about the role of nestedness and might contribute to expand our understanding of how evolution shapes mutualistic interactions and communities by placing the focus on the node-dependent properties rather than on global quantities.

In prokaryotes, the number of genes in different functional classes shows apparent universal scaling with the total number of genes that can be approximated by a power law, with a sublinear, near-linear, or superlinear scaling exponent. These dependences are gene class specific but hold across the entire diversity of bacteria and archaea. Several models have been proposed to explain these universal scaling laws, primarily based on the specifics of the respective biological functions. However, a population-genetic theory of universal scaling is lacking. We employ a simple mathematical model for prokaryotic genome evolution, which, together with the analysis of 34 clusters of closely related bacterial genomes, allows us to identify the underlying factors that govern the evolution of the genome content. Evolution of the gene content is dominated by two functional class-specific parameters: selection coefficient and genome plasticity. The selection coefficient quantifies the fitness cost associated with deletion of a gene in a given functional class or the advantage of successful incorporation of an additional gene. Genome plasticity reflects both the availability of the genes of a given class in the external gene pool that is accessible to the evolving population and the ability of microbes to accommodate these genes in the short term, that is, the class-specific horizontal gene transfer barrier. The selection coefficient determines the gene loss rate, whereas genome plasticity is the principal determinant of the gene gain rate.

We present a contact-based model to study the spreading of epidemics by means of extending the dynamic message-passing approach to temporal networks. The shift in perspective from node- to edge-centric quantities enables accurate modeling of Markovian susceptible-infected-recovered outbreaks on time-varying trees, i.e., temporal networks with a loop-free underlying topology. On arbitrary graphs, the proposed contact-based model incorporates potential structural and temporal heterogeneities of the contact network and improves analytic estimations with respect to the individual-based (node-centric) approach at a low computational and conceptual cost. Within this new framework, we derive an analytical expression for the epidemic threshold on temporal networks and demonstrate the feasibility of this method on empirical data.

Machine-learning algorithms can be fooled by small well-designed adversarial perturbations. This is reminiscent of cellular decision-making where ligands (called antagonists) prevent correct signaling, like in early immune recognition. We draw a formal analogy between neural networks used in machine learning and models of cellular decision-making (adaptive proofreading). We apply attacks from machine learning to simple decision-making models and show explicitly the correspondence to antagonism by weakly bound ligands. Such antagonism is absent in more nonlinear models, which inspires us to implement a biomimetic defense in neural networks filtering out adversarial perturbations. We then apply a gradient-descent approach from machine learning to different cellular decision-making models, and we reveal the existence of two regimes characterized by the presence or absence of a critical point for the gradient. This critical point causes the strongest antagonists to lie close to the decision boundary. This is validated in the loss landscapes of robust neural networks and cellular decision-making models, and observed experimentally for immune cells. For both regimes, we explain how associated defense mechanisms shape the geometry of the loss landscape and why different adversarial attacks are effective in different regimes. Our work connects evolved cellular decision-making to machine learning and motivates the design of a general theory of adversarial perturbations, both for in vivo and in silico systems.

During mitosis, human chromosomes are linearly compacted about 1000-fold by loop-extruding motors. Recent experiments have shown that condensins extrude DNA loops but in a “one-sided” manner. This contrasts with existing models, which predict that symmetric, “two-sided” loop extrusion accounts for mitotic chromosome compaction. We explore whether one-sided extrusion, as it is currently seen in experiments, can compact chromosomes by developing a mean-field theoretical model for polymer compaction by motors that actively extrude loops and dynamically turnover. The model establishes a stringent upper bound of only about tenfold for compaction by strictly one-sided extrusion. We confirm this result with stochastic simulations. Thus, strictly one-sided extrusion as it has been observed so far cannot be the sole mechanism of chromosome compaction. However, as shown by the model, other two-sided or effectively two-sided mechanisms can achieve sufficient compaction.

Recent success in deep neural networks has generated strong interest in hardware accelerators to improve speed and energy consumption. This paper presents a new type of photonic accelerator based on coherent detection that is scalable to large ($N\gtrsim {10}^6$) networks and can be operated at high (gigahertz) speeds and very low (subattojoule) energies per multiply and accumulate (MAC), using the massive spatial multiplexing enabled by standard free-space optical components. In contrast to previous approaches, both weights and inputs are optically encoded so that the network can be reprogrammed and trained on the fly. Simulations of the network using models for digit and image classification reveal a “standard quantum limit” for optical neural networks, set by photodetector shot noise. This bound, which can be as low as $50\mathrm{zJ}/\mathrm{MAC}$, suggests that performance below the thermodynamic (Landauer) limit for digital irreversible computation is theoretically possible in this device. The proposed accelerator can implement both fully connected and convolutional networks. We also present a scheme for backpropagation and training that can be performed in the same hardware. This architecture will enable a new class of ultralow-energy processors for deep learning.

Each living species carries a complex DNA sequence that determines their unique features and functionalities. It is generally assumed that life started from a random pool of oligonucleotide sequences, generated by a prebiotic polymerization of nucleotides. The mechanism that initially facilitated the emergence of sequences that code for the function of the first species from such a random pool of sequences remains unknown. It is a central problem of the origin of life. An interesting option would be a self-selection mechanism by spontaneous symmetry breaking. Initial concentration fluctuations of specific sequence motifs would have been amplified and outcompeted less abundant sequences, enhancing the signal to noise to replicate and select functional sequences. Here, we demonstrate with experimental and theoretical findings that templated ligation would provide such a self-selection. In templated ligation, two adjacent single sequences strands are chemically joined when a third complementary strand sequence brings them in close proximity. This simple mechanism is a likely side product of a prebiotic polymerization chemistry once the strands reach the length to form double-stranded species. As shown here, the ligation gives rise to a nonlinear replication process by the cooperative ligation of matching sequences which self-promote their own elongation. This process leads to a cascade of enhanced template binding and faster ligation reactions. A requirement is the reshuffling of the strands by thermal cycling, enabled, for example, by microscale convection. By using a limited initial sequence space and performing long-term ligations, we find that complementary sequences with an initially higher concentration prevail over either noncomplementary or less-concentrated sequences. The latter die out by the molecular degradation that we simulate in the experiment by serial dilution. The experimental results are consistent with both explicit and abstract theory models that are generated considering the ligation rates determined experimentally. Previously, other nonlinear modes of replication such as hypercycles have been discussed to overcome instabilities from first-order replication dynamics such as the error catastrophe and the dominance of structurally simple but fast-replicating sequences, known as the Spiegelman problem. Assuming that templated ligation is driven by the same chemical mechanism that generates prebiotic polymerization of oligonucleotides, the mechanism could function as a missing link between polymerization and the self-stabilized replication, offering a pathway to the autonomous emergence of Darwinian evolution for the origin of life.

In this paper, we propose a new approach to measuring ultrafast dynamics with free-electron lasers (FELs). Ultrafast experiments are among the most promising avenues of research at x-ray FELs, with the potential to reveal the chemical dynamics of charge separation, conical intersection crossing, and biologically mediated reactions. Pump-probe scanning is the standard approach to measure dynamics at x-ray FELs, but at the shortest timescales, and particularly for x-ray pump, x-ray probe experiments, the scans require challenging beam setups and can introduce systematic errors. Here, we propose an alternative approach using the randomness of the self-amplified spontaneous emission (SASE) process to drive many simultaneous pump-probe experiments on each shot. Measuring the fluctuations in the incident beam’s time profile on a shot-to-shot basis enables the reconstruction of ultrafast dynamics down to the coherence length of the FEL without the need for pump-probe scans. Because of similarity to ghost imaging, in which spatial properties are reconstructed by measuring the incident probe’s transverse properties, we call this “pump-probe ghost imaging.” In this paper, we describe the method and simulate an example experiment. We also describe an alternative implementation that uses only spectral measurements, avoiding the need for direct time-domain diagnostics and extending the method to the attosecond regime.

Analytical approaches to model the structure of complex networks can be distinguished into two groups according to whether they consider an intensive (e.g., fixed degree sequence and random otherwise) or an extensive (e.g., adjacency matrix) description of the network structure. While extensive approaches—such as the state-of-the-art message passing approximation—typically yield more accurate predictions, intensive approaches provide crucial insights on the role played by any given structural property in the outcome of dynamical processes. Here we introduce an intensive description that yields almost identical predictions to the ones obtained with the message passing approximation using bond percolation as a benchmark. Our approach distinguishes nodes according to two simple statistics: their degree and their position in the core-periphery organization of the network. Our near-exact predictions highlight how accurately capturing the long-range correlations in network structures allows easy and effective compression of real complex network data.

An algorithmically hard phase is described in a range of inference problems: Even if the signal can be reconstructed with a small error from an information-theoretic point of view, known algorithms fail unless the noise-to-signal ratio is sufficiently small. This hard phase is typically understood as a metastable branch of the dynamical evolution of message-passing algorithms. In this work, we study the metastable branch for a prototypical inference problem, the low-rank matrix factorization, that presents a hard phase. We show that, for noise-to-signal ratios that are below the information-theoretic threshold, the posterior measure is composed of an exponential number of metastable glassy states, and we compute their entropy, called the complexity. We show that this glassiness extends even slightly below the algorithmic threshold, below which the well-known approximate message-passing (AMP) algorithm is able to closely reconstruct the signal. Counterintuitively, we find that the performance of the AMP algorithm is not improved by taking into account the glassy nature of the hard phase. This result provides further evidence that the hard phase in inference problems is algorithmically impenetrable for some deep computational reasons that remain to be uncovered.

Time-dependent processes are often analyzed using the power spectral density (PSD) calculated by taking an appropriate Fourier transform of individual trajectories and finding the associated ensemble average. Frequently, the available experimental datasets are too small for such ensemble averages, and hence, it is of a great conceptual and practical importance to understand to which extent relevant information can be gained from $S(f,T)$, the PSD of a single trajectory. Here we focus on the behavior of this random, realization-dependent variable parametrized by frequency $f$ and observation time $T$, for a broad family of anomalous diffusions—fractional Brownian motion with Hurst index $H$—and derive exactly its probability density function. We show that $S(f,T)$ is proportional—up to a random numerical factor whose universal distribution we determine—to the ensemble-averaged PSD. For subdiffusion ($H<1/2$), we find that $S(f,T)\sim A/f^{2H+1}$ with random amplitude $A$. In sharp contrast, for superdiffusion $(H>1/2)$ $S(f,T)\sim B{T}^{2H-1}/f^2$ with random amplitude $B$. Remarkably, for $H>1/2$ the PSD exhibits the same frequency dependence as Brownian motion, a deceptive property that may lead to false conclusions when interpreting experimental data. Notably, for $H>1/2$ the PSD is ageing and is dependent on $T$. Our predictions for both sub- and superdiffusion are confirmed by experiments in live cells and in agarose hydrogels and by extensive simulations.

Recent work using tools from quantum information theory has shown that for small systems where quantum effects become prevalent, there is not one thermodynamical second law but many. Derivations of these laws assume that an experimenter has very precise control of the system and heat bath. Here we show that these multitude of laws can be saturated using two very simple operations: changing the energy levels of the system and thermalizing over any two system energy levels. Using these two operations, one can distill the optimal amount of work from a system, as well as perform the reverse formation process. What is more, using only these two operations and one ancilla qubit in a thermal state, one can transform any state into any other state allowable by the second laws. We thus have the result that the second laws hold for fine-grained manipulation of system and bath, but can be achieved using very coarse control. This brings the full array of thermal operations towards a regime accessible by experiment, and establishes the physical relevance of these second laws, potentially opening a new direction of studies.

Complex real-world phenomena across a wide range of scales, from aviation and Internet traffic to signal propagation in electronic and gene regulatory circuits, can be efficiently described through dynamic network models. In many such systems, the spectrum of the underlying graph Laplacian plays a key role in controlling the matter or information flow. Spectral graph theory has traditionally prioritized analyzing unweighted networks with specified adjacency properties. Here, we introduce a complementary framework, providing a mathematically rigorous weighted graph construction that exactly realizes any desired spectrum. We illustrate the broad applicability of this approach by showing how designer spectra can be used to control the dynamics of various archetypal physical systems. Specifically, we demonstrate that a strategically placed gap induces generalized chimera states in Kuramoto-type oscillator networks, tunes or suppresses pattern formation in a generic Swift-Hohenberg model, and leads to persistent localization in a discrete Gross-Pitaevskii quantum network. Our approach can be generalized to design continuous band gaps through periodic extensions of finite networks.

Autonomous, randomly coupled, neural networks display a transition to chaos at a critical coupling strength. Here, we investigate the effect of a time-varying input on the onset of chaos and the resulting consequences for information processing. Dynamic mean-field theory yields the statistics of the activity, the maximum Lyapunov exponent, and the memory capacity of the network. We find an exact condition that determines the transition from stable to chaotic dynamics and the sequential memory capacity in closed form. The input suppresses chaos by a dynamic mechanism, shifting the transition to significantly larger coupling strengths than predicted by local stability analysis. Beyond linear stability, a regime of coexistent locally expansive but nonchaotic dynamics emerges that optimizes the capacity of the network to store sequential input.

We demonstrate that the pulsed-time structure and high-peak ion intensity provided by the laser-ablation process can be directly combined with the high resolution, high efficiency, and low background offered by collinear resonance ionization spectroscopy. This simple, versatile, and powerful method offers new and unique opportunities for high-precision studies of atomic and molecular structures, impacting fundamental and applied physics research. We show that even for ion beams possessing a relatively large energy spread, high-resolution hyperfine-structure measurements can be achieved by correcting the observed line shapes with the time-of-flight information of the resonantly ionized ions. This approach offers exceptional advantages for performing precision measurements on beams with large energy spreads and allows measurements of atomic parameters of previously inaccessible electronic states. The potential of this experimental method in multidisciplinary research is illustrated by performing, for the first time, hyperfine-structure measurements of selected states in the naturally occurring isotopes of indium, ${}^{113,115}\mathrm{In}$. Ab initio atomic-physics calculations have been performed to highlight the importance of our findings in the development of state-of-the-art atomic many-body methods, nuclear structure, and fundamental-physics studies.

Machine learning with artificial neural networks is revolutionizing science. The most advanced challenges require discovering answers autonomously. In the domain of reinforcement learning, control strategies are improved according to a reward function. The power of neural-network-based reinforcement learning has been highlighted by spectacular recent successes such as playing Go, but its benefits for physics are yet to be demonstrated. Here, we show how a network-based “agent” can discover complete quantum-error-correction strategies, protecting a collection of qubits against noise. These strategies require feedback adapted to measurement outcomes. Finding them from scratch without human guidance and tailored to different hardware resources is a formidable challenge due to the combinatorially large search space. To solve this challenge, we develop two ideas: two-stage learning with teacher and student networks and a reward quantifying the capability to recover the quantum information stored in a multiqubit system. Beyond its immediate impact on quantum computation, our work more generally demonstrates the promise of neural-network-based reinforcement learning in physics.

Spatiotemporal correlations in brain activity are functionally important and have been implicated in perception, learning and plasticity, exploratory behavior, and various aspects of cognition. Neurons in the cerebral cortex are strongly interacting. Their activity is temporally irregular and can exhibit substantial correlations. However, how the collective dynamics of highly recurrent and strongly interacting neurons can evolve into a state in which the activity of individual cells is highly irregular yet macroscopically correlated is an open question. Here, we develop a general theory that relates the strength of pairwise correlations to the anatomical features of networks of strongly coupled neurons. To this end, we investigate networks of binary units. When interactions are strong, the activity is irregular in a large region of parameter space. We find that despite the strong interactions, the correlations are generally very weak. Nevertheless, we identify architectural features, which if present, give rise to strong correlations without destroying the irregularity of the activity. For networks with such features, we determine how correlations scale with the network size and the number of connections. Our work shows the mechanism by which strong correlations can be consistent with highly irregular activity, two hallmarks of neuronal dynamics in the central nervous system.

Multiplex networks have been intensively studied during the last few years as they offer a more realistic representation of many interdependent and multilevel complex networked systems. However, even if most real networks have some degree of directionality, the vast majority of the existent literature deals with multiplex networks where all layers are undirected. Here, we study the dynamics of diffusion processes acting on coupled multilayer networks where at least one layer consists of a directed graph; we call these directed multiplex networks. We reveal a new and unexpected signature of diffusion dynamics on directed multiplex networks, namely, that different from their undirected counterparts, they can exhibit a nonmonotonic rate of convergence to steady state as a function of the degree of coupling, resulting in a faster diffusion at an intermediate degree of coupling than when the two layers are fully coupled. We use synthetic multiplex examples and real-world topologies to illustrate the characteristics of the underlying dynamics that give rise to a regime in which an optimal coupling exists. We further provide analytical and numerical evidence that this new phenomenon is solely a property of directed multiplex, where at least one of the layers exhibits sufficient directionality quantified by a normalized metric of asymmetry in directional path lengths. Given the ubiquity of both directed and multilayer networks in nature, our results have important implications for studying the dynamics of multilevel complex systems.

We propose a theoretical framework for the study of spreading processes in structured metapopulations, with heterogeneous agents, subjected to different recurrent mobility patterns. We propose to represent the heterogeneity in the composition of the metapopulations as layers in a multiplex network, where nodes would correspond to geographical areas and layers account for the mobility patterns of agents of the same class. We analyze classical epidemic models within this framework and obtain an excellent agreement with extensive Monte Carlo simulations. This agreement allows us to derive analytical expressions of the epidemic threshold and to face the challenge of characterizing a real multiplex metapopulation, the city of Medellín in Colombia, where different recurrent mobility patterns are observed depending on the socioeconomic class of the agents. Our framework allows us to unveil the geographical location of those patches that trigger the epidemic state at the critical point. A careful exploration reveals that social mixing between classes and mobility crucially determines these critical patches and, more importantly, it can produce abrupt changes of the critical properties of the epidemic onset.