We consider percolation on interdependent locally treelike networks, recently introduced by Buldyrev et al., Nature 464, 1025, and demonstrate that the problem can be simplified conceptually by deleting all references to cascades of failures. Such cascades do exist, but their explicit treatment just complicates the theory -which is a straightforward extension of the usual epidemic spreading theory on a single network. Our method has the added benefits that it is directly formulated in terms of an order parameter and its modular structure can be easily extended to other problems, e.g. to any number of interdependent networks, or to networks with dependency links.

We study four Achlioptas type processes with "explosive" percolation transitions. All transitions are clearly continuous, but their finite size scaling functions are not entire holomorphic. The distributions of the order parameter, the relative size smax/N of the largest cluster, are doublehumped. But -in contrast to first order phase transitions -the distance between the two peaks decreases with system size N as N −η with η > 0. We find different positive values of β (defined via smax/N ∼ (p − pc) β for infinite systems) for each model, showing that they are all in different universality classes. In contrast, the exponent Θ (defined such that observables are homogeneous functions of (p − pc)N Θ ) is close to -or even equal to -1/2 for all models. Percolation is a pervasive concept in statistical physics and probability theory and has been studied in extenso in the past. It came thus as a surprise to many, when Achlioptas et al. [1] claimed that a seemingly mild modification of standard percolation models leads to a discontinuous phase transition -named "explosive percolation" (EP) by them -in contrast to the continuous phase transition seen in ordinary percolation. Following [1] there appeared a flood of papers [2-20] studying various aspects and generalizations of EP. In all cases, with one exception [20], the authors agreed that the transition is discontinuous: the "order parameter", defined as the fraction of vertices/sites in the largest cluster, makes a discrete jump at the percolation transition. In the present paper we join the dissenting minority and add further convincing evidence that the EP transition is continuous in all models, but with unusual finite size behavior.From the physical point of view, the model seems somewhat unnatural, since it involves non-local control (there is a 'supervisor' who has to compare distant pairs of nodes to chose the actual bonds to be established [21]). Also, notwithstanding [8], no realistic applications have been proposed. It is well known that the usual concept of universality classes in critical phenomena is invalidated by the presence of long range interactions. Thus it is not surprising that a percolation model with global control can show completely different behavior [22].Usually, e.g. in thermal equilibrium systems, discontinuous phase transitions are identified with "first order" transitions, while continuous transitions are called "second order". This notation is also often applied to percolative transitions. But EP lacks most attributes -except possibly for the discontinuous order parameter jump -considered essential for first order transitions. None of these other attributes (cooperativity, phase coexistence, and nucleation) is observed in Achlioptas type processes, although they are observed in other percolationtype transitions [23]. Thus EP should never have been viewed as a first order transition, and it is gratifying that it is also not discontinuous.Apart from the behavior of the average value m of the order parameter m, phase transitions can also b...

No abstract

Irreversible aggregation is revisited in view of recent work on renormalization of complex networks. Its scaling laws and phase transitions are related to percolation transitions seen in the latter. We illustrate our points by giving the complete solution for the probability to find any given state in an aggregation process (k + 1)X → X, given a fixed number of unit mass particles in the initial state. Exactly the same probability distributions and scaling are found in one dimensional systems (a trivial network) and well-mixed solutions. This reveals that scaling laws found in renormalization of complex networks do not prove that they are self-similar. PACS numbers: 89.75.Hc, 02.10.Ox, 05.70.Ln Droplets beget rain, goblets coagulate to make butter or cream, and dust particles stick together to form aggregates that can eventually coalesce into planets. At the microscopic level, irreversible aggregation of atoms and molecules creates many familiar forms of matter such as aerosols, colloids, gels, suspensions, clusters and solids [1]. Almost a century ago, Smoluchowski proposed a theory based on rate equations to describe processes governed by diffusion, collision and irreversible merging of aggregates [2]. The theory predicts how many small and large clusters exist at any given time and yields a mass distribution that depends on certain details such as the initial conditions, reactions present, relative rates, the presence or absence of spatial structure, etc. A key interest to physicists has been to derive scaling laws that characterize different universality classes [3, and references therein].By contrast, wide interest in complex networks [4-7] has emerged recently. Vast applications to physics, computer science, biology, and sociology [8-10, and references therein] continue to be vigorously investigated. An important question is whether or not complex networks exhibit self-similarity at different length scales and if they can be grouped into universality classes on that basis. Renormalization schemes for networks were proposed [11][12][13][14] to address this question. Scaling of the mass or degree distribution of the renormalized nodes was used to argue that many complex networks are selfsimilar. The semi-sequential renormalization group (RG) flow underlying the box covering of [11][12][13][14] was studied carefully in [15,16], where it was found that scaling laws may be related to an "RG fixed point" which was observed for a wide variety of networks. A convenient, fully sequential scheme called random sequential renormalization (RSR) was introduced [17]. At each RSR step, one node is selected at random, and all nodes within a fixed distance ℓ of it are replaced by a single super-node.We point out a simple mapping between RSR and irreversible aggregation on any graph. Hence any conclusion drawn for one process holds also for the other. Indeed, a local coarse-graining step to produce a new super-node represents one aggregation event, where a 'molecule' aggregates with all its neighbors within distance ℓ to p...

Complex networks are universal, arising in fields as disparate as sociology, physics, and biology. In the past decade, extensive research into the properties and behaviors of complex systems has uncovered surprising commonalities among the topologies of different systems. Attempts to explain these similarities have led to the ongoing development and refinement of network models and graph-theoretical analysis techniques with which to characterize and understand complexity. In this tutorial, we demonstrate through illustrative examples, how network measures and models have contributed to the elucidation of the organization of complex systems.= Ak k P ) ( , which appears as a straight line on a logarithmic plot. The continuously decreasing degree distribution indicates that low-degree nodes have the highest frequencies; however there is a broad degree range with non-zero abundance of very highly connected nodes (hubs) as well. Note that the nodes in a scalefree network do not fall into two separable classes corresponding to low-degree nodes and hubs, but every degree between these two limits appears with a frequency given by P(k). Figure reproduced with permission from Journal of Cell Science (Albert, R., 2005).

-We study a process termed agglomerative percolation (AP) in two dimensions. Instead of adding sites or bonds at random, in AP randomly chosen clusters are linked to all their neighbors. As a result the growth process involves a diverging length scale near a critical point. Picking target clusters with probability proportional to their mass leads to a runaway compact cluster. Choosing all clusters equally leads to a continuous transition in a new universality class for the square lattice, while the transition on the triangular lattice has the same critical exponents as ordinary percolation -violating blatantly the basic notion of universality.Percolation is a pervasive concept in statistical physics and an important branch of mathematics [1]. It typifies the emergence of long range connectivity in many systems such as the flow of liquids through porous media [2], transport in disordered media [3], spread of disease in populations [4], resilience of networks to attack [5], formation of gels [6] and even of social groups [7]. It also underlies a number of other critical phenomena -like the Ising order/disorder transition, which is a percolation transition on the set of spins with given sign [8].The phase transition in ordinary percolation (OP), where bonds or sites are added at random, represents a broad universality class. Recently Achlioptas et al.[9] made a simple modification by, at each step, selecting among two possibilities the link that leads to the slowest growth of large clusters. This global choice introduces a large length scale -the system size L -which can alter universality. Indeed, they concluded that an unusual, discontinuous transition (called "explosive percolation") emerges where a macroscopic cluster appears suddenly while at the same time scaling in other quantities is observed [10][11][12]. Various modifications of the rule have been made [13,14] -all finding evidence of a discontinuous transition. Although the claim for discontinuity in [9] was refuted later in [15], explosive percolation in that case does represent a new universality class.Here we discuss a percolation process that also contains a potentially large length scale in its definition, in this case the correlation length ξ. Our process has direct application to the study of complex networks. Instead of adding bonds randomly, we pick a random cluster and add bonds to its entire surface in order to link it to all adjacent clusters. Starting with the state where all clusters have size one, at each update t → t + 1 the process repeats until the entire lattice (or graph) is reduced to a single cluster. We call this "agglomerative percolation" (AP), in analogy with cluster growth by aggregation [16]. Thus if by chance a cluster of length scale ℓ is picked, links are added simultaneously at distances O(ℓ) apart.AP can be analyzed on any graph. It corresponds to random sequential renormalization [17] of a network, where a single cluster is identified as a 'super'-node that is a local coarse-graining of the graph. In this perspe...

Genes and gene products interact on several levels, forming transcriptional regulatory-, protein interaction-, metabolic- and signal transduction networks. Genetic, biochemical and molecular biology techniques have been used for decades to identify biological interactions; newly developed high-throughput methods now allow for the construction of genome-level interaction maps. In parallel, high-throughput expression data paired with computational algorithms can be used to infer networks of interactions and causal relationships capable of producing the observed experimental data. Graph-theoretical measures and network models are more and more frequently used to discern functional and evolutionary constraints in the organisation of biological networks. Perhaps most importantly, the combination of interaction and expression information allows the formulation of quantitative and predictive dynamic models. Some of the dominant experimental and computational methods used for the reconstruction or inference of cellular networks are reviewed, also the biological insights that have been obtained from graph-theoretical analysis of these networks, and the extension of static networks into various dynamic models capable of providing a new layer of insight into the functioning of cellular systems is discussed.

We develop a framework for simulating a realistic, evolving social network (a city) into which a disease is introduced. We compare our results to prevaccine era measles data for England and Wales, and find that they capture the quantitative and qualitative features of epidemics in populations spanning two orders of magnitude. Our results provide unique insight into how and why the social topology of the contact network influences the propagation of the disease through the population. We argue that network simulation is suitable for concurrently probing contact network dynamics and disease dynamics in ways that prior modeling approaches cannot and it can be extended to the study of less well-documented diseases.