🧬 biology

Multicritical Infection Spreading

Through large-scale Monte Carlo simulations in two and three dimensions, this study demonstrates that the multicritical contact process, where both dilution and spatially varying infection rates drive the transition, exhibits universal ultra-slow activated scaling consistent with the strong disorder renormalization group and places it in the same universality class as the multicritical quantum Ising model.

Original authors: Leone V. Luzzatto, Juan Felipe Barrera López, István A. Kovács

Published 2026-09-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Leone V. Luzzatto, Juan Felipe Barrera López, István A. Kovács

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine a world where a simple rule governs how a disease spreads: a sick person can infect their neighbors, but they can also recover and become healthy again. This is the essence of the contact process, a mathematical model used by scientists to understand how infections move through a population. In a perfectly uniform world where everyone is identical, this model predicts a clear tipping point. Below a certain rate of infection, the disease dies out; above it, the infection becomes a permanent, widespread presence. However, the real world is never uniform. People differ in their susceptibility, and the connections between them are not perfect. When scientists introduce this kind of randomness, or disorder, into the model, the rules change. The disease might get stuck in isolated pockets, or it might spread in a way that is incredibly slow and difficult to predict. For decades, researchers have studied two specific ways this randomness can alter the outcome, but they have struggled to understand what happens when both factors collide at a single, critical moment.

A team of physicists at Northwestern University has now mapped this elusive moment with unprecedented precision. They focused on a "multicritical point," a specific combination of infection rates and lattice dilution where the behavior of the system shifts in a fundamental way. To do this, they ran massive computer simulations on grids representing two and three dimensions, tracking millions of independent scenarios of infection spreading. Their goal was to see if the strange, ultra-slow behavior observed in these disordered systems matched the predictions of a powerful theoretical method known as the strong disorder renormalization group. This method, originally developed for quantum physics, suggests that at this critical point, the system behaves in a very specific, universal manner, governed by a set of hidden rules that apply regardless of the specific details of the infection.

The researchers found that the multicritical contact process does indeed follow these hidden rules. By carefully analyzing how the infection cluster grew, shrank, or survived over time, they measured the mathematical exponents that describe this growth. These numbers act like a fingerprint for the type of phase transition occurring. The team discovered that the fingerprints from their simulations matched the theoretical predictions for a disordered quantum system known as the quantum Ising model. This is a significant finding because it suggests that the messy, out-of-equilibrium world of a spreading infection and the abstract, equilibrium world of quantum magnets are governed by the same underlying physics at this critical point. The study confirms that the system belongs to a specific universality class, meaning its behavior is dictated by broad principles rather than the fine details of the model.

One of the most challenging aspects of this work was dealing with the sheer slowness of the process. In these disordered systems, time does not pass in a straight line; instead, the system evolves so slowly that the relevant changes happen on a logarithmic scale. It is as if the clock for the infection is ticking in a way that makes a million years feel like a single second. This made it difficult to distinguish the true, long-term behavior from short-term noise. The researchers overcame this by using a technique that allowed them to look at the relationships between different measurements, effectively canceling out the confusing effects of the initial conditions. They determined the precise infection rates where this critical behavior occurs: approximately 3.5585 for a two-dimensional grid and 5.600 for a three-dimensional one.

Previous attempts to understand this point had produced conflicting results. Some earlier studies suggested the system behaved differently, with numbers that did not fit the established theoretical framework. The new simulations, however, resolved these discrepancies. The team showed that the earlier conflicting results likely stemmed from the way the simulations were started and how the data was interpreted, particularly regarding the microscopic time scales involved. By using a much larger number of simulations and running them for longer periods, the authors were able to see the true asymptotic behavior. Their results align perfectly with the strong disorder renormalization group predictions, providing the first strong confirmation that these theoretical approximations are valid for this type of problem.

The implications of this work extend beyond just infection models. By proving that the disordered contact process and the disordered quantum Ising model share the same critical behavior, the researchers have opened a new door for understanding complex systems. It suggests that the tools used to study quantum magnets can be applied to understand how diseases spread in heterogeneous populations, and vice versa. The study does not offer a new vaccine or a way to stop a specific virus, but it provides a deeper, more accurate map of how order emerges from chaos in complex networks. It confirms that even in a world full of randomness and unpredictability, there are still universal laws that govern the tipping points where systems change from one state to another. The authors' work stands as a rigorous verification of these laws, settling a long-standing question about the nature of multicritical points in disordered systems.

Technical Summary: Multicritical Infection Spreading

Problem Statement
The contact process (CP) is a fundamental stochastic model describing infection spreading, exhibiting an out-of-equilibrium phase transition between an active and an inactive (absorbing) state. In the absence of disorder, this transition belongs to the directed percolation (DP) universality class. However, real-world systems often involve quenched disorder. Previous studies established that disorder can alter the critical behavior, leading to two distinct non-equilibrium phase transitions: a "generic" transition driven by random infection rates and a "percolation" transition driven by lattice dilution. These two transitions meet at a multicritical point (MCP). While the critical behavior at the generic and percolation transitions is understood to be governed by an infinitely disordered fixed point (IDFP) with ultra-slow, activated dynamical scaling, the universality class and critical exponents at the MCP have remained poorly understood. Specifically, existing numerical estimates for the MCP in two dimensions were inconsistent with Strong Disorder Renormalization Group (SDRG) predictions and, in some cases, violated the Harris criterion. Furthermore, no simulations of the MCP had been reported for three dimensions.

Methodology
The authors performed large-scale Monte Carlo (MC) simulations of the disordered contact process on critically diluted hypercubic lattices in two ( $d=2$ ) and three ( $d=3$ ) dimensions.

Model Setup: The system consists of sites that are either active (infected) or inactive (healthy). Infection spreads to neighbors at rate $\lambda$ , and sites heal at rate $\mu$ (set to 1). Disorder is introduced via bond dilution at the percolation threshold ( $p_c$ ).
Initial Conditions: Unlike previous studies that used fully active initial states, this work utilized a "spreading" initial condition starting from a single infected site. This allows for the measurement of the survival probability $P(t)$ , the number of active sites $N(t)$ , and the root-mean-square radius $R(t)$ .
Scaling Analysis: Due to the ultra-slow, activated scaling characteristic of IDFPs (where observables scale as powers of $\ln t$ $ln t$ rather than $t$ $t$ ), standard power-law analysis is insufficient. The authors employed:
1. Hyperscaling Relations: Utilizing the relation $P^2 R^d / N \sim \text{const}$ to locate the critical infection rate $\lambda_c$ without dependence on the unknown microscopic time scale $t_0$ .
2. Pairwise Scaling: Plotting observables against each other (e.g., $N$ vs. $P$ ) to extract exponent combinations (e.g., $(\psi\bar{\delta})^{-1}$ ) independent of $t_0$ .
3. Off-Critical Scaling: Analyzing the derivative of the logarithm of observables with respect to $\lambda$ ( $\partial_\lambda \ln N$ ) to estimate the order parameter exponent $\beta$ .
4. Extrapolation: Performing infinite-time extrapolations of effective exponents to the limit $[\ln(t/t_0)]^{-1} \to 0$ .

Key Contributions

First 3D Characterization: The paper provides the first large-scale Monte Carlo characterization of the multicritical contact process in three dimensions.
Resolution of 2D Discrepancies: The study re-examines the $d=2$ MCP, addressing inconsistencies in previous literature. The authors attribute previous discrepancies to the use of fully active initial conditions, which are highly sensitive to the microscopic time scale $t_0$ , and the use of only the largest connected component in prior simulations.
Methodological Rigor: The work demonstrates a robust methodology for extracting critical exponents in systems with activated scaling, leveraging hyperscaling and pairwise scaling to bypass the uncertainty of $t_0$ before determining individual exponents.

Results

Critical Rates: The authors determined the multicritical infection rates to be $\lambda_c^{(2)} = 3.5585(20)$ for $d=2$ and $\lambda_c^{(3)} = 5.600(5)$ for $d=3$ .
Critical Exponents: The extracted critical exponents (including $\beta$ $β$ , $\bar{\Theta}$ $\overset{ˉ}{Θ}$ , $\psi$ $ψ$ , and $\bar{\delta}$ $\overset{ˉ}{δ}$ ) for both dimensions are consistent with the predictions of the SDRG method applied to the multicritical quantum Ising model.
- In $d=2$ , the results satisfy the Harris criterion, unlike the earlier Monte Carlo results by Dahmen et al. [32].
- In $d=3$ , the results align with the only available theoretical estimates (SDRG).
Universality: The numerical estimates for the exponents in both dimensions match the SDRG predictions within statistical uncertainty, confirming that the multicritical contact process and the multicritical disordered quantum Ising model belong to the same universality class.
Dynamical Scaling: The simulations confirm the ultra-slow, activated dynamical scaling ( $\rho(t) \sim [\ln(t/t_0)]^{-\bar{\delta}}$ ) at the MCP, with exponents consistent with an IDFP.

Significance
The paper claims that its primary significance lies in confirming the universality class of the multicritical contact process. By demonstrating that the critical exponents obtained via Monte Carlo simulations align with SDRG predictions for the quantum Ising model, the work provides the first numerical confirmation of the multicritical SDRG results. This validates the approximations used in the numerical SDRG scheme. Furthermore, the observation that the disordered CP and the quantum Ising model likely share the same universality class in both $d=2$ and $d=3$ suggests a deep connection between non-equilibrium classical spreading processes and equilibrium quantum critical phenomena, implying that the dynamics of the former are governed by the same underlying activity clusters as the latter. The authors remain modest, noting that while their results are consistent with SDRG, systematic uncertainties related to the microscopic time scale $t_0$ and finite-time effects remain a challenge in such simulations.

More like this