Disorder can improve network stability under specific conditions, study finds

Northwestern researchers say disorder network stability can be improved when differences among components or their connections are introduced in the right places and amounts. Their study, published in Science, offers a framework for testing when variation helps a modeled system recover from a small disruption rather than becoming unstable.

The finding does not mean that randomness makes networks safer. The reported benefit depends on the system's dynamics, the location of the differences and their scale; moderate variation can help, while too much can undermine stability, according to Northwestern's account of the research.

How can disorder make a network more stable?

In the study, stability means a system returns toward a stable state after a small disturbance instead of allowing that disturbance to grow. The researchers examined systems near a stable state, then calculated whether minor disruptions would fade or intensify.

They compared networks whose components and connections were identical with networks that varied. The team found that heterogeneity—differences either among nodes or among the links between them—can improve stability under certain conditions.

A network's nodes are its individual components, while links are their connections. In a power-grid model, generators are nodes and transmission lines are links. In an ecological-network model, species are nodes, connected through relationships such as competition, cooperation and predation, Northwestern said.

What limits did the researchers find?

The effect is conditional. Northwestern said the node dynamics must be sufficiently rich for disorder to have a stabilizing role, and the outcome changes with the placement and amount of variation. Simplified models that represent each node with only one variable can miss the effect, the researchers said.

Northwestern's release says earlier network research often emphasized how components were connected and treated differences among components as imperfections. The new framework is intended to identify when those differences can be useful rather than harmful.

Which systems did the study examine?

The team tested the framework on models of power grids, neurons, flocks, architected materials and ecological networks. These are modeling results, not evidence that operating power grids, ecosystems or other deployed systems have already been made more stable with the method.

The researchers said the work could eventually inform efforts to design more robust power grids, architected materials and other interconnected systems. It may also help explain why varied components and uneven interactions are common in biological and ecological networks.

Northwestern also pointed to earlier work as context. A 2020 study by Motter's team reported that slight differences among generators could improve synchronization, while a 2025 study led by Arthur Montanari reported similar effects in models of flocking and drone swarms. The new study seeks to set out broader conditions for when such results arise.

Adilson Motter led the research, according to Northwestern. Arthur Montanari and Pietro Zanin were the co-first authors.

This story draws on original reporting from Phys.org.