For Decades Network Science Assumed Identical Parts Make a Stable System. Northwestern Physicists Proved the Opposite in Power Grids, Neurons, Flocks and Food Webs: The Right Amount of Mismatch Makes Them Harder to Knock Over, and Too Little Is as Dangerous as Too Much.

Engineers who build power grids, ecologists who model food webs and physicists who design new materials have long shared an assumption: a network runs most reliably when its parts are as alike as possible. Variation among generators, species or building blocks was treated as an imperfection to be engineered away. A study published in the journal Science by Northwestern University physicists says that assumption is wrong for a large class of real systems, and gives a mathematical framework for finding exactly how much mismatch makes a network most stable.

"Previous studies found a growing number of cases in which disorder, also called heterogeneity, irregularity or asymmetry, across a network's nodes can actually improve stability and desirable behavior," said Adilson Motter, the Charles E. and Emma H. Morrison Professor of Physics and Astronomy at Northwestern and director of its Center for Network Dynamics, who led the work. "We have seen this in important real-world systems, including power grids, metamaterials and brain computation. But we didn't know how widespread this effect was or which kinds of systems could benefit from it. Our new study answers those questions, explains why these differences can improve stability and even reveals why scientists overlooked this effect for so long."

The reason it was overlooked, the team argues, is the models. Network scientists have leaned on simplified descriptions such as the widely used Kuramoto model, which reduces each node to a single variable. "Disorder can stabilize networks, but only when the node dynamics are rich enough," Motter said. "Simplified models can inadvertently strip away the very stabilizing effect we want to capture." Postdoctoral researcher Arthur Montanari and graduate student Pietro Zanin, the study's co-first authors, built a general framework instead: analyze a system near a stable state, calculate whether small disturbances die out or grow, then compare networks of identical components against networks with varied components and connections, and identify the conditions under which heterogeneity wins.

They ran that framework on models of power grids, neurons, bird flocks, architected materials and ecological networks, and found two distinct routes by which disorder helps: differences among the nodes themselves, or differences among the links between them. Where the variation sits and how much of it there is both matter. "If you make the system more homogeneous, you lose stability," Montanari said. "But if you increase disorder too much, you also lose stability. Our framework can help pinpoint the level of disorder that helps the system achieve optimal stability." Strikingly, the differences often did not need to be designed at all. In many of the models, even randomly introduced variation beat the best perfectly uniform configuration. And when the disorder lived in the links rather than the nodes, even networks with simple node dynamics benefited.

The result builds on earlier hints from the same group. A 2020 Nature Physics paper found power generators synchronize more effectively when they run slightly differently from one another, and a 2025 Nature Communications study by Montanari found the same in models of flocking and drone swarms. The new work says those were not isolated cases but instances of a broad principle. It may also resolve an old paradox. "Since the 1970s, mathematical models have predicted that large, complex ecosystems should destabilize and collapse," Montanari said. "Yet very large and highly diverse ecosystems persist in nature. Our findings suggest that variation among mutually beneficial interactions, such as those between pollinators and flowers, could help explain this paradox." For engineers, the immediate application is architected materials, which are typically built from repeated, identical cells; deliberately varying the cells' shapes, sizes and orientations could unlock new behaviors. "When disorder enhances stability," Motter said, "the next challenge is figuring out how best to design it." The team has also released a website that lets users tweak the parameters and watch networks synchronize and organize.