Subsystem Resetting: TIFR researchers discover a new route to control phase transitions in complex systems

Researchers in the Department of Theoretical Physics at Tata Institute of Fundamental Research (TIFR), Mumbai, have discovered that instead of manipulating every component or modifying interactions in a many-body system, occasionally resetting just a small fraction can reshape how the entire system behaves macroscopically, including how it transitions from one phase to another. This counterintuitive approach, called subsystem resetting, offers a powerful, universal control strategy to tune collective behavior in complex systems ranging from magnets to neural networks.

This work by Anish Acharya, Rupak Majumder, and Prof. Shamik Gupta has been published in Physical Review Letters.

Resetting the few to control the many

In everyday life, a gentle nudge at the right moment can prevent a system from spiraling—resetting a frozen computer, rebooting a router, or tapping a malfunctioning machine. The researchers show that a similar idea works in physics, but with a deeper consequence. “The most surprising insight is that you can reshape the entire phase behavior of a many-body system without ever touching the interactions or the majority of its degrees of freedom,” says Prof. Shamik Gupta. “In statistical physics, the orthodoxy is that to change a phase transition, you must change couplings, external fields, geometry, or temperature. Yet this study shows that you can move, split, eliminate, or recreate phase transitions exactly as you like, simply by occasionally resetting a small fraction of the system to a chosen state.”

Crucially, this is not achieved by tuning the Hamiltonian or modifying the energy landscape. “This is not tuning the Hamiltonian. This is not changing the energy landscape by tuning the couplings. This is not enforcing global control,” Gupta explains. “It is something entirely different: a procedural intervention that replaces structural modification.”

At the heart of the mechanism is non-equilibrium dynamics. “Resetting breaks detailed balance, thereby inducing a nonequilibrium structure. It biases macroscopic order indirectly through the reset subsystem, exploits long-range interactions to propagate that bias, and turns memory effects into a stabilizing force,” Gupta adds. “Control emerges from an intricate interplay of noise, memory, and incomplete intervention.”

Why this matters

Many-body systems—from spins in a magnet to firing neurons in the brain—often exhibit sudden transitions, like water freezing into ice. Controlling such transitions usually demands fine-tuning across the entire system, which is a daunting task.

Subsystem resetting offers a surprisingly lighter alternative. The study shows that by choosing (1) what portion of the system to reset, (2) how often to reset, and (3) what state to reset to, researchers can shift critical points, alter the nature of phase transitions, or even reproduce the entire phase diagram of the bare dynamics without changing fundamental interactions. “One can dramatically reshape global behavior by intervening only in a small part of the system,” Acharya notes. “Moreover, interventions only occasionally in time are sufficient, reducing the energy budget of the protocol.”

Even though the intervention is local, its effects are global. “Long-range interactions transmit the influence of the reset subsystem to the non-reset part, effectively biasing the behavior of the entire system,” Acharya explains. “This reveals a general mechanism: in long-range systems or highly connected networks, control of influential subsystems cascades through the system.”

A universal and remarkably efficient strategy

Instead of tuning temperature, couplings, or noise, control is achieved through just three knobs: the reset frequency, the size of the reset subsystem, and the reset state itself. The team tested this idea on a variety of models and found the effect to be strikingly universal. “Our results hold across equilibrium spin models, nonequilibrium nonlinear oscillator systems, in mean-field and non-mean-field settings, and in systems with and without quenched disorder,” Majumder emphasizes. “This ubiquity implies that minimal-resetting control is broadly applicable, not restricted to idealized models.”

Even more remarkably, the results remain analytically tractable despite the presence of memory effects, and simulations match the theoretical predictions remarkably well. This approach offers a platform-agnostic route to control, without engineering couplings or altering global dynamics. “One may start seeing control in time and control of a subset as equally fundamental as tuning couplings,” Majumder says. “Stability can arise through occasional, targeted dynamical interventions, rather than through modifying the energy landscape.” This is particularly relevant for systems where interactions are fixed or difficult to manipulate, such as biological networks or trapped-ion hardware.

Broader impact

The work opens pathways to light-touch control in diverse real systems, including:

• Neural networks, where timed resetting could suppress pathological neural synchrony (e.g., Parkinson’s disease),
• Magnetic and quantum materials, potentially stabilizing phases over wider temperature ranges,
• Cold atom and ion-trap platforms, where resetting could be experimentally implemented,
• Complex interaction networks, where resetting only influential nodes may guide global behavior.

Looking ahead, the researchers are keen to see the idea tested in systems where failures are rare but catastrophic. “We would love to see subsystem resetting used to control cascades and instabilities in real-world networks—power grids, financial markets, and communication systems—using minimal targeted resets,” Gupta says. “One can not reset an entire power grid or an entire financial market, but one can certainly reset a handful of generators to force phase synchronization, reset the liquidity or leverage of a few institutions, or reset routing tables on key routers. Our work shows that controlling a small, influential fraction can reshape global behavior, and this is precisely what these real systems need.”