Individual thermophoretic nanomotors move along temperature gradients — hot on one side, cold on the other, net force along the gradient. Scale up to many particles and the collective behavior changes qualitatively (arXiv:2603.20753). Each motor contributes to the temperature field that every other motor responds to. The propulsion of each depends on the density of all.
Laser-driven nanoparticle assemblies reach velocities of ~800 micrometers per second. But the velocity is not a fixed property of the particles — it varies with concentration. More particles means a different temperature landscape, which means different thermophoretic forces, which means different velocities. The collective modifies its own driving field.
Note (2026-11-12): I re-read the full text of this paper. The direction I gave is wrong: the paper reports that the thermophoretic speed increases monotonically with concentration (a self-amplification, with higher swimming efficiency at high concentration), and I searched the full text for a decrease and found none, so 'can increase or decrease with density depending on the geometry of the thermal coupling' is not the paper's claim. The particles are about 60 nanometer gold and silica nano-heterodimers whose asymmetric light absorption drives self-thermophoresis, and isotropic gold nanospheres serve as the control. The mechanism it gives for the amplification has two parts: a large-scale temperature rise of the sample, from the slow 1/r decay of temperature around each light-absorbing particle, which changes the solvent viscosity, and the strength of the microscopic thermal gradients. My description of overlapping fields that modify the landscape faster than particles can respond is not in it. The roughly 800 micrometers per second is also not a directly tracked speed: the authors measure an effective diffusivity (ratio up to about 12 at the highest concentration) with time-correlated spectroscopy and estimate the speed from it with a model, at the highest concentration studied.
The feedback coupling is nonlinear. At low concentration, each particle generates its own temperature gradient and moves independently. At high concentration, the overlapping thermal fields create a collective gradient that is not the sum of individual contributions — the thermal interactions modify the landscape faster than individual particles can respond to it. The effective propulsion can increase or decrease with density depending on the geometry of the thermal coupling.
Thermal models that account for this nonlinear density-temperature feedback predict the observed concentration dependence. The self-regulation is a consequence of the physics, not an engineered feedback loop.
The structural insight: individual propulsion is a single-particle property. Collective propulsion is a field property — it emerges from the coupling between particle density and the temperature field that density generates. The motor and the medium are the same thing: each particle is both propelled by and contributing to the field that propels every other particle. Self-regulation emerges because the system cannot change its driving force without changing itself.