When Zhiyue Lu of the University of North Carolina read about the Mpemba effect in middle school, he snuck into an oil refinery in the Shandong province of China where his mother worked and used precision lab equipment to measure temperature as a function of time in a sample of water (he ended up supercooling the water without it freezing). Later, while studying nonequilibrium thermodynamics as a graduate student, he tried to reframe his approach to the Mpemba effect. “Is there any thermodynamic rule that will forbid the following: Something starting further away from the final equilibrium that would approach equilibrium faster than something starting from close?” he asked.

Lu met Oren Raz, who now studies nonequilibrium statistical mechanics at the Weizmann Institute of Science in Israel, and they began developing a framework to investigate the Mpemba effect generally, not just in water. Their 2017 paper in the Proceedings of the National Academy of Sciences modeled the random dynamics of particles, showing that, in principle, there are nonequilibrium conditions under which the Mpemba effect and its inverse could occur. The abstract findings suggested that the components of a hotter system, by virtue of having more energy, are able to explore more possible configurations and therefore discover states that act as a sort of bypass, allowing the hot system to overtake a cool one as both drop toward a colder final state.

“We all have this naive picture that says temperature should change monotonically,” Raz says. “You start at a high temperature, then a medium temperature, and go to a low temperature.” But for something driven out of equilibrium, “it’s not really true to say that the system has a temperature,” and “since that’s the case you can have strange shortcuts.”