Grab a rubber band, stretch it fast, and hold it against your top lip. It feels warm. Let it snap back and press it up there again, and it feels cool. I ran that four times at my desk before I trusted it.<\/p>\n
Stretching a solid can heat it. Letting it relax can cool it. Engineers call that the elastocaloric effect, and for more than a decade it has been the leading candidate to replace the compressor sitting in the back of your refrigerator.<\/p>\n
There has always been a catch. Something has to do the stretching, and that something has always been an electric motor.<\/p>\n
A team at the Karlsruhe Institute of Technology in Germany and the University of Tsukuba in Japan got rid of the motor. They published the result in Nature Energy<\/a> on August 28, and the thing doing the stretching now is heat.<\/p>\n Two strips of nickel-titanium alloy, mounted end to end on low-friction rails, linked by a small 3D-printed plastic block.<\/p>\n The first strip is 22 micrometers thick, a hair under a mil, which puts it right about where the heavy-duty foil in your kitchen drawer sits at 24. It is a shape memory alloy, the same family of metal that makes eyeglass frames spring back after you sit on them. Heat it above roughly 141 \u00b0F (60.8 \u00b0C) and it remembers the length it was set to and hauls itself shorter, pulling with about 1.2 pounds of force. That is the motor, and its only moving part is its own crystal structure.<\/p>\n The second strip is 26.5 micrometers thick, just over a mil, with a trace of iron alloyed in alongside the nickel and titanium. That one is superelastic, meaning it already sits in the springy phase at room temperature. Pull on it and its crystal structure flips over and dumps heat. Let go and it flips back and absorbs heat. That is the refrigerant, except it is a solid ribbon of metal instead of a gas in a pipe.<\/p>\n The cycle runs in four steps. Heat lands on strip one, which shortens and stretches strip two. Strip two warms up and touches a copper heat sink, which carries that heat away. Strip one then cools off and releases, so strip two snaps back and goes colder than the room around it. While it is cold, it touches a second copper block, which is the thing you actually wanted chilled. The whole loop takes 1.2 seconds and then starts over.<\/p>\n Every elastocaloric cooler built before this one used a motor, a hydraulic ram or some other electromechanical actuator, and all of those are designed to travel a long way. Superelastic alloy wants the opposite deal. It wants an enormous shove over a tiny distance, around 58,000 psi to start the transformation and 68,000 psi to finish it, across a stretch of roughly fifteen thousandths of an inch.<\/p>\n That mismatch shows up in one number. The commercial electromechanical actuator the team benchmarked against delivers 1.1 newtons of force per millimeter of stroke, and their shape memory film delivers 14.5.<\/p>\n Dr. Jingyuan Xu, who runs the ZEco Thermal Lab at Karlsruhe’s Institute of Microstructure Technology, says the group is “establishing a new approach to drive solid-state cooling”<\/a>. KIT is calling it a world first. The paper itself is more careful, and I appreciate that: a heat-driven elastocaloric cycle had been proposed and modeled before, and what is new here is that somebody built one at film scale and measured it.<\/p>\n You should know exactly what these numbers are before you get excited, because some of the coverage has run a little hot.<\/p>\n The 12.9-kelvin figure everyone is quoting, 23.2 \u00b0F, is what an infrared camera saw on the surface of the refrigerant film itself, with no heat exchangers attached. Bolt on the copper sink and the copper source and measure across the finished device, and the span is 4.0 K, or 7.2 \u00b0F. That run also used electricity to warm the actuator, by passing current straight through it, which is a fine way to prove a mechanism and is not what the headline promises.<\/p>\n So the team swapped in a real external heat source, a platinum microheater bonded to a copper plate and held at 266 \u00b0F. Device-level span dropped to 2.2 K. Call it 4 \u00b0F.<\/p>\n Cooling power at zero temperature lift came in at 2.79 milliwatts, and that is not a typo.<\/p>\n Bigger elastocaloric machines exist and they are not shy. Multi-cell architectures have reached 1,284 watts of cooling at zero lift, and macroscale devices have hit temperature spans of 75 K, about 135 \u00b0F. Every one of those runs on a motor. Karlsruhe went small deliberately, because thin films swap heat fast and cycle quickly, and because a walk-in freezer was never the target.<\/p>\n So why would anyone chase four degrees? Because of what sits at the other end of the road.<\/p>\n Navigant Consulting ran a study for the Department of Energy<\/a> that scored 17 alternatives to the vapor compression cycle on energy savings, fit, side benefits and cost. Elastocaloric cooling, which the report called thermoelastic, finished first, with the largest technical energy savings potential in the group at 2.68 quads a year. That works out to about 785 terawatt-hours.<\/p>\n Twelve years on, I can’t find an elastocaloric product an American consumer can go out and buy.<\/p>\n American cooling demand has not been waiting politely. Lawrence Berkeley National Laboratory<\/a> published its 2025 update in June: data centers took about 4.7 percent of US electricity in 2024, and the reference case has them at 649 terawatt-hours by 2030, or 11.8 percent, inside a range of 9.5 to 15.3 percent. The Congressional Research Service puts IT equipment at roughly half or more of a facility’s draw, with much of the remainder going to cooling.<\/p>\n People have gone to strange lengths over that remainder. Microsoft sank 855 servers in a sealed steel tube off the Orkney coast<\/a> in part to let the North Sea handle the heat,, and operators are now buying nuclear reactors by the pod<\/a> to feed the racks and the chillers alike. A cooler that eats the exhaust heat instead of adding to the bill is worth chasing at four degrees.<\/p>\n Sitting in the supplementary material is a layout that should interest anybody reading an automotive site. The team sketches an automotive configuration in which the actuator film is warmed by exhaust or drivetrain heat, and the refrigerant film chills sensitive on-board electronics nearby.<\/p>\n Think about how much heat a car throws away. A gasoline engine sends most of the energy in the tank out through the tailpipe and the radiator, all of it paid for and none of it used. An electric car has less to give away but still runs warm around the inverter and the pack. Every one of those is a heat source sitting inches from a control board that would rather be cooler.<\/p>\n The other layout in the paper is a desktop computer, with the actuator harvesting waste heat off the voltage regulator modules to chill the processor next door.<\/p>\n Both of those are drawings. Neither has been built.<\/p>\n Durability is the other open question, and the authors say so plainly. Their refrigerant film went past 2,000 cycles at 5.2 percent strain with no failure and no measurable fade, which is respectable for a proof of concept and thin for anything you would bolt into a car. TiNi films alloyed with copper and cobalt have run past 10 million cycles in other labs, and the paper points at those as the obvious upgrade for a future build.<\/p>\n Yi-Ting Hsiau built the device and ran the experiments, and Shuichi Miyazaki cold-rolled both films at Tsukuba. Nature Energy received their paper on September 11, 2025, accepted it on July 20, 2026, and published it on August 28, which is nearly a year of peer review for a machine that moves 2.79 milliwatts.<\/p>\n Image credit: Goan Zhou, Zexi Li \/ Nature<\/em><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":" Grab a rubber band, stretch it fast, and hold it against your top lip. It feels warm. Let it snap … <\/p>\nSo what did they actually build?<\/h2>\n
Motors are the wrong shape for this job<\/h2>\n
Four degrees is not a refrigerator<\/h2>\n
The Department of Energy flagged this back in 2014<\/h2>\n
Your car is on the list too<\/h2>\n