From palm-sized electronic circuits to massive data center racks, heat management continues to define the performance limits of virtually every energy conversion system. Heating and cooling now account for nearly half of global energy demand, and most of this relies on vapor-compression cycles that mechanically compress and expand gaseous refrigerants such as fluorocarbons. However, greenhouse gas emissions from refrigerant leakage and the bulk and noise of compressors are fundamentally incompatible with the increasing integration of microelectronic devices.
Thermoelectric conversion (Peltier devices), which has become a popular alternative for small-scale equipment, is likewise constrained by a thermodynamic ceiling of only 10–15% of the reverse Carnot efficiency, meaning most of the input electrical power is simply dumped back into the environment as additional waste heat.
In response to these cooling limitations, elastocaloric cooling—which exploits solid-state phase transformations—has drawn growing attention. When a tensile load is applied to a shape-memory alloy, its crystal structure transforms from the austenite phase to the martensite phase, releasing latent heat; when the load is removed, the reverse transformation occurs and heat is absorbed. This principle requires no refrigerant gas whatsoever and theoretically offers efficiency approaching 84% of the Carnot limit—an extraordinarily high potential.
However, existing elastocaloric cooling systems have required high-power electric motors or hydraulic pumps to deform the refrigerant alloy. The need to incorporate large mechanical drive systems just to achieve cooling—consuming substantial electrical power in the process—has been a major barrier to practical deployment of this technology.
On August 24, 2026, a joint research team—including Yi-Ting Hsiau and Jingyuan Xu of the Institute of Microstructure Technology at the Karlsruhe Institute of Technology (KIT), together with Shuichi Miyazaki of the University of Tsukuba's Division of Materials Science—published a new proof-of-concept that breaks through this drive-system barrier, in the British journal Nature Energy (DOI: 10.1038/s41560-026-02122-6). Instead of an electric motor, the team incorporated a "shape-memory alloy thin film that contracts when heated" as the actuator, converting thermal energy itself directly into the driving force needed to stretch the refrigerant film and complete a cooling cycle.
First author Yi-Ting Hsiau said, "The decisive moment for us was when we could actually measure the cold produced by a system driven purely by heat." This marks the establishment of a foundational experiment for a new form of solid-state thermal management that converts waste heat itself into the driving power for cooling.
A Coordinated Cycle of Heat and Stress Woven by Two Shape-Memory Alloys
The system built by the research team consists of two cold-rolled shape-memory alloy (SMA) thin films with different properties, connected in a planar arrangement. One serves as the drive unit, generating mechanical force from heat; the other functions as the cooling unit, generating a temperature change from applied stress.
At the core of the drive unit is a 22 μm-thick Ti₅₀.₁₈Ni₄₉.₈₂ alloy film, which exhibits a one-way shape-memory effect. This film is pre-stretched in its low-temperature martensite phase. When heated above the reverse-transformation finish temperature (A_f = 60.8 °C), beyond the start temperature (A_s = 47.1 °C), it transforms into the austenite phase while contracting back to its original length, generating a powerful tensile force. The deformation stress in the martensite state at room temperature (23 °C) is approximately 150 MPa, but this jumps sharply to about 440 MPa in the austenite state at 83 °C. This temperature-induced stress difference is what drives the actuator.
The paired cooling unit uses a 26.5 μm-thick Ti₄₉.₁Ni₅₀.₅Fe₀.₄ alloy film. This composition, with a small addition of iron (Fe), has its transformation finish temperature (A_f) tuned to 18.5 °C, meaning it consistently exhibits superelasticity (rubber-like reversible elastic deformation behavior) at room temperature.
When this room-temperature superelastic film is subjected to a tensile test, stress-induced martensitic transformation begins at approximately 400 MPa, and the crystal structure transition is completed within a transformation plateau region at around 470 MPa and 5.2% strain. The adiabatic temperature change (ΔT_ad) accompanying the phase transformation reaches 19.4 K in high-speed tensile tests of the material alone.
The two films are connected in series via 3D-printed polymer connectors that block heat conduction between them while accurately transmitting only longitudinal tensile force and displacement. The cooling cycle proceeds continuously through the following four stages:
- Heating and loading stage: When heat input raises the drive film's temperature above A_f, the drive film contracts, pulling the refrigerant film longitudinally. This induces stress-induced martensitic transformation in the refrigerant film, releasing latent heat and raising its temperature.
- Heat rejection stage: A heat sink in the cooling unit contacts the heated refrigerant film, dumping the released latent heat to the exterior.
- Cooling and unloading stage: When a heat sink contacts the drive film and cools it below A_s, the film's tension drops. The stretched refrigerant film then returns to its original length under its own superelastic restoring force (reverse transformation), absorbing heat from its surroundings and rapidly cooling below room temperature.
- Heat absorption stage: A heat source in the cooling unit contacts the now-cooled refrigerant film, absorbing heat from the target to be cooled.
The engineering advantage of this mechanism lies in how well the mechanical characteristics of the actuator and refrigerant are matched. Conventional electromagnetic motors and hydraulic cylinders are typically designed to maximize stroke length, making them poorly suited to the demands of thin-film refrigerants, which require "several newtons of force over an extremely short stroke of only a few hundred micrometers."
While commercially available compact electromagnetic actuators achieve a force-to-displacement ratio of only 1.1 N/mm, the team's shape-memory alloy thermal actuator achieved a far higher ratio of 14.5 N/mm. By directly connecting the thin films to one another without any additional reduction gears or linkage mechanisms, the team achieved a lighter, more compact drive system.
Temperature Span and Cooling Output Verified Under Joule Heating and External Heat Sources
To quantify the cooling performance of the coupled system, the research team conducted evaluations using two experimental setups: direct Joule heating via electrical current, and thermal conduction driving using an external heater. All measurement data shown represent averages and standard deviations from n=3 independent experiments.
First, under Joule heating drive—in which a 1.0 A current pulse was applied to the drive film to rapidly heat it from within (reaching a peak temperature of 86 °C)—the drive film reliably generated a stroke exceeding 300 μm, driving the refrigerant film to a strain of 4.5%.
In infrared thermography measurements of the refrigerant film alone (with the cooling unit's heat exchangers removed), the film exhibited a maximum temperature span of 12.9 K under loading and unloading rates of 0.09 s⁻¹ and 0.03 s⁻¹, respectively.
In measurements of the full device incorporating copper heat sinks and heat sources, the system maintained a steady-state device-level temperature span of 4.0 K at an operating frequency of 0.83 Hz (1.2 seconds per cycle)—a rise of +2.2 K on the hot side and a drop of -1.8 K on the cold side. The zero-temperature-lift cooling capacity, evaluated using the counter-heating method (in which a fine resistance wire wound around the cooling section measures the power needed to offset the temperature drop), was 2.79 mW, yielding a specific cooling power (SCP) of 4.43 W/g relative to the effective refrigerant mass (0.63 mg).
Next, the team conducted a verification test simulating actual waste-heat utilization by driving the system with an external heat source. Removing the Joule heating wiring, they periodically brought a copper heat-source block—embedded with a platinum microheater and held at 130 °C—into contact with the drive film. Driving the drive film using only thermal conduction from this external source, at loading and unloading rates of approximately 0.07 s⁻¹ and an operating frequency of 0.83 Hz, the steady-state device-level temperature span was 2.2 K (+1.2 K on the hot side, -1.0 K on the cold side).
Under these conditions, the cooling capacity was 2.09 mW, with a specific cooling power (SCP) of 3.32 W/g. The reasons the required temperature rose to 130 °C and the temperature span dropped by 1.8 K under external heat-source drive—compared to Joule heating, which generates heat uniformly throughout the film—are that thermal conduction across a solid interface is limited by contact thermal resistance, which restricts the rate of heat transfer into the film and reduces the achieved stroke to approximately 300 μm (about 4.1% strain).
Regarding thermodynamic efficiency, the coefficient of performance (COP) for the refrigerant unit alone—without work recovery—reached 2.61 (counter-heating method) and 2.88 (temperature gradient method). The overall system COP, based on total heat input including the drive unit, was 8.53 × 10⁻².
| Configuration | Drive method | Heating temperature (°C) | Operating frequency (Hz) | Device temperature span (K) | Specific cooling power SCP (W/g) | Cooling capacity (mW) | System COP |
|---|---|---|---|---|---|---|---|
| Cooling unit alone | Commercial electric actuator | Room temperature operation | 2.00 | 8.2 | Not measured | Not measured | 2.61 (refrigerant alone) |
| Coupled thermal drive system | Joule heating (internal generation) | 86 | 0.83 | 4.0 | 4.43 | 2.79 | 8.53 × 10⁻² |
| Coupled thermal drive system | External heat source (solid-state conduction) | 130 | 0.83 | 2.2 | 3.32 | 2.09 | Not measured |
Regarding durability, the team conducted a continuous 80-cycle (96-second) operational test under Joule heating drive and confirmed no degradation in either temperature span or maximum generated load. In material-level fatigue testing, the TiNiFe refrigerant film withstood over 2,000 cycles of continuous deformation under a tensile load of 5.2% strain without fracturing.
Comparison with Macroscopic Systems and the Current State of Elastocaloric Technology
Within the lineage of solid-state cooling technologies, elastocaloric cooling has been pioneered as a field with an exceptionally high theoretical thermal efficiency ceiling. Because it involves no compression of gaseous refrigerants, it carries no associated environmental burden, and its theoretical thermodynamic limit is calculated to reach up to 84% of the Carnot efficiency.
Large-scale elastocaloric systems using bulk materials (macroscopic solid alloy blocks or alloy tubes), which combine fluid heat-transfer media with active thermal storage mechanisms (regenerators), have reportedly achieved maximum temperature spans of up to 75 K and a substantial cooling capacity of 1,284 W under zero-temperature-lift conditions.
By contrast, the thin-film-based microscale cooling pursued by the research team represents a fundamentally different design philosophy.
Thin-film geometries have an overwhelmingly large surface-area-to-volume ratio (specific surface area), enabling rapid heat exchange through direct solid-solid contact alone, without any fluid heat-transfer medium. Prior research has reported high-speed cycling at up to 4 Hz and specific cooling power (SCP) reaching as high as 19 W/g.
However, previous thin-film systems have all required electromagnetic motors to drive deformation. While the concept of running macroscopic regenerative elastocaloric cooling using heat alone had been proposed theoretically through numerical simulation, this research team is the first to actually assemble and demonstrate the operation of such a system in a microscale thin-film structure.
In the demonstration, the standalone cooling unit—driven directly by an electric actuator—achieved a temperature span of 8.2 K at a frequency of 2 Hz and 4.5% strain. By contrast, the system with the coupled thermal-drive actuator saw this drop to 0.83 Hz and temperature spans of 4.0 K (Joule heating) and 2.2 K (external heat source).
This performance gap stems from the thermal time constant required to heat and cool the thermal actuator, which limits the rate of deformation and prevents the system from meeting adiabatic transformation conditions (maintaining fully adiabatic conditions generally requires a strain rate of roughly 0.1 s⁻¹ or higher). Even so, the significance of having opened up a physical pathway that completely decouples the system from large external mechanical drive systems—using heat alone to stretch and contract a solid and generate cooling near the freezing point—should not be understated.
Challenges Toward Practical Application and the Scale Gap in Cooling Capacity
The research team's achievement is a feasibility study that demonstrates the viability of a mechanism for generating cold from waste heat—it is not yet a completed device capable of immediately replacing existing cooling equipment. Several constraints and challenges must be squarely acknowledged from an engineering standpoint.
First is the low absolute cooling capacity. The cooling capacity obtained under external heat-source drive was only 2.09 mW (milliwatts)—several orders of magnitude below the tens to hundreds of watts of heat load continuously dissipated by personal computer CPUs or data center server racks.
While some overseas tech media outlets have presented this in exaggerated contexts, such as "a new technology that cools servers using data center waste heat," the conceptual diagrams of CPUs and automotive electronics included in the paper's supplementary materials are merely schematic illustrations of future possibilities—what was actually demonstrated in the laboratory was milliwatt-scale behavior from a single thin film.
Second, a fully power-free system has not yet been achieved. Although the main actuator that stretches the film has been converted to thermal drive, the reciprocating motion needed to bring the copper heat exchangers (heat sink and heat source) into and out of contact with the film—for dissipating and absorbing heat—still relies on an external servo linear actuator (an electric motor). Achieving true self-sustained operation will require developing a fully passive thermal switching mechanism, using approaches such as thermal expansion, bimetallic structures, or bistable mechanisms.
Third is the high temperature of the heat source required for operation. In the external heat-source drive experiment, the contact heating block had to be maintained at 130 °C. However, much of the low-grade waste heat emitted by integrated circuits and typical electronic equipment falls in the 60–80 °C range. Bridging this temperature gap will require precisely optimizing the actuator alloy's composition to shift the transformation temperature (A_f) further downward, along with fundamentally reducing the thermal resistance at the solid contact interfaces.
Furthermore, durability testing has so far only reached around 2,000 cycles, and the mechanical and functional fatigue resistance required for the millions to tens of millions of continuous cycles demanded by industrial applications has not yet been verified.
A Path Forward Through Multilayering and Materials Engineering
To elevate this milliwatt-scale proof of concept into a practical thermal management device, the research team has laid out a clear roadmap for improvement.
To boost cooling capacity, expanding the film area and adopting parallel stacking (parallelization) represents the most realistic approach. Since elastocaloric cooling capacity is proportional to the volume of alloy cycled, adopting a modular structure that layers dozens to hundreds of thin films together, exchanging heat through a shared heat exchanger, could raise cooling output from the milliwatt scale to the watt scale while keeping the footprint compact.
To expand the temperature span, introducing a multi-stage cascade (thermal storage) structure that connects multiple cooling elements in series is considered effective.
Regarding fatigue in the thin films themselves, excellent precedents already exist in the shape-memory alloy field. Thin-film materials made of titanium-nickel-copper (TiNiCu) alloys, or quaternary alloys with added cobalt (TiNiCuCo), have been reported to withstand over 10⁷ (ten million) repeated cycles under tensile deformation without functional degradation.
Because these fatigue-resistant compositions are compatible with the thin-film fabrication process used in this study, there remains ample room to incorporate them into next-generation prototypes to resolve the lifespan challenge.
Improving deformation speed to achieve a truly adiabatic response will also require structural mechanical innovations. Introducing bistable latch mechanisms or catch-and-release mechanisms—which release strain energy slowly accumulated through heat all at once—would allow the system to achieve instantaneous high-speed strain rates (above 0.1 s⁻¹) without being constrained by the rate of heat input, thereby maximizing the adiabatic temperature change of the refrigerant.
Jingyuan Xu, corresponding author at the Karlsruhe Institute of Technology, offered this outlook: "By scaling up this technology, we aim to build sustainable cooling systems that make use of the abundant untapped heat sources around us."
At present, no independent third-party replication or commercialization roadmap has been presented. Nevertheless, by driving solid-state phase transformation directly with thermal energy—without ever converting it into electrical power—and extracting a tangible temperature drop from that process, this research presents a genuinely new physical option in the field of thermal management.