A crystal structure can switch in roughly ninety picoseconds—and the entire switch-back cycle can finish within five nanoseconds. But the surprising part is that temperature alone cannot explain how fast it happens. Structural transformations in shape memory alloys enable applications including high force, high stroke actuation, sensing, mechanocaloric refrigeration, and waste heat harvesting.

In these alloys, the crystal structure changes between high-symmetric austenite at high temperatures and low-symmetric martensite at low temperatures. The transformations are diffusionless and reversible, so they are a priori fast, but their fundamental speed limits remain unexplored.

A fast cycle matters because application power is determined by the energy per cycle multiplied by the frequency, making the temporal limits worth exploring. The study probes a thin film of the prototype magnetic shape-memory alloy Ni-Mn-Ga and pushes understanding toward the sub-nanosecond time scale.

The film is heated with two-hundred-seventy femtosecond laser pump pulses, while crystal structure is probed with ninety-picosecond synchrotron X-ray pulses. A broad range of laser-to-X-ray delays and laser fluences gives the transformation rate and transformation time of the martensite-to-austenite transformation caused by laser heating.

Measurements and calculations convert each laser fluence into a corresponding sample-temperature evolution, while time-resolved strain measurements test thermal stress as a competing influence. Before studying dynamics, the structural transformation must be characterized quasi-statically as a function of sample temperature.

The experiments use epitaxial Ni-Mn-Ga films with relatively narrow, intense diffraction maxima, and the film is doped with three atomic percent copper to increase its transformation temperature. A fifty-nanometer thickness was chosen to promote homogeneous laser heating while keeping the thickness effect on transformation temperature small.

The film was grown on a silicon substrate with strontium titanate and chromium buffer layers of four and forty nanometers, respectively. This figure shows the film architecture and temperature-dependent structural evolution, with reciprocal space maps covering the martensite and austenite reflections.

Synchrotron-based time-resolved X-ray diffraction probes crystal-structure changes directly and non-invasively. The measurements span laser-to-X-ray delays from minus five hundred picoseconds to five thousand picoseconds to capture the transformation from martensite to austenite and back.

A cryogenic nitrogen gas stream sets the film's base temperature to one hundred fifty kelvin, guaranteeing complete return to martensite below the martensite-finish temperature of three hundred fourteen point one kelvin. The time-resolved reciprocal space maps are cropped into separate austenite and martensite regions, focused on the austenite and martensite peak positions.

Before the laser arrives, at minus sixty picoseconds, the film is martensitic: the martensite peak is visible, while only the chromium reflection appears in the austenite region. At one hundred forty picoseconds after excitation, the austenite intensity rises sharply and the martensite reflection disappears.

The disappearing martensite intensity indicates that a five millijoule-per-square-centimeter laser pulse transforms the complete X-ray-probed volume to austenite. At five nanoseconds, both regions have nearly recovered their initial intensity, showing that the transformation cycle is almost complete.

Across the top row, the central panel develops a broad, high-intensity austenite feature, whereas the corresponding bottom panel becomes nearly featureless. In the final column, both elongated reflections reappear, but with visibly weaker, more diffuse intensity than in the initial martensite map.

Time-resolved reciprocal space maps were recorded at five fluences between five and thirteen millijoules per square centimeter to test how excitation energy affects the dynamics. The martensite-to-austenite transformation is completed in less than two hundred picoseconds in every delay scan.

For the reverse austenite-to-martensite transformation, higher fluences make the transformation take much longer, and the delay range is too short to capture the complete cycle. Higher fluence produces a higher sample temperature, so thermal energy takes longer to dissipate and cool the film below the transformation temperature.

In the top and middle panels, the colored traces separate progressively during the cooling return: the dark-blue five-millijoule trace approaches the opposite phase fraction far earlier, while the red thirteen-millijoule trace remains near its transformed state through the plotted window.

The bottom panel clusters around the one-hundred-percent reference line, with no distinct extra signal between the two phases. Figure four resolves the laser-driven phase transformation on a dense time grid from roughly minus one hundred to three hundred picoseconds. Panel a tracks austenite intensity rising from near zero toward one hundred percent, while panel b shows martensite declining over the same delay window, across fluences from five to thirteen millijoules per square centimeter.

Dashed error-function fits summarize each time-dependent dataset, making the transformation dynamics quantifiable and supporting the paper’s focus on how rapidly the two phases exchange. The transformation-time data from martensite and austenite reflections show the same trend and differ by seventeen picoseconds on average.

Earlier measurements using a seven-nanosecond laser pulse are limited by the laser pulse length. With a two-hundred-seventy-femtosecond laser pulse, the observed transformation time is much shorter, about ninety picoseconds. The experiment is limited by the synchrotron X-ray pulse length, so the transformation may actually proceed even faster.

In the upper portion of the plot, the black seven-nanosecond data cluster near the horizontal seven-nanosecond guide, while the red points remain near the ninety-picosecond X-ray guide in the lower portion. Across the red series, the martensite and austenite markers closely overlap, whereas the black series shifts only modestly across the overheating axis.

The first four data points show negative overheating, meaning the calculated film temperature does not reach the austenite-start temperature required to initiate austenite formation. That result contradicts the observed complete transformation to austenite at the minimum fluence of five millijoules per square centimeter.

The temperature in the electronic, magnetic, and phononic systems therefore needs to be considered separately. The unexpected increase of transformation time with overheating indicates that temperature is not the only relevant factor. Mechanical stress is another decisive factor, originating from the surplus elastic energy required for straining austenite.

In this experiment, stress comes from laser-heated expansion of the thin film against the thick substrate, which remains at base temperature and does not expand. Thermal stress may therefore explain why transformation time increases with increasing fluence.

Figure six compares in-plane and out-of-plane lattice parameters after laser excitation across several fluences. The in-plane parameter stays nearly constant, while the out-of-plane parameter rises abruptly and then relaxes, indicating that heating produces a predominantly out-of-plane response.

Their ratio reveals the film’s compressive strain from epitaxial growth, and panel d shows that the additional compressive strain increases with laser fluence. In a second simulation starting from fourteen-M martensite, the structure shows lower distortion than L ten martensite, with a lattice-vector ratio close to zero point nine.

The fourteen-M structure is often considered nanotwinned N M martensite, so the simulation probes the first level of a hierarchical martensitic microstructure. The fourteen-M to austenite transformation starts around two picoseconds and finishes after four picoseconds, much faster than the N M to austenite transformation.

The simulations indicate that large distortion slows transformation, while a nanotwinned martensitic microstructure can increase transformation speed. Because the simulations contain only a single variant, substantially larger systems are needed to analyze nanotwin ordering and nucleation in real materials.

In the center panel, the directional features associated with fourteen-M gradually merge into the horizontal pattern labeled A, with the change concentrated between the dashed and solid vertical markers. Compared with the left panel, those bands reorganize over a visibly shorter time interval, while the substrate-constrained case in panel c shows the least pronounced directional structure.

The study analyzes five conditions that allow a fast, complete martensite-to-austenite-to-martensite transformation within five nanoseconds. The molecular-dynamics simulations suggest that large lattice distortion requires the collective movement of many atoms, which limits transformation speed.

The observed time remains longer than that of diffusionless transformations with low lattice distortion. Because the experiments are at the limit of the synchrotron probe duration, further experiments with higher temporal resolution, such as free-electron lasers, are needed.

The transformation is extraordinarily fast, yet its speed is constrained by more than heating: thermal stress matters, and large lattice distortion forces many atoms to move collectively. That identifies a structural limit on speed.