HomeArticlesMaterials Science

Shape Memory Alloys: Martensite, Austenite and Why Nitinol Remembers

Twinned martensite, detwinning, the one-way memory effect and superelasticity — how a diffusionless phase transition gives Nitinol its two party tricks.

mysimulator teamUpdated June 2026≈ 8 min read▶ Open the simulation

Two crystal structures, one composition

Nitinol — nickel-titanium, roughly 50/50 atomic percent — is the workhorse shape-memory alloy, and its behaviour comes entirely from a solid-state phase transformation between two crystal structures of the identical chemical composition. Above a characteristic temperature the stable structure is austenite, a cubic B2 (CsCl-type) lattice, highly symmetric and mechanically stiff. Below that temperature the stable structure is martensite, a lower-symmetry monoclinic B19′ lattice. Unlike most phase transitions, this one is diffusionless: no atom needs to migrate any significant distance, and the transformation completes in microseconds because it is really just a coordinated, cooperative shearing and shuffling of the existing lattice — a martensitic transformation, the same family of transformation (though a very different alloy) that hardens quenched steel.

live demo · the stress-strain hysteresis loop of a superelastic wire● LIVE

Four temperatures, one hysteresis loop

Because nucleating a new crystal structure needs to overcome an energy barrier, the transformation does not happen at one sharp temperature on cooling and warming alike — it lags, producing thermal hysteresis. Four transformation temperatures define the loop, always in this order on the temperature axis:

Ms   martensite start   — cooling austenite, martensite begins forming
Mf   martensite finish  — cooling continues, fully martensite
As   austenite start    — heating martensite, austenite begins forming
Af   austenite finish   — heating continues, fully austenite

typical ordering:           Mf  <  Ms  <  As  <  Af
typical hysteresis width Af − Mf : roughly 20-40 K for Nitinol

Below Mf the material is fully martensitic but, critically, it forms as twinned martensite: many small martensite variants (differently oriented plates of the same low-symmetry structure) self-accommodating so the macroscopic shape barely changes even though every atom has moved. Twinned martensite is soft and easily deformed — not because the material got weaker, but because an applied stress can simply grow the favourably oriented variants at the expense of others (a process called detwinning), producing a large apparent strain with almost no lattice distortion resisting it.

The one-way shape memory effect

Bend a Nitinol wire while it is cold, below Mf. What actually happens at the atomic scale is detwinning, not plastic slip: the martensite variants reorient to accommodate the bend, and the wire holds that shape because there is no thermodynamic driving force pushing the twins back — martensite is still the stable phase at that temperature. Heat the wire past Af and the entire crystal reverts to the single, high-symmetry austenite structure, which has only one possible orientation; since austenite has no memory of which martensite variant it detwinned into, the lattice necessarily returns to its original, high-temperature shape. That is the one-way shape memory effect — the wire remembers its hot shape, forgets its cold one, and needs to be deformed again each time you want to repeat the trick, because cooling back down just reforms self-accommodated twinned martensite, not the bent shape.

Superelasticity: the same physics, driven by stress instead of temperature

Hold the wire at a constant temperature above Af, so austenite is the stable phase, and pull on it instead. Because austenite is metastable relative to (stress-favoured) martensite under load, mechanical stress alone can drive the transformation — stress-induced martensite. As you load the wire, stress rises elastically in austenite until it hits a plateau: further strain now comes almost entirely from austenite converting to detwinned martensite at nearly constant stress, giving several percent of strain — 6-8% is typical for Nitinol, an order of magnitude beyond the elastic strain of ordinary steel — for very little additional load. Release the wire and, because the martensite is unstable the instant the stress is removed at this temperature, it transforms straight back to austenite and the wire springs back to its original shape. This full recovery of a large strain, with no heating step at all, is superelasticity (also called pseudoelasticity), and it is what makes Nitinol vascular stents self-expanding and eyeglass frames unbendable.

The loading and unloading curves do not retrace each other — there is a stress plateau on the way up and a lower stress plateau on the way down, enclosing a hysteresis loop whose area is the mechanical energy dissipated per cycle. That dissipation is a direct consequence of the same nucleation-barrier asymmetry that gives Ms ≠ As on the temperature axis; formally the two hysteresis loops (stress-driven and temperature-driven) are two cuts through the same stress-temperature phase diagram, related by a Clausius-Clapeyron-like slope dσ/dT that is roughly constant for a given alloy composition — meaning a hotter superelastic wire needs a higher stress plateau to transform, which is exactly the tunable knob device engineers use to set stent radial force at body temperature.

Twinning versus dislocation slip

The reason Nitinol recovers such large strains without damage, where an ordinary metal would yield permanently, is that detwinning is fully reversible: it moves twin boundaries, not dislocations, and moving a twin boundary back simply reverses the same crystallographic shear with no atoms left behind out of place. Ordinary plastic deformation, by contrast, moves dislocations through the lattice, and dislocation motion is not reversible — removing the load leaves permanent atomic-scale damage behind. Exceed Nitinol's recoverable strain limit (push detwinning too far, or load at too high a stress) and ordinary dislocation slip does eventually take over, which is why superelastic devices are engineered to stay safely inside the transformation strain window.

Frequently asked questions

What is the actual physical difference between martensite and austenite in Nitinol?

They are the same NiTi atoms in two different crystal structures: austenite is a cubic B2 lattice, stable at higher temperature and stiffer; martensite is a lower-symmetry monoclinic B19' lattice, stable at lower temperature. The transformation between them is diffusionless — a coordinated shearing of the existing lattice, not atoms migrating — which is why it happens almost instantly and is fully reversible.

Why do Nitinol devices spring back to shape without any heating?

That is superelasticity, not the shape-memory effect. Above the austenite-finish temperature Af, applying stress alone can drive austenite into detwinned martensite, absorbing several percent of strain at roughly constant stress. The moment the stress is removed, martensite is no longer thermodynamically favoured at that temperature, so the material snaps straight back to austenite and its original shape — no heating step needed.

Why is there a gap between the transformation temperatures on heating and cooling?

Nucleating a new crystal structure from the old one requires overcoming an energy barrier, so the transformation lags behind the equilibrium temperature in both directions — cooling has to undershoot before martensite nucleates (Ms, Mf) and heating has to overshoot before austenite nucleates (As, Af). That lag is thermal hysteresis, and the same nucleation asymmetry produces the stress hysteresis loop seen in superelastic loading and unloading.

Try it live

Everything above runs in your browser — open Shape Memory Alloy and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.

▶ Open Shape Memory Alloy simulation

What did you find?

Add reproduction steps (optional)