Shape Memory Alloy

The trick: a reversible solid-solid phase change

Most metals deform plastically by moving dislocations through their crystal lattice. Push hard enough on a copper bar and you bend it; the bar stays bent because dislocations have rearranged the atoms into a new equilibrium that has no memory of the original geometry. A shape-memory alloy cheats this by deforming through a different mechanism entirely — a reversible diffusionless transformation between two crystal structures. The low-temperature phase, martensite, accommodates strain by reorienting twin variants, not by gliding dislocations. The high-temperature phase, austenite, is a stiff parent lattice that remembers its trained shape. Cycle between the two by heating and cooling, and the macroscopic geometry follows.

The discovery is usually traced to William Buehler at the U.S. Naval Ordnance Laboratory in 1959, who was developing a new alloy for missile nose cones. The acronym Nitinol stands for Nickel Titanium Naval Ordnance Laboratory. Buehler noticed that a bar of his roughly equiatomic Ni-Ti, dropped on a workbench, made a different sound — duller and more lead-like — when cold than when it was warm from a soldering iron. Tracking down the difference led to the realisation that the alloy was undergoing a reversible phase change near room temperature. The mechanical effect — bend cold, heat, snap back — followed within a couple of years and the alloy was patented in the mid-1960s.

The two phases — and the temperatures that switch them

Nitinol exists in two crystallographically distinct phases that are stable in different temperature ranges and accommodate strain in completely different ways.

Phase	Crystal structure	Stable when	Modulus	Strain mechanism
Austenite (parent)	Cubic B2 (CsCl-type)	T > A_f	~75 GPa	Elastic; further deformation forms stress-induced martensite
Martensite	Monoclinic B19′	T < M_f	~28 GPa	Twin reorientation (detwinning) — large strain at constant stress
R-phase (intermediate)	Rhombohedral	Narrow band on cooling	~50 GPa	Small reversible strain; tightens hysteresis

The four transition temperatures — martensite-start M_s, martensite-finish M_f, austenite-start A_s, austenite-finish A_f — together define the operating envelope of the alloy. On cooling, austenite begins transforming at M_s and is fully martensite by M_f. On heating, martensite begins reverting at A_s and is fully austenite by A_f. The transition window is offset: typically A_s is 20 to 50 K above M_s, with the gap between A_f and M_f forming the thermal hysteresis loop that gives Nitinol its actuator-like behaviour. In commercial Nitinol the entire window can be placed anywhere from cryogenic temperatures to roughly 100 °C by tweaking composition and aging.

The shape-memory effect, step by step

The classical one-way shape-memory effect proceeds through a four-stage cycle:

1. Train. Constrain the part to its desired austenite geometry in a steel jig and anneal at roughly 500 °C for a few minutes (precise time and temperature are alloy-specific), then quench in water. The anneal sets the austenite memory.
2. Cool to martensite. Below M_f the alloy transforms into a self-accommodated twinned martensite. Because the twin variants are arranged so that the macroscopic strain averages to zero, the part remains visibly in its trained shape — there is no spontaneous distortion on cooling.
3. Deform below A_s. Apply load. Instead of dislocation glide, the alloy responds by detwinning — twin variants reorient to favour the load direction. Stress-strain shows a low, flat plateau between roughly 1 percent and 7 percent strain. On unloading, the part stays in the detwinned configuration; it looks plastically deformed.
4. Heat above A_f. Martensite reverts to austenite. The parent lattice has only one possible orientation, and the macroscopic shape follows back to the trained geometry. The visible "memory" is the consequence.

Two-way shape-memory alloys — which remember both a hot and a cold shape — are produced by training cycles that introduce oriented dislocations or precipitates favouring a specific martensite variant, so that cooling itself produces a macroscopic strain. The two-way effect is smaller (typically 2 percent rather than 8) and degrades faster with cycling, so most commercial parts use the one-way effect with a return spring providing the reset force.

Pseudoelasticity — superelastic recovery without heat

Operating above A_f gives a different and arguably more useful behaviour. The alloy is austenite at rest, but applying stress drives a stress-induced martensitic transformation. The phase change is itself the deformation mechanism — strain accumulates as the volume fraction of martensite grows, not as dislocations move. On unloading, the stress drops below the level needed to stabilise martensite, the reverse transformation runs, and the strain is recovered.

σ
↑
│   ┌─────────────────── loading plateau (austenite → martensite)
│   │
│  ╱│
│ ╱ │
│╱  └────────────────── unloading plateau (martensite → austenite)
└──────────────────────→ ε
       6 – 8 %

The macroscopic signature is a flag-shaped stress-strain loop with two flat plateaus separated by a hysteresis. The plateau strain — the difference between elastic onset and the end of the upper plateau — is the recoverable pseudoelastic strain, typically 6 to 8 percent in well-treated Nitinol. By comparison, the elastic strain limit of even high-strength steel is about 0.2 percent. That order-of-magnitude difference is what makes pseudoelastic Nitinol so useful in geometries that need to be deformed dramatically — folded, crimped, threaded through a catheter — and then released to recover.

Tuning the transition: composition and processing

The transition temperature is one of Nitinol's defining design parameters and one of its trickiest. Empirically, A_f drops by roughly 100 K for each additional 1 atomic percent of nickel above 50 at%, holding everything else constant. That sensitivity means melt composition control during alloy production must be tight, often to better than 0.05 at% Ni. Aging treatments alter the matrix composition by precipitating Ni4Ti3, which removes nickel from solution and shifts A_f upward; cold work introduces dislocations that raise the transition temperature and broaden the hysteresis.

Application	Mode	Target A_f	Why
Self-expanding stent	Pseudoelastic	5 – 15 °C below body T (37 °C)	Stay fully austenite in vivo so deployment is pure pseudoelasticity
Orthodontic archwire	Pseudoelastic	Around body T	Continuous low force across mouth temperature range
Eyeglass frame	Pseudoelastic	~ -10 °C	Always austenite at room temperature, large recoverable bending strain
Pipe coupling (CryoFit, Raychem)	One-way SME	Cryogenic (~ -100 °C)	Cool to fit oversize over pipe, warm to room T to shrink-grip
Thermal actuator (HVAC, valves)	One-way SME	40 – 80 °C	Switches state when ambient or process temperature crosses A_s
High-temperature aerospace actuator	One-way SME	150 – 400 °C	Requires ternary Ti-Ni-Hf or Ti-Ni-Pd

Worked example: how a self-expanding stent deploys

Consider a Nitinol stent designed to expand to an 8 mm internal diameter in a peripheral artery. The alloy is processed to give A_f = 27 °C — comfortably below the 37 °C body temperature. The stent is laser-cut from a Nitinol tube at the open geometry and given a shape-setting anneal so that 8 mm is its austenite memory.

1. Loading. At room temperature the stent is mechanically crimped down to about 2.5 mm outer diameter inside a polymer sheath. Because room temperature is at or just below A_f, much of the crimping strain forms stress-induced martensite — the alloy stores elastic-like energy without yielding.
2. Delivery. The sheathed device is threaded through the catheter to the lesion. Throughout transit, the constraint of the sheath holds martensite stable despite body temperature.
3. Release. The clinician retracts the outer sheath. The constraint disappears; stress drops below the level required to maintain martensite at 37 °C; the reverse transformation runs, and the lattice reverts to austenite. The stent recovers toward its 8 mm memorised diameter.
4. Chronic outward force. Once the stent meets the vessel wall, it cannot fully unfurl. The residual strain leaves the device perpetually trying to reach its austenite memory, generating a low, steady chronic outward force on the artery — well below the barotrauma threshold but enough to keep the lumen patent.

Compare with a balloon-expandable stainless-steel stent: there, deployment plastically yields the metal at high local strain rates, and chronic force is essentially zero — the stent is mechanically anchored, not actively pushing. The clinical consequences differ. Nitinol stents recover their shape after external deformation (important in superficial femoral arteries, where the leg's motion repeatedly crushes them) but their continuous outward push can also drive in-stent restenosis in some vessel beds.

Where shape-memory alloys show up

• Self-expanding cardiovascular stents. Carotid, peripheral, and increasingly coronary stents are laser-cut Nitinol with A_f below body temperature. The same physics powers neurovascular flow diverters for treating intracranial aneurysms.
• Orthodontic archwires. Nitinol archwires deliver continuous, low force across a wide range of tooth displacements because the pseudoelastic plateau is flat — the force the wire applies barely changes as teeth move into the arch form, unlike a stainless-steel wire whose force decays linearly with deflection.
• Eyeglass frames. "Kink-resistant" or "memory" frames use superelastic Nitinol bridges and temples. Sit on them, they fold flat; release, they snap back.
• Mars rover wheels. NASA Glenn's Spring Tires, developed for Perseverance and beyond after Curiosity's aluminium wheels tore on Gale Crater rocks, use a woven Nitinol mesh that takes percent-level strains and recovers without permanent damage. Operating temperature range on Mars: roughly -125 °C to 20 °C.
• Pipe couplings (CryoFit / Tinel-Lock). Manufactured oversize at room temperature in martensite, cooled below M_f, mechanically expanded, then slipped over the pipe joint. Warming to room temperature drives the SME — the coupling shrinks down onto the pipe with a precisely defined chronic radial force. Used on F-14 hydraulic lines from the early 1970s and the technology remains in service.
• MEMS microactuators. Thin-film Nitinol deposited by sputtering can be patterned into microvalves, microgrippers, and active fluidic devices that switch state in milliseconds at small length scales, where the heat capacity is low enough to allow rapid cycling.
• Robotic and biomimetic actuators. Nitinol wires used as artificial muscles in compliant robotic fingers, surgical end effectors, and aerospace control surfaces. Force-per-mass is high; bandwidth is limited by cooling rate.
• Vibration damping and seismic devices. Pseudoelastic SMAs dissipate energy in their stress-strain hysteresis without yielding, making them attractive for seismic isolation devices and as energy-absorbing braces in buildings — particularly because they self-centre after large displacements.

Failure modes and fatigue

The principal lifetime limit of pseudoelastic Nitinol is functional fatigue — gradual accumulation of residual strain, dislocation buildup, and decay of the recoverable strain after many transformation cycles. In stents this matters because pulsatile blood pressure cycles the device hundreds of millions of times over a patient's lifetime. Modern medical Nitinol is specified to allow safe strain amplitudes typically under 0.4 percent for 10⁸ cycles, far below the static recoverable maximum. Surface inclusions and tube-drawing defects act as fatigue initiation sites, and the lifetime is dominated by surface quality more than by bulk composition.

Other failure mechanisms include thermomechanical drift (transition temperature creeps with cycling because dislocations accumulate), corrosion (Nitinol's biocompatibility comes from a passive TiO2 surface layer — abrade or scratch it and corrosion accelerates), and nickel leaching from poorly passivated surfaces, which has driven decades of work on coatings and electropolishing routines.

Other shape-memory systems

• Cu-Al-Ni and Cu-Zn-Al. Copper-based SMAs. Cheaper than Nitinol, higher transition temperatures (up to ~200 °C), but smaller recoverable strain (~5%), lower fatigue life, and prone to intergranular fracture. Used in industrial actuators and pipe couplings.
• Fe-Mn-Si. Iron-based SMAs developed primarily as structural materials. Recoverable strain is small (typically < 2%) but the alloys are weldable, cheaper than Ni-Ti, and used as crimp couplings, civil-engineering reinforcement, and prestressed concrete tendons.
• High-temperature SMAs. Ti-Ni-Hf, Ti-Ni-Pd, Ti-Ni-Pt, and Ti-Ni-Zr push the transition temperature into the 200 – 600 °C range for aerospace actuators (chevron exhaust devices, variable geometry inlets).
• Magnetic shape-memory alloys. Ni-Mn-Ga and related compositions reorient martensite variants under magnetic field rather than heat. Bandwidth is kilohertz rather than hertz, but recoverable strain is smaller (~6% under field) and the materials are brittle.
• Polymeric shape-memory. Cross-linked polymers with a thermal glass transition can also "remember" shapes; the mechanism is different (rubbery vs glassy network entropy rather than martensitic transformation) but the macroscopic behaviour rhymes.

Common pitfalls when designing with SMAs

• Treating transition temperature as a single number. M_s, M_f, A_s, A_f are four distinct temperatures with a hysteresis loop between heating and cooling branches. Designs that rely on switching at "the" transition temperature get burned when ambient drift puts the alloy partly transformed.
• Ignoring stress dependence. Loading raises the effective transition temperature (Clausius-Clapeyron slope ~6 K/MPa for Nitinol). A stent crimped in a delivery system sees enough stress to stabilise martensite even at body temperature; that is the point, but it also means the device's "switching temperature" is not the same as its unstressed A_f.
• Underestimating the cooling-rate limit. An SMA actuator transforms quickly on heating (joule heating into a thin wire can run at hundreds of hertz), but cooling is set by convective or conductive heat removal. Bandwidth in air is typically less than a few hertz. Forced fluid cooling, thin-film geometry, or magnetic activation are the workarounds.
• Cold-working a finished part. Bending a shape-set Nitinol wire enough to introduce dislocations alters its transition temperatures and degrades fatigue performance. Forming should be done before final shape-set, or in martensite at moderate strain.
• Welding. Nitinol can be laser- or resistance-welded, but the heat-affected zone often has degraded transformation behaviour because of compositional shifts (TiO and TiC formation from atmosphere or contaminants) and grain growth. Welded joints rarely match the parent alloy's pseudoelastic strain capacity.
• Surface finish vs. fatigue. The microstructural perfection of the surface dominates fatigue life. Mechanical polishing, electropolishing, and pickling routines must be tuned for each alloy lot; a "looks shiny" surface is not the same as a fatigue-qualified one.