Diffusionless: atoms that keep every neighbour
The last guide split every structural change in two. A reconstructive transformation tears the old structure down — bonds break, atoms diffuse over long distances and re-sort into the new arrangement — which is slow and needs heat and time. A displacive transformation instead nudges every atom a little, all together, without breaking the bond network, so it can race through a crystal almost instantly. The martensitic transformation is the archetype of the displacive kind: the sharpest, most dramatic example of a polymorphic change carried out entirely by cooperative shear.
Diffusionless means exactly what it says: no atom wanders. Each one shifts by only a fraction of a bond length and — crucially — keeps the same set of neighbours it started with. Because nothing has to hop from site to site, the change does not wait on thermal diffusion at all. A martensite plate nucleates and grows at close to the speed of sound in the solid, roughly 1000 m/s, so a whole region flips in about 10^-7 seconds — and it will still go at liquid-nitrogen temperatures, where diffusion is frozen dead. This is a transformation by choreography, not by migration: millions of atoms each taking one small step in unison.
The textbook case is steel. Heat it and the iron sits as austenite (沃斯田鐵), the face-centred cubic gamma phase, which dissolves a lot of carbon in its roomy octahedral holes. Cool it slowly and the carbon has time to diffuse out while the iron reconstructs into soft ferrite plus cementite. But quench it — plunge it into water — and you slam the door on diffusion. The carbon is trapped in place, and the iron has only one escape left: shear. The face-centred lattice shears into a body-centred one, but the trapped carbon jams it slightly out of true, so the product is not quite cubic ferrite but body-centred tetragonal (BCT) martensite (麻田散鐵) — a supersaturated, strained solid solution frozen mid-escape.
The Bain distortion: the least-motion path from FCC to BCC
How can a face-centred cubic lattice become body-centred by shear alone, with every atom keeping its neighbours? The answer, spotted by Edgar Bain in 1924, is that an FCC cell already contains a hidden body-centred tetragonal cell — you just have to outline the atoms differently. Take austenite with a_gamma = 3.60 angstrom. Inside two neighbouring cells you can pick out a body-centred tetragonal box whose two horizontal edges are a_gamma / sqrt(2) = 2.54 angstrom and whose vertical edge is a_gamma = 3.60 angstrom, giving an axial ratio c/a = sqrt(2) = 1.414. Ferrite is that same body-centred box but with c/a = 1. So the whole trip from FCC to BCC is simply this: squeeze that ratio from 1.414 down to 1.
THE BAIN DISTORTION: an FCC cell already hides a body-centred cell
Outline the atoms of face-centred austenite differently and a
body-centred TETRAGONAL (BCT) cell appears inside it:
austenite gamma (FCC) the hidden BCT cell
a_gamma = 3.60 A a' = a/sqrt2 = 2.54 A
o---------o c' = a = 3.60 A
| o | ----> c'/a' = sqrt2 = 1.414
o---------o
BAIN STRAIN: squeeze c' down ~20%, spread a' out ~13%
c': 3.60 A --(-20%)--> 2.87 A = a_alpha
a': 2.54 A --(+13%)--> 2.87 A = a_alpha
--------
BCC alpha-iron (ferrite), a = 2.87 A
Every atom moves only a FRACTION of a bond and keeps all its
neighbours -- no diffusion, no bond-swapping. Trapped carbon
stops c' shrinking fully, leaving BCT martensite (c/a > 1).Put the numbers to it. To reach ferrite's a_alpha = 2.87 angstrom, the vertical axis must shrink from 3.60 to 2.87 angstrom — a contraction of about 20 percent — while the two horizontal axes stretch from 2.54 to 2.87 angstrom — an expansion of about 13 percent. Those are the classic Bain numbers, and notice how gentle they are: every atom moves only a fraction of a bond spacing, and no atom ever swaps a neighbour for a new one. That is exactly why the transformation can be diffusionless. When carbon is trapped in the mix it blocks the vertical axis from shrinking all the way, so c/a settles a little above 1 (around 1.02 to 1.08, rising with carbon content) — the tetragonality of martensite is literally the carbon holding the door ajar.
The shape change and the habit plane
A cooperative shear does something a diffusive change never does: it changes the SHAPE of the transformed region, not just its structure. Picture a deck of cards pushed sideways — the stack keeps every card, but the block tilts into a slanted prism. When a lens-shaped plate of martensite forms just under a polished metal surface, it tilts that patch of surface visibly; under a microscope you see surface relief, a tiny cliff where flat metal has heaved. That relief is the fingerprint of a displacive change — direct, visible proof that atoms sheared rather than reshuffled. There is a small volume change too (steel expands by a few percent on forming martensite), which is why quenched parts build up huge internal stresses and can crack.
Every plate lies flat along one particular crystal plane of the parent, and that plane is the habit plane — the flat interface between the untransformed austenite and the new martensite. It is special because it is an INVARIANT plane: as the shear sweeps through, this one plane is neither stretched nor rotated, so parent and product meet along it without a gap or a pile-up. That is what lets the interface glide fast — it is a glissile, semicoherent boundary, an interphase boundary that can move like a wave rather than by atoms hopping across. In iron-carbon steels the habit plane is often a high-index, irrational plane — close to {111}, {225} or {259} depending on carbon content — not one of the tidy low-index planes, which is your first hint that something subtle is happening inside the plate.
Orientation relationship, and the twins hiding inside
Because every atom kept its neighbours, the new lattice cannot be turned any old way relative to the old one — the two are locked together in a fixed orientation relationship. For FCC-to-BCC iron the best-known is the Kurdjumov-Sachs relation: a close-packed {111} plane of the austenite stays parallel to a {110} plane of the martensite, and a <110> direction in the parent lines up with a <111> direction in the product. (The Nishiyama-Wassermann relation is a close cousin, rotated by only about 5.3 degrees.) These are not guesses — you can measure the two lattices' orientations side by side with electron backscatter diffraction or with electron diffraction in a microscope and read the relationship straight off.
Here is the twist that makes martensite crystallography a real theory rather than a picture. The Bain strain alone does NOT leave any plane invariant — squeeze one axis 20 percent and stretch two others 13 percent and every single plane comes out either stretched or tilted. So the Bain strain by itself could never produce that clean, glide-ready habit plane. Nature's fix is to add a lattice-invariant shear: an extra shear that reshapes the plate but leaves the new crystal structure untouched, because it is carried by either fine internal twinning or ordinary slip. Twins and slip both re-stack a crystal without changing what the crystal IS — so they can adjust the plate's overall shape for free.
- Start with the Bain strain — the pure lattice deformation that turns FCC into BCT by compressing one axis about 20 percent and expanding the other two about 13 percent. On its own it leaves no undistorted plane, so it cannot be the interface.
- Add a lattice-invariant shear — fine transformation twins or slip — that changes the plate's shape but leaves the martensite lattice exactly as it is.
- Add a rigid-body rotation that swings the whole plate around.
- Tune all three together so their combined effect is an invariant-plane strain: exactly one plane ends up unstretched and unrotated. That plane is the predicted habit plane — and the same calculation spits out its irrational indices, the orientation relationship, and the visible shape change, all at once.
This is why a martensite plate is not a plain single crystal but is shot through with fine, parallel internal twins — under a transmission microscope they look like a herringbone or a stack of ultrathin mirror-image lamellae, each pair a twin of its neighbour. Think of pleating a strip of fabric: each pleat folds the cloth, yet the average line of the hem stays straight. High-carbon steels take the twinning route and form finely twinned plate martensite; low-carbon steels instead use slip and form dislocated lath martensite. Either way, the lattice-invariant shear is the plate's private accounting trick for making its outer shape fit an invariant habit plane.
Why one shear matters: hard knives and metals with memory
Now the payoff that made martensite famous: it is what makes quench-hardened steel hard. The carbon locked into the body-centred tetragonal lattice strains it severely, and the dense mesh of internal twins or dislocations gives any would-be gliding dislocation almost nowhere to go — so the steel resists deformation fiercely. But be precise about WHY. Carbon-free iron martensite is actually soft, barely harder than ferrite; the hardness comes from the trapped carbon and the fine substructure, not from the letters 'BCT'. That distinction is the whole difference between a razor blade and a paperclip, both made of iron.
Martensite also keeps unusual time. It starts forming at a fixed temperature Ms (martensite start), finishes near a lower Mf, and — because it is diffusionless — the amount you get depends on how far below Ms you have cooled, not on how long you wait there. Hold the steel steady just below Ms and the transformation simply stops; cool another 20 degrees and more plates snap in. This is called athermal behaviour, and Ms falls steeply as you add carbon (from around 500 degrees C in low-carbon steel down toward room temperature in high-carbon steel), which is why very high-carbon steels can retain soft austenite unless you chill them further. (Honest caveat: a few alloys show isothermal martensite that does grow with time, so 'always athermal' is a strong tendency, not an ironclad law.)
And here is the loveliest part: because the shear keeps every atomic neighbour, it can be run backwards. Warm a shape-memory alloy like nitinol (NiTi) and its martensite reverts cleanly to the parent phase along the very same orientation relationship — the plates un-shear and the metal springs back to a remembered shape. That is the basis of shape-memory and superelastic wires, of stents and eyeglass frames. And martensite is emphatically NOT a steel-only trick: the same diffusionless shear toughens zirconia ceramics, where a stress-driven tetragonal-to-monoclinic martensitic change expands by a few percent right at a crack tip and squeezes the crack shut. Any crystal that can shear from one polymorph to another, faster than its atoms can diffuse, can go martensitic.