Causes of Deformation in the Thermal Treatment of Steel Parts

When you austenitize steel, the crystal structure changes from body-centered cubic ferrite and pearlite to face-centered cubic austenite. Austenite packs atoms more efficiently, so the volume per unit mass drops during heating. On cooling, that same volume comes back, and when the part transforms to martensite it comes back larger than it started. A 1045 or 4140 section can grow several tenths of a percent in length, which on a 200 mm shaft is a visible number, not a rounding error.

Carbon content drives how much of this you get. Low-carbon steels such as 1018 form little martensite and move less. Medium-carbon grades like 4140 and 4340 transform more completely, so the volume change is larger. Alloying elements shift the transformation start and finish temperatures, which changes when the growth happens relative to the cooling curve.

The shape of the part decides where the growth lands. A thin wall cools fast and transforms early; a heavy boss cools slowly and transforms late. The two regions end up fighting each other, and the part bends toward whichever side finishes transforming last. Uniform section thickness is the single most effective fix, and it is decided at the design stage, not in the furnace.

Not every steel behaves the same way. Maraging and precipitation-hardening grades such as 17-4PH shrink slightly during aging instead of growing, because the transformation path is different. Tool steels with high carbide content move less in one direction and more in another after rolling, which is why directionality matters for long parts.

During heating and cooling, the surface and the core are never at the same temperature. The surface wants to expand or contract first; the core resists. That mismatch produces thermal stress, and once it passes the yield strength of the steel at that temperature, the deformation becomes plastic and permanent.

Quenching makes this worse because the cooling rate is high by design. A water-quenched 1045 part can see surface-to-core temperature differences of several hundred degrees in the first seconds. Oil and polymer quenchants slow the early stage and cut the gradient, at the cost of a shallower hardened case.

Section size sets the ceiling here. A 25 mm round bar in 4140 quenched in oil can usually be brought through without exceeding yield. The same steel at 100 mm diameter needs a different quenchant or a different hardenability grade, because the core simply cannot follow the surface.

Acute stress shows up as a crack. Chronic stress shows up as a bend or twist that appears after the part is already cool. Both come from the same source, and both respond to slower, more even cooling rather than to a tighter final tolerance.

Stress that was already inside the blank does not disappear when the part goes into the furnace. It relaxes, and relaxation means movement. Cold-drawn bar, cold-rolled plate, and heavily ground surfaces all carry stress from the operations that made them.

Machining adds its own layer. A heavy roughing pass with a dull tool leaves a tensile skin. The part holds its shape while the stressed skin is balanced against the core, then bows once heat treatment releases the balance.

Pre-hardening stock removal is the standard countermeasure. Leave 0.5–1.5 mm per side after roughing, stress-relieve at 550–650 °C, then finish. The number depends on how much material was removed and how asymmetric the part is.

Cold work is a related trap. A bent or straightened shaft carries stress at the bend. It looks straight on the bench and comes out of the quench curved. Straightening after heat treatment helps only if the stress is redistributed, not just reversed.

In practice, the three mechanisms overlap. A part with unequal sections transforms unevenly, cools unevenly, and releases stress unevenly, all at the same time. Separating them after the fact is difficult, so the useful work happens before heat treatment.

A quick diagnostic order helps. Check whether the part grew or shrank overall, since growth points to transformation and a pure bend points to stress. Check whether the distortion repeats across a batch, because a repeatable pattern is a design or process issue rather than a random furnace problem.

A batch that moves in different directions from part to part usually points to loading. Parts stacked too tightly cool at different rates on different faces. Spacing them on a fixture, or hanging them, often removes more variation than any change to the quench tank.

Fixtures cost money and floor space. For a one-off prototype they are rarely worth it. For a 10,000-part run, a fixture that holds a 0.05 mm straightness band usually pays for itself in the first few batches of scrap it prevents.