Guide
How much stock should you leave for heat treatment distortion, and when must finishing come after hardening?
Heat treatment distortion is the change in size and shape a steel part goes through when it is hardened: it grows or shrinks as its structure transforms, and it bows, twists or goes out of round as uneven cooling and released residual stress pull it about. There is no universal stock figure that covers it. The finish allowance on each surface has to cover the size change, the shape change, any scale or decarburized skin, and the cleanup the finishing process needs, and every surface that must hold a tight tolerance or fit has to be finished after hardening.
By Saola Manufacturing engineering team · Published · Updated
The usual complaint arrives after the lot comes back from the heat treater: bores that no longer take a gauge, shafts with runout that was not there before, flat plates that rock on a surface plate, a dark crust on every surface, a handful of parts with a fine crack at a corner, and a few that are now undersize after the scale was blasted off. Each of these has a different cause and a different fix, and most of them were decided on the drawing and in the routing long before the furnace. This guide sorts the symptoms, shows how to diagnose a rejected lot, explains how to build up a stock allowance from evidence instead of habit, and lists the notes that stop the problem from coming back.
On this page
What does heat treatment distortion look like, and how is it different from scale, decarburization or cracking?
Distortion is a dimensional problem: features are the wrong size or the wrong shape but the metal itself is sound. Scale and decarburization are surface problems, and a quench crack is a material failure that makes the part scrap whatever its dimensions.
Distortion research treats size changes and shape changes as two separate effects, and buyers often lump them together (the ASM Handbook chapter by Lübben, ASM Handbook Volume 4B, is organized around that split). Size change is the more or less uniform growth or shrinkage that comes from the change in steel structure on hardening. A carburized or hardened steel part usually grows, while a precipitation-hardening stainless grade contracts: the Precision Machined Products Association gives 0.0004 to 0.0006 in/in for 17-4 PH aged from Condition A to H900 (PMPA Craftsman’s Cribsheet 67). Shape change is everything that is not uniform: a shaft that bows, a ring that goes oval, a plate that cups, a bore that tapers. Size change is fairly repeatable for a given steel and process and can be compensated for. Shape change depends on the part geometry, the residual stress already in the part and how it was loaded and quenched, and it is what eats a stock allowance.
Scale is the oxide crust left when steel is heated in air or in a poorly controlled atmosphere. It costs metal: once it is blasted or pickled off, the surface underneath sits below where it started, which is one reason parts measure undersize after hardening. Decarburization is the carbon-depleted skin that often forms under the scale; it measures soft on a surface hardness test and wears or fatigues early, and it has its own guide, linked below. A quench crack is a brittle fracture that opens during or soon after quenching, usually at a sharp corner, a hole, a keyway or a sudden change in section. On a rejected lot, sort the parts into these four groups before arguing about who caused what.
What causes distortion, scale, undersize parts and quench cracks after hardening?
The biggest causes are uneven cooling of uneven sections, residual stress left by earlier machining, forming or straightening, and poor support or quench orientation in the furnace. Scale comes from the furnace atmosphere, and quench cracks come from a severe quench meeting a stress raiser or a delay before tempering.
Distortion is best treated as a property of the whole process chain, not of the furnace alone. Stress put into a bar by drawing, straightening or heavy roughing cuts relaxes when the part reaches hardening temperature, and the part moves to release it; that is why a part that measured straight after machining can bow without anything going wrong in the quench. A stress relief after rough machining, covered in the stress relieving guide, removes much of that stored movement before the final machining passes.
Geometry decides the rest. A thin web next to a thick boss cools at a different rate, transforms at a different time and pulls the part out of shape. A long slender part sags under its own weight at temperature if it is laid flat, and twists if it enters the quench at an angle. The quench itself is a trade-off: a faster quench reaches full hardness in a lower-hardenability steel but raises both distortion and cracking risk, while an oil, polymer or gas quench on a higher-hardenability grade moves the part less. A trade article from the Precision Machined Products Association lists seven causes of quench cracking, from overheating and an overly aggressive quenchant to poor design and a delay before tempering, warning that work not tempered right away can crack (PMPA, 7 causes for quench cracking of steel).
| Symptom on the returned lot | Likely causes, most common first | Check first |
|---|---|---|
| Bow, twist or runout on long parts | Released residual stress from bar or roughing; part laid flat or unsupported at temperature; uneven quench entry | Was a stress relief done after roughing? How were parts racked and quenched? |
| Rings or bores out of round, flat parts cupped | Unequal wall or section; asymmetric features; part quenched flat face down | Wall thickness variation on the drawing; load orientation on the heat treat record |
| Uniform growth, bores too small, pins too big | Normal transformation growth not allowed for in the pre-heat-treat size; a process or temper different from the one the allowance was set on | Hardness against the specified range; compare size change across the lot and against the pilot parts |
| Undersize after cleanup | Metal lost to scale and its removal; finish stock too small to clean up the shape change; contraction on aging of a precipitation-hardening grade | Furnace atmosphere on the record; stock left before heat treatment |
| Dark, flaking crust | Heating in air or a poorly controlled atmosphere; no protective wrap or vacuum on tool steels | Atmosphere or vacuum on the heat treat record; decarburization depth on a section |
| Fine cracks at corners, holes, keyways or markings | Sharp internal corners and abrupt section changes; quench too severe for the steel; delay before tempering; overheating; deep stamped part marks; parts packed so the quenchant cannot reach every face; seams left by too little stock removal | Magnetic particle test of the whole lot; time from quench to temper |
| Checked or crazed cracks on ground faces | Grinding burn: too aggressive a grind on hardened steel, or grinding before tempering | Etch or crack test of ground faces; grinding parameters |
How do you diagnose a heat-treated lot that fails inspection?
Measure the same features before and after heat treatment on a few serialized parts, check hardness, and crack-test the whole lot. Without before-and-after data on the same parts, nobody can say whether the machining, the material or the heat treatment moved the part.
Start with a diagnostic checklist a buyer or quality engineer can run on the returned lot:
- Hardness first. Check surface hardness on a few parts per the method on the drawing, for example Rockwell per ASTM E18. A part that is out of the hardness range is a heat treatment nonconformance whatever its dimensions, and a soft surface on otherwise hard parts points to decarburization.
- Size versus shape. Compare diameters, lengths and bores against the pre-heat-treat readings. If every part grew by a similar amount, it is size change and can be compensated for next time. If the scatter is large, or the parts bowed or went oval, it is shape change and the fix is in stress relief, geometry, support or quench.
- Runout and flatness, measured properly. Bow on a shaft is best read as runout between the functional datums, as defined in ASME Y14.5; reading it between centers that were themselves distorted gives a false answer.
- Cracks. Inspect every part of a cracked lot, not a sample, using magnetic particle testing per ASTM E709 on ferromagnetic steel. Quench cracks typically follow sharp corners and section changes. If a cracked part is broken open, a lab can often tell from oxide or temper color on the crack faces whether the crack was open before tempering; ask for that finding rather than guessing. Grinding cracks are shallow and form a checked pattern on the ground face.
- Skin. If scale or soft surfaces are present, have a cross-section measured for decarburization depth per ASTM E1077, because that depth sets part of the allowance for the next lot.
- Records. Ask for the heat treat record: steel and heat number, furnace atmosphere or vacuum, how parts were loaded, the quenchant, the temper cycle and the time between quench and temper. General process requirements such as these are the subject of SAE AMS2759, and naming it on the order makes the record a deliverable rather than a favor.
The decision on the lot follows from the sort. Distorted parts with enough stock left can often be recovered by finish grinding or hard turning; undersize, cracked or wrong-hardness parts usually cannot. Straightening hardened parts is possible on some geometries, but it puts stress back in and can start a crack, so it should be an agreed, recorded step followed by a stress relief, never a silent rework.
How much finish stock should you leave, and which features must be finished after hardening?
Leave enough stock per side to cover half the measured size change, half the measured bow or runout, the deepest scale or decarburization you will accept, and the cleanup the finishing process needs. Any surface that carries a fit, a bearing seat, a seal, a tight geometric tolerance or a hard wear face is finished after hardening; everything else can be finished before.
The number depends so heavily on steel, geometry, size and process that a figure quoted without that context is a guess. Growth can be large enough to plan around: a US patent on gear grinding gives the example of a carburized gear expected to grow about 0.012 in on a 12 in diameter, compensated for by offsetting the cut made before heat treatment (US Patent 8,353,739). For a new part, the reliable route is a pilot: machine a few parts to the pre-heat-treat drawing, serialize them, measure the critical features, heat treat them with the production process, measure again, and size the allowance from the worst part with a margin. After that, the allowance only needs revisiting if the steel source, the geometry or the heat treat process changes.
Too much stock is not the safe side either. Every extra amount on a hardened surface costs grinding or hard-turning time, raises the risk of grinding burn, and on a case-hardened part removes the hardest layer of the case, so the allowance has to be set together with the case depth that must remain after finishing.
| Allowance component | Where the value comes from | Note |
|---|---|---|
| Size change | Half the largest diametral growth or shrinkage measured on pilot parts | Repeatable; can instead be offset in the pre-heat-treat dimension |
| Shape change | Half the largest runout, bow or out-of-round measured on pilot parts | The largest and least predictable term; reduce it with stress relief and support |
| Scale and decarburized skin | Deepest depth measured on a cross-section, or the limit on the drawing | Near zero for vacuum or protective-atmosphere hardening |
| Finishing cleanup | The minimum the grinding, hard-turning or EDM pass needs to cut cleanly | Agree it with whoever finishes the part |
| Case depth to retain | Specified effective case depth after finishing | Case-hardened parts only; total allowance must leave the case |
| Feature or requirement | When to finish | How it is usually finished after hardening |
|---|---|---|
| Bearing seats, fits, seal diameters | After | Cylindrical grinding or hard turning |
| Dowel and precision bores | After (drill and ream undersize before) | Internal or jig grinding, or wire EDM |
| Flat datum faces, tight flatness or parallelism | After | Surface grinding |
| Profiles, slots and shapes in hardened tool steel | After | Wire or sinker EDM, which cuts fully hardened steel; on fatigue-critical faces, ask how the thin recast layer it leaves will be removed |
| Threads | Before, gauged after | Thread grinding only if the gauge check fails or the class demands it |
| Clearance holes, chamfers, non-critical pockets | Before | Not refinished |
| Nitrided or precipitation-hardened parts | Mostly before | Low distortion because there is no quench; allow for the small, predictable size change |
What should the drawing, RFQ and PO say so distortion and cracking do not recur?
Put the hardness requirement, the test method, the surfaces finished after heat treatment and the stock left on them, and the crack and skin limits on the drawing itself. A bare note such as “harden and grind” leaves every one of those choices to the shop and the heat treater.
- Hardness and method. The range, the scale and where it is tested, for example “Harden and temper to 50-55 HRC per ASTM E18, test location as shown.” The hardness scale guide covers which scale suits the part.
- Dimensions after heat treatment. State that all dimensions and geometric tolerances apply after heat treatment and finishing, and flag the surfaces to be finished after hardening with a symbol or a note.
- Pre-heat-treat stock. If you own the routing, give the pre-heat-treat size or the stock per side on those surfaces; if the supplier owns it, say so and ask them to state their allowance on the quote.
- Stress relief. “Stress relieve after rough machining” where the part is long, thin, asymmetric or cut from heavily drawn bar.
- Design for the quench. Generous internal radii instead of sharp corners, balanced sections, no holes close to edges, and part markings by a method that does not cut a notch; ask the supplier to flag features that raise cracking risk at quote stage.
- Skin and cracks. A decarburization limit on finished surfaces, “no scale on finished part,” and “100% magnetic particle inspection per ASTM E709, no cracks permitted” where cracking would be dangerous.
- Process and records. Name the process specification, such as SAE AMS2759 and the slash sheet for the steel, temper promptly after quench, and require a heat treat record and a hardness report with the shipment.
Straightening deserves its own line on the PO: either “straightening not permitted” or “straightening permitted, followed by stress relief and crack test.” Silence on the point is how straightened parts with hidden cracks reach an assembly line.
Who does what on a Saola order with a hardened steel part?
Heat treatment is not run at the lead CNC plant where the machining happens. When a drawing calls for hardening, the hardening is arranged with a partner plant we manage, and the quote names the production site and the country of origin for each step.
Rough and semi-finish turning and milling, the pilot parts and the pre-heat-treat measurements are done on the lathes and machining centers at our lead CNC plant, as described on the CNC turning and CNC milling pages. EDM and grinding/polishing equipment are on the lead CNC plant’s equipment list, together with a hardness tester, a height gauge and a 2D vision measuring machine, which is what makes finishing after hardening and the before-and-after comparison practical. Whether a given ground fit can be held there, or needs to be arranged elsewhere, is stated on the quote rather than assumed. Crack testing and decarburization sections need a lab, and the quote says where they would be done. Any coating after hardening goes to partner plants we manage through the surface finishing route.
The most useful package is the drawing with the hardness, finish-after-hardening and crack notes already on it, plus your intent on the routing and stock if you have one. Material certification is available on request, and ISO 9001 certification is in progress, not held. Send it through the quote form.
Frequently asked
Can heat treatment distortion be eliminated completely?
No. Size change comes from the transformation that makes the steel hard, so some movement always occurs. It can be reduced with stress relief, balanced geometry, good support, a gentler quench on a higher-hardenability steel, or a process without a quench such as nitriding, and the rest is handled by finish stock.
Is it cheaper to finish everything before hardening?
Only when the tolerances are loose enough to absorb the movement. If a fit or a datum face moves out of tolerance, the parts are scrap, so it is usually cheaper to grind or hard-turn the few critical surfaces than to tighten every feature before heat treatment.
Why did my parts come back undersize when steel grows on hardening?
Usually because scale consumed metal and was then blasted off, or because the finish stock was too small to clean up the shape change and the grinder had to go below size to clear the low spots. Precipitation-hardening stainless grades also contract slightly on aging.
Can a quench-cracked part be repaired?
Not reliably. A crack is a fracture that will grow in service, and grinding it out changes the dimensions and may leave the root. Scrap cracked parts, crack-test the rest of the lot, and fix the cause before the next run.
Who is responsible for distortion, the machinist or the heat treater?
It depends on who owned the routing and the allowance. If the drawing set the pre-heat-treat size, the designer owns it; if the supplier set the routing, they do. Writing the allowance and the finish-after-hardening surfaces on the drawing or quote settles this before the first lot.
Sources
- Basics of Distortion and Stress Generation during Heat Treatment (T. Lübben), ASM Handbook Volume 4B, Steel Heat Treating Technologies (abstract), ASM International, 2014
- AMS2759G, Heat Treatment of Steel Parts, General Requirements (scope), SAE International, 2019
- 7 Causes for Quench Cracking of Steel, Precision Machined Products Association, 2010
- Craftsman’s Cribsheet 67: Dimensional Contraction of 17-4 PH Stainless Steel, Precision Machined Products Association (PDF posted 2020)
- US Patent 8,353,739, Method for detecting and/or preventing grind burn, US Patent and Trademark Office, 2013
- ASTM E18-25, Standard Test Methods for Rockwell Hardness of Metallic Materials (scope), ASTM International, 2025
- ASTM E709-21, Standard Guide for Magnetic Particle Testing (scope), ASTM International, 2021
- ASTM E1077-14(2021), Standard Test Methods for Estimating the Depth of Decarburization of Steel Specimens (scope), ASTM International, 2021
- ASME Y14.5-2018 (R2024), Dimensioning and Tolerancing, ASME, 2018