# Decarburization

Surface carbon loss in steel: complete and partial layers, measurement endpoints, process control and finished-part acceptance.

Reviewed: 2026-11-02

Tags: production, heat-treatment, surface-integrity

The worked blank is **20.60 mm diameter after scale removal**, with **0.18 mm total decarburization depth** measured radially from that surface. Concentric finishing to **20.00 mm** removes **0.30 mm radially**, exceeding the depth by **0.12 mm**. Finishing to **20.30 mm** removes **0.15 mm**, potentially leaving **0.03 mm**. Verify minimum local removal, sampling and specified surface properties before accepting the part.

Total decarburization includes complete and partial carbon depletion. Measure perpendicular to the surface and state whether the endpoint uses structure, hardness or carbon content. [ISO 4885:2026, §3.9.2–3.9.3](https://www.iso.org/obp/ui?_escaped_fragment_=iso%3Astd%3Aiso%3A4885%3Aed-4%3Av1%3Aen)

State whether the limit applies to incoming stock, the heat-treated surface or the finished part.

## Carbon loss and oxide scale are different

Carbon depletion and oxidation may occur together, but they are different observations. ISO 4885 defines oxidation as reaction of oxygen with iron and oxide-forming alloying elements; scale is a surface layer formed during heat treatment in a non-protective atmosphere, usually oxide. Removing visible scale does not itself demonstrate that the underlying carbon-depleted zone has been removed. Conversely, appearance alone cannot establish a decarburization depth. [ISO 4885:2026, §3.9.11 and §3.9.13](https://www.iso.org/obp/ui?_escaped_fragment_=iso%3Astd%3Aiso%3A4885%3Aed-4%3Av1%3Aen)

Treat the steel grade, incoming surface, complete thermal history and furnace environment as part of the investigation. A later operation may expose or remove an earlier affected surface. Attribution requires evidence about those operations rather than assuming that the final hardening step caused every observed layer.

## Why the affected layer matters

A carbon-depleted surface can respond differently from the interior during hardening. Uddeholm's *Polishing of Mould Steel*, **edition 7, March 2021**, specifically identifies decarburization or recarburization during heat treatment as a source of hardness variation and polishing difficulty. Its guidance recommends protective heat-treatment routes for its tool steels. This is producer guidance, not a guarantee that any furnace labeled “vacuum” or “protective atmosphere” delivers an acceptable surface under every load and cycle. [Uddeholm polishing guide, heat-treatment section](https://www.uddeholm.com/app/uploads/sites/247/2024/09/Tech-Uddeholm-Polishing-EN.pdf)

For a finished part, connect the finding to the required surface function: specified hardness, wear behavior, contact surface, fatigue-sensitive feature or polishability. Do not infer a universal service-life penalty from one depth value. The importance depends on the material, final geometry and application requirements.

A low surface-hardness result is a reason to investigate, not a unique diagnosis of carbon loss. Check the actual test method and section evidence before assigning a cause or authorizing reprocessing. A correct core hardness likewise does not establish that the surface meets its own requirements.

## Define the depth and method

**ISO 3887:2023**, fourth edition published **September 2023**, defines decarburization and specifies three methods for determining depth in steel products. It replaced the 2017 edition. A laboratory instruction should identify the adopted edition, method, endpoint and reporting basis rather than merely request “a decarb test.” The public ISO scope establishes the standard's subject, not the full sampling or test procedure. [ISO 3887:2023](https://www.iso.org/standard/86689.html)

Complete and partial layers should not be silently interchanged. State whether the limit concerns complete, total or another expressly defined depth. Also state whether a reported result is a maximum at selected locations, an average or a local measurement. Those statistics can differ even on the same section.

Buehler's fastener-preparation note illustrates cross-sectional assessment around thread regions and explains that contrast depends on steel type and heat-treated structure. Its examples include internal oxidation and ferritic regions alongside the bulk microstructure. These examples support careful specimen preparation and location-specific interpretation; they do not establish acceptance limits for an unrelated shaft, spring or tool. [Buehler, metallographic preparation of fasteners](https://www.buehler.com/assets/solutions/technotes/Metallographic-preparation-of-Fasteners-Microscopic-assessment-and-analysis.pdf)

Preserve the edge during preparation and identify the original metal surface. Document the section orientation and inspected locations, especially near corners, threads and changes in section. Agree how rounded edges, scale, prior grinding and local surface damage affect the reference. An image-analysis number is meaningful only with that physical reference and the selected endpoint.

## Original machining-allowance example

Assume a hypothetical round steel blank measures **20.60 mm diameter** after the thermal operation and after scale removal. A representative cross-section shows a maximum **total decarburization depth of 0.18 mm**, measured radially inward from that same metal surface. For this arithmetic only, assume the layer is bounded by that depth everywhere around the functional cylinder.

Finishing to **20.00 mm diameter** removes (20.60 − 20.00)/2 = **0.30 mm radially**. Nominal removal exceeds the assumed affected depth by **0.12 mm**. Finishing instead to **20.30 mm** removes only **0.15 mm radially**, leaving a possible **0.03 mm** affected layer under the model.

The diameter reduction cannot be compared directly to a one-sided depth. A 0.30 mm diameter reduction is only 0.15 mm radial removal. The example also assumes concentric removal: eccentric stock, distortion or uneven machining could leave less removal at one location.

This calculation is a geometric screening exercise, not acceptance evidence. Actual disposition needs measurement uncertainty, relevant sampling, minimum local removal, final dimensions and the specified surface properties. A coupon's 0.18 mm result does not establish an everywhere-maximum on every production part. Removing the affected layer also does not automatically repair other heat-treatment problems.

## Process control and disposition

For a practical handoff, identify the grade and condition, thermal operations, surface protection, stock allowance, critical surfaces, applicable depth limit, test method and sampling locations. Preserve furnace/load records and the specimen's relationship to the parts it represents. Compare results at consistent manufacturing stages; a raw-stock limit and a finished-part limit may address different surfaces.

If a result exceeds the agreed limit, determine whether sufficient stock and approved processing remain to remove the affected material while meeting final requirements. Rehardening alone should not be assumed to restore the missing carbon. Any proposed restorative treatment requires its own qualified material/process basis and engineering disposition.

Use the actual purchase or drawing requirements to decide acceptance. Neither a producer's suggested furnace route nor a measurement standard supplies a universal permissible layer. The ISO vocabulary and depth-standard editions are identified above; the Uddeholm date is printed in the PDF. Buehler's cited note has no verified publication date here. Sources were reviewed on **2026-11-02 PST**.

## Related reference

[Quench hardening](../../processes/heat-treatment/quench-hardening.md)