# Hydrogen embrittlement

Hydrogen-related loss of mechanical performance: susceptibility, exposure, stress, delayed failure and the limits of process and material tests.

Reviewed: 2026-11-02

Tags: production, materials, hydrogen, failure-analysis

In the illustrative comparison, smooth specimens with **100 mm²** original gauge area reach **70 kN** without the selected hydrogen exposure and **49 kN** with it, or **700 and 490 MPa** nominal peak stresses. The **30%** difference is relative to this comparison. It gives neither a threshold for delayed failure nor allowable fastener stress. Test controls, loading history and the actual notch or thread-root geometry remain important.

WorldAutoSteel identifies susceptible material, available hydrogen and stress as conditions for delayed cracking in advanced high-strength steels. [WorldAutoSteel AHSS Guidelines, Delayed Cracking](https://ahssinsights.org/forming/testing-characterization-forming/delayed-cracking/)

Apply controls and tests to the actual steel, manufacturing history and service environment.

## Material, exposure and loading history

| Boundary | Information needed |
| --- | --- |
| Material susceptibility | Grade, strength/hardness, heat treatment, microstructure and manufacturing history |
| Hydrogen opportunity | Actual surface preparation, coating, joining or service environment |
| Stress | Assembly preload, applied loads, residual stress and local geometry |
| Time and temperature | Exposure, processing intervals, loading duration and operating conditions |
| Observation | Crack location, fracture condition, deformation and test/inspection history |

WorldAutoSteel notes that delayed cracking can appear after forming rather than during the stamping operation. Its discussion is specific to AHSS grades and processing conditions. An illustrative susceptibility range is not a universally safe hardness or strength boundary. [WorldAutoSteel, conditions and delayed cracking](https://ahssinsights.org/forming/testing-characterization-forming/delayed-cracking/)

Hydrogen introduced during manufacture and hydrogen exposure during service depend on the actual sequence and environment. A coating that changes corrosion behavior can also change the manufacturing route or hydrogen exposure; its appearance or corrosion rating does not establish freedom from HE.

## Delayed failure requires more than a visual check

NASA's historical *Fastener Design Manual* explicitly discusses internal hydrogen-related failures occurring after proof testing without an external indication of the hydrogen. That supports distinguishing an immediate proof-load result from later performance. The manual's **March 1990** processing examples are historical guidance, not current bake instructions for an arbitrary coating or fastener. [Barrett, NASA RP-1228, hydrogen-embrittlement section](https://ntrs.nasa.gov/api/citations/19900009424/downloads/19900009424.pdf)

A delayed crack or brittle-looking fracture is not a unique diagnosis. Overload, fatigue, material defects, corrosion or another mechanism may also explain the evidence. A photograph alone cannot establish HE as the cause; the available material, process and loading evidence limits the diagnosis.

ISO/TR 20491:2019 is a **technical report** on steel-fastener HE fundamentals. Its introduction distinguishes a mechanism of failure from a root cause and describes the involvement of the full fastener supply chain. Identifying hydrogen-assisted cracking still leaves the source, susceptibility, stress and control failure to explain. [ISO/TR 20491:2019 public introduction and scope](https://www.iso.org/obp/ui?_escaped_fragment_=iso%3Astd%3Aiso%3Atr%3A20491%3Aed-1%3Av1%3Aen)

## Test scope and interpretation

**ASTM F519-23** covers mechanical evaluation of plating/coating processes and relevant service-chemical exposures. Its public scope explicitly says it is **not** intended to rank the relative susceptibility of different steels. A process qualification using specified specimens therefore is not automatically a material-ranking experiment or proof that every finished component will perform identically. [ASTM F519-23, scope and significance](https://store.astm.org/f0519-23.html)

**ASTM F1624-12(2024)** addresses measurement of a hydrogen-embrittlement threshold in steel using incremental step loading. Interpretation depends on the specimen, material condition, environment, load sequence and definition of the reported threshold. An ordinary tensile peak load or an arbitrary short hold is not that standardized threshold. [ASTM F1624 public title and edition](https://store.astm.org/standards/f1624)

WorldAutoSteel also cautions that cracking in a simulative test does not itself demonstrate that parts will crack in an automotive environment: the required conditions may differ. Conversely, a favorable result in one exposure does not cover an untested environment or loading history. [WorldAutoSteel, test applicability](https://ahssinsights.org/forming/testing-characterization-forming/delayed-cracking/)

## Original example: a reported strength ratio is not a threshold

Assume two invented smooth tensile specimens, each with **100 mm² original gauge area**. One comparison test reaches **70 kN** peak load without the selected hydrogen exposure; the other reaches **49 kN** after that exposure. Ignore measurement uncertainty for this arithmetic only.

| Illustrative result | Calculation | Nominal peak engineering stress |
| --- | --- | --- |
| Comparison specimen | 70,000 N / 100 mm² | 700 MPa |
| Exposed specimen | 49,000 N / 100 mm² | 490 MPa |

The peak-load ratio is **49/70 = 0.70**, or a **30% reduction** relative to this comparison. That calculation does not establish an ASTM F1624 threshold, a guaranteed service reduction or proof that hydrogen alone caused the difference. Test control, replicates, material variability, exposure, strain rate and the chosen response metric remain essential.

A nominal area calculation also omits thread-root or notch stress concentration, residual stress and assembly load distribution. The 490 MPa result is not an allowable stress for a fastener. Neither one surviving specimen nor a mean result supplies a minimum guaranteed property for a production lot.

## Manufacturing and service controls need their own basis

The specified surface-preparation and coating sequence, process controls, material condition and any required hydrogen-relief treatment affect the process basis. Where a bake is required, its timing, temperature, duration and verification depend on the governing qualified specification. A general reference does not establish a universal bake, and reheating does not necessarily repair an already cracked component.

Process control and final-part evidence are related but different. A test's applicability depends on which lot or operation it represents, deviations and the conditions actually tested. Changing coating, cleaning, heat treatment or subcontract processing can change the question the earlier qualification answered.

For gaseous-hydrogen service, NASA's **2016** technical memorandum discusses dependence on pressure, temperature, gas purity and material condition, including surface finish and heat treatment. Compatibility data for a given service environment is distinct from a plating-process coupon assessment. [Lee and Woods, NASA/TM-2016-218602, publication record](https://ntrs.nasa.gov/citations/20160005654)

## Source versions and coverage

Reviewed **2026-11-02 PST**. ISO identifies **TR 20491:2019, edition 1, February 2019**. ASTM identifies **F519-23** and **F1624-12(2024)** on its public pages; those scopes do not reproduce the full testing procedures. NASA dates RP-1228 **March 1990** and TM-2016-218602 **April 1, 2016**. No separate issue date is assigned to the retrieved WorldAutoSteel page. The numerical example is original and bounded.

## Related reference

[Manufacturing process references](../../processes/index.md)