High-Temperature Hydrogen Attack (HTHA) 2026 — API RP 941 Nelson Curve Decoded

HTHA in depth: the mechanism, the Nelson curves, the Tesoro lessons and what inspection can and cannot do

Short answer: high temperature hydrogen attack (HTHA) happens when hydrogen at high temperature and pressure diffuses into carbon or low-alloy steel and reacts with carbides to form methane at grain boundaries. The result is decarburization, fissuring and eventually cracking. API RP 941 (8th edition, 2016, with Addendum 1 in 2020) uses empirical Nelson curves to set the operating limits for each steel. The U.S. Chemical Safety Board warns that HTHA is hard to find by inspection. Prevention through material selection and verified operating conditions comes first. Advanced ultrasonic testing supports that prevention but does not replace it.

This section replaces the generic template material earlier on this page with HTHA-specific engineering content, and corrects several statements that do not apply to this mechanism. Sources are API's public descriptions of RP 941, the CSB's Tesoro investigation and HTHA safety alert, and published industry papers. API RP 941 is copyrighted, so check curve positions and limits in a licensed copy.

Corrections to the earlier sections of this page

• Detection methods. The list above leads with UT thickness for "general thinning", long-range guided wave, and RT "for porosity". HTHA is not a thinning mechanism, and wall-thickness readings do not detect it. Guided wave screens for metal loss and is not used for HTHA. Radiography cannot see the microscopic fissuring of early HTHA. The techniques industry now relies on are volumetric ultrasonic methods (TOFD, PAUT and full matrix capture with the total focusing method), used in combination. They are described below.
• Prevention. Corrosion inhibitors, coatings and cathodic protection do not prevent HTHA. It is driven by hydrogen partial pressure, temperature, time and steel composition. The controls that work are material selection (steels with more chromium and molybdenum, or alloys not susceptible under the operating conditions), staying within verified operating limits, and, for carbon steel, the PWHT condition that RP 941's two carbon steel curves now distinguish.
• "Annual inspection cycles, RBI-driven." There is no standard annual cycle for HTHA. Inspection of susceptible equipment is prioritised by the owner from operating exposure and materials, and RBI is the owner's own engineering programme.
• API 571 section numbers. The 3rd edition of API RP 571 (2020) reorganised and renumbered its mechanisms, so the section numbers listed above do not match the current edition. Two different mechanisms are also shown under the same number. Look up mechanism entries by name in the current edition.
• Atlantis's role. Atlantis does not run API 581 calculations, material-upgrade cost comparisons or RBI recalibration; it supplies the NDE results that the owner's engineers use in that work. Atlantis does not provide RBI, fitness-for-service, or HTHA susceptibility and life assessments. Our role is the NDE, described at the end of this section. The "anonymised customer outcome" paragraph above names no actual project and should not be read as a case study.

The mechanism in plain terms

At elevated temperature, molecular hydrogen in the process stream breaks down at the steel surface, and atomic hydrogen diffuses into the metal. Inside the steel, it reacts with iron carbide (Fe₃C + 2H₂ → 3Fe + CH₄). Methane molecules are too large to diffuse back out, so they collect at grain boundaries and inclusions. Pressure builds in these tiny cavities. The cavities grow, join into fissures and, given time, link into cracks. As carbon is consumed, the steel loses strength and toughness.

Published descriptions of the damage share several features:

• An incubation period. Steel exposed to attack conditions can go through a period with no detectable change in structure before attack begins. No NDE technique can detect this stage. That is one reason HTHA cannot be managed by inspection alone.
• Two forms of damage. Surface decarburization can occur at higher temperatures and lower pressures. Internal decarburization and fissuring are the dangerous form, because they can develop below an apparently sound surface.
• Weld sensitivity. Welds and heat-affected zones, especially in carbon steel that has not been post-weld heat treated, are common sites of attack. This is why RP 941 now separates PWHT and non-PWHT carbon steel.
• Alloying helps. Chromium and molybdenum form more stable carbides that resist the methane reaction. That is why the Nelson curves for Cr-Mo steels sit at higher temperatures and pressures than the curves for carbon steel. The CSB's HTHA alert quotes API RP 571 as saying 300 series stainless steels and 5Cr, 9Cr and 12Cr alloys are not susceptible under conditions normally seen in refinery units.

API RP 941 and the Nelson curves

RP 941, Steels for Hydrogen Service at Elevated Temperatures and Pressures in Petroleum Refineries and Petrochemical Plants, is built on curves first compiled by G.A. Nelson. Each curve plots temperature against hydrogen partial pressure, and steel operated below its curve is expected to resist HTHA. The curves are empirical. As the CSB describes them, they are drawn beneath the operating points of reported HTHA occurrences, using data voluntarily submitted to API.

Some points about the editions:

• 8th edition, February 2016. This edition introduced two carbon steel curves, one for PWHT steel and one for non-PWHT steel. The CSB notes that the 2016 edition reported 13 new failures below the earlier carbon steel curve.
• Errata and Addendum 1 (2020). Standards distributors list RP 941:2016 with Addendum 1 (2020) as the current version. Confirm this against API's catalogue before citing it in a programme document.
• C-0.5Mo. Industry experience with carbon-half-moly steel led API, over successive editions, to stop giving it its own protective curve. Many older units still contain C-0.5Mo equipment that was originally justified against curves no longer recognised. Review these units first.
• What the curves do not tell you. A Nelson curve is a go/no-go line. It does not show how long it takes damage to start above the line, or how much damage exists after an excursion. Industry papers describe work on a revision with time-based curves, and on fitness-for-service methods for HTHA. Until those are published, treat any time-based assessment as engineering work, not a code rule.
• Inspection guidance has moved. The 8th edition contained an annex on inspection. Industry papers report that HTHA inspection guidance was moved into a new API RP 586 section on NDE for HTHA, and that the next RP 941 edition is expected to reference it instead. Check API's catalogue for the publication status of RP 586 before citing it.

Have carbon steel equipment in hydrogen service that your engineers have flagged for HTHA examination? Send the list for a quote.

What the Tesoro Anacortes investigation changed

On April 2, 2010, a carbon steel heat exchanger about forty years old ruptured in the naphtha hydrotreater at the Tesoro refinery in Anacortes, Washington. The explosion and fire killed seven workers. The CSB investigation found that HTHA had severely cracked and weakened the exchanger. The CSB also found that the damage occurred in the region of the carbon steel Nelson curve considered "safe", where RP 941 did not predict HTHA.

The CSB's recommendation to API asked for minimum mandatory requirements, a requirement to use inherently safer materials, verification of actual operating conditions, and a ban on carbon steel above 400 °F and 50 psia hydrogen partial pressure. In July 2016 the CSB judged that the 8th edition did not meet its recommendation and closed it as "Closed – Unacceptable Action". In its HTHA safety alert, the CSB gave owners four steps:

1. identify all carbon steel equipment in hydrogen service whose failure could harm workers or communities,
2. verify actual operating conditions for that equipment, meaning hydrogen partial pressure and temperature,
3. replace carbon steel equipment that operates above 400 °F and above 50 psia hydrogen partial pressure,
4. use inherently safer materials, such as steels with more chromium and molybdenum.

The CSB's guidance is not a regulation. Owners decide how to apply it under their own process safety management programmes. But it changes how inspection should be framed. The CSB concluded that inspection should not be relied on to identify and control HTHA, because successful detection depends heavily on the techniques used and the examiner's skill, and it found few inspectors with that level of expertise. Inspection is the check on a materials and operating-window programme, not a substitute for one.

Verifying operating conditions: the step most programmes underestimate

The Nelson curve check is only as good as its inputs. In practice, owners rebuild the exposure history for each piece of equipment:

• Hydrogen partial pressure. Calculate it from the actual gas composition at that location, not from design data for the whole unit.
• Metal temperature. Use metal temperature, not process temperature, including hot spots, bypass operation and end-of-run conditions.
• Transients. Startup, shutdown, catalyst regeneration and upsets can move equipment above the curve for short periods.
• PWHT records. These are needed for every weld, including repairs and nozzle additions made after the original construction.
• Material verification. PMI confirms that what is installed matches the materials of construction on record.

This work belongs to the owner's materials, process and integrity engineers. Its output is the list of equipment that needs replacement, closer operating limits or inspection, and that list is where an NDE scope starts.

HTHA inspection techniques and their limits

A 2012 joint industry project tested out-of-service samples with metallurgically confirmed HTHA. Published accounts report that the older ultrasonic methods (backscatter, velocity ratio and attenuation) gave very poor detection, and that samples cleared by those methods were later found damaged when sectioned. Industry then moved toward the techniques in the table below, used together and by examiners trained on real HTHA samples.

The techniques share some practical limits. None of them detects the incubation stage. Results depend heavily on the procedure and the individual examiner. On-stream examination of hot equipment is limited by the temperature ratings of probes and couplant. Encoded data should be kept so the next examination can be compared with this one. When the owner's engineers need sized damage for an engineering assessment, the NDE report should state the technique, coverage, sizing approach and limitations, and leave the interpretation to the engineers.

Planning a turnaround examination of non-PWHT carbon steel welds in hydrogen service? Ask for a technique and coverage proposal.

Regulatory context in the US and Canada

In the United States, HTHA-susceptible equipment in covered processes falls under OSHA's process safety management mechanical integrity requirements (29 CFR 1910.119(j)) and EPA's Risk Management Program. Under these rules, inspection and testing follow recognised and generally accepted good engineering practice. Owners commonly treat API RP 941, API RP 571 and the inspection codes as part of that practice, and the CSB's findings are a widely cited reference. In Canada, pressure equipment is regulated provincially, for example by ABSA in Alberta and TSSA in Ontario. Confirm with your jurisdiction how HTHA assessments and inspection results must be documented for registered equipment.

How Atlantis supports this

Atlantis performs advanced ultrasonic examination on equipment the owner's engineers have identified as HTHA-susceptible. As described on our phased array services page, that means encoded PAUT combined with TOFD and TFM. ASNT-certified technicians do the work under ASNT Level III procedure review, and we deliver encoded data with stated coverage and limitations. Results go to the owner's API 510 or API 570 inspector of record and to the engineers making the materials and fitness-for-service decisions. We do not perform HTHA susceptibility analysis, Nelson curve assessments, RBI or fitness-for-service, and we do not suggest that inspection alone can rule out HTHA. Owners building in-house UT capability can see our ASNT Level III training. Further reading: HTHA mitigation strategies and API 571 damage mechanisms.

Request a scope for encoded PAUT and TOFD on hydrogen service equipment.

HTHA questions engineers ask

What is high temperature hydrogen attack?

A damage mechanism in which hydrogen diffuses into hot carbon or low-alloy steel, reacts with carbides to form methane at grain boundaries, and causes decarburization, fissuring and cracking.

What is the current edition of API RP 941?

The 8th edition (2016), with errata and Addendum 1 (2020), according to standards distributors. Confirm with API's catalogue before citing it.

What is a Nelson curve?

An empirical temperature versus hydrogen partial pressure line for each steel type in RP 941, drawn beneath reported HTHA cases. Operation below the curve is expected to avoid HTHA.

Can HTHA occur below the Nelson curve?

Yes. The CSB found that the Tesoro exchanger failed in the "safe" region of the carbon steel curve, and the 2016 edition reported 13 failures below the earlier carbon steel curve.

What did the CSB recommend for carbon steel in hydrogen service?

Identify the equipment, verify actual hydrogen partial pressure and temperature, replace carbon steel operating above 400 °F and 50 psia hydrogen partial pressure, and use inherently safer Cr-Mo steels.

Which NDE methods detect HTHA?

Volumetric ultrasonic techniques used in combination: TOFD for screening, then PAUT and FMC/TFM for characterisation, carried out by examiners trained on HTHA. Thickness gauging, guided wave and RT do not detect early HTHA.

Can inspection prove equipment is free of HTHA?

No. The incubation stage cannot be detected, and the CSB advises against relying on inspection alone. Inspection supports a programme based on materials and verified operating conditions.

Does PWHT matter for HTHA?

For carbon steel, yes. RP 941's 8th edition gives separate curves for PWHT and non-PWHT carbon steel, so weld PWHT records are part of the assessment.