Hydrogen-induced cracking and how to prevent it

The most treacherous damage in a weld is not the flaw you can see the moment the arc stops. It is the crack that opens quietly hours later, while the weld cools — long after the inspector has gone home. That crack is hydrogen-assisted cold cracking (HACC), also called delayed cracking or underbead cracking: a planar discontinuity that forms not during welding but on the way down to room temperature, sometimes a day or two later, and one that usually stays invisible because it never breaks the surface. Once it exceeds the acceptance criteria, it becomes a defect and calls for rejection. In high-strength steel it is the most feared form of weld damage, because it is both hidden and able to grow under stress until it ends in sudden fracture. This article works through the factors behind cold cracking one at a time: where hydrogen enters the weld, how low-hydrogen filler metal must be stored and baked, the logic behind preheat and post-heat (hydrogen bake-out), and why the examination has to wait for a hold time instead of following straight on from the arc. Throughout, the reference point is the applicable code.

1. The four factors behind cold cracking

Hydrogen-assisted cold cracking never comes from a single cause. It needs several conditions present at the same time. The classical metallurgical model lists three necessary factors: diffusible hydrogen, a susceptible microstructure, and tensile stress. Low temperature and time activate those three, and this article treats them as a separate fourth multiplier, because that is the factor which drives examination timing in the field. Think of them as factors in a product: drop any one of them far enough and the crack does not form. The factors are:

Diffusible hydrogen. Hydrogen dissolved in the weld metal and the heat-affected zone (HAZ) in atomic form, free to move through the lattice.
Susceptible (crack-prone) microstructure. Usually hard, brittle martensite formed in the HAZ by rapid cooling.
Tensile stress. Residual stress from weld shrinkage, restraint, and external loads.
Low temperature and time. In practice, hydrogen cracking appears below roughly 200 °C. Above that temperature hydrogen diffuses out easily and toughness is high, so cracking is not expected. The risk window typically runs from about 150 °C down to ambient, and it grows the longer the weld sits there.

When these conditions coincide, hydrogen diffuses to points of stress concentration — grain boundaries, the spaces between martensite laths — embrittles the metal locally, and initiates a microcrack. The entire prevention strategy rests on breaking at least one of those factors: reduce the hydrogen, soften the microstructure, lower the stress, control the temperature. You do not have to control all four. But control none of them and the crack is close to inevitable.

2. Why "cold" and why "delayed"?

Hot cracking (solidification and liquation cracking) occurs at high temperature, while the weld is still between liquid and solid. Cold cracking is the opposite: the metal has solidified and cooled. Hydrogen diffusion and crack growth under stress both take time — which is why the crack shows up minutes, hours, or even a day or two after welding. The classic field experience is a weld that cracks overnight as it cools and is found the next morning. That delay is what dictates examination timing (see Section 8), and it is the one property that sets cold cracking apart from every other discontinuity.

The crack usually starts in the coarse-grained region of the HAZ, underneath the weld bead (underbead) and parallel to the weld axis, or at the root or the weld toe.

But HACC is not only a HAZ phenomenon. As the carbon equivalent of the base metal drops, the critical location shifts from the HAZ to the weld metal: in modern low-carbon, microalloyed steels the steel side behaves well and hydrogen cracking appears instead in the high-strength weld metal — particularly with filler metal in the 690 MPa class and above. In that case the crack runs transverse to the weld, often in a 45° chevron pattern. On jobs that use high-strength filler metal, this is now the dominant damage mode. The fix is to lower the H class of the filler metal and to set preheat and post-heat from the weld metal, not from the base metal alone; the base metal CE on its own will mislead you. The consequence for examination is the scan plan: a plan that scans only perpendicular to the weld will miss a transverse crack, so choose the scan directions accordingly.

Because the crack usually does not break the surface, visual examination (VT) will not find it — which makes it a discontinuity you have to search for with subsurface methods. The primary method for delayed cracking is UT/PAUT (TOFD where suitable), and MT for surface-breaking indications. Do not treat RT here as an equivalent method to fall back on: RT will not reliably find a tight planar crack in the HAZ or weld metal unless the crack happens to be favorably oriented to the beam. It is complementary, not sufficient on its own.

3. First factor: where does the hydrogen come from?

Hydrogen almost always enters the weld as moisture, which dissociates in the arc and enters the molten pool as atomic hydrogen. The main sources:

Moisture in the covering of shielded metal arc electrodes. Basic (low-hydrogen) coverings in particular are hygroscopic and pull moisture out of the air. The time an electrode may stay exposed to the atmosphere is limited by the filler metal class, the ambient relative humidity, and the atmospheric exposure table in the code — typically somewhere between a few hours and 9 hours. For low-hydrogen E70XX classes the common limit is 4 hours; for moisture-resistant "R" coverings it is 9 hours. Track this time in hours in the field records.
Contamination on the joint faces. Oil, grease, paint, protective primer, and rust (iron oxide plus bound water) all carry hydrogen.
Moisture and condensation. Water condensing on cold metal, rain, high relative humidity, damp shielding gas, or damp flux (SAW).
Wire and shielding gas. Damp wire feed, shielding gas drawing in air, oily wire surface.

In practice the biggest hydrogen sources are electrode and flux moisture together with joint cleanliness. That is why the first step in hydrogen control is always surface preparation and filler metal management — a battle won before the arc is even struck.

4. Diffusible hydrogen and how it is measured

Diffusible hydrogen (HD) is the amount of hydrogen that can move through the weld metal and therefore contribute to cracking. It is normally expressed in milliliters per 100 g of deposited weld metal (ml/100 g). Filler metals are classified against this value, and the designation uses the letter H:

H5 — ≤ 5 ml/100 g (very low hydrogen; critical, high-strength work)
H10 — ≤ 10 ml/100 g
H15 — ≤ 15 ml/100 g

This designation belongs to the EN/ISO consumable standards and is not limited to ISO 2560. The H designation also appears in ISO 14341, ISO 17632, ISO 18275, and ISO 26304. The AWS A5 series uses a different designation: H4 / H8 / H16 (≤ 4 / ≤ 8 / ≤ 16 ml/100 g respectively). If you work to AWS codes, look for the equivalent in that designation. The pairings H4 ↔ H5, H8 ↔ H10, and H16 ↔ H15 are approximate, not one-to-one; the limit values and the test conditions differ. The "R" suffix in AWS A5.1 (moisture-resistant covering) also matters in hydrogen management: R electrodes have a markedly longer permitted atmospheric exposure time.

The more hardenable and the thicker the material, the lower the H class you should choose (H5/H4). When you select filler metal, the H value in the manufacturer's classification designation is a parameter that directly governs cracking risk — not a decorative label. Do not confuse two things here: the test (determination) method is in ISO 3690, while classification is defined in the relevant filler metal standards (ISO 2560, ISO 14341, AWS A5.1, AWS A5.5, and so on). ISO 3690 is not a classification standard.

5. Second factor: hard microstructure and carbon equivalent (CE)

Hydrogen alone is not enough. What makes it dangerous is meeting a hard, brittle microstructure. That microstructure usually forms in the HAZ, when the coarse-grained region next to the fusion line transforms to martensite under rapid cooling. The faster the cooling and the higher the hardenability of the steel, the more martensite — and the higher the cracking risk.

The hardening tendency of a steel is predicted by its carbon equivalent (CE / CEV). The common IIW formula:

CE = C + Mn/6 + (Cr+Mo+V)/5 + (Ni+Cu)/15

This formula has a validity range. The IIW CEV is really only meaningful for C-Mn and low-alloy steels with carbon above about 0.18%. In modern low-carbon (C of about 0.12% or less) microalloyed steels it understates the risk; for those steels use Pcm (Ito-Bessyo) or the CET parameter used by EN 1011-2. If you work from EN 1011-2, know how its two methods differ: Method A is CET-based and Method B is CEV-based, and the two do not run on the same input.

As a rough rule, preheat demand rises with CE; for the IIW CEV, preheat becomes a serious question above roughly CE > 0.40. But CE alone is not a threshold. Preheat is set by four inputs together — CE, combined thickness, HD class (hydrogen scale), and heat input — with the degree of restraint also taken into account. In a thin section with H5 filler metal, preheat may be unnecessary at CE 0.45; in a thick section with H15 filler metal it may be required even at CE 0.35. That is why the EN 1011-2 nomograms run on hydrogen scale A–D (HD bands in ml/100 g), combined thickness, and heat input.

Calculate CE from the chemical composition on the material certificate (mill certificate / 3.1 document). In other words, before you make the first weld you can predict the risk from the analysis report and set the preheat in the WPS accordingly. Cooling rate, in turn, is governed by material thickness (combined thickness) and heat input: a thick section pulls the heat into itself, cools the weld quickly, and feeds martensite.

The concrete field output of this chain is HAZ hardness: CE + thickness + heat input → predicted HAZ hardness → hardness limit. For C-Mn steels, the hardness control approach in EN 1011-2 works to a limit on the order of 350 HV10; in wet H2S service (the scope of NACE/ISO 15156) the limit is much lower. AWS D1.1 also gives a HAZ hardness control method in its alternative preheat determination guidance; that method is limited to fillet welds. Take the hardness measurement on the PQR test coupon or on a production test coupon. The limit always depends on the code and the service condition.

6. Third factor: stress and restraint

The force that physically opens the crack is tensile stress. Weld metal shrinks as it cools. If the part is rigidly restrained — thick section, heavy joint, multiple weld runs, strong anchoring — that shrinkage cannot take place freely, and high residual tensile stress builds up in the weld. On top of that, add the external service loads and the stress concentration in the root geometry.

Ways to lower stress in the field: set up the welding sequence and direction plan so shrinkage balances out, avoid excessive restraint, reduce root stress with a suitable groove preparation and bead sequence, and apply PWHT (post-weld heat treatment / stress relief) where required. PWHT does more than relieve stress: it tempers (softens) martensite and helps some of the hydrogen escape — so it touches three factors at once. But PWHT arrives late. The crack may already have formed during cooling, before the part ever reaches PWHT temperature. That is why the primary defense inside the cooling window is preheat, not PWHT.

7. The backbone of prevention: low-hydrogen filler metal, baking and storage, and preheat

This section brings together the three practices that actually prevent cracking in the field.

Low-hydrogen filler metal. For high-strength and hardenable steels, choose basic-covered (low-hydrogen, e.g. E7018 class) electrodes, low-moisture flux, and filler metal in the H5/H10 classes (H4/H8 in AWS). Cellulosic electrodes (HD typically > 30 ml/100 g) are as a rule not used on hardenable or high-strength steels, or under codes that impose a low-hydrogen requirement — not in the root, not in the fill, not in the cap. Cellulosic practice in pipeline welding (API 1104) is not an exception to this: it is a separate application, separately qualified under its own code and managed with strict preheat and uninterrupted hot-pass discipline. In thick-section, hardenable material it is risky even in the root.

Baking and storage — filler metal management. Bake basic-covered electrodes in an oven at the temperature the manufacturer specifies. Typical code values: for AWS A5.1 low-hydrogen covered electrodes, roughly 260–430 °C (500–800 °F), typically 1–2 hours; for AWS A5.5 low-alloy electrodes, roughly 370–430 °C for at least 1 hour. SAW flux (on the order of 300 °C) and FCAW wires follow different regimes. The manufacturer's instruction governs in every case, and the stricter requirement prevails. Keep baked electrodes in a holding oven afterwards: AWS D1.1 requires at least 120 °C (250 °F) for low-hydrogen electrodes taken out of hermetic packaging and not used immediately; typical practice is 120–150 °C. A setting of 100 °C is below that floor and produces a nonconformity.

Electrodes taken out to the job are held in a heated quiver. An electrode that has exceeded the exposure time in the code's atmospheric exposure table may be rebaked only within the limit the manufacturer allows, and only once — not indefinitely. Rebaking a moisture-loaded electrode over and over, in the hope that it will dry out sooner or later, simply destroys the covering. A wet electrode is not rebaked at all; take it out of service and scrap it. Electrodes fresh out of a vacuum pack are likewise used according to the manufacturer's instruction. This discipline looks minor on paper, but it is the cheapest way to break the hydrogen factor.

Preheat and interpass temperature. Preheat does two jobs at once: it slows the cooling rate and so reduces martensite formation (the microstructure factor), and it gives hydrogen the time and the temperature to diffuse out (the hydrogen factor). Set the preheat temperature from the carbon equivalent, combined thickness, hydrogen level, and heat input.

One guiding reference for structural steel is the minimum preheat and interpass temperature tables in AWS D1.1 (the category-based Method A). Alongside that, the code's annex gives alternative preheat determination guidance: the HAZ hardness control method (limited to fillet welds) and the hydrogen control method (Method B). The letter and the status of that annex have changed from edition to edition. Verify from your own copy which annex letter applies in which edition and whether it is normative or informative — do not assume it carries the same status as the mandatory tables. For calculation-based approaches, use EN 1011-2 (preheat guidance for arc welding of ferritic steels), where the same split applies: Method A (CET-based) and Method B (CEV-based).

Preheat is not a matter of feel. Write it numerically in the WPS and verify it in the field with a contact thermometer/thermocouple or a temperature-indicating crayon. Do not confuse two separate requirements here:

The area to be heated. The preheat temperature shall be established over an area extending in all directions from the point of welding by the thickness of the thickest part, but in no case less than 75 mm (3 in.). The 75 mm here is not a measuring distance; it is the minimum width of the area that has to be heated. An inspector who misses that distinction can take a reading outside the heated area and accept a preheat that is not compliant.
The measurement. Measure the temperature preferably on the face opposite the heat source (the unheated face), adjacent to the groove. If you cannot reach the opposite face, remove the heat source, allow the section to equalize (common practice: about 1 minute for every 25 mm of pipe wall thickness or plate thickness), and measure afterwards. Applying the heat from the face opposite the one being welded is the recommended practice, because it assures that the whole section has been heated.

Verify the temperature immediately before striking the arc on each pass. Preheat is not a one-off operation: maintain it throughout welding and between passes. If welding is interrupted and the part falls below the minimum, preheat again. The heat shall be uniform across the section to be welded.

8. The last move on the hydrogen factor: post-heat and delayed examination

Post-heating / hydrogen bake-out. Holding the part at a set temperature as soon as welding finishes, without letting it fall to room temperature, accelerates the escape of diffusible hydrogen from the metal. Effectiveness depends entirely on the product of temperature × time × thickness, so writing a temperature without a duration gives you a recipe nobody can apply. Common practice is to hold at 250–350 °C for at least 1 hour per 25 mm of thickness (typical total minimum 2–4 hours, longer in thick sections). The lower end around 200 °C is only meaningful in thin sections and over long times; in thick sections the hydrogen diffusion coefficient at that temperature is far too low to give any practical benefit. Write the temperature and the time in the WPS numerically.

One condition must not be skipped: do not allow the part to cool below the preheat temperature before post-heating begins. This is not a "usually done this way" preference — it is the condition that makes the operation meaningful. Start post-heating as the uninterrupted continuation of preheat. Post-heating is not PWHT: its purpose is not stress relief or tempering, but hydrogen removal. On critical, thick, high-CE joints, post-heating attacks the first factor directly, so the low-temperature window that follows carries far less risk. It does not remove the temperature factor — the metal reaches ambient sooner or later — but the hydrogen that would have fed a crack in that window is already gone.

Delayed examination. Because cold cracking is delayed, examining a weld the moment it is finished and calling it clean is the most common and the most dangerous mistake: the crack may simply not exist yet. That is why a hold time before NDT is imposed on high-strength and hardenable steels. The minimum 48 hours provision in AWS D1.1 is not a general high-strength rule, and it is not even worded as "run NDT after 48 hours". The code requires that the acceptance criteria be based on final NDT performed at least 48 hours after completion of the weld. The scope is given as quenched and tempered (Q&T) steels — ASTM A514, A517 and A709 Gr. HPS 100W [690W] (depending on the edition, written as "Gr. 100/100W", with A852 also appearing) — and it applies to both statically and cyclically loaded structures. The clause number and the scope list change between editions, so stamp the edition you are using (for example AWS D1.1:2020 / 2025) in the examination plan. For other high-strength steels the same hold time is adopted as good engineering practice, even where the code does not require it. The wait lets hydrogen and stress finish their work at ambient temperature; the MT / UT / PAUT examination performed afterwards shows the real condition. Write the hold time into the job documentation and the WPS / inspection and test plan (ITP), and keep the record.

Q&T steels carry one more limit: preheat and interpass temperature are bounded not only by a minimum but also by the maximum values given by the manufacturer and the code. Excessive preheat and excessive interpass temperature degrade the tempered structure and reduce HAZ toughness and strength.

9. Integrated prevention strategy (decision chain in short)

A workable HACC prevention flow in the field looks like this:

Identify the material. Take the composition from the mill certificate, calculate CE (or CET/Pcm where needed), and determine the combined thickness.
Lower the hydrogen. Clean the joint of oil, rust, paint, and moisture. Select low-hydrogen filler metal in the correct H class. Run the oven–holding oven–quiver chain and track exposure time in hours.
Soften the microstructure and drive out hydrogen. Set preheat and interpass temperature in the WPS from CE, thickness, HD class, and heat input, and verify them in the field before every pass. Keep the heat input inside both the lower and the upper limits of the WPS. On Q&T steels the maximum heat input and the maximum interpass temperature are binding as well.
Manage the stress. Plan the welding sequence and the restraint. Apply post-heating and/or PWHT where required.
Examine at the right time. Observe the hold time the code prescribes (typically a minimum of 48 hours on Q&T steels), then search with MT and/or UT/PAUT, and write the possibility of transverse cracking into the scan plan.
Record it. Document preheat temperatures, filler metal baking logs, hold time, and NDT results in a traceable form. If a nonconformity comes out, raise an NCR.

This chain is a natural extension of the WPS and procedure qualification (see the standards below). No single step is sufficient on its own, but applied together they keep the cracking factors from coinciding.

Related standards

ASME BPVC Section IX — qualification of welding and brazing procedures and personnel performance (WPS/PQR/WPQ). A decrease of more than a specified limit (55 °C / 100 °F) in the qualified preheat is an essential variable (QW-406.1); an increase of more than 55 °C in the maximum interpass temperature is a supplementary essential variable (QW-406.3, which applies when notch toughness testing is required).
EN ISO 15614-1 — specification and qualification of welding procedures for metallic materials by procedure test (WPQR); ranges of qualification including preheat.
EN ISO 15609-1 — welding procedure specification (WPS); preheat, interpass temperature, heat input, filler material.
EN 1011-2 — recommendations for arc welding of ferritic steels; Method A (CET-based) and Method B (CEV-based) approaches to preheat and hydrogen cracking, hydrogen scale A–D.
AWS D1.1 — structural welding code, steel (stamp the edition year); minimum preheat and interpass temperature tables, the annex with alternative preheat determination guidance (HAZ hardness method and hydrogen control method), electrode storage, baking and atmospheric exposure provisions, and the 48-hour delayed NDT provision.
AWS D1.5 — bridge welding code; delayed examination and hold time provisions.
ISO 2560 — covered electrodes for manual metal arc welding of non-alloy and fine grain steels; classification and the H5/H10/H15 hydrogen designation.
AWS A5.1 / AWS A5.5 — carbon steel and low-alloy steel covered electrodes; the optional H4/H8/H16 designation, the "R" moisture-resistant suffix, and baking and storage requirements.
ISO 3690 — determination (test) method for diffusible hydrogen in weld metal; the test basis for classification (the classification itself sits in the filler metal standards).
EN ISO 5817 — quality levels for imperfections in fusion-welded joints; cracks are not acceptable at quality levels B, C, or D alike (with a limited exception for microcracks only).
EN ISO 6520-1 — classification and definition of weld imperfections (including cold/hydrogen cracking).
EN ISO 17640 — ultrasonic testing of welded joints; EN ISO 11666 acceptance levels.
EN ISO 17638 — magnetic particle testing of welded joints; EN ISO 23278 acceptance levels.
EN ISO 17636-1/-2 — radiographic testing of welded joints (film and digital).
ASME BPVC Section V — nondestructive examination methods (on work within the ASME scope).
EN ISO 3834-2/-3/-4 — quality requirements for fusion welding (comprehensive / standard / elementary); documenting filler metal management and preheat control inside the quality system.

Apply acceptance criteria and preheat/PWHT values according to the construction or service code the job is bound to and to the approved WPS. This article is for educational purposes and does not replace the official standard. Always work from the current edition of the applicable code or standard.

From the field

The real enemy of cold crack prevention in the field is haste, plus the habit of "that'll do". Three mistakes come up again and again. First, filler metal management slips. If the 7018s that came out of the oven in the morning are still sitting in an open box on the deck in the afternoon, your low-hydrogen claim for that weld is finished. Use a quiver, track exposure time in hours against the code's table, and do not try to rebake a wet electrode — set it aside. Second, preheat gets accepted because "it feels hot to the hand". Use a temperature crayon or a contact thermometer, preferably on the unheated opposite face and adjacent to the groove; if you cannot reach the opposite face, remove the heat source, wait for the section to equalize (about 1 minute per 25 mm), then read it. And do not mix this up: the 75 mm in the code is not a measuring distance, it is the minimum width of the area that has to be heated — preheat shall be established over an area extending in all directions from the point of welding by the thickness of the thickest part, and never less than 75 mm. Heating only the edge of the groove and leaving the inside of a thick section cold is an invitation to martensite. Third, and most dangerous of all: examining right after welding and issuing a clean report. In high-strength quenched and tempered (Q&T) steel, the delayed crack opens after you leave. Respect the hold time in the code (typically a minimum of 48 hours on Q&T steels) and run the MT or UT after that window. Breaking a single factor is enough to prevent the crack. But only the inspector who knows the material, measures it, records it, and signs off at the right time — not the moment the arc goes out — can say with any confidence which factor was broken.

Take it to the field: To keep the welding standards, the procedure logic, and the field control steps in this article in your pocket — completely offline and free — take a look at the Doawise Welding Eng Guide app.

At DoaWise we carry out welding engineering and inspection services to international standards (ASME Section IX, EN ISO 15614, AWS D1.1) — from WPS/PQR preparation to preheat and PWHT control, and from hydrogen management to delayed NDT examination. We manage cold cracking risk across material, procedure, and timing, and produce recordable, auditable results.