Corrosion of dissimilar steel joints... a cracking tale

by Trevor Gooch

Trevor Gooch graduated in industrial metallurgy at the University of Birmingham, and continued to obtain the degree of PhD from the University of London. Dr Gooch joined TWI in 1965. In 1980, Dr Gooch became Head of the Materials Department, embracing the welding characteristics of virtually all metallic materials. In 1994, he was made Materials Technology Manager for TWI, involved with joining characteristics of all materials used for construction.

Resistance of a completed weldment to environmental cracking may be significantly reduced relative to the base metals concerned. The local hardening causes sensitivity to hydrogen-induced cracking, while sensitisation can promote intercrystalline attack.

This paper considers the effects of material composition, welding procedure and PWHT conditions, and the environmental regimes likely to cause a practical problem.

Plant and associated facilities for the oil and gas industry are normally fabricated from a number of different materials, in order that the most cost-effective alloy can be used in any specific region. It is therefore frequently necessary to join carbon or low alloy steels to austenitic stainless steels. Such fusion welds are potentially susceptible to high or low temperature cracking during fabrication, and hence are made using an overalloyed consumable to tolerate dilution from both components of the joint without forming crack-sensitive microstructures. Either austenitic stainless steel or nickel-base fillers may be used, ^[1] appropriate types being readily available for all common fusion welding processes.

However, use of non-matching consumables means that a compositional gradient exists across the fusion boundary, and this leads to the local development of a range of microstructures which can have a profound effect on the behaviour of the completed weldment in corrosive service. This article reviews the situation to indicate the conditions under which a practical problem is likely to arise. Attention is paid to the effects of material composition, of welding procedure, and of conditions used for postweld heat treatment (PWHT) operations.

Fusion Boundary Microstructure

Fig.1 Schaeffler diagram, relating material microstructure to composition.

In practice, little concern regarding corrosion behaviour exists for the stainless steel side of the joint. The compositions of base metal, fusion boundary and bulk weld deposit are such that a stable austenite or austenite-ferrite structure is produced. The situation is, however, very different for the fusion boundary adjacent to the ferritic steel.

On Fig.1, points are plotted appropriate to representative C-Mn ferritic steels (point X) and overalloyed 309-type consumables (point Y). The former corresponds to a ferrite+carbide base metal structure, and the latter to a weld metal containing austenite+ferrite. Across the fusion boundary, the compositional gradient between these two points traverses a range of structure as indicated, and it will be noted in particular that a local region of virgin martensite will form ( Fig.2a). In addition, the tie line between the base steel and bulk weld metal analyses traverses a fully austenitic region.

Fig.2 Fusion boundary of austenitic stainless steel deposited on to ferritic steel, showing martensite on the weld metal side:

a) As-welded

b) PWHT

In principle, exactly the same construction can be used when a nickel-base filler such as ENiCrFe-2 is adopted (point Z). The width of the fusion boundary/stagnant zone at the pool periphery is essentially similar with the two consumable types, although possibly narrower with nickel-base fillers because of the generally lower melting point of these materials. The change in composition therefore takes place over much the same distance, so the compositional gradient is far steeper with nickel-base fillers. Nevertheless, exactly the same range of microstructure can arise as when stainless steel consumables are used. ^[1]

It might be expected that a postweld heat treatment operation specified for the ferritic component would temper any fusion boundary martensite ( Fig.2b), causing softening, and indeed this will happen for the martensite present in the as-welded condition. However, PWHT has three particular consequences, ^[2] viz:

1. Carbon migration will take place from the ferritic steel to the more alloyed weld metal ( Fig.3), and remarkably high carbon contents can be developed locally, even up to 1%, after extended exposure to elevated temperature ( Fig.4). This gives a high carbide density which will directly cause hardening. Carbon migration further removes chromium from the matrix, destabilising the austenite and raising the martensite start transformation temperature (Ms); in consequence, an area close to the fusion boundary will transform to virgin martensite on cool-out from the PWHT temperature.
   
   
2. The presence of nickel in stainless steel has a marked effect in depressing the ferrite-austenite transformation temperature (Ac _l). ^[1] Thus, in view of the compositional gradient, some regions of the fusion boundary will have an Ac _l point below the temperature used for heat treatment; they will therefore transform to austenite during PWHT, but will then reform martensite on coolout.
   
   
3. Carbon migration into the weld metal will lead to precipitation of chromium-rich carbides. With an austenite-ferrite structure, this develops primarily along phase boundaries, with only a minor effect on corrosion resistance, but the fully austenitic areas produced in high dilution deposits can display marked intergranular precipitation.

Fig.3 Secondary electron image of stainless steel weld metal/ferritic steel fusion boundary, showing carbon migration (appearing light-coloured) after PWHT.

Fig.4 X-ray line scans across stainless steel weld metal/ferritic steel fusion boundary after PWHT, showing peak in carbon content on the stainless steel side.

Hence joints between ferritic and austenitic steels will inevitably have a hard region from martensite and/or carbide formation at the fusion boundary, whether the joint is in the as-welded or the heat treated condition ( Fig.5). In addition, extensive fully austenitic areas can result from excessive weld metal dilution by base steel. These local microstructural variations have a major effect on the behaviour of the weldment in corrosive media.

Fig.5 Microhardness traverses across stainless steel/ferritic steel fusion boundaries:

a) As-welded

b) PWHT

To some degree, the specific welding conditions used will influence the fusion boundary region over which microstructural changes occur. High dilution runs associated with high current, high travel speed welding ( Table) will tend to have a wider austenitic region. Some reduction in peak as-welded hardness may be obtained by use of preheat to retard cooling and local hardening during welding, ^[3] although any benefit is limited by the inherent high hardenability of the alloyed dilution zone. However, regardless of welding conditions, flow in the molten weld pool is not uniform, especially as filler metal is transferred across the arc, so that the fusion boundary velocity will vary even in a single run. In consequence, marked variation can exist in the compositional gradient around a weld bead, regardless of welding conditions, as illustrated by Fig.6. Thus, for practical purposes, a risk of forming hard zones and fully austenitic regions must be noted in all weldments between ferritic and austenitic steels.

Effect of welding conditions on dilution for 4mm stainless steel MMA electrodes

Fig.6 Compositional gradients at two different fusion boundary locations: nickel-base weld metal/ferritic steel: electron probe microanalysis.

Effect of martensite

Hardened steel microstructures can suffer stress corrosion cracking (SCC) in a wide range of aqueous media. The problem is essentially one of hydrogen embrittlement, the hydrogen being picked up as a result of a cathodic corrosion reaction on the metal surface. The problem is well recognised in the oil and gas industry, since the presence of H ₂S in the service environment greatly increases the risk of cracking, as a consequence of the hydrogen atom combination reaction involved in the cathodic process being poisoned by the presence of a sulphide scale on the metal surface. The result is a tendency to develop particularly high hydrogen contents in the steel when exposed to sour H ₂S media.

Avoidance of hydrogen-induced SCC in specific environmental situations is normally based on control of material hardness, harder structures having significantly greater propensity for failure. The situation is exemplified by the NACE standard MR0175 which stipulates maximum permissible hardness levels for different materials operating under sour conditions. For ordinary ferritic steels, a hardness of 22HRC is cited, equivalent to 250HV. Martensite hardness decreases at lower carbon contents, but, even at say 0.01%C ( ie below the range found in normal steels and weld metals) the hardness of alloyed material will be around 300HV. Thus, the fusion boundary regions in dissimilar metal welds, will be appreciably harder than the NACE level, and welds between ferritic and austenitic steels exposed to sour media are highly likely to crack along the fusion boundary in association with the hard zone. ^[3-5]

In sour media, there will be a dependence of cracking risk on H ₂S partial pressure and pH, but these aspects have not been defined for dissimilar metal joints, and it is therefore prudent to regard the NACE MR0175 sour threshold of 0.0035 bar H ₂S as limiting. At the same time, this form of failure ( Figs.7 and 8) is not specific to sour conditions, but can occur in any environment promoting hydrogen ingress. Cracking may be encountered in sweet aqueous media, and under conditions of cathodic polarisation, although, in the writer's experience, such failures are rare, possibly occurring only under conditions where applied stress levels are particularly high.

Fig.7 Environmental cracking along the hard zone of a dissimilar metal weld.

Fig.8 Hydrogen-induced fusion boundary cracking: stainless steel/ferritic steel.

Cracking can arise also under gaseous conditions involving a high hydrogen partial pressure. Pick-up of hydrogen by steel is increased at higher temperatures, and, although hydrogen embrittlement per se may be negligible at operating temperature, the absorbed hydrogen can remain in the material if plant shutdown is sufficiently rapid. This then causes cracking along dissimilar metal weld interfaces at temperatures close to normal ambient. The effect has, perhaps, been most commonly encountered as 'disbonding' of weld deposited cladding in hydrocrackers/desulphurisers, ^[6] but can develop also in load-bearing structural joints.

From Fig.1 , the compositional gradient across the interface is steeper with nickel-base consumables than stainless steel fillers, so that the martensitic zone is narrower. Given variation in pool flow, it is possible with nickel-base consumables that locally martensite formation is negligible, and this means that a continuous path for hydrogen-induced SCC may not exist down the interface. Certainly, the resistance to hydrogen-induced environmental cracking at joints made using nickel-base fillers is higher than when stainless steels are used. Nevertheless, some risk of cracking must still be anticipated.

Clearly, when dissimilar metal joints are used in a fabrication, study of the service environment must be undertaken in terms of the likelihood of hydrogen ingress. If exposure to sour media is anticipated, preventative measures are essential. It may be possible to coat the weld area with an impervious paint system, avoiding contact with the service medium. However, coatings are unlikely to be completely reliable, and there is no doubt that a better approach is to design the fabrication such that dissimilar metal welds are not in a critical environmental situation.

Since fusion boundary hardening will persist after PWHT, the possibility of cracking must be recognised even in heat treated weldments. In principle, some advantage might be obtained from two stage heat treatment, in which the second cycle is carried out at an appreciably lower temperature than the first. The aim is to temper any previously formed martensite, while minimising the formation of fresh martensite. This approach has not been fully explored to determine the practical benefit, but it is unlikely to represent a complete panacea.

Fig.9 Intergranular cracking in a high dilution, fully austenitic region close to the fusion boundary: stainless steel weld metal/ferritic steel.

Fusion boundary austenite

Under conditions of continuous immersion in aqueous media at around normal ambient temperature, the problem of local weld metal sensitisation is frequently not recognised since galvanic action means that attack tends to take place preferentially on the ferritic parent steel and the weld metal is effectively cathodically protected. However, weld metal sensitisation can be significant either in conditions of intermittent wetting and drying as may be encountered in deadlegs of plant, or at higher temperatures in pressurised plant. In the latter case, electrochemical potential differences between steels of different alloy content may be less marked than at room temperature, while the ferritic steel may be effectively protected by a magnetite film. Once attack is initiated at the metal surface, it can be expected that local acidification will act to keep the sensitised grain boundaries active, leading to continuing penetration of the metal close to the fusion boundary.

A major factor in determining the occurrence of attack is the formation of a fully austenitic region close to the fusion boundary. As a weld metal, this tends to have a fairly coarse grain size, which will exacerbate the effect of carbide precipitation. If a reasonable ferrite level is obtained, the carbides form along ferrite-austenite boundaries, reducing the local precipitate density and associated chromium depletion. Avoidance of the problem thus depends greatly on the consumable composition and the level of dilution by the ferritic steel component. Fillers with a high chromium content are to be preferred for both stainless steel and nickel types, provided that, in the former case, the PWHT cycle does not induce excessive intermetallic formation and embrittlement.

However, using stainless steel consumables, the real key is to achieve an adequate ferrite content. This means that the filler composition must give tolerance to dilution, and hence the 309 type (23Cr/12Ni) is most commonly used, as indicated in Fig.1 . In addition, welding conditions must be carefully controlled to preclude excessive dilution. This may be difficult with manual welding, and correct operator training is required. High current levels should be avoided, especially in conjunction with high travel speeds, since, as illustrated by the results in the Table , dilutions approaching 50% can be induced. In this regard, it is good practice to butter the faying surface of the ferritic steel component prior to joint completion, with a reasonable degree of bead overlap, say 30-50%, to avoid excessive dilution. Further, the duration and peak temperature of the PWHT operation should be minimised as far as codes permit to limit the extent of the damaging carbon migration and carbide precipitation.

Again, if postweld heat treated dissimilar metal joints are used in a fabrication, close scrutiny of the expected environment is essential. Operation under continuous wetting at temperatures near normal ambient may be satisfactory, but joints in a deadleg, for example, should be positioned to be above the dew point during operation. For wetted service at temperatures above, say, 100°C, the welding procedure, consumable selection and PWHT cycle must be carefully controlled.

Fig.10 Intergranular cracking in fully austenitic weld metal joining HK40 austenitic steel to a ferritic steel.

A similar situation of sensitisation of high carbon austenitic weld metal can occur also if cast creep-resisting austenitic steels such as HK40 are welded to ferritic steel. In this case, carbon pickup by dilution from the austenitic steel can be sufficiently marked that environmental cracking arises in the as-welded condition at multipass welds which have experienced repeated weld thermal cycles ( Fig.10).

Concluding remarks

References