Avoiding Hydrogen Embrittlement Stress Cracking of Ferritic Austenitic Stainless Steels Under Cathodic Protection
P Woollin and A Gregori
Proceedings of OMAE 2004: 23rd International Conference on Offshore Mechanics and Arctic Engineering Vancouver, Canada 20-25 June 2004
The paper presents the results of a programme designed to define the material, stress and environmental factors controlling sensitivity of ferritic-austenitic stainless steels to hydrogen embrittlement stress cracking when exposed to cathodic protection. Factors examined in small and large-scale tests include microstructural coarseness, phase balance and hardness of a range of parent steels and welds. The results are presented in terms of threshold strain and normalised stress to develop hydrogen embrittlement stress cracks. The effects of microstructure and applied potential on crack initiation and propagation are described. Recommendations are made with respect to the strain/normalised stress levels for ferritic-austenitic steels under cathodic protection.
Duplex and superduplex stainless steels are attractive materials for use in subsea oil and gas developments due to their combination of pitting and stress corrosion resistance and high strength. In such applications, they are typically subject to cathodic polarisation either unintentionally due to contact with carbon steel structures under cathodic protection or intentionally due to the potential for pitting and crevice corrosion when in direct contact with seawater whilst containing hot produced fluids.
Several recent failures of duplex and superduplex components subsea have been attributed to hydrogen embrittlement stress cracking (HESC) (also known as hydrogen induced stress cracking (HISC)) arising from the application of cathodic polarisation and stress in an unacceptable combination for the microstructure in question. [1,2] A programme of testing was undertaken to define quantitatively the material, stress and environmental factors inducing cracking in duplex and superduplex stainless steel fabrications when cathodically polarised in seawater.
A range of 22%Cr duplex and 25%Cr superduplex stainless steels including parent pipe, plate, forgings and bars, welds (cross-weld and all weld metal specimens) and simulated heat affected zones (HAZs), Table 1 and Fig.1, were selected for constant load tensile testing under cathodic protection, to examine HESC behaviour. Simulated HAZs were produced by furnace heating at 1350°C followed by water quenching and Gleeble methods.
Table 1 Materials examined
|Material code and type||UNS No.||Hardness (HV5)||Austenite spacing (µm)||Ferrite fraction (%)||0.2% proof stress/UTS (MPa)||Third phases present|
|A: 22%Cr steel pipe
||56 ± 5
|B: Girth weld in 22%Cr steel pipe
|72 ± 6
61 ± 5
|K: 22%Cr steel forging
||46 ± 5
|C: 22%Cr steel pipe heat treated at 1350/1070°C to coarsen
||55 ± 5
|S: Simulated 1350°C HAZ in 22%Cr duplex pipe
||67 ± 4
|D: 0.5" superduplex bar
||53 ± 5
|E: 4.5" superduplex bar
|55 ± 5
|F: 160mm superduplex bar
|58 ± 6
|H: 12mm superduplex plate
||47 ± 3
|J: 25mm superduplex plate
||45 ± 5
|G: Butt weld in 12mm superduplex plate
|57 ± 7
49 ± 4
|I: Simulated 1350°C HAZ in superduplex bar
||61 ± 9
|P: Simulated 1350°C HAZ in superduplex plate
||55 ± 9
|* microstructure has secondary austenite, NA = not applicable, WM = weld metal, T = transverse, L = longitudinal
Fig.1a) 22%Cr duplex (UNS S32205) pipe A
Fig.1b) Superduplex (UNS S32760) bar F
The phase balance of each material was quantified by point counting. The average distance between austenite units ('austenite spacing') was measured by a linear intercept method. This distance does not represent the ferrite grainsize, which typically is substantially larger.
Vickers hardness measurements were made and tensile testing was performed on cylindrical specimens, to obtain conventional tensile data at a strain rate of 1.3 to 2x10-3 s-1 .
Small Scale Hydrogen Embrittlement Stress Cracking Tests
Cylindrical tensile specimens with a gauge length of 20mm and gauge diameter of 3.75mm were machined from each of the materials. Most specimens were pre-charged without applied stress, for two weeks prior to test, at the potential selected for the subsequent test but a few were tested without pre-charge for comparison.
Testing was under constant load conditions, in dead weight load frames. Strain was measured via a parallel extensometer. Each sample was held under tension in natural seawater at 20±3°C, and polarised by potentiostat. Various samples were tested at a range of stress for each material, to identify the threshold for HESC. Most testing was at an applied potential of -1100mV SCE but tests were also performed at other potentials, i.e. -900, -800 and -700mV SCE . Time to failure was recorded. Metallographic sectioning was performed after testing, to look for any cracks. Hydrogen measurements were made by vacuum hot extraction at 950°C on full-thickness piecescut from the tensile test samples after test.
Full Scale Welded Pipe HESC Tests
Wrought 22%Cr pipe (material A, Table 1) was used for full-scale welded pipe testing in a purpose-built, four-point bend load frame. Pipe samples were 4m long and the pipe had 15mm wt and 168mm OD. Two weld geometries were examined: (i) a pipe with a girth weld at mid-length and (ii) a pipe with a fillet weld to an external patch at mid-length (to simulate an anode attachment pad).
The girth-welded pipes, designated GW1 and GW2, were welded by mechanised TIG employing a superduplex filler wire. The fillet-welded pipe, designated FW1, was welded by manual TIG, employing superduplex wire. The stress concentration factors for the girth and fillet welded pipes were estimated to be about 2.6 and 2.8 respectively. Residual stresses were measured in locations away from the highly stressed test region, prior to testing, using the centre-hole drilling technique in the weld toe/HAZ area of the welded pipes.
The welded pipes were tested in natural flowing seawater at a temperature of 10±3°C with an applied potential of -1100mV SCE . The pipes were strain gauged. The load was increased incrementally in GW1 and FW1 to identify the threshold strain/stress for macroscopic HESC development, Table 2. The pipes were hydrogen pre-charged at -1100mV SCE , without load applied, for seven days. Each loading step was maintained for seven days and the pipe examined for the onset of cracking employing a binocular microscope. After 7 days exposure, the applied load was increased and the procedure repeated until macroscopic cracking initiated. The girth-welded pipe GW2 was held at 0.5% global strain, which was identified as just below the approximate threshold from test GW1, for 188 days (6months). After test, the pipes were sectioned in the highly stressed weld toe region to look for small cracks and to examine the crack morphology. Scanning electron microscopy was performed on the crack surfaces.
Table 2 Loading applied in four-point bend testing of welded pipes in seawater at 10±3°C, at -1100mVSCE .
|Girth-welded pipe GW1||Girth welded pipe GW2||Fillet-welded pipe FW1|
Small Scale Tests
Interpretation of HESC tensile test results
The constant load tensile HESC tests had four possible outcomes: (i) specimen did not crack or fail, (ii) specimen cracked but did not fail in 30 days ( Fig.2), (iii) specimen cracked and failed and (iv) the specimen failed by necking without any cracking. From each series of tests on a specific material, the threshold stresses and strains separating behavior types (i) and(ii), and (ii) and (iii) have been identified, Fig.3. These correspond to threshold conditions for crack initiation and propagation in 30 days in tensile test specimens. Considering results in terms of normalised stress gave good correlation between different materials, whereas simply plotting applied stress gave much greater scatter and no consistent trends. Threshold conditions measured in terms of strain also gave good correlation between the various materials.
Fig.2. Shallow cracks in superduplex bar F, after 30 days at 774MPa (20% strain during test)
Fig.3. Typical results of 30 day HESC dead-weight tensile tests. Example shown is material A tested at -1100mVsce
Correlation of Small Scale HESC data with microstructure
Figure 4 plots the threshold strains for crack initiation and propagation against austenite spacing, for each of the materials. The figure shows well-defined trends for both crack initiation and propagation, which are increasingly difficult as the austenite spacing decreases. The trend is most pronounced for crack propagation in 30 days. All materials initiated cracks at stresses between 87 and 104% of the 0.2% proof stress when polarised to -1100mVSCE . Propagation required normalised stresses of 92 to 138% of the proof stress.
Fig.4. Strain thresholds plotted against austenite spacing
It was found that when there was a bimodal distribution of austenite (e.g. Fig.1a), the fine, equiaxed austenite particles were apparently ineffective in preventing crack initiation and propagation. Elongated austenite islands were much more effective.
For the range examined, there was no strong relationship between threshold conditions for HESC in the parent steels and the ferrite volume fraction. However, there is, in general, a link between ferrite fraction and austenite spacing in weld metals, where the coarseness of the austenite islands varies over a fairly narrow range. In general for highly ferritic welds, reduced resistance to HESC would be anticipated. 
Also over the range examined, there was no strong relationship between threshold HESC conditions and average hardness. Again, in the general sense, sensitivity to cracking would be expected to increase with a substantial increase in hardness, e.g. due to cold working. Any effects of ferrite content and hardness are sufficiently weak in these tests that they were masked by the strong effect of austenite spacing. No distinction between 22%Cr duplex and 25%superduplex steels was found in terms of susceptibility to HESC.
Figure 5 plots the variation of hydrogen pick-up in 30 days with applied potential. It can be seen that no hydrogen pick-up was noted at -700 and -800mV SCE but hydrogen pick-up increased at increasingly negative potentials beyond -800 mV SCE .
Fig.5. Variation of total hydrogen content after 30 days testing and 14 day pre-charge, with applied potential
Strain development during test
The extensometer measurements indicated that strain developed rapidly at the start of the test and then at a steadily decreasing rate throughout the remainder of the test, Fig.6. Such time dependent strain development is known as 'low temperature creep'. 
Fig.6. Strain development in a tensile specimen due to low temperature creep at 98% of the 0.2% proof stress (steel E)
The work indicates that for smooth tensile specimens, the HESC behaviour of welds depends to some extent on the weld metal type. For welds in 22%Cr pipe made with superduplex filler, cracks initiated in parent steel first and in HAZat slightly higher strain, with crack initiation in the weld metal at a yet higher strain. The higher threshold stress for weld metal presumably reflects its higher strength compared to parent steel. In the one case where a cross-weld specimen failed in 30 days, the main crack developed from the HAZ, presumably as a result of the greater austenite spacing in the HAZ.
For welds in superduplex plate made with superduplex filler, cracks were initiated in weld metal, parent steel and simulated HAZ at similar strains, Fig.7. Four cross-weld specimens failed and, of these, three failed from cracks in the HAZ due to the greater austenite spacing of the HAZ. Crack propagation path was dominated by the direction of applied stress, with the crack path passing from HAZ into parent steel. The remaining specimen failed in the weld metal.
Fig.7. Variation of HESC threshold strains for parent steel, cross-weld, all-weld metal and simulated HAZ specimens
Overall, it is concluded that crack initiation in smooth tensile specimens from welded joints occurred in HAZ, adjacent parent steel, and weld metal at broadly similar stresses but with preferential initiation in the lowest strength region. Preferential crack propagation occurred from the HAZ due to the highest austenite spacing being in that region. It may be noted that the weld metals examined had fairly low ferrite content (38-49%) and that weld metals with higher ferrite contents, e.g. in an unreheated weld cap, might have lower resistance to HESC.
Data showing the effect of applied potential on thresholds for crack initiation and propagation in one material are shown in Fig.8. There was a pronounced trend of increasing threshold for crack initiation with less negative potentials, over the range examined. The threshold stress for crack propagation in 30 days was fairly insensitive but increased slightly with less negative potentials. No cracks were formed in any material at -700mV SCE and a few small cracks were noted at -800mV SCE but only at a stress approaching the UTS.
Fig.8. Effect of applied potential on threshold strains in superduplex bar D
Large Scale HESC Tests on Welded Pipes
Residual Stress Measurements
Maximum circumferential and axial tensile residual stresses of 447MPa and 309MPa were measured in the HAZ of girth welded pipe. The residual stress measurements on fillet-welded pipe showed a maximum axial tensile residual stress of404MPa in the HAZ about 1.5mm from the fusion line.
Full-scale HESC Testing
The global strain during testing, measured away from the weld, was constant and no strain due to low temperature creep was observed. Cracking occurred in GW1 after less than 12 hours at a global strain level of 0.85% corresponding to a nominal stress of 601MPa, calculated from the conventional stress-strain curve (normalised stress = 1.08). No cracking was observed after 7 days exposure at a global strain level of 0.63%, corresponding to a nominal stress level of 584MPa (a normalised stress of 1.05). No cracking was observed on metallographic sections through the HAZ region after 188 days (>6 months) testing of girth welded pipe GW2 at a global strain of 0.5%, corresponding to a nominal stress of 571MPa and a normalised stress of 1.03. Failure in pipe GW1 occurred at the weld toe, at the 12 o'clock position, and cracking propagated through the HAZ into the parent material ( Fig.9).
Fig.9. Crack at 12 o'clock position in girth-welded pipe GW1
In the fillet-welded pipe FW1, cracking occurred after two hours at a global strain level of 0.93%, corresponding to a nominal stress of 604MPa (normalised stress = 1.08). No cracking was observed after 7 days exposure at a strain level of 0.58%, corresponding to a nominal stress level of 579MPa (nominal normalised stress was 1.04). Failure of fillet-welded pipe FW1 occurred at the weld toe and in cap weld metal in the interpass area between two runs, close to the weld toe. Multiple initiation sites were observed near the 12 o'clock positions.
Factors Controlling Sensitivity to HESC in Tensile Tests
The data indicate that the austenite phase spacing is the key microstructural variable controlling HESC in ferritic-austenitic stainless steels. However, crack initiation in tensile specimens was fairly insensitive to austenite spacing and all ferritic-austenitic steels initiated HESC cracks at 87 to 104% of the conventional 0.2% proof stress. It may be noted that the ferrite spacing was broadly similar to the austenite spacing, so it is difficult to separate the effects of the two. Whilst the austenite spacing would be expected to control crack initiation, which is in ferrite, it is conceivable that ferrite spacing may be more significant with respect to crack propagation, as propagation requires the cracks to pass through the austenite phase. However, it has not been identified whether the propagation rate controlling factor is hydrogen charging and crack extension into ferrite, with austenite subsequently failing in a purely mechanical fashion, or whether hydrogen charging and crack extension through the austenite is controlling. Examination of the fracture faces was not conclusive in this respect, although it was noted that there was little evidence of ductile microvoid formation in the austenite. Very fine 'secondary' austenite, present typically as small equiaxed particles, is ineffective at blocking HESC cracks. Any effect of ferrite fraction over the range 40-67% and hardness over the range 240-300HV5 on HESC was masked by the dominant effect of austenite spacing.
Full Scale Pipe Tests
Macroscopic crack initiation and subsequent propagation, occurred at the same applied global strain (0.85-0.96%) in the large-scale samples. This is unexpected from the small-scale tensile test results, as crack propagation did not occur until a strain of around 19% was reached in the same pipe material. This indicates that stress intensity drives crack propagation rather than stress per se. The size of the highly stressed and highly hydrogen charged region is then significant, and it seems possible that the small size of the tensile specimens restricts the length of a surface crack that can develop and hence the stress intensity that can be developed ahead of it. There is much less restriction on surface crack length in the full-scale pipe. Indeed it was observed that crack extension along the surface, which would have highest hydrogen content, was much more rapid than into the thickness, where hydrogen content is lower, during the full-scale pipe tests. However, once surface propagation has occurred sufficiently, the stress intensity became sufficient for propagation into the bulk material, which has lower hydrogen level but may be actively hydrogen charged at the crack-tip.
Mechanism of HESC
The mechanism of HESC includes two discrete steps: initiation and propagation. Initiation occurs fairly rapidly, when the material is undergoing low temperature creep at a relatively high rate. Crack initiation occurs by dynamic plastic straining of hydrogen charged ferrite grains on the surface.
The observations indicate that the critical surface hydrogen content for crack initiation is fairly readily obtained in a few hours when polarised to -1100mV SCE and held at a stress above the threshold. At less negative potentials, crack initiation requires higher strains than at -1100mV SCE . This reflects a lower surface hydrogen concentration at the less negative potentials. The surface hydrogen concentration will reach an equilibrium level rapidly after the onset of charging and hence it is unlikely that the threshold for crack initiation will reduce substantially over longer periods.
Crack propagation requires a critical stress intensity and presumably a critical hydrogen content at the crack-tip, which may take a considerable length of time to develop over the required critical volume of material, e.g. in austenite, especially at sea bed ambient temperature due to low hydrogen diffusivity. Testing for 30 days is apparently inadequate for reliably determining a long-term crack propagation threshold in tensile specimens and 90 or 180 days may be more appropriate. Published diffusion coefficients suggest that several years would be required to get permeation of hydrogen through the full thickness of specimen. Therefore testing for sufficiently long to achieve a steady state hydrogen level is unrealistic and it is possible that the crack propagation threshold will continue to fall until a steady state is reached.
Strain Limitations for Future Cathodically Protected Developments
For material cathodically protected by aluminium alloy anodes, such that potential is around -1000 to -1100mV SCE , the following strain limitations are proposed for materials with ferrite in the range 35-65% and hardness below 300HV5:
- For parent material with austenite spacing ≥30µm, total strain should not exceed 0.5%. It is likely that this strain needs to be present only over a volume of material comparable with the austenite spacing.
- For material with austenite spacing ≤30µm, the total strain limit might be increased, e.g. to 0.75% or even 1% for material with ≤10µm austenite spacing but there is a substantial amount of scatter in data in this range, probably because fine secondary austenite is ineffective at blocking HESC. Hence, adoption of a limiting strain in excess of 0.5% requires caution and material specific testing is recommended.
For less negative potentials, limiting strains may be increased but for coarse material this effect was only significant at potentials less negative than -900mV SCE .
Application of these figures to design requires allowance for residual strain and locked-in strain from fabrication, the effects of stress concentrations and low temperature creep. In practice, this may be somewhat difficult, and for highly stressed components, it might be more appropriate to design such that all loading is displacement controlled. Under displacement control, low temperature creep leads to a reduction in stress and consequently it has been found that propagation of cracks is much more difficult, at least in small-scale specimens.  Nevertheless, plastic straining under CP can lead to crack initiation, on a fine scale, regardless of whether load or displacement control applies. However, it should be noted that although a clear distinction between load and displacement control exists in small scale tests, where the loading frame is much stiffer than the specimen and a small amount of creep can give substantial stress relaxation, the same is not true in large scale tests, where the specimen and test frame have much more similar stiffness and a small amount of creep gives much less, if any stress relaxation due to the elastic deflection of the loading frame. A similar situation may exist in practice and the distinction between load and displacement control in real structures is not a simple one.
Qualification of Materials and Designs
Qualification of ferritic-austenitic material for service under cathodic protection cannot follow practices developed for example for sour service, as ferritic-austenitic materials are inherently susceptible to HESC provided thatthe stress is high enough, whereas for sour service the susceptibility to cracking depends on the environment and the only requirement for the test stress is that it is conservative. If qualification testing is required to demonstrate the suitability for service of a ferritic-austenitic component or structure, then the stressing must be realistic. Due to the difficulties of (i) allowing for residual stress and stress locked-in during fabrication and (ii)understanding whether applied stressing is load-controlled, displacement-controlled or somewhere in between, the most robust approach is to test full-scale components or design details. If small scale testing is to be used, dead-weight tensile loading is recommended unless it can be demonstrated that stressing in practice is fully displacement controlled, in which case previous experience indicates that there is very much less risk of cracking.  Small scale testing is appropriate only if the maximum strains in service are known, such that the appropriate test strain can be defined. It may be appropriate to use finite element (FE) analysis for complex geometries. Any FE analysis will need to be elastic-plastic, using a representative stress-strain relationship.
At a simpler level, dead weight tensile tests could be used to establish whether ferritic-austenitic material has a certain level of resistance to HESC, say equivalent to or better than that expected for material with 20µmaustenite spacing. This would overcome difficulties associated with trying to interpret and quantify microstructures.
Conclusions and recommendations
The following conclusions and recommendations are drawn from the work:
- Ferritic-austenitic stainless steels are susceptible to hydrogen embrittlement stress cracking under cathodic protection. It is recommended that exposure to potentials more negative than about -850mV SCE should be avoided where practicable.
- When exposure of ferritic-austenitic stainless steels to potentials of -900 to -1100mV SCE cannot be avoided, it is recommended that highly stressed components be designed to be under displacement control where possible.
- For ferritic-austenitic stainless steels at potentials of -1000 to -1100mV SCE under load control, it is recommended that total strain be limited to less than 0.5%. This limit may be relaxed for material with austenite spacing less than 30µm, e.g. to 0.75 or 1% strain, or for potential between -850 and -950mV SCE but qualification testing would be required to establish the appropriate strain limit.
- The HESC cracking phenomenon is apparently controlled by plastic strain, with initiation occurring during the phase of most rapid low temperature creep, typically during the first few hours or days.
- Hydrogen embrittlement stress cracking susceptibility showed a strong correlation with austenite spacing but was only weakly influenced by ferrite fraction over the range 40-67% and hardness over the range 240-300HV5. No difference was found between the behaviour of 22%Cr duplex and 25%Cr superduplex grades.
- Cross-weld tensile specimens showed a tendency for HESC crack propagation preferentially from the HAZ, reflecting the higher austenite spacing in the high temperature HAZ. Crack initiation can occur in weld metal, parent steel and HAZ, depending on the proof strength of each region, with initiation being favoured in the lowest strength region.
- For welded ferritic-austenitic stainless steel components with a typical girth or fillet weld geometry, which are cathodically protected at potentials between -900 and -1100mV SCE , it is recommended that total global strain be limited to below 0.5%. A lower global strain limit might be appropriate for materials with coarse austenite spacing, e.g. ≥30µm, or severe weld toe geometries.
- It is recommended that testing to quantify resistance to HESC of ferritic-austenitic stainless steels employs full-scale component testing or small-scale dead-weight tensile testing. It is recommended that qualification testing for service accurately or conservatively reproduces the stressing to be seen in service, taking account of applied, residual and locked-in stress.
The following companies are thanked for sponsoring the work and for granting permission to publish: Amerada Hess, ARCO Exploration and Production Technology, BP International Ltd, Chevron Research and Technology Company, QinetiQ Ltd(formerly DERA), ExxonMobil, Industeel (Arcelor Group), Health and Safety Executive, Marathon Oil Company, Shell UK Exploration and Production, Statoil and Weir Materials and Foundries.
- Woollin P and Murphy W, 'Hydrogen embrittlement stress cracking of superduplex stainless steel', Proc conf Corrosion 2001, NACE International, paper 01018.
- Taylor T S, Pendlington T and Bird R: 'Foinaven super duplex materials cracking investigation'. Proc conf 'Offshore Technology Conference (OTC) 1999', Houston, 1999, paper OTC 10965, 467-480.
- Gooch TG, Leonard AJ, Gunn RN, 'Hydrogen cracking of ferritic-austenitic stainless steel weld metal', Proc.conf 'Duplex America 2000', 345-357.
- Boniardi M, La Vecchia G M, Roberti R, 'Stress relaxation behaviour in duplex and superduplex stainless steels', Proc conf 'Duplex Stainless Steels 94', Glasgow, TWI, paper 110.