Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Hydrogen Embrittlement Cracking of Superduplex Steel


Hydrogen Embrittlement Stress Corrosion Cracking of Superduplex Stainless Steel

P Woollin
Granta Park, Great Abington
Cambridge, CB1 6AL, UK


W Murphy
BP Exploration
Farburn Industrial Estate,
Dyce, Aberdeen, AB21 7PB,UK

Paper 01018 presented at Corrosion 2001, Houston, TX, USA, 12 - 16 March 2001


Two superduplex stainless steel hubs on a subsea manifold subject to cathodic protection failed as a result of hydrogen embrittlement stress corrosion cracking. A series of tests was performed to establish the threshold condition for cracking. The tests included: (i) constant-load smooth bend tests, (ii) constant-deflection smooth bend tests, (iii) constant-load pre-cracked bend tests, (iv) interrupted slow strain rate tensile tests, (v) constant-load tensile tests, (vi) approximately constant-strain tensile tests and (vii) full-scale hub tests. The testing identified a very marked difference in material response under load-control and displacement-control. Under constant-load conditions, tensile testing indicated a threshold stress for crack initiation and propagation in the hub material in 50 days of 545MPa, equivalent to an initial strain of 0.5%. Strain continued to develop over the test duration, due to low temperature creep, to 0.9% after 50 days. Full-scale hub tests confirmed that this threshold level was appropriate to the hubs and that residual stress in the hubs contributed to cracking. In displacement-controlled bend tests, with deflection prior to exposure, threshold strains of 2.1% and 8% were identified for crack initiation and propagation respectively. Comparison with previously published work and powder metallurgy pipe indicated that the hub material was particularly sensitive to hydrogen embrittlement stress corrosion as a consequence of its microstructure, which had coarse aligned grains and nitrides/carbonitrides. Ferrite volume fraction and hardness were apparently of secondary importance.


Ferritic-austenitic stainless steels (i.e. duplex and superduplex grades) are attractive materials for subsea oil and gas industry applications due to their combination of high strength and resistance to pitting corrosion and chloride stress corrosion cracking. Nevertheless, they may be subjected to cathodic protection in practice either due to concerns over pitting/crevice corrosion resistance on the seawater side, when internal fluid temperatures are high, or due to electrical contact with other cathodically protected systems. Typically, protection is via use of aluminium alloy anodes, so that in-service potentials may be of the order of -1000 to -1100mV SCE, although less negative potentials, of the order of -800mV, have been adopted in some instances, via use of potential limiting diodes. At the more negative potentials, generation of atomic hydrogen at the steel surface may lead to significant hydrogen pick-up. The body centred cubic ferrite phase is generally susceptible to hydrogen embrittlement, whilst face centred cubic austenite is much less susceptible but not immune. [1] Consequently, duplex and superduplex steels with mixed ferrite-austenite structure show susceptibility to hydrogen embrittlement and there is a considerably body of published literature on various aspects of the subject. [2-6]

Two distinct forms of embrittlement exist. When material has been bulk hydrogen charged, substantial loss of toughness and ductility may be recorded. [3,7] However, at normal ambient temperatures, bulk hydrogen charging of ferritic-austenitic steels proceeds very slowly under cathodic polarisation due to the low hydrogen diffusion coefficients. [3] Under such conditions, progressive cracking or hydrogen embrittlement stress corrosion cracking (HESCC) may occur, associated with development of a very high concentration of hydrogen at the steel surface with no significant bulk charging. [2,6] Under such conditions, conventional mechanical tests would indicate no significant loss of toughness or ductility.

Many operators have had good experience with use of cathodically protected ferritic-austenitic steels but a number of failures have been reported. Brittle fracture of bulk hydrogen-charged cold worked downhole tubulars in 25Cr duplex stainless steel was observed during removal and subsequent handling. [3] Failures of 25Cr bolts have been reported also. These were reportedly manufactured from steel with a high ferrite content, around 70%, and had been aged to increase strength via precipitation of the alpha prime phase. [6] It was considered at the time that these failures could be attributed in large part to the process route of the materials involved and that there was no practical concern for solution annealed and quenched materials with around 50% ferrite. [6] Indeed data have indicated that the threshold stress for HESCC in such material is well in excess of the proof stress and approaching the UTS. [6,8]

In 1996, HESCC was identified in two forged superduplex stainless steel hubs on the BP Foinaven field development, during final commissioning of a subsea manifold, in a water depth of approximately 400m [9] . The general arrangement of a hub is shown in Fig.1. Cracking occurred in the machined nib region, adjacent to a weld to 6" pipe. The manifold was subject to cathodic protection by aluminium alloy anodes, giving potentials around -1050mV SCE. The hubs were solution annealed and quenched, had approximately 50-50 ferrite-austenite phase balance and service stresses had not been applied, although pressure testing had been performed and locked-in/residual stresses had inevitably been introduced during fabrication and installation. However, gross plastic deformation had not occurred, throwing doubt on the applicability of previously published data. A programme of full-scale and small-scale tests was undertaken at TWI on behalf of BP to investigate the threshold condition for cracking in the hub material. This threshold condition had to be defined in terms of strain, so that it could be related to the actual hub geometry via appropriate stress analysis. These tests have been summarised elsewhere [9] and this paper presents a more detailed appraisal.

 Fig.1. General view of hub and two 6" pipes after removal from the manifold
Fig.1. General view of hub and two 6" pipes after removal from the manifold

Experimental programme


The majority of the testing was performed on material taken from the two failed UNS S32760 superduplex hubs. For comparison, tests were performed on 6" diameter powder metallurgy pipe also to UNS S32760. The hub and pipe materials were characterised by point counting the ferrite phase using a 25-point grid over 16 fields of measurement. For the hub material, a magnetic etch was used, consisting of fine iron oxide particles in a paraffin solution, to strain the ferrite phase, whilst the pipe material was etched electrolytically in 20%H 2SO 4 with 0.1g/l NH 4 SCN. Conventional tensile tests (12mm/hour crosshead displacement rate), reduced strain rate tensile tests (0.1mm/hour) and Vickers hardness measurements were made on both hub and pipe materials and impact tests performed on the hub material. Test specimens were removed from the nib area of the hubs, where failure had occurred. Transmission electron microscope studies were performed on hub material to characterise third phases present.


Seven test types were employed, including six small-scale test types on hub and pipe materials, and full-scale testing of two hubs. All tests were performed in seawater collected from the North Sea on the east coast of the UK. The water was replenished periodically. Cathodic polarisation was applied by potentiostat with a platinum counter electrode and a standard calomel or Ag/AgCl reference electrode. All potentials reported are relative to a standard calomel electrode.

Constant-load bend tests were performed on smooth specimens at 5°C and pre-cracked specimens at ambient temperature, i.e. 15-20°C, and atmospheric pressure, with the load being applied by a hanging weight. The applied potential was -1100mV in each case. Specimens had square cross-section, ranging from 10x10mm to 12.5x12.5mm and were 200mm long with the test face ground to 600grit in most cases. The specimens were taken from the hubs predominantly in locations similar to the failure location, to ensure the same microstructural alignment with respect to the direction of cracking, i.e. the aligned ferrite-austenite structure was through-thickness and in the crack growth direction. Comparative tests were performed on smooth specimens (i) with opposite microstructural alignment (i.e. specimens taken from the main body of the hub, along its axis, so that alignment of the ferrite and austenite was along the length of the specimen and perpendicular to the crack growth direction), (ii) taken across the weld between the hub and 6" pipe, with the weld profile left intact, and (iii) with a small surface notch, 0.5mm deep and 0.16mm wide, machined at mid-length. Specimens were loaded to a range of loads up to 150% of the load required to give the nominal proof stress in the outer fibre, as calculated using a linear elastic analysis. Applied potentials of -1100, -800 and -700mV were examined and no pre-charging was employed. A finite element analysis was adopted to estimate actual outer fibre stress and strain. This indicted that 150% of the nominal proof loading was equivalent to a true stress of 650MPa and associated with an outer fibre strain of 1.3%. No allowance in the calculations was made for stress relaxation due to low temperature creep. Test duration was for up to 50 days.

The pre-cracked specimens were similar geometry to the smooth ones but had a fatigue pre-crack located at mid-length and approximately 40% of the specimen thickness. Loading was in the same sense as the smooth specimens. Specimens were broken open after significant deflection of the load arm indicated crack extension (defined as 'failure') or after 40 days if no significant deflection was recorded, and the crack extension measured.

A series of constant deflection four-point bend and U-bend tests was also undertaken on hub material. Specimen dimensions were 100x12x3mm and outer fibre strains were up to 2.25% for four point bending and 17% for U-bending. Tests were performed at -1100mV and ambient temperature, at a pressure of 50 bar in an autoclave, for 30 days. Further exposure was adopted at ambient pressure and temperature, without unloading the specimens, as no cracking was observed after 30 days. Some of the four point bend specimens were given six additional strain increments of 0.1% every 2 days. At each step, the specimens were examined visually for cracking. Finally, specimens were left for a further 90 days at ambient pressure with regular inspections for cracking using dye penetrant. A second set of similar constant deflection bend tests was performed at -1100mV and ambient temperature and pressure but with hydrogen pre-charging at -1100mV, 50°C, ambient pressure for 1 or 7 days. Exposure was for 90 days with periodic inspections for cracking.

Conventional slow strain rate tensile (SSRT) testing was adopted to compare different microstructures, applied potentials and test temperatures. Although this provided useful relative performance data, it could not assist in definition of the threshold condition, which was required to assess the suitability of the hubs for further service. Interrupted slow strain rate tensile (ISSRT) tests were performed at -1100mV and ambient temperature and pressure, using specimens with 20x3.7mm cylindrical gauge length, pre-charged for 7 days at -1100mV, 50°C and ambient pressure. A test strain rate of 2.6x10 -6s -1 was used. The first test was a multi-step test, with the stress being increased in small steps to 595MPa, and after each step the specimen was removed and examined for cracking using dye penetrant. The test was stopped when cracking was observed. The second test was pulled in one step to the penultimate load of the first test (565MPa) whilst the third specimen was pulled in one step to 590MPa and the final specimen was taken in one step to 580MPa. After test, each specimen was examined visually with penetrant application and then sectioned for evidence of cracking.

Constant-load tensile tests were performed at ambient temperature and -1100mV on cylindrical specimens from hub material, with 20x3.7mm gauge length, which were pre-charged for 7 days at -1100mV, 50 bar pressure and ambient temperature. One test was performed on a specimen given 5% pre-strain. Extensometers were arranged in parallel with the specimens to allow measurement of strain development during test. Test durations were between 30 and 100 days. A number of tests were performed in the constant-load frames but with the stress regularly adjusted to give an approximately constant-strain, as low temperate creep occurred during the constant-load tests. The constant-strain tests employed a container of water on the load arm, from which water was removed periodically to keep the applied strain within 0.025% of the desired value.

Two full-scale tests were performed on hubs in seawater at ambient temperature and a potential of -1100mV. The nib region, where failure occurred in practice, was immersed in seawater and put in tension, after pre-charging the hubs for 7 days, by applying a bending moment to 6" pipe welded to the nib (see general hub and pipe arrangement in Fig.1). The area of maximum stress was examined periodically under an optical microscope for evidence of cracking. The applied load was raised incrementally with a dwell of 3 to 5 days at each load, in order to identify the load for onset of cracking. The second test was a repeat of the first but in the second test the initial load was fairly close to the threshold level identified in the first test, so that fewer load steps were required. The hub nib region was strain-gauged, with the likely crack initiation area left uncovered. To allow estimation of strain at the initiation site, comparative measurements were made on another hub loaded in air. Residual stress measurements at the initiation site were made on a duplicate hub, using the centre hole drilling technique.


Material Properties

The chemical analyses of the hub and pipe were within the UNS S32760 specification and the phase balances were within the normal range for ferritic-austenitic steels, being 50-57% and 46-52% respectively, Tables 1 and 2. The 0.2% proof stress of the pipe material was higher than for the hub, as was the hardness, although a significant range of properties was measured within the hubs in particular. All values were consistent with the material grade. The difference between proof stress values recorded in tests with extension rates of 12mm/hour and 0.1mm/hour was marked (about 20% and 14% higher at the higher strain rate for hub and pipe respectively), reflecting the low temperature creep behaviour of ferritic-austenitic steels. The impact energies recorded for the hub (44-90J at -25°C) were fairly low for a parent superduplex steel but room temperature elongation and reduction of area values were typical for the grade. The impact energy values recorded reflect the fairly coarse hub material microstructure ( Fig.2), which was much coarser than the powder metallurgy pipe. Average austenite island size was about 50x250µm. The microstructure was aligned approximately through the hub nib thickness in the area of crack formation in the failed hub components. TEM examination indicated the widespread presence of nitrides/carbonitrides within the ferrite grains and on grain and phase boundaries ( Fig.2).

Table 1 Chemical analysis

SampleElement, wt%
Hub 1 0.03 0.002 0.028 0.43 0.52 7.1 25.0 3.61 0.61 0.25
Hub 2 0.03 <0.002 0.023 0.40 0.51 6.8 25.2 3.66 0.66 0.24
Pipe 0.03 0.010 0.019 0.58 0.75 7.1 25.1 3.49 0.65 0.25
UNS S32760 ≤0.03 ≤0.01 ≤0.03 ≤1.0 ≤1.0 6.0-8.0 24.0-26.0 3.0-4.0 0.5-1.0 0.2-0.3

Table 2 Properties of superduplex hub and pipe materials

Hub Ferrite fraction
0.2% proof stress
0.2% proof stress
Reduction of area
Impact energy
50-57% (15 areas)
568-628N/mm 2 (12mm/hour)
476-505N/mm 2 (0.1mm/hour)
38% (12mm/hour)
64% (12mm/hour)
44-90J at -25°C (8 areas)
255-295 HV5 (11 areas)
Pipe Ferrite fraction
0.2% proof stress
0.2% proof stress
46-52% (3 areas)
634N/mm 2 (12mm/hour)
556N/mm 2 (0.1mm/hour)
275-312 HV5 (10 areas)
Fig.2. Microstructure of superduplex hub material, X500
Fig.2. Microstructure of superduplex hub material, X500


Mode of Cracking

In all tests, when cracking was induced it was predominantly by transgranular cleavage of the ferrite phase and followed the alignment of the microstructure in preference to the plane of maximum stress ( Fig.3). Cracks tended to arrest at ferrite/austenite boundaries. There was no relationship apparent between the cracks and third phases, i.e. nitrides/carbonitrides.

Fig.3. Crack initiation in displacement-controlled bend test on hub material at -1100mV, X80
Fig.3. Crack initiation in displacement-controlled bend test on hub material at -1100mV, X80

Constant load bend tests

Figure 4 illustrates the results of constant-load bend tests on hub material at -1100mV and 5°C with applied loads up to 150% of the calculated load required to reach the 0.2% proof stress in the outer fibre. No cracking was observed in any test. Higher loads could not be used as the specimen deflection became unacceptable. The presence of a weld profile or small surface notch did not induce cracking.

Fig.4. Results of smooth four point bend tests in seawater at 5°C and -1100mV
Fig.4. Results of smooth four point bend tests in seawater at 5°C and -1100mV

Constant load pre-cracked bend tests

Figure 5 shows results of bend tests on pre-cracked hub specimens at -1100mV, -800mV and -700mV and pipe specimens at -1100mV, at ambient temperature. Cracking was observed in a few tens of hours in most cases, with the pipe material having greater resistance to cracking them the hub. Sensitivity to cracking in the hub was greater at -1100mV than -800/-700mV. Approximate threshold stress intensities for cracking in 40 days were 20-25MPa √m (hub at -1100mV), 50-60MPa √m (hub at -800/-700mV) and 60-70MPa √m (pipe at -1100mV). In the pipe specimens, crack propagation was along the direction of microstructural alignment, i.e. perpendicular to the pre-crack. Microstructural alignment apparently had little effect on the stress intensity for propagation of cracks in the coarser hub material, although propagation was always along the aligned ferrite phase.

 Fig.5a) Results of four point bend tests on pre-cracked specimens of hub material tested in seawater at ambient temperature and -1100, -800 and -700mV
Fig.5a) Results of four point bend tests on pre-cracked specimens of hub material tested in seawater at ambient temperature and -1100, -800 and -700mV
Fig.5b) Results of four point bend tests on pre-cracked specimens from hub and pipe materials in seawater at ambient temperature and -1100mV
Fig.5b) Results of four point bend tests on pre-cracked specimens from hub and pipe materials in seawater at ambient temperature and -1100mVFig.5b) Results of four point bend tests on pre-cracked specimens from hub and pipe materials in seawater at ambient temperature and -1100mV

Constant displacement bend tests

The -1100mV constant-displacement bend test results are summarised in Fig.6. No cracks were seen in any of these specimens when they were removed from the 50 bar autoclave after 30 days. Further exposure at ambient conditions produced through-thickness cracking, observed only in U-bend specimens deflected to 8 or 17% strain and only after 7 days or more after removal from the autoclave. A number of very small cracks was observed in specimens at lower strains down to 1.3%, Fig.6. Cracking at the lowest strain levels of 1.3 to 1.6% was only observed in specimens given a succession of 0.1% strain increments during test and only after a minimum of 14 days exposure at ambient conditions. For the specimens that were pre-charged prior to test and then exposed under ambient conditions, without strain increments during test, very small cracks were only observed at strains of 2.1% and above.

Fig.6. Results of constant-displacement bend tests on hub material in ambient temperature seawater at -1100mV
Fig.6. Results of constant-displacement bend tests on hub material in ambient temperature seawater at -1100mV

Constant load tensile tests

The constant-load tensile tests on hub material at -1100mV and ambient temperature showed results that were strikingly different to those from the smooth bend tests, Fig.7. Cracking was observed in a few hours at stresses around the proof stress and a threshold stress of 545MPa was established for cracking in 1000 hours (i.e. slightly below the conventional 0.2% proof stress). At the onset of cracking, the specimen with the lowest failure stress of 545MPa gave a strain of 0.9% and shortly after loading, it had a strain of 0.5%. One specimen (549MPa) showed a small crack that had not propagated to failure in 40 days. Stresses around proof stress were associated with a marked degree of creep, with final strains of 1.2-2.5% being recorded for specimens loaded between 550 and 650MPa. Conventional tensile tests indicated strains of about 0.5 to 1.3% for the same stresses. The creep strain developed most quickly at the start of the test and the strain rate fell slowly over the 30-100 day test duration, although creep was still on going after this period. Strain rates averaged over the test duration were between 1x10 -4 and 1x10 -9s -1, depending on applied stress. A typical elongation versus time curve is given in Fig.8. The sample subjected to a pre-strain of 5% and tested at 600MPa did not crack and showed no appreciable creep. The approximately constant-strain tensile tests did not show cracking at strains of 0.5-1.2%, which were maintained by stresses between 372MPa falling to 362MPa and 587MPa falling to 521MPa. A specimen held at 1.3-1.5% strain (602MPa falling to 562MPa) failed in 24 hours.

Fig.7. Results of constant load tensile tests on hub material in seawater at ambient temperature and -1100mV
Fig.7. Results of constant load tensile tests on hub material in seawater at ambient temperature and -1100mV
Fig.8. Elongation of constant load tensile specimen at 549MPa, due to low temperature creep over test duration
Fig.8. Elongation of constant load tensile specimen at 549MPa, due to low temperature creep over test duration

Interrupted slow strain rate tests

The interrupted slow strain rate tests at -1100mV and ambient temperature indicated that cracking initiated in hub material between 580 and 590MPa, with strain between 0.7 and 1.3%.

Full-scale hub tests

Two full-scale tests showed cracking at applied strains of 0.25% in both cases, as determined by strain gauging of the test specimen and a duplicate hub loaded in air. Measurement indicated residual strain of 0.23% in the nib region of a comparable hub acting perpendicular to the crack plane, giving a total strain to crack of 0.48% for the two full size hub tests.


Test method

The work demonstrated the pronounced difference between load-controlled and displacement-controlled tests on ferritic-austenitic stainless steel under cathodic protection. Whilst crack initiation was induced at strains down to 2.1% in constant deflection bend tests, it required a long period of time (several hundred hours) and propagation only occurred at 8% strain and above. Crack initiation was encouraged by small increments of strain during test, reducing the threshold strain for cracking to 1.3%. When strain increments were not used, 2.1% strain was required for crack initiation. Constant-load tensile tests gave failures in 1-2 hours at strain levels of 1.3 to 1.5% and the lowest failure strain of 0.9% was obtained from a test lasting 1000 hours. Under constant-load conditions, crack initiation without propagation was only observed in one specimen. The difference between the test techniques is a result of low temperature creep, which leads to load relaxation in displacement-controlled tests. Low temperature creep proceeded fairly rapidly on load application and the rate then fell in constant-load tests, so that load relaxation would be expected to occur rapidly in a constant-displacement test and then more gradually at longer times, presumably reaching equilibrium after several hours or days. The rate of stress relaxation will also depend on the relative stiffness of the specimen and the test fixture. A fixture with high stiffness compared to the specimen will allow rapid stress relaxation in the specimen, whereas for a fixture whose stiffness more closely matches that of the specimen, stress relaxation in the specimen will be reduced by elastic springback in the fixture. This was the case in the full-scale tests where the stiffness of the loading frame was comparable with the hub and pipe test assembly. Interrupted SSRT tests produce a stress-strain regime that is intermediate between the two extremes of load and displacement control. Constant extension ensures that the specimen never reaches the equilibrium relaxed position associated with fixed displacement tests but at the slow strain rates adopted (2.6x10 -6s -1 was used in this instance and 1x10 -8s -1 has been suggested recently as appropriate for ferritic-austenitic steels under cathodic polarisation [10] ), some relaxation is inevitable (creep rates of 1x10 -4 to 1x10 -8s -1 were recorded for constant-load tests). This was reflected in the tests performed here by measurement of a threshold stress of 580-590MPa and threshold strain of ≤1.3% in ISSRT tests.

The results indicate that for cathodically protected ferritic-austenitic steel in a load-controlled application, threshold conditions can only be derived from constant-load tests. Constant-displacement tests or ISSRT tests may give non-conservative data for load-controlled applications due to stress relaxation. Conversely, constant-load tests do not reproduce ferritic-austenitic material behaviour under displacement control. When examining HESCC of ferritic-austenitic steel, it is essential to apply an appropriate test method. Further, it may be noted that susceptibility to HESCC seems to be significantly higher under load control than under displacement control.

Threshold for HESCC of ferritic-austenitic steel

The tests demonstrate a number of difficulties in establishing a stress and strain for HESCC crack formation in duplex/superduplex steel. Firstly, specimens may develop cracks that do not propagate, so it may be necessary to consider initiation and propagation separately. Approximate threshold stress intensities for propagation were derived fairly simply from pre-cracked specimens. However, the results for the hub material were such that fracture mechanics calculations indicated extremely small critical defect sizes. The data indicated that the hub material was significantly more susceptible than the pipe and more susceptible than a cast superduplex steel, which was the subject to a previous published study. [11] In smooth specimens required to examine initiation, strain is not simply a function of applied stress but also a function of time. In the hub material, the cracking threshold was 545MPa, i.e. around the proof stress. In order to model the actual in-service hub loading, it was necessary to consider strains, which implied that a threshold strain figure had to be derived from the constant-load tests. A conservative approach of taking the strain at the start of test (defined as 10 minutes after load application) was adopted in this instance, giving a threshold strain of 0.5%. This approach was supported by the full-scale tests, where the applied strain for crack formation (0.25%), measured from strain gauges, correlated well with the small-scale test threshold once an allowance had been made for residual strain in the full-scale hub forging (measured as 0.23% in the relevant orientation). This implies that the residual strain in the hubs contributed to crack formation.

For loading under displacement control, the test data indicated that the material has much higher tolerance than in load control, with 2.1% strain being required for crack initiation and 8% required for propagation in bend tests. Although finite element analysis was used to estimate stress in constant-load bend tests ( Fig.4), it was impossible to accurately relate the strains measured in these bend tests to actual stress levels as standard finite element analysis does not take account of stress relaxation. It may be noted that the constant-load bend tests did not induce cracking at loading nominally well above the proof stress. This indicates stress relaxation in the outer fibres, which were presumably constrained by the elastic bulk of the specimen cross-section. If testing is to be performed to establish threshold conditions, the results indicate that a 30 day test duration is appropriate for load controlled conditions (e.g. constant load tensile tests) but inadequate for displacement controlled conditions (e.g. constant displacement bend tests), for which a 90 day duration may be more appropriate. The bend test results indicate that hydrogen pre-charging reduces the time and strain required for crack initiation, i.e. cracking occurs more readily when strain is applied to hydrogen charged material rather than when strained material is then hydrogen charged. This is consistent with the rapid cracking in tensile tests, when hydrogen charged material is actively strained over several days due to creep, without stress relaxation. It is suggested that active strain in ferrite with a high concentration of hydrogen is required for cracking.

Interrupted slow strain rate tests allowed some definition of threshold conditions for cracking but fall between constant load and constant displacement tests, in terms of allowing some stress relaxation. This makes data derived somewhat difficult to relate to actual structures, although slow strain rate tests are in general useful for assessing relative susceptibilities of materials.

Practical implications of the data

The data obtained indicated that cracking of the hub material (both initiation and propagation) occurred at around the 0.2% proof stress. However, the presence of residual stress and locked-in fabrication stress combined with the presence of notches allowed development of HESCC in the hubs at low applied stress levels (maximum allowable applied stresses of 322 and 305MPa were derived for two hub geometries. [9] Design codes typically allow high stresses locally at stress concentrators and assume that local plastic deformation will be acceptable. This assumption was incorrect for these superduplex hubs when exposed to cathodic protection. High residual stresses are inevitable in ferritic-austenitic components due to the practice of quenching after solution annealing and further stresses develop at welds but they were exacerbated in the present case by the machining of the hubs which exposed areas with high tensile residual stresses.

There were a number of indications that the hub material microstructure had contributed to its sensitivity to cracking. Published data have demonstrated grain size and grain orientation effects, [12,13] such that coarse aligned microstructures would be expected to have fairly low resistance to cracking along the direction of alignment. This may explain the much lower resistance of the hub material to HESCC when compared to finer grained products studied here and in previous work. [6,8,11] The presence of third phases (nitrides and alpha prime) in the ferrite phase have been linked to enhanced cracking susceptibility [6,11,13] and a substantial population of nitrides/carbonitrides was present in the hubs, although it is noted that the presence of nitrides is more prevalent in coarse structures, so it is difficult to distinguish between the effects of nitrides and grain size.


  1. For the superduplex hub material examined, the threshold stress (545MPa) for HESCC under cathodic protection was slightly below the conventional 0.2% proof stress. This is much lower than previously reported data have indicated for duplex and superduplex steels.
  2. The results indicate a marked effect of microstructural features on sensitivity to HESCC of ferritic-austenitic steels. The low threshold for cracking in the hub material was apparently a result of the coarse aligned microstructure, perhaps with a contribution from the nitrides/carbonitrides present. Ferrite volume fraction and hardness were apparently of only secondary importance.
  3. Test method selection has a very significant influence on the measured threshold for HESCC in ferritic-austenitic steels under cathodic protection. In the present case, constant-load tests gave cracking to failure in a few hours at around the 0.2% proof stress, whilst constant-displacement tests required 8% strain for crack propagation and 2% for crack initiation over a period of several hundred hours. Similar differences in service behaviour would be expected for load-controlled and displacement-controlled loading.
  4. The different behaviour of ferritic-austenitic steels under load and displacement control may be accounted for by low temperature creep, which allows rapid stress relaxation in the latter case.
  5. In actual hub components, residual stress apparently contributed to crack development, so that the maximum allowable applied strain was 0.25%, which translated to maximum allowable applied stresses of 322 and 305MPa for two hub geometries. Cracking was induced in the hubs by exceeding these threshold values locally at a notch.


  1. Huang J H and Altstetter C J, 'Cracking of duplex stainless steel due to dissolved hydrogen', Met Trans A, 26A (May 1995), p1079.
  2. Mukai Y, Murata M and Wang J, 'Analytical evaluation and mechanism of hydrogen embrittlement SCC in duplex stainless steel welds', Welding International, 1992, 6 (2), p110.
  3. Sentence P, 'Hydrogen embrittlement of cold worked duplex stainless steel oilfield tubulars', Duplex stainless steels '91, p895 (Les Editions de Physique, 1991).
  4. Krishnan K N, Knott J F and Strangwood M, 'Hydrogen embrittlement during corrosion fatigue of duplex stainless steel', Hydrogen effects in materials, p689 (The Minerals and Metals Society, 1996).
  5. Gooch T G and Leonard A J, 'Hydrogen cracking of ferritic-austenitic stainless steel weld metal', Duplex America 2000, p345 (KCI Publishing, 2000).
  6. Francis R, Byrne G and Warburton G R, 'The effect of cathodic protection on duplex stainless steels in seawater', CORROSION 95, paper 290 (Houston, Tx: NACE International, 1995).
  7. Cohen L J R, Charles J A and Smith G C, 'Hydrogen embrittlement in thermally charged duplex stainless steels', Hydrogen effects on material behaviour, p363 (The Minerals, Metals and Materials Society, 1990).
  8. Olsson P, Bauer A D and Eriksson H, 'Hydrogen embrittlement of duplex grades UNS S32750 and UNS S31803 in connection with cathodic protection in chloride solutions', Stainless steel world '97, p607 (KCI Publishing, 1997).
  9. Taylor T S, Pendlington T and Bird R, 'Foinaven superduplex materials cracking investigation', Offshore Technology Conference 1999, p467.
  10. Turnbull A, Griffiths A and Reid T, 'Hydrogen embrittlement of duplex stainless steels - simulating service experience', CORROSION '99, paper 148 (Houston, Tx: NACE International, 1999).
  11. Pohjanne P and Festy D, 'Hydrogen embrittlement of duplex stainless steel and maraging steel in seawater: effect of pressure', CORROSION '94, paper 223 (Houston, Tx: NACE International 1994).
  12. Francis R, 'Hydrogen embrittlement in duplex stainless steels', Journal of Offshore Technology, 1994, p19.
  13. Zheng W and Hardie D, 'Effect of structural orientation on the susceptibility of commercial duplex stainless steels to hydrogen embrittlement', Corrosion, 47 (10), 1991, p792.
  14. Byrne G, Francis R and Warburton G, 'Variation in mechanical properties and corrosion resistance of different alloys within the generic designation UNS S32760', vide ref. 5, p113.

For more information please email: