Hydrogen Embrittlement of 8630M/625 Subsea Dissimilar Joints

Department of Metallurgy, Corrosion and Surfacing
TWI Ltd
Great Abington, Cambridge, UK

 

Paper presented at ASME 28th International Conference on Ocean, Offshore and Arctic Engineering, OMAE 2009, Honolulu, Hawaii, 31 May - 5 June 2009. Paper # OMAE 2009 - 80030.

Abstract

Following the failure of a small number of subsea dissimilar joints, there is significant interest in understanding the fracture mechanism(s) and in qualifying future items for avoidance of further failures. Subsea dissimilar joints typically comprise a butter weld deposit onto the joint face of an alloy steel forged hub/tee/elbow, postweld heat treated, and completed with a closure weld to a pipeline steel. The case examined in this paper, consists of a build up of alloy 625 onto an 8630M forging. Hydrogen embrittlement test methods are described and the results from the tests interpreted with respect to their implications for fabrication and service. The microstructure and chemistry across the dissimilar interface is found to be of fundamental importance, and this is illustrated by the relative performance of gas-tungsten arc (GTA), hot-wire GTA and friction welds. Guidance for fabrication, installation and service is given for avoidance of further hydrogen embrittlement failures.

1. Introduction

1.1 Background:

A common subsea dissimilar joint comprises a shop fabricated, postweld heat treated hot wire GTA nickel alloy butter weld deposit on a low alloy forging steel fitting, and an as-welded field GTA/manual metal arc nickel alloy closure weld between the butter and, typically, a micro-alloyed steel pipeline. Thus, the heat affected zone (HAZ) of the completion field weld lies wholly in the nickel alloy butter layers and no postweld heat treatment (PWHT) of the field weld is required. As for any fusion welded dissimilar joint, a narrow, partially-mixed zone (PMZ) is observed at the dissimilar interface, where the composition is graded between that of the parent forging and the 'bulk' chemistry of the first layer buttering runs. In order to prevent corrosion of piping systems subsea, cathodic protection is applied. This does inhibit corrosion, but can result in a certain amount of hydrogen charging. Whilst the vast majority of subsea joints have given successful service, a small number has failed at the dissimilar interface, with significant consequence.

The work presented in this paper is part of a larger programme of work conducted at TWI on a range of subsea dissimilar joints. The microstructure and failure modes in commercially produced hot-wire GTA buttered 8630M dissimilar joints, has been detailed elsewhere,^[1] but is briefly summarized in the following text to establish the terminology and highlight the critical zones in the present paper.

1.2 Microstructure of 8630M/625 dissimilar joints:

The interface can be divided into six different microstructural zones, which are summarized in Figure 1 moving from low alloy steel to butter deposit:

1. Parent material of the 8630M forging. This zone exhibited a body centred cubic (BCC) ferritic microstructure, with a small volume fraction of inclusions and carbides;
2. A fine decarburised zone within the 8630M grain coarsened HAZ (GCHAZ) close to the dissimilar interface, with fingers of fused material penetrating into prior austenite grain boundaries (labelled Zone Δ). Again, this zone was mostly ferritic and BCC;
3. An iron-rich martensite region mostly at inter-pass positions (where 'swirls' of trapped steel extend a few tens of microns into the butter weld - see schematic drawing in Figure 1) extending from the dissimilar interface into the PMZ (labelled Zone M). This zone exhibited a body centred tetragonal (BCT) martensitic microstructure;
4. A hard, particle-free, supersaturated carbon solid solution in an 'austenitic' region of the PMZ that has probably solidified in a planar manner (a 'featureless zone', labelled Zone Φ). This zone exhibited a face centred cubic (FCC) austenitic microstructure;
5. A region of the PMZ containing numerous high atomic number interdendritic particles (labelled Zone Π). This zone exhibited an FCC austenitic matrix;
6. The bulk of the weld with an alloy 625 chemistry diluted with a small proportion of the steel forging. This zone exhibited a FCC matrix.

Zones Φ and Π constituted the PMZ within the weld metal, with a chemistry graded between that of the parent metal and that of the bulk of the first butter layer. This PMZ mainly consists of Fe, Ni and Cr with proportions that were closer to that of the forging, close to the interface, and closer to the alloy 625 weld deposit further away. It has a width from a few tens to a few hundreds of microns (although mostly at the lower end of this scale). It should be noted than in addition to the inter-pass swirls described above, similar features are present at intra-pass positions, evidently due to periodic perturbations in weld penetration during welding.

1.3 Fracture morphologies:

When bend samples notched at the fusion line are mechanically tested under cathodic protection, failure occurs by a combination of fracture morphologies as follows:

• 'Terraced cleavage' fracture in Zone Φ of the PMZ away from inter-pass positions. Hydrogen embrittlement of austenitic material is not particularly common, but can occur under severe conditions.^[2] Given the presence of strain, the unusually high carbon content of Zone Φ^[1], its high hardness^[1] and the tendency for hydrogen to build up in this zone under hydrogen charging conditions^[3], it is apparent that local conditions satisfy the criteria for hydrogen embrittlement and cracking to occur within the 'austenitic' PMZ.
• 'Dimpled' fracture located predominantly at inter-pass and intra-pass positions, associated with Zone M cracking directly at the dissimilar interface or the martensite in this region. This fracture mode is exacerbated by hydrogen charging but can also occur in the absence of hydrogen.

Zone Φ and Zone M are particularly influential in terms of fracture mechanisms of subsea dissimilar joints under cathodic protection. Their microstructure and chemistry are influenced by welding process, procedure and subsequent PWHT. Thus, performance may be modified by subtle changes in any of the parameters. The potential magnitude of these changes has been explored in the present work.

2. Approach

The primary aim of the present work was to determine the factors that influence performance, in order to provide some guidance for fabrication of dissimilar joints to improve hydrogen embrittlement resistance. With this in mind, a series of dissimilar 8630M/alloy 625 deposits was fabricated using hot-wire GTA, GTA and friction welding. The microstructure and chemistry across the dissimilar interfaces were determined and, for some of the deposits, the hydrogen embrittlement performance was assessed.

3. Materials and fabrication

3.1 Commercially produced welds:

The joints (Figure 2) consisted of a build up of alloy 625 (ERNiCrMo-3) on a low alloy steel forging (8630M), deposited using hot wire GTA welding with an applied preheat. The buttered forging was given a postweld heat treatment (PWHT) of 2 hours at 1250F (676°C) and a completion weld made, again using alloy 625, to a steel pipe. The microstructure was examined in detail in previous work, together with fracture mechanisms^[1] and both are summarized above in Section 1. The environmental performance was investigated in the present paper.

3.2 Laboratory welds:

These consisted of a series of single layer, bead-on-plate deposits, each comprising of three overlapping beads, made using conventional GTA welding (i.e. not hot wire). Welding parameters employed are given in Table 1. A typical welding sequence is presented in Figure 3. Welds were made with a 3.2mm (0.126in) diameter alloy 625 (ERNiCrMo-3) wire. A total of four deposits were made using pulsed or constant current and comparatively low or high arc energies. Constant current was compared with pulsed current to see if this had any influence on the formation or otherwise of various welding features, especially intrapass swirls of diluted steel penetrating into the weld. A typical and an extended PWHT were performed: (i) 4 hours at 1230F (665°C), and (ii) 24 hours at 1230F (665°C). The aim was to assess the effect of the PWHT duration on carbon diffusion and to establish any potential interaction between the microstructure/chemistry of the PMZ and the carbon diffusion.

Table 1 Welding parameters employed for the bead-on-plate trials simulating the first three passes of the first butter layer of a dissimilar joint. Alloy 625 wire was employed for the weld.

TIG (cold wire)	Pulsed	Pulsed	Constant current	Constant current
Preheat/Interpass, °C		150/250
Wire diameter, mm (in)		3.2 (1.259)
Ar gas flow rate l/min	12
Arc voltage, V	13	14	14.8	12.6
Current, A peak Current, A background	240 80	291 90	291 90	240 80
Welding speed, mm/min	300	250	300	300
Wire feed speed, m/min	0.62	0.74	0.74	0.62

3.3 Friction welds:

A low alloy steel, 8630M, was joined to alloy 625 bar using rotary friction welding in an attempt to produce an abrupt interface. Microstructure and chemistry across the dissimilar interface were examined and performance was established in both as-welded and postweld heat treated conditions. Three PWHTs were employed: (i) 2 hours at 1250F (676°C), (ii) 10 hours at 1250F (676°C), and (iii) 10 hours at 1100F (590°C).

4. Experimental approach

4.1 Metallurgy across the dissimilar interfaces:

Microstructure and chemistry: Transverse and longitudinal sections were examined using light and scanning electron microscopy. Semi-quantitative elemental analysis was conduced using EDX in order to determine the profile of Fe, Ni, Cr, Mo and Nb across the dissimilar interface.

Carbon profile and hardness: The carbon profile was qualitatively established using an electron probe micro-analyser (EPMA) with a dedicated spectrometer for carbon. This analysis was carried out on commercially produced and laboratory welds. Vickers 5kg hardness values were taken on transverse sections from commercially produced and laboratory welds in the as-welded condition, and postweld heat treated conditions. The last bead of the run of three of the laboratory welds was tested in each case (ie there was no tempering from adjacent deposits). Hardness indents were targeted to the GCHAZ, both very close to the fusion boundary and more distant, and the PMZ and bulk of the deposit. Results were compared with data from commercially produced subsea dissimilar joints collected in previous work.^[1] Conventional Vickers hardness indents are too large to characterize hardness changes across the dissimilar interface, thus micro and nano-hardness profiles were performed across the dissimilar interface to establish any trends. Micro-Vickers hardness measurements were acquired with a low load (25g) and nano-hardness indentation were performed with a load of 4mN and thus the absolute values are not directly comparable with the Vickers results. Primarily, they serve to highlight the location of the peak hardness.

4.2 Environmental performance:

Specimen preparation and examination: Single edge notch bend (SENB) specimens were machined such that the dissimilar interface was positioned in the middle, transverse to the length of the specimen and notched using electro discharge machining (EDM), in the through-thickness direction. Specimens were hydrogen pre-charged prior to testing. After testing, the specimens were heat tinted at 572F (300°C) prior to being broken open for examination. Fracture face examination was carried out using light and scanning electron microscopy. The latter was coupled with energy dispersive X-ray spectroscopy, to help determine the location of the morphological features with respect to the microstructural zone.

Threshold stress intensity factor: Incremental, dwell loading tests were carried out in three point-bending to achieve stress intensity factor increments of 2.5 to 5 MPam^0.5. These tests were performed at 37.4F (3°C) in 3.5% NaCl aqueous solution under cathodic protection at -1100mV_SCE, applied by potentiostat. The data were used to estimate K_IH. For comparison, a limited amount of testing was performed in air using a similar methodology to the environmental tests, but without hydrogen pre-charging. A further conventional CTOD test was performed to produce comparative K_Q data. Finally, two notched specimens were simply broken open to reveal typical fractographic features: one of these was hydrogen pre-charged for two weeks, whilst the other was tested in the as-fabricated condition.

Crack growth resistance: Similar SENB specimens to those used for K_IH testing, were used to establish an 'R-curve' under slow strain rate loading conditions. An unloading compliance method was employed to establish a crack growth resistance curve following BS7448 Part 4. Again, testing was conducted at 37.4F (4°C) temperature in 3.5% NaCl under cathodic protection at -1100mV_SCE.

5. Results

5.1 Effect of weld fabrication parameters:

Microstructure and chemistry: The overall width of the PMZ (as determined from the tail-off in iron content from the fusion boundary to a constant level in the butter deposit) was similar for each weld: approximately 20-30µm. The general metallurgical features observed on sections extracted from bead-on-plate welds were mostly very similar to those observed in the commercially produced welds, though variations were found depending on the welding parameters:

• The PMZ thickness was typically a few tens of microns wide (Figure 5), but did not show any particular trend in width with welding parameters.
• The pulsed current lower arc energy deposit exhibited no discernible Zone Φ at the mid-bead position. A narrow Zone Φ was found for the pulsed current higher arc energy deposit, but this was significantly narrower than that of the two constant current deposits (Figure 5).
• Constant current parameters resulted in fewer intra-pass swirls, although their presence was not eliminated entirely.
• Lower arc energy tended to promote the formation of martensite (Zone M), located predominantly at interpass positions near swirls.
• Some evidence was found of fine martensite at mid-bead locations, when applying a low arc energy and constant current, although this might have been associated with intra-pass swirls;
• A higher proportion of inter-dendritic particles was found in the far PMZ when applying a low arc energy and constant current.

Fig. 5. SEM image and chemical profile across the dissimilar interface of laboratory weld fabricated using
a) a lower arc energy and pulsed current;

b) a higher arc energy and pulsed current;

c) a higher arc energy and constant current;

d) a lower arc energy and constant current. Dashed lines show limit of Zone Φ.

Fig. 5. SEM image and chemical profile across the dissimilar interface of laboratory weld fabricated using

The chemical profiles across the dissimilar interface at the mid-bead position, presented in Figure 5, varied depending on the welding parameters as follows:

• The Zone Φ was Fe-rich for all of the welds except for the lower arc energy constant current weld, which had a graded chemistry from Fe-base to Ni-base (similar to the commercially produced weld, albeit with greater dilution).
• Combining the lower arc energy and pulsed parameters virtually eradicated the particle-free region. In this case, Mo and Nb-rich particles were observed right up to the dissimilar interface. Selected particles were identified definitively as NbC using electron back scattered phase identification.
• At the inter-pass positions, the presence of partially-melted steel swirls penetrating into the weld metal resulted in periodic chemical variations, giving a range of composition between that of the 8630M forging and the alloy 625 butter.

The friction welds, after PWHT, exhibited a 100µm width zone of 'perturbed' microstructure and chemistry at the dissimilar interface (Figure 6). The microstructure varied across the dissimilar interface and can be divided into six zones from the low alloy steel to alloy 625 bar as follows:

1. Parent material of the 8630M forging exhibiting a fine grained BCC structure;
2. An HAZ within the 8630M along the fusion boundary;
3. The dissimilar region exhibited sequential thin layers of steel and alloy 625. This zone was less than 10µm width;
4. A fine grained region, about 30 to 40µm in width and exhibiting a chemistry close to alloy 625;
5. A region 50 to 60µm wide of slightly coarser grains enriched in Fe compared to alloy 625;
6. The bulk of the alloy 625.

5.2 Influence of Postweld Heat Treatment:

Carbon profile: EPMA analysis across the dissimilar interfaces exhibited a peak of carbon near the fusion boundary in the featureless region of the PMZ (Zone Φ), as shown in Figure 7. These results are qualitative, but are indicative of a significant concentration of carbon.

Hardness: After PWHT, an example of one of the commercially produced welds exhibited Vickers hardness values of 236HV in the bulk, 251HV in the HAZ, 253HV in the PMZ and 268HV in the bulk of the butter. It is noted that Vickers hardness indents are too large to characterise hardness changes across the dissimilar interface, but results indicate that the joint approximately met the maximum steel ISO 15156 hardness limit of 250HV for sour service. When employing microhardness at mid-bead positions, a peak of approximately 450µHV was found in Zone Φ of the PMZ and similar microhardness peaks were noted at inter-pass positions within the martensitic region (Zone M). Nano-indentation hardness measurements confirmed this trend and showed hardness values up to 650nHV in the martensite of the Zone M and up to 700nHV in Zone Φ (Figure 8). Nb and Mo-rich interdendritic particles observed in Zone Π showed dramatically higher hardness values (983nHV and 1183nHV), as would be expected for niobium or molybdenum carbides.

In laboratory welds, the highest as-welded hardness was found in the GCHAZ of the forging, which was reduced on PWHT. The hardness of a fully martensitic microstructure is anticipated to be in the region of 550HV in 8630M. Thus the values recorded (approximately 500HV) are consistent with a primarily untempered martensitic microstructure. Hardness tests performed in the second pass (in the tempered region) of the lower arc energy welds showed somewhat lower hardness (<430HV), consistent with some degree of tempering. Postweld heat treatment resulted in a reduction in GCHAZ hardness, with the softening being greater for the extended PWHT time. In summary:

• The as-welded condition showed high values of GCHAZ hardness with a maximum hardness of 515HV, consistent with a predominantly martensitic microstructure.
• Welds given an extended PWHT (24 hours at 1230F (665°C)) exhibited slightly lower GCHAZ hardness values than those given the shorter PWHT (4 hours at 1230F (665°C)), with a range of values from 218 to 222HV and 253 to 271HV, respectively;
• Welds made using pulsed and constant current parameters showed similar values for similar arc energies.
• Hardness in the deposit itself was reasonably consistent in all the welds, lying in the range 174-206HV. After PWHT, slightly greater values were observed, consistent with some precipitation hardening in the alloy 625 deposit with moderately greater values after the extended PWHT.

The micro-Vickers hardness profile measured across the dissimilar interface of the laboratory welds confirmed the overall trends of the macro Vickers hardness results with peak hardness in the forging GCHAZ in the as-welded condition, shifting, after PWHT, toward Zone Φ of the PMZ, with values between 400 and 500µHV as illustrated in Figure 9 for the laboratory weld fabricated using a higher arc energy than the commercially produced weld, and constant current. As noted previously, PWHT resulted in softening of the GCHAZ and hardening of the PMZ and bulk of the alloy 625 deposit. Similar results were found when employing nano-hardness and after PWHT, the peak of hardness was found in Zone Φ immediately adjacent to the fusion boundary (within approximately 2-10µm). For the lower arc energy weld, which did not exhibit any Zone Φ, no significant hardness peak was observed. Longer heat treatment resulted in greater hardness consistent with a pile-up of carbon due to the difference of carbon mobility at the PWHT temperature, between the BCC and FCC regions.

6. Environmental performance tests

6.1 Threshold stress intensity factor:

Commercially produced welds: The results of the threshold stress intensity tests using incremental dwell loads are shown in Table 2 (labelled: GTA - 2h @ 1250F). Both the threshold stress intensity factor (K_IH: maximum value applied without crack initiation) and the value at which cracking was observed (K_failure) are reported.

Table 2. Stress Intensity factor at the failure point K_failure and Threshold stress intensity factor K_IH

Fabrication	Environment	Kfailure (MPam0.5)	KIH (MPam0.5)
GTA - 2h @ 1250F	3.5%NaCl @ 3°C	60	57.5
GTA - 2h @ 1250F	3.5%NaCl @ 3°C	62	57
GTA - 2h @ 1250F	3.5%NaCl @ 3°C	64	54
GTA - 2h @ 1250F	Air @ 3°C	74	69
GTA - 2h @ 1250F	Air @ 3°C	Conventional CTOD KQ=76
FW - as-welded	3.5%NaCl @ 3°C	47.5	45
FW - 2h @ 1250F	3.5%NaCl @ 3°C	52.5	50
FW - 2h @ 1250F	3.5%NaCl @ 3°C	65	55
FW - 10h @ 1250F	3.5%NaCl @ 3°C	57.5	55
FW - 10h @ 1250F	3.5%NaCl @ 3°C	57.5	55
FW - 10h @ 1100F	3.5%NaCl @ 3°C	62.5	60
FW - 10h @ 1100F	3.5%NaCl @ 3°C	77.5	70

The lowest K_IH obtained during environmental testing and under cathodic protection was 54MPam^0.5 and corresponded to a K_failure of 64MPam^0.5. Note that this should not be viewed as a conservative threshold value, since only a limited number of tests was performed. The stress intensity factor for crack initiation for a specimen tested in air (eg with only that hydrogen present from fabrication) was 74MPam^0.5 (Corresponding K_IH=69MPam^0.5). For comparison, the stress intensity factor K_Q, determined by conventional CTOD testing, was 76MPam^0.5. This limited data set is only intended to provide indicative values. Even so these values illustrate that, in this instance, hydrogen from fabrication had less influence than hydrogen from the testing environment.

Typical fracture face morphologies are presented in Figure 10. Post-test examination of the environmentally-tested SENB specimens employed for threshold stress intensity factor and crack growth resistance curve determination showed similar fracture morphologies and exhibited plastic deformation around the edges of the samples, together with cracking along the dissimilar interface.

Extensive crack growth from the notch was observed and specimens showed fractographic features that were similar to those observed in previous work^[1] presented in Section 1; eg flat and dimpled Zone M cracking and cleavage in Zone Φ of the PMZ, together with a flat and relatively featureless fracture morphology at the interface, with no particularly discernible characteristic features other than a profile of cellular ridges (Figure 11).

The chemistry of the fracture face showed minimal amounts of dilution, consistent with fracture at or very close to the dissimilar interface. This cracking mode has been observed at dissimilar interfaces in vessels for high temperature hydrogen service, specifically at the interface between austenitic claddings and low alloy Cr-Mo steels. Here, rather than hydrogen charging through cathodic protection, thermal hydrogen charging is experienced and cracking is a risk upon cooling during shutdowns.^[3-4]

When tested in air (incremental step loading), the fracture faces exhibited mostly ductile features, and numerous isolated patches of Zone M cracking. Similar results were observed for the conventional CTOD test specimen and for an SENB specimen that was simply broken open. However, cleavage within Zone Φ of the PMZ was only found in specimens tested under hydrogen-charging conditions.

Friction welds: The results of the threshold stress intensity tests using incremental dwell loads are shown in Table 2. The lowest threshold stress intensity factor measured during environmental testing and under cathodic protection was 45MPam^0.5 in the as-welded condition, and the highest threshold stress intensity factor was 60MPam^0.5 for a specimen that was given a PWHT of 2 hours at 1100F (590°C). Specimens, postweld heat treated 2 hours and 10 hours at 1250F (676°C), exhibited threshold stress intensity factors of 50MPam^0.5 and 55MPam^0.5, respectively. These values are comparable to the hot wire GTA dissimilar welds given a similar PWHT, which exhibited a threshold stress intensity factor of 54MPam^0.5.

Typical fracture faces are presented in Figure 12. Post-testi examination revealed that fracture occurred by cleavage in the HAZ of the forging for friction welds in the as-welded condition. After PWHT of 2 hours at 1250F (676°C), the fracture faces exhibited some cleavage in the partly mixed region of alloy 625/8630M, immediately adjacent to the dissimilar interface and some micro-void coalescence. Similar results were found when employing a longer PWHT (eg 10 hours at 1250F (676°C)). PWHT for 10 hours at 1100F (590°C) resulted in a fracture mostly by micro-void coalescence, together with patches of flat and featureless cracking at the edge of the specimens, similar to the morphology noted for hot wire GTA welds.

6.2 Crack growth resistance:

Commercially produced welds: The R-curves are plotted as J-extension curves in Figure 13. The J_R curves established for the commercially produced weld were fairly 'flat', typical of brittle behaviour.^[5-7] This was confirmed by post-test examination. Note that a material showing a ductile behaviour (associated with the growth and coalescence of micro-voids) would exhibit a rising curve, where J increases with crack growth. A flat part of the curve was reached at 30 to 40N/mm, which is low. Fracture morphologies were once again cleavage in Zone Φ, Zone M cracking and flat and featureless 'disbonding'.

Friction welds: The results are plotted as J-extension curves in Figure 13. In the as-welded condition, the friction welds exhibited no resistance to crack growth. Cracking initiated at the dissimilar interface but quickly deviated and propagated into the HAZ of the forging.

After PWHT of 2 hours at 1250F (676°C), cracking occurred at or near the dissimilar interface. Quasi-cleavage was observed in the HAZ of the forging near the interface (Figure 14a) and cleavage was observed in the mixed region (MR), immediately adjacent to the fusion boundary. A large patch of micro-void coalescence was found in the mid-section along the notch (corresponding to the middle region of the rotary friction weld) of the specimens. The corresponding J_R curves were flat and plateau at 20N/mm, which represents very brittle behaviour.

A longer PWHT (eg 10 hours at 1250F (676°C)) resulted in fracture by 'flat' cleavage predominantly in the MR near the interface (Figure 14b) and faceted cleavage in mid-section along the notch (eg middle of the rotary friction welding). The environmental performance was slightly improved and the corresponding J_R curve 'plateaued' at about 38N/mm.

Fig. 14. SEM images showing the fracture morphologies from examples of the friction welds:

A rising J_R curve, was found for friction welds postweld heat treated for 10 hours at 1100F. This shape of J_R curve is typical of more 'ductile' behaviour associated with the growth and coalescence of micro-voids, where J increases with crack growth. Significant variation was observed for the two specimens postweld heat treated 10 hours at 1100F (R-curves separated by about 80 N/mm), but resistance to crack growth was, in any case, better than any other specimens tested in the present programme. Post-fracture examination confirmed the presence of a higher proportion of micro-void coalescence, together with flat and featureless 'disbonding' and patches of quasi-cleavage in the HAZ.

 

7. Discussion

7.1 Fundamental goals for improving performance:

Ultimately, the goal of the present work is to contribute to the development of modified fabrication procedures that give better resistance to hydrogen embrittlement in subsea dissimilar joints. The starting point is, of course, the performance of a typical commercially produced weld by the hot wire GTA weld presented in the present work.

Three brittle fracture morphologies were observed in the commercially produced welds in the present work, intimately linked with two key microstructural zones which evidently show susceptibility to hydrogen embrittlement. Thus, modification of both of these zones might, potentially, raise performance levels. The three main fracture morphologies are discussed further in the following text.

Firstly, Zone M cracking was observed directly at the dissimilar interface in 'swirls' of diluted steel penetrating into the weld metal and showing a highly diluted chemistry (promoting martensite formation). These swirls can be extensive because of the difference in melting temperature between alloy 625 (~1350°C) and steel (~1500°C). Both 'interpass swirls' and 'intrapass swirls' form, in the first layer of butter deposit: the former at interpass positions, and the latter anywhere at the dissimilar interface of the first pass butter beads, where penetration pertubations have occurred during welding. In both cases, a highly diluted (by the steel forging) chemistry can result, giving the potential for virgin martensite remaining after PWHT.

Virgin martensite will not only be susceptible to hydrogen embrittlement, but might also act to promote hydrogen ingress. Zone M cracking was also observed, albeit at higher stress intensities, in specimens that were tested in air without external hydrogen charging (via cathodic protection). The presence of martensite in Zone M is also a potential issue for tolerance to post-butter fabrication steps. For example, any undesirably high stresses as a consequence of the completion weld or during installation might initiate cracking in Zone M, which might act as subsequent initiation site for hydrogen embrittlement cracking in service. Thus, minimising the formation of swirls, and their associated martensite, should improve performance of the joints and at least reduce the risk of cracking during installation, under moderate plastic strain. Work on high temperature hydrogen service claddings has come to similar conclusions.^{[3, 8-9]}

Secondly, 'terraced cleavage' was found in Zone Φ of the PMZ. The terraced morphology can be explained by the cleavage being restricted to Zone Φ with significantly tougher (in the presence of hydrogen) material being present further into the butter deposit. Cleavage in Zone Φ can be explained, despite its austenitic stainless steel-like chemistry and structure, by its high carbon content, firstly as a result of dilution with the steel forging and secondly as a consequence of carbon pile-up in this zone during PWHT. For both carbon diffusion during PWHT and hydrogen diffusion under hydrogen charging conditions, diffusion and solubility effects are similar. In the steel forging, diffusion is very fast but solubility is low. However, in Zone Φ and the neighbouring butter material, diffusion is slow but solubility is high [10]. Consequently, carbon/hydrogen diffuses 'quickly' into Zone Φ, but subsequent diffusion into the bulk of the butter is much more slow, giving a pile up in Zone Φ. The consequence of increased carbon and hydrogen contents is greater susceptibility to hydrogen embrittlement.

Finally, disbonding at the dissimilar interface was noted in the present work, which has been observed historically in vessels for high temperature hydrogen service, specifically at the interface between austenitic claddings and low alloy Cr-Mo steels. Here, rather than hydrogen charging through cathodic protection, thermal hydrogen charging is experienced and cracking is a risk upon cooling during shutdowns. Asami and Sakai first coined the concept of three types of cracking at various positions across the dissimilar interface.^[4] They defined Type I cracking as occurring close to the interface itself and it is this fracture morphology that is closest to the disbonding found in the present work.

7.2 Effect of weld parameters:

Welding parameters affected the chemistry of Zone Φ and consequently the cracking in this region. Whilst most of the laboratory welds exhibited similar characteristics to the commercially produced welds, the lower heat input pulsed welding parameters resulted in significant alteration in observed microstructure. The extent of the featureless zone of the PMZ in the mid-bead position was minimised or even eliminated and microhardness across the dissimilar interface showed no significant peak. It would appear that the chemistry of the PMZ (which was neither the most nor the least diluted) together with the cooling rate, minimised the development of any planar solidification and thus the formation of the featureless band. Lower heat input resulted in a higher Ni content in Zone Φ immediately adjacent to the fusion boundary, less susceptible to hydrogen embrittlement than Fe-rich Zone Φ. Additionally, during PWHT the formation of a supersaturated solid solution is inhibited since carbon can diffuse readily to a regular array of Nb/Mo rich segregants developed during dendritic solidification, forming carbides.

The formation of interpass position swirls was promoted when using pulsed parameters and intra-pass position swirls were exacerbated by using constant current conditions. Overall, the formation of swirls was facilitated by the lower heat input parameters. The use of constant current parameters minimised the perturbations caused during welding and led to the formation of fewer intrapass swirls Constant current parameters resulted in fewer intra-pass swirls, although their presence was not eliminated entirely, and lower heat input tended to promote the formation of inter-pass swirls exhibiting martensite laths (Zone M). Increasing the heat input seems to be the best way to minimize the formation of swirls and mitigate the formation of martensite region. The complete elimination of swirls seems unlikely. However, moving to a lower carbon forging would seem to offer the most practical means of reducing the consequence of martensite in swirl positions, both by reducing hardness and ensuring a chemistry that is tempered during PWHT.

To summarize, lower heat input would improved the susceptibility to hydrogen embrittlement by producing a less susceptible microstructure at the dissimilar interface. On the other hand, it would promote the formation of swirls of partially diluted steel penetrating into the weld metal, favourable to martensite formation. This would impair the performance of the dissimilar interface. The dissimilar combination 8630M/625 has a high susceptibility to hydrogen embrittlement and the problem cannot be eliminated by controlling the welding parameters. Employing a lower carbon content forging has important benefits as no martensite would form at intra and inter-pass position and the formation of Zone Φ could be controlled by the welding parameters.

Some work carried out at TWI demonstrated that MIG welding could produce a Ni-rich Zone Φ and its environmental performance is being established.

7.3 Effect of PWHT:

In the as-welded condition, hard zones were predominantly observed in the GCHAZ of the forging, consistent with virgin martensite (although this could be partially tempered by subsequent beads).

Hydrogen embrittlement of Zone Φ of the PMZ was observed and resulted in cracking by a brittle mode, even though this region is 'FCC'. It is believed that in this zone a hard, carbon-supersaturated zone forms during PWHT. An extended PWHT, applied on simulated welds exacerbated the carbon migration, and had the tendency of shifting the peak of carbon, and consequently the peak of hardness further into the butter deposit; ie away from the featureless region of the PMZ. This is in addition to further softening of the GCHAZ of the forging.

It is important to dissociate the effect of the microstructure/chemistry of the PMZ resulting from a choice of welding parameters, from the effect of the carbon diffusion induced by the PWHT. This is, however, difficult to achieve since GTA weld exhibit a PMZ (Zone Φ can be eliminated by an appropriate choice of welding parameters, but Zone would remain). Rotary friction welding of a solid alloy 625 directly to the steel forging achieves a more abrupt interface with no real PMZ, even though a mixed region (MR) was still observed. Thus, the effect of the carbon diffusion on the performance can be investigated isolated from any interaction with the chemistry of Zone Φ that might affect the carbon activity. Tests performed demonstrated that friction welds performance was similar to commercially produced welds that had been given a similar PWHT. The performance of the friction welds were significantly affected by the PWHT, demonstrating the final means of minimising susceptibility is through optimisation of the PWHT. Optimisation of heat treatment to result in lower carbon contents in the critical regions, whilst achieving the softening required in the steel HAZ would decrease susceptibility of the joints to hydrogen embrittlement. Work performed on clad steels for high temperature hydrogen plant^[3] has demonstrated the potential improvements to be had from modification of PWHT procedure. Again, reducing the carbon content of the forging would be of benefit by reducing the peak carbon content in the PMZ (by lowering the activity gradient). The effect of PWHT on commercially welds is currently being investigated.

Appropriate qualification of fabrication procedures (on environmental performance) is necessary to give sufficient confidence in MIG or friction welding processes, since the relationship between chemistry, fabrication parameters, microstructure and performance is complex. Work is ongoing^[1] to define acceptance criteria and provide guidance on demonstrated solutions and enable confident design of robust dissimilar joints for subsea service.

8. Concluding remarks

The following conclusions are drawn:

1. The hydrogen embrittlement resistance of subsea dissimilar joints depends on three primary factors: (i) material selection, (ii) butter welding procedure, (iii) postweld heat treatment. Each of these governs the microstructure and chemistry across the dissimilar interface and unless specifically modified to take hydrogen embrittlement into account, can lead to a joint susceptible to environmental failure.
2. Two microstructural zones dominate fracture path in environmental tests: Zone Φ; a hard, carbon-supersaturated, austenitic, solid solution immediately adjacent to the fusion boundary, which fails by cleavage; and zone M, which contains martensite in highly-diluted weld material particularly at swirls of diluted steel penetrating into the butter deposit.
3. The present work demonstrated the difficulty of eliminating both Zone Φ and Zone M, simultaneously in arc welded butter deposits. Zone Φ could be eliminated by employing a low arc energy. However, Zone M, could only be limited by employing high arc energy. Thus, the elimination of a susceptible microstructure at the dissimilar interface, at least between 8630M and alloy 625 butter deposits, could not be prevented by modifying welding parameters alone.
4. Hydrogen embrittlement resistance was improved by using low temperature PWHT, limiting the carbon-supersaturation of Zone Φ.
5. The best environmental performance was obtained for a friction welded dissimilar joint given a PWHT of 10 hours at 1100F, achieved by a combination of the elimination of Zones Φ and M, and the mitigation of carbon diffusion during PWHT.

10. Acknowledgments

The work was funded by Industrial Members of TWI as part of the Core Research Programme. The authors acknowledge the contribution of Peter Sketchley, David Seaman, Jerry Godden, Sheila Stevens, Briony Holmes and Henryk Pisarski.

11. References

1. Beaugrand VCM, Smith LS and Gittos MF, 22-26 March, Atlanta, Corrosion 2009, NACE International.
2. Perng T and Altstetter C, ASTM STP 962, 1988.
3. Gittos MF, 16-20 March 2008, New Orleans, Corrosion 2008, NACE International.
4. Asami K and Sakai T, Trans. of the Iron and Steel Institute of Japan, Vol 21, issue 6, 1981.
5. ASTM STP 668.
6. ASTM STP 527.
7. Anderson TL, Fracture mechanics 3d ed., 2005, CRC press, Taylor and Francis group.
8. Wang Z, Xu B and Ye C, 1993, Welding Journal Research Supplement, August, p.398-s.
9. Gnriss G, 2001, Conference IIW/IIS Ljubljana, July.
10. Baboian R, Corrosion tests and standards, 2nd ed., ASTM, 2005.