Intergranular Stress Corrosion Cracking of Welded Steel

Understanding and Avoiding Intergranular Stress Corrosion Cracking of Welded Supermartensitic Stainless Steel

Paul Woollin
TWI Ltd,
Granta Park,
Great Abington,
Cambridge, CB1 6AL, UK

Paper presented at Corrosion 2007, Paper 07094, Nashville, Tennessee, 11-15 March 2007.

Abstract

Supermartensitic stainless steels (SMSS), including both lean grades (without molybdenum) and high grades (with 2% molybdenum), have shown sensitivity to intergranular corrosion or stress corrosion cracking (IGSCC) in weld HAZs during laboratory testing in chloride solutions acidified with CO ₂ at above about 100°C. ^[1-9] Also, lean SMSS grades have cracked in service when exposed to hot acidic brines. ^[4,6] This paper describes a Joint Industry Project designed to test existing understanding of the mechanism of sensitisation to IGSCC of welded SMSS. The project also sought to determine whether IGSCC could be reliably avoided via control of the welding process or by brief PWHT. Existing data were reviewed and two series of welding trials and corrosion tests were undertaken. The review and first series of tests allowed a hypothesis of the mechanism of sensitisation in SMSS to be developed, which was then tested via the second series of tests. Several SMSS grades were examined. Welds were made with a range of welding thermal cycles, by varying the heat input and interpass temperature and by use of active water cooling. Some welds were given PWHT at 650°C for 5 minutes. Stressed bend specimens from the welds were tested in 25% NaCl solution acidified with 10bar CO ₂ at 110°C. Some tests included a small addition of HCl, to acidify the solution to pH 3.3 prior to introduction of CO ₂ . From the results of the tests and measurements of the weld thermal cycles, it was concluded that a complete understanding of the mechanism of IGSCC in supermartensitic steels does not yet exist, although the main contributing factors are known. Consequently, where the service environment is sufficiently aggressive, the risk of IGSCC cannot be entirely eliminated by control of the welding process alone, although brief PWHT offers a practical means of avoiding IGSCC.

Introduction

A review of published information indicated that many but not all girth welds in SMSS, in the as-welded condition, are susceptible to stress corrosion cracking in hot acidic brines, by an intergranular mechanism following prior austenite grain boundaries. ^[1-6] The crack appearance suggests that sensitisation occurs at specific HAZ locations rather than throughout the HAZ. Specimens with a machined root or ones given brief PWHT at 650°C for 5 minutes were not found to be susceptible to cracking. ^[7] No clear links to HAZ hardness, HAZ microstructure or composition were found for the alloys tested and it was not clear how widely applicable the beneficial effect of PWHT was.

The most commonly observed macroscopic cracking mechanism follows prior austenite boundaries, therefore a hypothesis was developed which proposed that carbide precipitation on these boundaries provides the main mechanism for sensitisation. At least two thermal cycles are required, with the first thermal cycle forming fresh martensite and the second and subsequent cycles forming the damaging carbides.

Therefore, a series of welds was made with different heat inputs, and corrosion tests performed, to see if a distinction could be made between welds that were susceptible to IGSCC and those which were not, on the basis of welding procedure. Following this, a second series of welds was produced to further explore which aspects of welding procedure were critical with respect to sensitisation to IGSCC.

Phase I experimental programme

A series of GTA welds was produced, with a range of heat input, using Ar shielding and back purge gas, in three steels: a lean grade without Ti addition (steel C: 11Cr1.5NiCu), a high grade without Ti addition (steel D: 12Cr6Ni2Mo) and a high grade with Ti addition (steel A: 12Cr6.5Ni2.5MoTi), Table 1. Superduplex (Zeron 100X) and 22%Cr duplex (ER2209) filler wires were used for the high grade and lean grade steels respectively. For experimental convenience, fairly small pieces of steel ~150 x 200mm with thicknesses of 11-18mm, were used and single, double and multi-pass welds were produced. A maximum interpass temperature (IPT) of 150°C was specified to reflect typical industrial practice. The HAZ thermal cycles were recorded by thermocouples attached to the root surface. Details of the welding procedures used are given in Table 2a.

Table 1 Chemical composition of pipes used.

Table 2a Welding parameters used to prepare the mechanised GTA welds. No PWHT was used.

Table 2b STT GMA/PGMA girth welding parameters. Parts of welds NWA1, 2, 3 and NWB1, 2, 3 were given PWHT at 650°C for 5 minutes, and subsequently identified with a letter 'P'.

Four point bend tests on cross-weld specimens (100 x 15 x 3mm) from each weld root, with the weld root shape left as-welded, were performed in autoclaves at 110°C. The testing solution was 25% NaCl acidified to a calculated pH of 3.4 with 10 bar CO ₂ . The tests were performed for either 30 or 72 days. Specimens were deflected with the root in tension to give a strain equivalent to 100% of the parent material 0.2% proof stress measured at room temperature in the HAZ, i.e. a total strain of 0.57%.

Phase I results and discussion

It was found that none of the welds were susceptible to macroscopic IGSCC, although some very shallow corrosion/cracking, with an intergranular character, was found in some cases, Table 3a. One of the welds, WC-3 made in lean grade steel with a heat input of 1.4kJ/mm, and two passes with 86°C interpass temperature, showed evidence of microscopic IGSCC. Two welds showed features that were borderline, ie about 0.025mm deep. These were WD-7 in high grade steel, with approximately 1.6kJ/mm heat input, 15 passes and 40-147°C interpass temperature, and WC-5 in lean grade steel, made with two passes of 1.4 and 1.0kJ/mm and 85°C interpass temperature. Most other welds in the high grade steel showed very shallow penetrations along grain boundaries, with depth of <0.025mm.

Table 3a Summary of corrosion test results for high grade pipes (D&D2, 12Cr6Ni2Mo).

Examination of the thermal cycles revealed that fairly high interpass temperatures, in the range 80-150°C, had been used, especially for the multi-pass welds. This was caused by the small sample sizes, which gave fairly slow cooling and hence the welders typically had to wait some time for IPT to fall below 150°C. Also the HAZ cooling rates were fairly slow with times between 600 and 400°C (the expected approximate temperature range for damaging chromium carbide formation) of 30-45 seconds within about 2mm of the fusionline. Comparison of this work with the previous observation of IGSCC in pulsed GMA girth welds, made with heat input of about 0.5kJ/mm and a specified IPT of <150°C, in a variety of supermartensitic pipes ^[7] led to the development of a further hypothesis that, for pipe girth welds, rapid thermal cycles, ie low heat input and low interpass temperature, were likely to be most damaging with respect to IGSCC susceptibility. In additions, it was proposed that slower cooling allows an opportunity for 'healing', ie back diffusion of chromium into Cr-depleted zones formed adjacent to chromium carbides, whereas faster cooling does not. Typical arc welding thermal cycles for material geometries of the type examined here are probably, therefore, often sufficient to cause sensitisation to IGSCC but that 'healing' may readily occur, if this hypothesis is correct.

Phase II experimental programme

In order to test the above hypothesis, a programme of work, designed to induce shorter thermal cycles and to test the IGSCC susceptibility of the resulting welds, was initiated. This second piece of work looked at similar steels to the first programme but in larger pieces representative of pipe girth welds. The aim was to demonstrate that more rapid cooling gave welds that were more sensitive to IGSCC and also to look at the effect of brief PWHT at 650°C for 5 minutes on sensitivity to IGSCC in welds made with the shortest thermal cycles, which were expected to be the ones that would otherwise show IGSCC.

Two further pipe materials designated D2 (a high grade 12Cr6Ni2Mo, 18mm thick) and C2 (a lean grade type 11Cr1.5NiCu, 8mm thick), both without deliberate alloying with titanium, were studied, Table 1. A superduplex consumable (Zeron 100X) was used to weld the high-grade pipe (to produce a series of 'new welds A' or NWA), while the lean-grade pipe was welded employing a type ER2209 22% Cr duplex filler wire (NWB).

In each pipe, girth welds were made with as low a heat input as practically possible (0.3kJ/mm in the root, rising to 0.9kJ/mm in the cap), using an STT (surface tension transfer) GMA process for the root run and a pulsed GMA process for the second and fill passes, Table 2b. The STT-GMA process was selected as it gave a good, reproducible weld root bead profile without need for a backing strip. The shielding gases used were: (i) Ar-20%CO ₂ -2%O ₂ for the root pass in the lean grade pipe C2, (ii) Ar-38%He-2%CO ₂ for the remaining passes in lean grade pipe C2 and (iii) He-13.5%Ar-1.5% CO ₂ for the high grade pipe D2. Single pass, two pass and multi-pass welds were produced, with the pipe rotated (PA position) to keep heat input and bead profile as consistent as possible. The length of the pipe samples on each side of the weld was about 500mm. The weld root was protected from oxidation by use of argon purging in all cases. The weld HAZ thermal cycles on the pipe inner surface were recorded using thermocouples positioned in the HAZ typically at 5mm and 7mm from the joint centreline. Active cooling of the pipe using water sprayed onto the outer surface of the pipe about 80mm behind the weld torch was used for one series of weld, to increase the cooling rate. Water cooling could not be used for the root pass, as the water would have disturbed the welding. Two interpass temperatures were employed, namely 20°C and 140°C for the welds made without forced cooling, Table 2b.

A part of each weld obtained using low interpass temperature and forced cooling (NWA2-3 and NWB2-3, Table 2b), which, if the hypothesis is correct, were expected to be the welds with greatest sensitivity to IGSCC, was subject to PWHT at 650±10°C, as measured by thermocouples on the root, for 5 minutes, by induction heating. Also, the single pass welds were given PWHT, to check that it did not induce sensitisation to IGSCC in the as-formed martensite HAZ.

Four point bend tests on cross-weld specimens (100 x 15 x 3mm) from each weld, with the weld root shape left as-welded, were performed at 110°C. The testing solution was 25% NaCl acidified with 10 bar CO ₂ . The tests were performed for 90 days. The test solution was acidified to a pH of 3.3 by a small addition of HCl prior to test. The calculated pH during test was 3.2. Specimens were deflected to give a total HAZ strain of 0.57%.

Phase II results

Thermal cycles

The thermal cycle data showed the expected general trends of lower cooling rates for the higher interpass temperatures and higher cooling rates and lower peak temperatures when water cooling was used. Cooling rates were typically faster for a given peak temperature for the high grade steel, as it was substantially thicker than the lean grade (18mm versus 8mm) but this trend was not always observed clearly, as the highest peak temperatures were measured in the thinner, lean grade steel. This may have been an accident of the thermocouple location or a reflection of overall slower cooling in this material. When comparing welds in the high grade steel made with and without water cooling, maximum cooling rates of 23 and 4°C/s respectively were recorded for the high grade steel (NWA2/3 and NWA4) and 100 and 10°C/s (NWB2/3 and NWB4) for the lean grade. It may be noted that the highest second pass HAZ cooling rate of all welds (200°C/s for NWB5) was associated with the highest measured temperature (959°C). This is as expected and in general it is more suitable to compare cooling rates at a particular temperature of relevance to the phenomenon being studied, e.g. 600-400°C. This was not possible here, as the measured HAZ temperatures were typically fairly low due to the low heat input. It was found to be difficult to give an accurate comparison of the thermal cycles in the various welds due to the slightly different thermocouple positions with respect to the various welds and the different levels of reheating, which gave markedly different peak HAZ temperatures, presumably depending on the thicknesses of individual weld beads.

Figure 1 gives a comparison of the HAZ thermal cycles measured at about 5mm from the weld fusion line for GTA (WC3) and PGMA (NWB5) welds in lean grade steel. It can be seen that the peak temperatures were substantially higher and the cooling rates substantially slower in the GTA welds. The heat inputs for the GTA and PGMA welds shown were 1.4 and 0.4-0.9kJ/mm respectively and the interpass temperatures were room temperature and 86°C.

Weld characterisation

Metallographic sections revealed broadly similar microstructures in all cases. Each weld HAZ in the high grade steel had a well defined region with retained delta ferrite, containing between 5 and 10% ferrite locally, in a band about 100µm wide and 100-200µm from the fusionline, Figure 2. Between this band and the fusionline, a lower level of ferrite was present, again apparently on prior austenite boundaries but here the prior austenite structure is typically Widmanstatten. In the lean grade steel, the ferrite was only clearly defined in samples that had been given PWHT and was also less regular in distribution; some areas showed little or no ferrite, whereas other areas had a band of ferrite up to about 400µm wide, extending to the fusionline. Except for this ferrite, the as-welded HAZ microstructures consisted of fairly featureless martensite.

Corrosion test results

A summary of the results of the Phase II IGSCC tests is presented in Table 3b and 3c. The results are categorised as 'A': macroscopic cracking, 'B': microscopic cracking, 'C': very shallow intergranular corrosion, 'G': general corrosion or 'N': no corrosion. Figure 3 shows examples of features found on the sections taken through specimens after test.

Table 3b Summary of IGSCC corrosion test results for high grade pipe with Ti (A, 12Cr6.5Ni2.5MoTi).

Table 3c Summary of IGSCC corrosion test results for lean grade pipes (C&C2, 11Cr1.5NiCu).

Spalling of thick oxide in some areas of the HAZ and evidence of uniform corrosion was observed on lean-grade samples. This indicates that general corrosion had occurred on the lean grade material, to a depth of about 0.05-0.1mm, i.e. a corrosion rate of 0.2-0.4mm/year. Hence, it is concluded that the test environment was too aggressive than ideal for studying IGSCC in the lean grade material as general corrosion is likely to suppress IGSCC.

Visual and metallographic examination of the IGSCC test specimens showed macroscopic cracking with intergranular morphology in just two cases: (i) two-pass weld NWA2 made with forced cooling and 20°C interpass temperature in high grade steel (the highest cooling rate for the second pass of all welds), and (ii) two pass weld NWB5 made without forced cooling and with 20°C interpass temperature in lean grade steel. Microscopic intergranular cracking was found in several other welds, ie (i) a weld in lean grade steel, NWB6, consisting of two passes and made with 140°C interpass temperature, and (ii) five welds in high grade steel, NWA3, NWA5, NWA6 and NWA7, ie all welds except single pass weld NWA1 and NWA4, which was made with two passes and 20°C interpass temperature. It was noted that in three cases for high grade steel, NWA 5, 6 and 7, microscopic cracking was associated with ferrite in the HAZ at the weld toe, whereas the other cases showed microscopic cracking further from the weld toe (up to 1.5mm away). These three welds included two with five weld passes (20°C and 140°C interpass temperatures) and one with two passes (140°C interpass temperature), which would have had the slowest cooling and most re-heating of all the welds produced. No cracking, either macroscopic or microscopic, was found in the single pass welds.

The selection of welds for PWHT included those with most rapid cooling, which the hypothesis suggested would be the most damaging condition with respect to IGSCC. For the lean grade steel, this turned out to be incorrect. Nevertheless, none of the lean grade samples given PWHT showed any evidence of any intergranular corrosion or cracking in the 90 day test. Also PWHT of single pass welds, which would not be expected to sensitise during welding, did not induce sensitivity to IGSCC.

For the high grade steel, the weld with most rapid cooling was the one that gave macroscopic IGSCC. In this case, after PWHT, some 0.025mm deep intergranular features formed during the 90 day test, at a similar location to where macroscopic IGSCC formed in the as-welded condition, ie about 1.5mm from the weld toe. Macroscopic IGSCC did not occur in the PWHT sample. It is not clear if these microcracks are a cause for concern but they do indicate a need for further work to establish the robustness of the proposed PWHT when applied to the welds with greatest sensitivity to IGSCC. Neither of the other two PWHT samples showed any microscopic or macroscopic IGSCC, although one of the welds, NWA3, showed microscopic IGSCC in the as-welded condition.

Phase II discussion

As-welded condition

The results for the lean grade steel indicate that a fairly specific second thermal cycle is required for macroscopic IGSCC to occur but in this case it was not the shortest thermal cycle. Cracking did not occur when forced cooling was applied nor when 140°C IPT was employed nor in the single pass weld. Also, cracking did not occur in a broadly similar two-pass weld which was reheated to lower temperatures by the second pass. Although not quite as anticipated, this however does fit the hypothesis, if it is proposed that there is a thermal cycle 'window' in which sensitisation occurs. If the second cycle is too short, Cr-carbides and associated Cr-depleted zones cannot occur, and if it is too long, 'healing' occurs. However, it is noted that high temperatures were experienced during reheating of the HAZ in NWB5 and hence cracking apparently occurred in an area reheated to above 900°C. This indicates that full transformation to austenite did not occur during this short cycle even though the temperature exceeded by a considerable margin the typical published values of Ac1 for SMSS, ie 550-700°C. However, these values are typically measured with a slow heating rate, eg 10°C/s. ^[2] This is consistent with published data on the effect of heating rate on transformation in SMSS, which show that Ac1 may reach 850°C and Ac3 may reach 980°C during welding with a heating rate of 370°C/s. ^[8]

The results for the high grade steel fit the hypothesis fairly well, as the only weld showing macroscopic IGC on prior austenitic boundaries had the shortest second thermal cycle due to low heat-input, active cooling and an IPT of 20°C. No cracking occurred in a single pass weld and PWHT prevented formation of macroscopic IGC. However, a second phenomena is apparent; microscopic IGSCC developed in several welds made with more protracted thermal cycles and an IPT of 20°C or 140°C. This microscopic cracking was at the weld toe, where a band of ferrite existed, suggesting a second mechanism of sensitisation, associated with Cr-depleted zone formation as a result of carbide precipitation on ferrite-martensite boundaries, apparently requiring a longer thermal cycle and not being sensitive to IPT. This is consistent with formation of the delta ferrite phase boundaries at very high temperature (>1300°C) and the fact that different kinetics of precipitation could reasonably be expected for prior austenite boundaries and delta ferrite/martensite phase boundaries. One reason why such cracks were microscopic could be that the delta ferrite does not form a continuous network, whereas prior austenite boundaries are continuous. It is not clear whether such micro-cracks would extend over a longer period.

The work performed has shown an effect of a change in welding process from GTA to pulsed GMA, leading to macroscopic IGSCC in some of the PGMA welds but not in the GTA welds. The primary reason for the change of welding process was to shorten the thermal cycle, via reduced heat input, although reduced interpass temperature and active cooling were also used, and inevitably some change of weld root geometry and surface condition would have arisen, although efforts were made to minimise these. This provides evidence that shortening the weld thermal cycle in the context of arc welding encourages IGSCC, which supports a classical Cr-carbide sensitisation mechanism for both lean and high grades, although effects of weld surface condition and geometry cannot be ruled out.

The effect of PWHT at 650°C for 5 minutes

Postweld heat treatment has been reported to have a beneficial effect on the resistance of SMSS girth welds to IGSCC by several authors. ^[1,7] However, testing used to demonstrate this was of a fairly short, 30-day duration. The 90-day tests here caused one specimen to form microcracks even after PWHT. This supports reservations expressed by one end user with a 30 day test duration for qualification of SMSS for sweet service, as previous 30-day tests have not caused any suspicious features to form in welds given PWHT. ^[4] Further work would be required to demonstrate whether the 90-day test duration used here was responsible for the microcracking of this specimen after PWHT, whether it was because the weld was still sensitised to some degree or whether it was due to oxidation and would not propagate. In the present work, one PWHT cycle has been examined on girth welds, namely a nominal 650°C for five minutes. This is similar to the thermal cycle used by other authors but little work has been performed to demonstrate that this is the most appropriate thermal cycle and indeed whether the PWHT should vary for different steels.

Assuming that the Cr-carbide precipitation theory of sensitisation of SMSS to IGSCC is correct, the most likely mechanism by which PWHT is effective in eliminating sensitivity to IGSCC is by allowing chromium back-diffusion into the chromium-depleted zones. The chromium-depleted zone width has been estimated to be up to 20nm in lean grade material but may be <5nm in steel with about 6%Ni and 2%Mo. ^[5,6] Hence, in order for PWHT to be effective, the time and temperature must be sufficient for chromium to diffuse over a distance of this magnitude. Use of a simple x= √Dt calculation, based on published matrix diffusion coefficients in the range 4.9x10 ^-14 to 1.5x10 ^-13 cm ² s ^-1 for chromium in iron with 10-20%Cr, extrapolated from higher temperature data, ^[10] which relate to an austenitic microstructure, indicates a diffusion distance of about 40-70nm for five minutes at 650°C. Higher diffusion rates would be expected in the martensite and ferrite phases. Hence, this very simple calculation supports the proposed Cr-diffusion explanation of the effect of PWHT on eliminating sensitisation to IGSCC.

Practical implications

The likelihood of intergranular SCC occurring in as-welded supermartensitic steel in service will depend on a combination of the degree of sensitisation of the HAZ and the severity of the environment. Very aggressive environments will tend to cause general corrosion rather than intergranular SCC and more benign environments may cause neither.

One consequence of the proposed mechanism of sensitisation of prior austenite boundaries is that although higher heat input appears to have been beneficial, particularly for high grade steel, as it has apparently allowed 'healing', this would not apply if the heat input of any pass was sufficiently high to re-transform the root surface to austenite, and hence form fresh martensite on cooling. In this latter case, the material would then have been returned to a condition in which it may be sensitised by subsequent passes. Significant variation of heat input between passes may therefore contribute to sensitisation. For this reason, manual welding probably cannot be relied upon to give welds that are reliably free from sensitisation to IGSCC. In this respect, it is noted that when lean grade pipe failed by IGSCC in service, it typically coincided with the side of the weld on which the final capping pass was applied. ^[4] The lean grade steel used here was taken from such a failed pipe and it was found that heating of the HAZ to >Ac ₁ , would have occurred for several passes. The capping pass probably reheated material into the critical range for sensitisation.

In principle, it should be possible to select welding parameters that produce welds that are not sensitive to IGSCC, based on a knowledge of the critical thermal cycles that must be avoided. This work has hinted at the damaging thermal cycles, and published data for simulated HAZ microstructures give additional values of time and temperature for developing sensitivity to IGSCC. ^[11] However, the wide range of thermal cycles experienced in a multi-pass weld HAZ and the difficulty of accurately measuring or predicting them, imply that reliably avoiding IGSCC by quantitative means will be very difficult, if not impossible. An empirical approach, based on trying to avoid rapid cooling of the root during second and subsequent passes, superficially offers the best chance of producing as-welded welds that are not sensitive to IGSCC although sensitisation of the ferrite seems to occur for longer thermal cycles. However, even if such welds were produced and qualification corrosion tests showed acceptable performance, the lack of understanding of the range of applicability of the result, ie what range of heat input and interpass temperature would be acceptable in production welding, make this approach unreliable. Manual welding would presumably be more variable and have less chance of reliably producing acceptable welds than mechanised welding.

Surface condition and weld geometry probably contribute to susceptibility to intergranular corrosion/stress corrosion but there are no data to confirm this conclusively at present. In practice, efforts to achieve low oxidation levels and a smooth weld toe are considered prudent, to reduce the likelihood of initiating IGSCC.

It is recommended that PWHT should be applied to welds in supermartensitic stainless steel where there is a risk of intergranular SCC in service, ie in some hot acidic environments, provided that an appropriate qualification process has been undertaken. It may be noted that the environmental conditions required for IGSCC of lean and high grade steels have not been defined. Due to the limited information available, the use of welded supermartensitic stainless steel in the PWHT condition will require qualification on a case by case basis. There are insufficient data to conclude whether 90-day corrosion testing is required but it is advisable. The qualification process should consider the extremes of the range of PWHT thermal cycles that may be experienced, as the acceptable range has not been established. Qualification testing should address the effects of PWHT on toughness and sour service performance, in addition to IGSCC. Based on current data, a PWHT temperature of 620-660°C at the root is recommended and the heat treated zone should encompass the whole of the weld metal and HAZ. The maximum allowable cap temperature has not been established but the current work extended up to about 670°C. Heating and cooling should be fairly rapid. The most appropriate PWHT duration has not been established but there is fairly common agreement that 5 minutes is an appropriate duration.

Conclusions

• The results obtained do not provide a complete understanding of the mechanism of sensitisation to intergranular stress corrosion cracking in welded SMSS but are consistent with a mechanism based on formation of Cr-carbides and associated Cr-depleted zones.
• The most commonly observed macroscopic cracking mechanism follows prior austenite boundaries, suggesting that carbide precipitation on these boundaries provides the main mechanism for sensitisation. At least two thermal cycles are required, with the first thermal cycle forming fresh martensite and the second and subsequent cycles forming the damaging carbides. In the context of arc welding of pipes of the size examined here, fairly short reheating thermal cycles are apparently more damaging with respect to IGSCC on prior austenite boundaries than long thermal cycles. Higher interpass temperatures (about 140-150°C) seem to inhibit this form of sensitisation.
• Several welds developed microscopic IGSCC, some of which was at the weld toe and coincident with the band of retained delta ferrite at that location. This suggests that a second sensitisation mechanism might exist associated with carbide precipitation on delta ferrite/martensite phase boundaries. It is not clear whether this mechanism also requires two or more thermal cycles although it was apparently favoured by longer thermal cycles and greater numbers of welding passes.
• The significance of shallow cracks developed during tests of up to 90 days duration has not been established. They might be related to localised sensitisation caused by oxidation of the surface, in which case they will not propagate, or they might be associated with a microstructure with a low level of sensitisation or a discontinuous sensitised microstructure, in which case they might continue to propagate over longer periods.
• The results indicate that control of the welding process to reliably avoid sensitisation to IGSCC is not straightforward. Each steel will have its own range of critical thermal cycles for causing sensitisation. Mechanised welding giving reproducible thermal cycles for each weld pass provides a higher likelihood of avoiding sensitisation to IGSCC than manual welding, where a wider range of thermal cycles will inevitably result.
• Welds in supermartensitic stainless steel that experience more than one thermal cycle during welding should be given a brief PWHT at about 650°C for five minutes at the root if the environment that they will be exposed to is sufficiently aggressive to cause IGSCC of a sensitised HAZ microstructure.

Acknowledgements

The following organisations are thanked for sponsoring the work, for their technical input and for supply of materials: BP, ConocoPhillips, JFE Steel, Shell, Statoil, Sumitomo Metal Industries and TenarisNKK.

References

1. Coudreuse L, Ligier V, Lojewski Ch, Toussaint P, 'Environmental induced cracking (SSC and SCC) in supermartensitic steels (SMSS), Proc Conf, Supermartensitic Stainless Steels 2002, Belgium, KCI Publishing, pp163-172.
2. Woollin P and Carrouge D: 'Heat affected zone microstructure in supermartensitic stainless steels', Proc Conf Supermartensitic Stainless Steels 2002, Belgium, KCI Publishing, pp199-204.
3. Folkhard E, 'Welding metallurgy of stainless steels', Springer Verlag, 1988.
4. van Gestel W: 'Girth weld failures in 13Cr sweet wet gas flow lines', Proc conf Corrosion 2004, USA, NACE International, paper 04141.
5. Ladanova E and Solberg J K: 'Transmission electron microscopy investigation of precipitation reactions in coarse-grained heat affected zone in two 13%Cr supermartensitic stainless steels', Proc Conf, Supermartensitic Stainless Steels 2002, Belgium, KCI publishing, pp205-214.
6. Gregori A, Woollin P, van Gestel W: 'Analysis of in-service failures in girth-welded supermartensitic pipes', Proc conf Stainless Steel World 2003, Netherlands, KCI Publishing, p195.
7. Woollin P, 'Postweld heat treatment to avoid intergranular stress corrosion cracking of supermartensitic stainless steels', Proc conf 'Stainless Steel World 2005', KCI Publishing, Paper No. 5045.
8. Lippold J and Alexandrov B, 'Phase transformations during welding and postweld heat treatment of a 12Cr-6.5Ni-2.5Mo supermartensitic stainless steel', Proc conf 'Stainless Steel World 2004', KCI Publishing, paper P0447.
9. Amaya H, Kondo K, Taniyama A, Sagara M, Ogawa K, Murase T, Hirata H, Takabe H and Ueda, M: 'Stress corrosion cracking sensitivity of supermartensitic stainless steels in high chloride concentration environment', Proc conf Corrosion 2004, USA, NACE International, paper 04124.
10. Smithells Metals Reference Book, 6th Edition, Brandes, 1983.
11. Miyata Y, Kimura M, Nakamichi H, Sato K, Itakura N and Masamura K: 'Mechanism of intergranular stress corrosion cracking in HAZ for supermartensitic stainless steel', Proc conf Eurocorr '04, Nice, France.