P Woollin* and D Carrouge**
* TWI Ltd, Cambridge
** University of Cambridge, Department of Materials Science & Metallurgy
Paper presented at conference on Supermartensitic stainless steels, Brussels, Belgium, 3 - 4 October 2002
The use of supermartensitic stainless steels has expanded rapidly in recent years but characterisation of weld heat affected zone (HAZ) microstructure has been the subject of only limited study. Consequently, the current understanding of HAZs is limited, particularly with respect to corrosion, stress corrosion cracking and sulphide stress cracking behaviour. The paper describes the microstructures in the various regions of a supermartensitic stainless steel HAZ and considers effects of re-heating from subsequent passes and PWHT.
Supermartensitic stainless steels offer a combination of high strength, corrosion resistance and weldability that is particularly attractive for oil and gas flowlines. Compared to more traditional nickel-alloyed, soft martensitic stainless steels with around 13%Cr4%Ni, they have a lower carbon level (about 0.01% compared to ≥0.03%), an addition of molybdenum to improve corrosion resistance (typically 2%) and an increase of nickel content (typically to 5-6.5%). Lean supermartensitic grades, with little or no molybdenum and less nickel, are also available, as are intermediate grades.  Some grades also have a small titanium addition, to limit secondary hardening. Typical compositions are shown in Table 1.
Table 1 Chemical analyses of the range of steels examined
There is a substantial body of data indicating that crack-free welds with good HAZ toughness can be made without need for pre-heat and postweld heat treatment. [2,3] Hence, the steels are widely considered to have good 'weldability'. However, a number of HAZ corrosion and stress corrosion issues remain [4-6] and there is a need for improved understanding of the HAZ microstructure development in order to address these. Unfortunately, the HAZ microstructure is very fine scale and all relevant features cannot be fully resolved under an optical light microscope, which has hampered understanding of HAZ behaviour.
This paper reviews the current understanding of supermartensitic stainless steel HAZ microstructure and serves as an update to the paper presented by Gooch et al at the 1999 Supermartensitic Stainless Steel Conference. 
2. Parent steel microstructure
The parent steels are supplied in the tempered condition. Whilst precise details are proprietary, the steels typically contain several percent of stable revert austenite together with tempered martensite containing a population of carbides. Published literature indicates that it is possible to attain 20-30% stable austenite in 13%Cr steel with 4-6%Ni, if tempering is performed at about 100°C above Ac1. [7,8] Both chromium and molybdenum may be incorporated into carbides during tempering, together with minor elements such as V and Ti. For conventional nickel alloyed martensitic stainless steels, M2C and M7C3 (where M is a metal atom) form in the range 400 to 450°C, and contribute to secondary hardening, whilst M23C6 is dominant at above 500°C. M2C is stabilised by the presence of Mo, thus Mo alloying encourages secondary hardening on tempering. [8,9] It has been reported that tempering at 620-680°C of the coarse grained HAZ of a weld in supermartensitic stainless steel with 2.7%Mo predominantly gives molybdenum carbides. 
3. The various regions of a supermartensitic HAZ
3.1 General comments
The HAZ may be subdivided into several regions with different microstructures. [6,11,12] The sequence of transformation on heating is as follows. Below Ac1 some tempering, i.e. carbide precipitation, may occur. Above Ac1, martensite transforms to austenite and the proportion of austenite increases until the steel is fully austenitic at Ac3. Martensite remaining in this temperature interval will be tempered. There is then a single-phase austenite range until Ac4, when delta ferrite formation begins. The microstructure becomes fully ferritic at Ac5 and remains single phase until the melting point. On cooling, the resulting microstructure reflects the peak HAZ temperature attained and five distinct HAZ regions can be identified:
(i) that which was fully delta ferrite (>Ac5, which is >around 1370-1420°C, Table 2),
(ii) that which was a mixture of delta ferrite and austenite (Ac4-Ac5, which is around 1330-1370°C to 1350-1420°C),
(iii) that which was fully austenite (Ac3-Ac4, which is around 720-800°C to 1330-1370°C),
(iv) that which was a mixture of austenite and martensite (Ac1-Ac3, which is around 540-680°C to 720-800°C) and
(v) that which did not exceed Ac1 but which exceeded the minimum temperature for carbide precipitation, i.e. around 400°C.
Table 2 Transformation temperatures for three supermartensitic stainless steels, measured with a heating rate of 10°C/min
|Steel||Transformation temperature (°C)|
The characteristics of the various regions are illustrated in Figs.1-4 and discussed below. A typical microhardness distribution for a single pass TIG weld HAZ in 12Cr6Ni2Mo steel is shown in Fig.5.
3.2 HAZ that was fully delta ferrite at peak temperature
There are few published data available relating to Ac4 and Ac5 temperatures for supermartensitic stainless steels, although some values calculated using MTDATA have been reported,  indicating temperatures in the range 1240-1280°C for Ac4 and 1270-1300°C for Ac5 for a lean grade (11Cr1.5Ni) and a high alloy grade (12Cr6.5Ni2.5Mo) respectively. Subsequent measurements for the same steels have indicated that actual temperatures are somewhat higher than predicted, being1330-1370°C for Ac4 and 1350-1420°C for Ac5, Table 2, for a heating rate of 10°C/s.  These temperatures are very high and approaching the melting point. Consequently, the fully ferritic region is narrow, typically 100-200 microns for conventional TIG/MIG welding processes, Fig.1. This region exhibits grain growth due to rapid diffusion in the ferrite phase and, on subsequent cooling, the austenite retains a coarser grain size than the parent. However, identifying prior austenite and prior deltaferrite grain structures is difficult by optical light microscopy. Consequently, this region may also be referred to as the grain coarsened HAZ. On cooling, the austenite growth into the delta ferrite is predominantly Widmanstatten in nature with some allotriomorphic and intragranular austenite, [11,12] Fig.2 and 3. Austenite formation on cooling is apparently controlled by diffusion of substitutional elements as filaments of delta ferrite are retained between prior-austenite units to room temperature in the higher alloygrades and microanalysis indicates enrichment of the delta ferrite in Cr and Mo and depletion of Ni, Table 3. [11,12] Delta ferrite retention is probably a result of the low rate of back diffusion of substitutional elements into austenite during the transformation from ferrite to austenite on cooling, when compared to the opposite reaction on heating. Steels with higher levels of molybdenum and nickel retain higher volume fractions of delta ferrite than the lean alloys. The volume fraction of retained delta ferrite is small in actual HAZs, typically well below 10%, although preliminary studies have shown higher levels (exceeding 10% delta ferrite) in furnace heat treated samples and indicated a weak link to the cooling rate  Additional work is required to establish the effect of welding parameters and the resulting thermal cycle on HAZ microstructure. On cooling, ferrite is retained preferentially in pre-existing areas of segregation in the steel. Enerhaug et al reported partial melting in this region of the HAZ, which is presumably related to pre-existing segregation also. 
Fig.1. High temperature HAZ 12Cr3Ni steel, etched in Vilella's reagent, showing grain coarsened region that was fully ferritic at peak temperature
Fig.2. High temperature HAZ for 12Cr6.5Ni2.5Mo steel, etched electrolytically in 20%H2SO4, to show retained delta ferrite in grain coarsened region that was fully delta ferrite (A) and region that was mixed delta ferrite and austenite at peak temperature (B)
Fig.3. Different HAZ austenite morphologies, in region that was fully ferritic at peak temperature. The weld was deliberately made with high heat input to give a wide HAZ
Table 3 EDX microanalysis (wt%) of delta ferrite in weld HAZ of 12Cr6.5Ni2.5Mo steel and 11Cr1.5Ni steel. Measurements were made in region that was mixed delta ferrite and austenite at peak temperature but similar analyses apply tothe grain coarsened region that was fully delta ferrite at peak temperature.
| ||12Cr6.5Ni2.5Mo steel||11Cr1.5Ni steel|
||13.5 ± 0.9
||4.8 ± 1.0
||2.6 ± 0.4
||13.2 ± 0.7
||0.8 ± 0.3
||11.7 ± 0.3
||7.0 ± 0.4
||1.7 ± 0.2
||10.6 ± 0.3
||1.5 ± 0.2
|NA = not applicable
Little, if any, austenite is retained to room temperature after heating to in excess of Ac 5, so that the final microstructure consists of martensite with a small fraction of retained delta ferrite. Ms is of the order of 250°C for a 12Cr6Ni2Mo steel. 
3.3 HAZ that was mixed delta ferrite and austenite at peak temperature
The region of HAZ that was partially ferritic at peak temperature does not show any evidence of appreciable prior austenite grain coarsening, presumably indicating that the austenite remaining at peak temperature prevents deltaferrite grain growth. Due to the small temperature range between Ac4 and Ac5, this region is narrow, of the order of 50-100 microns wide for a typical TIG/MIG weld, Fig.2. Austenite reformation on cooling is apparently not Widmanstatten in nature in this area but identifying the growth morphology is difficult due to the fineness of the microstructure in this region. However, on cooling delta ferrite is once again retained to room temperature at a level that is typically less than 10%. In this case, the delta ferrite morphology is apparently intergranular, presumably with respect to prior austenite grains and enrichment of Cr and Mo and depletion of Ni is observed once more, Table 3. 
3.4 HAZ that was fully austenitic at peak temperature
The part of the HAZ that was fully austenitic at peak temperature may revert on cooling to a structure consisting entirely of virgin martensite or a mixture of martensite and stable austenite (remaining from the original parentsteel microstructure). Untempered, virgin martensite and retained austenite structures cannot be distinguished readily under an optical light microscope although colour-etching techniques can be useful. Depending on the extent of diffusion, previously stable austenite may remain stable or become unstable on cooling. The time at temperature required for sufficient diffusion to convert original stable austenite into unstable austenite (and hence martensite on cooling) has not been established.
This region typically shows a range of hardness and includes the hardest region of the HAZ, higher than that in the regions that are fully or partially delta ferrite at peak temperature, Fig 5. The primary factor controlling hardness of martensite is the carbon content (or carbon plus nitrogen content), with prior austenite grain size as a secondary factor.  Hardness increases with increasing carbon plus nitrogen and decreasing prior austenite grain size. Any delta ferrite will tend to lower hardness. Likely explanations for the hardness peak observed therefore include the finer austenite grain size in this region compared to the grain coarsened HAZ and the absence of ferrite. A further possible explanation is carbon and nitrogen enrichment of some of the martensite, as a consequence of partitioning of carbon and nitrogen between stable revert austenite and unstable austenite during heating. The last possibility is supported by the observation of highest hardness in simulated HAZ microstructures when some austenite is present rather than in fully martensitic structures.
3.5 HAZ region that was mixed austenite and tempered martensite at peak temperature
This region shows a mixture of dark etching grains, typically appearing darker than those of the parent steel presumably due to additional tempering and light etching grains of virgin martensite, Fig.4. There may also be some stable, revert austenite formed during heating. When the peak temperature exceeds Ac1 by a fairly small margin, partial transformation to austenite occurs. For furnace heat treatments of conventional nickel alloyed martensitic stainless steels, the austenite formed just above Ac1 is sufficiently enriched in austenite stabilising elements (e.g. C and N in particular) that its M s temperature is reduced to below room temperature and it remains stable on cooling. [7-9] The temperature range over which this occurs is from Ac1 to around Ac1 + 100°C. [7-9] At higher temperatures, the amount of austenite formed on heating increases, the enrichment of the austenite is consequently less pronounced and hence the fraction of stable, revert austenite decreases. For conventional nickelalloyed martensitic stainless steel, the stable, revert austenite decreases to zero for heating to around 200°C above Ac1. [7-9] There are no published data available relating to formation of revert austenite during a weld thermal cycle in supermartensitic stainless steel. However, it is expected that the kinetics of transformation will be such that an equilibrium level of revert austenite will not form during a typical welding cycle.
Fig.4. HAZ that was between Ac1 and Ac3 at peak temperature, showing tempered martensitic (dark) and virgin martensite (light), after etching in Vilella's reagent
The hardness of this region shows a transition from the parent steel value to the higher transformed HAZ levels, reflecting the variable mixture of tempered and untempered martensite, Fig.5. Softening in this region compared to parent steel, by about 10 VPN, has been reported.  This indicates that the additional tempering of untransformed martensite may outweigh the effect of formation of some virgin martensite but the extent of any such effect will presumably depend on the original tempering treatment of the parent steel.
Fig.5. Microhardness variation in single pass TIG bead-in-groove weld HAZ in 12Cr6Ni2Mo steel, made with superduplex filler wire. Arc energy = 0.35kJ/mm. Approximate locations of isotherms corresponding to the various phase transformation temperatures are shown
3.6 HAZ that did not exceed Ac1 at peak temperature
There will also be a degree of HAZ tempering below the Ac1 temperature, especially for the lean grades (Ac1 around 650-680°C), but this will be of a limited nature for the higher alloyed grades due to the low Ac1 temperature of these materials (about 550°C, Table 2). Again the extent of any softening will depend on the original tempering condition of the steel. A neural network for prediction of Ac1 of supermartensitic stainless steels as a function of heating rate has been developed.  This is available from the following site: www.msm.cam.ac.uk/map/neural
4. Reheating effects of a second thermal cycle
A second weld thermal cycle can have a number of effects on HAZ microstructure. When reheated above Ac1, a degree of grain refinement is observed in the grain coarsened region (i.e. the region that exceeded Ac5 during the first cycle) due to nucleation and growth of new austenite grains. In other parts of the HAZ exceeding Ac1, reformation of austenite will be expected, to a first approximation, to give a resultant microstructure comparable with that associated with a single thermal cycle. However, where a non-equilibrium level of stable austenite exists after one cycle, the second thermal cycle will increase the time available for diffusion and may reduce the level of such austenite.
A second cycle also allows an opportunity for carbide precipitation above and below Ac1 (in the approximate range 400°C-Ac3), whereas the low Ms temperatures of these steels will effectively prevent precipitation within any virgin martensite formed during the first weld cycle. This may be very significant with respect to intergranular corrosion of weld HAZs, which to date has been reported only in weld roots, i.e. areas subjected to multiple thermal cycles. [4,5] Cr23C6 precipitation on prior austenite boundaries occurs at above about 450-500°C, [8,9] so any sensitisation might be expected to develop in regions heated to between 450°C and the upper limit of Cr23C6 formation without adequate chromium mobility to prevent formation of depleted zones. Published information suggests the upper temperature for sensitisation is around 550-600°C, [14,17] although some effect of bulk composition is anticipated.
5. Postweld heat treatment
The effect of conventional PWHT on mechanical properties, in terms of softening through carbide precipitation and revert austenite formation is well established for nickel-alloyed 13%Cr steels with ≥0.03%C [7,9] but little specific information is available for the supermartensitic grades.  Maximum softening of conventional martensitic stainless steels, in general, is associated with Cr23C6 precipitation at above 500°C and for nickel free steels, tempering is typically performed at 700°C. For nickel alloyed supermartensitic grades, the low Ac1 temperature necessitates a somewhat lower tempering temperature. For tempering between 620 and 680°C a supermartensitic steel with 2.7%Mo was found to form predominantly molybdenum carbides,  which tend to give secondary hardening. Consequently, the Mo-alloyed supermartensitic steels are fairly resistant to softening on tempering. For 13%Cr4%Ni steels, tempering at about 600°C has been adopted but for supermartensitic steels with around 6%Ni and 2%Mo a brief treatment at 650°C has been used when PWHT was considered necessary. [3,6] This is considerably in excess of Ac1 according to measurements made at 10°C/min,  although it should be noted that Ac1 is heating rate dependent and generally increases with heating rate.  This brief 650°C PWHT was reported to increase resistance to sulphide stress cracking,  although the mechanism of improvement was not identified. In this respect, it may be noted that Enerhaug et al concluded that behaviour of HAZs in mildly sour environments was controlled by the surface oxide rather than the underlying microstructures.  However, in test specimens without surface oxide present, Hara and Asahi found that the presence of delta ferrite increased susceptibility to sulphide stress cracking.  The effect was attributed to precipitation of chromium carbides/nitrides at ferrite/martensite boundaries. Also, in laboratory studies, it has been reported that brief PWHT at 650°C increases resistance to intergranular corrosion, suggesting that healing of chromium-depleted zones may occur even in a very short time.  The diffusion distance for chromium is of the order of only 10nm for 5 minutes at 650°C, using the published diffusion coefficient for Cr in Fe  but the extent of any Cr depleted zone has not been established yet.
In order to be able to control reliably the mechanical and corrosion properties of supermartensitic weld HAZs, there is a need for additional data to establish the detailed effects of PWHT, in addition to the weld thermal cycle, on the microstructure and surface oxide in the HAZ. The microstructure and surface oxide will vary with location within the HAZ and between the variety of supermartensitic steels currently available.
6. Concluding remarks
Since the 1999 Supermartensitic Stainless Steel Conference, the industrial exploitation of these steels has increased substantially. In parallel, some efforts have been made to develop understanding of weld HAZ microstructure development and effects of PWHT. Nevertheless, there is still limited quantitative data available concerning the precise details of delta ferrite, austenite, martensite and carbide formation for the variety of steels available, which although of similar composition, may have significant differences in HAZ microstructure and hence properties.
Due to the lack of data on HAZ microstructure, it is very difficult, at present, to link the various features of a specific weld HAZ to the properties of relevance to service and to set meaningful quantitative limits, e.g. for weld procedure qualification purposes. To date, some efforts have been made to limit weld area hardness but there is an absence of data on the effect of hardness on properties relevant to service and it has not been demonstrated that hardness is a controlling parameter. Currently, interest is focussed on corrosion and stress corrosion properties of the weld HAZ but attention is also required in relation to mechanical integrity issues and in particular the effects of hydrogen. At present, there is no alternative to qualification of welding procedures for supermartensitic stainless steels via testing which simulates conservatively the likely installation and service conditions or by direct comparison with equivalent or less onerous service experience with the same grade.
The authors thank colleagues at TWI and University of Cambridge for assistance in preparing the paper and are particularly indebted to the late Trevor Gooch for his guidance and encouragement and to the late Alan Haynes, NiDI.
- AW Marshall and JCM Farrar: 'Welding of ferritic and martensitic 11-14%Cr steels', Welding in the World, vol. 45, No. 5/6, 2001.
- JJ Dufrane, E Franceschetti, J Heather and H van der Winden: 'Weldable 13%CR steel: the development of the components for a wet gas piping system', Stainless Steel World, April 1999, 27-33.
- J Enerhaug, PE Kvaale, M Bjordal, JM Drugli and T Rogne: 'Qualification of welded super 13%Cr martensitic stainless steels for the Asgard field', Proc conf 'Corrosion 99', NACE International, paper 587.
- P Woollin, D N Noble and B Lian: 'Weldable 13%Cr martensitic steels for pipeline application - preliminary studies', EPRG/PRCI 12th biennial joint technical meeting on pipeline research, Groningen, The Netherlands, May 1999.
- T Rogne, H I Lange, M Svenning, S Åldstedt, J K Solberg, E Ladanova, S Olsen, R Howard and R Leturno: 'Intergranular corrosion/cracking of weldable 13%Cr steel at elevated temperature', Corrosion 2002, Paper 02428, p1-14.
- J Enerhaug, Ø Grong and U M Steinsmo: 'Factors affecting initiation of pitting corrosion in supermartensitic stainless steel weldments', Science and Technology of Welding and Joining 2001, Vol. 6, No. 5, p330-338.
- T G Gooch, P Woollin and A G Haynes: 'Welding metallurgy of low carbon 13% chromium martensitic steels', Supermartensitic Stainless Steels 99, Paper S99 - 22, p188-195.
- E Folkhard: 'Welding metallurgy of stainless steels', Springer Verlag, 1984.
- O Bletton: 'The martensitic stainless steels', in 'Stainless steels', Les editions de physique, 1993, Editors Lacombe, Baroux and Beranger.
- JK Solberg, EV Ladanova, G Rorvik and PE Kvaale: 'Post weld heat treatment response of coarse grained heat affected zone in a supermartensitic stainless steel', Supermartensitic Stainless Steels 99, Paper S99 - 7, p56-63.
- D Carrouge: 'Phase transformations in the heat-affected zone of low carbon 13%Cr stainless steels', PhD thesis under preparation.
- D Carrouge, P Woollin: 'Microstructural change in high temperature heat-affected zone of low carbon weldable '13%Cr' martensitic stainless steels', Stainless Steel World 2002, Paper PO233, p61-67.
- D Peckner and IM Bernstein: 'Handbook of stainless steel', McGraw Hill 1977.
- R Perry: 'Stress corrosion resistance of high-tensile steels', Metallurgical Developments in High Alloy Steels, ISI Sp Report 86, 1964, p227-237.
- J E Truman, R Perry and G N Chapman: 'Stress-corrosion cracking of martensitic stainless steels', ISI, September 1964, p745-756.
- E H Phelps: 'Stress corrosion of ferritic-martensitic stainless steels', Metals Engineering Quarterly, May 1973, p44-49.
- H Kurahashi, T Kurisu, Y Sone, K Wada and Y Nakai: 'Stress corrosion cracking of 13Cr steels in CO2-H2S-Cl- environments', Corrosion, Vo.40, No. 4, April 1985.
- T Hara and H Asahi: 'Effect of δ-ferrite on sulfide stress cracking in a low carbon 13mass% chromium steel', ISIJ International, Vol 40 (2000), No. 11, pp1134-1141.
- EA Brandes: 'Smithells Metals Reference Book 6th Edition', Butterworths, 1983.