Sigma formation in austenitic stainless steel weld metal

TWI Bulletin, August 1971

Mr Willingham is a Scientific Officer in the Metallurgical Department. 

Mr Gooch is a Senior Scientific Officer in the same Department

It is well recognised that austenitic stainless steels, in common with certain other alloys, may suffer formation of sigma-type phases during exposure to temperatures between about 550°C and 900°C. These phases are basically FeCr intermetallic compounds, the precise composition and phase designation being dependent on the particular alloy. For example, in AISI 316 material, the phase may contain Mo, and it is known as X-phase.

The major effect of sigma in stainless steel is to lower the resistance to impact loading, although it may influence other properties, such as corrosion resistance. This embrittling effect is not usually noticeable at elevated temperatures, but it becomes evident below about 200°C. The presence of sigma is therefore important in high temperature stainless steel plant, which is periodically shut down, cooling to about room temperature, when shock loading may initiate a brittle fracture. This danger is probably more apparent than real, since there have been few reported failures directly attributable to sigma. However, the risk does exist, and it is of concern not only in plant intended for high temperature service, but also in material that must be heat treated within the sigma formation temperature range.

Factors affecting sigma formation have been extensively studied. It is well established that transformation is accelerated by the presence of ferrite, by strain, and by the presence of ferrite-forming elements in the material. Sigmatisation is also dependent on grain size and heat treatment conditions, including any heat treatment prior to exposure to the critical temperature range. However, such factors are seldom quantifiable, and at the present time it is difficult to predict how rapidly sigma will form in the practical situation, or what effect sigma formation is likely to have on service performance. To try to clarify the situation, a brief survey of available data has been made for common austenitic stainless steels ^[l-49] . Particular attention has been paid to weld metal behaviour, because sigma formation in weld metal is likely to be more rapid than elsewhere in a structure. The objectives have been two-fold, namely to obtain a measure of the practical effect of sigma formation during high temperature service or heat treatment, and to ascertain how rapidly sigma formation can occur.

Data obtained

Studies carried out on loss of impact resistance due to sigma formation have used a number of different testing procedures, such as Charpy V or Keyhole-notch, Izod, tensile-impact, etc. The testing method will affect the energy absorption recorded, and, to enable comparison to be made, all data obtained have therefore been rationalised into percentages of the original as-welded condition cited in each reference.

On this basis, Fig.1 - 6 give the effect of time and temperature on the reduction in impact resistance of the above stainless weld metals, the relevant references being cited for each figure. As-welded impact values may vary considerably, but are typically in the range 40 - 75 ft. lb. (55 - 100 J) ^[50] , as measured by the Charpy V notch method. A value of 40 ft. lb (55 J) prior to heat treatment can generally be regarded as conservative.

Approximate C-curves have been plotted on Fig. 1 - 6 to indicate the heat treatment required to reduce the impact properties to about 50% and 20% of the original values. The scatter bands associated with these C-curves have been outlined with continuous lines to indicate a fairly definite boundary, and with broken lines where the data are insufficient, or where variation between references is particularly marked.

The data shown in Fig. 1 - 6 pertain to weld metals with a maximum ferrite content of 10% and with no intentional cold working of the welds. They are thus relevant to general practice in indicating the probable embrittling effect of heat treatment on the alloys considered, but they should not be regarded as entirely definitive. Sigma formation in production welds may be more severe if such welds should contain more than 10% ferrite, or if they should experience straining during fabrication.

Further to Fig. 1 - 6, the rapidity with which sigma formation can occur is indicated in the Table with reference to both parent material and weld metal.

Fig.1. Effect of time and temperature on reduction of room temperature impact properties of type 308 weld metal, based on data from references ^[1-3]

Fig.2. Effect of time and temperature on reduction of room temperature impact properties of type 309 weld metal, based on data from references ^{[1, 6 and 10]}

Fig.3. Effect of time and temperature on reduction of room temperature impact properties of type 310 weld metal, based on data from references ^[3-6]

Fig.4. Effect of time and temperature on reduction of room temperature impact properties of type 316 weld metal, based on data from references ^[7-9]

Application of data

Secondly, a number of recent investigations have indicated that sigma formation may be considerably more rapid than is generally recognised. This is certainly indicated by the Table. Again, variation exists between individual studies, but there is ample evidence that sigma may be initiated in only a few minutes. In practical terms, it is highly improbable that the amount of sigma formed within this time-scale will have any significant effect on service behaviour. Nonetheless, once initiated, sigmatisation may be rapid, should the bulk material structure and composition promote such transformation. Thus, although the C-curves on Fig. 1 - 6 are drawn conservatively from available data, the chance of quicker and more severe sigmatisation must be regarded as a potential hazard.

Speed of transformation to sigma phase

Accepting the above comments, and despite the scatter associated with the C-curves, it is possible to draw certain general practical conclusions from the present survey. It appears that embrittlement is more rapid in 18/8 type materials than in the more highly alloyed AISI 309 and 310. Few data are available on AISI 321, but this alloy may be expected to behave in a similar fashion to AISI 347. Although the picture is incomplete, examination of ternary Fe-Cr-Ni equilibrium diagrams ^[18,30,51] suggests that the final equilibrium sigma content will be similar for all the weld metals considered. Thus the more rapid embrittlement in 18/8 alloys probably results from more rapid sigma initiation; this may well be a consequence of a generally higher ferrite content in 18/8 weld metal than in AISI 309 and 310 deposits.

Further, there can be no doubt that sigma will form in austenitic stainless steel during long term service between 550°C and 900°C, with a corresponding reduction in impact properties. If a minimum value of impact strength exists, it will be very low, and certainly below 5 ft lb (7 J). From examination of structures developed after extended periods at temperature, it is probable that in the compositions presently considered the equilibrium sigma content will be of the order of 30%. This would be quite sufficient to form a continuous network, and to cause drastic loss of fracture toughness. This situation must be faced, and it is obviously imperative that heavily sigmatised structures should not be subjected to any form of shock loading during shut-down periods.

The situation regarding postweld heat treatment is rather more encouraging. It is sometimes necessary to heat treat stainless steels in the sigmatising range, for example when tempering dissimilar metal joints or weld-clad material. The data in Fig. 1 - 6 show that with typical heat treatment times of up to 25 hr a 50% reduction in impact resistance is a strong possibility. However, the final impact fracture toughness of the material is unlikely to be so low as to cause concern, and thus for practical purposes sigma formation during a postweld heat treatment operation should not be significant. Obviously, caution must be applied; in some cases a 50% loss of toughness may be too severe, and sigma formation may be more rapid than in the studies cited in Fig. 1 - 6. Should doubt exist in a particular instance, the only recommendation that can be made at the present time is that realistic procedural trials should be carried out to determine what effect sigma is likely to have. These trials should make allowance for all the factors likely to enhance sigmatisation.

It is further remarked that such data as are available on loss of toughness are based on traditional impact tests. There is ample evidence from studies on ferrous steels that these tests are of limited relevance to service conditions ^[52] . More confidence could be placed on data accrued from more recently developed fracture mechanics tests ^[52] , but there appears to have been no work on sigmatisation employing such techniques. This constitutes a serious gap in available knowledge, and it is desirable that suitable studies should be undertaken.

Fig.5. Effect of time and temperature on reduction of room temperature impact properties of type 321 weld metal, based on data from references ^{[2 and 5]}

Fig.6. Effect of time and temperature on reduction of room temperature impact properties of type 347 weld metal, based on data from references ^{[2, 5, 7-9, 11-14]}

Conclusions

• Sigma may be initiated within a matter of minutes at suitable temperatures.
  
• With long term exposure, drastic loss of fracture toughness must be regarded as inevitable.
  
• Sigma formation during a postweld heat treatment operation will reduce material toughness, but the loss is unlikely to be of practical significance, unless factors accelerating sigma formation, such as high ferrite contents (>10%), high strain, etc. are operative.

References