NHML Resources - Martensite in Austenitic Stainless Steel Welds

Background

Martensite is a crystal structure that forms in steels during rapid cooling. Cooling rates are dependent on the particular chemistry of the steel. Certain conditions can be met that will cause martensite formation in austenitic stainless steels.

Long, needle-like clusters of crystals in the metal characterize martensite. The martensite crystals have a highly stressed body centered tetragonal structure. In austenitic stainless steels and the austenite phase of a magnetic (austenitic/ferritic) stainless, when martensite is present in a weld, don't expect it to pass the bend test. The weld will not have the toughness and ductility that we normally expect from a stainless steel, and in some circumstances there may be unanticipated corrosion.

Fillet Composition

With only one exception, adding alloying elements to steel allows martensite to form at a slower cooling rate. That exception is cobalt, which works the other way. With cobalt, martensite will form at a faster cooling rate.

We recognize that the actual composition of the weld fillet will vary along a line running across or bisecting the weld fillet. On the centerline, we expect the fillet composition to be closer to the wire composition. Approaching each base metal, the composition shifts from the filler metal composition towards each base metal. In welding metallurgy, we mark on the phase diagram the chemistry of one base metal and the chemistry of the filler wire. We find that the actual fillet compositions lie on the line drawn on the phase diagram. When the base metals are different, it takes different lines on the phase diagram to represent the compositions approaching each base metal.

Predicting Martensite

The results and microstructural consequences of this sort of exercise in physical metallurgy appeared in 1949 as the "Schaeffler Diagram". Our technology did not stand still and the Diagram iterated through several updates. The 1994 Winter Addendum to the ASME Code brought us the Welding Research Council's "WRC-1992 Diagram" which continues to be extensively used. Still, there has been a problem associated with manganese, which brought yet another modification into use. The analysis behind the modification appears, for example, in D.J. Kotecki's, "A martensite boundary on the WRC-1992 diagram" (Welding Journal , Vol. 78, No. 5, pp 180-192).

In lots of cases the 1% manganese line satisfactorily predicts martensite or not martensite. However, we sometimes encounter steels having higher manganese and quite often encounter much lower. High side examples include 1.0-1.5% manganese in a 309L filler wire, 4% in a 307, and 6% in the European 18 8 Mn filler wire. On the low side, if you are joining stainless to a modern carbon steel, the manganese can be quite low, perhaps even 0.3%.

The 1949 Schaeffler diagram - click to enlarge

(The 1949 Schaeffler diagram - click image to enlarge)

To enter the modified Diagram, we need to calculate two numbers roughly based on chromium and nickel content. We also have to apply a lot of welding "know-how" as to the mixing of the metals.

Calculate the

    nickel equivalent:

    • % nickel
    • + 35 x % carbon
    • + 20 x % nitrogen
    • + 0.25 x % copper

    chromium equivalent:

    • % chromium
    • + % molybdenum
    • + 0.7 x % niobium

The Amount of Nitrogen

One of the problems we have to also confront is the amount of nitrogen. The best shielding practice with a wire electrode might not introduce nitrogen. Flux cored electrodes tend to add nitrogen. For example, the metal in a flux cored electrode might analyze to 0.05% nitrogen while it deposits as 0.075% nitrogen. Most of us deal with the nitrogen by first plotting the points with only the known nitrogen and then we plot the nearby point based on our estimate of the actual nitrogen. The compositions at risk are those on the Diagram and below the indicated manganese brands. Within the bands the diagram is known to be imprecise.

In practice, you have to cope with the range of compositions between one base metal and the filler metal, and between the other base metal and the filler metal. You could put all of your predicted nickel and chromium equivalent compositions on the graph. However, examining the diagram we see that the compositions more at risk for forming martensite are those with low nickel and chromium equivalent numbers, coupled with low manganese. After a little practice with the diagram we know pretty well which compositions are likely to be at risk and we tend to plot only those numbers.

A warning: Notice that all of the alloying elements in the calculation contribute to hardenability. Should either base metal or the filler metal contain any unlisted elements which are known to contribute to hardenability then the Diagram doesn't apply. After decades of absence, tungsten is showing up mostly as a substitute for molybdenum and mostly in Russian and Chinese metals. In doing the calculation the practice is to lump the tungsten with the moly on a 1:1 basis. When forced to do it, welding metallurgists may tweak the calculations a little to reflect other unlisted elements, however, there isn't much of a research foundation on which to base such adjustments.

Ferrite Number

The 300 series of stainless steels is austenitic (non-magnetic) while the duplex stainless steels are mixed austenite and ferrite. In the field, the amount of ferrite is measured through its magnetic response. The portable meter is calibrated for % ferrite and it is called the "ferrite number". Martensite gives a magnetic response, but not as strongly as ferrite, so when it is present it contributes to the "ferrite number".

Recognize also that perfectly good austenitic stainless steel that has been heavily cold worked can become slightly magnetic and give a ferrite number even though ferrite nor martensite are present.

In the upper right side of the diagram there is a cluster of lines of constant ferrite number. The upper left line is 1% ferrite number which is nominally 99% austenite and 1% ferrite. The last line along the lower right side of the cluster is nominally 98% ferrite and 2% austenite.

Learning by Doing (and "cya")

Good record keeping means photocopying a bunch of diagrams and every time you confront the martensite issue you make a record of how you calculated the nickel and chromium equivalents. You plot these on the graph, draw the lines between and estimate an adjustment for nitrogen. You note the ferrite number and whether the bend test is pass or fail.

You don't just do the exercise for samples that fail the bend test. Otherwise you won't learn the limits for welds that pass every time! If testing the weld for its ferrite number is worthwhile then making the record is also worthwhile. The day will come when you note the manganese is low and you make yourself into a hero when you predict the need for a little preheat.

The WRC-1992 Diagram modified to reflect experience with manganese. Martensite is predicted to form with composition equivalents below and to the left of the manganese bands. Reprinted from Advanced Materials&Process, June 1000, p. 75 - click to enlarge

(The WRC-1992 Diagram modified to reflect experience with manganese. Martensite is predicted to form with composition equivalents below and to the left of the manganese bands. Reprinted from Advanced Materials&Process, June 1000, p. 75 - click image to enlarge)

There will be discrepancies. When it happens it most likely means that you misjudged the effect of cooling rate, or you didn't get the chemistry right, or the basic diagram is imperfect!

See our Industry Definitions for further insight.

 

<< return to Resources

News

Nuts & Bolts, V23

All News >>

Related Materials

Validation Compliance

Valid XHTML 1.0 Transitional Valid CSS!

who's online