Response to ‘A second set of comments on “Determination of M s temperature: methods,meaning and influence of the ‘slow-start’ phenomenon” by T. Sourmail and V. Smanio’

The present authors have recently proposed a method to identify the martensite start temperature, ¹ which has been the object of a discussion.^2,3 Bhadeshia has now provided a second set of comments ⁴ which are addressed below.

On the expansion coefficient of austenite, we agree with Bhadeshia's comments that determination of M_s is difficult, if not suspect, when dilatometry records exhibit significant non-linearity in the fully austenitic temperature range. (In this regard, it should be recalled that the data discussed in this particular example were from Ref. 5, in which no mention of this issue was made.) Clearly, neither method should be applied where significant non-linearity exists in the measured austenite expansion (as is the case in the example discussed). However, it is also the case that in most recordings obtained the dilatometry data do not show a perfectly linear austenite expansion. The purpose of our example was to illustrate the sensitivity of the offset method and the need to publish the temperature range used to define the tangent used as a reference. Either method would conceal the issue. As stated in our first reply ³ (paragraphs 3 and 4), it has never been claimed that the M_s,KM avoids the problem; rather, the discussion concerned only the set of parameters which should be published with the results.

Regarding Bhadeshia's comments ⁴ concerning Fig. 3 of Ref. 1, it should be emphasised that the purpose of the work was not to define or identify a ‘better’ method but simply to discuss the respective merits of the two methods. It is with this perspective that we have mentioned the ‘K–M method’ as giving a ‘better representation of the transformation as a whole’, and not simply as a better method.

We also agree that M_s,KM as identified in Fig. 3 of our response ³ ‘underestimates’ the measured martensite start temperature. However, as we have indicated, but perhaps insufficiently explained, there are cases where the measured martensite start temperature may differ from the real martensite start temperature as a result of experimental issues. In the course of the present discussion, we have demonstrated^1,3 that early measured transformation onset could be related to specimen heterogeneity and have distinguished two cases: (i) where the heterogeneity is extrinsic (temperature gradient in the specimen, surface effect, etc.), M_s,KM will be better adapted to approach the real M_s value of the material at hand; (ii) if, on the contrary, the heterogeneity is intrinsic, then indeed the offset method proposed by Bhadeshia will be best suited to identify M_s.

In his second set of comments, Bhadeshia ⁴ states that the real onset of martensitic transformations becomes more sensitive to experimental probes with better resolution. It is assumed that by ‘real’, reference is made to the martensite start temperature measured in ideal experimental conditions (regardless of intrinsic heterogeneities, which are unavoidable, but in the absence of any extrinsic factor). The conclusions of the present authors are not different, but concern the measured onset. Indeed, these are exactly the conclusions to be drawn from Table 2 of Ref. 3: the measured onset is highly sensitive to heterogeneities in the specimen.

As highlighted in our previous discussions, measured and real onsets may not coincide for a number of reasons (temperature gradient, influence of free surface, etc.). Thus, the stronger sensitivity of the offset method may, in a number of cases, only be efficient in measuring artefacts rather than actual transformation. A clear example is that of a longitudinal gradient which causes the specimen ends to be cooler than its centre. ¹ This effect is not normally visible in a standard experiment where the control thermocouple is welded at the centre of the specimen Figure 1. If, for example, it is supposed that the extremities of the specimen reach a temperature of 220°C or below (assume, for this example, a real M_s value of 220°C), when the area under the control thermocouple is still at 230°C, then the higher sensitivity of the offset method will lead to an estimate for M_s of 230°C (since this is the temperature measured by the instrument when the specimens extremities reach the ‘real’ M_s and thus began to expand). Clearly, this represents not the occurrence of a transformation at 230°C but experimental measurements issues. In this case, and contrary to the offset method, the ‘K–M method’ will provide an estimate which is closer to the ‘real’ M_s, as already highlighted.^1,3

OPEN IN VIEWER

1

Schematic illustration of the example discussed in the text, in which a temperature gradient in the specimen causes transformation at the extremities to take place before it occurs at the centre. Because temperature measurements are typically carried out only at the specimen centre, the onset of transformation is associated with the measured temperature (here, a temperature of 230°C or greater is likely to be identified as M_s using the offset method). The overestimation of M_s will become greater as the onset is measured with increased accuracy, but will not, for all that, be much better related to the actual M_s temperature of the material. The present authors have shown this hypothesis to be quantitatively consistent with measured gradient in the Baehr 805 dilatometer and with typical amplitudes for the ‘slow-start’ phenomenon ¹

In summary, our discussion does not aim at establishing a ‘better’ method but to assess the advantages of each method. It is clear and agreed that the offset method is best suited to determine the actual onset of transformation during cooling. However, when extrinsic heterogeneities are present (which is always the case in the experience of the present authors), this true expansion onset does not necessarily correspond to the ‘real’ martensite start temperature.

As to the practical aspects of the method and the use made of the M_s measurements within numerical simulation software packages, the present authors would suggest that this is a subjective matter that may best be left for users to decide.