Abstract
In a recent paper on the determination of the martensite-start temperature,1 a method is proposed to identify the temperature utilising a wider range of information from experiments than just the onset of transformation. The purpose of this note is to point out some difficulties with the scheme, complicated by an additional discussion on transformation which does not occur uniformly throughout the material.
We consider first a homogeneously transforming steel with a unique martensite-start (Ms) temperature. Such a temperature can be defined in terms of an arbitrarily chosen fraction 
of martensitic transformation, but will depend on the resolution of the experimental method. So, for example, acoustic emission is more sensitive than dilatometry and hence will lead to a higher Ms temperature. For this reason, it is wise to specify the Ms temperature along with a resolution, the simplest measure of which is the minimum detectable fraction 
consistent with the capability of the measuring instrument. Therefore, an additional parameter must be stated along with Ms in order for the work to be reproducible and useful.2
In the method proposed by Sourmail and Smanio,1 a portion of the dilatometer curve beginning with the fully austenitic state and ending with a mostly transformed sample is fitted to the Koistinen–Marburger equation3
is the volume fraction of martensite. A new Ms,KM is defined by fitting the ‘full’ experimental data to the equation. It is claimed that this procedure gives a better representation of the transformation as a whole.
However, equation (1) does not in practice fit in detail to dilatometric experimental data, so an algorithm is used which involves the fitting parameters Ms,KM, λ, the expansion coefficients of the parent and product phases βγ and 
and the lattice parameter of the martensite in the absence of alloying elements 
. It is argued that 
is not needed because it ‘plays a role symmetric to that of 
in allowing the total volume change to fit’, but this assumes that the transformation reaches completion.
A minimum of four parameters therefore must be stated in addition to Ms,KM for the method to be useful. But, as can be seen from the authors’ Table 4, although the parameters have physical meaning, they have fitted values which may not be physically reasonable, for example the very large βγ, which the authors have not compared with the experimental data available from dilatometry.
Turning now to the issue of heterogeneous transformation, it should be obvious that no method which assumes a homogeneous alloy, such as the unadulterated equation (1) should be misapplied. The method should first be adapted for non-uniform transformation, and the authors have done this by including a dispersion of Ms temperatures. The paper1 unfortunately implies that caution is therefore needed with the offset method, whereas in fact caution is needed with all methods including the one proposed by the authors when dealing with segregated transformations. After all, when using Ms,KM for a heterogeneous steel, the authors would need to specify further parameters including the nature of the dispersion and how the volume fractions of the different martensites evolve.
Finally, the comparisons given in Tables 6 and 7 are not meaningful because the Ms,KM temperatures have been stated in isolation without any of the fitting parameters which would allow readers to exploit the data. On the other hand, the Ms temperature is well-specified with the single additional parameter 
needed to define the resolution. It would even be interesting to see whether the fitting parameters are different for each of the cases illustrated, even when they come from the same alloy.
Sourmail and Smanio respond to these comments in an accompanying contribution. 4