Dear Hee-Eun Kim,
It is a bit difficult for me to understand your basic concept. What do you mean by "solidification start temperature"? In arc welding, there is always a competition between growth of the base metal and that of new grains. Do you mean "nucleation temperature" or the "average undercooling of the solidification front"?
Furthermore, I do not understand the difference between "solid-state nucleation" and "liquid-state nucleation". Formation of new grains can happen either by nucleation on seed particles which are existing in (or released to) the melt, or by fragmentation of the solid material. Nucleation of fcc on the fcc/liquid interface cannot be expected, because MICRESS assumes diffusion controlled interface kinetics when using "mob_corr" together with a high value of the physical interface mobility (which is reasonable in case of solidification of metallic materials). If nevertheless a significant undercooling at the front is present, it is either due to a numerical artefact connected with low grid resolution, or it is due to curvature. I both cases, nucleation of fcc on the fcc/liquid interface would not be realisitc.
The competition between the growth front evolving from the base metal and potential nucleation sites in the melt strongly depends on the morpholoty of the columnar front. If you expect nucleation close to the bottom of the melt pool, i.e. shortly after the transition from melting to solidification, I believe that it is absolutely necessary to include the melting process of a realistic microstructure of the base metal. In my work I found that in most cases the best and simplest approach is to obtain a consisten initial microstructure by MICRESS simulation. In case of an additive process it can be simply the resulting simulated microstructure of the previous layer. This approach can also include the formation of new grains by fragmentation (see e.g. Böttger et al 2023).
In case that nucleation is expected close to the top of the weld pool, simulation of melting is less important. Instead, one should start from a semi-solid microstructure which is already close to stationary growth. Starting from a planar interface is far from being realistic and always goes through a state of high undercooling before breaking up, which is not happening at all in reality.
Regarding your specific questions:
1.) Of course, for the competition of potential nucleation with growth of an existing columnar front, the critical radii and the density of the potential nucleation sites are the most relevant parameters. Unfortunately, it is close to impossible to experimentally determine such distributions (if not by reverse modelling). Furhtermore, if your simulation approach uses simplifications like simulation in 2D, the absolute value of the dendrite tip undercooling can be systematically wrong, requiring to adapt to that conditions by altering the particle radii of the nucleants.
2.) In the output of .TabNuc and also the .log-file, "1(4:9/10)" means that for seed type 1, number 9 of 10 potential nucleants existing in seed class 4 has been activated (nucleated). Thus, there are really only 5 classes.
3.) Gerally, the "Grain radius" of new seeds should (in most cases) be 0, so that nucleation starts with a "small grain" with an initial fraction of 2 x phMin. This ensures that the new grain can grow with the correct composition without violating the composition balance. The answer is different, if we talk about the "critical radius" which you are requested to input in case of the "analytical_curvature" model. Using "0" here would lead to an infinite critical undercooling which would prevent from any nucleation.
4.) Nucleation undercooling is always calculated as a function of temperature and local composition, i.e. the undercooling is the difference between the local temperature and the local liquidus temperature.
Bernd