Dear Taka,
Welcome to the MICRESS forum!
Broadening the an interface if high mobilities are applied is a quite normal process in MICRESS. Without having more details (like the driving file) I can give you only general information on which parameters can be used to prevent this artifact of the phase-field model.
First of all, I want to make sure that you are not fighting against a ghost: Are you really sure that the austenite-ferrite transformation is too slow with lower mobilities? How do you know that, do you have meaningful experiments, or do you just compare to global equilibrium (e.g. Thermo-Calc calculation)? You have to bear in mind that the system is very far from equilibrium after solidification, because the composition distribution of the interstitial elements is very inhomogeneous!
If you really need to go for higher mobilities, then the natural solution of the problem is to increase resolution, because the typical reason for interface broadening is that the diffusion length of at least some of the elements is not much bigger than the interface thickness, i.e. there are considerable concentration gradient inside the interface which correspond to a gradient of the driving force and which are the reason for the interface broadening. I know, that increasing resolution means increasing simulation time, and that in your case, the resolution has been chosen according to the needs of a solidification simulation, and deltaX is typically much thicker than the diffusion length in the solild phases!
There are more general and specific comments or measures which may help:
- use averaging of the driving force. It is specified by the avg parameter in the driving force options (phase interaction data). You should use maximum averaging:
avg 1
- interface stabilisation is proportional to the interfacial energy. Extremely small values for the interfacial energy, sometimes also triggered by an extreme anisotropy of the interfacial energy, increase the risk of interface broadening
- use the lowest possible interface thickness in cells. 3.5 cells is still working quite well, with small restrictions in the accuracy of curvature evaluation.
- in your example the substitutional elements are slow-diffusing compared to the interstitials. If you use redistribution_control and either nple or para as special redistribution model for the substitutional elements, then the effect of these elements for broadening of the interface will vanish. More information can be found in these threads:
paraequilibrium/nple
Para-equilibrium growth in MICRESS
MICRESS version 5.5 released 23.3.2010
Hopefully, one of the tips will help you!
Bernd
