CMSX4:Simulation Error

[Sudhee]

Dear Bernd,
I used MOBFe3 for the new calculations. The concentrations are going to negative values and also the diffusion values are decreased by large numbers, which later caused problems. I have attached the screen output (.scr), input file (.in), log-file (.log), linearisation output (.TabLin) and diffusion coefficient outputs (.TabD, .diff). and a screenshot of the conc. files. Please, let me know about the issue.
Thank you.

Best regards,
Sudhindra.

[Bernd]

Dear Sudhindra,

that means that negative diffusion coefficients are not an issue any more?
Then, we should not care any more about the combination of databases but focus on the other instabilities which we see in the simulation.

I can reproduce the negative compositions which you showed as attachments. The reason is the strong initial off-equilibrium situation which is not compatible with the nple model. The nple (as well as para) model requires a wide solubility range for the corresponding elements which is not given in this case. Thus, changing "nple" for "normal" removes the problem (apart from some few error messages at the very beginning).

What made the problem even worse in this case is the use of off-diagonal diffusion for carbon: The nple model produces "sharp" concentration gradients for the substitutional elements inside each phase which, in turn, lead to a strong unphysical flux of carbon. I know that I advised you at the beginning to do so - I did not have this problem in mind - however, in combination with nple it should never be done, and it also doesn't make sense...

Of course, without nple you can leave the off-diagonal terms for carbon diffusion although I now think it is not helpful.

Bernd

[Sudhee]

Dear Bernd,
Thank you for the reply.
Yes, the problem of negative diff. coeff. is solved.
I understand what you are suggesting and it eventually solved the problem. But, is it correct considering substitutional alloying elements as 'normal' instead of 'NPLE' (as C kinetics is slowed down as it has to match with that of Si and Mn) at least at these lower temperatures?

Is it fine, if I consider longer reliniearization intervals while keeping Mn and Si in NPLE mode? Just curious to know.

Thank you.

Best regards,
Sudhindra.

[Bernd]

Dear Sudhindra,

your thoughts about cross diffusion of carbon are theoretically right. However, given the extremely low diffusivity of the substitutional elements, their diffusion profiles at the interface are orders of magnitude smaller than the numerical interface thickness. Thus, cross diffusion of C cannot be solved at the given scale, and everything you get with "multi ggg" for carbon is just numerical error.

I think using longer relinearisation times would just hide the problem. You could argue, however, that the extreme off-equilibrium situation exists only at the very beginning. I would agree in the case that nple relinearisation works without error if you start it later (or do it less often). I would not agree if you still get error messages. However, in any case, you need to switch off the off-diagonal terms of carbon..

Bernd

[Sudhee]

Dear Bernd,
Sure. I got the point.
I did make the changes you explained and is working fine now. I will let you know if any other issues occur.
Thank you very much for the help.

Best regards,
Sudhindra.