Diffusionless transformation

dendritic solidification, eutectics, peritectics,....
Post Reply
G_azizi
Posts: 18
Joined: Wed Feb 17, 2021 7:59 pm
anti_bot: 333

Diffusionless transformation

Post by G_azizi » Fri Nov 18, 2022 12:21 am

Hi,

I am simulating peritectic solidification and in that the solid state transformation from delta-ferrite to austenite takes place. This transformation rate is too high (in order of millimeters/second). But in my simulations it's rate is very low. how should I change the "phase interaction data" parameters or other parameters to get such a transformation rate. Some people believe this transformation is diffusionless (controlled by short range diffusion or massive transformation).
is there any way in MICRESS to simulate such a diffusionless transformation? or in MICRESS all transformations are diffusion controlled.

Best

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Diffusionless transformation

Post by Bernd » Fri Nov 18, 2022 12:49 pm

Hi,

A diffusionless or "massive" transformation occurs if the driving force for transformation is so high that the pile-up of all elements at the interface can be overrun. Then, the front velocity can be much higher than in the diffusion-limited case. For each element, depending on the physical background, solute redistribution may still take place, or not. The first condition is called "NPLE" or "LENP" (no partitioning - local equilibrium), the second "para-equilibrium", the difference is how much driving force is essentially necessary to get into the condition.

In MICRESS, we handle that by using the "redistribution_control" option, which allows for definition of the corresponding condition for each element:

a) diffusion-controlled:
normal mob_corr
or
atc mob_corr (including an anti-trapping correction of diffusion)

In this case, diffusion control is assumed even if the driving force is so high that a diffusionless transformation could be possible

b) NPLE/LENP-condition
nple

c) Para-equilibrium
para (alternatively, the Thermo-Calc para-equilibrium model can be used with "paraTQ")

For the γ-α or δ-γ-transformation, typically, diffusion control by the element C is assumed. So, the choice in MICRESS would be to use "normal mob_corr" for C (and other fast diffusing interstitial elements like N, O, H, P), and use "nple" for all other elements. Then, the transformation should occur fast, only controlled by carbon diffusion.

PS: If you would use MICRESS without "redistribution_control" option, the δ-γ-transformation should be fast also, but potentially numerically unstable and with wrong kinetics.

Bernd

G_azizi
Posts: 18
Joined: Wed Feb 17, 2021 7:59 pm
anti_bot: 333

Re: Diffusionless transformation

Post by G_azizi » Tue Oct 03, 2023 10:55 pm

Hello Bernd,

Thanks a lot for your advices. I tried to use normal mob_corr to get high velocity for δ to γ (phase 2 to phase 1) transformation but it did not work and still the transformation rate is very low.
I applied very high undercooling (like what reported in literature) for γ nucleation (maximum nucleation T for phase 1 at interface). and again the rate is very low.
I tried to use nple and para-equilibrium but simulation stopped due to some error which I could not resolved.

I have attached the input file and the GSE file to this message. would you please take a look at it kindly and see what is wrong? I think para-equilibrium should solve the problem.
please not that I use MICRESS 6.4.

Thank you very much in advance.

Ghavam
Attachments
Fe-C.GES5
(90.03 KiB) Downloaded 92 times
Fe-0.13C.in
(22.66 KiB) Downloaded 110 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Diffusionless transformation

Post by Bernd » Wed Oct 04, 2023 12:42 am

Dear Ghavam,

If you think the transformation should be very fast because of a diffusionless transformation, you should not use "mob_corr", because this limits the transformation kinetics to "diffusion-limited". You say that nple or para/paratq gives numerical problems - which are the problems? If they cannot be solved, you can also try to use "normal" without "mob_corr", even if the kinetics may be not quantitatively correct.

Please note that currently I am on holidays and cannot observe the forum continuously, so it may take some days for me to answer...

Bernd

G_azizi
Posts: 18
Joined: Wed Feb 17, 2021 7:59 pm
anti_bot: 333

Re: Diffusionless transformation

Post by G_azizi » Wed Oct 04, 2023 10:21 pm

Thanks Bernd,

I could not run the simulation with para or nple. In other word, I was not able to make it work. would you please take a look at the file whenever you had the chance.

Many thanks

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Diffusionless transformation

Post by Bernd » Tue Oct 10, 2023 11:57 am

Hi Ghavam,

I think you have the following errors in your input file:

1.) You cannot have set C to stoichiometric when using the nple or para option in MICRESS. I also don't think it makes much sense otherwise.

2.) It is absolutely necessary to have some updating of thermodynamic data, e.g. all 0.01 seconds ("manual 0.01"), especially if interfaces are stagnating (so that there is no implicit updating by movement). As a consequence, phase 1 will get stable much earlier, so that you also should increase the maximum temperature for the 2nd seed type.

3.) You need to apply diffusion-limited interface kinetics ("normal mob_corr") not only to the 0/2 interface but also for 0/1.

Further, I would strongly advise you to do numerical testing with a much smaller simulation domain, otherwise you will loose too much time waiting for response, and use the final size only for production of results. I also recommend using the "global" relinearisation scheme for all interfaces for saving time (there would be even much better options available in newer MICRESS versions from 7.1 on).

Best wishes
Bernd

Post Reply