how to turn off the Gibbs-Thomson effect

ripening phenomena, dislocations, grainboundary topology
Ls_Thu
Posts: 34
Joined: Wed Jun 20, 2018 2:10 am
anti_bot: 333

Re: how to turn off the Gibbs-Thomson effect

Post by Ls_Thu » Tue Nov 26, 2019 2:16 am

Dear Bernd,

Thank you so much for your kind email !

The program is sent you by the private message, please check it !

The question is the values of 'Kinetic coefficient mu between phases 1 and 2 (1 and 3; 2 and 3) [ min. value ] [cm**4/(Js)]'. If we set 1, 1 and 1e-3, respectively . There would be a critical temperature (785C), higher than which the carbon content at point A would lower than point B, while lower than 785C the it would show the opposite results.

But if we set the valuse as 1e-4, 1e-4, 5e-7,respectively, which is consistent with the literature (The role of carbon diffusion in ferrite on the kinetics of cooperative growth of pearlite: A multi-phase field study, Acta, 2006), the carbon content at point A is always higher than B (which is opposite to the results we want).

So we want to know how the kinetic coefficient affect the carbon content in ferrite?

Thank you again for your kind help !

Regards,

Sai Li



PS: the program in section- 'Diffusion Data ' should be (apology that there are some missing in file):


# How shall diffusion of component 1 in phase 1 be solved?
1 1 diff
# Diff.-coefficient:
# Prefactor? (real) [cm**2/s]
0.19960
# Activation energy? (real) [J/mol]
1.39003E+05
# How shall diffusion of component 1 in phase 2 be solved?
1 2 diff
# Diff.-coefficient:
# Prefactor? (real) [cm**2/s]
0.41900
# Activation energy? (real) [J/mol]
1.08617E+05
# How shall diffusion of component 2 in phase 1 be solved?
2 1 diff
# Diff.-coefficient:
# Prefactor? (real) [cm**2/s]
0.15010
# Activation energy? (real) [J/mol]
2.60260E+05
# How shall diffusion of component 2 in phase 2 be solved?
2 2 diff
# Diff.-coefficient:
# Prefactor? (real) [cm**2/s]
6711.9
# Activation energy? (real) [J/mol]
3.06204E+05
end_diffusion_data
Attachments
1.PNG
1.PNG (29.61 KiB) Viewed 1584 times

Bernd
Posts: 1148
Joined: Mon Jun 23, 2008 9:29 pm

Re: how to turn off the Gibbs-Thomson effect

Post by Bernd » Tue Nov 26, 2019 2:30 pm

Dear Sai Li,

having seen your input file which you sent by PM, I found some issues which could be important for the results of your simulations:

1.) You have used very high values of the interface mobility which are numerically unstable or at least overestimate growth kinetics. What you definitively should do instead is to use the "mob_corr" option for correcting the physical value of the interface mobility for artifacts of the finite interface thickness. Here is more information on how to use this option.
When using the "mob_corr" option, you still need to make assumptions with respect to the physical interface mobility. My suggestion is to assume diffusion limited growth, which you do by setting the mobility input to a high value like 1,0 cm**4/(Js). Alternatively, if you have a reference with reliable data for the physical interface mobility, you can use them (together with mob_corr). However, please make sure that these values have not just been obtained by calibration of a phase-field model without taking into account proper physics and mobility correction!

2.) In case of Fe-Mn-C, diffusion of Mn is much slower than that of C. This has the consequence that C diffusion is limiting the transformation rate, while Mn is trapped or overrun (as the diffusion length of Mn is much smaller than the numerical interface thickness!). Therefore, you must make assumptions with respect to the redistribution behaviour of Mn (nple or para-equilibrium) in MICRESS. Therefore, you should use "redistribution_control" and chose either "nple" or "paraTQ" for Mn in all three phase interactions (and "normal" for C). Otherwise you would obtain interface kinetics which are undefined and would be somewhere between the two choices.

3.) In the current input file you did not apply any thermodynamic updating. This is not correct, because it "freezes" the initial equilibrium data which are far from equilibrium and also do not yet correctly include the effects of curvature and nple/paraTQ. To do so you just need to apply a sufficiently small global update interval (e.g. 1.E-6 s).

4.) In my opinion, 1.E-4 J/cm2 is an unrealistic value for the interface energy of the three phase interactions. Given the small length scale of you simulation, with this value you obtain about 100K curvature undercooling for the fcc/bcc interface! If you have no trustful literature values, I would propose 1.E-5 J/cm2 which is certainly more realistic and does not exaggerate curvature effects.

5.) I would suggest using mobility data from database which e.g. is available in our example .GES5-file TC2015a_FeCMn.GES5.

Having now more details of your approach, my conclusion is that during cementite dissolution the carbon composition in points A and B is determined by the kinetics of the fcc/cementite and bcc/cementite interfaces and by diffusion in fcc. If you assume diffusion limited transformation for all interfaces, then the carbon composition in A must always be lower than in B because of the diffusion process through fcc in case of A. Indeed, A only could be higher than B if you would assume that the interface mobility of bcc/cementite is much smaller than of both, fcc/cementite and fcc/bcc. You would need a good argument for such an assumption...

I also sent you my proposed input file (assuming diffusion limited kinetics for all interfaces) per PM.
Best wishes

Bernd

Ls_Thu
Posts: 34
Joined: Wed Jun 20, 2018 2:10 am
anti_bot: 333

Re: how to turn off the Gibbs-Thomson effect

Post by Ls_Thu » Tue Dec 03, 2019 10:29 pm

Dear Bernd,

Thank you so much for your kind reply and program !! It runs very well and much more quickly than the previous one !

We appreciate your help very much !

But there is a new problem when we change the system to the lamellar pearlite. The results show a dissymmetry phenomenon even if we set the same parameters (please see the attachement ). May I ask what would cause this phenomenon and is it normal (correct)?


Thank you again for your kind help !!

Regards,

Sai Li
Attachments
1.PNG
1.PNG (102.68 KiB) Viewed 1577 times

Bernd
Posts: 1148
Joined: Mon Jun 23, 2008 9:29 pm

Re: how to turn off the Gibbs-Thomson effect

Post by Bernd » Wed Dec 04, 2019 12:32 pm

Hi Sai Li,

Nice to hear that I could help you.

With respect to your unsymmetrical results I would say that symmetry can only be expected if the initial geometry was already symmetrical - was it? I mean, where did you place the initial seed of phase 1?

I assume that there is no temperature gradient in z-direction, and thus there is no reason why symmetry should be recovered once it was lost. To the contrary: I even would expect a sort of competition between the "tips" of phase 1 which enhances deviations from symmetry.

Bernd

Ls_Thu
Posts: 34
Joined: Wed Jun 20, 2018 2:10 am
anti_bot: 333

Re: how to turn off the Gibbs-Thomson effect

Post by Ls_Thu » Thu Dec 05, 2019 2:13 am

Dear Bernd,

Thank you so much for your reply !

The initial microstructure is attached, it is symmetry. The red part is austenite. And there is no temperature changing in whole system. what you mean is that it is reasonable ? Or we have to adjust teh parameters to make the microstructure symmetry?

Thank you again for your kind help !

Regards,

Sai Li
Attachments
1.PNG
1.PNG (46.54 KiB) Viewed 1576 times

Bernd
Posts: 1148
Joined: Mon Jun 23, 2008 9:29 pm

Re: how to turn off the Gibbs-Thomson effect

Post by Bernd » Thu Dec 05, 2019 11:37 pm

Dear Sai Li,

It seems that your initial microstructure was not sufficiently initialized: You see no interface regions (blue color) in the beginning, apart from close to the triple points. This, together with an initial situation which may be far from equilibrium, could cause the asymmetry. Please try to increase the number of steps for initialization of the interfaces in order to get initial interface profiles which are close to stationary shape.

Bernd

Ls_Thu
Posts: 34
Joined: Wed Jun 20, 2018 2:10 am
anti_bot: 333

Re: how to turn off the Gibbs-Thomson effect

Post by Ls_Thu » Tue Dec 10, 2019 2:00 am

Dear Bernd,

Thank you so much for your kind reply and help !

We have incresed the cells number of the interface but it is still as the previous image. But when we changed the Mn (component 2) from the nple mode to 'normal mob_corr' mode, the interface would occur. May I ask what we should do (or which order) could get initial interface profiles which are close to stationary shape?

Thank you again for your kind help !

Regards,

Sai Li

Bernd
Posts: 1148
Joined: Mon Jun 23, 2008 9:29 pm

Re: how to turn off the Gibbs-Thomson effect

Post by Bernd » Tue Dec 10, 2019 5:07 pm

Dear Sai Li,

I am a bit astonished that initialization of the interface does not work in your case! There should be no difference for initialization whether you use nple or normal for redistribution, because redistribution is not done during initialization. Thus, at least at time t=0, the interface should have a smooth shape. Please note, that it is much better to see whether the interfaces are smooth or not when using the .frac* outputs . In the .phas output you only can recognize whether there is a blue color in the center of the interface (indicating fractions between 0.3 and 0.7) or not.

Certainly, using "normal mob_corr" for Mn is not a solution at all. If the problem of the asymmetry comes from strongly off-equilibrium conditions, it could help using "equilibrium" together with a suitable temperature for initial equilibrium for defining the initial concentrations to at equilibrium at a higher temperature.

It would be easier to find the reason for this behavior if you could send your .dri file.

Best wishes
Bernd

Ls_Thu
Posts: 34
Joined: Wed Jun 20, 2018 2:10 am
anti_bot: 333

Re: how to turn off the Gibbs-Thomson effect

Post by Ls_Thu » Thu Dec 12, 2019 10:16 pm

Dear Bernd,

Thank you so much for your kind reply !

I have sent the dri.file to you~please check it !

As a matter of facet, we just changed the shape of the grains compared with the spheral model (we have discussed before). But it is very strange that the interface is missing at the planer model.

Thank you again for your kind help !! We appreciate it so much !

Have a good day !

Regards,

Sai Li

Bernd
Posts: 1148
Joined: Mon Jun 23, 2008 9:29 pm

Re: how to turn off the Gibbs-Thomson effect

Post by Bernd » Fri Dec 13, 2019 2:30 pm

Dear Sai Li,

Unfortunately, I cannot reproduce your problem. If I run your input file "as is" on my computer, it grows perfectly symmetrical. It is also growing much faster - are you sure you did not change anything (e.g. interface energies)? Can you send me a copy of the .in file which you got from your unsymmetrical run (just to be sure that input was correctly interpreted)?
The strange initialisation behavior is seemingly a bug, but it is not responsible for your problem. If you wish you can compensate it by increasing the "Number of steps to adjust profiles of initially sharp interfaces" to an untypical high value (~1000), but it seems not to be necessary. I have to check whether this bug has been already fixed meanwhile.

Best wishes
Bernd

Post Reply