vanadium partitioning in vanadium carbides

[pmecozzi]

I am currently running a simulation to study carbide precipitation at martensite lath boundaries during martensite tempering. The simulation starts with a ferritic matrix supersaturated with carbon and substitutional elements (in this case, Mn and V). VC precipitation (FCC_A1#2 in the Thermo-Calc GES file) is allowed to occur at the lath boundaries. The driving file used for the simulation is attached.
The simulation is performed at 923 K. The equilibrium phase fractions calculated using Thermo-Calc are reported in Table 1.

Stable Phases
BCC_A2#1 Moles Mass Volume Fraction
0.98684 55.09099 0.98919

Composition
Component Mole Fraction Mass Fraction
Fe 0.98104 0.98142
Mn 0.01871 0.01842
V 0.00017 0.00015
C 0.00007 0.00002

FCC_A1#2 Moles Mass Volume Fraction
0.01316 0.43699 0.01081

Composition
Component Mole Fraction Mass Fraction
V 0.45844 0.70350
C 0.46537 0.16838
Fe 0.07370 0.12399
Mn 0.00249 0.00413

Table 1 Equilibrium phase fraction at 923 K.

During this study, I have encountered the following issues:
Vanadium partitioning:
Vanadium partitions into the VC phase only in particles from seeds numbered 1 to 10. These particles originated from seeds, which nucleated simultaneously during the initial nucleation event. In contrast, particles that nucleated later in the simulation retain the same vanadium content as the ferrite matrix and do not exhibit any vanadium partitioning. This behavior can be observed in the vanadium concentration map shown in Figure 1.
 
Figure 1 VC precipitation during 0.5 s of holding at 923 K (driving file attached).


I also modified the nucleation settings by reducing the number of seeds nucleated simultaneously during each nucleation event to one. In this case, vanadium partitions only into the first two particles, which nucleated during the first two nucleation events. This behavior can be observed in the vanadium concentration map shown in Figure 2.
 

Figure 2 VC precipitation during 0.5 s of holding at 923 K, with modified nucleation setting.

Discrepancy with the equilibrium data
As shown in Figure 3, the VC phase (FCC_A1#2 in the GES file) appears to reach a phase fraction higher than the equilibrium value predicted by Thermo-Calc (see Table 1).

Figure 3 Carbide volume fraction during holding at 923 K.

Could you please provide an explanation for these two observations?
In particular, I would like to understand why vanadium partitioning appears to depend on the nucleation sequence and why the simulated VC phase fraction exceeds the equilibrium value predicted by Thermo-Calc. Could these discrepancies be related to an issue with the GES file used for the thermodynamic calculations?

Kind Regards,

Pina Mecozzi

[Bernd]

Dear Pina,

Only having the attached input file it is not so easy to understand what could happen here. The input file starts with reading in an initial microstructure with concentration distributions from files, but all that is overwritten by reading from a restart file. As I don't have all these files, I cannot know the exact input data and also cannot make a test run myself.

Anyway, what seems clear from the attached results is that the FCC_A1#2-phase does not grow with the correct vanadium composition. This could indeed be a problem of the .GES5-file, if the default major constituents have not been correctly set to C and V. You can check this in the .log-file output or in the standard (screen) output where the default major compositions are summarized at the end of initialisation. You should also check the initial equilibrium for phases 2/4 (written to the .log-output) for a correct composition of FCC_A1#2.

Another possible problem I noticed is in the nucleation data. You define an initial radius for the FCC_A1#2 seeds which is bigger than the grid spacing, forcing the seeds to have the (wrong) initial composition of BCC_A2#1. Instead, you should use an initial radius of zero, so that the seeds can start with correct composition.

A further point, maybe, is that you should define all components of phase 4 as stoichiometric and not just carbon. In any case, please remove the line "stoich_enhanced_on" because it does not make any sense in this context.

If these suggestions do not solve your problem, please send your latest input file together with all other required inputs so that I can try to reproduce the error and search for other solutions.

Best wishes
Bernd

[pmecozzi]

Dear Bernd,

Thank you for your fast reply.
I followed your suggestions, and the simulation is now running properly.
This is what I did:
First, I checked the log file for the equilibrium composition of the FCC_A1#2 phase, and it corresponded to the expected composition of the vanadium carbide at the prescribed temperature.
Then, I changed the nucleation input data by setting the initial radius of the nucleus to zero. Finally, I defined all three components of the FCC_A1#2 phase as stoichiometric.
The simulation is now running, and it seems that the composition of the carbide phase is correct.

Kind Regards,
Pina

[pmecozzi]

Dear Bernd,
As I mentioned in my previous reply, the simulation is now able to reproduce the correct carbide chemical composition.
I have therefore let the simulation run further to investigate the kinetics of carbide precipitation at the ferrite grain boundary and the resulting vanadium concentration in the ferrite matrix.
Unfortunately, as the holding time at 923 K increases, the carbides become unstable, as can be seen from the microstructural evolution shown in Figure 1.
Could you please give me some advice on how to resolve this issue?

Figure1a_time60s.jpg

Microstructure and vanadium concentration (at%) after 60 s holding at 923 K

Figure1b_time90s.jpg

Microstructure and vanadium concentration (at%) after 90 s holding at 923 K

Figure1c_time120s.jpg

Microstructure and vanadium concentration (at%) after 120 s holding at 923 K

Figure1d_time150s.jpg

Microstructure and vanadium concentration (at%) after 150 s holding at 923 K

I have attached the log file of this simulation.

Kind Regards

Pina

[Bernd]

Dear Pina,

The .log-file does not give any information when the VC-particles get unstable. Do you get error messages on standard (Screen) output?

WIthout the possibility to run the simulation myself, I can only make some crude guesses:

1.) The interface mobility of FCC_A1/FCC_A1#2 (interface 1/4) is simply too large. Why don't you use "mob_corr" to automatically adjust the numerical interface mobility to a given physical value (e.g. a large value like 1.0 for diffusion control)?

2.) Perhaps, the phase composition of V in FCC_A1 gets negative when the FCC_A1#2-particle gets "large" (and the phase fraction of FCC_A1 very small. More frequent updating of TQ could help in this case. If not you should use "penalty" (in section "Numerical Parameters"/"Concentration Solver") to avoid negative compositions.

3.) There could be an issue with averaging of driving forces. Try to use "avg 0.5".

4.) Maybe, grid resolution is not sufficiently high to resolve the diffusion fields of the substitutional elements (V and Mn) around the VC-particles.

Best wishes
Bernd

[pmecozzi]

Dear Bernd,

Thank you for your suggestions.
I will correct the input by applying each of your suggestions separately and check whether this helps to resolve the problem.
In any case, I have attached the driving file and all the files required to run the simulation. If you could run the simulation yourself, it may help you better understand the cause of the instability.
Thank you again for your help.

Kind Regards,

Pina