Hi Bernd,
Attached figure is the phase concentration (extracted from file e.g. *.c2Pha3)
since in the phase field model, the solute concentration follows C=f1*C1+f2*C2, where f1,f2 is the fraction of each phase and C1, C2 is solute concentration in each phase, there will be a concentration gradient across the interface especially if the solute diffusivity is small.
That is the case in my simulation of ferrite-to-austenite transformation in Fe-C-Mn alloy. The Mn is uniform initially (=Mn0) at both phases and the C in austenite (left) thus driving force is so high that the transformation approaches to NPLE (negligible partition local equilibrium). This is validated by the Mn concentration in austenite identical to the average Mn0. However, within the diffuse interface, the Mn in austenite has a spike. And the spike always follows the interface during migration,leaving the Mn in austenite equal to Mn0.
I think this is not real, how to avoid it?
Ben
Artifact?
Re: Artifact?
Hi,
why do you think the spike is not real? If we assume NPLE conditions (i.e. no para-equilibrium), we expect Mn to be redistributed inside the interface. In a sharp-interface approach this would correspond to a very sharp peak ahead of the interface, in a diffuse interface model with low resolution it is a spike inside the interface.
I admit, if no special models are used, this spike just mirrors the interface profile (equilibrium redistribution in each numerical grid cell) and thus does not lead to the correct driving force contribution of the slow element Mn. To improve that, you should use the NPLE model in MICRESS which transforms the spike to a plateau with the equilibrium composition close to the interface. You can also see this in the phase composition outputs of MICRESS. Then the driving force contribution of Mn is correctly calculated!
Bernd
why do you think the spike is not real? If we assume NPLE conditions (i.e. no para-equilibrium), we expect Mn to be redistributed inside the interface. In a sharp-interface approach this would correspond to a very sharp peak ahead of the interface, in a diffuse interface model with low resolution it is a spike inside the interface.
I admit, if no special models are used, this spike just mirrors the interface profile (equilibrium redistribution in each numerical grid cell) and thus does not lead to the correct driving force contribution of the slow element Mn. To improve that, you should use the NPLE model in MICRESS which transforms the spike to a plateau with the equilibrium composition close to the interface. You can also see this in the phase composition outputs of MICRESS. Then the driving force contribution of Mn is correctly calculated!
Bernd
Re: Artifact?
Hi Bernd,
Obviously a positive spike in austenite should be there during austenite-to-ferrite transformation.
However, here is ferrite-to-austenite transformation. A negative spike i.e. a valley exists in ferrite while the Mn in austenite--the growing phase should equal Mn0.
Therefore, this simulation modeled the NPLE with a PLE interface configuration.
Actually I want to model the transition from NPLE to PLE as transformation goes on. Is there a way?
By the way, what kind of special model can be used in MICRESS in this case .
Obviously a positive spike in austenite should be there during austenite-to-ferrite transformation.
However, here is ferrite-to-austenite transformation. A negative spike i.e. a valley exists in ferrite while the Mn in austenite--the growing phase should equal Mn0.
Therefore, this simulation modeled the NPLE with a PLE interface configuration.
Actually I want to model the transition from NPLE to PLE as transformation goes on. Is there a way?
By the way, what kind of special model can be used in MICRESS in this case .
Re: Artifact?
Hi Ben,
when you are using MICRESS without special model, the interface characteristics are close to para-equilibrium, if diffusion is slow (i.e. the theoretical diffusion peak is very small compared to the interface thickness. If there is no diffusion, there is no energy dissipation, and thus the total driving force is not reduced due to segregation of Mn. I guess the same is true on the length-scale of the physical interface thickness.
So, in order to have NPLE behaviour, you must use the NPLE model of MICRESS, or resolve correctly the Mn diffusion profile!
If you use the NPLE model with high diffusion coefficients or with very slow front movement, it should perfectly give you PLE behaviour - in fact our NPLE implementation is like a perfect antitrapping model! It has just the disadvantage that it is restricted to solution phases, where both phases can exist at the same composition. It would crash immediately with stoichiometric phases...
But please go on keeping your eyes open, as our special models are not yet tested very much!
Bernd
when you are using MICRESS without special model, the interface characteristics are close to para-equilibrium, if diffusion is slow (i.e. the theoretical diffusion peak is very small compared to the interface thickness. If there is no diffusion, there is no energy dissipation, and thus the total driving force is not reduced due to segregation of Mn. I guess the same is true on the length-scale of the physical interface thickness.
So, in order to have NPLE behaviour, you must use the NPLE model of MICRESS, or resolve correctly the Mn diffusion profile!
If you use the NPLE model with high diffusion coefficients or with very slow front movement, it should perfectly give you PLE behaviour - in fact our NPLE implementation is like a perfect antitrapping model! It has just the disadvantage that it is restricted to solution phases, where both phases can exist at the same composition. It would crash immediately with stoichiometric phases...
But please go on keeping your eyes open, as our special models are not yet tested very much!
Bernd
Re: Artifact?
Hi Bernd.
Previously I thought in NPLE model, the concentration of slow-diffusing elements within the interface is forced to the equilibrium values (NPLE tie line in the phase diagram) and the driving force is only contributed by the fast-diffusing elements e.g. C since only their concentrations are deviated from the equilibrium values.
Now when I set "nple" for substitutional element e.g. Mn in my simulation, this is still a artificial peak in the new phase.
So now I am confused how MICRESS deals with nple.
Previously I thought in NPLE model, the concentration of slow-diffusing elements within the interface is forced to the equilibrium values (NPLE tie line in the phase diagram) and the driving force is only contributed by the fast-diffusing elements e.g. C since only their concentrations are deviated from the equilibrium values.
Now when I set "nple" for substitutional element e.g. Mn in my simulation, this is still a artificial peak in the new phase.
So now I am confused how MICRESS deals with nple.
- Attachments
-
- 未命名.jpg (19.75 KiB) Viewed 4969 times
Re: Artifact?
Hi Ben,
it is correct that in the NPLE model equilibrium is assumed for Mn at the interface - but this does not mean that Mn does not contribute to the driving force! The reason for that is that the equilibrium is not the same as that if the front were not moving: Then we would have a positive Mn spike on the side of the austenite and a negative spike on the side of the ferrite, and the contribution to the driving force were 0. But when austenite is growing, there is only a negative spike on the side of the ferrrite, which is artificially "broadened" to the interface thickness by the NPLE model. Technically this is achieved in MICRESS by performing redistribtion in the interface with a fixed fraction of austenite of 1.-phMin and for ferrite of phMin.
Thus, the composition of austenite is practically equal to the mixture (=nominal) Mn composition while ferrite is at the equilibrium composition to this austenite. That means the compositions of both phases are shifted relative to the real equilibrium compositions (with not moving front), producing a driving force which acts against the movement. The front can move only if the driving force contribution of the interstitials is high enough to overcome those of the substitutionals.
By the way, what exactly shows the diagram you posted?
Bernd
it is correct that in the NPLE model equilibrium is assumed for Mn at the interface - but this does not mean that Mn does not contribute to the driving force! The reason for that is that the equilibrium is not the same as that if the front were not moving: Then we would have a positive Mn spike on the side of the austenite and a negative spike on the side of the ferrite, and the contribution to the driving force were 0. But when austenite is growing, there is only a negative spike on the side of the ferrrite, which is artificially "broadened" to the interface thickness by the NPLE model. Technically this is achieved in MICRESS by performing redistribtion in the interface with a fixed fraction of austenite of 1.-phMin and for ferrite of phMin.
Thus, the composition of austenite is practically equal to the mixture (=nominal) Mn composition while ferrite is at the equilibrium composition to this austenite. That means the compositions of both phases are shifted relative to the real equilibrium compositions (with not moving front), producing a driving force which acts against the movement. The front can move only if the driving force contribution of the interstitials is high enough to overcome those of the substitutionals.
By the way, what exactly shows the diagram you posted?
Bernd
Re: Artifact?
Hi Bernd,
It is the phase concentration (Mn in austenite).
The redistribution_control is chosen as C-normal, Mn-nple.
The interface is migrating. As you said, I expected the phase concentration profile should have been a horizontal line equal to nominal Mn concentration.
however I got a positive spike in the interface, similar to what I got previously with setting redistribution_control "Mn-normal".
That is why I am confused.
It is the phase concentration (Mn in austenite).
The redistribution_control is chosen as C-normal, Mn-nple.
The interface is migrating. As you said, I expected the phase concentration profile should have been a horizontal line equal to nominal Mn concentration.
however I got a positive spike in the interface, similar to what I got previously with setting redistribution_control "Mn-normal".
That is why I am confused.
Re: Artifact?
Hi Ben,
I agree that you should not see this strange and oscillating spike which you have there!
I checked our example Gamma_Alpha_NPLE, and everything seems correct with MICRESS Version6 (although here the transformation goes from austenite to ferrite). I know that older versions of MICRESS had still bugs in the implementation of the NPLE model!
Please check whether
- you are using MICRESS version 6
- the interface is not oscillating due to improper numerical conditions
- there is no considerable diffusion of Mn - otherwise you will get back the Mn spike in austenite (PLE)!
- phMin is sufficiently small (<=10-4)
Best wishes
Bernd
I agree that you should not see this strange and oscillating spike which you have there!
I checked our example Gamma_Alpha_NPLE, and everything seems correct with MICRESS Version6 (although here the transformation goes from austenite to ferrite). I know that older versions of MICRESS had still bugs in the implementation of the NPLE model!
Please check whether
- you are using MICRESS version 6
- the interface is not oscillating due to improper numerical conditions
- there is no considerable diffusion of Mn - otherwise you will get back the Mn spike in austenite (PLE)!
- phMin is sufficiently small (<=10-4)
Best wishes
Bernd
Re: Artifact?
Hi Bernd,
Unfortunately I am using version 5beta
. Maybe this is the reason why I got the artifact.
By the way, how can I get paraequilibrium thermodynamic data[entropy, slope etc] from MICRESS[with TQ database].
I found that regardless of nple or para setting, the thermodynamic data in the *.log file is the same. Actually, I want to get thermodynamic data for NPLE and PE, respectively.
could the driving force in para-eq case be calculated with the Local Equilibrium thermodynamics, rather than with Para-eq thermodynamics?
Ben
Unfortunately I am using version 5beta
. Maybe this is the reason why I got the artifact.By the way, how can I get paraequilibrium thermodynamic data[entropy, slope etc] from MICRESS[with TQ database].
I found that regardless of nple or para setting, the thermodynamic data in the *.log file is the same. Actually, I want to get thermodynamic data for NPLE and PE, respectively.
could the driving force in para-eq case be calculated with the Local Equilibrium thermodynamics, rather than with Para-eq thermodynamics?
Ben
Re: Artifact?
Hi Ben,
in MICRESS Version 6
there is the further option "paraTQ" available, which does exactly what you want! This functionality needs TQ version S / 7...
If you use "para" with thermodynamic data according to the local equilibrium assumption, it may nevertheless be quite correct in case of dilute alloys (no interaction between components). The pile-up for each element itself is disregarded according to the special redistribution characterisics of the "para" model. But with respect to the interaction between the elements the spikes are not removed, leading to an error in the thermodynamic description.This error will be the higher the stronger the interactions are. I guess that for low-alloyed steels it does not matter while for stainless steels the difference is relevant.
By the way, why don't you use MICRESS version 6? Don't you have a corresponding licence or is it just not installed?
Bernd
in MICRESS Version 6
there is the further option "paraTQ" available, which does exactly what you want! This functionality needs TQ version S / 7...If you use "para" with thermodynamic data according to the local equilibrium assumption, it may nevertheless be quite correct in case of dilute alloys (no interaction between components). The pile-up for each element itself is disregarded according to the special redistribution characterisics of the "para" model. But with respect to the interaction between the elements the spikes are not removed, leading to an error in the thermodynamic description.This error will be the higher the stronger the interactions are. I guess that for low-alloyed steels it does not matter while for stainless steels the difference is relevant.
By the way, why don't you use MICRESS version 6? Don't you have a corresponding licence or is it just not installed?
Bernd