Basics for Martensitic transformation
Basics for Martensitic transformation
Hello Bernd,
I am new to MICRESS. I am trying to simulate a microstructure evolution during cooling. Initially, I have a complete austenite microstructure which is cooled down at 20°C/s and is held at 170°C and is reheated again. I am now concentrating on the initial cooling stage. The Ms of the material is around 270°C. I expect austenite (30%) and martensite (70%) in the end microstructure when I cool it down to 170°C.
I started with one of the examples (Gamma_alpha_FeC_Acicular). I understood that we cannot mimic the martensitic transformation in MICRESS and so we consider it to be Acicular ferrite (the closest). I understood how this particular example works. I used my own Fe-C linearized phase diagram to define the slopes and C % in the FCC and BCC phases and created Gamma_alpha microstructure after cooling it down.
I didn't understand how to move forward to create the required microstructure (Gamma_Martensite), as at the end both austenite and martensite have same carbon content (what properties will be varying and how do I manipulate them when I want to create Gamma_Martensite). Could you please help me with this?
Thank you.
Sudhindra.
I am new to MICRESS. I am trying to simulate a microstructure evolution during cooling. Initially, I have a complete austenite microstructure which is cooled down at 20°C/s and is held at 170°C and is reheated again. I am now concentrating on the initial cooling stage. The Ms of the material is around 270°C. I expect austenite (30%) and martensite (70%) in the end microstructure when I cool it down to 170°C.
I started with one of the examples (Gamma_alpha_FeC_Acicular). I understood that we cannot mimic the martensitic transformation in MICRESS and so we consider it to be Acicular ferrite (the closest). I understood how this particular example works. I used my own Fe-C linearized phase diagram to define the slopes and C % in the FCC and BCC phases and created Gamma_alpha microstructure after cooling it down.
I didn't understand how to move forward to create the required microstructure (Gamma_Martensite), as at the end both austenite and martensite have same carbon content (what properties will be varying and how do I manipulate them when I want to create Gamma_Martensite). Could you please help me with this?
Thank you.
Sudhindra.
Re: Basics for Martensitic transformation
Dear Sudhindra,
it sounds that you did well and already created a two-phase microstructure with some similar morphology. As you used a linearized phase diagram, the "ferrite" has no real identity, so that you can call it also "martensite".
Now, the remaining problem is that martensite has the wrong composition. I see two options to address this problem:
1.) You alter thermodynamics (i.e. the linearized phase diagram) such that you get only a little bit of segregation and nearly identical C compositions. I don't like this approach so much because it incorrectly treats martensite as equilibrium phase (even global equilibrium), and the transformation will be quasi-massive (nearly no segregation, no drag by C pile up).
2.) You alter kinetics of diffusion or kinetics of solidification (or both) such that C is trapped. This is much closer to the real mechanism (although you still use local equilibrium assumption). On heating, the concentration difference would come back automatically as soon as diffusion gets fast enough.
What do you think?
Bernd
it sounds that you did well and already created a two-phase microstructure with some similar morphology. As you used a linearized phase diagram, the "ferrite" has no real identity, so that you can call it also "martensite".
Now, the remaining problem is that martensite has the wrong composition. I see two options to address this problem:
1.) You alter thermodynamics (i.e. the linearized phase diagram) such that you get only a little bit of segregation and nearly identical C compositions. I don't like this approach so much because it incorrectly treats martensite as equilibrium phase (even global equilibrium), and the transformation will be quasi-massive (nearly no segregation, no drag by C pile up).
2.) You alter kinetics of diffusion or kinetics of solidification (or both) such that C is trapped. This is much closer to the real mechanism (although you still use local equilibrium assumption). On heating, the concentration difference would come back automatically as soon as diffusion gets fast enough.
What do you think?
Bernd
Re: Basics for Martensitic transformation
Dear Bernd,
Thank you for the suggestions
The second option sounds good (it was also my idea).
Identifying the difference between Gibbs free energy of ferrite and austenite. At certain point it becomes '0' (point of diffusionless transformation, isn't it?). I am still trying to figure out how to couple this data to MICRESS and am not sure if this the right approach!!
Thank you for the suggestions
The second option sounds good (it was also my idea).
Identifying the difference between Gibbs free energy of ferrite and austenite. At certain point it becomes '0' (point of diffusionless transformation, isn't it?). I am still trying to figure out how to couple this data to MICRESS and am not sure if this the right approach!!
Re: Basics for Martensitic transformation
No, my second idea is a purely kinetic. Imagine carbon would not diffuse at all, then the two phases would always have the same carbon composition. In general, considering local equilibrium, this would mean that a higher undercooling is needed to obtain the transformation from gamma to alpha as compared to the case with fast diffusion of C (a similar effect is observed for the substitutional Elements already at much lower cooling rates, which is called "nple" or lenp).
The general condition for that is that the cooling rate is high enough to ensure that phase transformation is proceeding faster than diffusion. I think this is realistic at the low temperatures where the martensite transformation takes place.
The approach can also be easily transferred to a TQ-coupled simulation because no manipulation of the Gibbs energies is needed (a simple shift of Gibbs energies would still be possible using the "offset" keyword in the "'deltaG' options").
Bernd
The general condition for that is that the cooling rate is high enough to ensure that phase transformation is proceeding faster than diffusion. I think this is realistic at the low temperatures where the martensite transformation takes place.
The approach can also be easily transferred to a TQ-coupled simulation because no manipulation of the Gibbs energies is needed (a simple shift of Gibbs energies would still be possible using the "offset" keyword in the "'deltaG' options").
Bernd
Re: Basics for Martensitic transformation
Dear Bernd,
I have attached a driving file. Could you please tell me what is wrong with this file. No phase is growing. There was a slight trace of transformation at 30secs and then it disappears.
Thank you.
Best regards,
Sudhindra.
I have attached a driving file. Could you please tell me what is wrong with this file. No phase is growing. There was a slight trace of transformation at 30secs and then it disappears.
Thank you.
Best regards,
Sudhindra.
- Attachments
-
- FeCMn_noTQ_in.txt
- (23.04 KiB) Downloaded 260 times
Re: Basics for Martensitic transformation
Dear Sudhindra,
the problem seems to be the linearisation data of the phase diagram. In order to overrun Mn (whether in nple or para mode) you end up with a tie-line which has the original gamma composition as alpha composition. If you check your phase diagram data, that would lead to a tie-line with extremely high Mn composition in gamma, and to an extremely low equilibrium temperature. Even with "para" this effect completely prevents growth of alpha phase!
If you chose "nple" as redistribution model you can observe that the alpha phase is not vanishing (because here the back-direction is also prohibited) but never will grow. If you chose "normal", you indeed can get growth, but this is numerically wrong and corresponds to an "over-para" situation.
In my opinion, the linear phase diagram approach here gets to a limit because in order to overrun Mn a huge extrapolation of your phase diagram is used which is completely unrealistic.
I probably would rather try to couple to a database and use "nple" or "paraTQ"...
Bernd
the problem seems to be the linearisation data of the phase diagram. In order to overrun Mn (whether in nple or para mode) you end up with a tie-line which has the original gamma composition as alpha composition. If you check your phase diagram data, that would lead to a tie-line with extremely high Mn composition in gamma, and to an extremely low equilibrium temperature. Even with "para" this effect completely prevents growth of alpha phase!
If you chose "nple" as redistribution model you can observe that the alpha phase is not vanishing (because here the back-direction is also prohibited) but never will grow. If you chose "normal", you indeed can get growth, but this is numerically wrong and corresponds to an "over-para" situation.
In my opinion, the linear phase diagram approach here gets to a limit because in order to overrun Mn a huge extrapolation of your phase diagram is used which is completely unrealistic.
I probably would rather try to couple to a database and use "nple" or "paraTQ"...
Bernd
Re: Basics for Martensitic transformation
Dear Bernd,
Thank you so much.Yes, I solved the issue. Seems like, coupling with Thermocalc is the best to do when we have more than 1 alloying element.
Sudhindra.
Thank you so much.Yes, I solved the issue. Seems like, coupling with Thermocalc is the best to do when we have more than 1 alloying element.
Sudhindra.