Dear Bernd,
The calculation of dendritic growth was performed under consideration of two phases (liquid and fcc_a1). Unfortunately the concentration of Al (within the liquid) decreases at a early stage of solidification (fraction liquid = ca. 19 %; T = 1596 K). How to prevent such segregation artifacts? Of course results from microprobe indicate the maximum concentration of Al in the interdendritic regions.
I assume that changes in phMin, Interface thickness and/or phase interaction data (especially averaging driving force) would only affect segregation behavior at the very last stage of solidification (very small amount of fraction liquid).
Kind regards.
Martin
Segregation
Segregation
- Attachments
-
- Figure 1
- Al.jpg (21.79 KiB) Viewed 2282 times
-
- CMSX4_quer_UC_dri.txt
- Driving File
- (12.81 KiB) Downloaded 228 times
Re: Segregation
Hi Martin,
A good check whether the results are realistic or not is to compare to a Thermo-Calc Scheil simulation which I did with the alloy composition you specified:
As you can see, the Al composition in the melt is increasing until GAMMA_PRIME is formed, then it is decreasing.
But formation of GAMMA_PRIME is not included in your MICRESS simulation. Therefore we have to compare to a metastable Scheil simulation with only LIQUID and FCC:
Now you see that without GAMMA_PRIME formation, the Al composition in the melt is further increasing, but decreasing at a later stage.
The important question is now why in your MICRESS simulation the Al content is starting to decrease already at about T=1600K, while in the Scheil simulation it starts to decrease below T=1500K.
Of course, there could be numerical effects involved, if the interface thickness is not small compared to the diffusion length, if phMin is very big or, especially, if the interface profile would be destroyed due to poor numerical parameters ("semisolid" dendrite). But these effects would be especially strong at the beginning when solidification is fast - but here the agreement between the MICRESS and Scheil result seems to be quite reasonable...
My guess is that it rather has to do with the Scheil approximation itself, i.e. no diffusion in the solid and infinite diffusion in the liquid. Especially Al could be diffusing quite fast in FCC, you can look up the values in the .diff or .TabD outputs of MICRESS. This would explain a "back-diffusion" of Al into the dendrite.
Also, the non-infinite diffusion of Al in LIQUID could lead to a pile-up in front of the dendrite and, hence, a reduction of the segregation of this element.
This would mean that, at least up to the temperature where GAMMA_PRIME would appear, your results are realistic. If you compare to microprobe measurements, you should bear in mind that you measured Al compositions in the solid state (and maybe inside GAMMA_PRIME?), and that you did not allow GAMMA_PRIME to form in the simulation!
My advice is to repeat the MICRESS simulation without solid diffusion and to check whether this makes the difference.
Bernd
A good check whether the results are realistic or not is to compare to a Thermo-Calc Scheil simulation which I did with the alloy composition you specified:
- Scheil_all_phases.PNG (7.73 KiB) Viewed 2267 times
But formation of GAMMA_PRIME is not included in your MICRESS simulation. Therefore we have to compare to a metastable Scheil simulation with only LIQUID and FCC:
- Scheil_metastable.PNG (4.15 KiB) Viewed 2268 times
The important question is now why in your MICRESS simulation the Al content is starting to decrease already at about T=1600K, while in the Scheil simulation it starts to decrease below T=1500K.
Of course, there could be numerical effects involved, if the interface thickness is not small compared to the diffusion length, if phMin is very big or, especially, if the interface profile would be destroyed due to poor numerical parameters ("semisolid" dendrite). But these effects would be especially strong at the beginning when solidification is fast - but here the agreement between the MICRESS and Scheil result seems to be quite reasonable...
My guess is that it rather has to do with the Scheil approximation itself, i.e. no diffusion in the solid and infinite diffusion in the liquid. Especially Al could be diffusing quite fast in FCC, you can look up the values in the .diff or .TabD outputs of MICRESS. This would explain a "back-diffusion" of Al into the dendrite.
Also, the non-infinite diffusion of Al in LIQUID could lead to a pile-up in front of the dendrite and, hence, a reduction of the segregation of this element.
This would mean that, at least up to the temperature where GAMMA_PRIME would appear, your results are realistic. If you compare to microprobe measurements, you should bear in mind that you measured Al compositions in the solid state (and maybe inside GAMMA_PRIME?), and that you did not allow GAMMA_PRIME to form in the simulation!
My advice is to repeat the MICRESS simulation without solid diffusion and to check whether this makes the difference.
Bernd