Dendritic growth under high cooling rates

[PeGr]

Good morning everyone,

i am currently trying to simulate dendritic growth during solidification of X37CrMoV5 after EDM-processing. Due to EBSD-pictures, we do know that we are expecting austenite and ferrit after solidification.
By now, i am having a hard time to even get a stable denrite to grow, eventhough the cooling rate is not as high as it is supposed to be in the final simulation (will be around 1500 K/s).
I tried different mobility coefficients and interface energies out of various simulations which have been done before but all i get is diffuse grain growth.

Does anybody have experiences which parameters are crucial to get a stable dendrite to grow? And in what order of magnitude should these be selected?

Thanks in advance for the help!
Peter

[Bernd]

Dear PeGr,

I tried your simulation input files as you sent them using MICRESS Version 7.1, and the results seems not so bad. That means, you get solidification in form of s dendrite, nucleation of the peritectic phase at the interfaces between the dendrite and the liquid, and complete solidification with the peritectic grains overgrowing the primary delta phase. The only point is that your output interval is so small that you do not see anything of that...

In principle, getting a numerically stable setup of dendritic solidification for fast cooling boils down to having the correct grid resolution and a correct time scale for everything you request to be done periodically: Writing output, updating enthalpy, thermodynamic data, diffusion data, checking for nucleation, etc. must also be adapted to the time scale. While the first (the grid resolution) is already within reasonable range, the time scale is still missing, as said above. But you already have achieved what is the most difficult at the beginning, namely to get a stable dendrite without major numerical problems...

However, one very important question you still did not address: Which are the thermal conditions you want to simulate? You have chosen a high cooling rate, but no temperature gradient. From a physical point of view, the system has no control, because the growing dendrite is not slowing down as consequence of its growth. You should either consider a directional solidification by specifying a temperature gradient, or include latent heat. Otherwise, your dendrite would get faster and faster (if the boundary of the simulation domain would not stop it)!

Bernd

[PeGr]

Hello Bernd,

thank you for your fast response. You were absolutely right, as soon as i changed the time steps in the mentioned categories, i was able to see the dendrite grow, so thank you for the great tip!

Concerning the last passage of your response, i am not quite sure how to adress these suggestions. Perhaps you do have another suggestion, when i am explaining my simulation to you:

On top of the Heat affected Zone, which forms during electrical discharge machining and results in microstructural changes within this zone, there is another so called "white layer", which is known to be amorphous. The dendrite i am trying to simulate is now growing into this white layer druing cooling, which we know due to EBSD pictures. I am trying to simulate the dendrite from a "Bird's eye" point of view, which makes the temperature gradient (from my point of view) not necessary, since the temperature is the same in the whole domain. Please correct me if i am wrong.

Another issue is, that the dendrite is known to be 2µm in diameter maximum, so i am not sure how to stabilize the dendride as it reaches this size. By now, the dendride is growing into the liquid area, which technically right, but the surrounding phase of the dendride should not solidify cristallic but amorphous.

I am not sure, if Micress is capable of simulation these conditions, so please let me know if you think that this generally possible.

Thanks again for your help and all the best!

[Bernd]

Dear PeGr,

If "Bird's eye" view means that you are looking onto the top so that your simulation domain is along the isothermal plane, then you are right. But then, you cannot simulate nucleation of primary phase inside this plane, because the dendrites come from below. You thus need to set nuclei or initial grains in sort of a grid to resemble that, just as is done in the A006_CMSX4.dri example. This at the same time would solve your "2 µm" problem.

If I understand you correctly, you can see the dendrites experimentally, which means that they are crystalline. However, you observe some amorphous phase between the dendrites. Can you distinguish phases like fcc and bcc or martensite? Could it be that you get a massive solidification of fcc only because it has a lower diffusivity of carbon? In MICRESS you can have a massive transformation by using the "para" redistribution model for all elements including carbon.

Bernd

[PeGr]

Hello Bernd,

thank you for your support so far. You were totally right and I discarded the whole bird's eye view idea (since i had the wrong perception of it). I now changed it to the normal side-point-of-view.

After some research, i tried to add phase interaction inbetween phases 1/1, 2/2 and 3/3, which wasn't the case before. Since then, i now get a weird error which says:
"
Routine init calls routine initTwidth

tWidth_max( LIQUID: FCC_A1) = 1.1202830E-06 s
tWidth_max( LIQUID: BCC_A2) = 1.1202830E-06 s
tWidth_max( LIQUID: CEMENTITE) = 1.1202830E-06 s
tWidth_max( FCC_A1: FCC_A1) = ************* s

tWidth value too large.
With the chosen phase-field parameters,
tWidth must be less than -2.377377377377377E-014 s
STOP in routine initTwidth.
"

Not sure what to do about that or what causes it. Would you mind having a look?

P.S.: I know, that the kinetic coefficients are not right yet, i am still in an trial and error approach with these. In the end, i'd like to to come as close as possible to the experimental results of 4 dendrides on a width of 5µm and a length of around 20µm, all this under EDM conditions (cooling rate 1500 K/s)

[Bernd]

Dear PeGr,

The error message you get from MICRESS looks a bit strange, because input errors in the 1/1-, 2/2-, and 3/3-phase interactions are not correctly recognized. Instead of specifying a value for the interface energy, you input "avg 0.5". This is obviously wrong, but not directly recognized as an error. "-1" is used instead, leading to a negative time-step and the type of error you got at a later stage after input.

I will check whether this undesired error handling is still existing in the new MICRESS version...

Bernd

[PeGr]

Dear Bernd,

thanks for the fast response. I am currently using MICRESS 7.1. I recognizes that weird input configurations. Micress is normally asking me for both interface energy and kinetic coefficient. At 1/1 etc. it just wants to know the kinetic coefficient, even if i provide it with both numbers...

If i understand you right, there is nothing i can do at the moment and you will come back to me later?

Peter

[Bernd]

Dear Peter,

No, you misunderstood me - you just need to replace the wrong input "avg 0.5" by the correct value of the interface energy!

Bernd