Mysterious Disappearance of Phase 1:

ripening phenomena, dislocations, grainboundary topology
yamaka
Posts: 3
Joined: Thu Mar 05, 2020 7:57 pm
anti_bot: 333

Mysterious Disappearance of Phase 1:

Hello everyone!

I am having a bit of trouble with my grain growth file. I am trying to run a file with 1000 grains with 2 phases. However, every time I run the file, it seems that one of my phases disappears after the first step. Could I get some advice about why this is happening?
Attachments
GrainGrowth1000Benchmark2Proper.in

Bernd
Posts: 1155
Joined: Mon Jun 23, 2008 9:29 pm

Re: Mysterious Disappearance of Phase 1:

Hi yamaka,

In your example you define a two phase microstructure, and you want to see grain growth. However, there also can be a phase transformation between phases 1 and 2.
In your model, you have no coupling to concentration. That means, you have something like two phases of the same material, like graphite and diamond or like water and ice. The driving force for phase transformation is given by the difference between local temperature and equilibrium temperature, multiplied by the transformation entropy:

ΔG=-ΔTΔS

In your example, ΔT=1K, and ΔS=1 J/(cm**3 K), so ΔG = -1 J/cm**3.

This is a quite big value in comparison to curvature which is responsible for grain growth. As a consequence, phase 1 will completely transform to phase 2 in a quite short time. That is what you observe.

The fundamental question is what is your aim when defining two phases. Should the phase number be just like a marker for two different sets of grains, which otherwise ripen between each other normally? Then you just should set the temperature to 1000K (in order to have ΔG=0) and use identical parameters for 1/1, 1/2, and 2/2 interfaces (what you already do).
Or do you have really two phases in mind which do not interact at this time scale because slow diffusion of imaginary chemical components (e.g. austenite and ferrite in a steel)? Then you would expect ripening only inside each phase, while the phase boundaries cannot move. In that case you should set the interface mobility of the 1/2 interface to zero (temperature then doesn't matter because the interface cannot move anyway).

Just a further recommendation: For not losing too much time while testing, you can temporarily reduce the domain size (e.g. to 160 x 160 cells) without further changes. You can simply change back when you finished testing.

Best regards

Bernd