Hi MICRESS forum,
I'm using MICRESS to research SDAS coarsening on SX Ni-Al Binary systems (using a linearised phase diagram approach, not ThermoCalc), and am having issues with small liquid "bubbles" forming between secondary arms before they join up completely. This appears to replace the influence of ostwald ripening/coarsening. I have already enabled the anti trapping term (atc mob_corr), and I am using the following parameters:
Diffusion in solid: 2*e-9
Diffusion in liquid: 2*e-5
Interfacial energy: 1.6*e-5
Grid: 1000x1x1800
Grid spacing: 0.5 micron
Interface thickness: 4 cells
Any ideas as to how to reduce/eliminate this would be greatly appreciated!
Thanks
Ben
Dendrite "bubbles" forming
-
Ben_Guntrip
- Posts: 2
- Joined: Tue Jul 28, 2026 10:57 am
- anti_bot: 333
Dendrite "bubbles" forming
You do not have the required permissions to view the files attached to this post.
Re: Dendrite "bubbles" forming
Dear Ben,
Welcome to the MICRESS forum.
What you observe is an artefact of the phase-field method. Due to the interface thickness at the µm-level, coalescense is strongly exaggerated, leading to the observed liquid inclusions. This effect cannot be avoided by using ATC but only by reducing the interface thickness.
You can reduce (absolute) the interface thickness by two ways:
a) Increase resolution, i.e. reduce the grid spacing with simultaneous increasing of the number of grid cells in each direction. Indeed, the formation of such circular inclusions is a good indicator for a poor grid resolution. Increasing grid resolution is the best solution, but it comes with a huge increase of computation time.
b) Decrease the interface thickness (in cells), keeping the grid resolution untouched. This measure deteriorates the discretisation of the interface region (you should never go below 2.5 cells!), but comes without computational cost.
Deciding on the optimal value of the interface thickness η (in cells) is a controverse discussion (even in our house). My opinion is that 3 cells is perfect for alloy simulations (i.e. concentration coupled, like your case), while uncoupled simulations (e.g. grain growth) require an interface thickness of 4 cells. Thus, my advice would be to reduce the interface thickness to 3 cells as a first step. If this is not sufficient, try to increase resolution as much as required to remove the artefact.
Best wishes
Bernd
Welcome to the MICRESS forum.
What you observe is an artefact of the phase-field method. Due to the interface thickness at the µm-level, coalescense is strongly exaggerated, leading to the observed liquid inclusions. This effect cannot be avoided by using ATC but only by reducing the interface thickness.
You can reduce (absolute) the interface thickness by two ways:
a) Increase resolution, i.e. reduce the grid spacing with simultaneous increasing of the number of grid cells in each direction. Indeed, the formation of such circular inclusions is a good indicator for a poor grid resolution. Increasing grid resolution is the best solution, but it comes with a huge increase of computation time.
b) Decrease the interface thickness (in cells), keeping the grid resolution untouched. This measure deteriorates the discretisation of the interface region (you should never go below 2.5 cells!), but comes without computational cost.
Deciding on the optimal value of the interface thickness η (in cells) is a controverse discussion (even in our house). My opinion is that 3 cells is perfect for alloy simulations (i.e. concentration coupled, like your case), while uncoupled simulations (e.g. grain growth) require an interface thickness of 4 cells. Thus, my advice would be to reduce the interface thickness to 3 cells as a first step. If this is not sufficient, try to increase resolution as much as required to remove the artefact.
Best wishes
Bernd
-
Ben_Guntrip
- Posts: 2
- Joined: Tue Jul 28, 2026 10:57 am
- anti_bot: 333
Re: Dendrite "bubbles" forming
Thanks Bernd, this is very useful. I will apply the changes you advise.