Primary Dendrite Arms spacing (PDAS) prediction

[Bernd]

Hi Chamara,

that is really hard to tell - somewhere between 1.E-3 and 10. for the dendrite tip. For getting the end of solidification numerically stable, I often reduce it T-dependent down to maybe 1.E-8...

Bernd

[CharMIC]

Hej Bernd,

I am now trying with the systematic method that you suggest to get PDAS.

But I was wondering whether is there any way to add some noise to the initial grain interface.

e.g: define initial rectangular grain and add some nice to the interface profile. then depend on the solidification rate automatically selected dendrites will starts to grow.

Chamara

[Bernd]

Hi Chamara,

you can add noise in the dG-options:

# 'DeltaG' options: default
# avg ...[] max ...[J/cm^3] smooth ...[°] noise ...[J/cm^3] offset ...[J/cm^3]
avg 0.5 noise 20.0
...

Bernd

[CharMIC]

Hej Bernd,

I only want to add some noise to the interface at the beginning of the simulation, to the initial rectangular grain interface. So that it will not take longer time to grow dendrites.

But if I define that noise as you have mentioned, it will add noise during the whole simulation time?

Chamara

[Bernd]

Hi Chamara,

you are right.

What you also could do is read the initial rectangular grain from a file which is shaped with small irregularities which serve as noise. If you now how it must look like, you could prepare such a file by hand using a text editor, starting from an "empty" ASCII output created by DP_MICRESS from a .korn or .phas output.

Alternatively, you can set a high number of smaller rectangular grains instead, or use my original idea to set only one grain in the corner which starts forming branches.

Bernd

[CharMIC]

Hej Bernd,

I have a question.

The dendrite tips that resulted in the simulations looks sharp, like a triangle.

Is this due to the resolution is not fine enough ? I used 0.2µm cell dimension. Domain size 100µm X 100µm

[Bernd]

Dear Chamara,

This is certainly true.

However, the "sharp" shape also be due to a too low interface mobility. Did you use the automatic mobility ("mob_corr")? If not, you need to use the .driv output to check whether the interface mobility is at least inside a reasonable range.

Bernd

[CharMIC]

Hej Bernd,

I used the "mob_corr" option. But still I get the sharp interface.

Any suggestions.

I also tried to reduce the resolution for 2nm with a domain size of 50µm X 100µm

But MICRESS did not allow to run it. It said I will run out of memory and asked to reduce the size of the domain.

Is there any way to forcefully run it.

BR
Chamara

[Bernd]

Hi Chamara,

If you get such a message from MICRESS, you already have run out of memory, i.e. certain arrays could not be allocated with the specified size. If you say you tried with 25000 x 50000 cells, I am not astonished, because each individual double precision array already needs around 12 GB of memory, and there are many of them...

I would try to increase resolution step-wise by factors of 2, starting from the 200nm which you had before. How far you can go will be mainly a question of simulation time, probably long before you reach any memory limit.

Bernd

[CharMIC]

Hej Bernd,

I am ruining Linux version of the MICRESS. I get following error frequently now.
==========================================================================
Intermediate output for t = 5.00000E-02 s
CPU-time: 325736 s
Current phase-field solver time step = 6.06E-07 s
Unable to read from file Results_NiCrFeNb_PDAS_Try25/NiCrFeNb_PDAS_Try25G.korn
STOP in routine outScroll
==========================================================================
I have check the file NiCrFeNb_PDAS_Try25G.korn. It is in the right location.

I will send you the file through email since it is too large.

What could be the reason here. How do I trouble shoot it to find out the root of the error.

BR
Chamara