Noise at interface in phase-field

[Vitaliy]

Hello,

I was asking about the addition of a noise to phase-field calculations, and I was advised to use mask noise (normal_masked.i). I have created an example input file and added there a noise (lines 68-96) to the interface by using h function h = eta^3*(6.0*eta^2-15.0*eta+10.0) and then I used material_property_names dh:=D[h,eta] to make that function valid only at the interface. You may see some results in the attachment (I believe they are self-explanatory). However, regardless of the function and values of the noise, there are no changes in free energy (correspondingly in eta shape). In the attachment, there is only one example, however, I tried many different functions and values (even extreme values), but no effect on free energy. Can somebody help me with the addition of a noise to phase-field calculations?

  [./mask_function]

    type = DerivativeParsedMaterial

    f_name = mask_function

    args = 'eta '

    material_property_names = 'eps dh:=D[h,eta]'

    constant_names = 'const'

    constant_expressions = '20.2'

    function = 'const*dh*eps'

    derivative_order = 0

    outputs = exodus

  [../]

Thanks,

Vitaliy

[Daniel Schwen]

Vitaly,

try cranking up the amplitude to eleven (actually more like 1e3 - 1e4). There is a competition between the AC dynamics and the noise term, where AC tries to smooth out the noise while it is being added. Your mobility is quite high (3.3e3), compared to the interface width, which seems to strongly suppress the added noise. Technically you don't need to use the conserved noise system for a non-conserved order parameter (I guess I won't hurt though).

Daniel

The animation attached is your input run with Kernels/conserved_langevin/amplitude=2e3

[Vitaliy]

thanks, Daniel! I've adjusted amplitude for a desire result (see attachment), since we observe bush-like growth in certain conditions. I did not expect that the AC dynamics smooths so significantly a noise. Before by saying I used extreme values I meant 100 or 200, but not 1e3.

Thanks,

Vitaliy

[Daniel Schwen]

You might want to consider putting the conserved noise on the temperature field (as thermal fluctuations). Maybe that can be done in a more rigorous way than perturbing the phase parameter...

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/79a335ed-981a-41e5-9839-e45ab8fea995%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Vitaliy]

In our case temperature is 25 C and it does not change in experiments. The input I have attached in the first post, this is just a test case (clearly defined and fast to run), my real model is electrochemicaly driven, where the initiation of a bush-type dendrite is governed by surface morphology. However, from experiment we do not know anything about that initial roughness of the surface. So I believe by applying a noise to the surface (eta parameter) is a good approximation. However, we do have info from experiments about the radius of a bush and its progress in time, so I will try to fit parameters to get close possible to experiments.