Source and sink terms in Cahn-Hilliard equation

[mdh23rct]

Dear MOOSE experts,
Would you be kind to let me know the way of adding sink and source terms in the Cahn-Hilliard equation, to be specific in MOOSE!

I would also appreciate if you could provide me a step-by-step guideline to do so, if possible!

Thanks for your support and time.

Regards.

[GiudGiud]

Hello

How is the source or sink term calculated?
What does it depend on?

I would recommend looking through the phase field module for examples with such terms

Guillaume

[GiudGiud]

ok so because you need to solve for psi on every time step, psi needs to be a nonlinear variable.
Now that this is established, you cannot use psi as the variable for an auxkernel. auxkernels only act on auxiliary variables.

This:

[Functions]
  [./psi_func]
    type = ParsedFunction
    expression = 'psi>0.8'
  [../]
[]

should be this:

[Functions]
  [./psi_func]
    type = ParsedFunction
    expression = 'psi>0.8'
    coupled_variables = 'psi'
  [../]
[]

[mdh23rct]

Dear Dr. Guillaume,
Thank you for your reply. I have turned off the Auxkernel and added coupled variables. Unfortunately, it is not working. The parsed function is not recognizing the expression 'psi>0.8'. Because the expression can be a function of time or coordinate, but not of the non-conserved order parameter. What I can do now? Is there any advice from your side to solve this problem?

Thanks.

[GiudGiud]

Oh you’re right my bad. Functions cannot depend on variables.

so the user object that modifies the mesh cannot use anything else than a function ?

[mdh23rct]

Dear Dr. Guillaume,
Thank you for your reply. I am not an expert in MOOSE. That's why I don't know other ways to eliminate elements based on the non-conserved order parameter.

Actually, it should be straight forward to turn off the elements based on a solution variable. I did it in other solvers. Would you be kind to enquire, if any colleague of yours has solved such a problem earlier!!

Thanks.

[GiudGiud]

@dschwen @hugary1995 have you seen this done?

[GiudGiud]

I think this is the object to use
https://mooseframework.inl.gov/source/userobjects/CoupledVarThresholdElementSubdomainModifier.html

it works with a threshold on a variable, so you can use psi as the variable directly

[GiudGiud]

@jackeggemeyer you could request access to marmot. No guarantees and it will take some time, but best start early.

https://ncrcaims.inl.gov/Identity/Account/Login

[jackeggemeyer]

@GiudGiud Thanks! I applied a little bit ago, and my application is in review. My advisor wants me to recreate some kernels for practice. Would remaking this kernel be that difficult?
P.S.
I imagine you guys also can't give me much assistance due the kernel Im trying to recreate is restricted.

[laagesen]

I would say it would be a pretty straightforward task and a good first coding project. You can look at many of the kernels in the MOOSE phase-field module for examples. We can help out by giving feedback to your code and answering questions, as long as we are not copying/pasting code from other repos we should be fine.

[zahidhasan83]

Dear @jackeggemeyer ,
The recombination term depends upon the atomic fraction of species. If we talk about vacancies and interstititals, many paper assume the atomic fraction of interstitials negligible, thus the recombination is not included in the model. You can dump the vacancies at the grain boundaries using a sink term, and/or apply supersaturation of vancacies to generate voids inside grains. You can live with the source and sink terms considering model simplifications. Else, you generate your recombination kernel. If you need any help on setting up the model, please drop me a message. I will try to respond as soon as I can. Thanks.

[jackeggemeyer]

@laagesen & @zahidhasan83
Thank you so much for the insight, I am working on my .c and .h file for recombination right now. When I run into an issue that is not syntax debugging I will make a thread and tag your accounts if that's alright!

[jackeggemeyer]

@zahidhasan83 I was unable to contact you in any other way or tag you on a new thread so apologies for continuing this here if you are still able to help. I am still working on my Phase Field Model with sink and recombination. I have added source and sink using MaskedBodyForce however I can only get them to sink away from each other as seen in the images below.

Do you hyave any idea on why this occurs? I have swapped the signs, values and many other constants but I can only get sink to work in the opposite. Any suggestions to fix this? My input file that creates this is provided below

Also once I add recombination I can validate that recombination "works" from this paper https://www.sciencedirect.com/science/article/pii/S0927025622004839. However outside of the desnity of vacaccnies converging to a value of 0.01 which it should, the behavior of the system is poor. Convergnace at the begining of the simution also becomes really bad and gets better later in the simulation. This often creating cubes which really confuse me, or diffusing into the interface creating just one phase than randomly spindollaly decomposes multiple time steps later. It also likes to decompose into a mystery -1 phase which i cant control and happens very randomly. I have added my .c and .h files if that helps.

I understand this is a really loaded question but I am lost messing around with constants trying to get the correct behevior to happen. Do you guys have any suggestions that keep that phases more stable so they can sink arond the lattice instead of diffusing?

sinkandrecomEggemeyer.txt
GrandPotentialRecombinationHFILE.txt
GrandPotentialRecombinationCFILE.txt

[zahidhasan83]

@jackeggemeyer Dear Jack, please have my personal e-mail address: md.zahidhasan83@yahoo.de
I will break down the problem in small steps first. Before approaching the cluster problem, I would address a single void.

1. use the source term of vacancies and let the vacancies diffuse to the void. If the void grows in size, good.
2. next, do the opposite. if the void shrinks, its great.
3. Finally, introduce the recombination term in the model. I mean interstitial and vacancy recombination along with their source terms. Now see what happens.

The images you have attached look like redistribution of vacancies within the voids if I am correct. Larger voids are gowing at the expense of the small one. They are not all growing alltogether due to vacancy supersaturation.

For convergence, use automatic_scaling. Cube maybe related to the mesh resolution. I once had such a problem, mesh refinement solved the issue.

I will have a look at the files you have forwarded. Will come back as soon as I can.

Cheers,
Zahid