Hi. In my simulation, I'm getting a steady fluctuation in the value of diffusion coefficient reported in TabD file(Please see the 3rd column in the image below). I was expecting a steady state value for the diffusion coefficient. What could be the reason for this fluctuation? Should I be worried about this? Is this fluctuation an indication of any numerical instability?
Thanks and regards,
Deepu
Reg. fluctuation in diffusion coefficients
Reg. fluctuation in diffusion coefficients
- Attachments
-
- micress_fourm_diff_coeff_fluctuation_problem.png (16.2 KiB) Viewed 3264 times
Re: Reg. fluctuation in diffusion coefficients
Dear Deepu,
As it appears from your data, temperature is constant, therefore it must be a change of the average phase composition which causes the fluctuation. You should be able to see that with more detail in the .diff text output file.
Whether this should worry you or not depends on circumstances - you should in any case investigate where it comes from!
Bernd
As it appears from your data, temperature is constant, therefore it must be a change of the average phase composition which causes the fluctuation. You should be able to see that with more detail in the .diff text output file.
Whether this should worry you or not depends on circumstances - you should in any case investigate where it comes from!
Bernd
Re: Reg. fluctuation in diffusion coefficients
Dear Bernd,
Thanks for the reply. Yes, my average composition is fluctuating as shown in the snapshow below. What could be the reasons for this fluctuation?
Deepu
Thanks for the reply. Yes, my average composition is fluctuating as shown in the snapshow below. What could be the reasons for this fluctuation?
Deepu
- Attachments
-
- flustuation_composition.png (140.5 KiB) Viewed 3255 times
Re: Reg. fluctuation in diffusion coefficients
Dear Deepu,
oscillation seems to exist only or mainly for Mn composition, at the same frequency as updating of diffusion data - I guess that you update thermodynamic data also at that frequency. If this is the case, relinearisation is the trigger for this oscillation!
In general, there may be different reasons for such oscillations. To my experience, oscillations which involve relinearisation of thermodynamic data often vanish when the updating frequency is increased. Using "global" or "globalF" updating scheme, you should be able to afford that.
If this does't help, check out whether there are other irregularities: Do you get error messages, how does the concentration field of Mn looks like?
Bernd
oscillation seems to exist only or mainly for Mn composition, at the same frequency as updating of diffusion data - I guess that you update thermodynamic data also at that frequency. If this is the case, relinearisation is the trigger for this oscillation!
In general, there may be different reasons for such oscillations. To my experience, oscillations which involve relinearisation of thermodynamic data often vanish when the updating frequency is increased. Using "global" or "globalF" updating scheme, you should be able to afford that.
If this does't help, check out whether there are other irregularities: Do you get error messages, how does the concentration field of Mn looks like?
Bernd
Re: Reg. fluctuation in diffusion coefficients
Dear Bernd,
Thanks for the suggestion. I'm working it out now.
Out of curiosity, what will happen if I use database coupling and then use no linearization(relinearization mode=none), and still update diffusion coefficients? Does it mean my phase diagram is not updated, but my diffusion coefficient is?
How is the above strategy different from using a 'linearized phase diagram' description in micress?
Do you have any manual which tell what a particular error means? I'm meaning the error codes and the corresponding meaning.
Deepu
Thanks for the suggestion. I'm working it out now.
Out of curiosity, what will happen if I use database coupling and then use no linearization(relinearization mode=none), and still update diffusion coefficients? Does it mean my phase diagram is not updated, but my diffusion coefficient is?
How is the above strategy different from using a 'linearized phase diagram' description in micress?
Do you have any manual which tell what a particular error means? I'm meaning the error codes and the corresponding meaning.
Deepu
Re: Reg. fluctuation in diffusion coefficients
Dear Deepu,
With respect to your first question, you are right, in principle: If you specify database coupling for thermodynamic data with update options "none", data are calculated when they are used first time, and then not updated anymore afterwards. Nevertheless, the behaviour may be different for different linearisation options: if you chose "local" (default), each cell at the interface has its own set of linearisation parameters. This implies that newly formed cells get new linearisation data at the moment when they are created. This has the consequence that in this case a rapidly moving interface will be always "up to date" even without specifying any updating. This is different in case of "global" or "globalF" relinearisation modes, where newly formed cells also adopt the common set or linearisation parameters. Note furthermore that the updating schemes for "global relinearisation mode" and for each individual interface are additive, i.e. even if you specify "none" for the "global relinearisation mode" you may still define updating of specific interfaces.
Updating of diffusion data is completely independent from updating of thermodynamic data.
Even if using "global" relinearisation without updating, it is different from using a linear phase diagram in three respects:
1.) The format of the linearisation data corresponds to the "linearTQ" rather that to the "linear" format
2.) If not present at the beginning, linearisation data are calculated for the local conditions where the respective interface is formed by nucleation.
3.) Linearisation data are calculated for interfaces between grains, not phases. That means e.g. for case of precipitates in a matrix that each particle has its own set of parameters for its interaction with the matrix phase.
Unfortunately, we do not have a complete list of error messages in MICRESS. The reason is that they are changing rapidly and that they often are not conclusive for a user without insight to the code, but this is not a good excuse. I should provide a list as soon as possible, at least here in the forum
Bernd
With respect to your first question, you are right, in principle: If you specify database coupling for thermodynamic data with update options "none", data are calculated when they are used first time, and then not updated anymore afterwards. Nevertheless, the behaviour may be different for different linearisation options: if you chose "local" (default), each cell at the interface has its own set of linearisation parameters. This implies that newly formed cells get new linearisation data at the moment when they are created. This has the consequence that in this case a rapidly moving interface will be always "up to date" even without specifying any updating. This is different in case of "global" or "globalF" relinearisation modes, where newly formed cells also adopt the common set or linearisation parameters. Note furthermore that the updating schemes for "global relinearisation mode" and for each individual interface are additive, i.e. even if you specify "none" for the "global relinearisation mode" you may still define updating of specific interfaces.
Updating of diffusion data is completely independent from updating of thermodynamic data.
Even if using "global" relinearisation without updating, it is different from using a linear phase diagram in three respects:
1.) The format of the linearisation data corresponds to the "linearTQ" rather that to the "linear" format
2.) If not present at the beginning, linearisation data are calculated for the local conditions where the respective interface is formed by nucleation.
3.) Linearisation data are calculated for interfaces between grains, not phases. That means e.g. for case of precipitates in a matrix that each particle has its own set of parameters for its interaction with the matrix phase.
Unfortunately, we do not have a complete list of error messages in MICRESS. The reason is that they are changing rapidly and that they often are not conclusive for a user without insight to the code, but this is not a good excuse. I should provide a list as soon as possible, at least here in the forum
Bernd