Hi there,
I am trying to use MICRESS for precipitation at a boundary. My simulation works perfectly when there is no grain boundary diffusion. However, when I activate +b for boundary diffusion the simulation runs but no precipitation happens. So, I think I did not fully understand the grain boundary option.
As an example for one of my element diffusion, I used the following
### definition of bulk diffusion
2 1 diagonal d
# Diff.-coefficient:
# Prefactor? (real) [cm**2/s]
124.0000000000
# Activation energy? (real) [J/mol]
130500.0000000
### definition of grain boundary diffusion as extra option
2 1 diagonal +b
nbn
65250.0000000
100
Do these values make sense? I am assuming that I need to redefine the new activation energy for the boundaries (or is it the difference?). The precipitate would form by stoichiometric reaction. Is this the reason it does not work? I am defining the grain boundary diffusion for boundaries that are not associated with the precipitate.
grain boundary diffusion
Re: grain boundary diffusion
Dear khajezade,
You should define the difference in activation energy instead of the new activation energy (which however would be exactly the same in your specific case...).
In principle, the important influence of the grain boundary diffusion is to increase the diffusion coefficient within the interface region, most pronounced at the center of the interface (φ1=φ2=0.5). As it is difficult to estimate the combined effect of the change of activation energy and the re-scaling to the physical interface thickness, the maximum value (at the center of the interface) of the grain boundary diffusion coefficient is written to the .diff output. There you can check how big the effect really is.
Your finding that nucleation does not happen anymore with grain boundary diffusion could indicate that without GB-diffusion you had a pile-up of the relevant elements only within the interface. By broadening the pile-up, you may end up below the critical undercooling. If this is the reason, the difference should vanish when you increase grid resolution.
Bernd
You should define the difference in activation energy instead of the new activation energy (which however would be exactly the same in your specific case...).
In principle, the important influence of the grain boundary diffusion is to increase the diffusion coefficient within the interface region, most pronounced at the center of the interface (φ1=φ2=0.5). As it is difficult to estimate the combined effect of the change of activation energy and the re-scaling to the physical interface thickness, the maximum value (at the center of the interface) of the grain boundary diffusion coefficient is written to the .diff output. There you can check how big the effect really is.
Your finding that nucleation does not happen anymore with grain boundary diffusion could indicate that without GB-diffusion you had a pile-up of the relevant elements only within the interface. By broadening the pile-up, you may end up below the critical undercooling. If this is the reason, the difference should vanish when you increase grid resolution.
Bernd
Re: grain boundary diffusion
Thank you for your reply! I am going to check this with a higher resolution simulation. Having said this, I have introduced a precipitate on the boundary and tried it with the grain boundary diffusion. However, it looks like the simulation has been frozen to the initial setup. no evolution!
UPDATE: I checked *.diff file. All the numbers are zero. I increased the resolution by a factor of 2 and no change in my results. Again, turning off the grain boundary diffusion seems to work perfectly.
UPDATE: I checked *.diff file. All the numbers are zero. I increased the resolution by a factor of 2 and no change in my results. Again, turning off the grain boundary diffusion seems to work perfectly.
Re: grain boundary diffusion
Dear khajezade,
I did not take into account that you are currently not using diffusion data from database - thus you will not get the information about grain boundary diffusion in the .diff output like I said. I apologize for the not optimal implementation!
To get nevertheless a feeling about the effect and strength of grain boundary diffusion, you should have a look at the composition of component 2 in phase 1 (.c2pha1 output). If you would use 0 for the change of activation energy, there should be no effect at all, and precipitation should work. Once you start putting increasing values for the activation energy, there should be an effect in .c2pha1, which should explain your findings.
I don't know your setup, and whether there are composition gradients of element 2 at the 1/1-grain boundaries at the time of nucleation of your precipitates. By the way, is grain boundary diffusion in your case suppressing nucleation itself, or rather only successful growth of the precipitates?
Bernd
I did not take into account that you are currently not using diffusion data from database - thus you will not get the information about grain boundary diffusion in the .diff output like I said. I apologize for the not optimal implementation!
To get nevertheless a feeling about the effect and strength of grain boundary diffusion, you should have a look at the composition of component 2 in phase 1 (.c2pha1 output). If you would use 0 for the change of activation energy, there should be no effect at all, and precipitation should work. Once you start putting increasing values for the activation energy, there should be an effect in .c2pha1, which should explain your findings.
I don't know your setup, and whether there are composition gradients of element 2 at the 1/1-grain boundaries at the time of nucleation of your precipitates. By the way, is grain boundary diffusion in your case suppressing nucleation itself, or rather only successful growth of the precipitates?
Bernd
Re: grain boundary diffusion
Dear khajezade,
Thanks for sending me your input files per PM.
I essentially have overlooked the problem in the input syntax of grain boundary diffusion. For activating it you should just use "+b" and not "diagonal +b". The latter is misinterpreted and results in trying to use diffusion data from database. As there is none, diffusion essentially was switched off! I am sorry for not having seen that right away...
Apart from that, I have some remarks:
- Choosing 100 nm as physical interface width is extremely large (essentially above the numerical value you have!). A typical value would be 1 nm. Using 100 nm strongly increases grain boundary diffusion. Nevertheless, you should still start using smaller values for the activation energy.
- The molar volume of MG2SI should be given per mole of atoms, not formula units. That means you should divide it by 3
- I would advise to write much more types of output (.driv, .mueS, ...) and not use vtk format if not necessary. You can switch back for production at the end if needed.
Best wishes
Bernd
Thanks for sending me your input files per PM.
I essentially have overlooked the problem in the input syntax of grain boundary diffusion. For activating it you should just use "+b" and not "diagonal +b". The latter is misinterpreted and results in trying to use diffusion data from database. As there is none, diffusion essentially was switched off! I am sorry for not having seen that right away...
Apart from that, I have some remarks:
- Choosing 100 nm as physical interface width is extremely large (essentially above the numerical value you have!). A typical value would be 1 nm. Using 100 nm strongly increases grain boundary diffusion. Nevertheless, you should still start using smaller values for the activation energy.
- The molar volume of MG2SI should be given per mole of atoms, not formula units. That means you should divide it by 3
- I would advise to write much more types of output (.driv, .mueS, ...) and not use vtk format if not necessary. You can switch back for production at the end if needed.
Best wishes
Bernd
Re: grain boundary diffusion
hi Bernd,
Thank you for your reply. I am going to test your suggestions. I'll update you on this.
Thank you for your reply. I am going to test your suggestions. I'll update you on this.