Hello,
I am wondering about the effect of the setting of "# Number of steps to adjust profiles of initially sharp interfaces?". We are using initially measured composition fields. It is the microstructure of a superalloy containing gamma matrix and eutectic gamma prime phase. These eutectic regions are set in MICRESS initially as gamma prime and the rest as gamma. It means that the region around the gamma prime eutectic is oversaturated regarding gamma prime phase formation up to temperature close to the dissolution of theses particles. I allow for nucleation of gamma prime phase in the matrix. As expected the gamma phase is nucleating and growing. However, depending on the initial choice of the number of steps mentioned above, an unwished effect occurs: Setting this value to 1 seemed to be the choice giving the most realistic results. However, this leads to a complete dissolution of gamma prime in the second time step of the simulation followed by nucleation of my (600 allowed) gamma prime particles and a subsequent regrowth of the original gamma prime and the new gamma prime around. The problem is that the original gamma prime phase now consists of many different grains and I need to choose a large number of allowed gamma prime nuclei (600) in order to have enough newly formed (small) gamma prime phase around the original eutectics. This increases the calculations times massively and would not be required if the gamma prime would not disappear in the first step.
Maybe you have some ideas on how to prevent this?
Best regards,
Ralf
Effect of "Number of steps to adjust profiles of initially"
Re: Effect of "Number of steps to adjust profiles of initial
Hi Ralf,
I am not sure whether I understand the problem correctly: You read in the initial microstructure from file, i.e. the grain structure containing the grain regions for gamma prime, and the composition fields. That means, at the beginning, you have an initial microstructure which consists mainly of gamma phase (probably single crystal) and some gamma prime islands. The interfaces between gamma and gamma prime would be sharp if you would not use these initialisation steps to get a smooth interface profile. Is this correct so far?
With this approach, you may get trouble for different reasons:
- If you chose the number of steps "to adjust the initially sharp profiles" too large (say several 100's or 1000's), small gamma prime particles may just have vanished. This is an artifact which comes from the mixed calculation of curvature and interface stabilisation inside MICRESS and which can be avoided by choosing a small number of steps (~5-10). On the other hand, this "effect" can be used for fine-tuning the fraction of this phase (see here). But this is probably not your problem...
- If you read composition and grain structure from file, there is a strong risk that the data are not consistent in the interface region. This is especially the case if the information on the phase boundaries and on the compositions stem from different sources. But even if you obtain the positions of the phase boundary from the concentration mappings (e.g. by using a threshold value), the problem is that the concentration profile across the interface will neither correspond to a sharp nor to a diffuse interface as produced by the initialisation steps. In such cases, this inconsistency, which may lead to fast transformation in the first time steps, also strongly depends on the number of initialisation steps for the interface!
- Experimentally obtained concentration profiles often not only show a high level of noise (including negative values!), but also may be systematically biased. This has no strong consequences if only diffusion is calculated, but may lead to strange behaviour if phase transformations are included in the simulation!
What can you do? First of all, you can try to remove systematic inconsistencies in the data like noise, negative compositions, or shifts between concentration and grain distributions.
Furthermore it is important to stabilize the simulation in the first steps. This can be done in different ways, e.g.
- by use of the "max" option for the driving force in order to prevent extremely high values in the first steps
- reduction of the interface mobility in the first few steps to prevent dissolution of gamma prime before the system has "stabilized" (e.g. using "restart")
Please tell me whether this information was helpful for your kind of problems!
Bernd
I am not sure whether I understand the problem correctly: You read in the initial microstructure from file, i.e. the grain structure containing the grain regions for gamma prime, and the composition fields. That means, at the beginning, you have an initial microstructure which consists mainly of gamma phase (probably single crystal) and some gamma prime islands. The interfaces between gamma and gamma prime would be sharp if you would not use these initialisation steps to get a smooth interface profile. Is this correct so far?
With this approach, you may get trouble for different reasons:
- If you chose the number of steps "to adjust the initially sharp profiles" too large (say several 100's or 1000's), small gamma prime particles may just have vanished. This is an artifact which comes from the mixed calculation of curvature and interface stabilisation inside MICRESS and which can be avoided by choosing a small number of steps (~5-10). On the other hand, this "effect" can be used for fine-tuning the fraction of this phase (see here). But this is probably not your problem...
- If you read composition and grain structure from file, there is a strong risk that the data are not consistent in the interface region. This is especially the case if the information on the phase boundaries and on the compositions stem from different sources. But even if you obtain the positions of the phase boundary from the concentration mappings (e.g. by using a threshold value), the problem is that the concentration profile across the interface will neither correspond to a sharp nor to a diffuse interface as produced by the initialisation steps. In such cases, this inconsistency, which may lead to fast transformation in the first time steps, also strongly depends on the number of initialisation steps for the interface!
- Experimentally obtained concentration profiles often not only show a high level of noise (including negative values!), but also may be systematically biased. This has no strong consequences if only diffusion is calculated, but may lead to strange behaviour if phase transformations are included in the simulation!
What can you do? First of all, you can try to remove systematic inconsistencies in the data like noise, negative compositions, or shifts between concentration and grain distributions.
Furthermore it is important to stabilize the simulation in the first steps. This can be done in different ways, e.g.
- by use of the "max" option for the driving force in order to prevent extremely high values in the first steps
- reduction of the interface mobility in the first few steps to prevent dissolution of gamma prime before the system has "stabilized" (e.g. using "restart")
Please tell me whether this information was helpful for your kind of problems!
Bernd
Re: Effect of "Number of steps to adjust profiles of initial
Dear Bernd,
in fact this is no critical issue for us. As long as I am not allowing for nucleation of new gamma prime phase within the matrix, the microstructure is stabilized very easily by MICRESS in the first steps. This probably happens because we anyway limit the phase boundary mobility at lower temperatures in order to prevent an extreme growth of the primary gamma prime into the oversatured matrix.
The problem I had now was probably that I had to increase the phase mobility also at lower temperatures. Your idea to limit the driving force for the gamma prime phase is resolving the issue in principle, however I would have to remove this option at later times by restarting, because it also prevent the growth of the (in this case required) secondary gamma prime nucleis in the matrix.
I am only wondering why the new gamma prime grains attached to each other are not categorized?
Best regards,
Ralf
in fact this is no critical issue for us. As long as I am not allowing for nucleation of new gamma prime phase within the matrix, the microstructure is stabilized very easily by MICRESS in the first steps. This probably happens because we anyway limit the phase boundary mobility at lower temperatures in order to prevent an extreme growth of the primary gamma prime into the oversatured matrix.
The problem I had now was probably that I had to increase the phase mobility also at lower temperatures. Your idea to limit the driving force for the gamma prime phase is resolving the issue in principle, however I would have to remove this option at later times by restarting, because it also prevent the growth of the (in this case required) secondary gamma prime nucleis in the matrix.
I am only wondering why the new gamma prime grains attached to each other are not categorized?
Best regards,
Ralf
Re: Effect of "Number of steps to adjust profiles of initial
Dear Ralf,
"Categorization" is a MICRESS feature which was introduced in order to reduce the number of grains. High numbers of grains lead to a high administrative memory and CPU usage and can be avoided when several spatially separated grains are assigned to the same grain number. This allows for having huge numbers of precipitates, like gamma prime in your case, which behave like individual particles but are not counted as such.
In such a scenario, it is not wanted that particles which touch each other should coalesce and join to a single particle. Therefore, and also because removing existing interfaces would make the process much more complex, grains are only "categorized" if they do not yet touch!
More specifically, the conditions for categorization are:
1.) Grains must have the same phase, orientation (category), and reX energy (if applicable)
2.) Grains must have reached the end of their shield time and full size (or have been "killed" after end of shield time)
3.) Grains to be categorized must not touch each other
4.) Grains to be categorized not have both contact to a same third grain which has a grain number which in between the grain numbers of the first two grains (this is because then the interaction order would partially change, leading to a huge administrative effort...)
Criteria 3.) and 4.) could in principle be removed with a quite big programming effort. Up to now, there was not strong enough motivation to so so...
For preventing coalescence in case of very dense particles, it makes sense to define several orientation categories so that almost all particles are not in direct contact to particles of the same grain.
Bernd
"Categorization" is a MICRESS feature which was introduced in order to reduce the number of grains. High numbers of grains lead to a high administrative memory and CPU usage and can be avoided when several spatially separated grains are assigned to the same grain number. This allows for having huge numbers of precipitates, like gamma prime in your case, which behave like individual particles but are not counted as such.
In such a scenario, it is not wanted that particles which touch each other should coalesce and join to a single particle. Therefore, and also because removing existing interfaces would make the process much more complex, grains are only "categorized" if they do not yet touch!
More specifically, the conditions for categorization are:
1.) Grains must have the same phase, orientation (category), and reX energy (if applicable)
2.) Grains must have reached the end of their shield time and full size (or have been "killed" after end of shield time)
3.) Grains to be categorized must not touch each other
4.) Grains to be categorized not have both contact to a same third grain which has a grain number which in between the grain numbers of the first two grains (this is because then the interaction order would partially change, leading to a huge administrative effort...)
Criteria 3.) and 4.) could in principle be removed with a quite big programming effort. Up to now, there was not strong enough motivation to so so...
For preventing coalescence in case of very dense particles, it makes sense to define several orientation categories so that almost all particles are not in direct contact to particles of the same grain.
Bernd