(1)
when we define the nucleation condition, we set the maximum number of nuclei (N), and the max number of sumultaneous nucleation (n). Does that means that for every check, the program search all the favorable nucleation sites and then randomly selet n sites if available sites are enough?
And what means if 0 is set for automatic? all sites (if smaller than N) or N sites will be selected at the same check for nucleation?
(2)when stabilisation is used, that means curvature term is neglected until a sufficient size. what the sufficient size? I suppose the size when initially the delta_G exceeds the curvature driving force, is that right?
what the detail of kill_metastable? Since by using stabilisation, curvature is neglected and thus nuclei would grow by only delta_G before reaching the sufficient size. How to reach a metastable state?
(3) when nucleation occurs only at the interface of two different phases, I need to assign the two parent phases to the reference phase and substrate phase respectively?
(4) how to set saterated nucleation?
(5)for a pure metal solidification, how to simulate the nucleation rate change with undercooling and temperature? It seems the nucleation rates will increase with undercooling while independent of temperature for MICRESS.
nucleation issues
Re: nucleation issues
Hi Zhubq,
I will try to answer your questions:
1.) The maximum number of nuclei (N) is the number of grains of one seed type which will be allowed to nucleate during all the simulation. This is not very physical, and N should be set high enough in order not to interfere in the normal case. Sometimes one may wish only one seed to appear, then it makes sense.
The maximum number of sumultaneous nucleations (n) is the number of grains of all seed types to be nucleated in one nucleation check. A list of possible seeds is created and ordered with respect to the undercooling or driving force for each nucleus, and then the less favourable ones are discarded, if the number of possible nucleations for all seed types in the current check is higher than n (or if the total number of nucleations of one seed type N would be exceeded). The use of n again is not very physical, it is there for historical reasons, maybe it makes sense with very high checking frequencies. Setting it to "automatic" just removes the check for n, and the number of seeds per check and seed type is only restricted if N would be exceeded.
2.) No, that is not right. The question is whether the phase-field algorithm is capable of growing a small grain without "complete" interface. Typicyally, the central cell of the new grain needs at least a fraction >0.5 to be able to grow, in triple junctions it is more complicated. Actually, in MICRESS curvature is increasing linearly with the fraction of the central cell, i.e. the full curvature is experienced only if the grain reaches "full size" (one cell with fraction >1-phMin).
3.) Yes, the substrate phase should be the one which determines the extra curvature undercooling, the reference phase is the one where the undercooling is calculated. If this is not enough, a second substrate phase can be specified, if you have the newest beta-version of MICRESS (look here).
4.) I do not understand the question...
5.) In a pure metal (i.e. without concentration coupling) you still have to specify a equilibrium temperature, and thus the nucleation undercooling depends on temperature! If there is no temperature gradient, then - like in the case of a homogeneous alloy melt - the seed density model for heterogeneous nucleation would be appropriate in order not to get regular seeding patterns.
Bernd
I will try to answer your questions:
1.) The maximum number of nuclei (N) is the number of grains of one seed type which will be allowed to nucleate during all the simulation. This is not very physical, and N should be set high enough in order not to interfere in the normal case. Sometimes one may wish only one seed to appear, then it makes sense.
The maximum number of sumultaneous nucleations (n) is the number of grains of all seed types to be nucleated in one nucleation check. A list of possible seeds is created and ordered with respect to the undercooling or driving force for each nucleus, and then the less favourable ones are discarded, if the number of possible nucleations for all seed types in the current check is higher than n (or if the total number of nucleations of one seed type N would be exceeded). The use of n again is not very physical, it is there for historical reasons, maybe it makes sense with very high checking frequencies. Setting it to "automatic" just removes the check for n, and the number of seeds per check and seed type is only restricted if N would be exceeded.
2.) No, that is not right. The question is whether the phase-field algorithm is capable of growing a small grain without "complete" interface. Typicyally, the central cell of the new grain needs at least a fraction >0.5 to be able to grow, in triple junctions it is more complicated. Actually, in MICRESS curvature is increasing linearly with the fraction of the central cell, i.e. the full curvature is experienced only if the grain reaches "full size" (one cell with fraction >1-phMin).
3.) Yes, the substrate phase should be the one which determines the extra curvature undercooling, the reference phase is the one where the undercooling is calculated. If this is not enough, a second substrate phase can be specified, if you have the newest beta-version of MICRESS (look here).
4.) I do not understand the question...
5.) In a pure metal (i.e. without concentration coupling) you still have to specify a equilibrium temperature, and thus the nucleation undercooling depends on temperature! If there is no temperature gradient, then - like in the case of a homogeneous alloy melt - the seed density model for heterogeneous nucleation would be appropriate in order not to get regular seeding patterns.
Bernd
Re: nucleation issues
Hi Mr. Bernd,
I want to verify experimental results for Stabilor G. So for simulation, i took temp. at the bottom as liquidus temp. which i got from
thermo-calc (Scheil Simulation) and the simulation runs good. Now, the thing is if try to take one of the mould temp. from the expt. e.g. 973 or 1173 K as the temp. at the bottom, i get no simulation result at all.
What could be the prob. that i don't see any nucleation at all?
Thanks in advance
Raghav
I want to verify experimental results for Stabilor G. So for simulation, i took temp. at the bottom as liquidus temp. which i got from
thermo-calc (Scheil Simulation) and the simulation runs good. Now, the thing is if try to take one of the mould temp. from the expt. e.g. 973 or 1173 K as the temp. at the bottom, i get no simulation result at all.
What could be the prob. that i don't see any nucleation at all?
Thanks in advance
Raghav
Re: nucleation issues
Dear Raghav,
thanks for your question and for beeing back to the forum!
The information you gave us is a bit sparse. I understand that the simulation is running correctly if you start the MICRESS simulation with an initial bottom temperature corresponding to the liquidus temperature of your alloy and, I guess, you apply a cooling rate. If you change the initial temperature to a different value of 973 or 1173 K, you do not get nucleation. Which is the liquidus temperature of the alloy, is it above, below or in between the two temperatures?
Due to the high number of the nucleation parameters which can be defined in MICRESS, it can easily happen that the user is wondering why no nucleation is taking place, but - due to some inadequate parameters - nucleation erroneously is even not checked. Parameters which can prevent checking for nucleation are the temperature range for nucleation, the shield time and distance, the maximum number of seeds, the checking frequency etc. Sometimes it can even happen that, because there is no interface present before the first grain is nucleated, the simulation is passing the nucleation interval without having even a single time-step inside, because the automatic time-step is too big. Then, logically, there is no check of nucleation done.
In the beta version 5.408, there is a verbose mode for nucleation which is activated by specifying the verbose keyword:
nuleation verbose
This mode gives you information about when and how nucleation is checked.
Please check whether your problem is due to something like I described, or whether we have to search in another direction!
Bernd
thanks for your question and for beeing back to the forum!
The information you gave us is a bit sparse. I understand that the simulation is running correctly if you start the MICRESS simulation with an initial bottom temperature corresponding to the liquidus temperature of your alloy and, I guess, you apply a cooling rate. If you change the initial temperature to a different value of 973 or 1173 K, you do not get nucleation. Which is the liquidus temperature of the alloy, is it above, below or in between the two temperatures?
Due to the high number of the nucleation parameters which can be defined in MICRESS, it can easily happen that the user is wondering why no nucleation is taking place, but - due to some inadequate parameters - nucleation erroneously is even not checked. Parameters which can prevent checking for nucleation are the temperature range for nucleation, the shield time and distance, the maximum number of seeds, the checking frequency etc. Sometimes it can even happen that, because there is no interface present before the first grain is nucleated, the simulation is passing the nucleation interval without having even a single time-step inside, because the automatic time-step is too big. Then, logically, there is no check of nucleation done.
In the beta version 5.408, there is a verbose mode for nucleation which is activated by specifying the verbose keyword:
nuleation verbose
This mode gives you information about when and how nucleation is checked.
Please check whether your problem is due to something like I described, or whether we have to search in another direction!
Bernd
Re: nucleation issues
Hello Mr. Bernd
Actually, the liquidus temp. i got from Thermo-Calc for Stabilor G is 1227 K. I am getting proper simulation results if i take 1213 K as the temp. at the bottom. Somehow, by trial and error i realized that first grain appears at 1213 K.
But the experimental results ( at mould temp. 973 K, 1173 K) require me to have a simulation at the aforementioned temp. Actually, when i ran a simulation with 973 K i nearly got a undercooling of 247 K and for 1173 K of 54.4 K. For the first case, there was no simulation result but in the second case the result at the first time step (0.1s) continued to remain the same till thew last time step with absolutely no changes.
The cooling rate i have taken is -10 K/s
How can we manipulate or control the value of Undercooling? Is it only largely dependent on the temp. at the bottom?
Actually, the liquidus temp. i got from Thermo-Calc for Stabilor G is 1227 K. I am getting proper simulation results if i take 1213 K as the temp. at the bottom. Somehow, by trial and error i realized that first grain appears at 1213 K.
But the experimental results ( at mould temp. 973 K, 1173 K) require me to have a simulation at the aforementioned temp. Actually, when i ran a simulation with 973 K i nearly got a undercooling of 247 K and for 1173 K of 54.4 K. For the first case, there was no simulation result but in the second case the result at the first time step (0.1s) continued to remain the same till thew last time step with absolutely no changes.
The cooling rate i have taken is -10 K/s
How can we manipulate or control the value of Undercooling? Is it only largely dependent on the temp. at the bottom?
Re: nucleation issues
Dear Raghav,
what I do not understand is how you can perform a solidification experiment at 973K if the liquidus temperature is 1227K! Even if you would pour the melt (which initially has to be above 1227K to be completely liquid) into a mold of 973K, you would get nucleation at a higher temperature than 973K and, hence, at a lower undercooling, because the melt needs time to cool down! Thus, your scenario to start at a bottom temperature of 973 is highly unrealistic, and it is not astonishing that it leads to numerical problems (what I guess is what you get if you say that there is no simulation result).
But what do you mean by "the result remains the same" for 1173K? Does is mean that the nuclei vanished after nucleation and you have only pure liquid phase?
Essentially, the value of the nucleation undercooling is only dependent on temperature, as long as the melt composition is not changing!
Bernd
what I do not understand is how you can perform a solidification experiment at 973K if the liquidus temperature is 1227K! Even if you would pour the melt (which initially has to be above 1227K to be completely liquid) into a mold of 973K, you would get nucleation at a higher temperature than 973K and, hence, at a lower undercooling, because the melt needs time to cool down! Thus, your scenario to start at a bottom temperature of 973 is highly unrealistic, and it is not astonishing that it leads to numerical problems (what I guess is what you get if you say that there is no simulation result).
But what do you mean by "the result remains the same" for 1173K? Does is mean that the nuclei vanished after nucleation and you have only pure liquid phase?
Essentially, the value of the nucleation undercooling is only dependent on temperature, as long as the melt composition is not changing!
Bernd
Re: nucleation issues
Generally, even when the o/p appears (in the above conditions with high undercooling), the size of the nuclei does not change at all after the first time step. But, of course, as you mentioned its not practical at all to be trying above conditions.
Thanks
Raghav
Thanks
Raghav
Re: nucleation issues
Hi,
I tried to simulate simple recrystallization. I found that I cannot try nuclei with radius smaller than grid spacing when stabilisation is chosen. But I found in some examples nuclei radius could be set to 0 if stabilisation is chosen. What is the problem?
I still confused with 'stabilisation'. if the nucleus is smaller than grid spacing, the phase field value is assigned 'phasemin', the curvature is ignored. if the nucleus is bigger than a cell, curvature is recovered, right? So if the critical radius for nucleation (when curvature force=driving force) is more than one cell, after the nucleus grew bigger than one cell but still smaller than critical size, the nucleus will stay metastable as big as one cell.
I tried to simulate simple recrystallization. I found that I cannot try nuclei with radius smaller than grid spacing when stabilisation is chosen. But I found in some examples nuclei radius could be set to 0 if stabilisation is chosen. What is the problem?
I still confused with 'stabilisation'. if the nucleus is smaller than grid spacing, the phase field value is assigned 'phasemin', the curvature is ignored. if the nucleus is bigger than a cell, curvature is recovered, right? So if the critical radius for nucleation (when curvature force=driving force) is more than one cell, after the nucleus grew bigger than one cell but still smaller than critical size, the nucleus will stay metastable as big as one cell.
Re: nucleation issues
Dear zhubq,
I'm sorry for not answering before, I am on holidays and had no internet connection!
If you use stabilisation, it does not matter whether you input a radius of zero or any other value smaller than the grid spacing. In all those cases, a "small" grain with a fraction of 2.01 x phMin will be set. The code only checks whether the value is bigger or smaller than delta_X, so I don't understand why this makes a difference for you! Curvature is ignored at the beginning and switched on increasingly, proportional to the fraction of the central cell.
If you use stabilisation, the seed is supposed to grow. If the critical radius (like you defined it ) is bigger than the grid spacing, the seed will not grow, but also not vanish because it is metastable because of the artificial reduction of curvature. You should check whether this is the case in your example. You may terminate this stabilisation by using "kill_metastable". To avoid this situation (which may be due to a very small grid spacing, high interfacial energy or small stored energy) you can use "analytical_curvature" and specify a critical radius which is bigger than your critical radius. This is as if you had a big heterogeneous nucleation site which reduces curvature, but you still get a "small" grain with a fraction of 2.01*phMin.
If you use stabilisation with a initial radius bigger than the grid spacing, the situation is quite undefined. The seed consists of a cell (or several cells) with a sharp interface, so curvature is badly defined at the beginning. The seed will start growing or shrinking (or just get somehow diffuse), depending on the driving force. No stabilisation (i.e. curvature reduction) will be applied.
Sometimes, also a too high interface mobility can cause a strange nucleation behaviour.
Bernd
I'm sorry for not answering before, I am on holidays and had no internet connection!
If you use stabilisation, it does not matter whether you input a radius of zero or any other value smaller than the grid spacing. In all those cases, a "small" grain with a fraction of 2.01 x phMin will be set. The code only checks whether the value is bigger or smaller than delta_X, so I don't understand why this makes a difference for you! Curvature is ignored at the beginning and switched on increasingly, proportional to the fraction of the central cell.
If you use stabilisation, the seed is supposed to grow. If the critical radius (like you defined it ) is bigger than the grid spacing, the seed will not grow, but also not vanish because it is metastable because of the artificial reduction of curvature. You should check whether this is the case in your example. You may terminate this stabilisation by using "kill_metastable". To avoid this situation (which may be due to a very small grid spacing, high interfacial energy or small stored energy) you can use "analytical_curvature" and specify a critical radius which is bigger than your critical radius. This is as if you had a big heterogeneous nucleation site which reduces curvature, but you still get a "small" grain with a fraction of 2.01*phMin.
If you use stabilisation with a initial radius bigger than the grid spacing, the situation is quite undefined. The seed consists of a cell (or several cells) with a sharp interface, so curvature is badly defined at the beginning. The seed will start growing or shrinking (or just get somehow diffuse), depending on the driving force. No stabilisation (i.e. curvature reduction) will be applied.
Sometimes, also a too high interface mobility can cause a strange nucleation behaviour.
Bernd
Re: nucleation issues
Hello Mr. Bernd,
What would be the best idea to increase the number of grains nucleating in my simulation without changing my Geometry and starting the simulation from the melt? (Info: I am simulating Stabilor G and included phases Liquid and FCC_A1,
nucleation model chosen is Seed density)
I did variations in the values of different parameters like seed radius and density. Aslo, varying phase interaction parameters like kinetic coefficient and surface energy didn't help much.
Thanks,
Raghav
What would be the best idea to increase the number of grains nucleating in my simulation without changing my Geometry and starting the simulation from the melt? (Info: I am simulating Stabilor G and included phases Liquid and FCC_A1,
nucleation model chosen is Seed density)
I did variations in the values of different parameters like seed radius and density. Aslo, varying phase interaction parameters like kinetic coefficient and surface energy didn't help much.
Thanks,
Raghav