Hello, Sir,
I have a question about the nucleation in Micress.
In the seed-density nucleation model in MICRESS, the nucleated seeds are determined by free growth model. In an isothermal assumption, when the temperature of liquid decreasing, the big seed, for example seed A, nucleates at first, then there is a composition profile around it. with the decreasing of temperature, the seed C will nucleate while the critical undercooling is reached. My question is that if seed B with same size of C will nucleate or not in calculation of MICRESS? I mean does MICRESS considers the local constitutional undercooling when it check the seed nucleation.
Thanks a lot for your help.
Kind regards,
Feng
Nucleation in MIRESS
Nucleation in MIRESS
- Attachments
-
- Question.pptx
- (68.03 KiB) Downloaded 288 times
Re: Nucleation in MIRESS
Hi Feng,
of course, MICRESS does. Local composition and local temperature are used in order to calculate the driving force for nucleation. If nucleation ocurs on interfaces between different phases, the curvature contribution is also added.
Best wishes
Bernd
of course, MICRESS does. Local composition and local temperature are used in order to calculate the driving force for nucleation. If nucleation ocurs on interfaces between different phases, the curvature contribution is also added.
Best wishes
Bernd
Re: Nucleation in MIRESS
OK, thanks.
I have some other questions: in the seed-density model, it can be set up the seed size and number density, the seed is assumed as random distribution. the question (1) is can I set the seed in a determined position? (2) the seed number density is set with unit /cm3, but in a 2D simulation, how can I get the seed number in a 2D area, for example seed density is set as 1000/cm3, how can I get the numbers in 1 cm2 area.
I have some other questions: in the seed-density model, it can be set up the seed size and number density, the seed is assumed as random distribution. the question (1) is can I set the seed in a determined position? (2) the seed number density is set with unit /cm3, but in a 2D simulation, how can I get the seed number in a 2D area, for example seed density is set as 1000/cm3, how can I get the numbers in 1 cm2 area.
Re: Nucleation in MIRESS
Dear Feng,
(1) You cannot set seeds to determined positions using the seed density model. This would contradict the idea of a radius-density distribution. But you can apply this distribution to a region which is defined by a range in x, y- and z-coordinate.
(2) From the 3D-seed density values, a corresponding 2D value is calculated by assuming the same average distances between particles as in 3D:
rho2D=rho3D2/3
Bernd
(1) You cannot set seeds to determined positions using the seed density model. This would contradict the idea of a radius-density distribution. But you can apply this distribution to a region which is defined by a range in x, y- and z-coordinate.
(2) From the 3D-seed density values, a corresponding 2D value is calculated by assuming the same average distances between particles as in 3D:
rho2D=rho3D2/3
Bernd
Re: Nucleation in MIRESS
Thanks a lot.
Re: Nucleation in MIRESS
Hello Bernd,
I got another question on the seed-density nucleation model used in Micress in 2D simulation. Usually, we set a seed size and number density in 3D with /cm3, and the seed number can transfer to 2D by N(2D)=[N(3D)]^(2/3). but this equation is not additive in different seeds size. When there is only one type of seeds, it is ok, but when there are several types of seed size, there will be a problem, for example:
case 1: only one type of seeds with 1000000/cm3 (3D), it is 10000 /cm2 (2D)
case 2: there are two different size of seeds.
radius 1 um with 500000 /cm3 (3D), it is about 6300 /cm2 (2D)
radius 0.5 um with 500000 /cm3 (3D), it is about 6300 /cm2 (2D)
In case 1 it is ok. In case 2 the total number density of seed is same with case 1 in 3D, however when it is transferred to 2D the total number density is 12600, much more than that in case 1.
Can you tell me how you deal with this problem in your simulation?
Many thanks,
Feng
I got another question on the seed-density nucleation model used in Micress in 2D simulation. Usually, we set a seed size and number density in 3D with /cm3, and the seed number can transfer to 2D by N(2D)=[N(3D)]^(2/3). but this equation is not additive in different seeds size. When there is only one type of seeds, it is ok, but when there are several types of seed size, there will be a problem, for example:
case 1: only one type of seeds with 1000000/cm3 (3D), it is 10000 /cm2 (2D)
case 2: there are two different size of seeds.
radius 1 um with 500000 /cm3 (3D), it is about 6300 /cm2 (2D)
radius 0.5 um with 500000 /cm3 (3D), it is about 6300 /cm2 (2D)
In case 1 it is ok. In case 2 the total number density of seed is same with case 1 in 3D, however when it is transferred to 2D the total number density is 12600, much more than that in case 1.
Can you tell me how you deal with this problem in your simulation?
Many thanks,
Feng
Re: Nucleation in MIRESS
Dear Feng,
Simulations in 2D need always approximations, and the important question is which approximation suits best to the problem which has to be solved. Imagine the case that you have very few big seed particles, which trigger nucleation at low undercooling, and a huge number of small particles which get active only at a much higher undercooling. In a 2D simulation, we want to "project" all those particles of the 3D space onto our simulation plane which have an influence on the 2D plane. Obviously, the big particles have a much higher interaction length than the small ones: If they alone would be active, they would e.g. produce very large grains in an equiaxed solidification, while the small ones, when activated by large undercooling, would give a fine-grained structure. It is a good approximation to define the average distance between the particles of one class in 3D as interaction length.
This is the basic consideration for the implemented 2D-3D correction: The few bigger seeding particles have a high average distance, therefore they must be "over-represented" in a 2D simulation compared to a 3D simulation. The model assumption is that there is no "interaction" between the two seed classes.
Of course, if the classes get closer in seed radius, and it is more the total seed number rather than the individual class which determines nucleation, then this model is wrong. Then, it would be more appropriate to join the classes to form a bigger one.
By the way, in MICRESS there additionally is a statistical radius distribution inside each class (between the class limits), so that artifacts of very big classes are reduced.
In conclusion, in order to avoid or reduce 3D to 2D conversion errors in cases where quantitative 3D distributions are known, the seed classes should be chosen so big that there is no interaction between the classes, i.e. nucleation and competition between nuclei is dominated by only one class at a time. In simulations where distributions are obtained by calibration, the correctness of the resulting distribution is not important anyway.
Bernd
Simulations in 2D need always approximations, and the important question is which approximation suits best to the problem which has to be solved. Imagine the case that you have very few big seed particles, which trigger nucleation at low undercooling, and a huge number of small particles which get active only at a much higher undercooling. In a 2D simulation, we want to "project" all those particles of the 3D space onto our simulation plane which have an influence on the 2D plane. Obviously, the big particles have a much higher interaction length than the small ones: If they alone would be active, they would e.g. produce very large grains in an equiaxed solidification, while the small ones, when activated by large undercooling, would give a fine-grained structure. It is a good approximation to define the average distance between the particles of one class in 3D as interaction length.
This is the basic consideration for the implemented 2D-3D correction: The few bigger seeding particles have a high average distance, therefore they must be "over-represented" in a 2D simulation compared to a 3D simulation. The model assumption is that there is no "interaction" between the two seed classes.
Of course, if the classes get closer in seed radius, and it is more the total seed number rather than the individual class which determines nucleation, then this model is wrong. Then, it would be more appropriate to join the classes to form a bigger one.
By the way, in MICRESS there additionally is a statistical radius distribution inside each class (between the class limits), so that artifacts of very big classes are reduced.
In conclusion, in order to avoid or reduce 3D to 2D conversion errors in cases where quantitative 3D distributions are known, the seed classes should be chosen so big that there is no interaction between the classes, i.e. nucleation and competition between nuclei is dominated by only one class at a time. In simulations where distributions are obtained by calibration, the correctness of the resulting distribution is not important anyway.
Bernd