How are the initial grains with Voronoi created?

technical aspects of .dri file generation (e.g. debug mode ) etc...
Post Reply
longwen36
Posts: 3
Joined: Fri May 29, 2015 9:56 am
anti_bot: 333

How are the initial grains with Voronoi created?

Post by longwen36 » Fri May 29, 2015 12:35 pm

Hi, there

I. I would like to know how the initial grain set is implemented within the VORONOI construction routine in MICRESS. Anybody knows what the algorithm is?

II. The problem is that I want to utilize MICRESS to create an initial grain set with a given distribution, yet it seems in MICRESS you can only set the grains to be within a certain range (min size to max size), which can not be set to obey a certain distributtion, say, uniform, normal, and lognormal distributions, etc. Moreover, I would like to set the grain boundary a given thickness, but it just doesn't work, and the grain boudary seem to be unchaned no matter how I change the value of the thickness. In my simulation, I want to set the simulation box size to be 1um*1um, the grain size to be varying from 99~101nm( Or better, fixed at 100nm. The problem occurs when fixed at the certain value of 100nm, voids are inevitably created), and the thickness to be 5nm. I attach the input file here. Could some help me check out what is wrong with my input file? With DP_MICRESS, I visualized the data, and found the grain boudary thickness was not 5nm, and also the grain sizes were not within 99 nm~101 nm. What should I do? Any discussion or suggestions will be much appreciated.

Yours sincerely,

Venn Wu
Attachments
grain_105nm_in.txt
Input driving file
(9.28 KiB) Downloaded 326 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: How are the initial grains with Voronoi created?

Post by Bernd » Mon Jun 01, 2015 10:22 am

Dear longwen36,

Welcome to the MICRESS forum!

The Voronoi construction is the mostly used option when it comes to creating initial microstructures in MICRESS. It is a simple geometric rule how to construct the grain boundary between two overlapping grains. Each point of the boundary is constructed such that the ratio between the distances to the grain centers is the same as between the grain radii. It has also been discussed here.
Creating holes is not automatically avoided when using the Voronoi construction, because the background grain 0 is not included into this construction. To avoid that, one should use either sufficiently large grain radii, and/or set a single large grain as background before.
In your case (use of Voronoi), the effective grain size (without holes) only depends on the number of grains which you set, and not on their radii. If you use only a fixed radius (say 500 nm), you will automatically obtain a certain grain size distribution. This is due to the random selection of the grain positions and does not depend directly on the range of radii. Changing to an increasing range of radii creates more and more curved grain boundaries, but does not systematically change the grain size distribution. On the other hand, the grain size distribution can be changed by the definition of a minimal distance between the grain centers.
Unfortunately, it is not possible to directly achieve a given grain size distribution. If you know any tool which does that, please remember that it is also easily possible to read in initial virtual microstructures created by any image manipulation software or from experimental microstructure images.
I do not really understand your problem with the thickness of the grain boundaries. The interface thickness is defined at the very end of the driving file and must be given in grid cells. In your attached input file, 5 cells correspond to 1 nm interface thickness. Perhaps you use the wrong output type for checking the interface thickness. In case of grain growth, you can see the effective interface thickness in the .intf, .vel and .krum (i.e. interface, velocity and curvature) outputs, while e.g. .phas or .korn will always show a sharp interface.
We strongly recommend using "fd_correction" (optional input in the same line with the type of potential) and an interface thickness which is never smaller than 2.5 cells. In case of grain growth, we recommend 4-5 cells. So, to obtain an effective interface thickness of 5 nm with a nominal interface thickness of 5 cells, you should increase grid resolution to 1 nm.
Best wishes
Bernd

longwen36
Posts: 3
Joined: Fri May 29, 2015 9:56 am
anti_bot: 333

Re: How are the initial grains with Voronoi created?

Post by longwen36 » Mon Jun 01, 2015 11:42 am

Bernd wrote:Dear longwen36,

Welcome to the MICRESS forum!

The Voronoi construction is the mostly used option when it comes to creating initial microstructures in MICRESS. It is a simple geometric rule how to construct the grain boundary between two overlapping grains. Each point of the boundary is constructed such that the ratio between the distances to the grain centers is the same as between the grain radii. It has also been discussed here.
Creating holes is not automatically avoided when using the Voronoi construction, because the background grain 0 is not included into this construction. To avoid that, one should use either sufficiently large grain radii, and/or set a single large grain as background before.
In your case (use of Voronoi), the effective grain size (without holes) only depends on the number of grains which you set, and not on their radii. If you use only a fixed radius (say 500 nm), you will automatically obtain a certain grain size distribution. This is due to the random selection of the grain positions and does not depend directly on the range of radii. Changing to an increasing range of radii creates more and more curved grain boundaries, but does not systematically change the grain size distribution. On the other hand, the grain size distribution can be changed by the definition of a minimal distance between the grain centers.
Unfortunately, it is not possible to directly achieve a given grain size distribution. If you know any tool which does that, please remember that it is also easily possible to read in initial virtual microstructures created by any image manipulation software or from experimental microstructure images.
I do not really understand your problem with the thickness of the grain boundaries. The interface thickness is defined at the very end of the driving file and must be given in grid cells. In your attached input file, 5 cells correspond to 1 nm interface thickness. Perhaps you use the wrong output type for checking the interface thickness. In case of grain growth, you can see the effective interface thickness in the .intf, .vel and .krum (i.e. interface, velocity and curvature) outputs, while e.g. .phas or .korn will always show a sharp interface.
We strongly recommend using "fd_correction" (optional input in the same line with the type of potential) and an interface thickness which is never smaller than 2.5 cells. In case of grain growth, we recommend 4-5 cells. So, to obtain an effective interface thickness of 5 nm with a nominal interface thickness of 5 cells, you should increase grid resolution to 1 nm.
Best wishes
Bernd
Dear Bernd,

Thank you very much for your answer. There is indeed a mistake in my driving file. You mentioned that it was easy to create initial grain sets in obedience to any given distribution, how? Anys suggestive software or method? I guess I am not going to read in virtual microstructures from experimental images because they are often specific and lose the randomness.

Kind regards,

Venn Wu

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: How are the initial grains with Voronoi created?

Post by Bernd » Mon Jun 01, 2015 12:48 pm

Dear Venn,

I did not say it is easy to create an initial grain structure for a given distribution, I even believe, strictly speaking, it is not possible at all! I just said that it is quite easy to read in experimental microstructures.
I agree with you that reading experimental microstructures does not solve the problem in all cases because statistics are not sufficient. On the other hand, if you do not use experimental information, how can you know whether your initial distribution is realistic?
If you are interested in stationary grain growth distributions, the solution may be to start with a big number of grains with arbitrary initial distribution and observe the development of the size distribution during grain growth...
If we are talking about relatively few grains, one could also think about reading an experimental structure, but making manual corrections to improve the size distribution or meet other criteria.

Bernd

longwen36
Posts: 3
Joined: Fri May 29, 2015 9:56 am
anti_bot: 333

Re: How are the initial grains with Voronoi created?

Post by longwen36 » Wed Jun 03, 2015 4:21 am

Bernd wrote:Dear Venn,

I did not say it is easy to create an initial grain structure for a given distribution, I even believe, strictly speaking, it is not possible at all! I just said that it is quite easy to read in experimental microstructures.
I agree with you that reading experimental microstructures does not solve the problem in all cases because statistics are not sufficient. On the other hand, if you do not use experimental information, how can you know whether your initial distribution is realistic?
If you are interested in stationary grain growth distributions, the solution may be to start with a big number of grains with arbitrary initial distribution and observe the development of the size distribution during grain growth...
If we are talking about relatively few grains, one could also think about reading an experimental structure, but making manual corrections to improve the size distribution or meet other criteria.

Bernd
Dear Bernd,
Thank you very much for your kind suggestion. Reading in an experimental structure and making further corrections is OK. I read a few articles which describe creating grain sets of any given size distribution with voronoi polygons. So, that is why I raised this question to query if it is possible in MICRESS. Now, it seems the algorithms implemented in MICRESS is different. I may try some other methods. Thank you again!

Venn Wu

Post Reply