Capruring Sigma3 Twins in Synthetic Microstructure

[Sia]

I have 2D and 3D-EBSD data sets of a FCC material sample that its Dream3D statistics shows that frequency of sigma3 twins is substantially higher than other grain boundary misorientations ( high frequency for misorientation angle of 60 degree). But once I use the statistics of the EBSD sample which contains the high frequency of sigma3 twin boundaries (60 degree misorientations) to create a synthetic microstructure, the misorientation angle statistics of the synthetic microstructure does not have a peak at 60 degree, which means there is no high frequency of the twins.
What I did was this for 2D-EBSD: Once I get the statistics of 2D-EBSD, I use the StatsGenerator to generate a 3D statistically equivalent data using the grains size information from 2D-EBSD and its ODF, and MDF, and then modify the morphological statistics of ".dream3d" file that I have got from 2D-EBSD statistics. Then, I use the modified ".dream3d" file to create the synthetic microstructure.
for 3D-EBSD: I use the dream3d after statistics step, and use that to create a synthetic microstructure.
Unfortunately for none of the above I could get high frequency of sigma3 twins ( high frequency for misorientation angle of 60 degree) in synthetic microstructure. I am wondering if I have done something wrong or Dream3D is not capable of capturing that.

Thanks,
Sia

[Sia]

Hi all,

I am wondering if anybody knows whether or not Dream3D can capture high probability density of twins (high probability density of misorientation of 60 degree around [111]) when we create a synthetic microstructure from statistics of 3D EBSD of a FCC material. In my synthetic structures I am not able to capture twins while I have used 3D EBSD statistics to create the synthetic microstructure.

I would really appreciate it if anybody can help.

Best,

Sia

[Sean Donegan]

Hi Sia,

In principle, the Match Crystallography filter will eventually match the correct MDF that has been set either in StatsGenerator or calculated from experimental data.  The problem is that this process may take a very long time; the Match Crystallography filter operates for a user-set number of iterations, and will terminate after reaching that number even if the actual MDF has not been matched.  Matching the ODF is usually quite fast, but matching the MDF can take a large amount of time since the filter optimizes using a Monte Carlo approach, so the likelihood of getting a boundary with the correct misorientation angle and axis is dependent upon selecting two grains next to each other instead of just one in isolation.  Because this approach is used, it could take a large number of iterations to reach the desired MDF.

Additionally, using the synthetic builder in this fashion to go after twins will not generally result in twin boundaries that are morphologically flat.  To get the morphology correct, the Insert Transformation Phases filter should be used.  Try that filter instead for constructing a highly-twinned synthetic structure.

hope that helps,

-Sean

[Sia]

Sean,

Thanks alot for your response.
I did what you explained (about using "Insert Transformation Phases") in your notes to create small microstructure with the flat twins and it seems that it's working but there are still two problems. I created a microstructure with 10 grains and 4 twins are also inserted in the microstructure using the "Insert Transformation Phases"
1-When I open the csv file, it shows that the extra features are created, which are twins, but the information about them is incomplete. For example it writes down the Average Quaternion and Euler Angles for the twins while their EquivalentDiameter of all of them is 4, and the Centroids of all of them are the same, while Volume, aspect raio, and number of cells for each of these grains is Zero. Also the number of neighbors for all of the twins is zero and therefore no misorientation angle information for any of them is given.
2- When I open the ".xdmf" file, I can see the twins only when I set the Paraview to show based on "featureId" but with "IPFcolor" I can't see them, it's like they are still merged into parent grains.

I have attached the pipeline and the csv file as well as the ".xdmf" file here and hope you can help me out on this.

Thanks in advance,
Sia

[Sean Donegan]

Hi Sia,

Taking a quick look, the only thing I can think of is that you need to make sure to re-run the various statistics for features after doing the twin insertion, so that those statistics are correctly calculated on the newly formed twins.  This could explain why so much of the data in the CSV is zeros.  If that's not the case, I'll have to take a closer look at your pipeline.

-Sean

On Tuesday, November 17, 2015 at 1:47:12 PM UTC-5, Sia wrote:

[Michael Groeber]

Sia

What Sean suggested is true.  This is something that is often forgot when processing data in DREAM.3D.  If you make a calculation (volume, number of neighbors, etc) and then you insert new features or remove small features or anything like that, then the calculations you made previously are likely wrong/out-of-date.  The volumes and number of neighbors you have for the original 10 grains are likely wrong now, because some of the volume of those grains was lost to the twins.  These are just things that need to be considered when running increasingly complex pipelines.

Best

Mike

[Sia]

Hi Michael and Sean,

The problem with this filter is that when we use "Insert Transformation Phases" the filter assigns some of the cells to the new created features (Twins) but the Euler angle of those cells that are assigned to the new feature do not change. This means that if we open dream3d file in HDFView you will see that in the FeatureIds under CellData, the FeatureIds of these cells that belong to the new created features (Twins)  are changed but when you check the Euler angles that correspond to these cells in HDFview,  you see that their Euler angles of the cells remained unchanged (not transformed).

Do you have any idea how to resolve the problem.

Thanks!