Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

3D model with two crossing symmetry planes

Please login with a confirmed email address before reporting spam

Dear Comsol users,

I'm doing simulations with a 3D model, which has two crossing symmetry planes (xz and yz). Therefore it would be enough to simulate just a quarter of it.

Now I wonder how to set up the symmetry, because by adding a symmetry for two sides of the quarter I will only get three quarters overall. What about the fourth one? In order to find a solution of the problem, I also didn't find a way to set up a edge/point symmetry. In addition I don't know whether the result would be the same (calculating just one quarter and calculating all quarters)?

Another question:
Is there a huge reduction of calculation time if using symmetry? Or is it only to save time while drawing the model?

I couldn't attach a file, so please have a look to my model here: docs.google.com/file/d/0B4e6DBNbgmJjSWZ1NmdhdXhZSXc

9 Replies Last Post Jun 27, 2017, 3:10 a.m. EDT

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 5:36 a.m. EST
Hi

According to my experience, drawing only half of the model and using the symmetry boundary condition does not reduce the calculation time or memory significantly, at least in the diffusion model I had. I do not know other way than drawing half of your geometry and using symmetry as planned.

BR
Lasse
Hi According to my experience, drawing only half of the model and using the symmetry boundary condition does not reduce the calculation time or memory significantly, at least in the diffusion model I had. I do not know other way than drawing half of your geometry and using symmetry as planned. BR Lasse

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 5:55 a.m. EST
Thanks for your quick response.

Do you or anyone know how to get surface plots in the results section which are also showing the regions cutted by symmetry? Is there for instance an option of mirroring?
Thanks for your quick response. Do you or anyone know how to get surface plots in the results section which are also showing the regions cutted by symmetry? Is there for instance an option of mirroring?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 7:12 a.m. EST
Hi

You must define a Cut Plane under Results --> Data Sets node. Then under Results, add a 2D Plot Group, there a Data set to Cut Plane 1, then Surface . Was this clear?

BR
Lasse
Hi You must define a Cut Plane under Results --> Data Sets node. Then under Results, add a 2D Plot Group, there a Data set to Cut Plane 1, then Surface . Was this clear? BR Lasse

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 9:27 a.m. EST
I understood. Thanks a lot!
I understood. Thanks a lot!

Nagi Elabbasi Facebook Reality Labs

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 11:12 a.m. EST
Defining the two symmetry planes like you described is enough. You can think of it in the following manner: one symmetry plane gets you from 90 to 180 degrees and the other from 180 to 360.

Now when you reduce the model size via symmetry COMSOL might end up using smaller elements. If however you keep the element size the same as that used in the full model then the quarter model will have roughly a quarter of the degrees of freedom. That should lead to a 4 times faster solution or better. Usually better since solution time rarely scales with N (the number of degrees of freedom). It scales with N log(N) at best.

Nagi Elabbasi
Veryst Engineering
Defining the two symmetry planes like you described is enough. You can think of it in the following manner: one symmetry plane gets you from 90 to 180 degrees and the other from 180 to 360. Now when you reduce the model size via symmetry COMSOL might end up using smaller elements. If however you keep the element size the same as that used in the full model then the quarter model will have roughly a quarter of the degrees of freedom. That should lead to a 4 times faster solution or better. Usually better since solution time rarely scales with N (the number of degrees of freedom). It scales with N log(N) at best. Nagi Elabbasi Veryst Engineering

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 11:44 a.m. EST
Thanks for your reply.

I didn't understand how the second symmetry plane will bring me from 180° to 360°, because I can only select the bondaries from the first quarter. To get 360° I also have to add the boundaries from the second quarter, which is given by the first symmetry. But the boundaries of that are not visible/selectable. If I would see the part from the first symmetry and could select its boundaries as well, then I would agree.

Good to know, that the reduction by symmetry will theoretically speed up my calculations. But there is still the problem with the last quarter. :o(
Thanks for your reply. I didn't understand how the second symmetry plane will bring me from 180° to 360°, because I can only select the bondaries from the first quarter. To get 360° I also have to add the boundaries from the second quarter, which is given by the first symmetry. But the boundaries of that are not visible/selectable. If I would see the part from the first symmetry and could select its boundaries as well, then I would agree. Good to know, that the reduction by symmetry will theoretically speed up my calculations. But there is still the problem with the last quarter. :o(

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 23, 2014, 1:27 p.m. EST
Andre:

Though probably not how it is done numerically, think of it as an order of operation. If you were to mirror over one line of symmetry (say, y=0), you have half a unit cell. The other line of symmetry (x=0) has now ALSO been mirrored to run along the entire y-axis! Thus, when this is mirrored, you have the entire unit cell represented.

Keep in mind that COMSOL recognizes symmetry through the physics, not the actual geometry. Your full unit cell would normally have four external BCs related to the physics--generally some known field/potential or derivative. For example, in electromagnetics we use perfect electrical conductors (PECs) and perfect magnetic conductors (PMCs) to generate symmetry in the electrical and magnetic fields, repectively, which also act as anti-symmetrical boundaries if you switch the corresponding fields. If it is not a periodic structure, perfectly matched layer (PML) BCs are implemented on the outer boundaries of the symmetrical unit cell so that the field approaches 0 after a short distance. It is important that you understand these are NOT the same as geometrical symmetries! In COMSOL, you establish the geometrical symmetries by building a model with some known N-fold symmetry, and you establish the physical symmetries/anti-symmetries by selection of the proper BCs.

Let's assume your model is a periodic structure which has an E-field running along the extruded line (y), and is incident in -z (into the page). The BCs for your quadrant would be a symmetry BC in the +/- y faces (PEC) and anti-symmetry in the +/- x faces (PMC) (or you could think of it as having symmetry in H), both having the purpose of mirroring the fields at the boundaries without changing orientation. Thus the field and the structure are mirrored to simulate an infinite array. Even coupled behavior, such as the capacitive response between the cut center wires, will be manifest in the calculations.

HOWEVER, at off normal incidence, there is still geometrical symmetry, but no longer electromagnetic symmetry--incident E-field components have differing magnitudes and phases in each quadrant, and the neighboring unit cell has differing progagation (k) vectors--and so you CANNOT reduce the model into symmetrical representation.

Hope this helps shed some light!

-Bryan
Andre: Though probably not how it is done numerically, think of it as an order of operation. If you were to mirror over one line of symmetry (say, y=0), you have half a unit cell. The other line of symmetry (x=0) has now ALSO been mirrored to run along the entire y-axis! Thus, when this is mirrored, you have the entire unit cell represented. Keep in mind that COMSOL recognizes symmetry through the physics, not the actual geometry. Your full unit cell would normally have four external BCs related to the physics--generally some known field/potential or derivative. For example, in electromagnetics we use perfect electrical conductors (PECs) and perfect magnetic conductors (PMCs) to generate symmetry in the electrical and magnetic fields, repectively, which also act as anti-symmetrical boundaries if you switch the corresponding fields. If it is not a periodic structure, perfectly matched layer (PML) BCs are implemented on the outer boundaries of the symmetrical unit cell so that the field approaches 0 after a short distance. It is important that you understand these are NOT the same as geometrical symmetries! In COMSOL, you establish the geometrical symmetries by building a model with some known N-fold symmetry, and you establish the physical symmetries/anti-symmetries by selection of the proper BCs. Let's assume your model is a periodic structure which has an E-field running along the extruded line (y), and is incident in -z (into the page). The BCs for your quadrant would be a symmetry BC in the +/- y faces (PEC) and anti-symmetry in the +/- x faces (PMC) (or you could think of it as having symmetry in H), both having the purpose of mirroring the fields at the boundaries without changing orientation. Thus the field and the structure are mirrored to simulate an infinite array. Even coupled behavior, such as the capacitive response between the cut center wires, will be manifest in the calculations. HOWEVER, at off normal incidence, there is still geometrical symmetry, but no longer electromagnetic symmetry--incident E-field components have differing magnitudes and phases in each quadrant, and the neighboring unit cell has differing progagation (k) vectors--and so you CANNOT reduce the model into symmetrical representation. Hope this helps shed some light! -Bryan

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jan 24, 2014, 3:28 a.m. EST
Thank you for your very detailed reply. Now I've got it. ;o)

I also figured out how to get a 2D Plot in the xy-plane of the whole geometry. In the data sets node under results one have to define a 2D Mirroring with relation to a created cut plane (in xy). Defining another 2D Mirroring with relation to the first one will bring up the wanted 2D Plot after creating a 2D plot group and selecting the 2D Mirroring 2 as the data set.

Thanks a lot to all of you!
Thank you for your very detailed reply. Now I've got it. ;o) I also figured out how to get a 2D Plot in the xy-plane of the whole geometry. In the data sets node under results one have to define a 2D Mirroring with relation to a created cut plane (in xy). Defining another 2D Mirroring with relation to the first one will bring up the wanted 2D Plot after creating a 2D plot group and selecting the 2D Mirroring 2 as the data set. Thanks a lot to all of you!

Please login with a confirmed email address before reporting spam

Posted: 7 years ago Jun 27, 2017, 3:10 a.m. EDT
Dear Sir,
I have a small problem.I am working with an geometry which is periodic in 2D(consider a sided square box with a pillar in middle).Left right and front back are periodic.But top and bottom is not periodic and goes to infinity.

Which boundary condition should i use?You can have a look on mz geometry from the attachment.

Waiting for your kind reply.

Best Regards
Anisuzzaman Boni
Dear Sir, I have a small problem.I am working with an geometry which is periodic in 2D(consider a sided square box with a pillar in middle).Left right and front back are periodic.But top and bottom is not periodic and goes to infinity. Which boundary condition should i use?You can have a look on mz geometry from the attachment. Waiting for your kind reply. Best Regards Anisuzzaman Boni

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.