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 The information provided may be out of date.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.

Is it required to put Perfect Matched Layer (PML) alongside the scattering boundary condition?

Please login with a confirmed email address before reporting spam

I have a 3D RF problem. The whole domain is a box. I put a thin film (may have some patterning in the film) at the center, then I try to get the absorption of the thin film under normal incidence. The structure is periodic in the transverse plane, so I set Periodic boundary condition on the four sides, surrounding the transverse plane. For the top and bottom sides, I use scattering boundary condition. I set the excitation by explicitly defining the analytic formula for the excitation field when the thin film is not there.

The question I have is that do I need put PML alongside the scattering boundary condition in the top and bottom sides. Currently, I only use the scattering boundary condition, and without the PML.

I am not sure about the mechanism of how it works in COMSOL. In finite difference methods, like FDTD or FDFD, if total field/scattered field formalism is used, one will need to put PML layers outside the total/field/scattered field box. So, does anyone know whether for COMSOL, for a case using a scattering boundary condition, like here, it is required to use PML alongside the scattering boundary condition?

Thanks!

6 Replies Last Post 26 giu 2017, 05:29 GMT-4

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 5 giu 2013, 16:20 GMT-4
Well, I have a couple of thougths:

If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space.

On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial...

Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs....

Anywya, those were just some quick thoughts. Good luck!

--matt
Well, I have a couple of thougths: If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space. On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial... Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs.... Anywya, those were just some quick thoughts. Good luck! --matt

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 5 giu 2013, 16:20 GMT-4
Well, I have a couple of thougths:

If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space.

On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial...

Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs....

Anywya, those were just some quick thoughts. Good luck!

--matt
Well, I have a couple of thougths: If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space. On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial... Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs.... Anywya, those were just some quick thoughts. Good luck! --matt

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 5 giu 2013, 16:50 GMT-4
Hi Matt,

Thank you so much for your information! I do see that in some cases, there is some artifacts around the corners of the scattering boundary condition, which should correspond to the oblique scattering.

The PEC and PMC seem only available for normal incidence. For oblique incidence, the incidence field does not have the symmetry, in which case, Floquet boundary condition should be used for the four sides.

Thanks for the reply. And I have another question, if I use the port boundary condition to define the incidence (top and bottom sides), do the ports need to be far enough from the object?

Alex


Well, I have a couple of thougths:

If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space.

On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial...

Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs....

Anywya, those were just some quick thoughts. Good luck!

--matt


Hi Matt, Thank you so much for your information! I do see that in some cases, there is some artifacts around the corners of the scattering boundary condition, which should correspond to the oblique scattering. The PEC and PMC seem only available for normal incidence. For oblique incidence, the incidence field does not have the symmetry, in which case, Floquet boundary condition should be used for the four sides. Thanks for the reply. And I have another question, if I use the port boundary condition to define the incidence (top and bottom sides), do the ports need to be far enough from the object? Alex [QUOTE] Well, I have a couple of thougths: If you want to do it this way by launching the wave through the scattering BC and use scatt field formulation, I think you would ideally want the PML, but I don't know if comsol is happy with PML+scattering BC. I say that because if the device under test scatters things at angles other than normal, then those components would not normally be absorbed by the scattering BC. What I mean to say is that the scatt BC only absorbs perfectly for PLANE waves. Anything else could potentially be [artificially] reflected within your sim space. On the other hand, you can solve this problem much easier by using port BC and then still just specifying the analytic expression for a plane wave at the input port-->user defined. Reflection/Transmission problems are one of the best uses of port BC. The port absorbs perfectly the reflected and transmitted waves through the [periodic] sample. It also has handy S-parameters which makes transmission, reflection, and absorption trivial... Also, I know you said you were using the periodic BC. Be careful there, however, as I found I could only model an ifinte array of unit cell devices by using 2 PEC and 2 PMC boundary conditions. Which 2 they are depends on your E an H-field polarization. I only mention this because I struggled A LOT with periodic BC before I discovered that normal incidence plane waves are using the PEC and PMC pairs.... Anywya, those were just some quick thoughts. Good luck! --matt [/QUOTE]

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 5 giu 2013, 16:59 GMT-4
Yep, OK...I just mentioned the PEC and PMC because I thought you were only using normal incidence. It is definitely true that floquet is necessary....

As for the ports, well, I asked comsol support once about this and they said they had to be 1/2 wavelength away from the object being probed to avoid near field effects.

On the other hand, I was able to reproduce a bunch of other groups' results using ports that were closer to the object than that. It's one of those things were I certainly try to be at least 1/2 wavelength away from the DUT. Unfortunately that's not always the case (plus, as i'm sure you can guess, it increases the number of mesh elements/DOF, so that's unfortunate).

So, I guess the punch line is that officially you want the ports 1/2 wavelength away and no closer. However, unofficially, my experience taught me that you can do very good with ports that are closer. Having an analytic or otherwise known problem whose solution you can test this out on would be good if you feel you want to be closer than 1/2 wavelength, but don't want to just take my word for it ;)

Good luck again!

--matt
Yep, OK...I just mentioned the PEC and PMC because I thought you were only using normal incidence. It is definitely true that floquet is necessary.... As for the ports, well, I asked comsol support once about this and they said they had to be 1/2 wavelength away from the object being probed to avoid near field effects. On the other hand, I was able to reproduce a bunch of other groups' results using ports that were closer to the object than that. It's one of those things were I certainly try to be at least 1/2 wavelength away from the DUT. Unfortunately that's not always the case (plus, as i'm sure you can guess, it increases the number of mesh elements/DOF, so that's unfortunate). So, I guess the punch line is that officially you want the ports 1/2 wavelength away and no closer. However, unofficially, my experience taught me that you can do very good with ports that are closer. Having an analytic or otherwise known problem whose solution you can test this out on would be good if you feel you want to be closer than 1/2 wavelength, but don't want to just take my word for it ;) Good luck again! --matt

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 5 giu 2013, 20:52 GMT-4
Thanks Matt.

That is also the case with me. The wavelength in vacuum for the structure is much larger than the size of the object of interest.

Thanks for your replies!

Alex


Yep, OK...I just mentioned the PEC and PMC because I thought you were only using normal incidence. It is definitely true that floquet is necessary....

As for the ports, well, I asked comsol support once about this and they said they had to be 1/2 wavelength away from the object being probed to avoid near field effects.

On the other hand, I was able to reproduce a bunch of other groups' results using ports that were closer to the object than that. It's one of those things were I certainly try to be at least 1/2 wavelength away from the DUT. Unfortunately that's not always the case (plus, as i'm sure you can guess, it increases the number of mesh elements/DOF, so that's unfortunate).

So, I guess the punch line is that officially you want the ports 1/2 wavelength away and no closer. However, unofficially, my experience taught me that you can do very good with ports that are closer. Having an analytic or otherwise known problem whose solution you can test this out on would be good if you feel you want to be closer than 1/2 wavelength, but don't want to just take my word for it ;)

Good luck again!

--matt


Thanks Matt. That is also the case with me. The wavelength in vacuum for the structure is much larger than the size of the object of interest. Thanks for your replies! Alex [QUOTE] Yep, OK...I just mentioned the PEC and PMC because I thought you were only using normal incidence. It is definitely true that floquet is necessary.... As for the ports, well, I asked comsol support once about this and they said they had to be 1/2 wavelength away from the object being probed to avoid near field effects. On the other hand, I was able to reproduce a bunch of other groups' results using ports that were closer to the object than that. It's one of those things were I certainly try to be at least 1/2 wavelength away from the DUT. Unfortunately that's not always the case (plus, as i'm sure you can guess, it increases the number of mesh elements/DOF, so that's unfortunate). So, I guess the punch line is that officially you want the ports 1/2 wavelength away and no closer. However, unofficially, my experience taught me that you can do very good with ports that are closer. Having an analytic or otherwise known problem whose solution you can test this out on would be good if you feel you want to be closer than 1/2 wavelength, but don't want to just take my word for it ;) Good luck again! --matt [/QUOTE]

Please login with a confirmed email address before reporting spam

Posted: 7 years ago 26 giu 2017, 05:29 GMT-4
Dear sir,
I need a small help.I am also working on a geometry which left and right side is periodic but top and bottom has infinite boundary.I describe the unit cell surrounded by air .For left right ,i could use periodic boundary conditiom.But for the top and bottom, which boundary condition should i use.

Thanks in advance.

Best Regards
Anisuzzaman Boni
Dear sir, I need a small help.I am also working on a geometry which left and right side is periodic but top and bottom has infinite boundary.I describe the unit cell surrounded by air .For left right ,i could use periodic boundary conditiom.But for the top and bottom, which boundary condition should i use. Thanks in advance. 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.