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 The information provided may be out of date.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.

Some problems with the transmission and reflection model: Identity pairs

Please login with a confirmed email address before reporting spam

Hi all,
I want to build a model that calculate the transmission and reflection of light with the normal incidence. So I follow the model example ( Model ID: 6141 a Periodic Structure) to do such a modeling. I used the FEM model brewstersAngle3DTE.mph, except that I fixed incident angle theta=0, and I also the index of all the domains to be n_slab, since firstly I would like to check the light propagation in the homogenous dielectric medium, and the identity pairs that is used in the FEM structure brewstersAngle3DTE.mph

Then I got some problems, which has puzzled me for a long time. The problem is the following,
when n_slab=1, all the light is transmitted, as expected;
when n_slab=1.5, I found that 3.8 percent of the light is reflected, it starts to appear strange.
and when n_slab=2.5, I found that 15.51 percent of the light is reflected.
Since the index over all the domain is the same, so there should be no reflections, 15.51 percent is too large.

I donot think there is any problem with the mesh, since I use the same mesh strategy and use very dense mesh. I digged into the weak formulation of model, currently I am doubting the validity of constraints that applied on the destination boudary of the identity pair,
bnd.constrf = {0,0,0,0,0,0,0,{0;0;0;'if(src2dst_ip5,test(psi-src2dst_ip5(psi)),0)'}, ...
0,0};
bnd.constr = {0,0,0,0,0,0,0,{0;0;0;'if(src2dst_ip5,psi-src2dst_ip5(psi),0)'}, ...
0,0};

I attached the FEM here.
Any comments or hints? Thanks in advance


4 Replies Last Post 03.11.2011, 10:05 GMT-4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.04.2011, 10:28 GMT-4
Hi

what did you expect, ((n-1)/(n+1)^2, hence 4.00 and 18.37% reflectivity ? The slight difference must come from an integration issue, probably you should add the LM on the integration boundaries to get a better result, as I see no absorption in there.

But as you have defined n_slab on all parts, you might notice that the reflection comes from the PML layer, as normally one is assuming they have an index of 1, if I remember right.
Indeed you have given it the same index so one could expect no reflection, obviously not the case. I believe this was discussed some time ago in another thread, have a try of a search on the Forum, and check the KB (knowledge Base) on COMSOL's main site

--
Good luck
Ivar
Hi what did you expect, ((n-1)/(n+1)^2, hence 4.00 and 18.37% reflectivity ? The slight difference must come from an integration issue, probably you should add the LM on the integration boundaries to get a better result, as I see no absorption in there. But as you have defined n_slab on all parts, you might notice that the reflection comes from the PML layer, as normally one is assuming they have an index of 1, if I remember right. Indeed you have given it the same index so one could expect no reflection, obviously not the case. I believe this was discussed some time ago in another thread, have a try of a search on the Forum, and check the KB (knowledge Base) on COMSOL's main site -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.04.2011, 13:55 GMT-4
Thanks Ivar,
Indeed, I expect that reflection behaviers according to ((n-1)/(n+1))^2. I did find a previous thread called 'modeling of reflection/transmission', which is also referred to the model example ( Model ID: 6141 a Periodic Structure) , but the reflection is not discussed at all.
In my model, I donot have any absorptions. I forget to emphasis that the model can model the transmission perfect well when n_slab=1, the reflection I got is 1.0393e-007.
I agree that the integration will give me some problem, since it is always smaller than the input power P0. But we still got some reflection when the index is higher, and that is my main problem. Using scaling wavelength, lambda0/n_slab and keep all the index to be 1, I can have perfect transmission without any significant reflection. In my opinion, such scaling indicates that we donot have problems with the mesh and the PMLs , am I right?


Besides, what do you mean LM on the integration boundary? Do you have some model examples in which LM on the integration boundaries is implemented?

Thanks
Thanks Ivar, Indeed, I expect that reflection behaviers according to ((n-1)/(n+1))^2. I did find a previous thread called 'modeling of reflection/transmission', which is also referred to the model example ( Model ID: 6141 a Periodic Structure) , but the reflection is not discussed at all. In my model, I donot have any absorptions. I forget to emphasis that the model can model the transmission perfect well when n_slab=1, the reflection I got is 1.0393e-007. I agree that the integration will give me some problem, since it is always smaller than the input power P0. But we still got some reflection when the index is higher, and that is my main problem. Using scaling wavelength, lambda0/n_slab and keep all the index to be 1, I can have perfect transmission without any significant reflection. In my opinion, such scaling indicates that we donot have problems with the mesh and the PMLs , am I right? Besides, what do you mean LM on the integration boundary? Do you have some model examples in which LM on the integration boundaries is implemented? Thanks

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 25.04.2011, 16:49 GMT-4
Hi

sorry "LM" is my "short cut" for the Lagrange Multipliers, the difference could come from the precision of the integration, then the LM's should help

Well if you look at the field at the different boundaries, you will see that the source of the reflection is the PML interface so that means, for me that it does not accept a higher index, but from the underlying equations, the index is taken into account.

There are a few threads on the Forum about this, (I have just problems to locate them) try search
one comment is the last one of here www.comsol.eu/community/forums/general/thread/13247/
But there is another one more recent


--
Good luck
Ivar
Hi sorry "LM" is my "short cut" for the Lagrange Multipliers, the difference could come from the precision of the integration, then the LM's should help Well if you look at the field at the different boundaries, you will see that the source of the reflection is the PML interface so that means, for me that it does not accept a higher index, but from the underlying equations, the index is taken into account. There are a few threads on the Forum about this, (I have just problems to locate them) try search one comment is the last one of here http://www.comsol.eu/community/forums/general/thread/13247/ But there is another one more recent -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 03.11.2011, 10:05 GMT-4
Hi, how is your situation now?

Maybe you can try a thicker PML.

Good luck!
Hi, how is your situation now? Maybe you can try a thicker PML. Good luck!

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.