Discussion Forum

You can try only select the air domain for the far field domain node. I am not sure if this is correct rigorously because in this way the far field calculation surface you should select is part of the PML inner surface. But you can try it and see if you can get the results you expect (maybe for some scattered angles, it is a good approximation).

Another way is that you put a plane above the dipole and a plane inside the substrate, then calculate the far field through the calculated fields on the plane. But in this way, you do the far field calculation by yourself.
If you really want to use comsol's far-field calculation. An idea is that 1.you save your near-field results.2,build another empty model (I mean no glass substrate) and put the near-field results into this model 3.The far field for this model is homogeneous now, so run it and you can get the far field results. This is a bit more complicated and you need to be careful about every step.

Dear Xu Xu

Thank you very much for your reply , I really appreciate it.

I understand from your reply that I have three choices to try :

-the first choice is to calculate the far field only in the air domain
-the second is to calculate it into two parts one in air and the second in substrate.
- The third one is to calculate the near field for the dipole while I have both air and substrate then use this field to calculate the far field

The first two choices can be done easily , but I still don't understand how to do the third one , do I will use the near field as background field for the far field calculation , or how to implement this near field data on the blank model to calculate the far field



Thanks in advance

Hi all

I still don't trust the results that I have , does any one made this kind of calculations before

Thanks in advance

M.H.Elshorbagy

Hi M.H.

Not sure whether you have a solution yet, but what I might guess is that your current simulation is calculating the far-field distribution coming from the antenna and substrate combined. This means that, for instance, the antenna contribution will become smaller as your simulation domain gets larger and the scattered field is dominated by that from the 50% substrate.

If this is the problem, you might need to consider a way to isolate the scattered fields coming from just the antenna. An option could be to set up your simulation into two steps as per a superposition:
(1) create a first simulation step that calculates the total fields produced after your externally applied field is incident on the substrate without antenna. What you want from this simulation is the total fields in the volume/domain where your antenna would otherwise be located (i.e. if it wasn't removed from this simulation step).
(2) create a second simulation step to calculate the scattered field for the complete antenna + substrate, but you instead now specify the externally applied field as the total fields calculated in (1), and only in the domain of the antenna. Effectively, the antenna alone will now behave as the source of this simulation. This simulation should give you a far-field pattern that comes a bit more exclusively from just the antenna, and will hopefully look like the dipole on substrate far-field pattern you were looking for. A similar approach for this step could be to use no externally applied fields, and instead use a volume polarization distribution to impose the applied fields from (1) in (2).

Finally, I suspect the above approach may only work in frequency domain, but a first order approximation could also be done in time domain by calculating the total fields both with and without antenna, then taking their difference to find the approximate scattered field and subsequent far field profile. This won't work if the substrate-antenna interaction is too strong.

Hopefully this is relevant to your particular issue!

Cheers,
Ben

Dear Dr Ben

Thank you for your kind reply , I really appreciate your help . In fact I set several contacts with COMSOL people and we try your solution , but I didn't trust the result , while they told me that the equation that used by comsol to calculate the far-field is assuming that the resonant element is compelety embdded in air

while we have a different situation of two media one above the element (air) and another below it ( substrate ).

We still work on the problem , to set a solution for the far field in a non homogeneous media or multi- layer systems

I really appreciate you help and still need any experience or information to share


Thanks in advance

M.H.Elshorbagy

UCM Spain

I see there are a few other posts where it is concluded that the far field domain only works correctly for a homogeneous domain that surrounds the scatterer. However, I had a quick look into the accuracy of using two far field domains to calculate far field distributions from an 'antenna on substrate' model, and it didn't look too bad. Specifically, with my antenna positioned at the origin, I compared the far field calculation to the radial component of the time averaged Poynting vector at a wavelength away from my antenna.

Case 1: far field calculation
Two far field domains: one for each homogeneous hemisphere enclosing my antenna on substrate, with scattered field from the antenna calculated as per my first post. The scattered simulation is called ewfd2, and I called the two far field results Efar2 and Efar3, I then plot a 3D far field of the quantity:
(abs(ewfd2.Efar2x+ewfd2.Efar3x)^2+abs(ewfd2.Efar2y+ewfd2.Efar3y)^2+abs(ewfd2.Efar2z+ewfd2.Efar3z)^2)^0.5

Case 2: outgoing time averaged Poynting vector
I have a sufficiently large simulation domain to be able to fit a one wavelength radius sphere (radius = R) around my antenna.
I then make an 3D plot group with an isosurface of the quantity:
"(x^2+y^2+z^2)/(R^2), with levels at value 1.
This is my one wavelength radius sphere. To this isosurface I add first a color expression:
(ewfd2.Poavx*x+ewfd2.Poavy*y+ewfd2.Poavz*z)/R
Then I add second a deformation expression to the isosurface:
x component: (ewfd2.Poavx*x+ewfd2.Poavy*y+ewfd2.Poavz*z)/R*(x/R)
y component: (ewfd2.Poavx*x+ewfd2.Poavy*y+ewfd2.Poavz*z)/R*(y/R)
z component: (ewfd2.Poavx*x+ewfd2.Poavy*y+ewfd2.Poavz*z)/R*(z/R)
Provided one wavelength is sufficient to reach a far field regime (where 1/r decay dominates the E and and H-field), this should be the correct far field from the antenna

While I can't seem to attach pictures of these, I will say that there are quantitative differences between the two cases, but qualitatively they look similar (same general shape). Actually, the main issue is the Poynting vector ("true") calculation, which seems to have discontinuities imprinted in the radiation pattern from the mesh, but hopefully (*maybe) this would improve if I did denser meshing of the radius=R sphere.

Ben

Dear Dr Ben

First I would like to thank you for your kind help , I really appreciate it.

Then Your calculations and suggestions is very interesting to me , in fact it will help to much. We keep working on the problem until having a good results using several approaches. So if possible let us keep in touch I will let you knew the results .


Thanks to much ,

Sincerely

M.H.Elshorbagy

Hi,

I am working in a similar problem trying to calculate the far field in an inhomogeneous medium.
I do not think that you can use the Comsol far field calculator since the method works only in homogeneous media. What you could do though is calculate the near field of your antenna in
the homogeneous area in Comsol and then follow the techniques in this paper : DOI: 10.1021/acsphotonics.5b00559 . You will probably need to export the field to MatLab
to do that.

Best of luck and let me know about your progress.

Dear Konstantinos

Thank you very much for your kind response and help , i really appreciate it , I will try it and let you knew the results

Thanks too much

M.H.Elshorbagy

UCM Spain

Just to make sure that you saw this reference in the paper, I put it here.

www.lp2n.institutoptique.fr/Membres-Services/Responsables-d-equipe/LALANNE-Philippe

They have Matlab code that works directly with Comsol.

Dear Dr Konstantinos

Thanks too much for the Link

Sincerely

M.H Elshorbagy