Discussion Forum

Hi

have you checked that your mesh to wavelength is always respecting >3-5 elements per wavelength, also in the most dense material ?

There is an example somewhere on how to ensure the mesh is at least a few times smaller than the wavelength in all materials, and restricting h or the max mesh size with a variable depending on the wavelength and the index

I do not really remember any other phenomena that could induce you in error apart from that, bit I might also have missed an important point ;)
--
Good luck
Ivar

Thanks for your quick answer Ivar!

I've repeated the simulations being very carefully with the mesh, considering more than 3 elements per wavelength, and I'm still getting the same "bad" results.

I'll try to solve it, if something comes up I'll let you know.

Thanks and keep it up!

Hi there,

I'm not 100% sure, but I'm a little concerned about your boundary conditions on the sides (PMC and PEC) to be specific. If I were doing it, I'd at least enforce scattering BCs on the sides that don't have a PML (or the side where the plane wave launches). Ideally you'd like to have PMLs on every side, with appropriate absorption coefficients (PML should normally be of thickness >=wavelength....but smaller sections can do if you set the PML scale factor appropriately (that's what I read in comsol documentation anyway, I'm still myself unsure about it all).

By enforcing PMLs on all but the input sides, you won't have any reflections from the PEC or PMC boundaries. I'm thinking your results change because when you change the size of the box you get different reflection patterns from the sides and hence different standing wave effects in the region in and around the antenna.

Of course, adding more domains increases sim time and memory requirements, so I don't know if it's possible for you to do it. See the "dielectric scattering PML.mph" file in the comsol model files and tutorial documentation. It's a 2D model, but you'd just want to do the same thing in 3 dimensions.

Anyway, not sure if this helped, but hope so!

--Matt

Hi everybody,

I am almost working on the same model. I am trying to model a dipole antenna placed on a dielectric substrate and I put my structure inside air box. I added to PML boxes above and below air box, the side walls are defined as PEC whereas the top surface is defined as SBC with plane wave excitaion .
My question is : please, if I want to launch an x-polarized electric field perpendicular to my antenna, which excitation is true?

Ex=exp(-j*k0_rfw*z)

or

Ez=exp(-j*k0_rfw*x)

I appreciate any help.

Cheers

Ahmed

Hi there,

The first one is correct.

Ex=exp(-j*k*z) means the E field has only an x component (because you're only setting Ex) and exp(-j*k*z) means the wave travels in the + z direction.

In general:

E_component = Amplitude * exp (-j * (wavenumber_in_medium) * Direction_of_propagation)

In your case, the amplitude is 1 V/m, the wavenumber is set by comsol (k = 2*pi/lambda_0, or said another way k = 2*pi/(free_space_wavelength), and z is the direction of propagation.

Hope this helps to clear things up. Good luck with your model!

--Matt

Hi Matt,

Thanks a lot for your reply, now it is clear to me. Actually, I face a problem now: I am working on total field and when I run the simulation without PMl it finds a solution, but when I set the outer sphere as PML the simulation continue with no convergence. So please do you have any idea on this problem ? (I am working on COMSOL v3.4)

Should I set PML when solving for total field?


I appreciate your help and thanks in advance

Ahmed

Hi Ahmed,

I'm not sure I know what you mean by total field. Do you mean the norm of the E-Field? (sqrt(Ex^2+Ey^2+Ez^2))?

Or maybe you mean incident field plus scattered field?

If you uploaded your model, I could take a look and see if I see anything that looks out of sorts. If it's a sensitive piece of work (ip issues, etc...) then maybe at least a few screenshots with your geometry, setup, and errors may help.

In either case, you'll probably want the PML ON when you solve for total field. The idea behind the PML is to minimize non-physical reflections between the essentially artificial boundary that you create when you surround the antenna by air. Otherwise, without a PML, fields radiated by the antenna could hit the air box and reflect back into the domain surrounding the antenna.

Since in reality there would be no reflections, you use a PML to prevent these simulation artifact reflections.

What are the boundary conditions on the outermost sphere? They should be scattering, I think.

Ok let me know if this helps thanks!

--Matt

Hi Ahmed,

I'm not sure I know what you mean by total field. Do you mean the norm of the E-Field? (sqrt(Ex^2+Ey^2+Ez^2))?

Or maybe you mean incident field plus scattered field?

If you uploaded your model, I could take a look and see if I see anything that looks out of sorts. If it's a sensitive piece of work (ip issues, etc...) then maybe at least a few screenshots with your geometry, setup, and errors may help.

In either case, you'll probably want the PML ON when you solve for total field. The idea behind the PML is to minimize non-physical reflections between the essentially artificial boundary that you create when you surround the antenna by air. Otherwise, without a PML, fields radiated by the antenna could hit the air box and reflect back into the domain surrounding the antenna.

Since in reality there would be no reflections, you use a PML to prevent these simulation artifact reflections.

What are the boundary conditions on the outermost sphere? They should be scattering, I think.

Ok let me know if this helps thanks!

--Matt

Dear Matt,

Thanks a lot for your reply and your offer to help. Please find attached a simple model of mine, I didn't mesh to reduce the file size, so please mesh it and have a look too see what is wrong with it?

Thanks again mate.

Ahmed

Dear all,

As Matt correctly pointed, if you want to generate an x-polarized electric field propagating in the z direction you must use the Ex=exp(-j*k*z) definition.

Now I'm currently simulating the same dipole structure but with PML regions surrounding the whole air box. I'll also try to use SBC where I have the PEC and PMC BC. I'll let you know if something comes up.

Ahmed, since your model has been developed with COMSOL v3.4 I can not open it. Can you try to save it with another version?

Cheers!

Hi Jordi,

Thanks a lot for your explaination. Unfortunately, I don't have another version of COMSOl. Actually, I tried to enclose my model (antenna placed on dielectric substrate) by two spheres the inner is ear and the outer is PML and defined as scattering boundary condition with excitation. But the problem is the simulation continue with no convergence while when I remove PML I get solution but I think it is wrong!!


Cheers



Ahmed

Hi Ahmed,

Still simulating... by the way, where are you defining your electric field E=Ex*exp(-j*k0_rfw*z)? Are you using the option "Background electric field"? (I don't know if in v3.4 there's a similar option). Or you are defining an x-surface current in your sphere? or... something else?

Cheers!

Hi Jordi,

Actually, I don't have such option in my version. I usualy put this formula in the scalar variables and in the scattering boundary condion of the outer sphere I put "1" in the field of Ex , and "0" in the other fields. Do you think it's ture?


Cheers!

Hi Ahmed,

I looked at your model and I have a couple of things to mention that might help get you along:

1. In your specification of the relative permittivity of the metal bowtie, I'm not I see the real part as - - 5050 .....The two minus signs make it a positive. You're working in the GHz range, and I think the permittivity of metals in that range should be negative if you choose to specify it's electromagnetic properties with eps_r alone...

2. It looks like you're trying to launch the x-polarized field through a scattering boundary condition, I'm not sure that's possible....Take a look at the powerpoint in the link below....they show how to launch a plane wave with x-polarization....it's a little different than yours, but the process would be the same, changing only some of the inputs. Long story short, you might want to use a port BC to launch the wave.

www.comsol.com/showroom/documentation/model/6141/

I'm also not sure if you sent me the correct model, because in the above posts you mentioned a spherical PML but I'm only seeing a multi section box with the bowtie and substrate in the middle.

Also, what was the output you were going to extract vs. frequency?

Ok then, I hope that's all of some use!

--Matt

Hi Matt,

Thank you very much for your interesting in my case, I appreciate your help and support. Actually, I working on two models: one is consists of boxes (previously sent to you), and the other is Bowtie antenna placed on dielectric substrate and enclosed by two spheres (attached). I am looking forward to calculate the electric field in the feed gap of antenna vs. frequency. The (--) signs in the previous model is done by mistake , I really meant (-5050+i*1090).

Regarding launch the x-polarized field through a scattering boundary condition, I read it in a paper they explain this method !!
Thank you again for the link , I will see how do they use the excitation.

Please find attached (if you have time) the second model, any comments will be highly appreciated.

Kind Regards

Ahmed

Dear all,

Last week I wrote to COMSOL support about the issues related with this thread. As I find the answer very useful, I would like to share it with you. I hope that this will help a lot of people.

In the copied text there is first my question and then the COMSOL support answer. If you find it difficult to read please don't hesitated to say it.
______________________________________________________________________________________________
Dear Jordi,
Thanks for the request. Please find my comments below:

Methods of Excitation:
1.- Defining a surface current at the top of the air box, doing so I can be able to generate a plane wave with and electric field of E(1,0,0) (only Ex component) propagating in the z direction.

Ans: Using the surface current setting, one can excite plane waves.This is a generic technique for exciting plane waves and can be used in combination with PMLs.

2.- Using a "Background electric field" defined as E(1,0,0) in the Settings of the Electromagnetic waves physic choosing the option "scattered field" in the "Solve for" setting.

Ans: This option is useful for generating plane waves in situation when scattering from an isolated object in free space is to be studied. This can be specified only while using the Scattered field formulation.

3.- Defining a port at the top of the air box that would generate an electric field E(1,0,0) with the propagation constant defined as emw.k0 and a value at the "Port input power" leaving the "wave excitation at this port" in the On position.

Ans: The port is once again useful for generating plane waves. It also absorbs or acts as matched boundary only for waves of similar type. Any other type of wave is not absorbed by ports leading to unrealistic results.A major advantage of using ports is for readily calculating S-parameters or reflection and transmission co-efficient of the device/system.

4.- Using the "scattering boundary condition" on the top of the air box,choosing "plane wave" at the "wave type" option and generating an electric field E(1,0,0).

Ans: Scattering boundaries are just like ports. They can be set to excite and absorb plane waves. However, they do not calculate S-parameters.
Additionally, you can also use the Lumped Port boundary condition as an excitation. The lumped ports are used over boundaries bwtween two conductors and are lot easier to set-up when compared to generic ports. However, these can be used only when the distance between the conductors is smaller compared to the wavelength of excitation and might not be useful for optical sources.

Regarding the Perfectly matched layer:

1.- Defining two PML zones at both extremes (top and bottom of the air box) and PMC and PEC boundary conditions at its sides (PMC to the sides perpendicular to the magnetic field and PEC to the sides perpendicular to the electric field). Doing so is COMSOL considering that surrounding the optical antenna there are more antennas like the one I'm simulating?(Like a mirror)

Ans: Yes. It uses the image theory.

2.- Surrounding the whole air box with six blocks of PML (24 PML domains). Doing so is COMSOL considering that the antenna is alone and is not receiving influences from other antennas? If I use this option, is the plane wave affected for the PML zones?

Ans: Yes. This models a situation where in you are having a source in places in an infinite medium. If the PMLs are correctly set-up you should see that the E field amplitude reduces significantly as it passes through the PML.


I'm doing a frequency domain analysis, sweeping the value of the frequency. Doing so I expect to see a peak in the induced electric field in the antenna when the sweep reaches the resonance frequency of the antenna. The problem is that I'm getting different results for different dimensions of the air box and for the different methods above described. Since I'm not change de dimensions of the dipole antenna this is quite wired.

Ans: One of the reasons for this appears to be the fact that the PMLs might not be set-up correctly, which is why the resonant length is changing. Please see that the PML is wide enough that it contains at least 5-6 elements. Also, a box-like PML would work for plane waves. If the waves are spherical in nature, then a spherical PML would have to be used.

Hope this helps with your model set-up. Please feel free to let me know if you have any questions.
Regards,
COMSOL Support.
______________________________________________________________________________________________

Good luck!

Jordi

Hi Jordi,

Thanks a lot for these useful answers, they are very interesting and will help indeed.

Good Luck

Ahmed