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.

How to simulate using wave optics a model of big dimensions

Guillermo Martínez-Denegri

Please login with a confirmed email address before reporting spam

Hello,

I'm trying to simulate using the Wave Optics module what happens when a light source (450 nm) is intruduced in a device with certain geometry, how the electromagnetic field develops throughout the structures and what are the characteristics of it once the light goes out the structure. I'm modeling this as a 2D structure using Frequency Domain. The problem I have is that the geometry consists of an array of semicylinders with a diametre of 100 um, but I need to test what happens along many of them (let's say 20). Also, this structure has a thickness of 200 um but I need to add to my model a domain to take into account the environment (air). After that I will need to add certain dielectric multilayer structure to change the polarization, but first let's start with the geometry without the multilayer design. The problem I encounter is that my meshing must be of 450/6=75 nm each domain, this is too small, it takes ages to do the meshing and I suppose that if I run the simulation it will take much more.

Any suggestion of how I can simmulate this saving memory and time? At the moment I'm just trying with 3 cylinders, but even this is being impossible. The meshing needs to have this dimensions everywhere or far from the domain borders it can be bigger? I've thought of using in some areas Beam Envolope instead of wavelength domain. Where do you think I could use Beam Envolope without affecting my results?

Thanks in advanced!

Regards,

Guille


3 Replies Last Post 06.01.2019, 16:18 GMT-5
Ulf Olin COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 6 years ago 17.12.2018, 08:40 GMT-5

Hi Guilermo,

Problems that are large compared to the wavelength are heavy to solve. As a starting point, you should consider whether diffraction effects are important in your model. If not, the Ray Optics Module, is a better choice for you.

As a general advice, you should try to reduce the model domain as much as you can using symmetry considerations (mirror symmetry? periodic symmetry?) and by truncating the calculation domain using scattering boundary conditions or PMLs to absorb the outgoing waves.

You can try using a coarser mesh in parts of your structure where you don't have much power density. However, consider first if you can remove those parts completely from the calculation. If you make a coarser mesh in some parts, it is still good to make a mesh convergence study, where you refine the mesh in steps, to see that you have a result that seems to be converged to a physically sound solution.

The Beam Envelopes interface is good if the solution consists of mainly one or two propagation directions/modes that you know of in advance. If you have a scattering proplem, where the scattered wave propagates in many different directions, the Beam Envelopes interface will not help you.

If you consider coupling the regular Frequency Domain interface with the Beam Envelopes interface, you can find some information about this procedure in this blog post: https://www.comsol.com/blogs/2-methods-for-simulating-radiated-fields-in-comsol-multiphysics/

Good luck with your modeling project.

Best regards,

Ulf

Hi Guilermo, Problems that are large compared to the wavelength are heavy to solve. As a starting point, you should consider whether diffraction effects are important in your model. If not, the Ray Optics Module, is a better choice for you. As a general advice, you should try to reduce the model domain as much as you can using symmetry considerations (mirror symmetry? periodic symmetry?) and by truncating the calculation domain using scattering boundary conditions or PMLs to absorb the outgoing waves. You can try using a coarser mesh in parts of your structure where you don't have much power density. However, consider first if you can remove those parts completely from the calculation. If you make a coarser mesh in some parts, it is still good to make a mesh convergence study, where you refine the mesh in steps, to see that you have a result that seems to be converged to a physically sound solution. The Beam Envelopes interface is good if the solution consists of mainly one or two propagation directions/modes that you know of in advance. If you have a scattering proplem, where the scattered wave propagates in many different directions, the Beam Envelopes interface will not help you. If you consider coupling the regular Frequency Domain interface with the Beam Envelopes interface, you can find some information about this procedure in this blog post: https://www.comsol.com/blogs/2-methods-for-simulating-radiated-fields-in-comsol-multiphysics/ Good luck with your modeling project. Best regards, Ulf

Guillermo Martínez-Denegri

Please login with a confirmed email address before reporting spam

Posted: 6 years ago 17.12.2018, 09:09 GMT-5

Hello Ulf,

Thanks for your reply. Of course diffraction effects are important in my model, take into account that in the end what I want my structure for is to diffuse and polarize light which is going from one medium to air.

What do you mean by trying simmetry considerations? My model is periodic since it is an array but I don't understand how the "model domain" can be reduced? What will the advantages be and how to apply this reduction? I used PML for the "edges" of the air domain and Perfect Electric Conductor for the bottom of my geometry to achieve ideal reflexion. In general, scattering should not be a problem since I'm not considering roughness in any interface and the material i'm using is homogeneous.

Best regards,

Guille

Hello Ulf, Thanks for your reply. Of course diffraction effects are important in my model, take into account that in the end what I want my structure for is to diffuse and polarize light which is going from one medium to air. What do you mean by trying simmetry considerations? My model is periodic since it is an array but I don't understand how the "model domain" can be reduced? What will the advantages be and how to apply this reduction? I used PML for the "edges" of the air domain and Perfect Electric Conductor for the bottom of my geometry to achieve ideal reflexion. In general, scattering should not be a problem since I'm not considering roughness in any interface and the material i'm using is homogeneous. Best regards, Guille

Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 6 years ago 06.01.2019, 16:18 GMT-5

I'll add two suggestions to the above:

  1. If you have a large periodic structure (i.e., an array of many elements, equally spaced) and if you can convince yourself that its local physics behavior can be reasonably represented by an infinite array of such elements, then you may wish to employ Floquet boundary conditions. This approach will let you model just one unit cell rather than a large array, but as immersed in an (equivalent) infinite array. You have to set the Floquet conditions carefully, but it can be done. I've done it in the RF module. (I don't know if the Wave Optics module includes the same feature, but I'm guessing it does.)

  2. As another potentially useful technique, you may find it possible to extend your computations to greater distances via a near-field aperture-integration approximation: See https://www.comsol.com/community/exchange/672/ .

Good luck.

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
I'll add two suggestions to the above: 1. If you have a large periodic structure (i.e., an array of many elements, equally spaced) and if you can convince yourself that its local physics behavior can be reasonably represented by an *infinite* array of such elements, then you may wish to employ Floquet boundary conditions. This approach will let you model just one unit cell rather than a large array, but as immersed in an (equivalent) infinite array. You have to set the Floquet conditions carefully, but it can be done. I've done it in the RF module. (I don't know if the Wave Optics module includes the same feature, but I'm guessing it does.) 2. As another potentially useful technique, you may find it possible to extend your computations to greater distances via a near-field aperture-integration approximation: See https://www.comsol.com/community/exchange/672/ . 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.