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.

scattering : convergence issues at longer wavelengths

Please login with a confirmed email address before reporting spam

Hi,

I am trying to observe scattering from gold hemispherical nanoparticle (r = 50 nm) placed on substrate. I am modelling substrate using a slab (silicon). I am using scattered field formulation. I am attaching the convergence plot. I am trying to do parametric sweep from 400 nm to 800 nm. What I don't get is if it is converging for shorter wavelengths, why it is not converging (or taking longer) for longer ones. The issue starts beyond 600 nm. Can anyone please help? the system I am using has 16 GB RAM (unfortunately).

Regards

Hamza



1 Reply Last Post Dec 20, 2019, 11:38 p.m. EST
Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 5 years ago Dec 20, 2019, 11:38 p.m. EST
Updated: 5 years ago Dec 20, 2019, 11:43 p.m. EST

Without the details of your model, I can only guess. BUT, I can tell you that the BiCGStab solver often fails to converge in RF problems, especially if the mesh is not high quality or you use elements with linear discretization. If you are using linear elements, try quadratic. If you are using quadratic, consider cubic. Look at your mesh quality. In particular, make sure the minimum mesh quality is not very low. Re-do your mesh and/or clean up your geometry in a way to significantly improve your minimum mesh quality. Alternatively, if you have enough memory, change the solver to PARDISO. It is much more robust. if you don't have enough memory for PARDISO, change the solver to GMRES. It isn't fast, but it is much more likely to converge steadily than BiCGStab.

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
Without the details of your model, I can only guess. BUT, I can tell you that the BiCGStab solver often fails to converge in RF problems, especially if the mesh is not high quality or you use elements with linear discretization. If you are using linear elements, try quadratic. If you are using quadratic, consider cubic. Look at your mesh quality. In particular, make sure the minimum mesh quality is not very low. Re-do your mesh and/or clean up your geometry in a way to significantly improve your minimum mesh quality. Alternatively, if you have enough memory, change the solver to PARDISO. It is much more robust. if you don't have enough memory for PARDISO, change the solver to GMRES. It isn't fast, but it is much more likely to converge steadily than BiCGStab.

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.