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 The information provided may be out of date.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.

Trouble with meshing small structures

Please login with a confirmed email address before reporting spam

Hi,

I have rib waveguide whose rib height comes down to height of 0.5 nm and 4 um width. This is causing me some problems. I am using free triangular with smallest size of 4e-12, resolution of curvature to 0.2, resolution of narrow regions 1. Ideally, I would like to run this relative fast.

I get errors like,

Warning: The relative residual is greater than the relative tolerance.
Warning: Relative residual 2.5e-14

Any tips on how to get around this problem?


Thanks

6 Replies Last Post 12 oct. 2011, 10:12 UTC−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 11 oct. 2011, 08:51 UTC−4
Hi

1:10^6 in geoemtry size ratio is a lot !, not sure the mesher is managing to solve that in one go, I usually limit my work to 1:10'000 and I isolate small features with a "box or circle" so that I can mesh the content first and then the external part with a ration never exceeding 1:10'000

Such large differences might also lead to numerical errors, as "eps", or rather "sqrt(eps)" if double real numbers is used, is only around 1E-8

--
Good luck
Ivar
Hi 1:10^6 in geoemtry size ratio is a lot !, not sure the mesher is managing to solve that in one go, I usually limit my work to 1:10'000 and I isolate small features with a "box or circle" so that I can mesh the content first and then the external part with a ration never exceeding 1:10'000 Such large differences might also lead to numerical errors, as "eps", or rather "sqrt(eps)" if double real numbers is used, is only around 1E-8 -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 11 oct. 2011, 09:12 UTC−4
1:10^6 in geoemtry size ratio is a lot !, not sure the mesher is managing to solve that in one go, I usually limit my work to 1:10'000 and I isolate small features with a "box or circle" so that I can mesh the content first and then the external part with a ration never exceeding 1:10'000


Thank you for replying. I have some follow-up questions.

- I am not sure how to isolate these features? What do you mean with "box or circle"?

Do you mean I should try using box-shaped elements instead of free triangular?
Then how do I solve the boundary problem to the larger sized elements around the rib?
I am doing a mode analysis so this rib feature is very important for the mode confinement.

- It may be a silly question but ..how do you determine the geometry size ratio?
[QUOTE]1:10^6 in geoemtry size ratio is a lot !, not sure the mesher is managing to solve that in one go, I usually limit my work to 1:10'000 and I isolate small features with a "box or circle" so that I can mesh the content first and then the external part with a ration never exceeding 1:10'000[/QUOTE] Thank you for replying. I have some follow-up questions. - I am not sure how to isolate these features? What do you mean with "box or circle"? Do you mean I should try using box-shaped elements instead of free triangular? Then how do I solve the boundary problem to the larger sized elements around the rib? I am doing a mode analysis so this rib feature is very important for the mode confinement. - It may be a silly question but ..how do you determine the geometry size ratio?

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 11 oct. 2011, 10:12 UTC−4
Hi

no my "box" is simply to surround the "small feature" by an articifial internal boundary limit, so that you get a domain (or two) extra, you mesh the small features first, and then the larger ones. Just make it shire that the side of the box or sphere is not more than 1:10'000 larger than the small sizes. i.e. a 10nm item, is surrounded by a cube or a square (cercle or sphere) of some 10-100 um size

--
Good luck
Ivar
Hi no my "box" is simply to surround the "small feature" by an articifial internal boundary limit, so that you get a domain (or two) extra, you mesh the small features first, and then the larger ones. Just make it shire that the side of the box or sphere is not more than 1:10'000 larger than the small sizes. i.e. a 10nm item, is surrounded by a cube or a square (cercle or sphere) of some 10-100 um size -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 11 oct. 2011, 10:27 UTC−4

Hi

no my "box" is simply to surround the "small feature" by an articifial internal boundary limit, so that you get a domain (or two) extra, you mesh the small features first, and then the larger ones. Just make it shire that the side of the box or sphere is not more than 1:10'000 larger than the small sizes. i.e. a 10nm item, is surrounded by a cube or a square (cercle or sphere) of some 10-100 um size

--
Good luck
Ivar


I see, so are you are saying that you can mesh one domain at the time and that each domain shouldn`t be more than 1:10´000 larger then the neighbouring domains? By adding more domains this will improve the meshing, even if they are of the same or different material?
[QUOTE] Hi no my "box" is simply to surround the "small feature" by an articifial internal boundary limit, so that you get a domain (or two) extra, you mesh the small features first, and then the larger ones. Just make it shire that the side of the box or sphere is not more than 1:10'000 larger than the small sizes. i.e. a 10nm item, is surrounded by a cube or a square (cercle or sphere) of some 10-100 um size -- Good luck Ivar [/QUOTE] I see, so are you are saying that you can mesh one domain at the time and that each domain shouldn`t be more than 1:10´000 larger then the neighbouring domains? By adding more domains this will improve the meshing, even if they are of the same or different material?

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 11 oct. 2011, 12:17 UTC−4
Hi

yes, thats it, you add a few internal boundaries it doesnt really matter for the model, also when made of same material

--
Good luck
Ivar
Hi yes, thats it, you add a few internal boundaries it doesnt really matter for the model, also when made of same material -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 12 oct. 2011, 10:12 UTC−4
Okay, thanks I will try that.
Okay, thanks I will try that.

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.