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 The information provided may be out of date.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.

Why simulation doesn't work for structure with small features?

Please login with a confirmed email address before reporting spam

Dear all,

I am doing simulation with two equations in 3d. similar to the "laser heating" model. The laser is not moving and has a 3D intensity profile.

Dependent variables: T I
PDE in general Form.
Flux term
-k_fit*Tx -k_fit*Ty -k_conduct*Tz
0 0 -Iz+k_abs*I
Source Term
Iz*exp((-x^2-y^2)/sig^2)*I0
0
note: k_fit and k_abs are temperature(T) dependent functions. Temperature dependent conductivities and temperature dependent absorption.

During the test with a slab (10um*10um*0.25um) with a hole (r=3um), it works well. However, when I am trying a real simulation with a slab (10um*10um*0.25um) with many small holes (r=0.1um, hole seperation a=0.4um), it shows "singular matrix" errors if "normal mesh":
---------------------------------
Messages:
Singular matrix
Stack trace:
at linearization.cpp, row 2316, (
There are 2 void equations (empty rows in matrix) for the variable I
at coordinates: (-4.15e-006,-2.9471e-006,-5.16988e-022), (-4.15e-006,-3.44417e-006,-4.65289e-022), ...
)
at com.femlab.solver.FlSolver.femStatic(Native Method)
at com.femlab.solver.FemStatic.run(Unknown Source)
at com.femlab.server.FlRunner.run(Unknown Source)
at com.femlab.util.i.run(Unknown Source)
at com.femlab.util.aa.run(Unknown Source)
---------------------------------

If I use coarse mesh, the holes looks like rectangular, there is no error of "Singular matrix", but it can not find the solution. If I use finer mesh, it shows "out of mem". The simulations

The simulations are done on a Win7 32-bit computer: CPU is 2.93GHz, Mem 3GB. Comsol3.5a.

Can any one give any suggestions? Thanks in advance.

best,
Jiang


1 Reply Last Post 13 déc. 2010, 16:48 UTC−5
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 13 déc. 2010, 16:48 UTC−5
Hi

I'm not sure I can help alot, but I have a few comments to share:

1) singular matriux arrives generally for me when I have missed soem BC's (or not defined enough to give an unique solution)
2) there are the geoemtrical resolution, the mesh resolution and the relative error for geometrical analysis and the absolute reloution one for CAD import that must be distinguisehed for. The relative and abosolute errors can interfer with objects smaller than a few microns, sou you should set them to 1E-8 or smaller if you import small CAD objects, els some small features might simply "dissapear".

Then indeed with a coarse mesh a round hole wil become "square", if you use 2nd order shape functions this will help to "round" the edges, but the best is to increase the mesh density to > 6 nodes per circle edge.

Sometimes the graphical reolsution can play some games with you (have seldom happened to me) there is some settings in the options to tweak if this is the case, but this changes only the graphical rendering of the display, nothing to do with calculation errors

--
Good luck
Ivar
Hi I'm not sure I can help alot, but I have a few comments to share: 1) singular matriux arrives generally for me when I have missed soem BC's (or not defined enough to give an unique solution) 2) there are the geoemtrical resolution, the mesh resolution and the relative error for geometrical analysis and the absolute reloution one for CAD import that must be distinguisehed for. The relative and abosolute errors can interfer with objects smaller than a few microns, sou you should set them to 1E-8 or smaller if you import small CAD objects, els some small features might simply "dissapear". Then indeed with a coarse mesh a round hole wil become "square", if you use 2nd order shape functions this will help to "round" the edges, but the best is to increase the mesh density to > 6 nodes per circle edge. Sometimes the graphical reolsution can play some games with you (have seldom happened to me) there is some settings in the options to tweak if this is the case, but this changes only the graphical rendering of the display, nothing to do with calculation errors -- Good luck Ivar

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.