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 The information provided may be out of date.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.
Undefined value error with Scattering Boundary Condition
Posted 25 nov 2014, 07:38 GMT-5 RF & Microwave Engineering, Modeling Tools & Definitions, Parameters, Variables, & Functions Version 4.4 2 Replies
Please login with a confirmed email address before reporting spam
Dear all,
I try to set up an eigenfrequency simulation for an optical nanostructure, where its eigenmodes and corresponding Q-factors should be simulated. To do so, an air cylinder with scattering boundaries is placed around the structure. This method was already used in our research group with Comsol 3.5a, where it perfectly worked. If I try this in the same way as before in Comsol 4.4 I get an error that says that some undefined values where found (I put the complete error at the end).
Even if I take an working mph-file from 3.5a and load and rerun it in 4.4 I get this error.
To demonstrate it, I created a smaller simulation in Comsol 4.4 that should just calculate the optical eigenmodes of a silicon disk. It is attached to this message.
I know that I could also use a PML around the simulation volume (which I already did and which works), but in my particular case I would like to use a scattering boundary. You find the complete error at the end of this message.
Thank you.
Hannes
Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
There are 4992 equations giving NaN/Inf in the matrix rows for the variable comp1.E10.
at coordinates: (1.95739e-006,4.10639e-007,1.5e-006), (1.93312e-006,5.12863e-007,1.5e-006), (1.33784e-006,1.48666e-006,1.5e-006), (1.2579e-006,1.55489e-006,1.5e-006), (1.31557e-006,1.3701e-006,1.5e-006), ...
There are 1664 equations giving NaN/Inf in the matrix rows for the variable comp1.E20.
at coordinates: (1.23741e-006,1.44043e-006,1.5e-006), (1.27943e-006,1.26486e-006,1.5e-006), (1.52137e-006,1.29824e-006,1.40898e-006), (1.4484e-006,1.23084e-006,1.5e-006), (1.85009e-006,7.59702e-007,1.40843e-006), ...
There are 1664 equations giving NaN/Inf in the matrix rows for the variable comp1.E21.
at coordinates: (1.23741e-006,1.44043e-006,1.5e-006), (1.27943e-006,1.26486e-006,1.5e-006), (1.52137e-006,1.29824e-006,1.40898e-006), (1.4484e-006,1.23084e-006,1.5e-006), (1.85009e-006,7.59702e-007,1.40843e-006), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.
- Feature: Eigenvalue Solver 1 (sol1/e1)
I try to set up an eigenfrequency simulation for an optical nanostructure, where its eigenmodes and corresponding Q-factors should be simulated. To do so, an air cylinder with scattering boundaries is placed around the structure. This method was already used in our research group with Comsol 3.5a, where it perfectly worked. If I try this in the same way as before in Comsol 4.4 I get an error that says that some undefined values where found (I put the complete error at the end).
Even if I take an working mph-file from 3.5a and load and rerun it in 4.4 I get this error.
To demonstrate it, I created a smaller simulation in Comsol 4.4 that should just calculate the optical eigenmodes of a silicon disk. It is attached to this message.
I know that I could also use a PML around the simulation volume (which I already did and which works), but in my particular case I would like to use a scattering boundary. You find the complete error at the end of this message.
Thank you.
Hannes
Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
There are 4992 equations giving NaN/Inf in the matrix rows for the variable comp1.E10.
at coordinates: (1.95739e-006,4.10639e-007,1.5e-006), (1.93312e-006,5.12863e-007,1.5e-006), (1.33784e-006,1.48666e-006,1.5e-006), (1.2579e-006,1.55489e-006,1.5e-006), (1.31557e-006,1.3701e-006,1.5e-006), ...
There are 1664 equations giving NaN/Inf in the matrix rows for the variable comp1.E20.
at coordinates: (1.23741e-006,1.44043e-006,1.5e-006), (1.27943e-006,1.26486e-006,1.5e-006), (1.52137e-006,1.29824e-006,1.40898e-006), (1.4484e-006,1.23084e-006,1.5e-006), (1.85009e-006,7.59702e-007,1.40843e-006), ...
There are 1664 equations giving NaN/Inf in the matrix rows for the variable comp1.E21.
at coordinates: (1.23741e-006,1.44043e-006,1.5e-006), (1.27943e-006,1.26486e-006,1.5e-006), (1.52137e-006,1.29824e-006,1.40898e-006), (1.4484e-006,1.23084e-006,1.5e-006), (1.85009e-006,7.59702e-007,1.40843e-006), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.
- Feature: Eigenvalue Solver 1 (sol1/e1)
Attachments:
2 Replies Last Post 22 dic 2014, 07:13 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
I am also having similar problem. I am modelling three nanorods in a air box but scattering boundary condition could not be employed properly.
Undefined value found.
- Detail: Undefined value found in the equation residual vector.
There are 190232 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E10.
at coordinates: (1.3721e-008,1.13259e-008,4.1159e-008), (1.37873e-008,1.11166e-008,4.14483e-008), (1.34433e-008,1.14999e-008,4.12776e-008), (1.35503e-008,1.1087e-008,4.18412e-008), (1.32063e-008,1.14703e-008,4.16705e-008), ...
There are 149529 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E20.
at coordinates: (1.34968e-008,1.12935e-008,4.15594e-008), (1.3913e-008,1.13397e-008,4.16589e-008), (1.36225e-008,1.15166e-008,4.177e-008), (1.36759e-008,1.13101e-008,4.20517e-008), (1.35624e-008,1.09028e-008,4.1242e-008), ...
There are 149529 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E21.
at coordinates: (1.34968e-008,1.12935e-008,4.15594e-008), (1.3913e-008,1.13397e-008,4.16589e-008), (1.36225e-008,1.15166e-008,4.177e-008), (1.36759e-008,1.13101e-008,4.20517e-008), (1.35624e-008,1.09028e-008,4.1242e-008), ...
- Feature: Stationary Solver 1 (sol1/s1)
Undefined value found.
- Detail: Undefined value found in the equation residual vector.
There are 190232 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E10.
at coordinates: (1.3721e-008,1.13259e-008,4.1159e-008), (1.37873e-008,1.11166e-008,4.14483e-008), (1.34433e-008,1.14999e-008,4.12776e-008), (1.35503e-008,1.1087e-008,4.18412e-008), (1.32063e-008,1.14703e-008,4.16705e-008), ...
There are 149529 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E20.
at coordinates: (1.34968e-008,1.12935e-008,4.15594e-008), (1.3913e-008,1.13397e-008,4.16589e-008), (1.36225e-008,1.15166e-008,4.177e-008), (1.36759e-008,1.13101e-008,4.20517e-008), (1.35624e-008,1.09028e-008,4.1242e-008), ...
There are 149529 degrees of freedom giving NaN/Inf in the vector for the variable comp1.E21.
at coordinates: (1.34968e-008,1.12935e-008,4.15594e-008), (1.3913e-008,1.13397e-008,4.16589e-008), (1.36225e-008,1.15166e-008,4.177e-008), (1.36759e-008,1.13101e-008,4.20517e-008), (1.35624e-008,1.09028e-008,4.1242e-008), ...
- Feature: Stationary Solver 1 (sol1/s1)
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
I got a possible solution for the problem.
In Solver Configurations, go to your solver and there to the bullet point Eigenvalue Solver. There one has to enable the "Transform point" and set the value for the Point to the guess of the frequency where the solver should look for.
I recognized there that, in case the volume is just surrounded by scattering boundaries, the convergence may be harmed depending on the mesh one takes. So, if the simulation takes too long, try a different mesh, or, even better try to implement the symmetries of your system, if it has some.
Cheers,
Hannes
In Solver Configurations, go to your solver and there to the bullet point Eigenvalue Solver. There one has to enable the "Transform point" and set the value for the Point to the guess of the frequency where the solver should look for.
I recognized there that, in case the volume is just surrounded by scattering boundaries, the convergence may be harmed depending on the mesh one takes. So, if the simulation takes too long, try a different mesh, or, even better try to implement the symmetries of your system, if it has some.
Cheers,
Hannes