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.
Meshing of thin film electrodes for a capacitance sensor
Posted Feb 17, 2011, 9:06 p.m. EST 9 Replies
Please login with a confirmed email address before reporting spam
Greetings!
Im a graduate student at Hogskolen i Vestfold in Norway. As a part of my project I need to make a model of a capacitance sensor that is based on thin film Au electrodes on a Pyrex substrate covered with PTFE film.
The problem is that when I build electrodes with their real geometry I am unable to mesh them. The electrodes have a very, very high aspect ratio being only 100nm thick but about 10cm long. Meshing this as a free tetrahydral mesh takes unreasonably long - 2-3% per hour(never actualy finished meshing), so I tried using boundry layer meshing: selected boundry layer from mesh tab, selected the domains I wanted to mesh that way, on boundry layer properties selected the number of boundry layers to be 1 and first layer thickness equal to layers total thickness - 100nm. Meshing goes faster but still freezes at about 30% of creating innitial free mesh.
When I set the electrode thickness to a greater value like 10-100um instead of 100nm the model meshes with no problem and produces plausible C values.
Im confused... Is there any way to mesh this problem? The interdigital thin film electrode geametry is in the picture attached.
Any help apreciated.
Cheers!
- Fjodor
Im a graduate student at Hogskolen i Vestfold in Norway. As a part of my project I need to make a model of a capacitance sensor that is based on thin film Au electrodes on a Pyrex substrate covered with PTFE film.
The problem is that when I build electrodes with their real geometry I am unable to mesh them. The electrodes have a very, very high aspect ratio being only 100nm thick but about 10cm long. Meshing this as a free tetrahydral mesh takes unreasonably long - 2-3% per hour(never actualy finished meshing), so I tried using boundry layer meshing: selected boundry layer from mesh tab, selected the domains I wanted to mesh that way, on boundry layer properties selected the number of boundry layers to be 1 and first layer thickness equal to layers total thickness - 100nm. Meshing goes faster but still freezes at about 30% of creating innitial free mesh.
When I set the electrode thickness to a greater value like 10-100um instead of 100nm the model meshes with no problem and produces plausible C values.
Im confused... Is there any way to mesh this problem? The interdigital thin film electrode geametry is in the picture attached.
Any help apreciated.
Cheers!
- Fjodor
Attachments:
9 Replies Last Post Sep 18, 2012, 3:57 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hei Fjodor
Well why should you mesh the film in it thickness if it is really so "thin" ?
Do you really want to analyse what is happening in the thickness of the film ?
If not, try the thin film boundary layer BC, see the doc
The true difficulty is when you have more than 1 thin film stacked on the same boundary (currently, from my knowledge, not implemented in COMSOL)
Then when you have "simple" square/rectangular geometries, it's often easier to leave it a s a series of adjacent rectangles, with some common interiour boundaries. Then you can easier handle the meshing sequences.
--
Lykke til
Ivar
Well why should you mesh the film in it thickness if it is really so "thin" ?
Do you really want to analyse what is happening in the thickness of the film ?
If not, try the thin film boundary layer BC, see the doc
The true difficulty is when you have more than 1 thin film stacked on the same boundary (currently, from my knowledge, not implemented in COMSOL)
Then when you have "simple" square/rectangular geometries, it's often easier to leave it a s a series of adjacent rectangles, with some common interiour boundaries. Then you can easier handle the meshing sequences.
--
Lykke til
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hei Fjodor
Well why should you mesh the film in it thickness if it is really so "thin" ?
Do you really want to analyse what is happening in the thickness of the film ?
If not, try the thin film boundary layer BC, see the doc
The true difficulty is when you have more than 1 thin film stacked on the same boundary (currently, from my knowledge, not implemented in COMSOL)
Then when you have "simple" square/rectangular geometries, it's often easier to leave it a s a series of adjacent rectangles, with some common interiour boundaries. Then you can easier handle the meshing sequences.
--
Lykke til
Ivar
Thanks for the reply!
I dont actualy need to solve for what's going on inside the electrode - just the capacitance and how capacitance will be effected by placing things on the sensor.
I've checked the COMSOL help search and the doc folder - didn't find much. The knowledge base example 902 was some help but Im quite new at COMSOL and can't find how to make a composite object from two geometries...
Would this work if I just make the substrate a composite object with 2D electrodes on top of it covered by the PTFE film(this film serves as a dielectric)? If so, how can I create a composite object in COMSOL 4.0?
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
you have the model library thin_film_resistance model
So I assume, from the understanding of your comments, that you could probably get away with a 2 laer structure (+ surrounding air/vacuum) of bulk material below, a surface/boundary cutout region for the thin metallic film, and some thin domain/volume on top for the PTFE
--
Good luck
Ivar
you have the model library thin_film_resistance model
So I assume, from the understanding of your comments, that you could probably get away with a 2 laer structure (+ surrounding air/vacuum) of bulk material below, a surface/boundary cutout region for the thin metallic film, and some thin domain/volume on top for the PTFE
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
you have the model library thin_film_resistance model
So I assume, from the understanding of your comments, that you could probably get away with a 2 laer structure (+ surrounding air/vacuum) of bulk material below, a surface/boundary cutout region for the thin metallic film, and some thin domain/volume on top for the PTFE
--
Good luck
Ivar
Yes, the structure is basicaly as you describe: substrate with faces/boundaries to represent the thin film electrodes with a layer of PTFE on top, surrounded by air.
To find that model...
Regards
- Fjodor
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
its in the model library, but you should have other similar ones on the main COMSOL web site too, try a search
--
Good luck
Ivar
its in the model library, but you should have other similar ones on the main COMSOL web site too, try a search
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar
I have a similar problem.
have to mesh a thin layer and ended up with same problems.But I have to mesh it for the study of process change.
I'm doing it for mass transport analysis in PEM fuel cell- thin layers are the electrode films and catalytic layer.Please help
I have a similar problem.
have to mesh a thin layer and ended up with same problems.But I have to mesh it for the study of process change.
I'm doing it for mass transport analysis in PEM fuel cell- thin layers are the electrode films and catalytic layer.Please help
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
thin layers are tricky to mesh, to avoid that you explode the number of elements (and your RAM need) and in the same time to be able to resolve correctly your FEM model.
What I do, when analyising my mesh, is to consider the meshing as a time signal sampling. if you want to listen to some good classical music you should resolve your audio frequencies up to 16-20 kHz, hence with Nyquist use a 32 to 40 k sampling, at least, 2-3 times higher the better.
For FEM look at the gradients of your BC values dependent variables and the expected results, do your T change over a few mm or a few nm ? if so have a dense enough mesh in these regions to correctly catch the gradient of your dependent value (it means that you have some clue of the results, or that you go about it in a few iterative steps, correcting your first solved model based on the (partially wrong) results)
Mostly I mesh these thin regions manually, if they are "square" you select one surface and sweep through the thickness, with a dimensions node to define the distribution. You can get correct results with very uneven mesh (large in x,y thin in Z, provided that your dependent variables respect their gradient slope correspondingly. Sometimes you need to cut your thin domainjs into several parts, as the seep mesh have some limitations on the type of source - destination boundaries that must match well
--
Good luck
Ivar
thin layers are tricky to mesh, to avoid that you explode the number of elements (and your RAM need) and in the same time to be able to resolve correctly your FEM model.
What I do, when analyising my mesh, is to consider the meshing as a time signal sampling. if you want to listen to some good classical music you should resolve your audio frequencies up to 16-20 kHz, hence with Nyquist use a 32 to 40 k sampling, at least, 2-3 times higher the better.
For FEM look at the gradients of your BC values dependent variables and the expected results, do your T change over a few mm or a few nm ? if so have a dense enough mesh in these regions to correctly catch the gradient of your dependent value (it means that you have some clue of the results, or that you go about it in a few iterative steps, correcting your first solved model based on the (partially wrong) results)
Mostly I mesh these thin regions manually, if they are "square" you select one surface and sweep through the thickness, with a dimensions node to define the distribution. You can get correct results with very uneven mesh (large in x,y thin in Z, provided that your dependent variables respect their gradient slope correspondingly. Sometimes you need to cut your thin domainjs into several parts, as the seep mesh have some limitations on the type of source - destination boundaries that must match well
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar,I have also tried with "swept" as you have suggested in the thread.But I see no error while meshing but found those regions with warning "unmeshed" after running the case.please advise.
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Thanks Ivar..I will try and get back to you..