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 The information provided may be out of date.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.

Continuing Modeling Issues with 2D & 3D piezoelectric SAW model

Please login with a confirmed email address before reporting spam

Hello all,
I’m in desperate need of help with my model. The initial model setup is a 2D cross-sectional SAW device where I have silicon, a metallic electrode as my ground, a buffer layer, a piezoelectric layer (PZT-5), and gold electrodes. The current physics I have defined is with piezoelectric devices (with COMSOL version 4.2.0) with the silicon and buffer/protective layer (SrTiO3) as my linear elastic material model, the metallic electrodes and metallic layer as my electrical material layer, and piezoelectric material model with the PZT. I have defined parameters most of the materials in terms of elastic matrix D, Young’s Modulus, Bulk modulus, permittivity, and so forth. My main issue is when I run a frequency domain analysis; I consistently get the error of:
Failed to find a solution for the initial parameter. Singular matrix. There are 3906 void equations (empty rows in matrix) for…etc.
I have looked through the discussion forum consistently to see why I continue to get the errors I received and I thought it was simply a matter of fixing my boundary conditions but I thought I had accounted for all constraints for this simplified model. I do not know at this point where I have been going wrong for the past couple of months now. I’m attaching the current model. Any suggestions or comments would be greatly appreciated. Thank you.


--
Sarah


5 Replies Last Post 22 févr. 2013, 11:08 UTC−5
Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 8 févr. 2013, 13:30 UTC−5
Sarah,

Not sure if I understand what you are trying to achieve but below are some comments on your model:

- There are no mechanical constraints, so your system can move as a rigid body
- Model has two disconnected regions from structural mechanics point of view
- Are you sure you want to solve plane stress but not plane strain problem?

Regards,
Sergei
Sarah, Not sure if I understand what you are trying to achieve but below are some comments on your model: - There are no mechanical constraints, so your system can move as a rigid body - Model has two disconnected regions from structural mechanics point of view - Are you sure you want to solve plane stress but not plane strain problem? Regards, Sergei

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 12 févr. 2013, 09:42 UTC−5
Hello Sergei,

Thank you for your reply. I still have a few questions based on your comments about the model. But first ultimately what I wanted to do was first obtain the resonant frequencies of the SAW device, then study the changes and shifts in the device as I added a sensing layer to the center of the device and then when the sensing thin film material interacts with a gas. I wanted to be able to obtain results ultimately with my 3D model but I have scaled down to a 2D model to simplify the device and figure out where the issues I have had. I originally used an eigenfrequency study and obtained some various surface and potential plots at various frequencies I designated. Then for later models, I proceeded to add a frequency domain analysis.
From your comment, I thought that within the piezoelectric devices interfaces, that I had a mechanical constraint with setting a fixed constraint for my bottom layer (silicon) to prevent the device from tilting. I thought that was the main mechanical constraint I needed. In addition, I don’t understand how I still have disconnected regions from a structural mechanics points of view. How can I check this within the PZD physics interface and see this disconnect? In addition, I have been going back and forward trying to see if I need to be solving this as a plane strain problem vs. a plane stress figuring if this is having some effect on my model.

So I’m not sure if there’s also an effect of parameters that I used for the eignefrequency analysis is affecting the ability to do the frequency domain analysis and I did not fully remove these parameters or if there is something really wrong with the boundary conditions I have set since I continued to receive the failure to find an initial parameter and a singular matrix error even though I thought I had accounted for the BC for the model.

Any suggestions or comments would be helpful to the above would be greatly appreciated. Thank you. I’m also attaching some earlier eigenfrequency plots that I obtained at various frequencies too.


--
Sarah
Hello Sergei, Thank you for your reply. I still have a few questions based on your comments about the model. But first ultimately what I wanted to do was first obtain the resonant frequencies of the SAW device, then study the changes and shifts in the device as I added a sensing layer to the center of the device and then when the sensing thin film material interacts with a gas. I wanted to be able to obtain results ultimately with my 3D model but I have scaled down to a 2D model to simplify the device and figure out where the issues I have had. I originally used an eigenfrequency study and obtained some various surface and potential plots at various frequencies I designated. Then for later models, I proceeded to add a frequency domain analysis. From your comment, I thought that within the piezoelectric devices interfaces, that I had a mechanical constraint with setting a fixed constraint for my bottom layer (silicon) to prevent the device from tilting. I thought that was the main mechanical constraint I needed. In addition, I don’t understand how I still have disconnected regions from a structural mechanics points of view. How can I check this within the PZD physics interface and see this disconnect? In addition, I have been going back and forward trying to see if I need to be solving this as a plane strain problem vs. a plane stress figuring if this is having some effect on my model. So I’m not sure if there’s also an effect of parameters that I used for the eignefrequency analysis is affecting the ability to do the frequency domain analysis and I did not fully remove these parameters or if there is something really wrong with the boundary conditions I have set since I continued to receive the failure to find an initial parameter and a singular matrix error even though I thought I had accounted for the BC for the model. Any suggestions or comments would be helpful to the above would be greatly appreciated. Thank you. I’m also attaching some earlier eigenfrequency plots that I obtained at various frequencies too. -- Sarah


Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 13 févr. 2013, 10:27 UTC−5
Sarah,

In total, you have 14 domains. Domain #4 is active in the Piezoelectric Material Model, domains #1 and 2 are active in the Linear Elastic Model 1, and the rest domains are active in the Electrical Material Model. It means that you are not solving structural mechanics equations in the domains 2, 5-14.

If you plot displacement field, you will see that only domains 1, 3, and 4 are colored, while the rest domains are white, as shown in the attached Figure. You need add all “white” domains into Linear Elastic Model node.

You do not have any structural constraints in the (pzt) physics interface, only electrical constraints are imposed. Thus, rigid motion is not excluded.

I think that plane strain is more appropriate for you model.

Regards,
Sergei
Sarah, In total, you have 14 domains. Domain #4 is active in the Piezoelectric Material Model, domains #1 and 2 are active in the Linear Elastic Model 1, and the rest domains are active in the Electrical Material Model. It means that you are not solving structural mechanics equations in the domains 2, 5-14. If you plot displacement field, you will see that only domains 1, 3, and 4 are colored, while the rest domains are white, as shown in the attached Figure. You need add all “white” domains into Linear Elastic Model node. You do not have any structural constraints in the (pzt) physics interface, only electrical constraints are imposed. Thus, rigid motion is not excluded. I think that plane strain is more appropriate for you model. Regards, Sergei


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 févr. 2013, 16:49 UTC−5
Hello once again,

Once again I want to thank you for your suggestions and pointing out things with my model that I didn't really pay attention to and I have been steadily making improvements upon the suggestions and help with my 2D and subsequently 3D model. But it seems even with the changes I have made, when I run a frequency domain analysis I seem to still get these same commonly errors:

Failed to find a solution for all parameters,
even when using the minimum parameter step.
No convergence, even when using the minimum damping factor.
Returned solution is not converged

Or warning: inverted mesh elements near coordinates()…

I have been consistently going through the discussion forums to see how other people attempted to resolve these common issues and it seems to not working out for me. Once again, I would really appreciate any help and suggestions of what possibly I could be doing wrong.

Sarah



--
Sarah
Hello once again, Once again I want to thank you for your suggestions and pointing out things with my model that I didn't really pay attention to and I have been steadily making improvements upon the suggestions and help with my 2D and subsequently 3D model. But it seems even with the changes I have made, when I run a frequency domain analysis I seem to still get these same commonly errors: Failed to find a solution for all parameters, even when using the minimum parameter step. No convergence, even when using the minimum damping factor. Returned solution is not converged Or warning: inverted mesh elements near coordinates()… I have been consistently going through the discussion forums to see how other people attempted to resolve these common issues and it seems to not working out for me. Once again, I would really appreciate any help and suggestions of what possibly I could be doing wrong. Sarah -- Sarah


Sergei Yushanov Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22 févr. 2013, 11:08 UTC−5
Sarah,

I suspect that issues are related to physical formulation of the problem.
For example, what is the meaning of the antisymmetry boundary condition at the interior boundary? What is the meaning of the “Include geometrical nonlinearity” option in the Frequency Domain analysis? (Frequency Domain is linear analysis by definition)

It might be helpful to start with Comsol SAW Gas Sensor Model to perform Eigenfrequency first. Then, perform frequency sweep using Frequency Domain analysis to confirm that you are getting the same results in the sense that resonant frequencies coincide with eigenvalue solution and resonant field distribution is of the same shape as eigenfrequency mode. This might give you some insight and help to move gradually towards your actual model.

Regards,
Sergei
Sarah, I suspect that issues are related to physical formulation of the problem. For example, what is the meaning of the antisymmetry boundary condition at the interior boundary? What is the meaning of the “Include geometrical nonlinearity” option in the Frequency Domain analysis? (Frequency Domain is linear analysis by definition) It might be helpful to start with Comsol SAW Gas Sensor Model to perform Eigenfrequency first. Then, perform frequency sweep using Frequency Domain analysis to confirm that you are getting the same results in the sense that resonant frequencies coincide with eigenvalue solution and resonant field distribution is of the same shape as eigenfrequency mode. This might give you some insight and help to move gradually towards your actual model. Regards, Sergei

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.