Hi! I am trying to apply whole this approach onto 3D design of Si resonator with AlN 2um thickness on top of it. Soon as I go to "Show Default Solver" I get error of missing relative permitivity of Silicon. Do you know what could possibly be a problem ? Best
I guess you included Si into piezoelectric domain that requires electric properties. If so, you can create a linear elastic domain and select the Si part.
Dear Friend, I am working on same module and concept in COMSOL using surface and lamb waves in 3D. It is mandatory for surface waves that on four vertical sides bloch condition must be applied, traction-free condition should be imposed at top free boundary and perfectly matched layer should be applied at bottom surface. For Lamb waves just make the bottom surface traction-free instead of PML. I know how to apply bloch condition on four vertical sides however I did not find the way, how can I apply traction-free condition at top and Perfectly matched layer at bottom layer. When I enable PML inside comsol, it selects all domain, but for surface waves it is necessary only bottom layer should be PML. If you could please help me in this regard. Thank you for your support.
PML usually doesn't work well in eigenfrequency analysis. So my suggestion is to build a thick unitcell model (thicker than several times your interesting wavelength). On the top it is free by default, at the bottom you can set it to be fixed, free or symmetric, it doesn't matter since your model is thick enough. And you should get the surface wave modes on the top surface, as they evolve from Lamb modes as k increases. However they are hidden among numerous Lamb modes and hard to track. There's a way to filter it out by a method you can find it in this paper: aip.scitation.org/doi/full/10.1063/1.4902018 Good luck!
Thank you for your response. What about traction-free condition at the top surface, is it by default ? According to your view, what I understand is, PML cannot be applied in eigenfrequency module, I need to make my unit cell more thicker with top free boundary and bottom fixed, free or symmetric. What I read in this paper ( iopscience.iop.org/article/10.1088/1367-2630/18/8/083041) page 5, below table, line 3,4,5 they have plot dispersion graphs by applying my stated conditions on their unit cell. My goal is produce same result which is present inside the paper. I have made similar geometry, assigned same material properties but where i stuck is boundary condition. If you could please check the mentioned lines and dispersion graphs, and help me in this case. Thank You for your support
Yes the free boundary condition is by default in COMSOL if you don't assign another. You can definitely try PML at the bottom, for surface waves. it's only my experience that PML only works well in frequency response analysis. You can check the COMSOL manual on how to apply PML. There's another similar paper on surface waves in phononic structures you might find it useful: www.researchgate.net/publication/262231698_Evidence_of_a_Love_wave_bandgap_in_a_quartz_substrate_coated_with_a_phononic_thin_layer
Thank you for your help. I applied required BC, but dispersion graph which I get is straight lines corresponding to certain frequencies. Before applying this conditions, when I was simulating some other models, consecutively I was getting dispersion curve as full straight lines with gap in b/w. If you could tell me the reason why is it so. Thank you for your guidance
You choosed a very long edge as thickness and a short edge as length, which is a little confusing. It causes me think that you choosed wrong direction to compute wavenumber. But after all, this is the right way to compute dispersion curves. Leave the thickness direction wavenumber as 0.
I kind of figured out why use a long edge thickness, while a very short edge as wave direction. Because otherwise you won't be able to reach long along k axis. Is that the reason?
Hui Zhang yes the dispersion curves fold back and forth at k=0, pi/a so it'd be better to make "a" smaller to prevent too many folded curves. And also we only need good resolution along the thickness direction anyway…
Sure, I agree. Just one more question, have you tried to use 3D model to compute Lamb wave dispersion in COMSOL? I tried, but it couldn't give what I want well. Also, I tried to compute phase velocity from the wavenumber curve, it doesn't give correct answers as well. Most probably because in COMSOL, the colored lines in w-k diagram do not represent same modes. What do you think?
I tried 3D case, both work well, match theoretical Lamb wave dispersions. In 3D you will have additional shear horizontal modes. Be sure also to set periodic BC on the other pair of boundaries, although you are going to set ky =0. If you let it be free, it becomes a beam not a plate.
I tried 3D case for a periodic structure, it works well. For Lamb wave modeling, depending on how you choose the periodic block, it will give good and bad results. For example, if you choose a quite small area, with the length 0.1 of the thickness, it can give good results. I am not sure whether you computed the phase velocity from this or not. How did you plot the phase velocity then?
Will you please explain why are you using Floquet Boundary conditions?
Is this applicable for multilayer immersed in water
Hi! I am trying to apply whole this approach onto 3D design of Si resonator with AlN 2um thickness on top of it. Soon as I go to "Show Default Solver" I get error of missing relative permitivity of Silicon. Do you know what could possibly be a problem ?
Best
I am asking because it took more than an hour to do 1% of computation
I guess you included Si into piezoelectric domain that requires electric properties.
If so, you can create a linear elastic domain and select the Si part.
你好 我现在开始学comsol，想建立一个超声波探伤模型， 不知道是否也是像你一样建立一个solid mechanic 的模型再加一个周期性的力，还是用其他的模型？谢谢
這方法只能計算超聲波在週期性結構(或均勻介質)傳遞的特徵模態。如過要看超聲波遇到裂縫如何反射，可以用頻率響應模組給一個波源，搭配PML(防止模型的邊界反射超聲波)，或是在時域分析模組給一個隨時間變化的外力，看一個波包在固體中隨時間如何傳遞
谢了 我回头试一试
你好 我试了试没有成功，我用的是comsol4。4版本，comsol blog上面的教程基本上是5。2的。我想建立一个用超声波探伤的模型，用time dependent作为study，不知道具体选择哪个physics，想向你咨询一下
請問如果是Pwave,Swave或RayleighWave,要如何修改呢?
为什么我根据视频做出来的特征频率都偏大啊
Is this applicable to 3D solid model?
Yes.
yes it is
为什么我根据视频做出来的特征频率偏大呢
When the material depends on the wavelength ?
Can the dispersion curve be indentify by modal indentification(e.g. S0, A0 wave)?
I also tried the 3D model and didn't work. Has anyone managed to solve it? Any hints?
你好，我现在想通过comsol来calculate the lamb wave dispersion,看了你的视频，很有收获，不过你有matlab相关的代码，我想从理论上与comsol的想比较，看模拟的是不是正确的，能给我的话，不胜感激啊，studentsman@126.com
我也是需要这些，您有代码吗
感謝您的回覆
另外還有2個問題想請教您
a的變化量 對linbo3 128度y的影響 因為a值改變 波長λ也會改變 想請問這樣對嗎
頻散曲線圖 頻率取很多的話 折線會過於複雜難整理 想請教這方面有能改善的方法嗎
波長λ=2π/k 只跟k的值有關
a 只是決定曲線在哪裡反摺 (在k=π/a)
所以 a 小一點可以延後反摺 還是無法徹底避免
Dear Friend,
I am working on same module and concept in COMSOL using surface and lamb waves in 3D. It is mandatory for surface waves that on four vertical sides bloch condition must be applied, traction-free condition should be imposed at top free boundary and perfectly matched layer should be applied at bottom surface.
For Lamb waves just make the bottom surface traction-free instead of PML.
I know how to apply bloch condition on four vertical sides however I did not find the way, how can I apply traction-free condition at top and Perfectly matched layer at bottom layer. When I enable PML inside comsol, it selects all domain, but for surface waves it is necessary only bottom layer should be PML.
If you could please help me in this regard.
Thank you for your support.
PML usually doesn't work well in eigenfrequency analysis. So my suggestion is to build a thick unitcell model (thicker than several times your interesting wavelength). On the top it is free by default, at the bottom you can set it to be fixed, free or symmetric, it doesn't matter since your model is thick enough. And you should get the surface wave modes on the top surface, as they evolve from Lamb modes as k increases. However they are hidden among numerous Lamb modes and hard to track. There's a way to filter it out by a method you can find it in this paper: aip.scitation.org/doi/full/10.1063/1.4902018
Good luck!
Thank you for your response.
What about traction-free condition at the top surface, is it by default ? According to your view, what I understand is, PML cannot be applied in eigenfrequency module, I need to make my unit cell more thicker with top free boundary and bottom fixed, free or symmetric. What I read in this paper ( iopscience.iop.org/article/10.1088/1367-2630/18/8/083041) page 5, below table, line 3,4,5 they have plot dispersion graphs by applying my stated conditions on their unit cell. My goal is produce same result which is present inside the paper. I have made similar geometry, assigned same material properties but where i stuck is boundary condition.
If you could please check the mentioned lines and dispersion graphs, and help me in this case.
Thank You for your support
Yes the free boundary condition is by default in COMSOL if you don't assign another. You can definitely try PML at the bottom, for surface waves. it's only my experience that PML only works well in frequency response analysis. You can check the COMSOL manual on how to apply PML.
There's another similar paper on surface waves in phononic structures you might find it useful:
www.researchgate.net/publication/262231698_Evidence_of_a_Love_wave_bandgap_in_a_quartz_substrate_coated_with_a_phononic_thin_layer
Thank you for your help. I applied required BC, but dispersion graph which I get is straight lines corresponding to certain frequencies. Before applying this conditions, when I was simulating some other models, consecutively I was getting dispersion curve as full straight lines with gap in b/w. If you could tell me the reason why is it so.
Thank you for your guidance
You choosed a very long edge as thickness and a short edge as length, which is a little confusing. It causes me think that you choosed wrong direction to compute wavenumber. But after all, this is the right way to compute dispersion curves. Leave the thickness direction wavenumber as 0.
I kind of figured out why use a long edge thickness, while a very short edge as wave direction. Because otherwise you won't be able to reach long along k axis. Is that the reason?
Hui Zhang yes the dispersion curves fold back and forth at k=0, pi/a so it'd be better to make "a" smaller to prevent too many folded curves. And also we only need good resolution along the thickness direction anyway…
Sure, I agree. Just one more question, have you tried to use 3D model to compute Lamb wave dispersion in COMSOL? I tried, but it couldn't give what I want well. Also, I tried to compute phase velocity from the wavenumber curve, it doesn't give correct answers as well. Most probably because in COMSOL, the colored lines in w-k diagram do not represent same modes. What do you think?
I tried 3D case, both work well, match theoretical Lamb wave dispersions. In 3D you will have additional shear horizontal modes.
Be sure also to set periodic BC on the other pair of boundaries, although you are going to set ky =0.
If you let it be free, it becomes a beam not a plate.
I tried 3D case for a periodic structure, it works well. For Lamb wave modeling, depending on how you choose the periodic block, it will give good and bad results. For example, if you choose a quite small area, with the length 0.1 of the thickness, it can give good results. I am not sure whether you computed the phase velocity from this or not. How did you plot the phase velocity then?
How do I implement this to calculate dispersion relation for 3D blocks?
just apply periodic BC on both pairs of boundaries
劉庭瑋 i tried that but dispersion curves are just vertical lines,seems inaccurate
create different nodes of periodic BCs for each pairs of boundaries
劉庭瑋 tried that as well and same results,could you please check with simple block of aluminum and share the file or result.