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 The information provided may be out of date.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.

slab-waveguide simulation / scattering losses

Please login with a confirmed email address before reporting spam

Hello,

I'm attempting to simulate a simple slab waveguide (see attached). The waveguide has a core of n=1.5 surrounded by air (n=1) above and below. This design is surrounded by PML rectangles.
The boundary conditions and the size of the core are selected to be such that the electric field imposed on the three left boundaries (inside, to the right the PML, top to bottom) is a mode supported by the waveguide:

exp(-p0*y)
exp(-p0*d)*cos(h0*y)/cos(h0*d)
exp(p0*y)

The physics is based on the RF Module / photonic crystal model library in Comsol 3.5a.

I would like to know the core power input (left) and output (right). To this end, I have two questions:

1. I would like to take the core (middle subdomain), integrate the y-values and get a vector of the average power in each x value of the core. According to this vector I'd like to generate a graph of the decline of power from input to output as a function of propagation length x. Is this possible in Comsol?

2. I select the core input, and from Post Processing I choose "Power outflow, time average" (nPoav_rfwe), which seems a reasonable choice, but the results don't make sense to me (the input power -6.793381e-11 [W/m] is lower than the output -3.18875e-10 [W/m]; also, what do the signs mean? how can I simulate "time average" when the simulation is not time-dependent?). I would appreciate suggestions on how to measure the input and output power.

I'm attaching the model if anyone's interested.

Thanks in advance!


13 Replies Last Post 22.04.2013, 09:31 GMT-4
COMSOL Moderator

Hello Roy Shiloh

Your Discussion has gone 30 days without a reply. If you still need help with COMSOL and have an on-subscription license, please visit our Support Center for help.

If you do not hold an on-subscription license, you may find an answer in another Discussion or in the Knowledge Base.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.12.2010, 07:08 GMT-5
Hi roy, actually we can speak hebrew...

Anyway,few comments-

I have lower version, so i didn't open the mph file, but from the image i see that there are pml's only above and below- you should add ones also on the sides with pml cores and cladds (i.e. define the domains as pml's in x direction and suitable ref. index to each of the areas), in order to lower the reflactences.

Secondly, I would put another artificial Input and Output lines inside the structure, to avoid the effect of the pml.

For Question 1. you can use the "cross section plot paremeters" tool in the postprocessing tab. just choose the power and the coordinates of the line to see the changes in the power through the propagation.

For Q. 2.- what is the direction of the propagation? you can draw the phase through the device (according to the previous paragraph) to understand this, you might find that the power is opposite... it is also connected to the signs of the numbers. Abd for the meaning of the time average- it is a factor of 0.5 to the product of the field, since you assume in this model a sinusoidal behavior, which you need to take into account.

Hope I helped,

Avigdor

Hi roy, actually we can speak hebrew... Anyway,few comments- I have lower version, so i didn't open the mph file, but from the image i see that there are pml's only above and below- you should add ones also on the sides with pml cores and cladds (i.e. define the domains as pml's in x direction and suitable ref. index to each of the areas), in order to lower the reflactences. Secondly, I would put another artificial Input and Output lines inside the structure, to avoid the effect of the pml. For Question 1. you can use the "cross section plot paremeters" tool in the postprocessing tab. just choose the power and the coordinates of the line to see the changes in the power through the propagation. For Q. 2.- what is the direction of the propagation? you can draw the phase through the device (according to the previous paragraph) to understand this, you might find that the power is opposite... it is also connected to the signs of the numbers. Abd for the meaning of the time average- it is a factor of 0.5 to the product of the field, since you assume in this model a sinusoidal behavior, which you need to take into account. Hope I helped, Avigdor

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 12.04.2011, 09:50 GMT-4
Hello!

I have a very similar problem I do not know how to tackle. You seem to be more experienced than me in this regard, so I'd be more thank thankful for any hints.

The problem is the following. I would like to simulate the electric field (or the power) flow through a waveguide. (Typically a SOI-waveguide with 1.55um). For this purpose, I made a structure that resembles yours. In my case it's a 2D 'sandwitch' of air, Si and SiO2 in y-direction and the propagation direction chosen as x, however.

Then, I have defined two ports (1, 2) at x = 0 and x = L, which is still a rudimentary thing. However, here is where I am getting confused. In principle, I would like to excite the wave as TE10 (or TM01, or whatever other mode), however I do not know how, literally, I can 'buid up' this mode profiles at the entrance port 1. Even more, instead of building it, I would like COMSOL to do it, for example, by using the effective index method.

I do not know how I can accomplish that. How I can couple these calculations or perhaps, if (and how) I can feed the study of the wave propagation with the correct filed distribution at the input.

Once again, I will appreciate any comments.
my best regards
Oleg

PS. I'm using the 4.1, RF.
Hello! I have a very similar problem I do not know how to tackle. You seem to be more experienced than me in this regard, so I'd be more thank thankful for any hints. The problem is the following. I would like to simulate the electric field (or the power) flow through a waveguide. (Typically a SOI-waveguide with 1.55um). For this purpose, I made a structure that resembles yours. In my case it's a 2D 'sandwitch' of air, Si and SiO2 in y-direction and the propagation direction chosen as x, however. Then, I have defined two ports (1, 2) at x = 0 and x = L, which is still a rudimentary thing. However, here is where I am getting confused. In principle, I would like to excite the wave as TE10 (or TM01, or whatever other mode), however I do not know how, literally, I can 'buid up' this mode profiles at the entrance port 1. Even more, instead of building it, I would like COMSOL to do it, for example, by using the effective index method. I do not know how I can accomplish that. How I can couple these calculations or perhaps, if (and how) I can feed the study of the wave propagation with the correct filed distribution at the input. Once again, I will appreciate any comments. my best regards Oleg PS. I'm using the 4.1, RF.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17.05.2011, 16:26 GMT-4

Oleg,

What you are asking to do (I think) is almost covered in this example:

www.comsol.com/showroom/documentation/model/140/

The only problem is that the example uses the fundamental mode of the input waveguide. However, some tweaking of the first part of the simulation may force it to launch the TE10 mode instead.

Another option that may work: I've found that launching a plane wave at the boundary of the input guide actually settles down to the fundamental mode very rapidly. Perhaps launching a plane wave at an appropriate angle to the waveguide axis would result in a TE10 mode.

Good Luck,
Brian
Oleg, What you are asking to do (I think) is almost covered in this example: http://www.comsol.com/showroom/documentation/model/140/ The only problem is that the example uses the fundamental mode of the input waveguide. However, some tweaking of the first part of the simulation may force it to launch the TE10 mode instead. Another option that may work: I've found that launching a plane wave at the boundary of the input guide actually settles down to the fundamental mode very rapidly. Perhaps launching a plane wave at an appropriate angle to the waveguide axis would result in a TE10 mode. Good Luck, Brian

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 03.10.2011, 03:10 GMT-4
Hi,

I have another similar question on simulating the waveguide propagating modes using COMSOL 4.2. Is it possible to do it with combination of PML and ports?

For a lossless rectangular straight waveguide, the resulting Electric Field Norm should be uniform along the waveguide.

Most tutorial examples, like circulator, h-bend waveguide, waveguide adapter, use the perfect electric conductor boundary condition on the side wall of the waveguide and port boundary condition on the facet to excite and absorb the waveguide mode (TE/TM/TEM). The results are correct.

However, when it comes to the dielectric waveguide, e.g. silicon with air cladding, things become difficult. The 2D geometry is like this: a straight silicon waveguide is sandwiched by air cladding and the whole domain is boxed by PML (many reflections if I only use PML on the sidewall, the Electric Field Norm is not uniform).

The difficulty comes:
1. How to define port? I notice that port can only be defined at the outer boundary...But there is PML!
2. Is it reasonable to insert silicon waveguide into PML?

A possible solution is to user-define the E-field at the inner boundary of the PML instead of using port. Then, How to calculate the S-parameter? And is it possible to directly import the E-field of waveguide eigenmodes?

Looking forward to all the replies and helps.

Best,
Lu


Hi, I have another similar question on simulating the waveguide propagating modes using COMSOL 4.2. Is it possible to do it with combination of PML and ports? For a lossless rectangular straight waveguide, the resulting Electric Field Norm should be uniform along the waveguide. Most tutorial examples, like circulator, h-bend waveguide, waveguide adapter, use the perfect electric conductor boundary condition on the side wall of the waveguide and port boundary condition on the facet to excite and absorb the waveguide mode (TE/TM/TEM). The results are correct. However, when it comes to the dielectric waveguide, e.g. silicon with air cladding, things become difficult. The 2D geometry is like this: a straight silicon waveguide is sandwiched by air cladding and the whole domain is boxed by PML (many reflections if I only use PML on the sidewall, the Electric Field Norm is not uniform). The difficulty comes: 1. How to define port? I notice that port can only be defined at the outer boundary...But there is PML! 2. Is it reasonable to insert silicon waveguide into PML? A possible solution is to user-define the E-field at the inner boundary of the PML instead of using port. Then, How to calculate the S-parameter? And is it possible to directly import the E-field of waveguide eigenmodes? Looking forward to all the replies and helps. Best, Lu

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 03.10.2011, 04:55 GMT-4
Hi,

I'm not sure what you're trying to do, but I think the problem is simpler than you think: you can set Port to an inner boundary (in version 3.5a at least) - try clicking "Interior Boundaries" in Boundary Settings.

Roy
Hi, I'm not sure what you're trying to do, but I think the problem is simpler than you think: you can set Port to an inner boundary (in version 3.5a at least) - try clicking "Interior Boundaries" in Boundary Settings. Roy

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 03.10.2011, 06:02 GMT-4

Hi,

I'm not sure what you're trying to do, but I think the problem is simpler than you think: you can set Port to an inner boundary (in version 3.5a at least) - try clicking "Interior Boundaries" in Boundary Settings.

Roy


Hi Roy, Thank you very much for your reply.
What I am trying to do is to see how the EM waves propagating in a straight waveguide. Although it is simple, but I am facing serious reflection problem in my model.
in principle, a guided mode has uniform intensity along the waveguide but my simulation gives interference pattern...
[QUOTE] Hi, I'm not sure what you're trying to do, but I think the problem is simpler than you think: you can set Port to an inner boundary (in version 3.5a at least) - try clicking "Interior Boundaries" in Boundary Settings. Roy [/QUOTE] Hi Roy, Thank you very much for your reply. What I am trying to do is to see how the EM waves propagating in a straight waveguide. Although it is simple, but I am facing serious reflection problem in my model. in principle, a guided mode has uniform intensity along the waveguide but my simulation gives interference pattern...

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 03.10.2011, 14:29 GMT-4
Hi,

The mode of a single-mode slab-waveguide is not uniform! You have to solve the slab-waveguide equations from which you will get a solution that looks like:

E(y,z) = A(y)*exp(-i*beta*z)

A(y) = {
exp(-p0*y) .. y>d
exp(-p0*d)*cos(h0*y)/cos(h0*d) .. -d<y<d
exp(p0*y) .. y<-d

where h0 and p0 are constants found from the equation to be

h0 = sqrt( n1^2 * k0^2 - beta^2 )
p0 = sqrt( beta^2 - n2^2 * k0^2 )

So you have to excite the above field as your input wave. Look it up.

If you already did that and you're saying you have reflections, I would guess your PMLs are not set right or you have a "perfect electric conductor" boundary set somewhere, which of course reflects EM waves..

Roy
Hi, The mode of a single-mode slab-waveguide is not uniform! You have to solve the slab-waveguide equations from which you will get a solution that looks like: E(y,z) = A(y)*exp(-i*beta*z) A(y) = { exp(-p0*y) .. y>d exp(-p0*d)*cos(h0*y)/cos(h0*d) .. -d

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 04.10.2011, 04:11 GMT-4
Hi,

I think I find a solution. Thank you any way for your suggestion. As I started from the example "waveguide adapter (3D)", I prefer to stick on it at the moment. Because COMSOL is able to solve the boundary mode, i. e. waveguide cross section in 2D, and use this solution as a numerical mode input for the port. So that I can avoid typing the E-field with propagating constant for different frequency. Also because that the analytical equation is only an approximation.

Here are some remaining questions for discussion.

My geometry is simple, a silicon waveguide (w=0.3um, t=0.34um, l=1um), covered by air, and based on SiO2, all surrounded by PML. (If you see the cross section, it is like a tiger roll.) Input and output facets (entire facet) are set as ports. See the attached model.

Solving the boundary mode (at freq=2e14) at ports 1 and 2, I have a similar effective index for TE (or TM) mode as in 2D simulation. The only strange thing is that Neff becomes real instead of complex. Then, the frequency domain study gives that, only at freq=2e14, the TE (or TM) mode is uniform along the waveguide. I understand that for each frequency there is an effective index due to dispersion. Therefore, it is necessary to have a frequency sweep to solve the mode for each frequency and then the propagation. Does anybody know how to do it?? The parametric sweeping is only for global parameters as I understand. Am I wrong?

Best,
Lu
Hi, I think I find a solution. Thank you any way for your suggestion. As I started from the example "waveguide adapter (3D)", I prefer to stick on it at the moment. Because COMSOL is able to solve the boundary mode, i. e. waveguide cross section in 2D, and use this solution as a numerical mode input for the port. So that I can avoid typing the E-field with propagating constant for different frequency. Also because that the analytical equation is only an approximation. Here are some remaining questions for discussion. My geometry is simple, a silicon waveguide (w=0.3um, t=0.34um, l=1um), covered by air, and based on SiO2, all surrounded by PML. (If you see the cross section, it is like a tiger roll.) Input and output facets (entire facet) are set as ports. See the attached model. Solving the boundary mode (at freq=2e14) at ports 1 and 2, I have a similar effective index for TE (or TM) mode as in 2D simulation. The only strange thing is that Neff becomes real instead of complex. Then, the frequency domain study gives that, only at freq=2e14, the TE (or TM) mode is uniform along the waveguide. I understand that for each frequency there is an effective index due to dispersion. Therefore, it is necessary to have a frequency sweep to solve the mode for each frequency and then the propagation. Does anybody know how to do it?? The parametric sweeping is only for global parameters as I understand. Am I wrong? Best, Lu


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 05.10.2011, 06:40 GMT-4
Hello again,

I have been away for a while, but now I am back.

It took me some time to solve the problem, but now I think I have it running correctly as I wanted.
I will list the things that I did below in points, so that it's easier to follow.

Let's assume we work with a SOI waveguide of the following structure: SiO2 (bottom), Si (middle), air (top) - our model is a SIDE VIEW of the wavegiude, where the light enters from the left and leaves at the right.

1. Surround the wavegiude with appx. 1-2[um] PML layer all around.
2. Break the future port boundaries, let's say 1[um] below and 1[um] above the core, where you can assume there is almost no field.
3. Get two Boundary Mode Analysis nodes and one Frequency Domain under study. Make sure the Boundaries are calculated first.
4. Make the ports at the inner circumference of your model (outer will have your power eaten due to neighbouring PMLs).
5. Each port should be composed of three segments. Set excitation to port 1. Set 'numeric' to both.
6. In the Boundary Mode Analysis nodes, set port 1 (and 2, respectively), the refractive index as you expect it to be (whether this will give you TE, TM or higher order modes you should know in advance), number of modes eq. to 1 and the frequency you are interested in. (Lu Ding is right, if you want to account for dispersion you need to make a frequency sweep).
7. Very inportant. Remove the domains that exist between the ports' segments and the outer boundaries of your model from the Electromagnetic Study. This will automatically break the PMLs just behind the ports. If you fail to do it, some part of your power will be absorbed behind the ports giving your S-parameters no meaning at all.
8. Send wishes that there will be no errors...

I think that's it. Ai least it worked in my case. Of course, you can extand this model to anything you want, but it should work whenever you want to excite the fields at the ends of a dielectric 'sandwitch'.

Good Luck
Hello again, I have been away for a while, but now I am back. It took me some time to solve the problem, but now I think I have it running correctly as I wanted. I will list the things that I did below in points, so that it's easier to follow. Let's assume we work with a SOI waveguide of the following structure: SiO2 (bottom), Si (middle), air (top) - our model is a SIDE VIEW of the wavegiude, where the light enters from the left and leaves at the right. 1. Surround the wavegiude with appx. 1-2[um] PML layer all around. 2. Break the future port boundaries, let's say 1[um] below and 1[um] above the core, where you can assume there is almost no field. 3. Get two Boundary Mode Analysis nodes and one Frequency Domain under study. Make sure the Boundaries are calculated first. 4. Make the ports at the inner circumference of your model (outer will have your power eaten due to neighbouring PMLs). 5. Each port should be composed of three segments. Set excitation to port 1. Set 'numeric' to both. 6. In the Boundary Mode Analysis nodes, set port 1 (and 2, respectively), the refractive index as you expect it to be (whether this will give you TE, TM or higher order modes you should know in advance), number of modes eq. to 1 and the frequency you are interested in. (Lu Ding is right, if you want to account for dispersion you need to make a frequency sweep). 7. Very inportant. Remove the domains that exist between the ports' segments and the outer boundaries of your model from the Electromagnetic Study. This will automatically break the PMLs just behind the ports. If you fail to do it, some part of your power will be absorbed behind the ports giving your S-parameters no meaning at all. 8. Send wishes that there will be no errors... I think that's it. Ai least it worked in my case. Of course, you can extand this model to anything you want, but it should work whenever you want to excite the fields at the ends of a dielectric 'sandwitch'. Good Luck

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 10.10.2011, 22:43 GMT-4
hello Oleg,

Thank you very much for sharing your model.

As the frequency domain simulation takes too much time, now I switch to time domain to solve this propagation wave in waveguide problem. Because I need to do the parametric sweep on frequency as well as on the structure parameters.

From your description, your model is built in 2D plane with cross section view, assuming the SOI waveguide as a infinite wide slab. And you use the effective index instead of the bulk index. Am I correct?

By just reading your tips, I am still a bit confused about where the PML and ports are. Would you mind to upload a simple geometry graph for me to understand better?

Best
hello Oleg, Thank you very much for sharing your model. As the frequency domain simulation takes too much time, now I switch to time domain to solve this propagation wave in waveguide problem. Because I need to do the parametric sweep on frequency as well as on the structure parameters. From your description, your model is built in 2D plane with cross section view, assuming the SOI waveguide as a infinite wide slab. And you use the effective index instead of the bulk index. Am I correct? By just reading your tips, I am still a bit confused about where the PML and ports are. Would you mind to upload a simple geometry graph for me to understand better? Best

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18.10.2011, 08:01 GMT-4
Hello,

Again, it took me too long. I am sorry.

Answering you rquestions:
n_air = 1.00
n_SiO2 = 1.44
n_Si = 3.47

you assign them to the materials.

Then, you use the 'Boundary Mode Analysis' nodes, both linked to the specific ports, to figure out what kinds of modes the ports should excite or absorb on the 'Frequency Domain' study. In general N ports require N 'Boundary M.A.' nodes. ports are recognised by their index (1,2,3,...) and modes by their effective index of refraction. I recommend you to allow only one mode. You do it by specifying number of solutions equal to one and a relevant quess on the expected index of refraction a priori.

The result will be that you get a (transverse) mode to propagate, having some effective index, that will be n_cladding < n_eff < n_core.

If you want to account for the dispersion, you should probably parametrise the material indices somehow. (I haven't ried yet). Or as you say, try to make a time domain simulation (I also haven't tried).

You can find and attached drawing of where to assign which property. It is important that you exclude the area from behind the ports from the studies.

Good luck!
Oleg
Hello, Again, it took me too long. I am sorry. Answering you rquestions: n_air = 1.00 n_SiO2 = 1.44 n_Si = 3.47 you assign them to the materials. Then, you use the 'Boundary Mode Analysis' nodes, both linked to the specific ports, to figure out what kinds of modes the ports should excite or absorb on the 'Frequency Domain' study. In general N ports require N 'Boundary M.A.' nodes. ports are recognised by their index (1,2,3,...) and modes by their effective index of refraction. I recommend you to allow only one mode. You do it by specifying number of solutions equal to one and a relevant quess on the expected index of refraction a priori. The result will be that you get a (transverse) mode to propagate, having some effective index, that will be n_cladding < n_eff < n_core. If you want to account for the dispersion, you should probably parametrise the material indices somehow. (I haven't ried yet). Or as you say, try to make a time domain simulation (I also haven't tried). You can find and attached drawing of where to assign which property. It is important that you exclude the area from behind the ports from the studies. Good luck! Oleg


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 10.11.2011, 21:34 GMT-5
.....me too... the wave iinside the waveguide is like a standing wave...espacially when I have one waveguide and a tapered waveguide, in 3D model....
ps: r u LXS's Lu?hahah
.....me too... the wave iinside the waveguide is like a standing wave...espacially when I have one waveguide and a tapered waveguide, in 3D model.... ps: r u LXS's Lu?hahah

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.04.2013, 09:31 GMT-4
Hello,

Fortunately, I came across your message in a discussion board about simulating a planer waveguide with the Silicon core and Silicon di-Oxide Clad. Thanks for coming up with the idea of dividing the port in three different sections. I had a couple of follow-up questions in that regard:

On page 58 of RF User Guide pdf file, it was noted that exciting more than one port would turn off the s parameter calculation. Were you still able to compute the s11_rfw, s12_rfw, etc.? If yes, how?

Is it possible, from your experience, that I already have the field, and I want to calculate the s parameters from the ports of the geometry (i.e. none of the port is generating the wave)? I can not find a way to find the S variables in this case.

Thanks


Hello again,

I have been away for a while, but now I am back.

It took me some time to solve the problem, but now I think I have it running correctly as I wanted.
I will list the things that I did below in points, so that it's easier to follow.

Let's assume we work with a SOI waveguide of the following structure: SiO2 (bottom), Si (middle), air (top) - our model is a SIDE VIEW of the wavegiude, where the light enters from the left and leaves at the right.

1. Surround the wavegiude with appx. 1-2[um] PML layer all around.
2. Break the future port boundaries, let's say 1[um] below and 1[um] above the core, where you can assume there is almost no field.
3. Get two Boundary Mode Analysis nodes and one Frequency Domain under study. Make sure the Boundaries are calculated first.
4. Make the ports at the inner circumference of your model (outer will have your power eaten due to neighbouring PMLs).
5. Each port should be composed of three segments. Set excitation to port 1. Set 'numeric' to both.
6. In the Boundary Mode Analysis nodes, set port 1 (and 2, respectively), the refractive index as you expect it to be (whether this will give you TE, TM or higher order modes you should know in advance), number of modes eq. to 1 and the frequency you are interested in. (Lu Ding is right, if you want to account for dispersion you need to make a frequency sweep).
7. Very inportant. Remove the domains that exist between the ports' segments and the outer boundaries of your model from the Electromagnetic Study. This will automatically break the PMLs just behind the ports. If you fail to do it, some part of your power will be absorbed behind the ports giving your S-parameters no meaning at all.
8. Send wishes that there will be no errors...

I think that's it. Ai least it worked in my case. Of course, you can extand this model to anything you want, but it should work whenever you want to excite the fields at the ends of a dielectric 'sandwitch'.

Good Luck


Hello, Fortunately, I came across your message in a discussion board about simulating a planer waveguide with the Silicon core and Silicon di-Oxide Clad. Thanks for coming up with the idea of dividing the port in three different sections. I had a couple of follow-up questions in that regard: On page 58 of RF User Guide pdf file, it was noted that exciting more than one port would turn off the s parameter calculation. Were you still able to compute the s11_rfw, s12_rfw, etc.? If yes, how? Is it possible, from your experience, that I already have the field, and I want to calculate the s parameters from the ports of the geometry (i.e. none of the port is generating the wave)? I can not find a way to find the S variables in this case. Thanks [QUOTE] Hello again, I have been away for a while, but now I am back. It took me some time to solve the problem, but now I think I have it running correctly as I wanted. I will list the things that I did below in points, so that it's easier to follow. Let's assume we work with a SOI waveguide of the following structure: SiO2 (bottom), Si (middle), air (top) - our model is a SIDE VIEW of the wavegiude, where the light enters from the left and leaves at the right. 1. Surround the wavegiude with appx. 1-2[um] PML layer all around. 2. Break the future port boundaries, let's say 1[um] below and 1[um] above the core, where you can assume there is almost no field. 3. Get two Boundary Mode Analysis nodes and one Frequency Domain under study. Make sure the Boundaries are calculated first. 4. Make the ports at the inner circumference of your model (outer will have your power eaten due to neighbouring PMLs). 5. Each port should be composed of three segments. Set excitation to port 1. Set 'numeric' to both. 6. In the Boundary Mode Analysis nodes, set port 1 (and 2, respectively), the refractive index as you expect it to be (whether this will give you TE, TM or higher order modes you should know in advance), number of modes eq. to 1 and the frequency you are interested in. (Lu Ding is right, if you want to account for dispersion you need to make a frequency sweep). 7. Very inportant. Remove the domains that exist between the ports' segments and the outer boundaries of your model from the Electromagnetic Study. This will automatically break the PMLs just behind the ports. If you fail to do it, some part of your power will be absorbed behind the ports giving your S-parameters no meaning at all. 8. Send wishes that there will be no errors... I think that's it. Ai least it worked in my case. Of course, you can extand this model to anything you want, but it should work whenever you want to excite the fields at the ends of a dielectric 'sandwitch'. Good Luck [/QUOTE]

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.