Learning Forum

[Kian Goeloe]

Hi,

For my project, I trying to simulate the geometric phase of a Si nanoblock unit cell on top of a glass substrate. 

To be more specific:

I have a Si nanoblock with fixed dimensions of 145, 105 and 320 nm in length, width and height, respectively. The nanoblock is in the middle of a glass substrate with dimensions of 420 nm in both length and width. The whole structure is orientated with the normal in the x-direction. As a source, I want to use circular polarized light which is propagating in the x-direction with a wavelength of 630 nm. 

As the light propagates through the unit cell in transmittance (let's say for example left circular polarized light), the polarization changes to the opposite handedness (which in this example case will be right circular polarized light). This induces a phase to the light, which is dependent on the angle of the nanoblock (top view). Hence, I want to simulate a look-up-table where the phase (of the opposite polarization) of the light is plotted against the angle of the nanoblock. 

So in general, I try to simulate figure 1b from Wang, Bo et. al. Nano letters, 2016, 16.8: 5235-5240. 

I tried the S-parameters object library and the Polarization ellipse object library, but I'm questioning my simulation results. Can anyone help me out? Thanks in advance!! 

[Guilin Sun]

Without screenshots, I can only guess.

1: your structure. "a glass substrate with dimensions of 420 nm in both length and width". substrate is cosidered infinite in the periodic directions, thus, it should be larger enough and be outside of the periodic BCs.  Did you use periodic Bs in yz? and the substrate larger than periodicity in yz?

2: It seems you only need one wavelength, did you use a pulse or a cw source? if you want the frequency domain result, please use the pulse source. You can use a relatively narrow bandwidth but the pulse will be long. You can use a pulse centered at the single wavelength, and set the monitor to record only this wavelength. To avoid possible issue, you can also have a second monitor to record broadband, and see the spectrum is smooth, and this wavelengh has resoance or not. it can give you some hints, for example if the simulation time is long enough.

3: for the geometric phase, since the polarization will change, maybe you can simply monitor the EyEz phase change (and their magnitudes) and do some post processing. 

As a general tip in the forum, an image is worth more.

 

[Kian Goeloe]

Hi Guilin Sun,

Thank you for your quick reply. I really appreciate it!

I've added a picture of how the current structure looks like. 

To respond to your points: 

1. I do use periodic boundary conditions in y and z and PML in x. The substrate itself is larger than the simulation domain (to be specific, the simulation domain is 420x420 nm and the substrate is 450x450 nm). I'm not sure if it is large enough in this case. What do you recommend?
2. Yes, I'm only interested in one wavelength. Thanks for the tip. What I did is set the source to a single center wavelength of 630 nm with zero wavelength span. Moreover, I've unchecked the "optimize for short pulse" checkbox. 
3. In this case, is it preferable to simulate with linear polarized light separately and in the post processing combine the results? After following several examples I found on Ansys and Ansys forum, I believed that this was the way to do it. Also one of the reason why I thought that the S parameter object library would be helpful. Right now, I'm combining both sources to obtain circular polarized light (for the record, did not do that when I tried the S parameters object library. There I simulated two object libraries for y and z polarized light and tried to combine the result in post processing). 

And also thanks for the forum tip, I'm new at this, and eager to learn. If you need more information, please do not hesitate to reply. Thanks in advance! 

[Guilin Sun]

A1: I personally recommend much larger, say 1um by 1um. But your seems ok.

A2: you can do it. But I would not strongly recomend it, since a pulse is faster and you can get the single frequency result. And since the shorter wavelength in the source is smaller than 630nm, the mesh would be finer, so more accurate. With faster simulation and more accurate result, an intituitive choice is pulse. You can set your monitor to record this wavelength only.

A3: You can use two analysis group. But with more familiar with script, you can use one such analysis group, with two linearly polarized sources, and pick up both Ey and Ez at the same time. I think this way may be better: your device has polarization effect, eg, a Ey incident light can induce both Ey and Ez. In experiment, you use circularly polarized source, not linearly polarized light, right? using one analysis group with two polarized source, it counts all the contributions.What you think?