I'm trying to model a tapered GRIN fiber of more than 15cm. Using beam envelope methods with Comsol requires a computational infrastructure I do not have at this moment. Also, using BPM (form Rsoft for example) with a decent resolution takes more than a month for each sample, which makes it non-practical. So I was wondering if there is a method with lower computational cost that can reproduce the effects of a GRIN fiber being tapered down up to 3um (please note that at these diameters the graded core is no-longer effective and the fiber may behave as a no-core fiber)?
Hello Fellipe,
Unless you use a very dense grid (for some reason, unless Im wrong?) BPM can (generally) skip many wavelength between successive longitudinal sections and still yield accurate results. A ~15cm long piece should be computed within several minutes. However if you use a dx q dy of, say 0.02 microns and dz < 1micron, a very simple scenario starts to consume very long time and memory. Can you specify the grid and what do you simulate? is that a propagation of specific beam? what is the beam size and what is the fiber's data (diameters / core NA at the center / materials induices.... It is hard to answer without knowing all these...try to specify if possible.
Good luck.
Hi Yariv,
first of all, thank you very much for your interest. I have a tapered GRIN fiber that is spliced in between two singlemode fibers, so the input beam is a LP01. The geometry of the tapered GRIN fiber is as depicted in the attached picture where the SMF cladding has a white clad and the GRIN fiber has a gray cladding. Lt is the length of the taper (around 75mm) and Lg is the length of the non-tapered GRIN fiber (around 15cm, each side).
Please note at the outside I have an external refractive index that is usually around 1.3RIU.
About the grid: I'm using the grid suggested by Rsoft's BPM. That is:
dx = 0.04micron
dy = 0.04micron
dz = 0.1micron
So its similar to the grid you imagined. I did tried to do with a coarser grid, but I was not very confident since I'm not very familiar with BPM.
Thanks again!
Fellipe, from my experiance, this is a very very very dendse grig and I beleive this what causes your computer to perform so slowly. It is close to what I mentioned as a too dense grid. If your LP01 is in the vicinity of 9 micron (for example SMF-28 at 1550nm) then it is enough thay\t you sample your field with ~10 points, i.e. use dx = dy = 1micron of half of that. dz can be eazily taken as 5 to 10 microns and still give very accurate results. Start anywhara around these area and see how much it is faster. Also if not necessary, use scalar computation. If you need polarization effects use scalar first to get an impresion of the semi or full polarized light will give. I guess the BPM suggest such dense grid since historically tit was used for integrated optics q planar guides, where one needed to follow very sharp turns or so.. Hope this helped.
Yariv,
I did tried using the grid you suggested, by the results where far different from those obtained with better resolution. Anyway, I have seen I can find a resolution that is in between those two that can improve the velocity of the results. So far I'm using dx=dy=dz=0.1micron, although it is not as fast as I need (4 days for each run) I believe I can start working with that.
Thanks for your help.
- Badges
- Science topic