strongly guided fiber?

The long answer:

In any lossless optical system, the product of the area and projected solid angle of any single mode is always lambda**2/2. Stronger guiding equals a wider angular acceptance, which increases the projected solid angle. To keep a fibre single-mode with stronger guiding, you have to decrease the core diameter. This has the following disadvantages:

1. Tighter mechanical tolerances--if you halve the diameter, you have to align twice as accurately.

1. Higher losses at splices--you can't readily make the cladding index very much lower than that of fused silica (~1.46), so you have to increase the core index, which leads to higher reflection losses.

2. Higher losses due to microbending--roughness at the core-cladding boundary causes much higher scattering when the index discontinuity is large and the field strength is high, as in small-diameter, high-index fibre.

All of these things increase the cost of transmission systems, which is why very weak guiding is used.

Strong guiding is good for lots of things too:

1. More light per unit area--in a multimode fibre, you can pack more modes (and hence more light from an incoherent source) into the same fibre diameter with stronger guiding.

1. Lower bending loss--stronger guiding means you can turn tighter corners without losing light. Silicon-core waveguides used in optical ICs (at 1.55 micron wavelength) are about 0.45 microns wide and 0.2 microns tall, and you can make a 2-um radius bend with low loss. For fibre it isn't usually that big a deal, because bend radii are also limited by mean time to fracture (especially if the fibre can become wet).

2. Stronger interaction with perturbations, e.g. in sensors. This is the flip side of the increased microbend loss. Sometimes you _want_ increased sensitivity to what's going on at the surface.

The modelling thing can be fixed up to work with strongish guiding--though it gets much harder with silicon guides.

Cheers,

Phil Hobbs

Strongly guided fibre is good for multimode, because you can get more light into the fibre--more modes for a given diameter.

Phil Hobbs wrote: >

Minor correction: the invariant is n**2*A*Omega-prime. Forgot the n-squared.

Cheers,

Phil Hobbs

Thanks for a great explanation.

I have one theoretical question then. If you could dope the clad with Fluorine or something that could drop the index a bunch, keep the core the same size...you could have a strongly guided fiber without most of the problems you mentioned though right?

Thanks,

Pat.

Phil Hobbs wrote:

In article , Phil Hobbs responded to the query:

Phil,

Don't quarrel with anything you say -- but much of it is focused on equating "strongly guided" more or less to large V number (large delta-N times diameter product), rather than just to a large index step delta-N, period -- which could be an alternative interpretation of "strongly guided".

If you make this latter interpretation, would not some of the most basic consequences of increasingly strong guiding then include:

• Breakdown, to an increasing extent, of scalar wave function approximation and analysis; increasing need for more complicated vector mode solutions.

• Modes become increasing less "quasi TEM" in character.

• Increasing requirement for analysing in terms of radially and azimuthally symmetric modes rather than x and y polarized modes.

--AES

Tony,

Large V-numbers don't go with single-mode propagation, though...good SM guiding wants V-numbers of, iirc, about 1.8 to 2.40...the guide becomes multimode when V=first null of J0 ~ 2.405.

I was assuming that the OP was interested primarily in glass fibres, with NAs below about 0.2. The higher NA ones, up to about 0.4 NA, are (afaik) always multimode, so nobody expects to know what the fields are doing really.

The whole linearly-polarized scalar mode approximation goes south pretty fast when you have bigger index differences, it's true--I'm using guides with NAs greater than 2! I usually compute mode patterns by running a Gaussian beam into one end of a guide with a cladding whose absorption increases with radial distance, and then letting it rattle around numerically a few times to get the leaky modes stripped off. I really should write a proper mode solver, but ordinary FDTD does a pretty good job in an hour's run-time, which makes the time investment hard to justify.

This winter I finally got my optimizing FDTD code running on a 14-CPU Opteron cluster. The one main remaining problem is that there's a bug in the metal support, which makes it give The Wrong Answer. Ordinary FDTD assumes that the epsilon and mu of the materials are wavelength-independent, and so becomes unstable when n

The lowest-index material that isn't water soluble is magnesium fluoride, with N=1.38 or so. Thus you can't really go that low by this method. You'd win a little bit that way, since the microbending loss wouldn't be quite as bad, due to the larger diameter that this method could afford. The index contrast at the core-cladding boundary would still be large, though, so the scattering would still be strong.

Cheers,

Phil Hobbs