The fiber mode
A single-mode fiber guides one field pattern, its fundamental mode LP₀₁. For a step-index fiber with core radius a and numerical aperture NA, the mode depends on one number, V [1]:
The mode is a Bessel function J₀ in the core and a decaying K₀ in the cladding, whose parameters u and w follow from V [1]:
It is very nearly Gaussian [2], but not exactly. The Gaussian that couples to it best, of radius wm (the mode-field diameter is 2wm), couples 94.4 % at V = 1.2, 97.7 % at V = 1.6, 99.2 % at V = 2 and 99.7 % at V = 2.405. That is the most any TEM₀₀ beam can couple, through any lens, so the page couples into the exact mode: a Gaussian model of the mode would reach 100 %. (Marcuse’s fit, wm/a = 0.65 + 1.619 V−3/2 + 2.879 V−6, gives wm within 1 %; the page solves the mode instead.) A fiber with an 8.2 µm core and NA 0.14 has V = 2.33 at 1550 nm, a cutoff at 1500 nm and a 9.146 µm mode-field diameter. Real fibers are not ideal step-index profiles; their datasheets quote a measured mode-field diameter (usually by the Petermann II definition) and a cutoff wavelength. From those the page takes V = 2.405 λc/λ and sizes the step-index mode at that V to the datasheet’s diameter. The example, 10.4 µm with a 1260 nm cutoff, has V = 1.955 at 1550 nm, the mode of an 8.09 µm core, and couples at most 99.06 %. The page accepts 1.2 ≤ V ≤ 5, where the reference checks it; above 2.405 the fiber is no longer single-mode, and the page couples into LP₀₁ only.
Choosing the lens
The lens should focus the beam to the size of the mode. A collimated beam of radius w at the lens focuses to w′ = w f/√(f² + zR²), where zR is the input beam’s Rayleigh range [5], so the lens that matches the mode exactly is
For the example, a 2 mm beam at 1550 nm onto a 10.4 µm mode, that is 10.54 mm, and it couples 99.06 %, the mode’s best. The match is forgiving: a waist 10 % too large still couples 98.3 %, 25 % too large 95.2 %, 50 % too large 87 %. The coupling stays at 95 % or more for any focal length from 8.421 to 13.23 mm, so a catalogue lens in that range does nearly as well: an ideal 11 mm lens gives a 10.85 µm spot, and 98.9 %. The focused waist sits slightly short of f (here 10.9997 mm), which matters only for slow beams.
A real lens is not ideal. The simplest one, a plano-convex singlet, brings its outer rays to a focus short of its inner ones (spherical aberration), which spreads the spot. The page suggests one at the matching focal length and traces it with the Thin and thick lens calculator’s engine, the fiber at its best position on the axis: its curved side toward the collimated beam, the way round with less spherical aberration; in N-BK7 from 350 to 2000 nm, UV fused silica below and CaF₂ above, the glasses such lenses are usually made of; on the smallest common diameter at least twice the beam’s, which passes at least 99.97 % of its power, or failing that the widest at least 1.5 times it (at least 98.9 %); its coupling counts the light the lens clips. For the example it is a 6 mm N-BK7 lens with a 5.277 mm radius, 1.936 mm thick at the centre. Its focus has a Strehl ratio of 0.667, and with the fiber 83.8 µm nearer the lens than F′ it couples 88.1 %; flat side first it would couple 65.7 %. The aberration grows with the lens’s size at the same focusing angle: a 1 mm beam with f = 5.27 mm reaches the same mode at the same angle and couples 94.2 %, while a 5 mm beam at 1064 nm onto a 6.2 µm mode with a 980 nm cutoff, f = 22.88 mm, couples only 60 %. An aspheric lens or an achromatic doublet of the same focal length removes most of this loss: the suggestion is a lens to start from, not the best one.
Coupling efficiency
The power coupled into the mode is the overlap of the two fields on the fiber face [3][4], which the page integrates with the exact mode. With the mode taken as its best Gaussian, the overlap has closed forms. For two Gaussians whose waists coincide on the face it depends only on the ratio of their radii:
A lateral offset d, a tilt θ and a focus error Δz each lower it: d is how far from the core the beam’s centre meets the fiber face, Δz how far the face lies past the focused waist, and θ the angle between the beam and the fiber axis. For matched waists (w′ = wm = w) and one error at a time the losses take simple forms:
Lateral offset is the critical one: with an ideal 11 mm lens and the mode taken as Gaussian, the example loses 1 dB (21 % of the light) at a 2.55 µm offset, against 58 µm of focus error and 44.6 mrad of tilt. Offset and focus tolerances grow with the mode size and the tilt tolerance shrinks with it: a larger mode is easier to hit and harder to aim.
Coupling through a real lens
A singlet’s focus is not a Gaussian waist: spherical aberration spreads it and moves its best plane toward the lens. The second part of the page designs the singlet you describe and traces it with the Thin and thick lens calculator’s engine, then takes the overlap of the delivered field with the mode on the sphere about the fiber’s core, with the exact mode’s far field M in the angle α between each direction and the fiber’s axis. With the core moved by Δ across the axis, or the fiber turned, each plane wave reaches the core with an extra phase:
The second part starts from an 11 mm plano-convex lens of N-BK7, 6 mm across with a 1 mm edge, curved side toward the laser. The ideal lens of the same focal length overlaps the mode at 98.9 %; the singlet overlaps it best, at 88.7 %, with the fiber 84.4 µm closer to the lens than F′: aberrated light couples best at neither the paraxial focus nor the brightest plane, so the page places each fiber at its own best position. From there the singlet’s coupling halves with the core 4.8 µm off the axis (the ideal lens’s 4.5 µm), or with the fiber tilted by 74.2 mrad (77.9 mrad). Flat side first, the best overlap is 66.9 %; a bi-convex lens of the best form for the same f reaches 90 %, and an equi-convex one 85.1 %. The page counts the lens and the fiber end as uncoated: the two lens faces and a fused-silica fiber end pass 89.1 % of the light, so the coupling efficiency shown beside each overlap is that much lower, 79 % for the plano-convex singlet. Anti-reflection coatings recover most of the difference. A narrower beam suffers less aberration but focuses to a larger spot: with a 1 mm beam the same singlet is nearly perfect, yet overlaps the mode at only 65.3 %, because its 10.9 µm waist radius is twice the mode’s 5.2 µm; the first part matches it with a 5.27 mm lens.
Precision and limits
An independent Python reference integrates the overlap numerically on the fiber face, focuses the beam with the ABCD law, finds the matching focal length by bisection and computes the exact LP₀₁ mode from its Bessel-function equation. It agrees with the page’s closed forms for a Gaussian mode to 10⁻⁹ relative. For the exact mode, the reference’s best Gaussian, the coupling into the mode at any focal length, the matching focal length and the ends of the 95 % range agree to 10⁻⁶, and its far field, integrated directly, agrees with the page’s closed form to 10⁻⁷ of its peak.
Not included: Fresnel reflections in the first part, while the second part counts them at normal incidence for an uncoated lens and fiber end (the fiber face alone loses about 3.4 % for uncoated silica, 0.15 dB; an anti-reflection coating or index matching removes most of it), the aberrations of lenses other than spherical singlets (aspheres, doublets), clipping by a connector, polarization, and beams with M² > 1. The page refuses a focus steeper than 0.3 rad (half angle). The mode is that of an ideal step-index fiber with the given V: a real fiber’s index profile differs, so its mode can couple slightly more or less, especially far from alignment.
The plano-convex suggestion and the singlets of the second part are the Thin and thick lens calculator’s model, with its precision and limits (its own Python reference traces the singlet and integrates the focus and the overlap, offsets included, to 10⁻⁴; at F′ the lens design page’s two-dimensional fiber overlap agrees as closely, and the ideal lens matches the Gaussian closed form to 10⁻⁴; with the exact mode’s far field the singlet’s coupling, offsets included, agrees to within 10⁻⁴ of the light as well, and the ideal lens reaches the first part’s face overlap to 10⁻⁴. These are absolute differences: the page leaves out the far field where it is below 10⁻⁵ of its peak, which matters only for couplings that are themselves near 10⁻⁵ or below, such as a tilted fiber whose mode is far wider than the focus; there the relative error can reach several 10⁻⁴): the scalar focus of the traced singlet on axis, the exact mode’s far field taken on the sphere about the fiber’s core. Its glass, diameter and edge thickness (1 mm per 12.7 mm of diameter, at least 1 mm) are rules of thumb for a catalogue lens, not an optimum, and its focal length is the ideal lens’s: a plano-convex lens slightly longer can couple a little more, as its aberrated spot is wider. Expect a catalogue lens to differ in focal length, thickness and coating. The page gives no suggestion where that model does not hold: a lit pupil whose Fresnel number a²/(λf) is below 10, whose focus forms short of where the Debye integral puts it. The range of focal lengths near the match stops where focusing turns steeper than the paraxial limit.
References
- D. Gloge, “Weakly guiding fibers,” Appl. Opt. 10, 2252–2258 (1971).
- D. Marcuse, “Loss analysis of single-mode fiber splices,” Bell Syst. Tech. J. 56, 703–718 (1977).
- H. Kogelnik and T. Li, “Laser beams and resonators,” Appl. Opt. 5, 1550–1567 (1966).
- W. B. Joyce and B. C. DeLoach, “Alignment of Gaussian beams,” Appl. Opt. 23, 4187–4196 (1984).
- S. A. Self, “Focusing of spherical Gaussian beams,” Appl. Opt. 22, 658–661 (1983).