Single-Mode and Multimode Operation

Point a laser into a hair-thin glass fiber and one question decides almost everything about what it can do: does it carry a single transverse pattern of light, or hundreds at once? That choice fixes the usable bandwidth, the tolerance on every splice and connector, and whether a pulse arrives crisp or smeared after a hundred kilometres — and the answer is read off a single dimensionless number. This article consumes the mode structure worked out from the wave equation in the guiding articles — the discrete modes of a step-index cylinder, the numerical aperture, the total-internal-reflection condition that binds a ray to the core — and compresses the whole mode-counting problem into the normalized frequency, or V-number, $V=(2\pi a/\lambda_0)\mathrm{NA}$ . Two facts organize everything that follows. First, the fundamental mode $\mathrm{LP}_{01}$ has no cutoff — it propagates for any $V>0$ — while the next mode $\mathrm{LP}_{11}$ becomes bound only when $V$ climbs past $2.40483\ldots$ , the first zero of the Bessel function $J_0$ ; so $V<2.405$ means single-mode operation, and this number, inverted, hands the designer a core size and cutoff wavelength. Second, when $V$ grows large the modes proliferate, each occupying about one cell of transverse phase space, so their number swells as $M\approx V^2/2$ . A multimode fiber with a fat core and $V$ of a few tens supports many hundreds of modes and gulps light from cheap incoherent sources, but pays in modal dispersion; a single-mode fiber carries one spatial mode (two polarizations), banishes modal dispersion, and pays in a tiny core that demands micron alignment. The transition between these regimes, together with the near-Gaussian fundamental mode that often governs coupling in single-mode fibers, is central to fiber engineering.

Key takeaways

A step-index fiber's entire mode count is set by one dimensionless parameter, the normalized frequency $V=\dfrac{2\pi a}{\lambda_0}\,\mathrm{NA}=\dfrac{2\pi a}{\lambda_0}\sqrt{n_1^2-n_2^2}$ .
For a weakly guiding step-index fiber, $V<2.405$ (the first zero of $J_0$ ) means that only the fundamental LP $_{01}$ mode is guided — the fiber is single-mode; the fundamental itself has no cutoff.
The threshold defines a cutoff wavelength $\lambda_c=\dfrac{2\pi a\,\mathrm{NA}}{2.405}$ ; operating above $\lambda_c$ guarantees single-mode behaviour.
For $V\gg1$ the guided-mode count grows as $M\approx V^2/2$ (step-index) or $V^2/4$ (parabolic graded-index), both polarizations included.
Single-mode fibers kill modal dispersion and preserve spatial coherence, buying huge bandwidth; multimode fibers trade it for easy launching and more captured power.
Key idea: one number, the V-number $V=(2\pi a/\lambda_0)\mathrm{NA}$ , decides everything — $V<2.405$ is single-mode, and the mode count for large $V$ is $M\approx V^2/2$ .

The core insight

A single dimensionless parameter, the V-number $V=(2\pi a/\lambda_0)\mathrm{NA}$ , controls whether a fiber is single- or multimode: $V<2.405$ (the first zero of $J_0$ ) admits only the fundamental $\mathrm{LP}_{01}$ mode, while for $V\gg1$ the guided-mode count is $M\approx V^2/2$ .

Where this lives today

Every long-haul and metro optical link runs on single-mode fiber, and the V-number is why. By admitting one spatial mode, standard SMF (cutoff near 1.2 μm) eliminates modal dispersion at 1.31 and 1.55 μm, leaving only chromatic and polarization-mode dispersion — both compensable — and preserves the spatial coherence coherent detection and dense wavelength-division multiplexing demand. For more capacity per strand, operators step just past cutoff into the few-mode regime ( $2.405<V\lesssim5$ ), where space-division multiplexing sends an independent channel on each of a few LP modes, recovered with few-mode amplifiers and MIMO signal processing. Conversely, at the other extreme, multimode graded-index fiber with 50 μm cores dominates the data center, where cheap 850 nm VCSELs, loose tolerances, and short reaches make captured power matter more than bandwidth. The same reasoning, ported to rectangular geometry, dictates the dimensions of waveguides in silicon and silicon-nitride integrated photonics, where a single-mode rib waveguide guarantees the clean interference on-chip modulators, microring resonators, and directional couplers rely on. From transoceanic cables to a fingernail-sized coherent transceiver, the first design number written down is still $V$ .

The physical picture

Think of $V$ as counting how much transverse structure the guide can hold: it multiplies how many wavelengths span the core by how strongly the core grips the light. A wider core, a shorter wavelength, or a bigger index contrast each packs more standing-wave patterns across the fiber — more modes. A guided mode is a transverse pattern that reproduces itself on propagation; in the ray picture it is a bundle of rays bouncing at one discrete angle inside the acceptance cone, and $V$ measures that phase-space volume. The fundamental $\mathrm{LP}_{01}$ is a single smooth bell with no transverse nodes, so nothing forbids it at any $V$ — it always fits. The next mode, LP $_{11}$ , has a transverse node, and its guided solution appears only when $V$ is large enough to satisfy its cutoff condition. The magic number $2.405$ is simply the first zero of $J_0$ , the point at which the $\mathrm{LP}_{11}$ field can finally stop decaying in the cladding and become bound. Below it, exactly one solution survives: one mode.

Figure

Transverse mode families appear as V increases. The node-free LP₀₁ mode has no cutoff in the ideal weakly guiding step-index model; LP₁₁ reaches cutoff at V ≈ 2.405, while the LP₂₁ and LP₀₂ mode families reach cutoff at V ≈ 3.832. Thus the fiber is single-mode for V less than 2.405.

Push $V$ higher and the tidy staircase of cutoffs blurs into a smooth cloud of modes: rather than track each one, you count how many transverse states fit inside the guiding cone — the core area times the transverse-wavevector cone area, divided into phase-space cells and doubled for polarization. That geometric estimate, $M\approx V^2/2$ , is the second organizing law and the reason a fat multimode core carries so many modes. The next figure is what a designer reasons with: mode count versus V-number, single-mode window at the far left, a real multimode fiber marked at $V\approx37$ .

Figure

Guided-mode count M versus V-number for a step-index fiber, M ≈ V²/2. The shaded strip V less than 2.405 is the single-mode window where only LP₀₁ survives. A multimode fiber with a = 25 μm, NA = 0.20 at λ₀ = 0.85 μm reaches V ≈ 36.96 and M ≈ 683 modes.

The core equations

The whole subject rests on one definition. The normalized frequency bundles the core radius, wavelength, and index contrast into a single dimensionless group built from the numerical aperture — and every other result here is a statement about its value.

V=\frac{2\pi a}{\lambda_0}\,\mathrm{NA}=\frac{2\pi a}{\lambda_0}\sqrt{n_1^2-n_2^2} \qquad (1)

where: $V$ — normalized frequency (V-number), dimensionless; $a$ — core radius; $\lambda_0$ — free-space wavelength; $n_1,n_2$ — core and cladding indices; $\mathrm{NA}=\sqrt{n_1^2-n_2^2}$ . Physical reading: $V$ is the "electrical size" of the guide — large $V$ means a big, tightly guiding core holding many modes; small $V$ a weakly guiding one. The single-mode condition is the sharpest result in fiber optics: an exact threshold, not a rule of thumb. The number itself is a Bessel-function zero, derived in the first card below.

V<V_c=2.40483\ldots\ \approx 2.405\qquad\Longleftrightarrow\qquad\text{single-mode} \qquad (2)

where: $V_c$ — cutoff V-number, the first positive zero of $J_0$ . Physical reading: at $V=V_c$ the second mode $\mathrm{LP}_{11}$ reaches cutoff; for any smaller $V$ only $\mathrm{LP}_{01}$ propagates, so the fiber carries a single spatial mode. Inverting the condition for wavelength gives the quantity quoted on single-mode-fiber datasheets. Because $V$ falls as $\lambda_0$ rises, there is a cutoff wavelength above which the fiber remains single-mode.

\lambda_c=\frac{2\pi a\,\mathrm{NA}}{2.405} \qquad (3)

where: $\lambda_c$ — cutoff wavelength; single-mode for every $\lambda_0>\lambda_c$ . Physical reading: standard SMF is engineered with $\lambda_c\approx1.2\,\mu\mathrm{m}$ , safely below the 1.31 and 1.55 μm windows. When $V$ is large the cutoffs merge and mode-counting becomes statistical. Phase-space counting (the second card) gives a compact estimate scaling as $V^2$ , halved for a parabolic profile.

M\approx\frac{V^2}{2}\ \ (\text{step-index}),\qquad M\approx\frac{V^2}{4}\ \ (\text{parabolic graded-index}) \qquad (4)

where: $M$ — total guided-mode count, both polarizations. Physical reading: doubling $V$ quadruples the mode count; a graded profile supports half as many modes as a step-index fiber of the same $V$ , and equalizes their group velocities. Underneath the cutoff sits the mode machinery. Every guided mode splits $V^2$ between an oscillatory core part $u$ and an evanescent cladding part $w$ ; matching the field across the boundary fixes the split into one transcendental eigenvalue equation.

u^2+w^2=V^2,\qquad u\,\frac{J_1(u)}{J_0(u)}=w\,\frac{K_1(w)}{K_0(w)} \qquad (5)

where: $u=a\sqrt{n_1^2k_0^2-\beta^2}$ — core transverse wavenumber; $w=a\sqrt{\beta^2-n_2^2k_0^2}$ — cladding decay constant; $\beta$ — axial propagation constant; $J,K$ — Bessel and modified Bessel functions. Physical reading: a well-confined mode has most of $V^2$ in $w$ ; at cutoff $w\to0$ , $u\to V$ , and the field leaks away. Each root is one guided mode $\mathrm{LP}_{0m}$ . Finally, the shape of the one mode that matters most. The fundamental is nearly Gaussian, and its width — not the physical core — governs all coupling; the Marcuse fit gives that width from $V$ alone.

w_0\approx a\left(0.65+\frac{1.619}{V^{3/2}}+\frac{2.879}{V^{6}}\right),\qquad \eta=\left[\frac{2\,w_{01}w_{02}}{w_{01}^2+w_{02}^2}\right]^{2} \qquad (6)

where: $w_0$ — $1/e$ mode-field radius, $\mathrm{MFD}=2w_0$ ; $\eta$ — power coupling between two Gaussian mode fields of radii $w_{01},w_{02}$ . Physical reading: near cutoff a real fraction of power rides in the cladding, so $w_0>a$ — matching mode fields, not cores, sets splice loss. The contrast feeding $\mathrm{NA}$ is quoted as $\Delta=(n_1^2-n_2^2)/2n_1^2\ll1$ , so $\mathrm{NA}\approx n_1\sqrt{2\Delta}$ — the weak-guidance limit that collapses the vector modes into LP modes.

Derivation AFrom the wave equation to the V less than 2.405 single-mode rule

The goal is to show, from the cylindrical scalar wave equation, why the guided modes form a discrete ladder and why exactly one rung survives below $V=2.405$ . We follow the field from the Helmholtz equation into separated Bessel solutions, impose boundedness and boundary matching to get the eigenvalue equation, then locate the second mode's cutoff. The payoff: the threshold $2.40483\ldots$ is not an empirical fit but the first positive zero of a Bessel function arising from the cylindrical boundary conditions.

Step 1. In the weak-guidance limit $n_1\approx n_2$ , the vector modes collapse into scalar linearly polarized modes, and the transverse field $\psi(r,\phi)\,e^{i(\beta z-\omega t)}$ obeys $\nabla_T^2\psi+(n^2k_0^2-\beta^2)\psi=0$ with $k_0=2\pi/\lambda_0$ . Why: weak guidance removes the polarization coupling of the full Maxwell problem, leaving a single scalar Helmholtz equation that is tractable in closed form.

Step 2. Separate variables as $\psi(r,\phi)=R(r)\,e^{il\phi}$ with integer $l=0,1,2,\dots$ ; the radial factor satisfies Bessel's equation. Why: azimuthal single-valuedness forces integer angular order $l$ , which labels the mode's azimuthal node count.

Step 3. Boundedness at the axis and decay at infinity fix the radial solutions to $R(r)\propto J_l(ur/a)$ for $r<a$ and $K_l(wr/a)$ for $r>a$ , with $u^2+w^2=V^2$ . Why: a guided mode must carry finite power and stay bound, ruling out growing solutions and tying the two regions through the Pythagorean budget of equation (5).

Step 4. Continuity of $\psi$ and $\partial_r\psi$ at $r=a$ eliminates the amplitudes and yields the transcendental eigenvalue equation (5); its discrete roots $u_{lm}$ enumerate the modes $\mathrm{LP}_{lm}$ . Why: the boundary conditions quantize the continuum of $\beta$ into a discrete mode ladder — no matching, no discreteness.

Step 5. A mode reaches cutoff when $w\to0$ so $u\to V$ ; the characteristic equation then reduces to $J_{l-1}(V)=0$ . For $\mathrm{LP}_{11}$ ( $l=1$ ) this is $J_0(V)=0$ , first root $V=2.40483\ldots$ , so for $V<2.405$ the only bound solution is $\mathrm{LP}_{01}$ (no cutoff, since its threshold would be the $V=0$ zero of $J_{-1}$ ) with its two degenerate polarizations. Why: at cutoff the cladding field stops decaying, collapsing the mode condition onto a single Bessel zero — the threshold below which the fiber cannot support a second spatial pattern.

Go deeper · Derivation B · 1 minPhase-space counting of the mode number for large V

Derivation BPhase-space counting of the mode number for large V

The goal is to derive $M\approx V^2/2$ without solving the eigenvalue equation mode by mode, hopeless once hundreds are bound. The trick is a semiclassical phase-space count: guided modes occupy states inside a transverse-wavevector cone and across the core cross-section, and their number is the phase-space volume divided into cells of size $(2\pi)^2$ and doubled for polarization — the Weyl density-of-states counting, specialized to the fiber's acceptance cone.

Step 1. A ray is guided only if its transverse wavenumber stays inside the acceptance cone, $k_T\le k_{T,\max}=k_0\,\mathrm{NA}$ ; larger wavenumbers exceed the critical angle and leak away. Why: the numerical aperture is the boundary of the bound-ray cone, capping the transverse momentum a mode may carry.

Step 2. The transverse-wavevector area is a disk $A_k=\pi k_{T,\max}^2=\pi k_0^2\,\mathrm{NA}^2$ ; the real-space area is the core cross-section $A_r=\pi a^2$ . Why: each mode lives somewhere in the core and points somewhere inside the cone, so its phase space is the product of these two areas.

Step 3. Divide the phase-space volume by the cell $(2\pi)^2$ and double for polarization: $M\approx \dfrac{A_r A_k}{(2\pi)^2}\times2=\dfrac{a^2 k_0^2\,\mathrm{NA}^2}{2}$ . Why: $(2\pi)^2$ is the smallest phase-space area a distinct mode can occupy, so dividing by it counts them.

Step 4. Recognize $V=a k_0\,\mathrm{NA}$ from equation (1) and substitute: $M\simeq V^2/2$ for a step-index fiber. For an ideal parabolic graded-index profile, the corresponding large- $V$ mode count is $M\simeq V^2/4$ . Why: the graded-index profile changes the transverse density of states, giving approximately half as many guided modes as the step-index case at the same $V$ and reducing intermodal group-delay differences.

Go deeper · Derivation C · 1 minThe near-Gaussian fundamental mode and coupling efficiency

Derivation CThe near-Gaussian fundamental mode and coupling efficiency

The goal is to justify treating $\mathrm{LP}_{01}$ as a Gaussian beam of radius $w_0$ — the approximation behind every splice- and connector-loss calculation — and to read off the overlap efficiency that turns a mode-field mismatch into a number. The fundamental is not exactly Gaussian (a $J_0$ core joined to a $K_0$ tail), but close enough that the whole coupling problem reduces to matching one width, $w_0$ , given by the Marcuse fit of equation (6).

Step 1. The fundamental radial field is bell-shaped with no nodes ( $J_0$ core, decaying $K_0$ tail), and a Gaussian $\psi\propto e^{-r^2/w_0^2}$ reproduces it to about one percent over $1.2\lesssim V\lesssim2.4$ ; Marcuse's fit sets the width from $V$ alone, $w_0/a\approx0.65+1.619/V^{3/2}+2.879/V^{6}$ , with $w_0>a$ near cutoff. Why: as $V$ falls toward cutoff the cladding decay $w$ weakens and the mode spreads beyond the core — so the mode-field diameter, not the core, is the physically relevant size.

Step 2. The power coupling between two Gaussian mode fields is $\eta=[\,2w_{01}w_{02}/(w_{01}^2+w_{02}^2)\,]^2$ , equal to one only when the widths match. Why: splice and connector loss is set by mode-field overlap, so matching mode-field diameters minimizes loss — which is why datasheets quote MFD, not core size.

Limiting cases

The entire behaviour of a step-index fiber is organized by the value of $V$ :

Case	Result	Meaning
Single-mode	$0<V<2.405$	Only $\mathrm{LP}_{01}$ (two polarizations). No modal dispersion; bandwidth set by chromatic and polarization effects.
Few-mode	$2.405<V\lesssim5$	A handful of low-order modes, exploited in mode-division multiplexing and few-mode amplifiers.
Highly multimode	$V\gg1\Rightarrow M\approx V^2/2$	Hundreds to thousands of modes; large core eases launching but modal dispersion dominates the bandwidth budget.
Cutoff wavelength	$\lambda_c=\dfrac{2\pi a\,\mathrm{NA}}{2.405}$	Single-mode for $\lambda_0>\lambda_c$ ; standard SMF has $\lambda_c\approx1.2\,\mu\mathrm{m}$ , below both telecom windows.
Near cutoff ( $V\to2.405^{-}$ )	$w_0>a$	Mode field bulges into the cladding; splice tolerance tightens, microbend loss rises.

Common trapThe cutoff V = 2.405 is where the fundamental mode turns on, so below it no light is guided

This inverts the physics. $V=2.405$ is the cutoff of the second mode, $\mathrm{LP}_{11}$ — where a second spatial pattern first becomes bound. The fundamental $\mathrm{LP}_{01}$ has no cutoff at all: it propagates for every $V>0$ , right down to a vanishingly small core. So $V<2.405$ does not mean "no guiding"; it means "guiding of exactly one spatial mode", the whole point of single-mode fiber. A related trap: the mode-field diameter is generally larger than the core diameter, because the guided field extends evanescently into the cladding — so coupling, splice loss, and alignment tolerance are governed by the mode field, not the physical core (an 8.2 μm core can have a 10.4 μm mode field).

Worked examples

1 · Sizing a single-mode fiber at 1.55 μm (the concrete design)

Design a step-index fiber that is single-mode at $\lambda_0=1.55\,\mu\mathrm{m}$ with core index $n_1=1.4500$ and relative index difference $\Delta=0.003$ . Find the numerical aperture, the maximum core radius, and the mode-field diameter.

Step 1. Numerical aperture from the weak-guidance form: $\mathrm{NA}\approx n_1\sqrt{2\Delta}=1.45\sqrt{0.006}=1.45\times0.07746=\mathbf{0.1123}$ . The small index contrast gives a low NA, characteristic of single-mode fiber and its narrow acceptance cone.

Step 2. Impose the cutoff of equation (2) and solve for the largest single-mode core: $a_{\max}=\dfrac{2.405\,\lambda_0}{2\pi\,\mathrm{NA}}=\dfrac{2.405\times1.55\,\mu\mathrm{m}}{2\pi\times0.1123}=\dfrac{3.728}{0.7057}=\mathbf{5.28\,\mu m}$ , a maximum core diameter of $2a_{\max}\approx\mathbf{10.6\,\mu m}$ .

Step 3. Mode field from the Marcuse fit, equation (6), at the cutoff design $V=2.405$ : $w_0/a\approx0.65+1.619/2.405^{3/2}+2.879/2.405^{6}=1.10$ , so $w_0\approx1.10\times5.28=5.81\,\mu\mathrm{m}$ and $\mathrm{MFD}=2w_0\approx\mathbf{11.6\,\mu m}$ — wider than the 10.6 μm core.

Step 4. Reality check: SMF-28 has a core radius of about $4.1\,\mu\mathrm{m}$ and a measured $\mathrm{MFD}\approx10.4\,\mu\mathrm{m}$ at 1.55 μm. Using its specified numerical aperture of about $0.14$ gives $V\approx2.33$ at 1.55 μm, consistent with single-mode operation. NA = 0.1123, a_max = 5.28 μm (10.6 μm diameter), MFD ≈ 11.6 μm at cutoff.

2 · Counting modes in a multimode fiber and its bandwidth cost

A multimode fiber has core radius $a=25\,\mu\mathrm{m}$ and $\mathrm{NA}=0.20$ , operated at $\lambda_0=0.85\,\mu\mathrm{m}$ . How many modes does it guide, and what is the modal delay spread?

Step 1. Compute the V-number from equation (1): $V=\dfrac{2\pi(25\,\mu\mathrm{m})}{0.85\,\mu\mathrm{m}}(0.20)=\dfrac{157.08}{0.85}\times0.20=\mathbf{36.96}$ — deep in the multimode regime, over 15 times the single-mode cutoff.

Step 2. Count the modes with equation (4): $M\approx\dfrac{V^2}{2}=\dfrac{(36.96)^2}{2}=\dfrac{1366}{2}\approx\mathbf{683}$ modes, each with a slightly different group velocity. A parabolic profile of the same $V$ carries about $V^2/4\approx342$ .

Step 3. Estimate the intermodal delay spread. For a step-index profile the delay per unit length is $\dfrac{\Delta\tau}{L}\approx\dfrac{n_1\Delta}{c}$ , where $\Delta=\dfrac{\mathrm{NA}^2}{2n_1^2}=\dfrac{0.04}{2(1.46)^2}\approx0.00938$ . With $n_1\approx1.46$ : $\dfrac{\Delta\tau}{L}\approx\dfrac{1.46\times0.00938}{3\times10^8}\approx4.6\times10^{-11}\ \mathrm{s/m}\approx\mathbf{46\ ns/km}$ .

Step 4. Read off the bandwidth: a 46 ns/km spread caps the bandwidth–distance product at roughly $1/(2\times46\ \mathrm{ns/km})\approx\mathbf{11\ MHz\!\cdot\! km}$ — a few hundred metres at gigabit rates. A parabolic profile cuts this delay by about $\Delta/8$ , which is why graded-index fiber is used for multimode links. V ≈ 36.96, M ≈ 683 modes, delay ≈ 46 ns/km, bandwidth ≈ 11 MHz·km.

3 · Cutoff wavelength check

A fiber has $a=4.0\,\mu\mathrm{m}$ and $\mathrm{NA}=0.12$ . Is it single-mode at 1.31 μm?

Step 1. Cutoff wavelength from equation (3): $\lambda_c=\dfrac{2\pi a\,\mathrm{NA}}{2.405}=\dfrac{2\pi(4.0)(0.12)}{2.405}=\mathbf{1.254\,\mu m}$ . Equivalently $V=\dfrac{2\pi(4.0)}{1.31}(0.12)=\mathbf{2.30}<2.405$ at 1.31 μm — only $\mathrm{LP}_{01}$ propagates, and the operating $V$ sits just 4% below cutoff. λ_c = 1.25 μm; at 1.31 μm, V = 2.30 < 2.405, single-mode.

Physics of the portrait

Single- versus multimode operation is where wave optics, special functions, and a billion kilometres of deployed glass meet. The mode nomenclature and the very idea of engineering a fiber to carry one mode trace to one experimentalist who built the first single-mode guides and the first fiber lasers.

Origins · 1 minElias Snitzer and how this was discovered

Elias Snitzer

1925–2012 · glass waveguide modes, fiber lasers and amplifiers

Elias Snitzer was the American physicist who turned the mode theory of a dielectric cylinder into a working technology. In 1961, at American Optical, he published the definitive analysis of the modes of a cylindrical dielectric waveguide, establishing the $\mathrm{HE}$ , $\mathrm{EH}$ , $\mathrm{TE}$ , $\mathrm{TM}$ families and, in the weak-guidance limit, the linearly polarized combinations later systematized as LP modes. In the same period he demonstrated the first fiber supporting only the fundamental mode, giving "single-mode" operation physical reality, and pioneered neodymium-doped glass and fiber lasers and amplifiers. Two ideas became concrete: that mode content is a design variable fixed by core size and index contrast, and that a single-mode core could be pumped to amplify or lase — the seed of the erbium-doped fiber amplifier behind the internet's backbone and the rare-earth fiber laser that now cuts steel. The mode labels every fiber engineer uses today are, in large part, his.

History

1961

— Snitzer analyzes the modes of a cylindrical dielectric waveguide and demonstrates single-mode glass guiding

1964

— Snitzer and Koester demonstrate a neodymium-doped glass fiber amplifier

1966

— Kao and Hockham argue that purified glass fiber can reach a loss low enough for communications

1970

— Corning fabricates fiber below 20 dB per km, making long single-mode links practical

1971

— Gloge's weakly guiding approximation defines the LP modes and the V-number cutoff

Going further

Texts. Saleh and Teich, Fundamentals of Photonics, 2nd ed., Ch. 9 (Fiber Optics) · Hecht, Optics, 5th ed., §5.6 (Fiberoptics) · G. P. Agrawal, Fiber-Optic Communication Systems, Ch. 2 · Loudon, The Quantum Theory of Light, 3rd ed. (waveguide mode structure) · Foot, Atomic Physics (optical fibers in experiments) · Demtröder, Laser Spectroscopy (fiber coupling and modes). The LP-mode framework is Gloge's 1971 weakly guiding approximation, itself resting on Snitzer's 1961 cylindrical-waveguide analysis. Key numerical anchors: the first zero of $J_0$ is $j_{0,1}=2.40483$ ; the worked examples give SMF sizing ( $\mathrm{NA}\approx0.112$ , $a_{\max}\approx5.28\,\mu\mathrm{m}$ , $\mathrm{MFD}\approx11.6\,\mu\mathrm{m}$ ), the multimode case ( $V\approx36.96$ , $M\approx683$ , $\approx46$ ns/km), and the cutoff check ( $\lambda_c\approx1.25\,\mu\mathrm{m}$ , $V=2.30$ at 1.31 μm). All indices and datasheet figures (SMF-28 core 8.2 μm, MFD ~10.4 μm at 1.55 μm) are standard tabulated values; no fabricated constants are used.

Prerequisites. Total Internal Reflection and Optical Guiding · Waveguide Modes and the Wave Equation · Numerical Aperture and Acceptance Angle · Step-Index and Graded-Index Optical Fibers. Leads to. Chromatic and Modal Dispersion in Fibers · Fiber Attenuation and Loss Mechanisms · Polarization-Mode Dispersion and Birefringent Fibers · Coupled-Mode Theory and Directional Couplers · Integrated Photonic Waveguides and Circuits.

Ask the tutor: Give me a target wavelength, core size, and index contrast, and I'll tell you whether your fiber is single-mode, its cutoff wavelength, or how many modes it carries.

Does "single-mode" mean the light has only one degree of freedom?

No — it means one spatial mode, but that mode still carries two orthogonal polarizations. In a perfectly circular, stress-free fiber the two are exactly degenerate, so we count them as one mode. Real fibers have slight core ellipticity or frozen-in stress, which lifts the degeneracy and makes the polarizations travel at slightly different speeds — the residual birefringence that causes polarization-mode dispersion, a small but real limit on high-bit-rate links. So "single-mode" is shorthand for one spatial pattern, not one physical channel.

Why is the cutoff exactly 2.405 and not some rounder number?

Because it is a zero of a special function, not a design choice. In the weakly guiding limit the second mode $\mathrm{LP}_{11}$ reaches cutoff when its cladding field stops decaying, and its characteristic equation collapses to $J_0(V)=0$ , whose first positive root is $2.40483\ldots$ , rounded to 2.405. The cylindrical geometry selects $J_0$ ; a planar slab waveguide has cosine modes and integer-multiple cutoffs instead. The number is as fundamental to the round fiber as $\pi$ is to a circle.

If I double the core radius, what happens to the number of modes?

The mode count scales as $V^2$ , and $V\propto a$ , so doubling the core radius quadruples $M$ : a fiber with 683 modes at $a=25\,\mu\mathrm{m}$ carries roughly $2700$ at $a=50\,\mu\mathrm{m}$ . The same scaling near single-mode pushes you across the $V=2.405$ threshold — enlarging the core is the fastest way to turn a single-mode fiber multimode — why core radius is held to microns.

Why does multimode fiber have so much less bandwidth than single-mode?

Because its many modes travel at different group velocities. A launched pulse spreads its energy across hundreds of modes; the fastest and slowest arrive at different times, smearing the pulse — modal dispersion, growing linearly with distance. For step-index multimode fiber the spread is tens of ns/km, capping the bandwidth–distance product at tens of MHz·km. A single-mode fiber has one spatial mode, so there is nothing for modal dispersion to act on and its bandwidth is orders of magnitude higher. Graded-index fiber is the compromise: a parabolic profile nearly equalizes the mode velocities, recovering much of the bandwidth while keeping a large core.

Why does the datasheet quote a mode-field diameter instead of the core diameter?

Because coupling loss is set by the field, not the glass. A real fraction of the power travels in the cladding as an evanescent tail, so the $1/e$ mode-field radius $w_0$ exceeds the core radius $a$ — typically MFD ≈ 10.4 μm for an 8.2 μm core at 1.55 μm. When you splice two fibers or couple a laser, the transferred power is the overlap integral $\eta=[2w_{01}w_{02}/(w_{01}^2+w_{02}^2)]^2$ , set by the mode-field diameters. Matching MFDs minimizes loss, so the MFD is the number a designer needs.

What is the "few-mode" regime and why would anyone want it?

It is the narrow window just above cutoff, $2.405<V\lesssim5$ , where a fiber supports only a handful of well-separated modes — $\mathrm{LP}_{01}$ , $\mathrm{LP}_{11}$ , and perhaps one or two more. Rather than a nuisance, these extra spatial modes become independent data channels: space-division multiplexing sends a separate signal on each and untangles them with MIMO signal processing, multiplying a fiber's capacity by the number of modes. The mode-counting in this article is what tells a designer how many usable channels a given $V$ provides.