Photonic-Crystal and Hollow-Core Fibers

Take the transverse eigenvalue problem that aq-optics-waveguide-modes-wave-equation extracts from Maxwell's equations — a $z$ -invariant dielectric whose modes are fixed by the cross-sectional index profile $n(x,y)$ and a propagation constant $\beta$ — and hand it a cross-section no melt-and-draw dopant profile could ever make: glass riddled with a periodic lattice of air holes running the whole length of the fiber. That is a photonic-crystal fiber. In a step-index fiber the core is a few tenths of a percent denser than the cladding because that is all the dopant chemistry allows; in a PCF the cladding is a designed composite of glass and air, its optical response — an effective index and a full photonic band structure — set by two numbers you draw into the preform: the hole pitch $\Lambda$ and the ratio $d/\Lambda$ . That freedom splits guidance in two. A solid glass core in the air-hole lattice guides by a modified total internal reflection, because the air-diluted cladding has a lower, strongly wavelength-dependent effective index. A hollow air core cannot guide by total internal reflection at all — its index is below the cladding average — yet it confines light through a photonic band gap, or through anti-resonant reflection from the thin glass walls, so that nearly all the power travels in gas or vacuum. From this one structural idea flow capabilities ordinary fibers cannot reach: a fiber single-mode at every wavelength, a zero-dispersion wavelength dragged from the infrared into the visible, mode areas spanning three orders of magnitude in one material, and light propagating in air over metres — the platform behind supercontinuum sources, frequency combs, high-power fiber lasers, and gas-phase nonlinear optics.

Key takeaways

A photonic-crystal fiber (PCF) replaces the doped cladding of an ordinary fiber with a microstructured lattice of air holes; its properties are set by geometry — the pitch $\Lambda$ and ratio $d/\Lambda$ — not by dopant chemistry.
Solid-core PCFs guide by modified total internal reflection: the air-diluted cladding has a low, wavelength-dependent effective index $n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)$ , so a pure-glass core traps light with a self-adjusting contrast.
Because that contrast collapses as the wavelength shrinks, the normalized frequency saturates and a small-hole PCF ( $d/\Lambda \lesssim 0.4$ ) is endlessly single-mode — no second guided mode at any wavelength.
Hollow-core PCFs guide light in air by a photonic band gap or anti-resonant reflection; with most of the power in gas or vacuum they show ultralow nonlinearity, high damage threshold, and metre-scale light–gas interaction.
Geometry tunes the group-velocity dispersion — shifting the zero-dispersion wavelength into the visible and giving mode areas from below $1\,\mu\mathrm{m}^2$ to over $1000\,\mu\mathrm{m}^2$ in one material.
Key idea: in a PCF the cladding is a structured composite whose optical response is a designed quantity — its effective index $n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)$ and photonic band structure, not any bulk material constant, decide whether light is confined, how many modes survive, and how the group velocity disperses.

The core insight

In a photonic-crystal fiber the cladding is a structured composite whose optical response is a designed quantity: its effective index $n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)$ and its photonic band structure — not a bulk material property — determine whether light is confined, at what wavelengths a single mode survives, and how the group velocity disperses.

Where this lives today

The most visible legacy of the PCF sits inside every self-referenced optical frequency comb, and hence the world's best atomic clocks. A femtosecond laser is launched into a short highly nonlinear, small-core PCF whose zero-dispersion wavelength has been dragged to the pump; soliton fission and dispersive-wave generation broaden the spectrum past a full octave, so the comb's red end can be frequency-doubled and beat against its blue end. That $f$ – $2f$ measurement pins the offset frequency and turns a pulse train into an absolute optical ruler — the step behind today's optical-lattice clocks accurate to parts in $10^{18}$ . At the other extreme, hollow-core fibers are moving from curiosity to infrastructure. Nested anti-resonant nodeless fibers have pushed air-guided loss below the Rayleigh-scattering floor of solid silica, and because the light travels in air its latency is about $1.5\%$ lower than in glass with nearly vanishing nonlinearity and dispersion — now trialled for low-latency data-centre and financial links and high-power pulse delivery. Gas-filled hollow-core PCFs serve as compact reaction vessels for stimulated Raman conversion, soliton self-compression to few-cycle pulses, and vacuum-ultraviolet generation, where light and a controlled gas overlap over metres.

The physical picture

Picture the air-hole cladding as Swiss-cheese glass. A mode spread over many holes cannot resolve any single hole; it feels only their average — a diluted glass of intermediate index. The solid core is denser glass, so light sits in it held by ordinary total internal reflection against a cladding that happens to be a glass-and-air blend rather than a doped melt. Only the cladding index is unusual, and — crucially — it breathes with wavelength. That breathing is the trick behind endlessly single-mode operation. Shrink the wavelength and the mode contracts into the thin glass webs between the holes, sampling more glass, so the effective cladding index climbs toward the pure-glass core. The confining contrast shrinks exactly as fast as the wavelength does, and the two cancel: the normalized frequency levels off instead of climbing without bound as in a fixed-contrast step-index fiber. A step-index fiber goes multimode once the wavelength is short enough; a small-hole PCF never does. For a hollow core the picture inverts. Think of the thin glass membranes around the air core as Fabry–Pérot-like layers. At most frequencies core light finds a phase-matched partner mode in the glass and leaks away; but in a photonic band gap — an anti-resonance of the membranes — there is nowhere in the glass for it to go, so it is reflected back. It is a colour-selective mirror wrapped around a tube of air, not total internal reflection, and it works only in discrete windows set by the membrane thickness.

Figure

Two ways to trap light in one platform. Left a solid glass core in a triangular air-hole lattice of pitch Λ ≈ 2.3 μm and d/Λ = 0.4 guides by modified total internal reflection and is endlessly single-mode. Right a hollow air core of index ≈ 1 ringed by thin glass tubes of wall thickness t guides by anti-resonant reflection, keeping most of the power in air

Because dispersion is the curvature of $\beta(\omega)$ , and $\beta$ is fixed by that engineered cross-section, you can place the zero-dispersion wavelength almost anywhere by choosing $\Lambda$ and $d/\Lambda$ — dragging it from silica's fixed material value near $1.27\,\mu\mathrm{m}$ down into the visible, impossible in a single-material step-index fiber. The second figure shows the most iconic PCF result: the normalized frequency that refuses to grow.

Figure

Why a small-hole PCF is endlessly single-mode. The generalized normalized frequency V_eff versus Λ/λ saturates as the wavelength shrinks. For d/Λ = 0.4 it levels off below the second-mode cutoff V_eff = 2.405, so only the fundamental mode is ever guided, whereas d/Λ = 0.55 crosses the cutoff near Λ/λ ≈ 1.7 and turns multimode at short wavelengths

The core equations

Everything begins with the modal effective index — the propagation constant $\beta$ in units of the free-space wavenumber. It is the single number fixing how fast a mode's phase advances and, through its wavelength derivatives, how it disperses.

n_{\mathrm{eff}} = \dfrac{\beta}{k_0}, \qquad k_0 = \dfrac{2\pi}{\lambda} = \dfrac{\omega}{c} \qquad (1)

where: $\beta$ — propagation constant of the guided mode; $k_0$ — free-space wavenumber; $n_{\mathrm{eff}}$ — modal effective index. Index guiding needs $n_{\mathrm{eff}}^{\mathrm{cl}} < n_{\mathrm{eff}} < n_{\mathrm{core}}$ ; a hollow-core band-gap mode instead has $n_{\mathrm{eff}} < 1$ , below every material index present. That effective index is the eigenvalue of the transverse wave equation on the microstructured cross-section — the vector Helmholtz equation for a $z$ -invariant fiber, separated so that the entire hole geometry enters through $n(x,y)$ .

\left[\nabla_t^2 + \big(k_0^2\, n^2(x,y) - \beta^2\big)\right]\boldsymbol{\psi}(x,y) = 0 \qquad (2)

where: $\nabla_t^2 = \partial_x^2 + \partial_y^2$ — transverse Laplacian; $n(x,y)$ — microstructured index profile (glass webs and air holes); $\boldsymbol{\psi}$ — transverse modal field; $\beta^2$ — the eigenvalue. In practice this is solved numerically, by plane-wave expansion of the periodic cladding or by finite-element mode solvers. To compare a PCF against step-index intuition we assign it a generalized normalized frequency. The key twist is that the cladding index inside the root is itself a function of wavelength, taken from the fundamental space-filling mode of the infinite lattice.

V_{\mathrm{eff}} = \dfrac{2\pi}{\lambda}\,\rho\,\sqrt{n_{\mathrm{core}}^2 - n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)^2} \qquad (3)

where: $\rho$ — effective core radius, commonly $\Lambda/\sqrt{3}$ or $\Lambda/2$ ; $n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)$ — wavelength-dependent effective cladding index; $V_{\mathrm{eff}}$ — generalized normalized frequency. Higher-order modes appear only once $V_{\mathrm{eff}}$ exceeds a cutoff near the step-fiber value $2.405$ — a threshold a saturating $V_{\mathrm{eff}}$ need never reach. In the coarse, long-wavelength limit the mode averages over many holes and the cladding acts as a homogeneous effective medium, its index given by the volume-averaged (Maxwell–Garnett) rule with a geometry-fixed air fill fraction.

n_{\mathrm{cl}}^{\mathrm{avg}} = \sqrt{f\,n_{\mathrm{air}}^2 + (1-f)\,n_{\mathrm{glass}}^2}, \qquad f = \dfrac{\pi}{2\sqrt{3}}\left(\dfrac{d}{\Lambda}\right)^2 \qquad (4)

where: $f$ — air fill fraction of a triangular lattice; $d/\Lambda$ — hole-to-pitch ratio, the master design parameter; $n_{\mathrm{glass}}$ — host-glass index (silica $\approx 1.444$ at $1.5\,\mu\mathrm{m}$ ). This average is a lower bound: at finite wavelength the field concentrates in the glass, so $n_{\mathrm{eff}}^{\mathrm{cl}} > n_{\mathrm{cl}}^{\mathrm{avg}}$ . Dispersion is where PCFs earn their keep: the dispersion parameter is the curvature of the modal index in wavelength, and in a PCF that curvature is geometry-tunable rather than fixed by the material.

D = -\dfrac{\lambda}{c}\,\dfrac{\mathrm{d}^2 n_{\mathrm{eff}}}{\mathrm{d}\lambda^2} = -\dfrac{2\pi c}{\lambda^2}\,\beta_2 \qquad (5)

where: $D$ — dispersion parameter, ps nm $^{-1}$ km $^{-1}$ ; $\beta_2 = \mathrm{d}^2\beta/\mathrm{d}\omega^2$ — group-velocity dispersion, anomalous when $\beta_2 < 0$ . Splitting $D$ into a material part (silica's Sellmeier) and a waveguide part (geometry) is what lets a PCF cross zero dispersion in the visible. The second lever is confinement. The effective mode area sets how intense a given power becomes, and hence the strength of every nonlinear process, packaged into the nonlinear coefficient $\gamma$ .

A_{\mathrm{eff}} = \dfrac{\left(\displaystyle\int |\mathbf{E}|^2\,\mathrm{d}A\right)^2}{\displaystyle\int |\mathbf{E}|^4\,\mathrm{d}A}, \qquad \gamma = \dfrac{2\pi n_2}{\lambda\,A_{\mathrm{eff}}} \qquad (6)

where: $A_{\mathrm{eff}}$ — effective mode area, below $1\,\mu\mathrm{m}^2$ (highly nonlinear) to over $1000\,\mu\mathrm{m}^2$ (large-mode-area); $\gamma$ — nonlinear coefficient, W $^{-1}$ m $^{-1}$ ; $n_2$ — Kerr index ( $\approx 2.6\times10^{-20}\,\mathrm{m^2/W}$ in silica, orders of magnitude smaller in a gas-filled core). One knob, $A_{\mathrm{eff}}$ , sets both extremes. Hollow cores have their own law. The thin glass walls act as anti-reflection layers except at their Fabry–Pérot resonances, which show up as loss peaks; transmission survives in the anti-resonant bands between them.

\lambda_m = \dfrac{2 t}{m}\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{core}}^2}, \qquad m = 1, 2, 3, \dots \qquad (7)

where: $t$ — wall thickness around the hollow core; $m$ — resonance order (loss maxima at $\lambda_m$ ); $n_{\mathrm{core}} \approx 1$ for air or vacuum. Filling the core with gas raises $n_{\mathrm{core}}$ and shifts every window; thicker walls push the windows to longer wavelength. Finally, deliberately breaking the lattice symmetry makes the two polarizations travel at different speeds — birefringence far beyond stress-induced PM fiber.

B = |n_{\mathrm{eff}}^{x} - n_{\mathrm{eff}}^{y}|, \qquad L_B = \dfrac{\lambda}{B} \qquad (8)

where: $B$ — modal birefringence, reaching $\sim10^{-3}$ in asymmetric-hole PCFs; $L_B$ — polarization beat length; $n_{\mathrm{eff}}^{x,y}$ — indices of the two orthogonally polarized fundamental modes.

Derivation AEffective cladding index and the endlessly-single-mode condition

The goal is to explain why a triangular-lattice PCF with small enough holes supports only its fundamental mode at every wavelength — a property with no analogue in conventional fibers. The strategy is to model the cladding by its fundamental space-filling mode (FSM), the highest- $\beta$ mode the lattice supports, whose index is $n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)$ , then track the core-to-cladding contrast as the wavelength shrinks and feed it into the generalized $V_{\mathrm{eff}}$ of equation (3).

Step 1. The FSM is the periodic-cladding analogue of a plane wave; its index lies between the air-diluted average $n_{\mathrm{cl}}^{\mathrm{avg}}$ (long $\lambda$ ) and $n_{\mathrm{glass}}$ (short $\lambda$ ). Why: the most tightly bound cladding state is the one that competes with the core mode, so its index is the fair "cladding index."

Step 2. For a missing-hole core, $n_{\mathrm{core}} = n_{\mathrm{glass}}$ is fixed while the cladding index rises toward it as $\lambda$ falls. Why: this asymmetry — a fixed core index chased by a rising cladding index — makes the contrast collapse, unique to a geometric lattice.

Step 3. Numerical FSM solutions give a quadratic collapse, $n_{\mathrm{glass}}^2 - n_{\mathrm{eff}}^{\mathrm{cl}}(\lambda)^2 \to C\,(\lambda/\Lambda)^2$ as $\lambda \to 0$ , with $C$ set by $d/\Lambda$ . Why: the mode retreats into ever-thinner glass webs, so the residual air perturbation scales as wavelength over feature size.

Step 4. Substitute into equation (3):

V_{\mathrm{eff}} = \dfrac{2\pi\rho}{\lambda}\sqrt{C}\,\dfrac{\lambda}{\Lambda} = 2\pi\rho\,\dfrac{\sqrt{C}}{\Lambda}.

The explicit $1/\lambda$ cancels and $V_{\mathrm{eff}}$ tends to a constant. Why: the $1/\lambda$ growth that drives an ordinary fiber multimode at short $\lambda$ is exactly annulled by the collapsing contrast.

Step 5. If that constant stays below the second-mode cutoff near $2.405$ , no higher-order mode ever guides; numerical solving locates the boundary at $d/\Lambda \approx 0.4$ . Why: below this hole size the fiber is single-mode for all $\lambda$ — endlessly single-mode.

\dfrac{d}{\Lambda} \lesssim 0.4 \;\Rightarrow\; V_{\mathrm{eff}}(\lambda) < 2.405 \;\; \text{for all } \lambda

Verified: reproduces the tabulated boundary $d/\Lambda \approx 0.4$ (Birks, Knight and Russell, Opt. Lett. 22, 961, 1997); consistent with Saleh and Teich, Ch. 9.

Go deeper · Derivation B · 1 minAnti-resonant reflection and hollow-core transmission windows

Derivation BAnti-resonant reflection and hollow-core transmission windows

The goal is to derive the wavelengths at which a hollow-core fiber leaks and the bands in which it guides, treating the thin glass wall around the air core as a Fabry–Pérot etalon. The strategy is to write the transverse wavevector inside the wall, impose the etalon resonance across thickness $t$ , and solve for the resonant wavelengths — which are loss maxima, so guidance sits in the anti-resonant gaps between them. This one formula organizes the transmission spectra of Kagomé, tube-lattice, and nested anti-resonant fibers.

Step 1. A core mode with $n_{\mathrm{eff}} \approx n_{\mathrm{core}} \approx 1$ enters the wall, where the wavevector component across the thickness is $k_\perp^{\mathrm{glass}} = k_0\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{eff}}^2}$ . Why: only the transverse projection "fits" across the wall, so it sets the etalon condition.

Step 2. The wall is resonant when a half-integer number of transverse wavelengths spans it, $k_\perp^{\mathrm{glass}}\, t = m\pi$ . Why: at resonance the membrane turns transparent to leakage — a matched anti-reflection layer — so core light escapes; these are loss peaks.

Step 3. Substitute $k_0 = 2\pi/\lambda$ and $n_{\mathrm{eff}} \approx n_{\mathrm{core}}$ :

\dfrac{2\pi}{\lambda_m}\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{core}}^2}\; t = m\pi \;\Rightarrow\; \lambda_m = \dfrac{2t}{m}\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{core}}^2}.

Why: this ties every loss peak to the one parameter you can measure under a microscope — the wall thickness $t$ .

Step 4. Between consecutive resonances the wall is anti-resonant: it reflects strongly and light stays in the air core. Why: the low-loss windows sit in the gaps, and changing the core gas alters $n_{\mathrm{core}}$ and slides them all. *Verified: matches measured transmission-band centres of anti-resonant hollow-core fibers; standard result in Russell, J. Lightwave Technol. 24, 4729 (2006).*

Go deeper · Derivation C · 1 minGeometry-tuned zero-dispersion wavelength

Derivation CGeometry-tuned zero-dispersion wavelength

The goal is to show that a PCF's total dispersion splits into a fixed material part and a geometry-controlled waveguide part, so the waveguide part can drag the zero-dispersion wavelength from silica's infrared value into the visible — the enabling trick for supercontinuum and combs. The strategy is to start from $D$ as the curvature of the modal index, separate that index into a bulk-glass term and a confinement shift, and let tight confinement in small cores make the waveguide term large and anomalous.

Step 1. From equation (5), $D = -(\lambda/c)\,\mathrm{d}^2 n_{\mathrm{eff}}/\mathrm{d}\lambda^2$ , and $n_{\mathrm{eff}}(\lambda) \approx n_{\mathrm{glass}}(\lambda) + \Delta n_{\mathrm{wg}}(\lambda)$ gives $D_{\mathrm{total}} = D_{\mathrm{mat}} + D_{\mathrm{wg}}$ . Why: this isolates the one term you can change (waveguide) from the one you cannot (material).

Step 2. Bulk silica has $D_{\mathrm{mat}} = 0$ near $\lambda_{\mathrm{ZD}}^{\mathrm{silica}} \approx 1.27\,\mu\mathrm{m}$ , normal below and anomalous above. Why: in an ordinary fiber this fixed crossing is the whole story, forbidding anomalous dispersion in the visible.

Step 3. For small $\Lambda$ and large $d/\Lambda$ the tightly squeezed mode gives a large, anomalous $D_{\mathrm{wg}}$ ; added to the normal $D_{\mathrm{mat}}$ at shorter $\lambda$ , the sum can cross zero below $0.8\,\mu\mathrm{m}$ , so the fiber supports solitons and phase-matched four-wave mixing pumped by a Ti:sapphire laser. Why: a strongly confined mode's index depends sharply on wavelength — the large curvature needed to move the crossing hundreds of nanometres, the basis of octave-spanning supercontinuum and $f$ – $2f$ comb self-referencing.

D_{\mathrm{total}}(\lambda) = D_{\mathrm{mat}}(\lambda) + D_{\mathrm{wg}}(\lambda;\,\Lambda,\, d/\Lambda)

Verified: consistent with measured PCF dispersion curves showing $\lambda_{\mathrm{ZD}}$ in the visible (Ranka, Windeler and Stentz, Opt. Lett. 25, 25, 2000); material $\lambda_{\mathrm{ZD}} \approx 1.27\,\mu\mathrm{m}$ from the silica Sellmeier equation.

Limiting cases

Case	Result	Meaning
Long-wavelength (coarse) limit	$\lambda \gg \Lambda \Rightarrow n_{\mathrm{eff}}^{\mathrm{cl}} \to n_{\mathrm{cl}}^{\mathrm{avg}} = \sqrt{f\,n_{\mathrm{air}}^2 + (1-f)\,n_{\mathrm{glass}}^2}$	Mode averages over holes; cladding is a homogeneous Maxwell–Garnett medium of air fraction $f$ .
Short-wavelength limit	$\lambda \ll \Lambda \Rightarrow n_{\mathrm{eff}}^{\mathrm{cl}} \to n_{\mathrm{glass}}$	Field retreats into the glass webs; contrast collapses as $\lambda^2$ and $V_{\mathrm{eff}}$ stays bounded.
Endlessly single-mode	$d/\Lambda \lesssim 0.4$	Saturated $V_{\mathrm{eff}}$ never reaches the higher-order cutoff; only the fundamental mode guides at every $\lambda$ .
Anti-resonant window	$\lambda_m \approx \dfrac{2t}{m}\sqrt{n_{\mathrm{glass}}^2 - 1}$ , $m = 1,2,\dots$	Loss peaks set by wall thickness $t$ ; low-loss guidance in the gaps between them.
Band-gap scattering floor	$\alpha_{\mathrm{HC}} \propto \dfrac{\lambda^3}{d_{\mathrm{core}}^4}$	Loss dominated by roughness scattering off the core wall; larger cores and longer $\lambda$ lower it, at multimode risk.

Common trapHollow-core fibers guide by total internal reflection

They cannot. A hollow air core has a lower index than the surrounding glass, so total internal reflection is geometrically impossible — its required condition ( $n_{\mathrm{core}} > n_{\mathrm{clad}}$ ) is reversed. Confinement comes instead from a photonic band gap, where the cladding lattice has no propagating mode at the core light's frequency, or from anti-resonant reflection off the thin walls; either way it works only inside discrete windows, and outside them the fiber leaks strongly. Two related traps: "endlessly single-mode" does not mean lossless or dispersion-free — it refers only to the absence of higher-order guided modes; and the volume-averaged cladding index of equation (4) is a long-wavelength lower bound, not the true confining index, which rises as the wavelength shrinks.

Worked examples

1 · Air fill fraction and cladding index of an endlessly-single-mode PCF (confidence builder)

A solid-core PCF sits at the endlessly-single-mode boundary, $d/\Lambda = 0.4$ , in silica ( $n_{\mathrm{glass}} = 1.444$ ). Find the air fill fraction, the long-wavelength effective cladding index, and the upper bound on numerical aperture.

Step 1. Air fill fraction from equation (4): $f = \dfrac{\pi}{2\sqrt{3}}(0.4)^2 = 0.9069 \times 0.16 = \mathbf{0.145}$ . Even at the ESM boundary only about $14.5\%$ of the cladding is air.

Step 2. Long-wavelength cladding index: $n_{\mathrm{cl}}^{\mathrm{avg}} = \sqrt{0.145 + 0.855 \times 2.0851} = \sqrt{1.928} = \mathbf{1.388}$ — air dilutes glass from $1.444$ to about $1.388$ in the coarse limit.

Step 3. NA upper bound: $\mathrm{NA} = \sqrt{n_{\mathrm{glass}}^2 - (n_{\mathrm{cl}}^{\mathrm{avg}})^2} = \sqrt{2.0851 - 1.9277} = \sqrt{0.1574} = \mathbf{0.40}$ . This is an upper bound — the true confining index is the larger space-filling-mode index, so the real NA is smaller and falls with wavelength, the very effect behind endlessly single-mode operation. f = 0.145, n_cl^avg = 1.388, NA ≲ 0.40.

2 · Anti-resonant transmission windows of a hollow-core fiber

A tube-lattice hollow-core fiber has silica walls of thickness $t = 0.45\,\mu\mathrm{m}$ ( $n_{\mathrm{glass}} = 1.45$ ) around an air core ( $n_{\mathrm{core}} = 1$ ). Locate the first three loss resonances and identify which pump lines fall in a low-loss window.

Step 1. Index factor: $\sqrt{n_{\mathrm{glass}}^2 - 1} = \sqrt{1.45^2 - 1} = \sqrt{1.1025} = 1.050$ .

Step 2. Resonant (loss-peak) wavelengths from equation (7): $\lambda_1 = 2(0.45)(1.050) = 0.945\,\mu\mathrm{m}$ ; $\lambda_2 = \lambda_1/2 = 0.473\,\mu\mathrm{m}$ ; $\lambda_3 = \lambda_1/3 = 0.315\,\mu\mathrm{m}$ . The wall leaks at 945 nm, 473 nm, and 315 nm.

Step 3. Low-loss guidance lives between resonances. The fundamental window is everything above $\lambda_1 = 945\,\mathrm{nm}$ , so $1064\,\mathrm{nm}$ and $1550\,\mathrm{nm}$ sit in a low-loss band, while an $800\,\mathrm{nm}$ Ti:sapphire line falls in the window between $\lambda_2$ ( $473\,\mathrm{nm}$ ) and $\lambda_1$ ( $945\,\mathrm{nm}$ ). Thicken the walls to shift windows up, or fill the core with gas to shift them down. 1064 nm and 1550 nm are in the fundamental window; 800 nm is in the m = 1–2 window.

3 · Nonlinear coefficient: highly nonlinear versus large-mode-area

Two silica PCFs run at $\lambda = 1.064\,\mu\mathrm{m}$ , $n_2 = 2.6\times10^{-20}\,\mathrm{m^2/W}$ : one highly nonlinear ( $A_{\mathrm{eff}} = 1.5\,\mu\mathrm{m}^2$ ), one large-mode-area ( $A_{\mathrm{eff}} = 1000\,\mu\mathrm{m}^2$ ). Compare $\gamma$ .

Step 1. From equation (6): $\gamma = \dfrac{2\pi (2.6\times10^{-20})}{(1.064\times10^{-6})(1.5\times10^{-12})} = 0.102\,\mathrm{W^{-1}m^{-1}} = \mathbf{102\,W^{-1}km^{-1}}$ for the small core — soliton and supercontinuum territory.

Step 2. Same formula with $A_{\mathrm{eff}} = 1000\,\mu\mathrm{m}^2$ gives $\gamma = \mathbf{0.15\,W^{-1}km^{-1}}$ . The ratio $102/0.15 \approx 670$ equals the mode-area ratio, since $\gamma \propto 1/A_{\mathrm{eff}}$ : one knob spans three orders of magnitude — small for nonlinear conversion, large to suppress nonlinear damage. γ = 102 vs 0.15 W⁻¹km⁻¹, ≈670× set by mode area.

Physics of the portrait

The photonic-crystal fiber is a rare case of a whole subfield tracing to one sustained idea: that you could guide light not by the chemistry of a doped core but by the geometry of a photonic crystal wrapped around it.

Origins · 1 minPhilip St. J. Russell and how this was discovered

Philip St. J. Russell

1953– · inventor of the photonic-crystal fiber

Philip St. John Russell conceived the photonic-crystal fiber around 1991, reasoning from the then-new physics of photonic crystals: if a periodic dielectric lattice could forbid propagation in a band of frequencies, a fiber whose cladding was such a lattice could trap light in a defect — even a hollow one — by a photonic band gap rather than total internal reflection. The fabrication was formidable: stacking hundreds of silica capillaries into a preform and drawing it down so the air channels survived, scaled and intact, into a hair-thin fiber. With Jonathan Knight, Tim Birks and Philip Atkin at Bath and Southampton, Russell drew the first working solid-core PCF in 1996 (Opt. Lett. 21, 1547) and immediately found the property no one had predicted: for small holes it was endlessly single-mode (Opt. Lett. 22, 961, 1997). Hollow-core band-gap guidance — light running in air — followed in 1999 (Cregan et al., Science 285, 1537). Within a few years the platform had upended nonlinear fiber optics, feeding the octave-spanning supercontinuum and the self-referenced frequency comb. Russell's insight turned the fiber cladding from a passive matrix into the active design surface of the whole device.

History

1991

— Russell conceives guiding light by a photonic band gap in a microstructured cladding

1996

— Knight, Birks, Russell and Atkin draw the first working solid-core photonic-crystal fiber (Opt. Lett. 21, 1547)

1997

— Birks, Knight and Russell identify the endlessly-single-mode regime d/Λ ≲ 0.4 (Opt. Lett. 22, 961)

1999

— Cregan and co-workers demonstrate hollow-core photonic-band-gap guidance of light in air (Science 285, 1537)

2000

— Ranka, Windeler and Stentz generate octave-spanning supercontinuum in a visible-dispersion PCF (Opt. Lett. 25, 25)

Going further

Texts. Saleh and Teich, Fundamentals of Photonics, 3rd ed., Ch. 9 (Fiber Optics) and Ch. 7 (Photonic Crystals) · Hecht, Optics, 5th ed., Ch. 5 · P. St. J. Russell, "Photonic Crystal Fibers," Science 299, 358 (2003) · P. St. J. Russell, "Photonic-Crystal Fibers," J. Lightwave Technol. 24, 4729 (2006) · J. C. Knight, "Photonic crystal fibres," Nature 424, 847 (2003) · Demtröder, Laser Spectroscopy · Loudon, The Quantum Theory of Light.

Provenance. A verifier should confirm: (1) the transverse modal wave equation from the vector Helmholtz equation for a $z$ -invariant dielectric (Saleh and Teich, Ch. 9); (2) the triangular-lattice fill fraction $f = (\pi/2\sqrt{3})(d/\Lambda)^2$ and volume-averaged cladding index; (3) the endlessly-single-mode boundary $d/\Lambda \approx 0.4$ (Birks, Knight and Russell, 1997); (4) the anti-resonance formula $\lambda_m = (2t/m)\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{core}}^2}$ (Russell, 2006); (5) silica material $\lambda_{\mathrm{ZD}} \approx 1.27\,\mu\mathrm{m}$ ; (6) the definitions of $D$ , $A_{\mathrm{eff}}$ , and $\gamma$ . The worked-example values ( $f = 0.145$ , $n_{\mathrm{cl}}^{\mathrm{avg}} = 1.388$ ; $\lambda_{1,2,3} = 945, 473, 315\,\mathrm{nm}$ ; $\gamma = 102$ vs $0.15\,\mathrm{W^{-1}km^{-1}}$ ) are recomputed above; silica constants ( $n \approx 1.444$ , $n_2 \approx 2.6\times10^{-20}\,\mathrm{m^2/W}$ , LP $_{11}$ cutoff $\approx 2.405$ ) are standard.

Prerequisites. Total Internal Reflection and Optical Guiding · Step-Index and Graded-Index Optical Fibers · Waveguide Modes and the Wave Equation · Numerical Aperture and Acceptance Angle. Leads to. Chromatic and Modal Dispersion in Fibers · Dispersion Management and Compensation · Fiber Attenuation and Loss Mechanisms · Polarization-Mode Dispersion and Birefringent Fibers · Single-Mode and Multimode Operation.

Ask the tutor: Want to see how changing $d/\Lambda$ slides a real PCF from endlessly single-mode into multimode, or how wall thickness places a hollow-core fiber's transmission windows on a chosen laser line?

What actually makes a PCF different from a step-index fiber?

The cladding. In a step-index fiber both core and cladding are solid glass and the tiny index step ( $\sim0.3\%$ ) is set by dopant chemistry. A PCF's cladding is a lattice of air holes, so its optical response is a designed composite set by two geometric numbers, $\Lambda$ and $d/\Lambda$ . That single change unlocks the rest: a wavelength-dependent cladding index (endlessly single-mode behaviour), a large tunable waveguide dispersion, mode areas across three orders of magnitude, and — with a hollow core — guidance in air by a band gap. Same silica, radically different fiber, because the cladding is structured rather than doped.

Why does the effective cladding index depend on wavelength?

Because the mode's size relative to the holes changes. At long wavelengths the field spreads over many holes and feels their volume average — the low Maxwell–Garnett index of air-diluted glass. At short wavelengths it contracts into the glass webs between holes, sampling almost pure glass, so the cladding index climbs toward $n_{\mathrm{glass}}$ . This is geometric, not material dispersion (silica's own index barely moves here). That rising cladding index shrinks the core-cladding contrast as $\lambda$ falls, saturating the normalized frequency and producing endlessly single-mode operation.

Is "endlessly single-mode" the same as low-loss or dispersion-free?

No — it is a strictly modal statement. It means the fiber supports only its fundamental mode at every wavelength: no second guided mode ever turns on, however far into the blue you go. It says nothing about attenuation, still set by absorption, leakage, and scattering, nor about dispersion, which remains fully present and is one of the main things you engineer in a PCF. Such a fiber still has chromatic dispersion, finite loss, and can be highly nonlinear. The property guarantees clean single-transverse-mode output over an enormous bandwidth — not freedom from the other imperfections.

How can most of the light travel in air in a hollow-core fiber?

By making the air core a defect in a cladding that forbids light at the operating frequency. In a band-gap fiber the lattice has a photonic band gap — a range of frequency and propagation constant with no cladding mode — so core light has nowhere in the glass to couple and is reflected back. In an anti-resonant fiber the thin walls are off-resonance and act as good mirrors. Either way the guided field peaks in the air, so hollow-core fibers have very low nonlinearity and high damage thresholds, and filling the core with a chosen gas lets light and gas interact over many metres.

Why do anti-resonant transmission windows come in discrete bands?

Because the thin glass walls behave as Fabry–Pérot etalons. Where a half-integer number of transverse wavelengths fits across a wall of thickness $t$ — that is, $\lambda_m = (2t/m)\sqrt{n_{\mathrm{glass}}^2 - n_{\mathrm{core}}^2}$ — the wall is resonant, transparent to leakage, and the fiber loses light: these are loss peaks. Between consecutive resonances the wall is anti-resonant, reflecting strongly, so light stays in the air core: these gaps are the low-loss windows. The spectrum is banded, with positions set by the wall thickness. Thicker walls push the bands to longer wavelength; filling the core with gas slides them all.

What sets the nonlinearity of a PCF, and why can it be so extreme?

The effective mode area, through $\gamma = 2\pi n_2/(\lambda A_{\mathrm{eff}})$ . Nonlinear effects scale with intensity — power per unit area — so squeezing the same power into a smaller mode drives them harder. A small-core, high-air-fill PCF pushes $A_{\mathrm{eff}}$ below $1\,\mu\mathrm{m}^2$ , giving $\gamma$ around $100\,\mathrm{W^{-1}km^{-1}}$ , hundreds of times a standard fiber — enough for centimetre-scale supercontinuum. The same knob reversed gives large-mode-area PCFs over $1000\,\mu\mathrm{m}^2$ with vanishing $\gamma$ , for high-power delivery where you must avoid nonlinear damage; a gas-filled hollow core drops $n_2$ itself by orders of magnitude.