Transverse Cavity Modes (Hermite–Gaussian)

A stable laser resonator is a trap for light, and like every trap it has a spectrum. The obvious part is longitudinal — the ladder of axial frequencies spaced by the free spectral range. But a cavity also stamps a shape onto the light across the beam, and those transverse shapes are quantized just as sharply as the frequencies. This article borrows its mathematical skeleton from quantum mechanics: the paraxial wave equation for a beam on a Gaussian envelope is, term for term, the Schrödinger equation of a two-dimensional harmonic oscillator. So the self-reproducing transverse patterns are Hermite–Gaussian functions — Hermite polynomials times a Gaussian — labelled by two integers $m,n$ that play the role of oscillator quantum numbers. The lowest state, $\mathrm{TEM}_{00}$ , is the ordinary Gaussian beam: a single bright spot, the tightest the geometry allows. Climbing the ladder adds dark nodal lines — $m$ across $x$ , $n$ across $y$ — producing the lobed burn-paper patterns every laser technician recognizes. Every member shares one complex beam parameter $q$ , hence the same waist, Rayleigh range, and curvature; what changes with $m,n$ is the footprint, the lobe count, and, most consequentially, the Gouy phase, which grows as $(m+n+1)$ . That extra phase is the whole physics of transverse-mode splitting: two modes of equal axial index but different $m+n$ resonate at slightly different frequencies, set by the resonator's $g$ -parameters. And because higher-order modes spread wider and lose more to any aperture, a millimetre iris can filter them away and force clean single-mode output — the reason a well-built laser lases on one pattern out of the infinite ladder available.

Key takeaways

Transverse cavity modes are the stationary field patterns a stable resonator supports across the beam; for rectangular symmetry they are the Hermite–Gaussian functions $\mathrm{TEM}_{mn}$ , indexed by node counts $m,n$ .
All modes of one cavity share a single $q$ -parameter, so waist, Rayleigh range and curvature $R(z)$ are common — only the Hermite shape and the Gouy phase differ.
Higher orders carry an enhanced Gouy phase $(m+n+1)\psi(z)$ , shifting their frequencies by $\tfrac{(m+n+1)}{\pi}\arccos(\pm\sqrt{g_1g_2})$ of a free spectral range.
Higher orders are wider ( $\propto\sqrt{2m+1}$ ) and lose more to any aperture, so an intracavity iris suppresses them and enforces $\mathrm{TEM}_{00}$ with $M^2=1$ .
In cylindrical symmetry the natural basis is Laguerre–Gaussian, giving rings and, for $l\neq0$ , orbital angular momentum $l\hbar$ per photon.
Key idea: the transverse eigenmodes of a stable resonator are Hermite–Gaussian functions $\mathrm{TEM}_{mn}$ — one shared Gaussian $q$ -parameter dressed with Hermite polynomials $H_m,H_n$ and an order-dependent Gouy phase $(m+n+1)\psi(z)$ that fixes both their lobe shapes and their split resonant frequencies.

The core insight

The transverse eigenmodes of a stable resonator are Hermite–Gaussian functions $\mathrm{TEM}_{mn}$ : a common Gaussian $q$ -parameter dressed with Hermite polynomials $H_m,H_n$ and an order-dependent Gouy phase $(m+n+1)\psi(z)$ , which sets both their spot shapes and their split resonant frequencies.

Where this lives today

Nowhere is transverse-mode purity more demanded than in gravitational-wave detection. LIGO and Virgo circulate hundreds of kilowatts of $\mathrm{TEM}_{00}$ , and any admixture of higher-order modes — from mirror figure errors, thermal lensing, or misalignment — spoils the fringe contrast needed to sense a strain of $10^{-21}$ . The fix is a chain of "mode cleaner" ring cavities tuned so $\mathrm{TEM}_{00}$ resonates while every higher-order mode, sitting its Gouy-shifted $\arccos(\sqrt{g_1g_2})/\pi$ of an FSR away, is rejected off-resonance — the very splitting derived above put to work. The same physics powers a wide sweep of modern optics. Frequency-comb and cavity-QED experiments lock lasers to high-finesse cavities and must track transverse-mode families to avoid the wrong resonance; laser machining and eye surgery insist on $M^2\to1$ beams because focal spot scales as $\sqrt{2m+1}$ . The cylindrical cousins are tools in their own right: the Laguerre–Gaussian "doughnut" supplies the central null STED microscopy uses to beat the diffraction limit, its orbital angular momentum drives optical spanners, and mode-division multiplexing packs independent channels onto orthogonal modes down one fibre — each an application of the fact that a resonator quantizes light not just in frequency but in shape.

The physical picture

Treat the cavity as a periodic lens waveguide: every round trip refocuses the light, and only a field pattern that reproduces its own transverse shape under this endless refocusing can survive. Those survivors are the Hermite–Gaussian modes. The fundamental $\mathrm{TEM}_{00}$ is the single tightest self-consistent spot the geometry permits; the higher modes are its overtones, the standing "vibrations" of the same transverse resonator, each with more wiggles and a wider footprint. The Hermite polynomials are not pulled from a hat — they are forced by the same mathematics that governs the quantum harmonic oscillator. Strip the propagating Gaussian off the paraxial equation and what remains for the transverse structure is Hermite's differential equation, identical to the oscillator eigenvalue problem. The index $m$ is the oscillator quantum number counting nodes, and the Gouy phase $(m+n+1)\psi$ is the direct optical analogue of the oscillator energy $(N+1)\hbar\omega$ , zero-point term and all.

Figure

Intensity patterns of the first Hermite–Gaussian modes: TEM₀₀ is a single Gaussian spot, and each unit of m or n adds one dark nodal line, so TEM₂₁ shows a 3 × 2 grid of six lobes, while every pattern shares the same underlying Gaussian envelope and q-parameter

Higher-order modes cost more to sustain, and that is the practical heart of the story. They spread wider, overlapping a finite gain volume and any aperture less efficiently and diffracting more strongly, so a well-placed millimetre iris acts as a transverse-mode filter, bleeding round-trip loss into them while barely touching $\mathrm{TEM}_{00}$ . The extra Gouy phase is the other half: because each family accumulates a different phase per round trip, each resonates at a slightly different frequency, so even when many patterns fit under the gain curve they compete at distinct frequencies and loss discrimination picks the winner.

Figure

Resonance spectrum of a symmetric cavity with g₁ = g₂ = 0.5 and length 1.0 m: axial modes are spaced by the free spectral range Δν = c/2L = 150 MHz, while each unit of m+n adds a transverse offset of 50 MHz, one-third of the FSR, so the third transverse family of order q lands exactly on the fundamental of order q+1

The core equations

The single object that carries everything is the Hermite–Gaussian field: a Gaussian beam multiplied by two Hermite polynomials and given an enhanced Gouy phase. It solves the paraxial wave equation and reproduces its transverse shape under propagation, changing only scale and phase — read it as "the fundamental Gaussian, with $m$ nodes across $x$ and $n$ across $y$ ."

u_{mn}(x,y,z)=A_{mn}\,\frac{w_0}{w(z)}\,H_m\!\Big(\frac{\sqrt2\,x}{w(z)}\Big)H_n\!\Big(\frac{\sqrt2\,y}{w(z)}\Big)\exp\!\Big[-\frac{x^2+y^2}{w^2(z)}\Big]\exp\!\Big[-ik\frac{x^2+y^2}{2R(z)}\Big]\exp\!\big[i(m+n+1)\psi(z)\big] \qquad (1)

where: $m,n$ — transverse mode indices (nodes along $x$ and $y$ ); $H_m$ — Hermite polynomial of order $m$ ; $w(z)$ — $1/e^2$ intensity spot size of the shared Gaussian; $w_0$ — beam waist (minimum spot size); $R(z)$ — wavefront radius of curvature; $\psi(z)$ — Gouy phase; $k=2\pi/\lambda$ — wavenumber; $A_{mn}$ — normalization. The three exponentials are the Gaussian envelope, the spherical-wavefront phase, and the mode-dependent Gouy phase that splits the resonances. The Gaussian factors $w(z)$ , $z_R$ and $R(z)$ carry no dependence on $m$ or $n$ : they are fixed by one complex beam parameter $q(z)$ , itself fixed by the resonator geometry. Every $\mathrm{TEM}_{mn}$ of a given cavity therefore rides on the identical Gaussian scaffold; only the polynomials and the Gouy factor change with mode order.

w(z)=w_0\sqrt{1+\Big(\frac{z}{z_R}\Big)^2},\qquad z_R=\frac{\pi w_0^2}{\lambda},\qquad R(z)=z\Big[1+\Big(\frac{z_R}{z}\Big)^2\Big] \qquad (2)

where: $z_R$ — Rayleigh range; $\lambda$ — wavelength in the medium; $z$ — axial distance from the waist. The single line $1/q = 1/R - i\lambda/(\pi w^2)$ packages $R(z)$ and $w(z)$ together, which is why sharing $q$ means sharing all of these. The Gouy phase is the extra longitudinal phase a focused beam accumulates through its waist, beyond the plane-wave term $kz$ . For a higher-order mode it is enhanced by the integer factor $(m+n+1)$ — the analogue of the oscillator's zero-point-plus-quanta count, and the origin of transverse-mode splitting.

\psi(z)=\arctan\!\Big(\frac{z}{z_R}\Big),\qquad \Phi_{mn}(z)=(m+n+1)\,\psi(z) \qquad (3)

where: $\Phi_{mn}$ — total mode-order-dependent longitudinal phase; the " $+1$ " is the fundamental Gouy phase of $\mathrm{TEM}_{00}$ itself; each extra unit of $m+n$ adds one full $\psi(z)$ of accumulated phase across the beam. Feeding the enhanced Gouy phase into the round-trip resonance condition gives the frequencies: the axial index $q$ sets the coarse comb, and the transverse indices add a fractional offset governed by the $g$ -parameters through an arccosine of $\sqrt{g_1g_2}$ .

\nu_{qmn}=\frac{c}{2L}\left[q+\frac{(m+n+1)}{\pi}\arccos\!\big(\pm\sqrt{g_1 g_2}\big)\right] \qquad (4)

where: $\nu_{qmn}$ — resonant frequency of mode $(q,m,n)$ ; $q$ — longitudinal (axial) index; $L$ — cavity length; $c$ — speed of light in the medium; $g_i=1-L/R_i$ — resonator $g$ -parameters ( $R_i$ the mirror radii). The sign selects the branch appropriate to the stability region $0\le g_1g_2\le1$ . Subtracting adjacent transverse orders isolates the frequency step per unit of $m+n$ . Because it is $\arccos(\pm\sqrt{g_1g_2})/\pi$ of an FSR — a generally irrational fraction — the transverse comb rarely lines up with the axial comb, which is what lets a laser sort them out.

\Delta\nu_{\mathrm{trans}}=\frac{\Delta\nu_{\mathrm{FSR}}}{\pi}\,\arccos\!\big(\pm\sqrt{g_1 g_2}\big),\qquad \Delta\nu_{\mathrm{FSR}}=\frac{c}{2L} \qquad (5)

where: $\Delta\nu_{\mathrm{trans}}$ — frequency step per unit increase in $m+n$ ; $\Delta\nu_{\mathrm{FSR}}$ — longitudinal free spectral range. Special geometries make the fraction rational: confocal gives exactly $\tfrac12$ , and the $g=0.5$ symmetric cavity gives exactly $\tfrac13$ . Mode width and beam quality follow from the same Hermite structure: the second-moment radius grows as $\sqrt{2m+1}$ per axis, and the size-divergence product $M^2$ inherits that scaling exactly.

M^2_x=2m+1,\qquad M^2_y=2n+1,\qquad w_{\text{eff},x}=w(z)\sqrt{2m+1} \qquad (6)

where: $M^2$ — beam-propagation (times-diffraction-limit) factor, equal to 1 only for $\mathrm{TEM}_{00}$ ; $w_{\text{eff},x}$ — the effective second-moment spot size that a real focus and divergence obey. A larger $M^2$ means a bigger focal spot and faster divergence, in exact proportion.

Derivation AHermite–Gaussian modes from the paraxial wave equation

The goal is to show that the transverse eigenmodes are not a lucky guess but a forced consequence of the paraxial wave equation. We start from the slowly-varying-envelope equation, factor out the fundamental Gaussian carried by $q(z)$ , and demand that whatever transverse structure remains keep a fixed shape as the beam propagates. That single demand collapses the problem onto Hermite's differential equation — the equation of the quantum harmonic oscillator — fixing the polynomials, the integer indices, and the $(m+n+1)$ Gouy phase all at once.

Step 1. Write $E=U(x,y,z)\,e^{-ikz}$ and drop the second $z$ -derivative of the slowly varying envelope to get the paraxial equation $\nabla_\perp^2 U - 2ik\,\partial_z U = 0$ . Why: separating off the fast carrier $e^{-ikz}$ leaves an equation whose solutions are the beam shapes — the arena where cavity modes live.

Step 2. Insert the ansatz $U=\exp[-i(P(z)+k(x^2+y^2)/2q(z))]\,g(\sqrt2 x/w)\,h(\sqrt2 y/w)$ : a fundamental Gaussian (through $q$ ) and a longitudinal phase $P(z)$ times transverse structure functions $g,h$ . Why: the Gaussian already solves $\mathrm{TEM}_{00}$ , so peeling it off isolates the extra structure that distinguishes a higher-order mode.

Step 3. Requiring the transverse profile to be $z$ -independent forces $g'' - 2\xi g' + 2m\,g = 0$ with $\xi=\sqrt2 x/w$ — Hermite's equation, with bounded solutions only for integer $m$ , giving $g=H_m(\xi)$ and likewise $h=H_n$ . Why: this is exactly the 1D quantum-oscillator eigenvalue problem, so integer node counts and Hermite polynomials appear for the identical reason they do for oscillator levels.

Step 4. Matching the longitudinal phase fixes $P(z)=-i\ln[w_0/w(z)]-(m+n+1)\psi(z)$ , so the amplitude renormalizes as $w_0/w$ and the phase carries the integer factor $(m+n+1)$ . Why: consistency between the $q$ -evolution and the polynomial order generates the enhanced Gouy phase — the seed of transverse-mode splitting.

Step 5. Assembling the $w_0/w$ amplitude, the two Hermite polynomials, the Gaussian, the curvature phase, and the enhanced Gouy phase recovers Eq. (1). Why: every factor now traces to a step, so the boxed mode is the unique separable, shape-preserving solution family. *Verified: reproduces Saleh and Teich Eq. (3.3-8) and Siegman Ch. 16; the harmonic-oscillator correspondence gives the same Hermite structure.*

Go deeper · Derivation B · 1 minTransverse-mode splitting from the round-trip Gouy phase

Derivation BTransverse-mode splitting from the round-trip Gouy phase

The goal is to turn the enhanced Gouy phase into a measurable frequency splitting by imposing the resonance condition of a two-mirror cavity. A field is a mode only if it reproduces itself after one round trip, so its total accumulated phase must be an integer multiple of $2\pi$ . The plane-wave part gives the axial comb; the Gouy part, depending on $m+n$ , pushes each transverse family off the comb by an amount set purely by the geometry. Carrying this through yields Eq. (4) and, at confocal, the textbook half-FSR spacing.

Step 1. Demand round-trip self-consistency: $2kL - 2(m+n+1)\Delta\psi = 2\pi q$ , with $2kL$ the plane-wave phase and $2(m+n+1)\Delta\psi$ twice the one-way Gouy accumulation. Why: only fields whose phase closes on a multiple of $2\pi$ interfere constructively after a round trip, so this is the definition of a resonant mode.

Step 2. Evaluate the one-way Gouy accumulation between the mirrors: $\Delta\psi=\psi(z_2)-\psi(z_1)=\arccos(\pm\sqrt{g_1g_2})$ , sign fixed by the stability branch. Why: this compresses all the geometry into a single angle, so the transverse offset needs no explicit waist calculation.

Step 3. Substitute $k=2\pi\nu/c$ and solve for $\nu$ : $\nu_{qmn}=\tfrac{c}{2L}[q+\tfrac{(m+n+1)}{\pi}\arccos(\pm\sqrt{g_1g_2})]$ , with $c/2L=\Delta\nu_{\mathrm{FSR}}$ . Why: $q$ builds the coarse comb of spacing $c/2L$ , while the Gouy term adds a fixed fractional shift per unit of $m+n$ — the splitting made explicit.

Step 4. Check confocal $g_1=g_2=0$ : $\arccos 0=\pi/2$ , so $\Delta\nu_{\mathrm{trans}}=\tfrac12\Delta\nu_{\mathrm{FSR}}$ . Why: recovering the tabulated half-FSR degeneracy checks that the sign, the factor of two, and the arccosine are all correct. *Verified: matches Siegman Ch. 19 and Demtröder Ch. 5; the confocal $\tfrac12$ -FSR spacing is textbook-standard, and $g=0.5$ gives the $\tfrac13$ -FSR value used in the figures.*

Limiting cases

Case	Result	Meaning
Fundamental $m=n=0$	$u_{00}\propto\frac{w_0}{w(z)}e^{-(x^2+y^2)/w^2(z)}$	The plain Gaussian $\mathrm{TEM}_{00}$ : no nodes, minimum spot, minimum divergence, $M^2=1$ , Gouy phase just $\psi(z)$ .
Cylindrical symmetry	$\mathrm{TEM}_{mn}\longleftrightarrow\mathrm{LG}_{pl}$	Round apertures make Laguerre–Gaussian the natural basis: rings labelled by radial $p$ , azimuthal $l$ , with $2p+\lvert l\rvert=m+n$ sharing one Gouy phase and resonance.
Confocal resonator	$g_1=g_2=0\Rightarrow\Delta\nu_{\mathrm{trans}}=\tfrac12\Delta\nu_{\mathrm{FSR}}$	Transverse modes sit exactly halfway between axial modes; even and odd $m+n$ families collapse onto two combs — highly degenerate, hard to run single-mode.
Symmetric $g=0.5$	$\Delta\nu_{\mathrm{trans}}=\tfrac13\Delta\nu_{\mathrm{FSR}}$	A commensurate "re-entrant" case: every third transverse family lands on an axial mode, the geometry drawn in the second figure.
Large Fresnel number	$N_F=\frac{a^2}{\lambda L}\gg1$	Many transverse orders become nearly loss-degenerate; diffraction discrimination between them is weak and clean single-mode selection is difficult.

Common trapTransverse-mode splitting is not the same thing as longitudinal mode spacing, and higher orders do not move the waist

Two errors travel together. First, the indices $m,n$ do not change the waist location, the Rayleigh range, or the curvature $R(z)$ — all locked to the shared $q$ -parameter, which depends only on the geometry. What $m,n$ change is the footprint ( $\propto\sqrt{2m+1}$ ), the lobe count, and the Gouy phase. Second, transverse-mode splitting $\Delta\nu_{\mathrm{trans}}=(\Delta\nu_{\mathrm{FSR}}/\pi)\arccos(\pm\sqrt{g_1g_2})$ is a different frequency scale from the longitudinal FSR $\Delta\nu_{\mathrm{FSR}}=c/2L$ . Except in special geometries like confocal ( $\tfrac12$ ) or $g=0.5$ ( $\tfrac13$ ), the two are incommensurate — a transverse resonance rarely coincides with an axial one, which is what lets an iris plus mode competition isolate a clean $\mathrm{TEM}_{00}$ .

Worked examples

1 · Transverse-mode spacing of a real symmetric cavity (confidence builder)

A symmetric resonator has length $L=1.0\ \mathrm{m}$ and identical concave mirrors $R_1=R_2=2.0\ \mathrm{m}$ , at $\lambda=632.8\ \mathrm{nm}$ (HeNe). Find the free spectral range, the transverse-mode spacing, and where $\mathrm{TEM}_{01}$ sits relative to $\mathrm{TEM}_{00}$ .

Step 1. The $g$ -parameters are $g_1=g_2=1-L/R=1-1.0/2.0=0.5$ , so $\sqrt{g_1g_2}=0.5$ (inside the stability region) and the Gouy angle per pass is $\arccos(0.5)=\pi/3=1.0472\ \mathrm{rad}$ .

Step 2. The free spectral range is $\Delta\nu_{\mathrm{FSR}}=c/2L=(2.998\times10^8)/2.0=1.499\times10^8\ \mathrm{Hz}\approx\mathbf{150\ MHz}$ — the coarse axial spacing.

Step 3. The transverse step is $\Delta\nu_{\mathrm{trans}}=(\Delta\nu_{\mathrm{FSR}}/\pi)\arccos(0.5)=149.9\times(1/3)\approx\mathbf{50.0\ MHz}$ , so $\mathrm{TEM}_{01}$ ( $m+n=1$ ) sits $50\ \mathrm{MHz}$ above $\mathrm{TEM}_{00}$ at the same $q$ .

Step 4. Note the coincidence: since $\arccos(0.5)/\pi=1/3$ exactly, the spacing is a rational $\tfrac13$ of the FSR, so the $m+n=3$ family lands on the next axial mode ( $3\times50=150\ \mathrm{MHz}$ ) — a special commensurate "re-entrant" geometry, unlike the generic irrational case. Δν_FSR = 150 MHz, Δν_trans = 50 MHz = FSR/3, and TEM₀₁ sits 50 MHz above TEM₀₀.

2 · Lobe count, footprint, divergence, and beam quality of a higher-order mode

A Nd:YAG laser at $\lambda=1064\ \mathrm{nm}$ has fundamental waist $w_0=0.30\ \mathrm{mm}$ . If forced onto $\mathrm{TEM}_{21}$ , how many lobes appear, how much wider is the footprint per axis, and what are the far-field divergence half-angles?

Step 1. Lobe count is $(m+1)(n+1)=(3)(2)=\mathbf{6\ lobes}$ , 3 columns by 2 rows separated by 2 vertical and 1 horizontal nodal lines — the fourth panel of the first figure.

Step 2. The second-moment radius grows as $\sqrt{2m+1}$ : $\sqrt5=\mathbf{2.24\times}$ wider in $x$ and $\sqrt3=\mathbf{1.73\times}$ in $y$ , so $M^2_x=5$ , $M^2_y=3$ .

Step 3. The fundamental far-field half-angle is $\theta_0=\lambda/(\pi w_0)=(1064\times10^{-9})/(\pi\times0.30\times10^{-3})=\mathbf{1.13\ mrad}$ .

Step 4. A higher-order mode diverges $M=\sqrt{M^2}$ times faster: $\theta_x=\sqrt5\,\theta_0=2.52\ \mathrm{mrad}$ and $\theta_y=\sqrt3\,\theta_0=1.96\ \mathrm{mrad}$ . It focuses to a spot $\sqrt5\approx2.24$ times larger in $x$ than a clean $\mathrm{TEM}_{00}$ — which is why tight-focusing applications insist on the fundamental. 6 lobes; footprint ×2.24 (x) and ×1.73 (y); divergence 2.52 and 1.96 mrad versus 1.13 mrad fundamental.

3 · Mode degeneracy in a confocal cavity

A confocal resonator has $L=0.30\ \mathrm{m}$ with $R_1=R_2=L$ . Show that its transverse spectrum collapses to two interleaved combs, and give the effective line spacing.

Step 1. Confocal means $g_1=g_2=1-L/R=0$ , so $\arccos(0)=\pi/2$ and $\Delta\nu_{\mathrm{trans}}=\tfrac12\Delta\nu_{\mathrm{FSR}}$ .

Step 2. The FSR is $\Delta\nu_{\mathrm{FSR}}=c/2L=(2.998\times10^8)/0.60\approx\mathbf{500\ MHz}$ , so each unit of $m+n$ shifts the resonance by exactly $\mathbf{250\ MHz}$ .

Step 3. Because the step is precisely half an FSR, even- $m+n$ modes pile onto the axial frequencies and odd- $m+n$ modes onto the midpoints. The dense ladder of $(m,n)$ patterns collapses into just two combs, each of spacing $500\ \mathrm{MHz}$ but offset by $250\ \mathrm{MHz}$ .

Step 4. This degeneracy cuts both ways: it makes the confocal cavity a forgiving, mode-matched spectrum analyzer, but it makes single-transverse-mode laser operation hard, because $\mathrm{TEM}_{00}$ has no frequency isolation from its degenerate neighbours. Effective spacing 250 MHz; spectrum collapses to two interleaved 500 MHz combs.

Physics of the portrait

The Hermite–Gaussian mode arrived twice — first as nineteenth-century wave optics no one had a use for, then again in 1961 when Bell Labs, chasing the new laser, found that an open resonator sorts light into exactly these patterns.

Origins · 1 minHermite–Gaussian modes and how this was discovered

Hermite–Gaussian modes

the transverse eigenbasis u_mn = H_m(√2x/w) H_n(√2y/w) e^−(x²+y²)/w² of a stable resonator · laser resonator theory

When the first lasers fired in 1960 no one knew what transverse shape the light would take, because the mirrors were open — no waveguide walls to quantize the field. In 1961 A. G. Fox and Tingye Li answered it numerically, bouncing a trial field between two mirrors in a computer until it relaxed to a self-reproducing pattern — the first proof that an open resonator has discrete transverse eigenmodes. The same year Gary Boyd and James Gordon solved the confocal resonator analytically and found its eigenfunctions were precisely the Hermite–Gaussian functions, the family Louis Georges Gouy's 1890 focal-phase work foreshadowed and the quantum oscillator had made famous. Boyd and Herwig Kogelnik then generalized it to arbitrary stable cavities via the $g$ -parameters, and Kogelnik and Li's 1966 review "Laser Beams and Resonators" fused Gaussian beams, the $\mathrm{TEM}_{mn}$ modes, the Gouy phase, and the resonance formula into the framework still taught today. In 1992 Les Allen and co-workers showed the cylindrical cousins carry orbital angular momentum $l\hbar$ per photon, founding structured light.

History

1890

— Gouy discovers the focal phase anomaly, the extra half-cycle a converging wave gains through focus, now called the Gouy phase

1961

— Fox and Li compute self-reproducing modes of an open resonator numerically, proving open cavities have discrete transverse eigenmodes

1961

— Boyd and Gordon solve the confocal resonator analytically and identify its eigenfunctions as the Hermite–Gaussian modes

1966

— Kogelnik and Li publish "Laser Beams and Resonators," the synthesis of Gaussian beams, g-parameters, and the TEM_mn resonance formula

1992

— Allen and co-workers show Laguerre–Gaussian modes carry orbital angular momentum lħ per photon, launching the field of structured light

Going further

Texts. Siegman, Lasers (1986), Ch. 16–19 (the definitive treatment of Hermite–Gaussian and Laguerre–Gaussian modes) · Saleh and Teich, Fundamentals of Photonics, 3rd ed., Ch. 3 and Ch. 10–11 · Kogelnik and Li, "Laser Beams and Resonators," Applied Optics 5, 1550 (1966) · Demtröder, Laser Spectroscopy, Vol. 1, Ch. 5 · Hecht, Optics, 5th ed., §13.1 · Foot, Atomic Physics, Ch. 7.

Provenance. The HG mode form and its $(m+n+1)$ Gouy factor are recalled from the paraxial-beam and resonator articles and re-derived here from the paraxial wave equation; the resonance formula and confocal $\tfrac12$ -FSR limit follow Siegman Ch. 19 and Demtröder Ch. 5; $M^2=2m+1$ follows Siegman Ch. 17 and ISO 11146. A verifier should reconfirm: (1) the HG mode and Gouy factor from the paraxial equation; (2) the resonance formula and the confocal ( $\tfrac12$ ) and $g=0.5$ ( $\tfrac13$ ) special cases; (3) the worked numbers — 150 MHz FSR and 50 MHz step at $L=1.0\ \mathrm{m}$ , $g=0.5$ ; 500 and 250 MHz for the confocal $0.30\ \mathrm{m}$ cavity; 1.13, 2.52, 1.96 mrad divergences for $\mathrm{TEM}_{21}$ at $\lambda=1064\ \mathrm{nm}$ , $w_0=0.30\ \mathrm{mm}$ ; (4) lobe count $(m+1)(n+1)$ and spot scaling $\sqrt{2m+1}$ .

Builds on: Gaussian Beams · Optical Resonators and Cavity Modes · Resonator Stability and the g-Parameter Diagram · ABCD Matrix Beam Propagation. Leads to: Longitudinal Modes and Free Spectral Range · Cavity Q Factor and Photon Lifetime · Laser Threshold Condition.

Ask the tutor: Ask me to work out the transverse-mode spectrum, waist, or $M^2$ for any cavity you specify — just give me the length, mirror radii, and wavelength.

Do higher-order modes have a different waist, focus, or Rayleigh range than the fundamental?

No — this is the single most common confusion. Every $\mathrm{TEM}_{mn}$ of a cavity is built on the same $q(z)$ , so all share the identical waist location, waist size $w_0$ , Rayleigh range $z_R$ , and curvature $R(z)$ . What differs is the transverse shape: higher orders have more lobes, a wider footprint ( $\propto\sqrt{2m+1}$ ), and a larger Gouy phase. Imaged, the envelope tracks the same $w(z)$ curve for every mode; only the pattern inside it, and the resonant frequency, change with $m,n$ .

Why do Hermite polynomials appear at all — is the harmonic-oscillator link a coincidence?

It is not a coincidence; it is the same equation. Factor the fundamental Gaussian out of the paraxial equation and ask what transverse structure rides along without changing shape, and what remains is Hermite's equation $g''-2\xi g'+2m g=0$ — term for term the equation of the quantum oscillator's wavefunctions. Bounded solutions exist only for integer $m$ and are the Hermite polynomials, which is why optical mode indices count nodes exactly like oscillator quantum numbers. Even the Gouy phase $(m+n+1)\psi$ mirrors the oscillator energy $(N+1)\hbar\omega$ .

How is transverse-mode splitting different from the longitudinal free spectral range?

They are two independent frequency scales people often conflate. The free spectral range $\Delta\nu_{\mathrm{FSR}}=c/2L$ is the spacing between successive axial modes (increment $q$ , add one FSR). Transverse-mode splitting $\Delta\nu_{\mathrm{trans}}=(\Delta\nu_{\mathrm{FSR}}/\pi)\arccos(\pm\sqrt{g_1g_2})$ is the much finer spacing from incrementing $m+n$ at fixed $q$ , set by the $g$ -parameters. For most geometries it is an irrational fraction of the FSR, so transverse resonances weave through the axial comb without coinciding; confocal ( $\tfrac12$ ) and $g=0.5$ ( $\tfrac13$ ) are the exceptional commensurate cases.

What exactly is M-squared, and why is it equal to 2m+1?

$M^2$ is the times-diffraction-limit factor: the ratio of a real beam's space-bandwidth product (size times divergence) to an ideal Gaussian's. A perfect $\mathrm{TEM}_{00}$ has $M^2=1$ ; anything larger focuses to a bigger spot and diverges faster. For a pure Hermite–Gaussian mode $M^2_x=2m+1$ per axis, straight from the second-moment width: the Hermite recursion gives $\langle x^2\rangle=(2m+1)w^2/4$ , the same integral that gives an oscillator's position spread $\propto(2N+1)$ . So a $\mathrm{TEM}_{50}$ beam ( $M^2=11$ ) focuses to a spot $\sqrt{11}\approx3.3$ times larger than the fundamental.

When should I use Laguerre–Gaussian modes instead of Hermite–Gaussian?

Match the basis to the symmetry. Hermite–Gaussian modes are natural for rectangular symmetry — a resonator with astigmatism, a Brewster window, or a slab gain medium. Laguerre–Gaussian modes are natural for cylindrical symmetry — round mirrors and apertures, isotropic gain. Both are complete, so either expands in the other via $2p+\lvert l\rvert=m+n$ , sharing the same Gouy phase and resonant frequency. Choose Laguerre–Gaussian when you want the azimuthal phase $e^{il\phi}$ : the central null, or orbital angular momentum $l\hbar$ per photon.

How does an aperture select TEM₀₀ without simply killing the laser?

The trick is differential loss, not brute-force blocking. The fundamental is the narrowest mode, so a well-sized iris passes it with almost no clipping and negligible added loss. Every higher-order mode is wider ( $\propto\sqrt{2m+1}$ ) and pushes power to the edges where the iris cuts it, picking up substantial diffraction loss. Since lasing needs gain above loss, the aperture lifts the high modes' thresholds above the available gain while leaving $\mathrm{TEM}_{00}$ above its own — a clean fundamental beam with $M^2\approx1$ at a small cost in output power.