Gaussian Beams

A laser does not emit a plane wave, and it does not emit light from a point — it emits a Gaussian beam, a self-consistent bundle whose transverse profile stays Gaussian no matter how it is focused, expanded, or left to spread. That shape is not imposed by hand: it is the lowest-order solution of the paraxial Helmholtz equation, the approximation to the wave equation valid whenever a beam's angular spread is small. This article consumes that paraxial result directly. Write the field as a slowly varying envelope on the fast carrier $e^{ikz}$ , demand that the envelope stay confined about the axis, and the Gaussian falloff $e^{-r^2/w^2}$ drops out automatically — together with a wavefront that flattens at the waist and curves as the beam expands. Being the smoothest, lowest-loss field a resonator can hold, this mode is also a laser cavity's natural output. Two length scales run the whole show. The waist $w_0$ is the beam's narrowest radius; the Rayleigh range $z_R=\pi w_0^2/\lambda$ is the distance over which it stays roughly collimated before diffraction opens it into a cone of fixed half-angle $\theta=\lambda/(\pi w_0)$ . Their product $w_0\theta=\lambda/\pi$ is an invariant — you cannot make a beam both tightly focused and weakly diverging. All of this collapses into one complex number, the beam parameter $q(z)=z+iz_R$ , whose real part tracks distance from the waist and imaginary part fixes the spot size, and which propagates through any lens or gap by the same ABCD matrix that carries geometric rays.

Key takeaways

A Gaussian beam is the fundamental TEM $_{00}$ solution of the paraxial wave equation: its amplitude falls as $e^{-r^2/w^2(z)}$ and its intensity as $e^{-2r^2/w^2(z)}$ .
Two numbers fix everything — the waist $w_0$ and the wavelength $\lambda$ — giving the Rayleigh range $z_R=\pi w_0^2/\lambda$ , the curvature $R(z)$ , the divergence $\theta$ , and the Gouy phase.
The spot expands hyperbolically, $w(z)=w_0\sqrt{1+(z/z_R)^2}$ , opening into a far-field cone of half-angle $\theta=\lambda/(\pi w_0)$ ; the invariant $w_0\theta=\lambda/\pi$ is diffraction made quantitative.
Passing through a focus the beam slips an extra $\pi$ of Gouy phase, $\zeta(z)=\arctan(z/z_R)$ , relative to a plane wave — the shift that separates transverse cavity resonances.
Key idea: a Gaussian beam is completely specified by the complex parameter $q(z)=z+iz_R$ — its imaginary part fixes the beam size, its real part fixes the wavefront curvature — and $q$ transforms through any paraxial system by the same ABCD matrix that propagates rays.

The core insight

A Gaussian beam is completely specified by the complex parameter $q(z)=z+iz_R$ : its imaginary part fixes the beam size and its real part fixes the wavefront curvature, and $q$ transforms through any paraxial optical system by the same ABCD matrix that propagates rays.

Where this lives today

Every kilometre-scale gravitational-wave interferometer is, at bottom, an exercise in Gaussian-beam propagation. In LIGO and Virgo the laser field circulates in Fabry–Pérot arm cavities several kilometres long, and the beam must be mode-matched to each cavity's Gaussian eigenmode to a fraction of a percent — a mismatch dumps power into higher-order Hermite–Gaussian modes and spoils the strain sensitivity. Designers set the arm-cavity mirror radii to place a waist of a few centimetres, spreading the spot to sample a large mirror area (lowering thermal noise) without clipping the optics, and track the Gouy phase between components because it governs how alignment jitter couples into the readout. The same calculus, scaled down, runs through optical tweezers trapping a single cell, the femtosecond-laser cutting of LASIK surgery, satellite laser links whose beam expanders trade waist for divergence, and the fibre-coupling lenses on every photonics bench. Wherever coherent light is focused, guided, or resonated, the $q$ -parameter and the $ABCD$ law are doing the design.

The physical picture

Start at the waist, where the wavefronts are flat and the beam is at its narrowest radius $w_0$ . Diffraction will not let a field confined to width $w_0$ stay confined: it must acquire an angular spread of order $\lambda/w_0$ and open up. The tighter you squeeze the waist, the more violently it fans out — a small $w_0$ buys a short collimated region and a wide cone, a large $w_0$ the reverse. The crossover, where "collimated" turns into "diverging," is the Rayleigh range $z_R$ : the beam holds together over roughly $\pm z_R$ about the waist and beyond expands linearly into a cone. The first figure draws this envelope for a real helium–neon beam and marks the three numbers a designer reads off it.

Figure

Longitudinal envelope of a HeNe beam with λ = 633 nm and waist w₀ = 0.5 mm: the spot follows w(z) = w₀√(1 + (z ∕ z_R)²), reaching √2 w₀ ≈ 0.71 mm at one Rayleigh range z_R = 1.24 m and then opening into a cone of half-angle θ = λ ∕ (π w₀) ≈ 0.40 mrad

Now cut across the beam at fixed $z$ . The intensity is a bell curve, $I(r)=I_0\,e^{-2r^2/w^2}$ , and the "radius" $w$ is fixed by a convention worth memorizing: $w$ is where the intensity has fallen to $1/e^2\approx13.5\%$ of peak (where the amplitude reaches $1/e$ ). That circle encloses $1-e^{-2}\approx86.5\%$ of the total power. Beware the competing convention: the full width at half maximum is narrower, $\mathrm{FWHM}=w\sqrt{2\ln 2}\approx1.18\,w$ , and mixing the two silently corrupts every downstream number. The second figure carries all three markers.

Figure

Transverse intensity profile I(r) = I₀ exp(−2r² ∕ w²) at fixed z: the beam radius w is the 1 ∕ e² point where I drops to 13.5 percent of peak, the full width at half maximum is only 1.18 w, and the disc r less than w already captures 86.5 percent of the total power

Two mental moves generate every optical system. Free propagation adds real distance, $q\to q+d$ : the beam grows, the wavefront curves, the waist recedes behind you. A thin lens leaves the beam size unchanged but bends the wavefront, $1/q\to 1/q-1/f$ . Everything else — telescopes, focusing systems, mode-matching — is a composition of these two moves, which is why the ray matrices of geometric optics propagate Gaussian beams unchanged. Threaded through it all is the Gouy phase: gliding through its focus the beam slips an extra half-turn, a full $\pi$ relative to a plane wave, shifting the resonance frequencies of a cavity's transverse modes.

The core equations

The fundamental mode is one field expression carrying four ingredients: a Gaussian amplitude envelope, a peak that shrinks as the beam spreads (conserving power), a curved-wavefront phase, and the Gouy correction. Everything downstream is read off this line one factor at a time.

E(r,z)=E_0\,\frac{w_0}{w(z)}\exp\!\left[-\frac{r^2}{w^2(z)}\right]\exp\!\left[-ikz-ik\frac{r^2}{2R(z)}+i\zeta(z)\right] \qquad (1)

where: $E_0$ — amplitude at the waist centre; $r$ — radial distance from the axis; $z$ — axial coordinate, waist at $z=0$ ; $w(z)$ — $1/e^2$ -intensity radius; $R(z)$ — wavefront curvature radius; $\zeta(z)$ — Gouy phase; $k=2\pi/\lambda$ . The real Gaussian confines the light; the $w_0/w(z)$ prefactor dims the peak as the beam widens; the $r^2/R$ term is the parabolic approximation to a spherical wavefront. Detectors respond to intensity, so square the amplitude. The shape stays Gaussian but with a factor of two in the exponent, and the on-axis peak drops as $1/w^2$ as the same power spreads over a larger area.

I(r,z)=I_0\left[\frac{w_0}{w(z)}\right]^2\exp\!\left[-\frac{2r^2}{w^2(z)}\right],\qquad I_0=\frac{2P}{\pi w_0^2} \qquad (2)

where: $I_0$ — peak on-axis intensity at the waist; $P$ — total power; the factor $2$ comes from $I\propto|E|^2$ . Integrating over the plane fixes $I_0=2P/(\pi w_0^2)$ ; at $r=w$ the intensity is $I_0/e^2$ , and the disc $r<w$ holds $1-e^{-2}\approx86.5\%$ of $P$ . The single most important propagation law is how the spot grows with distance — a hyperbola, nearly constant near the waist and asymptotically linear far away. The whole curve is set by $w_0$ and one derived scale.

w(z)=w_0\sqrt{1+\left(\frac{z}{z_R}\right)^2} \qquad (3)

where: $w_0$ — minimum spot radius (waist); $z_R$ — Rayleigh range, where $w=\sqrt2\,w_0$ . Near the waist $w\approx w_0$ ; far away it is linear, $w\approx (w_0/z_R)\,z$ . That scale is the Rayleigh range, the axial distance over which the beam stays effectively collimated. It grows as the square of the waist, so doubling $w_0$ quadruples the collimated region — the most useful trade in beam design.

z_R=\frac{\pi w_0^2}{\lambda},\qquad b=2z_R \qquad (4)

where: $\lambda$ — wavelength in the medium; $b=2z_R$ — confocal parameter, or depth of focus. At $z=z_R$ the cross-sectional area has doubled; a tighter waist buys a shorter, less forgiving focal region. The same complex solution that gives $w(z)$ also gives the wavefront curvature and Gouy phase. The curvature is infinite (flat) at the waist, passes through a minimum radius $2z_R$ at the Rayleigh point, then straightens toward $R\approx z$ far away, as if radiating from the waist.

R(z)=z\left[1+\left(\frac{z_R}{z}\right)^2\right],\qquad \zeta(z)=\arctan\!\left(\frac{z}{z_R}\right) \qquad (5)

where: $R(z)$ — wavefront radius, $\to\infty$ at the waist and $\to z$ far away; $\zeta(z)$ — Gouy phase, running $-\pi/2$ to $+\pi/2$ for a total shift of $\pi$ across the focus. Higher Hermite–Gaussian modes carry an enlarged term $(m+n+1)\arctan(z/z_R)$ that spectrally separates them. Far from the waist the hyperbola becomes a cone whose half-angle is the far-field divergence. Multiplying by the waist gives a constant fixed by wavelength alone — the diffraction limit as an invariant.

\theta=\frac{\lambda}{\pi w_0},\qquad w_0\,\theta=\frac{\lambda}{\pi} \qquad (6)

where: $\theta$ — half-angle divergence to the $1/e^2$ radius, valid only in the far field $z\gg z_R$ . With the product pinned to $\lambda/\pi$ , tight focus and low divergence are irreconcilable — you trade one for the other but never beat their product. Finally, all of this — $w(z)$ , $R(z)$ , and their evolution — packages into one complex number whose reciprocal splits into curvature and size, the object that makes beam design a matrix computation.

\frac{1}{q(z)}=\frac{1}{R(z)}-i\,\frac{\lambda}{\pi w^2(z)},\qquad q(z)=z+iz_R,\qquad q_2=\frac{A q_1+B}{C q_1+D} \qquad (7)

where: $q$ — complex beam parameter; $\mathrm{Re}\,q=z$ — signed distance from the waist; $\mathrm{Im}\,q=z_R$ — the invariant Rayleigh range; $\begin{pmatrix}A&B\\C&D\end{pmatrix}$ — ray matrix. Free space gives $q\to q+d$ , a thin lens gives $1/q\to 1/q-1/f$ , and any cascade is the matrix product applied by the same bilinear rule.

Derivation AGaussian solution of the paraxial wave equation

The goal is to show that the Gaussian beam is not an ansatz plucked from a hat but the forced consequence of the paraxial approximation, and that the complex parameter $q$ appears the moment you insist on a solution with a quadratic-in- $r$ phase. We start from the Helmholtz equation, strip off the fast axial oscillation to expose a slowly varying envelope, feed in the simplest phase that can carry both curvature and confinement, and let the equation dictate how it evolves. What emerges is $q(z)=z+iz_R$ and, from its real and imaginary parts, every quantity in the core-equations section.

Step 1. Write the field as an envelope on a carrier, $E=\psi(r,z)\,e^{-ikz}$ , substitute into the Helmholtz equation $(\nabla^2+k^2)E=0$ , and drop the second axial derivative $\partial_z^2\psi$ because the envelope changes slowly over a wavelength. This leaves the paraxial equation $\nabla_\perp^2\psi-2ik\,\partial_z\psi=0$ . Why: discarding $\partial_z^2\psi$ is the entire paraxial approximation — it is legitimate precisely when the beam's angular spread is small, the same condition that makes $\theta\ll 1$ .

Step 2. Try the ansatz $\psi=\exp[-i(P(z)+k r^2/2q(z))]$ , a phase quadratic in $r$ with two unknown functions: a purely axial piece $P(z)$ and a complex $q(z)$ multiplying $r^2$ . Why: a complex coefficient on $r^2$ is the minimal object that can be a real curvature and an imaginary Gaussian width at once — the real part becomes $1/R$ , the imaginary part becomes the $1/w^2$ confinement.

Step 3. Insert the ansatz and collect powers of $r$ . The $r^2$ terms force $\mathrm{d}q/\mathrm{d}z=1$ , so $q(z)=q_0+z$ , while the $r^0$ terms give $\mathrm{d}P/\mathrm{d}z=-i/q$ . Choosing $q_0=iz_R$ puts the waist at $z=0$ and makes $q(z)=z+iz_R$ . Why: the choice $q_0=iz_R$ is what makes the envelope finite and Gaussian at $z=0$ rather than singular; a real $q_0$ would place the waist elsewhere or describe a diverging spherical wave.

Step 4. Read off physics by separating $1/q=1/R-i\lambda/(\pi w^2)$ , which delivers $R(z)$ and $w(z)$ exactly as quoted, while integrating $\mathrm{d}P/\mathrm{d}z=-i/q$ produces both the amplitude prefactor $w_0/w$ and the Gouy phase $\arctan(z/z_R)$ . Why: this is the payoff — one complex $q$ simultaneously encodes the two real observables a lab measures, the spot size and the wavefront curvature, and the leftover axial integral is exactly the anomalous Gouy shift.

Go deeper · Derivation B · 1 minFocusing a collimated Gaussian beam with a thin lens

Derivation BFocusing a collimated Gaussian beam with a thin lens

The goal is the single most-used result in laser engineering: where a lens brings a beam to its tightest point and how small that point can be. We treat an incoming beam whose own waist $w_0'$ sits at the lens, so its parameter there is purely imaginary, $q_1=iz_R'$ with $z_R'=\pi w_0'^2/\lambda$ . We apply the lens as a curvature kick, propagate to where the wavefront flattens again, and identify the new waist. The lesson — smaller spot demands a larger input beam and shorter focal length — governs laser cutting, two-photon microscopy, and optical trapping alike.

Step 1. Apply the lens: $1/q_2=1/q_1-1/f=-1/f-i/z_R'$ . The real part jumps from zero to $-1/f$ , but the imaginary part is untouched, so the beam size at the lens plane is exactly what it was. Why: a thin lens is infinitely thin, so it cannot resize the beam where it sits — it can only stamp a new curvature onto the wavefront, which is the whole content of $1/q\to 1/q-1/f$ .

Step 2. Propagate a distance $d$ past the lens, $q_3=q_2+d$ , and locate the new waist by demanding a flat wavefront there, $\mathrm{Re}\,q_3=0$ . Solving gives $d=f/[1+(f/z_R')^2]$ , which tends to $f$ whenever $z_R'\gg f$ . Why: the waist is by definition where the curvature is infinite, i.e. where $\mathrm{Re}\,q=0$ ; for a well-collimated input the focus therefore lands essentially in the focal plane, matching the geometric-optics expectation.

Step 3. Read the new Rayleigh range from the imaginary part, $z_R=\mathrm{Im}\,q_3=z_R' f^2/(f^2+z_R'^2)$ , and convert to a spot size through $z_R=\pi w_0^2/\lambda$ . In the collimated limit $z_R'\gg f$ this collapses to $w_0=\lambda f/(\pi w_0')$ . Why: the focal spot shrinks with a bigger input beam and a shorter lens because both raise the numerical aperture — the same diffraction bookkeeping that pins $w_0\theta=\lambda/\pi$ reappears as the focusing formula.

Go deeper · Derivation C · 1 minSelf-consistent mode of a two-mirror resonator

Derivation CSelf-consistent mode of a two-mirror resonator

The goal is to explain why a laser emits a Gaussian beam at all: the resonator selects it. A field survives many round trips between two curved mirrors only if it reproduces itself — same spot, same curvature — after a full loop, its wavefronts arriving at each mirror already matched to that mirror's curvature so they reflect straight back. We turn "reproduces itself" into a fixed-point condition on $q$ , find that a real Rayleigh range exists only inside a bounded region of mirror geometries, and recover the stability criterion and the mode's waist — the bridge from beam optics to real cavity design and mode matching.

Step 1. Impose round-trip self-consistency: with the cavity's round-trip ray matrix $\begin{pmatrix}A&B\\C&D\end{pmatrix}$ , the mode must satisfy $q=(Aq+B)/(Cq+D)$ , a quadratic in $1/q$ . Why: an eigenmode is a field the cavity leaves unchanged up to a phase and loss, so its $q$ must be a fixed point of the round-trip transformation — anything else reshapes on each pass and eventually walks off the mirrors.

Step 2. Solve the quadratic and demand that $\mathrm{Im}\,q=z_R$ be real and nonzero, which requires the discriminant to have the right sign. In terms of the $g$ -parameters $g_i=1-L/R_i$ this is precisely $0\le g_1 g_2\le 1$ . Why: a real $z_R$ is the difference between a physically confined beam and a mode that diffracts away every pass; the condition $0\le g_1g_2\le1$ is therefore literally the boundary between a working laser and a lossy one.

Step 3. With a stable geometry, back out the waist size and location, $w_0^2=(\lambda L/\pi)\sqrt{g_1 g_2(1-g_1g_2)}\,/\,(g_1+g_2-2g_1g_2)$ , and $z_1=g_2(1-g_1)L/(g_1+g_2-2g_1g_2)$ . Why: these tie the beam parameter to hardware — mirror radii and spacing — so matching an external beam's $q$ to the mode's $q$ becomes the recipe for coupling light into the cavity.

Limiting cases

The propagation law $w(z)=w_0\sqrt{1+(z/z_R)^2}$ and its partners collapse into simple, memorable behaviour in each regime.

Case	Result	Meaning
Near field, $z\ll z_R$	$w\approx w_0$ , $R\to\infty$	Nearly collimated with flat wavefronts; the beam mimics a plane wave over the depth of focus $\sim 2z_R$ .
Rayleigh point, $z=z_R$	$w=\sqrt2\,w_0$ , $R=2z_R$ (minimum)	Spot area has doubled and the wavefront curvature is tightest — the boundary between near and far field.
Far field, $z\gg z_R$	$w\approx\theta z=\tfrac{\lambda}{\pi w_0}z$ , $R\approx z$	Linear expansion into a fixed cone; wavefronts are essentially spherical, centred on the waist.
Plane-wave limit, $w_0\to\infty$	$z_R\to\infty$ , $\theta\to0$	Infinite waist gives infinite Rayleigh range and zero divergence — the ideal plane wave, physically unattainable.
Point-source limit, $w_0\to0$	$z_R\to0$ , $\theta\to\infty$	Vanishing waist diverges instantly — diffraction's opposite, equally idealized, extreme.

Common trapThe divergence angle applies everywhere along the beam, and the beam radius is the FWHM

Both halves are wrong and both bite hard. First, $\theta=\lambda/(\pi w_0)$ is a far-field statement, valid only for $z\gg z_R$ ; applied near the waist it predicts spreading where the beam is actually flat — inside one Rayleigh range the beam is collimated, not fanning at angle $\theta$ . Second, the beam radius $w$ is the $1/e^2$ -intensity ( $1/e$ -amplitude) radius, not the full width at half maximum: $\mathrm{FWHM}=w\sqrt{2\ln 2}\approx1.18\,w$ . Mixing the conventions misplaces focal spots, mis-sizes apertures (the $1/e^2$ disc holds $86.5\%$ of the power, not $50\%$ ), and corrupts every $z_R$ and $\theta$ downstream.

Worked examples

1 · Focusing a Nd:YAG beam for laser machining (confidence builder)

A $10\ \mathrm{W}$ Nd:YAG laser at $\lambda=1064\ \mathrm{nm}$ has a collimated beam of waist radius $w_0'=3\ \mathrm{mm}$ striking a lens of focal length $f=100\ \mathrm{mm}$ . Find the focal spot radius, the depth of focus, and the peak intensity at focus.

Step 1. Check the input is well-collimated over the lens. Its Rayleigh range $z_R'=\pi w_0'^2/\lambda=\pi(3\times10^{-3})^2/(1.064\times10^{-6})\approx26.6\ \mathrm{m}$ dwarfs $f=0.1\ \mathrm{m}$ by over two orders of magnitude, so $w_0=\lambda f/(\pi w_0')$ applies and the focus sits essentially in the focal plane.

Step 2. Focal spot: $w_0=\dfrac{\lambda f}{\pi w_0'}=\dfrac{(1.064\times10^{-6})(0.100)}{\pi(3\times10^{-3})}=1.13\times10^{-5}\ \mathrm{m}$ , so the beam focuses to $w_0\approx11.3\ \mu\mathrm{m}$ — roughly ten wavelengths across.

Step 3. Depth of focus: $2z_R=2\pi w_0^2/\lambda=2\pi(1.129\times10^{-5})^2/(1.064\times10^{-6})=7.5\times10^{-4}\ \mathrm{m}$ , i.e. $2z_R\approx0.75\ \mathrm{mm}$ — the axial window over which the cut stays sharp. Peak intensity: $I_0=2P/(\pi w_0^2)=2(10)/[\pi(1.129\times10^{-5})^2]\approx\mathbf{5.0\times10^{10}\ W/m^2}$ . A modest $10\ \mathrm{W}$ becomes metal-vaporizing purely by concentration into an $11\ \mu\mathrm{m}$ waist.

2 · Rayleigh range, divergence, and far-field size of a HeNe beam

A helium–neon laser at $\lambda=633\ \mathrm{nm}$ has a waist $w_0=0.5\ \mathrm{mm}$ . Find its Rayleigh range, far-field divergence half-angle, and beam radius $10\ \mathrm{m}$ downstream.

Step 1. Rayleigh range: $z_R=\pi w_0^2/\lambda=\pi(5\times10^{-4})^2/(6.33\times10^{-7})\approx\mathbf{1.24\ m}$ — collimated over roughly a metre either side of the waist.

Step 2. Divergence: $\theta=\lambda/(\pi w_0)=(6.33\times10^{-7})/[\pi(5\times10^{-4})]\approx4.0\times10^{-4}\ \mathrm{rad}=\mathbf{0.40\ mrad}$ — a far-field number, not to be applied inside the first $1.24\ \mathrm{m}$ .

Step 3. At $z=10\ \mathrm{m}$ , well beyond $z_R$ , use the full law: $w(10)=w_0\sqrt{1+(10/1.24)^2}=5\times10^{-4}\sqrt{1+65.0}\approx4.06\times10^{-3}\ \mathrm{m}$ . The spot radius is about $4.1\ \mathrm{mm}$ (diameter $\approx8.1\ \mathrm{mm}$ ) — the waist has grown eightfold, matching the far-field estimate $\theta z=(4.0\times10^{-4})(10)=4.0\ \mathrm{mm}$ .

3 · Reading the spot size off the complex parameter

A beam at $\lambda=1064\ \mathrm{nm}$ has its waist $w_0=1\ \mathrm{mm}$ at $z=0$ . Use the $q$ -parameter alone to find the beam radius $5\ \mathrm{m}$ past the waist, then confirm it against $w(z)$ .

Step 1. At the waist $q$ is purely imaginary: $z_R=\pi w_0^2/\lambda=\pi(10^{-3})^2/(1.064\times10^{-6})\approx2.95\ \mathrm{m}$ , so $q(0)=2.95\,i\ \mathrm{m}$ . Free propagation adds real distance, so at $z=5\ \mathrm{m}$ , $q=5+2.95\,i$ (metres).

Step 2. Invert: $1/q=\bar q/|q|^2=(5-2.95i)/33.7=0.148-0.0876\,i\ \mathrm{m^{-1}}$ . The imaginary part is $-\lambda/(\pi w^2)$ , so $w^2=\lambda/(\pi\cdot0.0876)=3.87\times10^{-6}\ \mathrm{m^2}$ , giving $w\approx1.97\ \mathrm{mm}$ .

Step 3. Cross-check with the closed form: $w(5)=w_0\sqrt{1+(5/2.95)^2}=10^{-3}\sqrt{3.87}=\mathbf{1.97\ mm}$ . The two agree exactly — $q$ carries the same information as $w(z)$ and $R(z)$ together, but propagates through any lens or gap by simple arithmetic.

Physics of the portrait

The Gaussian beam is a nineteenth-century phase anomaly and a twentieth-century laser problem stitched into one formalism. The name attached to that stitching belongs to the engineer who showed that a beam of light and a ray of light obey the same matrices.

Origins · 1 minHerwig Kogelnik and how this was discovered

Herwig Kogelnik

b. 1932, Austrian-American physicist and engineer · laser beam and resonator theory

Herwig Kogelnik trained in Vienna and Oxford and spent his career at Bell Laboratories, where in the mid-1960s the newly invented laser posed a sharp question: what does a laser beam do as it propagates, focuses, and resonates between mirrors? Building on Boyd and Gordon's 1961 confocal-resonator mode analysis, Kogelnik and Tingye Li synthesized the answer in their 1966 review "Laser Beams and Resonators," one of the most cited papers in optics. Its two enduring gifts are the complex beam parameter $q=z+iz_R$ , which fuses spot size and wavefront curvature into a single number, and the demonstration that $q$ transforms through any system by the very same $ABCD$ matrix geometric optics uses for rays — so a diffracting Gaussian beam can be traced with the bookkeeping of ray optics. Kogelnik also formalized the mode-matching prescription for coupling an external beam into a cavity or fibre, still a daily tool of every laser laboratory. He later did pioneering work in integrated optics and holography, but the $q$ -parameter and the $ABCD$ law remain the formalism this article is built around.

History

1890

— Gouy observes the axial phase anomaly a converging wave acquires in passing through its focus

1960

— Maiman operates the first laser, producing the first man-made coherent optical beam

1961

— Boyd and Gordon derive the Gaussian eigenmodes of a confocal optical resonator

1966

— Kogelnik and Li publish "Laser Beams and Resonators," codifying the complex q-parameter and the ABCD law for beams

1990

— Siegman introduces the M² factor quantifying a real beam's departure from an ideal Gaussian

Going further

Texts. Saleh and Teich, Fundamentals of Photonics, 3rd ed., Ch. 3 (Beam Optics) · Siegman, Lasers, Ch. 16–17 (the definitive treatment of Gaussian beams and the ABCD law) · Hecht, Optics, §13 (lasers and beam propagation) · Demtröder, Laser Spectroscopy, Vol. 1, Ch. 5 (resonators and Gaussian modes) · Foot, Atomic Physics, appendix on laser beams · Kogelnik and Li, Laser Beams and Resonators, Applied Optics 5, 1550 (1966).

Provenance. The field and intensity expressions, the propagation law $w(z)$ , the Rayleigh range $z_R=\pi w_0^2/\lambda$ , the curvature $R(z)$ , the Gouy phase $\arctan(z/z_R)$ , the divergence $\theta=\lambda/(\pi w_0)$ , and the $q$ -parameter and ABCD rules follow Saleh and Teich (Ch. 3), Siegman (Ch. 16–17), Kogelnik and Li (1966), Hecht, and Demtröder. A verifier should recompute: the HeNe numbers ( $z_R\approx1.24\ \mathrm{m}$ , $\theta\approx0.40\ \mathrm{mrad}$ , $w(10\,\mathrm{m})\approx4.1\ \mathrm{mm}$ ); the Nd:YAG focusing spot $w_0=\lambda f/(\pi w_0')\approx11.3\ \mu\mathrm{m}$ , depth of focus $2z_R\approx0.75\ \mathrm{mm}$ , and peak intensity $\approx5.0\times10^{10}\ \mathrm{W/m^2}$ ; the $q$ -parameter cross-check $w(5\,\mathrm{m})\approx1.97\ \mathrm{mm}$ ; the enclosed-power figure $1-e^{-2}\approx0.865$ ; and the conversion $\mathrm{FWHM}=w\sqrt{2\ln2}\approx1.18\,w$ . All values are standard textbook results; none are fabricated.

Prerequisites. Paraxial Helmholtz equation · Fourier optics and diffraction · Optical Resonators and Cavity Modes. Leads to. ABCD Matrix Beam Propagation · Transverse Cavity Modes (Hermite–Gaussian) · Resonator Stability and the g-Parameter Diagram.

Ask the tutor: ask me to trace your own beam through a lens or cavity with the $q$ -parameter, and I will walk each $ABCD$ step and read off the waist, divergence, and depth of focus.

Why is the beam radius defined at the 1/e² intensity point

Because it is the most useful compromise and it keeps the mathematics clean. At $r=w$ the intensity has dropped to $1/e^2\approx13.5\%$ of peak, the disc inside encloses $1-e^{-2}\approx86.5\%$ of the power, and it is the natural scale of the Gaussian itself (the amplitude is $E_0/e$ there) — so $w$ genuinely marks "where the beam is." The competing full-width-at-half-maximum measure is narrower, $\mathrm{FWHM}=w\sqrt{2\ln2}\approx1.18\,w$ , and common in spectroscopy, so always check which convention a datasheet means before plugging into $z_R$ or $\theta$ .

What is the physical meaning of the Rayleigh range

The Rayleigh range $z_R=\pi w_0^2/\lambda$ is the distance from the waist over which the beam stays effectively collimated. At $z=z_R$ the spot has grown to $\sqrt2\,w_0$ , so its area has exactly doubled, and the wavefront curvature reaches its tightest value $R=2z_R$ . It divides the beam into two worlds: for $z\ll z_R$ the beam is nearly a plane wave with flat wavefronts, and for $z\gg z_R$ it fans into a spherical wave at the divergence angle $\theta$ . The usable depth of focus is $b=2z_R$ , and because $z_R\propto w_0^2$ , doubling the waist quadruples the collimated region — the central trade-off in every focusing decision.

How can the same ABCD matrices work for both rays and Gaussian beams

Because both are captured by a single quantity that transforms as a ratio. A ray carries a height and slope $(y,y')$ , and any paraxial element maps them linearly through its $ABCD$ matrix, so the ratio $y/y'$ transforms by the bilinear rule $(Aq+B)/(Cq+D)$ . The beam parameter $q$ plays exactly that role for a beam. Kogelnik and Li showed the correspondence is exact — free space gives $q\to q+d$ , a thin lens gives $1/q\to1/q-1/f$ , identical to the ray matrices — so you can trace a diffracting beam with the arithmetic of geometric optics.

What is the Gouy phase and why does it matter

The Gouy phase $\zeta(z)=\arctan(z/z_R)$ is an extra advance a focused beam gains over the plane-wave phase $kz$ as it passes through its waist, running from $-\pi/2$ before the focus to $+\pi/2$ after — a total anomalous shift of $\pi$ . Physically it reflects the spread of transverse wavevectors needed to confine the beam: each off-axis component travels slightly farther, and their collective phase slips through the focus. It matters because it shifts a cavity's resonant frequencies, and because higher-order modes carry an enlarged shift $(m+n+1)\arctan(z/z_R)$ that spectrally separates the transverse modes.

If I want the smallest possible focal spot, what should I change

The focusing result $w_0=\lambda f/(\pi w_0')$ gives three levers: shorten the wavelength (why photolithography pushes into the ultraviolet), fill the lens with a larger input beam $w_0'$ for higher numerical aperture (why beam expanders precede a focusing objective), and shorten the focal length. All three widen the cone of converging light, and diffraction bounds the result through $w_0\theta=\lambda/\pi$ . The unavoidable cost of a smaller spot is a shorter depth of focus, since $2z_R\propto w_0^2$ .

Does the beam seeming to speed up near a focus violate anything

No. The Gouy phase makes the on-axis phase advance faster than $kz$ near the focus, which can look like a superluminal crest, but a phase feature carries no information — the same subtlety that makes a phase index below one harmless. Energy and signals travel with the group and Poynting flux, which stay subluminal. The Gouy shift is a real, measurable effect — it displaces the peak on-axis field slightly from the geometric focus and shifts cavity resonances — but it describes how a confined beam's wavefronts are arranged, not anything outrunning light.