Gaussian Beam Divergence (θ = λ/πw₀)

A ray has no width, and every equation in the optics shard next door — the thin lens, Snell's law, Brewster's angle — was built on that convenience. A laser beam has nothing but width. Kogelnik and Li worked out in 1966 what the field actually looks like, and the answer is a Gaussian profile that narrows to a waist and opens out again, never becoming a point and never becoming parallel. The far-field half-angle it opens at is $\theta = \lambda/(\pi w_0)$, and it depends on nothing but the wavelength and the width of the beam's own waist.

Read the equation as a statement about what you cannot have. A small waist and a small divergence are the same request made twice, and the product $w_0\theta = \lambda/\pi$ is fixed. Every beam expander, every lens, every fibre in the chain trades one against the other, and none of them reduces the product. If you want a beam that is still narrow a kilometre away, the only lever is to make it wide before it leaves — which is why a laser rangefinder has a telescope on the front and why the retroreflector experiments on the Moon fire through a two-metre mirror.

Now the thing that causes almost every real error in laser arithmetic, and it is not the physics. This page returns a HALF-angle, measured from the beam axis, to the $1/e^2$ contour, in the far field. There are three other perfectly standard ways to quote the same beam. The full angle is twice this. The full width at half maximum of a Gaussian is $2w\sqrt{\ln 2 / 2} = 1.1774\,w$, so a $1/e^2$ width is 1.699 times an FWHM width, and a divergence quoted at FWHM is 1.699 times smaller than one quoted at $1/e^2$. Combine the two choices and there is a factor of 3.4 between the largest and smallest honest number for one physical beam.

A datasheet frequently does not say which it used. Beam profilers report whichever the operator selected in a menu, and the selection does not always make it onto the certificate. So when a measurement disagrees with this page by 2, by 1.7 or by 3.4, nothing is broken — you are comparing conventions, and the fix is a phone call rather than a new optic. I would rather say this plainly on every page of this shard than have one reader re-cut a lens over it.

The second trap is quieter. $w_0$ is a RADIUS. A beam profiler reports a diameter, a catalogue prints a diameter, and the number you are holding is therefore usually twice what this equation wants. Entering a diameter here halves the divergence, which is not an obviously wrong answer and is exactly why the mistake survives.

Two limits on the result itself. It is a FAR-FIELD figure: close to the waist the beam is not diverging at anything like this rate, and the Rayleigh range page is where that transition lives. And it assumes a perfect Gaussian, which no real laser is. A real beam diverges by a factor $M^2$ more, and that factor is measured rather than assumed.