Aerial view of the LIGO Hanford Observatory in the desert of eastern Washington, with the central building and the two 4 kilometre arms running away to the horizon
Unit 15 · Advanced + Gravitational Waves · 75 min

Laser Interferometer Observatories

How a strain of one part in a thousand million million million is measured
Image: LIGO Laboratory

Turns the Michelson interferometer of Unit 2 into a gravitational-wave detector and explains the optics of Advanced LIGO: 4 kilometre arms made effectively 1,144 kilometres long by Fabry-Perot cavities, power and signal recycling, and 40 kilogram mirrors on four-stage pendulums. The noise budget is taken term by term, seismic, thermal, quantum, and Newtonian, with squeezed light as the way past the shot-noise limit, and the unit ends with matched filtering, what five sigma means, and the network of detectors that locates a source on the sky.

Preview: the beginning of the unit

Unit 13 ended with the number that this unit has to explain: the first gravitational wave ever measured changed each 4 kilometre arm of a detector by \(2 \times 10^{-18}\) metres and the difference of the two arm lengths by \(4 \times 10^{-18}\) metres, a four-hundredth of the diameter of a proton. Unit 14 described the black holes that sent it. This unit is about the machine in between. It takes the Michelson interferometer of Unit 2, built to find an ether wind, and shows how the same instrument, with mirrors that hang free, light that is stored for milliseconds, and a noise budget worked out term by term, became the detector that measured a strain of \(10^{-21}\) on 14 September 2015.

1 The interferometer as a detector

The Michelson interferometer of Unit 2 splits a beam of light at a half-silvered mirror, sends the two halves along two perpendicular arms of equal length \(L\) to a mirror each, and brings them back together, where they interfere. It detects any effect that changes one arm relative to the other, and a gravitational wave does exactly that. Unit 13 showed that a wave arriving along the normal to the plane of a ring of free masses stretches the ring along one axis and squeezes it along the perpendicular axis, each by the fraction \(h/2\), and the two arms of an interferometer are two diameters of that ring. If the plus polarization of the wave is lined up with the arms, each arm changes its length by \begin{equation} \Delta L = \frac{h}{2}\,L , \label{eq:dL} \end{equation} one arm growing while the other shrinks, so that the difference of the two arm lengths changes by \(hL\). For the wave to do this the mirrors must be free to move, which is why every mirror of a gravitational-wave detector hangs from a pendulum rather than standing on a slab of stone: at every frequency well above its own swing frequency (about 1 hertz, as Section 3 explains) a pendulum lets the mirror move freely along the beam, so that it follows the geodesic deviation of Unit 13 as a free mass would.

The round trip of the light along an arm that has grown by \(\Delta L\) is longer by \(2\Delta L\), so light of wavelength \(\lambda\) completes \(2\Delta L/\lambda\) more cycles on the way and its phase on return is shifted by \begin{equation} \Delta\phi = 2\pi\,\frac{2\Delta L}{\lambda} = \frac{4\pi}{\lambda}\,\Delta L \label{eq:dphi} \end{equation} for one round trip. In the other arm the sign is reversed, so the phase difference between the two returning beams, which the interference measures, changes by twice this amount. In the detectors of this unit the arms are adjusted so that the two beams cancel each other at the output, which is called the dark fringe: with no wave, no light reaches the photodetector, and a wave lets a small amount through whose variation follows the strain. A small change in the light is easier to measure when the output is almost dark than when it is bright, which is one reason for the choice; the other is that at the dark fringe almost all of the light returns towards the laser, where Section 2 will put it to use.

Why are the arms 4 kilometres long? A given strain produces a phase shift proportional to \(L\), while the disturbances that move the mirrors, the shaking of the ground and the thermal motion of the glass, are displacements of a fixed size. The strain that a displacement noise \(\delta x\) imitates is \(\delta x/L\), so a longer arm pushes every displacement noise down by the same factor. The limit comes from the light itself: if the round trip takes longer than half a period of the wave, the arm is already shrinking again while the light is still on its way out, and the two halves of the trip cancel. The derivation box below works this out, and Figure 1 shows the result: the best arm length for a wave of frequency \(f\) is a quarter of its wavelength, \(L = c/(4f)\), which is 750 kilometres at 100 hertz. No one can build that, and Section 2 describes the trick that makes 4 kilometres behave like a thousand.