The 100-metre Green Bank Telescope in West Virginia under a blue sky, one of the radio telescopes that time the pulsars of the NANOGrav array
Unit 16 · Advanced + Gravitational Waves · 70 min

Pulsar Timing Arrays and Space Detectors

A galaxy-sized detector, the evidence of 2023, and LISA
Image: NRAO/AUI/NSF, Wikimedia Commons, CC BY 4.0

Lays out the gravitational-wave spectrum from kilohertz to nanohertz and explains how an array of millisecond pulsars timed to a hundred nanoseconds over decades becomes a detector the size of the Galaxy, with the Hellings-Downs curve as the fingerprint of a background of waves. The evidence announced in June 2023, the supermassive black hole binaries behind it, LISA with its arms of 2.5 million kilometres, and what LISA Pathfinder proved in 2016 close the program with a look at where the field is going and where the student can go next.

Preview: the beginning of the unit

The detectors of Unit 15 hear waves with frequencies between about 10 and 1,000 hertz, which is the band of merging stellar black holes and neutron stars. Most of the universe’s gravitational radiation lies elsewhere. Two black holes of a thousand million solar masses circle each other once in a few years and send out waves with periods of years, far below anything a 4 km arm can sense, and the millions of white dwarf pairs of the Milky Way radiate at millihertz frequencies that no instrument on the ground will ever reach. This unit presents the two detectors built for those bands: an array of millisecond pulsars spread across the Galaxy, which reported evidence for a background of nanohertz waves in June 2023, and the three spacecraft of LISA, adopted by the European Space Agency in January 2024 for a launch in 2035. It closes with the other frontiers of the field and with the question of where a student who has reached the end of this program can go next.

1 The gravitational-wave spectrum

A gravitational wave from a binary has the frequency \(f_{\mathrm{gw}} = 2f_{\mathrm{orb}}\) of Unit 13, and Kepler’s third law ties the orbital frequency to the mass and the separation of the pair. The largest frequency a binary can reach is set by the last stable orbit of Unit 14, \(f_{\mathrm{gw}} \approx c^3/(6^{3/2}\pi G M)\) for a total mass \(M\), which is \(4{,}400\unit{Hz}\) for \(M = \Msun\) and falls in proportion to \(1/M\). A pair with a total mass of \(10^{6}\,\Msun\) therefore never radiates above about \(4\unit{mHz}\), and a pair of \(10^{9}\,\Msun\) in total never above about \(4\unit{\mu Hz}\); a few thousand years before they merge, two black holes of that size radiate at tens of nanohertz, with a period of about a year. Figure 1 is read with this rule from right to left: the heavier the source, the lower its frequency. Each band of frequencies needs its own kind of detector, as the rest of this section explains.

The gravitational-wave spectrum as characteristic strain against frequency, over thirteen powers of ten. The three solid curves are the sensitivities…
Figure 1. The gravitational-wave spectrum as characteristic strain against frequency, over thirteen powers of ten. The three solid curves are the sensitivities of the detectors of this unit and of Unit 15: the NANOGrav 15-year array (data file of the NANOGrav noise-budget paper, Zenodo record 8092346), LISA (sky-averaged sensitivity of Robson, Cornish, and Liu, Classical and Quantum Gravity 36, 105011, 2019), and the design sensitivity of Advanced LIGO (analytic fit of Sathyaprakash and Schutz, Living Reviews in Relativity 12, 2, 2009). The grey lines are sources computed from the formulas of Units 13 and 14: the background of supermassive black hole binaries with the amplitude measured in 2023, the inspiral of two black holes of a million solar masses at a redshift of about one, the inspiral of GW150914, and the white dwarf pair HM Cancri. A source is detectable where its line lies above the detector curve.

The vertical axis of Figure 1 is the characteristic strain \(h_c\), a quantity chosen so that one plot can hold both the noise of a detector and the signal of a source. For a detector it is built from the amplitude spectral density \(\sqrt{S_n(f)}\) of Unit 15, which has the units \(\unit{Hz^{-1/2}}\), and the characteristic strain of the noise is \(\sqrt{f\,S_n(f)}\), a dimensionless number. For a source it is the strain amplitude multiplied by the square root of the number of cycles the wave spends near the frequency \(f\), because the matched filter of Unit 15 gains in signal-to-noise ratio in proportion to that square root. The number of cycles differs enormously between the bands, which is why the figure is drawn this way: a background with a period of a year completes only about fifteen cycles in a fifteen-year data set, while a white dwarf pair at millihertz frequencies completes hundreds of thousands in a few years. The ratio of the source line to the detector line at each frequency is then roughly the signal-to-noise ratio that frequency contributes, which is why the figure can be read at a glance. Table 1 lists the three bands of the figure with their sources.

The full unit is part of the program

Want to see a complete unit first? Unit 3, Simultaneity, Time Dilation, and Length Contraction, is free to read.