Skip to main content
Sandbox Physics
Black hole ringdown and the no-hair test

From two tones to a no-hair test

Sandbox Physics ·

Two measured tones give four numbers, two frequencies and two damping times, but a Kerr black hole has only two, its mass and spin. The no-hair test asks whether one mass and one spin predict all four. In practice one lets the second tone's frequency drift by a fraction δf\delta f and checks whether the data still favour δf=0\delta f=0. A passed test bounds how large a departure from Kerr can be; a single failed fit is first a question about noise, start time and missing modes.

Open the black hole ringdown Lab

This chapter builds on fitting a ringdown. It uses the grid posterior and profile likelihood from there.

What the no-hair statement claims

In general relativity a black hole that has settled down is a Kerr black hole, fixed by its mass MM and angular momentum J=χM2J=\chi M^2, with electric charge negligible for astrophysical holes. Every quasinormal frequency is then a function of those two numbers alone, ωlmn=ωlmn(M,χ)\omega_{lmn}=\omega_{lmn}(M,\chi). Dreyer and colleagues turned this into a test: measure several tones and check that one (M,χ)(M,\chi) fits them all. Another theory of gravity, or a different kind of compact object, would in general break that agreement. Dreyer et al., Black hole spectroscopy: testing general relativity through gravitational wave observations, arXiv:gr-qc/0309007v1

Two tones over-determine the hole

One mode gives two measured numbers, ff and τ\tau, and two unknowns. It can always be fitted by some (M,χ)(M,\chi) and so tests nothing. A second mode adds two more numbers with no new unknowns:

2πf220 GMc3=Re Mω220(χ),GM/c3τ220=∣Im Mω220(χ)∣,2πf221 GMc3=Re Mω221(χ),GM/c3τ221=∣Im Mω221(χ)∣.\begin{aligned} 2\pi f_{220}\,\frac{GM}{c^3}&=\mathrm{Re}\,M\omega_{220}(\chi), & \frac{GM/c^3}{\tau_{220}}&=\lvert\mathrm{Im}\,M\omega_{220}(\chi)\rvert,\\ 2\pi f_{221}\,\frac{GM}{c^3}&=\mathrm{Re}\,M\omega_{221}(\chi), & \frac{GM/c^3}{\tau_{221}}&=\lvert\mathrm{Im}\,M\omega_{221}(\chi)\rvert . \end{aligned}

Each mode on its own defines a point in the (M,χ)(M,\chi) plane. The test is whether the two points coincide within their uncertainties.

Worked example 1. Take the tones of a Kerr hole with M=68.5 M⊙M=68.5\,\mathrm{M}_\odot and χ=0.69\chi=0.69: Q220=3.254Q_{220}=3.254, f220=249 Hzf_{220}=249\,\mathrm{Hz} and Q221=1.052Q_{221}=1.052, f221=244 Hzf_{221}=244\,\mathrm{Hz}. Invert each with the Berti, Cardoso and Will fits. For 221221 the coefficients are q1=0.1q_1=0.1, q2=0.5436q_2=0.5436, q3=−0.4731q_3=-0.4731 and f1=1.3673f_1=1.3673, f2=−1.026f_2=-1.026, f3=0.1628f_3=0.1628. Berti, Cardoso and Will, On gravitational-wave spectroscopy of massive black holes with the space interferometer LISA, Table VIII Then (1−χ)−0.4731=(1.052−0.1)/0.5436=1.752(1-\chi)^{-0.4731}=(1.052-0.1)/0.5436=1.752, so χ=0.694\chi=0.694 and M=69.1 M⊙M=69.1\,\mathrm{M}_\odot. The fundamental gave χ=0.692\chi=0.692 and M=68.9 M⊙M=68.9\,\mathrm{M}_\odot. The two points agree to within the fits' stated errors of one to two percent, as they must for a Kerr input.

Parameterising a departure

Real data do not give four sharp numbers. The standard approach keeps the Kerr model but frees the overtone:

f221=f221Kerr(M,χ) (1+δf1),τ221=τ221Kerr(M,χ) (1+δτ1).f_{221}=f_{221}^{\mathrm{Kerr}}(M,\chi)\,(1+\delta f_1),\qquad \tau_{221}=\tau_{221}^{\mathrm{Kerr}}(M,\chi)\,(1+\delta\tau_1).

The fundamental pins down (M,χ)(M,\chi), and the overtone then tests the prediction. Isi and colleagues used exactly this form for GW150914, with flat priors δf1∈[−0.5,0.5]\delta f_1\in[-0.5,0.5] and δτ1∈[−1,1]\delta\tau_1\in[-1,1]. Isi, Giesler, Farr, Scheel and Teukolsky, Testing the no-hair theorem with GW150914, arXiv:1905.00869v2 They found the damping time poorly constrained, and that fixing δτ1=0\delta\tau_1=0 left the frequency result unchanged. The Lab therefore frees only δf1\delta f_1 and computes

p(δf1∣d)∝∑M,χLprof(M,χ,δf1),p(\delta f_1\mid d)\propto\sum_{M,\chi}L_{\mathrm{prof}}(M,\chi,\delta f_1),

on a grid of 4141 values of δf1\delta f_1 from −0.5-0.5 to 0.50.5, 4141 masses and 3434 spins.

Reading the interval

Table 4.1 Overtone frequency shift in the Lab presets, synthetic Kerr records, median and 5 to 95 percent range.
Presetρ\rhoδf1\delta f_1
Overtone needed at the peak14140.06  [−0.44, 0.46]0.06\;[-0.44,\,0.46]
A loud event80800.09  [−0.06, 0.31]0.09\;[-0.06,\,0.31]
A fast-spinning remnant30300.11  [−0.38, 0.48]0.11\;[-0.38,\,0.48]
A faint overtone3030−0.02  [−0.45, 0.41]-0.02\;[-0.45,\,0.41]

Every record here is Kerr, so the true value is 00, and every interval contains it. At ρ=14\rho=14 the interval fills almost the whole prior range: the data barely constrain the overtone's frequency. Only the loud event narrows it to a band of about 0.370.37.

For GW150914 Isi and colleagues report δf1=−0.05±0.2\delta f_1=-0.05\pm0.2 at 68 percent credibility. They also give a Bayes factor of 1.751.75 for the no-hair model over the free model. Isi, Giesler, Farr, Scheel and Teukolsky Our synthetic numbers are not a check on theirs. They use two detectors, coloured noise and sampled amplitudes; we use one white-noise record and a profile likelihood.

Worked example 2. A Bayes factor multiplies prior odds into posterior odds. With even prior odds, a factor of 1.751.75 gives a no-hair probability of 1.75/(1+1.75)=0.641.75/(1+1.75)=0.64. That is a mild preference: the data sit comfortably with Kerr and still leave room for a departure.

Why one deviation is not new physics

Suppose a fit returns an interval that excludes δf1=0\delta f_1=0. Check these causes before anything else:

A departure becomes interesting only when it survives changes of start time, noise estimate and mode content, and when other events show the same sign and size.

The overtone debate as a case study

The GW150914 overtone shows how a single analysis turns into a long discussion. Isi and colleagues reported the overtone at 3.6σ3.6\sigma in 2019, in a paper published in Physical Review Letters. Cotesta and colleagues found no evidence for it after the peak and argued the earlier claim was noise-dominated. Isi and Farr replied that they could not reproduce that result and that their measurement was robust to analysis choices. Isi and Farr, Revisiting the ringdown of GW150914, arXiv:2202.02941v2 arXiv lists no journal publication for either of the 2022 papers. Both nonlinear-mode papers appeared in Physical Review Letters in 2023, but they analyse simulations, not detector data. Whether quadratic modes can be measured in a real event is an open question.

Lab tasks

  1. In the default preset, read δf1\delta f_1 and compare its width with the prior range [−0.5,0.5][-0.5,0.5].
  2. Switch to the loud event. Check that zero is still inside and measure how much the interval has narrowed.
  3. Step through noise draws 11 to 2020 in the advanced group at ρ=80\rho=80, and count how many intervals exclude zero.

Exercises with explained answers

  1. Why does a single damped sinusoid test nothing about the no-hair theorem?
  2. Suppose one record gives δf1=0.30  [0.05, 0.48]\delta f_1=0.30\;[0.05,\,0.48] at t0=0t_0=0 and 0.02  [−0.40, 0.45]0.02\;[-0.40,\,0.45] at t0=2Mt_0=2M. What do you conclude?
  3. With prior odds of 1:101:10 against the no-hair model, what posterior probability does a Bayes factor of 1.751.75 give?
  4. Isi and colleagues give δf1=−0.05±0.2\delta f_1=-0.05\pm0.2 at 68 percent. If the posterior were Gaussian, what would its 90 percent interval be, and how much of the prior range would it cover?
Answers
  1. Two measured numbers, ff and τ\tau, can always be matched by some (M,χ)(M,\chi) on the Kerr curves. A test needs more measured numbers than unknowns.
  2. The departure does not survive a small change of start time, so it is more likely a fluctuation or an early-time effect than a property of the remnant. Repeat with other noise draws and start times before saying anything else.
  3. Posterior odds are 1.75×0.1=0.1751.75\times0.1=0.175, so the no-hair probability is 0.175/1.175=0.150.175/1.175=0.15. The Bayes factor alone cannot overturn strong prior odds.
  4. A Gaussian's 90 percent half-width is 1.6451.645 times its 68 percent half-width, so 0.2×1.645=0.330.2\times1.645=0.33 and the interval is about [−0.38, 0.28][-0.38,\,0.28]. That covers two thirds of the prior range [−0.5,0.5][-0.5,0.5]. The real posterior need not be Gaussian, and the paper's Fig. 4 shows its 90 percent bounds directly.

To see these steps on public data, follow recomputing ringdown tones, a synthetic signal and a slice of GW150914.

Try it in the Lab

Compare the overtone frequency-shift curve in the default and loud-event presets; note how much narrower it gets and whether it still contains zero. Open the black hole ringdown Lab

Sources

  1. Dreyer et al., Black hole spectroscopy: testing general relativity through gravitational wave observations, arXiv:gr-qc/0309007v1
  2. Berti, Cardoso and Will, On gravitational-wave spectroscopy of massive black holes with the space interferometer LISA, Table VIII
  3. Isi, Giesler, Farr, Scheel and Teukolsky, Testing the no-hair theorem with GW150914, arXiv:1905.00869v2
  4. Cotesta, Carullo, Berti and Cardoso, Analysis of ringdown overtones in GW150914, arXiv:2201.00822v2
  5. Mitman et al., Nonlinearities in black hole ringdowns, arXiv:2208.07380v2
  6. Cheung et al., Nonlinear effects in black hole ringdown, arXiv:2208.07374v3
  7. Isi and Farr, Revisiting the ringdown of GW150914, arXiv:2202.02941v2

Keep reading

What this page links to directly

See this series on the content map →

Black hole ringdown and the no-hair test · Reading path

From a ringing black hole to a no-hair test

  1. ArticleWhy does a newly merged black hole ring like a bell?
  2. ArticleDamped tones and the Kerr spectrum
  3. ArticleWhy a ringdown fit needs the overtone near the peak
  4. You are hereFrom two tones to a no-hair test
  5. ArticleRecomputing ringdown tones, a synthetic signal and a slice of GW150914
  6. ArticleHow to read the 2019 GW150914 overtone paper and the debate that followed
Something broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy