Echoes strike back!

Black Hole echoes, as rendered by Kaća Bradonjić (@physartphil)




In a post last year, I talked about our search for “echoes from the abyss” with Jahed Abedi and Hannah Dykaar in the Advanced LIGO gravitational wave observations, which are smoking guns for Planck-scale structure near black hole event horizons. Such structures are not expected in classical General Relativity, but may be motivated by various versions of the black hole information paradox such as the fuzzball models of black holes in string theory, or the infamous firewall paradox. Most surprisingly,  we found that the evidence seen for (a toy model of) echoes in first observational run of LIGO data can only be seen in <1% of random noise realizations.




While this was widely covered by science journalists (see here for some links), two of my favourites were by Zeeya Merali on Nature News and Sabine Hossenfelder @skdh on Aeon.

Headline from Sabine's article about "Echoes from the Abyss"


And then, there was the response from the community. While the theorists were beside themselves with excitement (our two papers are cited close to a hundred times in 16 months), we got a long silence from observers/experimentalists with one exception. A group of of LIGO collaboration members in Albert Einstein Institute (AEI) in Hanover were quick to express (a very healthy and deserved) scepticism. We responded to their comments in our “Holiday Edition!”, but the real question was whether our results could be reproduced by people who had more experience with LIGO data.

The latter was a long time coming, partly because big collaborations work slowly, and partly because of other exciting discoveries (such as the first binary neutron star merger, seen by the LIGO collaboration). In the meantime, I also organized a workshop on “Quantum Black Holes in the Sky?”, along with Vitor Cardoso and Samir Mathur, where we discussed various observational and theoretical aspects of  black hole echoes, and more.


The AEI group finally released their analysis last December (which was later updated with a 4th black hole merger event in February), and lo and behold, they found that the evidence seen for echoes (using the same dataset and model), is only seen in 2±1% of random noise models (i.e. within 1σ of what we reported). Surprisingly though, they went on to say

“The reduced significance is entirely consistent with noise, and so we conclude that the analysis of Abedi et al. does not provide any observational evidence for the existence of Planck-scale structure at black hole horizons.” !!!

Needless, to say that this didn’t make too much sense to us, a point that I made publicly on arXiv, and on a (let’s say less diplomatic) exchange on facebook, with Thomas Dent.

What they really are saying is that, since General Relativity has been such a successful theory for the past 100 years, we don’t really think echoes are there, and we need really strong evidence (e.g., p-value of 10-6) to be convinced otherwise. Fair enough, but that is a very subjective statement. Someone like me may argue that we have known for nearly fifty years that if you consider quantum mechanics, something funny is happening at the black hole horizons. Why is the entropy proportional to horizon area? How could information get out of the horizon of evaporating black holes? We can also explain the scale of dark energy (the infamous 10-120), by assuming a “quantum equilibrium” at the horizons of astrophysical black holes.  So for me the bar is probably lower. Therefore, it is important to separate the objective statistical statements (e.g. p-value) which only depend on data, from subjective “priors” that varies from theorist to theorist.

For more, have a look at Sabine’s update on the state of echoes and controversy in Quanta Magazine.


James O’Brien for Quanta Magazine

Black Hole Echology

Out of the “tentative evidence” for echoes and the resulting controversy, emerged the need for a clearer understanding of what echoes should really look like. In a recent arXiv preprint: “Black Hole Echology: The Observer’s Manual”, Qingwen Wang and I provided the most comprehensive study of echo templates, and their model dependence (and independence), for a spinning black hole.

Nominal echoes from GW150914 event, for different wall distance to the horizon

Qingwen Wang









We made some surprising discoveries, e.g., that the echoes decay as a power-law ∝ time-4/3, not exponentially, as we had originally assumed, or that the signal below the superradiance frequency is insensitive to initial conditions. Echoes_Western_Mar_2018.001

These findings set the stage for the new observational evidence for echoes that was about to emerge.

Echoes Strike Back!

There is a quote that I have often heard in Physics and Astronomy gatherings, but I don’t know who it can be attributed to:

“If you have to argue about statistics, it means that you need more data.”

and indeed that will be the way to unambiguously settle the argument about the significance of black hole echoes.

First came the surprising results by Conklin, Holdom, and Ren from University of Toronto, who developed a “model-agnostic” search for echoes by cross-correlating data from the two detectors and looking for periodic signals. This was very complementary to our original search, as it assumed very little about the exact template, but looked for repeating echoes that lasted much longer.  Indeed, they think they see echoes in 5 of the LIGO/Virgo events that we did not find (or look for) echoes in, with p-values 0.2%-0.8% (roughly 3σ evidence).


The grand finale came about, after I gave a talk about echoes at Yukawa Institute in Kyoto, during the CosPa 2017 meeting.


During the meeting, both Shinji Mukohyama and Lam Hui asked me whether I expected to see echoes from the binary neutron star merger GW170817, which had made headlines a couple of months earlier. I first dismissed the idea as it was a very different frequency regime from what we had for binary black holes, and given the lack of any detectable post-merger signal by LIGO/Virgo, it wasn’t even clear when a black hole remnant would form, if at all.




However, it then occurred to me that we might have an opportunity to probe a very different regime, consisting of the first few harmonics of the echo chamber. This is at too low a frequency for binary black hole mergers, but is squarely within the LIGO sensitivity band. Indeed, a simplified version what the Toronto group did, with proper inverse noise weighting, gave a huge and surprising signal for echoes at 4.2σ (or p-value of 10-5) exactly where you expected for the mass and spin of a black hole remnant of GW170817 binary neutron star merger.

cross-spectrum of the two LIGO detectors (after Wiener filtering and summing over integer multiples of the frequency). The peak at 72 Hz and 1 second is the most significant in the diagram, at 4.2σ.
Same cross-power spectrum as above, but plotted at 1 second after the merger. Blue region shows the expected range for echo frequency.

For more details, you should read our paper, watch my talk, and/or see my presentation slides.

So, is this just a statistical or systematic fluke? Or, could it be the beginning of the end for the black hole information paradox, as well as our first crack at the quantum gravity nut? Only time will tell.

One thought on “Echoes strike back!

Comments are closed.