Thursday, 11 February 2016


Today we announced it to the world: we did it! We detected gravitational waves!

This has taken some time to prepare. Since the 0.2-second-long blip appeared in the data in mid-September it has taken five months to verify and write up the results. Before that, it took two decades to build the LIGO detectors. But that’s nothing. The event we observed — two black holes whirling around each other before merging into one — occurred over a billion years ago.

In the same spirit of advance preparation, I wrote this in January, so that today I could concentrate on more urgent matters. Like partying. I wrote quite a lot on what we have found, and what it means, and also what it has been like to be part of a major scientific discovery. (The word “major” is meant as an understatement, in an effort to pretend that we’re cool about all this. We’re not.) I'll post more over the next few days. But for now...

There are in fact three incredible observations here.

First, we observed gravitational waves. That’s what we’ve been trying to do for decades. It had to happen eventually, so long as the detector was as sensitive as promised. (It wasn't -- it was better!) The only question was when, and when depended on just how many loud sources there are out there. There could have been so few that we would have had to run the detectors at their most sensitive for five years before we found anything — which could have been as late as 2025. Or there could be so many that we found something in the first three months. What we didn’t expect was that we would detect a signal before we even started. More about that later.

For gravitational-wave enthusiasts, the big question was not whether we would ever detect gravitational waves, but what the source would be. Producing a gravitational wave that we can detect requires a stupidly violent, intense astronomical event — we need massive objects moving extremely fast in tiny spaces. A likely option would be two neutron stars in close orbit around each other. Take the sun and squash it into a ball twenty kilometres across, and you’ve got something pretty much like a neutron star. That is a very small, dense star. Since neutron stars are so small, they can orbit each other very close without touching — so close that they can orbit hundreds of times every second. That is fast. We now have two massive objects moving very fast in a small space -- a perfect gravitational-wave source. And the great thing about neutron-star binaries is that we have seen several already in the universe. None in such tight orbits as I just described, but we assume that we have only seen a tiny fraction of the total number out there.

In principle we can do better than that. We can put two black holes in orbit. If the sun were a black hole, it would be only three kilometres across. Now that is dense! Also, neutron stars are limited to be no more than twice as massive as the sun. For black holes there is no limit. And the more massive they are, the stronger the signal. The only problem with binary black holes is that we have never seen any. Sure, that’s not so surprising, given that black holes are, well, black, but that’s no excuse when you’re asking for a billion dollars to build a detector. You either know the source is there, or you don’t. And when it came to binary black holes, we just did not know for certain if any existed [1].  

Until now. LIGO just heard two of them crash into each other. They were both about thirty times more massive than our sun. (To be precise, which is what we’ve spent the last five months struggling to be: one was 36 times the mass of the sun, and the other 29. Give or take a few suns.) To make one of these black holes you would have to take thirty of our suns, and squash them together, and squash them all into a ball about 200 kilometres across. Don’t forget, we didn’t just find those two black holes — we also found the black hole that resulted from the two of them eating each other. In fact, we witnessed the millisecond-long gulp. The final black hole is about 60 times the mass of our sun. During the last rapid orbits (we heard the two black holes circle each other about five times in 0.2 seconds) and merger, the binary gave off three suns worth of energy. What that means is, if you took three copies of our sun, and turned them all instantly into pure energy in an incredible nuclear reaction, that would be the amount of energy released when these two black holes collided. In terms of energy, this makes it among the brightest events humanity has ever observed. If this energy had been released as light, for an instant it would have been the brightest thing in the sky. Instead, the energy was locked in gravitational waves, which interact so weakly with the rest of the universe that they passed through the earth completely unnoticed -- except, of course, for the blip that peeked above the noise of the massive experiment built with the express purpose of noticing it.

The signal in both detectors.
Top: The data
Middle: General-relativity calculations of the signal from two black holes.
Bottom: The data minus the predicted binary-black-hole signal. 

Let’s recap. We’ve just directly observed gravitational waves for the first time. Also the first ever binary black hole. Also black holes tens of times as massive as our sun. Anything else?

Well, as a matter of fact, yes. After its monumental collision, the final black hole bled away its scars as a last dying dribble of gravitational waves. Immediately after merger the black hole was like a big distorted spinning spacetime jelly. It quickly settled down into the squashed-ball shape of a spinning black hole (not as round as a ball, but not as a flat as a pancake), and as it settled down it released more gravitational radiation, like a bell ringing as it settles down after being struck.

Why is that radiation so interesting? Encoded in it are the properties of the final black hole. More than that, this signal is the closest we can imagine to measuring a signal from a black hole. We have never seen a black hole, but we have heard one.

That is incredible observation number three: we have measured a signal from a black hole.

In the future, when we find much stronger signals, we will be able to use that part of the signal alone to probe in detail the physics of black holes. That used to be our dream, but no more. It won’t be long now.

This all comes back to the point I hinted at earlier: we saw this signal before we even started looking. The first run of the detector, where we would officially take “science data” to analyse, started on September 18th. Two weeks before there was a final “engineering run”, a dress rehearsal where data was collected, verified and analysed, to make sure that all of the equipment and processes were working correctly and in place. That was when the signal was observed — on September 14th, during the last engineering run.

Needless to say, the beginning of the official “observing run” was a little more hectic than originally planned. But data continued to be collected until January 12th. Much of that data is still being analysed. As you will have realized by now, if there were more signals in that data, we already know about them. But just as before, the details will have to wait until the analysis is finished.

Where do I fit into all this? I just happen to study binary black holes. I assumed I was going to have to wait a lot longer before my work would be of any use at all. We all knew that the first observations would be neutron-star binaries. Maybe a binary where one object was a neutron star and the other a black hole. But binary black holes? We didn't even know for sure if they existed.

So the first detection was a very pleasant surprise indeed. And it gives me a perfect opportunity to tell you all about binary black holes, and how we study them. Stay tuned!

Next: What it feels like to detect gravitational waves.

More science: How to decode gravitational waves from black holes.

Explanations: Why bother trying to explain gravitational waves?
                         Is spacetime really curved?


1. We know that some black holes exist, of course. We can see other stars orbiting around them. Unfortunately, that only tells us that individual black holes exist, not whether they ever end up in binaries. We also have not previously found any especially big black holes. We know that black holes a few times more massive than our sun exist. But what about fifty times bigger? A hundred times bigger? We have no idea. (We do know that extremely massive black holes exist, millions and billions of times more massive than our sun. But there was only minimal evidence for black holes in the "intermediate" range, tens of times the mass of the sun.)


  1. Congratulations! (also to the engineers who built the marvellous machines)

  2. Absolutely awesome.
    Please keep the posts coming!
    I still cannot fathom how you can have a resolution of E-18 fathom.
    And, pray tell, what are the odds of 2 supermassive ones colliding? Would that break the detector? :)

    1. The mirrors only move a fraction of the size of an atomic nucleus, which you'd expect would be impossible to measure -- but there are many moles worth of atoms in one mirror, so on average we can measure the movement to very high precision.

      Supermassive black holes are millions of times more massive, and so the merger signal will be at a frequency millions of times lower, i.e., outside the sensitivity band of the detector. That's what the space-based LISA detector is supposed to solve -- it would be sensitive to much lower frequencies.

    2. Crystal clear!

  3. Is there an estimate of h+ and hx at the source?
    Or you cannot reconstruct that with only 2 detectors?

    1. Gravitational waves are only defined far from the source (strictly speaking, at infinity, but 1.3 billion light years is a decent approximation), so there is no estimate of h+ and hx at the source.

      If you're asking about whether we know h+ and hx separately at the detectors (which is more likely), in principle we can, but not for this source, due to a combination of its sky position and orientation. If you look at Figs. 2 and 4 of the parameter estimation paper, you'll see that these are not measured well.

    2. Thanks, Mark. Somehow I cannot access the link on DCC.
      Is this paper on arxiv?

    3. It's here. The other link doesn't work for me, either, and I can't edit the comment, so this is the link to use.

  4. This comment has been removed by a blog administrator.

    1. This comment has been removed by the author.

  5. I thoroughly enjoyed reading this. Thanks Mark.

    It sounds like the LIGO team have also confirmed that general relativity not only explains weak field gravity - but also gravity when it's at its most dynamic. Which is also quite remarkable!

    1. It's the most extreme test of GR so far (or if you prefer, the test that considers GR at its most extreme). I don't think anyone expected it to fail for signals of this strength, though. The real test will be when we have detectors like the Einstein Telescope, where we can hope for much stronger signals, and can put really tight constraints on the theory. That's right, Einstein -- we're coming after you!

  6. The Einstein telescope sounds very interesting. It'll be exciting now to see which experiment takes the first step towards quantum gravity. It's starting to look as though we're not going to go beyond the standard model at the LHC. But, would love to be wrong.

  7. Can you comment more on the residual in your plot (also in the PRL)? It's the difference between the num. rel. evolution and the data right? And so it has amplitude of roughly "0.5" (presumably also 10e-21 as the top frame is). But that seems too big based just on eye-balling the top frames...I'd expect a maximum residual of maybe 0.2.

    1. Yes, the residual is the difference between the data and the general relativity waveform; it is equivalent here to use either a specific NR waveform, or a waveform model with parameters within the uncertainty ranges that we found. It's hard to guess by eye what the residuals should be -- near the signal peak, for example, small differences in amplitude and phasing between the data and the model, which are not easy to determine in the figure, could still give a residual as large as 0.5.

      I haven't looked at this myself, but there is a set of python notebooks here, where you can play with the data analysis yourself.

    2. Thanks for the link. Very interesting and I'm glad that the LSC has made it available. However, it doesn't, as far as I can tell by searching in the page for 'residual,' show how the residual was computed. I suppose what I'm after would be some sort of normalized residual. Afterall, a naive look at the plot might suggest as high as a 50% residual, but I'm sure that it must be much smaller.

    3. The important thing is that the residual looks like noise -- it averages out to close to zero. If you match the original data against possible GW signals, you'll find an agreement with a signal-to-noise ratio of 24 for the BBH signal we found. If you do the same test against the residual, you'll find nothing significant -- subtracting the theoretical BBH waveform has removed any meaningful signal from the data.

  8. I was amazed by the engineering feat of measuring a strain of 10^{-21}.

    Many of the other numbers are equally astonishing. For instance, the energy flux at earth can be estimated using the radiated source energy as E=3 M c^2 (M = solar mass) and r = 1 billion light years. Assuming a signal of length \Delta t =0.1 sec, then the average flux (energy per unit time per unit area) passing you and I at 09:50:45 UTC on Sept 14, 2015 was F=E/(4 \pi r^2 \Delta t), which if I got the numbers right, is about 10^{-2} W m^{-2} = 10^1 erg cm^{-2}.

    This is a HUGE energy flux. The equivalent flux of energy in every day life would correspond to jet level noise close to the engine (sound pressure level of >100 Pa), enough to blow your ear drums.

    This begs the question: how could this flux just pass through us? Is it that the space-time metric sort of gulped, but we as inhabitants of the metric did not feel the energy flow through us?

    1. The energy flux from the GWs is indeed huge. If the same energy had been in light, it would have been among the brightest events we have every observed. But gravity interacts extremely weakly with matter, and the GWs impart very little of their energy. The detectors do not absorb energy from the GWs, only measure the changes in lengths due to spacetime being stretched and squashed.

    2. Mark: The detectors do absorb energy, but its negligible.

    3. Sorry, I should have said that it was negligible. The first sentence ("impart very little of their energy") was right, but in the second it seems I made a zeroth order approximation, i.e., a mistake.

  9. Mark, probably this is discussed in one of the papers, but were the detectors sensitive enough in S5 and S6 to see this event if it happened then? I would think yes (from fig 4 of 1602.03844) which shows that the detectors were sensitive to a distance 10 times that of GW150914.
    Also were any bar detectors operating at the time of GW150914?

    1. The Initial LIGO detectors were not sensitive enough. This same event would have had an SNR of less than 8 (I can't remember the exact value off the top of my head, somewhere between 4 and 6, I think), and that would not have been registered as a detection, or even a "trigger" for a possible event. One of the things that made it easier to believe we had seen something in the later weeks of September was that the sensitivity volume (over time) of Advanced LIGO in the weeks of the engineering runs surpassed the volume of all of S5 and S6 put together.


[Note: comments do not seem to work from Facebook.]