Friday, March 21, 2014

Direct or indirect?

This week has seen the potentially momentous result from the BICEP2 experiment indicating the detection of gravitational waves from the inflationary era of the Universe, just a tiny fraction of a second after the Big Bang. It's a fantastic result, and if/when confirmed by other experiments (e.g., Planck) will be huge leap in developing our understanding of the beginnings of the Universe. Many other people have discussed the background (that's just a scattering of a few of the many links to some scientific and more general descriptions of the results) and potential implications of the results, and a few areas for some considered scepticism, but I wanted to briefly talk about whether this classes as a direct or indirect detection of gravitational waves. I'm mainly interested in this because, to be clear up front, I'm part of a large scientific collaboration (the LIGO Scientific Collaboration [LSC]) that is currently trying for direct gravitational wave detection using a set of specially designed detectors/observatories (LIGO, Virgo and GEO600) here on Earth. I should also point out the views I'm giving are entirely my own and definitely not those of the LSC.

I should note that on BICEP2's FAQ the word "direct" gets used in the answer to the question "Have you detected a gravitational wave?" to which they answer "The frequency of the cosmic gravitational waves is very low, so we are not able to follow the temporal modulation. However, we are indeed directly observing a snapshot of gravitational waves through their imprints on matter and radiation over space." Whether this fits into my description of direct or indirect below is another question!

What do I mean by indirect or direct detection? Well in 1993 Hulse and Taylor won the Nobel Prize in Physics for their earlier observation of a pulsar in a neutron star binary system, which was losing energy exactly as predicted through the emission of gravitational waves. This has always been said to be an indirect detection of gravitational waves, i.e., it wasn't physically measuring the waves themselves, but was inferring their presence through the energy they carry away as observed by the binary system's evolution (since their original observations this effect has been measured in many other binary neutron star systems, which also provide other tests of general relativity). With the gravitational wave detectors (such as the aforementioned LIGO, Virgo and GEO600) they aim to directly detect the waves by actually seeing their effect in stretching and squeezing the distance between parts of the detectors. So, the former uses some observations to measure the properties of a source (the orbital evolution of a binary system) and from that infer the presence of gravitational waves, whilst the later directly measures their effect within a detector system. [On a slight aside there could be much discussion on the semantics of "direct" observation/detection - in pretty much all observations (including a persons senses) you could say that you're variously removed/abstracted by a number stages from directly measuring/experiencing the effect of something. In scientific observations it's pretty much always the case that you're having to use proxies to convey some information to you. In most astronomy photons are counted by a CCD, processed by a computer and then displayed, whilst in particle physics you're often measuring the decay of one particle through the products it produces, which themselves are relayed to you through tracks left on silicon detectors, or energy deposited in calorimeters. However, in most cases using "direct" observation/detection is probably a fair term.] 

So, in the case of the BICEP2 results, where they're measured the imprint of gravitational waves in the cosmic microwave background (CMB), where does that fit on the scale (if there is some scale in between!) of direct or indirect detection? Initially I was biased against calling this a direct detection. As mentioned above this is mainly due to working as part of a collaboration hoping to soon directly detect gravitational waves with ground-based detectors. I (not wanting to speak for the rest of the collaboration) would like us to be the first to claim a direct detection, so there's a level of guardianship (or unjustified feeling of ownership!) over that claim. However, I think (obviously I'm not the sole arbiter) the CMB measurements deserve the right to be called more than an indirect detection, so for now I'll go with the compromise of semi-direct detection (as used by Andrew Jaffe here).

So, why not indirect? Well, the gravitational waves that are observed in the CMB have (redshifted) frequencies of order 10-17 Hz, which corresponds to wavelengths of ~1 Gigaparsec. To measure such waves you'd need a detector about the size of the Universe. There's obviously no way you could build a physical detector to measure that, so using the CMB's the only way to do it - it is the only "detector" you could have available. In this sense they don't seem to fit with the indirect pulsar binary system paradigm above. [Note that there are also efforts to measure gravitational waves with frequencies around 10-9 Hz using astrophysical objects (in this case pulsars) as the components of a "detector".]

But, why only semi-direct then? This is maybe a technicality that could be argued over, but I suppose it comes down to the basic fact that despite the CMB being the only way to perform the measurement of ultra-low frequency waves you still aren't physically measuring the wave in a detector on Earth (another example might be dark matter, who's effects are imprinted in various astronomical observations, but you still want to see them in a detector on Earth to claim detection). You're also having to use the effect of the gravitational waves on density perturbations, which in turn affect the light intensity, which then affects the CMB polarisation signal received; in a laser interferometric detector the gravitational wave affects the position of mirrors, which in turn effect the phase of reflected and detected light, which you could argue (an I may be pushing it here) is a step less removed than the case with the CMB. There's also the case (which may not be entirely relevant in a direct/indirect argument) that given that the CMB polarisation signal (by the very nature of how it had to be formed during a short period in recombination when photons could diffuse far enough that they would encounter different temperature regions, but that there were still enough free electrons to scatter off and give a polarisation signal) was imprinted within a short space of time, it is just a single snapshot of the gravitational wave signal. Gravitational wave detectors on the other hand (including those using pulsars) are able to measure the variations as the waves pass them, so give a complete time series of the signal. My hand wavy analogy (also implied on the BICEP2 FAQ) is that the CMB measurement is like seeing a photograph of the shadows of water waves on a ripple tank, whereas gravitational wave detectors are like continuously measuring the position of a cork floating on top of the tank.
Shadows of waves on a ripple tank. Analogous to the imprints of gravitational waves in the CMB polarisation? [Credit]
Whether the BICEP2 result is indirect, direct or semi-direct detection of gravitational waves it doesn't take away from the fantastic work they've done and it's still an amazing feat of observation and analysis.

Anyway, that's my view. What do you think?


  1. I think an interferometer is more direct, but that directness doesn’t matter too much. The BICEP2 result is like measuring temperature through analysing ice cores, compared to looking at a thermometer. (best analogy I could ad lib there ;) They’re both methods of ascertaining “temperature”, but the context is totally different and the results contribute to our knowledge in different ways.

    An alternative way to distinguish might be “local” and “non-local”, with Hulse/Taylor and B-mode detections being non-local. Admittedly “LIGO makes first local detection of GWs!” isn’t the best headline ever.

  2. Disagree. The Hulse-Taylor-Weisberg detection is more 'direct' than BICEP, if anything. Both Hulse-Taylor and BICEP aren't measuring a passing GW, they are measuring a natural phenomenon (orbital frequency evolution or B-mode polarization) which is supposedly caused by GW. Then the question is how clear is the link between GW and the observed phenomenon, and how much more information you need to infer that it really was GW and not some other confusing factor.
    For the binary pulsar it's very simple to draw the link and you have the perfectly exact correspondence of GW emission with high-quality observations made over many years and the absence of any reasonable alternative hypothesis. Can't be anything else.
    For BICEP you have only one observing frequency and only a narrow range of multipoles, and a complex cosmological history to imprint GW onto the microwaves, and *many* confusing alternatives: leakage, lensing, dust, phase transitions (, topological defects (, magnetic fields ( - all of which have to be investigated and ruled out before you can be at all confident that a GW caused the observed result.
    And don't get bogged down in semantics: astronomers observe, physicists detect something directly if a signal is due to that thing interacting with the detection apparatus.