Wednesday, May 21, 2014

What is direct evidence and does the BICEP2 measurement prove that gravity must be quantized?

Fast track to wisdom: Direct evidence is relative and no, BICEP doesn’t prove that gravity must be quantized.

In the media storm following the BICEP announcement that they had measured the polarization of the cosmic microwave background due to gravitational waves, Chao-Lin Kuo, member of the BICEP team was widely quoted with saying:
“This is the first direct image of gravitational waves across the primordial sky.”

As of lately, it’s been debated whether BICEP has signals from the early universe at all, or whether their signal is mostly produced by matter in our own galaxy that hasn’t been properly accounted for. This isn’t my area of research and I don’t know the details of their data analysis. Let me just say that this kind of discussion is perfectly normal to have when data are young. Whether or not they actually have seen what they claimed, it is worthwhile to sort out exactly what it would mean if the BICEP claims correct, and that is the purpose of this post.

The BICEP2 results have variously been reported as the first direct evidence of cosmic inflation, direct proof of the theory of inflation, indirect evidence for the existence of gravitational waves, the first indirect detection of the gravitational wave background [emphasis theirs],the most direct evidence of Albert Einstein’s last major unconfirmed prediction, and evidence for the first detection of gravitational waves in the initial moments of the universe.

Confused already?

What is a direct measurement?

A direct measurement of a quantity X is if your detector measures quantity X.

One can now have a philosophical discussion about whether not human senses should account for as the actual detector. Then all measurements with external devices are indirect because they are inferred from secondary measurements, for example the reading off a display. However, for what physicists are concerned the reading of the detector by a human is irrelevant, so if you want to have this discussion, you can have it without me.

An indirect measurement is if your detector measures Y and you use a relation between X and Y to obtain X.

A Geiger-counter counts highly energetic particles as directly as it gets, but once you start thinking about it, you’ll note that we rarely measure anything directly. A common household thermometer for example does not actually measure temperature, it measures volume. A GPS device does not actually measure position, it measures the delay between signals received from different satellites and infers the position from that. Your microphone doesn’t actually measure decibel, it measures voltage. And so on.

One problem in distinguishing between direct and indirect measurements is that it’s not always so clear what is or isn’t part of the detector. Is the water in the Kamiokande tank part of the detector, or is the measurement only made in the photodetectors sourrounding the water? And is the Antarctic part of the IceCube detector?

The other problem is that in many cases scientists do not talk about quantities, they talk about concepts, ideas, hypotheses, or models. And that’s where things become murky.

What is direct evidence?

There is no clear definition for this.

You might want to extend the definition of a direct measurement to direct evidence, but this most often does not work. If you are talking about direct evidence for a particle, you can ask for the particle to hit the detector for it to be direct evidence. (Again, I am leaving aside that most detectors will amplify and process the signal before it is read out by a human because commonly the detector and data analysis are discussed separately.)

However, if you are measuring something like a symmetry violation or a decay time, then your measurement would always be indirect. What is commonly known as “direct” CP violation for example would then also be an indirect measurement since the CP violation is inferred from decay products.

In practice whether some evidence is called direct or indirect is a relative statement about the amount of assumptions that you had to use to extract the evidence. Evidence is indirect if you can think of a more direct way to make the measurement. There is some ambiguity in this which comes from the question whether the ‘more direct measurement’ must be possible in practice or in principle, but this is a problem that only people in quantum gravity and quantum foundations spend sleepless nights over...

BICEP2 is direct evidence for what?

BICEP2 has directly measured the polarization of CMB photons. Making certain assumptions about the evolution of the universe (and after subtracting the galactic foreground) this is indirect evidence for the presence of gravitational waves in the early universe, also called the relic gravitational wave background.

Direct measurement of gravitational waves is believed to be possible with gravitational wave detectors that basically measure how space-time periodically contracts and expands. The slowing down of the rotation period in pulsar systems is also indirect evidence for gravitational waves, which according to Einstein’s theory of General Relativity should carry away energy from the system. This evidence gave rise to a Nobel Prize in 1993.

Evidence for inflation comes from the presence of the gravitational wave background in the (allegedly) observed range. How can this evidence for inflation plausibly be called “direct” if it is inferred from a measurement of gravitational waves that was already indirect? That’s because we do not presently know of any evidence for inflation that would be more direct than this. Maybe one day somebody will devise a way to measure the inflaton directly in a detector, but I’m not even sure a thought experiment can do that. Until then, I think it is fair to call this direct evidence.

One should not mistake evidence for proof. We will never prove any model correct. We only collect support for it. Evidence – theoretical or experimental – is such support.

Now what about BICEP and quantum gravity?

Let us be clear that most people working on quantum gravity mean the UV-completion of the theory when they use the word ‘quantum gravity’. The BICEP2 data has the potential to rule out some models derived from these UV-completions, for example variants of string cosmology or loop quantum cosmology, and many researchers are presently very active in deriving the constraints. However, the more immediate question raised by the BICEP2 data is about the perturbative quantization of quantum gravity, that is the question whether the CMB polarization is evidence not only for classical gravitational waves, but for gravitons, the quanta of the gravitational field.

Since the evidence for gravitational waves was indirect already, the evidence for gravitons would also be indirect, though this brings up the above mentioned caveat about whether a direct detection must not only be theoretically possible, but actually be practically feasible. Direct detection of gravitons is widely believed to be not feasible.

There have been claims by Krauss and Wilzcek (which we discussed earlier here) and a 2012 paper by Ashoorioon, Dev, and Mazumdar that argues that, yes, the gravitational wave background is evidence for the quantization of gravity. The arguments in a nutshell say that quantum fluctuations of space-time are the only way the observed fluctuations could have been large enough to produce the measured spectrum.

The problems with the existing arguments is that they do not carefully track the assumptions that go into it. They do for example make assumptions about the coupling between gravity and matter fields being the usual coupling. That is plausible of course, but these are couplings at energy densities higher than we have ever tested. They also assume, rather trivially, that space-time exists to begin with. If one has a scenario in which space-time comes into being by some type of geometric phase transition, as is being suggested in some approaches to quantum gravity, one might have an entirely different mechanism for producing fluctuations. Many emergent and induced gravity approaches to quantum gravity tend not to have gravitons, which raises the question of whether these approaches could be ruled out with the BICEP data. Alas, I am not aware of any prediction for the gravitational wave background coming from these approaches, so clearly there is a knowledge gap here.

What we would need to make the case that gravity must have been perturbatively quantized in the early universe is a cosmic version of Bell’s theorem. An argument that demonstrates that no classical version of gravity would have been able to produce the observations. The power of Bell’s inequality is not in proving quantum mechanics right - this is not possible. The power of of Bell’s inequality (or measuring violations thereof respectively) is in showing that a local classical, ie “old fashioned”, theory can not account for the observations and something has to give. The present arguments about the CMB polarization are not (yet) that stringent.

This means that the BICEP2 result is strong support for the quantization of gravity, but it does not presently rule out the option that gravity is entirely classical. Though, as we discussed earlier, this option is hard to make sense of theoretically, it is infuriatingly difficult to get rid of experimentally.


The BICEP2 data, if it holds up to scrutiny, is indirect evidence for the relic gravitational wave background. It is not the first indirect evidence for gravitational waves, but the first indirect evidence for this gravitational wave background that was created in the early universe. I think it is fair to say that it is direct evidence for inflation, but the terminology is somewhat ambiguous. It is indirect evidence for the perturbative quantization of gravity, but cannot presently rule out the option that gravity was never quantized at all.


  1. One can always break up a detector in tinier parts and argue that any measurement is indirect. (The Geiger counter detects electrical current pulses caused by ions which have been ionized by hitting with previous ions, one of which was an atom which interacted with a beta particle which has been caused by Compton effect from the gamma ray you want to measure). So how about: a measurement is more direct if it applies the detector in a regime where it has been better studied and known to be reliable, uses less and simpler post-processing of data and so on. Most pioneering experiments will be considered indirect, and will be considered "direct" if and when they become mainstream.

  2. Thank you for this very nice overview post. I have one question though:

    When you write that in the analysis for classical gravity people make "assumptions about the coupling between gravity and matter fields being the usual coupling", what do you mean by "usual coupling"?

    I would assume that the most "usual" way to couple classical gravity to quantum matter are the semi-classical Einstein equations. But to the best of my knowledge this coupling is only very poorly understood for relativistic quantum fields on a curved space-time? What approximations enter there?

    I would also be very grateful if you could provide me with some reference for these calculations for classical gravity.

  3. Okay, I am sorry. Somehow I overlooked the link to the Ashoorioon paper. Guess I should look at it first :).

  4. Excellent analysis!
    Philosophy aside there is parallel confusion we use as evidence to apply to social theory.
    This carries over to the two major camps as to how we do science and fund it each dismissing the other side.
    Germany since 1830 set the world into our era of deeper science. In the 70's few saw Einstein as more at Princeton than a relic of the past coasting in a pipe dream of unified theories.
    That strikes me as some sort of evidence there are social gravitational waves.
    That the data finds 3 to 1 Higgs decay for detection or the dark side I regard as rather solid evidence. It should predict limits on the wide varieties of Higgs decays observable.
    I suspect, to three or four levels we cannot simply mirror super particles as a standard theory but new classes other than fermions or bosons or something neutral. I mean a rayon is rather heavy to be called a lepton.

  5. Oops tauon spell check error, but rayon is about as good a term as anything.

  6. Nothing to say Sabine as you presented the question in a perfect way! Compliments!!!

  7. " if you are measuring something like a symmetry violation or a decay time, then your measurement would always be indirect." A space group P3(1)21 alpha-quartz single crystal has all its atoms in congruent right-handed helices. For space group P3(2)21, left-handed helices. Load a geometric Eötvös experiment with these enantiomorphic test masses. A net non-zero output is direct measurement of vacuum chiral symmetry breaking toward hadronic matter.

    Zero assumptions. The Equivalence Principle is or is not violated. Everything other than unit cell geometry cancels.

  8. Gravity is indeed quantized with photons and all particles mediating the mass, but I don't think just the observation of gravitational waves in any form can prove it. In rigorous 4D general relativity the 4D gravitational wave must be a stationary object (exactly as the BICEP2 observations indicated) and there is no evidence for fact, just these giant artifacts are somehow quantized in their mass/energy.

  9. Sabine,
    On reading this essay again beholding the starry sky in awe of my insignificance yet absolute and transcended in wisdom awakening to that in science or philosophy to what on first reading more than I could see or know, this is a post of most excellent philosophy also.
    I will post this small step epiphany on my blog if anyone desires to view my arguements - if philosophy in some sense can be "quantized " as well.

  10. Holger,

    Yes, you have a point there. There are always additional 'levels of trust' that we just don't discuss because we think they are well understood, or we discuss them separately. I originally wrote in this blogpost there is no such thing as a direct measurement. But then language is always formed by its actual use rather than by meaningful definitions, so instead I was trying to capture the way that the word is actually used. Best,


  11. Andre,

    I mean universal minimal coupling to the perturbatively quantized fluctuations. These are not the semi-classical Einstein-equations. You have there the quantized fluctuations coupling to the stress-energy of the quantum fields. Just from this you can estimate the order of magnitude of certain effects (that's basically what's done in the Krauss & Wilzcek paper, in the other paper it's more subtle, it is stated in the text after Eq (22)). As I said, that is all well and plausible, but I think one should pin this down explicitly as an assumption. Best,


  12. Indirect and direct distinction seems a little subjective, though. I prefer to think of WMAP as a photo of our baby universe. That makes BICEP2 an ultrasound for the anxious excited parents :-)There remains much interpretation but amazing.
    Really wonderful post. Thanks.

  13. "I think it is fair to say that it is direct evidence for inflation" -- didn't you see Raphael Flauger's devastating presentation at Princeton which suggests the entire BICEP2 signal can be explained by gravitational lensing and dust? Didn't you see this scientific paper from today, indicating the same thing?
    Haven't you heard -- BICEP2 is sinking like a stone tossed onto a pond.

  14. Bar,
    the call for a new philosophy of cosmology does seem to suggest going back to the drawing board which in this case is or is close to the drawing board.
    Evidence may not fall into the usual debate as to what is subjective or objective.
    Consciousness at least as thought may be defined as some interval in isolation (or half observation).
    To know we are mystically all interconnected is to lose the wisdom of thought that senses such social uncertainty bonds as we in totality become detatched from other minds in the name of social unity.

    So a physical object as M mentioned a stone into water leaving waves vanishes into depths beyond white or pink noise in a higher integration over infinite matrices and conformal algebras that we agree defines matter.

  15. Redshift looks back in time, z = [(α_now)/(α_then)] - 1. "α" is an atomic spectral line frequency. Larger z is older from us, younger from the Big Bang at infinite z.
    "+30&" sets z, then enter for data.

    z = 173, 7.26 megayears old, 201°C. Chemistry cooks panspermia.
    z = 135, 10.6 megayears old, 98°C. Livable under pressure.
    z = 110, 14.5 megayears old, 29°C. Summer day.
    z = 100, 16.8 megayears old, 2°C. Arctic and Antarctic Oceans.
    z = 85, 21.5 megayears old, -38°C. Large bodies retain accretion and radioactive decay heat.

    15 - 20 megayears is a short span for life to evolve. It is also 7.8×10^69 cubic miles in which to do it.

  16. This comment has been removed by the author.

  17. Mohammad: Deleted your comment - too many links to strange sources. Please read comment rules.

    Otto: Deleted your reply to Mohammad because it made no sense without the original comment.

  18. "How can this evidence for inflation plausibly be called “direct” if it is inferred from a measurement of gravitational waves that was already indirect?"

    It can't. End of story. You can maybe kid yourself it ain't so, but you can't kid anybody else. As for Bell, read this:

    "...Bell's own view of his theorem, however, was quite different. He instead took the theorem as establishing an "essential conflict" between the now well-tested empirical predictions of quantum theory and relativistic local causality. The goal of the present paper is, in general, to make Bell's own views more widely known and, in particular, to explain in detail Bell's little-known mathematical formulation of the concept of relativistic local causality on which his theorem rests..."

  19. John,

    Your condescending snarl is pointless. I am simply conveying how a word is used. Whether you like or do not like the way it is used is entirely irrelevant. Personally, I don't think it makes a lot of sense, but then I don't think it makes much sense naming a city 'New York' either. Best,


  20. It's just simple logic, Sabine. If you have indirect evidence for something from which you infer something else, you don't have direct evidence for the latter.

  21. In order to make it clearer what BICEP found, one has to know why it was looking for B-modes in the first place. No, it's not to confirm inflation -- read on.

    Namely, very few people noticed it, but Linde has revealed in his April audio-interview to Financial Times, that Guth recanted inflation altogether "as a theory which never worked" (listen to it here:

    This is what Guth's wikipedia page says about the affair:

    So Linde actually never was after something that didn't work. Otherwise you could call him stupid, and we all know he's anything but stupid. Perhaps it's only now that he opened his cards, but that's OK, Nobels are at stake and scientists do play politics just like everyone else.

    Indeed, Linde's theory has very little to do with original inflation. In fact, one could argue that Linde's theory in its simplicity rather ridicules inflation's gazillion failures that eventually forced Guth to use more than 100 pages to recant.

    Also, "chaos inflation" is an oxymoron.

    So knowing those background facts on the motivation of main players, we're actually talking BICEP confirming Linde's theory on origin of universes (properly called).

    The Linde theory requires multiverse (as in: infinitely many universes). This in turn kills Genesis and all religion, or, "bye bye God" -- as Linde said vividly in the above interview -- and mysteriously vanished from all the media.

    Then the witch-hunt against Linde and BICEP came into its full swing.

    Question is, how much are we helping science by buying into most ridiculous accusations, such as the "Boltzmann in Wonderland" tale by one Carroll again, or by the "dusty slide affair", statistics critique (though BICEP is statistics-related less than particle physics is say, GPS-related).

    People in general tend to talk too much, and physicists are gifted speakers. So far, this barrage of fairy-tales against BICEP (and thus against Linde and multiverse) has played into the hands of the religion alone.

    Interesting to see how Christians managed to throw a bone amongst physicists. Shockingly, most of the starved ones (those not on big project grants) seem to be biting on it like mad dogs...

  22. Here is another take on the direct vs. indirect question.


COMMENTS ON THIS BLOG ARE PERMANENTLY CLOSED. You can join the discussion on Patreon.

Note: Only a member of this blog may post a comment.