Monday, March 17, 2014

First Direct Evidence for Cosmic Inflation

That was the title of the BICEP2 presentation today. Gives you some idea about the magnitude of the result, if it holds up: it really is astonishingly exciting.

Unfortunately it was so exciting that we in Helsinki couldn't even access the Harvard server and so couldn't watch any of the webcast at all. It seems the same was true for most other cosmologists around the world. So my comments here are based purely on a preliminary reading of the paper itself, and a distillation of the conversations occurring via Facebook and the like.

Firstly, the headline results: the BICEP team claim to have detected a B-mode signal in the CMB at exceedingly high statistical significance. Their headline claim is

$r=0.2^{+0.07}_{-0.05}$, with $r=0$ disfavoured at $7.0\sigma$

That is frankly astonishing. Here's the likelihood plot:

BICEP2 constraint on the tensor-to-scalar ratio r. 

(All figures are taken from the paper avalaible here.)

The actual measurement of the BB power spectrum looks like this:

The black points are the new measurements, the other coloured points are the previously available best upper limits. The solid red curve is the theoretical expectation from lensing (the relatively boring contribution to BB), the dashed red curve that dies off is the theoretical expectation from a model with inflationary gravitational waves and $r=0.2$, and the other dashed red curve (were they short of colours?!) is the total.

They've also done a pretty good job of eliminating other foreground sources (dust, synchrotron emission etc.) as possible explanations for the signal seen, which means it is much more likely that the signal is actually due to primordial gravitational waves from inflation. In doing this, it helps that the signal they see is actually as large as it is, since there's less chance of confusing it with these foregrounds (which are much smaller).  [Update: I'm not an expert here, apparently some others were less convinced about the removal of foregrounds. Not sure why though – I'd have thought other systematic errors were far more likely to be a problem than foregrounds.]

So far so good. In fact — and I really can't stress this enough — this is an extraordinary, wonderful, unexpected result and huge congratulations to the BICEP team for achieving it. It will mean a lot of happy theorists as well, because we finally have something new to try to explain!

However, it is very important that as a community we remain skeptical, particularly so when - as here - the result is one that we would so desperately love to be true. Given that, I'm going to list a serious of things that are potentially worrying/things to think about/things I don't understand. (Some of these are not things I noticed myself, but were points raised by Dave Spergel, Scott Dodelson and other experts at the ongoing live discussion on Facebook.) Doubtless these are questions the BICEP team will have thought about themselves; perhaps they already have all the answers and will tell us about them in due course — as I said, no one I know was able to watch the webcast live.

  • In the BB-spectrum plot above, the data seem to be showing a significant excess above expectations for multipoles about $\ell\sim200-350$. What's going on with that?
  • This is particularly noticeable in another figure (Fig. 9) in the paper:
  • From the above figure, preliminary results of the cross-correlation with Keck don't show the excess at high-$\ell$ (a reason to believe it might go away), but the same cros-correlation also shows less power at lower $\ell$ (which is a bit confusing).
  • At lower values of $\ell$ the EE power spectrum also shows an excess (Fig. 7):
  • All the above points put together suggest that perhaps there is some leakage in the polarisation maps coming from the temperature anisotropy — a large part of the analysis work is concerned with accounting for and correcting for any such leakage, of course, the question is to what extent independent experts will be satisfied that these methods worked.
  • Although the headline figure is $r=0.2$, they rather confusingly later say that when the best possible dust model is used for foreground subtraction, this becomes $r=0.16^{+0.06}_{-0.05}$. But if this the the best possible dust model, why is this not the quoted headline number? Is this related somehow to the power excess at $\ell\sim200-350$?
  • If $r$ is as large as they have measured why was it not seen by Planck? Actually this is a fairly complicated question: the point being that if the tensor amplitude is so large, it should make a non-negligible contribution to the temperature power spectrum as well, which would have affected Planck's results. Planck had a constraint $r<0.11$, but this specifically assumed that the primordial power spectrum had a power-law form with no running (sorry about the technical jargon, unfortunately not enough time to explain here today). So BICEP suggest one way around this tension is to simply introduce a running, but it seems (but this bit was not entirely clear to me from the paper) that you need a fairly large value of the running for this explanation to fly. And if you've got a large running then you have to worry about why not a running of the running, a running of the running of the running and so on ad infinitum - in fact how do we know that the power-law expansion form of the $P(k)$ is the correct way to go at all?
  • Besides, are there viable inflationary models that predict both large $r$ as well as large running (or non-power-law form of the primordial power)? Given the vast array of inflationary models, the answer to this question is almost certainly yes, but people may consider some other explanations more worthwhile ...
Phew. There are probably lots of other things to think about, but that's about all I can manage today. It's been a very exciting day!


  1. Incidentally, one thing I mentioned in my preview post was the possibility that the B-mode signal could come from cosmic defects rather than inflation. I have now been informed via Facebook by Dani Figueroa (one of the authors of the paper I cited) that defects could not hope to produce an $r$ value as large as 0.2 - his estimate is about an order of magnitude smaller. So if this is real it is pretty definitely inflation that did it!

  2. Thanks for the insightful discussion- great job!

  3. Thank you Dr.Sesh for your great effort.
    I have recently read an article about a paper by James B. Dent, Lawrence M. Krauss and Harsh Mathur called "Killing the Straw Man: Does BICEP Prove Inflation?" saying that the source of these gravitational waves may not be inflation but a phase transition occurred after it. What do you think?

    Here is the link of the article:

    1. Interestingly, that was one of the papers I was planning to mention in my next blog post. Unfortunately that paper is spoiled by the fact that the authors were in such a hurry to put it out. Essentially, what it says is that if the constraints on isocurvature perturbations were the same as they were a year ago when Planck first published their data, their strange model for production of gravitational waves could still be allowed by the data, but if they were tighter their "straw man" model would be ruled out. But they don't check whether the constraints on isocurvature perturbations have themselves been altered by the BICEP2 data - they have, so their model is ruled out before they begin and the paper becomes a little boring.

      Anyway, more on this later when I have a bit more time.