BICEP2: Two Months Later (and the Morning After)

It's two months since the BICEP2 team announced it had seen the fingerprints of gravitational waves in the microwave background, thus apparently opening a portal into the universe ten trillion, trillion, trillionths of a second after the Big Bang. In the last week, however, the mood among cosmologists has taken on a morning-after tone, with a wave of doubt rolling through the community. It's possible that the cosmology community is slowly waking up to find itself in an unfamiliar Las Vegas hotel room with a throbbing headache, hazy memories of the night before, and a fresh tattoo reading "r=0.2".

 

What's the issue?

The second thoughts are about how the BICEP2 analysis accounts for "foregrounds", which is to say, things between us and what we're looking at. In this case, the question is: how might dust in our own galaxy interfere with the detection of tell-tale signs of gravitational waves in the microwave sky?

The BICEP2 team concluded that foregrounds contribute around 20% of their signal, which leaves plenty of room to make a confident claim to have detected gravitational waves. 

By far the best way to isolate foreground dust is to use the microwave equivalent of a colour photo – but unfortunately the BICEP2 image is monochromatic. Consequently, the BICEP2 team could not extract foregrounds using just their own data. Instead, they presented a slate of indirect methods, in order to arrive at a reasonable estimate. One of these approaches uses the Planck satellite's measurements of the microwave sky, and it's this one that has come in for serious scrutiny. 

Firstly, it's become clear that the BICEP2 team snagged the data they needed –  which has not been formally released by the Planck collaboration –  from a "teaser" image in a presentation posted online. This is certainly unorthodox, but does not immediately undermine the BICEP2 result. The issue burst into the public domain when Dan Falkowski (who often discusses -- and disseminates -- particle physics rumours on his blog Resonaances) not only drew attention to the "data-scraping", but claimed that the BICEP2 team had misinterpreted the images and would be revising their paper.

The BICEP2 people hotly denied they had made a mistake, and didn't even concede there was a mistake to argue about.

Separately, in a talk at Princeton (see video and slides), Raphael Flauger presented a virtuoso re-analysis of the foregrounds, including an estimate of the extra uncertainty injected by the fact that information was grabbed from a PDF file rather than from raw data. His measured conclusion was that the BICEP2 result is possibly overly optimistic. 

The blogosphere and science media has kicked into overdrive as the debate rages. Sesh Nadathur, Peter Woit and Sean Carroll all provide good summaries, while the Washington Post provides a great analysis of the current state of play.

 

So Where Are We Now?  

This is not (yet) a show-stopper, but the debate shines a light on a weak spot in BICEP2's claim to have seen the fingerprints of gravitational waves.

In the long run, we need more data. Planck data was only part of the BICEP2 team's estimation of the foregrounds, and new data (being gathered as you read this) should provide a much better answer over the next 12 months. 

 

Is This How Science is Done Now?

Apparently, yes it is. This is science in the age of the internet, and the world gets to watch in real time. We are caught between two powerful forces -- on the one hand, as Lyman Page (a Princeton astrophysicist and microwave background expert) says at the end of Flauger's talk:

So this is not – we all know, this is not sound methodology. You can’t bank on this, you shouldn’t. [...] You just can’t, you can’t do science by digitizing other people’s images.

But on the other hand, does anyone really expect us just to sit and wait?

As far as the screen-scraping is concerned, there are precedents -- a few years ago, the Pamela satellite was rumoured to have seen an excess of high energy positrons in cosmic rays that might have been due to dark matter in our galaxy. The Pamela team showed a slide at a conference and a couple of enterprising individuals snapped photos and extracted the datapoints -- hundreds of papers quickly followed. Given the cameras that live inside our phones, the ubiquity of video at big conferences, these days "teasers" effectively amount to an unofficial data release. So people who drop hints about unpublished results in conference talks while coyly flashing a visual aid should not be surprised at the consequences. [As it turns out, the Pamela excess is real, but can explained by less glamorous mechanisms than dark matter]. 

One might also ask why Planck doesn't just release the sky map used by BICEP2 and be done with it. That's because what they showed was a work-in-progress: maps like this are not "raw" data, but the end-product of a long and painstaking analysis, and we can't demand that anyone turn over a half-finished product. 

As many people have pointed out, the BICEP2 results have not gone through peer review. On the other hand, many other people (including me) also pointed out that over the last two months the BICEP2 papers have been dissected by hundreds of scientists, so they are getting more stringent, open-air scrutiny than any journal could provide (given that important journal papers might still only go past three referees). Moreover, many recent announcements (Planck, WMAP, the Higgs) were made before undergoing peer review, so this is the new normal. (It follows the near ubiquitous practice among the astro-and particle physics communities of posting full "preprints" on arxiv.org before you send your article to an actual journal.)

 

The Worst-Case Scenario

Whatever happens, the BICEP2 observations are by far the most precise measurements of the microwave background ever made. Even if the claimed detection of gravitational waves evaporates, the technological strides that underpin BICEP2 (and similar experiments now gathering data) will bring dramatic progress in cosmology. Even in the worst-case scenario, we are not looking at a re-run of the faster-than-light neutrinos flap which was traced to a loose cable and a dodgy clock and became entirely uninteresting once those problems were solved.  

 

My Own Guess

Personally, I would not be surprised if BICEP2 had overestimated the strength of the gravitational wave signal, even if I am not expecting it to vanish completely when the dust has settled (if you will pardon the pun). Cosmologists use the parameter "r" to describe the strength of the gravitational wave background. Before BICEP2, indirect measurements suggested that r was no more than 0.1, but BICEP2 prefers a higher number. A higher value of r would be fantastic for me and my fellow theorists, but it almost seems too good to be true. 

On the other hand, if BICEP2 is correct, it successfully probes the universe at energies a trillion times higher than we can reach at the LHC. Any intuition we might claim to have about physics at these scales is tenuous at best, so we will simply have to wait and see what develops. 

So even if the honeymoon is over, cosmology and gravitational waves are not yet headed for a Vegas-style quickie divorce. On the other hand, perhaps they need a restorative breakfast at the hotel buffet and a heart-to-heart about where they go from here.  (And they may yet need some touch-ups on that tattoo).


Live from New York City

If you are in Auckland on May 31 (or in New York City on May 30, when you can see it in the flesh) the University of Auckland is partnering with New York's World Science Festival to present a simulcast of a Festival programme on the BICEP2 results, Ripples from the Big Bang. Moderated by Brian Greene it brings together John Kovac, one of the leaders of BICEP2, Alan Guth and Andrei Linde, who played key roles in the development of inflation, microwave background experimentalist Amber Miller, and Princeton theorist Paul Steinhardt. The New York event will be streamed live followed by a local Q&A, with me providing the Answers. Free entry, but ticket required for entry