Monday, 12 May 2014

Is BICEP wrong?

The BICEP claim of detecting the primordial B-mode in the polarization of the Cosmic Microwave Background was a huge news. If confirmed, it would be an evidence of gravity waves produced during cosmic inflation, and open a window on physics at an incredibly high energy scale of order 10^16 GeV. Possible implications were described in detail in some 300 papers triggered by the BICEP announcement.  But, among this understandable excitement, we have been aware that the signal has to be confirmed by other experiments before the discovery is established. Back then, Résonaances precisely estimated the chances of the signal being true at `fifty-fifty'.  It appears it's the latter fifty that's gaining an upper hand...

Barring a loose cable, the biggest worry about the BICEP signal is that the collaboration may have underestimated the galactic foreground emission. BICEP2 performed the observations at only one frequency of 150 GHz which is very well suited to study the CMB, but less so for polarized dust or synchrotron emission. As for the latter, more can be learned by going to higher frequencies, while combining maps at different frequencies allows one to separate the galactic and the CMB component. Although the patch of the sky studied by BICEP is well away from the galactic plane, the recently published 353 GHz polarized map from Planck demonstrates that there may be significant emission from these parts of the sky (in that paper the BICEP patch is conveniently masked, so one cannot draw any quantitative conclusions). Once the dust from the BICEP announcement had settled, all eyes were thus on precision measurements of the galactic  foreground. The rumors that have been arriving from the Planck camp were not encouraging, as they were not able to confirm the primordial B-mode signal. It seems that experts now put a finger on what exactly went wrong in BICEP.

To estimate polarized emission from the galactic dust, BICEP digitized an unpublished 353 GHz map shown by the Planck collaboration at a conference.  However, it seems they misinterpreted the Planck results: that map shows the polarization fraction for all foregrounds, not for the galactic dust only (see the "not CIB subtracted" caveat in the slide). Once you correct for that and rescale the Planck results appropriately, some experts claim that the polarized galactic dust emission can account for most of the BICEP signal. The rumor is that the BICEP team has now admitted to the mistake [Update: this last statement is disputed and outwardly denied].

Note that we should not conclude that there is no observable tensor modes in the CMB. Indeed, the tensor to scalar ratio of order 0.1 is probably consistent with the existing experimental data, and may be responsible for a part of the B-mode signal detected by BICEP. New data from Planck, POLARBEAR, ACTpole, and Keck Array should clarify the situation within a year from now.  However, at this point, there seems to be no statistically significant evidence for the primordial B-modes of inflationary origin in the CMB.

46 comments:

  1. I heard that the fact that the BICEP region was masked in the Planck release was not deliberate politics - the Planck paper was in preparation well before the BICEP results, and the masking regions already determined.

    ReplyDelete

  2. If the BICEP2 results go the way of "faster-than-light" neutrinos, will there be massive media fanfare and public soul-searching by physicists to match the massive media fanfare and childish grand-standing when the BICEP2 results were released?

    Would we learn anything, or just keep repeating the same old mistakes?

    ReplyDelete
  3. Thanks for the very important update, and please keep us posted on further info on this topic. What is very interesting is to compare the polarization fraction map on your blog post (the one used by the BICEP2 team) with the new map on page 6 of this Planck team paper:
    http://arxiv.org/pdf/1405.0871v1.pdf

    It's basically the same map, but with one huge difference -- a large fraction of the areas that are yellow in the previous map are now red. That means there's many more regions with 20% polarization rather than just 10% polarization. In short, much more dust polarization than previously accounted for by BICEP2.

    ReplyDelete
  4. Thanks for the very important update, and please keep us posted on further info on this topic. What is very interesting is to compare the polarization fraction map on your blog post (the one used by the BICEP2 team) with the new map on page 6 of this Planck team paper:
    http://arxiv.org/pdf/1405.0871v1.pdf

    It's basically the same map, but with one huge difference -- a large fraction of the areas that are yellow in the previous map are now red. That means there's many more regions with 20% polarization rather than just 10% polarization. In short, much more dust polarization than previously accounted for by BICEP2.

    ReplyDelete
  5. It's worth keeping in mind that the "polarisation fraction" maps produced by Planck aren't signal dominated in the BICEP2 region. That is, the instrument noise is actually larger than the signal (dust or primordial) along those lines of sight (this is why they were blocked out in the recent Planck paper). Therefore, even if Planck measures a dust fraction of 0.1 or 0.2 along that specific line of sight it is hard to tell whether that is genuine dust or instrument noise.

    Planck will be able to tell statistically what the noise is along a *typical* BICEP2-like line of sight, but its measurement along BICEP's precise line of sight won't be definitive. Of course if the typical dust fraction is large, that will be very worrying for BICEP2.

    ReplyDelete
  6. Have a look at: http://arxiv.org/abs/arXiv:1404.1899. Where they consider an additional foreground. They have the following remark in their conclusion: "It has not escaped our attention that the lower part of Loop I, in particular the additional loop structure identified by Wolleben (2007), crosses the very region of the sky from which the BICEP 2 experiment has recently
    detected a B-mode polarisation signal (Ade et al. 2014). This has been ascribed to primordial gravitational waves from inflation
    because ``available foreground models" do not correlate with the BICEP maps. The new foreground we have identi ed is however not included in these models. Hence the cosmological signi cance if any of the detected B-mode signal needs further investigation. Forthcoming
    polarisation data from the Planck satellite will be crucial in this regard."

    ReplyDelete
  7. Someone above insinuated censorship, or at least a purposeful delay in release of data.

    If the sky research did get censored due to politics, we must remember that politics swings many ways.

    So a patch of sky "masked in order to hide proof A" could also be the patch "masked in order to delay confirmation of proofs A, B...".

    Politics is a business of patience and deceptively traded interests. Physics is a science of patience and deceptively traded interests.

    Business rules the world, so I call bs as for any conspiracy as implied by someone in the above.

    ReplyDelete
  8. I'm not involved with the BICEP2 team or result, however this rumor is bullshit, the BICEP2 team very much still stands behind their results. I think there is a lot of propaganda coming out from the larger (more well-funded) Planck and Princeton groups.

    ReplyDelete
  9. The point here is really lack of 'sciencemanship' Neither BICEP nor Planck can truly claim a discovery without each other's maps. BICEP did f* up, they admitted it. (This is also like the 'hot pics of Pamela' affair...) Planck is not interested in giving them all the credit either. ACT/SPT need to stir sufficient doubt to be relevant, aka Princeton. Bottom line, is not clear what will happen to the signal, but it's buried in politics right now...and btw, there's another paper from Planck on its way...

    ReplyDelete
  10. Read the BICEP2 paper. Their result does not rely on any data from Planck, including unpublished conference data, which is truly a joke to even consider as being used by a veritable scientific collaboration. This specific rumor is wholly without basis.

    ReplyDelete
  11. There is of course the embarrassing possibility that (the multi-million pound taxpayer funded) Planck project failed to collect sufficient data to make any strong enough claim regarding BICEP2.

    We will see...

    ReplyDelete
  12. Dear anonymous of 12 May 2014 23:13, I suggest that you yourself "Read the BICEP2 paper." Footnote 33 refers to the slides of a conference talk (http://www.rssd.esa.int/SA/PLANCK/docs/eslab47/ Session07_Galactic_Science/47ESLAB_April_04_11_25_ Bernard.pdf) from which they digitized the dust map that became their DDM2 dust model. Did they "rely" on this dust model? Not exactly. But most of their dust models relied on speculation about polarization rather than actual data. Furthermore, they only argue that their data cannot be explained by dust alone or by synchrotron alone; they never give statistical evidence against the combination of synchrotron and dust.

    I desperately want the signal to be primordial but the reasons to have doubts seem strong.

    ReplyDelete
  13. Anonymous

    erm, the BICEP2 result is strongly dependedent on a accurate dust map polarization model, at least of the part of the sky they measured. Their result is obviously dependent on Planck not discovering a huge polarization signal from trivial sources.

    ReplyDelete
  14. Has anyone come up with preliminary estimates of how this correction will impact the bottom line value for "r"?

    The best fit of the Planck data to date IRRC was about r=0.01 but it was not inconsistent at the two sigma level with r=.10 to .11.

    Assuming that the margin of error remains the same, any big reduction in the BICEP-2 best fit data point would kill its statistical significance, but wouldn't necessarily render the find irrelevant.

    For example, a result of say r=0.1 after reanalysis wouldn't be inconsistent with existing Planck data at two sigma and would still be a pretty strong hint in favor of more than minimal tensor modes.

    ReplyDelete
  15. One comment: indeed, the issue affects only BICEP's DDM2 foreground emission model. But DDM2 is what they single out in the paper as their best model. Recall that in the original analysis using DDM2 shifts the central value of r from 0.2 down to 0.16. Now we know that their best model underestimates the foreground, so we know the significance must go down further. By how much, I don't know. Various rumors place the significance of the corrected signal between 0 and 2 sigma.

    ReplyDelete
  16. The quotes from BICEP's Clement Pryke at sciencemag don't help to give one confidence that the result will hold up, he admits they used data from the Planck team which they didn't fully understand

    http://news.sciencemag.org/physics/2014/05/blockbuster-big-bang-result-may-fizzle-rumor-suggests

    ReplyDelete
  17. 2 sigma significance of corrected signal is probably a fluctuation. What was the significance in the original BICEP paper of the uncorrected signal?

    ReplyDelete
  18. It's kind of appalling that places like : http://news.sciencemag.org/physics/2014/05/blockbuster-big-bang-result-may-fizzle-rumor-suggests are treating Jester as a primary source. I am glad to see that he rewrote his earlier statements about BICEP acknowledging the error in response to the comments by Pryke, but it is obvious that this is more gossip than fact at this point.

    ReplyDelete
  19. From the abstract of the paper:

    "We report results from the BICEP2 experiment, a Cosmic Microwave Background (CMB) polarimeter specifically designed to search for the signal of inflationary gravitational waves in the B-mode power spectrum around l=80. . . . We find an excess of B-mode power over the base lensed-LCDM expectation in the range 305σ. . . . We also estimate potential foreground signals and find that available models predict these to be considerably smaller than the observed signal. These foreground models possess no significant cross-correlation with our maps. Additionally, cross-correlating BICEP2 against 100 GHz maps from the BICEP1 experiment, the excess signal is confirmed with 3σ significance and its spectral index is found to be consistent with that of the CMB, disfavoring synchrotron or dust at 2.3σ and 2.2σ, respectively. The observed B-mode power spectrum is well-fit by a lensed-LCDM + tensor theoretical model with tensor/scalar ratio r=0.20+0.07−0.05, with r=0 disfavored at 7.0σ. Subtracting the best available estimate for foreground dust modifies the likelihood slightly so that r=0 is disfavored at 5.9σ."

    Note that taken together the abstract implies that the error distribution is strongly non-Gaussian. If the distribution were Gaussian, r=0 would have been disfavored at only 4.0σ.

    ReplyDelete
  20. Dear anonymous, there are 2 important reasons to consider rumours on this blog more reliable that journals such as sciencemag: 1) the author of this blog is a physicist; 2) the author of this blog is not a journalist.

    ReplyDelete
  21. Is it correct then that BIECP2 was systematically unable to precede PLANCK? Surely in their methodology before getting funded someone would have asked about how they would measure the foreground, the answer would have been "We will have to wait for PLANCK!" So in this sense this very modest telescope was never designed to compete for a Nobel Prize but was always complementary to PLANCK?

    ReplyDelete
  22. dear Anonymous of 12 May 2014 23:27 - using a pdf from a talk - without even knowing exactly what it is (cf. reaction from BICEP people inhttp://news.sciencemag.org/physics/2014/05/blockbuster-big-bang-result-may-fizzle-rumor-suggests) - is incredible, to say the least ! how can such a paper have been accepted ?

    ReplyDelete
  23. "Moreover, Pryke says, conversations with members of the Planck team leave it uncertain exactly what is in the key plot. "It is unclear what that plot shows," he says."

    This is some kind of joke, right? RIGHT??!!!!

    ReplyDelete
    Replies
    1. I have done this, but as a temporary expedient. A big danger is exactly what may have happened here - a conference presentation may be VERY sloppy about the difference between the graphs they show, and what they say those graphs show. (I have seen puzzling graphs that simply had to have different units from what the caption said.) In a presentation, this is more or less OK, but it shouldn't (and generally doesn't) make it into print.

      Delete
  24. Dear Anonymous, sciencemag (unlike Jester) both avoided to make unfounded claims which were then edited away , and also asked members of the team and quoted their sources. Not only is that better journalism, it would have also been better science than what was written here originally.

    It could be that all of this speculation will turn out to be true, and the significance of bicep will weaken or disappear. The problem with what was written initially is that it offered no insights and no sources, just made accusations which (maybe true) but were retracted.

    ReplyDelete
  25. Nick, my understanding is that the amount of polarized foreground in the BICEP patch is a bit of a surprise. That region is rather clean in temperature maps, apparently it is less clean in polarization. But, right, it was always clear that for a fully reliable estimation of the foregrounds we need measurements at several frequencies, and Planck is by far best suited to do that.

    ReplyDelete
  26. That the BICEP2 team did not know the meaning of "not CIB subtracted" (as indicated in the post) is of course a joke. So what information do we have from Planck that alter the picture about the strength of the CIB in the region of interest?

    ReplyDelete
  27. Pierre Ramond once quipped that "rumors can travel faster than the speed of light because they carry no information." Perhaps this blog lends support to Ramond's theory, which incidentally was formulated before the advent of the blogosphere. It is disappointing to see the author of this blog appropriate for himself the role of shadow spokesperson for the Planck Collaboration, to which he is in no way connected. I found it hard to understand how someone with no expertise in CMB foregrounds should find it legitimate to be interviewed by Science magazine based on the "rumors that have been coming in." Let us wait till Planck is ready to present findings that have been carefully checked and abstain from speaking on behalf of the Planck collaboration based on the claim of being a clearinghouse for Planck rumors. Martin Bucher (bucher@apc.univ-paris7.fr)

    ReplyDelete
  28. There are references in the comments to an original version of the blog entry. What is the difference with the edited version? In the interest of 'blogging ethics' Jester should consider adding accounts of significant edits. It would remove a blemish in this otherwise fantastic blog.

    ReplyDelete
  29. The difference between physics blogging and science journalism may lie in the eye of the beholder.
    To be (a whip of god disguised in a gadfly) or not to be (Jester)? Keep blowing wind of scientific scrutiny and some itching powder on the blogosphere Adam!

    ReplyDelete
  30. "using a pdf from a talk - without even knowing exactly what it is incredible, to say the least !"

    Yes, I was also staggered when I read this. Upfront on the BICEP2 paper should have read "These results are crucially dependent on a PDF we found which was used in a talk by someone else which we don't really understand." Because - no matter if this rumour is confirmed or not - that nows appears to be the truth of the situation. Staggering.

    ReplyDelete
  31. This rumor sounds a lot like the "faulty cables" case in the making, remember that one?

    A lot's (and I mean a lot's) of politics involved here too again. After all the political crap seen each time when some (otherwise perfectly sane group of physicists) even touches on a paradigm topic, I'm not even sure I like science any more.

    For it all seems to boil down to which of the opposing sides has more muscle to flex (if you prefer plain English: the moneys).

    And who can beat the particle physics gang money-wise? Only the Chinese Army, my friends!

    ReplyDelete
  32. As far as I can tell, the recent Planck data has no information in it concerning the BICEP2 results. The only new morsel is that one or two of the template maps that they used probably underestimated the amplitude of that map by up to a factor of 2. All of the template maps that they used were being presented simply as evidence that it couldn't possibly be dust, and there was no detected trace of these template maps in the BICEP2 data.

    Instead of there being effectively zero chance that dust is a problem, now that has to be modified to "it seems incredibly unlikely to be dust, but we won't be able to conclusively rule that out until we get more data from planck/spt/act/polarbear"

    ReplyDelete
  33. Good stuff Jester. I have a hunch that the end consequence of all this will be entitled Is Inflation Wrong?. It's said to be responsible for the uniformity in the CMBR. And the non-uniformity! There's a nail-and-shoe GR/black hole chain-of-logic wherein you end up wondering why it was ever needed, and thinking big bang cosmology is better without it.

    ReplyDelete
  34. Anonymous 13:55

    "Incredibly unlikely" that it's dust? All it takes is a factor of 2: if instead of the 5% polarization from dust they assumed it's 10%, that's the whole signal. Some experts think this is easily plausible. The only _data_ they used to support this was the misread Planck plot.

    As other commenters here have said, it is "staggering" that BICEP was so cavalier about such an obvious foreground.




    ReplyDelete
  35. "so cavalier with an obvious foreground"?

    Read the paper. What they actually wrote about polarized foregrounds is:

    "The main uncertainty in foreground modeling is currently
    the lack of a polarized dust map. (This will be alleviated soon by the next Planck data release.) In the meantime we have therefore investigated a number of existing models and have formulated two new ones. A brief description of each model is as follows:"

    They then fit to all the models they could dig up at the time. And added two new ones to exhaust all the information available to them, including the infamous DDM2. They conclude with:

    "All of the the models except FDS make explicit predictions
    of the actual polarized dust pattern in our field — presumably
    with varying probabilities of success. We can therefore search for a correlation between the models and our signal by taking cross spectra against the BICEP2 maps. Figure 6 shows the resulting BB auto and cross spectra - note that the autospectra are all well below the level of our observed signal and that the cross spectra are consistent with zero. We also not that the DDM2 model auto spectrum (which is the highest) contains uncorrected noise bias from the polarization fraction and angle maps."

    Is this cavalier? They used what they knew. We don't know much more now. We'll have to see.

    ReplyDelete
  36. ""Incredibly unlikely" that it's dust? All it takes is a factor of 2: if instead of the 5% polarization from dust they assumed it's 10%, that's the whole signal. Some experts think this is easily plausible. The only _data_ they used to support this was the misread Planck plot."

    This comment really over-simplifies the situation. The strongest evidence that dust is not the dominant contribution doesn't come from the subtraction of auto-correlated foregrounds, as suggested above, and for which the knowledge is weak. The strongest evidence is the lack of cross-correlation, and the likelihood analysis of the spectral index. We should be careful of sweeping judgement yet.

    ReplyDelete
  37. Novac himself has denied they are admitting any weakness in their analysis other than what they made clear in the original publication.

    http://www.newscientist.com/article/dn25558-rumours-swirl-over-credibility-of-big-bang-ripple-find.html#.U3JhvnZZhyk

    ReplyDelete
  38. jaybeegee: "Novac himself has denied they are admitting any weakness in their analysis other than what they made clear in the original publication."

    From the original publication:

    "The main uncertainty in foreground modeling is currently the lack of a polarized dust map."

    "All of the the models except FDS make explicit predictions
    of the actual polarized dust pattern in our field — presumably
    with varying probabilities of success."

    "The probability that each of these models reflects reality is hard to assess"

    These are some of the weaknesses in the analysis admitted in the original publication. It is not clear that very much is new.

    ReplyDelete
  39. I would put money (if I had any) on Planck having a lot more to say about both foregrounds and axes of beauty like the ecliptic.

    ReplyDelete
  40. I'm eagerly awaiting the appearance of a new movie on youtube.

    THE SCENE: in front of Linde's house.

    Door opens --- by the way, does Renata K *always* accompany Linde to the door when the bell rings? --- and someone from PLANCK is standing there with a cream pie in his hand.

    BUT before he can squish it in Andrei's face, Renata steps up and punches him in the stomach. Fade to credits.

    ReplyDelete
  41. "The rush to present the BICEP2 data may be a desperate attempt to cover up the significance of the CMB anisotropies verified as cosmic by Planck, plus the pressure put on scientists by the release of the documentary The Principle (exploring challenges to the Copernican Principle"

    No.

    ReplyDelete
  42. The PDF has just come out for the Princeton talk on BICEP2 foregrounds.

    http://www.pctp.princeton.edu/pcts/SpecialEventSimplicity2014/Simplicity.pdf

    Read it and frown if you’re a BICEP2 advocate.

    Go to page 39 of the presentation. One corrected model shows 7 out of 9 BICEP2 observation points within the area predicted by a combination of dust foregrounds and gravitational lensing. The other corrected model also shows 7 out of 9 BICEP2 observation points within the area predicted by a combination of dust foregrounds and gravitational lensing.

    If you use the BICEP2/Keck combination for the observation points, then it’s basically 9 out of 9, with nothing that needs to be explained by cosmic inflation.

    A separate estimate using column densities finds the same thing.

    ReplyDelete
  43. Jester?
    Lot of entangled Boltzmann brains here except for Marni and Robert.
    See newscientist article today near where they cited your blog
    Best

    ReplyDelete