Thursday, March 21, 2013

What Planck has seen

Update at 16:30 CET: I've now had a chance to listen to the main science briefing, and also to glance at some of the scientific papers released today, albeit very briefly. So here are a few more thoughts, though in actual fact it will take quite some time for cosmologists to fully assess the Planck results.

The first thing to say – and it's something easy to forget to say – is just what a wonderful achievement it is to send a satellite carrying two such precise instruments up into space, station it at L2, cool the instruments to a tenth of a degree above absolute zero with a fluctuations of less than one part in a million about that, spin the satellite once per minute, scan the whole sky in 9 different frequency bands, subtract all the messy foreground radiation from our own galaxy and even our solar system, all to obtain this perfect image of the universe as it was nearly 14 billion years ago:

The CMB sky according to Planck.

So congratulations and thanks to the Planck team!

Now I said all that first up because I don't want to now sound churlish when I say that overall the results are a little disappointing for cosmologists. This is because, as I noted earlier in the day, there isn't much by way of exciting new results to challenge our current model of the universe. And of course physicists are more excited by evidence that what they have hitherto believed was wrong than by evidence that it continues to appear to be right.

There are however still some results that will be of interest, and where I think you can expect to see a fair amount of debate and new research in the near future.

Firstly, as I pointed out earlier, Planck sees the same large scale anomalies as WMAP, thus confirming that they are real rather than artifacts of some systematic error or foreground contamination (I believe Planck even account for possible contamination from our own solar system, which WMAP didn't do). These anomalies include not enough power on large angular scales ($\ell\leq30$), an asymmetry between the power in two hemispheres, a colder-than-expected large cold spot, and so on.

The problem with these anomalies is that they lie in the grey zone between being not particularly unusual and being definitely something to worry about. Roughly speaking, they're unlikely at around a 1% level. This means that how seriously you take them depends a lot on your personal prejudices priors. One school of thought – let's call it the "North American school" – tends to downplay the importance of anomalies and question the robustness of the statistical methods by which they were analysed. The other – shall we say "European" – school tends instead to play them up a bit: to highlight the differences with theory and to stress the importance of further investigation. Neither approach is wrong, because as I said this is a grey area. But the Planck team, for what it's worth, seem to be in the "European" camp.

The second surprise is the change in the best-fit values for the parameters of the simplest $\Lambda$CDM model. In particular the Hubble parameter is lower than WMAP's, which was already getting a bit low compared to distance-ladder measurements from supernovae. This will be a cause for concern for the people working on distance-ladder measurements, and potentially something interesting for inventive theorists.

And finally, something close to my own heart. A few days ago I wrote a post about the discrepancy in the integrated Sachs-Wolfe signal seen from very rare structures, and pointed out that this effect had now been confirmed in two independent measurements. Almost immediately I had to change that statement, because one of those independent measurements had been partially retracted.

Well, the Planck team have been on the case (here, paper XIX), and have now filled that gap with a second independent measurement (as well as re-confirming the first one). The effect is definitely there to be seen, and it is still discrepant with $\Lambda$CDM theory (though I'll need to read the paper in more detail before quantifying that statement).

So there's a ray of hope for something exciting.

11:30 am CET: Well, ESA's first press conference to announce the cosmological results from Planck has just concluded. The full scientific papers will be released in about an hour, and there will be a proper technical briefing on the results in the afternoon (this first announcement was aimed primarily at the media). However, here is a very quick summary of what I gathered is going to be presented:
  • The standard Lambda Cold Dark Matter Model continues to be a good fit to CMB data
  • However, the best fit parameters have changed: in particular, Planck indicates slightly more dark matter and ordinary (baryonic) matter than WMAP did, and slightly less dark energy. (This is possibly not a very fair comparison – my hunch is that the Planck values are obtained from Planck data alone, whereas the "WMAP values" that were quoted were actually the best fit to WMAP plus additional (non-CMB) datasets.)
  • The value of the Hubble parameter has decreased a bit, to around 67 km/s/Mpc. Given the error bars this is actually getting a bit far away from the value measured from supernovae, whch is around 74 km/s/Mpc. I think the quoted error bars on the measurement from supernovae are underestimated.  
  • The Planck value of the spectral tilt is a bit smaller than, but consistent with, what WMAP found.
  • There is no evidence for extra neutrino-like species.
  • There is no evidence for non-zero neutrino masses.
  • There is no evidence for non-Gaussianity.
  • There is no evidence for deviations from a simple power-law form of the primordial power spectrum.
  • No polarisation data, and therefore no evidence of gravitational waves or their absence, for around another year.
  • There is evidence for anomalies in the large-scale power, consistent with what was seen in WMAP. We'll have to wait and see how statistically significant this is – the general response to the anomalies WMAP saw could be summarised as "interesting, but inconclusive"; I don't think Planck is going to do a lot better than this (and the bigging-up of it in the press conference might have had more to do with the lack of other truly exciting discoveries), but I'd love to be surprised!
That's about all I got out of the media briefing. Obviously we are all waiting for more details this afternoon! 


  1. what about 'dark flow?

    1. There was no mention of anything about the dark flow. Presumably Planck can say something about this because of their SZ measurements, but I don't know what. The papers are now online here, but there are 29 of them, so I can't make any more precise statements just yet!

  2. So does anyone know Why they did not search for B-modes first,
    as a top priority ? If found, this would be a great victory for inflation, and if not, a truly revolutionary result empowering the cyclic universe to center stage.

    1. Searching for the B-mode signal is hard. It's very small and the galactic foreground is much more difficult to remove. It will probably take them another year of working on the data before they can announce any results there, and even then I would advise you not to bet on a positive detection. Meanwhile there is lots of very interesting science that can be done more easily with the temperature power spectrum, lensing, ISW etc.

      Incidentally, a non-detection of a B-mode signal by no means rules out inflation. "Small-field" models of inflation (in which the inflaton field value does not exceed the Planck scale) generally do not predict any detectable gravitational waves (though some exceptions can be constructed).

  3. Thanks for the insights. I have quoted part of your post at length at my blog.

  4. "Incidentally, a non-detection of a B-mode signal by no means rules out inflation."

    What would definitely rule out inflation?

    1. Gosh, tough question. Almost nothing.

      There are many things whose existence, or lack thereof, could rule out some class of inflation models, while favouring other classes of models. These include gravitational waves (the B-mode signal), non-Gaussianity, deviation from a power-law power spectrum, and so on. None of these are seen in Planck. Certain other things, like relatively large curvature, might be hard to explain, though you can get models for that too.

      But inflation as a paradigm? I'm struggling to think of anything that we don't already know doesn't exist. (I also don't know of any fully self-consistent, theoretically sound, believable model of inflation, but that's a different story.)

  5. Sesh,

    I would've expected that Planck was designed
    (< 2009)with the sensitivity to detect B-modes based upon existing inflation models, most of which place the inflaton mass between GUT & reheating scales ~ 10^15 Gev. This is 4 decades below the Planck mass. I find it hard to believe that inflation takes place above the Planck scale.
    Also, if B-modes are so hard to detect, why are there ongoing terrestrial efforts to detect them, e.g. BICEP2, & QUIET @ the S.Pole ? Surely their detection schemes have a higher noise level to contend with than Planck ?

    1. It looks like you've misunderstood. Which is fair enough, because my comments were probably a little cryptic. So let me clarify.

      First up, there's a distinction to be made between the inflaton field value, $\phi$, and the energy scale of inflation, $V(\phi)$, which is proportional to $\phi^4$. "Inflation taking place above the Planck scale" would mean $V(\phi)>M_P^4$; there are no models in which this happens. Indeed, the fact that gravitational waves have not already been seen places an upper limit on the scale of inflation quite a bit lower.

      However, there are plenty of models of models in which $\phi$ itself has to take values larger than the Planck mass. Such models are viewed with suspicion by many field theorists; there are many theoretical consistency issues that I won't go into here. Other models ("small-field models") don't have these problems but may have others.

      Planck's design sensitivity means that it could detect gravitational waves if the tensor-to-scalar ratio $r$ is $>0.05$ or so. A well-known result, called the Lyth bound, shows that in order to get such large values of $r$, the inflaton must take values $\phi>M_P$ (except there is actually a tiny bit of wiggle room here too). Small-field models – which many consider more natural candidates for the correct model of inflation – give values of $r$ orders of magnitude smaller, such that they can never be detectable, not by Planck or any future successors. All inflation models Planck constrains are large-field ones.

      I don't know enough about the experiments you mention to make a detailed comment about their capabilities compared to Planck. However, the ultimate limitation facing attempts to detect gravitational wave backgrounds is that our galaxy is a source of B-mode polarisation which drowns small enough primordial B-mode signals out.