When the Planck science team announced their results in March, they also put out a great flood of papers. You can find the list here; there are 29 of them, plus an explanatory statement.
Except if you look carefully, only 28 of the papers have actually been released. Paper XI, 'Consistency of the data', is still listed as "in preparation". Now, what this paper was supposed to cover was the question of how consistent Planck results were with previous CMB experiments, such as WMAP. We already knew that there were some inconsistencies, both in the derived cosmological parameters such as the dark energy density and the Hubble parameter, and in the overall normalization of the power seen on large scales. We might expect this missing paper to tell us the reason for the inconsistencies, and perhaps to indicate which experiment got it wrong (if any). The problem is that at present there is no indication when we can expect this paper to arrive – when asked, members of the Planck team only say "soon". I presume that the reason for the delay is that they are having some unforeseen difficulty in the analysis.
However, if you were paying attention last week, you might have noticed a new submission to the arXiv that provided an interesting little insight into what might be going on. This paper by Anne Mette Frejsel, Martin Hansen and Hao Liu – the authors are at the Niels Bohr Institute in Copenhagen, and in fact all three recently visited Bielefeld for our Kosmologietag workshop – applied a particular consistency check to Planck and WMAP data ... and found WMAP wanting.
The test they applied is really pleasingly simple. Suppose you want to measure the CMB temperature anisotropies on the sky using your wonderful satellite – either WMAP or Planck. Unfortunately, there's a great big galaxy (our galaxy) in the way, obscuring quite a large fraction of the sky:
Except if you look carefully, only 28 of the papers have actually been released. Paper XI, 'Consistency of the data', is still listed as "in preparation". Now, what this paper was supposed to cover was the question of how consistent Planck results were with previous CMB experiments, such as WMAP. We already knew that there were some inconsistencies, both in the derived cosmological parameters such as the dark energy density and the Hubble parameter, and in the overall normalization of the power seen on large scales. We might expect this missing paper to tell us the reason for the inconsistencies, and perhaps to indicate which experiment got it wrong (if any). The problem is that at present there is no indication when we can expect this paper to arrive – when asked, members of the Planck team only say "soon". I presume that the reason for the delay is that they are having some unforeseen difficulty in the analysis.
However, if you were paying attention last week, you might have noticed a new submission to the arXiv that provided an interesting little insight into what might be going on. This paper by Anne Mette Frejsel, Martin Hansen and Hao Liu – the authors are at the Niels Bohr Institute in Copenhagen, and in fact all three recently visited Bielefeld for our Kosmologietag workshop – applied a particular consistency check to Planck and WMAP data ... and found WMAP wanting.
The test they applied is really pleasingly simple. Suppose you want to measure the CMB temperature anisotropies on the sky using your wonderful satellite – either WMAP or Planck. Unfortunately, there's a great big galaxy (our galaxy) in the way, obscuring quite a large fraction of the sky:
 |
| The CMB sky as seen by Planck in the 353 GHz channel. Obviously there's a lot of foreground in the way. (This is not the best frequency for viewing the CMB, by the way. I chose it only because it illustrates the foregrounds quite nicely!) |
Now, as I've mentioned before, there are clever ways of removing this foreground and getting to the underlying CMB signal. The CMB signal is what is interesting for cosmologists, because that is what gives us the insight into fundamental physics. Foregrounds are about messy stuff to do with the distribution of dust in our galaxy: the details are complicated, but the underlying physics is not that interesting (ok, maybe it is, but to different people). Anyway, using their clever techniques (and measurements of the CMB+foreground at several different frequencies), the guys at Planck or WMAP can come up with the best map they can that they think represents the CMB with foreground removed.
 |
| Planck's SMICA map of the CMB. |
The map above shows the Planck team's effort. Well actually they produced four different such "CMB only" maps, constructed by four different methods of removing the foregrounds. These are known as the SMICA, SEVEM, NILC and Commander-Ruler maps, the names indicating the different foreground-removal algorithms used. For some reason, Commander-Ruler appears not to be recommended for general use. WMAP on the other hand produced only one, known as the Internal Linear Combination or ILC map. (Planck's NILC is meant to be a counterpart to WMAP's ILC.)
Now, although the algorithms used to produce these maps are, as I said, very clever, the resultant maps are never going to be completely foreground-free. Let's express this as the equation
map = CMB + noise
where the "noise" term includes foregrounds as well as instrument noise, systematics and other contaminants. If you have more than one map, they see the same fundamental CMB, but the noise contribution to each is different. So you can subtract one from the other to get a new map consisting of their difference:
difference = map1 – map2 = noise1 – noise2.
Since most of the residual noise should be due to the galactic foreground, most of the features in the difference map should be around the galactic plane. If the foreground removal has been reasonably successful, these features should also be small. And for the Planck maps, that is in fact what Frejsel, Hansen and Liu find:
So the various different methods used by Planck seem to give self-consistent answers.
The same is not true, however, for WMAP. Of course WMAP only use the one method of removing foreground, but they did provide different maps based on the data they had collected after 7 years of operation and after 9 years. The ILC9–ILC7 difference map looks quite different:
Most of the difference appears well away from the galactic plane, as you can see in the right-hand figure, where the galaxy is masked out. So there is some important source of noise that is not foregrounds – so probably some systematics – that has affected the WMAP ILC map. Even more importantly, it is some kind of systematic effect that has changed between the 7-year and the 9-year WMAP data releases, meaning that the ILC9 and ILC7 maps do not appear to be consistent with each other. Frejsel et al. discuss a method of quantifying this, but I won't go into that here because the impression created by the images is both dramatic enough and entirely in line with the quantitative analysis.
As you might have expected, the same method shows that WMAP's ILC9 map is thoroughly inconsistent with the various Planck maps (the picture here is even worse than that between the two ILC maps). But perhaps surprisingly, ILC7 is perfectly consistent with Planck. So it appears that whatever might have affected the WMAP results only affected the final data release.
I guess one should be careful not to make too much of a fuss about this. The results from Planck and WMAP are, generally speaking, in pretty good agreement, except for some problems at the very largest scales. It is also true that the WMAP team themselves do not use the ILC map for most of their analysis (except for the low multipoles, $\ell<32$ – that is, the very largest scales!). But I'm sure this paper will provoke some head-scratching among the WMAP team as they try to figure out what has happened here. Oh, and if you are cosmologist using the ILC9 map for your own analysis, you should probably check whatever conclusions you draw using some other maps before publishing!
All in all, I think I'm rather looking forward to Planck's consistency paper when it does finally come out!
 |
| NILC–SMICA, NILC–SEVEM and SMICA–SEVEM difference maps. Figure from arXiv:1305.4033. |
So the various different methods used by Planck seem to give self-consistent answers.
The same is not true, however, for WMAP. Of course WMAP only use the one method of removing foreground, but they did provide different maps based on the data they had collected after 7 years of operation and after 9 years. The ILC9–ILC7 difference map looks quite different:
 |
| ILC9–ILC7 difference map on the left, and with a galactic mask overlaid on the right. Figure from arXiv:1305.4033. |
Most of the difference appears well away from the galactic plane, as you can see in the right-hand figure, where the galaxy is masked out. So there is some important source of noise that is not foregrounds – so probably some systematics – that has affected the WMAP ILC map. Even more importantly, it is some kind of systematic effect that has changed between the 7-year and the 9-year WMAP data releases, meaning that the ILC9 and ILC7 maps do not appear to be consistent with each other. Frejsel et al. discuss a method of quantifying this, but I won't go into that here because the impression created by the images is both dramatic enough and entirely in line with the quantitative analysis.
As you might have expected, the same method shows that WMAP's ILC9 map is thoroughly inconsistent with the various Planck maps (the picture here is even worse than that between the two ILC maps). But perhaps surprisingly, ILC7 is perfectly consistent with Planck. So it appears that whatever might have affected the WMAP results only affected the final data release.
I guess one should be careful not to make too much of a fuss about this. The results from Planck and WMAP are, generally speaking, in pretty good agreement, except for some problems at the very largest scales. It is also true that the WMAP team themselves do not use the ILC map for most of their analysis (except for the low multipoles, $\ell<32$ – that is, the very largest scales!). But I'm sure this paper will provoke some head-scratching among the WMAP team as they try to figure out what has happened here. Oh, and if you are cosmologist using the ILC9 map for your own analysis, you should probably check whatever conclusions you draw using some other maps before publishing!
All in all, I think I'm rather looking forward to Planck's consistency paper when it does finally come out!
