Tuesday, April 9, 2013

The universe as seen by Planck - Days Three and Four II

[Continued from yesterday...]

In the first piece of this post I covered the implications of Planck for the paradigm of inflation. This piece covers the rest.

The anomalies
This is what the CMB would look like in an unphysical Bianchi universe. A worry for our physicality is that this unphysical Bianchi universe seems to fit the data better than a physical \(\Lambda\)CDM universe.

It would be impossible to provide an overview of this conference without mentioning the features and anomalies that Planck has chosen to draw significant attention to. I have a bunch of notes that I've written down that I might one day turn into a new blog post, but I'm not going to delve into them now.

These features and anomalies are clearly going to become a contentious issue in cosmology for the next few years. In fact, the words believer, atheist and agnostic were even being used by speakers during talks regarding whether the anomalies are real or statistical effects. Each time someone declared themselves an anomaly atheist or anomaly agnostic, someone in the audience inevitably spoke up and passionately defended the significance of the questioned anomaly.

The list of potential anomalies is long. There is the cold spot, the anomalously low quadrupole, the hemispherical asymmetry, the statistical difference between the odd and even multipoles at large scales, there is the dipole modulation, there is the general lack of power at large scales, there is the feature in the temperature power spectrum at small scales, the fact that the universe seems to be in an unphysical Bianchi model and there is the "axis of evil" (to name a few).

Pick your side. Atheist, believer or agnostic. The great anomaly wars of cosmology are about to begin (another inevitable consequence of an observational, rather than experimental science, I suppose - i.e. there is only a finite quantity of information available to us, so for some observables we can't just do the experiment again to check who is right).

What should we make of Planck vs SPT and Planck vs the local universe?

The cosmological results Planck has obtained for quantities like the expansion rate of the universe and the overall density of matter in the universe seem to be slightly discrepant with some other measurements of these quantities.

I've already discussed in some detail the nature of, and possible resolutions to, the discrepancy between Planck's CMB measurements and the abundance of galaxy clusters detected by Planck.

There is also a discrepancy between Planck and local measurements of the expansion rate of the universe. The value for the expansion rate inferred from Planck appears to be smaller than what was measured locally using supernovae and Cepheids. Unfortunately no real insight was gained regarding these discrepancies during the conference. One interesting, and mildly concerning, result was that already two of the local measurements of this expansion rate have been "corrected" to produce a smaller value. Although, thankfully, both of the groups claim that these corrections were in the pipeline before Planck released its results.

Finally there is a more curious discrepancy. This is the apparent discrepancy between Planck and the South Pole Telescope (SPT). This is curious because it is a discrepancy between two CMB measurements. In principle SPT should be measuring a sub-set of the same sky that Planck measures. Therefore any discrepancies are harder to understand or put down to misunderstanding the theory (and thus more interesting).

I overheard an interesting conversation during the conference. It was between a senior figure in SPT, a senior figure in WMAP and a senior figure in Planck. They were discussing this discrepancy. What I overheard was that, when Planck's data is examined on the same patch of sky that SPT can see, the data points from SPT and from Planck lie on top of each other. This is very interesting because it suggests that the source of this discrepancy is not coming from either Planck or SPT making a mistake. It is either just sample variance (i.e. a rare fluke), or something more interesting.

On Friday, John Carlstrom, from SPT was part of a discussion panel where he was asked to address this discrepancy. His explanation for the discrepancy suggested two sources (both of which probably combine to cause the change). One is that the power in the WMAP temperature anisotropies is systematically slightly larger than Planck's. SPT is a ground based telescope, which means that it cannot measure the whole sky. Therefore, to do cosmology, SPT needed to combine its small-scale measurements with WMAP's large-scale measurements. Therefore, WMAP's small, but non-zero surplus of power relative to Planck will affect SPT's cosmology. Note that the WMAP people at the conference were taking this small surplus very seriously. On the large scales Planck and WMAP should see exactly the same CMB. The fact that they don't indicates that one of the groups has missed some systematic effect in their telescope (my impression from the conference was that WMAP thought it was them).

The second effect that Carlstrom discussed is an extra "rounding of the small-scale peaks" in Planck as compared to SPT. The implication seemed to be that this could partially be due to the effects of lensing on the CMB by local structures. This lensing effect is real and expected and measured, but it will still vary from line of sight to line of sight. It is conceivable that there is simply less structure to provide lensing along SPT's line of sight, which would mean less rounding of the small-scale peaks, which would mean SPT would require less matter than Planck to describe its measurements.

Of course the probability of this happening is included in both SPT and Planck's error bars, so if there is an excess or shortage of lensing along any line of sight, it would require some sort of new physics in the local universe along that line of sight. Such a conclusion should certainly be approached cautiously, but it is worth observing that all of these discrepancies between Planck and other measurements seem to have potential explanations in modified late-universe physics (i.e. this SPT discrepancy, the cluster abundance discrepancy and the local measurements of the expansion rate).

The late-universe is nowhere near as tightly constrained as the early universe.

Time will tell...

What comes next for CMB science?

An artist's depiction of the Cosmic Origins Explorer (COrE). Hopefully one day COrE will do to CMB polarisation what Planck did to CMB temperature; suck it dry.

Planck has more or less exhausted the temperature anisotropies in the CMB as a useful probe of cosmology. In principle a new telescope could be made that could measure microwave radiation with even better resolution, but the primordial CMB has almost no fluctuations on scales smaller than Planck's best resolution and foregrounds are already dominant on Planck's smallest scales.

So what else is there? Do we give up on the CMB and find other sources of information about the early universe?

Not just yet. Firstly, Planck will not exhaust the polarisation of the CMB. Planck was not initially designed to be a polarisation sensitive telescope and so its polarisation resolution is well below its temperature resolution. There is scope for a future polarisation telescope that will be able to exhaust the polarisation the same way that Planck has exhausted the temperature. An example of such a telescope is the COrE (Cosmic Origins Explorer) telescope proposed to ESA by European collaborations.

In the meantime both the Atacama Cosmology Telescope and South Pole Telescope have attached polarisation detectors to their CMB telescopes and have begun taking high resolution measurements of the small-scale polarisation spectrum. There is also Planck's impending polarisation measurements. The future holds exciting prospects for constraining cosmology using the CMB's polarisation.

But there is another source of potential information lurking in the CMB that is often forgotten. The spectrum of Planck is equivalent to an almost perfect blackbody, but it can't be completely perfect. At some level, there must be some distortion. The best measurements to date of the blackbody spectrum of the CMB were made by COBE (by the FIRAS instrument) in the 1990's. These measurements were rightfully hailed as one of humanity's greatest achievements at the time. The famous plot produced by FIRAS where the theoretical blackbody curve and experimental data points fit so closely that you can't even see the error bars, truly is impressive. However, technology has come a long way in twenty years. Why should we not measure the frequency spectrum of the CMB to even better accuracy?

I can tell you why we should.

Firstly, there is a guaranteed signal. The same physics that takes the power spectrum of temperature anisotropies and reduces the power at small scales, will also introduce a spectral distortion to the blackbody spectrum. On Friday we had a talk from Rashid Sunyaev, another old-school Soviet cosmologist (though Sunyaev is more on the astrophysics side and less on the field theory side than the other two), who told us that this spectral distortion would be measurable by a proposed future telescope called PIXIE.

The prominent physicist from the former Soviet Union that sponsored this post was Rashid Sunyaev. If you stick with the blog long enough you can collect the full set (be warned, there are a lot of them).

So that gives the guaranteed signal, but spectral distortions in the CMB are also a really interesting probe of a variety of new physics possibilities. Firstly, there is the possibility of detecting the effects of the decays of exotic particles. If an exotic particle were to exist in the early universe and was to interact weakly enough with the rest of the constituents of the universe then it would not be in thermal equilibrium with them. If it were to then decay into other particles that did interact with the rest of the universe, then this would appear like an injection of additional thermal energy. If the CMB formed before the universe could reach thermal equilibrium again, then this would introduce a spectral distortion in the CMB that PIXIE could potentially measure. Also, if the primordial density perturbations had a large enough amplitude on very small scales, primordial blackholes would form. These primordial blackholes would also be out of thermal equilibrium with the rest of the universe and would also emit radiation. This again would distort the blackbody spectrum of the CMB.

This raises the interesting possibility that PIXIE could further constrain inflation on distance scales that no other probe could ever realistically constrain. The problem is that the fluctuations on these scales would need to be much larger than what we observe on the larger scales already probed. When you combine this with the fact that on the scales we have observed, the power seems to decrease at smaller scales, it does make the prospects of these spectral distortions existing less plausible. But that doesn't mean they definitely wouldn't be there.

Finally, the SZ effect (where S stands for the same Sunyaev), which is scattering of the CMB off hot gas inside clusters and galaxies, would also introduce a spectral distortion. If PIXIE were to map out this spectral distortion over the whole sky this would give a wonderful map of the large scale structure of the universe.

There is one last source of information lurking inside the CMB. On Wednesday night at the conference a public session was held. One of the attendees asked a wonderful question. He understood that the CMB we see is just an image of a surface, billions of light years away. He wanted to know how long we needed to wait until the image we see changed. I don't think the panel quite understood his question because their answer just related to the inevitable redshifting of the CMB's temperature, but as well as that the actual values of the anisotropies will change. This is because, if we wait long enough, we're literally looking at another patch of the primordial hydrogen plasma with different over/under-densities.

Thankfully Lyman Page was listening and understood the question, and, during his talk on Friday he answered it. The answer to this question is present in Stuart Lange's senior thesis and I think it's really cool. The answer is that, if we wait billions of years, the CMB will have completely changed. You can see a simulated film of this below (note the scale in years). This might then sound like it is just a cute curiosity, nice but unobservable. But it isn't quite. Using very ambitious estimates, Lange estimates in his thesis that even in a timeframe of the order of 100 years the minute differences that will have occurred might just be detectable. If they were, we could then do real-time cosmology by calculating the statistical properties of this difference map. Those will be cool days.

This is how the CMB anisotropies would change over time if we sat and watched for long enough (obviously this is only a simulation, we don't know how the real universe will change with time). What a cool senior thesis topic!


And I'm done. Hopefully I will get around to writing something about the anomalies at some point. However, given how contentious they appear to have become I want to be careful and precise whenever I do choose to write about them.

My lasting impressions of the conference were the following:

  • A measurement that a parameter is zero, is still a measurement. Planck has told us some genuine facts about the very early universe. Those facts might not be extravagant, but they are facts. It's really quite amazing that as a species we've managed through observation and reasoning to learn anything about what the laws of physics are like at those incredibly high energies.
  • Planck might not have measured a primordial bispectrum, but the fact that they have detected, with reasonable significance, the secondary bispectrum that arises from a lensing-ISW correlation is important to remember. It means is that a power spectrum is no longer enough to fully describe the statistical properties of a Planck CMB map. The era of non-Gaussianity has begun.
  • All of the discrepancies between Planck and other experiments potentially relate to late-universe effects. The early universe is quite well constrained now, but the late-universe is not at all. Could we be starting to see the signs of new physics in the late-universe? It is true that the most significant of these discrepancies, the under-abundance of galaxy clusters in ACT, SPT and Planck's cluster samples, could be the result of biases in the mass measurements of these clusters. However as Dick Bond said in the panel-discussion on Friday afternoon "ignore clusters at your peril". Clusters have been the unheeded heralds of new physics before, might they be trying to tell us something interesting again?
The only unquestionably statistically significant anomaly from Planck. Ignore clusters at your peril...

(Note that, in the 1990's there was a "missing mass" problem relating to the masses of galaxy clusters. People were convinced that the universe was flat, but if it was that required more energy density in the universe than what was present in clusters, hence the "missing mass". That missing energy density is what we now call dark energy, but it took the supernovae measurements before people were convinced that the cluster measurements weren't just wrong.)

  • Finally, just because no deviations from \(\Lambda\)CDM have yet been measured, does not mean that no deviations exist. Even in the early universe there is still plenty of room for surprises.

Twitter: @just_shaun


  1. Thanks a lot for the exhaustive summaries of the first post-Planck conference! Your remarks here "that already two of the local measurements of this expansion rate have been 'corrected' to produce a smaller value" of the Hubble constant puzzle me, though: after the low Planck value had come out I contacted the lead authors of http://arxiv.org/abs/1208.3281 which last year had been widely hailed as the last word on H0 - and they don't yield to Planck. And then there is the new paper http://arxiv.org/abs/1303.5341 making a good case for taking the "local" values seriously and pointing out 'tensions' already with WMAP. Which seem to have grown now ...

    1. Thanks Daniel.

      Wendy Freedman, was at the conference last week and was certainly still standing firmly by the measurements in that paper, so you're right. In fact, she was saying that given the enormous difference in the types of measurements being made of this one parameter (i.e. H_0), the agreement is remarkable (and any disagreement between the values was not significant, in her opinion).

      In this table I think it was UGC 3789 and RXJ1131-1231 who had or were going to revise their measurements of H_0 down (though I might be remembering badly).

      Hopefully the slides come out soon, but the first mention of "corrected" H_0 measurements came in Efstathiou's talk on Tuesday and the second came in Pain's talk on Wednesday. They each showed updated versions of the table I've linked to above in this comment.

      I wasn't meaning to imply that this was going to go away soon. My apologies if that's the impression I gave.

  2. Thank you very much Shaun, for the series of posts on the conference. They are very interesting and informative, and partly beyond my understanding. :)

    I remember that the WMAP team announced they have seen a very strong quadruple signal. If I have understood your post correctly, they are gone in the Planck mission?

    And about the dipole modulation, I might have read from somewhere that the dipole can be sourced by a change in the reference frame. Is it true?

    I totally have no idea what is an "unphysical Bianchi model". Would you mind referring to some articles, explaining its relevance with the Planck results?

    1. Thanks Lingfei.

      I'm not sure what you're referring to regarding the quadrupole. My understanding is that both WMAP and Planck see an anomalously small quadrupole. That signal is there in both.

      I don't know much about the anomalous dipole modulation signal (its potential cause or its detection). However you are right that our motion with respect to the CMB would cause a dipole and modulation of all the multipoles. The dipole our motion causes is very large and was detected a long time ago. The modulation was actually detected for the first time by Planck, in this paper. It is along the direction of the main dipole. However the anomalous dipole modulation appears to be along another axis and appears to be bigger (I think).

      I hope that helps.

      I've asked someone to help with a coment about the Bianchi model. I'll have a go if he doesn't get back to me, but in the meantime, the paper it is from is this one.

    2. Andrew Pontzen asserts that the particular Bianchi model the Planck team used is inconsistent with nucleosynthesis, therefore unphysical. Other Bianchi models can be consistent with nucleosynthesis. I don't have a reference, but he has studied Bianchi models himself in the past, so I'm sure you could find something among the list of his papers on InSpire.

    3. I had the impression that the model itself wasn't physically consistent, not that it just wasn't consistent with other observations. For example, it seems to require the density of the universe to be multi-valued (?).

      The guy I've asked for feedback from did the calculation in Planck, so hopefully he does get back to me. We'll see...

    4. As Sesh says, the Bianchi models studied by Andrew Pontzen are more realistic than the simple models that we studied in the Planck analysis. However, the non-physical setting studied in the Planck analysis refers to the situation where the parameters describing the Bianchi model are decoupled from their standard meaning. For example, the energy density used to describe the Bianchi model differs to the usual cosmological value, although physically they should be the same.

      The reason for performing a non-physical study was to make comparisons with the analysis of WMAP data, where a Bianchi component was seen previously. In these previous studies the Bianchi parameters were not fitted simultaneously with the other cosmological parameters and thus were allowed to vary from their standard values. Interesting, we do indeed see a very similar Bianchi component in Planck data in this non-physical setting.

      However, we go beyond this setting in Planck and also perform a more physical analysis where the parameters describing the Bianchi model are coupled to the usual cosmology. In this setting, we find no evidence for a Bianchi component in Planck data.

    5. So many replies within days. Thank you all.

      Shaun, I think you are right on the dipole modulation, and I previously confused dipole and dipole modulation. About the quadrupole anomaly, I remember there were quite some open discussions around 1 year ago, on possibly a strong signal along a certain axis. I don't remember any settlement on that, so perhaps it's just some preferential hypothesis.

      Sesh, it seems that if you do two fits, one with the constraint on nucleosynthesis and one without it, the one without this constraint would always be statistically favored. In that case, people would call it excluded instead of unphysical, and people will not continue to talk about it. So for me, there must be some other reason to call it unphysical, as is explained by Jason.

      Jason, thank you for the clarification. So the previous analysis did not fully consider the dependence of the Bianchi parameters. Now Planck does and we find the simplest universe does not need such anisotropies.