This week’s entry covers the exciting ACT data release, the mutual tension between ACT, Planck, and SPT, and a flexible parametrization of black hole shadows based on a class of curves known as limaçons. Have a nice weekend!
#1 2007.07288: The Atacama Cosmology Telescope: DR4 Maps and Cosmological Parameters by the Atacama Cosmology Telescope collaboration
This has been an exciting week as the Atacama Cosmology Telescope (ACT) collaboration, which some of you might remember I already discussed in my Week 14 entry, finally released their data release 4 (DR4) products, including maps and constraints on cosmological parameters. ACT is a ground-based Gregorian telescope located in the Atacama Desert in Chile. Despite being ground-based, its relatively low latitude (low as in being close to the Equator) of ~23°S allows ACT to map a good portion of the sky, 17000 deg^2 which is almost half the sky. Besides measuring the anisotropies in temperature fluctuations of the Cosmic Microwave Background, ACT also presented exquisite measurements of E-mode polarization, and of course of temperature-E-mode cross-correlations. Why is it exciting that the ACT results are coming out just now? Because with all the hype around cosmic tensions (such as the H0 tension), it has become really important that the Planck results be scrutinized by an independent experiment. ACT was the first experiment that could do this, and their results overall provide an independent confirmation of the soundness of the Planck results. From my point of view, this is huge.
In my opinion, there are three key take-away points from this week’s paper:
1) the overall confirmation of the goodness of Planck’s results. In this sense, Fig. 11 is very instructive, considering both ACT alone and various combinations of ACT with external priors and/or datasets (including WMAP and Planck). Perhaps one of the most interesting cosmological parameter constraints is that of H0=67.9±1.5 from ACT alone. While the error bar is about 3 times larger than the Planck error bar, the central value is in excellent agreement with the latter. This lends support, if any support was needed, to the fact that the H0 tension has nothing to do with some systematic in the Planck data (again, I don’t think anybody seriously believed the contrary, but it is always good to have independent confirmation).
2) however, it’s not all fun and games. There is a mild tension between ACT and Planck in both the baryon fraction Omegabh2, and the scalar spectral index ns. While this tension is very mild, it is worth understanding where it comes from. It basically comes from the fact that because of its limited sky fraction, ACT cannot measure the first peak well. There is then a fundamental Omegabh2-ns degeneracy, and along this degeneracy ACT data happens to prefer values of these two parameters respectively lower and higher than the Planck value. Including external information on the amplitude of the first peak (e.g. from WMAP large-scale data) allows one to break this degeneracy, and then everything is consistent with Planck. Using Eq.(14), the tension with Planck was estimated to be at the level of 2.7σ.
3) The (unphysical) Alens parameter, artificially enhancing or reducing the lensing-induced smoothing of the acoustic peaks, is consistent with Alens=1. This lends support to the idea that the Planck internal Alens tension (of which you can read more in my Week 21 entry), and by extension the “curvature tension”, might actually be a statistical fluctuation in Planck data (you can read more about various points of view in this “curvature debate” in 1908.09139, 1910.00483, 1911.02087, 2002.06892, and 2003.04935).
I definitely look forward to more data which will be coming both from ACT and Advanced ACT!
#2 2007.08496: Quantifying the global parameter tensions between ACT, SPT and Planck by Will Handley and Pablo Lemos
This very interesting paper expands a bit more on point 2) above, regarding the overall consistency of ACT with Planck. In the previous paper, using the metric given by their Eq.(14), the results of ACT and Planck were found to agree (or to be in tension, whichever way you prefer to look at it in this grey zone) within 2.7σ. In general, comparing high-dimensional Bayesian probability distributions is rather challenging, as it is challenging to define a notion of distance between distributions in higher dimensions, but such a notion is crucial in order to assess the consistency or tension between different experiments whose results cannot simply be condensed in one number (such as H0). In the past years, a lot of work has gone into assessing the concordance or discordance between datasets living in many dimensions, and the Handley-Lemos duo also played an important role in this, for instance in 1902.04029. In this week’s paper, Handley (whose office - outside pandemic hours - happens to be almost across mine) and Lemos try to quantify the overall level of concordance between Planck, ACT, and SPT. The commonly used Bayes ratio is argued to be problematic due to its intrinsic dependence on prior choices, and therefore what is used is the so-called suspiciousness statistic, developed by Handley and Lemos in 1902.04029, which naturally factors out this prior dependence and allows for a more prior-independent assessment of the tension.
The results are very interesting, and nicely summarized in Table I. While Planck and SPT are not in tension, there is a 2.6σ tension between ACT and Planck (in agreement with the findings of the ACT collaboration), a 2.4σ tension between ACT and SPT, and a 2.8σ tension between ACT and Planck+SPT (the latter two can be safely combined as they are consistent). I suspect that the tension between ACT and Planck is coming in an important measure from the tension between Omegabh2 and ns I discussed previously, due to the intrinsic degeneracy between the two which can only be broken by an accurate measurements of the first acoustic peak. In any case, this tension, which can be qualified as being mild-to-moderate, casts some doubts on the legitimacy of the ACT-Planck dataset combination, a combination which instead was considered in the ACT paper I discussed above. While I found this paper very interesting, I do have a small concern in that of course Handley and Lemos could not use the full ACT likelihood (which was only released yesterday I believe), but only a Gaussian approximation thereof (were the relevant numbers already available so early on, well before the likelihood was released? I honestly don’t remember!). This is argued to be a good approximation given Fig. 1, and the good agreement with the 2.7σ tension found by ACT reinforces the goodness of this approximation. Overall, another reason why I really liked this paper is because of its also stressing a simple, important, but often overlooked message (which I have often stressed in this blog): you cannot combine datasets which are in tension with each other (a point which was also remarked rather strongly, in a different context within the scope of the discussion concerning the “curvature tension”, by Handley in 1908.09139 and by Di Valentino et al. 1911.02087).
#3 2007.06723: On the approximation of the black hole shadow with a simple polar curve by Joseph R. Farah et al.
Black hole (BH) shadows are a recurrent theme in this blog, so let me just remind the reader that the BH shadow is the apparent (i.e. gravitationally lensed) image of the photon sphere (the region of space-time where gravity is so strong that photons travel in unstable, not necessarily circular, orbits). In other words, the BH shadow is the closed curve on the sky separating capture from scattering orbits, and is not a direct image of the event horizon. Or, equivalently, the BH shadow is the observer’s line-of-sight into the BH. As we await higher-resolution observations of BH shadows from VLBI projects beyond what the Event Horizon Telescope was able to achieve, it becomes important to develop flexible model fitting and inference tools which can allow one to rapidly assess the consistency of the Kerr metric with observations of BH shadows. Many earlier studies have condensed the information on Kerr BH shadows using summary statistics such as mean diameter, asymmetry (or deviation from circularity) and displacement from the origin, although the latter is for all intents and purposes unobservable. Others have used more complex approaches making use of Legendre expansions or principal components. Another interesting approach, lying somewhere in between, consists instead of parametrizing the shadow using a family of simple curves. A seminal paper in this sense was that of de Vries at this link (as far as I know it is not on the arXiv). This week’s paper by Farah and collaborators basically expands on de Vries’ approach.
The family of polar curves consideres in this week’s paper is known as limaçons (which I suspect is related to the French word for “snail”), and is formed by tracing a fixed point on a circle as it rotates around a second circle of equal radius but displaced by one diameter (this Wikipedia animation is rather instructive in this sense). The resulting figure does bear some resemblance to a snail, or to a cardioid (a heart) which indeed is a particular case of a limaçon. Farah and collaborators compare the fitting capabilities of a displaced ellipse, a limaçon, and a generalized limaçon. They find that the limaçon generally provides a better fit than the displaced ellipse while at the same time requiring one less parameter, while the generalized limaçon does even better especially at high spin. These parametric curves are also useful in that they naturally capture degeneracies between BH parameters one gets by observing the shadow alone. More precisely, there are mutual degeneracies in the mass-spin-inclination angle parameter space which prevent one from precisely determining all three these quantities at once, but can be lifted if independent measurements of one of them are available. The findings of this week’s paper will definitely be very useful for flexibly parametrizing and testing deviations from the Kerr metric, especially in light of future observations of BH shadows.