From CMB-S4 wiki
Jump to: navigation, search

Back to SLAC-2017 main page

Clusters and High ell science

Chairing: David Spergel

General context

Cosmology focused part - mnu and w

  • 11.08 - 11.18 Nick Battaglia - mnu and w from lensing-calibrated clusters slides
  • 11.20 - 11.30 Mat Madhavacheril - mnu and w from CMB-calibrated clusters slides
  • 11.32 - 11.42 Colin Hill - mnu and w from tSZ power spectrum slides
  • 11.44 - 11.49 Lindsey Bleem - Cluster Simulation tools at ANL here
  • 11.50 - 12.00 Christian Reichardt - forecasting cosmology from cluster lensing, and kSZ with 4MOST File:20170227 cmbs4.pdf
  • 12.02 - 12.12 David Alonso - growth function from kSZ, comparison to DESI/Euclid File:KSZgrowth.pdf

Astrophysics focused part - reionization and cluster astrophysics

  • 12.14 - 12.24 Simone Ferraro - reionization case from kSZ slides
  • 12.26 - 12.36 Nick Battaglia - cluster astro case from tSZ+kSZ slides
  • 12.38 - 12.48 Jim Bartlett - Probing the Circumgalactic Medium File:Bartlett.pdf
  • David Spergel - few minute wrap-up

Notes from session

Notes from Steve Allen's talk:

  • 1 arcminute resolution provides an enormous increase in cluster discovery potential relative to 3 arcminutes
  • 1 arcminute would greatly increase galaxy science, and the size of the community served.

Here is the strawman conversion between size of telescope and FWHM as a function of channel that Mat and Nick were looking at (provided by Darcy Barron/Mike Niemack). Just a strawman. Ignore the noise column.

Tel size.png


  • Can do extraordinary things in cluster astro and cluster cosmology with sufficiently small beam.
  • Current leading catalogs are X-ray, optical, SZ. Need a well-defined selection function, and mass-observable relation. This used to be a problem. But, now we split into two parts: relative mass (x-rays) and absolute mass (lensing).
  • Current best constraints are Mantz et al from 220 ROSAT clusters, z<0.5, with Chandra follow-up and WtG WL masses. Similar results come from from Planck clusters. Currently have 15% on DE w from clusters.
  • Radical change coming. Optical: DES/HSC, Euclid, LSST. Mm: SPT-3G, ACT, S4. End of 2017 eROSITA for X-ray.
  • Strengths for optical: cluster finding, photo-z, WL mass calibration. Strengths for mm: high-z clusters, CMB-WL mass calibration. Strengths for X-rays: cluster-finding, low-scatter mass proxies.
  • Target: 100k clusters; far stronger in combination.
  • Optical clusters are limited to z=1.2. X-ray eROSITA find all clusters at low-z, some up to 1.5. Unique discovery space for CMB is at z=1.5 and above, up to z=3. A beam of 3’ loses much of the science.
  • On galaxy cluster astrophysics: we can find virialized haloes out to where they first formed. Things to study: Impact of environment on triggering and quenching of star formation and AGN activity. Evolution of feedback process. Integrated history of star formation.
  • The majority of SPT papers came from cluster astrophysics/cosmology and galaxy astrophysics. (Spergel - ACT too)
  • Conclusion: with 1’ we can be transformative.
  • Douglas Scott: dubious about cluster cosmology. Allen: results have held up.


  • Forecasting constraining power of mass-calibrated SZ cluster counts
  • Summary of code: non-white-noise, 1.5 uK’ at 90 and 150, includes extragalactic foregrounds, marginalizes over scaling relation. Planck tau prior 0.01.
  • Need to calibrate SZ masses, here use optical WL to get mass errors as function of mass. Assume HSC-like coverage and use optical WL out to z=1 or z=2.
  • Going from 5m to 6m to 7m does not affect counts and forecasts, but 3m is much worse
  • Main conclusion: neutrino mass is well constrained by clusters +Planck (as good as CMB lensing +DESI), using both z=1 and z=2
  • Completely independent and complementary to lensing and DESI. Forecast a better w than DESI BAO.
  • Going from 5m to 7m gets about 2x more clusters.
  • Optical surveys are required to obtain redshifts.
  • Requirements: couple percent on w, and 2.5-3 sigma on mnu minimal mass.
  • Allen: likely can do even better, and may want higher res, if look at w model that varies at high z.


  • Showing forecasts with new code using CMB-lensing-calibrated masses. CMB lensing takes over at high z.
  • You can do better on w if you use internal CMB lensing mass calibration up to z=3, compared to WL-calibration up to z=1.
  • Find same scaling with size as for WL-calibration: want to be >= 5m.
  • Science targets: roughly 35 meV for mnu, 2.5% on w. This is without DESI, just with Planck.
  • Atmosphere and tau-limit means the improvement going from 5-7m is only 5% for mnu. CVL on tau gives more improvement going from 5m to 7m (40%).
  • For w, you get ~30% improvement going from 5m-7m (with atmosphere and Planck tau)
  • If you use 5m with internal P only, do nearly as well as with T+P, which mitigates worries about temperature contamination.
  • Two calibration methods to get competitive neutrino mass, and interesting w limits.
  • Request from Charles to talk about resolution not size. Jo/others: the resolutions are now posted.
  • Allen – looking at w(z) could push to larger telescope.
  • Bond – what do we need to get ‘gold sample’? Might be higher resolution. Nick agrees.
  • Carlstrom, Dodelson: show constraint on sigma8(z)


  • Looking at information in the tSZ power spectrum. There is also wealth of info in the full pdf, but only showing spectrum forecasts today.
  • Includes tSZ trispectrum.
  • Can get down to 25 meV on neutrino mass with CVL tau. If marginalize over gas physics get e.g. 50 meV. Does a lot better than Planck.
  • Includes atmospheric noise and foregrounds - has post-component-separation noise, of which CIB is the main one.
  • To improve forecasts, should include Planck low-ell and include non-Gaussian pdf. Target is to combine with cluster counts.
  • Promising additional probe of mnu and w, and similar measurement requirements to cluster counts.
  • Suzanne: how orthogonal are these constraints. Colin - complementary, so should add together.


  • Describes simulations from HACC code at Argonne: N-body simulations. Building emulator to get 1% halo mass function. Getting ready to do realistic cluster work.
  • Bartlett– really need to calibrate mass function. This is a big challenge. Bleem – yes, these sims will be important for that.


  • Polarization wins for cluster lensing below 2 uK/amin, good as foregrounds are simpler. Most biases considered, e.g. offset centres, are sub-percent.
  • S/N increases by 1.7 going from 3’ to 2’. This is consistent with findings of Mat/Nick.
  • CMB lensing systematics much easier to deal with than optical WL calibration.
  • Promise of kSZ with southern spectro survey 4MOST, first light in 2021.


  • Forecasting measurement of growth of structure from kSZ. There are hints of problems in current data so this could be useful.
  • Constraints depend on the quality of the overlapping spectro survey (e.g. DESI better than BOSS).
  • Strength of kSZ compared to other data: you can measure growth at low-z where RSD is limited. Beats DESI and at low-z Euclid.
  • But need to know details about how you make the measurement, as there are many uncertainties about extracting the signal and e.g. measuring tau (v and tau are degenerate).
  • Pros: This is an alternative measurement with different systematics. Cons: Drawbacks are systematics in y-tau relation and wrong kSZ profile model.
  • Conclusion: growth from kSZ may not be main science driver of S4 but adds important information that has different systematics.
  • Spergel – can get more information at higher-z if you include extra statistics


  • Showing forecasts on patchy reionization
  • We would like to be able to tell different models apart
  • kSZ demands more frequencies if you want to do it in power spectrum.
  • Can get 10% constraints on duration of reionization.
  • Can also use higher order statistics - trispectrum. Lets you separate the reionization part from the late-time kSZ.
  • Going from 3’ to 2’ improves by factor 4. Prefer deep narrow patch.
  • kSZ gets different scales to 21cm – CMB can do bubble power spectrum.
  • Dunkley – clarify scales that kSZ can do that 21cm can’t, to help science case. Ferraro - yes, and should compare to e.g. HERA kSZ can see smaller scales.


  • What are the astrophysics things we want to know? Feedback and non-thermal pressure support. These things are not well known. In galaxy evolution these parameters are thrown into simulations.
  • Particular parameters of interest are global feedback efficiency and fraction of non-thermal pressure support.
  • 4m to 5m sees big improvement on these, but less so for 6m. Targeting percent level precision on these things as science requirements.
  • Need spectro surveys to get redshifts but could use other estimators.
  • Staggs: why do we care? Spergel: Astronomers care about AGN feedback and galaxy formation. Big question is the amplitude of feedback. Would like to know how it evolves with redshift for different galaxies. Also for cosmology, if we want to use small-scale data in WL surveys, we need to know the role of feedback.
  • Allen – would be good to be clearer on synergies with Athena satellite. Show complementarity to that.


  • SZ can have big impact on biggest question in galaxy formation. Galaxy formation is very inefficient. Less than 10% of baryons make stars. We don’t know why. We can find out why.
  • Bulk of baryons is in the CGM (circumgalactic medium). Keeping it there requires efficient feedback.
  • So, need to observe CGM, rather than galaxies. SZ is unbeatable way to do this. Need higher redshifts – push out to z=2. Study CGM versus mass and z, and by galaxy type and property to high z.
  • Need halo masses of objects which we use CMB lensing for. S4 can get to much lower halo masses, where star formation is taking place.
  • Dust is an issue though. Contaminates signal and needs to be separated from SZ. Example has 220 GHz as highest frequency.
  • Dust from the halos dominates – will need multiple frequencies. Might need channels not available from ground.
  • Community who want to answer these galaxy formation questions don’t know that we have this great machine, should make clear.
  • Spergel – can we use CCAT’ to measure some of them to address the dust? Maybe, will hear more tomorrow. Or Herschel? Jim: you want big samples so need wide area.

Spergel wrap-up

  • cluster science is so rich. We can’t lose it so need to determine what size telescope we need. It is unique science that other data cannot be do. Majority of papers from S4 will be from this IF we have high enough resolution.
  • Need to explain our case to our astrophysics colleagues, to convey what we can tell them about galaxy formation. Need to articulate it more broadly to astrophysics community.
  • Keating – is this really what we should be doing with polarimeter? Spergel - Remember CMB lensing relies on polarimetry. We are limited by calibration, and CMB lensing will let us calibrate the maps.
  • Comment: Having neutrino mass from two probes is huge benefit and cross-check is significant value (in particular to persuade particle physicists).
  • Bond – can we split into delta nu

Action items/Next steps

Summarize action items here

High level:

  • Define neutrino mass requirement from cluster counts
  • Define sigma(w(z)) requirement from counts, and growth function requirement from kSZ
  • Define reionization requirement
  • Define astrophysics requirements (e.g. on feedback parameters)
  • Then define measurement requirements for these things

Details that came up in talks include:

  • Extend cluster forecasts to extended w models, check resolution requirements
  • Look at 4m-5m range for calibrated cluster count forecasts
  • Include scatter in scaling relation for cluster counts
  • Combine cluster counts and SZ pdf together
  • Compare patchy reionization constraints clearly to 21cm
  • ...