CMB-S4 Clusters / SZ

CMB-S4 Summer collaboration meeting,

2 August 2023

Speakers:

• Lindsey Bleem
• Will Coulton
• Iñigo Zubeldia
• Erwin Lau
• Aleksandra (Ola) Kusiak
• Matthew Johnson
• Chun-Hao To

​

zoom-link

Organisers:

Srinivasan Raghunathan & Boris Bolliet

From SPT to CMB-S4 Cluster Catalogs and Cosmology�

Lindsey Bleem

Argonne National Laboratory

2

11-11.10 a.m.

(remote)

3

24

Cluster finding and

number count likelihood

Iñigo Zubeldia

38

11.20-11.25 a.m.

(remote)

IoA/KICC, Cambridge

39

Improvements to MMF cluster detection method

• Iterative noise covariance estimation: boosts signal-to-noise and gets rid of ILC-like bias in cluster tSZ observable.

​

(IZ, Rotti, Chluba & Battye, 2204.13780)

​

• Foreground deprojection with spectrally constrained MMF: with moment expansion, highly effective at removing CIB bias in tSZ observable.

​

(IZ, Chluba & Battye, 2212.07410)

​

• Implemented in SZiFi, the Sunyaev-Zeldovich iterative Finder.

github.com/inigozubeldia/szifi

New cluster number count code: cnc

Easy-to-use, flexible cluster number count code:

​

​

• Unbinned, binned, and extreme value likelihoods.
• Arbitrary number of mass proxies, with the possibility of each cluster in the sample having different combinations of them.
• Links mass to mass proxies with model with arbitrary number of layers, each layer allowing for correlated scatter.
• Redshift measurement uncertainties.
• Unconfirmed detections.
• Stacked data (modelled consistently with cluster detection).
• Written in Python, interfaced with Cobaya.
• Fast (Planck cnc likelihood in 1s on a laptop) and accurate (biases less than 0.2 sigma).
• Publicly available soon, paper in prep.

With Boris Bolliet

Baryon Pasting

Erwin Lau (SAO)

on behalf of the Baryon Pasting Team

Daisuke Nagai, Hironao Miyatake, Arya Farahi, Ken Osato, Masato Shirasaki, Han Aung, Naomi Gluck, Andrew Hearin, Tae-Hyeon Shin, Anja von der Linden, Zhuowen Zhang, Matt Ho, Michelle Ntampka, Ismael Mendoza, Phil Mansfield, Camille Avestruz, Marina Ricci, and many more

41

11.25-11.35 a.m.

Baryon Pasting Model

Goal:

Maximize the scientific returns of multiwavelength (SZ+optical+X-ray) surveys of galaxy clusters and LSS.

Challenges:

● Halo-Gas Connection: modeling of SZ and X-ray profiles from ICM to CGM.

● Baryonification: constraining baryonic effects with cross-correlations.

Solution:

Develop a simple, physically-motivated computationally efficient method for modeling multi-properties of clusters, groups, and galaxies.

Gas Clumping

Non-thermal Gas Motions

Gas Heating

by SNe/AGN feedback + mergers

Cool-Core modeling

Triaxial Shapes of DM+Gas

Polytropic Gas in Hydrostatic Equilibrium in a spherical DM potential

​

Cooling +

Star Formation

Also see Florian Kéruzoré’s JASC talk on Monday

Calibrating Baryon Pasting Model with X-ray observations

43

McDonald+13,17:

X-ray measurements of gas density profiles of the Chandra-SPT sample

Vikhlinin+06, Sun+09, Lovisari+15:

measurements of the relation between mass of gas and total mass (DM+gas+stars)

Baryon Pasting gas model describes X-ray observations: gas density profiles, Mgas-M Relations well (Flender, Nagai, McDonald 17)

Baryon Pasted Multi-wavelength Maps

• Realistic mock maps in X-ray and tSZ, e.g. using lightcones generated from large-scale DM N-body simulations.
• Fast and parallelized: ~150 halos per second per core
• Convenient to explore impact of astrophysics by varying parameters of feedback, non-thermal pressure in our BP model.

DM surface density

X-ray Surface Brightness

Compton-y (thermal SZ)

cosmoDC2 Lightcone : Korytov+2019

~440deg²

Forward Modeling SZ Cluster Physics with Baryon Pasting

Feedback and ICM non-thermal pressure (turbulence) change tSZ power spectrum at different scales (e.g., Shaw+10, Battaglia+12).

Can be easily modeled in BP maps.

​

Thermal SZ angular power spectrum

Particle-based Baryon Pasting: tSZ and kSZ maps

Osato & Nagai 23 arxiv:2201.02632

46

Particle-based baryon pasting allow us to model gas outside of halos. Important for kSZ!

(~40 times slower than halo-based painting )

Cosmology & Astrophysics with

Cross-survey & Cross-correlations

Cross angular power spectra in tSZ (CMB-S4), X-ray (eROSITA), and Weak Lensing (Rubin-LSST) will lead to improved constraints on cosmology and astrophysics

tSZ+Weak Lensing+X-ray

Shirasaki, Lau, & Nagai 2020

X-ray Angular Power Spectrum

​

• X-ray angular power spectrum from eFEDS (eROSITA early data release) consistent with ROSAT on large scales and Chandra/COSMOS on small scales.
• The BP model (Not a fit!) calibrated with Chandra-SPT density profiles agrees with measurements.
• Cycle 25 Chandra archival proposal has been accepted to constrain S₈ with X-ray power spectra.

140 sq. deg.

Lau+23 MNRAS

arxiv: 2204.13105

​

​

Scatter in ICM profiles biases power spectra

• Intrinsic scatter in gas density and pressure profiles lead to scale-dependent bias in power spectra!
• BP maps are useful for quantifying systematics in power spectrum measurements.

Xray-Xray

tSZ-tSZ

Xray-tSZ

Power spectra measured from Baryon Pasted X-ray and SZ maps

Orientation & Projection Bias in Cluster SZ-selection

• Given a mass, the halos aligned along the LOS pick up larger integrated Y.
• BP enables modeling of extrinsic scatter in cluster scaling relations due to triaxiality and projection on the map-level.

Tae-Hyeon Shin (Stony Brook)

Triaxial SZ profiles pasted to the z=0.5 snapshot of Erebos N-body sim by B. Diemer.

​

Auto-differentiable Baryon Pasting: Diffgas

• Baryon Pasting is being upgraded with auto-differentiation and GPU acceleration.
• Better models of halo Mass Accretion History dependent processes: e.g., non-thermal pressure.
• Combining with auto-differentiable galaxy modelling (A. Hearin+) for modeling cross-correlation with galaxies.

51

Naomi Gluck (Yale)

Non-thermal pressure fraction profile with Diffgas

Summary

• Multiwavelength cross-correlations of clusters and groups will provide competitive constraints on cosmology and astrophysics.
• Baryon Pasting is a power tool for forward modeling Halo-Gas connection for cross-survey cross-correlation cosmology
• Slack: @Erwin Lau

Baryon Pasting Gas Model

Polytropic equation of state in cluster cores and outskirts (Ostriker+05; Shaw+10, Flender+17)

​

​

​

Star formation : stellar mass fraction (e.g., Giodini+09, Leauthaud+11, Budzynski+13)

​

​

Dynamical heating from DM and energy feedback from AGN and SNe

​

​

​

Model of merger-induced non-thermal pressure fraction (Nelson+14, Lau+09,13, Green+20)

A physically-motivated parameterized model of gas in DM halos:

​

54

Projected-fields kSZ

Aleksandra (Ola) Kusiak

Columbia University

55

11.35-11.40 a.m.

(remote)

Work with Colin Hill, Boris Bolliet, Shivam Pandey, Will Coulton, Fiona McCarthy, Kristen Surrao, Alex Krolewski, Simone Ferraro & many others; ACT and DES teams

Projected-fields kSZ²-LSS estimator

Main idea: foreground-cleaned blackbody CMB temperature map contains kSZ information

kSZ signal traces the overall mass distribution, and thus can be detected by cross-correlating it with any large-scale structure (LSS) field, e.g. galaxies, galaxy/CMB lensing

But <kSZ x LSS> vanishes! (electron velocity)

Projected-field kSZ²-LSS:

• Construct a clean T map & apply Wiener filter
• Cross-correlate with projected (2D) LSS tracer
• But <T x LSS> vanishes!
• Solution: measure <T² x LSS>

No accurate redshift estimates needed!

kSZ in comparison with primary CMB and noise (from Bolliet et al. (2022))

Doré et al. (2004); DeDeo et al. (2005); Hill et al.(2016), Ferraro et al. (2016), Kusiak et al. (2021), Bolliet et al. (2022), Patki et al. (2023)

kSZ²-LSS estimator

kSZ-induced temperature shift in the CMB:

projected galaxy overdensity:

W_g(𝜂) =b_g(𝜂) *p(𝜂) - projection kernel

Redshift distribution of LSS tracer

Gas density profile

Triangle power spectrum:

2

x

See Raagini Patki’s work on including all terms in the bispectrum

Doré et al. (2004); DeDeo et al. (2005); Hill et al.(2016), Ferraro et al. (2016), Kusiak et al. (2021), Bolliet et al. (2022), Patki et al. (2023)

What can we get from the kSZ?

​

​

​

​

​

​

Caution: <T_CMB² x LSS> receives important contribution from CMB lensing that must also be accounted for (Hill+2016, Ferraro+2016)

Large scale limit: baryon abundance can be constrained!

​

Halo model: shape of gas density profile!

Upcoming CMB experiments!

Doré et al. (2004); DeDeo et al. (2005); Hill et al.(2016), Ferraro et al. (2016) Kusiak et al. (2021), Bolliet et al. (2022), Patki et al. (2023)

Projected-fields kSZ with Planck x unWISE

​

Projected-fields kSZ with unWISE and Planck

CMB:

• LGMCA map (tSZ-deprojected)
• Planck SMICA map

unWISE catalog (Krolewski et al. 2020):

• Based on WISE and NEOWISE
• Over 500 million galaxies on the full sky
• 3 subsamples: blue (z=0.6), green (z=1.1), and red (z=1.5)

New aspects of the analysis:

• Included the magnification bias contributions
• Asymmetric quadratic estimator (multiplying two differently-cleaned CMB maps instead of squaring one map) to increase S/N
• (LGMCA*SMICA) x unWISE, instead of (LGMCA²) x unWISE
• New ell-dependent CIB cleaning method (using <kSZxg>=0) and extensive testing for foreground contamination
• Validating with different map combinations

unWISE blue

Kusiak et al. (2021)

Overall S/N ~5.5

+Red (highest redshift kSZ detection)

Blue (z~0.6): (f_b/0.158) (f_free/1.0) = 0.65 +/- 0.24�Green (z~1.1): (f_b/0.158) (f_free/1.0) = 2.24 +/- 0.23�Red (z~1.5): (f_b/0.158) (f_free/1.0) = 2.87 +/- 0.56

No missing baryons!

Projected-fields kSZ with unWISE and Planck

Kusiak et al. (2021)

Measuring SZ with unWISE and newest ACT DR6 data

​

Projected-fields kSZ for unWISE and ACT DR6

ACT DR6 Component-separated T maps from Coulton et al. 2023

​

Measurement is lensing dominated

Data has CIB, but not tSZ (“the asymmetric method’’)

• CIB cleaning using the fact that <kSZxg>=0

Theory curve is not a fit!

Analyze in the halo model (Bolliet et al. 2022) using unWISE HOD (Kusiak et al. 2022) with class-sz to constrain the gas density profile

​

Prospects for projected-fields with S4

CMB-S4 – huge S/N, clearly distinguish between different gas density profiles, measure kSZ² “internally” with CMB lensing

-> Turn into tight constraints on ICM thermodynamics (galaxy evolution, feedback, etc)

Bolliet et al. (2022)

Backup slides

Looking ahead: Halo model kSZ² x LSS

• Halo model hybrid bispectrum

X - LSS tracer

Implemented in class-sz (Bolliet)

Fourier transform of the gas density profile/tracer

Liner matter power spectrum

1 halo

2 halo

3 halo

kSZ² halo model forecasts

Forecasts:

​

• AdvACT, SO and CMB-S4 temperature maps
• Galaxy density (unWISE)
• Galaxy lensing (DES, VRO, Euclid)
• CMB lensing (SO, CMB-S4)

​

kSZ²-unWISE forecasts

AdvACT: SNR ~ 17

SO: SNR ~ 61

For three gas density profiles:

• NFW
• Adiabatic
• AGN

We have constrained unWISE HOD (2203.12583)!

Bolliet et al. (2022)

kSZ Tomography

68

11.40-11.50 a.m.

(remote)

Kinetic Sunyaev Zel’dovich Effect

69

Velocity Reconstruction

T-g-g Bispectrum and Cosmology

• Correlating the reconstruction with LSS gives equivalent information to the kSZ-galaxy-galaxy bispectrum (see 1810.13423).

Cosmology:

• Primordial dipole: potential first direct measurement (1707.08129, 1806.01290).
• PNG: S4 x LSST could yield f_NL < 1 (1810.13424).
• CMB anomalies: S4 x LSST could confirm power asymmetry (1904.10981).
• Modified gravity: possible 2 x better constraints than T or g alone (1906.04208).
• Baryon-DM Isocurvature: possible 100 x better constraints (1908.08953, 2208.02829).

​

X

=

X

X

Realizing Cosmology Constraints

• Validation of several analysis pipelines based on quadratic estimators and maximum likelihood estimators including a variety of systematics.

​

• QE pipeline: ReCCO
• MaxL pipelines: 2205.15779, 2305.08903
• Nbody: 1806.01290, 2010.07193

https://github.com/jcayuso/ReCCO

Next step is data!

Planck x unWISE (preliminary)

• CMB: Planck SMICA
• Galaxies: unWISE Blue Sample - 80 million galaxies (credit: Krolewski).
• Mask: 58% sky cut, including galactic plane, point sources, unWISE cuts.

​

W/ Richard Bloch

Planck x unWISE (preliminary)

• Quadratic estimator:

​

[Mpc^-1]

W/ Richard Bloch

Planck x unWISE (preliminary)

+noise

[v²/c²]

Systematics: redshift distribution, optical depth (synthetic CMB)

CMB sample variance + Systematics (synthetic CMB)

W/ Richard Bloch

Planck x unWISE (preliminary)

• Evidence this may be correlation between dust residuals in the CMB and unWISE galaxies - still investigating.

​

+noise

[v²/c²]

Systematics: redshift distribution, optical depth (synthetic CMB)

CMB sample variance + Systematics (synthetic CMB)

W/ Richard Bloch

Conclusions

• Primordial dipole (dipole in frame of CMB without aberration) is highly correlated with the L=1 component of the reconstruction.
• Even with systematics, this result provides best constraint on size of the primordial dipole - of order the size of the quadrupole.
• Final results pending investigation of foreground residuals.

​

• Reconstruction noise level for S4 is ~5-10 x lower.
• More redshift information available from other galaxy surveys, different systematics, many other avenues to investigate.
• Future looks promising!

​

Secondary CMB for BSM

Not Blackbody!

Couples to monopole!

Small-scale power!

2307.15124

W/ Dalila Pirvu + Junwu Huang

79

Combination of S4 clusters and LSST galaxies

​

Chun-Hao To (OSU, CCAPP)

80

11.50-11.58 a.m.

(remote)

Collaborator: Elisabeth Krause (U of Arizona)

Synergies between LSST and S4

• Overlap region: fsky~=0.3�
• LSST Y10
• Bright galaxies with good photoz �→ lens galaxies �
• Galaxies with good shape measurement �→ source galaxies ��
• CMB-S4: �~40k clusters with S/N>5 and z<1.5�

Nvist by LSST

Expected footprint for S4 wide (Chilean site)

Using information from all 3 different tracers of LSS

• Analysis program: �To jointly analyze all 2pt functions from clusters, lens galaxies, and shears.
• Testing CDM on different environments.
• Different tracers suffer from different systematics. �→ Combination of probes gives the best constraints
• Have been applied on real data (DES) �→ Two ongoing studies in DES Y3/Y6 and ACT will be pathfinder studies for this analysis in CMBS4 era. �
• What cosmological information do we expect to get in the CMBS4 era?

To, Krause+ 2021

Forecast

• A minimal analysis:
• Only using galaxies, clusters, and shears in the overlapping area and redshifts.
• Only analyze two-point correlation functions and cluster abundances.
• Consider all expected systematics in galaxy surveys:�Photo-z uncertainties, shear calibration biases, and intrinsic alignments.
• Imposing scale cuts to remove nonlinear scales.�
• Simulated likelihood analyses:
• Non-Gaussian posteriors → Fisher forecast does not work.
• Generate simulated data and analyze it using MCMC technique�→ 48 parameters : Sampling takes two months with 128 cpus

Neural network based sampler

LINNA (2203.05583):

• Automatically build theory emulators and sample posteriers. �
• Two months → couple of hours. �
• Reaching LSST Y10 accuracy for this analysis.

To+ 2022

Analysis Setup

• Setup:
• Code implemented in CosmoLike
• tSZ cluster samples: S/N>5 and z=0.2-1.5
• Scale cut:
• Most of the two-point correlations:�
• Cosmic shear: same as DESC SRD
• Cluster lensing : 0.2
• 40 Nuisance parameters:
• 10 lens galaxy linear bias
• 10 Lens photo-z bias/ 5 Source photo-z bias
• 5 Multiplicative bias
• 2 Intrinsic alignment (NLA model)
• 8 cluster mass-observable relation.
• Cosmological parameters:
• w0-wa-nu CDM model (8 parameters)

Samples and analyses:

• S/N>5 clusters based on �https://arxiv.org/pdf/2112.07656.pdf �
• LSST galaxies and shears based on DESC-SRD. �
• Three analyses:

​

S4 clusters

Cluster + 3x2pt

Cluster focused

Galaxy focused

Forecast Results

• Clusters and LSST 3x2pt yields comparable constraints on neutrino mass and dark energy equation of states.�
• Their combination lead to �

Does the gain come from increased S/N?

• FOM with 3 parameters scales as S/N^3. �
• More data → better constraints. �
• Gains in constraining power from clusters mostly comes from degeneracy breaking.

Cluster + 3x2pt

Conclusions:

• We consider conservative analyses of LSST galaxies, shears, and CMB-S4 clusters. �
• Using newly developed likelihood inferences tool, we forecast the constraining power of these analyses with simulated likelihood analysis. �
• Combining S4 cluster and LSST 3x2pt analyses improves constraints on dark energy equation of states relative to each individual probes. �
• This combination leads to competitive constraints on neutrino mass�[𝞼(∑m_𝞶) = 26 meV].

​

90

Other topics for the parallel session

Boris Bolliet & Srinivasan Raghunathan

91

• Relativistic SZ
• Joint ymap + cluster counts
• kSZ 2pt and 4pt
• Rotational kSZ
• Ultra high-z clusters/proto clusters
• Compton-ymap cross-correlations
• +++