Intro to cosmology with galaxy surveys

Seminar given at Imperial College on Nov 4 2020, to an audience of astrophysicists (not necessarily cosmologists). This is astate of affairs of some aspects of observational cosmology, with a focus on a comparison between the 2010s and the 2020s.

Transcript

1. Cosmology with galaxy surveys Boris Leistedt Imperial College London, 4 November 2020
2. P Cosmology in the 2010s and 2020s Galaxy surveys Data analysis challenges Disclaimer – there are many other: 1) interesting questions in cosmology, 2) types of surveys/missions that can help answer those questions, 3) data analysis challenges. 2/58
3. O Cosmology in the 2010s and 2020s Galaxy surveys Data analysis challenges 3/58
4. S Well-established standard model... but recent data hint at discrepancies, potential new physics, and some fundamental questions still unaddressed. We need more data 4/58
5. S ΛCDM : Isotropy and homogeneity on large scales Universe geometrically flat (space-time metric) Components: photons, baryons, leptons, dark matter Gaussian anisotropies / initial conditions = quantum fluctuations stretched by initial phase of inflation Gravity governed by general relativity Cosmological constant (dark energy) explains recent accelerated expansion Very predictive and successful given number of parameters (∼ 6 − 10) and assumptions Other assumptions: three neutrinos, minimum neutrino mass, no running of primordial power spectrum, gravitational waves negligible, no topological defects, recombination understood, simple reionisation, etc 5/58
6. M ‘Early’ universe (z > 1000): Cosmic microwave background Primordial nucleosynthesis (abundances of H, He, Li) ‘Late’ universe (0 < z < 3): Clustering of galaxies, including baryon acoustic oscillations Weak lensing of galaxies Clustering of neutral hydrogen in quasar Lyman alpha forest Type Ia supernovae Clustering, abundances, & lensing of galaxy clusters Strong lensing 6/58
7. R , , Figure: astropy 7/58
8. CMB 2 10 100 500 1000 1500 2000 2500 Multipole 0 1000 2000 3000 4000 5000 6000 DT T [µK2] 2 10 100 500 1000 1500 2000 2500 Multipole −100 −50 0 50 100 DT E [µK2] 2 10 100 500 1000 1500 2000 Multipole 0.01 0.1 1 10 20 30 40 DEE [µK2] 101 102 103 Multipole L 0.0 0.5 1.0 1.5 [L(L + 1)]2/(2π) Cφφ L [107 µK2] CMB Planck data power spectra Very stringent constraints on most ΛCDM parameters Currently supported by other experiments (ACT, SPT) Figure: Planck Collaboration et al. (2018a). “Planck 2018 results. I. Overview and the cosmological legacy of Planck”. In: : 10.1051/0004-6361/201833880. eprint: arXiv:1807.06205 8/58
9. P z = 0 Clustering of matter as function of scale (fourier transform of 3D correlations) Different data sets constraint different scales, epochs, physics Cross-checks essential for looking for new physics Figure: Planck Collaboration et al. (2018a). “Planck 2018 results. I. Overview and the cosmological legacy of Planck”. In: : 10.1051/0004-6361/201833880. eprint: arXiv:1807.06205 9/58
10. 2020s improve by an order of magnitude Figure: eBOSS Collaboration et al. (2020). The Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological Implications from two Decades of Spectroscopic Surveys at the Apache Point observatory. eprint: arXiv:2007.08991 10/58
11. Any questions? 11/58
12. T 1. Data analysis is fun 2. No good theory for dark matter or dark energy 3. Multiple ΛCDM assumptions relaxed = weakened model constraints 4. Tensions between current data sets, with no viable explanation 5. Some physics/assumptions yet to be tested: curvature, modified gravity, inflationary physics, neutrino masses, etc 12/58
13. E Early and late universe data disagree on (projected) Hubble constant Many possible explanations (extensions of ΛCDM), but none better constrained or motivated theoretically Figure: L. Verde, T. Treu, and A. G. Riess (2019). “Tensions between the Early and the Late Universe”. In: : 10.1038/s41550-019-0902-0. eprint: arXiv:1907.10625 13/58
14. G 0.70 0.75 0.80 0.85 S8 ≡ σ8 √ Ωm/0.3 KiDS-1000 3 × 2pt Cosmic shear + GGL Cosmic shear + galaxy clustering Cosmic shear Galaxy clustering Planck 2018 TTTEEE+lowE BOSS+KV450 (Tr¨ oster et al. 2020) DES Y1 3 × 2pt (DES Collaboration 2018) KV450 (Hildebrandt et al. 2020) DES Y1 cosmic shear (Troxel et al. 2018) HSC pseudo-C (Hikage et al. 2019) HSC ξ± (Hamana et al. 2020) MAP + PJ-HPD CI M-HPD CI nominal σ 8 is r.m.s of matter density fluctuations within 8h−1Mpc, so mostly normalisation of P(k) at 0.1 < k < 2 Early and late universe data disagree on growth of structure Many possible explanations (extensions of ΛCDM), but none better constrained or motivated theoretically Figure: Catherine Heymans et al. (2020). KiDS-1000 Cosmology: Multi-probe weak gravitational lensing and spectroscopic galaxy clustering constraints. eprint: arXiv:2007.15632 14/58
15. C Figure: eBOSS Collaboration et al. (2020). The Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological Implications from two Decades of Spectroscopic Surveys at the Apache Point observatory. eprint: arXiv:2007.08991 15/58
16. E Many crucial questions: what energy scale? what fields involved? etc But few imprints in data, and difficult to extract Figure: Planck Collaboration et al. (2018a). “Planck 2018 results. I. Overview and the cosmological legacy of Planck”. In: : 10.1051/0004-6361/201833880. eprint: arXiv:1807.06205 16/58
17. T 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 z 0.2 0.3 0.4 0.5 0.6 0.7 0.8 f 8 6dFGS SDSS MGS SDSS LRG 6dFGS +SnIa GAMA VIPERS BOSS DR12 WiggleZ DR14 quasars FastSound Constraints on growth from a range of surveys (from anisotropic 3D clustering measurements) Much room for imprints of exotic gravity / dark energy / dark matter. Figure: Planck Collaboration et al. (2018a). “Planck 2018 results. I. Overview and the cosmological legacy of Planck”. In: : 10.1051/0004-6361/201833880. eprint: arXiv:1807.06205 17/58
18. Any questions? 18/58
19. P What cosmology question do you think is most important? 1. Looking for new properties of dark matter/dark energy 2. Measuring growth (σ 8) 3. Measuring neutrino masses 4. Testing alternatives to general relativity 5. Probing inflation / early universe physics 6. Measuring curvature 19/58
20. O Cosmology in the 2010s and 2020s Galaxy surveys Data analysis challenges 20/58
21. S Photometry and spectroscopy are complementary Some exciting missions are happening in the 2020s 21/58
22. E : SDSS- Ideal: accurate, dense survey of tracers of LSS density fluctuations Even better: trace velocities, split by galaxy type/dark matter halos, etc 22/58
23. S Measured P galaxy (clustering) and P matter (lensing) depend on: scale k redshift z line of sight angle µ galaxy bias b Sensitivity: Primordial physics : largest volumes/scales, high redshift, high bias Gravity/growth and expansion : all scales (number of modes) Neutrinos: small scales 23/58
24. G Main criteria for cosmology: Sky area & depth =⇒ volume, density, & redshift range Spectral resolution =⇒ object types & redshifts Time resolution (cadence) =⇒ variability, supernovae, etc Constrained by detector technology and observing strategy 24/58
25. P (CCD ) ( ) Figures from SDSS & DES official galleries 25/58
26. S P 4000 6000 8000 10000 λ [ ˚ A] 100 101 4000 6000 8000 10000 λ [ ˚ A] 100 101 Fictitious SDSS spectroscopy and DECALS photometry (unrealistic noise) 26/58
27. F SDSS ( ) Right: constraints on Baryon Acoustic Oscillations from galaxy clustering Figure: SDSS official gallery 27/58
28. S A L Good classifications Good redshifts Yields 3D clustering Relatively shallow Requires targeting (from photometry) Inherits some systematics from photometry 28/58
29. M P SDSS- I,II,III/BOSS 2000-14 10k deg2 700k z < 0.2 galaxies 500k 0.2 < z < 0.5 galaxies 1m z > 0.5 galaxies 100k z < 2 QSOs 200k z > 2 QSOs SDSS-IV/eBOSS 2014-18 8-10k deg2 600k 0.6 < z < 1 galaxies 500k z > 0.9 QSOs DESI 2020-25 14k deg2 17m 0.6 < z < 1.6 ELGs 6m 0.4 < z < 1.0 LRGs 2.4m 1 < z < 2.1 QSOs 10m 0.05 < z < 0.4 bright galaxies 29/58
30. P Cute but useless for cosmology Millions of those = precision cos- mology 30/58
31. T −135 −90 −45 0 45 90 135 180 −15 0 15 30 45 60 75 90 0 30 60 90 120 150 180 210 240 270 300 330 0 −15 0 15 30 45 60 75 KIDS DES Galactic Plane MzLS+BASS DECaLS DECaLS ATLAS ATLAS −15 0 15 30 45 60 75 Blue is SDSS (ugriz), used for SDSS/BOSS/eBOSS spectroscopic targeting. DES (grizY) & KiDS (ugriZYKsH) are best current photometric surveys. DeCaLS+MzLS+BASS are deeper grz imaging for DESI targeting Next generation: LSST & Euclid Figure: Arjun Dey et al. (2018). “Overview of the DESI Legacy Imaging Surveys”. In: : 10.3847/1538-3881/ab089d. eprint: arXiv:1804.08657 31/58
32. P A L Depth Uncertain redshifts and types 2+1D clustering Complex selection function P Nobj SDSS 2000-14 10k deg2 ugriz 106 galaxies DES 2012-18 5k deg2 grizY 107−8 galaxies KiDS+VIKING 2011-18 1.5k deg2 ugriZYHKs 107−8 galaxies LSST 2022-32 18k deg2 ugrizy 109 galaxies Euclid (space) 2028+ 15k deg2 VIS + YJK 109 galaxies Redshift precision depends on bands. Usable area/numbers depends on quality, uniformity, etc 32/58
33. P Smearing due to photometric redshift precision (idealized) Robust statistical techniques are critical Figure: N. Benitez et al. (2008). “Measuring Baryon Acoustic Oscillations along the line of sight with photometric redshifs: the PAU survey”. In: : 10.1088/0004-637X/691/1/241. eprint: arXiv:0807.0535 33/58
34. LSST : 0 1 2 3 4 z 0 5 10 15 20 dN/dΩ dz [arcmin−2] LSST galaxies WκCMB Very idealised. Figure: Emmanuel Schaan, Simone Ferraro, and Uroˇ s Seljak (2020). Photo-z outlier self-calibration in weak lensing surveys. eprint: arXiv:2007.12795 34/58
35. LSST : D 100 1000 10−6 10−5 Cgκg Galaxy × galaxy lensing g0 κgj g1 κgj g2 κgj g3 κgj g4 κgj g5 κgj g6 κgj g7 κgj g8 κgj g9 κgj 10−5 0.001 Cgi gi Clustering i = 0 i = 1 i = 2 i = 3 10−4 0.0002 0.0003 Cgi gi+1 i = 4 i = 5 i = 6 i = 7 100 1000 10−4 6 × 10−5 Cgi gi+2 i = 8 i = 9 100 1000 10−8 10−7 10−6 10−5 Cκg κg Shear tomography κgi κgi+0 κgi κgi+1 κgi κgi+2 κgi κgi+3 κgi κgi+4 κgi κgi+5 κgi κgi+6 κgi κgi+7 κgi κgi+8 κgi κgi+9 100 1000 2 × 10−5 3 × 10−5 4 × 10−5 CgκCMB Galaxy × CMB lensing κCMB g0 κCMB g1 κCMB g2 κCMB g3 κCMB g4 κCMB g5 κCMB g6 κCMB g7 κCMB g8 κCMB g9 + a few others! Tradeoff between redshift accuracy, density of objects, etc Figure: Emmanuel Schaan, Simone Ferraro, and Uroˇ s Seljak (2020). Photo-z outlier self-calibration in weak lensing surveys. eprint: arXiv:2007.12795 35/58
36. P Which is best: photometric or spectroscopic surveys? Science cases: 1. Primordial physics 2. Modified gravity / growth 3. Dark energy 4. Geometry / curvature Criteria: volume (sky area × depth), scales involved, number density of objects, redshift accuracy, etc 36/58
37. E DESI Figure: Mariana Vargas-Magana et al. (2019). Unraveling the Universe with DESI. eprint: arXiv:1901.01581 37/58
38. P LSST Galaxy surveys constrain regimes unattainable by CMB 10−3 10−2 10−1 100 101 k [Mpc−1] 1.6 1.8 2.0 2.2 2.4 P(k) ×10−9 HZ spectrum Inflationary spectrum Linear feature Logarithmic feature Localized feature 101 102 103 ωlin [Mpc] 10−4 10−3 10−2 10−1 σ(Alin) BOSS DESI Euclid Future LSS-CVL Planck CMB-S3 CMB-S4 CMB-CVL Non-Gaussianity: LSST+DESI could reach σ(f NL,local ) ∼ 0.1 − 1 Neutrino masses (and hierarchy): σ(Mν ) ∼ 10−2 Figure: Anˇ ze Slosar et al. (2019). Scratches from the Past: Inflationary Archaeology through Features in the Power Spectrum of Primordial Fluctuations. eprint: arXiv:1903.09883 38/58
39. Any questions? 39/58
40. O Cosmology in the 2010s and 2020s Galaxy surveys Data analysis challenges 40/58
41. S Galaxy surveys face two major unsolved problems: Mitigating spatially varying selection function in both spectroscopic and photometric surveys Redshift distributions of photometric surveys New methodologies required. 41/58
42. S - Sophisticated statistical methods needed to achieve science goals 42/58
43. S & Non-uniformity cannot be avoided. Complex effect on data. Critical for correlation studies (= most of observational cosmology) Pre-KiDS/DES: could be ignored with minor data cuts! Unavoidable survey non-uniformities (DES SV data) Figure: B. Leistedt et al. (2015). “Mapping and simulating systematics due to spatially-varying observing conditions in DES Science Verification data”. In: : 10.3847/0067-0049/226/2/24. eprint: arXiv:1507.05647 43/58
44. LSST PentagonDitherPerSeason 27.35 27.40 27.45 27.50 27.55 r-band Coadded Depth Best scenario, with idealised observing strategy optimized for uniformity! Figure: Humna Awan et al. (2016). “Testing LSST Dither Strategies for Survey Uniformity and Large-Scale Structure Systematics”. In: : 10.3847/0004-637X/829/1/50. eprint: arXiv:1605.00555 44/58
45. M Currently: correct data maps or power spectra (ad-hoc models) −20 0 20 40 DEC [deg] Galaxy Density Map Corrected Galaxy Density Map 0 100 200 −20 0 20 40 DEC [deg] NN Selection Mask 0 100 200 RA [deg] Quadratic Selection Mask 0 100 200 Linear Selection Mask 0.9 1.0 1.1 n/n DECALS data: model based on observing conditions maps, learned from correlations with density field. Figure: Mehdi Rezaie et al. (2019). “Improving Galaxy Clustering Measurements with Deep Learning: analysis of the DECaLS DR7 data”. In: : 10.1093/mnras/staa1231. eprint: arXiv:1907.11355 45/58
46. L Anomalous eBOSS ELG density reproduced with injections Model select fuction directly by Injecting simulated galaxies in real images Running the real data pipeline Measuring properties and detection of simulated objects. Figure: Hui Kong et al. (2020). “Removing Imaging Systematics from Galaxy Clustering Measurements with Obiwan : Application to the SDSS-IV extended Baryon Oscillation Spectroscopic Survey Emission Line Galaxy Sample”. In: : 10.1093/mnras/staa2742. eprint: arXiv:2007.08992 46/58
47. B ... Injection technique limited by 1. realism of simulations 2. computational load of repeated injection-extraction process (+inability to run on single-epochs) New approach: learn result of injections with machine learning then emulate selection function over whole survey, without extra new injection or pipeline runs. Can also reduce impact of simulation realism. Figure: Hui Kong et al. (2020). “Removing Imaging Systematics from Galaxy Clustering Measurements with Obiwan : Application to the SDSS-IV extended Baryon Oscillation Spectroscopic Survey Emission Line Galaxy Sample”. In: : 10.1093/mnras/staa2742. eprint: arXiv:2007.08992 47/58
48. Any questions? 48/58
49. W ? Galaxies at different redshifts can look very similar in photometry λ[cm−8] Arbitrary Flux 103 104 Lyman Break Balmer Break Balmer Break @ z=1.1 Sy1.8 QSO SB Sc Ell a) 1 2 3 4 −0.5 0.0 0.5 1.0 1.5 redshift Observed i−z [AB] Sy1.8 QSO SB SB+eml. SB+ext. Sc Ell b) Figure: Mara Salvato, Olivier Ilbert, and Ben Hoyle (2018). The many flavours of photometric redshifts. eprint: arXiv:1805.12574 49/58
50. D -z Due to different SEDs, priors, treatment of errors. Figure: S. J. Schmidt et al. (2020). Evaluation of probabilistic photometric redshift estimation approaches for LSST. eprint: arXiv:2001.03621 50/58
51. R Three methods: 1) from sub-sample of galaxies with spectroscopic redshifts Issues: spectroscopic redshifts not 100% accurate or representative 2) from photometric redshift PDFs. Issues: SEDs/priors/data errors biased 3) spatial cross-correlations. Issues: ignores photometry, limited redshift range + problem of getting uncertainties on redshift distributions 51/58
52. U S S D Aleatoric uncertainties Data biases M Epistemic uncertainties Model biases Hierarchical models to the rescue! 52/58
53. H First implementation with existing data β α ɣ Object properties: z, t, m, … Model photometry Observed photometry Model SEDs (corrected) Base SED templates Linear SED corrections Prior distributions p(z, t, m, …) Photometric 
 zeropoint/error biases Nobj Ntemplates New approach in development with physical SED models & priors Goal: exploit full statistical power of photometric surveys to constrain galaxy formation + N(z) for cosmology Figure: Boris Leistedt et al. (2018). “Hierarchical modeling and statistical calibration for photometric redshifts”. In: : 10.3847/1538-4357/ab2d29. eprint: arXiv:1807.01391 53/58
54. C Galaxy surveys: great cosmological constraints, but many data analysis challenges. Exciting times for astrostatistics: large data sets, new modeling/inference methods, good software 54/58
55. Any questions? 55/58
56. R I Awan, Humna et al. (2016). “Testing LSST Dither Strategies for Survey Uniformity and Large-Scale Structure Systematics”. In: : 10.3847/0004-637X/829/1/50. eprint: arXiv:1605.00555. Benitez, N. et al. (2008). “Measuring Baryon Acoustic Oscillations along the line of sight with photometric redshifs: the PAU survey”. In: : 10.1088/0004-637X/691/1/241. eprint: arXiv:0807.0535. Collaboration, eBOSS et al. (2020). The Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological Implications from two Decades of Spectroscopic Surveys at the Apache Point observatory. eprint: arXiv:2007.08991. Collaboration, Planck et al. (2018a). “Planck 2018 results. I. Overview and the cosmological legacy of Planck”. In: : 10.1051/0004-6361/201833880. eprint: arXiv:1807.06205. Collaboration, Planck et al. (2018b). “Planck 2018 results. VI. Cosmological parameters”. In: : 10.1051/0004-6361/201833910. eprint: arXiv:1807.06209. 56/58
57. R II Dey, Arjun et al. (2018). “Overview of the DESI Legacy Imaging Surveys”. In: : 10.3847/1538-3881/ab089d. eprint: arXiv:1804.08657. Heymans, Catherine et al. (2020). KiDS-1000 Cosmology: Multi-probe weak gravitational lensing and spectroscopic galaxy clustering constraints. eprint: arXiv:2007.15632. Kong, Hui et al. (2020). “Removing Imaging Systematics from Galaxy Clustering Measurements with Obiwan : Application to the SDSS-IV extended Baryon Oscillation Spectroscopic Survey Emission Line Galaxy Sample”. In: : 10.1093/mnras/staa2742. eprint: arXiv:2007.08992. Leistedt, B. et al. (2015). “Mapping and simulating systematics due to spatially-varying observing conditions in DES Science Verification data”. In: : 10.3847/0067-0049/226/2/24. eprint: arXiv:1507.05647. Leistedt, Boris et al. (2018). “Hierarchical modeling and statistical calibration for photometric redshifts”. In: : 10.3847/1538-4357/ab2d29. eprint: arXiv:1807.01391. 57/58
58. R III Rezaie, Mehdi et al. (2019). “Improving Galaxy Clustering Measurements with Deep Learning: analysis of the DECaLS DR7 data”. In: : 10.1093/mnras/staa1231. eprint: arXiv:1907.11355. Salvato, Mara, Olivier Ilbert, and Ben Hoyle (2018). The many flavours of photometric redshifts. eprint: arXiv:1805.12574. Schaan, Emmanuel, Simone Ferraro, and Uroˇ s Seljak (2020). Photo-z outlier self-calibration in weak lensing surveys. eprint: arXiv:2007.12795. Schmidt, S. J. et al. (2020). Evaluation of probabilistic photometric redshift estimation approaches for LSST. eprint: arXiv:2001.03621. Slosar, Anˇ ze et al. (2019). Scratches from the Past: Inflationary Archaeology through Features in the Power Spectrum of Primordial Fluctuations. eprint: arXiv:1903.09883. Vargas-Magana, Mariana et al. (2019). Unraveling the Universe with DESI. eprint: arXiv:1901.01581. 58/58
59. R IV Verde, L., T. Treu, and A. G. Riess (2019). “Tensions between the Early and the Late Universe”. In: : 10.1038/s41550-019-0902-0. eprint: arXiv:1907.10625. 59/58