LFI frequency maps: data analysis, results and future challenges Davide Maino Università degli Studi di Milano, Dip. di Fisica New Light in Cosmology from the CMB 22 July - 2 August 2013, Trieste Davide Maino LFI frequency maps: data analysis, results and future challenges 1/42
Outline 1 Planck Maps 2 Data Processing Overview 3 Internal Validation 4 Scientific Analysis 5 Conclusions & Perspectives Davide Maino LFI frequency maps: data analysis, results and future challenges 2/42
Planck-LFI 30 GHz Davide Maino LFI frequency maps: data analysis, results and future challenges 3/42
Planck-LFI 44 GHz Davide Maino LFI frequency maps: data analysis, results and future challenges 4/42
Planck-LFI 70 GHz Davide Maino LFI frequency maps: data analysis, results and future challenges 5/42
Operations Planck was launched on 14 May 2009 and the current data release includes temperature maps based on the first 15.5 months of observations Planck is still alive and LFI is currently taking data HFI was switched off on January 2012, beyond the nominal period and now is supporting LFI providing essential H/K telemetry (e.g. 4K cooler) LFI is expected to continue at least till September 2013 Data are of incredible quality! Davide Maino LFI frequency maps: data analysis, results and future challenges 6/42
Data Processing Overview DPC approaches data reduction with specific tasks aiming to estimate and correct instrumental systematic effects There are three main logical levels: Level 1: H/K and Science telemetry from the satellite are transformed into raw timelines and stored into dedicated databases with the associated time information Level 2: instrument information is gathered and ingested into the Instrument Model, removal of systematic effects, flag data of suspected quality, photometric calibration and creation of maps and ancillary products Level 3: more science here with component separation, power spectra estimation and extraction of cosmological parameters Each step is internally validated (with dedicated sims) and most of the DPC work is spent cross-checking internally and between the two instruments Davide Maino LFI frequency maps: data analysis, results and future challenges 7/42
From raw data to maps Raw data from L1 AD/C & 1Hz spikes L1 raw TOI diode Focal plane geometry Planck velocity Attitude history file AD/C Non linearity correction & 1 Hz Spikes Removal Differentiated TOI Detector pointing Computation Check, Fill Gaps & flag Subsampled data Sky/load mean & GMF evaluation GMF values Pointing pipeline Detector pointing diode weights Differentiation & diode combination Differentiated TOI Detector pointing Planck velocity Galaxy mask Phase binned data Dipole Fit Dipolefit Raw gain table Iterative calibration Raw gain table Differentiated TOI Gain Application Final Gain table OSG Smoothing Calibrated TOI Detector pointing Galaxy mask Noise characteristics MapMaker (MADAM) LEVEL 2 pipeline Planck Symposium - 28 Jan 2004 Davide Maino LFI frequency maps: data analysis, results and future challenges 8/42
Beams reconstruction LFI M-Beams boresight direction in Planck FOV error boxes and positions magnified by a factor 100 0.1 0.05 V coordinate (in-scan) 0 Jupiter 1 Jupiter 2 Jupiter 3 Jupiter 4 Stacked J1J2 Stacked J3J4 Stacked J1J2J3J4-0.05-0.1-0.1-0.05 0 0.05 0.1 U coordinate (cross-scan) Beams contours (-20dB/LFI) from planet transit Pointing reconstruction: error smaller that 0.5 Davide Maino LFI frequency maps: data analysis, results and future challenges 9/42
Effective Beams Band FWHM [ ] e Ω [arcmin 2 ] 30 32.239±0.013 1.320±0.031 1189.51±0.84 44 27.01±0.55 1.034±0.031 833±32 70 13.252±0.033 1.223±0.026 200.7±1.0 Scanning beams are used to compute effective beams which gives pixel-by-pixel the beam shape projected in the sky when scanning strategy is accounted for. Fundamental point for source extractions and beam window function Davide Maino LFI frequency maps: data analysis, results and future challenges 10/42
Photometric Calibration This is the most important and critical aspect of data reduction pipeline. The LFI one works as follows estimate dipole amplitude and alignment of dipole for any direction create the expected level of the dipole signal accounting for the full-beam shape fit real data with dipole with iterative approach (raw gains) filter raw gains or use 4K to calibrate data Davide Maino LFI frequency maps: data analysis, results and future challenges 11/42
Photometric Calibration Davide Maino LFI frequency maps: data analysis, results and future challenges 12/42
Map-making Calibrated TOI for each radiometer are input of madam map-making code, together with pointing data The algorithm is a maximum-likelihood destriping and estimates in this fashion the amplitude of the 1/f -noise baseline, subtract from the timelines and then simply bins the resulting TOI into a map Maps produced at different levels: Frequency maps, HR maps and Survey maps Low resolution maps used for the computation of the noise covariance matrix Davide Maino LFI frequency maps: data analysis, results and future challenges 13/42
Systematic effects Null tests: primary tool to see systematic effect residual w.r.t. white noise level Sims: Assess their impact on TOI using in-flight H/K data. This approach provides a powerful tool to check for systematics Davide Maino LFI frequency maps: data analysis, results and future challenges 14/42
44 GHz Davide Maino LFI frequency maps: data analysis, results and future challenges 15/42
70 GHz Davide Maino LFI frequency maps: data analysis, results and future challenges 16/42
LFI Internal Validation - Null tests C l [µk 2 ] 10-2 10-1 10 0 TT SS TT jk EE SS EE jk C l [µk 2 ] 10-2 10-1 10 0 TT SS TT jk EE SS EE jk C l [µk 2 ] 10-2 10-1 10 0 TT SS TT jk EE SS EE jk 10 1 10 2 10 3 l 10 1 10 2 10 3 l 10 1 10 2 10 3 Quality of LFI maps is assessed and verified by a set of null-tests in an almost automatic way Several data combination (radiometer, horn-pairs, frequency) on different time-scales (1 hour, survey, full-mission): difference at horn level at even/odd surveys clearly reveals side lobe effect Null-test power spectra are used to check total level of system. effects to be compared w.r.t. white noise level and systematic effects analysis l Davide Maino LFI frequency maps: data analysis, results and future challenges 17/42
LFI Internal Validation - Check Spectra l(l +1)C l 44 /2π [ µk 2] 0 2000 4000 6000 70 vs 44GHz α =1.006 ± 0.017 l =[2, 400] 4000 5000 0 1000 2000 3000 6000 l(l +1)C 70 l /2π [ µk 2] l(l +1)C l 30 /2π [ µk 2] 0 2000 4000 6000 70 vs 30GHz α =0.994 ± 0.022 l = [50, 400] 4000 5000 0 1000 2000 3000 6000 l(l +1)C 70 l /2π [ µk 2] l(l +1)C l 30 /2π [ µk 2] 0 2000 4000 6000 44 vs 30GHz α =0.979 ± 0.027 l = [50, 400] 4000 5000 0 1000 2000 3000 6000 l(l +1)C 44 l /2π [ µk 2] Compute power spectra in multipole range around the first acoustic peak removing the unresolved point source contributions. Spectra are consistent within errors. 30 and 44/70 have different approaches to gain applied Hausman test assess consistency at 70 GHz showing no statistically significant problem Spectra from horn-pairs and from all 12 radiometers: χ 2 analysis shows compatibility with null-hypothesis Davide Maino LFI frequency maps: data analysis, results and future challenges 18/42
Maps Characteristics Uncertainty Applies to Method Gain calib All sky WMAP dipole Zero level All sky Galactic Cosecant model Beam All sky GRASP models via FeBecop CC non-cmb ground/flight bandpass leakage Resid. Sys All sky Null-tests Property 30 44 70 Frequency [GHz] 28.4 44.1 70.4 Noise rms/pixel [µk CMB ] 9.2 12.5 23.2 Gain Uncert 0.82% 0.55% 0.62% Zero Level Uncert [µk CMB ] ±2.23 ±0.78 ±0.64 Davide Maino LFI frequency maps: data analysis, results and future challenges 19/42
Component Separation Davide Maino LFI frequency maps: data analysis, results and future challenges 20/42
Component Separation Microwave emission is dominated by CMB from 70 to 143 GHz At low and high frequency foregrounds emissions dominates not only along the galactic plane: synchrotron, free-free, dust Forerground components have distinct spectral shapes different from CMB Multi-frequency observations are fundamental to separate such components Davide Maino LFI frequency maps: data analysis, results and future challenges 21/42
Planck CMB sky Davide Maino LFI frequency maps: data analysis, results and future challenges 22/42
Planck vs WMAP Davide Maino LFI frequency maps: data analysis, results and future challenges 23/42
From maps to Power Spectrum Given a map of CMB sky T (ˆn) decompose in spherical harmonics a lm = d ˆn T (ˆn) Ylm (ˆn) Without instrumental noise power spectrum is given by C l = 1 2l + 1 l m= l a lm 2 For real experiments, with finite beam size, noise and incomplete sky coverage C l = l M ll F l B 2 l Cth l + N l Davide Maino LFI frequency maps: data analysis, results and future challenges 24/42
Planck TT Angular Power Spectrum Davide Maino LFI frequency maps: data analysis, results and future challenges 25/42
Planck Cosmological Parameters Planck 2013 results. XVI. Cosmological Parameters Shape, positions and heights of peaks strongly depends on cosmological parameters Instead of comparing theory (C th l ) with data on defined grid points in parameter space, a Monte Carlo Marchov Chain approach is used: random walk with Metropolis-Hasting chain samples Davide Maino LFI frequency maps: data analysis, results and future challenges 26/42
Planck Cosmological Parameters Davide Maino LFI frequency maps: data analysis, results and future challenges 27/42
Hubble Constant H 0 H 0 = (67.3 ± 1.2)km s 1 Mpc 1. Lower than previous estimation based on astrophysical observations. Higher age of the Universe 13.817 ± 0.048 Gyr Davide Maino LFI frequency maps: data analysis, results and future challenges 28/42
Ω m P/Pmax 0.0 0.2 0.4 0.6 0.8 1.0 Planck+WP+highL SNLS combined SNLS SiFTO SNLS SALT2 Union 2.1 0.1 0.2 0.3 0.4 Ω m SN data are almost consistent within each other Union 2.1 is consistent with Planck and SNLS combined shows some tension Residual systematics? Davide Maino LFI frequency maps: data analysis, results and future challenges 29/42
Dark Energy 1.0 0.8 Planck+WP+BAO Planck+WP+Union2.1 Planck+WP+SNLS Planck+WP 1.6 0.8 P/Pmax 0.6 0.4 0.2 0.0 2.0 1.6 1.2 0.8 0.4 w wa 0.0 0.8 Planck+WP+BAO Planck+WP+Union2.1 Planck+WP+SNLS 1.6 2.0 1.6 1.2 0.8 0.4 w 0 w = p/ρ = 1 cosmological constant. All data-sets consistent with this but the inclusion of SNLS w(a) = w 0 + w a (1 a): the (w 0, w a ) = ( 1, 0) is within 68% CL but for the inclusion of SNLS Davide Maino LFI frequency maps: data analysis, results and future challenges 30/42
Intriguing inconsistencies WMAP reported some anomalies on large angular scales: Quadrupole/Octupole alignment Low variance Hemispherical asymmetry Dipolar power modulation Cold Spot Planck revises these anomalies and provides new robust results Davide Maino LFI frequency maps: data analysis, results and future challenges 31/42
Quadropole/Octupole alignment Search for axis of maximum momentum dispersion independently for l = 2 and l = 3 CMB map derived from different component separation methods Alignment varies between 9 and 13 (depending on comp. sep.) less significant (below 98%) Davide Maino LFI frequency maps: data analysis, results and future challenges 32/42
Hemispherical asymmetry Analysis of WMAP data reveal power spectrum asymmetry: power in hemisphere centred in (θ, φ) = (110, 237 ) (close to ecliptic plane) is larger than in the opposite for l = 2 40 Other methods (Minkowski functionals, N point correlation functions) show similar evidence Planck reviews such evidences using different CMB component separated maps and N point correlation functions Davide Maino LFI frequency maps: data analysis, results and future challenges 33/42
Hemispherical asymmetry North: almost 99.9% of simulations has larger structure of 4-pt correlation function South: in line (54-57%) with (Gaussian) simulations Davide Maino LFI frequency maps: data analysis, results and future challenges 34/42
Low-l deficit ΛCDM is not a good fit for Planck data for 20 < l < 40 Already present in WMAP but with Planck at 2 2.5 σ level Davide Maino LFI frequency maps: data analysis, results and future challenges 35/42
Dipole Modulation Gordon et al. (2005) proposed a phenomenological model for large scale power in WMAP data described by a multiplicative dipole modulation d = (1 + Ap n)s iso + n WMAP-5yr found A = 0.072 ± 0.022 and (l, b) = (224, 22 ) at 3.3 σ level Davide Maino LFI frequency maps: data analysis, results and future challenges 36/42
Dipole Modulation Probability distribution 0 5 10 15 20 Commander NILC SEVEM SMICA 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Modulation amplitude, A Commander NILC SEVEM SMICA WMAP Dipole amplitude consistent between different CMB maps with A [0.065, 0.078] ± 0.022 Dipole direction in agreement also with WMAP with significance 2.9 3.5 σ Davide Maino LFI frequency maps: data analysis, results and future challenges 37/42
Interpretations of anomalies Many features observers by WMAP are seen by Planck: inter-instrument consistency rules out intrinsic systematic effects Microwave sky is neither Gaussian nor isotropic but component separation is robust Possible local origin: removing ISW originated with a volume at z < 0.3 alleviates most of the anomalies Most interesting would be their cosmological origin: candidates models with compact topology although no evidence is shown anisotropic Bianchi VII h model is a good template and when subtracted from the data many of the large-scale anomalies are no longer present but...... Bianchi provides parameters incompatible with ΛCDM Davide Maino LFI frequency maps: data analysis, results and future challenges 38/42
Conclusions Standard ΛCDM provides a good fit to data for l > 40 Planck provides a unique window on the early universe and shades light on content of the Universe: Ω m, Ω Λ, Ω b Possible new physics from tension on some parameters values (e.g. H 0 ) Spectrum shape 20 < l < 40 is a real feature of CMB anisotropies. Not a decisive significance but is an anomaly. Davide Maino LFI frequency maps: data analysis, results and future challenges 39/42
What s next? Polarisation data will open new possibilities: direct measure of τ, new hints on tensor modes, break some parameter degeneracies with temperature only data Foregrounds polarisation maps will provides informations on, e.g., emission mechanisms and structure of galactic magnetic field Challenging due to the tiny CMB polarisation signal but... Davide Maino LFI frequency maps: data analysis, results and future challenges 40/42
What s next? Angular power spectra of TE (left) and EE (right) from Planck data combinations. f sky = 0.4, no foreground cleaning, uniform weight of channels Davide Maino LFI frequency maps: data analysis, results and future challenges 41/42
Davide Maino LFI frequency maps: data analysis, results and future challenges 42/42