Blind Search for Variability in Planck Data
AAstronomy in Focus, Volume 1XXIXth IAU General Assembly, August 2015Piero Benvenuti, ed. c (cid:13) Blind Search for Variability in Planck Data
J¨org P. Rachen , Elina Keih¨anen and Martin Reinecke Department of Astrophysics/IMAPP, Radboud University Nijmegen, The Netherlands Department of Physics, University of Helsinki, Finland MPI for Astrophysics, 85741 Garching, Germany
Abstract.
The sky is full of variable and transient sources on all time scales, from milliseconds todecades.
Planck ’s regular scanning strategy makes it an ideal instrument to search for variablesky signals in the millimetre and submillimetre regime, on time scales from hours to severalyears. A precondition is that instrumental noise and systematic effects, caused in particular bynon-symmetric beam shapes, are properly removed. We present a method to perform a full skyblind search for variable and transient objects at all Planck frequencies.
Keywords. space vehicles: instruments, methods: statistical, techniques: miscellaneous
1. Introduction
Planck (Tauber et al. 2010, Planck Collaboration I. 2011) provides 8 full sky surveysat the frequencies of its LFI instrument (30, 44, and 70 GHz), and 5 full surveys at the6 HFI frequencies in the range 100–857 GHz. The satellite rotates with 1 rpm around anaxis kept fixed for a pointing period (PID) of about 50 minutes, then shifts along theecliptic by 2’. This keeps a point source near the ecliptic in the main beam (2 FWHM) forabout 5 −
30 PIDs (depending on frequency) for each survey; for sources near the eclipticpoles the coverage can be much larger. Analysing the time information in Planck datafor a particular sky direction thus allows to search for variability in the sky signal.
2. Mapping of time ordered information
Variability mapping is based on four-dimenional Healpix (G´orski et al. 2005) constructscalled 4D-maps, which record for every sky pixel k all contributions of a given detectorat times t j and beam orientation b j , where the index j refers to Planck PIDs. To con-struct an average sky signal S k free from beam orientation effects, we use the ArtDeco beam deconvolution code (Keih¨anen & Reinecke 2012). A variability map is then a two-dimensional Healpix map of a quantity X k = 12 sgn ( ˆ χ − ) N k (cid:104) ˆ χ − − ln ˆ χ (cid:105) where ˆ χ k ≡ χ k N k = 1 N k (cid:88) j ( I kj − ( S ∗ b ) kj ) σ k with σ kj = σ n + ξ ( S ∗ b ) kj .N k is the number of entries for pixel k in the 4D signal map I kj , ( S ∗ b ) kj is the beam-reconvolved average 4D-map based on the ArtDeco map S k , σ n is the detector white noise,and ξ combines all instrumental fluctuations which factor on the signal (e.g., calibration,inaccuracies of the beam model). The definition of X is motivated by the Chernoff boundon the CDF P N ( χ ), with X (cid:54) − ln(1 − P N ( χ )) for ˆ χ (cid:62)
1, and
X > ln P N ( χ ) for1 a r X i v : . [ a s t r o - ph . I M ] D ec J.P. Rachen, E. Keihanen & M. Reinecke
Figure 1.
Histogram of an X-map for the LFI 44 GHz detectors for ξ = 0 .
01, and detectornoise n s as given in Planck Collaboration II. (2014), thin dotted lines show the distributionexpected for pure Gaussian noise. The tail to large values of X indicates the presence of truesky variability. A tail to large negative values of X would indicate an overestimation of ξ , anoffset of the peak from X = 0 an incorrect estimation of n s . ˆ χ <
1. The distribution of X-values over a sky map is indicateive not only of variability,but also to incorrect estimations of noise and instrumental variations (see Fig. 1).
3. Time resolved Planck fluxes, status and outlook
As the analysis of time variations in the sky is essentially background-free, our methodprovides a way to extract time resolved Planck fluxes down to a time resolution ofa few hours. If the position of a variable point source is known, the residual 4D-map, R kj = I kj − ( S ∗ b ) kj , can be beam-deconvolved to the source position by a simple divisionand thus provide an estimate of the flux variation, (cid:104) ∆ S ( t j ) (cid:105) k = R kj /b kj , where b kj is a beamfactor map expressing the measured flux of a unit emitter at the source position in abeam centered at pixel k and time t j . This method is used, e.g., to extract time-resolvedPlanck fluxes for co-eval monitoring of blazars with the F-GAMMA program (Fuhrmannet al. 2007, Rachen et al. 2015).Our analysis tools work on unified data structures, which ensures that all methods forvariability analysis and flux extraction are applicable to LFI and HFI data in the sameway. Interfaces for intitial 4D-mapping are in place at both the LFI and HFI DPC. Weexpect that all Planck frequencies will be analysed for variability, and science resultsprepared for publication well before the Planck Legacy release. References
Fuhrmann, L., Zensus, J.A., Krichbaum,T.P., Angelakis, E., Readhead,A.C.S. 2007,
AIPC , 921,249Keih¨anen, E., Reinecke, M. 2012,
A&A
ApJ , 622, 759Planck Collaboration I. 2011,
A&A
A&A th Texas Symposium on Relativistic Astrophysics, GeneveTauber, J.A., Mandolesi, N., Puget J.-L., et al. 2010,