LISA Data Analysis: progress and open issues

Transcription

1 LISA Data Analysis: progress and open issues Alberto Vecchio University of Birmingham LISA 6-8 February, 2008

2 Summary We have a vanilla demonstration of the capabilities of data analysis for the key sources: population of white dwarfs in the Galaxy, supermassive black hole binaries, extreme-mass ratio inspirals (EMRIs) We have a standard framework where to test and compare analysis tools, and evaluate LISA s science capabilities At the conceptual level, we know how to tackle the analysis for the expected LISA sources (with the exception of isotropic stochastic signals) Huge range of technical issues are way open

3 Outline The LISA data analysis challenge A selection of recent progress (largely taken from the Mock LISA Data Challenges): Super massive black holes Extreme-mass ratio inspirals Some open issue: stochastic signals

4 Global analysis from MLDC-2

5 LISA data analysis GW astronomy using: Six one-way Doppler readouts Modulations due to the motion of constellation Full information encoded in 3 time series (A, E, T), linear combinations of the waveforms: h+ = h = M5/3 [πf (t)]2/3 Q+ (geometry) cos ΦGW (t) DL M5/3 [πf (t)]2/3 Q (geometry) sin ΦGW (t) DL A.Vecchio, LISA Astro-GR@ComoMilano 7 Feb. 2008

6 Data rate Data rate is not an issues here (very different with respect to LIGO, GEO, Virgo, etc.) A.Vecchio, LISA 7 Feb (courtesy K. Danzmann)

7 Mock LISA Data Challenges Why: How: - - Demonstrate that LISA science requirements can be met Understand data analysis quantitatively (e.g. implications for design requirements) Kick-start development of LISA data analysis computational infrastructure Encourage development and comparison of analysis techniques Coordinated, voluntary effort in the LISA community Periodically issue challenge data sets with participants returning best fit parameters

8 Genesis of MLDC AMIGOS LISA Data-Analysis Planning Meeting LISA Computational Infrastructure LISA Simulator Synthetic LISA LISACode LIGO experience with LAL, etc Open development MLDCs Sensitivity Generator LISA Calculator Astrogravs Prove that LISA data analysis is feasible Establish common playground for experiments Encourage data-analysis development

9 Involvement MLDC Task Force: Keith Arnaud Stas Babak John Baker Matt Benacquista Neil Cornish (WG-1B co-chair) Jeff Crowder Curt Cutler Sam Finn Steffen Grunewald Shane Larson Tyson Littenberg Eric Plagnol Ed Porter B.S. Sathyaprakash Michele Vallisneri (co-chair) Alberto Vecchio (co-chair) Jean Yves Vinet MLDC Participants: NASA Ames U. of Auckland Chinese Academy of Sci., Beijing U. Birmingham U. Southampton U. Texas Brownsville Caltech/NASA JPL U. Cambridge Cardiff U. Carleton College U. Glasgow NASA Goddard Albert Einstein Institut Golm Albert Einstein Institut Hannover U. Illes Balears Indian Inst. of Tech., Kharagpur IMPAN Warszaw Montana State U.

10 Task Force Charter Specify pseudo-lisa model Plan challenge progression Identify standard source models Specify data format (lisaxml) Prepare and distribute training and challenge datasets Develop software infrastructure Compile challenge submissions Common framework where to evaluate and compare analysis tools (and therefore address LISA science capability)

11 MLDCs History Jan 2006 work begins! Dec 2006 Challenge 1 results due; presented at GWDAW-11 [CQG 24, S529 (2007), plus several by participants] Jun/Jul 2007 Challenge 2 results due; presented at Amaldi (arxiv.org/ ). Challenge 1B released Jun 2006 Challenge 1 datasets released at 6th LISA Symposium [proceeds, gr-qc/ ] Jan 2007 Challenge 2 datasets released [CQG 24, S551 (2007)] Dec 2007 Challenge 1B results due, presented here. Challenge 3 released

12 MLDC 1 MLDC 2 MLDC 1B MLDC 3 more... EMRIs SMBHs Galactic binaries Verification binaries Unknown isolated Unknown interfering Isolated [restricted 2PN] Galaxy (3x10 7 sources) Number unknown (4-6) over Galaxy + EMIRs [restricted 2PN] Isolated Several over Galaxy + SMBH Verification binaries Unknown isolated Unknown interfering Isolated [restricted 2PN] Isolated Galaxy (6x10 7 sources) chirping +/- Over Galaxy (loud sources removed) [+spins and precession] Number unknown (4-6), together and weaker Bursts from cosmic string cusps Isotropic stochastic background

13 MLDC 1 MLDC 2 MLDC 1B MLDC 3 more... EMRIs SMBHs Galactic binaries Verification binaries Unknown isolated Unknown interfering Isolated [restricted 2PN] Galaxy (3x10 7 sources) Number unknown (4-6) over Galaxy + EMIRs [restricted 2PN] Isolated Several over Galaxy + SMBH Verification binaries Unknown isolated Unknown interfering Isolated [restricted 2PN] Isolated Galaxy (6x10 7 sources) chirping +/- Over Galaxy (loud sources removed) [+spins and precession] Number unknown (4-6), together and weaker Bursts from cosmic string cusps Isotropic stochastic background

14 Some more info MLDC official site: astrogravs.nasa.gov/docs/mldc MLDC taskforce wiki: Mailing lists: (formulation) (participants) LISAtools software (including full MLDC pipeline): lisatools.googlecode.com

15 A range of analysis technique Coherent match-filter based Coherent through stochastic sampling (Reversible Jump Markov Chain Monte Carlo) Time-frequency and pattern recognition Hilbert transform Tomographic reconstruction Single-stage searches Multiple-stage hierarchical searches Many LAL algorithms applied to LISA (mock) data

16 Stellar-mass binaries Demonstrated: - Galaxy regression (reach confusion noise level) - Source identification

17 Galaxy regression

18 SMBHs SMBH over Galaxy restricted 2PN waveform unknown number Range of SNR (~ )

19 SMBHs Statistical errors in a real analysis (of mock data) Systematic errors (accuracy of waveforms)

20 SNR is not all: SMBH-1 even with large SNR position can be way off

21 Bimodal distributions: SMBH rad ~ 0.5 o

22 Weak(er) signal + galaxy regression: SMBH-2

23 Weak(er) signal + galaxy regression: SMBH-2 ΔM/M ~ 10-3 Δμ/μ~ 10-2 ΔΩ ~ 2 o x5 o ΔD/D~10-2

24 Systematic vs statistical errors Cutler & Vallisneri, 2007 arxiv:

25 Systematic vs statistical errors Cutler & Vallisneri, arxiv:

26 Summary of errors statistical systematic* [3.5PN - 3PN*] systematic [3.5PN - 3PN] ΔM/M x x Δη/η x x Δθsky 1 o 0.1 o 0.1 o normalised at SNR = 1000

27 Higher harmonics M 1 = M 2 = 10 7 M z = 1 Trias and Sintes, 2008; Arun et al. 2007, arxiv: ; Cornish & Porter, in preparation?

28 Spins sky position no precession no precession precession precession precession no precession Lang & Hughes, 2007 distance n

29 Precessing SMBH in MLDC3 total instrument noise precessing SMBHs (circular orbit) regressed galaxy MLDC-3

30 LISA Performance Evaluation Goals: Task Force 1. to develop reliable (vetted) codes for calculating LISA s parameter estimation accuracy for different sources (e.g., by the Fisher matrix approach, which is accurate in the high-snr limit, or more generally by MCMC methods) 2. to develop source population models, so the codes developed in (1) can be run in Monte Carlo fashion over a representative sampling of sources. S. Babak J. Baker L. Barack E. Berti N. Cornish (WG 1B co-chair) C. Cutler (co-chair) J. Gair E. Porter B.S. Sathyaprakash A. Sintes (co-chair) M. Trias C. van den Broek M. Volonteri H. Wang A. Vecchio, LISA Astro-GR@ComoMilano 7 Feb. 2008

31 EMRIs Isolated Barack&Cutler (2004) waveforms SNR ~ 50

32 EMRIs We do have techniques that work at SNR~50 for isolated sources (At least) 2 independent coherent analysis approaches Time-frequency pattern tracking Key (future) tests: - lower SNR (~20-30) [in MLDC3] - over galaxy

33 Stochastic signals Two classes in terms of signal properties Anisotropic: by far the dominating contribution (white dwarfs in the Galaxy) Isotropic: - - cosmological background from the early Universe foreground from populations of astrophysical sources at z ~ 1

34 Anisotropy Well understood problem (Cornish, 2001; Ungarelli & AV, 2001; Hideaki & Atsushi, 2005; Atsushi & Hideaki, 2005), without demonstration yet However, this analysis is done routinely with groundbased data Upper-limit map Pulsar hardware injection Abbott et al [LIGO Scientific Collaboration] PRD 76, (2007)

35 Cosmological stochastic backgrounds

36 Isotropic foregrounds Extra-galactic WDs Cosmic population of EMRIs Farmer & Phinney, 2004 Barack & Cutler, 2004b

37 GW response insensitive to GWs sensitive to GWs Tinto et al, 2001

38 T as noise calibrator? T 2 = σ 2 n A 2 = a σ 2 n + σ 2 h!"!!$!"!!#!"!!* T 8-09:;:<:;231/,<-9 >,-!?@A!B B! " C? CD?D 8-09:;:<:;234!=,;567!"!!)!"!!(!"!!' A a Tinto et al., 2001; Hogan & Bender, 2001!"!!&!"!%"!"!#!"!$!"!%!"!! +,-./

39 T as noise calibrator? not quite T 2 = σ 2 n A 2 = a σ 2 n + σ 2 h no prior knowledge on a prior knowledge (10% level) on a

40 T as noise calibrator, take 2 ) α!"!!$!"!!#!"!!* T 8-09:;:<:;231/,<-9 >,-!?@A!B B! " C? CD?D T (f) 2 = σ 2 n A(f) 2 = a ( f ( f f 0 f 0 ) β T (f) 2 + σ 2 h ( f f 0 ) γ 8-09:;:<:;234!=,;567!"!!)!"!!(!"!!'!"!!& a A!"!%"!"!#!"!$!"!%!"!! +,-./

41 T as noise calibrator: possibly ( f ) α T (f) 2 = σ 2 n A(f) 2 = a ( f f 0 f 0 ) β T (f) 2 + σ 2 h ( f f 0 ) γ However it assumes: β < 0 symmetric noise

42 Conclusions We have a first demonstration of the basic capabilities of data analysis that underpin LISA science We have a standard framework where to: (i) test and compare analysis tools, and (ii) evaluate LISA science performance At the conceptual level, we know how to tackle the analysis for the expected LISA sources (with the exception of isotropic stochastic signals) A plethora of technical issues are open and require active stusy