CW signals: plans for S2

1. CW signals: plans for S2 June 16-17 the group held a f2f meeting to discuss plans for analyses of S2 data (and beyond). Slides from the various presentations can be found at www.lsc-group.phys.uwm.edu/pulgroup/ The content of this presentation is mostly cut-and-paste from these.

2. Coherent search methods • Time domain Bayesian analyses: • extend to all known isolated pulsars emitting above ~ 50 Hz (Glasgow). • special search for Crab (Glasgow) • Metropolis-Hastings Markov Chain Monte Carlo method for parameter estimation and small parameter space searches (Carleton) • Frequency domain analyses: • extend to inspect a larger parameter space (non-targeted search) (AEI+UWM) • extend to search for signals from a pulsar in a binary system, will be applied to Sco X-1 (Birmingham)

3. Non-coherent search methods • Unbiased searches (Michigan) • Hough search (AEI) • Stack-slide (Hanford)

4. TDS: Known pulsars (R. Dupuis, G. Woan, GLASGOW) • Perform similar statistical analysis as S1 for all known isolated radio pulsars with fGW > 50 Hz. • Total of 34 pulsars • Computational cost minimal. • Processing time on Medusa should be around 15 min per pulsar. Another 15 minutes on a workstation to calculate posterior probability. • Automatic characterization of noise in small freq. band • Would like to perform a coherent analysis using all IFOs for every pulsar. • Preliminary results should be available by August LSC.

5. Isolated pulsars with fGW > 50 Hz #PULSARFREQUENCY (HZ) 1 B0021-72C 173.708218966053 2 B0021-72D 186.651669856838 3 B0021-72F 381.15866365655 4 B0021-72G 247.50152509652 5 B0021-72L 230.08774629150 6 B0021-72M 271.98722878893 7 B0021-72N 327.44431861755 8 J0024-7204V 207.90 9 J0030+0451 205.53069927493 10 B0531+21 30.2254370 11 J0537-6910 62.055096 12 J0711-6830 182.117237666390 13 J1024-0719 193.71568669103 14 B1516+02A 180.06362469727 15 J1629-6902 166.64990604074 16 B1639+36A 96.362234564 17 J1709+23 215.94 18 J1721-2457 285.9893480119 19 J1730-2304 123.110289179797 20 J1744-1134 245.4261237073059 21 J1748-2446C 118.5382530563 22 B1820-30A 183.82343541777 23 B1821-24 327.40566514989 24 J1910-5959B 119.6487328457 25 J1910-5959C 189.48987107019 26 J1910-5959D 110.6771919840 27 J1910-5959E 218.7338575893 28 B1937+21 641.9282611068082 29 J2124-3358 202.793897234988 30 B2127+11D 208.211688010 31 B2127+11E 214.987407906 32 B2127+11F 248.321180760 33 B2127+11H 148.293272565 34 J2322+2057 207.968166802197 Black dots show expected upper limit using H2 S2 data. Red dots show limits on ellipticity based on spindown.

6. Crab pulsar (M. Pitkin, R. Dupuis, G. Woan, GLASGOW) RMS phase deviations (rad) RMS frequency deviations (Hz) Time interval (months) • Young pulsar with large spindown and timing noise • Timing noise not irrelevant: Frequency residuals will cause large deviations in phase if not taken into account.

7. More on the Crab • Removing the timing noise • Use Jodrell Bank monthly crab ephemeris to recalculate spin down params every month. • Calculate phase residuals between radio observations and fixed spin down parameters.Then add another heterodyning phase to data processing to remove timing noise. • Open Issues • Could add another parameter  = - IEM / IGW (I. Jones paper) • No glitches during S2 but in the event of a glitch we would need to add an extra parameter for jump in GW phase • 60 Hz, filters • Preliminary results should be available by August LSC

8. Time Domain Searches MCMC's (N. Christensen) Definitely useful to nail down signal parameter values around candidates Might be useful for small parameter space searches, e.g. the Middleditch optical pulsar in SNR1987A. On-going work to efficiently search over additional parameters (frequency, fdot, wobble angle) in low signal-to-noise conditions. Automation of TDS (G. Santostasi) Plan to automate the data processing done in the time domain search making it an on-line analysis (perhaps LDAS?) Just starting to look at this.

9. FDS: parameter space searches (M.A. Papa, X. Siemens – AEI , UWM) • Have modified the software used for the targeted S1 search to search for arbitrary ranges of f0 (with arbitrary resolution), areas in the sky and spin-down parameter values (now only 1st order) • The limiting factor is computational time need to decide how to spend the computing cycles that we have • We have decided to pursue two types of searches: • short observation time (low sensitivity), no spin-down paramaters, wide band (300-400 Hz), all sky search • longer observation time (higher sensitivity), 1 spin-down param, small area search (galactic plane/first spiral arm), low frequencies (~ 70 Hz), small bands Note: different choices could be made in order to produce the best ULs.

10. FDS: Sco X1 (C. Messenger, A. Vecchio – Birmingham) • Sco X-1: • Frequency: not known a priori – search on a wide band (~ 100 Hz) • Doppler effects: • Source position: known (no search over sky position) • Orbit: circular (e < 3x10-4), therefore 3 search parameters (projected semi-major axis, orbital period, initial phase on the orbit) • For T < 1 month period is not a search parameters • Search over only 2 parameters (flat parameter space) • Spin-down/up: number of spin-down parameters depend on the integration time, but for T < 10 days signal monochromatic • Physical scenarios: • Accretion induced temperature asymmetry (Bildsten, 1998; Ushomirsky, Cutler, Bildsten, 2000) • R-modes (Andersson et al, 1999; Wagoner, 2002)

11. PSR J1939+2134 Sco X-1 (Rejean’s plot) Upper-limit on Sco X-1 Sco X-1 and LIGO/GEO sensitivity PULG F2F Meeting, UWM, 16th - 17th March, 2003

12. Unbiased All-Sky Search (Michigan)[ D. Chin, V. Dergachev, K. Riles ] Analysis Strategy: (Quick review) • Measure power in selected bins of averaged periodograms • Bins defined by source parameters (f, RA, d) • Estimate noise level & statistics from neighboring bins • Set “raw” upper limit on quasi-sinusoidal signal on top of empirically determined noise • Scale upper limit by antenna pattern correction, Doppler modulation correction, orientation correction • Refine corrected upper limits further with results from explicit signal simulation Unbiased, All-sky CW Search -- Michigan Group

13. Preliminary Data Pipeline Diagram for Unbiased All-Sky CW Search Creation of Power Statistic Simulated Data Raw Data • Loop over frequency and sky: • Determine search range and control sample range • Determine upper limit on detected power • Apply efficiency corrections (Doppler, AM, orientation • Determine limit on h0 and store Determine efficiency corrections • S2 Numbers: • 59 days = 1416 hours = 84,960 minutes = 5.1 × 106 seconds  Single FFT: 0.2 μHz bins • Power sum for 1-minute FFT’s: 0-2000 Hz x 60 bins/Hz x 4 Bytes/bin = 480 kB (17mHz bins) • Power sum for 30-minute FFT’s: 0-2000 Hz x 1800 bins/Hz x 4 Bytes/bin = 14.4 MB (0.56 mHz bins) Overview:

14. What will be ready by August LSC meeting? • Expect bare-bones baseline limits over samples of clean frequency ranges with semi-analytical efficiency corrections • Primary goal in next two months What is unlikely to be ready by August: Graceful handling of problematic frequency ranges Efficiency corrections based on full MC simulations Refinements under consideration but not in baseline: Non-uniform weighting of SFT’s Explicit consideration of spin-down parameters Shifting frequency search range vs SFT

15. Hough Transform (B. Krishnan, A. Sintes, (map) – AEI) • start with SFTs with baseline of ~ 30 minutes (or more if stationarity of data allows it). • for every SFT (a) select frequency bins i such |SFT|ia2/<|SFT|2>a exceeds some threshold  time-frequency plane of zeros and ones. • Hough transform: track patterns of “ones” consistent with what one would expect from a signal. Something should be added regarding status

16. Stack slide (M . Landry, G. Mendell – Hanford) • new proposal presented in May at a meeting devoted to new data analysis proposals • based on P.R. Brady, T. Creighton Phys.Rev. D61 (2000) 082001 • it is a hierarchical search: the coherent stage could be performed by either resampling and FFT-ing or by the F statistic. The incoherent search consists in summing appropriately (by choosing the appropriate search frequency values) the outcome of the first stage. • as a first step, the idea is to implement a stack-slide that starts from SFTs. Similar to Hough, but without peak selection.