Noise characterization for LISA Julien Sylvestre Massimo Tinto - PowerPoint PPT Presentation

Noise characterization for LISA Julien Sylvestre Massimo Tinto Caltech/JPL GWDAW 2003 1 of 8

Noise characterization with one IFO • Measured signals are the sum of (possibly continuous) gravitational wave signals and of instrumental noise • The instrumental noise must be estimated in the presence of the gravitational wave “noise” • Look for TDI measurements that are insensitive to GW but that are still affected by the instrumental noises 2 of 8

Review of TDI • The LISA array forms 12 phase measurements ›› 6 one-way measurements along the arms ›› 6 inter-bench measurements, 2 per spacecraft • Form linear combinations of delayed phase measurements to cancel overwhelming laser and optical bench noises • Many possible TDI combinations; all can be generated by only four combinations: α, β, γ, ζ 3 of 8

ζ as a GW shield • The fully symmetric Sagnac TDI combination ζ is free of laser and optical bench noise, and is much less sensitive to GW than the other TDI combinations Tinto, Armstrong and Estabrook, Phys. Rev. D 63 , 021101 (2001) 4 of 8

Noise measurements with ζ • Can form four GW-free spectra: S ζζ , S αζ , S βζ , S γζ ›› Definition: S ab = E[a b * ] • How well can an arbitrary spectrum be approximated as a linear combination of these four spectra? S αα = a S αζ + b S βζ + c S γζ + d S ζζ • Well-defined linear approximation problem. Solution is a least-squares fit • J. S. and M. Tinto, PRD 68 , 102002 (2003) 5 of 8

Example: S αα −40 10 |spectrum| (Hz −1 ) −41 Real 10 −42 10 Estimated −43 10 −4 −3 −2 10 10 10 f (Hz) 0.2 0.1 spectrum error 0 −0.1 −0.2 −0.3 −4 −3 −2 10 10 10 f (Hz) 6 of 8

Application: stochastic background −20 10 Confusion noise from galactic binaries 1 −21 10 1 σ upper limit on S XX strain −22 10 S XX −23 10 1yr, SNR = 5 −24 10 −4 −3 −2 10 10 10 f (Hz) 1. Bender and Hils, Class. Quantum Grav. 14 , 1439 (1997) 7 of 8

Conclusion • Below ~5 mHz, the spectrum of the instrumental noise can be esti- mated using ζ with a relative error of ~20% • This leads to a negligible loss in SNR (~0.1%) when doing matched filtering • Directly translates into a ~20+% sensitivity loss when searching for a stochastic background using excess noise • It seems possible to use ζ to construct estimators of the instrumental noise that are good enough for the data analysis, both for signal detection and for signal estimation 8 of 8