Status of VIRGO Sipho van der Putten
2 Contents Introduction to gravitational waves VIRGO Pulsars: gravitational waves from periodic sources Pulsars in binary systems Analysis approach
3 Introduction to Gravitational Waves ‘Ripples’ in space-time due to accelerating masses distorting space- time Two polarizations: ‘+’ & ‘x’ Measured in strain: h=dL/L Extremely weak effects: Supernova (10 kpc, ~10 M solar ): h~ Rotating deformed neutron star (~10 kpc, ~1 M solar, 100 Hz): h~ Ring of free falling masses L
4 VIRGO Science Run VSR1 complete Oct ‘07 Combined LIGO & VIRGO data taking ~5 months data Current upgrade (VIRGO+) underway 5x better at low freq & 2x better at high freq Nikhef: Electronics, IMC VSR2 in mid 2009: Combined run with eLIGO 1 year of data Next upgrade AdvVirgo 10x better in sensitivity than initial VIRGO design 2010 to 2012
5 Periodic sources of gravity waves Pulsars: spinning neutron stars Emitting GWs requires quadrupole moment; symmetry axis is not rotation axis Neutron stars: Isolated Binary systems
6 Neutron stars in binary systems 2/3 of NS (f>10Hz) in a binary system Mass transfer: Spin-up: f increases Many parameters : Orbital: Sky Position: Source: …. Doppler shift due to orbit of binary system: non stationary frequency Our goal: All-sky search for neutron stars in binary systems
7 Doppler shift Fixed point emits stationary frequency Rotation of Earth (daily motion) df/f~10 -6 Earth’s orbit (yearly motion) df/f~10 -4 Pulsars in binary system df/f~10 -3 Hulse-Taylor system All shifts included
8 Analysis binaries: spectral filtering Non stationary frequency: FFT → power spread out, bad S/N Spectral filtering: Identify the signal in the data f sig = Hz f sig = – 5*10 -4 t Hz
9 Conceptual approach Split the data in time stretches Second order spectral filter Each matching filter: f 0,a,b,t avg Pattern recognition using all the information available
10 Simulations Simulated waveform: Hulse-Taylor system Divide in time slices 1286 s Simple FFT each slice Time (s) Frequency (Hz) h rec Time (s) Frequency (Hz) h re c P(noise)=100 P(sig)
11 Spectral filtering Spectral filters: 2607 filters, applied ~10 9 times Threshold 4x noise level CPU time: 1 day for ~10 h data (0-400 Hz) Many improvements in efficiency possible Non-stationary frequency not an issue anymore To do: Investigate higher order filters→ longer FFTs Pattern recognition Time (s) Frequency (Hz) h re c P(noise)=100 P(sig)
12 Conclusions VIRGO & LIGO have 5 months of science-grade data VIRGO+ & eLIGO upgrade underway and will be ready in 2009 Analysis of binary pulsars: Conceptual approach to analysis: spectral filtering S/N 0.1 easy with 2 nd order filter Time dependent frequency no problem Todo: test idea on simulations and real data (higher frequency signal, more noise) Todo: implement pattern recognition
13 P(noise)=10 4 P(sig) Time (s) Frequency (Hz) h re c