Data analysis challenges in gravitational-wave astronomy ric - - PowerPoint PPT Presentation

Data analysis challenges in gravitational-wave astronomy

Éric Chassande-Mottin*

for the LIGO Scientific Collaboration and the Virgo Collaboration

* CNRS, AstroParticule et Cosmologie, Paris France

Outline

Gravitational waves Direct detection of GW with large-scale interferometers Searches for GW transients and related data analysis challenges Multimessenger astronomy with GW Future detectors Conclusions

Gravitational waves GW

• Propagating distorsions of space-time

metric

• Predicted by General Relativity
• Propagate at the speed of light
• Transverse and quadrupolar (in far field)
• Two polarizations (+ and x)
• Dimensionless strain amplitude h
• Sources of GW
• Produced by accelerated mass
• Rapid changes in shape and orientation
• f massive objects
• Large mass and density, relativistic

motion → astrophysical sources

plus cross

Indirect evidence

• Binary pulsar PSR B1913+16
• Orbital decay → energy loss due to GW
• In agreement with GR to ~0.2 %
• Hulse & Taylor's Nobel prize

Binary orbit will continue to decay

• ver 300 millions years until coalescence
• GWs from binary systems
• Estimate with quadrupole formula
• For a binary close to coalescence

R=20 km, M=1.4 Msun, f=400 Hz, d=15 Mpc

Direct detection of GW

• Michelson interferometer
• test mass displacement due to GW→

phase shift measurement

• Sees mixture of both polarizations
• Large aperture: not directional
• more like a ear than an eye!
• 1D time series (not a 2D image)

Sensitivity of interferometric GW detectors

• High-precision metrology
• Measurement limitations
• Fundamental sensing and

displacement noises

• “Technical” noises (controls,

electronics, acoustic, etc.)

• Observable freq. band
• From few 10 Hz to few kHz

Detector highlights

Long folded arms using Fabry- Long folded arms using Fabry- Perot cavities Perot cavities Suspended instrument Suspended instrument High-Q material High-Q material Ultra-high vaccum Ultra-high vaccum High-power stabilized laser High-power stabilized laser Long folded arms using Fabry- Long folded arms using Fabry- Perot cavities Perot cavities Suspended instrument Suspended instrument High-Q material High-Q material Ultra-high vaccum Ultra-high vaccum High-power stabilized laser High-power stabilized laser

3 km

Worldwide network of GW detectors

GEO 600 Germany Virgo Italy LIGO US Since 2007, partnership and data exchange agreement

Sources of gravitational waves

We will be interested in transient sources in this presentation Data analysis is challenging for the other GW sources too!

Sources of gravitational wave transients

• Catastrophic astrophysical events

the “violent Universe”

• Efficient production of GWs
• compact objects: neutron stars (NS)
• r black holes (BH)
• bulk motion at relativistic velocities
• Some degree of asymmetry
• Binary mergers (BBH, BNS)
• post-Newtonian chirps + numerical relativity
• Supernova core collapses
• numerical simulations. no comprehensive

view of the collapse. few predictions, robustness?

• … and others (e.g. star quakes,

cosmic strings, etc)

• 3 joint LIGO – Virgo science runs

~2 yrs total

• NS-NS = 1% total mass emitted in GW

horizon is ~ 20 – 40 Mpc

• Core Collapse SN = 10-8 M c2

galactic SN are observable

Achieved sensitivity and data takings

strain sensitivity (Hz-1/2) frequency (Hz) 103 102 10-21 10-23 10

S5 S6

VSR 1 2 3 4

Searches for GW transients: basic ideas

Expected signal is Expected signal is unknown unknown Excess in time-frequency maps Excess in time-frequency maps (wavelets) (wavelets) Expected signal is Expected signal is known known (inspiralling binaries) Matched filtering Matched filtering

Time series analysis Time series analysis rare transients with low signal to noise ratio rare transients with low signal to noise ratio

Challenges with real-world data (1)

• Non-stationary and non-

Gaussian

• zoo of instrumental glitches →

background has heavy tails

• Data quality is a key issue
• Veto known artifacts
• Cross-correlation with >100 auxiliary

channels

• Trade-off: maximize “efficiency”

(fraction of glitches that get vetoed) and minimize “dead time” (volume of vetoed data)

• Safety checks with “hardware”

injection of fake GW signals

• 70 DQ flags, efficiency 90% for loud

glitches

p

• w

e r l a w loud glitches

bulk of the glitch population

Challenges with real-world data (2)

• Due to instrument complexity, comprehensive noise

modelling is out of reach

• Background estimation is also a key issue: “time-slide”

analysis

• Exploit availability of multiple detectors
• Apply non-physical (> 1 s) time-shifts to data stream and repeat analysis

→ Reference background distribution of noise-only events

• Compare distribution of non time-shifted (“zero-lag”) events to reference

to get confidence (probability of occurrence)

• Limitation of the number of time-slides (1 s – 1 day)

Worldwide network of GW detectors

time delay scale factor phase shift

all detectors receive the same polarizations but GW couples differently according to the antenna patterns

• rientation

phase shift – scaling (antenna patterns) position time delay (propagation) this property is specific to GW and can be used to eliminate noise events

Challenges with real-world data (3)

Background rejection using multiple detectors

• require time coincidence and

phase consistency

• “coherent veto”: signal vs null

P is the projector onto noise or null space linear combining of data from each detector so that GW signatures cancelled in the sum coherent and incoherent projected energies

signal space is a 2D plane!

GW: on and off-diag. terms cancels (are of same order) Glitch: off-diag. terms much smaller than on-diag.

Selection of results

• Latest “all-sky” burst search
• S5-VSR1 & S6-VSR 2/3: 2 yrs
• bservation total
• Transients (< 1s) in 64 Hz– 5 kHz
• Search with coherent WaveBurst
• No GW candidate event
• Upper-limits on the rate of bursts

estimated using generic waveforms

standard candle EGW=1 Msun c2

detectable GW energy at a given distance 10 kpc: EGW= 3 x 10-8 Msun c2 (comparable to CC SN) 15 Mpc: EGW= 10-1 Msun c2 (comparable to black-hole binary merger)

Stress test with “blind injections”

• Blind injection challenge
• fake signals secretly added to the

data to test the detector and analysis

• Nick name “Big Dog”
• Event (inspiral) successfully

recovered as a detection candidate with a FAR < 1/7000 y.

• The process was a valuable end-

to-end test of our analyses Analysis → Paper draft on event candidate → Internal review → Detection committee → Envelope

• pening

arXiv:1111.7314

http://www.ligo.org/science/GW100916

Envelope opening Envelope opening

Connection to high-energy astrophysics

low medium high energy range Gamma-ray burst and their afterglow Soft-gamma repeaters Anomalous X-ray pulsars Pulsar glitches low high energy range

neutrino neutrino electromagnetic electromagnetic

GW GW

Prompt search for EM counterparts

• Low-latency analysis (30 mins)
• Transfer, qualify and search the data
• Select candidates, reconstruct source direction (3

detectors)

• Error region made of disconnected islands

10 to 100 sq degrees

• Wide-field robotic telescopes (but not only)
• Observe asap + subsequent nights
• Galaxy mass targetting, FOV ~ few sq deg

EM EM

alert message alert message

Example of GW error region

2nd generation of “advanced” detectors

• x 10 sensitivity improvement
• x 103 observable volume
• Neutron star binaries ~ 140 Mpc
• Typ. few tenths of detectable binaries

per year

• Black binaries ~ 1 Gpc
• First science data ~ 2015
• Installation is under way
• Kagra (Japan), LIGO India
• x 10 improvement in angular resolution

x 10

Summary

• A new window on the Universe opens with GW observations
• Our searches cover a wide range of expected sources
• We are prepared to detect a signal
• Developing synergy with high-energy astrophysics
• With the 2nd generation of instruments, the next decade will

probably see the 1st direct detection of GW Stay tuned! Stay tuned!