Florent RobinetLaboratoire de l’Accélérateur Linéaire robinet@lal.in2p3.fr

Detection of gravitational-wave signals from binary black hole mergers

2

2

3

The second one is usually better than the first one!

4

GW151226

5

GW151226

Rewind

6

On Sep. 14, 2015, both LIGO detectors detected the gravitational-wave signal produced by the coalescence of a binary black hole system:

GW150914

7

Inspiral Merger Ringdown

8

0.25 0.50 0.75 1.000.00

LIGO - LivingstonLIGO - Livingston

LIGO - HanfordLIGO - Hanford

0.25 0.50 0.75 1.000.00

9

10

11

12

13

Gμν=Rμ ν−12

gμ ν R=8 πG

c4 T μν=0

Add a small perturbation to the Minkowski metric:

gμν=ημν+hμν |hμ ν|≪1

– obeys a plane-wave equation– the wave propagates at the speed of light– 2 degrees of freedom: and

h

h+ hx

14

Gravitational waves

L±Δ L

h=2Δ LL

h+

hx

Gμν=Rμ ν−12

gμ ν R=8 πG

c4 T μν=0

Add a small perturbation to the Minkowski metric:

gμν=ημν+hμν |hμ ν|≪1

– obeys a plane-wave equation– the wave propagates at the speed of light– 2 degrees of freedom: and

h

h+ hx

15

Gravitational waves

L±Δ L

h=2Δ LL

h+

hx

When they propagate, gravitational waves– do not interact with matter– are attenuated by 1/r

→ Gravitational waves are the perfect probe! → BUT... h~10–21

Gravitational wave detectors

16

Poutput (t)=P input

2[1+C cos (δ ϕ( t))]

Poutput (t)≃Pinput

2[1+C cos(δϕOP)−C sin(δ ϕOP)×δϕGW (t )]

δϕ(t )=δϕOP+δϕGW

Michelson interferometer

L

→ A gravitational wave is detected as a power variation

Gravitational wave detectorsLIGOsuspensions

Virgosuspensions

O(100 kW)

Kilometer sca

leSuspended test masses

Virgo interferometer

Detector sensitivityThe detector's sensitivity: amplitude spectrum densityThe detector's sensitivity: amplitude spectrum density

18

Fundamental noises:– f < 10 Hz seismic noise→– 10 Hz < f < 200 Hz thermal noise→– f > 200 Hz quantum shot noise→

+ technical noise (environment, scattered light, control...)

19

LIGO Livingston, USALIGO Livingston, USA

LIGO Hanford, USALIGO Hanford, USA Virgo Pisa, ItalyVirgo Pisa, Italy

20

Scientific runs

02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21

S5 S6 O1

VSR1 VSR2VSR3

VSR4 VSR5

O2 O3

LIGO

Virgo VSR6

K1

KAGRA

LIGO

Indi

a

I1

?

?

?

initial enhanced advanced

21

LIGO sensitivity

2010 (enhanced)

2015 (advanced, O1)

2018 (advanced, O3)

2020+2020+

22

23

LIGO O1 run: Sep 12, 2015 Jan 19, 2016→

GW150914analysis

weeks

~ 50 days ofcoincident time

24

h(t) data are transferred to computing centers (< 1 s)where low-latency searches are conducted

25

h(t) data are transferred to computing centers (< 1 s)where low-latency searches are conducted

Hannover:– unmodeled burst search (cWB)

Caltech:– CBC modeled search (GstLAL)– unmodeled burst search (olib)

Virgo:– CBC-modeled search (MBTA)

Online searchesGW151226

26

Online searchesGW151226

Modeled search

27

Online searchesGW151226

~ 950 yr-1

28

Online searchesGW151226

Detected after ~ 1 min

29

Online searchesGW151226

30

Online searches

31

+9 min

+12 min

32

0.50 1.00 1.50 2.000.00

LIGO - LivingstonLIGO - Livingston

LIGO - HanfordLIGO - Hanford

0.50 1.00 1.50 2.000.00

GW150914

0.50 1.00 1.50 2.000.00

LIGO - LivingstonLIGO - Livingston

LIGO - HanfordLIGO - Hanford

0.50 1.00 1.50 2.000.00

0.50 1.00 1.50 2.000.00

LIGO - LivingstonLIGO - Livingston

LIGO - HanfordLIGO - Hanford

0.50 1.00 1.50 2.000.00

GW151226GW150914

33

Modeled search (offline)Theoretical input:Theoretical input:– 90s: CBC PN waveforms (Blanchet, Iyer, Damour, Deruelle, Will, Wiseman, …)– 00s: CBC Effective One Body “EOB” (Damour, Buonanno)– 06: BBH numerical simulation (Pretorius, Baker, Loustos, Campanelli)

The intrinsic waveform parameters:The intrinsic waveform parameters:– Masses:

– Spins and orbital angular momentum:

The waveform models used for the search:The waveform models used for the search:– Inspiral, PN3.5 for – Inspiral/Merger/Ringdown EOB + numerical relativity for– Spins and orbital angular momentum are aligned Template bank match-filtering technique→Template bank match-filtering technique→

M tot<4 M sun

M tot>4 M sun

S⃗ tot=S⃗1+ S⃗2+ J⃗

M tot=M1+M 2

34

Modeled search (offline)

35

2 modeled searches are conducted offline: pyCBC and GstLAL

They use the same template bank but they differ when it comes to:– estimate the background– rank the events

Modeled search

36

Full O1 analysis, final calibration

Iterative process → remove the loudest event (GW) from background before

estimating the second-loudest event significance

pyCBC GstLAL

Modeled search

37

Full O1 analysis, final calibration

Iterative process → remove the loudest event (GW) from background before

estimating the second-loudest event significance

pyCBC GstLAL

Modeled search

38

Full O1 analysis, final calibration

pyCBCGW151226

● SNR = 13.0

● Significance > 5.3 σ

Monitoring noise200,000 auxiliary channels are used to monitor the instrument– environmental sensors– detector sub-systems– detector control

Noise injection campaigns are conducted to identify the detector's response to different noise stimulation

Multiple transient noises were identified during the run → Anthropogenic noise → Earthquakes → Radio-frequency modulation …→

Option #1: fix the detectorOption #2: remove transient

events in the data

No indication of external disturbanceat the time of the event

39

Parameter estimationBayesian inference + MCMC analyses using accurate waveform models

40

Parameter estimation

41

Black hole population

42

Source location

43

DL=420−180+150 Mpc

DL=440−190+180 Mpc

z=0.09−0.04+0.03

z=0.09−0.04+0.03

90% credible region for sky location: → GW150914 = 230 deg2

→ GW151226 = 850 deg2

GW150914:GW151226:

Event rate

44

– Simulated signals with parameters drawn from astrophysical populations– Noise distribution from CBC searches

p(m1 ,m2)∝m1−1m2

−1

underestimates the rate

p(m1)∝m1−2.35

overestimates the rate

p(m2)=cte

R=9−240Gpc−3 yr−1

Detection rate

45

False-alarm rate = 1 / 100 yearsN>10N>35

N>70

Waveform

46

Testing GR

47

Opportunity to study GR in a strong-field regime (update GW150914 results)Opportunity to study GR in a strong-field regime (update GW150914 results)Test #1 signal waveform/GR consistency: residual compatible with noise→

Test #2 BBH parameter consistency before/after merger: excellent→

Test #3 deviation from PN waveforms: constraints on PN coefficients→

Test #4 consistency with the least-damped quasi-normal-mode of the remnant black hole→

Test #5 theory with massive graviton: best constraints on the graviton mass→

Upper bounds onthe fractional deviation in PN coefficients

Conclusions & outlook → 22ndnd gravitational-wave event was detected gravitational-wave event was detected :

– lower masses than GW150914 (compatible with known X-ray binaries)– at least one companion black hole is spinning– population of binary black holes

→ Observations consistent with GR

→ Gravitational-wave events can be detected with a minute latency: first step in GW astronomy

The future is promisingThe future is promising

→ Additional O1 results will be released soon (CBC, burst sources, pulsars)

→ Improved detector sensitivity

→ More events: population, new types of sources (supernovae, neutron stars)

→ New instruments will join the network (Virgo this year!)

48

THANK YOUTHANK YOU

Event significanceThe background of a gravitational-wave search is estimated using the time-slide techniqueAssumption = uncorrelated noise between detectors

Coincidence time window

+δ t

True experiment = noise + signal

List of background events

Fake experiment = noise only

List of candidates

time time

A very large number of fake experiments can be simulatedusing multiple offsets

→ background estimated using a fake experiment of O(100,000 years)

Hanford

Livingston