Gravitational-wave Astronomy and Astrophysics with LIGO and Virgo

Chris Pankow (CIERA/Northwestern)for the LIGO and Virgo Collaborations

DCC: G1701947

• The LIGO and Virgo Collaborations: 1000+ scientists, engineers, and others spread amongst 50+ academic institutions world wide (presence on all continents except Africa and Antarctica)

• Collectively develop and operate a network of three kilometer scale interferometers (LIGO Hanford, LIGO Livingston, Virgo), and a 600m pathfinder/technology testbed interferometer (GEO600)

• Two kilometer scale interferometers under construction (KAGRA collaboration, Japan) or in design process (LIGO India)

The LVC: Who We Are2

R. Williams (Caltech)

VIRGO

Advanced Interferometer Spectral Sensitivity3

Seismic motion shot noise

power line couplings

calibration lines

Low Frequency: motion of the Earth

coupling into motion of the test masses

High Frequency: uncertain photon arrival times at photodetector

Monochromatic Lines: calibration lines, 60 Hz power

line and harmonics thereof

Left: Spectral sensitivity during O1

Bottom: Typical spectral sensitivity

during late O2

GW Signal Detection Primer4

Putative strain is embedded in detector noise — cross correlate the model with the data to extract a signal-to-

noise ratio (SNR, ρ) statistic — this maximizes the likelihood (probability of signal vs probability of noise)

Searches maximize likelihood analytically for speed and

over masses/spins by brute force (template banks)

arxiv:1606.04856

d(t) = h(t) + n(t)

GW Signal Detection Primer5

Putative strain is embedded in detector noise — cross correlate the model with the data to extract a signal-to-

noise ratio (SNR, ρ) statistic — this maximizes the likelihood (probability of signal vs probability of noise)

Searches maximize likelihood analytically for speed and

over masses/spins by brute force (template banks)

arxiv:1606.04856

d(t) = h(t) + n(t)

H1+L1 SNR = 13

Remnant mass ~47 M☉

dL = 880 Mpc z = 0.18

PRL 118, 221101

GW1701046

Dealing with Multiple Event Categories: Being unsure of the intrinsic source populations and origins, we calculate the event rates for distributions of events according to uniform in the logarithm of component mass and

according to the stellar initial mass function: p(m1) ∝ m12.35

Basic Pop. Modeling: Assume p(m1) ∝ m1-α and fit observations

to model — flat in log is becoming disfavored, and lower

bound on rates may jump up

End of O1 After GW170104

After GW170104: 12 - 213 Gpc-3yr-1 Power law (α=-2.35) only: 40 - 213 Gpc-3yr-1

7

Dealing with Multiple Event Categories: Being unsure of the intrinsic source populations and origins, we calculate the event rates for distributions of events according to uniform in the logarithm of component mass and

according to the stellar initial mass function: p(m1) ∝ m12.35

After GW170104: 12 - 213 Gpc-3yr-1 Power law (α=-2.35) only: 40 - 213 Gpc-3yr-1

End of O1 After GW170104

Towards Pop. Distributions: Constraints on the primary mass

distribution from current observations (assumes power law

from previous slide)

8

high magnitude, aligned spins disfavored — hints of

anti-alignment, but difficult to measure spin magnitude

Precession: mostly consistent with prior assumptions

9

Not as much spin information here

Precession: distribution tilted further towards hints of precession

0

2

4

6

8

10

12

14

SNR

Hanford Livingston Virgo

16

32

64

128

256

Freq

uenc

y[H

z]

0.46 0.48 0.50 0.52 0.54 0.56

Time [s]

�1.0

�0.5

0.0

0.5

1.0

Whi

tene

dSt

rain

[10-

21]

0.46 0.48 0.50 0.52 0.54 0.56

Time [s]0.46 0.48 0.50 0.52 0.54 0.56

Time [s]

�5

0

5

�5

0

5

�2

0

2

�no

ise

0.00.51.01.52.02.53.03.54.04.55.0

Nor

mal

ized

Am

plitu

de

GW170814 (1709.09660) The first three detector network

detection H ~ 7.3 / L ~ 13.7 / V ~ 4.4

10

All spin posteriors in context: → Mass ratio usually means second spin is unconstrained → Most distributions have significant weight near zero spin

→ GW searches do not include precession effects (could lead to observational bias)

GW150914 LVT151012 GW151226

GW170104

11

Farr, et al. 2017 (Nature) Assume various simple distributions of spin magnitude. Use these as

proxies for low, high, and flat. Using existing observations, infer a virtual “population” giving rise to those distributions in effective spin.

FI: Flat mag., isotropic dir. FA: Flat mag. aligned dir.

HI: High mag, isotropic dir. HA: High mag, aligned dir LI: Low mag, isotropic dir LA: Low mag, aligned dir

Current detections: Bayes Factors (evidence for one model versus another) already disfavors high/flat mag. aligned

spin

12

Farr, et al. 2017 (Nature) Assume various simple distributions of spin magnitude. Use these as

proxies for low, high, and flat. Using existing observations, infer a virtual “population” giving rise to those distributions in effective spin.

FI: Flat mag., isotropic dir. FA: Flat mag. aligned dir.

HI: High mag, isotropic dir. HA: High mag, aligned dir LI: Low mag, isotropic dir LA: Low mag, aligned dir

Additional detections: After 10 additional detections from

each model

13

Zevin, et al. 2017 (submitted to ApJ) Similar statistical procedure, but use mass distributions to determine how many mass observations are required to confidently discern a given BH

natal kick prescription

Kick Prescription: Full: vBH = vNS

Fallback: vBH = (1-f) vNS Proportional: vBH = MNS/MBH vNS

β = fraction of field vs. clusters

Distinguishability: Cluster populations are very similar

across prescriptions, so β becomes easier to measure —

Kick disruption changes the mass population density and so some prescriptions (proportional) can

be distinguished easier than others even with only mass

14

Distinguishability: Cluster populations are very similar

across prescriptions, so β becomes easier to measure —

Kick disruption changes the mass population density and so some prescriptions (proportional) can

be distinguished easier than others even with only mass

Zevin, et al. 2017 (submitted to ApJ) Similar statistical procedure, but use mass distributions to determine how many mass observations are required to confidently discern a given BH

natal kick prescription

Kick Prescription: Full: vBH = vNS

Fallback: vBH = (1-f) vNS Proportional: vBH = MNS/MBH vNS

15

Astrophysics Implications16

Ap. JL. L22 2016

Environment Metallicity: GW150914 / GW170104 — must have

been formed in Z < 0.5 Zsolar — what are the implications for delay time?

A Stochastic “Foreground”?: If rates are on the high end, BBH could

form confusion background which is detectable by advanced detectors

Environment Metallicity: GW150914 / GW170104 — must have

been formed in Z < 0.5 Zsolar — what are the implications for delay time?

Astrophysics Implications17

Ap. JL. L22 2016

Stochastic Foreground?: If rates are on the high end, BBH could

form confusion background which is detectable by advanced detectors

O1 IMBHB Upper Limits18

https://arxiv.org/abs/1704.04628

Only sensitive to near solar mass IMBH, future

low frequency improvements are limited by finances and >1000 Msolar likely out of reach before third generation.

Maximum reach at z ~0.5, average of z ~0.25

Minimum rate is within “a factor of a few” of a nominal rate of 0.1

GC-1yr-1

Part 2: Astrophysics with Neutron Stars

O1 BNS / NSBH Upper Limits20

Compact Sources: Only BBH detections so far, NSBH and BNS remain elusive, but expected to constrain

models in the next few years

https://arxiv.org/abs/1607.07456201520172018 201520172018

Ozel & Freire (2016)

Maximum (reliably) measured NS mass

~2.01 +/- 0.04

21

Binary dynamics is affected by the EoS

through the tidal deformation of

neutron stars

…but is highly subdominant… enters into the

analytical phase expression used by searches and “PE”

at O((v/c)10)

This is subordinate to the masses, mass

ratio, spins, spin interactions, …

Hinderer, et al. (2009)

k2: Tidal Love number — encodes the rigidity of the star

22

The presence of matter can potentially drastically change the near merger and post merger effects

BH: “quasi-normal” ringdown modes

PMNS: EoS dependent post-

merger oscillations

Read, et al. (2009)

merger rem

nant post-m

erger oscillations increasing com

pactness

collapse to black

hole

23

Clark, et al. (2014) Reconstructing Post-merger physics with generic burst

detection algorithms

24

Clark, et al. (2014) Reconstructing Post-merger physics with generic burst

detection algorithms

25

1. Pre-merger: measure Mtot

2. Post-merger: measure fpeak

3. Constrain allowed EOS

fpeak

Mtot

= a ·R2

1.6 + b ·R1.6 + c

!f-mode

⇠r

M

R

Clark, et al. (2015) Measuring frequencies of oscillations in post merger NS

1508.05493

Clark, et al. (2014) Reconstructing Post-merger physics

with generic burst algorithms SFHo EoS

27

Clark, et al. (2014) Reconstructing Post-merger physics

with generic burst algorithms Shen EoS

28

color scale is density

Clark, et al. (2015) Reconstructing Post-merger

physics

nonlinear coupling between oscillation modes

fpeak

fspiral

What is the intersection of astrophysics and neutron

star interiors?

Metzger and Berger (2012)

Post-merger content and EM emission

sensitive to the configuration and size of

NS(es)

30

Metzger (2017)

31

compactness: C = GM/Rc2 M*: baryonic

mass of the NS a, b, c, d: fit coefficients

Dietrich & Ujevic (2017)

Dynamically ejected matter is sensitive to the EoS!32

The Next ~5 Years33

O1, H1/L1 (2015): 60-80 Mpc

The Next ~5 Years34

O2, H1/L1 (2016-2017): 60-90 Mpc (top of green)

The Next ~5 Years35

O2 (2017): ~25 Mpc (top of cyan)

The Next ~5 Years36

O3+, H1/L1/V1 (2018?): H1/L1 100+ Mpc? V1 30-50 Mpc?

The Next ~5 Years37

→ 4 and 5 detector networks by mid 2020s

→ 2019 LIGO to meet design sensitivity (blue curve)

→ iKagra online before 2020 → LIGO India tentatively for

2024

Concluding Remarks38

• O2 has provided two additional BBH detections

• GW170104, relatively heavy BBH at z ~= 0.2

• GW170814, first three instrument detection — critical improvements in sky localization realized

• O1 upper limits on BNS and NSBH are reaching into the observational and model space

• Future detections of binaries with neutron star components may constrain the EoS via tidal parameters in the GW waveform

• Post merger content is also important, but is weaker and harder to detector — definitive detection of PM frequencies would allow for constraints on the EoS as well as physical post-merger oscillations

• Astrophysics through joint electromagnetic and gravitational observations should provide independent and correlating information about post merger activity and connect back to the arrangement and density of matter inside component NS

Extra Slides

dL (Mpc)

0.0

0.2

0.4

0.6

0.8

1.0

Cum

ulat

ive

Frac

tion

✓JN (rad)

020

040

060

080

010

000.0

0.2

0.4

0.6

0.8

Cum

ulat

ive

Frac

tion

0.5

1.0

1.5

2.0

2.5

Pankow, et al. 2016 (ApJ) Examine what benefit, if any, constraining information from

electromagnetic observations would have on putative gravitational wave event parameter estimation

Source orientation: Breaking distance / inclination degeneracy does allow for noticeable

improvement in either given the other

Mc (M�)

0.0

0.2

0.4

0.6

0.8

1.0

Cum

ulat

ive

Frac

tion

q a1 ✓1 (rad)

10�2

10�1

100

101

102

0.0

0.2

0.4

0.6

0.8

Cum

ulat

ive

Frac

tion

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.5

1.0

1.5

2.0

2.5

3.0

Pankow, et al. 2016 (ApJ) Examine what benefit, if any, constraining information from

electromagnetic observations would have on putative gravitational wave event parameter estimation

Physical params: Since the orientation and physical parameters are “separable” in their response in the instrument, we get very modest improvement in only spin determination

Alternative Hypothesis (H1): data are distributed as in null, after subtraction of the

signal model (h)

Null Hypothesis (H0): Data samples are uncorrelated Gaussian noise with

variance proportional to S(f)

Basic Terminology42

observations: Putative strain from gravitational wave is embedded in detector noise

Noise power spectrum: Autocorrelation of the noise in the frequency domain — “limiting factor” of the sensitivity of the

instrument

Noise weighted inner product: frequency-domain cross-correlation between two quantities

Inferred Rates / Probability of Astrophysical Origin43

arxiv:1302.5341

Likelihood of obtaining ensemble of ranking statistics xi with two categories of events: background (terrestrial) and

foreground (astrophysical) Λfg,bg ~ expected counts from each

category pfg, pbg - modeled or measured, for

astrophysical distribution of binaries pfg ~ ρ-4

Inferred Rates / Probability of Astrophysical Origin44

Obtain probability of astrophysical origin by

marginalizing against the counts

Model 2 ranking stat. Model 1 ranking stat.

LVT151012 ~ 87% probable foreground

GW150914 suppressed since > 99% probable and far to the

right

Mass: Distributions cover roughly similar intervals — too few detections to infer which channel contributes what in this space

Isolated Galactic Field: Binaries are born and evolve together (mass transfer and

double SN)

Dynamical: Welcome to the “mosh pit”: interacting compact objects lead to light component ejection and 3-body hardening

Globular clusters (z<0.1): < 20 Gpc-3yr-1 — Rodriguez, et al. 2016

Globular clusters (local): 5 Gpc-3yr-1 — Askar, et al. 2016

Galactic Nuclei (z<0.3)— 5 Gpc-3yr-1 Antonini & Rasio, 2016

AGN (local)— 5 Gpc-3yr-1 Bartos et al., 2016

StarTrack (z<0.1): 200 Gpc-3yr-1 — Belczynski, et al. 2016

COMPAS (extrapolation): 300 Gpc-3yr-1 — Stevenson, et al.

2017Chemically homogenous

(z~0.1-0.5): 10-20 Gpc-3yr-1 — Mandel & de Mink, 2016

Astrophysics Implications46

Ap. JL. L22 2016