Presets
Playback
Binary
Playback and template
Layers
Model check
The science
What is modelled
Two compact objects (black holes or neutron stars) spiral together and radiate gravitational waves. At leading (quadrupole, Newtonian-orbit) order the wave frequency obeys f(τ) = (1/π) (5/(256 τ))3/8 (G Mc/c3)-5/8, where τ is the time left before the merger and the chirp mass is Mc = (m1m2)3/5/(m1+m2)1/5. So f ∝ τ-3/8: the pitch climbs faster and faster. The amplitude grows as h = (4/r)(G Mc/c2)5/3(π f/c)2/3 and the phase is Φ(τ) = −2 (5 G Mc/c3)-5/8 τ5/8.
The detector sees h = F+ h+ + F× h× with h+ ∝ (1+cos2ι)/2 cosΦ and h× ∝ cosι sinΦ (ι = inclination, F from the polarisation angle for a source overhead).
Interferometer readout
A wave of strain h makes one arm longer and the other shorter by ΔL = h L/2 each (L = 4 km), so the arm difference is h L. Light bouncing about N≈280 times in each arm cavity (wavelength 1064 nm) gains the phase Δφ = (4π/λ) N h L. For h = 10-21 this is about 10-8 rad.
Matched filter
The data is d = h + n with Gaussian noise of one-sided power spectrum S(f). The template is correlated with the data in the frequency domain, z(t) = 4 Re ∫ d̃(f) h̃*(f)/S(f) e2πift df, and divided by the template norm, ρopt2 = 4∫|h̃|2/S df. If the template is right, the peak of this SNR(t) equals ρopt (plus noise of unit variance), and the peak time gives the merger time. Changing the template chirp mass shows how much SNR is lost to a wrong guess.
Honest simplifications
- The inspiral is leading-order post-Newtonian. Real analyses use higher orders, spins and full numerical relativity; the last cycles here are schematic.
- Merger and ringdown are glued on at the innermost stable circular orbit: a damped sinusoid at
fring ≈ 0.0845 c3/(G M)(final spin 0.69 assumed) with a schematic amplitude boost of 2.6. Real peak amplitudes come from numerical simulations. - The noise spectrum is an illustrative analytic fit (about 5×10-24 / √Hz near 200 Hz), not a measured curve. Noise is Gaussian and stationary; real detectors have glitches. The chirp is shown only for the last 2.6 s (or from 20 Hz), at a sampling rate of 4096 Hz.
- Only one detector and a source straight overhead are modelled, so no sky localisation. The ripple picture is the quadrupole pattern with arbitrary visual wave speed; sizes and speeds are not to scale. Arm motion is drawn enormously exaggerated.
Numerical checks on this page
Chirp mass of a published pair; phase from numerical integration of 2πf dt against the closed form; the −3/8 slope by finite differences; the ISCO frequency against the textbook rule f = 4397 Hz/(M/M☉); and a noise-free matched filter whose peak equals the independently summed optimal SNR, lands at the injected time, and halves when the distance doubles.