Interactive visualization of a laser interferometer gravitational wave detector: watch how spacetime ripples from merging black holes stretch and squeeze the detector arms, producing an interference signal
Gravitational waves are ripples in spacetime produced by accelerating massive objects, predicted by Einstein's General Relativity in 1916. The strongest sources are compact binary mergers. The wave has two polarization modes: h+ stretches one axis while compressing the perpendicular axis; h× does the same rotated 45°. The strain h = ΔL/L is typically ~10⁻²¹ for astrophysical sources. On September 14, 2015, LIGO detected the first gravitational wave signal (GW150914) from two black holes merging ~1.3 billion light-years away.
LIGO uses a Michelson interferometer with 4 km arms. A laser beam is split: half travels along each arm and returns. Without gravitational waves, destructive interference produces a dark fringe. When a GW passes, the differential arm-length change is ΔL = h × L. With L=4km and h~10⁻²¹, ΔL~4×10⁻¹⁸m — about 1/1000 the proton diameter! Fabry-Pérot cavities effectively increase arm length to ~1200km. In this teaching model, the lower panel shows a demodulated readout with additive noise and simple smoothing, rather than a full matched-filter pipeline.
Multi-messenger astronomy: GW170817 detected simultaneously in GW and EM, confirming neutron star mergers as the origin of heavy elements. Black hole physics: confirmed stellar-mass black hole binaries. Tests of General Relativity. Cosmology: standard sirens for Hubble constant. 2017 Nobel Prize to Weiss, Barish, and Thorne.
The top canvas shows a Michelson interferometer. Watch the arms alternately stretch (red) and compress (blue). The middle canvas shows h(t). The bottom shows a demodulated detector readout with optional noise. Toggle Filtered Signal to show a simple smoothed trace of that readout. Use Chirp to see a binary inspiral. Merger shows inspiral + merger + ringdown. Noise Only demonstrates the detector without a signal. The SNR shown here is only a simple toy-model amplitude-to-noise estimate; real searches use more sophisticated matched filtering.