EMRI Dark Siren H₀ — Interactive Figures
Jasper Seehofer — Master Thesis, Heidelberg University
This page provides browser-explorable versions of key figures from the EMRI dark siren H₀ inference pipeline. Each figure is fully interactive: hover over data points for exact values, toggle legend entries to isolate traces, and zoom or pan to explore regions of interest.
- Combined H₀ Posterior Joint posterior over the dimensionless Hubble parameter h from all detected EMRI events, with 68% and 95% credible intervals and Planck / SH0ES reference bands.
- Sky Localization Map Mollweide projection of detected EMRI event sky positions, colour-coded by signal-to-noise ratio. Hover for per-event SNR, redshift, and luminosity distance.
- Fisher Matrix Error Ellipses Cramer–Rao parameter uncertainty ellipses for selected pairs of EMRI parameters. 1σ and 2σ contours shown per event; zoom in to compare event-to-event spread.
- H₀ Posterior Convergence Two-panel diagnostic: combined posterior sharpening as more events are included (left) and credible-interval width vs event count with a 1/√N reference (right).
- Mz Improvement Explorer Three-panel interactive answer to "did adding the redshifted MBH mass channel Mz tighten the H₀ posterior?" — bootstrap-banded metric curves (HDI width, relative precision, fractional improvement, effective event gain) vs. number of detections, with a per-N posterior overlay and a live improvement summary strip.
- Single-Event Detail Explorer For one EMRI detection at a time, six interactive panels show all candidate host galaxies with and without the BH-mass cut: per-galaxy weights, side-by-side comparison, and the resulting per-event L(h) curves. Use the dropdown to switch between representative events.
- Closure Test Overlay Combined H₀ posteriors from independent runs with htrue ∈ {0.60, 0.65, 0.70, 0.73, 0.75} overlaid on a single axis — each posterior should peak near its own injection truth. Demonstrates pipeline closure across injection cosmologies.
- Catalog Coverage & Host Counts GLADE+ catalog coverage vs luminosity distance plus per-event candidate-host counts. Visualises how the BH-mass cut reduces the host pool (median reduction ≈ 64 % across the 1473-event campaign).