Orateur
Description
We present tfwaves, a Python package that represents long-lived gravitational-wave signals directly in the time-frequency domain. From the amplitude and phase of a time-domain waveform, it computes an analytical approximation of the Short-Time Fourier Transform, together with an approximate LISA response. The resulting representation is sparse: each harmonic occupies only a few frequency bins per time segment. Two consequences follow. First, non-stationarity is taken into account naturally, since the noise is described locally in each time-frequency pixel rather than by a single stationary spectrum over the whole observation. Second, the likelihood becomes cheap enough that parameter estimation with MCMC methods is practical for signals that stay in the LISA band for months to years, such as those from extreme mass-ratio inspirals (EMRIs) and stellar-mass black hole binaries.
We focus on the EMRI case. The analytical time-frequency waveform and response reproduce the numerically computed STFT of FEW waveforms passed through the full LISA response, with overlaps greater than 95% for the EMRI sources of the Mojito Lite simulated dataset, and its accuracy depends on the time-frequency resolution and on the set of retained harmonics, together with injection-recovery MCMC results on those sources. Results obtained from the frequency-domain framework of the package for stellar-mass black hole binaries will be shown briefly for comparison.
| Which working group does your abstract concern? | Méthodes d'analyse des données / Data analysis methods |
|---|