A pipeline for searching and fitting instrumental glitches in LISA data

TL;DR

This paper introduces a Bayesian pipeline using advanced MCMC techniques to detect, characterize, and differentiate instrumental glitches from astrophysical signals in LISA data, enhancing analysis accuracy.

Contribution

It presents a novel Bayesian method employing Reversible Jump MCMC and parallel tempering for glitch detection and characterization in LISA data.

Findings

Successfully identifies diverse glitch morphologies

Accurately infers parameters of Massive Black Hole Binaries

Demonstrates robustness on simulated LISA data

Abstract

Instrumental artefacts, such as glitches, can significantly compromise the scientific output of LISA. Our methodology employs advanced Bayesian techniques, including Reversible Jump Markov Chain Monte Carlo and parallel tempering to find and characterize glitches and astrophysical signals. The robustness of the pipeline is demonstrated through its ability to simultaneously handle diverse glitch morphologies and it is validated with a 'Spritz'-type data set from the LISA Data Challenge. Our approach enables accurate inference on Massive Black Hole Binaries, while simultaneously characterizing both instrumental artefacts and noise. These results present a significant development in strategies for differentiating between instrumental noise and astrophysical signals, which will ultimately improve the accuracy and reliability of source population analyses with LISA.