Robust Bayesian inference with gapped LISA data using all-in-one TDI- ∞

Abstract

The Laser Interferometer Space Antenna (LISA), an European Space Agency L-class mission, is designed to detect gravitational waves in the millihertz frequency band, with operations expected to begin in the next decade. LISA will enable groundbreaking studies of astrophysical phenomena such as massive black hole mergers, extreme mass ratio inspirals, and compact binary systems. A key challenge in analyzing LISA’s data is the significant laser frequency noise, which must be suppressed using time-delay interferometry (TDI) during on-ground processing. Classical TDI mitigates this noise by algebraically combining phase measurements taken at different times and spacecraft. However, data gaps caused by instrumental issues or operational interruptions complicate the process. These gaps affect multiple TDI samples due to the time delays inherent to the algorithm, rendering surrounding measurements unusable for parameter inference and degrading overall data quality. In this paper, we apply the recently proposed variant of TDI known as TDI- to astrophysical parameter inference, focusing on the challenge posed by data gaps. TDI- frames the LISA likelihood numerically in terms of raw measurements, marginalizing over laser phase noises under the assumption of infinite noise variance. Additionally, TDI- is set up to incorporate and cancel other noise sources beyond laser noise, including optical bench motion, clock noise, and modulation noise, establishing it as an all-in-one TDI solution. The method gracefully handles measurement interruptions, removing the need to explicitly address discontinuities during gravitational-wave template matching. We integrate TDI- into a Bayesian framework, demonstrating its superior performance in scenarios involving data gaps. Compared to classical TDI, the method preserves signal integrity more effectively and is particularly interesting for low-latency applications, where the limited amount of available data makes data gaps particularly disruptive. The study’s results highlight the potential of TDI- to enhance LISA’s scientific capabilities, paving the way for more robust data analysis pipelines.