The first year of LISA Galactic foreground
Riccardo Buscicchio, Federico Pozzoli, Daniele Chirico, Alberto Sesana · 2025
Read The first year of LISA Galactic foreground by Riccardo Buscicchio, Federico Pozzoli, Daniele Chirico, Alberto Sesana (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-03604.txt.
Introduction
Riccardo Buscicchio and colleagues rehearse how the space antenna LISA will untangle the hum of countless white-dwarf pairs in our galaxy. That galactic foreground is both signal and noise for LISA's hunt for massive black hole mergers. A global-fit code named bahamas rebuilds the foreground from a simulated first year of data.
Data are handled in a time-frequency form, and several likelihood, sampling, and segmentation choices are compared. Evidence for a non-stationary foreground model becomes overwhelming within a month of data. The rebuild stays robust when extra foregrounds from beyond the galaxy are added. Early rehearsal matters because foreground mistakes could bias every other source LISA hopes to measure.
Catalog record
| CREATOR | Riccardo Buscicchio, Federico Pozzoli, Daniele Chirico, Alberto Sesana |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-03604
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The first year of LISA Galactic foreground Preprint: APS/123-QED Riccardo Buscicchio, Federico Pozzoli, Daniele Chirico, Alberto Sesana Abstract Galactic white-dwarf binaries play a central role in the inference model for the Laser Interferometer Space Antenna. In this manuscript, we employ the bahamas codebase to characterize, in a global-fit fashion, the reconstruction of the Galactic foreground during the first year of observation. To account for its statistical properties, we represent the data in time–frequency domain, and characterize the effectiveness of multiple approaches, e.g. statistically viable likelihoods, sampling schemes, segmentation widths, and gaps density. Our analysis yields consistent results across, with overwhelming evidence in favor of a non-stationary model in less than a month of data. Moreover, we show robustness against the presence of additional extragalactic foregrounds, and test the suitability of our approximations on the more complex simulated data in the Yorsh data challenge. I Introduction The Laser Interferometer Space Antenna (LISA) Amaro-Seoane et al. 2017 is a pioneering space-based observatory designed to detect gravitational waves (GWs) in the millihertz frequency band. Over its mission lifetime, LISA is expected to observe a wide variety of astrophysical systems, ranging from mergers of massive black hole binaries (MBHBs) across the cosmic history to inspirals of Galactic, stellar-origin compact objects Colpi and others. 2024 . LISA will resolve thousands nearly monochromatic double white dwarfs (DWDs), offering insights into the Galactic population of compact binaries. The LISA datastream will be signal-dominated, with numerous GW sources persistently emitting in its frequency band. This motivates the development of so-called global fits Littenberg and Cornish 2023 ; Katz et al. 2025 ; Deng et al. 2025 ; Strub et al. 2024 . The majority of Galactic WD binaries will remain unresolved due to source confusion, hence piling up in an incoherent Galactic foreground (GF) Nelemans 2009 , dominant over the instrumental noise at frequencies \approx 0.5 – 3\mathrm{mHz} . Recent studies have highlighted expectations for an extragalactic foreground (EF) Perego et al. 2025 ; Boileau et al. 2025 ; Hofman and Nelemans 2024 ; Staelens and Nelemans 2024 , generated by the (entirely unresolved) extra-Galactic DWD population. By contrast to the EF, the GF originates primarily from sources concentrated toward the Galactic center in a strongly anisotropic distribution. Therefore, due to LISA orbital motion around the Sun, a time-dependent, quasi-periodic amplitude modulation —often referred to as cyclostationarity — is induced on the observed stochastic signal. Modelling and inferring on it (i) adds discriminating power with respect to the instrumental noise, and (ii) provides inference capability on the MW structure and morphology, where electromagnetic observations are dominated by dust extinction. In addition, non-Gaussianity arising from the Poisson nature of the GF at high frequencies Buscicchio et al. 2025 and spurious global fit residuals Rosati and Littenberg 2024 , may allow for additional discriminating power as shown with heavy-tailed likelihood models in literature Karnesis et al. 2025 . In previous work Pozzoli et al. 2025a , we proposed a full frequency-domain approach to account for cyclostationarities. We did so by modeling off-diagonal terms in the stochastic process covariance matrix. We applied the method to study the detectability of backgrounds from MW satellites Rieck et al. 2024 . Our study highlighted a key limitation: at least close-to-one year of LISA data had to be available to reach sufficient frequency resolution to infer on the target spectral cross-correlations. Global-fit analyses and transients detection in low latency (e.g. massive black-hole binaries) may not be compatible with such requirement: rapid noise inference on time-segmented data, potentially accomodating for the presence of gaps, are essential. In this work, we extend our analysis to the time–frequency domain, representing data segments as short-time Fourier transforms (STFTs). Approaches have been already proposed in literature, suited to a time–frequency GF representation. Some of them rely on wavelets decomposition Digman and Cornish 2022 , some other focus on astrophysically-motivated templates Criswell et al. 2025a , pixelation Hindmarsh et al. 2025 , or spherical harmonic decomposition Criswell et al. 2025b ; Banagiri et al. 2021 to describe sources anisotropy and the induced GF modulation. We instead turn our attention to quantifying the evidence in favour of a cyclostationary GF during the first year of LISA operations, and to assess the impact of data segmentation, scheduled and unscheduled gaps, likelihood models, stochastic samplers, evidence estimators, and the presence of an EF component. In addition, we test our approximations against a more realistic dataset, the Yorsh data challenge LDC , containing a full simulation of LISA instrumental noise and a Galactic population of DWDs. Individual studies are performed through a series of inferences with the publicly available code bahamas Pozzoli et al. 2025b , adapting the GF modulation described in Buscicchio et al. 2025 . The paper is organized as follows. In Section II.1 , we first introduce the basic conventions adopted in our analysis, then describe the time-frequency representation, as implemented in bahamas ,in Section II.2 . In Section III we present our results. First, in Section III.1 we consider a simulated GF in the presence of instrumental noise, only. Then, we quantify the impact of data gaps and of an additional EF in Sections III.2 and III.3 , respectively. Finally, we apply our methodology to Yorsh in Section III.4 . We conclude and outline future prospects in Section IV . II Data model II.1 Conventions for LISA We perform our analysis assuming time-delay interferometric (TDI) data. TDI is a post-processing technique that suppresses laser frequency noise by combining delayed single-link measurements from LISA Tinto and Dhurandhar 2005 . We adopt the noise-uncorrelated A,E, and T channels and for simplicity we assume analytical, Keplerian orbits yielding equal, constant light-travel times between spacecrafts. Consequently, their motion is fully determined by the initial orbital parameters: \alpha_{0} , which sets the initial phase of the LISA barycenter, and \beta_{0} , defining the initial rotation in the constellation plane (i.e., the orientation of the triangular configuration in the ecliptic at t=0 ). In the following analyses, we set \alpha_{0},\beta_{0} to zero. We use first-generation TDIs hence safely neglect cross-correlation given the assumed orbits. Introducing additional complexity in the LISA constellation dynamics, such as armlength breathing or unequal link noises, introduces further non-stationarities that have not yet been fully explored in literature, although are expected to be subdominant with respect to the GF Kume et al. 2025 . We model the GF envelope following Buscicchio et al. 2025 ; Pozzoli et al. 2025a , and compute it under the low-frequency approximation by averaging the LISA response over the MW sky distribution, modelled as a two-dimensional Gaussian. Our approach yields an analytical model extremely fast to evaluate, thus allowing for a large number of inferences. II.2 STFT representation To capture the time-varying frequency content of the cyclostationary GF signal we divide the full data stream into N_{\mathrm{chunk}} non-overlapping segments and perform an STFT on each. The time-frequency representation let us define a likelihood that accounts for the signal and noise stationarity, locally in each chunk. The total log-likelihood is expressed as the sum of the log-likelihoods for each chunk: \displaystyle\log\mathcal{L}(\bm{{d}}|\theta)=\sum_{c=1}^{N_{\mathrm{c}}}\log\mathcal{L}_{W,G}^{\mathrm{chunk}}(\bm{{d}}^{c}|\theta), (1) where \bm{d}^{c} denotes the data segment corresponding to the c- th chunk, and {\cal L}_{W,G}^{\rm chunk} represent the two likelihood models considered for each single chunk, conditioned