LVK O4a Data: Gravitational-Wave Breakthroughs

• LVK O4a Data is a comprehensive dataset from the fourth gravitational-wave observing run, featuring enhanced detector sensitivity and optimized analysis pipelines.
• The dataset employs advanced calibration, noise gating, and multiple pipelines like PyCBC and GSTLAL to identify compact binary mergers with high signal-to-noise ratios.
• Open data releases via GWOSC empower multi-messenger and population astrophysics studies, enabling refined cosmological measurements and insights on standard sirens.

The LIGO-Virgo-KAGRA (LVK) O4a data set encompasses the first half of the fourth gravitational-wave observing run, covering 24 May 2023 through 16 January 2024. The campaign featured fundamentally improved detector sensitivity, newly optimized analysis methodologies, and an unprecedented volume of open data releases. The O4a epoch resulted in a significant expansion of confidently identified compact-object merger candidates, refined upper limits across all classes of gravitational-wave (GW) transients, and a sharp increase in multi-messenger and population-astrophysics reach.

1. Observing Run Configuration and Data Quality

O4a operations were dominated by the two LIGO 4 km interferometers: Hanford (H1) and Livingston (L1). Virgo was offline for O4a commissioning; KAGRA joined at low sensitivity only briefly and did not contribute to the main data set. The formal O4a observing span was 237 days, with H1 and L1 duty cycles of 67.5 % and 69.0 %, yielding a network coincident live time of ≈120 days. The calibrated strain data, released via the Gravitational Wave Open Science Center (GWOSC), employed the C00 (v1) pipeline with time- and frequency-domain calibration errors of ≲2 % amplitude and ≲2° phase below 2 kHz. Hourly photon-calibrator line injections tracked systematic response variations (Collaboration et al., 25 Aug 2025).

A curated set of ≈40 auxiliary channels per detector and science-mode data-quality flags (CAT1 for compact-binary coalescence, CAT2 for bursts, etc.) were included. Glitch removal was enforced by time-domain gating on identified short-duration transients, and by subsequent auxiliary-channel–based vetoes. Noise transients (“glitches”) were exhaustively characterized: at LLO (Livingston), the dominant groupings were seasonal microseism-driven artifacts (“12 Hz arches,” “20 Hz arches”), correlating with ground motion (Pearson r up to 0.64). At LHO (Hanford), the leading populations were broadband 20–50 Hz events, tightly linked to electrostatic-drive feedback, exhibiting two-state rate behavior modulated by control configuration (Ferreira et al., 3 Dec 2025). These insights enabled targeted gating and vetoing strategies.

2. Compact Binary Coalescence Detections and the GWTC-4.0 Catalog

O4a, together with the preceding engineering run, produced 128 new GW candidates with astrophysical probability $p_{\rm astro} \geq 0.5$ not vetoed during event validation, as collated in GWTC-4.0. Of these, 86 surpassed false-alarm rate (FAR) $<1$ yr$^{-1}$ and had detailed Bayesian parameter estimation. The offline analyses employed four independent pipelines: CWB-BBH (coherent burst search for BBH), GSTLAL, MBTA, and PyCBC (matched-filter searches). The catalog features 84 BBH mergers, two NSBH systems, and expands the population of well-measured events across the parameter space: BBH total masses span $14–236\,M_\odot$, component spins are typically consistent with $\chi_{\rm eff}\approx 0$ but with minority positive and negative outliers, and signal-to-noise ratios (SNR) exceed 30 for the first time (notably GW230814_230901, SNR = 42.1) (Collaboration et al., 25 Aug 2025).

Event selection combined FAR and $p_{\rm astro}$ criteria. Bayesian parameter estimation leveraged PE samples made available through GWOSC, supporting population synthesis, cosmological inference, and multi-messenger follow-up. A representative BNS candidate from a redshift-corrected, population-weighted search (GW231109_235456) achieved SNR = 9.7, FAR ≈ 1/50 yr (1/10 yr trials-adjusted), and demonstrated the utility of targeted banks for revealing sub-threshold events (Niu et al., 11 Sep 2025).

3. Sensitivity Improvements and Injection Campaigns

Detector noise during O4a was reduced over O3a, particularly in the 100–400 Hz band (multi-IFO ASD ≃ 3×10$^{-25}$ /√Hz at 200 Hz). Massive Monte Carlo injection campaigns were conducted, drawing $>4.3\times10^8$ simulated events covering source-frame masses $1–1000\,M_\odot$, spins, redshift $z<3$, etc., and streamed through to detection pipelines (Essick et al., 14 Aug 2025). Detection efficiency for arbitrary populations was estimated via importance sampling, using:

$$\epsilon(\Lambda) = \frac{1}{N_{\rm inj}} \sum_{e \in {\rm found}} \frac{p(\theta_e|\Lambda)}{p(\theta_e|\Lambda_{\rm inj})}$$

Sensitive volumes, effective ranges, and selection-function biases were quantified as a function of mass, redshift, and sky position. Subtle effects, such as diurnal sensitivity modulation (∼20 % amplitude; night-time maxima), and sky-location selection biases were characterized explicitly.

MBTA's O4a pipeline enhancements included an expanded template bank (total mass up to 500 $M_\odot$), two-band filtering to reduce computational cost, improved gating and $\chi^2$-based glitch rejection, single-detector analysis formalism, and source-type–dependent trials factoring. These improvements reduced low-latency retractions and boosted overall detection efficiency (Alléné et al., 8 Jan 2025).

4. Searches for Non-CBC Signals: Bursts, Long-Duration Transients, and Continuous Waves

Burst searches (for transient signals up to a few seconds, 16–4096 Hz) utilized Coherent WaveBurst (cWB) and other pipelines, with machine-learning post-processing. Four analysis pipelines (2G+XGB, 2G+GMM, XP+XGB), leveraging XGBoost and GMM classifiers, found no statistically significant non-CBC candidates (highest non-CBC IFAR = 1.75 yr < detection threshold of 100 yr). Sensitivity for short ad-hoc waveforms at 50 % efficiency, $h_{\rm rss}^{50\,\%}$, improved by $2–10\times$ over O3 depending on morphology and frequency; e.g., for $f_0=235$ Hz, $Q=9$ sine-Gaussians, $$h_{\rm rss}^{50\,\%}\approx (0.6–0.8)\times10^{-22}$$ Hz$^{-1/2}$ (Collaboration et al., 16 Jul 2025). Astrophysical reach now includes the detection of rapidly rotating core-collapse supernovae (CCSN) throughout the Milky Way for optimistic models.

Long-duration searches ($1–1000$ s) targeted unmodeled GW emission from, e.g., magnetars and eccentric coalescences. O4a delivered $\sim$30 % improved $h_{\rm rss}$ sensitivity over O3. For example, for $1.4+1.4\,M_\odot$ eccentric binaries ($e=0.2–0.6$), 90 % rate upper limits improved to $R_{90\,\%}=(9.1–18)\times10^3$ Gpc$^{-3}$ yr$^{-1}$ (Collaboration et al., 16 Jul 2025).

Continuous-wave searches (CW) leveraged semi-coherent GPU-accelerated “fasttracks” analysis of binary neutron star signals in the 100–350 Hz band, targeting orbital periods 7–15 d, $a_p=5–15$ light-sec. No credible astrophysical signals were found. The 95 % strain sensitivity depth $D^{95\%}$ at 200 Hz reached $23.3\pm0.6$ Hz$^{-1/2}$, corresponding to $h_0\simeq1.6\times10^{-25}$, a factor $\sim$1.5 better than O3a (Collaboration et al., 21 Nov 2025).

5. Stochastic Background Limits and Cosmological Constraints

Stochastic background analyses combined O4a with previous runs, applying cross-correlation estimators, Bayesian inference, and model-independent double-peaked spectral parameterizations (Miritescu et al., 1 Dec 2025, Collaboration et al., 30 Oct 2025). O4a delivered the tightest upper limits to date:

• Flat spectrum ($\alpha=0$): $\Omega_{\rm GW}(f)\le2.8\times10^{-9}$ (20–58.2 Hz).
• Compact-binary–distributed spectrum ($\alpha=2/3$): $$\Omega_{\rm GW}(25\ \mathrm{Hz})\le2.0\times10^{-9}$$ (20–86.8 Hz).

No evidence was found for double-peak or other nontrivial spectral signals. O4a enables exclusion of broad, shallow inter-peak valleys at high amplitude for double-peaked backgrounds.

Parameterized early-Universe backgrounds are strongly constrained: phase transitions at $T_*\gtrsim10^9$ GeV and $\beta/H_*\lesssim3$ (for $\alpha\gtrsim0.1$) are excluded; cosmic string tensions $G\mu\gtrsim10^{-15}$–$10^{-9}$ are ruled out depending on the loop model; domain wall annihilation at $T_{\rm ann}\ll10^9$ GeV is restricted; stiff equation-of-state epochs and parity-violating sources are tightened as well. Models for scalar-induced GWs and primordial black holes contributing $>1\%$ of dark matter for $\mu\lesssim30\,M_\odot$ are excluded (Collaboration et al., 30 Oct 2025).

6. Multi-Messenger and Cosmological Results

No significant neutrino counterparts were detected by Super-Kamiokande for 56 O4a GW triggers (mainly BBH/NSBH). The $90\%$ CL per-event fluence upper limits were set at $5.0\times10^7$ cm$^{-2}$ (Fermi–Dirac 5 MeV spectrum), $1.2\times10^7$ cm$^{-2}$ (Fermi–Dirac 8 MeV), and $2.0\times10^7$ cm$^{-2}$ (flat), with stacked limits improved by $\sim7.5\times$. Only very nearby ($D\ll10$ Mpc) BNS mergers would yield detectable low-energy neutrino fluence (Machado, 15 Sep 2025). These non-detections challenge optimistic neutrino-production scenarios and demonstrate world-leading fluence sensitivity in the MeV–TeV range.

A galaxy-catalogue–driven “dark siren” analysis combined five new O4a BBH events with 10 previous GW events, yielding $H_0=70.4^{+13.6}_{-11.7}$ km s$^{-1}$ Mpc$^{-1}$ ($68\%$ CI) (Bom et al., 24 Apr 2024). Inclusion of GW170817 (bright siren) and jet constraints tightens to $H_0=68.0^{+4.4}_{-3.8}$ km s$^{-1}$ Mpc$^{-1}$, achieving a $6\%$ standard-siren measurement and $10\%$ lower uncertainty than GW170817 alone.

7. Data Accessibility, Open Science, and Future Prospects

O4a data products—including calibrated strain (H1/L1, 16 or 4 kHz), DQ bitmasks, auxiliary regression channels, software pipelines, curated hardware-injection logs, and full GWTC-4 catalog releases—were made available via GWOSC (Collaboration et al., 25 Aug 2025). Tutorials and code examples (e.g., using gwpy, PyCBC, Bilby) facilitate external analysis. Quality masks (bitwise, 1 Hz) encode all relevant veto logic, and open-source notebooks support both standard and advanced analysis workflows.

As O4 continues into O4b and approaches O5, ongoing detector upgrades—real-time veto integration, noise reduction, expanded three- and four-detector networks—and continued population-targeted and machine-learning-informed search pipelines are projected to further improve sensitivity and up the sub-threshold discovery rate, particularly for rare populations and multi-messenger sources. Open data and reproducible computation remain foundational to LVK science, enabling rigorous tests of fundamental physics and further population astrophysics advances.