Toward Low-Latency, High-Fidelity Calibration of the LIGO Detectors with Enhanced Monitoring Tools

TL;DR

The paper addresses the need for low-latency, high-fidelity calibration of LIGO detectors to support rapid gravitational-wave science. It details the evolution of calibration practices, the DARM loop modeling framework, and the use of absolute references (including Pcals and Ncal) to reconstruct accurate strain measurements, with Gaussian Process Regression used to quantify systematic errors. A centerpiece is the O4 deployment of a low-latency calibration pipeline (~3 s latency) and the CalMonitor real-time calibration-monitoring system, along with spectral-line subtraction and improved reliability, enabling near real-time scientific analyses and uncertainty quantification. The work charts a path toward real-time calibration for O5 and next-generation detectors, highlighting hardware upgrades, time-dependent sensing/actuation modeling, and the necessity of robust monitoring and error estimation to support rapid multi-messenger astronomy and stringent tests of fundamental physics.

Abstract

Accurate and reliable calibration of the Advanced LIGO detectors has enabled a plethora of gravitational-wave discoveries in the detectors' first decade of operation, starting with the ground-breaking discovery, GW150914. In the first decade of operation, the calibrated strain data from Advanced LIGO detectors has become available at a lower latency and with more reliability. In this paper, we discuss the relevant history of Advanced LIGO calibration and introduce new tools that have been developed to enable faster and more robust calibrated strain data products in the fourth observing run (O4). We discuss improvements to the robustness, reliability, and accuracy of the low-latency calibration pipeline as well as the development of a new tool for monitoring the LIGO detector calibration in real time.

Figures (6)

• Figure 1: Schematic visualizing the evolution of several aspects of LIGO calibration from O1 through O4. This graphic focuses on five specific components of the LIGO calibration procedure which are discussed primarily in section \ref{['sec:currentwork']}.
• Figure 2: An example of line subtraction from data in the LIGO Hanford detector. This figure shows the amplitude spectral density (ASD) zoomed in on the 17.1 Hz Pcal line. The data with no line subtraction applied is shown in red. The data with line subtraction performed using a static 128 second running median method is shown in blue. The data with line subtraction performed using a static 4096 second running median method is shown in orange. The adaptive switching method that dynamically changes the length of the running median based on whether a secular change is detected in the $T_j$ before taking the median is shown in green. The adaptive switching method yields the best results and is the method implemtened during O4.
• Figure 3: Simplified workflow for the data products associated with CalMonitor. The red containers indicate timeseries data streams that are required as inputs to CalMonitor. The data products produced by CalMonitor are broadcast using the Kafka event streaming platform and then aggregated and stored in an InfluxDB backend using the ligo-scald software package. The details of these steps are omitted from this summary diagram, but the end result of calibration metric aggregation in a database is represented by the green cylinder. The Grafana web application is used to query InfluxDB and display calibration metrics in a real-time, user-friendly interface.
• Figure 4: Time series of the magnitude and phase of calibration systematic error as measured by the monitoring lines (see eq:calsyserror) in the LIGO Hanford detector. The top plots show the magnitude of the calibration systematic error and the bottom plots show the phase. The plots on the left are for a time period when the detector was fully thermalized and no thermalization effects were measured in the detector sensing function. The plots on the right are for a time when the detector was actively thermalizing and these effects are measured in the low-frequency monitoring lines as a relatively large systematic error in the calibration.
• Figure 5: Violin bode plots of the magnitude and phase of calibration systematic error as measured by the monitoring lines (see eq:calsyserror) at the LIGO Hanford detector over the same time period as figure \ref{['fig:monplots_time']}. The top plots show the systematic error in the calibration when the detector was fully thermalized and no thermalization effects were measured in the detector sensing function. The bottom plots show the systematic error for a time when the detector was actively thermalizing. The larger spread in systematic error at the lower frequency monitoring lines is a result of the larger systematic error in the sensing function model during thermalization. The frequencies of the monitoring lines are shown along the top of each bode plot. The box on each violin plot extends from the first to third quartile. The black line shows the median value. The whiskers stretch from the box to the most distant data point that falls within 1.5 times the interquartile range of the box. Any points beyond the whiskers are considered outliers.