ArchGEM: an Advanced Data Analysis Tool for Analyzing Scattered Light Noise in LIGO

Author(s)

McGowan, Kaylah, Nichols, Shania, Soni, Siddharth, Chatterjee, Chayan, Gonzalez, Gabriela, Holley-Bockelmann, Kelly, Jani, Karan

Abstract

Scattered light is one of the most common sources of non-stationary noise at low frequencies in Advanced LIGO detectors. It appears as arch-like features in time-frequency spectrograms, produced when stray light reflects from moving surfaces and recombines with the main interferometer beam. In this study, we present ArchGEM, an automated framework for identifying and characterizing these arches and recovering the physical properties of the scattering surfaces. ArchGEM combines a prominence-based peak-finding method with a Gaussian Mixture Model clustering approach to capture a range of scattered-light morphologies across different detector conditions. We apply ArchGEM to scattered light glitches across Advanced LIGO observing runs O3 (2019--2020) and O4 (2023--2024). We find that the average frequency distributions of this noise span 15--25 Hz in O3a and O4, but increase to 20--40 Hz during O3b. Typical inferred surface velocities are 0.2--0.5 $μ$m/s, and inferred surface displacements are 0.1--0.3 $μ$m. The Gaussian Mixture Model performs most consistently for complex or overlapping features, with mean frequency offsets within 5 Hz of the Gravity Spy baseline. Our results show that ArchGEM provides a practical tool for detector characterization by linking observed spectrogram features to the motion of scattering surfaces and helping guide future mitigation of scattered light noise in current and next-generation interferometers. By quantifying the temporal and spectral behavior of scattered light, ArchGEM provides a robust framework for diagnosing noise sources and guiding targeted mitigation strategies in future detector upgrades.

Figures

Caption

Q-scan of scattered light noise observed in the calibration strain channel at LIGO Livingston during the Third Observing Run (O3). The plot highlights the distinct arch-like morphology of the scattered light noise, indicating periodic motion of a scattering surface within the interferometer environment. This noise predominantly affects the gravitational-wave data between 20–50 Hz.

Caption

Flowchart of the \textsc{ArchGEM} algorithm. The input requires auxiliary channel data, GPS event time, and a duration window. We select our triggers classified by GravitySpy as scattered light events within the specified duration, followed by the generation of Q-scans and spectrograms to enhance the morphology of the scattered light. The data-filtering step extracts the arches from the Q-scan using frequency and energy filtering. The data-processing step sends the filtered data to two methods: Find Peaks and GMM. The output of \textsc{ArchGEM} provides insights into the dynamics of the scattering sources, such as frequency versus time and maximum frequency versus velocity plots, along with key characteristics of the scattered light arches.

Caption

Time--frequency representations for a simulated scattering event. The left panel displays a constant-$Q$ (Q-transform) map of the simulated scattered-light noise, highlighting periodic arch-like features in the 20--30~Hz range consistent with scattering surface motion. The right panel shows a standard short-time Fourier-transform spectrogram of the same data, illustrating the relative amplitude variations over time and frequency.

Caption

Box plot comparing $f_{\mathrm{max,avg}}$ distributions for Find Peaks, GMM, and GravitySpy methods across observation runs O3a, O3b, and O4. The Gravity Spy catalog values (derived from Omicron trigger parameters) provide a stable baseline across all runs, with a consistent lower frequency distribution, while the Find Peaks and GMM methods show variability. In O3a and O3b, Find Peaks and GMM exhibit higher frequencies with broader ranges, particularly in O3b. In O4, both methods show reduced distributions, indicating a systematic decrease in maximum frequency. The comparison to the GravitySpy values suggests that observed variability in $f_{\mathrm{max,avg}}$ over time is not due to baseline shifts but may reflect method-specific or environmental factors influencing these two methods across runs. At Livingston, changes in commissioning state and scattered-light coupling paths between O3 and early O4 may also contribute to run-to-run shifts in the recovered frequency distributions.

Caption

Residuals for $f_{\mathrm{max,avg}}$ ($f_{\mathrm{ARCHGEM}} - f_{\mathrm{GS}}$) for scattered-light event classifications across observing runs O3a, O3b, and O4. Each distribution compares the average maximum frequency recovered by the \textsc{ArchGEM} pipeline (using \textit{Find Peaks} or \textit{GMM}) against corresponding \textsc{Gravity Spy} estimates. A median near zero indicates close agreement between the two methods, while positive or negative offsets reflect systematic over- or underestimation by \textsc{ArchGEM}. Consistent with prior analyses, the standard deviation of the mean relative error ranges from approximately $2.5$--$4.9$~Hz across runs. The \textit{GMM} method demonstrates narrower spreads and smaller residuals than \textit{Find Peaks}, highlighting its robustness and stability under varying noise conditions, particularly during O3b and O4.

Caption

This figure presents violin plots comparing the performance of Method 1 and Method 2 in analyzing scattering events for four variables: (a) Scattering Frequency [Hz], (b) Average Maximum Frequency [Hz], (c) Scattering Surface Movement Distance [$\mu$m], and (d) Scattering Surface Velocity [$\mu$m\,s$^{-1}$]. Each subplot visualizes the distribution of values for each method, with "Method 1" and "Method 2" indicated on the x-axis. The violin plots highlight the differences in data distribution between the two methods, providing a comprehensive comparison of their effectiveness.