Narrow spectral artifact investigation and mitigation in LIGO data from the fourth LIGO-Virgo-KAGRA observing run

This paper details the updates to software tools and the successful investigation and mitigation of narrow spectral artifacts in LIGO data during the fourth observing run, aiming to minimize non-astrophysical noise bands to enhance the discovery potential for persistent gravitational-wave signals.

The Gist

Imagine the LIGO detectors as the world's most sensitive microphones, designed to hear the faintest whispers of the universe—specifically, the ripples in space-time caused by colliding black holes or spinning neutron stars. These "whispers" are incredibly quiet. The problem is that our universe is also full of loud, annoying static.

This paper is a report card on how the LIGO team cleaned up that static during their fourth major listening session (called "O4"). Here is the breakdown of what they did, using simple analogies.

The Problem: The "Hum" in the Room

Think of the data LIGO collects as a giant, continuous recording. Scientists are looking for a specific, pure musical note (a gravitational wave) that lasts for a long time. However, the recording is filled with "lines"—persistent, narrow spikes of noise that look like musical notes but are actually just the building shaking, electrical equipment buzzing, or cameras humming.

If you are trying to hear a specific violin solo, and there is a refrigerator humming at the exact same pitch, you can't hear the violin. In LIGO's case, these "refrigerator hums" are called narrow spectral artifacts. They can hide real cosmic signals or trick scientists into thinking they found something when they didn't.

The Tools: The Detective's Toolkit

To find these hums, the team upgraded their software detective tools.

• Fscan: Think of this as a high-powered microscope for sound. It breaks the data down into tiny frequency slices (like looking at a rainbow through a very fine prism) to spot even the faintest, narrowest hums. They updated this tool to be faster, more interactive, and better at spotting patterns that change over time.
• STAMP-PEM & StochMon: These are like wide-angle lenses. They look at broader chunks of sound to find noise that affects the whole "room" rather than just a single note. They also check if the two LIGO detectors (in Washington and Louisiana) are hearing the same noise. If they are, it's likely a local problem (like a power line), not a signal from space.

The Case Studies: Catching the Culprits

The paper details several specific "criminals" they caught and neutralized during the O4 run. Here are a few examples:

1. The Heater That Was Too Hot

• The Crime: A strange "comb" of noise (many notes spaced evenly apart) appeared in the data.
• The Clue: The noise would vanish and reappear randomly.
• The Solution: The team realized the noise was linked to a heater on a specific mirror (the "OM2"). When the heater was turned on, the noise appeared. By rewiring the heater's controller, they silenced it. It was like realizing a noisy fan was only on when a specific light switch was flipped.

2. The Camera Shutter

• The Crime: Another "comb" of noise, this time related to a camera taking pictures of the laser beam.
• The Solution: The camera was snapping photos at a rate that created a rhythmic hum. The engineers changed how the camera operated during sensitive listening times, and the noise stopped.

3. The Flowing Water

• The Crime: A series of hums that seemed to drift in pitch.
• The Solution: After a long investigation, they found the culprit was a flow meter on a cooling system for the main laser. The electrical signal from the meter was leaking into the data. They rewired the power supply to isolate the meter, and the hum disappeared.

4. The "Ghost" Cameras

• The Crime: A persistent hum near 30 Hz (the speed of a TV frame rate).
• The Solution: They found three video cameras in the laser room that were running 24/7, even though they weren't needed for the experiment. These cameras were humming at 29.97 Hz. When the team unplugged them, the noise vanished. It turned out they had been leaving the "TVs" on in the control room the whole time.

5. The "Double-Tone" Timing

• The Crime: A new, loud noise appeared near 960 Hz that was heard by both LIGO detectors.
• The Solution: This was caused by a new timing system update. Because it was synchronized to the GPS clock at both sites, it sounded exactly the same in both detectors. They couldn't just turn it off because it was needed for the system to work. Instead, they decided to move the frequency of the noise up to a higher pitch (1920 Hz) where it wouldn't interfere with the specific signals they were hunting for.

The Result: The "Do Not Listen" Lists

Even after fixing what they could, some noise remains. To help scientists searching for real signals, the team created two "Blacklists":

1. Lines Lists: A detailed catalog of every known "hum" for the continuous wave searches. If a search finds a signal on a frequency on this list, they know to ignore it because it's just a known noise source.
2. Notch Lists: A slightly coarser list for searches looking for a background "hiss" of gravitational waves. It tells them which frequency bands to cut out of their analysis to avoid false alarms.

The Bottom Line

The paper concludes that while they successfully identified and silenced many annoying noises (like the cameras and heaters), some stubborn problems remain, particularly those caused by complex interactions between different parts of the machine (like "intermodulation," where two noises mix to create a third, unwanted noise).

The key takeaway is that to hear the universe, you first have to make sure your own house isn't making noise. The team spent a lot of time unplugging unnecessary devices, rewiring connections, and upgrading their software to ensure that when they hear a "whisper" from space, they know it's really a whisper and not just a refrigerator humming.

Technical Summary

Technical Summary: Narrow Spectral Artifact Investigation and Mitigation in LIGO O4 Data

Problem Statement
Gravitational-wave (GW) detectors, specifically the Laser Interferometer Gravitational-wave Observatory (LIGO), are subject to non-astrophysical noise sources that manifest as narrow spectral artifacts, or "lines," in the strain data $h(t)$. These artifacts are particularly detrimental to searches for continuous waves (CW) from rapidly rotating, non-axisymmetric neutron stars, which require high sensitivity to nearly monochromatic signals. Furthermore, these lines can bias cross-correlation analyses used in stochastic gravitational-wave background (SGWB) searches by accumulating power in specific frequency bins. During the fourth LIGO-Virgo-KAGRA observing run (O4), which yielded more sensitive data than all previous runs combined, the Detector Characterization (DetChar) group faced the challenge of identifying, characterizing, and mitigating a multitude of these persistent spectral artifacts to maximize the potential for astrophysical discovery.

Methodology
The investigation relied on a suite of software tools designed for data quality monitoring and artifact characterization, primarily utilizing high-resolution frequency-domain data.

• Fscan: A Python-based software package rewritten for O4 to replace legacy tools. It generates normalized and averaged spectra, spectrograms, and coherence between the strain channel $h(t)$ and approximately 75 auxiliary channels per detector. Fscan operates on daily, weekly, and monthly cadences using Short Fourier Transform (SFT) data (typically 1800s duration). Key features include:
  • Spectral Averaging: Utilizing both normalized averages and inverse-variance weighting to enhance the visibility of narrow features.
  • Coherence Analysis: Computing coherence between $h(t)$ and auxiliary channels to localize noise sources.
  • Persistency Metrics: Using a Z-statistic to track the time-dependent nature of lines, helping to correlate artifacts with detector configuration changes.
  • Line Counting: Employing a topographic prominence algorithm to robustly count lines and identify "combs" (families of lines with common frequency spacing).
• STAMP-PEM: A complementary tool designed for SGWB analyses, operating at coarser frequency resolution (0.1 Hz) to identify broadband or moderately narrow features impacting cross-correlation measurements.
• StochMon: A tool that mimics isotropic SGWB analysis to monitor coherence between spatially separated detectors (Hanford and Livingston), identifying statistically significant coherence outliers that violate the assumption of uncorrelated noise.
• Investigation Workflow: The team utilized a prioritization scheme based on astrophysical impact (prolificness, sensitivity of affected bands, and impact on known pulsars) to allocate resources. Investigations involved correlating artifact behavior with detector configuration changes, analyzing historical data to identify onset times, and using portable magnetometers and auxiliary channel data to trace coupling mechanisms.

Key Contributions and Results
The paper details the identification and, where possible, mitigation of several specific narrow spectral artifacts observed during O4, primarily in the LIGO Hanford (H1) detector:

1. OM2 Heater Driver Comb: A comb artifact with $\approx 1.66$ Hz spacing was traced to the output mirror 2 (OM2) optic heater. Changing the electrical connections to the Beckhoff controller successfully mitigated the artifact.
2. HWS Camera Shutter Comb: Combs with spacings near $\approx 5$ Hz and $\approx 1$ Hz were identified as originating from the Hartmann Wavefront Sensor (HWS) camera shutter rates. Mitigation was achieved by altering the camera's operation during low-noise runs.
3. PSL Flow Sensor Combs: A series of combs near $\approx 9.48$ Hz were traced to flow meter sensors in the Pre-Stabilized Laser (PSL) chiller system, installed during upgrades prior to O4. Reworking the power supply to isolate the noise successfully mitigated the artifact.
4. Violin Mode Intermodulation: Excess line contamination surrounding the fundamental violin modes ($\approx 500$ Hz) was found to be consistent with intermodulation arising from non-linear effects in the interferometric readout electronics, specifically between violin modes and calibration lines.
5. Temporary Calibration Line Intermodulation: The addition of new temporary calibration lines exacerbated line contamination, particularly when violin modes were excited. These temporary lines were subsequently turned off.
6. Aliasing Artifacts: In-band aliasing artifacts above 400 Hz were identified due to insufficient anti-alias filtering in the digital readout path. Installation of additional filtering mitigated these lines.
7. Near-30 Hz and Near-100 Hz Comb Families: The near-30 Hz family was traced to NTSC-timed video cameras in the PSL enclosure; powering them off eliminated the artifact. The near-100 Hz family remains under investigation.
8. Suspected Beckhoff Comb: A comb near $\approx 11.9$ Hz, observed in both detectors, is suspected to originate from the Beckhoff control system, though it has not yet been mitigated.
9. DuoTone Artifacts: New lines near 960 Hz, introduced by a redesign of the readout electronics, were identified as highly coherent between detectors due to GPS synchronization. A configuration change to move these frequencies to $\approx 1920$ Hz was proposed to reduce their impact on low-frequency searches.

Data Quality Products
The paper outlines the generation of two critical data quality products for narrowband persistent GW searches:

• Lines Lists: Used primarily for CW searches, these catalogs vetted artifacts based on the identification of non-astrophysical sources. By the end of O4, 2,399 artifacts were vetted in Hanford (covering 17.21% of the 10–2000 Hz band, reduced to 5.92% if violin mode regions are excluded) and 449 in Livingston (6.37%).
• Notch Lists: Optimized for SGWB cross-correlation analyses, these lists exclude frequency bins impacted by instrumental artifacts, coherence outliers, or large temporal variability. They are coarser in frequency resolution (1/32 Hz) than lines lists.

Significance and Claims
The paper asserts that the systematic investigation and mitigation of narrow spectral artifacts are essential to maximize the discovery potential for continuous gravitational waves and to ensure the accuracy of stochastic background measurements. The authors highlight that while significant progress was made in O4, including the successful mitigation of several major combs and the identification of intermodulation mechanisms, challenges remain. Specifically, the paper notes that:

• Contamination can persist for months while investigations unfold.
• Intermodulation problems affecting broad frequency bands remain largely unsolved.
• Automation of the identification and investigation workflow is necessary to handle the volume of data in future observing runs.
• Rigorous characterization of new electronics prior to installation and the minimization of non-essential powered devices are critical strategies for minimizing noise coupling.

The work emphasizes that minimizing the number of frequency bands containing non-astrophysical artifacts is a prerequisite for the successful detection of new classes of gravitational-wave signals.