Automated Evaluation of Environmental Coupling for Advanced LIGO Gravitational Wave Detections

Author(s)

Helmling-Cornell, Adrian, Nguyen, Philippe, Schofield, Robert, Frey, Raymond

Abstract

The extreme sensitivity required for direct observation of gravitational waves by the Advanced LIGO detectors means that environmental noise can potentially contaminate gravitational wave signals. Consequently, environmental monitoring efforts have been undertaken and novel noise mitigation techniques have been developed which have helped keep environmental artifacts from influencing gravitational wave detections for the $90$ gravitational wave events detected from 2015--2020 by the aLIGO detectors. The increasing rate of gravitational wave detections due to detector sensitivity improvements requires sophisticated, reliable and automated ways to monitor and assess the degree of environmental coupling between gravitational wave detectors and their surroundings. We introduce a computational tool, PEMcheck, for quantifying the degree of environmental coupling present in gravitational wave signals using data from the network of environmental monitoring sensors. We study its performance when applied to the $79$ gravitational waves confidently detected in LIGO's third observing run and test its performance in the case of extreme environmental contamination of gravitational wave data. We find that PEMcheck's automated analysis identifies only a small number of gravitational waves that merit further study by environmental noise experts due to possible contamination, a substantial improvement over the manual vetting that occurred for every gravitational wave candidate in previous observing runs. Overall, PEMcheck works as intended. Consequently, PEMcheck will play a critical role in event validation during LIGO's fourth observing run.

Figures

Vibrational coupling between \ac{LLO}'s \ac{HAM} 6 vacuum chamber Z-axis accelerometer and \ac{DARM} as measured prior to \ac{O3}. The vibrational coupling here is likely driven by stray light scattering off of the septum window dividing this vacuum chamber and the adjoining \ac{HAM} 5 chamber and then recombining with the main beam~\cite{septumalog}. The large fraction of upper limit estimates, rather than measured values, for the vibrational coupling is because \ac{PEM} injections could not be increased to an amplitude such that the injection was visible in the \ac{DARM} data without saturating the accelerometer signal. Each sensor count corresponds to $\unit[6.1]{\mu m~s^{-2}}$ of acceleration in the vertical direction~\cite{PEMpage}.

Vibrational coupling between \ac{LLO}'s \ac{HAM} 6 vacuum chamber Z-axis accelerometer and \ac{DARM} as measured prior to \ac{O3}. The vibrational coupling here is likely driven by stray light scattering off of the septum window dividing this vacuum chamber and the adjoining \ac{HAM} 5 chamber and then recombining with the main beam~\cite{septumalog}. The large fraction of upper limit estimates, rather than measured values, for the vibrational coupling is because \ac{PEM} injections could not be increased to an amplitude such that the injection was visible in the \ac{DARM} data without saturating the accelerometer signal. Each sensor count corresponds to $\unit[6.1]{\mu m~s^{-2}}$ of acceleration in the vertical direction~\cite{PEMpage}.


\ac{ASD} of the Z-axis \ac{HAM} 6 accelerometer during the background time. Background \ac{PEM} sensor data is used to compute the tuned \ac{CF} of section~\ref{sssec:tunedasd}. $\unit[45.75]{s}$ of sensor data were used to calculate this \ac{ASD}.

\ac{ASD} of the Z-axis \ac{HAM} 6 accelerometer during the background time. Background \ac{PEM} sensor data is used to compute the tuned \ac{CF} of section~\ref{sssec:tunedasd}. $\unit[45.75]{s}$ of sensor data were used to calculate this \ac{ASD}.


Top: Comparison of the \ac{CF} interpolated from the data in figure~\ref{fig:example_cf} to the interpolated, tuned \ac{CF} computed by analyzing the \ac{DARM} background \ac{ASD}. The untuned and tuned predictions for the \ac{ASD} largely agree, except near $30$ and $\unit[90]{Hz}$. The points where environmental contributions to \ac{DARM} are overestimated and need tuning are derived from upper limit estimates of the chamber motion to \ac{DARM} coupling. These frequencies are marked with black arrows. Bottom: Comparison of the \ac{DARM} $t_{\mathrm{back}}$ \ac{ASD} with the estimated contribution from the \ac{HAM} 6 accelerometer. The result of the tuning procedure is to constrain the predicted environmental contribution from a sensor to be not greater than the observed \ac{DARM} \ac{ASD}. As in the top plot, black arrows mark where the untuned \ac{CF} predicts that the \ac{DARM} \ac{ASD} should exceed its observed value to the \ac{HAM} 6 accelerometer data. These are the points which require tuning to reconcile the predicted and actual \ac{DARM} \ac{ASD}s.

Top: Comparison of the \ac{CF} interpolated from the data in figure~\ref{fig:example_cf} to the interpolated, tuned \ac{CF} computed by analyzing the \ac{DARM} background \ac{ASD}. The untuned and tuned predictions for the \ac{ASD} largely agree, except near $30$ and $\unit[90]{Hz}$. The points where environmental contributions to \ac{DARM} are overestimated and need tuning are derived from upper limit estimates of the chamber motion to \ac{DARM} coupling. These frequencies are marked with black arrows. Bottom: Comparison of the \ac{DARM} $t_{\mathrm{back}}$ \ac{ASD} with the estimated contribution from the \ac{HAM} 6 accelerometer. The result of the tuning procedure is to constrain the predicted environmental contribution from a sensor to be not greater than the observed \ac{DARM} \ac{ASD}. As in the top plot, black arrows mark where the untuned \ac{CF} predicts that the \ac{DARM} \ac{ASD} should exceed its observed value to the \ac{HAM} 6 accelerometer data. These are the points which require tuning to reconcile the predicted and actual \ac{DARM} \ac{ASD}s.


The results of the $c$-statistic calculation for each spectrogram tile for the \ac{LLO} \ac{HAM} 6 accelerometer during GW190707. The value of $1-c$ is plotted for each tile so that time-frequency tiles with the highest likelihood of contamination, as calculated by \textsc{PEMcheck}, appear brighter. The time-frequency tiles enclosed by the red box track the evolution of the \ac{GW} signal as predicted by \ac{NR}. The lowest $c$-statistic found within the \ac{GW} track occurs at $\sim\unit[0.1]{s}$ and $\sim\unit[78]{Hz}$ prior to the event's $t_c$. A minimum $c$-statistic of $0.52$ indicates that environmental disturbances witnessed by this sensor do not couple to the \ac{GW} strain data during the event.

The results of the $c$-statistic calculation for each spectrogram tile for the \ac{LLO} \ac{HAM} 6 accelerometer during GW190707. The value of $1-c$ is plotted for each tile so that time-frequency tiles with the highest likelihood of contamination, as calculated by \textsc{PEMcheck}, appear brighter. The time-frequency tiles enclosed by the red box track the evolution of the \ac{GW} signal as predicted by \ac{NR}. The lowest $c$-statistic found within the \ac{GW} track occurs at $\sim\unit[0.1]{s}$ and $\sim\unit[78]{Hz}$ prior to the event's $t_c$. A minimum $c$-statistic of $0.52$ indicates that environmental disturbances witnessed by this sensor do not couple to the \ac{GW} strain data during the event.


Histogram of the minimum $c$-statistic calculated by \textsc{PEMcheck} for all GWTC-2.1 and GWTC-3 confident events using the pre-\ac{O3} coupling data. The \textsc{PEMcheck} analysis identifies $6$ of the $149$ \ac{O3} \ac{GW} events as having a minimum $c$ below $0.2$

Histogram of the minimum $c$-statistic calculated by \textsc{PEMcheck} for all GWTC-2.1 and GWTC-3 confident events using the pre-\ac{O3} coupling data. The \textsc{PEMcheck} analysis identifies $6$ of the $149$ \ac{O3} \ac{GW} events as having a minimum $c$ below $0.2$


Cumulative distribution of the $c$-statistic found for each \ac{PEM} channel studied by \textsc{PEMcheck} for each \ac{GW} event in \ac{O3} using the pre-\ac{O3} coupling data. Each light grey trace corresponds to the \ac{CDF} of the $c$-statistic for an individual \ac{GW} event, while the black trace denotes the mean number of sensors with that particular value of $c$ or less. GW190707 as well as the three events with the lowest $c$ are highlighted.

Cumulative distribution of the $c$-statistic found for each \ac{PEM} channel studied by \textsc{PEMcheck} for each \ac{GW} event in \ac{O3} using the pre-\ac{O3} coupling data. Each light grey trace corresponds to the \ac{CDF} of the $c$-statistic for an individual \ac{GW} event, while the black trace denotes the mean number of sensors with that particular value of $c$ or less. GW190707 as well as the three events with the lowest $c$ are highlighted.


\ac{CDF} of the $c$-statistic for each run of \textsc{PEMcheck} at \ac{LHO} with GW200115's properties. The \ac{CDF} marked ``foreground" corresponds to \textsc{PEMcheck} run at the actual time of the \ac{GW} event.

\ac{CDF} of the $c$-statistic for each run of \textsc{PEMcheck} at \ac{LHO} with GW200115's properties. The \ac{CDF} marked ``foreground" corresponds to \textsc{PEMcheck} run at the actual time of the \ac{GW} event.


Below: constant-Q transform of \ac{LLO} \ac{GW} strain data during a thunderclap witnessed. The 31 red time-frequency tracks overlaid on the thunderclap data represent 31 hypothetical \ac{GW} time-frequency tracks with the properties given in table~\ref{tab:190521g_props} vetted by \textsc{PEMcheck}. Above: the lowest value of $c$ found by \textsc{PEMcheck} for the series of hypothetical \ac{GW} tracks overlaid on thunderclap data. The channel with the lowest $c$-statistic for a given trial is denoted by the marker style. The bold line corresponds to $c=0.2$.

Below: constant-Q transform of \ac{LLO} \ac{GW} strain data during a thunderclap witnessed. The 31 red time-frequency tracks overlaid on the thunderclap data represent 31 hypothetical \ac{GW} time-frequency tracks with the properties given in table~\ref{tab:190521g_props} vetted by \textsc{PEMcheck}. Above: the lowest value of $c$ found by \textsc{PEMcheck} for the series of hypothetical \ac{GW} tracks overlaid on thunderclap data. The channel with the lowest $c$-statistic for a given trial is denoted by the marker style. The bold line corresponds to $c=0.2$.


References
  • [1] Abbott B et al. 2016 Phys. Rev. Lett. 116 061102 (Preprint 1602.03837)
  • [2] Abbott B et al. 2017 Phys. Rev. Lett. 119 161101 (Preprint 1710.05832)
  • [3] Abbott B et al. 2019 Phys. Rev. X 9 031040 (Preprint 1811.12907)
  • [4] Abbott R et al. 2021 arXiv e-prints arXiv:2108.01045 (Preprint 2108.01045)
  • [4] Abbott R et al. 2021 arXiv e-prints arXiv:2108.01045 (Preprint 2108.01045)
  • [5] Abbott R et al. 2021 arXiv e-prints arXiv:2111.03606 (Preprint 2111.03606)
  • [5] Abbott R et al. 2021 arXiv e-prints arXiv:2111.03606 (Preprint 2111.03606)
  • [6] Acernese F et al. 2015 Class. Quantum Gravity 32 024001 (Preprint 1408.3978)
  • [7] Acernese F et al. 2019 Phys. Rev. Lett. 123 231108
  • [8] Akutsu T et al. 2021 Progress of Theoretical and Experimental Physics 2021 05A101 (Preprint 2005.05574)
  • [9] Aasi J et al. 2015 Class. Quantum Gravity 32 074001 (Preprint 1411.4547)
  • [10] Buikema A et al. 2020 Phys. Rev. D 102 062003 (Preprint 2008.01301)
  • [11] Ganapathy D et al. 2023 Phys. Rev. X 13 041021
  • [12] Nguyen P et al. 2021 Class. Quantum Gravity 38 145001 (Preprint 2101.09935)
  • [13] Glanzer J, Soni S, Spoon J, Effler A and González G 2023 Class. Quantum Gravity 40 195015 (Preprint 2304.07477)
  • [14] Soni S et al. 2021 Class. Quantum Gravity 38 025016 (Preprint 2007.14876)
  • [15] Janssens K et al. 2023 Phys. Rev. D 107 022004 (Preprint 2209.00284)
  • [16] Davis D et al. 2021 Class. Quantum Gravity 38 135014 (Preprint 2101.11673)
  • [17] Abbott B P et al. 2016 Class. Quantum Gravity 33 134001 (Preprint 1602.03844)
  • [18] Schofield R et al. H1 L1 PEM vetting report for the Aug. 17, 2017 BNS candidate https: //dcc.ligo.org/LIGO-T1800508/public
  • [19] The LIGO Scientific Collaboration and The Virgo Collaboration 2018 Data quality report user documentation https://docs.ligo.org/detchar/data-quality-report/
  • [20] The LIGO-Virgo-KAGRA Collaboration 2022 dqrtasks user documentation https://detchar. docs.ligo.org/dqrtasks/
  • [21] Abbott R et al. 2023 Physical Review X 13 011048 (Preprint 2111.03634)
  • [22] Abbott B P et al. 2018 Living Reviews in Relativity 21 3 (Preprint 1304.0670)
  • [23] Pem central URL https://pem.ligo.org/
  • [24] Malaquias-Reis J and Effler A Pem sensor work at ey https://alog.ligo-la.caltech.edu/ aLOG/index.php?callRep=64021
  • [25] Clara F and Helmling-Cornell A Pem microphone upgrade https://alog.ligo-wa.caltech.edu/ aLOG/index.php?callRep=69448
  • [26] Effler A et al. 2015 Class. Quantum Gravity 32 035017 (Preprint 1409.5160)
  • [27] Arai K, Barnum S, Fritschel P, Lewis J and Waldman S Output Mode Cleaner Design https: //dcc.ligo.org/LIGO-T1000276/public
  • [28] Nguyen P 2022 Environmental Noise in Gravitational Wave Detectors and the Search for Gravitational Wave Signals Associated with Gamma-Ray Burst During LIGO’S Third Observing Run Ph.D. thesis University of Oregon
  • [29] Schofield R Ham5/6 vibration coupling much reduced after septum window removal, 35.5 hz darm peak is from residual coupling in the area https://alog.ligo-wa.caltech.edu/aLOG/index. php?callRep=66470
  • [30] Nguyen P et al. Pre-o3 pem injection summary and links to individual logs https://alog. ligo-wa.caltech.edu/aLOG/index.php?callRep=49521
  • [31] Schofield R, Austin C, P N and Effler A Llo magnetic injections - o3 couplings https://alog. ligo-la.caltech.edu/aLOG/index.php?callRep=45371
  • [32] Short R, Glanzer J, Helmling-Cornell A, Effler A and Schofield R Pem injections successfully completed - will inform hvac experiments https://alog.ligo-wa.caltech.edu/aLOG/index. php?callRep=69745
  • [33] Effler A et al. Pem injections status https://alog.ligo-la.caltech.edu/aLOG/index.php? callRep=64609
  • [34] Gracedb — the gravitational-wave candidate event database URL https://gracedb.ligo.org/
  • [35] Macleod D et al. 2021 SoftwareX 13 100657
  • [36] Abbott R et al. 2021 Phys. Rev. Lett. 126 241102 (Preprint 2101.12248)
  • [37] Abac A G et al. 2023 arXiv e-prints arXiv:2308.03822 (Preprint 2308.03822)
  • [37] Abac A G et al. 2023 arXiv e-prints arXiv:2308.03822 (Preprint 2308.03822)
  • [38] Messick C et al. 2017 Phys. Rev. D 95 042001 (Preprint 1604.04324)
  • [39] Sachdev S et al. 2019 arXiv e-prints arXiv:1901.08580 (Preprint 1901.08580)
  • [39] Sachdev S et al. 2019 arXiv e-prints arXiv:1901.08580 (Preprint 1901.08580)
  • [40] Adams T et al. 2016 Class. Quantum Gravity 33 175012 (Preprint 1512.02864)
  • [41] Nitz A H, Dal Canton T, Davis D and Reyes S 2018 Phys. Rev. D 98 024050 (Preprint 1805.11174)
  • [42] Dal Canton T and Harry I W 2017 arXiv e-prints arXiv:1705.01845 (Preprint 1705.01845)
  • [42] Dal Canton T and Harry I W 2017 arXiv e-prints arXiv:1705.01845 (Preprint 1705.01845)
  • [43] Hooper S et al. 2012 Phys. Rev. D 86 024012 (Preprint 1108.3186)
  • [44] Chu Q 2017 Low-latency detection and localization of gravitational waves from compact binary coalescences Ph.D. thesis Western Australia University
  • [45] Bohé et al. 2017 Phys. Rev. D 95 044028 (Preprint 1611.03703)
  • [46] Klimenko S, Mohanty S, Rakhmanov M and Mitselmakher G 2005 Phys. Rev. D 72 122002 (Preprint gr-qc/0508068)
  • [47] Klimenko S et al. 2016 Phys. Rev. D 93 042004 (Preprint 1511.05999)
  • [48] Abbott R et al. 2021 Phys. Rev. X 11 021053 (Preprint 2010.14527)
  • [49] LIGO Scientific Collaboration and Virgo Collaboration 2019 GCN 25012 URL https://gcn. gsfc.nasa.gov/other/GW190707q.gcn3
  • [50] Abbott R et al. 2022 Astron. Astrophys. 659 A84 (Preprint 2105.15120)
  • [51] Huber M C and Davis D 2022 arXiv e-prints arXiv:2208.13844 (Preprint 2208.13844)
  • [51] Huber M C and Davis D 2022 arXiv e-prints arXiv:2208.13844 (Preprint 2208.13844)
  • [52] Hannam M et al. 2022 Nature 610 652–655 (Preprint 2112.11300)
  • [53] Payne E et al. 2022 Phys. Rev. D 106 104017 (Preprint 2206.11932)
  • [54] Abbott R et al. 2023 Astrophys. J. Suppl. Ser. 267 29 (Preprint 2302.03676)
  • [55] Farr W M, Gair J R, Mandel I and Cutler C 2015 Phys. Rev. D 91 023005 (Preprint 1302.5341)
  • [56] Abbott R et al. 2020 Astrophys. J. Lett. 896 L44 (Preprint 2006.12611)
  • [57] Dwyer M, Kijbunchoo N, Merfeld K, Sigg D and McCarthy R Comment to “replacing the failing squeezer laser” https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=51244
  • [58] Chatterji S, Blackburn L, Martin G and Katsavounidis E 2004 Class. Quantum Gravity 21 S1809– S1818 (Preprint gr-qc/0412119)
  • [59] Chatterji S 2005 The search for gravitational wave bursts in data from the second LIGO science run Ph.D. thesis Massachusetts Institute of Technology
  • [60] Brown J C 1991 J. Acoust. Soc. Am. 89 425–434
  • [61] Ball M et al. New beating-shaker technique localizes coupling site of 48 hz darm peak, and black glass eliminates it https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=52184
  • [62] Arain M A and Mueller G 2008 Optics Express 16 10018
  • [63] Nguyen P and Schofield R Coupling estimates from pem injections correctly predict llo darm glitch near s190510g from thunder-driven vibration at ey https://alog.ligo-la.caltech. edu/aLOG/index.php?callRep=46025
  • [64] Abbott R et al. 2020 Phys. Rev. Lett. 125 101102 (Preprint 2009.01075)
  • [65] Abbott R et al. 2021 SoftwareX 13 100658 (Preprint 1912.11716)
  • [66] The LIGO Scientific Collaboration and The Virgo Collaboration 2018 ligocam user documentation https://docs.ligo.org/detchar/data-quality-report/tasks/ligocam-dqr.html
  • [67] Punturo M et al. 2010 Class. Quantum Gravity 27 194002
  • [68] Reitze D et al. 2019 Cosmic Explorer: The U.S. Contribution to Gravitational-Wave Astronomy beyond LIGO Bull. Am. Astron. Soc. vol 51 p 35 (Preprint 1907.04833)