Summary: correlations, noise estimates, and movies suggest that some scattering noise seen at LHO and LLO is due to micron-scale relative motion of test and reaction masses. The maximum frequency is increased by multiple reflections, possibly between the ESD traces on the RM and the HR coating on the ETM face. Driving R0 to reduce this relative motion, and/or offloading to ISI or HEPI, may mitigate the scattering noise.
Timesh, Matthew, Adrian, Jenne, Anamaria, Robert
At LHO we are noticing more scattering shelves in DARM with the seasonal increase in the microseism. The noise reaches nearly 100 Hz, several hundred times higher in frequency than the 0.15 Hz seismic motion produced by counter-propagating ocean waves. This requires high velocities of the reflector relative to the cavity and multiple strong reflections, but I think it can be accounted for by standard scattering mechanisms.
The relative velocity of the test and reaction masses is correlated with scattering arches and can account for their frequency spacing
The relative motion between test and reaction masses is not measured, but, at low frequencies, it should be similar to the relative motion of the penultimate masses, L2 and R2, measured by BOSEMs. Figure 1 shows that the relative velocity of the EX PUMs during high microseism reached about 5 microns per second, fast enough to produce the concurrent ~10 Hz gap between scattering arches. This large velocity is associated with the DARM and tidal offloading signals sent to L1 and L2. Figure 1 also shows that the ETMX L2 length witness signal correlates in time and amplitude with the scattering arches in DARM. DetChar scattering tools, and DetChar members have also identified this and similar channels as correlated with scattering arches, e.g. https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=53887 and https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=46057
Movies of test masses during lock show light amplitude modulation correlated with relative motion of test and reaction chains
Light modulation is evident in movies of H1 ETMX and ETMY in lock, suggesting that nearby moving reflectors are reflecting/scattering a lot of light into the test mass. To better understand the motion that produces the visible modulation, I took the difference between the pixel values in successive frames of several movies and calculated the standard deviation of the differences in pixel values to provide a single number indicating the degree of frame-to-frame variation. Figure 2 shows that the light variation in a movie of ETMY is correlated with the slope of the local ETMY L2 length witness, but not with the L2 length witness of ETMX. This supports the argument that the motion between the chains is modulating light so much that it is visible. I posted the movie of ETMY used in Figure 2, sped up by a factor of ten so that the light modulation would be more obvious (https://youtu.be/WkNR89ItXF8 ). The correlation in Figure 2 is for the central region of this movie, but the variation was similar in the 4 close-ups that I examined. Frame analyses of other movies supported this conclusion. I also made movies during 0.1 Hz injections at R0, and spectra of the changes in the movie frames showed a strong peak at 0.2 Hz.
Multiple reflections rather than harmonics
Figures 1 & 2 show that scattering arches are stacked evenly in frequency. Two mechanisms can account for this equal spacing, first, multiple reflections, and second, harmonics of a modulation frequency. These can be distinguished by whether or not the spacing changes as the amplitude of motion changes (https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=31898 ). If the spacing is associated with harmonics of the modulation frequency, then the spacing of the arches doesn’t change as the amplitude of the motion changes. In Figure 1 and in 0.1 Hz ISI injections that I made, the spacing of the arches gets smaller as the amplitude decreases, consistent with the multiple reflection hypothesis and not the harmonic series hypothesis.
The higher-frequency arches from multiple reflections would have relatively smaller amplitudes because of losses at each reflection. Figure 3 shows that each successive shelf in DARM during high microseism has roughly 0.3 of the amplitude of the preceding lower frequency shelf. This suggests a power factor of roughly 0.1 per reflection, more like partial specular reflection than a more lossy path that requires scattering with each cycle, such as scattering from a beam spot to a shiny metal surface and back.
Reflections from the ESD traces can roughly account for the scattering noise and may also provide a multi-reflection mechanism
Hiro’s estimate of scattering noise from the ESD traces (https://dcc.ligo.org/LIGO-T1500455) assumed much smaller relative motion between the cavity and the reaction mass. An updated estimate using the measured motion suggests that the noise is visible or nearly visible in DARM. In addition, photographs in Figure 4, suggest that there is a lot of additional scattered light at the radius of the ESD traces that may end up scattering into the TEM00 mode.
A possible multi-reflection path that could increase the maximum frequency due to ETM-RM relative motion by a factor of several would be multiple reflections between the ESD traces and the HR coating on the front of the test mass. I think a 0.1 power factor per cycle is not unreasonable for this HR-ESD optical path.
There are four alternative paths that I considered that don’t work as well: back and forth between the TM and its cage, the TM and the TMS, the TM and the ACB, or the TM and chamber walls. These other paths would likely have more losses with each reflection then the proposed TM-RM path. In addition, the arch spacing would not fit the witnessed velocity as well as the reaction mass path does because the test mass moves half as much relative to these alternate reflectors (offloading low frequency DARM at L1 and L2 moves the main and reaction chains roughly equal amounts in opposite directions). These other paths may nevertheless produce noise (reaching half the frequency) that doesn’t currently dominate.
Possible mitigation
Driving at R0 in order to minimize the relative motion of the two chains would mitigate noise from the test mass - reaction mass path. Jenne suggested offloading to the ISI or HEPI, which could also mitigate noise from the cage, ACB and the TMS path. I think Jenne, Sheila, Jeff and others are starting to think about this.
Some noise at LLO may have a similar source
While the DARM offload system is different at LLO, at least some of the noise associated with the microseism appears to have a similar source as the noise at LHO. Figure 5 shows that the relative motion of the test and reaction chains predicts the scattering arch spacing and is correlated with the appearance of shelves in DARM.
Because of the non-linear nature of scattered light noise, this low frequency motion may also increase scattering noise from higher frequency motion, even when the scattering arches from the ~0.1 Hz motion are not visible, and mitigation may thus help with, e.g., anthropogenic-band scattering noise. Finally, it might be worth looking into using the TM-RM velocities to improve past data.
Movies of light modulation at the test masses, with illuminated stills for orientation:
ETMY side view used in frame analysis discussed above: https://youtu.be/WkNR89ItXF8
ETMY 45 degree view: https://youtu.be/JghBSjQ2xV4
ETMX 45 degree view: https://youtu.be/EHyjzE8XIXM
