KAGRA underground environment and lessons for the Einstein Telescope

The KAGRA gravitational-wave detector in Japan is the only operating detector hosted in an underground infrastructure. Underground sites promise a greatly reduced contribution of the environment to detector noise thereby opening the possibility to extend the observation band to frequencies well below 10 Hz. For this reason, the proposed next-generation infrastructure Einstein Telescope in Europe would be realized underground aiming for an observation band that extends from 3 Hz to several kHz. However, it is known that ambient noise in the low-frequency band 10 Hz–20 Hz at current surface sites of the Virgo and LIGO detectors is predominantly produced by the detector infrastructure. It is of utmost importance to avoid spoiling the quality of an underground site with noisy infrastructure, at least at frequencies where this noise can turn into a detector-sensitivity limitation. In this paper, we characterize the KAGRA underground site to determine the impact of its infrastructure on environmental fields. We find that while excess seismic noise is observed, its contribution in the important band below 20 Hz is minor preserving the full potential of this site to realize a low-frequency gravitational-wave detector. Moreover, we estimate the Newtonian-noise spectra of surface and underground seismic waves and of the acoustic field inside the caverns. We find that these will likely remain a minor contribution to KAGRA’s instrument noise in the foreseeable future.

Figure 1

Positions where seismic noise was measured moving a Trillium 20s day by day together with a representation of the X arm with the noise sources (red rectangles). Green star: KAGRA seismometer used as reference clock. Light blue stars: positions relative to the X arm. Lilac stars: positions relative to the Y arm. To be noted that position 1 was not used in the analysis because it served only to synchronize the Trillium 20s with KAGRA’s seismic sensors.

Figure 2

Location of the GIFH10 seismic station with respect to the KAGRA detector. The orange lines represent the 3 km long arms of KAGRA.

Figure 3

Top: spectra of seismic noise as measured along the interferometer arm (see Fig. 1 for position references). Bottom: zoom taken on 10–30 Hz. Distances are with respect to the BS. Data were taken from November ${27}^{\mathrm{th}}$ to December $6^{\mathrm{th}}$, 2019. In each selected position, the sensor was deployed on the floor surface and left in acquisition for 1 to 2 days, with a sampling of 250 Hz. The excess noise observed at position 2 is produced by a compressor, which is switched off during science mode.

Figure 4

Seismograms relative to the different locations at which the Trillium was set (see Fig. 1). In position 2 the presence of noise due to the compressor is clearly visible. Each plot was made taking 18 consecutive hours of data. Position 2 starts from 6∶00 a.m. on November 27, 2019; Position 3 starts from 2∶00 a.m. on December 2, 2019; Position 4 starts from 6∶00 a.m. on November 28, 2019; Position 5 starts from 3∶00 a.m. on November 29, 2019; Position 6 starts from 2∶00 a.m. on December 4, 2019; Position 7 starts from 2∶00 a.m. on December 3, 2019.

Figure 5

We show here the spectral variations of the vertical channels of the Trillium 120QA deployed in KAGRA (over a year of quiet times) together with the global low-noise and high-noise models (red dashed lines) [29] and the ${10}^{\mathrm{th}}$, ${90}^{\mathrm{th}}$ (white lines), and ${50}^{\mathrm{th}}$ (black line) percentiles. Upper left: X-end mass upper cavern. Upper right: Y-end mass upper cavern. Bottom: X-input mass upper cavern.

Figure 6

Comparison between seismic spectra at KAGRA (positions 2 and 5 of Fig. 1) and Virgo.

Figure 7

Histogram of seismic speeds given by the first ranked peak (up to 0.6 Hz) and by the third one (from 0.6 Hz to 1 Hz).

Figure 8

NN estimate from surface Rayleigh waves—data from GIFH10 sensor (left), from body waves—data from position 5 (right), and from cavern acoustic fields—data from microphone ACO TYPE 7146NL (bottom). The black lines are the ${50}^{th}$ percentile, while the white lines represent the ${10}^{\mathrm{th}}$ and the ${90}^{\mathrm{th}}$ percentiles.

Figure 9

Acoustic noise in Virgo (red line) compared to the acoustic noise recorded by the microphone (ACO TYPE 7146NL) close to the BS (see Fig. 1). The black line is the ${50}^{\mathrm{th}}$ percentile, the white lines represent the ${10}^{\mathrm{th}}$ and the ${90}^{\mathrm{th}}$ percentiles.

Figure 10

Modeled acoustic and seismic NN spectra (${90}^{\mathrm{th}}$ percentiles). The KAGRA sensitivity curve represents an optimistic scenario (130 Mpc horizon for neutron-star binaries) for the upcoming science run O4 [4]. The seismic spectrum used for the pos 5 NN spectrum is not representative of conditions during observing runs. It is only shown to illustrate what effect infrastructure might potentially have on NN.

Figure 11

The 9 peaks of Eq. (4) in the portion of the wave vector space analyzed at 0.5 Hz. They are represented both in the polar plane (left) and in the Cartesian plane (right).

Figure 12

Values of the 9 median ranked speeds (left). Comparison of the DOA from the first three peaks (blue: first peak, orange: second peak, and green: third peak) with the DOA inferred from the horizontal channels (red lines). See in the text to further explanations (right).