This paper compares ways to reduce quantum noise in the KAGRA gravitational‑wave detector after its planned post‑O5 upgrade. Quantum noise comes from the basic quantum jitter of light. It limits how far and how well detectors like KAGRA can see signals from colliding neutron stars and black holes. The authors test a set of practical methods and report which work best under different noise conditions.

Quantum noise shows up in two main forms: shot noise at high frequencies and radiation‑pressure noise at low frequencies. Both come from tiny vacuum fluctuations of the electromagnetic field. A standard tool to fight this is squeezed vacuum: a special light state that reduces uncertainty in one quantity at the cost of increasing it in the other. To reduce quantum noise across the detector’s full frequency band, it helps to make that reduced‑uncertainty direction change with frequency. That is called frequency‑dependent squeezing (FDS). In practice FDS is usually made by sending squeezed light through a long, detuned, over‑coupled optical cavity, often a few hundred metres long. (LIGO used FDS in its O4 run and saw a large increase in detections.)

The authors compare five schemes for KAGRA post‑O5: frequency‑independent squeezing (FIS, no frequency rotation), FDS made with a standard filter cavity (FC), FDS using an amplitude filter cavity (AFC), FDS using a frequency‑dependent beam splitter (FDBS), and a two‑mode or EPR (Einstein–Podolsky–Rosen) squeezing scheme that builds the frequency dependence into two entangled beams. They model realistic losses and KAGRA’s other noises, notably suspension thermal noise from multi‑stage pendulums and mirror thermal noise, which limit low‑frequency sensitivity.

Key results are concrete. The standard filter‑cavity FDS (FC) outperforms the AFC and FDBS versions at all frequencies in their models. Whether FIS or FC gives the best sensitivity depends on what limits the detector at low frequency. If low frequencies are dominated by classical noises (for example suspension or mirror thermal noise), then FIS gives the largest range for detecting binary neutron star (BNS) mergers. If instead low frequencies are dominated by quantum noise, then the FC scheme gives the largest BNS range. The EPR two‑mode scheme performs best for heavy binary systems in their comparison.