Improving the sensitivity to gravitational-wave sources by modifying the input-output optics of advanced interferometers

To improve the sensitivity of laser-interferometer gravitational-wave (GW) detectors, experimental techniques of generating a squeezed vacuum in the GW frequency band are being developed. The squeezed vacuum generated from nonlinear optics has a constant squeeze angle and squeeze factor, while optimal use of squeezing usually requires a frequency dependent (FD) squeeze angle and/or a homodyne detection phase. This frequency dependence can be realized by filtering the input squeezed vacuum or the output light through detuned Fabry-Perot cavities. In this paper, we study FD input-output schemes for signal-recycling interferometers, the baseline design of Advanced LIGO and the currently operational configuration of GEO 600. Complementary to a recent proposal by Harms to use FD input squeezing and ordinary homodyne detection, we explore a scheme which uses an ordinary squeezed vacuum, but FD readout. Both schemes, which are suboptimal among all possible input-output schemes, provide a global noise suppression by the power squeeze factor. At high frequencies, the two schemes are equivalent, while at low frequencies the scheme studied in this paper gives better performance than the Harms scheme, and is nearly fully optimal. We then study the sensitivity improvement achievable by these schemes in Advanced LIGO era (with 30-m filter cavities and current estimates of filter-mirror losses and thermal noise), for neutron star binary inspirals, for low-mass x-ray binaries, and known radio pulsars. Optical losses are shown to be a major obstacle for the actual implementation of these techniques in Advanced LIGO. On time scales of third-generation interferometers, such as EURO/LIGO-III (similar to2012), with kilometer-scale filter cavities and/or mirrors with lower losses, a signal-recycling interferometer with the FD readout scheme explored in this paper can have performances comparable to existing proposals.