On-Chip Optical Squeezing

We report the observation of all-optical squeezing in an on-chip monolithically integrated CMOS-compatible platform. Our device consists of a low-loss silicon nitride microring optical parametric oscillator (OPO) with a gigahertz cavity linewidth. We measure 1.7 dB (5 dB corrected for losses) of sub-shot-noise quantum correlations between bright twin beams generated in the microring four-wave-mixing OPO pumped above threshold. This experiment demonstrates a compact, robust, and scalable platform for quantum-optics and quantum-information experiments on chip.

Figure 1

(a) Schematic of the generation of intensity-correlated signal and idler beams via four-wave mixing in the on-chip microring optical parametric oscillator (OPO). (b) Top: Optical spectrum analyzer scan of the light coupled out of the chip, at a pump power just above the parametric oscillation threshold. The ${\mathrm{Si}}_3{\mathrm{N}}_4$ ring is pumped at 1549.6 nm, generating signal and idler modes at 1540.2 and 1559.2 nm, respectively, which are 15 cavity modes away from the pump wavelength, as can be seen in the bottom figure. Bottom: Transmission spectrum of the ring for transverse-electric (TE) polarization.

Figure 2

Experiment setup. The on-chip microring resonator is pumped by a continuous-wave laser followed by an erbium-doped fiber amplifier (EDFA) from JDS Uniphase and a bandpass filter (BPF). To ensure efficient coupling in and out of the chip, inverse tapers are used [29]. A lensed fiber is used for input coupling. The output from the waveguide is collected with a high-NA objective lens. The diffraction grating spatially separates the pump and the twin beams. After blocking the pump, the beams are focused on a balanced detector. The rf output of the balanced detector is sent to an electrical spectrum analyzer. A small fraction of the beams is sent to an optical spectrum analyzer (OSA) to monitor the onset of parametric oscillation. The inset shows a microscope image of the ring resonator coupled to the bus waveguide. The shape of the fabricated ring is noncircular to fit the resonator within one field of the electron-beam lithography tool and, hence, avoids stitching errors. VOA: Variable optical attenuator. M1, M2: Mirrors. BS: Beam splitter.

Figure 3

Linearity of the balanced detection system from 0.5 to 5 MHz. The data are dark-noise corrected and normalized to the shot-noise spectrum at a base local oscillator (LO) power of $P_0=7.8\mu \mathrm{W}$, corresponding to the horizontal solid line at 0 dB. The dashed lines depict the mean optical power in the corresponding spectrum referenced to the base LO power $P_0$, in decibels. There is significant technical noise at low frequencies, which is clearly evident at relative powers above 10 dB. The data have more fluctuations at high frequencies due to higher electronic noise and consequently lower dark-noise clearance for the base LO power $P_0$. The dark noise varies from 5 dB below the 0-dB reference at 0.5 MHz to 1 dB below the reference at 5 MHz. Around 3 MHz, the detection system is linear over more than 20 dB, which is also substantiated by the linearity of the shot-noise calibration measurement in Fig. 4.

Figure 4

Shot-noise calibration showing the linearity of the noise power in the photocurrent difference with optical power, as expected from theory. The data are taken at a Fourier sideband frequency of 3 MHz, using a spectrum analyzer with a resolution bandwidth of 30 kHz and a video bandwidth of 100 Hz. Data points are corrected by subtracting the electronic noise of the balanced detection system. Error bars are extracted from the raw data.

Figure 5

Intensity-difference squeezing. (a) The variance of the signal and idler photocurrent difference normalized to the shot-noise level (SNL) on a linear scale. The dashed line at 1 represents the SNL, and the solid line represents the intensity-difference fluctuations. The dark-noise clearance is 16 dB at 0.5 MHz and decreases to 5 dB at 5 MHz. The data are taken with a pump power of 93 mW. The power in the signal and idler beams is $45\mu \mathrm{W}$, which requires low-dark-noise detectors commercially unavailable at higher frequencies. Squeezing at low frequencies is masked by excess technical noise below approximately 2 MHz, that can be clearly seen in Fig. 3. (b) Variation of squeezing with attenuation at a sideband frequency of 3 MHz. The $x$ axis is the transmittivity of the variable optical attenuator, $\eta$. The $y$ axis is the intensity-difference noise in shot-noise units, that is, the variance of the signal and idler photocurrent difference compared to the shot-noise level. Error bars are determined from the standard deviation of the measured data points.