On-chip, photon-number-resolving, telecommunication-band detectors for scalable photonic information processing

Integration is currently the only feasible route toward scalable photonic quantum processing devices that are sufficiently complex to be genuinely useful in computing, metrology, and simulation. Embedded on-chip detection will be critical to such devices. We demonstrate an integrated photon-number-resolving detector, operating in the telecom band at 1550 nm, employing an evanescently coupled design that allows it to be placed at arbitrary locations within a planar circuit. Up to five photons are resolved in the guided optical mode via absorption from the evanescent field into a tungsten transition-edge sensor. The detection efficiency is $7.2\pm 0.5\%$. The polarization sensitivity of the detector is also demonstrated. Detailed modeling of device designs shows a clear and feasible route to reaching high detection efficiencies.

Figure 1

(a) Schematic of the evanescently coupled photon counting detector. The detection layer is deposited on top of an ultraviolet (UV) laser–written silica waveguide structure. The calculated internal mode profile can also be seen at the front of the device schematic. (b) Microscope image of the fabricated TES on the optical waveguide. The TES dimension is 25 $\mu$m $\times$ 25 $\mu$m $\times$ 40 nm. The tungsten TES bisects the waveguide in the center. The niobium wiring attaches to the top and bottom of the TES and exits off the chip to the right. Additional arrows simply serve as alignment marks and guide.

Figure 2

(a) Plot showing how the overlap between the mode supported by the regions with and without the detector (blue) and the absorption coefficient ($\alpha /{\mathrm{cm}}^{-1}$, red) of both TE (dashed) and TM (solid) modes varies as the thickness of the layer is altered for a $25-\mu \mathrm{m}$-wide detector. The maximum absorption coefficient for the TM mode occurs at around 45 nm, and the preferential absorption of the TM-like mode is clear. (b) Plot showing how the absorption for the TE (dashed) TM (solid) mode changes as the length of the detector is increased while keeping the volume constant at $25\mu {\mathrm{m}}^3$. (c) The modeled normalized intensity distribution of the TM mode in the region without the detector. (d) The modeled normalized intensity distribution of the TM mode in the region with the detector for a 40-nm thickness. A slight distortion in the vertical direction can be seen along with a small amount of intensity near the detector layer.

Figure 3

Experimental scheme. A 35-kHz pulsed laser at 1550 nm delivers attenuated weak coherent state pulses to the waveguide device, which is situated inside a dilution refrigerator. The input light's polarization can be modified via a fiber polarization controller. The waveguide device along with the TES is cooled to about 12 mK. The TES is voltage biased, and the electrical output is fed into a SQUID circuit (not shown). The SQUID output is amplified at room temperature, and the signal is measured with data acquisition electronics.

Figure 4

Experimental results. (a) Photon pulse height distribution for a measured coherent state with $\langle n\rangle =0.986$ for the optimal TM polarization (${\eta }_{\text{det}}=7.2\%$). The inset shows the photon pulse-height distribution for the anti-optimal TE polarization. The measured mean photon number was $0.086$ (${\eta }_{\text{det}}=0.65\%$). (b) Photon pulse height distribution for the TES not evanescently coupled to the waveguide ($\langle n\rangle =1.03$; ${\eta }_{\text{det}}=0.0056\%$). (c) Electrical TES output traces for different numbers of photons in the weak laser pulse; the photon-number-resolving capability is clearly visible here. The TES recovery time is about 10 $\mu$s due to its low transition temperature. Note that this detector was optimized for its photon-number-resolving capability and the evanescent coupling strength, not for its recovery time.