Integrated transition edge sensors on lithium niobate waveguides

We show the proof-of-principle detection of light at 1550 nm coupled evanescently from a titanium in-diffused lithium niobate waveguide to a superconducting transition edge sensor. The coupling efficiency strongly depends on the polarization, the overlap between the evanescent field, and the detector structure. We experimentally demonstrate polarization sensitivity of this coupling as well as photon-number resolution of the integrated detector. The combination of transition edge sensors and lithium niobate waveguides can open the field for a variety of new quantum optics experiments.


I. INTRODUCTION
Integrated photonic circuits are widely used to realize compact and complex quantum optics experiments. ey enable scalable creation and processing of quantum states which can be used in communication, computation and simulation protocols to potentially outperform classical systems 1 . Lithium niobate is an established platform in the eld of classical integrated optics because of its high second-order susceptibility and electro-optic properties 2 . Titanium indi used waveguides in lithium niobate are surface-guiding and o er low-loss waveguiding in both polarization directions (TE and TM), across a broad frequency range, making them ideal for a range of quantum optical circuits. For example, integrated single-photon sources using parametric down conversion 3,4 can be combined with electro-optic modulators 5,6 and active or passive routing to enable a variety of applications in quantum optics 7 . In addition, low-loss ber pigtailing, by a aching single-mode bers directly to the waveguide end-faces ("bu -coupling"), enables high overall system e ciency 3 and compatibility with the existing communication network infrastructure.
Until now, detection from lithium niobate chips has been restricted to ber-coupled detectors, which adds an extra interface and associated losses. Of the various types of bercoupled detectors, those based on the breakdown of superconductivity o er the highest e ciency at telecom wavelengths and can be tailored for low timing-ji er, di erent photon numbers, or even photon-number resolution [8][9][10][11][12][13][14] . However, the positioning of ber-coupled detectors with respect to an integrated optical circuit is limited, and their scaling towards complex circuitry is challenging.
Integrated detectors, using the coupling from an evanescent eld of a waveguide into an on-chip detector, enable more complex circuitry, as they can be deposited at di erent positions inside the optical circuit. Furthermore, in an inline geometry, non-detected (and non-sca ered) photons rea) Electronic mail: jan.philipp.hoepker@upb.de main inside the waveguide; this geometry potentially allows for further processing of undetected photons 15 . On platforms such as silicon or III-V semiconductor waveguides the integration of superconducting nanowire single photon detectors (SNSPDs)  or transition edge sensors (TESs) 39,40 has been realized. However, the integration of single-photon detectors on lithium niobate waveguides is challenging 41,42 .
TESs operate at a transition stage between a superconductive and a normal resistance state 43,44 . When voltage-biased at their transition, the detector works as a micro-calorimeter, which is sensitive to temperature changes introduced by single photon absorption even at infrared wavelengths. Using their weak electron-phonon coupling, TESs made of thin-lm tungsten have shown remarkable properties in terms of detection e ciency (95 %-98 %) and energy resolution (0.29 eV-0.42 eV) 12,13 .
In this le er, we report on the rst proof-of-principle evanescent single-photon detection with a transition edge sensor on a lithium niobate waveguide.
is completes the toolbox for integrated quantum optics on this platform, adding integrated detection. We rst show the waveguide fabrication and detector fabrication in section II. In section III, we describe simulations which estimate the detection efciency. In section IV we show our experimental results including photon-number resolution up to six photons, system detection e ciency measurements for both polarizations, and rst results for the energy resolution and decay time.

II. WAVEGUIDE AND DETECTOR FABRICATION
e waveguide fabrication starts with an 80 nm titanium deposition on a congruent lithium niobate wafer using ebeam evaporation, followed by a positive photoresist development. Under vacuum contact lithography and subsequent wet etching, 5 µm, 6 µm, and 7 µm wide titanium stripes are formed.
ese stripes are di used into the lithium niobate substrate at 1060 • C creating an index gradient from 2.211 to 2.214 in TE-and 2.132 to 2.138 in TM-polarization at 1550 nm wavelength [45][46][47] .
is way both polarizations (TE and TM) can be guided with losses below 0.02 dB/cm. A 2.5 cm long sample consisting of 70 waveguides is cut and its end-faces are polished. To ensure the waveguide quality, loss measurements are executed using an interferometric technique described by Regener and Sohler 48 , where similar loss values were achieved.
A er ensuring the low-loss waveguiding, the TESs are deposited. Our integrated TES devices comprise a homogeneous 20 nm thick tungsten layer, deposited using magnetron spu ering. As 25 µm x 25 µm devices show high yield for ber-coupled TESs, the same structure is chosen in this work. We place three detectors per waveguide on top of a 2 nm amorphous silicon layer, which does not e ect the optical properties of the detectors or waveguide, and additional niobium-contact pads for wire-bonding using photo lithography. A micrograph image of one device is shown in gure 1.
One bene t of titanium in-di used waveguides on lithium niobate is the mode diameter which closely matches optical ber. With this, a single-mode ber ferrule can be directly glued to the polished waveguide end-face using a UVsensitive adhesive. In this device, we used standard singlemode ber (not polarization-maintaining), in order to simplify the pigtailing process. A precise, motorized alignment setup is used to ensure good coupling between the two interfaces. A thin layer of UV-glue with a ber-matching refractive index is deposited in between the waveguide end-face and the ber-ferrule and symmetrically cured using a ring of UV-LEDs for pre-curing and a UV-gun to fully harden the glue. To minimize displacement while cooling the pigtailed sample inside a cryostat, the thickness of the glue layer must be minimized and a symmetric spreading of the glue is preferable. At room-temperature a theoretical maximum coupling to ber of 92 % can be achieved with the given mode overlap 3 .

III. SIMULATIONS
As the guided modes inside the waveguide were optimized to match the mode diameter of a standard single-mode ber, the overlap of the waveguide mode and the tungsten layer is very small. erefore, the 25 µm x 25 µm x 20 nm device only sees a small part of the mode, as illustrated in gure 2. Finite-Element-Method and Finite-Di erence Beam Propagation Simulations were executed using a comercial modesolver to estimate the detector e ciency for the given structure, using the refractive index of the waveguides based on Edwards 45 , Jundt 46 , and Strake 47 as well as independently measured values for the refractive index of tungsten. From these simulations a strong polarization dependence in the absorption was calculated, with values of 1.3 % ±0.6 % in TMpolarization and 0.16 % ±0.06 % in TE-polarization. Although the absorption and therefore device e ciency is small, modications to the structure can be implemented to enhance the absorption as well as multiplexing several detetors 40 .

IV. RESULTS
We rst tested the device under ood illumination, which showed successful electrical connection and optical response for the 25 µm x 25 µm x 20 nm devices 42 . Following this, the device was pigtailed and installed inside a dilution refrigerator (DR), to investigate the sensitivity to the evanescent eld. e packaged device is robust and remains functional a er several temperature cycles. e pigtailed bers from the sample were spliced and connected from inside the DR to external FC/PC connectors and connected to an a enuated pulsed 1550 nm laser. As both end-faces of the waveguide were pigtailed, stable transmission through the waveguide at cold temperatures could also be veri ed. An overall transmission of 43 % at room temperature and 8 % ±2 % at 0.01 K was measured for both polarization directions. From this, a coupling e ciency to the waveguide of 66 % at room temperature and 28 % at 0.01 K is estimated. A er optimizing the detector output, photon traces with di erent peak heights corresponding to di erent photon numbers per pulse could clearly be observed, as shown in gure 3. e detector response for two detectors on the same waveguide was measured for di erent polarization se ings and both coupling directions. We observed a strong polarization sensitivity but a small in uence of the in-coupling direction, which veri es the evanescent coupling. When measuring sca ered light, one would expect a large in uence on coupling direction, as the detectors are placed not symmetrically along the waveguide, as illustrated in gure 4 b).
From the measured photon traces of each detector and for each polarization se ing, a mean photon trace can be calculated as a template 49 . From the convolution of each trace and the template, we construct histograms for individual detectors, polarization se ing, and in-coupling direction, as shown in gure 5. Also, an average (1/e) decay time of 1.6 µs can be calculated from a mean photon trace. By using a Gaussian t or by manually se ing individual thresholds, the histogram can be used to calculate a mean photon number. In addition, we can determine the energy resolution of the detectors from the histogram using the mean photon energy of the zero and one photon peak and its FWHM. We found an average energy resolution of 0.33 eV ±0.05 eV, which is similar to other platforms.
We measured the expected mean photon number using a spli ing-ratio method as illustrated in gure 4. In a rst step, the spli ing ratio between the two arms is investigated, using low a enuation on A enuator 1 and high a enuation on Attenuator 2, which leads to high transmission at the calibrated Powermeter 1 and a lower but still measurable signal at the calibrated Powermeter 2. Next, Powermeter 2 is exchanged with a ber that feeds into the cryostat and is pigtailed at the waveguide end-face. Using a high a enuation on A enuator 1, while keeping A enuator 2 xed, reduces the output in both arms. Measuring the output at Powermeter 1 can be used to calculate the single-photon ux in the other arm. From this, as well as an expected mean photon number, the maximum and minimum system detection e ciency (SDE) is calculated, depending on the polarization. is gives 0.23 % ±0.04 % and 0.06 % ±0.01 % for one detector and 0.19 % ±0.04 % and 0.07 % ±0.01 % for a second tested detector. e in-coupling direction has no in uence on the SDE within the measurement accuracy. e TES response was optimized by changing the polarization se ing and monitoring the mean photon number using an in situ histogramming. From our room-temperature loss measurements and simulations the maximum SDE was related to TM-polarization and the minimum SDE to TEpolarization. Further experiments using polarization maintaining bers should be executed to verify this assumption. From transmission measurements through the pigtailed waveguide, we subtract coupling losses into the chip and calculate a corrected maximum detection e ciency of 0.7 % per detector. Compared to simulations, the lower e ciency can be a ributed to a simpli ed model used to calculate the absorption in the simulations and to the polarization se ing accuracy.

V. CONCLUSION
In this paper, we demonstrated the rst proof-of-principle detection of single photons using TESs on lithium niobate waveguides. From the measured photon traces a histogram was calculated to con rm the photon-number resolution of the detectors. Light was coupled in from both directions into two detectors on the same waveguide and a strong polarization sensitivity was observed. Also, values for the average response decay time and energy resolution were calculated.
As an outlook, the detection e ciency can be signi cantly increased by changing the detector and/or waveguide geometry. As already shown on silica-on-silicon waveguides, additional tungsten ns (metallic absorptive strips deposited on the waveguide to conduct heat to the TES) can be used to increase the interaction length with the waveguide 40 . is can increase the evanescent coupling by at least one order of magnitude. Furthermore, the system detection e ciency can be enhanced by detector multiplexing.