Introduction

On-chip generation of diverse optical quantum states has been widely demonstrated, highlighting its considerable potential for advancing integrated quantum information processing applications1,2,3,4,5,6,7,8. As the field advances, the scalable and configurable generation and manipulation of quantum states across various degrees of freedom become increasingly essential5,6,9,10,11,12,13,14,15,16,17,18,19. Photon lifetime is one of the critical degrees of freedom in optical quantum states, predominantly determined by the bandgap of a two-level system or the cavity quality factor (Q factor) of a resonator-based photon emitter20. Photons with long lifetimes, on the order of nanoseconds, typically exhibit relatively narrow linewidths, ranging from a few MHz to several hundred MHz. They are indispensable for aligning with the atomic transition linewidths to enable effective photon-atom interactions, crucial for quantum repeater-based long-distance quantum secure communications21,22,23. On the other hand, to maximize the rate of quantum communication systems, the lifetime of single-photons τ must be compressed, since the communication bandwidth is proportional to 2π/τ, as governed by Fourier reciprocity13. Short-lived photons are also required in quantum optical coherent tomography24,25. Furthermore, the ability to configure photon lifetime is vital for many applications. Dynamically tunable photon lifetime schemes have been utilized as quantum random number generators10. Moreover, sources with configurable photon lifetime are highly promising for frequency-encoded qubits, a key component in scalable quantum information processing technologies26.

On the integrated photonics platform, research has shown that the control of photon lifetime can be realized by integrated modulation modules to shape spectra of single-photon wave packets9,12,13,14,15,16. However, implementing modulation after photon generation complicates the quantum system, introducing extra losses. In recent years, optical microresonators with spontaneous nonlinear effects have been extensively demonstrated as quantum light sources1,2,3,4,5,6,7,8,27,28,29,30,31,32,33,34,35,36. It has been shown that different photon lifetimes can be achieved by setting distinct coupling gaps between the resonator and the bus waveguide20. Still, modifying the lifetime of a photon becomes challenging once the devices of a quantum light source are fabricated. More importantly, the fundamental trade-off between efficiency and bandwidth imposes significant constraints on the design of high-performance configurable photon lifetime quantum light sources. That is, to generate photons with a short lifetime, the duration of light-matter interaction must be reduced (related to an increase of the bandwidth of the cavity), which in return severely deteriorates the generation efficiency35,36. Therefore, it is crucial to find other photon lifetime adjustment schemes that do not sacrifice the efficiency or the photon-pairs generation rate (PGR).

Recent advances in parity-time (PT)-symmetric systems have attracted substantial attention due to their easy implementation yet profound implications in photonics37,38,39,40,41. The non-Hermitian property of PT-symmetric systems, especially at the exceptional point (EP), where the eigenvalues and corresponding eigenvectors coalesce, gives rise to a range of unconventional light-matter interaction phenomena42,43. These properties enable wide applications in areas such as lasing44, enhanced sensing45,46, coherent perfect absorption47, optical frequency combs48,49, and beyond. In addition, we have shown in a recent work that by leveraging PT symmetry and breaking the uniformity of the resonance linewidth of the coupled microcavities, the trade-off between efficiency and bandwidth has been relieved for classical applications such as nonlinear optical signal processing50. This indicated that the trade-off between short photon lifetime and large PGR may be relieved via a similar linewidth-manipulation method. Indeed, integrated resonance-manipulated devices51,52,53,54 have been studied for several potential quantum applications, such as spectral-uncorrelated photon-pairs generation51,52 or high-coincidence efficiency photon-pairs sources53. Nevertheless, achieving direct and large-range control of cavity photon lifetime remains elusive.

In this work, we demonstrate a lifetime-configurable quantum light source by leveraging a dual-ring coupled PT-symmetric system. The PT symmetry introduces an extra degree of freedom through unbalanced losses to control the decay rates of microresonators, making it naturally suitable for adjusting the lifetime of photons. The maximum loss contrast near and away from the EP enables orders of magnitude control of the photon lifetime. Meanwhile, the PGR has maintained a moderately high value thanks to the undegraded pump enhancement via the selective linewidth manipulation technique. Experimentally, we demonstrate a near 20-fold variation in photon lifetime, from 7.1 ± 1.5 to 129.6 ± 1.9 ps. Energy-time entangled photon pairs are generated, exhibiting a typical raw two-photon interference visibility of 87.1 ± 1.1%. A typical heralded second-order autocorrelation \({g}_{H}^{\left(2\right)}\left(0\right)\) of 0.069 ± 0.001 is also obtained.

Results

System design and principle of operation

We demonstrate the configurable photon lifetime quantum light source using spontaneous four-wave mixing (SFWM) in a dual-ring coupled PT-symmetric system, as sketched in Fig. 1a. The main ring (red) with a radius \({R}_{1}\) is coupled to the auxiliary ring (blue) with a radius \({R}_{2}\). The intrinsic decay rates of the main and auxiliary rings are \({\gamma }_{1}\) and \({\gamma }_{2}\), respectively, and the coupling strength between the two rings is \(\kappa\). The auxiliary ring is coupled to a bus waveguide with a coupling decay rate \({\gamma }_{{{{\rm{c}}}}}\), resulting in a total decay rate \({\gamma }_{2}+{\gamma }_{{{{\rm{c}}}}}\). When \({\gamma }_{{{{\rm{c}}}}} \, \gg \, {\gamma }_{1,2}\), the decay rates of the two coupled rings become severely unbalanced, and a passive PT-symmetric system is thus formed47,50,55. Given that the radius ratio is set to \({R}_{1}=2{R}_{2}\), the free spectral ranges (FSRs) of the two rings (denoted as \({{FSR}}_{1}\) and \({{FSR}}_{2}\), respectively) differ by a factor of 2, as depicted by the intracavity photon density of state (DOS) in Fig. 1b. A thermal tuning module is fabricated upon the auxiliary ring to adjust the detuning between the two rings. When a voltage is applied to the thermal tuning electrode, the resonance of the auxiliary ring gradually aligns with that of the main ring, consequently reducing the Q factors of the corresponding resonances of the main ring (see Fig. 1c(ii) and d). The resonances influenced by this adjustment are designated as the signal and idler resonances, whereas the central resonance, which remains unchanged due to the FSR difference of the two rings, functions as the pump resonance (see Fig. 1d(iii)). The purpose of this design is twofold. On one hand, the symmetrically reduced Q factors of the signal and idler resonances result in shorter lifetimes of paired photons. On the other hand, the unaffected high-Q pump resonance ensures high SFWM efficiency, thereby maintaining a high PGR. Physically, this phenomenon arises because SFWM primarily occurs in the main ring, where the pump light is predominantly confined50, while the generated photons are emitted through the bus waveguide. The photon lifetime of the emitted photons is largely determined by the coupling decay introduced by the auxiliary ring.

Fig. 1: Coupled PT-symmetric microresonators system and the concept of configurable photon lifetime.
Fig. 1: Coupled PT-symmetric microresonators system and the concept of configurable photon lifetime.
Full size image

a Sketch of the coupled dual-ring system. The main ring (red) with a radius of \({R}_{1}\) is coupled to the auxiliary ring (blue) with a radius of \({R}_{2}\). The intrinsic decay of the main and auxiliary ring is \({\gamma }_{1}\) and \({\gamma }_{2}\), respectively. The coupling strength between two rings is \(\kappa\). The thermal tuning module is fabricated upon the auxiliary ring. The auxiliary ring is coupled to the bus waveguide with the coupling decay rate \({\gamma }_{{{{\rm{c}}}}}\). Photon pairs are generated via the spontaneous four wave-mixing (SFWM) that occurs in the main ring. b Normalized intracavity photon density of state (DOS) of the main ring, with yellow arrows indicating the thermal tuning of the resonance of the auxiliary ring (redshift in wavelength axis). c Photon DOS evolutions (i) without (w/o) and (ii) with (w/) PT-symmetry. d Detailed DOS of photon at different detuning statuses marked by the dashed lines in c (ii).

Without loss of generality, we consider in the following the evolution of signal resonance since the signal and idler resonances act similarly. According to temporal coupled-mode theory (TCMT), the non-Hermitian Hamiltonian of the system reads41,50

$$\hat{H}=\left[\begin{array}{cc}{\omega }_{1}-i\frac{{\gamma }_{1}}{2} & \kappa \\ \kappa \hfill & {\omega }_{2}-i\frac{{\gamma }_{2}+{\gamma }_{{{{\rm{c}}}}}}{2}\end{array}\right],$$
(1)

where \({\omega }_{1}\), \({\omega }_{2}\) are the resonant frequencies of the main ring and the auxiliary ring, respectively. Define \({{{\rm{\delta }}}}\omega={\omega }_{2}-{\omega }_{1}\) as the detuning between the two rings. It is worth noting that the Hamiltonian is defined in a lossy PT-symmetric configuration by factoring out an average loss \(({\gamma }_{1}+{\gamma }_{2}+{\gamma }_{{{{\rm{c}}}}})/4\), in which the unbalanced losses compose an equivalent gain-loss mechanism in this passive PT-symmetric system47,55. Solving the complex eigenfrequencies of the Hamiltonian when two rings are aligned, i.e., under the condition of δω = 0, they can be expressed as41,50

$${\omega }_{\pm }=\left[{\omega }_{1}-\frac{i}{4}\left({\gamma }_{2}+{\gamma }_{{{{\rm{c}}}}}+{\gamma }_{1}\right)\right]\pm \frac{1}{4}\sqrt{16{\kappa }^{2}-{\left({\gamma }_{2}+{\gamma }_{{{{\rm{c}}}}}-{\gamma }_{1}\right)}^{2}}$$
(2)

Discussion simplifies when the intrinsic losses of the two rings are identical, i.e., \({\gamma }_{1}={\gamma }_{2}\), which can be approximated in our design and fabricated devices. The EP occurs at the degeneracy point of complex eigenfrequencies, i.e., \(\kappa={\gamma }_{{{{\rm{c}}}}}/4\) (detailed analyses are given in Supplementary Note 1).

Consider a common situation that occurs in coupled ring-systems when the coupling between two rings is strong and ring-bus waveguide coupling loss is weak, i.e., \(\kappa > {\gamma }_{{{\mathrm{1,2}}}} \, \gg \, {\gamma }_{{{{\rm{c}}}}}\)10. The system exhibits a typical mode splitting at \({{{\rm{\delta }}}}\omega=0\), as mapped in Fig. 1c(i). In contrast, the loss of the coupling decay \({\gamma }_{{{{\rm{c}}}}}\) in our design is much larger than the intrinsic losses \({\gamma }_{{{\mathrm{1,2}}}}\). When the coupling coefficient is set to be \(\kappa \approx {\gamma }_{{{{\rm{c}}}}}/4\), the system operates near the EP, as mapped in Fig. 1c(ii). A merged single resonance with a significantly low-Q factor is formed at \({{{\rm{\delta }}}}\omega=0\), which is naturally suitable for generating short-lived photons.

The photon lifetime adjustment process is achieved by changing the detuning between two rings using a thermal tuning module, as illustrated in Fig. 1a, b and d. Initially, the auxiliary ring with resonant frequency \({\omega }_{2}\) is misaligned with the main ring \({\omega }_{1}\). The signal and idler resonances can be regarded as resonances from a single ring under quasi-critical coupling with a linewidth of \(\Delta \omega={2\gamma }_{1}\) when detuning is large, e.g., \({{{\rm{\delta }}}}\omega \approx {{FSR}}_{1}/2\) (the high-Q pump resonance functions in the same way. Detailed analyses refer to Supplementary Note 2)50,56. Therefore, the signal and idler resonances are in high-Q state in this case, as shown in Fig. 1d(i), and the photon lifetime is calculated by20,57,58.

$${\tau }_{{{{\rm{high}}}}-{{{\rm{Q}}}}}=\frac{1}{\Delta \omega }=\frac{1}{{2\gamma }_{1}}$$
(3)

As the voltage applied to the thermal tuning electrode of the auxiliary ring gradually increases, the resonance of the auxiliary ring approaches that of the main ring (Fig. 1d(ii)) until they are completely aligned (\({{{\rm{\delta }}}}\omega=0\)), as shown in Fig. 1d(iii). In case (iii), the signal and idler resonances operate near the EP59, with a linewidth of \(2{{{\rm{Im}}}}\left({\omega }_{+}\right)\approx {\gamma }_{{{{\rm{c}}}}}\), according to Eq. (2). At this point, the photon lifetime is defined as the inverse of the mode linewidth by

$${\tau }_{{{{\rm{low}}}}-{{{\rm{Q}}}}}=\frac{1}{{\gamma }_{{{{\rm{c}}}}}}$$
(4)

It can be derived from Eqs. (3) and (4) that the maximum achievable contrast of the photon lifetime is \({\gamma }_{{{{\rm{c}}}}}/2{\gamma }_{1}\). Thus, increasing the coupling decay rate of the auxiliary and the bus-waveguide or reducing the intrinsic loss of the main ring can enhance the contrast ratio in the tunability of the photon lifetime.

Device characterizations

We design the device on a low-loss integrated silicon nitride-on-insulator (SNOI) platform to experimentally demonstrate the configurable photon lifetime quantum light source. The radius of the main ring is designed as \({R}_{1}=228\,\mu {{{\rm{m}}}}\) (with a FSR close to 100 GHz to match the International Telecommunication Union’s (ITU) dense wavelength division multiplexer (DWDM) channels), and the radius of the auxiliary ring is \({R}_{2}=114\,\mu {{{\rm{m}}}}\). The transmission spectra from 1534 nm to 1540 nm of the system under different voltages, i.e., 0 V and 6.6 V, are plotted in Fig. 2a, b, respectively (for details of spectra characterizations, see Methods and Supplementary Note 3). Under the 0 V bias, the two rings are misaligned, and the overall transmission spectrum exhibits a series of quasi-critical-coupled high-Q resonances from the main ring (\({{FSR}}_{1}\) = 0.78 nm). Between each two high-Q resonances, there are auxiliary ring resonances (yellow lines) with a relatively low extinction ratio (\({{FSR}}_{2}\) = 1.56 nm) due to the auxiliary ring being in a state of over-coupling with the bus waveguide. Under the bias of 6.6 V, the resonances of the main ring and the auxiliary ring are aligned, resulting in a distribution of alternating low-Q and high-Q resonances in the transmission spectrum. Resonances that are three \({{FSR}}_{1}\) away from the pump resonance on both sides are selected as the signal and idler resonances to avoid imperfect pump rejection filtering, which might cause large noise photon counts. To be concrete, the resonance centred at 1536.9 nm (C51 DWDM channel) is chosen as the pump resonance. The resonances centred at 1534.6 nm (C48) and 1539.3 nm (C54) are chosen as the signal and idler resonances, respectively, as shown by the dashed areas in Fig. 2a, b.

Fig. 2: Spectra characterizations.
Fig. 2: Spectra characterizations.
Full size image

a System transmission spectrum under 0 V bias (left axis). A series of high-Q resonances is from the main ring, and the highlighted yellow lines are the resonances of the auxiliary ring. The free spectral ranges (FSRs) of the main ring is \({{FSR}}_{1}\) = 0.78 nm, and the auxiliary ring is \({{FSR}}_{2}\) = 1.56 nm. b System transmission spectrum under 6.6 V bias (left axis). The green circles in (a) and (b) are the marks of the high-Q factors of the main ring’s resonances (right axis). The orange circles in (b) are the low-Q factors of merged resonances. The dashed grey regions represent the locations of signal (C48), pump (C51), and idler (C54) resonances. Under 0 V bias, the Q factors of the three resonances are: 1.53 × 105, 1.53 × 105, and 1.43 × 105, respectively; Under 6.6 V bias, the Q factors are 8.03 × 103, 2.05 × 105, and 1.12 × 104, respectively.

The Q factors of all resonances are fitted and marked by the green (high-Q) and orange (low-Q) circles (the right axis). Before and after thermal tuning, the Q factors of the resonances in the main ring that coupled to the auxiliary ring exhibit a significant reduction, whereas the unaffected resonances keep a similar high-Q factor under quasi-critical coupling. According to TCMT, the system parameters are extracted by the fittings: \({\gamma }_{1}={\gamma }_{2}=3.0\,{{{\rm{GHz}}}}\), \({\gamma }_{{{{\rm{c}}}}}=146.8\,{{{\rm{GHz}}}}\), \(\kappa=45.5\,{{{\rm{GHz}}}}\). It can be seen that \(\kappa={\gamma }_{{{\rm{c}}}}/4\) is approximately satisfied, indicating that the system operates near the EP.

Correlated photon-pairs generation

Next, we characterize the photon lifetime of the correlated photon pairs generated from the lifetime configurable device. Figure 3a plots the experimental setups for photon-pairs generation. The pump light from a continuous wave (CW) laser is passed through an erbium-doped fibre amplifier (EDFA), a variable optical attenuator (VOA), and a polarization controller (PC), and then into a 90/10 beam splitter (BS). 10% of the pump enters a power metre (PM) to monitor the input power entering the chip. The remaining 90% of the pump is filtered out using a high-isolation (with an isolation of ≥ 120 dB) band-pass DWDM at the C51 channel and then coupled into the chip. The band-pass DWDM is employed here to filter out noise photons at the signal and idler wavelengths of the preceding system. The second band-stop DWDM is set after the generation of correlated photon pairs as the pump rejection filter. Figure 3b gives the test system of correlated photon pairs. A tunable bandwidth filter (TBF) is applied to separate the signal and idler wavelengths, and then send the photons to the superconducting nanowire single-photon detectors (SNSPDs, with a detection efficiency of about 90% and a dark count rate of about 30 Hz). The output signals of SNSPDs are further connected to a time-to-digital converter (TDC) to record the coincidence events.

Fig. 3: Experimental setups.
Fig. 3: Experimental setups.
Full size image

a Generation of correlated photon pairs from PT symmetry-enabled lifetime tunable device. b Test of correlation properties (coincidence counts). c Characterization of energy-time entanglement with two-photon interference using a folded Franson interferometer. d Test of heralded single-photon using Hanbury Brown-Twiss (HBT) interferometer. CW continuous wave, EDFA erbium-doped fibre amplifier, VOA variable optical attenuator, PC polarization controller, BS beam splitter, PM power metre, DWDM dense wavelength division multiplexer, DUT device under test, VS voltage source, TBF tunable bandwidth filter, SNSPDs superconducting nanowire single-photon detectors, TDC time-to-digital converter, UMI unbalanced Michelson interferometer.

The coincidence counts (CCs) of generated photon pairs under the bias voltages of 0 V and 6.6 V are recorded in Figs. 4a, b, respectively. A 1000-time Monte Carlo simulation is used to fit the coincidence count histograms32. A composite distribution of exponential and Gaussian functions is applied to numerically examine the lineshapes60,61, since the timing jitters of the detection systems are included in the coincidence peaks. The total standard deviations of the two counting histograms are \({w}_{1}=153.7\pm 1.3 \ {{{\rm{ps}}}}\) and \({w}_{2}=83.0\pm 1.0 \, {{{\rm{ps}}}}\), respectively. Accordingly, the fittings giving the lifetimes \(\tau\) of long and short-lived photons are 129.6 ± 1.9 ps and 7.1 ± 1.5 ps, respectively, and the timing jitter of the detection system \(\sigma\) is 82.8 ± 0.9 ps. The uncertainties are extracted from the Monte Carlo simulations. A near 20-fold photon lifetime adjusting contrast is achieved, matching the Q factor variations in spectral tests given earlier (detailed analyses of the extraction of photon lifetime are given in Supplementary Note 4). From TCMT, the expected photon lifetimes before and after the alignment are \({\tau }_{{{{\rm{high}}}}-{{{\rm{Q}}}}}=167.7 \, {{{\rm{ps}}}}\) and \({\tau }_{{{{\rm{low}}}}-{{{\rm{Q}}}}}=6.5 \, {{{\rm{ps}}}}\), respectively, according to the relation between decay rate and photon lifetime (Eqs. (3) and (4)). Here, the photon lifetime predicted for long-lived photons deviates from the experimental results. This discrepancy arises because, in the long-lifetime regime, the additional losses introduced by the auxiliary ring to the main ring are neglected in TCMT. Nevertheless, the expressions derived from the TCMT can be used as a simple model to roughly approximate the tunable range of the photon lifetime.

Fig. 4: Experimental results of photon-pairs generation and lifetime adjustment.
Fig. 4: Experimental results of photon-pairs generation and lifetime adjustment.
Full size image

a Coincidence counts without thermal tuning (0 V bias). b Coincidence counts with thermal tuning (6.6 V bias). The yellow lines are the superimposed 1000-time Monte Carlo simulations for carrying out the photon lifetimes and the timing jitters of the detection system. The total standard deviations are \({w}_{1}=153.7\pm 1.3\,{{{\rm{ps}}}}\) and \({w}_{2}=83.0\pm 1.0\,{{{\rm{ps}}}}\), respectively. The grey backgrounds are for ease of visualization. c and d are the coincidence count rates with respect to the on-chip pump powers without and with thermal tuning, respectively. The error bars are derived by a Poissonian photon distribution. e and f are the corresponding CARs. The error bars are determined by repeated measurements of accidental counts.

Next, Fig. 4c, d record the peak values of CCs with respect to the on-chip pump powers, from 0.25 mW to 14.12 mW. The CC rates scale with the square of the pump power, aligning with the theoretical predictions for the PGR in SFWM35. The error bars are derived from Poissonian photon-counting statistics28,32. The corresponding coincidence-to-accidental ratios (CARs) are then recorded in Fig. 4e, f. Under the 0 V bias, the maximum CAR reaches 31, while under the 6.6 V bias, the CAR reaches 170.

It is noteworthy that the coincidence count rate increases surprisingly, even though the Q factors of signal and idler resonances are decreased when comparing the states before and after the rings are aligned. This can be attributed to the over-coupling of signal and idler resonances with an enhanced photon escape probability, which can be effectively modelled by taking into account the different Q factors of evolved three resonances in the SFWM process35,36,62,63,64. Theoretical calculation reveals an on-chip PGR of high-Q case (0 V bias) of 85 kHz (under 14.12 mW on-chip pump); and a higher PGR of 140 kHz is expected after the linewidth manipulation (6.6 V bias), agreeing well with the experimental results in Fig. 4c, d (~1.65 times larger PGR in the linewidth-manipulated case), excluded the system losses and detection efficiency of SNSPDs (see Methods and more details in Supplementary Note 4).

Energy-time entanglement and heralded single-photon generation

We perform the energy-time entanglement of the short-lived correlated photon pairs generated by CW pumping. The quantum entanglement can be characterized through a Franson-type two-photon interference experiment, by violating Bell’s inequality65,66. As shown in Fig. 3c, we measure two-photon interference fringes using a folded Franson interferometer configuration implemented via an unbalanced Michelson interferometer (UMI)28,32,67,68. This arrangement can be regarded as a folded version of the standard Franson interferometer, in which the two unbalanced interferometers for the signal and idler photons are effectively combined into a single interferometer, employing the symmetry of the original configuration. Figure 5a, b give the CCs of destructive and constructive two-photon interference, respectively. The single-side count rates of signal and idler photons are recorded in the upper panel of Fig. 5c, which are near constants (about 40.9 kHz and 39.6 kHz on average, respectively), indicating that single-photon interference does not occur. The measured interference fringe with respect to the phase is further shown in the lower panel of Fig. 5c. The green circles are the coincidence events in 20 seconds, and the dashed green curve is the fit using a 1000-time Monte Carlo method28,32. The interference visibility is calculated as 87.1 ± 1.1% (without subtracting the accidental coincidence counts), which exceeds 70.7%, indicating effective energy-time entanglement violating the Bell inequality65,66,69.

Fig. 5: Short-lived photon pairs test of two-photon interference and energy-time entanglement.
Fig. 5: Short-lived photon pairs test of two-photon interference and energy-time entanglement.
Full size image

a Two-photon destructive interference with respect to delay time. b Two-photon constructive interference with respect to delay time. c Single side count rates of signal and idler photons (upper panel), as well as the two-photon interference fringe with a 1000-time Monte Carlo method fit (lower panel). The error bars are derived by Poissonian photon distribution.

Lastly, the heralded single photons can be obtained based on the photons generated in pairs. The heralded second-order autocorrelation \({g}_{H}^{\left(2\right)}\left(\tau \right)\) is measured by a Hanbury Brown–Twiss (HBT) interferometer70, as shown in Fig. 3d. In the experiment, the signal photons are detected directly by SNSPD, while the idler photons are detected with a delay time after passing through the 50:50 beam splitter, and the threefold coincidence events are recorded by TDC. The \({g}_{H}^{\left(2\right)}\left(\tau \right)\) of heralded single photons (idler) and their heralding photons (signal) is given in Fig. 6. The zero-delay value \({g}_{H}^{\left(2\right)}\left(0\right)\) is measured to be 0.069 ± 0.001 with a heralding rate of 256 kHz under an 11 mW on-chip pump, indicating that it can be used as a short-lifetime heralded single-photon source.

Fig. 6: Measurement of heralded second-order autocorrelation.
Fig. 6: Measurement of heralded second-order autocorrelation.
Full size image

\({g}_{H}^{\left(2\right)}\left(0\right)\) is tested to be 0.069 ± 0.001, showing good single photon properties. The error bars are derived by a Poissonian photon distribution.

Discussion

We experimentally demonstrated a lifetime configurable quantum light source on a SNOI chip harnessing PT-symmetry. By designing a coupled dual-microring structure with a radius ratio of 2:1 and fabricating a thermal tuning electrode on the auxiliary ring to control the detuning between the rings, the device that operates near or away from the EP can realize efficient lifetime adjustment without sacrificing SFWM efficiency. A near 20-fold photon lifetime tuning range from 7.1 ± 1.5 ps to 129.6 ± 1.9 ps has been achieved. The results for photon lifetimes are obtained from the Monte Carlo simulations of the coincidence count histograms, which are very close to the reciprocals of the linewidths obtained in the spectral measurements.

As a proof-of-concept device, the tuning range of the photon lifetime in the experiment is limited by the relatively low intrinsic-Q factor of the main microresonator. Further reducing the waveguide loss, for example, by widening the waveguide of the microring, can enhance the intrinsic-Q factor of the microcavity, thereby allowing for a larger range of photon lifetime adjustment (more details refer to Supplementary Note 5). As verified by the experimental results, we emphasize that our design realizes independent photon lifetime manipulation of the pump and signal (idler) resonances, enabling the adjustment of photon lifetime while maintaining a high PGR. The trade-off between photon lifetime and SFWM efficiency is mitigated. For practical applications, this is highly beneficial for achieving high-quality quantum light sources with configurable lifetimes, not limited to single photon sources but also the squeezed light sources, quantum frequency combs, and more. We envision that the lifetime-configurable quantum light sources based on the PT-symmetrically coupled microring system, due to their unique characteristics, can significantly benefit quantum information technologies, including applications in high-speed quantum key distribution71,72,73, frequency-encoded quantum communication74, quantum computing75, and beyond.

Methods

Chip fabrication

The chip is fabricated by LIGENTEC (Switzerland) using a standard thick silicon nitride waveguide process. The waveguide dimensions are 1000 nm × 800 nm (width × height), supporting the fundamental mode only.

Spectra characterizations

A tunable continuous wave (CW) laser with an output power of −15 dBm is scanned from 1520 nm to 1580 nm to characterize the spectra of the coupled ring system. The fibre-to-chip coupling loss is estimated to be 4 dB per facet. A complete analysis of spectra and Q factor fittings can be found in Supplementary Note 3.

Coincidence measurements

For coincidence measurement between the signal and idler photons, we use a high-resolution time-to-digital convertor (TDC) with a time resolution of 1 ps. The timing jitters of signal and idler channels are tested to be \({\sigma }_{1}=68.4\pm 0.3 \, {{{\rm{ps}}}}\) and \({\sigma }_{2}=49.1\pm 0.3 \, {{{\rm{ps}}}}\), and the total timing jitter in coincidence measurement can be calculated as \({\sigma }_{{{{\rm{test}}}}}=\sqrt{{\sigma }_{1}+{\sigma }_{2}}=84.1 \, {{{\rm{ps}}}}\) (see Supplementary Note 4 for detailed information), which is very close to the fittings of coincidence histograms. The CAR values in Fig. 4c, d are calculated by \({CAR}=(C-A)/A\), where \(C\) and \(A\) are the overall and accidental coincidence counts obtained from the peak and background of the coincidence counting spectra.

Photon-pairs generation rate

We theoretically calculate that a PGR = 85 kHz in the high-Q case (0 V bias), and a PGR = 140 kHz in the linewidth-manipulated case (6.6 V bias), respectively, under 14.12 mW pump power, using the following method35,36,62,63

$${PGR}=\frac{8{v}_{g}^{4}}{{\omega }^{3}}{\left(\frac{2\gamma P}{L}\right)}^{2}\frac{{Q}_{{{{\rm{p}}}}}^{4}{Q}_{{{{\rm{s}}}}/{{{\rm{i}}}}}^{3}}{{Q}_{{{{\rm{e}}}},{{{\rm{p}}}}}^{2}{Q}_{{{{\rm{e}}}},{{{\rm{s}}}}/{{{\rm{i}}}}}^{2}}$$

where \(\omega\) is the resonance frequency, \({v}_{g}\) is the group velocity, \(\gamma\) is the nonlinear coefficient, \(P\) is the pump power, \(L=2\pi {R}_{1}\) is the circumference of the main ring, respectively. \({Q}_{{{{\rm{p}}}}}\) is the loaded Q factor of the pump resonance, \({Q}_{{{{\rm{s}}}}/{{{\rm{i}}}}}\) is the loaded Q factor of the signal (idler) resonance, \({Q}_{{{{\rm{e}}}},{{{\rm{p}}}}}\) is the external Q factor of pump resonance, \({Q}_{{{{\rm{e}}}},{{{\rm{s}}}}/{{{\rm{i}}}}}\) is the external Q factor of the signal (idler) resonance, respectively. Given that the intrinsic Q of the resonator is \({Q}_{{{\mathrm{int}}}}\), the relation between loaded Q, and external, intrinsic Q is \(1/{Q}_{{{{\rm{loaded}}}}}=1/{Q}_{{{\mathrm{int}}}}+1/{Q}_{{{{\rm{e}}}}}\). The Q factors are obtained from the spectral analyses (please refer to Supplementary Note 3). The key parameters used here are: \(\omega=2\pi c/\lambda\) (where \(c\) is the light velocity in vacuum, λ = 1535 nm is the corresponding wavelength), \({v}_{g}=c/2.1 \, {{{\rm{m}}}}{s}^{-1}\), \(\gamma=1 \, {W}^{-1}{{{{\rm{m}}}}}^{-1}\), \(P=14.12 \, {{{\rm{m}}}}W\), \(L=2\pi \times 228 \, \mu{{{\rm{m}}}}\).