Quantum Pulse Gate: Mode-Selective Conversion

• Quantum pulse gate is a method leveraging spectrally engineered sum frequency generation to selectively convert specific broadband modes of ultrafast quantum pulses.
• It employs tailored pump pulse shaping and group-velocity matching in nonlinear waveguides to achieve high fidelity and efficient mode conversion.
• It enables advanced applications such as temporal-mode tomography, high-dimensional quantum key distribution, and programmable time-frequency mode sorting.

A quantum pulse gate (QPG) is a method for accessing the intrinsic broadband spectral mode structure of ultrafast quantum states of light by means of spectrally engineered sum frequency generation (SFG). In its original formulation, a weak multimode quantum pulse and a strong, shaped classical gating pulse interact in a nonlinear waveguide such that only the spectral broadband mode matched to the gate is upconverted to a new frequency, while orthogonal modes are minimally converted. In the subsequent literature, the QPG became a central device for temporal-mode (TM) selective manipulation, full modal characterization, high-dimensional state tomography, multi-output demultiplexing, and programmable time-frequency mode sorting at the single-photon level (Eckstein et al., 2010, Ansari et al., 2017, Serino et al., 2022, Serino et al., 2024).

1. Modal degree of freedom and conceptual scope

Ultrafast quantum light pulses possess a rich intrinsic broadband spectral mode structure. In the language used for biphoton states, these orthogonal broadband spectral modes are often identified with Schmidt modes; in the later temporal-mode literature, they are treated as field-orthogonal wave-packet modes that span an infinite-dimensional Hilbert space. Hermite-Gaussian functions in frequency appear repeatedly as a convenient modal basis for both analysis and pulse shaping (Eckstein et al., 2010, Serino et al., 2022).

The QPG was introduced to overcome a specific measurement and control problem. Ordinary detectors and narrowband filters cannot resolve the underlying broadband modes: they either respond to all of them indiscriminately or, in the case of very narrow filters, reduce the signal dramatically and destroy pulse properties. The QPG instead provides selective and efficient access to individual broadband modes while preserving their ultrafast characteristics and orthogonality. This is the essential distinction between mode-selective frequency conversion and conventional spectral filtering (Eckstein et al., 2010).

Within quantum information processing, the importance of this modal selectivity follows from the structure of the encoding space. Temporal modes can support arbitrary superpositions, mutually unbiased bases (MUBs), and high-dimensional alphabets. A single selected mode is therefore not merely a spectral feature but a projective degree of freedom in a high-dimensional Hilbert space. This suggests why QPGs recur in work on state tomography, high-dimensional quantum key distribution (HD-QKD), and time-frequency mode sorting (Ansari et al., 2017, Serino et al., 2022, Serino et al., 2024).

2. Theoretical formulation as mode-selective frequency conversion

The QPG is based on $\chi^{(2)}$ three-wave mixing in the strong-pump regime. For SFG, the relevant joint spectral distribution is written as

$$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$

where $\alpha(\omega_\mathrm{p})$ is the spectral amplitude of the gating pulse and $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$ is the phasematching function determined by dispersion and poling. In the original proposal and later generalization, this transfer function is Schmidt decomposed into orthonormal input and output pulse modes,

$$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \sum_j \kappa_j \varphi_j(\omega_\mathrm{i}) \psi_j(\omega_\mathrm{o}),$$

with $\sum_j \kappa_j^2 = 1$ (Eckstein et al., 2010, Brecht et al., 2011).

Broadband mode operators are then defined by

$$\hat{A}_j = \int d\omega_\mathrm{i}\,\varphi_j(\omega_\mathrm{i})\hat{a}(\omega_\mathrm{i}), \qquad \hat{C}_j = \int d\omega_\mathrm{o}\,\psi_j(\omega_\mathrm{o})\hat{c}(\omega_\mathrm{o}).$$

In this representation, the SFG interaction becomes a sum of independent channels, and in the single-mode limit the time-integrated Hamiltonian reduces to a beamsplitter-like transformation on one pair of pulse modes,

$$\hat{H}_{\mathrm{int}} = \hbar \theta \left(\hat{A}_0 \hat{C}_0^\dagger + \hat{A}_0^\dagger \hat{C}_0\right),$$

with conversion efficiency

$\eta = \sin^2\theta.$

The original QPG paper gives the same physical interpretation in terms of a mode-selective beamsplitter for broadband spectral modes, with the effective coupling governed by the overlap with the shaped gate and the nonlinear interaction strength (Eckstein et al., 2010, Brecht et al., 2011).

The practical criterion for ideal QPG operation is that only one Schmidt coefficient is significant. In that regime, the device acts selectively on a single input mode, frequency-converting it into a fixed output mode while leaving orthogonal input modes unaltered. A common misconception is to identify the QPG with generic frequency conversion; the defining feature is not upconversion alone, but engineered single-mode selectivity in the Schmidt basis (Eckstein et al., 2010, Brecht et al., 2011).

3. Spectral engineering, group-velocity matching, and single-output implementations

Mode selectivity is obtained by engineering the SFG transfer function. The original proposal emphasizes a narrow phasematching function, specifically a phase-matching bandwidth much narrower than the gate pulse spectrum, so that the transfer function becomes separable and unwanted frequency correlations are frozen out. Later work reformulates the same requirement as group-velocity matching together with broad bandwidth, which makes the phasematching function effectively horizontal or vertical in the joint spectral plane and promotes single-mode conversion (Eckstein et al., 2010, Serino et al., 2022).

The canonical implementation uses a periodically poled lithium niobate (PPLN) waveguide. The original implementation example specifies a PPLN waveguide of $8 \times 5\,\mu\mathrm{m}^2$, length $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$0 mm, and poling period $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$1, operated at $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$2 for phasematching. The wavelengths are input at 1550 nm, gating pulse at 870 nm with spectral FWHM 0.635 nm / 2 ps pulse, and output at 557 nm. In this setting, pulse shaping of the bright SFG pump beam allows different orthogonal broadband modes to be addressed individually and extracted with near unit efficiency, and the paper illustrates mode selectivity for up to the 10th broadband mode with fidelity exceeding 99% (Eckstein et al., 2010).

Programmable pulse shaping is the operational control layer of the QPG. Changing the gate from a Gaussian spectrum for $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$3 to a first-order Hermite-Gaussian for $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$4, and so on, tunes the device to a different input broadband mode. In the generalized treatment, the output mode is set by phasematching while the input mode is arbitrarily controlled by the gating pulse. This asymmetry is central to the distinction between the QPG and its reverse counterpart, the quantum pulse shaper (QPS), discussed below (Eckstein et al., 2010, Brecht et al., 2011).

4. Measurement tomography and calibrated operation

The transition from a conceptual mode-selective converter to a calibrated measurement device was established by temporal-mode measurement tomography of a QPG. In that work, weak coherent states in well-defined temporal modes are used as probes, and a full set of measurement operators is reconstructed for a device operating on a 7-dimensional space. The abstract reports an average fidelity of 0.85 to a theoretically ideal device, followed by calibrated high-dimensional TM state tomography with 0.99 fidelity (Ansari et al., 2017).

The formalism treats each pump setting $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$5 as defining a measurement operator,

$$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$6

with measured mean converted photon number

$$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$7

for an input probe state $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$8. Reconstruction is performed by measurement tomography via linear inversion (weighted least squares), under Hermiticity and positive-semidefiniteness constraints. Diagonalization of the reconstructed operators yields the measured eigenmodes and conversion efficiencies, so the QPG is characterized as an actual POVM rather than assumed to be an ideal projector (Ansari et al., 2017).

The experimental implementation reported 5 or 7 Hermite-Gaussian modes, probe states drawn from a full set of $$G(\omega_\mathrm{i}, \omega_\mathrm{o}) = \alpha(\omega_\mathrm{p})\,\phi(\omega_\mathrm{i}, \omega_\mathrm{o}),$$9 MUBs, and, for $\alpha(\omega_\mathrm{p})$0, a total of 3136 measurements. The detailed results include filtered measurement-operator purity and fidelity of 0.92 and 0.91 in 5D, and 0.81 and 0.85 in 7D. Using the characterized measurement operators in the reconstruction step, the reported state-tomography fidelities rise to 0.991 in 5D and 0.988 in 7D. The practical implication stated in the work is that small imperfections in single-mode selectivity can be compensated via proper operator calibration (Ansari et al., 2017).

This calibration result addresses another recurrent misconception: an experimental QPG need not behave as an ideal one-mode projector to be useful in high-dimensional protocols. What matters is that the implemented measurement operators are known accurately enough for subsequent state estimation and protocol analysis.

5. Multi-output quantum pulse gates and programmable mode sorting

The single-output QPG is intrinsically limited to one temporal mode at a time. The multi-output quantum pulse gate (mQPG) removes that limitation by combining custom poling patterns with multipeak pump shaping so that several phase-matched output channels coexist in one device. In the 2022 realization, alternating poled and unpoled regions create multiple phase-matching peaks, and the overall transfer function becomes a sum of matched input-output mappings. Each output channel corresponds to a POVM element

$\alpha(\omega_\mathrm{p})$1

with detection probabilities

$\alpha(\omega_\mathrm{p})$2

Experimentally, this mQPG demultiplexed five-dimensional TMs of single photons with an average fidelity of $\alpha(\omega_\mathrm{p})$3, operated on any basis from a set of 6 five-dimensional MUBs, and enabled resource-efficient state tomography with an average fidelity of $\alpha(\omega_\mathrm{p})$4 (Serino et al., 2022).

A further extension demonstrated programmable time-frequency mode sorting of single photons with a multi-output QPG that can switch between pulse modes, frequency bins, time bins, and their superpositions. That device was characterized through detector tomography in 3 and 5 dimensions and reached a fidelity up to $\alpha(\omega_\mathrm{p})$5 at the single-photon level. The same work introduced a “fancy” frequency-bin mode-sorting approach with linear scaling, $\alpha(\omega_\mathrm{p})$6 pump bins for $\alpha(\omega_\mathrm{p})$7 modes versus $\alpha(\omega_\mathrm{p})$8 for standard, and reported full operation in $\alpha(\omega_\mathrm{p})$9 and $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$0 (Serino et al., 2024).

Architecture	Demonstrated task	Reported performance
Single-output QPG	Full modal characterization and calibrated TM state tomography	Average fidelity 0.85 to a theoretically ideal device in 7D; tomography with 0.99 fidelity
Multi-output QPG	Demultiplexing five-dimensional TMs and complete decoding for 6 five-dimensional MUBs	Average fidelity $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$1; state tomography $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$2
Programmable mQPG mode-sorter	Switching among pulse modes, frequency bins, time bins, and superpositions	Fidelity up to $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$3 at the single-photon level

The detector technology in these demonstrations also matters to the reported performance. In the 2022 mQPG work, the CCD spectrograph reconstruction yielded average fidelity $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$4 and purity $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$5, whereas the ToF spectrograph gave $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$6. The same paper reports selectivity per basis of 61%–78% with ToF and 92% measured with CCD. This indicates that the distinction between internal mode selectivity and measurement-system resolution is operationally important when assessing mQPG performance (Serino et al., 2022).

6. Applications, complementary devices, and operational limits

The QPG was proposed as a tool for preparation of pure heralded single photons by purifying mixed-mode photons from typical parametric down-conversion (PDC) sources, for spectral-temporal demultiplexing of ultrafast pulses into orthogonal quantum channels, and for manipulation of multi-mode squeezed states relevant to quantum metrology. In later work, the mQPG is explicitly positioned as a receiver for high-dimensional quantum key distribution, as well as a device for efficient quantum state tomography. The programmable sorter extends this role to pulse modes, frequency bins, time bins, and superpositions, and the 2024 work identifies potential use as a Hadamard receiver (Eckstein et al., 2010, Serino et al., 2022, Serino et al., 2024).

The broader program of engineered frequency conversion also includes the quantum pulse shaper (QPS), introduced as the difference-frequency generation (DFG) analogue of the QPG. The QPG selects an arbitrary pulse mode from a multimode input state, whereas the QPS enables the generation of specific pulse modes from an input wavepacket with Gaussian-shaped spectrum. In the generalized account, the QPG and QPS together allow arbitrary manipulations of the pulse-mode structure of ultrafast pulsed quantum states. A plausible implication is that the pair functions as a read/write interface for pulse-mode quantum optics: the QPG performs mode-selective extraction or measurement, while the QPS performs tailored mode generation (Brecht et al., 2011).

Operational limits are stated explicitly in the literature. For single-output devices, the limitation is one TM at a time. For multi-output devices, scalability is limited by pump bandwidth, phase-matching, and detection resolution. The 2024 sorter work also makes clear that different detection architectures trade fidelity against real-time capability: the CCD spectrograph is high-fidelity but slow readout, while the time-of-flight spectrograph is real-time with lower frequency resolution. Within those constraints, the QPG family defines a consistent technical architecture: engineered $\phi(\omega_\mathrm{i}, \omega_\mathrm{o})$7 waveguides, modal programmability through pulse shaping, and calibrated POVM-level characterization for high-dimensional time-frequency quantum information (Serino et al., 2022, Serino et al., 2024).