---
title: "How a pulse gate approaches a single-mode measurement"
description: "Connect sum-frequency transfer kernels to pump shaping, phase matching and detection effects, then explain finite-bandwidth mode cross-talk."
canonical_url: "https://sandboxphysics.com/stories/time-frequency-quantum-optics/conversion-kernels/"
language: "en"
page_type: "story"
alternate_languages:
  zh-CN: "https://sandboxphysics.com/zh/stories/time-frequency-quantum-optics/conversion-kernels/"
last_modified: "2026-10-28"
---

[Time-frequency quantum optics](https://sandboxphysics.com/stories/time-frequency-quantum-optics/)

L4 · Chapter 7

# How a pulse gate approaches a single-mode measurement

Sandbox Physics · 2026-10-28

A pulse gate converts the selected input packet to another color. Detecting that color reads the mode. Establishing single-mode selection requires the whole input-output kernel and the output modes accepted by the detector.

## Before you begin

Singular values from chapter three, conditioning from chapter five and projection probabilities from chapter six.

[Review the preceding chapter](https://sandboxphysics.com/stories/time-frequency-quantum-optics/tomography-and-holdouts/)

- Distinguish a source amplitude from a device kernel.
- Obtain a detection effect and conditional state from conversion.
- Separate coherent output selection from unresolved multi-output detection.

## Meet the strong gate pulse in a nonlinear medium

Sum-frequency generation converts an input photon to a higher-frequency output under a strong gate pulse. The gate supplies the frequency difference. Dispersion and periodic poling determine which frequencies add coherently. Shaping the gate changes input-mode selection; spectral optics separate the converted light from the residual fields.

Write the amplitude kernel from input to output angular frequency as $K(\omega_o,\omega_i)$. In weak conversion it is proportional to gate spectrum $\alpha(\omega_o-\omega_i)$ times phase matching. This is a measurement-device kernel. The source amplitude $F$ describes pair production.

$$
K(\omega_o,\omega_i)\propto\alpha(\omega_o-\omega_i)\Phi(\omega_o,\omega_i)
$$

This model uses an undepleted strong classical gate. At higher conversion, full propagation and time-ordering dynamics determine the amplitude.

Reading: A. Eckstein · L. Serino et al.

## Count the output modes accepted by one click

In a discrete kernel, columns label input modes and rows output modes. Equal-efficiency acceptance of all outputs gives input detection effect $E=K^\dagger K$. Singular-value decomposition identifies independent input channels. Only one nonzero singular value makes the effect a rank-one projector times efficiency.

A detector accepting one coherent output mode first projects onto that mode. A detector accepting several distinguishable outputs without reading their labels adds their effects. A pure source does not by itself repair a multimode herald effect.

$$
K=U\Sigma V^\dagger,\qquad E=V\Sigma^2V^\dagger,\qquad E_{\mathrm{fail}}=I-E
$$

Physical conversion requires singular values at most one, so failure is also a positive effect. An output acceptance operator can include detection loss.

$$
\widetilde\rho_s=FE^{\mathsf T}F^\dagger,\qquad p=\operatorname{Tr}\widetilde\rho_s
$$

The transpose follows from our convention: signal rows, idler columns. A rank-one effect recovers coherent projection.

## Finite output width admits extra modes

Narrow output phase matching with a slowly varying gate spectrum makes the kernel approximately separable: one output shape times one input shape. One channel dominates. A wider accepted output lets different output frequencies select slightly different input superpositions. Unresolved detection then accepts multiple input modes.

The two-mode experiment uses the published independent Schmidt-channel conversion law. The selected mode and its orthogonal partner enter separate converted outputs, which the herald pools. The coupling is fixed at a quarter turn. An adjustable second Schmidt coefficient sets the first through normalization. These illustrative coefficients define the teaching model; a waveguide prediction would require material dispersion and geometry.

$$
\eta_k=\sin^2(g\kappa_k),\qquad g=\frac\pi2,\qquad \kappa_0^2+\kappa_1^2=1
$$

The kernel rows are the adjoints of the selected and orthogonal modes, weighted by square roots of their conversion probabilities. Unresolved detection adds their effects.

Reading: A. Eckstein · L. Serino et al.

## A multi-output sorter needs detector calibration

Several phase-matching peaks can direct input modes to different output colors. The programmable sorter study introduced in 2024 probes the outputs with known input states to reconstruct their measurement effects. This is detector tomography. State tomography asks what state arrived; detector tomography asks what the instrument actually measured.

Device comparisons need conversion efficiency, cross-talk, output resolution and uncertainty. Normalizing only successful clicks can conceal photons that never converted. An explicit inconclusive outcome keeps efficiency visible in the complete measurement.

Reading: L. Serino et al.

## Try the prediction

### [Admit a second converted mode](https://sandboxphysics.com/labs/quantum-pulse-gate/)

1. Establish a purity and success reference with ideal coherent selection.
2. Choose “A second mode slips through”, expand the kernel settings and change the second Schmidt coefficient.
3. Compare conditional purity and herald success. Explain what a successful-click-only normalization would hide.

## Exercises and explanations

1. Two orthogonal channels convert with probabilities $0.8$ and $0.2$. If output labels are ignored, is the detection effect rank one?

   Read the explanation

   No. Both singular values are nonzero, so the effect has two positive weights. Applied to an entangled pure pair, this unresolved herald generally leaves a mixed signal.

2. Does increasing gate energy always improve success and selectivity together?

   Read the explanation

   No. Strong conversion requires its full dynamics, and different channels can respond differently. Scaling a weak-conversion kernel can even produce probabilities above one. Establish a physical conversion operator before interpreting performance.

[Previous: Reconstruct coherence from clicks and predict a new measurement](https://sandboxphysics.com/stories/time-frequency-quantum-optics/tomography-and-holdouts/) [Next: Recompute a Gaussian coherence model before reading the experiment](https://sandboxphysics.com/stories/time-frequency-quantum-optics/paper-benchmarks/)

## Sources and calculation scope

1. [A. Eckstein, B. Brecht and C. Silberhorn, A Quantum Pulse Gate based on Spectrally Engineered Sum Frequency Generation, arXiv:1007.3215v2; Optics Express 19, 13770–13778 (2011)](https://arxiv.org/pdf/1007.3215v2)

   Equations (1)–(4), Figure 2 and the heralding application define the pump-times-phase-matching kernel, its Schmidt channels and ideal mode selection. Our finite two-channel Schmidt kernel is a dimensionless teaching reduction, not the paper’s PPLN parameter calculation.

2. [L. Serino et al., Programmable time-frequency mode-sorting of single photons with a multi-output quantum pulse gate, arXiv:2410.03606v1; DOI 10.1364/OE.544206](https://arxiv.org/html/2410.03606v1)

   Section 2 and Figure 2 specify the transfer kernel and the ideal/projective versus nonideal POVM description. Sections 3–4 explain detector tomography and output resolution. This work has an Optics Express journal DOI; the version read here is the 2024 preprint.