Single-Copy Homodyne Protocol

• Single-copy homodyne protocol is a method that uses one bosonic state with fixed quadrature measurements to efficiently benchmark quantum throughput and preserved entanglement.
• The approach compresses channel action into a two-dimensional subspace using coherent state overlaps and semidefinite programming to derive quantitative entanglement bounds.
• It applies across various settings—from optical channel testing to non-Gaussian entanglement detection—while relying on calibrated homodyne detection and specific phase references.

A single-copy homodyne protocol denotes a protocol in which one copy of a bosonic state, or one use of a quantum channel or linear network, is interrogated by homodyne detection in each run. In optical-channel benchmarking, the phrase is closely associated with the minimal-resource scheme that probes a device with two coherent states and two homodyne quadratures in order to lower-bound the entanglement preserved by the device and hence its quantum throughput (Killoran et al., 2010). The same expression is also used in broader contexts, including generic passive-linear-network metrology, digital homodyne of stationary cavity modes, and non-Gaussian entanglement detection from randomized homodyne data (Triggiani et al., 2021, Strandberg et al., 2023, Straeter et al., 27 Jun 2026).

1. Operational meaning and scope

Across the literature, “single-copy” means that each experimental run uses one copy of the probe state, or one use of the device under test, with no entanglement between different uses, no multiple passes through the same network, and no joint measurements across copies. “Homodyne” means that the readout is a quadrature measurement selected by a local-oscillator phase. Repetition is still required to estimate probability distributions, moments, or witness values, but the per-run quantum primitive is fixed and minimal.

Context	Single-copy meaning	Representative result
Quantum throughput benchmarking	Two coherent inputs, two homodyne quadratures	Lower bound on Negativity from minimal data (Killoran et al., 2010)
Passive linear-network metrology	One squeezed coherent probe, one pass, homodyne on all outputs	Heisenberg scaling without auxiliary networks (Triggiani et al., 2021)
Stationary bosonic modes	One cavity copy yields one scalar homodyne sample after repeated weak probes	Single-shot statistics of homodyne reproduced digitally (Strandberg et al., 2023)
Non-Gaussian CV entanglement	One copy per run, randomized homodyne, classical U-statistics	$p_3$-PPT witnesses from homodyne data (Straeter et al., 27 Jun 2026)

In this usage, the phrase does not denote a single universal protocol. Rather, it denotes a family of protocols that replace multi-copy interferometry, joint measurements, or full process tomography with fixed homodyne settings and classical post-processing.

2. Minimal-resource benchmarking of quantum throughput

The most explicit minimal homodyne construction is the quantum-throughput protocol for optical channels, memories, and repeaters (Killoran et al., 2010). The operational question is how much entanglement, or more generally how many quantum correlations, a device can still support between a trusted qubit and an optical mode after the optical mode has passed through the device.

The reference entangled state is a qubit–mode state

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$

with $\alpha\in\mathbb{R}$. If the device is $\mathcal{E}$, the output state is

$$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$

The protocol quantifies “quantumness” through the Negativity

$$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$

normalized so that a maximally entangled two-qubit state has $\mathcal{N}=1$. The quantum throughput can be expressed either as the absolute output entanglement $\mathcal{N}(\rho_{AB})$ or as an approximate relative fraction

$$\mathrm{Throughput}\approx \frac{\mathcal{N}(\rho_{AB})}{\mathcal{N}(|\psi\rangle\langle\psi|)}.$$

The entangled state is not physically generated. Instead, the protocol is rewritten in a prepare-and-measure picture. Alice chooses $|\alpha\rangle$ or $$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$0 with probability $$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$1, sends that coherent state through the device, and Bob measures one of two homodyne quadratures,

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$2

For each input $$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$3, Bob estimates the first moments

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$4

and the variances

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$5

This is the single-copy or minimal homodyne scenario in its strictest form: only two input coherent states and only two homodyne settings. Earlier work by Häseler–Moroder–Lütkenhaus had already used a similar two-coherent-state / two-homodyne setting to certify whether a channel is entanglement-breaking or not. The 2010 throughput work upgrades that setting from a qualitative yes/no test to a quantitative lower bound on preserved entanglement via semidefinite programming (Killoran et al., 2010).

3. Effective two-dimensional reconstruction and semidefinite optimization

The protocol compresses the relevant channel action into a two-dimensional effective mode subspace. First, the measured first moments define effective output coherent amplitudes

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$6

Second, the measured variances bound the purity of each conditional output state $$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$7. Writing the largest eigenvalue as $$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$8, the protocol uses

$$|\psi\rangle_{AB} =\frac{1}{\sqrt{2}} \left( |0\rangle_A|\alpha\rangle_B + |1\rangle_A|-\alpha\rangle_B \right),$$9

Third, defining the overlap of the effective coherent states by

$\alpha\in\mathbb{R}$0

the overlap of the dominant eigenvectors of $\alpha\in\mathbb{R}$1 and $\alpha\in\mathbb{R}$2 is bounded as

$\alpha\in\mathbb{R}$3

with explicit formulas in terms of $\alpha\in\mathbb{R}$4.

These quantities define a $\alpha\in\mathbb{R}$5 effective mode subspace spanned by the dominant eigenvectors $\alpha\in\mathbb{R}$6 and $\alpha\in\mathbb{R}$7. The corresponding projector is

$\alpha\in\mathbb{R}$8

and the projected state is

$\alpha\in\mathbb{R}$9

By strong monotonicity of the Negativity under LOCC,

$\mathcal{E}$0

and because $\mathcal{E}$1 is not completely known, the protocol minimizes the Negativity over all projected $\mathcal{E}$2 states compatible with the homodyne data: $\mathcal{E}$3

The feasible set is defined by positivity, $\mathcal{E}$4, bounds on diagonal matrix elements in the effective subspace, and lower bounds on the off-diagonal coherences derived from the known input overlap

$\mathcal{E}$5

The overlap $\mathcal{E}$6 is fixed to the upper bound $\mathcal{E}$7, which is shown to give the minimal entanglement. The objective is to minimize $\mathcal{E}$8. Since the trace norm of the partial transpose can be cast as an SDP by standard trace-norm techniques, the entire optimization is efficiently solvable (Killoran et al., 2010).

A nonzero minimum proves that no separable qubit–mode state can reproduce the observed homodyne data together with the known input coherence. Equivalently, the device is not entanglement-breaking and therefore has strictly positive quantum throughput.

4. Effective channel parameters, certified regimes, and limitations

For derivation of the bounds, the channel is treated agnostically: the method assumes no channel model beyond the observed first and second moments. For interpretation and plots, the paper introduces a phenomenological Gaussian-like parametrization. With $\mathcal{E}$9 and $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$0, the output means define an effective transmittivity

$$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$1

and the symmetric excess noise in shot-noise units is

$$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$2

assumed equal for both quadratures and both inputs in the examples.

The quantitative bounds are strongest in the low-noise, low-loss regime. For $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$3 and $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$4, the lower bound on output entanglement is essentially tight and matches the entanglement of the input qubit–cat state. For no loss, nontrivial entanglement bounds can be certified up to about $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$5 shot-noise units. For $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$6, the tolerable excess noise drops to about $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$7. Above those regimes, entanglement may still be present and may still be verified qualitatively by earlier methods, but the quantitative lower bound can become zero. In a specific thermal-loss model where $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$8 and $$\rho_{AB}=(\mathsf{I}\otimes \mathcal{E})(|\psi\rangle\langle\psi|).$$9 are computed exactly rather than bounded from homodyne data, nontrivial entanglement bounds remain available for excess noise larger than $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$0 of vacuum (Killoran et al., 2010).

The method is explicitly semi-device-dependent. It assumes trusted, calibrated homodyne detection with known shot-noise level; a stable phase reference so that the definitions of $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$1 and $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$2 are meaningful and $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$3; stationary channel statistics during data acquisition; single-mode behavior of the channel at the signal frequency; and known input overlap $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$4. The quantitative bounds weaken at higher noise mainly because of the truncation to a two-dimensional mode subspace and because the bounds on $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$5 and $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$6 become conservative as the noise increases.

5. Broader realizations of single-copy homodyne protocols

The same terminology appears in several other technical settings. In passive linear-network metrology, a single-mode squeezed coherent state

$$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$7

is injected once into an arbitrary $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$8-mode passive network $$\mathcal{N}(\rho)=\|\rho^{T_A}\|_{Tr}-\operatorname{Tr}\rho,$$9, and each output channel is measured by homodyne detection. The scheme is non-adaptive, requires no auxiliary network and no prior coarse estimate of $\mathcal{N}=1$0, and achieves Heisenberg scaling when the local-oscillator phases satisfy

$\mathcal{N}=1$1

with $\mathcal{N}=1$2 (Triggiani et al., 2021).

For stationary bosonic modes, “digital homodyne” or “qubitdyne” replaces a propagating-field detector by repeated weak interactions of a cavity mode with a two-level system. A single cavity copy is probed many times by qubits, and the weighted sum

$\mathcal{N}=1$3

is shown numerically to reproduce single-shot homodyne statistics for the original cavity state. Each full sequence yields one homodyne sample from one cavity copy (Strandberg et al., 2023).

For non-Gaussian continuous-variable entanglement detection, randomized balanced homodyne data are converted into unbiased U-statistic estimators of the partial-transpose moments

$\mathcal{N}=1$4

These feed linear and quadratic $\mathcal{N}=1$5-PPT witnesses, with the quadratic witness also yielding a dimension-free lower bound on the entanglement negativity. The protocol estimates $\mathcal{N}=1$6 and $\mathcal{N}=1$7 up to additive error $\mathcal{N}=1$8 at Fock cutoff $\mathcal{N}=1$9 from

$\mathcal{N}(\rho_{AB})$0

measurements at fixed confidence, and numerical demonstrations reach $\mathcal{N}(\rho_{AB})$1 empirical one-sided detection probability from $\mathcal{N}(\rho_{AB})$2 to $\mathcal{N}(\rho_{AB})$3 homodyne measurements for states with $\mathcal{N}(\rho_{AB})$4 (Straeter et al., 27 Jun 2026).

Communication and verification settings supply further variants. A verifiable homodyne construction replaces the usual trusted-coherent-LO assumption by a three-detector architecture and a squashing map, yielding explicit error bounds between implemented statistics and ideal quadrature moments for arbitrary local-oscillator pulses (Kato et al., 2022). In DV-QKD with homodyne detection, time–phase BB84 states are measured by balanced homodyne detection on two time bins, no shared reference phase is required, and the selection of decoding basis can be performed after measurement (Primaatmaja et al., 2021). In subcarrier-wave quantum communication, the carrier itself acts as the local oscillator, leading to a self-homodyne or self-heterodyne architecture on RF sidebands (Melnik et al., 2019). In a related free-space KMB09 model, each transmitted polarization qubit is read out by homodyne interference with a local oscillator, with simulated system efficiency $\mathcal{N}(\rho_{AB})$5 for $\mathcal{N}(\rho_{AB})$6 and increased QBER under turbulence and depolarizing noise (Jamal et al., 2024).

Imaging and foundational tests also adopt the same logic. Single-pixel homodyne imaging reconstructs the full complex transmission $\mathcal{N}(\rho_{AB})$7 of an object from Hadamard-shaped local-oscillator modes and one balanced homodyne detector, with reported phase precision $\mathcal{N}(\rho_{AB})$8 radians and relative amplitude precision $\mathcal{N}(\rho_{AB})$9 (Cuozzo et al., 2022). A loophole-free Bell proposal uses homodyne-heralded mesoscopic comb states and single-copy homodyne Bell measurements, with CHSH violation robust down to line transmission $$\mathrm{Throughput}\approx \frac{\mathcal{N}(\rho_{AB})}{\mathcal{N}(|\psi\rangle\langle\psi|)}.$$0 for a $$\mathrm{Throughput}\approx \frac{\mathcal{N}(\rho_{AB})}{\mathcal{N}(|\psi\rangle\langle\psi|)}.$$1 state (Etesse et al., 2013). Adaptive dyne state preparation likewise treats one entangled single-rail copy at a time and uses the single-shot phase record to deterministically prepare

$$\mathrm{Throughput}\approx \frac{\mathcal{N}(\rho_{AB})}{\mathcal{N}(|\psi\rangle\langle\psi|)}.$$2

with adaptive dyne outperforming simple homodyne and heterodyne in the presence of realistic feedback limitations (Pozza et al., 2014).

6. Conceptual significance

This suggests that “single-copy homodyne protocol” is best understood as a methodological template rather than a single named algorithm. Its recurring ingredients are a fixed and experimentally standard quadrature measurement, minimal or finite measurement settings, and classical post-processing that extracts more structure than a raw homodyne trace would appear to provide. In the quantum-throughput setting, the output is a semidefinite lower bound on preserved Negativity from only two coherent probes and two quadratures (Killoran et al., 2010). In metrology, the output is Heisenberg-limited Fisher information from a single squeezed probe and multi-channel homodyne readout (Triggiani et al., 2021). In non-Gaussian entanglement detection, the output is a $$\mathrm{Throughput}\approx \frac{\mathcal{N}(\rho_{AB})}{\mathcal{N}(|\psi\rangle\langle\psi|)}.$$3-PPT witness and a negativity lower bound computed from randomized homodyne snapshots (Straeter et al., 27 Jun 2026).

A second recurring feature is that these protocols are generally not device-independent. They typically trust the characterization of the homodyne apparatus, the local-oscillator phase reference, or the input state preparation. The verifiable-homodyne construction makes this limitation explicit by showing how to certify closeness to an ideal homodyne model even when the local oscillator is arbitrary (Kato et al., 2022). The resulting landscape is therefore technically unified by a tradeoff: single-copy homodyne protocols avoid multi-copy entangled inputs, photon-number-resolving detection, or full tomography, but they usually do so by relying on carefully calibrated quadrature detection and mathematically controlled post-processing.