Homodyne-Based Bell Tests in CV Systems

• Homodyne-based Bell tests are frameworks that use continuous-variable quadrature measurements, converting continuous outcomes into binary values to violate Bell inequalities.
• They employ optimized binning strategies and advanced state engineering techniques, such as squeezed states and non-Gaussian breeding, to achieve robust violation margins even under loss.
• Hybrid implementations and GKP encodings extend these tests to multipartite systems, enabling device-independent quantum protocols with practical loss resilience.

A homodyne-based Bell test is an experimental and theoretical framework for violating Bell inequalities using continuous-variable (CV) quadrature measurements, typically realized with balanced homodyne detection. Unlike traditional Bell tests that use discrete-variable (DV) systems and photon-counting, homodyne-based tests leverage high-efficiency measurement of field quadratures, providing unique access to CV nonlocality, loss resilience, and the potential for device-independent quantum protocols.

1. Theoretical Foundations: CHSH Inequality with Homodyne Quadratures

The canonical homodyne-based Bell test adapts the CHSH scenario to continuous-variable optics. Each measurement station (Alice, Bob) measures a rotated quadrature $$X(\theta) = a\,e^{-i\theta} + a^\dagger e^{i\theta}$$, controlling the phase $\theta$ of a strong local oscillator. Outcomes are typically binned into dichotomic ($\pm1$) results, often by taking the sign of the measured value or within optimized threshold intervals. The normalized correlation for settings $(\theta_A, \theta_B)$ is

$$E(\theta_A, \theta_B) = \langle \mathrm{sign}[X_A(\theta_A)]\,\mathrm{sign}[X_B(\theta_B)] \rangle,$$

and the CHSH parameter is given by

$$S = |E(\theta_1, \theta_2) + E(\theta_2, \theta_3) + E(\theta_3, \theta_4) - E(\theta_1, \theta_4)|.$$

Local hidden variable theories impose $S \leq 2$, while quantum mechanics predicts a maximum of $2\sqrt{2}$ under ideal conditions (Thearle et al., 2018, Oudot et al., 2024).

Homodyne detection grants direct access to field quadratures, allowing correlation functions to be formulated from low-order quadrature moments or by binning continuous outcomes for use with standard inequality functionals.

2. Experimental Architectures and State Engineering

Homodyne-based Bell tests have been implemented across a range of photonic and hybrid architectures:

Four-mode entanglement and Gaussian resources: Two squeezed vacuum states generated by optical parametric oscillators (OPOs) are interfered on beam splitters, producing EPR-type entanglement; further mode mixing and polarization rotation yield four entangled modes for detection. Alice and Bob locally mix their orthogonally polarized beams via polarization optics and measure the quadratures by homodyne detection. Optimal violation is typically observed for local oscillator phases $\theta_A \in \{\pi/8,3\pi/8\}$ and $\theta_B \in \{0,\pi/4\}$, with observed $B = 2.31 \pm 0.02$ using $1.1$ dB of input squeezing and total efficiencies $\eta \approx 95$--$98\%$ (Thearle et al., 2018).

Iterative non-Gaussian "breeding" schemes: States approaching macroscopic Schrödinger-cat superpositions are prepared via homodyne heralding. Cat amplitude and purity are tunable, and violations up to $S \simeq 2.8$ are achievable with moderate squeezing and multi-step heralding, robust to transmission losses down to $\eta \sim 0.74$ (Etesse et al., 2013).

Hybrid circuits and heralded state preparation: Machine-learned photonic circuits using squeezers, beam splitters, and threshold heralding yield robust Bell violations with modest resources (e.g., $\sim 4$ dB squeezing, only two beam splitters), maintaining violation margins $S > 2$ with fiber losses up to $\sim 8$ km and threshold detector efficiency $\geq 25\%$ (Lanore et al., 2024).

Millimeter-wave domain: The use of spin-wave squeezing in ferromagnetic materials (YIG) enables entangled photon pair generation at GHz frequencies. Homodyne interferometry discards the large thermal background and establishes Bell violation at ambient temperature with high-rate entangled pair production (Salmon et al., 2020).

3. Binning, Correlators, and Bell Parameter Optimization

A crucial technical element is the mapping of continuous homodyne outcomes to binary values suitable for Bell-inequality evaluation. Several binning strategies are observed:

• Sign-binning: Assign $a = +1$ if $x \geq 0$, $-1$ otherwise. This approach is straightforward but not always optimal for state subspaces such as the $\{|0\rangle,|2\rangle\}$ qubit (Oudot et al., 2024).
• Optimized thresholding: Multiple intervals or thresholds are numerically optimized to maximize the violation for a given state, sometimes exploiting the structure of photon-number distributions in bounded-Fock subspaces.
• State-dependent binning: For GKP encodings, periodic binning implements approximate Pauli logical measurements: each outcome $q_\theta$ is mapped to $Z_L$ or $X_L$ via a partition into intervals offset by $\sqrt{\pi}$ (Yang et al., 22 Jan 2026).
• Moment-based correlators: For Gaussian states, quartic correlators ($\langle X^2Y^2 \rangle$) reduce to combinations of second moments, with Bell parameters computed from inferred photon-number correlations (Thearle et al., 2018).

Numerical and analytical optimization over state parameters, binning intervals, and measurement setting phases is standard practice to approach or maximize violation under loss, noise, and other constraints (Oudot et al., 2024).

4. Robustness, Loopholes, and Loss Tolerance

Homodyne-based Bell tests offer significant advantages in terms of efficiency and robustness:

• Detection efficiency: Homodyne detectors routinely achieve $\gtrsim 99\%$ quantum efficiency; as a result, the detection loophole is trivial to close in purely homodyne scenarios (Plick et al., 2018, Thearle et al., 2018, Oudot et al., 2024).
• Loss thresholds: For optimized states in $\{|0\rangle,|2\rangle\}$ subspaces, violations persist for overall transmission $\eta \gtrsim 0.75$; for mesoscopic cats, down to $\eta \sim 0.74$ (Etesse et al., 2013, Oudot et al., 2024). In multipartite GKP schemes, the threshold is $\eta_\text{min}\sim 0.78$ for $N=3$ GHZ/MABK-type Bell tests (Yang et al., 22 Jan 2026).
• Hybrid schemes: Combining homodyne with photon-counting detection further relaxes efficiency thresholds and permits violation even when photon counters are highly inefficient, with the detection threshold pushed arbitrarily low for on-off detectors (Araújo et al., 2011).
• Loopholes: While detection and fair-sampling loopholes are addressable, the locality loophole requires time-resolved, spacelike-separated measurement stations and fast setting randomization, which is not always implemented (Thearle et al., 2018, Töppel et al., 2014, Etesse et al., 2013).

5. Multipartite Extensions and GKP Encodings

Bipartite homodyne-based Bell tests in traditional CV encodings are fundamentally bounded (CHSH violations in the qubit $\{|0\rangle,|2\rangle\}$ subspace reach $S\approx 2.15$, broader Fock subspaces up to $S\approx 2.74$), but approaches based on GKP code encodings yield qualitatively different behavior.

• No-go for bipartite CHSH with GKP: The binning structure of Pauli measurements realized via homodyne yields observables that cannot violate the CHSH inequality for logical Bell states, regardless of squeezing (Yang et al., 22 Jan 2026).
• Multipartite nonlocality: GKP-encoded GHZ and $W$ states, probed by logical $X_L$ and $Y_L$ (via quadrature with periodic binning), can consistently violate MABK- and Cabello-type multipartite Bell inequalities with finite squeezing ($\sim$6–8 dB), provided per-mode transmission $\eta\gtrsim0.75$--0.80.
• Analytical framework: The overlap of GKP code projectors with periodically binned quadrature projectors is analytically tractable, yielding explicit dependence of the violation on squeezing, loss, and thermal noise.

This suggests that while bipartite CV nonlocality with solely homodyne detection is fundamentally bounded, access to the high-efficiency logical Pauli readout via GKP binning gives robust multipartite certifications of nonlocality under realistic photonic conditions (Yang et al., 22 Jan 2026).

6. Practical Implementations, Applications, and Outlook

Homodyne-based Bell tests are experimentally accessible and provide unique platforms for device-independent quantum protocols, quantum key distribution, and certified randomness (Thearle et al., 2018, Oudot et al., 2024, Lanore et al., 2024). Key features and future directions include:

• Resource-efficient circuits: Compact circuits with few squeezers and beam splitters combined with homodyne heralding enable violations with modest squeezing ($<4$ dB), high loss-resilience, and standard optical technologies (Lanore et al., 2024).
• Hybrid Bell-state measurements: The combination of photon subtraction (on-off detection) and quadrature windowing (homodyne conditioning) suppresses multi-photon errors, outperforming even ideal photon-number-resolving detectors in linear optical protocols, and enabling high-fidelity teleportation and entanglement swapping in photonic quantum interconnects (Asenbeck et al., 2024).
• Continuous monitoring: Time-continuous Bell measurements, realized via continuous homodyne monitoring and feedback, admit deterministic entanglement generation and CHSH violation when the detection efficiency exceeds $50\%$ (Hofer et al., 2013).
• Scaling and multimode systems: The combination of multi-pixel homodyne detection and large-scale, frequency-comb-based entanglement generation enables exploration of high-dimensional and multi-party nonlocality, though in practice observed violation margins can be small and require robust data acquisition (Plick et al., 2018).

Limitations include the need to close remaining loopholes (particularly locality), the relatively modest violation margins in two-party, Gaussian-resource scenarios, and the complexity of state engineering for optimal non-Gaussianity and loss tolerance.

7. Summary Table: Homodyne Bell Test Paradigms

Paradigm	Max $S$ Observed/Predicted	Loss Thresholds
Pure Gaussian, sign-binning, 2-party	$S\sim1.03$	$\eta \gtrsim 0.99$
$\{|0\rangle,|2\rangle\}$ qubit, optimized	$S\sim2.15$	$\eta \sim 0.77$
Cat/comb states, breeding via homodyne heralding	$S\sim2.8$	$\eta \sim 0.74$
Hybrid (homodyne+photon count), weak N00N	$S\sim2.4$	$\eta_\text{SPD}\sim0.58$
GKP codes, multipartite, periodic binning	$S_N>2$ (with $N\geq3$)	$\eta\gtrsim0.75$
RL-optimized circuits with CV resources	$S\sim2.07$	$\eta_{\text{SPD}}\sim0.25$

Values are for best-case or representative scenarios; see references for detailed parametric dependencies.

• (Thearle et al., 2018) — Demonstration of violation using four-mode entangled states and homodyne; analysis of experimental regimes and future improvements.
• (Etesse et al., 2013) — Iterative cat breeding, large violations, and loophole closure analysis.
• (Lanore et al., 2024) — RL-based optimal circuit synthesis and loss tolerance for homodyne-only Bell tests.
• (Oudot et al., 2024) — Systematic analysis of binning, Fock subspaces, and robustness for realistic homodyne Bell tests.
• (Salmon et al., 2020) — Extension to millimeter-wave regime and discrimination of thermal noise via homodyne.
• (Yang et al., 22 Jan 2026) — Bell tests with GKP encodings; no-go for bipartite CHSH, multipartite violations quantified.
• (Asenbeck et al., 2024) — Hybrid Bell-state measurement outperforming PNR detection.
• (Hofer et al., 2013) — Formalism for time-continuous Bell measurement and feedback.
• (Plick et al., 2018) — Multi-pixel homodyne and frequency-comb systems for high-dimensional Bell tests.
• (Araújo et al., 2011, Töppel et al., 2014, Brask et al., 2012) — Hybrid measurement scenarios, amplifiers, and efficiency thresholds.
• (Das et al., 2021) — Analysis of single-photon schemes and the necessity of setting-dependent local oscillator strengths.

Homodyne-based Bell tests establish the viability and versatility of continuous-variable and hybrid protocols for fundamental and applied quantum nonlocality, bridging the gap between foundational studies and scalable quantum photonics.