Photon Subtraction in Two-Mode Squeezed Vacuum

• The paper demonstrates that photon subtraction from a two-mode squeezed vacuum effectively de-Gaussifies the state, boosting its entanglement properties.
• It details a methodology using weak beam splitters and heralded photon detection to conditionally subtract photons and modify photon number statistics.
• The analysis shows that symmetric photon subtraction optimally enhances entanglement, with practical implications for high-fidelity continuous-variable protocols like teleportation and QKD.

A two-mode squeezed vacuum (TMSV) state is the prototypical continuous-variable entangled resource. Photon subtraction from a TMSV is a heralded, non-Gaussian operation in which one or more photons are conditionally removed from one or both modes—typically by weakly tapping the modes with highly transmissive beam splitters and post-selecting on the detection of specific photon numbers in the ancillary ports. This operation "de-Gaussifies" the TMSV, modifying both its photon number statistics and phase-space structure, and is of central operational importance in quantum information processing tasks such as entanglement distillation, high-fidelity continuous-variable (CV) teleportation, and the conditional generation of squeezed Fock states.

1. Structure and Formalism of Two-Mode Squeezed Vacuum

Let $A, B$ denote bosonic modes with annihilation operators $\hat{a}, \hat{b}$. The TMSV is generated via the unitary operator $$\hat{S}(r) = \exp[r(\hat{a}\hat{b} - \hat{a}^\dagger \hat{b}^\dagger)]$$, acting on the two-mode vacuum $|0,0\rangle$, resulting in

$$|\Psi_{\rm TMSS}\rangle = \hat{S}(r)\,|0,0\rangle = \sqrt{1-\lambda^2} \sum_{n=0}^\infty \lambda^n |n\rangle_A \otimes |n\rangle_B, \quad \lambda = \tanh r.$$

This state features perfect photon-number correlations between modes and serves as the Gaussian resource state for CV entanglement protocols. The reduced state of either mode is thermal, with mean photon number $\bar{n} = \lambda^2/(1-\lambda^2)$ (Navarrete-Benlloch et al., 2012).

2. Photon Subtraction Operations and State Transformation

Photon subtraction on a TMSV is performed by interaction with a weakly reflective beam splitter (transmissivity $T$), which couples the target mode to a vacuum ancilla. Detection of exactly $k$ photons in the reflected ancilla implements the nonunitary operation $(\hat{a})^k$ on mode $A$, and analogously for mode $B$. For subtraction of $k$ and $l$ photons from $A$ and $B$, respectively, the (normalized) post-subtraction state is

$$|\Psi_{\rm sub}^{(k,l)}\rangle = \sum_{n=\max(k,l)}^\infty \sqrt{q_n^{(k,l)}}\, |n-k\rangle_A \otimes |n-l\rangle_B,$$

where

$$q_n^{(k,l)} = \frac{\lambda^{2(n-\max\{k,l\})}\, \binom{n}{k} \binom{n}{l}}{\mathcal{N}_{\rm sub}^{(k,l)}}, \quad \mathcal{N}_{\rm sub}^{(k,l)} = \sum_{n=\max(k,l)}^\infty \lambda^{2(n-\max\{k,l\})} \binom{n}{k}\binom{n}{l}.$$

This state is manifestly non-Gaussian, with higher-order Fock terms weighted by the action of the annihilation operators. In the limit of ideal photon-number resolving detectors and vanishing reflectivity, the protocol approaches the pure, idealized photon-subtraction operation [(Navarrete-Benlloch et al., 2012); (Bartley et al., 2012)].

3. Entanglement Enhancement and Optimality Conditions

The action of photon subtraction on a TMSV modifies the Schmidt coefficients, typically increasing the entropy of entanglement (as measured by the von Neumann entropy of the reduced state) or logarithmic negativity: $$E_{\rm sub}^{(k,l)} = -\sum_{n=\max(k,l)}^\infty q_n^{(k,l)} \log q_n^{(k,l)}.$$ Numerical and analytic studies establish that, for a fixed total number of subtractions $k+l$, maximum entanglement enhancement occurs when the operation is balanced (i.e., symmetric subtraction $k = l$). However, in realistic settings with losses or inefficiencies, single-photon subtraction on a single mode may be optimal for the entanglement-gain rate, defined as $R = P \cdot \Delta E_N$, where $P$ is the success probability of the heralding event [(Navarrete-Benlloch et al., 2012); (Bartley et al., 2012)]. Single-photon subtraction ($k=1, l=0$ or $k=0, l=1$) is found to achieve the largest nonclassical enhancement per event; subtracting additional photons does not yield further gain and can even reduce entanglement (Bandyopadhyay et al., 2012).

The entanglement gain depends sensitively on the losses (combined into the effective efficiency $\eta$), the squeezing parameter $r$, and the subtraction beam splitter transmissivity. Optimal conditions for maximizing entanglement gain rate are analytically and numerically characterized in the literature (Bartley et al., 2012).

4. Physical Realizations and Virtual/Indistinguishable Schemes

Heralded photon subtraction is routinely implemented using a high-transmissivity beam splitter and avalanche photodiode (APD) clicks, requiring only weak coupling and photon detection. More advanced variants include "virtual" photon subtraction—achieved via non-Gaussian post-selection on measurement outcomes, which circumvents the need for high-purity photon-number resolving detectors (Zhao et al., 2017). This approach, which weights measurement outcomes according to the retrodicted subtraction statistics, is particularly useful in continuous-variable quantum key distribution (CV-QKD), where it enables entanglement enhancement and improved tolerable noise.

Parity-controlled or indistinguishable subtraction schemes, realized for example by symmetrically coupling both modes to a common lossy channel (as in the waveguide trimer), allow conditional removal of a total of $N$ photons without identifying the subtraction site. These protocols produce "even" or "odd" photon-subtracted TMSVs with distinct parity-dependent Fock structure and nonclassicality (Datta et al., 2024).

5. Quantum Properties: Non-Gaussianity, Phase-Space Structure, and Squeezed Fock States

Photon subtraction converts the Gaussian TMSV resource into a non-Gaussian state enriched in higher photon-number components. Quantitative measures of non-Gaussianity include the entropy distance to the closest Gaussian state with identical first and second moments. Notably, there is no monotonic relation between non-Gaussianity and entanglement; highly non-Gaussian states are not always the most entangled for a given energy or total photon number (Navarrete-Benlloch et al., 2012).

The Wigner functions of photon-subtracted TMSVs develop significant negative regions, especially under single-photon subtraction, indicating strong nonclassicality. For symmetric subtraction, the Wigner function displays characteristic negative lobes along the squeezed quadratures, which can be reconstructed by quantum state tomography (Bashmakova et al., 2024). These states also support phase-space vortices and controlled parity structure, which can be tuned via the number and pattern of subtracted photons [(Bandyopadhyay et al., 2012); (Datta et al., 2024)].

A major application is in conditional state engineering: subtracting $n$ photons from one mode of a TMSV projects the other mode onto a squeezed Fock state $S(r)|n\rangle$, provided the setup parameters satisfy certain matching conditions ("universal solution regime"). For $n=1$, this constraint is automatically satisfied; for $n>1$, input squeezing, beam splitter parameters, and optical gate strengths must be tailored for optimal fidelity and probability (Korolev et al., 2023, Bashmakova et al., 2024). Both beam splitter and controlled-Z (CZ) gate implementations are feasible, with trade-offs between resource cost, fidelity, and sensitivity to imperfect detection.

6. Application to Quantum Communication and Quantum Information Protocols

Photon-subtracted TMSVs are superior entanglement resources for various CV protocols, including teleportation, entanglement distillation, and QKD. In Braunstein–Kimble quantum teleportation, symmetric photon subtraction enhances both the Einstein–Podolsky–Rosen (EPR) correlations and the teleportation fidelity:

• For symmetric subtraction ($k=l$), EPR variance decreases and sum squeezing increases (i.e., state becomes more entangled and suitable for high-fidelity teleportation).
• For asymmetric or single-mode subtraction, EPR correlations are degraded, and teleportation fidelity is always below that of the original TMSV [(Wang et al., 2014); (Arora et al., 2024)].

In CV-QKD, virtual photon subtraction extends transmission distance and increases tolerable channel excess noise, with optimal operating points at single-photon subtraction and moderate squeezing (Zhao et al., 2017). Selective photon subtraction also increases "energy-efficiency"—the ratio of entanglement gain to mean total photon number—over photon addition and other de-Gaussification operations (Navarrete-Benlloch et al., 2012).

7. Limitations, Experimental Considerations, and Outlook

Energy growth induced by photon subtraction must be carefully accounted for in benchmarking nonclassical advantages: if total injected energy is fixed across different states, the advantage of photon subtraction for single-parameter estimation is reduced, though multi-parameter and correlated metrological scenarios may still benefit (Samantaray et al., 2018). Experimental imperfections—including loss, finite detector efficiency, and imperfect mode-matching—affect both success probability and fidelity; these can be modeled and optimized within comprehensive theoretical frameworks [(Bartley et al., 2012); (Bashmakova et al., 2024)].

Photon subtraction, especially in its heralded and parity-controlled variants, has become an essential and operationally accessible non-Gaussian resource for quantum technologies. Its capacity to systematically manipulate entanglement, nonclassicality, and quantum correlations is central to the ongoing development of photonic quantum networks, CV error correction, and advanced state engineering [(Navarrete-Benlloch et al., 2012); (Korolev et al., 2023); (Datta et al., 2024)].