betaFeedback

WorksPhysics and Astronomy

Phase-conjugate Gaussian encoding for continuous-variable quantum key distribution with a single phase-sensitive amplifier

Abstract

Phase-conjugate encoding is naturally compatible with shared phase-sensitive amplification and joint detection, offering a compact optical receiver architecture for continuous-variable quantum key distribution (CVQKD). Here, we develop a unified encoding and security framework for two-mode Gaussian CVQKD with tunable phase-conjugate intermode correlations matched to a joint receiver comprising a single phase-sensitive amplifier (PSA) and one balanced photodetector. We present both coherent-state and squeezed-state prepare-and-measure protocols that use the same transmitted ensemble. The two preparations can be implemented through classically precoded Gaussian displacements of coherent sidebands or a directly generated two-mode squeezed vacuum, respectively, without an actual optical mode converter. The encoding and detection scheme separates low-variance joint quadratures for key generation from high-variance joint quadratures carrying private random modulation. This complementary modulation can reduce Eve's Holevo information without changing the Alice–Bob mutual information. Numerical results show that optimized correlated coherent-state encoding achieves higher asymptotic key rates at long distances and extends the positive-key range compared with optimized single-mode coherent-state modulation. The optimal intermode correlation increases with transmission distance, and the range advantage persists under the symmetric correlated two-mode Gaussian attacks considered here. These results provide a theoretical basis for the design and implementation of PSA-enhanced two-mode CVQKD with phase-conjugate encoding.

Introduction

Quantum key distribution (QKD) enables two distant parties, Alice and Bob, to establish secret keys using quantum states and authenticated classical communication [1, 2]. Continuous-variable (CV) QKD encodes information in the quadratures of optical fields and can be implemented with Gaussian-modulated coherent or squeezed states and coherent receivers [3, 4]. The intrinsic quantum state and information encoding used to prepare Alice’s ensemble are defining features of CVQKD protocols and strongly influence their achievable performance [3, 5, 6, 7, 8, 9, 10, 11, 12]. Although most CVQKD protocols use single-mode encoding, multimode signals with engineered intermode correlations allow information to be encoded in collective quadratures and accessed through joint measurements [13], which can be realized with a compact optical receiver.

A concrete example is phase-conjugate two-mode encoding, in which two real modulation variables XX and PP are mapped onto the coherent-state pair ∣α⟩1∣α∗⟩2|\alpha\rangle_{1}|\alpha^{*}\rangle_{2}, with α=(X+iP)/2\alpha=(X+\mathrm{i}P)/2 in shot-noise units. The two modes therefore carry displacements (X,P)(X,P) and (X,−P)(X,-P), respectively. The encoded information can be measured simultaneously through the collective quadratures x^+=(x^1+x^2)/2\hat{x}_{+}=(\hat{x}_{1}+\hat{x}_{2})/\sqrt{2} and p^−=(p^1−p^2)/2\hat{p}_{-}=(\hat{p}_{1}-\hat{p}_{2})/\sqrt{2}, which commute and have mean values of 2X\sqrt{2}X and 2P\sqrt{2}P, respectively. Cerf and Iblisdir showed that jointly measuring this phase-conjugate pair provides more information about the encoded displacements than using two identical coherent-state copies [14]. In their proposed realization, a balanced beam splitter maps x^+\hat{x}_{+} and p^−\hat{p}_{-} onto complementary quadratures of two output modes, which are then measured by homodyne detection. Niset et al. subsequently proved and experimentally demonstrated that joint measurements on Gaussian-distributed phase-conjugate coherent-state pairs reconstruct ∣α⟩|\alpha\rangle with higher fidelity than any strategy that measures the two modes separately, even when the local operations are constrained through classical communication [15]. Arném et al. further established the optimality of phase-conjugate coherent-state encoding for simultaneous estimation of XX and PP under a global energy constraint. They showed that phase-conjugate encoding with joint detection simultaneously attains the quantum Cramér–Rao bounds for both variables [16]. In CVQKD, Su et al. demonstrated phase-conjugated subcarrier modulation and phase-sensitive joint detection using a single balanced photodetector, connecting correlated Gaussian encoding to compact joint optical reception [17].

However, the practical benefit of this joint measurement is limited by finite detection efficiency and electronic noise, which reduce the useful correlations available for reconciliation. Receiver-side optical preamplification can mitigate these detection penalties by amplifying the signal before photodetection [18, 19]. A phase-sensitive amplifier (PSA) amplifies selected quadratures without the added-noise requirement of phase-insensitive amplification [20]. Phase-sensitive amplification was considered in early squeezed-state quantum cryptography [21] and has subsequently been investigated for Gaussian coherent-state protocols [18], discrete modulation [22, 23], two-way communication [24], unidimensional modulation [25], and self-referenced CVQKD [26].

Phase-conjugate encoding offers a natural route to combining phase-sensitive preamplification with joint detection, since a single two-mode PSA can amplify a selected pair of commuting collective quadratures that match the observables accessed by the joint receiver [27]. For phase-conjugate frequency sidebands, these quadratures can be read out using a central local oscillator (LO) and one balanced photodetector (BPD), followed by electrical quadrature demodulation, as demonstrated experimentally in PSA-enhanced high-speed CVQKD [28]. The compatibility between paired encoding, amplification, and joint detection motivates a unified analysis of how the transmitted intermode correlations and Alice’s quantum-state preparation affect the achievable security and performance.

In this work, we consider tunable phase-conjugate Gaussian encoding matched to a joint receiver with a single phase sensitive preamplifier and construct both coherent-state and squeezed-state preparations from a common entanglement-based representation. We relate the encoding to independently prepared asymmetric ensembles and show that the two-mode encoding is practically more appealing because the mode conversion occurs at the source, avoiding physical mode conversion before the joint receiver. Moreover, using coherent states, Alice’s implementation is further simplified by directly generating the correlated sidebands through classical precoding and modulation. Using a trusted PSA-enhanced receiver characterized by its net gain and noise figure, we compare correlated coherent-state encoding with single-mode coherent-state modulation under the same receiver-noise conditions. Numerical results show that the single-mode benchmark achieves higher rates at short and intermediate distances, while correlated encoding gives higher rates at long distances and extends the positive-key range. We further examine performance under correlated two-mode Gaussian attacks and find that the advantage in maximum transmission distance over the single-mode coherent-state benchmark persists. Together with the analysis of detector imperfections and mode-conversion loss, these results provide a theoretical basis for the design and implementation of practical two-mode CVQKD with phase-conjugate encoding and PSA-enhanced joint reception.

Phase-Conjugate Gaussian Encoding

Consider a zero-mean two-mode Gaussian ensemble with tunable intermode correlation specified by a nonnegative parameter cc. Write the covariance matrix (CM) of this ensemble in physical-mode ordering (x1,p1,x2,p2)(x_1,p_1,x_2,p_2) as

ΓS1,S2=(VI2cZcZVI2),(1)\Gamma_{S_1,S_2} = \begin{pmatrix} V I_2 & c Z \\ c Z & V I_2 \end{pmatrix}, \tag*{(1)}

where I2I_2 is the identity matrix and Z=diag⁡(1,−1)Z=\operatorname{diag}(1,-1). We restrict the family to V>1V>1 and 0≤c≤V−10\le c\le V-1 [29]; the two endpoints correspond to uncorrelated modes and maximal classical phase-conjugate correlation, respectively.

A common entanglement-based (EB) source provides a purification of (1), as illustrated in Figure 1(a). Alice starts with two independent pure EPR pairs, each with variance u=V2−c2u=\sqrt{V^2-c^2}. She holds one mode of each pair, denoted by A1A_1 and A2A_2, and applies opposite squeezing Sr=diag⁡(e−r,er)S_r=\operatorname{diag}(e^{-r},e^r), and S−r=diag⁡(er,e−r)S_{-r}=\operatorname{diag}(e^r,e^{-r}) to the outgoing modes, where

Figure 1. The EB source and two equivalent preparations. (a) The purification of S1,S2S_1,S_2 with EPR and asymmetric squeezing. (b) Coherent-state preparation uses classical precoding and direct modulation, no optical mixer required. (c) Oppositely squeezed states are independently Gaussian-modulated and mixed by BB. (d) Equivalently, a TMSV source is followed by correlated modulation driven by classical precoding.

r=14ln⁡V+cV−c.(2)r=\frac{1}{4}\ln\frac{V+c}{V-c}. \tag*{(2)}

Alice then mixes these two modes b1b_1 and b2b_2 using the balanced transformation

B=12(I2I2−I2I2)(3)B=\frac{1}{\sqrt{2}}\begin{pmatrix} I_{2} & I_{2} \\ -I_{2} & I_{2} \end{pmatrix} \tag*{(3)}

producing the signal modes S1,S2S_{1},S_{2} with CM (1). The four-mode state {A1,A2,S1,S2}\{A_{1},A_{2},S_{1},S_{2}\} is pure and constitutes the EB representation of the source. The full purification is detailed in Appendix A.

The domain 0≤c≤V−10 \le c \le V-1 admits both coherent-state and squeezed-state preparations from the same EB source, as described below.

Coherent-state preparation

When Alice performs a generalized heterodyne measurement on mode A1A_{1} of the EB source, characterized by the equivalent measurement noise CM

M=diag⁡(m,m−1),m=u−sV−c−1,(4)M=\operatorname{diag}(m,m^{-1}),\qquad m=\frac{u-s}{V-c-1}, \tag*{(4)}

and s=e−2rs=e^{-2r}, she remotely prepares a coherent state in b1b_{1}, as derived in Appendix A. The measurement on A2A_{2} uses the exchanged entries diag⁡(m−1,m)\operatorname{diag}(m^{-1},m) and likewise prepares a coherent state in b2b_{2}. The average CMs of b1b_{1} and b2b_{2} are diag⁡(V−c,V+c)\operatorname{diag}(V-c,V+c) and diag⁡(V+c,V−c)\operatorname{diag}(V+c,V-c), respectively. They describe two independently prepared, oppositely oriented asymmetric coherent-state ensembles. Finally, the balanced transformation BB introduces the correlation in (1).

Since a passive beam splitter maps a product of coherent states to another product of coherent states, Alice can implement the physical mode transformation directly through classical precoding. An explicit realization goes as follows: Alice uses independent zero-mean Gaussian variables X,PX,P of variance cc, together with four mutually independent Gaussian variables n1x,n2x,n1p,n2pn_{1x},n_{2x},n_{1p},n_{2p} of variance ε=V−c−1\varepsilon=V-c-1, which are also independent of X,PX,P. Alice prepares the product state ∣α1⟩⊗∣α2⟩|\alpha_{1}\rangle\otimes|\alpha_{2}\rangle with

xˉ1=X+n1x,pˉ1=P+n1p,xˉ2=X+n2x,pˉ2=−P+n2p,(5)\begin{aligned} \bar{x}_{1}&=X+n_{1x}, & \bar{p}_{1}&=P+n_{1p},\\ \bar{x}_{2}&=X+n_{2x}, & \bar{p}_{2}&=-P+n_{2p}, \tag*{(5)} \end{aligned}

and αj=(xˉj+ipˉj)/2\alpha_{j}=(\bar{x}_{j}+i\bar{p}_{j})/2. When ε=0\varepsilon=0, the two amplitudes become exact phase conjugates, corresponding to the boundary discussed in § 0.2 C. For ε>0\varepsilon>0, individual amplitudes are no longer exact conjugates, and the ensemble has the tunable phase-conjugate correlation.

In practice, Alice can synthesize the required displacements (5) directly on two coherent-state sources, such as two sidebands of a single-mode laser, as illustrated in Figure 1(b).

Squeezed-state preparation

For the same EB source, Alice can instead perform ordinary heterodyne measurements on both her modes with

Coh.: coherent; Sq.: squeezed. M1M_{1} is the measurement-noise CM on A1A_{1}; A2A_{2} exchanges its diagonal entries. See Eq. (4) for mm. u=V2−c2u=\sqrt{V^{2}-c^{2}} and s=e−2rs=e^{-2r}. The coherent preparation at ρ=1\rho=1 is understood as the m→∞m\to\infty homodyne limit. At ρ=0\rho=0, the two preparations coincide.

measurement-noise CM M=I2M=I_{2} and remotely prepare two squeezed states in b1b_{1} and b2b_{2} with CMs diag⁡(s,s−1)\operatorname{diag}(s,s^{-1}) and diag⁡(s−1,s)\operatorname{diag}(s^{-1},s), respectively. The average CMs of b1b_{1} and b2b_{2} are exactly the same as in the coherent preparation.

In the PM implementation, Alice prepares two oppositely squeezed vacuum states, applies required Gaussian displacements, and mixes the modes with BB, as illustrated in Figure 1(c). The intrinsic outgoing state is therefore a two-mode squeezed vacuum (TMSV) with CM

Γ∗=(cosh⁡(2r)I2sinh⁡(2r)Zsinh⁡(2r)Zcosh⁡(2r)I2).(6)\Gamma_{*}=\begin{pmatrix}\cosh(2r)I_{2}&\sinh(2r)Z\\\sinh(2r)Z&\cosh(2r)I_{2}\end{pmatrix}. \tag*{(6)}

An equivalent implementation can directly generate the TMSV through a suitable nonlinear source and then add correlated displacement accordingly [4]; in this case, Alice also uses only classical precoding, similar to the coherent-state protocol above, but the single laser source must be replaced by a TMSV source, as illustrated in Figure 1(d).

Boundary cases

At c=0c=0, one has u=Vu=V, and s=1s=1. The two preparations coincide and produce the physical ensemble of two independent thermal Gaussian states with CM Γ0=VI4\Gamma_{0}=V I_{4}, which is the source for two parallel single-mode coherent-state GG02-like protocols [30].

At c=V−1c=V-1, one has ε=0\varepsilon=0, and the coherent-state implementation consists of exact phase-conjugate pairs

∣ψX,P⟩=∣α⟩1⊗∣α∗⟩2,α=X+iP2,(7)|\psi_{X,P}\rangle=|\alpha\rangle_{1}\otimes|\alpha^{*}\rangle_{2},\qquad\alpha=\frac{X+iP}{2}, \tag*{(7)}

where XX and PP are independent zero-mean Gaussian variables of variance VA=V−1V_{A}=V-1 known to Alice. The two-mode ensemble has maximal classical correlation:

Γ1=((VA+1)I2VAZVAZ(VA+1)I2).(8)\Gamma_1 = \begin{pmatrix} (V_A+1)I_2 & V_A Z \\ V_A Z & (V_A+1)I_2 \end{pmatrix}. \tag*{(8)}

The average CMs of two modes before mode mixing are diag⁡(1,2V−1)\operatorname{diag}(1,2V-1) and diag⁡(2V−1,1)\operatorname{diag}(2V-1,1), which is a pair of oppositely oriented unidimensional modulations [8]. In this case, coherent-state protocols must deal with the estimation problem of unmodulated quadratures.

For squeezed preparation at the same boundary, the source becomes the one discussed in the PCTB minimum-leakage protocol [31], where

u=2V−1,s=12V−1.(9)u=\sqrt{2V-1}, \qquad s=\frac{1}{\sqrt{2V-1}}. \tag*{(9)}

The squeezed decomposition modulates all quadratures and therefore has no unmodulated directions of the kind encountered in the coherent decomposition above.

Define a normalized correlation strength

ρ=cV−1,0≤ρ≤1.(10)\rho=\frac{c}{V-1}, \qquad0\leq\rho\leq1. \tag*{(10)}

Table 1 summarizes the two PM decompositions throughout the allowed correlation range.

ρ\rhoPMM1M_{1}IntrinsicModulation
[0,1]Coh.[m00m−1]\begin{bmatrix}m&0\\0&m^{-1}\end{bmatrix}[11]\begin{bmatrix}1\\1\end{bmatrix}[V−c−1V+c−1]\begin{bmatrix}V-c-1\\V+c-1\end{bmatrix}
0Coh.I2I_{2}[11]\begin{bmatrix}1\\1\end{bmatrix}[V−1V−1]\begin{bmatrix}V-1\\V-1\end{bmatrix}
1Coh.[∞000]\begin{bmatrix}\infty&0\\0&0\end{bmatrix}[11]\begin{bmatrix}1\\1\end{bmatrix}[02V−2]\begin{bmatrix}0\\2V-2\end{bmatrix}
[0,1]Sq.I2I_{2}[ss−1]\begin{bmatrix}s\\s^{-1}\end{bmatrix}[V−c−sV+c−s−1]\begin{bmatrix}V-c-s\\V+c-s^{-1}\end{bmatrix}
0Sq.I2I_{2}[11]\begin{bmatrix}1\\1\end{bmatrix}[V−1V−1]\begin{bmatrix}V-1\\V-1\end{bmatrix}
1Sq.I2I_{2}[u−1u]\begin{bmatrix}u^{-1}\\u\end{bmatrix}[1−u−1u2−u]\begin{bmatrix}1-u^{-1}\\u^{2}-u\end{bmatrix}

Table 1. PM decompositions of the same ensemble. Intrinsic and modulation variances are in (x,p)(x,p) order for b1b_{1} before mixing; b2b_{2} exchanges the entries.

Joint Reception and Security Model

To connect the source encoding to Bob’s receiver, we introduce the joint quadrature operators

x^±=x^1±x^22,p^±=p^1±p^22.(11)\hat{x}_{\pm}=\frac{\hat{x}_1\pm\hat{x}_2}{\sqrt{2}}, \qquad\hat{p}_{\pm}=\frac{\hat{p}_1\pm\hat{p}_2}{\sqrt{2}}. \tag*{(11)}

With [x^j,p^k]=2iδjk[\hat{x}_j,\hat{p}_k]=2i\delta_{jk}, we have two pairs of commuting observables [x^−,p^+]=[x^+,p^−]=0[\hat{x}_{-},\hat{p}_{+}]=[\hat{x}_{+},\hat{p}_{-}]=0, and each pair can be measured jointly. The transformation in (11) is simply BTB^{T} with BB given in (3). Namely,

BT(x1p1x2p2)=(x−p−x+p+).(12)B^{T}\begin{pmatrix}x_1\\p_1\\x_2\\p_2\end{pmatrix} = \begin{pmatrix}x_-\\p_-\\x_+\\p_+\end{pmatrix}. \tag*{(12)}

For the source described by (1), the joint quadrature variances are

Var⁡(x^−)=Var⁡(p^+)=V−c,\operatorname{Var}(\hat{x}_{-})=\operatorname{Var}(\hat{p}_{+})=V-c,
Var⁡(x^+)=Var⁡(p^−)=V+c.(13)\operatorname{Var}(\hat{x}_{+})=\operatorname{Var}(\hat{p}_{-})=V+c. \tag*{(13)}

The mode correlation cc thus allocates different variances to two commuting pairs. Unless otherwise stated, the low-variance pair (x−,p+)(x_{-},p_{+}) supplies the key data, while the complementary pair (x+,p−)(x_{+},p_{-}) is used for parameter estimation. The quadrature selection rule is motivated by the numerical results on system performance detailed in Section IV.

For the coherent preparation in (5), Alice’s corresponding records are the displacements of the joint quadratures (x−,p+)(x_{-},p_{+}), namely,

ax=xˉ1−xˉ22=n1x−n2x2,ap=pˉ1+pˉ22=n1p+n2p2.(14)a_x=\frac{\bar{x}_1-\bar{x}_2}{\sqrt{2}}=\frac{n_{1x}-n_{2x}}{\sqrt{2}}, \qquad a_p=\frac{\bar{p}_1+\bar{p}_2}{\sqrt{2}}=\frac{n_{1p}+n_{2p}}{\sqrt{2}}. \tag*{(14)}

Each has variance ϵ=V−c−1\epsilon=V-c-1. The common variables XX and PP cancel from these records and instead contribute to the high-variance pair. For the squeezed preparation, the displacement variance is ϵ=V−c−s\epsilon=V-c-s.

At the receiver, Bob sets the pump phase of an ideal two-mode PSA to amplify (x−,p+)(x_{-},p_{+}). Its symplectic transformation in the physical-mode ordering is [4]

SG=(μI2−νZ−νZμI2),μ=G+G−12,ν=G−G−12.(15)S_G=\begin{pmatrix}\mu I_2&-\nu Z\\-\nu Z&\mu I_2\end{pmatrix}, \qquad \mu=\frac{\sqrt{G}+\sqrt{G}^{-1}}{2}, \qquad \nu=\frac{\sqrt{G}-\sqrt{G}^{-1}}{2}. \tag*{(15)}

where G≥1G\geq1 is the power gain. In the joint-quadrature basis (x−,p−,x+,p+)(x_{-},p_{-},x_{+},p_{+}), the transformation is

BTSGB=diag⁡(G,1/G,1/G,G).(16)B^{T}S_GB=\operatorname{diag}(\sqrt{G},1/\sqrt{G},1/\sqrt{G},\sqrt{G}). \tag*{(16)}

The PSA thus amplifies both commuting key quadratures (x−,p+)(x_{-},p_{+}) by G\sqrt{G} and deamplifies the complementary pair (x+,p−)(x_{+},p_{-}) by 1/G1/\sqrt{G}.

When the two correlated modes are the two frequency sidebands at ω0±Ω\omega_0\pm\Omega modulated onto a single continuous-wave laser, Bob can measure the joint quadratures using a central LO at ω0\omega_0 and a single BPD, followed by analog-to-digital conversion and digital I/Q demodulation [17, 28]. Let θ\theta denote the relative phase between the signal and LO after proper phase locking. The I/Q demodulation results of the photocurrent are proportional to x+cos⁡θ+p+sin⁡θx_{+}\cos\theta+p_{+}\sin\theta and −p−cos⁡θ+x−sin⁡θ-p_{-}\cos\theta+x_{-}\sin\theta, respectively. Bob switches the LO phase between π/2\pi/2 and 0 to select the amplified pair (x−,p+)(x_{-},p_{+}) or the complementary pair (x+,p−)(x_{+},p_{-}), and distills keys from the former.

For zero correlation, c=0c=0, both quadrature pairs can be used to generate keys, giving the sifting probability pkey=1p_{\mathrm{key}}=1, and the coherent-state protocol becomes equivalent to two parallel GG02 links. In this case, Bob switches the PSA pump phase along with the LO phase so that each measured quadrature pair is amplified with the same gain and noise figure. For nonzero correlations, Bob fixes the PSA pump phase to amplify the low-variance quadrature pair only as described above.

Figure 2(a) shows the complete receiver consisting of one two-mode PSA that amplifies the chosen key quadrature pair, and one BPD for joint quadratures readout. One may argue that the correlated source encoding is not truly necessary since we can place the mode mixer BB at Bob before the joint receiver, as in Figure 2(b). This holds for identical independent thermal-loss channels and ideal mode mixing, but it need not hold for general multimode channels, which require dedicated effort for the security proof and protocol design. But even for iid channels, a practical distinction remains. Direct phase-conjugate encoding occurs at the transmitter, avoiding additional receiver-side mode conversion and its potential insertion loss and calibration requirements.

Figure 2. Phase-conjugate encoding and an equivalent implementation. (a) Alice directly prepares the correlated sideband pair and Bob uses one two-mode PSA, a central LO, and one BPD with RF I/Q demodulation. (b) Independently prepared, oppositely oriented asymmetric modes undergo the balanced transformation BB before the same joint receiver. For identical independent phase-insensitive channels and ideal mixing, the two routes have the same per-mode information quantities.

We characterize a practical PSA by the same net power gain GG and coherent-input linear noise figure FF for both key quadratures q∈{x−,p+}q \in\{x_{-},p_{+}\}. These parameters include the PSA module’s insertion loss but exclude the BPD noise. Assuming linear amplification, the effective quadrature model for both preparation schemes can be written as

qout=G qin+z,Var⁡(z)=G(F−1),(17)q_{\mathrm{out}}=\sqrt{G}\,q_{\mathrm{in}}+z,\qquad\operatorname{Var}(z)=G(F-1), \tag*{(17)}

where F−1F-1 is the PSA added-noise variance referred to its input. Using a trusted receiver model and denoting the BPD efficiency by η\eta and electrical noise variance velv_{\mathrm{el}} [32], the total receiver noise referred to the PSA input is given by

χrec=F−1+1−η+velηG.(18)\chi_{\mathrm{rec}}=F-1+\frac{1-\eta+v_{\mathrm{el}}}{\eta G}. \tag*{(18)}

An ideal PSA has F=1F=1 and suppresses the detector contribution by 1/G1/G [18, 28]. The complementary quadrature is assumed to be independently calibrated for parameter estimation, and the required channel covariances are taken as known in the asymptotic analysis below.

Each physical mode passes through an identical independent thermal-loss channel,

ΓB=T(ΓS1,S2+χlineI4),(19)\Gamma_{B}=T\left(\Gamma_{S_{1},S_{2}}+\chi_{\mathrm{line}}I_{4}\right), \tag*{(19)}

where TT is the transmittance, χline=(1−T)/T+ξ\chi_{\mathrm{line}}=(1-T)/T+\xi with excess noise ξ\xi, and ΓB\Gamma_{B} is the signal CM at the receiver input, before the PSA.

The asymptotic key rate per mode is [33, 34, 35]

K=pkey12max⁡{β(IABx−+IABp+)−χBEx−,p+,0},(20)K=p_{\mathrm{key}}\frac{1}{2}\max\left\{\beta\left(I_{AB}^{x_{-}}+I_{AB}^{p_{+}}\right)-\chi_{BE}^{x_{-},p_{+}},0\right\}, \tag*{(20)}

where pkeyp_{\mathrm{key}} is the probability of selecting the quadratures {x−,p+}\{x_{-},p_{+}\}, and β\beta is the reconciliation efficiency. For both preparations, the mutual information satisfies IABx−=IABp+I_{AB}^{x_{-}}=I_{AB}^{p_{+}}, and is found to be

IAB=12log⁡2[1+εs+χline+χrec/T],(21)I_{AB}=\frac{1}{2}\log_{2}\left[1+\frac{\varepsilon}{s+\chi_{\mathrm{line}}+\chi_{\mathrm{rec}}/T}\right], \tag*{(21)}

where ε=V−c−s\varepsilon=V-c-s with the squeezed quadrature variance ss, and s=1s=1 for coherent preparation. Appendix B details the calculation of Holevo information.

To allow correlations between the physical channels, we assume symmetric channels with common transmittance TT and excess noise ξ\xi, and use a two-mode entangling cloner with input state characterized by [36]

ΓE1E2=(ωI2CECEωI2),CE=diag⁡(g1,g2),(22)\Gamma_{E_{1}E_{2}}= \begin{pmatrix} \omega I_{2} & C_{E}\\ C_{E} & \omega I_{2} \end{pmatrix}, \qquad C_{E}=\operatorname{diag}(g_{1},g_{2}), \tag*{(22)}

where ω=1+Tξ/(1−T)\omega=1+T\xi/(1-T) is the ancillary mode variance. The allowed correlation is determined by

ω±g1>0,ω±g2>0,\omega\pm g_{1}>0,\qquad\omega\pm g_{2}>0,
(ω+g1)(ω+g2)≥1,(\omega+g_{1})(\omega+g_{2})\ge1,
(ω−g1)(ω−g2)≥1.(23)(\omega-g_{1})(\omega-g_{2})\ge1. \tag*{(23)}

The channel output is now

ΓB=TΓS1,S2+(1−T)ΓE1E2,(24)\Gamma_{B}=T\Gamma_{S_{1},S_{2}}+(1-T)\Gamma_{E_{1}E_{2}}, \tag*{(24)}

and the total added noise at the PSA input for quadrature {x−,p+}\{x_{-},p_{+}\} is

N1=(1−T)(ω−g1)+χrec,N2=(1−T)(ω+g2)+χrec.(25)\begin{aligned} N_{1} &= (1-T)(\omega-g_{1})+\chi_{\mathrm{rec}},\\ N_{2} &= (1-T)(\omega+g_{2})+\chi_{\mathrm{rec}}. \tag*{(25)} \end{aligned}

Numerical Results and Discussion

Optimized phase-conjugate encoding

Using the normalized correlation defined by (10) and the parameters outlined in Table II, Figure 3(a) compares the phase-conjugate encoded coherent-state and squeezed-state protocols with a standard single-mode coherent-state benchmark (GG02). For coherent encoding, the ensemble variance V≤100V \le100 and correlation ρ\rho are jointly optimized at each distance, whereas the single-mode benchmark uses symmetric modulation and optimizes only VV. The probability of selecting the joint quadratures {x−,p+}\{x_{-},p_{+}\} is pkey=1/2p_{\mathrm{key}} = 1/2 in both correlated protocols, whereas the single-mode case has pkey=1p_{\mathrm{key}} = 1. The squeezed-state encoding uses the same optimized parameters (V,c)(V,c) as the coherent encoding and improves performance because of a higher IABI_{AB}. At 100 km it requires 8.17 dB of intrinsic squeezing and gives 2.26×10−32.26 \times10^{-3} bits/mode. Correlated coherent encoding extends the maximum distance of the single-mode benchmark from 152.6 to 171.9 km with the optimal correlation changing from ρ=0\rho= 0 at zero distance to ρ≃0.965\rho\simeq0.965 at the maximum distance, as shown in Figure 3(b). Although the singlemode protocol gives higher rates at short and intermediate distances, the correlated coherent encoding overtakes it at approximately 133.3 km. As discussed in § 0.2 C, the correlated coherent preparation reduces to GG02 in the boundary case with ρ=0\rho=0 and pkey=1p_{\mathrm{key}}=1. This suggests that the best practice with the proposed two-mode coherent-state CVQKD is using zero correlations and pkey=1p_{\mathrm{key}}=1 for distances below the performance crossing point and using strong correlations and pkey=1/2p_{\mathrm{key}}=1/2 for distances beyond.

Figure 3. Optimized secret key rate based on phase-conjugate encoding and joint reception. (a) Secret key rates for correlated squeezed-state and coherent-state preparations and the single-mode coherent-state benchmark under V≤100V \le100. The correlated coherent encoding jointly optimizes (V,c)(V,c), while the single-mode benchmark optimizes VV with symmetric modulation. The squeezed-state preparation uses the same optimized (V,c)(V,c) as the correlated coherent encoding. Both correlated preparations use only the low-variance quadratures for key distribution (pkey=1/2p_{\mathrm{key}} = 1/2), whereas the single-mode benchmark uses both (pkey=1p_{\mathrm{key}} = 1). (b) Left axis: key modulation variances ε=V−c−1\varepsilon= V - c - 1 and V−1V - 1 for the correlated and single-mode coherent encodings, respectively. Right axis: the optimal correlation ρ\rho (green dotted). The orange dotted continuation denotes the squeezed-state preparation using the same optimized parameters as the coherent preparation. (c) Secret key rates at 100 km versus the maximal VV used for optimization, denoted by Vmax⁡V_{\max}. The vertical dotted line marks Vmax⁡=100V_{\max} = 100, used in (a) and (b). Other parameters are given in Table II.

ParameterDefault setting
Fiber loss coefficient0.2 dB/km
Input-referred excess noiseξ=0.05\xi= 0.05 SNU
Reconciliation efficiencyβ=0.95\beta= 0.95
BPD efficiencyη=0.72\eta= 0.72
Electronic-noise variancevel=1.19v_{\mathrm{el}} = 1.19 SNU
PSA net power gainG=10G = 10 dB
PSA noise figureNF=1.3\mathrm{NF} = 1.3 dB
Key-setting probabilitypkey=1/2p_{\mathrm{key}} = 1/2

Table II. Default parameters used in the numerical studies unless stated otherwise.

The optimized (V,ρ)(V,\rho) also determines the key modulation variance ε=V−c−1\varepsilon=V-c-1 of the correlated coherent encoding, which is compared with the single-mode modulation variance VA=V−1V_A=V-1 in Figure 3(b). The optimal modulation variances approach finite values in both cases and strong correlation is needed for the correlated protocols. Figure 3(c) further confirms the performance of the correlated coherent protocol is insensitive to the maximal VV beyond 100 used above, checked at 100 km. The squeezed-state preparation benefits slightly from the larger VV, at the cost of a larger intrinsic squeezing.

Equation (14) shows that XX and PP cancel in the low-variance records and appear only in the high-variance quadratures. For the independent channels in (19), varying XX and PP therefore does not change the mutual information IABI_{AB} in the quadratures (x−,p+)(x_-,p_+), as in (21), but provides randomization of the complementary quadratures (x+,p−)(x_+,p_-). This is equivalent to applying an additional random-displacement channel to Eve’s conditional states without changing Bob’s key statistics. The data-processing inequality then implies that χBE\chi_{BE} cannot increase under this operation [37]. Numerical checks at fixed modulation level ε\varepsilon confirm that χBE\chi_{BE} decreases and approaches saturation as the variances of XX and PP increase. A closely related effect was identified for trusted preparation noise in squeezed-state CVQKD, where noise in the anti-squeezing quadrature can reduce Eve’s Holevo information without changing IABI_{AB} [38]. In the decoupled mode basis, our additional private modulation plays a similar role in the high-variance quadrature of each mode.

Performance under correlated attacks

We next compare the correlated coherent-state protocol under iid and correlated two-mode Gaussian attacks, using the model in (22). At each distance, Alice optimizes (V,c)(V,c) separately for the two attack models under V≤100V\leq100, as in Figure 3. The iid reference fixes g1=g2=0g_1=g_2=0. For correlated attacks, Eve minimizes the key rate over (g1,g2)(g_1,g_2) within the physical domain in (23).

Figure 4 shows that allowing correlations in Eve’s ancillary modes lowers the optimized key rate. Compared with the iid result, the rate decreases by 17.9% at 10 km, 3.32% at 50 km, and 0.387% at 100 km. The results also show that the preference for stronger source correlations at longer distances persists under correlated attack. Numerical scans at various distances, with the corresponding optimized source parameters held fixed, support that the optimal attack satisfies g1=−g2>0g_1=-g_2>0 (see Figure 7 in Appendix C). In the decoupled mode basis, these correlations represent phase-sensitive environmental noise, with opposite orientations in the two decoupled modes.

Figure 4. Correlated coherent-state encoding under iid and correlated two-mode Gaussian attacks. Blue curves give the secret key rate (left axis), and green curves give the optimal source correlation ρ\rho (right axis). Dashed curves use g1=g2=0g_1=g_2=0; solid curves minimize the rate over the attack domain in Eq. (23). The source parameters (V,c)(V,c) are optimized separately for each attack model under V≤100V\leq100. Both cases use pkey=1/2p_{\mathrm{key}}=1/2 and the parameters in Table II.

Figure 7. Correlated attack scans at fixed optimized coherent-state sources. Panels (a) and (b) show the key rates at 1 and 10 km, with (V,c)=(100,34.98)(V,c)=(100,34.98) and (100,82.82)(100,82.82), respectively. Solid black curves mark the physical boundaries, dashed lines indicate g2=−g1g_2=-g_1, open circles mark the iid attack, and stars mark the refined minima. All other parameters are those of Figure 4.

PSA compensation of detector noise

According to (18), a practical PSA suppresses the detector noise contribution by 1/G1/G but also adds noise. Let the detector noise be χ0=(1−η+vel)/η\chi_0=(1-\eta+v_{\mathrm{el}})/\eta; then the total receiver noise is χrec=F′−1+χ0/G\chi_{\mathrm{rec}}=F'-1+\chi_0/G. The PSA reduces the total receiver noise when

F<1+χ0(1−G−1).(26)F < 1+\chi_0(1-G^{-1}). \tag*{(26)}

For large enough gain, the total receiver noise consists of mainly the PSA added noise since χrec→F−1\chi_{\mathrm{rec}}\to F-1 and the PSA benefits only when its NF satisfies F<1+χ0F<1+\chi_0. Lower NF is therefore important once the detector contribution has been sufficiently suppressed.

Figure 5 examines the combined effect of gain and NF for the coherent-state preparation with the detector considered here, with η=0.72\eta=0.72 and vel=1.19v_{\mathrm{el}}=1.19 SNU. The color scale shows the rate change ΔK=K(G,F)−K(1,1)\Delta K=K(G,F)-K(1,1) relative to the no-PSA case, keeping the source and channel unchanged. For V=54V=54 and c=49.5c=49.5 at 50 km, the region with ΔK>0\Delta K>0 agrees with the noise-reduction condition in (26). The 10 dB, 1.3 dB-NF module increases the key rate from 3.81×10−33.81\times10^{-3} to 7.26×10−37.26\times10^{-3} bits/mode.

Figure 5. Key-rate change relative to the case without using a PSA for correlated coherent-state encoding at 50 km, with V=54V=54, c=49.5c=49.5, η=0.72\eta=0.72, and vel=1.19v_{\mathrm{el}}=1.19 SNU. The black solid curve marks ΔK=0\Delta K=0, and the green dashed curve gives equality in Eq. (26); the two boundaries coincide within the scan resolution. The star marks a net gain of 10 dB and NF = 1.3 dB. The source remains fixed, with ξ=0.05\xi=0.05, β=0.95\beta=0.95, and pkey=1/2p_{\mathrm{key}}=1/2; all receiver noise is trusted.

To isolate the PSA effect on the detector loss and noise, we can examine the NF requirement in (26) by setting vel=0v_{\mathrm{el}}=0 or η=1\eta=1, respectively. For loss alone at η=0.72\eta=0.72, a net gain of 10 dB requires NF below 1.303 dB to satisfy this condition. The 1.3 dB-NF considered here therefore gives almost no net improvement in terms of compensating the detector loss. For the noisy detector with η=1\eta=1, the NF threshold is 3.162 dB and for both η=0.72\eta=0.72 and vel=1.19v_{\mathrm{el}}=1.19 SNU, the NF threshold increases to 4.529 dB at the same 10 dB gain.

Receiver-side mode conversion

Direct generation of the correlated modes avoids the receiver-side mode converter in Figure 2(b). To see the impact of a lossy converter, we model it as ideal mixing followed by equal trusted losses of transmittance τm\tau_m in the two modes. The combined converter and PSA then have G′=τmGG'=\tau_mG and F′=F/τmF'=F/\tau_m, with total receiver noise referred to the converter input

χrec′=1−τm+χrecτm.(27)\chi'_{\mathrm{rec}}=\frac{1-\tau_m+\chi_{\mathrm{rec}}}{\tau_m}. \tag*{(27)}

Figure 6. Effect of receiver-side mode-conversion loss on coherent-state (left) and squeezed-state (right) preparations of the same average state. Rates are normalized to their zero-loss values. The source is optimized for coherent preparation without the converter under V≤100V\leq100, using the same iid calculation as Figure 3, and then held fixed for both preparations. The PSA has a net gain of 10 dB and NF = 1.3 dB, with η=0.72\eta=0.72, vel=1.19v_{\mathrm{el}}=1.19 SNU.

Figure 6 shows the resulting key-rate reduction for both coherent and squeezed preparations. At each distance, the coherent source is optimized under the same condition as in Figure 3, and its parameters are then fixed while the conversion loss is varied. An additional loss of 1 dB reduces the coherent-state rate by 19.7% at 50 km and 18.1% at 100 km, with similar reductions for squeezed-state preparation.

Note that the source–receiver equivalence used for this conversion comparison is restricted to iid channels; unequal or generally correlated multimode channels need not commute with the mode mixer.

Conclusion

We have developed a tunable phase-conjugate Gaussian encoding and security framework for two-mode CVQKD with a single PSA and joint detection using one BPD. We present both coherent-state and squeezed-state protocols based on the same two-mode source with tunable intermode correlations. The encoding and detection scheme separates joint quadratures used for key generation from complementary quadratures carrying private random modulation, which can reduce Eve’s information without changing the Alice–Bob mutual information under the channel models considered here. The correlated coherent-state encoding gives higher rates at long distances compared to the single-mode coherent-state benchmark and extends the maximal distance under the same receiver-noise conditions. The preferred intermode correlation increases with transmission distance, and the distance advantage persists when the encoding is reoptimized against the symmetric correlated two-mode Gaussian attacks. The practical benefit of correlated encoding is direct compatibility with the shared PSA and joint measurement. The coherent preparation uses classically precoded Gaussian displacements of coherent sidebands, while the squeezed preparation applies correlated displacements to a directly generated two-mode squeezed vacuum. Neither implementation requires an additional optical mode converter, preserving a rate advantage in a nonideal implementation with finite mode conversion loss.

The EB source and generalized heterodyne measurements

The EB source in Figure 1(a) starts from two independent pure EPR pairs. Each pair has CM

ΓEPR(u)=(uI2kZkZuI2),k=u2−1.(28)\Gamma_{\mathrm{EPR}}(u)=\begin{pmatrix} uI_{2} & kZ \\ kZ & uI_{2} \end{pmatrix}, \qquad k=\sqrt{u^{2}-1}. \tag*{(28)}

After the opposite squeezing and mode mixing defined in § 0.2, the four-mode CM in the ordering (A1,A2,S1,S2)(A_{1},A_{2},S_{1},S_{2}) can be written as

ΓEB=(uI20C~0/2−C0/20uI2C~0/2C~0/2C0T/2C~0T/2VI2cZ−C0T/2C~0T/2cZVI2),(29)\Gamma_{\mathrm{EB}}=\begin{pmatrix} uI_{2} & 0 & \widetilde{C}_{0}/\sqrt{2} & -C_{0}/\sqrt{2} \\ 0 & uI_{2} & \widetilde{C}_{0}/\sqrt{2} & \widetilde{C}_{0}/\sqrt{2} \\ C_{0}^{T}/\sqrt{2} & \widetilde{C}_{0}^{T}/\sqrt{2} & VI_{2} & cZ \\ -C_{0}^{T}/\sqrt{2} & \widetilde{C}_{0}^{T}/\sqrt{2} & cZ & VI_{2} \end{pmatrix}, \tag*{(29)}
C0=kZSr=kdiag⁡(s,−1/s),C_{0}=kZ S_{r}=k\operatorname{diag}(\sqrt{s},-1/\sqrt{s}),
C~0=kZS−r=kdiag⁡(1/s,−s),\widetilde{C}_{0}=kZ S_{-r}=k\operatorname{diag}(1/\sqrt{s},-\sqrt{s}),

which is a purification of the average signal state ΓS1,S2\Gamma_{S_{1},S_{2}}. A generalized heterodyne measurement can be implemented by mixing the input mode with vacuum on a beam splitter of transmittance τA\tau_{A}, homodyning orthogonal output quadratures, and rescaling the outcomes to obtain unit gain for the input quadratures. The equivalent measurement noise CM is

M=diag⁡(m,m−1),m=1−τAτA.(30)M=\operatorname{diag}(m,m^{-1}), \qquad m=\frac{1-\tau_{A}}{\tau_{A}}. \tag*{(30)}

Applying this measurement to A1A_{1} in Eq. (29), the conditional CM of b1b_{1} before mode mixing is [39]

γb1∣a1=γb1−C0T(uI2+M)−1C0=diag⁡(us,u/s)−C~0T(uI2+M)−1C~0=diag⁡(vm,vm−1),(31)\gamma_{b_{1}|a_{1}}=\gamma_{b_{1}}-C_{0}^{T}(uI_{2}+M)^{-1}C_{0} =\operatorname{diag}(us,u/s)-\widetilde{C}_{0}^{T}(uI_{2}+M)^{-1}\widetilde{C}_{0} =\operatorname{diag}(v_{m},v_{m}^{-1}), \tag*{(31)}

where

γb1=Sr(uI2)SrT=diag⁡(us,u/s)(32)\gamma_{b_{1}}=S_{r}(uI_{2})S_{r}^{T}=\operatorname{diag}(us,u/s) \tag*{(32)}

describes the b1b_{1} mode before Alice’s measurement, and

vm=sum+1u+m.(33)v_{m}=s\frac{um+1}{u+m}. \tag*{(33)}

Requiring vm=1v_{m}=1 gives

m=u−sus−1=u−sV−c−1.(34)m=\frac{u-s}{us-1}=\frac{u-s}{V-c-1}. \tag*{(34)}

which agrees with Eq. (4). We have used us=V−cus=V-c and u/s=V+cu/s=V+c for mm. These relations follow from the source parameters defined in § 0.2,

u=(V−c)(V+c),s=e−2r=V−cV+c.(35)u=\sqrt{(V-c)(V+c)}, \qquad s=e^{-2r}=\sqrt{\frac{V-c}{V+c}}. \tag*{(35)}

The measurement on A2A_{2} likewise gives γb2∣a2=I2\gamma_{b_{2}|a_{2}}=I_{2}. Thus the conditional state is a product of coherent states, before and after the passive mixing BB.

Holevo information with trusted receiver noise

Model Eve’s attack by a two-mode entangling cloner that injects modes e=(E1,E2)e=(E_{1},E_{2}) with CM Γe=ΓE1E2\Gamma_{e}=\Gamma_{E_{1}E_{2}} in Eq. (22), and correlations satisfying Eq. (23). Eve couples each injected mode to the corresponding signal mode through a beam splitter of transmittance TT and keeps both reflected outputs and the modes purifying Γe\Gamma_{e}. These modes together form her system EE.

Let A=(A1,A2)A=(A_{1},A_{2}) denote Alice’s modes and B=(B1,B2)B=(B_{1},B_{2}) Bob’s at channel output but before the receiver. From the EB source in Eq. (29), their CM is

ΓAB=(ΓACABCABTΓB),ΓA=uI4,(36)\Gamma_{AB}=\begin{pmatrix}\Gamma_{A} & C_{AB}\\ C_{AB}^{T} & \Gamma_{B}\end{pmatrix}, \qquad\Gamma_{A}=uI_{4}, \tag*{(36)}
ΓB=TΓS+(1−T)Γe,CAB=TCAS,\Gamma_{B}=T\Gamma_{S}+(1-T)\Gamma_{e}, \qquad C_{AB}=\sqrt{T}C_{AS},

where ΓS=ΓS1,S2\Gamma_{S}=\Gamma_{S_{1},S_{2}} and CASC_{AS} is the cross-covariance part in Eq. (29). Since Eve purifies ABAB, her unconditional entropy is S(E)=S(AB)S(E)=S(AB).

For the channel model considered here, Bob’s transformation BTB^{T} would decouple the two correlated modes and, with proper local phase rotations and reordering, reduce ΓAB\Gamma_{AB} to two independent Gaussian blocks:

γj=(uI2TC0TC0TγB,j),j=1,2,(37)\gamma_{j}=\begin{pmatrix} uI_{2} & \sqrt{T}C_{0}\\ \sqrt{T}C_{0}^{T} & \gamma_{B,j}\end{pmatrix}, \qquad j=1,2, \tag*{(37)}
γB,j=Tdiag⁡(V−c,V+c)+(1−T)Dj,\gamma_{B,j}=T\operatorname{diag}(V-c,V+c)+(1-T)D_{j},

where C0C_{0} is defined in Eq. (29) and

D1=diag⁡(ω−g1,ω−g2),(38)D_{1}=\operatorname{diag}(\omega-g_{1},\omega-g_{2}), \tag*{(38)}
D2=diag⁡(ω+g2,ω+g1).D_{2}=\operatorname{diag}(\omega+g_{2},\omega+g_{1}).

Given the total receiver noise χrec\chi_{\mathrm{rec}} in Eq. (18) and d=(1+χrec)−1d=(1+\chi_{\mathrm{rec}})^{-1}, the CM used for conditioning is [40]

Mj=Rd(γj⊕I2)RdT,(39)\mathcal{M}_{j}=\mathcal{R}_{d}(\gamma_{j}\oplus I_{2})\mathcal{R}_{d}^{T}, \tag*{(39)}
Rd=I2⊕(1−dI2−dI2dI21−dI2).\mathcal{R}_{d}=I_{2}\oplus\begin{pmatrix}\sqrt{1-d}I_{2} & -\sqrt{d}I_{2}\\ \sqrt{d}I_{2} & \sqrt{1-d}I_{2}\end{pmatrix}.

The mode ordering in Mj\mathcal{M}_{j} is (Aj,Fj,Yj)(A_{j},F_{j},Y_{j}), where FjF_{j} is the unused trusted receiver mode and YjY_{j} is measured. Writing

Mj=(γAjFjKjKjTγYj).(40)\mathcal{M}_{j}=\begin{pmatrix}\gamma_{A_{j}F_{j}} & K_{j}\\ K_{j}^{T} & \gamma_{Y_{j}}\end{pmatrix}. \tag*{(40)}

the homodyne conditional CM is [4]

γ~j=γAjFj−Kj(ΠγYjΠ)+KjT,(41)\widetilde{\gamma}_j=\gamma_{A_jF_j}-K_j\left(\Pi\gamma_{Y_j}\Pi\right)^{+}K_j^{T}, \tag*{(41)}

where Π=diag⁡(1,0)\Pi=\operatorname{diag}(1,0) and ++ denotes the Moore–Penrose pseudoinverse.

Denote the two symplectic eigenvalues of γj\gamma_j and γ~j\widetilde{\gamma}_j as νj,±\nu_{j,\pm} and ν~j,±\widetilde{\nu}_{j,\pm}. The two-mode Holevo information is

χBEx−,p+=∑j=12∑σ=±[h(νj,σ)−h(ν~j,σ)],(42)\chi_{BE}^{x-,p+}=\sum_{j=1}^{2}\sum_{\sigma=\pm}\left[h(\nu_{j,\sigma})-h(\widetilde{\nu}_{j,\sigma})\right], \tag*{(42)}
h(ν)=ν+12log⁡2ν+12−ν−12log⁡2ν−12.h(\nu)=\frac{\nu+1}{2}\log_2\frac{\nu+1}{2}-\frac{\nu-1}{2}\log_2\frac{\nu-1}{2}.

The mode number normalization and the factor pkeyp_{\mathrm{key}} are applied separately in Eq. (20).

Correlated-attack optimization

Figure 7 examines the correlated attack plane over the physically allowed region of (g1,g2)(g_1,g_2) at 1 and 10 km, fixing the corresponding source parameters from Figure 4 at (V,c)=(100,34.98)(V,c)=(100,34.98) and (100,82.82)(100,82.82). The color scale gives the rate Kraw=pkey(βIAB−χBE)K_{\mathrm{raw}}=p_{\mathrm{key}}(\beta I_{AB}-\chi_{BE}), without reoptimizing the source at each attack point. The refined minima lie on g1=−g2g_1=-g_2, with g1≃1.48416g_1\simeq1.48416 and 0.2712690.271269 at the two distances, respectively.

References

  1. [1]V. Scarani, H. Bechmann-Pasquinucci, N. J. Cerf, M. Dušek, N. Lütkenhaus, and M. Peev, The security of practical quantum key distribution, Reviews of Modern Physics 81, 1301 (2009).DOI
  2. [2]S. Pirandola, U. L. Andersen, L. Banchi, M. Berta, D. Bunandar, R. Colbeck, D. Englund, T. Gehring, C. Lupo, C. Ottaviani, J. L. Pereira, M. Razavi, J. Shamsul Shaari, M. Tomamichel, V. C. Usenko, G. Vallone, P. Villoresi, and P. Wallden, Advances in quantum cryptography, Advances in Optics and Photonics 12, 1012 (2020).
  3. [3]F. Grosshans and P. Grangier, Continuous variable quantum cryptography using coherent states, Physical Review Letters 88, 057902 (2002).pubmed.ncbi.nlm.nih.gov/11863782
  4. [4]C. Weedbrook, S. Pirandola, R. García-Patrón, N. J. Cerf, T. C. Ralph, J. H. Shapiro, and S. Lloyd, Gaussian quantum information, Reviews of Modern Physics 84, 621 (2012).DOI
  5. [5]N. J. Cerf, M. Lévy, and G. Van Assche, Quantum distribution of Gaussian keys using squeezed states, Physical Review A 63, 052311 (2001).DOI
  6. [6]A. Lev​​errier and P. Grangier, Unconditional security proof of long-distance continuous-variable quantum key distribution with discrete modulation, Physical Review Letters 102, 180504 (2009), erratum: Phys. Rev. Lett. 106, 259902 (2011).
  7. [7]C. Weedbrook, S. Pirandola, S. Lloyd, and T. C. Ralph, Quantum cryptography approaching the classical limit, Physical Review Letters 105, 110501 (2010).DOI
  8. [8]V. C. Usenko and F. Grosshans, Unidimensional continuous-variable quantum key distribution, Physical Review A 92, 062337 (2015).DOI
  9. [9]C. S. Jacobsen, L. S. Madsen, V. C. Usenko, R. Filip, and U. L. Andersen, Complete elimination of information leakage in continuous-variable quantum communication channels, npj Quantum Information 4, 32 (2018).
  10. [10]M. S. Winnel, N. Hosseinidehaj, and T. C. Ralph, Minimization of information leakage in continuous-variable quantum key distribution, Physical Review A 104, 012411 (2021).DOI
  11. [11]H. Q. Nguyen, A. A. E. Hajomer, I. Derkach, N. Jain, U. L. Andersen, and T. Gehring, Effects of imbalanced modulation in continuous-variable quantum key distribution, in CLEO 2023, Technical Digest Series (Optica Publishing Group, 2023) p. FM1A.6.DOI
  12. [12]Ö. Erkilic, B. Shajilal, L. O. Conlon, A. Walsh, A. Das, S. Kish, T. Symul, P. K. Lam, S. M. Assad, and J. Zhao, Enhanced continuous-variable quantum key distribution via adaptive signal processing, Communications Physics 8, 406 (2025).DOI
  13. [13]V. C. Usenko, L. Ruppert, and R. Filip, Entanglement-based continuous-variable quantum key distribution with multimode states and detectors, Physical Review A 90, 062326 (2014).DOI
  14. [14]N. J. Cerf and S. Iblisdir, Phase conjugation of continuous variable quantum variables, Physical Review A 64, 032307 (2001).DOI
  15. [15]J. Niset, A. Acín, U. L. Andersen, N. J. Cerf, R. García-Patrón, M. Navascués, and M. Sanburn, Superiority of entangled measurements over all local strategies for the estimation of product coherent states, Physical Review Letters 98, 260404 (2007).pubmed.ncbi.nlm.nih.gov/17678072
  16. [16]M. Arnhem, E. Karpov, and N. J. Cerf, Optimal estimation of parameters encoded in quantum coherent state quadratures, Applied Sciences 9, 4264 (2019).DOI
  17. [17]Z. Su, J. Wang, D. Cai, X. Guo, D. Wang, and Z. Li, Experimental demonstration of phase-sensitive multimode continuous variable quantum key distribution with improved secure key rate, Photonics Research 11, 1861 (2023).DOI
  18. [18]S. Fossier, E. Diamanti, T. Debuisschert, R. Tualle-Brouri, and P. Grangier, Improvement of continuous-variable quantum key distribution systems by using optical preamplifiers, Journal of Physics B: Atomic, Molecular and Optical Physics 42, 114014 (2009).
  19. [19]Y. Shaked, Y. Michael, R. Z. Vered, L. Bello, M. Rosenbluh, and A. Pe’er, Lifting the bandwidth limit of optical homodyne measurement with broadband parametric amplification, Nature Communications 9, 609 (2018).DOI
  20. [20]C. M. Caves, Quantum limits on noise in linear amplifiers, Physical Review D 26, 1817 (1982).DOI
  21. [21]M. Hillery, Quantum cryptography with squeezed states, Physical Review A 61, 022309 (2000).DOI
  22. [22]H. Zhang, J. Fang, and G. He, Improving the performance of the four-state continuous-variable quantum key distribution by using optical amplifiers, Physical Review A 86, 022338 (2012).DOI
  23. [23]Y. Guo, R. Li, Q. Liao, J. Zhou, and D. Huang, Performance improvement of eight-state continuous-variable quantum key distribution with an optical amplifier, Physics Letters A 382, 372 (2018).DOI
  24. [24]Y.-C. Zhang, Z. Li, C. Weedbrook, S. Yu, W. Gu, M. Sun, X. Peng, and H. Guo, Improvement of two-way continuous-variable quantum key distribution using optical amplifiers, Journal of Physics B: Atomic, Molecular and Optical Physics 47, 035501 (2014).
  25. [25]L. Huang, Y. Zhang, Y. Huang, T. Jiang, and S. Yu, Improvement of unidimensional continuous-variable quantum key distribution systems by using a phase-sensitive amplifier, Journal of Physics B: Atomic, Molecular and Optical Physics 52, 225502 (2019).DOI
  26. [26]J. Qi, J. Peng, W. Liu, C. He, and M. Zhang, Performance improvement of self-referenced continuous-variable quantum key distribution via optical amplifiers, Laser Physics Letters 18, 085202 (2021).DOI
  27. [27]Z. Tong, C. Lundström, P. A. Andrekson, C. J. McKinstrie, M. Karlsson, D. J. Blessing, E. Tipsuwannakul, B. J. Puttnam, H. Toda, and L. Grüner-Nielsen, Towards ultrasensitive optical links enabled by low-noise phase-sensitive amplifiers, Nature Photonics 5, 430 (2011).DOI
  28. [28]J. Liao, Z. Chen, J. Wang, H. Xiao, X. Guo, Z. Li, and D. Wang, Experimental demonstration of high-speed continuous variable quantum key distribution enhanced by phase-sensitive amplifier, npj Quantum Information 11, 105 (2025).DOI
  29. [29]R. Simon, Peres-horodecki separability criterion for continuous variable systems, Physical Review Letters 84, 2726 (2000).pubmed.ncbi.nlm.nih.gov/11017310
  30. [30]F. Grosshans, G. Van Assche, J. Wenger, R. Brouri, N. J. Cerf, and P. Grangier, Quantum key distribution using Gaussian-modulated coherent states, Nature 421, 238 (2003).DOI
  31. [31]Z. Zhao and D. Wang, Deterministic minimum-leakage continuous-variable quantum key distribution with phase-conjugated twin beams (2026), arXiv:2607.12432 [quant-ph].arxiv.org/abs/2607.12432
  32. [32]V. C. Usenko and R. Filip, Trusted noise in continuous-variable quantum key distribution: A threat and a defense, Entropy 18, 20 (2016).
  33. [33]I. Devetak and A. Winter, Distillation of secret key and entanglement from quantum states, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 461, 207 (2005).
  34. [34]M. Navascués, F. Grosshans, and A. Acín, Optimality of Gaussian attacks in continuous-variable quantum cryptography, Physical Review Letters 97, 190502 (2006).DOI
  35. [35]R. García-Patrón and N. J. Cerf, Unconditional optimality of Gaussian attacks against continuous-variable quantum key distribution, Physical Review Letters 97, 190503 (2006).
  36. [36]C. Ottaviani, S. Mancini, and S. Pirandola, Gaussian two-mode attacks in one-way quantum cryptography, Physical Review A 95, 052310 (2017).DOI
  37. [37]G. Lindblad, Completely positive maps and entropy inequalities, Communications in Mathematical Physics 40, 147 (1975).DOI
  38. [38]A. n. Oruganti, I. Derkach, R. Filip, and V. C. Usenko, Continuous-variable quantum key distribution with noisy squeezed states, Quantum Science and Technology 10, 025023 (2025).
  39. [39]M. G. Genoni, L. Lami, and A. Serafini, Conditional and unconditional Gaussian quantum dynamics, Contemporary Physics 57, 331 (2016).DOI
  40. [40]S. Yamano, T. Matsuura, Y. Kuramochi, T. Sasaki, and M. Koashi, General treatment of Gaussian trusted noise in continuous variable quantum key distribution (2023), arXiv:2305.17684 [quant-ph].arxiv.org/abs/2305.17684

Paper details

vlnc:W_dmd34h2j
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents