AI 中文总结
研究在制备与测量框架内比较不同光接收机用于量子通信和随机性认证，通过半定规划利用接收机统计量界定相关量，得出光子计数、连续变量接收机及混合接收机在不同能量下的性能表现，并能为多种量子协议提供接收机级比较。
AI 中文摘要
光接收机的选择决定了传输态的哪些特性在观测数据中可见，从而影响不同量子协议的性能。本文在相同的制备与测量框架内，以半设备无关随机性生成作为主要案例研究，比较连续变量、光子计数和混合接收机。测量设备未作表征，源由其纯信号态的Gram矩阵描述。利用观测到的接收机统计量对与测量设备相关但与输入无关的经典边信息约束下的$H_{\min}(B|X,\Lambda)$进行界定。对于固定的Gram矩阵，通过精确的半定规划获得该界。结果表明，光子计数对固定模相位编码是相位盲的，连续变量接收机在中等能量下给出最高认证熵，混合接收机在低能量下表现更好。相同的接收机统计量还为离散调制连续变量量子密钥分发、量子读取、隐蔽通信和量子签名验证提供了接收机级别的比较。
英文摘要
The choice of optical receiver determines which properties of the transmitted states remain visible in the observed data and therefore affects the performance of different quantum protocols. We compare continuous-variable, photon-counting and hybrid receivers within the same prepare-and-measure framework, using semi-device-independent randomness generation as the main case study. The measurement device is left uncharacterised, while the source is described by the Gram matrix of its pure signal states using an energy-derived overlap constraint, a magnitude-Gram benchmark or the full complex Gram matrix of a certified coherent phase-shift-keyed constellation. Within this framework, the observed receiver statistics are used to bound $H_{\min}(B|X,Λ)$ against classical side information correlated with the measurement device but independent of the input. For a fixed Gram matrix, this bound is obtained from an exact semidefinite program, with complex multi-input cases treated in block-real form and checked through the corresponding dual certificate. Photon counting alone is phase blind for fixed-modulus phase encoding and therefore certifies no worst-case randomness. Continuous-variable receivers give the highest certified entropy at moderate energy, while under the nominal source calibration a hybrid receiver performs better at low energy when the beacon-region label is retained in the output. The same receiver statistics also provide receiver-level comparisons for discrete-modulated continuous-variable quantum key distribution, quantum reading, covert communication and quantum-signature verification, without replacing the full security analysis required for each protocol.
Comments16 pages, 14 figures
