arXiv:2610.16286
Metasurface quantum graphs for polarization-independent certified randomness
Xing Lin, Wenwen Liu, Wange Song, Sixin Chen, Shuang Zhang
Abstract
Quantum random number generators are crucial for information security, yet existing implementations are hindered by bulky optics, device imperfections, and state alignment constraints. Here, using a compact metasurface platform, we achieve polarization-independent certified randomness by establishing a quantitative quantum graph framework. By mapping multi-channel nanophotonic responses onto graph vertices, we assign metasurface measurements an operational information-theoretic meaning, directly connecting weighted graph degrees to conditional min-entropy estimation. The resulting graph-certified measurements yield robust private randomness even under entanglement-based attacks without complex optics. Experimentally, we fabricate 4- and 6-channel dielectric metasurfaces to implement high-dimensional positive operator-valued measure graphs, successfully certifying randomness with heralded single photons and Poissonian light. By integrating complex quantum measurements into a single passive layer, our metasurface quantum graphs provide a scalable route for integrated quantum technologies.
AI Summary
The paper addresses whether a compact, fixed photonic measurement can generate private randomness for arbitrary input polarization, including inputs correlated with an adversary. It implements multi-outcome polarization POVMs using dielectric metasurfaces and represents their measurement elements as vertices of a weighted graph, with edges determined by Hilbert–Schmidt overlaps and vertex weights incorporating channel efficiencies. The polarization Hilbert space remains two-dimensional; the number of measurement outcomes increases. For an $m$-photon input, the proposed graph analysis gives $p_{\mathrm{guess}}\leq\max_i D_i^{-m}$, where $D_i$ is a weighted vertex degree. Accounting for accepted detection patterns, the stated entropy bound is $H_{\min}(A|E)\geq m\log_2(\min_iD_i)-\log_2g_m$, with $g_m$ correcting for retaining only events in which $m$ distinct ports click. The security argument allows Eve to purify the input and extends the tensor-product measurement analysis to coherent attacks across rounds. It relies on characterized measurement operators, rather than device-independent certification. Experiments use fabricated four- and six-channel TiO$_2$ metasurfaces, calibrated optically, with heralded single photons and attenuated laser pulses. Toeplitz hashing yields reported certified randomness of 0.6153 and 0.8059 bits per detected heralded photon, respectively. For Poissonian illumination near the reported optimum of $\mu\approx0.4$, the corresponding yields are approximately 0.1192 and 0.1220 bits per response. Stated practical limitations include unequal channel efficiencies, fabrication imperfections, losses, and reduced entropy from multiphoton components, which are treated conservatively in the source model. Single-detector time multiplexing reduces detector count but limits throughput through detector dead time and the approximately $2.5\,\mu\mathrm{s}$ acquisition frame.
Central claim may not hold: The claimed single-photon entropy per detected response is false for an allowed H–V–D device: an explicit pure input yields 0.4150 bits rather than the certified lower bound of 0.5850 bits because loss conditioning is omitted. The mixed-Fock formula averages entropy contributions, although conditional min-entropy for a mixture tagged to Eve requires taking the negative logarithm of the averaged guessing probability. For the symmetric equal-efficiency configuration, the paper's own entropy formula is log2 binomial(N,m)−m and peaks near N/3, contradicting its claimed maximum near N/2. The Poissonian certification requires characterized photon statistics and a detection model, so it does not establish security for the broadly described uncharacterized or arbitrarily controlled source. The multiphoton proof bounds same-port tensor-power effects rather than the accepted distinct-port measurement and does not justify its proposed postselection correction.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 46 pages, 12 figures