arXiv:2610.06247
High-Fidelity Remote Graph State Preparation for Blind Quantum Computation
Jiawei Cai, Rex Fleur, Benedikt Tissot, Wolfgang Löffler, Tzula B. Propp
Abstract
Measurement-based quantum computation (MBQC) relies on entangled graph states, yet existing remote state preparation (RSP) protocols prepare only separable states, requiring subsequent entangling gates on the remote server. Here, we introduce Remote Graph State Preparation (RGSP), a framework that prepares arbitrary graph states directly from a single high-dimensional photonic qudit. By encoding multiple qubits and their graph connectivity into the photon's structured phase profile, RGSP can reduce or, for small computations, completely eliminate the need for server-side entangling operations among qubits. We show that under a reasonable noise model the resulting state fidelity is independent of the graph topology. Under cumulative fiber phase drift, we demonstrate that this fidelity is significantly enhanced by a ``highest-weight-first'' mode reordering strategy. Finally, we show that RGSP enables reduced-SWAP universal blind quantum computing on arbitrary topologies, reducing qubit overhead e.g., for a six-qubit Quantum Fourier Transform from $420$ RSP qubits with a standard brickwork topology to $48$ RSP qubits. These results establish RGSP as a resource-efficient, topology-invariant primitive for quantum-secured cloud computing.
AI Summary
The paper asks whether remote preparation can supply entangled MBQC resources directly, rather than product states requiring subsequent server-side entangling gates. It proposes remote graph-state preparation (RGSP): a single photon occupies $2^n$ coherent modes, with phases $\phi_x=\sum_i\theta_i x_i+\pi\sum_{i<j}A_{ij}x_ix_j$ encoding local rotations and graph adjacency. Mode-conditioned photon–memory interactions, memory Hadamards, and photonic which-mode erasure transfer this phase pattern to an $n$-qubit register, up to known corrections. Analytical derivations establish the state-transfer construction and topology-independent fidelity under the assumed phase-only noise model: $F=|\sum_xw_xe^{i\epsilon_x}|^2$, because the programmed graph phases cancel from the overlap. The paper analyzes independent Gaussian noise and cumulative Wiener drift, and numerically studies a highest-Hamming-weight-first mode ordering intended to improve fidelity with loss-compensated mode weights. It also describes graph-edge toggling through additional quadratic phases and sequential multi-client phase accumulation. For blind computation, a proposed tube-graph architecture prepares connected subgraphs optically and uses server-generated entanglement between them. A compiler-based comparison reports 48 remotely prepared qubits for a six-qubit QFT, versus 420 for a brickwork implementation. Trap isolation, position verification, and multi-party delegation are discussed as applications. The paper explicitly acknowledges exponential optical-mode scaling, motivating preparation and stitching of small subgraphs rather than entire large computations. Fidelity analysis assumes independently compensated amplitude loss and phase noise as the remaining decoherence source; the ordering benefit depends on the mode weights used. Multi-client phase composition alone is explicitly not a security proof.
Central claim may not hold: The claimed highest-weight-first improvement uses pre-compensated input weights instead of the uniform conditional output weights, under which the paper's phase-noise model predicts no mode-ordering dependence of graph-state fidelity. The stated Wiener model makes coherence depend on temporal separations rather than absolute arrival times, so transmitting important modes earlier is not by itself a valid explanation of improved fidelity. The headline QFT resource comparison is not reproducibly reconciled with the width-dependent counting equation, although this omission does not establish that the counts are numerically wrong. The QFT4 and QFT5 resource reductions should be 82.2% and 85.7%, respectively, rather than the table's 79.1% and 81.4%.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.