arXiv:2610.26603
Beyond Fidelity: Resolving Errors in Photonic Seed State Generation
Sara Bartolucci, Patrick Birchall, Jacob Bulmer, Sam Gartenstein, Mihir Pant, Terry Rudolph, Chris Sparrow, Andrzej Veitia
Abstract
An important aspect of many linear optical quantum computing architectures is probabilistically generating small dual-rail qubit entangled states, or seed states. While many schemes to achieve this have been discovered, attention has been primarily focused on minimizing input resources and maximizing success probability rather than establishing a detailed understanding of linear-optical error performance. In this work, we develop a framework for resolving sensitivities to physically distinct errors in linear optical state generation, allowing us to identify precise ways in which these errors affect the properties of the output state. We use this framework to identify seed state generators in which destructive many-photon interference prevents detector loss and detector noise from causing a falsely successful heralding event. We show that this effect suppresses errors, for example cutting the effective loss on the output of particular Bell state generators in half relative to a standard scheme. We also present a new Bell state generator in which both detector loss and detector noise do not affect the quality of the output Bell state to first order, and a 4GHZ generator with >75% higher success probability than a standard circuit. In addition we provide some useful identities, bounds and invariants for studying linear optical seed state generation.
AI Summary
The paper asks how physical errors in heralded linear-optical seed-state generation translate into logical errors and leakage, and when circuit structure can suppress false heralding. It develops a subspace-resolved description tracking correlated vacuum, multiphoton, Pauli and distinguishability errors, with response coefficients giving their leading-order probabilities conditioned on successful detection. Channel-commutation identities relate input errors to detector errors and unavoidable output errors. In particular, $c_{L_{\rm in}\to(V,I,\ldots)}=1+c_{L_{\rm det}\to(V,I,\ldots)}\geq1$. When every interferometer input contains one photon, corresponding multiphoton and combined addition–loss Pauli-response bounds are $3/2$ and $1/4$, respectively. Calculations propagate error-modified Fock states or detector POVMs through each circuit and project the heralded outputs onto the chosen subspaces. Many-photon interference can eliminate detection patterns adjacent to successful patterns, preventing first-order detector loss or noise from falsely heralding success. A new six-photon Bell generator has both protections and succeeds with probability $1/18$; detector-loss protection halves leading-order single-qubit vacuum leakage from input loss relative to the standard generator. Other circuits convert detector errors into correlated leakage detectable by subsequent fusions. A twelve-photon ballistic four-qubit GHZ generator succeeds with probability $10/729$, over 75% above the $1/128$ benchmark, with detector-loss errors leaking all four qubits. An example shows that seed-state fidelity can reverse circuit rankings relative to fidelity after fusion. The analysis is primarily leading order, excludes internal interferometer imperfections and unbalanced internal loss, and uses a two-internal-mode distinguishability model. Subspace probabilities omit coherences and intra-subspace structure; their use as a reduced state representation requires additional assumptions or twirling.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 18+24 pages, 13 figures