arXiv:2610.03168
Self-testing ideal quantum measurements
Debashis Saha, Zhen-Peng Xu, Adán Cabello
Abstract
A fundamental prediction of quantum mechanics is that observables can be measured through ideal measurements, that is, measurements that give the same outcome when repeated and do not disturb any observable that is jointly measurable. In terms of quantum instruments, such an ideal measurement is described by projective effects together with the associated Luders state-update rule. Here, we show that quantum mechanics allows for self-testing any ideal measurement. Using solely the output statistics of bipartite Bell tests in which a non-demolition measurement is sometimes added, it is possible to certify in a device-independent manner whether the added measurement is ideal.
AI Summary
Can an uncharacterized device be certified to implement both projective measurement effects and the associated Lüders state update using only observed correlations? This paper gives a device-independent self-testing construction for ideal instruments $\mathcal P_c(X)=\Pi_cX\Pi_c$, with arbitrary outcome number and projector ranks, up to input and outcome-dependent output extraction isometries. The protocol compares bipartite Bell-test statistics with and without an intervening instrument on Bob’s system. The initial test certifies $|\phi_d^+\rangle$ and Alice’s measurements; conditional tests after outcome $c$ certify the projected maximally entangled state, with outcome probability $p(c)=\operatorname{rank}(\Pi_c)/d$. Keeping Alice’s extraction fixed connects these tests to the instrument’s unnormalized Choi operators. Perfect strategies for context-projector games built from merged Peres-24 Kochen–Specker sets provide the construction for ranks at least four. Rank-two and rank-three outcomes use additional CHSH-based tests, requiring two additional Bob settings if no rank-three outcome occurs, and four otherwise. Rank-one outcomes require only Alice–instrument correlations: their output-state freedom is absorbed into the allowed extraction. Certification is therefore modulo outcome-dependent output isometries, rather than in a fixed physical output basis; Bob’s measurements may also depend on $c$. The main theorem assumes exact perfect correlations. A robustness statement bounds the extracted instrument’s generalized Choi trace-distance error by $\delta_1+\delta_2$, given input-state and unnormalized post-measurement self-testing errors $\delta_1$ and $\delta_2$. It does not supply explicit bounds translating observed Bell-score deviations into these errors.
Central claim may not hold: The allowed outcome-dependent output isometries certify only an equivalence class that includes nonrepeatable measure-and-reset instruments, so the abstract's certification of physical ideality is stronger than the formal result. The specified rank-three construction gives 7√2/6 rather than the asserted 4√2/3 in Eq. (B3-2), although replacing Alice's setting x_1 by x_0 repairs this calculation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 11 pages; 2 figures