arXiv:2610.37347
Measurement incompatibility and noiseless quantum codes
Yu-Xuan Zhang, Jing-Ling Chen
Abstract
We construct a rank-one minimal informationally complete measurement (MIC) with an invertible dual operator and prove that, for arbitrary finite input and output dimensions, a channel universally preserves measurement incompatibility if and only if its input space contains a noiseless code subspace of the full output dimension. Preserving the compatibility threshold of a finite family of binary measurements already guarantees such a code. Furthermore, for every $d\ge2$, there exists a family of binary measurements that is incompatible on the full space but compatible after compression to any proper subspace. These results resolve the conjecture of Chitambar and Zhang and the existence question posed by Uola and coauthors. We also give an explicit construction of exact encodings and show that, for a sufficiently small shift in the compatibility threshold, the transmission error of an approximate encoding is bounded by a constant times the square of the shift. These results establish a connection between measurement compatibility, quantum steering, and quantum error correction.
AI Summary
The paper asks which finite-dimensional channels preserve incompatibility of every finite measurement family under their Heisenberg adjoint, and whether incompatibility can disappear upon compression to every proper subspace. It proves that a channel $\mathcal N:\mathbb C^D\to\mathbb C^d$ universally preserves incompatibility exactly when there is an isometry $W:\mathbb C^d\to\mathbb C^D$ satisfying $\mathcal N(W\rho W^\dagger)=\rho$ for every state. This resolves the Chitambar–Zhang conjecture: the channel must contain a full-output-dimensional noiseless code, requiring no recovery. The proof constructs a rank-one minimal informationally complete measurement with an invertible dual operator. From its $d^2$ effects it builds $d^2$ binary measurements, one depending on a parameter $s$, whose joint-measurability threshold is exactly $s=1$. Expanding the complement of a Stinespring range in this operator basis shows that preserving this threshold forces a noiseless encoding, and gives an explicit construction. The same family, at a suitable parameter, is incompatible in dimension $d$ but compatible after compression to every proper subspace, answering the existence question posed by Uola and collaborators for every $d\ge2$. For a threshold shift $\delta$, the construction yields an approximate isometric encoding with diamond-norm transmission error at most $2c^2\delta^2$, where $c$ depends only on the chosen measurement. This requires the image threshold to lie below the parameter endpoint and $c^2\delta^2<1$. The measurement count is not shown optimal; whether fewer measurements suffice and whether the stability bound can be improved remain open. The results also translate to steering through the incompatibility–steering correspondence.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 5 pages, no figures