arXiv:2609.40343
Quantum de Finetti theorems for states and channels in any distance measure
Liuhang Ye, Bjarne Bergh, Nilanjana Datta
Abstract
Standard finite quantum de Finetti theorems approximate the $k$-system marginals of permutation-invariant states of $n$-systems by mixtures of independent and identically distributed (iid) states, usually in trace distance. We prove both standard and Renner's exponential de Finetti theorems in the stronger form of operator inequalities, implying bounds in every Schatten norm and for every quantum Rényi divergence satisfying data processing. In the standard case, for fixed local dimension, our $k/n$ error bound in max-relative entropy improves on the previously best known $k^2/n$ scaling, even in the classical setting. The operator-inequality approach is particularly suited to study channel de Finetti representations because operator order between Choi states is equivalent to completely positive (CP) order between the underlying channels. For permutation-covariant channels $N^{(n)}:A^{\otimes n}\to B^{\otimes n}$, where $d_A=\dim A$, we prove that the $k$-system reduced channel is CP-dominated by a mixture of tensor-power channels with error $O(k/\sqrt{n})$ and polynomial dependence on the local dimensions, addressing a question raised by Berta et al. [Math. Program. 194, 781-829 (2022)]. Under the no-signalling condition, we also prove an exponential channel de Finetti theorem where the approximating mixture consists of Choi-almost-iid channels, whose normalized Choi states are almost-iid in the sense of Mazzola-Sutter-Renner. In the case of $r$ defects, the representation error is at most $\mathrm{poly}(n)\bigl(2d_A^4k^3/(nr^2)\bigr)^{(r+1)/2}$ and decays exponentially in $n$ for a suitable choice of parameters $r$ and $k$.
AI Summary
Finite quantum de Finetti theorems usually approximate symmetric-state marginals in trace distance. This paper asks whether analogous representations hold in stronger distinguishability measures and whether channel versions can avoid the exponential-in-$k$ loss from converting Choi-state trace distance to diamond norm. For a permutation-invariant $n$-partite state of local dimension $d$, it proves $(1-\varepsilon)\rho^{(k)}\leq\int\omega_\sigma^{(k,r)}\,d\mu(\sigma)$, where the atoms are almost-iid with at most $r$ defects and $\varepsilon\leq(n-k+1)^{d^2-1}e^{-(n-k)(r+1)/n}$. For $r=0$, the atoms are iid and $\varepsilon\leq d^2k/n$, improving the previous $k^2/n$ max-relative-entropy scaling at fixed dimension, including classically. Operator domination yields bounds for all Schatten norms and suitable divergences, including data-processing Rényi divergences. For permutation-covariant channels, the reduced channel is CP-dominated by an iid-channel mixture with error $O(k/\sqrt n)$ at fixed dimensions and polynomial dimension dependence; diamond distance is at most twice this error. Without no-signalling, reduction means supplying maximally mixed discarded inputs. Under no-signalling, an almost-iid-channel representation has error at most $4(n+1)^{3d_A^2d_B^2/2-1}[2d_A^4k^3/(nr^2)]^{(r+1)/2}$. The proofs compute Haar-averaged defect-sector operator identities. Channel representations additionally enforce trace-preserving Choi marginals through completion constructions; the exponential result uses constrained Haar integration and marginal repair preserving almost-iid support. Almost-iid channels are defined through their Choi states. The standard channel bound vanishes for $k=o(\sqrt n)$. The exponential bound requires substantial discarding and does not cover retaining nearly all systems; that regime remains open.
GPT-6 Astra later helped improve the estimation of several error bounds in the appendices, and Claude Fable 5.1 was used to improve parts of the introduction.
Error found: The corollary's purified-distance smoothing bound is false: for pure ρ and σ=ρ, D_max^ε(ρ‖ρ)=log(1−ε²)>log(1−ε), so its correction term must instead be log(1−ε²). The smoothing proof incorrectly asserts that λρ is feasible without restricting λ relative to the smoothing radius; under purified distance, feasibility requires √(1−λ)≤ε.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 52 pages