Pith. sign in

REVIEW 3 major objections 4 minor 23 references

Global vs. Product Observables in Bipartite Quantum Systems: The Sharp Bound

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read In bipartite quantum systems, restricting probes to product operators costs at most $\sqrt{2}\,\min\{n,m\}$ in trace norm, and this constant is asymptotically optimal.

desk verdict A likely-correct sharp constant sqrt(2) for trace vs product-observable norms, with a fixable conjugation error in the Khintchine proof and a typo in the stated constant. read the letter →

arxiv 2608.06235 v1 pith:PH64NCKH submitted 2026-08-06 quant-ph math.FA

Zhi Li , Xiaofei Shi This is my paper · ORCID
classification quant-phmath.FA MSC 46B2846L5281P4560B20 PACS 03.67.Mn
keywords tracenorminjectivetensorproductobservablesnoncommutativeKhintchineinequalityHaarunitarycanonicalanticommutationrelationsquantumdatahidingbipartitecorrelation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks how much detecting power is lost when a bipartite quantum system is probed only by product operators—one local measurement on each side—instead of an arbitrary joint measurement. Its answer is a sharp universal bound: for every operator $z$ on $\mathbb{C}^n\otimes\mathbb{C}^m$, the global trace norm $\|z\|_1$ is at most $\sqrt{2}\,\min\{n,m\}$ times the product-probe norm $\|z\|_\varepsilon$, and no smaller constant factor can work in general, since the ratio approaches $\sqrt{2}$ as the smaller dimension grows. This closes the gap between the previously known linear bound $2\min\{n,m\}$ and the trivial lower bound $\min\{n,m\}$, showing that the natural conjecture that the lower bound is sharp is false. The upper bound rests on a new $L_1$ noncommutative Khintchine inequality whose random coefficients are the entries of a Haar-distributed unitary; the lower bound uses a fermionic CAR construction whose singular-value spectrum follows the quarter-circle law. The same sharp constant is shown to govern bipartite correlation measured by joint versus product observables, and it improves the universal quantum data-hiding ratio to $\sqrt{2}\,\min\{n,m\}$.

What carries the argument

The upper-bound argument combines two estimates. Proposition 3.1 bounds the trace norm of a block matrix $z=\sum E_{ij}\otimes x_{ij}$ by $\sqrt{n}\,\|(x_{ij})\|_{L_1[R+C]}$, where the row-plus-column norm is the infimum, over decompositions $x_{ij}=a_{ij}+b_{ij}$, of the trace norm of $(\sum a_{ij}a_{ij}^*)^{1/2}$ plus the trace norm of $(\sum b_{ij}^*b_{ij})^{1/2}$. Theorem 3.2 is the Haar-unitary $L_1$ noncommutative Khintchine inequality: $\mathbb{E}_{w\in U(n)}\|\sum w_{ij}x_{ij}\|_1 \ge (1/\sqrt{2}\,n)\|(x_{ij})\|_{L_1[R+C]}$, proved through its dual Fourier-synthesis statement that every coefficient family of row-plus-column norm at most 1 appears as the first-order Fourier coefficients of a function $F:U(n)\to M_m$ with pointwise norm at most $\sqrt{2n}$, built by singular-value truncation and iteration. For the lower bound, the paper forms the CAR matrix $C=\sum E_{ij}\otimes c_{ij}$ from $n^2$ fermionic modes and polar-decomposes it into a partial isometry $V$ that flattens all singular values to 1; the structural fact $T_V(a)^2=0$ for every $a$, together with the quarter-circle limiting distribution of the singular values of $\sqrt{2/n}\,C$, makes the rank of $V$ fill almost all of the ambient $n\cdot 2^{n^2}$ dimensions, forcing the ratio $\|V\|_1/(n\|V\|_\varepsilon)$ to tend to $\sqrt{2}$.

What would settle it

Compute the first few moments of the empirical singular-value distribution of $\sqrt{2/n}\,C$ for small $n$ (say $n=3$ or 4) by exact symbolic Wick expansion: the paper predicts Catalan moments $2,5,14,\ldots$ up to $O(n^{-1})$. If the moments deviate, the quarter-circle step and the $\sqrt{2}$ lower bound fail; separately, a numerical optimization of $\|z\|_1/\|z\|_\varepsilon$ over Hermitian $z$ for fixed $n=m$ could search for violations of the upper bound.

Watch

Extended reading notes

Core claim

The central result is Theorem 1.1: for every $n,m\ge 1$ and every $z\in M_n\otimes M_m$, $\|z\|_1\le \sqrt{2}\,\min\{n,m\}\,\|z\|_\varepsilon$, where $\|\cdot\|_1$ is the trace norm, the supremum of $|\operatorname{tr}(yz)|$ over all contractive joint observables $y$, and $\|\cdot\|_\varepsilon$ is the injective tensor norm, the same supremum restricted to product observables $a\otimes b$. The constant is optimal in the asymptotic sense $\lim_{n\to\infty}\sup_{m\ge 1} C_{n,m}/\min\{n,m\}=\sqrt{2}$, where $C_{n,m}$ is the worst-case norm ratio. In particular, the SWAP-operator lower bound $\min\{n,m\}$ is not attainable as a universal constant; the exact worst-case loss from restricting to product measurements is $\sqrt{2}$ times the smaller local dimension.

Load-bearing premise

The lower bound depends on the appendix's fermionic calculation that the flattened CAR partial isometry still squares to zero and that its singular-value spectrum converges to the quarter-circle law; if the signs or normalization in that combinatorial pairing count are wrong, the extremal constant could shift.

Editorial extensions

If this is right

  • The previously bracketed gap between global and product probing is now pinned down: the worst-case ratio is asymptotically $\sqrt{2}\,\min\{n,m\}$, not the conjectured $\min\{n,m\}$.
  • For every bipartite state, the trace norm of the correlation operator $\rho_{AB}-\rho_A\otimes\rho_B$ is at most $\sqrt{2}\,\min\{n,m\}$ times the largest product-operator correlation function, and the constant remains asymptotically sharp among Hermitian correlation operators with vanishing partial traces.
  • The data-hiding ratio against local operations and the larger classes of LOCC and separable measurements is at most $\sqrt{2}\,\min\{n,m\}$, improving the previous $2\sqrt{2}\,\min\{n,m\}$ bound in the chain $\min\{n,m\}\le R_{SEP}\le R_{LOCC}\le R_{LOCC\to}\le R_{LO}\le \sqrt{2}\min\{n,m\}$.
  • The entire upper-bound argument has been machine-checked in a formal proof assistant, so the inequality itself, the block-matrix estimate, and the Khintchine inequality carry a verified certificate.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the singular-value-flattening step that turns the CAR matrix into a partial isometry is a transferable trick; the same flatten-then-prove-square-zero strategy could yield sharp constants in other norm-comparison problems where spectra are otherwise uneven.
  • Editorial inference: the second local dimension in the lower-bound witness is $2^{n^2}$, far larger than the first dimension; if a polynomial-dimensional witness exists, the $\sqrt{2}$ asymptotics would apply in more realistic finite-size settings, and the paper's open question on dimension reduction is the natural next test.
  • Editorial inference: because the lower-bound argument never uses randomness, one could try replacing the CAR construction with random Gaussian or Haar-random matrices; the quarter-circle law is the only spectral ingredient, so such an ensemble might reproduce the same $\sqrt{2}$ constant with a simpler proof.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper determines the sharp universal comparison between the trace norm and the injective tensor norm associated with the trace norms on the two factors of a bipartite matrix algebra. For every z in M_n ⊗ M_m it proves ||z||_1 ≤ √2 min(n,m) ||z||_ε, and shows that the constant √2 is asymptotically optimal as n → ∞ (with m allowed to grow very fast). The upper bound is obtained from a new L1 noncommutative Khintchine inequality for random coefficients that are entries of a Haar unitary matrix, proved via a dual bounded Fourier-synthesis theorem. The lower bound uses a CAR-based construction, polar decomposition to flatten singular values, and a quarter-circle law for the singular-value distribution of the CAR matrix. Applications are given to bipartite correlation measures and to quantum data hiding, improving the known upper bound from 2√2 min(n,m) to √2 min(n,m). The paper also states that the upper-bound argument has been machine-checked in Lean 4.

Significance. If the proof is correct, this is a clean and nontrivial resolution of a natural norm-comparison problem. The constant √2 (rather than 1 or 2) is a genuine surprise, and the upper-bound technique—a sharp Haar-unitary noncommutative Khintchine inequality—is likely to be useful elsewhere. The lower-bound construction, which combines CAR algebras with a functional-calculus flattening step and a moment-method proof of a quarter-circle law, is conceptually interesting. The applications to correlation measures and data hiding are immediate and relevant. The claimed Lean formalization is a genuine strength, though the manuscript does not provide enough detail to verify it independently. The main caveat is that the printed upper-bound proof contains a conjugation error in the Fourier-coefficient extraction that is load-bearing; this is local and repairable, but it must be fixed before the paper can be accepted.

major comments (3)
  1. [§3.2, Eqs. (3.6)–(3.8)] The Fourier-coefficient convention is inconsistent. Equation (3.6) defines y_ij = E_w[w_ij F(w)], but then Eq. (3.8) cannot be correct: with S(w) = √n ∑ w_ij y_ij, Schur orthogonality gives E_w[w_ij S(w)] = 0 and E_w[\bar w_ij S(w)] = y_ij/√n, not y_ij. Consequently, the claim in the proof of Theorem 3.3 that "Schur orthogonality gives y^{(0)}_ij = y_ij" is false as stated. The correct extraction requires a conjugate: y_ij = E_w[\bar w_ij F(w)] (with S built from \bar w_ij, or with an equivalent normalization). This is load-bearing because it is the mechanism by which the first-order Fourier data of F are matched to the given family (y_ij).
  2. [§3.2, Eq. (3.40)] The duality step in the proof of Theorem 3.2 is only valid under the conjugate coefficient convention. With the printed definition y_ij = E_w[w_ij F(w)], the expression on the left of Eq. (3.40) is ∑ tr(y*_ij x_ij) = E_w tr(F(w)^* ∑ \bar w_ij x_ij), not E_w tr(F(w)^* ∑ w_ij x_ij). The printed equality would hold if y_ij = E_w[\bar w_ij F(w)] were used. Thus the proof as written does not connect the supremum over y_ij to the norm ∥(x_ij)∥_{L1[R+C]} via Eq. (2.7).
  3. [§3.2, Theorem 3.2 and Eq. (3.45)] The constant in Theorem 3.2 is displayed as "1√ 2n", which is ambiguous between 1/(√2 n) and 1/√(2n). The proof via Theorem 3.3, whose bound is ∥F∥∞ ≤ √(2n), yields E∥∑ w_ij x_ij∥_1 ≥ (1/√(2n)) ∥(x_ij)∥_{L1[R+C]}. Only this version combines with Proposition 3.1 to produce the chain ∥z∥_1 ≤ √2 n ∥z∥_ε in Eq. (3.45); if the constant were instead 1/(√2 n), the chain would give √2 n^{3/2}. The statement should state explicitly \frac{1}{\sqrt{2n}} and the chain in Eq. (3.45) should be aligned with it.
minor comments (4)
  1. [§3.3] The claimed Lean formalization would be more useful with a commit hash or a precise pointer to the declaration `UpperBound.upper_bound`, especially because the printed proof has the conjugation issue; quoting the exact formalized statement of Theorem 3.2 would help resolve the ambiguity in the constant.
  2. [Appendix A, Eqs. (A.17)–(A.19)] The statement that noncrossing pairings have Wick sign +1 and all other permutations are of lower order is standard but is asserted rather than justified; a short explanation or reference for the sign convention would improve readability.
  3. [§5.2, Eq. (5.19)] The norm ∥·∥_M is defined for Hermitian h, but Proposition 5.3 applies it to arbitrary z; the extension to non-Hermitian operators by taking absolute values should be stated explicitly.
  4. [Throughout] The typesetting of constants such as "1√ 2n" and "√ 2n" is ambiguous; the authors should use explicit parentheses or \frac throughout, particularly in the abstract, Theorem 3.2, and Eq. (3.45).

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the upper bound is derived from an independent Haar-unitary Khintchine inequality, the lower bound from an explicit CAR construction, and the only self-citations are non-load-bearing or machine-checked.

full rationale

Walking the derivation chain: the upper bound in Theorem 1.1 is assembled from Proposition 3.1 (an elementary row/column block-matrix estimate) and Theorem 3.2 (a Haar-unitary noncommutative Khintchine inequality). Theorem 3.2 is proved through its dual Theorem 3.3, whose proof is an explicit truncation-and-iteration construction over first-order Fourier coefficients; no step in that proof assumes the final norm comparison ||z||_1 <= sqrt(2) min{n,m} ||z||_epsilon. The lower bound is a separate constructive witness: the CAR matrix C, its polar part V, the square-zero identity from Proposition 4.1, the U(n)xU(n)-invariance argument fixing ||T_V(a)||_2, and the combinatorial moment computation in Theorem 4.3 leading to the quarter-circle law. None of these inputs is defined in terms of the target ratio ||z||_1/||z||_epsilon, and the target constant sqrt(2) is not smuggled into the lower-bound construction. The only self-citations are [LS26], the authors' Lean formalization of the upper bound, and [LFH26], mentioned only as a previous collaboration in the acknowledgements. [LS26] is machine-checked evidence and therefore independent support under the review rules; [LFH26] is not load-bearing. The skeptic's reported inconsistency in the Fourier-coefficient definitions (eqs. 3.6/3.8/3.40, conjugate placement and the displayed 1/(sqrt(2)n) versus the proof's sqrt(2n) bound) is a correctness or consistency concern in the printed derivation, not a circularity: Theorem 3.2 is not assumed as the conclusion, and correcting the coefficient definitions would not make the derivation depend on Theorem 1.1. Overall, the central claims are self-contained against external or machine-checked inputs, so the circularity burden is low.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

No fitted parameters: the theorem is parameter-free and universal over dimensions n,m; the polynomial q in (4.8) is an exact interpolant on a finite spectrum, not a fitted quantity. No new physical entities are postulated; the fermionic modes and the partial isometry V are mathematical constructions from known objects.

assumptions (7)
  • standard math Haar moment formula for U(n), eq. (3.15), from Collins 2003 and Collins-Sniady 2006.
    Used in Proposition 3.4 to compute E(S*S)^2.
  • standard math Operator concavity of sqrt and operator convexity of t^2.
    Used in Proposition 3.1 and Proposition 3.4 for matrix inequalities.
  • standard math Duality between L1[R+C] and L-infinity[R intersect C] under the trace pairing, eq. (2.7).
    Used to prove Theorem 3.2 from Theorem 3.3.
  • domain assumption Fermionic Wick formula for CAR, eq. (A.17).
    Gives the Gaussian pairing rule for traces of products of creation and annihilation operators in the Fock representation.
  • standard math Method of moments and compact support imply weak convergence (Marchenko-Pastur and Mingo-Speicher).
    Used in Theorem 4.3 to pass from Catalan moments to the quarter-circle law.
  • standard math Schur's lemma and uniqueness of the U(n) x U(n)-invariant quadratic form on M_n.
    Used in Corollary 4.2 to determine the norm of T_V(a).
  • standard math Existence of a polynomial interpolating 1/sqrt(lambda) on the nonzero spectrum of C*C.
    Used to realize the partial isometry V as C_q(C*C).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Global vs. Product Observables in Bipartite Quantum Systems: The Sharp Bound." pith.science (2026). https://pith.science/paper/PH64NCKH

@misc{pith2026260806235,
  author       = {Pith},
  title        = {Pith review of: Global vs. Product Observables in Bipartite Quantum Systems: The Sharp Bound},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PH64NCKH}},
  note         = {Machine review of arXiv:2608.06235}
}
abstract

To probe a bipartite quantum system, one may use arbitrary global operators or restrict to product operators acting separately on the two subsystems. We determine the sharp universal comparison between the resulting norms. For every $z\in M_n\otimes M_m$, we prove $\|z\|_1\leq\sqrt{2}\min\{n,m\}\|z\|_\varepsilon$, where $\|\cdot\|_1$ is the trace norm and $\|\cdot\|_\varepsilon$ is the injective tensor norm associated with the trace norms on $M_n$ and $M_m$. To prove the upper bound, we establish an $L_1$ noncommutative Khintchine inequality whose random coefficients are the entries of a Haar unitary. We also show that the coefficient $\sqrt{2}$ is sharp. As applications, we show that the same sharp constant governs the gap between bipartite correlation measured in trace norm and that measured by a correlation function, and obtain an improved universal upper bound for quantum data hiding. The upper bound has also been formalized and machine-checked in Lean.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 12 canonical work pages

  1. [1]

    Szarek, and Andreas Winter

    Guillaume Aubrun, Ludovico Lami, Carlos Palazuelos, Stanis aw J. Szarek, and Andreas Winter. Universal gaps for XOR games from estimates on tensor norm ratios. Communications in Mathematical Physics , 375(1):679--724, 2020

    work page 2020

  2. [2]

    Fernando G. S. L. Brand \ a o and Micha Horodecki. Exponential decay of correlations implies area law. Communications in Mathematical Physics , 333(2):761--798, 2015

    work page 2015

  3. [3]

    Some properties of crossings and partitions

    Philippe Biane. Some properties of crossings and partitions. Discrete Mathematics , 175(1--3):41--53, 1997

    work page 1997

  4. [4]

    Willian H. G. Corr \^e a, Ludovico Lami, and Carlos Palazuelos. Maximal gap between local and global distinguishability of bipartite quantum states. IEEE Transactions on Information Theory , 68(11):7306--7314, 2022

    2022

  5. [5]

    Moments and cumulants of polynomial random variables on unitary groups, the Itzykson--Zuber integral, and free probability

    Beno \^i t Collins. Moments and cumulants of polynomial random variables on unitary groups, the Itzykson--Zuber integral, and free probability. International Mathematics Research Notices , 2003(17):953--982, 2003

    2003

  6. [6]

    Integration with respect to the Haar measure on unitary, orthogonal and symplectic group

    Beno \^i t Collins and Piotr \'S niady. Integration with respect to the Haar measure on unitary, orthogonal and symplectic group. Communications in Mathematical Physics , 264(3):773--795, 2006

    2006

  7. [7]

    DiVincenzo, Debbie W

    David P. DiVincenzo, Debbie W. Leung, and Barbara M. Terhal. Quantum data hiding. IEEE Transactions on Information Theory , 48(3):580--599, 2002

    2002

  8. [8]

    The Grothendieck inequality for bilinear forms on C^* -algebras

    Uffe Haagerup. The Grothendieck inequality for bilinear forms on C^* -algebras. Advances in Mathematics , 56(2):93--116, 1985

    work page 1985

Show all 23 references
  1. [9]

    Grothendieck type norms for bilinear forms on C^* -algebras

    Uffe Haagerup and Takashi Itoh. Grothendieck type norms for bilinear forms on C^* -algebras. Journal of Operator Theory , 34(2):263--283, 1995

    1995

  2. [10]

    On the best constants in noncommutative Khintchine -type inequalities

    Uffe Haagerup and Magdalena Musat. On the best constants in noncommutative Khintchine -type inequalities. Journal of Functional Analysis , 250(2):588--624, 2007

    2007

  3. [11]

    A unified framework for locally stable phases

    Zhi Li, Raz Firanko, and Timothy H Hsieh. A unified framework for locally stable phases. arXiv preprint arXiv:2605.00088 , 2026

    2026 arXiv

  4. [12]

    In \'e galit \'e s de Khintchine dans C_p (1<p< )

    Fran c oise Lust-Piquard. In \'e galit \'e s de Khintchine dans C_p (1<p< ) . Comptes Rendus de l'Acad \'e mie des Sciences. S \'e rie I. Math \'e matique , 303(7):289--292, 1986

    1986

  5. [13]

    Non-commutative Khintchine and Paley inequalities

    Fran c oise Lust-Piquard and Gilles Pisier. Non-commutative Khintchine and Paley inequalities. Arkiv f \"o r Matematik , 29(2):241--260, 1991

    1991

  6. [14]

    Ultimate data hiding in quantum mechanics and beyond

    Ludovico Lami, Carlos Palazuelos, and Andreas Winter. Ultimate data hiding in quantum mechanics and beyond. Communications in Mathematical Physics , 361(2):661--708, 2018

    2018

  7. [15]

    Lean formalization for `` Global vs

    Zhi Li and Xiaofei Shi. Lean formalization for `` Global vs. Product Observables in Bipartite Quantum Systems: The Sharp Bound ''. Code repository, 2026. Available at https://github.com/xf-shi/Tensor_Norm

    2026

  8. [16]

    Marchenko and Leonid A

    Vladimir A. Marchenko and Leonid A. Pastur. Distribution of eigenvalues for some sets of random matrices. Mathematics of the USSR-Sbornik , 1(4):457--483, 1967

    1967

  9. [17]

    Mingo and Roland Speicher

    James A. Mingo and Roland Speicher. Free Probability and Random Matrices , volume 35 of Fields Institute Monographs . Springer, New York, 2017

    2017

  10. [18]

    Distinguishability of quantum states under restricted families of measurements with an application to quantum data hiding

    William Matthews, Stephanie Wehner, and Andreas Winter. Distinguishability of quantum states under restricted families of measurements with an application to quantum data hiding. Communications in Mathematical Physics , 291(3):813--843, 2009

    2009

  11. [19]

    Grothendieck 's theorem for noncommutative C^* -algebras, with an appendix on Grothendieck 's constants

    Gilles Pisier. Grothendieck 's theorem for noncommutative C^* -algebras, with an appendix on Grothendieck 's constants. Journal of Functional Analysis , 29(3):397--415, 1978

    1978

  12. [20]

    Introduction to Operator Space Theory , volume 294 of London Mathematical Society Lecture Note Series

    Gilles Pisier. Introduction to Operator Space Theory , volume 294 of London Mathematical Society Lecture Note Series . Cambridge University Press, Cambridge, 2003

    2003

  13. [21]

    Grothendieck 's theorem, past and present

    Gilles Pisier. Grothendieck 's theorem, past and present. Bulletin of the American Mathematical Society , 49(2):237--323, 2012

    2012

  14. [22]

    Quantum XOR games

    Oded Regev and Thomas Vidick. Quantum XOR games. ACM Transactions on Computation Theory (ToCT) , 7(4):1--43, 2015

    2015

  15. [23]

    Terhal, David P

    Barbara M. Terhal, David P. DiVincenzo, and Debbie W. Leung. Hiding bits in Bell states. Physical Review Letters , 86(25):5807--5810, 2001

    2001

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.