REVIEW 4 minor 1 cited by
A new proof of Lee's conjecture on the Frobenius norm via the matrix Cauchy-Schwarz inequality
T0 review · 0 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read This paper proves, using only Cauchy-Schwarz, that for all complex square matrices A and B the Frobenius norm of A+B is at most sqrt((1+sqrt(2))/2) times the Frobenius norm of |A|+|B|.
desk verdict A clean, checkable elementary proof of Lee's Frobenius-norm conjecture, but the theorem was already proven in 2022, so the contribution is pedagogical simplification rather than new mathematics. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is Lemma 2.2, a trace inequality valid for any contraction Q and positive semidefinite X, Y: 4|Tr(QXY)| <= t Tr($X^{2}$+$Y^{2}$) + (1/t) Tr(XY+YX) for all t>0. It is proved by splitting the trace into a product of two factors, applying the matrix Cauchy-Schwarz inequality from Lemma 2.1, and then using the contraction bound after cyclically permuting the trace. The parameter t is left free until the final step, where t = $\sqrt$(2)-1 balances the two sides so that the right-hand side becomes a perfect square proportional to || |A|+|B| ||$_F^{2}$. This trace lemma is what carries the whole proof, converting the polar-decomposition expansion of ||A+B||$_F^{2}$ into the conjectured bound.
What would settle it
Compute random small cases: choose n=2 or 3, random contractions Q and random positive semidefinite X, Y, and test whether 4|Tr(QXY)| <= t Tr($X^{2}$+$Y^{2}$) + (1/t) Tr(XY+YX) for all t>0; any violation would refute the proof's key lemma. A direct counterexample to the conjecture would be a pair A,B with ||A+B||_F > $\sqrt$((1+$\sqrt$(2))/2) || |A|+|B| ||_F, which a numerical search over small complex matrices could attempt to find.
Extended reading notes
Core claim
The paper's central claim is that the conjecture holds, with the sharp constant $\sqrt$((1+$\sqrt$(2))/2). Writing A = U|A| and B = V|B|, and setting Q = U*V, X = |B|, Y = |A|, the proof reduces the desired norm inequality to Lemma 2.2: for every contraction Q and every positive semidefinite X, Y, 4|Tr(QXY)| is bounded by t Tr($X^{2}$+$Y^{2}$) + (1/t) Tr(XY+YX) for any t>0. Choosing t = $\sqrt$(2)-1 yields 2||A+B||$_F^{2}$ <= ($\sqrt$(2)+1)|| |A|+|B| ||$_F^{2}$, which is exactly the conjecture after dividing by 2. The argument for Lemma 2.2 runs through the matrix Cauchy-Schwarz inequality and the contraction condition ||Q||_infty <= 1, and nothing else is needed.
Load-bearing premise
The proof stands on Lemma 2.2, the four-line trace inequality for contractions and positive semidefinite matrices; if that inequality were false for even one instance, the derivation of the conjecture would not go through, even though the conjecture itself could still be true.
Editorial extensions
If this is right
- The bound holds uniformly in the matrix dimension n, because Lemma 2.2 is independent of n and no step in the proof uses dimension-specific information.
- The free parameter t in Lemma 2.2 shows why the constant is optimal: the choice t = sqrt(2)-1 is exactly where the diagonal and cross-term contributions balance.
- A reader can now derive the conjecture entirely from Cauchy-Schwarz and the arithmetic-geometric mean inequality, with no angle-based geometry required.
- The trace inequality itself is a reusable tool for bounding other quadratic expressions in matrix sums, not just the one used in this conjecture.
Reading between the lines
- Editorial inference: the same polar-decomposition and trace-lemma strategy could be applied to other Schatten p-norms, though the paper does not address whether the constant would remain optimal there.
- Editorial inference: because Lemma 2.2 is stated for any contraction, it may hold under weaker hypotheses such as numerical radius at most one; verifying this would require fresh work.
- Editorial inference: a natural testable extension is to replace the Frobenius norm by the Hilbert-Schmidt norm of infinite-dimensional operators, where the trace inequality and polar decomposition both remain valid.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves Lee's conjecture on the Frobenius norm: for any n x n complex matrices A and B, ||A+B||_F <= sqrt((1+sqrt(2))/2) || |A| + |B| ||_F. The proof reduces the conjecture to a trace inequality (Lemma 2.2) via polar decompositions A = U|A|, B = V|B|, and then applies the Hilbert-Schmidt Cauchy-Schwarz inequality together with the AM-GM inequality. The argument is self-contained and yields the conjectured optimal constant.
Significance. The paper provides a clean, elementary proof of a previously established conjecture (originally proved by Lin and Zhang). The derivation is transparent and the constant emerges naturally from the trace inequality. The proof is fully explicit and appears to be correct, with no hidden assumptions or circular reasoning. While the result is not new, the methodology is elegant and may be of pedagogical interest; the trace inequality in Lemma 2.2 could also be useful in other contexts. The paper is suitable for a journal that publishes short, elementary proofs of known results.
minor comments (4)
- [Section 2, Lemma 2.2] The proof of Lemma 2.2 is correct but terse. The second application of Lemma 2.1 with parameter 1 uses S* = YQ and T = XQ*, and the final contraction bounds Tr(Q*Y^2Q) <= Tr(Y^2) and Tr(QX^2Q*) <= Tr(X^2) follow from ||Q||_∞ <= 1. Please spell out these substitutions explicitly to improve readability.
- [Abstract] The statement that the proof relies 'solely on the Cauchy-Schwarz inequality' is slightly inaccurate, since the AM-GM inequality and the contraction property of partial isometries are also used. Please rephrase to 'mainly' or 'elementary inequalities'.
- [Section 2, Lemma 2.1 proof] The word 'Combing' should be changed to 'Combining'.
- [Section 2, Proof of Conjecture 1.1] Before applying Lemma 2.2, it would be helpful to explicitly note that W = U*V is a contraction because U and V are partial isometries (or unitary in the invertible case).
Circularity Check
No significant circularity: the proof is an independent derivation from Cauchy-Schwarz and trace inequalities.
full rationale
The paper's derivation chain is self-contained. Lemma 2.1 is the matrix Cauchy-Schwarz inequality followed by the scalar AM-GM inequality, both stated and proved in the text. Lemma 2.2 is then derived from Lemma 2.1 with explicit factorization Tr(QXY) = Tr((Y^{1/2}QX^{1/2})(X^{1/2}Y^{1/2})) and uses the contraction property QQ* <= I, Q*Q <= I to bound the resulting traces. The proof of Conjecture 1.1 substitutes the polar decompositions A = U|A|, B = V|B|, sets Q = U*V, X = |B|, Y = |A|, and applies Lemma 2.2; the final choice t = sqrt(2)-1 is an explicit algebraic optimization and is not fitted to the target inequality. Prior work by Lin and Zhang is cited only as historical context, not as a load-bearing premise; the proof does not invoke their theorem or any uniqueness result. There is no fitted input renamed as a prediction, no ansatz imported via citation, and no known result merely renamed. The only minor caveat is that the proof of Lemma 2.2 is terse, but every intermediate step is standard and valid, so this is a clarity concern rather than circularity. Therefore the circularity score is 0.
Assumptions & free parameters
assumptions (4)
- standard math Every complex square matrix A has a polar decomposition A=U|A| with U unitary and |A| positive semidefinite.
- standard math The matrix Cauchy-Schwarz inequality: |Tr(S*T)| <= (Tr(S*S))^(1/2) (Tr(T*T))^(1/2) for all matrices S,T.
- standard math The arithmetic-geometric mean inequality for nonnegative reals: 2ab <= t a^2 + (1/t) b^2 for a,b >=0 and t>0.
- standard math For a contraction Q with operator norm at most 1 and a positive semidefinite matrix Y, Tr(Q*Y^2Q) <= Tr(Y^2); similarly Tr(QX^2Q*) <= Tr(X^2).
Cite this review
Pith. "Pith review of A new proof of Lee's conjecture on the Frobenius norm via the matrix Cauchy-Schwarz inequality." pith.science (2026). https://pith.science/paper/D6VK43OR
@misc{pith2026250702684,
author = {Pith},
title = {Pith review of: A new proof of Lee's conjecture on the Frobenius norm via the matrix Cauchy-Schwarz inequality},
year = {2026},
howpublished = {\url{https://pith.science/paper/D6VK43OR}},
note = {Machine review of arXiv:2507.02684}
}
abstract
In 2010, Eun-Young Lee conjectured that if $A,B$ are two $n\times n$ complex matrices and $\left|A\right|, \left|B\right|$ are the absolute values of $A, B$, respectively, then \[ \|A+B\|_F\le \sqrt{\dfrac{1+\sqrt{2}}{2}}\|\left|A\right|+\left|B\right|\|_F, \] where $\|\cdot\|_F$ is the Frobenius norm of matrices. This conjecture has been proven by Lin and Zhang [J. Math. Anal. Appl. 516 (2022) 126542] by studying inequalities for the angle between two matrices induced by the Frobenius inner product. In this paper, we present a new proof of the same result, relying solely on the Cauchy-Schwarz inequality.
Forward citations
Cited by 1 Pith paper
-
Sharp perturbation bounds on the Frobenius norm of subunitary and positive polar factor
Optimal Frobenius-norm perturbation bounds for subunitary and positive polar factors are derived from convex analysis, along with strengthened Lee, AM-GM, and Cauchy-Schwarz inequalities.
Reference graph
Works this paper leans on
-
[2]
Lee, Rotfel’d type inequalities for norms, Linear Algebra Appl
E.-Y. Lee, Rotfel’d type inequalities for norms, Linear Algebra Appl. 433 (2010) 580-584
2010
-
[3]
J. Lin, Y. Zhang, A proof of Lee’s conjecture on the sum of absolute values of matrices, J. Math. Anal. Appl. 516 (2022) 126542. 4
work page 2022
-
[1]
Bhatia, Matrix Analysis, Springer, New York, 1997
R. Bhatia, Matrix Analysis, Springer, New York, 1997
1997
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.