REVIEW 4 major objections 4 minor 21 references
Matrix Ordering through Spectral and Nilpotent Structures in Totally Ordered Complex Number Fields
T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read A total order on complex numbers supports a new partial order on all square matrices, extending Loewner's order to non-Hermitian systems.
desk verdict The SNO order is not antisymmetric, so the paper's central theorem is false; the Jordan block computations are interesting but the framework collapses. 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 machinery is the lexicographic total order z1 ≤ z2 if Re z1 < Re z2 or (Re z1 = Re z2 and Im z1 ≤ Im z2), together with the spectral-and-nilpotent representation R(X) built from a Jordan decomposition. Ordering is carried by weak majorization on sorted eigenvalue lists and by the generalized dominance order on partitions of Jordan-block sizes, whose quantitative work is done by the generalized dominance ordering distance D_{p,q}(j). The same total order on complex numbers feeds the affine T-transformations used to prove Schur-convexity and the monotonicity criteria.
What would settle it
Find a 2x2 non-Hermitian matrix pair X ⪯SN Y and a complex polynomial f with f′(λk)≠0 at all eigenvalues such that f(X) ⋠SN f(Y); by Corollary 4 and Theorem 7(A), such a pair would directly contradict the claimed monotonicity, and it can be computed by comparing the Jordan block sizes of f(X) and f(Y).
Extended reading notes
Core claim
The central claim is that SNO is a partial ordering on same-sized square matrices, defined through a unique spectral-and-nilpotent representation R(X) that lists eigenvalues in the new complex total order and then lists the sizes of Jordan blocks for each eigenvalue in decreasing order. Matrix comparison first compares the spectral list by weak majorization with complex entries; if the spectral lists are equal, it compares the nilpotent lists by a lexicographic dominance order on partitions. The paper further claims that this order is reflexive, antisymmetric, and transitive, that it reduces to Loewner order when restricted to Hermitian matrices with equal representations, and that monotone increasing or decreasing complex functions satisfying certain derivative and difference-sum conditions preserve SNO. Finally, it derives an operator-convexity characterization for analytic functions using 2x2 block matrices, extending the Hansen–Pedersen method from Hermitian to general matrices.
Load-bearing premise
The load-bearing premise is that the lexicographic total order on real and imaginary parts is a meaningful comparison for eigenvalues, even though it does not respect complex multiplication, so the paper must modify multiplication compatibility in Section 2.2.2 and rely on derivative conditions to carry the ordering through functions.
Editorial extensions
If this is right
- If SNO is a valid partial order, matrix inequalities can compare non-Hermitian matrices by jointly constraining their spectra and their non-diagonalizable parts, not just their Hermitian representatives.
- Complex-valued majorization and the complex Schur–Ostrowski criterion would give practical tests for when a symmetric function of eigenvalues is ordered under SNO.
- The Jordan-block transformation rules would let engineers and physicists predict how functions applied to a non-normal matrix reshape its nilpotent structure, based only on derivatives at the eigenvalues.
- The Hansen–Pedersen type convexity equivalence would yield operator-convexity criteria for analytic functions on general square matrices, with applications in non-Hermitian quantum systems.
- Since the Loewner order is a special case when the matrices are Hermitian, the SNO framework inherits the classical theory and embeds it in a larger order-theoretic setting.
Reading between the lines
- One implication the author leaves implicit is that the lexicographic complex order is not compatible with field multiplication, so SNO's monotonicity results likely require the function's derivative conditions to do the real work; a natural test is whether a small perturbative example with eigenvalues differing only in imaginary parts preserves the claimed inequalities.
- The SNO order may also offer a way to rank exceptional points in non-Hermitian physics, since it tracks Jordan-block sizes explicitly, though the paper does not develop this application.
- A concrete extension would be to check whether SNO compares with existing pseudospectrum-based comparison frameworks for non-normal matrices, for instance whether SNO-monotone functions are also stable under pseudospectral perturbations.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a lexicographic total order on the complex numbers and uses it to define the Spectral and Nilpotent Ordering (SNO) on square matrices, based on a weak-majorization comparison of spectral lists and a dominance comparison of Jordan block sizes. It claims that SNO is a partial order extending the Loewner order to arbitrary matrices, and it develops complex-domain majorization, a Schur-Ostrowski criterion, formulas for Jordan blocks of f(X), and monotonicity and convexity criteria for functions under SNO.
Significance. If the central construction were valid, the paper would offer a general framework for comparing non-Hermitian matrices, which is a topic of current interest. The paper does make its objects explicit: the total order, the SNO definition, and the Jordan-block size formulas in Lemmas 8 and 10 are concrete and checkable. However, the main claim that SNO is a partial order fails because of a simple antisymmetry counterexample, and several transfer arguments used in later sections rely on unjustified blockwise or affine-combination steps. There are no machine-checked proofs or reproducible code artifacts. The paper's significance is therefore currently limited by the correctness of its foundational ordering.
major comments (4)
- [Definition 1 and Theorem 1] The relation ⪯SN is not antisymmetric, so Theorem 1 is false. Definition 1 declares X1 ⪯SN X2 whenever Eq. (24) holds, i.e. whenever [λ(X1)] ⪯w [λ(X2)] under weak majorization. Weak majorization is reflexive, so any two matrices with identical spectral lists satisfy both X1 ⪯SN X2 and X2 ⪯SN X1 through the first disjunct, without any use of the nilpotent comparison in the second disjunct. For example, take A = [[0,1],[0,0]] and B = [[0,0],[0,0]]. Both have spectral part [0,0], so Eq. (24) gives A ⪯SN B and B ⪯SN A; yet A is not similar to B because their Jordan types are [2] and [1,1], respectively. The proof of Theorem 1 handles the case a = c only under the second disjunct, and it incorrectly treats the first disjunct as a strict comparison. This invalidates the central claim that SNO is a partial order.
- [Theorem 8, Eqs. (162) and (167)] The proof of the convexity characterization assumes that an SNO inequality between block-diagonal matrices implies the same SNO inequality between the corresponding diagonal blocks. From Eq. (162), the author concludes f(CHXC) ⪯SN CHf(X)C by comparing only the upper-left blocks of two block-diagonal matrices. SNO is defined on the combined spectral data of the whole matrix, and no monotonicity property for taking direct summands is stated or proved. The same unsupported step is used in Eq. (167) for tX + (1−t)Y. Therefore the equivalence in Theorem 8 is not established.
- [Section 4.1, Lemma 4] The affine T-transform argument is not valid over C. The proof asserts that yi − xi, xj − yj, and yi − yj are all > 0 + 0ι and defines β by Eq. (31), but coordinatewise inequalities do not follow from majorization of sorted complex vectors; for example, x = (2,2,0) and y = (3,1,0) satisfy x ≺ y while y2 − x2 < 0. The formula also degenerates when yi = yj. Standard T-transform proofs require β ∈ [0,1], whereas the paper explicitly relaxes this to affine combinations without supplying a valid complex analogue. Since Lemma 6 and Theorem 2 both rely on Lemma 4, the complex Schur-Ostrowski criterion is not supported.
- [Section 2.2, Eqs. (5)-(8) and (12)-(13)] The multiplication and division rules for the lexicographic total order are stated without proof. Because C with the lex order is not an ordered field, the modified compatibility conditions in Eqs. (12)-(13) are additional axioms rather than consequences of Eqs. (1)-(2). These conditions are used in later arguments, including the T-transform analysis in Section 4 and the sign analysis in Theorem 2. The paper should either prove their consistency with the lex order or state clearly that they are imposed separately; as written, the algebraic foundation of the complex majorization framework is incomplete.
minor comments (4)
- [Section 2.2.1, Eq. (4)] The sentence 'For any real numbers z1, z2, and z3' should say 'complex numbers' to match the surrounding context.
- [Theorem 2, Eq. (50)] The imaginary part of the product is written as ℑ(ϵ)DR_{i,j}(f) + ℜ(ϵ)DR_{i,j}(f); the second term should be ℜ(ϵ)DI_{i,j}(f).
- [Section 3.3, Loewner order remark] The claim that the Loewner order is a special case of SNO is not substantiated; for Hermitian matrices, Loewner order is not equivalent to weak majorization of the eigenvalue lists alone, and the sentence needs a precise statement with a proof or a citation.
- [Theorem 3, proof of majorization preservation] The sentence 'We first show that any complex-valued function f(z) is monotone increasing ... will have ≺w preserving property' is false without the additional difference-sum condition Eq. (64); the proof should clearly state that Eq. (64) is an extra hypothesis and should justify why it is compatible with x ≺w y.
Circularity Check
One supporting majorization-preservation theorem reduces to its own conclusion; the SNO partial-order claim has a separate antisymmetry flaw.
-
self definitional
[Section 4.2, Theorem 3, Eq. (64) and its proof; similarly Theorem 4, Eq. (70)]
"and the following difference sum conidtion for the increasing function f f (xi) − f (yi) ≤ P_{i−1}_{j=1} (f (yj) − f (xj)), (64) where i = 2, . . . , n. Then, the function f is ≺w preserving. ... from the monotone increasing property of the function f and the difference sum condition provided by Eq. (64), we have [partial sums] ... Therefore, we have f (x) ≺w f (y) given x ≺w y."
Condition (64) is not an independent hypothesis: for each i ≥ 2 it rearranges to Σ_{j=1}^i f(x_j) ≤ Σ_{j=1}^i f(y_j), and monotonicity supplies the i = 1 partial sum. These are exactly the partial-sum inequalities defining f(x) ≺w f(y) in Eq. (57). Thus the theorem assumes the conclusion it claims to establish. Theorem 7 later invokes Theorem 3 to conclude f(X) ≺SN f(Y), so the SNO-monotonicity step inherits the same circularity by construction.
full rationale
The paper's main structure—lexicographic total order on C, SNO built from weak spectral majorization plus nilpotent dominance, the complex-domain Schur–Ostrowski criterion, and the Hansen–Pedersen style convexity equivalences—is mostly a transcription of known arguments rather than a circular derivation. The Schur–Ostrowski proof follows the standard T-transform route, and Theorem 8 follows the standard 2×2 block-matrix route; those parts have independent mathematical content. The one clear circularity is in the majorization-preservation theorems used later for SNO monotonicity: Theorem 3 (and Theorem 4) imposes a 'difference sum condition' that is a rearrangement of the target partial-sum inequalities defining f(x) ≺w f(y). With monotonicity giving the first partial sum, condition (64) is exactly the conclusion f(x) ≺w f(y), and the proof simply reads the condition back out as the result. Theorem 7 then invokes this condition to assert f(X) ≺SN f(Y), so that portion of the claimed framework reduces to its own conclusion by construction. I do not count the self-citation [19] for the spectral mapping theorem as circular, because the stated formula is a standard and externally checkable result. I also do not count the apparent failure of SNO antisymmetry (equal spectral lists compare in both directions via Eq. (24) regardless of nilpotent data) as circularity; it is a correctness defect rather than a reduction of a prediction to its input. Score 6 reflects partial, not total, circularity: the central ordering and the convexity equivalences have independent mathematical content, but one load-bearing 'prediction'—≺w-preservation, and hence the SNO-monotonicity case built on it—is true by construction.
Assumptions & free parameters
assumptions (6)
- domain assumption A total ordering on complex numbers defined by real part first, then imaginary part.
- domain assumption The spectral and nilpotent representation R(X) is unique and faithfully records the matrix up to similarity.
- ad hoc to paper The weak majorization over complex vectors with the lexicographic order is a partial order.
- ad hoc to paper The generalized dominance order ⊴ for partitions of different integers can be used to compare Jordan block structures across different matrices.
- ad hoc to paper An analytic function f applied to a matrix via the spectral mapping theorem preserves the SNO order in monotonicity theorems.
- ad hoc to paper The proof of convexity in Theorem 8 assumes that inequalities on block diagonal matrices imply inequalities on their diagonal blocks.
Cite this review
Pith. "Pith review of Matrix Ordering through Spectral and Nilpotent Structures in Totally Ordered Complex Number Fields." pith.science (2026). https://pith.science/paper/MMYITTXW
@misc{pith2026250110603,
author = {Pith},
title = {Pith review of: Matrix Ordering through Spectral and Nilpotent Structures in Totally Ordered Complex Number Fields},
year = {2026},
howpublished = {\url{https://pith.science/paper/MMYITTXW}},
note = {Machine review of arXiv:2501.10603}
}
read the original abstract
Matrix inequalities play a pivotal role in mathematics, generalizing scalar inequalities and providing insights into linear operator structures. However, the widely used L\"owner ordering, which relies on real-valued eigenvalues, is limited to Hermitian matrices, restricting its applicability to non-Hermitian systems increasingly relevant in fields like non-Hermitian physics. To overcome this, we develop a total ordering relation for complex numbers, enabling comparisons of the spectral components of general matrices with complex eigenvalues. Building on this, we introduce the Spectral and Nilpotent Ordering (SNO), a partial order for arbitrary matrices of the same dimensions. We further establish a theoretical framework for majorization ordering with complex-valued functions, which aids in refining SNO and analyzing spectral components. An additional result is the extension of the Schur--Ostrowski criterion to the complex domain. Moreover, we characterize Jordan blocks of matrix functions using a generalized dominance order for nilpotent components, facilitating systematic analysis of non-diagonalizable matrices. Finally, we derive monotonicity and convexity conditions for functions under the SNO framework, laying a new mathematical foundation for advancing matrix analysis.
Reference graph
Works this paper leans on
-
[2]
A. W. Marshall and I. Olkin, Inequalities: Theory of Majorization and Its Applications . Academic Press, 1979
work page 1979
-
[1]
Bhatia, Matrix Analysis, 2nd ed
R. Bhatia, Matrix Analysis, 2nd ed. Springer Science & Business Media, 2013
work page 2013
-
[3]
S. Boyd, L. E. Ghaoui, E. Feron, and V . Balakrishnan,Linear Matrix Inequalities in System and Control Theory. Society for Industrial and Applied Mathematics, 1994
work page 1994
-
[4]
M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information , 10th ed. Cam- bridge University Press, 2010
work page 2010
-
[5]
L. Vandenberghe and S. Boyd, “Semidefinite programming,” SIAM Review, vol. 38, no. 1, pp. 49–95, 1996
work page 1996
-
[6]
Random Tensor Inequalities and Tail bounds for Bivariate Random Tensor Means, Part I
S.-Y . Chang, “Random tensor inequalities and tail bounds for bivariate random tensor means, part i,” arXiv preprint arXiv:2305.03301, 2023
work page Pith review arXiv 2023
-
[7]
Random Multiple Operator Integrals
——, “Random multiple operator integrals,” arXiv preprint arXiv:2210.09392, 2022
work page Pith review arXiv 2022
-
[8]
Random Parametrization Double Tensors Integrals and Their Applications
S. Y . Chang, “Random parametrization double tensors integrals and their applications,”arXiv preprint arXiv:2205.03523, 2022
work page Pith review arXiv 2022
Show all 21 references
-
[9]
Random double tensors integrals,
——, “Random double tensors integrals,” arXiv preprint arXiv:2204.01927, 2022
2022 arXiv
-
[10]
Multivariate Mond-Pecaric method with applications to hypercomplex function Sobolev embedding,
S.-Y . Chang, “Multivariate Mond-Pecaric method with applications to hypercomplex function Sobolev embedding,” arXiv preprint arXiv:2405.17203, 2024
2024 arXiv
-
[11]
Generalized converses of operator Jensens inequalities with applications to hypercomplex func- tion approximations and bounds algebra,
——, “Generalized converses of operator Jensens inequalities with applications to hypercomplex func- tion approximations and bounds algebra,” arXiv preprint arXiv:2404.11880, 2024
2024 arXiv
-
[12]
Generalized Choi-Davis-Jensen’s operator inequalities and their applica- tions,
S. Y . Chang and Y . Wei, “Generalized Choi-Davis-Jensen’s operator inequalities and their applica- tions,” arXiv preprint arXiv:2403.04892, 2024
2024 arXiv
-
[13]
Exceptional topology of non-hermitian systems,
E. J. Bergholtz, J. C. Budich, and F. K. Kunst, “Exceptional topology of non-hermitian systems,” Reviews of Modern Physics, vol. 93, no. 1, p. 015005, 2021
2021
-
[14]
L. N. Trefethen and M. Embree, Spectra and Pseudospectra: The Behavior of Nonnormal Matrices and Operators. Princeton University Press, 2005. 49
2005
-
[15]
Topological phases of non-hermitian systems,
Z. Gong, Y . Ashida, K. Kawabata, K. Takasan, S. Higashikawa, and M. Ueda, “Topological phases of non-hermitian systems,” Physical Review X, vol. 8, no. 3, p. 031079, 2018
2018
-
[16]
Non-hermitian physics,
Y . Ashida, Z. Gong, and M. Ueda, “Non-hermitian physics,” Advances in Physics, vol. 69, no. 3, pp. 249–435, 2020
2020
-
[17]
R. A. Horn and C. R. Johnson, Matrix Analysis. Cambridge University Press, 1990
1990
-
[18]
A simple proof of the Jordan decomposition theorem for matrices,
I. Gohberg and S. Goldberg, “A simple proof of the Jordan decomposition theorem for matrices,” The American Mathematical Monthly, vol. 103, no. 2, pp. 157–159, 1996
1996
-
[19]
Operator characterization via projectors and nilpotents,
S.-Y . Chang, “Operator characterization via projectors and nilpotents,” arXiv preprint arXiv:2411.11883, 2024
2024 arXiv
-
[20]
Jensen’s operator inequality,
F. Hansen and G. K. Pedersen, “Jensen’s operator inequality,” Bulletin of the London Mathematical Society, vol. 35, no. 4, pp. 553–564, 2003
2003
-
[21]
Matrix analysis: matrix monotone functions, matrix means, and majorization,
F. Hiai, “Matrix analysis: matrix monotone functions, matrix means, and majorization,” Interdisci- plinary Information Sciences, vol. 16, no. 2, pp. 139–248, 2010. 50
2010
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.