Pith. sign in

REVIEW 4 major objections 4 minor 15 references

Hermitian Maps: Approximations and Completely Positive Extensions

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

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

Pith's one-line read The paper claims that every Hermitian map has a completely positive extension on an ancilla whose minimal dimension equals the rank of its Choi matrix.

desk verdict The paper's main claim about minimal ancilla dimension is false, and the examples are unreliable; the correct parts are standard. read the letter →

arxiv 2506.09631 v1 pith:JOJMGOWD submitted 2025-06-11 math.FA math-phmath.MP

Mohsen Kian , Mohsen Rostamian Delavar This is my paper · ORCID
classification math.FAmath-phmath.MP MSC 47A6515A6915A60
keywords HermitianmapcompletelypositiveChoimatrixCP-distanceJordandecompositionHilbert-SchmidtnormCPextensionauxiliaryspace
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 studies Hermitian linear maps between matrix algebras, the class of operations that preserve the Hermitian property but need not be physically realizable as quantum channels. It claims three results: a sharp lower bound on the Hilbert–Schmidt norm of the negative component in any decomposition of a Hermitian map as a difference of two completely positive maps; a proof that the positive part of the Jordan decomposition gives the best completely positive approximation in that norm; and an explicit construction of a completely positive extension on an auxiliary space, with the claim that the minimal auxiliary dimension equals the rank of the Choi matrix. If these results hold, they would give a canonical quantitative handle on how far a Hermitian operation is from being physical, and a way to simulate it by a physical operation on a larger system.

What carries the argument

The load-bearing object is the Choi–Jamiołkowski matrix $C_\Phi=\sum_{i,j}E_{ij}\otimes\Phi(E_{ij})$, which identifies a linear map with a matrix in $M_m\otimes M_n$ and converts complete positivity into positive semidefiniteness. The proofs run through the Jordan decomposition $C_\Phi=C_\Phi^+-C_\Phi^-$ and the spectral representation $\Phi(X)=\sum_i\lambda_i A_i X A_i^*$, where $\operatorname{vec}(A_i)$ are the eigenvectors of $C_\Phi$. For the extension, the paper takes one auxiliary basis vector per nonzero eigenvalue, weights the Kraus operators by $|\lambda_i|$, and encodes eigenvalue signs in a Hermitian matrix $Q$; the partial-trace identities $\operatorname{tr}_m[(X^T\otimes I_n)\operatorname{vec}(A_i)\operatorname{vec}(A_i)^*]=A_iXA_i^*$ carry the reconstruction.

What would settle it

Take the 2×2 transposition map $\Phi(X)=X^T$. Its Choi matrix has eigenvalues $1,1,1,-1$ and rank 4, so the paper's theorem predicts minimal ancilla dimension 4. But writing $\Phi=\Phi_+-\Phi_-$ with $\Phi_+(X)=X^T+\operatorname{tr}(X)I$ and $\Phi_-(X)=\operatorname{tr}(X)I$, both completely positive, and defining $\Psi(Y)=\Phi_+(Y_{11})\otimes E_{11}+\Phi_-(Y_{22})\otimes E_{22}$ with $Q=\operatorname{diag}(1,-1)$, yields an extension with $k=2$. This shows the claimed equality $k_{\min}=\operatorname{rank}(C_\Phi)$ fails for this map.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the Choi matrix $C_\Phi$ of a Hermitian map $\Phi:M_m\to M_n$ completely dictates both its approximation and its extensibility. The paper proves that in any decomposition $\Phi=\Phi^{(1)}-\Phi^{(2)}$ into completely positive maps, $\|\Phi^{(2)}\|_{\mathrm{HS}}\ge \sqrt{k}\,d_{\mathrm{CP}}(\Phi)$, where $d_{\mathrm{CP}}(\Phi)=-\lambda_{\min}(C_\Phi)$ and $k$ is the multiplicity of that minimal eigenvalue, with equality attained by the Jordan decomposition. It further proves that the best completely positive approximation to $\Phi$ in the Hilbert–Schmidt norm is the map $\Phi_+$ whose Choi matrix is the positive part of $C_\Phi$. Finally, it constructs, from the spectral decomposition $C_\Phi=\sum_i\lambda_i u_i u_i^*$, a completely positive map $\Psi:M_m\otimes M_k\to M_n\otimes M_k$ and a Hermitian matrix $Q\in M_k$ satisfying $\Phi(X)=\operatorname{tr}_k[\Psi(X\otimes I_k)(I_n\otimes Q)]$, and claims the minimal such $k$ is $\operatorname{rank}(C_\Phi)$.

Load-bearing premise

The minimal-ancilla result rests on the unproved assertion, made at the end of Theorem 2.7's proof, that the spectral decomposition gives the smallest possible auxiliary space; nothing in the proof rules out a smaller dimension.

Editorial extensions

If this is right

  • Every Hermitian map $\Phi$ admits a completely positive extension of the form $\Phi(X)=\operatorname{tr}_k[\Psi(X\otimes I_k)(I_n\otimes Q)]$, with $\Psi$ built explicitly from the eigenvectors of $C_\Phi$.
  • The Hilbert–Schmidt distance from $\Phi$ to the nearest completely positive map is exactly $\|C_\Phi^-\|_{\mathrm{HS}}$, realized by the positive part of the Jordan decomposition.
  • Any CP decomposition $\Phi=\Phi^{(1)}-\Phi^{(2)}$ must pay at least $\sqrt{k}\,d_{\mathrm{CP}}(\Phi)$ in Hilbert–Schmidt norm for the negative component, so the CP-distance and the multiplicity of the minimal Choi eigenvalue together quantify the non-physical content.
  • When $C_\Phi$ is block diagonal, the auxiliary dimension can be reduced to the largest rank among the blocks, and equality with the total rank holds only for indecomposable Choi matrices.

Reading between the lines

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

  • Because any Hermitian map can be written as a difference of two completely positive maps, the block construction with $Q=\operatorname{diag}(1,-1)$ extends every such map using an ancilla of dimension 2; this indicates that the claimed equality between minimal ancilla size and Choi rank cannot hold in general.
  • The lower-bound theorem depends only on the negative eigenvalues of $C_\Phi$, so a natural refinement would replace the rank by the number of distinct sign patterns or the CP-rank of the positive and negative parts.
  • The explicit spectral construction remains a valid extension even when the ancilla is not minimal; it could serve as a starting point for practical simulation of Hermitian operations in quantum error mitigation, where the ancilla cost is the relevant resource.
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

4 major / 4 minor

Summary. The manuscript studies Hermitian linear maps on matrix algebras through their Choi matrices. It claims three main results: (i) Theorem 2.2 gives a lower bound of the form sqrt(k) d_CP(Phi) on the Hilbert-Schmidt norm of the negative component in any decomposition of a Hermitian map as a difference of two completely positive maps, together with a sharpness assertion; (ii) Theorem 2.5 asserts that the positive part of the Jordan decomposition is the best completely positive approximation in the Hilbert-Schmidt norm; and (iii) Theorem 2.7 asserts that the minimal ancilla dimension needed to extend a Hermitian map to a completely positive map through equation (2.5) equals the rank of the Choi matrix, with Theorem 2.9 refining the statement for block-diagonal Choi matrices. The paper also presents several worked examples.

Significance. The ancilla-dimension claim is the paper's headline contribution and is motivated by quantum broadcasting and dilation problems. The constructive part of Theorem 2.7, showing that an extension exists with an ancilla of dimension equal to the Choi rank, is essentially correct and could be useful. Theorem 2.5 is also essentially correct and gives a clean characterization of the best CP approximation in Hilbert-Schmidt norm. However, the minimality statement in Theorem 2.7 is not proved and is, in fact, false: every Hermitian map admits such an extension with a two-dimensional ancilla. Several examples also contain incorrect eigenvalue and Choi-matrix computations. Because the central claim is invalid, the paper cannot be accepted in its current form.

major comments (4)
  1. [Theorem 2.7, final paragraph] The proof asserts that the dimension k = r is minimal because it corresponds to the number of nonzero eigenvalues of C_Phi, but no lower bound on the ancilla dimension is proved, and the claim is false. For any Hermitian map Phi, write Phi = Phi_+ - Phi_- with Phi_+, Phi_- completely positive (for instance, via the Jordan decomposition of C_Phi). Set k = 2, Q = diag(1,-1), and define Psi : M_m tensor M_2 -> M_n tensor M_2 by Psi(Y) = Phi_+(Y_11) tensor E_11 + Phi_-(Y_22) tensor E_22, where Y_ii = (I_m tensor E_ii) Y (I_m tensor E_ii). This Psi is completely positive as a composition of compressions, CP maps, and ampliations. For Y = X tensor I_2, the two diagonal blocks are both X, so tr_2[Psi(X tensor I_2)(I_n tensor Q)] = Phi_+(X) - Phi_-(X) = Phi(X). Thus every Hermitian map has an extension of the form (2.5) with k = 2. For the transposition map on M_2, rank(C_Phi) = 4, so k_min = rank(C_Phi) is directly contradicted. The claimed equality k_min = rank(C_Phi) is therefore invalid.
  2. [Theorem 2.2, proof and sharpness statement] After the spectral decomposition, the proof states: 'Let lambda_min = -d_CP(Phi) be the only negative eigenvalue.' This is not true in general. If C_Phi has additional negative eigenvalues, then C_-Phi = d_CP(Phi) P_min + C_rest, so ||C_-Phi||_HS^2 = k d_CP(Phi)^2 + ||C_rest||_HS^2 > k d_CP(Phi)^2. By Lemma 2.1, any CP map Phi_2 in a difference decomposition Phi = Phi_1 - Phi_2 must satisfy C_Phi2 >= C_-Phi, hence ||Phi_2||_HS >= ||C_-Phi||_HS > sqrt(k) d_CP(Phi). Therefore the asserted sharpness, 'there exists a decomposition where equality is achieved,' is false whenever C_Phi has two distinct negative eigenvalues. A concrete example is C_Phi = diag(1,-1,-2) (with m = 1, n = 3), for which d_CP(Phi) = 2 and k = 1 but the negative part has norm sqrt(5) > 2.
  3. [Example 2.4] The Choi matrix displayed in Example 2.4 has eigenvalues 2, 2, 1, -1, not 2 (multiplicity 2) and -1 (multiplicity 2). Consequently the multiplicity of lambda_min is k = 1, not k = 2. Moreover, the displayed C_-Phi is not the Jordan negative part: the negative part corresponding to the eigenvalue -1 is the orthogonal projection (1/2)[[0,0,0,0],[0,1,-1,0],[0,-1,1,0],[0,0,0,0]], whose Hilbert-Schmidt norm is 1, not sqrt(2). The example therefore does not illustrate Theorem 2.2 as stated.
  4. [Example 2.10] The submap Phi_2(X_2) = X_2 - X_2^* has a Choi block, with the manuscript's ordering, equal to [[0,0,0,0],[0,1,-1,0],[0,-1,1,0],[0,0,0,0]], whose eigenvalues are 2, 0, 0, 0, not +/-1. The displayed skew-symmetric block [[0,0,1,0],[0,-1,0,0]] is not Hermitian and cannot be the Choi matrix of a Hermitian map. In addition, the Q computed later as diag(2,0,1,-1) does not recover Phi_2: for the block-2 Kraus terms, tr(e_1 e_1^* Q) = 2 and tr(e_2 e_2^* Q) = 0, so the block-2 contribution is 2 A_{2,1} X_2 A_{2,1}^*, which is not X_2 - X_2^*. The example therefore does not satisfy the extension identity (2.5).
minor comments (4)
  1. [Proof of Theorem 2.7] The sentence 'It is evident that Phi is completely positive' should refer to Psi, not Phi.
  2. [Keywords] The keyword 'Choi matrxi' is a typo for 'Choi matrix'.
  3. [Example 2.8] The phrase 'WE define' should be capitalised as 'We define'.
  4. [Theorem 2.9] The definition Q = sum_i V_i Q_i V_i^* requires a convention for the coordinates e_j when different blocks use the same basis vector; as written, the formula can produce conflicting diagonal entries, as illustrated by Example 2.10.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the main constructions are self-contained; the flawed minimality claim in Theorem 2.7 is an unsupported assertion, not a circular derivation.

full rationale

I walked the paper's derivation chain. Theorem 2.2 derives the lower bound on the negative CP component from Lemma 2.1 and the spectral decomposition of the Choi matrix; no fitted parameter is renamed as a prediction. Theorem 2.5 proves optimality of the positive Jordan component via a direct Hilbert-Schmidt norm computation and the positivity of traces of products of positive semidefinite matrices. Theorem 2.7 constructs a CP extension with auxiliary dimension r = rank(C_Phi) by spectral decomposition, the vec identity, and partial-trace calculations, none of which assumes the claimed minimality. The only potentially circular-looking element is the definition of d_CP, which is attributed to the first author's earlier work [9]; however, the definition is merely used as notation and the proofs do not invoke any theorem from [9] as a load-bearing premise. The final sentence of Theorem 2.7, 'The dimension k = r is minimal, as it corresponds to the number of nonzero eigenvalues in the spectral decomposition of C_Phi, ensuring the smallest possible auxiliary space,' is an unproved lower-bound claim, and it is in fact false (a two-dimensional ancilla construction extends every Hermitian map), but a false or unsupported assertion is not circular reasoning: it does not make the conclusion equivalent to an input by construction. The block-diagonal extension in Theorem 2.9 also follows by applying Theorem 2.7 blockwise, and its minimality statement inherits the same unsupported nature without adding a circular step. Since every proved equation in the paper is derived from standard spectral and partial-trace identities rather than from the conclusion being assumed, the paper contains no significant circularity.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No free parameters are fitted. The paper relies on standard matrix analysis; the only nonstandard premise is the asserted minimality of the Choi rank as an ancilla dimension, which is false.

assumptions (4)
  • standard math The Choi-Jamiolkowski isomorphism identifies Hermitian maps with Hermitian Choi matrices and CP maps with PSD Choi matrices, preserving the Hilbert-Schmidt norm.
    Used throughout; see Section 1 and the proofs of Theorems 2.2, 2.5, and 2.7.
  • standard math The identity (A⊗B)vec(C)=vec(BCA^T) and the partial trace identity tr_m[vec(B)vec(C)^*]=BC^* hold.
    Invoked in the proof of Theorem 2.7, after equation (2.6).
  • standard math Lemma 2.1, the minimality of A_- among PSD B with A+B≥0, is proven in the paper and is correct.
    Central to Theorem 2.2.
  • ad hoc to paper The unproved assertion that the rank of C_Φ is a lower bound on the ancilla dimension in Theorem 2.7.
    This premise is not proven and is false: k=2 suffices for all Hermitian maps.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hermitian Maps: Approximations and Completely Positive Extensions." pith.science (2026). https://pith.science/paper/JOJMGOWD

@misc{pith2026250609631,
  author       = {Pith},
  title        = {Pith review of: Hermitian Maps: Approximations and Completely Positive Extensions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JOJMGOWD}},
  note         = {Machine review of arXiv:2506.09631}
}
read the original abstract

This study investigates Hermitian linear maps, focusing on their decomposition into completely positive (CP) maps and their extensions to CP maps using auxiliary spaces. We derive a precise lower bound on the Hilbert-Schmidt norm of the negative component in any CP decomposition, proving its attainability through the Jordan decomposition. Additionally, we demonstrate that the positive part of this decomposition provides the optimal CP approximation in the Hilbert-Schmidt norm. We also determine the minimal dimension of an auxiliary space required to extend a Hermitian map to a CP map, with explicit constructions provided. Practical examples illustrate the application of our results.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 14 canonical work pages

  1. [1]

    Ando,Positive map as difference of two completely positive or super-positive maps, Adv

    T. Ando,Positive map as difference of two completely positive or super-positive maps, Adv. Opr. Theory,3(1) (2018), 53–60

    work page 2018

  2. [2]

    Bengtsson, and K

    I. Bengtsson, and K. ˙Zyczkowski,Geometry of Quantum States: An introduction to quantum entanglement, Cambridge University Press, 2006

    work page 2006

  3. [3]

    Buscemi, M

    F. Buscemi, M. Dall’Arno, M. Ozawa, and V. Vedral,Direct observation of any two-point quantum correlation function, arXiv:1312.4240v1, (2013)

    arXiv 2013

  4. [4]

    M. -D. Choi,Completely positive linear maps on complex matricesLinear Algebra Appl.,10(3) (1975), 285–290

    work page 1975

  5. [5]

    Dadkhah, M

    A. Dadkhah, M. Kian, and M. S. Moslehian,Decomposition of tracial positive maps and applications in quantum information, Anal. Math. Phys.,14(2024), 48

    work page 2024

  6. [6]

    K. H. Han, S.-H. Kye, and E. Størmer,Infinite dimensional analogues of Choi matrices, J. Funct. Anal,287 (8) (2024), 110557

    work page 2024

  7. [7]

    G. Homa, A. Ortega, M. Koniorczyk,Choi Representation of Completely Positive Maps in Brief, Z. Naturforsch., 79(12) (2024), 1123–1133

    work page 2024

  8. [8]

    R. A. Horn, and C. R. Johnson,Topics in matrix analysis, Cambridge University Press, 1991

    work page 1991

Show all 15 references
  1. [9]

    Kian,Deviation from complete positivity: Structural insights and quantum information applications, arXiv:2506.03773, (2025)

    M. Kian,Deviation from complete positivity: Structural insights and quantum information applications, arXiv:2506.03773, (2025)

    2025

  2. [10]

    W. A. Majewski,On Positive Maps in Quantum Information, Russian J. Math. Phys.21(2014), 362–372

    2014

  3. [11]

    A. J. Parzygnat, J. Fullwood, F. Buscemi, and G. Chiribella,Virtual quantum broadcasting, Phys. Rev. Lett. 132(2024), 110203

    2024

  4. [12]

    Piotrowski, L

    J. Piotrowski, L. Skowronek,New Tools for Investigating Positive Maps in Matrix Algebras, Linear Algebra Appl.,662(2023), 39–48

    2023

  5. [13]

    Siudzi´ nska, S

    K. Siudzi´ nska, S. Chakraborty, and D. Chru´ sci´ nski,Interpolating between Positive and Completely Positive Maps: A New Hierarchy of Entangled States, Linear Algebra Appl.,662(2018), 39–48

    2018

  6. [14]

    Størmer,Positive linear maps of operator algebras, Springer, 2013

    E. Størmer,Positive linear maps of operator algebras, Springer, 2013

    2013

  7. [15]

    Watrous,The theory of quantum information, Cambridge University Press, 2018

    J. Watrous,The theory of quantum information, Cambridge University Press, 2018. Mohsen Kian: Department of Mathematics, University of Bojnord, P. O. Box 1339, Bojnord 94531, Iran Email address:kian@ub.ac.ir Mohsen Rostamian Delavar: Department of Mathematics, University of Boj...

    2018

Pith tools

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