Pith. sign in

REVIEW 3 major objections 3 minor 31 references

Noncommutative partial convexity via $\Gamma$-convexity

T0 review · 3 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Γ-convex sets are separated by monic Γ-pencils whenever the matrix convex hull of their Γ-image is closed, and unconditionally for bounded operator sets.

desk verdict A genuinely new framework for noncommutative partial convexity with clean separation theorems; the matrix result is conditional in an important way, but the operator result and concrete y2 constructions make it a solid paper. read the letter →

arxiv 1908.05949 v1 pith:YL6T7T6I submitted 2019-08-16 math.FA

Michael Jury , Igor Klep , Mark E. Mancuso , Scott McCullough , James Eldred Pascoe This is my paper · ORCID
classification math.FA MSC 46N1047L0752A30
keywords Γ-convexitypartialconvexitybiconvexitybilinearmatrixinequalitiesnoncommutativepolynomialsfreesemialgebraicsetsEffros-Winklertheorem
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 establishes a Hahn–Banach separation theorem for $\Gamma$-convex sets, a family of noncommutative sets defined by stability under compressions by isometries that intertwine a fixed tuple of symmetric polynomials $\Gamma=(\gamma_1,\dots,\gamma_r)$. The certificate is a monic $\Gamma$-pencil, a matrix-valued expression $I+\sum_j A_j\gamma_j(x)$ that is positive semidefinite on the whole set but fails at the point being separated. In the finite-matrix setting the theorem requires the matrix convex hull of $\Gamma(K)$ to be closed; in the operator setting boundedness plus strong-operator-topology closure makes that hypothesis automatic. This gives a common dual description for spectrahedra, biconvex sets, and domains defined by bilinear matrix inequalities.

What carries the argument

The engine is the $\Gamma$-pair condition $V^*\Gamma(X)V=\Gamma(V^*XV)$, which specifies exactly which compressions a $\Gamma$-convex set must survive and makes the $\Gamma$-convex hull commute with the map $\Gamma$: $X\in \Gamma\text{-co}(K)$ if and only if $\Gamma(X)\in \operatorname{matco}(\Gamma(K))$. This identity transfers separation of $Y$ from $K$ to separation of $\Gamma(Y)$ from the matrix convex hull of $\Gamma(K)$, where the Effros–Winkler theorem produces monic linear pencils; composing those with $\Gamma$ yields monic $\Gamma$-pencils. In the operator setting, a compactness principle for bounded operator tuples shows that $\operatorname{opco}(\Gamma(K))$ is SOT-closed whenever $K$ is bounded and SOT-closed, supplying the missing closure. The final section adds a slice characterization: for $\Gamma=\{x,y,y^2\}$, a free set is $\Gamma$-convex exactly when each slice $\{X:(X,Y)\in K(n)\}$ is convex in the ordinary sense, and explicit pencils show the TV sets $\{1-x^2-y^{2d}\succeq 0\}$ are positivity sets of single monic $y^2$-pencils.

What would settle it

In the operator setting, the decisive test is a bounded, SOT-closed, operator $\Gamma$-convex set $K$ containing $0$ with $\Gamma(0)=0$ and $0\in\operatorname{opco}(\Gamma(K))$, together with some $Y\notin K$; if every finite monic $\Gamma$-pencil that is positive semidefinite on $K$ is also positive semidefinite at $Y$, then Theorem 3.8 is refuted. In the matrix setting, the analogous test is a $\Gamma$-convex $K$ with $\operatorname{matco}(\Gamma(K))$ closed and a point $Y\notin K$ that no monic $\Gamma$-pencil of matching size separates.

Watch

Extended reading notes

Core claim

The paper claims that every sufficiently closed $\Gamma$-convex set containing the origin is described by monic $\Gamma$-pencils. Theorem 2.4 states that if $K$ is $\Gamma$-convex, $0\in K$, $\Gamma(0)=0$, and $\operatorname{matco}(\Gamma(K))$ is closed, then for each $Y\notin K$ of size $\ell$ there is a monic $\Gamma$-pencil $L$ of size $\ell$ with $L(K)\succ 0$ and $L(Y)\not\succeq 0$. Theorem 3.8 removes the separate closedness hypothesis in the operator setting: for bounded, SOT-closed operator $\Gamma$-convex sets, every outside point is separated by a finite monic $\Gamma$-pencil, so $K$ is the intersection of the positivity sets of the monic $\Gamma$-pencils containing it. Corollary 1.6 then yields a single monic operator $\Gamma$-pencil $L$ with $K=\{X:L(X)\succeq 0\}$ when $0$ lies in the interior of the convex hull of $\Gamma(K)$.

Load-bearing premise

The finite-matrix theorem rests on the assumption that the collection of all matrix compressions of the image $\Gamma(K)$ is closed—which the paper notes is not automatic—while the operator version obtains that closure from boundedness plus strong-operator-topology closure, leaving unbounded sets outside the theorem.

Editorial extensions

If this is right

  • In the operator setting, every bounded, SOT-closed $\Gamma$-convex set containing $0$ with $\Gamma(0)=0$ is exactly the intersection of the positivity sets of the monic $\Gamma$-pencils that contain it, with no extra closedness hypothesis on the hull.
  • When $0$ lies in the interior of the convex hull of $\Gamma(K)$—automatically the case when the real span of the coordinates of $\Gamma$ contains no polynomial that is positive on $K$, e.g. for multilinear $\Gamma$—a single monic operator $\Gamma$-pencil alone cuts out $K$.
  • For any regular free polynomial $p$ whose positivity set is operator $\Gamma$-convex, every outside matrix point is separated by a monic $\Gamma$-pencil of the same size, so $D_p$ is an intersection of $\Gamma$-pencil positivity sets; in particular this applies to the TV sets $1-x^2-y^{2d}$.
  • For $\Gamma=\{x,y,y^2\}$, $\Gamma$-convexity is equivalent to ordinary convexity of each $x$-slice at fixed $Y$, giving a free analog of partial convexity with explicit pencils for the TV screen examples.

Reading between the lines

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

  • A natural next step, not pursued in the paper, is to track how the minimal size of a separating $\Gamma$-pencil grows with the size of the outlier; the matrix theorem gives size $\ell$, while the operator theorem only promises some finite size, and the gap may encode geometric complexity of $\Gamma$.
  • For bilinear matrix inequality feasibility sets, the $xy$-convex results imply that a point outside the feasibility region admits a symbolic $xy$-pencil certificate; this could turn BMI infeasibility into a finite algebraic witness in control applications, though the paper does not address computational cost.
  • The $y^2$ slice characterization suggests a broader dictionary: for other $\Gamma$, $\Gamma$-convexity may correspond to convexity along noncommutative curved directions determined by $\Gamma$, and testing this on $\Gamma=\{x,y,xy+yx\}$ would give a free analog of biconvexity with bilinear separating pencils.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper introduces Γ-convexity for free sets: given a tuple Γ=(γ_1,...,γ_r) of symmetric free polynomials that includes the coordinate variables, a free set K⊆S(C)^g is Γ-convex if it is closed under V^*XV whenever V is an isometry satisfying V^*Γ(X)V=Γ(V^*XV). The main finite-dimensional result, Theorem 2.4, shows that if Γ(Y) lies outside the matrix convex hull of Γ(K), then a monic Γ-pencil separates Y from K, and consequently, when matco(Γ(K)) is closed, every Y∉K is separated by a monic Γ-pencil. Section 3 extends the theory to bounded, SOT-closed operator Γ-convex sets, proving an operator-level Effros–Winkler theorem (Theorem 3.8) and representation of K as an intersection of positivity sets of monic Γ-pencils without a separate closedness hypothesis for opco(Γ(K)). Section 4 characterizes y2-convex sets as free sets that are convex in x and gives explicit monic y2-pencils for the sets TV_d.

Significance. If the results hold, the paper gives a substantial extension of Effros–Winkler Hahn–Banach separation to constrained noncommutative polynomial inequalities, with natural applications to partial convexity and bilinear matrix inequalities. The proofs are mostly rigorous and transparent: Theorem 2.4 reduces cleanly to the classical Effros–Winkler theorem; Theorem 3.3 uses a compactness principle for bounded operator tuples (Lemma 3.2) to obtain SOT-closedness of opco(Γ(K)); and Proposition 4.2 provides explicit, checkable monic y2-pencils. The paper is also honest about the hypotheses in the operator setting. The main caveats are a notational inconsistency in the definition of matco, a local but real proof defect in Theorem 3.5, and the fact that the finite-dimensional separation theorem is conditional on a closedness hypothesis that fails for natural unbounded Γ-convex sets; these issues affect the presentation and scope of the main claims more than the validity of the individual theorems.

major comments (3)
  1. [Section 2.1, Proposition 2.2, Theorem 2.4] The notation matco(S) is defined in §2.1 as the levelwise closed matrix convex hull, but Proposition 2.2 and Theorem 2.4 use matco(Γ(K)) as the ordinary (not necessarily closed) matrix convex hull. As written, Proposition 2.2 is false if matco denotes the closure: for the unbounded y2-convex set K(n)={(X,Y): Y≽0, X≽Y, ker(Y)⊆ker(X)} with Γ=(x,y,y^2), the point (1,0)∉K has Γ(1,0) in the closure of matco(Γ(K)) but not in the ordinary matco(Γ(K)). Please denote the ordinary hull by matco and its closure by \overline{matco} throughout, and restate Theorem 2.4 accordingly.
  2. [Theorem 3.5, proof] In the proof of Theorem 3.5, the equality M(Y)=M(Y')⊕I is not valid for a general Y, because Y need not be block diagonal with respect to the decomposition H=V(C^N)⊕(V(C^N))^\perp. The desired conclusion still follows by compression: M(Y')=(I⊗V)^*M(Y)(I⊗V); if M(Y) were positive semidefinite, then M(Y') would be, contradicting Proposition 3.7(d). Please replace the displayed equality with this compression argument.
  3. [Theorem 1.4, Remark 2.5, abstract] The closedness hypothesis in Theorem 2.4 is not a harmless technicality. For Γ=(x,y,y^2) and K(n)={(X,Y): Y≽0, X≽Y, ker(Y)⊆ker(X)}, K is a free y2-convex set containing 0 with Γ(0)=0; however matco(Γ(K)) is not closed, and no monic y2-pencil separates (1,0) from K, since L(1,0)=I+A would be a limit of L(1,y)≽0 as y→0+. This example should be discussed, and the abstract's unqualified claim that Γ-convex sets "are delineated by linear pencils" should be qualified to the cases where the stated hypotheses hold, or to the operator theorem.
minor comments (3)
  1. [Theorem 2.4, proof] In the perturbed pencil after the application of Theorem 1.3, the coefficient should be (1−t)A_j, not tA_j: from L'(z)=I+ΣA_jz_j one gets tI+(1−t)L'(z)=I+Σ(1−t)A_jz_j.
  2. [Section 2.1] The sentence defining matco(S) as the closed matrix convex hull conflicts with the later use of matco for the unclosed hull; in addition to fixing the notation, please check that all occurrences in Propositions 2.2 and 2.3 and Theorem 2.4 use the intended meaning.
  3. [References] Lemma 3.2 is cited from [Man+] as appearing "to appear"; if a published version now exists, please update the citation with full publication data.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central separation results are reductions to the external Effros-Winkler theorem, and the sole self-cited lemma is an independent compactness black box.

full rationale

The derivation chain is not circular. Theorem 2.4 proves the finite-dimensional Gamma-Hahn-Banach separation by reducing to the classical Effros-Winkler theorem (Theorem 1.3), an external benchmark: 'Since Gamma(Y) / in matco(Gamma(K))(ell) and matco(Gamma(K)) is a closed matrix convex subset of S(C)^r containing 0, Theorem 1.3 implies there is a monic linear pencil... Thus L' = M composed with Gamma is a monic Gamma-pencil.' The characterization in Proposition 2.2, Gamma-co(K) = Gamma^{-1}(matco(Gamma(K))), is a direct consequence of the definition of a Gamma-pair, not a hidden assumption of the target. In the operator setting, Theorem 3.8 uses Theorem 3.3, whose SOT-closedness argument invokes [Man+, Lemma 4.5]; although [Man+] is by a co-author, the lemma is a parameter-free compactness principle about bounded operator sequences that does not contain the separation conclusion, so it qualifies as independent support rather than a self-citation carrying the circular load. The boundedness and closedness hypotheses are explicit in Theorems 2.4 and 3.8, and Remark 2.5 openly flags the difficulty with non-closed matco(Gamma(K)); the paper nowhere claims that hypothesis is automatic. No fitted parameter is renamed as a prediction, no uniqueness theorem from prior work is invoked to forbid alternatives, and no known result is merely relabeled.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper rests on two external mathematical inputs: the classical Effros-Winkler theorem for matrix convex sets, and a compactness lemma (Lemma 3.2) from the companion preprint [Man+] by one of the authors. It also assumes structurally that the defining tuple Γ contains the original coordinate polynomials x_1,...,x_g, which makes Γ-pairs encode compressions of X itself. There are no free parameters fitted to data and no invented empirical entities; the new objects (Γ-convex sets, Γ-pencils) are definitions that constitute the paper's contribution.

assumptions (3)
  • standard math Effros-Winkler matricial Hahn-Banach separation theorem (Theorem 1.3) for matrix convex sets.
    Used as the engine for both Theorem 2.4 and Theorem 3.5; cited to [EW97] via [HM12, HKM17].
  • domain assumption Lemma 3.2 (Mancuso): bounded sequences of operator tuples admit subsequences that become SOT-convergent after conjugation by unitaries.
    Pulled from the companion preprint [Man+] and used to prove SOT-closedness of opco(Γ(K)) in Theorem 3.3. Not re-proved in this paper.
  • domain assumption Γ contains the coordinate polynomials: γ_j = x_j for 1 ≤ j ≤ g (Section 1.2).
    This is the structural hypothesis that makes Γ-pairs encode compressions of X itself; it is essential in Prop 2.2 and hence in all separation results. The theory does not cover Γ tuples that omit the original variables.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Noncommutative partial convexity via $\Gamma$-convexity." pith.science (2026). https://pith.science/paper/YL6T7T6I

@misc{pith2026190805949,
  author       = {Pith},
  title        = {Pith review of: Noncommutative partial convexity via $\Gamma$-convexity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YL6T7T6I}},
  note         = {Machine review of arXiv:1908.05949}
}
abstract

Motivated by classical notions of partial convexity, biconvexity, and bilinear matrix inequalities, we investigate the theory of free sets that are defined by (low degree) noncommutative matrix polynomials with constrained terms. Given a tuple of symmetric polynomials $\Gamma$, a free set is called $\Gamma$-convex if it closed under isometric conjugation by isometries intertwining $\Gamma$. We establish an Effros-Winkler Hahn-Banach separation theorem for $\Gamma$-convex sets; they are delineated by linear pencils in the coordinates of $\Gamma$ and the variables $x$.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 27 canonical work pages

  1. [1]

    Agler, J.E

    J. Agler, J.E. McCarthy: Pick interpolation for free holomorphic functions. Amer. J. Math. 137 (2015), no. 6, 1685--1701

    work page 2015

  2. [2]

    VanAntwerp, R.D

    J.G. VanAntwerp, R.D. Braatz: A tutorial on linear and bilinear matrix inequalities, Journal of Process Control 10 (2000) 363--385

    work page 2000

  3. [3]

    Balasubramanian, S

    S. Balasubramanian, S. McCullough: Quasi-convex free polynomials, Proc. Amer. Math. Soc. 142 (2014) 2581--2591

    work page 2014

  4. [4]

    J.A. Ball, G. Marx, V. Vinnikov: Noncommutative reproducing kernel Hilbert spaces. J. Funct. Anal. 271 (2016) 1844--1920

    work page 2016

  5. [5]

    Barvinok: A course in convexity, Graduate Studies in Mathematics, 54

    A. Barvinok: A course in convexity, Graduate Studies in Mathematics, 54. American Mathematical Society, Providence, RI, 2002

    work page 2002

  6. [6]

    Blekherman, P.A

    G. Blekherman, P.A. Parrilo, R.R. Thomas (editors): Semidefinite optimization and convex algebraic geometry, MOS-SIAM Series on Optimization, 13. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2013

    work page 2013

  7. [7]

    S. Boyd, L. El Ghaoui, E. Feron, V. Balakrishnan: Linear matrix inequalities in system and control theory, SIAM Studies in Applied Mathematics, 15. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 1994

    work page 1994

  8. [8]

    Choi, E.G

    M.D. Choi, E.G. Effros: Injectivity and operator spaces, J. Funct. Anal. 24 (1977) 156--209

    work page 1977

Show all 31 references
  1. [9]

    Davidson, A

    K.R. Davidson, A. Dor-On, O.M. Shalit, B. Solel: Dilations, inclusions of matrix convex sets, and completely positive maps, Int. Math. Res. Not. IMRN (2017) 4069--4130

    2017

  2. [10]

    Davidson, M

    K. Davidson, M. Kennedy: Noncommutative Choquet theory, preprint https://arxiv.org/abs/1905.08436

    1905 arXiv

  3. [11]

    Dym, J.W

    H. Dym, J.W. Helton, S. McCullough: Non-commutative polynomials with convex level slices, Indiana Univ. Math. J. 66 (2017) 2071-–2135

    2017

  4. [12]

    Effros, S

    E.G. Effros, S. Winkler: Matrix convexity: operator analogues of the bipolar and Hahn-Banach theorems, J. Funct. Anal. 144 (1997) 117–-152

    1997

  5. [13]

    Evert: Matrix convex sets without absolute extreme points, Linear Algebra Appl

    E. Evert: Matrix convex sets without absolute extreme points, Linear Algebra Appl. 537 (2018) 287--301

    2018

  6. [14]

    Evert, J.W

    E. Evert, J.W. Helton: Arveson extreme points span free spectrahedra , preprint https://arxiv.org/abs/1806.09053

    arXiv

  7. [15]

    Fuller, M

    A.H. Fuller, M. Hartz, M. Lupini: Boundary representations of operator spaces and compact rectangular matrix convex sets, J. Operator Theory 79 (2018) 139--172

    2018

  8. [16]

    Hay, J.W

    D.M. Hay, J.W. Helton, A. Lim, S. McCullough: Non-commutative partial matrix convexity, Indiana Univ. Math. J. 57 (2008) 2815–2842

    2008

  9. [17]

    Helton, I

    J.W. Helton, I. Klep, S. McCullough: The matricial relaxation of a linear matrix inequality, Math. Program. Ser. A 138 (2013) 401--445

    2013

  10. [18]

    Helton, I

    J.W. Helton, I. Klep, S. McCullough: The tracial Hahn-Banach theorem, polar duals, matrix convex sets, and projections of free spectrahedra, J. Eur. Math. Soc. (JEMS) 19 (2017) 1845–-1897

    2017

  11. [19]

    Hartz, M

    M. Hartz, M. Lupini, The classification problem for operator algebraic varieties and their multiplier algebras. Trans. Amer. Math. Soc. 370 (2018), no. 3, 2161--2180

    2018

  12. [20]

    Helton, S

    J.W. Helton, S. McCullough: Every convex free basic semi-algebraic set has an LMI representation, Ann. of Math. (2) 176 (2012) 979--1013

    2012

  13. [21]

    M.T. Jury, I. Klep, M.E. Mancuso, S. McCullough, J.E. Pascoe: -Convex Polynomials , preprint

  14. [22]

    Kaliuzhnyi-Verbovetskyi, V

    D.S. Kaliuzhnyi-Verbovetskyi, V. Vinnikov: Foundations of free noncommutative function theory, Mathematical Surveys and Monographs, 199. American Mathematical Society, Providence, RI, 2014

    2014

  15. [23]

    Kanev, C

    S. Kanev, C. Scherer, M. Verhaegen, B. De Schutter: Robust output-feedback controller design via local BMI optimization, Automatica J. IFAC 40 (2004) 1115--1127

    2004

  16. [24]

    Mancuso: Inverse and Implicit Function Theorems for Noncommutative Functions on Operator Domains, to appear in J

    M.E. Mancuso: Inverse and Implicit Function Theorems for Noncommutative Functions on Operator Domains, to appear in J. Operator Theory

  17. [25]

    Passer, O.M

    B. Passer, O.M. Shalit: Compressions of compact tuples, Linear Algebra Appl. 564 (2019) 264--283

    2019

  18. [26]

    Passer, O.M

    B. Passer, O.M. Shalit, B. Solel: Minimal and maximal matrix convex sets, J. Funct. Anal. 274 (2018) 3197--3253

    2018

  19. [27]

    Paulsen: Completely bounded maps and operator algebras, Cambridge Studies in Advanced Mathematics, 78

    V. Paulsen: Completely bounded maps and operator algebras, Cambridge Studies in Advanced Mathematics, 78. Cambridge University Press, Cambridge, 2002

    2002

  20. [28]

    M. Popa, V. Vinnikov: H^2 spaces of non-commutative functions. Complex Anal. Oper. Theory 12 (2018) 945--967

    2018

  21. [29]

    Popescu: Invariant subspaces and operator model theory on noncommutative varieties, Math

    G. Popescu: Invariant subspaces and operator model theory on noncommutative varieties, Math. Ann. 372 (2018) 611--650

    2018

  22. [30]

    Salomon, O.M

    G. Salomon, O.M. Shalit, E. Shamovich: Algebras of bounded noncommutative analytic functions on subvarieties of the noncommutative unit ball, Trans. Amer. Math. Soc. 370 (2018) 8639--8690

    2018

  23. [31]

    Zalar: Operator Positivstellensätze for noncommutative polynomials positive on matrix convex sets, J

    A. Zalar: Operator Positivstellensätze for noncommutative polynomials positive on matrix convex sets, J. Math. Anal. Appl. 445 (2017) 32--80

    2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.