Pith. sign in

REVIEW 1 major objections 3 minor 19 references

$\mathbb{Z}^2$ is flexibly stable in the operator norm

T0 review · 1 major / 3 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read This paper proves that the group Z² is flexibly stable in the operator norm: every pair of almost-commuting unitary matrices is asymptotically close, after a negligible enlargement of the matrices, to a pair that exactly commutes.

desk verdict Clean, novel Z² flexible stability proof; Section 3 has a load-bearing unsupported operator-norm upgrade of Eckhardt's HS construction. read the letter →

arxiv 2607.17578 v1 pith:MR7C7K6H submitted 2026-07-20 math.OA math.GR

classification math.OAmath.GR MSC 46L0547A5520F69
keywords flexiblestabilityoperatornormalmostcommutingunitarieswindingnumbergroupC*-algebrasdimensiongrowth
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

A classical 1983 counterexample showed that Z² is not stable in the operator norm: certain pairs of unitary matrices have commutators tending to zero, yet stay uniformly far from any commuting pair of the same size. This paper shows the obstruction vanishes once the matrices are allowed to grow by an asymptotically negligible amount. For every almost-commuting pair in dimension d_n, there is a commuting pair in dimension d_n + o(d_n) whose compression back to d_n almost equals the original pair. Since Z² is not stable, this yields the first separation between flexible stability and stability in any metric context. The same circle of ideas also produces groups that are very-flexibly stable but not flexibly stable, making both inclusions strict.

What carries the argument

The winding-number invariant w(u,v) of a pair of unitaries: the winding number around the origin of the determinant path t ↦ det((1−t)uv + tvu), which for almost-commuting pairs equals (1/2πi)Tr(log(vuv*u*)). It is additive under direct sums and equals ±1 for the cyclic shift/phase pair; the proof cancels the winding number of an arbitrary almost-commuting pair by adding such blocks, then applies the completeness theorem for the winding-number obstruction in the operator norm (zero winding number implies approximability by commuting pairs).

What would settle it

Take the standard shift/phase pair (S_n, Ω_n) with commutator norm tending to zero and winding number −1, and compute the minimal excess dimension k_n − n needed for a commuting pair to approximate it in operator norm; if some c > 0 with k_n − n ≥ c n exists along a subsequence, the paper's claim that an o(d_n) enlargement suffices is false.

Watch

Extended reading notes

Core claim

The central discovery is that the winding-number invariant of a pair of almost-commuting unitaries — the winding number of the determinant curve traced by the line segment from uv to vu — is an integer 'charge' that can be killed by a small direct sum. By appending |w(u_n,v_n)| copies of the standard shift/phase pair, which carries winding number −1 or +1, the total winding number becomes zero while the dimension increases by only o(d_n). A cited theorem then guarantees that zero winding number suffices for approximation by commuting unitaries in the enlarged dimension. Thus the only obstruction to stability is the winding number, and it is removable at negligible cost.

Load-bearing premise

The entire construction leans on the cited theorem that pairs of almost-commuting unitaries with zero winding number can be approximated by commuting unitaries in the operator norm; if that completeness statement fails, the flexible-stability result collapses.

Editorial extensions

If this is right

  • Flexible stability and stability are not equivalent in the operator norm; Z² is the first concrete witness.
  • For the group C*-algebra C(T²) (the torus), the flexible-stability constant is D = 1: an asymptotically negligible enlargement always suffices.
  • The construction gives a quantitative bound: the extra dimension is |w(u_n,v_n)| m_n with m_n → ∞ chosen slowly, so the correction cost scales with the winding number.
  • The hierarchy stability ⊂ flexible stability ⊂ very-flexible stability is strict in the operator norm, since the paper's second theorem provides finitely generated groups that are very-flexibly stable but not flexibly stable.
  • Because Z² is amenable and abelian, the known equivalence between flexible and ordinary stability in the normalized Hilbert–Schmidt norm does not transfer to the operator norm.

Reading between the lines

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

  • The same cancellation trick is likely to extend to some higher-rank abelian groups; the paper explicitly leaves Z³ open, and a full treatment would require higher-dimensional analogues of the winding number.
  • The explicit dependence of the added dimension on the total winding number suggests a quantitative 'stability cost' that may be compared with K-theoretic invariants or almost flat K-theory.
  • If flexible stability is robust enough, the groups eG_p constructed here could serve as testbeds for whether operator-norm flexible stability is characterized by trace approximations in a way analogous to the Hilbert–Schmidt case.
  • A natural experiment is to compute the constant D(C*(Z³)); whether it equals 1, exceeds 1, or is infinite would calibrate how much the Z² result generalizes.
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

1 major / 3 minor

Summary. The paper proves that Z^2 is flexibly stable in the operator norm: every sequence of pairs of unitaries whose commutator tends to zero in the operator norm can be approximated, after passing to a space of dimension d_n + o(d_n) and compressing, by a pair of commuting unitaries. The proof uses the winding-number invariant w(u,v), a cancellation of this invariant by adding suitable Voiculescu-type blocks of dimension m_n, and the Gong--Lin/Eilers--Loring--Pedersen theorem. The paper also claims a second main result, Theorem C / Corollary 3.4: certain finitely generated amenable groups are very-flexibly stable but not flexibly stable in the operator norm, as an operator-norm analogue of a result of Eckhardt for the normalized Hilbert--Schmidt norm.

Significance. Theorem A is a clean and substantial result. It shows that the notion of flexible stability is genuinely weaker than ordinary stability in the operator norm, and it gives the first separation between these two notions in any metric context. The proof is elegant: the dimension increment |w_n|m_n/d_n tends to 0, the winding number is explicitly cancelled, and the external approximation theorem is applied correctly. If the second result were fully justified, it would strengthen Eckhardt's Hilbert--Schmidt separation to the operator norm and give the strictness of both implications stability --> flexible stability --> very-flexible stability. However, the current manuscript does not supply the operator-norm asymptotic representations needed for Theorem 3.1, so Theorem C is not yet established.

major comments (1)
  1. [3, proof of Theorem 3.1] The proof relies on the assertion that 'the proof of [8, Theorem 2.14] provides maps phi_n: Gamma -> U_{d_n}' satisfying operator-norm asymptotic multiplicativity and pointwise trace convergence. This is the only source of the contradiction. Eckhardt's stated theorem concerns the normalized Hilbert--Schmidt norm; Hilbert--Schmidt-small errors need not be small in the operator norm (an error can be supported on a subspace of dimension o(d_n) with entries of size 1). No construction or operator-norm estimate is reproduced. Since the desired conclusion is non-flexible stability in the operator norm, this gap is load-bearing. Proposition 3.2 and Lemma 3.3 are correct conditional on the existence of such maps; the missing ingredient is the existence proof for phi_n.
minor comments (3)
  1. [2.4, definition of D(A)] The symbol pi is used both for the quotient map and for unitary representations; this is potentially confusing and should be renamed.
  2. [3, Corollary 3.4] The very-flexible stability part of Corollary 3.4 relies on the external preprint [13]. Since the separation result depends on this, the final version should state the precise theorem from [13] or indicate its status.
  3. [3, Theorem 3.1] The sentence 'For completeness, we give the short argument' is inaccurate because the construction of the asymptotic representations is omitted. The paragraph should either include the construction or refer explicitly to the part of Eckhardt's proof that yields operator-norm estimates.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: Theorem A is self-contained given external inputs; Theorem C has an external-evidence gap, not a circular step.

full rationale

The central derivation for Z^2 is not circular. Theorem 2.3 takes arbitrary almost-commuting unitaries u_n,v_n, computes the winding invariant w(u_n,v_n), shows w_n/d_n -> 0 (Lemma 2.1), and then cancels the winding obstruction by adjoining copies of Voiculescu's matrices (S_m, Omega_m) with known winding number +/-1. The dimension cost is |w_n| m_n = o(d_n), so k_n/d_n -> 1. The resulting pairs X_n,Y_n have zero winding and small commutator, so the Gong-Lin / Eilers-Loring-Pedersen theorem (Theorem 2.2) applies verbatim. This is a genuine reduction to an external theorem, not an import of the conclusion: the paper never assumes that (u_n,v_n) are close to commuting; it constructs corrected pairs in a slightly larger dimension and then invokes the external approximation theorem. Voiculescu's theorem is used only to establish non-stability, and the flexible-stability definition from Becker-Lubotzky [1] is used as terminology, not as a load-bearing premise. The second main result (Theorem 3.1 / Corollary 3.4) rests on the assertion that 'the proof of [8, Theorem 2.14] provides maps phi_n' satisfying an operator-norm asymptotic representation and trace convergence. This is an external citation and not a fitted parameter or a renamed version of the paper's own conclusion, so it is not circular. There is a legitimate technical concern that Eckhardt's stated theorem is for the normalized Hilbert-Schmidt norm and the operator-norm upgrade is asserted rather than proved; that is a correctness/evidentiary gap, not a circularity. No step in the paper reduces by construction to its own inputs.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

No numbers are fitted to data. The central construction depends on the winding-number invariant and on external theorems (Gong–Lin/ELP, Eckhardt's maps, Fournier-Facio–Willett). The m_n sequence is an auxiliary existential choice, not a fitted parameter. No new physical or mathematical entities are postulated.

free parameters (1)
  • m_n (auxiliary block dimension sequence)
    Chosen with m_n → ∞ and |w_n| m_n / d_n → 0; existence follows from Lemma 2.1. This is a construction choice, not a fitted constant, and any sufficiently slow growth works.
assumptions (5)
  • domain assumption Gong–Lin / Eilers–Loring–Pedersen theorem (Theorem 2.2): if u_n, v_n are almost commuting unitaries in fixed dimension and w(u_n, v_n) = 0 eventually, then they are approximated by commuting unitaries in the same dimension.
    Used as a black box in the proof of Theorem 2.3; the paper contributes the dimension-enlargement step, not this theorem.
  • domain assumption Eckhardt's construction provides maps φ_n : Γ → U_{d_n} with operator-norm almost multiplicativity and trace convergence to τ (proof of [8, Theorem 2.14]).
    This is the key transfer step from Hilbert–Schmidt to operator norm; the paper cites the proof rather than reproducing it, and it is load-bearing for Theorem 3.1.
  • domain assumption Fournier-Facio–Willett theorem ([13]): every separable unital C*-algebra with the local lifting property and residual finite-dimensionality is very-flexibly stable in every normalized unitarily invariant matrix norm.
    Used for the very-flexible stability half of Corollary 3.4; cited from an arXiv preprint.
  • domain assumption Eckhardt's group eG_p is finitely generated, residually finite, solvable, amenable, and has a normal subgroup Λ for which the trace τ = 1_Λ is not a pointwise limit of finite-dimensional traces.
    These properties are imported from Eckhardt's work and are needed for both halves of Corollary 3.4.
  • standard math C*(eG_p) is nuclear and residually finite-dimensional, hence has the local lifting property.
    Standard C*-algebra facts used to apply the Fournier-Facio–Willett theorem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of $\mathbb{Z}^2$ is flexibly stable in the operator norm." pith.science (2026). https://pith.science/paper/MR7C7K6H

@misc{pith2026260717578,
  author       = {Pith},
  title        = {Pith review of: $\mathbbZ^2$ is flexibly stable in the operator norm},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MR7C7K6H}},
  note         = {Machine review of arXiv:2607.17578}
}
abstract

A cornerstone of stability theory is Voiculescu's 1983 counterexample: he constructed a sequence of pairs of unitary matrices whose commutators converge to zero in the operator norm, but whose distances from the set of commuting unitary pairs remain bounded away from zero. Namely, the group $\mathbb{Z}^2$ is not stable in the operator norm. We prove, somewhat surprisingly, that stability is restored after an asymptotically negligible enlargement of the dimension. That is, the group $\mathbb{Z}^2$ is flexibly stable in the operator norm. This provides the first example, in any context, of a flexibly stable group that is not stable. Building on a construction of Eckhardt, who produced finitely generated amenable groups that are very-flexibly stable but not flexibly stable in the normalized Hilbert-Schmidt norm, we show that the same groups exhibit the analogous separation in the operator norm: they are very-flexibly stable but not flexibly stable.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 4 linked inside Pith

  1. [1]

    Becker and A

    O. Becker and A. Lubotzky. Group stability and Property (T).J. Funct. Anal., 278(1):108298, 20, 2020

  2. [2]

    Bowen and P

    L. Bowen and P. Burton. Flexible stability and nonsoficity.Trans. Amer. Math. Soc., 373(6):4469–4481, 2020

  3. [3]

    Chapman and A

    M. Chapman and A. Lubotzky. Stability of homomorphisms, coverings and cocycles I: equivalence.Eur. J. Math., 11(1):Paper No. 13, 38, 2025. Z2 IS FLEXIBLY STABLE 13

  4. [4]

    Chapman and A

    M. Chapman and A. Lubotzky. Stability of homomorphisms, coverings and cocycles II: examples, applications and open problems.Adv. Math., 463:Paper No. 110117, 38, 2025

  5. [5]

    Dadarlat

    M. Dadarlat. Obstructions to matricial stability of discrete groups and almost flat K-theory.Adv. Math., 384:Paper No. 107722, 29, 2021

  6. [6]

    A. Dogon. Flexible Hilbert-Schmidt stability versus hyperlinearity for property (T) groups.Math. Z., 305(4):Paper No. 58, 20, 2023

  7. [7]

    Dor-On, L

    A. Dor-On, L. Hall, and I. Kachkovskiy. On almost commuting unitary matrices. arXiv:2510.03674, 2025

  8. [8]

    Eckhardt

    C. Eckhardt. Residually finite amenable groups that are not Hilbert-Schmidt stable. arXiv:2501.07791, 2025

Show all 19 references
  1. [9]

    Eckhardt and T

    C. Eckhardt and T. Shulman. On amenable Hilbert-Schmidt stable groups.J. Funct. Anal., 285(3):Paper No. 109954, 31, 2023

  2. [10]

    S. r. Eilers, T. A. Loring, and G. K. Pedersen. Morphisms of extensions ofC∗-algebras: pushing forward the Busby invariant.Adv. Math., 147(1):74–109, 1999

  3. [11]

    R. Exel. The soft torus and applications to almost commuting matrices.Pacific J. Math., 160(2):207–217, 1993

  4. [12]

    Exel and T

    R. Exel and T. Loring. Almost commuting unitary matrices.Proc. Amer. Math. Soc., 106(4):913–915, 1989

  5. [13]

    Fournier-Facio and R

    F. Fournier-Facio and R. Willett. The local lifting property, property fd, and stability of approximate representations.arXiv:2603.18456, 2026

  6. [14]

    Gerhold and O

    M. Gerhold and O. Shalit. Dilation distance and the stability of ergodic commutation relations.arXiv:2406.05864, 2024

  7. [15]

    Gong and H

    G. Gong and H. Lin. Almost multiplicative morphisms and almost commuting ma- trices.J. Operator Theory, 40(2):217–275, 1998

  8. [16]

    Hadwin and T

    D. Hadwin and T. Shulman. Stability of group relations under small Hilbert-Schmidt perturbations.J. Funct. Anal., 275(4):761–792, 2018

  9. [17]

    A. Ioana. Stability for product groups and property (τ).J. Funct. Anal., 279(9):108729, 32, 2020

  10. [18]

    Lazarovich, A

    N. Lazarovich, A. Levit, and Y. Minsky. Surface groups are flexibly stable.J. Eur. Math. Soc. (JEMS), 27(4):1739–1768, 2025

  11. [19]

    Voiculescu

    D. Voiculescu. Asymptotically commuting finite rank unitary operators without com- muting approximants.Acta Sci. Math. (Szeged), 45(1-4):429–431, 1983. Department of Mathematics, Weizmann Institute of Science, Rehovot, Is- rael Email address:alex.lubotzky@mail.huji.ac.il Schoo...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.