Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Local Statistics of Singular Values for Products of Truncated Unitary Matrices

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper proves that the local singular-value statistics of products of truncated unitary matrices undergo a three-phase transition governed by a single modified depth-to-width ratio.

desk verdict The truncated-unitary three-phase transition is real and mostly proven, but the theorems drop the standing dimensional constraint and are undefined for some parameter choices. read the letter →

arxiv 2506.14193 v1 pith:ZQVIDCXF submitted 2025-06-17 math.PR

Yandong Gu , Dang-Zheng Liu This is my paper · ORCID
classification math.PR MSC 60B2015B52
keywords truncatedunitarymatricessingularvaluesdeterminantalpointprocessphasetransitioncriticalkerneldepth-to-widthratioAirysine
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's central claim is that the local singular-value statistics of products of independent truncated unitary matrices are controlled by one modified depth-to-width ratio, $\Delta_{M,n}=\sum_{j=0}^M \frac{1}{n+v_j}-\sum_{j=1}^M \frac{1}{m_j}$. As $\Delta_{M,n}\to\infty$ the correlation kernel becomes Gaussian; as $\Delta_{M,n}\to\gamma\in(0,\infty)$ it converges to explicit critical kernels; and as $\Delta_{M,n}\to0$ it returns to the standard Airy kernel at the edge and the sine kernel in the bulk. This gives products of truncated unitary matrices the same three-phase diagram previously found for products of complex Gaussian matrices. A sympathetic reader would care because singular values of such products appear in wireless communications and chaotic dynamics, and the result reduces many separate limiting statements to a single parameter.

What carries the argument

The load-bearing object is the modified depth-to-width ratio $\Delta_{M,n}$ of (1.3), which combines all layer widths $n+v_j$ and row counts $m_j$ into one parameter. It enters the kernel through sums of digamma functions, and the proofs analyze the explicit double contour integral representation (1.2) of the correlation kernel. In the high-DWR regime a steepest-descent argument with a scaled contour through $1-k$ yields the Gaussian kernel; in the moderate regime Euler's reflection formula transforms the kernel into an integral whose residue evaluation produces the $\theta$-function critical kernels; in the low regime a cubic saddle at a point $z_0$ defined by the rational equation (1.17) gives the Airy kernel, and a separate parameterization of the limiting density yields the sine kernel. The two critical kernels (1.7)-(1.8) are named limit objects that interpolate between GUE statistics and classical Gaussian behavior.

What would settle it

Set $m_j=n+v_j$ for all but one factor so that the dimensional constraint $n\le\sum_j(m_j-n-v_j)$ fails while still letting $\Delta_{M,n}\to\gamma\in(0,\infty)$, and compute local spacing statistics of the squared singular values numerically; if the limiting statistics are not $K_{\mathrm{crit}}^{\mathrm{(bulk)}}$ or $K_{\mathrm{crit}}^{\mathrm{(edge)}}$, the claimed universality of the phase transition fails at that boundary. Alternatively, verify the convergence in (1.6) directly for a product with $M\sim n^2$ by Monte Carlo simulation of the kernel.

Watch

Extended reading notes

Core claim

The discovery is a complete characterization of local singular-value correlations for $Y_M=T_M\cdots T_1$, where each $T_j$ is an $(n+v_j)\times(n+v_{j-1})$ truncation of a Haar-distributed unitary matrix. The squared singular values form a determinantal point process with kernel $K_n(x,y)$ written as a double contour integral, and the paper proves three uniform limits for this kernel. In the high-depth regime the rescaled kernel tends to $\frac{1}{\sqrt{2\pi}}e^{-\eta^2/2}$, so singular-value positions become independent Gaussian fluctuations; in the moderate-depth regime it tends to the bulk and edge critical kernels $K_{\mathrm{crit}}^{\mathrm{(bulk)}}$ and $K_{\mathrm{crit}}^{\mathrm{(edge)}}$ defined through a Jacobi $\theta$ function and a gamma-function contour integral; and in the low-depth regime it tends to the Airy kernel at the soft edge and the sine kernel in the bulk. The paper also states the same three-phase transition for products of complex Ginibre matrices and for products of Ginibre with inverse Ginibre matrices.

Load-bearing premise

Everything rests on the explicit double contour integral kernel for the determinantal point process, which is only valid when the dimensional constraint $n\le\sum_{j=1}^M(m_j-n-v_j)$ holds; if that constraint fails, the kernel formula and with it all three theorems are not justified.

Editorial extensions

If this is right

  • If the theorems are correct, the single parameter $\Delta_{M,n}$ predicts which of the three statistics regimes applies, so fixed-$M$ and growing-$M$ limits are unified rather than treated as separate cases.
  • In the deep-narrow regime, local singular values become effectively independent: the normalized correlation function converges to a Gaussian, so deterministic scaling plus independent normal fluctuations emerges.
  • In the balanced regime, the critical kernels provide the interpolation between GUE and Gaussian behavior for truncated unitary products, and they coincide with the critical kernels found for Ginibre products.
  • In the shallow-wide regime, the product ensemble exhibits GUE universality despite $M\to\infty$: the Airy kernel at the soft edge and the sine kernel in the bulk.
  • The paper's remark extends the same three-phase statement to products of complex Ginibre matrices and to Ginibre-inverse-Ginibre products, including the spherical ensemble as a special case.

Reading between the lines

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

  • A natural extension not pursued in the paper is testing whether the same modified ratio controls local statistics for products with non-Haar or weakly dependent truncations; the kernel argument would likely need a new determinantal structure.
  • If the critical edge kernel also appears in last-passage percolation models as the cited literature suggests, the phase transition here may be a random-matrix manifestation of a hard-to-soft transition visible in other probabilistic models.
  • A concrete testable consequence of Theorem 1.1 is that eigenvalue repulsion disappears in the deep-narrow limit; simulations of spacing distributions for $M\gg n$ should look essentially independent, while $M\sim\gamma n$ should show the critical kernel's signature.
  • The explicit edge scaling in Theorem 1.3(i), with $z_0$ and $\lambda_M$, suggests a finite-$M$ correction rate of order $\rho_{M,n}^{-1/2}$; a numerical check of that rate could confirm that the low-depth regime is reached quickly as $\Delta_{M,n}\to0$.
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 / 5 minor

Summary. This paper studies the local singular-value statistics of products Y_M = T_M ... T_1 of truncated unitary matrices. It introduces the modified depth-to-width ratio Delta_{M,n} in (1.3) and proves three asymptotic regimes for the correlation kernel of the determinantal point process of squared singular values, whenever the explicit double-contour formula (1.2) applies: for Delta_{M,n} to infinity the kernel becomes Gaussian (Theorem 1.1), for Delta_{M,n} to gamma in (0, infinity) it converges to the critical bulk and edge kernels (Theorem 1.2), and for Delta_{M,n} to 0 it converges to the Airy and sine kernels (Theorem 1.3). The proofs use steepest-descent analysis of the kernel (1.2), together with a new parameterization of the limiting density in the low-DWR regime, and the paper also states a simplified edge limit (Theorem 3.1).

Significance. The announced three-phase transition is a natural and valuable counterpart, for the truncated-unitary ensemble, of the Ginibre-product phase transition established in [4,25]. The modified DWR Delta_{M,n} is a useful unifying parameter, and the explicit low-DWR parameterizations leading to Theorem 1.3 and Theorem 3.1 are new. The paper gives detailed steepest-descent estimates for the main bulk and high-DWR cases and tracks error terms in several places. However, several load-bearing ingredients are currently missing or stated too broadly, so the results cannot be accepted in the present form.

major comments (4)
  1. [Section 1.2, Eqs. (1.1)-(1.3)] Theorems 1.1-1.3 do not state the dimensional condition (1.1) as a hypothesis, although the proofs rely on the determinantal kernel (1.2), which the paper itself says is valid only when (1.1) holds. The parameter regimes in the theorems do not force (1.1): for M=1, v_1=0, m_1=3n/2, one has Delta_{M,n}=4/(3n) to 0, so Theorem 1.3(i) applies as written, but n <= m_1 - n - v_1 = n/2 fails, and equation (1.17) has no positive solution (the cross-multiplied equation gives z = -3n). Hence rho_{M,n}, lambda_{M,n}, and the statement of Theorem 1.3(i) are undefined in an allowed regime; Theorems 1.1 and 1.2 inherit the same problem through their use of kernel (1.2). The fix is to impose (1.1) as a standing assumption or to handle the non-DPP regime by other methods; without this, the theorems overreach.
  2. [Section 2.2, Theorem 1.2(ii)] The proof of the critical edge limit u=0 is omitted, with the text stating that it is similar to u in (0,1) and only the scaling functions need to be substituted. This is not a negligible variant: the contour analysis and the estimates leading to Lemma 2.2 are specific to the bulk scaling (1.11), while the edge statement (1.13)-(1.14) requires its own verification. Since Theorem 1.2(ii) is one of the three main limits announced in the abstract, the omitted proof must be supplied.
  3. [Section 3.1, Lemma 3.4] Lemma 3.4 is the global estimate that controls all contributions outside the local saddle-point domains in the proof of Theorem 1.3(ii), and its proof is explicitly omitted. The reference to Lemma 2.1 of [26] is not sufficient, because the present parameterization via the Fuss-Catalan-type density and the kernels eK_n is different; the monotonicity and steepest-descent details need to be checked in this setting.
  4. [Section 3.2, Theorem 3.1] Theorem 3.1 is presented as a theorem but its proof is only sketched. If it is intended as a corollary of Theorem 1.3(i) for the simplified kernel, the reduction should be made explicit; if it is an independent result, the omitted contour-deformation and saddle-point estimates should be provided. As written the reader cannot verify the displayed normalization in (3.10).
minor comments (5)
  1. [General] Throughout the text there are numerous typographical errors (e.g., 'T runcated', 'intoduce', 'propoties', 'etablishing', 'Togeter', 'denstiy', 'Stielties'); these should be corrected.
  2. [Eq. (1.12)] Equation (1.12) has a misplaced parenthesis in the exponential factor; it should read e^{(g(xi)-g(eta))[nu]} K_n(g(xi),g(eta)).
  3. [Section 2.2, Eq. (2.44)] In equation (2.44), the differential dw is written ambiguously after the t-integral; the order of integration should be displayed with parentheses.
  4. [Section 3.1, Eqs. (3.38)-(3.39) and (3.14)] The two displays labeled (3.38) and (3.39) are identical, and the definitions of Sigma-tilde_+ and Sigma-tilde_- in (3.14) both use the same interval for theta rather than distinguishing the two conjugate branches; these appear to be typos.
  5. [References] The bibliography entry for [4] contains a stray '255202 (2014)' at the end.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the three-phase limits are genuine asymptotic analyses of the explicit kernel (1.2), with only a scope-of-validity caveat.

full rationale

No circular step is present. Each theorem is obtained by steepest-descent and saddle-point analysis of the explicit double-contour kernel (1.2), cited from [22], and the target kernels are not assumed as inputs: the Gaussian limit follows from F''(1-k;k) = -rho^2_{M,n}(k) and the expansion (2.19); the critical limit follows from the t^2-coefficient gamma' computed in (2.35), which is derived rather than imposed; and the Airy and sine limits follow from cubic and quadratic saddle-point expansions with rho_{M,n}, lambda_{M,n} and z0 defined through the spectral equation (1.17). The modified DWR Delta_{M,n} is an asymptotic parameter appearing in the kernel, not a parameter fitted to the claimed limit kernels. The critical kernels in (1.7)-(1.8) are imported from [4] and [25] as definitions only; [25] is coauthored by the second author, but the convergence proofs here do not invoke [25] as a black box for the truncated-unitary limit, so this self-citation is not load-bearing. Two caveats should be weighed separately from circularity: the theorems omit the standing dimensional constraint (1.1), and low-DWR parameters can violate it (e.g., M=1, v_1=0, m_1=3n/2 gives Delta_{M,n}->0 but n <= m_1-n-v_1 fails), in which case rho_{M,n}, lambda_{M,n} and z0 in (1.15)-(1.17) are not well-defined; and the u=0 edge-critical proof is omitted in Section 2.2 ('so we omit it') while Theorem 3.1 is only sketched. These are completeness and validity gaps, not circular reductions.

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

No numerical constants are fitted to output; the parameters rho_{M,n}, lambda_M, z_0, gamma', x_*, and c_2 are all explicit functions of the model data (m_j, n, v_j). The invented_entities list is empty: the critical kernels and Airy and sine limits are existing objects. All novel content rests on the DPP kernel imported from prior work and on standard asymptotic analysis.

assumptions (4)
  • domain assumption The eigenvalues of Y_M^* Y_M form a determinantal point process with the explicit kernel (1.2).
    The whole proof analyzes this kernel; the DPP property holds only under the dimensional constraint (1.1), n <= sum_{j=1}^M (m_j - n - v_j), as characterized in [12] and used in Section 1.1.
  • domain assumption The limits defining the regimes exist, e.g., Delta_{M,n} to gamma and gamma' = lim in (1.10).
    Theorems 1.1-1.3 assume these limits; no rates or verifications are given.
  • standard math Stirling's formula, digamma expansions (2.1)-(2.3), and standard steepest-descent estimates can be applied uniformly in the many-parameter limit.
    Used throughout Sections 2 and 3 for the gamma functions in (1.2).
  • standard math The contour deformations (e.g., Sigma_global in Section 2.3, C and Sigma in Section 3) do not pick up unwanted residues, and the signed real parts of F and f_{M,n} have the monotonicity asserted in Lemmas 2.1-2.3 and 3.2-3.4.
    These lemmas control the non-local parts of the integrals; some proofs are only sketched (Lemma 3.4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Local Statistics of Singular Values for Products of Truncated Unitary Matrices." pith.science (2026). https://pith.science/paper/ZQVIDCXF

@misc{pith2026250614193,
  author       = {Pith},
  title        = {Pith review of: Local Statistics of Singular Values for Products of Truncated Unitary Matrices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZQVIDCXF}},
  note         = {Machine review of arXiv:2506.14193}
}
read the original abstract

This paper investigates local spectral statistics of singular values for many products of independent large rectangular matrices, sampled from the ensemble of truncated unitary matrices with the invariant Haar measure. Our main contribution establishes a universal three-phase transition in these statistics, demonstrating an interpolation beween GUE statistics and classical Gaussian behavior. While such transition was previously known for products of complex Gaussian matrices\cite{ABK19}\cite{LWW23}, the current work provides the complete characterization in the truncated unitary matrix setting.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Edge statistics for singular values of products of rectangular complex Ginibre matrices

    math.PR 2025-07 conditional novelty 5.0 of 10

    For products of rectangular complex Ginibre matrices, the depth-to-width ratio Delta is the sharp threshold: Delta tending to 0 gives Airy kernel edge statistics, Delta tending to infinity gives Gaussian edge fluctuations.

Reference graph

Works this paper leans on

32 extracted references · 16 canonical work pages · cited by 1 Pith paper

  1. [25]

    Liu, D.-Z., Wang, D., Wang, Y.: Lyapunov exponent, universality and phase transition for products of random matrices. Comm. Math. Phys. 399(3), 1811–1855 (2023)

    2023

  2. [26]

    Liu, D.-Z., Wang, D., Zhang, L.: Bulk and soft-edge universality for singular values of products of Ginibre random matrices. Ann. Inst. Henri Poincar´ e Probab. Stat. 52(4), 1734– 1762 (2016)

    2016

  3. [1]

    Ahn, A.: Fluctuations of β-Jacobi product processes. Probab. Theory Relat. Fields 183, 57–123 (2022)

    2022

  4. [2]

    Ahn, A., Peski, V.R.: Lyapunov exponents for truncated unitary and Ginibre matrices. Ann. Inst. Henri Poincar´ e Probab. Stat. 59(2), 1029–1039 (2023)

    work page 2023

  5. [3]

    Journal of Physics A 47(39), 395202, 35 (2014)

    Akemann, G., Burda, Z., Kieburg, M.: Universal distribution of Lyapunov exponents for products of Ginibre matrices. Journal of Physics A 47(39), 395202, 35 (2014)

    2014

  6. [4]

    EPL (Europhys

    Akemann, G., Burda, Z., Kieburg, M.: From integrable to chaotic systems: Universal local statistics of Lyapunov exponents. EPL (Europhys. Lett.) 126(4), 40001 (2019). 255202 (2014)

    2019

  7. [5]

    Acta Phys

    Akemann, G., Ipsen, J.R.: Recent exact and asymptotic results for products of independent random matrices. Acta Phys. Polon. B 46(9), 1747–1784 (2015)

    work page 2015

  8. [6]

    Akemann, G., Ipsen, J.R., Kieburg, M.: Products of rectangular random matrices: Singular values and progressive scattering. Phys. Rev. E 88(5), 052118 (2013)

    work page 2013

Show all 32 references
  1. [7]

    Akemann, G., Kieburg, M., Wei, L.: Singular value correlation functions for products of Wishart random matrices. J. Phys. A 46(27), 275205 (2013)

    2013

  2. [8]

    Alexeev, N., G¨ otze, F., Tikhomirov, A.: Asymptotic distribution of singular values of powers of random matrices. Lith. Math. J. 50(2), 121–132 (2010)

    2010

  3. [9]

    Cam- bridge Studies in Advanced Mathematics 118

    Anderson, G.W., Guionnet, A., Zeitouni, O.: An Introduction to Random Matrices. Cam- bridge Studies in Advanced Mathematics 118. Cambridge University Press, Cambridge, 2010

    2010

  4. [10]

    Bellman, R.: Limit theorems for non-commutative operations. I. Duke Math. J. 21, 491–500 (1954). 23

    1954

  5. [11]

    arXiv preprint arXiv:2503.22801 (2025)

    Berezin, S., Strahov, E.: Last-passage percolation and product-matrix ensembles. arXiv preprint arXiv:2503.22801 (2025)

    2025 arXiv

  6. [12]

    Claeys, T., Kuijlaars, A. B. J., Wang, D.: Correlation kernels for sums and products of random matrices. Random Matrices Theory Appl. 4(4), 1550017 (2015)

    2015

  7. [13]

    Springer Series in Solid-State Sciences 104

    Crisanti, A., Paladin, G., Vulpiani, A.: Products of Random Matrices in Statistical Physics. Springer Series in Solid-State Sciences 104. Springer, Berlin, 1993

    1993

  8. [14]

    Forrester, P.J.: Eigenvalue statistics for product complex Wishart matrices. J. Phys. A 47(34), 345202 (2014)

    2014

  9. [15]

    Forrester, P.J., Liu, D.-Z.: Singular values for products of complex Ginibre matrices with a source: hard edge limit and phase transition. Commun. Math. Phys. 344(1), 333–368 (2016)

    2016

  10. [16]

    Furstenberg, H., Kesten, H.: Products of random matrices. Ann. Math. Stat. 31, 457–469 (1960)

    1960

  11. [17]

    Gorin, V., Sun, Y.: Gaussian fluctuations for products of random matrices. Amer. J. Math. 144(2), 287–393 (2022)

    2022

  12. [18]

    In Operator Algebra and Dynamics 157–186

    Haagerup, U., M¨ oller, S.: The law of large numbers for the free multiplicative convolu- tion. In Operator Algebra and Dynamics 157–186. Springer Proc. Math. Stat. 58. Springer, Heidelberg, 2013

    2013

  13. [19]

    Hanin, B., Paouris, G.: Non-asymptotic results for singular values of Gaussian matrix products. Geom. Funct. Anal. 31, 268–324 (2021)

    2021

  14. [20]

    arXiv preprint arXiv:2503.07872 (2025)

    Hanin, B., Jiang, T.: Global universality of singular values inproducts of many large random matrices. arXiv preprint arXiv:2503.07872 (2025)

    2025 arXiv

  15. [21]

    Jiang, T., Qi, Y.: Spectral radii of large non-Hermitian random matrices. J. Theor. Probab. 30(1), 326–364 (2017)

    2017

  16. [22]

    and Stivigny, D.: Singular value statistics of matrix products with truncated unitary matrices

    Kieburg, M., Kuijlaars, A.B.J. and Stivigny, D.: Singular value statistics of matrix products with truncated unitary matrices. Int. Math. Res. Not. IMRN 11, 3392–3424 (2016)

    2016

  17. [23]

    Random Matrices Theory Appl

    Kuijlaars, A.B.J., Stivigny, D.: Singular values of products ofrandom matrices and polyno- mial ensembles. Random Matrices Theory Appl. 3(3), 1450011 (2014)

    2014

  18. [24]

    Kuijlaars, A.B.J., Zhang, L.: Singular values of products of Ginibre random matrices, multiple orthogonal polynomials and hard edge scaling limits. Commun. Math. Phys. 332(2), 759–781 (2014)

    2014

  19. [27]

    Liu, D.-Z., Wang, Y.: Phase transitions for infinite products of large non-Hermitian random matrices. Ann. Inst. Henri Poincar´ e Probab. Stat. 60(4), 2813–2848 (2024)

    2024

  20. [28]

    Fields Institute Mono- graphs 35

    Mingo, J.A., Speicher, R.: Free Probability and Random Matrices. Fields Institute Mono- graphs 35. Springer, New York, 2017

    2017

  21. [29]

    Random Matrices Theory Appl

    Neuschel, T.: Plancherel-Rotach formulae for average characteristic polynomials of products of Ginibre random matrices and the Fuss-Catalan distribution. Random Matrices Theory Appl. 3(1), 1450003 (2014)

    2014

  22. [30]

    Random Matrices Theory Appl

    Neuschel, T.: Spectral densities of singular values of products of Gaussian and truncated unitary random matrices. Random Matrices Theory Appl. 9(04), 2050014 (2020). 24

    2020

  23. [31]

    (eds.): NIST handbook of math- ematical functions

    Olver, F.W.J., Lozier, D.W., Boisvert, R.F., Clark, C.W. (eds.): NIST handbook of math- ematical functions. U.S. Department of Commerce, National Institute of Standards and Technology, Washington, DC; Cambridge University Press, Cambridge 2010. With 1 CD- ROM (Windows, Macinto...

    2010

  24. [32]

    Tulino, A.M., Verd´ u, S.: Random matrix theory and wireless communications. Found. Trends Commun. Inf. Theory. 1(1), 1-182 (2004). 25

    2004

Pith tools

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