Pith. sign in

REVIEW 4 major objections 6 minor 19 references

A remark on Chebyshev rational functions, multipoint Pad\'e approximants and Noise

T0 review · 4 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

Pith's one-line read Chebyshev rational functions provide explicit multipoint Padé approximants to 1/√(x²−1), interpolating both the function and its derivative at each prescribed node.

desk verdict Core multipoint Padé result is correct and new; Thm 6.3's proof is a real gap, and the noise section is numerical only — worth refereeing after fixes. read the letter →

arxiv 2606.13965 v2 pith:IPKIFIDN submitted 2026-06-11 math.CA hep-thmath-phmath.CVmath.MP

classification math.CAhep-thmath-phmath.CVmath.MP MSC 30E0533C4742C0530B7033F05
keywords multipointPadéapproximantsChebyshevrationalfunctionsBlaschkeproductscontinuedfractionsinterpolationnoiseFroissartdoubletsanalyticcontinuation
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 shows that the Chebyshev rational functions—rational analogues of Chebyshev polynomials built from finite Blaschke products—give explicit multipoint Padé approximants to the function φ(x)=1/√(x²−1), which has a branch cut on [−1,1]. For distinct nodes a_k outside the cut, the ratio U_{n−1}/T_n matches both φ and its derivative at every a_k, which is exactly the condition for a type [n−1/n] multipoint Padé approximant at double nodes. The paper proves three-term recurrence relations and an R_II-type continued fraction for these functions, and establishes locally uniform convergence to φ on the complement of the cut under a divergence condition on the Blaschke parameters. It also gives numerical evidence that adding small random noise to the interpolation data triggers a breakdown with spurious poles and zeros, and that these spurious pairs accumulate near the interpolation nodes rather than on the natural boundary.

What carries the argument

The key object is the finite Blaschke product f_n(z)=∏_{k=1}^n (z−c_k)/(1−c_k z), where the c_k are related to the interpolation nodes by the Joukowski map x=½(z+1/z). The rational functions T_n and U_n are defined through f_n: T_n=½(f_n+1/f_n), and U_n=(f_{n+1}−1/f_{n+1})/(z−z^{−1}). This machinery carries the argument because the ratio U_{n−1}/T_n equals φ(x)·(1−f_n²)/(1+f_n²); since f_n(c_i)=0, the ratio matches φ and φ′ at each a_i. The same identity turns convergence into the standard Blaschke-product problem of whether f_n→0 in the unit disk, which is governed by the divergence of ∑(1−|c_k|).

What would settle it

For the triangular node set a_{n,k}=e^{2πik/n}, compute c_{n,k} using the branch with |c_{n,k}|<1 and evaluate ∑_{k=1}^n (1−|c_{n,k}|); if this sum does not tend to infinity as n grows, then the proof of Theorem 6.3 cannot hold and the claimed convergence for the roots-of-unity scheme collapses.

Watch

Extended reading notes

Core claim

The central claim is Theorem 5.2: for the Chebyshev rational functions constructed from a finite Blaschke product with zeros at c_k = a_k − √(a_k²−1), the ratio U_{n−1}/T_n interpolates φ(x)=1/√(x²−1) and its derivative at each a_k. Equivalently, it is a multipoint Padé approximant of type [n−1/n] at the double nodes {a_k, a_k}. The proof uses the identity (U_{n−1}/T_n)(x) = φ(x)·(1−f_n(z)²)/(1+f_n(z)²), where z is the inverse Joukowski variable; at x=a_k, z=c_k and f_n vanishes, forcing the ratio to φ(a_k), with a short calculation giving the derivative match. For real a_k with ∑(1−|c_k|)=∞, the paper proves locally uniform convergence to φ on C∖[−1,1] and a convergent continued fraction ex

Load-bearing premise

For complex nodes, the construction needs a branch of √(a_k²−1) such that every c_k lies inside the unit disk and the Blaschke divergence condition holds; the paper does not specify this branch, and its roots-of-unity convergence argument relies on a summability assertion that is not demonstrated.

Editorial extensions

If this is right

  • The functions U_{n−1}/T_n provide an explicit, computable family of multipoint Padé approximants to 1/√(x²−1) for any choice of distinct nodes outside [−1,1].
  • Under the condition ∑(1−|c_k|)=∞, the associated R_II-type continued fractions converge locally uniformly, giving a continued-fraction representation of 1/√(x²−1).
  • The determinant formula in Theorem 5.3 lets one construct double-node multipoint Padé approximants for general analytic functions from divided-difference data.
  • The numerical noise study shows a breakdown phenomenon quantitatively similar to the single-point Padé case, with the new feature that spurious poles and zeros migrate to the interpolation nodes as noise grows.
  • Symmetrically arranged interpolation nodes appear less sensitive to noise than asymmetric arrangements, suggesting the geometry of the node set controls the stability threshold.

Reading between the lines

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

  • If convergence extends to complex nodes under a suitable branch choice—which the paper does not specify—the same rational family could serve as a practical tool for analytic continuation from scattered finite data, particularly in settings where only a finite number of function values are known.
  • The observed dependence of the breakdown order's logarithmic slope on node distribution suggests an optimization problem: arranging interpolation nodes to maximize the noise threshold; testing whether symmetric or circular layouts are optimal is a direct extension of the numerics.
  • The explicit identity T_n²+(1−x²)U_{n−1}²=1 could be used to derive rigorous a posteriori error bounds in the presence of noise, converting the numerical evidence into a provable statement about where the approximant is reliable.
  • The migration of spurious poles and zeros toward the interpolation nodes suggests a noise-detection heuristic: clusters of nearby poles and zeros in a computed multipoint Padé approximant may flag corrupted data points.
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

4 major / 6 minor

Summary. The paper studies rational Chebyshev functions built from finite Blaschke products with prescribed poles outside [-1,1]. It derives three-term recurrences (Theorems 3.1, 3.2) and a Thiele-type continued fraction (Corollary 3.3), extends the construction to complex nodes (Section 4), and proves that the ratio U_{n-1}/T_n is a multipoint Padé approximant of type [n-1/n] to phi(x)=1/sqrt(x^2-1) at double nodes (Theorem 5.2). It also gives determinant formulas for general multipoint Padé approximants (Theorem 5.3), proves locally uniform convergence for real nodes satisfying sum(1-|c_k|)=infinity (Theorem 6.1) and for the triangular node set of n-th roots of unity (Theorem 6.3), and presents numerical experiments on noise-induced breakdown of multipoint Padé approximants (Section 7).

Significance. If all claims are correct, the paper provides one of the few explicit, convergent families of multipoint Padé approximants for a function with a branch cut, together with computable recurrences and a continued fraction. The identity [T_n]^2+(1-x^2)[U_{n-1}]^2=1 (Prop. 4.3) is particularly clean and is the algebraic backbone of the interpolation result. The convergence theorem for the real-node case (Thm 6.1) follows from standard Blaschke-product facts and is a nice extension of the classical Chebyshev-polynomial convergence. The numerical study of noise is qualitative but illustrates a plausible extension of the Costin–Dunne–Meynig phenomenon. However, as detailed below, the proof of Theorem 6.3 has a gap, the branch convention for complex nodes is not stated, and there are indexing and sign errors in auxiliary statements. With these repaired, the paper would be a solid contribution to explicit multipoint Padé approximation.

major comments (4)
  1. [Section 6, Theorem 6.3] The proof of Theorem 6.3 is not valid as written. The product f_n(w) = ∏_{k=1}^n (w-c_{n,k})/(1-c_{n,k}w) is claimed to be a finite Blaschke product because the set {c_{n,k}} contains conjugate pairs. While the full product is indeed Blaschke (denominators can be rearranged to 1-\bar c w when the set is conjugation-invariant), the individual factors are not Blaschke factors for complex c: |(w-c)/(1-cw)| can exceed 1 inside D. Consequently the estimates |f_n(w)| ≤ exp(-∑(1-|b_{n,k}(w)|)) and 1-|b_{n,k}(w)| ≥ C(1-|c_{n,k}|) are unjustified. Moreover, the assertion that ∑(1-|c_{n,k}|)→∞ for the n-th roots of unity is only sketched via the phrase “fill in the unit circle”; this requires a proof. The theorem may be true, but the presented argument does not establish it. Please supply a correct proof, e.g., by applying the standard Blaschke-product estimate to the paired conjugate factors and
  2. [Sections 2, 4, and 5, Theorem 5.2] The branch of the square root is not specified for complex nodes, and the proof of Theorem 5.2 contains a branch-convention slip. In Section 2, z(x) is defined as the unique Joukowski root with |z|>1. Then for x=a_i one has z(a_i)=1/c_i, not z(a_i)=c_i. The proof states “x=a_i implies z=c_i,” which is incorrect under the stated convention. The interpolation claim of Theorem 5.2 is likely true and can be obtained from Prop. 4.3 and the partial fraction expansions (24)–(27), but the proof as written is not rigorous. Please state explicitly that c_k is chosen with |c_k|<1, i.e., the inner Joukowski root, and give a correct derivation of the interpolation conditions, or use the algebraic identity (23).
  3. [Section 3, Remark after Theorem 3.2] The initial conditions for the recurrence (14) are inconsistent. From the definition U_n(cost)=sinθ_{n+1}(t)/sint, one obtains U_0(x)=sinδ_1(t)/sint = -sqrt(a_1^2-1)/(x-a_1), not U_0=0. What the remark calls U_1 is in fact U_0. This indexing error propagates to the initialization of the recurrence and potentially to the continued fraction (15). Please correct the initial values and verify that the subsequent formulas for U_n are indexed consistently.
  4. [Theorem 5.3, eq. (45)] The determinant formula in Theorem 5.3 is false as stated. Expanding the displayed (n+1)x(n+1) determinant along the last row gives detM - ∑ detM_i/(x-a_i) (with M_i defined as in the proof after eq. (43)), not detM + ∑ detM_i/(x-a_i). For n=1 the theorem would assert P_1(x)/ω_1(x) = φ'(a1) - φ(a1)/(x-a1), which does not equal T_1(x)=(1-a1 x)/(x-a1). The sign error originates in the linear system (42): the right-hand side should have a minus sign. This theorem needs to be corrected or explicitly repairable; it is a standalone result but is advertised as part of the paper's contributions.
minor comments (6)
  1. [Abstract] The abstract contains a typo: “RA TIONAL” should be “RATIONAL”.
  2. [Section 3] In the first paragraph, “they they showed” is a duplicated word.
  3. [Proposition 4.3] The proof states z-z^{-1} = -2√(x^2-1), but with the convention in Section 2, z-z^{-1} = 2√(x^2-1). The squared identity is unaffected, but the sign should be corrected for consistency.
  4. [Theorem 6.1] The phrase “standard fact about Blaschke products” is terse. Since the argument is in the exterior domain, specify that one applies the Blaschke theorem in the variable w=1/z and then transfers the conclusion back to z.
  5. [Theorem 6.3] For even n, the n-th roots of unity include ±1, which lie in [-1,1] and are excluded by the standing assumption. The remark about disregarding ±1 should be made precise in the triangular setting, where the number of effective nodes then depends on n.
  6. [Section 7] The noise experiments are described qualitatively. While the paper makes clear this is numerical evidence, adding a brief description of how the threshold n_c is extracted from the pole-zero plots would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the multipoint Padé identification is an algebraic theorem derived from explicit Blaschke-product definitions; convergence uses standard facts, and the only self-citation is a benchmark.

full rationale

The central claim (Theorem 5.2) is not circular. The functions T_n and U_n are explicitly defined from finite Blaschke products f_n(z)=∏(z-c_k)/(1-c_k z), and the multipoint Padé property is proved via the identity (23), [T_n]^2+(1-x^2)[U_{n-1}]^2=1, together with the error formula giving the double-node vanishing. No Padé interpolation is assumed as an input. The recurrence relations, partial-fraction expansions, and continued fractions are algebraic consequences of the same definitions. The convergence theorems rest on the standard Blaschke-product criterion under ∑(1-|c_k|)=∞, not on a result imported from the authors' prior work. The noise section cites the coauthored paper [9] only as a known benchmark for comparing the paper's own numerical experiments; those experiments are computed directly and do not use [9] as an ingredient. There is a genuine proof gap in Theorem 6.3: the factors b_{n,k}(w)=(w-c_{n,k})/(1-c_{n,k}w) are not Blaschke factors for non-real c, so the estimate 1-|b_{n,k}(w)| ≥ C(1-|c_{n,k}|) and the asserted divergence of ∑(1-|c_{n,k}|) need additional justification. That is a correctness risk, not circularity. No load-bearing input is renamed as a prediction, and no central result is forced by a self-citation chain.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

Main derivations are self-contained: T_n and U_n are defined explicitly via Blaschke products, and the interpolation and convergence theorems use standard facts about Blaschke products, divided differences, and continued fractions. No fitted parameters enter the mathematics. The noise experiments use ad hoc node choices and fitted threshold slopes, which are listed as free parameters. No new physical or mathematical entities are invented.

free parameters (1)
  • slope of noise-breakdown order vs. log10(1/epsilon) = not reported; dashed fits in Fig. 1
    The paper claims the threshold order n_c grows logarithmically in epsilon with a slope depending on node distribution, but the slopes are only shown as dashed fits and never quantified. The central numerical claim therefore depends on unpublished fitted values.
assumptions (4)
  • standard math Finite Blaschke products with zeros c_k converge locally uniformly to 0 in D whenever sum(1-|c_k|) = infinity.
    Used in Theorem 6.1 to conclude f_n(z)^2 -> 0 and hence U_{n-1}/T_n -> phi on compact subsets of C\[-1,1].
  • standard math For each a in C\[-1,1] there is a unique c with |c|<1 satisfying a=(c+1/c)/2; the branch of sqrt(a^2-1) is chosen so c=a-sqrt(a^2-1).
    Used to define f_n as a Blaschke product and to evaluate at z=c_i in Theorem 5.2; Section 4 does not restate this branch convention.
  • domain assumption Interpolation nodes a_k are distinct and lie outside [-1,1].
    Assumed in Sections 3-6 so poles of T_n do not lie on the branch cut and the interpolation points are regular; also ensures the c_k are distinct for the Blaschke product.
  • ad hoc to paper In Theorem 6.3, the triangular node sets satisfy sum(1-|c_{n,k}|) -> infinity and the associated c_{n,k} fill the limiting curve in the required sense.
    The proof asserts this for 'n-th roots of unity' but the text is inconsistent with node sets outside [-1,1] and the summability argument is not fully demonstrated. The convergence conclusion depends on this condition.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A remark on Chebyshev rational functions, multipoint Pad\'e approximants and Noise." pith.science (2026). https://pith.science/paper/IPKIFIDN

@misc{pith2026260613965,
  author       = {Pith},
  title        = {Pith review of: A remark on Chebyshev rational functions, multipoint Pad\'e approximants and Noise},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IPKIFIDN}},
  note         = {Machine review of arXiv:2606.13965}
}
abstract

Motivated by the recent interest in multipoint Pad\'e approximants in the physics community, we discuss Chebyshev rational functions and show how they give rise to multipoint Pad\'e approximants in exactly the same way that Chebyshev polynomials produce Pad\'e approximants. We present recurrence relations for Chebyshev rational functions, as well as the underlying continued fraction of Thiele type (also known as $R_{II}$ type). Finally, we provide numerical evidence illustrating the effects of noise on this interpolation scheme and show that a phenomenon similar to that recently observed by Costin, Dunne, and Meynig for Pad\'e approximants also occurs in the multipoint setting.

Figures

Figures reproduced from arXiv: 2606.13965 by the authors.

Figure 1
Figure 1. This plot shows the breakdown Pad´e order nc for the function ϕ(x) = 1/ √ x 2 − 1, as a function of the noise strength ϵ, for two different configurations of interpolation nodes. The black squares are for nodes distributed evenly around the circle with radius 3/2, and the black circles are for nodes distributed evenly on the interval [2, 4] of the real line. Dashed lines show fits to these two sets of threshold poin… view at source ↗
Figure 2
Figure 2. The plots show the zeros (open triangles ∆) and poles (solid circles •) of the type [18/19] multipoint Pad´e approximant for the interpolation problem (47), with the input interpolation nodes ai placed on the positive real axis, as indicated by the open circles ◦. The first spurious poles appear with noise level ϵ ≈ 10−57 , and note that as the noise level increases the Pad´e poles and zeros migrate to the vicinity … view at source ↗
Figure 3
Figure 3. The plots show the zeros (open triangles ∆) and poles (solid circles •) of the type [18/19] multipoint Pad´e approximant for the interpolation problem (47), with the input interpolation nodes ai placed on the positive imaginary axis, as indicated by the open circles ◦. The first spurious poles appear with noise level ϵ ≈ 10−61, and note that as the noise level increases the Pad´e poles and zeros migrate to the vicin… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The plots show the zeros (open triangles ∆) and poles (solid circles •) of the type [19/20] multipoint Pad´e approximant for the interpolation problem (47) with the input interpolation nodes ai placed symmetrically on the positive and negative imag￾inary axis, as indic…
Figure 5
Figure 5. Figure 5: The plots show the zeros (open triangles ∆) and poles (solid circles •) of the type [19/20] multipoint Pad´e approximant for the interpolation problem (47) with the input interpolation nodes ai placed symmetrically around a circle of radius 3/2, as in￾dicated by the op…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 1 canonical work pages

  1. [9]

    Dunne, and Max Meynig,Noise effects on Pad´ e approximants and conformal maps, J

    Ovidiu Costin, Gerald V. Dunne, and Max Meynig,Noise effects on Pad´ e approximants and conformal maps, J. Phys. A55(2022), no. 46, 464007, available at2208.02410

  2. [1]

    N. I. Akhiezer,Theory of approximation, Frederick Ungar Publishing Co., New York, 1956 (eng)

  3. [2]

    Christopher Aubin, Thomas Blum, Maarten Golterman, and Santiago Peris,Pad´ e approxi- mants and g-2 for the muon, Proceedings of ScienceLA TTICE2012(2012), 176, available at1210.7611

  4. [3]

    Baker and P

    George A. Baker and P. R. Graves-Morris,Pad´ e approximants, 2nd ed., Encyclopedia of mathematics and its applications ; v. 59, Cambridge University Press, Cambridge [England], 1996 (eng)

  5. [4]

    2, 187– 211

    Laurent Baratchart and Maxim Yattselev,Multipoint pad´ e approximants to complex cauchy transforms with polar singularities, Journal of Approximation Theory156(2009), no. 2, 187– 211. CHEBYSHEV FUNCTIONS, MULTIPOINT PAD ´E APPROXIMANTS, AND NOISE 19 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -1 0 1 2 3 (a) -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -1 0 1 2 3 (b) -1.5 -1.0 -0.5 ...

  6. [5]

    Jay, and Patrick R

    Thomas Bergamaschi, William I. Jay, and Patrick R. Oare,Hadronic structure, conformal maps, and analytic continuation, Phys. Rev. D108(2023Oct), 074516

  7. [6]

    1, 85–88

    Daniel Bessis,Pad´ e approximations in noise filtering, Journal of Computational and Applied Mathematics66(1996), no. 1, 85–88. Proceedings of the Sixth International Congress on Computational and Applied Mathematics

  8. [7]

    36, 365202

    Daniel Bessis and Luca Perotti,Universal analytic properties of noise: introducing the J- matrix formalism, Journal of Physics A: Mathematical and Theoretical42(August 2009), no. 36, 365202

Show all 19 references
  1. [8]

    3, 501–519, available athttps://londmathsoc.onlinelibrary.wiley.com/ doi/pdf/10.1112/jlms/50.3.501

    Peter Borwein, Tam´ as Erd´ elyi, and John Zhang,Chebyshev Polynomials and Markov- Bernstein Type Inequalities for Rational Spaces, Journal of the London Mathematical Soci- ety50(1994), no. 3, 501–519, available athttps://londmathsoc.onlinelibrary.wiley.com/ doi/pdf/10.1112/jl...

  2. [10]

    Maxim Derevyagin,The Jacobi matrices approach to Nevanlinna-Pick problems, J. Approx. Theory163(2011), no. 2, 117–142. MR2754986

  3. [11]

    ,A note on Wall’s modification of the Schur algorithm and linear pencils of Jacobi matrices, J. Approx. Theory221(2017), 1–21. 20 M. DEREVYAGIN AND M. MEYNIG -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -4 -2 0 2 4 (a) -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -4 -2 0 2 4 (b) -1.5 -1.0 -0.5 0.0 0.5 1....

  4. [12]

    Derevyagin and Alexei S

    Maxim S. Derevyagin and Alexei S. Zhedanov,An operator approach to multipoint Pad´ e approximations, J. Approx. Theory157(2009), no. 1, 70–88

  5. [13]

    2, 199–214

    Jacek Gilewicz and Maciej Pindor,Pad´ e approximants and noise: A case of geometric series, Journal of Computational and Applied Mathematics87(1997), no. 2, 199–214

  6. [14]

    1, 285–297

    ,Pad´ e approximants and noise: rational functions, Journal of Computational and Applied Mathematics105(1999), no. 1, 285–297

  7. [15]

    A. A. Gonˇ car and Giermo Lopes L.,Markov’s theorem for multipoint Pad´ e approximants, Mat. Sb. (N.S.)105(147)(1978), no. 4, 512–524, 639

  8. [16]

    Mourad E. H. Ismail and David R. Masson,Generalized orthogonality and continued fractions, J. Approx. Theory83(1995), no. 1, 1–40

  9. [17]

    1, 221–240 (eng)

    Herbert Stahl,On the convergence of generalized Pad´ e approximants, Constructive approxi- mation5(1989), no. 1, 221–240 (eng)

  10. [18]

    Alexei Zhedanov,Explicit multipoint rational interpolation Pad´ e table for exponential and power functions, Group theory and numerical analysis, 2004, pp. 285–298. Workshop on Group Theory and Numerical Analysis, CRM, Montr´ eal, May 26–31, 2003

  11. [19]

    Zhedanov,Pad´ e interpolation table and biorthogonal rational functions, Elliptic integrable systems, 2004, pp

    Alexei S. Zhedanov,Pad´ e interpolation table and biorthogonal rational functions, Elliptic integrable systems, 2004, pp. 323–363. CHEBYSHEV FUNCTIONS, MULTIPOINT PAD ´E APPROXIMANTS, AND NOISE 21 -2 -1 0 1 2 -2 -1 0 1 2 (a) -2 -1 0 1 2 -2 -1 0 1 2 (b) -2 -1 0 1 2 -2 -1 0 1 2 ...

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.