REVIEW 4 major objections 4 minor 12 references
Half-Iterates and Delta Conjectures
T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Exact constant bridges two algorithms for Abel's equation
desk verdict A transparent, clearly labeled conjecture note whose delta values all rest on unverified asymptotic expansions; the general formula is new but the numerical evidence is thin. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the additive offset $\delta$ between the two algorithm outputs, defined by $\delta = g(x)-\tilde{g}(x)$ for any admissible $x$. The argument uses Abel's equation $g(\theta(x))=g(x)+1$ and Julia's equation $\lambda(\theta(x))=\theta'(x)\lambda(x)$, whose reciprocal approximates $g'$; the ML value $\tilde{g}(x)$ comes from an asymptotic expansion of the iterates $x_n=\theta^n(x)$, with a logarithmic correction term such as $-\frac{3}{8}\frac{\ln n}{n^{3/2}}$ for the cubic family. The conjectures are derived experimentally by matching the constant difference at carefully chosen points.
What would settle it
Compute both $g_{13}(x)$ and $\tilde{g}_{13}(x)$ to high precision at a third point, for example $x=1/4$, and test whether $g_{13}(1/4)-\tilde{g}_{13}(1/4)$ equals $-(3/4)\ln(2)$; any deviation beyond the numerical precision would disprove the independence of $x$. Similarly, for a parameter triple $(\tau,\sigma,\rho)$ not among those tested, the predicted formula can be checked against a fresh high-precision computation.
Extended reading notes
Core claim
The paper's central conjecture is that the difference $\delta$ between the EJ solution and the ML principal solution of Abel's equation is a constant, independent of the starting point $x$, with an exact formula determined by the first two nonlinear coefficients of $\theta$. For the cubic $\theta_{13}(x)=x(1-x^2)$, it asserts $\delta_{13}=g_{13}(1/2)-\tilde{g}_{13}(1/2)=g_{13}(1/\sqrt{3})-\tilde{g}_{13}(1/\sqrt{3})=-(3/4)\ln(2)$. For the general polynomial family $\theta(x)=x(1-\sigma x^\tau-\rho x^{2\tau})$, Addendum III conjectures $\delta=((\tau+1)/(2\tau)+\rho/(\tau\sigma^2))\ln(1/(\tau\sigma))$. The paper offers high-precision numerical evidence at two test points per map, and it shows that the same formula applies to several transcendental maps such as the Fresnel cosine.
Load-bearing premise
The whole delta conjecture rests on the asymptotic expansions that define the ML limit, such as the expansion of $\sqrt{2}x_n$ with its $\ln(n)/n^{3/2}$ term; these expansions are cited from an earlier paper and are not re-derived here.
Editorial extensions
If this is right
- If the delta conjecture holds, the faster EJ algorithm, corrected by the conjectured $\delta$, yields the principal solution without evaluating the slower ML limit.
- The conjectured formulas extend to transcendental maps like the Fresnel cosine, giving exact offsets for a wider class than polynomials.
- The observation that $\delta$ depends only on $\tau$ and the first two coefficients $c_1,c_2$ would mean higher-order Taylor coefficients are irrelevant to the correction.
- For maps like $\theta_{14}(x)=x(1-x+x^2)$, the conjecture $\delta_{14}=0$ implies that EJ already gives the principal solution, a useful simplification.
- Knowing $\delta$ makes possible an explicit half-iterate $\theta^{[1/2]}(x)=g^{-1}(g(x)+1/2)$ using the corrected EJ function.
Reading between the lines
- One could test whether the delta formula extends to non-polynomial maps with the same first two nonlinear coefficients, predicting e.g. that $\sin(x)$ and its cubic truncation share the same $\delta$.
- If the conjectured dependence on only $(\tau,c_1,c_2)$ is true, it suggests a renormalization-type universality: the principal solution's offset is a fixed function of three 'effective' parameters.
- A proof might be attempted by showing that the difference $g-\tilde{g}$ is invariant under iteration and hence constant, using the asymptotic expansions to evaluate it at the fixed point.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies two algorithms for solving Abel's equation g(θ(x)) = g(x) + 1 for maps θ with θ(0)=0, θ'(0)=1: the Écalle-Jagy (EJ) series method and the Mavecha-Laohakosol (ML) limit method. It conjectures exact values for the additive offset δ between the EJ solution and the 'principal' solution computed by ML, for several cubic and related maps. The central example is θ13(x)=x(1-x²), for which the paper claims δ13 = g13(1/2) − \tilde{g}13(1/2) = g13(1/√3) − \tilde{g}13(1/√3) = −(3/4)ln2. More generally, Addendum III conjectures δ = ((τ+1)/(2τ) + ρ/(τσ²)) ln(1/(τσ)) for θ(x)=x(1−σx^τ−ρx^{2τ}). The evidence is high-precision numerical matching at two x-values per map, and no proofs are given. The paper also discloses that a software error invalidated all earlier EJ results for g14, g15, and g16 in a previous draft.
Significance. If the conjectures are correct, they give an exact closed-form correction that would allow the faster EJ algorithm to reproduce the principal solution, effectively subsuming the ML method's 'intrinsicality' advantage. The conjectures are crisp, falsifiable, and potentially useful for future work on iterates and Abel equations. The paper is honest in disclosing a prior software error and in labeling its statements as conjectures. However, the results are entirely numerical, rest on unproved asymptotic expansions, and are tested at only two points per map; the paper does not provide reproducible code or machine-checked proofs. As a conjectural research note the paper has value, but as a finished journal article it needs substantial strengthening of its evidence base.
major comments (4)
- [Section 1, definition of \tilde{g}13(x)] The defining limit for \tilde{g}13(x) uses the asymptotic expansion √2 x_n = n^{-1/2} − (3/8) ln(n) n^{-3/2} + o(n^{-3/2}), cited from reference [9] without derivation. Every δ value in the paper is the difference between an EJ-based value and an ML limit built from such an expansion; if the coefficient 3/8, or the analogous coefficient (τ+1)/(2τ) in Addendum III, were even slightly wrong, the expression inside the limit would diverge logarithmically and all δ conjectures would shift. This is a load-bearing input, not a presentation detail. The paper should either derive these expansions from the iteration of θ, or provide strong independent numerical verification, such as a convergence study showing that the limit stabilizes to the claimed precision for several n and that the resulting \tilde{g} values are consistent across multiple starting points.
- [Section 1, 'Clarification'] The disclosed software error invalidated every EJ-based result about g14, g15, and g16 in an earlier draft. The current manuscript presents corrected-looking EJ series and δ values for these maps, but gives no details of the correction, no independent verification, and no code (the EJ notebook is 'forthcoming'). Given that a software bug already corrupted results for three of the main examples, the reader cannot currently distinguish reliable numerical output from another subtle implementation error. The authors should provide a reproducible implementation or an independent cross-check, for example comparing the EJ series at several x-values with high-precision direct iteration and with the ML limits at additional points.
- [Sections on δ13, δ14, δ15, δ16] The central conjectures assert that δ is a constant independent of the initial x, but the evidence for this is only two test points per map (e.g., x=1/2 and x=1/√3 for θ13). The paper states that EJ finds g(x)+δ with δ independent of x; if this is a theorem about Abel equation solutions, it should be stated and cited (or proved), since all solutions of g(θ(x))=g(x)+1 on the relevant interval need not obviously differ by a constant. If it is an empirical claim, two points are far too few to support a global constant. Because the exact value δ13 = −(3/4)ln2 is meaningful only if the same δ holds for every x, this point needs to be established rigorously or tested much more extensively.
- [Addendum II, 'In essence'] The far-reaching claim that δ is completely determined by τ and the two coefficients c1 and c2 is inferred from a small set of examples with matching coefficients. This is a strong generalization that goes well beyond any derivation in the paper, and the listed pairs are exactly the data used to formulate it. To avoid overfitting, the conjecture should be stated with precise definitions (which coefficients are meant, what happens when c3 is nonzero) and tested on at least one map that was not used in its formulation, ideally with a different τ and nonzero c2. Without such a test, the 'completely determined' assertion is not yet supported.
minor comments (4)
- [General] The manuscript does not number its equations or display equations, which makes it difficult to refer precisely to the many limit formulas and conjectures; adding numbered displays would improve readability.
- [Section 1, footnote/software error] The phrase 'inexplicably, over time' in the clarification of the software error is not needed and could be replaced by a neutral description of how the indexing error arose.
- [Addendum II] The sentence 'In essence, δ seems to be completely determined...' is informal; if this is intended as a conjecture, it should be labeled as such and stated separately from the examples.
- [References] The citation of the author's own previous arXiv versions [7] and [8] as sources for algorithms may be confusing because the current paper is itself arXiv:2506.07625; clarifying the version history would help the reader locate the cited material.
Circularity Check
No significant circularity: the delta conjectures are empirical fits to independently computed numerical limits, not derivations that presuppose their own conclusions.
full rationale
The paper's central claims are conjectured identities for the additive offset δ between the EJ and ML solutions of Abel's equation. These identities are inferred from high-precision numerical values obtained by two distinct algorithms: EJ solves Julia's equation via formal series, while ML evaluates an asymptotic limit for iterates. The delta value is then the observed difference between these independently computed numbers, e.g. δ13 = g13(1/2) − g̃13(1/2) = −(3/4)ln(2). This is a numerical observation, not a deduction in which the conclusion is assumed in the premises. The asymptotic expansions defining g̃(x) are cited from the author's prior work [9], but they are parameter-free statements about iterates that can be checked independently and do not themselves presuppose the delta conjectures. The references [7] and [8] are earlier versions of the same paper and provide additional observed examples, but the main conjecture for each family is computed directly here rather than taken as a premise. The paper honestly labels its results as conjectures and states 'Formal proofs would be good to see someday.' The disclosed software error in Section 1 affected an earlier draft, not the current computations, and is a correctness concern rather than a circularity. The main vulnerability is the reliance on unproved asymptotic expansions from [9]; if any log coefficient were wrong, the delta values would shift. That is a correctness risk, not a circularity, because the expansions are not defined in terms of the delta values and are externally checkable. Overall, the paper is an experimental numerical study whose conjectures are fitted to data rather than circularly derived; no step reduces, by construction or self-citation, to its own input.
Assumptions & free parameters
assumptions (3)
- domain assumption The Abel equation g(θ(x))=g(x)+1 has a unique principal solution (up to additive constant) for θ analytic with θ(0)=0, θ'(0)=1 and the stated Taylor form.
- domain assumption The asymptotic expansion of the iterates x_n used to define \tilde{g}(x) is valid to the required order.
- domain assumption The EJ algorithm yields g(x)+δ with δ independent of x, while ML yields the principal solution g(x).
Cite this review
Pith. "Pith review of Half-Iterates and Delta Conjectures." pith.science (2026). https://pith.science/paper/JOKQTOUS
@misc{pith2026250607625,
author = {Pith},
title = {Pith review of: Half-Iterates and Delta Conjectures},
year = {2026},
howpublished = {\url{https://pith.science/paper/JOKQTOUS}},
note = {Machine review of arXiv:2506.07625}
}
abstract
The vivid contrast between two competing algorithms for solving Abel's equation $g(\theta(x)) = g(x) + 1$, given $\theta(x)$, is easily sketched. EJ is faster and more efficient, but ML evaluates a limit characterizing the principal solution $g(x)$ directly. EJ finds $g(x)+\delta$, where $\delta$ is possibly nonzero but independent of $x$. If we were to know an exact expression for $\delta$, then the "intrinsicality" of ML would be subsumed by EJ. Filling this gap in our knowledge is the aim of this paper.
Figures
Reference graph
Works this paper leans on
-
[9]
S. R. Finch, Generalized logistic maps and convergence, arXiv:240 9.15175
-
[1]
N. G. de Bruijn, Asymptotic Methods in Analysis , North-Holland, 1958; MR0099564 / Dover, 1981; MR0671583
work page 1958
-
[2]
F. Bencherif and G. Robin, Sur l’it´ er´ e de sin( x), Publ. Inst. Math. (Beograd) 56(70) (1994) 41–53; MR1349068; http://eudml.org/doc/25612 2
work page 1994
-
[3]
S. Mavecha and V. Laohakosol, Asymptotic expansions of iterat es of some classical functions, Appl. Math. E-Notes 13 (2013) 77–91; MR3121616; http://www.emis.de/journals/AMEN/2013/2013.htm
work page 2013
-
[4]
S. R. Finch, What do sin( x) and arcsinh( x) have in common? arXiv:2411.01591
-
[5]
´Ecalle, Th´ eorie it´ erative: introduction ` a la th´ eorie des invariants holomorphes, J
J. ´Ecalle, Th´ eorie it´ erative: introduction ` a la th´ eorie des invariants holomorphes, J. Math. Pures Appl. 54 (1975) 183–258; MR0499882
work page 1975
- [6]
-
[8]
S. R. Finch, Half-iterates of x exp(x), x + 1/x and arcsinh( x), arXiv:2506.07625v2
Show all 12 references
-
[10]
S. R. Finch, Exercises in iterational asymptotics III, arXiv:250 3.13378
-
[11]
Jagy, Half iterate of x2 + c, http://math.stackexchange.com/questions/208996/half-iterate-of-x2c (2012)
W. Jagy, Half iterate of x2 + c, http://math.stackexchange.com/questions/208996/half-iterate-of-x2c (2012)
2012
-
[12]
Kouznetsov, Super sin, Far East J
D. Kouznetsov, Super sin, Far East J. Math. Sci. 85 (2014) 219–238; http://mizugadro.mydns.jp/PAPERS/
2014
-
[13]
S. R. Finch, Iterational asymptotics of sine and cosine, http://community.wolfram.com/groups/-/m/t/3362137. Steven Finch MIT Sloan School of Management Cambridge, MA, USA steven finch math@outlook.com
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.