Pith. sign in

REVIEW 2 major objections 3 minor 35 references

On Siegel's problem and Dwork's conjecture for $G$-functions

T0 review · 2 major / 3 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Some G-functions of order 2 cannot be written as polynomials in algebraic pullbacks of hypergeometric functions.

desk verdict A strong, credible negative answer to Siegel's problem for G-functions, with a genuine new mechanism and an infinite family; the one real soft spot is a terse unproved finiteness claim in the infinitude proof. read the letter →

arxiv 2502.02147 v1 pith:F6HL3IBW submitted 2025-02-04 math.NT math.AGmath.CA

Javier Fresán , Yeuk Hay Joshua Lam , Yichen Qin This is my paper · ORCID
classification math.NTmath.AGmath.CA MSC 11J9134M3514D0711G18
keywords G-functionshypergeometricfunctionsSiegel'sproblemDwork'sconjecturelocalsystemsadjointtracefieldShimuracurvesTannakiancategories
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

G-functions are power series with algebraic coefficients that satisfy a linear differential equation and whose denominators grow at most exponentially. Siegel's problem, as formulated in [FR22], asks whether every G-function can be written as a polynomial expression in functions of the form $\mu(z)\cdot {}_{p+1}F_p[\mathbf{a};\mathbf{b}|\lambda(z)]$ with $\lambda$ algebraic and $\lambda(0)=0$. The paper answers this in the negative: it constructs G-functions of differential order 2, the smallest possible order, that cannot be expressed this way. It also shows that infinitely many non-equivalent such G-functions exist, even up to reparametrizations of the projective line and rank-one twists, answering a question raised in [Kra96] and providing new counterexamples to Dwork's conjecture.

What carries the argument

The load-bearing machinery is a Tannakian reformulation. The category $\mathbf{G}$ consists of regular-singular connections on $\mathbb{P}^1$ that arise as direct summands of relative de Rham cohomology of smooth proper families over $\mathbb{Q}(z)$. The subcategory $\mathbf{H}$ is generated by connections of the form $\pi_{2*}\pi_1^*H$, where $H$ is a hypergeometric connection and $\pi_1,\pi_2$ are finite Galois covers in a correspondence over $\mathbb{P}^1$, together with finite-monodromy connections. The crucial step is a Lie algebra version of Goursat's lemma from [FJ21], which implies that a simple object of $\mathbf{G}$ with non-commutative simple Lie algebra must, if it lies in $\mathbf{H}$, be Lie-generated by a single pullback of a hypergeometric connection. The Beukers–Heckman computation of differential Galois groups, Theorem 2.4.1, then forces that hypergeometric connection to have rank two and Lie algebra $\mathfrak{sl}_2$. The invariant that ultimately separates $\mathbf{H}$ from $\mathbf{G}$ is the adjoint trace field: by [MR03], it is invariant under passing to finite-index subgroups, and by [Kat90], the adjoint trace field of a hypergeometric connection is contained in a cyclotomic field. Shimura curve local systems have non-abelian adjoint trace fields, giving the obstruction.

What would settle it

A computational search over rational parameters (a1,a2,b1) for rank-two hypergeometric connections that returns an adjoint trace field Q(sqrt(D)) with odd squarefree D>=7 would disprove Proposition 2.3.2 and thereby the key trace-field obstruction used in Theorem 1.1.3; alternatively, exhibiting infinitely many rank-two hypergeometric local systems with the same adjoint trace field K would falsify the finiteness assertion behind Theorem 5.3.1.

Watch

Extended reading notes

Core claim

The central discovery is that the Tannakian category $\mathbf{H}$ generated by hypergeometric connections and their algebraic pullbacks is strictly contained in the category $\mathbf{G}$ of differential modules of geometric origin. The proof's key criterion (Lemma 4.1.3) shows that if a rank-two object of $\mathbf{G}$ with differential Galois group $\mathrm{SL}_2$ lies in $\mathbf{H}$, then it must have the same adjoint trace field as some rank-two hypergeometric connection. However, hypergeometric adjoint trace fields are always contained in cyclotomic fields and, in particular, cannot equal $\mathbb{Q}(\sqrt{D})$ for an odd squarefree integer $D\ge 7$ (Proposition 2.3.2). The paper exhibits rank-two local systems of geometric origin attached to rational Shimura curves whose adjoint trace fields are non-abelian cubic fields, so these cannot lie in $\mathbf{H}$. By Corollary 3.2.7, such local systems give actual G-functions of order 2 that are not expressible in the form allowed by Siegel's problem. Theorem 5.3.1 then produces infinitely many such G-functions using the two-parameter family of local systems of [LL23] together with the André–Pink–Zannier theorem in the cases proved by [RY21].

Load-bearing premise

The infinitude result (Theorem 5.3.1) rests on the unproved assertion that only finitely many rank-two hypergeometric local systems have a given adjoint trace field K; if that finiteness failed, the pigeonhole argument would collapse and only the finitely many counterexamples of Section 4 would remain.

Editorial extensions

If this is right

  • The class of G-functions is strictly larger than the class of polynomial expressions in algebraic pullbacks of hypergeometric functions, even when restricted to differential order 2.
  • Dwork's conjecture, which predicted that every order-2 G-function with infinite monodromy is an algebraic pullback of a hypergeometric function, is false in a strong way: infinitely many non-equivalent counterexamples exist.
  • The explicit series $F(z)=1-\frac{5}{2952}z^2-\frac{889}{726192}z^3-\cdots$, defined by equation (1.1.2), is a concrete G-function of order 2 that cannot be written in the hypergeometric form of Question 1.1.1.
  • The adjoint trace field provides a general certificate of non-hypergeometricity: if a rank-two local system of geometric origin has an adjoint trace field that is non-abelian or equal to $\mathbb{Q}(\sqrt{D})$ for odd squarefree $D\ge 7$, then it cannot be an algebraic pullback of a hypergeometric system.

Reading between the lines

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

  • The infinitude theorem (Theorem 5.3.1) depends on the assertion, stated without proof in Section 5.3, that only finitely many rank-two hypergeometric local systems have a given adjoint trace field $K$; if this finiteness fails, only the finite counterexamples of Section 4 would remain.
  • A similar trace-field obstruction may work for higher rank, since hypergeometric adjoint trace fields are cyclotomic for any rank; the Goursat argument might generalize to rank $n$ with simple differential Galois group, potentially yielding higher-order non-hypergeometric G-functions.
  • One could computationally test the finiteness assertion by enumerating rational parameter pairs $(a_1,a_2,b_1)$ for rank-two hypergeometric systems and checking how many distinct adjoint trace fields occur, which would either support or refute the pigeonhole step behind the infinitude result.
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

2 major / 3 minor

Summary. The paper answers in the negative Siegel's problem for G-functions, as formulated by Fischler and Rivoal. It constructs G-functions of differential order 2 that cannot be written as Q-polynomial expressions in algebraic pullbacks of hypergeometric functions, and it further claims to produce infinitely many such functions up to reparametrization and rank-one twist. The strategy is Tannakian: a category H generated by hypergeometric connections under algebraic correspondences is compared with the category G of geometric origin, and an obstruction is derived from adjoint trace fields. The main technical tools are a Lie-algebra Goursat lemma, Beukers--Heckman monodromy computations, and results on Fuchsian groups. The paper also gives an explicit example and connects the results to Dwork's conjecture and a question of Krammer.

Significance. If the results stand, the paper settles a natural formulation of Siegel's problem for G-functions and provides the first unconditional counterexamples of minimal differential order. The proofs combine several deep ingredients, and the paper is careful to state the external theorems on which it relies. The explicit example in (1.1.2) is a valuable concrete artifact. The main advertised infinitude theorem, however, depends on a currently incorrect or at least unproved finiteness statement in Section 5.3, which must be repaired before the full strength of the paper is established.

major comments (2)
  1. [Section 5.3, proof of Theorem 5.3.1, after Eq. (5.3.1)] The assertion that 'there are only finitely many rank-two hypergeometric local systems with adjoint trace field K' is false as stated. Tensoring any rank-two hypergeometric local system H by a torsion rank-one local system does not change the adjoint representation ad0(H) and therefore does not change the adjoint trace field; such twists again give hypergeometric local systems, so there are infinitely many rank-two hypergeometric local systems with a given adjoint trace field K. The pigeonhole step only needs the weaker statement that there are finitely many possible local systems ad0(H), i.e., finitely many adjoint representations. That weaker statement is plausible from rigidity: the values lambda+lambda^{-1}, mu+mu^{-1}, nu+nu^{-1} in Proposition 2.3.2 must lie in K, and there are finitely many roots of unity whose real part generates a subfield of K. Please replace the assertion by this corrected finiteness statement and supply the proof, and adjust the subsequent phrase 'the same hypergeometric local system' to refer to the same adjoint representation.
  2. [Section 5.3, proof of Theorem 5.3.1, final paragraph] The proof fixes 'an infinite sequence of points p1,p2,... in ~M(Q)\Sexc' without justification. The second assertion of Theorem 5.3.1, namely the existence of infinitely many non-equivalent G-functions, depends on ~M(Q) being infinite outside the finite exceptional set. This is not stated in Proposition 5.1.1 and does not follow formally from the existence of a dominant etale morphism to M0,4, since a finite etale cover of a rational curve can have few rational points. Please add a proof or a precise reference establishing that ~M(Q) contains infinitely many points, or state this as a hypothesis in Theorem 5.3.1.
minor comments (3)
  1. [Proposition 2.3.2, proof] The trace formula '2+lambda+lambda^{-1}' is the trace on the full adjoint representation gl2, whereas the adjoint trace field is defined via ad0; since the two differ by the constant 1, the field generated is the same. Please clarify this distinction in the text.
  2. [Section 5.3, after Eq. (5.3.1)] The sentence 'It follows straightforwardly from the rigidity of hypergeometric connections and Proposition 2.3.1 that there are only finitely many rank-two hypergeometric local systems with adjoint trace field K' should be removed or replaced by the corrected statement, because as written it is not only unproved but false; see the first major comment.
  3. [Remark 1.1.4] The 'Andre--Chunovsky--Katz theorem' is invoked without a reference; the surrounding discussion in [And00, §3] is cited, but a precise pointer would help the reader.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction found; the derivation is self-contained relative to external rigidity, trace-field, and unlikely-intersection inputs, with only minor non-load-bearing self-citations.

full rationale

After walking the derivation from Question 1.1.1 through the Tannakian reformulation (Definitions 3.1.1 and 3.2.1, Corollary 3.2.7), the key criterion (Lemma 4.1.3), the Shimura and Teichmuller curve constructions, and the infinite family argument (Section 5), I find no step in which a claimed prediction is equivalent by construction to its inputs. The category H is defined independently of the non-representability conclusion, and Lemma 4.1.3 derives constraints on any hypergeometric connection that could Lie-generate a rank-two object, using Beukers-Heckman and Maclachlan-Reid, rather than assuming the desired conclusion. Proposition 2.3.2 independently restricts the possible adjoint trace fields of hypergeometric connections, and the Shimura curve counterexamples are obtained from external classifications and constructions (Voight, Krammer, Takeuchi, McMullen). The only self-citations, namely the Lie-algebra Goursat lemma from Fresan-Jossen [FJ21] and the Lam-Litt family [LL23] underlying Proposition 5.1.1, are used as external inputs rather than as substitutes for the main theorems; neither citation asserts or presumes that the examples are non-hypergeometric. The unproved finiteness assertion in Section 5.3 that there are only finitely many rank-two hypergeometric local systems with a given adjoint trace field K is a substantive proof gap that should be filled explicitly, but it is not a circular reduction: a missing justification is not a self-referential input. Thus the paper is not significantly circular, and the minor self-citations do not carry the central claims.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

No free parameters or invented entities. The arguments rely on standard deep theorems from the literature; the Lam-Litt input is self-cited but is an independent theorem with proof, and its use does not make the central claim circular.

assumptions (6)
  • standard math Beukers-Heckman classification of differential Galois groups of hypergeometric connections (Theorem 2.4.1).
    Used to deduce that an irreducible hypergeometric connection Lie-generating SL2 must itself be rank two with Galois group SL2.
  • standard math Katz's rigidity theorem and geometric origin of hypergeometric connections (Theorem 3.5.4 and Theorem 5.4.4 of [Kat90]).
    Used in Proposition 3.2.2 to show H is contained in G and in Proposition 2.3.2 to show trace fields of hypergeometric systems are cyclotomic.
  • standard math Maclachlan-Reid invariant trace field of Fuchsian groups (Lemma 2.2.5).
    Used to make adjoint trace fields invariant under finite-index subgroups, a key step in Lemma 4.1.3.
  • standard math Takeuchi classification of arithmetic triangle groups (Theorem 5.2, [Tak77]).
    Used in Proposition 4.3.4 to rule out hypergeometric monodromy groups commensurable with Shimura-curve monodromy groups over Q for discriminants other than (1) and (2)(3).
  • standard math Richard-Yafaev André-Pink-Zannier theorem for Shimura varieties of abelian type (Theorem 5.2.1).
    Used in Section 5.3 to obtain the infinite family by showing that the alternative would force a Zariski-dense intersection with a generalized Hecke orbit, contradicting the theorem.
  • standard math Lam-Litt construction of geometric local systems on P^1 minus four points (Theorem 1.1.4 and 1.1.6 of [LL23]).
    Provides the family V of rank-two local systems with prescribed trace fields and monodromy; this is a published independent theorem, not a restatement of the present claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On Siegel's problem and Dwork's conjecture for $G$-functions." pith.science (2026). https://pith.science/paper/F6HL3IBW

@misc{pith2026250202147,
  author       = {Pith},
  title        = {Pith review of: On Siegel's problem and Dwork's conjecture for $G$-functions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F6HL3IBW}},
  note         = {Machine review of arXiv:2502.02147}
}
abstract

We answer in the negative Siegel's problem for $G$-functions, as formulated by Fischler and Rivoal. Roughly, we prove that there are $G$-functions that cannot be written as polynomial expressions in algebraic pullbacks of hypergeometric functions; our examples satisfy differential equations of order two, which is the smallest possible. In fact, we construct infinitely many non-equivalent rank-two local systems of geometric origin which are not algebraic pullbacks of hypergeometric local systems, thereby providing further counterexamples to Dwork's conjecture and answering a question by Krammer. The main ingredients of the proof are a Lie algebra version of Goursat's lemma, the monodromy computations of hypergeometric local systems due to Beukers and Heckman, as well as results on invariant trace fields of Fuchsian groups.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 33 canonical work pages

  1. [1]

    G -functions and geometry , volume E13 of Aspects of Mathematics

    Yves Andr\' e . G -functions and geometry , volume E13 of Aspects of Mathematics . Friedr. Vieweg & Sohn, Braunschweig, 1989

    work page 1989

  2. [2]

    S\' e ries G evrey de type arithm\' e tique

    Yves Andr\' e . S\' e ries G evrey de type arithm\' e tique. I . T h\' e or\`emes de puret\' e et de dualit\' e . Ann. of Math. , 151(2):705--740, 2000

    work page 2000

  3. [3]

    G -functions, motives, and unlikely intersections---old and new

    Yves Andr\' e . G -functions, motives, and unlikely intersections---old and new. Preprint arXiv:2501.09867, 2025

    arXiv 2025

  4. [4]

    Monodromy for the hypergeometric function _nF_ n-1

    Frits Beukers and Gert Heckman. Monodromy for the hypergeometric function _nF_ n-1 . Invent. math. , 95(2):325--354, 1989

    work page 1989

  5. [5]

    Bouw and Martin M \"o ller

    Irene I. Bouw and Martin M \"o ller. Differential equations associated with nonarithmetic Fuchsian groups. J. Lond. Math. Soc. , 81(1):65--90, 2010

    work page 2010

  6. [6]

    \'E l \'e ments de math \'e matique

    Nicolas Bourbaki. \'E l \'e ments de math \'e matique. Alg \`e bre . Chapitre 8. Modules et anneaux semi-simples. Berlin: Springer, 2nd revised ed. of the 1958 original edition, 2012

    work page 1958

  7. [7]

    Fields of definition

    Frank Calegari. Fields of definition. Blog post, https://www.galoisrepresentations.com/2024/12/07/fields-of-definition/, December 2024. Accessed: 2025-01-21

    work page 2024

  8. [8]

    On the classification of rank-two representations of quasiprojective fundamental groups

    Kevin Corlette and Carlos Simpson. On the classification of rank-two representations of quasiprojective fundamental groups. Compos. Math. , 144(5):1271--1331, 2008

    work page 2008

Show all 35 references
  1. [9]

    Construction et classification de certaines solutions alg \'e briques des syst \`e mes de G arnier

    Karamoko Diarra. Construction et classification de certaines solutions alg \'e briques des syst \`e mes de G arnier. Bull. Braz. Math. Soc. (N.S.) , 44(1):129--154, 2013

    2013

  2. [10]

    Representability of G -functions as rational functions in hypergeometric series

    Thomas Dreyfus and Tanguy Rivoal. Representability of G -functions as rational functions in hypergeometric series. Bull. Sci. math. , 199:103542, 2025

    2025

  3. [11]

    The early history of the hypergeometric function

    Jacques Dutka. The early history of the hypergeometric function. Arch. Hist. Exact Sci. , 31:15--34, 1984

    1984

  4. [12]

    Differential operators with nilpotent p-curvature

    Bernard Dwork. Differential operators with nilpotent p-curvature. Amer. J. Math. , 112(5):749--786, 1990

    1990

  5. [13]

    A non-hypergeometric E -function

    Javier Fres \'a n and Peter Jossen. A non-hypergeometric E -function. Ann. of Math. , 194(3):903--942, 2021

    2021

  6. [14]

    Exponential motives

    Javier Fres \'a n and Peter Jossen. Exponential motives. Available at http://javier.fresan.perso.math.cnrs.fr/expmot.pdf, 2024

    2024

  7. [15]

    On S iegel's problem for E -functions

    St\' e phane Fischler and Tanguy Rivoal. On S iegel's problem for E -functions. Rend. Semin. Mat. Univ. Padova , 148:83--115, 2022

    2022

  8. [16]

    Nicholas M. Katz. On the calculation of some differential Galois groups . Invent. math. , 1(87):13--61, 1987

    1987

  9. [17]

    Nicholas M. Katz. Exponential sums and differential equations , volume 124 of Annals of Mathematics Studies . Princeton University Press, 1990

    1990

  10. [18]

    Nicholas M. Katz. Rigid local systems , volume 139 of Annals of Mathematics Studies . Princeton University Press, 1996

    1996

  11. [19]

    Hodge loci and atypical intersections: conjectures

    Bruno Klingler. Hodge loci and atypical intersections: conjectures. Proceedings Motives and Complex Multiplication, to appear, 2017

    2017

  12. [20]

    An example of an arithmetic Fuchsian group

    Daan Krammer. An example of an arithmetic Fuchsian group. J. reine angew. Math. , 473:69--86, 1996

    1996

  13. [21]

    Geometric local systems on the projective line minus four points

    Yeuk Hay Joshua Lam and Daniel Litt. Geometric local systems on the projective line minus four points. Compos. Math. , to appear, 2023

    2023

  14. [22]

    Canonical representations of surface groups

    Aaron Landesman and Daniel Litt. Canonical representations of surface groups. Ann. of Math. , 199(2):823--897, 2024

    2024

  15. [23]

    Finite braid group orbits on SL_2 -character varieties

    Yeuk Hay Joshua Lam, Aaron Landesman, and Daniel Litt. Finite braid group orbits on SL_2 -character varieties. Preprint, arXiv 2308.01376 , 2023

    2023 arXiv

  16. [24]

    Billiards and T eichm \"u ller curves on H ilbert modular surfaces

    Curtis McMullen. Billiards and T eichm \"u ller curves on H ilbert modular surfaces. J. Am. Math. Soc. , 16(4):857--885, 2003

    2003

  17. [25]

    Canonical models of Shimura curves

    James S Milne. Canonical models of Shimura curves . Notes available at https://www.jmilne.org/math/articles/2003a.pdf, 2003

    2003

  18. [26]

    o ller. Variations of H odge structures of a T eichm \

    Martin M \"o ller. Variations of H odge structures of a T eichm \"u ller curve . J. Amer. Math. Soc. , 19(2):327--344, 2006

    2006

  19. [27]

    Colin Maclachlan and Alan W. Reid. The arithmetic of hyperbolic 3-manifolds , volume 219 of Graduate Texts in Mathematics . Springer-Verlag, 2003

    2003

  20. [28]

    Generalised A ndr \'e -- P ink-- Z annier conjecture for S himura varieties of abelian type

    Rodolphe Richard and Andrei Yafaev. Generalised A ndr \'e -- P ink-- Z annier conjecture for S himura varieties of abelian type. Preprint, arXiv:2111.11216, 2021

    2021 arXiv

  21. [29]

    Height functions on Hecke orbits and the generalised Andr\'e--Pink--Zannier conjecture

    Rodolphe Richard and Andrei Yafaev. Height functions on Hecke orbits and the generalised Andr\'e--Pink--Zannier conjecture. Compos. Math. , 160(11):2531--2584, 2024

    2024

  22. [30]

    \"U ber einige Anwendungen diophantischer Approximationen

    Carl Ludwig Siegel. \"U ber einige Anwendungen diophantischer Approximationen . A bhandlungen der P reu ischen A kademie der W issenschaften. P hysikalisch-mathematische K lasse , 1, 1929. Reprinted in Gesammelte Abhandlungen I, 209--266

    1929

  23. [31]

    Commensurability classes of arithmetic triangle groups

    Kisao Takeuchi. Commensurability classes of arithmetic triangle groups. J. Fac. Sci. Univ. Tokyo Sect. IA Math , 24(1):201--212, 1977

    1977

  24. [32]

    A hyperelliptic saga on a generating function of the squares of Legendre polynomials

    Mark van Hoeij, Duco van Straten, and Wadim Zudilin. A hyperelliptic saga on a generating function of the squares of Legendre polynomials . Journal of Experimental Mathematics, to appear , 2023

    2023

  25. [33]

    Shimura curves of genus at most two

    John Voight. Shimura curves of genus at most two. Math. Comp. , 78(266):1155--1172, 2009

    2009

  26. [34]

    Quaternion algebras

    John Voight. Quaternion algebras . Springer Nature, 2021

    2021

  27. [35]

    Arithmetica Infinitorum , volume I

    John Wallis. Arithmetica Infinitorum , volume I. Opera Mathematica, 1656

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.