Transformation formula of Dwork's $p$-adic hypergeometric function

Yusuke Nemoto

REVIEW 4 major objections 4 minor 16 references

Transformation formula of Dwork's $p$-adic hypergeometric function

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

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

Pith's one-line read This paper proves an inversion formula for Dwork's p-adic hypergeometric function: replacing $t$ by $t^{-1}$ multiplies the function by the explicit factor $((-1)^{d+1}t)^l$, with a stated sign rule for $p=2$.

desk verdict Genuine extension of Wang's transformation formula to all d, but the central derivation has an exponent off by a factor of N; as written, Theorem 1.2 does not follow. read the letter →

arxiv 2505.24215 v1 pith:AZJ6XPDC submitted 2025-05-30 math.NT

Yusuke Nemoto This is my paper · ORCID

classification math.NT MSC 33C2033E50

keywords Dwork'sp-adichypergeometricfunctiontransformationformulaFrobeniusactiondeRhamcohomologyschemefinitefieldfunctionsunitrootTatealgebra

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

Dwork's p-adic hypergeometric function $F^{\mathrm{Dw}}_{a,\ldots,a}(t)$ is a $p$-adic analytic function built from the classical hypergeometric series by dividing by its $p$-Frobenius transform. This paper proves a transformation formula under inversion of the variable: replacing $t$ by $t^{-1}$ multiplies the function by an explicit power of $(-1)^{d+1}t$, with the exponent determined by the parameter $a$ and the prime $p$. The result holds for every dimension $d \ge 1$ when $a \in N^{-1}\mathbb{Z}$, $p \nmid N$, and $0

What carries the argument

The carrying object is the hypergeometric scheme $U_t: (1-x_0^N)\cdots(1-x_d^N)=t$ and the quotient $V^*_t: w^N=z_1\cdots z_d(1-z_1)^{N-1}\cdots(1-z_d)^{N-1}(z_1\cdots z_d-t)^{N-1}$, $w\ne0$. The decisive input is the Frobenius congruence stated as Theorem 2.2: for indices $p j_k\equiv i_k\pmod N$, the Frobenius pullback satisfies $\Phi(\omega_{j_0,\ldots,j_d})\equiv p^d F^{\mathrm{Dw}}_{a}(t)^{-1}\omega_{i_0,\ldots,i_d}$ modulo the unit-root submodule. Under the explicit inversion map $\iota:(z_i,w)\mapsto(z_i^{-1},v)$, the holomorphic differential $\omega_n$ is pulled back to a scalar multiple $(-1)^{(d+1)n-1}\xi^{-\varphi_{N,d}(n)}t_0^{N-n}\omega_n$; comparing the Frobenius eigenvalue obtained from Theorem 2.2 on $U$ and on the inverted scheme $\widehat U$ yields the power of $t$ and the sign in the formula.

What would settle it

Take $p=3$, $d=1$, $N=2$, $a=1/2$; here $l=1$ and $(-1)^{d+1}=1$, so Theorem 1.2 predicts $F^{\mathrm{Dw}}_{1/2,1/2}(t)=t\,F^{\mathrm{Dw}}_{1/2,1/2}(t^{-1})$ in $\mathbb{Z}_3\langle t,t^{-1},h(t)^{-1}\rangle$. Computing both sides as power series in $t$ to order $t^6$ with $p$-adic precision $2$ would either confirm the identity to that order or locate the first coefficient where it fails; a failure would refute the theorem, while agreement to arbitrarily high order would support it.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 1.2: under the conditions $a\in N^{-1}\mathbb{Z}$, $p\nmid N$, and $0<a<1$, the identity $F^{\mathrm{Dw}}_{a,\ldots,a}(t)=((-1)^{d+1}t)^l F^{\mathrm{Dw}}_{a,\ldots,a}(t^{-1})$ holds in $\mathbb{Z}_p\langle t,t^{-1}, h_{a,\ldots,a}(t)^{-1}\rangle$ for odd $p$, where $l\in\{0,\ldots,p-1\}$ is determined by $a+l\equiv0\pmod p$. For $p=2$ the same identity holds up to sign, and the extra minus sign appears exactly when $d$ is even and the Dwork prime $a'$ satisfies $a'\equiv1\pmod2$. The proof constructs an isomorphism between the scheme $V^*_{t^N}$ and its inverted analogue $V^*_{t^{-N}}$ and compares the eigenvalues of the $p$-power Frobenius on their $d$-th de Rham cohomology groups; the two eigenvalue descriptions are expressed in terms of $F^{\mathrm{Dw}}$ at $t$ and at $t^{-1}$, forcing the transformation.

Load-bearing premise

The proof rests on a cited theorem (Theorem 2.2) that gives a congruence for the Frobenius action on the differential forms of the hypergeometric scheme; if that congruence fails for the completed coefficient rings or for the inverted variable, the transformation formula would not follow.

Editorial extensions

If this is right

Conjecture 1.1 is now proven for all $d\ge1$ in the rational-parameter range, extending the known $d=0$ case and the conditional $d=1$ case.
Corollary 3.2 extends the inversion formula to the iterated function $F^{\mathrm{Dw},f}_{a,\ldots,a}(t)$, with the exponent $l$ taken modulo $q=p^f$.
For $p=2$, the transformation is no longer ambiguous: the extra minus sign appears exactly when $d$ is even and $a'\equiv1\pmod2$.
The finite-field analogue in Appendix A gives a $t\leftrightarrow t^{-1}$ transformation for hypergeometric functions over $\mathbb{F}_q$; the paper shows it implies a special-value relation for Dwork's function at roots of unity.
The formula yields a functional equation for the relevant Frobenius traces on hypergeometric varieties at reciprocal arguments, usable in computations of zeta functions.

Reading between the lines

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

An extension the paper does not pursue: remove the rationality assumption $a\in N^{-1}\mathbb{Z}$. The proof needs $N$ to write the scheme equations and the $N$-th root $\xi$, so a deformation or approximation argument from rational parameters to arbitrary $a\in\mathbb{Z}_p$ may establish the full conjecture.
The same Frobenius-eigenvalue comparison could be run on non-diagonal parameter tuples $(a_0,\ldots,a_d)$ whenever the relevant cohomology class is one-dimensional, yielding inversion formulas beyond the equal-parameter case.
Read backward, the finite-field appendix hints at a Gauss-sum identity underlying the $p$-adic sign; tracing the power of $t$ and the $p=2$ sign through that identity could yield explicit congruences for truncated hypergeometric sums.
For $p=2$, the parity condition $a'\equiv1\pmod2$ suggests a cohomological explanation of the extra sign; checking whether the same condition governs unit-root zeta functions of the hypergeometric curve would test that explanation.

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.

Desk Editor's Note

read the letter

Main take: this paper is a real advance in p-adic hypergeometric transformations, but the proof as written has a load-bearing exponent error. The good news is that the error looks repairable.

The result itself is what the abstract promises: for a rational a with denominator N (p ∤ N), Conjecture 1.1 holds for all d ≥ 1, plus a p = 2 sign criterion. That goes beyond Wang's d=0 and d=1 under the same hypotheses, and the cohomological mechanism — comparing Frobenius eigenvalues through the isomorphism ι — is a genuine new application of Asakura's Theorem 2.2. The finite-field appendix (Prop. A.1) is also new and the specialization to equal parameters is a clean finite analogue. I credit the paper for the explicit sign analysis, which is careful and consistent.

The soft spot is in the heart of the proof of Theorem 1.2. The author sets t = t0^N, and after applying ι^* to (3.6) obtains, up to signs and constants, t0^{p(N-m)} F(t0^N)^{-1} = t0^{N-n} F(t0^{-N})^{-1}. That gives F(t0^N)/F(t0^{-N}) proportional to t0^{(p-1)N - (pm-n)} = t0^{Nl}, with l = p a_m - a_n. The paper displays t0^{p a_m - a_n} = t0^l instead. Since the theorem's right-hand side is ((-1)^{d+1} t)^l with t = t0^N, the required power of t0 is Nl. The displayed formula is off by a factor of N in the exponent. For N=1 it would be fine, but N is arbitrary. This is an internal inconsistency, not a failure of the cited theorem. The sign analysis afterwards is unaffected because it only uses m-n and l modulo 2, so the fix appears to be replacing t0^l by t^l throughout the final equation. But as written, the proof does not establish Theorem 1.2.

Minor issues: the pullback formula (3.3) is stated without derivation, and the appendix's claim that Proposition A.1 implies a special case of the p-adic transformation is only sketched. Both are minor compared to the exponent error.

Who this is for: people working on Dwork's p-adic hypergeometric functions and their cohomological interpretations. It deserves a serious referee: the result is likely true, the method is sound in outline, and the error is fixable. I would send it to review but with the clear expectation that the exponent be corrected and the derivation verified.

Referee Report

4 major / 4 minor

Summary. The paper proves a transformation formula for Dwork's p-adic hypergeometric function under inversion of the argument: for a in N^{-1}Z with p not dividing N and 0<a<1, the author shows that F^Dw_{a,...,a}(t) = ((-1)^{d+1}t)^l F^Dw_{a,...,a}(t^{-1}) in the relevant Tate algebra for every d at least 1, with a signed version for p=2. The proof applies Asakura's Frobenius congruence (Theorem 2.2) to the hypergeometric scheme and to the inverted family, uses an explicit isomorphism between the associated quotient schemes V*_{t_0^N} and V*_{t_0^{-N}}, and compares the resulting Frobenius eigenvalues. An appendix gives a finite-field analogue and a remark connecting it to Corollary 3.2.

Significance. If correct, Theorem 1.2 is a substantial generalization of Wang's earlier d=1 result, removing the condition p>N and covering all d at least 1. The proof is conceptually clean: the inversion isomorphism is explicit, the unit-root submodules are preserved, and the parity bookkeeping is concrete. A particular strength is that the proof does not assume the desired transformation formula; the Frobenius congruence is imported from independent work. The paper is concise but relies heavily on Asakura's Theorem 2.2, and several base-change details are left implicit. I also checked the potential exponent discrepancy raised in the review: it does not actually land, because the proof sets t=t_0^N before the final comparison, so the displayed t^{p a_m-a_n} equals t_0^{N(p a_m-a_n)}.

major comments (4)

[Appendix A, Proposition A.1] The 'In particular' assertion does not follow from the first displayed formula. Substituting alpha_i=alpha and beta_i=epsilon into the first assertion gives a right-hand side with numerator parameters alpha,...,alpha and denominator parameters alpha^2,...,alpha^2, not epsilon,...,epsilon. Hence the claimed equality d+1F_d(alpha,...,alpha; epsilon,...,epsilon; t) = alpha((-1)^{d+1}t) d+1F_d(alpha,...,alpha; epsilon,...,epsilon; t^{-1}) is not a consequence of (A.1) unless alpha^2=epsilon. The appendix needs a corrected statement and proof, or the claim should be removed.
[Section 1.1 and proof of Theorem 1.2] The coefficient field is introduced as the fraction field of W(F_p), but the entire argument requires mu_N to be contained in K, both for the characters chi_{i_0,...,i_d} in Section 2 and for the element xi with xi^N=-1 in equation (3.2). This fails for many N with p not dividing N over Q_p, for example p=3 and N=5. Please clarify that K is the fraction field of W(\bar F_p), or another extension containing the N-th roots of unity, and explain how the resulting identity over the extension descends to the Z_p statement of Theorem 1.2.
[Section 3, equation (3.6)] The passage from Theorem 2.2 on U to the congruence on the inverted family bU is too terse. Since bU is obtained from U by the base change tau=t_0^{-N}, one must verify that the Frobenius sigma(t_0)=t_0^p induces the required sigma(tau)=tau^p, and that the Dwork function appearing in the congruence is F_a^Dw(t_0^{-N})^{-1}. The one-line citation to [15, Proposition 4.11(3)] does not supply this verification; please expand this step.
[Section 3, proof of Theorem 1.2] After applying eι^* to (3.6), the proof implicitly uses two facts: eι^* commutes with the Frobenius Φ, and the quotient isomorphism (2.1) allows one to equate the coefficients of omega_{n,...,n} modulo V, i.e., that V ∩ bBK omega_{n,...,n} = 0. These facts are not stated or proved in the text. They are plausible, but because they are load-bearing for the final equality, they should be justified explicitly.

minor comments (4)

[Section 3, final display] To forestall a possible misreading of the final comparison: since t=t_0^N by the notation fixed before (3.6), the factor t^{p a_m-a_n} equals t_0^{N(p a_m-a_n)}, so there is no missing factor of N in the proof.
[Section 3, equation (3.1)] The factor N^{d+1} in the pullback formula (3.1) should be typeset unambiguously; as rendered, 'N d+1' is easy to misread.
[Section 3, p=2 case] In the p=2 parity discussion, the equivalence (a_n)' = a_m = 1 - m/N ≡ 1 (mod 2) with m ≡ 0 (mod 2) uses that N is odd; this is correct but should be stated explicitly for the reader.
[References] References [2], [3], and [4] are arXiv preprints; please update to published versions if they are available at the time of final submission.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity detected: the proof relies on an independent Frobenius congruence from Asakura and derives the transformation formula without assuming it.

full rationale

The paper's central derivation does not assume the transformation formula it proves. Theorem 1.2 is established by comparing Frobenius actions on de Rham cohomology of V*_t and V*_{t^{-1}}, using Theorem 2.2, quoted from Asakura [3, Theorem 4.6], which is an independent result describing the Frobenius congruence Phi(omega_{j0,...,jd}) = p^d F^Dw_a(t)^{-1} omega_{i0,...,id} modulo a unit-root submodule. This cited theorem is not derived from the present paper and does not contain Conjecture 1.1 as an input; it is a genuinely external cohomological statement about the Dwork hypergeometric function. The construction of the isomorphism iota between V*_{t^N} and V*_{t^{-N}} is explicit, and the subsequent eigenvalue comparison is a direct computation from the cited Frobenius congruence. No parameter is fitted to the target formula, no prediction is a renamed input, and no self-citation is load-bearing: the only citations used essentially are Asakura's theorem, Wang's conjecture/prior cases, and Otsubo's finite-field identities, none of which are by the present author. The finite-field appendix independently states a transformation formula over F_q and derives it from Otsubo's results, again without assuming the p-adic theorem. Any apparent issue in the proof, such as the role of the exponent N in t0, would be a matter of internal correctness rather than circularity, since the claimed derivation is not equivalent to its assumptions by construction. Therefore the circularity score is 0.

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

The proof introduces no free parameters and no ad hoc entities. All inputs are the parameter a subject to the stated conditions, plus imported theorems: Dwork's congruence, Asakura's Frobenius eigenvalue formula, and Wang's equality of Tate algebras. The only subtle background assumption is the existence of xi, which requires the residue field to be algebraically closed. The ledger is therefore light; the cost of the result is mostly paid by the cited papers [3], [5], and [15].

assumptions (5)

domain assumption Frobenius congruence for the hypergeometric scheme (Theorem 2.2, cited from Asakura [3, Theorem 4.6]).
This theorem is the bridge from Frobenius eigenvalues on H^d_dR to Dwork's p-adic hypergeometric function. The proof applies it to both U and its t-inverted counterpart at equation (3.6) without re-derivation.
domain assumption Dwork's congruence and the Tate-algebra property of F^Dw_a(t) (from [5, Theorem 2] and [1, Corollary 2.3]).
The theorem is stated in a completed Tate algebra, so F^Dw_a(t) must be known to be an invertible element of that algebra; this is imported from earlier literature.
domain assumption Equality of the two Tate algebras under inversion (from Wang [15, Proposition 4.11(3)]).
This equality makes the involution f(t) maps to f(t^{-1}) well-defined and justifies comparing F^Dw_a(t) with F^Dw_a(t^{-1}) inside one ring.
standard math Standard results on algebraic de Rham cohomology used in Proposition 3.1: localization exact sequence, perfect pairing, and D-module isomorphisms.
These are background facts invoked without proof; they support identification of the cohomology groups and the unit-root quotient at (2.1).
domain assumption The base field K is large enough to contain an N-th root of -1 when N is even, needed for xi in (3.2).
The map iota requires xi with xi^N=-1, which exists if the residue field is algebraically closed. If K were merely an unramified extension of Q_p, this could fail, so the base field must be the fraction field of W of the algebraic closure.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Transformation formula of Dwork's $p$-adic hypergeometric function." pith.science (2026). https://pith.science/paper/AZJ6XPDC

@misc{pith2026250524215,
  author       = {Pith},
  title        = {Pith review of: Transformation formula of Dwork's $p$-adic hypergeometric function},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AZJ6XPDC}},
  note         = {Machine review of arXiv:2505.24215}
}

abstract

In this paper, we give a transformation formula of Dwork's $p$-adic hypergeometric function between $t$ and $t^{-1}$. As an appendix, we introduce a finite analogue of this transformation formula, which implies the special case of the above transformation formula.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 13 canonical work pages

[1]

Asakura,Frobenius action on a hypergeometric curve and an algorithm for computing values of Dwork’sp-adic hypergeometric functions, Springer Proc

M. Asakura,Frobenius action on a hypergeometric curve and an algorithm for computing values of Dwork’sp-adic hypergeometric functions, Springer Proc. Math. Stat.,373Springer, Cham, 2021, 1–45

2021
[2]

A generalization of the Ross symbols in higher K-groups and hypergeometric functions I

M. Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions I, arXiv:2003.10652

work page Pith review arXiv 2003
[3]

Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions II, arXiv:2102.07946

M. Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions II, arXiv:2102.07946

arXiv
[4]

Asakura, N

M. Asakura, N. Otsubo,On the adelic Gaussian hypergeometric function, arXiv:2408.08012

arXiv
[5]

Dwork,p-adic cycles, Inst

B. Dwork,p-adic cycles, Inst. Hautes ´Etudes Sci. Publ. Math. No.37(1969), 27–115

work page 1969
[6]

Fuselier, L

J. Fuselier, L. Long, R. Ramakrishna, H. Swisher, F-T. Tu,Hypergeometric functions over finite fields, Mem. Amer. Math. Soc.280(2022), no. 1382, vii+124 pp

work page 2022
[7]

Greene,Hypergeometric functions over finite fields, Trans

J. Greene,Hypergeometric functions over finite fields, Trans. Am. Math. Soc.301(1) (1987), 77–101

work page 1987
[8]

Hartshorne,On the De Rham cohomology of algebraic varieties, Inst

R. Hartshorne,On the De Rham cohomology of algebraic varieties, Inst. Hautes ´Etudes Sci. Publ. Math. No.45(1975), 5–99

work page 1975

Show all 16 references

[9]

N. M. Katz,Exponential Sums and Differential Equations, Ann. of Math. Stud.,124Prince- ton University Press, Princeton, NJ, 1990, xii+430 pp

1990
[10]

Koblitz,The number of points on certain families of hypersurfaces over finite fields, Compositio Math.48(1983), no

N. Koblitz,The number of points on certain families of hypersurfaces over finite fields, Compositio Math.48(1983), no. 1, 3–23

1983
[11]

Le Stum,Rigid cohomology, Cambridge Tracts in Math.,172

B. Le Stum,Rigid cohomology, Cambridge Tracts in Math.,172. Cambridge University Press, Cambridge, 2007, xvi+319 pp

2007
[12]

McCarthy,Transformations of well-poised hypergeometric functions over finite fields, Fi- nite Fields Appl.18(6), 1133–1147 (2012)

D. McCarthy,Transformations of well-poised hypergeometric functions over finite fields, Fi- nite Fields Appl.18(6), 1133–1147 (2012). TRANSFORMATION FORMULA OF DWORK’Sp-ADIC HYPERGEOMETRIC FUNCTION 11

2012
[13]

Otsubo,Hypergeometric functions over finite fields, Ramanujan J.63(2024), no

N. Otsubo,Hypergeometric functions over finite fields, Ramanujan J.63(2024), no. 1, 55– 104

2024
[14]

L. J. Slater,Generalized hypergeometric functions, Cambridge University Press, Cambridge, 1966, xiii+273 pp

1966
[15]

C. H. Wang,Congruence relations forp-adic hypergeometric functions bF (σ) a,...,a(t)and its transformation formula, Manuscripta Math.169(2022), no. 3-4, 565–602

2022
[16]

C. H. Wang,On transformation formulas ofp-adic hypergeometric functionsF (σ) a,...,a(t)and bF (σ) a,...,a(t), arXiv:2104.14092. Graduate School of Science and Engineering, Chiba University, Yayoicho 1-33, Inage, Chiba, 263-8522 Japan. Email address:y-nemoto@waseda.jp

arXiv

Pith tools

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

[1] [1]

Asakura,Frobenius action on a hypergeometric curve and an algorithm for computing values of Dwork’sp-adic hypergeometric functions, Springer Proc

M. Asakura,Frobenius action on a hypergeometric curve and an algorithm for computing values of Dwork’sp-adic hypergeometric functions, Springer Proc. Math. Stat.,373Springer, Cham, 2021, 1–45

2021

[2] [2]

A generalization of the Ross symbols in higher K-groups and hypergeometric functions I

M. Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions I, arXiv:2003.10652

work page Pith review arXiv 2003

[3] [3]

Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions II, arXiv:2102.07946

M. Asakura,A generalization of the Ross symbols in higher K-groups and hypergeometric functions II, arXiv:2102.07946

arXiv

[4] [4]

Asakura, N

M. Asakura, N. Otsubo,On the adelic Gaussian hypergeometric function, arXiv:2408.08012

arXiv

[5] [5]

Dwork,p-adic cycles, Inst

B. Dwork,p-adic cycles, Inst. Hautes ´Etudes Sci. Publ. Math. No.37(1969), 27–115

work page 1969

[6] [6]

Fuselier, L

J. Fuselier, L. Long, R. Ramakrishna, H. Swisher, F-T. Tu,Hypergeometric functions over finite fields, Mem. Amer. Math. Soc.280(2022), no. 1382, vii+124 pp

work page 2022

[7] [7]

Greene,Hypergeometric functions over finite fields, Trans

J. Greene,Hypergeometric functions over finite fields, Trans. Am. Math. Soc.301(1) (1987), 77–101

work page 1987

[8] [8]

Hartshorne,On the De Rham cohomology of algebraic varieties, Inst

R. Hartshorne,On the De Rham cohomology of algebraic varieties, Inst. Hautes ´Etudes Sci. Publ. Math. No.45(1975), 5–99

work page 1975