Split-prime supercongruence at the mixed CM point ( 1 6 , 1 3 , 1

Alex Shvets · 2026

Read Split-prime supercongruence at the mixed CM point ( 1 6 , 1 3 , 1 by Alex Shvets (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-19773.txt.

Introduction

Shvets proves a split prime supercongruence at the complex multiplication point one sixth, one third, one. For a sequence built from a cubed hypergeometric series, he shows a congruence modulo the fourth power of the prime for every split prime from seven upward. The exponent four beats the weight three prediction of three, an extra factor traced to complex multiplication enhancement at j equals zero. A matching inert prime obstruction is also established, via a formal parameter congruence and a Cartier parity law modulo p.

Catalog record

CREATORAlex Shvets
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE107248
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.19773

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Split-prime supercongruence at the mixed CM point ( 1 6 , 1 3 , 1 ) (\tfrac{1}{6},\tfrac{1}{3};1) Thanks: ORCID: 0009-0005-9802-379X Alex Shvets Abstract . For the mixed CM point ( a , b , c ) = ( 1 6 , 1 3 , 1 ) (a,b,c)=(\tfrac{1}{6},\tfrac{1}{3},1) , set A n mix := 108 n ​ [ z n ] ​ F 1 2 ​ ( 1 6 , 1 3 , 1 , z ) 3 A_{n}^{\mathrm{mix}}:=108^{n}[z^{n}]{}_{2}F_{1}(\tfrac{1}{6},\tfrac{1}{3};1;z)^{3} . We prove unconditionally that for every split prime p ≥ 7 p\geq 7 , p ≡ 1 ( mod 3 ) p\equiv 1\pmod{3} , and every m ≥ 1 m\geq 1 , A m ​ p mix ≡ A m mix ( mod p 4 ) . A_{mp}^{\mathrm{mix}}\equiv A_{m}^{\mathrm{mix}}\pmod{p^{4}}. The exponent 4 4 exceeds the generic weight- 3 3 Hodge-gap prediction of 3 3 from the Roberts–Rodriguez Villegas framework [ 14 ] ; the additional factor of p p is a CM enhancement traced to the j = 0 j=0 elliptic curve. We also prove the matching unconditional inert-prime obstruction. For inert primes p ≥ 5 p\geq 5 , equivalently p ≡ 2 ( mod 3 ) p\equiv 2\pmod{3} , the split-style formal-parameter congruence cannot hold on the q q -side, and the coefficient sequence itself satisfies an explicit Cartier parity law modulo p p : one Frobenius step exchanges the two Eisenstein branches and two steps return to the original branch. The proof proceeds via the modular realization on Γ 0 ​ ( 3 ) \Gamma_{0}(3) with C mix = 3 ​ E 5 , χ 0 , χ 3 − 27 ​ E 5 , χ 3 , χ 0 C_{\mathrm{mix}}=3E_{5,\chi_{0},\chi_{3}}-27E_{5,\chi_{3},\chi_{0}} and parameter t = u / ( 1 + 27 ​ u ) 2 t=u/(1+27u)^{2} , u = η ​ ( 3 ​ τ ) 12 / η ​ ( τ ) 12 u=\eta(3\tau)^{12}/\eta(\tau)^{12} . Lagrange–Bürmann reduces the supercongruence to three Cartier identities Λ p ​ ( C mix ​ U p ℓ ) ≡ 0 ( mod p 4 ) \Lambda_{p}(C_{\mathrm{mix}}U_{p}^{\ell})\equiv 0\pmod{p^{4}} , ℓ = 1 , 2 , 3 \ell=1,2,3 . The key technical input is a length-three Witt–Cartier pole estimate at the tame elliptic stack point P − : u = − 1 / 27 P_{-}:u=-1/27 , j = 0 j=0 , where μ 3 \mu_{3} -equivariance of the canonical Frobenius lift, combined with p ≡ 1 ( mod 3 ) p\equiv 1\pmod{3} , forces K ≡ 1 ( mod 3 ) K\equiv 1\pmod{3} in the residue calculation and produces the extra digit. A short Atkin–Lehner intertwining transports the resulting classicality from C 0 C_{0} to u ​ C 0 uC_{0} and yields the unconditional mixed Cartier cancellation, which is the exact Katz–Dwork input for the main theorem. 1. Introduction Let A_{n}^{\mathrm{mix}}:=108^{n}[z^{n}]{}_{2}F_{1}\!\left(\tfrac{1}{6},\tfrac{1}{3};1;z\right)^{3}. The sequence begins 1,\ 18,\ 864,\ 55152,\ 4035906,\ 320012532,\ 26749991016,\ldots. The main result is the following. Theorem (Main theorem) . For every split prime p\geq 7 , p\equiv 1\pmod{3} , and every m\geq 1 , A_{mp}^{\mathrm{mix}}\equiv A_{m}^{\mathrm{mix}}\pmod{p^{4}}. The proof establishes the scalar Katz–Dwork congruence \Lambda_{p}\!\left(C_{\mathrm{mix}}(q)H_{\mathrm{mix}}(q)^{pm}\right)\equiv C_{\mathrm{mix}}(q)H_{\mathrm{mix}}(q)^{m}\pmod{p^{4}} for every m\geq 1 , from which the displayed coefficient congruence follows by Lagrange–Bürmann. This is proved as Theorem 4.33 . The exponent 4 is one more than the generic weight- 3 Hodge-gap exponent predicted by the Roberts–Rodriguez Villegas framework [ 14 ] ; the extra factor of p is a CM enhancement attached to the j=0 split-ordinary elliptic curve. The same level of enhancement appears in earlier work for related but distinct families: Long–Ramakrishna obtain p^{6} for a Van Hamme-type sum at the j=0 point [ 11 ] , and Zudilin records analogous CM-driven gains [ 18 ] . For the present sequence in this exact form, the result appears to be new. The matching negative result is the inert-prime obstruction. Theorem (Inert obstruction) . If p\geq 5 and \chi_{3}(p)=-1 , the p^{4} formal-parameter congruence cannot hold. More precisely, for m=p^{a}m_{0} with p\nmid m_{0} and \beta(m_{0})\not\equiv 0\pmod{p} , v_{p}(c_{mp}^{\mathrm{mix}}-c_{m}^{\mathrm{mix}})=0. Thus the obstruction occurs already on the q -side, in the Eisenstein coefficient tower. There is also a coefficient-level first-digit theorem. For \varepsilon\in\{\pm 1\} define A_{m}^{(\varepsilon)}:=[q^{m}]\bigl(C_{0}(q)-27\varepsilon\,u(q)C_{0}(q)\bigr)H_{\mathrm{mix}}(q)^{m}. Then for every prime p\geq 5 , every r\geq 0 , and every m\geq 0 , A_{mp^{r}}^{\mathrm{mix}}\equiv A_{m}^{(\chi_{3}(p)^{r})}\pmod{p}. In particular, if p\equiv 2\pmod{3} , then A_{mp^{2r}}^{\mathrm{mix}}\equiv A_{m}^{\mathrm{mix}}\pmod{p},\qquad A_{mp^{2r+1}}^{\mathrm{mix}}\equiv A_{m}^{(-1)}\pmod{p}, and A_{p}^{\mathrm{mix}}\equiv 72\pmod{p} . These are proved as Theorems 5.2 , 5.3 , 5.4 and 5.5 . The dichotomy comes from the second Eisenstein piece E_{5,\chi_{3},\chi_{0}} : its Euler factor is compatible with the p^{4} tower at split primes and acquires a unit-root factor at inert primes; on the coefficient sequence this is seen as branch exchange modulo p . Method The modular foundation is the order drop L^{(3)}=(S-108)L^{(2)} for the symmetric-cube recurrence specialized at (1/6,1/3;1) from the Mao–Tian framework [ 13 ] , together with the level- 3 dictionary u(\tau)=\frac{\eta(3\tau)^{12}}{\eta(\tau)^{12}},\qquad t=\frac{u}{(1+27u)^{2}},\qquad C_{\mathrm{mix}}=(1-27u)C_{0}, where C_{0}=3E_{5,\chi_{0},\chi_{3}} and uC_{0}=E_{5,\chi_{3},\chi_{0}} . Lagrange–Bürmann gives A_{m}^{\mathrm{mix}}=[q^{m}]C_{\mathrm{mix}}H_{\mathrm{mix}}^{m} with H_{\mathrm{mix}}=q/t , and a Katz–Dwork reduction converts the supercongruence to three Cartier identities \Lambda_{p}(C_{\mathrm{mix}}U_{p}^{\ell})\equiv 0\pmod{p^{4}},\qquad\ell=1,2,3, where U_{p}=\log(t(q)^{p}/t(q^{p}))\in pq\mathbf{Z}_{(p)}[[q]] . The main estimate is a length-three Witt–Cartier pole estimate at the tame elliptic stack point P_{-}:u=-1/27 , j=0 , with effective \mu_{3} -stabilizer ( Proposition 4.13 ). Working in the formal completion [\operatorname{Spf}\mathbf{Z}_{p}[\zeta_{3}][[r]]/\mu_{3}] and using the descended canonical Frobenius lift constructed via a fine prime-to- 3 atlas, \mu_{3} -equivariance forces the residue index K to satisfy K\equiv 1\pmod{3} . For K\geq 2 this gives K\geq 4 , which after the valuation count carried out in Proposition 4.13 yields the desired p^{4-\ell} bound. The same estimate, applied to both C_{0} and uC_{0} , gives classicality of g\Lambda_{p}(C_{0}\phi_{p}^{\ell}/p^{\ell}) and g\Lambda_{p}(uC_{0}\phi_{p}^{\ell}/p^{\ell}) as weight- 5 , \chi_{3} forms modulo p^{4-\ell} . A short Atkin–Lehner intertwining T_{p}w_{3}^{*}=\chi_{3}(p)w_{3}^{*}T_{p} then converts the joint classicalities into the bridge identity \beta_{\ell}=\gamma_{\ell}/27 between the scalars on the two sides; at split primes \chi_{3}(p)=1 the intertwining is untwisted. This cancels the two Eisenstein components in C_{\mathrm{mix}}=C_{0}-27uC_{0} and gives the unconditional mixed Cartier identities \Lambda_{p}(C_{\mathrm{mix}}U_{p}^{\ell})\equiv 0\pmod{p^{4}} for \ell=1,2,3 , which is the exact Katz–Dwork input for the main theorem. Notes on the literature The first-digit congruence A_{p}^{\mathrm{mix}}\equiv 18\pmod{p^{3}} that follows from our main theorem is, via Clausen’s identity and the substitution z\mapsto 108t , equivalent to a special case of Beukers [ 4 , Ex. 5.2] on the Clausen sum \sum_{n=0}^{p-1}(3n)!(2n)!/((n!)^{5}\,108^{n})\bmod p^{3} ; analogous Beukers-method results for binomial sums at other parameter specializations appear in Z.-H. Sun and D. Ye [ 16 ] . Our argument differs in route, not in conclusion at this depth. The novelty is the p^{4} -strengthening at all m\geq 1 , the inert-prime obstruction, and the local stack-level mechanism producing the extra digit. Outline Section 2 records the modular setup and the order drop. Section 3 performs the Katz–Dwork reduction to three Cartier identities. Section 4 proves the Witt–Cartier estimate and the Atkin–Lehner bridge leading to mixed Cartier cancellation. The main theorem ( Theorem 4.33 ) follows directly from this cancellation via Katz–Dwork reduction. Section 5 proves both the unconditional inert q -side obstruction and the coefficient-level inert Cartier parity law modulo p . Section 6 summarizes the computational verification. Notation and central objects We fix the notation used below. Sequence and hypergeometric function. Set A_{n}^{\mathrm{mix}}:=108^{n}[z^{n}]\,{}_{2}F_{1}\!\left(\tfrac{1}{6},\tfrac{1}{3};1;z\right)^{3},\qquad F_{\mathrm{mix}}(t):=\sum_{n\geq 0}A_{n}^{\mathrm{mix}}t^{n}={}_{2}F_{1}\!\left(\tfrac{1}{6},\tfrac{1}{3};1;108t\right)^{3}.

Read the complete text