We prove that the generalized Fermat equation x^3+y^5=z^7 has no solution in nonzero coprime integers, and we make explicit how the proof extends the prior work cited below. Dahmen-Siksek had established the signed local descent and eliminated the three cases in which the associated degree-seven algebra is reducible. Putz had then proved that the remaining irreducible case can involve only seven septic fields: six pure fields and one exceptional field. The missing step was to exclude those seven fields. For the six pure fields we construct an explicit Fano resolvent, descend to a smooth plane quartic of genus three over \mathbb{Q}(\sqrt{-7}), and prove that its rational-parameter locus consists only of five points above the branch values. For the exceptional field we combine the modularity and conductor results of Pacetti-Villagra Torcomian with level lowering over \mathbb{Q}(\sqrt{5}) and a finite Hecke comparison at 29. We also reconstruct the three reducible-sector arguments of Dahmen-Siksek as independently replayable calculations, including their database-free identification of the quadratic field, and add an intrinsic formal-group treatment at the ramified prime. The paper labels every major step as a literature input, reconstructed input, or new argument. All project-specific finite calculations are supplied as exact certificates, programs, inputs, and authenticated logs.
There are no nonzero primitive integers x,y,z satisfying
Suppose that a nonzero primitive solution exists and reorder it as in (?). The descent algebra A_\eta belongs to exactly one of the four sectors in (?). Proposition 3.1 excludes [1,6]. Proposition 4.2 excludes [2,5]. Proposition 5.1 excludes [3,4]. Proposition 6.2 excludes [7]. Every possible factorization is therefore impossible, a contradiction.
The polynomial f_\eta has no rational root. Hence the sector [1,6] is empty.
The rational points on D are exactly
with independent signs in the second family.
Map D to the genus-two curve
The étale algebra of the right side is \mathbf Q\times\mathbf Q\times L_4, where
An exact 2-cover descent [BruinStollDescent] gives three fake-Selmer classes, whose components at 0, 4, and \vartheta are
The first class cannot lift to a nonendpoint rational point on D. Indeed, in this class x=-3ta^2 and x-4=tb^2, whereas the additional condition on D requires x(x-4)/(-35) to be a square. This would make 3/35 a square, which is impossible already over \mathbf Q_7. The endpoint x=4 does not lift either, since it would require \beta^2=105. The other two classes map to elliptic curves over L_4.
Their ranks are respectively 3 and 2, their torsion is trivial, and elliptic Chabauty [BruinEC] retains only the rational x-coordinates
For the rank-three curve, completeness is certified by saturation at every prime dividing the Chabauty index
Substitution into (?) gives precisely the six points stated.
The sector [2,5] is empty.
The sector [3,4] is empty.
Let J_+=\operatorname{Jac}(C_+). Then
The corresponding parameter values are 0,1,\infty,\infty,\infty.
Put D_i=[P_i-P_1] and H_0=\langle D_2,D_3,D_4,D_5\rangle. Appendix ? gives the complete containing-space and local stopping argument in the genus-three framework of [BPS], including the 7-adic kernel witness and the literal 5-adic obstruction. Together with the independently checked lower bound, it proves
The independent 2-saturation witness shows that D_2,D_3,D_4,D_5 form a basis modulo 2. Thus H_0 has odd index. The existence of the fifth-division class used below can be seen from the branch divisor. Write A=(t)_0/5, a K_7-rational divisor of degree three, and set B=[A-3P_1]. The pole divisor is (t)_\infty=7P_3+7P_4+P_5, so
Consequently the class E=3B-4D_3-4D_4 satisfies
Set H_1=\langle D_2,E,D_4,D_5\rangle. Equation (Proof) gives H_0\subseteq H_1. Four independent local C_{25} rows have determinant 4 modulo 5, and J_+(K_7)[5]=0; hence H_1 is 5-saturated. Its index is therefore coprime to 400, and H_1\to J_+(K_7)/400J_+(K_7) is surjective. This is the coefficient coverage used in the sieve.
At the split primes
both reductions of the curve are enumerated, as are the images of every reduced point in the relevant local quotients of J_+. The same global coefficient vector must occur at both places. In addition, rationality of t(Q) imposes equality of the two projective reductions of t. The resulting finite sieve has the following exact counts:
The final classes use exactly the five vectors modulo 25
and their projection at the two primes above 23 consists precisely of the five ordered residue pairs of P_1,\ldots,P_5.
The combined 23-adic logarithm matrix has size 6\times4 and rank 4. The resulting two independent logarithmic functionals vanish on H_0. Since H_0 has finite index, they vanish on all of J_+(K_7): for any P\in J_+(K_7), a nonzero integer multiple of P lies in H_0, and the logarithm takes values in characteristic zero. At the disks of P_2,P_3,P_4,P_5, the Siksek tangent determinants modulo 23 are
This is the number-field Chabauty criterion of [Siksek]; the values are units, so each disk contains only its reference point among common Coleman zeros.
At P_1, write the two local parameters as 23u,23v, so that the whole product residue disk is parametrized by (u,v)\in\mathbf Z_{23}^2. For the two annihilator integrals F_1,F_2, use the normalized equations F_1/23=0 and (F_2-6F_1)/23^2=0. The second numerator is integrally divisible by 23^2: its linear coefficients vanish modulo 23 before scaling, and all higher terms supply the extra factor. The reductions are
They have exactly two simple reduced zeros,
with determinants 20 and 3. Multivariate Hensel lifting gives one exact common zero in each of these two classes and none elsewhere. The first is P_1. The arithmetic retains seven digits after both normalizing divisions: an omitted integrated term of exponent k\ge12 has valuation at least k-v_{23}(k)\ge12, and the coefficient arithmetic retains the required precision. Multiplication by 23 gives the local coordinates modulo 23^8. The canonical parameter denominator is a unit at both lifted endpoints, so its evaluation preserves those eight digits. At the second zero, the two parameter values modulo 23^8=78{,}310{,}985{,}281 are
Their difference has exact 23-adic valuation 5, with first nonzero quotient digit 9. Hence this common Coleman zero does not have rational t. The five reference points are therefore the entire rational-parameter locus.
The sector [7] is empty.
Let d_v=\dim_{\mathbf F_2}J(K_v)/2J(K_v). Suppose actual local divisor classes give r_v independent fake images and k_v independent classes in the fake kernel. If r_v+k_v=d_v, these images span the full fake image and the fake kernel has dimension k_v.
Apply rank–nullity to f_{K_v}. The two exhibited subspaces give lower bounds whose sum already equals the dimension of its domain. For a finite place of odd residue characteristic, d_v=\dim J[2](K_v); for v\mid2, d_v=3[K_v:\mathbf Q_2]+\dim J[2](K_v), by [BPS, Remark 11.6]. This is the stopping criterion of [BPS, Lemma A.24(b)].
\operatorname{Cl}(L)[2]=0.
The degree-56 field L has an S_4-type Galois closure over the degree-14 field E; write F for the corresponding degree-42 field, N for the degree-84 field with N/F quadratic, and T for the degree-21 subfield with F=T(\sqrt{-7}). The Brauer–Kuroda relation \zeta_E\zeta_N=\zeta_L\zeta_F (all 24 permutation-character identities are verified exactly) and the exact regulator comparison give
where J is the index of the unit correspondence; exact 2-saturation of that correspondence gives v_2(J)=27. It therefore suffices that h(E), h(F) and h(N) are odd. h(E)=1 is certified unconditionally (bnfcertify). For T: any unramified quadratic extension of T split at the real places has discriminant d_T^2 and splits every principal prime of T; the 3{,}216 certified principal primes of norm at most 30\,000 enter the unconditional local-corrected discriminant bound of Brueggeman–Doud [BrueggemanDoud, Theorem 2.4(1)], evaluated with interval arithmetic, and the bound exceeds \log d_T^2 by a margin of 0.0407; hence no such extension exists and h(T) is odd. F/T is ramified at one finite and three real places with unit Hilbert-symbol rank 3, and N/F is ramified at four primes with unit symbol rank 3; the ambiguous-class formula gives h(F) and then h(N) odd. The steps are the prior-evidence jobs E_class_one, T21_prime_principality, T21_norm_export, T21_directed_Arb_bound, F_odd_class, pair_ambiguous, J27_images, J27_chain, J27_terminal, S4_integral_relation and class_field_bindings of the companion. Under evidence/v3/repository/results/, the relevant runs are:
2026-08-26_l2_relative_s4_class_gate: the acceptance contract, whose initial status was NO_GO;
2026-08-26_t_odd_odlyzko_exact_local_certificate_v1: the discriminant certificate;
2026-08-26_t_odd_odlyzko_downstream_implication_replay_v1: the completed conditions of the contract (summaries/T_ODD_DOWNSTREAM_IMPLICATION_REPLAY_V1.md, certificates/downstream_implication_decision_v1.json).
The true 2-Selmer group and the Mordell–Weil group satisfy
Every true Selmer class has fake image in Q=B\oplus\langle c\rangle. If its image were c+b, choose a point of H_0 with fake image b. Subtracting its global Kummer class would give an everywhere locally Kummer lift of c, contradicting the obstruction at 5. Thus the true Selmer image lies in B. The comparison is injective, so the Selmer dimension is at most four. The independent point lower bound is four. Since J(K)[2]=0, the Kummer exact sequence
proves all three assertions.
The text on this page is from arXiv:2609.26996v1 by Peter Chocian, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2609/2609.26996.jsonl.