P-adic Fields (pCF)

Origin. James Ax and Simon Kochen, "Diophantine problems over local fields I–II" (American Journal of Mathematics 87, 1965; part II subtitled "A complete set of axioms for p-adic number theory") and "Diophantine problems over local fields III. Decidable fields" (Annals of Mathematics 83, 1966), and Yuri Eršov independently ("On the elementary theory of maximal normed fields", Doklady Akademii Nauk SSSR 165, 1965, with sequels in Algebra i Logika), axiomatized the first-order theory of ℚ_p and proved it decidable. Angus Macintyre, "On definable subsets of p-adic fields" (Journal of Symbolic Logic 41, 1976), proved quantifier elimination in the language with the power predicates. Alexander Prestel and Peter Roquette, Formally p-adic Fields (Lecture Notes in Mathematics 1050, 1984), generalized the theory to p-valued fields of arbitrary fixed p-rank. Jan Denef proved Serre's rationality conjecture for Poincaré series (Inventiones Mathematicae 77, 1984) and the cell decomposition theorem ("p-adic semi-algebraic sets and cell decomposition", Journal für die reine und angewandte Mathematik 369, 1986).

Models. The p-adically closed fields: fields elementarily equivalent to ℚ_p, standing to ℚ_p as real closed fields stand to ℝ and algebraically closed fields of fixed characteristic to ℂ — the third classical field theory with completeness, decidability, and quantifier elimination, the valuation in the structural position the order occupies in RCF. The algebraic numbers inside ℚ_p form a countable model and, by model completeness, an elementary substructure, as ℝ_alg does for RCF. Four conditions determine the theory: henselianity, characteristic 0, residue field 𝔽_p, value group a Z-group — and this reduction of a valued field to its residue field and value group is the Ax–Kochen–Eršov method, the starting point of the model theory of valued fields.

Formalism.

Language: +, ×, −, 0, 1, with the valuation carried by a predicate for the valuation ring ℤ_p (or a divisibility relation). Macintyre's language adds Pₙ for each n ≥ 2.

Axioms (p-rank 1):

• valued field of characteristic 0
• henselian
• residue field 𝔽_p
• value group a Z-group — elementarily equivalent to ⟨ℤ, +, <⟩
• v(p) = the least positive element of the value group

Completeness and decidability (Ax–Kochen II–III, Eršov): the axioms are complete and recursive, so Th(ℚ_p) is decidable. Model complete in the valued-field language.

Quantifier elimination (Macintyre 1976): in the ring language with ℤ_p and the predicates Pₙ for the n-th powers, every formula is equivalent to a quantifier-free one. Ax–Kochen, and primitive-recursively Paul Cohen (1969), had eliminated quantifiers in a language with a cross-section from the value group; the cross-section defines infinite discrete subsets of ℚ_p, which no formula of Macintyre's language can, and discarding it restores the parallel with the reals. Definable sets are the p-adic semialgebraic sets.

The Ax–Kochen principle: for every sentence φ of the valued-field language there is a finite set of primes outside which ℚ_p ⊨ φ iff 𝔽_p((t)) ⊨ φ. Mixed and equal characteristic agree at almost every prime, sentence by sentence.

Artin's conjecture: Artin conjectured that ℚ_p is C₂ — every form of degree d in more than d² variables has a nontrivial zero. Lang (1952) proved 𝔽_p((t)) is C₂; the transfer gives, for each d, a finite set of primes outside which ℚ_p satisfies the degree-d case. This is the Ax–Kochen theorem. Terjanian ("Un contre-exemple à une conjecture d'Artin", Comptes Rendus 262, 1966): a quartic form in 18 > 4² variables over ℚ₂ with only the trivial zero. The exceptional primes are real, "almost all" is exact, and only degrees 1, 2 (classical), and 3 (Lewis) hold for every p.

Cell decomposition (Denef 1986): every semialgebraic set splits into finitely many cells on which the valuation of each polynomial is read off from linear data, so p-adic integrals over semialgebraic sets reduce to geometric series. Consequences: rationality of Igusa's local zeta functions without resolution of singularities, and of the Poincaré series counting points modulo p^m — Serre's conjecture, Denef (1984). The analytic Serre–Oesterlé series followed from subanalytic cell decomposition: Denef and van den Dries, "p-adic and real subanalytic sets" (Annals of Mathematics 128, 1988).

p-adic Hilbert 17 (Kochen 1967, Roquette 1971): the Kochen operator in place of the square; model completeness characterizes the integral definite rational functions over ℚ_p.

Continuations, named: p-adic subanalytic and semialgebraic geometry; P-minimality, the p-adic counterpart of o-minimality; motivic integration (Kontsevich 1995; Denef and Loeser), which lifts the p-adic integrals to a geometry independent of p.

Symbols.

Metatheory. Complete, model complete, decidable (Ax–Kochen, Eršov), with quantifier elimination once the Pₙ are named (Macintyre) — and the need to name them is a definability fact in the same position as Presburger arithmetic's congruences ≡ₙ: the n-th powers are existentially definable but not quantifier-free definable, and QE is exact about the difference. Decidability has the same explanation as RCF's: ℕ is a subset of ℚ_p but not a definable one, so nothing codes sequences and Gödel's argument has nothing to work with. The transfer principle is what completeness delivers in use, as the Lefschetz principle is for ACF: a sentence moves between ℚ_p and 𝔽_p((t)) at almost every prime. The theory is NIP and unstable, beside RCF and against stable ACF. Terjanian's form measures the principle's precision: the transfer is sentence-by-sentence and almost-everywhere, and at p = 2 the two local fields genuinely disagree about quartics.

Applies to. Diophantine problems over local fields — forms in many variables, where the Ax–Kochen theorem is the result. Igusa's local zeta functions and point counting modulo p^m. Subgroup and representation growth of groups, through rationality of the associated p-adic integrals (du Sautoy, Segal). Motivic integration and uniform-in-p arithmetic geometry. p-adic semialgebraic and subanalytic geometry, as real algebraic geometry stands to RCF.

Limitations. The bare valued-field language does not eliminate quantifiers; the Pₙ or an equivalent are forced. Adding a cross-section back preserves decidability but makes infinite discrete sets definable, and the parallel with the reals is lost. The transfer to 𝔽_p((t)) holds at almost every prime only (Terjanian), and the exceptional set depends on the sentence. 𝔽_p((t)) itself is not covered: no complete axiomatization of Th(𝔽_p((t))) is known, and its decidability is open. Each prime has its own theory — nothing is uniform in p at the object level, and "for all p" statements live in the metatheory or in the motivic lift. Finite extensions of ℚ_p need the Prestel–Roquette apparatus of p-rank, with a prime element fixed for completeness. The field has no order: everything real-semialgebraic needs RCF (see Real Closed Fields.md), and everything about varieties over algebraically closed fields needs ACF (see Algebraically Closed Fields.md).

© 2026 Lingenic LLC