nLab p-adic number

-adic numbers

Context

Algebra

Arithmetic

pp-adic numbers

Idea

For pp any prime number, the pp-adic numbers (or pp-adic rational numbers, for emphasis) form a field p\mathbb{Q}_p that completes the field of rational numbers with respect to a metric, called the pp-adic metric. As such they are analogous to real numbers. A crucial difference is that the reals form an archimedean field, while the pp-adic numbers form a non-archimedean field.

Fields of pp-adic numbers play a role in non-archimedean analytic geometry that is analogous to the role of the real numbers/Cartesian spaces in ordinary differential geometry.

Moreover, as such they serve as an approximation technique in arithmetic geometry over prime fields 𝔽 p\mathbb{F}_p (see e.g. Lubicz).

There have long been speculations (see the references below) that this must mean that pp-adic numbers also play a central role in the description of physics, see p-adic physics.

Definition

We first recall the definition and construction of the p-adic integers

and then consider

Recollection of the pp-adic integers

Let Z\mathbf{Z} be the ring of integers and for every q0q\neq 0, qZq\mathbf{Z} its ideal consisting of all integer multiples of qq, and Z/qZ\mathbf{Z}/q\mathbf{Z} the corresponding quotient, the ring of residues mod qq.

Let now pZ +p\in \mathbf{Z}_+ be a prime number. Then for any two positive integers nmn\geq m there is an inclusion p nZp mZp^n \mathbf{Z}\subset p^m\mathbf{Z} which induces the canonical homomorphism of quotients ϕ n,m:Z/p nZZ/p mZ\phi_{n,m}:\mathbf{Z}/p^n\mathbf{Z}\to \mathbf{Z}/p^m\mathbf{Z}. These homomorphisms for all pairs nmn\geq m form a family closed under composition, and in fact a category, which is in fact a poset, and moreover a directed system of (commutative unital) rings. The ring of p-adic integers Z p\mathbf{Z}_p is the (inverse) limit of this directed system (in the category Ring of rings).

Regarding that the rings in the system are finite, it is clear that the underlying set of Z p\mathbf{Z}_p has a natural topology as a profinite (Stone) space and it is in particular a compact Hausdorff topological ring. More concretely, Z p\mathbf{Z}_p is the closed (hence compact) subspace of the cartesian product nZ/p nZ\prod_{n} \mathbf{Z}/p^n\mathbf{Z} of discrete topological spaces Z/p nZ\mathbf{Z}/p^n\mathbf{Z} (which is by the Tihonov theorem compact Hausdorff) consisting of threads, i.e. sequences of the form x=(...,x n,...,x 2,x 1)x = (...,x_n,...,x_2,x_1) with x np nZx_n\in p^n\mathbf{Z} and satisfying ϕ n,m(x n)=x m\phi_{n,m}(x_n) = x_m.

The kernel of the projection pr n:Z pZ/p nZpr_n: \mathbf{Z}_p\to\mathbf{Z}/p^n\mathbf{Z}, xx nx\mapsto x_n to the nn-th component (which is the corresponding projection of the limiting cone) is p nZ pZ pp^n\mathbf{Z}_p\subset\mathbf{Z}_p, i.e. the sequence

0Z pp nZ pZ/p nZ0 0 \longrightarrow \mathbf{Z}_p\stackrel{p^n}\longrightarrow \mathbf{Z}_p\longrightarrow \mathbf{Z}/p^n\mathbf{Z} \longrightarrow 0

is an exact sequence of abelian groups, hence also Z p/p nZ pZ/p nZ\mathbf{Z}_p/p^n\mathbf{Z}_p\cong \mathbf{Z}/p^n\mathbf{Z}.

An element uu in Z p\mathbf{Z}_p is invertible (and called a pp-adic unit) iff uu is not divisible by pp.

Let UZ pU\subset\mathbf{Z}_p be the group of all invertible elements in Z p\mathbf{Z}_p. Then every element xZ px\in \mathbf{Z}_p can be uniquely written as s=up ns= u p^n with n0n\geq 0 and uUu\in U. The correspondence xnx\mapsto n defines a discrete valuation v p:Z pZ{}v_p:\mathbf{Z}_p\to \mathbf{Z}\cup\{\infty\} called the p-adic valuation and nn is said to be the pp-adic valuation of xx. Of course, v p(0)=v_p(0)=\infty as required by the axioms of valuation. The norm induced by the valuation is (up to equivalence) given by |x| p=p v p(x){|x|}_p = p^{-v_p(x)}, and this in turn induces a metric

d(x,y)=|xy| p, d(x,y) = {|x-y|}_p,

making the ring Z p\mathbf{Z}_p a complete metric space and in fact a completion of Z\mathbf{Z}, in that dd is a complete metric, and Z\mathbf{Z} is dense in it.

Concretely, a pp-adic integer xx may be written as a base-pp expansion

x= n0a np nx = \sum_{n \geq 0} a_n p^n

with a n{0,1,,p1}a_n \in \{0, 1, \ldots, p-1\}. Addition and multiplication are performed with carrying as in ordinary base-pp arithmetic, carried infinitely far to the left if xx is written as a na n1a 1a 0\ldots a_n a_{n-1} \ldots a_1 a_0.

As an endomorphism ring

Algebraically, the ring of pp-adic integers is isomorphic to the endomorphism ring hom(Z(p ),Z(p ))\hom(\mathbf{Z}(p^\infty), \mathbf{Z}(p^\infty)) where Z(p )\mathbf{Z}(p^\infty) is the Prüfer group Z(p )[1/p]/\mathbf{Z}(p^\infty) \coloneqq \mathbb{Z}[1/p]/\mathbb{Z}. In particular, Z(p )\mathbf{Z}(p^\infty) is tautologically a Z p\mathbf{Z}_p-module.

Relatedly, the additive group of pp-adic integers is Pontrjagin dual to Z(p )\mathbf{Z}(p^\infty). Observe that Z(p )\mathbf{Z}(p^\infty) embeds in S 1S^1 as the set of all roots of unity of order p np^n, and that every character χ:Z(p )S 1\chi: \mathbf{Z}(p^\infty) \to S^1 factors through this embedding Z(p )S 1\mathbf{Z}(p^\infty) \hookrightarrow S^1.

The pp-adic numbers proper

The field of pp-adic numbers Q p\mathbf{Q}_p is the field of fractions of the p-adic integers Z p\mathbf{Z}_p. The pp-adic valuation v pv_p extends to a discrete valuation, also denoted v pv_p on Q p\mathbf{Q}_p. Indeed, it is still true for all xQ px\in \mathbf{Q}_p that they can be uniquely written in the form p nup^n u where uUu\in U (the same group UU as before), but now one needs to allow nZn\in \mathbf{Z}. One defines the metric on Q p\mathbf{Q}_p by the same formula as for Z p\mathbf{Z}_p. It appears that Q p\mathbf{Q}_p is a complete field (in particular locally compact Hausdorff) and that Z p\mathbf{Z}_p is an open subring.

The distance dd satisfies the “ultrametric” inequality

d(x,z)sup{d(x,y),d(y,z)} d(x,z) \leq sup\{d(x,y),d(y,z)\}

Concretely, a pp-adic number xx may be written as nka np n\sum_{n \geq k} a_n p^n, with only finitely many negative powers of pp occurring. If k<0k \lt 0, the expansion is conventionally displayed as

x=a 1a 0.a 1a kx = \ldots a_1 a_0.a_{-1} \ldots a_k

with finitely many terms to the “right” of the “decimal” point. Again such expressions are added and multiplied with carrying as in ordinary arithmetic.

Properties

Basic properties

Proposition

An element x px\in \mathbb{Z}_p is invertible precisely if x 00x_0 \neq 0.

As a field completion

Ostrowski's theorem implies there are precisely two kinds of completions of the rational numbers: the real numbers and the pp-adic numbers.

Theorem

(Ostrowski's theorem)

Any non-trivial absolute value on the rational numbers is equivalent either to the standard real absolute value, or to the pp-adic absolute value.

Topological disconnectedness and G-topology

While the pp-adic numbers are complete in the p-adic norm, that topology is exotic: p\mathbb{Q}_p is a locally compact, Hausdorff, totally disconnected topological space.

For that reason the naive idea of formulating p-adic geometry in analogy to complex analytic geometry as modeled on domains in p n\mathbb{Q}_p^n, regarded with their subspace topology, fails (for instance there would be no analytic continuation), as also all these domains are totally disconnected.

Instead there is (Tate 71) a suitable Grothendieck topology on such affinoid domains – the G-topology – with respect to which there is a good theory of non-archimedean analytic geometry (“rigid analytic geometry”) and hence in particular of p-adic geometry. Moreover, one may sensibly assign to an pp-adic domain a topological space which is well behaved (in particular locally connected and even locally contractible), this is the analytic spectrum construction. The resulting topological space is equipped with covers by affinoid domain under the analytic spectrum are called Berkovich spaces.

Pontryagin duality

Earlier we observed that as an additive compact Hausdorff topological group, the inverse limit Z p=lim n/(p n)\mathbf{Z}_p = \lim_{\leftarrow n} \mathbb{Z}/(p^n) is dual to the discrete Prüfer group Z(p )[1/p]/\mathbf{Z}(p^\infty) \coloneqq \mathbb{Z}[1/p]/\mathbb{Z} that is isomorphic to a direct limit of finite cyclic groups lim n/(p n)\lim_{\to n} \mathbb{Z}/(p^n). The canonical inclusion [1/p]Q p\mathbb{Z}[1/p] \to \mathbf{Q}_p induces an isomorphism Z(p )Q p/Z p\mathbf{Z}(p^\infty) \to \mathbf{Q}_p/\mathbf{Z}_p, in fact an isomorphism of Z p\mathbf{Z}_p-modules, so there is an exact sequence

0Z piQ pqZ(p )0.0 \to \mathbf{Z}_p \stackrel{i}{\hookrightarrow} \mathbf{Q}_p \stackrel{q}{\to} \mathbf{Z}(p^\infty) \to 0.

This exact sequence is Pontrjagin self-dual in the sense that the map Q pQ p \mathbf{Q}_p \to \mathbf{Q}_p^\wedge induced from the pairing

Q p×Q pmultQ pq[1/p]//\mathbf{Q}_p \times \mathbf{Q}_p \stackrel{mult}{\to} \mathbf{Q}_p \stackrel{q}{\to} \mathbb{Z}[1/p]/\mathbb{Z} \hookrightarrow \mathbb{R}/\mathbb{Z}

fits into an isomorphism of exact sequences

0 Z p i Q p q Z(p ) 0 0 Z(p ) q Q p i Z p 0\array{ 0 & \to & \mathbf{Z}_p & \stackrel{i}{\to} & \mathbf{Q}_p & \stackrel{q}{\to} & \mathbf{Z}(p^\infty) & \to & 0 \\ & & \downarrow & & \downarrow & & \downarrow & & \\ 0 & \to & \mathbf{Z}(p^\infty)^\wedge & \stackrel{q^\wedge}{\to} & \mathbf{Q}_p^\wedge & \stackrel{i^\wedge}{\to} & \mathbf{Z}_p^\wedge & \to & 0 }

where the commutativity of the squares can be traced to the fact that qq is a Z p\mathbf{Z}_p-module homomorphism, and where the vertical isomorphisms on left and right come from Pontrjagin duality. The middle arrow is then an isomorphism by the short five lemma for topological groups, which holds by protomodularity of topological groups.

This self-duality figures in Tate’s thesis; for more, see ring of adeles.

Function field analogy

function field analogy

number fields (“function fields of curves over F1”)function fields of curves over finite fields 𝔽 q\mathbb{F}_q (arithmetic curves)Riemann surfaces/complex curves
affine and projective line
\mathbb{Z} (integers)𝔽 q[z]\mathbb{F}_q[z] (polynomials, polynomial algebra on affine line 𝔸 𝔽 q 1\mathbb{A}^1_{\mathbb{F}_q})𝒪 \mathcal{O}_{\mathbb{C}} (holomorphic functions on complex plane)
\mathbb{Q} (rational numbers)𝔽 q(z)\mathbb{F}_q(z) (rational fractions/rational function on affine line 𝔸 𝔽 q 1\mathbb{A}^1_{\mathbb{F}_q})meromorphic functions on complex plane
pp (prime number/non-archimedean place)x𝔽 px \in \mathbb{F}_p, where zx𝔽 q[z]z - x \in \mathbb{F}_q[z] is the irreducible monic polynomial of degree onexx \in \mathbb{C}, where zx𝒪 z - x \in \mathcal{O}_{\mathbb{C}} is the function which subtracts the complex number xx from the variable zz
\infty (place at infinity)\infty
Spec()Spec(\mathbb{Z}) (Spec(Z))𝔸 𝔽 q 1\mathbb{A}^1_{\mathbb{F}_q} (affine line)complex plane
Spec()place Spec(\mathbb{Z}) \cup place_{\infty} 𝔽 q\mathbb{P}_{\mathbb{F}_q} (projective line)Riemann sphere
p() p()p\partial_p \coloneqq \frac{(-)^p - (-)}{p} (Fermat quotient)z\frac{\partial}{\partial z} (coordinate derivation)
genus of the rational numbers = 0genus of the Riemann sphere = 0
formal neighbourhoods
/(p n)\mathbb{Z}/(p^n \mathbb{Z}) (prime power local ring)𝔽 q[z]/((zx) n𝔽 q[z])\mathbb{F}_q [z]/((z-x)^n \mathbb{F}_q [z]) (nn-th order univariate local Artinian 𝔽 q \mathbb{F}_q -algebra)[z]/((zx) n[z])\mathbb{C}[z]/((z-x)^n \mathbb{C}[z]) (nn-th order univariate Weil \mathbb{C} -algebra)
p\mathbb{Z}_p (p-adic integers)𝔽 q[[zx]]\mathbb{F}_q[ [ z -x ] ] (power series around xx)[[zx]]\mathbb{C}[ [z-x] ] (holomorphic functions on formal disk around xx)
Spf( p)×Spec()XSpf(\mathbb{Z}_p)\underset{Spec(\mathbb{Z})}{\times} X (“pp-arithmetic jet space” of XX at pp)formal disks in XX
p\mathbb{Q}_p (p-adic numbers)𝔽 q((zx))\mathbb{F}_q((z-x)) (Laurent series around xx)((zx))\mathbb{C}((z-x)) (holomorphic functions on punctured formal disk around xx)
𝔸 = pplace p\mathbb{A}_{\mathbb{Q}} = \underset{p\; place}{\prod^\prime}\mathbb{Q}_p (ring of adeles)𝔸 𝔽 q((t))\mathbb{A}_{\mathbb{F}_q((t))} ( adeles of function field ) x((zx))\underset{x \in \mathbb{C}}{\prod^\prime} \mathbb{C}((z-x)) (restricted product of holomorphic functions on all punctured formal disks, finitely of which do not extend to the unpunctured disks)
𝕀 =GL 1(𝔸 )\mathbb{I}_{\mathbb{Q}} = GL_1(\mathbb{A}_{\mathbb{Q}}) (group of ideles)𝕀 𝔽 q((t))\mathbb{I}_{\mathbb{F}_q((t))} ( ideles of function field ) xGL 1(((zx)))\underset{x \in \mathbb{C}}{\prod^\prime} GL_1(\mathbb{C}((z-x)))
theta functions
Jacobi theta function
zeta functions
Riemann zeta functionGoss zeta function
branched covering curves
KK a number field (K\mathbb{Q} \hookrightarrow K a possibly ramified finite dimensional field extension)KK a function field of an algebraic curve Σ\Sigma over 𝔽 p\mathbb{F}_pK ΣK_\Sigma (sheaf of rational functions on complex curve Σ\Sigma)
𝒪 K\mathcal{O}_K (ring of integers)𝒪 Σ\mathcal{O}_{\Sigma} (structure sheaf)
Spec an(𝒪 K)Spec()Spec_{an}(\mathcal{O}_K) \to Spec(\mathbb{Z}) (spectrum with archimedean places)Σ\Sigma (arithmetic curve)ΣP 1\Sigma \to \mathbb{C}P^1 (complex curve being branched cover of Riemann sphere)
() pΦ()p\frac{(-)^p - \Phi(-)}{p} (lift of Frobenius morphism/Lambda-ring structure)z\frac{\partial}{\partial z}
genus of a number fieldgenus of an algebraic curvegenus of a surface
formal neighbourhoods
vv prime ideal in ring of integers 𝒪 K\mathcal{O}_KxΣx \in \SigmaxΣx \in \Sigma
K vK_v (formal completion at vv)((z x))\mathbb{C}((z_x)) (function algebra on punctured formal disk around xx)
𝒪 K v\mathcal{O}_{K_v} (ring of integers of formal completion)[[z x]]\mathbb{C}[ [ z_x ] ] (function algebra on formal disk around xx)
𝔸 K\mathbb{A}_K (ring of adeles) xΣ ((z x))\prod^\prime_{x\in \Sigma} \mathbb{C}((z_x)) (restricted product of function rings on all punctured formal disks around all points in Σ\Sigma)
𝒪\mathcal{O} xΣ[[z x]]\prod_{x\in \Sigma} \mathbb{C}[ [z_x] ] (function ring on all formal disks around all points in Σ\Sigma)
𝕀 K=GL 1(𝔸 K)\mathbb{I}_K = GL_1(\mathbb{A}_K) (group of ideles) xΣ GL 1(((z x)))\prod^\prime_{x\in \Sigma} GL_1(\mathbb{C}((z_x)))
Galois theory
Galois groupπ 1(Σ)\pi_1(\Sigma) fundamental group
Galois representationflat connection (“local system”) on Σ\Sigma
class field theory
class field theorygeometric class field theory
Hilbert reciprocity lawArtin reciprocity lawWeil reciprocity law
GL 1(K)\GL 1(𝔸 K)GL_1(K)\backslash GL_1(\mathbb{A}_K) (idele class group)
GL 1(K)\GL 1(𝔸 K)/GL 1(𝒪)GL_1(K)\backslash GL_1(\mathbb{A}_K)/GL_1(\mathcal{O})Bun GL 1(Σ)Bun_{GL_1}(\Sigma) (moduli stack of line bundles, by Weil uniformization theorem)
non-abelian class field theory and automorphy
number field Langlands correspondencefunction field Langlands correspondencegeometric Langlands correspondence
GL n(K)\GL n(𝔸 K)//GL n(𝒪)GL_n(K) \backslash GL_n(\mathbb{A}_K)//GL_n(\mathcal{O}) (constant sheaves on this stack form unramified automorphic representations)Bun GL n()(Σ)Bun_{GL_n(\mathbb{C})}(\Sigma) (moduli stack of bundles on the curve Σ\Sigma, by Weil uniformization theorem)
Tamagawa-Weil for number fieldsTamagawa-Weil for function fields
theta functions
Hecke theta functionfunctional determinant line bundle of Dirac operator/chiral Laplace operator on Σ\Sigma
zeta functions
Dedekind zeta functionWeil zeta functionzeta function of a Riemann surface/of the Laplace operator on Σ\Sigma
higher dimensional spaces
zeta functionsHasse-Weil zeta function

References

The pp-adic numbers had been introduced in

  • Kurt Hensel, Über eine neue Begründung der Theorie der algebraischen Zahlen Jahresbericht der Deutschen Mathematiker-Vereinigung 6 (3): 83–88. (1897)

A standard reference is

Review in the context of p-local homotopy theory is in

  • Dennis Sullivan, pp. 9 of Localization, Periodicity and Galois Symmetry (The 1970 MIT notes) edited by Andrew Ranicki, K-Monographs in Mathematics, Dordrecht: Springer (pdf)

Review of the use of pp-adic numbers in arithmetic geometry includes

  • David Lubicz, An introduction to the algorithmic of pp-adic numbers (pdf)

A formalization in homotopy type theory and there in Coq is discussed in

pp-adic differential equations are discussed in

The development of rigid analytic geometry starts with

  • John Tate, Rigid analytic spaces, Invent. Math. 12:257–289, 1971.

p-adic homotopy theory is discussed in

Last revised on November 26, 2021 at 08:23:03. See the history of this page for a list of all contributions to it.