An introduction to p-adic L-functions — Lean blueprint

3.7. Locally analytic functions and distributions🔗

Finally we sketch locally analytic distributions, the dual of locally analytic functions, which extend the Mahler correspondence from \OL[[T]] to all everywhere-convergent power series on the open unit disc. This section is used only peripherally (to study values of \zeta_p near s = 1).

Definition3.7.1
uses 0used by 1XL∃∀N

The p-adic open unit ball is B(0,1) = \set{z \in \Cp : \abs{z} < 1}. An L-valued function on B(0,1) is rigid analytic if it is given by a power series \sum_n a_n T^n \in L[[T]] everywhere convergent on B(0,1) (i.e. \abs{a_n} r^n \to 0 for all r < 1); write \rp \subset L[[T]] for these. A rigid analytic function is bounded if the \abs{a_n} are bounded.

The bounded rigid analytic functions form \OL[[T]] \otimes_{\OL} L, which by Theorem 3.4.4 is \sM(\Zp, L): measures on \Zp are precisely the bounded rigid analytic functions on B(0,1). It is natural to drop boundedness, extending the correspondence to all of \rp.

Definition3.7.2
uses 0used by 1XL∃∀N

A function f : \Zp \to L is locally analytic at z if there exist n_z \geq 0 and a_k(z) \in L with \sum_{k} a_k(z)(x-z)^k = f(x) for all x \in z + p^{n_z}\Zp; it is locally analytic if so at every z \in \Zp. Write \cCla(\Zp, L) for the L-vector space of such functions. Setting \cC^{n\text{-an}}(\Zp, L) to be those with a uniform radius n_z = n — a Banach space under \norm{f}_n = \sup_{z} \sup_k \abs{a_k(z)} p^{-nk} — one has \cCla(\Zp, L) = \varinjlim_n \cC^{n\text{-an}}(\Zp, L) with the direct limit topology.

Locally analytic functions are continuous, so \cCla(\Zp, L) \subset \cC(\Zp, L) densely (locally constant functions are locally analytic), though the locally analytic topology is finer than the induced one.

Definition3.7.3
Statement uses 2
Statement dependency previews
Preview
Definition 3.4.1
Loading preview
Statement dependency preview content is loaded from the Blueprint HTML cache.
Used by 4
Reverse dependency previews
Preview
Theorem 3.7.4
Loading preview
Reverse dependency preview content is loaded from the Blueprint HTML cache.
XL∃∀N

The space \sDla(\Zp, L) of locally analytic distributions on \Zp is the continuous dual \mathrm{Hom}_{\mathrm{cts}}(\cCla(\Zp, L), L), using Definition 3.7.2. We write \int_{\Zp} \phi \cdot \mu := \mu(\phi) and, since the binomial polynomials Definition 3.4.1 are locally analytic, extend the Mahler transform by \Am_\mu(T) = \int_{\Zp}(1+T)^x \cdot \mu = \sum_n (\int_{\Zp}\binomc{x}{n}\cdot\mu) T^n \in L[[T]].

Theorem3.7.4
Statement uses 3
Statement dependency previews
Preview
Theorem 3.4.4
Loading preview
Statement dependency preview content is loaded from the Blueprint HTML cache.
used by 0XL∃∀N

The Mahler transform induces a bijection \sDla(\Zp, L) \xrightarrow{\ \sim\ } \rp \subset L[[T]], an isomorphism of Fréchet spaces. This extends Theorem 3.4.4 and rests on Definition 3.7.3 and Definition 3.7.1.

Proof for Theorem 3.7.4
uses 0

This is a theorem of Amice–Colmez. Both sides are inverse limits of Banach spaces. On the function side, \cCla = \varinjlim_n \cC^{n\text{-an}}, so \sDla = \varprojlim_n \sD^{n\text{-an}} with each \sD^{n\text{-an}} a Banach space under the strong dual topology. On the power-series side, B(0,1) is the increasing union of the closed discs B(0,r), r < 1, so \rp = \varprojlim_{r < 1} \cO(B(0,r)) is an inverse limit of Banach spaces of analytic functions; both carry Fréchet topologies. The growth condition \abs{a_n} r^n \to 0 defining \rp is exactly dual to the radius-p^{-n} analyticity norms on \cC^{n\text{-an}}, so the term-by-term Mahler map is a topological isomorphism at each level and hence in the limit.

Restricting a measure \mu \in \sM(\Zp, L) to \cCla(\Zp, L) gives a locally analytic distribution \widetilde\mu; by density of \cCla this is injective, so \sM(\Zp, L) \subset \sDla(\Zp, L). Comparing Theorem 3.4.4 with Theorem 3.7.4, this matches the inclusion of bounded power series \OL[[T]] \otimes_{\OL} L \subset \rp. Every operation of the toolbox carries over verbatim to distributions.

Proposition3.7.5
uses 1used by 0XL∃∀N

There is a multiplicative analogue for \Zpx. The weight space is \cW(\Cp) = \mathrm{Hom}_{\mathrm{cts}}(\Zpx, \Cp^\times); using \Zpx \cong \mu_{p-1} \times (1 + p\Zp), evaluation at a topological generator of 1 + p\Zp identifies it with p - 1 copies of B(0,1), \cW(\Cp) = \bigsqcup_{\nu \in (\mu_{p-1})^\vee} U_\nu. A measure \mu on \Zpx is the bounded rigid analytic function F_\mu(\chi) = \int_{\Zpx} \chi \cdot \mu on \cW, multiplicative convolution becoming pointwise multiplication; a pseudo-measure Definition 3.6.1 is then a rigid function on \cW with at worst a simple pole at the trivial character.

Proof for Proposition 3.7.5
Proof uses 2
Proof dependency previews
Preview
Theorem 3.4.4
Loading preview
Proof dependency preview content is loaded from the Blueprint HTML cache.

The decomposition \Zpx \cong \mu_{p-1} \times (1+p\Zp) splits a continuous character into its restriction to the torsion part \mu_{p-1} (one of p-1 characters \nu) and to the pro-p part 1+p\Zp; evaluating the latter at a fixed topological generator lands in B(0,1), giving the disjoint-union description. The pairing \chi \mapsto \int_{\Zpx} \chi \cdot \mu is, by the multiplicative analogue of Theorem 3.4.4 (Amice), a bounded rigid function on \cW, and convolution dualises to multiplication. For a pseudo-measure \lambda = \mu/([a]-[1]) with a a topological generator, \int_{\Zpx} \chi \cdot ([a]-[1]) = \chi(a) - 1 vanishes only when \chi(a) = 1, i.e. for the trivial character; hence \lambda is analytic away from the trivial character, where it may have a simple pole, by Proposition 3.6.6.