Uniqueness of topology for the $p$-adic integers
L. Corwin
DOI: https://doi.org/10.1090/S0002-9939-1976-0414773-1
1976-02-01
Abstract:It is shown that the only Hausdorff topologies on Z , the p-adic integers, which make it into a locally compact Abelian group are the />adic and discrete topologies. The key ingredient in the proof is a structure theorem for certain LCA groups which may be of independent interest. Recently Kallman proved [3] that a compact simple Lie group has a unique nondiscrete Hausdorff topology which makes it a locally compact topological group. Groups with this property seem to be fairly rare. In this paper, we show that the p-adic integers enjoy the same property. Our main tool is a structure theorem involving unique divisibility. Let G be a locally compact Abelian group (the group operation is written additively), and let p be a prime. G is said to be uniquely divisible by p if for every a G G, the equation px = a has exactly one solution. Theorem 1. Let S be any set of primes, and suppose that G is uniquely divisible by allp G S. Then G can be written as Gx X G2 X G3, where: (a) G] is a vector group; (b) G2 = lines (G2p; K2 p), a restricted direct product of groups, where G2 p is isomorphic to a group HA „ of the following form: A is a cardinal, and HA,P ̂ IIa<^ Qp is the subgroup U„ez Ua<AP~'"^p topologized by letting Ha<A Zp be an °Pen Subgroup K2p S* TIa<,4 Zp(c) G3 has a compact open subgroup C uniquely divisible by all p G S. Note. In the special case where 5 is the set of all primes, C is a divisible subgroup and, hence, an algebraic direct summand; it is also clear that C is a topological direct summand. Hence G3 = C X G4, where G4 is a divisible torsion-free discrete subgroup and, hence, a product of copies of Q. The result in this case is a theorem of Mackey's [5]. Proof of Theorem 1. We may always write G = G0 x Gx, where G0 contains a compact open subgroup K0 and Gx is a vector group; see, e.g., [2]. That takes care of Gx, and we may now worry about G0. So consider K0 and its dual K0 . For p G S, x \-+ px is a continuous 1-1 map from K0 onto its image. Given a G KQ , we may define a character B0 on pK0 by B0(px) = a(x). Extend B0 to a character B on K0 (see [2, Corollary 24.12]); then pB = a. It follows that the p-primary component of K0 is Received by the editors October 23, 1974 and, in revised form, November 28, 1974. AMS (MOS) subject classifications (1970).Primary 22B05.