Home Mathematics Hausdorff dimension in R-analytic profinite groups
Article Publicly Available

Hausdorff dimension in R-analytic profinite groups

  • Gustavo A. Fernández-Alcober EMAIL logo , Eugenio Giannelli and Jon González-Sánchez
Published/Copyright: September 14, 2016

Abstract

We study the Hausdorff dimension of R-analytic subgroups in an infinite R-analytic profinite group, where R is a pro-p ring whose associated graded ring is an integral domain. In particular, we prove that the set of such Hausdorff dimensions is a finite subset of the rational numbers.

1 Introduction

The study of Hausdorff dimension in profinite groups was initiated by Abercrombie in [1], and has attracted special attention in recent times; see, for example, [2, 3, 4, 7, 9, 13]. If G is a countably based profinite group, a (normal) filtration of G is a descending series {Gn}n of (normal) open subgroups of G which is a base of neighbourhoods of 1. Throughout the paper, we always assume that G is infinite. Then we can define a metric on G by

d(x,y)=inf{|G:Gn|-1xy-1Gn},

and the topology induced by d coincides with the original topology in G. The metric d defines a Hausdorff dimension function on all subsets of G, which we denote by hDim{Gn}. As shown in [4, Theorem 2.4], if the filtration {Gn}n is normal and H is a closed subgroup of G, then

hDim{Gn}(H)=lim infnlog|H:HGn|log|G:Gn|.

In [4], Barnea and Shalev studied the Hausdorff dimension in p-adic analytic pro-p groups with respect to the filtration {Gpn}n, and proved that

hDim{Gpn}(H)=dim(H)dim(G)

for every closed subgroup H of G. Here, if G is an analytic group, dim(G) denotes the dimension of G as an analytic manifold.

In this paper we study Hausdorff dimension in R-analytic profinite groups, where R is a pro-p ring whose associated graded ring is an integral domain. We recall that a pro-p ring R is a Noetherian commutative ring with maximal non-zero ideal 𝔪 such that R/𝔪 is a finite field of characteristic p, and R is complete in the 𝔪-adic topology. The requirement for the associated graded ring to be a domain is the natural setting in the theory of analytic groups over pro-p rings, see [5, Chapter 13]. Other fundamental references for analytic groups in the general case where R is not necessarily a principal ideal domain are [10] and [11].

If G is an R-analytic profinite group, then G is virtually a pro-p group, but the subgroups Gpn need not to form a filtration, since they need not be open in G. We consider instead the natural filtration induced by a standard open subgroup of G. As we will see, the Hausdorff dimension of a closed subgroup H of G is the same with respect to the natural filtrations induced by all open standard subgroups of G. This allows us to define the concept of standard Hausdorff dimension of H, denoted by hDimStd(H). We get the following theorem, in the spirit of the result of Barnea and Shalev.

Main Theorem

Let G be an R-analytic profinite group and let H be an R-analytic subgroup of G. Then

(1.1)hDimStd(H)=dim(H)dim(G).

An R-analytic subgroup H of G is a subgroup that is also an R-analytic submanifold of G; then H is closed in G. A crucial remark is that while the converse is true in a p-adic analytic group, i.e. every closed subgroup is analytic, it need not hold for an arbitrary pro-p ring R. For example, 𝔽p[[td]] is a closed subgroup of 𝔽p[[t]] for every positive integer d, but it is not analytic for d>1, even if it is an 𝔽p[[t]]-analytic group in its own right.

Observe that (1.1) implies that the R-analytic spectrum of G, defined by

SpecR(G)={hDimStd(H)H is an R-analytic subgroup of G},

is finite and consists of rational values. On the other hand, the spectrum of 𝔽p[[t]] corresponding to all closed subgroups is the full interval [0,1], as shown in [4, Lemma 4.1]. Thus our main theorem is pointing to the fact that most closed subgroups of 𝔽p[[t]] are non-analytic.

Notation

We write X(d) to denote the cartesian product of d copies of a set X. The symbol o indicates that a subset of a topological space is open.

2 Preliminaries

Throughout this paper R is a pro-p ring whose associated graded ring is an integral domain. Hence R is an integral domain, and we write K for its field of fractions. If 𝔪 is the maximal ideal of R, then R/𝔪 is a finite field of characteristic p. We will denote by q the cardinality of R/𝔪. Set H(n)=dimR/𝔪(𝔪n/𝔪n+1) for all n. For large enough n, H(n) coincides with a polynomial P(n), called the Hilbert polynomial of R (see [6, Chapter 8, Theorem C]).

Let G be an R-analytic group. Without loss of generality, we assume that the manifold structure is given by a full atlas. A subgroup H of G is an analytic subgroup if it is also an analytic submanifold of G. We adopt Serre’s definition of submanifold [12, Section 3.11]. Every analytic subgroup is closed in G. The converse is not true in general, but every open subgroup H of G (actually, any open subset) is a submanifold when considered with the restrictions of the charts of G, and dimH=dimG.

An R-analytic group S with a global chart given by a homeomorphism

ϕ:S(𝔪N)(d)

is called an R-standard group of level N if ϕ(1)=0 and, for all j{1,,d}, there exist Fj(X,Y)R[[X,Y]] (where X and Y are d-tuples of indeterminates), without constant term, such that

(2.1)ϕ(xy)=F(ϕ(x),ϕ(y))for every x,yS.

Here, F=(F1,,Fd) is called the formal group law associated to S. In these circumstances, we also say that (S,ϕ) is a standard group.

By using (2.1), F defines a new group structure on (𝔪N)(d), other than its natural additive structure, and then ϕ is an isomorphism between S and ((𝔪N)(d),F). Every R-analytic group contains an open (and so analytic) R-standard subgroup S (see [5, Theorem 13.20]). As a consequence, profinite R-analytic groups are countably based.

The formal group law of a standard group satisfies the following result (see [5, pp. 331–334]).

Lemma 2.1

Let F be a formal group law of dimension d associated to a standard group S. Then

F(X,Y)=X+Y+G(X,Y),

where every monomial involved in G has total degree at least 2, and contains a non-zero power of Xr and of Ys for some r,s{1,,d}. Moreover,

F(X,I(X))=0=F(I(X),X)

for some I(X)=-X+H(X)R[[X]], and every monomial involved in H has total degree at least 2.

It follows that (𝔪N+n)(d) is a subgroup of ((𝔪N)(d),F) for all n0. This allows us to introduce a special type of filtrations in an R-analytic group.

Definition 2.2

Let G be an R-analytic group and let S be a standard open subgroup of G, with global chart (S,ϕ). Then for every positive integer n0, and for every subset AS, we define

𝔪ϕnA=ϕ-1(𝔪nϕ(A)).

We say that {𝔪ϕnS}n is the natural filtration of G induced by S.

Observe that

𝔪ϕnS=ϕ-1((𝔪N+n)(d)),

which implies that 𝔪ϕnS is an open subgroup of G, and also that {𝔪ϕnS}n is a filtration of G. Actually, we have 𝔪ϕnSS (see [5, Proposition 13.22]). Here N is such that S is a standard group of level N.

Lemma 2.3

Let (S,ϕ) be an R-standard group. Then:

  1. For all x,yS and nN, we have ϕ(xy-1)(𝔪n)(d) if and only if ϕ(x)-ϕ(y)(𝔪n)(d).

  2. For every n0, we have |S:𝔪ϕnS|=|(𝔪N)(d):(𝔪N+n)(d)|=qdf(n), where f(n)=i=NN+n-1H(i).

  3. ϕ is an isometry between the group S with the metric induced by the filtration {𝔪ϕnS}n and the group ((𝔪N)(d),+) with the metric induced by the filtration {(𝔪N+n)(d)}n.

Proof.

(i) Since ϕ(1)=0, we may assume that xy. Let k be such that ϕ(xy-1)(𝔪k)(d)(𝔪k+1)(d). Then since

ϕ(x)=ϕ(xy-1y)=F(ϕ(xy-1),ϕ(y))=ϕ(xy-1)+ϕ(y)+G(ϕ(xy-1),ϕ(y)),

by Lemma 2.1, we have

ϕ(x)-ϕ(y)ϕ(xy-1)(mod(𝔪k+1)(d)).

Hence also ϕ(x)-ϕ(y)(𝔪k)(d)(𝔪k+1)(d), and (i) follows.

(ii) Observe that (i) implies that

(2.2)xy-1𝔪ϕnSϕ(x)-ϕ(y)(𝔪N+n)(d),

or what is the same,

x𝔪ϕnS=y𝔪ϕnSϕ(x)+(𝔪N+n)(d)=ϕ(y)+(𝔪N+n)(d).

This proves (ii), since ϕ is a bijection.

(iii) According to the definition of the metric associated to a filtration, it is clear that (ii) and (2.2) together imply that ϕ is an isometry. ∎

Hausdorff dimension can be defined for any subset of a metric space, see [8, Chapter 2] for its definition and main properties. We need the following two lemmas about Hausdorff dimension in countably based profinite groups.

Lemma 2.4

Let G be a countably based profinite group with filtration {Gn}nN. Let H be a closed subgroup of G, and let U be a non-empty open subset of H. Then

hDim{Gn}(H)=hDim{Gn}(U).

Proof.

The proof is a straightforward consequence of [8, Section 2.2]. ∎

Let G be a countably based profinite group and let S be an open subgroup of G. If {Sn}n is a filtration of S, we can calculate the Hausdorff dimension of XS with respect to the metric induced by {Sn}n in S or in G, which we denote by hDim{Sn}S(X) and hDim{Sn}G(X), respectively. Our next lemma shows that there is no need to make this distinction in the notation.

Lemma 2.5

Let G be a countably based profinite group and let S be an open subgroup of G. If {Sn}nN is a filtration of S, then

hDim{Sn}G(X)=hDim{Sn}S(X)

for every XS.

Proof.

Let dS and dG be the metrics induced by {Sn}n in S and G. Then dS(x,y)=|G:S|dG(x,y) for all x,yS, and the identity map from (S,dS) to (S,dG) is bi-Lipschitz. Now the result follows from [8, Corollary 2.4]. ∎

3 Proof of the main theorem

In this section we first prove that the Hausdorff dimension of a closed subgroup in an analytic profinite group with respect to a natural filtration is independent of the standard subgroup. Then we prove the main theorem of our paper about Hausdorff dimension of analytic subgroups.

Theorem 3.1

Let G be an R-analytic profinite group, and let (S,ϕ) and (T,ψ) be two open standard subgroups of G. Then

hDim{𝔪ϕnS}(H)=hDim{𝔪ψnT}(H)

for every closed subgroup H of G.

Proof.

Let us write Sn and Tn for 𝔪ϕnS and 𝔪ψnT, and N(S) and N(T) for the levels of S and T. We first show that there exist non-negative integers a and b such that

(3.1)Sn+aTnSn-b

for every nb. Since the charts ϕ and ψ belong to the full atlas of G, the two functions ψϕ-1|ϕ(ST) and ϕψ-1|ψ(ST) are analytic. Since ϕ(ST) is open in ϕ(S), it follows that ψϕ-1 can be evaluated in (𝔪)(d) for some . By [5, Lemma 6.45], and observing that ψϕ-1(0)=0, there exists a natural number k such that

ψϕ-1((𝔪n+k)(d))(𝔪n)(d)

for every n0. This implies that Sn+aTn for n0, by choosing

a=max{k-N(S)+N(T),0}.

Arguing similarly with ϕψ-1, we get (3.1).

From Lemmas 2.4 and 2.5, we get

hDim{Tn}G(H)=hDim{Tn}G(HT)=hDim{Tn}T(HT).

Since {Tn} is a normal filtration of T, we can apply [4, Theorem 2.4] to calculate the last Hausdorff dimension, and we readily obtain (since G is infinite) that

hDim{Tn}(H)=lim infnlog|H:HTn|log|G:Tn|.

It follows that

hDim{Tn}(H)lim infnlog|H:HSn+a|log|G:Sn+a|-log|Tn:Sn+a|
lim infnlog|H:HSn+a|log|G:Sn+a|(1-log|Sn-b:Sn+a|log|G:Sn+a|)-1
=hDim{Sn}(H).

Notice that the last equality holds because, by (ii) of Lemma 2.3, for n large enough we have

logq|Sn-b:Sn+a|=di=N+n-bN+n+a-1P(i),

which is a polynomial in n of degree degP, and

logq|G:Sn+a|=logq|G:S|+d(C+i=NN+n+a-1P(i)),

where C is independent of n, which is a polynomial of degree degP+1 by the Euler–MacLaurin formula. By swapping S and T, the result follows. ∎

Definition 3.2

Let G be an R-analytic profinite group and let H be a closed subgroup of G. Then the standard Hausdorff dimension of H, hDimStd(H), is the Hausdorff dimension of H calculated with respect to the natural filtration induced by any given standard open subgroup of G.

We need the following lemma before proving our main theorem.

Lemma 3.3

Let R be a pro-p ring and let K be its field of fractions. Let E be a vector subspace of dimension e of K(d). Then, for every NN, the Hausdorff dimension of (mN)(d)E in R(d) with respect to the filtration {(mn)(d)}nN is e/d.

Proof.

Let ={w1,,we} be a basis of E, and let A be the e×d matrix over K whose rows are the vectors in . We may assume that A is in reduced echelon form. Set W=w1,,weR. Let k be such that the product of all denominators of the entries of A belongs to 𝔪k. One readily checks that

(3.2)𝔪n+kW(𝔪n)(d)E𝔪nW.

The required Hausdorff dimension is the limit inferior of the sequence

cn=log|(𝔪N)(d)E:(𝔪n)(d)E|log|R(d):(𝔪n)(d)|,

where nN. Now, from (3.2), for nN+k, we have

|𝔪N+kW:𝔪nW||(𝔪N)(d)E:(𝔪n)(d)E||𝔪NW:𝔪n+kW|.

Since W is a free R-module of rank e, we have

dimR/𝔪(𝔪nW/𝔪n+1W)=eH(n),

and consequently

e(H(N+k)++H(n-1))d(H(1)++H(n-1))cne(H(N)++H(n+k-1))d(H(1)++H(n-1)).

Now, since H(n)=P(n) for large enough n, we have

limnH(N+k)++H(n-1)H(1)++H(n-1)=limnP(N+k)++P(n-1)P(1)++P(n-1)=1,

since both sums in the last limit are polynomials in n of degree degP+1 and with the same leading coefficient. Arguing similarly with the upper bound for cn given above, we conclude that limncn=e/d, as desired. ∎

We are now ready to prove our main theorem.

Proof of the Main Theorem.

Let dim(G)=d and dim(H)=e. Since H is an R-analytic submanifold of G, there exist an open subset U of H containing 1, and a chart (V,ϕ) of G such that UV and ϕ(U)=Eϕ(V), for some vector subspace E of K(d) of dimension e.

From the proof of [5, Theorem 13.20], there exist a natural number N and an open subgroup S of G such that SV and (S,ϕ) is a standard subgroup of G of level N. Since USoUoH, we have

hDimStd(H)=hDim{𝔪ϕnS}(US),

by Lemma 2.4. Since ϕ is an isometry between S and ((𝔪N)(d),+) by Lemma 2.3, we get

hDim{𝔪ϕnS}(US)=hDim{(𝔪n)(d)}(ϕ(US)).

Now since ϕ(US)=E(𝔪N)(d), Lemma 3.3 yields that

hDim{(𝔪n)(d)}(ϕ(US))=ed,

which completes the proof. ∎

Corollary 3.4

Let G be an R-analytic group of dimension d. Then

SpecR(G){0,1d,,d-1d,1}.

In particular, the R-analytic spectrum of G is finite and consists of rational numbers.


Communicated by Andrea Lucchini


Funding statement: The first and third authors are supported by the Spanish Government, grants MTM2011-28229-C02 and MTM2014-53810-C2-2-P, and by the Basque Government, grants IT753-13 and IT974-16. The second author gratefully acknowledges financial support by the GRECA research group and by the ERC Advanced Grant 291512.

References

[1] L. Abercrombie J., Subgroups and subrings of profinite rings, Math. Proc. Cambridge Philos. Soc. 116 (1994), 209–222. 10.1017/S0305004100072522Search in Google Scholar

[2] M. Abért and B. Virág , Dimension and randomness in groups acting on rooted trees, J. Amer. Math. Soc. 18 (2005), 157–192.. 10.1090/S0894-0347-04-00467-9Search in Google Scholar

[3] Y. Barnea and B. Klopsch , Index-subgroups of the Nottingham group, Adv. Math. 180 (2003), 187–221. 10.1016/S0001-8708(02)00102-0Search in Google Scholar

[4] Y. Barnea and A. Shalev , Hausdorff dimension, pro-p groups, and Kac–Moody algebras, Trans. Amer. Math. Soc. 349 (1997), no. 12, 5073–5091. 10.1090/S0002-9947-97-01918-1Search in Google Scholar

[5] J. Dixon , M. Du Sautoy , A. Mann and D. Segal , Analytic Pro-p Groups, 2nd ed., Cambridge University Press, Cambridge, 1999. 10.1017/CBO9780511470882Search in Google Scholar

[6] D. Eisenbud , Commutative Algebra with a View Toward Algebraic Geometry, Grad. Texts in Math. 150, Springer, Berlin, 2004. Search in Google Scholar

[7] M. Ershov , On subgroups of the Nottingham group of positive Hausdorff dimension, Comm. Algebra 35 (2006), 193–206. 10.1080/00927870601041532Search in Google Scholar

[8] Falconer K., Fractal Geometry: Mathematical Foundations and Applications, John Wiley and Sons, New York, 1990. 10.2307/2532125Search in Google Scholar

[9] G. A. Fernández-Alcober and A. Zugadi-Reizabal , GGS-groups: Order of congruence quotients and Hausdorff dimension, Trans. Amer. Math. Soc. 366 (2014), 1993–2017. 10.1090/S0002-9947-2013-05908-9Search in Google Scholar

[10] A. Jaikin-Zapirain and B. Klopsch , Analytic groups over general pro-p domains, J. Lond. Math. Soc. (2) 76 (2007), 365–383. 10.1112/jlms/jdm055Search in Google Scholar

[11] A. Lubotzky and A. Shalev , On some Λ-analytic pro-p groups, Israel J. Math. 85 (1994), 307–337. 10.1007/BF02758646Search in Google Scholar

[12] J.-P. Serre , Lie Algebras and Lie Groups, Lecture Notes in Math. 1500, Springer, Berlin, 1992. 10.1007/978-3-540-70634-2Search in Google Scholar

[13] O. Siegenthaler , Hausdorff dimension of some groups acting on the binary tree, J. Group Theory 11 (2008), 555–567. 10.1515/JGT.2008.034Search in Google Scholar

Received: 2015-5-15
Revised: 2016-8-16
Published Online: 2016-9-14
Published in Print: 2017-5-1

© 2017 by De Gruyter

Downloaded on 12.1.2026 from https://www.degruyterbrill.com/document/doi/10.1515/jgth-2016-0040/html
Scroll to top button