- Home
- Documents
*INTERSECTIONS OF ANALYTICALLY AND GEOMETRICALLY INTERSECTIONS OF ANALYTICALLY AND GEOMETRICALLY...*

prev

next

out of 12

View

1Download

0

Embed Size (px)

transactions of the american mathematical society Volume 343, Number 1, May 1994

INTERSECTIONS OF ANALYTICALLY AND GEOMETRICALLY FINITE SUBGROUPS OF KLEINIAN GROUPS

JAMES W. ANDERSON

Abstract. We consider the intersection of pairs of subgroups of a Klcinian group of the second kind K whose limit sets intersect, where one subgroup G is analytically finite and the other / is geometrically finite, possibly infinite cyclic. In the case that J is infinite cyclic generated by M, we show that either some power of M lies in G or there is a doubly cusped parabolic element Q of G with the same fixed point as M. In the case that J is nonelementary, we show that the intersection of the limit sets of G and J is equal to the limit set of the intersection G n J union with a subset of the rank 2 parabolic fixed points of K . Hence, in both cases, the limit set of the intersection is essentially equal to the intersection of the limit sets. The main facts used in the proof are results of Beardon and Pommerenke [4] and Canary [6], which yield that the Poincaré metric on the ordinary set of an analytically finite Kleinian group G is comparable to the Euclidean distance to the limit set of G .

1. Introduction

The purpose of this paper is to investigate the intersection of certain sub- groups of Kleinian groups, with an eye towards understanding the relationship between the intersection of their limit sets, which is topological information, and the limit set of their intersection, which is algebraic information. We con- centrate here on Kleinian groups of the second kind, where we require one subgroup to have finite area quotient (analytic finiteness) and the other to be geometrically finite, possibly infinite cyclic. As all finitely generated Kleinian groups are analytically finite, by Ahlfors' finiteness theorem, these results apply to a large class of groups. In the case that the geometrically finite subgroup is cyclic, we obtain the following result, which is proven as Theorems 4.1 and 4.2.

Let K be a Kleinian group of the second kind, let G be an analytically finite subgroup of K, and let M be an infinite order element of K which has a fixed point in A(G). Then, either M" G C7 for some n > 0 or there is a doubly cusped parabolic element Q of G with the same fixed point as M.

For a nonelementary geometrically finite subgroup, we obtain the following result, which is proven as Theorem 5.2.

Received by the editors February 26, 1992. 1991 Mathematics Subject Classification. Primary 30F40; Secondary 57M50, 20H10.

©1994 American Mathematical Society 0002-9947/94 $1.00 + 125 per page

«7

License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use

88 J. W. ANDERSON

Let K be a Kleinian group of the second kind, let G be a nonele- mentary analytically finite subgroup of K, and let J be a nonele- mentary geometrically finite subgroup of K. Then, G n J is geometrically finite and K(G) n A(J) = A(G n J) U P(G, J). Moreover, P(G, /) C Q(G n /).

Here, P(G, J) is the set of points whose stabilizer in K has rank 2 but whose stabilizers in both G and J have rank 1. These points are isolated in £l(G n J), and are a precise measure of the difference between A(G) n A(J) and A(G n J).

There are two interesting features of the results proven in this paper. First, we are provided with information about how all finitely generated and some in- finitely generated groups can intersect with geometrically finite groups; namely, that this intersection is always geometrically finite, and hence finitely generated. We pay a small price for this information, namely that the groups involved are subgroups of some common group of the second kind. We also get exact in- formation about the intersection of the limit sets of the groups and its relation to the limit set of the intersection; in particular, this intersection is as large as possible, with the only exceptional points lying in the set P(G, J).

Analytic finiteness is a condition which is purely planar and does not involve the 3-dimensional action of G. So, the proofs of these two results use infor- mation about how the groups act of the Riemann sphere C, making use of the Poincaré metric on the ordinary set of a nonelementary Kleinian group.

The general problem of understanding subgroups of a Kleinian group and their intersections has been investigated by a number of authors; we will at- tempt to give a brief chronology. Thurston showed that every finitely generated subgroup of a geometrically finite Kleinian group of the second kind is itself ge- ometrically finite; for a proof, see [11]. Hempel [7] proved that the intersection of two finitely generated subgroups of a geometrically finite group of the second kind is again finitely generated, and hence geometrically finite. The author [2] extended this result to two finitely generated subgroups of any Kleinian group of infinite covolume. So, part of Theorem 5.2 can be thought of as extending Hempel's original result to the case of a geometrically finite and an analytically finite subgroup of a Kleinian group of the second kind.

However, we are interested in more than the group theory; we would also like to have some concrete information about the relationship between the action of the groups on the Riemann sphere and their group theory. This line of work also has a history. Maskit [9] proved the analogues of the two main theorems given here in the case that the subgroups are component subgroups of a Kleinian group; in this case, the set P(G, J) is empty.

If we remove the assumption that the groups be of the second kind, stronger hypotheses are needed on the subgroups involved. Susskind [13] proved the analogues of the two main theorems in the case that the subgroups are geomet- rically finite; these results were extended to higher dimensions by Susskind and Swarup [14]. Soma [12] showed that the conclusions of both the main theo- rems hold for two function groups which are subgroups of a Kleinian group. These results were extended by the author [1] to any pair of topologically tame subgroups of a Kleinian group.

License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use

INTERSECTIONS OF SUBGROUPS OF KLEINIAN GROUPS 89

Acknowledgments

The first result is the main result of my doctoral thesis from SUNY Stony Brook; I would like to thank my advisor, Bernie Maskit, for his support and encouragement. The second result is work done while I was a postdoctoral fel- low at the Mathematical Sciences Research Institute in Berkeley; this work was supported in part by NSF grant DMS 8505550. I would also like to thank Dick Canary for several helpful discussions and the referee for several careful read- ings of the early drafts and for pointing out that the original proof of Corollary 3.4 was incorrect.

2. Preliminaries

In this section, we define our terms and present some basic results we will make use of later. Our reference for the basics of Kleinian groups is [10].

A Kleinian group G is a discrete subgroup of PSL2( C ), which we consider as acting on the Riemann sphere C by Möbius transformations. The action of G decomposes C into two disjoint sets. The ordinary set ¿1(G) is the set of points of C at which G acts discontinuously; if Q(G) is nonempty, we say that G is of the second kind. The connected components of ¿1(G) are the components of G. Let ¿Io (G) denote ¿1(G) minus the fixed points of elements of finite order.

The complement of ¿1(G) in C is the limit set, denoted A(G). A Kleinian group G is nonelementary if A(G) contains at least 3 points; otherwise, G is elementary. An elementary group is either finite with empty limit set, contains a purely parabolic subgroup of finite index and has limit set a single point, or contains a loxodromic cyclic subgroup of finite index and has limit set a pair of points. For a complete description of elementary groups, see [10]. Alternatively, for a nonelementary G, we can think of A(G) as the set of accumulation points of the orbit of any point in C under G.

For any set X in C, define the stabilizer of X in G to be the subgroup stabG(X) = {g G G : g(X) = X} of elements of G which keep X invariant; the stabilizer of a component A of G is denoted G^ . For a subgroup H of G, we say X is precisely invariant under 77 in G if h(X) = X for all A G 77 and g(X) n X is empty for all g G G - 77.

Let r be a Möbius transformation which does not fix oo. The isometric circle It of T is the circle {z g C : |r'(z)| = 1} ; the radius of It is r(T) and the center of It is r-1(oo). The following lemma of Shimizu and Leut- becher lays out constraints on the radii of isometric circles of elements in certain Kleinian groups.

Lemma 2.1 [10]. Let G be a Kleinian group containing P(z) = z + 1 and let g be an element of G not fixing oo. Then, r(g) < 1.

For a Kleinian group G, let ext(G) denote the common exteriors of the isometric circles of all elements of G not fixing oo . If g is an element of G not fixing oo, then g(e\t(G)) is disjoint from ext(G), as g maps lg to lg-i, taking the region exterior to Ig onto the region interior to Ig-i .

A point of approximation of G is a limit point of G which behaves like a fixed point of a loxodromic element of G. More precisely, we say x is a point

License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use

90 J. W. ANDERSON

of approximation of G if there existjnfinitely many distinct elements gm of G so that, for every compact set C in C - {x} , there is a constant cq > 0 so that \Sm(x) - gm(z)\ > ec > 0