This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Two Applications of Topological Fixed Point Theory

Chaitanya Ambi111Chennai Mathematical Institute, H1, SIPCOT IT Park, Siruseri, Kelambakkam 603103, India.
Email: [email protected]
Abstract

We give a new proof of Cartan’s fixed point theorem using topological fixed point theory. For an odd dimensional, simply connected and complete manifold having non-positive curvature, we further prove that every isometry with finite order not only fixes a point, but also maps a pencil of geodesics to itself.

We apply similar techniques to the automorphism group of a smoothly bounded domain in the complex plane having connectivity three or more. We show that any prime dividing the order of this finite group must divide a certain integer depending only on the connectivity of the domain.

Keywords: Topological Fixed Point Theory, Automorphism Groups
MSC[2020]: 57S05, 57S25

1 Introduction.

The geometric properties of a complete Riemannian manifold MM with non-positive curvature are intimately linked with its fundamental group. Specifically, the structure of the isometry group Isom(M)Isom(M) has been a field of intense investigation. It is not easy to ascertain whether a geometric property is, in fact, topological. As an attempt in this direction, we present a new proof of the following version of Cartan’s fixed point theorem (see [6], Chap. 6, Thm. 23, p. 164). A stronger version of the same for precompact subgroups of Isom(M)Isom(M) is known (see [2], Thm. 1.4.6, p. 21).

Theorem 1.1 (E. Cartan).

Consider a complete, simply connected Riemannian manifold MM of non-positive curvature. If σIsom(M)\sigma\in Isom(M) is an element of finite order, then

Fix(σ).Fix(\sigma)\not=\emptyset.

The above theorem can be proved by considering the centre of mass of the orbit of σ\sigma. In this article, we present a proof using topological fixed point theory. Our approach also reveals the following fact which is interesting by its own. By a pencil of geodesics, we shall mean a (non-intersecting) family of geodesics any two of which lie within a finite Hausdorff distance of one another.

Corollary 1.1.

In addition to the hypotheses of Cartan’s Theorem 1.1, assume that dimM\dim M is odd. Then there exists a pencil Γ\Gamma of geodesics such that

σ(Γ)=Γ.\sigma(\Gamma)=\Gamma.

Next, certain domains in the complex space n\mathbb{C}^{n} also become complete manifolds with non-positive curvature when equipped with the Bergmann metric. In this vein, we consider the group of conformal automorphisms of a bounded planar domain (with smooth boundary) having finite connectivity (see [5], Def. 4.2.2, p. 85). It is well-known that such a group is finite whenever the connectivity exceeds two (see [5], Thm. 12.2.3, p. 263). We prove the following result about such domains (where Aut(.)Aut(.) denotes the group of conformal automorphisms of the respective domain).

Theorem 1.2.

Let Ωk\Omega_{k}\subset\mathbb{C} be any bounded planar domain having connectivity k+1k+1 (with k2k\geq 2) whose boundary consists of finitely many smooth Jordan curves. If pp is a prime integer dividing the order of Aut(Ωk)Aut(\Omega_{k}), then we have

k3kmodp.k^{3}\equiv k\mod p.

As far as the author knows, it is an open problem to characterise the finite groups which may occur as the automorphism groups of planar domains. In particular, by [4], we have the following sharp upper bound for k2k\geq 2.

#Aut(Ωk)2k for k4,6,8,12,20,\displaystyle\#Aut(\Omega_{k})\leq 2k\quad\text{ for }k\not=4,6,8,12,20,
#Aut(Ω4)12,\displaystyle\#Aut(\Omega_{4})\leq 12,
#Aut(Ω6),#Aut(Ω8)24,\displaystyle\#Aut(\Omega_{6}),\#Aut(\Omega_{8})\leq 24,
#Aut(Ω12),#Aut(Ω20)60.\displaystyle\#Aut(\Omega_{12}),\#Aut(\Omega_{20})\leq 60.

Theorem 1.2 shows that there are further restrictions on the possible order of an element in Aut(Ωk)Aut(\Omega_{k}). Hence, the subgroups of Aut(Ωk)Aut(\Omega_{k}) are constrained by the value of kk. For instance, the cyclic group of order 77 cannot occur as a subgroup of Aut(Ωk)Aut(\Omega_{k}) for any k<6k<6 (otherwise, we would have 7|(k3k)7|(k^{3}-k)). Note that the upper bound mentioned above does not exclude this possibility as #Aut(Ω5)10\#Aut(\Omega_{5})\leq 10.

2 Proof of Theorem 1.1.

By the Cartan-Hadamard Theorem (see [6], Chap. 6, Thm. 22, p. 162), MM is diffeomorphic to n\mathbb{R}^{n}, where n=dimMn=\dim M. We compactify MM by adding its ideal boundary M()M(\infty) to obtain the Gromov compactification

M¯:=MM().\bar{M}:=M\cup M(\infty).

We refer the reader to [1], Chap, 3, pp. 22-32 for the details. In fact, M¯\bar{M} is homeomorphic to the closed unit ball 𝔹nn\mathbb{B}^{n}\subset\mathbb{R}^{n}.

Next, let σIsom(M)\sigma\in Isom(M) have a finite order. It is known that σ\sigma induces a homeomorphism σ¯\bar{\sigma} of M¯𝔹n\bar{M}\approx\mathbb{B}^{n}.

We shall argue inductively to reduce the dimension nn. By Brouwer’s fixed point theorem, σ¯\bar{\sigma} has a fixed point xM¯x\in\bar{M}. If xM()x\in M(\infty), we observe that (see [1], Chap. 3, p. 30) σ¯\bar{\sigma} maps a horosphere centred at xx to itself. This horosphere in M¯\bar{M} is diffeomorphic to 𝔹n1\mathbb{B}^{n-1}. So, we may repeat the whole argument to keep reducing the dimension. At each stage, the restriction of σ¯\bar{\sigma} fixes a point lying either in MM or on M()M(\infty).

If the latter case were to occur until the dimension reaches two, the argument would yield a geodesic γM\gamma\subset M satisfying

σ(γ)=γ.\sigma(\gamma)=\gamma.

But now, γ\gamma is isometric with \mathbb{R} (equipped with the standard Euclidean metric). Any isometry of the latter either fixes a point or has infinite order. As σ\sigma is of finite order, we get a fixed point as asserted. This completes the argument.

2.1 Proof of Corollary 1.1.

As in the proof of Theorem 1.1, the extended map σ¯\bar{\sigma} restricts to a homeomorphism of M()M(\infty). Since we want to find a family of undirected geodesics (and not geodesic rays), we may well disregard the directions to obtain the quotient of M()M(\infty) by /2\mathbb{Z}/2\mathbb{Z} action. As M()𝕊n1M(\infty)\approx\mathbb{S}^{n-1}, the quotient space is homeomorphic to the real projective space Pn1\mathbb{R}P^{n-1}. This space also has an action induced by that of σ¯\bar{\sigma}.

Assume the contrary that the above action induced by σ¯\bar{\sigma} has no fixed points. Let mm\in\mathbb{N} be the smallest power of σ¯\bar{\sigma} whose Lefschetz number is nonzero. It is clear that mm cannot exceed the order of σ¯\bar{\sigma} because (n1)(n-1) is even and so

λ(IdPn1)=χ(Pn1)=1.\lambda(Id_{\mathbb{R}P^{n-1}})=\chi(\mathbb{R}P^{n-1})=1.

If m1m\not=1, choose any prime pp which divides mm. Set τ¯:=σ¯m/p\bar{\tau}:=\bar{\sigma}^{m/p}. For the action induced by τ¯\bar{\tau} on the quotient space mentioned above, we must have

0=λ(τ¯)λ(τ¯p)modp.0=\lambda(\bar{\tau})\equiv\lambda(\bar{\tau}^{p})\quad\mod p.

Here, we have used the modp\mod p Lefschetz Theorem (see [3], Chap. 9, Thm. 3.4, p. 231). Thus, we arrive at the contradiction that

01=χ(Pn1)modp.0\equiv 1=\chi(\mathbb{R}P^{n-1})\quad\mod p.

Hence, the action induced by σ¯\bar{\sigma} itself fixes some point. This fixed point corresponds precisely to a pencil Γ\Gamma of geodesics as asserted.

3 Proof of Theorem 1.2.

Proof.

Select an arbitrary element σAut(Ωk)\sigma\in Aut(\Omega_{k}) having a prime order pp. By [5] (see Thm. 1.4.3, p. 22), we have

#Fix(σ)2.\#Fix(\sigma)\leq 2. (1)

Since Ω:=Ωk\Fix(σ)\Omega:=\Omega_{k}\backslash Fix(\sigma) is not compact, we shall work with generalised Lefschetz numbers (as in [3], Def. 3.1, p. 421). Note that Ω\Omega is a Lefschetz space (by [3], Chap. 5, Thm. 4.1, p. 423).

Now, the restriction of σ\sigma to Ω\Omega is an automorphism without fixed points. Therefore, the generalised Lefschetz number λ(σ)\lambda(\sigma) of σ\sigma must vanish. As Ω\Omega has the homotopy type of a wedge of (k+#Fix(σ))(k+\#Fix(\sigma)) circles, we obtain (using [3], Chap. 9, Prop. 4.7. p. 233)

χ(Ω)=1k#Fix(σ)0modp.\chi(\Omega)=1-k-\#Fix(\sigma)\equiv 0\mod p.

In view of Eq. (1), this implies that pp must divide either kk or one of k±1k\pm 1. Thus, the assertion of the Theorem follows. ∎

Acknowledgement.

The author would like to thank Chennai Mathematical Institute for support by a post-doctoral fellowship.

References

  • [1] Werner Ballmann, Mikhael Gromov, and Viktor Schroeder. Manifolds of nonpositive curvature, volume 61 of Progress in Mathematics. Birkhäuser Boston, Inc., Boston, MA, 1985.
  • [2] Patrick B. Eberlein. Geometry of nonpositively curved manifolds. Chicago Lectures in Mathematics. University of Chicago Press, Chicago, IL, 1996.
  • [3] Andrzej Granas and James Dugundji. Fixed point theory. Springer Monographs in Mathematics. Springer-Verlag, New York, 2003.
  • [4] Maurice Heins. On the number of 1-1 directly conformal maps which a multiply-connected plane region of finite connectivity p(>2)p(>2) admits onto itself. Bull. Amer. Math. Soc., 52:454–457, 1946.
  • [5] Steven G. Krantz. Geometric function theory. Cornerstones. Birkhäuser Boston, Inc., Boston, MA, 2006. Explorations in complex analysis.
  • [6] Peter Petersen. Riemannian geometry, volume 171 of Graduate Texts in Mathematics. Springer, New York, second edition, 2006.