INSTANTONS, RATIONAL MAPS, AND HARMONIC MAPS Martin. A. Guest 2 In these lectures I shall discuss the space Hol(S ,Fk) of holomorphic maps from the Rie- 2 mann sphere S to the space Fk, where Fk is a certain compact (singular) projective variety. Such holomorphic maps are automatically “rational”. The spaces Fk are subvarieties of the infinite dimensional manifold ΩUn, the loop group of the unitary group. In fact they form a filtration of the space of algebraic loops, the “Mitchell-Segal filtration”. Because of this, it 2 turns out that Hol(S ,Fk) plays a role in problems of gauge theory and topology. Lecture I is a brief survey of the problems in Yang-Mills theory and the theory of harmonic 2 maps which motivate the discussion of Hol(S ,Fk). Lecture II gives some basic properties of 2 Fk and Hol(S ,Fk). Applications to instantons are presented in Lecture III, and to harmonic maps in Lecture IV. 2 In Lecture III I shall show that the space Hol(S ,Fk) is a “good approximation” to the 2 2 space Map(S ,Fk) of continuous maps from S to Fk (at least in the case n = 2, and for basepoint preserving maps of a fixed degree). By a correspondence of M. F. Atiyah and S. K. Donaldson, this result may be translated into an approximation theorem for instantons, and in fact re-proves and extends an earlier result of M. F. Atiyah and J. D. S. Jones. 2 In lecture IV I shall discuss harmonic maps from S to Un, which, via a formulation of 2 K. Uhlenbeck and G. Segal, correspond to certain holomorphic maps from S to Fk. I shall show how the natural action of the (complex) loop group on such harmonic maps agrees with an apparently more complicated action, the “dressing action”, which arose earlier from the theory of integrable systems. Combining this with some ideas from elementary Morse Theory, one can obtain nontrivial deformations of harmonic maps. I shall illustrate this by giving a simple proof that the space of harmonic maps from S2 to S4 of fixed energy is a path connected space. In describing the spaces Fk I am reporting on work of S. Mitchell, A. Pressley, W. Richter, 2 and G. Segal. My interest in Hol(S ,Fk) owes much to discussions with A. Pressley. The results on the dressing action in Lecture IV are joint work with Y. Ohnita, and the idea of using the dressing action to obtain results on the connectivity of spaces of harmonic maps came from discussions with N. Ejiri and M. Kotani. This work was supported by the Japan Society for the Promotion of Science and the U.S. National Science Foundation. I am very grateful to the faculty, staff, and students of Tokyo Metropolitan University for their kind assistance during the academic year 1990/1991. Typeset by AMS-TEX 1 2 MARTIN. A. GUEST Lecture I: Instantons, rational maps, and harmonic maps §1.1 The moduli space of instantons. 4 Let P → S be a principal G-bundle, with c2(P ) = −d, where G is a compact simple Lie group and d is a nonnegative integer. Let Ad be the space of (smooth) connections ∇ in P → S4. The Yang-Mills functional is defined by Z 2 YM : Ad → R,YM(∇) = ||F (∇)|| S4 where F (∇) is the curvature of ∇. The critical points of YM (“Yang-Mills connections”) are the solutions of the equation (1) d∗F (∇) = 0. It is not easy to find solutions to this equation. However, the critical points of YM which are absolute minima (“Yang-Mills instantons”) turn out to be the solutions of the simpler equation (2) ∗F (∇) = F (∇). Let Id be the subset of Ad consisting of solutions to (2). The “based gauge group” G is the group of all automorphisms of P → S4 which are the identity over ∞ ∈ S4. This acts freely on Ad and on Id, and the submanifold Md = Id/G of Cd = Ad/G is called the moduli space of (framed) G-instantons of charge d over S4. All these concepts extend to principal G-bundles P → M 4 where M 4 is a compact oriented Riemannian manifold of dimension 4. Since the work of Donaldson in 1984 we know that the 4 corresponding moduli space Md(M ,G) is a fundamental object which reflects deep properties of M 4 and G. However, in these lectures, we shall be concerned mainly with the simplest 4 4 4 case M = S ,G = SU2; even in this case the moduli space Md = Md(S , SU2) is quite nontrivial! §1.2 Basic properties of the moduli space of instantons. 4 The first interesting examples of SU2-instantons on S were obtained by the “’t Hooft 4 construction” (around 1977): given a collection {q1, . , qd} of distinct points in R , it is possible to write down an explicit solution of (2). Shortly afterwards, Atiyah, Hitchin and 4 Singer showed that Md(S ,G) is a finite dimensional manifold, and computed its dimension. 4 For example, dim Md(S , SUn) = 4nd. Then Atiyah, Drinfeld, Hitchin and Manin gave a linear algebraic description of all the solutions to (2) (the “ADHM construction”), for the compact simple classical groups. What can be said about the space Md? For d = 1, it is known that there is a diffeomor- phism ∼ 5 M1 = SO3 × {x ∈ R | ||x|| < 1}. For d = 2 (see [Ha],[Hh],[Au]) there is a homotopy equivalence 5 M2/SU2 ' Gr2(R ). 4 (In general, G acts on Md(M ,G), although the action is not usually free.) For any d, it is ∼ known that Md is connected (see [Ta1]), and it is known that π1Md = Z/2Z (see [Hu1]). However, it is not very easy to get this kind of explicit topological (or geometrical) information about Md from the ADHM construction. INSTANTONS, RATIONAL MAPS, AND HARMONIC MAPS 3 §1.3 Morse theoretic principles. A different way of obtaining topological information about Md is suggested by Morse the- ory. We have a functional YM : Cd → R, whose critical points are the Yang-Mills connections, and for which Md is the set of absolute minima. This can be compared with another, simpler, situation. If P, Q are two points of a Riemannian manifold M, we have the energy functional E :ΩM → R on (a component of) the set of smooth paths from P to Q. The critical points of E are the geodesics connecting P and Q, and of course the shortest geodesics constitute the absolute minima of E. It is a classical fact that Morse theory applies to E. A simple consequence of Morse theory is that if Mmin denotes the set of shortest geodesics (in a fixed component), then the induced maps HiMmin → HiΩM in homology, and πiMmin → πiΩM in homotopy, are isomorphisms for i < n, where n + 1 is a lower bound for the index of any nonminimal critical point. For example, if M = Sn and P, Q are the north and south poles of the n-sphere, then n ∼ n−1 n−1 n (S )min = S and we obtain an inclusion S → ΩS which induces isomorphisms in dimensions less than 2n − 3. This is the “Freudenthal Suspension Theorem”. A direct generalization of this example can be made when M is a compact symmetric space: Mmin is also a compact symmetric space (for suitable P, Q), and in 1958 Bott used the equivalence πiMmin → πiΩM to obtain his famous “Periodicity Theorem”. Now, the classical Morse theory does not apply to YM in the same way that it applies to E, but there are some formal analogies betwen the two situations. In the version of the Yang-Mills equations for a principal G-bundle P → M 2, where M 2 is a compact oriented Riemann surface, Atiyah and Bott made this analogy very precise. This suggests the first basic principle: 4 4 APPROXIMATION PRINCIPLE: Md(M ,G) should approximate Cd(M ,G) in homology and homotopy, up to some dimension which increases with d. In the case of E :ΩM → R, this approximation principle is valid and it allows one to study 4 ΩM by studying the simpler space Mmin. In the case of YM : Cd(M ,G) → R the emphasis 4 4 is switched, as it is Cd(M ,G) which is simpler than Md(S ,G). In fact: 4 3 ∗ 3 Proposition [AJ]. The space Cd(M ,G) has the homotopy type of ΩdG = Mapd(S ,G). ∗ 3 3 Here, Mapd(S ,G) denotes the set of smooth (or continuous) maps f : S → G such that ∼ f(∞) = e (the identity element of G), and such that in π3G = Z the class [f] corresponds 3 3 ∼ ∼ 3 ∼ to d. Note that ΩdG is connected, and that π1ΩdSU2 = π4SU2 = π4S = Z/2Z, so the approximation principle is at least consistent with the information given earlier on Md. The second basic principle (again from the analogy with geodesics) is: 4 LOWEST INDEX PRINCIPLE: The space Md(M ,G) of minimal (index zero) Yang-Mills 4 connections should “generate” all Yang-Mills connections in Cd(M ,G). This was shown to hold for the Yang-Mills problem over a Riemann surface, in the work of Atiyah and Bott just mentioned. For the Yang-Mills problem over M 4, very little is known about nonminimal Yang-Mills connections; in fact, existence of such connections for 4 4 M = S ,G = SU2 was proved only recently, by Sibner, Sibner, and Uhlenbeck and by Sadun and Segert, as well as by Parker in the case of a perturbed metric.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages28 Page
-
File Size-