# Harmonictoriinspheresandcomple feb/papers/sphere/paper.pdfآ maps of 2-tori as doubly periodic...

date post

01-Aug-2020Category

## Documents

view

0download

0

Embed Size (px)

### Transcript of Harmonictoriinspheresandcomple feb/papers/sphere/paper.pdfآ maps of 2-tori as doubly periodic...

Harmonic tori in spheres and complex projective spaces

F. E. Burstall

Introduction

A map φ : M → N of Riemannian manifolds is harmonic if it extremises the energy functional: ∫

|dφ|2 dvol

on every compact subdomain of M . Harmonic maps arise in many different contexts in Geometry and Physics (for an overview, see [16,17]) but the setting of concern to us is the following: take M to be 2-dimensional and N to be a Riemannian symmetric space of compact type. In this case, the energy is conformally invariant so that we may take the domain to be a Riemann surface and the methods of complex analysis may be brought to bear. Moreover, the symmetric nature of the target allows us to reformulate the harmonic map equations in a gauge-theoretic way so that harmonic maps may be viewed as simple analogues of Yang–Mills fields.

This paper treats harmonic maps of a 2-torus into a sphere Sn or a complex projective space CPn and makes use of the ideas and methods of two separate developments in the theory of harmonic maps. The first, and more recent, of these is the soliton-theoretic approach which has its roots in the Pinkall–Sterling classification [24] of constant mean curvature 2-tori in R3 (which are equivalent, via the Gauss map, to non-conformal harmonic 2-tori in S2). Pinkall–Sterling showed that all such maps could be constructed from solutions to a family of finite-dimensional completely integrable Hamiltonian ordinary differential equations. There followed a rapid development and extension of these ideas [5,20] which culminated in a rather general theory of harmonic maps into symmetric spaces due to Burstall–Ferus–Pedit–Pinkall [9]. This theory distinguishes special harmonic maps of R2

into a symmetric space called harmonic maps of finite type which are constructed from commuting Hamiltonian flows on finite-dimensional subspaces of a loop algebra. Viewing maps of 2-tori as doubly periodic maps of R2, these authors prove:

Theorem A non-conformal harmonic map of a 2-torus into a rank one symmetric space is of finite type.

In particular, this result accounts for all non-conformal harmonic 2-tori in Sn and CPn

but excludes the conformal harmonic (i.e., branched, minimal) tori.

The second development of importance to us is the well-established twistor theory of har- monic maps which goes back to Calabi’s study [11,12] of minimal surfaces and, especially, minimal 2-spheres in Sn. Recall [12,18] that a harmonic map of a Riemann surface into

Research partially supported by EEC contract SC1-0105-C

1

a sphere or complex projective space has a sequence of invariants which are differentials measuring the lack of orthogonality of iterated derivatives of the map. These invariants have the following properties:

1. The first invariant is the obstruction to conformality.

2. The first non-zero invariant is a holomorphic differential.

3. There is a number N (depending on the target) such that, if the first N invariants vanish, then all the invariants vanish.

If all these invariants vanish, the map is variously called pseudo-holomorphic [11], super- minimal [7], or isotropic [18]. In this case, the harmonic map is covered by a horizontal holomorphic map into an auxiliary complex manifold, a twistor space, and the study of isotropic harmonic maps is therefore reduced to a problem in Algebraic Geometry.

In particular, since the Riemann sphere admits no non-vanishing holomorphic differentials, any harmonic 2-sphere in Sn or CPn is isotropic. This is the basis of Calabi’s classification of minimal 2-spheres in Sn [11,12] and the classification theorem for harmonic 2-spheres in CPn [8,14,18,21].

We see from these results that, as far as harmonic 2-tori are concerned, we can treat two extremes of behaviour: on the one hand, when the first invariant is non-zero, we obtain the harmonic map from soliton-theoretic ODE; on the other hand, when all invariants vanish, we obtain the harmonic map from a holomorphic curve in a twistor space. It is the purpose of this article to treat all the intermediate cases and thus account for all harmonic 2-tori in spheres and projective spaces. Our main results are Theorems 3.4 and 4.4 which can be briefly summarised as follows: a conformal non-isotropic map is covered by a map into a twistor space and this map, instead of being holomorphic, is constructed from soliton-theoretic ODE.

That such a picture obtains is already indicated by the analysis by Ferus–Pedit–Pinkall– Sterling of minimal, non-superminimal tori in S4 [20] which was generalised to Sn and CPn

by Bolton–Pedit–Woodward [6]. In these papers, the harmonic maps under consideration are characterised by the vanishing of all invariants except the last. In this case, the harmonic map is covered by a map into a flag manifold which can be shown to arise from commuting Hamiltonian ODE. One may view the flag manifold as a twistor space and the covering map, while no longer horizontal holomorphic, satisfies a first order condition.

The situation is similar for all non-isotropic 2-tori in spheres and complex projective spaces: any such map is covered by a “twistor lift” which is a map into a homogeneous space. Moreover, this lift can be constructed from commuting ODE and so is, in an appropriate sense, of finite type. The lift is constructed from iterated derivatives of the harmonic map by the usual method of twistor theory. The only novelty here is the nature of our twistor space and the map into it so obtained. Our twistor spaces are k-symmetric spaces which are analogues of the familiar Riemannian symmetric spaces where where the involutive isometries are replaced by isometries of finite order k [23]. The k-symmetric spaces form a class of reductive homogeneous spaces that includes both Riemannian symmetric spaces (k = 2) and flag manifolds. The use of the latter as twistor spaces for Riemannian symmetric spaces is well-known [10] but our use of other k-symmetric spaces in this way seems to be new.

2

As for the twistor lifts, these are examples of what we have called primitive maps: a map into a k-symmetric space, k > 2, is primitive if it satisfies a first-order equation of Cauchy– Riemann type which arises from the geometry of the k-symmetric space. Primitive maps are maps which are “f -holomorphic with respect to a horizontal f -structure” in the sense of Black [4] and, as such, enjoy a number of interesting properties: they are harmonic maps and their harmonicity is preserved under homogeneous projection. Thus primitive maps project onto harmonic maps. In our case, the converse is true: we show that any conformal harmonic map of a Riemann surface into Sn or CPn is covered by a primitive map into a suitable k-symmetric space.

To construct such maps from commuting ODE, we make use of the results of [9]: recall that a harmonic map φ : R2 → G into a Lie group is essentially the same as a loop of flat connections d +Aλ with Aλ of the form

Aλ = (λ− 1)α ′ + (λ−1 − 1)α′′,

λ ∈ S1, where α′ is a gC -valued (1, 0)-form on R2 with complex conjugate α′′. In [9], a method is given for constructing such loops of flat connections by solving commuting ODE on finite-dimensional subspaces of the based loop algebra. To treat harmonic maps into a Riemannian symmetric space G/K, one uses the totally geodesic Cartan embedding G/K →֒ G. The composition of a harmonic map into G/K with the Cartan embedding is also harmonic and so gives rise to a loop of flat connections. Moreover, if appropriate initial conditions are chosen for the commuting ODE, one constructs from the flows a harmonic maps which factors through the Cartan embedding in this way.

It transpires that we are able to treat primitive maps ψ : R2 → G/H into a k-symmetric space in a similar fashion. Firstly, there is an obvious analogue of the Cartan embedding G/H → G of any k-symmetric space into its group of isometries. For k > 2, this map is no longer totally geodesic so that its composition with a primitive map need not be harmonic. However, the structure equations for a primitive map are identical to those for a harmonic map into a Riemannian symmetric space and this enables us to construct a loop of flat connections. Moreover, a simple extension of the arguments of [9] suffices to show that commuting flows with the right initial conditions give rise to primitive maps.

We therefore arrive at the notion of a primitive map of finite type. The main result of [9] gives a simple criterion for a loop of flat connections on a torus to arise from commuting ODE from which we can deduce sufficient conditions for a primitive map to be of finite type. In our applications to twistor lifts, these conditions amount to the non-vanishing of one of the invariants of the underlying harmonic torus.

We have developed the theory of primitive maps of finite type in rather more generality than is necessary for our applications to harmonic tori in spheres and complex projective spaces. We have done this for two reasons: firstly, this theory provides the natural frame- work for our results and, secondly, because we hope it will find applications elsewhere. In this latter regard, it is already known that primitive maps into a full flag manifold are closely related to affine Toda fields [6].

Acknowledgements I wish to thank J. Bolton, F. Pedit and L. Woodward for informing me of their work [6] and J.-P. Bourguignon for suggesting the term “primitive” for the maps with which we shall be concerned below.

3

Notation Throughout this work, when a Lie group is denoted by an upper case letter, its Lie algebra will be denoted

*View more*