Dressing orbits of harmonic maps (Q1913362): Difference between revisions

In recent years some progress in the theory of harmonic maps from a Riemann surface to a compact Riemannian symmetric space has come from the systematic use of soliton methods, and in particular of a ``spectral parameter'' = ``loop parameter''. This allows one to rephrase the harmonic map equations in terms of one equation, a zero-curvature condition (for the associated ``extended framings''). Another advantage of the use of the loop parameter is the possibility of considering deformations of harmonic maps by ``dressing''. As a matter of fact, from work of the reviewer, \textit{F. Pedit} and \textit{H. Wu} [Commun. Anal. Geom. 6, No. 4, 633-668 (1998; Zbl 0932.58018)] it is known that the class \(\mathcal H\) of harmonic maps from a given Riemann surface to a compact Riemannian symmetric space falls into many (infinite-dimensional) dressing orbits relative to the positive loop group \(G_+\). Some of these orbits have received almost all of the attention since many of their corresponding harmonic maps are of ``finite type'', i.e., they can be constructed by algebro-geometric means, Lax equations or the Adler-Kostant-Symes scheme. The reviewer and \textit{H. Wu} [J. Reine Angew. Math. 440, 43-76 (1993; Zbl 0779.53004)] observed that one can obtain all tori of constant mean curvature by dressing from the standard cylinder, if one works not with one, but with two parallel circles on the Riemann sphere. The very interesting and well-written paper under review discusses Lax equations and dressing from ``vacuum solutions'' for ``primitive'' harmonic maps \(\mathbb{R}^2\to G/U\), where \(G/U\) is a (\(k\)-)symmetric space. In this case the ``vacuum solutions'' are the maps \(f^A\colon\mathbb{R}^2\to G/U\) defined by \(f^A(z)=\exp(zA+\overline{zA})K\), where \(A\) is an element of the complexification of the Lie algebra of \(G\) satisfying \([A,\overline A]=0\). The paper gives a careful introduction to primitive harmonic maps and their extended framings in a double loop group setting. This includes Iwasawa-type decompositions and dressing. The main results of the paper concern the orbit \(O_A\) of \(f^A\) under the dressing action. It is shown that every primitive harmonic map of ``semisimple finite type'' lies in some \(O_A\). Here a harmonic map is said to be of finite type if it can be constructed from a polynomial Killing field as in a paper by \textit{F. E. Burstall, D. Ferus, F. Pedit} and \textit{U. Pinkall} [Ann. Math., II. Ser. 138, No. 1, 173-212 (1993; Zbl 0796.53063)]. The harmonic map is said to be of semisimple finite type if the lowest-order term of this polynomial Killing field is always semisimple. As a corollary to the main result, the authors obtain the following: Every nonisotropic harmonic torus in a sphere or complex projective space is dressing equivalent to a vacuum solution. Finally, it is shown that the orbits \(O_A\) admit a hierarchy of commuting flows and that a harmonic map in \(O_A\) is of finite type iff its orbit under the hierarchy is finite-dimensional.