Compactness results for triholomorphic maps
Costante Bellettini,Gang Tian
DOI: https://doi.org/10.4171/JEMS/860
2019-01-01
Journal of the European Mathematical Society
Abstract:We consider triholomorphic maps u from an almost hyper-Hermitian manifold M-4m into a (simply connected) hyperKahler manifold N-4n . This entails that the map u is an element of W-1,W-2 satisfies a quaternionic del-bar equation. We work under the assumption that u is locally strongly approximable in W-1,W-2 by smooth maps; such maps are almost stationary harmonic in a suitable sense (in the important special case that M is hyperKahler as well, they are stationary harmonic). We show, by means of the bmo-h(1) duality, that in this more general situation the classical epsilon-regularity result still holds and we establish the validity, for triholomorphic maps, of the W-2,W-1-conjecture (i.e. an a priori W-2,W-1-estimate in terms of the energy). We then address compactness issues for a weakly converging sequence u(l) -> u(infinity) of strongly approximable triholomorphic maps u(l) : M -> N with uniformly bounded Dirichlet energies. The blow-up analysis leads, as in the usual stationary setting, to the existence of a rectifiable blow-up set Sigma of codimension 2, away from which the sequence converges strongly. The defect measure Theta (x)H4m-2 left perpendicular Sigma encodes the loss of energy in the limit and we prove that for a.e. point on Sigma the value of Theta is given by the sum of the energies of a (finite) number of smooth non-constant holomorphic bubbles (here the holomorphy is understood with respect to a complex structure on N that depends on the chosen point on Sigma). When M is hyperKahler this quantization result was established by C. Y. Wang [41] with a different proof; our arguments rely on Lorentz spaces estimates. By means of a calibration argument and a homological argument we further prove that whenever the restriction of Sigma to an open ball B subset of M \ Sing(u infinity) is covered by a connected (4m - 2)-dimensional Lipschitz graph, then actually Sigma boolean AND B is a smooth submanifold and it is pseudo-holomorphic for a (unique) almost complex structure on M (with Theta constant on this portion); moreover the bubbles originating at points of Sigma boolean AND B are all holomorphic for a common complex structure on N.