The Monge-Ampère equation for non-integrable almost complex structures (Q2002832)

Let \((M^{2n}, J, g)\) be a compact almost Hermitian \(2n\)-manifold and denote by \(\omega = g(J \cdot, \cdot)\) its associated Kähler form. By Yau's Theorem, if \(J\) is integrable and \((M, J, g)\) is Kähler, for any smooth volume form \(\text{vol}^{(F)} = e^F \omega^n\) satisfying the normalisation condition \(\int_M \text{vol}^{(F)} = \int_M \omega^n\), there exists a unique smooth function \(\varphi\) satisfying the conditions \[ (\omega + \sqrt{-1} \partial \bar \partial \varphi)^n = \text{vol}^{(F)},\qquad \omega + \sqrt{-1} \partial \bar \partial \varphi > 0,\qquad \sup_M \varphi = 0\tag{\(\ast\)}.\] The result has been generalised by the second and third author in [J. Amer. Math. Soc. 23, 19--40 (2010; Zbl 1208.53075)] to arbitrary Hermitian manifolds up to addition to \(F\) of a (uniquely determined) constant \(b\). In this paper this result is proved to be true on arbitrary compact almost Hermitian manifolds, provided that the operator \(\varphi \mapsto \sqrt{-1} \partial \bar \partial \varphi\) is replaced by the operator \(\varphi \mapsto \frac{1}{2}(d (J d \varphi))^{(1,1)}\), where \((\cdot)^{(1,1)}\) denotes the natural pointwise projection onto the \(2\)-forms of bidegree \((1,1)\) with respect to the complex structure \(J_x\), \(x \in M\). Note that if \(J\) is integrable, this reduces to the classical definition of the \(\partial \bar \partial\) operator. The result is crucially based on the following theorem, whose proof represents the hardest part of the whole paper: ``Given a compact almost Hermitian manifold \((M, J, g)\), for any pair of smooth real functions \(F\) and \(\varphi\) satisfying \((\ast)\), there exist a priori \(\mathcal C^\infty\) estimates on \(\varphi\) depending only on \((M, J, g)\) and bounds for \(F\).'' The main result of this paper provides a positive solution to a problem posed by Gromov, provided that \(\partial \bar \partial \varphi\) is understood as \((\frac{1}{2} d (J d \varphi))^{(1,1)}\). The original statement of the problem, where \(\partial \bar \partial \varphi\) is understood as \(\frac{1}{2} d (J d \varphi)\), is known to have a negative answer by the results of \textit{P. Delanoë} [Osaka J. Math. 33, No. 4, 829--846 (1996; Zbl 0878.53030)] and \textit{M. Warren} and \textit{Y. Yuan} [Commun. Pure Appl. Math. 62, No. 3, 305--321 (2009; Zbl 1173.35388)].