The deformed Hermitian-Yang-Mills equation, the Positivstellensatz, and the solvability (Q6058888)

The deformed Hermitian-Yang-Mills equation is an analogue of the special Lagrangian graph equation in the context of Kähler geometry. It is a second-order nonlinear equation for a Hermitian metric on a line bundle over a given Kähler manifold \((X, \omega)\), of the form \[ \arctan \lambda_i = \hat\theta, \] where \(\lambda_i\) are the eigenvalues for the curvature \(\chi\) of the Hermitian metric, relative to the given Kähler metric. This equation fits into a family of Hessian-type equations in Kähler geometry, which has recently received much attention. The behaviour of the equation depends on the range of the phase angle \(\hat\theta\), and the natural generality is \(\hat\theta > \frac{(n-2)\pi}{2}\), known as supercritical phase. Some notable results in this field are due to \textit{T. C. Collins} et al. [Camb. J. Math. 8, No. 2, 407--452 (2020; Zbl 1442.14124)], who established a formal infinite-dimensional GIT picture for this equation, related the question to algebro-geometric stability conditions, and proved the existence of the solution under the assumption of the existence of a \(C\)-subsolution (a notion originally introduced by Székelyhidi). The recent important work of \textit{G. Chen} [Invent. Math. 225, No. 2, 529--602 (2021; Zbl 1481.53093)] proves the existence result under a uniform variant of the stability condition. A technical gap in the literature is when \(C\)-subsolutions are supercritical; in the work of Collins et al. [loc. cit.] there was a non-optimal assumption that \(\hat\theta\geq (n - 2 + \frac2n) \frac\pi2\). The goal of this paper is to improve this to the natural generality \(\hat\theta > \frac{(n-2)\pi}{2}\), for complex dimension \(3, 4\). The strategy involves finding a continuity path which goes outside of the deformed Hermitian Yang-Mills equation, into more general Hessian-type equations, and prove the same kind of a priori estimates. One needs to technically generalize the notion of \(C\)-subsolutions, which the author encodes into the notion of \(\Upsilon\)-cones, defined iteratively by inequalities on the eigenvalues, and morally amounting to taking cones over cones. The paper has a heavily computational flavour.