[Paper Review] Subelliptic Estimates

David Catlin, John P. D’Angelo|ArXiv.org|May 29, 2009

Geophysics and Gravity Measurements5 references4 citations

TL;DR

This paper establishes a precise geometric characterization of the optimal subelliptic estimate parameter $\epsilon$ for the $\overline{\partial}$-Neumann problem on $(0,1)$-forms in complex dimension three. Using Catlin's method of constructing plurisubharmonic functions with large Hessian and Kohn's subelliptic multipliers, the authors show that $\epsilon$ is determined by the order of contact of complex curves with the boundary, and construct explicit domains where $\epsilon$ achieves any prescribed value in $(0, \frac{1}{4}]$. The key contribution is a complete solution to the subelliptic estimate problem in three dimensions via geometric analysis.

ABSTRACT

We discuss some aspects of the theory of subelliptic estimates.

Motivation & Objective

• To clarify the relationship between the geometry of the boundary and the largest possible $\epsilon$ in subelliptic estimates for the $\overline{\partial}$-Neumann problem.
• To resolve the long-standing problem of determining the optimal $\epsilon$ in three complex dimensions, where the theory is significantly more complex than in two dimensions.
• To demonstrate that for any $\epsilon_0 \in (0, \frac{1}{4}]$, there exists a smooth pseudoconvex domain in $\mathbb{C}^3$ where the subelliptic estimate holds precisely up to $\epsilon_0$.
• To provide explicit constructions of defining functions using holomorphic functions with controlled log-ratios to realize arbitrary $\epsilon_0$ values.
• To extend the understanding of subelliptic estimates beyond the two-dimensional case, where the theory is simpler and commutator-based.

Proposed method

• The authors use Catlin's method of constructing uniformly bounded, plurisubharmonic functions $\Phi_\delta$ whose complex Hessian blows up like $\delta^{-2\epsilon}$ near the boundary, which implies a subelliptic estimate of order $\epsilon$.
• They analyze the order of contact $\mathbf{T}$ between complex curves $\gamma_t(\zeta)$ and the boundary defined by $r = 0$, where $r$ is a defining function involving holomorphic functions $f$ and $g$ with $\log|f| = \lambda \log|g|$.
• The parameter $\epsilon$ is shown to be the reciprocal of the order of contact $\mathbf{T} = 2m_1 + \frac{2(1-\lambda)m_1(m_2-1)}{(m_2-1)\lambda + 1}$, derived from the asymptotic behavior of $|r(\gamma_t(\zeta))|$ as $|\zeta| \to 0$.
• By choosing $f(z) = z^p$, $g(z) = z^q$, and $\lambda = p/q$, they construct polynomial defining functions that realize rational $\epsilon_0 = 1/\mathbf{T}$.
• For irrational $\epsilon_0$, they use functions vanishing to infinite order, such as $f(\zeta) = \exp(-p/\sqrt{-\zeta})$, to achieve arbitrary real $\epsilon_0 \in (0, \frac{1}{4}]$.
• The sufficiency of the Hessian growth condition is established via Theorem 7.1, which links $H(\Phi_\delta) \geq c\delta^{-2\epsilon}$ to the existence of a subelliptic estimate of order $\epsilon$.

Experimental results

Research questions

• RQ1 What is the maximal possible $\epsilon$ for which a subelliptic estimate holds at a boundary point in $\mathbb{C}^3$, and how does it relate to the geometry of the boundary?
• RQ2 Can one construct smooth pseudoconvex domains in $\mathbb{C}^3$ such that the subelliptic estimate holds exactly up to a prescribed $\epsilon_0 \in (0, \frac{1}{4}]$?
• RQ3 How does the order of contact $\mathbf{T}$ of complex curves with the boundary determine the optimal $\epsilon$ in subelliptic estimates?
• RQ4 Can the subelliptic estimate fail at $\epsilon_0$ even when it holds for all $\epsilon < \epsilon_0$, and how is this behavior realized geometrically?
• RQ5 To what extent can the subelliptic estimate parameter $\epsilon$ be controlled by the choice of holomorphic functions in the defining function of the domain?

Key findings

• For any $\epsilon_0 \in (0, \frac{1}{4}]$, there exists a smooth pseudoconvex domain in $\mathbb{C}^3$ such that the subelliptic estimate holds with $\epsilon = \epsilon_0$ but fails for any larger $\epsilon$, demonstrating sharpness of the estimate.
• The optimal $\epsilon$ is the reciprocal of the order of contact $\mathbf{T} = 2m_1 + \frac{2(1-\lambda)m_1(m_2-1)}{(m_2-1)\lambda + 1}$, which depends on the homogeneity degrees $m_1, m_2$ and the ratio $\lambda = \log|f| / \log|g|$.
• When $\lambda = 1$, $\mathbf{T} = 2m_1$, so $\epsilon = 1/(2m_1)$; when $\lambda \to 0$, $\mathbf{T} \to 2m_1m_2$, so $\epsilon \to 1/(2m_1m_2)$, showing the full range of possible $\epsilon$ values.
• For rational $\epsilon_0$, the domain can be defined by a polynomial defining function $r = |f|^{2m_1} - |g|^{2m_2}$ with $f(z) = z^p$, $g(z) = z^q$, and $\epsilon_0 = 1/\mathbf{T}$.
• The method of weighted $L^2$ estimates using $\Phi_\delta$ with $H(\Phi_\delta) \geq c\delta^{-2\epsilon}$ provides a sufficient condition for subelliptic estimates of order $\epsilon$, and this condition is sharp in the constructed examples.
• The failure of effectiveness in Kohn’s algorithm for finding subelliptic multipliers is demonstrated via a counterexample (Proposition 4.4), showing that the algorithm does not always yield the optimal $\epsilon$.