Sharpness of Hörmander's condition $a_q$ ($Z(q)$): necessity, characterization of closed range, and minimal boundary regularity for the $L^2$ $\bar\partial$-theory on bounded domains

This problem concerns the sharpness of the foundational solvability criterion in Hörmander's 1965 paper "$L^2$ estimates and existence theorems for the $\bar\partial$ operator" (Acta Mathematica 113). That paper proved (Theorems 3.2.5 and 3.4.1), for a relatively compact domain $\Omega$ with $C^3$ boundary in a complex manifold, that the pointwise Levi-form sign condition $a_q$ — at every boundary point the Levi form has at least $n-q$ positive or at least $q+1$ negative eigenvalues, a condition now denoted $Z(q)$ — implies closed range of $\bar\partial$ in $L^2_{0,q}$, finite-dimensional $L^2$-cohomology, and Sobolev-regular solutions, while necessity of the Levi signature was proved in the paper only for the local basic estimate (Theorem 3.2.1, Corollary 3.2.3). The subsequent literature documents that the question of how sharp this criterion is remains open. Pointwise $Z(q)$ is not necessary for closed range (annuli between smooth pseudoconvex domains have closed range in every degree although the inner boundary carries the Levi signature $(0,n-1)$, which violates $Z(n-1)$; Shaw 1985, Hörmander 2004, Laurent-Thiébaut–Shaw 2013); Hörmander (2004) proved that a nondegenerate failure of $Z(q)$ at a single boundary point forces either an infinite-dimensional null space or non-closed range; Harrington–Raich (2022) obtained the first necessary conditions, but only for families of strong closed range estimates, and only in degenerate $Z(q)$-type form per boundary component; Laurent-Thiébaut–Shaw (2013) characterized degree-one closed range in $\mathbb{C}^2$ modulo complement topology. Necessity for bare closed range, a full geometric characterization of the closed-range property, and the extension of the $Z(q)$-theory from $C^3$ to $C^2$ boundaries remain open.

Background

Let $\Omega$ be a bounded, relatively compact open set whose boundary is of class $C^k$ for some finite $k$, in an $n$-dimensional complex manifold. The Cauchy–Riemann operator $\bar\partial$ is the portion of the exterior derivative that raises the bidegree of a complex differential form: it maps a $(0,q)$-form to a $(0,q+1)$-form, and solving $\bar\partial u=f$ with estimates is the central analytic problem of several complex variables. Hörmander's 1965 Acta Mathematica paper founded the $L^2$-theory of this equation: one seeks solutions $u$ in the Hilbert space $L^2_{0,q}(\Omega)$ of square-integrable $(0,q)$-forms. The structural property on which everything rests is that $\bar\partial:L^2_{0,q}(\Omega)\to L^2_{0,q+1}(\Omega)$ has closed range (that is, a closed subspace as its range). Closed range is equivalent to a priori estimates of the form $\|u\|^2\le C(\|\bar\partial u\|^2+\|\bar\partial^*u\|^2)$ modulo the null space of the $\bar\partial$-Neumann Laplacian $\square=\bar\partial\bar\partial^*+\bar\partial^*\bar\partial$ — the self-adjoint operator governing the natural boundary value problem for $\bar\partial$, where $\bar\partial^*$ is the Hilbert-space adjoint — and to the Hausdorff separation property of the $L^2$-Dolbeault cohomology, the quotient of $\bar\partial$-closed by $\bar\partial$-exact forms.

The geometry enters through the Levi form. If $\rho$ is a $C^2$ defining function for $\Omega$ (so $\partial\Omega=\{\rho=0\}$, $\nabla\rho\neq0$ there, and $\rho<0$ inside), the Levi form at a boundary point is the Hermitian form $i\partial\bar\partial\rho$ restricted to the complex tangent space, the maximal complex subspace of the real tangent space, of complex dimension $n-1$. A domain is pseudoconvex when its Levi form is positive semidefinite at every boundary point. Pseudoconvexity is sufficient — and, for solvability in every degree at once, also necessary — for the equation $\bar\partial u=f$ to be solvable in $L^2$ on a bounded domain (sufficiency is Hörmander's weighted $L^2$ method; the necessity of pseudoconvexity for simultaneous solvability in all degrees is classical and is recorded, with the 1965 attribution, by Harrington–Raich (2015)). For a single fixed degree, pseudoconvexity is no longer necessary, and it is this per-degree problem on non-pseudoconvex domains that Hörmander addressed with the condition $a_q$: at every point of $\partial\Omega$ the Levi form has at least $n-q$ positive eigenvalues or at least $q+1$ negative eigenvalues. Later literature, in the monograph treatment of the $\bar\partial$-Neumann problem, denotes this condition $Z(q)$; Chakrabarti–Harrington–Raich (2026) record that it is necessary and sufficient for $\tfrac12$-subelliptic estimates (a priori estimates in which the $\bar\partial$-Neumann operator gains half a derivative in the Sobolev scale) on $(0,q)$-forms, so it is exactly the borderline case of the theory. The borderline character follows from linear algebra: if the Levi form is nondegenerate at a point, so that its numbers $p$ of positive and $m$ of negative eigenvalues satisfy $p+m=n-1$, then $Z(q)$ fails at that point if and only if $(p,m)=(n-q-1,q)$, the signature that Hörmander (2004) later called excluded.

Hörmander's main boundary theorem (1965, Theorems 3.2.5 and 3.4.1) states that if $\partial\Omega$ is of class $C^3$ and satisfies $a_q$ at every point, then $\bar\partial$ has closed range in $L^2_{0,q}$, the null space of $\square$ on $L^2_{0,q}$ is finite dimensional, and solutions gain derivatives in the Sobolev scale (the $L^2$-based differentiability spaces $W^s$ of functions with square-integrable weak derivatives up to order $s$). Necessity was established in the paper only for the local theory: its Theorem 3.2.1 and Corollary 3.2.3 derive the Levi-eigenvalue signature from the local basic estimate, and as Harrington–Raich (2015) record, $Z(q)$ is equivalent to solvability of $\bar\partial u=f$ with $u$ in the Sobolev space $W^{1/2}$, while necessity for the global closed-range conclusion was left open. The paper's applications — Bergman-kernel asymptotics (§3.5, Theorem 3.5.1), that is, the boundary behavior of the reproducing kernel of the Hilbert space of square-integrable holomorphic functions, and a cohomology-with-bounds theorem for coverings of $\mathbb{C}^n$ by pseudoconvex sets (§2.4, Theorem 2.4.1) — take the closed-range property as a hypothesis.

The source itself poses no open-problem list; the sharpness question is documented as open by the subsequent literature, and two structural gaps in the 1965 theory anchor it. First, condition $a_q$ is defined (Definition 3.3.4 of the source) for $C^2$ defining functions — the Levi form itself is a $C^2$ notion — while the proofs of the global theorems use the implicit function theorem and orthonormal frames with controlled commutators (the frame choice preceding Proposition 3.1.3), which need $C^3$ boundaries, so the sufficiency theory sits one regularity class above the category in which its own hypothesis is defined. Second, necessity of the Levi signature is proved in the source only for the local basic estimate, and necessity for the global closed-range conclusion, together with an exact characterization of when the range of $\bar\partial$ is closed, has remained open since.

The subsequent landscape is well mapped on the sufficiency side. Shaw (1985) proved solvability and regularity on annuli between weakly pseudoconvex domains. Hörmander (2004) himself returned to the problem: he proved that for the spherical shell $\{R_0<|z|<R_1\}\subset\mathbb{C}^n$ the range of $\bar\partial$ on $L^2$ functions is closed even though the inner boundary carries the signature $(0,n-1)$ that violates $Z(n-1)$ (for $n=2$ this is the violation of $Z(1)$), computed the null space of $\square$ on $L^2_{0,n-1}$ explicitly, classified the confocal-ellipsoid shells (ellipsoids sharing the same foci) with the same multiplier property, and proved the excluded-signature theorem: a nondegenerate failure of $Z(q)$ at a single boundary point forces either an infinite-dimensional null space or non-closed range. Harrington–Raich (2015) introduced weak $Z(q)$, a metric-dependent generalization of $Z(q)$ and of $(q-1)$-pseudoconvexity (a convexity notion requiring the sum of the $q+1$ smallest eigenvalues of the Levi form to dominate its trace on a subbundle of rank at most $q$), which still implies closed range for $C^3$ boundaries. Harrington–Raich (2022) obtained the first necessary conditions, but only for families of strong closed range estimates, a stable weighted strengthening of closed range. Laurent-Thiébaut–Shaw (2013) characterized degree-one closed range in $\mathbb{C}^2$ modulo the topology of the complement, and Chakrabarti–Harrington–Raich (2026) proved vanishing theorems for $L^2$-cohomology on smooth $Z(q)$-domains carrying a smooth weight whose complex Hessian has a prescribed number of positive eigenvalues and interacts with the Levi form through a common subbundle of positive directions. On the boundary-regularity axis, the pseudoconvex side has been pushed to $C^1$ and Lipschitz (Harrington 2009; Saber 2011; Harrington–Zeytuncu 2019), and for annulus-type domains, closed range in degree zero survives non-smooth holes (Chakrabarti–Laurent-Thiébaut–Shaw 2018), with pseudoconvexity of the pieces of a Lipschitz annulus characterized spectrally by Fu–Laurent-Thiébaut–Shaw (2017); but the $Z(q)$-theory on not-necessarily-pseudoconvex domains remains at $C^3$. What no later work provides is a necessary and sufficient geometric characterization of the closed-range property, a necessity theorem for bare closed range, or a $C^2$ extension of the $Z(q)$-theory — the three facets of the problem stated below.

Problem Statement

This is the open sharpness question for the criterion of Hörmander (1965) in the $L^2$-theory of the Cauchy–Riemann operator, documented as open by the subsequent literature. Let $\Omega$ be a relatively compact domain in an $n$-dimensional complex manifold ($n\ge2$) whose boundary is of class $C^k$ for some finite $k$, and fix $1\le q\le n-1$.

(i) Necessity and characterization. Determine geometric necessary conditions for the closedness of the range of $\bar\partial:L^2_{0,q}(\Omega)\to L^2_{0,q+1}(\Omega)$. In particular: is a degenerate failure of $Z(q)$ — a boundary point where the Levi form is degenerate (some eigenvalue zero) and $Z(q)$ fails — compatible with closed range together with finite-dimensional $L^2$-cohomology, or does it force the full 1965 conclusion to fail; and if a necessary condition survives, must it be a $Z(q)$-type sign condition, or a weaker condition coupled to a weighted or potential-theoretic one as in the known necessary conditions for families of strong closed range estimates? Clarify the exact interplay between pointwise Levi geometry (including degenerate signatures, which the excluded-signature theorem of Hörmander (2004) does not cover), global topology (annuli, holes, connectedness of the complement), and the closed-range property; and characterize the closed-range property in degree $q$ geometrically by necessary and sufficient conditions.

(ii) Minimal boundary regularity. Decide whether the conclusions of Hörmander (1965), Theorems 3.2.5 and 3.4.1 — the basic estimate, closed range of $\bar\partial$ in $L^2_{0,q}$, and finite-dimensional $L^2$-cohomology under condition $a_q$ — and the representative weakening of the sufficiency theory, weak $Z(q)$, remain valid for $C^2$ boundaries, the largest category in which $a_q$ (Definition 3.3.4 of the source, stated for $C^2$ defining functions) and hence the Levi form are defined; the current theorems require $C^3$ boundaries. A $C^2$ counterexample on which one of the corresponding conclusions fails resolves (ii) negatively.

References

1. Lars Hörmander, "$L^2$ estimates and existence theorems for the $\bar\partial$ operator", Acta Mathematica 113 (1965), 89–152. DOI: 10.1007/BF02391775. URL: https://doi.org/10.1007/BF02391775.
2. J. J. Kohn, "Harmonic integrals on strongly pseudo-convex manifolds, I", Annals of Mathematics (2) 78 (1963), 112–148. DOI: 10.2307/1970506. URL: https://doi.org/10.2307/1970506.
3. J. J. Kohn, "Harmonic integrals on strongly pseudo-convex manifolds, II", Annals of Mathematics (2) 79 (1964), 450–472. DOI: 10.2307/1970404. URL: https://doi.org/10.2307/1970404.
4. Mei-Chi Shaw, "Global solvability and regularity for $\bar\partial$ on an annulus between two weakly pseudo-convex domains", Transactions of the American Mathematical Society 291 (1985), 255–267. DOI: 10.1090/s0002-9947-1985-0797058-3. URL: https://doi.org/10.1090/s0002-9947-1985-0797058-3.
5. Lars Hörmander, "The null space of the $\bar\partial$-Neumann operator", Annales de l'Institut Fourier 54 (2004), 1305–1369. DOI: 10.5802/aif.2051. URL: https://doi.org/10.5802/aif.2051.
6. Michael Christ, "Global $C^\infty$ irregularity of the $\bar\partial$-Neumann problem for worm domains", Journal of the American Mathematical Society 9 (1996), 1171–1185. DOI: 10.1090/s0894-0347-96-00213-5. URL: https://doi.org/10.1090/s0894-0347-96-00213-5.
7. Phillip S. Harrington, "Sobolev estimates for the Cauchy–Riemann complex on $C^1$ pseudoconvex domains", Mathematische Zeitschrift 262 (2009), 199–217. DOI: 10.1007/s00209-008-0369-7. URL: https://doi.org/10.1007/s00209-008-0369-7.
8. Sayed Saber, "The $\bar\partial$-Neumann operator on Lipschitz $q$-pseudoconvex domains", Czechoslovak Mathematical Journal 61 (2011), 721–731. DOI: 10.1007/s10587-011-0021-2. URL: https://doi.org/10.1007/s10587-011-0021-2.
9. Christine Laurent-Thiébaut and Mei-Chi Shaw, "On the Hausdorff property of some Dolbeault cohomology groups", Mathematische Zeitschrift 274 (2013), 1165–1176. DOI: 10.1007/s00209-012-1111-z. URL: https://doi.org/10.1007/s00209-012-1111-z.
10. Shaban Khidr and Osama Abdelkader, "The $\bar\partial$-equation on an annulus between two strictly $q$-convex domains with smooth boundaries", Complex Analysis and Operator Theory 8 (2014), 1151–1172. DOI: 10.1007/s11785-013-0309-3. URL: https://doi.org/10.1007/s11785-013-0309-3.
11. Phillip S. Harrington and Andrew Raich, "Closed range for $\bar\partial$ and $\bar\partial_b$ on bounded hypersurfaces in Stein manifolds", Annales de l'Institut Fourier 65 (2015), 1711–1754. DOI: 10.5802/aif.2972. URL: https://doi.org/10.5802/aif.2972. (Also arXiv:1106.0629.)
12. A.-K. Herbig and J. D. McNeal, "On closed range for $\bar\partial$", Complex Variables and Elliptic Equations 61 (2016), 1073–1089. DOI: 10.1080/17476933.2016.1139579. URL: https://doi.org/10.1080/17476933.2016.1139579. (Also arXiv:1410.3559.)
13. Siqi Fu, Christine Laurent-Thiébaut, and Mei-Chi Shaw, "Hearing pseudoconvexity in Lipschitz domains with holes via $\bar\partial$", Mathematische Zeitschrift 287 (2017), 1157–1181. DOI: 10.1007/s00209-017-1863-6. URL: https://doi.org/10.1007/s00209-017-1863-6. (Also arXiv:1609.07454.)
14. Debraj Chakrabarti, Christine Laurent-Thiébaut, and Mei-Chi Shaw, "On the $L^2$-Dolbeault cohomology of annuli", Indiana University Mathematics Journal 67 (2018), 831–857. DOI: 10.1512/iumj.2018.67.7307. URL: https://doi.org/10.1512/iumj.2018.67.7307. (Also arXiv:1602.03896.)
15. Phillip S. Harrington and Yunus E. Zeytuncu, "$L^p$ mapping properties for the Cauchy–Riemann equations on Lipschitz domains admitting subelliptic estimates", Complex Variables and Elliptic Equations 64 (2019), 369–385. DOI: 10.1080/17476933.2018.1434628. URL: https://doi.org/10.1080/17476933.2018.1434628.
16. Phillip Harrington and Andrew Raich, "Strong closed range estimates: necessary conditions and applications", Indiana University Mathematics Journal 71 (2022), 1047–1078. DOI: 10.1512/iumj.2022.71.8926. URL: https://doi.org/10.1512/iumj.2022.71.8926. (Also arXiv:1904.09345.)
17. Phillip S. Harrington and Andrew Raich, "Boundary invariants and the closed range property for $\bar\partial$", Differential Geometry and its Applications 82 (2022), article 101871. DOI: 10.1016/j.difgeo.2022.101871. URL: https://doi.org/10.1016/j.difgeo.2022.101871. (Also arXiv:1805.05793.)
18. Phillip S. Harrington and Andrew Raich, "Maximal estimates for the $\bar\partial$-Neumann problem on non-pseudoconvex domains", The Journal of Geometric Analysis 34 (2024), article 237. DOI: 10.1007/s12220-024-01684-9. URL: https://doi.org/10.1007/s12220-024-01684-9. (Also arXiv:2201.02679.)
19. Debraj Chakrabarti, Phillip S. Harrington, and Andrew Raich, "The $\bar\partial$-problem on $Z(q)$-domains", Pacific Journal of Mathematics 342 (2026), 1–35. DOI: 10.2140/pjm.2026.342.1. URL: https://doi.org/10.2140/pjm.2026.342.1. (Also arXiv:2403.04756.)