Several Complex Variables: Foundations

Job: Hartogs extension, domains of holomorphy, the \(\bar\partial\)-equation, and Oka–Cartan coherence — the local-to-global analytic toolkit for \(n\ge 2\).

Core

Holomorphic functions on \(\mathbb{C}^n\)

\(f:U\to\mathbb{C}\) (\(U\subset\mathbb{C}^n\) open) is holomorphic if it is holomorphic in each variable separately (then automatically \(C^\infty\) and locally given by convergent power series — Osgood / Hartogs theorem on separate analyticity under mild hyp., e.g. local boundedness).

Cauchy integral on polydiscs; identity theorem: agreement on a nonempty open set \(\Rightarrow\) agreement on connected components. Zero sets of nonconstant \(f\) are thin (complex codimension \(1\) analytic sets) — never isolated for \(n\ge 2\) unless \(f\equiv 0\).

Wirtinger operators \(\partial/\partial z_j\), \(\partial/\partial\bar z_j\); holomorphy \(\Leftrightarrow\) \(\bar\partial f=0\).

Hartogs’ extension phenomenon

Theorem (Hartogs). Let \(n\ge 2\), \(\Omega\subset\mathbb{C}^n\) a domain, \(K\subset\Omega\) compact with \(\Omega\setminus K\) connected. Then every \(f\in\mathcal{O}(\Omega\setminus K)\) extends uniquely to \(\tilde f\in\mathcal{O}(\Omega)\).

Classical figure. Holomorphic functions on a nbhd of the Hartogs figure in the bidisc extend to the full bidisc. Consequently \(\mathcal{O}(\mathbb{C}^n\setminus\{0\})=\mathcal{O}(\mathbb{C}^n)\) for \(n\ge 2\) — no isolated singularities.

Removable sets. Analytic sets of codimension \(\ge 1\) (with codim \(\ge 2\) for bounded holom. functions, Riemann / Hartogs-type removability) are often removable for holomorphic or meromorphic extension under growth control.

Domains of holomorphy

\(\Omega\subset\mathbb{C}^n\) is a domain of holomorphy if there is no larger domain \(\Omega'\supsetneq\Omega\) to which every \(f\in\mathcal{O}(\Omega)\) extends. Equivalently (Cartan–Thullen): \(\Omega\) is holomorphically convex — for compact \(K\subset\Omega\), the holomorphic hull $$ \hat K_\Omega={z\in\Omega:|f(z)|\le\sup_K|f|\ \forall f\in\mathcal{O}(\Omega)} $$ is compact in \(\Omega\).

Notion Role
Holomorphic convexity \(K\mapsto\hat K\) compact in \(\Omega\)
Pseudoconvexity local psh defining functions; Levi form \(\ge 0\) on analytic tangent space
Levi problem pseudoconvex \(\Leftrightarrow\) domain of holomorphy (solved affirmatively: Oka, solution in full generality mid-20th c.)
Stein manifold holomorphically convex + holomorphic separation + countable topology; closed complex submanifolds of \(\mathbb{C}^N\)

Plurisubharmonic (psh) functions: u.s.c. \(u\) with \(i\partial\bar\partial u\ge 0\) as a current (or \(u\circ\gamma\) subharmonic for complex lines \(\gamma\)). Strictly psh defining functions characterize strictly pseudoconvex boundaries.

The \(\bar\partial\)-equation

For a \((0,q)\)-form \(u\), solve \(\bar\partial v=u\) when \(\bar\partial u=0\).

Setting Result
Polydisc / \(\mathbb{C}^n\) Dolbeault lemma: local exactness
Pseudoconvex / Stein Hörmander \(L^2\) estimates: solve \(\bar\partial\) with \(L^2\) bounds weighted by psh functions
Compact Kähler Hodge theory replaces pure \(L^2\) existence; harmonic representatives

Hörmander. On pseudoconvex \(\Omega\), for \(\bar\partial\)-closed \(u\in L^2_{0,q}(\Omega,e^{-\varphi})\) with \(\varphi\) strictly psh (weights), there exists \(v\) with \(\bar\partial v=u\) and \(\|v\|^2\le\int |u|^2/(\text{Levi lower bound})\). This is the analytic engine behind Cartan B and vanishing theorems.

Overdetermined system: \(\bar\partial^2=0\) is the compatibility condition; ellipticity in the complex-tangential directions ties to microlocal analysis of \(\bar\partial_b\) on boundaries.

Oka–Cartan and coherence of \(\mathcal{O}\)

A sheaf \(\mathcal{F}\) of \(\mathcal{O}_X\)-modules on a complex space \(X\) is coherent if it is locally finitely presented and the kernel of any map \(\mathcal{O}_U^p\to\mathcal{F}|_U\) is locally f.g. (standard definition: locally f.g. and relations of finitely many sections are locally f.g.).

Theorem (Oka coherence). The structure sheaf \(\mathcal{O}_{\mathbb{C}^n}\) (and \(\mathcal{O}_X\) on complex manifolds / reduced complex spaces under standard constructions) is coherent. Ideals of analytic sets are coherent (Rückert Nullstellensatz / Cartan).

Cartan’s Theorems (Stein \(X\)).

Theorem Statement
A For coherent \(\mathcal{F}\), stalks \(\mathcal{F}_x\) are generated by global sections \(\Gamma(X,\mathcal{F})\)
B \(H^q(X,\mathcal{F})=0\) for all \(q\ge 1\) and all coherent \(\mathcal{F}\)

Consequences. Surjectivity \(\mathcal{O}(X)^p\to\mathcal{O}(X)^q\) criteria; holomorphic convexity characterizations; existence of embeddings of Stein manifolds into \(\mathbb{C}^N\); Cousin problems solvable on Stein spaces (additive Cousin \(\Leftrightarrow\) \(H^1(\mathcal{O})=0\); multiplicative with line bundles).

Oka principle (pointer). On Stein spaces, topological complex vector bundles admit holomorphic structures; many homotopy classes of continuous maps are represented by holomorphic maps (Grauert, Gromov elliptic geometric structures).

See it

Cousin problems. Additive Cousin: prescribe principal parts on a cover; obstruction in \(H^1(X,\mathcal{O})\). On Stein \(X\), Cartan B \(\Rightarrow\) solvable. Multiplicative Cousin / Mittag-Leffler for zeros: obstruction in \(H^1(X,\mathcal{O}^\times)\), related to Picard / line bundles. For \(n=1\), Runge / Weierstrass / Mittag-Leffler are the classical special cases; SCV explains them via coherence + vanishing.

Analytic sets. Zero loci of finitely many holomorphic functions are locally exactly the zero loci of coherent ideal sheaves; dimension theory and Noetherianity of \(\mathcal{O}_n\) (stalks are Noetherian local rings — Rückert) parallel algebraic geometry.

Check

Show that the punctured polydisc \(\Delta^n\setminus\{0\}\) is not a domain of holomorphy for \(n\ge 2\).

By Hartogs, every \(f\in\mathcal{O}(\Delta^n\setminus\{0\})\) extends to \(\Delta^n\), so the holomorphic hull of compact sets surrounding \(0\) can escape every compact of the punctured domain; equivalently all functions extend to a strictly larger domain.

Why does Cartan B imply that every holomorphic vector bundle on a Stein manifold is generated by global holomorphic sections?

Apply Theorem A (or B to the sheaf of the bundle and to kernels of surjections from free sheaves): coherence of the sheaf of sections + vanishing of \(H^1\) of relation sheaves yields global generation and freeness criteria; finite generation from A, and for the dual: surjections \(\mathcal{O}^N\to E\) globally after choosing generators of fibers.

Pitfalls

  • Separate holomorphy without continuity can fail to imply joint holomorphy in pathological cases; local boundedness repairs this.
  • Pseudoconvexity of the boundary is not the same as holomorphic extendability across a boundary point (local peak functions, etc.).
  • Coherence fails for \(\mathcal{C}^\infty\) structure sheaves in the same algebraic sense — analytic coherence is special to \(\mathcal{O}\).
  • Non-Stein manifolds (e.g. compact positive-dimensional) have abundant coherent sheaf cohomology; Cartan B characterizes Stein among complex manifolds (with separation hyp.).

Refs

  • Hörmander, An Introduction to Complex Analysis in Several Variables
  • Gunning–Rossi, Analytic Functions of Several Complex Variables
  • Grauert–Remmert, Theory of Stein Spaces / Coherent Analytic Sheaves
  • Range, Holomorphic Functions and Integral Representations in Several Complex Variables

Next

SCV index · \(\bar\partial\)-Neumann & Kohn Estimates · Stein & Oka Principle · CR Geometry & \(\bar\partial_b\) · Analytic Methods & \(L^2\) · Complex geometry foundations · Complex analysis · Commutative algebra · Homological algebra · AG foundations