Automorphism groups and the Fefferman mapping theorem

Intuition Beginner

Every region in space has a collection of symmetries: the ways you can reshuffle it onto itself without tearing or stretching in the forbidden directions. For a region in several complex dimensions the allowed reshufflings are the holomorphic self-maps that have holomorphic inverses. Collect all of them and you get the region's symmetry group. A round ball has a huge supply of these symmetries; a lumpy region has very few. So the size and shape of the symmetry group is already a fingerprint of the region.

Two regions that a holomorphic dictionary can translate between must have matching symmetry groups, the same way two languages with a perfect dictionary share the same grammar. This gives a quick test: if the symmetry groups have different sizes, no dictionary exists. That single idea, counting symmetries, already separates the ball from the box-shaped product of discs in two complex variables. They look similar but their symmetry groups have different dimensions, so no holomorphic dictionary can match them.

The deeper question is what a dictionary does at the edge of a region. A holomorphic map is defined inside, but does it stay smooth as you walk out to the boundary? Fefferman's theorem says yes, for nicely curved regions: the dictionary extends smoothly all the way to the rim, so the boundaries themselves get matched.

Visual Beginner

Two smoothly bounded blobs in complex space sit side by side, joined by a two-way arrow labelled "holomorphic dictionary". Inside each blob a cloud of curved arrows depicts its symmetry group, the self-maps sliding the blob onto itself; the ball's cloud is dense and uniform, the lumpy blob's cloud is sparse. A dashed boundary shell wraps each blob, and the dictionary arrow is drawn continuing right up to and across the shells, annotated "extends smoothly to the rim (Fefferman)". A small inset contrasts the round ball against the square product of two discs: a tally beneath each reads "8 symmetries" and "6 symmetries", with a red cross on an arrow between them marking "no dictionary — counts differ".

Worked example Beginner

Take the unit disc in one complex variable. Its symmetries are the Möbius maps that send the disc to itself: rotations about the centre, plus maps that slide the centre to any other interior point. Count the dials. A rotation has one dial, the angle. Sliding the centre to a new point uses two dials, the two coordinates of that point. So the disc's symmetry group has three dials in total.

Now stack two discs to make a square-cornered region in two complex variables, the product of two discs. Each disc factor contributes its own three dials, and you may also swap the two factors. That gives six dials.

Compare with the round ball in two complex variables. A direct count of its symmetries gives eight dials, because the ball can slide any interior point to any other in more independent ways than the product can.

What this tells us: six dials versus eight dials. The two regions have symmetry groups of different sizes, so no holomorphic dictionary can match them. The ball and the box are genuinely different complex shapes, a fact first noticed by Poincaré.

Check your understanding Beginner

Formal definition Intermediate+

Let $\Omega \subseteq {\mathbb{C}}^n$ be a domain (open and connected). Its holomorphic automorphism group is

$$\mathrm{Aut}(\Omega )=\{\Phi :\Omega \to \Omega :\Phi \text{ biholomorphic}\},$$

a group under composition. We topologise it by uniform convergence on compact subsets of $\Omega$ (equivalently the compact-open topology); a sequence ${\Phi }_j\to \Phi$ when ${\Phi }_j\to \Phi$ uniformly on every compact $K⋐\Omega$, together with the analogous convergence of inverses.

Definition (proper action). A continuous action of a topological group $G$ on a locally compact space $X$ is proper if the map $G\times X\to X\times X$, $(g,x)\mapsto (gx,x)$, is proper (preimages of compacta are compact). Concretely, for compact $K,L\subseteq X$ the set $\{g\in G:gK\cap L\ne \varnothing \}$ is relatively compact.

Definition (strongly pseudoconvex, smoothly bounded). A domain $\Omega ⋐{\mathbb{C}}^n$ is smoothly bounded and strongly pseudoconvex if there is a defining function $\rho \in C^{\infty }$ near $\overline{\Omega }$ with $\Omega =\{\rho <0\}$, $d\rho \ne 0$ on $\partial \Omega$, and the Levi form ${\sum }_{j,k}{\rho }_{z_j{\bar{z}}_k}(p)w_j{\bar{w}}_k$ positive definite on the complex tangent space $T_p^{1,0}\partial \Omega$ for every $p\in \partial \Omega$ 06.10.09.

Definition (Bergman transformation law). For a bounded domain the Bergman kernel $K_{\Omega }(z,w)$ 06.10.09 transforms under a biholomorphism $\Phi :{\Omega }_1\to {\Omega }_2$ by

$$K_{{\Omega }_1}(z,w)=J_{\Phi }(z)K_{{\Omega }_2}(\Phi z,\Phi w)\overline{J_{\Phi }(w)},$$

where $J_{\Phi }=\det (\partial \Phi /\partial z)$ is the holomorphic Jacobian. On the diagonal $K_{{\Omega }_1}(z,z)=|J_{\Phi }(z){|}^2{K}_{{\Omega }_2}(\Phi z,\Phi z)$, so a biholomorphism is an isometry of the Bergman metrics of 06.10.12.

A map $\Phi :\overline{{\Omega }_1}\to \overline{{\Omega }_2}$ extends to a $C^{\infty }$ diffeomorphism of the closures if there is a diffeomorphism of $\overline{{\Omega }_1}$ onto $\overline{{\Omega }_2}$, smooth up to the boundary in the real sense, restricting to $\Phi$ on ${\Omega }_1$.

Counterexamples to common slips

• $\mathrm{Aut}(\Omega )$ is a real Lie group, not a complex one: the maps are holomorphic but the group parameters (rotation angles, translation amounts) are real. For $\Omega ={\mathbb{C}}^n$ the group is the complex affine group, infinite supply, and the action is not proper, so unboundedness genuinely matters in H. Cartan's theorem.
• The smooth extension of Fefferman's theorem needs strong pseudoconvexity. On a smoothly bounded but only weakly pseudoconvex domain a biholomorphism can still extend continuously, but the proof mechanism (the $(-\rho {)}^{-(n+1)}$ singularity and the subelliptic gain) degrades, and the clean $C^{\infty }$ statement is what fails first.
• A biholomorphism need not extend smoothly when the boundary is merely $C^1$: the theorem is a regularity statement that consumes the smoothness of $\partial \Omega$. Boundary smoothness of $\Phi$ is bootstrapped from boundary smoothness of $\partial \Omega$, not assumed.

Key theorem with proof Intermediate+

Theorem (Fefferman mapping theorem, 1974). Let ${\Omega }_1,{\Omega }_2⋐{\mathbb{C}}^n$ be smoothly bounded strongly pseudoconvex domains and let $\Phi :{\Omega }_1\to {\Omega }_2$ be a biholomorphism. Then $\Phi$ extends to a $C^{\infty }$ diffeomorphism $\overline{\Phi }:\overline{{\Omega }_1}\to \overline{{\Omega }_2}$ of the closures. In particular its boundary restriction $\partial \Phi :\partial {\Omega }_1\to \partial {\Omega }_2$ is a smooth CR-diffeomorphism.

Proof. The argument couples the boundary asymptotics of the Bergman kernel to the invariant-metric geodesic structure, then bootstraps regularity through the subelliptic estimates.

Step 1 — the transformation law sets up the comparison. The diagonal Bergman kernels satisfy $K_{{\Omega }_1}(z,z)=|J_{\Phi }(z){|}^2{K}_{{\Omega }_2}(\Phi z,\Phi z)$. Fefferman's expansion 06.10.09 gives $K_{{\Omega }_i}(z,z)={\phi }_i(z)(-{\rho }_i{)}^{-(n+1)}+{\psi }_i\log (-{\rho }_i)$ with ${\phi }_i,{\psi }_i\in C^{\infty }(\overline{{\Omega }_i})$ and ${\phi }_i>0$ on $\partial {\Omega }_i$. Matching leading singularities across the transformation law forces $-{\rho }_1\asymp |J_{\Phi }{|}^{-2/(n+1)}(-{\rho }_2\circ \Phi )$ near $\partial {\Omega }_1$, so $|J_{\Phi }|$ is bounded above and below near the boundary: $\Phi$ neither collapses nor blows up the boundary distance.

Step 2 — the invariant metric controls the boundary behaviour. By the transformation law $\Phi$ is an isometry of the Bergman metrics, and on a strongly pseudoconvex domain the Bergman (and Kobayashi) metric of 06.10.12 blows up like $(-\rho {)}^{-1}$ in the complex-normal direction and $(-\rho {)}^{-1/2}$ in complex-tangential directions. Geodesics for these complete metrics that head toward $\partial {\Omega }_1$ have a definite normal/tangential anisotropy; the isometry $\Phi$ carries them to boundary-approaching geodesics of ${\Omega }_2$ with the same anisotropy. Hence $\Phi$ sends boundary points to boundary points in a way compatible with the defining functions: ${\rho }_2\circ \Phi$ extends continuously to $\partial {\Omega }_1$ and vanishes there.

Step 3 — first boundary regularity via the kernel. Because $K_{{\Omega }_2}(\cdot ,w)$ is, for fixed interior $w$, smooth up to $\partial {\Omega }_2$ away from $w$ and reproduces holomorphic $L^2$ functions, the identity ${\Phi }^{\ast }{K}_{{\Omega }_2}(\cdot ,w)=\overline{J_{\Phi }}{K}_{{\Omega }_1}(\cdot ,{\Phi }^{-1}w)/J_{\Phi }({\Phi }^{-1}w)$ expresses each component of $\Phi$ through Bergman kernels of the two domains. The boundary smoothness of those kernels propagates to a first finite order of boundary smoothness for $\Phi$.

Step 4 — bootstrap by the $\bar{\partial }$-Neumann estimates. On a strongly pseudoconvex domain the Bergman projection $P=I-{\bar{\partial }}^{\ast }N\bar{\partial }$ satisfies condition R: $P$ maps the Sobolev space $W^s(\Omega )$ to $W^s(\Omega )$ for every $s$, because the $\bar{\partial }$-Neumann operator $N$ gains $1$ derivative (the subelliptic $1/2$-estimate, 06.10.09). Writing the components of $\Phi$ as $P$-images of compactly supported data, condition R upgrades the finite boundary regularity of Step 3 to all Sobolev orders, and Sobolev embedding gives $\overline{\Phi }\in C^{\infty }(\overline{{\Omega }_1})$. Applying the same to ${\Phi }^{-1}$ makes $\overline{\Phi }$ a diffeomorphism of closures; restricting to $\partial {\Omega }_1$ gives a smooth map intertwining the CR structures, a CR-diffeomorphism. $\square$

The kernel-and-metric route is Fefferman's original 1974 argument ^{[Fefferman 1974]}; Steps 3-4 are the Bell-Ligocka condition-R streamlining ^{[Bell-Ligocka 1980]}, which replaces the most delicate asymptotic analysis by the single regularity property of the Bergman projection.

Bridge. This theorem builds toward the boundary classification of strongly pseudoconvex domains, which appears again in 06.10.16, by converting an interior biholomorphism into a smooth boundary CR-equivalence. The foundational reason the extension is smooth is that the Bergman kernel's boundary singularity is a CR-invariant locked to the Levi form, so the transformation law forces $\Phi$ to respect that singularity all the way to the rim — this is exactly the mechanism by which an $L^2$-interior statement becomes a $C^{\infty }$-boundary statement. The central insight is that the invariant metric of 06.10.12 is complete and geodesically rigid near a strongly pseudoconvex boundary, so an isometry cannot fray at the edge; putting these together, the boundary map is determined and smooth, and the bridge is the pair (kernel asymptotics of 06.10.09, metric completeness of 06.10.12) that this unit was created to host. The same coupling generalises: it is dual to the Wong-Rosay rescaling of 06.10.16, where a single domain with runaway symmetry collapses to the ball model.

Exercises Intermediate+

Lean formalization Intermediate+

Mathlib does not formalise the holomorphic automorphism group of a domain in ${\mathbb{C}}^n$, the Bergman kernel and its transformation law, or the subelliptic estimates behind the Fefferman mapping theorem. A proposed signature, in Lean 4 / Mathlib syntax, sketching the target statements:

-- Sketch only; no current Mathlib coverage. See lean_mathlib_gap.
import Mathlib.Analysis.Complex.Analytic
import Mathlib.Topology.Algebra.Group.Basic

namespace Codex.RiemannSurfaces.SeveralComplexVariables

variable {n : ℕ} (Ω : Set (Fin n → ℂ))
  (hΩ : IsOpen Ω) (hb : Bornology.IsBounded Ω)

-- The automorphism group: biholomorphic self-maps, with the
-- compact-open topology making it a topological group.
structure Aut where
  toFun  : (Fin n → ℂ) → (Fin n → ℂ)
  bihol  : Biholomorphic Ω toFun
  maps   : Set.MapsTo toFun Ω Ω

-- H. Cartan: the action on a bounded domain is proper.
theorem aut_action_proper :
    ProperAction (Aut Ω) Ω := by
  sorry

-- Fefferman mapping theorem (statement only).
theorem fefferman_mapping
    {Ω₁ Ω₂ : Set (Fin n → ℂ)}
    (h₁ : StronglyPseudoconvex Ω₁) (h₂ : StronglyPseudoconvex Ω₂)
    (Φ : (Fin n → ℂ) → (Fin n → ℂ))
    (hΦ : Biholomorphic Ω₁ Φ) (hmap : Set.BijOn Φ Ω₁ Ω₂) :
    ∃ Ψ : (Fin n → ℂ) → (Fin n → ℂ),
      ContDiffOn ℝ ⊤ Ψ (closure Ω₁) ∧
      Set.EqOn Ψ Φ Ω₁ ∧ Set.BijOn Ψ (closure Ω₁) (closure Ω₂) := by
  sorry

end Codex.RiemannSurfaces.SeveralComplexVariables

The proof depends on names absent from Mathlib: the biholomorphic-map predicate, Montel's theorem in several variables for the proper action, the Bergman kernel and its transformation law, condition R for the Bergman projection, and the subelliptic $\bar{\partial }$-Neumann estimates. Each is a candidate Mathlib contribution; until then this unit ships with lean_status: none.

Advanced results Master

The mapping theorem sits at the centre of a web of structure theorems organising the automorphism groups and boundary geometry of strongly pseudoconvex domains.

H. Cartan's theorem and the Lie-group structure. For a bounded domain $\Omega ⋐{\mathbb{C}}^n$, $\mathrm{Aut}(\Omega )$ with the topology of uniform convergence on compacta is a real Lie group acting smoothly and properly on $\Omega$ ^{[Cartan 1935]}. The proof runs through Montel's theorem (a uniformly bounded family of holomorphic maps is normal) to make the group locally compact, H. Cartan's uniqueness theorem to embed isotropy subgroups into ${\mathrm{GL}}_n(\mathbb{C})$, and the Bochner-Montgomery theorem (a locally compact group acting effectively and smoothly is a Lie group) to finish. Properness is the geometric heart: orbits are closed and isotropy groups compact, so the quotient $\Omega /\mathrm{Aut}(\Omega )$ is well behaved. The dimension is bounded by $n^2+2n$, attained only by the ball.

The model groups. The ball ${\mathbb{B}}^n$ has $\mathrm{Aut}({\mathbb{B}}^n)=\mathrm{PU}(n,1)$, the projective unitary group of a Hermitian form of signature $(n,1)$, acting transitively with isotropy $U(n)$ at $0$; thus ${\mathbb{B}}^n=\mathrm{PU}(n,1)/U(n)$ is a rank-one Hermitian symmetric space of non-compact type, the complex hyperbolic space. The polydisc has $\mathrm{Aut}({\Delta }^n)=(\mathrm{Aut}\Delta {)}^n⋊S_n$, dimension $3n$, acting only coordinate-wise: it is not transitive for $n\ge 2$. Poincaré's 1907 observation ^{[Poincaré 1907]} that ${\mathbb{B}}^2\ncong {\Delta }^2$ is read off either way — dimension $8\ne 6$, or transitivity versus its failure.

Fefferman's theorem and CR boundary equivalence. The mapping theorem ^{[Fefferman 1974]} reduces biholomorphic equivalence of smoothly bounded strongly pseudoconvex domains to CR-equivalence of their boundaries: $\Phi$ extends to a smooth CR-diffeomorphism $\partial \Phi$, and conversely a CR-diffeomorphism of the boundaries extends holomorphically inward (the CR-extension and Hartogs phenomena). This converts a global function-theory problem into the local-differential-geometry problem of classifying real hypersurfaces in ${\mathbb{C}}^n$ up to CR-equivalence.

The Chern-Moser invariants. Chern and Moser ^{[Chern-Moser 1974]} solved that local problem: a strongly pseudoconvex real hypersurface admits a formal normal form under CR-equivalence, and the obstruction to flattening it to the sphere is a tensor-valued curvature, the Chern-Moser invariant, the CR analogue of the Riemann curvature tensor. Two strongly pseudoconvex boundaries are CR-equivalent precisely when their Chern-Moser invariants match. Composed with Fefferman's theorem, this gives a complete invariant for biholomorphic equivalence of the domains, and locates the hypersurface inside the Cartan-geometry framework of parabolic geometries modelled on $\mathrm{PU}(n,1)/P$.

Bergman, Szegő, and the parabolic invariant. The leading boundary coefficient of the Bergman kernel, ${\phi }_0=n!{\pi }^{-n}J[\rho ]$ on $\partial \Omega$ 06.10.09, and the logarithmic coefficient $\psi$, are CR-invariants; $\psi$ is Fefferman's parabolic invariant, the ancestor of CR $Q$-curvature. The mapping theorem is the statement that these invariants, carried by the kernel, transport correctly under biholomorphism, and that their correct transport forces the boundary map to be smooth. The Szegő-kernel singularity 06.10.09 carries the same information one normal-derivative milder, so either reproducing kernel detects the CR structure.

Synthesis. The automorphism group is the global symmetry fingerprint and the Chern-Moser tensor is the local boundary fingerprint, and the Fefferman mapping theorem is the bridge welding them: it shows that an interior biholomorphism is a boundary CR-equivalence, so global equivalence of strongly pseudoconvex domains is local CR-equivalence of their boundaries. The foundational reason this works is that the Bergman kernel's boundary singularity of 06.10.09 is a CR-invariant locked to the Levi form, and the transformation law forces a biholomorphism to respect it up to the rim; this is exactly the mechanism converting an interior $L^2$-statement into $C^{\infty }$-boundary smoothness, with the subelliptic $\bar{\partial }$-Neumann estimates supplying the bootstrap. Putting these together, the invariant metric of 06.10.12 is the geodesic skeleton that an isometry cannot fray at the boundary, which is dual to the rescaling picture of 06.10.16 where runaway symmetry collapses a domain to the ball. The central insight is that maximal symmetry, $\dim \mathrm{Aut}=n^2+2n$, singles out the ball, and every other strongly pseudoconvex domain is pinned by its Chern-Moser boundary curvature; the bridge is the kernel-and-metric coupling this unit hosts, the foundational reason the dangling forward references of 06.10.09, 06.10.12, and 06.10.16 all point here.

Full proof set Master

Proposition (H. Cartan uniqueness). Let $\Omega ⋐{\mathbb{C}}^n$ be a bounded domain, $p\in \Omega$, and $\Phi :\Omega \to \Omega$ holomorphic with $\Phi (p)=p$ and $d{\Phi }_p=\mathrm{id}$. Then $\Phi =\mathrm{id}$.

Proof. Center coordinates at $p=0$. Expand $\Phi (z)=z+P_m(z)+O(|z{|}^{m+1})$ where $P_m$ is the first non-vanishing homogeneous term of degree $m\ge 2$ (if none, $\Phi =\mathrm{id}$ already). The $k$-fold iterate satisfies ${\Phi }^{\circ k}(z)=z+k{P}_m(z)+O(|z{|}^{m+1})$ by induction, since composing adds the leading terms. The iterates ${\Phi }^{\circ k}$ map the bounded $\Omega$ into itself, so by the Cauchy estimates on a fixed ball $B(0,r)⋐\Omega$ the degree-$m$ Taylor coefficients of ${\Phi }^{\circ k}$ are bounded uniformly in $k$. But those coefficients equal $k$ times the coefficients of $P_m$, which is bounded in $k$ only if $P_m=0$. Therefore $P_m=0$, a contradiction unless there is no such term, giving $\Phi =\mathrm{id}$. $\square$

Proposition (properness of the action on a bounded domain). For $\Omega ⋐{\mathbb{C}}^n$ bounded, the action of $\mathrm{Aut}(\Omega )$ on $\Omega$ is proper; in particular all isotropy groups are compact.

Proof. Let $K,L⋐\Omega$ be compact and let ${\Phi }_j\in \mathrm{Aut}(\Omega )$ with ${\Phi }_j(K)\cap L\ne \varnothing$. The family $\{{\Phi }_j\}$ is uniformly bounded (it maps into the bounded $\Omega$), so by Montel's theorem a subsequence converges locally uniformly to a holomorphic $\Phi :\Omega \to \overline{\Omega }$. The same applies to ${\Phi }_j^{-1}$, giving a limit $\Psi$. On the overlap, ${\Phi }_j(K)\cap L\ne \varnothing$ keeps the limit from pushing $K$ out of $\Omega$, and $\Phi \circ \Psi =\mathrm{id}=\Psi \circ \Phi$ by passing to the limit in ${\Phi }_j\circ {\Phi }_j^{-1}=\mathrm{id}$, using Hurwitz to keep the limits open. Hence $\Phi \in \mathrm{Aut}(\Omega )$ and the subsequence converges in $\mathrm{Aut}(\Omega )$. Thus $\{\Phi :\Phi K\cap L\ne \varnothing \}$ is relatively compact, which is properness. The isotropy of $p$ is the case $K=L=\{p\}$, a closed subgroup of a compact set, hence compact. $\square$

Proposition (the ball isotropy is $U(n)$). The isotropy subgroup $\{\Phi \in \mathrm{Aut}({\mathbb{B}}^n):\Phi (0)=0\}$ equals $U(n)$, acting linearly.

Proof. By Cartan uniqueness the map $\Phi \mapsto d{\Phi }_0$ is injective on the isotropy. A holomorphic self-map of ${\mathbb{B}}^n$ fixing $0$ satisfies the Schwarz lemma $|\Phi (z)|\le |z|$, and an automorphism satisfies it both ways, forcing $|d{\Phi }_0(v)|=|v|$ for all $v$, so $d{\Phi }_0\in U(n)$. Conversely each $U\in U(n)$ is a linear automorphism of ${\mathbb{B}}^n$ fixing $0$. Injectivity plus this surjectivity onto $U(n)$ give the isotropy $\cong U(n)$, and since $d{\Phi }_0=U$ already is an automorphism, Cartan uniqueness applied to $\Phi \circ U^{-1}$ shows $\Phi =U$ is itself linear. $\square$

Theorem (Fefferman mapping theorem). Statement as in the Key theorem section.

Proof. The four-step argument of the Intermediate Key-theorem section is the proof: the transformation law and Fefferman boundary asymptotics 06.10.09 set the comparison and bound $|J_{\Phi }|$; the complete invariant metric 06.10.12 makes $\Phi$ a geodesic-preserving isometry that carries boundary to boundary; the Bergman-kernel representation gives a first finite order of boundary regularity; and condition R for the Bergman projection — valid on strongly pseudoconvex domains by the subelliptic $\bar{\partial }$-Neumann estimate 06.10.09 — bootstraps to $C^{\infty }(\overline{{\Omega }_1})$. Applying the conclusion to ${\Phi }^{-1}$ makes $\overline{\Phi }$ a diffeomorphism of closures, and the boundary restriction is a CR-diffeomorphism. The Bell-Ligocka form ^{[Bell-Ligocka 1980]} packages Steps 3-4 into the single condition-R property. $\square$

Corollary (biholomorphic equivalence is boundary CR-equivalence). Two smoothly bounded strongly pseudoconvex domains ${\Omega }_1,{\Omega }_2⋐{\mathbb{C}}^n$ are biholomorphic if and only if $\partial {\Omega }_1$ and $\partial {\Omega }_2$ are CR-diffeomorphic.

Proof. ($\Rightarrow$) The mapping theorem extends a biholomorphism to a smooth CR-diffeomorphism of boundaries. ($\Leftarrow$) A CR-diffeomorphism $f:\partial {\Omega }_1\to \partial {\Omega }_2$ of strongly pseudoconvex boundaries extends holomorphically to a neighbourhood by the CR-extension theorem (holomorphic functions on the pseudoconvex side extend by Hartogs/Bochner), and the extension restricts to a biholomorphism ${\Omega }_1\to {\Omega }_2$ because it is a local biholomorphism matching boundaries and is proper. Hence the two notions coincide. $\square$

Proposition (dimension bound singles out the ball). Among smoothly bounded strongly pseudoconvex domains $\Omega ⋐{\mathbb{C}}^n$, ${\dim }_{\mathbb{R}}\mathrm{Aut}(\Omega )\le n^2+2n$, with equality iff $\Omega$ is biholomorphic to ${\mathbb{B}}^n$.

Proof. The Chern-Moser theory bounds the dimension of the CR-automorphism group of a strongly pseudoconvex hypersurface by that of the sphere, $\dim \mathrm{PU}(n,1)=n^2+2n$, with equality forcing the hypersurface to be CR-equivalent to the sphere (vanishing Chern-Moser curvature). By Fefferman's mapping theorem $\mathrm{Aut}(\Omega )$ embeds into the CR-automorphism group of $\partial \Omega$, so the same bound holds, and equality forces $\partial \Omega$ CR-spherical, hence $\Omega \cong {\mathbb{B}}^n$ by the Corollary. $\square$

Connections Master

• Szegő kernel and Fefferman boundary asymptotics 06.10.09. This unit hosts the proof of the Fefferman mapping theorem that 06.10.09 forward-references. The boundary expansion $K(z,z)=\phi (-\rho {)}^{-(n+1)}+\psi \log (-\rho )$ and the subelliptic $\bar{\partial }$-Neumann estimate proved there are the analytic inputs: the transformation law transports the CR-invariant singularity, and condition R (from the same subelliptic estimate) bootstraps boundary smoothness of the biholomorphism.

• Invariant metrics: Carathéodory, Kobayashi, Bergman 06.10.12. The biholomorphism is an isometry of the invariant metrics, and the completeness and anisotropic boundary blow-up established there are what force boundary-approaching geodesics to map to boundary-approaching geodesics. The metric rigidity is the geometric half of the mapping-theorem proof, complementing the kernel-asymptotics half.

• Wong-Rosay theorem and boundary rigidity 06.10.16. Wong-Rosay is the companion rigidity: where Fefferman extends a map between two domains, Wong-Rosay collapses a single domain with non-compact automorphism group to the ball. Both run on the same coupling of Bergman/Kobayashi metric estimates and strong-pseudoconvexity scaling; 06.10.16 cites this unit for the mapping theorem it uses to identify the scaling limit.

• Bergman kernel and Bergman metric 06.10.08. The transformation law $K_{{\Omega }_1}=J_{\Phi }(K_{{\Omega }_2}\circ \Phi )\overline{J_{\Phi }}$ and the Bergman metric as $\partial \bar{\partial }\log K$ are the constructions this unit transports across a biholomorphism; the automorphism group acts by Bergman isometries, the structural fact behind Cartan's Lie-group theorem.

• Pseudoconvexity and the Levi form 06.10.03. Strong pseudoconvexity — positive-definite Levi form on the complex tangent space — is the hypothesis that makes the kernel singularity exactly $(-\rho {)}^{-(n+1)}$, the $\bar{\partial }$-Neumann operator subelliptic, and the Chern-Moser normal form available; dropping it degrades every step of the proof.

Historical & philosophical context Master

Henri Poincaré opened the subject in 1907 with Les fonctions analytiques de deux variables et la représentation conforme ^{[Poincaré 1907]} (Rend. Circ. Mat. Palermo 23, 185-220), where he observed that the unit ball and the bidisc in ${\mathbb{C}}^2$ are not biholomorphically equivalent — the first proof that the Riemann mapping theorem has no several-variable analogue, since two of the most natural simply connected domains already fail to be equivalent. Poincaré's tool was effectively the automorphism group: the ball's symmetry group is transitive and larger than the bidisc's, so no dictionary can match them. This made the automorphism group the first invariant of complex-analytic shape in higher dimension and set the programme of classifying domains by their symmetries.

Henri Cartan, in his 1935 Sur les groupes de transformations analytiques ^{[Cartan 1935]} (Hermann), proved that the automorphism group of a bounded domain is a real Lie group acting properly, supplying the structural foundation — the uniqueness theorem, the normal-family compactness, and the Lie-group conclusion. The decisive analytic advance came in Charles Fefferman's 1974 The Bergman kernel and biholomorphic mappings of pseudoconvex domains ^{[Fefferman 1974]} (Invent. Math. 26, 1-65): his boundary asymptotic expansion of the Bergman kernel let him prove that biholomorphisms between smoothly bounded strongly pseudoconvex domains extend smoothly to the closures, finally connecting the interior function theory to the boundary geometry. Shiing-Shen Chern and Jürgen Moser's 1974 Real hypersurfaces in complex manifolds ^{[Chern-Moser 1974]} (Acta Math. 133, 219-271) had simultaneously built the local CR invariants that classify the boundaries, so Fefferman's theorem reduced the global equivalence problem to their tensors. Steven Bell and Ewa Ligocka's 1980 simplification ^{[Bell-Ligocka 1980]} (Invent. Math. 57, 283-289) recast the whole proof around the single regularity property (condition R) of the Bergman projection, the form in which the theorem is usually taught.

Bibliography Master

@incollection{Cartan1935Groupes,
  author    = {Cartan, Henri},
  title     = {Sur les groupes de transformations analytiques},
  series    = {Actualit{\'e}s Scientifiques et Industrielles},
  number    = {198},
  publisher = {Hermann},
  year      = {1935}
}

@article{Poincare1907Fonctions,
  author  = {Poincar{\'e}, Henri},
  title   = {Les fonctions analytiques de deux variables et la repr{\'e}sentation conforme},
  journal = {Rend. Circ. Mat. Palermo},
  volume  = {23},
  year    = {1907},
  pages   = {185--220}
}

@article{Fefferman1974Bergman,
  author  = {Fefferman, Charles},
  title   = {The {B}ergman kernel and biholomorphic mappings of pseudoconvex domains},
  journal = {Invent. Math.},
  volume  = {26},
  year    = {1974},
  pages   = {1--65}
}

@article{ChernMoser1974,
  author  = {Chern, Shiing-Shen and Moser, J{\"u}rgen K.},
  title   = {Real hypersurfaces in complex manifolds},
  journal = {Acta Math.},
  volume  = {133},
  year    = {1974},
  pages   = {219--271}
}

@article{BellLigocka1980,
  author  = {Bell, Steven R. and Ligocka, Ewa},
  title   = {A simplification and extension of {F}efferman's theorem on biholomorphic mappings},
  journal = {Invent. Math.},
  volume  = {57},
  year    = {1980},
  pages   = {283--289}
}

@article{Bell1981ConditionR,
  author  = {Bell, Steven R.},
  title   = {Biholomorphic mappings and the $\bar\partial$-problem},
  journal = {Ann. of Math.},
  volume  = {114},
  year    = {1981},
  pages   = {103--113}
}

@book{KobayashiTransformation,
  author    = {Kobayashi, Shoshichi},
  title     = {Transformation Groups in Differential Geometry},
  series    = {Classics in Mathematics},
  publisher = {Springer},
  year      = {1972}
}

@book{KrantzFTSCV,
  author    = {Krantz, Steven G.},
  title     = {Function Theory of Several Complex Variables},
  edition   = {2},
  series    = {AMS Chelsea Publishing},
  volume    = {340},
  publisher = {American Mathematical Society},
  year      = {2001}
}