06.10.16 · riemann-surfaces / several-variables

Wong-Rosay theorem and boundary rigidity

shipped3 tiersLean: none

Anchor (Master): Wong 1977 (Invent. Math. 41); Rosay 1979 (Ann. Inst. Fourier 29); Pinchuk 1991; Greene-Kim-Krantz *The Geometry of Complex Domains*

Intuition Beginner

Imagine a bounded region $Ω$ in several complex variables and the collection of all biholomorphic self-maps of $Ω$ — its symmetries. For most regions this symmetry collection is small and tame: a few rotations, maybe a reflection. But for the round ball, the symmetry collection is enormous. You can slide any interior point toward the boundary by a symmetry, the way a Möbius map of the disc can push the centre out toward the edge. A symmetry collection that can push a fixed interior point arbitrarily close to the boundary is called non-compact: the symmetries run off to infinity.

The Wong-Rosay theorem says this richness is a fingerprint. If a region's symmetries can shove one interior point all the way out to a boundary point that is nicely curved (strongly pseudoconvex), then the region had to have been the ball all along. Curvature plus runaway symmetry forces the most symmetric possible shape.

So a single accumulation point of one orbit, sitting at a well-curved part of the boundary, decides the entire region up to a change of holomorphic coordinates. That is what we mean by boundary rigidity.

Visual Beginner

A bounded domain drawn with a sequence of interior points marching toward one boundary point $p$ , each image obtained from the previous by a symmetry of the domain. Near $p$ the boundary is shown bulging outward (strongly pseudoconvex). A second panel shows the same marching sequence after zooming in repeatedly near $p$ , with the rescaled pictures converging to the round ball. An arrow labelled "scaling" connects the two panels.

Worked example Beginner

Take the unit disc in one complex variable, $Ω = {∣ z ∣ < 1}$ , the simplest model of "the ball." Fix the centre point $0$ . The maps

φ_{a} (z) = \frac{z + a}{1 + a z}, - 1 < a < 1,

are symmetries of the disc for every real $a$ between $- 1$ and $1$ . Each one sends $0$ to the point $a$ .

As $a$ climbs from $0$ toward $1$ , the image point $φ_{a} (0) = a$ marches straight toward the boundary point $1$ . So the orbit of the centre under these symmetries accumulates at the boundary. The symmetry collection is non-compact, and the boundary point $1$ is a perfectly curved point of the circle.

What this tells us: the disc already exhibits the Wong-Rosay pattern — a runaway orbit hitting a well-curved boundary point. The theorem is the converse in higher dimensions: whenever you see this pattern, the region must be a ball. The disc is not an accident; it is the one-variable instance of a rigidity that holds in every dimension.

Check your understanding Beginner

Formal definition Intermediate+

Let $Ω \subset C^{n}$ be a domain (open and connected). Its automorphism group is

Aut (Ω) = {Φ : Ω \to Ω ∣ Φ biholomorphic},

a topological group under composition, topologized by uniform convergence on compact subsets. By Cartan's theorem $Aut (Ω)$ is a real Lie group acting properly and real-analytically on $Ω$ whenever $Ω$ is bounded.

The group $Aut (Ω)$ is non-compact when it is not compact in the compact-open topology; equivalently, by properness of the action, there exist a point $x_{0} \in Ω$ and a sequence $Φ_{j} \in Aut (Ω)$ with $Φ_{j} (x_{0}) \to p \in \partial Ω$ . Such a $p$ is an orbit accumulation point (a boundary point of the orbit $Aut (Ω) \cdot x_{0}$ ).

A boundary point $p \in \partial Ω$ of class $C^{2}$ is strongly pseudoconvex when, for a local defining function $ρ$ with $Ω = {ρ < 0}$ and $d ρ (p) \neq = 0$ , the Levi form

L_{p} (ρ; v) = j, k = 1 \sum n \frac{\partial ^{2} ρ}{\partial z _{j} \partial z ˉ _{k}} (p) v_{j} \overset{v}{ˉ}_{k}

is positive definite on the complex tangent space $T_{p}^{C} (\partial Ω) = {v \in C^{n} : \sum_{j} \partial_{z_{j}} ρ (p) v_{j} = 0}$ . This is the curvature notion of 06.10.03.

The unit ball is $B^{n} = {z \in C^{n} : ∣ z ∣^{2} = \sum_{j} ∣ z_{j} ∣^{2} < 1}$ ; every boundary point of $B^{n}$ is strongly pseudoconvex, and $Aut (B^{n}) = PU (n, 1)$ acts transitively on $B^{n}$ with non-compact isotropy directions, so its automorphism group is non-compact.

Counterexamples to common slips

The polydisc $Δ^{n}$ ( $n \geq 2$ ) also has non-compact automorphism group, but its orbit accumulation points lie on the Levi-flat parts of the distinguished boundary, never at a strongly pseudoconvex point. So the polydisc is not a counterexample to Wong-Rosay; it shows the strong-pseudoconvexity hypothesis cannot be dropped.
Non-compactness of $Aut (Ω)$ alone does not force the ball. The annulus-like and Reinhardt examples accumulate at weakly pseudoconvex or non-smooth points.
The conclusion is biholomorphic equivalence, not isometric or affine equivalence: $Ω ≅ B^{n}$ as complex manifolds, generally not by an affine map.

Key theorem with proof Intermediate+

Theorem (Wong 1977; Rosay 1979). Let $Ω \subset C^{n}$ be a domain whose automorphism group $Aut (Ω)$ is non-compact. Suppose there is a point $x_{0} \in Ω$ and automorphisms $Φ_{j} \in Aut (Ω)$ with $Φ_{j} (x_{0}) \to p \in \partial Ω$ , where $\partial Ω$ is $C^{2}$ -smooth and strongly pseudoconvex in a neighbourhood of $p$ . Then $Ω$ is biholomorphic to the unit ball $B^{n}$ .

Proof (scaling method, after Pinchuk). The argument rescales the domain near $p$ by stretching against the boundary curvature, takes a normal-family limit, and identifies the limit with an unbounded realization of the ball.

Step 1 — normalization near $p$ . Since $\partial Ω$ is $C^{2}$ and strongly pseudoconvex at $p$ , after an affine complex-linear change of coordinates and translation taking $p$ to $0$ , the defining function has the form

ρ (z) = 2 Re z_{n} + ∣ z^{'} ∣^{2} + o (∣ z ∣^{2}), z = (z^{'}, z_{n}), z^{'} = (z_{1}, \dots, z_{n - 1}),

so that the complex tangent directions are $z^{'}$ and the Levi form is the standard Hermitian form $∣ z^{'} ∣^{2}$ . The model domain frozen at $p$ is the Siegel half-space

H_{n} = {z \in C^{n} : 2 Re z_{n} + ∣ z^{'} ∣^{2} < 0},

which is biholomorphic to $B^{n}$ via the Cayley transform $w = (w^{'}, w_{n}) \mapsto (\frac{2 w ^{'}}{1 + w _{n}}, \frac{w _{n} - 1}{w _{n} + 1})$ (computed in the Full proof set below), mapping $B^{n}$ onto $H_{n}$ .

Step 2 — the scaling sequence. Write $q_{j} = Φ_{j} (x_{0}) \to 0$ . Let $π_{j}$ be the boundary projection of $q_{j}$ to $\partial Ω$ and let $δ_{j} = dist (q_{j}, \partial Ω) \to 0$ . Choose affine maps $A_{j}$ (a unitary rotation aligning the real normal at $π_{j}$ with the $z_{n}$ -axis, a translation to $π_{j}$ , followed by the anisotropic dilation $z^{'} \mapsto δ_{j}^{- 1/2} z^{'}$ , $z_{n} \mapsto δ_{j}^{- 1} z_{n}$ ) so that $A_{j} (q_{j})$ is a fixed base point and $A_{j} (Ω)$ has defining function $A_{j}$ -pushed-forward $ρ$ . The anisotropic dilation is dictated by the strong pseudoconvexity: the normal direction $z_{n}$ scales like distance, the complex tangential directions $z^{'}$ like the square root, exactly matching the quadratic term in $ρ$ .

Step 3 — convergence of the rescaled domains. The $C^{2}$ strong-pseudoconvexity bound makes the $o (∣ z ∣^{2})$ remainder in $ρ$ vanish under the anisotropic dilation. Hence the rescaled domains $Ω_{j} = A_{j} (Ω)$ converge, in the local Hausdorff sense on compact sets, to the Siegel half-space $H_{n}$ . This is the analytic core: the defining functions $δ_{j}^{- 1} (ρ \circ A_{j}^{- 1})$ converge in $C_{loc}^{2}$ to $2 Re z_{n} + ∣ z^{'} ∣^{2}$ .

Step 4 — the rescaled automorphisms form a normal family. Set $Ψ_{j} = A_{j} \circ Φ_{j} : Ω \to Ω_{j}$ . Each $Ψ_{j}$ is biholomorphic onto its image and $Ψ_{j} (x_{0}) = A_{j} (q_{j})$ is bounded away from $0$ and from $\partial H_{n}$ . The Kobayashi-metric estimates of 06.10.12 give uniform interior bounds on $Ψ_{j}$ on compact subsets of $Ω$ (the Kobayashi metric of $Ω$ is locally bounded below, and the distance-decreasing property controls the images). By Montel's theorem in several variables a subsequence converges uniformly on compacta to a holomorphic map $Ψ : Ω \to \overline{H_{n}}$ .

Step 5 — the limit is a biholomorphism. The same Kobayashi estimates applied to the inverses $Ψ_{j}^{- 1}$ show $Ψ$ is non-degenerate and that $Ψ (Ω) \subseteq H_{n}$ with a holomorphic inverse defined on $H_{n}$ ; properness of the rescaling forces $Ψ$ to be onto. Thus $Ψ : Ω \to H_{n}$ is a biholomorphism. Composing with the Cayley transform of Step 1 gives $Ω ≅ B^{n}$ . $□$

Bridge. This rigidity builds toward the automorphism-group classification that appears again in 06.10.13, where the Fefferman mapping theorem shows that any biholomorphism between smoothly bounded strongly pseudoconvex domains extends smoothly to the boundary; Wong-Rosay is the foundational reason a single such domain with extra symmetry collapses to the model. The central insight is that the scaling limit identifies the local boundary geometry with the global complex structure: the anisotropic dilation matched to the Levi form is exactly the mechanism that converts a curvature statement at one point into a biholomorphism of the whole domain, and the bridge is the Kobayashi metric of 06.10.12, which both bounds the rescaled maps and survives the limit. The same scaling generalises: it identifies the asymptotic Bergman geometry of 06.10.08 near a strongly pseudoconvex point with the constant-curvature Bergman metric of the ball, and putting these together, strong pseudoconvexity at an orbit limit is dual to maximal symmetry of the whole domain.

Exercises Intermediate+

Exercise 3 (medium, short-answer).

Explain why the polydisc $Δ^{2} \subset C^{2}$ , although it has non-compact automorphism group, does not contradict the Wong-Rosay theorem.

Hint

Where on $\partial Δ^{2}$ can an orbit accumulate, and is that point strongly pseudoconvex?

Answer

$Aut (Δ^{2})$ contains the product Möbius maps $φ_{a} (z_{1}) \times φ_{b} (z_{2})$ , so the orbit of $(0, 0)$ accumulates on the distinguished boundary ${∣ z_{1} ∣ = ∣ z_{2} ∣ = 1}$ and on the two flat faces ${∣ z_{j} ∣ = 1}$ . Every boundary point of $Δ^{2}$ is Levi-degenerate: along the flat faces the boundary is foliated by complex discs, so the Levi form is not positive definite. There is no strongly pseudoconvex accumulation point, and the hypothesis of Wong-Rosay is not met. Rubric: full credit for identifying Levi-degeneracy of $\partial Δ^{2}$ as the reason the hypothesis fails.

Exercise 5 (medium, short-answer).

Why is Montel's theorem (normal families) the right tool to extract the limit map $Ψ$ in Step 4, and what replaces it if one only worked with bounded domains in one variable?

Hint

Normal families need a uniform bound on compacta; the Kobayashi metric supplies it.

Answer

The rescaled maps $Ψ_{j}$ are holomorphic into a fixed target neighbourhood of $H_{n}$ and are uniformly bounded on each compact subset of $Ω$ because the Kobayashi distance is distance-decreasing and locally bounded below on $Ω$ . Montel's theorem in $C^{n}$ then gives a locally uniformly convergent subsequence. In one variable the same role is played by the classical Montel theorem for bounded families; the several-variable statement is identical once a local uniform bound is in hand, which is precisely what the invariant metric provides. Rubric: full credit for citing the uniform bound from the Kobayashi metric and the resulting normal family.

Exercise 6 (hard, short-answer).

Use Cartan's uniqueness theorem (a holomorphic self-map of a bounded domain fixing a point with identity differential there is the identity) to argue that the isotropy group of a point in $B^{n}$ is compact, and explain how this is consistent with $Aut (B^{n})$ being non-compact.

Hint

The isotropy is determined by its action on the tangent space; compare with the full transitive action.

Answer

By Cartan's uniqueness theorem an automorphism of $B^{n}$ fixing $0$ is determined by its differential at $0$ , and that differential must preserve the ball, so it lies in the unitary group $U (n)$ ; the isotropy at $0$ is therefore isomorphic to $U (n)$ , which is compact. Non-compactness of $Aut (B^{n})$ comes from the transitive directions: the boundary-pushing maps (the analogues of $φ_{a}$ ) form a non-compact family moving $0$ toward $\partial B^{n}$ . The group is a non-compact Lie group with compact isotropy, which is exactly the structure $PU (n, 1) / U (n)$ of a rank-one symmetric space. Rubric: full credit for identifying the isotropy as $U (n)$ via Cartan uniqueness and locating non-compactness in the transitive part.

Exercise 7 (hard, short-answer).

Sketch why strong pseudoconvexity, rather than mere pseudoconvexity, is essential in Step 3 of the proof for the rescaled domains to converge to $H_{n}$ .

Hint

What does the Levi form being positive definite (not just semidefinite) give the anisotropic dilation?

Answer

The anisotropic dilation with $z^{'} \mapsto δ^{- 1/2} z^{'}$ is calibrated to a Levi form that is positive definite: each tangential direction carries a strictly positive quadratic coefficient, so a single uniform $δ^{- 1/2}$ scaling renders all of $∣ z^{'} ∣^{2}$ of order one in the limit. If the Levi form were only semidefinite (weak pseudoconvexity), some tangential directions would carry vanishing quadratic terms and higher-order contact with the boundary; no single power-law dilation matches all directions, the limit model is not $H_{n}$ , and indeed it need not be the ball — weakly pseudoconvex orbit limits give other model domains. Strong pseudoconvexity is what makes the parabolic scaling uniform and the limit the standard Siegel half-space. Rubric: full credit for connecting positive-definiteness to a uniform power-law dilation and the $H_{n}$ limit.

Advanced results Master

The scaling method that proves Wong-Rosay is a general technique for boundary-orbit problems; the theorem is its first and cleanest output. Three refinements and one structural generalization organize the rigidity around it.

Rosay's localization. Wong's 1977 proof assumed $Ω$ bounded with globally $C^{2}$ strongly pseudoconvex boundary. Rosay's 1979 contribution is the localization: only a one-sided $C^{2}$ neighbourhood of a single strongly pseudoconvex orbit-accumulation point $p$ is needed, with no global hypothesis on $Ω$ or even boundedness. The scaling is purely local at $p$ , and the normal-family limit only sees the boundary near $p$ ; the Kobayashi-metric bounds are interior. This is why the theorem applies to unbounded and irregular domains provided one good boundary point carries the orbit.

Klembeck's Bergman-curvature theorem. Klembeck 1978 showed that on a smoothly bounded strongly pseudoconvex domain, the holomorphic sectional curvature of the Bergman metric of 06.10.08 tends to the constant negative value $- 4/ (n + 1)$ of the ball's Bergman metric as the point approaches the boundary. This is the differential-geometric shadow of Wong-Rosay: near a strongly pseudoconvex point the Bergman geometry is asymptotically that of $B^{n}$ . Combined with the scaling limit, it explains why the rescaled automorphisms converge to ball automorphisms: they are converging Bergman isometries of a geometry that is becoming the ball's.

The Greene-Krantz semicontinuity and the orbit dichotomy. For a smoothly bounded domain, the set of orbit-accumulation points of $Aut (Ω)$ is either empty (the group is compact, $Ω \neq = B^{n}$ ) or, when a strongly pseudoconvex point appears, forces $Ω ≅ B^{n}$ . Greene-Krantz semicontinuity of automorphism groups under $C^{\infty}$ perturbation of the boundary refines this: a domain close to a non-ball strongly pseudoconvex domain has automorphism group no larger than the limit, so the ball is rigidly isolated in moduli. The Bun Wong-Rosay rigidity and the semicontinuity together give the automorphism-group characterization of the ball among smoothly bounded strongly pseudoconvex domains: $B^{n}$ is the unique one with non-compact automorphism group.

Beyond strong pseudoconvexity: finite type and model domains. When the orbit accumulates at a weakly pseudoconvex point of finite type $m$ (in the sense of D'Angelo), the scaling method still runs, but the parabolic dilation must be replaced by a non-isotropic dilation matched to the multi-type, and the limit is no longer the ball but a Catlin-D'Angelo model domain of the form ${2 Re z_{n} + P (z^{'}) < 0}$ for a weighted-homogeneous plurisubharmonic polynomial $P$ . The Bedford-Pinchuk theorem (1991, 1994) classifies the finite-type domains in $C^{2}$ with non-compact automorphism group: they are biholomorphic to the model ${2 Re z_{2} + ∣ z_{1} ∣^{2 m} < 0}$ . Wong-Rosay is the $m = 1$ case of this hierarchy.

Synthesis. The Wong-Rosay theorem is the statement that for strongly pseudoconvex boundaries, curvature is dual to symmetry: putting these together, a positive-definite Levi form at one orbit-accumulation point combined with a non-compact automorphism group forces the whole domain to be the most symmetric model, the ball. The scaling method is the foundational reason this works — it identifies the asymptotic boundary geometry with the global complex structure by a power-law dilation matched to the Levi form, and this is exactly the mechanism that recurs throughout the rigidity theory. The Bergman-curvature limit of Klembeck and the orbit-accumulation dichotomy of Greene-Krantz are the differential-geometric and the moduli-theoretic faces of the same fact: the ball is the unique smoothly bounded strongly pseudoconvex domain whose symmetries run off to a strongly pseudoconvex boundary point. The central insight is that the finite-type generalization, where the parabolic scaling becomes a weighted dilation and the ball is replaced by a Catlin-D'Angelo model, identifies the whole rigidity program with the classification of homogeneous model domains, and Wong-Rosay is the order-one stratum of that classification.

Full proof set Master

Proposition (Cayley equivalence of ball and Siegel half-space). The unit ball $B^{n}$ is biholomorphic to the Siegel half-space $H_{n} = {z \in C^{n} : 2 Re z_{n} + ∣ z^{'} ∣^{2} < 0}$ .

Proof. Define the Cayley map $C : B^{n} \to C^{n}$ by

C (w^{'}, w_{n}) = (\frac{2 w ^{'}}{1 + w _{n}}, \frac{w _{n} - 1}{w _{n} + 1}),

holomorphic on ${w_{n} \neq = - 1} \supset B^{n}$ . Write $z = C (w)$ , so $z^{'} = 2 w^{'} / (1 + w_{n})$ and $z_{n} = (w_{n} - 1) / (w_{n} + 1)$ , with inverse $w_{n} = (1 + z_{n}) / (1 - z_{n})$ and $w^{'} = 2 z^{'} / (1 - z_{n})$ . Substitute these into $1 - ∣ w^{'} ∣^{2} - ∣ w_{n} ∣^{2}$ and clear the common denominator $∣1 - z_{n} ∣^{2}$ :

1 - ∣ w^{'} ∣^{2} - ∣ w_{n} ∣^{2} = \frac{∣1 - z _{n} ∣ ^{2} - 2∣ z ^{'} ∣ ^{2} - ∣1 + z _{n} ∣ ^{2}}{∣1 - z _{n} ∣ ^{2}} .

Using $∣1 - z_{n} ∣^{2} - ∣1 + z_{n} ∣^{2} = - 4 Re z_{n}$ , the numerator equals $- 4 Re z_{n} - 2∣ z^{'} ∣^{2} = - 2 (2 Re z_{n} + ∣ z^{'} ∣^{2})$ , so

1 - ∣ w^{'} ∣^{2} - ∣ w_{n} ∣^{2} = \frac{- 2 ( 2 Re z _{n} + ∣ z ^{'} ∣ ^{2} )}{∣1 - z _{n} ∣ ^{2}} .

Hence $∣ w ∣ < 1$ if and only if $2 Re z_{n} + ∣ z^{'} ∣^{2} < 0$ , so $C$ carries $B^{n}$ bijectively onto $H_{n}$ . The displayed inverse $C^{- 1} (z) = (2 z^{'} / (1 - z_{n}), (1 + z_{n}) / (1 - z_{n}))$ is holomorphic on ${z_{n} \neq = 1} \supset H_{n}$ and satisfies $C^{- 1} \circ C = id$ , so $C$ is a biholomorphism $B^{n} ≅ H_{n}$ . $□$

Proposition (Cartan uniqueness, used in the limit identification). Let $D \subset C^{n}$ be a bounded domain, $a \in D$ , and $F : D \to D$ holomorphic with $F (a) = a$ and $d F_{a} = id$ . Then $F = id_{D}$ .

Proof. Without loss of generality $a = 0$ . Expand $F (z) = z + P_{k} (z) + (higher order)$ , where $P_{k}$ is the first non-vanishing homogeneous term of degree $k \geq 2$ in the Taylor series, should one exist. The iterates satisfy $F^{(m)} (z) = z + m P_{k} (z) + (higher order)$ by induction: the degree- $k$ term accumulates linearly because $d F_{0} = id$ contributes no lower-order mixing. Boundedness of $D$ gives, by the Cauchy estimates on a fixed polydisc inside $D$ , a uniform bound $∥ P_{k}^{(m)} ∥ = ∥ m P_{k} ∥ \leq M$ independent of $m$ . Letting $m \to \infty$ forces $P_{k} \equiv 0$ , contradicting the choice of $k$ . Hence no non-zero homogeneous term of degree $\geq 2$ appears and $F (z) = z$ . $□$

Theorem (Wong-Rosay, full statement). Statement as in the Key theorem section: a domain $Ω \subset C^{n}$ with non-compact automorphism group and a $C^{2}$ strongly pseudoconvex orbit-accumulation point is biholomorphic to $B^{n}$ .

Proof. The Key-theorem proof supplies the scaling limit $Ψ : Ω \to H_{n}$ . The two propositions above complete it: the Cayley equivalence converts $H_{n}$ to $B^{n}$ , giving $Ω ≅ B^{n}$ , and Cartan uniqueness pins down the limit map by showing that any two scaling limits differ by an automorphism of $H_{n}$ fixing the base point with identity differential, hence are equal. The local hypotheses (Rosay) suffice because every step — the normalization, the dilation, the $C_{loc}^{2}$ convergence of defining functions, and the Kobayashi bound on the rescaled maps — is carried out in a fixed neighbourhood of $p$ and on compact subsets of $Ω$ . $□$

Corollary (ball characterization). Among smoothly bounded strongly pseudoconvex domains in $C^{n}$ , the ball $B^{n}$ is the unique one with non-compact automorphism group.

Proof. A smoothly bounded domain has compact boundary; if $Aut (Ω)$ is non-compact, properness of the action gives an orbit-accumulation point $p \in \partial Ω$ , which is strongly pseudoconvex by hypothesis. Wong-Rosay gives $Ω ≅ B^{n}$ . Conversely $B^{n}$ has non-compact $Aut (B^{n}) = PU (n, 1)$ . $□$

Connections Master

Pseudoconvexity and the Levi form 06.10.03. The strong-pseudoconvexity hypothesis is the positive-definiteness of the Levi form at the orbit-accumulation point. The entire scaling construction is the parabolic dilation matched to this Levi form; without the positive-definite Levi form there is no canonical model domain and the theorem fails.
Bergman kernel and Bergman metric 06.10.08. Klembeck's theorem realizes Wong-Rosay differential-geometrically: the Bergman metric approaches constant negative holomorphic curvature near a strongly pseudoconvex point, so the rescaled automorphisms are converging Bergman isometries of a geometry becoming that of the ball. The Bergman metric is the canonical invariant metric that detects the rigidity.
Invariant metrics: Carathéodory, Kobayashi, Bergman 06.10.12. The Kobayashi metric supplies the uniform interior bounds that make the rescaled automorphisms a normal family in Step 4 of the proof. Distance-decreasing under holomorphic maps is the property that controls the limit map and forces it to be a biholomorphism.
Szegő kernel and Fefferman boundary asymptotics 06.10.09. The boundary asymptotics of reproducing kernels at a strongly pseudoconvex point are governed by the same Levi-form model as the scaling limit; the Fefferman expansion and the Wong-Rosay limit both encode the local CR geometry of $\partial Ω$ near $p$ .
Automorphism groups and the Fefferman mapping theorem 06.10.13. Wong-Rosay is the rigidity statement for a single domain; the Fefferman mapping theorem is the rigidity statement for maps between two such domains. Both rest on the strongly pseudoconvex model geometry and the smooth boundary behaviour of biholomorphisms.

Historical & philosophical context Master

B. Wong proved the original theorem in 1977 in Characterization of the unit ball in $C^{n}$ by its automorphism group ^{[Wong 1977]} (Invent. Math. 41, 253-257), under the hypotheses that $Ω$ is bounded with globally $C^{2}$ strongly pseudoconvex boundary. Wong's argument used the boundary asymptotics of the Bergman metric — Klembeck's then-recent result that the Bergman holomorphic sectional curvature tends to that of the ball at a strongly pseudoconvex point — to show that the orbit-accumulating automorphisms converge to ball automorphisms.

Jean-Pierre Rosay removed the global hypotheses two years later, in 1979, in Sur une caractérisation de la boule parmi les domaines de $C^{n}$ par son groupe d'automorphismes ^{[Rosay 1979]} (Ann. Inst. Fourier 29, 91-97). Rosay localized the statement to a one-sided neighbourhood of a single strongly pseudoconvex orbit-accumulation point, dropping boundedness and global regularity. The combined statement is now standard as the Wong-Rosay theorem.

The proof presented here is the scaling method of S. Pinchuk, developed through the 1980s and surveyed in The scaling method and holomorphic mappings ^{[Pinchuk 1991]} (Proc. Symp. Pure Math. 52, Part 1, 151-161). Pinchuk's dilation argument replaced the Bergman-curvature input by a direct normal-family limit of rescaled automorphisms, and it generalizes far beyond the ball: matched non-isotropic dilations at finite-type points produce the Catlin-D'Angelo model domains, and the Bedford-Pinchuk classification of finite-type domains in $C^{2}$ with non-compact automorphism group is its descendant. Klembeck's curvature computation appeared in 1978 ^{[Klembeck 1978]} (Indiana Univ. Math. J. 27, 275-282), and the consolidated modern account, including semicontinuity of automorphism groups and the orbit-accumulation dichotomy, is in Greene-Kim-Krantz, The Geometry of Complex Domains (Birkhäuser, 2011) ^{[Krantz GeomComplexDomains]}.

Bibliography Master

@article{Wong1977,
  author  = {Wong, B.},
  title   = {Characterization of the unit ball in {$\mathbb{C}^n$} by its automorphism group},
  journal = {Invent. Math.},
  volume  = {41},
  number  = {3},
  year    = {1977},
  pages   = {253--257}
}

@article{Rosay1979,
  author  = {Rosay, Jean-Pierre},
  title   = {Sur une caract{\'e}risation de la boule parmi les domaines de {$\mathbb{C}^n$} par son groupe d'automorphismes},
  journal = {Ann. Inst. Fourier (Grenoble)},
  volume  = {29},
  number  = {4},
  year    = {1979},
  pages   = {91--97}
}

@incollection{Pinchuk1991,
  author    = {Pinchuk, Sergey},
  title     = {The scaling method and holomorphic mappings},
  booktitle = {Several Complex Variables and Complex Geometry, Part 1},
  series    = {Proc. Sympos. Pure Math.},
  volume    = {52},
  publisher = {Amer. Math. Soc.},
  year      = {1991},
  pages     = {151--161}
}

@article{Klembeck1978,
  author  = {Klembeck, Paul F.},
  title   = {K{\"a}hler metrics of negative curvature, the Bergman metric near the boundary, and the Kobayashi metric on smooth bounded strictly pseudoconvex sets},
  journal = {Indiana Univ. Math. J.},
  volume  = {27},
  number  = {2},
  year    = {1978},
  pages   = {275--282}
}

@book{GreeneKimKrantz2011,
  author    = {Greene, Robert E. and Kim, Kang-Tae and Krantz, Steven G.},
  title     = {The Geometry of Complex Domains},
  series    = {Progress in Mathematics},
  volume    = {291},
  publisher = {Birkh{\"a}user},
  year      = {2011}
}

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

@article{BedfordPinchuk1994,
  author  = {Bedford, Eric and Pinchuk, Sergey},
  title   = {Convex domains with noncompact automorphism groups},
  journal = {Russian Acad. Sci. Sb. Math.},
  volume  = {82},
  year    = {1994},
  pages   = {1--20}
}

Prerequisites

06.10.03
06.10.08
06.10.12

Tier anchors

beginner: Krantz *Function Theory of Several Complex Variables* Ch. 12 informal; Greene-Krantz survey on automorphism orbits
intermediate: Krantz *Function Theory of Several Complex Variables* Ch. 11-12; Pinchuk *The scaling method and holomorphic mappings* (Proc. Symp. Pure Math. 52)
master: Wong 1977 (Invent. Math. 41); Rosay 1979 (Ann. Inst. Fourier 29); Pinchuk 1991; Greene-Kim-Krantz *The Geometry of Complex Domains*

References

Wong 1977 — Characterization of the unit ball in C^n by its automorphism group · Invent. Math. 41, 253-257 (originator: a bounded domain with non-compact automorphism group and a C^2 strongly pseudoconvex orbit-accumulation point is the ball)
Rosay 1979 — Sur une caractérisation de la boule parmi les domaines de C^n par son groupe d'automorphismes · Ann. Inst. Fourier (Grenoble) 29, 91-97 (removes Wong's boundedness/global hypotheses; local strong-pseudoconvexity at the orbit limit suffices)
Pinchuk 1991 — The scaling method and holomorphic mappings · Proc. Symp. Pure Math. 52, Part 1, 151-161 (the dilation/scaling proof; normal-family limit of rescaled automorphisms)
Krantz — Function Theory of Several Complex Variables · AMS Chelsea 340, 2nd ed., Ch. 11-12 (automorphism groups, boundary rigidity, statement of Wong-Rosay)
Greene-Kim-Krantz — The Geometry of Complex Domains · Birkhäuser Progress in Mathematics 291, 2011, Ch. 8-9 (scaling method, orbit accumulation, semicontinuity of automorphism groups)
Bun Wong / Klembeck — Bergman metric near a strongly pseudoconvex boundary point · Klembeck 1978, Indiana Univ. Math. J. 27, 275-282 (the Bergman metric approaches constant negative holomorphic curvature at a strongly pseudoconvex point)

Estimated time

beginner: 12m
intermediate: 35m
master: 55m