Circle-domain uniformization and removable-boundary rigidity

Source and verification

This article explains family 071 of the October 6, 2026 OpenAI collection. The research conclusions are attributed to its manuscripts. The source review covers statements, introductions, and the proof mechanisms described below; it is not an independent verification of the complete proofs.

Overview

Circle-domain uniformization seeks a conformal coordinate system in which every hole is round or shrinks to a point. Family 071 pairs an unrestricted existence claim with a rigidity theorem under a boundary-removability hypothesis. The main issue is controlling infinitely many boundary components during a limiting construction.

🏷️ Preliminaries: conformal models and boundary components

A domain is a nonempty connected open set. The Riemann sphere is $\hat{\mathbb{C}}=\mathbb{C}\cup \{\infty \}$. A circle domain is a domain in this sphere whose complementary connected components are closed round disks or single points. A conformal equivalence is a bijective holomorphic map with holomorphic inverse, interpreted in spherical coordinates near infinity.

For a simply connected proper plane domain with at least two boundary points, the Riemann mapping theorem gives a disk model. Several holes introduce moduli: an annulus, for example, has a conformally meaningful ratio between its radii. Circle-domain uniformization asks for round holes without requiring them to have prescribed radii or positions.

A compact set $E$ is conformally removable if every orientation-preserving homeomorphism of the sphere that is conformal outside $E$ is Möbius. A Möbius map has the form $(az+b)/(cz+d)$ with $ad-bc\ne 0$. Removability begins with a globally defined homeomorphism. A map between circle domains is initially defined only on the domains; extending it across their boundaries is a separate problem.

🏷️ The two manuscript claims

Family 071, as stated in the manuscripts

Every domain in $\hat{\mathbb{C}}$ is conformally equivalent to a circle domain. Moreover, if a circle domain has conformally removable boundary, every conformal equivalence from it to another circle domain is the restriction of a Möbius map. (OpenAI, 2026) (OpenAI, 2026a)

The existence statement imposes no countability or separation assumption on the holes. The second statement is conditional on removability; it does not give Möbius uniqueness for every circle domain. Its hypothesis is not an assumed extension of the given conformal map.

🏷️ Dirichlet energy as a detector of unwanted holes

For a real function $u$ on a plane domain, its Dirichlet energy is

$$\mathcal{E}(u)=\int |\nabla u{|}^{2}dA.$$

This quantity is conformally invariant: under a conformal change of coordinates, the squared derivative from the gradient cancels the area Jacobian. It therefore provides information that survives passage to conformal models.

The elementary principle behind a barrier argument is worth deriving. If an absolutely continuous function changes from zero to one along an interval of length $L$, then Cauchy–Schwarz gives

$$\begin{array}{rl}1& ={\left|{\int }_0^L{u}^{\prime }(t)dt\right|}^2\\ & \le L{\int }_0^L|u^{\prime }(t){|}^{2}dt.\end{array}$$

If many parallel intervals require this change, integration across the intervals gives a positive lower bound for the two-dimensional energy. Thus a construction of barriers with energy tending to zero can rule out a limiting component with a substantial projection onto an axis. The manuscript has to construct the barriers simultaneously and justify the relevant boundary traces; the elementary inequality alone does neither.

🏷️ Architecture of the uniformization argument

Finite circle-domain uniformizations give a starting point, but a locally uniform limit of maps need not preserve the desired geometry of every hole. A limiting complementary component can contain a round disk together with an additional protrusion. Other components may not have a distinguished disk at all.

The existence manuscript separates these obstructions. It marks countably many components and associates a limiting disk $D_b$ to each marked label $b$. It then seeks

$$K_b=\{\begin{array}{ll}{D}_b,& b\text{ marked},\\ \text{a singleton},& b\text{ unmarked},\end{array}$$

where $K_b$ is the actual complementary component in the limit.

The source’s transfer theorem retains finite collections of analytic tests on larger and larger compact cores while making their added energy and boundary oscillations small. A barrier-versus-path-law alternative supplies the tests: either an end can be cheaply separated from an interior anchor, or a controlled probability distribution of paths reaches it. A crossing argument then limits the ends that cannot be separated to a countable collection.

Small real barriers collapse unmarked components. For a marked component, tests with prescribed additive period detect protrusions through a boundary integral. The crucial selection argument retains all earlier tests while new ones are added. Without this compatibility, solving one boundary problem could undo another.

🏷️ The rigidity companion

The companion uses the retained-test construction to obtain a Sobolev extension statement across removable, totally disconnected sets. It then handles disk components with masks and estimates the remaining traces on almost every line. A flux inequality, compared with source and target area bounds, supplies the rigidity conclusion.

This explains why saying “extend the map and apply removability” would omit the hard part. The source builds enough extension and trace control to make that final use of removability legitimate. The proof outline here records those dependencies; it does not verify the full transfer, selection, or trace arguments.

🏷️ Key steps and proof outlines

Retaining tests while changing the conformal model

The finite-transfer theorem is stronger than compactness of normalized conformal maps. Its input is a finite compatible list of scalar coordinates, a compact core, and an error tolerance $\epsilon$. Its output is a finite bordered surface whose conformal structure agrees with the original one near the core. Every old coordinate is transferred unchanged there, while the sum of its added Dirichlet energies is below ${\epsilon }^2$. Boundary oscillations are controlled in ${\ell }^2$, and prescribed periods of multivalued coordinates are preserved. These are three different controls: agreement fixes previously established information, energy controls new regions, and periods detect topology. See “Finite transfer and normalized limits” in the existence manuscript. (OpenAI, 2026a)

The proof strategy first constructs compatible bands and terminal disks around the ends. It modifies the geometry outside the retained core so that extending the finite coordinate list costs little energy. Finite circle-domain uniformization then transports both geometry and tests. A normal-family limit gives a conformal map, but the limiting tests supply the information that the map alone lacks. In particular, one cannot apply an arbitrary exhaustion and assume that its boundary oscillations vanish.

Why a diagonal selection needs stable certificates

A useful model of the selection is to choose tolerances ${\epsilon }_j$ with ${\sum }_j{\epsilon }_j<\infty$. At stage $j$, the retained list contains every test introduced before stage $j$. Later stages must preserve its conclusion with an error bounded by the remaining tail. Thus a test added at a fixed stage survives the limit. The source’s actual compatible selection also retains path-law exclusions and period information; the summable-error model explains the role of its stability requirements.

For an unmarked end, the desired certificate is a real barrier separating that end from an interior anchor at arbitrarily small energy. The barrier theorem obtains it from lower bounds for weighted lengths of all allowed paths. Its weights consist of an $L^2$ density in the domain and ${\ell }^2$ costs at ends. The length functional is converted into a truncated distance function, with cutting and joining arguments ensuring compatibility with the earlier tests. A dual path law handles the alternative in which no inexpensive barrier exists. The subsequent crossing argument bounds the exceptional ends by a countable set, which becomes the marked set.

From certificates to round complementary components

For an unmarked limiting component, suppose a coordinate projection has positive length. Almost every transverse line through that projection then sees a definite barrier change. The one-dimensional Cauchy–Schwarz calculation above, integrated over those lines, forces positive energy. Retained barriers have energy approaching zero, giving a contradiction. Both coordinate projections must have length zero; connectedness then forces the component to be a point.

Marked components require a different test. A real barrier would try to collapse the disk that is supposed to survive. A periodic coordinate instead has increment one around its marked hole and zero around the others. Its boundary integral can detect an extra protrusion while respecting the disk’s required period. The stream argument uses these tests to identify the entire component with the limiting disk. This division of labor is the reason the proof carries both real and periodic coordinates.

flowchart TD
 A[Finite compatible tests] --> B[Transfer with small tail energy]
 B --> C[One normalized limit retaining all tests]
 C --> D[Real barriers collapse unmarked ends]
 C --> E[Period tests exclude marked protrusions]
 D --> F[Circle domain]
 E --> F

The rigidity companion’s trace estimate

The companion must control traces across countably many disks before removability can be used. Its ball-synthesis lemma has disjoint sets $E_j\subset B_j$, each occupying a fixed proportion of a ball of radius $r_j$, and coefficients $a_j\ge 0$. Duality and the maximal theorem give

$${\Vert \sum _j\frac{a_j}{r_j}{1}_{B_j}\Vert }_2\lesssim {\left(\sum _j{a}_j^2\right)}^{1/2}.$$

Indeed, pairing the sum with $g\ge 0$ bounds each ball average by a constant times $Mg$ on $E_j$. Cauchy–Schwarz separates the $a_j$ from the $L^2$ integrals of $Mg$; disjointness lets those integrals be summed once. The $L^2$ maximal inequality closes the bound. Applied to disk masks, this turns square-summable geometric errors into a trace estimate. Sobolev extension across removable dust and the resulting almost-every-line trace control then feed the final length–area argument. Circle-domain existence alone would not provide these analytic estimates. (OpenAI, 2026b)

See Also

• Family 072: Brennan and integral means studies another way of measuring conformal boundary distortion, through powers of the derivative rather than the geometry of complementary components.

🏷️ Manuscripts and verification scope

• Removable Boundaries and Rigidity of Circle Domains — the pinned manuscript; its directory supplies the source and citation metadata.
• Koebe’s Circle-Domain Conjecture — the pinned manuscript; its directory supplies the source and citation metadata.

The repository provides a formalization-scope page. Its scope description lists circle-domain existence and removable-boundary rigidity. These are statements about the repository’s documentation: no Lean build, axiom audit, or comparator-equivalence check was performed for this article.

OpenAI collection gives the reading sequence and source policy; coverage register records which families have articles and which remain to be written.

🐻  OpenAI 2026a. Koebe’s Circle-Domain Conjecture.

🐻  OpenAI 2026b. Removable Boundaries and Rigidity of Circle Domains.