Inscribed Square Problem¶

The open square-peg conjecture that every planar Jordan curve contains four distinct points forming a nondegenerate Euclidean square, without requiring square order to match curve order.

Version: v1 · 2026-08-30 · History
Domain-specific #: 2077
Origin domain: geometry
Subdomain: inscribed configurations
Aliases: Square Peg Problem, Toeplitz Conjecture

Core Idea¶

The Inscribed Square Problem, also called the Square Peg Problem or Toeplitz Conjecture, asks whether every Jordan curve in the Euclidean plane contains the four vertices of a square of positive side length. A Jordan curve is the image of an injective continuous map \(\gamma:S^1\to\mathbb{R}^2\). “Inscribed” requires only that all four vertices lie on that image. The filled square need not lie inside the bounded region, and the order in which the curve visits the vertices need not agree with their cyclic order around the square.^[1]^[2]

The statement is a universal existence claim over arbitrarily rough simple closed curves. Smooth, convex, polygonal, locally monotone, and several lower-regularity or symmetry-controlled classes are known to contain such a square, but the general Jordan-curve case remains open. The difficulty is not finding squares on familiar shapes. It is preventing an existence witness from degenerating when regular curves approximate a wild curve.^[3]^[2]

This problem has a stable abstraction-level identity because it fixes a host class, a finite metric configuration, a nondegeneracy condition, and a universal quantifier. It then organizes a large research program around configuration spaces, parity or intersection arguments, equivariant topology, compactness with scale control, integration, and symplectic/Floer methods. A positive result for a restricted curve class is not the conjecture's general solution; it is progress obtained by adding enough structure to prevent the square from disappearing.

Structural Signature¶

Arbitrary planar Jordan curve + four distinct curve points + Euclidean square constraints + positive side length + unrestricted curve-order convention -> prove a witness exists for every curve, or exhibit a counterexample.

Let \(C=\gamma(S^1)\) for an injective continuous \(\gamma:S^1\to\mathbb{R}^2\). The claim is that there exist distinct points \(p_1,p_2,p_3,p_4\in C\), labeled in cyclic order around the square, and \(s>0\) such that

\[ \|p_i-p_{i+1}\|=s\quad(i\bmod 4), \qquad \|p_1-p_3\|=\|p_2-p_4\|=\sqrt{2}\,s. \]

Equivalently, the diagonals share a midpoint, have equal length, and are perpendicular. The parameters \(t_i\in S^1\) satisfying \(p_i=\gamma(t_i)\) are not required to appear around \(S^1\) in square order. This is important because cyclic and noncyclic inscriptions behave differently in configuration-space arguments.

The mandatory roles are:

host: the image of an arbitrary planar Jordan curve, not merely a smooth or convex loop;
guest configuration: the four-vertex Euclidean metric pattern of a square;
placement: four distinct points of the host matching that metric pattern;
nondegeneracy: side length \(s>0\), excluding four coincident limiting vertices;
quantification: at least one witness for every member of the host class;
status boundary: special-class theorems versus the unresolved all-Jordan-curves statement.

Recognition test. A result belongs to this problem only if it controls all four vertices, the exact square metric relations, and positive scale on a declared curve class. A rectangle of unspecified aspect ratio, an approximate square, a square touching rather than vertex-inscribed in the curve, or a theorem only for smooth curves does not by itself settle the full conjecture.

What It Is Not¶

It is not the graph-theoretic node Planarity. Planarity asks whether graph edges can be embedded without crossings and is characterized by forbidden minors. The square-peg problem starts with an already planar simple closed curve and asks for a four-point metric configuration on it.

It is not the general Inscribed Polygon Problem. Triangles, rectangles, cyclic quadrilaterals, regular polygons, and higher-dimensional configurations have different existence theorems and obstructions. Even the Rectangular Peg Problem is distinct: prescribing every rectangle aspect ratio is stronger on smooth curves, but results with a smoothness hypothesis do not settle arbitrary Jordan curves.^[4]

It is not a question about placing a square entirely inside the Jordan domain, fitting the largest square, or requiring the square's sides to lie on the curve. Only the four vertices must lie on the curve. It is also not the table theorem, which makes a level placement statement under different hypotheses.

It is not solved by density or numerical observation. Sampling a curve may find many approximate squares while missing pathology at arbitrarily small scales. Nor is it solved by smoothing and passing to a subsequence unless one proves a positive lower bound on square side length.

Finally, it is not a prime. The portable skeleton is an embeddability/existence question, but the literal roles—Jordan curve, Euclidean square, side and diagonal equations, cyclic ordering, curve regularity, and configuration-space boundary—remain geometric.

Scope of Application¶

The home scope is plane geometry and geometric topology, especially inscribed-configuration problems. Research varies the regularity of the host curve, the target quadrilateral, the ordering convention, and the ambient dimension. The canonical problem holds the host at “all planar Jordan curves” and the guest at “one nondegenerate square.”

Known positive regimes explain the boundary. Emch treated curves built from analytic arcs. Stromquist proved existence under a local monotonicity condition that includes convex curves, polygons, and suitable piecewise \(C^1\) curves.^[5]^[3] Smooth Jordan curves are covered, and Greene and Lobb proved the stronger result that every smooth Jordan curve inscribes a rectangle similar to any prescribed rectangle.^[4]

Low-regularity progress is also literal. Tao proved the conjecture when the curve is the union of two endpoint-matching graphs whose Lipschitz constants are strictly less than one; Greene and Lobb later extended the square conclusion to constants below \(1+\sqrt2\).^[2]^[6] These theorems enlarge the solved class but preserve a regularity margin.

Variants include cyclic versus noncyclic inscriptions, prescribed rectangles, other cyclic quadrilaterals, higher-dimensional curves, symmetric continua, and discrete analogues. They are related research programs, not aliases. The node excludes self-intersecting loops unless a variant explicitly relaxes the Jordan condition.

Clarity¶

The node clarifies four easily conflated choices. First, the host is a Jordan curve, not its filled interior. Second, inscription is a vertex condition, not containment of the square. Third, the square must be nondegenerate. Fourth, square order and curve order are separate.

It also clarifies what “open” means. The problem is not open for every curve anyone draws. It is open only because continuous injective curves can be extremely irregular. Many large and natural subclasses are settled. Reporting a new regularity theorem should therefore state exactly which curve class is added and why the proof controls scale.

The metric formulation separates shape from parametrization. Reparametrizing \(\gamma\) does not change its image or its inscribed squares. In contrast, conditions such as local monotonicity or a Lipschitz-graph representation introduce extra structure, and their constants cannot be dropped silently.

Manages Complexity¶

The problem compresses a vast search over four-tuples on a curve into a configuration-space zero or intersection question. One can encode pairs of points by midpoint, direction, and chord length, or encode ordered quadruples and impose equal-side and diagonal equations. Symmetries of the square reduce redundant labels, while transversality makes isolated solutions countable in generic smooth cases.

This formulation lets topology replace coordinate search. Parity, mod-2 intersection, obstruction theory, or non-embeddability results can force a solution without locating it explicitly. Greene and Lobb's smooth rectangular theorem converts the rectangle family into a symplectic/Lagrangian problem and uses non-embeddability of a suitable Klein bottle.^[4]

At the same time, the abstraction isolates the exact place where naive compactness fails. Approximating curves may have squares, but the solution set has a boundary stratum where four vertices collide. A proof must separate genuine configurations from that diagonal or provide a quantitative invariant that survives the limit.

Abstract Reasoning¶

Several reusable inferences follow. If \(\gamma_n\to\gamma\) uniformly and each \(\gamma_n\) has an inscribed square of side \(s_n\), compactness can produce a limiting quadruple. But it proves the desired result only if \(\inf s_n>0\). When \(s_n\to0\), all four vertices may converge to one point. Thus “approximate by smooth curves” is a proof plan only after a scale-separation lemma.

Generic parity arguments can also fail at nongeneric curves. A transverse intersection count may be stable under perturbation while solutions collide, become nontransverse, or escape into the degenerate boundary. Compactifying the configuration space records these escape routes rather than deleting them.

Regularity assumptions work because they restrict local geometry. A curve represented as a graph with controlled slope cannot oscillate arbitrarily near one point, which helps rule out pathological families of ever-smaller almost-squares. But proving one Lipschitz threshold does not license the endpoint or larger constants; the inequality is part of the theorem.

Finally, stronger variants can settle a special square case without approaching the general host class. Every smooth curve inscribing every rectangle shape includes squares, yet “smooth” remains the decisive limitation. Strength in the guest variable does not compensate for weakness in the host quantifier.

Knowledge Transfer¶

Within geometry, the problem transfers a standard workflow to other peg problems: encode a target configuration by distances or ratios, form a configuration space of curve points, quotient label symmetries carefully, identify degeneracy strata, and seek an intersection or obstruction that cannot vanish. This workflow applies to rectangles, rhombi, triangles, cyclic quadrilaterals, and higher-dimensional analogues, but each target has distinct constraints.

It also transfers the discipline of quantifier accounting. “Every host contains some guest” differs from “one host contains every guest shape,” and a theorem may strengthen one quantifier while weakening another. Greene–Lobb's smooth rectangular theorem is stronger in prescribed shape and narrower in curve regularity than the original conjecture.

Outside geometry, the general lesson that limits of witnesses may lose feasibility belongs to Compactness, Degeneracy, and Boundary Effects. The exact square-peg name and methods should not be exported metaphorically. The principal catalog transfer is to Embeddability: a fixed metric guest configuration must admit a faithful placement into every host in a geometric class.

Examples¶

Circle. Every choice of an orthogonal pair of diameters supplies a square, so a circle has infinitely many. This illustrates abundance in a highly symmetric host but says little about irregular curves.

Polygon. Polygons fall within solved regularity regimes and therefore have an inscribed square. The square's edges need not lie inside the polygon, and the four curve vertices need not be visited in square order. The theorem is about boundary membership.

Smooth prescribed rectangle. For a smooth Jordan curve and any rectangle \(R\), Greene and Lobb produce an inscribed rectangle similar to \(R\). Taking \(R\) to be a square proves the square case for smooth curves, but not for all Jordan curves.^[4]

Two controlled graphs. If the curve is the union of two graphs with common endpoint values and both Lipschitz constants below \(1+\sqrt2\), an inscribed square exists.^[6] The explicit threshold is part of the solved-class definition.

Degenerate limiting sequence. Let smooth approximations acquire smaller-scale wiggles. Suppose each approximation carries a square entirely in one wiggle and its side length tends to zero. Uniform convergence of curves and vertices yields four coincident limiting points, not a square. This is the canonical failed proof pattern.^[2]

Nonexample: self-crossing loop. A continuous closed parametrized loop with a transverse self-intersection is not a Jordan curve. It may contain square vertices, but it does not instantiate the canonical host class.

Structural Tensions¶

T1 — Smooth approximation versus positive scale. Smooth curves are solved and approximate arbitrary Jordan curves, but their witnessing squares may collapse. Diagnostic: require a uniform positive lower bound or an invariant excluding the diagonal.

T2 — Cyclic versus noncyclic order. Some configuration-space arguments naturally count vertices in curve order, while the problem permits any curve order. Diagnostic: declare the labeling quotient and prove that the counted configuration matches the intended statement.

T3 — Generic parity versus exceptional curves. Transverse solutions can be counted on generic curves; nongeneric limits can merge or develop multiplicity. Diagnostic: control compactification boundaries and invariance under perturbation.

T4 — Strong target theorem versus weak host theorem. Prescribing every rectangle aspect ratio is stronger than finding a square, but only on smooth hosts. Diagnostic: compare both host and guest quantifiers before calling one theorem a generalization.

T5 — Topological freedom versus metric rigidity. Jordan-curve topology is extremely permissive, while a square imposes exact Euclidean equalities. Diagnostic: keep metric constraints explicit; topological equivalence of curves does not transport squares.

Structural–Framed Character¶

The node is structurally precise but strongly field-framed. Its logical skeleton is universal embeddability of a finite configuration, and its proof obstacles involve compactness, symmetry, and boundary escape. Yet literal recognition requires a planar Jordan curve and a Euclidean square with exact side and diagonal relations.

It is not merely a historical topic. The statement remains stable across more than a century of primary research, and new methods repeatedly target the same host, guest, and degeneracy boundary. Still, the recurrence is internal to geometry and allied topology, so classification remains domain-specific.

Structural Core vs. Domain Accent¶

The structural core is: for every host in a class, determine whether a fixed structured guest has a nondegenerate embedding, and prevent witnesses from escaping through a boundary under limits. The domain accent fixes the host as a planar Jordan curve, the guest as the four-point Euclidean square metric, and the escape as side length tending to zero.

Embeddability captures the feasibility gate. Compactness captures subsequential limits but not preservation of nondegeneracy. Symmetry reduces label redundancy. None of these nodes, singly or together, states the Toeplitz claim or its cyclic-order and regularity boundaries without reconstructing the candidate.

The Inscribed Square Problem presupposes Embeddability. It asks whether the metric four-point configuration of a square admits a nondegenerate placement into every member of a constrained host class. The proposed DAG edge is therefore composition/presupposes/strict to prime:embeddability, rather than subsumption: the named conjecture is a proposition about universal embeddability, not a subtype of the abstract feasibility property.

Continuity, Compactness, Symmetry, Boundary Effects, and Degeneracy are important method or failure-mode relations. They remain prose-only to avoid turning proof ingredients into taxonomic parents. Planarity is not proposed because its live node is graph-theoretic crossing-free drawability.

Relationships to Other Abstractions¶

Current abstraction Inscribed Square Problem Domain-specific

Parents (1) — more general patterns this builds on

Inscribed Square Problem presupposes Embeddability Prime

The Inscribed Square Problem presupposes Embeddability.

Hierarchy paths (2) — routes to 2 parentless roots

Inscribed Square Problem → Embeddability → Constraint

Show alternative path (1)

Neighborhood in Abstraction Space¶

Inscribed Square Problem sits in a sparse region of the domain-specific corpus (96^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Planarity is the strongest frozen semantic match but concerns graph embeddings and forbidden minors. Birthday Problem and Graph Coloring are named problems with different probability and constraint structures. Matching, Emptiness Problem, and P versus NP Problem share search or existence language but none supplies the Jordan-curve metric configuration. Nine-Point Conic is an inscribed-geometry neighbor with a different theorem identity. Problem Representation describes how an encoding changes available search; it does not parent the mathematical proposition itself.

Outside the catalog, distinguish the Square Peg Problem from the Rectangular Peg Problem, table theorem, inscribed-square optimization, squares inscribed in triangles, circumscribed squares, approximate-square detection, and digital or discrete analogues. Preserve “Toeplitz conjecture” as the geometry name, not the unrelated Toeplitz conjectures of operator or matrix theory.

References¶

[1] Matschke, Benjamin. “A Survey on the Square Peg Problem.” Notices of the American Mathematical Society 61, no. 4 (2014): 346–352. Authoritative survey of the statement, configuration-space methods, solved cases, generic criteria, and related problems. registry ↩

[2] Tao, Terence. “An Integration Approach to the Toeplitz Square Peg Problem.” Forum of Mathematics, Sigma 5 (2017): e30. Defines the nondegenerate problem, explains small-square boundary failure, and proves the two-graph Lipschitz-constant-below-one case. registry ↩a ↩b ↩c ↩d

[3] Stromquist, Walter. “Inscribed Squares and Square-Like Quadrilaterals in Closed Curves.” Mathematika 36, no. 2 (1989): 187–197. Proves smooth and local-monotonicity regimes and records the remaining general question. registry ↩a ↩b

[4] Greene, Joshua Evan, and Andrew Lobb. “The Rectangular Peg Problem.” Annals of Mathematics 194, no. 2 (2021): 509–517. Proves that every smooth Jordan curve inscribes a rectangle of every prescribed similarity class. registry ↩a ↩b ↩c ↩d

[5] Emch, Arnold. “On Some Properties of the Medians of Closed Continuous Curves Formed by Analytic Arcs.” American Journal of Mathematics 38, no. 1 (1916): 6–18. Early positive theorem for analytic-arc regularity. registry ↩

[6] Greene, Joshua Evan, and Andrew Lobb. “Square Pegs Between Two Graphs.” Commentarii Mathematici Helvetici (online first, 2026). Extends the two-graph square theorem to Lipschitz constants below \(1+\sqrt2\) and uses Jordan Floer homology. registry ↩a ↩b

[7] Chambers, Gregory R. “On the Square Peg Problem.” 2022. Proves a positive result for Jordan curves quantitatively close to a \(C^2\) Jordan curve. registry

[8] “Inscribed square problem,” Wikipedia, frozen revision 1336213497, 2026-02-02. Discovery provenance only; status and mathematical claims were independently checked against the sources above. registry