Inscribed square problem
Does every closed curve in the plane that does not cross itself pass through all four corners of some square?
Summary
Why it is open
Otto Toeplitz posed the problem in a talk in 1911. The conference report says he had solved it only for convex curves, and Benjamin Matschke notes that Toeplitz seems never to have published a proof [1]. In modern terms the conjecture says that every Jordan curve, a continuous closed curve in the plane that does not cross itself, contains the four vertices of a square of positive side length [1, 3].
Most proved cases assume that the curve is regular enough [1, 3]. Terence Tao describes the proofs as "homological": they use algebraic topology, and many of them show that a suitably generic curve inscribes an odd number of squares [3]. The natural route to the general case is to approximate a rough curve by smooth ones, each of which inscribes a square, and pass to a limit. But the limit of those squares can be a single point, and in 2014 Matschke wrote that "so far there is no argument known that can deal with this problem" [1]. Tao calls this degeneration "the primary reason" the regularity hypotheses cannot be removed [3]; Joshua Greene and Andrew Lobb call it shrinkout [7].
Matschke names one kind of statement that would be enough: if every smooth curve enclosing a disc of radius r inscribed a square of side at least εr, for some fixed ε > 0, the approximation argument would extend the result to every continuous curve [1].
No institution or prize sets the bar. A resolution would be a proof for every Jordan curve, or a single Jordan curve with no inscribed square.
What is known
Early cases (1913–1944). Arnold Emch gave a proof for sufficiently smooth convex curves in 1913 and, in 1916, for piecewise analytic curves with finitely many inflection points and other singular points at which one-sided tangents exist [1]. Later authors completed the convex case [1, 3]. In 1929 L. G. Schnirelman proved the problem for a class of curves slightly larger than the twice continuously differentiable (C²) curves; an extended version appeared after his death, in 1944 [1]. His argument deforms an ellipse, which inscribes exactly one square up to symmetry, into the given curve; smoothness keeps the squares from shrinking to a point during the deformation, so the parity of their number is preserved [1].
Wider classes (1989–2014). In 1989 Walter Stromquist proved the problem for locally monotone curves [1, 2]: near each of its points such a curve moves strictly monotonically in some fixed direction [1]. His abstract notes that the condition holds for convex curves, polygons and, with certain restrictions, piecewise C¹ curves, and that "for other curves, the question remains open" [2]. Other proofs cover curves symmetric about a line or a point, piecewise linear curves, C¹ curves and curves of bounded curvature without cusps [1, 3]. Matschke proved it for curves with no inscribed "special trapezoids" of some fixed size, a condition met by an open and dense set of curves [1]. He also proved that a closed curve lying in the annulus 1 ≤ |x| ≤ 1 + √2 and going around its hole, so that it cannot be shrunk to a point within the annulus, inscribes a square of side at least √2 [1].
Lipschitz graphs and nearby curves (2017–2026). Tao proved that a curve formed by the graphs of two functions over one interval, meeting at the endpoints, inscribes a square if both functions have Lipschitz constant (a bound on their slope) below 1 [3]. In a paper published in 2026, Greene and Lobb raised the bound to 1 + √2, and showed that with constant 1 the curve inscribes rectangles of every shape [8]. Gregory Chambers proved in 2025 that a Jordan curve inscribes a square if it lies within 1/(10κ) of a C² Jordan curve whose maximum curvature is κ, in a precise sense that includes a winding condition [6].
Rectangles (2014–2023). Every Jordan curve inscribes some rectangle, by an argument of Vaughan [1, 9]. In 2014 Matschke called the version with a prescribed aspect ratio "wide open, even for smooth or piecewise linear curves" [1]. In a paper posted in 2020 and published in 2021, Greene and Lobb proved that every smooth Jordan curve inscribes a rectangle similar to any given rectangle [4]. Their proof uses Shevchishin's theorem that the Klein bottle has no smooth Lagrangian embedding in C², four-dimensional complex space; Lagrangian surfaces are a special class of surfaces studied in symplectic geometry [4]. They then extended the result to every cyclic quadrilateral, one whose vertices lie on a circle [5]. Smoothness matters there: for every cyclic quadrilateral that is not an isosceles trapezoid, some continuous Jordan curve, even a triangle, does not inscribe it [7].
Recent preprints (2024–2026). The results in this paragraph are in preprints, not yet peer reviewed. In 2024 Greene and Lobb built a Floer homology, an invariant from symplectic geometry, generated by inscribed rectangles, and proved that a rectifiable Jordan curve (one of finite length) inscribes a square if it encloses more than half the area of a circle of the same diameter [7]. Tomohiro Asano and Yuichi Ike, using microlocal sheaf theory, report a proof that every rectifiable Jordan curve and every locally monotone curve inscribes rectangles of every shape; they write that this is, to their knowledge, "the first result that gives an affirmative answer to the square peg problem" for all rectifiable curves [10]. Their preprint was posted in December 2024 and revised in January 2026; an earlier revision fixed an error in its Section 5 [10]. In April 2026 Greene and Lobb cited it as solving the rectangular peg problem for rectifiable curves, and proved that every Jordan curve, with no regularity assumed, inscribes θ-rectangles (rectangles whose diagonals meet at angle θ) for a set of angles θ of measure at least A/R², where A is the enclosed area and R half the diameter [9]. Two September 2026 preprints identify further curves that meet the Asano–Ike criterion: curves of finite p-variation with 1 ≤ p < 2, a notion from rough-path theory, and curves meeting local variation or conformal-map conditions [11, 12]. In 2024 Cole Hugelmeyer reported a proof of a periodic version of the problem that Tao had proposed and noted was open even for regular curves [3, 13].
Main approaches
Parity and obstruction theory. Count inscribed squares modulo 2, or with signs, and show that the count cannot change as the curve is deformed [1]. Strongest evidence: the proofs for smooth, locally monotone and generic curves [1, 2]. Main difficulty: for a general curve the degenerate solutions, in which the four points coincide, cannot be separated from the squares, which Matschke calls "the basic reason" these methods have not solved the problem [1].
Integrals. Tao attaches integrals to the curve, such as the signed area under it, which have "more stability properties under limits than intersection numbers" [3]. Strongest evidence: the Lipschitz-graph theorem [3]. Main difficulty: the proof applies the Jordan curve theorem to an auxiliary curve that the slope bound guarantees to be simple, and Tao writes that he does not know how to adapt the argument to the general case [3].
Symplectic geometry and sheaves. Inscribed rectangles correspond to intersection points of two Lagrangian tori in C², and invariants from Floer homology or microlocal sheaf theory keep track of those intersections when a rough curve is approximated by smooth ones [4, 7, 10]. Strongest evidence: the theorems for smooth curves [4, 5], the two-graph theorem [8], and spectral invariants that Greene and Lobb describe as "robust under approximation", giving "control over shrinkout for rectangles for a wide class of curves" [7]. Main difficulty: the results for rough curves are so far in preprints [7, 9, 10], and for an arbitrary Jordan curve the April 2026 result guarantees θ-rectangles only for a set of angles of measure at least A/R² [9].
Large squares. Show that some inscribed square has size bounded below, so that it cannot vanish in a limit [1]. Strongest evidence: the annulus theorem [1] and the area condition of Greene and Lobb [7]. Main difficulty: Matschke expects this to need "more geometric than merely topological ideas" [1].
Open sub-questions
- A wider annulus. Can the outer radius 1 + √2 in Matschke's annulus theorem be increased by some small ε > 0, a question he recorded as open in 2014 [1]?
- The Asano–Ike criterion. Which Jordan curves, beyond the rectifiable, locally monotone and finite p-variation curves and those covered by Asano and Ike's 2026 conditions, admit the "continuous Legendrian lift" their theorem requires [10, 11, 12]?
- The square in the measure result. For which classes of Jordan curves that are not rectifiable can the set of angles in Greene and Lobb's April 2026 theorem be shown to contain π/2, the angle of a square [9]?
- Tao's combinatorial model. Is Tao's "almost purely combinatorial" conjecture on the sign patterns of sums y₁ + y₂ + y₃ of numbers from finite sets, which he called a tractable toy model and could not fully resolve in 2017, true [3]?
- Trapezoids. Does every Jordan curve inscribe every isosceles trapezoid, the trapezoidal peg problem for continuous curves, the only cyclic quadrilaterals not ruled out by triangles [1, 9]?
Work on this in Papex
If you have a testable answer to one of the sub-questions above, you can draft it as a private hypothesis in Papex, or prepare the draft from Claude or ChatGPT with the Papex connector. You review the exact version, sign it and decide whether to publish it; a public record that names this problem is listed on this page. A Papex record shows who signed which version and when. It is not peer review and does not mean the hypothesis is correct.
Key references
- Matschke, B. (2014). A survey on the square peg problem. Notices of the American Mathematical Society 61(4), 346–352. 10.1090/noti1100 · Checked 16 Oct 2026
- Stromquist, W. (1989). Inscribed squares and square-like quadrilaterals in closed curves. Mathematika 36(2), 187–197. 10.1112/S0025579300013061 · Checked 16 Oct 2026
- Tao, T. (2017). An integration approach to the Toeplitz square peg problem. Forum of Mathematics, Sigma 5, e30. 10.1017/fms.2017.23 · Checked 16 Oct 2026
- Greene, J. E. & Lobb, A. (2021). The rectangular peg problem. Annals of Mathematics 194(2), 509–517. 10.4007/annals.2021.194.2.4 · Checked 16 Oct 2026
- Greene, J. E. & Lobb, A. (2023). Cyclic quadrilaterals and smooth Jordan curves. Inventiones mathematicae 234(3), 931–935. 10.1007/s00222-023-01212-6 · Checked 16 Oct 2026
- Chambers, G. R. (2025). On the square peg problem. Discrete & Computational Geometry 73(4), 1144–1153. 10.1007/s00454-025-00720-x · Checked 16 Oct 2026
- Greene, J. E. & Lobb, A. (2024). Floer homology and square pegs. Preprint, arXiv:2404.05179 (version 2, 13 July 2024), not peer reviewed. Source · Checked 16 Oct 2026
- Greene, J. E. & Lobb, A. (2026). Square pegs between two graphs. Commentarii Mathematici Helvetici, published online 2 June 2026. 10.4171/cmh/619 · Checked 16 Oct 2026
- Greene, J. E. & Lobb, A. (2026). Jordan curves inscribe a positive measure of rectangles. Preprint, arXiv:2604.17116 (version 1, 18 April 2026), not peer reviewed. Source · Checked 16 Oct 2026
- Asano, T. & Ike, Y. (2024). The rectifiable rectangular peg problem. Preprint, arXiv:2412.21057 (version 3, 5 January 2026; the version 2 comment reads "An error in Section 5 fixed"), not peer reviewed. Source · Checked 16 Oct 2026
- Li, X. & Pan, Y. (2026). Rectangular pegs on Jordan curves of finite p-variation. Preprint, arXiv:2609.04918 (version 1, 4 September 2026), not peer reviewed. Source · Checked 16 Oct 2026
- Asano, T. & Ike, Y. (2026). Liouville regular Jordan curves. Preprint, arXiv:2609.11598 (version 1, 10 September 2026), not peer reviewed. Source · Checked 16 Oct 2026
- Hugelmeyer, C. (2024). A solution to the periodic square peg problem. Preprint, arXiv:2407.20412 (version 1, 29 July 2024), not peer reviewed. Source · Checked 16 Oct 2026
Source
Open, as listed by Wikipedia · Unsolved problems in mathematics. Checked 08 Oct 2026.
The status is the source’s. Papex does not decide whether a problem is solved.
Records on this problem
No public record addresses this problem yet.