Erdős Problem #353
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (not automatically the live result).
Subject areas: geometry.
Read this status correctly
The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.
The question in plain language
Erdős Problem 353 asks about shapes forced inside a very large measurable subset of the plane. Let \(A\) be a measurable subset of \(\mathbb{R}^2\) with infinite Lebesgue measure. Must four points of \(A\) be the vertices of an isosceles trapezoid having area \(1\)? The same question was posed for an isosceles triangle, a right-angled triangle, a cyclic quadrilateral, and a convex polygon whose sides are all congruent. The original question is recorded in the maintained Erdős Problems entry and is attributed there to Erdős's 1983 source Er83d.
The trapezoid part is an existence statement. It does not say that every four points in \(A\) work, that the shape is unique, or that a finite search will find it. “Infinite measure” is also stronger than merely being unbounded: an unbounded set can have finite or even zero measure. The measurable-set hypothesis is used by the proof through density arguments.
A small example
Here is an exact finite check of the geometry. Take, in cyclic order,
The two horizontal bases have lengths \(2\) and \(1\), and the height is \(2/3\). Thus the trapezoid area is
The two nonparallel sides have squared lengths
so they are equal. A direct rational shoelace calculation gives area \(1\), and the two leg-squared values are both \(25/36\); these are the exact outputs of the finite arithmetic check for this example. This verifies one concrete isosceles trapezoid, not the Erdős assertion: its four points form a specially chosen finite set, not an arbitrary measurable infinite-measure set.
What is known
The primary resolution is Junnosuke Koizumi, “Isosceles trapezoids of unit area with vertices in sets of infinite planar measure,” arXiv:2501.01914v1, submitted 3 January 2025; the paper is also associated with DOI 10.1090/proc/17322. Its Theorem 2, stated in the Introduction on PDF page 1 and proved in Section 3 beginning on PDF page 4, says exactly that every measurable \(S\subseteq\mathbb{R}^2\) of infinite Lebesgue measure contains the four vertices of an isosceles trapezoid of area \(1\).
The proof is constructive in mechanism, although not a finite construction from a description of \(S\). It chooses a density point near a small disk and another density point far away. For a large scale factor \(R\), it considers points that remain in \(S\) after scaling by \(R^{-1}\). A radius-dependent rotation \(f\) is chosen so that \(p\), \(f(p)\), \(R^{-1}f(p)\), and \(R^{-1}p\) form an isosceles trapezoid. The angle is selected so that
and the map preserves area. Density and a contradiction argument then force both \(p\) and \(f(p)\) into the required large-measure subset. The same paper proves stronger hypotheses for the isosceles- and right-triangle variants: unbounded measurable sets of positive measure suffice.
Kovač and Predojević’s 2024 paper supplies the positive cyclic-quadrilateral result and the negative congruent-sides result, which explains why the catalogue groups several related questions under one entry. Those are useful neighboring results, not substitutes for Koizumi’s trapezoid theorem.
GCL's current state
The frozen catalogue packet is teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry number: "353", whose recorded Git blob is 4e60263a80af6207aca494d2e1388e44fa372889. It records informal state proved, last update 2026-06-21, and formal state Lean, with formalized state yes dated 2026-07-22. The catalogue is evidence about its ledger state, not GCL mathematical certification.
The pinned formal artifact is google-deepmind/formal-conjectures commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob 08f75db840a35b75d31b5888bd4a17455380d004. Its main theorem quantifies over every \(A : \mathrm{Set}\,\mathbb{R}^2\), assumes measurability and volume A = ⊤, and asks for four members forming an IsIsoscelesTrapezoid whose convex-hull volume is \(1\). It also records triangle, right-angle, cyclic, and congruent-side variants. Every theorem body in the pinned file is by sorry; therefore the file is a formal statement and interface, not a replayed proof.
What remains
The mathematical proposition has a readable primary proof, but a bounded successor task remains valuable. First, compare the formal predicate IsIsoscelesTrapezoid and convex-hull volume with Koizumi’s geometric convention, including ordering, distinctness, parallel-side, and degeneracy requirements. Second, replace the sorry in only the main trapezoid theorem, or document the exact missing lemmas if that is not yet tractable. Third, separately inspect the three Koizumi variants and the Kovač–Predojević variants; their formal statements should not be treated as automatically equivalent to the prose source.
Acceptance should require a source-linked semantic comparison, a compiling proof with no sorry for the selected theorem, and a check that the hypotheses are no stronger or weaker than the cited theorem. Failure of any comparison is a falsification of equivalence, not evidence against the mathematics. Stop after the selected theorem, its definitions, and the required measure/geometry lemmas have been audited; do not turn this article into an exhaustive Lean build or a new claim about publication.
Sources and claim boundaries
This article was prepared from the original GCL issue #1623, the frozen commits and blobs named above, the live Problem 353 page, Koizumi’s arXiv v1, and the cited Erdős and Kovač–Predojević sources. The downloaded arXiv v1 PDF used for the source check had SHA-256 c96a5d4197e1b4ab0a4b17c23bc0f00a530a067ddbb79b109d7ee6024bfcfcaf; arXiv supplies the version identifier rather than a Git commit.
The external paper establishes a mathematical result in its own source. The catalogue records accepted coverage according to its own ledger. The formal file records a theorem interface with sorry. None of these layers alone is GCL certification, and this article supplies no Lean certification. Original SOURCE RESULT/1 citation: authenticated original RESULT/1.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1623 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.