Skip to content

Erdős Problem #397

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: disproved (not automatically the live result).
Subject areas: number theory, binomial coefficients.

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

Write \(C(k)=\binom{2k}{k}\) for the central binomial coefficient. Erdős Problem 397 asks whether there are only finitely many equalities

\[ \prod_i C(m_i)=\prod_j C(n_j) \]

when all the indices \(m_i\) and \(n_j\) are distinct. The original wording does not spell out a fixed number of factors, positivity convention, or whether “distinct” means distinct within each side or across both sides. The natural reading for the problem is a finite pair of disjoint lists of positive integers, with order irrelevant. The later formal encoding uses disjoint finite subsets of \(\mathbb{N}\), which makes that interpretation precise but also admits \(0\) and empty sets.

The source question is printed on page 77 of P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980). The maintained Erdős Problems entry gives the same question and records a negative answer. The exact GCL assignment and its source packet are in issue #1667.

A small example

The maintained entry gives a family. For every integer \(a\geq 2\), set \(c=8a^2+8a+1\). It claims

\[ C(a)C(2a+2)C(c)=C(a+1)C(2a)C(c+1). \]

Take \(a=2\), so \(c=49\). The two index sets are \(\{2,6,49\}\) and \(\{3,4,50\}\); they are disjoint and both sums are \(57\). Exact factorial arithmetic gives

\[ C(2)=6,\quad C(6)=924,\quad C(49)=25477612258980856902730428600, \]

and

\[ C(3)=20,\quad C(4)=70,\quad C(50)=100891344545564193334812497256. \]

Multiplication on both sides gives the identical integer

\[ 141247882363789870668737496158400. \]

This is reproducible with exact arithmetic using comb(2*k,k) from Python’s standard math library; no floating-point approximation is involved. It verifies one finite collision and shows why the original finiteness claim cannot survive if the displayed family holds for every \(a\). It does not, by itself, prove the identity for all values of \(a\), establish the intended domain convention, or certify the formal Lean theorem.

What is known

The scanned primary source on page 77 records the problem as an open question about distinct indices. It is a source for the original statement, not evidence that the question was solved there. The maintained page, last edited 12 January 2026, attributes the negative answer to Neel Somani and says that further families occur in the discussion. Its formula is mathematically checkable without trusting a numerical approximation.

Indeed, the standard ratio

\[ \frac{C(t+1)}{C(t)}=\frac{2(2t+1)}{t+1} \]

reduces the displayed identity to a polynomial identity. After cancellation, it is equivalent to

\[ (4a+1)(4a+3)(c+1)=2(2a+1)^2(2c+1), \]

which holds exactly when \(c=2(2a+1)^2-1=8a^2+8a+1\). Since \(c\) strictly increases with \(a\), the family supplies infinitely many different collisions under the disjoint-positive-index interpretation. This is an elementary verification of the claimed mechanism, not a claim that GCL has certified a theorem.

A 2013 MathOverflow discussion supplies an earlier explicit nontrivial collision and exploratory arguments. Those posts are useful corroboration and context, but they are not the original Erdős–Graham source and do not replace a named proof publication for the later family.

GCL's current state

The frozen catalogue packet is teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, whose data/problems.yaml blob is 4e60263a80af6207aca494d2e1388e44fa372889. Its entry for 397 records informal status disproved, status disproved (Lean), and dates 10 January 2026 and 21 November 2025 for the associated fields. The live entry agrees with that snapshot’s substantive direction, though the ledger is a dated catalogue record rather than mathematical certification.

The pinned FormalConjectures file is at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob e0ec4d360291b31b7a95d48c615700ec9fd932df. It encodes an equivalence between answer(False) and finiteness of the set of disjoint pairs of finite subsets of \(\mathbb{N}\) whose central-binomial products agree. The theorem ends with by sorry; the file’s presence and “formalized” metadata therefore do not establish a checked proof or semantic equivalence with the original wording.

At the time of this audit, issue #1667 has no comments and no coordinator-posted RESULT/1. V2 ledger state, not this article, decides accepted coverage.

What remains

The main documentary obligation is provenance. The original book gives the question, while the live entry gives the later family but attributes it to “Somani, using ChatGPT” without a conventional paper, theorem number, or proof manuscript. A bounded successor should preserve that attribution exactly, derive the family from the displayed ratio identity, and compare its hypotheses with the original page and the Lean encoding.

The formal route also needs a semantic check: decide whether \(\mathbb{N}\) should be replaced by positive integers, whether empty sets and the index \(0\) are admissible, and whether pairwise disjoint finite sets exactly represent “the \(m_i,n_j\) distinct.” A formal proof of the elementary family or of infinitude would be useful, but a theorem ending in sorry remains a statement source, not a proof result.

Sources and claim boundaries

The exact finite calculation and ratio reduction are audit evidence only. They do not confer GCL mathematical certification, publication authority, or a stronger status than the dated ledger and sources support.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1667 — 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.

Back to catalogue · Compendium introduction