Skip to content

Erdős Problem #420

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

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

The divisor-counting function \(\tau(n)\) gives the number of positive divisors of an integer. If \(n!\) is a factorial, then increasing \(n\) changes its prime factorization, and with it the number of divisors. Erdős Problem 420 asks about the growth of the divisor count over a short interval of consecutive factorials, rather than only one step as in Problem 419.

For a nonnegative real-valued function \(f(n)\), define

\[ F(f,n)= \frac{\tau\bigl((n+\lfloor f(n)\rfloor)!\bigr)} {\tau(n!)}. \]

The floor is important: the factorial argument is an integer. The question has several logically distinct parts. Does \(F((\log n)^C,n)\) tend to infinity for all sufficiently large fixed constants \(C\)? Is the collection of values of \(F(\log n,n)\) eventually dense throughout the real interval \((1,\infty)\)? What if \(f\) increases monotonically to infinity but never exceeds \(\log n\)? A convergence claim, a liminf claim, and a density claim are different. None follows simply from the existence of infinitely many large values.

A small example

The prime factorization of a positive integer \(N=\prod p^{a_p}\) gives

\[ \tau(N)=\prod_{p\mid N}(a_p+1). \]

For \(n=5\), the base factorial is \(5!=120=2^3\cdot3\cdot5\), so \(\tau(5!)=16\). With a one-step shift, \(6!=720=2^4\cdot3^2\cdot5\) and \(\tau(6!)=30\). Thus

\[ F(1,5)=\frac{30}{16}=\frac{15}{8}. \]

With a two-step shift, \(7!=5040=2^4\cdot3^2\cdot5\cdot7\) has \(60\) divisors, giving

\[ F(2,5)=\frac{60}{16}=\frac{15}{4}. \]

The computation is exact, using integer prime exponents and rational arithmetic. The two numbers show how an additional factorial factor can change divisor counts quite abruptly. No conclusion about asymptotic density follows from two finite values.

For fixed \(n\), the function \(t\mapsto\tau((n+t)!)\) is nondecreasing in nonnegative integers \(t\), because each new factorial is obtained by multiplying by an integer, which increases prime exponents without decreasing any. Consequently \(F(f,n)\geq1\) and is monotone as the integer shift increases. This elementary statement does not imply uniform growth as \(n\) tends to infinity: the denominator changes simultaneously.

What is known

In the original Erdős–Graham Old and New Problems and Results in Combinatorial Number Theory (1980), printed page 83, the authors consider factorial divisor ratios after adding several factors. They already knew substantial growth for shifts comparable to square roots of \(n\) and raised questions about much shorter, logarithmic shifts. The text is directly visible in the historical scan.

The main relevant later primary paper is P. Erdős, S. W. Graham, A. Ivić and C. Pomerance, On the number of divisors of \(n!\), in Analytic Number Theory, Progress in Mathematics 138 (1996), pp.337–355, publisher record. Their work on the factorial divisor function also resolves adjacent Problem #419, but the claims for #420 have to be stated separately.

The maintained Erdős Problems #420 page attributes three important results to the 1996 paper. For any fixed \(c>0\) it reports

\[ \liminf_{n\to\infty}F(c\log n,n)=1. \]

It also reports that \(F(n^{4/9},n)\) tends to infinity, and that if \(f(n)=o((\log n)^2)\) then \(F(f,n)\sim1\) for almost all integers \(n\). The qualifier “almost all” does not mean every sufficiently large integer; a sparse subsequence could behave very differently. These theorems leave the intermediate regime involving \((\log n)^C\) and the value-density questions unresolved in their stated form.

The maintained discussion also connects large jumps of the ratio to primes in short intervals. A single large prime entering the factorial can nearly double the divisor count; multiple such entries can multiply the effect. Heuristics and conditional statements involving prime gaps are informative but are not unconditional solutions of the precise density questions.

GCL's current state

The original audit issue #1690 pins the teorth catalogue at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml Git blob 4e60263a80af6207aca494d2e1388e44fa372889. This frozen entry records an open question and no formal solution as of August 2025. The maintained problem page also says OPEN on the 13 October 2026 source inquiry.

There is no Problem 420 statement file in the issue-pinned FormalConjectures provider snapshot at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, and its absence is recorded in the original issue. A newer external formal file, if one appears, would not automatically become part of that source lock. This article therefore does not claim a formal statement, Lean proof, or completed proof replay.

The authenticated source-return reference on the original issue is authenticated original RESULT/1. A source audit and its finite example are documentary evidence; accepted coverage is determined only through the Compendium's V2 admissions ledger.

What remains

The most useful small successor task is a reproducible finite program to compute factorial divisor ratios by two independent methods. For integers \(5\leq n\leq100\) and shifts \(0\leq t\leq10\), compute \(\tau((n+t)!)/\tau(n!)\) from prime exponent vectors; separately obtain the same rational by multiplying the one-step ratios from \(n\) to \(n+t-1\). Confirm exact equality for every case and check \(F(1,5)=15/8\) and \(F(2,5)=15/4\).

The second part is source-semantic: transcribe the relevant 1996 theorem and corollary hypotheses and distinguish convergence for almost all \(n\), the liminf for logarithmic shifts, and full convergence for polynomial shifts. The successor passes if all finite identities agree and the statement table matches the primary text; it fails if an integer floor is omitted, a denominator exponent is wrong, or an almost-everywhere result is silently promoted to everywhere. Stop at this small arithmetic and source-concordance task, rather than claiming that finite observations resolve the remaining density problem.

Sources and claim boundaries

Original statement: Erdős–Graham 1980, printed p.83. Related primary results: Erdős–Graham–Ivić–Pomerance, 1996. Maintained status and theorem summaries: Erdős Problem 420, accessed 2026-10-13. Original GCL source issue: #1690. Frozen teorth catalogue.

Finite factorial computations and monotonicity in the shift are independently checkable. The asymptotic results are attributed to a published external paper; the remaining conjectures are not announced as solved. No mathematical novelty, formal replay, or accepted editorial status is claimed.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1690 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: STATEMENT_CAPTURE_REQUIRED.
  • Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider 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