Erdős Problem #427
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (not automatically the live result).
Subject areas: number theory, primes.
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
List the primes in increasing order as \(p_1=2,p_2=3,p_3=5,\ldots\). Suppose someone fixes a starting position \(n\) and a positive integer modulus \(d\). Erdős Problem 427 asks whether one can always choose a nonempty consecutive block of primes beginning immediately after that position whose sum is divisible by \(d\). In symbols, is there always a \(k\geq1\) such that
The starting position is fixed; one may change the length of the block, not skip primes or rearrange them. Divisibility says that the sum has remainder zero after division by \(d\). The modulus cannot be zero, because the sum is strictly positive.
This simple-sounding question joins two different subjects. It is about the remainders of partial sums, but its proof depends on the remarkable existence of arbitrarily long runs of consecutive primes in one specified reduced residue class. The answer to Erdős's original question is affirmative, by an argument attributed to Cédric Pilatte using a theorem of D. K. L. Shiu.
A small example
Take \(n=2\) and \(d=6\). We must start at the third prime, \(p_3=5\). The first four primes after our starting position are \(5,7,11,13\). Their sum is
Hence \(k=4\) works. The calculation is exact integer arithmetic. It demonstrates what a valid witness looks like but says nothing about every starting point and modulus.
There is a useful elementary observation: if we add a prime that is congruent to \(1\) modulo \(d\), the partial sum's remainder increases by \(1\). Imagine encountering a run of \(d\) consecutive primes all congruent to \(1\) modulo \(d\). Whatever the residue of the sum immediately before that run, adding some number from \(0\) through \(d-1\) of those primes makes the result divisible by \(d\). If the sum is already divisible at the beginning, stop there. The preceding sum must be nonempty; we arrange this by choosing the run sufficiently far after the original starting point.
For example, if the preceding sum has residue \(r=4\) modulo \(6\), adding two primes each congruent to \(1\) modulo \(6\) gives residue \(4+2=0\) modulo \(6\). The important part is not that a particular small run exists, but that arbitrarily late runs of the required length exist for every modulus. That last assertion is Shiu's theorem, not a consequence of this small calculation.
What is known
The problem is recorded in P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980). It now has a positive resolution from a deeper published prime-number theorem.
D. K. L. Shiu proved Strings of Congruent Primes, Journal of the London Mathematical Society 61(2) (2000), pp. 359–373. Shiu showed that for any modulus \(q\), any residue class \(a\) coprime to \(q\), and any fixed block length \(\ell\geq1\), there are infinitely many blocks of \(\ell\) consecutive primes all congruent to \(a\) modulo \(q\). The publisher's abstract expressly states the existence of arbitrarily long strings. This is stronger than finding many isolated primes in the same residue class: no primes of other classes may intervene inside the selected block.
The maintained Erdős Problems #427 entry, last edited January 22, 2026, credits Cédric Pilatte with the deduction. Apply Shiu with \(q=d\), \(a=1\) and \(\ell=d\). Choose a run
of consecutive primes, each congruent to \(1\) modulo \(d\), with \(m>n+1\). This is possible because Shiu's result supplies infinitely many such runs.
Let \(S=p_{n+1}+\cdots+p_{m-1}\). Choose \(r\) in the range \(1\leq r\leq d\) with \(S\equiv r\pmod d\), taking \(r=d\) when the remainder is zero. Put \(t=d-r\), so \(0\leq t<d\). The sum of the original prefix followed by precisely \(t\) primes from the selected run then satisfies
If \(t=0\), the added sum is empty, but the overall original prefix still has positive length since \(m>n+1\). Otherwise the final prime used is \(p_{m+t-1}\). This small residue calculation proves the target statement conditional on Shiu's published theorem. No assumption about the distribution of arbitrary prime gaps is additionally needed.
GCL's current state
The original source assignment #1697 pins the teorth catalogue at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file data/problems.yaml, Git blob 4e60263a80af6207aca494d2e1388e44fa372889. The frozen entry says proved. The maintained mathematical page says PROVED (LEAN), consistent with a solved mathematical problem and an externally reported Lean verification.
The pinned FormalConjectures 427.lean, Git blob 8473d6d80391c695667f4497953f44cf0f3b1d77, formalizes a positive-length prime-prefix divisibility property, an explicit Shiu theorem and an implication from that theorem. Its prime indices are zero-based inside a finite interval, so careful translation is needed to match the one-based mathematical notation. The pinned file's theorem bodies are by sorry despite an external formal-proof link. This article does not claim the external Lean development was compiled, axiom-audited or mathematically certified by GCL.
The authenticated source return is original issue #1697, RESULT/1 comment #6157367255. That comment establishes documentary source custody only. The V2 editorial ledger separately governs accepted article coverage.
What remains
For mathematical exposition, the strongest useful bounded task is to formalize the deduction from Shiu, treating Shiu's result as an explicit premise, and separately document the premise's original journal source. Check the modulus \(d=1\), the exclusion of \(d=0\), the positivity of the stopping length, and the final index when the number \(t\) of primes added from the Shiu run is zero.
An executable fixture can calculate prime prefixes for all \(0\leq n\leq20\) and \(1\leq d\leq20\), stopping at the first divisible nonempty sum and recording \(k\) and the exact quotient. Such a bounded search is not a proof of the universal statement; it only checks the indexing and arithmetic. Acceptance requires correct witnesses plus a source-to-formal-quantifier comparison and a proof of the elementary conditional residue lemma. Falsification includes an off-by-one index, an inadmissible empty overall sum, or treating a pinned placeholder as a compiled theorem. Stop after that interface verification rather than claiming a new prime-number theorem.
Sources and claim boundaries
Historical statement: Erdős–Graham (1980). Primary theorem: Shiu (2000), DOI 10.1112/S0024610799007863. Maintained deduction and status: Problem #427. Source custody: GCL issue #1697, teorth catalogue, pinned formal statement, and authenticated source audit.
The residue-class implication and displayed sample are independently checked. Shiu's deep theorem and any external Lean verification are credited to their sources, not newly certified by GCL. No original research result, protected publication or V2 editorial admission is asserted.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1697 — 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.