Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Retain the odd dimension and cyclic matrix from [[additive_combinatorics/adamczewski_2026_erdos1/corollary_2_2|Corollary 2.2]].
Statement
For a real and integers , the bound forces
Proof
Abbreviate the integer coordinates as (). For an index at which is largest, the inequality yields
so every coordinate lies in and every lies in .
The real coordinate also satisfies . Suppose ; then forces , while forces . Define the integer vector
At index , ; at index , ; and at all remaining indices the original inequalities apply. This contradicts [[additive_combinatorics/adamczewski_2026_erdos1/lemma_2_1|Lemma 2.1]], since . So , and since satisfies the same hypotheses, also .
For each , checking the nine possible pairs under
shows that or . Once a zero occurs, every later term is zero. Thus the nonzero terms of the tail form an alternating initial segment , whose consecutive pairs cancel, so the tail sums to when is even and to when is odd.
If the tail sums to , the total is , of absolute value below . In the other case the tail sum is . If , then gives , which together with gives and hence . If , it gives , hence and . The case belongs to the first alternative. Therefore the total coordinate sum always has absolute value less than .
Source and dependencies
An explanation of the proof of Erdős Problem 1, preliminary exposition with no named author (erdosproblems.com, 2026), §2, Lemma 2.3, pp. 2–3. The edition read is named on the source card. The final two sign cases make explicit the source's last one-line estimate.
Bears on. #1.