Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Theorem 3.2 and its proof, preprint pp. 5--6; Proposition 3.1 and Lemma 3.1, pp. 4--5. Read on the rendered pages.

Statement

Theorem 3.2 (p. 5): "Let {an}n=1∞\{a_n\}_{n=1}^{\infty} and {bn}n=1∞\{b_n\}_{n=1}^{\infty} be two sequences of integers such that an∤ bna_n\not|\,b_n for every nn. Suppose

lim inf⁡n→∞∣bn∣an=0andlim⁡n→∞bnan−1an=0.\liminf_{n\to\infty}\frac{|b_n|}{a_n}=0\qquad\text{and}\qquad \lim_{n\to\infty}\frac{b_n}{a_{n-1}a_n}=0 .

Then S=∑n=1∞bna1…anS=\sum_{n=1}^{\infty}\frac{b_n}{a_1\ldots a_n} is irrational."

The standing convention (p. 3) applies: the ana_n are integers greater than 11. The paper calls a series with an∤bna_n\nmid b_n for every nn non-degenerate (p. 3); then every bnb_n is nonzero. The introduction (p. 2) states the theorem in that language: a non-degenerate series is irrational if lim inf⁡∣bn∣/an=0\liminf|b_n|/a_n=0 and bn=o(anan−1)b_n=o(a_na_{n-1}). The paper presents it as a variant of Corollary 3.2 (p. 5), Oppenheim's criterion, which assumes ∣bn∣<an|b_n|<a_n for every nn and lim inf⁡(∣bn∣+1)/an=0\liminf(|b_n|+1)/a_n=0.

Proof sketch (pp. 5--6)

The proof rests on Proposition 3.1 (p. 4): for a non-degenerate series, lim inf⁡N→∞∣SN∣=0\liminf_{N\to\infty}|S_N|=0 already forces SS to be irrational. That proposition combines Lemma 2.1 with Lemma 3.1, which says that a block ∑n=MN−1bn/(aM⋯an)\sum_{n=M}^{N-1}b_n/(a_M\cdots a_n) cannot vanish when $a_{N-1}\nmid b_{N-1}$: a rational SS would make qSNqS_N integers tending to 00 along a subsequence, hence two indices M<NM<N with SM=SN=0S_M=S_N=0 and a vanishing block. Theorem 3.2 supplies the small tails: at an index NN where ∣bN∣/aN|b_N|/a_N is small and every later ∣bn∣/(an−1an)|b_n|/(a_{n-1}a_n) is small, the tail SNS_N is bounded by its first term plus a geometric series, so it is small.

Uses

The proof of Theorem 5.1 (p. 9) appeals to Theorem 3.2 in the case where the tails decrease at more than half the indices of a dyadic range, after showing that ana_n then grows fast.

Bears on. No catalog problem directly; a tool in the proof of Theorem 5.1.