Wiki
Wiki

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

Updated


Statement

Setting (p. 292). Fix an integer k≥2k\ge2. For positive integers M≤NM\le N, Sk(M,N)S_k(M,N) is the family of sets A⊂[0,M]A\subset[0,M] such that every positive integer n≤Nn\le N is a+bka+b^k with a∈Aa\in A and bb a positive integer, and fk(M,N)=min⁡{∣A∣:A∈Sk(M,N)}f_k(M,N)=\min\{\lvert A\rvert:A\in S_k(M,N)\}. Mk=Mk(N)M_k=M_k(N) is the smallest integer MM for which Sk(M,N)S_k(M,N) is non-empty.

Proposition (p. 292). With B=[N1/k]B=[N^{1/k}], the integer part of N1/kN^{1/k},

Bk−(B−1)k−1≤Mk≤(B+1)k−Bk−1.B^k-(B-1)^k-1\le M_k\le(B+1)^k-B^k-1.

Before the statement the paper gives, as examples, that for an integer S≥4kS\ge4k one has Mk(Sk−1)=Sk−(S−1)k−1M_k(S^k-1)=S^k-(S-1)^k-1 and Mk(Sk+1)=(S+1)k−Sk−1M_k(S^k+1)=(S+1)^k-S^k-1; in both examples MkM_k equals the upper bound (with B=S−1B=S-1 and B=SB=S respectively). The paper does not prove these examples.

Source. Wenguang Zhai, The additive completion of kkth powers, J. Number Theory 79 (1999), 292--300, doi:10.1006/jnth.1999.2441: the setting and the Proposition on p. 292, the proof in Section 2 on p. 294. The edition read is identified on the source card.

Read depth. Claims checked: the setting and the statement were read clause by clause on the printed pages. The short proof was read. Nothing here is independently reviewed.

Proof pointer

Section 2, p. 294. For the upper bound, the interval of integers {0,1,…,(B+1)k−Bk−1}\{0,1,\ldots,(B+1)^k-B^k-1\} completes the kkth powers up to NN, since each n≤Nn\le N lies between consecutive kkth powers bk≤n≤(b+1)kb^k\le n\le(b+1)^k. For the lower bound, the integer n=Bk−1n=B^k-1 can only use some b≤B−1b\le B-1, which forces an element a≥Bk−(B−1)k−1a\ge B^k-(B-1)^k-1.

Bears on

  • Problem 33: at k=2k=2 the Proposition gives 2B−2≤M2(N)≤2B2B-2\le M_2(N)\le2B with B=[N1/2]B=[N^{1/2}]: a finite set completing the squares b2b^2, b≥1b\ge1, up to NN has an element at least 2B−22B-2, and some such set lies in [0,2B][0,2B]. It concerns finite completions only and decides neither question of the problem.