Wiki
Wiki

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

Updated


Source. Robert I. Saye, On two conjectures concerning the ternary digits of powers of two, J. Integer Seq. 25 (2022), Article 22.3.4, 9 pp., as identified on the source card: Lemma 1, stated on p. 3, proved in Section 5, pp. 7--8.

Read depth. Claims checked: the statement and the notation it uses (Section 2, p. 2) were read clause by clause on the print. The proof was read but not checked step by step; nothing here is independently reviewed.

Statement

Notation (p. 2). For integers a,ba,b and a positive integer kk, a≡kba\equiv_k b means a≡b(mod3k)a\equiv b\pmod{3^k}. For aa with ternary expansion a=∑i=0nai3ia=\sum_{i=0}^{n}a_i3^i, the kk-th ternary digit is dk(a)=ak−1d_k(a)=a_{k-1}, so d1(a)d_1(a) is the least significant digit.

Lemma 1 (p. 3). Let kk be a positive integer and put uk=2⋅3k−1u_k=2\cdot3^{k-1}. Then:

(i) uku_k is the least positive integer uu with 2u≡k12^{u}\equiv_k1;

(ii) for i,j∈Ni,j\in\mathbb N, if 2i≡k2j2^i\equiv_k2^j then ii and jj differ by a multiple of uku_k;

(iii) for i,j∈Ni,j\in\mathbb N,

dk+1(2iuk+j)≡dk+1(2j)+i d1(2j)(mod3).d_{k+1}\bigl(2^{iu_k+j}\bigr)\equiv d_{k+1}\bigl(2^j\bigr)+i\,d_1\bigl(2^j\bigr) \pmod 3.

The paper notes (p. 3, footnote 1) that uk=φ(3k)u_k=\varphi(3^k), so part (i) says that 22 has the full order φ(3k)\varphi(3^k) modulo 3k3^k. In part (iii) d1(2j)d_1(2^j) is 11 or 22, so as ii runs over 0,1,20,1,2 the (k+1)(k+1)st digit takes all three values while, by part (i), the last kk digits stay fixed; this is the step the paper's search uses (p. 3).

Proof pointer

Section 5 (pp. 7--8). The key fact, proved by induction on kk by cubing, is that 2uk≡1+3k(mod3k+1)2^{u_k}\equiv1+3^k\pmod{3^{k+1}}. Part (i) follows by induction on kk, ruling out uk−1u_{k-1} and 2uk−12u_{k-1} as the order modulo 3k3^k; part (ii) reduces to part (i) by cancelling a power of two, a unit modulo 3k3^k; part (iii) follows by expanding (1+3k)i(1+3^k)^i modulo 3k+13^{k+1} and multiplying by 2j2^j.

Dependencies

None outside the paper.

Bears on

  • Problem 406: only as the tool behind the computer search on the main result's page. The lemma itself says nothing about which powers of two avoid the digit 22.