The background here is a puzzle about reordering numbers. A "permutation" of a set of numbers is just a way of listing them in some order. The puzzle asks: can you rearrange the positive integers so that no three numbers in their new positions form an evenly spaced increasing or decreasing sequence? It turns out you cannot do this with all the positive integers at once. Researchers showed long ago that you can split the integers into three groups, each of which can be safely reordered this way. The open question, posed by Erdos and Graham, is whether two groups are enough.
A natural candidate for a two-way split uses powers of 2 as dividing points. One of the two sets, called S_A, collects numbers from certain intervals between consecutive powers of 2, specifically the intervals sitting above even powers of 2. The paper investigates whether S_A can be admissibly reordered. The authors develop several theoretical tools to understand what any valid reordering would have to look like, including constraints on how large chunks of numbers can be arranged relative to each other. These tools rule out certain organizational strategies that might seem promising.
The main result is that S_A cannot be admissibly reordered, so this particular natural split does not solve the two-set problem. The authors reduce the question to a concrete finite puzzle about ordering numbers in a specific interval, show that the puzzle has no solution when the interval size falls into a particular pattern (multiples of 8), and verify that all the relevant cases fit this pattern. The proof uses a chain of logical deductions about arithmetic progressions, and the authors also confirmed the result independently using computer-based logic solvers. The broader two-set question remains open.