Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Fix and . Let be a sufficiently large integer and let
Let be the centered representatives of the inverses of modulo . There are constants and depending only on , a real with and integer , and:
- a retained set , ;
- a witness , ;
- a symmetric progression , where and the are positive integers;
such that , a translate of is proper and lies in , and, writing ,
For real dilates, the coordinate bounds are rounded down. In particular, when , (1) gives for sufficiently large . The retained set and the witness have distinct roles. The external theorem supplies , though the argument below only needs the weaker consequence .
Source: published PDF, pp. 7–8. This expands and repairs the positive-input, symmetrization, properness, and volume steps needed to use the exact external CFP interface.
Bears on. Problem 297.
Proof
For large , , so the elements of have distinct residues. Inversion preserves distinctness. No member of is zero. At least half of is positive or at least half is negative. Reflect the latter half if necessary, obtaining a set of size , and a sign .
Apply Theorem 3 with and . Indeed for large . Let be its constants and choose
Then , , and eventually. This real value of is permitted in the external statement, and makes its dilation an integer. The retained set loses at most , so has size at least . Reflect the output back by . It gives , and a progression with such that a translate of lies in and is proper.
Write the coordinate map of as an affine integer map on an integer box, and choose the coordinate preimage of 0. Recenter at that preimage. There are nonnegative integers such that
Here the affine constant has vanished exactly. Delete any coordinate of zero width. Put , giving . At least one coordinate remains, since contains a unit and is nonempty. Because is an integer, this recentering represents the same -fold sum progression on the box .
Set and . Then
since . Thus the image of the smaller box, which is a translate of , lies in . It is proper because the enclosing coordinate map is injective. The unshifted coordinate box of this smaller progression contains the box defining , since , so itself is proper.
The reduced progression has points: for . All subset sums of lie between and , so their number is at most . Comparison proves (1). As and ranges over the fixed finite set , its right side is less than eventually for every such .