Write and let be the given finite coloring of the edges of the complete graph on the positive integers. Construct an increasing sequence and nested infinite sets as follows. Set . Having chosen , take , partition by the color of , and use the infinite pigeonhole principle to choose an infinite set on which this color is constant, say . Every element of exceeds .
The infinite pigeonhole principle applied again to gives an infinite set of indices and a color such that for all . For in , nesting gives , so . Thus the infinite set is a monochromatic complete graph:This proves the required infinite case of Ramsey's theorem; finiteness of the color set is used at both applications of the infinite pigeonhole principle.
Pull the given finite coloring back to a finite coloring of unordered pairs by puttingThe ordering makes this a well-defined function on the edges of a complete graph. By Ramsey's theorem, some infinite set has all its pairs of one color under . ConsequentlyPossible coincidences between these products cause no difficulty, since the original finite coloring assigns each positive integer a single color.
Use the cyclic logarithmic coloringHere the floor function places in the unique half-open real interval . If , thenbecause . For any real and , is either or . Applying this to the logarithms of and shows that their bin indices differ by or , and hence have different residues in modular arithmetic modulo . This proves the required separation, including both endpoints of the prescribed ratio interval. The half-open bins remove any ambiguity at their boundaries.
Let be the cyclic logarithmic coloring from part (iii). Define a finite coloring of the positive integers byThe inner logarithm is at least whenever , so this is defined everywhere. Suppose an increasing infinite sequence made every product with distinct indices monochromatic. Fix an index with and write . The sequence is unbounded, so choose with . Both ordered products are among the supposedly monochromatic values, but their logarithms satisfyIndeed, the first inequality is equivalent to , and the second uses . Part (iii) therefore gives , a contradiction. Thus the assertion with all distinct ordered indices is false. The obstruction needs an infinite unbounded sequence; it does not assert that every pair of different positive integers gives different colors.
Articles by others on the same topic
There are currently no matching articles.