稳定连续模式与链簇

A consecutive permutation pattern is stable if its occurrence distribution on words is unchanged when the multiplicities of the letters are permuted. We prove that the stable patterns are precisely the chain patterns of Elizalde and Noy: every connected system of overlapping occurrences must impose a total order on its positions. For a pattern of length m, it suffices to count pairs of occurrences in words on multisets of size at most 2m-1 with one doubled letter and all other letters single. For m ≥ 3, this length bound is sharp. A finite two-occurrence overlap criterion gives an equivalent structural test. Stable patterns also have invariant occurrence-position distributions; their marked-position enumerators are nonnegative combinations of products of elementary symmetric functions. The characterization supplies nonmonotone stable patterns of every length at least five, disproving a conjecture of Chen, Fang and Kitaev. We give weighted generating functions and the classification through length five. Corresponding criteria treat joint distributions for families beginning with their smallest letter. Two individually stable patterns of length five furnish an explicit failure of joint stability.