Maximum zero strings of Bell numbers module primes (Q1065845)

The nth Bell number \(B_ n\) can be defined as the number of ways of partitioning a set of n objects into subsets. The author gives an elegant proof that for every prime p there is a run of p-1 consecutive \(B_ n's\) that are all congruent to zero mod p. The method determines explicitly the starting point of one such run for each p. (Since \(B_ 0=1\) and \(B_ n\) is known to satisfy a linear recurrence of degree p (mod p), no longer run of consecutive zeros is possible.) The same result is obtained by a similar method for the sequence \(A_ n\), where \(A_ n\) is the difference between the number of ways of partitioning a set of n objects into an even number of subsets and the number of ways of partitioning it into an odd number of subsets. Until now such results have been obtained only on the assumption (supported by numerical and heuristic evidence but so far unproved) that the known period \((p^ p-1)/(p-1)\) for \(B_ n\) is minimal. For \(B_ n\) there is just one run of p-1 consecutive zeros per minimal period mod p and for \(A_ n\) there are just two such runs (except for \(p=2)\) separated by a half-period. \(\{\) Reviewer's remark: It is also true that the minimal period of \(A_ n\) (mod p) is twice the length of the minimal period of \(B_ n\) for \(p\neq 2.\}\)