The Riemann rearrangement theorem states that a convergent real series $sum_n a_n$ is absolutely convergent if and only if the value of the sum is invariant under all rearrangements $sum_n a_{p(n)}$ by any permutation $p$ on the natural numbers; furthermore, if a series is merely conditionally convergent, then one may find rearrangements for which the new sum $sum_n a_{p(n)}$ has any desired (extended) real value or which becomes non-convergent. In recent joint work with Andreas Blass, Will Brian, myself, Michael Hardy and Paul Larson, based on an exchange in reply to a Hardy’s MathOverflow question on the topic, we investigate the minimal size of a family of permutations that can be used in this manner to test an arbitrary convergent series for absolute convergence.  Specifically, we define the rearrangement number $rr$, a new cardinal characteristic of the continuum, to be the smallest cardinality of a set $P$ of permutations of the natural numbers, such that if a convergent real series $sum_n a_n$ remains convergent to the same value after any rearrangement $sum_n a_{p(n)}$ by a permutation $p$ in $P$, then it is absolutely convergent. The corresponding rearrangement number for sums, denoted rr_Sigma, is the smallest cardinality of a family $P$ of permutations, such that if a series $sum_n a_n$ is conditionally convergent, then there is some rearrangement $sum_n a_{p(n)}$, by a permutation $p$ in $P$, for which the series converges to a different value. We investigate the basic properties of these numbers, and explore their relations with other cardinal characteristics of the continuum. Our main results are that b≤ rr≤ non(M), that d≤ rr_Sigma, and that b≤ rr is relatively consistent.

MathOverflow question | JDH blog post about this talk