We present quantum mechanical (QM) studies of carbon nanotube (CNT) fracture using two different semiempirical methods. One proposed mechanism for CNT fracture - based mainly on studies with empirical potentials - involves an aggregation of Stone-Wales defects followed by a ring-opening step whereby a bond between two 5-membered rings is severed. We have performed QM studies which instead predict that this bond is a particularly strong one, and that the failing bonds lie within the pentagons. We also explore why empirical bond-order potentials (in particular, a potential of Brenner and coworkers) predict qualitatively different fracture mechanisms than quantum mechanical calculations do. (C) 2003 Elsevier B.V. All rights reserved.