Ref[1] points out that certain nanowires can carry enormous current densities (~ 10^11[A/cm^2]) which vaporize macro-sized wires.
In metals, this equates to ballistic electron speeds of ~ 100 km/sec - approximately the same as (0-Amp) random thermal electron velocity - far greater than a diffusive electron current drift velocity ~ 1 mm/sec - far less than relativistic speeds. When the wire diameter approaches 1 nm, nearly ballistic electon speeds are possible over lengths of several microns. In some nanowire and e-m field distributions, electrons attain inductive (not kinetic!) energies > 1 MeV. Collisions with protons or nuclei can overcome the potential barrier (0.78 MeV) allowing neutron formation. Unless large (AC or DC) current flows are induced, conduction electrons will not acquire significant inductive energy - i.e., they will not acquire large "effective mass" - a term commonly misunderstood as relativistic mass. Here "effective mass" is a not a scalar, but a vector quantity measuring electron coupling to the inductive energy of the total current. It is large in direction of large current flow, while small normal to it. This my attempt at a semi-classical check on Widom-Larsen theory. It looks quite reasonable to me, but I could be mistaken. I would appreciate corrections or criticisms. Thanks, Lou Pagnucco [1] "Stability of Metal Nanowires at Ultrahigh Current Densities" http://arxiv.org/abs/cond-mat/0411058