Peak effect in polycrystalline vortex matter.

We report a study of the peak-effect phase diagram of a strongly disordered type-II superconductor V-21 at. %Ti using ac magnetic susceptibility and small-angle neutron scattering (SANS). In this system, the peak effect appears only at fields higher than 3.4 T. The sample is characterized by strong atomic disorder. Vortex states with field-cooled thermal histories show that both deep in the mixed state, as well as close to the peak effect, there exist no long-range orientationally ordered vortex lattices. The SANS scattering radial widths reveal vortex states ordered in the sub-mum scale. We conjecture that the peak effect in this system is a remnant of the Bragg glass disordering transition, but occurs on submicron length scales due to the presence of strong atomic disorder on larger length scales.

Out-of-equilibrium nucleation in the solidification of helium

We observed nucleation in a first-order phase transition in which a nonequilibrium (metastable) solid phase nucleates and grows from overpressured superfluid 4He in preference to the equilibrium solid phase. After a time varying from a fraction of a second to tens of minutes, the stable phase nucleates independently from the liquid, and the nonequilibrium solid phase melts. We examined the possibility of accounting for these events in terms of the known differences in the interfacial free energies between the superfluid and the two solid phases (stable and metastable). The experimental results are not consistent with this explanation, however, and we discuss other possibilities.

A neural model for category learning.

We present a general neural model for supervised learning of pattern categories which can resolve pattern classes separated by nonlinear, essentially arbitrary boundaries. The concept of a pattern class develops from storing in memory a limited number of class elements (prototypes). Associated with each prototype is a modifiable scalar weighting factor (lambda) which effectively defines the threshold for categorization of an input with the class of the given prototype. Learning involves (1) commitment of prototypes to memory and (2) adjustment of the various lambda factors to eliminate classification errors. In tests, the model ably defined classification boundaries that largely separated complicated pattern regions. We discuss the role which divisive inhibition might play in a possible implementation of the model by a network of neurons.