The V(T,I,H) characteristics have been measured for various multi-prob
e aluminium nanostructures. For several samples (pairs of voltage prob
es) the resistance at the top of the R(T) dependence can increase the
corresponding normal state value. The effect is not fully reproducible
, being dependent upon the cooling history. The magnitude of the resis
tance 'bump' can be reduced by a strong measuring current and/or by an
external magnetic field. Two models have been proposed to explain the
observed phenomena. Both models involve the existence of the N/S inte
rface within the resistive transition. The inevitably formed N/S bound
ary potential step gives rise to additional resistance at the interfac
e. For a static model, the formation of the N/S boundary is purely det
ermined by the distribution of imperfections in the sample, while for
a dynamic one, the N/S interface is periodically established with the
average frequency of the thermal fluctuations Gamma and relaxes during
the quasiparticle relaxation time tau(Q). Both models could give the
'order of magnitude' agreement with the experiment, while the static
one is simpler. The bad reproducibility of the experimental results (d
ependence upon the cooling history) is attributed to the dramatic sens
itivity of the mesoscopic-size samples transport properties to the dis
placement of a few impurity atoms. (C) 1997 Academic Press Limited.