and strong localization) are each affected by disorder in a
similar way. For boundary roughness the influence is weak near the band edge,
and increases as the energy increases. However, the influence of the island
disorder is strongest on the highest propagating modes and does not depend on
the mode (subband) number. Since the total conductance increases with the
number of propagating modes, then, in general, the average conductance
increases with the energy. But note that the average conductance, as a function
of energy, always drops when a new subband opens due to the enhanced
intersubband scattering. This was not observed for the case of boundary
roughness only. For the quasi-ballistic regime the conductance quantization
deteriorates very rapidly as the number of scattering events increases. The
average conductance decays exponentially as a function of wire length (for
), for any kind of disorder, which is considered as an additional
confirmation of the exponential localization of electron states. Anderson
localization is very effective in reducing the carrier mobility in narrow
quantum wires. This effect acts strongly against the predicted high mobility
for quantum wires (Sakaki 1980).
The conductance fluctuations of narrow quantum wires depend, in general, on the
length of the wire, and are therefore not universal conductance fluctuations
(UCF). However, the case with rough edges can show a universal region, but only
for energies in the first subband. The value for 
is not far from the
UCF value for metallic quasi-1D systems (
) and depends on the
energy. An increase in the localization length extends this region of
constant fluctuations. Ando and Tamura (1992) have predicted that for
wider wires than we have, a much broader region of universal conductance
fluctuations will appear. We find, on the contrary, that for island disorder
short universal regions can exist only for higher energies and the actual value
for approaches the UCF as the energy increases. The universal
region, if it exists, appears for wires of length , and the
fluctuations reach a maximum in this region.
Changing the cross-section of the leads makes no qualitative difference to
the conductance of a system consisting of a disordered wire attached to two
perfect leads (Nikolic 1983). Some small quantitative changes are observed
only when 
(which will be reported elsewhere). In any case the
conductance becomes independent of 
when 
. This should be expected,
since for large values of , transverse modes in the leads are densely
distributed, and there are many of them contributing to the total conductance
(Szafer and Stone 1989).