In this Thesis we present novel, robust and general algorithms for combining planewave
density-functional theory with the Landauer-Buttiker transport formalism.
The method automates this process with minimal user input to allow a high throughput
of calculations. We make use of a maximally-localised Wannier function basis
to describe systems using short-ranged Hamiltonians. Further, these Hamiltonians
may be used as "building-blocks" to create model Hamiltonians of much larger
(10,000+ atom) systems, thus allowing electronic transport properties of structurally
complex systems to be determined with first-principles accuracy. A similar building-block
method is applied to construct model dynamical matrices from those of smaller
systems, from which the lattice thermal conductivity Kl may be inferred.
The methods were applied to investigate the thermoelectric properties of (110), (111)
and (211) Si nanowires (SiNWs) that contain axial heterostructures of Ge. Their
performance is measured by the figure of merit, [equation included here], where S, G , Ke and T
are the Seebeck co-efficient, electronic conductance, electronic contribution to the
thermal conductance and average temperature between the sample's contacts, respectively.
We find the thermoelectric power factor S2G is reduced by the presence
of heterostructures, however, as a result of the differences between phonon density
of states in the Si and Ge regions, low Kl values (< 0.1 nWK-1) are reported. Thus greater values of zT are found compared to the pristine SiNW case. Of the growth
directions studied, the (111) direction is found to display the greatest values of zT,
with values as large as three in systems with periodic arrangements of heterostructures.
More modest values of 1.6 are found in structures that model disorder in the
heterostructure length, which may occur experimentally; this is still a factor of four
greater than the pristine case. In addition, we observe that trends in S2G, KI and
zT that are predicted for systems containing a single heterostructure can often be
used to predict trends in systems with many heterostructures.