Mesoporous silica materials with linear and interconnected pore systems and tuneable pore diameters were synthesized for their use as template material for the nanocasting of thermoelectric materials, which allows us to synthesize nanowire network structures with a controllable diameter and interconnectivity. It was shown that the template’s pore diameter could be easily enlarged by ageing them at a higher temperature or using hydrophobic swelling agents. Mesoporous silica with interconnected pores and a high pore volume, such as KIT-6 and FDU-12, were the most promising template materials for the synthesis of electrically conductive nanowire composite materials as they create more percolation paths than templates with linear pores, such as SBA-15. The surface of the silica template could be chemically modified to alter the interactions with the host molecules. Next, the mesoporous silica templates were used to synthesize SrTiO3 and Bi1-xSbx nanocomposites.
The A sol-gel precursor solution of Sr(NO3)2 and Ti(OH)4 complexated with chelating agents
was impregnated in a mesoporous silica template. Since the charge density of both metal ions
is very different, using the same complexing agent for both metal species favoured the impregnation of the Ti complex in the silica template. It was found that the use of complexing agents with more carboxylic acid functionalities for the Ti precursor slowed down its impregnation. Depending on the pH, impregnation time and the combination of complexing agents, Ti and Sr were impregnated at a similar rate into the template. However, the pore loading remained low, while the reproducibility to obtain a Ti/Sr 1:1 ratio inside the template was insufficient. A higher pore loading and better control of the impregnation kinetics of a mixed precursor solution could be achieved when the double salt Sr(OXA)2Ti.2H2O was formed in the template’s pore system. Herefore, Sr(NO3)2 was impregnated in the pore channels, treated with oxalic acid to form Sr oxalate and subsequently reacted with TiCl4 to form the double salt Sr(OXA)2Ti.2H2O. However, we did not manage to completely avoid leaching of the double salt out of the pores during the synthesis. As a consequence, no SrTiO3 nanowires were
obtained in this way. Nevertheless, the impregnation of a mixed precursor solution of SrCl2 and TiOCl2 in diluted HCl may avoid the problem of leaching. Subsequently, a heat treated in an ammonia atmosphere at 400 °C may be used to remove all chlorides, followed by a thermal
treatment under reducing conditions to form metallic oxygen deficient SrTiO3

A new nanocasting method to impregnate an aqueous bismuth precursor solution into the pore
channels of a mesoporous silica template was developed. The procedure consisted of two
mechanisms to impregnate the pores and avoid deposition of material outside the template
instead of inside. Namely, the two-solvent method and incipient method. Firstly, when the silica
template was dispersed in a non-polar solvent, such as octane, and an aqueous precursor was
added, due to their immiscibility, the precursor migrated into the pore channels. Secondly, by
adding less precursor solution than the pore volume of the template, it was assumed that all
precursor went inside the pores instead of being deposited on the template’s exterior surface.
Consequently, to fill the pores with metal salt instead of the precursor solution, the non-polar
solvent was refluxed during the impregnation while the water was removed from the system
using a Dean Stark separator. The method enabled us to synthesize mesoporous metal oxides
and metalloids, such as TiO2, Bi1-xSbx, Ni, TiS2, etc.
An unusually high pore loading of a Bi salt precursor enabled us to make bismuth nanowire
composite materials. The samples were reduced at 230 °C in a flow of formic acid or hydrazine
hydrate vapour and subsequently sintered using a SPS device. Evidence of size quantization
in the bismuth nanowire composites was obtained from measurements of the Hall effect and
resistivity of the sintered samples. An activated behaviour was observed, following a law n(T)
= n0 e-Eg/2kBT valid for intrinsic semiconductors, with Eg the band gap of the semiconductor. A
value of Eg = 45 meV corresponded very well to the theoretical value of the band gap opened
by size-quantization effects in Bi nanowires of about 20 nm diameter. Furthermore, the chargecarrier concentration measured on the nanowire sample was an order of magnitude smaller
than that of bulk Bi, which provides additional confirmation of size quantized nanostructures.
Bi1-xSbx nanowire composites, on the other hand, were synthesized by impregnating a mixed
precursor solution of BiCl3 and SbCl3, followed by a reduction at 230 °C and sintering at
245 °C using SPS. We discovered that the short-circuits within the material could be avoided
by synthesizing a nanocomposite embedded in a matrix with a resistivity higher than the
nanowires, which lead to the enhancement of the Seebeck coefficient compared to bulk Bi1-xSbx alloys due to size quantization.
Although we were able to prove that the Seebeck coefficient could be improved in a
nanocomposite material, we believe that the further enhancement of the TE properties could
be achieved. Firstly, the reduction time should be shortened to limit leaching. This could be
achieved either by using a more reactive reducing agent, such as anhydrous hydrazine rather
than hydrazine hydrate. Ideally, a vertical split tube furnace should be used to enhance the
contact with the hydrazine vapour or any other reducing agent. Also, the temperature of the
hydrazine hydrate bubbler could be increased to increase the reducing agent’s vapour
pressure, which might speed up the reduction. Secondly, we expect that the electrical resistivity
could be further reduced by enhancing the pore loading through the reaction of the
impregnated precursor with trimethylbismuth (Eq.1), as demonstrated in this research for
trimethylantimony. By impregnating the pores with 50 v% Bi0.88Sb0.12 and adding 10 v%
Bi0.76Sb0.24 matrix, for instance, a treatment of the composite powder with Me3Bi, would cause
a complete filling of the pores with Bi0.94Sb0.06 while the matrix would have a composition of
Bi0.88Sb0.12. Since no voids would be present anymore and only a short reduction procedure
would be required, it is expected that the resistivity would drop, while the amount of nanowires
contributing to the TE efficiency enhancement would increase. Also, doping of the nanowires
should be considered to control the carrier concentration. Since the latter is very dependent
on the wire diameter and the temperature, the carrier density should be adjusted to 1018 cm-1 ,
which should lead to the highest thermopower of the nanowires. Namely, it is important thatthe resistivity of the nanowires remains lower than the matrix at all temperatures. Only under
those circumstances short-circuits through the bulk matrix would be avoided.
BiCl MeMgBr BiMe MgBrCl THF
3 +3 ® 3 +3 Eq. 1
Thirdly, something which was not studied in this work was the morphology of the template.
Namely, it is desirable to use spherical particles as they pack better than elongated or irregular
particles, which is the case for respectively SBA-15 and KIT-6 mesoporous silica. This may
also affect the electrical conductivity of the composite. It has been reported that the morphology
of mesoporous silica materials can be tuned by the use ethanol as co-solvent and CTAB as
co-surfactant or by changing the stirring conditions. [3-6] Moreover, any effort to enhance the
pore volume and pore diameter of the template would also benefit the electrical conductivity.
Specifically, localization of the charge carriers in Bi1-xSbx nanowires must be avoided.
Therefore, mesoporous silica with large pores (10-30 nm) is recommended. While FDU-12 is
most suitable to synthesize a template material with large pores, its pore volume is generally
lower than for KIT-6. Typically, a maximum porosity of 70 % can be obtained with FDU-12,
whereas KIT-6 with a porosity up to 82 % and an average pore diameter of 13 nm was
prepared. According to the classic effective medium theory for a binary thermoelectric system,
the figure of merit of a composite cannot exceed the zT of any of its constituents. [7] In the
most basic approximation, the volume fraction and their individual zT value will affect the
macroscopic zT of the composite. It can be understood from this simplified situation that the
occurrence of non-thermoelectric material in a composite will only reduce the overall
performances. In the case studied here, the silica causes parasitic heat losses through the
structure while it does not contribute to the thermoelectric effect. For this reason it is desirable
to use mesoporous silica with the highest possible porosity.
In case KIT-6 with a pore volume of 1.8 mL/g or a porosity of 82 % is used as template material
and the maximum pore loading of Bi of 60 v% is achieved, it was calculated that the composite
consisted of 41 v% nanowires. This the highest amount that can be achieved for a fixed amount
of 17 v% matrix in the composite. Only in case the pores would be completely filled by means
of the gas phase impregnation of Me3Bi, a value as high as 69 v% may be achieved.
The thermopower enhancement through the spin Seebeck effect was demonstrated for the
first time in a bulk nanocomposite sample of Ni-Pt. It is expected that the effect of nanostructuring could be further exploited by making smaller nanoparticles and higher surface areas of the magnetic material, for example depositing Pt nanoparticles onto mesoporous Ni to enhance the interaction area. Alternatively, individual nanowires of Ni and Pt could be mixed and compressed yielding a composite with a very high interaction surface.