First in this section a brief description of the best materials employed nowadays for thermoelectric applications will be performed, followed by traditional and new possible applications for devices made with these materials.
State of the art materials
As discussed earlier, in general the best materials for thermoelectric applications are highly doped semiconductors, even though some exceptions exist such as intermetallic compounds. A convenient way to classify the materials is according to their working temperature, i.e., the temperature corresponding to their maximum “ZT”, as described in Figure 1.14.
Low and medium temperature materials
For low temperatures (up to 450 K), the best materials employed are tellurium and bismuth compounds such as Bi2Te3. By doping this material with antimony and selenium “p” and “n” materials are obtained, respectively. For medium temperatures, between 500 and 800 K, the reference material is PbTe with partial substitution of lead by tin and of tellurium by selenium. Even though it has a lower “ZT”, it’s higher melting point allows the using of it without chemical stability problems . Doping can be obtained by a large choice of impurity atoms such as Na, Au, Ti, and O for “p” type and Zn, Cd, In, Bi, and Cl for “n” type.
It should be noticed that both tellurium and lead-based materials are highly toxic [11,12] and thus new materials should be developed for a safe usage of thermoelectric devices at low to medium temperatures.
Some other classes of materials are currently studied as being possibly well suited for low and medium temperatures having the advantage of being constituted of materials relatively abundant and presenting a low toxicity.
It can be cited for example some metals silicides such as CrSi2, FeSi2, MnSi1.7 and Mg2(Si,Sn) . The best performances were measured for the two latter. Solid solutions of Mg2Si and Mg2Sn were reported having an “n” type behavior with a “ZT” of 1.3 at around 700 K . For the “p” type, the best performances were for MnSi1.7, which presented for the same temperature range a “ZT” of 0.7.
There are other materials for the medium temperature range having interesting properties, such as skutterudites and clathrates. Both of them share a possible low thermal conductivity due mostly to their complex crystalline structure, presenting large voids that when filled with heavy atoms can establish local soft phonon modes causing a lowering of the thermal conductivity .
Skutterudites have the property of a glass-like lattice thermal conductivity because of the presence of loose atoms having more than one metastable position in the interior of the crystalline cell. They were reported having a “ZT” approaching 1 [10,17].
Similarly, clathrates have a very low thermal conductivity mostly due to the very large size of the lattice unit cell and consequently large open structure inside of it where guest atoms can be incorporated . Theoretical studies for optimized clathrates compositions showed a possible “ZT” of 1.7 at 800 K . However the high degree of complexity of these materials such as a large number of different atoms and possible stoichiometry make their synthesis and industrial application problematic.
High temperature materials: Silicon-Germanium (SiGe)
As illustrated in Figure 1.14, the last class of materials is those for high temperature applications, represented basically by Si-Ge alloys.
Figure 1.15: Thermal conductivity of SiGe as a function of Ge content .
Silicon, when doped, has a high electrical conductivity, but also a high thermal conductivity. Silicon and germanium form a solid solution with no intermetallic compounds, thus it is possible to mix the two atoms in any different concentrations. By mixing both atoms to form a solid solution a great reduction of the thermal conductivity can be observed (Figure 1.15). Moreover, for high doping levels such as for degenerate SiGe, very little changes of the carriers mobility is observed . These results imply that a relatively high figure of merit can be achieved by including Ge to Si. The state of the art values of “ZT” for conventional doped SiGe materials are around 1 for “n” type and 0.6 for “p” type materials using the composition Si0.8Ge0.2 at temperatures around 1200 K . This material has the great advantage of being non-toxic. Moreover, silicon is the second more abundant element on earth . Germanium however is much rarer and expensive. As a conclusion, further investigations should be made in order to improve the “ZT” of SiGe materials, allowing a further reduction of the Ge content. This issue is the core of this work. By including nanoparticles inside a SiGe matrix, a reduction on the thermal conductivity and an increase on the “ZT” are expected, and a possible reduction of the Ge content for the same figure of merit can be predicted.
Bulk materials-based devices
Applications for thermoelectric devices will be grouped in this document in three major working modes. The first one, the generator mode, is based fundamentally on the Seebeck effect, where a temperature gradient is employed to produce electric power in the form of an electrical current. The second major application is the cooling mode, which is basically the reverse of the generator mode and can be also thought as a heat pump. It is generally described by the Peltier effect. The last mode of operation is the sensor mode, where changing of the heat flux in a certain spatial region can be detected also based on the Seebeck effect.
Although thermoelectric generators have been used since the 1950’s, it always remained a “niche” application. The reason for the low-scale utilization of generators lays in its low efficiency (around 5 %), making it suitable for applications where the basic requirements are not the cost but rather their reliability . This reliability is guaranteed mostly because of the simplicity of the operating mode of this type of device, which contains no moving parts.
The traditional applications for thermoelectric modules are basically concentrated on the field of military and spatial applications, such as the RTG.
RTG is the most successful application of thermoelectric generators using other materials than Bi2Te3. This energy production system is considered to be the only one capable of powering long-lasting spacecraft such as those employed for interplanetary missions [23,24].
RTG working principle is based by coupling a thermoelectric module with a heat source originated from the decay of radioactive isotopes, commonly Pu-238, which has a specific energy release of 0.57 W/g and a half-life period of 87.7 years . Once again, the great advantage of this type of powering system is its high reliability (independent of solar radiation) and long life.
Examples of well-known space probes that use RTG as powering system are the probes Pioneer, Voyager, Apollo 11, Galileo and more recently New Horizons.
Typically each RTG can provide power of around hundreds of watts. The general-purpose heat source-RTG (GPHS-RTG), which until 2006 was the current RTG employed by NASA nominally generates around 250 watts at the beginning of its mission .
The GPHS-RTG is considered of main importance in the context of this work. It employs doped SiGe as thermoelectric materials in a range of temperature of around 1273 K at the hot side and 566 K at the cold side . This material will be further used as a reference material in order to evaluate the performance of the SiGe-based materials produced in this work.
Concerning potential future applications, both low and high power (microwatts for thin film applications and kilo watts for bulk materials) generators can be built, and the increase of the global concern of environmental issues can be considered a driving force for research on this field  . Others applications possibly viable from an economic point of view are those concerning heat waste. In these cases, where the heat is normally not re-used, the heat source can be considered as a “free” energy source, and cost considerations concern only the materials and the device production costs. For example, only 25% of the energy from the combustion in an automobile engine is used as mechanical work for moving the vehicle and 40% is lost in the form of hot exhaust gases .
The world market concerning cooling thermoelectric devices is considered to be ten times more developed than the generator one .
Even though the energetic efficiency of the system remains low as compared to traditional fluid compressing systems, additional advantages exist for using thermoelectric coolers rather than traditional ones. Because thermoelectric devices use no compressors they are lighter, smaller and silent-operating. They also have the advantage of a more precise temperature control .
Several applications already exist nowadays, and in opposition to generators applications that are concentrated basically in the military and spatial domains, thermoelectric coolers are employed in more varied fields.
Some examples of commercially available products are: consumer products such as car refrigerators, portable picnic coolers, heated/cooled automobile seats (Figure 1.17), laboratory equipment (cold plates, cold chambers) and others .
The most important example concerning macroscopic thermoelectric sensors in terms of applications is the thermocouples employed for temperature measurements (Figure 1.18). These systems have a big precision and are largely employed in industrial and laboratory applications. It is made basically of only one junction of two thermoelectric materials. Usually two metallic alloys are employed. There are several types of thermocouples, each one being more adapted for a specific temperature and for the ambient conditions.
The working mode is based on the Seebeck effect as demonstrated in Figure 1.3 and the reference temperature is the room temperature. The difference of temperature between the each sides of the junction produces a voltage drop that can be easily measured with a voltmeter. Calibration tables are widely available on internet with the Seebeck coefficient of both thermocouples materials with different temperatures.
Thermoelectric thin films devices
As for bulk applications, applications for thermoelectric thin films can also be divided in three major fields, i.e., generators, cooling devices and sensors. The expression “thin film thermoelectric generators” can lead to some misunderstanding and a more precise description of the type device is necessary. In the most part of the time it refers to any device where the thickness of the “p-n” thermoelectric junction is on the micrometer range. It can be produced by printing, sputtering, Molecular Beam Epitaxy (MBE) and Chemical Vapor Deposition (CVD), for example.
In this work a discussion will be presented concerning basically applications related to the microelectronics industry, which correspond to the materials and techniques employed during the production of our samples, i.e., silicon-based materials, CVD, lithography techniques and others.
Thin film thermoelectric generators
The main application of thin film thermoelectric generators related to the microelectronics industry is called “on-chip energy harvesting”. These devices could take advantage of the high energy density dissipated on the so-called hot spots of microprocessors, which could attain values up to 100-300 W/cm2. By integrating these devices into chips, the heat wasted could be directly converted and employed to power the microprocessor.
A simulation work has been reported using a superlattice structure of Bi2Te3 and Sb2Te3 as thermoelectric materials with a ZT larger than 2 (Figure 1.19). In this reference the thermoelectric module was both considered to be directly grown on the Si substrate (die) and on an integrated heat spreader (IHS) . The calculations showed that by using this material it is possible to harvest up to 30 mW of power from a heat flux of 200 W/cm2.
Figure 1.19: Example of a Bi2Te3 and Sb2Te3 thermoelectric module for “on-chip energy harvesting”. a) module integrated on the HIS. b) module integrated on the silicon die. Image from .
Some other references of integration of both BiSbTe and Si-based materials can be found in . The Si-based materials however produced a lower conversion efficiency because of its lower ZT (considered to be 0.1–0.2 by the authors in the conditions of this work, i.e., at ambient temperatures).
All of these reported works are still in the research step, and some problems related to the integration of these devices in industrial applications exist. One of the major problems is related to the difficulty of integrating the BiSbTe-based materials on the silicon die because of the lack of compatibility between the two materials.
It can be seen here an example of the importance of obtaining a higher ZT at low and medium temperatures for Si-based materials, which is the goal of this thesis. The ways of increasing this material’s ZT will be presented and discussed in section 3 – .
Thin film cooling devices
Similarly to the “on-chip energy harvesting” applications, thermoelectric modules could be integrated in microelectronic devices to serve as an “on-chip cooling” device in order to keep the devices operation in the optimal temperature range thus reducing thermal noise and current leakage . Due to size of these components, no other cooling systems are possible rather than those based on thermoelectricity. By using thermoelectric devices, it should be possible to locally cool hot spots of hundreds of micrometers of diameter on chips .
A work has been published where a thermoelectric module was integrated on a state-of-the-art electronic package . This was achieved using a Bi2Te3 superlattice-based material grown by metal-organic CVD (MOCVD) on GaAs substrates (Figure 1.20). It is interesting to notice that this was the same material studied for the “on-chip energy harvesting” devices. The authors reported a cooling of up to 15 °C at a heat flux of 1,300 W.cm-2.
Figure 1.20: on-chip cooling device integrated on a silicon chip package. Image modified from .
In the case of this device also it could also be interesting for different reasons if the thermoelectric material employed was based on Si. To cite some of those, fewer production steps could be necessary, since the thermoelectric cooler could be grown directly on the chip. Also, it would be more interesting in terms of environmental care since no toxic elements would be employed (such as Bi and Te).
Thin film thermoelectric sensors
The last field of application for thermoelectric thin film devices is the one of thermal microsensors. These devices have the advantage of being reliable, inexpensive and produced using integrated circuit technology. Even though these devices work basically measuring the changings of the potential created by a changing of temperature via the Seebeck effect, they are often employed to measure non-thermal variables, such as radiation, pressure, position, flow and chemical reactions [34–36].
This is achieved by two transduction steps. The first one is the transduction of non-thermal to thermal signals, and the second one is the transduction of thermal to electrical signals, which is accomplished through the thermoelectric device. Another advantage of this method is that the power needed for creating the electrical signal comes directly from the thermal signal, thus no external power is necessary .
Examples of already existing integrated thermoelectric microsensors include IR-radiation, vacuum, gas flow and heat flux sensors. Different materials have been proposed for these sensors, such as Bi2Te3 and Sb2Te3 films but also silicon, silicon/germanium and multi quantum wells structures (MQW). Lower performances were observed for Si-based devices, mostly because of their large thermal conductivity, but these materials are still interesting because of their technological potential .
The main problem cited (high thermal conductivity) could be avoided by reducing the material’s thermal conductivity via nanostructuration, which is the next topic of this document.
As an example, a recent work has been published by Ziouche et al. in reference  showing the fabrication of a planar infrared microsensors (µSIR) using a CMOS technology employing as thermoelectric materials SiGe-based QDSL grown by CVD. Very interesting results were obtained, and the authors observed a sensitivity improvement of around 28% due to the material nanostructuration.
Increasing the thermoelectric properties via nanostructuration
The establishment of thermoelectric materials science was accomplished in the middle of the XX century with the understanding of the figure of merit ZT and the development of functional thermoelectric devices and materials, mostly based on Bi2Te3.
Even though niche applications were developed, only incremental gains were obtained on the ZT of the employed materials, without new breakthrough discoveries that could direct the scientific research towards a higher ZT.
In the 90’s decade however new proposals were made, and a big hope of increasing the performance of thermoelectric materials was lied on the advent of nanotechnology. By nanostructuring the materials two major contributions to the increase of the ZT were thought to be possible, the first one is the increase of the thermoelectric power factor by quantum confinement effects and the second one is the decreasing of the thermal conductivity by phonon scattering.
Power factor improvement
In the year of 1993 a work was published , containing theoretical studies showing that great increases of the ZT could be obtained by the nanostructuration of materials, providing the theoretical basis and encouraging further researches on this field.
The basic phenomenon allowing the increase of the Seebeck coefficient comes from the changing of the density of states (DOS) of the material when the size is reduced from a 3-D solid to quantum wells (2-D), nanowires (1-D) or quantum dots (0-D), as represented in Figure 1.21.
A possible theory is that the changes on the material’s density of states causes a higher asymmetry around the Fermi level between the hot and cold electrons energy, resulting on an higher average carriers energy and larger number of carriers, leading to an large Seebeck coefficient and electrical conductivity. However, until nowadays some controversy exists on whether the Seebeck coefficient improvement is possible or not by nanostructuring the material, with the possible mechanisms for that not being fully understood .
Several groups attempted to validate this theory by producing Quantum Well SuperLattices (QWSL) structures of different thermoelectric materials such as Bi2Te3/Sb2Te3, PbTe/PbSexTe1–x, GaAs/AlxGa1−xAs and others . For example, for the materials PbTe/Te and PbTe/PbSe, claims were made of a measured increase of the Seebeck coefficient . However, some discussions have been made stating that actually no increasing was observed and the observed values came actually from calculation errors. Some results have also been published concerning the observations of electron filtering, but at the same time a reduction of the electronic conductivity was observed, canceling the effect over the global figure of merit of the material .
Even if some controversy exist related to the changes on the Seebeck coefficient when nanostructuration occurs, when researchers begun trying to verify these theories by creating experimental low-dimensional materials, an interesting factor was observed, the reduction of the thermal conductivity when compared to bulk materials.
This phenomenon, which was not the initial motivation for nanostructuring thermoelectric materials became the core of one of the major research field nowadays for increasing the materials figure of merit.
Thermal conductivity reduction
The basic idea of the research on the increase of the materials’ ZT is to decouple the thermal conductivity to the electrical conductivity. In order to do this, a further study on the thermal conductivity is made. On Equation 1.22 it was shown the contribution of the phonons relaxation time “ to the lattice thermal conductivity “λl”. The relaxation time depends on the collision mechanisms, which scatter the phonons responsible for the heat transport.
The relaxation time can be divided in different parts, corresponding to the different scattering sources present in a material, as described by the Matthiesen law: 1.25 Where τa is the intrinsic inharmonic contribution, τd is the solid solution contribution, τnp is the contribution from nanoparticles inside the matrix and τgb is the contribution due to grain boundaries.
By nanostructuring the material it can thus be possible to change the thermal conductivity by creating new interfaces with the matrix.
It should be noticed that the relative size of each one of these defects will cause a different interaction with different frequency phonons. Long wavelength phonons will mostly interact with grain boundaries and nanometric inclusions and short wavelength phonons will interact mostly with atomic defects such as alloying and dopant atoms (Figure 1.22).
In the case of polycrystalline materials, the grain boundary acts as a natural scattering site for phonons but also for electrons. This explains why polycrystalline materials have typically a smaller thermal and electrical conductivity than a monocrystalline solid with the same doping level and stoichiometry.
Table of contents :
CHAPTER I – THERMOELECTRICITY AND NANOSTRUCTURATION
1 – PRINCIPLES OF THERMOELECTRICITY
1.1 – HISTORY OF THERMOELECTRICITY
1.2 – THE THERMOELECTRIC EFFECTS
1.2.a – The Seebeck effect
1.2.b – The Peltier effect
1.2.c – The Thomson effect
1.2.d – The Kelvin Relationships
1.3 – PRINCIPLES OF THERMOELECTRIC CONVERTERS
1.4 – MAIN THERMOELECTRIC PARAMETERS
1.4.a – Electrical conductivity (σ)
1.4.b – The thermal conductivity (λ)
1.4.c – Seebeck Coefficient (S)
1.4.d – Ideal carriers concentration
2 – THERMOELECTRIC APPLICATIONS
2.1 – STATE OF THE ART MATERIALS
2.1.a – Low and medium temperature materials
2.1.b – High temperature materials: Silicon-Germanium (SiGe)
2.2 – BULK MATERIALS-BASED DEVICES
2.2.a – Generators
2.2.b – Cooling devices
2.2.c – Thermoelectric sensors
2.3 – THERMOELECTRIC THIN FILMS DEVICES
2.3.a – Thin film thermoelectric generators
2.3.b – Thin film cooling devices
2.3.c – Thin film thermoelectric sensors
3 – INCREASING THE THERMOELECTRIC PROPERTIES VIA NANOSTRUCTURATION
3.1 – POWER FACTOR IMPROVEMENT
3.2 – THERMAL CONDUCTIVITY REDUCTION
4 – QUANTUM WELLS AND QUANTUM DOTS SUPERLATTICES
4.1 – INTRODUCTION TO QUANTUM CONFINED STRUCTURES AND SUPERLATTICES
4.2 – GENERAL APPLICATIONS FOR QUANTUM WELLS AND QUANTUM DOTS
4.3 – THERMOELECTRIC APPLICATIONS FOR QWSL’S AND QDSL’S
5 – CONCLUSION
CHAPTER II – THE CVD GROWTH OF QUANTUM DOTS SUPERLATTICES
1 – INTRODUCTION
2 – THE CVD GROWTH
2.1 – GENERALITIES
2.2 – NUCLEATION AND GROWTH MECHANISMS
2.3 – THE GROWTH RATE LIMITING FACTOR
2.4 – OUR CVD TOOL
2.5 – SI AND SIGE THIN FILM GROWTH
2.6 – TI AND MO PRECURSORS
2.7 – DELIVERY SYSTEM FOR NON-GASEOUS PRECURSORS
2.7.a – TiCl4 evaporation system
2.7.b – MoCl5 sublimation
3 – THE GROWTH OF TI-BASED SILICIDE/SIGE QDSL
3.1 – INTRODUCTION
3.2 – CVD DEPOSITION OF TI-BASED NANO-ISLANDS
3.2.a – The role of deposition temperature
3.2.b – The role of the substrate Ge content
3.2.c – The role of deposition duration
3.2.d – The role of the precursor partial pressure
3.2.e – Role of substrate crystallinity
3.2.f – Conclusion
3.3 – EMBEDDING THE NANO-ISLANDS AND FORMATION OF QUANTUM DOTS
3.3.a – Low temperature embedding: nanowires growth
3.3.b – High temperature embedding
3.4 – TI/SIGE QDSL GROWTH
4 – THE GROWTH OF MO-BASED SILICIDE/SIGE QDSL
4.1 – INTRODUCTION
4.2 – CVD DEPOSITION OF MO-BASED NANO-ISLANDS
4.2.a – Role of deposition temperature
4.2.b – Role of Ge content
4.3 – MO/SIGE QDSL GROWTH
5 – CONCLUSION
CHAPTER III – CHARACTERIZATION OF QDSL FOR THERMOELECTRIC APPLICATIONS
1 – INTRODUCTION
2 – STRUCTURAL CHARACTERIZATION
2.1 – INTRODUCTION
2.2 – XRD
2.3 – TEM ANALYSIS
2.3.a – Ti-based QDSL: “n”-doped monocrystalline samples
2.3.b – Ti-based QDSL: “n”-doped polycrystalline samples
2.3.c – Ti-based QDSL: “p”-doped monocrystalline samples
2.3.d – Ti-based QDSL: “p”-doped polycrystalline samples
2.3.e – Mo-based QDSL: “n”-doped monocrystalline samples
2.3.f – Mo-based QDSL: “n”-doped polycrystalline samples
2.3.g – Mo-based QDSL: “p”-doped monocrystalline samples
2.3.h – Mo-based QDSL: “p”-doped polycrystalline samples
2.4 – CONCLUSION
3 – THERMOELECTRICAL CHARACTERIZATION
3.1 – INTRODUCTION
3.2 – TI-BASED QDSL: N-DOPED SAMPLES
3.2.a – Monocrystalline QDSL
3.2.b – Polycrystalline QDSL
3.3 – TI-BASED QDSL: P-DOPED SAMPLES
3.3.a – Monocrystalline QDSL
3.3.b – Polycrystalline QDSL
3.4 – DISCUSSION ABOUT THE THERMOELECTRICAL CHARACTERIZATION RESULTS OF THE TI-BASED QDSL
3.5 – MO-BASED QDSL: N-DOPED SAMPLES
3.5.a – Monocrystalline QDSL
3.5.b – Polycrystalline QDSL
3.6 – MO-BASED QDSL: P-DOPED SAMPLES
3.6.a – Monocrystalline QDSL
3.6.b – Polycrystalline QDSL
3.7 – DISCUSSION ABOUT THE THERMOELECTRICAL CHARACTERIZATION RESULTS OF THE MO-BASED QDSL
4 – CONCLUSION