PREPARATION AND PROPERTIES OF SOME COMPLEX OXIDES FOR
THERMOELECTRIC ENERGY CONVERSION
By
Chang Liu

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Materials Science and Engineering
2012

ABSTRACT
PREPARATION AND PROPERTIES OF SOME COMPLEX OXIDES FOR
THERMOELECTRIC ENERGY CONVERSION
By
Chang Liu
During the past decades, thermoelectric (TE) materials have received renewed attention for potential applications in power generation, waste heat harvesting and solid-state
cooling/heating. With the introduction of new material systems and advanced preparation
techniques, the performance of state-of-the-art candidate materials has been greatly improved. Oxide-based materials, as newcomers to this field, have attracted growing interests
due to their low production cost, high thermal stability and low toxicity.
In the present work, the author reports upon the following three oxides with complex
crystal structures, with emphasis on the synthesis techniques and transport property study.
In delafossite-type copper aluminum oxide (CuAlO2 ), magnesium was used as the dopant
to substitute aluminum atoms to improve the compound’s high temperature performance.
Powder processing and advanced sintering techniques were employed to fabricate bulk samples. Characterization results showed simultaneous improvement in the electrical and thermal properties of hot-pressed samples.
Sodium-rich sodium cobalt oxide (Nax CoO2 ) was investigated for structural study and
thermoelectric characterization. Both wet-chemical and electrochemical techniques were
adopted to intercalate sodium ions into the compound, which is expected to improve cryogenic TE performance. Starting from the setup of Na battery, a in-situ Seebeck coefficient
measuring system was proposed, based on the fact that precise sodium concentration control
was successfully achieved.

The focus of calcium cobalt oxide (Ca3 Co4 O9 ) is to develop an innovative synthesis
approach. Using agarose as a template, a sol-gel chemistry route involving mild acetates salt
were developed. Comparing to traditional solid-state reaction, finer product particles, lower
reaction temperature and more effective doping were achieved. Stoichiometric and Yb-doped
samples were successfully prepared and characterized.
Our study confirmed that oxides are a promising category of materials for a wide range
of thermoelectric applications. Even though they are not as sophisticated as traditional TE
materials at present, they showed great potential and interesting physical properties that
would attract interests towards future research.

Copyright by
CHANG LIU
2012

To my parents, for giving me life and all the amazing things that come with it.
And
To Yushu, for being the most amazing of them all.

v

ACKNOWLEDGMENTS

Pursuing a doctoral degree is not just about exploring into the scientific field. It is also, if
not more, an evolution of an individual’s understanding about himself/herself. Throughout
this journey, no one can stand alone to face the pain, confusion and hesitation. Luckily,
there have been a lot of helping hands along my way, pointing out the direction and raise
me up. Hereby, I would like to acknowledge them with my sincerest appreciation.
It has been great honor and precious experience to work with Dr. Donald Morelli.
I learned from him not only experimental skills and research abilities, but also positive
attitudes and kindness. I am also very grateful to my committee members Dr. Eldon Case,
Dr. Tim Hogan and Dr. Jeff Sakamoto for their support and guidance throughout the
program.
I would like to thank my lab-mates Dr. Ponnambalam Vijayabarathi, Dr. Long Zhang,
Dr. Chen Zhou, Dr. Eric Skoug, Hui Sun, Xu Lu, Hao Yang, Gloria Lehr, Steve Bonna
for their help during my study. I would also like to express my thank to my collaborators
Dr. Fei Ren, Dr. Hsin Wang, Colin Blakely, Ezhiylmurugan Rangasamy, Travis Thompson,
Peng Gao, Robert Schmidt, Dr. Bhanu Mahanti, Dat Do, Ed Timm, Karl Dersch and
Brian Wright for their generous assistance and special thank to Dr. Jihui Yang for his
enlightenment and advice.
Last but not least, I am extremely thankful to Bo Xu and Shaowen Ji, Chenling Huang
and Jinglei Xiang, my dearest friends, for helping me get through the most difficult times
and become a stronger person.

vi

PREFACE

My major research interest in materials development is to incorporate novel techniques to
achieve higher preparation efficiency and better material performance. In the present work
on thermoelectric oxides, my research covered a relatively wide range of material systems,
pursuing distinct research goals and involving a variety of synthesis methods. I choose not
to arrange all the preparation processes into one introductory chapter, as otherwise it would
be inconvenient for the readers to navigate among chapters. Instead, I present a general
introduction of contemporary material fabrication strategies and characterization techniques
in an individual chapter, and distributed detailed preparation methods in respective chapters
with corresponding materials systems, for a smoother reading experience. In addition to my
primary work on the complex oxide thermoelectrics, I also investigated a full-Heusler type
alloy Fe2 VAl. It is included in this book as an appendix.

vii

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . .

ix

LIST OF FIGURES

. . . . . . . . . . . . . . . . .

x

.
.
.
.

. . . . . . . . . . . . .

Chapter 1 Introduction to Thermoelectricity
1.1 Thermoelectric Effects and Applications . .
1.2 Efficiency of Thermoelectric Modules . . . .
1.3 Electrical and Thermal Transport Properties

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

1
2
6
7

Chapter 2 Development in Thermoelectric Materials
2.1 New Materials and Approaches . . . . . . . . . . . .
2.2 Oxide Thermoelectrics . . . . . . . . . . . . . . . . .
2.2.1 Motivation of Present Work . . . . . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

11
11
16
18

Chapter 3 Experimental Techniques . . . .
3.1 Sample Preparation . . . . . . . . . . . . .
3.1.1 Powder Processing . . . . . . . . .
3.1.2 Sintering . . . . . . . . . . . . . . .
3.2 Characterization Principles and Techniques
3.2.1 Structural Analysis . . . . . . . . .
3.2.2 Transport Property Measurement .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

19
19
20
22
22
22
23

Chapter 4 Copper Aluminum Oxide . . . . . . . . . .
4.1 Background and Motivation . . . . . . . . . . . . . .
4.2 Experimental Details . . . . . . . . . . . . . . . . . .
4.3 Results and Discussion . . . . . . . . . . . . . . . . .
4.3.1 Composition and Structural Characterization
4.3.2 Transport Property Measurement . . . . . . .
4.4 Summary and Future Work . . . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

27
27
28
29
29
34
37

Chapter 5 Sodium Cobalt Oxide . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Wet-chemical Intercalation Study . . . . . . . . . . . . . . . . . . . . . . . .

40
40
43

viii

.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

44
45
47
51
51
54
55

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

57
57
59
62
62
68

Conclusions and Future Work . . . . . . . . . . . . . . . . . . . .

69

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

73

Appendix A Iron-Vanadium-Aluminum . . . . . . . . . .
A.1 Background and Motivation . . . . . . . . . . . . . .
A.2 Experimental Details . . . . . . . . . . . . . . . . . .
A.3 Results and Discussion . . . . . . . . . . . . . . . . .
A.3.1 Composition and Structural Characterization
A.3.2 Transport Property Measurement . . . . . . .
A.4 Summary . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.

73
74
74
76
76
76
81

. . . . . . . . . . . . . . . . .

85

5.3

5.4

5.2.1 Experimental Details . . . . . . . . . . . . . . . . . . .
5.2.2 Composition and Structural Analysis . . . . . . . . . .
5.2.3 Transport Property Characterization . . . . . . . . . .
Electrochemical Intercalation Study . . . . . . . . . . . . . . .
5.3.1 Experimental Details . . . . . . . . . . . . . . . . . . .
5.3.2 Proposed in-situ Seebeck coefficient Measuring System
Summary and Future Work . . . . . . . . . . . . . . . . . . .

Chapter 6 Calcium Cobalt Oxide . . . . . . . . .
6.1 Background and Motivation . . . . . . . . . . .
6.2 Experimental Details . . . . . . . . . . . . . . .
6.3 Results and Discussion . . . . . . . . . . . . . .
6.3.1 Calcination Condition and Doping Study
6.4 Summary and Future Work . . . . . . . . . . .
Chapter 7

BIBLIOGRAPHY

. . . . . . . . . . . . .

ix

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

LIST OF TABLES

Table 6.1

Density of hot-pressed (Ca1-x Ybx )3 Co4 O9 samples, where M, ρ0 , ρ
represent molecular weight, theoretical density and actual density. .

x

68

LIST OF FIGURES

Figure 1.1

Figure 1.2

Figure 1.3

Figure 1.4

Annual world energy consumption from 1820 to 2010. Reproduced
from [1] and [2]. For interpretation of the references to color in this
and all other figures, the reader is referred to the electronic version
of this dissertation. . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

Two operation modes of thermoelectric uncouples: (A) power generator mode and (B) thermoelectric active cooler mode . . . . . . . .

4

Estimated temperature-dependent efficiency for thermoelectric devices with different ZT values: (red) ZT = 0.3, (orange) ZT = 1.0,
(green) ZT = 2.0, (blue) ZT = 8.0, (purple) ZT = ∞. (A) Power
generator mode and (B) Refrigerator mode . . . . . . . . . . . . . .

7

Intercorrelation between Seebeck coefficient, electrical conductivity,
thermal conductivity and ZT with carrier concentration in a classic
semiconductor. Reproduced from [3]. . . . . . . . . . . . . . . . . .

10

Figure 2.1

ZT versus temperature of some important materials at their peak performance in an chronologically order: (a) Yb0.2 Co4 Sb12 , (b) PbTe,
(c) SiGe, (d) AgSbTe2 -GeTe (TAGS), (e) Bi2 Te3 , (f) CeFe4 Sb12 ,
(g) CsBi4 Te6 , (h) AgPb18+x SbTe20 (LAST), (i) Yb14 MnSb11 , (j)
Ag(PbSn)m SbTe2+m (LASTT), (k) Ybx Ga8-x Ga16 Ge30 , (l) Tl-PbTe,
(m) Hf0.6 Zr0.4 NiSn0.98 Sb0.02 (half-Heusler), (n) NaPb20 SbTe22 (SALT),
(o) Fe4 Sb12 , (p) Zn4 Sb3 , (q) Ba0.08 La0.05 Yb0.04 Co4 Sb12 (Skutterudites) and (r) PbTe-PbSe. . . . . . . . . . . . . . . . . . . . . . . . . 12

Figure 2.2

Electronic density of states for a) RuAl2 and b) Fe2 VAl. Near the
fermi energy level, a) shows a real gap due to hybridization; b) shows
a deep well, or a pseudo-gap. Figure reproduced from [4] [5]. . . . .

15

Schematic DOS in materials with different dimensions: 3-D bulk materials, 2-D nanofilm, 1-D nanowire and 0-D quantum dots. Reproduced from [6] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15

Figure 2.3

xi

Figure 2.4

ZT of recently developed thermoelectric oxides: (a) Ca3 Co4 O9 whisker,
(b) NaCo2 O4 , (c) Bi2 Sr2 Co2 Oy whisker, (d) NaM2 O4 (M=Co, Cu),
(e) NaCo2 O4 , (f) NaCo2 O4 , (g) In2 O3 -SnO2 , (h) MMnO3 (M=Ca,
Bi), (i) CaMO3 (M=Mn, In), (j) ZnO-Al, (k) SrMO3 (M=Ti, Nb),
(l) ZnO-Al nanovoid, (m) ZnO-Al nanovoid. Reproduced from [7] . . 18

Figure 3.1

Schematic flowchart of typical sol-gel synthesis process of powders .

21

Figure 3.2

Schematic illustration of intercalation. Ions can be extracted (A) or
inserted (B) between layers structures. . . . . . . . . . . . . . . . . .

21

Figure 3.3

Sample wiring schematic diagram in a steady-state cryostat . . . . .

24

Figure 4.1

XRD patterns of as-fired Cu1-x Mgx AlO2 (0% x 10%) powders . .

30

Figure 4.2

XRD patterns of cold-pressed Cu1-x Mgx AlO2 (0% x 10%) . . . . .

32

Figure 4.3

XRD patterns of hot-pressed Cu1-x Mgx AlO2 (0% x 10%) . . . . .

33

Figure 4.4

XRD patterns of PECS treated Cu1-x Mgx AlO2 (0% x 10%) . . . .

34

Figure 4.5

SEM images (gentle-beam mode) of hot-pressed CuAlO2 (A), hotpressed CuAl0.98 Mg0.02 O2 (B), hot-pressed CuAl0.925 Mg0.075 O2 (C),
PECS-sintered CuAlO2 (D), PECS-sintered CuAl0.98 Mg0.02 O2 (E)
and PECS-sintered CuAl0.925 Mg0.075 O2 (F). . . . . . . . . . . . . .

35

Figure 4.6

Electrical resistivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%) . .

36

Figure 4.7

Seebeck coefficient of hot-pressed Cu1-x Mgx AlO2 (0% x 5%) . . .

37

Figure 4.8

Power factor of of hot-pressed Cu1-x Mgx AlO2 (0% x 5%) . . . . .

38

Figure 4.9

Thermal diffusivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%) . . .

38

Figure 4.10

Thermal conductivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)

.

39

Figure 4.11

ZT of hot-pressed Cu1-x Mgx AlO2 (0% x 5%) . . . . . . . . . . . .

39

Figure 5.1

Crystal structure of Nax CoO2 [8] . . . . . . . . . . . . . . . . . . . .

41

Figure 5.2

Phase diagram of Nax CoO2 ; reproduced from [9] . . . . . . . . . . .

42

xii

Figure 5.3

Figure 5.4

XRD patterns of Nax CoO2 powders with different nominal starting
sodium concentration and PECS-sintered Nax CoO2 (x=0.8) . . . . .

45

XRD patterns of Nax CoO2 (x=1.0) as-fired powders and PECS treated
pellets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

Figure 5.5

Electrical resistivity of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical
intercalated and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 :
(1) x≈0.71, (2) x≈0.75, (3) x≈0.80, (4) x≈0.85, (5) x≈0.88, (6)
x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10) x≈1.0 . . . . . 47

Figure 5.6

Seebeck coefficient of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical
intercalated and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 :
(1) x≈0.71, (2) x≈0.75, (3) x≈0.80, (4) x≈0.85, (5) x≈0.88, (6)
x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10) x≈1.0 . . . . .

49

Figure 5.7

Thermal conductivity of (A) PECS treated Na0.74 CoO2 , (B) wetchemical intercalated and PECS treated Na1.0 CoO2 and reported
S-Nax CoO2 : (1) x≈0.71, (2) x≈0.75, (3) x≈0.80, (4) x≈0.85, (5)
x≈0.88, (6) x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10) x≈1.0 50

Figure 5.8

ZT of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical intercalated
and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 : (1) x≈0.71,
(2) x≈0.75, (3) x≈0.80, (4) x≈0.85, (5) x≈0.88, (6) x≈0.89, (7)x≈0.96,
(8) x≈0.97, (9) x≈0.99 and (10) x≈1.0 . . . . . . . . . . . . . . . .

51

Original battery setup using cast Na0.74 CoO2 powders and sodium
metal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

Improved T-cell battery setup using Na0.74 CoO2 as both working and
counter electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

Starting and stopping voltage plotted on a complete discharge curve,
reproduced from [10] . . . . . . . . . . . . . . . . . . . . . . . . . .

54

(a) XRD patterns of Na0.74 CoO2 before (blue curve) and after (red
curve) electrochemical intercalation. (b) shift of the primary peak
towards the higher angle . . . . . . . . . . . . . . . . . . . . . . . .

55

Figure 5.13

Proposed in-situ Seebeck coefficient measuring system . . . . . . . .

56

Figure 6.1

Chemical formula of Agarose . . . . . . . . . . . . . . . . . . . . . .

60

Figure 5.9

Figure 5.10

Figure 5.11

Figure 5.12

xiii

Figure 6.2

Figure 6.3

Figure 6.4

Figure 6.5

Sol-gel synthesis of Ca3 Co4 O9 : (A) solution, (B) gel, (C) calcined
powders, (D) sintered pellet . . . . . . . . . . . . . . . . . . . . . .

61

XRD pattern of solid-state reaction Ca3 Co4 O9 powders fire at various temperature. Asterisks indicate impurity peaks identified as unreacted Co3 O4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

XRD pattern of sol-gel synthesis Ca3 Co4 O9 powders fire at various
temperature. Asterisks indicate impurity peaks identified as unreacted Co3 O4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

XRD pattern of as-fired, PECS treated and annealed Ca3 Co4 O9 powders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65

Figure 6.6

XRD patterns of undoped, 2% and 5% Yb doped (Ca1-x Ybx )3 Co4 O9
fired at 850 ◦ C and 900 ◦ C via solid state reaction. Blue asterisks
indicate unreacted Yb2 O3 and red asterisks indicate Ca3 Co2 O6 phase. 66

Figure 6.7

XRD patterns of undoped, 1%, 2%, 4%, 8% and 15% Yb doped
(Ca1-x Ybx )3 Co4 O9 fired at 850 ◦ C via sol-gel synthesis. Asterisks
indicate impurity phases. . . . . . . . . . . . . . . . . . . . . . . . .

67

Figure A.1

High voltage electric arc-melter . . . . . . . . . . . . . . . . . . . . .

75

Figure A.2

XRD patterns of (Fe1-x Cox )2 VAl (5% x 40%) . . . . . . . . . . .

77

Figure A.3

Seebeck coefficient of (Fe1-x Cox )2 VAl (5% x 40%) . . . . . . . . .

78

Figure A.4

Electrical resistivity of (Fe1-x Cox )2 VAl (5% x 40%) . . . . . . . .

79

Figure A.5

Power factor of (Fe1-x Cox )2 VAl (5% x 40%) . . . . . . . . . . . .

80

Figure A.6

Total thermal conductivity of (Fe1-x Cox )2 VAl (5% x 40%) . . . .

81

Figure A.7

Electronic thermal conductivity of (Fe1-x Cox )2 VAl (5% x 40%) . .

82

Figure A.8

ZT of of (Fe1-x Cox )2 VAl (5% x 40%) . . . . . . . . . . . . . . . .

83

xiv

Chapter 1
Introduction to Thermoelectricity
The energy consumption of human society has grown substantially, especially since the 1940s
(Figure 1.1). In contrast, traditional and primary energy resources, such as oil and coal, are
draining away with the increasing demand. This contradiction is putting on increasing
pressure on every aspect of human society and placing an urgent call for solutions. One
approach to achieve the goal is to discover new energy resources. The development of solar,
wind, nuclear and biomass energy has shown promising progress, even though none of them
can compete with the cost and convenience of fossil fuels. An alternative answer to the same
question would be enhancing efficiency in the use of existing energy resources.
In most cases, poor efficiency occurs due to heat loss during energy transformation.
Taking combustion engine vehicles for example, only up to 30% of total energy effectively
serves as driving power, while the majority is lost in the form of hot exhaust. Therefore,
a big opportunity is presented to reuse this lost energy through waste heat recovery. The
emergence of thermoelectricity (TE) technology provides almost the only feasible option to
realize this idea. During the past decades, thermoelectric technology has received increasing
1

EJ (1018 J) per year

600
500
400
300
200
100
0
1820 1840 1860 1880 1900 1920 1940 1960 1980 2000
Year
Figure 1.1: Annual world energy consumption from 1820 to 2010. Reproduced from [1]
and [2]. For interpretation of the references to color in this and all other figures, the reader
is referred to the electronic version of this dissertation.

attention and scrutiny for potential applications in energy harvesting, as well as solid-state
heating and cooling. Every story has a beginning. For thermoelectricity, it all starts with
the fundamental thermoelectric effects discovered centuries ago.

1.1

Thermoelectric Effects and Applications

In 1821, T. J. Seebeck firstly discovered that electrical potential can be established at a
junction of two dissimilar metals when the circuit was placed in temperature gradient [11].
Electric carriers, electrons or holes, diffuse from the hot end to the cold end and create a
voltage difference under the influence of temperature gradient. The migration stops once a
balance is reached between the thermal driving force and electrical fields. Shown in Equation
1.1, the ratio of Seebeck voltage to the temperature difference is defined as Seebeck coefficient,
2

which is also known as thermoelectric power or thermopower.

S=

∆V
∆T

(1.1)

The widely used thermocouples for temperature measurement make use of this effect. In
most cases, systems with single charge carriers display better TE performance as carriers
with opposite charge would cancel each other out in double carriers systems, as is shown in
Equation 1.2, where σ p and σ n are electrical conductivity contribution of holes and electrons.
S p and S n are Seebeck coefficient contribution of hole and electrons, and are opposite in
sign.

S=

σp Sp + σn Sn
σp + σn

(1.2)

A reverse effect was found by J. Peltier a decade later [12]. He observed that, when
˙
electric current passed through a heterogeneous junction, certain amount of heat Q can be
released or absorbed depending on the current direction. This thermodynamically reversible
Peltier effect is similar but fundamentally distinct from Joule heating, which is irreversible.
The Peltier coefficient was defined to indicate the ratio of generated heat to the passing
current, following Equation 1.3. Based on this effect, solid-state cooling or heating modules
can be used for heat management.

Π=

˙
Q
I

(1.3)

In 1851, the third TE effect was predicted by W. Thomson (Lord Kelvin), which repre3

sents a reversible heating/cooling effect produced by electric current in magnetic field. This
effect was afterwards demonstrated and recognized as the Thomson effect, which describes
the heating/cooling when current passes through a conductor in temperature gradient. He
also revealed the interconnection among the three main TE effects by proposing the first
(Equation. 1.4) and second (Equation. 1.5) Kelvin Relations.

µ = T dS/dT

(1.4)

Π
T

(1.5)

S=

In addition, the Nernst Ettingshausen effect can happen when a conductor with passing
electric current is placed in an environment where magnetic field and temperature gradient
are present at the same time.
Heat Source
n-type
_

Active Cooling

p-type

n-type
_

+

_

+

_

+
_

p-type

+
_

+

+

Heat Sink

Heat Rejection

(A)

(B)

Figure 1.2: Two operation modes of thermoelectric uncouples: (A) power generator mode
and (B) thermoelectric active cooler mode
The application of thermoelectric energy generator (TEG) [Figure 1.2, A] started with
4

deep-space missions, which requires reliable, compact and rugged power source. Contemporary TE materials, including lead telluride and GeTe-AgSbTe alloys (TAGS), has been used
to fabricate radioisotope thermoelectric generators (RTG) for spacecrafts. These attributes
also attracted automotive manufacturers to use TE modules for waste heat retraction in
combustion engine powered vehicles. A exhaust pipe mount two-staged TE module has been
demonstrated by General Motors on its Chevrolet Suburban [13]. A similar energy recovery
TEG was also proposed by BMW as a key component of its EfficientDynamics technology
to improve fuel efficiency [14].

The Peltier solid-state coolers [Figure 1.2, B] are especially attractive because there are
no moving parts or liquid refrigerant involved the these types of devices, which demonstrate
higher reliability and flexibility. The products can be made small in size and are easy to tailor
for desired use environment, such as automobiles. Even though the low efficiency remains
a major obstacle, Peltier modules has been deployed in commercial products. BMW was
the first among car manufacturers to have started to use TE coolers in car seats to achieve
targeted area climate control [14].

Going beyond bulk materials, applications in small dimension have also been explored
[15]. A thin film superlattice coating was demonstrated to realize in active chip-scale heat
management, with higher efficiency than conduction cooling and higher reliability than liquid
based solutions. A concept of on-chip heat recycle module was also proposed and tested
recently [16] [17].
5

1.2

Efficiency of Thermoelectric Modules

Even with all the benefits, whether the industry and the general market would accept thermoelectric modules is simply decided by their efficiency. As for cooling/heating devices,
TE products are expected to exhibit comparable performance to compete with conventional
technologies.
The conversion efficiency of any heat engine is limited by the Carnot efficiency ηc . Therefore the efficiency of a thermoelectric power generator can be expressed as a fraction of ηc
in Equation 1.6:

η = ηc ·

1 + ZTavg − 1

(1.6)

1 + ZTavg + Tcold /Thot

where ZT is the figure-of-merit, indicating the properties of thermoelectric materials (see
below). In the heat-pump mode (refrigeration), the coefficient of performance (COP ) of
thermoelectric modules can be expressed as Equation 1.7:

COP =

Tavg
Q
=
·
W
Thot − Tcold

1 + ZTavg − 1
1 + ZTavg + 1

−

1
2

(1.7)

where T avg =(T cold +T hot )/2 is the average temperature.
Supposing that, in both senarios, ZT avg is temperature-independent and Tcold is fixed
at 300 K, the coefficient of conversion for a generator and the COP for a refrigerator versus
ZT are plotted in Figure 1.3. It is generally accepted that, to compete existing technologies,
ZT is expected to exceed 2.0 to justify the manufacturing cost of materials and devices for
practical applications.
6

Conversion Efficiency

0.8
0.6
0.4
0.2
0.0

0

200

400

600 800 1000 1200 1400
T K

(A)
3.5
3.0
COP

2.5
2.0
1.5
1.0
0.5
0.0
0

100

200

300
T K

400

500

(B)
Figure 1.3: Estimated temperature-dependent efficiency for thermoelectric devices with different ZT values: (red) ZT = 0.3, (orange) ZT = 1.0, (green) ZT = 2.0, (blue) ZT = 8.0,
(purple) ZT = ∞. (A) Power generator mode and (B) Refrigerator mode

1.3

Electrical and Thermal Transport Properties

Figure-of-merit

ZT =

σ · S2
·T
κe + κl
7

(1.8)

A dimensionless figure-of-merit (ZT ) is expressed as Equation 1.8 as an unit-less indicator
of the TE property of a single material. As was mentioned, this is the single parameter
that characterizes the performance of a material. Generally, ideal TE materials possess high
electrical conductivity, larger Seebeck coefficient and low thermal conductivity to exhibit
large ZT. Higher operation temperature is also a positive factor in increase the parameter.
Strategies to enhance ZT will be discussed in detail in the next chapter.

Electrical Conduction

Electrical conductivity σ can be expressed as Equation 1.9:

σ = neµ

(1.9)

where n is carrier concentration, e is the electronic charge and µ is the carrier mobility.
Electrical conductivity is proportional to carrier concentration. On the other hand, the
Seebeck coefficient of a strongly degenerated system with single parabolic band model is
given in the Pisarenko relation in Equation 1.10 [18]:
8π 2 k 2
π
S=
· m∗ T ( )2/3
3n
3eh2

(1.10)

where h is Planck’s constant, k is Boltzmann’s constant, m∗ is the carrier effective mass
and n is the carrier concentration. Seebeck coefficient therefore is inversely to the carrier
concentration. As we pursuit large total electronic contribution, defined as the power factor
(σS 2 ), an optimized n is required to balance the two parameters.
8

Thermal Conduction
Total thermal conductivity κ includes two parts: lattice thermal conductivity κl and electronic thermal conductivity κe . κl is phonon transfer in the form of lattice vibrations, while
κe is caused by heat-carrying charge carrier migration, which is strongly tied to electrical
conductivity.

κe = L0 σT

(1.11)

κe can be calculated using the empirical Wiedemann-Franz law (Equation 1.11). L0 is
the Lorenz number, which equals 2.44 × 10−8 W ΩK −2 . As κ is a measurable parameter, κl
can be separated following Equation 1.12.

κl = κ − κe

(1.12)

Carrier Concentration and Mobility
As was discussed, transport properties are all inherently related to the electronic carrier
concentration. As shown in Figure 1.4, to get high σ without hurting S, improving carrier
mobility is preferred to increasing carrier concentration. Therefore, good TE materials are
usually semiconductors with moderate carrier concentration (1019 -1020 cm-3 ).

9

𝛔

ZT

arbitrary units

|S|

𝜅
1019
1020
Carrier Concentration (cm-3 )

Figure 1.4: Intercorrelation between Seebeck coefficient, electrical conductivity, thermal conductivity and ZT with carrier concentration in a classic semiconductor. Reproduced from [3].

10

Chapter 2
Development in Thermoelectric
Materials
In this chapter, some classic approaches and new concepts in thermoelectric material development are introduced, along with representative materials systems. Special attention is given
to oxides with complex crystal structure. A brief overview of the benefits and challenges of
this new category of materials is presented.

2.1

New Materials and Approaches

Leaving the device construction steps out of account, all of the questions in TE technology
have been boiled down to the search for materials with large ZT. The dilemma, as was previously discussed, lies in the interrelation between fundamental physical transport properties.
Traditionally, effort was made to reach a compromise among those parameters to achieve an
optimized integration. Starting from the 1990s, however, novel materials have been developed with the arrival of new ideas, strategies and preparation techniques. A summary of the
11

peak performance versus temperature of some important thermoelectric materials is shown
in Figure 2.1:

1.8

(h) 2004

(r) 2011

(n) 2009

1.6

(q) 2011

(j) 2006
1.4

(l) 2008

ZT

(d) 1972

1.2

(o) 2010

(a) 1855

(p) 1997

1.0
(e) 1995
0.8

(k) 2008
(i) 2006

(m) 2009

(b) 1957

(f) 2011

(c) 1964

(g) 2004
200

400

600

800

1000

1200

1400

Temperature (K)

Figure 2.1: ZT versus temperature of some important materials at their peak performance in an chronologically order: (a) Yb0.2 Co4 Sb12 , (b) PbTe, (c) SiGe, (d) AgSbTe2 GeTe (TAGS), (e) Bi2 Te3 , (f) CeFe4 Sb12 , (g) CsBi4 Te6 , (h) AgPb18+x SbTe20 (LAST),
(i) Yb14 MnSb11 , (j) Ag(PbSn)m SbTe2+m (LASTT), (k) Ybx Ga8-x Ga16 Ge30 , (l) Tl-PbTe,
(m) Hf0.6 Zr0.4 NiSn0.98 Sb0.02 (half-Heusler), (n) NaPb20 SbTe22 (SALT), (o) Fe4 Sb12 , (p)
Zn4 Sb3 , (q) Ba0.08 La0.05 Yb0.04 Co4 Sb12 (Skutterudites) and (r) PbTe-PbSe.
In 1995, the concept of Phonon Glass/Electron-Crystal (PGEC) materials was proposed
by Slack [19]. In ideal PGEC materials, phonons are expected to propagate in the manner
as if they are in amorphous materials, such as glasses. Meanwhile, they retain the electrical
conduction behavior as crystals. The thermal and electric contribution are expected to be
separately tunable. Even though perfect materials with such behavior are not yet available,
progress has been made to bring the idea closer to reality.
12

Minimization of Thermal Conductivity

Forming solid solutions through atomic substitution is a traditional way to reduce lattice
thermal conductivity. Bigger or heavier atoms induce size and mass contrast between dopants
and hosts atoms and break down matrix regularity. Phonons are therefore strongly scattered
by the increased complexity in the solid-solutions [20]. Usually two or more compounds
sharing similar crystal structures are selected, as in this case a wide solid-solution range
is expected. This method has been widely applied to materials that already possess good
thermoelectric performance while further reduction of thermal conductivity is needed, such
as Bi2 Te3 -Bi2 Se3 [21] and PbTe-PbSe [22]. Recent discovery of copper-based diamondlike compounds revealed a new approach: discovering materials with low intrinsic thermal
conductivity [23]. These compounds usually consist naturally complex structures, large unit
cells and heavy elements. By intentionally creating doubled, tripled or even quadrupled unit
cells, complex crystal structures with many atoms per unit cell can be realized, which can
contribute to additional thermal conductivity reduction [24].
Another category of materials with phonon scattering agents are skutterudite [25] [26]
and calthrates [26]. These compounds possess open space in their crystal structure. Some
external atoms, such as rare-earth elements, can be deliberately inserted into those hollow
cages to become ”rattlers”. These weakly bonded atoms can effectively disturb phonon propagation and provides additional reduction in thermal conductivity [27]. Although electrical
properties may be affected by the phonon scattering atoms, the electron carrier transport
properties is usually reduced by less than one magnitude, which is lower than that of thermal
conductivity reduction and beneficial to the improvements of ZT.
Advanced powder processing techniques, involving energized ball-milling and rapid sin13

tering, can produce smaller grain size and finer microstructure, which can introduce more
interfaces that serve as phonon scattering agents [28].

Enhancement of Power Factor
Admitting that great progress has been made to in minimizing materials’ thermal conductivity, it can never be indefinitely reduced. The limits of lattice thermal conductivity comes
from the minimum phonon mean free paths, which can be smaller than the interatomic spacing [29]. Therefore, the enhancement of power factor is an equivalently important approach.

S=

π2k2T
·
3e

d ln σ(E)
dE

(2.1)
E=Ef

Mott equation (Equation 2.1) provides a simple description of the Seebeck coefficient
[30]. Assuming that electron relaxation time is constant, the density of states (DOS) is
proportional to σ(E). Subsequently S can be greatly promoted if abrupt change of the
DOS happens near the Fermi energy level. Such band structures do exist in some bulk
materials, caused by hybridization between transition metals and main group elements. As
shown in Figure 2.2, the DOS near the Fermi level displays sharp feature, creating a deep
gap in narrow-band semiconductors such as RuAl2 [4], or a pseudo-gap in semimetals such
as Fe2 VAl [5]. Due to this hybridization, unusual transport properties and enhanced TE
performance are expected to appear. The author has conducted a research project on the
doping effect study in Heusler-type Fe2 VAl alloy. The results are presented in tthe Appendix
A.
For polycrystalline materials, defects near the grain boundary can cause serious reduction
in electron transport. By improving the electrical conduction near the boundaries, higher
14

semimetal-like resistivity. the of the major objectives proposal is to gain a gain a
semimetal-like resistivity. One of One major objectives of this of this proposal is to deeper
understanding of this hybridization the nature nature of electronic the neighborhood of
understanding of this hybridization gap andgap and theof electronic states instates in the neighbor
the pseudogap through detailed transport investigations both both experimental
the pseudogap region region through detailed transport investigations experimentally and
theoretically and careful electronic structure calculations.
theoretically and careful ab initioab initio electronic structure calculations.
a) 
(a)

a)

4
3
2
1
0

4

Ef

3
2
1
0
-10
-5

-10

-5
0

0
5

5
10

10

10

10
Density of States (states/eV)

5

Density of States (states/eV)
Density of States (states/eV) 

Density of States (states/eV) 
Density of States (states/eV)

Density of States (states/eV)

5

b)

b)
(b)
8

8
6
4
2

Ef

6
4
2
0

0

-10-5

-10

Energy (eV)
Energy (eV)
Energy (eV)

-5 0

0 5

5

Energy (eV)
Energy (eV)
Energy (eV) 

Figure 2.2: Single particle states statesRuAl2 a) RuAland b) TheVAl. The former is an example of an inter
Figure 16. Electronic density of of a) for a) RuAl 2 and b) Fe2 former is an fermi energy
Figure 16. Single particle density ofdensityfor states for and b)2Fe2VAl. Fe2 VAl. Near theexample of an intermetallic
level, a)true hybridization-gap; an example of a b) shows aor a pseudo-gap. Data adapted from
with a shows a real the due the latter an example well, deep well, a pseudo-gap. Data adapted from Wei
with a true hybridization-gap; gap latterto hybridization; deep of a deep well, oror a pseudo-gap. Figure Weinert and
reproduced from [4] [5].
Watson62.Watson62.

Criticalfactor can to obtaining the As was reported in some intermetallicwe the such as location o
to obtaining the unusual transport behavior that we seek is seek is the
power Critical also be achieved. unusual transport behavior that alloys, location of EF (in
Figure 16 indicated by the line at line at since the electrons near the Fermi Fermi an
Figure 16 indicated by the vertical vertical E = 0), E = 0), since the electrons near thelevel in lev
CoSi [31] [32] and Ni3 Al [33], boron segregation along the grain boundaries can improve the

28
28
mechanical strength and boundary conductivity. Silver particle inclusion at grain boundaries
have also been reported in a variety of materials to improve interfacial conduction and
therefore greatly reduced electrical resistivity [34].

3-D

2-D

DOS

DOS

DOS

DOS

Nanostructure Inclusion

1-D

0-D

Figure 2.3: Schematic DOS in materials with different dimensions: 3-D bulk materials, 2-D
nanofilm, 1-D nanowire and 0-D quantum dots. Reproduced from [6]

The concept of low dimensional materials was introduced to the thermoelectric commu15

nity by Hicks and Dresslhaus in the 1990s [6]. Nanotechnology provides a viable approach
to include nano-scale microstructures during materials fabrication and can simultaneously
achieve optimization in thermal conductivity and power factor. Theoretical minimum value
of lattice thermal conductivity can be estimated by kl = 1/3Cv l, where the phonon mean
free path l is limited by the interatomic spacing. By reducing materials dimension to 2-D
(quantum wells), 1-D (quantum wires) or even 0-D (quantum dots), lattice spacing can be
pushed to the extent that they are comparable to interatomic distance (Figure. 2.3).

S=

π2k2T
·
3e

d ln g(E) d ln v 2 (E) d ln τ (E)
+
+
dE
dE
dE

(2.2)
E=Ef

Expanded Mott and Jones relation (Equation 2.2) shows the possibility to improve electrical properties. The power factor enhancement can be realized when Fermi energy level is
positioned near a rapid change in the DOS, as shown in Figure 2.3. Energy filtering effect
has also been proposed and demonstrated to serve the same purpose by properly adjusting
the mechanism of electron scattering in nano-structured materials [35].

2.2

Oxide Thermoelectrics

Most contemporary TE materials to date are binary or ternary chalcogenides and pnictides. In spite of their good properties, these materials have inevitable drawbacks. They are
usually chemically or mechanically unstable at elevated temperature, which prevents them
from achieving better performance with increased temperature gradient. Most of them are
sensitive to oxidization, and therefore require extra packaging and insulation for practical
use. In addition, the toxic and expensive elements they consist of make them less desirable
16

to manufacture. All these challenges have hindered the widespread use of thermoelectric
technology. Yet they have also motivated pursuit for new candidate materials at the same
time. Oxides have attracted great attention in recent years due to their thermal stability,
nontoxicity and low production and maintenance cost. In addition, low density and good
mechanical strength qualifies oxides as preferable materials in device construction.
Although possessing large Seebeck coefficient, oxides have long suffered from poor electrical conductivity and high thermal conductivity. With the the discovery of large Seebeck
coefficient in NaCo2 O4 [36], large power factor in single crystal Nax CoO2 [37] and low thermal conductivity in polycrystalline samples [38], development of oxide TE has received a
strong boost over the past decades. More importantly, the concept of complex oxides involving nanoscale building blocks revealed a new class of materials. In simple systems, the
inherently interrelated S, σ and κ are difficult to fine tune. With multiple atom species and
crystalline fields in the complex hybrid structures, thermal and electrical properties may
be individually adjusted with weaker intercorrelation. Recently reported ZT values of both
p-type and n-type candidates are summarized in Figure 2.4.
Current research of p-type TE oxides has focused on sodium and calcium based cobalt
oxides and their derivative compounds [39]. The maximum ZT of single crystal NaCo2 O4 was
reported to surpass unity in 2001 [40], while polycrystalline counterparts reached ZT =0.8
[41]. Single crystal and polycrystalline Nax CoO2 with various sodium concentration has been
investigated for high-temperature TE applications. Calcium cobalt oxides (Ca3 Co4 O9 [42],
Ca2 Co2 O5 [43]) and the modulated layered Bi2 M3 Co2 Oy (M = Sr, Ca, and Ba) also display
promising performance and even better stability [44] [45]. In contrast, n-type materials have
lagged behind in achieving large ZT. Al-doped ZnO, a wide gap semiconductor with simple
17

1.2

p-type - single crystal

(c)

p-type - polycrystalline
1

n-type - polycrystalline

(a)

(b)

0.8

(f)

ZT

(e)
(m)

0.6

(d)
(l)

0.4
(k)

(j)
0.2
(g)
0
1992

(h)
1994

(i)
1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.4: ZT of recently developed thermoelectric oxides: (a) Ca3 Co4 O9 whisker, (b)
NaCo2 O4 , (c) Bi2 Sr2 Co2 Oy whisker, (d) NaM2 O4 (M=Co, Cu), (e) NaCo2 O4 , (f) NaCo2 O4 ,
(g) In2 O3 -SnO2 , (h) MMnO3 (M=Ca, Bi), (i) CaMO3 (M=Mn, In), (j) ZnO-Al, (k) SrMO3
(M=Ti, Nb), (l) ZnO-Al nanovoid, (m) ZnO-Al nanovoid. Reproduced from [7]
structure, offers competitive power factor compared to conventional materials, but limited
by very high thermal conductivity [46]. Newly emerging heavily doped strontium titanium
oxide (SrTiO3 ) and its layered derivatives (SrO)(SrTiO3 )m , however, show great potential
in reaching better performance [47].

2.2.1

Motivation of Present Work

In this thesis, three p-type complex oxide-based systems are investigated as potential thermoelectric materials, with respective research goals and approaches. Polycrystalline samples
were chosen due to the considerably lower preparation cost and lower thermal conductivity.
18

Chapter 3
Experimental Techniques
In this chapter, an overview of techniques used to prepare polycrystalline thermoelectric
oxides is presented. Characterization methods used to examine material structural properties
and evaluate transport performance are briefly introduced, with emphasis on equipment
setup and measuring principles.

3.1

Sample Preparation

With active exploration of material systems for thermoelectric applications, various fabrication methods are utilized to produce single crystal and polycrystalline products. Commercially produced single crystals are usually grown from polycrystalline powders using
the Czochralski process or the Bridgman technique [48]. Metallic compounds are usually
produced via direct fusion of pure elements and annealing. Alternative methods, such as
melt-spinning [49] and mechanical alloying [50], have been developed to finetune and improve
the properties of existing material systems.
19

3.1.1

Powder Processing

Solid-state reaction is the most widely used synthesis method in powder processing. Reactant powders were mixed according to desired stoichiometry using low-speed stationary
mills or high-energy ball-milling. The former is convenient and effective in mixing, while the
latter offers shorter processing time and additional particle size reduction effect. Inert gas
atmosphere is required for oxygen or water sensitive samples during mixing. The precursor
mixtures are then calcined at desired temperature in preferred gas environment to activate
chemical reactions.
Solid-state reaction is an economic and fast synthesis technique and was used to calcine
reactant mixtures in all three projects. Due to the slow diffusion rate in solids, however,
the reaction speed and product homogeneity are limited. In the CuAlO2 work, low-speed
powder mixing, powder calcination and cold-pressing were carried out in Dr. Case’s lab
at Michigan State University. Nax CoO2 and Ca3 Co4 O9 samples were processed using the
SPEX SamplePrep 8000M mixer with glass vials and polyethylene grinding media.
Sol-gel process is a wet-chemical technique, providing finer product particles and improved preparation efficiency. Owing to improved reaction interface and higher reactivity,
sol-gel synthesis also has facilitated easier and more effective doping, which sometimes cannot
be achieved through traditional solid-state reactions.
Figure 3.1 demonstrats a typical sol-gel process. Reactants are first dissolved in clear
solvent. Subsequently, a integrated network is formed with the help of binders/template. The
as-formed gels are then dried to remove solvents in the precursor and turn to xerogels with
much lower porosity. A pre-calcination step is required to burn out the organic components.
The final product is obtained after calcination at desire temperature. Most reported sol-gel
20

1

2

Sol

3

Gel
Reactant A

Xerogel
Reactant B

4

Pre-calcined
Precursor

Polymer

Product

Product

Figure 3.1: Schematic flowchart of typical sol-gel synthesis process of powders

processing methods involve harsh organic acid. In this research, we take a simpler approach
with mild reactants include metal acetate salts, agarose and water.
Chemical intercalation is a technique wildly used in battery technology. Strictly
speaking, it serves to modify a well-set material system by inserting or extracting ions from
a matrix under the influence of concentration gradient or electrical field. In layered structures
like Nax CoO2 , or LiCoO2 , Na and Li ions can enter and leave the framework, as shown in
Figure 3.2. Using either wet-chemical or electro-chemical approach, ions can be inserted
from or extracted to a certain compound to reach desired ’doping’ level. Within certain
limits, this is process can be viewed as reversible without damaging the structure.

A

B

Figure 3.2: Schematic illustration of intercalation. Ions can be extracted (A) or inserted (B)
between layers structures.

21

3.1.2

Sintering

As-produced powders are consolidated into a bulk sample using a variety of sintering techniques. Cold-pressing, also know as pressureless sintering, is the most intuitive consolidation method. Powders were placed in stainless steel dies and pressed at room temperature
to form loosely bonded green compacts. The compacts were then calcined above critical
temperature to eliminate the pores between powder particles.
Hot-pressing simultaneously applies pressure and heat to powders, resulting in more
effective diffusion and higher final sample density. However, since both cold-pressed and
hot-pressed take hours to ramp to target temperature, grain growth may occur and lead to
augmented thermal conductivity in certain systems. In the CuAlO2 project, hot-pressing
was performed by Dr. John Colling at Dr. Gopalan Srinivasan’s lab in Oakland University.
Pulsed electric current sintering (PECS), also known as spark plasma sintering,
applies pulsed DC current through powders pre-pressed in graphite or tungsten carbide dies.
The internally generated heat, unlike conduction from external heating elements, enables
greatly faster ramping rate and therefore shorter sintering time. Samples can preserve their
fine microstructures while achieving high density. PECS processing was finished in Dr. Tim
Hogan’s lab at Michigan State University.

3.2
3.2.1

Characterization Principles and Techniques
Structural Analysis

Powder X-ray Diffraction (XRD) served for crystal structure determination and phase
identification. All samples were pulverized into fine powders and scanned using a Rigaku
22

MiniFlex II X-ray diffractometer(Cu Kα , λ=0.154 nm, 30kV, 15mA). Phase identification
was accomplished with Jade XRD analysis software (version 9.0) and MDI mineral database,
Scanning Electron Microscopy (SEM) was used for microstructure observation and
phase detection. The Two field-emission scanning electron microscope (JEOL JSM-7500F).
In gentle-beam (GB) mode, separated phases appear in distinct grey scale due to the different
average molecular weight. Energy-dispersive X-ray spectroscopy (EDS) is also a good indication of phase separation based on atomic ratio measurement. SEM work was performed
at the Center for Advanced Microscopy at Michigan State University.
For more accurate results, which is crucial in determining the sodium concentration in
Nax CoO2 , an inductively coupled plasma-mass spectrometry (ICP-MS) was used.
Mr. Colin from the Department of Chemistry offered help in ICP analysis.

3.2.2

Transport Property Measurement

Electrical Conductivity and Seebeck Coefficient
Low-temperature measurement is carried out in a lab-built steady state cryostat system
cooled by liquid nitrogen. A bar-shaped sample was first cut out of the as-made bulk pellets
and attached to a copper base using silver epoxy (Figure 3.3). Two copper strips were glued
using the same epoxy onto one side of the bar as probe contacts. Two type-T thermocouple
were soldered onto the copper stripes. A metal-film resistor (Rh =1000 Ω) was wrapped
with copper foil and adhered to the top of sample bar with the same silver epoxy. This
type of epoxy is used because it is both electrically and thermally conductive. Finally,
two current wires are attached at the top and the bottom of the sample. These wires are
electrically isolated from the heater. The entire unit was sealed in a high vacuum (<10−5
23

Ih
Heater

Vh

Copper
foil

Thot

Is
Voltage
probes

Sample

Vs

Tcold

Copper base

Figure 3.3: Sample wiring schematic diagram in a steady-state cryostat
torr) chamber. Liquid nitrogen is flowed into the copper base, which is equipped with a
temperature sensor and its own heater. Control of the base temperature is achieved using
a temperature controller and balancing the cooling effect of liquid nitrogen with Joule heat
from the base heater.
Measurement of the transport properties is performed as follows: first, the base temperature is set to a specified value (say 80 K). Next, the sample current (I s ) is energized.
The resulting resistive voltage V across the sample is detected using the copper legs of the
two thermocouples. From the values of V and I s the resistivity of the sample is calculated
using ρ = (V /Is )(A/L), where A is the cross-sectional area of the specimen and L is the
distance between the voltage probes. The sample current is then turned off and the heater
current is energized. Joule heat from the heater flows down the sample and results in a
24

temperature gradient along its length. When a steady-state condition is achieved, the temperature difference along the sample (∆T = Thot − Tcold ) is measured by measuring the
voltages across the ”hot” and ”cold” thermocouples respectively, and using a calibration
table to convert these values to temperatures. The thermal conductivity of specimen is then
2
given by κ = (Ih Rh /∆T )(L/A). Simultaneously, the Seebeck coefficient Vs /∆T is calcu-

lated. Finally, the sample heater current is turned off, and the copper base temperature is
set to the next value (say 90 K). In this way, curves of the temperature dependence of the
three relevant thermoelectric parameters are determined. The experiment is controlled by a
computerized LabView software sequence. S and σ are measured in the same manner using
a commercial system (ULVAC ZEM-3) above 300 K.

Thermal Conductivity
As was discussed, thermal conductivity measurement below 300 K is carried out simultaneously in the cryostat. In contrast, in the high temperature measurement is executed in a
different way to eliminate radiation loss. κ can be expressed as Equation 3.1, where α, ρ
and c p are thermal diffusivity, density and specific heat capacity respectively. The density
is obtained using the Archimedes law, and c p can be determined using differential scanning
calorimetry (DSC). Thermal diffusivity is measured in a laser flash thermal diffusivity analyzer (LFA, Anter Flashline X-platform). Samples were sliced into thin disks with 1-2 mm
in thickness.

κ = α · ρ · cp

(3.1)

Alternatively, c p of solids with minimal electron contribution has good approximation
25

with the Dulong-Petit law when the solids are above their Debye temperature. N, R and
M in Equation 3.2 refer to the number of atoms in one molecule, the ideal gas constant
and molar mass of the molecule. Using this expression, we can estimate materials specific
heat in the high temperature range with fairly accurate fit. Dr. Hsin Wang from the High
Temperature Materials Laboratory (HTML) at Oak Ridge National Laboratory provided
generous assistance in high-temperature transport property measurement.

cp =

Cp,m
3·R·N
=
M
M

(3.2)

Carrier Concentration and Mobility
Concentration and mobility of electrical carriers play an important role for us to understand
samples’ transport mechanism. Low-temperate Hall measurement can be performed using
a commercial apparatus, VersaLab from Quantum Design based on an assumption of one
band model. Knowing sample thickness, Hall voltage and magnetic field, both parameters
can be calculated following Equation 3.3 and 3.4:

Ey
V t
1
= H =
jx B
Ix B
ne
R
µ= H
ρ
RH =

26

(3.3)
(3.4)

Chapter 4
Copper Aluminum Oxide
Copper aluminum oxide (CuAlO2 ) is a delafossite-type compound with layered structure. It
recently emerged as a new candidate for high-temperature thermoelectric applications. In
the present research, magnesium was selected as a dopant to substitute for aluminum atoms
and improve thermoelectric performance. Three different sintering techniques were adopted
to prepare bulk materials for structural study and transport property measurement. The
doping effect and preparation conditions are studied and discussed.

4.1

Background and Motivation

Delafossite structure is named after French mineralogist Gabriel Delafosse by Friedel, who
discovered copper iron oxide (CuFeO2 ) in 1873. Since then, a variety of compounds with
similar crystal structure have been discovered and prepared. Some of them received close
investigation for battery technologies and catalytic applications, including copper aluminum
oxide (CuAlO2 ). It has been widely applied in the photocatalytic hydrogen generation in
the form of transparent conductive oxides (TCOs) [51], thanks to its exceptional chemical
27

stability and small band-gap energy. Noticing these properties, Koumoto suggested CuAlO2
as a potential candidate for thermoelectric applications [52].
The delafossite structure belongs to quasi-two-dimensional crystal systems. CuAlO2 is
constructed by alternative stacking of Al3+ O6 octahedra conducting sheets and monovalent
Cu+ cations bonded in between. Depending on the stacking orientation, two different crystal
configurations can be identified as the rhombohedral 3R-(R3m) polytype with three layers
in one unit cell and the hexagonal 2H-(P6/mmc) polytype with two [53].
CuAlO2 is a natural p-type semiconductor with moderate band-gap energy on the order
of 1.9 eV, which is relatively small among oxides [54]. In addition, its complex layered
structure facilitates reduction of thermal conductivity. Previously, iron [55] and calcium [56]
have been used as dopants to improve electrical conductivity. In this work, magnesium was
chosen to substitute the aluminum sites. Given its similar atomic size to aluminum, higher
doping level was expected to be reached for fine control over transport properties of this
material.

4.2

Experimental Details

A series of CuAlO2 powders with various level of of magnesium doping were produced using
traditional solid-state reactions. Aluminum oxide (Al2 O3 , Baikowski, 0.35 µm), cuprous (I)
oxide (CuO, Sigma-Aldrich, <5 µm) and magnesium oxide (MgO, Matheson, Coleman and
Bell, 5 µm) powders were mixed according to nominal composition Cu1-x Mgx AlO2 (x= 0,
0.01, 0.02, 0.05, 0.075 and 0.1) with cylindrical alumina grinding media (12 mm in diameter
and length) for 12 hours. The resultant mixtures were then dried in air at 353 K for 12 hours,
spread onto rectangular alumina trays and fired in an electrical tube furnace (Carbolite CTF28

12/75/700) at 1393 K for 20 hours with a heating and cooling rate of 5 K/min. Continuous
argon (Ar) flow of approximately 10 SCFH (cubic feet per hour at standard conditions) was
applied during the entire reaction. As-fired powders were ground before consolidation.
To produce bulk samples for transport characterization, three distinct sintering techniques were used. Cold-pressing was first applied. Approximately 4 g of as-prepared powders were pressed in stainless steel dies under a pressure of 23 MPa at room temperature,
followed by 12 hour calcination under the same condition as powder preparation. A hotpressing treatment was also applied on the same powders. Samples were consolidated in air
for 3 hours in a 10 mm graphite dies. Peak pressure and temperature were 27.6 MPa and
1323 K. In pulsed electric current sintering (PECS) process, as-prepared powders were prepressed in a half inch graphite dies and pressed in flowing argon. The samples were heated
to 1473 K at a rate of 80 K/min, dwelled for 15 minutes in argon flow and naturally cooled
down to room temperature. Pressure was ramped to 60 MPa at the rate of 30 MPa/min 5
minute after the temperature began increasing, and released 1 minute after cooling started.

4.3

Results and Discussion

Structural analysis was performed on all as-fired powders, cold-pressed, hot-pressed and
PECS treated pellets. Due to the limitation of apparatus availability, transport data were
only collected from hot-pressed samples.

4.3.1

Composition and Structural Characterization

Density of bulk samples was determined using Archimedes’ principle and compared to the
theoretical density of 5.10 g/cm3 . Both hot-pressing and PECS yielded highly consolidated
29

pellets. The average relative density of hot-pressed samples is 95.9%, which is slightly lower
than the value of 98.4% of by PECS treated samples. In comparison, the maximum density
reached among pressureless sintered samples was only 65%. Density measurement results
were confirmed in SEM images shown in Figure 4.5.

XRD and phase analysis
All samples were examined by X-ray diffraction. Patterns and phase identification results are
shown in below. In all as-fired powder samples (Figure 4.1), delafossite phase was dominant
phase.

CuAl0.9Mg0.1O2

Intensities (a.u.)

CuAl0.925Mg0.075O2
CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2
CuAl0.99Mg0.01O2
CuAlO2
10

20

30

40
50
2 θ (deg)

60

70

80

Figure 4.1: XRD patterns of as-fired Cu1-x Mgx AlO2 (0% x 10%) powders

This result confirmed the reaction steps (reaction 4.1 and 4.2) suggested by Tonooka [57]:
30

spinel-type was initially formed as an intermediate before delafossite phase was achieved.

CuO + Al2 O3 −→ CuAl2 O4
CuO + CuAl2 O4 −→ 2 CuAlO2 +

(4.1)
1
O
2 2

1
Cu2+ + Ox −→ Cu1+ + h· + O2
O
2

(4.2)
(4.3)

Following the Kr¨ger-Vink notation, expression 4.3 shows the charge state changes in
o
this two-step reaction.

MgO + Al2 O3 −→ MgAl2 O4

(4.4)

1
O
2 2

(4.5)

MgO + 2 CuAlO2 −→ MgAl2 O4 + Cu2 O

(4.6)

CuO −→ Cu2 O +

However, the amount of secondary phases, Cu2 O and spinel-type CuAl2 O4 , increased as
Mg doping level rose to above 7.5%. As MgO can also react with Al2 O3 , it competed with
CuO and formed spinel phase compound MgAl2 O4 , following either reaction 4.4 and 4.5 or
4.6. In addition, the peaks of Cu2 O may come from the decomposition of residual CuO in
argon environment due to the low partial pressure (reaction 4.6).
Since cold-press treatment occurred under the same conditions as powder calcination,
their XRD patterns matched those of the corresponding precursor powders well, indicating
that the composition remained unchanged.
Conversely, the relative intensity of secondary phases notably increased after hot-pressing
in air, especially for heavily doped samples. The phase change can be viewed as the reverse
process of reaction 4.2 due to exposure to oxygen at elevated temperature. Unfortunately,
31

CuAl0.9Mg0.1O2

Intensities (a.u.)

CuAl0.925Mg0.075O2
CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2
CuAl0.99Mg0.01O2
CuAlO2
10

20

30

40
50
2 θ (deg)

60

70

80

Figure 4.2: XRD patterns of cold-pressed Cu1-x Mgx AlO2 (0% x 10%)

the near coincidence of the lattice constants of iso-structural MgAl2 O4 and CuAl2 O4 made
it difficult to determine the exact phase of spinel that exists in the samples. We can also
expect the heavily doped hot-pressed samples to have poor electrical conductivity because
spinel phases are highly electrically insulating.

PECS pressed pellets preserved and even improved the phase purity of the original powder
samples. Even in the 10% Mg doped sample, only a small amount of secondary phases was
detected. This can be attributed to the argon environment and higher temperature, which
is preferred for the formation of delafossite phase.
32

CuAl0.9Mg0.1O2

Intensities (a.u.)

CuAl0.925Mg0.075O2
CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2
CuAl0.99Mg0.01O2
CuAlO2
10

20

30

40
50
2 θ (deg)

60

70

80

Figure 4.3: XRD patterns of hot-pressed Cu1-x Mgx AlO2 (0% x 10%)

Microstructure analysis

Scanning electronic microscopy images of polished surfaces from hot-pressed and PECSsintered pure CuAlO2 , CuAl0.98 Mg0.02 O2 and CuAl0.925 Mg0.075 O2 are shown in Figure 4.5.
In hot-pressed and PECS treated samples, grains are closely interconnected, indicating very
low porosity. The contrast in color produced in gentle-beam mode reflects the difference
of average molecular weight in those areas, indicating phase separation. As the randomly
scattered insulating spinel phase in Figure 4.5(B) started to form interconnected network
in Figure 4.5(C), electrical conductivity dropped dramatically due to the structural change.
In contrast, phase uniformity was maintained in PECS-treated samples, which is consistent
with the results from x-ray diffraction analysis.
33

CuAl0.9Mg0.1O2

Intensities (a.u.)

CuAl0.925Mg0.075O2
CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2
CuAl0.99Mg0.01O2
CuAlO2
10

20

30

40
50
2 θ (deg)

60

70

80

Figure 4.4: XRD patterns of PECS treated Cu1-x Mgx AlO2 (0% x 10%)

4.3.2

Transport Property Measurement

High temperature electrical and thermal property measurements were performed on hotpressed samples. Figure 4.6 shows the electrical resistivity. The inverse proportionality
between ρ and T, and the positive value of S indicate that all samples are p-type semiconductors. The electrical resistivity first decreased as Mg content rose and then increased after
the secondary phases became significant. at 450 K, the reduction is about 60%, from 6.6
Ω·cm to 2.8 Ω·cm. The difference narrowed as the temperature increased, indicating this is
the doping effect.
In Figure 4.7, Seebeck coefficient displayed a similar trend. Power factor was calculated
from ρ and S ; The 1% and 2% doped samples have larger power factor than undoped CuAlO2 ,
while 5% doped one was lower due to poor electrical conductivity. The 7.5% and 10% doped
34

Figure 4.5: SEM images (gentle-beam mode) of hot-pressed CuAlO2 (A), hot-pressed
CuAl0.98 Mg0.02 O2 (B), hot-pressed CuAl0.925 Mg0.075 O2 (C), PECS-sintered CuAlO2 (D),
PECS-sintered CuAl0.98 Mg0.02 O2 (E) and PECS-sintered CuAl0.925 Mg0.075 O2 (F).

35

7

CuAl0.95Mg0.05O2

6
CuAl0.98Mg0.02O2

ρ (Ω ⋅ cm)

5

CuAl0.99Mg0.01O2
4

CuAlO2

3
2
1
0
400

450

500

550

600

650

700

750

Temperature (K)

Figure 4.6: Electrical resistivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)

samples could not be tested due to high electrical resistivity.

Thermal conductivity and ZT of some of the hot-pressed samples were plotted in Figure
4.10. At all temperatures, thermal conductivity decreased as Mg content went up. Figureof-merit of all samples increased with temperature. 2% Mg doped sample has the largest
ZT value at all temperature, followed by 1% doped and pure CuAlO2 . The 5% doped
sample has a worse performance due to the existence of secondary phases and poor electrical
conductivity. It can be expected that Mg doping can improve the thermoelectric performance
of delafossite-type CuAlO2 as long as sintering conditions are controlled to minimize the
formation of secondary phases.
36

800

CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2

S (μV • K-1)

700

CuAl0.99Mg0.01O2
CuAlO2

600

500

400

450

500

550

600

650

700

750

Temperature (K)

Figure 4.7: Seebeck coefficient of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)

4.4

Summary and Future Work

In this work, CuAlO2 was doped with Mg to substitute for the Al to improve its electrical
conductivity and reduce its thermal conductivity. Powder samples were prepared using
ball-milling and solid-state reaction. Cold-pressing, hot-pressing and PECS were used to
successfully fabricate bulk samples. X-ray diffraction and SEM results shows that inert gas
environment is preferred for pure delafossite phase formation. High temperature transport
properties were measured, which reveals that Mg doping can increase power factor and reduce
thermal conductivity. The formation of competing phases, primarily spinel phase, greatly
reduced the electrical conductivity.

37

Power Factor (10-5 W m-1 K-2)

10

CuAl0.95Mg0.05O2
CuAl0.98Mg0.02O2

8

CuAl0.99Mg0.01O2

6

CuAlO2

4

2

0
400

450

500

550

600

650

700

750

Temperature (K)

Figure 4.8: Power factor of of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)

0.15

CuAl0.99Mg0.01O2

CuAl0.95Mg0.05O2

0.10

CuAl0.98Mg0.02O2

CuAl0.925Mg0.075O2
α (cm2 • s-1)

CuAl0.9Mg0.1O2

CuAlO2

0.05

0.00
300

400

500

600

700

Temperature (K)

Figure 4.9: Thermal diffusivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)
38

0.6
CuAl0.98Mg0.02O2

0.5
κ (W ⋅ cm-1 ⋅ K-1)

CuAl0.9Mg0.1O2
CuAl0.925Mg0.075O2

CuAl0.99Mg0.01O2

0.4

CuAl0.95Mg0.05O2

CuAlO2

0.3
0.2
0.1
0.0
300

400

500

600

700

Temperature (K)

Figure 4.10: Thermal conductivity of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)

0.006
CuAl0.95Mg0.05O2
0.005

CuAl0.98Mg0.02O2

ZT

0.004

CuAl0.99Mg0.01O2

0.003

CuAlO2

0.002
0.001
0.000
400

450

500

550

600

650

700

750

Temperature (K)

Figure 4.11: ZT of hot-pressed Cu1-x Mgx AlO2 (0% x 5%)
39

Chapter 5
Sodium Cobalt Oxide
Nax CoO2 has long been considered as a competitive candidate for thermoelectric applications
due to its misfit structure. In this chapter, we took different approaches to investigate this
material system, expand its field of applications and better understand its thermoelectric
properties. Solid-state reaction, wet-chemical and electrochemical techniques were used to
facilitate this work.

5.1

Background and Motivation

Sodium cobalt oxide (Nax CoO2 ) is an exceptionally versatile materials system that display
interesting physical properties, due to its special hybrid crystal structure involving both
coordinated and disordered nanoblocks. NaCo2 O4 was firstly synthesized in the 1970s [58].
It was then investigated for electrodes in Na batteries [59]. Superconductivity was also
discovered near 4.5 K in Na0.35 CoO2 ·1.3H2 O [60]. It even has insulator phases [9].
The basic frame of Nax CoO2 is constructed of tilted CoO6 octahedra sheets, with charge
balancing sodium ions randomly distributed between the layers, as shown in Figure 5.1. The
40

the-art thermoelectric material Bi2 Te3 [2].
Fig. 2(b) shows the magnetic susceptibility w and
the specific heat C of a polycrystalline sample of
NaCo2 O4 [7,9]. The susceptibility is relatively large

Figure 5.1: Crystal structure of Nax CoO2 [8]
Fig. 1. Crystal structure of NaCo2 O4 :

evaluated by the va
approaches 35–50 m
orders of magnitud
conventional metals.
clearly indicate that t
the effective mass)
NaCo2 O4 :
Singh [11] calcula
NaCo2 O4 ; and pred
According to his calc
composed of Co 3d
spread along the ou
large Fermi surface
a1g þ eg bands spread
make small hole p
density of states of th

concentration of sodium ions determines not only the lattice spacing between layers, but also
the charge state of Co (Co3+ and Co4+ ). Due to the weak Van der Waals type bonding
between Na+ and CoO2 , sodium can be easily inserted or extracted from the structure and
a series of compounds with various sodium concentration can be produced.
Depending on the stacking manner of CoO6 layers and sodium concentration x, these compounds were classified into four major phases [61], as shown in Figure 5.2. The γ-Nax CoO2
(0.55 x 0.74) with hexagonal structure contains two layers in one unit cell. Three-layered
phases include α-Nax CoO2 (0.9 x 1.0) with rhombohedral structure, α -Nax CoO2 (x=1.0)
with monoclinic structure and β-Nax CoO2 (0.55 x 0.6) with orthorhombic structure. The
electrical charge state in the CoO6 layer is controlled by the sodium concentration. Transport properties show strong dependence on the sensitive crystal structure change in the
compound. Therefore these four thermodynamically stable phases render distinct structural
and electronic properties.
Starting from late 1990s, Nax CoO2 has been considered as a promising TE candidate
due to its large Seebeck coefficient and high electrical conductivity. In most cases, compounds that display metallic conduction behavior usually render poor thermopower. How41

60

Temperature	
  (K)

50
Paramagnetic
metal

40

Curie-Weiss metal

Charge oerdered
insulator

30
20
10

Magnetic
Order

H2O intercalated
Superconductor

0
0

0.1

0.2

0.3

0.4
0.5
0.6
Na	
  Content	
  x

0.7

0.8

0.9

1

Figure 5.2: Phase diagram of Nax CoO2 ; reproduced from [9]
ever, Na0.5 CoO2 seemed to violate this trend by showing low resistivity and large Seebeck
coefficient. Although an explicit explanation remains unclear, a combined effect of effective
mass enhancement and electron entropy change is speculated to simultaneously contribute
to this phenomenon.
Sodium-rich (0.74<x<1.0) γ-Nax CoO2 samples, however, did not receive much attention
until the discovery of unusually large Seebeck coefficient in the region at around 100 K [62].
Part of the reason lies in the difficulty in sample preparation, since Na ions can easily
escape due to their high volatility. In 2006, using a wet-chemical intercalation technique,
Lee and et al. successfully prepared single crystal sodium cobalt oxides, with x ranging from
0.75 to 1.0. Transport characterization revealed a 40-fold enhancement of S in Na0.88 CoO2
while electrical resistivity remained low. The calculated ZT exceeded 0.13 at 100 K. Thus,
42

sodium-rich Nax CoO2 was suggested as a potential material for cryogenic Peltier cooling.
As was mentioned in previous sections, Nax CoO2 (0.85<x<1.0) is a mixture consisting
both Na0.85 CoO2 and Na1.0 CoO2 phases, which exhibit distinct molecular configuration and
physical properties. The alleged x value indicates the ratio between the two different phases.
As the valence of Co is determined by the amount of balancing Na ions, Co possesses both
3+ and 4+ valence status in Na0.85 CoO2 , and solely 3+ in Na1.0 CoO2 . The diversity of Co
charge state and complexity of crystalline field in this mixed region are suggested to give rise
to the enhanced thermoelectric performance, which makes this materials a very interesting
system.
Thin flakes of single crystal sodium cobalt oxide have been prepared in sodium chloride
(NaCl) flux [63]. Larger crystals are reported to grow using the floating-zone technique.
Polycrystalline samples of various sodium concentration were also synthesized using polymerized method [64] and most commonly, traditional solid-state reaction.

5.2

Wet-chemical Intercalation Study

Our wet-chemcial intercalation study was inspired by both low temperature Peltier cooling
device applications and exceptionally high Seebeck coefficient reported in sodium-rich single
crystal Nax CoO2 . The application of thermoelectric materials in cryogenic refrigerators was
proposed by Ioffe in 1957 [65]. Heat absorption can be produced by applying electricity
on a TE module in a certain direction. Compared to conventional fluid-based compressors,
solid-state cooling devices provide smaller size, better stability and lower operation range
(10 K - 150 K). However, low efficiency is still the limiting factor. Lee’s discovery greatly
inspired us to use this materials for potential cryogenic Peltier cooling application [62].
43

5.2.1

Experimental Details

Powder Sample Synthesis
Two stable compositions of Nax CoO2 powders, γ-phase Na0.74 CoO2 , were synthesized using solid-state reaction.

For the former, high purity sodium carbonate (Na2 CO3 ) and

cobalt(II,III) oxide (Co3 O4 ) powders were weighed at a nominal Na/Co molar ratio of
x=0.74, 0.81 and 0.9, with an additional 10% of Na2 CO3 (x’=0.8, 0.9 and 1.0) in order
to compensate the loss of sodium during reactions. The powders were blended in polyethylene jars with alumina grinding media in air. The resultant homogenous mixtures were then
fired at 1133 K for 12 hours with a heating rate of 10 K/min.

Wet-chemical Intercalation
To increase Na concentration, 2 g of sodium metal was ground and dissolved in 75 mL of
tetrahydrofuran (THF) in a pressure tube. Na0.74 CoO2 powders were placed on a fritted
disk and inserted into the tube, which is sealed tightly and placed in a flask heated to 323 K.
Benzophenone was then added to form a blue solution. The reaction is typically carried out
at 352 K for 4 days. After 4 days of reaction, treated powders are washed with acetonitrile
and dried. All powders and bulk samples are stored in evacuated desiccators.

Bulk Sample Fabrication
PECS was used to prepare Na0.74 CoO2 and Na1.0 CoO2 . Approximately 3 g of each of asfired powders was ball-milled and pre-pressed into a 12.7 mm graphite die. In a chamber
with flowing argon, powders were heated to 1073 K at a rate of 80 K/min, dwelled for 15
minutes and cooled down to room temperature at a rate of 10 K/min.
44

5.2.2

Composition and Structural Analysis

Intensities (a.u.)

NaxCoO2 x=0.8
PECS

NaxCoO2 x=1.0
powder
NaxCoO2 x=0.9
powder

NaxCoO2 x=0.8
powder
15

25

35

45

55

65

75

2 θ (deg)

Figure 5.3: XRD patterns of Nax CoO2 powders with different nominal starting sodium
concentration and PECS-sintered Nax CoO2 (x=0.8)
Figure 5.3 shows the XRD pattern of as-fired Nax CoO2 powders and PECS treated
Nax CoO2 (x’=0.8) pellet. Here x’ indicates the initial Na/Co ratio in reactant mixtures,
which is 10% more than target sodium concentration x. All three powder samples exhibited
almost identical pattern corresponding to Na0.74 CoO2 despite the different starting values,
which indicated that Na0.74 CoO2 is a thermodynamically stable phase above the synthesis
temperature used. PECS treated sample also shows the sample pattern without shift or phase
change, meaning that the composition of the sample was preserved during consolidation.
In Figure 5.4, the lower curve shows the XRD pattern of wet-chemical intercalated powder
samples with x≈1.0 according to results of ICP analysis. However, it clearly consists two
different phases, both Na0.74 CoO2 and Na1.0 CoO2 , indicating mixed phases existing in the
45

Intensities (a.u.)

Na1.0CoO2
PECS

Na1.0CoO2
powder
15

25

35

45

55

65

75

2 θ (deg)

Figure 5.4: XRD patterns of Nax CoO2 (x=1.0) as-fired powders and PECS treated pellets

samples. The higher pattern belongs to the bulk sample after powders were sintered using
PECS. As is shown, no visible phase change can be identified, indicating that the Na1.0 CoO2
phase is also stable even after high-temperature treatment. The change of relative intensity
between peaks can be attributed to the realignment of preferred orientation under pressure.
Preferred crystal orientation induced during sintering process is a well-known issue in
polycrystalline materials. Due to the strong anisotropic characteristics in layered oxides,
transport properties of bulk samples can be different along or perpendicular to the compressing directions, if strong orientation is present. In our study, some samples were measured
along different directions but did not show significant difference in electrical and thermal
transportation properties. Therefore, we assume that the sintering condition we applied to
the powder samples did not induced significant texturing effect in the bulk samples. Similar
46

results were shown in the Ca3 Co4 O9 system, which will be discussed in the following chapter.

5.2.3

Transport Property Characterization

Two PECS treated bulk samples with different measured x values (x=0.74 and 1.0) were cut
into bars for low-temperature transport measurement. Their electrical/thermal conductivity
and Seebeck coefficient were measured to compare with reported data of single crystal SNax CoO2 with various sodium content from [62].

1000
1
2
3
4
5
6
7
8
9
10
A
B

ρ (mΩ ⋅ cm)

100

10

1

0.1
50

100

150

200

250

300

350

Temperature (K)

Figure 5.5: Electrical resistivity of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical intercalated and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 : (1) x≈0.71, (2) x≈0.75,
(3) x≈0.80, (4) x≈0.85, (5) x≈0.88, (6) x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10)
x≈1.0

Figure 5.5 shows the electrical resisitivity of PECS sintered Na0.74 CoO2 and wet-chemical
intercalated Na1.0 CoO2 with reference results. For all samples, ρ generally rises with in47

crease of sodium concentration. During the entire measuring range, the resistivity of our
Na0.74 CoO2 polycrystalline sample is in the vicinity of 1 Ω·cm with increasing trend versus temperature. This result matched the values in their single crystal counterpart, further
confirming the correct sodium concentration. The situation is similar for Na1.0 CoO2 . Its
resistivity curve is very close to Na0.99 CoO2 reported single crystal sample, indicating that
our Na1.0 CoO2 sample could contain small amount of Na0.74 CoO2 , which is verified later by
Seebeck coefficient and thermal conductivity. As was discussed previously, the upper limit of
single phase Nax CoO2 is x≈0.85. Above that threshold, the resistivity goes up very rapidly.
Between x=0.85 and x=1.0, samples are always two phase mixtures: conducting Nax CoO2
(x 0.85) and highly insulating Na1.0 CoO2 phase, in which all cobalt atoms are Co3+ .
Seebeck coefficient are plotted in Figure 5.6. In the reference samples, when x starts
to exceed 0.8, dramatically increased S with sodium content can be observed. In these
enhanced sample, S peaks at around 80 K. The largest S was found in S-Na0.99 CoO2 , which
exceeded 350 µV/K. In our Na0.74 CoO2 sample, the S value was consistent with reported
data, displaying in increase curve laying between reported Na0.71 CoO2 and Na0.75 CoO2 . In
contrast, Na1.0 CoO2 exhibited the largest value of 370 µV/K at around 100 K, beating single
crystal Na0.97 CoO2 . This increase is considered to be contributed by existence of insulating
Na1.0 CoO2 phase.
The thermal conductivity changed dramatically with increasing sodium concentration,
as shown in Figure 5.7. The thermal conductivity of our Na0.74 CoO2 sample displayed
a decreasing trend versus temperature. Its κ started a slightly higher (13 W/cmK) than
corresponding single crystal samples at 80 K, but rapidly reduced to less than 6 W/cm at
room temperature. On the other hand, ”fully” doped Na1.0 CoO2 possesses a much higher
48

400
1
2
3

S (μV ⋅ K-1)

300

4
5
6

200

7
8
10

100

A
B
0

50

100

150

200

250

300

350

Temperature (K)

Figure 5.6: Seebeck coefficient of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical intercalated and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 : (1) x≈0.71, (2) x≈0.75, (3)
x≈0.80, (4) x≈0.85, (5) x≈0.88, (6) x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10)
x≈1.0

thermal conductivity than partially doped counterparts (Nax CoO2 , x<1.0). The curve of
our Na1.0 CoO2 sample is about 50 W/cmK and quite flat with temperature change. As all
the sodium sites are occupied, the layer in the compound loses its irregularity and becomes
more like a well bonded system. The phonon scattering mechanism originating from the
randomly distributed sodium ions disappears. As the average x exceeded 0.85, the poorly
thermally conductive system is gradually added with highly conductive ”impurities”, giving
rise to the abruptly increased thermal conductivity.
Calculated ZT was also plotted in Figure 5.8. Although we successfully increased the
sodium concentration in Nax CoO2 powders and produced bulk samples without causing
49

250
200
3

κ (W ⋅ cm-1 ⋅ K-1)

150

4

100

5

50

7

20

8
10

15

A
10
5

B

50

100

150

200

250

300

350

Temperature (K)

Figure 5.7: Thermal conductivity of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical intercalated and PECS treated Na1.0 CoO2 and reported S-Nax CoO2 : (1) x≈0.71, (2) x≈0.75,
(3) x≈0.80, (4) x≈0.85, (5) x≈0.88, (6) x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10)
x≈1.0

phase changing, neither of our as-fired sample or intercalated samples reached desired ZT
as the sodium concentration was not tuned to the optimized region, which is in the vicinity
of x=0.85.
In this part of our work, we demonstrated that using the organic solvent based wetchemical intercalation technique, we can successfully increase the sodium concentration from
a thermodynamically stable phase Na0.74 CoO2 . The modified product exhibited altered
thermoelectric performance. However, we lack control over inserted amount of sodium using
this method. As the Seebeck coefficient is extremely sensitive to the sodium concentration,
precise intercalation control is desired.
50

0.15
1
3
4

0.10

5

ZT

6
7

0.05

8
A
B

0.00
50

100

150

200

250

300

350

Temperature (K)

Figure 5.8: ZT of (A) PECS treated Na0.74 CoO2 , (B) wet-chemical intercalated and PECS
treated Na1.0 CoO2 and reported S-Nax CoO2 : (1) x≈0.71, (2) x≈0.75, (3) x≈0.80, (4)
x≈0.85, (5) x≈0.88, (6) x≈0.89, (7)x≈0.96, (8) x≈0.97, (9) x≈0.99 and (10) x≈1.0

5.3
5.3.1

Electrochemical Intercalation Study
Experimental Details

Battery Construction

The first attempt of battery construction is shown in Figure 5.9. As-fired Na0.74 CoO2 powders were ball-milled, dispensed in propylene carbonate along with polytetrafluoroethylene
as the binder. The thick mixture was then cast on to a thin aluminum foil. After drying, this
layer served as the working electrode. The counter electrode, which serves as the source of
Na ion, is made by applying a thin pure sodium metal layer on the copper sheet. As sodium
51

Copper
foil
Aluminum
board
PE 

Sodium
metal

Na+

e-

Na0.74CoO2
Aluminu
m foil

Figure 5.9: Original battery setup using cast Na0.74 CoO2 powders and sodium metal
is extremely reactive, the metal was carefully handled in a glovebox. The two electrode
layers were separated by a piece of glass fiber filter, which allows ions to travel through without causing shorting. To compress the electrodes together for good contacts, two aluminum
bricks with screws were used. Between the electrodes and aluminum, insulating Teflon blocks
were used. The entire assembly was immersed in the electrolyte, which consisted of 1 mol/L
sodium perchlorate propylene carbonate solution. Electrodes were connected to a computer
controlled power supply to control electrical current and voltage.

Counter
electrode

Reference 
electrode

Working 
electrode

Figure 5.10: Improved T-cell battery setup using Na0.74 CoO2 as both working and counter
electrodes
52

Due to the high reactivity of sodium metal, we improved the battery setup by adopting
a sealed t-cell module, as shown in Figure 5.10. In this simpler setup, both target and
source of sodium are Na0.74 CoO2 pellets. The key issue to successfully build this device
is to control the density of this pellet. While needing to maintain mechanical strength for
battery assembly, the pellet has to contain sufficient porosity to allow sodium ion transport.
Starting with as-fired Na0.74 CoO2 powders, we hot-pressed them in an 1/4” graphite die at
750 ◦ C for 15 mins. The thickness of sintered pellets was well controlled to achieve desired
density. According to calculation, pellets with approximately 70% of theoretical density was
successfully produced, which was robust enough to survive the battery construction without
crushing.

Results

The obtained discharging curve is shown in Figure 5.11.
The the system started at open circuit voltage of approximately 2.55 V, which corresponded to the x=0.74 phase. The discharge stopped at approximately 2.40 V, which represented Na0.76 CoO2 . After the intercalation, pellets were taken out, crushed and washed
with ethanol. Dried powders were examined by ICP, which showed sodium concentration
change from x=0.74 to x=0.76. This value is consistent with the stopping voltage during
electrochemical intercalation. The XRD pattern in Figure 5.12 also qualitatively confirmed
successful phase change. The primary peak shifted towards higher angle, indicating shrunk
lattice and reduced layer spacing caused by enhanced attraction between CoO2 layers and
increased sodium ions.
53

4
3.8
3.6

Voltage V

3.4
3.2
3

~2.55 V

2.8
2.6
2.4
2.2
2
0.45

~2.40 V
0.55

0.65

0.75

0.85

Na concentration x

Figure 5.11: Starting and stopping voltage plotted on a complete discharge curve, reproduced
from [10]

5.3.2

Proposed in-situ Seebeck coefficient Measuring System

In 2011, Berthelot et al. conducted an interesting electrochemical investigation of Nax CoO2
(0.5 x 1.0). In this in-situ study, the shift of a single peak in the Nax CoO2 X-ray diffraction pattern was tracked while external sodium ions were inserted to the compound under
electrical driving force [10]. This simultaneous modification/characterization experiment
inspired us to track the Seebeck coefficient change during electrochemical intercalation

An in-situ electrical conductivity and Seebeck coefficient measuring system was designed,
as shown in Figure 5.13. Using this setup, we will be able to study the Seebeck coefficient
change as a function of intercalation.
54

Intensity (arb. units)

(b)

Before EC
After EC

15

Intensity (arb. units)

*	
  

15.5

16

16.5

17

(a)

Before EC
After EC

15

25

35

45

55

65

75

2 theta (degree)

Figure 5.12: (a) XRD patterns of Na0.74 CoO2 before (blue curve) and after (red curve)
electrochemical intercalation. (b) shift of the primary peak towards the higher angle

5.4

Summary and Future Work

In this study, we used both wet-chemical and electrochemical intercalation technique to
increase the sodium concentration in Nax CoO2 , in order to improve the low-temperature TE
performance. Wet-chemical treatment successfully increased the sodium content in powder
samples. Electrochemical route was demonstrated to be more precise in control the amount
of sodium to be doped into the compounds. An in-situ Seebeck coefficient measuring system
55

Reference
electrode

Thermocouple

Counter
electrode

Working
electrode
Heater

Figure 5.13: Proposed in-situ Seebeck coefficient measuring system
is proposed to simultaneously study the structure-transport relation at various temperature.

56

Chapter 6
Calcium Cobalt Oxide
The improvement of Nax CoO2 though sodium concentration control is practically limited by
its thermal stability. Due to the volatility of sodium, ideal composition can be difficult to
obtain and stoichiometry can deviate after cycling. More stable materials with comparable
performance are strongly desired. The mis-fit compound calcium cobaltite, consisting of a
similar CoO2 framework, has shown better stability. In this chapter, we report a newly
developed sol-gel method to synthesize Ca3 Co4 O9 . This wet-chemical technique is also
proven to be more effective in doping in the Ca-Co-O system, in comparison to traditional
solid-state reaction. Ytterbium substituted powders and bulk samples were produced and
analyzed for structural and transport properties.

6.1

Background and Motivation

Like most misfit-structured compounds, calcium cobalt oxide is constructed of stacked CoO2
sheets. Instead of randomly arranged Na+ ions, an insulating Ca2 CoO3 sub-lattice is present
between the conducting CoO2 sheets. Depending on the Ca/Co ratio and sub-lattice com57

position, Ca-Co-O system possesses multiple phases with respective thermally stable ranges.
Ca3 Co4 O9 and Ca2 Co2 O5 are more stable at lower temperature. They are actually two
variations of the same structure, except that Ca3 Co4 O9 has vacancies in 25% of the calcium
sites [66]. In comparison, Ca3 Co2 O6 oxide system is the major phase above 1200 K. Woermann pointed out the cobalt tends to be present in the divalent state at high temperature
while trivalent cobalt starts to appear below the 1200 K threshold [67].

Successful preparation of single crystal and polycrystalline samples have been extensively
reported [43] [68]. Solid-state reaction and sol-gel method involving nitric acid (HNO3 ) or
citric acid (C6 H8 O7 ) were two commonly used approaches. As the electrical conductivity is
primarily contributed by the CoO2 layer, doping is expected to partially change the charge
state of Co atoms from Co4+ to Co3+ in CoO2 layers and thereby improve the compound’s
conductivity. Doping on either Ca sites or Co sites has been investigated and has demonstrated the ability to change the concentration and mobility of the compound’s electrical
carriers, some of which resulted in improvements in the power factor. Dopants that have
been reported include Bi [69] [70], Y [68], Na [71], Eu [72] and Sr [73]. Double doping has
also been reported, such as Na/Bi [69] and Ag/Ba [74].

Based on previous research, we have developed a new sol-gel synthesis method to prepare polycrystalline Ca3 Co4 O9 sample. With the help of this effective technique, rare-earth
element ytterbium was selected to dope this compound, with the expectation to improve the
transport properties in this material system.
58

6.2

Experimental Details

A matrix study of calcination temperature (ranging from 700 ◦ C to 950 ◦ C) for the solgel method was finished. As solid-state synthesis has been widely used in preparing calcium
cobalt oxide, a detailed study of preparation temperature and time of solid-state reaction was
carried out parallel to the new sol-gel method, in order to compare the effect and condition
between these two approaches.

Solid-state synthesis
The initial attempt of synthesis was via traditional solid-state reaction. Undoped Ca3 Co4 O9
powders were prepared following the similar procedures as those used for sodium cobalt
oxide. High purity calcium carbonate (CaCO3 ) and cobalt(II,III) oxide (Co3 O4 ) were mixed
at the Ca/Co molar ratio of 1:1. The powder mixtures were then heated up in air to
desired temperature (700◦ C - 950◦ C) with a ramping rate of 20 K/min and calcined at peak
temperature in air for 4 hours, followed by furnace cooling. To produce partially Yb doped
samples ((Ca1-x Ybx )3 Co4 O9 ), Yb2 O3 was added to replace CaCO3 according to desired
stoichiometry. The remaining preparation steps were identical to pure samples.

Sol-gel Synthesis
An agarose-templated formula was developed using agarose ([C24 H28 O10 (OH)8 ]n , SigmaAldrich, Figure 6.1), calcium acetate monohydrate (Ca(CH3 CO2 )2 ·H2 O, Sigma-Aldrich,
M=176.18 g/mol) and cobalt(II) acetate tetrahydrate (Co(CH3 CO2 )2 ·4H2 O, Sigma-Aldrich,
M=249.08 g/mol). Ytterbium acetate (Yb(CH3 CO2 )3 , Sigma-Aldrich, M= 394.08 g/mol)
was used as the source of dopant. All reactants were weighted according to desired molar
59

HO
HO
O OH

HO

O OH
O
O

O

O
O

HO

O

O

OH

O
OH
n

Figure 6.1: Chemical formula of Agarose

ratio of (Ca1-x Ybx )3 Co4 O9 (0 x 0.15).

For one batch of gel, 3 g of reactant powder mixture was dissolved in 15 g of 5% agarose
water solution. In practice, powders are first dissolved in 14.25 g of water to form completely
clear solution. 0.75 g agarose is thereafter added to the solvent, since agarose has very limited
solubility at room temperature. The resulting emulsion was then well shaken and heated
until bubbling using a microwave oven with intermittent de-gas to help dissolve agarose.
When the suspension turned into clear yet thick solution, it was transferred to a refrigerator
to form the gel. A successful precursor should be homogeneous and transparent without
visible agglomerate.

As formed gels were sliced into small pieces and dried in flowing air for 48 hours to remove
moisture. The dried xerogels (Figure 6.2-B) were then pre-calcined at 400 ◦ C to burn out the
organic templates. The foaming products were then ball-milled into fine powders and coldpressed for calcination. Pellets were calcined under the same conditions as in the solid-state
reaction process.
60

(A)

(B)

(C)

(D)

Figure 6.2: Sol-gel synthesis of Ca3 Co4 O9 : (A) solution, (B) gel, (C) calcined powders, (D)
sintered pellet

Consolidation

Due to the availability of equipment, both PECS and hot-press were employed to produce
bulk samples in this research. PECS was initially used on powders made with solid-state
reactions. Three g of powders were ball-milled and pre-pressed into a 12.7 mm graphite
die, and then transferred to argon-filled chamber. Powders were heated to 1153 K (set
temperature) at a rate of 80 K/min, dwelled for 15 minutes under pressure of 60 MPa and
cooled to room temperature at a rate of 10 K/min. Argon flow was applied during the entire
process. An additional air-annealing process at 850 ◦ C was applied after PECS treatment
in order to ensure correct composition, which will be discussed in the results part.
61

Hot-press was applied to powders obtained via the sol-gel route. As-fired powders were
also ball-milled and pressed in a 12.7 mm graphite die. Samples were heating up to 870 ◦ C
(set value) at a rate of 15 K/min in argon environment. At the peak temperature, samples
were held under pressure of 75 MPa for an hour, followed by furnace cooling and additional
air-annealing process.

6.3
6.3.1

Results and Discussion
Calcination Condition and Doping Study

Calcination conditions
Reactant precursors prepared via both solid-state reactions and sol-gel synthesis were calcined at 700 ◦ C, 750 ◦ C, 800 ◦ C, 850 ◦ and 900 ◦ C, to compare the activation temperature
over the products phase formation. As-fired products were pulverized into powders and
examined using XRD.
Figure 6.3 shows the XRD patterns of SSR samples. At 700 ◦ C, the reaction can be
triggered to start forming Ca3 Co4 O9 phase. At 750 ◦ C, the reactions were much more
effective, with only small amount of unreacted Co3 O4 that can be observed. At 800 ◦ C,
phase pure Ca3 Co4 O9 can be obtained and the pattern remained clean as the temperature
went up to 900 ◦ C. An extended calcination of 12 hours at 750 ◦ C was later carried out,
small amount of impurity phases can still be detected, indicating that 800 ◦ C is the lowest
temperature required for complete formation via solid-state reaction.
The same temperature dependent XRD patterns of sol-gel samples were shown in Figure
6.4. At 700 ◦ C, impurity phases can still be seen. However, phase pure Ca3 Co4 O9 samples
62

900C

Intensities (a.u.)

850C

800C

750C

*

*

*

*
700C
10

*
20

*

30

40

50

60

2 θ (deg)

Figure 6.3: XRD pattern of solid-state reaction Ca3 Co4 O9 powders fire at various temperature. Asterisks indicate impurity peaks identified as unreacted Co3 O4

can be obtained at as low as 750 ◦ C, without detectable peaks from the reactant residues.
In another words, the minimum required reaction temperature of the sol-gel methods is 50
◦C

lower than that of SSR. Compared to the micron-level particle-particle interface, sol-gel

method can produce much finer precursor particles and more surface area, which requires
to lower reaction energy to start the reaction. The closer distance between reactants also
reduced the need of mass diffusion. We also tried to increase the firing temperature to
950 ◦ C, which resulted in dramatic phase change. Ca3 Co4 O9 decomposed into CoO and
Ca3 Co2 O6 .
Although we set the peak temperature to be lower than 900 ◦ C during consolidation
processes, both in PECS and hot-pressed samples, we still observed phases change when
63

Intensities (a.u.)

950C
900C
850C
800C
750C
700C
10

*

*
20

30
40
2 θ (deg)

*
50

60

Figure 6.4: XRD pattern of sol-gel synthesis Ca3 Co4 O9 powders fire at various temperature.
Asterisks indicate impurity peaks identified as unreacted Co3 O4

scanning the sintered pellets. As is plotted in Figure 6.5, phase pure Ca3 Co4 O9 powders
partially decomposed to Ca3 Co2 O6 after PECS treatment. One possible reason is in accuracy
of temperature monitoring during the process. The actual temperature where the samples
locate can be higher for as much as 50 K than set value as the thermocouple is placed in
the die yet not in the core. Another possibility is that the oxygen deficient environment
in the either PECS or hot-press can play a role in changing the preferred phase formation.
Therefore, an additional air-annealing processes was applied to pellet sample, which restored
the Ca3 Co4 O9 phase in the samples.
64

Intensities (a.u.)

Annealed

PECS treated

As-fired
10

20

30

40

50

60

2 θ (deg)

Figure 6.5: XRD pattern of as-fired, PECS treated and annealed Ca3 Co4 O9 powders

Doping study

In order to modify the material to increase power factor and reduce thermal conductivity,
ytterbium was selected to substitute for calcium. Using solid-state reaction, 2% and 5%
doped samples were fired at 850 ◦ C and 900 ◦ C. All four samples patterns are plotted
along with undoped Ca3 Co4 O9 in Figure 6.6. At 850 ◦ C, Ca3 Co4 O9 and Yb2 O3 phases
can be observed, indicating that Yb2 O3 remained as a separate phase and failed to react
with CaCO3 and Co3 O4 . Interestingly, in both 900 ◦ C samples, Ca3 Co2 O6 phase showed
strong peaks, which only appeared when the firing temperature was over 900 ◦ C in undoped
samples. These results demonstrate that using oxides of rare-earth elements in solid-state
reaction cannot achieve effective doping in this system.
65

Intensities (a.u.)

5%
900C

*
*

5%
850C
2%
900C

*

*

*
*

*

*
*
*

2%
850C

*

*

*
*

*

*

0%
850C
10

20

30

40

50

60

2 θ (deg)

Figure 6.6: XRD patterns of undoped, 2% and 5% Yb doped (Ca1-x Ybx )3 Co4 O9 fired at
850 ◦ C and 900 ◦ C via solid state reaction. Blue asterisks indicate unreacted Yb2 O3 and
red asterisks indicate Ca3 Co2 O6 phase.
In contrast, rare-earth element doping using this sol-gel method was achieved and the
X-ray diffraction results are shown in Figure 6.7. In all sample (0% x 15%), the Ca3 Co4 O9
is the primary phase, suggesting that Yb occupied the Ca sites and preserved the crystal
structure. Only starting from 8% and above do detectable impurity phases become notable.
The success of sol-gel methods is due to the improved reaction interfaces. In solid-state
reactions, the particle sizes of reactant powders are in vicinity of microns if not bigger.
The mass transportation rate and reaction activity are strongly limited by this parameter.
In comparison, sol-gel method created much finer particles and molecular level interfaces.
Reactants were first dissolved in solvents to form homogeneous solutions and then turned into
gels, where reactants were confined in porous frameworks. The greatly increased surface area
66

*

*

*

8%
Intensities (a.u.)

15%

*

*

*

4%

*

*

*

2%
1%
0%
10

20

30
40
2 θ (deg)

50

60

Figure 6.7: XRD patterns of undoped, 1%, 2%, 4%, 8% and 15% Yb doped
(Ca1-x Ybx )3 Co4 O9 fired at 850 ◦ C via sol-gel synthesis. Asterisks indicate impurity phases.
reduced the activation energy and the distanced required by mass transportation. Thus, rareearth element doping was realized, which was impossible in traditional solid-state reactions.
Also, the condition for Ca3 Co4 O9 formation is lowered, which resulted in lower reaction
temperature and shorter reaction time.

Density
The measured density values of hot-pressed (Ca1-x Ybx )3 Co4 O9 (0% x 15%) are listed in
Table 6.1. Assuming that all ytterbium atoms entered the calcium sites, the theoretical
density of doped samples are calculated by substituting the mass of calcium atoms with
that of ytterbium. All samples reached very high density (ρaverage =98.3% of theoretical
67

value). All these samples were cut into small bars for high temperature transport property
measurement.
Table 6.1: Density of hot-pressed (Ca1-x Ybx )3 Co4 O9 samples, where M, ρ0 , ρ represent
molecular weight, theoretical density and actual density.
Doping level
0%
1%
2%
5%
7.5%
10%

6.4

M (g/mol)
499.90
503.89
507.88
515.86
531.82
559.75

ρ0 (g/cm3 ) ρ (g/cm3 )
4.68
4.71
4.75
4.83
4.98
5.24

4.584
4.604
4.679
4.742
4.890
5.171

ρ/ρ0
98.0%
97.7%
98.5%
98.2%
98.3%
98.7%

Summary and Future Work

In this study, an acetate salt based agarose-templated sol-gel synthesis method has been
developed. Higher efficiency and more effective doping in the preparation of Ca3 Co4 O9 was
achieved using this new method. The structure and transport properties of partially Yb
doped samples were investigated. Considering the moderate solubility of other acetate salts,
this agarose-templated sol-gel synthesis methods can be applied to the fabrication of a series
of rare-earth element doped compounds, such as Er, Dy and Gd. Mg and Sr may also be
used to substitute Ca to form solid solution to achieve thermal conductivity reduction.

68

Chapter 7
Conclusions and Future Work
In this work, we aimed to develop new complex oxides for potential thermoelectric applications. A variety of preparation techniques and analysis methods were used to study the TE
properties of of following three material systems: copper aluminum oxide (CuAlO2 ), sodium
cobalt oxide (Nax CoO2 ) and calcium cobalt oxide (Ca3 Co4 O9 ).
CuAlO2 , a delafossite-type compound, possesses a natural layered structure and small
band gap energy, which can be favorable for high-temperature thermoelectric application. To
improve its electrical conductivity and reduce its thermal conductivity, magnesium (Mg) was
used to substitute for the aluminum site. Undoped and doped powder samples were prepared
using ball-milling and solid-state reaction in flowing argon gas environment. Cold-pressing,
hot-pressing and pulsed electric current sintering were used to fabricate bulk samples. X-ray
diffraction and SEM results shows that inert gas environment is preferred for pure delafossite
phase formation. High temperature transport properties were measured, which reveals that
Mg doping can increase power factor and reduce thermal conductivity. The formation of
competing phases, primarily spinel phase, greatly reduced the electrical conductivity.
69

The second material system we investigated was Nax CoO2 . Inspired by the exceptionally
large cryogenic power factor in sodium-rich (x>0.74) samples, we attempted to synthesize
polycrystalline powders with high sodium concentration. Starting with the traditional solidstate reaction, we adopted wet-chemical intercalation techniques to successfully increase
sodium content in powder samples from x=0.74 to x=1.0. The results were verified by both
ICP analysis and transport properties. We also applied electrochemical intercalation to
serve the same purpose. Accurate sodium concentration change was achieve by controlling
the discharge current and voltage in a modified T-cell. ICP and XRD confirmed successful
sodium insertion from Na0.74 CoO2 to Na0.76 CoO2 . These results were in good agreement
with the starting and finishing voltages during discharging. Based on these results, we
showed the feasibility and proposed a setup of using the electrochemical technique to realize
in-situ tracking of Seebeck coefficient during the intercalation process.
Last but not least, we developed a novel sol-gel synthesis method to prepare pure and
rare-earth element doped Ca3 Co4 O9 , which is more efficient than solid-state reactions and
more environment-friendly than wet-chemical methods involving nitric acid. Using agarose as
a template, the water solutions consisting of Co(Ac)2 and Ca(Ac)2 were made into clear gels.
After drying, pre-calcination and calcination, phase pure powder samples can be obtained
at lower firing temperature than that of SSR. More importantly, partially Yb doped samples
were successfully made, which was not realized in SSR involving Yb2 O3 . As-fired powders
were consolidated into bulk pellets using a hot-press, which caused phase change, probably
due to temperature fluctuation and inert gas environment. Additional air-annealing restored
the desire composition.
Even though great progress has been made in the development of these new yet promising
70

candidates, some major issues should be addressed in future study. First, the performance
of n-type materials needs improvement. The ZT of state-of-the-art n-type materials only
shows about half the value of their p-type counterparts. New materials with comparable
performance are in great demand. Second, theoretical calculation is needed for complex oxides. The electrical conduction mechanism of these strong correlated systems cannot be fully
explained by classic semiconductor theories. Deeper understanding would assist the discover
and design of better materials. Lastly, problems encountered in devices fabrications are
also waiting for improved solutions, such as large contact resistance and thermal expansion
mismatch between oxide legs and substrates.

71

APPENDICES

72

Appendix A

Iron-Vanadium-Aluminum

Ternary metallic alloy Fe2 VAl with a pseudogap in its energy band structure has received
intensive scrutiny for potential thermoelectric applications. Due to the sharp change in the
density of states profile near the Fermi level, interesting transport properties can be triggered
to render possible enhancement in the overall thermoelectric performance. Previously, this
full-Heusler type alloy was partially doped with cobalt at the iron sites to produce a series
of compounds with n-type conductivity. Their thermoelectric properties in the temperature
range of 300 K to 850 K were reported. In this research, efforts were made to extend
the investigation on (Fe(1-x) Cox )2 VAl to the low-temperature range. Alloy samples were
prepared by arc-melting and annealing. Seebeck coefficient, electrical resistivity, and thermal
conductivity measurements were performed from 80 K to room temperature. The effects of
cobalt doping on the electronic and thermal properties are discussed.
73

A.1

Background and Motivation

Iron-vanadium-aluminum (Fe2 VAl) has been recognized as a potential thermoelectric (TE)
material with large power factor [75]. This compound displays a sharp change in the density
of states near the Fermi energy level, which is believed to give rise to its large Seebeck
coefficient. It has been shown that such a deep pseudogap in the electronic band structure is
induced by the hybridization between the aluminum’s s and p states and transition-metal’s
d orbitals.
Doping has been widely used as an effective approach to fine tune the band structures
of semiconductors and semimetals. As holes and electrons simultaneously contribute to
the total conductivity in semimetals according to calculation results [4], both n-type and
p-type doping have been demonstrated to successfully reduce electrical resistivity and promote thermopower in this system. Doping using Sn, Si, and Ge at the Al site has been
reported to greatly enhance the power factor [76] [77]. Ti and Mo were also used to substitute for V atoms, which resulted in improved low-temperature ZT [78] [79]. Thermoelectric
properties of (Fe1-x Cox )2 VAl in the high-temperature range have been reported by other
research groups [80]. Herein, we present electrical and thermal transport measurement results of (Fe1-x Cox )2 VAl in the low-temperature range (80 K to 300 K), in order to better
understand the fundamental physical properties of this compound.

A.2

Experimental Details

Due to the high melting point of vanadium and iron, samples were prepared using an arcmelting technique. High-purity elements were weighed according to desired stoichiometry,
74

placed on a water-cooled copper crucible under owing argon, and melted with current-induced
electric arc. To achieve homogeneity, melted chunks were flipped over and remelted for three
times, followed by additional annealing. As-cast ingots were wrapped with graphite foil,
sealed in evacuated quartz tubes, and then soaked at 900 ◦ C for 48 h. Due to its relatively
low melting point of, aluminum is likely to vaporize during the melting process, which is
considered to be the main reason for material loss. As was suggested by Mizutani et al.,
even slight deviation from stoichiometry in the compound can lead to a dramatic change
in transport properties and thermoelectric performance [81]. Therefore, additional Al was
carefully and accordingly added between each remelting step to compensate the loss and
ensure stoichiometric composition [82].

Figure A.1: High voltage electric arc-melter

For phase identification, samples were ball-milled into fine powders and scanned by pow75

der X-ray diffraction (XRD, Rigaku MiniFlex II). A rectangular bar was cut out of each
sample using a spark wire cutter to perform transport property measurement. Electrical
conductivity (σ) and Seebeck coefficient (S) were obtained from 80 to 300 K in a steady
state cryostat system cooled by liquid nitrogen. Thermal conductivity (κ = κe + κl ) was
measured simultaneously in the same apparatus.

A.3
A.3.1

Results and Discussion
Composition and Structural Characterization

XRD and phase analysis
The fluorescence effect of cobalt produces strong background noise. To enhance the signalto-noise ratio in XRD patterns, slow scanning speed (0.1 ◦ min) was adopted to minimize
this effect. As shown in Figure A.2 (left), all sample patterns matched the reference data
of the full-Heusler structure without detectable secondary phases, suggesting that Fe was
successfully substituted by Co. On closer inspection of one high-angle peak near 81◦ in Figure
A.2 (right), a slight yet systematical shift towards lower angle can be observed, indicating
monotonically expanded lattice due to the size difference between Fe and Co atoms.

A.3.2

Transport Property Measurement

Electrical Transport Properties
Seebeck coefficient of the stoichiometric sample is positive and exhibits weak yet positive
dependence of temperature over the range of measurement, consistent with previous reports [83]. Even with small amount of Co substitution, S immediately changes sign as
76

Intensities (a.u.)

(Fe0.6Co0.4)2VAl

(Fe0.8Co0.2)2VAl

(Fe0.9Co0.1)2VAl

(Fe0.95Co0.05)2VAl
20

40

60

80

81

82

2 θ (deg)

Figure A.2: XRD patterns of (Fe1-x Cox )2 VAl (5% x 40%)
shown in Figure A.3. This transition reflects the sudden increase in the concentration of
negative carriers due to doping. For all doped samples, the absolute value of Seebeck coefficient (|S|) still grows with temperature in the range of measurement. At each temperature
point, interestingly, |S| first increases with doping level x up to 10% and then declined.
Heavily doped samples (x > 20%) exhibit even smaller |S| than the pure sample. As was
indicated in the formula of Seebeck coefficient in two-carrier systems [84], the enhancement
in thermopower is due to the optimized carrier density in the 10% Co doped sample. Going
beyond that doping concentration, Seebeck coefficient maintained a small magnitude associated with the metal-like conduction due to large electron concentration. A maximum value,
S= -97.89 µV/K, was obtained in the 10% doped sample at 300 K.
77

100

S (μV ⋅ K-1)

50

(Fe0.6Co0.4)2VAl
(Fe0.8Co0.2)2VAl

0

(Fe0.9Co0.1)2VAl
(Fe0.95Co0.05)2VAl

-50

-100
100

150

200

250

300

Temperature (K)

Figure A.3: Seebeck coefficient of (Fe1-x Cox )2 VAl (5% x 40%)

The electrical resistivity (ρ = 1/σ) in pure Fe2 VAl sample displays semiconductor-like
behavior, which modestly decreases with the increasing temperature. As each Co atom
provides one extra electron when substituting Fe, ρ is significantly suppressed by over 50%
after additional electrical carriers are introduced into the system: the higher the doping
level is, the lower the resistivity. In addition, Co substitution converted the conduction
behavior into metallic-type, as shown by the positive temperature coefficient of resistivity.
This phenomenon can be explained by the gradual transition from a two-carrier carrier
system to a single carrier system dominated by electrons.
Power factor (S 2 /ρ) represents the electrical contribution to the figure-of-merit. The
power factors of all samples, including pure Fe2 VAl, are illustrated in Figure A.5. A maximum value of 26 µW/cmK2 is obtained in the 10% Co doped sample at 300 K, due to its
78

0.0005

ρ (Ω ⋅ cm)

(Fe0.6Co0.4)2VAl
0.0004

(Fe0.8Co0.2)2VAl
(Fe0.9Co0.1)2VAl

0.0003

(Fe0.95Co0.05)2VAl

0.0002

0.0001
100

150

200

250

300

Temperature (K)

Figure A.4: Electrical resistivity of (Fe1-x Cox )2 VAl (5% x 40%)
large Seebeck coefficient and low electrical resistivity. Doping greatly increases the electron
concentration n, which reduced the electrical resistivity. S initially increased as the compensating effect and then reduced when the electron concentration exceeded the optimal
range. Thus, an optimized point is achieved when the product of these two factors reach its
maximum at x = 10%. Considering the fact that the power factor curve is still ascending,
optimized performance may be obtained at a higher temperature.

Thermal Conductivity
Total thermal conductivities (κ = κe + κl ) in pure and doped samples are plotted in Figure
A.6. With increasing cobalt content, κ is monotonically reduced because Co atoms serve
as point defects to scatter phonons. In the 40% doped sample, the reduction is over 50%
through the entire range of measurement. Electronic and lattice contribution is separated
79

Power Factor (10-6 W m-1 K-2)

30
(Fe0.6Co0.4)2VAl
(Fe0.8Co0.2)2VAl
20

(Fe0.9Co0.1)2VAl
(Fe0.95Co0.05)2VAl

10

0
100

150

200

250

300

Temperature (K)

Figure A.5: Power factor of (Fe1-x Cox )2 VAl (5% x 40%)
following the Wiedemann-Franz law (κe = L0 σT ), where L0 is the Lorentz number. As
shown in Figure A.7, κe significantly increase with doping, due to the increase in extrinsic
electron concentration. κl displays the opposite trend because the mass difference between
Fe and Co leads to an augmentation in lattice disorder. Since the κl is approximately one
order of magnitude higher than κe , the total thermal conductivity still significantly decreases.
Despite of the considerable reduction, the total thermal conductivity is still approximately
one magnitude higher than those of the skutterudite and Bi2 Te3 based materials, which is
due to the low molecular weight of atomic species and relatively simple crystal structure.

Dimensionless Figure-of-merit, ZT
Figure-of-merit (ZT ) of (Fe1-x Cox )2 VAl is computed and plotted in Figure A.8. For all
samples, ZT increases with temperature in the range of measurement. The highest value
80

0.4

(Fe0.9Co0.1)2VAl

(Fe0.8Co0.2)2VAl
κ (W ⋅ cm-1 ⋅ K-1)

(Fe0.6Co0.4)2VAl

(Fe0.95Co0.05)2VAl

0.3

0.2

100

150

200

250

300

Temperature (K)

Figure A.6: Total thermal conductivity of (Fe1-x Cox )2 VAl (5% x 40%)
of 0.034 is obtained in the 10% doped sample at 300 K. Although Co doping has doubled
the performance of the stoichiometric compounds, the efficiency of this material is still
much lower than contemporary candidates, mainly due to its high thermal conductivity.
Further reduction of thermal conductivity would be an viable method for additional ZT
enhancement. Using heavier atoms to form solid solution may further lower the lattice
thermal conductivity. Powder processing and advanced sintering techniques will also be
adopted to create fine microstructure to serve the same purpose [85] [86].

A.4

Summary

In this research, partially cobalt doped Heusler alloy (Fe1-x Cox )2 VAl was prepared with an
arc-melting technique. Single-phase samples were confirmed by XRD results in all samples.
81

0.03

(Fe0.6Co0.4)2VAl

κe (W ⋅ cm-1 ⋅ K-1)

(Fe0.8Co0.2)2VAl
(Fe0.9Co0.1)2VAl
0.02

(Fe0.95Co0.05)2VAl

0.01

100

150

200

250

300

Temperature (K)

Figure A.7: Electronic thermal conductivity of (Fe1-x Cox )2 VAl (5% x 40%)
Excessive electrons introduced by n-type doping not only significantly reduced electrical resistivity and increased Seebeck coefficient, but also succeed in reducing thermal conductivity.
In spite of the large power factor in this compound, large thermal conductivity prevents ZT
values for practical applications. Future investigation would focus on thermal conductivity
reduction though isovalent substitution with heavy atoms and powder processing to create
fine microstructure.

82

0.04
(Fe0.6Co0.4)2VAl

ZT

0.03

0.02

(Fe0.8Co0.2)2VAl
(Fe0.9Co0.1)2VAl
(Fe0.95Co0.05)2VAl

0.01

0.00
100

150

200

250

Temperature (K)

Figure A.8: ZT of of (Fe1-x Cox )2 VAl (5% x 40%)

83

300

BIBLIOGRAPHY

84

BIBLIOGRAPHY

[1] V. Smil. Energy Transitions: History, Requirements, Prospects. ABC-CLIO, LLC.,
2010.
[2] BP Statistical Review of World Energy June 2012. BP, 2012.
[3] J.G. Snyder and E.S. Toberer. Complex thermoelectric materials. Nature Materials,
7(2):105–114, 2008.
[4] M. Weinert and R.E. Watson. Hybridization-induced band gaps in transition-metal
aluminides. Physical Review B, 58(15):9732–9740, 1998.
[5] R. Weht and W.E. Pickett. Excitonic correlations in the intermetallic Fe2 VAl. Physical
Review B, 58(11):6855–6861, 1998.
[6] M.S. Dresselhaus, G. Chen, M.Y. Tang, R.G. Yang, H. Lee, D.Z. Wang, Z.F. Ren, J.P.
Fleurial, and P. Gogna. New Directions for Low-Dimensional Thermoelectric Materials.
Advanced Materials, 19(8):1043–1053, April 2007.
[7] A. Ootaki.
Development of peak performance of thermoelectric oxides.
http://www.mm.kyushu-u.ac.jp/lab-02/pages/research/research-top.html ,
September 2008.
[8] I. Terasaki. Transport properties and electronic states of the thermoelectric oxide
NaCo2 O4 . Physica B: Condensed Matter, 328(1-2):63–67, April 2003.
[9] J. Foo, Y. Wang, S. Watauchi, H. Zandbergen, T. He, R. Cava, and N. Ong. Charge
Ordering, Commensurability, and Metallicity in the Phase Diagram of the Layered
Nax CoO2 . Physical Review Letters, 92(24):247001, June 2004.
[10] R. Berthelot, D. Carlier, and C. Delmas. Electrochemical investigation of the P2Nax CoO2 phase diagram. Nature Materials, 10(1):74–80, December 2010.
85

[11] J.T. Seebeck. Abh. K. Akad. Wiss., May 1821.
[12] J.C. Peltier. Ann. Chim., May 1834.
[13] J. Yang and F.R. Stabler. Automotive Applications of Thermoelectric Materials. Journal of Electronic Materials, 38(7):1245–1251, February 2009.
[14] State of the art prototype vehicle with a thermoelectric
http://www1.eere.energy.gov/vehiclesandfuels/resources/proceedings/2012thermoelectrics-presentations.html , March 2012.

generator.

[15] H. B¨ttner, G. Chen, and R. Venkatasubramanian. Aspects of Thin-Film Superlattice
o
Thermoelectric Materials, Devices, and Applications. MRS Bulletin, 31(03):211–217,
March 2006.
[16] G.S. Nolas, J. Poon, and M. Kanatzidis. Recent developments in bulk thermoelectric
materials. MRS Bulletin, 31(3):199–205, 2006.
[17] Y. Zhou, S. Paul, and S. Bhunia. Harvesting Wasted Heat in a Microprocessor Using
Thermoelectric Generators: Modeling, Analysis and Measurement. In 2008 Design,
Automation and Test in Europe, pages 98–103. IEEE.
[18] W.H. Nam, Y.S. Lim, S.M. Choi, W.S. Seo, and J.Y. Lee. High-temperature charge
transport and thermoelectric properties of a degenerately Al-doped ZnO nanocomposite.
Journal of Materials Chemistry, 22(29):14633–14638, 2012.
[19] G.A. Slack. CRC Handbook of Thermo-electrics. August 1995.
[20] F. Ioffe, editor. Physics of Semiconductors. August 1960.
[21] I.V. Gasenkova and T.E. Svechnikova. Structural and Transport Properties of Sn-Doped
Bi 2 Te3–x Sex Single Crystals. Inorganic Materials, 40(6):570–575.
[22] I. Kudman. Thermoelectric properties of p-type PbTe-PbSe alloys. Journal of Materials
Science, 7(9):1027–1029, September 1972.
[23] S.R. Brown, S.M. Kauzlarich, F. Gascoin, and J.G. Snyder. Yb14 MnSb11 : New High
Efficiency Thermoelectric Material for Power Generation. Chemistry of Materials,
18(7):1873–1877, April 2006.
86

[24] E.J. Skoug, J.D. Cain, and D.T. Morelli. High thermoelectric figure of merit in the
Cu3 SbSe4 -Cu3 SbS4 solid solution. Applied Physics Letters, 98(26):261911, 2011.
[25] W. Braun, D.J.and Jeitschko. Preparation and structural investigations of antimonides
with the LaFe4 P12 structure. Journal of the Less Common Metals, 72(1):147–156, July
1980.
[26] G.S.S Nolas, G.A. A Slack, and S.B. Schujman. Chapter 6 - Semiconductor clathrates:
A phonon glass electron crystal material with potential for thermoelectric applications,
volume 69 of Semiconductors and Semimetals. Elsevier, 2001.
[27] J.R. Sootsman, D.Y. Chung, and M.G. Kanatzidis. New and Old Concepts in Thermoelectric Materials. Angewandte Chemie International Edition, 48(46):8616–8639,
November 2009.
[28] B. Yu, Q. Zhang, H. Wang, X. Wang, H. Wang, D. Wang, H. Wang, J.G. Snyder, G. Chen, and Z.F. Ren. Thermoelectric property studies on thallium-doped
lead telluride prepared by ball milling and hot pressing. Journal of Applied Physics,
108(1):016104, 2010.
[29] D. Cahill, S. Watson, and R. Pohl. Lower limit to the thermal conductivity of disordered
crystals. Physical Review B, 46(10):6131–6140, September 1992.
[30] N.F. Mott and H. Jones. The Theory of the Properties of Metals and Alloys . August
1958.
[31] Z.J. Pan, L.T. Zhang, and J.S. Wu. Electronic structure and transport properties of
doped CoSi single crystal. Journal of Applied Physics, 101(3):033715, 2007.
[32] H. Sun, D.T. Morelli, M.J. Kirkham, H.M. Meyer, and E. Lara-Curzio. The role of boron
segregation in enhanced thermoelectric power factor of CoSi1−x Bx alloys. Journal of
Applied Physics, 110(12):123711, 2011.
[33] D.A. Muller, S. Subramanian, P.E. Batson, J. Silcox, and S.L. Sass. Structure, chemistry and bonding at grain boundaries in Ni3 Al. The role of boron in ductilizing grain
boundaries. Acta Materialia, 44(4):1637–1645, April 1996.
[34] P.H. Xiang, Y. Kinemuchi, H. Kaga, and K. Watari. Fabrication and thermoelectric
properties of Ca3 Co4 O9 /Ag composites. Journal of Alloys and Compounds, 454(12):364–369, April 2008.
87

[35] B. Paul, A.V. Kumar, and P. Banerji. Embedded Ag-rich nanodots in PbTe: Enhancement of thermoelectric properties through energy filtering of the carriers. Journal of
Applied Physics, 108(6):064322, 2010.
[36] J.J. Molenda, C. Delmas, and P. Hagenmuller. Electronic and electrochemical properties
of Nax CoO2−y cathode. Solid State Ionics, 9-10:431–435, December 1983.
[37] I. Terasaki, Y. Sasago, and K. Uchinokura. Large thermoelectric power in NaCo2 O4
single crystals. Physical Review B, 56(20):12685–12687, 1997.
[38] H. Yakabe, K. Kikuchi, I. Terasaki, Y. Sasago, and K. Uchinokura. Thermoelectric
properties of transition-metal oxide NaCo2 O4 system . In Proceedings of the 16th International Conference on Thermoelectrics, pages 523–527, Dresden, August 1997. IEEE.
[39] K. Koumoto, I. Terasaki, T. Kajitani, M. Ohashi, and R. Funahashi. Oxide Thermoelectrics. In Thermoelectric Handbook - Macro to Nano. August 2006.
[40] K. Fujita, T. Mochida, and K. Nakamura. High-temperature thermoelectric properties
of Nax CoO2-δ single crystals. In Proceedings of the 20th International Conference on
Thermoelectrics, pages 168–171, 2001.
[41] M. Ohtaki, Y. Nojiri, and E. Maed. In Proceedings of the 19th International Conference
on Thermoelectrics, page 190, Babrow, Wales, August 2000.
[42] M. Sopickalizer, P. Smaczynski, K. Kozlowska, E. Bobrowskagrzesik, J. Plewa, and
H. Altenburg. Preparation and characterization of calcium cobaltite for thermoelectric
application. Journal of the European Ceramic Society, 25(12):1997–2001, 2005.
[43] R. Funahashi, I. Matsubara, H. Ikuta, T. Takeuchi, U. Mizutani, and S. Sodeoka. An
oxide single crystal with high thermoelectric performance in air. Japanese Journal of
Applied Physics Part 2-Letters, 39:L1127–L1129, 2000.
[44] R. Funahashi, I. Matsubara, and S. Sodeoka. Thermoelectric properties of Bi2 Sr2 Co2 Ox
polycrystalline materials. Applied Physics Letters, 76(17):2385, 2000.
[45] W. Shin and N. Murayama. Thermoelectric properties of (Bi, Pb)-Sr-Co-O oxide. Journal of Materials Research, 15(2):382–386, February 2000.
[46] M. Ohtaki, T. Tsubota, K. Eguchi, and H. Arai. High-temperature thermoelectric
properties of (Zn1−x Alx )O. Journal of Applied Physics, 79(3):1816–1818, 1996.
88

[47] S. Ohta, T. Nomura, H. Ohta, and K. Koumoto. High-temperature carrier transport and
thermoelectric properties of heavily La- or Nb-doped SrTiO3 single crystals. Journal of
Applied Physics, 97(3):034106–034106–4, 2005.
[48] G. Garbarino, F. Morales, M. N´nez Regueiro, M. Armand, and P. Lejay. Transport and
u˜
pressure properties of Na0.5 CoO2 single crystals with metallic ground state. Europhysics
Letters, 81(4):47005, January 2008.
[49] D.G. Ebling, A. Jacquot, M. J¨gle, H. B¨ttner, U. K¨hn, and L. Kirste. Structure
a
o
u
and thermoelectric properties of nanocomposite bismuth telluride prepared by melt
spinning or by partially alloying with IV–VI compounds. Physica Status Solidi, 1(6):238,
November 2007.
[50] R.B. Song, T. Aizawa, and J.Q. Sun. Synthesis of Mg2 Si1−x Snx solid solutions as
thermoelectric materials by bulk mechanical alloying and hot pressing. Materials Science
and Engineering: B, 136(2):111–117, 2007.
[51] H. Kawazoe, M. Yasukawa, H. Hyodo, M. Kurita, H. Yanagi, and H. Hosono. P-type
electrical conduction in transparent thin films of CuAlO2 . Nature, 389(6654):939–942,
October 1997.
[52] K. Koumoto, H. Koduka, and W.S. Seo. Thermoelectric properties of single crystal
CuAlO2 with a layered structure. Journal of Materials Chemistry, 11(2):251–252, 2001.
[53] A. Buljan, P. Alemany, and E. Ruiz. Electronic structure and bonding in CuMO2 (M
= Al, Ga, Y) Delafossite-type oxides: an ab − initio study. The Journal of Physical
Chemistry B, 103(38):8060–8066, September 1999.
[54] X. Nie, S.H. Wei, and S. Zhang. Bipolar Doping and Band-Gap Anomalies in Delafossite
Transparent Conductive Oxides. Physical Review Letters, 88(6), January 2002.
[55] K. Park, K. Ko, J. Seong, and S. Nahm. Microstructure and high-temperature thermoelectric properties of polycrystalline CuAl1−x Mgx O2 ceramics. Journal of the European
Ceramic Society, 27(13-15):3735–3738, 2007.
[56] K. Park, K.Y. Ko, and W.S. Seo. Effect of partial substitution of Ca for Al on the
microstructure and high-temperature thermoelectric properties of CuAlO2 . Materials
Science and Engineering: B, 129(1-3):1–7, April 2006.
[57] K. Tonooka, K. Shimokawa, and O. Nishimura. Properties of copper-aluminum oxide
films prepared by solution methods. pages 129–133, 2002.
89

[58] M. Jansen and R. Hoppe. Notiz zur Kenntnis der Oxocobaltate des Natriums. physica
status solidi (RRL) – Rapid Research Letters, 408(2):104–106, September 1974.
[59] M. Thomas, P. Bruce, and J. Goodenough. Lithium mobility in the layered oxide
Li1−x CoO2 . Solid State Ionics, 17(1):13–19, July 1985.
[60] K. Takada, H. Sakurai, E. Takayama-Muromachi, F. Izumi, R.A. Dilanian, and
T. Sasaki. Superconductivity in two-dimensional CoO2 layers. Nature, 422(6927):53–55,
March 2003.
[61] C. Fouassier. Sur de Nouveaux Bronzes Oxygenes de Formule Nax CoO2 ) Le Systeme
Cobalt-Oxygene-Sodium. J Solid State Phys, 1973.
[62] M. Lee, L. Viciu, L. Li, Y. Wang, M.L. Foo, S. Watauchi, R.A. Pascal Jr, R.J. Cava,
and N.P. Ong. Large enhancement of the thermopower in Nax CoO2 at high Na doping.
Nature Materials, 5(7):537–540, June 2006.
[63] Y. Takahashi. Single-crystal growth, crystal and electronic structure of NaCoO2 . Journal of Solid State Chemistry, 172(1):22–26, April 2003.
[64] T. Nagira, M. Ito, and S. Katsuyama. Thermoelectric properties of Nax Co2 O4 prepared by the polymerized complex method and hot-pressing. Materials transactions,
44(1):155–160, 2003.
[65] F. Ioffe. Semiconductor thermoelements, and Thermoelectric cooling. Infosearch, ltd.,
1957.
[66] S. Li, R. Funahashi, I. Matsubara, K. Ueno, S. Sodeoka, and H. Yamada. Synthesis and
Thermoelectric Properties of the New Oxide Materials Ca3-x Bix Co4 O9+δ . Chemistry
of Materials, 12(8):2424–2427, August 2000.
[67] E. Woermann and A. Muan. Phase equilibria in the system CaO-cobalt oxide in air.
Journal of Inorganic and Nuclear Chemistry, 32(5):1455–1459, 1970.
[68] H.Q. Liu, Y. Song, S.N. Zhang, X.B. Zhao, and F.P. Wang. Thermoelectric properties of
Ca3-x Yx Co4 O9 ceramics. Journal of Physics and Chemistry of Solids, 70(3-4):600–603,
2009.
[69] Y. Masuda, D. Nagahama, H. Itahara, T. Tani, W.S. Seo, and K. Koumoto. Thermoelectric performance of Bi- and Na-substituted Ca3 Co4 O9 improved through ceramic
texturing. Journal of Materials Chemistry, 13(5):1094–1099, March 2003.
90

[70] Y. Liu, Y. Lin, L. Jiang, C.W. Nan, and Z. Shen. Thermoelectric properties of Bi3+
substituted Co-based misfit-layered oxides. Journal of Electroceramics, 21(1):748–751,
2008.
[71] C. Liu, G. Chen, J. Pei, L. Cui, D. LU, N. Zhou, and H. Xian. Preparation and
characterization of the new oxides Ca2−x Nax Co2 O5 . Physica B: Condensed Matter,
403(10-11):1808–1812, May 2008.
[72] D. Wang, L. Chen, Q. Yao, and J. Li. High-temperature thermoelectric properties of
Ca3 Co4 O9 with Eu substitution. Solid State Communications, 129(9):615–618, March
2004.
[73] F. Delorme, F.C. Martin, P. Marudhachalam, D. Ovono, and G. Guzman. Effect of Ca
substitution by Sr on the thermoelectric properties of Ca3 Co4 O9 ceramics. Journal of
Alloys and Compounds, 509(5):2311–2315, February 2011.
[74] F.P. Zhang, Q.M. Lu, and J.X. Zhang. Synthesis and high temperature thermoelectric
properties of Bax Agy Ca3−x−y Co4 O9 compounds. Journal of Alloys and Compounds,
484(1-2):550–554, September 2009.
[75] Y. Nishino, M. Kato, S. Asano, K. Soda, M. Hayasaki, and U. Mizutani. Semiconductorlike behavior of electrical resistivity in Heusler-type Fe2 VAl compound. Physical
Review Letters, 79(10):1909–1912, 1997.
[76] E.J.J Skoug, C. Zhou, Y. Pei, and D.T Morelli. High Thermoelectric Power Factor Near
Room Temperature in Full-Heusler Alloys. In Journal of Electronic Materials, pages
1221–1223, 2009.
[77] Y. Nishino, S. Deguchi, and U. Mizutani. Thermal and transport properties of the
Heusler-type Fe2 VAl1−x Gex (0 x 0.20) alloys: Effect of doping on lattice thermal conductivity, electrical resistivity, and Seebeck coefficient. Physical Review B,
74(11):115115, 2006.
[78] H. Matsuura, Y. Nishino, U. Mizutani, and S. Asano. Doping effects on thermoelectric
properties of the pseudogap Fe2 VAl system. Journal of the Japan Institute of Metals,
66(7):767–771, 2002.
[79] H. Nakayama, N. Ide, and Y Nishino. Thermoelectric Properties of P-Type Heusler
Compounds (Fe2-x Cox )(V1-y Tiy )Al. Materials Transactions, 49(8):1858, August 2008.
91

[80] W. Lu, W. Zhang, and L. Chen. Thermoelectric properties of (Fe1−x Cox )2 VAl Heuslertype compounds. Journal of Alloys and Compounds, 484(1-2):812–815, 2009.
[81] Y. Nishino, H. Kato, and M. Kato. Effect of off-stoichiometry on the transport properties
of the Heusler-type Fe2 VAl compound. Physical Review B, 2001.
[82] C. Lue, C. Chen, J. Lin, Y. Yu, and Y. Kuo. Thermoelectric properties of quaternary
Heusler alloys Fe2 VAl1−x Six . Physical Review B, 75(6):064204, February 2007.
[83] Y. Nishino. Unusual electron transport in Heusler-type Fe2 VAl compound. In Intermetallics, pages 1233–1241. Nagoya Inst Technol, Dept Mat Sci & Engn, Showa Ku,
Nagoya, Aichi 4668555, Japan, 2000.
[84] J. Yang, D.T. Morelli, G.P. Meisner, W. Chen, J.S. Dyck, and C. Uher. Effect of
Sn substituting for Sb on the low-temperature transport properties of ytterbium-filled
skutterudites. Physical Review B, 67(16):165207, 2003.
[85] M. Mikami and A. Matsumoto. Synthesis and thermoelectric properties of microstructural Heusler Fe2 VAl alloy. Journal of Alloys and Compounds, 2008.
[86] M. Mikami. Effect of Bi addition on microstructure and thermoelectric properties of
Heusler Fe2 VAl-sintered alloy. Journal of Alloys and Compounds, 2008.

92