YTTERBIUM BASED ALLOYS FOR SPACE-BASED THERMOELECTRIC COOLING
APPLICATIONS
By
Gloria Jeannine Lehr

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Materials Science and Engineering – Doctor of Philosophy
2014

ABSTRACT
YTTERBIUM BASED ALLOYS FOR SPACE-BASED THERMOELECTRIC COOLING
APPLICATIONS
By
Gloria Jeannine Lehr
Solid-state thermoelectric devices are of great interest for researchers in many fields due
to their potential to increase the efficiency of electrical power generation as well as provide
localized heating or cooling in inconvenient environments. Improving the thermoelectric
properties of the materials used in these devices is critical to improve their efficiencies and thus
feasibility for niche applications. Unfortunately, several of the underlying materials properties
are inversely correlated making the thermoelectric figure of merit difficult to improve. In
addition, the performance of traditional semiconducting materials is degraded as the operating
temperature is decreased. Thus, a unique approach must be taken to achieve larger thermoelectric
figures of merit at low temperatures. The research presented here investigates the use of
intermediate valence Yb-based compounds for Peltier cooling in the cryogenic regime.
These Yb-based intermediate valence compounds demonstrate large Seebeck coefficients
at low temperatures, which is essential to a large figure of merit. The Seebeck coefficient is
related to the fluctuating Yb valence, a relationship that is investigated in several compounds and
utilized to maximize the thermoelectric figure of merit. Solid solutions are synthesized in order
to reduce the lattice thermal conductivity as well as alter the size of the unit cell. Changes in the
unit cell volume may contribute to a change in the average Yb valence, and thus be utilized to
tune the magnitude and peak temperature of the Seebeck coefficient. By reducing the lattice
thermal conductivity and optimizing the Seebeck coefficient, enhancements in the thermoelectric
figure of merit can be achieved.

To my family and friends for their unwavering support,
endless encouragement and unconditional love.

iii

ACKNOWLEDGEMENTS

I am extremely grateful for the guidance and support I have received over the years from
my advisor and mentor, Dr. Donald Morelli. I would like to thank him for the opportunities he
has afforded me to not only conduct this research but also to travel near and far to present and
publish my results. I would also like to thank current and former members of the Morelli lab for
their hard work, collaboration, and helpful discussions throughout my graduate studies:
Vijayabarathi Ponnambalam, Kevin Zhou, Eric Skoug, Chang Liu, Hui Sun, Steven Boona, Xu
Lu, Winston Carr, Spencer Waldrop, and Jared Williams. In addition, many thanks to Travis
Thompson for his insightful conversations and unique perspective.
I am also very thankful for my doctoral committee members Drs. Martin Crimp, S.D.
Mahanti, and K. Subramanian for their counsel and guidance along the way. Collaborators Dr.
Joseph Heremans and Hyungyu Jin (Ohio State University) have been critical to my success as a
researcher and are greatly appreciated, as well as Dr. Reza Loloee (MSU-Physics), Karl Dersch
(MSU-ECE) and the many other faculty, staff and students at MSU who have assisted me.
I am indebted to Dr. Christopher Fasano (Monmouth College), without whom I would
not have attended graduate school. He instilled in me a passion for science and provided me with
the strong foundation upon which this work is built.
Finally, I would like to acknowledge the Air Force Office of Scientific Research for
funding this project under Multi-University Research Initiative (MURI), “Cryogenic Peltier
Cooling,” under Contract Number FA9550-10-1-0533.

iv

TABLE OF CONTENTS

ABSTRACT ................................................................................................................................... ii
ACKNOWLEDGEMENTS ........................................................................................................ iv
TABLE OF CONTENTS ............................................................................................................. v
LIST OF FIGURES .................................................................................................................... vii
Chapter 1 Introduction................................................................................................................. 1
1.1 Thermoelectric Effects .......................................................................................................... 1
1.1.1 Seebeck Effect ............................................................................................................. 1
1.1.2 Peltier Effect ................................................................................................................ 3
1.1.3 Thomson Effect ........................................................................................................... 3
1.2 Thermoelectric Devices and Applications ............................................................................ 4
1.2.1 Power Generation ........................................................................................................ 6
1.2.2 Thermoelectric Refrigeration ...................................................................................... 8
1.3 Thermoelectric Performance ............................................................................................... 10
1.3.1 Contraindicated Properties ........................................................................................ 11
Seebeck Coefficient ............................................................................................... 12
Thermal Conductivity ............................................................................................ 13
1.3.2 Approaches for Enhancing zT ................................................................................... 14
Chapter 2 Intermediate Valence Effects ................................................................................... 19
2.1 Intermediate Valence........................................................................................................... 19
2.2 Theoretical Understanding .................................................................................................. 20
Chapter 3 Methodology .............................................................................................................. 26
3.1 Synthesis.............................................................................................................................. 26
3.1.1 Arc Melting ............................................................................................................... 26
3.1.2 Homogenization ........................................................................................................ 27
3.1.3 Chemical Etching ...................................................................................................... 27
3.1.4 Ball Milling ............................................................................................................... 28
3.1.5 Densification ............................................................................................................. 29
3.2 Characterization .................................................................................................................. 30
3.2.1 Density....................................................................................................................... 30
3.2.2 X-ray Diffraction ....................................................................................................... 31
3.2.3 Scanning Electron Microscopy ................................................................................. 31
3.3 Property Measurement ........................................................................................................ 32
3.3.1 Transport Properties .................................................................................................. 32
3.3.2 Magnetic Properties ................................................................................................... 33
Chapter 4 YbAl3-Based Solid Solutions .................................................................................... 35
4.1 Introduction ......................................................................................................................... 35
4.2 Pure YbAl3 .......................................................................................................................... 37
4.2.1 Synthesis .................................................................................................................... 37
4.2.2 Characterization......................................................................................................... 38
4.2.3 Transport Properties .................................................................................................. 40
4.3 Yb1-xScxAl3 .......................................................................................................................... 40
4.3.1 Synthesis .................................................................................................................... 40
v

4.3.2 Characterization......................................................................................................... 41
4.3.3 Transport Properties .................................................................................................. 45
4.4 Yb1-xErxAl3 .......................................................................................................................... 49
4.4.1 Synthesis .................................................................................................................... 49
4.4.2 Characterization......................................................................................................... 50
4.4.3 Transport Properties .................................................................................................. 52
4.5 Yb1-xLuxAl3 ......................................................................................................................... 56
4.5.1 Synthesis .................................................................................................................... 57
4.5.2 Characterization......................................................................................................... 58
4.5.3 Transport Properties .................................................................................................. 59
4.6 Summary ............................................................................................................................. 63
Chapter 5 YbAl2-Based Solid Solutions .................................................................................... 66
5.1 Introduction ......................................................................................................................... 66
5.2 Yb1-xScxAl2 .......................................................................................................................... 67
5.2.1 Synthesis .................................................................................................................... 68
5.2.2 Characterization......................................................................................................... 69
5.2.3 Transport Properties .................................................................................................. 72
5.2.4 Magnetic Properties ................................................................................................... 76
5.3 Summary ............................................................................................................................. 81
Chapter 6 YbCu2Si2-Based Solid Solutions .............................................................................. 84
6.1 Introduction ......................................................................................................................... 84
6.2 YbCu2Si2-xGex ..................................................................................................................... 86
6.2.1 Synthesis .................................................................................................................... 86
6.2.2 Characterization......................................................................................................... 87
6.2.3 Transport Properties .................................................................................................. 90
6.2.4 Magnetic Properties ................................................................................................... 93
6.3 Yb1-xLaxCu2Si2 .................................................................................................................... 96
6.3.1 Synthesis .................................................................................................................... 96
6.3.2 Characterization......................................................................................................... 97
6.3.3 Transport Properties .................................................................................................. 99
6.3.4 Magnetic Properties ................................................................................................. 103
6.4 Summary ........................................................................................................................... 106
Chapter 7 Conclusions.............................................................................................................. 108
REFERENCES .......................................................................................................................... 113
REFERENCES .......................................................................................................................... 114

vi

LIST OF FIGURES

Figure 1-1: Schematic of thermocouple where A and B are different materials. ........................... 2
Figure 1-2: Thermoelectric unit operating in refrigeration mode ................................................... 4
Figure 1-3: Thermoelectric Device: Composed of many unicouples, these devices can be
fabricated in many shapes and sizes. .................................................................................. 5
Figure 1-4: Of all of the energy produced in the United States, waste heat accounts for nearly
two-thirds. From Lawrence Livermore National Lab and Department of Energy (2013).
[13] ...................................................................................................................................... 6
Figure 1-5: Power Factor is optimized for highly doped semiconductors. Adapted from Snyder
and Toberer Nature (2008). [21]....................................................................................... 12
Figure 1-6: Thermal conductivity of the Si-Ge alloy as a function of Si concentration. .............. 16
Figure 2-1: A) Ideal delta function, B) Lorentzian function, C) Lorentzian function superimposed
on a background DOS (~E1/2). .......................................................................................... 22
Figure 3-1: Sample mounting configuration for transport property measurements ..................... 33
Figure 4-1: Unit cell diagram of YbAl3. ....................................................................................... 36
Figure 4-2: The Ytterbium-Aluminum binary phase diagram by Gschneider and Calderwood
[79]. ................................................................................................................................... 37
Figure 4-3: XRD patterns of each step in the YbAl3 synthesis process as labeled: A) after the
nominal YbAl9 was arc melted - * indicate peaks from the pure Al phase, B) after sample
is etched in NaOH to remove excess Al, C) after YbAl3 is ball milled, D) after
densification using PECS – unlabeled peaks are oxide impurity phase. .......................... 39
Figure 4-4: The Sc-Al phase diagram showing the congruent melting of the ScAl3 phase [84]. . 41
Figure 4-5: XRD pattern for the Yb1-xScxAl3 solid solution where A) x = 0.0, B) x = 0.25, C) x =
0.50, D) x = 0.75, E) x = 1.0 – unlabeled peaks are from the oxide impurity phase. ....... 43
Figure 4-6: Lattice parameters for the Yb1-xScxAl3 solid solution where 0≤x≤1 at room
temperature. Closed symbols come from literature data [79, 81]. .................................... 43
Figure 4-7: SEM images of fractured surfaces for A) YbAl3 and B) ScAl3. ................................ 44
Figure 4-8: Electrical resistivity from 80K to 350K for the Yb1-xScxAl3 solid solution where x =
0.0, 0.25, 0.5, 0.75, and 1.0. Dashed lines are guides to the eye. ..................................... 45

vii

Figure 4-9: Seebeck coefficient from 80K to 350K for the Yb1-xScxAl3 solid solution where x =
0.0, 0.25, 0.5, 0.75, and 1.0 ............................................................................................... 46
Figure 4-10: Power Factor from 80K to 350K for the Yb1-xScxAl3 solid solution where x = 0.0.
0.25, 0.50, 0.75, and 1.0. ................................................................................................... 47
Figure 4-11: Total thermal conductivity from 80K to 350K for the Yb1-xScxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ........................................................................... 48
Figure 4-12: Dimensionless figure of merit from 80K to 350K for the Yb1-xScxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ........................................................................... 48
Figure 4-13: The Erbium-Aluminum binary phase diagram by Gschneider and Calderwood [85].
........................................................................................................................................... 50
Figure 4-14: XRD pattern of the Yb1-xErxAl3 solid solution at room temperature where A) x = 0,
B) x = 0.25, C) x = 0.50, D) x = 0.75. and E) x = 1.0. ...................................................... 51
Figure 4-15: SEM images of fractured surfaces for ErAl3. .......................................................... 51
Figure 4-16: Electrical resistivity from 80K to 350K for the Yb1-xErxAl3 solid solution where x =
0.0, 0.25, 0.50, 0.75, and 1.0. ............................................................................................ 52
Figure 4-17: Seebeck coefficient from 80K to 350K for the Yb1-xErxAl3 solid solution where x =
0.0. 0.25, and 0.50. ............................................................................................................ 53
Figure 4-18: Power factor from 80K to 350K for the solid solution Yb1-xErxAl3 where x = 0.0,
0.25, and 0.50. ................................................................................................................... 54
Figure 4-19: Lattice thermal conductivity from 80K to 350K for the Yb1-xErxAl3 solid solution
where x = 0.0, 0.25, and 0.50. ........................................................................................... 54
Figure 4-20: Total thermal conductivity from 80K to 350K for the Yb1-xErxAl3 solid solution
where x = 0.0, 0.25, and 0.50. ........................................................................................... 55
Figure 4-21: The dimensionless figure of merit from 80K to 350K for the Yb1-xErxAl3 solid
solution where x = 0.0, 0.25, and 0.50 .............................................................................. 56
Figure 4-22: The Lutetium-Aluminum binary phase diagram by H. Okamoto [86]. ................... 57
Figure 4-23: XRD pattern of the Yb1-xLuxAl3 solid solution at room temperature where A) x =
0.0, B) x = 0.25, C) x = 0.50, D) x = 0.75, and E) x = 1.0. ............................................... 58
Figure 4-24: SEM images of fractured surfaces for LuAl3. .......................................................... 59
Figure 4-25: The electrical resistivity from 80K to (300K) 350K for Yb1-xLuxAl3 where
x=
0.0, 0.25, 0.50, 0.75, and 1.0. ............................................................................................ 60

viii

Figure 4-26: The Seebeck coefficient from 80K to (300) 350K for the Yb1-xLuxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ........................................................................... 60
Figure 4-27: The power factor from 80K to (300K) 350K for the Yb1-xLuxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ........................................................................... 61
Figure 4-28: The lattice thermal conductivity from 80K to (300K) 350K for Yb1-xLuxAl3 where x
= 0.0, 0.25, 0.50, 0.75, and 1.0. ........................................................................................ 62
Figure 4-29: The total thermal conductivity from 80K to (300K) 350K for the Yb1-xLuxAl3 solid
solution where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ............................................................. 62
Figure 4-30: The thermoelectric dimensionless figure of merit from 80K to (300K) 350K for
Yb1-xLuxAl3 where x = 0.0, 0.25, 0.50, 0.75, and 1.0. ...................................................... 63
Figure 5-1: Unit cell diagram of YbAl2. ....................................................................................... 66
Figure 5-2: The XRD patterns at room temperature following the each of the steps in
synthesizing YbAl2. A) after arc melting nominal YbAl where unlabeled peaks indicate a
pure Yb phase, and B) pure YbAl2 phase after 24 hour heat treatment. ........................... 69
Figure 5-3: The XRD patterns for the Yb1-xScxAl2 solid solution where A) x = 0.60, B) x = 0.70,
C) x = 0.80, D) x = 0.85, E) x = 0.90, F) x = 0.95, and G) x = 1.0. .................................. 70
Figure 5-4: The unit cell volume as a function of Sc concentration at room temperature for Yb1xScxAl2 where x = 0.0≤x≤1.0. Closed circles are data from literature [57]..................... 70
Figure 5-5: SEM images of fractured surfaces for A) Yb0.60Sc0.40Al2 and B) ScAl2. ................... 71
Figure 5-6: The electrical resistivity from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid
solution where x = 0 [93], 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0. ............................. 72
Figure 5-7: The Seebeck coefficient from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid
solution where x = 0 [93]. 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0 .............................. 73
Figure 5-8: The power factor from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid solution
where x = 0 [93], 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0. ........................................... 74
Figure 5-9: The total thermal conductivity from 2K (80K) to 300K (350K) for the Yb1-xScxAl2
solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95. ....................................... 74
Figure 5-10: The zT from 2K(80K) to 300K(350K) for the Yb1-xScxAl2 solid solution where x =
0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0. ........................................................................ 76
Figure 5-11: The corrected magnetic susceptibility data (points) and the corresponding ICF
model fits (lines) from 2K to 300K for the Yb1-xScxAl2 solid solution where x = 0.60,
0.70, 0.80, 0.85, 0.90, and 0.95 ......................................................................................... 78

ix

Figure 5-12: Comparison of the temperatures at which the magnitude of the Seebeck coefficient
and the magnetic susceptibility are maximized as a function of Sc concentration. .......... 78
Figure 5-13: The average Yb valence as a function of temperature obtained from the ICF model
for the Yb1-xScxAl2 solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95. ....... 79
Figure 5-14: The Seebeck coefficient as a function of the average Yb valence for the Yb1-xScxAl2
solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95. ....................................... 80
Figure 6-1: Unit cell diagram of YbCu2Si2. .................................................................................. 85
Figure 6-2: XRD patterns for the YbCu2Si2-xGex solid solution................................................... 87
Figure 6-3: The unit cell volume increases linearly as the Ge concentration is increased in the
YbCu2Si2-xGex solid solution. Solid circles represent literature data [92]. ....................... 88
Figure 6-4: SEM images of fractured surfaces for A) YbCu2Si2 and B) YbCu2Ge2. ................... 89
Figure 6-5: The electrical resistivity for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50,
0.75, 1.0, 1.25, 1.50, 1.75 and 2.0. .................................................................................... 90
Figure 6-6: The Seebeck coefficient for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50,
0.75, 1.0, 1.25, 1.50, 1.75 and 2.0. .................................................................................... 91
Figure 6-7: The power factor for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50, 0.75,
1.0, 1.25, 1.50, 1.75 and 2.0. ............................................................................................. 91
Figure 6-8: The thermal conductivity for the YbCu2Si2-xGex solid solution where x = 0, 0.25,
0.50, 0.75, 1.0, 1.25, 1.50, 1.75 and 2.0. ........................................................................... 92
Figure 6-9: The zT for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50, 0.75, 1.0, 1.25,
1.50, 1.75 and 2.0. ............................................................................................................. 93
Figure 6-10: The corrected magnetic susceptibility for the YbCu2Si2-xGex solid solution where
0≤x≤2. Solid lines are fits to the ICF model. .................................................................. 94
Figure 6-11: The average Yb valence as a function of temperature for the YbCu2Si2-xGex solid
solution where 0≤x≤2. ..................................................................................................... 94
Figure 6-12: The Seebeck coefficient vs. average Yb valence for the YbCu2Si2-xGex solid
solution where 0≤x≤2. ..................................................................................................... 95
Figure 6-13: Comparison of the temperatures at which the magnitude of the Seebeck coefficient
and the magnetic susceptibility are maximized as a function of Ge concentration. ......... 96
Figure 6-14: XRD patterns collected at room temperature for the Yb1-xLaxCu2Si2 solid solution
where A) x = 0, B) x = 0.20, C) x = 0.40, D) x = 0.60, E) x = 0.80, and F) x = 1.0. ........ 97

x

Figure 6-15: The unit cell volume increases linearly as the La concentration is increased in the
Yb1-xLaxCu2Si2 solid solution. Solid circles represent literature data............................... 98
Figure 6-16: SEM images of fractured surfaces for LaCu2Si2...................................................... 98
Figure 6-17: The electrical resistivity for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05,
0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0. ............................................. 99
Figure 6-18: The Seebeck coefficient for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05,
0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0. ........................................... 100
Figure 6-19: The power factor for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10,
0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0. .................................................... 101
Figure 6-20: The thermal conductivity for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05,
0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0. ........................................... 102
Figure 6-21: The zT for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10, 0.15, 0.20,
0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0. ...................................................................... 102
Figure 6-22: The corrected magnetic susceptibility for the Yb1-xLaxCu2Si2 solid solution where x
= 0, 0.05, 0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0............................ 104
Figure 6-23: The average Yb valence as calculated from the ICF model from 2K to 300K for
Yb1-xLaxCu2Si2 where x=0, 0.05, 0.10, 0.15, and 0.20. .................................................. 104
Figure 6-24: The Seebeck coefficient as a function of the average Yb valence for the Yb1xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10, 0.15, and 0.20. ................................ 105
Figure 7-1: The average Yb valence vs. the unit cell volume for three solid solutions shows that
the Yb valence state is dependent on the available space in the unit cell. ...................... 110
Figure 7-2: The Seebeck coefficient vs. average Yb valence for each of the Yb-based
intermediate valence compounds investigated in this work. .......................................... 111

xi

Chapter 1 Introduction
Energy efficiency has become a widespread household concern, whether it is lighting,
appliances, or vehicle fuel efficiency. The US Department of Energy and other government
agencies spend millions of dollars on research to improve the efficiency of different energy
processes. One way to improve energy efficiency is to recapture the heat that is expelled as a
wasted byproduct from a generator or other device and convert it into useable electricity.
Conversion of this wasted heat to electricity can be achieved by incorporating solid-state
thermoelectric devices into the system. These thermoelectric devices can also be used as
localized heaters/refrigerators when a current is passed through the device creating a temperature
gradient. Thermoelectric devices are important for niche applications that require precise
temperature control, such as medicinal storage units, laboratory equipment, and even high-end
computing systems.

1.1 Thermoelectric Effects
The thermoelectric effects that are implemented in solid-state thermoelectric devices today were
discovered long ago, but were not completely understood for quite some time. Three
fundamental effects that were discovered in the 19th century are known as the Seebeck Effect,
Peltier Effect, and the Thomson Effect. Fully understanding how these effects work is critical to
optimizing thermoelectric materials and thus increasing the efficiency of thermoelectric devices.

1.1.1 Seebeck Effect
The Seebeck effect was first discovered in 1821 when Thomas Johann Seebeck found that a
circuit made of two different metals deflected the needle of a magnet when the junctions were

1

held at different temperatures. Seebeck believed that the magnetic field was created directly by
the temperature gradient and that this effect could explain Earth’s magnetic poles [1]. It was later
discovered that this magnetic field was actually created by an electric current flowing through
the circuit. Despite Seebeck’s incorrect interpretation of the effect, he is still credited with the
discovery that a temperature gradient at the junctions of two dissimilar materials causes an
electrical current to flow [2–6].

Figure 1-1: Schematic of thermocouple where A and B are different materials.
The Seebeck effect is the basis for the thermocouple, which consist of two wires of
dissimilar materials joined together. In order for a thermocouple to work, the two wires create a
circuit, across which a voltage is measured. Figure 1-1 shows a schematic of a thermocouple,
where the temperature difference (Th-Tc) is determined by measuring the voltage across two
different materials (A and B). This measured voltage and a proportionality constant, known as
the Seebeck coefficient, can be used to calculate the temperatures of the two junctions using the
following formula:
Equation 1
where V is the voltage measured across the circuit of the two different materials, S is the Seebeck
coefficient, Th is the hot side temperature, and Tc is the cold side temperature.
2

1.1.2 Peltier Effect
In 1834 Jean Charles Athanase Peltier discovered that a current passed through a circuit of two
dissimilar materials caused one junction to be heated and the other cooled [3–5]. This
heating/cooling effect was switched when the direction of the current was reversed. The heat
generated at a junction per unit time, Q, can be described as:
Equation 2
where Π is the Peltier coefficient for materials A or B and I is the electric current from A to B
[3,7]. This Peltier refrigeration (and heating) was very clearly demonstrated in 1838 by Heinrich
Lenz when he placed a drop of water at the junction of two wires, made of bismuth and
antimony, and then passed a current through the circuit. Passing the current through the junction
in one direction caused the water to freeze while reversing the direction of the current caused the
ice to melt [3–5].

1.1.3 Thomson Effect
Sir William Thomson, also known as Lord Kelvin, also contributed to the field of
thermoelectrics by improving the interpretation and understanding of the Seebeck and Peltier
effects. Thomson noted that the Peltier effect is the reverse of the Seebeck effect, and that these
coefficients are thermodynamically related by the Kelvin relation as:
Equation 3
where T is the temperature in Kelvin [3]. This relationship is used for simplicity, and today, only
the Seebeck coefficient is used in thermoelectric studies and reports. Thomson also observed that
heating or cooling occurs in different segments of a single material as a current is passed through
the material as long as a temperature gradient is maintained across the material. This is known as
the Thomson effect and can be thought of as a continuous version of the Peltier effect.
3

1.2 Thermoelectric Devices and Applications
Based on these fundamental effects, a thermoelectric device can operate either as a
thermoelectric generator (TEG) or as a thermoelectric cooler (TEC). These devices are all solidstate, meaning that there are no moving parts in the system. Currently, thermoelectric devices are
not widespread commercially due to their low efficiencies, which are limited by the materials
used in the device. There is substantial interest in thermoelectrics research as there is a long list
of potential applications for both TEGs and TECs fueling the research on material
improvements.

Figure 1-2: Thermoelectric unit operating in refrigeration mode

Thermoelectric generators and thermoelectric cooling devices are both designed in the
same manner. In order to operate, these devices require two dissimilar materials arranged as a
circuit. If a temperature gradient is applied across this circuit, a current is established as
4

discussed in the explanation of the Seebeck effect (Section 1.1.1). If a current is applied to the
circuit, one of the junctions is cooled and the other junction is heated as discussed in the Section
1.1.2 on the Peltier effect. These two dissimilar materials are known as n-type and p-type. N-type
materials have electrons as the dominant charge carrier, and a negative Seebeck coefficient.
Conversely, p-type materials use holes as the dominant charge carrier and have a positive
Seebeck coefficient. These n- and p-type materials make up the two parallelepiped legs, shown
in Figure 1-2, known as a unicouple. Any number of these unicouples may be arranged together,
electrically in series and thermally in parallel to create a thermoelectric device, generally in an
arrangement similarly to that shown in Figure 1-3 [8]. Arranging these unicouples thermally in
parallel helps to maintain the temperature gradient across each element of the device, while
placing them electrically in series allows for an increased voltage output.

Figure 1-3: Thermoelectric Device: Composed of many unicouples, these devices can be
fabricated in many shapes and sizes.
5

1.2.1 Power Generation
Thermoelectric devices operating in power generation mode convert a temperature gradient into
an electrical current. This phenomenon occurs because of the Seebeck effect. TEGs have been
widely used over the past 50 years in space exploration missions such as Voyager 1 and 2,
Galileo, the Mars Curiosity Rover, and many other satellites/spacecraft [9,10]. In these
applications, a radioactive material provides the heat source, which TEGs convert into a usable
electrical current to power the sensors and other electronics aboard the spacecraft [9]. TEGs are
optimal for this type of application because they are reliable; having an infinite theoretical
lifetime with no moving parts and the power they provide does not fluctuate as long as the
temperature gradient is maintained. These devices are also known to perform well in harsh
environments, such as the vacuum of deep space [11,12].

Figure 1-4: Of all of the energy produced in the United States, waste heat accounts for nearly
two-thirds. From Lawrence Livermore National Lab and Department of Energy (2013). [13]
6

Apart from space applications, TEGs have also gained attention terrestrially for their
environmentally friendly operation, which can improve the efficiency of any energy related
process. Of all of the energy produced in the United States, whether it be from renewable
resources or fossil fuels, only 40% is actually utilized, while the remaining 60% is wasted,
typically in the form of heat, as seen in Figure 1-4 [13]. This statistic demonstrates the impact
that TEGs could have in reducing the cost of electricity or increasing the efficiency of countless
processes if integrated on a large scale. Current TEGs are capable of converting between 5% and
10% of wasted heat into electricity [3,11,12,14,15]. While this number sounds small, it is much
more significant when we consider that the heat needed for these TEGs to operate is essentially
free, as it is already being produced and expelled as waste.
Many industries could benefit greatly if thermoelectric efficiencies were improved
enough to make them commercially viable solutions. One such application that is widely
researched is in the automotive industry [12], where TEGs could convert waste vehicle exhaust
heat into electricity. The average automobile on the road releases roughly 2/3 of its energy as
heat through the exhaust and radiator; only the remaining 1/3 is actually available to power the
car and drive auxiliary components [16]. The electricity produced by the TEGs could be used to
power accessories, such as the radio, power windows and locks, navigational systems, etc., that
currently draw off of the engine and reduce the vehicle’s fuel efficiency. There are many
challenges to successfully integrating TEGs into an automotive system. For example, the
temperature of the exhaust gas dramatically fluctuates based on the driving cycle, while each
thermoelectric module has a small optimal temperature range. Therefore, when the exhaust gas is
not at the ideal temperature, the TEG does not produce much or any electricity, rendering the
entire device an unnecessary added weight. In addition, the thermoelectric materials must be able
7

to handle thermal cycling without degradation of their thermoelectric or mechanical properties.
The mechanical strength of the thermoelectric materials involved is another issue that must be
addressed. These TEG devices would be attached under the car (similarly to a muffler);
therefore, they must endure many bumps and vibrations. Many of today’s best thermoelectric
materials are rather brittle and would not likely withstand this type of environment. Despite all of
the engineering challenges, the fuel efficiency returns could be great enough that many
automotive companies are investing significant time and money into overcoming these hurdles.
Enhancing the power conversion efficiency of thermoelectric materials would also make
TEGs an economic option for recovering waste heat from industrial and manufacturing power
plants [12]. There is a significant amount of waste heat that is expelled from these plants in gas
or liquid form because it is no longer at an adequate temperature for conventional power
generation systems. However, TEGs optimized for this temperature range could be utilized to
regain some of this lost energy and increase the amount of power produced with zero excess fuel.
As the materials in TEGs become cost effective enough to manufacture and/or efficient enough
to justify their cost, integration of these thermoelectric power generators is inevitable. In the last
year, the first commercially available thermoelectric power generator was produced and
announced for industrial applications by Alphabet Energy. This waste heat generator is capable
of generating electricity at approximately 2.5% efficiency [17]. Further research is still needed to
improve the thermoelectric properties of the materials and enhance the electrical output of these
generators.

1.2.2 Thermoelectric Refrigeration
Thermoelectric cooling/refrigeration, also known as Peltier cooling/refrigeration, operates
similarly to the power generators, only in reverse. These thermoelectric coolers (TECs) function
8

based on the Peltier effect, which was explained in Section 1.1.2. A current is supplied to the
device and one side is cooled while the other side is heated. Reversing the direction of the
current switches which side of the device heats and which side cools. TECs are currently not
widespread; however, in niche applications where precise temperature control is the priority,
rather than energy efficiency, these solid-state devices are often implemented.
A few examples where TECs are used are: in laboratory equipment where specimens
must be held at a very specific temperature, in many wine coolers, for the heating and cooling of
upscale automotive seats, and as coolers for expensive computer systems [12,18]. There are
many other applications where TECs could be used if their efficiencies were better. One
particular application, at which this study is directed, is for cooling infrared sensors located on
satellites orbiting the earth [12,19].
While space is fundamentally thought of as “cold” it is also a vacuum, and therefore any
heat that is produced does not diffuse away from the source, causing electronics in space to
become quite hot. This problem is addressed via cooling systems that are implemented in all
satellites. Infrared sensors, however, pose a slightly more difficult problem, because in order to
operate properly they must be cooled to cryogenic temperatures. Previously, liquid helium or
nitrogen was used to cool such devices, requiring large pressurized dewars, which, regardless of
capacity, have a relatively short lifetime. The current state-of-the-art technology for cooling
these infrared sensors utilizes mechanical cooling systems known as cryocoolers [20]. These
cryocoolers operate on fundamental thermal cycles, such as Joule-Thomson, that require
compression and expansion of gasses for cooling. While these mechanical cryocoolers are
successful at the given task, there are many issues with the basic idea of using mechanical
devices in space. The nature of mechanical systems is that over time the moving parts wear out
9

and break down, becoming a major concern when the primary function of the device is to operate
in orbit. The moving parts in this device also create vibrations that are not desirable for sensitive
components, such as infrared sensors. It is also very difficult to scale down these devices to
smaller sizes without greatly reducing the efficiency. Finally, mechanical cryocoolers are very
expensive to produce and tune, require costly electronics, and they also add a significant cost to
the launch, as they are quite heavy [20]. Despite this list of negative attributes, to date
mechanical cryocoolers are the best solutions currently available for cooling infrared sensors.
This work is aimed at improving the thermoelectric properties of thermoelectric materials
at low temperatures, in hopes of replacing these mechanical cryocoolers with cryogenic TECs.
TECs can solve many of the problems that plague mechanical cryocoolers because of their all
solid state nature [5,6,21]. There are no moving parts to wear out and break down, and therefore,
no unwanted vibrations are added to the infrared sensor system. These TECs would weigh less
and likely cost much less to produce, making them a very attractive option. Finally, unlike
mechanical cryocoolers, the efficiency of a TEC scales with the size of the device, and therefore
devices can be custom made for each infrared sensor detector as necessary. However, the
thermoelectric properties of typical thermoelectric materials are dramatically reduced below
room temperature and have essentially zero efficiency in the cryogenic range. Therefore, a new
approach to thermoelectrics with unique materials is reported here in the study of ytterbiumbased thermoelectric alloys.

1.3 Thermoelectric Performance
In order to understand and compare the performance of thermoelectric materials, a single value,
known as the thermoelectric dimensionless figure of merit, zT, is used. This value incorporates
the electrical conductivity (σ), thermal conductivity (κ), the Seebeck coefficient (S), and
10

operating temperature (T) as:
Equation 4
This value is useful because all thermoelectric materials, regardless of optimal operating
temperature, can be directly compared. The zT of each material can also be used to determine the
theoretical conversion efficiency (η) of a TEG or coefficient of performance (φ) of a TEC [22].
Equation 5

Equation 6
Although the zT is a fairly simple equation, it is made up of properties that are intertwined with
one another, making it difficult to tune them independently and optimize the zT value.

1.3.1 Contraindicated Properties
Electrical Conductivity
Several of the properties in the zT equation are “contraindicated,” in other words, optimizing one
property ultimately decreases the value of another property. As can be seen in Figure 1-5 [21], as
the number of charge carriers (either electrons or holes) is increased, the electrical conductivity
increases [21]. This is because there are a larger number of charge carriers (n), and the electrical
conductivity goes as:
Equation 7
where e is the charge of the carrier and μ is the carrier mobility.

11

Figure 1-5: Power Factor is optimized for highly doped semiconductors. Adapted from
Snyder and Toberer Nature (2008). [21]

Seebeck Coefficient
Figure 1-5 also shows that as the number of charge carriers increases the Seebeck coefficient
decreases. For metals or degenerate semiconductors the Seebeck coefficient can be expressed as :
Equation 8
where kB is the Boltzmann constant, h is Planck’s constant, and m* is the effective mass of the
charge carrier [20]. This expression clearly shows that increasing n has a detrimental effect on
the Seebeck coefficient. There is also a relationship between the effective mass and the carrier
concentration, in which if one is increased the other is decreased. This affects both the Seebeck
coefficient and the electrical conductivity, as a large effective mass is good for the Seebeck
coefficient, but bad for electrical conductivity [21]. Both the Seebeck coefficient and the

12

electrical conductivity fall in the numerator of the zT value, and S2σ is referred to as the power
factor. Because of these contraindicated properties, there is an optimal range for the number of
carriers in which the power factor is maximized in a thermoelectric material, generally between
1019 and 1021 carriers/cm3, which is typically found in heavily doped semiconductors or
semimetals.
The Seebeck coefficient is also proportional to the derivative of the electron density of
states at the Fermi level as described by the Mott relation [4]:
Equation 9
Here it is apparent that as the electrical conductivity is increased the Seebeck coefficient is
decreased. However, the rate of change of the DOS is also directly related to the Seebeck
coefficient. This description is useful, because it gives an idea of the type of band structure that
is desired for a good thermoelectric material. Ideally, the Fermi level would pass through the top
of a flat conduction band (or bottom of the valence band for p-type). If looking at a plot of the
density of states, the Fermi level should be near the peak, but slightly offset in order to pass
through the steepest part of the density of states peak, as will be further discussed in Chapter 2.
Mahan and Sofo [23] describe the best DOS as a Dirac delta function, however, this is not
achievable in real materials.

Thermal Conductivity
Not only are there contraindicated properties within the power factor, but the electrical
conductivity and thermal conductivity are also entwined. An ideal thermoelectric material would
have a large electrical conductivity with a small thermal conductivity. This is an unrealistic goal
because the same charge carriers that are needed to create a large electrical conductivity also
contribute to the thermal conductivity. The thermal conductivity can be defined as:
13

Equation 10
where κe is the electronic contribution to the thermal conductivity and κl is the lattice
contribution. The electronic contribution can be calculated from the Wiedemann-Franz law:
Equation 11
where L is the Lorenz number, taken as 2.4x10-8 WΩK-2 for metallic or nearly metallic materials.
This relationship demonstrates that increasing σ also increases κe thus increasing the total value
of κ, which is in the denominator of the zT equation. Therefore arbitrarily increasing σ is
generally counterproductive to the overall thermoelectric efficiency of a material.

1.3.2 Approaches for Enhancing zT
The complicated relationships between the thermoelectric properties limit the possible
approaches to enhance the figure of merit, making such a task quite difficult. However, several
methods have proven successful for optimizing the thermoelectric efficiency of materials. In
order to optimize the power factor, the ideal number of carriers for a compound must be found. It
is generally faster and more reliable to obtain results experimentally than by theoretical
calculations. Changing the carrier concentration or carrier mobility is the only way to alter the
electrical conductivity, and thus tuning these properties is the best means available for
optimizing the power factor. Secondly, reducing the thermal conductivity of a compound is a
very important task for enhancing the zT of a thermoelectric material. However, because of the
correlation between the electronic component of the thermal conductivity and the electrical
conductivity, the primary focus must be on reducing the lattice component of the thermal
conductivity. Below are several approaches for reducing lattice thermal conductivity, which may
be implemented independently or in combination to achieve the best results.

14

Creating solid solutions or alloys is achieved by substituting elements of different sizes
and masses for one or more of the original elements in the compound. Typically, solid solutions
are made with atoms that have the same number of conduction electrons as the host atom
(isoelectronic substitution), to prevent changes to the electronic properties. However, the band
structure may be still be altered resulting in systematic changes in the carrier concentration and
mobility. Thus, solid solutions can result in changes to the electronic properties. These changes
can have either positive or negative effects on the power factor. The substitutional atoms in a
solid solution are point defects, which scatter phonons, thus disrupting the flow of the lattice
thermal conductivity [24-26]. The larger the mass difference, the more effective the scattering is:
Equation 12
where r is the scattering rate, c is the impurity concentration, Δm is the mass difference, and

is

average mass of the atoms in the crystal [26]. Thermodynamic calculations can be used to
determine the solubility of a substitutional atom in a compound. Experimentally, candidate
substitutional atoms can be found by looking for compounds with the same crystal structure and
similar lattice parameter. Although this is not a guarantee, it is common that these two different,
but isostructural compounds, will form a full solid solution. A classic example is that of Si-Ge
alloys, which form a complete solid solution and reduce the lattice thermal conductivity via point
defect scattering and thus increases the figure of merit. Figure 1-6 demonstrates this reduction in
the thermal conductivity as Si and Ge are alloyed. A minimum in the thermal conductivity is
observed near x=0.5 where the number of point defects is maximized [27].

15

Thermal Conductivity
(W/cmK)

1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0

0.2

0.4
0.6
Si concentration

0.8

1

Figure 1-6: Thermal conductivity of the Si-Ge alloy as a function of Si concentration.

Doping is a similar method to solid solutions in that one of the host atoms is replaced;
however, in this case, it is replaced with an element with one more or one less electron. The
dopant concentration can be tailored to alter the number of charge carriers in order to achieve a
carrier concentration in the optimal range. Optimizing this carrier concentration alters the
Seebeck coefficient and electrical conductivity to maximize the power factor. Doping can also
create disruptions in the phonon propagation, similarly to solid solutions. Dopant atoms are
usually not significantly different from the host atom in size or mass, only in the number of
electrons, and are typically used in low concentrations. Therefore the phonon scattering effect is
not as strong as in solid-solution alloying, but there is still a measurable decrease in the lattice
contribution of the thermal conductivity for many doped samples.
Finally, another approach to enhancing the zT is through reduction of the lattice
component of the thermal conductivity by reducing the grain size. Particle sizes can be reduced
by powder processing [28] samples and using rapid sintering techniques that prevent grain
growth [29,30]. High-energy ball mills are often used to create powders with particles in the
16

nanometer range. These powders must then be densified without significant grain growth. Hot
pressing and pulsed electric current sintering are two common ways to densify powders. The
time, temperature, and pressure of these processes can be customized for each alloy to achieve
high densities with minimal grain growth. This processing can reduce the lattice thermal
conductivity as the grain boundaries cause disruption of phonon propagation. Smaller grain sizes
means less time between scattering events and thus a higher frequency of disruptions and
reduced lattice thermal conductivity [31].
While these methods for increasing power factor and reducing the lattice thermal
conductivity have proven to be successful approaches for traditional semiconducting
thermoelectric materials, further difficulties exist at lower temperatures. As previously discussed,
the focus of this project is to create thermoelectric materials that operate successfully at
temperatures in the cryogenic range. Generally speaking as temperature decreases in
semiconductors, the number of available carriers is reduced because there is less energy to
promote carriers across the band gap. This results in a reduction to the electrical conductivity
[29]. The Seebeck coefficient also decreases, approaching zero at zero Kelvin. These changes in
the electronic properties with decreasing temperature generally result in extremely low power
factors. The reduced electrical conductivity also results in a decreased electronic component of
the thermal conductivity; however, reduced temperature also means a decrease in the phononphonon scattering in the lattice [32]. Therefore, the total thermal conductivity may actually be
enhanced at low temperatures before approaching zero at zero Kelvin, further reducing the zT
value.
Due to these fundamental issues with traditional semiconductors, a new approach must be
taken in order to discover and develop materials that are optimized for low temperature

17

performance. Metals are generally ignored in the search for good thermoelectric materials
because of their low Seebeck coefficient and large thermal conductivity; however, a particular
group of compounds, known as intermediate valence materials, may offer some unique
properties that can overcome the issue of small Seebeck coefficient at low temperatures despite
their otherwise metallic character [5,23].

18

Chapter 2 Intermediate Valence Effects
Theoretical and experimental work over the past 60 years has shown that optimizing the
thermoelectric effects and achieving large zT in materials at low temperatures is even more
challenging than in high temperature materials [12,15,33]. At low temperatures, not only are the
Seebeck coefficient and electrical conductivity reduced, but also the temperature itself factors in
to reduce the zT value. It has become apparent that traditional doped semiconductors are not
necessarily the best materials for low temperature applications. One promising alternative is to
use intermediate valence alloys. These materials typically have metallic resistivities, but
occasionally also show unusually large peaks in the Seebeck coefficient at low temperatures
[23,33–36]. This high electrical conductivity, in addition to the large Seebeck coefficient, result
in some of the largest power factors ever reported at room temperature and below [23].
Unfortunately, many of these intermediate valence systems have relatively simple unit cells, and
therefore also have very large lattice and total thermal conductivities, reducing the overall zT
value.

2.1 Intermediate Valence
Intermediate valence compounds contain one atom in the host matrix that has a fluctuating
valence, meaning that one electron state can have either n or n+1 electrons at any given moment.
Intermediate valence compounds are not to be confused with mixed valence compounds. Mixed
valence compounds have a single element that can be in one of two different valence states (two
different numbers of valence electrons) according to which lattice site it occupies [37-39]. In an
intermediate valence compound, each atom of consequence (Yb in this case) fluctuates in time
between two different valence states, regardless of lattice site [39–42]. It is common to refer to a
19

non-integer value as the valence state of this element, which is actually the average valence. For
instance if an atom fluctuates between the 2+ and 3+ valence state, it may have an average
valence value of 2.5 if the fluctuating electron is found to exist, on average, in the localized
electron-energy state exactly the same amount it is found in the conduction band. However, if the
atom favors a divalent state, it may be closer to an average valence state of 2.2 as the electron is
more frequently found in the localized state than in the conduction band, and so on. Any noninteger value between 2 and 3 may be assigned to this atom in an intermediate valence
compound.
The valence state of an atom can to some extent be determined by the volume available in
the lattice for the atom to occupy. This volume can be altered via external or internal pressures.
External pressures of several GPa are applied in order to reduce the size of the unit cell and can
often force atoms into a valence state with fewer electrons [43-46], which has a smaller atomic
radius (more of the electrons move to the conduction band). Chemical pressures can also be used
to change the size of the unit cell [39,46–53]. This method allows for expansion or contraction of
the lattice by substitution of atoms of larger or smaller atomic radius relative to the host. The
ability to stretch the unit cell allows researchers to not only examine the effects of smaller unit
cells, but also of larger unit cells on the valence fluctuations.

2.2 Theoretical Understanding
Intermediate valence effects are most commonly found in compounds containing rare-earth
elements, and more specifically Ce and Yb-based compounds [5,23,33,34,39,54–59]. These two
elements can be thought of as opposites because when Ce fluctuates between two valence states
(4+ and 3+) it has an f-shell that is completely empty or contains just one electron, while Yb (2+
and 3+) fluctuates between a completely full f-shell or one containing a single hole. The
20

dominant carrier in Ce-based compounds is the hole, while in Yb-based compounds, electrons
are the dominant charge carrier. Thus, one electron in the f-shell of a Ce-based compound (3+
valence state) is effectively the opposite of one hole in the f-shell of a Yb-based compound (3+
valence state). The simple inverse relationship between Ce- and Yb-based compounds has
resulted in many theoretical models and explanations that are applicable to both types of
systems.
Previous researchers have done a significant amount of work on understanding the effect
that these intermediate valence fluctuations have on different Ce- and Yb-based systems. One of
the most commonly observed effects of intermediate valence systems is a minima in the
electrical resistivity and peaks in the magnetic susceptibility which are reminisant of Kondo-like
systems. In addition to these interesting features in the electrical resistivity and magnetic
susceptibility, intermediate valence compounds are known for large peaks in their Seebeck
coefficient in the low temperature regime [33,34,39,58–60]. These peaks are generally attributed
to a large peak in the electron density of states (DOS) near the Fermi level. While the Dirac delta
function is the ideal electronic transport distribution, it is not a distribution found in real
materials. One of the closest approximations found in nature to this delta function is in rare-earth
containing compounds which have f-electron levels [23].
Figure 2-1 demonstrates a Dirac delta function, a lorentzian distribution, as well as a
more realistic distribution of a lorentzian function superimposed on, or “hybridized” with, a
background signal. The f-electron levels are bound tightly about the nucleus, giving rise to the
narrow lorentzian like peak in the DOS (Figure 2-1B). The background contribution shown in

21

Density of States

A

B

C

Energy
Figure 2-1: A) Ideal delta function, B) Lorentzian function, C) Lorentzian function
superimposed on a background DOS (~E1/2).
Figure 2-1C is 10% of the f-level peak and is approximately proportional to the more classical
E1/2 distribution that represents the DOS signal from the other orbitals in the system. This
background signal can be detrimental to the maximum possible Seebeck coefficient and thus
reduce the maximum zT [23,35]. Calculations by Mahan and Sofo show that background DOS
contributions of greater than 1% of the f-shell peak are quite detrimental to the maximum zT
[23]. In fact, a background contribution of only 10% can reduce the maximum possible figure of
merit to only 25% of its original value [23]. Ideally, the sharp singularity in the DOS is very near
the Fermi energy, which maximizes the thermoelectric figure of merit. A simple examination of
Equation 9, describing the Seebeck coefficient using the Mott relation, shows that the value of
the Seebeck coefficient is dependent on the derivate with respect to energy of the density of

22

states at the Fermi level. Therefore, if the Fermi energy intersects the singular Lorentzian peak at
its steepest point, then the Seebeck coefficient will be maximized.
The magnitude of the Seebeck coefficient is greatly dependent on the electronic transport
distribution as discussed above; however, the concentration of f-ions can alter the Seebeck
coefficient as well [35]. By altering the number of magnetic moments (or f-ions), it is possible to
achieve large values of Seebeck coefficient while providing a means to tailor the temperature at
which the Seebeck coefficient peaks. Diluting the system via isoelectronic substitution typically
weakens the hybridization strength, thus causing a shift in the peak towards lower temperature
without significantly reducing the Seebeck coefficient (particularly when compared to
compounds without intermediate valence behavior) [33,35]. Both concentrated and dilute
quantities of Yb atoms result in interesting features in the magnetic and thermoelectric
properties, due to the small spatial reach of the 4f wave functions, which are on the order of onetenth the interatomic spacing [35]. This means that neighboring Yb atoms have little direct
interaction, and therefore even dilute systems demonstrate strong intermediate valence effects.
Due to the strong f-shell interactions with the conduction electrons in even dilute compositions,
it is possible to study and understand these intermediate valence effects across a wide spectrum
of concentrations as well as to maximize the other thermoelectric properties (which may benefit
from dilute Yb concentrations) without dramatically affecting the Seebeck coefficient.
Isoelectronic substitutions cannot only shift the temperature dependence of the Seebeck
coefficient, but substitutions can also create internal chemical pressure, which can have dramatic
effects on the temperature dependence and ultimately the magnitude of the Seebeck coefficient.
Both internal (chemical) and external pressures can force the Yb ion toward a more stable
trivalent 4f13 configuration [34], while chemical expansion has the opposite effect, resulting in an
23

intermediate valence or nearly divalent state for the Yb-ion [57]. The Seebeck coefficient reflects
the magnetic stability of the Yb-ions and can be theoretically described as one of four types of
temperature dependent shapes, as defined by Zlatic and Monnier [34]. These different types refer
to the temperature dependencies of the Seebeck coefficient as having (a) a positive maximum at
low temperatures and a negative minimum at high temperatures, which are nearly the same in
magnitude, (b) two negative minima with a small positive maximum separating them, (c) a small
negative minimum at low temperatures as well as a large negative minimum at high
temperatures, and (d) a single negative minimum at high temperatures [34]. Here, high
temperature is defined to be near 100K while low temperatures are <30K. By categorizing the
shape of the Seebeck coefficient it is possible to understand the magnetic regime of the Yb-ion as
magnetic with a stable 4f13 configuration (type-a), non-magnetic with an unstable intermediate
valence configuration (type-d), or somewhere in between (types-b&c). By applying pressure to
the system, the compound can be transformed from a type-d shape to a type-c or b shape, and
eventually all of the way to a type-a shape as the magnetic moments are stabilized and no longer
in the intermediate valence regime [34]. The changes induced by internal pressure/expansion can
be optimized in many Yb-based compounds to achieve large peaks in the Seebeck coefficient
over a broad temperature range by achieving a shape between type-c and type-d. This shape
would ideally have a shoulder on the low temperatures side of a large negative minimum,
resulting in the largest overall magnitude of the Seebeck coefficient (near 100K) which is
maintained over a relatively broad temperature range. The concentration of substitutional atoms
required to achieve this type of optimization is different for each composition and may be
determined experimentally, but the theoretical concepts for this approach can be understood by
the single-impurity Anderson model [34].

24

While there is much interesting physics to be explored and understood in these models,
the goal here is two-fold: 1) synthesize new materials on a large enough scale to have
mechanically robust samples for thermoelectric measurements using common processing
techniques and 2) control the magnitude and temperature dependence of the thermoelectric
properties of these materials using chemical substitution according to the above discussion. This
research will implement these concepts in real Yb-based alloys in order to enhance the
thermoelectric properties in the heretofore unexplored cryogenic temperature range. While these
studies are in their nascent stages, it is hoped that the understanding gained from using this
approach will help one day to lead to useful Peltier coolers.

25

Chapter 3 Methodology
3.1 Synthesis
Sample preparations were different for each solid solution; therefore, a broad description of each
of the methods used in this research is given below. Further details and any alterations to the
process are discussed in the appropriate chapter for each solid solution.

3.1.1 Arc Melting
Stoichiometric quantities (unless otherwise noted in corresponding chapters) of pure elements
were weighed out using a Mettler Toledo AL54 digital scale to within ±0.0005g for each base
compound of the solid solutions. The pure elements were obtained from Sigma Aldrich or Alfa
Aesar, while some of the rare earth elements used in this research originated from Ames National
Laboratory. The appropriate quantities of elemental pieces were then placed in the trough of a
copper base, or hearth, for arc melting. Elements with higher vapor pressure and lower melting
temperatures (such as Yb) were placed at the bottom of the trough and high melting temperature
elements were arranged on top. This arrangement was implemented as an attempt to prevent
vaporization and non-homogeneous losses. The arc-melting chamber was first evacuated via
rotary pump and then backfilled to a positive pressure of 1-2atm with flowing argon gas. This
pump and purge process was completed at least 5 times to ensure a pure argon environment for
melting. A water-cooled tungsten electrode with a current of approximately 100 amperes was
struck against the copper base to create an arc. This arc was then slowly passed over the raw
elements in order to melt them together. After the molten piece cooled, the ingot was flipped and
remelted. Any extraneous chunks that did not participate in the first melt were moved close to
the large chunk and then melted together. After the preliminary melts in which all of the

26

elemental pieces were physically joined together, the ingot was flipped and re-melted four or
more times in order to increase the diffusion process and encourage homogeneity within the
ingot.

3.1.2 Homogenization
Some of the arc-melted ingots were not fully homogenized from multiple melts and therefore had
to undergo a high-temperature heat treatment. The arc-melted ingots were broken into small
chunks, or pulverized if possible, using pliers and/or a mortar and pestle. The resulting
chunks/powders were loaded into an alumina crucible, which was sealed under vacuum in a
carbon coated quartz tube. The quartz tubes were carbon coated by pyrolizing a small amount of
acetone in the bottom portion of the tube using a methane/oxygen torch. This torch was also used
to seal the quartz ampoules while under a vacuum of at least 10-5 Torr. These tubes were then
hung vertically in a Thermolyne 21100 tube furnace for a set temperature profile. Each material
required different temperature profiles as will be discussed in the corresponding chapters.

3.1.3 Chemical Etching
Several of the samples were made by using an aluminum flux processing technique, similar to
that used for single crystal growth [61,62]. In single crystals, the stoichiometric elements of the
compound are melted within a flux material, which can be an element in the compound or an
additional element that does not react with the elements in the compound. This process generally
takes several days in which the elements are heated above the melting temperature of the flux
and the desired compound and then slowly cooled to allow crystal growth. The flux material
generally has a much lower melting temperature than that of the desired compound, thus as the
sample is cooled the desired phase solidifies while the flux remains molten. After the crystal is
grown, the sample is held above the melting temperature of the flux and the excess flux is
27

removed by centrifuging or chemical etching. In this study, the aluminum flux technique was
used in an arc melter to achieve the desired phase in a polycrystalline sample rather than a single
crystal grown in a furnace. Sodium hydroxide (NaOH) was found to react selectively with Al,
while acting as an inert liquid to the desired phase, and was chosen as the chemical etchant. After
arc melting, the ingot was broken into small chunks in order to increase the rate of penetration of
5M NaOH into the entirety of the ingot. These chunks were laid in a small plastic mesh bin to
increase ease of access and drainage. Samples were left in the NaOH for at least 24 hours
beneath a ventilation hood. After 24 hours, the sample chunks were rinsed in a series of distilled
H2O baths to remove all remnant NaOH. This process was determined to be complete once the
pH level was 7 using pHydrion papers. The samples were then dried using a Thermolyne Hot
Plate Oven at 100oC. The dried samples were then ground using an agate mortar and pestle. The
powder was sieved through a 60mesh screen to remove any un-etched pieces. The un-etched
sample was much more difficult to grind and therefore the two compositions (desired and unetched) were very easy to separate physically. Un-etched pieces were returned to the NaOH bath
for another 24 hours and the process was repeated. The sieved powder was then characterized
and further processed to achieve the final sample.

3.1.4 Ball Milling
Some of the synthesis procedures, such as etching and high-temperature homogenization
required that powder-processing techniques be used. Although some samples were dense and
homogenous upon synthesis, all samples were powder processed in order to remain consistent.
Samples were initially ground by hand using a mortar and pestle prior to undergoing ball milling.
Stoichiometric quantities of each base compound were weighed out and then loaded and sealed
with parafilm under argon into a stainless steel vial containing six stainless steel ball bearings;
28

two with diameter 0.5in and four with diameter 0.25in. Ball milling was done in a high-energy
SPEX Sample Prep Mixermill 8000M vibratory mill. Milling times are reported in the following
chapters. After ball milling, the sample was removed from the jar and passed through a 60mesh
screen to remove any agglomerates. This ball milling process not only created well-mixed
powders, but also reduced the particle size, and in some cases even caused mechanical alloying
of the two compounds as established by x-ray diffraction (XRD).

3.1.5 Densification
Two different processes were used to densify the ball-milled powders. Primarily, pulsed electric
current sintering (PECS), also known as spark plasma sintering, was used, however some
samples were formed using traditional hot press methods. PECS processing was used because of
its high temperature capabilities combined with short processing times [29]. During the PECS
process, the sample powder is heated up within the die while pressure is applied and a current is
simultaneously pulsed through the sample. This current causes very localized heating within the
sample, which induces rapid sintering, making the densification process happen more quickly.
Samples were pressed in a GT Advanced Technologies (formerly Thermal Technology LLC)
SPS Model 10-4 system using a stainless steel rod in a graphite die while under an argon or
vacuum atmosphere. Different compositions required varying time and temperature profiles that
will be reported in the following chapters. Some samples were formed using hot pressing in
order to compare results and as an attempt to increase the purity of the sample in which oxides
were a problem. Further elaboration on the oxide impurities will appear in subsequent chapters.
The hot press used was a Thermal Technology Model HP200-14020-23G, which also used a
graphite die with a stainless steel rod under vacuum or argon. The only difference between hot
pressing and PECS is that there is no current pulsed through the sample during hot pressing.
29

PECS generally resulted in higher density samples in less time, and was therefore the primary
process used for densification of the samples reported in this work.

3.2 Characterization
While the primary focus of this thermoelectric research is to enhance the thermoelectric
properties of materials at low temperatures, it is critical to know the composition and purity of
the samples being measured. Without this information, it is impossible to draw conclusions about
the different effects that occur in the materials because of the many variables that can influence
thermal and electronic transport. Several characterization techniques are used to examine the
density and purity of each sample made.

3.2.1 Density
The density of each sample disk was measured after powder processing and densification, using
Archimedes’ method. This method uses the mass of the sample in air (MA), the mass of the
sample in 200 proof ethanol (ME), and the density of the ethanol (ρE) based on temperature (as
obtained from Lange’s Handbook of Chemistry, 10th edition [63]) to determine the sample
density (ρ) in g/cm3.

Equation 13
This measured value can be compared to the theoretical value for the given composition in order
to determine the percent density of the sample. The theoretical density can be obtained from
literature or calculated for each sample if the lattice parameter(s), mass of atoms and unit cell
structure are known. All samples that were studied for their thermoelectric properties had
measured densities of at least 95% of theoretical value.

30

3.2.2 X-ray Diffraction
Powder x-ray diffraction (XRD) was the main characterization technique used to identify phase
purity in samples. XRD was conducted using a Rigaku MiniFlexII Benchtop system using Cu Kα
radiation. Patterns were collected on each sample following every step in the synthesis process.
These patterns were then compared to existing powder diffraction files in the Jade9 database. Xray diffraction is considered a reliable method for detecting crystalline phases of greater than
three atomic percent of the total composition. It is possible to have secondary phases in a sample
that do not appear in the patterns but may affect the electronic properties or appear in other
characterization methods if this phase is only a small fraction of the total composition or exists in
an amorphous state. XRD patterns were used not only to detect impurity phases in the samples,
but also to calculate the lattice parameter. The plane corresponding to each peak was indexed and
using the d-spacing and the crystal structure, the lattice parameter(s) could be calculated. These
lattice parameters are important for comparing the average valence with the size of the unit cell.

3.2.3 Scanning Electron Microscopy
Scanning electron microscopy (SEM) was used to find the average grain sizes of densified
samples and to identify any impurity phases that could be observed. The microscope was a JEOL
JSM-6610LV from JEOL Ltd. Tokyo, Japan. Secondary electron images reported here show
fractured surfaces for different samples. While fractured surfaces are not ideal for measuring
grain sizes, this method proved most convenient and allowed for a rough approximation of the
average grain size. Backscatter mode was used to observe differences in the average atomic
weight, indicating areas of the sample that have different atomic composition. While these
images were gathered to qualitatively identify impurity phases, they are not reported in this
work. Finally, the different elements in the sample were mapped using energy dispersive x-ray
31

spectroscopy (EDS) using an Oxford Instruments Aztec system, software version 2.1a, using a
20mm2 silicon drift detector crystal and an ultrathin window. The resulting patterns were again
used to qualitatively and semi-quantitatively identify the elements that made up the different
phases but are not published in this report.

3.3 Property Measurement
3.3.1 Transport Properties
Measurement of the three thermoelectric transport properties (electrical resistivity ρ = 1/σ, the
Seebeck coefficient S, and thermal conductivity κ) was performed from 80 to 350 K under
vacuum in a Janis model ST-100 liquid nitrogen flow cryostat. Electrical resistivity was
measured with an uncertainty of approximately 3% using a four-probe DC technique. The
Seebeck coefficient and thermal conductivity are measured using a steady state technique in
which a small metal film resistor is affixed at one end of the sample while the other end attached
to a copper base which is then anchored to the cryostat cold tip. When this resistor is energized,
heat flows down the sample and a steady state temperature difference is developed, which is
sensed by a pair of copper-constantan thermocouples. These thermocouples are soldered to
copper contacts which are attached to the sample using a silver epoxy. The resistor is externally
wrapped in a copper foil to provide an electrically conductive pathway at the top of the sample.
The electrical conductivity is measured by passing a current through the sample and measuring
the voltage difference at the thermocouples. A schematic of this sample measurement
configuration is shown in Figure 3-1. From this, the lattice thermal conductivity can also be
determined by simple subtraction. All contacts to the sample are made using silver epoxy, and
thermal and electrical voltages are sensed using Keithley multimeters. From a set of
measurements on standards it was determined that this method gives an uncertainty in both
32

Seebeck coefficient and thermal conductivity of approximately 5%. The Wiedemann-Franz law,
as previously described, was used to calculate the electronic contribution to the thermal
conductivity.

IHeater

ISample

VH/TH
VS
Vc/TC

Figure 3-1: Sample mounting configuration for transport property measurements

Transport property measurements were extended down to 2K by colleagues at Ohio State
University using the thermal transport option (TTO) in a Physical Properties Measurement
System (PPMS) by Quantum Design. The results of this technique return similar accuracies as
that of the nitrogen flow cryostat reported above.

3.3.2 Magnetic Properties
To assist in the identification of the valence state of Yb, magnetization measurements were
performed on some of the samples. The magnetic susceptibility of select samples was determined
by measuring the magnetic moment of the sample in a Quantum Design superconducting
33

quantum interference device (SQUID) magnetometer in the Physics and Astronomy Department
at MSU. The moment was collected in a magnetic field of 5000 gauss from 2K to 300K. Small
pieces (typically <0.5g) were placed inside of a small plastic pouch and then slid into a plastic
straw. The straw was inserted into the SQUID chamber that was then evacuated and cooled to
2K. Thermal equilibrium was achieved by allowing the sample to sit at each temperature point
for approximately 15 minutes before the magnetic moment was measured.
The magnetic data collected on these samples had to be corrected for oxide impurities
found in the samples. This oxide phase contributed a paramagnetic signal that overwhelmed the
very low temperature features in the data. This impurity signal could be removed by calculating
the quantity of impurity by fitting the low temperature (2-10K) Curie tail from the inverse
moment versus temperature data to a line and subtracting it from the magnetic moment. Doing so
left a magnetic susceptibility curve with a peak at low temperatures, characteristic of
intermediate valence compounds.
After correction, this data was analyzed using the interconfigurational-fluctuation (ICF)
model, proposed by Sales and Wohlleben [40]. This model is used to fit the magnetic
susceptibility curve for Yb-based materials using two variables, the excitation energy, and the
spin fluctuation temperature. By fitting these two variables to the data, the average Yb valence
was extracted as a function of these variables. This information allowed for a direct comparison
the average Yb valence and the other data collected, including the Seebeck coefficient.

34

Chapter 4 YbAl3-Based Solid Solutions
4.1 Introduction
Yb-based compounds are very interesting for low temperature thermoelectric research due to
their large power factors at and below room temperature. One of the most well known Yb-based
materials is YbAl3, because it has the largest power factor reported for any bulk material at room
temperature [23,64–66]. This large power factor is a strong impetus to further study and pursue
YbAl3 as a potential thermoelectric material. While the thermoelectric properties have been
measured by many groups [64,65,67–69], minimal experimental work has been done to optimize
these thermoelectric properties. This is primarily due to the difficulty in obtaining large single
crystals of this compound, and synthesis of large polycrystals has gone essentially
uninvestigated. The large thermal conductivity in this compound must be reduced in order to
take advantage of the advantageous power factor. A simple but common approach to reduce this
thermal conductivity is via solid solution formation [25,31], taking advantage of the
aforementioned scattering of phonons by substituting atoms of differing masses.
First-principle calculations have shown that there are many isostructural alloys that could
be synthesized as full solid solutions with YbAl3. Substitution on the Al site could come in the
form of Mg, Cu, Zn, In and Sn [70] while Ho, Er, Tm [71], Sc [72] and Lu [72,73] could be
substituted on the Yb site. There is also a vacant octahedral site in the cubic structure that could
be filled with a small atom such as Mn, as was reported by He et al. [67]. Magnetic properties
have been collected on several of these alloys, revealing strong changes in the magnetic
susceptibilities as the Yb atoms are diluted. The low-temperature peak in the magnetic
susceptibility, which is characteristic of intermediate valence systems, is reduced and shifted

35

toward higher temperatures in both Lu [74] and Sc [72] substituted systems. These changes in
magnetic susceptibility indicate a change in the Yb valence and suggest the Seebeck coefficient
may be dramatically altered, although the Seebeck coefficient is not reported in the literature for
these compositions. The work described in this section investigates the effects of Sc, Er, and Lu
substitution on the thermoelectric properties of YbAl3 and has been published in Intermetallics
and the Journal of Electronic Materials [75,76]. Attempts were also made to form solid solutions
with In and Sn substituting for Al; however, no phase pure compounds were successfully
synthesized and thus could not be reported.
YbAl3 has a cubic crystal structure [77–80], as shown in Figure 4-1, which is relatively
simple in structure therefore has a very large lattice thermal conductivity. In order to reduce this
thermal conductivity, solid solutions are formed to introduce defects into the lattice and scatter
phonons. The three alloys investigated here are isostructural to YbAl3 and thus form complete
solid solutions. The synthesis procedure for each family of compounds was quite complex and
will be discussed in the following section. In order to overcome obstacles in the synthesis, solid

Figure 4-1: Unit cell diagram of YbAl3.
36

solutions were formed by first creating the individual binary compounds (YbAl3, ScAl3, ErAl3,
and LuAl3) and then using powder processing to create compounds of varying compositions.

4.2 Pure YbAl3
4.2.1 Synthesis
The nominal composition YbAl9 was arc melted in order to achieve an Al rich environment and
encourage the growth of the YbAl3 phase. A peritectic reaction is shown in the phase diagram
[79], see Figure 4-2, indicating that stoichiometric YbAl3 forms from two separates phases, solid
YbAl2 and liquid YbAl3+Al, upon cooling. As discussed in previous studies [64,65,81,82], this
YbAl2 phase is unavoidable by simply arc melting stoichiometric quantities. Assuming perfect
stoichiometery was maintained during the arc melting process, a prolonged heat treatment at just
below the peritectic temperature would allow for the homogenization of the YbAl3 phase.

Figure 4-2: The Ytterbium-Aluminum binary phase diagram by Gschneider and Calderwood
[79].
37

However, this is an extremely slow diffusion process as the YbAl2 phase has already crystallized
at this temperature and is not a realistic synthesis procedure. Thus, an Al rich environment was
created to force the compound to maintain the YbAl3 phase without allowing YbAl2 to form.
This approach of using an Al rich environment is commonly found in the literature for growing
single crystals [65,68,73,74,82,83], however, it is a novel approach to use a flux while arcmelting for bulk polycrystalline samples.
After arc-melting in the aluminum rich environment the sample was then etched in NaOH
as previously described in Section 3.1.3. This etching process removed all of the excess Al from
the sample, leaving only the desired YbAl3 phase, which was then easily powdered. Etching with
NaOH is also done in single crystal growth; however, the nature of single crystals prevents the
excess Al being incorporated into the bulk and is only found on the surface, and therefore does
not require breaking the single crystals into smaller pieces.
The powder resulting from the etching step was then ball milled for 40 minutes.
Densification was done using PECS at 800oC for 30 minutes under 60 MPa of pressure in an
argon atmosphere. Several different approaches were taken to densifying these samples in an
attempt to reduce the oxide phase that formed during sintering. Samples were processed using
shorter pressing times, lower temperatures, and under vacuum rather than argon; however none
of these precautions improved the quality of the samples. Hot pressing, rather than PECS, also
proved futile in an attempt for higher purity samples.

4.2.2 Characterization
YbAl3 forms in the Cu3Au or L12 crystal structure, which is a cubic structure with Yb on the
corners and Al on the faces as previously shown in Figure 4-1. XRD was used at each step in the

38

synthesis procedure to determine the phases present in the sample and ensure consistency
throughout many batches. Figure 4-3 demonstrates the XRD patterns that result from a single
batch of YbAl3 at each step in the processing chain. The arc melted nominal YbAl9 composition
(Figure 4-3A) shows the characteristic peaks of YbAl3 in addition to peaks corresponding to pure
Al metal indicated by *, as expected from the Al rich arc-melting environment. After etching, the
powder shows sharp peaks representative of only the desired YbAl3 phase (Figure 4-3B). After
ball milling the pattern does not show any additional peaks, however the characteristic peaks are

(321)

(320)

(222)

(311)

(310)

(300)

(220)

(211)

D

C

B

*
*

A
10

(210)

(200)

(111)

(110)

(100)

Normalized Intensity (Arbitrary Units)

broadened (Figure 4-3C). This broadening could be a result of several factors including reduced

20

30

40

*
50
2θ

60

*
70

80

90

Figure 4-3: XRD patterns of each step in the YbAl3 synthesis process as labeled: A) after the
nominal YbAl9 was arc melted - * indicate peaks from the pure Al phase, B) after sample is
etched in NaOH to remove excess Al, C) after YbAl3 is ball milled, D) after densification
using PECS – unlabeled peaks are oxide impurity phase.

39

particle size, defects in the crystal structure such as dislocations, point defects or concentration
gradients, and finally strain, which may occur from the high energy ball milling process. After
densification, the peaks narrow again (Figure 4-3D), suggesting grain growth, reduced strain
and/or homogenization of the crystal structure during the sintering step. The XRD pattern also
reveals several un-desired peaks that can be indexed to a Yb-oxide phase. The oxide phase was
resilient to several different synthesis techniques and the samples were used as the baseline for
the solid solution.

4.2.3 Transport Properties
The thermoelectric properties of YbAl3 have been reported in several different papers
[64,65,67,83], each showing very large power factors. The data collected in this work is
consistent with the literature data, and is reported in each of the sections below for comparison
for the solid solutions.

4.3 Yb1-xScxAl3
The scandium atom is ideal for creating a solid solution because although it forms the
isostructural compound ScAl3, there is a large mass mismatch between Sc and Yb [72,73,80]. A
full solid solution of Yb1-xScxAl3, where x=0, 0.25, 0.50, 0.75, and 1.0, was created to study the
effects of the solid solution on the lattice thermal conductivity and overall thermoelectric figure
of merit.

4.3.1 Synthesis
The YbAl3 base compound was synthesized as described above, separately from ScAl3. .
According to the phase diagram[84] shown in Figure 4-4, the ScAl3 phase forms congruently and
thus the stoichiometric quantities of Sc and Al were arc melted. The resulting ingot was then
40

powdered and heat-treated for 24 hrs at 900oC to ensure homogeneity The arc melted sample
showed no obvious signs of phase separation or impurities. Stoichiometric quantities of YbAl3
and ScAl3 were then ball milled together for 40 minutes, sieved, and densified. PECS processing
was done for 30 minutes at 800oC under 60MPa pressure in an argon atmosphere.

Figure 4-4: The Sc-Al phase diagram showing the congruent melting of the ScAl3 phase [84].

4.3.2 Characterization
Figure 4-5 shows the single phase XRD patterns from the Yb1-xScxAl3 solid solution with minor
oxide impurities, consistent with those observed in the pure YbAl3 samples. These XRD patterns
41

also indicate that the oxide impurity phase is reduced as Sc is increased in the system, indicating
that Yb has a higher tendency to oxidize in lower oxygen content environments than Sc.
Majewski et al. reported that single-phase samples were not achieved for compositions
0.5<x<1.0. However, the entire range of concentrations x = 0, 0.25, 0.50, 0.75, and 1.0 were
successfully synthesized in this work, except for a minor Yb-oxide impurity phase. The patterns
in Figure 4-5 show that the crystal symmetry is not altered by the addition of Sc into the YbAl3
system; however, the lattice parameter is affected as expected from isostructural alloying. The
XRD patterns show a shift in the peaks toward larger angles, indicating that as the Sc
concentration is increased, the lattice parameter is reduced. This can be more clearly seen by
calculating the lattice parameter from these peaks and plotting versus Sc concentration, and is
displayed in Figure 4-6. Literature data for 0≤x≤0.90 [72] and x=1.0 [80] are represented by the
closed circles and show good agreement with the lattice parameters calculated in this study.
The end compounds of this solid solution (x = 0 and 1.0) were investigated via SEM.
Secondary electron images in Figure 4-7 show grain size to be on the order of several microns.
EDS maps of the sample, not included here, also qualitatively confirmed the oxide impurity
phase that was identified by XRD.

42

(222)

(311)

(310)

(300)

(220)

(211)

(210)

(200)

(111)
(110)

(100)

Normalized Intensity (Arbitrary Units)

E

D

C

B

A
10

20

30

40

50
2θ

60

70

80

90

Figure 4-5: XRD pattern for the Yb1-xScxAl3 solid solution where A) x = 0.0, B) x = 0.25, C)
x = 0.50, D) x = 0.75, E) x = 1.0 – unlabeled peaks are from the oxide impurity phase.

Lattice Parameter (Å)

4.22
4.2

4.18
4.16
4.14
4.12
4.1
4.08
0

0.25

0.5
Sc concentration

0.75

Figure 4-6: Lattice parameters for the Yb1-xScxAl3 solid solution where 0≤x≤1 at room
temperature. Closed symbols come from literature data [79, 81].
43

1

Figure 4-7: SEM images of fractured surfaces for A) YbAl3 and B) ScAl3.

44

4.3.3 Transport Properties
The thermoelectric transport properties were measured on each sample in the solid solution using
the techniques described in Section 3.3.1. While the primary purpose of creating a solid solution
was to reduce the lattice thermal conductivity, strong changes were also observed in the
electronic properties. As shown in Figure 4-8, the electrical resistivity of the YbAl3 sample
rapidly increases from low temperatures until approximately 300K where the curve plateaus. The
trend and magnitude of the resistivity for this sample closely matches reported values in the
literature by Rowe et al [83]. The addition of scandium causes the resistivity to decrease greatly
and display a more linear temperature dependence. In fact, the resistivity of the pure ScAl3
sample is comparable to that of many good metals, at approximately 11 μΩcm at room
temperature.

Electrical Resistivity (μΩcm)

140
x=0.0

120

x=0.25
x=0.50

100

x=0.75
80

x=1.0

60
40
20
0
0

50

100

150
200
250
Temeprature (K)

300

350

400

Figure 4-8: Electrical resistivity from 80K to 350K for the Yb1-xScxAl3 solid solution where
x = 0.0, 0.25, 0.5, 0.75, and 1.0. Dashed lines are guides to the eye.

45

0
x=0.0

Seebeck Coefficient (μV/K)

-10

x=0.25

-20

x=0.50

-30

x=0.75

-40

x=1.0

-50
-60
-70
-80
-90
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-9: Seebeck coefficient from 80K to 350K for the Yb1-xScxAl3 solid solution where
x = 0.0, 0.25, 0.5, 0.75, and 1.0

Figure 4-9 shows a Seebeck coefficient for YbAl3 that is consistent with other bulk
polycrystalline samples reported by Rowe et al. [83]. This Seebeck coefficient decreases as Sc
concentration increases. However, this reduction in the Seebeck coefficient does not occur as
rapidly as the reduction in electrical resistivity, particularly for the x=0.25 composition which is
barely altered. However, at the x=1.0 composition, the Seebeck coefficient is essentially zero,
which is again characteristic of a good metal. The rapid reduction in the electrical resistivity and
the relatively slow reduction of the Seebeck coefficient with respect to Sc concentration results
in significant enhancements in the power factor. The power factor is increased across the entire
temperature range for the x=0.25 composition, while the x=0.50 and 0.75 compositions show
moderate enhancements in the power factor above 130K and 200K respectively, as shown in
Figure 4-10.
46

120

Power Factor (μW/cmK2)

x=0.0
x=0.25

100

x=0.50
80

x=0.75
x=1.0

60
40
20
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-10: Power Factor from 80K to 350K for the Yb1-xScxAl3 solid solution where x =
0.0. 0.25, 0.50, 0.75, and 1.0.
The purpose of creating a YbAl3-ScAl3 solid solution is to take advantage of the mass
mismatch between Yb and Sc, in order to reduce the lattice thermal conductivity. The difference
between the atomic masses of Yb (173.05amu) and Sc (44.96amu) should cause a large increase
in the phonon scattering rate, according to Equation 12. The nearly metallic resistivity results in
a very large electronic component to the thermal conductivity that grows rapidly as the electrical
resistivity is dramatically reduced with increasing Sc content. However, the lattice thermal
conductivity is reduced for x=0.25, 0.50, and 0.75, creating an opposing effect on the total
thermal conductivity. Figure 4-11 shows that the total thermal conductivity is increased for all
concentrations of scandium because the reduction in the lattice thermal conductivity is not large
enough to counteract the increase in the electronic component. Fortunately, the increase in
thermal conductivity is very minimal in the x=0.25 sample.

47

Thermal Conductivity (W/cmK)

0.90
x=0.0

0.80

x=0.25

0.70

x=0.50

0.60

x=0.75

0.50

x=1.0

0.40
0.30
0.20
0.10
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-11: Total thermal conductivity from 80K to 350K for the Yb1-xScxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0.
0.25
x=0.0
x=0.25

0.20

x=0.50
x=0.75

0.15
zT

x=1.0
0.10

0.05

0.00
0

50

100

150
200
250
Temperature (K)

300

350

Figure 4-12: Dimensionless figure of merit from 80K to 350K for the Yb1-xScxAl3 solid
solution where x = 0.0, 0.25, 0.50, 0.75, and 1.0.
48

400

The enhanced power factor in combination with the minimally increased thermal
conductivity of the x=0.25 sample results in an overall enhancement of the zT over the entire
temperature range. As can be seen in Figure 4-12, the zT is increased by nearly 40% at room
temperature. This is a significant improvement to this widely studied low-temperature
thermoelectric material. Unfortunately, while the large electrical conductivity allows for a very
large power factor, this conductivity also contributes to the large thermal conductivity, which is
the limiting factor in further improving the thermoelectric efficiency of this material.

4.4 Yb1-xErxAl3
While Sc was an ideal element for reducing lattice thermal conductivity because of the large
mass mismatch, it also strongly affected the electronic properties. The Seebeck coefficient and
resistivities became more metallic as Sc was added. Another substitutional atom with a small
mass mismatch but more similar electronic properties is erbium. Erbium also forms the
isostructural compound ErAl3 that is used to create a full solid solution. No experimental work
on the YbAl3-ErAl3 system was found in the literature; however, first-principle calculations
show that a full solid solution between these two compounds is possible and the electronic
structure should not be significantly altered [71].

4.4.1 Synthesis
Similarly, to the Yb-Al phase diagram, ErAl3 shown in Figure 4-13 [85] also undergoes a
peritectic reaction and therefore an Al rich environment is necessary to achieve a phase-pure
sample. The same synthesis technique of arc melting, etching and powder processing, as
described for the YbAl3 compound was utilized to synthesize ErAl3, this time using the nominal
composition ErAl5. A full solid solution of Yb1-xErxAl3 was synthesized for x=0.0, 0.25, 0.50,
0.75, and 1.0.
49

Figure 4-13: The Erbium-Aluminum binary phase diagram by Gschneider and Calderwood
[85].

4.4.2 Characterization
The XRD patterns clearly show the L1 crystal structure for the densified samples for all
compositions. The shift in the characteristic XRD peaks is significantly smaller than that of the
Sc samples, because Er is much closer in size to Yb than Sc, so the lattice parameter is not
significantly altered. The XRD patterns shown in Figure 4-14 reveal that there is an abrupt and
dramatic increase in the oxide phase of the x = 0.75 and 1.0 samples. While it is believed that the
role of this oxide impurity phase is minimal for low Er concentration samples, it is likely the
transport properties are influenced by the impurity phase in the x = 0.75 and 1.0 samples. The
50

(321)

(222)
(320)

(311)

(310)

(300)

(220)

C

(211)

(210)

(200)

(111)

(110)

D
(100)

Normalized Intensity (Arbitrary Units)
10

E

B
A
20

30

40

50
2θ

60

70

80

90

Figure 4-14: XRD pattern of the Yb1-xErxAl3 solid solution at room temperature where A) x =
0, B) x = 0.25, C) x = 0.50, D) x = 0.75. and E) x = 1.0.

Figure 4-15: SEM images of fractured surfaces for ErAl3.
51

thermoelectric properties were measured on these samples, but they reflect this higher-thannormal oxide phase and are not reported in all cases.
Figure 4-15 shows the SEM micrograph for the ErAl3 fractured surface. This micrograph
shows a larger grain structure than that of YbAl3 and ScAl3; however, grains are still on the
order of several microns in diameter.

4.4.3 Transport Properties
The electrical resistivity of the samples was strongly decreased by the addition of Er. However,
samples with Er concentrations x=0.75 and 1.0 had higher than expected electrical resistivity, as
shown in Figure 4-16, which is suspected to arise from the large insulative oxide phase that
appears in the XRD patterns. For this reason, the transport properties of these two compositions
are not associated with phase pure samples and not reported for the rest of the thermoelectric
properties.

Electrical Resistivity (μΩcm)

140
x=0.0

120

x=0.25
x=0.50

100

x=0.75

80

x=1.0

60
40
20
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-16: Electrical resistivity from 80K to 350K for the Yb1-xErxAl3 solid solution where
x = 0.0, 0.25, 0.50, 0.75, and 1.0.
52

0

x=0.0

Seebeck Coefficient (μV/K)

-10

x=0.25

-20

x=0.50

-30
-40
-50
-60
-70
-80
-90
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-17: Seebeck coefficient from 80K to 350K for the Yb1-xErxAl3 solid solution where
x = 0.0. 0.25, and 0.50.

Figure 4-17 shows that the Seebeck coefficient was also reduced with the addition of Er
into the system, but not as strongly for the x=0.50 sample as occurred in the Sc substituted
x=0.50. This rapid decrease in the electrical resistivity and small drop in the Seebeck coefficient
results in a gigantic power factor value of nearly 140 μW/cmK2 between 120K and 300K for
both the x=0.25 and 0.50 samples. Figure 4-18 shows that Er substitution nearly doubles the
original YbAl3 power factor.
With a very promising power factor, the thermal conductivity of these samples is of
utmost importance. The reduction to the lattice thermal conductivity is not expected to be very
large in comparison to the Sc substituted samples because the mass difference between Er and
Yb is quite small. However some reduction in the lattice component is revealed above 150K and

53

160

x=0.0
Power Factor (μW/cmK2)

140

x=0.25

120

x=0.50

100
80
60
40
20
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Lattice Thermal Conductivity (W/cmK)

Figure 4-18: Power factor from 80K to 350K for the solid solution Yb1-xErxAl3 where x =
0.0, 0.25, and 0.50.
0.12
x=0.0
x=0.25

0.10

x=0.50
0.08
0.06
0.04
0.02
0.00
0

50

100

150
200
250
Temperature (K)

300

350

Figure 4-19: Lattice thermal conductivity from 80K to 350K for the Yb1-xErxAl3 solid
solution where x = 0.0, 0.25, and 0.50.
54

400

200K for the x=0.50 and 0.25 samples respectively, as can be seen in Figure 4-19. Unfortunately,
the resistivity of these Er substituted samples is so rapidly reduced that the electronic component
of the thermal conductivity is quite large and overwhelms any reduction of thermal conductivity
from the lattice component. The total thermal conductivity is increased for all Er concentrations
and is shown in Figure 4-20.

Thermal Conductivity (W/cmK)

0.35
x=0.0
0.30

x=0.25

0.25

x=0.50

0.20
0.15
0.10
0.05
0.00
0

50

100

150

200

250

300

350

400

Temperature (K)
Figure 4-20: Total thermal conductivity from 80K to 350K for the Yb1-xErxAl3 solid solution
where x = 0.0, 0.25, and 0.50.

Combining the strongly enhanced power factor with the increased thermal conductivity
results in an enhancement in the zT value for the Er x = 0.25 sample. The thermal conductivity of
the x = 0.50 sample is too large to achieve an increase in the zT despite the doubled power factor,
as can be seen in Figure 4-21. While higher Er concentration samples are not reported here, the
trends indicate that each of the thermoelectric properties would become metallic, thus reducing
power factor while increasing thermal conductivity. Therefore there is no reason to expect that
55

there would be any enhancement in the zT for high concentration Er samples.
0.20
0.18

x=0.0

0.16

x=0.25
x=0.50

0.14
zT

0.12
0.10
0.08
0.06
0.04
0.02
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-21: The dimensionless figure of merit from 80K to 350K for the Yb1-xErxAl3 solid
solution where x = 0.0, 0.25, and 0.50

4.5 Yb1-xLuxAl3
A third element, lutetium, was also used to create a solid solution with YbAl3. This solid solution
was synthesized with the goal of reducing the lattice thermal conductivity without having
significant affects on the electronic properties. Similarly to the previous solid solutions, LuAl3
and YbAl3 are isostructural compounds, thus forming a full solid solution. Previous researchers
collected magnetic data on Yb1-xLuxAl3 single crystals in order to better understand and analyze
the theoretical explanations of the YbAl3 intermediate valence system [74]. While this study
shows that a full solid solution is possible for this system, no bulk polycrystalline samples have
been made and no thermoelectric properties were measured.

56

4.5.1 Synthesis
Similarly to the Yb-Al phase diagram, the Lu-Al phase diagram [86] shown in Figure 4-22
demonstrates a peritectic reaction, although the transition temperatures have not been accurately
determined in this case. In order to achieve a phase pure sample an Al-rich environment was
again utilized to synthesize the phase pure LuAl3, starting with the nominal LuAl5 composition.
By arc-melting followed by etching, pure LuAl3 was synthesized and subsequently ball milled
together with YbAl3 powder to create Yb1-xLuxAl3 where x = 0.0, 0.25, 0.50, 0.75, and 1.0.

Figure 4-22: The Lutetium-Aluminum binary phase diagram by H. Okamoto [86].

57

4.5.2 Characterization
LuAl3 forms in the L12 crystal structure as confirmed by the XRD pattern. The XRD patterns,
shown in Figure 4-23, also revealed the same small oxide impurity peaks as seen in the other
YbAl3 solid solutions but are not as overwhelming as those seen in the Yb1-xErxAl3 solid
solution. No shift is observed in the characteristic peaks because the change in the lattice
parameter is minimal due to the similarity in size of Yb and La.
The SEM micrograph, shown in Figure 4-24, shows grain sizes on the order of several
microns. This micrograph is very similar to those of the YbAl3 and ScAl3 other base

10

(320)
(321)

(222)

(310)
(311)

E

(300)

(220)

(211)

(210)

(200)

(111)

(110)

(100)

Normalized Intensity (Arbitrary Units)

compounds.

D
C
B
A
20

30

40

50
2θ

60

70

80

90

Figure 4-23: XRD pattern of the Yb1-xLuxAl3 solid solution at room temperature where A) x
= 0.0, B) x = 0.25, C) x = 0.50, D) x = 0.75, and E) x = 1.0.

58

Figure 4-24: SEM images of fractured surfaces for LuAl3.

4.5.3 Transport Properties
Substituting even a small amount of Lu for Yb causes a dramatic decrease in the electrical
resistivity. Similar to the Sc and Er solid solutions, at x=1.0, the resistivity (Figure 4-25) is as
low as a very good metal. The magnitude of the Seebeck coefficient also decreases with
increasing Lu content as can be seen in Figure 4-26. This decrease in Seebeck coefficient is not
as rapid as that of the electrical resistivity, therefore there is an increase in the power factor for x
= 0.25 and 0.50 across the entire temperature range as can be seen in Figure 4-27. This power
factor for the Lu x = 0.25 sample is nearly double that of pure YbAl3, replicating the results
achieved by the Er solid solution.

59

Electrical Resistivity (μΩcm)

140
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

120
100
80
60
40
20
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-25: The electrical resistivity from 80K to (300K) 350K for Yb1-xLuxAl3 where
x = 0.0, 0.25, 0.50, 0.75, and 1.0.
0
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

Seebeck Coefficient (μV/K)

-10
-20
-30
-40
-50
-60
-70
-80
-90
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-26: The Seebeck coefficient from 80K to (300) 350K for the Yb1-xLuxAl3 solid
solution where x = 0.0, 0.25, 0.50, 0.75, and 1.0.
60

160
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

Power Factor (μW/cmK2)

140
120
100
80

60
40
20
0
0

50

100

150
200
250
Temeperature (K)

300

350

400

Figure 4-27: The power factor from 80K to (300K) 350K for the Yb1-xLuxAl3 solid solution
where x = 0.0, 0.25, 0.50, 0.75, and 1.0.

While the reduction to the electrical resistivity helps to enhance the power factor, it also
causes an increase in the electronic component of the thermal conductivity. The Lu substitution,
does however induce some phonon scattering and reduces the lattice component of the thermal
conductivity particularly at higher temperature, as shown in Figure 4-28. With the reduction in
lattice thermal conductivity offsetting only some of the increase in the thermal conductivity from
the electronic contribution, the total thermal conductivity, plotted in Figure 4-29, is increased for
all concentrations of Lu.
Fortunately, the increase in the power factor is large enough to overcome the increase in
the thermal conductivity for the Lu x=0.25 sample and the zT value is increased across the entire
temperature range. As seen in Figure 4-30, the increase in the thermal conductivity for all other
concentrations of Lu is too large and the zT is reduced.
61

Lattice Thermal Conductivity (W/cmK)

0.1
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-28: The lattice thermal conductivity from 80K to (300K) 350K for Yb1-xLuxAl3
where x = 0.0, 0.25, 0.50, 0.75, and 1.0.

Thermal Conductivity (W/cmK)

0.5
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-29: The total thermal conductivity from 80K to (300K) 350K for the Yb1-xLuxAl3
solid solution where x = 0.0, 0.25, 0.50, 0.75, and 1.0.
62

0.2
x=0.0
x=0.25
x=0.50
x=0.75
x=1.0

0.18
0.16
0.14
zT

0.12
0.1
0.08
0.06
0.04
0.02
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 4-30: The thermoelectric dimensionless figure of merit from 80K to (300K) 350K for
Yb1-xLuxAl3 where x = 0.0, 0.25, 0.50, 0.75, and 1.0.

4.6 Summary
This chapter has demonstrated a new synthesis method for tri-aluminide compounds that allows
for the fabrication of large polycrystals of YbAl3 and related compounds for thermoelectric
measurements. This work has investigated how substitution on the Yb site of YbAl3 affects the
thermoelectric transport properties. Generally speaking, isoelectronic substitutions create alloys
that are more metallic in electrical and thermal conductivity, while the Seebeck coefficient is
minimally altered for very small substitutional concentrations. This small reduction in the
Seebeck coefficient and large drop in the electrical resistivity is a general characteristic of this
family of compounds and results in enhanced power factor values. The already large power
factor achieved by YbAl3 is doubled for the Er and Lu solid solutions where x=0.25 and

63

enhanced across the entire temperature range for each solid solution where x≤0.50. However,
the reduced electrical resistivity also caused a large increase in the electronic contribution to the
total thermal conductivity, which could not be completely compensated for by the reduction in
the lattice thermal conductivity. While Sc substitution created the largest reduction in lattice
thermal conductivity, the large total thermal conductivity is still the limiting factor in these
compositions and must be further reduced in order to achieve additional improvement in the zT.
The thermoelectric figure of merit is enhanced for x = 0.25 samples in each of the solid
solutions, while the x ≥ 0.50 samples are limited by the increased thermal conductivity. ZT
values between 0.18 and 0.20 were achieved for compositions with x = 0.25 for each alloy of Sc,
Er and Lu with Yb, which is an enhancement of approximately 25% when compared to the pure
YbAl3 polycrystalline sample.
Creating solid solutions between YbAl3 and (Sc, Er, Lu)Al3 resulted in enhance zT
values, however there is still further optimization that must be done in order to achieve a material
that is adequate for low-temperature thermoelectric applications. Reduction of the thermal
conductivity of these materials would be necessary to achieve a material with a zT feasible for
applications. Two potential approaches for further study are reducing grain size and substituting
on the Al site. While this work used high energy ball milling to reduce particle size, neither this
process nor the pressing procedure were optimized to maintain small grain sizes in the final
densified puck. By reducing the final grain size, it is possible to further scatter phonons without
disrupting the electronic transport properties. Isoelectronic substitution on the Al site would
create additional scattering sites within the lattice, providing a means to further reduce the lattice
thermal conductivity. As mentioned in the introduction of this chapter, attempts were made to
synthesize YbIn3 and YbSn3 compounds for substitution; however, phase-pure samples were not
64

achieved. There is theoretical evidence that these phases should form and create a full solid
solution with YbAl3; however, it appears that they may require single crystal synthesis, which is
beyond the scope of this work. Substitution on the Al site may also have a dissimilar affect on
the electronic properties than the Yb site substitution, potentially creating a means to tailor the
electrical resistivity and limit the electronic contribution to the thermal conductivity.
The thermoelectric properties of the well-known intermediate valence compound YbAl3
were improved by forming isoelectronic solid solutions with ScAl3, ErAl3, and LuAl3. This work
demonstrated that not only is this compound promising for low-temperature thermoelectrics, but
other Yb-based compounds may also be enhanced by simple solid solution alloying. The
following chapters will further investigate Yb-based compounds and the effects of solid solution
alloying on the thermoelectric properties due to changes in the Yb valence via magnetic
investigations.

65

Chapter 5 YbAl2-Based Solid Solutions
5.1 Introduction
YbAl2 is an intermediate valence compound that has been studied for the effects of external and
internal pressures on the average Yb valence [45,46,49,50,53]. The valence of the Yb atom,
which is approximately 2.2 at room temperature for the base YbAl2 composition, can be easily
affected by chemical substitution. By substituting a smaller atom on the Yb site, the size of the
unit cell can be reduced and the Yb atom is forced into a higher valence state with a smaller
radius. The atomic structure of this compound is the C15 cubic Laves MgCu2 structure
[45,50,79,87,88] as shown in Figure 5-1. This structure is more complex than the YbAl3
structure, thus it is expected to have an intrinsically lower lattice thermal conductivity.

Figure 5-1: Unit cell diagram of YbAl2.

66

While the relationship between Yb valence and magnetic properties has been heavily
investigated, no research has been done on the effect of these valence changes to the Seebeck
coefficient and other thermoelectric properties. YbAl2 itself is a very poor thermoelectric
material because of its nearly metallic Seebeck coefficient. However, if the average Yb valence
and the Seebeck coefficient are directly connected as hypothesized, it may be possible to
increase this Seebeck coefficient through chemical pressure. This work investigates the changes
in thermoelectric properties as Sc is substituted for Yb in YbAl2 and the relationship between
these changes and the average Yb valence as calculated via magnetic properties. The results
reported here have been previously published in the Journal of Applied Physics [89].

5.2 Yb1-xScxAl2
The solid solution between YbAl2 and ScAl2 is ideal because there is not only a large mass
mismatch that can cause a reduction in the lattice component of the thermal conductivity, but Sc
and Yb also have a large size mismatch which can be used to tailor the average Yb valence [49].
It is well understood that intermediate valence systems show changes in the Seebeck coefficient
depending on the magnetic state of the f-ions, or in this case the Yb atoms [33,34]. Substitution
on this site causes variation in the magnetic susceptibility through compression or expansion of
the unit cell, resulting in a change in the average Yb valence and the Seebeck coefficient. There
is also a large mass mismatch between the Sc and Yb atoms, as previously discussed in Section
4.3, which have the potential to induce a large reduction in the lattice thermal conductivity. Thus,
chemically substituting Sc for Yb could be used to both optimize the Seebeck coefficient as well
as reduce the lattice thermal conductivity through phonon scattering. Previous research has
investigated the average Yb valence [49] and the magnetic susceptibility [57] of these solid

67

solutions, however these properties have not been experimentally compared to the Seebeck
coefficient and other thermoelectric properties.

5.2.1 Synthesis
The high vapor pressure of Yb makes synthesizing pure YbAl2 quite difficult, as maintaining the
stoichiometric amount of Yb is nearly impossible. Just as excess Al was used to make YbAl3, in
this case excess Yb was necessary to force the molten material to crystallize in the YbAl2 phase
due to the volatility of pure Yb metal. The nominal composition YbAl was arc melted only
enough to combine the two pure materials and begin the alloying process. The sample was then
subjected to a heat treatment that allowed the YbAl2 phase to homogenize while the excess Yb
vaporized. This was done by sealing the sample in a tube as described in Section 3.1.2, and
heating it to 980oC at a rate of 0.3oC/min and then holding at that temperature for 24 hours. The
tube was then quenched in a room temperature water bath. This process resulted in a high quality
sample that could be easily pulverized and used to create the solid solution with ScAl2. The
ScAl2 binary compound was synthesized via direct arc melting of the stoichiometric quantities of
Sc and Al metals. The ScAl2 sample was then pulverized and mixed with the YbAl2 binary
compound in stoichiometric quantities. The binary compounds were ball milled together for 20
minutes before they were sieved through a 60-mesh screen and densified. PECS was performed
for 30 minutes under 60 MPa of pressure at 800oC. This PECS procedure resulted in densities of
at least 95% for all compositions with the exception of pure YbAl2, which was not synthesized.
This sample was ignored as there is literature data on the thermoelectric properties, none of
which are particularly interesting [90,91]. The values given in the literature [90] are however
used for the following discussions in order to provide a basis for comparison. Large

68

concentrations of Sc gave the most interesting thermoelectric results, thus Yb1-xScxAl2 samples
with concentrations of x = 0.40, 0.60, 0.65, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0 were formed.

5.2.2 Characterization
The phase purity of the samples was confirmed at each step in the process as can be seen in
Figure 5-2, which shows the step-by-step XRD patterns of the YbAl2 synthesis process. In these
patterns, excess Yb peaks are clearly visible after the arc melting process but then disappear after
the heat treatment process leaving only the desired YbAl2 phase. The peaks are indexed
according to the Miller indices observed via Gӧrlach et. al [88]. The XRD patterns for the
densified samples of the solid solutions are shown in Figure 5-3. A clear shift in the peaks
indicate that the unit cell is shrinking as Sc concentration is increased, which is expressed more
clearly in terms of the unit cell volume in Figure 5-4 as a function of Sc concentration. The

10

(444)
(551)

(440)
(531)

(511)

(422)

(331)

(620)
(533)

B

(400)

(222)

(311)

(220)
(111)

Relative Intensity (arbitrary units)

values for the unit cell volume measured in this work match quite well with the data in the

A
20

30

40

50
2ϴ

60

70

80

90

Figure 5-2: The XRD patterns at room temperature following the each of the steps in
synthesizing YbAl2. A) after arc melting nominal YbAl where unlabeled peaks indicate a pure
Yb phase, and B) pure YbAl2 phase after 24 hour heat treatment.
69

(620)
(533)
(444)
(551)

(440)
(531)

(511)

(422)

(331)

G

(222)
(400)

(220)

(111)

Relative Intensity (Arbitrary uits)

(311)

literature [57] which are represented by the filled circles.

F

E
D
C
B
A

10

20

30

40

50
2ϴ

60

70

80

90

Figure 5-3: The XRD patterns for the Yb1-xScxAl2 solid solution where A) x = 0.60, B) x =
0.70, C) x = 0.80, D) x = 0.85, E) x = 0.90, F) x = 0.95, and G) x = 1.0.

Unit cell Volume (Å3)

500
490
480
470
460
450
440
430

0

0.2

0.4
0.6
Sc concentration (x)

0.8

1

Figure 5-4: The unit cell volume as a function of Sc concentration at room temperature for
Yb1-xScxAl2 where x = 0.0≤x≤1.0. Closed circles are data from literature [57].
70

A

B

Figure 5-5: SEM images of fractured surfaces for A) Yb0.60Sc0.40Al2 and B) ScAl2.

71

The microstructures of the x=0.40 and x=1.0 samples were examined using SEM. These images
are shown in Figure 5-5. Grains on the order of several microns are observed, similarly to those
previously discussed in the tri-aluminide series. Small impurity phases were qualitatively
identified by EDS as a Yb3Al5O12 impurity phase. This phase is assumed to be rather
insignificant towards the transport properties; however, it is taken into account later when
examining the magnetic properties.

5.2.3 Transport Properties
The thermoelectric properties were measured on each of the samples in the solid solution from
80K to 350K. Several of the samples showed interesting trends at low temperatures and thus
were measured by collaborators at Ohio State University from 2K to 300K. Literature data [90]
was used for the x=0 sample for comparison to the samples measured in this study.

Electrical Resistivity (μΩcm)

50

40

30

20
x=0
x=0.70
x=0.85
x=0.95

10

x=0.60
x=0.80
x=0.90
x=1.0

0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 5-6: The electrical resistivity from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid
solution where x = 0 [93], 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0.
72

The electrical resistivity has a non-zero minimum for each of the samples measured at
low temperatures shown in Figure 5-6. This trend is reminiscent of a Kondo-like behavior, but
this detail is not explored further here. Above 100K the electrical resistivity follows an
increasing linear trend with similar slopes for each of the compositions, with the exception of x =
0. The electrical resistivity increases with increasing Sc content up to x = 0.80, before it reaches
a maximum and starts decreasing. The pure ScAl2 sample has the lowest electrical resistivity of
the solid solution, similar to the metallic ScAl3 compound in the Yb1-xScxAl3 solid solution.
The Seebeck coefficient also shows very interesting trends at low temperature with high
concentration of Sc. The Seebeck coefficient reaches its largest magnitude of about 50μV/K at
approximately 40K when the Sc concentration is x = 0.90. The Seebeck coefficient plotted in

Seebeck Coefficient (μV/K)

0
-10
-20
-30
-40
x=0
x=0.70
x=0.85
x=0.95

-50
-60
0

100

200
Temperature (K)

x=0.60
x=0.80
x=0.90
x=1.00

300

400

Figure 5-7: The Seebeck coefficient from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid
solution where x = 0 [93]. 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0
Figure 5-7 shows an increase in the magnitude as well as a shift towards lower temperatures with
increasing Sc content, except for x=0.95, where the magnitude is reduced from the x=0.90
73

175
x=0
x=0.70
x=0.85
x=0.95

Power Factor (μW/cmK2)

150
125

x=0.60
x=0.80
x=0.90
x=1.0

100

75
50
25
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 5-8: The power factor from 2K (80K) to 300K (350K) for the Yb1-xScxAl2 solid
solution where x = 0 [93], 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0.

Total Thermal Conductivity (W/cmK)

0.35
0.30
0.25
0.20
0.15
0.10
0.05

x=0.60

x=0.70

x=0.80

x=0.85

x=0.90

x=0.95

0.00
0

50

100

150
200
250
Temperature (K)

300

350

Figure 5-9: The total thermal conductivity from 2K (80K) to 300K (350K) for the Yb1xScxAl2 solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95.
74

400

sample. The peaks in the Seebeck coefficient are large, especially in comparison to the literature
values reported for YbAl2; however, they are also very narrow. Combining the sharp peaks in the
Seebeck coefficient with the decreased electrical resistivity, the power factor rises to enormous
values within a very small temperature range, as shown in Figure 5-8. The power factor reaches a
value near 160µW/cmK2, which is even larger than that reported for the YbAl3 solid solutions.
This tremendous power factor at cryogenic temperatures (40K) is significant and the largest
value achieved in this temperature range by a bulk thermoelectric material to date. The total
thermal conductivities for all samples in the solid solution except, for x=0 and 1.0, are shown in
Figure 5-9. The x = 1.0 sample in not shown here because its thermal conductivity is greater than
0.6 W/cmK, making the scale difficult to observe the differences in the lower Sc concentration
samples, and there is no literature data on the thermal conductivity of the x = 0 sample. As
expected, the total thermal conductivity of the sample increases as the Sc concentration increases
and becomes more metallic due to the increasing electronic contribution. Unfortunately, further
reduction of the lattice thermal conductivity is likely to have little to no affect on the total
thermal conductivity as the electronic contribution makes up a significant portion of the total
thermal conductivity and cannot be reduced without detrimental affects to the power factor
The dimensionless figure of merit for these compositions follows a similar trend to the
power factor, with a very narrow peak at low temperatures. The zT, shown in Figure 5-10, clearly
demonstrates that the Sc content can be used to tailor the temperature range that optimizes the
zT. There is little difference between the thermal conductivities in the cryogenic temperature
range, thus the zT peaks occur at the same temperatures as the power factor and Seebeck
coefficient. Unfortunately, each of the compositions results in very sharp peaks in the zT, which
is not ideal for applications. Despite the dramatic increase in the Seebeck coefficient and power
75

factor at low temperatures, the zT value is still quite low due to the electronic contribution to the
thermal conductivity. Despite the low zT value, these compositions provide a clear demonstration
that alloying is a promising method for increasing Seebeck coefficient, power factor, and zT in
intermediate valence systems, and for controlling the temperature at which these parameters are
optimized. In order to better understand the underlying properties causing these low temperature
peaks, magnetic measurements were done to investigate the average Yb valence as a function of
temperature and composition.
0.09
x=0.60
x=0.80
x=0.90
x=1.0

0.08
0.07

zT

0.06

x=0.70
x=0.85
x=0.95

0.05

0.04
0.03
0.02

0.01
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 5-10: The zT from 2K(80K) to 300K(350K) for the Yb1-xScxAl2 solid solution where
x = 0.60, 0.70, 0.80, 0.85, 0.90, 0.95, and 1.0.

5.2.4 Magnetic Properties
The magnetic moment was measured for each of the samples as a function of temperature. This
data was collected as described in Section 3.3.2. Sales and Wholleben [40] previously proposed a
model to fit the magnetic susceptibility of Yb-based intermediate valence compounds. This
76

model is known as the Interconfigurational Fluctuation (ICF) model, and is used here to extract
the average Yb valence from the data. This model takes into account a temperature-dependent
ratio of the Yb2+ and Yb3+ state, which have two independent energies (Eex = E3+-E2+), angular
momentums (J3 or J2), and effective magnetic moments as determined by Hund’s rules (μ3 or μ2).
By fitting two variables, the excitation energy (Eex) and the spin fluctuation temperature (Tsf), to
the temperature dependent magnetic susceptibility data, the average temperature-dependent Yb
valence (ν(T)) can be determined. For the ICF model, the susceptibility may be expressed as:

which can be simplified for the Yb case. When Yb is in the 4f14 (divalent) state, J2=0 and μ2=0,
thus the model is reduced to a more simplified version using the values J3=7/2 and μ3=4.54 μB
for Yb in the 4f13 (trivalent) state as follows:

A correction was done to remove the background signal that was observed from the oxide
impurity phase. The phase was identified as Yb3Al5O12, which has a Yb valence of 3+. The large
moment associated with the trivalent Yb ion creates a Curie tail at low temperature in the data.
The Curie tail was fit between 2K and 10K according to M=C/T, where M is the magnetic
moment, C is the Curie constant, and T is the temperature. The Curie constant is obtained from
the slope of the M vs. 1/T curve. This impurity signal was then subtracted from the total moment
and used to estimate the mass of the impurity phase. From the corrected moment and the
corrected sample mass, the magnetic susceptibility was found per Yb atom, a quantity that could

77

Magnetic Susceptibility (emu/mol(Yb))

0.035

x=0.60
x=0.70
x=0.80
x=0.85
x=0.90
x=0.95

0.030
0.025
0.020

0.015
0.010
0.005
0.000
0

50

100

150
200
Temperature (K)

250

300

Figure 5-11: The corrected magnetic susceptibility data (points) and the corresponding ICF
model fits (lines) from 2K to 300K for the Yb1-xScxAl2 solid solution where x = 0.60, 0.70,
0.80, 0.85, 0.90, and 0.95

Temperature of Peak (K)

300
250

T (max Seebeck)

200

T (max susceptibility)

150

100
50
0
0.5

0.6

0.7
0.8
Sc concentration (x)

0.9

1

Figure 5-12: Comparison of the temperatures at which the magnitude of the Seebeck
coefficient and the magnetic susceptibility are maximized as a function of Sc concentration.
be fit to the ICF model. The resulting curves were fit to the model by adjusting the two model
parameters. The corrected data and their fits are shown in Figure 5-11.
78

The low temperature peaks in the magnetic susceptibility data are characteristic of
intermediate valence systems. It can be seen that the temperatures at which these peaks occur
correspond very closely with the temperatures at which the Seebeck coefficient peak for each
composition; see Figure 5-12. This correlation provides strong evidence that the large Seebeck
coefficient is driven and controlled by the valence fluctuations of the Yb atom. These interesting
transport and magnetic data call for a direct investigation of the relationship between the average
Yb valence and the Seebeck coefficient. Using the ICF model fits, the average Yb valence can be
plotted as a function of temperature, as shown in Figure 5-13. The compression of the unit cell
resulting from increased Sc content causes the Yb atom to have a less available space and thus
the Yb atom spends more time in the trivalent state, which has fewer electrons (4f13) and a
smaller radius. The average Yb valence changes toward the trivalent state rather rapidly, and at
3.0
2.9

Average Yb Valence

2.8
2.7
2.6
2.5
x=0.60
x=0.70
x=0.80
x=0.85
x=0.90
x=0.95

2.4
2.3

2.2
2.1
2.0
0

50

100

150
Temperature (K)

200

250

300

Figure 5-13: The average Yb valence as a function of temperature obtained from the ICF
model for the Yb1-xScxAl2 solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95.

79

high Sc content the average Yb valence reaches 2.8 by 100K, as can be seen in Figure 5-13. The
Seebeck coefficient can be plotted against this average Yb valence, which results in a revealing
trend, shown in Figure 5-14.

Seebeck Coefficient (μV/K)

0.0
-10.0
-20.0
-30.0
x=0.60
x=0.70
x=0.80
x=0.85
x=0.90
x=0.95

-40.0
-50.0
-60.0
2

2.2

2.4
2.6
Average Yb Valence

2.8

3

Figure 5-14: The Seebeck coefficient as a function of the average Yb valence for the Yb1xScxAl2 solid solution where x = 0.60, 0.70, 0.80, 0.85, 0.90, and 0.95.
This figure demonstrates that the magnitude of the Seebeck coefficient is maximized,
independent of composition, near an average Yb valence of 2.5. This result gives experimental
evidence of the effects of Yb valence on the Seebeck coefficient as well as provides a specific
goal for tailoring the Seebeck coefficient via tuning the average Yb valence in intermediate
valence compositions. The narrow peak in the Seebeck coefficient can be understood by
examining Figure 5-13 with respect to the relationship shown in Figure 5-14. If it holds true, as
evidenced here, that the Seebeck coefficient peaks at the temperature where the average Yb
valence curve passes near 2.5, it can be assumed that the Seebeck coefficient will only stay large

80

for temperatures where the average Yb valence remains near 2.5. The slopes of the average Yb
valence curves are quite steep for large Sc content, and therefore the average Yb valence of each
composition is only near 2.5 for small (or narrow) temperature ranges, consistent with the narrow
peaks that are observed in the Seebeck coefficient. The temperature at which the Seebeck
coefficient peaks also shifts towards lower temperatures as Sc is added to the system. This shift
again can be explained by the average Yb valence curve, which increases more rapidly as the Sc
content is increased, thus reducing the temperature at which an average Yb valence of 2.5 is
achieved.
The relationship between the average Yb valence and Seebeck coefficient, as well as the
shape of the average Yb valence vs. temperature curve demonstrated here, provide new insight
into the desired characteristics of Yb-based intermediate valence compounds. One might
speculate that the average Yb valence curve should have a magnitude near 2.5 and be flat across
the entire temperature range to achieve a large Seebeck coefficient across the entire temperature
range, meeting the needs of a good thermoelectric material for low-temperature applications.

5.3 Summary
The Yb1-xScxAl2 solid solution does not show significant promise as a material for low
temperature thermoelectric applications; however, this system does reveal important trends
between the Seebeck coefficient and magnetic properties. YbAl2 is an intermediate valence
system with nearly metallic thermoelectric properties. Methods for altering the average Yb
valence have been widely investigated in the literature, yet no work had been done to investigate
the relationship between these valence changes and the Seebeck coefficient, which is known to
be affected by the fluctuating Yb valence state. In this work, a new synthesis method for YbAl2
was developed that suppressed the formation of the YbAl3 phase resulting in a fairly high-purity
81

sample; further, bulk polycrystalline samples were large enough to carry out thermoelectric
property measurements.
The thermoelectric properties of YbAl2 were significantly enhanced by forming a solid
solution with ScAl2. Large peaks in the Seebeck coefficient and intrinsically low electrical
resistivity resulted in tremendous power factors, reaching a value of 160μW/cmK2 near 50K.
This large power factor is unfortunately paired with a large thermal conductivity resulting in a
modest zT of 0.08 at 50K. The large peaks in the Seebeck coefficient are also quite narrow,
further rendering these materials useless for applications.
The magnetic measurements on these samples showed trends similar to those shown in
the literature, with a large peak in the magnetic susceptibility at low temperatures, typical of
intermediate valence systems. These curves were fit by the ICF model and thus the average Yb
valence was extracted for each sample as a function of temperature. This average valence was
then directly compared to the temperature dependent Seebeck coefficient, yielding some of the
most interesting and insightful results of this solid solution. The absolute value of the Seebeck
coefficient, which is known to be strongly related to the intermediate valence state of Yb, is
maximized in each sample where the average Yb valence is near 2.5, regardless of composition.
A deeper understanding of the fundamental physics of this relationship would be very desirable,
but is beyond the scope of this work.
YbAl2-ScAl2 is an ideal system for demonstrating the effects of chemical pressure on the
average Yb valence and thus showing the relationship between the Yb valence and the Seebeck
coefficient. While there is little potential for thermoelectric applications using this particular
compound due to the dominance of the electronic contribution to the thermal conductivity, the

82

new understanding regarding the relationship between average Yb valence and Seebeck
coefficient has provided a guide for enhancing other intermediate valence Yb-based compounds.
Chemical pressure may be used to systematically study the thermoelectric properties of
intermediate valence systems as a function of average Yb valence, and an average Yb valence
near 2.5 is ideal for this and possibly other similar systems. This experimental evidence has
offered new insight into low temperature intermediate valence compounds and provided a
specific goal for tailoring other Yb-based compounds with potential as low-temperature
thermoelectrics.

83

Chapter 6 YbCu2Si2-Based Solid Solutions
6.1 Introduction
YbCu2Si2 has been reported several times in the literature [54,55,92-95] as an intermediate
valence system with a relatively large Seebeck coefficient at low temperatures. This compound is
promising because there are several solid solutions that can be investigated to maximize the
thermoelectric figure of merit. Solid solutions are used to alter the lattice parameter and thus the
average Yb valence as well as reduce the lattice thermal conductivity via disruption of phonon
propagation. The transport properties, particularly the Seebeck coefficient, can be compared to
the average Yb valence. As shown in the last chapter, the work on the YbAl2-ScAl2 solid
solution demonstrated a maximum in the magnitude of the Seebeck coefficient when the average
Yb valence is near 2.5. YbCu2Si2-based alloys provide another system on which the versatility of
this behavior can be tested. The first solid solution investigated here is YbCu2Si2-YbCu2Ge2. The
unit cell volume should be increased as Ge is substituted for Si, as it has a larger atomic radius,
and thus the average Yb valence should be reduced as it is allowed more space to assume the
divalent state. The second solid solution reported here is YbCu2Si2-LaCu2Si2. Alloying with La
may also create lattice expansion, as the La atom has a larger atomic radius than the Yb atom.
YbCu2Si2 has an average Yb valence of approximately 2.75 at room temperature, and therefore
expanding the unit cell should cause the average Yb valence to shift toward a more divalent
state, ideally approaching the target average Yb valence of 2.5.
YbCu2Si2, and its isostructural compounds used for alloying, have a tetragonal unit cell
of the ThCr2Si2-type crystal structure [44,47,48,92,96-99], shown in Figure 6-1. This structure
has a smaller lattice thermal conductivity than the YbAl3 and YbAl2 compounds previously

84

investigated in this report because of a more complex phonon spectrum. This difference suggests
that alloying may not have the same impact on reducing the lattice thermal conductivity as it has
in the previous samples because this value is already greatly reduced due to the intrinsic
complexity of the atomic structure. However, creating solid solutions is anticipated to have a
strong influence on the Seebeck coefficient, which can be tailored to achieve maximum values
for the zT and may still have some effect on the thermal conductivity. The results of the Yb-La
solid solution study are currently pending publication in the Journal of Electronic Materials
[100], while the YbCu2Si2-YbCu2Ge2 results are drafted for future publication.

Figure 6-1: Unit cell diagram of YbCu2Si2.

85

6.2 YbCu2Si2-xGex
YbCu2Si2 is isostructural with YbCu2Ge2, and the solid solution between these compounds has
been investigated here. This solid solution is interesting because there is no change in the
concentration of Yb atoms in the compound, but the size of the unit cell is altered instead by
substitution on the Si site. The Ge atom is larger than the Si atom, therefore an expansion in the
lattice parameter is expected which can be used to tailor the average Yb valence. Ge is also
heavier than the Si atom and thus a reduction in the lattice thermal conductivity is expected.
By expanding the unit cell, the average Yb valence should shift toward a more divalent
state from the nominal 2.75 of the pure YbCu2Si2 sample. If results from the YbAl2-ScAl2 solid
solution hold true for other Yb-based intermediate valence compounds, the Seebeck coefficient
should be maximized when the average Yb valence is approximately 2.5. The YbCu2Si2YbCu2Ge2 solid solution should provide an interesting series on which to evaluate the Seebeck
coefficient and its relationship with the average Yb valence.

6.2.1 Synthesis
Samples were formed in the solid solution YbCu2Si2-xGex where x=0, 0.25, 0.50, 0.75, 1.0, 1.25,
1.50, 1.75, and 2.0. Similar to the methods previously utilized, the two base compounds were
initially synthesized and then mixed together before densification to form the desired
compositions. Both the YbCu2Si2 and YbCu2Ge2 base compounds were formed by direct arc
melting of the stoichiometric quantities. The two base compounds were then powdered and
mixed together by ball-milling stoichiometric quantities for 20 minutes and followed by sieving
using a 60mesh screen. The powder was then densified using PECS processing at 950oC under
60MPa of pressure for 30 minutes. This time and temperature profile resulted in samples with
densities >95% of the theoretical value.
86

6.2.2 Characterization
Each sample was characterized using XRD, as shown in Figure 6-2. The systematic shift in the
large 2θ range toward lower angles as Ge content is increased is characteristic of an expansion in
the unit cell. The change in the unit cell volume is plotted in Figure 6-3, showing the linear
increase in size as the Ge concentration in increased. Also shown Figure 6-3 is the unit cell
volume for each of the end compounds as reported in the literature [92]. The increase in the unit
cell volume is the primary driving force for a change in the average Yb valence, and thus the

(312)

(215)

(116)
(220)

(123)
(024)

(105)
(211)

(110)

(200)

(103)
(004)
(112)
x=2.0

(101)

Relative Intensity (Arbitrary units)

motivation for studying this solid solution.

x=1.75

x=1.50
x=1.25
x=1.0
x=0.75
x=0.50
x=0.25
x=0.0

10

20

30

40

50
2ϴ

60

Figure 6-2: XRD patterns for the YbCu2Si2-xGex solid solution.

87

70

80

90

Unit Cell Volume (Å3)

175
170
165
160
155
150
145
0

0.25

0.5

0.75
1
1.25
Ge concentration (x)

1.5

1.75

2

Figure 6-3: The unit cell volume increases linearly as the Ge concentration is increased in the
YbCu2Si2-xGex solid solution. Solid circles represent literature data [92].
Figure 6-4 shows the micrographs for the end compounds YbCu2Si2 and YbCu2Ge2
collected via SEM. These figures show grain sizes on the order of several microns, consistent
with other Yb-based compounds processed in the same manner as discussed in the above
chapters. Small quantities of oxygen were observed via EDS even though secondary phases are
not visible in the XRD patterns. The amount of oxygen detected here is small and thus can be
ignored for the thermoelectric transport properties; however, it is taken into account for the
magnetic measurements as discussed in the following sections. The oxide impurity is assumed to
be Yb2O3, an impurity commonly found in Yb-based compounds, and qualitatively confirmed by
EDS measurements.

88

A

B

Figure 6-4: SEM images of fractured surfaces for A) YbCu2Si2 and B) YbCu2Ge2.

89

6.2.3 Transport Properties
The electrical resistivity of this solid solution decreases systematically as Ge is added to the
system as shown in Figure 6-5. The YbCu2Si2 sample has a fairly large electrical resistivity
compared to some intermediate valence compounds, which is detrimental to the power factor of
this base compound. However, the pure YbCu2Ge2 sample has a resistivity of approximately
20μΩcm at room temperature. This reduced electrical resistivity is good for the power factor, but
creates a large electronic contribution to the thermal conductivity.
80

x=0
x=0.25
x=0.50
x=0.75
x=1.0
x=1.25
x=1.50
x=1.75
x=2.0

Electrical Resistivity (Ωcm)

70
60
50
40
30
20
10

0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-5: The electrical resistivity for the YbCu2Si2-xGex solid solution where x = 0, 0.25,
0.50, 0.75, 1.0, 1.25, 1.50, 1.75 and 2.0.

Broad maxima in the Seebeck coefficient for each composition are shown in Figure 6-6.
As Ge is added to the system, the magnitude of the Seebeck coefficient is reduced and the
temperature at which the maxima occur shifts toward a higher temperature. This shift in
magnitude toward higher temperatures is opposite of that seen in the YbAl2-ScAl2 solid solution,
90

0
x=0
x=0.25
x=0.50
x=0.75
x=1.0
x=1.25
x=1.50
x=1.75
x=2.0

Seebeck Coefficient (μV/K)

-10
-20
-30
-40
-50
-60
-70
-80

-90
0

50

100

150
200
250
Temperature (K)

300

350

400

Power Factor (μW/cmK2)

Figure 6-6: The Seebeck coefficient for the YbCu2Si2-xGex solid solution where x = 0, 0.25,
0.50, 0.75, 1.0, 1.25, 1.50, 1.75 and 2.0.
150
140
130
120
110
100
90
80
70
60
50
40
30
20
10
0

x=0
x=0.25
x=0.50
x=0.75
x=1.0
x=1.25
x=1.50
x=1.75
x=2.0

0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-7: The power factor for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50,
0.75, 1.0, 1.25, 1.50, 1.75 and 2.0.
91

where the peak of the Seebeck coefficient was shifted towards lower temperatures as the size of
the unit cell decreased. This result supports the hypothesis that increasing the unit cell volume
has the opposite effect on the temperature dependence of the Seebeck coefficient as compressing
the unit cell. Combining the reduction in the electrical conductivity with the diminished Seebeck
coefficients, the power factor, shown in Figure 6-7, is enhanced across the entire temperature
range for x ≤ 1.0 and at higher temperatures for higher concentrations of Ge.
Substituting Ge for Si makes that samples more metallic, however the solid solution also
increases the phonon scattering rate. The lattice thermal conductivity is reduced as the electronic
component increases, resulting in a decrease in the total thermal conductivity for x < 1.0 across
the entire temperature range, as can be seen in Figure 6-8. The total thermal conductivity is
significantly increased in samples with large Ge concentration as the electronic component
overwhelms any reduction in the lattice thermal conductivity originating from point defects.

Thermal Conductivity (W/cmK)

0.45
x=0
x=0.25
x=0.50
x=0.75
x=1.0
x=1.25
x=1.50
x=1.75
x=2.0

0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-8: The thermal conductivity for the YbCu2Si2-xGex solid solution where x = 0, 0.25,
0.50, 0.75, 1.0, 1.25, 1.50, 1.75 and 2.0.
92

The zT for these materials is enhanced for x≤1.0 resulting from the increased power
factor combined with the decrease in the thermal conductivity. Higher concentrations of Ge also
result in increased zT when compared to the YbCu2Si2 sample, particularly at higher
temperatures; however, the values are not as large as those achieved by the x = 1.0 sample. By
varying the Ge content, a maximum zT of 0.16 can be achieved across the temperature range
150K to 350K, as seen in Figure 6-9.
0.20
x=0
x=0.25
x=0.50
x=0.75
x=1.0
x=1.25
x=1.50
x=1.75
x=2.0

0.18
0.16
0.14

zT

0.12
0.10
0.08
0.06
0.04
0.02
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-9: The zT for the YbCu2Si2-xGex solid solution where x = 0, 0.25, 0.50, 0.75, 1.0,
1.25, 1.50, 1.75 and 2.0.

6.2.4 Magnetic Properties
The expansion of the unit cell is suspected to cause a change in the average Yb valence, which
can be determined by fitting the ICF model to the magnetic susceptibility. The magnetic moment

93

Magnetic Susceptibility (emu/mol(Yb))

0.010

x=0.0
x=0.75
x=1.50

0.008

x=0.25
x=1.0
x=1.75

x=0.50
x=1.25
x=2.0

0.006

0.004

0.002

0.000
0

50

100

150
200
Temperature (K)

250

300

Average Yb Valence

Figure 6-10: The corrected magnetic susceptibility for the YbCu2Si2-xGex solid solution
where 0≤x≤2. Solid lines are fits to the ICF model.
3.0
2.9
2.8
2.7
2.6
2.5
2.4
2.3
2.2
2.1
2.0

x=0
x=0.75
x=1.50

0

x=0.25
x=1.0
x=1.75

50

100

x=0.50
x=1.25
x=2.0

150
Temperature (K)

200

250

300

Figure 6-11: The average Yb valence as a function of temperature for the YbCu2Si2-xGex
solid solution where 0≤x≤2.
was collected for each sample and then corrected for a small quantity of Yb2O3, as determined by
fitting the low temperature Curie tail. After removing this background, the resulting magnetic
94

susceptibility per mole of Yb is shown in Figure 6-10, along with the ICF model fits. The peak in
the magnetic susceptibility shifts towards higher temperatures as the Ge concentration is
increased. Figure 6-11 shows a decrease in the average Yb valence across the entire temperature
range as the Ge content is increased, as hypothesized for an expansion in the unit cell. This plot
also shows a relatively flat trend near an average Yb valence value of 2.5, particularly for the
x=1.0 sample, explaining the large broad peaks observed in the Seebeck coefficient.

0
x=0

Seebeck Coefficient (μV/K)

-10

x=0.25

-20

x=0.50

-30

x=0.75

-40

x=1.0
x=1.25

-50

x=1.50

-60

x=1.75

-70

x=2.0

-80
-90
2

2.2

2.4
2.6
Temperature (K)

2.8

3

Figure 6-12: The Seebeck coefficient vs. average Yb valence for the YbCu2Si2-xGex
solid solution where 0≤x≤2.

The fits from the ICF model were used to compare the average Yb valence and the
Seebeck coefficient data, as shown in Figure 6-12. This figure supports the results found in the
YbAl2-ScAl2 solid solutions, where the Seebeck coefficient is maximized in each composition
when the average Yb valence is near 2.5. The shift in the temperature at which the Seebeck
coefficient peaks is also related to the temperature at which the magnetic susceptibility peaks, as
95

shown in Figure 6-13. This result further supports the strong relationship between the Seebeck
coefficient and the magnetic state of the Yb atoms in the compound.
400
T(max Seebeck)
T(max susceptibility)

Temperature (K)

350
300
250
200
150
100
50
0
0

0.5

1
Ge content (x)

1.5

2

Figure 6-13: Comparison of the temperatures at which the magnitude of the Seebeck
coefficient and the magnetic susceptibility are maximized as a function of Ge concentration.

6.3 Yb1-xLaxCu2Si2
6.3.1 Synthesis
The solid solution Yb1-xLaxCu2Si2 where x=0.05, 0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90,
0.95, and 1.0 were synthesized by first forming the two ternary compounds, YbCu2Si2 and
LaCu2Si2, independently. These base compounds were synthesized by arc melting stoichiometric
quantities of the elements. The LaCu2Si2 compound was homogenized via heat treatment after
the arc melting process, and then ball milled. This high temperature homogenization process was
done in for 48 hours at 750oC. Stoichiometric quantities of the base ternary compounds were
balled milled together for 20 minutes to achieve the correct composition. The powder was then
sieved and densified using PECS processing. PECS was done under 60MPa of pressure at a
temperature of 950oC for 30 minutes.
96

6.3.2 Characterization
X-ray diffraction was done following each step of the synthesis process. The final XRD patterns
for select compositions in the solid solution are reported in Figure 6-14. XRD patterns were also
collected on the other compositions, which show only the characteristic peaks with no visible
secondary phase, but were not included here in order to simplify the plot. A shift in the peaks is
clearly visible at higher angles, indicating a change in the size of the unit cell. The unit cell
volume for each composition is shown in Figure 6-15, consistent with the literature data for the
base compounds YbCu2Si2 [92] and LaCu2Si2 [101] represented as filled circles.
The microstructure of the LaCu2Si2 is shown in Figure 6-16. The grains shown here are

10

(312)

(215)

(220)
(116)

(123)
(024)

(211)
(105)

(200)

(110)
(103)
(112)
(004)

(101)
F
E
D
C
B

20

30

40

50
2ϴ

60

70

(312)

(215)

(116)
(220)

(123)
(024)

(105)
(211)

(200)

(110)
(103)
(004)
(112)

A
(011)

Relative Intensity (Arbitrary units)

on the order of several microns in diameter and appear to be similar to the YbCu2Si2

80

90

Figure 6-14: XRD patterns collected at room temperature for the Yb1-xLaxCu2Si2 solid
solution where A) x = 0, B) x = 0.20, C) x = 0.40, D) x = 0.60, E) x = 0.80, and F) x = 1.0.
97

Unit Cell Volume (Å3)

180
175
170
165
160
155
150
0

0.2

0.4
0.6
La concentration (x)

0.8

1

Figure 6-15: The unit cell volume increases linearly as the La concentration is increased in
the Yb1-xLaxCu2Si2 solid solution. Solid circles represent literature data.

Figure 6-16: SEM images of fractured surfaces for LaCu2Si2.

microstructure. Small impurity phases are noticeable in the SEM images although the XRD
patterns do not show any extraneous peaks. Again, these impurities are oxide phases and are
assumed to be less than 5% based on the limitations of XRD.
98

6.3.3 Transport Properties
The electrical resistivities of these samples is not affected much by low concentrations of La,
however moderate quantities (x>0.20) causes a significant drop in the electrical resistivity, as can
be seen in Figure 6-17. The pure LaCu2Si2 sample has a different trend in the resistivity than the
samples with even dilute Yb concentrations. It is suspected that the magnetic moment of the Yb
atom allows the electrical resistivity to become very small, an effect that is not observed in
LaCu2Si2.
80
x=0
x=0.05
x=0.10
x=0.15
x=0.20
x=0.40
x=0.60
x=0.80
x=0.85
x=0.90
x=0.95
x=1.0

Electrical Resistivity (μΩcm)

70
60
50
40
30
20

10
0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-17: The electrical resistivity for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05,
0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.

The Seebeck coefficient data shown in Figure 6-18 reveals that it is strongly dependent
on the La concentration, and as La is added to the composition, the magnitude of the Seebeck
coefficient decreasey. However, as the La concentration increases, it is also apparent that the

99

0
x=0
x=0.05
x=0.10
x=0.15
x=0.20
x=0.40
x=0.60
x=0.80
x=0.85
x=0.90
x=0.95
x=1.0

Seebeck Coefficient (μV/K)

-10
-20
-30

-40
-50
-60
-70
-80
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-18: The Seebeck coefficient for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05,
0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.
temperature at which the Seebeck coefficient peaks is shifted towards lower temperatures. The
Seebeck coefficient must approach the origin at zero Kelvin, therefore the peaks for high
concentrations of La are assumed to occur below the measured temperature. While the
magnitude of the Seebeck coefficient decreases with increased La, this reduction is relatively
small in samples with La concentrations up to x=0.60. Therefore it is possible to optimize the
temperature and magnitude of the Seebeck coefficient by tuning the La concentration.
Because there is little change in the electrical resistivity and Seebeck coefficient at low
La concentrations, the power factor is not dramatically changed, as can be seen in Figure 6-19. In
the previous chapters, a reduction to the electrical resistivity has been beneficial to the power
factor; however, in this case while the electrical resistivity is reduced with increasing La
concentrations, the Seebeck coefficient is also reduced. Thus a reduction in the power factor is
100

Power Factor (μW/cmK2)

140
130
120
110
100
90
80
70
60
50
40
30
20
10
0

x=0
x=0.05
x=0.10
x=0.15
x=0.20
x=0.40
x=0.60
x=0.80
x=0.85
x=0.90
x=0.95
x=1.0
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-19: The power factor for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10,
0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.
observed for large concentrations of La subsitituion.
Creating solid solutions has also been a successful method for reducing the lattice
thermal conductivity of samples. Unfortunately, once again the thermal transport in these
samples is dominated by the electronic component rather than lattice component, and is therefore
much more strongly related to the electrical conductivity, which was shown above to be very
large. Using the Weidemann-Franz law, it is possible to calculate the electronic contribution,
which is nearly 90% of the total thermal conductivity. As La is added to the system the lattice
thermal conductivity is reduced; however, electronically the samples become more metallic,
thus, the thermal conductivity gets much larger, particularly in the high La concentration
samples, shown in Figure 6-20. Small concentrations of La do slightly reduce the lattice thermal
conductivity while the electronic resistivity is not significantly changed, therefore a small
101

0.50
x=0
x=0.05
x=0.10
x=0.15
x=0.20
x=0.40
x=0.60
x=0.80
x=0.85
x=0.90
x=0.95
x=1.0

Thermal Conductivity (W/cmK)

0.45
0.40
0.35
0.30

0.25
0.20
0.15
0.10
0.05
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-20: The thermal conductivity for the Yb1-xLaxCu2Si2 solid solution where x = 0,
0.05, 0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.
0.16
x=0
x=0.05
x=0.10
x=0.15
x=0.20
x=0.40
x=0.60
x=0.80
x=0.85
x=0.90
x=0.95
x=1.0

0.14

0.12

zT

0.10
0.08
0.06
0.04
0.02
0.00
0

50

100

150
200
250
Temperature (K)

300

350

400

Figure 6-21: The zT for the Yb1-xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10, 0.15, 0.20,
0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.
102

reduction in the total thermal conductivity can be seen for samples with La concentration x≤0.40,
particularly in the low temperature region.
Combining these transport properties, the dimensionless figure of merit is shown in
Figure 6-21. The peak of the zT reflects the shift towards lower temperatures in the Seebeck
coefficient, as expected. More significantly, this plot demonstrates that small concentrations of
La give the best results, reaching zT values of nearly 0.15 at 150K. While this value is enhanced
from the pure YbCu2Si2 sample and does show a fairly broad peak, it is not as large as that
achieved by the Ge substitution and the large electronic contribution to the thermal conductivity
is once again the limiting factor in the potential for this material as a thermoelectric.

6.3.4 Magnetic Properties
In order to better understand the peak in the magnitude of the Seebeck coefficient, measurements
were performed to collect the magnetic moment. This data was again corrected for an oxide
impurity phase which contributes a paramagnetic moment to the background. The corrected
moment was used to find the magnetic susceptibility of each of the samples and then fit to the
ICF model to extract the average Yb valence. Figure 6-22 shows that as the La concentration is
increased the magnetic signal is significantly decreased because the Yb concentration is small.
Therefore, these compositions were very difficult to fit to the model, which requires strong
signals from the Yb atoms. Samples with small concentrations of La were investigated; however
the corrected data used to obtain the average Yb valence was only reliable for samples x≤0.20.
As can be seen in Figure 6-23, increasing temperature causes an increase in the average Yb
valence from 2.2 to 2.8, contradictory to the literature, which suggests that the Yb valence has a
much flatter trend, growing from ~2.5 to 2.8 [40]. This discrepancy can be assumed to originate
from the correction factor and any error in the fits to the model which may
103

Magnetic Susceptibility (emu/mol(Yb))

0.010

x=0.0
x=0.20
x=0.85

0.008

x=0.05
x=0.40
x=0.90

x=0.10
x=0.60
x=0.95

x=0.15
x=0.80

0.006

0.004

0.002

0.000
0

50

100

150
200
Temperature (K)

250

300

Figure 6-22: The corrected magnetic susceptibility for the Yb1-xLaxCu2Si2 solid solution
where x = 0, 0.05, 0.10, 0.15, 0.20, 0.40, 0.60, 0.80, 0.85, 0.90, 0.95, and 1.0.

Average Yb Valence

3.0
x=0.05
x=0.10
x=0.15
x=0.20

2.8
2.6
2.4
2.2
2.0
0

50

100

150
200
Temperature (K)

250

300

Figure 6-23: The average Yb valence as calculated from the ICF model from 2K to 300K for
Yb1-xLaxCu2Si2 where x=0, 0.05, 0.10, 0.15, and 0.20.
arise from this correction. One may readily notice that the fits are quite poor when compared to
those reported above on the YbAl2-ScAl2 and YbCu2Si2-YbCu2Ge2 solid solutions. However, it
is still possible to assume that the trend in the average Yb valence from smaller to larger values
104

is accurate as there is little change in the trends for all of the x≤0.20 samples and the Seebeck is
also maintained in each of these compositions. Assuming these values for the average Yb
valence are accurate, Figure 6-24 shows again that the magnitude of the Seebeck coefficient is
maximized when the average Yb valence is near 2.5. The Seebeck coefficient is associated with
average Yb valences near or above 2.5 because an average Yb valence of less than 2.5 is only
observed below 80K, a temperature range in which the transport properties are not measured.
The trend observed here may be considered further evidence in support of the previous findings
regarding the relationship between the average Yb valence and the Seebeck coefficient.

-40

Seebeck Coefficient (μV/K)

x=0
-45

x=0.05

-50

x=0.10
x=0.15

-55

x=0.20

-60
-65
-70

-75
-80
2

2.2

2.4
2.6
Average Yb Valence

2.8

Figure 6-24: The Seebeck coefficient as a function of the average Yb valence for the Yb1xLaxCu2Si2 solid solution where x = 0, 0.05, 0.10, 0.15, and 0.20.

105

3

6.4 Summary
YbCu2Si2 was alloyed with YbCu2Ge2 and LaCu2Si2 in an attempt to further investigate the
affects of chemical substitutions on the average Yb valence and the corresponding changes to the
Seebeck coefficient. These substitutions caused an expansion in the unit cell, providing more
space for the Yb atoms. This increased lattice parameter was expected to decrease the average
Yb valence, as Yb2+ has a larger radius than Yb3+. The magnetic susceptibility of these solid
solutions was collected and fit to the ICF model to extract the average Yb valence. XRD of the
YbCu2(Si,Ge)2 solid solution showed that unit cell was expanded and the magnetic fits
demonstrated a decrease in the average Yb concentration as Ge was added to the system. In the
La solid solution, the lattice parameter was also increased with increasing La content as seen in
XRD; however, the dispersion of Yb atoms causes a reduced signal in the magnetic moment and
the average Yb valence could not be accurately determined.
The substitutions of Ge and La have opposing affects on the shift in the peaks of the
Seebeck coefficient. The YbCu2(Si,Ge)2 solid solution shows a shift towards higher temperatures
in the peak Seebeck coefficient as Ge content is increased. This shift can be explained by looking
at the temperature at which the average Yb valence is near 2.5. Since the average Yb valence
decreases with increasing Ge content, a higher temperature is required to achieve a value near
2.5. The Seebeck coefficient of the La substituted samples shows a shift towards lower
temperatures, more similarly to that seen in the YbAl2-ScAl2. It is suspected that this shift
towards lower temperatures is a result of the dilution of Yb atoms, as proposed by Zlatić and
Monnier [34]. However, the lack of reliable results on the average Yb valence prevents a
complete explanation for this trend in Seebeck coefficient toward lower temperatures.

106

The Ge substitution in the YbCu2Si2 solid solution not only shifts the peak in the Seebeck
coefficient towards higher temperatures, but it also alters the electrical and thermal
conductivities. The power factor reaches a maximum value of 140μW/cmK2 at 80K for the x=1.0
sample (YbCu2SiGe). The large thermal conductivity of this sample does not make it the optimal
composition across the entire temperature range. Instead, lower Ge concentrations (x=0.50) have
the best zT at the lowest temperature ranges, while x=1.0 is ideal above 200K. A zT value of
approximately 0.15 can be achieved at 100K and above using varying compositions where x≤1.0.
Substitution of La for Yb in the YbCu2Si2 compound alters each of the thermoelectric
properties in a way that enhances the zT for low La concentrations, reaching values of 0.15 at
150K. Smaller concentrations of La result in the lowered thermal conductivities while the power
factors are slightly enhanced due to the reduced electrical resistivity. However, none of these La
substituted samples exceed the zT values of those seen in the Ge solid solution.
While the zT values are enhanced from that of the original YbCu2Si2 sample, they are still
not sufficiently large for most applications. However, the broad peak in the zT that can be
achieved by varying concentrations of Ge or La is potentially valuable for niche Peltier cooling
applications. Unfortunately there is little work that could be done further enhance this zT because
most of the thermal conductivity comes from the electronic contribution due to the already
complex unit cell. Therefore, grain size engineering and further isoelectronic substitution would
be unlikely to significantly reduce the lattice thermal conductivity. Thus the only remaining
method for increasing the zT is further enhancement of the power factor.

107

Chapter 7 Conclusions
The goal of this work was to synthesize new materials using common processing
techniques through which the thermoelectric properties could be controlled by chemical
substitution. Yb-based compounds were investigated because of their unique properties in the
cryogenic temperature range, an area that has not been heavily investigated. By studying the
unique intermediate valence behaviors that many Yb-based materials exhibit it is possible to
control the thermoelectric properties via a newfound understanding of the underlying physical
concepts and their dependence on temperature and stoichiometry.
Synthesis of these Yb-based materials required an innovative approach due to the
volatility of pure Yb metal. As discussed in Chapters 4 and 5, new methods which required nonstoichiometric quantities of the constituent elements resulted in high-purity samples. The
trialuminide compounds were formed using an aluminum flux arc melting method which
prevented the dialuminide phase from crystallizing as the sample cooled from the molten phase.
This excess aluminum was then removed via chemical etching leaving porous chunks of pure
YbAl3 which were then powder processed and sintered. The dialuminide samples discussed in
Chapter 5 was formed by arc melting with excess Yb to compensate for the non-homogenous
losses that occur during the arc melting process. A high temperature heat treatment allowed any
remaining Yb metal to vaporize and only the desired YbAl2 phase remained. These samples were
also powder processed and densified. These two new synthesis techniques allowed high purity
polycrystalline samples (particularly in comparison to the polycrystalline samples previously
reported in the literature) to be investigated.

108

Thermoelectric properties were measured on three different series of solid solutions.
YbAl3 was alloyed with ScAl3, ErAl3 and LuAl3 resulting in increased power factor and zT
values for each solid solution where 25% of the Yb atoms were substituted. The solid solution
between YbAl2 and ScAl2 showed unique peaks in the Seebeck coefficient near 50K, particularly
for samples with large quantities of Sc. These peaks greatly enhanced the zT value in the
cryogenic temperature range. The peak in the Seebeck coefficient also shifted toward lower
temperatures as the Sc content was increased. These dialuminide samples were further
investigated via magnetic property measurements in order to evaluate the average Yb valence in
the range where the magnitude of the Seebeck coefficient was largest. This investigation
revealed a strong relationship between the peak temperatures of the Seebeck coefficient and the
magnetic susceptibility as well as between the maximum Seebeck coefficient and an average Yb
valence of approximately 2.5. The final series that was investigated in this work was the
YbCu2Si2 series in which solid solutions with YbCu2Ge2 and LaCu2Si2 were formed. An increase
in the zT was observed in the YbCu2Si2-LaCu2Si2 solid solution at low temperatures, while the
zT of the YbCu2Si2-YbCu2Ge2 samples were enhanced across the entire temperature range. The
peak in the Seebeck coefficients of these two solid solutions behaved in different manners, where
Ge caused the peak to shift towards higher temperatures, the La substitution caused a shift
toward lower temperatures. The average Yb valence of these compounds was determined using
magnetic data. Reminiscent of the dialuminide series, the temperature at which the Seebeck
coefficient is maximized matches closely with the temperature at which magnetic susceptibility
peaks. Also, the magnitude of the Seebeck coefficient is maximized when the average Yb
valence is near 2.5.

109

These results provide evidence that the Seebeck coefficient is maximized when the f-shell
electrons are in a state of maximum fluctuation, a state which may be achieved by tailoring the
unit cell volume through chemical substitution. This chemical substitution causes an expansion
or contraction of the unit cell and as the unit cell volume is altered, the Yb atom adapts to the
new environment and adjusts the number of electrons in the f-shell. If the volume is reduced the
Yb atom loses an electron to better fit into the available space, thus becoming more trivalent,
while if the volume is increased the Yb atom becomes more divalent as it retains a full f-shell.
This trend is clearly shown in Figure 7-1 where the average Yb valence is plotted vs. the unit cell
volume for the YbAl2-ScAl2 solid solution as well as both the of the YbCu2Si2 solid solutions.

Average Yb Valence

430

435

440

Unit Cell Volume (A3)
445
450
455
460

3.0
2.9
2.8
2.7
2.6
2.5
2.4
2.3
2.2
2.1
2.0

465

470

475

YbCu2(Si,Ge)2
(Yb,La)Cu2Si2
(Yb,Sc)Al2

145

150

155
160
165
3
Unit Cell Volume (A )

170

175

Figure 7-1: The average Yb valence vs. the unit cell volume for three solid solutions shows
that the Yb valence state is dependent on the available space in the unit cell.
This figure shows a non-linear relationship in which the average Yb valence transforms from
nearly trivalent to the divalent state as the unit cell volume is increased. This trend is apparent in
the two solid solutions that were fully studied and limited results from the YbCu2Si2-LaCu2Si2

110

solid solution also support this relationship. This figure provides conclusive evidence that the
average Yb valence of these systems can be controlled by altering the unit cell volume via
chemical substitutions, a method which may be utilized in future work on intermediate valence
compounds.

Seebeck Coefficient (μV/K)

0
-10
-20
-30
-40
-50
-60
(Yb,Sc)Al2

-70

YbCu2(Si,Ge)2

-80

(Yb,La)Cu2Si2

-90
2

2.2

2.4
2.6
Average Yb Valence

2.8

3

Figure 7-2: The Seebeck coefficient vs. average Yb valence for each of the Yb-based
intermediate valence compounds investigated in this work.

It has been well established that the Seebeck coefficient is strongly correlated with the
average Yb valence. This correlation is explicitly shown in Figure 7-2, where the absolute value
of the Seebeck coefficient is maximized for each solid solution when the average Yb valence is
near 2.5. This plot summarizes several of the figures shown in the previous chapters, because it
is important to note that while the magnitude of the Seebeck coefficient may be dependent on
composition (indicated by different symbols), it always peaks when the fluctuations in the Yb
valence are maximized. Figure 7-2 provides strong motivation for further investigation of Yb-

111

based compounds and a means to tailor the thermoelectric properties to optimize the Seebeck
coefficient via solid solution alloying. Understanding the underlying physics of this behavior is
an outstanding problem and would be an interesting question to investigate in the future.

112

REFERENCES

113

REFERENCES

[1] T. J. Seebeck and A. v (Arthur) Oettingen, Magnetische Polarisation der Metalle und Erze
durch Temperatur-Differenz, 1822-1823. Hrsg. von A.J. v. Oettingen. Leipzig W.
Engelmann, 1895.
[2] S. Kasap, “Thermoelectric effects in metals: Thermocouples,” Can. Dep. Electr. Eng. Univ.
Sask., 2001.
[3] D. T. Morelli, “Thermoelectric Devices,” Encyclopedia of Applied Physics, vol. 21.
WILEY-VCH Verlag GmbH & Co KGaA, 1997.
[4] T. M. Tritt, “Thermoelectric Materials: Principles, Structure, Properties, and Applications,”
Encyclopedia of Materials: Science and Technology. Elsevier, 2002.
[5] G. Mahan, B. Sales, and J. Sharp, “Thermoelectric Materials: New Approaches to an Old
Problem,” Phys. Today, vol. 50, no. 3, pp. 42–47, 1997.
[6] J. Zheng, “Recent advances on thermoelectric materials,” Front. Phys. China, vol. 3, no. 3,
pp. 269–279, Jul. 2008.
[7] D. M. Rowe, Ed., Thermoelectrics Handbook: Macro to Nano, 1st ed. CRC Press, 2005.
[8] “Thermoelectric Overview,” Noah Precision, LLC, 2011. [Online]. Available:
http://www.lauda-noah.com/thermoelectric-overview.html. [Accessed: 14-Oct-2014].
[9] P. Dyches, “Radioisotope Thermoelectric Generator (RTG),” Solar System Exploration.
[Online]. Available: http://solarsystem.nasa.gov/rps/rtg.cfm. [Accessed: 25-Sep-2014].
[10] K. Tate, “Nuclear Generators Power NASA Deep Space Probes (Infographic),” Space.com.
[Online]. Available: http://www.space.com/13702-nuclear-generators-rtg-power-nasaplanetary-probes-infographic.html. [Accessed: 25-Sep-2014].
[11] J. W. Sharp, “Thermoelectric and Energy Conversion Devices,” in Encyclopedia of
Condensed Matter Physics, F. Bassani, G. L. Liedl, and P. Wyder, Eds. Oxford: Elsevier,
2005, pp. 173–180.
[12] S. B. Riffat and X. Ma, “Thermoelectrics: a review of present and potential applications,”
Appl. Therm. Eng., vol. 23, no. 8, pp. 913–935, Jun. 2003.
[13] A. J. Simon, “U.S Energy Flow Chart.” Lawerence Livermore National Lab and
Department of Energy.
[14] G. Chen, M. S. Dresselhaus, G. Dresselhaus, J.-P. Fleurial, and T. Caillat, “Recent
developments in thermoelectric materials,” Int. Mater. Rev., vol. 48, no. 1, pp. 45–66, Feb.
2003.
[15] F. J. DiSalvo, “Thermoelectric Cooling and Power Generation,” Science, vol. 285, no.
5428, pp. 703–706, Jul. 1999.
[16] L. Laird, “Could TEG Improve Your Car’s Efficiency?,” Energy.gov, 16-Aug-2010.
[Online]. Available: http://energy.gov/articles/could-teg-improve-your-cars-efficiency.
[Accessed: 25-Sep-2014].
[17] “Alphabet Energy Introduces The World’s Most Powerful Thermoelectric Generator,”
Alphabet Energy. [Online]. Available: http://www.alphabetenergy.com/pressreleases/alphabet-energy-introduces-worlds-powerful-thermoelectric-generator/. [Accessed:
09-Oct-2014].
[18] R. E. Simons and R. C. Chu, “Application of thermoelectric cooling to electronic
equipment: a review and analysis,” in Sixteenth Annual IEEE Semiconductor Thermal
Measurement and Management Symposium, 2000, 2000, pp. 1–9.
114

[19] L. Yershova, V. Volodin, T. Gromov, D. Kondratiev, G. Gromov, S. Lamartinie, J. P.
Bibring, and A. Soufflot, “Thermoelectric Cooling for Low Temperature Space
Environment,” in Proceedings of the 7th European Workshop on Thermoelectrics,
Pamplona, 2002.
[20] T. M. Davis, J. Reilly, and F. L. B. J. Tomlinson, “Air Force Research Laboratory
Cryocooler Technology Development,” in Cryocoolers 10, R. G. R. Jr, Ed. Springer US,
2002, pp. 21–32.
[21] G. J. Snyder and E. S. Toberer, “Complex thermoelectric materials,” Nat. Mater., vol. 7, no.
2, pp. 105–114, Feb. 2008.
[22] H. J. Goldsmid and R. W. Douglas, “The use of semiconductors in thermoelectric
refrigeration,” Br. J. Appl. Phys., vol. 5, no. 11, p. 386, Nov. 1954.
[23] G. D. Mahan and J. O. Sofo, “The best thermoelectric,” Proc. Natl. Acad. Sci., vol. 93, no.
15, pp. 7436–7439, Jul. 1996.
[24] A. F. Ioffe, “On thermal conduction in semiconductors,” Il Nuovo Cimento Ser. 10, vol. 3,
no. 4, pp. 702–715, Apr. 1956.
[25] A. . V. Ioffe and A. F. Ioffe, “Thermal conductivity of semiconductor solid solutions,” Sov.
Phys. Solid State, vol. 2, pp. 719–728, 1960.
[26] R. Berman, Thermal Conduction in Solids. Oxford: Oxford University Press, 1976.
[27] P. G. Klemens, “Thermal Resistance due to Point Defects at High Temperatures,” Phys.
Rev., vol. 119, no. 2, pp. 507–509, Jul. 1960.
[28] C. Suryanarayana, “Mechanical alloying and milling,” Prog. Mater. Sci., vol. 46, no. 1–2,
pp. 1–184, Jan. 2001.
[29] Z. A. Munir, U. Anselmi-Tamburini, and M. Ohyanagi, “The effect of electric field and
pressure on the synthesis and consolidation of materials: A review of the spark plasma
sintering method,” J. Mater. Sci., vol. 41, no. 3, pp. 763–777, Feb. 2006.
[30] M. Omori, “Sintering, consolidation, reaction and crystal growth by the spark plasma
system (SPS),” Mater. Sci. Eng. A, vol. 287, no. 2, pp. 183–188, Aug. 2000.
[31] H. J. Goldsmid and A. W. Penn, “Boundary scattering of phonons in solid solutions,” Phys.
Lett. A, vol. 27, no. 8, pp. 523–524, Sep. 1968.
[32] S. Kasap, “Chapter 5: Semiconductors,” in Principles of Electronic Materials and Devices,
3 edition., Boston: McGraw-Hill Science/Engineering/Math, 2005.
[33] N. Oeschler, S. Hartmann, U. Köhler, M. Deppe, P. Sun, and F. Steglich, “Thermoelectric
Power of Correlated Compounds,” in Properties and Applications of Thermoelectric
Materials, V. Zlatić and A. C. Hewson, Eds. Springer Netherlands, 2009, pp. 81–90.
[34] V. Zlatic and R. Monnier, “Theory of the thermoelectricity of intermetallic compounds with
Ce or Yb ions,” Phys. Rev. B, vol. 71, no. 16, Apr. 2005.
[35] M. B. Maple, L. E. DeLong, and B. C. Sales, “Chapter 11 Kondo effect: Alloys and
compounds,” in Handbook on the Physics and Chemistry of Rare Earths, vol. Volume 1, J.
and L. E. Karl A. Gschneidner, Ed. Elsevier, 1978, pp. 797–846.
[36] M. D. Daybell and W. A. Steyert, “Localized Magnetic Impurity States In Metals: Some
Experimental Relationships,” Rev. Mod. Phys., vol. 40, no. 2, pp. 380–389, Apr. 1968.
[37] B. Batlogg, E. Kaldis, and P. Wachter, “Mixed valence compounds with hopping
conductivity,” J. Magn. Magn. Mater., vol. 3, no. 1–2, pp. 96–98, Mar. 1976.
[38] E. Kaldis, A. Schlegel, P. Wachter, and C. Zurcher, “Electronic structure of divalent-,
trivalent- and mixed valence rare earth compounds,” J. Magn. Magn. Mater., vol. 3, no. 1–
2, pp. 1–3, Mar. 1976.
115

[39] J. M. Lawrence, P. S. Riseborough, and R. D. Parks, “Valence fluctuation phenomena,”
Rep. Prog. Phys., vol. 44, no. 1, p. 1, Jan. 1981.
[40] B. C. Sales and D. K. Wohlleben, “Susceptibility of Interconfiguration-Fluctuation
Compounds,” Phys. Rev. Lett., vol. 35, no. 18, pp. 1240–1244, Nov. 1975.
[41] L. L. Hirst, “Theory of magnetic impurities in metals,” Phys. Kondens. Mater., vol. 11, no.
3, pp. 255–278, 1970.
[42] L. L. Hirst, “The configuration‐based approach to magnetic impurities in metals,” in AIP
Conference Proceedings, 1975, vol. 24, pp. 11–15.
[43] E. Colombier, D. Braithwaite, G. Lapertot, B. Salce, and G. Knebel, “High-pressure
transport and microcalorimetry studies on high quality YbCu_{2}Si_{2} single crystals,”
Phys. Rev. B, vol. 79, no. 24, p. 245113, Jun. 2009.
[44] K. Alami-Yadri, H. Wilhelm, and D. Jaccard, “Pressure-induced magnetically ordered
Kondo lattice state in YbCu2Si2,” Eur. Phys. J. B - Condens. Matter Complex Syst., vol. 6,
no. 1, pp. 5–11, Nov. 1998.
[45] H. Nowatari, Y. Saiga, Y. Kato, K. Iwata, S. Katano, T. Fujiwara, Y. Uwatoko, and M.
Kosaka, “Effect of Pressure on the Electrical Resistivity of YbAl2,” J. Phys. Soc. Jpn., vol.
76SA, no. Supplement A, pp. 80–81, 2007.
[46] C. Dallera, E. Annese, J.-P. Rueff, A. Palenzona, G. Vankó, L. Braicovich, A. Shukla, and
M. Grioni, “Valence changes in YbAl2 under pressure: a resonant inelastic X-ray emission
study,” J. Electron Spectrosc. Relat. Phenom., vol. 137–140, pp. 651–655, Jul. 2004.
[47] A. Grytsiv, D. Kaczorowski, V. H. Tran, A. Leithe-Jasper, and P. Rogl, “Crystal structure
and physical properties of Yb-based intermetallics Yb(Cu,Ag)2(Si, Ge)2,
Yb(Cu1−xZnx)2Si2 (x = 0.65, 0.77) and Yb(Ag0.18Si0.82)2,” J. Alloys Compd., vol. 504,
pp. 1–6, 2010.
[48] M. M. A.-E. H. Winkelmann, “Direct observation of a magnetically ordered state in YbCu_
{2} Si_ {2} under high pressure,” Phys Rev B, vol. 60, no. 5, 1999.
[49] M. Sancrotti, I. Abbati, E. Puppin, Z. X. Shen, I. Lindau, R. Eggenhöffner, A. Iandelli, and
G. L. Olcese, “Yb mean valence in chemically compressed Yb1-xScxAl2 pseudobinary
alloys: An X-ray absorption spectroscopy investigation,” Solid State Commun., vol. 74, no.
10, pp. 1131–1135, Jun. 1990.
[50] L. Braicovich, B. De Michelis, A. Fasana, E. Puppin, M. Sancrotti, E. Vescovo, R.
Eggenhöffner, A. Iandelli, G. L. Olcese, and A. Palenzona, “Chemical substitution effect on
Yb valence in the surface region of Yb1−xMxAl2 (= Sc, Ca): A photoemission
investigation,” Solid State Commun., vol. 79, no. 1, pp. 75–79, Jul. 1991.
[51] I. Abbati, B. DeMichelis, A. Fasana, M. Sancrotti, A. Iandelli, G. L. Olcese, and R.
Eggenhöffner, “Chemical pressure effects and valence fluctuations in Yb0.6Ca0.4Al2 and
YbAl2 alloys: A photoemission investigation,” Solid State Commun., vol. 65, no. 7, pp.
743–747, Feb. 1988.
[52] R. Eggenhoffner, M. Sancrotti, I. Abbati, E. Puppin, Z. X. Shen, I. Lindau, A. Iandelli, and
G. L. Olcese, “Yb valence in Yb1-xCaxAl2 pseudobinary alloys via Yb LIII X-ray
absorption spectroscopy,” Solid State Commun., vol. 74, no. 9, pp. 1009–1012, Jun. 1990.
[53] R. Eggenhoffner and G. L. Olcese, “Chemical pressure behaviour of intermediate valent
compounds in the resonance fermi liquid model,” J. Common Met., vol. 154, no. 1, pp. 89–
98, Oct. 1989.

116

[54] K. Alami-Yadri, D. Jaccard, and D. Andreica, “Thermopower of Yb Heavy Fermion
Compounds at High Pressure,” J. Low Temp. Phys., vol. 114, no. 1–2, pp. 135–149, Jan.
1999.
[55] B. C. Sales and R. Viswanathan, “Demagnetization due to interconfiguration fluctuations in
the RE-Cu2Si2 compounds,” J. Low Temp. Phys., vol. 23, no. 3–4, pp. 449–467, May 1976.
[56] B. C. Sales, “A model for the thermodynamic properties of metallic rare earth systems with
an unstable valence,” J. Low Temp. Phys., vol. 28, no. 1–2, pp. 107–127, Jul. 1977.
[57] A. Iandelli and G. L. Olcese, “Structural and magnetic investigations on some RAl2-MAl2
laves phase systems,” J. Common Met., vol. 111, no. 1–2, pp. 145–156, Sep. 1985.
[58] B. Coqblin, B. Chevalier, and V. Zlatić, “Thermoelectric Power and Thermal Transport of
Anomalous Rare-Earth Kondo Compounds,” in Properties and Applications of
Thermoelectric Materials, V. Zlatić and A. C. Hewson, Eds. Dordrecht: Springer
Netherlands, 2009, pp. 91–117.
[59] J. Flouquet and H. Harima, “Heavy fermion material: Ce versus Yb case,” arXiv e-print
0910.3110, Oct. 2009.
[60] V. Zlatić, B. Horvatić, I. Milat, B. Coqblin, G. Czycholl, and C. Grenzebach,
“Thermoelectric power of cerium and ytterbium intermetallics,” Phys. Rev. B, vol. 68, no.
10, pp1-11, Sep. 2003.
[61] P. C. Canfield and Z. Fisk, “Growth of single crystals from metallic fluxes,” Philos. Mag.
Part B, vol. 65, no. 6, pp. 1117–1123, Jun. 1992.
[62] P. C. Canfield, “Solution Growth of Intermetallic Single Crystals: A Beginner’s Guide,”
Prop. Appl. Complex Intermet., pp. 93–111, 2010.
[63] N. A. Lange and J. A. Dean, Lange’s handbook of chemistry, 10th ed. [New York:
McGraw-Hill, 1967.
[64] D. M. Rowe, G. Min, and L. Kuznestsov, “Electrical resistivity and Seebeck coefficient of
hot-pressed YbAl3 over the temperature range 150-700K,” Philos. Mag. Lett., vol. 77, no.
2, pp. 105–108, Feb. 1998.
[65] D. M. Rowe, V. L. Kuznetsov, L. A. Kuznetsova, and G. Min, “Electrical and thermal
transport properties of intermediate-valence YbAl3,” J. Phys. Appl. Phys., vol. 35, no. 17,
pp. 2183–2186, Sep. 2002.
[66] G. Min and D. M. Rowe, “A novel thermoelectric converter employing Fermi gas/liquid
interfaces,” J. Phys. Appl. Phys., vol. 32, no. 6, p. L26, Mar. 1999.
[67] T. He, T. G. Calvarese, J.-Z. Chen, H. D. Rosenfeld, R. J. Small, J. J. Krajewski, and M. A.
Subramanian, “Origin of low thermal conductivity in alpha;-Mn: enhancing the ZT of
YbAl3 and CoSb3 through Mn addition,” in 24th International Conference on
Thermoelectrics, 2005. ICT 2005, 2005, pp. 434 – 439.
[68] S. Ohara, G. F. Chen, and I. Sakamoto, “Effect of pressure on transport properties of
mixed-valence compound YbAl3,” J. Alloys Compd., vol. 323–324, pp. 632–635, Jul. 2001.
[69] C. L. Foiles, “Thermopower of GdIn3 and YbAl3,” J. Appl. Phys., vol. 52, no. 3, pp. 2217–
2219, Mar. 1981.
[70] J. Zhou, Z. Sun, X. Cheng, and Y. Zhang, “Phase stability and electronic structures of
YbAl3−xMx (M=Mg, Cu, Zn, In and Sn) studied by first-principles calculations,”
Intermetallics, vol. 17, no. 12, pp. 995–999, Dec. 2009.
[71] J. Zhou, B. Sa, and Z. Sun, “First-principles investigations on phase stability and electronic
structures of Yb1−xMxAl3 (M = Ho, Er and Tm) alloys,” Intermetallics, vol. 18, no. 12,
pp. 2394–2398, Dec. 2010.
117

[72] R. E. Majewski, A. S. Edelstein, A. T. Aldred, and A. E. Dwight, “Susceptibility
measurements on the valence fluctuation system (Yb,Sc)Al3,” J. Appl. Phys., vol. 49, no. 3,
pp. 2096–2097, Mar. 1978.
[73] S. Lee, J. Park, P. Canfield, and D. Lynch, “Optical properties and electronic structures of
single crystalline RAl3 (R=Sc, Yb, and Lu),” Phys. Rev. B, vol. 67, no. 7, Feb. 2003.
[74] E. D. Bauer, C. H. Booth, J. M. Lawrence, M. F. Hundley, J. L. Sarrao, J. D. Thompson, P.
S. Riseborough, and T. Ebihara, “Anderson lattice behavior in Yb1-xLuxAl3,” Phys. Rev.
B, vol. 69, no. 12, p. 125102, Mar. 2004.
[75] G. J. Lehr and D. T. Morelli, “Synthesis, crystal structure, and thermoelectric properties of
the YbAl3-ScAl3 solid solution,” Intermetallics, vol. 32, no. 0, pp. 225–229, Jan. 2013.
[76] G. J. Lehr and D. T. Morelli, “Thermoelectric Properties of Yb1−x (Er,Lu) x Al3 Solid
Solutions,” J. Electron. Mater., vol. 42, no. 7, pp. 1697–1701, Jul. 2013.
[77] J. L. Moriarty, J. E. Humphreys, R. O. Gordon, and N. C. Baenziger, “X-ray examination of
some rare-earth-containing binary alloy systems,” Acta Crystallogr., vol. 21, no. 5, pp.
840–841, Nov. 1966.
[78] F. G. Meng, L. G. Zhang, H. S. Liu, L. B. Liu, and Z. P. Jin, “Thermodynamic optimization
of the Al–Yb binary system,” J. Alloys Compd., vol. 452, no. 2, pp. 279–282, Mar. 2008.
[79] K. A. Gschneidner and F. W. Calderwood, “The Al-Yb (Aluminum-Ytterbium) system,”
Bull. Alloy Phase Diagr., vol. 10, no. 1, pp. 47–49, Feb. 1989.
[80] Y. Harada and D. . Dunand, “Microstructure of Al3Sc with ternary transition-metal
additions,” Mater. Sci. Eng. A, vol. 329–331, no. 0, pp. 686–695, Jun. 2002.
[81] T. P. Braun and F. J. DiSalvo, “Transport properties of intermetallic thulium compounds,”
J. Alloys Compd., vol. 305, no. 1–2, pp. 43–48, Jun. 2000.
[82] A. Hiess, J. X. Boucherle, F. Givord, J. Schweizer, E. Lelièvre-Berna, F. Tasset, B. Gillon,
and P. C. Canfield, “Magnetism in intermediate-valence YbAl3 : a polarized neutron
diffraction study,” J. Phys. Condens. Matter, vol. 12, no. 6, pp. 829–840, Feb. 2000.
[83] D. M. Rowe, G. Min, and V. L. Kuznetsov, “Thermoelectric properties of hot-pressed
YbAl3 compound over temperature range 150-800 K,” in XVI International Conference on
Thermoelectrics, 1997. Proceedings ICT ’97, 1997, pp. 528 –531.
[84] J. Murray, “The Al-Sc (aluminum-scandium) system,” J. Phase Equilibria, vol. 19, no. 4,
pp. 380–384, 1998.
[85] K. A. Gschneidner and F. W. Calderwood, “The Al−Er (Aluminum-Erbium) system,” Bull.
Alloy Phase Diagr., vol. 9, no. 6, pp. 676–678, Dec. 1988.
[86] H. Okamoto, “Al-Lu (aluminum-lutetium),” J. Phase Equilibria, vol. 14, no. 3, pp. 394–
395, Jun. 1993.
[87] S. J. Lee, S. Y. Hong, I. R. Fisher, P. C. Canfield, B. N. Harmon, and D. W. Lynch,
“Optical properties and electronic structure of single crystals of LuAl_{2} and YbAl_{2},”
Phys. Rev. B, vol. 61, no. 15, pp. 10076–10083, Apr. 2000.
[88] T. Görlach, C. Pfleiderer, K. Grube, and H. v. Löhneysen, “Low-temperature properties of
YbAl_{2},” Phys. Rev. B, vol. 71, no. 3, p. 033101, Jan. 2005.
[89] G. J. Lehr, D. T. Morelli, H. Jin, and J. P. Heremans, “Enhanced thermoelectric power
factor in Yb1-xScxAl2 alloys using chemical pressure tuning of the Yb valence,” J. Appl.
Phys., vol. 114, 2013.
[90] H. J. van Daal, P. B. van Aken, and K. H. J. Buschow, “The seebeck coefficient of YbAl2
and YbAl3,” Phys. Lett. A, vol. 49, no. 3, pp. 246–248, Sep. 1974.

118

[91] W. Franz, F. Steglich, W. Zell, D. Wohlleben, and F. Pobell, “Intermediate Valence on
Dilute Europium Ions,” Phys. Rev. Lett., vol. 45, no. 1, pp. 64–67, Jul. 1980.
[92] N. D. Dung, T. D. Matsuda, Y. Haga, S. Ikeda, E. Yamamoto, T. Ishikura, T. Endo, S.
Tatsuoka, Y. Aoki, H. Sato, T. Takeuchi, R. Settai, H. Harima, and Y. Ōnuki, “de Haas-van
Alphen Effect and Fermi Surface Properties in High-Quality Single Crystals YbCu2Si2 and
YbCu2Ge2,” J. Phys. Soc. Jpn., vol. 78, no. 8, Aug. 2009.
[93] B. C. Kuzhel, I. D. Shcherba, and I. I. Kravchenko, “The contribution of valence unstable
ytterbium states into kinetic properties of YbNi2−xGe2+x and YbCu2−xSi2+x,” J. Alloys
Compd., vol. 425, no. 1–2, pp. 54–58, Nov. 2006.
[94] D. Andreica, K. Alami-Yadri, D. Jaccard, A. Amato, and A. Schenck, “Transport properties
and μSR spectroscopy of Yb(NixCu1−x)2Si2,” Phys. B Condens. Matter, vol. 259–261, pp.
144–145, Jan. 1999.
[95] K. Alami-Yadri, H. Wilhelm, and D. Jaccard, “Transport properties of Yb-compounds at
high pressure,” Phys. B Condens. Matter, vol. 259–261, pp. 157–158, Jan. 1999.
[96] E. V. Sampathkumaran, L. C. Gupta, and R. Vijayaraghavan, “NMR studies in ICF systems
of RECu2Si2 compounds,” J. Phys. C Solid State Phys., vol. 12, no. 20, p. 4323, Oct. 1979.
[97] C. N. R. Rao, D. D. Sarma, P. R. Sarode, E. V. Sampathkumaran, L. C. Gupta, and R.
Vijayaraghavan, “Valence fluctuation in some Yb intermetallics by X-ray photoemission
and X-ray absorption,” Chem. Phys. Lett., vol. 76, no. 3, pp. 413–415, Dec. 1980.
[98] T. D. Matsuda, N. D. Dung, Y. Haga, S. Ikeda, E. Yamamoto, T. Ishikura, T. Endo, T.
Takeuchi, R. Settai, and Y. Ōnuki, “Fermi surface properties of YbCu2Si2,” Phys. Status
Solidi B, vol. 247, no. 3, pp. 757–759, 2010.
[99] T. K. Hatwar, R. M. Nayak, B. D. Padalia, M. N. Ghatikar, E. V. Sampathkumaran, L. C.
Gupta, and R. Vijayaraghavan, “X-ray absorption spectroscopic study of mixed valence
systems EuCu2Si2, YbCu2Si2 and Sm4Bi3,” Solid State Commun., vol. 34, no. 8, pp. 617–
620, May 1980.
[100] G. J. Lehr and D. T. Morelli, “YbCu2Si2-LaCu2Si2 solid solutions with enhanced
thermoelectric power factor,” J. Electron. Mater., pending.
[101] I. L. Spain, F. Steglich, U. Rauchschwalbe, and H. D. Hochheimer, “Energy dispersive in
situ high pressure study of stoichiometric and non-stoichiometric rare earth
coppersilicides,” Phys. BC, vol. 139–140, pp. 449–452, May 1986.

119