THE DESIGN, DEVELOPMENT, FABRICATION AND
TESTING OF A 100 WATT SKUTTERUDITE
THERMOELECTRIC GENERATOR
By
Matthew Lyle

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Mechanical Engineering
2011

ABSTRACT
THE DESIGN, DEVELOPMENT, FABRICATION AND TESTING OF A 100 WATT
SKUTTERUDITE THERMOELECTRIC GENERATOR
By
Matthew Lyle
Thermoelectric technology is a method of renewable, alternative energy that utilizes the Seebeck
effect to convert some of the thermal energy in a temperature gradient to electricity. The optimal
temperature range for skutterudite thermoelectric devices is around 650°C, making them ideal
for high temperature applications. At this temperature range, the skutterudite thermoelectrics
have a device-level conversion efficiency of about 9% [1]. As these devices are still in the
development stage, testing that simulates real-world conditions is necessary to assess the
feasibility of implementing skutterudite thermoelectric technology with current processes.

A standardized procedure to test the skutterudite thermoelectric devices has been
established to reduce variability in device fabrication and generator assembly. This procedure
includes a measurement and tracking system to aid in establishing relationships between
component properties and thermoelectric performance. In addition, a technology has been
developed to electrically bypass any failed devices to preserve overall power generation.

Results indicate that additional efforts are needed to address the high level of thermal
stresses the devices experience during operation. Several methods to reduce thermal stresses and
investigate potential stressors are proposed. In addition, the successful performance of the
electrical bypass technology suggests that it is indeed a viable method of bypassing individual
devices for experimental tests. Additional testing and improvements can be made as necessary to
implement this technology in the envisioned 1 kW skutterudite thermoelectric generator.

ACKNOWLEDGEMENTS
The discoveries and accomplishments detailed in this Thesis are the result of hard work,
continual support and expert guidance from a number of individuals. I would like to first and
foremost thank my advisor, Dr. Harold Schock who believed in me from our first meeting and
has shown nothing but support and encouragement during my time at MSU. In addition, his solid
methodology and perceptive insights shared during the course of this project have been
instrumental in addressing critical problems and determining next steps.

I would also like to thank my colleagues at MSU who have worked me on this project, shared
their expertise and included me as a friend. First of all, Trevor Ruckle, who spearheaded the
module design and created the generator design and the testing process, has taken me under his
wing for the past 2 years and has tirelessly answered my many questions, saved me from
countless mistakes and has redefined the meaning of “Practical Wisdom.” Dr. Jeff Sakamoto,
who invented the individual couple testing procedure and Ryan Maloney have both made
significant contributions in effectively integrating the aerogel insulation technology with the
thermoelectric devices as well as providing an in-depth analysis and explanation of the material
interactions from a material science perspective. Edward Timm, who created the flexible hot
shoe design, and Dr. Long Zhang have both devoted substantial efforts to material synthesis and
testing and have also identified and implemented improvements to the fabrication and testing
processes, as well as to the material itself. Kevin Moran and Timothy Stewart developed the
Couple Bypass Technology and have made substantial, ongoing improvements to its design and
control. Dr. George Zhu has also played a prominent role in drafting electronic designs and
highlighting electrical considerations. Finally, Jeff Higel and Gary Keeney have performed

iii

exceptional machining work for this project and have gone out of their way to make me feel
welcome here at MSU.

I would like to thank my parents, Stephen and Katrina Lyle for their support and guidance over
the course of my life. They have always been willing to lend a hand, offer advice, or simply be a
listening ear. My sister, Rebecca and my brother, Daniel have also encouraged me, believed in
me and have both taught me invaluable life lessons that I will never forget.

The incredible support and encouragement that my wife, Katie, has shown me during this
journey has entrenched in me a commitment for excellence and a grateful appreciation for the
best life partner I could ever have hoped for.

Finally, and most importantly, I would like to thank my Lord and Savior, Jesus Christ, without
whom, I am truly less than nothing.

iv

TABLE OF CONTENTS
LIST OF TABLES ................................................................................................................... vii
LIST OF FIGURES ................................................................................................................ viii
LIST OF SYMBOLS AND ACRONYMS ..................................................................................x
1. Introduction.............................................................................................................................1
1.1. TE Overview ....................................................................................................................1
1.1.1. History.......................................................................................................................2
1.1.2. How TEDs work ........................................................................................................3
1.2. Project Scope....................................................................................................................4
1.3. Outline of Thesis ..............................................................................................................5
2. Literature Review ....................................................................................................................7
2.1. Previous Work ..................................................................................................................7
2.1.1. Ideal Location ............................................................................................................8
2.1.2. Cost Analysis .............................................................................................................9
2.1.3. Finite Element Analysis ........................................................................................... 12
2.2. Previous Tests ................................................................................................................ 13
2.3. Research Questions ........................................................................................................ 15
2.4. Summary ........................................................................................................................ 15
3. Fabrication Process ............................................................................................................... 16
3.1. Materials Synthesis ......................................................................................................... 16
3.1.1. Element Mixing ....................................................................................................... 17
3.1.2. Synthesis and Annealing .......................................................................................... 18
3.1.3. Powder Processing ................................................................................................... 18
3.2. Leg Fabrication .............................................................................................................. 19
3.2.1. Hot Press ................................................................................................................. 20
3.2.2. Dicing ...................................................................................................................... 22
3.3. Couple Assembly ........................................................................................................... 24
3.3.1. Hot Shoe Design ...................................................................................................... 25
3.3.2. Bonding ................................................................................................................... 26
3.3.3. Individual Couple Testing ........................................................................................ 29
3.4. Module Assembly ........................................................................................................... 30
3.4.1. Module Board .......................................................................................................... 30
3.4.2. Insulation ................................................................................................................. 32
3.4.3. Cold Plate ................................................................................................................ 34
3.5. Measurements ................................................................................................................ 35
3.5.1. Tracking System ...................................................................................................... 36
3.5.2. Electrical Resistivity and Short Circuit..................................................................... 37
3.6. Summary ........................................................................................................................ 39

v

4. Electronic CBT ..................................................................................................................... 40
4.1. Overview ........................................................................................................................ 40
4.2. Design ............................................................................................................................ 42
4.2.1. Hardware ................................................................................................................. 44
4.2.2. Software .................................................................................................................. 49
4.2.3. Data Acquisition and Supervisory Control ............................................................... 49
4.3. Fabrication ..................................................................................................................... 51
4.4. Summary ........................................................................................................................ 53
5. Testing .................................................................................................................................. 54
5.1. Generator Design ............................................................................................................ 54
5.1.1. Module Housing ...................................................................................................... 55
5.1.2. Heat Source ............................................................................................................. 57
5.1.3. Cooling System ....................................................................................................... 59
5.1.4. Instrumentation & Data Acquisition System ............................................................ 59
5.2. Testing Process ............................................................................................................... 62
5.3. Post-processing Data ...................................................................................................... 63
5.4. Summary ........................................................................................................................ 65
6. Results & Discussion ............................................................................................................ 66
6.1. Power Results ................................................................................................................. 67
6.2. CBT Results ................................................................................................................... 67
6.3. Insulation Comparison Test ............................................................................................ 72
6.3.1. Test Setup ................................................................................................................ 72
6.3.2. Results ..................................................................................................................... 74
6.4. Individual Couple Testing Results .................................................................................. 76
6.5. Summary ........................................................................................................................ 77
7. Conclusions ........................................................................................................................... 78
7.1. Research Achievements .................................................................................................. 78
7.2. Conclusions of Research ................................................................................................. 79
7.3. Recommendations for Future Work ................................................................................ 80
7.4. Summary ........................................................................................................................ 84
REFERENCES ......................................................................................................................... 86

vi

LIST OF TABLES
Table 2-1 WAVE Simulation Results ..........................................................................................9
Table 2-2 Cost Analysis Results ................................................................................................ 10
Table 6-1 Maximum ΔT and Hot Side Temperatures for Insulation Comparison Test ................ 76

vii

LIST OF FIGURES
Figure 1-1 Thermoelectric Couple ...............................................................................................3
Figure 2-1 Payback Period Estimation ....................................................................................... 11
Figure 2-2 FEA Simulation of TE Couple ................................................................................. 13
Figure 3-1 Skutterudite Metallization Layers ............................................................................. 20
Figure 3-2 Skutterudite Hot-Pressed Ingots ............................................................................... 22
Figure 3-3 Ingot Grinding Diagram ........................................................................................... 23
Figure 3-4 Current TE Couple Design ....................................................................................... 24
Figure 3-5 Unpopulated Module Board ..................................................................................... 31
Figure 3-6 Module Board with EASs, Ribbon Cable Connectors, Tracking Number,
Mounting Bolts and Clamping Bolts ......................................................................................... 31
Figure 3-7 Aerogel Form........................................................................................................... 32
Figure 3-8 Module Board after adding Aerogel Mold, Mica Film and Soldered Couples
(Side View) ............................................................................................................................... 33
Figure 3-9 Module Board after adding Aerogel Mold, Mica Film and Soldered Couples
(Bottom View) .......................................................................................................................... 33
Figure 3-10 Module with Cold Plate .......................................................................................... 35
Figure 3-11 TE Couple Tracking Diagram ................................................................................ 37
Figure 3-12 Electrical Resistivity Measurement Jigs.................................................................. 38
Figure 4-1 EAS Placement Schematic [19] ................................................................................ 43
Figure 4-2 Microprocessor Board .............................................................................................. 44
Figure 4-3 EAS Activation Signal and Low Side Voltages ........................................................ 45
Figure 4-4 Additive Principle of Low Side Voltages.................................................................. 46
Figure 4-5 Benefits of Floating Power Supply ........................................................................... 48

viii

Figure 4-6 CBT Schematic [20]................................................................................................. 51
Figure 4-7 Installation of the Microprocessor Board and Ribbon Cable onto Module ................ 52
Figure 5-1 Partially Assembled TEG ......................................................................................... 56
Figure 5-2 TEG with CBT Hardware......................................................................................... 56
Figure 5-3 TEG with Air Diffuser ............................................................................................. 58
Figure 5-4 Fully Assembled TEG Mounted to Airtorch and Exhauster ...................................... 58
Figure 5-5 Completed 100 W Test Setup ................................................................................... 62
Figure 6-1 TE Couple Status as Predicted by CBT .................................................................... 68
Figure 6-2 Post-Test Leg Status................................................................................................. 69
Figure 6-3 CBT Prediction Accuracy ........................................................................................ 71
Figure 6-4 Modules Used in Insulation Comparison Test .......................................................... 72
Figure 6-5 Module Voltages during Insulation Comparison Test ............................................... 75

ix

LIST OF SYMBOLS AND ACRONYMS
ΔT

Change in Temperature

°

Degree

μm

Micrometer

AWG

American Wire Gauge

Ba

Barium

C

Celsius

CAN

Controller Area Network

CBT

Couple Bypass Technology

Ce

Cerium

CFH

Cubic Feet per Hour

CFM

Cubic Feet per Minute

cm

Centimeter

Co

Cobalt

CTE

Coefficient of Thermal Expansion

EAS

Electrically Activated Switch

ERS-APU

Energy Recovery System - Auxiliary Power Unit

Fe

Iron

FR

Flame Retardant

FEA

Finite Element Analysis

gal

Gallon

I

Current

kg

Kilogram

x

kPa

Kilopascal

kW

Kilowatt

MJ

Megajoule

ml

Milliliter

mm

Millimeter

mV

Millivolt

MPG

Miles per Gallon

MSU

Michigan State University

OTR

Over the Road

P

Power

REAS

Resistance across EAS

rpm

Revolutions per Minute

RTG

Radioisotope Thermoelectric Generator

Sb

Antimony

Sn

Tin

TE

Thermoelectric

TED

Thermoelectric Device

TEG

Thermoelectric Generator

V

Volt

VAL

EAS Activation-Signal-to-Low-Side Voltage

W

Watt

Yb

Ytterbium

xi

Chapter 1
1. Introduction
The entropy generation that prevents processes from being reversible is often caused in part by
waste heat. Opportunities to improve the overall efficiency of these processes can be realized by
converting some of the energy in the waste heat to useful work. As industry standards and
governmental regulations become more aggressive, this opportunity becomes even more
appealing. In addition, accurately identifying and implementing viable sources of renewable,
alternative energy is necessary to lessen global dependence on fossil fuels and establish a system
of sustainable energy production.

1.1. TE Overview
Thermoelectric (TE) technology is one area of waste heat energy recycling that has received
renewed interest in the last few decades. Thermoelectric Devices (TEDs) utilize the Seebeck
effect to convert some of the thermal energy in a temperature gradient to electricity. Much
progress has already been made with low temperature thermoelectric generators (TEGs) and
several variations of low temperature TEGs made from materials such as bismuth telluride are
currently being manufactured and sold. Recently, research efforts have focused more heavily on
non-heritage, high temperature thermoelectric generators, which feature significant benefits but
also greater challenges. High temperature TEGs open the door to a wider range of applications
than low temperature TEGs alone. However, the higher operating temperatures of these devices
incur higher levels of thermal stresses, placing the thermoelectric components at higher risk of

1

structural failure. Mitigating this risk is necessary to produce durable, reliable, high temperature
TEGs that can undergo several thermal cycles without significant degradation.

1.1.1. History
In 1821 AD, Thomas Johann Seebeck discovered that a compass magnet could be deflected by
joining two dissimilar metals in a circuit and exposing the junctions to different temperatures. It
was later realized that this magnet deflection was induced by an electric current in the circuit.
This effect was further investigated and explained by Jean-Charles Peltier, who discovered that
an electrical current would heat or cool the junctions of the circuit, and by William Thomson
(Lord Kelvin) who summarized that a given material exposed to a temperature gradient will
either absorb or release heat when subjected to an electrical current [2-3].
Thermoelectrics have played a prominent role in NASA’s Radioisotope thermoelectric
generators (RTGs) which have powered several exploratory spacecraft such as Voyager 1 and
Voyager 2 [4-6]. RTGs have also been used for some terrestrial applications, including
lighthouses and navigation beacons [7]. The terrestrial applications for RTGs have been limited,
due to the low system efficiencies and the inherent risk associated with using radioactive
materials for the heat source. Recently, however, advances in material science and a greater push
for “green” energies have sparked a renewed interest in thermoelectric technology, particularly
for the automotive industry, which is facing ever-tightening fuel economy and emissions
regulations.

2

1.1.2. How TEDs work
Modern TE technology utilizes the discoveries made by Seebeck, Peltier and Lord Kelvin to
convert some of the thermal energy in a temperature gradient to electricity. Materials that drive
current toward the hot side when exposed to a temperature gradient are referred to as N-type,
whereas materials that drive current away from the hot side are referred to as P-type. Joining the
hot sides of an N-type material and a P-type material creates a thermoelectric circuit, able to
power a load between the cold sides of the two materials (Figure 1-1). This unit is commonly
referred to as a thermoelectric couple. Combining several of these couples in series allows for
greater power values.

Figure 1-1 Thermoelectric Couple

3

There are several other technologies that exist to convert waste heat into useful work, including
steam turbines and Sterling engines. Compared to these competing technologies, thermoelectrics
have several advantages, including:

High reliability (>250,000 hours)
Silent operation
No vibration
Small electromagnetic signature
Precise temperature control (fractions of a degree)
Position independent
Small and lightweight
Environmentally “green”
Direction of heat pumping is fully reversible
No chlorofluorocarbons, chemicals or compressed gases (nothing to replenish)
Able to function in environments too severe or sensitive for conventional refrigeration

However, it should also be noted that current thermoelectric technology has lower efficiencies
compared to competing technologies [8].

1.2. Project Scope
Much of the research in the thermoelectrics area has been devoted to synthesizing and testing
various thermoelectric materials. This is a rather appropriate place to focus initial research
efforts, as the thermal and electrical properties of the TE materials determine the feasibility of
producing TEDs on a manufacturing scale. To validate the findings from the materials research,

4

a logical next step is to conduct experiments with the thermoelectric materials that simulate realworld conditions. As well as evaluating the performance of the thermoelectric material, these
experiments can also uncover potential roadblocks that may hinder implementation of
thermoelectric technology with the present energy infrastructure.

One of the most potentially lucrative applications for TE technology is integration with over-theroad (OTR) trucks. As much as 35% of the energy generated by the vehicle’s engine can escape
out the tailpipe as waste heat, creating a significant opportunity to reclaim energy [9]. The
temperature and flow rate of exhaust gases from OTR trucks provides an ideal heat source and
the engine’s cooling system can be modified to act as the heat sink. In addition, the high duty
cycle and large idling time experienced by most OTR trucks offers a sizable opportunity for
energy recovery. Thus, to help simulate real-world conditions for the experiment, the heat source
used in these experiments will, as much as practically possible, simulate the exhaust of an engine
used in OTR trucks.

The scope of this Thesis is to experimentally evaluate the thermal, electrical and physical
properties of the skutterudite TEDs by conducting performance tests that simulate real-world
conditions of the potential OTR truck application for TE technology.

1.3. Outline of Thesis
The present chapter introduces the idea of using TE technology to improve process efficiency by
means of waste heat recovery. A brief history and explanation of how TEDs work is given, as
well as the advantages and disadvantages of competing technologies. In addition the scope of
this Thesis is defined.

5

The second chapter covers previous work that has been done in the area of thermoelectrics. The
results of a WAVE simulation of ideal generator location are presented, followed by a projected
cost analysis for mass production of skutterudite TEGs. A finite element analysis of the TE legs
is reviewed, as well as previous generator tests that have been conducted with the skutterudite
material. Lastly, the research questions are posed.

The third chapter describes the fabrication methods for the TEDs. In addition, the measurement
and tracking system is explained.

The fourth chapter introduces the Couple Bypass Technology and describes the design used.
Fabrication techniques are also discussed, as well as the procedures of integrating the technology
with the TEDs.

The fifth chapter details the generator assembly procedure along with the test fixture and
instrumentation. The testing process is outlined and the data collection and post-processing
methods are discussed.

The sixth chapter covers the results of 100 W test, as well as the details and results of the followup insulation comparison test. The results of the Couple Bypass Technology are also presented,
along with the results of the individual couple testing.

The seventh chapter highlights the research achievements made and illustrates the implications
of the testing results. In addition, specific “next steps” are suggested to continue research efforts
in a way that builds on the knowledge gained from these tests. Finally, the research questions
posed in chapter two are answered.

6

Chapter 2
2. Literature Review
Renewed interest in the field of thermoelectrics has led researchers to investigate a host of
various thermoelectric topics. This Thesis and the tests described in the following chapters build
on the foundation of previous work laid by researchers exploring the more theoretical aspects of
thermoelectrics. The topics covered in this chapter are especially relevant to the experiments
detailed in this Thesis.

2.1. Previous Work
While thermoelectrics is still a relatively new field, substantial efforts have been devoted to
investigating and developing thermoelectric materials, optimizing generator designs and
improving the durability of the thermoelectric couple. To guide research and development
efforts, a goal has been established to produce a reliable 1 kW skutterudite thermoelectric
generator with an efficiency of ≈ 10%. To meet that goal, an incremental approach has been
adopted. With this approach, a smaller power goal is set. A generator is assembled from the
existing couple design and then tested. Results are evaluated and changes are made to the TE
couple and generator designs as needed. The generator is tested again and changes from the
previous test are noted. This information is used to make further improvements until the
generator meets or exceeds the power goal. A new power goal is then set and the generator is
scaled up to meet it. This generator scaling often reveals new areas for improvement and the

7

cycle repeats itself. Power goals of 5 W, 25 W and 50 W have been met by previous tests.
Previous 100 W tests produced no more than 75 W despite repeated iterations. This prompted a
change from the previous square generator design to the new, round generator design (see
section 5.1).

In addition to these experiments, simulations have been conducted help guide research efforts in
achieving the aggressive goal of producing a reliable, 1 kW TEG. These theoretical studies have
helped to predict expected operating conditions, explain couple breakages that occur during
operation and provide a realistic benchmark for generator longevity.

2.1.1. Ideal Location
Much work has already been done studying the many factors involved in integrating TEGs into
the exhaust system of OTR trucks. One area of particular interest is the effect of generator
location in the exhaust system on thermoelectric performance. A simulation was performed at
MSU to compare the amount of power generated for three different TEG configurations: one
cylinder per TEG unit, three cylinders per TEG unit and six cylinders per TEG unit [10]. This
was achieved by using Ricardo WAVE Engine System Performance Simulation software to
model a Cummims ISX engine under the three TEG configurations. Table 2-1 below highlights
the findings of this study, which predicted that the design that results in the highest amount of
TEG heat energy input is the one cylinder per TEG configuration. A greater amount of system
heat energy input results in a higher amount of heat energy converted to electrical energy.

Implementing a one cylinder per TEG design validates the power generation goals for this
experiment. These results indicate that to achieve a total of 1 kW of thermoelectric power, it is

8

more practical to install six 170 W TEGs, each fed by a different cylinder, rather than one, 1 kW
TEG, fed by all six cylinders. Thus, achieving 100 W of thermoelectric power marks a
significant milestone in producing a 1 kW TEG system.

Table 2-1 WAVE Simulation Results

2.1.2. Cost Analysis
As one of the long term goals for skutterudite thermoelectric research is to develop a robust, lowmaintenance TEG that can be integrated with current OTR truck engine designs, a
comprehensive cost analysis study is necessary to evaluate the feasibility of mass-producing
TEGs. Determining the payback period for the generator also serves as a realistic benchmark for
longevity and cycle testing. Intuitively, a viable generator design must perform longer than the
payback period to validate the financial venture necessary to purchase, install and maintain the
TEGs.

A preliminary cost analysis was estimated for two skutterudite TEG sizes: 1 kW and 5 kW [11].
Specific dollar amounts were assigned by dividing the generator into subsystems, evaluating
material costs based on an annual production volume of 10,000 units and using standard

9

estimates for material-to-labor cost ratios. Table 2-2 contains the cost estimates for each
subsystem for the 1 kW and the 5 kW generator designs.

Table 2-2 Cost Analysis Results

In addition, the payback period was estimated under the following assumptions:

A 1kW TEG with belt integrated motor-generator is used to convert waste heat to
electrical energy
The cost of diesel fuel is $4/gal and has an energy density of 38.6 MJ/liter
The OTR truck has the following fuel economies: 5 MPG base, 5.02 MPG with TEG
The TEG has a 10% electrical energy conversion efficiency
Vehicle idle consumes 0.8291 gal/hour without TEG versus 0.249 gal/hour with TEG
Vehicle operates 300 days per year
Vehicle spends 8.3 hours on road and 8 hours in idle per day
Vehicle travels 150,000 miles per year

This study found that 120 gallons of fuel would be saved every year from exhaust heat energy
recovery and 1,392 gallons of fuel would be saved every year from idle reduction. At $4 a
gallon, this translates to a yearly savings of $6,048 per year or $42,336 over the 7 year (≈ 1

10

million miles) theoretical life of the engine. Similar estimates were made with the 5 kW TEG. As
seen in Figure 2-1, the 1 kW generator is estimated to pay for itself within 1 year and the 5 kW
generator is estimated to pay for itself within 3 years.

These results have specific implications for this experiment. While the purpose of the test
described in this Thesis is to evaluate total power generation, device integrity should not be
ignored. Even if the power goals set for this test are met, further design improvements will need
to be made if the thermoelectric components being tested fail at a higher rate than mandated by
the payback period. In addition, performance tests will need to be conducted to quantify the
amount of device degradation that occurs with repeated cycling.

Figure 2-1 Payback Period Estimation
For interpretation of the references to color in this and all other figures, the reader is
referred to the electronic version of this Thesis

11

2.1.3. Finite Element Analysis
The interface between the skutterudite and the metallization layers on the P-type leg (see section
3.2) has been identified as a high-risk area to device durability. This has been confirmed by
initial generator power testing which has shown that the most common couple failure mode
during operation is a complete material break at the hot-side skutterudite and Material B
interface. A finite element analysis (FEA) of the TE couple was conducted at MSU to investigate
the effects of thermal cycling on the skutterudite and metallization layer interfaces [12]. A
SIMULIA FEA software package, ABAQUS, was used to model the TE couple and simulate
thermal cycling. The results of this analysis indicated that the large difference in yield strength
between skutterudite and Material B caused the P-type skutterudite to plastically deform when
heated, increasing stress levels at the interface. Figure 2-2 shows the stresses on the couple and
the bulge between the skutterudite and Material B at operating temperature (about 500°C).
Although the stress levels between the metallization layers are higher than those between the
skutterudite and Material B, the lower yield strength and ultimate strength of skutterudite makes
the skutterudite and Material B interface more vulnerable to failures than the interface between
Material A and Material B.

One method of reducing the stress levels at the skutterudite and Material B interface that was
proposed in the study was to reduce the cross-sectional area of the TE leg, thereby reducing the
amount of material subject to thermal expansion. This proposal was accepted and a smaller leg
design that featured half the cross-sectional area of the previous leg design was implemented for
this experiment.

12

Figure 2-2 FEA Simulation of TE Couple

2.2. Previous Tests
While these topics all relate to skutterudite TEDs, their results are all based on simulations and
estimations rather than experiments. To verify these predictions, experiments have been
performed to evaluate the thermoelectric and mechanical performance of the TEDs at various
stages of their development. As changes and improvements to the design of the TEDs have been
made, the test fixtures and methods to conduct these experiments have undergone changes and
improvements as well. Thus, the testing procedures described in the following chapters reflect
the changes that have been made through several iterations of tests and improvements.

The first experiments that were conducted tested the performance of a single module. After this,
multiples modules were tested under a parallel wiring configuration so that if one module failed,
the electrical circuit for the other modules would be uninterrupted. Total power for these tests
was estimated to be the sum of the individual power produced by each module. The TEDs
13

consistently demonstrated a module-level thermal efficiency of about 4%. As experiments were
conducted that tested 5, 10 and even 20 modules, it became common for at least 1 module to fail
during each test. These module failures prompted the suggestion that a method of bypassing
failed couples was needed in order to test generators with a series wiring configuration. As a
result, significant efforts went into developing the Couple Bypass Technology (see chapter 4).

As a result of the FEA described in the previous section, the cross-sectional area of the TE legs
was reduced from 7 mm x 3.5 mm to 3.5 mm x 3.5 mm in order to reduce the amount of thermal
stress at the hot side metallization layers. To evaluate the effects of reducing the cross-sectional
area, three tests have been conducted since the implementation of the smaller leg size design.
The first of these tested a six-couple module, of which one couple was purposefully broken prior
to the test. This module was cycled to a hot side temperature of about 685°C. The second tested a
single 10-couple module which was cycled to a hot side temperature of about 570°C. The third
test cycled two 10-couple modules to 565°C, cooled to room temperature and then cycled again
to 700°C. One of the couples in this test was purposefully broken before the test to evaluate the
performance of the Couple Bypass Technology. None of the 34 couples that were unbroken prior
to these tests broke, even after 19 were cycled a second time in the 2-module test. These results
seemed to support the proposal that reducing leg cross-sectional area results in lower stress
concentrations at the skutterudite and metallization layer interface.

The success rates of these tests prompted the development of the 100 W test described in this
Thesis, using the same fabrication methods and testing procedures of these preliminary tests. The
only changes made to the 100 W test were the module insulation type and the brazing agent used
to bond the legs to the hot shoe. As only a few modules were being tested, commercial insulation

14

Fiberfrax Durablanket-S from Unifrax was sufficient to achieve the desired ΔT for these
preliminary tests. However, the test described in this Thesis uses the same heat source and
contains 20 modules, requiring a more effective insulation to maintain similar ΔT levels. In
addition, a brazing agent comparison test conducted before the 100 W test found that couples
bonded with a new brazing paste produced slightly more power than couples bonded with the
standard brazing foil. Thus, unlike the couples in the preliminary tests, the couples used in the
100 W test were insulated with mica films and aerogel rather than the Durablanket insulation and
were bonded using the paste as the brazing agent, rather than the foil.

2.3. Research Questions
Building on the results of these studies and experiments, the 100 W test was designed to answer
the following questions:

Is the current skutterudite thermoelectric generator design able to reliably achieve a
power level of 100 W? If not, what roadblocks need to be overcome to consistently
achieve this power level?
Can couple failures be accurately and reliably detected and bypassed to allow
thermoelectric generators to be wired in series?

2.4. Summary
A number of previous studies at MSU helped to shape the experiment detailed in this Thesis. A
WAVE simulation validated the power goals set for the experiment. A cost analysis established
the generator longevity goals and the results of a finite element analysis study prompted a leg
size reduction, necessitating changes to the fabrication process described in the next chapter.
15

Chapter 3
3. Fabrication Process
In order to utilize the thermoelectric properties of materials such as skutterudite, it is beneficial
to create individual devices which pair P-type material with N-type material. Manufacturing and
testing these individual devices (known as “couples”) in a standardized, repeatable manner
requires implementation of a fabrication process than minimizes variability. As advancements
have been made in TE research, the fabrication process for the TE couples has undergone many
changes to incorporate the improvements. The following procedures reflect the current
fabrication process used at MSU.

3.1. Materials Synthesis
An essential feature of the TEDs is the TE material itself. Under operating conditions, the ideal
TE material has a low value of thermal conductivity and high values of electrical conductivity
and thermopower. A figure of merit, ZT, has been established to assess the electrical and thermal
properties of a material for use in thermoelectrics:

ZT

S2

*T

Equation 3-1 Figure of Merit [13]
Where σ is the electrical conductivity, κ is the thermal conductivity, T is the absolute
temperature and S is the Seebeck coefficient (or thermopower), which is a measurement of the

16

voltage differential created by a material exposed to a temperature differential, typically
expressed in units of microvolts per Kelvin. As a general rule, a thermoelectric couple should
have a minimum ZT value of 1 to be considered a feasible material for thermoelectric power
generation. Peak values of N-type and P-type skutterudite at 550°C were measured to be
approximately 1.15 and 0.85 respectively [14].

While the first traces of skutterudite were discovered naturally, the procedure for synthesizing
skutterudite is relatively straight-forward, eliminating the need to procure skutterudite materials
from an external source. The procedures detailed in this section describe how the skutterudite is
currently synthesized from basic elements at MSU.

3.1.1. Element Mixing
Synthesizing the skutterudite material is a multi-step process which currently takes 4 days to
complete. N-type and P-type skutterudites are synthesized in 100-200 gram batches. In order to
minimize oxidation, the process is performed almost entirely in either a vacuum or a glove box
with an inert atmosphere.

In the first step of this process, the individual elements are weighed on an electronic balance
(±0.005 grams) and placed in a tapered glassy carbon crucible. These pure elements are acquired
from Alfa Aesar and include Sb ingot (99.9%), Co powder (99.9%), Fe powder (99.999%), Ba
pieces (99.9%), Ce pieces (99.9%) and Yb pieces (99.9%) [15]. This crucible is then placed in a
50 mm diameter quartz tube, removed from the glove box and immediately placed on a high
vacuum sealing line, which consists of a Varian Turbo Mini Pumping Station (Turbo V70LP)
backed by an Edwards RV5 pump. This vacuum sealing line can accommodate quartz tubes

17

ranging from 10 mm to 50 mm in diameter and can reach a vacuum pressure of 1x10
achieve a pressure of ≈ 4x10

-6

-6

torr. To

torr, the compound is held under vacuum for about 24 hours

before the quartz tube is sealed by placing a slightly smaller diameter quartz plug inside the tube
and using a torch to flame seal the tube.

3.1.2. Synthesis and Annealing
On the second day of the process, the sealed quartz tube is removed from the vacuum line and
placed in a tube furnace where it is ramped up to 1100°C at a rate of 10°C per minute and held at
temperature for about 15 hours. The tube is then air quenched to ensure smaller grain size of the
skutterudite precursors. To convert these precursors into a homogeneous skutterudite compound,
the mixture is then annealed for 48 hours at 740°C and 700°C for the P-type and N-type
skutterudites, respectively.

3.1.3. Powder Processing
The last step is to powder process the skutterudite ingot for hot pressing. After the compound has
annealed for 2 days, the quartz tube is moved into the inert atmosphere double glove box and is
broken to release the glassy carbon crucible. The ingot is then removed from the crucible and is
crushed and ground for five minutes using a motorized mortar and pestle. The powder is then
sieved for 30 minutes using a 75 μm test sieve. Any material that doesn’t pass through the sieve
is reground for 5 minutes and sieved again. This is repeated until approximately 99% of the
skutterudite material has passed through the 75 μm test sieve (typically 3 iterations). All the
powder is then collected and placed inside a 500 ml stainless steel ball mill jar along with 7

18

stainless steel balls (20 mm diameter) where it is ball milled for 3 hours at 110 rpm. This reduces
the average particle size to approximately 34.9 μm. The skutterudite powder is then removed
from the grinding jar and stored in a glass jar until hot pressing.

3.2. Leg Fabrication
While skutterudite has great thermoelectric properties, it fails to form an adequate bond with the
copper hot shoes. To address this, metallization layers are added on the top and bottom of the TE
legs to provide buffers between the skutterudite material and the hot shoe and cold shoe,
respectively. The metallization layers for the P-type legs consist of Material B, followed by
Material A and copper foil. The metallization layers for the N-type legs consist of Material A
followed by copper foil (Figure 3-1).

Material A forms a good bond with the copper foil, which forms a good bond with both the
solder used on the cold side and the brazing agent used on the hot side. As the linear coefficient
of thermal expansion (CTE) of Material A is roughly the same as the linear CTE of the N-type
skutterudite, no additional metallization layers are necessary. However, the P-type skutterudite
has a linear CTE approximately 33% larger than that of Material A, which would result in high
levels of shear stress at the skutterudite and Material A interface during operation. Thus, Material
B is used as a buffer between the P-type skutterudite and the Material A layer. Material B has a
linear CTE roughly the same as P-type skutterudite, but has a greater yield strength by a factor of
6. This enables Material B to better withstand the high levels of shear stress caused by the
difference in linear CTE values.

19

Figure 3-1 Skutterudite Metallization Layers

3.2.1. Hot Press
These metallization layers are sintered with the skutterudite in a Thermal Technology HP20014020-23G hot press system. This system can achieve a temperature of 2200°C in a vacuum or
inert gas and can exert a force of up to 100 tons on a 12 inch diameter by 12 inch high work
zone. The temperature and force in the work zone are programmable for simultaneous and
independent automatic control. A typical hot press run for a single ingot takes about 6 hours.

To prepare the materials for the hot press, a cylindrical, graphite, powder pressing die is
internally lined with Grafoil, a flexible graphite material by GrafTech International. The die has
an outer and inner diameter of 6 inches and 2 inches, respectively. After lining with Grafoil, a 2
inch diameter graphite plug is inserted into the bottom of the die such that the bottom of the plug
is flush with the bottom of the die. A 2 inch diameter copper foil is then placed inside the die and
glued on top of the plug. The assembly is then moved into the double glove box to complete the
preparation in an inert atmosphere.

20

To prepare the N-type skutterudite, Material A powder is first weighed and then placed in the
powder pressing die on top of the copper foil. The powder is then cold pressed to 3 tons with a
Carver Lab hydraulic press. The die is rotated 90° and cold pressed again. After this, a 2 inch
diameter Material C foil is placed on top of the Material A powder. The N-type skutterudite
powder is then measured and placed on top of the Material C foil. The powder is cold pressed,
rotated and cold pressed again. In the same manner, another layer of Material C foil is placed on
top of the skutterudite powder and another layer of Material A powder is placed on top of the
Material C foil. After the final cold pressing, a 2 inch diameter graphite plunger is placed on the
powder before the die is loaded into the hot press.

To prepare the P-type skutterudite, Material A powder is first weighed and then placed in the
powder pressing die on top of the copper foil. The powder is then cold pressed to 3 tons, rotated
and pressed again in the same manner as the N-type leg. A 50-50 mixture of Material B powder
and P-type skutterudite is added on top of the Material A layer and is cold pressed. A layer of
pure P-type skutterudite is added next, followed by another layer of the 50-50 Material B and Ptype skutterudite mixture and a final layer of Material A, cold pressing after each layer is added.
After the final cold pressing, a 2 inch diameter graphite plunger is placed on the powder before
the die is loaded into the hot press.

The die is then loaded into the hot press and the work zone is vacuumed for approximately 1
hour to achieve a pressure of 100 millitorr. After this pressure is reached, the work zone is
backfilled with high-purity argon gas (99.999%) to atmospheric pressure and is then vacuumed
overnight. To begin the hot press run, the temperature is ramped up to 250°C over 20 minutes to
draw out any residual oxygen from the graphite. After this, the work zone is backfilled with

21

argon gas again until a positive pressure of 1 psi is achieved. The argon is then allowed to vent to
the atmosphere and a gas flow rate of ≈ 40 liters/hour is maintained for the remainder of the run.
Over the next 20 minutes, the temperature is ramped up to 500°C and the hydraulic ram is
ramped up to 15,500 kg. After this, the temperature is ramped up to 730°C in 20 minutes under
constant force. This is held for 2 hours before the ram is released and the temperature is ramped
down to 600°C in 20 minutes. The solid is then allowed to cool to room temperature over 2
hours. The hot press run compresses the material from a height of about 10 mm in powder form
to about 7 mm. This solid is called an ingot and is given a unique number for tracking purposes
(Figure 3-2).

Figure 3-2 Skutterudite Hot-Pressed Ingots

3.2.2. Dicing
After hot pressing, the ingots are diced into legs using a KO Lee grinder/slicer, which is isolated
inside a soft wall clean room. The grinder’s programmable controller and variable high-speed
motor are used to automatically and precisely cut the skutterudite ingots into legs (Y and Z axis
± 0.0001 inch). The diamond cutting blade is 0.017 inches thick and is actively cooled with
Omni-cut FC-100 coolant. Figure 3-3 illustrates the cutting diagram for the ingots. It takes about
3 hours to cut both 2 inch diameter ingots into 3.5 mm x 3.5 mm legs.
22

To prepare the legs for dicing, both ingots are crystal bonded to a graphite-covered steel plate.
When the bond sets, the plate is transferred to the KO Lee grinder/slicer, where it is magnetically
locked into place. The cutting blade is then adjusted to the correct position and the dicing
program is started. After dicing, the steel plate is removed from the grinding machine and placed
on a hot pad where it is heated to about 285°C to soften the glue. The legs are then removed from
the steel plate and washed in acetone to remove the crystal bond residue. Each N-type and P-type
leg pair, designated by matching numbers (see section 3.5.1), is stored in a numbered, glass vial
until the bonding process

Figure 3-3 Ingot Grinding Diagram
23

3.3. Couple Assembly
There are several desirable characteristics for a TE couple that have been identified and
incorporated into the couples that are currently produced at MSU. An ideal TE couple contains
the structural framework to achieve a large temperature gradient while maintaining an electrical
circuit that is free from shorts and has high electrical conductivity for maximum power
production. A heat collector (commonly referred to as a hot shoe) is used to electrically connect
the hot sides of a P-type leg and an N-type leg. The legs should be a standard size and shape to
ease fabrication efforts and improve unit modularity. The couple must be physically robust
enough to withstand reasonable handling, which is necessary to assemble the couples into
modules. These TE couples can be thought of as the building blocks of a thermoelectric
generator; greater generator power values must include more TE material, which is accomplished
by integrating more couples into the generator design. Figure 3-4 illustrates the current couple
design.

Figure 3-4 Current TE Couple Design
24

3.3.1. Hot Shoe Design
A vital component of the TE couple is the hot shoe. The ideal hot shoe design maintains a solid
bond with the legs, conducts a maximum amount of heat and minimizes electrical resistance
between the hot sides of the two legs. Depending on the mode of heat transfer, the hot shoe may
also need to allow for differences in thermal expansion between the two legs.

When the primary heat transfer mode is conduction, as is the case with many of the current lowtemperature TEG designs, the hot shoe is shaped to maximize heat transfer from the heat source
to the legs. It must also be electrically insulated from the heat source to prevent a short circuit.
This is the likely hot shoe design for TEGs which must operate in an inert atmosphere.

As convection is the primary mode of heat transfer in the experiment, a new hot shoe design was
needed to more effectively transfer heat from the working fluid to the TE legs. A finned design
was proposed which increased the surface area of the hot shoe by a factor of 15 for the same
cross-sectional area. More importantly, the new design featured an inverted “U” shaped bridge in
the middle of the hot shoe which allowed the shoe to flex (Figure 3-4). This flexibility in the hot
shoe is important to reduce thermo-mechanical stresses due to different rates of thermal
expansion between the P-type and N-type legs. A study by Sakamoto et al. was conducted to
estimate the difference in thermal expansion between the legs when heated to a ΔT of 600°C
[16]. This difference was compared to the amount of deflection the hot shoe was able to achieve
via the inverted “U” shaped bridge. It was concluded that the hot shoe design was more than able
to achieve the 10 microns of deflection needed to accommodate the differences in thermal
expansion between the N-type and P-type legs at this temperature.

25

The hot shoes are fabricated using the following method. 1/2 x 3/8 inch copper bar stock is cut
and milled into 19/64 inch lengths and deburred. After this, four 0.06 inch wide slots are cut into
the copper blocks with a band saw, creating 4 thinner outer fins and 1 thicker middle fin. The hot
shoes are deburred again and placed fin-side down in a metal jig. The jig is then placed in the
KO Lee grinder/slicer described in section 3.2.2 which grinds a slot down the middle fin. This
cut forms the inverted “U” shaped bridge which provides the shoe’s flexibility. The hot shoes are
then removed from the jig and stored in an acetone bath until the bonding process.

3.3.2. Bonding
The next step in the couple assembly process is to physically bond a skutterudite leg pair to a hot
shoe. Currently, the bonding agent used in the fabrication process is Fusion 1000 brazing paste
(STL-1000B-658). The skutterudite legs are housed in a graphite jig and aligned with the copper
hot shoes with the bonding agent between the leg and hot shoe interfaces. The hot side is heated
with a 3 inch diameter oxygen substrate button heater while the cold side is actively cooled with
water. The entire bonding procedure is performed in an inert atmosphere inside a Kurt J. Lesker
stainless steel vacuum chamber and takes about 2 hours.

The current bonding run procedures begin by selecting 23 leg pairs and sanding both ends of
each leg with 400 grit sandpaper to remove oxidation. The legs are then washed in 100% ethanol
for at least two minutes and then washed again in fresh ethanol. Meanwhile, 23 copper hot shoes
are selected and sanded with 400 grit sandpaper to remove oxidation and any burs incurred by
the machining and grinding process. These hot shoes are then etched with tracking numbers
corresponding to the numbers of the 23 selected leg pairs, after which the hot shoes are washed
in ethanol for at least two minutes and then washed again in fresh ethanol.

26

A 3 inch diameter graphite jig is used to align the legs with the hot shoes. The jig can
accommodate 23 leg pairs at a time and is divided into four sections: the base, the leg mount, the
hot shoe mount and the cap. The base supports the leg mount which houses the leg pairs. The leg
mount has 2 features that allow the legs to thermally expand with minimal amounts of stress:
spring-loaded ceramic rods and spacers made of Grafoil. When properly loaded into the jig, the
skutterudite legs rest in paired holes on ceramic rods that are constrained by die compression
springs, allowing the legs to expand axially when heated. The Grafoil spacers align the legs away
from the inner wall between each hole pair. As the hot side is heated, the copper hot shoe
expands laterally until the maximum temperature is reached, melting the paste and bonding the
leg to the hot shoe. As the system cools down, the copper hot shoe contracts, pulling the nowbonded legs inward. If the legs are positioned too close to the inner wall before the bonding run,
this inward contraction by the hot shoe will cause shear stress at the leg and hot shoe interface,
possibly resulting in a break. The Grafoil spacers give the legs room to move with the
contracting hot shoe by initially positioning them on the outer edges of the holes.

While the components are being washed in ethanol, forty-six 1/16 x 3/32 inch Grafoil spacers are
cut and placed into their respective slots on the leg mount. After the second ethanol bath, the legs
are dried and placed into the circular holes such that a P-type leg will be bonded with a samenumbered N-type leg. A bead of brazing paste is then placed on the top surface of each leg via a
plastic syringe and the hot shoe mount is placed on the leg mount. The hot shoes are dried and
then placed on top of the leg pairs, such that the number etched onto each hot shoe matches the
number of the corresponding leg pair. The cap is then placed on top of the hot shoe mount and
the entire jig is placed in the bonding apparatus inside the vacuum chamber. This apparatus is
designed to heat the top of the jig while actively cooling the bottom. The jig is placed on a water-

27

cooled, spring-loaded copper plate and is insulated with two Fiberfrax Duraboard blocks that are
coated with a fireproof sealant. The jig is then secured in place with a 3 inch diameter oxygen
substrate button heater which is tightened down two turns after the springs under the cooling
plate begin to compress. Two more Duraboard blocks and several Durablanket strips are used to
insulate the button heater and the top of the jig. The control thermocouple is then secured in a
designated hole in the cap, just above the hot shoes. The lid of the vacuum chamber is then
bolted down and the chamber is vacuumed with a BOC Edwards XDS10 dry scroll pump for 5
minutes. The chamber is then backfilled with a 95/5 argon/hydrogen gas mixture to atmospheric
pressure and is vacuumed for another 5 minutes. This gas mixture is used in place of high-purity
argon gas to reduce the amount of residual oxygen molecules in the chamber by bonding them
with hydrogen. The chamber is again backfilled with the gas mixture which maintains a
continuous flow of 8 cubic feet per hour (CFH) for the remainder of the bonding run. Once the
pressure inside the bonding chamber reaches 2 kPa (gage), the one-way atmosphere valve is
opened to allow the gas mixture to vent. The cooling water is then turned on and the bonding run
program is initiated.

The button heater is controlled by a programmable Heatwave Labs temperature controller. The
bonding run program ramps the temperature up to 700°C in 29 minutes and then dwells for 15
minutes before ramping down to room temperature over 20 minutes. After the system has
sufficiently cooled, the jig is removed from the vacuum chamber and the cap and hot shoe mount
are removed. If any couples are found to have broken legs after the bonding run, the location of
the breakages on the leg and the location of the broken legs in the jig are documented. The nonbroken couples are then stored by ingot numbers.

28

3.3.3. Individual Couple Testing
To assess theoretical power potentials and cycling durability for individual couples and to
evaluate the effect that changes in the fabrication process have on power production, a couple
testing stand was fabricated. The testing stand is enclosed inside a Kurt J. Lesker Pyrex bell jar
station to perform the tests in an inert atmosphere. The couple testing stand is similar to the
bonding station; a water-cooled, spring-loaded copper plate actively cools the cold side of the
couple while a button heater, controlled by a Heatwave Labs temperature controller raises the hot
side temperature to a set value. In the couple testing stand, however, the couple is not placed in a
graphite jig. Instead, the hot shoe is in direct contact with the button heater, allowing more
precise control over couple hot side temperatures.

To prepare a couple for testing, the couple legs are soldered to separate copper blocks which act
as heat sinks and the whole assembly is placed on the copper cooling plate in the bell jar station.
Voltage leads and thermocouples are placed such that the voltage difference across each leg is
measured along with the hot and cold side temperatures. The control thermocouple is placed in a
designated hole in the hot shoe and the couple is insulated with Durablanket. The bell jar is then
sealed and the station is purged of oxygen using the same vacuum and backfill method described
in the previous section. After backfilling the station with the 95/5 argon/hydrogen gas mixture a
second time, the gas flow rate is adjusted to maintain a continuous flow of about 4 CFM
throughout the duration of the test. After the test stand reaches a pressure of 0.5 kPa (gage), the
one-way atmospheric valve is opened and the testing program is started. The temperature
controller ramps the temperature up to 656°C in 30 minutes and then dwells for 15 minutes

29

before ramping down to room temperature over 25 minutes. After the hot side temperature has
sufficiently cooled, this cycle is repeated another 8 times.

Eighteen couples of the same and similar material used in the 100 W test were tested using this
method.

3.4. Module Assembly
As the power produced by an individual couple is relatively limited (typically around 0.5 to 1
W), it is advantageous to combine several couples into a modular subassembly that electrically
connects the couples in series, enhances durability and rigidity, establishes a common cold side
heat exchanger and improves ease of generator installation. For example, given an average
couple power output of 0.5 W, a 100 W generator would require installing 200 couples.
However, if these couples were assembled into 10-couple modules, a 100 W generator would
require installing only 20 modules. Thus, the next step in the fabrication process is to assemble
the TE couples into 10-couple modules.

3.4.1. Module Board
At the heart of the TE module is the module board. The circuit design of this high-temperature,
2-layer, FR-4 board connects the couples in series and houses 10 electrically activated switches
(EASs), ten 0603 resistors and 3 ribbon cable connectors to accommodate the Couple Bypass
Technology components described in Chapter 4. As is seen in Figure 3-5, the module board
contains 20 holes which align with the legs of 10 couples and 2 holes for 4-40 brass mounting
bolts.

30

Figure 3-5 Unpopulated Module Board

The first steps in the module assembly process are to solder the ribbon cable connectors to the
module board with high-temperature, lead-free solder and the heads of the mounting bolts in
their corresponding holes with SnPb 60/40 solder. The board is then etched with a tracking
number as well as the ingot numbers of the legs that will be used in the module. Four 0.055 inch
diameter through holes are then drilled into the module board, using the clamping bolt holes in
the cold plate discussed in section 3.4.3 as a guide. After four 00-90 bolts are inserted though
these holes, 00-90 nuts are used to fasten them into place and the bolt heads are secured to the
module board with epoxy. After the epoxy has set, the cold plate is removed and the module
board is transferred to the soldering station. Figure 3-6 shows the module board at this point.

Figure 3-6 Module Board with EASs, Ribbon Cable Connectors, Tracking Number,
Mounting Bolts and Clamping Bolts
31

3.4.2. Insulation
As the power generated by a TED is directly proportional to the ΔT across the skutterudite leg,
exceptional insulation of the TE couples is critical to overall generator performance. As such,
considerable emphasis has been placed on researching various insulation materials and casting
methods. One high performance insulation technology that is receiving serious consideration is
titania-opacified aerogel [17]. This silica-based insulation features excellent conductive and
convective insulation properties and the added titania effectively scatters infrared radiation. As
the aerogel material is very porous and delicate, quartz fibers are added to improve durability.
This combination of silica-based aerogel, titania and quartz fibers are referred to in this Thesis
simply as “aerogel.” While aerogel fabrication methods are beyond the scope of this Thesis, it
should be noted that fabrication procedures have been modified from a cast-in-place method to a
form-casting method, greatly easing fabrication efforts. In addition to the aerogel forms, 0.875
by 4.5 inch mica films are used to improve module durability and provide additional insulation
by overlapping gaps between modules. A finished aerogel form is shown in Figure 3-7.

Figure 3-7 Aerogel Form

Once the aerogel forms are ready, 10 TE couples are sanded and placed in the slots in the mica
films and aerogel forms. The assembly is then rotated such that the legs protruding through the
aerogel are facing upwards. The module board is then placed on top of the aerogel form, aligning

32

the holes with the legs. A soldering iron is used to heat individual pockets, which are filled with
SnPb 60/40 solder until flush with the module board. An intermittent pattern is used to more
evenly apply the heat and to avoid unintentional de-soldering of existing components. Once all
the pockets are filled, flux remover spray is used to clean the solder pads. Heat shrink wrap is
then added to the mounting bolts to electrically insulate the cold plate from the bolts. Figure 3-8
and Figure 3-9 show the module at this point. As these pictures were taken after the 100 W test
was conducted, significant oxidation of the copper hot shoes can be seen in these figures. The
clamping bolts are not shown for Figure 3-9.

Figure 3-8 Module Board after adding Aerogel Mold, Mica Film and Soldered Couples
(Side View)

Figure 3-9 Module Board after adding Aerogel Mold, Mica Film and Soldered Couples
(Bottom View)

33

3.4.3. Cold Plate
The function of the cold plate is to direct the cooling fluid as close to the solder pads as possible
while providing a physical barrier between the cooling fluid and the electronics on the module
board. The cold plate also comprises the frame of the round generator design; each plate is
designed to cover 18° of the 5.9 inch diameter of the generator, completing the round shape
when all 20 modules are fixed in place.

The cold plates are fabricated using the following process. 3/8 x 1 inch aluminum bar stock is cut
and milled into 125 mm lengths. A 9° angle is machined on both sides and ten 1.25 mm pockets
are milled on the inner face to accommodate the CBT switches. In addition, ten 0.75 mm pockets
are milled on the inner face to accommodate the resistors. Two 5 mm diameter through holes are
drilled for the mounting bolts, four 0.055 inch diameter through holes are drilled for the
clamping bolts and three 1.5 mm slots are milled for the ribbon cable connectors. Two 1/4
diameter holes are drilled and tapped lengthwise through the plate for the cooling fluid. The cold
plates are then anodized at Great Lakes Metal Finishing Inc. to provide electrical insulation from
the TEDs.

Once the plates have been anodized, they are etched with a module number and bolted to the
module board, aligning the CBT switches with their corresponding slots. The four 0.055 inch
diameter through holes in the cold plate are then used to align the holes for the clamping bolts in
the module board, as described in section 3.4.1. After the TEGs have been soldered to the
module board, the cold plate is coated with a thermal compound (Céramique by Arctic Silver)
and is again bolted to the module board, with hard-fiber washers electrically insulating the
mounting bolts and nuts from any electrically conductive “hot spot” that may have been exposed

34

during assembly. After the clamping bolts are re-bolted, the module is ready for generator
assembly. The final module fabrication steps, which connect the CBT hardware to the module,
are completed after the generator is mounted to the Airtorch. Figure 3-10 shows the module at
this point. As this picture was taken after the 100 W test, significant oxidation of the copper hot
shoes can be seen in this figure.

Figure 3-10 Module with Cold Plate

3.5. Measurements
As TEG fabrication is still in the early stages of development, many cause-and-effect
relationships have yet to be determined. To gain a holistic understanding of the diverse factors
that affect TEG performance, a tracking system was established and various properties and
measurements such as electrical resistivity were recorded. As these cause-and-effect correlations
are developed, more intelligent decisions can be made to optimize the fabrication process and TE
couple design.

35

3.5.1. Tracking System
To enable investigation of the relationship between both thermoelectric and mechanical
performance and various leg, couple and module properties, a numerical tracking system was
established. Properties such as ingot number, location in the ingot, electrical resistivity and
bonding run information are recorded and documented for each leg. A number is assigned to
each leg, which is also used to pair N-type legs and P-type legs for bonding into TE couples.
Figure 3-11 illustrates the leg numbering system with respect to ingot location.

During the bonding process, each skutterudite leg pair is matched with a copper hot shoe etched
with the same number. Module boards are etched with the ingot and module numbers and the
cold plate is etched with the module number as well. Thus, at any point during fabrication or
assembly, it is possible to track the status of an individual leg, from its location in the ingot, to its
location in a module to its location in the generator itself. This is crucial for identifying any
effect that factors such as material concentrations, hot press conditions, bonding run conditions
and electrical resistivity may have on thermoelectric performance.

36

Figure 3-11 TE Couple Tracking Diagram

3.5.2. Electrical Resistivity and Short Circuit
It has been proposed that there is a possible correlation between electrical resistivity and
thermoelectric performance. Higher resistance values require higher voltages to achieve the same
current, reducing the amount of power a TE couple can produce. Thus, identifying processes or
fabrication methods that result in lower resistance values for the couples can result in higher
power values for the TEG. To better understand the effect that each step of the fabrication
process has on the electrical resistivity of the legs, electrical resistivity measurements are taken
after the legs are cut from the ingot, after the legs are bonded to the hot shoes and after the
couples have been assembled into modules.

37

Two separate jigs are used to measure the electrical resistivity of the legs and the couples. The
operating method of both of these jigs is the same: two copper leads provide the electrical
contact between the legs and a Kepco power supply. One or both of the copper leads is springloaded to secure the leg/couple in the jig. The voltage drop across the leg/couple is measured
with a multi-meter while applying a current of 1 amp with the power supply. These two values
are then used to calculate the electrical resistance of the leg/couple. The electrical resistivity of a
module is measured without a jig and instead uses alligator clips attached to the mounting bolts
to complete the circuit. Figure 3-12 shows the loaded and unloaded jigs for measuring the
electrical resistivity of the TE legs and couples.

Figure 3-12 Electrical Resistivity Measurement Jigs

38

After the electrical resistivity measurements, the modules undergo an electrical short circuit test.
Because of the high levels of electrical resistivity of the thermoelectric legs compared to the
aluminum material that comprises the cold plate, an electrical short circuit would result in almost
no power generation, based on voltage divider principles. As such, it is critical to prevent shorts
in the circuit connecting the thermoelectric couples. After module fabrication and just before
installation in the generator, the modules are all tested for short circuits by placing one lead of a
multi-meter on the mounting bolt and the other lead at various locations on module, including the
clamping bolts and the surface of the cold plate, and measuring the resistance. If all
measurements result in infinite resistance, the module is ready for installation in the generator.
Any discovered short circuits are electrically insulated and the module is tested again.

3.6. Summary
The process of fabricating components for the TEG is three-fold. First, pure elements are mixed,
placed in a furnace and then annealed to synthesize the skutterudite compounds. This compound
is pulverized, layered with film and powdered metals and hot pressed into ingots. These ingots
are then cut into TE legs. In the second stage, the legs are arranged into TE couples, consisting of
one P-type leg, one N-type leg and a common hot side heat exchanger. These components are
bonded together with brazing paste. During the final stage, the couples are arranged into TE
modules. Each module houses 10 couples, all wired in series, and provides each with a common
cold side heat exchanger. The module also houses the CBT, insulation and cooling infrastructure.

At each stage, electrical resistivity measurements are taken to aid in establishing correlations
between both thermoelectric and mechanical performance and leg, couple and module properties.
This is facilitated via the established tracking system.

39

Chapter 4
4. Electronic CBT
The power produced by thermoelectric devices is directly proportional to the temperature
difference between the hot side and the cold side of the thermoelectric legs. In order to achieve
the desired power output of 0.5 W per couple, a considerable temperature difference of about
475°C is required. This results in a significant temperature gradient of over 1700°C per inch.
These extreme conditions in which the thermoelectric couples must operate pose a considerable
risk to the integrity of the thermoelectric legs. As the couples are wired in series, this risk
jeopardizes the overall power of the generator; if just one couple in a 10-couple module fails
structurally, it breaks the electric circuit for the other nine couples, effectively negating their
contribution to the overall power. Higher voltages require multiple modules to be wired in series,
multiplying the power loss caused by a broken couple. To address this risk, the Couple Bypass
Technology (CBT) was developed.

4.1. Overview
The CBT is an electronic solution that incorporates the use of electrically activated switches
(EASs) to protect against couple failures. When a particular EAS is closed, it establishes a lowresistance path in parallel with the corresponding thermoelectric couple. This is useful when a
couple degrades to the point where it is no longer contributing power to the circuit, but is acting
more like a load due to an increase in internal resistance that accompanies couple degradation.
More importantly, the low-resistance path established by a closed EAS enables the generator

40

circuit to bypass a couple that has failed to the point that it acts as an open circuit, allowing
utilization of the power produced by the other non-broken couples in the series circuit. However,
experiments have shown that about 0.5 W -the amount of power each couple is predicted to
produce during operation- is absorbed by a closed EAS. This power must be supplied by the
other couples in the series; if one couple fails, the power produced by another couple is absorbed
by the closed EAS while bypassing the failed couple. Thus, while the EASs provide a way to
bypass degrading or failed couples, they place an additional load on the circuit in the process.
Because of this, CBT for future generators should not be viewed as a way to accommodate
fragile TE material or sloppy fabrication and installation procedures, but rather as a short-term
solution to maintain a closed circuit and minimize power losses in the TEG until adequate repairs
can be made.

As skutterudite research is still in the early stages, design improvements such as material
robustness and bond durability are ongoing, compelling present generator testing research to be
completed with the present TE couple design, which is known to have integrity liabilities at the
hot side interface. To account for these liabilities, the present CBT design monitors the status of
every couple during generator testing and is able to bypass any individual couple that either has
degraded to the point of acting as a load on the circuit, or has physically fractured such that it
acts as an open circuit. Both of these failures compromise total power production, which is the
primary goal for the 100 W test. Thus, for the present generator testing research, the CBT is
designed to bypass any failed couple as needed to minimize power losses. As skutterudite
materials technology becomes more advanced, such that couple breakages during operation are
rare, this design can be simplified to monitor and short entire modules, rather than individual
couples, reducing the complexity and power requirements of the CBT.

41

4.2. Design
The primary objective of the CBT is to provide the capability to bypass any TE couple that fails
during the 100 W test. In order to short any couple, one EAS is wired in parallel with each
couple and is closed only when a failure is present (Figure 4-1). To detect a couple failure,
measurements are taken across every couple to obtain the voltage differential. In addition, the
voltage change across each module is measured directly using a voltage divider.

When a couple slowly degrades during operation, the internal resistance of the TE legs gradually
increases. Once the couple has degraded to the point that the measured voltage differential across
the couple falls below a specified threshold for a certain amount of time, the corresponding EAS
is closed, allowing the current to flow through the EAS, which has lower resistance than the
failing couple. While this bypassing reduces the voltage drop across the couple, the voltage
change across the EAS is still negative and is on the order of 20-60 mV, resulting from the
following relation:

VEAS

I * REAS

Equation 4-1 Voltage Load for Failed couple [18]

42

Figure 4-1 EAS Placement Schematic [19]

43

4.2.1. Hardware
The present CBT design features 2 circuit boards: the module board and the microprocessor
board. The module board (described in section 3.4.1) houses the EASs as well as 3 ribbon cable
connectors for communicating with the microprocessor board. The microprocessor board is
populated with a Freescale MC9S08DZ60 microcontroller and supporting circuitry, including 3
ribbon cable connectors to enable communication with the module board. Depending on the
testing objectives, this microprocessor board can be programmed to control the EASs
independently and automatically, or with a supervisory controller. Communication with the
optional supervisory controller and data acquisition system is achieved via a CAN (Controller
Area Network) transceiver. Future CBT designs that utilize module-level, rather the couple-level
bypassing methods will feature simplified circuitry and reduced power requirements, enhancing
the feasibility of powering the microprocessor board internally. Figure 4-2 displays the top view
of the microprocessor board.

Figure 4-2 Microprocessor Board

A significant hardware consideration that needed to be addressed when designing the CBT was
how to supply sufficient EAS activation-signal-to-low-side voltage for each of the 200 seriesconnected EASs. The resistance across the EAS, denoted REAS, is a function of the difference
44

between the activation signal voltage and the low side voltage, which are indicated in Figure 4-3.
Larger values of this voltage difference, denoted VAL, correspond to smaller values of REAS and
thus, a smaller voltage drop across the EAS. Ideally, this voltage difference should be around 812 V to minimize REAS and the power absorbed by a closed EAS.

Figure 4-3 EAS Activation Signal and Low Side Voltages
If a couple is functioning normally and the corresponding EAS is open, the VAL term is
insignificant since no current is passing through the EAS. The 12 V power supply for the
microprocessor board supplies the activation signal voltage for a closed EAS. Thus, the
activation signal voltage for a closed EAS is 12 V. However, as the couples and modules are
wired in series, the low side voltage increases after each non-broken couple. This is illustrated in
Figure 4-4.

45

Figure 4-4 Additive Principle of Low Side Voltages
As the low side voltage for the first couple in Figure 4-4 is tied to ground and is therefore zero,
the value of VAL for the first EAS is simply the activation signal voltage, 12 V. Assuming that
couple 1 is non-broken and thus producing power, the low side voltage for the second couple is
the value of low side voltage 1, plus the voltage produced by couple 1. Thus, the value of VAL

46

for the second EAS is 12 V minus the voltage produced by couple 1. Following the same logic
and assumptions, the value of low side voltage 200 is the sum of the voltages produced by all of
the previous couples. This reduces the value of VAL for EAS 200 to 12 V minus the voltage
produced by couples 1 through 199. This increasing low side voltage steadily decreases the value
of VAL which raises the value of REAS and thus, the voltage drop across the EAS. As each
couple is expected to produce an open-circuit voltage of about 0.15 V, this adding of low side
voltages is significant, reducing the value of VAL to zero by the 80th couple.

To account for this incrementally increasing value of low side voltage, a floating power supply
was installed in each microprocessor board to establish a local floating ground. The floating
power supply uses a DC-DC convertor to provide a 12 V supply to the microprocessor board
with a floating ground tied to the positive terminal (or output voltage) of the previous module.
For example, if the first module in the generator produces 1.5 V, the value of the local ground for
the second module is 1.5 V, relative to the system ground. Additionally, the power supply is
offset by 1.5 V, relative to the system ground. Thus, while the local voltage values of power
supply and ground for the microprocessor board are 12 V and 0 V respectively, the voltage
values relative to the system ground are 13.5 V and 1.5 V, respectively. So while the absolute
ground and supply voltages are different for each microprocessor board in the generator, they all
maintain a voltage difference of 12 V. More importantly, by offsetting the ground and supply
voltages by the positive terminal voltage of the previous module, the floating power supply
effectively “resets” the low side voltage of the first couple in each module back to a local value
of 0 V, regardless of the increase of low side voltage in the previous module. This sets the VAL
value of the first EAS in each module back to 12 V (Figure 4-5).

47

While the increase in low side voltage across each module is thus isolated from the subsequent
modules in the generator, the voltage gain across each functioning couple increases the low side
voltage for the subsequent couples within that module. However, as the voltage produced by
each couple is limited, the total increase in low side voltage within a single module is not
significant. Assuming that each couple in a module produces 0.15 V, the first EAS in a module
would have a VAL value of 12 V and the last EAS would have a VAL value of 10.65 V. These
both fall within our ideal range, resulting in low values of REAS and thus, a minimal voltage
drop across the EAS. Minimizing the voltage drop across the EASs is critical as the very purpose
of the CBT is to facilitate maximum power production from the TE couples by minimizing
power losses

Figure 4-5 Benefits of Floating Power Supply

48

4.2.2. Software
The task of designing and coding the programming logic to enable the microcontrollers to
monitor the differential voltages across the couples and control the EAS activation was
outsourced by MSU. After the code was written in C and compiled, PE Micro hardware and
software was used to flash the code onto the microprocessor board at MSU.

Under the current program, the microprocessor boards power up in the default mode of
automatic control and actively monitor the differential voltages across each couple, the
OPEN/CLOSED status of each EAS and the overall voltage change across the entire module. If
the differential voltage of any couple drops below the default threshold value, a counter is
started. If the voltage remains below the threshold value until the counter reaches the default
delay count, the microcontroller closes the corresponding EAS. Once an EAS has been activated,
it will not reset until the module is powered down, or is manually instructed to reset by the
supervisory controller.

4.2.3. Data Acquisition and Supervisory Control
For current testing purposes, additional coding is included in the software for data acquisition
and optional, manual control of EAS activation. This communication is achieved via a dSpace
AutoBox and the CAN. The present configuration of the data acquisition program incrementally
requests information from the microcontrollers, cycling through all 20 modules every 500
milliseconds. The information collected and displayed for real-time observation during testing
includes the differential voltages across each couple, the OPEN/CLOSED status of each EAS
and the overall voltage change across the entire module. Because all of this information is

49

transferred through a single CAN Bus, the requests are addressed to identify which module is
transmitting the information.

The external controller can also send addressed commands to the microcontrollers to manually
open or close specific EASs as needed for testing purposes. There are three commands available
to the external controller: close EAS and remain in manual mode, open EAS and remain in
manual mode and open EAS and switch to automatic mode. When an EAS is in manual mode, it
will not change its open/closed status until it is powered down, unless instructed to do so by the
supervisory controller. When the EAS receives a command from the supervisory controller to
open and switch to automatic mode, the microcontroller again begins monitoring the voltage
differential across the corresponding couple and will automatically close the EAS if the voltage
again drops below the threshold for a specified amount of time. Figure 4-6 shows the schematic
for the entire CBT system.

In order to sustain sufficiently high values of V AL, each of the microprocessor boards has a
floating power supply which ties the floating ground to the positive terminal of the previous
module (see section 4.2.1). Because of the difference in local ground values relative to the
system ground, each of the microprocessor boards contains an isolated CAN transceiver to
enable communication with the supervisory controller.

50

Figure 4-6 CBT Schematic [20]

4.3. Fabrication
The module board fabrication is currently outsourced to Hughes Electronics where the EASs and
resistors are added to the high temperature FR-4 circuit board. The ribbon cable connecters are
added at MSU before the board is populated with the TE couples and mounted to the cold plates
(see section 3.4.1). The 4-layer, FR-4 microprocessor boards are currently fabricated at
Semiconductor Hybrid Assembly Inc. where they are populated with the microcontrollers,
isolated CAN transceivers, ribbon cable connectors and supporting circuitry and hardware. The
microcontrollers are flashed at MSU before the microprocessor boards are installed onto the
modules.

51

The final assembly step for module fabrication is to install the microprocessor board onto the
module. This is completed after the generator is mounted to the Airtorch to prevent damage to
the microprocessor boards and accompanying ribbon cables (see section 5.1.2). After the
generator has been secured, a second nut is added to each mounting bolt. This is to provide
adequate spacing for the microprocessor boards, which are oriented such that the ribbon cable
connectors on the microprocessor boards align laterally with the ribbon cable connectors on the
module boards. The microprocessor boards are then bolted to the outside of the module using a
third nut to secure the boards to the mounting bolts. Three Parlex Corporation 8-channel ribbon
cables are then inserted into the ribbon cable connectors of each module board and the
corresponding microprocessor board. Figure 4-7 illustrates this installation process with a single
module.

Figure 4-7 Installation of the Microprocessor Board and Ribbon Cable onto Module

52

4.4. Summary
To achieve the high voltages levels required for commercial feasibility, it is necessary to wire
many TE couples in series. This configuration places the power production of the entire circuit at
risk in the event of a failed couple. The Couple Bypass Technology is an electronic solution that
incorporates the use of electrically activated switches to provide a low-resistance path in parallel
with any couples that fail during operation. This not only protects against open circuit failures,
but also power losses that occur from extensive couple degradation. The CBT allows higher
voltage generators to be constructed without the risk of a failed couple compromising the power
production of the entire generator.

53

Chapter 5
5. Testing
As stated earlier, the focus of this Thesis is to apply the knowledge gained from theoretical
thermoelectric research to an experiment simulating real-world conditions. Experiments tend to
reveal characteristics of the specimens being tested that are not predicted by theoretical research.
As such, it is expected that several iterations of testing, analyzing and quantifying the results,
adjusting the testing parameters as needed and then testing again will be necessary to fully
understand and control the behavior and performance of these TEDs. This large-scale, multipleunit testing described here represents the first of its kind with skutterudite thermoelectrics.

5.1. Generator Design
The area of application in question for TEG implementation is the exhaust system of OTR
trucks. The potential waste heat recovery opportunity lies in transferring the heat in the exhaust
flow to the TE legs, via the hot shoe heat collectors. One of the most effective modes of
transferring heat from a fluid to a surface is jet impingement. However, while jet impingement
features superior heat transfer characteristics, it also brings several components of the current
TEG design that are subject to oxidation at the high operating temperatures into direct contact
with the working fluid. This direct contact renders the current TEG design unsuitable for jet
impingement with flows containing oxygen, such as Diesel exhaust. However, the tremendous
heat transfer benefits associated with jet impingement merit further investigation of ways to
integrate skutterudite TE technology in oxygen-rich flows without causing oxidation.

54

Consequently, the current generator design uses a flow divider to channel fluid streams onto each
of the hot shoes, enabling investigation of the effects of using jet impingement as the method of
heat transfer.

5.1.1. Module Housing
The round TEG design is made up of 2 endplates and the 20 modules themselves. The 2
endplates are 6.5 inch diameter aluminum rings with 40 tapped through holes for coolant access,
4 tapped through holes for mounting and 2 coolant ports. The endplates feature an internal
channel that connects all 40 through holes with the 2 coolant ports. The two 9° angles cut into
each cold plate enable the modules to be assembled in a circle, with the hot shoes facing the
center (Figure 5-1). The modules are bolted to the endplates, using o-rings to ensure a water-tight
seal. The mica films are staggered such that they overlap one another to mitigate air leaks
between the modules. After both endplates have been secured, Durablanket insulation strips are
placed at both ends to further prevent heat loss. This unit is referred to as a generator and is fully
assembled after it is secured to the Airtorch and final components have been installed (Figure
5-2).

55

Figure 5-1 Partially Assembled TEG

Figure 5-2 TEG with CBT Hardware
56

5.1.2. Heat Source
To derive logical values for heat source flow rate and temperature, corresponding values from a
Cummins ISX engine [21] were used. These conditions were simulated with a 12 kW MHI
Airtorch (MTA925-12) and controller. The Airtorch is capable of achieving temperatures of
925°C with an air flow rate of 14.4 cubic feet per minute (CFM). A steel, cylindrical air diffuser
featuring 20 rows of five 0.785 inch diameter holes was bolted to the outlet of the Airtorch to
provide the jet impingement. A mica gasket was placed between the air diffuser and the Airtorch
to seal the interface. The TEG was then bolted concentrically to the Airtorch around the air
diffuser such that none of the hot shoes came in contact with the air diffuser (Figure 5-3). A
high-alumina ceramic gasket was placed between the air diffuser and the generator to seal the
interface. A stainless steel nozzle was then bolted to the outlet of the generator, using a mica
gasket to seal the interface (Figure 5-4). The nozzle features a vented exhauster which connects
to a building exhaust hose. The vents prevent a pressure drop inside the generator when the
building exhaust hose is pumping at a greater rate than the gas is being supplied. The Airtorch
and the nozzle were insulated with Durablanket to minimize heat loss to the surroundings.

As mentioned in section 5.1, the current TEG design exposes several components that are subject
to oxidation at higher temperatures to the working fluid. In OTR truck applications, this working
fluid is oxygen-containing Diesel exhaust. As the purpose of this experiment is not to test
methods of insulating the oxygen-sensitive components from the flow, but rather to test the
performance of the TEDs under temperature and flow conditions similar to those in the exhaust
system of an OTR truck, nitrogen gas was used as the working fluid. This allowed the test to be
performed with environmental conditions similar to air, but without the oxidation effect.

57

Figure 5-3 TEG with Air Diffuser

Figure 5-4 Fully Assembled TEG Mounted to Airtorch and Exhauster

58

5.1.3. Cooling System
The endplates described in section 5.1.1 function as the coolant distributer and receiver for the
generator. After the coolant supply lines are attached to the coolant ports of the upstream
endplate, cooling fluid is able to access the entire endplate via the channel that connects all 40
through holes.

The bolts used to fasten the modules to the endplates are banjo bolts which feature a lateral
through hole in the threads which intersects with an axial hole through the shaft. This allows
fluid in the endplate channel to enter the shaft of the bolt laterally and exit axially, which is
exactly how the coolant accesses the 2 lengthwise through holes in each cold plate. The coolant
exits the cold plate into the downstream endplate in the same manner. Attaching the coolant
return lines to the coolant ports of the downstream endplate completes the coolant loop.
To test the TEDs over a greater ΔT range, a 5-ton Temptek CF series chiller was used to create a
closed-loop cooling cycle with a 50/50 propylene glycol-water mixture as the coolant.

5.1.4. Instrumentation & Data Acquisition System
After the generator is secured to the Airtorch, the microprocessor boards and ribbon cables are
installed as described in section 4.3. Nineteen staggered copper strips are then soldered on to the
mounting bolts of sequential modules to connect all 20 modules in series. At this point, the
generator hardware is completely installed.

To enable heat transfer analysis and to gain a greater understanding of the forces at work inside
the generator during operation, a data acquisition system was implemented. Omega OMB-DAQ-

59

56 Data Acquisition Modules and the accompanying Personal Daqview software system were
used for the data acquisition. Measurements of all properties were recorded every three seconds
during the test. All temperatures were measured using Omega K-type and T-type standard
thermocouples. Module voltages were measured with 22 AWG wire soldered to the mounting
bolts of each module. Organized by system, the following properties were measured:

Pre-heater

Nitrogen temperature (K-type thermocouple)
Nitrogen flow rate, measured using a Meriam 50MH10 laminar flow element

Heater

Nitrogen temperature after Airtorch heater (K-type), grid of 2

TEG

TED hot shoe temperature (K-type), located on couple 6 in module 20
TED cold side temperature (K-type), located in the module board below couple 6 in
module 20
Stream wall temperature (K-type), located next to couple 6 in module 20
Module voltage

Exhaust

Nitrogen temperature (K-type), grid of 5

60

Coolant

Coolant flow rate, measured using two Omega FLR 1000 flow sensors
Inlet coolant temperature (T-type)
Outlet coolant temperature (T-type)

During the test, the power produced by the generator is obtained by measuring the generator
voltage at different currents. A Kepco power supply is used to back load the generator at
specified current levels. The current is incremented and the voltage for each current is
documented to find the maximum power according to the common relation: P VI . The
theoretical voltage at which maximum power occurs can be calculated using the following
equation:

VO.C .
2

PMax

Equation 5-1 Maximum Power Calculation Equation
Where PMax is the maximum power and VO.C. is the open-circuit voltage.
To visualize the amount of power produced by the generator, an incandescent light bulb bank
was fabricated. The light bulb bank was wired in parallel and could accommodate up to 24
individual 14 V, 4.9 W light bulbs. To account for voltage uncertainty of the generator, the bulbs
were tested to estimate maximum voltage before failure, which occurred around 32 volts.

In addition, every 500 milliseconds, the CBT software documented the voltage differential across
each couple, as well as the overall voltage change across each module. A visual couple status
indicator was created to display the status of each couple in the generator. Couples that
61

performed above the voltage threshold were indicated by a green light and couples that had
fallen below the voltage threshold were indicated by a red light. A counter was implemented in
both data acquisition systems to enable synchronized data comparison between the two systems.

The final generator assembly can be seen in Figure 5-5.

Figure 5-5 Completed 100 W Test Setup

5.2. Testing Process
Up to this point, the processes described have reflected current fabrication and assembly
procedures. The following describes the process used during the 100 W test:

1. Attach building exhaust hose to the vented exhauster and turn on the pump
2. Set nitrogen flow rate to 12 CFM
3. Set chiller temperature to -6°C and coolant flow rate to ≈ 8 liters/min
4. Begin recording measurements from data acquisition system

62

5. Begin continuous monitoring and testing of the CBT via the supervisory controller
6. Ramp Airtorch temperature up to 925°C over 30 minutes
7. Once Airtorch is at temperature, increase nitrogen flow rate to maximum possible while
maintaining Airtorch temperature ≥ 915°C; adjust chiller temperature if necessary
8. Measure generator open-circuit voltage and divide by 2 to estimate maximum power
voltage
9. Back load generator with power supply at current levels above and below maximum
power voltage estimation to calculate maximum generator power
10. Use maximum power values obtained in step 8 to determine the number of light bulbs
needed for light bulb bank
11. Load generator with light bulb bank and photograph
12. Ramp Airtorch temperature down to room temperature in 30 minutes
13. After generator temperature has fallen below 75°C, close nitrogen flow
14. Stop recording measurements from data acquisition system
15. Turn off chiller
16. Turn off building exhaust hose pump

5.3. Post-processing Data
As with previous tests, the data collected from the 100 W test were compiled and post-processed
to obtain the following information:

Power Levels
o As power generation is the main function of the TEG, one of the most important
post-test calculations is the amount of power that was produced by the generator

63

Steady-state hot side and cold side temperatures
o Calculating these temperatures allows evaluation of the effectiveness of different
TED insulation types and generator insulation
o This calculation can also help assess how realistic the test was by comparing
these values to known exhaust temperature values for target engines.

Maximum and average ΔT
o As the power generation of TEDs is a function of ΔT, it is necessary to obtain
these values to normalize performance

Heat transfer
o Calculating the amount of heat transferred through the TEDs is necessary to
calculate thermal efficiency
o Calculating the heat transfer of the nitrogen gas and the cooling fluid allows an
evaluation of the TED insulation

Module voltage
o Establishing a relationship between voltage production and ΔT is critical to
determine optimal operating temperatures
o Module voltage data can also pinpoint couple breakage times, helping to
establish relationships between environmental conditions and couple failures

64

One of the most important post-processing operations is calculating the thermal efficiency of the
TEG. Thermal efficiency is defined as the useful energy output of a device divided by the energy
input. Knowing the thermal efficiency of the TEDs allows for a normalized performance
comparison between different thermoelectric materials and aids in predicting the level of energy
recovery possible by implementing TE technology with existing processes.

To calculate the thermal efficiency of the generator, the mass flow rate and the inlet and outlet
temperatures of the cooling water are used to calculate the rate of heat transfer to the water.
Assuming that a negligible amount of heat transfer occurs across the aerogel insulation, all of the
heat gained by the water must pass through the legs of the TE couples. Dividing the electrical
energy production rate of the generator (the useful energy output) by the rate of heat transfer to
the water plus the electrical production rate of the generator (the energy input) gives the thermal
efficiency of the TEDs. Previous TEG designs have yielded thermal efficiencies of about 4%.

5.4. Summary
The testing process for the TEDs is accomplished by assembling the modules into a round
generator and heating the hot sides of the TE couples with nitrogen gas via jet impingement.
Concurrently, the cold sides of the TE couples are actively cooled with a propylene glycol-water
mixture. This provides the ΔT necessary for the couples to generate a voltage and allow testing
of the TEG. The testing process follows a standardized procedure to measure maximum power
generation and temperature values. A vast network of instrumentation for data acquisition is used
to measure and calculate these values. A standardized method of data post-processing provides
calculations such as average ΔT and thermal efficiency.

65

Chapter 6
6. Results & Discussion
The previous 100 W generator tests were performed with TEDs made up of legs with twice the
cross-sectional area of the legs used in the test described above. While the greater cross-sectional
area of these legs allowed greater current flow and thus greater power generation per leg, it also
increased hot-side interface stresses due to thermal expansion, resulting in material breakage
during testing (see section 2.2). This material breakage prevented the generator from producing
the targeted 100 W. The cross-sectional area of the legs used in this test was reduced for the twofold purpose of lowering the current to a more practical value and more importantly, to reduce
the thermal stresses during operation in order to prevent material breaks at the hot-side interface.

As described in section 2.2, two single-cycle tests of 1 module each and one dual-cycle test of 2
modules were conducted prior to the 20-module, 100 W test to assess the material integrity of the
smaller leg size. None of the couples broke during these tests, seemingly indicating lower levels
of thermal stress. Apart from the brazing agent used and the number of couples tested, the only
major deviation from these tests that was made for the 100 W test was the use of Aerogel
insulation with mica film inserts instead of Durablanket. Thus, similar success rates were
anticipated for the 100 W test.

66

6.1. Power Results
As the test process was initiated, it was noted that two couples had already failed and had been
automatically bypassed via the EASs. It was concluded that these couples likely broke during
generator installation. As the generator continued to heat up, more couples failed. By the time the
hot shoes had reached 450°C, the CBT was reporting that over half the couples had already
broken. This unexpected mass couple failure made it impossible to realize the goal of producing
100 W. In addition, voltage measurements for some of the modules had railed to various values,
eliminating the opportunity to accurately measure power generation. Thus, the power
measurement steps were omitted and the cool down process was initiated.

After the test, 156 of the 200 couples were found to have a least one broken leg. Of those 156
broken couples, 155 had a broken P-type leg and 8 had a broken N-type leg. All leg breakages
occurred at the hot side skutterudite and metallization layer interface.

6.2. CBT Results
During the test, it was unclear how the CBT was performing. The visual couple status indicator
was displaying a much higher number of broken couples than was expected and the displayed
module voltages were different than the sum of the displayed voltages across their individual
couples. It was determined after the test that this discrepancy was caused by a difference in the
methods used by the microcontrollers to measure the voltage differential across individual
couples and to measure the voltage change across the modules. In addition, an inside trace failed
on module 4 which severed the communication to the entire module and railed the measured
voltage differential for each couple in that module to a value above the fail threshold.

67

As the generator was cooling down, a picture was taken of the visual couple status indicator just
before the CBT control system was shut down (Figure 6-1). According to the indicator, 140
couples had fallen below the voltage threshold and therefore had at least one broken leg. After
the generator was disassembled, 156 couples were found to have broken legs. Figure 6-2
displaces a grid showing the actual leg status of each couple.

Figure 6-1 TE Couple Status as Predicted by CBT

68

Figure 6-2 Post-Test Leg Status

69

All of the couples that were predicted by the indicator to have broken legs were found to have at
least one broken leg. In addition, 16 couples that were predicted by the indicator to have no
broken legs were found to have at least one broken leg. This discrepancy is likely due to the fact
that the CBT software was deactivated before the generator had finished cooling. As the hot
shoes of the couples were still around 400°C when the picture was taken, it is quite possible that
the leg breakages in the 16 couples that were incorrectly predicted to have no broken legs
occurred after the picture was taken, either while the generator was cooling, or during generator
disassembly. This is supported by the fact that no couples that were found to be unbroken were
incorrectly predicted by the indicator to have broken legs. In the best case scenario, the CBT
correctly predicted the integrity status of 100% of the couples, assuming that each of the 16
incorrectly predicted couples broke after the picture was taken and excluding module 4, which
lost communication. In the worst case scenario, the CBT correctly predicted the integrity status
of 91.6% of the couples, assuming that each of the 16 incorrectly predicted couples broke before
the picture was taken and excluding module 4. Figure 6-3 displays a grid showing the indicator’s
predicted status and the actual status of each couple

70

Figure 6-3 CBT Prediction Accuracy

71

6.3. Insulation Comparison Test
As the insulation type and brazing agent used were the only major changes that were made from
the earlier generator tests that had 100% couple success rates, another test was performed to
evaluate the effect of insulation type on couple structural integrity. Another 3 modules were
fabricated from TE material similar to what was used in the 100 W test. One module was
insulated with aerogel and a mica film, another was insulated with just aerogel and the third was
insulated with Durablanket, the type of insulation used in the previous 2-module tests (Figure
6-4).

Figure 6-4 Modules Used in Insulation Comparison Test

6.3.1. Test Setup
As many variables as possible were left unchanged for the insulation comparison test. Similar TE
material was used, the same bonding process and module assembly methods were used and the
same CBT setup and control procedures were used. As the 17 modules not being tested were

72

needed to complete the generator framework, they were included in the test as well but were not
connected to the microprocessor boards, power, or load. Including these 17 modules in the
insulation comparison test also helped to reduce variability from the 100 W test.

One significant deviation from the previous test was the way in which power was measured. The
three modules were not wired in series as in the previous test; instead, each module was
connected to a separate load. During the 100 W test, as well as most of the tests before it, the
modules were back loaded with a Kepco power supply at incremental current levels and the
corresponding generator voltages were measured to calculate the generated power. For this test,
power was measured by sequentially loading each individual module with a 3 level resistor bank:
0.11 ohms, 0.44 ohms and 0.77 ohms. The current was measured for each of these resistance
levels and power generation was calculated using the common relation: P VI .

The testing procedure was similar to that of the 100 W test, but with a different method of
calculating power generation. The light bulb visualization was not used either, as the purpose of
this test was not total power generation, but insulation comparison. Additional thermocouples
were installed in the hot and cold shoes of the center couple in each module to measure the
variation of ΔT between the different insulation types. The three modules being tested were
spaced in the generator such that each test module had a non-tested “dummy” module on each
side. The modules were placed in the following order: Durablanket-insulated module, dummy
module, aerogel-and-mica-insulated module, dummy module, aerogel-only-insulated module.
The three modules being tested were connected to their respective microprocessor boards and the
CBT supervisory control software was modified to accommodate 3 modules, instead of 20.

73

6.3.2. Results
The voltage measurements for the insulation comparison test provide an insightful look at the
effect of CBT on module voltage over the duration of the test. For the beginning portion of the
test, the voltages for all three modules steadily rose as the copper hot shoes climbed to a
temperature of about 420-450°C. As seen in Figure 6-5, the voltages of the two modules with
aerogel insulation peaked and started to decline after 1600 seconds. As there was a steady
external resistance of 0.11 ohms on the generator during this portion of the test, this voltage drop
was due to an increase of internal resistance in one or more of the thermoelectric legs, an
indication of gradually failing structural integrity. Recall that the CBT software is written such
that once a couple reaches a certain negative voltage for a specified amount of time, the
corresponding EAS is closed, giving the circuit a low-resistance path in parallel with the failing
couple. The sharp voltage jumps in the graph indicate EAS activation times for the module. The
sharp increase in voltage occurs since the module current is no longer forced to pass through the
high resistance legs of the failing couple, but can instead pass through the low-resistance EAS.

As the generator was heating up, it was discovered that one of the couples in the Durablanketinsulated module was already broken. It is likely that this break occurred during generator
assembly. After the hot shoes reached a temperature of about 420-450°C, 7 couples in the
aerogel-only insulated module broke, while 8 couples in the aerogel-and-mica insulated module
broke. During the cool down portion of the test (not shown in Figure 6-5), 1 couple in the
Durablanket-insulated module broke. Thus, over the entire test, the module with Durablanket
insulation had only one couple break that was not broken before the test began.

74

Figure 6-5 Module Voltages during Insulation Comparison Test

While the modules that were insulated with aerogel experienced much higher couple failure rates
than the module insulated with Durablanket, it is doubtful that the insulation materials
themselves are responsible for the breakages. Instead, it is likely that the insulation type directly
affects additional variables that in turn, directly affect the thermal stress on the couples. While
these additional variables are still unknown, there seems to be a correlation between temperature
and couple failures. As seen in Table 6-1, the modules that were insulated with aerogel
experienced significantly higher hot side temperatures and maximum ΔT values than the module
insulated with Durablanket.

75

Table 6-1 Maximum ΔT and Hot Side Temperatures for Insulation Comparison Test

6.4. Individual Couple Testing Results
Couples that were individually tested (see section 3.3.3) had a ΔT of about 600°C and produced
an average of ≈ 1 W. This is almost twice the power produced from the average couple with
similar ΔT values tested in the generator. One reason for this discrepancy is the higher
resistances that accumulate when combining couples into modules. However, electrical
resistance measurements of modules are typically only 10-20% higher than the sum of the
electrical resistance measurements of the couples in the module. In addition, none of the 18
couples that were cycled in the individual couple test stand failed, or even suffered moderate
degradation. Thus, it has been theorized that the compression experienced by the TE couples
during the individual couple testing procedure may help improve power output and material
integrity.

76

6.5. Summary
During the 100 W test, over 75% of the TE couples failed, increasing the difficulty and
nullifying the usefulness of any power calculations. After the test, it was discovered that 155 Ptype legs had broken and 8 N-type legs had broken, all at the hot side skutterudite and
metallization layer interface. As the most significant design change from the previous, smaller
and highly successful generator tests was the insulation type, another test was conducted with 3
modules, each with a different insulation type. The two modules that were insulated similar to
the modules of the 100 W test had a 75% couple failure rate, while the module that was insulated
similar to the smaller tests had an 11% failure rate during the test. However, the two modules
that were insulated similar to the modules of the 100 W test had a significantly higher maximum
ΔT and hot side temperatures than those of the module insulated similar to the smaller tests.
Further investigation is needed to determine the true cause of the couple breakages.

77

Chapter 7
7. Conclusions
The experiments that are performed early on in a research area are rarely conducted without a
few unexpected results. Further investigation of these apparent “anomalies” can often lead to
new discoveries and a better understanding of the science involved. As Richard Buckminster
Fuller noted, “There is no such thing as a failed experiment, only experiments with unexpected
outcomes” [22]. Thus, while the 100 W test may not have reached the stated goal of producing
100 W, there are other results and discoveries that we can learn from these “unexpected
outcomes.”

7.1. Research Achievements
A procedure for generator fabrication, assembly and testing has been established. This
procedure provides a method to accomplish the following tasks in a standardized,
repeatable manner:
o Material Synthesis
o Leg Fabrication
o Bonding Process
o Module Formation
o CBT Integration
o Generator Assembly and Testing

78

A tracking system has been established. This system can be used to track an individual
leg from its location in the hot-pressed ingot to its eventual position in the TEG. This
facilitates the opportunity to evaluate the effect that various, adjustable parameters have
on TE performance.

The CBT has been successfully tested with supervisory control. The system was able to
accurately detect couple failures and successfully establish a low-resistance path through
EASs in parallel with the failed couples to maintain the closed-loop circuit and thus,
overall power generation.

A possible correlation between insulation type and couple structural integrity has been
discovered. Current observations and assumptions suggest that a module’s insulation type
likely affects a third variable (such as ΔT) which in turn, directly affects the structural
integrity of the couples in that module.

7.2. Conclusions of Research
From the outcomes of these generator tests and those previously conducted, a number of
conclusions can be drawn regarding the optimal performance of TE generators.

Couple Bypass Technology

Until skutterudite material advancements are made such that the number of couple failures
becomes insignificant, a method of bypassing failed couples will be necessary to measure series

79

power generation. Even after such advancements are made, a module bypass system will likely
be needed in any practical implementation of TE technology to mitigate the effects of a broken
couple and allow power production for the rest of the generator until repairs can be made. The
CBT tested in these experiments has shown great potential to meet these bypass requirements
and should be included in future generator tests to further evaluate and improve the accuracy and
automation of the technology. After material advancements are made, the CBT can be easily
modified to monitor and bypass on a module level, rather than a couple level.

Couple Failures

The single greatest roadblock currently facing skutterudite research is the high level of thermal
stresses at the interface between the hot shoe and the skutterudite. Each of the observed failures
occurred between the skutterudite and the upper metallization layer: Material B for the P-type
and Material A for the N-type. While these thermal stresses compromise the structural integrity
of both N-type and P-type legs, the failure rate of the P-type legs was significantly higher than
that of the N-type leg. Thus, it is logical to assume that the composition of the metallization
layers also affects the structural integrity of the legs. The high failure rate observed in these tests
must be addressed in order for TE generators to be feasible for practical application.

7.3. Recommendations for Future Work
While much work remains in TE research, efforts made in the following areas will build on the
framework already established and provide additional insight into the results of these tests.

80

Couple Bypass Technology

As described above, the CBT tested in these experiments has shown great potential to meet the
bypass requirements necessary to test multiple modules wired in series without jeopardizing
overall power generation. To further develop this useful technology, additional testing and
improvements should be made that focus on automatic control. Up to this point, all CBT testing
has featured supervisory control which included manual resetting of the EASs. Testing the
automatic mode of the CBT would be useful to either validate the current system, or to expose
any programming errors or system bugs. Additionally, the discrepancy between reported module
voltages and the sum of the reported individual couple voltages warrants a closer look at the two
different methods used by the microcontroller to make these measurements. Testing will be
needed to assess the accuracy of each method. Following these tests, design changes may be
necessary to improve the accuracy of one or both methods. Efforts should also be made to
condense the hardware currently populated on two separate boards onto one control board. This
would eliminate the ribbon cables and accompanying connectors as well as the losses and
vulnerabilities associated with them. Finally, as the CBT was developed and integrated after the
round generator design was implemented, it has been suggested that a new generator design
could be developed that takes CBT thermal conditioning and spacing needs into account from the
very beginning.

Insulation Effect

The results of the insulation comparison test seemingly indicate a correlation between insulation
type and couple breakages. This correlation is likely due to a third variable affected by the
insulation type, such as ΔT or hot side temperature. Additional tests should be conducted to

81

verify the findings of the insulation comparison test and investigate possible causes of the couple
breakages in order to determine methods of reducing the number of these failures.

Brazing agents

The high success rates of the previous generator tests of this leg size conflict with the low
success rates of the generator tests discussed in this Thesis. As mentioned before, the only major
deviations from the previous generator tests were the insulation type and the brazing agent.
Preliminary results from the insulation comparison test seem to indicate a correlation between
insulation type and couple success rates. However, it would also be insightful to run a side-byside brazing agent comparison test to investigate the effect of the different brazing agents on
couple success rates. While the failures occur at the metallization layers rather than the leg and
hot shoe interface, the uncertainty of the physical and chemical reactions that occur at high
temperatures warrant an investigation of the effect of the brazing agent on couple integrity.

Leg Cross-Sectional Area

The level of thermal stresses experienced by the TE legs during testing is a function of the crosssectional area of the legs. FEA studies of the interface of the skutterudite and metallization layers
indicate that legs with larger cross-sectional areas experience higher levels of stress caused by
thermal expansion [23]. While these studies have already prompted the reduction of the TE leg
size (see section 2.2), a further reduction of the leg size could reduce the levels of thermal stress
even further, improving leg survival rates. This reduction in leg size is not without
disadvantages, however. The amount of current a leg produces for a given load is proportional to
the cross-sectional area. Thus, for a given voltage, a leg with a larger cross-sectional area will

82

produce higher current, and thus greater power, than a leg with a smaller cross-sectional area. To
achieve similar power levels then, a greater number of smaller cross-sectional area legs is
needed, which increases couple fabrication cost and time. Additional tests should be conducted
to evaluate the success rates of legs with smaller cross-sectional areas. This will aid in
determining if the reduction in thermal stresses gained by reducing the cross-sectional area of the
leg size outweighs the increase in fabrication cost and time.

Compression Testing

While the success rates of the generator tests vary, the success rates of the individual couple
testing (described in section 3.3.3) are consistently 100%, even after multiple cycles. In addition,
the power produced by the average couple during individual couple testing is about twice the
power produced by the average couple during generator testing. One of the main differences
between the generator tests and the individual couple tests is the method of heat transfer. The
generator tests rely on forced convection via jet impingement, while the individual couple test
uses direct conduction, which places the couple being tested under compression. It has been
hypothesized that the greater power values and high success rates of the individual couple testing
may be a result of this compression. Two options for testing these theories have been proposed:

The generator test setup could be modified to place the couples under compression
during testing

The individual couple test stand could be modified to use jet impingement instead of
direct conduction as the method of heat transfer

83

7.4. Summary
The research questions posed in chapter two:

Is the current skutterudite thermoelectric generator design able to reliably achieve a
power level of 100 W? If not, what roadblocks need to be overcome to consistently
achieve this power level?
o The high level of thermal stresses at the skutterudite and Material B interface
compromises the integrity of the P-type leg during operation. The resulting
breakages in the P-type legs are the single greatest roadblock to consistently
producing higher power levels in skutterudite TEGs. While the current generator
design is fully capable of achieving 100 W, attaining this power level consistently
will require design improvements that drastically reduce the thermal stresses at
the skutterudite and metallization layers in the legs.

Can couple failures be accurately and reliable detected and bypassed to allow
thermoelectric generators to be wired in series?
o Preliminary results from the CBT tested in these experiments indicate that the
technology is fully capable of bypassing broken couples when operated under
supervisory control. Additional testing will be needed to assess the automatic
control capabilities and the measurement accuracy of the technology. In addition,
compiling the hardware onto a single board would ease fabrication and assembly
efforts and improve module durability. However, the CBT shows great promise in
fulfilling the bypass requirements for current and future skutterudite TEGs.

84

REFERENCES

85

REFERENCES
[1] J. Sakamoto, H. Schock, T. Caillat, J. P. Fleurial, R. Maloney, M. Lyle, T. Ruckle, E. Timm,
L. Zhang, “Skutterudite-based thermoelectric technology for waste heat recovery: Progress
towards a 1kW generator,” submitted to Science of Materials: Special Edition on
Thermoelectrics, Dec, 2010, accepted Feb, 2011, pp. 2-3, 11*
[2] Yunus A. Cengel and Michael A. Boles, Thermodynamics: An Engineering Approach, 6th
Ed, McGraw-Hill, New York, NY, 2008, p 650-651
[3] Caltech, “Brief History of Thermoelectrics,”
http://www.thermoelectrics.caltech.edu/history_page.htm, (accessed March 25, 2011)
[4] J. Sakamoto, et al, “Skutterudite-based thermoelectric technology for waste heat recovery,”
p. 3*
[5] G. L. Bennett, “Space Application,” In CRC Handbook of Thermoelectrics, ed. D.M. Rowe,
CRC Press LLC, Boca Raton, FL, 1995
[6] Robert D. Abelson, Tibor S. Balint, Kathryn E. Marshall, Heros Noravian, James E.
Randolph, Celeste M. Satter, George R. Schmidt, James H. Shirley, “Enabling Exploration with
Small Radioisotope Power Systems,” 2004, available from
http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/40856/1/04-10.pdf (accessed March 25, 2011)
[7] R.C. O’Brien, R.M. Ambrosi, N.P. Bannister, S.D. Howe, H.V. Atkinson, “Safe radioisotope
thermoelectric generators and heat sources for space applications,” Journal of Nuclear Materials
377, Issue 3, (July 15 2008), pp. 506-521, available from
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6TXN-4SCKRXJ2&_user=1111158&_coverDate=07%2F15%2F2008&_rdoc=1&_fmt=high&_orig=gateway&_o
rigin=gateway&_sort=d&_docanchor=&view=c&_searchStrId=1693536156&_rerunOrigin=goo
gle&_acct=C000051676&_version=1&_urlVersion=0&_userid=1111158&md5=1153b61c17e8
8f29e9b5f895ae238ca0&searchtype=a (accessed March 25, 2011)
[8] Harold Schock, Eldon Case, Tim Hogan, Mecouri Kanatzidis, John Miller, Jim Novak, Fang
Peng, Edward Timm, Tom Shih, Bin Zhu, Thierry Caillat, Jeff Sakamoto, Wayne Eckerle, Todd
Sheridan, Charles Cauchy, “Thermoelectric Conversion of Waste Heat to Electricity in an IC
Engine Powered Vehicle: Phase 1 Final Report,” submitted to US Department of Energy, 2005,
p. 5
[9] Ibid, p. 4
[10] Ibid, pp. 12-27

86

[11] Harold Schock, Eldon Case, Tim Hogan, Jeff Sakamoto, Fang Peng, Edward Timm, Trevor
Ruckle, Jim Winkelman, Tom Shih, Mecouri Kanatzidis, Thierry Caillat, Jean-Pierre Fleurial,
Christopher Nelson, “Thermoelectric Conversion of Waste Heat to Electricity in an IC Engine
Powered Vehicle: Phase 2 Final Report,” submitted to US Department of Energy, 2009, pp. 3340
[12] Andreas Panayi, “Thermoelectric Module – Finite Element Analysis,” Project Report,
Thermoelectrics, Harold Schock, MSU, 2010
[13] Rama Venkatasubramanian, Edward Siivola, Thomas Colpitts, Brooks O'Quinn, “Thin-film
thermoelectric devices with high room-temperature figures of merit,” Nature, 413, 2001, pp.
597-602, available from
http://www.nextremethermal.com/media/pdf/Nature413597602October11%202001.pdf
(accessed March 25, 2011)
[14] J. Sakamoto, et al, “Skutterudite-based thermoelectric technology for waste heat recovery,”
p. 9*
[15] Ibid, p. 4*
[16] Ibid, pp. 12-13*
[17] Ibid, p. 5*
[18] K. Moran, “Use of shorting (switches) in the thermoelectric generator,” Project Report,
Thermoelectrics, Harold Schock, MSU, 2011, p. 4
[19] Ibid, p. 3
[20] Ibid, p. 8
[21] Harold Schock et al. “Thermoelectric Conversion of Waste Heat to Electricity in an IC
Engine Powered Vehicle: Phase 1 Final Report,” pp. 4, 13
[22] Thinkexist, “Richard Buckminster Fuller Quotes,”
http://thinkexist.com/quotation/there_is_no_such_thing_as_a_failed_experiment/185928.html,
(accessed March 25, 2011)
[23] Andreas Panayi, “Thermoelectric Module – Finite Element Analysis,” pp. 5-10

*Represents internal page number of report. Article was accepted but not published at press time.

87