{Mm

.
.n... ,
,.

. mm
5 1.

.13 .n
.w
.a .w

A". .42:

2.22 .
3.11
. t...

.I... . i

anal 2......
a.
.e

.t .o
. .. A... .1.
fr....... .

 

... a ..
.. £333....
3&531nnfiufinnwuu
3.99%.“1stz. .
333.. an. i
:41. ..
. . 2.9. 331.: §
.1 . sag
ull33~¥finu Q44]..IC<- .
.95 .. «may and
t. {r .2
Arm... mm, . .3
.. .mwnhm {afflirufihu
4% 1.! turn )
r5!!!
«It;
ufivt hung”
.t. $.11
‘ . it . ‘

wan ‘ .14.
Vii-[:1 .-

El.

,gwaiflég gianefiu ,

afihfiifiw

amiémg.

 

This is to certify that the
thesis entitled

A NEW SYSTEM DEVELOPED TO CHARACTERIZE
THERMOELECTRIC DEVICES FOR POWER
GENERATION APPLICATIONS

presented by

JARROD L. SHORT

has been accepted towards fulfillment
of the requirements for the

Master of degree in Electrical and Computer
Science Engineering

L __.___’
rap——

 

Major Prdfeflor’s Signature
4~28—06

Date

MSU is an Affirmative Action/Equal Opportunity Institution

LIBRARIES
MICHIGAN STATE UNIVERSITY
EAST LANSING, MICH 48824-1048

n-.--—.-.-.-.-.—..-._._.-.

~-I-¢-0-.-.-O-I-O-0-0-l-o-l-l-o-o-~0-- - - a .

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05—c1/C-———IRC/DateDue.lndd-p. 15 ‘

A NEW SYSTEM DEVELOPED TO CHARACTERIZE THERMOELECTRIC
DEVICES FOR POWER GENERATION APPLICATIONS

By

Jarrod L. Short

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
Department of Electrical and Computer Engineering

2005

ABSTRACT

A NEW SYSTEM DEVELOPED TO CHARACTERIZE THERMOELECTRIC
DEVICES FOR POWER GENERATION APPLICATIONS

By

Jarrod L. Short

This document describes work accomplished in the Pulsed Laser Deposition and
Transport Characterization Laboratory of Michigan State University (MSU) where
a new measurement system has been developed for testing thermoelectric
power generation devices based on new materials. Previous work done at MSU
has focused on the search, synthesis, and characterization of new thermoelectric
materials. This work extends that effort with a focus on developing the materials
into working devices. Device fabrication can be as challenging and important as
finding more efficient thermoelectric materials; so, present and future work is
weighted more heavily on the fabrication and testing of thermoelectric devices
made from the previously synthesized materials, namely the Pb-Sb-Ag-Te based
“LAST” material [10,41]. Thermoelectric devices have their own unique
properties apart from the materials that make them, and the measurement
systems currently used at MSU to investigate the raw materials lack the abilities
to test fabricated devices. Therefore, a new measurement system was
developed at MSU to investigate the properties of thermoelectric devices used
for power generation applications. This thesis presents an introduction to
therrnoelectrics, as well as the process of developing the new device

measurement system, its abilities and limitations, and results from the system.

Copyright by
JARROD LEON SHORT
2005

ACKNOWLEDGEMENTS

For God’s graces, my families love, support and
motivation, advisor’s patience and guidance, Hogan
Hero’s, and friend’s confidence whom have all inspired
me to accomplish more than I ever thought possible~

thank you.

iv

TABLE OF CONTENTS

LIST OF TABLES ................................................................................ ix
LIST OF FIGURES ............................................................................... x
1 INTRODUCTION TO THERMOELECTRICS 1
1.1 How Thermoelectric Technology Works ................................................ 1
1.2 Thermoelectric Applications ................................................................... 4
1.3 Renewed Interest in Thermoelectrlcs, TFD’s ......................................... 6
2 FUNDAMENTALS OF THERMOELECTRIC TECHNOLOGY 8
2.1 Seebeck Effect .......................................................................................... 8
2.2 Peltler Effect ............................................................................................. 9
2.3 Thomson Effect ...................................................................................... 10
2.4 Thermocouples ...................................................................................... 1 1
2.5 Thermocouples and Thermoelectric Devices ...................................... 17
3 THERMOELECTRIC MATERIALS 19
3.1 Ideal Material Properties ........................................................................ 19
3.2 Engineering Thermoelectric Materials ................................................. 20
3.2.1 Improving Efficiency, Lowering Thermal ConductiVity ............... 21

3.2.1.1 Lattice Thermal Conductivity Reduction, Atom Substitution ............... 25

3.2.1.2 Lattice Thermal Conductivity Reduction, Grain Boundaries ............... 25

3.2.1.3 Lattice Thermal Conductivity Reduction using QSL’s ......................... 26

3.2.1.4 Lattice Thermal Conductivity Reduction, Bandgap Engineering ......... 28

3.3

4.1
4.2
4.3
4.4

5.1
5.2
5.3

3.2.2 Improving Efficiency, Increasing Electrical Conductivity ............ 28

Measurements of Bulk Materials .......................................................... 29
3.3.1 Electrical Conductivity Measurements of Bulk Materials ........... 30
3.3.2 Therrnopower Measurements on Bulk Materials ..................... 31
3.3.3 Thermal Conductivity Measurements of Bulk Materials ............ 33
THERMOELECTRIC DEVICES 34
Why Investigate thermoelectric devices at MSU ................................. 34
Thermoelectric Efficiency and Carnot Cycles ..................................... 35
Thermoelectric Device Efficiency ......................................................... 36
Measurements of Devices using Z-Meters ........................................... 40

A NEW THERMOELECTRIC DEVICE TESTING SYSTEM (DTS) 42

DTS Objectives ....................................................................................... 42
DTS Capabilities ..................................................................................... 45
DTS Hardware ......................................................................................... 46
5.3.1 DTS Chamber Bell-jar and Baseplate .................................. 47
5.3.2 Measurement and Supply Equipment ................................... 51
5.3.3 Power Out, Electrical Loads ............................................... 53
5.3.4 Device Cold Side Temperature Stage Design ........................ 55
5.3.5 DTS Cold Side Stage Electrical Connections, PCB .................. 58
5.3.6 Cold Plate Stage Assembly ................................................ 60
5.3.7 Cold Side Temperature Adjustment ..................................... 63
5.3.8 Temperature Gradient Heaters ................. . ......................... 66

vi

5.4
5.5

5.6

5.7

5.8

6.1
6.2
6.3
6.4

5.3.8.1 Temperature Gradient Heater, Large Devices .................................... 67

5.3.8.2 Temperature Gradient Heaters, Small Devices .................................. 72
DTS Hardware Costs .............................................................................. 74
DTS Software .......................................................................................... 74 ‘
DTS Device Measurements ................................................................... 76
5.6.1 Device Seebeck Measurements .......................................... 77
5.6.2 Device Resistance Measurements ....................................... 81
5.6.3 Device Figure of Merit Measurement Analysis ........................ 83
DTS Measurements of a Commercial Device ....................................... 87
DTS Measurements of LAST Devices ................................................... 97
5.8.1 Calculated Performance .................................................... 97
5.8.2 LAST Fabrication Methods and Results ................................ 99
5.8.2.1 LAST devices fabricated with silver paint ........................................... 99
5.8.2.2 LAST devices fabricated with silver filled epoxy ............................... 103
5.8.2.3 LAST devices fabricated with various solders .................................. 105
5.8.2.4 LAST devices fabricated with deposited contacts and solder ........... 107
5.8.2.5 LAST devices fabricated by diffusion bonding .................................. 108
USING AN IR CAMERA FOR DEVICE TESTING 110
Why Infrared ......................................................................................... 1 10
How Infrared Cameras Measure Temperature ................................... 111
The Infrared Camera used at MSU for Thermoelectrics .................... 114
Using the IR camera ............................................................................. 115
6.4.1 IR measurements on bulk materials .................................... 116
6.4.2 IR Measurements on a Commercial Device ......................... 118

vii

7 CONCLUSIONS 121
a FUTURE WORK 122
8.1 Autonomous operation ........................................................................ 122
8.2 Software ................................................................................................ 122
8.3 Analysis ................................................................................................ 123
8.4 Sample Stage ........................................................................................ 123
8.5 Larger devices ...................................................................................... 123
8.6 Hardware ............................................................................................... 123
APPENDIX A .................................................................................................... 125
APPENDIX B .................................................................................................... 126
APPENDIX C .................................................................................................... 128
APPENDIX D .................................................................................................... 129
APPENDIX E .................................................................................................... 132
APPENDIX F .................................................................................................... 134
APPENDIX G .................................................................................................... 136
APPENDIX H .................................................................................................... 140
REFERENCES .................................................................................. 143

6.4.3 IR Measurements on a LAST Device .................................. 119

viii

Table 1
Table 2
Table 3
Table 4
Table 5

Table 6

Table 7
Table 8
Table 9

Table 10

Table 11

LIST OF TABLES

Baseplate conflat style feedthrough ports. .................................... 48
Measurement instruments used. ................................................... 52
Device measurement and required instmmentation ...................... 52
DTS cost, approximate .................................................................. 74
Seebeck under pressure. .............................................................. 87

ZTAW, and efficiency for Tellurex device, T-cold = 75°C, T-hot

= 150°C. ....................................................................................... 97
Summary of LAST module results. ............................................... 99
Data for computing contact resistance from silver paint .............. 100
Room temperature thermoelement data for calculating ZT. ........ 101
TC vs. IR dT comparison, no vacuum (750 Torr) ........................ 116
dT comparison, 8x10'2 Torr vacuum. .......................................... 117

ix

LIST OF FIGURES

Figure 1 Thermoelectric device materials. .................................................... 2
Figure 2A-C Thermoelectric devices in three different operating modes. ........... 2
Figure 3 Thermoelectric couple as a majority carrier device. ........................ 3
Figure 4 Seebeck effect, for polarity shown, 03,, > (13 for T2 = T1 + AT. ........ 8
Figure 5 Peltier effect, 61,, > 613. ................................................................... 9
Figure 6 Thomson effect on single homogeneous material. ....................... 11
Figure 7 Single-ended thermocouple. ......................................................... 12
Figure 8 Thermocouple measurement. ....................................................... 13
Figure 9 Differential thermocouple. ............................................................. 15
Figure 10 dV/dT for type K, E, and S thermocouples. ................................... 16
Figure 11 Comparing type E, K, and S thermocouples. ................................ 17
Figure 12 Thermocouple based thermoelectric device. ................................ 18
Figure 13 Ideal thermoelectric materials. ...................................................... 21
Figure 14 Sintering affects on grain size, least sintering (left), most

(n'ght). ........................................................................................... 26
Figure 15 Electrical conductivity test. ............................................................ 30
Figure 16 Therrnopower and thermal conductivity measurements ................ 32
Figure 17 ZT comparisons with LAST. .......................................................... 34
Figure 18 Efficiency versus gradient and load. ............................................. 38
Figure 19 Efficiency with varied percent contact resistance .......................... 39
Figure 20 DTS System Concept. .................................................................. 44
Figure 21 DTS with IR camera. ..................................................................... 44

Figure 22
Figure 23
Figure 24
Figure 25
Figure 26
Figure 27
Figure 28
Figure 29
Figure 30
Figure 31
Figure 32
Figure 33
Figure 34
Figure 35
Figure 36
Figure 37
Figure 38
Figure 39
Figure 40
Figure 41
Figure 42
Figure 43

Figure 44

DTS open bell-jar. ......................................................................... 45
Bell-jar to Baseplate Self-aligning leg. .......................................... 48
Device testing system baseplate (Bottom view). ........................... 49
Device testing system pump-down time ........................................ 50
DTS load resistor .......................................................................... 54
DTS custom cold plate. ................................................................. 56
DTS large cold plate. .................................................................... 57
Custom PCB and small sample stage ........................................... 59
Sample stage with embedded temperature sensor. ...................... 61
Spring-loaded assembly. .............................................................. 62
Large Device Cold Plate Assembly. .............................................. 63
Cold plate stage with added heater. ............................................. 64
32-pin EM interference with power and data wires. ...................... 65
EM interference with cold side heater and data leads ................... 65
Early stage design under test. ...................................................... 67
Series heaters for uniform dT. ...................................................... 68
IR view of burnt aluminum housed style heater. ........................... 70
Parallel heaters in OFHC for large size devices. .......................... 71
TB-175 characteristics. ................................................................. 72
Small devices under test. .............................................................. 73
DTS Software flowchart. ............................................................... 75
Nonlinear effects of time delay for dVdT on Seebeck results ........ 78
Seebeck results, steady-state versus non-steady-state. ............... 78

xi

Figure 45
Figure 46
Figure 47
Figure 48
Figure 49
Figure 50
Figure 51
Figure 52
Figure 53
Figure 54
Figure 55

Figure 56

Figure 57
Figure 58
Figure 59
Figure 60
Figure 61
Figure 62
Figure 63
Figure 64
Figure 65

Figure 66

Type K thermocouple device. ....................................................... 80
Type K device Seebeck vs. Temperature. .................................... 81
Tellurex device resistance versus temperature (300-400°K). ....... 82
CZ1-1.0-127-1.27. ........................................................................ 87
Raw data collection, temperature and voltage versus time. .......... 89
Tellurex load error curve, T-cold = 50°C, T-hot = 100°C ............... 89
Tellurex load curve, T-cold = 50°C, T-hot = 100°C. ...................... 90
Tellurex load error curve, T-cold = 75°C, T-hot = 125°C ............... 90
Tellurex load curve, T-cold = 75°C, T-hot = 125°C. ...................... 91
Tellurex load error curve, T-cold = 75°C, T-hot = 150°C ............... 91
Tellurex load curve, T-cold = 75°C, T-hot = 150°C. ...................... 92

ZTAVe and efficiency for Tellurex device, T-cold = 323°K, T-hot

= 373°K. ....................................................................................... 93
Tellurex device dT changes, T-cold = 50°C, T-hot near 100°C ..... 94
Thomson coefficient for Tellurex device ........................................ 95
Thomson heat effects. .................................................................. 96
LAST module 3 day test .............................................................. 102
Soldered contacts ....................................................................... 108
Closest camera to device distance is 11cm. ............................... 115
LAST pellet (left), IR thermal image (right) .................................. 116
N16 dT profile. ............................................................................ 118
Tellurex device. ........................................................................... 118
LAST made device ...................................................................... 119

xii

Figure 67
Figure 68
Figure 69

Figure 70

Cold plate support plate. ............................................................. 126
Large stage assembly aluminum stage ....................................... 127
PCB wiring Diagram .................................................................... 128
Temperature sensor trial. ............................................................ 130

xiii

1 Introduction to Thermoelectrics

Fossil fuels are predicted to be nearly extinct by the late twenty-first century, so
present day consumer technologies are frequently designed to operate
electronically at low power. Electrical energy sources are more environmentally
friendly and are often created from abundant sources of sunlight, heat, water, or
air. Several technologies exist using electrical energy as an alternative to fossil
fuels, and thermoelectrics are one type that converts energy between forms of

heat and electricity.

1.1 How Thermoelectric Technology Works

Solar powered calculators, cars, houses, and satellites are just a handful of
examples demonstrating alternative energy sources born from semiconductor
technology. Solar powered devices work by using semiconductor materials to
convert light energy into electricity, while thermoelectric devices use
semiconductor materials to convert energy between heat flow and electricity.
Thermoelectric devices are typically made of two pieces of semiconductor
material connected electrically in series and thermally in parallel. A common
configuration of a device shown in figure 1 utilizes a metallic conductor at the
junction between N and P-typ'e semiconductor elements that are sandwiched

between two electrically insulating, thermal conducting plates.

Ceramic
Metal

Semiconductor
elements

Metal
Ceramic

 
 

NP

  

Figure 1: Thermoelecth device materials.

In figures 2A—C, electrical current flows into the bottom of one thermoelectric
material, through a top electrically conducting material, down through another
thermoelectric material, and back to the voltage source or sink. As shown in
figure 2A, if a DC source is applied to a thermoelectric device, Peltier heating and
cooling at the ohmic metal-semiconductor junction results in one side becoming
hot, and the other cold. If the direction of the applied current is reversed, the
heat flows in the opposite direction (figure 28). As true heat pumps,
thermoelectric devices are reversible and can also convert the flow of heat into
electricity. If a temperature gradient is applied to a thermoelectric material,

forcing heat to flow, then this energy is converted to electricity (figure 20).

A E 9
Heat a'rawn in Heat pushed out Heat alpplied

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

      

V-applied V-applied

Figure 2A-C: Thermoelectric devices in three different operating modes.

Thermoelectrics are majority carrier devices usually made of extrinsic
semiconductors where the free charge carrier density is adjusted to best convert
energy between forms of heat and electricity. The majority carrier is responsible
for the electrical and a portion of the thermal currents. If a temperature gradient
is applied across a device, a thermal current forms consisting of phonons and
free charge carriers. The charge carriers at the hot side have a higher thermal
energy than those at the cold side, so they preferentially diffuse to the cold side
of the device thus contributing to charge and thermal currents. Under open
circuit conditions, the carriers accumulate at the cold side forming a potential that
produces an electric field. This electric field continues to increase until an

equilibrium condition is reached and the net charge flow goes to zero [1].

To best utilize the majority carrier effects, thermoelectric devices are made of
repeated patterns of P and N type semiconductor materials placed electrically in
series and thermally in parallel. The simplest device is made from one piece of

each type of material, and is known as a thermoelectric couple (figure 3).

Applied Hea
iii iii
N P

 

 

 

 

 

 

 

 

 

1 ed

Figure 3: Thermoelectric couple as a majority carrier device.

The amount of heat that a device can pump is proportional to the size of the
device. For a fixed length of the N and P legs of the device, an increase in the
thermal cross-sectional area of the legs will cause the device to generate more
electricity, or pump larger quantities of heat. If the cross-sectional area is
increased by cascading couples electrically in series and thermally in parallel,
then there is an increase in the voltage across the device. If the cross-sectional
area is increased by increasing the dimensions of each of the two legs, there is
an increase in the current through the device. A thermoelectric device made
from a single couple would produce voltages too small for most present day
applications. In order to make a device that sources a practical voltage, the leg
cross-sectional areas are kept small and several couples are cascaded

electrically in series and thermally in parallel.

1.2 Thermoelectric Applications

There are many applications for thermoelectrics, from medical to automotive, and
consumer electronics to space exploration. Depending on the application, cost,
size, weight, efficiency, power out and reliability, all factor in to deciding how and
when to apply thermoelectrics as a means of cooling or for power generation.
Thermoelectric technology is not always an appropriate substitute for other
technologies, but many times they can be. Some examples are provided here,

and many others exist [4].

Chlorofiuorocarbon (CFC) chillers are often used for cooling purposes in industry,
and are often near twenty percent efficient [2]. However, chilling units can be
mechanically complex and require routine expensive maintenance. For example,
conventional chiller systems can be quite acoustically and electromagnetically
noisy and cause interference with other people and equipment. Because
thermoelectrics operate on DC and have no moving parts, they offer less
electromagnetic interference, are cleaner for the environment, and require less
routine maintenance. Most industry fabricated thermoelectrics for cooling are
presently limited to efficiencies of five to ten percent, and therefore are less
efficient than CFC chillers. Nevertheless, thermoelectric chillers are a popular
alternative to overcoming many of the problems encountered with conventional

CFC based chiller systems.

Within the auto industry, thermoelectric climate control is available [2]. Wasted
heat energy to brakes, engine, and exhaust systems are potential places
thermoelectrics could be utilized to harness waste heat and turn it into electricity.
With automobiles becoming equipped with more electronics and converting
traditional mechanical systems to electrical systems, such as in belt-less
engines, thermoelectrics may be quite convenient in power recovery systems.
Thermoelectrics are not presently suitable for use in the automotive industry as
power recovery systems because the efficiency does not justify the expense.
Should the efficiency improve and fabrication costs decrease, many more

applications are expected.

Many space exploration missions involving nuclear powered satellites are able to
use thermoelectrics to convert heat produced from radioactive energy into
electricity. Because of the long half life on several radioactive materials, images
and information can be acquired about our galaxy for several years or decades.
In fact, thermoelectric generators have been used for decades by NASA, either
using radioisotope or nuclear fission reactor fuels [3]. Solar power based
systems are often limited as the means to energy supply when missions are
distant from or in an indirect path with the sun, so thermoelectric based systems

are often a reliable alternative.

1.3 Renewed Interest in Thermoelectrics, TFD’s

Many other applications are available for thermoelectrics [4]. The need for
thermoelectrics is not new, but their low conversion efficiencies have kept them
out of many commercial markets and therefore out of the consumer eye.
Theoretical predictions and experimentally verified improvements in conversion
efficiency have spurred a renewed interest in recent years. Many new materials
and manufacturing processes have been discovered, however much of the
renewed interest in thermoelectric technology has grown from exploiting quantum
effects to improve conversion efficiency [21]. Such quantum effects have been
realized in thin film devices (TFD) and have shown efficiencies far superior to
traditional bulk materials. As charge carriers are confined, the allowed energies

for those carriers is reduced, and an increase in the density of states at the

remaining allowed energies results. There is an increasing sharpness of the
density of states at discrete energies as the charge carriers are increasingly

confined to two, one, or zero dimensions.

Due to rapid progress in TFD technologies, TFD research and development has
attracted many efforts in the thermoelectric community since its first appearance
in the late twentieth century. However, like traditional bulk grown materials,
TFD’s have their own set of new challenges to overcome whether used for power
generation or cooling applications. Such challenges include the needs to
improve the manufacturing cost and time, and increase the ability to have large
temperature gradients across them. The power produced from a thermoelectric
generator is directly proportional to the temperature gradient established, and

higher gradients are required to yield more power out.

Efforts from quantum based TFD research have redirected some bulk material
research to achieve properties similar to TFD’s. This redirected focus may lead
to bulk materials containing quantum structures that improve efficiencies without
the need for complex manufacturing processes used for TFD's. There are a
couple groups investigating the incorporation of nanodots into bulk materials,

including MSU.

2 Fundamentals of Thermoelectric Technology

2. 1 Seebeck Effect

The basics for how most thermoelectric technology works can be explained by
the fundamental relations contributed by scientists Seebeck, Peltier and
Thomson. In 1822, T. J. Seebeck discovered the first of three main
thermoelectric effects, which is described with reference to figure 4 [5].

B

 

 

 

 

 

 

Figure 4: Seebeck effect, for polarity shown, a A > a B for Tz= T, + AT.

Given two different materials, A and B, joined with theirjunctions maintained at
temperatures T1 and T2 , with a break at To, an open circuit voltage is created if
T1 ¢ T2. The details for the type of gradient between T1 and T2 do no reflect the

open circuit voltage, Vopen , for homogenous materials. The generated voltage is

generally linear for small temperature gradients in homogenous materials, and is
proportional to the temperature gradient through a temperature dependent

Seebeck coefficient, aAB (T') , as,

Vopen = “A3 (7"le '— TZ) (V) (1)

The “Seebeck coefficient” refers to a junction of two dissimilar materials where
each material contributes an absolute Seebeck coefficient “thermopower",

ax(T') , to yield the net Seebeck coefficient. More generally in terms of the

thermopower of material A, aA(T'), and material B, aB(T'):

T'=T2 T'=T2
V0,... = (aA(T')—aB(T'))dT'= jaAatT'W (V) (2)
T'=T1 T‘=T1

In addition to the Seebeck coefficient being temperature dependent, the type of
materials forming the junction will dictate whether the Seebeck coefficient is a

positive or negative value.

2.2 Peltier Effect

The Peltier effect was discovered in 1834 by J. C. A. Peltier and can be
demonstrated by passing a current through the junction of two different materials
[5]. Peltier observed that under these conditions, heat, P0, is absorbed or
liberated at the junction. For an applied current through two materials where

“A > (23, heat is liberated at the right junction and absorbed at the left.

 

Figure 5: Peltier effect, a A > a B .

Peltier discovered this effect to be reversible, in that if the direction of current is
reversed, so is the direction of heat exchange at the boundaries of materials A

and B. The Peltier heat, PQ, is related to the current, I , by the Peltier coefficient

as,

PQ = 2: MI (W) (3)

The Peltier coefficient relates to the Seebeck effect through Kelvin relations as,

7’ AB = TaAB (WK) (4)

2.3 Thomson Effect

The third and final thermoelectric effect discussed is the Thomson effect,
discovered in 1854 [5]. Essentially, when an electrical current passes through a

homogeneous material subjected to a temperature gradient, a Thomson heat is

generated, TQ. The direction of heat flow (absorbed or emitted) is related to the

type of material, the temperature dependence of the thermopower, and the
direction of current flow relative to the temperature gradient. For a metal wire
shown in figure 6, heat is liberated when the current flows toward regions of

colder temperature [6].

10

     

G

‘Appiied
Figure 6: Thomson effect on single homogeneous material.

The Thomson coefficient, 2', relates the Thomson heat and electrical current

through,
TQ = -2' [AT (W) (5)

where TQ > 0 implies heat is liberated, and can be readily related through Kelvin

relations with the Seebeck effect as,

da

 

r=Td—T (WK) (6)
0r
T'=T ,
a(T)= I “2]": )dT' (WK) (7)
T'=0
2-4 Thermocouples

With the fundamentals of thermoelectrics established, it is possible to better

understand how thermoelectric devices operate through a brief explanation of

11

how thermocouples work. A thermocouple is a device that produces a voltage
proportional to the difference in temperature along its length by use of the
Seebeck effect. There exist eight NIST standardized thermocouple types which
have been well characterized, but a thermocouple can essentially be made of
any two dissimilar electrical conductors [7]. To form a single-ended
thermocouple, the ends of each conductor are joined at a single point to ensure
good electrical and thermal connection between the conductors. Figure 7 will be

used with the following calculations in determining temperature, where T1 is the

temperature to be determined, VTC is the measured thermocouple voltage, and
To is a reference temperature for the thermocouple not equal to T1.

TO

+ A T1

V \

TC M
% 3 A

 

 

 

 

 

 

 

 

Figure 7: Single-ended thermocouple.
Using Kirchoff’s voltage law, it’s possible to sum up contributions from each
material’s generated thermoelectric voltage when a temperature gradient exists

across the materials from To to T1.

T'=T1 T'=To T'=T1 T'=T1
VTC= [aA(T')dT'+ [aB(T')dT'= jaA(T')dT'- [aB(T')dT'
T'=T0 T'=T1 T'=T0 T'=T0

12

T'=T1 T'=T1
= [loci (T') —aB(T')ldT'= laABU'W' (8)
T '=T0 T'=T0
The above voltage depends on the temperature, T0, of the meter measuring the
voltage. The Seebeck coefficient data, 0/130"): for standard thermocouples is

typically given relative to a reference junction maintained at 0°C; however

equation 9 allows a thermocouple to be referenced to any temperature [8].

T' = Tref T'= Tjunction
VMeasured = [aAB(T')dT'+ IaAB(T')dT' (9)
T'=OOC T'=OOC

Each of the integrals in equation 9 can be solved separately using a standard

thermocouple look up table or a curve fit to the standard data [7].

A common question is why there is no Seebeck voltage contribution due to

extension wires from the measurement instruments to reference junction.

Voltage
Measuring

Instrument T
E 0

C: + A T

r: :1 v *

1:1 :1 I TC />%
, / B

_Ta Extension wire (E)

 

 

 

 

 

 

 

 

 

Figure 8: Thermocouple measurement.

Previous calculations assume the extension wires are made of the same material
and that the temperature gradient along the wires from measurement instrument
to the reference junction are equal allowing Seebeck contributions to cancel.

Adding the measurement wires into the derivation yields,

13

T'=To T'=T1 T'=T0 T'=Ta
VTC= jaE(T')dT'+ [aA(T')dT'+ [aB(T')dT'+ [a5(r')dr'

T'=Ta r'=r0 T'=7'1 T'=T0
T'=To T'=T1 T'=T0 T'=T0

= [aE(T')dT'+ [aA(T')dT'+ ja3(T')dT'- jag(T')dT'
T'=Ta T'=To T'=T1 T'=Ta

T'=T1 T'=To T’=T1 T'=T1

= [aA(T')dT'+ [aB(T')dT'= IaA(T')dT'- [6130 ')dT '
T'=T0 T'=T1 T'=T0 T'=To

T'=T1

T'=T0

This is the same result as found previously without the extension wiring
contributions. If the thermocouple material changes from sample to instrument,
the temperature at the point of material change must be known. This point of
material change is often referred to as the reference junction, and requires use of

another temperature sensor such as a well calibrated diode or platinum resistor.

The single-ended thermocouples of figures 7 and 8 are used if an absolute
temperature is to be measured. If a temperature gradient is to be measured with
single-ended thermocouples, two are needed, and this results in four
measurement wires. However, a differential thermocouple can also be used to
obtain the temperature gradient, and only uses two wires. Figure 9 shows an

example of a differential thermocouple, made of materials A and B surrounded

by shaded isothermal regions T0, T1, and T2 where T0 at T1 ¢ T2.

14

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 9: Differential thermocouple.
Following the same procedures as for the single-ended thermocouple, the

temperature gradient can be obtained by integrating the Seebeck coefficient for

the thermocouple over the range of temperature defined.

T'=T1 T'=T2 T '=To
VTC = jaA(T')dT'+ jaB(T’)dT'+ jaA(T')dT'

T'=T0 T'=T1 T'=T2

T'=T1 T'=T2

= [aA(T')dT'+ jaB(T')dT'

T'=T2 T'=T1

T'=T1 T'=T1

= [aA(T')dT'— [aB(T')dT'

T'=T2 T'=T2

T'=Tl T'=Tl
= flaA(T')-aB(T')]dT'= I0A3(T')dT‘ (11)
T'=T2 T'=T2

For small temperature gradients, the last integration can be linearly approximated

. T +T
sothat llde-eOEaAB( l ZJEaAB(TAverage)5a/18(Tl)EaAB(T2)- Here,

 

0.413(7'1) and 0.1302) indicate aAB as a function of temperature. This

approximation yields a simple result for obtaining the temperature gradient that is

independent of the reference temperature.

15

 

T1 —T2 = VTC (12)
“AB (T A verage)

As with the single-ended thermocouple, a reference temperature must be

measured by another sensor in a region local to the gradient in order to identify

aAB (TA verage) -

For measuring large temperature gradients with a differential thermocouple, if
either the average, hot or cold side temperatures is known, the other
temperatures may be found from a fitting technique. For example, figure 10
shows how it may be possible to use the type K thermocouple with a linear

approximation, but a type E thermocouple may require a polynomial fit [9].

 

0.12 VV'VV'U'VVTVIVV'VIr'VrrYV‘U‘"VU'I‘V'Ir‘j

o V/dT, Type K
0.1 ; ° dV/dT, Type E
x dVIdT, Type S

 

IAPIJL

 

l

0.08

.......oooeeeeeeeeeo....
o

0.06

dVIdT (mVIK)

0-04 l-OOOOOOOOoooOoOOOOOOOOOO00000000000000 _

I

0.02 L .‘
i .

xxXXXXXXXXXXXXXXXXXXXXX
X

 

 

I
lllllllllllllllllljjllllJlllllllllllll

300 400 500 600 700 800 900 1000
Temperature (K)

 

Figure 10: dVIdT for type K, E, and S thermocouples.

Once the range of temperature measurements is known, it is possible to choose
one thermocouple over another depending on the temperatures measured,

sensitivity of the thermocouple, cost, and susceptibility to oxidation. For

16

example, figure 11 shows how it may be useful to use a type E thermocouple

over a type K or S for its increased sensitivity.

 

 

 

 

 

 

so-
'OTeK .0 3
70: VP 0 _,
g :OTypeE .3’ :
: o '
:60: xTypeS o... -I
at : :
Q 502- -I
s = =
g 40.” 1
5 : I
o L .:
.2 30: 1
220;- a
o i :
4:10.” 1
l- : 3
0:- 'i
llLlILLJlLJLlJ‘JIIJJJ+L

 

 

 

200 400 600 800 1000 1200 1400
Temperature (K)

Figure 11: Comparing type E, K, and S thermocouples.

2.5 Thermocouples and Thermoelectric Devices

A single couple thermoelectric device functions as a thermocouple, such that if a
temperature gradient is applied across its two dissimilar materials, a Seebeck
voltage is generated that is proportional to the gradient. Consider the following
example of a thermoelectric device made from several cascaded thermocouples
thermally in parallel and electrically in series, also known as a “thermopile”, in
figure 12, where materials A and B alternate. The shaded regions are

isothermal.

l7

”AV/[WI]! T1

IIIIIII

fiWM/IMJQ

       
 

 

Voltage
Measuring—-
Instmrnent

 

Figure 12: Thermocouple based therrnoelectrlc device.

The thermocouples are identical and each produce equivalent voltages, and the

extension wires between T0 and instrument have canceling Seebeck voltages.

Therefore, the measured voltage is just one thermocouple voltage repeated four
times.

T'=T1

VMeasured = (4) [“113 (PVT. (13)
T'=T0

So for a thermoelectric device of X-couples, the voltage produced scales
proportionally with the Seebeck coefficient X-times. The electrical power out
increases for the device with increased number of couples, but so must the
power in to establish the same temperature gradient across them. Therefore, the
efficiency as the ratio of power out to power in remains constant for devices
made of a single couple, as well as for X-couples. Thermocouples are made of
metals which have low thermoelectric efficiencies. Therefore, other materials are
used which have much better thermoelectric performance. Reasons for why

some materials have better performance than others are discussed further.

18

3 Thermoelectric Materials

3.1 Ideal Material Properties
The efficiency of a thermoelectric material is proportional to a property known as
the figure of merit, 2. For power generation applications, the efficiency ideally

relates to the figure of merit as follows [10].

 

[THot "'TCoId]

 

 

T H
”ideal = 0’ ’ (14)
[2__1_[THot ’TCold]+_4_]
2 THot ZTHot

The figure of merit is further related to the material properties: Seebeck

voltage, a(T) , electrical conductivity, 0(T), and thermal conductivity, x(T) , which

are all functions of temperature (T) [11].

= amamz
rm

Z (15)

Since the material properties change as a function of temperature, so must the
efficiency. Each known thermoelectric material has a range of temperatures it
performs most efficiently in. Different materials are suited for different
temperature applications, and to date there is no single material suited for all

applications.

The figure of merit shows that an ideal thermoelectric material should be a
thermally insulating electrical conductor. Unfortunately for thermoelectrics,
nature’s elements are found as conductors, non-conductors or semiconductors

for which electrical and thermal properties are directly coupled. Hence, good

19

electrical conductors are poor thermal insulators and good thermal insulators are
poor electrical conductors. This limitation makes thermoelectric materials
unique, so nature’s materials need to be engineered to possess the ideal

thermoelectric properties.

3.2 Engineering Thermoelectric Materials

Engineering thermoelectric materials for increased efficiency often requires a
lowering of the thermal conductivity or increasing the electrical conductivity. One
approach to developing a good thermoelectric material involves starting with a
good electrical conductor, such as a metal, and engineering the material to
decrease its thermal conductivity. A second approach is to start with a low
thermally conductive material, such as an oxide, and increase its electrical
conductivity. However, analytical and experimental research shows that the
most efficient thermoelectric materials occur as semiconductors and semimetals
[12,13]. As shown in figure 13, semiconductors and semimetals tend to yield the
highest figure of merit. One important characteristic of materials with high

thermoelectric figures of merit is that they share a common free charge carrier

density on the order of 1018 - 1021 cm'3 [6].

20

 

 

 

\ Semi- Semi- I
. Inswators‘ \Conductors Metals Metals i,
:5 E ‘5‘ '
15 .9 3%, I
a g a ’
g (D E r/ ’.
O 0 0 'I I
_ (i 0 l I"
(D a ’ ’
g- 8 E / ,- ’ bi
13 3 5 "
.93 m J:
LIJ U) I—
I I I
I I i

 

 

 

 

 

 

 

LogICarn’er Density)

Figure 13: Ideal thermoelectric materials.

As shown in figure 13, increasing the thermoelectric power causes a reduction in
electrical conductivity. An increase in electrical conductivity leads to an increase
in the electronic contribution in thermal conductivity. It is a challenge to find
efficient thermoelectric materials because the parameters are dependent on the
free charge carrier density, and while one parameter is optimized with a low
carrier density, another requires a high density. There are several ways to
enhance material properties to satisfy one parameter while having little or no
affect on the other, and some of these techniques are described further. Such
techniques include new approaches to material synthesis, as well as a variety of

fabrication methods.

3.2.1 Improving Efficiency, Lowering Thermal Conductivity

Since the efficiency of thermoelectric materials requires a low thermal

conductivity, an understanding of how to achieve this is discussed. Thermal

21

conductivity is the ability for a material to conduct heat. Both the phonons and

the charge carriers can conduct heat through the material, therefore

K Total = K Electronic + ’C Lattice (16)
At very low temperatures, the atoms vibrate very slowly, so the lattice thermal
conductivity can be very small or negligible compared to the electronic
contribution. At sufficiently high temperatures the lattice vibrates much more, so
the lattice thermal conductivity becomes dominant. Between these extremes,
both mechanisms contribute to the thermal conductivity of a material. When
engineering a thermoelectric material for low thermal conductivity, it is often
possible to reduce one component of the thermal conductivity without

appreciably affecting the other material parameters.

When manipulating characteristics of thermoelectric materials to reduce the
thermal conductivity, it is useful to recall the Wiedemann-Franz Law. This law is
usually applied to metals, and states that the electrical conductivity and thermal
conductivity are often proportional to temperature through a constant [57].
Hence, if the materials electrical conductivity is increased, the electronic thermal
conductivity will increase proportionally with temperature. This constant of

proportionality applies for temperatures above the Debye temperature, 60, and

is generally referred to as the Lorentz number, L, for the material [14].

2
L=£=(s+2)[5é] TforT>dD (17)
0' e

22

Here, 3, kg and q are the scattering parameter, Boltzmann constant as

k3 = 8.616x10‘5 eV/K, and electron charge as q =1.6x10‘19C respectively. The

scattering parameter indicates the type of dominant scattering: s = O for pure
phonon, s =1for mixed and s = 2 for ionized impurity scattering [15]. For metals,

the Lorentz number is 2.45x10‘8 W-Q/deg2 [57], or it can be calculated directly as

shown below [16].

L=[k—B]2 [(S+3))[Fs+2]_ (5+2)2Fs+12 (18)
q (3+1) Fs (s+1)2F_,2

Th e Fermi integrals, F,, are found with respect to the reduced chemical potential,

 

 

a: , and referenced from the conduction band edge as,

w .
x!

F.- = [We (19)

The reduced chemical potential is found directly from the thermopower, and for

an N-type material is given by,

_ _q_ (3+2)Fs+1
4’[a/rgiri(.s+1) F, i] (20)

For the LAST materials at some arbitrary temperature, 400K, with mixed

 

-3 W0

scattering, the Lorentz number can be calculated as 2.28x10 ‘7
K

[17]. From

this . the lattice and electronic contributions may be decoupled to give the

electronic contribution from Wiedemann-Franz as K'Electmnic = LaT. Using

literature data for the LAST material with a measured overall thermal conductivity

23

of 2.1—VIE, this results in xgkamnic = 0.684—W— and

m mK

KLam'ce = me, — Kmemonic = 2.1— 0.684 = 1.416%. This result agrees well with

the optimal figure of merit requirement on thermal conductivity where higher

values of ZT have been observed when M ~ 0.5 [27].
K Lattice

The thermal conductivity can be adjusted by limiting or altering the dominant
scattering mechanism, but the type of scattering must be known. Most metals
and semimetals have an increasing lattice contribution and decreasing electronic
contribution with increasing temperature; so for high temperatures, the goal is
usually to reduce the dominant lattice contribution. Many times in thermoelectric
material engineering, the goal is to reduce either the thermal or electronic
contribution without inversely affecting the other. Due to the Wiedemann-Franz
Law, it is not always best to reduce the electronic component of the thermal
conductivity for high charge carrier density materials since it will adversely affect
the electrical conductivity and have little or no affect on the figure of merit.
Therefore, much effort is placed on reducing the thermal conductivity of a
material through the lattice component, and this can be achieved through tuning

of the scattering mechanisms.

24

3.2.1.1 Lattice Thermal Conductivity Reduction, Atom Substitution

One way to reduce the lattice thermal conductivity involves substitution of one of
atom with another. The mass fluctuation increases the scattering of phonons

and thus decreases the lattice thermal conductivity. For example, Skutterudites

are minerals of (CO,NI)AS3-X composition that exhibit good thermoelectric

properties [18]. They have high thermopower and electrical conductivity, but also
have a high thermal conductivity. Therefore the lattice thermal conductivity of
Skutterudites is commonly reduced by substituting particular lattice site atoms,

adding atoms into the unit cell voids, or both.

3.2.1.2 Lattice Thermal Conductivity Reduction, Grain Boundaries

A second way to reduce the lattice thermal conductivity involves phonon
scattering at grain boundaries. Phonon scattering is increased at the grain
boundaries, thus decreasing the lattice thermal conductivity. However, grain
boundaries also scatter charge carriers causing a lowering of the electrical
conductivity. This trade-off effect in the conductivities can sometimes be partially
decoupled through different fabrication methods with cold or hot pressing [19].
Such fabrication methods involve grinding the raw elements down to a particular
particle size, and then pressing the ground material into a desired shape. The
material now has an increase in grain boundaries, where a grain of the material
is a region of continuous crystal growth without any defects. Perfect crystalline

materials would have no grain boundaries, and would allow an ideal path for

25

charge carriers, as well as phonons. Polycrystalline materials are a collection of
crystalline regions that are touching, but discontinuous; and discontinuous
implies that the atomic lattice contains one or more defects. The material grain
size can be adjusted to scatter the phonons but promote charge carrier travel.

By sintering the pressed material, grains begin to grow together. The growth can
be controlled by the initial particle size, pressing force, and sintering conditions of

time, temperature, and atmosphere [20].

me ,x
9 OW Oi
$788? F

Figure 14: Sintering affects on grain size., least sintering (left), most (right).

 

 

 

 

 

 

 

 

 

 

 

3.2.1.3 Lattice Thermal Conductivity Reduction using QSL’s

A third way to achieve reduced lattice thermal conductivity is through phonon
scattering with least affect on the charge carriers. Calculations have been
predicted and experimentally verified to show this is possible through reduction of
the dimensionality of the material to that of 2-D, 1-D, O-D quantum confined
structures [21]. 2-D quantum superlattice structures (QSL) can be made in a
variety of ways [22,23], and one popular approach is through molecular beam
epitaxial growth (MBE). MBE is a process by which layers of atoms are grown in
precise layers, and this allows tuning of the thermoelectric properties through
quantum confinement. This tuning allows for higher in-plane charge carrier

mobilities leading to reduced scattering effects, and thus a reduced lattice

26

thermal conductivity [24]. The high mobility desired in thermoelectric materials is
across plane, but mobility is taken as inversely proportional to a weighted

average of the scattering effects [25]. The mobility and thermal conductivity can

 

be indirectly related to ZT through the B-factor, ZT ~ B where B ~ 'u [21]. A
Klattice

second approach to reducing the dimensionality involves growing 1-D nanowires
that are smaller in diameter than the mean phonon wavelengths. Such
nanowires inhibit phonon travel while promoting charge carrier movement,
leading to low thermal conductivities with high electrical conductivity, and higher
figure of merits. For example, superlattice (SL) P-type BizTeg based materials
have shown substantial improvements in the figure of merit, ZT = 2.4, compared
to the respective bulk value of 1.14 [26]. Although QSL devices have potential to
offer higher efficiencies than bulk materials, the technology is relatively new
compared to bulk made devices, so they bring forth many new challenges and
opportunities for future research. One such challenge is that QSL devices are
usually so thin that the overall thermal conductance of the device can be large,
so it is difficult to establish a high temperature gradient. QSL also faces
challenges of reducing fabrication costs and time, as they are presently more
involved processes than required for bulk material processing. QSL technology
opens many new doors of exciting research for thermoelectric applications, and
such obstacles must be addressed if they are ever to dominate bulk material

fabricated devices in the marketplace.

27

3.2.1.4 Lattice Thermal Conductivity Reduction, Bandgap Engineering

A fourth way to reduce the lattice thermal conductivity is to synthesize materials
with small band gaps, since small band gap materials usually have low lattice
thermal conductivities [27]. However, thermoelectrics are majority carrier
devices, and materials with small band gaps may have problems of keeping a
high thermopower before minority carrier generation occurs. A relationship of
optimal band gap to maximum thermopower, which occurs at some maximum
temperature, can be approximated by the following [28].

Egap =2quaxamax (21)

Here, q, is the electron charge of 1.6x10'19 C, Egg], is the band gap, amax is the
maximum thermopower, and Tmax is the temperature at which the thermopower

peaks before minority carriers play a significant role.

3.2.2 Improving Efficiency, Increasing Electrical Conductivity

Good thermoelectric materials should have a high electrical conductivity. It is
possible to increase the conductivity of semiconductors with impurities, but the
choice and concentration of impurities is critical for good material properties. The
reasons for this are similar to discussion earlier for ideal thermoelectric materials,
and the relationship can be derived further with respect to the free charge carrier

density. The electrical conductivity, considering a single carrier, gives

2k T
a=qy.n=§”—;[ —§—]%( (m"'i/2Fr/ (22)
71'

28

where [1,. is the mobility in the x direction, m; is the density of states effective

mass, k3 is Boltzmann’s constant (8.6‘l6x10'5 eV/K), T is temperature, it, is
Plank’s reduced constant, q is the electronic charge (1 .602x10'19 C), and F); is

the Fenni-Dirac function of order one half as defined in section 3.2.1. As
mentioned, the choice and concentration of impurities is critical for improving
thermoelectric performance. For example, the impurity should be a shallow
donor so as not to act as a mid-band hopping site for valence band carriers into
the conduction band, and vice-versa. As heat is provided to a semiconductor,
this energy can cause an electron to change energy levels. If the appropriate
amount of heat is provided for good thermoelectric performance, the charges will
scatter in the band, but not from band to band. Carriers generated from band to
band are undesirable because they generally oppose the net flow of electrical

current.

3.3 Measurements of 8qu Materials

Measurement of electrical and thermal conductivities, and Seebeck coefficient
obtained in the Pulsed Laser Deposition and Transport Characterization

Laboratory at MSU are discussed. Results of system measurement accuracy
have been verified and are readily available elsewhere, so only the processes

are briefly described here [17, 45, 50].

29

3.3.1 Electrical Conductivity Measurements of Bulk Materials

Electrical conductivity is taken using a four point probe technique [29]. This
technique is usually used when measuring highly conductive materials, such as
those used for thermoelectrics, since a two point probe would contribute
dominating lead and contact resistances. The general technique involves
applying current to a sample, and measuring the ohmic voltage produced. The
sample conductivity is found from the measured dimensions, voltage and known

current supplied.

_fl_ -1 -1
d—VA (Q m ) (23)

Here, 0', I, V, I, and A are the conductivity, current supplied, voltage measured,

length and cross-sectional area of the sample, respectively.

Although thermoelectric materials are good electrical conductors, the ends of the
sample under test are often nickel electroplated to ensure an even current

distribution throughout the sample.

Ni—plating _ fl
Sample R ..

~.____ |+

Figure 15: Electrical conductivity test.

30

Measurements made at low temperatures are often made using wires of 1-3 mil
copper that are adhered with indium solder or conductive paint. For high
temperature measurements, the technique is the same, but lead wires are often
made of 1 mil constantan, chromel, or platinum, and adhered with the silver
paint. Copper leads are rarely used for high temperature measurements due to
its reactivity and oxidation effects. The supplied current produces an ohmic
voltage drop across the material, and thermoelectric materials produce an
additional voltage due to the Seebeck effect. The Seebeck effect is due to DC
electron flow, and the ohmic voltage can be isolated by applying an AC current to
inhibit temperature gradients. However as commonly used is this lab, a Keithley
model 2400 sourcemeter and 2182 nanovoltmeter have been configured to
reverse the polarity of a DC current fast enough to show comparable results to

AC testing utilizing a lock-in amplifier [45].

3.3.2 Thermopower Measurements on Bulk Materials

The thermopower of a material is generally determined by measuring the
temperature gradient and voltage simultaneously by use of single ended
thermocouples.

(IV
a _ d_f (V/K) (24)

For low temperatures, 80-400K, type T thermocouples of 1 or 3 mil diameter wire
are used to measure the temperature gradient. Both single-ended and

differential thermocouples have been used, however single-ended styles are

31

more accurate for measuring thermopower of a sample since the location of
temperature gradient and electric potential are monitored in the exact same
location [46]. For high temperature applications, type E, K and S thermocouples
of 1 and 3 mil diameter wire are commonly used to measure the temperature

gradient.

Thennooouple
Figure 16: Thermopower and thermal conductivity measurements.

To establish the temperature gradient, 3 small surface mount resistor or platinum
sense resistor is prepared adhered to one side of the sample with a metallic paint

or solder, as shown in figure 16 above.

The Maldonado technique is commonly used for rapid measurement of
thermopower and thermal conductivity on bulk materials [30, 45, 50]. This
technique provides a pulsing heat as opposed to the traditional steady-state
approach and offers several advantages. First, this method does not require the
long waiting time to approach steady-state. Second, the thermopower is taken
as a linear fit of several measurements of dVIdT, not just a single steady-state

value.

32

3.3.3 Thermal Conductivity Measurements of Bulk Materials

The thermal conductivity is also obtained through the Maldonado technique,
simultaneously with thermopower. All that is required to obtain thermal
conductivity in addition to the thermopower is the measurement of the heater

power. Then, thermal conductivity is obtained as,

Kzii. (JV—J (25)
dT-A m-K

where x, P, l, A, and dT are thermal conductivity, power delivered to the resistive
heater, length of the sample along direction of heat flow, cross sectional area of
the sample, and temperature gradient respectively. This measurement relies on
all power generated in the heater to flow through the sample by means of
conduction. However, the supplied heat also travels by means of convection,
conduction to the measurement wiring and radiation to the cooler ambient

environment, so accurate measurements must account for these. To minimize

-6 .
the losses, the samples are sealed under 10 Torr vacuum for testing to
minimize convection losses, lead wires are kept long and small in cross sectional
area to reduce conductive losses, and the sample stage housing is electroplated

and polished to reduce radiation losses. Losses are not avoidable, but can be

reduced so their affect on the measurement is secondary or negligible [45].

33

4 Thermoelectric Devices

4. 1 Why investigate thermoelectric devices at MSU

New thermoelectric compounds are being developed at MSU for purposes of
power generation, such as the LAST material, and are exhibiting excellent
properties for temperature ranges of 500-700°K. In fact, the LAST material
presently out performs many popular materials for power generation in this

temperature region.

 

 

 
   
 
 
    

 

 

 

2 -°-TAGS-80 (GeTe based p-type)
' . +TAGS-85 (GeTe-based p-type)
+Skutterudite (CoSb—based n-type)
. -°-SiGe (p-type)
1.5 - -°-SiGe (n-type)
* ‘ +PbTe (alkali doped)
-*-LAST (PbTe-based n-type)
I -°-BLAST (PbTe-based p-type)
IG 1 - -
.
0.5 l .‘
i A J . A . J . . . l

 

 

 

o J - . l . . . I . . . 1 . .
200 400 600 800 1000 1200 1400
Temperature (K)

Figure 17: ZT comparisons with LAST.

The next steps for these materials involve developing them as a working device.
As presented by others in the thermoelectric community, challenges in device
fabrication and testing provide an entire new area in thermoelectrics. Initial
research was exploratory synthesis to find a promising material with high ZT.
The focus at MSU is now drawn toward improving fabrication methods and

expedited testing on large quantities of devices. In order to begin testing, it is

34

important to know how devices operate, what properties does the thermoelectric
community commonly use to characterize device performance, and how does

one measure these properties.

4.2 Thermoelectric Efficiency and Carnot Cycles

Thermoelectric power generators ideally operate as Carnot cycles, and Carnot
cycles are defined as reversible processes in which energy can be converted
between work and heat. In a Carnot cycle for which a heat engine operates
between a hot and cold reservoir, such as a thermoelectric device, some of the
heat energy is converted to work through a chemical process and the other is
rejected to the cold reservoir. This phenomenon comes from thermodynamic
experiments, and the conversion of energy between work and heat for a given
engine has never been shown to be 100%. As explained by Lord Kelvin, the
ratio of heat transfer between the hot and cold reservoirs is proportional to the
absolute temperatures of the reservoirs. Therefore, the conversion efficiency of
a Carnot process, 77C, is limited by the heat transfer or absolute temperatures [5].

77 AT
C 1':
T Hot

 

(26)

Thermoelectric generators operate ideally as Carnot engines, but further
limitations arise from adverse effects of irreversible Joule and Thomson heating
that prevents them from being fully reversible. Nevertheless, thermoelectric
generating processes are commonly considered to be Carnot limited since this

represents the upper limit to their efficiency when the adverse effects are

35

 

minimized. Thermoelectric device efficiencies are commonly referred to in
literature in terms of their ideal performance, or “Carnot” efficiency. A more
realistic value of thermoelectric device efficiency includes limitations from
material properties, and is often referred to as the “reduced efficiency". The

reduced efficiency is generally defined in terms of the Carnot efficiency.

4.3 Thermoelectric Device Efficiency

In addition to being Carnot limited, thermoelectric devices are also limited by the
temperature dependent properties of the materials that make them. For power
generation applications, efficiency is the ratio of power out to power in [31]. The
power out is defined as the electricity given to some fixed load due to some

power from an applied heat source.

2
P - I R
”Ideal = ElectricaIOut = Load (27)

QSuppIiedln K(T)AT + 01.10; (T)THotl - 'i'l 2RModuIe

The heat flow given by the denominator of the previous equation can be defined
separately in order to identify how the heat distributes throughout the device.

K(T)AT 9 Thermal conduction
aHO,(T)TH0,I -) Peltier effect

éIZR -) Joule heating

In describing heat flow through a thermoelectric device, it is usually convenient to

know how much heat goes in, and how much passes through due to thermal

36

conduction, K(T)AT , that was not converted to electricity, aH0,(T)TH0,l . When

the device converts the heat to electricity and electrical current flows, the
resistance causes Joule heating. For understanding the heat flow
mathematically, the device can be split into two sections, the part that the heat
flows in to, and the other half that the heat flows out of. Because the entire
device experiences current flow and therefore Joule heating, half of the Joule
heat can be considered to flow toward the hot side and half to the cold side. In
this regard, the half Joule heating that flows to the hot side can be thought of as

contributing to the total heat sourced into the device.

The efficiency can also be maximized by adjusting the thermoelement lengths
and cross sectional areas. When this is done, the efficiency can be written as a
function of the figure of merit and Carnot efficiency, where the Peltier and thermal

conduction terms are disguised in the figure of merit, ZTH0,.

ML]
_ P Electrical Out __ RModule
”Ideal — Q . _ 77C 2

Supplied In [ RLoad +1]

RModule +[—Rlfli-+l]—( AT J
Z T Hot RModuIe 2THot

Thus, the device efficiency varies as a function of temperature gradient and load.

 

 

(28)

 

 

 

37

 

Efficiency (%)

 

 

T-cold = sojc (323°K)§
o 5 1o 15 20
Load ((2)

 

Figure 18: Efficiency versus gradient and lead.
A more realistic device will include electrical contact resistances due to interfaces
between the device materials and metal interconnects. This efficiency including

contact resistance, 77%,, can be found as,

 

 

( RLoad J
PEI ' 1014 R
77wcr : Qectrtea t = 77C (F 02 Module AT 7 (29)
Supplied In + + (1., +1)_[1+ 2[ RChs ]J[ J
ZTHot RModule 2THot
where R0,, is the contact resistance of the hot side of the device and
F = RLoad + RContact (30)

 

RModule

Figure 19 shows an example of how contact resistance affects efficiency when

the temperature gradient is held constant across some fictitious material.

38

 

-°-ldeal Efficiency (96)
--10% contact resistance
-°-20°/o contact resistan

 

3 ' ' ' ' r ' ' ' ' r ' --30% contact resistance
——40% contact resistance
25 -—50% contact resistan

-°—60% contact resistance
+70% contact resistance
-'-80% contact resistance
-—90% contact resistance

    

,.A‘
Q
......
‘
.‘
Q

 

 

 

 

 

 

 

Efficiency (7.) with varied
percent contact resistance

 

 

 

 

Load (a)

Figure 19: Efficiency with varied percent contact resistance.

The equations given previously apply for devices under small gradients, < 10°K.

For devices operated under load conditions and larger temperature gradients, the

total heat flow1 is more accurately described with consideration of the Thomson

heat, QThomson = @TATI .
1
QLoad = I"'(TlAT + aHot(T)THotI “EIZRModule - GTATI (31)

where O is the ratio of Thomson heat that flows toward the cold side of the
device [32]. This ratio ranges from 0 to 1 and depends on the homogeneity of

the materials used and temperature difference applied, but is typically treated as

 

I Note, the notation used to describe heat flow through a device when operated under load conditions
changes from QSuppIied In '9 QLoad -

39

Joule heating where the heat is distributed evenly along the length of the material

and half contributes to each of the hot and cold sides. r is the Thomson

coefficient, 2' = T Ave (Ci—g. Rewriting for large temperature gradients,

1
QLoad = K(T)AT + aEhot(T)THot1 "EIZRModule (32)

where aEh0,(T) is the Seebeck and Thomson effect combined as,

 

 

rAT
aEhor (T) = aHot (T) - —— (33)

2THot

In similar form, the figure of merit changes to,
2
MT
0(T)[au 1(T) - J

Z = «Danae? = ° 2THot (34)

K(T) ’((T)

and the efficiency is as stated previously for mm, but with Thomson effects

disguised in the figure of merit.

4.4 Measurements of Devices using Z-Meters

As shown previously, thermoelectric device efficiency depends on the figure of
merit, so a device's figure of merit is of primary interest. Measurement
techniques and instruments have been designed and improved by the
thermoelectric community to efficiently and accurately measure a device's figure
of merit. A “Z-Meter” is an instrument that determines the figure of merit for a

thermoelectric material or device, and several types exist [33]. One type

40

 

simultaneously measures electrical conductivity, thermal conductivity and
Seebeck coefficient as a function of temperature, and determines the figure of
merit by calculation of such measured parameters. A second type, designed by
Harman, calculates the figure of merit from a ratio of two voltages [34]. Harman’s

method applies an AC current to a sample which causes an ohmic voltage, V ,
and then applies a DC current to measure the steady-state additive ohmic and
Seebeck voltages. The Seebeck voltage, Va, is extracted and used with the
ohmic voltage to calculate the figure of merit, ZT.

ZT_£/_a;:VDC—VAC =VDC _1 (35)

VP VAC VAC

 

or more commonly written as a ratio of measured resistances,

ZT = 529—1 (36)
RAC

Harman’s method has shown to give good results in comparison with the first
type of Z-Meter when near adiabatic conditions are achieved. This method has

shown promising results for bulk material and device measurements [35,36].

This thesis focuses on developing a new measurement system for characterizing
devices made for power generation applications, so particular attention is given
to systems previously designed for high temperature measurements. One
particular type of Z-Meter was established to test bulk PbTe samples up to 973°K
[33]. This Z-meter design along with several others from previous device
measurement systems have lead to the development of a new system that can

test devices under large gradients in addition to the single pellets and couples

41

that Z-Meters are commonly used for. The new system offers many features that

enable it to be more flexible than prior designs built by others [37.38.39.40].

5 A New Thermoelectric Device Testing System (DTS)

5. 1 DTS Objectives

Due to recent progress in discovering the LAST material, a new measurement
system has been developed for use in the Pulsed Laser Deposition and
Transport Characterization Laboratory at MSU to characterize thermoelectric
devices used for power generation applications. A new system was needed
because previous measurement systems in the lab were used for bulk material
measurements and lacked the ability to fully characterize a device. The method
chosen for characterizing such devices determined the objectives of system

design.

The primary objective of this thesis was to build a system to test the efficiency of
devices made from new thermoelectric materials synthesized at MSU, primarily
from the LAST material [41]. Such testing must allow for optimized LAST
performance, which requires 900°K to be established on the device hot side.
Device characterization should also include to ability to test under varied

temperature gradients and load conditions.

A secondary objective was to use an infrared (IR) camera to obtain the

temperature profile across a device. The profile of a device is important because

42

it is useful for understanding if the heat distribution is linear, as most simplified 1-
dimensional thermoelectric modeling and analysis assumes. The profile results
may eventually lead to a more accurate thermoelectric model for simulating
device operation. As with developments in Z-Meter designs, the IR camera was
expected to provide an opportunity to test devices more efficiently through non-

contact measurements.

A third objective was to design a system to be large enough to test devices of
various sizes and types (single layer or stacked). Furthermore, the system
chamber was to be large enough and adaptable to become a multi-testing

system. In fact, two independent tests have been run at the same time in this

1 . . .
system , and there exrsts much more room, electrical connections and unused

ports to allow the system to adapt for even more testing.

With the above objectives set forth early on in the design process, a sketch was

given, in the late summer of 2003, to indicate what the system might look like...

 

l Steady-state and AC methods on different modules.

43

 

1=Chlller

2=H015t

3=Vacuum Pump Unit
4=Bell Jar

5=IR camera
6=|nstrument Panel
7=IR Camera Computer

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4 8=System Computer
6
/
., , Illa
a. I
1
2 3

 

 

 

 

 

 

 

 

Figure 20: DTS System Concept.

A new system was progressing with part ordering, developing schematics,
assembling, and testing. A year later, the system was ready to begin testing

(figures 21 and 22).

 

Figure 21: DTS with IR camera1.

 

1 IR camera from Indigo Systems, division of Flear International, model mid-infrared range Merlin

44

 

Figure 22: DTS open bell-jar.

5.2 DTS Capabilities

The primary objective for the system was to test thermoelectric devices made
from the LAST material for future implementation as a 1kW thermoelectric
generator (TEG). Several unicouples fabricated with the LAST material have
been tested, some with hot side up to 900°K under open circuit conditions. A
power generation device from Tellurex (model CZ1-1.0-127-1.27HT) has been
tested under varying load and gradient conditions up to 448°K. Depending on
device size and load conditions, the cold side temperature has been varied from
290-400°K. Larger size devices can also be tested, but the sample stage may
need to be reconfigured for increased cooling capacity and supplied heater
power. It may be useful to test larger size devices to identify how the efficiency

of a 3-dimensional device scales according to typical 1-dimensional calculations.

45

A second objective for this system was to use an IR camera to obtain the
temperature profile across a thermoelectric device. Several devices have been
viewed with the camera and the temperature profile was obtained. Results are

discussed further in section 6.

A third objective was to be able to build a system that could test devices of
various sizes, and to be a multi-testing system. The system has tested
unicouples and commercial devices. Due to the large chamber size, the system
has tested two devices at one time, one with a steady-state technique and the
other with an AC method [42]. The second measurements performed using the
AC method were taken using two separate lock-in amplifiers and a different
computer. In addition to testing two modules at once, the system is configurable
for testing individual bulk material samples as well. A separate sample stage
could be designed for use in this system to test materials at higher temperatures,

and this is left for future work.

5.3 DTS Hardware

The design of the device testing system is made up of several main components:
vacuum chamber, data acquisition equipment, cold plate, chiller, cold and hot
side heaters, temperature sensors, sample stage, cart, vacuum system, hoist,
infrared camera, and 400 MHz PC. The choice of components and their

implementation into the system design is discussed in further detail.

46

5.3.1 DTS Chamber Bell-jar and Baseplate

The system chamber is made of an 18 inch ID, 0.25 inch thick stainless steel
bell-jar and 22 inch OD baseplate, both custom made and assembled onto a
custom rack by the Kurt J. Lesker company (KJL) in Pennsylvania. The rack has
(2) 22 inch instrument bays and houses everything except the computer, infrared
camera, rough pump and chiller. The large bell-jar has a #8 polished interior and
brushed exterior finishes, is 30 inches high, and was selected to allow devices of
various size arrays and heights to be measured. The bell-jar has two 6 inch
conflat ports, one contains a 6 inch (2 inch viewable) sapphire window for
infrared imaging, and the other is unused. The sapphire window was chosen
specifically for mid-range IR transmission, the range defined by an IR camera
purchased for co-use in this research. Sapphire transmits greater than 80% of
wavelengths in the range of 1-5um, while a typical quartz window only transmits
a fraction of this‘. The camera was chosen to detect wavelengths in the mid-IR
range because the temperatures of the devices tested produce wavelengths in
the middle of the infrared spectrum, near 3-5um. The bell-jar self-aligns to the
baseplate with four plastic tipped steel angle iron legs to align itself when being

closed to the baseplate. The bell-jar seals to the baseplate'with a Buna type 0-

ring. The bell-jar is attached to a standard duty electric hoist2 for easy sample

loading and unloading.

 

1 Per information requested with the sapphire viewport purchased from the Kurt J. Lesker Company
2 Kurt J. Lesker Company Part number: HSTST

47

 

Figure 23: Bell-jar to Baseplate Self-aligning leg.

The custom baseplate has several various sized conflat style vacuum flange

ports, and uses are specified in table 1.

Table 1: Baseplate conflat style feedthrough ports.

feedthrough

solid copper rods

O.D. stainless steel tubes

vacuum gauge
gauge

vacuum pump

 

Figure 24 shows a sketch of the baseplate with all various sized feedthroughs.

48

R11.000"

(6) 2.75" OFF
6" from center
30 deg apart

  
 
  
 

  

6' ' CFF
(wntered) ,,
+- 4

o (4) #1024 —-o 45 F

 
 

 

     
    

4.5" OFF

(45 deg)
. i ©

1.50L1

 

 
 

(4) 3/8-16 “j

4) M“ —-o

    

Figure 24: Device testing system baseplate (Bottom view).

Conflat flanges were chosen to give the best vacuum for large chamber size in a
reasonable period of time. The system has a BOC Edwards dry scroll pump1 to

. -2 .
achieve sub 10 Torr vacuum pressures for this system chamber, and connects

to a EXT 255 series turbo molecular pump with a QF-25 flange and 48 inch long
braided hydro-form bellow. The scroll pump has an adjustable bypass switch
setting which is used to pump large humid chambers with less wear on the pump,
and then is switched to pump the chamber further at full speed. The turbo pump

is air cooled, has a feedback controller, connects directly to the chamber

 

1
BOC Edwards Part Number: XDS-C Dry Pump.

49

baseplate through a 6 inch conflat flange, and is directly attached onto the

baseplate’s centered 6 inch conflat flange to reduce vacuum conductive losses.
A dry system pumps to mid 10'6 Torr in just a couple of hours, and after many
uses of testing, still achieves this in just a few hours. Depending on sample
under test, the system pumps to upper 10'7 Torr range over a period of a few

days. The analysis used for determining device efficiency assumes zero
convection heat losses, so a testing environment providing UHV conditions

considerably simplifies the analysis for device measurements.

 

100 '- U I r I I I

START
10 PPUMP’NG turbo tum, ON '1

P1
1 - /@s I
0.1 -
0.01 - .

Vacuum (Torr)

0.001 -
CONTINUE

CONSTANT —> -
PUMPING *

I

0.0001

fi. 1
10‘ . l . l 1 sub 19’6 Torr

O 20 40 60 80 100 120 140

Time (minutes)

 

 

 

Figure 25: Device testing system pump-down time.

Vacuum pressures are monitored with a thermocouple gauge, BA hot filament

ion gauge, and standard gauge controller1. The controller has two DB-9
connecters, one for gauge feedback, and the other for computer control;

however, the system is currently not setup to computer control the vacuum

 

l KJL Vacuum controller part number: KJL4400

50

assembly. The controller also has a feature of taking an action at two user
specified set points (SP1 and SP2) of vacuum pressure. The EXT 255 Turbo is

recommended for turn ON at vacuum pressures of 7 Torr or lower, so SP1 is set

to 10'1 Torr to turn on the turbo pump1. SP2 is available to reduce the turbo

pump speed or power in half when vacuum pressures become lower than

desired, but is not presently used.

5.3.2 Measurement and Supply Equipment

In order to determine what measuring equipment was needed for this system, it
was first appropriate to decide what transport properties of devices and samples
would be measured. The desired measurements were determined from tests
performed in the MSU Pulsed Laser Deposition and Transport Characterization
Laboratory on thermoelectric compounds, as well as described in the previous
section of this paper [31, 43]. Raw data collection is fully automated with a PC
and National Instruments LabView ® software, version 7.0. This allows for
obtaining large amounts of data in relatively short periods of time. Several
Keithley brand instruments were chosen for their high accuracy and reliability,

and are conveniently interchangeable with other lab equipment. Furthermore,

many of the lab instruments are GPIB controlled by a computer2, and already
had several communication programs written in LabView® that were readily

adapted for use in this system. One of the power supplies on this system is older

 

I EXT Compound Molecular pump instruction manual, page 21
2 National Instruments PCI-GPIB card part number: 778032-01

51

and analog controllable only, so a digital to analog converter with USB interface
is used to control the power supply1.

Table 2: Measurement instruments used.

 

 

 

 

 

 

 

 

 

 

Quantity Mfg, model # Description

1 Keithley, 2400 Current source meter

4 Keithley, 2182 Nanovoitmeter

1 Keithley, 2002 Multi-meter

1 Sorenson, DCSSO-18E 1kW power supply

1 Sorenson, EMSll 250W power supply

1 Measurement Computing, PMD- Digital-to-Analog converter, USB
1208LS

1 National Instruments, GPIB card GPIB interface card

1 indigo Systems, Merlin infrared camera

1 Gateway, PC Personal computer

 

 

 

 

 

Table 3: Device measurement and required instrumentation.

 

 

 

 

 

Parameter Measurements Meters Required
Efficiency Power Out {V-OUT} 2182#1: measures V-OUT
Power lN {l-lN, V-IN} 2002: measures l-lN & V-IN
DCS#1: provides power lN
ZT dT-Open {V-TC} 2182#2 for differential: measures V-TC or
dT-closed {V-TC} 2182#2 & 3 for single-ended: measures V-
TC or
Indigo infrared Camera
Seebeck Voltage 2182#1 : measures V-OUT
dT 2182's: same as for ZT
Resistance Voltage 2400front: supplies l-IN
Current 2182#1: measures V-OUT

 

 

 

 

 

 

1 Measurement Computing part number: PMD-1208LS

52

The thermal conductance of the device is a calculated parameter found from the
gradient and power supplied, or can be determined through ZT and device

resistance for some average temperature. All leads from instrument to baseplate

feedthrough1 mate with a 32-pin military style connectorz.

5.3.3 Power Out, Electrical Loads

The baseplate contains a 2.75 inch power feedthrough3 which has four copper
rods, each rod of 0.25 inch diameter, see figure 26. This style feedthrough was
chosen to given the least resistance between device and load. Many of the
devices tested will be single couples, also known as unicouples, on order of
milliohms, so any contact resistance between device and load may greatly

influence the power out across the load, and thus efficiency calculations. The

rods have been silver plated4 for even less resistance and oxidation. The wiring
from sample stage to feedthrough copper rod is braided 12 AWG, soldered at
one end to a spade terminal and screwed onto the copper rods, and each screw
is vented for quicker vacuum. The connection from load to atmospheric side of
the feedthrough is made with a screwed on banana jack for convenient load
switching during testing. Each load then has a male banana, and the contact
resistance seen at the banana to banana is less than 10mn. This may be
significant for unicouples of small dimensions whose overall resistance may be

comparable.

 

Kurt J. Lesker 32-pin feedthrough part number: IFDRG327013
Newark 32-pin military connector part number: 93F9354

Kurt J. Lesker power feedthrough part number: EFT1243253
Cool—Amp silver plating part number: 1233-500 1/4

buts)—

53

Silver-
plated
Rods

Banana
Jack
Power
Resistor

 

Figure 26: DTS load resistor.

A resistor switch box was originally going to be fabricated for autonomous
operation in load switching, but one could not be readily located giving less than
a 10-20mQ switching contact resistance. This contact resistance may be
considerable when testing unicouples with 10-20mfl resistances, for which it is
assumed all power from module transfers to the load under load testing
conditions. There exist lC's with switching loads, but these are generally for RF
use and have resistances much higher than that of a thermoelectric device

whose resistances are generally several orders of magnitude smaller.

The load resistors are 25W ceramic, axial style power resistors1 with standard
aluminum housing. Power resistors are chosen because of their low resistances
with 1% tolerances. Also, since these are designed to dissipate a large amount
of power but are only typically dissipating milliwatts, their resistance does not
change much when flowing small currents through them at room temperature; so,

only the voltage needs to be measured. To confirm this, a few tests have been

 

] Huntington Electric Inc. axial style, 25W, power resistors, some standard order through Digi-key Corp.

54

performed monitoring current and voltage compared to just voltage, and both
agree. It should be noted that each load has been tested using a four-point
method, as well as the overall resistance of the wiring and connections to the
device, for more accurate efficiency calculations. The wiring resistance
contributes 65m!) of resistance to the load, and should be reduced for efficiency
testing of unicouples by increasing the lead wire diameter above the presently

used 12 AWG copper.

5.3.4 Device Cold Side Temperature Stage Design

The cold side of the device is cooled with a CFC based chiller1 of set point 20°C

with adjustable pressure, and thus flow rate. The chiller is supplied with a special

coolant to inhibit bacterial growth over time, and thus require less maintenancez.
The chiller induces some vibration along the supply line to the baseplate, and the
temperature can fluctuate due to tolerances of the chiller. Therefore, a vibration
and thermal oscillation damper was placed inline between chiller and chamber
connection. The damper was made by placing 25 feet of 0.375 inch OD coiled
copper tubing in a 5-gallon bucket submerged in 80le. of concrete. Although
this reduced the flow rate, it also reduced the vibration substantially and
decreased the temperature fluctuation from 05° to 0.1 ° with sufficient heat

pumping capacity for testing done so far to date.

 

1 Kurt J. Lesker Company chiller part number: ATCKTR6001
2 Kurt J. Lesker Company chiller coolant: Hexafluid

55

The chiller connects to a 2.75 inch conflat water feedthrough1 made with 0.25
inch OD stainless steel tubing atmospheric side, and 0.25 inch VCO fittings on
vacuum side. VCO style fittings were chosen to be sufficient for this UHV
system, so VCR fittings were not required. The first cold plate was custom made

by the MSU Physics Machine Shop, figure 27. it is made of 0.5 inch thick OFHC
and 0.25 inch OD copper tubing, and all is nickel platedz. The cold plate
connects with two VCO female fittings made of a nut and glands, which mate to
two 4.75 inch compressed, flexible, convoluted tubes with 0.25 inch male VCO

4
connectors .

TOP VIEW 1/ "VCOfigs Ag-soldered
Silversoldersealedatthisend to er of ooppertublng

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

O l 48 drill and a? J

i O I

2. 01' 1004 FJ 3. ' 3.. "
L 1:2--- 1/ "ooppertubing
2.00—-0 r welded tooopperblock
O ' I
Wm hole—7
(5/16" after nickel plating) l‘
.5
SIDE VIEW FRONT VIEW
0 d ill

 

 

 

Figure 27: DTS custom cold plate.

 

' Kurt J. Lesker Company water feedthrough part number: LFT322MCTE
2 Nickel plating done by Adams Plating, located in Lansing, MI
3 Kurt J. Lesker Company VCO nut and gland part numbers: SS-4-VCO-4, SS-4-VCO-3
4 HE. Lennon convoluted tubing part number: 321—4-X-6FMO

56

The cold plate sits atop four vented standoffs that fix to the baseplate, and made
to allow the cold plate to be raised and lowered. The height adjustment was
implemented for direct view line with the bell-jar sapphire viewport used for

infrared imaging with varied device heights.

A second cold plate was made to accommodate larger sized devices. This

second cold plate is made of 0.5 x 3.5 x 6 inch aluminum with 0.375 inch OD

stainless steel tubing1, and has two silver soldered VCO female fittings attached
by the MSU Physics Machine Shop. The cold plate was drilled for attaching the

sample mounting stage and for standoffs to a baseplate support plate, figure 28.

()2 VCO

(1/ " nut ()8: gland)“
1. 5"

3/8" SS tubing

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 28: DTS large cold plate.

 

I McMaster Carr cold plate part number: 35035106

57

Because it is not convenient to disconnect the cold plate each time to remove a
sample, a sample stage was designed that screws to the cold plate for easy
sample changing. The cold plate and sample stage were made thick to avoid
warping, and ground and polished when fabricated to ensure good thermal
contact. Although polished flat, a thin layer of Apeizon H thermal grease is used
between cold plate and stage to ensure a better thermal contact. The stage
consists of a 0.125 inch thick aluminum, and custom printed circuit board (PCB)
to attach electrical leads from device to instrument feedthrough. Aluminum was
chosen for the stage because of its high thermal conductivity, availability and

ease of machining.

5.3.5 DTS Cold Side Stage Electrical Connections, PCB

The single-sided PCB has 15 electrical traces made from 0.03125 inch thick
copper clad board. Two extra wide traces are also shown available for soldering
the device leads to the baseplate feedthrough wiring, to the load for reduced

contact resistance, figure 29.

58

Custom PCB with 30 pin
Edge Card Connector
(15 Pb/Sn tinned traces)

(T1 ~60L‘l Wde load traces
75{ Aluminum plate

underPCB
; — e—zoo" e

il

 

 

 

 

 

 

 

 

 

 

 

 

 

Same 2’ "2‘50"
placed
- q 1.00" G-
550" \\ J
(4) drill for #8-32
1] " inch from sides

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 29: Custom PCB and small sample stage.

The PCB was designed using Layout software from Cadence, and then

fabricated by the MSU ECE shop staff and T-tech prototyping machine. The

PCB is made for a 30-pin edge card connector1, but only half of the pins are
used for the single-sided PCB. Other PCB’s were made using ferric chloride
etchant and permanent marker. Also, photoresist boards were experimented
with and worked well, but the processing chemicals and developing time made
this an inconvenient choice as well compared to the ECE shop router. The PCB
was then GE varnished® to a 0.125 inch thick aluminum stock, ground and
polished, for good mechanical bond and thermal contact for the thermocouple
reference junctions. The aluminum plate was used to allow the device and PCB
to be one piece, as opposed to mounting the device under test directly onto the
cold plate. This design limits the cold side of device testing to the melting point

of the varnish and PCB, but most device testing is generally done at

 

l Sullins Electronic Corp. edge card connector part number: EZClSDRXH

59

temperatures below such melting points. However, another design could easily
be adapted for testing at higher temperatures, as may be required for bulk
material measurements. The final PCB stage assembly, shown in figure 29, gets
spring loaded with the device and screwed onto the small custom cold plate, and

is good for testing devices up to 1.5 x 1.5 inches.

5.3.6 Cold Plate Stage Assembly

The final assembled design is made of the PCB based stage assembly and cold
plate. The PCB aluminum stage has a Pt-103 sensor placed inside it to monitor
the cold side temperature and provide a thermocouple reference temperature.
The Pt-103 was initially placed next to the device under test, and adhered to the
stage with silver paint, however, it was later determined that best results were
obtained with the sensor placed inside the stage just below the device under test,
experiencing the same heat flow with the device. To achieve this, a section of
the aluminum was milled out, and only 1mm thick of aluminum remains between

the device under test and sensor tip. The sensor and its lead wires were placed

in the milled out region and then covered with an alumina epoxy1 to ensure good

heat flow from device to cold plate figure 30.

 

l Aremco ceramabond 668

60

Side view Top view
Aluminum 0 O
Alumina M l g
l

o 4mmO

 

 

 

 

 

‘_—1

 

 

 

 

 

 

Figure 30: Sample stage with embedded temperature sensor.

The local temperature of the device cold side is desired, and the sensor would -
ideally be placed in direct contact with the device; however, a 1mm region of

aluminum is in between the sensor and device. Due to its high thermal

conductance1, the temperature drop across the 1mm thick region of aluminum is

considered negligible compared to the gradient across the device under test.
The epoxy has a thermal conductivity rated 85% of alumina2 when prepared

properlya, near 17 W/m-K. The temperature drop across the alumina epoxy is
considered negligible since it is not directly between device and sensor.
Between the device cold side surface and temperature sensor is a thin layer of
Apeizon-H thermal grease to adhere the device to aluminum stage plate. The
thermal grease has a thermal conductivity near 0.2 Wlm'K (0.002W/cm-K), is
recommended for use to 240°C (513°K), does not melt but becomes harder as
temperature increases [44]. The grease has excellent stiction properties, cleans

easily with acetone, and sometimes is preferred over metallic paints which can

 

1 Aluminum thermal conductivity from Goodfellow, as 237 W/m-K (0-100°C)
2 Alumina thermal conductivity from Goodfellow, as 20-25 W/m-K (0-100°C)

3 As stated 85% of alumina by Aremco engineer through a phone conversation

61

lose their bond due to coefficient of thermal expansion differences. It is important
not to apply too much thermal grease or an undesirable temperature gradient
could occur between the cold side temperature sensor and device surface. For
example, a device of size 1.44 in2 (9.29 cmz) covered with the grease, applied of
just 100um (0.01cm) thickness, it is possible to have a appreciable temperature
drop,

10.0W

AT=QlK= 2 25.000°.
9.290 /
(0.002W/cm*K)( cm 00100")

The cold plate assembly is shown below in figure 31.

 

Wing-nut
6 ring

uminum plate
#8-32 Screw
Quartz Tube

OFHC Heater Block
Device Under Test

Circuit Board
Aluminum Stage plate

Cold Plate

 

Figure 31: Spring-loaded assembly.

This cold plate assembly design works well, but has much room for improvement.
One inconvenience in the design lies in the spring-loading of the device and
stage to cold plate. The same screws are used for spring loading the device,
gradient heater, and stage into place. During sample loading, the device,
gradient heater or other wiring is often bumped and needs repair before testing
begins. Therefore, another stage assembly was fabricated which uses the large

cold plate, allows independent attachment of each section of the stage assembly,

62

and is slightly thicker. For this altered design, the device is spring loaded onto a
0.25 inch thick aluminum stage assembly with PCB, 0.125 inch thick where the
device is mounted. This stage assembly then screws down to the cold plate that
sits atop a large 10.5 inch diameter support plate that bolts to the main
baseplate. This support plate was designed to utilize the real estate of the

baseplate for testing larger devices (Appendix B).

Support plate for cold plate (BOTTOM)
- bolts to baseplate of belljar
- 1/8" thick stainless steel

0

Cold plate (MIDDLE) ' 5.2500"
- McMaster H I

 

 

 

 

 

 

 

 

- aluminum
0 o , o 0
6,1111'
0 O
K449” Stage plate (TOP)
° / ° - aluminum
*3.5000"—'~

Figure 32: Large Device Cold Plate Assembly.

5.3.7 Cold Side Temperature Adjustment

it is sometimes necessary to operate power generation devices above a cold
side temperature of 300°K for optimal performance. Therefore, a cold side

heater was made to accomplish this, Figure 33. The heater is made using two

63

cartridge heaters are from McMaster1 that are silver epoxied2 into a 0.5 inch

thick by 2 inch square OFHC block. The heaters are controlled with a Sorenson
power supply3 that can provide them with up to 25W each, 50W total in parallel.

\llfing-nut

S ring

A uminum plate
#832 Screw
Quartz Tube

OFHC Heater Block
Device Under Test

PCB
Aluminum Stage plate

OFHC Heater Block
Glass slide

Cold Plate

 

Figure 33: Cold plate stage with added heater.

The heat required to achieve the desired cold side temperature depends on the
device characteristics and testing conditions. The cold side heater block was
initially mounted to the cold plate, and the temperature controlled by adjusting
power to the cold side heaters. A large amount of power was required to achieve
temperatures above 350°K for some devices under certain testing conditions.
Such high power caused electromagnetic interference with the data collection

wiring, figures 34 and 35.

 

l McMaster Carr 0.25 inch diameter, 200 W, cartridge style heater
2 Aremco silver filled epoxy, pyroduct 558
3 Sorenson power supply model number: DCS60-18E

64

- , ' Power leads

   

l" _; ._ Data wiring
, j 32-pin
l connector

 

 

 

 

 

 

 

 

 

 

 

 

160 _ ......... , ......... 6 ......... o
0W :1”
155 _- .......,,° ° 0 .. <25 -
I 000(5) o '1 0.05
150 '- ll 0 o o °o° ‘ z
: 0%..0 o e o 4 a
-145: °°%o&° , 0° :-01 §
é : (b O o a <
I; 140 ’LCoid side °oo 0 j a;
: heater power<40W,=40W (9 ° - 0.15%
. O O ‘ A
135 '- ° 5
: ° 0 .
: o - -0.2
130 :l 0 0T (K) using Pt-sensorfl '
E L0 Module VoltageM .0 .
125 L; a a L . a i . 1 ......... l ......... 0.25
1550 1600 1650 1700
Data Point #

Figure 35: EM interference with cold side heater and data leads.

Two adjustments were made to the system design in order to correct this
problem. First, the amount of power required to the cold side heaters was
reduced by increasing the thermal resistance between cold plate and cold side
heater. The thermal resistance was increased by simply placing a glass

microscope slide between the heater and cold plate. Second, a 2.75 inch conflat

65

style electrical feedthrough1 was added to the system to allow separation of the

power supplied to the heaters and data collection wires.

5.3.8 Temperature Gradient Heaters

In order to test thermoelectric devices for power generation applications, a
temperature gradient must be produced. The gradient needs to be uniform

across the device and must often be sustained for different load testing
conditions. As shown through the heat balance equation, Q = aTI — —:-IZR + KAT ,

changing the load current, I, affects the gradient, AT, across the device for some
fixed heater power, Q. Therefore, the gradient heater must be able to provide
enough heat uniformly for various load conditions. Depending on the size of the
device tested, the size and type of heater may need to be adjusted. For
example, if the heater surface area is much small than the device, the heat may
not distribute evenly across the device; and, if the heater is too large, the heat
radiated from the exposed surface area of the heater may induce undesired
lateral temperature gradients on the device. Therefore, several heaters were
investigated that varied in size, ability to supply high power at high temperatures
under ambient and UHV conditions, and ability for use with laboratory available

50V/50A power supplies.

 

I Kurt J. Lesker company electrical feedthrough part number: FTT0181053

66

5.3.8.1 Temperature Gradient Heater, Large Devices

First, a single cartridge style heater1 was embedded into a rectangular OFHC
block with silver paintz. The OFHC heater block was sized to fit the cartridge
style heater. The sample device for testing was a power generation device3
whose cross sectional area was 1.2 in2. This device was chosen because

reference data was readily available. The heater block covered 50 percent of the

total area of the device, figure 36.

 

Figure 36: Early stage design under test.

The open circuit voltage was recorded under a 20° gradient with cold side at
300°K, and the Seebeck measured was 300uV/K per couple, a 21 percent
difference from the expected 380uV/K [43]. The voltage collected from the
device appeared clean from the 2182 nanovoltmeter as no spikes or random

noise were observed, and the measurement repeated well. Next suspected was

 

I Omega Engineering 0.25 inch diameter cartridge style heater
2 Ted Pella company silver paint
3 Tellurex power generation device, model CZl-l.0-127-1.27

67

that the gradient was not being measured correctly, however the thermocouples
appeared to be adhered securely to the device under test, and the voltage to
temperature conversion look up table for the thermocouple seemed correct.
Next, it was presumed that the heat was not uniformly spread out across the
device through the alumina plate to the device pellets. So, a 0.0156 inch thick
piece of OFHC, the same size of the device, was silver painted between the
copper block and device to spread out the heat, as shown in figure 36. This

improved the agreement, but not by much, only to 31 OuV/K.

Next, a gradient heater was made from two 25W power resistors with aluminum
housing connected in series, silver painted and spring loaded to the top of the
device, figure 37. Results did not show a significant change with and without
spring loading, but the devices under test are usually spring loaded to ensure an

even thermal contact from run to run.

 

Figure 37: Series heaters for uniform dT.

68

The 25W resistors were chosen to deliver enough power to the device to
generate temperature gradients as given in the reference data, near 150°K. To
achieve 150°K, the amount of power is proportional to the thermal conductance
of the device, neglecting losses. Based on thermal conductivity data for BizTea
[27], measured pellet dimensions and the number of pellets, the thermal
conductance was found to be 341 mW/K. Therefore, the amount of heat required
was found to be near 51W, and two resistors would achieve this if supplied the
appropriate power. The resistors were also chosen for their size, which nearly
covered the device surface by greater than 95 percent. Finally, the resistors
were an appropriate load for the given power supply when connected in series or
parallel. With these resistors as the heater, 3 small 10° gradient was established
and voltage measured from the device which yielded a corresponding Seebeck
of 354pV/K, less than a 3 percent difference from the expected 365uV/K.
Although operated beyond the recommended temperature limits, the heaters
were used to create a larger gradient, near 100°, but the measured Seebeck
decreased to 320uV/K after an hour of testing. The test was stopped, device and
heaters inspected, heater resistors gave the correct resistance, but no flaws were
obvious; however, a smell of burnt electronics was noticed. The heaters were
used once more, and an infrared camera was used to measure the temperature
gradient in comparison with the thermocouples. Although the results in absolute
temperature between thermocouples and camera did not agree, for reasons
discussed in section 6, the camera discovered that one of the heaters (left) was

hotter than the other (right), figure 38.

69

 

Figure 38: lR view of burnt aluminum housed style heater.

in fact, the heaters were taken out and inspecte‘d again, and one was found to
have cracked and come loose from its aluminum housing, inhibiting heat transfer,
and thereby appearing hotter on the infrared image. The heaters were no longer
used, and replaced in later designs with ones that could deliver high power at
high temperatures. Nevertheless, the importance of uniform heat transfer was

made clear for accurate Seebeck measurements.

Another OFHC block was made which allowed two cartridge style heaters1 to be
embedded in parallel, figure 39. These heaters are wire wound resistors in
stainless steel sheaths good to greater than 600°C (873°K), designed for 120V
and were specified for good to 100W power ratings. The heaters are silver
painted into the OFHC for a good thermal contact. The OFHC is then spring
loaded to the top of the device under test. A 22mm 0.0. quartz tube was used to
separate the spring loaded aluminum plate and OFHC heater to reduce thermal
losses. With a small 20° gradient, cold side at 310°K, the Seebeck measured

322uV/K, a 15% difference from the expected 380uV/K.

 

l McMaster Carr 0.25 inch diameter, 100W, cartridge style heaters

70

 

Figure 39: Parallel heaters in OFHC for large size devices.

The heaters work well, but were too high of resistance, and limit the power
supply voltage to deliver them only 71W when connected electrically in parallel
for combined reduced resistance. Although these heaters are rated for higher
temperatures, they can not achieve the high temperatures desired, above 500°K,
under atmospheric conditions, with the system power supply used. Under UHV,
the heaters are sufficient to produce the high power at high temperatures,
however when placed in the OFHC block, are too large for the majority of devices
tested. Therefore, a substrate style heater1 was purchased which can supply
high temperatures at high power and is an appropriate load for the system power

supplyz.

 

LTB-l 75 model substrate heater from Heat Wave Labs
Sorenson DCS60-18E power supply

71

 

 

 

 

 

 

 

 

14 L+V°W39° (V) +P°W9HW1 ‘ 140
_ —+—Current (A) 1
3: C J 120
g c -‘ 100 g:
0 10 i- 3
g I -' so i
8 8 - : ..
3 ‘ - 50
g : 1
L 1

 

 

 

4 i....i....1.. 20
1150 1200 1250 1300 1350 1400 1450 1500
Temperature (K)

Figure 40: TB-175 characteristics.

The heater size is appropriate for testing smaller devices of 0.5in2, and can be
placed into an OFHC block to distribute heat evenly for larger devices. The
heater is 0.75 inch diameter, made of a tungsten filament embedded in an
alumina potting, with surrounding molybdenum shielding to reduce power lost

due to radiation. This heater has yet to be used in the system.

5.3.8.2 Temperature Gradient Heaters, Small Devices

For small single couple devices, thin film platinum resistive temperature sensors1
were first used as gradient heaters. Gradients up to 550°K have been
established across 8mm high devices made from the LAST material [41]. The

heater is made of two resistors connecter in series, and both are silver painted to

 

l
Pt-S 00 platinum thin film sensor from Omega Engineering

72

the devices under test. The devices tested were made of metallized alumina

plates silver paintd to the LAST material pellets.

 

Figure 41: Small devices under test.

The heaters work well, and have supplied enough power to achieve hot side
temperatures near 900°K under open circuit and UHV conditions. The Pt-103
resistors are made as a thin film device, and a protective coating covers the
resistive material. However, the coating on the heater tends to burn at higher
temperatures, near 900°K. The heaters were used repeatedly by adhering them
with silver paint from device to device. While preparing the devices for testing,
adhering heater to device, silver paint tended to creep from the heater to device
pellets creating an undesirable conductive path to the pellets. Therefore, the Pt-
103 sensors satisfied the gradient heater requirements for small devices, but
could not be used for repeated testing. Although a thin film device, the Pt-103
sensors are too expensive to purchase for each device tested. The substrate
style heater discussed in the previous section for testing larger devices is also

appropriate for testing small devices.

73

5.4 DTS Hardware Costs

The following table gives the approximate cost of the system components.

Table 4: DTS cost, approximate.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Quantity Item Supplier Cost (5)
1 18 inch stainless steel vacuum sudassembly KJL 26,164
(cart, chamber, hoist, chiller, vacuum system,
vacuum gauges and controller)
4 2182 Nanovoitmeter Keithley 8,800
1 2002 Multimeter Keithley 2,500
1 2400 Sourcemeter Keithley 2,000
1 DC860-18E power supply Sorenson 2,000
1 EMS ll power supply MSU ECE shop 1,000
1 DAC Measurement Supply 100
and Computim

1 PC MSU ECE shop 400
1 Sapphire viewport KJL 1,000
1 IR camera (optional) Indigo Systems 50,000
1 Water feedthrough KJL 120
1 4-rod power feedthrough KJL 262
1 32-pin feedthrough KJL 367
1 8-pin power feedthrough KJL 112
3 Conflat plate nut and silver plated bolt sets KJL 250
1 Assortment of power resistors Huntington Electric 75
1 Small cold plate MSU Phys Mach shm 120
1 Large cold glate McMaster Carr 30
2 Aluminum stages MSU Phys Mach shop 40
3 OFHC heater blocks MSU Phys Mach shop 120
1 TB-175 substrate heater Heat Wave Labs 600
1 PCB MSU ECE shop 40
1 Cold plate flex connectors KJL 156
1 Cold plate VCO fittings KJL 30
1 GPIB interface PCI card National Instruments 445
1 Chiller damper Custom 100
7 GPIB cables MSU ECE shop 140
4 Cartridge style heaters McMaster 100
1 Cartridge style heaters Omega Engineering 30
1 Nickel plating Adam’s Plating 40
1 TAV vent valve BOC Edwards 270
1 Conflat and QF-25 bellow, clamps, o~rings Varian inc. 300
Total w/o camera 47,823

 

5.5 DTS Software

The DTS is not fully automated, so some operations require user feedback

during the test. Data collection of device voltage, temperatures and power

74

 

supplied are fully automated, but the changing load conditions and temperature
control are manually adjusted. The measurement equipment is controlled
through a PC with LabView® National Instruments Software, version 7.0.
LabView® was chosen because it is compatible with the other MSU
measurement systems from which several programs and simple subroutines
were readily available [45]. The system is intended to measure the properties of
devices described in section 5.6. The flowchart for how the system operates is

provided below, figure 42.

 

   
   

 

  
 

  
 
  

 

 

INITIALIZE ESTABLISH COLD “#135:
Insmuuau‘rs :a 5'95 TEWER‘TURE Resistance

 

   

APPLY L , L
RLOAD I V
l l

   

GENERATE AT

     
  

 
 
 
 
 

MEASURE
AT, HEATER
POWER &
MODULE
VOLTAGE

AT AT
STEADY-
STATE?

SAVE DATA
AS TEXT FILE

     
  
  
 

 

 

YES

 

CALCULATE
MODULE
PARAMETERS

 

   

CHANGE
RLOAD

 

 

 
 

 

   

Figure 42: DTS Software flowchart.

The DTS LabView® programs are located on the DTS computer with the system,
C:lProgram FilesWationaI InstrumentleabVlEW 7.0luser.lib, and the library

is named DTS programs.vi. The main program to operate the system is named

75

DTS Main program.vi. Operating instructions for the DTS can be found in

Appendix H.

5.6 DTS Device Measurements

The DTS measures device Seebeck, resistance, efficiency and power out for
varying load and gradient conditions. These parameters are functions of
temperature, and most of a temperature gradient. Therefore, careful
consideration must be placed on measuring the gradient. The placement of the
temperature sensors and materials used to make the testing stage apparatus
was found to be critical to accurate temperature measurement. Several
measurements have been performed until the proper placement was identified
and materials used. Several of the experiments that lead to the success of the

most accurate sensor placement are listed in (Appendix D).

Single-ended type T thermocouples were used for measuring temperatures 290-
500°K, and type E for ranges of 300-900°K. Pt-100 platinum sensors were also
used to monitor the temperatures to characterize the thermocouples in initial

system development.

76

5.6.1 Device Seebeck Measurements

One of the most important properties of a good thermoelectric device is its overall
Seebeck coefficient. Therefore, accurate Seebeck measurements are critical for
assessing device performance. As discussed for measurements on bulk
materials in section 3.3.2, the relationship between voltage and temperature is
linear through the thermopower. Unlike measurements made on bulk materials
where single ended thermocouples are used to measure the gradient and voltage
produced at the same local point on the sample, the device Seebeck coefficient
is found from the device open circuit voltage, and separate thermocouples are
fixed to the hot and cold side of the device to measure the temperature gradient.
The device voltage and temperature sensors are not in the same location, so
they do not track each other in time. Thus, it is important to allow the device to
achieve steady-state conditions before data collection, or else measurement
errors can occur [46]. The time required to reach steady-state varies depending
on testing conditions and device characteristics. When the voltage and
temperature gradient do not track each other in time, a non-linear slope change

result can occur, figure 43.

77

0.01 1 U V I V t V l I V U I V—V—V—‘I '- 1 T I V V V
I

 

-0.012

'1—‘“'

-0.014 :-

-0.016

t. I

   

-0.018

VI'V

Module Voltage (V)

.092 LSlope change due to ——>

' delocalized thermocouple

.0322 {and voltage leads

llljlllll

 

 

0.024*4421.mi1...1...1.....12.."
80 1 00 120 140 1 60 1 80 200

dT (K)

 

Figure 43: Nonlinear effects of time delay for dVdT on Seebeck results.

The resulting effects on the Seebeck coefficient are provided in figure 44.

 

 

 

 

 

 

 

 

 

 

20 .. vrvv'VT-H'I . 360
l ° C” (Kll stmwwma
f ' °
' ,0. ,.e O .0 .Q... 1. .9. .0 0’. ° " 340
15 I. ‘5...“ 0’. '3..¢.. . e. . . 0. ~. ’0...“ ‘9 ”—9 g
_ ‘0 . ‘ n.
. f; 0 TEP (uVIK) per couple usmg Pt—sensofii g
' 0 320 a
A 0 ‘
5 10 - ,o’ 8
E .o " §
° W lags dT, so Seebeck is 1 300 3r
_ 0 smaller until steady-state ‘ ‘3
5 '- O 1 Z
a 280
I l
0 .W;.l....l...al....l....l....l..a. 260

 

 

0 5 10 15 20 25 30 35 40

Time (minutes)

Figure 44: Seebeck results, steady-state versus non-steady-state.

The Seebeck coefficient for non-steady-state conditions is shown as the first 15

minutes of data collection, and steady-state conditions are shown thereafter.

78

The data collected is a little noisy because the system chamber was open and
under atmospheric conditions, and tends to decrease substantially for tests done

under vacuum.

in order to verify the system measures device Seebeck coefficient accurately, a
standard reference device was needed; however, no such standard was found
readily available. Nevertheless, Seebeck coefficients for a variety of
thermocouples are well characterized and were therefore used as a reference
[7]. in particular, a device made of type K thermocouples was used as a
reference for characterizing the percent error in measurement of the Seebeck

coefficients.

Three type K thermocouples were connected electrically in series, and thermally
in parallel to make a thermoelectric device, figure 45. The device was used as a
reference for characterizing system errors in Seebeck measurements. Any
number of thermocouples could have been used for this test, but three were
chosen to increase the voltage magnitude detected to that of other devices
typically tested, and more than three would have served the same purpose but

been more time consuming to assemble. The thermocouple wire is 10 mil

diameter from the Omega Engineering Company1. The thermocouple junctions

were made by spark welding, and then thermally anchored to two metallized

‘

1
Omega Engineering type K thermocouple wire part numbers: SPCH-OlO-SO and SPAL-OlO-SO

79

alumina plates with alumina filled epoxy1, spaced 2mm apart. A quartz tube was
used to support the top plate, and is bonded to the quartz tube with the epoxy as
well. A slit along the quartz tube length was made to allow manipulation of the

wires so as not to cause any shorts. The starting and ending thermocouple leads
are soldered to the bottom cold metallized alumina plate. Two 3mil teflon coated

copper leads were soldered with the thermocouple leads to monitor the voltage.

 

Figure 45: Type K thermocouple device.

The Seebeck coefficient for the device was found at various temperature
gradients up to 220°K, as well as varying the cold side temperature from 295-
340°K. The results agree well with expected reference [7]. and are shown in

figure 46.

 

 

l
Aremco ceramabond 668

80

30-..”... ...
' -e~- Measured module voltage (mV)
: +Expected module voltage (mV)

 

V V r V V V I V VT

 

 

......,
A%error|'120

 

 

  
 
  

 

25 -
- 10

 

20- ____..o

Module voltage (mV)
(Measured, Expected)
1%) 10110 :ueaaed

15

Measured Seebeck = 116.8iVIK j _20
Expected Seebeck = 120.91VIK

10 A l l l A l A l l l l l l A l I LLLJ .l l l l I l‘

80 100 120 140 160 180 200 220 240
dT (K)

 

 

 

Figure 46: Type K device Seebeck vs. Temperature.

Since there are three couples that make this device, dividing the slope by 3 gives
the corresponding Seebeck value for the thermocouple, near 40uVlK [7]. A three

percent error in Seebeck measurements was on a type K thermocouple device.

5.6.2 Device Resistance Measurements

A second important device parameter is its overall resistance at various
temperatures. The temperature dependent resistivity and dimensions of the
materials used to make a thermoelectric device are usually known within some
tolerance. However, parasitic electrical contact resistance plays a crucial role in

device efficiency and is usually quite different than the material efficiency.

A standard four probe method is used to measure the device resistance, as done

for electrical conductivity discussed in section 3.3.1. In order to verify the system

81

measures device resistance accurately, a standard reference device was
needed; however, no such standard was found readily available. Therefore, a
commercial device, with temperature dependent data provided, was used as a
reference for calibration of resistance measurements from 300-400°K [43].

Results are provided in figure 47.

 

O
0
IILALJ

 

Z;
eaueiagip %

 

Resistance (ohms)
A

alnammlaaaa

 

 

 

 

: -°-Measured resistance (ohms) '20
2 . *Teliurex reference data ‘

- --° - - BizTe3 Marlow reference data 1 -25
0 --.--% difference (measured vs. Tellurex ref data‘ . ‘ 30

 

260 280 300 320 340 360 380 400 420

Temperature (K)

Figure 47: Tellurex device resistance versus temperature (300-400’K).

The device resistance was measured lower than the reference data by 5-10

percent. This percent difference may be due to manufacturing tolerances, as no

two devices are made exactly the same1. Approaching the results as a result of
manufacturing differences, this indicated the conductivity of the materials used in
making this particular device is higher, or that the temperature was measured
incorrectly. If the resistance is lower and conductivity is higher for the pellets of

the device tested, then the device Seebeck should be lower. The device

 

1
For phone conversation with Tellurex employees.

82

Seebeck was measured lower, but could also have been a result of measuring
temperature incorrectly. A second Pt-100 temperature sensor was used to verify
the measurement of the average temperature measured by the first sensor, and
both were in good agreement. Although the device tested was not a first choice
for reference, it did indicate that the system measures device resistance within
10 percent. A second system measured the AC resistance of the exact same
device at room temperature, and found good agreement, within 1 percent, with

measurements taken by this system [42].

5.6.3 Device Figure of Merit Measurement Analysis

As shown in section, 4.3, the figure of merit, ZT, is directly related to the device
conversion efficiency. Therefore, knowing ZT for a device can be one of the
most important measurements performed by this system. The figure of merit for
thermoelectric materials is commonly found by calculation of the three measured
parameters discussed in section 3.3, however this requires careful consideration
to the losses accounted for in accurate measurements of thermal conductivity.
To avoid this measurement, the figure of merit measured for thermoelectric
couples has been found using a variety of methods, some of which do not require
measurement of thermal conductivity [47,48]. The method chosen uses a
modified open-short circuit test [47], similar to the technique discovered by Al-
Obaidi and Goldsmid (herein referred to as method 1) [48]. Method 1 determines
the figure of merit through a ratio of open and short circuit condition thermal

conductance under small temperature gradient conditions. When a small

83

temperature gradient is applied across a device under open circuit conditions,

heat transport occurs by thermal conduction, QOpen = KAT. If the device is short

circuited, an additional Peltier heat contributes to the conduction giving a total

heat flow of QShm = KAT + S A BIT . For short circuit conditions,

RModuIe

 

S A BAT = IRMOduIe, so S A BIT = S A B[ JT. if the temperature gradient is

held constant for open and short circuit measurements and Joule effects are
neglected, the total heat flow under short circuit conditions may be rewritten in

terms thermal conductance to yield the figure of merit.

27‘ = _I_<S_ho_"_ -1 (37)
KOpen

Method 1 gives a quick approximate result for the figure of merit, but has several
limitations. First, method 1 is used for obtaining the figure of merit under small
temperature gradients. Most thermoelectric devices are used under high
temperature gradients, so they must also be tested under high gradients.
Second, method 1 neglects Joule and Thomson heating effects which are
inherent with current flow in thermoelectric materials. Third, method 1 requires
short circuit conditions, an impractical testing condition for thermoelectric devices
with low resistances. For these reasons, method 1 was adapted by Min and
Rowe to allow for testing under varied load conditions (herein referred to as
method 2). Method 2 extends the use of method 1 by determining the figure of
merit for non-short circuit conditions including Joule effects due to Peltier current.

Starting with the heat flow at the device hot side,

84

1 2
QLoad = “Load THotI '— 51 RModule + KAT Load (38)

given open circuit conditions,

QOpen = KAT Open and VOpen = “OpenAT Open (39)
and load circuit conditions from Kirchoff’s laws,

01 Load AT Load = (RLoad + RModule )1 = RTotaII (40)
the figure of merit is determined.

AT Open _
ZT = ”Load . (41

2
[( RModule I 2THot J _(TH01 - T Cold ][ RModule J
RTotaI T Hot + TCold THO! + TCold RTotaI

Analysis for this approach is shown in Appendix E. The figure of merit can also

 

 

be found for a fixed temperature gradient and varied heat source.

QLoad _1
QOpen , 42

2
[ 2THot )( RModuIe J _ 1(TH01 - TCoId ][ RModule J
T Hot + TCoId RTotaI 2 THot + TCold RTotal

Analysis for this approach is shown in Appendix F. Here, subscripts “Load"

ZTAve =

 

 

indicate effects under load conditions, and “Open” refers to effects when the
device is open circuited. The device resistance, R Module, is given at the average
temperature under load conditions. Therefore, the device resistance is found at
temperatures equivalent to the load circuit conditions, but with no load attached.
Min and Rowe give their analysis for steady-state conditions, and neglect heat

losses due to conduction, convection and radiation. Method 2 may be used for

85

testing under small temperature gradients, or when the material parameters are
held constant with respect to some temperature. For testing under large
gradients, the material parameters may not be approximated as constant, losses

must be accounted for, and the analysis must be modified.

The losses must be accounted for since they are no longer considered equal or
negligible for large gradient testing on large devices. With the heat supplied to
the device hot side fixed for open and load circuit conditions, the temperature
gradient across the device decreases for the loaded condition due to the
additional Peltier effect. Therefore, conduction losses down measurement wires
from the hot side, radiation and convection are all higher for the open circuit
condition. The difference in losses between open and load circuit conditions is
proportional to the difference in gradients. If the difference in temperature

gradients is kept small, the change in losses may be neglected:
QOpen = QLoad + AQLosses : QOpen = QLoad

Depending on testing conditions, losses for each condition should be calculated

to verify if they can be neglected to simplify the analysis.

Method 2 was further improved by including the Thomson effects that are usually
neglected for small temperature gradients (herein referred to as method 3).
Larger commercial devices tested under large temperature gradients have shown
that Thomson effects should not be ignored [32]. Depending on the device

tested and the percent of measurement accuracy required, the Thomson effect

86

may or may not have a significant affect on the results. Method 3 further adapts
the analysis to testing under large temperature gradients, but still holds the
material parameters constant. The analysis for incorporating material

temperature dependencies presents an opportunity for future work.

5. 7 DTS Measurements of a Commercial Device

A commercially available power generation device from Tellurex Corporation,
model CZ1-1.0-127-1.27, was tested under varied atmospheres, cold side

temperatures, temperature gradients, and load conditions.

Figure 48: 021 -1.0-127-1.27.

The results provided are for atmospheric conditions under a closed bell-jar,

assumed 750 Torr. The device was previously tested under vacuum pressures

from 10'3 to 10'6 Torr, but the expected Seebeck voltages did not agree well.

Table 5: Seebeck under pressure.

 

 

 

 

 

Pressure T-cold T-hot Seebeck Seebeck %diff
(Torr. Pa) (K) (K) measured (uVIK) expected (uVIK)
5X10-6, 6.66X10-4 311.8 334.4 314.6 380 17.2
7x10-1, 93.3 310.9 334.7 327 380 14
L 756,1.01X10+5 311.6 334 345 380 9

 

 

 

 

 

 

 

87

Research on the pressure dependencies of some Bi2Te3 thermoelectric
compounds may explain these results, but this has yet to be investigated further
[49]. By comparisons with reference data provided, it was determined through
various experiments that the devices were intended for use under atmospheric

conditions.

For testing this device, the gradient heater was made from that described in
section 5.3.8.1. The cold side stage used for testing this device is described in
section 5.3.6. For testing under varied load conditions, the device was allow to
reach near steady-state conditions for accurate data collection. Device size,
gradient, ambient pressure, load, average temperature and several other
conditions affect the steady-state time, so care must be taken when measuring
devices under varied conditions. An example of steady-state testing is shown in
figure 49 of raw data collected from the system on the Tellurex device while

testing at a constant temperature gradient and varied load conditions.

88

 

 

 

 

     

 

 

+Cold side temperature (K)
~45 Hot Side temperature(K) . Module Voltage (V) 4
400 dT (K)
. <——
. 3.5
E 350 ,.... _
2. 300 : 7 : : = = = = ‘ : : : = : = = : = : = = ‘ ’ . : ; ; : : ; : ; : : .- s‘ : : s -: 3
O '- d
‘3. E a 2 5%
O 250 E Adjusting dT : ' E.
E i No load 1 2 2
£200 _(eeeeeeseeeeeoeeeeeee ‘ 1m : 9-
3 150 - - 1-5‘8
g-100 1 3
5
'— 50 fpwu. 0~5
0 :w. 1 1 LA I L 0

 

 

 

0 200 400 600 800 1000 1200
Time (minutes)

Figure 49: Raw data collection, temperature and voltage versus time.

The following load curves for the Tellurex device are provided as figures 50

 

 

 

 

 

 

 

through 55.
2.5 ...,...,...,...,...o
: +V-module (V) j
. —°—V-module: Ref 1
2 % difference V-modulel '1 '10 33
I 1 a.
r J
2 1.5. “2, g
.2 ' ‘ a
3 '- 1
'o 8
° <
T 1 " -l -30 g
> r 1
i ' 8-
. - C
. . a
0.5 - . .40
o .4 a n 14 A a l A a l I a L #1 L a a j A a n l n a A l a n 1 . .50

 

 

0 2 4 6 8 1 0 12 1 4 16
Load (ohms)

Figure 50: Tellurex load error curve, T-coid = 50°C, T-hot = 100°C.

89

2.5 .,.fi,. -,

.°
N

 

 

 

 

 

 

 

 

I +P-module (W): Ref +V-moduie (V): Ref 3
C —°—l-module (A): Ref +V-module (V) f 1 0.6
2 _' +l-module (A) ’ 1
- +P-module(W) = J 0.5
S . ' I
I 1'5 ' J 0.4
s :
8 = :
g 1 0.3
> ’ g
-j 0.2
f 0.1
. . J . . .L 1 . . . l . i . l . . A ' o

 

 

8 10 12 14 1 6
Load (ohms)

 

Figure 51: Tellurex load curve, T-cold = 50°C, T-hot = 100°C.

 

 

 

 

 

 

 

2.5 ...,4-fi,...,.f.,... 0
+V-module (V) ‘
+V-module: Ref .

2 1 % difference V-module ‘ '10
I
E 1.5 - - -20
o .
'5 . .
B I. .
o _ .
5 1 - - -30
>
0.5 '- .‘ .40
o b a 1 a l A a a L n a a l 1 a A l a a a L. n n 144+l a n a ‘ .50

 

 

 

0 2 4 6 8 10 12 14 16
Load (Ohms)

Figure 52: Tellurex load error curve, T-coid = 75°C, T-hot = 125°C.

90

(M) alnpow-d ‘lvl alnpouI-l

elnpoui-A eaueaeglp °/.

V-module (V)

V-module (V)

P
s:

 

 

 

l

 

3 ............... - - - - -2 -
I +I-module (A): Ref -—e—~V-modula (V): Refl I
I +P-module (W): Ref +deul° (V) f ‘

 

 

 

 

 

 

25 S—-°—i-module(A) .106 g
:+P-module(W) 3 8.

2 _ - 0.55
: . 1 °
: : E

15 - g- :04"
P / O

1: /'+ . - as:
r / ’ a

/

05- : = ‘02;

0ps-n1.L.1...1...1L..L..nll-L" “0.1
0 2 4 s s 10 12 14 16

Load (ohms)

Figure 53: Tellurex load curve, T-cold = 75°C, T-hot = 125°C.

 

 

 

 

 

 

3.5 7.......,..r,...,fi. o
+V-module(V)
“ew V-module: Ref -- ”0 -
m~% difference V-modulel ' '. '10
‘ i
2 i .
P a: , ' ‘
t . ’ - -30
1.5 _- I
1 _ -‘ -40
d
0.5iggr).1...i.a.l...11..l..21...|...‘.50
0 2 4 6 8 10 12 14 16
Load (ohms)

Figure 54: Tellurex load error curve, T-cold = 75°C, T-hot = 150°C.

91

empow-A 0000103“) %

4 vvvvvvvvvvv I V’- I l

 

 

 

 

 

 

 

 

 

 

E "—i-module(A): Ref ' V-moduie (V): Ref L 1.2
3.5 : +P-modula (W): Ref +V-modulew) j
; *r—l-module (A) 0 5
3 :+P-module(W) M / 1 1 8
E : “a..- - a-
0 2 5 _- r: ' J T ‘ 6i . "' 0.8 ‘5
i ' - ‘ - a:
‘l': 2 I" i 0.6 g
> : o >. ,...,, _ . Q
n a .r ‘Q *~ a-.- 1 5
1'5 I- ” '2 i - 1 0.4%
I x '“ ‘ ‘5‘ ~\;:; 1 O 2
0.5'.§i.1...1...1...1...i...1....1,,l
0 2 4 6 8 10 12 14 16

Load (ohms)

Figure 55: Tellurex load curve, T-cold = 75°C, T-hot = 150°C.

The Tellurex device has been tested under various load and temperature
conditions. The percent errors in voltage and current measurements under load
conditions ranged from 15-22 percent, and some of this error can be accounted
for. First, this device has a tolerance of difference from device to device
manufactured, and will therefore yield different results than the reference data.
This would agree with the resistance measurements taken previously given in
figure 47. Second, the reference data provided is in plot form, and was extracted
to the best of the author's ability, but one should expect some error. Third, the
deviations in open circuit voltage measurements merely transfer through for all

load conditions.

92

As shown in figure 56, the figure of merit can be computed using the open-load
circuit test with a fixed heat supply [47].

1.2

 

 

vvvvvvv I v v v I v v v I v v v I I V T 2.5
I +Measured ZT ‘

. +ZT (Bi2Te3 from Marlow) :
1 H '1 2

 

 

 

 

’ ? 15 g
g 9 v a. E
< c
l: - <— A. . a
M -
-1 ‘53
‘ E

 

 

~0— Measured efficiency (%)
+ Efficiency (Bi2Te3 from Marlow) 0.5

 

 

 

T-cold=323°K, T-hot=373‘K

a 4 l a

6 8 10 12
Load (ohms)

 

Figure 56: ZTAV. and efficiency for Tellurex device, T-cold = 323°K, T-hot = 373°K.

The efficiency of figure 56 is calculated based on ideal effects because the
contact resistance of the device is included in the measured device resistance,
and the wiring resistance between device and load of 65m!) is included in the
load resistance. The figure of merit drops suddenly for a near short circuit

condition, so one might want to investigate further.

Thermal losses due to conduction, convection, and radiation for open and load
circuit conditions have been computed [45]. For the experimental setup
provided, the difference in heat losses due to these three mechanisms was found
to be near equal, and their effect negligible for this test. See Appendix G, for

heat loss calculations.

93

 

 

 

 

 

 

 

 

 

50-..,...,...,...,...,20
- 1
I -‘ 1s 3
~ . -i
40 _ : g
g I T 14 3
v 44 b ”A : ‘ '
'6 CE“) :.. a,
- <—— . +(dT-open)-(dT-load)| i .1.
42 - , . . 8
' ‘ ;—--——S " 10 3
4° 5 '1 8
' Open circuit dT = 56.5 j
38 n n a I a a n l n g; n . .4 a a L l 6
0 2 4 e s 10

Load (ohms)

Figure 57: Tellurex device dT changes, T-cold = 50°C, T-hot near 100°C.

The Thomson heat generated is not accounted for in the analysis leading to the
figure of merit. For the results presented, the effects of Thomson heating are
compared, per couple, relative to the other load circuit effects to determine what
inaccuracy this may cause on the figure of merit. A worst case scenario is
considered under high electrical current outputs, and figure 51 is used for the
165m!) load condition.

QLoad = K(T)AT + aHor (T)TH011 _-;_12RModule -%TATI

The right most term is the Thomson heat, found from Thomson coefficient

as: = Tag—6%, and is calculated from reference data obtained for the Seebeck

voltage [43].

94

 

 

 

 

 

 

 

 

A 280 7. . - - . r s 250

g . —°—Thermopower per pellet (uVIK)] . 4 _.
> .

E. 260 ’ ’ Thomson Coefficient per pellet (uVIK‘ , §
§ l- . 3
a : j :
. 240 - . n
a - . 3
a . . a?
5 22° - - 200 a“
a - . 3
° 2
E - .,
o 200 - .3
'E I ._.
0 ,. g
3 130 - .2.
0 ’ .

2 . —y=6.40466304+0.577425159x R=1 . 3

160 . . . l 4 4 1 A 1 . . . . l . A . . 1 1 150
300 350 400 450
Temperature (K)

F lgure 58: Thomson coefficient for Tellurex device.

of}:

For an average temperature of 75°C (348°K), 2' = 20 . The Thomson heat is

calculated for the 165m!) load current of 1.4A as TQ = —z' IAT a 14mW per pellet

(28mW per couple) is liberated since current flows with the gradient. The
Thomson heating is not a junction effect like the Peltier, however only one half
contributes to the hot side, acting as an additional 14mW “from” the hot side.

The Peltier heat is found as PQ = SHotlTHO, =115mW per pellet (230mW per

couple). The Joule heat per couple is calculated from the average device

resistance value obtained from figure 47 as 4.40 divided into 127 couples. The
Joule heat per couple is JQ = I 2 £7 = 68mW, 34mW additional “from” the hot

side.

95

 

 

 
 
 
 
 
 
  
  
 

 

 

 

 

 

 

 

 

10 V V V 1 I V V V V I V _I.=Lr I V V - 1
-°-Peltler heat (W)
-*-Joule heat (W)
3 . -°-Thomson heat (W)
: T-cold=323K, T-hot=373K ‘
. R—module @ Tave = 4.80hms .
6 _ Seebeck @ Thot = 442uVIK -
E - Thomson @ Tave = 440 UVIKl
on P 4
g .
:I: 4 . ..
2 _ Take 1I2 values for
_ / hot side heat total
r- % 3 3 3 3 : 3 ‘ s c - -
o a 1 1 A I 1‘ T E w A _. I _ L
0 5 10 15 20
Load (ohms)

Figure 59: Thomson heat effects.
For the figure of merit analysis, the heat flow conditions for open and load circuit

testing are assumed equal, thereby canceling the thermal conduction effect from

both sides. Therefore, the Thomson effect contributes

f ITQI _ 14mW
lTQl+|PQ|+JQ _ 278mW

 

= 5% and is neglected for this analysis. This result may

give insight to why the figure of merit is lower for the 165m!) condition, and the
incorporation of Thomson heating into the figure of merit analysis is left for future

work.

The efficiency was also tested under higher temperature conditions, and the

following results were obtained neglecting Thomson effects.

96

 

5.l

Th
dr;
de

me

Table 6: ZT,". and efficiency for Tellurex device, T-cold =3 75°C, T-hot = 150°C.

 

 

 

Load ((1)1 ZT Efficiency (%)
1 5 0.53 1 .5
1 0 0.49 1 .7

 

 

 

 

 

5.8 DTS Measurements of LAST Devices

Thermoelectric devices have been fabricated from the LAST materials and tested
by the DTS measurement system. The LAST material thermoelectric properties
have been measured at various temperatures, and the data has been used to
determine the appropriate size of the thermoelements to fabricate a device to
yield maximum efficiency [50, 51]. Various fabrication techniques with the LAST

material and results of devices tested are discussed.

5.8.1 Calculated Performance

The size of the thermoelements used in fabricating a thermoelectric device can
dramatically affect the device efficiency. In fact, the conversion efficiency
decreases as leg length decreases because the ratio of electrical contact to
material resistance increases, as well as an increase in the ratio of thermal

conductivity of the material to the ceramic layers used [51,52].

 

‘ Wiring resistance adds 65m0 to this value, and was done so for ZT and efficiency provided.

97

77C

 

 

'7: 2 (43)
AL 2 1 4 1+ [:0
[1.2 c] 2-_,,.[_] 4L
ACLO 2 ZTHO, 1+2 AL,
ACLO

Here nC,/i,1C,Lc,Lo,ZTH0,, pc, pare the Carnot efficiency, thermal conductivity
of the thermoelement materials, thermal conductivity of the electrical and thermal
contact materials, thickness of the ceramic layers, thickness of the
thermoelement length, figure of merit at the hot side temperature, electrical
contact resistivity, and thermoelement resistivity, respectively.

However, the device thermoelements can not be made too long or the rate of

heat flow would decrease, as well as the power out.

azATvo
2
(2pLo{1+ 2pc)[1+ %]
pLo AcLo

Here a,ATo , A0 are the Seebeck coefficient, temperature gradient and cross

 

(44)

PEIectricaIOut =

 

sectional area of the thermoelements, respectively. This shows there is a trade-
off between device efficiency and the power produced when designing a device

for optimal performance.

A thermoelectric material and device modeling program has been used to
determine the dimensions of the thermoelements needed for maximum efficiency
for the LAST materials [51]. Using the modeling program developed for

LabView®, Efficiency vs. current density.vi, and given temperature dependent

98

data for P and N type materials, the optimal thermoelement sizes were

determined.

5.8.2 LAST Fabrication Methods and Results

A few fabrication methods have been investigated for use with the LAST material
and are discussed; however as the summary of results table below shows, the

best method may not have been discovered yet.

Table 7: Summary of LAST module results.

 

 

 

 

 

 

Contact °/. contact Expected ZT Expected Measured Measured
method resistance (T -ave=600K) efficiency ZT efficiency
Ag-paint 99 1.3 10 0.017 < 1
Ag-epoxy 99 - - - -
Solder 40 - - - -
Diffusion bond 15 0.4 4 0.020 < 1

 

 

 

 

 

 

5.8.2.1 LAST devices fabricated with silver paint

LAST devices were initially fabricated with a silver paint from the Ted Pella
Company. There is a coefficient of thermal expansion (cte) mismatch between
the thermoelement and contact materials, however the paint is able to withstand
temperature cycling without any noticeable peeling or cracking up to 900°K. The
silver paint is an excellent thermal and electrical conductor; however, the paint
contributes a dominating contact resistance that substantially degrades device
performance. When using silver paint, the contact resistance on one such device

was found from data of table 10 in the following manner at room temperature.

99

 

 

Table 8: Data for computing contact resistance from silver paint.

 

 

 

 

 

Material P-(LAST) N-(Ag-Sn-Sb-Te)
Electrical conductivity (S/cm) 1,500 1,200
Length (cm) 0.80 0.80

Area (cm"2) 0.25 0.32
Resistance (m0) 2.13 2.08

 

 

 

 

 

The device resistance was measured as described in section 5.6.2,and

measured near 20, giving the percent contact resistance as,

0.00421

 

 

R
%RContact ____ (I _ Thermoelements J x1 00 = [I _

)x1 00 = 99.89% (45)
R0verall

In order to lower the contact resistance, the silver painted devices were annealed
at 400°C for 2 hours in a 54 sccm 95% Ar, 5% H atmosphere. After annealing,
the device resistance decreased to 0.950, but still yielded a contact resistance
about 99.55%. One such device fabricated in this manner gave a figure of merit
near 0.01 under a 200° gradient, and the result was confirmed by another
temperature measurement system using the open-short thermal conductance
method [54]. The expected figure of merit for the device was found from

previous data taken on the individual thermoelements, Table 9, and by

 

21/2+fl1/2 2 pk “2
using 2 = 1 1: 2 ,where 7 = [44] [53]. The expected figure of
7 2K2

merit for an average temperature of 400°K was calculated to be 0.48.

100

 

Table 9: Room temperature thermoelement data for calculating ZT.

 

 

 

Material P-(LAST) N-(Ag-Sn-Sb-Te)
Thermal Conductivity (W/m°K) 8 2 ~ 2
Figure of merit (400K) 0.21 0.25

 

 

 

 

 

When applying the silver paint, if too much was applied, some tended to spill
over the edges of the thermoelement decreasing the effective length and
increasing the overall contact area. However, if not enough silver paint is
applied, some pores are formed in the contact when the drying occurs and
alcohols evaporate. Furthermore, pores form while the paint dries and cause
pores to form in the paint weakening the contact mechanically, thermally and
electrically. As evidence of this, another silver paint fabricated device was
repeatedly tested under various temperature gradients. The hot and cold side
temperatures were ranged up to 900°K and 300°K, respectively. The device

temperature gradient was cycled once a day for three days in UHV conditions.
The system UHV conditions were measured daily, 1X10.5 Torr on the first day,
2><10'6 Torr the second day, and 5><10'7 Torr the third day before the testing

chamber was released back to atmosphere. On the last day of testing, vacuum

conditions were released from 5X10'7 Torr to 750 Torr over a 30 minute period.

After each of the temperature cycles, the device resistance and Seebeck were

measured, and decreased after each time of cycling.

101

180

 

 

 

 

.Vacufim level: ' ' ' ' fl j ' 'KF1'253'R2A' '(éflisi').

day 1 -> 10-5Torr KF2229R1D (LAST) .

Lday 2 -> 2e-6Torr .

A 160 "day 3 -> 5e-7Torr ‘
g r Day 1, dT=1 8" for this Seebeck <
3 ' Resistance @ 300K = 0.950hms ‘
g 140 ' dT-max = 20° for day 1 run ‘

ooOOOCOOOOOOOOOOOOOOOOOOOOOOOOOOOO

5 _ Day 2, dT=23° for this Seebeck ,
120 .. Resistance @ 300K = 0.79ohms ..

g L dT max = 600° for day 2 run -
g L Day 3, dT=29° for this Seebeck ‘
100 - Resistance @ 300K = 0.550hms :

I dT max = 600° for day 3 run ,

80 L.L.L.L;ALL4.JL....l....l..-.
0 10 20 30 40 50 60

Data point it

Figure 60: LAST module 3 day test.

Due to the decreasing device resistance and Seebeck, it was suspected that
silver atoms from the paint moved into the thermoelements and contaminated the
device, as with the affect of copper contacts to BizTea based devices [11].
Therefore, the system was released to atmospheric conditions. At atmosphere
the device resistance increased unexpectedly to 1.20. The device was
disassembled and the electrical conductivity of the thermoelements were
measured at room temperature by a separate system, but the thermoelements
did not show any change in room temperature electrical conductivity from the
initial measurements made prior to device fabrication. During disassembly, it
was noticed that the top conducting plate removed easier than usual. It was
therefore suspected that the device contacts were changing during the testing,
such that as the device was cycled in temperature under UHV, any air pores

formed in the silver painted contact when initially fabricated were able to escape.

102

As the pores decreased and density of the contact material increased, the

contact resistance decreased.

The electrical and mechanical properties of silver painted contacts for

thermoelectric device fabrication with the LAST materials does not seem

appropriate. To improve upon this fabrication method, a silver filled epoxy1 was
purchased from Aremco for use in device fabrication. This epoxy was chosen
because it was good to high temperatures, offered low electrical resistance, and

is applicable with simple curing conditions.

5.8.2.2LAST devices fabricated with silver filled epoxy

A silver filled epoxy from Aremco was used to fabricate devices of the LAST
materials. Dilute solutions of phosphoric acid to clean etch the surface, applied
epoxy layer thickness, dilute version of the epoxy, and drying and baking times
were varied for device fabrication. The silver filled epoxy followed the same
contact resistance trend as the silver paint. The epoxy also began peeling at
temperatures above 420°K, indicating a mismatch in coefficient of thermal
expansion (cte); however, the epoxy has a room temperature cte of 17.3 inlin/°C
and Pb-Te alloys are near 18in/in/°C [54]. It is not presently understood why the

peeling occurs for some of the prepared samples but not others.

 

] Aremco silver filled epoxy pyroduct 597.

103

The silver epoxy fabricated devices formed a good mechanical contact, as the
material broke before the bond during stress testing. However, the epoxy gave a
dominating electrical contact resistance worse than that of the silver paint. The
overall device resistance ranged from 40-10000, depending on how the devices

were prepared and cured, and gave a best case contact resistance of,

 

 

mam, = 1— RThe’moe’ememS x100 = (1— O'006)x100 = 99.98%
R0verall 4’0

. . . . 1
To reduce the contact reSIstance, gentle polishing and several acrds were

experimented with to etch the thermoelement surfaces before applying the

epoxy. Also, deionized water, acetone, and alcohols2 were experimented with in
removing the various acids etchants before applying the epoxy. The best results
for low contact resistance were found by first etching the material with a dilute
phosphoric acid, cleaning with DI water, and then following up with the
manufacturers recommended epoxy curing procedure of a 2 hour room
temperature dry, and 2 hour 200°F bake. While drying, some of the devices
fabricated tended to have air pores that burst while baking. Therefore, the
devices were dried and hot plate baked under spring loading, in atmospheric
conditions under a hood. Spring loading was applied to make the contact as thin
as possible for a low thermal resistance, as well as inhibit the epoxy from forming
contact weakening pores as found with the silver painted devices. The force of
spring loading was not measured directly, but estimated to be less than 1psi.

The majority of thermoelements prepared with the dilute phosphoric acid and

 

1 Phosphoric, comet bond flux, and Kester flux acids.

2 Methanol and Isopropanol

104

cleaned with DI water gave average device resistances of 40-200(), but no
consistency was found in the process toward obtaining 400 or lower each time.
With hopes to reduce the contact resistance, devices were annealed on a hot
plate under a hood at 500°F for an additional 2-8 hours under atmospheric
conditions. After annealing, the contact resistance results did not improve and
the epoxy peeling became worse for some of the contacts. These devices were
not further tested under varied load and gradient conditions due to the high

electrical contact resistance.

As with the silver paint in fabricating devices using the LAST materials, the silver
epoxy does not seem a good choice for making electrical contacts. Others have
reported that the size and structure of the silver particles in a paint or epoxy has
shown to dramatically influence the contact resistance, and this may have been
the case for the paint and epoxy chosen for these experiments [55]. Other paints
and epoxies may lead to improved performance, and such experiments are

opportunities for future work.

5.8.2.3 LAST devices fabricated with various solders

Different fluxes, solders and soldering temperatures1 were experimented with for

fabricating devices made of the LAST materials, and best results were obtained

 

I Comet bond flux, Kester flux, Zn-based and PbSn solders, and temperatures from 400—700°F

105

through use of a comet bond flux1, Bi-Sn-Sb based solder2 with soldering
temperatures near 450°F. The soldered devices were not intended for testing at
high temperature gradients, moreover for improving the contact resistance. A
solder showing low contact resistance was considered a possible candidate for
making cold side electrical contacts to the LAST materials. The completely
soldered devices gave a lower overall resistance, ranging from 38-50mfl, giving
a contact resistance of near 40 percent depending on how the devices were
cleaned and soldered. Some of the LAST specimens cracked due to thermal
stress after using a soldering iron to apply the solder. To reduce the stress, a hot
plate was used to slowly raise the thermoelements to near the solder melting
point, and then a solder iron was used for additional heat to melt and apply the
solder. The device resistance was measured, and contact resistance extracted
based on the material properties. The contact resistance results gave good

agreement with testing done using a transmission line method (TLM) experiment.

Varied load and gradient testing has not been performed on these devices due to
the high contact resistance. Of the solders, fluxes and preparation methods
demonstrated, none seemed to give a reasonable contact resistance to the LAST
materials. The electrical characteristics of the Bi-Sn-Sb solder are proprietary,
and there seems to be a lattice mismatch between the solder and LAST

materials leading to the high contact resistance. To further lower the contact

 

1 Comet bond flux is from Kapp Alon and Wire
2 Bi-Sn-Sb based solder is from Tellurex Corporation

106

resistance, different metals were electron beam and sputter deposited onto the

LAST material as an intermediate step to the solder.

5.8.2.4 LAST devices fabricated with deposited contacts and

solder

LAST based devices were fabricated with electron beam (ebeam) deposited and
sputter deposited contacts using materials such as silver, gold, and tungsten.
The deposited metals were to provide an intermediate step between solders,
paints and epoxies. Tungsten was chosen because of the success it has given
with PbTe fabricated devices [56]. Silver and gold were chosen as a combination
of layers to offer high electrical conductivity from the silver, and a thin layer of

gold for reduced oxidation effects.

The silver and gold were ebeam deposited onto plasma etched LAST materials.
The plasma etching was made with an ebeam prior to deposition to remove any
oxide layers. The LAST material was at room temperature during deposition,
and deposition took place under sub 10'5 Torr atmospheric conditions. The
materials seemed to deposit well, and gave interesting results when combined
with the Bi-Sn-Sb solder to form a contact. For the silver and gold deposited
contacts, the gold and or silver and gold seemed to ball up with the solder, figure

61.

107

 

Figure 61: Soldered contacts.

Because of the dissolved intermediate deposited contact, no devices were

fabricated with this method of materials.

Next, Tungsten was ebeam deposited onto a new pellet of the LAST material.
Attempts to solder at various temperatures and with different fluxes were
unsuccessful. The solder did not wet the surface well and balled onto the
soldering iron tip. When trying to use a hot plate, the solder again bailed up and
did not spread evenly for a uniform contact. No devices were fabricated with this
method of materials. Attempts of soldering with an intermediate metal contact
were not unsuccessful because it initiated a more involved investigation of

diffusion bonded contacts.

5.8.2.5 LAST devices fabricated by diffusion bonding

Devices have been successfully made by diffusion bonding contacts of nickel
and antimony. Several other materials have been investigated, but this
combination gives the best present results. The contact resistance was

measured very small, giving only a 15 percent contact resistance as verified

108

through TLM measurements. The device resistance was measured to be 18mfl,
Seebeck measured 120pV/K, and figure of merit was found to be approximately
0.02. The measurement of the figure of merit requires steady-state conditions,
however this device did not appear to reach a steady-state condition even after a
two day period for each of the load and open circuit conditions. For the testing
conditions applied to this device, calculations show that if the gradient is
measured incorrectly by just 0.1 °, ZT can vary by 0.05. In addition to testing the
device using the open load circuit method, results for the device resistance and
figure of merit were found to be in good agreement with an AC method using two
lock-in amplifiers [42]. With the applied gradient, the expected ZT was found
from literature to be near 0.3, but this is not guaranteed to yield the same results
for the materials used in this device [41]. Temperature dependent data was not
taken on the materials prior to device fabrication and testing, so a precise

predicted figure of merit was not known.

The diffusion bonding methods for making contacts to the LAST material are
showing much promise over the initial attempts with paints, epoxies, and solders.
More devices are presently being fabricated with nickel-antimony diffusion
bonded contacts, and their properties will then be measured with the use of the

new device testing system.

109

6 Using an IR Camera for Device Testing

The temperature gradient applied across a thermoelectric device is one of the
measurements required to characterize its figure of merit [47]. Most device
gradients are measured by direct contact with the sample, by conduction of heat
to the junction of a thermocouple, diode or platinum resistor sensor; however,
each has limitations. The use of an infrared (IR) camera will be investigated as a
non-contact measurement of temperature gradients used for calculating the

figure of merit based on the open load circuit testing method.

6. 1 Why infrared

Direct contact measurements of the temperature gradient may involve
thermocouples, diodes or platinum resistive sensors. Diodes and platinum
resistive sensors are generally too large to measure small enough regions of
interest on smaller objects such as thermoelectric pellets or couples.
Thermocouples are also limited to a spatial resolution dictated by the size of the
junction of materials used to make it or the material that bonds it to the surface
for which it is placed upon. Furthermore, thermocouples conduct heat away from
an object that can distort the temperature profile or cause other undesirable
effects. Larger thermocouples are easier to prepare for sample testing, but suffer
smaller spatial resolution of temperature and have more conductive losses.
Decreasing the size of the thermocouple wires reduces both of these effects, but

many difficulties arise in sample preparation because they are small to handle

110

and manipulate. Errors in temperature measurements vary depending on the
size ratio of specimen under test and thermocouple contact. Also, the purity of
the materials used to make the thermocouple can affect its performance, so
materials are often sold as matched pairs based on their purity. Thermocouples
are fairly inexpensive compared to other tools used for obtaining temperature
gradients in thermoelectric testing, but do not offer as many features as more

recent technologies.

IR cameras provide an alternate method of measuring temperature that alleviates
many of the difficulties associated with direct contact measurements employing
thermocouples. In addition to measuring temperature by conduction,
temperature can be measured by the amount of electromagnetic radiation
emitted from an object. IR cameras are tools used to measure electromagnetic
radiation, and thus temperature without direct contact to the device under test.
However as with thermocouples, measurements with IR cameras offer new

challenges that must be overcome for obtaining accurate measurements.

6. 2 How Infrared Cameras Measure Temperature

All objects continuously emit and absorb energy in the form of electromagnetic
radiation. An object at thermal equilibrium with its surroundings emits and

absorbs energy at equal rates, but an object hotter than its surroundings emits
energy. For solid objects, the radiation is a function of the objects temperature

and surface characteristics, and is commonly referred to as thermal radiation.

111

Objects that emit and absorb energy perfectly without any reflections are called
blackbodies, and have an emissivity value of one. All other objects emit or
absorb a fraction of the energy involved as governed by the temperature
dependent emissivity value, 8(7) of the object. The emissivity is also frequency
dependent with a value typically given as a weighted average for the range of
corresponding wavelengths, a. The total amount of energy emitted from an
objects surface may be calculated relative to a black body by use of the Stefan-

Boltzmann law. This law states that the emitted energy, MD, is a function of

temperature whose magnitude is proportional through a constant, 053 [5].

 

 

R(T) = caSBT4 3V7 (46)
m
27I5k84 -8
053 = = 5.67x10 (47)
15h3c2 K4m2

Here Osaka." and c are the Stefan-Boltzmann constant, Boltzmann constant,

Planck's constant and the speed of light, respectively. For objects at
temperatures of 10-5000°K, the radiation predominantly occurs in the IR part of
the spectrum of 0.2-500nm wavelengths, so IR detectors are commonly used to
measure temperature. IR cameras have detectors made up of an array of IR
sensors to monitor temperature, and each sensor of the array represents a pixel

in an acquired thermal image.

Each material has a unique emissivity and therefore emits a different amount of

thermal energy for a particular temperature. Since IR cameras rely on the

112

magnitude of thermal energy hitting a sensor to yield a corresponding
temperature, two objects at the same temperature of different emissivity may be
measured at different temperatures. Thermoelectric devices are made of several
materials: thermoelectric material, electrical interconnects, solders, paints, and
thermal conduction layers. The device emissivity can be normalized by coating
the surfaces with a material that doesn’t greatly affect the material properties
such as electrical or thermal conductivities. The coating material must also be
able to withstand high temperatures, from room temperature to 900K for our
testing purposes. Therefore, a Rustoleum® flat black high temperature spray
paint has been used for this purpose, and a few coats are applied on a narrow
region of the device. it seems best to only apply a limited number of coatings, or
else the luster increases and decreases the emissivity further from the desired
black body effect. Now, the device is heated to a particular gradient also
measured by thermocouples, and the camera emissivity is adjusted to
correspond with the thermocouples. The thermocouple junctions are not directly
visible with the IR camera in the present testing apparatus, so the surrounding
areas of the thermocouples are viewed. The camera emissivity with use of the

black spray paint gave best agreement with the thermocouples when set to 0.7.

113

6.3 The Infrared Camera used at MSU for Thermoelectrics

The IR camera1 used in the Pulsed Laser Deposition and Transport
Characterization Laboratory at MSU uses a cooled lnSb detector array and is
designed for use with 3-5um wavelengths, corresponding to temperatures of
approximately 273-1000°K. Due to the high measurement range, this camera is
an ideal choice for testing thermoelectric devices for power generation
applications. It was purchased with interchangeable 25mm and 50mm lenses,
and has focal length adapters to allow precise viewing frOm several inches and
beyond. The resolution varies depending on which lens and adapter are used. It
also has intensity filters for measuring above 573°K. The camera is part of a
system that includes a PC with Windows 2000 ® based environment and

Thermagram ® software for a user friendly system. Further specifications are

available from the manufacturer's websitez.

The camera is not directly positioned near the devices since they are enclosed in

a vacuum environment, so a viewport window made of sapphire has been

chosen. Quartz3 transmits 0.18-2.5um wavelengths, whereas sapphire will
transmit > 90% of wavelengths in range of 1-5pm. The total transmission of

radiation from sample, 911,0,”de , to camera is, ‘Prransmifled = 0-9‘PEmmed-

 

l Mid-range Merlin from Indigo Systems Company, a division of Flear International.
2 http://www.indigosystems.com/product/rnerlinhtml
3 Quartz transmission from Goodfellow

114

6.4 Using the IR camera

Different lenses and focal length adapters are provided with the camera to

maximize the magnification for a given field of view and distance to object. With
the present testing conditions, there exists about 4.5 inches between device and
IR camera, and this includes space between device to sapphire, and IR camera

only 0.5 inches away from the closed bell-jar viewport.

 

Figure 62: Closest camera to device distance ls 11cm.

With present lenses, maximum magnification of 0.5 is obtained with the 50mm
lens and 1 inch focal length adapter (referred to as an extender ring). An infrared
image may not look good visually, but can still be very good thermally. However,
devices tested appear quite clear, especially those made with the LAST material
of larger sizes. The devices by Tellurex are difficult to obtain a clear image, but
shows up thermally very well when each pixel temperature is profiled across a

specified region of interest.

115

6.4.1 IR measurements on bulk materials

 

Figure 63: LAST pellet (left), IR thermal image (right).

Results of comparing measurements of the temperature gradient using
thermocouples to IR camera are shown in the tables below. Best agreement with

the thermocouples was found on the non-painted surface under atmospheric

 

 

 

 

conditions.
Table 10: TC vs. IR dT comparison, no vacuum (750 Torr).
dT1 de (IR camera, dTa (IR camera, TC vs. IR w/o TC vs. IR w/
(thermocouples) w/o paint) painted surface) paint (%) paint (%)
10.15 12 26.4 18.2 160.1
35 40 58 14.3 65.7
68 92 107 35.3 57.4

 

 

 

 

 

 

However, with reduced convection (8x10.2 Torr) the result is just the opposite

and better agreement was found with the thermocouples on the painted surface.

116

 

Table 11: dT comparison, 8x10'z Torr vacuum.

 

 

 

 

 

 

dT, de (IR camera, dT3 (IR camera, TC vs. lR w/o TC vs. IR w/
(thermocouples) w/o paint) painted surface) paint (%) paint (%)
4.5 2.7 4 40.0 1 1.1
16 10.4 14 35.0 12.5
34.2 11.2 29 67.3 15.2
57.2 42 48 26.6 16.1
87.3 60.9 71.3 30.2 18.3

 

 

 

 

 

 

 

More testing should be done to decrease the atmospheric pressure to see if
results between thermocouple and IR improve further. Although the absolute
temperatures do not agree between camera and thermocouples, the key feature
of the IR camera is that it measures several points along the gradient as opposed
to just two from the thermocouple. A temperature profile versus pellet length is
provided in figure 64. For this plot, there exist 105 data points collected across
the 7mm high pellet, corresponding to a resolution of 21 pixels/mm (0.095°lpixel).
This is not necessarily the defined resolution for the camera, but is for the
configuration used for this measurement. This configuration of 1inch focal length
adapter and 50mm lens was chosen to yield the highest magnification for an

object 4.5 inches away from the detector.

117

 

 

 

 

 

  

 

100 .......,....,..fi.fis..,....,....,....,.
- + Temperature (F) TC's
i ° Temperature (F) Bare .
95f x Temperature (F) Faintmi.
E .
2 90- _'
3 ' j
E ‘l
a 85- i
s . .
d)
i- .
sol -'
75 1....J+Llu....|....I....I....I....1...

 

 

 

0 1 2 3 4 5 6 7
Distance (mm)

Figure 64: N16 dT profile.

6.4.2 IR Measurements on a Commercial Device

IR measurements have also been used on the large Tellurex power generation

device. Figure 65 shows the device tested under the assembly of figure 39.

4

 

Figure 65: Tellurex device.

The IR measurements are compared to that as taken with thermocouples
previously and show good agreement. Using the infrared camera with an

average device temperature of 310°K, the figure of merit was measured as ZT =

118

0.65. The thermocouples were used in a different test with average temperature
of 340°K and gave ZT = 0.55. For the BizTea based Tellurex device, the
temperature dependent figure of merit follows a parabolic curve for the
temperature range tested. This may explain why the figure of merit is higher for
the lower temperature test measured with the IR camera than the one measured

with the thermocouples at higher temperatures.

6.4.3 IR Measurements on a LAST Device

The IR camera has been used to test thermoelectric devices made from the
LAST material as well. Results are for an early fabricated silver painted device
are shown at steady-state conditions, and no black spray paint was used to
normalize the material emissivity differences. Notice how the thermal image of
figure 66 (right) has a odd thermal distribution from light (hot) to dark (cold) areas

in a non-linear fashion.

 

Figure 66: LAST made device.

This uneven thermal image temperature distribution is due to changes in

emissivity between the materials used. Thermocouples were used to measure

119

the temperature gradient as well, and did not show any unordinary behavior such
as this. If the thermal profile were real, and emissivities all matched, then the
device ends would be cooler than the middle section, not possible for the steady-
state condition. The figure of merit was not determined from use of the IR
camera in its early uses, so no ZT data is provided here. This device was used
to show the capability of testing with an IR camera. Further devices made from
diffusion bonding methods described earlier will soon be tested using the IR

camera .

The IR camera has shown use for testing bulk materials, commercial devices,
and devices made from the new LAST materials. The technique needs to be
investigated further for use under varied vacuum conditions. The IR camera has
obtained good agreement with the use of thermocouples, and provides the
temperature profile in addition to the gradient by acquiring several points along
the length of the sample. This data may also be useful for providing additional
insight with thermal modeling techniques that divide the thermoelectric device

thermoelements into a finite number of small sections.

120

7 Conclusions

A new measurement system has been developed for testing thermoelectric
devices based on new materials for power generation applications. The process
of developing the device testing system (DTS), it’s features and capabilities have
been discussed. This system allows measurement of several device properties
that previous systems in the MSU lab could not accommodate. This system is
unique because it allows for testing of larger devices in addition to bulk samples
and single couples that commercial systems, such as Z-Meters, are designed for.
The capability of non-contact measurements using an IR camera has also been
proposed. The system measurement tolerances have been given by testing of
reference materials. Device resistance and Seebeck versus temperature, device
performance under varied load and gradient conditions, figure of merit and
efficiency have all been measured by this system. Several device fabrication
methods for the new LAST material have been shown, as well as their results.
The results of fabricating devices of the LAST material are helping to provide
insight for additional device improvements. Some of the system limitations have
been presented and provide opportunities for extending its capabilities through

future work.

121

8 Future Work

The main topic ideas for future work are present below with brief descriptions.
Several of these ideas have been presented throughout the paper, and are just

reiterated in summary here.

8.1 Autonomous operation

In order to control the temperature gradient across the device, a
temperature feedback system would need to be incorporated. This could
be satisfied through use of a temperature controller, such as those
available through Lakeshore Cryogenic industries‘ of through additional

system programming.

The load changing can be automated with a switchbox of low contact
switch resistance. With such modifications to the system, near

autonomous operation may be achieved.

8.2 Software

The software used presently works, but may be modified with more

autonomous features.

 

' Lakeshore model 330 series temperature controllers.

122

8.3

8.4

8.5

8.6

Analysis

The calculations for heat losses and their affects on high gradient changes
should be considered. Also, non-linear temperature gradients and their
affect on the Joule and Thomson effects at large gradients should be

derived for the analysis leading to the figure of merit.

Sample Stage

A new sample stage should be developed to test samples at higher
temperatures that present conditions, limited by the melting of PCB used
with GE varnish and soldered leads. New bonding materials to fix heater
to device should be investigated, matching coefficient of thermal
expansion. This new sample stage may also be designed to test bulk

samples.

Larger devices

A larger sample stage, higher cooling capacity chiller, and larger heater is
required for testing devices required for the 1kW generator. The system

can presently can test devices up to 1.51n2.

Hardware

An automated vent valve with Nitrogen backflll for clean venting should be

installed on the Turbo pump for more efficient venting.

123

RS-232 should be utilized on the vacuum gauge controller for vacuum
dependent testing on materials and devices. This could also be used to
control the flow of a backfill gas.

A new DAC controller for PC to old EMSII analog controlled power supply
is needed since original one malfunctioned. The same DAC should not be
purchased as it seemed to cause fluctuations in the power supply voltage.
An infrared camera mounting stage should be designed and attached to
the bell-jar housing, as any vibration caused by the system or user would

be in unison while viewing.

124

APPENDIX A

On a microscopic scale, semi-conductors are of crystalline form, for which the
atoms are not randomly placed to make up the material, but bonded with some
periodic pattern. Consider a semi-conductor material subjected to energy in the
form of heat. The atoms are chemically bonded to each other fixing their center
of mass, so they can not move very far unless more energy is given to break
their bond with neighboring atoms. Suppose just enough energy is applied to
elastically displace the atoms from their original positions in the lattice, hence no
bonds are broken. The atoms will begin to vibrate around their center of mass as
a function of the energy induced, and the effect of one atom’s movement onto
another carries on until the energy is distributed and damped by each interaction
of neighboring atoms. Each atom has a distinct associated mass. If two atoms,
one small and one large, are subjected to some fixed energy, the smaller atom
will displace more than the larger because the larger can not change it's
momentum as easily. This is analogous to striking a billiard ball with a cue stick,
compared to applying the same energy with a cue stick to a bowling ball. The
energy that disrupted atoms from their natural positions can be quantized and

represented as a particle called a phonon [57].

125

APPENDIX B

 

 

 

f) e 10.50"
(4) e9/32"
4.50" ——e>1
1.00"
i O
= % 0.75"
0 45 ~
(4)#1c>24—/ 1 f
1.375" ‘91 14‘
3.75"
4’ II
0 O

 

Figure 67: Cold plate support plate.

126

The stage is made of 0.25 inch thick aluminum, ground and polished upon
completion. All holes are #8-32 unless othenNise specified, and the red inner
holes are counterbore only as affirrned in the side view. All dimensions are
symmetric unless otherwise specified, and the 1/16 inch diameter hole was

placed along the horizontal at 1.75 inch vertical.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IQE 3.50"
—=— ‘flfl'
, o o
(8) #8-32
"sh-150'?“—
¢1/16" 5_ .
4‘50“ \. (DJ (b
7/ !
2'4'5" 0 O "2.1.”?i
1%”035" 1.75 1
. 9 fl ‘3’ l

 

 

 

 

 

 

I
0.75%— * ._.. {.100}
SIDE i i m: Mi 1 l

_ L 0.125" I 015..

counterbore

 

 

 

 

 

 

Figure 68: Large stage assembly aluminum stage.

The cold plate is as shown in Figure 28: DTS large cold plate.

127

APPENDIX C

The following figure is the wiring diagram for the device testing system, which

includes wiring to 32-pin military connector and power feedthrough ports on the

baseplate from the PCB and heaters. The gradient heater and load are wired

directly, so no PCB pin is designated for these.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32 O Pin1
Pin 14
81 O
PCB Pin 32-Pin Mil U__se Meter
1 A Thermocouple + (cold) 2182 #1 (+)
2 B Thermocouple - (cold) 2182 #1 (-)
3 C Thermocouple + (hot) 2182 #2 (+)
4 D Thennocouple - (hot) 2182 #2 (-)
5 E Module/Load V + 2182 #4 (+)
6 F Module/Load V - 2182 #4 (-)
7 G Pt-100 l + (hot) 2400 (rear)
8 H Pt-100 I - (hot) 2400 (rear)
9 J Pt-100 l + (cold) 2002 HI (front)
10 K Pt-100 I - (cold) 2002 L0 (front)
11 L Pt-100 V + (cold) 2002 SENSE HI (red)
12 M Pt-100 V - (cold) 2002 SENSE LO (blk)
13 N Pt-100 V + (hot) 2182 #3 (+)
14 P Pt-100 V - (hot) 2182 #3 (-)
V Heater l + (cold) DCSGO-18E (+)
f Heater l - (cold) DCS60-18E ( -)
EC_B Eov_ve_r
Wide Traces Feedthru
1 Load Attachment
2 Load Attachment

ANN-F

Figure 69: PCB wiring Diagram.

Heater l + (hot)
Heater l - (hot)

128

APPENDIX D

1. Differential thermocouples: not accurate because of reasons stated earlier
in the thermocouple section of this document.
2. Single-ended type E, 1mil bare

a. Placed under the heater atop of the device: forces heater to not be
in good thermal contact with the device, and I shorted by the Ag-
paint holding the heater in place.

b. Placed on side of alumina: no noticeable difference as opposed to
atop of heater. However, must be adhered to each device, not
convenient for device loading and unloading.

3. Single-ended type T, 10mil coated and leads twisted together

a. GE varnished to stage next to device: measuring too cold
compared to actual temperature across device. 80 placed atop of
alumina on device bottom plate.

b. GE varnished between device electrical leads, more accurate.
Placed a Pt-100 sensor next to the TC to monitor and compare.
This Pt-100 is used as a reference for the cold side thermocouples,
which if the device and reference temperature of the electrical leads
on circuit board are the same temperature, should read near zero if
no gradient exists.

c. Placed inside alumina tube and silver epoxied into the heater block,
and verified the temperature by placing a Pt-100 sensor atop of the

heater block. The Pt-100 and TC agreed within 10%, but this was

129

still not accurate enough. Next the Pt-100 was placed inside the
heater with silver epoxy 4mm away from it, in a 3/8 inch deep hole

in the heater block that is 0.5 inches thick.

TOP VIEW

Heater #1
Quartz tube spacer

Pt-100
Thermocouple
Heater #2

OFHC Heater Block

 

SIDE VIEW

Thermocouple

Quartz tube spacer
Pt-100

Heaters
OFHC Heater Block

 

1"—1.2600'——-1

Figure 70: Temperature sensor trial.

4. An infrared camera was used to verify both sensors, but is not taken as
absolute rather the gradient, only for comparison. The gradient was in
best agreement at an emissivity setting of 0.8, and agreed within 10% of
the platinum sensors.

5. The bottom Pt-100 was mounted on the aluminum stage bottom in an
alumina filled epoxy, which was milled out to allow wiring and sensor to be
embedded without affecting the smooth thermal interface: best agreement
between sensors and thermocouples.

a. The cold side Pt-100 was using the curve on the 2002, and the hot

side was found from a 4-wire measurement and linearly fit data

130

from Omega, but the linear curve gave 3° inaccuracy so was

regenerated as a polyfit with one tenth of a degree accuracy.

131

APPENDIX E

Equating the heat supplied for open and load testing conditions gives,

1 2
[aLoad THotI - E I RModule + KAT Load :l

K

 

AT Open =

Substituting for the load circuit current,

 

 

a Load ATLoad J _ _1_ [ a Load AT Load
RT otal RTotaI

2
2 J RModule 1' KAT Load
K

a Load THO! [

 

AT Open =

Factoring out a ATLoad to the other side, and moving K into the numerator,

 

 

ATOW “L d 1 0L d 2 RM d1
= aLoad THot 00 - —ATLoad oa 0 u e +1
AT Load (RTotaI )K 2 RTotaI K

2
a . . R
——L9—"—?"— and multiply numerator and denominator by JBd—‘d-‘i

RToial )K RModuIe

Factor out the

and rewrite to obtain,

2
ATOPe" _ a Load [RModule )[T _ 1 AT [RModule II
- Hot Load
AT Load RModuleK RTotaI 2 RTotaI

Substitute for the figure of merit and factoring the equation,

 

 

2
ATOPC’" _ THot RModuIe _ _1__ RModule
._. Z AT Load +1
AT Load RTotaI 2 RTotal

Multiply the numerator and denominator by The = TH“ :TCOI‘I . and replace

 

AT Load With T Hot — T Cold .

132

 

 

 

2
ATOPe" = ZTA [[ RModuIe ][ 2THO! ]_ [ THO! - TCoId I RModule J +1
ve
AT Load RTotaI THot + T Cold THO! + TCold RTotal

133

APPENDIX F

For the following analysis, ideal conditions of zero losses are assumed. Again,
starting from the heat balance equation,

1 2
QLoad = a Load T Hall "' 5 I RModuIe + KAT Load

open circuit conditions,
QOpen = KAT Open

V Open = a OpenAT Open
and load circuit conditions from Kirchoff’s laws,

a Load AT Load = (RLoad + RModule )1 = RTotal I
the formula for ZT is obtained by the following method. First, the open and load
circuit conditions are coupled into a single equation.

1 2
“Load THotI " 5 I RModule '1' KAT Load

QLoad =
QOpen KAT Open

 

Substituting for the device current,

 

 

 

2
aLoadATLoad J 1 (‘1 Load ATLoad ]
a]. dTHt( -— RMdI +KATL d
QLoad = 0a 0 (RTotal) 2 (RTotal) 0 u e 00
QOpen KAT Open

and maintaining a constant temperature gradient allows

ATOM" = ATLoad = AT = THO, — TCOM. Also, thermal conductance of the device

does not change significantly in open circuit conditions compared to load

conditions, as it is a material property only dependent on temperature.

134

 

 

2 2
a L (1 TH 1 (I d
0" 0’ “—(T Hot ‘T Cold )[ L00 ] RModule
QLoad _ (RTOMI) 2 (RTotal)

+ 1
QOpen K

Next, multiply by Mend moving K into the numerator to get
Module

2 2 2
QLoad __ a Load THo! RModu/e _l T -T aLoad RModu/e 1
- ( Hot Cold) K R +
QOpen RModuleK RTotal 2 RModule ) Total

 

 

2
Using the relationship for the figure of merit, Z = m, and multiplying the
RModuleK

2 T T . T T
( Hot '1' Cold) ’ and usmg TAve = Hot '1' Cold ’

2(T Hot + T Cold ) 2

 

 

first term by

 

 

2
QLoad = ZTA ( 2TI-Iot I RModule )_ _1_[ THot — TCold ][ RModule ] + 1
ve
QOpen THor + TCold RTotal 2 T Hot + TCold RTotaI

Finally, rewriting in terms 01 ZTAve'

QLoad _ 1
QOpen

Z T A ve = 2
[ 2THot ]( RModule ]_ 1[THot " TCold ]( RModule )
T Hot '1' TCold RTotal 2 THot '1' TCold RTotal

 

 

 

135

API

Call

des

Cal

'mt'

. n e C II\ (.1 1

APPENDIX G

Calculations for heat losses contributed from testing of the Tellurex device

described in section 5.7 are provided below.

 

Calculations for heat losses in thermoelectric device testing
in the DTS system with a Tellurex device.

 

k = thermal conductivity
1 = length

a = cross sectional area
K = thermal conductance
Q = heat

sa = surface area

em = emissivity

q = quartz

c = copper

co = constantan
ch = chrome]

w = wiring

to = thermocouple
p = pellets (device)
t = total

on = conduction
cv = convection

r = radiation

 

 

Conduction Loss Calculations

 

 

--> Thermal conductivity, k, data for near room temperature, from Goodfellow
--> Quartz heat insulator:

W
kq := 1.46—
mK

lq := 2.54cm

136

mm 2 mm 2
aq 2:1: 22—) —1r- 20—
2 2

 

 

Kq := kq-fl-
lq
--> Copper wiring fi'om heater:
kc := 401-1-
m-K
1c := 25.4cm
ac := 1r-(lmm)2
Kc := 2-kc-fi
lc
--> Thermocouple wiring:
kch := I9-w—
mK
lch := 25.4cm
(.050&:m)2
ach := 1r-
2
Kch := kch-31h-
lch
kco := 19.5l
mK
lco := 25.4cm
.05082m 2
aco := 1r-
2
Kco := kco-Ec—O-
lco

 

For open circuit conditions:

 

Th := 423K

Tc := 331K

dT := Th — Tc

--> quartz losses
Qq := Kq-dT

Qq = 0.353w

--> copper losses
Qc := Kc-dT

Qc = 0.923w

--> thermocouple losses
th ;= (Kch + Kco)-dT

th = 2.857x 10‘ 3w

--> conduction total loses
ant g: Qq + Qc + th
ant = 1.278W

137

 

For 0.165 ohm load circuit conditions:

 

Th := 403K

Tc := 33(K

dT := Th — Tc

Qq := Kq-dT

Qc := Kc-dT

th := (Kch + Kco)-dT
Qt := Qq + QC + th
Qt = 1.003W

 

 

Convection Loss Calculations

 

 

--> Device pellets, 4 sides of each as 1x1.5mm:
sap := (127)-(4)-(1mm)-(l.5mn)

sah := (I)-(4-1.2in-0.5in + 1.2in-1.2in)

saw := (2)1t-(0.O2in)-lch

saq := n-22mmlin

--> Note: 1 mm of mercury is 1 Torr

sat := saq + saw + sah + sap

sat = 58.057cm2

--> Convection total losses, assuming argon atmosphere

--> hconv = an approximate convection transfer coefficient for atmospheric conditions
found from:

hconv := 10—
m2-K

 

For open circuit conditions:

 

Qformula := hconv -(373K - 323K)
cht := Qformula-sat
cht = 2.903W

 

For 0.165 ohm load circuit conditions:

 

Qformula := hconv -(357K —- 323K)
cht := Qformula-sat
cht = 1.974W

138

 

Radiation Losses

 

--> Stefan Boltzmann constant, sbc

sbc :-_- 5.67-10— 81—-

K4-m2
--> emissivity of copper, emc
emc := 0.03

--> emissivity of Bismuth Telluride, guess as close to graphite, 0.98
emp := 0.92

--> outer surface area of the pellets, osap

osap := (60)-(l)-lmml.5mn

 

For open circuit conditions:

 

Tc2 := 323K
Th2 2: 373K
--> radiation losses for the copper

Qrc := sbc fl—]~(Th24 - Tc24)-sah

 

_ '(2 - emc)
—-> radiation losses for the outside pellets, with only outside face exposed
Qrp := sbc--—cmp—-jl-('l"1124 - Tc24)-osap
_ (2 - emp)
Qrt := Qrc + Qrp
Qn = 0.06W

 

For 0.165 ohm load circuit conditions:

 

Tc3 := 323K

Th3 := 357K

--> radiation losses for the copper

Qrc := sbc-i -(Th24 — T024)-sah
_ (2 - emc)

--> radiation losses for the outside pellets, with only outside face exposed

Qrp := psbc emp ]-(Th24 - Tc24)-osap

_ (2 - emv)
Qrt := Qrc + Qrp

Qrt = 0.06W

on = 0.06W

 

139

 

APPENDIX H

STARTING THE SYSTEM

._.}

. Prepare a device for testing (see device mounting instructions)

2. With a vented system chamber at atmospheric conditions, open the bell-
jar with the yellow hoist operating switch, located in the left instrument
panel of the system cart.

3. Insert the device testing stage, and screw it into place, either with the
provided #8-32 screws or spring loaded assembly. Add thermal grease
between stage and cold plate heater as necessary.

4. Solder the device load leads to the 12 AWG feedthrough wiring.

5. Attach the edge card connector to the device testing stage PCB.

6. Alligator connect the dT heater leads to the feedthrough lead wiring (larger
devices only).

7. Ensure all wiring is in place properly and nothing inhibits the bell-jar from
sealing to the baseplate.

8. Close the system chamber with the hoist operating switch. The bell-jar
sealing O-ring should press evenly on the baseplate, and the baseplate to
be-jar spacing is near 1/16 of an inch.

9. Ensure the Turbo pump manual vent valve is closed, a black plastic screw
located on the pump housing side toward the front of the system.

10.Tum ON the rough pump.

11.Tum ON the switch for the Turbo pump and turbo cooling fan, located on

the front panel of the right side instrument bay of the system cart.

140

 

12.The BA gauge turns ON when the first set point is reached.

13. Monitor vacuum until desired, then turn OFF using the switch on the
vacuum gauge controller.

14.Turn ON the chiller, and adjust the flow rate using the bypass valve
located between the chiller supply and return lines of the chiller, observe
the pressure gauges on the chiller and chiller vibration damper located just
next to the chiller. This bypass needs to be adjusted manually for the
desired device cold side testing temperature.

15. Using LabView®, open the DTS Main.vi in the DTS programs library, and
set the desired testing conditions and fill in entries for the device
information.

16.Adjust the gradient and cold side heater power on the fly during testing,

and pause data collection for changing the loads.

141

STOPPING THEfS YS TE_M

17. Stop the DTS Main.vi by pressing the STOP button in the program.
Stopping the VI using the main LabView® menu will not reset the
instruments properly, and may cause problems for starting a future run.

18.Tum OFF the turbo switch, leave the fan running however.

19.Turn OFF the chiller.

20.When vacuum has returned to 10'3 — 10’2 Torr ranges, turn OFF the turbo
fan and rough pump.

21 .Vent the system chamber by unscrewing the manual vent valve on the
turbo pump housing.

22.Wait until the system bell-jar to baseplate spacing is back to 1/16 of an
inch, and the air hissing sound has stopped from the vent valve.

23. Open the bell-jar chamber using the hoist switch.

24. Unsolder load leads, detach edge card connector, remove the stage
mounting screws and remove the sample testing stage from the system.

25.Close the bell-jar and keep closed when the system is not in use.

142

REFERENCES

 

[1]

[2]
[3]

[4]

[5]

[6]

[7]
[8]
[9]

[10]

Harman and Honig, Thermoelectric and Thermomagnetic Eflects and Applications,
McGraw-Hill, Lincoln Laboratory Publications, 1967.

http://www.commerg2fl.carrier.com, April 10, 2005.

Gary L. Bennett, “Space Applications”, Chap. 41, CRC Handbook of
Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York, pg. 503, 1995.

Section H, CRC Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press
LLC, New York, 1995.

C. L. Foiles, Encyclopedia of Physics, VCH Publishers, Inc., New York, pg. 1263,
1991.

Heikes, R. R., Ure, R. W., Thermoelectricity: Science and Engineering, Interscience
Publishers, New York — London, 1961.

httn://srdz_it§.nist.gcw/its90/downlogd/downlogd.htrnl, April 1, 2005.
http://wwwmepdx.edu/~gerry/epub/pdf/thermocouplepdf, February 3, 2004.

A T type thermocouple consists of copper and constantan wires, K type is chromel
and alumel, E type consists of platinum and chromel, S type is platinum/10%
rhodium and platinum, and J type is iron and constantan.

Gao Min and David M. Rowe, “Peltier Devices as Generators”, Chap. 38 in CRC

Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York,
pg.482,l995.

[l l] H. J. Goldsmid, Electronic Refrigeration, Plenum, New York, 1964.

[12] Mahan, G. D., “Figure of Merit for thermoelectrics”, J. of App. Phys., Vol. 65, pg.

1578-1583, 1989.

[13] Mahan, G. D. and Sofo, J. 0., “The best thermoelectric”, Proc. from the Nat. Acad.

ofSci. U. S. A., App. Phys. Sci., Vol. 93, pg. 7436-7439, 1996.

[14] Z. H. Dughaish, “Lead telluride as a thermoelectric material for thermoelectric

power generation”, Physica B, Vol. 322, pg. 205-223, 2002.

[15] Slack and Hussain, “The maximum possible conversion efficiency of silicon-

gerrnanium thermoelectric generators”, J. of App. Phys., Vol. 70, pg. 2695, 1991.

143

 

[16] Tim Hogan, Sim Loo, Fu Guo, Jarrod Short, “Measurement Techniques for
Thermoelectric Materials and Modules”, Mat. Res. Soc. Sym. Proc., Vol. 793, pg.
405-411, 2004.

[17] Jarrod Short, Sim Loo, Sangeeta Lal, Kuei Fang Hsu, Eric Quarez, Mercouri G.
Kanatzidis, Timothy P. Hogan, “Hall Effect Measurements on New Thermoelectric
Materials”, Mat. Res. Soc. Sym. Proc., Vol. 793, pg. 323-332, 2004.

[18] Geoff D. Staneft, Paul D. Asimow and Thierry Caillat, “Synthesis and
thermoelectric properties of Ce(Ruo,67Rho,33)4Sb12”, Mat. Res. Soc. Conf Proc., Vol.
793, 2004.

[19] S. Seal, S.C. Kuiry, P. Georgieva, and A. Agarwal, “Manufacturing Nanocomposite
Parts: Present Status and Future Challenges”, Mat. Res. Soc. Bulk, pg. 16-21,
January 2004.

[20 S. Bhattacharya, Terry M. Tritt, Y. Xia, V. Ponnambalam, S. J. Poon, and N.
Thadhani, “Grain structure effects on the lattice thermal conductivity of Ti-based
half-Heusler alloys”, App. Phys. Lett., Vol. 81, No. l, 2002.

[21] L. D. Hicks and M. S. Dresselhaus, “Effect of quantum-well structures on the
thermoelectric figure of merit”, Phys. Rev. B, Vol. 47, No. 19, pg. 727-731, 1993.

[22] T. C. Harman, P. J. Taylor, M. P. Walsh, B. E. LaForge, “Quantum Dot Superlattice
Thermoelectric Materials and Devices”, Science, Vol. 297, No. 27, 2002.

[23] R. Venkatasubramanian et al., “MOCVD of BizTe3, szTe3 and their superlattice
structures for thin-fihn thermoelectric applications”, J. of Crystal Growth, Vol. 170,
pg. 817-821, 1997.

[24] R. Venkatasubramanian et al., “Thin-film thermoelectric devices with high room-
temperature figures of merit”, Nature, Vol. 413, pg. 597-602, 2001.

[25] Streetman and Banerjee, Solid State Electronic Devices, 5'” ed., Prentice Hall, NJ,
2000.

[26] R. Venkatasubramanian et al., “Thin-film thermoelectric devices with high room-
temperature figures of merit”, Nature, Vol. 413, pg. 597-602, 2001.

[27] C. M. Bhandari, “Minimizing the Thermal Conductivity”, Chap. 38 in CRC

Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York,
pg.56,1995.

144

 

[28] HI. Goldsmid and J .W. Sharp, “Estimation of the Thermal Band Gap of a
Semiconductor from Seebeck Measurements”, J. of Elec. Mat, Vol. 28, No. 7,
1999.

[29] I. A. Nishida, “Measurement of Electrical Properties”, Chap. 15 in CRC Handbook
of Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York, pg. 158-180,
1995.

[30] O. Maldonado, “Pulse method for simultaneous measurement of electric

thermopower and heat conductivity at low temperatures”, Cryogenics, Vol. 32, No.
10, 1992.

[31] Cobble, Milan H., “Calculations of Generator Performance”, Chap. 39 in CRC
Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York,
pg. 489-501, 1995.

[32] Gao Min, D M Rowe, and K Kontostavlakis, “Thermoelectric figure-of-merit under
large temperature differences”, J. of Phys. D, Vol. 37, pg. 1301-1304, 2004.

[33] Hugh H. Woodbury, Lionel M. Levinson, and Robert S. Lewandowski, “Z-Meters”,
Chap. 17 in CRC Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press
LLC, New York, pg. 181, 1995.

[34] T. C. Harman, “Special Techniques for Measurement of Thermoelectric Properties”,
Letters to the Editor, Amer. Inst. of Phys, pg. 1373, 1958.

[35] Hideo Iwasaki, Mikio Koyano, and Hidenobu Hon', “Evaluation of the figure of
Merit of Thermoelectric Materials by Harman Method”, J. of App. Phys, Vol. 41,
pg. 6606-6609, 2002.

[36] Hideo Iwasaki, Shin-ya Yokoyama, Tomoharu Tsulcui, Mikio Koyano, Hidenobu
Hori, and Seijirou Sano, “Evaluation of the figure of Merit of Thermoelectric
Modules by Harman Method”, J. of App. Phys, Vol. 42, pg. 3707-3708, 2003.

[3 7] Edward J. Kroliczek and M. Nelson Ross, “Design and Operation of Couple
Testers”, IEEE Conf Proc., 1960.

[3 8] Larry Selwitz and Gerhard Stapfer, “The Development and Use of an Integrated
Thermoelectric Life and Performance Testing System”, IEEE Conf Proc., 1960.

[39] Ikushi Yoshida and Isamu Yunoki, “Automatic Measuring System of

Thermoelectric Properties with a Desk-top Computer”, Proc. of the Sixth Int. Conf
on Thermoelectric Energy Conversion, pg. 39-42, 1986.

145

 

[40] Van Wynsberghe, R., “Rapid simultaneous measurement of thermoelectric material
properties”, Proc. of the Sixth Int. Conf on Thermoelectric Energy Conversion, pg.
28-34, 1986.

[41] Kuei Fang Hsu et al., "Cubic Ang(m)SbTe(2+m) Bulk Thermoelectric Materials
with High Figure of Merit", Science, Vol. 303, 2004.

[42] Adam Downey and Timothy P. Hogan, “Circuit model of a thermoelectric module
for AC electrical measurements”, Int. Conf on Thermoelectrics Proc., 2005.

[43] http://www.tellurex.com/12most.html July 12, 2004.
[44] http://www.apiezoncom/greasetablehtm February 10, 2005.

[45] T. Hogan, N. Ghelani, S. Loo, S. Sportouch, S.-J. Kim, D.-Y. Chung, M. G.
Kanatzidis, Proc. from the Int. Conf on Thermoelectrics, pg. 671-674, 1999.

[46] C. Wood, D. Zoltan, and G. Stapfer, “Measurement of Seebeck coefficient using a
light pulse”, Rev. of Sci. Inst, Amer. Inst. of Phys, Vol. 56, No. 5, pg. 719-722,
1985.

[47] Gao Min and D M Rowe, “A novel principle allowing rapid and accurate
measurement of a dimensionless thermoelectric figure of merit”, J. of Meas. of Sci.
and T ech., Vol. 12, pg. 1261-1262, 2001.

[48] T. A. Al-Obaidi and H. J. Goldsmid, “Determination of the Thermoelectric Figure
of Merit from Thermal Conductivity Measurements”, Energy Conversion, Vol. 9,
pg. 131-132,1969.

[49] T. J. Scheidemantel and J. V. Badding, Chemistry, Physics, and Materials Science

of Thermoelectric Materials, pg. 203-218, Kluwer Academic/ Plenum Publishers,
New York, 2003.

[50] Sim Loo, Jarrod Short, Kuei Fang Hsu, Mercouri Kanatzidis, Tim Hogan, “High
Temperature Measurement System Design for Thermoelectric Materials In Power
Generation Application”, Mat. Res. Soc. Sym. Proc., Vol. 793, pg. 375-3 83, 2004.

[51] T. Hogan and Tom Shih, “Modeling and Characterization of Power Generation
Modules Based on Bulk Materials”, CRC Handbook on Thermoelectrics, 2005.

[52] Gao Min and David M. Rowe, “Peltier Devices as Generators”, Chap. 38 in CRC

Handbook of Thermoelectrics edited by D. M. Rowe, CRC Press LLC, New York,
pg.485,1995.

146

 

[53] Heikes, R. R., Ure, R. W., Thermoelectricity: Science and Engineering, Interscience
Publishers, New York - London, pg. 512-513, 1961.

[54] Sim Loo, “Transport Characterization of New High Temperature Quaternary
Thermoelectric Materials And Devices”, Ph.D. thesis, Michigan State University,
2003.

[55] Atsushi Suzuki, Yasuhiro Hasegawa, Yoshiaki Ishikawa, Takashi Komine, Hiroyuki
Morita, and Hajime Shirai , “Investigation of physical and electric properties of
silver paints as a binder for thermoelectric materials”, Rev. of Sci. Inst., Vol. 76,

2005.
[56] H. E. Bates and Martin Weinstein, “ The Preparation and Properties of Segmented
Lead Telluride-Silicon-Gerrnanium Thermoelements”, IEEE Thermoelectrics Conf

Proc. (Fourth), 1966.

[5 7] C. Kittel, Chap. 4, Introduction to Solid State Physics, Seventh ed., John Wiley and
Sons, Inc., 1996.

147

   

ljlijjijjiji111]]