qufizm «mun.

uni-:91.
I.

m «mph.
“usfnvis... l

.33... w... .
: 4.11.1?31 .6... .
Au. 3! s It! all}?!
g, tug-1.4 ‘

3......
i .3-

Ir. ‘1
2111‘ a 1.! '41...
but. 2...}...

i..-

. a. 3.2.;

it!!! 3L3...

.5... 1.32.: 1 than:

iiummiést... i... s...
.

$4.m.4¢.z.l!!s....
:.... "33.33"

:1!!!»ng-

 

 

. (.9

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

THERMOELECTRIC MATERIAL CHARACTERIZATION AND
MODULE FABRICATION

presented by

Jonathan James D’Angelo

has been accepted towards fulfillment
of the requirements for the

Doctor of degree in Electrical Engineering
Philosophy

 

 

\

(p / '
Major Professor’s Signature

9-49 2007

 

Date

MSU is an Afiinnative Action/Equal Opportunity Employer

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5’08 K:/Proj/Acc&Pres/CIRCIDateDue.indd

 

 

THERMOELECTRIC MATERIAL CHARACTERIZATION AND MODULE FABRICATION

By

Jonathan Iames D'Angelo

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILSOPHY

Electrical Engineering

2009

ABSTRACT
THERMOELECTRIC MATERIAL CHARACTERIZATION AND MODULE FABRICATION
By

Jonathan James D'Angelo

Thermoelectric generators operate with no moving parts to convert heat flow to
electricity. Significant interest in developing new materials in recent years has led
to the discovery of several promising thermoelectrics; however there can be
considerable challenges in developing the materials into working devices. Testing
and feedback is needed at each step to gain valuable information for identification of
difficulties in; quality of the materials and modules, electrical contacts to the
materials, repeatability in fabrication, and longevity of the devices. The goal Of this

dissertation is to show results in all of these categories.

Acknowledgements
I would like to thank the generous support of the Office of Naval Research and
Strategic Environmental Research and Development Program. I would also like to
acknowledge the support and guidance given to me by my advisor, Dr. Timothy
Hogan. Lastly, I would like to say thank you to my mother and father; thank you

mom, this is for you.

iii

TABLE OF CONTENTS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LIST OF TABLES vi
LIST OF FIGURES vii
1.0 Introduction 1
1.1 The “Thermomagnetic Effect” 1
1.2 Thermoelectric Devices 2
1.3 Advantages of Thermoelectrics ............... 5
2.0 Literature Review 7
2.1 The Foundations of Thermoelectrics 7
2.2 Properties Of Good Thermoelectric Materials 8
2.3 B Factor; 3D and 2D 9
2.4 Large 2 value Contributions 15
2.5 Electronegativity, Mobility, and Weighted Mobility 16
2.6 BizTe3 17
2.7 IV-VI Semiconductors (PbTe) ...... 18
2.8 SiGe 18
2.9 Bismuth Antimony 19
2.10 Skutterudites 19
2.11 Clathrates 20
2.12 Half-Heusler Alloys 21
2.13 TAGS 21
2.14 Quantum Well, Supperlattice Materials 21
2.15 Quantum Dots 22
2.16 Electrical Conductivity 22
2.17 Thermopower, Mott Equation 23
2.18 Thermal Conductivity 24
2.19 Contact Resistance 26
2.20 Measuring Contact Resistance 28
3.0 Ultra High Temperature Measurement System (UHT) 31
3.1 Experimental Setup 31
3.2 System Characterization 32
3.3 N-Type Material Annealing Experiment 36
3.4 Ingot Uniformity 45
3.5 Hot Pressed UHT Results 47

 

iv

3.6 CAST to Hot Pressed Property Comparison ...... 49

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4.0 Diffusion Bonding 52
4.1 Diffusion Bonding Introduction 52
4.2 Diffusion Furnace 61
4.3 304 Stainless Steel Diffusion Bonds to LAST and LASTT Material ....................... 63
4.3.1 First Bond Experiment 63
4.3.2 Unicouple Fabrication with 304 Stainless Steel 64
4.4 410 Stainless Steel Diffusion Bonds to LAST and LASTT Material ....................... 76
4.5 Diffusion Bonds of Sno.3Feo.3Te to LAST and LASTT Materials 105
4.6 Diffusion Bonds of BizTe3 to LAST and LASTT Materials 106
4.7 Diffusion Bonds of Chromium, Iron and Indium Layers Between Nickel and
LAST and LASTT ' 110
4.8 Diffusion Bonds Of Copper Beryllium to LAST and LASTT Materials ................ 119
4.9 Diffusion Bonds of 316 Stainless Steel to LAST and LASTT Materials .............. 121
4.10 Diffusion Bonds Of Molybdenum tO LAST and LASTT Materials ....................... 179
4.11 Diffusion Bonds of 347 Stainless Steel to LAST and LASTT Materials ........... 179
4.12 Diffusion Bonds of ‘Maganin’ to LAST and LASTT Materials 179
5.0 Module Fabrication and Measurements 180
5.1 Long Term Module Testing System (LTMTS) 180
5.1.1 System Fabrication 180
5.1.2 Experimental Procedure 184
5.1.3 Results 187
5.2: LAST and LASTT Modules -- 192
6.0 Conclusions and Future Directions 225
6.1 Conclusions 225
6.2 The Next Step, Spark Plasma Sintering 226
References 228

 

LIST OF TABLES

Table 1: Common thermoelectric materials, maximum operating temperature,
maximum figure Of merit (Z), and their temperature of Zmax. 5

Table 2: Table of Energy Band gap and BEg values for typical thermoelectric
materials. 15

 

Table 3: U values and electronegativity differences for some thermoelectric
materials. 17

Table 4: Changes in thermoelectric properties of PbTe from brazed materials[61]. 58

 

 

 

Table 5: Stainless steel properties. 63
Table 6: Module AX output results. 197
Table 7: Bland BK segmented module results 207
Table 8: SnTe unicouple measurements. 224

 

vi

LIST OF FIGURES

 

Figure 1: Thermoelectric Device. 2

Figure 2: Thermoelectric heating device (E-Electrons in n—type leg, H-Holes in p-type
leg). . 3

 

 

Figure 3: Thermoelectric power generation device. 4

Figure 4: Z3DT versus reduced chemical potential. This plot shows the optimal
Chemical potential for a certain B factor Value to give a maximum Z3 DT. .......... 11

Figure 5: Reduced chemical potential as a function ofthe B factor ....... 12

 

Figure 6: Plot of Z3DT as a function of the 8 factor at optimal reduced chemical

 

 

 

 

 

 

 

 

potential. 12
Figure 7: B'vs. the optimal chemical potential... 14
Figure 8: ZZDT as a function OfB’. 15
Figure 9: Models ofthe density Ofstates for different dimensionalities. ....................... 24
Figure 10: Thermoelectric module power generation circuit. 26
Figure 11: Efficiency degradation due to contact resistance []. 28
Figure 12: Room temperature scanning probe. 29
Figure 13: N leg thermoelectric to 316 stainless steel scan. 30
Figure 14: Sample with mounting configuration. 32
Figure 15: Comparison of NIST data and system-measured data. 33

 

Figure 16: Percent error from NIST data and measured data of electrical

 

conductivity. 34
Figure 17: Measured thermopower and reference thermopower of platinum. .......... 35
Figure 18: Percent error between measured and reference thermopower. ................ 36

Figure 19: ETN207 sample 1 before the high temperature measurement, electrical
conductivity scan. 38

 

Figure 20: Electrical conductivity anneal at 700K. 39

 

vii

Figure 21: Thermopower anneal at 700K. 39

 

Figure 22: Scan of ETN207 after temperature cycling and measurement of
thermoelectric properties. RT o=544.68 (S/cm). 40

 

Figure 23: ETN207_1_423 room temperature scan after anneal at 423°C .................... 41

Figure 24: Room temperature electrical conductivity scan of ETN207_2_423. The

 

 

 

 

conductivities of each side of material are shown ....... 42
Figure 25: ETN207_3_423 room temperature scan 43
Figure 26: ETN207_4_423. - - - - 43
Figure 27: Room temperature scan of ETN207 as cast. 44

Figure 28: Plot of electrical conductivities ofsamples measured from each coin. 45
Figure 29: Plot of thermopowers measured from samples of each coin. ....................... 46
Figure 30: Plot of power factors measured from samples of each coin. ......................... 47

Figure 31: The Change in electrical conductivity while pumping down atmosphere.
48

 

Figure 32: Change in electrical conductivity Of HP12 during venting of system. ....... 48
Figure 33: Comparison of electrical conductivity Of cast and hot pressed material. 49
Figure 34: Comparison of thermopower of cast and hot pressed material. ................. 50

Figure 35: A schematic depiction Of a Creep Curve of strain versus time at a constant
stress, tr is the total time to rupture []. 55

 

Figure 36: Diffusion bond B will have better interface or contact resistance due to
the reduction of the size of the grains in the material. 56

 

Figure 37: Self-diffusion activation energy as a function of the melting
temperature[]. 60

 

Figure 38: Diffusion bonding furnace. 61

 

Figure 39: Diffusion furnace pumping system with computer controlled pressure
mechanism. 62

 

Figure 40: Unicouple Al N leg scan. 64

 

viii

 

 

Figure 41: Unicouple A1 P leg scan. ............ 65

 

 

 

 

Figure 42: Unicouple A2 N leg scan. 66
Figure 43: Unicouple A2 p leg scan. 67
Figure 44: Unicouple A3 N leg scan. ....... .- 68
Figure 45: Unicouple A3 P leg scan. 69
Figure 46: ZT measurement with and without solder. 70

 

 

Figure 47: ZT measurement with copper wire added. ZT dropped down to 0.0602
even though overall resistance of the unicouple was improved. 71

 

Figure 48: Unicouple A4 P leg scan, there are visible cracks along the length Of the

 

 

 

 

 

 

 

 

sample. 72
Figure 49: Unicouple run U N leg scan. .. 73
Figure 50: Unicouple run U P leg scan.- 74
Figure 51: Scan of N leg from unicouple run X. This shows a high contact resistance

of the n-type material to the 304 stainless steel. 75
Figure 52: Module Legs Cast in R300. 76
Figure 53: R300 module bonding setup. 77
Figure 54: Completed module, top view. 77
Figure 55: Completed module, bottom view. 77

 

Figure 56: This diffusion boat has a set screw for both n and p legs for during

 

 

 

 

 

diffusion bonding. 78
Figure 57: The unicouple on the left had the starting pressure of 420 Torr, and the

unicouple on the right had a starting pressure of 332 Torr. 79
Figure 58: Unicouples built with starting pressure of 450 Torr. 80
Figure 59: Side view of a 12 leg module. 80
Figure 60: Top view of 12 leg module. 81
Figure 61: Power Factor for ETN126. 82

 

ix

#9.... _l

 

Figure 62: shows the 14 unicouples bonded at 700°C for 4 hours. 82

 

 

 

 

 

 

 

Figure 63: Unicouple BZ voltage scan Of the N leg contact. - 83
Figure 64: Unicouple 82 P leg scan. ...... 84
Figure 65: Stage 1 of bond run D. ..... 85
Figure 66: Run D final unicouples. 85
Figure 67: Unicouple D1 n leg scan ............ 86
Figure 68: Unicouple D1 p leg scan ..... 87
Figure 69: Unicouple D5 p leg scan. 88
Figure 70: Unicouple run G stage 1 results. 89

 

Figure 71. Unicouple run G final results. The dot on the material marks the p-type
material. 89

 

Figure 72: Unicouple G1 contact scan of the p-type leg. 90

 

Figure 73: Unicouple G3 N leg contact scan. The ZT for this unicouple was 0.08..... 91

 

 

 

 

 

 

 

Figure 74: Unicouple G3 P leg contact scan. 92
Figure 75: Unicouple G4 N leg contact scan. 93
Figure 76: Unicouple G4 P leg contact scan. 94
Figure 77: Unicouple G5 N leg contact scan. 95
Figure 78: Unicouple G5 P leg scan. 96
Figure 79: Unicouple G6 N leg contact scan. 97
Figure 80: Unicouple G6 P leg contact scan. - - 98
Figure 81: Unicouple G7 N leg contact scan. 99

 

Figure 82: Unicouple G7 contact scan for the p-type leg. The measured ZT for this
unicouple was 0.04 100

 

Figure 83: Unicouples fabricated from run H 100

 

I“

 

Figure 84: Unicouple run Z, N leg scan. 103

 

 

 

 

 

 

 

 

 

 

Figure 85: Scan of N leg from run AU. 104
Figure 86: Bonding results of LAST and LASTT to Sn-Fe—Te alloy. 105
Figure 87: Voltage scan of LAST bonded to a SnFeTe alloy. 106
Figure 88: Segmented leg of LAST and BI2T63, measurement and data prepared by
Dr. Chun-l Wu. 108
Figure 89: Segmented leg of LASTT and BizTe3, measurement and data prepared by
Dr. Chun-I Wu. 109
Figure 90: Picture of failed bonds of BizTE3 to the LAST material. 110
Figure 91: Nickel to LAST bond with a thin film Of chromium-iron. 111
Figure 92: Contact scan ofunicouple run Q. 112
Figure 93: Unicouple run R contact scan. 1 13
Figure 94: Scans of LASTT bonded directly to nickel. 1 14
Figure 95: Contact scan Ofa bond from unicouple run T. 115

 

Figure 96: Experimental setup for bonding LAST/LASTT to nickel with an indium

 

 

foil. 1 16
Figure 97: Contact scan of the n-type leg from unicouple run BE. 117
Figure 98: Voltage scan of the n-type leg to an iron-chromium alloy ............................ 118

Figure 99: Photo of the failed diffusion bond of LAST and LASTT to a copper

 

 

beryllium alloy. 119
Figure 100: Setup of the legs before the deposition of W and Au. 120
Figure 101: Picture of setup after deposition with gold coating. 120

 

Figure 102: Results of bonding experiment after adding gold and tungsten to the

 

 

copper beryllium alloy. 121
Figure 103: Unicouple run AA N leg scan 122
Figure 104: Unicouples for run AA. 122

 

xi

 

_

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 105: Unicouple run AB N leg scan. 123
Figure 106: Unicouple AB P leg scan. A 124
Figure 107: Unicouple Run AC N leg scan. 125
Figure 108: Unicouple AC P leg scan. 126
Figure 109: Unicouple 2 from AC run N leg scan. 127
Figure 110: N leg scan from a unicouple run AC. ....128
Figure 111: P leg scan from unicouple run AC. 129
Figure 112: Unicouple run AD N leg scan. ..... 130
Figure 113: Module legs after a sputtering deposition of 316 stainless steel ............ 131
Figure 114: Scan of copper soldered to LAST with a thin film Of 316 sputtered to the
LAST. 131
Figure 115: Scan ofNi-316-LAST bond. 133
Figure 116: Bonds of 0.02" thick 316 stainless steel to LAST material. ....................... 134
Figure 117: Scan of unicouple run AG N leg to 316. 134
Figure 118 Scan of Cu-316-LASTT bond 135
Figure 119: Picture Of bonded hot pressed module. 136
Figure 120: Scan Of hot pressed leg before and after bonding. 137
Figure 121: Fractured legs bonded to 316 stainless steel. 138
Figure 122: Scan of P leg contact to 316 stainless steel. 139
Figure 123: Crack of leg with high contact resistance. 140
Figure 124: Scan of p-type hot pressed material bonded to 316 stainless steel ....... 141
Figure 125: Scan of N leg bond to tin-tellurium coated 316 stainless steel. ............... 142
Figure 126: Scan of MSUHP-Z to 316 stainless steel. 143
Figure 127: Contact scan of P leg to 0.02" stainless steel. 144

 

xii

Figure 128: Photograph of new bonding device.

 

145

Figure 129: Scan of P leg bonded to 316 stainless steel with new bonding device. 146

Figure 130: Unicouples bonded from runs AT and AS.

 

Figure 131: Fractured sample from run AT. ......

 

Figure 132: Scan of ETN174 bonded to 316 stainless steel. The horizontal line

represents a contact resistance of ZOqucmZ.

 

146

147

148

Figure 133: Photograph ofbond of LAST to 316 with a soak temperature Of 800°C.

 

Figure 134: Scan of P leg from unicouple run AW

 

Figure 135: N leg scan of bonding run AW.

 

Figure 136: Scan ofhot pressed 316 and LAST.

 

Figure 137 Scan of N leg contact scan from run AX

 

 

Figure 138: EDS maps of unicouple run AX bond.

Figure 139: Unicouple run BG P leg scan.

 

Figure 140 BG sample 2 P leg scan

 

 

Figure 141 Unicouple run BG sample 2 N leg scan.

Figure 142: Unicouple run BH N leg scan.

 

Figure 143: Unicouple run BH P leg scan.

 

Figure 144: Three p-types with 316 stainless bonded to both ends.

Figure 145 Unicouples fabricated from run B].

 

 

Figure 146: 811 P leg scan.

 

 

Figure 147: 811 N leg scan.

Figure 148: 312 P leg scan.

 

Figure 149: B12 N leg scan.

 

xiii

149

150

151

152

153

154

156

157

158

159

160

161

161

162

163

164

165

 

Figure 150: BB N leg scan. 166

 

 

 

 

 

Figure 151: 813 P leg scan. 167
Figure 152: BJ4 P leg scan. 168
Figure 153: 814 P leg scan. 169
Figure 154: The unicouples were nickel plated for preparation to be coated with
QS/B4. 170
Figure 155: Head piece ofspring loaded bonding device. 171
Figure 156: Long view of spring load device from main head down shaft. ................. 171

Figure 157: Base mechanism of spring loaded device which houses the spring. ..... 172

Figure 158: Scan of p-type contact to 316 stainless steel with spring loaded bonding
device. 173

 

Figure 159 Continued: EDS of LASTT bonded to 316 stainless steel before annealing.
175

 

Figure 160 Continued: EDS of LASTT to 316 after an anneal at 620°C for 4 days. ..176

Figure 161: Post anneal scans of LASTT to 316 contact. 177

 

Figure 162: Before and after scans Of quenching experiment. 178

 

Figure 163: Long Term Module Testing System. A) Gas port; B) KF-40 quick connect;
C) ISO Flange; D) Main Chamber; E) Chiller feed throughs for cold plate; F)

 

 

Vacuum Port. 180
Figure 164: Schematic drawing of long-term module testing system. .......................... 181
Figure 165: Heater Resistance as a function of temperature. 182
Figure 166: Nickel-Chromium meandering line heater 182

 

Figure 167: Complete system. A) Nano Voltmeter; B) Main Chamber; C) Current
Meter; D) Temperature Monitor; E) Power Supply; F) Chiller; G) Argon Gas
Cyhnden 184

 

Figure 168: Test Setup. A)Thermocouple; B)Bolt heater connection; C)Nickel-
chromium heater; D)Thermoelectric module E)Module leads to feed-through
solder connection; F)Thermally grounded leads. 185

 

xiv

Figure 169: Power to hot side heater and corresponding temperature. Note: At
heater power 30Watts the chiller temperature was increased from 30°C to
35°C. All temperatures are measured under an open circuit condition. .......... 186

Figure 170: Comparison of voltage measured by the LTMTS and published Tellurex
results 187

 

Figure 171: Comparison ofthe current measured by the LTMTS and Tellurex ........ 188
Figure 172: Comparison ofthe power measured by the LTMTS and Tellurex .......... 189

Figure 173: Hot side and cold side temperatures of module during month long test.
After one week of operation the heater power was increased to raise the hot
side temperature, and then it was maintained for the duration of the
experiment. 190

 

Figure 174: Power output of the Tellurex module over one month of operation....190

 

 

 

 

Figure 175: Hand made heater being tested in the LTMTS. 191
Figure 176: Top View ofheater in LTMTS. 192
Figure 177: Front view of experimental setup of heater in the LTMTS. ....................... 192
Figure 178: Completed cold side contacts Of chromium to aluminum nitride. ......... 193
Figure 179: Contact with PbSn solder. 194
Figure 180: Modules fabricated from unicouples AX and AW. 195
Figure 181: Test set-up for module power output test. 195

 

Figure 182: Mobile module testing setup. This set up can be used in the lab and also

 

 

 

 

 

for public demonstrations. 198
Figure 183: Unicouple BF built into module. 199
Figure 184: Unicouple module BG2. 200
Figure 185: Unicouple module BG2 front View. 200
Figure 186: Picture of experimental setup. 201
Figure 187: Reading from current sensor (1 volt is a reading of 10 amps). ............... 201
Figure 188: Hot side temperature reading. 202

 

XV

Figure 189: Cold side temperature reading. 202

 

 

 

Figure 190: Cold side contact for a 36 leg module. 202
Figure 191: 32 leg module. 203
Figure 192: The electrical resistance was tested and it was an open circuit. ............ 203

Figure 193: The red marks show were there is an electrical connection between

 

 

 

unicouples. 204
Figure 194: Photographs ofunicouples taken out ofthe 32 leg module. ..................... 205
Figure 195: Unicouples built from runs B] and BK. 206
Figure 196: Completed module. 206
Figure 197: Module after one week ofoperation. 207

 

Figure 198: Scan of N leg ofa module that ran for one week under a 585°C gradient.
208

 

Figure 199: EDS of contact. Top is LAST, then solder junction (Bi-Sn-Sb), Bottom

 

 

BizTe3. 209
Figure 200: EDS ofsame junction as figure above, but at a higher magnification, the

picture shows an area approximately 200nm across. 210
Figure 201: Module before drop testing. 211

 

Figure 202: After the module dropped 9' the base aluminum nitride cracked. ........ 212

 

 

 

 

 

Figure 203: Module built with solder for contacts. 212
Figure 204: Cold side substrate for the module called Ghost Rider. 214
Figure 205: Cold side electrodes soldered to cold side substrate. 215
Figure 206: Front view of Ghost Rider’s legs soldered to cold side 215
Figure 207: Thermoelectric legs soldered down to cold side. 216
Figure 208 Ghost Rider 216

 

xvi

Figure 209: Ghost Rider output voltage and output current for a temperature
gradient across the module of AT 2 184°C.

 

Figure 210: Ghost Rider l-V Characteristics .......................................

 

 

Figure 211: Plot of the resistance compared to the AT.

Figure 212: Module "Jen" built with cast materials and coated with QS/B4. ..........

Figure 213: Output voltage and current for the four leg module “Jen” with a

 

temperature gradient of 589°C.

Figure 214: Module “Jen”, side view.

 

Figure 215: Module “Ien” rear view.

 

Figure 216: Sixteen leg module called "Tape".

 

Figure 217: Four leg module using nickel electrodes and titled "0001”. ..................

Figure 218: Module 7-09-08-2Leg.

 

Figure 219: Nuraddin's SnTe unicouple after running in the LTMTS.

 

xvii

216

217

217

218

219

219

220

221

222

223

224

1.0 Introduction

1.1 The “Thermomagnetic Effect”

In early 1820, Thomas Seebeck was experimenting with the possibility Of a
relationship between heat and electricity [1]. His experimental setup was a current
loop formed by the junctions at both ends ofa copper and a bismuth wire. Seebeck
heated up one of the copper-bismuth junctions while the other remained at room
temperature. What he measured was a magnetic field radiating from the wire. He
believed that the temperature gradient was causing the magnetic field. He called
this the “Thermomagnetic Effect." This is was wrong. The magnetic field that he
measured was produce by the current flowing due to the temperature gradient in
the current loop, however the relationship between current flow and magnetic field
was not known at that time. lfone ofthe wires is cut, then no magnetic field results;
instead a voltage can be measured at the open circuit terminals. This voltage is the
Seebeck coefficient for the junction formed by those metals, and is a temperature
integral sum of the thermopowers (or absolute Seebeck coefficients) of each metal.

In 1834 when another scientist by the name of Peltier discovered that when
electricity is forced to flow through a junction of dissimilar metals, heat would
either be absorbed or liberated at the junction based on the direction of current flow
[2]-

Later Thomson showed that heating or cooling along the length ofa single

material also takes place when electrical current flows through the material. This

became known as the Thomson Effect. Today's thermoelectric devices are based on

these scientific principles.

1.2 Thermoelectric Devices

Metal Contacts

n g r;.
“f ‘1 . <1" . I . 2 ‘l:"
. ‘1» "H‘ i r. :' M C“

”I

   

Figure 1: Thermoelectric Device.

Figure 1 shows a single junction (unicouple) between n-type and p—type
thermoelectric material in a standard thermoelectric device. Both the n-type and p-
type materials are doped to degeneracy and form ohmic contacts to the electrodes;
thus although it is an n-type to p-type structure it does not form a diode and these

devices exhibit linear current vs. voltage profiles.

Heat

Current |+

Figure 2: Thermoelectric heating device (E-Electrons in n—type leg, H—Holes in p-type
leg).

By supplying electrical power to this device as shown in Figure 2 heat can be
pumped from one side of the device to the other to establish a temperature gradient.
Heat will flow through the p-type leg in the same direction as the electrical current
flow through that leg. The carriers, or holes, in the p—type material move in the
direction of current flow and also carry heat from the bottom of the device to the
top. The electrons on the n-type side are also moving in the same upwards direction
along the device, also carrying heat. This makes the top side ofthe device hot and
cools the bottom of the device as shown in Figure 2. With this same layout and
reversing the current direction, the top of the device will then be cooled following

the same principles.

Heat

(E) G)
l l

+

Figure 3: Thermoelectric power generation device.

These devices, thus serve as heat pumps. If instead of supplying electrical power to
the device in order to create a temperature gradient, a temperature gradient is
supplied to the device, then electrical power can be generated. The larger the
temperature gradient the more voltage produced by the module. A portion Of the
heat supplied to the module is carried to the cold side of the device by the electrons
in the n-type and the holes in the p-type. This makes one side a positive potential
and the other a negative, and can be attached to an electrical circuit and be used as a
power supply. There are a number of advantages for using thermoelectric devices
as listed in the next section, however the most significant disadvantage is the
relatively low efficiency in thermal to electric energy conversion (typically 5% for

Bi2Te3 based devices).

1.3 Advantages of Thermoelectrics

The science ofthermoelectrics was born some time ago, now a number Of

companies are based on their fabrication and application. Most thermoelectrics are

used for cooling and heating devices. A list of advantages includes[3];

1.

2.

5.

6.

No moving parts for long device lifetimes and little maintenance

They can be made into small or large devices without loss Ofefficiency
The same module can be used as a power generator and a cooler

They can Operate in any orientation

They give very little electronic noise

They can be used for accurate temperature control

Common thermoelectric materials and their operating temperatures are given in

Table 1.

Table 1: Common thermoelectric materials, maximum Operating temperature,

maximum figure of merit (Z), and their temperature of Zmax.

 

Zmax (K'l) USBfUI Range TZmax (K)

 

 

 

Bi2Te3 3x10'3 < 500 K 300
PbTe 1.7x10'3 < 900 K 650
Si-Ge 1x10'3 < 1300 K 1100

 

 

 

 

 

 

The figure of merit, Z, is used to evaluate thermoelectric materials, has units of

(1 / K) and is based on the material properties of electrical conductivity, Seebeck

coefficient (or thermopower), and thermal conductivity. The unitless figure Of

merit, Z'I‘is given by

ZT=————-T (1)

where Sis the thermopower, o is the electrical conductivity, K is thermal
conductivity, and Tis temperature. To obtain a high ZT the thermopower and
electrical conductivity must be high, and the thermal conductivity should below.
Maximizing ZT can be challenging as these materials properties are not independent
ofone another. Usually a material that has a higher electrical conductivity will have
a low thermopower, and a high thermal conductivity. This behavior is seen in most
conductors, such as copper. For insulators, such as glass, a low thermal conductivity
and a high thermopower is typically accompanied by a very low electrical
conductivity. Optimization Of the three material properties is generally found in
heavily doped semiconductors [4].

Modules that use bismuth telluride as their thermoelectric materials Operate
near room temperature, and solder contacts can be used to bond both the hot and
cold side electrical-mechanical junctions. The devices built here at Michigan State
University need to have bonds on the hot junctions that can withstand rigorous
temperature cycling and high temperature operations where solder contacts are not
possible. The Chemistry Department at Michigan State University has discovered

several chalcogenide based materials that have shown promising thermoelectric

properties [5]. Recently compounds in the Agl-bemeTe2+m and

Ag(Pb1-xSnx)meTez+m systems have exhibited outstanding thermoelectric

properties for power generation applications [6].

 

 

7

 

Lead antimony silver tellurium or also know as LAST and lead-antimony-silver-tin-
tellurium, or LASTT, are the n-type and p-type materials (respectively) used for

fabrication of the thermoelectric devices. The purpose of this dissertation is to give
the research of both high and low temperature electrical contacts Of metals to LAST
materials. The contact resistance drops the efficiency ofthe device and thus should

be minimized.

2.0 Literature Review

2.1 The Foundations of Thermoelectrics

For fabrication of thermoelectric refrigeration devices there are many important
contributions for the development of the modules. Metals with the highest
thermopowers are needed. These are antimony and bismuth alloys [7].

There is a great importance of the electrical contacts at the junctions. Large
contact resistances between the thermoelectric and the metal electrode are
detrimental to the efficiency of the thermoelectric module.

The minimization of the thermal conductivity is important. It has an
electronic and lattice component. The electronic contribution can be small for
insulator, and lightly doped semiconductors. For metals and highly doped
semiconductors the electronic contribution is amplified. The lattice component
comes from phonon scattering in the lattice.

For materials to have the highest figure of merit possible it is desirable to
have a ratio Of the carrier mobility to the thermal conductivity to be large, and the

effective mass of the carriers to be as large as possible. Therefore semiconductors

7

 

used for thermoelectric refrigeration should be chosen with the highest mean
atomic weights for reduced lattice thermal conductivity contributions, and with

large effective masses of the charge carriers. Goldsmid suggested in 1954 that

Bi2Te3 as a potential candidate for thermoelectric devices and it is still used today.

The highest ZT for room temperature materials was seen in

(Bi2Te3)o.25(Sb2Te3)(szSe3)0.o3 [8]. This alloy has a z of3.4x10'3K'1 and a room

temperature resistivity of1.1m£2cm.

2.2 Properties of Good Thermoelectric Materials [9]

Thermoelectrics (TE) need electronic bands near the Fermi level with many
valleys preferably away from the Brillouin zone boundaries to improve
thermopower and mobility. Such high degeneracy is typically found in crystals with
high symmetry. TE's elements with large atomic numbers and large spin-orbit
coupling helps to lower the overall thermal conductivity. Compositions with more
than two elements also lowers the thermal conductivity. A low average
electronegativity difference between the elements improves mobility. A large unit

cell with many atoms will lower the thermal conductivity. Having an energy gap

equal to about 10kBT, where Tis the operating temperature of the thermoelectric,

and k3 is Boltzman’s constant improves the thermopower. The band gap for room

temperature materials should range from 0 to 0.3eV.

2.3 B Factor; 3D and 2D

Chasmar and Stratton first introduced the B parameter in 1959 [1 O]. The B

parameter is directly proportional to the BEg, which is the energy gap divided by the

thermal energy. ZT=F(B,BEg), where fl=1/kBT.

 

1 2k T 3” k2
B = y (i) mxmym: '_B'qu (2)

3x2 122 ex,

where m is the effective mass in the its corresponding crystallographic direction, px

is the mobility in the x direction, or it can be set to the direction of majority
transport, and y is the degeneracy of the bands for the carries in the material.

A high B factor translates to a high figure of merit. In the formula shown
above, a highly anisotropic material can be utilized if the highest mobility direction
can be chosen to correspond to the electrical current flow direction. This approach
may counter attempts to increase the B factor through large band degeneracy, which
is commonly found in highly symmetric crystal structures, for example cubic. It is

also beneficial to have many atoms in the unit cell. Highly anisotropic effective

masses are not found in such materials. Bi2Te3 has a band degeneracy Ofsix.

Below is shown a calculation of Z through the Chemical potential and B factor
for the 3D case [11]. The following assumption are made for the calculations; a
general, anisotropic, one-band material, constant relaxation times, and parabolic

bands

hzkfi hzkz2
+
2m

 

Ek k k -h2k3 3
(mm , w

is the energy for parabolic bands.

The electrical conductivity is defined as,

0=—— ;2(2k——B——T)y2(mx m ymz)'l/é F/qutx (4)

The thermopower and electronic contribution ofthe thermal conductivity are as

 

 

 

follows;
F
S=_£ll §_/_2__§* (5)
‘16 3 Fl/
/2
2 5/2 1/2 25F2
Th ZkBT mvmz 7 3/2
Ke=-—— 2 ' XkB —F5//2— (6)
h mx 2 6Fl/2
0" i
F.- =E(c*)=f—x.—dx (7)
Pg +1

5 = 7}? is the reduced Chemical potential relative to the edge of the conduction
B

band, qe is the magnitude ofthe electronic charge [1.602 x 10’19 (C)], mx,my,mZ are

the effective mass components, and ,ux is the mobility in the x—direction. This shows

the thermopower gives a measure Of the reduced chemical potential, which can be

used in conjunction with measured electrical conductivity to gain insight on the

I
l mymZ /2
factorof px(mxmymz 2=qu m
X

Therefore Z is given as;

 

where T is the relaxation time.

2
Z=_;g_‘7_ (3)
KC +Kph

10

.f_

 

 

.z.D7‘= (9)
‘ 2wfi2

— 6"] #2

 

I 7
—+—F.
8 25¢

This calculation does not take degeneracy into account.

 

 

 

 

3 L . ...... . r r . . . - r r . . .
I- ll
25: Srrxh 3
. .0 . .
0 °. .
I: 0. .9 :
2 :- .° 0. _
I" l- o: .0. I
. o ' 2
g 1.5 _- .' -;
N C .0. ... II
. .- ;* °- :
1 :' 0. Opt 1
3 l
- 1
0.5 - ..
- 1
I v 1
O ......... l .........
-5 0 5

Figure 4: Z3DT versus reduced chemical potential. This plot shows the optimal
chemical potential for a certain B factor Value to give a maximum Z3 DT.

11

 

 

 

0..
2 _ . .
- o
’ \
1 _. \ -
l- \ I
' O, u
.. 0 .. -
e ' . -
I \ I
'3 I- \ .
JJ‘ -1 '_ .2
. o. -
I- \
P
- \
-2 _ . -
l' \- -
C \. I
I \ H
I \ u
4 D i I

0.001 0.01 0.1 1 10 100
B

Figure 5: Reduced chemical potential as a function ofthe B factor.

 

100 = i I 11"!“ r1 IIIIII' f I’ lll’ml t I IIIIII’I I TT‘I’II:
10 r '. 1
I . ,0" E
E l
n .
I\
<9: 1 :- , 1
N -
I ‘ :
. / .
L / .
0.1 ? fl/ 1
= i
4 '1
/ u
Ill-l I l lllllll l 1 1111111 1 l llllll

 

 

 

O_O1 . . ......I . . ..
0.001 0.01 0.1 1 10 100
B

Figure 6: Plot of 2307' as a function of the B factor at optimal reduced chemical
potential.

12

f

A proposed method to increase OfZTin some materials is by preparing them
in the form ofa multilayered super lattice [11]. These structures will increase Z due
to the fact that the electrons are now confined to move in two dimensions. This also
may reduce the phonon thermal conductivity and therefore increase Zwith the
interface scattering of the phonons.

For the 2D case the energy band functions are as follows;

hzkg + LIZ/<3 + l'IZJr2

 

 

 

E k ,k , = (10)
( x ) 2mx 2’72), 2nzza2
The electrical conductivity, thermopower and the electronic contribution to the
thermal conductivity are;
2k T 3/2
a =—'—(h, B ) (m m ) Emu. (11)
San/egg) (12)
qe F0
2 5/2 1/2 2
Th ZkBT my 4Fl
Ke=——— ——'-' k8 3F2———-— (13)
4.750 h2 mx F0

The reduced Chemical potential in the 2D case is;

'12” 2
C [=C-2 2)/k37 (14)
2% 02

 

 

2
2F *
(f-C) F0
22DT= 0 (15)
Lug—55—
F0

13

 

 

. I ZkBT 1.231“
B =21ra( hz )Jmey q—K——,ux (16)

ph
The reduced chemical potential now can be varied by both doping and changing the
layer a Of the quantum well structure. Recall that in the 3D case, the only means to
increase B in an anisotropic crystal is to choose the current direction to be the

direction of the highest mobility. The situation is more complex for the 2D quantum

well. Z20T not only can be Optimized by choosing the best current direction, but

also by narrowing layers and choosing the best orientation to make the layers. A

layer thickness of 10A is optimal for Bi2Te3 materials.

 

1 I I U'Ill'l I I U1'U'I' r I 1“

 

 

 

- -1
0 ‘i- 1'
" T
I- J
- 1
u -1 :- i‘.‘\ j
0.
O ' q
-k 1- \ -1
OJ . \\
-2 _ _
Z \R I
-3 - \ -
I '\
: :
4 - l l j lJllll I I l Ill-ll l l I l‘jll‘
0 1 1 10 100

B!

Figure 7: B'vs. the optimal chemical potential.

14

 

1O I I I "I‘Il I’ I I IIIT'I’ ‘ '1 'Y'I'

 

 

0.1 l n n llllll 1 j 1 nannl - n n lllll
0.1 1 10 100

B!

 

Figure 8: ZznT as a function OfB'.

2.4 Large Z value Contributions
Materials with a large effective mass, high carrier mobility, and low lattice

thermal conductivity will yield high Zvalues. Semiconductors with an energy gap Of

10kBTis good for an efficient thermoelectric, BEg~10 [12].

Table 2: Table of Energy Band gap and BEg values for typical thermoelectric

 

 

 

 

 

 

materials.
Material Eg(eV) Temp of Operation (K) Eg/ka
Bi2Te3 0.16 300 9.7
PbTe 0.5 650 8.9
AngTez 0.3 650 5.4
GeTe 0.4 700 6.6
SiGe 0.7 1100 7.4

 

 

 

 

 

 

15

2.5 Electronegativity, Mobility, and Weighted Mobility

As discussed earlier, mobility is a contribution to improving the figure of
merit of thermoelectric materials. High mobilities come from direct bandgap
semiconductors, but the effective mass is small. A small effective mass will reduce
the B factor. An indirect band gap material would be ideal for a thermoelectric
material.

Goldsmid and Mahan stressed the importance of the weighted mobility [13].

:1: 3’
U=,u[m /mo]2 (17)

The weighted mobility, U, is defined by m*, the effective density Ofstates mass for

the conduction band, mg, the free electron mass, and p, the mobility. Better

thermoelectrics exhibit large weighted mobilities. Uvalues of 100 or above are
typical for thermoelectrics. Compounds with U values of 100 or above typically have
an average electronegativity below 0.5. Mobility is high in compounds where there
is no charge motion associated with phonons in purely covalent lattice. In
compound semiconductors made of two or more elements with differing
elecronegativity values there is an electron transfer between the ions. The amount
of Charge transfer increases with the electronegativity difference. Therefore it key
to have low average electronegativities for thermoelectric materials made with

multiple elements.

16

 

Table 3: U values and electronegativity differences for some thermoelectric

 

 

 

 

materials.
U value (cmz/V°sec) Electronegativity Difference (AX)
n-type p-type
PbTe 200 79 0.53
BIZTE3 406 608 0.38

 

 

 

 

 

 

2.6 Bi2Te3 [14]

Bismuth Telluride, and its alloys exhibit the best ZT near room temperature.
Bismuth telluride's melting point is 585°C. When the percentage composition Of

tellurium exceeds 62.5%, it goes from being a p-type material to n-type material.

Bismuth telluride is usually alloyed with szTe3, which has the same crystal

structure. There is a S1 alloy, Bi35b32Te60, and a 82 alloy, Bi1oSb3oTe60. The 51
alloy has no ZT change if transport is measured parallel or perpendicular to the
cleavage plane. This is not the case for the 52 alloy, which has a higher ZT on the

plane parallel to the cleavagee plane. The n-type solid solution optimal figure Of

merit carrier concentration is 2x1019cm'3. The p-type solid solution optimal figure

of merit carrier concentration is 4x1019cm'3. The alloying of the bismuth telluride

to the antimony telluride can reduce the room temperature thermal conductivity as
low as 1W/m'K. The low thermal conductivity is the main contribution for it high

ZT. Bismuth telluride's both n-type and p-type materials have equally good

17

thermoelectric properties. Its crystal structure is considered to be rhombohedric,
but it is easier to represent it as a layered crystal with hexagonal symmetry. Both
the conduction and valance bands have six equivalent minima. The energy bands

are nonparabolic, and the effective masses are temperature dependent. The band

gap for undoped Bi2Te3 is 0.16eV.

2.7 IV-VI Semiconductors (PbTe) [15]
Group lV-VI semiconductors all have a N aCI crystal structure. PbTe has a

maximum ZT around 700K. The room temperature band gap is 0.31eV. The
standard alloy for lead telluride is Pb1-xSnxTe. With x >0.25 there is a ferroelectric

transition of the alloy at 100K, which is a spontaneous electric polarization of a
material. The composition with the highest figure of merit is when x=0.25. Lead
telluride’s ZT exceeds 1 at 800K. The use Of a few tenths of 1% of sodium can make

p-type lead telluride. Lead telluride is the base composition of the LAST materials.

2.8 SiGe [16,17]

Silicon has one of the highest weighted mobilities at room
temperature, however is also has a large value of thermal conductivity, 100W/m0K
at room temperature, which decreases with increasing temperature. It was then
later discovered that alloying silicon with germanium decreases the thermal
conductivity by a large amount. SiGe's minimum resistivity is at 1100K, which is

also the point of its maximum ZT. Silicon germanium ingots are fabricated by zone

18

leveling, which is time consuming, or by hot pressing. The highest ZT mixture is

Si0.7Ge0.3 and boron and phosphorus are used for doping.

It was first used in space for the SNAP-10A nuclear reactor, and has been the
preferred space born thermoelectric since 1976. Silicon germanium Operates at
temperatures up to 1300K for years without degradation of its properties. It has
accumulated over 250 million device hours without a single known failure.

With silicon germaniums high operating temperatures, thermal expansion
issues, Silicon germanium is 3-4 times stronger in compression than in tension.

Therefore contact materials should have a larger CTE than Si-Ge.

2.9 Bismuth Antimony [18,19]

Bismuth antimony is a semimetal with a rhombohedral structure and a

carrier concentration Of1018cm'3. It has six electron ellipsoids and is a multivalley

material. Alloying bismuth antimony with group IV materials will add donors or
acceptors, and modify the band structure. Alloying with group V materials can
modify the band structure and Change the carrier concentration. The ZTis 0.6 at

room temperature with around 10-15% antimony.

2.10 Skutterudites [2 0-2 3]

Skutterudites are a natural occurring material that was first discovered in
Skutterud, Norway. It has a cubic crystal structure with 32 atoms per unit cell. The
atoms are arranged in an open, cage like structure. Atoms can be placed inside

these cages to reduce the thermal conductivity.

19

Skutterudites meet the basic conditions for a high ZT. They have a large unit
cell, heavy constituent atom masses, a low electronegativity difference between
atoms, and have the highest reported mobilities among all know materials. Filling
the voids in the cage like structures has been the focus of efforts to improve the
thermoelectric properties. Lanthanum, cerium, and thallium partially filled
Skutterudites have been reported to possess relatively good electronic properties

while also possessing much lower thermal conductivites than the unfilled cages of

CoSb3.

2.11 Clathrates [24,25]

Transport properties of semiconductor type-I Si and Ge Clathrates reveal low
thermal conductivity values along with relatively high Seebeck coefficients and
electrical conductivity. The crystal structure of these clathrate compounds consists
of a framework of group-IV atoms having several polyhedral cages, each of which
can incorporate relatively large guest atoms, not unlike Skutterudites. Thermal
conductivities of these crystalline compounds are similar in temperature
dependences and magnitudes to that of amorphous materials. The experimental
trends in thermal conductivity are similar to those found in Skutterudites. These are
semiconductors with good electron properties that can be synthesized with a

glasslike thermal conductance.

Group IV elements can also form Clathrates of type X8E26, where E

represents a group IV element and X represents an atom in the voids formed by

20

Ar

elements E. The X atoms do not enter the lattice substitutionally but are interstitial,

residing inside the oversize cage.

2.12 Half-Heusler Alloys [26]
Half-Heusler alloys have a MNiSn composition where M is a group IV

transition metal (Hf, Zr, or Ti). The alloys have a band gap of 0.1-0.5eV. Room

temperature values of power factor are comparably to that of Bi2Te3, however the

room temperature thermal conductivity is approximatley 10 W/m°K. They have a
M gAgAs type crystal structure, forming three interpenetrating face centered cubic

sublattices with one Ni sublattice vacant.

2.13 TAGS [27]

TAGS is GeTe alloyed with 10-20% AngTez. TAGS has a very low lattice

thermal conductivity. The two most well known TAGS compounds are called TAGS-
80 and TAGS-85. The number represents the percentage of GeTe. TAGS-80 has a ZT

of 1.85 at 800K. TAGS-85 has a ZT of 1.35 at 800K. TAGS has been used in

numerous space applications, for example pioneer 10 and 11.

2.14 Quantum Well, Supperlattice Materials [28,29,30]

As discussed earlier, supperlattice (SL) materials are phonon
blocking/electron transmitting superlattices. Structures utilize the acoustic
mismatch between the superlattice components to reduce the lattice thermal
conductivity, rather than using the alloying approach which can scatter carriers as

well as phonons. Layers as thick as 10A Offer significantly higher in-plane carrier

21

 

 

mobilities parallel to the superlattice interfaces, than alloys, due to the near absence

of alloy scattering and random interface carrier scattering.

P—type Bi2Te3/Sb2Te3 SL structure have a ZTof 2.4 with a 32.2K gradient

across the module. The enhancement achieved by controlling the transport of
phonons and electrons in the superlattice. The thickness of the heterojunction
layers influences the mobility of the samples.

A type one quantum well (QW) is a semiconductor with a narrow banggap
sandwiched between two materials with larger band gaps. Impurities are put in the
barrier materials so the conducting layer is free of impurities, unlike in bulk
materials where impurities cause scattering. These devices have promising

applications but are very expensive and difficult to fabricate.

2.15 Quantum Dots [31]

Quantum dot supperlattice structures have a delta function distribution of
the density of states, and discrete energy levels due to three-dimensional quantum
confinement. This is a more favorable carrier scattering mechanism, and a much
lower lattice thermal conductivity. Devices have been grown by molecular beam

epitaxy, with compositions of PbSeTe/PbTe.

2.16 Electrical Conductivity
The electrical conductivity has a direct relation to the dopant concentration

Of a semiconductor material.

0=q(#n‘n+up-p) (18)

22

In the above equation, n and p are the dopant concentrations ofa semiconductor

material. The concentration of dopants in LAST materials is around 1019 cm6. At

this concentration, the sample is degenerately doped and shows metallic
temperature dependence for the electrical conductivity (decreasing o with

increasing temperature). We have experimentally found electrical conductivity

values in the range of800 t01500 S-cm'1 to correspond to some ofthe highest ZT

samples for the LAST and LASTT materials we are investigating.

2.17 Thermopower, Mott Equation

The thermopower ofa device, or also known as the absolute Seebeck
coefficient is the open circuit voltage ofa semiconductor under the influence ofa
temperature gradient. The thermopower ofa material is a measure ofthe local
entropy, and is influenced by effective masses and the energy dependence ofthe

density of states. For LAST materials we have experimentally found values near

100p.V-K'1 at room temperature tend to yield the highest ZT materials. When

adjusting the composition and doping of LAST, a reduction in the carrier
concentration causes an increase in the thermopower, however also corresponds to

a decrease in the electrical conductivity.

_ Jrzsz _ dlna(E)
- 3e dE

S

 

|E=Ef (19)

The equation above is the Boltzmann transport theory equation for thermopower,
the Mott equation [32]. The electrical conductivity in the equation is a function of

the Fermi energy. The thermal power is a measure of the asymmetry in electronic

23

structure and scattering rates near the Fermi level, finding materials with electrical

conductivities that have strong asymmetrical energy dependence in a small energy

interval near Ef is a way to approach high thermopower. This has lead to searching

for materials with complex structures and compositions. Fabricating materials with

2D, 1D, and 0D structures ofthe density Ofstates give a large value in dln0(E)/dE.

   

n E) "03) “(El HUS)

 

 

 

 

E , E . - E
Bulk (3D) B Quantum Well (2D) Quantum Wire (10) Quantum Dot (on)

Figure 9: Models of the density of states for different dimensionalities.

2.18 Thermal Conductivity
The thermal conductivity is the measurement of a materials ability to
conduct heat. From the Z T equation it can be seen that a low thermal conductivity

material helps to increase the figure of merit. For the materials studied in this
thesis, a thermal conductivity near 1W-m‘1-K’1 is commonly seen for the high ZT
samples. There are two contributions tO thermal conductivity.

24

K = Ke + K, (20)

The first part is the electronic contribution and the second part is the lattice (or
phonon) contribution to the thermal conductivity. For metals, the electronic
contribution to the thermal conductivity was found to obey the Weidemann-Franz
law

Kg = L - UT (21)
where L is the Lorentz number, 0 is the electrical conductivity of the material, and T
is the temperature. This indicates that a trade off between high electrical
conductivity and low thermal conductivity must be considered for optimizing the ZT
of a sample.

The lattice thermal conductivity can also be adjusted through mass
fluctuations within the crystal lattice. This can be accomplished through the choice
of multielement semiconductor materials containing high and low mass elements,
and through alloying of materials. More recently, the prediction of ZT
enhancements through quantum confinement [33] has led to thin film materials that
exhibit high ZT values primarily through a significant reduction in thermal
conductivity [34]. The LAST and LASTT bulk materials developed by the Kanatzidis
group at Michigan State University and Northwestern University also exhibit
endotaxially imbedded nanostructures which form during ingot growth through a
spinodal decomposition mechanism [6]. These nanostructures are believed to
reduce electron-phonon interaction by preferentially scattering phonons while

allowing electrons to flow through them.

25

 

 

2.19 Contact Resistance

Contact resistance reduction is crucial for building an efficient thermoelectric
power generation module. The contact resistance is measured at the interface of
two dissimilar materials which electrical current is passing through. Thermoelectric
devices have a semiconductor-metal interface where a detrimental non-ohmic
behavior can arise from the Schottky effect. Thermoelectric power generation

devices are linear and reversible devices, therefore low resistance ohmic contacts

 

are desired. The losses due to joule heating at the contacts is a non-reversible
energy loss component that reduces the overall efficiency of the device as shown in

the equations below.

Q}.

 

RchZ

R

cc2

   

/

   

 

 

._ T +

l]

 

 

 

 

Figure 10: Thermoelectric module power generation circuit.

Equation (6) is the thermoelectric efficiency ofa module.

2
P0 = 1 R0 (23).

P0 is the electrical output.

I
Q, = KAT+SThl-EIZ(R+RC) (24),

Qh is the heat input into a thermoelectric device.

"C = _“ (25).

Equation (5) is the equation for carnot efficiency.

 

[u = J(l+6)2+(l+5)ZTav (26),
where

R Rccl + RCC2 + Rchl + Rch 2

 

 

5 = 7% = Rn+Rp (27)"
And the complete module efficiency becomes [35],
= #776
(”My +(#+1+5)_m (28).
27}, 2

Figure 11 shows plots of the relationship between ZT and module efficiency at
different values of 6. The values for (S for fabrication ofa efficient (~20%)

thermoelectric devices is between 0.0 and 0.1.

27

25

 

 

 

 

 

20'—
°\° I
v 15 - -
>~ .
U .
C .
.9 - -
_(_J 10 - _
t . .
“J I
5'— _'
0 “11.1I4_.4_LALL..1.4..L.LJ LlsLJ.
0 0.5 1 1.5 2 2.5 3

ZT
av

Figure 11: Efficiency degradation due to contact resistance [36].

The purpose of this project is to develop a method of making electrical contacts that

have low contact resistance and are mechanically strong.

2.20 Measuring Contact Resistance
After bonding runs, a measure of the contact resistance is needed to determine how
well the contact material bonds to the LAST materials. A room temperature

scanning probe was built to do a voltage scan across the contact.

28

 

Figure 12: Room temperature scanning probe.

A discontinuity in slope of the voltage across the junction indicates significant
contact resistance. For the power generation devices investigated in this study we
wish to have a voltage drop across all contacts which is less than one tenth of the
voltage drop across the full length of the thermoelectric leg. The room temperature
scanning probe is a quick and accurate way to determine the quality of a contact

material.

29

Unicouple AB N Leg Scan
140 ...,...,...,...,...,...

 

120

100

80

60

Voltage (IN)

40

V r1 v—rv 'f'T'j rrI—V 1". I 7T1

20'- ; .2
. . .

b d
I u
0 A J .L j A J 14 A l l A L J A 1 L l L L A

-2 0 2 4 6 8 1O

 

 

 

Position (mm)

Figure 13: N leg thermoelectric to 316 stainless steel scan.

As shown in Figure 13 LAST to 316 stainless steel show room temperature contact

resistivities below 20p.Q"cm2 which corresponds to less than a 9pV drop across the

junction for these samples. The 316 stainless has a well matched coefficient of
thermal expansion to that Of LAST, which will reduce thermal stresses during high
temperature Operation.

The contact resistance can change as a function of temperature, and is
therefore of interest to measure contact resistivities as a function of temperature
[3 7]. A system is presently being assembled by my colleague, Muhammad Farhan

for temperature dependent measurements of the contact resistances.

30

3.0 Ultra High Temperature Measurement System (UHT)

Measurements ofthe thermoelectric properties must be done at the operating
temperatures of these devices. Typical thermoelectric power generation modules
operate anywhere from 300K to 1000K. Systems described in the literature for
characterization of thermoelectric materials commonly utilize the four probe
measurement configuration to minimize the errors associated with the contacts
[38-43]. The system designed in our lab also utilizes the four probe configuration in
a manner that allows for simultaneous electrical conductivity and thermopower
measurements and has the capability Of measurements from room temperature to

1000°C. Accuracy of the system was evaluated by using a reference platinum wire.

3.1 Experimental Setup [44]

The high temperature system is made up of four main components; furnace,
computer, measurement stage, and a vacuum pump. For uniform sample heating, a
12009C split tube furnace with a 2 foot long heating zone was used. With such a
large furnace capacity, the maximum natural drift cooling rate is less than 2°C per
minute. The slow cooling rate allows for accurate measurements of the
thermopower and electrical conductivity from high temperatures (~800-900K)

tTM

back down to room temperature. The ProboSta measurement platform and

wiring terminal is a purchased item from NorECs. Platinum wiring is used due to
the fact that it remains stable in inert or non-inert atmospheres [45], and type S
thermocouples are used. The base of the system remains at room temperature to

allow for solder connections of all electrical leads needed for measurement. The

31

 

voltages measured for thermopower and electrical conductivity are in the nano to
microvolt range. For measuring these voltages, Keithley 2182 nanovolt meters are
used. Currents are sourced with a Keithley 2400 source meter. The measurements
of electrical conductivity and thermopower needs three 2182 meters. A Varian dry
scroll and turbo pump is used to pump the system down to pressures below
ZOuTorr. The system is for measuring bulk samples only. The minimum sample

dimensions are on the order of 2x2x2mm.

 

Figure 14: Sample with mounting configuration.

As shown in Figure 14, two thermocouples, a heater and two currents leads are
attached to the sample for measurement. The diameter of the platinum wire used is

.003 inches, which was found to be the smallest diameter that could be easily

handled.

3.2 System Characterization

To test the accuracy of the system platinum was used. The data was compared to

NIST data of platinum.

32

 

 

 

 

 

 

 

 

 

l 1 I I I I r

9104 - ' _

0 Measured Elect. Cond. (S/cm)
4 I NIST Elec. Cond. (S/cm)

810 - ..
E
.2 4
g 710 - _
E 6104 - -
O
O 4
46' 510 - o -
(D ’q
_ ..
LU 4 ’104 I- ...' -

o.’.
310‘ _ - a. _
2 104 I I I I I I I
200 300 400 500 600 700 800 900 1000
Temp (K)

Figure 15: Comparison of NIST data and system-measured data.

33

 

 

5 I 1 I T I l I

>5

.'"._'.'

.2

"'6

3 4- ..
U

C

O

0

To 3*-

.2 I
.2

0

2

UJ 2_ _
L-

8

L

UJ

.o—a

a w- -
O

S—

G.)

o.

O l L l J I I l

 

 

 

200 300 400 500 600 700 800 900 1000

Temperature (K)

Figure 16: Percent error from NIST data and measured data of electrical
conductivity.

34

 

0 Measured TEP uV/K
I TEP Platinum (uV/K) from Meas. Sci. Technol. 12 (2001) 264-272 (Roberts)

 

 

 

 

 

 

 

'4 fl 1 I i I l I
I

-6 l— .1
2
> '8 "' -I
3
L
‘é’

-10 - -
8 't.
O o
E '0.
a -12 l— ‘ n-

o
'C I
l.— 0
he
-14 .. I. -
o
o
-16 1 l 1 L 1 l i
200 300 400 500 600 700 800 900 1000
Temp (K)

Figure 17: Measured thermopower and reference thermopower of platinum.

3S

 

1.5 '- I-
L—
O
t:
”J
E
8 1 - -
h
0)
0.
CL
E
0'5 P -

 

 

 

0 l l l l l l

200 300 400 500 600 700 800 900 1000

 

Temperature (K)

Figure 18: Percent error between measured and reference thermopower.

The percent error for both the thermopower and electrical conductivity is below

four percent over the 300K to 900K temperature range.

3.3 N-Type Material Annealing Experiment

The alloys of lead, tellurium, silver, and antimony (LAST) are proven
excellent thermoelectric materials for temperatures ranging from room
temperature to 800K [46,47]. The material is either used as cast or hot pressed.
Ingots of LAST are cast in a quartz tube and then machined into thermoelectric legs
for use in fabrication of thermoelectric power generation modules. Annealing of
these materials may improve the properties of the LAST material. Properties such
as electrical conductivity (S / cm) and thermoelectric power (uV/ K) are improved.

Annealing of thermoelectric materials is a quite prevalent technique. Some

36

examples found in the literature include long thermal anneals at temperatures of

BOO-850°C that were used to obtain B-FeSiz phase from the metallic a-FeSiz phase

[48]. The samples were cut and annealed in a quartz ampoule filled with argon gas.

A comparison of as grown and annealed samples of BaBAuxSi46-x was done to show

a chemical composition variation [49]. Annealing was used to improve the thermal

stability of the thermoelectric material Zn4Sb3, and healed certain defects and

surface states, which were created either during the synthesis or with the removal

of elemental zinc [50]. Thermal investigation revealed an enhanced heat capacity

for nc-Zn4Sb3, which could be reduced to the bulk value of Zn4Sb3 by annealing at

190°C. Cold compacted and sintered Pb1—xSnxTe samples were examined for their

annealing effects [51]. Annealing of these materials under controlled pressures of
Te and Pb can be used to change the carrier concentration. The carrier
concentration can be varied by changing the annealing temperature, on account of
the variations of the solubility limits of the Te and Pb with temperature. For this
study, the sample's thermoelectric properties were measured before and after
annealing to compare property changes.

To investigate annealing effects, a sample from an n-type ingot, ETN207was
prepared for measurement. The room temperature electrical conductivity before

the measurement was 147.4 (S/cm].

37

 

2000

 

I I I I 1 r x
o
o
o
o
o
o
1500 - .0. -l
o
O
A ..
> .0
e ..
a) 1000 - . ..
O) O
(u o
'3: o
O o
> .°
.0
500 . . a
o
o
o
O I I I I I I I

 

 

 

-1 0 1 2 3 4 5 6 7

Position (mm)

Figure 19: ETN207 sample 1 before the high temperature measurement, electrical
conductivity scan.

The sample was first taken to 400K and cooled to let the silver paste contacts cure.
The probe separation was then adjusted so the room temperature conductivity
matched the room temperature scanning probe electrical conductivity. It is
assumed that the sample does not change when cycling to 400K. The sample was
then cycled three times to 700K and back to room temperature. Twice the sample
was held at 700K to see if the sample would anneal. A typical result of an annealed
sample is that the conductivity of the sample will increase while the thermopower
would remain the same. This sample however had a conductivity that decreased

and an increase in the thermopower.

38

Electrical Conductivity (S/cm)

Thermopower (pV/K)

150

140

130

120

110

100

 

 

Time of Anneal

 

660

680 700

720 740

Temp (K)

Figure 20: Electrical conductivity anneal at 700K.

-240

-260

~280

-300

-320

 

 

I I

I I

Time of Anneal

 

660

I
680 7

l
l
l
l

V
7

I
20 740

Temp (K)

Figure 21: Thermopower anneal at 700K.

39

 

 

 

The sample was removed from the measurement system and there was a noticeable
coating on the outside of the sample. This coating suggests that there was an

outgassing or evaporation of material during the measurement run.

 

600 I I f I I I I
500 - ,0 a
O
.0
C
400 - .‘° -
A 0
>1
V o
g, 300 L ..' 1
(U o
“-5-; .-
O
> 200 . .' -
O
O
C
O
100 . .' -
O
C
C
0 I l l I I I I

 

 

 

-1 o 1 2 3 4 5 6 7
Position (mm)

Figure 22: Scan of ETN207 after temperature cycling and measurement of
thermoelectric properties. RT o=544.68 (S/cm).

The sample properties changing during the temperature cycling was also an issue.
Most samples measurements repeat their properties through temperature cycling.
Four of the ETN207 were annealed at 423°C for 24 hours under a flow of Ar/HZ gas.
These samples were then scanned at room temperature for their electrical
conductivity. One of these samples was then used for a temperature dependent run
to see how the properties changed as a function of temperature compared to the un-

annealed ETN207.

40

 

Four samples were annealed at 423°C for 24 hours. The samples
were named as follows; ETN207_sample number_annealed temperature. After
annealingETN207_1_423 had a room temperature conductivity of 388.68 [S/cm].
The sample's conductivity before the anneal was 147 (S/cm). Sample
ETN207_1_423 was mounted in the UHT for measurement of its temperature

dependent properties. The measurements were taken from room temperature to

700K.

 

 

 

 

O I I I I I I
O
O
-100 - o "
O
I
O
-200 - ' -
I
A O
0

>3 -300 . .0 .
a) O
0) .°
S 400 - . -
— O
s

-soo - ,0 -

O
O
O
O
-600 . ,0 -
O
O
O
_700 l l l 1 l J
o 1 2 3 4 5 5 7

Position (mm)

Figure 23: ETN207_1_423 room temperature scan after anneal at 423°C.

Sample ETN207_2_423 showed a large crack about 4.5mm into the sample. Below is

a plot of the room temperature scan.

41

f .

 

 

 

 

0 I I I I I DebooooOOW‘
o=312(S/cm)
A '5000 p- _
>
3;
(D
O)
59
B
> -1104 t -
0=41O(S/cm)
.OOQOOOOOOOOOOOOOOOCOO.
’1.5104 I J I J l . l
3 4 5 6 7

-1 o 1 2
Position (mm)

Figure 24: Room temperature electrical conductivity scan of ETN207_2_423. The
conductivities of each side of material are shown.

Sample ETN207_3_423 had a room temperature electrical conductivity of

429.24 (S/cm) after the anneal at 423°C.

42

Voltage (pV)

Voltage (pV)

 

 

 

 

I I I I I I I "
.0
0
-100 . ,o' -
0
0
0
-200 - .' -
0
.0
0
-300 - ' -
0
.0
-400 . .0 -
0
o ' 0: 429.24 (S/cm)
-5oo - o' _
.0
.0
0
500 . , -
0
-700 I I I I I I I
-1 o 1 2 3 4 5 5 7
Position (mm)
Figure 25: ETN207_3_423 room temperature scan.
100 I I I I I I I
0 )- .01..
0
.0
-100 - . -
0
0
0
'200 - o=457.79(S/cm) .' -
+ 0
.0
-300 - o '
.0
0
-400 '°
0
0
0
0
-500 . .. 4
a...
_600 I II I 1 l L l

 

 

 

Position (mm)

Figure 26: ETN207_4_423.

43

 

A second cast sample (ETN207_5_Cast] was measured at room temperature for its

electrical conductivity, ETN207_5_CAST.

-200

 

 

 

 

 

I I I m I I
o
o
o
-400 _ ' -
.0
A 0:24367 (S/cm) . °
> - - -
3 600 .0
o
d)
U) o.
a ,-
6 -800 .. .
> o
o
-1000 - o _
o
o
.0
0'.
4200 0} I I I I I
0 1 2 3 4 5 6 7

Position (mm)

Figure 27: Room temperature scan of ETN207 as cast.

The “as cast” ETN207 had conductivities ranging from 147 to 243 (S /cm]. The

conductivities of the annealed samples are all around 400 [S/cm).

44

3.4 Ingot Uniformity

To test the uniformity ofan ingot, samples were measured in each "coin" of
an ingot; A typical ZOO-gram ingot is 22mm in diameter and 56mm tall. Seven—
millimeter coins were cut from the ingot, and labeled A to D with the A coin
corresponding to the top ofthe vertically cooled ingot. Samples from each coin

were measured in the UHT system.

 

 
 
 
 
 
 

1200 -
1000 ‘ ’ETN161A2
IENT161A8
800 ETN161BZ
-— ETN16188
600 _ *ETN161D2

 

 

0 ETN161D8

  

 

400 '

200 ‘

Electrical Conductivity (Scm‘)

 

 

300 350 400 450 500 550 600 650 700

Temperature (K)

Figure 28: Plot of electrical conductivities of samples measured from each coin.

These samples may have differing compositions or may exhibit cracks. If the
compositions were different, the thermopower should also be different. Cracks

would show less of a change in the thermopower.

45

 

 

 

 

 

-100
l o ETN161A2
J.-. g I ENT16‘IA8
? -150 ETN161BZ
: ETN161D2
% _200 ‘ XETN16‘ID8
O
5 I
g _250 . 5.3??WQX ’
a - . _ , .. ,
.: 451ih...
'— -300 ~ “III-q.
-350 I I I I I ‘l’ l I

 

 

300 350 400 450 500 550 600 650 700

Temperature (K)

Figure 29: Plot of thermopowers measured from samples of each coin.

ETN16132 from the B coin has the highest room temperature thermopower.
ETN161A8 has the highest thermopower at 700K. Higher thermopower can result
from an increase in nanostructuring in this sample, and/or from the doping level in
this sample resulting in a Fermi level located in a more rapidly changing portion of
the density of states (DOS). For example, a Fermi level located at a peak in the DOS
results in a small thermopower since the density of states is symmetrical about the
Fermi level, however slightly off from the DOS peak leads to a larger thermopower.
This sample may contain nanostructuring that would improve thermopower. As
mentioned earlier, nanostructuring would give a low dimensionality energy

dependence of the density of states, which in turn leads to a larger thermopower.

46

 

 

 

 

 

 

 

9ETN161A2
35 _ IENT161A8
ETN16182
30 - WM r ETN161D2
g." . flffir—wfihw " x ETN161D8
X I '- .lrkfl\_.%lip
E 25 - - - - ...-~m,~,.a.gag
g "I --:n.
1 20
'5
3 15
I!
u.
3 10
a
O
l 5 -(
O I I T I I I I I

300 350 400 450 500 550 600 650 700
Temperature (K)

Figure 30: Plot ofpower factors measured from samples of each coin.

Overall, ETN161A8 has the largest power factor. The rest of the samples lie from

16-20 (uW/cm'KZ] at 300K and around 10 (uW/cm'Kz] at 700K. The cast LAST

materials show to have similar properties throughout the ingot With one sample

having above average properties.

3.5 Hot Pressed UHT Results

A hot pressed material labeled, MSUHP12, was measured in the UHT and
exhibited interesting properties with pressure. During pumping, the electrical
conductivity of the material increased and when the pressure returns to atmosphere

the electrical conductivity returned to a value that was close to its original value.

47

Pumping of HP12

 

 

 

 

28 I I I
E 26 - -
2
El
3‘ 24 - -
IE
.9;
8
'2 22 l- -I
o
0
T3 20 - 3
.5
8
[I] 18 i- .0. -1
. 0
0
16 L l l
10 100 1000 10“ 105
Time (s)

Figure 31: The change in electrical conductivity while pumping down atmosphere.

 

 

 

 

Venting of HP12
28 , ,
0 0 0 00000.

A 26 - . -
5
\ 0
9, 24 . . _
2;. 0
E 22 - .0 '
*6 '-.
3 20 - -
C
o
2 18 t -
(U
.2
L
~06 16 h -
2
LL] 14 F q

12 l l

10 100 1000 104
Time (s)

Figure 32: Change in electrical conductivity of HP12 during venting of system.

48

It is thought that the moisture content may be affecting the electrical properties of
the material [52], however other hot pressed samples did not exhibit this pressure

dependent effect.

3.6 CAST to Hot Pressed Property Comparison

To compare the properties of a cast sample with the properties of a hot
pressed sample, an ingot (ETN191) was measured for its electrical conductivity and
thermopower prior to crushing it into powder. It was then hot pressed and both
ETN191 and its corresponding pressed MSUHP32-N sample were measured to see

how the properties changed before and after pressing.

 

    
 

1200 fir I I 1 I I I— I- I T I I I I r _I I I I I I I I I

I 4

. I .
’E‘ 1000 - ° -
s : :
V l :1
3 800 h -
IE ; 1
g . ETN191 .
.0 600 "‘ III
C . -l
8 . .
s 400 :- .
‘05 II
0
d.) ' .1
[I] 200 - -

 

 

 

 

200 300 400 500 600 700

Temperature (K)

Figure 33: Comparison of electrical conductivity of cast and hot pressed material.

49

 

-1OO fITI' Ifif'. III'IIIIIIIITIII—rj'

0
I l J J l

-150

I l I I If I

    

g .
E I .. ETN191 i
“3’ -200 - -
Q. .- .
0

g l

m .

.C II

'— —25o - -

 

 

 

-300 IJEIILIILJlLlllklllllllllllll

200 300 400 500 600 700 800

Temperature (K)

Figure 34: Comparison ofthermopower of cast and hot pressed material.

It is typical that the hot pressed materials have a lower room temperature electrical
conductivity than the cast materials. The room temperature thermopower did
increase for the hot pressed material but then matches the cast material at 700K.
The hot pressed material has an overall lower electrical conductivity. The hot
pressed materials nearly match the electrical properties of the cast materials at
700K. These results suggest the powder processing steps effectively remove
carriers from the LAST material. Defects and / or oxides on the surfaces of the

powder particles may cause this. The hot pressed legs may be cascaded with cast

materials or hot pressed Bi2Te3 for an efficient thermoelectric leg from room

50

 

temperature to 700K. As discussed earlier, the hot pressed materials have a higher

strength than the LAST and are more uniform in composition.

51

4.0 Diffusion Bonding

4.1 Diffusion Bonding Introduction

Diffusion bonding is a method of bonding two similar or dissimilar metals,
alloys, or nonmetals. This is achieved by pressing together the two materials to be
bonded either in vacuum, an inert gas, or a reducing atmosphere, and increasing the
temperature until atomic diffusion across the interface bonds the two materials
together. Diffusion bonding experiments typically yield the best results in an
atmosphere with the least amount of oxygen present. This helps to reduce the
amount of oxides present at the interface of the two materials, which tend to
degrade the bond. There are many control parameters investigated in these
diffusion bonding including; temperature, time, and pressure. The bonding
temperature should be 50-80% of the melting point of the more fusible metal [53].
Tamman's rule states that the importance of diffusion is made evident at
temperatures about 80 to 90 percent of the absolute melting point of the material
considered [54]. (Depending on the physical process that you are considering,
diffusion can be very significant at 0.6 of the melting temperature]. The
temperature aids in the inter-diffusion of the atoms across the interface of the two
materials and the deformation of the interface surfaces. The bonding pressure fills
voids and pushes out any oxides that may reside on either of the two surfaces.
Bonding times are found experimentally and can range from minutes to hours.
Some of the criteria for choosing a metal interconnect is one that has a low electrical

conductivity, 3 high thermal conductance, and a well matched coefficient of thermal

52

expansion to LAST and LASTT. To minimize the stresses induced by bonding,
contact materials have been chosen with comparable coefficients of thermal

expansion (CTE). Most solid materials expand upon heating.

A—l = alAT, (29)
10
V0

Equations [29) and (30) show the linear relationship between the CTE, a, and the

 

change in either length or the entire volume ofa material [55]. Thermal expansion
is the increase in the average distance between atoms. LAST has a coefficient of
thermal expansion around 21ppm / K. Choosing a contact material with a similar
CTE helps reduce thermal stresses associated with the bonding process.

Many considerations must be taken into account when designing a
thermoelectric module, including the strength of the materials used. For cast LAST
materials, a fracture strength of ~ 15 MPa has been reported [56]. To increase the
strength of these materials a separate group at MSU are investigating powder
processing and hot pressing of LAST. Through these processes, the grain sizes
within the hot pressed samples can be made much smaller (~ 5pm] than in the cast
samples (~500um]. This helps to increase the strength of the samples to ultimately
make a more robust module [5 7]. Diffusion along grain boundary paths is the
dominant diffusion channels, thus through powder processing the diffusion bonding
to the LAST materials is expected to improve. Materials that are placed in
environments at elevated temperatures and exposed to static mechanical stresses
would undergo a plastic deformation process known as creep [58].

53

 

 

£=— (31]
[0
d8
'=— 32
8 dt ( l
n
5', =k a bexp( QC) (33)
GS RT

Equation [31) is the definition ofcreep strain. Equation (32] gives the creep strain
rate. Equation (3 3) gives the relationship gives the relationship of applied stress 0, ..

temperature Tand grain size GS to the steady-state creep rate as. The values ofk

and b are material dependent constants. The value ofb is typically 2 to 3, and n is

between 1 and 6 depending on the creep mechanism. QC is the creep activation
energy, and Tis temperature. Equation (33] shows that the smaller the grain size,
the higher the creep strain rate. Higher creep rates mean the interface surfaces will

deform more during bonding allowing better interdiffusion ofmaterial across that

interface [59].

54

 

Rupture

 

 

 

, Primary

V

 

 

V

Creep Strain, 8

Secondary

 

Instantaneous deformation

 

 

 

Time, t
Figure 35: A schematic depiction of a Creep Curve ofstrain versus time at a constant
stress, tr is the total time to rupture [60].
In Figure 35 is a curve of creep strain vs. time [60]. Primary or transient creep
occurs first followed by secondary creep or also termed steady-state creep rate is
constant where the slope of the curve tends toward zero. This is when the material
retains its ability to deform. The tertiary creep is the materials structural failure.

This is normally termed as rupture.

55

T ‘mfi

 

 

 

 

 

(\7 \ \IW /

j \ /
\\\ ’/-\/ ;\_} \\\\\ /¥“‘\
B \fi 71 \/

KKK
1 1. l -

i

2

 

 

 

 

 

Figure 36: Diffusion bond B will have better interface or contact resistance due to
the reduction of the size of the grains in the material.

Figure 36 is a schematic that illustrates a reduction ofgrain size in the
thermoelectric material. With the smaller grain sizes, interdiffusion of material at
the interface occurs more readily along the grain boundaries. The reason for the
high diffusivity along grain boundaries is the high concentration of defects,
especially vacancies, at the grain boundaries. The problem with interdiffusion is that
contact material may migrate into the thermoelectric material and change its
thermoelectric properties.

Similar thermoelectric materials, such as lead telluride have been extensively
investigated. Studies have focused on the poisoning of PbTe from different metals
as contacts [61]. Initial studies on lead telluride based modules showed degradation
of the power generation capabilities over time and Eiss [61] believed it to be
associated with failure of the bond between the lead telluride and the metal shoes

used as electrical contacts. He studied the poisoning ofthe thermoelectric legs by

56

 

inter-diffusion of contact material between the lead telluride and the braze zone or
shoes, and the cracking that would result from stresses generated during the
bonding process. He observed the effects were inversely proportional to the
melting temperature of the metal contact. The studies were conducted by
measuring electrical conductivity and thermopower changes to the materials before
and after bonding. Also measurements were done after thermal cycling to see if
more changes to the lead telluride properties have occurred. He noted the changes
in these properties and documented which elements changed, or "poisoned" the
thermoelectric material more. Shown in Table 4 is an example of how PbTe’s

thermoelectric properties changed with the addition ofa contact metal.

57

Table 4: Changes in thermoelectric properties of PbTe from brazed materials[61].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Thermopower Electrical Conductivity
Material Temp °C
% Change % Change

106 -56 -52
n-PbTe to Cu

164 -50

104 -100 164
p-PbTe to Cu

154 -100

97 6 -5
n-PbTe to Fe

158 7

99 94

152 73 261
p-PbTe to Fe

96 68

149 91

 

Stress can arise from the differences in thermal expansion of the contact

metal and the thermoelectric material. Eiss considered a bond by brazing, and it

was apparent that below the brazing temperature the component having the higher

coefficient of expansion would be in tension and that the stress level will continually

increase as the bond temperature is lowered [61]. In addition it was shown that the

lowest electrical contact was formed when the braze layer was heated to a softening

58

 

 

state where significant deformation occurs. lfthe interdiffusion of the contact
material into the LAST degrades its properties, diffusion barriers will be needed.

lnterdiffusion can also be a source of stress at the contact [62]. The stress
generates if the total vacancy flux occurs due to a difference of the intrinsic diffusion
coefficients, also known as the Kirkendall Effect [63].

Diffusion barriers are used to aid in stopping migration of contact atoms into
the bulk material [64,65]. Long term studies of the contacts will be undertaken to
ascertain the stability of the contacts and to help determine ifpoisoning of the
samples is occurring. ln Ono's paper, a study was done on six different elements as
diffusion barriers of copper into silicon. Copper has been ofgreat interest as a
interconnect material because of its lower resistivity and higher electromigration
resistance than aluminum [64]. Titanium, chromium, niobium, molybdenum,
tantalum, and tungsten were investigated as the diffusion barrier layers. These
elements are listed in Figure 37 in the order of self-diffusion coefficients in

transition metals between 400-600°C [66].

59

 

160

120

80

Q (kcal/mole)

4O

 

Tle)

 

 

   

I l l
O 1000 2000 3000 4000

0

Figure 37: Self-diffusion activation energy as a function ofthe melting

temperature[67].
D = D0 exp(iQ—) (34)
kT
J' = -D - VC (35)

Equation (34) shows the exponential dependence of diffusivity D for elements as a
function of the self diffusion activation energy Q. According to Figure 37; tungsten,
tantalum, molybdenum, and niobium have the highest values of Q which yields the
lowest values of diffusivity. Equation (35] gives the relationship between mass
diffusion flux, 1 as a function of the diffusivity. The lower the diffusivity, the less
mass flux will be. Therefore, tungsten and tantalum are candidates for diffusion
barriers. ln Ono’s paper, the diffusion metal was deposited on a silicon wafer by

sputtering, followed by a layer of copper. By studying the resistance changes after

60

annealing, they concluded that tantalum and tungsten were the two best diffusion

barriers of copper into the silicon.

4.2 Diffusion Furnace

A three zone diffusion furnace was designed and assembled for an operating
temperature up to approximately 1000°C with a controlled atmosphere (typically
vacuum or an argon hydrogen mixture]. For full operating procedures, see

reference [68].

 

Figure 38: Diffusion bonding furnace.

The furnace interior temperature is measured by a thermocouple that is placed into
the center zone of the furnace, and the Ar/Hz gas mixture 95 /5 is used as a forming

gas to help remove surface oxides that could degrade the quality of the bonds.
Temperature control of the diffusion furnace is accomplished with

temperature controllers for each of the three zones of the furnace.

61

A pressure controller was added to the diffusion furnace to create a flow of the
argon-hydrogen through the system. The gas would flow at around 1 liter per
minute while maintaining a total pressure of 800 Torr in the system at all times.

This would replenish the reducing gas during the entire bonding run.

 

Figure 39: Diffusion furnace pumping system with computer controlled pressure
mechanism.

Diffusion bonding experiments were labeled with letters. The first

experiment was labeled with A and then the second B and so on.

62

4.3 304 Stainless Steel Diffusion Bonds to LAST and LASTT Material

Table 5: Stainless steel properties.

 

 

 

 

 

 

Stainless Composition CTE Melting Room Thermal
Steel (K‘l) Point Temperature Conductivity
(°C) . Electrical (W/m°l(]
Conductivity
(S/cm)
304 FeCr18Ni10 18.0 1400- 14,000 16.3
1455
316 FeCr18Ni10Mo3 16- 1370- 13,000-14,000 16.3
18 1400
410 FeCr12 10- 1480- 14,000-18,000 18-24
12 1530

 

 

 

 

 

The above table shows the different stainless steels and their corresponding
properties. These stainless steels were chosen for their CTE's being approximately

to that of LAST and LASTT.

4.3.1 First Bond Experiment

A bonding experiment of 304 stainless steel to LAST and LASTT was done.
The bonding temperature was 700°C for 4 hours. Thirteen unicouples were placed
in the boat initially and only 8 bonded, for a bonding yield of better than 50%.

The first bonding experiment of LAST and LASTT to 304 stainless steel
showed promising results, so a second bond was attempted. The bond was done at

700°C with a 4 hour ramp and 4 hour soak. Only 5 out ofthe 13 unicouples bonded.

63

 

4.3.2 Unicouple Fabrication with 304 Stainless Steel

For unicouple run A, ETN133 and ETP16 were bonded to 304 stainless steel using a

4 hour soak at 700°C. Below are results for this bonding run.

 

0.0002 'ITI I l I I I TTT I I I I II I I I I II—I ' I I I I

r Unicouple A1 -
r N Leg Contact Scan ‘
' Jonathan D'Angelo a! '
' 5123/06 M '
0.00015 - _, -
. or ~1353Scm . s .

 

 

 

 

- ZT: 0.0672 -

E ' l

g 0.0001 - d

E I I
5

> d

510'5 - J

II ...V 1

l- . ... q

L J39 I]

0 -i

j l l L l J I l l 1 LJ 1 l l l j l l I L l l l I I I l L
0 1 2 3 4 5 6

Position (mm)

Figure 40: Unicouple Al N leg scan.

64

Voltage (V)

0.00035

 

0.0003

0.00025

0.0002

0.00015

0.0001

5 10'5

   
 
 
  
  

..m

LLJlllllllllllllllllllJllPLj

IIfi'IIII'j—IIr'TrfI'III‘IIrIIr

Unicouple A1

P Leg Contact Scan
Jonathan D'Angelo
5IZ3I06

o ~800$cm"
ZT: 0.0672

 

 
 

1 2

Position (mm)

4 5

Figure 41: Unicouple A1 P leg scan.

65

l

[1111.11lnllllnllanlnlLLllnnlI144

 

O)

Voltage (V)

0.0003

0.00025

0.0002

0.00015

0.0001

5 10‘5

 

IIIITfITT'IIIr'IIII'IfifIIIII

I Unicouple A2

- N Leg Contact Scan 0’:
'_ Measured By Jonathan D'Angelo o .1.
_ 5/12/06 {66“

0 ~ 1,400 3cm"

 

II'IIII'IIII'ITII'II

 

 

0 1 2 3 4 5

Position (mm)

Figure 42: Unicouple A2 N leg scan.

66

LlelllllllllllllllllLllllllJ

 

O)

 

0.000356......---.....,.fi.t....,...-
UnicoupleA2
P Leg Contact Scan
0.0003 Measured By Jonathan D'Angelo .0.
5/12/06 4’

000025 0 ""7808CITI-1

 

lLlllJlllllJJJllllllllllllJllllllJ

 

 

E 0.0002
0
O)
.9
§ 0.00015

0.0001

0;
may
5 10 5 Jpn"
61".
air”! .
owlllllllLLlLLLLJLlllll[All]
0 1 2 3 4 5 6

Position (mm)

Figure 43: Unicouple A2 p leg scan.

The room temperature ZT for unicouple A2 was measured to be 0.015. The

expected value was approximately 0.3.

67

Voltage (V)

0.0003

0.00025

0.0002

0.00015

0.0001

5 10'5

0

 

1-
r

 

EFITIjIIII'IIII'IIII'III

O

firIII'IIIIIIIIIrIIII'ITj‘I'IIII

Unicouple A3

N Leg Contact Scan .
Jonathan D'Angelo .0°
5I18I06 '

   

0 ~ 1,200 5cm"

artM

fl"
lllllllllllllllllllllllIll

llllllllllllllllllllllllllLl

 

 

1 2 3 4 5

Position (mm)

Figure 44: Unicouple A3 N leg scan.

68

O)

 

0.0005 ...,....,........fi....,....
* Unicouple A3

I P Leg Contact

_ Jonathan D'Angelo
0.0004 h55/18/06

C 0 ~ 600 Scm‘1

 

A 00003 - _~ -

a I I’..:“‘ II

a) 633*. .

8’ . ‘i' . , .
E) . ...0'

0.0002 - I. -

' ”o. .

' 5’ ‘

. I”? .

_ 0* 0’ . _.

0.0001 - M .1

 

 

 

Position (mm)

Figure 45: Unicouple A3 P leg scan.

Unicouple A3 has a room temperature ZT of 0.08. A lead-tin solder was added to the
top of the electrode in an attempt to lower to unicouple’s resistance. This decreased
the module resistance 2mg}. This however, did not significantly change the ZT for
the module giving a new ZT of 0.079 after adding the solder bridge.

A second experiment was done to decrease the resistance from the metal
electrode by soldering a copper wire down to the 304 stainless steel and the ZT was

tested again.

69

Resistance (Ohms)

 

0.01 _
- Unicouple A3
: o o ZT Meter
00095 .' . Jonathan D'Angelo
I ZT: 0.08016 . 5I15I06
0.009 - . .
; o o 00-
0.0085 _-
0.008 _-
0 0075 -_ Solder added to unicouple bridge
' - ZT : 0.078845
: C g,
n O
0.007 -
" 0
: . ’ 0 O on
0.0065
0.001 0.01 0.1 1 10 100
Frequency (Hz)

Figure 46: ZT measurement with and without solder.

70

 

 

 

 

0.0088 I I I I IIII' I I I I IIII' I I I IIIII' I I I IIIII

I . . Unicouple A3 I

- ZT Meter -

' Jonathan D'Angelo ‘

0.0087 - -

. 5I17l06 -

C 0 ZT=0.0602 I

. 3rd Run -

13 00085 f Copper wire was added '2
E - . to unicouple bridge ,
O h . c d
V L to see IfZT improves .
8C3 0.0085 - Begun :It did not -
P I

g h I
g I . I
r «I

a 0.0084 b -
1- J

1 - :

. +- o -

00083 _ . . . ‘

- 1

0.0082 L I l Allin] l l l jljlll .l I l l lllll I I I I III.
0.001 0.01 0.1 1 10

Frequency (Hz)

Figure 47: ZT measurement with copper wire added. ZT dropped down to 0.0602
even though overall resistance of the unicouple was improved.

As shown in Figure 47 the ZT does not improve with the lowering of the electrode
resistance through this technique. This may have been caused by the module
resistance being dominated by contact resistance.

When the forming gas was not used, and a lower pressure was used during
bonding (30 minute ramp to 700°C and 4 hour soak at 700°C), no bonds were

formed between 304 stainless steel and LAST.

71

 

 

0.0005

 

....,..1f,....1.....r.fi.....
' UnicoupleA4

: P Leg Contact Scan

. Jonathan D'Angelo
0.0004 -5/23/06

 

. d~9005<:m'1 .

- ZT: 0.05426 -
A 0.0003 - .
a p . . - ' d
a.) c“: 5:.
a ’ ‘1 I
E : o. ‘
> 0.0002 1- . ~

P 3..

r . .v

I ,...-~ .’ ' 2

0.0001 - -
i- ..
i

 

 

OQWJIILlLljgjLAPAIIIHLIILLJ
0 1 2 3 4 5 6

 

Position (mm)

Figure 48: Unicouple A4 P leg scan, there are visible cracks along the length ofthe
sample.

For unicouple bonding run A, most of the contacts show a low contact resistance,
however most of the thermoelectric legs did show signs of cracks in the voltage
scans. The cracks in the legs increase the resistance of the unicouple and decrease
the strength. For module construction, unicouples based on LAST and LASTT should
have room temperature ZTs of approximately 0.3 and have no cracks in the legs of
the material.

Unicouple run I was a two stage bond of ETN137 and ETPZS to 304 stainless
steel. The two stage run has a 30 minute ramp for the n—type bond and 4 hour ramp

for the p-type material. Both bonds have soak times of4 hours at 700°C. Only 6 of

72

20 unicouples were fabricated from this run. The reason for the low bonding yield

was not clear, so another bonding attempt was made.

Unicouple run U is a bond of ETN142 and ETP26 to stainless steel 304. The
coefficient of thermal expansion of 304 stainless steel, 18ppm, is a closer match the

LAST and LASTT, ~21ppm than that ofstainless steel 410, which has a CTE of

10ppm.

U1 unicouple N leg

250 I I I I I

200 -
#5..”

150 t .

 

Voltage (W)
‘5.

 

 

 

100 - :5" i
.:
50 '- —1
M
000000.
0 J I L 1 1
-1 0 1 2 3 4 5

Position (mm)

Figure 49: Unicouple run U N leg scan.

73

U1 unicouple p leg

 

 

 

 

350 I I I I I
300 - ...o -
.0
250 '- Pl... .1
O
5‘ ,.-‘
g 200 " .. "'
a) J'a'
0)
g 150 _ ,.d:° -
g «0" ,r
.0
100 h .,.o -
I
50 1" u:
I
O O...
.00...
O I I l l
-1 0 1 2 3 4 5

Position (mm)

Figure 50: Unicouple run U P leg scan.

The bonding of LAST and LASTT to 316 shows a relatively low contact resistance

below SOuQcmz, and the material has fewer cracks after bonding. This suggests the

CTE mismatch may cause cracks in the material rather than the pre-bond
thermoelectric material having indigenous cracks.

Furnace run V and W were done simultaneously. One boat was a bond of
ETN149 and ETP24 to stainless steel 410 and the other boat was a bond to 304
stainless. The bonding run was only done in one thermal treatment stage consisting
of a 4 hour ramp and a 5 hour soak at 700°C. The 304 stainless steel interconnects

bonded 4 unicouples of4 put in the furnace while none of the 410 stainless steel

74

electrodes based unicouples bonded. The 304 stainless bonds well with only a one
stage bond.

Furnace run X was a bond of ETN149 and ETP33 to 304 stainless steel. A one
thermal treatment stage bond consisting of a 4 hour ramp and a 5 hour soak was
used. Nine out of 15 unicouples were bonded, however a significant contact
resistance was found at the LAST to stainless steel 304 bond as indicated in Figure
51: Scan of N leg from unicouple run X. This shows a high contact resistance ofthe

n-type material to the 304 stainless steel Figure 51.

 

 

 

 

700 T l I I l
non-0"...“
600 - M H
500 F u
’>" 400 - -
3-
8) 300 l- all
59.
g 200 - -
100 - ..
o
0 '- CI
_100 l 1 I l 1
-1 0 1 2 3 4 5

Position (mm)

Figure 51: Scan of N leg from unicouple run X. This shows a high contact resistance
of the n-type material to the 304 stainless steel.

75

4.4 410 Stainless Steel Diffusion Bonds to LAST and LASTT Material

As indicated in the last sectionm bonding of stainless steel to the LAST
material has shown promising results. Development ofa module with diffusion
bonded contacts to stainless steel 410 was done using a 30 minute ramp time to
700°C and a soak time of 4 hours at 700°C. The total soak time was 4 hours. The
furnace was filled with argon to a pressure of 350Torr and then sealed off. During
the rise in temperature the furnace would reach a pressure of near atmosphere to
inhibit outgassing of the thermoelectric material at such high temperatures.

For this experiment a four leg mold of R300 was built to cast the module legs

before bonding.

 

Figure 52: Module Legs Cast in R300.

The R300 is a thermal insulator primarily used in furnace insulation. The purpose
of the R300 is two fold. First is to keep the legs from outgassing during operation,

and second is for mechanical support. Figure 52 shows a picture of the legs cast into

76

the R300. All of the electrical connections were to be diffusion bonded at one time.

Figure 53 shows how the module was setup to be bonded.

 

Figure 53: R300 module bonding setup.

 

Figure 54: Completed module, top View.

 

Figure 55: Completed module, bottom view.

All eight connects of the module were successfully bonded. Figure 54 and Figure 55

show pictures of both the hot side and cold side interconnects. Unfortunately no

77

measurements were done on this module, because the contacts on the hot side fell
offduring the polish ofits surface. This experiment was an example ofbonding a
complete module in one step. This can speed up the fabrication process, however it
also requires that all contacts must bond to complete a module. Bonding individual
unicouples and then fabricating an entire module would prove to be a superior

method.

 

Figure 56: This diffusion boat has a set screw for both n and p legs for during
diffusion bonding.

The first bonding experiment done with this diffusion boat shown in Figure 56 used
a 30 minute ramp time to 700°C, and a 5 hour soak time. The starting pressure of
the furnace was 332 Torr of argon. All four unicouples in the experiment bonded,
but the p-type material showed signs of degradation. At low atmospheric pressure

and the bonding temperature used, the thermoelectric legs are known to outgas

78

during bonding, therefore another experiment was run using an intial pressure of
approximately 420 Torr.

The soak temperature, ramp time, and soak time remained the same as
before. The unicouples formed at the higher pressure showed less degradation at

the interfaces as shown in Figure 57.

 

Figure 57: The unicouple on the left had the starting pressure of 420 Torr, and the
unicouple on the right had a starting pressure of 332 Torr.

The room temperature ZTs of the unicouples were measured using a direct ZT

measurement technique [69] and showed a ZT of 0.2. The stainless steel used was
410, (FeCr12.5]. The 410 stainless steel has a electrical conductivity of 17,857

S/cm. A calculation of the resistance of the metal interconnects shows that its
resistance should be 33711.9. Resistance in the micro-ohms is the threshold for
metal interconnects. The bonding of the 13 unicouples was done using a 30 minute
ramp to 700°C followed by a 4 hour soak at 700°C, and a starting pressure of 414
Torr of argon. Only 6 of the 13 unicouples bonded. Also 5 of the six bonded show

signs of outgassing. The unicouples that did not bond did not have adequate

79

mechanical pressure applied to the legs to allow for bonding. The next bonding run,

less unicouples were attempted at bonding, and a higher starting pressure was used.
A follow up bond ofseven unicouples was done with the same conditions

attempted to bond the 13 unicouples, but with a higher starting pressure of around

450 Torr of argon.

 

Figure 58: Unicouples built with starting pressure of 450 Torr.

The measured ZT on these unicouples was 0.16 at room temperature, and a 12 leg

module was built from them as shown in Figure 59.

 

Figure 59: Side view ofa 12 leg module.

80

 

Figure 60: Top view of 12 leg module.

A bond run of 7 unicouples was done at 700°C, ramp of 30 minutes, soak of4
hours, and a starting pressure of 420 Torr of argon. Five of the 7 unicouples
bonded. The material used for this experiment showed visible surface cracks. Such
cracks at the bonding interface inhibit the diffusion bonding process [70].

A bonding experiment of ETP19 and ETN126 to 410 stainless steel was
carried out. Figure 6lshows the power factor for ETN126 which peaks at 15 at
around 500K and the sample exhibited signs of evaporation or sublimation after the

measurement. The properties of ETP19 were not measured.

81

 

20 ....,....,fi-..,---.,....,
LAST (N-type) May 4,2006 1

Made by Ed Timm -
Measured by Adam Downey .

Z T measured in separate
system @300K = 0.11

  

 

 

   

 

r5; 15 .. wwflmuonuuuhu... _.
E . O Q I . ' I . . ‘
2 . ' ° . . . .
a .
3 . .
3 1o - -1
‘5 .

«I

u.

'5 BEFOPE RUN; AFTER RUN“ ‘

3 . . ’ “47.... '

g 5 - ‘ _

L V s. weible subl'mauon
o I I I I l I I I I I I I I I l I I I I I I I I I I
300 400 500 600 700 800

Temperature (K)

Figure 61: Power Factor for ETN126.

Eighteen unicouples were loaded and 14 bonded successfully. The four unicouples
that did not bond did have bonds between the p—type material and the metal

electrode, however the n-type material to metal electrode did not bond.

 

Figure 62: shows the 14 unicouples bonded at 700°C for 4 hours.

Unicouple run B was a bond of n-type LAST and p-type LASTT to 410
stainless steel. The bonding temperature was 700°C for 4 hours with a 30 minute

ramp.

82

Voltage (V)

 

0.0007 I I I I I I I T I I j T I I I I I rj—I I I I
Unicouple 82 o i
.1

 

 

N Leg Contact Scan _.I .
0.0006 Jonathan D'Angelo M j
5I23I06 .. “.00" :
00005 0 ~ 1,14OScm" ........¢~‘* .1
ZT: 0.07117 M“. :
0
0.0004 . '3
0.0003 -
I
0.0002 .'
0.0001 1
not. .
000‘. :
O W 1 1 1 1 I 4 1 1 1 l 1 1 1 1 l 1 1 1 1
0 1 2 3 4 5

Position (mm)

Figure 63: Unicouple B2 voltage scan ofthe N leg contact.

83

 

 

 

 

0.0005 I I I I I f I T I T I I I I I I Ifir 1' l r I I Id

: Unicouple 82 g.

_ P Leg Contact Scan .

. Jonathan D'Angelo ,m " .

0.0004 b 5I23I06 , .3229 -

' __ -1 .92: '1

- o 442$cm “A; .

' ZT: 0.07117 £00m ' '

.60 '

A 0.0003 t- .M" -
2, - . o o ‘
8) t I 'l
.9 O -
:3 l- .. .
0.0002 l- o' ... H

L .... .

: 0 .0".

.. .0. .o' .

0.0001 - .J’,’ -

I ....396¥‘ d

. afi‘é‘.‘ .

0 m I l I I I I L A J L J l I I I j l I I I I

0 1 2 3 4 5

Unicouple BZ resistance is 11.8mQ and the room temperature ZT is 0.071 which is
higher than other unicouples with resistance above 10mg. The higher ZT in this
case may be contributed to better thermopower or lower thermal conductivity. The

measured ZTs of the other unicouples in bonding run B were as follows; 83-0.077,

B4-0.057, BS-0.062, and Bé-0.076. The ZTs are still lower than what is needed for

Position (mm)

Figure 64: Unicouple 82 P leg scan.

good module performance, but repeatability is also of great importance.

Unicouple run C used the same material as run B, but a 4 hour ramp was used

instead of 30 minutes. Only the p-type bonded to the 410 stainless steel. Thus

experiments using two stages during the bonding were investigated.

84

For unicouple run D the first stage of the bonding the n-type material
(ETN134) was bonded to the 410 stainless steel using a a 30 minute ramp time
followed by 4 hours of annealing at 7009C. After stage 1 of this bonding experiment,
the samples were removed from the furnace and inspected. Seven of the eight legs
were found to have bonded to the 410 stainless steel as shown in Figure 66. The
stainless steel was then lightly polished and the p-type material (ETP32) configured
on the electrodes. The unicouples were then placed back in the furnace for stage 2

of the bonding using a 4 hour ramp time followed by 4 hours of annealing at 7009C.

 

Figure 65: Stage 1 of bond run D.

 

Figure 66: Run D final unicouples.

Run D finished with 7 completed unicouples as shown in Figure 67. Large voltage
drops in the scans shown for the n-type and p-type legs of the unicouples as shown
in Figures 68 and 69 indicate significant contact resistance and cracking of the
samples.

85

Voltage (pV)

400

300

200

100

  

 

I
A

I
p I
I I
n -
I u
I I
I . 1
I d
n u:
I . I
I u
I d
r -I
0.. "
b I
in I
l' 1
i- d
L n
I . . q
I 1
- I“ c 1
v

D1 N Leg scan 2

 

O
a
N
(A)
A
U!

Position (mm)

Figure 67: Unicouple D1 n leg scan.

86

D1 P Leg scan 2

 

 

 

 

600 P j I 1 I I—r f I I I I I I I l I T I I l I I I I l I I j—I
- 1
E -l
500 "
400 "'
9 I
:1
7"; l-
g) 300 L- . 2'
6 .3
> p fia
I ‘t I
200 F 1&0 I-
I fig 13" . -
e” I
1 00 " / .-
2 ° I
0 - 1 I i l I I l J l l L l I I I l

 

 

-1 0 1 2 3 4 5

Position (mm)

Figure 68: Unicouple D1 p leg scan.

The ZT of the D1 unicouple was 0.077 and for unicouple D3 the ZT was 0.12. The
resistance of this D3 unicouple was 15mg, which was significantly lower than the
other unicouples in this experiment. ZT measurement for unicouples D4 and D5

were 0.078 and 0.099 respectively.

87

 

  
   
   

 

 

800....,.........,.........'
Unicouple DS ”’0' :

700 P Leg Contact Scan . _;
Jonathan D'Angelo .00 -

5I30/06 .0. ‘

O -

60° 0 ~ 250 3cm" 9. ;

ZT: 0.0987 .o‘ 3

500 .0 .-

S‘ ' 2
3 4
. Ii

:33. 400 .. -
>° .0' :
300 no 1

.4' -

200 9" ‘

.0

100 1

' i

0 l 1 l L J l L I L I l l l l J 4 L
o 1 2 3 4 5

Position (mm)
Figure 69: Unicouple D5 p leg scan.
D5 shows a good example of material that does not exhibit significant

cracking along the length of the leg. The contact resistance is approximately 500

ufl-cm2 and the ZT is close to 0.1.

Unicouple run G started with a tellurium deposition onto the n-type legs. The bond
was done with a 30 minute ramp and a 4 hour soak at 700°C. For this initial

bonding stage, all of the bonds were successful as shown in Figure 71.

88

 

Figure 70: Unicouple run G stage 1 results.

The second stage of bonding was also had 100% success as shown in Figure 72.

 

Figure 71. Unicouple run G final results. The dot on the material marks the p—type
material.

89

Voltage (pV)

 

500 ........-,.-...........___

' Unicouple G1 7'
Z P Leg Contact Scan
. Jonathan D'Angelo
400 -6/8/06

7. o~ 750$cm" .e" .
- ZT: 0.051 .‘ 0

 

300

200

I J_ l l J l

100

I I ‘l I I I I l I I ff FT

lLlLll

 

o
”flungw

W I
O LLllllllllllllllllli+lli

0 1 2 3 4 5

 

Position (mm)

Figure 72: Unicouple G1 contact scan ofthe p-type leg.

90

 

3m frII I IITI'IrI I l I II I I I I II

I Unicouple G3 ..

 

 

 

- N Leg Contact Scan “’9 §
' Jonathan D'Angelo
250 - '-
- 613/06 .9" M“ .
' 0 ~ 1240 .‘t‘ocm'1 ' '
. . ..o .
200 l- . n-
I . . u
’>‘ e e ’
3 - e e ‘
85 150 '- g d
m I I
‘6 0‘
> . e
. o .0.
100 - .0 ..
I 0"” "
50 - .0" -
D . . .
I d
0 M. L n n n I n 1 1 a L L 4 j_l J L 4 n 1 'l
0 1 2 3 4 5

Position (mm)

Figure 73: Unicouple G3 N leg contact scan. The ZT for this unicouple was 0.08.

91

Voltage (pV)

500 I I I f I I f jfi 1 I I’ I I I l I If I I r I I I I
' Unicouple G3 :
_ P Leg Contact Scan 1
. Jonathan D'Angelo ,1} -
400 - SIB/06 “,9 -
I , . q
- 0 ~ 600 Scm'1 “9., "J .
I a. II
. 0‘ .. .
300 - . . °' -
I u . . . q
F . Q I i
l' .. . .O. Ii
200 b ". 0'. j
P
r ” 0. 0" '4
' C. o. o. 4
I . . It
I .. ... u
. I
I .“g I
F ..:IG’ -
I a”. d
0 W.: I l L l 1 PL 1.; i L J A l l L I L I ‘
O 1 2 3 4 5

 

 

 

 

Position (mm)

Figure 74: Unicouple G3 P leg contact scan.

92

Voltage (pV)

500

400

300

200

100

0

 

 

 

 

I I I I I I 1—Ii I if I I I I I I I I I I I I I
* Unicouple G4 ‘
Z N Leg Contact Scan :
. Jonathan D'Angelo “fizz” .
- 6I8I06 o...“ .
I 0 ~ 720 Sam" ”if” I
; ZT: 0.088 flee: .
.' o ‘.
. a‘ ' .
: .9" :
- e’eagé .4
I: .3. I :
I ‘I‘aeflv I
I . I
I- 'i
Ml...lL4..LL44..I.L..I
0 1 2 3 4 5

Position (mm)

Figure 75: Unicouple G4 N leg contact scan.

93

Voltage (uV)

 

500 I I I II I I I I I I I I rT I I I I I I T I I

_ Unicouple G4

_ P Leg Contact Scan .0":

.. Jonathan D'Angelo o”.‘
400 - 6/8/06 ’5;

I 0 ~ 700 Sam“

300

 

200 '-
100 '-

O

in.
E ‘5‘.”
useeev¢99¢a¢o

OMIIIAIIAJIILIIILIIIII
O 1 2 3 4

 

Position (mm)

Figure 76: Unicouple G4 P leg contact scan.

94

lllllllllllllLJlllllllJl

 

OI

Voltage (pV)

250

 

. . . . .
' Unicouple GS
: N Leg Contact Scan
. Jonathan D'Angelo
200 -6flN06

I 0 ~ 2000 Sam"
. ZT: 0.065

150 '-

 

100 - '9'

o

0.3“...

50"

 

o
oo 0
mante-

0 I I I I j I l I I I I l I I 4 I I I I L I

I I I I I T r I l I I I I I I I.‘r
I

 

 

0 1 2 3 4

Position (mm)

Figure 77: Unicouple GS N leg contact scan.

95

Voltage (pV)

700

600

500

400

300

200

100

own..-

 

L
D
-
h
I

I

 

0

. . . . r . . . . . .
. Unicouple GS

I P Leg Contact Scan
_' Jonathan D'Angelo
I 6I9I06

'. 0 ~ 700 3cm"
: ZT: 0.065

1 2 3

Position (mm)

Figure 78: Unicouple G5 P leg scan.

96

IJJILLJiJII

4

I I

 

 

Voltage (pV)

 

 

 

 

500 I T I I ' I I I I I I— I 1 T l I I I I I I I r r
Unicouple 66 '
E N Leg Contact Scan 1
i Jonathan D'Angelo .
400 {619106 «223' -
... ' -
.o~7408cm" ’ .
' ZT: 0.087 ‘ “’ *
" 'i
300 - we“ ..
. .. .
. . .
2 o .
200 P . -
r f 4
in ... . d
C .o' E: 1
100 - .° . o' .
b ‘63. u
' fl". 1
0 m. I . . l . 1 4 4 . . 1 . . . . I . . 4 T
0 1 2 3 4 5

Position (mm)

Figure 79: Unicouple G6 N leg contact scan.

97

Voltage (uV)

 

800 A I I I I l’ I I I I l I I rfi I I I I I l I T r r

 

 

 

 

Unicouple GB

700 P Leg Contact Scan 0 -
Jonathan D'Angelo .' 7
6/9/06 . 3

600 0 ~ 560 3cm" ,' . '3
ZT: 0.087 .° .‘ ;

500 ‘ -
400 '2
-l

l

300 -
o 1

1

200 ‘ 1
. . E

100 o 1
fl .0 .

0 19!:.:?;;;:;:n..n I 1 1 A n I n 1 j n :

0 1 2 3 4 5

Position (mm)

Figure 80: Unicouple G6 P leg contact scan.

98

Voltage (pV)

400 ’

350

300

250

200

150

100

50

 

 

r I I I I I I III I I I rT' I I I I I I I I‘

I I I I I I I

1 2 3

Position (mm)

Unicouple G?
N Leg Contact Scan
Jonathan D'Angelo
6I7I06

0 ~ 1020 Sam"

I II I l I I

4

Figure 81: Unicouple G7 N leg contact scan.

99

I

 

IIIIIJILLIII

IJILII

I‘ll

 

 

   
 
  

 

 

350 I I I I II I I I ' I I I I I I I I I r I I I I .
Unicouple G7 “.9-

P Leg Contact Sean M

300 Jonathan D'Angelo ”03» 1

6/7/06 .90 M '

700 s " "ya .3

250 O " cm ‘

”05'” .

.0". 1

A o -l
33: 200 o. 0 1
o o .

§ .0 ”I o ‘
I: . .. I
(>3 150 .0 . 1
. Q q

I. e ° 9 I

100 .° 00. -

.. ..¢‘sm :

a" .

50 -

0 l l I I I I l J L I I J I I I I l I J I I %
0 1 2 3 4 5

Position (mm)
Figure 82: Unicouple G7 contact scan for the p-type leg. The measured ZT for this
unicouple was 0.04
Furnace run H was a 2 stage run of ETN137 and ETPZS to 410 stainless steel.
OUt of 20 initially put in the diffusion furnace to be bonded, 15 unicouples were

completed.

 

 

 

 

Figure 83: Unicouples fabricated from run H.

100

Unicouple run] was a two stage bonding to form a unicouple of ETN14O and
ETP25 to 410 stainless steel. Twenty two unicouples were fabricated from this
runand then used to fabricate modules as will be discussed later.

Unicouple run K was the bonding of ETN141 and ETP25 to 410 stainless
steel. With the use of the two stage bonding technique 19 total unicouples were
fabricated. Two consecutive bonding runs have shown results of good bonds with
different ingot materials. Unicouples from both bonding runs of I and K were used
in modules.

Unicouple run L was the bonding of ETN141 and ETP26 to 410 stainless steel
again. This time each 410 stainless steel was polished before bonding. The final
results of the experiment were no bonds of either the LAST or LASTT material to the
polished 410 stainless steel. Thus either the rough surfaces of the 410 stainless steel
assisted in the bonding process, or there were contaminants introduced during the
polishing process.

With unicouple run M a five hour soak was used. The extra hour was added
to possibly aid in annealing of the material along with diffusion bonding the contact.
ETN141 and ETP26 to 410 stainless steel and the two stage bonding technique was
used. Only 5 out 18 unicouples showed bonding. This low yield may have been
caused by a low furnace pressure, which in the past showed poor bonding results.
Also the bolts on the bonding device were not double checked to make sure they

were tightened down before bonding.

101

Unicouple run N was a bond of the same material as unicouple run M. This
time each bolt was well tightened on the bonding device and the two stage run was
used. Ten out of 13 unicouples successfully bonded.

Unicouple run 0 was a bond of slow cooled ETN143 to 410 stainless steel. A
30 minute ramp with a 5 hour soak at 700°C was used. Only 3 of the 9 legs bonded
to the 410 stainless. To investigate the influence of the mechanical pressure on the
bonding process, the was redone with all bolts further tightened increasing the
torque from 6 to 7 inch pounds. Four unicouples were bonded from this run. One of
the goals from unicouple run 0 was to see if slow cooled material (ETN143 was a
slow cooled ingot) would exhibit less cracks throughout ingots. Slow cooled ingots
have shown less cracking throughout the ingot.

Unicouple run P was a bond of ETN142 and ETP26 to 410 stainless steel. The
bonding surfaces were cleaned with acetone and methanol before loading into the
diffusion boat. Only 7 out of 14 unicouples bonded suggesting that surface
contaminants are not the cause of poor bonding yield.

Unicouple run Z was a bond of ETN151 and ETP33 to 410 stainless steel. A
two stage bond was used with a 30 minute ramp for the n-type bond and 4 hour

ramp for the p-type bond. Both used a 5 hour soak at 700°C.

102

 

400 I l I l T T

350 - f... -
A i 1."
> +5.“ 9 ;°'
3 250 - gee-34' J
a.) .— .’
E? 4 la’
3: 200 . ,.¢ g. _
O 1‘43” I
> o

39;?»
1 50 "r f

 

 

 

100 - “73:: ° -
M
50 I 1 I 1 l 1
-1 O 1 2 3 4 5 6

Position (mm)

Figure 84: Unicouple run Z, N leg scan.

The scan in Figure 84 shows a low contact resistance, but the bonding yield for this
run was below 50%. The best conclusion would be that there was not enough
mechanical pressure on each unicouple to allow bonding.

Unicouple run AU was the first run to use the new background gas pressure
control system in the diffusion furnace. The run was a bond of ETN173 and ETP40
to 410 stainless steel. The 410 stainless was being tried again because it has a
higher electrical conductivity than the 316, but the CTE is around 12ppm which is
lower than the CTE for 410 stainless steel. Only one of the legs came out bonded,

however none of the legs exhibited signs of outgassing. This showed that a flow of

103

the argon-hydrogen forming gas could be used during bonding runs to minimize

evaporation of the thermoelectric materials.

 

 

 

 

AU scan
300 T I l I I l
250 - o=1,224.5 (S/cm) . .4'
I . .
I
E 200 - 3 -
8, -’
0
g 150 ‘ '
s - a '
'0
O
.0
100 - . ‘ -
. O
I
O
O O
50 L l l l l l
-1 0 1 2 3 4 5 6

Position (mm)

Figure 85: Scan ofN leg from run AU.

Figure 85 shows a scan of a n-type leg that was part of bonding run AU. This
confirms no change in the electrical properties of the material. Samples in the past

have outgassed if the pressure was too low. A visual inspection showed no change

in the physical properties.

104

4.5 Diffusion Bonds of Sno,3Feo.3Te to LAST and LASTT Materials

An alloy of tin, iron and tellurium was made, Sno,3Fe0.3Te, and used at the

interface between stainless steel and the thermoelectric materials during a bond at

700°C with a 4 hour soak

 

Figure 86: Bonding results of LAST and LASTT to Sn-Fe-Te alloy.

Since the p-type material deformed during the bonding process a room temperature
scan of the contact was not possible. The n-type leg and contact was scanned for its

room temperature contact resistance shown below.

105

 

 

 

 

0.00045 , , , , ,

0.0004 - o _

0.00035 - -
2; . p...

O
0

2% 0.0003 .. . ‘H. q
—>6 ‘0’

0.00025 _ .’ .

I"
00002 ’
0.00015 . . 1 . .
—1 0 1 2 3 4 5

Position (mm)

Figure 87: Voltage scan of LAST bonded to a SnFeTe alloy.

As shown in Figure 87 the LAST material properties have changed. During the
bonding process, the iron, tin, and tellurium may have diffused and /or reacted with

the thermoelectric material causing the electrical properties to change [71].

4.6 Diffusion Bonds of Bi2Te3 to LAST and LASTT Materials

For higher efficiency devices, segmentation of the LAST and LASTT material
with Bi2Te3 can be used. A direct bonding attempt of LAST to Bi2Te3 was
investigated using a ramp time of 30 minutes and a soak time of4 hours at a

temperature of 450°C. The furnace pressure was set to 400 Torr of an Ar/Hz

106

atmosphere. The result of the experiment showed the LAST did not bond to the

Bi2Te3. This may be due to CTE mismatch between the two materials, not enough

bonding time, or not high enough bonding temperature or some combination of

these. The bonding temperature is limited to the melting point ofthe Bi2Te3, which

is approximately 300°C lower than the melting point of LAST material.

A colleague in our laboratory, Dr. Chun-l Wu also investigated a solder bond

between LAST material and Bi2Te3 using a lead-tin solder. Both the LAST and

Bi2Te3 were nickel plated before soldering, and an excellent bond was formed as

shown by the voltage scan in Figure 88. The room temperature ZT of the segmented

leg was 0.144. Similarly, an excellent solder bond between Bi2Te3 and LASTT could

be formed with the same process as shown in Figure 89.

107

 

600 Iii'TIYIII'II'I'II'I'ITYVIUUIIIUI'I".

 

 

 

: :
500 t LAST .,.-' .
; 0.. d
L ... d
O
I . . I
400 , .-' .
A l' .. d
> P .0
:x. . .0.
I? l ,0“
g0 300 _- ’0 -:
g I ... I
> : - ' " 2
100 '- .‘
.- cl
0 .111IllllnnlellillnlJiHLllLLllll+nll‘
0 1 2 3 4 5 6 7

Position (mm)

Figure 88: Segmented leg of LAST and Bi2Te3, measurement and data prepared by
Dr. Chun-l Wu.

108

 

350_.'...,.......fi......r.......t..tt.,.
: 0‘:
300'- LASTT ,0”. 5
: 0’... 2
. . .
250- _ .-°' 0
l- o. .
_ | .0.
9 I a...
1200. I ’0 _
v , g a
30 C ’0". j
a 150 -. P'B'zTe3 .- .-
§ . .0 .
I i.0°. j
100: ’0' -
; .0” l .
. o. l .
501......" l d.
' l
0 hl....J....|....1.#L11....JLPL.1....1.'

 

 

 

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

Figure 89: Segmented leg of LASTT and Bi2Te3, measurement and data prepared by
Dr. Chun-l Wu.

The length ofthe Bi2Te3 in this case is only 2mm. The ideal length of the material is

dependent on its properties and ZT may be maximized with adjusting the relative

geometry of either the Bi2Te3 or the LAST.

-A second attempt to bond Bi2Te3 to LAST at 450°C for 4 hours again

resulted in no direct bonds as shown in Figure 90.

109

 

Figure 90: Picture of failed bonds of Bi2Te3 to the LAST material.

4.7 Diffusion Bonds of Chromium, Iron and Indium Layers Between Nickel and
LAST and LASTT

This section describes bonding studies using pre-deposited layers of chromium and
iron between the thermoelectric materials and a nickel metal electrode. A unicouple
bonding run was done with a pre-deposition of chromium-iron onto n-type material.
The bonding was done in two separate stages. Stage one bonding of the n-type
material showed strong bonds. One of the samples exhibited melting indicating the
formation of some eutectic during the bond. A contact scan of the LAST to the

nickel is shown below.

110

 

160 I I r l I ..I’°'
140 - o a
o
.0
..
120 I- ... '0le '1
o 5'
on.“ >
A 100 - a... f/‘Ir' '-
> 0
3
o
g) 80 '- 0 mil EL "‘
E ”o
'5 ° '_
> 60 - .
o
40 _ . o a _
I

 

 

 

 

Position (mm)

Figure 91: Nickel to LAST bond with a thin film of chromium-iron.

The contact scan shows a high contact resistance as seen in Figure 91 but also a
change in the electrical conductivity of the LAST. Some of the chromium-iron film
may have diffused into the LAST material causing a doping change of the material.
The p-type material bonded well, but there was some deformation of the material.
Room temperature ZT measurements gave values of 0.08, 0.05, and 0.05.
Unicouple run Q was a bond of ETP31 and ETN147 to an indium coated

nickel electrode. The bond was done at 639°C for 10 minutes with a 2 hour ramp.

Only the n-type material bonded.

111

 

 

 

 

60
I r I I ....
o
50- ° -
O
40- -
9 . -"
3 30. ... ..
Q) 0
U)
2 20- .
O o
>
10i- -i
o
0- 0000000 -i
_10 1 J l l
-1 0 1 2 3 4 5

Position (mm)

Figure 92: Contact scan of unicouple run Q.

Unicouple run R was a bond of p-type material directly to nickel. The bond

was done at 725°C with a 4 hour ramp and a 4 hour soak. Only one ofthe legs

bonded with a contact scan shown in Figure 93.

112

160 T l I l I

120 .. .2”. _

 

 

 

 

31‘ 100 - 5% r'. -
‘1’ 80 - -
5’ I z"
6 o

60 - c -
> >2 “I:

40 - " _

20 - .1

0 l

-1 4 5

 

Position (mm)

Figure 93: Unicouple run R contact scan.

Furnace run 5 was a repeat of run R, but with a 6 hour soak and the bonding
temperature was 725°C. All of the legs bonded, but showed significant contact

resistances as indicated in the scans shown in Figure 94.

113

3500 I I I r I I I

3000 l. K d

2500

 

2000 -

1500

Voltage (pV)

1000

500

 

 

 

 

Position (mm)

Figure 94: Scans of LASTT bonded directly to nickel.

The high bonding temperature combined with the long soak time may have allowed
- nickel to diffuse into the bulk material. The diffusion of nickel would change the
properties of the LASTT material. The bond may be a mechanical success but does
not work electrically.

Furnace run T was a bond of ETP26 directly to nickel but with a lower
bonding temperature. The bond was done at 700°C with a 5 hour soak and a 4 hour
ramp. Below shows the room temperature scan of the contact and there is some
change in the electrical properties of the thermoelectric material in the first ~

1.5mm near the junction. This could result from diffusion of nickel into the

114

thermoelectric material causing a doping effect, or causing a mechanical change in

the material that results in significant cracking near the junction.

 

 

 

 

2500 I I I I r
2000 - :1?
A 1500 — .‘" .
> :..:
3 O
a) °
C, 1000 - _
9 = ‘
500 L- i” . -
: 0°.
0 - W - ~ I -
-500 l J l l I
-1 0 1 2 3 4 5

Position (mm)

Figure 95: Contact scan ofa bond from unicouple run T.

An experiment was done with an indium and gold deposition onto nickel to
help prevent oxidization of the nickel before bonding runs. For this experiment
ETN149 and ETP33 were used and 500nm of in and 150nm of gold was deposited
onto the nickel. A 4 hour ramp and a 1 hour soak at 640°C was used. The
experiment yielded no bonds. To test if the gold layer was inhibiting the bonding
process, another experiment was done at a similar bonding temperature.

Unicouple run BD was a bonding run with a film of indium to bond the LAST

to nickel in the configuration shown in Figure 96. The ramp time was 2.5 hours

115

with a soak temperature of628°C for 1.5 hours. Both n-type and p-type materials

were bonded in the same run.

LAST

   

<—— Indium Foil
Nickel

Figure 96: Experimental setup for bonding LAST/LASTT to nickel with an indium
foil.

Unicouple run BE was a bond like before with ETN179 to nickel with an
indium foil just like what is shown in Figure 96. The bond was done with a 30
minute ramp and a 1.5 hour soak at 628°C. Unicouples were fabricated and the

contact resistance is shown below.

116

 

 

 

 

700 I I I
.0
600 - .0 ..
0
.0
.0
500 — .0“ ..
0
> .o’
3. 400 - .o -
0
8 --
:93 300 - ,o' .
O 0
> 0
0
200 - -
0
100 I- ..
0
OGWl L l
0 1 2 3 4

Position (mm)

Figure 97: Contact scan ofthe n-type leg from unicouple run BE.

After trying to measure the module all ofthe bonds failed after several temperature
cycles.

A bond of an iron-chromium alloy to ETN151 and ETP33 along with another
nickel indium bond was done. One boat had 4 unicouples to be bonded with the
iron-chromium alloy and the other boat had 4 unicouples to be bonded with indium
to nickel. The bond was done with a 4 hour ramp and 4 hour soak at 700°C. Only
one p-type leg bonded with the indium to the nickel, and only one bond with the
iron-chromium alloy.

117

1000 I I I I I I

 

 

 

 

800 - ‘ ." ..
O
A \
O
3:; 500 - ,3 -
0 -"
E 400 . / -
> .
O
O
0 W I I I I
-1 0 1 3 4 5 6

Position (mm)

Figure 98: Voltage scan of the n-type leg to an iron-chromium alloy.

The bond between the iron-chromium has a high contact resistance and the scan
shown in Figure 98 shows some non-linear regions possibly indicating cracks which
are not parallel to the interface, or a distribution to the impurity concentration
throughout the material.

Unicouple run AM was a bond of ETN171 to pure iron. Four samples were
used and the bonding was done with a 30 minute ramp and 4 hour soak at 700°C.
Unicouple run AM showed that the LAST material does not bond directly to pure
iron under these conditions. There may have been an oxide layer on the iron that

prevented bonding.

118

4.8 Diffusion Bonds of Copper Beryllium to LAST and LASTT Materials
Unicouple run Y was a bond of ETN149 and ETP33 to a copper beryllium

alloy. The bond was done with a 4 hour ramp and a 5 hour soak at 700°C. During

the bonding run the LAST and LASTT formed a eutectic with the copper beryllium

and everything melted as shown in Figure 99.

 

Figure 99: Photo of the failed diffusion bond of LAST and LASTT to a copper
beryllium alloy.

The use of a copper beryllium alloy for the electrode is attractive because of
the relatively low electrical conductivity, relatively close matching of the coefficient
of thermal expansion with LAST and LASTT, and the availability of such alloys. The
use of copper beryllium alloys however, would require a suitable diffusion barrier to
prevent the eutectic formation. To that end, an electron beam deposition of 75.1mm
of tungsten followed by 60nm of gold on the copper beryllium interconnects was
studied. A quartz crystal monitor was used to monitor film thickness during the

deposition process.

119

 

Figure 100: Setup ofthe legs before the deposition ofW and Au.

 

Figure 101: Picture of setup after deposition with gold coating.

120

The bond was done at 700°C with a 4 hour ramp and 2 hour soak. A comparison of
Figure 99and Figure 102 does show a reduction in the melting of the samples with
the diffusion barrier, however it was not sufficient for the prevention of the eutectic

formation.

   

Figure 102: Results of bonding experiment after adding gold and tungsten to the
copper beryllium alloy.

4.9 Diffusion Bonds of 316 Stainless Steel to LAST and LASTT Materials
Unicouple run AA was a bond of ETN151 and ETP28 to 316 stainless steel.
The 316 stainless used has a thickness of 0.18” to reduce the resistance of the
interconnect. The bond was done in two stages; 30 minute ramp for the n-type
bond and a 4 hour ramp for the p-type bond. Both stages had a soak time of 5 hours
at 700°C and 4 out of 5 unicouples bonded. The room temperature ZT ofone of the
unicouples was 0.104 with a room temperature resistance below 5mg. These
results were among the highest ZT and lowest resistance of any unicouple studied in

this research.

121

 

 

 

 

160 I I I I
140 _ _
120 — _
g‘ 100 — .° —
7? I
g 80 _ . 4
% '..-
> 60 - . -
40 — -
20 — 1
o l I J I
-2 0 2 4 5 8

Position (mm)

Figure 103: Unicouple run AA N leg scan.

As shown in Figure 103 the n-type leg looks to have a crack in the center of the leg,

but still the overall unicouple resistance was below 5mg.

 

Figure 104: Unicouples for run AA.

122

Unicouple run AB was a bond of ETN153 and ETP28 to 316 stainless steel
and was done to investigate the repeatability ofthe results from unicouple run AA.

Unicouple AB n leg scan

 

 

 

 

140 I I I I I
120 - -
100 - ..
9
3 80 _ —
a)
O)
.53 60 - .
o
>
40 - -
20 '- I u
0
0 X l I k
-2 O 2 4 8 10

Position (mm)

Figure 105: Unicouple run AB N leg scan.

123

Unicouple AB p leg

 

 

 

 

300 I I I I F
250 — .3 -
0
A 200 .. .
>
3
g 150 . 4
£9
E
100 - .1
50 . -
I
I
0 J l I I
-2 0 2 4 8 10

Position (mm)

Figure 106: Unicouple AB P leg scan.

The scans shown in Figure 105 and Figure 106 show low contact resistances once
again for the bonding of LAST and LASTT to 316 stainless steel.

Unicouple run AC was a bond of ETN153 and ETP28 to 316 stainless steel.
The two stage run was used as before but this time the soak time was 6 hours at a

temperature of 700°C. The experiment was used to see how a longer soak time

would affect bonding results.

124

Module AC Unicouple 1 N
500 I I I I

 

"I.

450 r _
9
3
8, 400 - 'l
.59
6
>

350 - ' -

 

 

300 l l l l

 

Position (mm)

Figure 107: Unicouple Run AC N leg scan.

Figure 107 shows a contact scan ofthe n-type leg ofone ofthe unicouples from

bonding run AC. The contact resistance is below 20uQcm2, and there are no cracks

along the length of the sample. Below is a scan of a p-type leg from the same

unicouple which shows a contact resistance near zero and the leg has no cracks.

125

Module AC Unicouple 2 P leg

 

 

 

 

“550 r 1 r I
.0
-600 . 0: -
l
0’
A -650 - I; -
> O
3-
g, -700 _ .-
£9
6 .
> ’750 '- —I
-800 p 1
-350 JI I I
—2 0 2 4 6 8

Position (mm)

Figure 108: Unicouple AC P leg scan.

Two unicouples from run AC were put into a 4 leg module. The room temperature
ZT of the four leg module was 0.039. The individual unicouple ZT’s were not

measured.

126

Module AC Unicouple 2 N leg
-860

 

-880 - \

-900 - o

-920 -

-960 -

Voltage (0V)
is
O

-980 -

-1000 -

 

 

 

-1020 I I I I
Position (mm)

Figure 109: Unicouple 2 from AC run N leg scan.

Unicouple AC-2 had a room temperature ZT of 0.06.

127

Voltage (uV)

 

-300

 

 

 

I l I I
~350 L-
O
-400 -
-450 l-
_500 I l I I
-2 0 2 4 6

Position (mm)

Figure 110: N leg scan from a unicouple run AC.

128

 

 

 

 

50 I I I I
O — d
-50 _. -
3‘ -100 - o -
E:
(D
D) “150 '-' u-
5.9
§ -200 — -
-250 — ..
-300 - I a
W
_350 I I I I
-2 0 2 4 5 8

Position (mm)

Figure 111: P leg scan from unicouple run AC.

Figure 111 is a scan ofa p leg from a unicouple made in run AC. As shown in the
contact scans, many of the legs from this bonding run did exhibit cracks.
Unicouple run AD was a bond of ETN156, slow cooled material, to 316

stainless steel. A two stage run was used with a soak time of 700°C for 6 hours.

129

Unicouple Run AD N Leg scan

 

 

 

 

500 I I I I I
a
400 I- g -i
3 300 - -
a)
CI
5.9
‘5 200 - -
>
100 - ..
0 f! i I I
-2 0 2 4 6 8 10

Position (mm)

Figure 112: Unicouple run AD N leg scan.

This result shows the slow cooled material bonds well to the 316 stainless steel.
Unicouple run AE was a bond of ETN153 and ETP28 to 316 stainless steel.
The stainless steel strips were cut to 1mm thick. The bond was a two stage bond at
700°C for 6 hours, however only one unicouple bonded.
For a more intimate interface between the stainless steel and the

thermoelectric material, 500 nm of 316 stainless steel was sputtered onto LAST and

LASTT.

130

 

Figure 1 13: Module legs after a sputtering deposition of 316 stainless steel.

The sputtering of the stainless steel was done for 30 minutes. One of these legs was

then soldered to copper to test the contact resistance as shown in Figure 114.

JD_ss_thin_PbSn_Copper_01

 

500 l l I l I I l
,0
400 - a? -
/
300 - o. -

200 - ’5‘ .

Voltage (pV)

 

 

 

0 - m ..
-100 I L AL I I I I
-1 0 1 2 3 4 5 6 7

Position (mm)

Figure 114: Scan of copper soldered to LAST with a thin film of 316 sputtered to the
LAST.

131

These stainless steel samples were then brazed to nickel plated copper electrodes,
however the samples melted during the brazing process. A XXX nm thick stainless
steel layer (5 hour sputtering of 316 stainless steel) was also investigated , however
as before, the brazing of the legs caused the LAST and LASTT material to melt. In
addition, some ofthe XXX nm 316 stainless steel coated legs were first annealed at
700°C for six hours and then brazed to the nickel plated copper. Once again all of
the samples melted during the brazing process.

Unicouple run AF was a bond of ENT154 to a 0.02" thick 316 stainless steel
foil. A 30 minute ramp and a 6 hour soak at 700°C was used. Stage one ofthe bond
was a success. A 200pm layer ofbraze was placed on top ofthe stainless steel layer
and then bonded to nickel at 700°C for 90 minutes. The nickel brazed to the
stainless steel without forming a eutectic with the LAST. Below is a scan ofthe

nickel-316 stainless-LAST bond.

132

Unicouple AF N

 

400 I I I I I I

350 - -

300 '- q
.0

 

 

 

’>‘ 250 - -
3 4"
g, 200 .. -
£9.
g 150 - .
100 _ .
50 — J
O
0 1 l w l I I l
—1 0 1 2 3 4 5 6

Position (mm)

Figure 115: Scan of Ni-316-LAST bond.

As shown in Figure 115 there is significant contact resistance between the stainless
steel and the LAST material. This may have occurred after the brazing of the nickel
to the stainless steel, possibly caused by CTE mismatches among the materials used..

The bonding conditions for unicouple AG were a repeat ofthose used in run

AF and 8 out of 8 unicouples bonded.

133

 

3UU

 

200‘ .MM-

 

 

 

9 O
3 150 _ -
OJ
U)
3 100 - -
O
>
50 - -
O — all
0000000
_50 I I I I I
-1 0 1 2 3 4 5

Position (mm)
Figure 117: Scan of unicouple run AG N leg to 316.
Once again the large voltage discontinuity is shown between the LAST and 316

stainless steel. Thus the large contact resistance was not caused by cracking during

the brazing step. Unicouple run AH was a bond of 6 p-types to 0.02” thick 316

134

stainless steel. The bond was done with a 4.5 hour ramp and a 6 hour soak at 700°C.
After the bonding of the stainless steel to the LAST and braze of copper to the 316

stainless was done. The brazing temperature was 700°C for 90minutes.

Unicouple AH sample 1 #200

 

 

 

 

700 I I I I I I I
.C
600 - / .
500 - f .
’>" 400 - 17‘ .
E
8, 300 - .
(U
:2
g 200 - .-
100 _ -
000000
0 - 000000! -
_100 I I I I I I I
-1 0 1 2 3 4 5 6 7

Position (mm)

Figure 118 Scan of Cu-316-LASTT bond

After the scan was finished, the LASTT cracked off of the 316 stainless. Also shown
from the contact scan in Figure 118 the contact resistance is high. The 0.02"
stainless steel is not a good contact to either LAST or LASTT.

Unicouple run‘Al was a bond of 0.02” thick 316 stainless steel to n-type LAST.
The 316 surface was roughened before bonding to create more bonding surface
area. The samples bonded to the rough surfaces, but after brazing all of the samples

melted.

135

Since previous experiments (name the unicouple runs) showed low contact
resistances, the processing conditions for fabrication of 316 stainless steel foil
versus 316 stainless steel bulk materials was questioned. The 0.5mm 316 stainless
steel may not be formed the same way as the thicker, 4.6mm, stainless steel. The
different casting techniques of the stainless may change the properties of the
diffusion bonded LAST to 316.Unicouple run A] was a bond of 4.6mm thick 316
stainless steel to a hot pressed sample made from powder of ETN154. A 30 minute
ramp and 6.5 hour soak at 700°C was used. All bonds attempted were successful.

Scans of legs before bonding were done and then the legs were placed into
the furnace for bonding and then scanned again. The bond was done with a 30

minute ramp and 6.5 hour soak at 700°C.

 

Figure 119: Picture of bonded hot pressed module.

136

MSU HP1 to 316 Stainless

 

 

 

 

6000 h I I i l I I I 'fiil I I I f T I I I l f u r 1 I‘
b . I
I- . d
l- . fi
5000 '- 0 _
I . ‘
l- . ..
F c. ‘
4000 - 0° .1
C Pre-Bond .. 1
C
S : 0"40 \ .0 ..
3 3000 .- éfi :
a) I 4f» .1
9 “ 0* .
= I- “ .1
g 2000 7 .11. Post-Bond j
I . .
: 316 Stainless . 0~400 l
1000 - .* .3
' /\ 0 ..
: 1‘ Q :
0 _ ___/ 1
_1000 I. I l L I 1 1 J I I I l I l I l l l 1 l l . .1
'2 O 2 4 6 8 10

Position (mm)

Figure 120: Scan ofhot pressed leg before and after bonding.

Figure 120Error! Reference source not found. shows how the conductivity
increased by 10 times after bonding. After scanning the sample it was broken

intentionally to test the strength of the mechanical bond.

137

 

Figure 121: Fractured legs bonded to 316 stainless steel.

Breaking of the bonds was not easy, in-fact the leg fractured before the bond. After
the leg fractured approximately 80% of the bonded contact remained on the
stainless steel for the p-type leg, and approximately 30% of the n-type leg remained

bonded to the stainless steel.

138

 

 

 

 

 

1000 I I I I I t I I r T I I I I I I 1 I I I 1 I 1
- 1
800 " "'
i- d
g 600 '- -
3 I- .
s. . .
E
6
> 400 - -
200 " -'
- -1
O I .l I I_I_ I I I j. I I I l l I I 1 I L I I I I
-2 0 2 4 6 8 10

Position (mm)

Figure 122: Scan of P leg contact to 316 stainless steel.

The contact resistance in Figure 122 is approximately 19uQcm2. This was a scan

from the p-type leg before it was cracked off in the bond strength test A second

unicouple was scanned and had a higher contact resistance, 87chm2. When it was

cracked only about 40% of the material remained bonded as shown in Figure 124.

139

 

Figure 123: Crack of leg with high contact resistance.

Unicouple run AK was a bond of ETN164, EETP38, and 0.18" thick 316
stainless steel. This was done in a two stage bond. Stage 1 was a 30 minute ramp,
and 6.5 hour soak at 700°C for bonding the n-type leg to the stainless steel. Stage 2
added the p-type leg using a 4 hour ramp and 6 hour soak at 700°C. Only 3 modules
bonded. This could be a direct result of non-uniform mechanical pressure on the
individual legs during bonding.

Run AL was a bond of MSUHP4, p-type hot pressed material, to 316 stainless
steel. A 4 hour ramp, 6 hour soak at 700°C was used. The bond results showed that
the p-type hot pressed material bloated about 10-15% larger than their original

volume, but still bonded. Below is a scan of the contact resistance.

AL Ptype hot pressed

 

 

400 I' 1 i 1 1' i 1 I i I I l I T 1 r t

350

300

MSUHP4 0~720 S/cm 3;?“

A 250
> i;
3 ,3"
CD .2
5’ 200 ‘ ’53..
6
>

150

316 Stainless Steel if?
100
50 ‘1‘
0 l L L l l I l l l L I I. l l l
0 2 4 6 8 10

Position (mm)

Figure 124: Scan ofp-type hot pressed material bonded to 316 stainless steel.

Figure 124 shows that the p-type hot pressed material makes a good contact to the
316 stainless steel.

Run AN was a bond of tin-tellurium coated 316 stainless steel to ETN171. A 30
minute ramp and 6 hour soak at 700°C was used. 2 out of 4 samples bonded. The

two that bonded were strong mechanical bonds.

141

AN N scan 1

 

 

 

 

 

350 I I I I I ...
0’.
300 l- -i
250 - ..
9
3 200 - -
Q)
8’
g 150 F' ..
>
100 - .
50 - ..
-::- 3:: W
0 W ""l l 1
-1 0 1 2 3 4 5

Position (mm)

Figure 125: Scan of N leg bond to tin-tellurium coated 316 stainless steel.

The 316-LAST contact scan is shown in Figure 125 has a voltage jump over 275uV
which is a contact resistance well above ZOuQ-cmz. This shows with an addition of

tin-tellurium to the contact surface increases the contact resistance.
Unicouple run A0 was a bond of MSUHP-Z to 0.02” thick 316 stainless steel.
The bond was done with a 30 minute ramp and 6.5 hour soak at 700°C. The results

of the run were bloated hot pressed legs and high contact resistance.

142

A0 N scan 1

 

 

 

 

 

1.4104 , , , , To. .
.0
1.2104 - "“' a
O
1104 - -
’5
3 8000 p I
(D
(3)
$3 6000 . -
O
>
4000 - a
2000 _ _
0 H I l l J
-2 0 2 4 6 8 10

PosMon(nun)

Figure 126: Scan ofMSUHP-Z to 316 stainless steel.

Unicouple run AP was a bond of ETP38 to 0.02" thick 316 stainless steel. A 4
hour ramp with a 6.5 hour soak at 700°C was used. Two out of the 2 samples

bonded, but exhibited high contact resistance as shown in Figure 126.

143

 

 

 

 

300 r I I I I
I
250 - -
A 200 - -
>
3-
(D
C) 150 - -
E
B
> 100 - ..
50 - -
H
0 I l l l 1
—2 0 2 4 6 10

Position (mm)

Figure 127: Contact scan of P leg to 0.02" stainless steel.

The contact resistance of the p-type LASTT to thin 316 stainless is around

ZOOuQ-cmz. This is too high for an efficient thermoelectric device.

Several results have shown that all of the bonds do not come out successful
when other bonds from the same run do. This may be due to the fact that the
individual mechanical bonding pressure from unicouple to unicouple may not be
constant. A new bonding device was developed to individually monitor the

mechanical pressure of each leg.

144

 

Figure 128: Photograph of new bonding device.

This bonding device bonds one leg at a time. A torque wrench was used to maintain
better uniformity in the mechanical pressure used from leg to leg.

Unicouple run AQ was the first bonding run to use the new bonding device.
The bond was ETP38 to 316 stainless steel. The bond was done with a 4 hour ramp
and 6.5 hour soak at 700°C. The contact scan of the bond shows that it was a low

resistance contact as seen in Figure 129.

AQ

 

 

 

 

200 l T l I T
150 - I ..
3 100 - 4
(D
0')
£9
'5 50 - .
>
O - 2 -
_50 l l l l l
-2 0 2 4 6 8 10

Position (mm)

Figure 129: Scan of P leg bonded to 316 stainless steel with new bonding device.

The bonding procedure was repeated in unicouple run AT to fabricate many
unicouples using the new bonding device. The two stage bonding technique was

used and 26 total unicouples were bonded from the two runs.

 

.':'17:???“21'5533I‘Tr’i-‘Z—v‘jm...Ian-.3". _' 1 -Z.

  

2‘3“" an“ ;, '
7‘ l ' 23‘, "

.
12. s
r _

 
  

       

Figure 130: Unicouples bonded from runs AT and AS.

146

 

Figure 131: Fractured sample from run AT.

In Figure 131 is a picture of a fractured unicouple. The p-type leg fractured leaving
material bonded to the stainless, however the n-type leg cracked off without leaving
LAST material behind. The diffusion of the LAST may be slower than the LASTT and
a longer bonding time may be necessary. To reduce the total resistance of the metal
interconnect, unicouples from run AT were machined to have most of the stainless
steel cut off, followed by a brazing of nickel plated copper to the electrode.
Unfortunately all of the unicouples were destroyed during the machining step. The
cut was done too close to the contact.

Unicouple run AV was a bond of 9 ETN174 samples to 316 stainless steel.
The argon-hydrogen gas was flowing at 1 liter per minute with a chamber pressure
of 770 Torr. A 30 minute ramp and 6.5 hour soak at 700°C was used. Below is a

contact scan of a sample bonded in the run.

147

 

 

 

 

 

500 ...,...,...,...,.1.,,fl
C
400 -
3:; 300:
(D
U) ..
L“ .
6 200 -
> .
: 316 Stainless Steel
100 -
i <
0 .M I I I I I I
-2 0 2 4 6 8 10

Position (mm)

Figure 132: Scan of ETN174 bonded to 316 stainless steel. The horizontal line
represents a contact resistance of ZOuQ-cmz.

Unicouple run AW was a bond of ETN173 and ETN174 to 316 stainless steel
with a soak temperature of 800°C. The ramp time was 1 hour with a 6.5 hour soak.
The higher soak temperature was used to see how it effects the bonding results. The
higher bonding temperature did not melt the samples but they did change in color

as seen in Figure 133.

148

 

Figure 133: Photograph of bond of LAST to 316 with a soak temperature of 800°C.

Figure 133 shows how the material bonded to the 316 but then the leg cracked off
while preparing the sample for a contact scan. Stage 2 of AW used samples of
ETP40with a 4.5 hour ramp and 6.5 hour soak at 700°C. Only 3 total unicouples
were fabricated from this bonding run. Future experiments would not use the soak

temperature of 800°C.

149

Voltage (pV)

350

300

250

200

150

100

50

AW P-Leg scan

 

 

 

 

bIII'IIIIIIrrIII'IFIIIII
r

b

' O
I

I

- 0
II

I

I

b

I

h

P

-

- O

I

'_ O

' w
.IIIIIIIILIIIII4LLIIIIIL
-2 0 2 4 6 8 10

Position (mm)

Figure 134: Scan of P leg from unicouple run AW.

150

AW N-Leg scan

 

 

 

 

 

600 ...,..fi,. .
500'- J

A 400'- l
> P 'l
3 ' 1
a, I I
m 300' "‘
4g ' . :

E

200- 'l

I .

100- -

. d
0.1-fig“ . a

-2 o 2 4 6 8

Position (mm)

Figure 135: N leg scan ofbonding run AW.

The conductivity of the n-type material after bonding at 800°C was 902.9 8 / cm.
This gives insight on how the material reacts at higher temperatures for long
periods of time. Even though the properties did not change from bonding run AW,
the yield was low suggesting longer bonding times and / or higher mechanical
pressure on the interface may be needed.

A hot pressed experiment was done with powdered 316 stainless and
powdered LAST. The hope was that the stainless steel and LAST would sinter
during the hot pressing run. A voltage scan across the interface between the hot

pressed stainless steel to LAST is shown in Figure 136.

151

 

 
    
    

 

 

 

I r I
8000 - Hot Pressed 15 -
. 0:42.? (S/cm) .
3 6000 - -
CD . .
C)

59 . .
'5 4000 - _
> I Ii
l- P .
2000 - ‘ 1
. Pressed Stainless Steel 316 .

. 4" 0:32.84 (S/cm)

0 . I I I I I I I I I I I I I I I I I I 4

-2 0 2 4 6 8

Position (mm)

Figure 136: Scan ofhot pressed 316 and LAST.

The conductivity of cast stainless steel 316 is nearly 13,000 S/cm, but the hot
pressed powder done in this experiment was only 32.84 S / cm. The large differences
in the optimal pressing and Sintering conditions of stainless steel compared to those
for LAST are believed to be the cause of the relatively low electrical conductivity for
the stainless steel. Annealing of this sample at 700°C for 6.5 hours did not improve
the stainless steel, but rather lowered its conductivity to 3 S / cm.

Unicouple run AX was a bond of ETN173 and ETN174 to 316 stainless steel.

The soak temperature of the run was 720°C for 12 hours. The bonding time was

152

taken to an extreme to see how the bonding results. The pressure was maintained

at 780 Torr with a flow of 1LPM ofan Ar/Hz forming gas.

 

 

 

 

AX-N
350 p' I I I l I I I I I I r I I I I I I I l I I W] I I I I1 1 I I I I I r I I ‘
I .C :
0' .
300 ’- ..o‘° 1
0.. ~
.
250 "' —i
.'. .. :
’>" . I .
3: 200 - ,' 1
CD .- " ..
g) I . I
t 150 - ..
O " ..
> : .1
100 I- ! .:
: . I
I o” .
50 _- .ooooo -
0hIIIIIIIIIIIIIIIIIIIIIIIIILIIIIIIIIIIIII

-0.5 O 0.5 1 1.5 2 2.5 3 3.5

Position (mm)

Figure 137 Scan of N leg contact scan from run AX

153

Figure 138: EDS maps of unicouple run AX bond.

    
   

.
. \

‘ ‘4
.-‘ ...
‘-<
c

. . . ' . .4 iv:
1. - ‘ n: \
48*:‘tu'rv‘t- "1’
‘.'I. t‘rz“ ‘35-‘61?de-
- .7. urn-r (w 'k =3
i a =er
[J's-"I5: "

7.
.‘t-‘u A_

  
  

 

    
 

  

;-‘:-‘é:.fifl I
gr”; '.
- ‘4:

 

154

 

    

1.7

Figure 138 Contiued: EDS maps of unicouple run AX bond.

Error! Reference source not found. shows an EDS mapping of the contact
between the stainless steel and LAST material after a 12 hour bonding at 720°C.
The contact area appears to have plastically deformed.

Unicouple run BA was a bond of ETN 176 and E'I‘P42 to 316 stainless steel. A
two stage bond was used. Both stage soak temperatures were 700°C for 4.5 hours.
Unicouple run BB was similar to bonding run BA but used all hot pressed materials.
It was a bond of HPMSU12-N and HPMSU18-P to 316 stainless steel. Stage 1 of
bonding used a 30 minute ramp and 6.5 hour soak at 700°C. It had a flowing gas
rate of 1LPM maintained at a pressure of 790 Torr. None of the bonds resulted from

this stage of the bonding. Stage two also showed no bonds between the p-type

155

material and the 316. No conclusions were made as ofyet on why the hot pressed

material did not bond to the 316.

Unicouple run BC was a bond of ETN179 to 316 stainless steel. Stage one of
the bond was done with a 30 minute ramp and 6.5 hour soak at 720°C. Stage 2 of

the bond was done with a 4.5 hour ramp and 6.5 hour soak at 720°C. Unicouple run

86 fabricated 8 out of 11 modules.

 

-100 - -

-200 b ‘\“~“ _

-300 l- \ a

400 _ '\ -

Voltage (pV)

 

 

 

\
_500 1 1 1 l l
-2 0 2 4 6 8 10

Position (mm)

Figure 139: Unicouple run BG P leg scan.

156

Voltage (pV)

BG_P_leg_sample_2

 

 

 

 

250 j ]_ T T
villi"
200 - . _‘
150 - ..
100 l- ..
50 l- -
0 l J l l
-2 0 2 4 6

Position (mm)

Figure 140 BC sample 2 P leg scan

157

10

BG_N_leg_sample_2

 

 

 

 

1000 I T T l I
900 ‘ \ A
800 - . ..
’>‘ 700 _ \ -
3 .
an
o, 600 '- ‘ .-
59 .
'6 i
> 500 - ‘ -
400 - i ..
300 . ‘ .
9
M
200 J l l l l
-2 o 2 4 6 8 10

Position (mm)
Figure 141 Unicouple run 36 sample 2 N leg scan
Unicouple run BH was a bond of MSUHPZO and ETP44 to 316 stainless steel. The

bond was done with a 4.5 hour ramp and 6.5 hour soak at 720°C. BH run completed

with 5 unicouples bonded. One unicouple was scanned while the other were

machined for segmentation with Bi2Te3.

158

Voltage (pV)

BH_N_Ieg_sample_1

 

 

 

 

 

0 I I f I I.
-200 - _
—400 - -
'600 '- u-I
~800 - -
-1000 - -
-1200 - ; ..
.
-1400 __J-I-n 1 I I I
-2 0 2 4 6 8 10

Position (mm)

Figure 142: Unicouple run BH N leg scan.

159

BH_P_leg_sample_1
'1 450 I I I I I

 

 

 

 

4500 - \ .
‘1550 h \ u
‘5
9 -1600 - .
3
g, -1650 - _
£9
g 4700 - _
O
4750 .. ~ -
-1800 - \ -
_1850 l l l l I
-2 o 2 4 6 3 10

Position (mm)

Figure 143: Unicouple run BH P leg scan.

Unicouple run Bl was a bond of ETP44 with stainless steel bonded on both
sides. The bond was done with a 4.5 hour ramp and 6.5 hour soak at 720°C. During
the run there was a flow of argon-hydrogen at a pressure of 790 Torr.

Unicouple run B] was a bond of HPMSUZO-N and ETP44 to stainless steel

316. The bond was done with a 4.5 hour ramp and a 6.5 hour soak at 7209C.

160

 

Figure 145 Unicouples fabricated from run B].

161

Voltage (pV)

250

200

150

100

50

 

 

 

 

Position (mm)

Figure 146: 811 P leg scan.

162

10

Voltage (pV)

1000

900

800

700

600

500

400

300

200

 

 

 

Position (mm)

Figure 147: 811 N leg scan.

163

 

Voltage (pV)

250

200

150

100

50

BJ2-P-Leg

 

 

1 I

 

 

Position (mm)

Figure 148: BIZ P leg scan.

164

10

Voltage (pV)

1200

1000

800

600

400

200

 

 

 

 

10

 

Position (mm)

Figure 149: BIZ N leg scan.

165

Voltage (pV)

BJ3-N-Leg

 

1200 I I I I
1000 - -‘.
800 - ‘
600 ..
400 -

200 -

 

0 I l l J_

 

 

-2 0 2 4 6

Position (mm)

Figure 150: 313 N leg scan.

166

Voltage (pV)

BJ3-P-Leg scan 1

 

1300

1250 -

1200

1150 -

1100 -

1050

 

I I I

I

l

i

‘F—'

 

 

1000

Position (mm)

Figure 151: 313 P leg scan.

167

6

10

Voltage (pV)

BJ4-P-Leg scan 1

200 —‘ l l l I

 

150 - _

100 - .1

50- ‘ -
\

 

 

 

 

0 - _
\
_50 1 1 1 1 1
-2 0 2 4 6 8 10

Position (mm)

Figure 152: 814 P leg scan.

168

BJ4—N-Leg scan 1

 

 

 

 

1000 I I I I ’—
4'
900 i- " -
800 - ..
S‘ 700 - -
3;
CD
3 600 - . -
:5 f
> 500 - {I _
400 . 3' ..
.‘
o.
300 b .- ..
O
200 g 1 1 1 1
-2 0 2 4 6 8 10

Position (mm)

Figure 153: BM P leg scan.

As shown in the figures above most of the scans show good contact resistance and
the material does not have many cracks. The average resistance of the unicouples in
run B] was 10mg. The unicouples are nickel plated and then coated with QS/B4.
The QS / B4 is used to inhibit outgassing of the unicouple legs during high
temperature operations. The legs are nickel plated to protect the unicouples from
the Q5 / B4, due to the fact that it is very basic and eats away at the legs during

curing.

169

 

 

Figure 154: The unicouples were nickel plated for preparation to be coated with
S / B4.

Unicouple run BK was a bond of MSUHP24 and ETP44 to 316 stainless steel.
The bond was done with a 4.5 hour ramp and 6.5 hour soak at 720°C. The gas flow
was at 1 LPM at a pressure of 800 Torr. All 4 unicouples attempted successfully
bonded.

A bonding run ofseven HPMSU30—N to 316 stainless steel was done. The 316
stainless steel was polished with 800 grit sandpaper. The MSUHP30-N was then
polished lightly with 1200 grit sandpaper. The torque on the bonding device was 6
inch-lbs. The first stage of bonding was done at 720°C for 6 hours with a 4 hour
ramp. Unfortunately, none of the samples bonded after this first stage. Either it was
the hot pressed material not being able to bond to the 3 16 or the polishing of the
316 and the LAST caused poor bonding conditions. Past results have also shown

failed bonds when polishing of the legs was involved.

A new bonding device was developed to have a spring loaded mechanical
pressure during bonding. Using this device (Figure 155), a sample of ETP44 was

bond to 316 stainless steel at 720°C with a 4.5 hour ramp and 6.5 hour soak.

170

 

Figure 156: Long view of spring load device from main head down shaft

171

 

Figure 157: Base mechanism of spring loaded device which houses the spring.

The spring loaded bonding structure was fabricated to keep the spring in the cold
section of the furnace. This would help assure a constant pressure on the leg and

stainless during the bonding run.

172

ETP44 NEW Spring Load

 

 

 

 

200 I I I I I I I
O
C
C
o"
150 - if -
9
3.:
(D
o) 100 " ’ -1
59 a'
$5 r”
50 - ’ .
We
0 l l l l l I
-1 1 2 3 4 5 6 7

Position (mm)

Figure 158: Scan of p-type contact to 316 stainless steel with spring loaded bonding
device.

Figure 158 shows the scan of a resulting bond using the spring loaded device. The
new bonding device is good for bonding, but because of its physical structure only 3

or 4 legs can be bonded at once.

A hot pressed leg, HP30, was placed into the spring loaded bonding device to
be bonded to 316 stainless steel. The hot pressed material plastically deformed due
to creep. The small grain size in the hot pressed material allows it to creep which

can be advantageous in forming intimate contact at the junction.

173

An EDS study of p—type LASTT to 316 stainless steel was donebefore and
after a post bonding annealing step. Figure 159 shows the EDS of the sample before
annealing.

Figure 159: EDS of LASTT bonded to 316 stainless steel before annealing

 

 

174-

 

 

 

   

. u . . - ~-
. ‘ . - w.
- . 1 a
a -'
.1 ~
.3", .
, .
ls .. " . ‘
I .5. '
. ’ ‘ A
.- '~i- ' '
‘ 4
H v
a 1 l'
- .‘ ' ' .
. I '
. .1 . . --

. . ‘ I}. ~' .
-. - ‘.. G a" -i '3 . —
.‘ l ' iv. . I . . I -' .

' Te ' ‘5 ' ‘ ' Pb

Figure 159 Continued: EDS of LASTT bonded to 316 stainless steel before annealing.

Figure 160: EDS of LASTT to 316 after an anneal at 620°C for 4 days.

  

519mm Imloe‘l 2mm Scale Bar

175

 

 

Fe Cr

Figure 160 Continued: EDS of LASTT to 316 after an anneal at 620°C for 4 days.

The anneal was done at 620°C for 4 days in an argon-hydrogen atmosphere at 800
Torr. As seen from the EDS maps above, the nickel and chromium may have

176

diffused into the LASTT material. Lead looks to have diffused into the 316 stainless

 

   

 

 

 

steel.
P-316 post 4 Day Anneal
300 I I l l l I I
250 i-
A 200 -
>
3
a)
o, 150 -
£9.
6
> 100 .
5O -
a .
0 f." 1 ¥ 1 l l l
-1 0 1 2 3 4 5 6 7

Position (mm)

Figure 161: Post anneal scans of LASTT to 316 contact.

The SALT, Nao,8szoSbo,4Te22, material has exhibited a high ZT [72], and an

attempt was done to bond the SALT material to 316 stainless steel using 720°C for 6
hours with a 4 hour ramp. The torque on the set screw on the bonding ram rod was

set to 61bs-inches. The gas was set to a pressure of 800 Torr and a flow of 1 LPM of

Ar/Hz gas was used. The Na0_8szoSbo,4Te22 material did not bond to the stainless

steel and a longer bonding time may be necessary.

177

A quenching experiment was done ofa diffusion bond of316 to LAST. An N
leg was sealed in a quartz tube and placed into a furnace for eight hours at 600°C.
After the soak it was quenched in a water bath. Contact scans were done before and

after the quench to see the effects of the rapid temperature change.

1000‘YUITI‘IYTV‘IUTyI'ITI
I

800 - [,1 -

 

   
  

' Quenched Scan ‘

 

Pre—Quenched Scan

3 600
a)
c»
9 as!”
'5 400 a5 -
> .
.
- 4
200 -+
- s

 

 

 

6 8 10

Position (mm)

Figure 162: Before and after scans of quenching experiment.

The quenched sample exhibited many cracks in the sample showed by the voltage
scan in Figure 162. The different colors in the scans are the measurements that wre
taken across the entire sample. This proves if the module were to undergo rapid

changes in temperature, it would destroy the module.

178

4.10 Diffusion Bonds of Molybdenum to LAST and LASTT Materials

Six n-type LAST legs were coated with a 20nm molybdenum film as a
diffusion barrier. These samples were then attempted to be bonded to 316 stainless
steel. The bond was done with a 30 minute ramp and 6.5 hour soak at 700°C. No

bonds resulted from the experiment.

4.11 Diffusion Bonds of 347 Stainless Steel to LAST and LASTT Materials
Unicouple run AR was a bond of ETN137 to stainless steel 347. The 347 was
used because it has a lower resistance than stainless steel 316 along with having a
CTE of 18 ppm/K. The bond was done with a 30 minute ramp and 6.5 hour soak at
700°C. The cast n-type material did not bond to the 347 stainless steel.
Unicouple run AS was a bond ofHP n-type to 347 stainless steel. The run
conditions were the same as run AR. The hot pressed LAST did not bond to the 347

stainless steel.

4.12 Diffusion Bonds of ‘Maganin' to LAST and LASTT Materials

An alloy of copper, manganese, and nickel was attempted to be bonded to
LAST and LASTT. The alloys name is Manganin and has a coefficient of thermal
expansion (CTE) of 13-19 ppm/K. ETN154 was bonded to the Manganin with a 4.5
hour ramp and 6.5 hour soak at 720°C. The end result was a melt of both the

Manganin and LAST.

179

5.0 Module Fabrication and Measurements
5.1 Long Term Module Testing System (LTMTS)

5.1.1 System Fabrication

The system’s main chamber is a stainless steel cylinder 5 inches in diameter.
The structure ofthe main chamber is designed to pump down to pressures below a
milli-Torr, and to be pressurize slightly above atmosphere. The flanges on either
end of the main chamber have 150 flange connections. With the quick connect
flange, it is easy to add or remove gas, electrical, or any other connection needed for
the system. The stage of the system, which is also the cold plate, is welded directly
to the main chamber flange. The cold plate is a liquid cooled plate that is controlled

by a chiller. The chiller has a maximum cooling capacity of 215 Watts.

 

Figure 163: Long Term Module Testing System. A) Gas port; B] KF-40 quick connect;
C) 150 Flange; D] Main Chamber; E) Chiller feed throughs for cold plate; F]
Vacuum Port.

180

   
   
         
  
    

Inside Outside

Heater Power

Module Output

TC Cold
TC Hot

Module

 

   

Heater /
I I I A

Cold Plate Water Port

7

Figure 164: Schematic drawing of long-term module testing system.

The heater for the hot side of the thermoelectric is fabricated out of a 140m
thick nickel-chromium foil. During operation of the heater from room temperature
to 150°C, the heater resistance varies from 29 to 1.89. As shown in Figure 165
below the nickel-chromium heater has a temperature dependent coefficient of
resistance of 0.001755 Q/K. The nickel chromium heater has a very small

dependence on temperature.

181

2.5

 

 

 

 

2-
(D
0
g
37.” 1.5—
U)
G)
I
(U
CD
I
0.5—
0
20

40

60 80 100 120 140 160

Heater Temperature (C)

Figure 165: Heater Resistance as a function of temperature.

 

Figure 166: Nickel-chromium meandering line heater.

Figure 166 shows the nickel-chromium heater. The electrodes are connected to the

heater with a nut and bolt assembly. The leads to the heater are attached to the cold

182

plate to keep the wire insulation from melting during operation. The heater is
powered by a 1200W power supply. Thermocouples are attached to the heater to
measure the temperature for feedback to the power supply. The placement of the
hot side thermal couple is at the base of the heater next to the module’s hot side.
The cold side thermocouple is placed next to the module on the cold plate (shown in
Figure 164). Computer control and data acquisition is maintained using software
developed in the LabViewTM programming environment. The software is used to
monitor the temperatures ofthe hot side and cold side of the module and to set the
current that is being supplied to the hot side heater.

A Keithley 2182 Nanovolt Meter is used to measure the output voltage to the
load. A True RMS AC/DC Mini Clamp Meter measures the current to the load by
magnetic pickup from the load wires. Combinations of resistors are used for loads.

The system uses a tri-scroll dry pump with a base pressure of 1mTorr. High
vacuum is not needed for this system. A tank of argon is used to pressurize the main

chamber with an inert gas.

183

 

 

Figure 167: Complete system. A) Nano Voltmeter; B) Main Chamber; C) Current
Meter; D) Temperature Monitor; E) Power Supply; F) Chiller; G) Argon Gas
Cyhnden

5.1.2 Experimental Procedure

The module is placed on the cold plate using a thermal grease. For this study
a Tellurex power generation module was used as a reference sample[73].
Thermocouples were placed on the hot side and cold side to monitor their
temperatures. The module leads were soldered to 14 gauge copper wire feed-
throughs. The nickel-chromium heater is bolted to a separate pair of 14 gauge

copper wire feed-throughs.

184

 

A B C D E E

Figure 168: Test Setup. A)Thermocouple; B)Bolt heater connection; C)Nickel-
chromium heater; D)Thermoelectric module E)Module leads to feed-through
solder connection; F)Thermally grounded leads.

The standard procedure is that the cold plate is placed into the main chamber, and
the chamber is initially pumped down with the roughing pump. The chamber is
then backfilled with argon gas followed by pumping down to 1mTorr and this
backfill/pump down procedure is repeated at least three times to ensure a better
oxygen free vacuum. After sufficient pumping, the system is filled with argon to
approximately 780 Torr. The chiller is initially set to 30°C. Since the heater is
custom made, an initial slow heating is done to monitor the power input vs.

temperature of the nickel-chromium heater as shown in Figure 169.

185

16,,
140
120 .
40

 

Hot Side Temperature (C)

0 S 10 15 20 25 30 35 40

Power to Heater (W)

Figure 169: Power to hot side heater and corresponding temperature. Note: At
heater power 30Watts the chiller temperature was increased from 30°C to

35°C. All temperatures are measured under an open circuit condition.

The effect of heat flow through the module is clearly seen in Figure 169 by a change
in the cold side temperature (chiller) by 5°C. Under an open circuit condition, the
input power to the heater of 37 watts gives the maximum operating temperature of
the Tellurex module, 150°C. Once a load is applied to the module the Peltier Effect
will begin to decrease the hot side temperature, and heater power must be
increased to maintain the hot side temperature of 150°C. Under short circuit
condition, the hot side temperature drops to 113°C with a heater power of 40 watts.
For testing under a matched load condition, 5(9), the heater power must be

increased to 70 watts to obtain a hot side temperature of 150°C.

186

5.1.3 Results

For characterizing the system's measurement accuracy 3 commercial power
generation module was used [73]. Over a period ofone month different loads were

attached and the corresponding voltages, currents, and powers were measured.

  

 

5 _
I
Z
0
DD
I'll
i:
O
>
O Tellurex Voltage (V)
I Measured Voltage (V)
o i . ‘ i ' ‘
20 25

 

10

15
Load (9)

Figure 170: Comparison of voltage measured by the LTMTS and published Tellurex
results.

187

 

1200

1000

800

600

Current (mA)

400

200

._.____ .. _J

  
  

0 Tellurex Current (mA)

  

I Measured Current (mA)

 

 

T 2 1 1 1

0 S 10 15 20 25

Load (S2)

Figure 171: Comparison ofthe current measured by the LTMTS and Tellurex.

188

 

 

e
- O
9 I I "o,
14 4 ° I I °
, e
O. .0
5. 9 I O’O
. 0 ’0
A 1.2 : . I ....
3 \ OI I .....I
v _j 0 I
a; 1 .: ’ I
g. o
g t I
o 08.! ’
n. E 0
: 0
0.6 7.
3 O Tellurex Power (W)
;6
0.4 J.
g I Measured Output Power (W)
0.2 “
o i » r-
0 5 10 15 20 25

Load (9)

Figure 172: Comparison ofthe power measured by the LTMTS and Tellurex.

As shown in the above figures, the LTMTS is capable of accurate measurements of
thermoelectric power generation devices. After the characterization of the LTMTS, a

long-term test was done on the same module with the results shown in Figure 173.

189

 

i
i
0

H
A
.0
O
1

f3

+Temperature Hot C
80.0 “ +Temperature Cold C
60.0 “

 

Temperatur

40.0 ‘

20.0“ I I.- II'H'IW I—\'

0.0 r . . , T 1
1 6 11 16 21 26 31

 

 

Days

Figure 173: Hot side and cold side temperatures of module during month long test.
After one week of operation the heater power was increased to raise the hot
side temperature, and then it was maintained for the duration of the

 

 

experiment.
3.00 ~
2.50 3
‘3‘ 2.00 ‘H
V
E 1.50 ~
3
o. 1.00 -
0.50 1
0.00 1 . , , , ,
1 6 11 16 21 26 31
Days

Figure 174: Power output ofthe Tellurex module over one month of operation.

190

 

 

Figure 174 shows the measured power output of the purchased Tellurex module
over a month long period oftime. These results give confidence in an accurate long
term measurement can be done with the LTMTS.

A small heater for individual unicouples was built using a nickel—chromium

wire wrapped around an alumina substrate and then cast in an alumina adhesive.

 

Figure 175: Hand made heater being tested in the LTMTS.

The heater shown in Figure 175 was tested at 600°C for 48 hours. This showed that
the heater could be used for testing unicouples for long periods of time. Another
heater was tested for 120 hours at 600°C and also worked without degrading. A
photo of the heater is shown below after being strapped down to an insulation

encapsulated module.

191

 

Figure 176: Top view of heater in LTMTS.

 

Figure 177: Front view of experimental setup of heater in the LTMTS.

5.2: LAST and LASTT Modules

Metalizing the cold side alumina plates has been a concern. A thermally
conductive and electrically insulating material is needed as a substrate which can be
metalized to define the electrodes. A patent was found that is used to metalize
aluminum nitride [74]. The patent explains how to deposit chromium to the

192

aluminum nitride followed by anneals in air and then argon. This method was used
to fabricate the cold side contacts of the modules built here. The chromium was
sputtered to the aluminum nitride with a shadow mask. The first sputtering stage
was done at 300 Watts for 1 hour. It was then annealed at 450°C for 2.5 hours in air.
The second sputtering stage was also done at 300 Watts for one hour. Without
breaking vacuum tungsten was e-beamed to the chromium for 1/2 hour at 450mA.
The contacts were then electroplated with nickel and then annealed at 250°C for 4

hours with an argon-hydrogen atmosphere.

 

Figure 178: Completed cold side contacts of chromium to aluminum nitride.

Later it was discovered that initial depositions of the chromium at 350 Watts for 30
minutes anneals better to the aluminum nitride. After the initial anneal in air for 2
hours at 450°C the chromium contacts turn a brown color, which is known to be the
color of chromium oxide. The plating step of nickel showed best results when

plated for 15 minutes at 1.3 volts for each individual contact. After nickel plating

193

one more anneal was done at 250°C for 4 hours under vacuum. These contacts

worked best for modules built.

 

Figure 179: Contact with PbSn solder.

Unicouples from runs AW and AX were nickel plated, soldered into four leg
modules and then coated with QS/B4. Then the modules were tested in air for

power output.

194

 

Figure 181: Test set-up for module power output test.

195

The module made from AX unicouples had a resistance of 33mg. The module made
from unicouple run AW had a resistance of 276mg. The AW module resistance is

too high for a measurement so only the AX module was used in this setup.

196

 

Table 6: Module AX output results.

 

Day Time Th (°C) Tc (°C) AT Voc (mV) lsc (mA)

1 16:15 481.11 70.02 411.09 179 84.8
1 18:40 485.84 72.69 413.15 188 89.1
2 9:25 484.6 71.67 412.93 193.3 90

2 13:15 484.4 71.29 413.11 194.2 90.3
5 10:00 491.3 65.01 426.29 196.2 86.6
5 12:04 599.01 85.55 513.46 235.9 99

5 18:58 603.83 88.42 515.41 228.5 87.1
6 5:25 603.31 88.33 514.98 228.1 79.5
7 10:20 600.46 88.33 512.13 228.1 79.5

 

 

The test setup had a lot of parasitic resistances so actual short circuit current was
higher than the values shown in Table 6. The output voltage under open circuit
conditions is in good agreement with thermopower measurements of these

materials, which is approximately 200mV.

197

Unicouple run AY and AZ was a bond of MSUHPlZ and ETP40 to 316

stainless steel. This was a two stage run. All ofthe legs to be bonded were cut to
lengths of 5mm because samples will be soldered to Bi2Te3 after bonding in order

to form segmented legs. Stage 1 had a 30 minute ramp and stage two had a 4.5 hour
ramp and both stages had a 6.5 hour soak at 700°C. The gas flow rates for both
stages were 1LPM with a pressure of 790 Torr. A module made from these runs had
an open circuit voltage of 192mV with a 206.2°C temperature gradient across the

four leg modules.

 

Figure 182: Mobile module testing setup. This set up can be used in the lab and also
for public demonstrations.

Unicouple run BF was a bond of four ETN179 to 316 stainless steel samples. Stage
one of the bond had a 30 minute ramp and 6.5 hour soak at 700°C. The gas was
flowing at 1LPM with a pressure of 790 Torr. A11 4 ETN179 bonded to the 316.

198

Stage 2 of bonding run BF was a bond of p-type material (ETP ??) to the 316 to
complete the unicouple. The bond was done with a 4.5 hour ramp and 6.5 hour soak
at 700°C. The pressure was 790 Torr and the gas flowed at 1LPM. One unicouple

was fabricated from run BF and was built into a module.

 

 

 

Figure 183: Unicouple BF built into module.

The module was tested in the mobile testing device shown in Figure 182. The short
circuit current out of the module was up to 3.3 Amperes which was the highest
current reported out of a module in the labs to that day.

One module was fabricated, BG2, and had a room temperature resistance of
9.08mQ. This room temperature resistance was comfirmed with the room
temperature scans shown in Figure 140 and Figure 141. The module was put into

the mobile testing system.

199

 

Figure 184: Unicouple module BG2.

 

Figure 185: Unicouple module BG2 front View.

200

 

Figure 186: Picture of experimental setup.

,. ,4.» »~ nut-H"

.5
g

 

Figure 187: Reading from current sensor (1 volt is a reading of 10 amps).

201

 

Figure 188: Hot side temperature reading.

 

Figure 189: Cold side temperature reading.

Module BG2 produced 3 amps with a AT of S70.48°C. This experiment showed good
repeatability with the results from module BF. Modules can be built from different
bonding runs and show similar results.

A cold plate with contacts for 18 unicouples was fabricated using the

chromium deposition technique onto aluminum nitride.

 

Figure 190: Cold side contact for a 36 leg module.

202

The unicouples were soldered to the cold side contact, but during this fabrication
step 2 of the unicouple were destroyed so only 16 unicouples were used in the

device.

 

Figure 191: 32 leg module.

The module was the filled with QS/ B4 and then had its resistance tested. No current
could flow through the device indicating an open circuit condition at one or more of

the junctions.

 

Figure 192: The electrical resistance was tested and it was an open circuit.

203

The conductance between adjacent unicouples was tested and the figure below

shows the only places were the module was connected.

 

Figure 193: The red marks show were there is an electrical connection between
unicouples.

Only five connections were sustained through a few hours of operation. The module
was broken apart and studied. One ofthe largest construction failures was that the
individual unicouples were not nickel plated before being coated with the QS/B4.

This may be one of many other reasons for the breakdown ofthe device.

 

Figure 194: Photographs of unicouples taken out of the 32 leg module.

The photos above show oxidization developing between the stainless steel and the
hot pressed LAST material. This is due to an oxidization of the iron in the stainless
steel.

Using unicouples fabricated in runs B] and BK, 4 modules were built.

205

 

Figure 195: Unicouples built from runs Bland BK.

Each module was nickel plated and then coated in QS/ B4. Boron nitride was

sprayed on top of each module to keep the electrodes from shorting to the heater.

 

Figure 196: Completed module.

The first module tested cracked during operation. A later study was done to

determine why the module cracked. The second module tested was segmented with
Bi2Te3. The table below shows its results.

Table 7: B] and BK segmented module results

 

THot TCold Voltage Current

 

613°C 30.92°C 408.7mV 0A

 

632°C 27°C 1.7mV 1.12A

 

 

 

 

 

 

The max power calculated for this module was 114mW. For a 4 leg device the
expected power is around SOOmW. The module ran under this gradient for one

week.

 

Figure 197: Module after one week of operation.

The module was dismantled and a scan was done on the legs.

207

BK Module N-Leg to BiTe
6000 , r I

 

I l T r
5000 '- J '1
O
A 4000 . ° 4
> .
:1.
v o
g, 3000 _ . 1
2
O
> 2000 .. f... 1
O
O
1000 — ' .-

 

 

 

Position (mm)

Figure 198: Scan of N leg ofa module that ran for one week under a 585°C gradient.

As shown in Figure 198 the leg resistance increased. The properties ofthe leg

changed after the diffusion bond. The module resistance went from 10mg to 50m§2
after operating for 7 days. EDS was done on the interface between the Bi2Te3 and

LAST.

208

 

 

 

. u - .
l V, J , ‘34 . . ‘ . ,
1 ‘ O
’ . I .. ..r. .
I; .. - .0 0' g < ‘. ‘ :, '. .
n - ‘ a-
. § 3 ‘- - '7') ’ °“‘.
. I . - c‘ 0" I q. 0“ ,‘I ‘§ I
”1" ~-.." ° w 3“
v ‘ I .
. “ ‘ .- ,
~ . :-,:' ‘ ‘ r
I.‘ ' f.

 

   

Image is 1.6mm Across

Figure 199: EDS of contact. Top is LAST, then solder junction (Bi-Sn-Sb), Bottom
Bi2Te3.

209

 

 

 

Figure 200: EDS of same junction as figure above, but at a higher magnification, the
picture shows an area approximately 200nm across.

210

As shown in the figures above, bismuth looks to have diffused from the Bi2Te3 and

the solder into the LAST material. This could be the reason why there is a change in
the conductivity of the material.

A completed module underwent a testing of its mechanical stability by being
dropped from increasing heights starting from 1 foot to 10 feet onto a concrete
floor. To test the effects of each drop the resistance was documented followed by

photographs.

 

Figure 201: Module before drop testing.

The initial module resistance was 70mg. The drops between 1 foot and 8 feet did
not change module resistance or damage the module. After dropping the module 9
feet the module resistance changed to 87mg. The figure below shows how the

aluminum nitride cracked off of the bottom of the module.

211

 

Figure 202: After the module dropped 9' the base aluminum nitride cracked.

A module was fabricated with 2 unicouples that was soldered together with a
BiSbSn solder. ETN197 and ETP44 were used to build the module. The module is

pictured below.

 

Figure 203: Module built with solder for contacts.

It was first thought to only use this module for a low delta T, but with the QS/ B4
around the legs, the module could be used to higher temperatures than the melting
point of the solder. If the solder melts the electrical contact on the hot side will then

be liquid and allow for CTE mismatches between the LAST and the hot side
212

electrode. The QS/B4 would maintain the mechanical integrity of the module. More
modules were built by this technique and will be discussed later.

A unicouple was soldered together at both the cold and hot side contacts
with ETN152 and ETP44. A nickel-chromium-aluminum nitride cold side substrate
was used. The hot side was nickel and was solder to the legs with a Bi-Sn-Sb solder.

The room temperature resistance of the module was 6.4mQ. The unicouple was

placed into the LTMTS for measurement of the output power. With Th=135.26°C

and Tc=29.7°C the short circuit current was 700mA. The hot side temperature was

raised to 196.3°C and the cold side temp was 30.34°C. The short circuit was 1.06
amps. The hot side temperature was set to 221.23°C and the corresponding open
circuit voltage and short circuit current were 53.1 mV and 1.26 Amps respectively.
The hot side temperature remained around 225°C for 5 days to see if the output of
the module would change.

On the fifth day the hot side temperature was 225°C and the cold side was
29°C. The short circuit current was 1.27 Amps. Over the 5 days of operation the
modules output did not decrease. There was no change in the current output, which
gives evidence that module did not degrade over the 5 days of operation. After one
week of operation the hot side temperature was 214°C and the short circuit current
was 1.22 Amps. Slowly the hot side temperature was brought down to room
temperature. The initial resistance of the module was 6.4m§2 and after the test the

resistance changed to 7.6mQ.

213

 

After the good results of "Unicouple Built," a second soldered module was
built named "Ghost Rider." The substrate for the module was a metalized

aluminum nitride.

 

Figure 204: Cold side substrate for the module called Ghost Rider.

Figure 204 shows the nickel plated chromium contact that was deposited to
aluminum nitride. Nickel plated copper was then soldered to the substrate as

shown below.

214

 

Figure 205: Cold side electrodes soldered to cold side substrate.

First the four legs of the unicouple were solder to the cold side electrodes before

soldering on the copper hot side electrodes.

 

Figure 206: Front view of Ghost Rider's legs soldered to cold side.

215

 

Figure 207: Thermoelectric legs soldered down to cold side.

The legs were then nickel plated and the copper was soldered to the hot side. The
entire module was coated with QS/B4 with the completed module shown in Figure

208.

 

Figure 208 Ghost Rider

Figure 209: Ghost Rider output voltage and output current for a temperature
gradient across the module of AT = 184°C.

216

Voltage (V)

0.35 - .

 

 

Itj'IrIIIITrIIIIII'IIIr'IrI

 

  
 
  
 

—G— R=38.35mOhms Delta T=180K
1,... R=43.67mOhms Delta T=286K
~--0 R=48.6m0hms Delta T=408K

—0—- R=55.58m0hms Delta T=486K
—O— R=57.98m0hms Delta T=580K

0.3

 

0.25

 

 

0.2

   
 

0.15

0.1

0.05

 

Current (A)

Figure 210: Ghost Rider l-V Characteristics.

0.06 f I r I I I' '7' fit I I I T I I I I I l I I I I

 

0.055

I

[j I

0.05

I I I

0.045

1

Resistance (Q)

0.04 -

 

l 1 l__LI

q

 

 

0.035 --"-+L-IL---I---tn....
100 200 300 400 500 600

Delta T (K)

 

Figure 211: Plot of the resistance compared to the AT.

217

lllllllll ll

llll

 

Jilllllllljlj

  

 

 

The straight line fit in Figure 211 gives a 0 AT resistance of 28.85mQ. The actual
room temperature resistance of the module was measured using a four probe
configuration to be 15mg. This means there was around 15mQ of parasitic
resistances in the measurement setup (part of the measurement system, and
external to the module). For 3 AT of 570°C the short circuit current of the module
would have been closer to 7.2 amps, rather than the 5.5 amps measured if the
parasitic resistance was zero.

Another 4 leg module was fabricated based on soldered junctions and the
module was titled "Ien". For this module nickel instead of copper for the hot side

electrodes was used, and the module resistance at room temperature was 24.1mQ.

 

Figure 212: Module "Ien" built with cast materials and coated with QS/B4.

218

612°C-23°C
0.4

 

     

0.35
o_3 Resistance=91.77m§2
0.25

0.2

Voltage

0.15 f-
0.1 i
0.05 :-

 

 

 

Current

Figure 213: Output voltage and current for the four leg module "jen" with a
temperature gradient of 589°C.

 

With the higher module resistance, the short circuit current is lower for "jen" than
for "Ghost Rider", but the open circuit voltage is higher for “jen”. After running the
measurement the room temperature resistance of "jen" increased to 136mg. The

QS / B4 was removed and the legs investigated.

 

Figure 214: Module "jen", side view.

219

 

Figure 215: Module "Ien" rear View.

As shown in Figure 214 and Figure 215 no visible cracks or degradation of the'legs
could be seen, however increases in junction resistances may not be visibly
apparent, and cutting marks left from removal of the Q5 / B4 may have masked
cracks in the legs.

A simplified technique for fabrication of the cold side electrodes was also
investigated. Using a double-sided thermally conductive tape the cold side
electrodes were attached to the aluminum nitride substrate. With the ease for
making contact to the aluminum nitride a 8 unicouple module was fabricated as

shown in Figure 216.

220

 

Figure 216: Sixteen leg module called "Tape".

The "Tape" module had a room temperature resistance of 160mg. The expected
resistance of the module was 80mg. Some of the solder contacts may not have been
bonded well. This module ran for 93 hours with a hot side temperature of 225.4°C
and cold side temp of 25.1°C. The short circuit current was 440mA. With
approximately the same AT the four leg module "Ghost Rider" was giving close to
1.5A of short circuit current.

A 4 leg module with 6.35mm thick nickel was also built, and the room
temperature resistance was 27mg. This resistance is close to the resistance of the

module "jen".

221

 

 

 

Figure 217: Four leg module using nickel electrodes and titled "0001".

Module “0001” had an open circuit voltage of 23.4mV with a hot side temperature of
966°C. The hot side temperature was increased to 982°C and the short circuit
current was measured, 270mA. With the hot side at 118.6°C, the short circuit
current increased to 353mA. For modules previously built, the short circuit current
at this AT was nearly 1Amp. This module was not showing good results so it was
shut down.

A module was built with ETP52 and MSUHP-41-N. The resistance of the
module was 362mg. This modules resistance is very high; one leg of the module
typically is only 3.18m9, however the hot pressed material is known to have a
lower electrical conductivity at room temperature (then increasing with increasing
temperature). The room temperature resistance of 362m9 was still an order of
magnitude too high. Normally a hot pressed leg has a room temperature resistance

of approximately 20mg. The legs of the module may have cracked during the

222

soldering process. Results in the past have shown legs cracking during soldering,
but during recent constructions this has not been the case.

A module was built with another set oflegs from MSUHP-41-N and ETPSZ.
For the p-type material Cerromatrixm solder was used, and for the n-type leg, the

BiSnSb solder alloy was used. The 0.25" nickel was used as the hot side electrode.

 

Figure 218: Module 7-09-08-2Leg.

The module was placed in the LTMTS. With a hot side temperature of 247.1°C and
cold side temperature of 184°C the open circuit voltage was 122.9mV and short
circuit current was 354mA. Based on these measurement the calculated resistance
of the module at this AT was 346mg. The expected resistance of the module for this
temperature gradient is XX ma.

Dr. Nurraddin Matchanov built a unicouple using SnTe as the contact
material. The unicouple's resistance was 6mg. It was put into the LTMTS for

testing.

223

Table 8: SnTe unicouple measurements.

 

 

 

 

 

 

 

T HOt Side COId Side Open Circuit Short Circuit Unicouple
emperature Temperature Voltage (mV) Current (A) ReSIstance
(°C) (°C) (m9)
191.1 16.8 - 1.3 -
248 17.4 82.8 1.97 41.6
316 18.2 113.5 2.8 40.5
381.8 19.6 150 3.16 47.8
497 21.9 194.4 3.8 51.15

 

 

 

 

 

After running the module there showed no visible signs of degradation of the

unicouple. The unicouple was not coated in QS/B4 and measurements were taken

in an argon atmosphere.

 

Figure 219: Nuraddin's SnTe unicouple after running in the LTMTS.

 

6.0 Conclusions and Future Directions

6.1 Conclusions

This dissertation has been a record of experiments done in diffusion bonding,
temperature cycling, and long term tests. Diffusion bonds were attempted to many
different materials. Bonds to 410 and 316 stainless steel produced good results.
Some modules were fabricated with a direct soldering ofthe hot side electrode, and
have given high short circuit currents at maximum ATs.

These contacts were subjected to temperature cycling. Contacts such as
nickel showed that the material fractures after temperature cycling, while bonds to
316 stainless steel do not. Mechanical pressure showed an importance during
bonding. Many bonding runs did not have uniform mechanical pressure. Runs that
kept close measurement of individual mechanical bonding pressure showed 100%
bonding yield.

Bonds were studied to distinguish if the bonding material poisons or
degrades the thermoelectric material. Measurements of the thermoelectric
properties along with EDS plots were done. EDS showed that mostly tellurium,
chromium, and nickel would readily diffuse through contact materials as well as the
thermoelectric leg itself.

Diffusion barriers of molybdenum were studied but showed to not bond
with the LAST material. Other experiments in the lab have shown if SnTe is added
to the contact, molybdenum coated LAST bonds to stainless steel.

A system has been built to measure contact resistances at high temperatures

(700K), by other teammates in the lab [reference Farhan's thesis].

225

Spring loaded module designs have been investigated and built. None of
these modules entered the measurement phase due to high resistances. Recent
results have shown high contact resistances with a pressed contact of LAST to
coppen

In all, the main challenge is the LAST and LASTT material. Contacts to these
materials have been made and shown to have low contact resistance, but during
operation or fabrication of these contacts the LAST material has a propensity to
fracture. Initial investigations were done to powder process this material and hot
press to increase the strength and uniformity of its properties, but this showed to be
a challenge. Significant increases (three fold) in sample fracture strength has been
made through powder processing, however additional challenges of sample bloating

and good repeatability require further studies.

6.2 The Next Step, Spark Plasma Sintering

Spark plasma sintering (SP8) is an innovative technique for pressing and
sintering powders. Powders are loaded into a graphite die and while pressing the
material a large DC current is passed through the die and material. Unlike hot
pressing that uses external heating elements to the graphite die, the die and the
powder charge are heated due to the electrical current. This allows for rapid
heating and cooling of the material. Also a novel idea is to press an entire
thermoelectric part at once. The n-type, p-type, and electrode can be pored into the
die at once and sintered all together. The tooling of the graphite die would govern

the temperatures of the different zones to be sintered. Work has been done on

AngmeSem+2 materials in an SPS system, which is a cousin to the LAST material

226

 

[75]. These materials have shown comparable results to that ofthe LASTT

materials. SPS will also speed up the production of material as it was shown in the
SPS ofB-FeSiz [76]. Lastly, Zhou has SPS a high lead content LAST material and has
shown outstanding ZTs [77]. Their materials showed ZTs above 1 and improvement
of the thermal conductivity after a 30 day anneal. Perhaps the best feature of SPS

processing is the rapid throughput such that a variety of parameter adjustments can

be investigated in a relatively short amount of time.

227

References

 

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[10]

[11]

[12]

[13]

Thomas johann Seebeck,
ht : hem. h.hu'i.ac.il ~eu eniik histo

   

seebeckhtml. 8/29/2009.

   

Introduction into Thermoelectrics,
http://www.thermoelectrics.com/in troduction.htm#ref3

Sim Loo, "Transport Characterization of New High Temperature Quaternary
Thermoelectric Materials And Devices", Ph.D. thesis, Michigan State
University, 2003.

A. F. loffe, “Semiconductor Thermoelements and Thermoelectric Cooling",
lnfosearch Limited, London, 1957.

M. G. Kanatzidis, “The Role of Solid-State Chemistry in the Discovery of New
Thermoelectric Materials,” in Solid State Physics, ed. H. Ehrenreich, F.
Spaepen, vol. 69, pp 51-100, Academic Press, New York, 2001.

Kuei Fang Hsu et al., “Cubic Angme Te2+m Bulk Thermoelectric Materials with
High Figure of Merit,” Vol. 303, Science, www.5ciencemag.org

H.]. Goldsmid, "The use of semiconductors in thermoelectric refrigeration,"
British journal of Applied Physics, pages 386-390, 1954.

M.H. Ettenberg, W. A. jesser and PD. Rosi, “A New n-type and Improved p-
type Pseudo-ternary (Bi2Te3) (szTe3) (szSe3) Alloy for Peltier Cooling,”

Proc. 15th Int. Conf. On Thermoelectrics 52-56, 1996.

MC. Kanatzidis, “The role of Solid—State Chemistry in the Discovery of New
Thermoelectric Materials,” in Solid State Physics, ed. H. Ehrenreich, F.
Spaepen, Vol. 69, pp 51-100, Academic Press, New York, 2001.

RR Chasmar and R. Stratton, ]. Electron. Control 7, 52 (1959).

L.D. Hicks, M.S. Dresselhaus, "Effect of quantum-well structures on the
thermoelectric figure of merit,” Phy. Rev. B, Vol. 47, no. 19, pp. 12727, 1993.

G. Mahan, “Good Thermoelectrics," Solid State Physics, vol. 51,pp 82-155,
1997.

GA. Slack, CRC Handbook of Thermoelectrics, ed. D.M. Rowe, 1995, pp. 407-
440.

228

 

 

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

[2 5]

[26]

l2 7]

H. Scherrer and S. Scherrrer, CRC handbook of Thermoelectrics, pp. 21 1-237

G. Mahan, "Good Thermoelectrics," Solid State Physics, vol. 51, pages 113-
114, 1997

GB. Vining, CRC Handbook of Thermoelectrics, pp. 329-337

G. Mahan, “Good Thermoelectrics,” Solid State Physics, vol. 51, pages 115—
116, 1997

AL. jain, “Temperature Dependence of the Electrical properties of bismuth
antimony alloys," Physical Review, Vol. 114, No. 6, june 15, 1959.

G. Mahan, “Good Thermoelectrics," Solid State Physics, vol. 51, pages 116,
1997

G. Mahan, "Good Thermoelectrics,” Solid State Physics, vol. 51, pages 117-
119, 1997

GS. Nolas, D.T. Morelli, T.M. Tritt, “Skitterudites: A Phonon-Glass-Electron
Crystal Approach to Advanced Thermoelectric Energy Conversion
Applications,” Annual Review Material Science, pp. 89-116, 1999.

GS. N olas, et al., “High figure of merit in partially filled ytterbium
Skutterudites materials," Appl. Phys. Letts. Vol 77, No. 12,2000

G.A. Lamberton, et al., "High figure of merit in Eu-filled CoSb3-based
Skutterudites," Applied Physics Letters, Vol. 80, No. 4, 2002

GS. N olas, et al., "Semiconducting Ge Clathrates: Promising candidates for
thermoelectric applications, Applied Physics Letter, Vol. 73, No. 2, july 13,
1998

GS. Nolas, et al., “Transport properties of polycrystalline type-I Sn
Clathrates,” Physical Review B, Vol. 65, 165201, 2002

S. Bhattacharya, et al., “Effect of Sb doping on the thermoelectric properties

of Ti-based half-Heusler compounds, TiNiSn1-bex," Applied Physics Letters,
Vol 77, No. 16, 16 October 2000.

E. A. Skrabek and D.S. Trimmer, CRC Handbook of Thermoelectrics, ed. D.M.
Rowe, 1995, pp. 267-274

229

 

 

[28]

[2 9]

[3 0]

[31]

[3 2l

[3 3]

[34]

[3 5]

[36]

[3 7]

[38]

[39]

[40]

R. Venkatasubramanian, “Thin-film thermoelectric devices with high room-
temperature figure of merit,” Nature, 2001

R. Venkatasubramanian, et al., “Low temperature organometallic epitaxy and
its application to superlattice structures in thermoelectrics,” Appl. Phys. Lett.
75, 1104-1106 (1999)

R. Venkatasubramanian, “Low temperature chemical vapor deposition and
etching apparatus and method,” US Patent No. 6071351 (6 june 2000).

TC. Harman, et al., “Quantum Dot Superlattice Thermoelectric Materials and
Devices,” Science, Vol. 297, Sep. 27, 2002

N.F. Mott, “The Theory ofthe Properties of Metals and Alloys," Dover
Publications, New York, NY.

L. D. Hicks, M. S. Dresselhaus, "Effect of Quantum-Well Structures on the
Thermoelectric Figure of Merit," Physical Review B, Vol. 47, No. 19, pp.
12 727-12 731, 1993.

T. C. Harman, M. P. Walsh, B. E. LaForge, G. W. Turner, “Nanostructured
Thermoelectric Materials,” Journal of Electronic Materials, 34 (5), pp. L19-
L22, 2005.

D.M. Rowe, “Handbook of Thermoelectrics,” CRC Press, 1995, pp 496- 497

M.H. Cobble, “Calculations of Generator Performance,” CRC Handbook of
Thermoelectrics, CRC Press, 1995.

T. Caillat, et 31, “Development of a High Efficiency Thermoelectric Unicouple

for Power Generation Applications," 18th International Conference on
Thermoelectrics, 1999.

F. Chen, “Low-frequency ac measurement of the Seebeck coefficient," Rev. Sci.
Instrum, vol. 72, no. 11, 2001, pp. 4201-4206.

A. T. Burkov, A. Heinrich, P. P. Konstantinov, T. N akama, K. Yagasaki,
"Experimetal set-up for thermopower and resistivity measurements at 100-
1300K, “ Meas. Sci. Technol, vol. 12,2001, pg. 264-272.

A. L. Pope, R. T. Littleton, T. M. Tritt, "Apparatus for the rapid measurement of

electrical transport properties for both “needle-like” and bulk materials,” Rev.
Sci. Instrum, vol. 72, no. 7, july 2001, pp. 3129-3131.

230

 

 

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

R. T. Littleton IV, et. al., “High Temperature Transport Probe for
Thermopower and Resistivity Measurement," Materials Research Society
Symposium Proceedings, 1998.

Z. Zhou, C. Uher, "Apparatus for Seebeck coefficient and electrical resistivity
measurements of bulk thermoelectric materials at high temperature,” Rev.
Sci. Instrum, vol. 76, 2005.

M. Putti, M. R. Cimberle, A. Canesi, C. Foglia, A. Siri, "Thermopower
measurements of high-temperature superconductors: Experimental artifacts
due to applied thermal gradient and a technique for avoiding them,” Phys.
Rev. B, vol. 58, no. 18, Nov. 1998.

Downey, Adam Darwin, “Advancing Thermoelectric Research with New
Measurement Systems and Thermoelectric Modeling for AC Electrical
Measurements,” Ph.D. Dissertation, Michigan State University, 2006.

Merker ]., Lupton D., Topfer M., Knake H., “High Temperature Mechanical
Properties of the Platinum Group Metals,” Platinum Metals Review, Volume
45, Number 2, 1 April 2001 , pp. 74-82(9).

Kuei Fang Hsu et al., “Cubic AngmeTe2+m Bulk Thermoelectric Materials
with High Figure of Merit," Vol. 303, Science, www.5ciencemag.org

M. G. Kanatzidis, “The Role of Solid-State Chemistry in the Discovery of New
Thermoelectric Materials,” in Solid State Physics, ed. H. Ehrenreich, F.
Spaepen, vol. 69, pp 51-100, Academic Press, New York, 2001.

N. Uchitomi et al., “Characterization of a B-FeSiz p-n junction formed by the
PECS method,” Thin Solid Films 461, pp 174-178, Science Direct, 2004.

N. Iaussaud et al., “n- and p- Type behavior of the gold-substituted type-l

clathrate, BasAuxSi46-x (x=5.4 and 5.9),” C.R. Chimie 8, pp 39-46, Science
Direct, 2005.

S. Schlecht et al., "Nanoscale Zinc Antimonides: Synthesis and Phase
Stability,"
Inorganic Chemistry, 45, pp 1693-1697, 2006.

Y. Gelbstein et al,. "Powder metallurgical processing of functionally graded p-

Pb1-xSnxTe materials for thermoelectric applications,” Physica B 391, 2007,
pages 256—265

231

 

[52]

[53]

[60]

[61]

[62]

[63]

[64]

[65]

T. Pham Hong, et al., "Water Absorption in a Glass-Mica-Epoxy Composite l:
Influence on Electrical Properties," IEEE Transactions on Dielectric and
Electrical Insulation Vol. 16, No. 1; February 2009.

AA. Shirzadi, H. Assadi and ER. Wallach, “Interface Evolution and Bond
Strength when Diffusion Bonding Materials with Stable Oxide F ilms,” Suface
and Interface Analysis 2001; 31:609-618.

1. Cadoff, E. Miller, "Thermoelectric Materials and Devices," Reinhold
PublishingCorporation, 1960, pp. 68.

W. D. Callister jr, "Material Science and Engineering," Third Edition, john
Wiley & Sons, INC., 1994, pp. 646.

 

F. Ren, E.D. Case, E.j. Timm, et al., Phil. Mag. Lett. 86 (2206) p. 673.

DA. Porter and KB. Easterling, "Phase Transformations in Metals and Alloys,”
1993, Chapman and Hall.

W. D. Callister, jr. "Materials Science and Engineering An Introduction,” Sixth
Addition, john Wiley & Sons, Inc. 2003, pp. 221-224.

S.-X. Li et al., “A Probabilistic Model for Prediction of Bonding Time in
Diffusion Bonding,” Materials Science and Engineering A 407 (2005) 250-
255

W. D. Callister, jr. "Materials Science and Engineering An Introduction,” Sixth
Addition, john Wiley & Sons, Inc. 2003, pp. 222.

A. L. Eiss, "Studies of Lead Telluride Thermoelectric Elements," Hittman
Associates, Inc. Baltimore, Maryland.

Y. Kuru et al., “lnterdiffusion and Stress Development in Thin Film Diffusion

' Couples,” Materials Science Forum Vols. 524-525 (2006) pp. 801-806.

A. Smigelskas and E. Kirkendall, Trans. AIME, 171, 130 (1947).

H. Ono, T. Nakano, and T. Ohta, “Diffusion Barrier Effects of Transition Metals
for Cu / M /Si Multilayers (M=Cr, Ti, Nb, Mo, Ta, W),” Appl. Phys. Lett. 64 (12)
21 March 1994.

Kyung-Hoon Min, Kyu-Chang Chun, and Ki-Bum Kim, "Comparative Study of

Tantalum and Tantalum Nitrides as Diffusion Barrier for Cu Metallization," j. -
Vac. Sci. Technology B 14(5), Sep/Oct 1996.

232

 

[66]

[67]

[68]

[69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

[77]

R. C. Weast, M. j. Astle, and W. H. Beyer, “CRC Handbook of Chemistry and
Physiscs,” 64th edition, CRC, Boca Raton, Florida, 1983, p. F-47.

R. j. Borg, G. j. Dienes, “An Introduction to Solid State Diffusion," Academic
Press 1988, pp. 81.

jonathan D’Angelo, "Low Resistance Contacts to Thermoelectric Materials,”
MS. Thesis, Michigan State University, 2006.

A. D. Downey, T. P. Hogan, B. Cook, “Characterization of thermoelectric
elements and devices by impedance spectroscopy,” Review of Scientific
Instruments, Vol. 78, No. 9, 2007.

G. Q. Wu, et al., “The effects of various finished surfaces on diffusion
bonding," Modeling and Simulation in Materials Science and Engineering, 16,
2008

B. l. Boltaks, “Diffusion in Semiconductors," Academic Press, New York, 1963,
page.57

P. F. P. Poudeu, j. D’Angelo, A. Downey, j. L. Short, T. P. Hogan, M. G.
Kanatzidis, "High thermoelectric figure of merit and nanostructuring in bulk

p-type Na1-bemeyTem+2," Angewandte Chemie-lnternational Edition, Vol.
45, No. 23, pp. 3835-3839, 2006

www.tellurex.com. 3-9-2009

Arledge, et al., “Adherent metal coating for aluminum nitride surfaces,"
5,382,471, january 17, 1995

K.F. Cai, et al., “Preparation and thermoelectric properties of AngmeSem+2
materials,” journal of Solid State Chemistry 181 (2008), pages 1434-1438.

K. N ogi, T. Kita, "Rapid production of B-FeSiz by spark-plasma sintering,”
journal ofMaterials Science 35, 2000, pages 5845-5849.

Min Zhou, et al., "Nanostructured AngmeTem+2 System Bulk Materials with
Enhanced Thermoelectric Performance,” jACS Articles, 03/08/2008.

233