. .x. ..
La...»
:3...
.1“:-

tfifivl
V a...

“may

. w. .
.. . m3»... 9..
. 3
I. u A

bfi

«fink.
a.

t
!E

323512?

2??

1

.26 \
‘5“!!!
.1?

lug
ttsolilss

a...»

. ..
‘
33 t4
:fii
. a.“
3.; U
imuzii
| l1..§t\
I“... A! 1|!- ‘tclti xii!)
sp‘ In. A o: i‘
V 91.1.33}: It). ~
:1. n in. ‘ ~
x ”vita... 2&2... .
n pl. ‘5
‘ ..

.. ‘
rt... ”rill. Huh“...
.$:(!? r. ‘
.1."‘- In l..'.
32.35 ‘\IIOC:I
.10 ~orq‘o‘»

.1 .4..A.|o|~

..
.233».
19‘“:1
19.1111. 0 A
:51 ‘
‘0.

9 .v.. .01.:

.. «1....

zit»: s ti: 1... v

, 1.2.2.... 3 v m... v.5.
13:21,. a." .

 

 

 

 

LHNTARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

THERMOELECTRIC MODULE CONTACT FABRICATION
AND ANALYSIS AT ELEVATED TEMPERATURES BY HIGH
TEMPERATURE VOLTAGE PROFILING SYSTEM (HTVPS)

presented by

MUHAMMAD FARHAN

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Electrical Engineering

 

 

 

W

Major Profeséor’s Signature
8 * 24 - 09

Date

MSU is an Affirmative Action/Equal Opportunity Employer

-.u.;~'.~.v

 

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:lProleoc&Pres/CIRCIDateDuo.indd

 

‘— 1. - -—— .
I ...EmL. -£—._

 

THERMOELECTRIC MODULE CONTACT FABRICATION AND ANALYSIS
AT ELEVATED TEMPERATURES BY HIGH TEMPERATURE VOLTAGE
PROFILING SYSTEM (HTVPS)

By

MUHAMMAD FARHAN

A DISSERTATION
Submitted to
Michigan State University

In partial fulfillment of requirements
For the degree of

DOCTOR OF PHILOSOPHY
Electrical Engineering

2009

ABSTRACT
THERMOELECTRIC MODULE CONTACT FABRICATION AND ANALYSIS
AT ELEVATED TEMPERATURES BY HIGH TEMPERATURE VOLTAGE
PROFILING SYSTEM (HTVPS)
By

Muhammad Farhan

High temperature bulk thermoelectric (TE) materials Ag1.bemeTe2+m and

Ag(Pb1.xSnx)meTe2+m (also known as LAST and LASTT) developed at Michigan

State University have exhibited a figure of merit, ZT, of 1.7 at 700 K. Thermoelectric
power generation devices based on these materials are being developed at the Electronic
Material, Pulse Laser Deposition and Transport Characterization laboratory. For optimal
performance, low resistance high temperature electrical contacts are required.

In this research work a new high temperature voltage profiling system has been
designed and fabricated to investigate the temperature dependence of the hot Side contact
resistance. The system can measure contact resistance and electrical conductivity of a
sample from room temperature to 800 K. The system was used to characterize bonds of
tin-tellurium, braze and solder to hot pressed LAST, cast LAST and LASTT type
thermoelectric materials.

The second research effort focuses on spring loaded TE modules and their output
parameters (open circuit voltage and short circuit current) from 300- 800 K. This has
helped to explore a practical TE generator configuration. Also InGa based liquid contacts

were investigated to circumvent coefficient of thermal expansion (CTE) mismatch

problems with hot side metal electrodes. The results are in good agreement with the
expected output based on module modeling efforts.

The high temperature voltage profiling system has been utilized to evaluate a
variety of potential electrode materials including brass, copper, nickel and stainless steel.
During this investigation we developed SnTe based bonds which proved to be our
optimum solution in terms of low contact resistance, high power output and sustainable

operational endurance.

ACKNOWLEDGEMENTS

I am thankful to Almighty Allah for giving me the courage to accomplish this task. I am
deeply grateful to my advisor Dr. Tim Hogan for all the guidance and patience without
which it all would not be possible. My gratitude for my fellow lab mates specially Ajmal
Khan for his cooperation during research and course work, Nuraddin Matchanov for
closely collaborating on experiments, Jarrod Short for introducing me to thermoelectrics

and Jonathan D’Angelo for useful inputs and allowing me to use his equipment.

I am thankful to my mother for her untiring efforts, my brother for his constant

persuasion and sisters for all the care.
Finally, financial support from Higher Education Commission Pakistan, National Science

Foundation for nanowire research, Office of Naval Research and SERDP for

thermoelectric research is gratefully acknowledged.

iv

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... viii
LIST OF FIGURES ................................................................................ ix
Chapter 1: Introduction to Thermoelectric Research .................................... 1
1.1 Thermoelectric Materials ........................................................... 4
1.2 Seebeck Effect ....................................................................... 4
1.3 Peltier Effect ......................................................................... 5
1.4 Thompson effect ...................................................................... 6
1.5 Figure of Merit(ZT) ................................................................. 7
1.6 Thermoelectric applications ........................................................ 8
1.7 Lead-Antimony—Silver-Tellurium (LAST) TE Material ...................... 10
1.7.1 Lead-antimony-Silver-tellurium (LAST) Mechanical
Properties .................................................................. 12
1.7.1.1 Hardness .......................................................... 13
1.7.1.2 Young’s Modulus ............................................... 13
1.7.1.3 Weibull analysis and bending strength ....................... 14
1.8 Calculation for thermoelectric generator out put and efficiency
with and without contact resistance ............................................ 15
Conclusion ........................................................................... 20
Chapter 2: Thermoelectric Module Bonding and Diffusion Barriers. . . . . . .21
Introduction ......................................................................... 21
2.1 Diffusion Bonding ................................................................ 22
2.2 Diffusion Furnace ................................................................. 24
2.3 Room Temperature Scanning Probe system ................................... 25
2.4 LAST to LAST bonding ........................................................... 26
2.5 Room temperature scan of stainless steel bonded unicouples ............... 37
2.6 Soldered VS. Silver paste contacts ................................................ 44
2.7 Diffusion barriers in thermoelectric — previous efforts ....................... 48
2.7.1 Brass diffusion barrier coatings ........................................ 52
2.7.2 Diffusion bonding with coatings for Ni, Cu and steel ............... 55
2.8 Diffusion bonding with Sn and Te powder ...................................... 59
2.9 Electrical contact resistance measurement ..................................... 64
Conclusion ......................................................................... 69
Chapter 3: InGa and Spring Loaded Modules ............................................. 70
Introduction ........................................................................ 70
3.1 InGa Module Fabrication ......................................................... 70
3.1.1 Coatings on TE legs ...................................................... 77
3.1.2 SS-Ni-Au coating ......................................................... 79
3.1.3 Mo-Au coating ............................................................ 80
3.1.4 Mo-Ni coating ............................................................ 81

3.2

3.3
3.4

Chapter 4:

4.1
4.2
4.3

4.4

Appendix A:

Al
A2

A3

A.4

Spring Loaded Modules ............................................................ 85
3.2.1 Surface asperities and contact resistance .............................. 85
3.2.2 Effect of Surface Film .................................................... 89
Room Temperature Spring Loaded module parameters ...................... 90
Spring Loaded High Temperature Short Circuit Current

Measurements ...................................................................... 96
3.4.1 New heater location .................................................... 101
3.4.2 Spring Loaded Hot Pressed Module ................................... 104
3.4.3 Spring Loaded Ni Electrode Module ................................. 105
3.4.4 Spring Loaded Silver (Ag) Electrode Module ...................... 107
3.4.5 Spring Loaded Module with tungsten (W) coated samples ....... 108
3.4.6 Spring Loaded and InGa Based Combined Module. . . . .............1 11
Conclusion ........................................................................ 1 14

High Temperature thermoelectric contact investigation using

Voltage Profiling System (HTVPS) ......................................... 115
Introduction ......................................................................... 1 15
System Description ................................................................ 116
System Operation ................................................................ 120
High Temperature contact investigation........................................121
4.3.1 SnTe based Modules ................................................... 123

4.3.1.1 SnTe based cast n-type bonded modules ................... 124
4.3.1.2 SnTe based cast p-type bonded module. 143
4.3.1.3 SnT e based hot pressed bonded modules ................... 159
4.3.2 Prema Braze based diffusion bonded modules ..................... 163
4.3.2.1 Premabraze bonded modules with p-type cast
TE material ...................................................... 163
4.3.2.2 Premabraze bonded module with n-type cast
TE material ........................................................ 167
4.3.3 Low temperature solder (Cerromatrix) bonded module. . ........... 174
4.3.3.1 P-type LASTT bonded using Cerromatrix
solder with Cu .................................................... 181
Energy Dispersive Spectroscopy (EDS) analysis ............................ 191
Conclusion .......................................................................... 193
Nanowire Synthesis ............................................................ 194
Introduction ...................................................................... 194
Experimental setup ................................................................ 194
Germanium Oxide (G602) nanowires ........................................ 196
A21 Experimental details ................................................... 196
Hotplate Synthesis of Ge02 nanowires ...................................... 199
A31 Experimental procedure ................................................ 199
A32 Results and Discussion ................................................ 200
A33 Growth Mechanism .................................................... 201
Indium Antimonide (InSb) Nanowires Synthesis ........................... 207

vi

 

A.5 Catalyst Assisted Zinc Oxide (ZnO) Nanowires Synthesis ................ 211

A6 Catalyst Free Zinc Oxide (ZnO) Nanowires Synthesis ..................... 212
Conclusion ........................................................................ 216

References ............................................................................................. 217

vii

Table 2-1.

Table 2-2.

Table 2-3.

Table 2-4.

Table 2-5.

Table 2-6.

Table 2-7.

Table 3-1.

Table 3-2.

Table 4-1.

Table A-1.

Table A-2.

Table A-3.

LIST OF TABLES

LAST to LAST bonding summary ............... ............................. 37
Unicouple contact resistance values ............................................ 42
Brass coatings with diffusion bonding .......................................... 53
Diffusion bonding with coatings for Ni, SS316 and Cu ..................... 56
SnTe based diffusion bonding ................................................... 60
TE legs resistance values ......................................................... 61
Hot side electrode resistance values ............................................ 61
Summary results for InGa based modules .................................... 84
Summary results for spring loaded modules .................................. 113
EDS results for dark and shiny regions indicating Cu and oxygen rich
regions ............................................................................. 192
EDS results indicating Ge and 0 ratio ........................................ 205
EDS results indicating In and Sb ratio ........................................ 209
EDS result indicating Zn and 0 ratio .......................................... 212

viii

Figure 1—1.
Figure 2-1.
Figure 2-2.
Figure 2-3.
Figure 2-4.
Figure 2-5.
Figure 2-6.

Figure 2-7.

Figure 2-8.

Figure 2-9.

Figure 2-10.
Figure 2-11.
Figure 2-12.

Figure 2-13.

Figure 2-14.
Figure 2-15.
Figure 2-16.
Figure 2-17.
Figure 2-18.

Figure 2-19.

LIST OF FIGURES

Thermoelectric unicouple ........................................................ 16
Room temperature scanning probe system .................................... 26
LAST to LAST bonding ......................................................... 27
LAST—LASTT bond .............................................................. 28
LASTT to LASTT bond using Sb and LAST powder ....................... 29
LAST-LASTT direct bonding ................................................... 30
High temperature scan for LAST to LASTT bonding ........................ 31
Room temperature voltage scan of the LASTT to LASTT bond using a
solder foil at the interface ........................................................ 32
ETP 47 bonded to ETP 51 using a Sn/Sb/Pb solder foil ..................... 33
LASTT to LASTT bond using solder .......................................... 34
ETP 47 bonded to HPMSU 35 using SnTe powder .......................... 35
ETP 47 bonded to ETP 51 using SnTe powder .............................. 36
Six unicouples tested for room temperature contact resistance ............ 38

Room temperature and 700 K scan indicating increased contact

resistance ........................................................................... 43
Soldered contact thermoelectric (ETP 38) legs ............................... 44
Silver paste contact measurements for six samples ........................... 45
Hot pressed sample bonded to Ni electrode using solder ................... 46

Hot pressed sample bonded to Cu electrode using solder ................... 47
ETP 53 samples solder contacted on hot plate ................................ 48
LASTT to brass bonding ......................................................... 54

ix

Figure 2-20.
Figure 2-21.
Figure 2-22.
Figure 2-23.
Figure 2-24.
Figure2-25.
Figure 2-26.
Figure 2-27.
Figure 2-28.
Figure 2-29.
Figure 2-30.
Figure 2-31.
Figure 3-1.

Figure 3-2.

Figure 3-3.

Figure 3-4.

Figure 3-5.

Figure 3-6.
Figure 3-7.
Figure 3-8.
Figure 3-9.

Figure 3-10.

EDS results for Zn diffirsion from brass to LAST ............................. 55
LAST bonded to Ni electrode ................................................... 58
LASTT bonded to Ni electrode ................................................. 58
Carpenter steel 19—2 bonded to ETP 52 using SnTe .......................... 62
Room temperature scan for carpenter steel bonded to ETP52 ............... 63
SEM image for austenitic steel and HPMSU 41 .............................. 63
Carpenter steel bonded to HPMSU 41 ......................................... 64
Calculation of contact resistance from voltage vs position plot ............ 65
Curve fit for contact resistance measurement ................................. 66
Scan with 0.2 mm step size ...................................................... 67
Scan with 0.1 mm step size ...................................................... 67
Superimposed scans of 0.1 and 0.2 mm step size ............................. 68
Ni electrode based InGa liquid contact ........................................ 72
Stainless Steel 316 electrode for 12mm TE legs .............................. 73
Output for 12mm long TE legs .................................................. 74
Stainless steel 316 electrode with 12 mm TE legs ............................ 75
Module first measured under Ar, then three cycles with different
electrodes measured in open air ................................................. 76
Au and Ni coated TE legs based unicouple output ........................... 78
Mo coated TE legs based module tested twice in open air .................. 79
Mo-Au sputtered TE legs module ............................................... 81
Mo-Ni plated TE legs measured under Ar and ambient air .................. 82
Module prepared without any coating on TE legs ............................ 83

Figure 3-11.

Figure 3-12.

Figure 3-13.

Figure 3-14.

Figure 3-15.

Figure 3-16.
Figure 3-17.
Figure 3-18.
Figure 3-19.
Figure 3-20.
Figure 3-21.
Figure 3-22.
Figure 3-23.
Figure 3-24.
Figure 3-25.
Figure 3-26.
Figure 3-27.
Figure 3-28.
Figure 3-29.
Figure 3-30.
Figure 3-31.

Figure 3-32.

Bulk electrical interface with current lines passing through asperities
(points of contacts between surfaces) .......................................... 86

A-spot making an angle with opposite surface due to conical asperity. . ..88

Electric lines of force in (a) high resistance film (b) low resistance

fihn .................................................................................. 90

Spring loaded design for RT scan ............................................... 91

(a) Sample holder for polishing (b) TE legs secured inside holder

(0) Leco 2000 Polisher with felt cloth on right side plate .................... 93
Spring loaded ETP53 with Ni plated Cu electrode ............................. 94
Spring loaded ETN 224 with Ni plated Cu electrode ........................... 95
Spring loaded hot pressed sample with Ni plated Cu electrode ............... 96
High temperature spring loaded unicouple measurement system ............ 97
Spring loaded apparatus on module testing system (LTMTS) ................ 97
First spring loaded module with Ni plated Cu electrode ...................... 98
Second spring loaded module measured in open air ........................... 99
First module measured inside LTMTS under Ar .............................. 100
Spring loaded module with heater directly on unicouple ..................... 101
Temperature dependent electrical conductivity for ETN 164 ............... 102
Spring loaded module with highest recorded current ......................... 103
Spring loaded module with MSUHP94 ......................................... 104
Spring loaded module with Ni electrode ....................................... 105
Spring loaded Ni electrode based module measured under Ar .............. 106
Spring loaded Ag electrode module ............ . ................................. 107
Spring loaded tungsten coated module measured under Ar .................. 109
Spring loaded module with tungsten coating measured in air ............... 110

xi

Figure 3-33. Combined spring loaded and InGa liquid contact based module ........... 111

Figure 3-34. Spring loaded InGa based Ag electrode module .............................. 112
Figure 4-1. High Temperature voltage profiling system ................................. 119
Figure 4-2. Probe pin holder ................................................................. 1 19
Figure 4-3. New sample holding arrangement ............................................. 120
Figure 4-4. NIST SRM 1461 room temperature and high temperature plots .......... 121
Figure 4-5. Bonds and materials investigated for high temperature contact

Figure 4-6.

Figure 4-7.
Figure 4-8.

Figure 4-9.

Figure 4-10.

Figure 4-11.

Figure 4-12.

Figure 4-13.

resistance measurement ......................................................... 122

ETN 72 bonded to Carpenter Steel 19-2 using SnTe (a) Room
Temperature scan (b) scan at 100 0C. (c) Scan at 200 °C ((1) scan at
300 °C. (e) Scan at 400 °C (0 Scan at 550 °C ................................. 127

ETN 72 bonded to carpenter steel 19-2 using SnTe ........................ 129
Stainless steel 316 bonded to ETN 153 with SnTe ......................... 130

Stainless steel 316 bonded to ETN 153 using SnTe. (a) Pre anneal

room temperature scan (b) Scan at 550 °C indicating increase in

contact resistance. (c) After anneal room temperature scan indicating a
permanent increase in contact resistance accompanied with increase in
electrical conductivity ........................................................... 132

ETN 218 bonded to Ni electrode using SnTe. (a) Pre-anneal room
temperature scan (b) 500 °C scan indicating increase in contact

resistance. (0) Afier anneal room temperature scan indicating contact
resistance value returned to pre anneal value ................................ 136

ETN 153 bonded to Ni electrode using SnTe coin. Bonding

temperature 615 °C. (a) Pre-anneal room temperature scan (b) 450 °C
scan after 12 hrs (c) 450 °C scan after 48 hrs (d) 450 °C scan after

5 days indicating increased contact resistance .............................. 140

ETP 50 bonded to carpenter steel 19-2 using SnTe. As the temperature
is increased electrical conductivity and contact resistance decrease. . ....143

ETP 53 bonded to Ni using SnTe. (a) Pre-anneal room temperature

scan (b) 500 °C scan indicating decrease in contact resistance
(C) After anneal room temperature scan indicating a permanent

xii

Figure 4-14.

Figure 4-15.

Figure 4-16.

Figure 4-17.

Figure 4-18.

Figure 4-19.

Figure 4.20.

Figure 4-21.

Figure 4-22.

decrease in contact resistance .................................................. 145

ETP 51 bonded to Ni using SnTe coin. Contact resistance increases
with temperature with a permanent increase after cooled down to
room temperature .................................................................. 148

ETP 53 directly bonded to stainless steel 316. Contact resistance
increases with measurement temperature accompanied with
permanent increase after cooling down to room temperature. . . . . . 149

ETP 52 bonded to carpenter steel using SnTe. (a) Pre-anneal room
temperature scan (b) First anneal at 600 °C. contact resistance

increased (c) Room temperature scan after lSt anneal. Contact

resistance reduced to pre-anneal value ((1) Second anneal run at

600 °C. Contact resistance increased (e) Room temperature scan after
second annealing run. Contact resistance reduced to 1St pre-anneal

value ............................................................................... 150

ETP 52 bonded to stainless steel 316 using SnTe. (a) Pre-anneal

run with low contact resistance (b) Post anneal room temperature

scan indicating no change in contact resistance (c) Second anneal

run at 600 °C for 12 hr (d) Second anneal nm at 600 °C for 24 hr (6)
Second anneal run at 600 °C for 48 hrs (0 Second anneal run at 600

°C for 60 hrs indicating no appreciable change in contact resistance. . ..155

MSUHP 35 bonded to carpenter steel using SnTe. Electrical
conductivity increases and contact resistance decreases with
increasing temperature ......................................................... 159

MSUHP37 bonded to stainless steel 316 using SnTe. There is
permanent increase in electrical conductivity and contact resistance
remains same after thermal cycle ............................................. 160

HPMSU 35 coated with Mo bonded to M0 coated stainless steel
electrode. Electrical conductivity increases and contact resistance
decreases with increasing measurement temperature ....................... 161

MSUHP 37 bonded to carpenter steel. Electrical conductivity
improves with increasing measurement temperature. No appreciable
change in contact resistance during or after annealing ..................... 162

ETP53 bonded to N1 plated Cu using Premabraze 616. (a) Pre anneal
room temperature scan (b) Scan at 450 °C indicating increase in

contact resistance (c) Post anneal room temperature scan indicating

a permanent increase in contact resistance ................................... 164

xiii

Figure 4-23.

Figure 4-24.

Figure 4-25.
Figure 4-27.

Figure 4-28.

Figure 4-29.

Figure 4.30.

Figure 4-31.

Figure 4-32.

Figure 4-33.

Ni electrode bonded to ETP 53 using Premabraze 616 ( pre anneal
scan indicating high resistance junction (b) 450 °C scan indicating

increase in contact resistance (c) Post anneal scan ......................... 165
ETP 51 bonded to Ni electrode using Premabraze 616 ..................... 166
Ni coated brass bonded to ETP 53 using Premabraze 616. There is a
permanent increase in junction resistance afier annealing cycle .......... 166
ETP 51 bonded to stainless steel 316 using Premabraze 616 .............. 167

ETN 157 bonded to Ni coated brass using Premabraze6l6.(a) Pre

anneal scan(b) 450 °C scan indicating increase in junction resistance

(c) Post anneal scan indication junction resistance approaching back

to pre-anneal value .............................................................. 168

Ni electrode bonded to ETN 218 using Premabraze61 6.(a) Pre-anneal

scan (b) 450 °C scan after 6 hr indicating increase in junction

resistance (c) 450 °C scan after 24 hr indicating no appreciable change
than 6 hr scan. ((1) Post anneal scan indicating junction resistance

reduced back to pre-anneal value .............................................. 170

ETN 218 bonded to Ni electrode using Premabraze616. (a) Pre-anneal
scan indicating a low contact resistance junction (b) 450 °C scan
indicating no appreciable change in contact resistance (c) Post armeal

scan indicating increase in contact resistance ............................... 173

ETN 213 bonded to Ni plated Cu using Cerromatrix solder (a) Pre-

anneal scan (b) 100 °C scan (c)(d) Scan indicating gradual increase in
contact resistance with increasing temperature. (e) Contact resistance
increasing with temperature. (0 Post anneal scan indicating a

permanent increase in contact resistance as compared to pre-anneal

scan (g) Second anneal 600 0C scan indicating increase in contact
resistance as compared to first anneal high temperature scan

(h) Post anneal scan after second run indicating after every anneal

there is permanent increase in contact resistance higher than previous
cycle ............................................................................... 175

Ni plated Cu and Ni plated ETN 207 leg bonded using Cerromatrix
solder. Contact resistance increases with increasing temperature. . . . . l 80

ETP 52 bonded to Cu electrode using Cerromatrix solder. (a) Pre
anneal scan indicating high contact resistance (b) (c) ((1) Contact
resistance decreases with increasing temperature from 100 °C to
500 °C. (e) (0 Electrical conductivity drops from ~ 500 S/cm to

xiv

110 S/cm after 12 hr anneal at 600 °C. (g) (h) Electrical conductivity
remains same for one and half days and contact resistance does not
deteriorate. (i)(j )(k)(l) Electrical conductivity remains same for four
days and contact resistance does not deteriorate. (m) Post anneal
room temperature scan indicating permanent increase in contact
resistance (n)(p) No improvement in contact resistance with
increasing temperature (q) Second post anneal room temperature

scan. Contact resistance increased two fold ................................. 182
Figure 4-34. Cu electrode with ETP 52 (Ag0_9Pb9$n9$b0_6Te20) ...................... 191
Figure 4-35. Fractured sample leg indicating dark and shiny region ..................... 192
Figure A-l. Nanowire synthesis setup ....................................................... 195
Figure A-2. GeOz Nanowires SEM image .................................................. 198
Figure A-3. Quantitative result indicating Ge and 02 ratio ............................... 198
Figure A—4. (a)TEM diffraction pattern (b) after two minutes of irradiation

diffraction pattern is lost and NW turns amorphous ........................ 203
Figure A-S. SEM images of GeOz NWs on Au coated (a) Aluminum nitride (AlN)

surface (b) Aluminum oxide( AL203) surface. Wires are straighter

(needle-like) and more dense (c) quartz surface, nanowires have non

uniform thickness (d) silicon (Si) substrate, nanowires are sparse and

irregular ............................................................................ 204
Figure A-6. Raman spectra showing GeOz NWs are trigonal ............................ 205
Figure A-7. TEM image showing Au eutectic ............................................... 206
Figure A-8. Hotplate setup for nanowire synthesis ........................................ 206
Figure A-9. Powder X-ray diffraction spectrum indicating GeOz peaks .............. 207
Figure A-10. InSb Nanowire SEM image .................................................... 209
Figure A-ll. InSb NW with a diameter of ~200 nm ........................................ 210
Figure A-12. InSb SAED showing zincblend structure ..................................... 210

XV

Figure A—l3.
Figure A—l4.
Figure A-lS.
Figure A-16.

Figure A-l7.

ZnO nanowires grown in tube finnace ....................................... 212
ZnO NWs grown on hot plate ................................................. 214
EDS quantitative results for ZnO ............................................. 214
TEM selected area diffraction image ......................................... 215
ZnO NW length ~ 6 pm ........................................................ 215-

xvi

Chapter 1: Introduction to Thermoelectric Research

Chalcogenide based thermoelectric materials Ag1.bemeTez+m and

Ag(Pb1_xSnx)meTe2+m (also known as LAST and LASTT) developed at Michigan

State University have exhibited promising thermoelectric properties for power generation
applications [1]. Thermoelectric generators consisting of this material are expected to
operate at hot side temperature of ~ 900 K and cold side temperature of ~300 K. This
thermoelectric research is a collaborative effort between the Electrical and Computer
Engineering, Material Science and Engineering, and Mechanical Engineering
departments at Michigan State University and the Chemistry department at Northwestern
University. For this project a goal of Material Characterization and Pulsed Laser
Deposition Lab at Michigan State University Electrical and Computer Engineering
Department is to study these new materials through characterization and develop them
into thermoelectric generators. Material characterization involves transport measurements
which include electrical conductivity, thermal conductivity, therrnopower, and ZT

measurements. A thermoelectric generator consists of an array of unicouples each having

one n-type leg of Ag1.bemeTe2+m (LAST) and one p-type leg of

Ag(Pb1.xSnx)meTe2+m (LASTT) electrically interconnected with electrode. These

electrode-thermoelectric material contacts need to be low resistance for higher efficiency
and power output. Typical acceptable total contact resistance values are less than 10% of

the total thermoelectric legs resistance. The purpose of this thesis is to investigate

methods of bonding and high temperature contact behavior of thermoelectric material to
metal electrodes. The electrodes on hot side are typically more difficult to fabricate as
they must operate at temperatures where common solders are molten and the issue of
differing coefficient of thermal expansions (CTE) between the thermoelectric material
and the electrode material becomes critical. As part of this effort various diffusion
bonding processes have been investigated to develop low resistance bonds. These low
resistance bonds need to be stable at high temperature which brings in the issue of
elemental diffusion from the metal electrode. To circumvent this problem, diffusion
barriers have been studied on the electrode and on the thermoelectric materials. Liquid
contacts consisting of InGa were used to avoid coefficient of thermal expansion
mismatch challenge between thermoelectric material and metal electrode. Spring loaded
modules were tested under ambient conditions and inert environment for short circuit
current measurements. To investigate thermoelectric contact behavior at high temperature
a new high temperature voltage profiling system has been designed and fabricated. In this
system a voltage probe pin scans across the metal electrode-thermoelectric leg contact.
Data obtained from this measurement is used to determine the sample resistance at
various temperatures, to monitor the sample uniformity, to see if any cracks in the sample
exist, and to measure the contact resistance as a function of temperature.

This dissertation is divided into four chapters. The first chapter describes
fundamental thermoelectric phenomenon with discussion of LAST and LASTT material
properties along with a review of the literature related to contact resistance studies. The
second chapter includes diffusion bonding results and diffusion barrier coatings on metal

electrodes. The third chapter discusses liquid contacts and spring loaded modules output

short circuit current measured under ambient conditions and inert environments. In the
fourth chapter the design of a voltage profiling system is given along with experimental
results of high temperature contact resistance measurements. Due to funding constraints,

my research focus has changed. My initial research efforts in this group were towards

nanowire synthesis for sensor applications. As part of that effort, GeOz, ZnO and InSb

nanowires were synthesized by vapor-liquid-solid (VLS) mechanism. Moreover ZnO and

GeOz nanowires were grown through direct heating on hot plate under ambient

conditions. These results are mentioned in appendix A of this dissertation.

1.1 Thermoelectric Materials

Thermoelectric (TE) generators convert heat energy into electrical energy when a
thermal gradient is maintained across the device. The process is reversible such that
thermoelectric coolers (also known as Peltier coolers) operate by supplying electrical
energy to the thermoelectric module resulting in a temperature gradient being established
across the module. These devices are rugged and have no moving parts which make them
ideally suitable for harsh environments and unmanned places. The focus of this research
is on TE generators, and this chapter begins with descriptions of the governing

thermoelectric effects of the Seebeck effect, Peltier effect, and the Thomson effect.

1.2 Seebeck Effect

If the two materials are joined at two junctions such that a closed loop electrical
circuit is formed there will be a current in the close loop given that there is a temperature
difference between the two junctions. Thomas Johann Seebeck originally observed this
phenomenon when a compass needle placed close to such a loop moved when junctions
were maintained at different temperatures. The current resulted in a magnetic field which
moved the compass needle. He also observed that magnitude of deflection depended
upon the amount of heat provided at the junction and type of material used. Typical
voltages from the Seebeck effect range from a few microvolts per Kelvin for metals to
several hundred microvolts per Kelvin in semiconductors, and into the millivolts per
Kelvin for insulators. The Seebeck coefficient is the voltage developed for a one Kelvin
temperature difference maintained between the ends of the conductors. Seebeck

coefficient is given as:

dV
Ctr—FY;— (V/K) (l)

and by integration, the voltage can be found as:

T

2
V: jadr (V) (2)

T1

1.3 Peltier Effect

The Peltier effect was discovered by Jean Peltier in 1834 and states that heat is
liberated or absorbed when electric current passes through the interface between two
different conductors. It is a reversible effect, such that if the direction of the current is
reversed, the direction of the heat exchange at the boundaries of material will also

change. The Peltier effect is based on the fact that heat current accompanies electric

current in homogenous conductors. The Peltier coefficient ['1 AB is related to individual

Peltier coefficients as:

“AB = HA {I}? (3)

The Peltier coefficient is related to absorption or liberation of heat through [2]:

dQ .
WEAR , <4>

where

i=5?—

5
dt ()

Here q (q = -1.602 x 10.19 C for an electron) is the charge. Above equation can be

integrated to give:

AQ=HAB Aq (6) '

This is related to the Seebeck coefficient as:

HAB =TaAB (V) (7)

1.4 Thompson effect

The Thompson effect was discovered by William Thompson (Lord Kelvin) in
1854 and states that heat is liberated when an electrical current is passed through a
homogenous conductor which is subjected to a temperature gradient. The heat is
absorbed or emitted based on type of material, direction of the current flow relative to
temperature gradient and therrnopower of the conductor [3].

The Thompson coefficient ([3) and Seebeck coefficient (a) are related as:

fl = :—;‘—.T (WK) (8)
or
a = Ijfldr' (WK) (9)
0 T.

Thompson coefficient can be related to the Thompson heat and Thompson current as:

_£_Q_=_
TQ _ dT ,BIAT (W) (10)

For materials like Zn and Cu, the higher temperature end of the conductor is at a higher
potential, so heat will be liberated if current moves from hotter end to the colder end. It is

known as positive Thompson effect. For materials like Ni, Fe and Co, the colder end will

be at higher potential, so heat will be absorbed when current moves from hotter end to

colder end, known as negative Thompson effect.

1.5 Figure of Merit (ZT)

Thermoelectric materials are characterized by a dimension less figure of merit ZT.
Z T = ——T (l 1)

Here a is the absolute Seebeck coefficient, a is electrical conductivity and K is thermal

conductivity. It indicates that a good thermoelectric material will have high electrical
conductivity and low thermal conductivity. Thermal conductivity of a material is its
ability to conduct heat. Both phonons and charge carriers participate in heat conduction in

metals and semiconductors.

Ktotal = Kelectrical + Klattice

Materials like metals which are good electrical conductors are also good thermal
conductors and materials like quartz which have low thermal conductivity also have low
electrical conductivity whereas; thermoelectric materials need to have low thermal
conductivity and high electrical conductivity. At low temperature there is very little
vibration of lattice and thermal conduction is through charge carriers and at higher
temperature lattice vibration dominates and much of the heat conduction is through
phonons. In between these extremes both charge carriers and phonons participate in
thermal conductivity. The relationship between electrical and thermal conductivity can be

given by the Wiedemann Franz Law:

5 = LT (WQK'I) (12)
0'

Here L is proportionality constant and is known as Lorenz number.

For metals:

2 k _ -
L=i=-’5—(—B)2 =2.44x108 (WQK 2) (13)
OT 3 e

This value of Lorenz number is applicable for temperatures above the Debye 9]) value.

This law indicates that charge carriers are involved in both electrical and thermal
conduction. For LAST material with a typical room temperature value of electrical

conductivity of 1850 S/cm [1], thermal conductivity of 2.3 W/m-K [1] and the value of
-8 —2 . -2
the Lorenz number of 2.28 x 10 WQK [4] gives the value of Klauice= 1.27 WQK .

We can easily now calculate the value of Kelectronic=1-O3 WQK-z'

1.6 Thermoelectric Applications

Fossil fire] has been the energy work horse for more than a century, however, this
resource will deplete one day, forcing humans to look for alternate energy resources.
Thermoelectric generators and coolers typically are small in size and have no moving
parts and are more reliable because of their rugged solid state construction. Their silent
operation makes them ideally suited for military applications like submarine power
generators where silence is paramount for undetected submarine operation. In areas like

space exploration where human intervention is not possible on unmanned missions and

there is no oxygen available, radioactive isotope thermoelectric generators (RTG) are
used. TE generators are also eco-fiiendly unlike hydrocarbon where exhaust gases are
harmfirl to the environment. Thermoelectric coolers have been commercially
manufactured to serve as heat pumps. They can enhance cooling of the CPU in computers
for overclocking. Thermoelectric coolers can be cascaded for applications requiring
larger temperature gradients for example a six stage cooling device capable of achieving
below 170 K has been manufactured [3]. These devices can provide distributed cooling
and heating. Another potential application area is automotive industry. About 65% of fuel
energy is lost as waste heat in a typical diesel vehicle [5]. If waste heat recovery can be
utilized to meet the accessory load requirement of over the road vehicles, more fuel
energy can be used for propulsion. A typical alternator on a diesel vehicle can consume
10 kW of fuel chemical energy to produce 1 kW of output. If the alternator is replaced by
an integrated TE radiator, overall vehicle efficiency can improve by 3.3 % [5]. However
this task has a number of challenges that must be considered in design which include
additional weight of TE modules, increased complexity in power management design,
engine backpressure concern and aerodynamic drag increase. Application of
thermoelectrics within the medical field have included TE based cooler box
(Thermoelectric Cooling America Corporation, TECA) to carry intravenous solutions.
Other applications include cold plates for tissue freezing, cooling tank for beverages,
Peltier stirring cold plates, liquid chiller, cooling of infrared detectors, air conditioning
for small enclosures, thermal blankets for treating hypothermia, and therrnoelectrically

driven wrist watches.

1.7 Lead-Antimony-Silver-Tellurium (LAST) TE Material

Different TE materials are suitable at different temperatures i.e. their ZT values

peak in a certain temperature range and no single TE material can be used for entire

temperature range. Traditional materials used near room temperature include BizTe3,

szTe3 and BizSe3 are used up to temperature of 400 °C [6]. These alloys have ZT better

than unity. From 400 °C to 600 °C range, historically PbTe based materials are best suited
and above this range SiGe has commonly been the best material. In the last decade new
materials such as skutterudites, calathrates, and half-husler alloys have been reported
along with nanostructured variations of traditional materials [6]. Goldsmid and Nolas
have discussed various new materials and have indicated their preference of higher power
factor over lower thermal conductivity for TB materials [6]. High power factor can be
obtained by higher effective mass or higher mobility. As for lattice thermal conductivity,
it can be reduced by preferential scattering by grain boundaries. Scattering from the grain
boundary can reduce thermal conductivity even if grain size is larger than phonon mean
free path [6]. Larger grain size may not affect carrier mobility unless career free path
becomes comparable to grain size. Bhindari [7] has discussed various methods to reduce
TE material thermal conductivity that include introduction of point defects in alloys
(alloy disorder), resonant scattering, scattering by charge carriers and scattering by grain
boundaries. Most of these efforts relate to limiting the phonon mean free path [7].

For LAST material, low thermal conductivity is believed to be due to the
structural and compositional inhomogeneity at the nanoscopic level [8], which is

responsible for interface phonon scattering by nanostructures of Ag/Sb. The size of these

10

‘1 -—..--I -

nanostructures was found to be ~ 5-20 nm [9]. Another possible reason for low thermal
conductivity of LAST material is microcracks in bulk material [10]. Androulakis et al [9]
attributed LAST low thermal conductivity to Pb/ Sn regions which reside in the
surrounding Pb matrix without disturbing its electronic structure. Quarez et al studied
LAST material and found that material has resonant states near Fermi level and Ang

microstructures in Pb/T e matrix may be responsible for higher thermopower in LAST

[11]. Earlier it was proposed that AngTez — PbTe is a solid solution [12], however

Quarez et al showed this is not the case and there are nanostructure periodic domains

[1 1]. There is large anisotropy in carrier effective mass for such systems [1]. For p-type

material Ag(Pb1-y-Sny)meTe2+m, the TB parameters can be controlled by adjusting

Pb/Sn ratio and by tuning the Ag/ Sb nanostructures. Increasing Sn increases the electrical
conductivity and thermopower has been found to vary linearly with Sn concentration
reaching approximately 280 (uV/K) at 680 K [9]. For cast samples the electrical
conductivity decreases with increasing temperature and at about 650 K is 22%-28% of
room temperature electrical conductivity [9]. The magnitude of the thermopower
appreciably increases in the 330 K to 550 K range which is attributed to an increase in the
effective mass with increasing temperature [9]. Lattice thermal conductivity values for
LASTT were found to be ~ 0.70 W/m-K at room temperature and ~ 0.43 W/m-K at 640

K [9]. This indicates that there is significant change in lattice dynamics at higher

temperature and it deviates from Debye-Peierls law of Klatfice ~ UT [9]. The highest ZT

ll

achieved for a p-type material was ~l .45 for Ag0_5Pb6Sn28b0.2Te10 composition at 627

K [9]. Bothe LAST and LASTT materials have NaCl structure (F m3m symmetry) [1].
To increase strength and sample homogeneity, powder processing and hot

pressing have been investigated. Some of the initial hot pressed samples showed bloating

after annealing. Both p-type and n-type ingots release C02 at annealing and p-type

sample release Te or Tez on annealing under vacuum at high temperature [13]. It is

assumed that carbon enters the system during powder processing or from graphite dye in
the hot press system [13]. Hot pressed samples also show increasing electrical
conductivity with increasing temperature. Kosuga et a1 [14] also reported an increase in
electrical conductivity with increasing temperature for hot pressed materials as did Zhou

et al [15] for spark plasma sintering.

1.7.1 Lead-antimony-silver-tellurium (LAST) Mechanical Properties

During thermoelectric materials applications thermo-mechanical stresses are
generated from thermal gradient across the TB material, thermal expansion mismatch
between TE legs-hot side electrode, mechanical vibrations and thermal cycling. These
stresses can create micro and macro-cracks resulting in material failure. For better
understanding and to avoid mechanical failure under thermo-mechanical stresses LAST
material mechanical properties that include hardness, Young’s Modulus, Weibull

modulus and bending strength were investigated in [13] and [16].

12

1.7.1.1 Hardness

Hardness (H) of a material is its resistance to shape change when a force is
applied. It has been used as a tool to characterize material mechanical properties as a
function of alloy compositional changes [13]. Vickers indentation technique is often used
to determine hardness where material’s ability to resist plastic deformation from a

standard source (load) is measured in Pascals. Hardness is also associated with

machinability and wear of materials e. g. wear in ceramics is related to hardness by Kc/H

where Kc is the fi'acture toughness. Hardness is also related to compressive fracture

strength ~ H/3 [13]. Wang et al found that material machinability is correlated to both
Young’s Modulus and hardness [17]. Since for thermoelectric generator fabrication
LAST material will require machining of legs from ingots, hardness is an important

material property to be considered. The mean values of hardness (H) for cast type LAST
(AganbSbcTed) with mole fraction ranges 0.006 <a< 0.043, 0.417 < b < 0.480, 0.011 <
c < 0.050, 0.496 < d < 0.517 were found to be 0.548 to 0.922 GPa [6]. For comparison,

PbTe hardness falls in the range of 0.339 to 0.451 GPa and Zinc antimonide has the

highest hardness value of 2.24 GPa.

1.7.1.2 Young’s Modulus

Young’s Modulus (E) is the measure of stiffness of a material and is defined as

the ratio of uniaxial stress to uniaxial strain:

E=0’/8

l3

Here a is the stress on the material when it is subjected to strain a and E is young’s

modulus. Since thermoelectric generators are subjected to thermo-mechanical stresses
during operation (0 = E 8), Young’s Modulus is an important parameter to characterize

material response. For cast type LAST material Young’s Modulus was measured using
microindentation (indentation loads of 2.94 N and 0.49 N) and nanoindentation.
Measured values ranged from 24.6 GPa to 71.2 GPa [13] for different specimens. Kosuga
et a1 [14] measured Young’s Modulus for hot pressed LAST material and values ranged
between 27.6 and 54.2 GPa. For comparison, Young’s Modulus value for PbTe is 58 GPa

and SiGe has highest value of 137-145 GPa [16].

1.7.1.3 Weibull Analysis and bending strength

For brittle materials Weibull modulus is a measure of the distribution of flaws in
the material which is used to obtain information about the scatter in strength expected
from a set of specimen [l3]. Bending strength (or fracture strength) represents upper limit
of normal stress at which fracture or plastic deformation of material takes place. The

Weibull relationship for fracture strength distribution is given as:

lnln(l/(1-Pf)) = mln(o/oo) = mln o — mln 00

Here m is Weibull modulus, 00 is a scaling factor and PfIS probability of failure.

Weibull modulus for LAST material was found to be 3.16 which is a small number that
corresponds to extensive scatter in fracture strength whereas desirable modulus range is
10-30. The bending strength for a cast type LASTT material was measured to be 15.3

MPa and for hot pressed LAST material its value was determined to be 51.6 MPa [16].

14

For comparison, the bending strength for BizTe3 ranges fi'om 8 to 166 MPa. No bending

strength data could be found for PbTe material.

1.8 Calculation for thermoelectric generator out put and efficiency with and

without contact resistance

Min and Rowe have described the effect of contact resistance in comparison to
ideal module calculations [3] for a thermoelectric generator. For an ideal module power

output P is given by

_ (2adT) 2
4R

Here dT is the temperature difference between hot side and cold side of module, a is

P (14)

Seebeck coefficient and R is the total resistance of the module given by

L (15)
R=2 —
pA

Here L is the length, A is the area of thermoelement and p is electrical resistivity.

Replacing R in equation will give

 

 

2
P = Ad Ta (16)
2Lp

To account for thermal and electrical contact resistances:

Pr = P 2 (17)

(1+ n/L)(1+ 2rw)

here

(l+2rw)=AT/dT (18)

15

Copper strip

1

‘TT
I AT dT

LC P 1/ j
T] r/

Ceramic plate

LC L.

 

 

 

 

Copper
interconnect

Figure 1-1. Thermoelectric unicouple.

Here AT is the temperature difference when two alumina plates are attached to top and
bottom of the module and r = zl/xlc where A is thermal conductivity of thermoelements and
Ac is thermal conductivity of contacting layers. Also w = Lc/L where Lc is the thickness
of contacting layer as shown in Figure 1-1. In equation (1 7) term n/L is taken from the

expression:
L n
R =2 — 1+ — 19
a p A( L) ( )
Here Ra = R +Rc and RC: 4po/A and n=2 pc/ p where Rc is electrical contact resistance

and pc is contact resistivity. The efficiency of an ideal thermoelectric generator is given

by

 

 

-l
T -T T —T
n=[h 6]2—1[h 0}. 4 (20)

16

Efficiency of a realistic thermoelectric generator including thermal and electrical contact

resistance is given by:

-1

T -T T -1 1+3
n=[ hr C] (1+2rw)2 2—%[l—i]+[ 4 ] L (21)

h Th ZTh 1+2rw

 

 

Here Z is thermoelectric figure of merit and T h and T c are hot side and cold side

temperatures respectively.
Cobble [3] has given a description of thermoelectric efficiency with and without
including contact resistance. The efficiency of a thermoelectric generator without contact

resistance is given by
2: =P/Q (22)

Where electrical power out put is

P=12R (23)
0
And heat input is given by
Q=KAT+aTI—%12R (24)

Following relationships are used for above equations:

I = V / R t (25)
Where V is open circuit voltage and R, is total resistance given by

R t = R + R0 (26)
If it is thermal conductivity then total thermal conductance is

K = A" 7n + 1p yp (27)

17

here
7i=Ai/Li ,z=n,p

If we define

ATzTh —TC
and
a=|a |+ a j
n P

 

R=,0n/}In+pp/yp
V = aAT
p = R0 / R

We define Carnot efficiency as

nc=AT/Th

The expression (22) can be written as

MC

”t = 2 2
[1010: +1) /a Th +(p +1)—nc /2]
In above expression

102:2" pn +2npplr+2ppn /\11+,1ppp =KR(‘P)

where ‘1’ = 7n / 7p

Thus 77, =77, (l1, ‘1’)
Thermoelectric figure of merit is given by
Z=a2 /KR= 2011)

Efficiency is then given by

18

(28)

(29)

(30)

(31)

(32)

(33)

(34)

(35)

(36)

(37)

(38)

(39)

#776. (40)

nt:[(,u+l)2/ZTh +(y+l)-nc/2]

 

For thermoelectric efficiency with contact resistance included the input heat in expression

 

 

(24) is now given by
Q=KAT+aT I—lIZR—IZR (41)
h 2 ch
With total resistance R, given by
Rt =R +R0 +Rc (42)
and
Rc = Rch + Rcc =Rchl + Rch2 + Rccl + Rcc2 (43)
Let us define
6=6C=RC/R and 50h =Rch/R (44)
Efficiency is now given as
#770
n, =— 2 2 (45)
[KR(;1+1+6 j/a Th +(,u+1+6)—¢cnc/2]
We can write above expression in terms of figure of merit Z
#77
C (46)

77 :
t [(21+1+6)2/2Th+(,u+l+6)—(anc/2]

l9

Conclusions

LAST and LASTT are promising thermoelectric materials with high ZT and good
thermoelectric properties. The properties of LAST and LASTT materials have been
investigated and prototype modules have been assembled for contact studies and power

output characterization.

20

Chapter 2: Thermoelectric Module Bonding and Diffusion Barriers

Introduction

Electrical contact resistance causes a drop in thermoelectric module efficiency
and overall increase in module resistance [18] so it is desirable to have low contact
resistance. Typically the contact resistance should be less than the 10% of the resistance
of the thermoelectric materials making up the module. The selection of a suitable hot side
contact material depends on various factors which include high thermal and electrical
conductivity and coefficient of thermal expansion (CTE) matching with the
thermoelectric material. Yarnashita et al [19] studied the effect of different electrodes
(Cu, Ag, Zn, Ni, Pt, Au, Mo, Al) work function on the effective Seebeck coefficient of
bismuth telluride compounds and found that it had negligible effect. They also found that
both n-type and p-type legs display similar (linear) behavior for all electrode types;
indicating ohmic contacts being formed in all cases. Another important consideration is
the diffusion coefficients of the electrode elements into the thermoelectric material which
is governed by Fick’s laws and depends upon flux concentration gradients. Some

materials like Cu quickly diffuse in bismuth telluride and deleteriously alter the

thermoelectric properties of Bi2Te3. To avoid this, diffusion barriers are commonly used

and have been investigated as part of this research effort for LAST and LASTT materials.
Thermoelectric bonds should be able to survive thermal cycling and remain stable at high

temperatures. In this lab we have experimented with diffusion bonding, soldering, liquid

21

contact with InGa and spring loaded modules. Here I will discuss solder and diffirsion

bonding results.

2.1 Diffusion Bonding

Diffusion bonding is used to bond metals with metals, alloys and ceramics. The
diffusion mechanism involved in this bonding takes place along grain boundaries and
through vacant lattice sites and can be described by:

- /kT
D=D0 C ( E )

Here Do is diffusion coefficient and it represents the amount of solute migrating across a

unit cube of solvent in unit time with unit concentration gradient [20]. E is the activation
energy required to move atoms from one site to another. This expression shows that
diffusion is temperature dependent process. During diffusion bonding mating surfaces are
pressed together and heated to bond surfaces through interdiffusion of atoms across this
interface. Pressure and heat during bonding helps to eliminate voids at the surfaces which
are filled by diffusion. These voids may also contain surface oxides, which are purged
during the process. This process is carried out in an inert or reducing atmosphere to help
avoid oxidation. The interface bond can be mechanically strong or weak depending upon
the materials used. The applied temperature is usually 0.75 Tm where Tm is the melting
point of lower melting point material being bonded [21]. However, the bonding
temperature at the same time should be high enough to enable dissolution of surface
oxide fihns.

In some cases solder bonding is used instead of diffusion bonding when an

appropriate solder can be found. Some diffusion bonding processes also involve a

22

transient liquid phase [21]. Often this interlayer is needed to bond dissimilar metals or
alloys. At high enough temperature this intermediate layer becomes molten. A liquid also
posses surface energy and tend to draw up into droplet shapes where its surface energy is
higher than its volume energy. Surface of liquid acts as elastic skin and is in the state of
tension where under isothermal conditions surface energy is equal to surface tension. The
liquid will wet the surface if all three surface tension forces are in equilibrium. These
surface tension forces are between the liquid droplet and solid substrate, liquid droplet
and the atmosphere and the substrate and the atmosphere [22]. When an intermediate
layer of filler or solder material is used, it will form intermetallic phases after heating and
this part of the bond will have different mechanical strength, often less than the original
material. In bimetallic systems, formation of intermetallic compounds can be
accompanied by an increased porosity that takes place due to different rate of diffusion of
species and the layer thickness is less than lOOum [22]. Good wetting of the surface
enhances the formation of filler between layers and results in good bond. For diffirsion
bonding the mating surfaces should be smooth to minimize the size of interface voids.
Moreover, if oxides can not be purged in a reducing atmosphere during bonding then
plastic flow of asperities should be strong enough to scrub away the surface to remove
oxides by mechanical dispersion [21]. Grain boundary diffusion can also play an
important role in diffusion bonding such that an increased number of grain boundaries
enhance atomic diffusion across the interface [23]. More than one mechanism are
involved in diffusion bonding [22] that include creep of the surface asperities, plastic
yielding of surface asperities, volume and surface diffusion and as mentioned above,

grain boundary diffusion at the bond interface. The four main process variables are

23

temperature, pressure, time and surface condition as creep and diffirsion are temperature
and time dependent. Plastic deformity depends upon temperature, pressure and the
materials being bonded. Surface asperities’ height and frequency determines the initial
surface contact and influence the bonding rate [22]. In our experiments an interlayer is
used that melts and fills the voids, wets and disrupts thin oxide films on mating surfaces

for a better bond.

2.2 Diffusion Furnace

Bonding experiments are carried out in a three zone diffusion system [23]
designed and assembled by Jonathan D Angelo. The three zone furnace has a 12 inch
long center zone with 6 inch zones one on each side through which an 8’ long quartz tube
extends. Approximately 2’ of tube extends out each end of the furnace to allow for
cooling of the tube before the O—ring seals on the ends of the tube. Quartz tube has an
inside diameter of over 5” to accommodate larger module assembly. The bonding range
for LAST based materials is ~ 700 °C. Furnace consists of 8 feet long quartz tube. Quartz
tube end caps provide a vacutun seal and the chamber is evacuated using a dry pump to
less than 10 mTorr vacuum. The 9” isoflange end caps can be easily removed and angle
brackets are provided to hold them in place for back filling with Ar gas. The vacuum
pump and pressure gauge are attached to the vacuum port end-cap and a valve is used to
isolate the firmace from the vacuum pump. Three PID controllers are used to control the
temperature of the three zones. The controller can be set to ramp and soak time as desired

by user and typical ramp time is 4 hrs followed by soak time of 4 hrs. To maintain

24

isothermal region and firrther isolate the center zone circular disks of stainless steel are

inserted in the quartz tube as heat shield.

2.3 Room Temperature Scanning Probe system

For room temperature scanning of thermoelectric material’s electrical
conductivity and contact resistance of unicouples a room temperature scanning probe
system
(RTSP) is used as shown in Figure 2-1 below. This system gives voltage vs. position
slope (d V/dx) of TE specimen under test. The slope is then used to calculate electrical
conductivity using the following equation:

0-; 1
d_VArea
dx

 

S/cm (1)

System consists of one Keithly 2004 current source and Keithly 2002 nanovoltrneter.
Current is sourced through the sample at approximately 33 Hz with 50% duty cycle
square wave function. Voltage is measured by a probe scanning along the sample surface.
The current is flipped approximately 20 times at each measurement point. This is done to
avoid any Seebeck voltage development in the sample. It also cancels out voltage offset
in measurement wires developed due to voltage gradient. The system has three axis (X,

Y, Z) of motion to adjust the sample and then scan the measuring pin across the sample.

25

 

Figure 2-1. Room temperature scanning probe system.

2.4 LAST to LAST bonding

To circumvent the issue of CTE mismatch, it is ideal to use a hot side electrode
with same coefficient of thermal expansion (CTE) as the thermoelectric (TE) element.
This makes LAST a suitable candidate as hot side electrode. However, LAST has
inherently low thermal and electrical conductivity compared to metals like Cu, Ag or Fe.
It is possible to heavily dope LAST to increase thermal and electrical conductivity but
then it might change the coefficient of thermal expansion of material. To check this
hypothesis two pieces of thermoelectric material were bonded using an interface paste
consisting of Sb powder + LAST powder mixed in ethylene glycol. Bonding was carried

out at 700 °C with a 2 hr soak time.

26

LAST-LAST
Sb-I-Ethylene glycol+LAST powder

 

N
0'
O
O
l

”.000”. ...

LAST

N
O
O
O
I

P
_x
U"
C
O
U

‘
: t
31000 E’ F Rc > 200 linem:z

3
> : LAST

0|

0

O
I
r

i
ll] Iljlllll

0 :- ...”...0000000

-500’..i1...l...1.n.1...
0 2 4 6 8 10

Position (mm)

lLllll

 

 

 

Figure 2-2. LAST to LAST bonding.

Figure 2-2 above shows an n-type to n-type LAST material bonding experiment. Room
temperature scan indicates a high resistance bond. Mechanically it was weak bond and

broke off easily.

27

LAST-LASTT

2500 _ . §9+Ettv3°9°filvc£l+E€ST gorse: .

 

JLJ

looooooooooeoooe

2000 2

I‘ll.

1500 3

I
l

4 Re > 200 pflcmz

lillll

. 1

Voltage (11V)
8
CD
CD

 

 

 

500 - r
.0 1
- .000.... :
0 _- «0' bonding 1
F juncflon
_500 . . . l l . . I . I L 1 . I . r . . .
O 2 4 6 8 10

Position (mm)

Figure 2-3. LAST-LASTT bond.

Figure 2-3 above shows a p-type to n-type cast materials bond at ~ 700 °C. Antimony
(Sb) and LAST powder mixed in ethylene glycol were used as intermediate bonding
material and bonding time was 2 hrs. It appears that Sb and LAST did not melt and
diffuse during diffusion bonding process. Very high contact resistance at junction makes
this unsuitable for any application.

Figure 2-4 below shows two p-type cast samples bonded at ~ 700 °C in the same
run as above. In case of p-type samples a mixture of ethylene glycol and LAST powder
reacts with thermoelectric material during bonding process and allows interdiffusion to
make a successful bond. The small kink at the junction may be due to interdiffusion of

extra antimony (Sb) into joining materials. It can be seen that slope is different at either

28

side junction indicating different electrical conductivities. It may be due to antimony

diffusion to different batch materials (ETP 47 and ETP 51).

 

 

 

 

[1...1....1.............................
160 1 LASTT-LASTI' :
' Sb+Ethylene glycol+LAST powder ... Z
140 - ...e' .3

u C
120 - .0. .'
A . o .
3100 '- LASTT i'. 1
o I ° I
E 80 E- .' LASTT -:
§ : 2 ' :
50 . Rc<25uncm .' ..
. o q
40 .' J. bonding junction '1
20 '- '° .'
0 E .... I
1 .

-1 0 1 2 3 4 5 6 7

Position (mm)

Figure 2-4. LASTT to LASTT bond using Sb and LAST powder.

29

LAST-LASTT (HPMSU N-25-CAST p-type)
7000 direct bonding

 

 

 

 

6000 :- °’-j

. .0 .

I o' I

5000 - ‘ .0. -
S : ! .0 5
a : : .0' 3
g, 4000 :- 1 ....cast p type 1
3 : ‘ e. - :
25 3000 - LAST ,.- -
> : HPMSU N25 . :
I ..." bonding temp 1

2000 E' ....0 t ~ 750 c ‘:
1000 L °"° Rc < 25 uflcmz _'
2 4 6 8 10

Position (mm)

Figure 2-5. LAST-LASTT direct bonding.

Figure 2-5 above shows direct bonding between n—type hot pressed samples with a cast p-
type sample with bonding temperature ~ 700 °C for 2 hrs where a low resistance bond is
obtained as evident by a negligible jump in voltage across the interface. Although this is a
n-type to p—type junction, no interlayer formation is observed and one ohmic interface is
measured. The same module was tested in the high temperature voltage profiling system

where scan taken at 700 K again show low contact resistance as seen in Figure 2-6.

30

 

 

 

 

 

 

2000..........4.......f
High Temp -LAST-LASTT ~700K ,
O
1500 L ...“ .‘
t LAST ,.° :
5" I (MSUHP N-25) .o' 1
31000 )- ..0° 1
O F 0.. LASTT -
g, : RC < 25 p.96"!2 . A (p-type)
3 500 - ... ..
> : ..-
. .... ‘
0 L .- -
i bonding junction I
-500 . . . . r . . . r 4 . . r . . . 1 d

2 0 2 4 6 8
Position (mm)
Figure 2-6. High temperature scan for LAST to LASTT bonding.
A number of bonding experiments were carried out to directly diffusion bond two p-type

samples together without any interlayer however efforts proved to be unsuccessful and

the bonds were often too fragile and broke off easily.

31

350 .1.

 

 

 

 

E ETP 50-ETP 47 ., 3
30° 3 Sn/Pble foil 3
250 E" bonding temp ~ 650 °C .... '.
’>" 5. ..' 3
3200 : 2 ... :
§150 _ Rc< 25 119cm .... _:
E 100 i ’3’
50 L ....eee .:
I ,v" bonding junction .
0 _- O. ..

2 4 6 8 10

Position (mm)

Figure 2-7. Room temperature voltage scan of the LASTT to LASTT bond using a

solder foil at the interface.

For the next experiment a solder foil was used as an interface bonding material
consisting of Sn/Pb/Sb solder foil 0.006 mm thick (Goodfellow) placed between two cast
p-type samples and bonded at ~ 650 °C. A low resistance bond was achieved as seen in
Figure 2-7. A similar experiment was carried out using a Sn/Pb/Sb foil but this time
bonding temperature was 750 °C. As we can see fiom the room temperature scan in
Figure 2-8 a low resistance bond is obtained and there is no kink at the junction which is

slightly different from previous plot.

32

250 . .

 

 

I 1 T l .
ETP47 to ETP51 ..0'
Snle/Pb foil T ~ 750C ..-
200 - ,- -
.0
O...
2 "
$150 b Rc < 25 an cm .... _
a .
a .o' ETP51
O) .0
§ 100 - ,-° 4
O o
> .°
O.+.
50 .-' -
ETP47 ...-
..-
O Ix.’ J l l l l -

 

 

 

-2 O 2 4 6 8 1O 12
Position (mm)

Figure 2-8. ETP 47 bonded to ETP 51 using a Sn/Sb/Pb solder foil.

Figure 2-9 shows a bond between two p-type cast samples at ~ 650 °C. A low melting

point Cerromatrix solder (Bi4ng28.5Sn14.58b9.0) was used between the samples as an

interface bonding layer. The initial process was carried out on a hotplate under ambient
conditions to wet the sample mating surfaces. Sample was then placed inside diffusion
furnace and bonded at 650 °C and 800 Torr pressure. The solder was expected to reflow
under the high temperature conditions and completely wet the mating surfaces but room
temperature scan indicates that it is a high resistance bond. However, inspection of the

interface under an optical microscope revealed significant number of voids and gaps at

33

the interface suggesting poor wetting and/or oxides at the interface. From above

experiments it appears that solder foil is a better choice as intermediate bonding material.

 

 

 

ETP 50- ETP 47
350 PblSn/Bile cerromatrix solder
300 E; .........
E m"' ETP 47
250 ;- Rc ~ 160 ”mm2
3200 E ,'
a: .0. . . .
E150 :' ,o' bonding junction
— I o
g 100 E' ETP 50...-
. ..
50 E- ..' bonding temp ~ 650 °C
I e
0 - o.
2 4 6 8 10
Position (mm)
Figure 2-9. LASTT to LASTT bond using solder.

LASTT bonding with hot pressed material using SnTe powder also showed low contact

resistance junctions. For this powdered Sn50% by weight was added to Te50% by weight

and mixed for about 15 minutes by shaking in a glass container. The mixture was then

layered between LASTT and an n—type hot pressed LAST leg. Bonding was carried out at

~ 750 °C for about 30 minutes. As shown in room temperature scan in Figure 2-10 it is

low resistance junction. Since hot pressed samples have very low room temperature

electrical conductivity (< 100 S/cm) as compared to cast p-type samples which have

34

conductivity in the range of 1800 S/cm, the hot pressed sample has a relatively steep

slope on the voltage vs. position plot of Figure 2-10.

 

7000 l l l I l f
O
ETP47 to HPN35 o’
6000 Sn+Te powder .... '-
. O
5000 _ bonding T 750 C .. _
...
g 4000 - .-’ -
EL- .3
g, 3000 - ...? -
3 .’
> 2000 - .0 HP N35 ‘
O
2 O
0 F W 4
ETP47

 

 

 

4000 1 1 1 1 1 1
-2 0 2 4 6 8 10 12

Position (mm)

Figure 2-10. ETP 47 bonded to HPMSU 35 using SnTe powder.

35

250 .

 

 

 

ETP47 to ETP51 ..-
Sn-I-Te powder ‘
200 b bonding T ~ 750C 0 -
,a
:-
2
3150 . Rc < 25 pflcm .3 ..
O
8: 0'.
g ....
£100 ' interface '0’ '
.0
o‘
a ETP51
50 - -
ETP47 a...
‘4-
0 J I 1 I l

 

-2 0 2 4 6 8 1O 12
Position (mm)

Figure 2-11. ETP 47 bonded to ETP 51 using SnTe powder.

Sn and Te powder mixture was also tried for LASTT to LASTT bonding and it produced

low resistance bonds as shown in the Figure 2-11 above. The results for LAST to LAST

and LASTT bonding are summarized in Table 2-1 below.

36

Table 2-1. LAST to LAST bonding summary.

 

 

 

 

 

 

 

 

 

 

 

 

 

TE Material Bonding material T (°C) comments

Cast LAST Sb+ ethylene glycol 700 High resistance bond
Cast LAST + LAST powder

Hot pressed n-type Direct bonding 700 Low resistance bond
Cast LASTT

Cast LAST Sb+ ethylene glycol 700 High resistance bond
Cast LASTT +LAST powder

Cast LASTT Sb+ ethylene glycol 700 Low resistance bond
Cast LASTT +LAST powder

Cast LASTT Pb/Sn/Sb foil 650 Low resistance bond
Cast LASTT

Cast LASTT Cerromatrix 650 High resistance bond
Cast LASTT Solder

Hot pressed n-type Sn+Te powder 750 Low resistance bond
Cast LASTT

Cast LASTT Sn+Te powder 750 Low resistance bond
Cast LASTT

 

2.5 Room temperature scan of stainless steel bonded unicouples

Initial experiments in the bonding fumace included direct bonding of stainless

steel 316 with cast LAST thermoelectric 5mm x 5mm x 7mm legs without any

intermediate bonding material. Six unicouples were scanned for room temperature

contact resistance. Unicouples were bonded by Jonathan D’Angelo and analysis was

carried out in collaboration with Jarrod Short in this lab.

37

 

Figure 2-12. Six unicouples tested for room temperature contact
resistance. (a) Low resistance p-type bond (b) N—type sample cracked (c) High.
resistance bond for both types of legs. (d) Low resistance bond for n-type.

(e) P-type sample cracked (f) High resistance bond for both legs.

38

 

Voltage (11V)

1000

§

§

§

§

 

 

 

 

 

' Voltage (11V): P
D Voltage (uV):N

 

(IN = 500 S/cm ,

   
 
  

1

 

 

 

 

 

 

 

  

 

 

6 10 12
Position (mm)
‘ Voltage (11V): P-type
0 Voltage (11V): N-type '
Up = 417 S/cm '
c = 714 S/cm
N J
(b)
2 4 6 8 10 12 14

Position (mm)

39

 

 

 

 
 

 

 

 

  
  
 

 

 
 

 

 

 

 

 

 

 

 

 

14

 

Figure 2-12. Continued.
560 ' ' ‘ ' ' ' ' 1
' Voltage (uV): P
480 - 0 Voltage (11V): N
A400 - 0'
>
3320 . P
81
53 240 -
O _ . 2
> 160 Rd, — 18811.!) cm , ‘
80 . rrcN=119110~c1112 ,. ‘
O 43191125125}.1.1.-:5:1 (c) _
0 2 4 5 8 10 12
Position (mm)
1200 1 . , , , . T
1000 . ‘ Voltage (pV): N ,
0 Voltage (11V): P
800 - . .
9 0' = 424 S/cm ,4:
N ...s‘
3: 600 - ., e .
w .
or 2 :2” ,
2'3 400 I'RcN = 205119°cm x-‘V: VJ»
>0 2 .. 3 M
200 ~RcP=292pQ-cm .
0P=857 S/cm _
-200 I 1 1 1 1 r 1 (d)
2 4 6 8 10 12

Position (mm)

40

Figure 2-12. Continued.

1600 ,

 

 

 

 

. Voltage (”V): P
E Voltage (”V): N
1200 - - q
9
3.
: 800 L = 700 S/cm -
an P
2 R = 75 °cm2
g 400 - cP “‘2

   

RCN = 29811521113 .. . " ...’.-- , , _ _".{_fg,,_1:..=-.111
, ,5. ON = 2000 S/cm -

(e)

 
 

   

 

 

 

500. . . . . . .

VOItage (14V): P
E VOltage (IN): N

 

 

   

J;
8

 

 

Voltage (11V)
N t»
O o
O o

é

 

 

 

Position (mm)

41

Table 2-2 below summarizes the results for room temperature scanning of direct bonding

to stainless steel 316 to LAST material.

Table 2-2. Unicouple contact resistance values.

 

 

Unicouple 0' 0' R Ron ch
# (swing/03113) R" (m0) (uQ-cmz) (uQ-cmz) cm“? R‘ (m0)
1 500 625 6.3 5.1 130 80 N0 22
2 714 417 4.4 7.6 < 20 90 N 19
3 733 947 4.3 3.3 120 200 P 11
4 424 857 7.5 3.7 200 300 N0 23
5 2000 700 1.6 4.5 300 80 P 8
6 1920 890 1.7 3.6 100 < 20 Both 6.6

 

It can be seen that there is no clear pattern for n-type or p—type contact resistance, either
one can be high in a given unicouple making it a high resistance module. Two out of six
modules have less than 10 (m9) total resistance. Four out of six modules have cracks
after bonding as shown by a sharp increase in the voltage scan. These cracks may have
occurred during cutting of the legs, or during the bonding process. However in high
temperature contact resistance scans, direct stainless steel bonded modules have shown

higher resistance at the junction as shown in Figure 2-13.

42

 

400 ..

 

 

 

 

 
 
 

 

 

 

3 0 E ETP 45-33 315 ,.
5 ;' Room Temp Scan .3 E
300 :— ,.° .1
I O .1
9 250 :- .’ d:
5 0' :
o 200 '- ' .:
g t 3' ETP 45 3
-5 150 - ,- .3
> . bonding .0 :
100 :- junction.’ 2 i
i ' Rc < 25 cm 1
50 E ' p9 -:
0 E stainless steel 316 3
0 2 4 6 8 10
Position (mm)
500 --.
; ETP 45-88316 / 3
400 -_ High Temp scan ~7OOK _-
:00. x’ ~:
0 . e" ETP 45 -
3 I ..0' .
L5 200 - bonding .- j
> junction’,’
100 _' stainless 2 '2
3 steel 316 Re ~ 62 pncm ;
0 - n11].-1llama-Innnnlnnnnl-nL-l111-:
-1 0 1 5 7

Position (mm)

Figure 2-13. Room temperature and 700 K scan indicating increased contact resistance.

43

2.6 Soldered vs. silver paste contacts

As mentioned in previous section room temperature scanning of thermoelectric
legs provides important information about contact resistance and material integrity.
Contacts for these measurements were made to the samples using a soldering iron. The
rapid heating of the sample in the vicinity of the soldering iron was a suspected cause of
cracking. To test this hypothesis six legs were first solder contacted and measured. These
were compared with six legs from same batch (ETP 38) which were contacted using
silver paste. Since silver paste is applied at room temperature no thermal shock to the
sample occurs. Figure 2-14 below shows scan from six legs using soldering iron at ~ 477

K. Four out of six samples showed cracks.

 

 

 

 

 

 

 

250 I I I I I I I
Sample 1 P-Type cast ingot ETP 38
a sample 2 . . .
200 - _ Sample 3 d
I SamPle 5
+ Sample 6
$150 - d
a + + + (t rh
0 at + + + i
3100- --+’+*fl*# QED??? -
c 3+4 3 - r: a L?
> *1“ r - ,5.
50 - 5 4 +. "Fig”? . 1 ‘
+¥+::¥3l3§\ '
* : 1249i
0 11 0 5 d E U -- 4 Out of 6 samples showed cracks
- - (solder measurement)

 

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

Figure 2-14. Soldered contact thermoelectric (ETP 38) legs.

Figure 2-15 shows six samples contacted using silver paste. It can be seen that only one

out six samples showed cracks as compared to four out of six for soldered contact

 

 

 

 

 

 

 

 

samples.
200 I l I ' I I I
. Sample 1 P-Type cast ingot ETP 38
I Sample 2
0 Sample 3
150 Sample 4 11 " . .
Sample 5 11 . 1 ‘ 3
a O Sample 6 ....l: . : : : 0 3 3
:100 b .l ““f:' ,1 -
w I - “ ,3 3 .1 , i) a V.
g n u' ‘2‘ ’33 4
§ ' 4 1 ‘33? «>5 ‘3
50 r '5' ‘ ‘5'3 it” '
u 3 3": wwwfi
I . l O 3:51‘9553}
g g g g I <> if 1 out of 6 samples showed cracks
0 - (silver-paste measurement) '
0 1 2 3 4 5 6 7

Position (mm)

Figure 2-15. Silver paste contact measurements for six samples.

Although silver paste shows better results for room temperature scanning
contacts, soldered contact are preferred for unicouple cold side junctions because of its
superior mechanical strength and low contact resistance as compared to silver paste. To
circumvent thermal shock due to the soldering iron we investigated the use of a hot plate
and a slower heating rate. The cold side electrode (Cu or Ni) was then first coated with
solder and placed on top of hot plate. The TE leg was then placed on top of the soldered

electrode and the hot plate was switched on. The hot plate was heated slowly by

45

gradually increasing the set point. Using this process a hot pressed sample was bonded to

Ni and Cu electrodes as shown in Figure 2-16 and Figure 2-17 respectively.

 

 

 

 

 

500 . I I . . I
MSU Hot Pressed-l (N-Type) “.0009”...
LAST-Ni cold side solder “,0” A Ni
400 - v, '1 r 1 _
‘ 2
E. 300 ' Rc < 20 “II-cm _
3; a = 28 S/cm
31"
L; 200 - _
>
100 - ,3 Current: 1 mA -
,.°° size~2x2x7mm
0 - ° . _
0 l 2 3 4 5 6 7

Position (mm)

Figure 2-16. Hot pressed sample bonded to Ni electrode using solder.

46

 

350 . .

  

MSU Hot Pressed-l (N-Type) WWW
300 _ LAST-Cu cold side solder .33] Cu _
”:3 03.... j
250 - .9 .03 I _
0’.
E5200 - .3... ch < 20 [in-cm2 -
0 .9
w o = 32 S/cm
g 150 " .9.
g .033
100 r -
93.3.
50 - ’03. ..
.0" Current: 1 mA
03 .
0 ' 3' srze~2x2x7mnr

 

 

 

0 1 2 3 4 5 6 7 8
Position (mm)

Figure 2-17. Hot pressed sample bonded to Cu electrode using solder.

Three samples from ETP 38 cast p-type were solder contacted for room temperature

scanning using hotplate and showed no cracks as shown in Figure 2-18 below.

47

 

300 1 1

 

 

 

ETP 38 hot plate solder
250 - 1113511151135” _
3111113?-
200 "' IM -
{filmy ”o
9 5:» II/
a: 150 p I ’00 o u
g ”:3 ‘9’
an . we?“ (09’
E 100 L - I ...«r' I
> I»)
50 P T: ...Q 0 u
0 " 95:99.... 1‘ " -I
-50 I l l J 1 I I
-1 1 2 4 6 7
0 POSitign (mm)

Figure 2-18. ETP 53 samples solder contacted on hot plate.

2.7 Diffusion barriers in thermoelectric - previous efforts

As mentioned above it is desirable to have low contact resistance between the
thermoelement and electrode. The hot side contact should be able to withstand thermal
cycling thus electrodes with thermal expansion near that of the thermoelectric materials
are desired. It is well known that element like Cu, Ni and Cr can degrade properties of
bulk TE materials, especially telluride based thermoelements [24]. Kendall et al studied
PbTe diffusion bonding to AIS 300 stainless steel [25]. They found the mechanical
failure due to thermal expansion mismatch can be eliminated however they also
discovered elements like Cr in stainless steel can diffuse into the material and deteriorate

the thermoelectric properties. A thin diffusion barrier of Fe, M0 or W can help to prevent

48

this diffusion. Weinstein et a1 [26] used braze tin telluride between PbTe and Fe
electrode. A good bond was obtained by diffusion and subsequent solidification of the tin
telluride braze. It behaved well up to elevated temperatures of 600 °C. However, the
braze joint failed after thermal cycling in which joint was first raised to 600 °C and then
brought back to 100 °C [26]. The failure occurred at the tin-telluride-iron interface and
was due to the formation of a brittle layer. The n-type PbTe was easily bonded to Fe ingot
by diffusion bonding at 858 °C, which is 15 °C lower than the eutectic temperature of Fe
and PbTe. Kendall et al [25] found that by adding a small amount of antimony the brittle
layer at tin-tellurium-iron interface would not form and the joint could sustain thermal
cycling. It was also found that if ferrous metal has trace amount of carbon, manganese
and silicon as is the case with commercial ingots, it would generate cracks in PbTe bulk

material even if the braze was good with no brittle layer. These cracks were mainly due

to CTE mismatch of PbTe (21 x 10-6in/in/0C) and iron ingot (14 x 10'6in/in/°C). When

Fe ingot was replaced by stainless steel, the problem of CTE mismatch was removed.
However, it was revealed that doping agents in PbTe were eroded in the braze region
when stainless steel electrode was used. The reason could be the chromium in stainless
steel diffusing through molten tin-telluride and passing into bulk PbTe and then reacting
with free tellurium in p-type material. This would result in increased electrical resistance
[25]. Nickel also dissolves in tin telluride and reacts with PbTe. This forms a eutectic
mixture and lowers the working temperature of module. The rate of diffirsion is higher at
the hot side and this diffusion also occurs even if braze material is not used and electrode
and bulk material is bonded through diffusion bonding process. As a solution to these

problems, thin 0.002 inch thick foils of iron, molybdenum and tungsten were deposited

49

on stainless steel electrode through diffusion bonding. Unicouple were fabricated and
tested for 750 hrs and 36 thermal cycles with no degradation observed [25]. It was found
that presence or absence of tin telluride braze layer had no effect on degradation of
performance of thermoelectric device. Other alloys having similar range of CTE like
constantan, brass, silicon bronze, aluminum bronze or beryllium copper can also be used
as hot side electrodes. Although this literature shows satisfactory results but in practice
PbTe thermoelectric module have been assembled with spring loaded configuration
especially for space applications. Probably diffusion bond with PbTe has not been
successful in long term operation.

As described above the purpose of the diffusion barrier is to prevent diffusion of
contaminations from the hot side electrode into the thermoelement and at the same time
prevent reaction between electrode and thermoelement. The need is to provide a barrier
layer which is thin enough not to interfere as CTE mismatch between thermoelement and
electrode (which are generally both chosen to have similar CTE) and at the same time
thick enough to prevent diffusion of contaminants. Refractory materials are suitable as
barrier layers however; these materials have very low coefficient of thermal expansion
e.g. [W (4.5), Mo (4.8), Ta (6.3), Ni (7.3)] um/m-K at 25 °C. Therefore the barrier should
be thin enough that forces arising from barrier thermal properties are negligible as
compared to the thermal forces exerted by larger masses of electrode and thermoelement.
Krebs [27] found that even if the barrier material has larger tensile strength, but it is thin
enough, it will not have deleterious effect on thermoelement and a refractory element (W,

Mo, Ta) diffusion barrier of about 0.05 to 0.5 mils thickness can be used with electrode

of 10 to 40 mils thickness having CTE of 18 x 106/°C. An alternate way to address this

50

problem is to use an intermediate layer of a low resistance semiconductor which will
allow diffusion of contaminants into the intermediate layer but will not be able to cross
into the thermoelement. Contaminants will form a solid solution at the intermediate layer
until the barrier layer saturates. Such barrier layer has life of about five years [27]. This
was tested under accelerated condition. The barrier layer has to be thick enough to absorb
contaminants and thin enough that it does not start acting as separate thermoelement with
uncontrollable properties. Suitable thickness of such barriers is found to be about 10 to 20
mils [27]. Formation of such polycrystalline thermoelement barrier of same material as
main thermoelement can be obtained by sputtering or vapor deposition. Probable cause of
slow diffusion of contaminants into poly crystalline segment may be due to increased
number of grains and therefore impurities have to take a longer path along the grain
boundaries. Another way to improve the performance of TE device is to use both above
mentioned permeable and impermeable layers. Impermeable layers can be of tungsten,
tungsten carbide, molybdenum and silicon carbide [27]. The permeable barrier can
consist of materials like bismuth-telluride, lead-telluride, (lead telluride-germanium
telluride), lead-tin-telluride and alloys of (lead telluride-lead selenide) [27]. However,
microcrystalline structure of permeable layer needs to be different than the bulk
thermoelement. Kerbes used Ni as the bonding material between lead-telluride and
stainless steel. A 0.05 to 0.15 mils thick layer of Ni was deposited on a stainless steel
electrode and annealed in a reducing atmosphere for about 10 minutes at about 800 °C.
This allowed the diffusion of Ni into stainless steel. Ni was then coated with tungsten

layer of 0.1 to 0.5 mils thick and then another layer of nickel was deposited. Bonding was

51

carried out in an inert atmosphere. This formed a eutectic alloy of nickel-lead-telluride. A

reasonable thickness of such an alloy is 0.1 to 0.10 mils.

2.7.1 Brass diffusion barrier coatings

The following diffusion barrier and bonding experiments were carried out in
collaboration with Dr. Nuraddin Matchanov in this lab. Some materials like Mn/Pd
despite having high CTE can not be used because it undergoes transformation at higher
temperature (300- 350 °C) that includes volume change and high electrical resistivity
[28]. Since LAST material is very close in stoichiometry to PbTe, diffirsion barriers

mentioned in literature can also be tested. Yellow brass with 33% zinc has CTE of ~18.9

-6 . . . . . .
x 10 rn/rn/C WhICh rs very close to LAST. Followrng coatings were tested on 2 mm x 5

mm x 12 mm brass pieces.

52

Table 2-3.

Brass coatings with diffusion bonding.

 

 

 

 

 

 

 

 

 

 

 

 

Electrode Thermoelectric Bonding T(°C) comments
coating material material
Ni plated brass LASTT Pb/Sn/Sb foil 650 High resistance
with Ti/Zr/Ni bond and Zn
coating diffusion was
observed around
sample
Ni plated brass LASTT Pb/Sn/Sb foil 700 No bond
Ni plated brass LAST Pb50/Sn50 700 Bonded but brass
solder reacted with LAST
Brass coated LAST Pb/Sn/Sn 700 Mechanically good
with Mo. bond but Zn
Substrate cold appeared on LAST.
when coated Mo fihnpeeled off.
Brass coated LAST Direct bonding 700 Zn appeared on
with TiN LAST
Brass coated LASTT Direct bonding 650 No bond
with Ti/T iN/T i
in PLD system
Brass coated LAST Direct bonding 650 Zn appeared on
with LAST
Ti/TiN/Ti/Sb
Ni plated brass LAST Direct bonding 700 No bond
coated with
TiN in PLD
system
Brass coated LAST and SnTe 720 No bond
with Mo LASTT
Ni coated brass Hot pressed SnTe 610 GoOd bond. Cu and
LAST Zn diffusion
observed

 

 

 

 

 

Figure 2-19 below shows a bond between p-type cast sample and 2mm x5mm

x12mm brass sample. As mentioned above brass was chosen as hot side electrode

because it has a close CTE (18.9 x 10.6 cm/cm/°C) match with LASTT.

53

 

LASTT(ETP 50)- brass
Sanb/Sb foil

ETP 50 o

,‘ Rc ~ 343 pncm2

1)

bonding
junction

 

molybdenum (Mo)
coated brass

bonding temp ~ 650 °C

1 2 3 4 5

 

6 7 8 9

Position (mm)

Figure 2-19. LASTT to brass bonding.

In this particular run Sn60/Pb39/Sb1 solder foil was used as intermediate bonding

material and the bonding was done at 650 °C for 2 hrs in a flowing Ar/I-12 gas under 800

Torr. It is a high resistance bond and has non linear slope at the junction as shown in the

plot. Similar coatings of Ti/TiN have not been able to stop Zn diffusion from brass to

LAST material. Following is the EDS result (Figure 2-20) for brass/Ti/TiN/Ti/LAST

bonding carried out at 720 °C for 4 hrs under Ar/Hz. EDS was canied out 4mm from the

junction towards the sample. It can be clearly seen that Zn has diffused through coating

on electrode into the thermoelement leg.

54

ZnK 31.56
c
T 1

 

Figure 2-20. EDS results for Zn diffusion from brass to LAST.

2.7.2 Diffusion bonding with coatings for Ni, Cu and steel

In addition to brass other electrodes with different coatings as listed below were

also tried for diffusion bonding:

55

Table 2-4.

Diffusion bonding with coatings for Ni, 88316 and Cu.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lBondingBondind
Thermoelectric Temp Time
Material Contact Material (°C) (Hours) Results
LAST(Hot
pressed) SnTe-thTe/Nickel 615 4 Successful Bond
LAST(Hot
pressed) SnTe/Nickel 615 4 Successful Bond
LAST(Hot Good bond, but Cu and Ag
pressed) PB616/SS316 615 4 diffusion was observed
LAST(Hot Bonded but high contact
pressed) Pb616/N i 610 4 resistance
Mechanically weak bond, Cu
LAST(Hot and Ag diffusion was
pressed) PB6l6/FeMnC 610 4 observed
LAST(Hot SnTe/SS316 layer Good bond, high contact
pressed) (2.85 mm)/Copper 610 4 resistance
Good bond, high contact
LAST(Hot SSB 16(347)(~ 0.5-1 resistance (oxide layer
pressed) mm)/N i coated brass 720 4 between SS and Ni)
SS316(347) piece(~ Good bond, high contact
LAST(Hot 0.5-1 mm)/N i coated resistance (oxide layer
pressed) Cu 720 4 between SS and Ni)
FeMnC piece(~ 0.5-
LAST(Hot 1 mm)/Ni coated
pressed) brass 720 4 No bond
LAST(Hot FeMnC piece(~ 0.5-
pressed) 1 mm)/Ni coated Cu 720 2 No Bond
LAST and Good bond, high contact
LASTT Si/Mn/Si/FeMnCi 750 4 resistance
LAST and Good bond, high contact
LASTT Mn/FeMnC 720 6 resistance
LAST and Good bond, high contact
LASTT Mn/Si/Mn/Si/SS3 16 750 4 resistance
LAST and
LASTT Mn deposited/S8316 720 6 N—legnot bonded
LAST and
LASTT Mn deposited/N i 720 6 N leg melted
LAST and
LASTT In solder/FeMnC 810 4 Mechanically weak bond
No Bond (single Stage
LAST and unicouple run)
LASTT In solder/88316 810 5

 

56

 

Table 2-4. Continued.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

lBondindBondind
Thermoelectric Temp Time
Material Contact Material (°C) (Hours) Results
LAST and No Bond (single stage
LASTT In solder/Mn/SS316 700 5 unicouple run)
LAST and
LASTT FeMnC 800 1 No bond
LAST and SnTe
LASTT powder/FeMnC 800 l Mechanically good bond
LAST and No bond (single Stage
LASTT SnTe powder/S8316 800 l unicouple run)
LAST and Mo(~0.9 mm)/Mo
LASTT (~0.9 mm)/Cu 720 2 Samples melted
LAST and Mo/SnTe low contact resistance bond,
LASTT powder/Mo/Cu 650 16 Cu diffusion was observed
Mo(~1.4 mm)/PbSn Bonds well and low contact
LAST and solder/Mo (~l .4 resistance, Cu diffusion was
LASTT mm)/Cu 650 16 observed
Bonds well and low contact
LAST and resistance, Cu diffusion was
LASTT Mo/In solder/Mo/Cu 650 16 observed
LAST and Mo(~1.4mm)/Mo Bonds well but high contact
LASTT (~l .4 mm)/Cu 680 2 resistance
Bonds well, but high contact
LAST and Mn(~200nm)/In/3 l6 resistance
LASTT Stainless Steel 750 4
LAST and Mo(~900nm)/In/316 Bonds well, but high contact
LASTT Stainless Steel 760 4 resistance
Good bond, high contact
LAST and Mo/Sb/3l6 Stainless resistance (single stage
LASTT Steel 770 4 unicouple run)
Good bond, high contact
LAST and Mn/Sb/3l6 Stainless resistance (single stage
LASTT Steel 810 4 unicouple run)

 

57

 

Voltage (pV)
on
O

300

250

Voltage (IN)
3 a B
O O O

U!
C

 

 

Ni substrate- LAST
50I50 PblSn solder
bonding temp~ 600 C

 
 
 
 
  

 

2x5x12mmNi

/ substrate

 

 

—£ I I 1 l 1 1

O 1 2 3 4 5 6
Position (mm)

Figure 2-21. LAST bonded to Ni electrode.

 

Ni substrate- LASTT
- 50I50 PblSn solder
bonding temp ~ 600 C

/ Rc~ 70 uncm2

 

 

 

-2 0 2 4 6 8

Position (mm)

Figure 2—22. LASTT bonded to Ni electrode.

58

10

Figure2-21 and Figure 2-22 show voltage scans of bonded 2mm x 5mm x 12 mm Ni
electrodes as the hot side contact. These were done using a low temperature bond of
50/50 Pb/Sn solder. Bothe n-type and p-type legs were bonded in the same run. Nickel is

not a suitable electrode material for LAST as it reacts at high temperature (~ 700 C) and

6

CTE mismatch (Ni ~ 13 x10' /K at 20 C and LAST ~ 21x10'6/K) may lead to thermal

stresses inducing cracks during operation.

2.8 Diffusion bonding with Sn and Te powder

The most promising bonding experiments in terms of low contact resistance were
canied out using Sn + Te powder (Sn70%Te30%) cold pressed coins (2000 psi) in
oxygen free environment of glove box. Diffusion bonding was carried out in Ar+5%H
environment at 820 Torr with ramping time of 4 hr, soaking time of 0.5 hr , heating rate

of 2.7 °C/min and cooling time of 0.83 °C/min. Samples were pressed under spring

pressure of 6.3 Kg/cm2 during bonding. Electrode coatings and 4-probe unicouple

resistance are given below.

59

 

Table 2-5.

SnTe based diffusion bonding.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Electrode p-type leg Bonding 4-probe TE legs Contact
T(0C) Unicouple resistance resistance
Resistance (m9) (m9)
(m0)
Ni ETN 153 615 3.5 2.78 0.71
ETP 55
Ni ETN 153 615 3.89 2.78 1.10
ETP 55
Ni ETN 153 615 4 2.78 1.21
ETP 55
Ni coated ETN 153 615 4.3 2.78 1.52
Cu ETP 55
SS 316 ETN 153 615 4.3 2.78 1.22
ETP 55
SS316 ETN 159 650 5 3.46 1.24
ETP55
SS 316 ETN159 650 5.14 3.46 1.38
ETP52
Cu coated ETN 159 650 5.5 3.46 1.94
with SS316 ETP55
Cu coated ETN 157 650 5.6 4.26 1.24
with SS316 ETP55
iron ETN 153 615 5.82 2.78 3
ETP55
SS 316 ETN 157 650 5.9 4.26 1.3
ETP52
SS316 ETN 157 650 6 4.26 1.4
ETP55
Ni ETN 157 650 6.49 4.26 2.22
ETP52
Ni coated HPMSU 46, 615 35.7 29.16 6.0
brass ETP52

 

 

 

 

 

 

60

 

For Table 2-5 resistance values of thermoelectric legs are given in Table 2-6 below. Leg

size is 5mm x 5mm x 7mm.

Table 2-6. TE legs resistance values.

 

 

 

 

 

 

 

 

TE material a (S/cm) Resistance (m9)
ETP 55 2400 1.16

ETP 52 2400 1.16

ETN 153 1726 1.62

ETN 157 900 3.1

ETN 159 1200 2.3

HP 94 100 28

 

 

 

 

For Table 2-5 resistance values for hot side electrodes are given in Table 2-7 below:

Table 2-7. Hot side electrode resistance values.

 

 

 

 

 

 

 

 

 

Electrode resistance (p.11)
Ni 80

Cu 20

Ni plated Cu 50

SS 316 350

SS 316 plated Cu 100

Iron (Fe) 50

Brass . 60

 

 

 

LAST and LASTT based unicouples bonded using compacted Sn+Te powder
discs were checked with room temperature scans for contact resistance and SEM images

for junction analysis. As it can be seen from following SEM image (Figure 2-23) that

61

there are no cracks at the junction for austenitic steel (carpenter steel 19-2) with LASTT

(ETP 52) and no intermediate layer was observed in the SEM images after bonding.

 

Figure 2-23. Carpenter steel 19-2 bonded to ETP 52 using SnTe.

The room temperature scan in Figure 2-24 shows that it is a low resistance contact at the

junction of carpenter steel and ETP 52 as there is no abrupt voltage jump. We also do not

observe any crack in the thermoelectric material.

62

cs 192-shre-ETP 52

 

 

 

 

0 " I I Juneau: .
C
'20 ' 0' austenitic steel -
A .3 (19-2)electrode
- Q _I
>3 4O ETP 52 .3
also , ' _
3 e
g -80 - ... _
.
'100 ' .. _
O
.0
—120 - ..
....
-140 ' l I I
0 4 8 10

6
Position (mm)

Figure 2-24. Room temperature scan for carpenter steel bonded to ETP52.

SEM image in Figure 2-25 shows a crack free interface with no intermediate layer

between cast LAST and austenitic steel using Sn+Te compacted coin.

 

Figure 2-25. SEM image for austenitic steel and HPMSU 41.

63

The room temperature scan for carpenter steel bonded to hot pressed sample in Figure 2-
26 shows a low resistance bond at the junction but the slope is non linear which might be

due to diffusion of Sn into the hot pressed material.

carpenter steel -SnTe-HPMsu41

 

5000 I I I l I
0 - ,.......... -
.. austenitic
-5000 - . -
... steel
4 e .
E '110 ' HPMSU 41 ,e' = "
217-1.5 104 - .‘ _
8’ .'
g -2104 - . -
> 3
-2.5104 - . -
-310“ - '3 -
O
-3.5104 ——F - - - -

 

 

 

-2 0 2 4 6 8 10
Position (mm)

Figure 2-26. Carpenter steel bonded to HPMSU 41.

LAST and LASTT based unicouples bonded using compacted Sn+Te powder discs were
tested by Nuraddin Matchanov for short circuit current in module testing system and

consistently produced > 8 A currents which is reported elsewhere [29].

2.9 Electrical contact resistance measurement
In a typical thermoelectric module we want hot side contacts to be ohmic in

nature or a schottky barrier height of zero. For ohmic contacts resistance is independent

64

of voltage. This implies that work function of metal should be close to or larger than the
electron affinity plus bandgap of p-type semiconductor and the metal work function
should be less than or equal to the electron affinity of an n-type semiconductor.

For our measurement purposes, we use curve fitting for the slope of the voltage vs
position plot. In a typical experimental setup, current is sourced through the
thermoelectric sample and voltage is measured using Keithley nanovoltrneters. Current is
flipped approximately 20 times for each measurement to avoid Peltier effects. The
following formula is then used to measure the contact resistance.

V—V
12:21
1

C

x Area

Here V2 and V] are the voltages at the junction of metal—semiconductor.

 

 

V

 

Position

Figure 2-27. Calculation of contact resistance from voltage vs position plot.

The Figure 2-27 above shows how to take values of V1 and V2 from the voltage vs.

position plot. The dark line is the actual voltage slope and dotted line is the slope if there

had been no contact resistance. The area is the cross sectional area of thermoelectric leg

65

in contact with the metal electrode. Care has to be taken as to which point to select as the

end of electrode and start of slope because that will cause variation in the V2 —V1.

A special case would be if the slope is not linear at the end rather shows a curve

as indicated in Figure 2-28 below. In this case the part of the slope which is linear should

be considered for measuring voltage difference because if V2-V1 is directly measured

from two nearest data points separating electrode and thermoelement (in Figure 2-28 this
value is shown to be 135 uV), that will result in erroneous value of contact resistance.
However, if the slope is curve fitted, the voltage difference is 200 uV which corresponds

to more accurate COHtflCt resistance value.

 

   

 

 

 

500 , Y r r
400 - A
a 300 - q
0
g V2
3 200 - _
>
100 - voltage jump across _
contact ~ 135 uV
0 b .
-2 O 2 4 6 3

Position (mm)

Figure 2-28. Curve fit for contact resistance measurement.

66

500

 

 

 

 

 

 

 

 

.0
O
.0
400 - scan with 0.2 mm ... ..
step size .,'
O
A O
a. 300 " .0... -i
‘6' .
a: .-
3 O
'3 200 - -
>
O
100 ' electrode' '0" '
y
y
0 w...” I I I
-2 O 2 4 6 8
Position (mm)
Figure 2-29. Scan with 0.2 mm step size.
500 , , , .
I
400 _ scan with 0.1 mm _
step size
>5 300 - r
0 O
3’ .4'
'5 200 - -
> O
i O
100 - l ' .
electrode! /
0 I . I I I
-2 O 2 4 6 8

Position (mm)

Figure 2-30. Scan with 0.1 mm step size.

67

Another aspect to consider is the step size of scanning pin. That directly corresponds to
sensitivity of measurement: smaller the step size, more accurate will be the value of
measured resistance. Figure 2-29 shows a metal electrode-thermoelement scan with step
size of 0.2 mm and Figure 2-30 shows same sample with 0.1 mm step size.

We can see that in both cases curve fit gives negligible contact resistance. However,
when two plots are superimposed as shown in Figure 2-31, the step size makes a
difference in measurement for high resistance area (shown by the circle). The 0.1 mm

scan gives a voltage difference of 58 IN and 0.2 mm scan shows ~ 80 pV difference.

500 | I

 

I l . i

 

0 0.1 mm step size
400 _ D 0.2 mm step size

 

 

3300 - -
0

5’ I

3 200 - -
>

100- / - -

0 g I I
-2 0 2 4 6 8
Position (mm)

 

 

_

 

Figure 2-31. Superimposed scans of 0.1 and 0.2 mm step size.

The sensitivity of the system is also dictated by the radius of scanning pin. For example
in our high temperature voltage scanning system we can use scanning pogo pin with a tip

radius of 0.53 mm (manufactured by Otsby Pylon), that means any voltage gap across the

68

contact less than 0.53 m can not be detected. However, system also has the provision to

use 0.08 mm tip radius pogo pin to improve sensitivity.

Conclusions

A number of bonding and diffusion barrier coating experiments were carried out.
Stainless steel (SS316) has been promising but due to slight mismatch in CTE, causes
microcracks in LAST sample after bonding. Another emerging option is LASTT as hot
side electrode but that will need heavy doping to make it more thermally and electrically
conductive. Brass has a close CTE match but the zinc tends to diffuses into the LAST
material to form ZnTe, even when diffusion barriers of Ni, Mo, Ti and TinN were used.
Low contact resistance in the range of 4 to 7 (m9) have been obtained using cold pressed
SnTe coins based diffusion bonding and have proven to be a better Option for LAST

diffusion bonded unicouples.

69

Chapter 3: InGa and Spring Loaded Modules

Introduction

Due to coefficient of thermal expansion (CTE) mismatch problems it was decided
to use InGa as liquid contact between the LAST material and metal electrodes. Dumke et
al found that the surface of InGa eutectic is enriched with indium and they
experimentally evaluated that with bulk Ga concentration of 83.5 % the surface
concentration of Ga was less than 6% [30]. Gallium is highly reactive metal and dissolves
most metals at high temperature however Ni and Au moderately react at high temperature
with gallium [31] and therefore have been tried in our experiments. We used 75.5

Ga/24.5 In alloy by Goodfellow with melting point of 15.7 0C. This eutectic has high

. . . 4
electrical conductrvrty (3.4 x 10 S/cm) [32], wets most surfaces, makes a low contact

resistance interface with variety of materials [33] and is known to flow after a surface
stress of 0.5N/m is applied [34]. Molten InGa surfaces in air primarily consist of oxides
of Ga [35] which is an n-type semiconductor. In ambient air it has an adsorbed film of
water as it has high surface energy (630 dynes/cm) [32]. The room temperature electrical

properties of InGa are stable in time afier the initial surface oxide is formed.
3.1 InGa Module Fabrication

Module legs consisted of n-type AngmeTem+2 (LAST) and p-type Ag(Pb1-

ySny)meTe2+m (LASTT) which were cut to 5 mmx5mmx7 mm and polished to remove

oxides. Two 2mm x 5mm x12mm Cu pieces were used as cold side contacts and were

mounted on an AlN piece using double sided carbon tape. Care was taken to prevent

7O

short circuit at cold side by removing the double sided tape between Cu pieces. Low
temperature Ceromatrix solder was used to mount both TE legs on the Cu pieces. InGa
liquid was then applied on the top surface of each leg. Different metal electrodes were
investigated. First a 2mm x5mm x12 mm copper piece was nickel plated. It was lightly
polished with 1200 grit sand paper to produce a shiny surface. Residues left on surface
from the Ni plating solution were found to cause poor wetting of InGa on the electrode
surface. After acetone cleaning InGa is applied using a wooden applicator and the
electrode is placed on top of already mounted TE legs. A thick binding solution of QS/B4
by F ortafix is prepared and applied around the legs such that it also engulfs the electrode
sides thus holding it in place and providing necessary mechanical support. Two types of
measurements were canied out, one in ambient air and other under Ar atmosphere inside
the Long Term Module Testing System (LTMTS) [36]. For ambient air short circuit
current measurements a thick short wire was used to short two legs of the unicouple.
Initial measurements were carried out inside the LTMTS. A 2mm x5mm x12 mm Ni
electrode was used. Nickel has lower electrical and thermal conductivity than Cu but
previous experience showed that LAST readily reacts with Cu at high temperature and
that Cu has tendency to oxidize. Two unicouples consisting of LASTT (ETP 53) and
LAST (ETN 224) were prepared. InGa was used as liquid contact between Ni electrodes
and TE legs. QS/B4 was used to hold the electrodes in place on top of legs. F our-probe

resistance for the unicouple module was 14 (m0). Temperature was then ramped up to

600 °C and the maximum short circuit current was measured as 6.4 (A) for Th= 600 °C

and To: 30 °C. Output short circuit current then started decreasing indicating that InGa

71

reacted with LAST material which could be seen upon sample removal. Figure 3-1 shows
the power output for this module. After thermal cycling the 4—probe resistance was 45
(m9). The measuring wires for LTMTS system were about one meter long and had 14
(m0) resistance whereas matched load resistance per unicouple is typically < 10 (m0).
The measurement configuration was then changed by measuring the sample in air and

using a short thick Cu wire which greatly increased the measured short circuit current.

InGa liquid contact
0.7 ~——~~ -_ r i l

 

'I

dT =59? °C

4 leg module

5x5x7 mm legs

Ni electrode
quoated with 0834

 

__J____ .-

__ l_fifi_‘n_z __

 

 

2 3
Resistance ((2)

Figure 3-1. Ni electrode based InGa liquid contact.
Different geometrical configurations for the electrodes were investigated

primarily to increase the contact surface area. One such configuration is shown in Figure

3-2, where 12 mm long TE legs were used such that 5mm was inside the electrode and 7

72

mm outside. The 5mm leg portion inside the electrode would give five times the contact

surface area compared to that of a regular 7mm leg surface.

 

 

 

 

 

 

 

 

 

 

Figure 3-2. Stainless Steel 316 electrode for 12mm TE legs.

A single unicouple module was assembled with 5mm x5mm x12mm legs of
LASTT (ETP 53) and LAST (ETN 221) and InGa was used as liquid contact. A recess
was machined in the electrode to enclose a part (5mm) of 12 mm leg with InGa as shown
in Figure 3-2. The module was run inside the LTMTS and the maximum current reached
was 4.0 (A) for AT of 560 °C. Figure 3-3 shows the power output for this module. After
thermal cycling module resistance was 45(mfl). The output current was steady for a
longer duration than with Ni electrode however; current started decreasing after one hr
due to InGa reaction with stainless steel. This is because TE legs with InGa permanently
bonded with stainless steel electrode and without InGa the TB legs remained separate

from SS 316 electrode up to 600 °C.

73

InGa liquid contact

1 l

0.15

 

AT =560 °c

SS 316 electrode
single unicouple

0.1 5x5x12 mm legs

 

 

Power (W)

 

0.05

W‘,
1”
’2.
1';

 

 

O
p
L

l 1

L0, 0.5 1 1.5 "‘2
Resistance ((2)

Figure 3-3. Output for 12mm long TE legs.

A similar module was prepared with 12mm long TE legs and InGa liquid contact.
It was measured in open air with a smaller measurement wire (R < lmfl). Starting 4-
probe resistance was R ~ 7.5 (m0). The short circuit current plot is shown in Figure 3-4.
The maximum current reached was 6.47 (A) at 684 °C and current stayed at this level for

about four hrs which is longer than with Cu or Ni electrode. Afier thermal cycling

module 4-probe resistance increased to R ~ 30 (m0).

74

SS 316-InGa-ETN 221, ETP 53 (12 mm legs)
7 7% __ ’— T T

max current~ 6. 47 A
at 684°C

 

    
  
    

short circuit current (A)
N w -b 01 O)

A
T

 

 

O __ 1 . _l_ 1 1 LL .-
0 1 00 200 300 400 500 600 700
Temperature (hot side) °C

Figure 34. Stainless steel 316 electrode with 12 mm TE legs.

For metal electrodes copper has the second highest electrical (59.6 x 108 S/cm)

and thermal (400 W/m-K) conductivity after silver. Copper also has a relatively large
coefficient of thermal expansion; 18 (ppm/°C), which is also of interest since LAST and
LASTT have coefficients of thermal expansion greater than 20 (ppm/°C). Previous
experiments proved that copper readily reacts with LAST material and oxidizes thus
increasing the contact resistance. Nickel plating of the copper electrodes was used and it
was found to significantly improve the stability of the contacts; however InGa moderately
reacts with Ni. InGa was used as liquid contact and QS/B4 was used to minimize
sublimation and to mechanically hold the Cu electrode in place. Four probe resistance for
the unicouple module was 5.5 (mfl). Short circuit current was measured with module

inside the LTMTS under argon and the maximum value of the current wa35.23 (A).

75

Current was measured by first shorting the current wires in series with the module and
using hand held current clamp meter (EXTECH 380942).The output current plot is

shown in Figure 3-5 below [indicated by (x) line].

ET P 55, ETN 224- lnGa- Ni plated Cu

 

 

T 7 i

 
 
  
  
   

l
8 » ]
max current 8.45 A l
S. 5 '
t“ l
E 6 [ measured in 1
a open air i
3 I
'5 4 4
e i .
.5 5
s 2 ' .
'5 1‘ measured inside

LTMTS under Ar }

0 ' - 1 __ . i _ 1 1 . 1'
0 1 00 200 300 400 500 600 700 800
Temperature (hot side) °C

 

 

Figure 3-5. Module first measured under Ar, then three cycles with different

electrodes measured in open air.

The current started decreasing as InGa reacted with LAST material. Module was
cooled down, taken out and the QS/B4 was removed by using a screw driver to replace
with new electrode. After removal of the electrode, the surfaces of the thermoelectric legs
were polished, cleaned and new InGa was applied. Second time module was tested using
a short thick Cu wire with a resistance ~ 800p!) in open air. Starting 4-probe resistance

for module was ~ 7.5 (m0). Current reached 8.13 (A) for hot side temperature of 700 °C

76

 

[indicated by (0) line in Figure 3-5]. However, cold side temperature also increased to
120 °C at this point which effectively created a AT of 580 °C. To see if the results could
be repeated the module was cooled down, QS/B4 removed, and a new electrode was put
in place with new QS/B4 coating. Legs were polished for clean surface. This time short
circuit current reached only 6.0 (A) at hot side temperature of 550 0C [indicated by (El)
line in Figure 3-5]. The module was cooled down and the QS/B4 was removed for
inspection. It was revealed that electrode was not properly put in place and only a fraction
of total electrode area was covering TE leg surfaces thus not providing full contact. New
electrode was replaced, legs were polished clean, and new QS/B4 coating was applied
with InGa as liquid contact. Module was tested third time in open air and current reached
8.45 (A) at 688 °C [indicated by (0) line in Figure 3-5]. This module proved that LAST is
a robust material capable of surviving rough handling, multiple thermal cycling and still

can maintain its TE properties.

3.1.1 Coatings on TE legs

Several experiments were done to find suitable diffusion barriers to prevent an
InGa reaction with the LAST material at high temperatures. P-type LASTT (ETP 55) and
n-type LAST (ETN 224) materials were used to make legs of 5mm x5mm x7 mm size
which were then polished and sputtered with Au (100 nm), Ni plated to few micron
thickness and again Au sputtered (100 nm). Sputtering was carried out at room
temperature. The reason for first Au coat was that Ni plating did not hold on to LAST
surface well. However this pre-coating did not serve the purpose and during InGa

application of module fabrication it was observed that Ni film was still peeling off along

77

with Au fihn thus exposing LAST material surface. Probably due to room temperature

sputtering Au films did not bond to LAST surface. Short circuit current (156) reached to
9.00 (A) at hot side temperature of 662 °C as shown in Figure 3-6 below. For this module
we first time observed 8.62 (A) for hot side temperature of 600 °C which is the proposed

operating temperature for LAST based modules. Coatings did not help and current started

decreasing at high temperature.

ETP 55, ETN 224-InGa-Ni plated Cu

 

 

 

 

10 [ eeeee r T 1 r
2 8 :_ max current 9 00 A
*5 o ' 8.63 A .
g 6 at 662 C !
a i '7
0 1
.5
2 4 ~ ,
.8 l
‘5
5 2 L ,
0 . . - i L l L j
0 100 200 300 400 500 600 700

Temperature (hot side) °C

Figure 3-6. Au and Ni coated TE legs based unicouple output.
Next module was tested with a sputtered Mo coating on TE legs. Mo was

sputtered with substrate temperature ~ 150 °C and coating thickness was < one micron.

The coating did not stick well to the surface and could be brushed off with thumb

78

pressure only. Moreover, there was black soot like contamination on target holder inside
the PVD system that might have altered the Mo coating properties. Same module was
tested twice in open air to check the results. Each time current reached 7.5 (A) as shown
in Figure 3-7 below. The M0 coating was not able to prevent InGa reaction with LAST

material and the current started decreasing in both cases.

ETP 55, ETN 224 (Mo coating)- InGa-Ni plated Cu
8 f l * 1' T I l 1

 

\l

h current 7.5 A j
at 633 °C

short circuit current (A)
N 00 -h 0'! O)

—‘L
T

 

i
___l_q.n.___l__, 4%; a

_l l

0 J l l L
0 100 200 300 400 500 600 7
Temperature (hot side) °C

o l

0

Figure 3-7. Mo coated TE legs based module tested twice in open air.

3.1.2 SS-Ni-Au coating

In next experiment same legs were first coated with stainless steel using pen
plating system and stainless steel solution, then Ni plated and again Au sputtered (150
nm). The idea was that stainless steel might help to hold the Ni plating on the LAST

surface. A coating of QS/B4 was put around the TB legs and InGa was used as liquid

79

contact. The resulting module was tested in open air and a measured short circuit current
of 2.5 (A) at 341 °C hot side temperature was obtained. After the run the module was cut
open to remove QS/B4 with electrode and the thin film of stainless steel-Ni was still
intact but appeared dull, easily peeled off and perhaps did not allow direct contact of

InGa with LAST material.

3.1.3 Mo-An coating

In this experiment TE legs (ETP 55 and ETN 224) were coated on the ends with
M0 by sputtering at a substrate temperature of 150 oC and then Au sputtered (150 nm) at
room temperature. The module was QS/B4 coated and InGa was used as a liquid contact.
The Au coating was investigated here as a possible method for reducing the interface
resistance between LAST and Ni plated Cu. The short circuit current reached 7.12 (A) for
660 °C hot side temperature as shown in Figure 3-8 and then started decreasing right
away. It was the third experiment in which a Mo coating did not help prevent an InGa

reaction with LAST material.

80

ETP 55, ETN 224- InGa-Ni plated Cu

 

 

 

 

 

 

8 {7.7 l l f l .
7 [ max current 7.12 A * ’ ‘
$6 , at 660 °c -l
E r ,
€55? .
3
vI-I 4 ” 2}
.::3~ .
0
E 2 -
.C
a
1 z
0 1

o 100 200 300 400 500 600 700
Temperature (hot side) °C

Figure 3-8. Mo-Au sputtered TE legs module.

3.1.4 Mo-Ni coating

In this Mo coating experiment a module was tested inside the LTMTS under an
inert environment (Ar) of one atmosphere to see if the degradation rate could be slowed
(assuming oxidation at the interface as the cause). For this unicouple module LASTT
(ETP 5 5), LAST (ETN 224) TE legs were polished, Mo sputtered and Ni plated using
pen plating system. Before exposing the module to a large temperature gradient, the room
temperature four-probe resistance was measured to be ~ 4.5 (m9). As mentioned above
measuring wires of the LTMTS were cut short to reduce wire resistance to ~ 4 (m0). The
same module was then tested again in open air with short Cu wire (R < 1 m9). Module

output is shown below in Figure 3-9. For open air measurement starting 4-probe

81

resistance was R ~ 6.7 (m0). It can be seen that the open air measurement still performed

better due to low resistance measurement wires. This shows that open air oxidation of

interface is less

wires.

short circuit current (A)

0A

NOD-kalmflm

“VT 4 -Lr_

l
g l l L l i
0 1 00 200 300 400 500 600 700

deleterious for module out put than higher series resistance of measuring

 

ETP55, ETN 224 - InGa-Ni plated Cu

1 l l T T

max current ~ 7 A
at 576 °C /

# T—"—T _fi

   
   
 

measured in open I
— air 4

l

 

 
   
 

measured inside LTMTS -

_l

Temperature (hot side) °C

Figure 3-9. Mo-Ni plated TE legs measured under Ar and ambient air.

In previous experiments various coatings were investigated on the TB legs but

they were not successful in stopping the reaction between InGa and LAST rather most of

them increased resistance at the interface. An InGa layer between the TB legs and the

metal electrode without any coating on LAST surface was investigated. The same module

used in the previous experiment was cooled down, the QS/B4 removed, TE leg surfaces

polished to remove any contaminants from previous experiments, and prepared again

82

using a new Ni plated Cu electrode and coated with InGa. Module was taken high in

temperature and maximum short circuit current of 10.5 (A) was observed at 672 oC hot

side temperature. It was for the first time that any LAST based module crossed 9.0 (A)

within hot side temperature of 600 °C as shown in Figure 3-10 below. The short circuit

current then began decreasing, so the module was cooled down and the QS/B4 removed.

InGa could still be seen on the n-type leg, however the p-type leg permanently bonded to

the electrode. For the hot side temperature of 672 °C the cold side was 113 °C, thus the

effective AT was 559 °C and the maximum short circuit current obtained was 10.5 A.

_u _.L
m C N
l

short circuit current (A)
CD

N

0 ..
0

ETP 55, ETN 224- InGa-Ni plated Cu

 

A

 

1

max current 10.5 A
at 672 °C

l . l l T

 

 

 

 

100 200 300 400 500 600 700
Temperature (hot side) °C

Figure 3-10. Module prepared without any coating on TE legs.

83

Table 3-1.

Summary results for InGa based modules.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Electrode Material Air/Ar Th Tc 1 Comments
& coatings (°C) (°C) sc (A)
Ni ETP53 Ar 600 30 6.4 4 leg module
ETN 224
Stainless ETP 53 Ar 600 40 4.0 12 mm legs
Steel 316 ETN 221
Stainless ETP 53 Air 684 62 6.47 12 mm legs
Steel 316 ETN 221 (second run)
Ni plated ETP 55 Ar 660 74 5.23 7 mm legs
Cu ETN 224 first run
Ni plated ETP 55 Air 700 120 8.13 7 mm legs
Cu ETN 224 Second run
Ni plated ETP 55 Air 550 34 6.0 7 mm legs
Cu ETN 224 Third run
Ni plated ETP 55 Air 688 107 8.45 7 mm legs
Cu ETN 224 Fourth run
Ni plated Au-Ni-Au Air 662 103 9.0 7 mm legs
Cu ETP 55 coatings did not
ETN 224 work
Ni plated Mo on Air 633 83 7.57 7 mm legs
Cu ETP 55 2 runs
ETN 224
Ni plated SS-Ni-Au Air 341 28 2.5 7mm legs
Cu ETP 55 Dull colored SS
ETN 224 film
Ni plated Mo-Au Air 660 91 7.12 7mm legs
Cu ETP 55
ETN 224
Ni plated Mo-Ni Ar 624 61 5.90 7 mm legs
Cu ETP 55 First run
ETN 224
Ni plated Mo-Ni Air 576 57 7.00 7 mm legs
Cu ETP 55 second run
ETN 224
Ni plated ETP 55 Air 672 113 10.5 7 mm legs
Cu ETN 224

 

 

 

 

 

 

 

84

 

3.2 Spring Loaded Modules

The history of spring loaded modules is at least fifty years old [3 7]. Space
launched radio isotope generators (RTGs) used SNAP series thermoelectric converters.
These generators used PbTe and SiGe based modules and all PbTe based modules were
spring loaded [3]. Although we were not able to find much published data for spring

pressure used, Fritts [3 7] recommended 100 psi spring pressure.

3.2.1 Surface asperities and contact resistance

When two solid surfaces are pressed against each other, they make contact only at
certain spots because no matter how flat the surfaces are; there will be surface micro
roughness. Due to this roughness there will be micro peaks, valleys and geometrical
shapes which will come in contact. Thus the actual contact area is much smaller than the
nominal surface contact area. The asperities of the contact spots undergo elastic and/or
plastic deformation depending upon the surface hardness, applied pressure and
temperature [38]. As the actual contact between surfaces is limited to these spot,
electrical current lines become increasingly distorted and bundle together to pass through

these spots as shown in Figure 3-11 below.

85

 

 

 

Figure 3-11. Bulk electrical interface with current lines passing through

asperities (points of contacts between surfaces).

Due to these spots (called a-spots) actual volume of contact reduces and gives rise to
constriction resistance as current is constricted through these a-spots. The other factor
causing the increase in interface resistance is insulating film (usually oxide films) on the
material surface. The combination of these two resistances is called contact resistance.
For circular a-spots the equipotential surfaces in the contact members consist of

ellipsoids:

 

-——2—+ 2 2:1 (2.1)

Where ,u is the vertical axis of the vertical ellipsoidal surface and (r and z) are cylindrical

coordinates. The resistance between these equipotential surfaces is then:

,u
R=,0/27rIdy/(az+,112)=(p/27ra)tan'l (,u/a) (2.2)
o

86

Here [1 represents resistivity of conductor. Ifspreading resistance R, is defined as

constriction resistance for very large p then:

R, = p/4a (2.3)
The total constriction resistance is then:

RC = p/2a (2.4)
If two surfaces consist of dissimilar metals then

RC: (P1 + my 4a (25)

Spreading resistance for elliptical a-spots with semi axis a and b is given by:

00
Rs(a.b)=p/2n [cw/«a2 wszzwznm (2.6)
O

Spreading resistance for square a-spots with side length 2L is given by:
RS= 0.434p /L (2.7)

In practice, a-spots are considered circular for all practical purposes. In real contacts
surfaces mate where there are clusters of a-spots and their combined effects give the

contact resistance:

Rc =p(l/2na+l/2a) (2.8)

Here n is the number of a-spots, a is the mean radius of a-spots and a is radius of cluster.
When the surfaces are firmly pressed against each other, a-spots (asperities) go through
plastic deformation and these deformed asperities have slightly larger contact area. If one
of the two mating surface material is softer than the other then load F applied to the

contact and contact area Ac are related by:

87

F= AcH (2.9)
Here H is the hardness of softer material. This expression reveals the true area of
mechanical contact is independent of area of nominal contact for two surfaces. In other
words it is independent of dimensions of contacting bodies and only depends upon
applied force (load) and hardness of the materials. If insulating films on surface are not

present then contact resistance in terms of above expression can be given as:

R={ pznrrH/4F}1/ 7- (2.10)

Where 1] is an empirical coefficient and for a clean surface its value is unity. It is a simple
expression saying that with increased applied force contact resistance will decrease. This
can be explained as the force is increased, there is plastic deformation of a-spots and they
become more flatten and actual contact area is increased. As a-spots flatten the
constriction of electrical lines of force also cases and they have larger cross section
available to pass through between the two surfaces. Moreover, as the load is increased,
depending upon the surface hardness, more spots will come into contact thus increasing
the actual contact area.

When a-spot makes contact with opposite surface there is angle between the spot

and opposite surface depending upon the shape of asperity as shown in Figure 3-12 below
contact
x /<
Figure 3-12. A-spot making an angle with opposite surface due to conical asperity.
In this case the spreading resistance is given as:

88

R =(p/4a)tan{(7r+20)/4} (2.11)

3.2.2 Effect of Surface Film

Film coating on a conductor may change its properties. Ifthere is inter-diffusion
across the films of opposite surfaces, the contact resistance will normally increase due to
an inter-metallic compound formation under pressure. For insulating oxide films or thin
weakly conducting films, conduction will only take place once the film is ruptured under
force to enable metal to metal contact. A change in the surface hardness due to the
presence of a metal film may also change its contact resistance. Usually surfaces are
electroplated with nickel to stop inter-diffusion between mating surfaces. Ifthe
electroplated material is such that the surface hardness is reduced as compared to original
surface, then the contact resistance will decrease. It is also possible that an electroplated
film has higher conductivity than the original surface. Coatings also help to stop
oxidation thus improving contact resistance. However, if the fihn is porous it may not be
able to perform this protective role under long term operation. The contact resistance
depends upon the relative resistance of the deposited film to that of the substrate and the
radius of a-spots. As shown in the Figure 3-13 if the fihn resistance is higher, electric
lines of force will be constricted and contact resistance will be higher. In this case the
film resistance overshadows constriction resistance. Ifthe deposited film resistance is
lower than the substrate, then the electric lines of force will spread more in the film and

the underlying metal and electric contact resistance will be low.

89

 

 

 

 

 

 

// \\
// \\

Figure 3-13. Electric lines of force in (a) high resistance fihn (b) low resistance film.

3.3 Room Temperature Spring Loaded module parameters

To study room temperature spring loaded parameters a new apparatus was
designed as shown in Figure 3-14 below. The apparatus was designed such that it could
fit in room temperature scanning system that allowed contact resistance and electrical
conductivity measurements. Two types of springs were used in these experiments:

compression springs and die springs, both purchased from Mcmaster Carr Inc.

90

Compression springs are soft and have force range of 10-15 lbs for l”-l .5” length
Whereas; die springs are hard and have a range of up to 30 lbs force for same length. With
multiple experiments it was found that surface roughness had more impact than spring
load for a low resistance module. A better-polished surface with less spring force would

give low contact resistance.

 

Figure 3-14. Spring loaded design for RT scan.

Initial experiments with cast ingot p-type material showed high contact resistance.
The first few polishing attempts did not produce a good result as it was not possible to
hold 5mm x5mm x 7mm samples by hand and get flat surface during polishing. The
sample would inevitably tilt and there would be unparallel surface mating between the

electrode and the TB leg, leaving large gaps. To solve this problem, a two leg stainless

91

steel sample holder was designed as shown in Figure 3-15 (a) and (b) below. Legs were
firmly secured to avoid any movement during polishing which also helped to remove any
size difference between n-type and p-type legs as both were polished in the same run. A
sequential polishing procedure was used that included a first polishing step with 800 grit
sand paper, a second polishing step with 1200 grit san paper and third step of using a 0.3
micron liquid alumina solution on felt cloth. All three steps were canied out on a Leco

2000 polisher as shown in Figure 3-15 (0) below.

92

 

Figure 3-15. (a) Sample holder for polishing (b) TE legs secured inside holder (c)

Leco 2000 Polisher with felt cloth on right side plate.

The best p-type result after polishing is shown in Figure 3-16. The contact

resistance is 187 ufl cm2 which is very high as compared to our goal of less than ~25

uflcmz. This contact resistance goal is based upon 10% of total thermoelectric leg

resistance. For a p-type leg with electrical conductivity of 1400 S/cm, the total leg

93

resistance for a 5mm x5mm x 7mm leg is ~ 2 (m9). If we define 6 as the ratio of contact
resistance to total leg resistance, then for this case 8 = 0.374. However it can be noted
that 4-probe resistance for a single leg is 2.7 (m0). Figure 3-17 below shows the result

for LAST (ETN 224) cast n-type TE leg with spring loaded contacts. Here again contact

resistance is 615 110cm2 which is higher than acceptable limit of 10% of leg resistance.

For electrical conductivity of 800 S/cm, total leg resistance for 5mm x 5mm x 7mm
LAST (ETN 224) leg is 3.5 (m0). For this leg, the ratio of contact resistance to leg
resistance 5 = 0.70 and the four-probe resistance for the same leg is 7.1 (m9). So if we
combine the individual 4-probe resistance of both n-type and p-type leg, we will get a
total resistance of 9.8 (m0) for this module. We have observed that modules with total 4-

probe resistance < 10 (m0) give output in the range of 7 to 8 (A) short circuit current.

 

 

 

 

250 l I 1 i
0 ~ 1400 Slcm ETP 53B ...-
2 O
200 _ch ~ 187 uncm ....“ a
spring force ~ 44 lbs ....
A O
3150 - ,-" —
0 Q.
s .-
3 o
g 100 ~ .. 4
.0
5O __ 4 probe R ~ 2.7 m9
Ni coated Cu electrode
0 O
~———6«oooe 1 1 1

-2 O 2 4 6 8
Position (mm)

Figure 3-16. Spring loaded ETP53 with Ni plated Cu electrode.

94

 

700 . .
0’ ~ 833 Slcm
600 §Pfin9 force ~ 44.28 lbs ....- 4
4-probe R ~ 7.1 mt: ...-

500 _Ni coated Cu electrode ..o" fl
....“

o ETN 224

.h
o
o
l
.0

l
O
L

300

Voltage (IN)

I
l

200
ch ~ 615 "mm2
100 _ ~

 

 

[—

O W1 L 1
-2 O 2 4 6 8 10
Position (mm)

Figure 3-17. Spring loaded ETN 224 with Ni plated Cu electrode.

As described in literature [3 7], initial module resistance for spring loaded modules
is very high as compared to diffusion bonded modules. This is because surface oxide
films on TE legs and the electrode initially prevent good electrical contact. However,
when the module is heated close to 400 0C range, these oxides dissipate to produce direct
mating between the surfaces, giving good electrical contact. Figure 3-18 below shows a

plot for spring loaded hot pressed n-type LAST (MSUHP 94). Its contact resistance of

2 . . . .
250 uQcm rs much lower than that of a cast 1ngotwh1ch had a contact resrstance of 61 5

uQcmz, whereas, the electrical conductivity of hot pressed leg is ~ 100 S/cm and that of
the cast TE n-type leg is ~ 800 S/cm. The ratio of contact resistance to leg resistance is 8

= 0.036. We speculate that this improved contact resistance may be because of the

95

smaller grain size of the hot pressed (5-50 micron) as compared to the ~ 700 micron grain

size of the cast material.

3.4

 

 

2000 . we .- . . LL_ a- mm“
0' ~ 102 Slcm
i __ 2 ,0.
1500 | RG 250 pncm .
[ spring force ~ 50 lbs .'
9 4 probe R ~ 25 m!) .-
3 [ Ni plated Cu electrode .°
3,1000 . o a.
3 ‘ e
T) .'
> C
500 . ,- MSU HP 94
..
0'.

O W l ' — . —-—¢ _____1_____i
-1 0 1 2 3 4 5 6 7

Position (mm)

 

Figure 3-18. Spring loaded hot pressed sample with Ni plated Cu electrode.

Spring Loaded High Temperature Short Circuit Current Measurements

After several room temperature scans, a high temperature unicouple output

measurement apparatus was designed as shown in Figure 3-19 below. It uses two springs

and can accommodate a locally made Nichrome (N i80/Cr20) wire-based heater which is

assembled by first winding Nichrome wire on a square piece of alumina, then

sandwiching it between another two alumina plates. The three pieces are held together by

Fortafix QS/B4 high temperature binder. These heaters have a short life and after a few

runs they either fail electrically or alumina plates crack under stress at high temperature.

96

The first design did not have the four supporting pillars with the middle plate and had
difficulty in maintaining balance on the unicouple electrode. The bottom plate is made of

Cu for good thermal conductivity which helps keep the cold side temperature down.

 

Figure 3-19. High temperature spring loaded unicouple measurement system.

 

Figure 3-20. Spring loaded apparatus on module testing system (LTMTS).

97

As mentioned before, due to the high resistance of the measurement wires of the
LTMTS, a thick, short wire was used with individual unicouples during short circuit
current measurement. If this wire was only soldered to the cold side electrode, it would
fall off at high temperature making measurements impossible. To overcome this problem,
the soldered joint on the cold side was also coated with QS/B4 for mechanical support.
However, this permanent short circuit wire does not allow an open circuit voltage
measurement during operation. The first spring-loaded holder did not allow good contact
between the electrode and TE legs. The first module tested with this holder was measured
under ambient conditions and consisted of LAST (ETN 221), LASTT (ETP 55) and Ni

plated Cu electrode. The output is shown in Figure 3-21.

spfingloaded
ETP55, ETN 221- Ni plated Cu

 

 
   

5
max current ~ 4.27 A

g 4 “‘ at 553 °c “
"E starting R ~ 27 m!)
g 3 _ —
3
0
§ 2 _ -
'8
t
.8 1 ~ —
ta

0

 

 

 

O 1 00 200 300 400 500 600
Temperature (hot side) °C

Figure 3-21. First spring loaded module with Ni plated Cu electrode.

98

For this first module starting four probe resistance was R~ 27 (m0), maximum
current obtained was 4.27 (A) at 553 0C with corresponding cold side temperature of 55
°C, whereas; expected (simulated) value of short circuit current at this temperature using
an iterative technique is 11.1 (A). At 553 0C the heater failed and it was not possible to
make any further measurements. The second module tested used an improved design for
the module holder that included four pillars passing through the middle plate. This helped
to align and keep the two plates parallel thus applying uniform pressure on the hot-side
electrode under load. The second module consisted of LASTT (ETP 55) and LAST (ETN
184) with Ni plated Cu electrode and its short circuit current reached 5.82 (A) for cold
side temperature of 43 °C as shown in Figure 3-22 below. For these values, the expected

short circuit current calculated using iterative technique is 10.69 (A).

spring loaded
ETP 55, ETN 184- Ni plated Cu

1 l

6 ,1 r -3
max current 5.82 A

at 500 °C
starting R ~ 29.7 mg

 

     
 
  

Ch

 

l

.h
r
-mnznlzzz

h)
___T__tfl-

in .---L-

measured in open air

short circuit current (A)
or

 

A
o 1 —.—

fio 260 360 460 560 600
Temperature (hot side) °C

Figure 3-22. Second spring loaded module measured in open air.

99

The starting resistance of this module was R ~ 30 (m0), but when the module was
measured after the thermal cycle, resistance dropped to 17 (m0) while still under spring
load. Probably, the materials expanded during run and developed more a-spots, yielding
better conductivity. The springs were further compressed to increase the load and the 4-
probe module resistance dropped to 13.7 (mil) indicating that even after thermal cycle
under ambient condition, pressing the load can break oxide films and increase the true
contact area between TE material and electrode. The module was again checked for
resistance the next day and the value did not change. This suggested that modules can

survive room temperature oxidation.

spfingloaded
6 ETP 55, ETN 184 - Ni plated Cu

 

  
 
   

max current ~ 5.97 A
at 500 °C

01

.b
I

measured inside LTMTS _

short circuit current (A)
N 00

A
l

 

 

 

0 1 00 200 300 400 500 600
Temperaure (hot side) °C

Figure 3-23. First module measured inside LTMTS under Ar.

100

The second spring loaded module was again measured inside the LTMTS under
Ar with a new Ni plated Cu electrode. The open circuit voltage reached 190 mV with
short circuit current of 5.67 (A) as shown in Figure 3-23. However, module could not
sustain the output current and it started decreasing. The module was cooled down and
taken out of measurement system. After cooling, the 4-probe resistance was 20 (m9).
When the module was removed from the spring loaded structure, resistance increased to
86 (m0), confirming the idea that even with high resistance surfaces, spring loading can

improve the contact area. No sign of Cu diffusion was observed visually.

3.4.1 New heater location

In previous modules the heater was not placed directly on the module, but rather
the middle of the Cu plate contacted the module and the heater was placed on top of the

Cu plate. Because of this indirect heating of the module, heat losses were high.

Spring Loaded

 

   
  

 

 

7 [fl ETP 55, ETN 221- Ni plated Cu
6 starting R ~ 43 m 0 1
2 spring force ~ 14 lbs '
:f 5 , heater placed on top
E z of sample
:1 4 F
0
.3 3 T _
t 2 1 4.
o 1 ~
.2 l
(0 1 i. a
O L l l l i l .
100 200 300 400 500 600 700

Temperature (hot side) °C

Figure 3-24. Spring loaded module with heater directly on unicouple.

101

A third module was prepared with LASTT (ETP 55), LAST (ETN 221) and Ni
plated Cu. The heater was placed directly on top of the unicouple for better heat transfer.
The output is shown in Figure 3-24 above. Maximum short circuit current obtained was
5.24 (A) at 490 0C. However, since now the heater was much closer, the cold side
temperature also rapidly increased to 71 °C. The current started decreasing after the
module reached 490 oC. Expected short circuit current using an iterative technique for
these values of hot side and cold side temperature is 10.43 (A).

A fourth module was prepared with LASTT (ETP 55) and LAST (ETN 164). The
n—type leg (ETN 164) had good temperature dependent electrical conductivity values as

shown in Figure 3-25 below.

 

ETN 164 electrical conductivity
1400 . . a— fi 1 a

1200

..__ '_'I A
O

..L
C3
C)
C)

.
. .M,..H
: O
. .
I O
1

o ; 1 i I
800 _H .p m..§flhmiflflw, ., m, ..g _j
; 9 .

600 ~-

400 L , . I. . 7’9.9..§.,, . _5

Elect. Cond. (Slcm)

200 i i 1 . 1 L . l
300 350 400 450 500 550 600 650
Furnace Temperature (K)

 

Figure 3-25. Temperature dependent electrical conductivity for ETN 164.

102

As mentioned before, to measure open air short circuit current, the short circuit
Cu wire was coated after soldering with QS/B4 so that at higher temperature the solder
would not melt on the cold side and hold the wire for measurement. The module was
coated with QS/B4 to avoid sublimation during operation. The module was spring loaded
and heater was placed immediately on top of the unicouple. This module was measured
in open air and short circuit current reached 10.0 (A) for a hot side temperature of 589 OC
and corresponding cold side temperature of 91 °C as shown in Figure 3-26 below.
Expected short circuit current for these values of temperatures using iterative technique is

10.72 (A), however, current started decreasing immediately.

spring loaded
ETP 55, lETN 164- Ni plated Cu

 

.3
N

max current ~10.0 A
at 589 °C

.3
O
1
\

    
  

short circuit current (A)
C)

 

   

 

 

   

 

0 1 00 200 300 400 500 600
Temperature (hot side) °C

Figure 3-26. Spring loaded module with highest recorded current.

103

3.4.2 Spring Loaded Hot Pressed Module

Next a spring loaded module was prepared with a hot pressed n-type material
(MSUHP 94) which had room temperature electrical conductivity of ~ 100 (S/cm). High
temperature measurements for electrical conductivity and thermopower are not available
at the moment. However, based on previous hot pressed sample data of electrical
conductivity and thermopower we assumed that module output would be low. The heater
was placed immediately on top of the unicouple and the module was measured in open

air. The output values for short circuit current are shown in Figure 3-27 below.

spfingloaded
ETP 55, HP 94-Ni plated Cu

max current ~ 6.77 A
at 585 °C

V

 

short circuit current (A)
N 0) h 01 O)
F l I l i
l l 1 41

A
I
l

 

 

O

 

o 100 200 300 400 500 600
Temperature (hot side) °C

Figure 3-27. Spring loaded module with MSUHP94.

104

As shown above, the short circuit current reached 6.77 (A) for a cold side temperature of
83 °C. This is the highest short circuit current observed for any hot pressed module in our

lab, however, current started decreasing immediately at this temperature.

3.4.3 Spring Loaded Ni Electrode Module

Previous modules used Ni plated Cu for better electrical conductivity but Cu is
notorious for its high reactivity and a thin Ni coating cannot stop Cu diffusion through
TE material at high temperature. A thick Ni electrode of 5mm x 5mm x 12mm (R ~ 33.5
110) was used as a hot side electrode for next experiment. This module was spring loaded
and coated with QS/B4 to avoid sublimation during heating and it reached an output short
circuit current value of 8.00 (A) for a hot side temperature of 515 0C and cold side

temperature of 62 °C as shown in Figure 3-28 below.

 

 

 

 

spring loaded

10 g _ ETP52, ETN 225 - thick Ni fi_g

1 max current ~ 8 A i

i o 1

g 3 ._ at 515 C ..

E J

g 6 . ,

3 l

0 l

8 4 ;. 1
'8 .

‘5 1

s 2 - i

l

0 |

 

0 1 00 200 300 400 500 600
Temperature (hot side) °C

Figure 3-28. Spring loaded module with Ni electrode.

105

The expected value of short circuit current using iterative technique for these
values of hot side and cold side temperature is 10.71 (A), however; current started
decreasing after this temperature.

A second module was prepared with ETN 225, ETP 52 H, thick Ni electrode and
measured inside the module testing system (LTMTS) under Ar. This module reached
6.13 (A) for hot side temperature of 600 oC and cold side temperature of 57 0C as shown
in Figure 3-29 below. The low current can be attributed to the high resistance of
measurement wires of the module testing system. The current started decreasing at 600

°C and this module was brought down to room temperature.

spfingloaded
ETN 225, ETP 52 ll- thick Ni

 

 

 

 

7 I T I I 1 r

6 max current ~ 6.13 A
2 L at 600 °c 2
rug: 5 — L
t
3 4 P a
.3 3 . _
'6
t 2 — measured in LTMTS .
.2 underAr
m

1 ~ a

O 4 l l 1 g 1 1

O 100 200 300 400 500 600 700

Temperature (hot side) °C

Figure 3-29. Spring loaded Ni electrode based module measured under Ar.

106

3.4.4 Spring Loaded Silver (Ag) Electrode Module

Silver (Ag) has the lowest electrical resistivity (15.87 nQ-m) and highest thermal
conductivity (429 W/m-K). Modules based on Ag electrode are supposed to perform

better than Ni based modules due to these properties. However Ag reacts with sulfur (S)

in air and turns into tarnish brown color silver-sulfide (Ang) and reacts with tellurium in

TE material to form silver-telluride (Ange) which is an n-type semiconductor. The next

module was prepared with silver electrode using LAST (ETN 225) and LASTT (ETP
52H). This module was measured in open air and its short circuit cmrent reached 8.00 (A)
for a hot side temperature of 507 °C and a cold side temperature of 78 °C as shown in
Figure 3-30 below. The expected short circuit current using iterative technique for these
temperatures is 10.40 (A). The current started decreasing at 507 °C and the module was

cooled down to room temperature.

spring loaded
10 ETN 225, ETP 521", Ag electrode

 

max current ~ 8.0 A

. .
gs» atsoec )
E .
0
H l
u l
.5 .
e4» -1!
'5
g . .
52L 1

l
0’ '

 

 

o 100 200 300 400 500 600'
Temperature (hot side) °C

Figure 3-30. Spring loaded Ag electrode module.

107

3.4.5 Spring Loaded Module with tungsten (W) coated samples

In all previous spring loaded experiments current could not be sustained at its
peak value. EDS has shown that Te comes off at the TB leg surface and reacts with the
electrode. Kortier [39] has successfully used tungsten (W) with PbTe thermoelectric legs
as hot side electrode since tungsten is non reactive to Te. However, tungsten makes a

stable bond with oxygen at higher temperatures which is difficult to remove. Moreover

-6 .
tungsten has a very low co-efficient of thermal expansion (CTE) ~ (4.5 x 10 m/K)

-6
which makes it unsuitable for diffusion bonding with LAST (with CTE of 22 x 10

in/K). However, since spring loaded modules are supposed to have no permanent bond
between TB legs and electrode, tungsten can still work. For this experiment, LAST and
LASTT legs were coated with tungsten (~10 nm) then Ti (~ 200 nm) and Au film inside
the physical vapor deposition (PVD) system through e-beam evaporation of target
material. Ti film was coated as it is more resistant to oxidation than tungsten and Au film
was deposited to increase surface conductivity. A spring loaded module was prepared
using these coated TE legs (ETN 225 and ETP 52) and thick Ni electrode and measured
inside the module testing system (LTMTS) under Ar. The reason for Ar atmosphere was

to provide an oxygen free environment. The module out put is shown in Figure 3-31.

108

spfingloaded
r ETP51?“ ETN225- thick Ni

 

(J1

coated with W,Ti, Au _, 'E

.5

(JD

A

 

short circuit current (A)
N

O
1

measured under Ar

 

 

 

I
—\.

0 100 200 300 400 500 600
Temperature (hot side) °C

Figure 3-31. Spring loaded tungsten coated module measured under Ar.

It appears that coated film has high resistance as current initially only reached 3.4
(A) at 561 °C. The temperature was decreased and increased again to see thermal cycling
effect and current started increasing again and reached 4.3 (A) for about the same hot side
temperature. This led to the module being cooled down and opened to inspect the surface.
A dark green film could be seen indicating tungsten-titanium-oxide (WTiO). Also, there
were cracks on the film and we postulate that fiom cracks tellurium seeped out and was
responsible for late increase in output current. The heater failed during the second cycling
and it was not possible to do any further measurements.

The same module was again tested in open air with a new Ni coated Cu electrode

and TE leg surfaces were left untouched fi'om previous run to see the post run effect of

109

coated film. Short circuit output current is shown in Figure 3-32 below. Current could not

be sustained at high temperature and module was cooled down.

spring loaded
W,Ti,Au coated ETN 225, ETP 52- Ni plated Cu

7 fi T 7 T l— fi 1

 

  
   
  
  

 

l I
2 6 ‘ max current ~ 6.22 i
2' 5 ,»- at 582 °c 4
c i i
e l l
S 4 )
o l
'3 3 ;. J
'6
t 2 r
4% : measuren in open air ;

 

o 100 200 300 400 500 600
Temperature (hot side) °C

Figure 3-32. Spring loaded module with tungsten coating measured in air.

After this run module was opened to investigate hot side surfaces. There was no
indication of Cu diffusion into TE leg, however tungsten coating was black like a thick
carbon fihn and rather a depression could be seen in Ni coated Cu electrode indicating

strong oxidation.

110

3.4.6 Spring Loaded and InGa Based Combined Module

Experiments were carried out to see the combined effects of using InGa under
spring load. For this a module was prepared with LAST (ETN 221), LASTT (ETP 55)
polished legs, InGa was used as liquid contact and Ni plated Cu was used as hot side
electrode. The module was spring loaded and measured under Ar inside the LTMTS.

Before high temperature cycling module 4-probe resistance was R ~ 8.5 (m0).

ETP 55, ETN 224- InGa-Ni plated Cu

 

 

 

 

8 V I I I 1

7 __ max current ~ 7.52 A _
A 0 ‘il
5. 6 _ at 583 C B/ A
t" /
C I;
g 5 —- measured inside / . i
3 LTMTS - ,
:1: 4 ~- / ~
./
"'3 3 i— ,i measured in
E 2 i open air _
.c
m

1 .. i

0 1 1 1 1 L

0 100 200 300 400 500 600

Temperature (hot side) °C

Figure 3-33. Combined spring loaded and InGa liquid contact based module.

As shown in Figure 3-33 above, the short circuit current reached 7.52 (A) for a hot side
temperature of 5 83 °C and a cold side temperature of 33 °C with an open circuit voltage

of 212 (mV). The current started decreasing at this temperature and module was cooled

111

down to room temperature. The same module was then measured in open air with a
starting resistance of 12.0 (mil). As shown in Figure 32, short circuit current reached
6.47 (A) for a hot side temperature of 570 °C with a cold side temperature of 42 °C. This
low output current is probably due to InGa reaction with LAST and LASTT in previous
high temperature cycle. A similar experiment was carried outwith silver (Ag) electrode
combining InGa and spring loaded configuration. Two open air high temperature runs
were carried out as shown in Figure 3-34 below. In the first run measurement wire broke
at 440 °C and experiment could not be continued. In the second run the output current

reached 5.93 (A) at 600 °C with corresponding cold side temperature of 106 0C.

spfingloaded
ETP 55, ETN 221- InGa-Ag electrode

l I l T l l

 

 

   
   

h 01
. WT

short circuit current (A)
00

open air measurement i

 

O 1 00 200 300 400 500 600 700
Temperature (hot side) °C

Figure 3-34. Spring loaded InGa based Ag electrode module.

112

The module was cooled down and after a visual inspection it did not appear that InGa

reacted with Ag but a thick layer could be seen on module’s surface facing the electrode

which probably consisted of high resistance AgTe.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3-2. Summary results for spring loaded modules.

Electrode Material Air/Ar Th Tc I Comments
& (°C) (°C) “0‘"
coatim

Ni plated Cu ETP 55 Air 553 55 4.27 First design
ETN 221

Ni plated Cu ETP 55 Air 500 43 5.82 Improved design
ETN 184 First run

Ni plated Cu ETP 55 Ar 500 31 5.67 second run
ETN 184

Ni plated Cu ETP 55 Air 490 71 5.24 Heater on
ETN 221 unicouple

Ni plated Cu ETP 55 Air 589 91 10.72 Thick short wire
ETN 224

Ni plated Cu ETP 55 Air 585 83 6.77 Hot pressed n-
HP 94 type

Ni ETP 52 Air 515 62 8.00 Thick Ni with
ETN 225 33.5 ufl

Ni ETP 52 Ar 600 57 6.13 Measured inside
ETN 225 LTMTS

Ag ETP 52 Air 606 78 8.0 Silver electrode
ETN 225

Ni (Thick, W-Ti-Au Ar 561 51 3 .4 Measured inside

33.5 1152) ETP 52 LTMTS.
ETN 225 First nm

Ni plated Cu W-Ti-Au Air 582 67 6.22 Second run with
ETP 52 different
ETN 225 electrode

Ni plated Cu ETP 55 Ar 583 33 7.52 Measured inside
ETN 224 LTMTS, 1St run

Ni plated Cu ETP 55 Air 570 42 6.47 Second run

with InGa) ETN 224
Ag (with ETP 55 Air 600 106 5.93 Silver electrode
InGa) ETN 221

 

 

 

 

 

 

 

113

 

Conclusion

A series of experiments were carried out to obtain short circuit current for LAST and
LASTT material using InGa liquid contacts and a spring loaded configuration. Both types
of module fabrication techniques produced high output currents of 10.5 (A) and 10.0 (A)
respectively. Different types of coatings were tested on TE leg surfaces to prevent
reaction between LAST, LASTT and InGa at high temperature. In case of spring loaded
modules tungsten films were tested but these efforts have not been able to sustain current
at their peak value during high temperature operation. The spring loaded design looks
more promising since it has been employed in the past for PbTe based module. Future
experiments will include using a tungsten electrode to prevent tellurium reactivity with

hot side electrode.

114

Chapter 4: High Temperature thermoelectric contact investigation

using Voltage Profiling System (HTVPS)

Introduction

The compound Pb-Sb—Ag-Te (LAST) has shown promising thermoelectric
properties for power generation application [1]. TE power generation modules made with
this compound operate at ~ 900 K. High electrical conductivity, low thermal conductivity

and low contact resistance are important characteristics of thermoelectric modules. At

. . . . 2
room temperature contact reSIStance acceptable limit is below 20 uQ-cm for 5 mm x5

mm x 7 mm LAST based TE legs (within 10% of total leg resistance) and various
methods are under investigation to achieve this goal [23]. However, contact resistance is
a function of temperature as rate of diffusion changes with temperature. Moreover, at
high temperature materials undergo thermal stresses which may induce micro cracks in
the material resulting in degradation of TE module performance. Electrical conductivity
of LAST TE material decreases with increasing temperature while contact resistance
tends to increase. To investigate electrical conductivity and contact resistance of
thermoelectric modules at high temperatures, a novel voltage profiling system is designed
and fabricated. The system is designed to scan across the individual samples for in situ
monitoring of voltage variations along a sample from which electrical conductivity and
contact resistivity can be determined. The motor motion, furnace temperature, and data
acquisition are automated with LabVlEW National Instruments software. To confirm the

validity of the system NIST reference sample SRM 1461 was measured at room

115

temperature and ~ 700 K, showing electrical conductivity values of 12585 S/cm and 9505
S/cm which is in good agreement with reference data Thermoelectric unicouples were

then scanned for high temperature contact resistance measurement.

4.1 System Description

The system is designed to measure electrical conductivity and contact resistances
within the range of 300K to 800K. The system works on a four-probe measurement
principle. A variable magnitude square wave function switching current is supplied at 33
Hz with a 50 % duty cycle. Usually for LAST material samples, a 100 m(A) current is
used. Current is sourced through a Keithley 2400 source meter and voltage is measured
by Keithley 2002 Multimeter. The unique feature of this system is the use of bellows. A
20DAM series (20DAM10D2U-K) digital linear actuator DC stepper motor by Portescap
is used for bellows movement. It has linear travel of 0.0254 mm per step with maximum
travel distance of 15 mm. It uses a 12 V power supply and produces a maximum force of
20 N. A Lini Stepper controller along with NI USB-6501 24 line digital 1/0 is used for
motor interface to a PC. A detailed system sketch is shown in Figure 4-1. To create the
scanning motion, the bellows is positioned on top of the motor. Bellows movement is
transferred through an alumina tube, which is coupled to the bellows by a one-inch
stainless steel interface at the bottom. At the top end of this tube is the probe tip which
drags across the sample with up-down movement of the bellows. The probe tip design is
an important aspect of this apparatus. Initially a tungsten wire wrapped around an
alumina tube was used. But it did not effectively make contact with sample in motion that

resulted in erroneous data. Our previous experience of using pogo connector pins with a

116

room temperature scanning probe was a success. The same idea is applied here and a new
probe pin holder was designed as shown in Figure 4-2. The collar design firmly holds on
top of the alumina tube with the help of a set screw and another two set screws hold the
pogo connector in place. The connector is pushed against the sample and tightened with
the set screws. The spring inside the connector ensures a good connection with the
sample even if the sample surface is not perfectly flat.

The sample holder is a round nickel disc with five holes to accommodate the top
edge of the alumina tubes. Wires for current sourcing and voltage measurements are run
through four alumina tubes. Alumina is selected as it can withstand high temperatures. At
the base an aluminum cage is designed to hold these tubes with provision for addition of
a fifth tube. This aluminum cage is fastened to the base plate where bellows are
positioned on top of the DC motor. The sample holder and alumina tubes are inserted into
a quartz tube to maintain vacuum and sustain a high temperature inside the furnace. A
PID controlled ceramic furnace by Omega is used which can go up to ~ 982 °C with RS-
232 interface used to automate the furnace. There is also provision for using a local
heater inside the quartz tube. The sample is placed so that it touches the heater. This
allows visual inspection at high temperature of pogo pin movement across the sample
contact. This local heater is custom made with Ni80/Cr20 thin wire with a heater
resistance in the range of ~ 15 Q. A type K thermocouple is used to measure furnace
temperature. Since the sample is under vacuum, there is a difference between sample
temperature and furnace temperature. For accuracy of data, the sample temperature is
also measured using a single ended Pt/PtRh Type-S thermocouple. Silver paste is used to

attach the thermocouple and measurement wires to the sample. Initially we used low

117

temperature silver past for high temperature measurements but this resulted in erroneous
measurements. To overcome this problem a 2” x 1” stainless steel rectangle was designed
as shown in Figure 4-3. The sample is now pressed under a screw during operation which
helps to maintain good contact of the measuring voltage and current wires with the TB
sample. Lately we have started using high temperature silver paste by SPI Supplies and
wire contacts have performed better than with low temperature silver paste. Omega RTD
Pt (1009) is used along with two Keithley 2182 Nanovoltrneters for reference and sample
temperature measurement. A Keithley 2601 source meter is used to source 10 11A to the
RTD. For better noise immunity, twin axial cable and connectors are used. The vacuum
system consists of a Varian V70 Turbo pump and a dry scroll-roughing pump. This
enables the system to achieve better than milli-Torr range vacuum. All materials have the
tendency to outgas under vacuum at high temperature which has caused a problem with
measurement as the Au from measuring probe would come off at higher temperature and
deposit on the LAST material. Also, the motor would not be able to pull down the probe
after upward motion as bellows had to be moved against the vacuum suction power. It
was also noticed that backfilling the bonding furnace apparatus in this lab with non
reactive gas like Ar improves the results. Now the system is equipped with an Ar gas

backfilling option. [BEE-488 GPIB is used to control all equipment through a PC.

118

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample ,Z..3‘ probe
holder '
Alumina tubes —":_’ Quartz
‘ Tube
Furnace —’§
PC ...... .. Vacuum
l Almninum
cage ——> .
Source meter Ar
Voltrneter Bellows
Motor Stepper motor
controller

 

 

 

 

Figure 4-1. High Temperature voltage profiling system.

 

-1
--

.--d
-

 

h

 

C

Figure 4-2. Probe pin holder.

 

 

 

b-Cid
p---

 

119

  
 
 
 
  

Pressing

screw Stainless
4— steel
tru tur
TE sample ..., S C e
Measuring pogo

 

pin movement
across contact

 

Electrode—>

Figure 4-3. New sample holding arrangement.

4.2 System Operation

For all voltage measurements, the current through the sample is rapidly switched
from positive to negative values (at a frequency of greater than 30 Hz) to avoid thermal
offset voltages [40] which can be a significant source of error. The current is flipped
more than 20 times and the magnitude of the readings is averaged for a single
measurement. Voltage is scanned along the surface of the sample and plotted against
position (X). The electrical conductivity (0) is then determined by calculating the slope

dV/dX along with sample dimensions using following expression:

I S

cm
—— - area
dx

This method is more accurate than typical four-probe measurement as the voltage profile

is fitted to many points along the sample instead of just two points. To ascertain the

120

accuracy and validity of equipment a NIST SRM 1461 sample was cut into 2.9rrnn x 2.9
mm x 9.7 mm dimensions for electrical conductivity (0) measurements. The sample was
first scanned at 294 K and then at ~700 K. As expected there is a considerable decrease in
electrical conductivity at higher temperatures. As shown in Figure 4-4 measured values
are 12686 S/cm at room temperature and 8606 S/cm at ~ 700 K, which are in good

agreement with reference data.

 

 

 

 

 

 

 

 

ms‘r SRM 1451 NIST SRM 1461
20 . , j , , 25 . , , , '__
Sample size= 2.9 x ' Sample size= 2.9 x A ’
2.9 x 9.7 mm 20 I 2.9 x 9.7 mm o .
15 - current=100mA . 4 current= 100mA ‘
5‘ slope = 9.44 5‘ slope = 12.51
3 o=12585 Slcm . 315 "a: 9505 Slcm 0, ‘
8.1m T~294K .- - 8: T~708K
g g 10 . .
> > .
5 - . " 5 b ..
o _3~ - e . 1 o a - ,- . -
cs 0.5 1 1.5 2 2.5 -o.5 o 0.5 1 1.5 2 2.5

Position (mm) Position (mm)

Figure 4-4. NIST SRM 1461 room temperature and high temperature plots.

4.3 High temperature contact investigation

High temperature contact resistance measurements were carried out to determine
the viability of different diffusion bonding alloys using the high temperature voltage
profiling system (HTVPS) described above. The flow chart in Figure 4-5 shows the
different bonding materials investigated for high temperature contact resistance

measurement.

121

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

High Temperature
contact resistance
investigation
SnTe based Premabraze Cerromatrix
diffusion 616 based solder based
bonding bonding bonding
Cast n-type Cast n-type Cast n-type
Bonding bonding bonding
using SnTe using using
premabraze cerromatrix
Cast p-type Cast p-type Cast p-type
bonding bonding bonding
using SnTe using using
premabraze cerromatrix
Hot
pressed n-
type
bonding
using SnTe

 

 

 

 

 

 

Figure 4-5. Bonds and materials investigated for high temperature contact resistance

measurement.

122

4.3.1 SnTe based modules

SnTe is a p-type semiconductor extensively studied for PbTe and

(AngTe2)0.15(GeTe)0.85 (commonly known as TAGSSS) based TE modules [41].

Seetoo[42] used a stoichiometric Sn and Te powder mixture at the bonding site to
generate sufficient heat through an exothermic reaction which is local to the bonding site
and does not affect the entire TE leg. We follow Seetoo [42] for our bonding procedures.

Sn and Te powders are added in a ratio of 70:30 respectively, mixed thoroughly
and then cold pressed in a dye producing circular discs of ~ 500 microns thick and 6 mm
in diameter. The prepared disc is placed between the hot shoe (comprised of either
stainless steel 316 or Ni) and LAST material. This arrangement is pressed under load in a
custom-made diffusion bonding apparatus and inserted into a three zone fiimace operated
under Ar. Temperature is raised to 888 K and is allowed a bonding time of ~ 4hrs. The
exothermic reaction can be described as:

Sn +Te , SnTe +Q (4.2)

Here Q is the heat generated during the process. This reaction is initiated at about 571 K

[42]. As mentioned previously LAST material has a linear CTE of 22 x 10.6 (ppm/°C)

and a suitable material is needed at the hot side junction to match this property and avoid

shear stress on the mating surfaces during high temperature operation. CTE for SnTe is

-6
21.3 x 10 (ppm/°C) [40] which is closely matched to that of LAST material. The

thermal conductivity of SnTe is ~ 10 W/m-K above 100 K and its electrical conductivity

. 4 . . . .
1s ~ 2 X 10 S/cm [43]. LAST unicouples diffusron bonded usmg cold pressed SnTe were

123

tested for high temperature contact resistance in high temperature voltage profiling

system (HTVPS) and the results are presented in next section.

4.3.1.] SnTe based cast n-type bonded modules

Carpenter high expansion 19-2 alloy has a nominal composition of (0.55C, 1.0

Mn, 0.208i, 2Cr, 19Ni, Bal Fe), electrical resistance value of 79 uQ-cm and is used for

hot side electrode because it has a CTE of (20 x 10.6 /°C) which is closely matched to the

CTE of LAST material (22 x 10.6 /°C). This helps to avoid shear stresses during heating

of unicouple. As can been seen fiom the following plots it forms a high resistance bond
even at room temperature and as the temperature is increased, contact resistance also
increases. This is probably the reason that in literature [41] and [42] SnTe is used only for
p—type semiconductor TE legs. However, this is not the case in [44] where SnTe was used
to bond both n-type and p-type PbTe legs with W and Ta as the hot shoe with satisfactory
results. Figure 4-6 below shows the progression with temperature for an n-type cast TE
leg ETN 72 bonded to Carpenter Steel 19-2 using cold pressed SnTe. Bonding
temperature was 650 0C and bonding was carried out under Ar for 0.5 hr. As the
temperature was increased inside the HTVPS, the contact resistance decreased. It can also
be observed that the temperature dependent contact resistance is strongly dependent upon
the initial room temperature value of contact resistance. In this particular case there is no
‘break’ between the electrode and TE leg, rather there is a high resistance region at the
junction which is probably due to SnTE diffusion into the TE leg during the bonding
operation. This high temperature region can probably be attributed to oxygen in cold

pressed SnTe as the coin was pressed in open air in this particular case.

124

.A . _‘£““nhnnl

Figure 4-6. ETN 72 bonded to Carpenter Steel 19-2 using SnTe
(a) Room temperature scan (b) Scan at 100 °C. (c) Scan at 200 0C
((1) Scan at 300 °C. (6) Scan at 400 °C (0 Scan at 550 °C.

125

ETN 72A bonded to carpenter steel19-2
using cold pressed SnTe

 

 

 

 

 

 

400 bOI'Idil'Er T3650 °C
scan Temp=25 °C /
300 F 0’ -
9 o =1212 Slcm ..I'
f 200 - e -
a: e
§
§ R =400uflcm20
100 " ° 1 high resistance
0 junction
0 - easel-IO. i -
I l l l (a)
O 2 4 6 8
Position (mm)
ETN 72A bonded to carpenter steel19-2
using cold pressed SnTe
300 bondinng=650 °c
scan T ~ 100 °C /_
250 . 0 ~ 975 Slcm -
o'
A200 " i '
> {If
E:
g 150 - . -
g 0
§ 100 - J -
e
50 - 0 RC = 360 pincm2 ~
0 -____-m - - (b) -
-2 O 2 4 6 8

 

 

 

Position (mm)

126

Figure 4-6. Continued.

ET N 72A bonded to carpenter steel19-2
using cold pressed SnTe

bonding r=550 °c
l I l

 

 

 

 

500 .
scan T ~ 200 °C
400 _ 0 ~ 650 Slcm I, _
i 300 - -
0
g e
'5 200 - . -
> e
100 - 0 RG = 390 pflcm2 .
M
0 l l l 1 (C)
-2 O 2 4 6 8
Position (mm)

ETN 72A bonded to carpenter steel19-2
using cold pressed SnTe

500 bgnding T=§§0 °C 1

 

 

 

 

scan T ~ 300 °C
400 _ 0 ~ 620 Slcm _
A e.
3 300 - .0' -
V e
o
3 I
- 200 -
g i o
e
.. 2
10° , Rc = 152 uQcm
O E “.4 A r (d)
-2 0 2 4 6
Position (mm)

127

mil

800

600

Voltage (pV)
A
O
O

800
700
600

O
O

voltage (uV)
00 h 01
O O
O O

200
100

Figure 4-6. Continued.

ETN 72A bonded to carpenter steel19-2
using cold pressed SnTe

bonding T-650 °C
I -

 

 

scan Temp= 400 °C

a = 600 Slcm 0..

. 2
..e' RC - 170 uflcm

and”. .

(e) ..

 

 

0 2 4 6
Position (mm)

ETN 72A bonded to carpenter steel19-2
using cold pressed SnTe

bonding T=650 °C

 

 

1 l I I

scan T ~ 550 °C 0

0' ~ 287 Slcm ..0’

.e
e
e.

e
e
e
0..
0 R6 = 175 uncm
M . .

2
(f)

‘

 

 

0 2 4 6
Position

128

ET N 72 bonded to Carpenter Steel 19-2

 

 

 

 

 

 

 

 

using SnTe. Bonding T ~ 615 °C
I I I I
800 - ' Voltage (1.1V) 400 °c .. -
‘3 Voltage (pV) 300 °C .'
0 Voltage (pV) 200 °c ..' DUDE
600 P X Voltage (11V) 100 °C .' DUE -
5‘ Voltage (pV) 25 °C ..DDDD , .250
:L . O , ’
V . U00
3, 400 - .°;g§°xxxxx><’ffx _
g @QQQIXXX- * i
0 Bee 5000‘] 1"
> 5 xxx M . +7 '
0§§§++i +
51*
200 - 3% "
-.g
0 - W ..
l l I l
-2 0 2 4 6 8
Position (mm)
Figure 4-7. ETN 72 bonded to carpenter steel 19-2 using SnTe.

Figure 4-7 above shows a second sample ETN 72 bonded to carpenter steel using SnTe

coin. It emphasizes the same point mentioned in previous sample description that as the

temperature increases during measurement, electrical conductivity decreases and contact

resistance increases. However, since the initial contact resistance is high there is no

appreciable change in contact resistance with temperature. Another difference from the

previous sample is that there is a clean ‘break’ between electrode and sample at the

junction whereas in the previous sample there was a high resistance region at the junction

but not a clean ‘break’ (or direct voltage jump). This shows that even with the same

bonding conditions and material it is difficult to reproduce diffusion bonding results.

129

stainless steel 316-SnTe-ETN 153

 

 

 

 

 

 

 

800 1” I I l T W 7
Voltage (11V) 500 °C ‘
600 k D Voltage (uV) 25 °c a
5: o ~ 220 Slcm "
a 400 — g
81 I
g i "7:27:
o in
> [2311
200 I“ C, " fiqIJCU—j—
2 MUCH—l” 0 ~ 800 Slcm
R ~ 450 on em a;
-. 2
0 ~ ETTTT“ R ~ 255 on em .1
I
;_ l 1 l l J
-2 0 2 4 6 3

Position (mm)

Figure 4-8. Stainless steel 316 bonded to ETN 153 with SnTe.

Figure 4-8 above shows a SnTe based diffusion bond between stainless steel 316 and
LAST (ETN 153) with a bonding temperature of 615 °C. A SnTe coin was pressed in
open air and probably had more oxygen which resulted in a high resistance bond. This
follows a similar trend of decreasing electrical conductivity and increasing contact
resistance with increasing temperature. The next logical step is to determine how thermal
cycling would affect contact resistance permanently. As shown in following three plots of
Figure 4-9 the contact resistance initially increases with temperature as compared to a
room temperature scan. As it is cooled to room temperature after thermal cycling, there is
a permanent increase in contact resistance. This may be attributed to thermal expansion

and contraction with changing temperature. Another interesting observation is that

130

electrical conductivity has actually increased after thermal annealing. This has been

previously observed for hot pressed samples but not for east material.

131

Figure 4-9. Stainless steel 316 bonded to ETN 153 using SnTe.

(a) Pre anneal room temperature scan (b) Scan at 550 °C indicating
increase in contact resistance. (c) After anneal room temperature

scan indicating a permanent increase in contact resistance accompanied

with increase in electrical conductivity.

132

500

400 -

9 300

Voltage (11

..x
O
o

-100

1200
1000

800

Voltage (pV)
O)
O
O

.5
O
O

200

0

$831 GSnTe-ETN 157

 

 

I

0 ~ 740 Slcm
scan T ~ 25 °C
bonding T ~ 600 °C

’ i voltage jump

a!"

z

/’

j~115uV

l

 

2

4

Position (mm)

3831 6-SnTe-ETN 1 57

 

L

 

-2

0 ~ 212 Slcm

scan T ~ 550 °C

annealed for 2 hrs

I

f

271 1.1V

voltage jump

Cb)

 

0

2

4

' Position (mm)

133

 

 

Figure 4-9. Continued.

350
300
250

N
O
0

Voltage (pV)
3 a?
O O

 

 

 

 

SS316-SnTe-ETN 157

. 6 ~ 1330 Slcm .
scan T ~ 25 °C a

' after 6 hr anneal
_ at 550 C -
P d

i
.. I voltage jump _

~ 145 uV
- w -
(C)
-2 0 2 4 6 8

Position (mm)

134

10

A similar experiment was carried out with LAST (ETN 157) bonded to Ni using SnTe.
This time the SnTe coin was prepared inside a glove box oxygen free environment and
bonding was carried out at 650 oC. As the following plots in Figure 4-10 show there is
increase in contact resistance at high temperature. However, the pre-annealing contact
resistance is very low and after one annealing cycle the contact resistance is back to the

pre-annealing value thus indicating that module can sustain multiple thermal cycles.

135

Figure 4-10. ETN 218 bonded to Ni electrode using SnTe.
(a) Pre-anneal room temperature scan (b) 500 °C scan indicating increase
in contact resistance. (c) After anneal room temperature scan indicating

contact resistance value returned to pre anneal value.

136

Ni-SNTe-ETN218

 

 

 

 

140 I l I T I I .
e
... e
120 _ o 1200 Slcm .. _
scan T ~ 25 °C .
100 - .o‘ .
S 0’.
a. 80 I' .0 .
0 e
c»
g 60 - O -
o
> 40 - 0 -
.e
20 - .3 -
' R < 25 Mcmz
0 - c (a) -
-1 0 1 2 3 4 5 6
Position (mm)
Ni-SNTe-ETN218
500 l u r l l 1
0 ~ 300 Slcm
e
400 " scan T ~ 500 °C .9 ‘
after 6 hr anneal ..
s "
3; 300 - .0. -
a ...
.8 e‘
'8 200 - ' -
> .0
e
e
100 - ... .
' Rc~120pflcm2 (b)
0 M’l m I J l "
-1 O 1 2 3 4 5 6

 

 

 

 

Position (mm)

137

 

Figure 4-10. Continued.

Ni-SNTe-ETN218

 

 

 

250 l l I l I I
0 ~ 1200 Slcm .
200 - 0'
scan T ~ 25 °C ... '
3‘ after 6 hr anneal ...
3150 _ at 500 C ...... _
e
U ..
3.3 e
g 100 - . -
0..
50 ' ... 2 "
use. R < 25 no em
C
(C)
O I I I I I I
-1 O 1 2 3 4 5

Position (mm)

138

 

Previous modules were tested for short periods of time. In order to run a module long
term testing is required. For this purpose a low contact resistance ETN 153-SnTe-Ni
electrode unicouple was tested for five days under Ar at 450 °C. A SnTe coin was
prepared in a glove box to avoid oxidation which produced low contact resistance
diffusion bond as shown in Figure 4-11. The unicouple had a 4-probe resistance of 4.5
(m9). It can be seen that for the first two days there is negligible change in contact

resistance and by the fifth day contact resistance has increased to 250 uflcmz.

139

Figure 4-1 1. ETN 153 bonded to Ni electrode using SnTe coin.
Bonding temperature 615 0C. (a) Pre-anneal room temperature scan.
(b) 450 0C scan after 12 hrs. (0) 450 °C scan after 48 hrs. ((1) 450 °C

scan after 5 days indicating increased contact resistance.

140

ETN 153-SnTe-Ni electrode

 

 

 

 

 

 

 

 

300 I I l I
.e
scan T ~ 25 °C 0
250 — o~ 11oo Slcm .’
e
..e
3200 — e'
a "
5 o'
'o 150 -
>
100 -
fi
(60
50 I I I I
-2 0 2 4 6
position (mm)
ETN 153-SnTe-Ni electrode
1000 I l l I
scan T 1- 450 °C
800 - after 12 hr anneal
0 ~ 280 Slcm
€600 - e
o I
°’ o'
5 e
E 400 - e
,e’
e
200 - .‘
e
z (b)
O - I I I I
-2 0 2 4 6

Position (mm)

141

Figure 4-11. Continued.

ETN1£H$flbJfldmflflme

 

 

 

 

1000 , . I .
scan T ~ 450 °C
800 _ after 48 hr anneal
o ~ 300 Slcm f.
a 600 - If
° e’
G
g 400 - .’
O
> f
200 - ,0"
o e ...... (c)
-2 o 2 4 6

Position (mm)

ETN 153-SnTe-Ni electrode

 

 

 

 

1000 If I I .
scan T ~ 450 °C
800 . after 5 day anneal ’
o ~ 300 Slcm ...e.
9600 - ..
3
3,400 .
.8
‘o
> 200 .
e 2
o - .... R ~ 225 no em
(d)
'200 1 ' l I
-2 0 2 4 6

Position (mm)

142

4.3.1.2 SnTe based cast p-type bonded module

SnTe bonding is traditionally used for p-type material in literature [41], [42] since
SnTe itself is a p-type semiconductor. In the first sample LASTT (ETP 50) was bonded to
carpenter steel 19-2 at 615 °C using cold pressed SnTe coin. This coin was prepared in
open air and had excessive oxygen. Room temperature contact resistance was high and it
decreased as temperature was increased as shown in Figure 4-12. A major difference
between n-type bond and p-type bond using SnTe is that with n-type cast material, if
initial room temperature contact resistance was high it would not appreciably change with

temperature but for p-type material it still improves with increasing temperature.

 

 

 

 

 

 

 

' Voltage (pV) 600 °C
D o
. Voltage (W) 500 DC ETP 50-SnTe-carpenter steel
1000 Voltage (W) 400 C I I
' Voltage (W) 300 °C
* Voltage (W) 200 °C
800 :3 Voltage (UV) 100 oC .**x$f+** -
O O +++f . C»
Voltage (1.1V)25 C ++++T><Z:::Aééeggém
+ . / A
E 6‘” ' + 32453338” '5’ -
v + 0 Q <20“)
31 a? i I; ‘“ 06’
%400 . _ 2% -
> I 35]ch
1. BED...
200 " L] .. u
130'
I.
0 - IflIII-Ii -
I I I I
-2 0 2 4 6 8

Position (mm)

Figure 4-12. ETP 50 bonded to carpenter steel 19-2 using SnTe. As the

temperature is increased electrical conductivity and contact resistance decrease.

143

A similar behavior is observed for ETP 53 bonded to Ni electrode using SnTe cold
pressed coin indicating that this phenomenon is electrode material independent as shown

in Figure 4-13 below.

144

Figure 4-13. ETP 53 bonded to Ni using SnTe. (a) Pre-anneal room
temperature scan (b) 500 oC scan indicating decrease in contact resistance.
(c) After anneal room temperature scan indicating a permanent decrease in

contact resistance.

145

Ni-SNTe-ETP53

 

 

 

 

 

 

bonding T ~ 615 °C
160 I I I I I I
.e
140 _ o ~ 1750 Slcm ... -
120 scan T ~ 25 °C .’ _
e
a 100 - .0. -
o .0.
e: 30 . .e .
i! .0
a 0'
> 60 - . -
40' e ch~150pncm2 "
20 - ' I -
0 - see... i (a) -
-1 O 1 2 3 4 5 6
Position (mm)
Ni-SNTe-ETP53
800 I I I I I I ‘—
o ~ 300 Slcm o'
e
scan T ~ 500 °C 0
600 F after 6 hr anneal ..
e
9 at 500 °C 0
1 e
v4oo - o -
e e
. D g.
> 200 - ,0 -
e
e
e
O. 2
O _ ”... Rc ~100 p!) cm _
(b)

 

 

Position (mm)

146

Figure 4-13. Continued

 

Ni-SNTe-ETP53
120 [ I l I I I r—
0' ~ 1800 Slcm ,’
100 ‘ RT scan after 6 hr .0
80 _ anneal at 500 °C .’

. e
a, 60 - ......
8’
.. e
> 0
.e
20 - e
0 R ~ 60 at) cm2

 

(C)

 

 

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

147

However there were two occasions when SnT e based p-type bond showed different
behavior and in these instances the contact resistance increased with increasing
temperature. This might have been caused by excess oxygen present during bonding. The

effects are shown in Figure 4-14.

 

 

 

 

 

 

Ni-SnTe-ETP 51
350
. Voltage (11V) 25 00 I . _: .3
300 Voltage (11V) 600 °C E" ..
Voltage (pv) 25 °C (after anneal) FEW
D ,.

250 - DD -
5. 200 , -95 f -
e :1.

g, 150 - of -
§ 100 - If -
50 - d
0 - M ‘

-50 i l I l

 

 

 

Position (mm)

Figure 4-14. ETP 51 bonded to Ni using SnTe coin. Contact resistance increases
with temperature with a permanent increase after cooled down to room

temperature.

This can be compared with direct diffusion bonding of p-type material to stainless steel
316 as shown in Figure 4-15 below where no SnTe interface layer was used. As with
results in Figure 4-14 the contact resistance increases with increasing temperature with a

permanent increase in contact resistance after it is cooled down to room temperature.

148

stainless steel 316- ET P 53

 

 

 

500 0 Voltage (1.1V) 25 C 1
D Voltage (pV) 400 C ' '
Voltage (11V) 25 C (after anneal) I333
400 - B3 -
300 - -F -
a J
5
3 200 - -
>
1:?
100 e c ’ Ive-0'" -
“TV‘.....
:3-*
o . 00993335306 .
I I L I I_ I I

 

 

 

2 3 4 5 6 7 8 9 10
Position (mm)

Figure 4-15. ETP 53 directly bonded to stainless steel 316. Contact resistance
increases with measurement temperature accompanied with permanent increase

after cooling down to room temperature.

A third type of behavior was observed for p-type material diffiision bonded using SnTe in
which contact resistance increases with measurement temperature but then comes back to
pre-annealed value. This is different from previous behavior where thermal cycling

resulted in permanent increase in contact resistance. As shown in Figure 4-16 the module

was cycled twice and maintained its room temperature contact resistance value.

149

Figure 4-16. ETP 52 bonded to carpenter steel using SnTe. (a) Pre-anneal room _
temperature scan (b) First anneal at 600 °C with increase in contact resistance.
(c) Room temperature scan afier first anneal. Contact resistance reduced to
pre-anneal value (d) Second anneal run at 600 °C. Contact resistance increased.

(e) Room temperature scan after second annealing run. Contact resistance reduced

to first pre-anneal value.

150

ETP 52-SnTe-carpenter steel

 

 

 

 

r~ 1 °c
120 . . ””2““ . 5 5. . .
O
100 _ scanT 25 C ..
pre-anneal run .0
80 O.
O
5‘ e
53 60 -
O
5' e
.0
20 - e‘
.e
o - M "‘t:"2"’“n"'“2
(a)
_20 I I I I I I I

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

 

 

 

 

ETP 53-SnTe-carpenter steel
bondn T ~ 615 °C
300 [ | l g I I I I
scan T ~ 500 °c .'
250 - after annealing for 24hrs .0
.0
200 - O
E .'
e _ O
E 150 .
3 ...
>
100 -
O
O
.0
50 - 2
R ~ 100 119 cm (b)
0 . W . . . .

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

151

Figure 4-16. Continued.

ET P52- SnTe-carpenter steel
bonding 1' ~ 515 °c
I l

 

 

 

 

 

 

 

 

 

140 I I
scan T ~ 25 °C
120 ' after 1st anneal at
600 C for 24 hr .
100 - .0
.0
a so - .e
8:
a 60 - e
3
>
40 - .
.0
20 - '
f ~ 2
o .. an... R 25 "9 cm
I I I I (C)
-2 0 2 4 6
Position (mm)
ETP 52-SnTe- carpenter steel
400 u o l I l .
350 _scanr~5oo c .0
second anneal at 600 °C .
300 - for 36 hr ..
O
9 250 - .’
a .-
8, 200 - e
g .0
° 0
> 150 - ’
100 .0.
O
O
50 " 2
R ~ 115 an em (d)
o I ‘1' I I
-2 O 2 4 6

Position (mm)

152

Figure 4-16. Continued.

 

 

 

 

ET P52-SnTe-carpenter steel
bonding r ~ 615 °c
1 00 I I l I .
scan T ~ 25 °C 0
80 _ after second anneal .
at 6000 for 48hrs
O
60 - '
E O
0
g 40 - .
-° 0
> . .
20 .. . O
. 2
o - eeeeeeee R~25 Mom
(6)
_20 J I I I
0 1 2 3 4

Position (mm)

153

To ascertain the viability of SnTe based diffusion bonding for LASTT material a four day
measurement with two high temperature thermal cycles was carried out. In this module
stainless steel 316 was bonded to ETP 52 using cold pressed SnTe coin. Stainless steel
316 is used because of its close CTE match (17.5 pm/m/°C) with LAST material. It has
relatively low thermal conductivity (21.5 W/m-K) and relatively high electrical resistivity
( 740 nan) in comparison to metals like Ag or Cu. Bonding was carried out at 615 °C.
As can be seen from Figure 4-17 below, the contact resistance did not increase during the

annealing run and the module successfully survived two thermal cycles.

154

Figure 4-17. ETP 52 bonded to stainless steel 316 using SnTe. (a) Pre-anneal
run with low contact resistance (b) Post anneal room temperature scan
indicating no change in contact resistance. (0) Second anneal run at 600 °C for
12 hr (d) Second anneal run at 600 °C for 24 hr. (e) Second anneal run at 600 0C
for 48 hrs (f) Second anneal run at 600 0C for 60 hrs indicating no appreciable

change in contact resistance.

155

 

350

300

250

200

150

Voltage (11V)

100

50

250

200 -

150

100

Voltage (IN)

50

$831 6-SnTe-ET P52

 

 

 

 

 

 

 

 

 

Position (mm)

156

I I I I I I I
scan T ~ 25 °C
._ pre -anneal run .0
O
O
- O
O
.0
. .0
O
O
- O
O
. 0’
O
c O
(a)
' I I I I I I
1 2 3 4 5 6 7 8 9
Position (mm)
8831 6-SnTe-ET P52
I I I I I
scan T ~ 25 °C
(RT scan after 36hr _
anneal at 6000)
F A
p .l
(b)
I I I I I
-2 0 2 4 6 8 10

Figure 4-17. Continued.

 

 

 

 

 

SS316~$nTe-ETP52
600 I I I I I
scan T ~ 600 °C
500 " (after 12 hr at sooc ...
second anneal run) ..
400 '-
3 300 a
CI
3
E 200 -
100 '-
o .
(C)
'100 I I I I I
-2 0 2 4 6 8
Position (mm)
83 316-SnTe-ETP52
500 I l I I T

400 _ scan T ~ 600 °C
24 hr second anneal
run at 600 C

300 -

Voltage (pV)
h)
0
O
I

 

 

 

O
o - d
(d)
_100 I I I, I I
-2 0 2 4 6 8

Position (mm)

157

Figure 4-17. Continued.

SS316-SnTe-ETP52
500 I l I

 

I I

scan T ~ 600 °C

400 _ annealed at 6006 for 48 hr
2nd anneal run under Ar

0)
O
O
I
’0

Voltage (11V)
‘0

N
O
O
I
e.

100 -

 

0'
“a“. (e)

 

 

-2 0 2 4 6 8
Position (mm)

88 316- SnTe-ETP 52
500 I l I I

10

 

scan T ~ 600 °C
400 - annealed at 600 °c

anneal run ..
300 - 0.

Voltage (IN)
N
8
O

100 - e.

(f)

 

'100 I I I I I l

for 60 hrs. second ..

 

 

Position (mm)

158

4.3.1.3 SnTe based hot pressed bonded modules

Hot pressed n-type materials was diffusion bonded using SnTe. Hot pressed
specimens have higher density (close to 95 % of theoretical density) and show increase in
electrical conductivity with increasing temperature. Diffusion bonded samples using
SnTe show an improvement in contact resistance with increasing temperature. Figure 4-
18 below shows plots for MSUHP 35 bonded to carpenter steel using SnTe cold pressed
coin. This behavior is unique to hot pressed samples because for east p-type diffusion
bonded samples we observed decrease in contact resistance but electrical conductivity
also decreased with temperature, whereas for hot pressed samples both electrical

conductivity and contact resistance improve with temperature.

HPMSU 35- SnTe-carpenter steel

 

 

 

 

 

 

 

 

4
3 10 I I I I
2 104 Voltage (NV) 600 00 e
'5 ' Voltage (pV) 500 °c ,' .e ‘
g .3.
Voltage (pV) 400 °C . ;.
4
s 4’ Voltage (11V) 200 °C .25 . +++
5 1 5 104 - ‘9‘ Voltage (pV) 100 °C .efJIJ ..
3: ' Voltage (IN) 25 0c; “git
a 4 .I;;:+ ,
a 1 10 '- .9:T . _ _ -
> ...;‘3: ‘ i I
9.; i: . A; ' ‘
5000 ' .24: AA}
.:i v [‘6’ .I . DD DBDEDCDC
rgg3EHEDDDCD .00....
o - eaceneeettieeWUIfi-"”"’”” -
'5000 L 4 I I
-2 0 2 4 6 8

Position (mm)

Figure 4-18. MSUHP 35 bonded to carpenter steel using SnTe. Electrical conductivity

increases and contact resistance decreases with increasing temperature.

159

A similar behavior was observed for stainless steel 316 electrode bonded to MSUHP 37.
In this case a permanent increase in hot pressed material conductivity is observed after
annealing cycle. Module started with a low contact resistance and it did not change
during annealing and maintained its value after the annealing cycle of 24 hrs as shown in

Figure 4-19 below.

stainless steel 316-SnTe-MSUHP37

 

7000 ' Voltage (uV) 25°C

 

 

 

 

 

 

 

I
‘3 Voltage (uV) 600 °C .
6000 Voltage (pV) 25 °C (after anneal) . ‘
O
5000 h e. I.
Q
.. 4000 - .' ‘
a O
z 0
a, 3000 - .' "
g 0
o 0.
> 2000 - 0 ” '
. a
o 0:43;“
1000 _ ..e _ _ L j we ..
. fail—flirt};
0 ~ —“5' ”I d
-1000 ‘ ' ' l l
-2 0 2 4 6 8 10

Position (mm)

Figure 4-19. MSUHP37 bonded to stainless steel 316 using SnTe. There is

permanent increase in electrical conductivity and contact resistance remains same

afier thermal cycle.

For hot pressed samples SnTe based results can be compared with that of diffusion
bonded with molybdenum coating. For this experiment HPMSU 35 sample legs and

stainless steel 316 electrodes were coated with molybdenum inside physical vapor

160

deposition (PVD) system. Bonding was can’ied out at 750 °C. As can be seen from Figure

4-20 below that as temperature is increased electrical conductivity also increases and

2
contact resistance decreases (fiom 1000 u!) cm2 for 25 °C to 1750 p!) cm for 600 °C).

This phenomenon is similar to what we observe for SnTe based modules except that

initial contact resistance is very high as compared to SnTe based diffusion bonds.

HPMSU 35 bonded to SS 316
SSIMo/MOIHPMSU 35

 

bonding T ~ 750 °C
1

5600 ' Voltage (pV) 600°C
0 V I v °
4800 otage(u )500 0C _ _

Voltage (pV) 400 C ‘

4000 ' Voltage (pV) 300 °c

 
   

 

 

3‘ ‘ Voltage (pV) 200 °C
1|33200 ; Voltage (pV) 100°C '
°’ ' Volta e v 25°C
% 2400 Q (l1 ) _
>
1600 - '
800 - 4
o - w 4

 

 

 

Position (mm)

Figure 4-20. HPMSU 35 coated with Mo bonded to M0 coated stainless steel
electrode. Electrical conductivity increases and contact resistance decreases with

increasing measurement temperature.

In a similar experiment MSUHP 37 was directly diffusion bonded to carpenter steel

without any interface coatings. The bonding temperature was ~ 800 °C and low contact

161

resistance bonds were obtained. Module was taken up to 450 °C for 24 hrs and no
appreciable change was observed in contact resistance. Module was cooled down to room

temperature and scanned and contact resistance value was same as before annealing as

shown in Figure 4-21 below.

 

 

 

 

 

 

 

 

’ Voltage (uV) 25 °C
6 10‘ D Voltage (pV) 450 °c .
0 Voltage (0V) 25 °C (after anneal) .
5 1o4 - . -
.0
4 Ni coated brass-prema .'
4 1° ' braze 616-ETP 53 .e ‘
a 3 104 _ bonding T ~ 430 °C .. _
o e.
5' . .'
E 2 10 l- .O ‘0‘» 0000' 1‘0 "
0009'"
1 104 - 3.06%” -
’0‘”
.30
0 - W _
_1 104 I I I I I
-2 0 2 4 6 8 10

Position (mm)

Figure 4-21. MSUHP 37 bonded to carpenter steel. Electrical conductivity
improves with increasing measurement temperature. No appreciable change in

contact resistance during or after annealing.

162

 

 

4.3.2 Prema Braze based diffusion bonded modules

A number of diffusion bonding experiments were carried out using braze as
interface between hot side electrode and LAST and LASTT material. For this Premabraze
616 was purchased from Lucas Milhaupt Inc, WI. It has a nominal composition of
(Ag61.5/In14.5/Cu balance) by weight with melting point of 620 °C. However bonding
with LAST material is carried out at around 450 °C as Premabraze 616 makes eutectic at
this temperature with LAST material (probably tellurium). First I will discuss bonding of

p-type material using Premabraze 616.

4.3.2.1 Premabraze bonded modules with p-type cast TE material

Four samples prepared with Premabraze were measured inside high temperature
voltage profiling system after bonding. First sample is Ni plated Cu electrode bonded to
ETP 53. As it can be seen in Figure 4-22 that at room temperature there is a high
resistivity region and gap of (300 uV) at the junction. The probable reason is Ag and Cu
diffusion into LAST material during bonding or CTE mismatch between TE material and
electrode or a combination of both. This behavior was observed for three out of four
samples. As the sample is taken high in temperature this voltage gap at the junction
increases about three times. This behavior is very different from bonding using SnTe
where contact resistance decreases with increasing measurement temperature. After the
module is cooled down to room temperature there is a permanent increase in the contact
resistance as voltage gap increased fiom 500 uV pre-anneal to 800 uV post anneal. All
Premabraze bonding were carried out under vacuum in a new system using rotary pump

that can reach mTorr range.

163

Ni plated Cu-Premabraze616- ETP 53 Ni plated Cu-Premabrazo616- ETP 53
bonding T ~ 430 °C bonding T ~ 430 °C

 

 

 

 

 

 

 

 

 

 

400 I I I I I I 2000 T r I r I I I
~ 0 scan T ~ 450 °c
300 Sean T 25 C M," after 6 hr anneal
" l ' 1500 - -
t e
S .
3 200 - l - E1000
0 o l- . .
5 .fi 300 uV 5 O 1500 ”v
o 100 - .0 l‘ -1 o 0
> e > 500 r 0 , .
e
0 - é.-- _ - .
(a) 0 - ...-III! (b) a
_1 00 I I I I I I I I I I I I I I
-1 0 1 2 3 4 5 6 7 -1 0 1 2 3 4 5 6 7
Position (mm) Position (mm)

Ni plated Cu—PremabrazeB‘IB- ETP 53

 

 

 

 

 

 

~ 0
1000 . b'onding T .430 'C ‘
scan T ~ 25 °c “.00“
800 _ after 450 C anneal _
for 6 hr
:6... , .
7; o-
a O 800 uV
% 400 l" . _.
> e
200 - 0 I .
e i (C)
0 - we". 1 .
-1 0 1 2 3 4 5 6
Position (mm)

Figure 4-22. ETP53 bonded to NI plated Cu using Premabraze 616. (a) Pre
anneal room temperature scan (b) Scan at 450 °C indicating increase in contact
resistance (c) Post anneal room temperature scan indicating a permanent increase

in contact resistance.

A similar behavior was observed for Ni electrode bonded to ETP 53 and ETP 5 1 using
Premabraze 616 as shown in Figure 4-23 and Figure 4-24. The junction resistance

increases with increasing temperature and the electrical conductivity drops as can be

164

observed from relative conductivity slopes of Figure 4-24, Figure 4-25 and Figure 4-26.
As the sample is cooled to room temperature, the junction resistance reduces as compared

to high temperature but still manages a higher value than pre-anneal run value.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Ni-prema braze-ET P 53 Ni-Prema braze- ETP 53
500 . ?°"d,'"9T, ‘3," c' , 1200 . midi”: 43? c a
e
scan T ~ 25 0c 1000 L scan T 450 0C .. 1
400 - pre-anneal run a after 24 hr anneal ..
800 - .. -
.C
a 5 600 . .
a 200 _ ’ 350 UV .. a 400
5 ' g . . 650 uV ~
0 l
> 100 - ’ - > 200 . l A
e .e
0 ' W— ’ ”‘— * ‘ 0 meeeeOO.——.. ' - — -
-100 l L I I I I ga) '200 l I l 1 l (b)
-1 0 1 2 3 4 5 6 7 1 2 3 4 5 6 7
Position (mm) Position (mm)
Nl-Prema braze-ETP 53
~ 0
600 fbondingT 4.30 C I
scan T ~ 25 °C
500 h post anneal scan -
400 r -
E
g 300 - .ul .
g 0
§ 200 - ' 360 uV -
100 - a I 4
0 *M———. . . (c)
0 2 10

4
Position (mm)

Figure 4-23. Ni electrode bonded to ETP 53 using Premabraze 616 ( pre anneal
scan indicating high resistance junction (b) 450 °C scan indicating increase in

contact resistance (0) Post anneal scan.

165

Ni-prema braze-ETP 51

 

 

 

 

 

 

 

bonding 1' ~ 430 °c
3000
' Voltage (W) 25 °C (after anneal) '
V I ° D
2500 otage (11V) 420 C 058 _
Voltage (pV) 25 C D; U
E C]
2000 - d a D o D .
E 1500 r J -
'3 Q C
£31000 - Gees... .
> W
500 i
I- ‘ a ‘ -1
.9'8!
Cl
C]
O - Inhiiac -
_500 l l l I 1 J
-1 0 1 2 3 4 5 6

Position (mm)

Figure 4-24. ETP 51 bonded to Ni electrode using Premabraze 616.

700

600

500

.h
C
O

300

Voltage (IN)

200

100

NI coated brass-prams braze 616

ETP 53, bonding T ~ 430 °c

 

 

 

 

 

 

 

I I I I
_ ' Voltage (pV) 25 °C -
:1 Voltage (pV) 25 °C after anneal E
,_ (Us -
J ,.
C
:5 00°“
- tr 1 .eeeO”.. -
.0
I- ... -
- m I l .
-2 0 2 4 6 3

Position (mm)

Figure 4-25. Ni coated brass bonded to ETP 53 using Premabraze 616.

There is a permanent increase in junction resistance after annealing cycle.

166

stainless steel 316-prom blue 616
ETP 51, bonding T ~ 430 °C

 

 

 

 

 

 

 

1 I I I I
1600 h o -
' Voltage (pV) 500 C
3 Voltage (11V) 25 °C ...“.
O
O
1200 _ ..." -
a".
a a".
v 0
3,800 - . .
é o
0 3:1":
> Q‘jfjffl'tfl
400 - ”33‘ -
t
o - engages-ED .
l 1 l l I
-2 o 2 4 6 8

Position (mm)

Figure 4-27. ETP 51 bonded to stainless steel 316 using Premabraze 616.

4.3.2.2 Premabraze bonded module with n-type cast TE material

Premabraze 616 seems to perform better with n-type material as compared to p-
type material. Although a high resistance junction is observed but we do not see a crack
in the leg 2 mm high from the junction as was the case with p-type material. It can be
seen from Figure 4-28 below that as the measurement temperature is increased, junction
resistance increases and then reduces back to pre anneal value after it is cooled down. For
n-type bonding fiirnace temperature is maintained at ~ 450 °C which is about 20 °C
higher than for p-type TE material with Premabraze 616. Probably Sn in p-type material

helps to make eutectic at lower temperature than for n-type material.

167

Ni coated brass-prema braze-ET N 157

Ni coated brass-Prema braze-ETN 157

 

 

 

 

 

 

 

 

' ~ ° B di T ~ 450 °C
350 . bonding T .450 TC . 1400 . ' onfi ng r 1 1 '
e o
300 P pre anneal run ... q 1200 _ scan T~450 c O. 4
scan 1' ... 25 °c ... after 24 hr anneal ..'
250 - ... 4 1000 . 0'” 240 Slcm .0. -
s ... 9 0..
5200 - .0 . 3 800 - l
a . O .0.
g 150 - , . 5 600 - ...-° .
> 100 .' voltage jump~ g 400 ..'
' 13° "V ' l 275 W
50 - W i (a) .l 200 "' ...... l (b) -i
0 I I I I I o I I I I I I I
0 1 2 3 4 6 7 -1 0 1 2 3 4 5 6 7

Position (mm)

Position (mm)

Ni coated brass-prema braze-ETN 157

350

bonding T ~ 450 °C

 

300 -
250 ~

I

1

0 ~ 950 Slcm

scan T ~ 25 °C ”
after 24 hr anneal e' 1

at 450C /

s
5200
o
a
g 150
O

.0’

O

100 r .

voltage jump ~
104 IN

50-

 

 

(C)

)-

4 l l l

 

2 4 6 8 1 0
Position (mm)

12

Figure 4-28. ETN 157 bonded to Ni coated brass using Premabraze6l6.(a) Pre

anneal scan(b) 450 °C scan indicating increase in junction resistance (c) Post

anneal scan indication junction resistance approaching back to pre-anneal value.

A similar behavior was observed for Ni electrode bonded to ETN 218 whereas previous

module was for Ni coated brass (19 x 10.6 in/K) which has significantly different

168

coefficient of thermal expansion(CTE) than Ni (13 x 10‘6 in/K). This indicates that this

behavior is independent of electrode material as shown in Figure 4-29 below.

169

Figure 4-29. Ni electrode bonded to ETN 218 using Premabraze616.

(a) Pre-anneal scan (b) 450 °C scan after 6 hr indicating increase in junction
resistance (c) 450 °C scan after 24 hr indicating no appreciable change than
6 hr scan. ((1) Post anneal scan indicating junction resistance reduced back to

pre-anneal value.

170

 

 

300

250

Voltage (pV)
01 ES 3 B
O O O O

O

Ni-Prema braze-ETN21 8

 

 

 

 

I
A

600

500 -

400

3‘3 200

Vol

100

—100

 

 

 

 

Position (mm)

171

bonding T ~ 450 °
I I I I I I h
_ o~760 Slcm ... _
scan T ~ 25 °C .'
. pre-anneal scan ..' -
- R ~ 130 on em2 -
eeeeeOO (a)
0 1 2 3 4 5 6
Position (mm)
Ni-Prema braze-ETN 218
bonding T ~ 450 °C
I I T I I T
o~3oo Slcm ,
scan T ~ 450 °C ,
' after 6hr anneal ,' ‘
_ R~175uflcm2 _
I I I I I I (b)
1 0 1 2 3 - 4 5 6

Figure 4-29. Continued.

Ni-Prema braze-ETN 218

 

 

 

 

 

bonding T ~ 450 °C
600 I l I I l l
0 ~ 350 Slcm .
500 - .
scan T ~ 450 °C 0'
400 t- after 24 hr anneal ..
S‘ -’
3 300 - .-
o ,-
0, 0
g 200 - .
a
100 - ..'
0 _ R~1tsottnem2
(C)
_100 I I I I I I
-1 0 1 2 3 4 5
Position (mm)
Ni-Prema braze-ETN 218
bonding T ~ 450 °C
400 l r l l l l
o~700 Slcm ..o'
300 ”scanT~25°C .-°
... after 24 hr anneal .°°
3200 _ at450C
o ,'
g’ .
g 100 r .'
> ' 2
0 R ~ 130 on cm
((0

 

 

 

_100 I I I I I L
-1 0 1 2 3 4 5

Position (mm)

172

Another sample prepared with ETN 218 and Ni electrode using premabraze 616 depicted
low contact resistance. However once through thermal cycle it also showed high contact

resistance as shown in Figure 4-30 below.

 

 

  
 
  

 

 

 

 

 

 

 

 

 

 

300 , "LPN"? huge-BIN 2‘18 ‘ 500 r Nl-Prema'braze-ETN'zw .
250 L ° " 90° 5’°"‘ 1 o ~ 400 Slcm ..
Scan 1 ~ 25 °c 400 - scan 1 ~ 450 °c .’ -
9 200 ' pre anneal 3C3" ‘ after 24 hr anneal ...
e150 . _ E300 . ..o' .
O O
5' 100 . . g ,o'
>° bonding r o 200 ' .° ‘
>
50 - i __ 400°C . e
100 a .. -
0 - ...—1‘ .
(a) m‘ (b)
-50 I I I I I I I 0 I I I I I I
-1 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Position (mm) Position (mm)
500 T Ni-Prema braze-ETN 2:18 I
0 ~ 700 Slcm
400 iscan T ~ 25 °C after ..I/ '
6 hr run at 4500 .0
a 300 - ." -
31 .0.
g 200 - .e .
> e
100 . l 2 .
' i R ~ 335 poem
0 I. I”. I l I I (IC) J
-1 0 1 6 7

Position (ml‘i‘I)
Figure 4-30. ETN 218 bonded to Ni electrode using Premabraze616. (a) Pre-
anneal scan indicating a low contact resistance junction (b) 450 °C scan indicating
no appreciable change'in contact resistance (c) Post anneal scan indicating

increase in contact resistance.

173

4.3.3 Low temperature solder (Cerromatrix) bonded module

The first incident of relatively high short circuit current (7 Amps) for LAST and
LASTT based modules (5mm x5mm x7 mm legs) were observed for solder based hot
side junctions. To ascertain its viability for long term module it was necessary to
investigate high temperature contact resistance for multiple heating cycles and longer
durations. For this Ni plated Cu electrode was bonded with LAST (ETN 213) using
Cerromatrix solder having a composition of 4SBi/28.SPb/14.SSn/9Sb (by weight) and
measured inside high temperature voltage profiling system (HTVPS). Melting point for
this solder is 280 to 440 °F. As shown in Figure 4-31 module was first raised to 500 °C,
brought back to room temperature and then again raised to 600 °C and cooled down to
room temperature over a two days period of time. There is increase in contact resistance
with increasing temperature and as the module is cooled down there is permanent
increase in contact resistance. As the module is cycled second time we see even a further
increase in contact resistance as shown in the room temperature scan after second
annealing. We believe that Cerromatrix makes eutectic with the LAST material at the
higher temperature and contraction during solidification of the Cerromatrix is also a
possible cause of cracking in the thermoelectric material. Another contributing factor

might be diffusion of Sn or Sn reacting with Te to form SnTe at the junction.

174

 

Figure 4-31. ETN 213 bonded to Ni plated Cu using Cerromatrix solder

(a) Pre-anneal scan (b) 100 °C scan (c)(d) Scan indicating gradual increase

in contact resistance with increasing temperature. (e) Contact resistance
increasing with temperature. (1) Post anneal scan indicating a permanent increase
in contact resistance as compared to pre-anneal scan (g) Second anneal 600 °C
scan indicating increase in contact resistance as compared to first anneal high
temperature scan (h) Post anneal scan after second run indicating after every
anneal there is permanent increase in contact resistance higher than previous

cycle.

175

Ni plated Cu-cerometrix solder-ETN 213
160 I I I l I T +-

 

140 _ scan T ~ 25 °C
pre-anneal scan

o~1100 Slcm ‘

..s

N

O
I

Voltage (pV)
co 8
O O

 

 

 

I"
60 - I -
40 - .
20 - _
(a)
0 N I I I l I -
-1 0 1 2 3 4 5 6 7

Position (mm)

Ni plated Cu-cerometrix solder-ETN 213

 

 

 

 

250 I I I I 1 I I
0 ~ 870 Slcm .0
0
20° ' scan T~100°C .' ‘
.0
A 0
>3 150 - ..' -
O
6 ,-
g 100 _ . .
o .0
> O
O
50 - .0 .
O
O
O
0 - moses (b) -
-1 2 . . 3 4 5 7
Posutlon (mm)

176

Figure 4-31. Continued.

300 Ni plated Cu-cerometrix solder-ETN213

 

 

 

o l 665 'SIcm' ' ,-
0
25° ' scan T~ 200 °c ,o‘ ‘
O
.. 200 . ..' .
> O
3 150 . ,' -
8: .'
g 100 .. .o ..
§ -'
50 l- g. -
' 2
0 _ ....... R ~ BOW-cm _
50 (c)

 

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

ONi plated Cu-cerometrix solder-ETN 213

 

I I s, I I I I ..
300 _ 0' ~ 665 cm ... _
250 - scan T ~ 300 °c ’0' -
S 200 - .o’ -
3: ...
g, 150 - . -
33 o.
3 100 - .. ..
> o
50 - 2 _
R~ 162 pQ-cm
O - w -
50 (d)

 

 

 

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

177

Figure 4-31. Continued.

500Ni plated Cu-cerometrix solder-ETN 213

0 ~ 500 Slcm

 

400 ' scanT~400°C .0 '
(annealed for 12 hrs) .0.
_ o

00
O
O

Voltage (pV)
N
O
O
'o
o

100 r- 0. ..
0‘ 2
R~167 pfl-cm

o #n - - - 5e)
-1 0 1 2 3 4 5 6 7
Position (mm)

 

 

 

Ni plated Cu-cerometrix solder-ETN 213

 

200 . . . . . . .
0' ~ 1000 Slcm .0.

150 - scan T ~ 25 °C ,0“ ‘
A (after annealing to .’
i100 _ 500 C for 5 hrs) .0. d
o .o'
E "
E 50 " I '

R ~ 95 pfl-Clflz
0 - w -
(f)

 

 

 

2 3 4
Position (mm)

178

Figure 4-31. Continued.

800

600

Voltage (pV)
J:
O
O

N
O
O

350
300
250

3200

g, 150

s

5 100

>

50

Ni plated Cu-cerometrix solder-ETN 213

 

 

 

scan T ~ 600 °C .0.

(annealed for
' 12 hrs-2nd run) ..I.
_ 0' ~.390 Slcm ’

R~ 552 .nn-cm2
- ell-Io
I I I I (g)
-2 O 2 4 6

Position (mm)

Ni plated Cu-cerometrix solder-ETN 213

 

 

'(annealed to 600 °C
_for 24 hrs-2nd run)

scan T ~ 25 °C

0 ~ 392 Slcm

0 ~ 1200 Slcm

R~140 no -cm2
~

41 I I I

 
  
 
 
 

(h)

 

2 4
Position (mm)

179

 

 

A similar behavior was observed for another Cerromatrix based module as shown in
Figure 4-32 below. In this case both Cu electrode and unicouple legs were coated with Ni
before bonding. However, this did not have a significant effect on the high temperature

scans.

 

 

Cu- Cerromatrix solder-ETNZO‘I
1 000 I I I I I I I
Voltage (pV) 300 °C

800 - ' Voltage (W) 200 °C
’ Voltage (pV) 100 °C
+ Voltage (W) 25 °C

400 - '0'
. .
’ o?‘
' Q?+ ‘
... ctr
200 - 1
¢

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

 

 

 

§

 

Voltage (uV)

 

 

 

Figure 4-32. Ni plated Cu and Ni plated ETN 207 leg bonded using Cerromatrix

solder. Contact resistance increases with increasing temperature.

180

4.3.3.1 P-type LASTT bonded using Cerromatrix solder with Cu

P-type bonding with Cerromatrix shows a different behavior than n—type and is
similar to SnTe bonding in which contact resistance reduces with increasing temperature.
For this experiment Cu electrode was bonded with ETP 52 using Cerromatrix solder. In
the first annealing run module was taken to 600 0C under Ar for five days. It was then
cooled down and again annealed to 600 °C for two days under vacuum. Successive
annealing and cooling induced permanent increase in contact resistance. As can be seen
from Figure 4-33 below that as the temperature increases, electrical conductivity drops
and contact resistance decreases. After about 12 hrs annealing at 600 °C, electrical
conductivity drops to about ~ 110 S/cm and then stays at that value for next three days.
After the module was cooled to room temperature, the electrical conductivity increased to
2300 S/cm from the 1800 S/cm pre-annealed value. However, there was permanent
increase in contact resistance after it was cooled down. We believe it is because during
cooling LASTT material, Cerromatrix solder and electrode contract at different rates due
to difference in coefficient of thermal expansion (CTE) and cause shear stresses at the
surface which results in large contact resistance. In the second annealing run under
vacuum electrical conductivity drops to ~ 110 S/cm equivalent to previous annealing run
under Ar but unlike the first run, the contact resistance does not improve with increasing
temperature. As expected, after the second annealing run the room temperature electrical
conductivity increases close to 2000 S/cm and the contact resistance also increases as

compared to previous run under Ar.

181

Figure 4-33. ETP 52 bonded to Cu electrode using Cerromatrix solder.
(a) Pre anneal scan indicating high contact resistance (b) (c) (d) Contact
resistance decreases with increasing temperature from 100 °C to 500 °C.
(6) (0 Electrical conductivity drops from ~ 500 S/cm to 110 S/cm after
12 hr anneal at 600 °C. (g) (h) Electrical conductivity remains same for
one and half days and contact resistance does not deteriorate. (i)(j )(k)(l)
Electrical conductivity remains same for four days and contact resistance
does not deteriorate. (rn) Post anneal room temperature scan indicating
permanent increase in contact resistance (n)(p) No improvement in contact
resistance with increasing temperature (q) Second post anneal room

temperature scan. Contact resistance increased two fold.

182

140 -

Voltage(|.IV)
s on on 3 R?
O O O O O

200

150

Voltage (IN)
8
O

01
0

Ni plated Cu electrode

soldered(cerometrix) to ETP5

2-II

 

scan T ~ 25 °C

' c ~ 1739 Slcm

 

I I

voltage jump across

T

I

contact ~ 40 IN

(a)

 

2

3

Position (mm)

4

5

Ni plated Cu electrode soldered

 

 

.50'

(cerometrix) to ETP52-ll
Scan T ~ 100 °C
0
_ 0 ~ 1538 Slcm ...0.
.o‘

 

Position (mm)

183

do
n .r
. voltage jump across
contact ~ 50 pV
- cocoa
l l I L l I [(b)
O 1 2 3 4 5 6

 

 

 

Figure 4-33. Continued.

Ni plated Cu electrode
250 soldered(cerometrix) to ETP52-II

 

scan T ~ 200 °C
0 ~ 1290 Slcm

200 -

Voltage (IN)
8 a
O O

01
O
I

’ voltagejump across
0 _ 3 contact ~ 25 IN

(C)
_50 I I I I I I

 

 

 

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

Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

 

 

 

 

350 l | l | l I I
O
300 _ scanT 500 C
0 ~ 714 Slcm
250 -
3:200 -
3150 .
g
o 100 -
>
50 -
0 I —‘ R~ 37pn-cm2 (d)
-50 J I I I I I I
-1 0 1

2 _ .3 4
PosItIon (mm)

184

Figure 4—33. Continued.

Ni plated Cu electrode
500 soldered(cerometrix) to ETP52-II

scan T ~ 600 °C

400 _(annealed for

1 hr at 600 °C)
0 ~ 512 Slcm

 

 
 
 
  

300

N

O

O
I

Voltage (pV)
8
O

 

 

 

_100 I I I I I I I
- 0 2 . . 3 4
Position (mm)
Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

scan T ~ 600 °C
(annealed for

12 hrs at 600 °C)
0 ~ 110 Slcm

2000

 

1 500

 
 
  

1000 -

500 -

Voltage (pV)

 

 

 

(f)
_500 I I I I I I J
-1 O 1 2 3 4 5 6 7

Position (mm)

185

Figure 4-33. Continued.

2000

1 500

A
O
O
O

500

Voltage (uV)

-500
0

1 200

1 000

0) on
O O
O 0

Voltage (uV)
.5
O
O

200

Ni plated Cu electrode

soldered(cerometrix) to ETP5

2-II

 

scan T ~ 600 °C
(annealed for 24 hrs

I at 500 °C)

 

o ~114 Slcm

I I I

I

O

c’
I

I

(g)

 

2 3

4

Position (mm)

5

Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

6

 

 

scan T ~ 600 cc

at 600 °C)
0 ~ 127 Slcm

- (annealed for 36 hrs

I

I
a

I

(h)

 

2 3

4

Position (mm)

18

6

 

 

Figure 4-33. Continued.

Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

 

 

 

 

1000 -scan T~ ~600 °C .
(annealed at .0
800 -600 0C for 48 hrs) 0..
9 0 ~ 1 25 Slcm 0'
g 600 r ’
g Q
a: 400 - ...
> 0'...
200 -
f
’0
0 I“ (1)
o 1 4

Position (rim)

Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

 

scan T~ ~600 °C
1000 -(annealed at
600 °C for .
- 3 days) ,1,“

c~127 Slcm ,1;

Voltage (IN)
on co
8 8

.p.
8

 

N
8
‘5

(i)

O

 

 

2 3
Position (mm)

187

Figure 4-33. Continued.

Ni plated Cu electrode

 

soldered(cerometrix) to ETP52-ll

 

 

 

 

 

 

 

1000 - scan T ~ 600 °C ,
(annealed at 600 °C .g’
800 - for 4 days) o/
/
3 ./
a 600 _0'~127 Slcm a
o ./
E .-
3 400 *- e/
> .o/
200 - 7/
1/‘I
0 - ’ ‘ (k)
-1 1_ , 2 3
Posltlon (mm)
Ni plated Cu electrode
1200 soldered(cerometrix) to ETP52-ll
scan T ~ 600 °C
1000 - (annealed at 600°C f.
for 4-1 I2 days) /
800 -
E 0' ~ 119 Slcm ’/
0 ,
g 600 - /‘.
§ 400 - j
9,,
200 - 9’
0 - 3“ . . . <1)
-1 0 1 2 3

Position (mm)

188

Figure 4-33. Continued.

Ni plated Cu electrode
soldered(cerometrix) to ETP52-ll

 

 

250 - Room Temp scan after I a
annealing at 600 °C for
200 rtive days -
E
v 150 - 0 ~ 2300 Slcm -
81
g 100 - voltagejump across 4
> . contact ~ 175 IN
50 - -
e
0 I. l I I I I I (m)I .

 

 

 

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

Ni plated Cu electrode

 

soldered(cerometrix) to ETP52-ll

I

scan T ~ 600 °C
second run-annealed

at 600 °C for 12 hrs
o~117 Slcm

 
 
  
 

N
O
O
O
I

IN)
a:
o
O

 

 

 

_500 I I I '4 I I I
-1012 3 4 5 6

Position (mm)

189

Figure 4-33. Continued.

Ni plated Cu electrode
soldered(licerometrix) to ETP52- II

 

  
 
  

 

 

 

 

 

 

 

 

2000 _ scan T~ ~5'oo °c' ' 4
second run-annealed
1500 - at 500 °c for 24 hrs -
31000 - -
g, 0 ~ 133 Slcm
g 500 - -
o
>
O - III- ..
(p)
_500 I I I I I
1 2 3 4 5 6 7
Position (mm)
Ni plated Cu electrode
2500 soldered(cerometnx) to ETP52-ll
scan T ~ 25 °C
2000 - "I- -
E 1500 ' 5 ~ 2000 Slcm
on
011000 - -
.3;
o
> 500 - d .
0 - .5 -
(q)
_500 I I I I
-2 O 2 4 6 8

Position (mm)

190

4.4 Energy Dispersive Spectroscopy (EDS) analysis of fractured surface

The sample was then fractured to investigate Cu diffusion through EDS analysis.

Dark
region

 

Figure 4-34. Fractured sample leg indicating dark and shiny region.

As can be seen from Figure 4-37, there are two clear regions of dark and shiny silver
color on both sides of fractured leg. EDS was carried out on the sample but we were not
certain about the results as it was indicating abnormal values of constituent elements. We
again annealed another sample under same conditions for approximately 18 hrs. Figure 4-
35 shows the broken sample part with both dark and shiny regions. EDS results for this

sample are shown in Table 4-1.

191

  
   

 
   

Shiny
region

Figure 4—35. SEM picture of new p-type leg.

Table 4-1. EDS results for dark and shiny regions indicating Cu and oxygen rich

regions.

atomic %
9

 

192

Conclusion

A new high temperature voltage profiling system has been developed to investigate
contact resistance at ~ 800 K. A number of SnTe, Premabraze and Cerromatrix based
modules were tested for contact resistance. Among these SnTe based p-type bonds
showed best results in terms of low contact resistance at high temperature. For n-type
material there is variation in results that include good results for hot pressed n-type using
SnTe and direct bonding of hot pressed with stainless steel. Soldered bonds initially show
promising results but contact resistance increases after thermal cycling. Based on these

results SnTe is better option for both type of LAST thermoelectric material.

193

Appendix A: Nanowire Synthesis

Introduction

My initial research efforts were focused on GeOz nanowire synthesis for sensor
applications. Nanowires were grown using the vapor-liquid-solid (V LS) mechanism. This
idea was proposed by Ellis and Wagner in 1964 [45]. In the VLS mechanism a liquid
catalyst droplet sits on top of a nanowire and absorbs semiconductor material as the
catalyst supersaturates, the semiconductor material drops out to form another layer of
nanowire. This process continues with the supply of base material either through a gas

source or fi'om surface diffusion of atoms. The catalyst droplet also involves some

amount of base material that makes a eutectic; Ger with melting point of 361°C. There

is also evidence that nanowires grow by vapor-solid-solid (V SS) mechanism [46]. If the
growth takes place below the eutectic temperature, then the catalyst (usually Au) will
remain a solid. Nanowire grown through this mechanism usually have thinner diameter.
Whether nanowires will grow through VLS or VSS mechanism in a particular experiment

depend upon the grth temperature, vapor pressure, source element and catalyst used.

[46]. I will discuss nanowire synthesis setup followed by high temperature GeOz and

ZnO synthesis and finally low temperature synthesis of Ge02 nanowires.

A.1 Experimental setup

In our lab the experimental arrangement consists of a quartz tube and horizontal
tube furnace which is a hollow cylindrical shape with resistive heating coils covered in

thermal insulation. Furnace has outside diameter of 8 inches, inside diameter of four

194

inches and is 12” long. A 3.35” diameter quartz tube extends through the cylindrical
furnace for uniform heating of samples as shown in Figure A-l. The two ends of the
quartz tube have stainless steel end caps with double ‘0’ rings. One end of the tube is
connected to a rotary pump which can create vacuum less than 100 mTorr in the tube.

The other end is connected to two gas supply lines through mass flow controllers ('MFC)

for control of Ar or 02 supplies. Type ‘K’ thermocouple is used to measure the furnace

temperature. In a typical arrangement source material is placed on a 2.5” x 6” quartz plate
at the upstream end and a substrate for nanowires is placed few inches downstream inside
the furnace heating zone. The arrangement has also been made to place a separate
thermocouple inside the quartz tube to know more accurately the substrate temperature
during growth. As a result of the low pressure environment in the tube there is always a
temperature difference between furnace temperature outside the tube and actual substrate

temperature inside the tube.

 

Figure A-l. Nanowire synthesis setup.

195

A.2 Germanium Oxide (GeOz) nanowires

Oxide NWs have been found to have different bandgap than bulk oxides [47] e. g.
bulk CuO has bandgap of 1.85 eV and CuO nanowires were measured with 2.05 eV
bandgap. Nanowires are better candidates for sensors because of higher surface to

volume ratio which allow larger surface activities and accumulation of charge species

which can modify transport properties [48]. Germanium dioxide (GeOz) nanowires have

been reported with a photoluminescence signal an order of magnitude higher than GeO
powder [49,50]. They also have higher refi'active index (n=1.63) than silicates [51] which
makes them attractive for optoelectronic applications and they exhibit a bandgap of 2.44

eV. This material can be used as optical waveguides [52] and our research group has

investigated GeOz nanowires as substrate for surface enhanced Raman Spectroscopy

(SBRS) [53].

A.2.l Experimental details

In this lab nanowires were grown on silicon (110) and quartz substrates. The
substrate was cleaned in acetone, then methanol and DI water, all through sonication.
Three different types of catalyst were tested. A gold film in the range of 2 to 15 nm was
either sputtered deposited or eVaporated using e-beam evaporator. Gold colloidal
particles on silicon substrate have also been used as catalyst. To deposit the gold collides,
the silicon substrate is first coated with poly L lysine which is hydrophilic for Au

particles. This helps the gold particles to stick and to spread uniformly on substrate.

196

Platinum (Pt) wire was also used successfully as catalyst. The substrates were placed on
quartz plate and inserted into the quartz tube within hot furnace zone. Source material of
pure germanium was placed approximately four to five centimeters upstream. The quartz
tube was evacuated to less than 100 mTorr and an argon/hydrogen gas (95/5) was
introduced at 50 sccm. The pressure inside the tube was raised to about 100 to 150 Torr
and the tube was then heated to 850 °C for about 30 minutes. After this a 60% Ar and
40% oxygen mixture was introduced through multi function control (MF C) unit. An
alternate method is to switch off the mechanical rotary pump and open the valve to allow
air to rush in the vacuum tube. The tube was then heated for about 10 to 15 minutes. The
firrnace was switched off and tube was allowed to cool off under Ar flow. A white wool-
1ike grth was observed on the silicon or quartz substrate. SEM analysis showed that
more dense nanowires were obtained if substrate was covered during the process. It may
be because of the increased vapor pressure at the point of growth. Higher growth is
observed at the edges of the substrate. This higher growth is known as edge effect where
due to surface roughness of edges more source material is trapped resulting in dense
growth [54]. Nanowire diameter ranges from 40 nanometers to micron size. Energy
dispersive spectroscopy (EDS) analysis confirmed that atomic ratio of germanium to
oxygen is close to 1:2. Some of the SEM images show gold particle at the end of
nanowires which indicate that growth is taking place through VLS mechanism. The kink
in nanowires as shown in Figure A-2 can be result of stacking faults and dislocations

[5 5]. Lugstein et al have explained kink in Si nanowires by overall change in pressure
during growth [56, 5 7]. We believe change in pressure is more appropriate reason in our

case as we suddenly change pressure by opening valve to inlet oxygen at high

197

 

temperature. This, however, is not the case every time. Straight nanowires have been

obtained by allowing mixture of oxygen/ argon gas in the chamber at high temperature.

10 pm

 

Figure A-2. GeOz Nanowires SEM image.

 

Figure A-3. EDS spectra indicating Ge and 0; ratio.

198

A.3 Hotplate Synthesis of GeOz nanowires

Several routes of GeOz nanowire synthesis have been reported including physical
evaporation [58], laser ablation [59], thermal oxidation [60], carbon nanotube confined
reaction [61] and thermal annealing of Ge under an oxidizing environment [62]. These
synthesis procedures require heating above the sublimation temperature of GeO (710 °C)
to generate sufficient vapor pressure of source material. Nanowire growth can occur
rapidly under these typical growth conditions and for added control, vacuum growth
chamber are commonly used which might include process gas control instrumentation,
vacuum pumps and furnaces (internal or external to the chamber). We have developed a
low cost, low temperature, hot plate nanowire synthesis method with minimum process

variables for good repeatability.

A.3.l Experimental procedure

A hotplate capable of reaching ~ 350 °C temperature was used to grow G502

nanowires on the following substrates: quartz, alumina, AlN and Si wafers each coated
with 9 to 28 nm of gold. Hotplate setup is shown in Figure A-8. To fabricate the
nanowires a gold (Au) coated substrate was placed on a hotplate with the Au coating
facing upward and a small piece of an undoped Ge wafer was placed on top of the
substrate. Aluminum foil was used to cover the sample on the hotplate taking care not to
touch the samples with the aluminum. The hotplate was then heated to ~ 350 °C for about

2 hrs then switched off and allowed to cool down to room temperature before the

199

substrate was removed. Through an optical microscope, a white wool-like growth could

be observed on the substrate.

A.3.2 Results and Discussion

The product was characterized by JEOL 6400 SEM, EDS (Oxford Instruments),
J EOL 3010 TEM, Powder X ray diffraction (Rigaku), and Raman Spectroscopy (EZ
Raman). Nanowires were tens of micron long and ~50 to 200 nm in diameter. For TEM
analysis a Cu TEM grid (Pelco 100 mesh with lacy carbon) was placed on an AlN piece
having nanowires on its surface and the TEM grid was moved several millimeters to
scrape some of the nanowires onto the grid. The obtained diffraction pattern during TEM
imaging is shown in Figure A-4. After two minutes of imaging, the diffiaction pattern
was lost which is in agreement with what has been reported elsewhere [5 8] and can be

attributed to unstable and sensitive nature of nanowires under electron irradiation. SEM

images of GeOz nanowires grown on aluminmn nitride (AlN), alumina (A1203), quartz,

and silicon (Si) wafer are shown in Figure A-S. It can be clearly seen that growth on the
Si wafer and quartz are sporadic whereas on alumina and aluminum nitride there is
uniform layer of nanowires. A closer look reveals that these later substrates have thin
wires with more uniform morphology. We believe it is because alumina and aluminum
nitride have rough surfaces providing more edges'for a dense growth. EDS results in

Table A-1 shows the ratio of germanium and oxygen (24:71) confirming that nanowires

are GeOz. However, EDS is susceptible to some error for oxygen concentration. Rarnan

peak in Figure A-6 indicates that nanowires are trigonal in structure.

200

A.3.3 Growth Mechanism
In Addition to GeOz we also synthesized ZnO nanowires using a hotplate which

grew only on the zinc source and any effort to grow nanowires on catalyst coated
substrate was not successful. This can be explained by self catalytic nucleation and tip

growth process [63] which clearly is not a VLS growth mechanism. Whereas in the case

of G602, nanowires could be synthesized only on a catalyst coated substrate and not on

the source itself. It appears that a catalyst is necessary for low temperature grth of

GeOz because melting point of Ge is 915 °C but melting point of Au-Ge eutectic is 356

°C (at 28 at.% Ge) [64] which is well below the sublimation point of germanium. There

are previous reports of hotplate metal oxide nanowire growth. Fook et a1 [65] synthesized

WOx nanowires on a hotplate by heating a tungsten foil and collecting nanowires on a

glass slide without any catalyst. They have proposed the solid-vapor-solid (SVS) growth

mechanism for this synthesis process. Yanwu et al [63, 66] reported metal oxide NW

including ZnO, CuO, a-Fe203 and C00,; nanostructures through direct heating of the

source material on a hotplate without using any catalyst. They have ascribed the solid-

liquid-solid (SLS), self catalytic nucleation and tip growth mechanism for their

nanowires. For the GeOz nanowires we have investigated we believe our synthesis

mechanism is different from these reports because of the Au catalyst presence. We
believe it is VLS mechanism where first GeO evaporates from Ge source. GeO then

travels in vapor form to the substrate and forms eutectic with Au droplet. The captured

201

GeO combines with oxygen fi'om air to form the GeOz nanowire. Because of close

proximity of source and catalyst the GeO vapors quickly supersaturate the Au droplet
which is necessary for crystalline structures [67]. No growth was observed when the
substrate was placed farther from the Ge source which indicates the GeO rapidly

condenses out of the gas phase. This could be caused by low vapor pressure of the GeO,

and/or the time to formation of GeOz in the gas phase. Nanowire growth is only

observed when the source is in intimate contact with catalyst coated substrate.

It may also be possible to have simultaneous oxide assisted growth. Lee and
coworkers [68] proposed oxide assisted growth (OAG) for semiconducting NWs with no
metal catalyst present. Under OAG, the oxides serve as catalyst and this mechanism can
co-exist with the VLS mechanism in same synthesis experiment. TEM (Figure A-7)
shows a metal catalyst at the tip of most nanowires; however the same metal catalyst is
not visible for all the NWs observed. It is possible that the OAG mechanism coexists
with VLS mechanism or that the gold catalyst is absorbed into the nanowire during
growth or that the nanowires without Au tips are broken wires. Allesandro et al [69]

heated a Ge wafer to investigate the formation of sub oxides and found that at low

temperatures sub stoichiometric Ger is formed which is unstable and forms a poor

interface with Ge. As the temperature is increased G602 becomes the dominant species

up to 300°C, for temperatures greater than 300°C the oxidation rate starts decreasing and

beyond 400°C the amount of GeOz reduces and that of sub-oxide species increases. As

202

our experiments were for a substrate temperature near 350°C, Ger is expected, possibly

clustering on the substrate and serving as nucleation sites for oxide assisted growth.

 

Figure A-4. (a)TEM diffraction pattern (b) after two minutes of irradiation
diffraction pattern is lost and NW turns amorphous.

203

 

Figure A-5. SEM images of GeOz NWs on Au coated (a) Aluminum nitride (AlN)

surface (b) Aluminum oxide( AL203) surface. Wires are straighter (needle-like) and

more dense (c) quartz surface, nanowires have non uniform thickness (d) silicon (Si)

substrate, nanowires are sparse and irregular.

204

Table A-1.

EDS results indicating Ge and 0 ratio.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Element App Intensity Weight% Weight% Atomic%
Conc. Corm. Sigma
0 K 0.25 0.9644 31.35 4.44 71.80
Ge L 0.28 0.6883 48.42 4.58 24.44
Au M 0.11 0.6635 20.22 5.06 3.76
Totals 100.00
2500 , 1 1 1
438
2000 — _
1
E1500 — i —
fl .
5
LE ‘ [I
1000 ' a 1 —1
{1 .1 WI
500 ‘ .1 I 1* MMWWJ ~
Ni
O 1 l l 1
0 500 1000 1500 2000 2500

Wavenumber(c m-1 )

Figure A-6. Raman spectra showing GeOz NWs are trigonal.

205

 

 

Figure A-7. TEM image showing Au eutectic.

Ge wafer piece
substrate

Hot plate 0 O

 

 

 

Figure A-8. Hotplate setup for nanowire synthesis.

206

 

1200 — 101 102 J
1000 ~ ‘ —

800 ~ I

 

Intensity

600 — z ‘ .

400 ~
I

 

 

 

50 60 70

Figure A-9. Powder X-ray diffraction spectrum indicating GeOz peaks.

A.4 Indium Antimonide (InSb) Nanowires Synthesis

4
Irer is a direct bandgap semiconductor (0.18eV) with high mobility (8x10

2 . . .
cm N -s) [70]. It is a good candidate for detector arrays operatrng 1n infrared wavelength

[71]; magneto resistive sensors [72] and high speed electronic devices [73]. InSb
nanowires have been studied extensively due their high mobility, direct bandgap, phase
change nature and expected high figure of merit (ZT) for thermoelectric applications

[74]. Figure of merit can be defined as

ZT=T S ZU/K.

Here S is Seebeck coefficient; T is temperature, ais electrical conductivity and K‘ is

thermal conductivity. For higher ZT, materials should have higher electrical conductivity

207

and lower thermal conductivity. Phonons are responsible for heat transfer in
semiconductors at the higher temperature range and theoretically it was predicted that
one-dimensional structure has higher phonon scattering with its dimensions becoming
comparable to phonon mean free path [75]. This helps to reduce thermal conductivity and
improve the thermoelectric efficiency of one dimensional nanowires. High mobility of
electrons in Irer makes it an ideal candidate for high-speed devices.

Various synthesis methods have been reported in the literature. Electro-
deposition, chemical vapor deposition (CVD), pulsed laser deposition (PLD) and physical
vapor deposition (PVD) are widely used techniques. For this experiment, the vapor-
liquid-solid (VLS) mechanism using a PVD technique was used to grow InSb NWs. As
described in the beginning of this chapter the VLS mechanism involves a metal catalyst
which serves as a seed for NW growth. We used Au thin film (10 nm) as catalyst for InSb
nanowires. Pure InSb (99.99 %) source was placed in a quartz boat and thin Au coated Si
(100) substrate was placed upside down on top of the quartz boat. It was placed inside a
quartz tube fumace and evacuated to mTorr range. The Ar was allowed to flow into the
chamber at 20 sccm. The temperature was maintained at 650 °C for about 1 hr and then
allowed to cool down to room temperature. Nanowires were collected into solution by
sonicating the substrate. A micropipette was used to place a drop of isopropanol on lacey
carbon film coated Cu grid for TEM analysis. Selected area electron diffraction (SAED)
patterns revealed that nanowires are single crystalline zincblend structure and the lattice
constant is 6.7 A which is very close to literature value of 6.48 A. Energy dispersive
spectroscopy (EDS) was carried out to determine the constituent elements. Nanowires

had oxidized, but it was unknown whether, they absorbed oxygen during synthesis or

208

only an outer layer of atmospheric oxygen developed after taking out the sample from
tube fumace. Trace oxygen in the system could also be a possible source of the nanowire
oxidation. The nanowires were tens of micron long and ranged in diameter from 100 to
200 nm. The good InSb material should have In and Sb ratio of 1:1. Whereas, EDS
showed that our nanowires were 49:1 ratio and could not be used for thermoelectric
applications. However, they can still be used for gas sensor applications where Sb can be

considered as doping to InO NWs.

.. y ‘K \
v ‘3“. (luv b H.» ‘5. D

   

- In
.. ‘ 4

'11». “
.- '3‘.

  
   
  

I

   

Figure A-l0. InSb Nanowire SEM image.

Table A-2. EDS results indicating In and Sb ratio.
%

 

209

 

Figure A-ll. InSb NW with a diameter of ~200 nm.

d =0.6772 nm
29 = 0.210
r =3.7O mm

 

Figure A-12. InSb SAED showing zincblend structure.

210

A.5 Catalyst Assisted Zinc Oxide (ZnO) Nanowires Synthesis

ZnO nanowires are a primary candidate for photonic applications including UV
detectors and lasing. These are wide bandgap (3.7 eV) semiconductors showing both
piezoelectric [76] and thermoelectric properties [77]. It is suitable for optoelectronic
applications because of wide band gap and is transparent to visible light. ZnO naowires
grow in a variety of shapes including nanobelts, nanosprings, nanoshells, nanocombs and
nanocages [78]. A number of synthesis techniques have been reported in literature
including chemical vapor deposition [79], metal oxide vapor phase epitaxy [80], physical
vapor deposition [81], template assisted growth [82] and laser ablation [83].

In this lab zinc oxide nanowires were grown using the VLS mechanism in the
quartz tube furnace. Zinc powder was placed upstream at a distance of 4-5 cm from the
substrate. The substrate was cleaned with acetone, methanol and DI water using
sonication. A thin gold film was deposited by thermal evaporation or sputtering. The tube
was evacuated to less than 100 mTorr and then Ar 99.99% was introduced in the tube to
bring the pressure to ~ 10 Torr. The tube was heated up to 650 °C for about an hour. The
furnace was then switched off and allowed to cool down. White wool like substance was
obtained on the substrate and the source material turned from gray to whitish in color.

SEM analysis showed that length is in tens of micron and diameter range is 100-200 nm.

211

 

Figure A-13. ZnO nanowires grown in tube furnace.

 

Table A-3. EDS result indicating Zn and 0 ratio.
Element App Intensity Weight Weight% Atomic%
Conc Corrn. % Sigma
0 K 2.36 0.7236 4.42 0.45 9.19
Si K 57.99 1.1429 68.82 0.96 81.53
C1 K 1.49 0.6652 3.03 0.19 2.85
Zn K 4.57 0.8702 7.12 0.79 3.62
Au M 6.82 0.5565 16.61 0.83 2.81
Totals 100.00

 

 

A.6 Catalyst Free Zinc Oxide (ZnO) Nanowires Synthesis

The hot plate synthesis approach provides a low cost method compared to the

quartz tube synthesis process described in the previous section. With hot plate we do not

212

need vacuum system or gas control for nanowire synthesis. For this particular case NWs
are grown without using any catalyst which also helps to reduce cost. Pure Zn (99.99%)
30-mesh was placed in custom made quartz boat, covered from all sides but not sealed,
and placed on top of a hot plate under atmospheric conditions. The material was heated at
480 °C for 2 hrs. Densely packed uniform ZnO NWs grow axially on Zn source. Wires
are < 100 nm in diameter and few micron in length. First attempt was made to sonicate
NWs in ispropanol and a pipette was used to place a drop on a Cu grid, but that did not
prove to be successful. Then Zn source was left inside vial containing is0propanol and
sonicated again. This time few wires landed on grid. Also, wires were scratched from
surface directly onto a Cu grid. However, this increased the probability of adding
contarninations to sample. There are other reports of such oxide nanowire growth on
hotplates [84], [63] but it has been stated that due to high vapor pressure and low melting
point, control of ZnO nanowires under ambient condition growth is challenging [63].
Whereas, we have been able to grow dense NWs repeatedly through hot plate process. It
is assumed that nanowires grow by vapor-solid (VS) mechanism [84] and it can also be
explained by self catalytic nucleation [63]. Zinc has melting point of about 420 °C. When
hot plate temperature reaches this value, zinc melts and oxidizes immediately under
ambient conditions. Since the melting point of zinc oxide is much higher, it solidifies and
serve as nucleation site for further growth [63]. The usual VLS mechanism is ruled out
because the typical indicator of VLS mechanism is a catalyst (Au) tip at the end of

nanowire, whereas, no catalyst is used for hotplate nanowire growth.

213

 

Figure A—l4. ZnO NWs grown on hot plate.

 

 

 

Figure A-lS. EDS spectra for ZnO nanowires.

214

 

 

Figure A-l6. TEM selected area diffraction image.

 

Figure A-l7. ZnO NW length ~ 6 pm.

215

Conclusion

Oxide nanowires were synthesized using two different arrangements. GeOz, InSb and

ZnO nanowires were synthesized in tube firmace using the VLS mechanism. ZnO and

G602 nanowires were also synthesized using a hotplate under ambient conditions.

Hotplate ZnO nonowires follow the self-catalytic growth model and GeOz nanowires

follow VLS growth mechanism. We suspect that oxide assisted growth mechanism may

also be simultaneously present with VLS mechanism for GeOz nanowire hotplate

synthesis. To the best of our knowledge it is the first instance of hotplate Ge02 nanowire

growth.

216

 

[1]

[2]

[3]
[4]

[5]

[6]

[7]

[8]

[9]

[10]

References

Kuei Fang Hsu, Sim Loo, Fu Guo, Wei Chen, Jeffrey S. Dyck, Ctirad Uher, Tim
Hogan, E. K. Polychroniadis, and Mercouri G. Kanatzidis, “Cubic
Abe(m)SbTe(2+m) bulk Thermoelectric Materials with High Figure of Merit”,
Science, vol. 303, pp. 818-821, 2004.

VA Drebushchak, “The Peltier Effect”, Journal of Thermal Analysis and
Calorimetry, vol. 91, pp. 311-315, 2008.

D.M Rowe, CRC Handbook of Thermoelectrics. Boca Raton: CRC Press, 1995.

J. Short, Sim Loo, S La], K F Hsu, E Quarez, M.G Kanatzidis, T Hogan, “Hall
Effect Measurement on New Thermoelectric Materials”. Mat. Res. Soc. Sym.
proc, vol. 793, pp. 323-332, 2004.

Douglas T Crane, “Potential Thermoelectric applications in diesel vehicles”. In
Proceeding of the Ninth Diesel Engine Emission Reduction (DEER) conference.
August 23-24, 2003. (available at http://www.bsst.com/technica1 _papers).

HG Goldsmid and G S Nolas, ”A review of the new thermoelectric materials,” in
Proceedings of the IEEE 20:}: International Conference on Thermoelectric, 2001,

pp. 1-6.

C M Bhindari, “Minimizing the thermal conductivity”. in CRC Handbook of
Ihermoelectrics. Ed. D M Rowe. CRC Press: Boca Raton, FL, 1995.

D. Bilc, S D Mahanti, Eric Quarez, Kuei-Fang Hsu, Robert Pcionek and M G
Kanatzidis, “Resonant States in the Electronic Structure of the High Performance
Therrnoelectrics AngmeTe2+mz The Role of Ag—Sb Microstructures”, Phys. Rev.
Lett. vol. 93, p. 146403, 2004.

John Androulakis, Kuei Fang Hsu, Robert Pcionek, Huijun Kong, Ctirad Uher,
Jonathan J. D'Angelo, Adam Downey, Tim Hogan, and Mercouri G. Kanatzidis,
“Nanostructuring and High Thermoelectric Efficiency in p-Type Ag(Pb1_
ySny),,,SbTe;_;+,,,. ”, Adv. Mater. vol. 18, pp. 1170—1173, 2006.

Jane Y. Howe, Hsin Wang, and Jihui Yang. “Structure of Polycrystalline
Thermoelectric Bulk Material AngmeTe2+m”. Microsc Microanal, 13(Suppl 2),
2007.

217

 

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

E. Quarez, K F Hsu, R Pcionek, N Frangis, E K Polychroniadis, and M G
Kanatzidis, “Nanostructuring, Compositional Fluctuations, and Atomic Ordering

- in the Thermoelectric Materials AngmeTe2+m. The Myth of Solid Solutions” J.

of Am. Chem Soc. vol. 127(25), pp. 9177-9190, 2005.

Ono, T.; Takahama, T.; his, T, “The Thermoelectric Properties of AngTez-
AgBiTez, -PbTe and -SnTe Systems”, J. Phys. Soc. Jpn. vol. 17, p. 1070, 1962.

Fei Ren, “Mechanical Characterization and Powder Processing of Lead Telluride
based Thermoelectric Materials”, PhD Dissertation, Deparrnent of Material
Science and Chemical Engineering, Michigan State University, 2007.

Atsuko Kosuga, Masayoshi Uno, Ken Kurosaki and Shinsuke Yamanaka,
“Thermoelectric Properties of Stoichiometric Ag1_ Pblgstezo (x = 0, 0.1, 0.2)”,
Journal of Alloys and Compounds, vol. 391, pp. 288-291, Apr 2005.

Heng Wang, J ing-F eng Li, Ce-Wen Nan, Min Zhou, Weishu Liu, Bo-Ping Zhang,
and Takuji Kita. “High-performance Ago,3Pb13+beTe20 thermoelectric bulk
materials fabricated by mechanical alloying and spark plasma sintering”, Appl.
Phys. Lett. vol. 88, p. 092104, (2006).

Bradley Hall, “Powder processing, powder characterization, and mechanical
properties of LAST and LASTT thermoelectric materials”, Masters Thesis,
Deparrnent of Material Science and Chemical Engineering, Michigan State
University, 2008.

G. Wang, W. Pan, M. N. J iang, J. Chen, and Y. M. Luo, “Investigation of the
physical and mechanical properties of hot-pressed machinable Si3N4/h-BN
composites and FGM”, Mater. Sci. Eng. B. vol. 90, p. 261, 2002.

IA Drabkin, L B Ershova, Energy Climate Technology (EC 7) conference, Bergen
April 17-18, 2008, P1-09-1.

O Yamashita, H Odahara, and S Tomiyoshi, “Effect of metal electrode on
thermoelectric power in bismuth telluride compounds”, J. Materials Science, vol.
39, pp. 5653-5658, 2004.

Peter Thomas and Houldcroft, Welding Process Technology, Cambridge:
Cambridge University Press, 1977.

M G Nicholas, Joining processes: Introduction to Brazing and Diflusion bonding,
The Netherlands: Kluwer Academic Publishers, 1998.

Giles Hurnpston and David M Jacobson, Principles of Soldering, ASM
International Publications, 2004.

218

 

I

 

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[311

[32]

[33]

[34]

[35]

Jonathan D Angelo, “Low resistance contacts to Thermoelectric Materials”,
Masters Thesis, Deaprtment of Electrical and Computer Engineering, Michigan
State university, 2006.

N C Miller, “Method of manufacturing a bonded electrical contact for
thermoelectric semiconductor element”, US Patent 3879838, Apr 29, 1975.

L F Kendall, and J H Bredt, “Diffusion barrier for semiconductive thermoelectric
generator elements”, US Patent 3650844, Mar 21, 1972.

M. Weinstein and A. I. Mlavsky, “Bonding of Lead Telluride to Pure Iron
Electrodes”, Rev. Sci. Instrum. vol. 33, p. 1119, 1962.

T R Krebs, “Stable bonded barrier leyer-Telluride thermoelectric devices”, US
Patent 3859143, Jan 7, 1975.

M P Ravdel and O I Evdokimova, “Alloys With a High Coefficient of Thermal
Expansion Based on the Mn-Pd System”, Metal science and heat treatment, vol.
16, pp. 403-405, 2004.

TP Hogan, MRS Spring meeting April 13-17, 2009, San Francisco, CA.

MF Dumke, T A Tombrello, R A Weller, RH Housley, and EH Cirlin, “Sputtering
of the gallium-indium eutectic alloy in the liquid phase”, Surf Sci. vol.124, pp.
407-422, 1983.

Wade, K.; Banister, A. J. The chemistry of aluminum, gallium, indium and
thallium, vol. 12; Pergamon texts in inorganic chemistry; Pergamon Press: New
York, 1973.

D Zmic, and D S Swatik, “On the resistivity and surface tension of the eutectic
alloy of gallium and indium”, J. less—common met. vol. 18, pp. 67-68, 1969.

S J French, D J Saunders, and G W Ingle, “The system gallium indium”, J Phys.
Chem. vol. 42, pp. 265-274, 1938.

C. P. Rhodes, J .W. Long, M. S. Doescher, J. J. Fontanella, D. R. Rolison,
“Nanoscale Polymer Electrolytes: Ultrathin Electrodeposited Poly(Phenylene
Oxide) with Solid-State Ionic Conductivity”, J. Phys. Chem. B. vol. 108, pp.
13079-13087, 2004. '

Ryan C. Chiechi, Emily A. Weiss, Michael D. Dickey, and George M.
Whitesides, “Eutectic Gallium-Indium (EGaIn): A Moldable Liquid Metal for
Electrical Characterization of Self-Assembled M’onolayers”, Angew. Chem. vol.
120, pp. 148-150, 2008. .

219

 

 

 

 

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

Jonathan D Angelo, “Electrical contact fabrication and electronic characterization
of thermoelectric materials”, PhD dissertation, Department of Electrical and
Computer Engineering, Michigan State University, 2009.

LB Cadoff, and Edward Miller, Thermoelectric Material and Devices, New York:
Reinhold Publishing Corporation, 1960.

Ronald S Tirnsit, “Electrical contact resistance: Fundamental principles” in
Electrical Contacts: principles and applications, Paul G Slade, Ed. New York:
Marcel Dekker Inc, 1999. pp. 1—72.

Kortier, William E Muller, John J Eggers, Philip E, “Thermoelectric Module”,
US Patent 4211889, Jul 8, 1997.

Belson H S, and Houston B, “Thermal expansion of tin-telluride”, J Appl. Phys.
vol. 41, pp. 422-424, 1970.

A Singh, S Bhattacharya, C Thinaharan, D K Aswal, S k Gupta, J V Yakhmi, K
Bhanumurtlri, “Development of low resistance electrical contacts for
thermoelectric devices based on n-type PbTe and p-type TAGS-85
((AngTeflo,15(GeTe)o_35)”, J.Phys.D: Appl. Phys. vol. 42, p. 015502, 2009.

Winston R Seeto, Roman K Weide, and Telesphore L Charland, “Exothermic
Bonding of Thermoelectric Couples US Patent 3808670, May 7, 1974.

D H Harnmon, “Thermal conductivity of SnTe between 100 K and 500 K”, J.
Appl. Phys. vol.37. No.8, p.3181, July 1966.

H E Bates, M Weinstein, “Thermoelectric bonding study: The bonding of PbTe
and PbTe—SnTe with non-magnetic electrodes”, NASA-CR-290, Sepl, 1965.

Wagner, RS. and WC. Ellis, “Vapor-liquid-solid mechanism of single crystal
growth”. App. Phys. Lett. Vol. 4, pp. 89-90, 1964.

Volker Schimidt and Ulrich Gosele, “How nanowires grow”, Science, vol. 316,
pp. 698-699, May 4, 2007.

M Kaur, K P Muthe, S K Despand, S Choudhury, J B Singh, N Verma, S K
Gupta, J and V Yakhmi, “Growth and branching of CuO nanowires by thermal
oxidation of copper”, J Crystal Growth, vol. 289, pp. 670-675, 2006.

J Q, Meng X M, J iang Y, Lee C S and Lee S T, “Fabrication of Germanium-Filled

Silica Nanotubes and Aligned Silica Nanofibers”, Adv.Mater. vol. 15, pp. 70-73,
2003.

220

 

[49]

[50]

[51]

[52]

[53]

[54]

[55]

[56]

[57]

[58]

[59]

[60]

[61]

X. C. Wu, W. H. Song, B. Zhao, Y. P. Sun and J. J. Du, “Preparation and
photoluminescence properties of crystalline GeOz nanowires”, Chem. Phys. Lett.
vol. 349, pp. 210-214, 2001.

Zacharias, and M.; F auchet, P. M, “Light emission fi'om GeO and GeOZ
nanocrystals”, J. Non-Cryst. Solids, vol. 227-230, pp. 1058-1062, 1998.

Chun-I Wu and T P. Hogan, “Growth of GeOz nanowires by thermal annealing”,
MRS Symp Proc, vol 951, 2007.

 

Z Y Lin and BK Garside, “Low-loss GeOz optical waveguide fabrication using
low deposition rate rf sputtering”, Appl. Opt. vol. 21, pp. 4324-4328, 1982.

M A Khan, “Substrates for surface enhanced Raman spectroscopy”, PhD
dissertation, Department of Electrical and Computer Engineering, Michigan State
University, 2008.

H Pan, S Lim, C Poh, H Sun, XWu, Y F eng and J Lin, “Growth of silicon
nanowires by thermal evaporation”, Nanotechnology, vol. 16, pp. 417-421, 2005.

Xuan Wang, Qingwen Li, Zhibo Liu, J in Zhang, Zhongfan Liu, and Rongrnin
Wang, “Low temperature growth and properties of ZnO nanowires”. Appl. Phys.
Lett. vol. 84, No. 24, pp. 4941- 4945, 2004.

A Lugstein, M Steinmair, Y J Hyun, G Hauer, P Pongratz, and E bertagnolli.
“Pressure induced orientation control of the growth of epitaxial silicon
nanowires”. Nano letters. vol. 8, No. 8, pp. 2310- 2314, 2008.

 

Youn-Joo Hyun, Alois Lungstein, Mathias Steinmair, Emmerich Bertagnolli and
Peter Pongratz. “Orientation specific synthesis of kinked silicon nanowires grown
by vapor-liquid-solid mechanism”, Nanotechnology, vol. 20, p. 125606, 2009.

Z G Bai, D P Yu, H Z Zhang, Y Ding, Y P Wang, X Z Gai, Q L Hang, G C
Xiong, S. Q Feng, “Nano scale GeOz nanowires synthesized by physical
evaporation”, Chem Phys Let. vol. 303, pp. 311-314, 1999.

Y H Tang, Y F Zhang, N Wang, I Bello, C S Lee, and S T Lee, “Germanium
dioxide whiskers synthesized by laser ablation”, Appl Phys Lett, vol. 74, pp.
3824-3826, 1999.

H Y Dang, J Wang, and SS Fan, “the synthesis of metal oxide nanowires by
directly heating metal samples in appropriate oxygen atmosphere”,
Nanotechnology, vol. 14, pp. 738-741, 2003.

Y J Zhang, J Zhu, Q Zhang, Y J Yan , N L Wang, and X Z Zhang, “Synthesis of

GeOz nanorods by carbon nanotube template”, Chem Phys. Lett. vol. 317, pp.
504-509, 2000.

221

[62]

[63]

[64]

[65]

[66]

[67]

[68]

[69]

[70]

[71]

[72]

Y Su, X Liang, S Li, Y Chen, Q Zhou, S Yin, X Meng and M Kong, “self-
catalytic VLS growth and optical properties of single crystalline GeO2 nanowire
arrays”, Materials Letters Vol.62. pp 1010-1013, 2008.

Y Zhu, C H Sow, T Yu, Q Zhao, P Li, Z Shen, D Yu and J T L Thong, “Co-
synthesis of ZnO-CuO Nanostructures by Directly Heating Brass in Air”, Adv
Funct. Mater. Vol. 16, pp. 2415-2422, 2006.

M. Hansen and K Anderkol958 Constitution of Binary Alloys., (New York:
McGraw-Hill) p 206.

F C Cheong, B Varghese, Y Zhu, E P S Tan, L Dai, V B C Tan, C T Lim and C H
Sow, “W0 3-x nanorods synthesized on a thermal hotplate”, J Phys Chem C, vol.
111, pp. 17193-17199, 2007.

T Yu, Y Zhu, X Xu, K S Yeong, Z Shen, P Chen, C T Lim, J T L Thong and C H
Sow, “Substrate-friendly synthesis of metal oxide nanostructures using a
hotplate”, Small, vol. 2, No.1, pp. 80-84, 2006.

B. A. Wacaser, K. A. Dick, J. Johansson, M. T. Borgstrdm, K. Deppert and L.
Samuelson, “Preferential interface nucleation: An expansion of the VLS growth
mechanism for nanowires”, Adv. Mater. vol. 21, pp. 1 53-1 65, 2009.

R Q Zhang, and Y Lifshitz, S T Lee, “Oxide assited growth of semiconducting
nanowires”. Adv. Mater. vol. 15, No. 7-8, pp.635-640, 2003.

A Molle, M N K Bhuiyan, G Tallardia, and M Fanciulli, “Formation and stability
of germanium oxide induced by atomic oxygen exposure”, 2006 Mater. Sci.
Semicond. Process, vol. 9, pp. 673-678, 2006.

Xueru Zhang, Yufeng Hao, Guowen Meng, and and Lide Zhang, “Fabrication of
Highly Ordered InSb Nanowire Arrays by Electrodeposition in Porous Anodic
Alumina Membranes”, Journal of The Electrochemical Society, vol. 152(10),
C664-C668, (2005).

H. D. Su, S. Z. Chang, S. C. Lee, and T. P. Sun, “High temperature InAs infrared
detector based on metal-insulator-semiconductor structure”, Electron. Lett. , vol.
31, No.1], pp. 918-920,1995.

S. A. Solin, T. Thio, D. R. Hines, and J. J. Heremans, “Enhanced Room-

Temperature Geometric Magnetoresistance in Inhomogeneous Narrow-Gap
Semiconductors”, Science, vol. 289, pp. 1530-1532, 2000.

222

 

 

 

 

 

[73]

[74]

[75]

[76]

[77]

[78]

[79]

[80]

[81]

[82]

[83]

[84]

T. Ashley, A. B. Dean, C. T. Elliott, G. J. Pryce, A. D. Johnson, and H. Willis,
“Uncooled high speed InSb field-effect transistor”, Appl. Phys. Lett. vol. 66, pp.
481-483, 1995.

Vedernikov M.V, Uryupin O.N, Goltsman B.M, Ivanov Yu.V, and Kumzerov
Yu.A, “Experimental study of thermoelectric properties of InSb nanowires”, in
Proceedings of IEEE 20:}: International Conference on Thermoelectrics, 2001, pp.
361-363.

L D Hicks, and M. S. Dresselhaus, “Thermoelectric figure of merit of a one-
dimensional conductor”, Phys Rev B, vol. 47, pp.16631- 16634, 1993.

Savarimuthu Philip Anthony, Jeong In Lee, and J in Kon Kim. “Tuning optical
bandgap of vertically aligned ZnO nanowire arrays grown by homoepitaxial
electrodeposition”. Appl. Phy. Lett. vol. 90, p. 103107, 2007.

Chul-Ho Lee, Gyu-Chul Li, Yuri M Zuev and Philip Kim, “Thermoelectric Power
measurement of wide bandgap semiconducting nanowires”. Appl. Phy. Lett. vol.
94,p.022106,2009.

Z L Wang, “Zinc oxide nanostructures: growth, properties and applications”,
Journal of Physics-Condensed Matter, vol. 16, pp. R829-R85 8, 2004.

J J Wu, and S C Liu, “catalyst-free growth and characterizationof ZnO nanorods”,
Journal of Physical ChemistryB, vol. 106, pp. 9546-9551, 2002.

W I Park, D H Kim, S W Jung and G C Yi, “Metalorganic vapor-phase epitaxial
growth of well aligned ZnO nanorods”, Applied Physics Letter. vol. 80, pp. 4232-
4234, 2002.

J Y Li, X L Chen, H Li, M He and Z Y Qiao,” Fabrication of zincoxide
nanorods”, Journal of Crystal Growth. vol. 233, pp. 5-7, 2001.

M J Zheng, L D Zheng, G H Li and W Z Shen, “Fabrication and optical properties
of large scale uniform zincoxide nanowires arrays by one-step electrochemical
deposition technique”, Chemical PhsicsmLetter, vol.363, pp. 123-128, 2002.

H J Son, K A J eon, C E Kim, J H Kim, K H Y00 and S Y Lee, “Synthesis of ZnO
nanowires by pulsed laser deposition in furnace”, Applied surface Science:
Photon Assisted Synthesis and Processing of Functional Materials-E MRS H
Symposium. vol. 253, pp.7848-7850, 2007.

Y W Zhu, T Yu, FC Cheong, X J Xu, C T Lim,V B C Tan, J T L Thong and C H

Sow, “Large-scale synthesis and field emission properties of vertically oriented
CuO nanowire films”, Nanotechnology, vol. 16, pp. 88-92, 2004.

223