.4»qu F..?.21
96.5“”... .. : ....
.3 z r
. a. . h»... .
. n .
V2? 2 3...“... mm.
a:.h:.:mn.u.€a..f no: I
fwd... 2.9.
.; J1....ﬁ.y.w
. b ‘2

.

. . ﬁlm! h
_,. imam»?
a?

3m

 

 

 

 

 

 

 

 

iii}:
.Ix5...:...v:, ..
infill... “21¢

.w Stvumr ‘sd 7
5"! ‘

.3...
.x 3..., A damn;
I. 515(3)!
1:! Llonzlb

.flt .

, 1.. Art...
gwnhmﬁi

hie:

 

 

 

 

‘ Ti-iE‘S.’ V

300?

This is to certify that the
thesis entitled

Low Resistance Contacts to Thermoelectric Materials

presented by

Jonathan James D’Angelo

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in Electrical and Computer
Science Engineering

 

 

/ V/ﬁwpz—I
Major Professor? Signature
M 6:}?! I i 2006

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

LIBRARY
Michigan State
University

 

 

 

 

-,-.-,-,-.-,-

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:ICIRC/DateDue.indd-p.1

LOW RESISTANCE CONTACTS TO THERMOELECTRIC MATERIALS
BY

Jonathan James D’Angelo

A THESIS
Submitted to
Michigan State University
In partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE

Department of Electrical and Computer Engineering

2006

ABSTRACT
LOW RESISTANCE CONTACTS TO THERMOELECTRIC MATERIALS
By
Jonathan James D’Angelo
Thermoelectric generators operate with no moving parts to convert heat

ﬂow to electricity. Low electrical contact resistance is essential for the fabrication
of high efﬁciency thermoelectric generators. These contacts must be stable to
high temperatures and through thermal cycling. Here we present the fabrication
procedure and characterization of several contacts to Pb-Sb—Ag-Te (LAST) and
Pb—Sb-Ag-Sn-Te (LASTI') compounds. Several measurement systems have
been developed for characterization of the materials and contact resistances. A
diffusion bonding system has also been developed and utilized for investigating
various metal contacts. Contact resistances less than SOuQ-cm2 have been
achieved for several diffusion bonded metals, and solderjunctions. In addition,
sputtering and e—beam deposition of various metal contacts were investigated.
The development of these systems, and the results obtained is presented along
with the processes for fabricating an entire thermoelectric power generation

module.

Table of Contents

LIST OF TABLES ................................................................................................ vii
LIST OF FIGURES ............................................................................................. viii
Chapter 1 Overview .......................................................................................... 1
1.1 The “Therrnomagnetic Effect” ...................................................................... 1
1.2 Advantages of Thennoelectrics ................................................................... 2
1.3 Current Research ........................................................................................ 2
1.4 Thermoelectric Devices ............................................................................... 4
Chapter 2 Characterizing Thermoelectric Materials ......................................... 7
2.1 What Makes a Good Thermoelectric? ......................................................... 7
2.2 Electrical Conductivity ................................................................................. 7
2.3 Therrnopower .............................................................................................. 8
2.4 Thermal Conductivity ................................................................................... 9
2.5 Contact Resistance ................................................................................... 10
Chapter 3 Depositing and Measuring Contacts .............................................. 14
3.1 Introduction ................................................................................................ 14
3.2 Transmission Line Model ........................................................................... 14
3.3 Sputtering .................................................................................................. 16
3.4 Electron Beam (E-Beam) Deposition ......................................................... 18
3.5 Deposition Masks ...................................................................................... 19

iii

3.6 Deposition of Contacts .............................................................................. 24

3.7 Electron Beam Deposition of Tungsten ..................................................... 25
3.8 Initial Contact Measurements .................................................................... 25
3.9 Annealing of Contacts ............................................................................... 27
Chapter 4 Investigating Different Metals for Electrical Contacts ..................... 30
4.1 Antimony Contacts .................................................................................... 30
4.2 Cerromatrix TLM Measurements ............................................................... 31
4.3 Other TLM material Testing ....................................................................... 32
Chapter 5 Bonding of Contacts to Thermoelectric Material ............................ 34
5.1 Diffusion Bonding ...................................................................................... 34
5.2 SnTe Diffusion Bond .................................................................................. 36
5.3 Nickel Antimony Bonds .............................................................................. 37
5.4 Antimony Thin Film Diffusion Bonding ....................................................... 38
5.4.1 TLM Measurement of Nickel Antimony Contacts ................................ 40

5.5 Other Bond Experiments ........................................................................... 41
Chapter 6 Diffusion Furnace ............................................................................... 43
6.1 Designing a Diffusion Furnace System ...................................................... 43
6.2 Unicouple Boat .......................................................................................... 49
6.3 Furnace Run 1 ........................................................................................... 51
6.4 Experiment 2 ............................................................................................. 53
6.5 Furnace Operation ..................................................................................... 54
6.6 Diffusion Bonds and Results ...................................................................... 57

iv

Chapter 7 Module Fabrication ............................................................................. 59

7.1 Indium Thin Film Modules .......................................................................... 59
7.2 Initial Module Fabrication ........................................................................... 60
7.3 Indium Foil Modules ..................................................... 64
7.4 Indium Foil Module Run ............................................................................. 66
Chapter 8 Diffusion Bonded Contacts Investigations and Alloy Bonds ............... 69
8.1 Improved Contact Resistance Measurement System ................................ 69
8.2 Indium Antimony Alloy ............................................................................... 70
8.3 Silver Diffusion Bonding ............................................................................ 71
8.4 Antimony to LASTT Contact Scans ........................................................... 73
8.5 BiSb Alloy Contact Scan ............................................................................ 75
8.6 Gold Silver Alloy Bond ............................................................................... 77
Chapter 9 New Diffusion Bonding Techniques and Stainless Steel Bonds ......... 79
9.1 Higher Pressure During Bonding ............................................................... 79
9.2 Diffusion Boat Improvements ..................................................................... 80
9.3 Initial Stainless Steel Diffusion Attempts ................................................... 82
9.4 New Unicouple Boat .................................................................................. 85
9.5 New Reducing Gas .................................................................................... 86
9.6 Stainless Steel Bond Investigations ........................................................... 86
Chapter 10 Summary and Future Work .............................................................. 88
10.1 Summary ................................................................................................. 88
10.2 Future Work ............................................................................................. 88

APPENDIX A ...................................................................................................... 90

APPENDIX B ...................................................................................................... 95

BIBLIOGRAPHY ................................................................................................. 99

vi

LIST OF TABLES

Table 1: Common thermoelectric materials, their maximum operating

temperature, maximum efﬁciency (Z), and the temperature of Zmax. ................... 3
Table 2: Contact resistance of different metals to LAST and LASTT samples...33
Table 3: List of diffusion bonds and results ......................................................... 57
Table 4: 4 Leg unicouple resistances and calculated contact resistances. ......... 67

Table 5: Diffusion bonds and corresponding contact resistances ....................... 87

vii

LIST OF FIGURES

Figure Intro 1: Thermoelectric Device ................................................................... 4

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

in p-type leg) ......................................................................................................... 5
Figure Intro 3: Thermoelectric power generation device ....................................... 6
Figure Intro 4: Efﬁciency degradation due to contact resistance[9] .................... 13

Figure 1: Diagram of planar contacts to a bulk material. L1 and L2 can be any

lengths chosen by the user as long as one contact does not interfere electrically

with another contact. ........................................................................................... 15
Figure 2: Above is pictured the plasma of the sputtering process ....................... 18
Figure 3: Transmission Line Mask 1 ................................................................... 20
Figure 4: TLM Mask 2 ......................................................................................... 22
Figure 5: Example of deposited contacts to LAST material ................................. 23
Figure 6: 32 Sample Holder ............................................................................... 24
Figure 7: Tungsten-Gold Contact IV ................................................................... 26
Figure 8: Annealed IV scan vs Preannealed ...................................................... 28
Figure 9: Gold Tungsten Anneal 400°C 1 hour ................................................... 29

Figure 10: Antimony to LASTT bond. Three stainless steel weights are shown on

the right. .............................................................................................................. 30
Figure 11: Antimony to LAS'IT interface cross section ....................................... 31
Figure 12: TLM solder contact pad ..................................................................... 32

viii

Figure 13: Diffusion bond B will have better interface resistance due to the

reduction in the size of the grain boundaries. ...................................................... 35
Figure 14: SnTe interface with the LAST ........................................................... 36
Figure 15: Nickel antimony alloy bonded to LAST .............................................. 38

Figure 16: Antimony thin ﬁlm to nickel diffusion bond. In this photo, the sample

was mounted in cutting wax. ............................................................................... 39
Figure 17: Annealing data .................................................................................. 39
Figure 18: TLM contacts of Ni-Sb on LAST ........................................................ 40
Figure 19: TLM contacts after anneal ................................................................. 41
Figure 20: Antimony-LAST interface after bond .................................................. 42
Figure 21: Diffusion furnace with quartz tube ..................................................... 44
Figure 22: Gas inlet end cap .............................................................................. 44
Figure 23: Endcap design exploded view ........................................................... 45
Figure 24: Vacuum connect end cap and pressure gauge attachment ............... 46
Figure 25: TC gauge for diffusion furnace .......................................................... 47
Figure 26: Temperature Controllers ................................................................... 48

Figure 27: Overall diffusion bonding system showing the quartz tube, the inlet

endcap, and the thermocouples for each zone on top of the furnace .................. 49
Figure 28: Preliminary drawings for unicouple boat ........................................... 50
Figure 29: Unicouple Boat .................................................................................. 51
Figure 30: Design diagram ................................................................................. 52
Figure 31: Run 1 results ...................................................................................... 52
Figure 32: N and P material to antimony bar all to a nickel strip ........................ 53

ix

Figure 33: Bonded unicouple with antimony bar to nickel .................................. 54

Figure 34: Indium thin ﬁlm on nickel ................................................................... 59
Figure 35: Indium-Nickel unicouple .................................................................... 60
Figure 36: Cold plate and interconnects ............................................................. 61
Figure 37: Unicouples in boat before diffusion bonding ...................................... 61
Figure 38: Unicouples after removal from the furnace ....................................... 62
Figure 39: 20 bonded unicouples for module fabrication ..................................... 63
Figure 40: Module fabrication on the hotplate .................................................... 63
Figure 41: Completed Module ............................................................................. 64
Figure 42: Indium foil bond .................................................................................. 65
Figure 43 : Indium foil bond small pieces ........................................................... 65
Figure 44: Module formed from the unicouples shown in Figure 43. .................. 66
Figure 45: Indium foil module pre-bake. ............................................................. 66
Figure 46: Modules made with left over unicouples ........................................... 67
Figure 47: Room temperature scanning probe ................................................... 69

Figure 48: Room temperature probe scan at 100Elm resolution across a solder to

LAST material junction. ....................................................................................... 70
Figure 49: 1Sit alloy diffusion bond ...................................................................... 71
Figure 50: N and P bonds to silver ...................................................................... 72
Figure 51: Silver to N-type LAST voltage proﬁle scan across contact ................. 73
Figure 52: Antimony to LAST contact Scan ....................................................... 74
Figure 53: LAST-Antimony-LAST Contact Scan ................................................ 75
Figure 54: Phase diagram for bismuth - antimony alloy []. ................................. 76

Figure 55: BiSb alloy to LAST contact scan ........................................................ 77
Figure 56: Gold silver alloy bonds ...................................................................... 78
Figure 57: Stainless steel bonding block ............................................................ 79
Figure 58: LAST bonded directly to stainless steel ............................................ 80
Figure 59: Added quartz pieces for weights during diffusion bonding ................ 81
Figure 60: First diffusion bon of LAST and LASTT to stainless steel ................. 82
Figure 61: N-type bond at 725°C ......................................................................... 82
Figure 62: P-type bond at 725°C ......................................................................... 83
Figure 63: P-type to stainless steel contact scan ............................................... 84
Figure 64: N-type to stainless steel contact scan ............................................... 85
Figure 65: New unicouple boat .......................................................................... 86
Figure A 1: ZnClz pre-etch + Sb powder, no bond (out-gassing) ......................... 90
Figure A 2: Sb, good bonds (no out-gassing) ...................................................... 90
Figure A 3: Bile (55/45) on Ni, no bonds .......................................................... 91
Figure A 4: Bile to LAST, good bond ................................................................ 91
Figure A 5: Bile to Ni (No LAST), Bi/Sb blackens on Ni ................................... 91
Figure A 6: Sb only to LASTT, Good Bonds ....................................................... 92
Figure A 7: Sb on Ni (No LAST), Not Bad- A little blackening ............................ 92
Figure A 8: Bi/Sb (55l45) on Ti (no LAST), Good bond between metals ............ 92
Figure A 9: Bile (55l45) on W (no LAST), no bond between metals ................ 93
Figure A 10: In (0.5mm thick) on Ni, Consumed samples .................................. 93
Figure A 11: In on Ni (No LAST) Good bond between metals ............................ 93
Figure A 12: In on Nickel to LAST (prealloyed), Good Bond .............................. 93

xi

Figure A 13: In 1mm thick on Ni, Good Bond ..............................

Figure A 14: Stainless steel to LAST, good Bond .......................

Figure B 1: N-type to Stainless Steel, 690°C 1 Hour, 138uﬂcm2
Figure B 2: P-type to Stainless Steel, 690°C1 Hour, 195tiocm2
Figure B 3: P-type to Stainless Steel, 630°C 4 Hours, ~5chm2

Figure B 4: N-type to Stainless Steel, 700°C 4 Hours, ~5pﬂcm2

xii

....................... 94

....................... 94

....................... 95

....................... 96

....................... 97

....................... 98

Chapter 1 Overview
1.1 The “Thermomagnetic Effect”

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

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

current ﬂow. [2].

Later Thomson showed that heating or cooling along the length of a single
material also takes place when electrical current ﬂows through the material. This
became known as the Thomson Effect. Today’s thermoelectric devices are

based on these scientiﬁc principles.

1.2 Advantages of Thermoelectrics

The science of therrnoelectrics was born some time ago, but only in recent
days has it grown to become available commercially. Most therrnoelectrics are
used for cooling and heating devices. A thermoelectric cooler can be bought at
your local stores. A list of advantages is [3];

1. No moving parts for long device lifetimes and little maintenance

2. They can be made into small or large devices without loss of efficiency

3. The same module can be used as a power generator and a cooler

4. They can operate in any orientation

5. They give very little electronic noise

1.3 Current Research

The goal of the transport characterization laboratory here at Michigan
State University is to build a twenty percent efﬁcient, one kilowatt generator. The
generator will be operating under a temperature gradient of ~580 degrees (room
temperature to 600°C). Common thermoelectric materials and their maximum

operating temperatures is given in Table 1 below.

Table 1: Common thermoelectric materials, their maximum operating temperature,
maximum efﬁciency (2), and the temperature of Zmax.

 

Zmz,x (K'l) Uscﬁtl Range Tm",x (K)

 

 

 

BizTe3 3x10“3 < 500 K 300
PbTe 1.7x10'3 < 900 K 650
Si-Ge 1x103 < 1300 K 1100

 

 

 

 

 

 

Most bismuth telluride modules operate near room temperature where
solder contacts can be used on both the hot and cold side junctions. The
devices built here at Michigan State University need to have bonds on the hot
side of the device that can withstand rigorous temperature cycling and high
temperature operations where solder contacts are not possible.

The Chemistry Department at Michigan State University has discovered
several chalcogenide based materials that have shown promising thermoelectric
properties [4]. Recently compounds in the Ag1.bemeTe2+m and
Ag(Pb1.xSnx)meTe2+m systems have exhibited outstanding thermoelectric
properties for power generation applications [5]. Lead antimony silver tellurium
or also know as LAST and Iead-antimony-silver—tin-tellurium, or LASTT, are the
n-type and p-type materials (respectively) used for fabrication of the

thermoelectric devices. The purpose of this thesis is to investigate both high and

low temperature contacts to this material. The contact resistance of a
thermoelectric device is a loss to the efﬁciency of the device (through ohmic
heating) and thus should be minimized. As part of this thesis, many materials
have been tested and various contact resistance measurement systems were

developed.

1.4 Thermoelectric Devices

Metal Contacts

N P
——

Figure Intro 1: Thermoelectric Device

 

 

 

 

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

diode and these devices exhibit linear current vs. voltage proﬁles.

Heat

 

T T
® ®

‘

 

 

 

 

 

 

Current I”

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

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

will then be cooled following the same principles.

Heat

® GD
l l

+

Figure Intro 3: Thermoelectric power generation device

These devices, thus serve as heat pumps. If instead of supplying electrical
power to the device in order to create a temperature gradient, a temperature
gradient is supplied to the device, then electrical power can be generated. The
larger the temperature gradient the more voltage produced by the module. The
heat that is supplied by the module is carried to the cold side of the device by the
electrons in the n-type and the holes in the p-type. This makes one side a
positive potential and the other a negative. This can be attached to an electrical
circuit and be used as a power supply. The concept is relatively simple, but
these devices do have their limits. There efﬁciencies are very low (typically ~5%

for BizTea based devices).

Chapter 2 Characterizing Thermoelectric Materials
2.1 What Makes a Good Themtoelectric?
Every thermoelectric material is rated by its ﬁgure of merit, Z T. Z T is a

dimensionless quantity given by

S20
ZTZT'T (1)

where S is the thennopower, c is the electrical conductivity, x is thermal
conductivity, and T is temperature. To obtain a high ZT the therrnopower and
conductivity must be high along with a low thermal conductivity. All of these
quantities are temperature dependent. All of three of these quantities are
measured here at MSU. Maximizing ZT can be challenging as these materials
properties are not independent of one another. Usually a material that has a
higher electrical conductivity will have a high the thermal conductivity. This
behavior is seen in most conductors, such as copper, and insulators, such as
glass. Optimization of the three material properties is generally found in heavily

doped semiconductors [6].

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

of a semiconductor material.

0' = (101" '"+/1p 'P) (2)
In equation (2), n and p are the dopant concentrations of a semiconductor

material. The concentration of dopants in LAST materials is around 1019 cm'3.

At this concentration, the sample is degenerately doped and shows metallic
temperature dependence for the electrical conductivity (decreasing o with
increasing temperature). We have experimentally found electrical conductivity
values in the range of 800 to1500 S-cm'1 to correspond to some of the highest

2 T samples.

2.3 Therrnopower

The therrnopower of a device, or also known as the Seebeck coefﬁcient is
the open circuit voltage of a semiconductor under the inﬂuence of a temperature
gradient. The therrnopower of a material is a measure of the local entropy, and
is inﬂuenced by effective masses and the energy dependence of the density of
states. For LAST materials we have experimentally found values near 100uV-K'1
at room temperature tend to yield the highest 2 T materials. When adjusting the
composition and doping of LAST, a reduction in the carrier concentration causes
an increase in the therrnopower, however also corresponds to a decrease in the

electrical conductivity.

S = 7:2sz . dln0'(E)
36 dE

Equation above is the Boltzmann transport theory equation for therrnopower, the

 

 

H; (3)

 

Mott equation [3]. The electrical conductivity in the equation is a function of the

Fermi energy. The therrnopower is directly proportional to the derivative of the

density of states. Current research is looking for more complex structures with

materials that have 2D, 1D and OD structures.

2.4 Thermal Conductivity

The thermal conductivity is the measurement of a materials ability to
conduct heat. From equation (1) it can be seen that a low thermal conductivity
material helps to increase the ﬁgure of merit. For the materials studied in this
thesis, a thermal conductivity near 1W-m'1-K'1 is commonly seen for the high 2 T

samples. There are two contributions to thermal conductivity.

KZKe+KI (4)

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

law
[(6 =L°U°T (5)

where L is the Lorentz number, a is the electrical conductivity of the material, and
T is the temperature. This indicates that a trade off between high electrical
conductivity and low thermal conductivity must be considered for optimizing the

ZT of a sample.

The lattice thermal conductivity can also be adjusted through mass
ﬂuctuations within the crystal lattice. This can be accomplished through the
choice of multielement semiconductor materials containing high and low mass
elements, and through alloying of materials. More recently, the prediction of ZT
enhancements through quantum conﬁnement [7] has led to thin ﬁlm materials
that exhibit high 2 T values primarily through a signiﬁcant reduction in thermal
conductivity [8]. The LAST and LASTT bulk materials developed at Michigan
State University also exhibit endotaxially imbedded nanostructures which form
during ingot growth through a spinodal decomposition mechanism [5]. These
nanostructures are believed to reduce electron-phonon interaction by

preferentially scattering phonons while allowing electrons to ﬂow through them.

2.5 Contact Resistance

Reduction in the contact resistance is crucial for building an efﬁcient
thermoelectric device. The contact resistance comes from the interface of to
dissimilar metals or a semiconductor to a metal. Also for a good thermoelectric
device all electrical contacts must be ohmic. An ideal ohmic contact is a perfect
source and sink of both holes and electrons. A nonlinear contact would have
diode behaviors. The losses due to IZR from the contact have to be reduced.
Below is the derivation of how the contact resistance directly affects the

efﬁciency of a device.

10

 

 

 

 

 

 

 

 

Po

 

'7 - Q. <6)

Pa = 12R.) (7),

1
Qh : KAT + SThI —§12(R + RC) (8),

 

'11: f+6)2+ (1+6)ZTav (10),

11

_£_R +R +Rh1+Rch2

 

 

 

6 cc] cc2 c
R Rn+Rp <11),
Z M.
2 .
(”2171-6) +(ﬂ+1+5)_(1+5)77¢ (12).
h

Equation (12) gives the overall efﬁciency of a device. Figure (lntro.4)
below shows plots of the relationship between contact resistance and module

efﬁciency.

12

 

 

 

 

0-25 I I I I. I.
R E 5:00
5:?5 (6:01
02 .. ............. .............. .............. .............. ”o‘f‘l'r'é .......... A
’;/""'O 4:9
// ‘5'- i
I’ '0... .0:-
f /:I 0'. ' .’,’o"10"
0.15 " 'f‘,‘5",“‘"WV/".129”, """""" ‘
: : - -’-" "' 5:11
i /’ I:o o"$.. -
. ’/"' , ’01:...
:- ' ’0. .’.o':'.
:I’P’. '41"
- ’42:. 42:"? : :
3 III." ' 44“". 3 § 5
I. .:‘:'
If./ “ g i i
0.05 - ------- “‘ ----------- -------------- ------------- Th = 800K ----
‘ = ‘ Tc = 300K
. '
0 i 1 1 1
0 05 I 1.5 2 25 3
ZT
avg

Figure Intro 4: Efﬁciency degradation due to contact resistance[9]

As is further described in this thesis various contact metals have been
investigated for contacts to LAST and LAS'IT materials. A description of the
diffusion bonding system, and the results obtained from it are discussed. In
addition, several measurement systems speciﬁcally developed for testing the

contact resistance have been developed, and are described in this thesis.

13

Chapter 3 Depositing and Measuring Contacts
3.1 Introduction

Electrical contacts to our thermoelectric devices are needed to ensure an
overall promising efﬁciency. The goal of this thesis is to investigate low
resistance contacts to new thermoelectric materials of Ag1.bemeTe2.m and
Ag(Pb1-xSnx)meTe2.m. As part of this study, systems were developed to
measure contact resistances, and a furnace system was developed for diffusion
bonding of various metal contacts to the thermoelectric materials. The materials
investigated were all contacts to LAST (lead antimony silver tellurium) and
LASTT (LAST with tin) materials. The contact metals to be measured were also

deposited to the materials by different methods which are described below.

3.2 Transmission Line Model

The contact resistance of a junction can be obtained by measuring the
position dependent voltage proﬁle through the junction. This is a common
technique for determining the contact resistance for fabricated devices, however
for small contact resistances, the discontinuity in the voltage across the junction
can be very small. Under such conditions the transmission line model (T LM)
technique [10] is a more accurate measurement of the contact resistance. This
technique uses planar contacts to bulk materials to measure contact resistance.
By measuring the resistance between these pads, a simple calculation can then

be done to obtain the overall contact resistance in units of pﬂcmz.

14

R1

   

Figure 1: Diagram of planar contacts to a bulk material. L1 and L2 can be any lengths chosen by
the user as long as one contact does not interfere electrically with another contact.

Once this setup is achieved, two simple resistance measurements are
needed followed by the calculation of the contact resistance.
: R2 'L1_R1'L2
C _ (1)
2(LI L 2 )

 

This will give you the contact resistance of your contact metal to your bulk
material. This calculation can be easily derived for any material. The only
assumption this calculation uses is that the material along lengths L1 and L2 are
uniform in electrical conductivity. If they are not, the ratio of the two resistances
will not be correct and then the calculation would yield an improper contact
resistance.

The TLM method is more accurate, however the conﬁguration of the
contacts for the TLM method is not always amiable to the ﬁnal device
conﬁguration. Therefore, both measurement techniques are valuable in the
course of developing thermoelectric generators. Several contact materials have
been investigated using the TLM method by depositing the contact materials with

sputtering and electron beam evaporation techniques.

15

3.3 Sputtering

Sputtering is a method of depositing thin ﬁlms to a substrate by repetitively
accelerating ionized atoms of a noble gas (typically argon) into a target material.
The energy of these accelerated argon atoms are transferred to the atoms in the
target, which expels them from the target surface, and are collected onto the
substrate. This technique has been likened to a game of atomic billiards. A
noble gas is used as the sputtering atoms since it will not easily be incorporated
into the deposited ﬁlm. The atoms of argon are first ionized in a plasma and then
are accelerated by a large electric potential.

There are two main types of sputtering, Direct Current (DC) and Radio
Frequency (RF). DC sputtering can only be used for conducting materials, while
RF can be used for both conducting and insulating materials. Both methods
were used in this investigation.

Sputtering rates and ﬁlm thickness are dependent on the argon pressure
and power used. These yields can be found in most litterateurs, or can be found
experimentally. Various sputtering systems check pressures at different
locations and may have varying power capabilities, and electrode conﬁgurations,
therefore sputtering yields can change from system to system.

Sputtering instructions are as follows;

1. Inside chamber loosen the dark space shield retaining screws and slide the

dark space shield off the surface.

16

2. Loosen the 4-target hold down screws and rotate the hold down ring to a
position where the through holes are directly beneath the screw heads. The
hold down ring can now be taken off.

3. Remove old target and install new one

4. Replace the target hold-down ring over the 4 protruding screws and rotate
the ring so that the screw heads are not aligned with the through holes.
Tighten down the screws to secure the hold down ring over the target.

5. Replace the dark shield and tighten the dark space shield retaining screws.

6. IMPORTANT! Visually inspect the gap between the dark space shield and
the target hold—down ring; it should measure at least 1116”

7. Pump down system to a vacuum of lower than 20mTorr (vacuum level is
user determined)

8. When achieved... close hi-vac and open high wac throttle, and process gas.

9. Set gas pressure by adjusting the gas inlet. Want about 5mTorr of argon.

10.Tum on power supply and adjust the output until plasma is observed.

11.Let target wear in for 5-10 minutes.

12.Sputter to desired ﬁlm thickness. Rates depend on pressures and power.

17

 

Figure 2: Above is pictured the plasma of the sputtering process

The sputtering plasma is a bright purple glow. After the process is completed, let
the system cool for about a half hour, and then shut off the pump and vent the

system.

3.4 Electron Beam (E-Beam) Deposition

Electron beam deposition is an evaporation technique for depositing thin
ﬁlms. This is a good deposition technique for many materials, but does not work
well for some materials. In the physical vapor deposition (PVD) system located
in the Engineering Research Complex Cleanroom, a quartz crystal monitor is
used to accurately measure ﬁlm thickness during deposition. This e-beam
system is able to hold up to four different evaporation sources at one time.

An electron beam is guided by a magnetic ﬁeld and controlled by a set of
deﬂection plates to melt the target material. Once the melted material enters the

vapor phase it follows a direct path up through the PVD chamber and deposits

onto a substrate located on a rotating platen. Within this deposition system,
sample must be attached to the rotating platen in a way that allows the platen to
be ﬂipped upside down during deposition. Because of this, and the size of our

substrates, special masks had to be developed for deposition.

3.5 Deposition Masks

The ﬁrst mask developed for contact resistance research was the

Transmission Line Mask shown in Figure 3.

19

—-- 9.8m -'-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5.0 rrm
—-— ---—-
5.0 rrrn I
I O J 1
:1 2.8 rrm
I 2.0 nm :1
* 4.0 rrm
[II
+ 6.0 nm 42.1 rrm
1 I:
8.0 m I j
; I:
13.4 mm 1.0 mm
O O I

 

 

 

—-— --— 2.0m

 

 

Figure 3: Transmission Line Mask 1

This mask was designed as a shadow mask such that the contact material would
deposit only in the ordered pattern needed for the TLM measurements. The
contact spacing was varied from 2mm to 8mm in 2mm increments. Each contact
has an area of 2mm2. The ﬁrst mask was design not only as the mask, but also
doubled as the sample holder. The mast has 4 screw holes at the comers that
are used to screw down the mask and also press the sample to the platen.

Several challenges with this design were discovered during its use, such as the

20

need for the sample to be perfectly ﬂat to keep the sample held well in place.
Placing and centering the sample while screwing the mask into place was
challenging, and the sample had a tendency to crack under the pressure. The
mask did serve its purpose well until the second TLM mask was designed.
Because of the challenges associated with this mask design, a second design

was developed as shown in Figure 4.

21

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

15.00rrm
r C1 I
I 5.00m
I 4'09"“ - j I 30.00mm
3.00m . !
imam I i S :2 mm 1050""
[ +1.75rrm
ﬂ ‘—
l l I
C) O ____________
___ 8.08m
, ﬂ
___ A U
30.00rrm
”“ ﬂ
___ K/
9.75m ____ ______
CD CD
3.50m

 

 

Figure 4: TLM Mask 2

Wrth the 2"d TLM Mask the sample is held in place from the side with a spring-
loaded setscrew. The spring would allow for any expansion or contraction of the

sample during the deposition. Due to the nature the set screw only making a

22

small pressure contact, the possibility of cracking the sample while setting it into
place is very minimal.

The ﬁrst mask also presented the problem that the contacts deposited
would only have an area 1mm by 2mm. This meant that your sample size could
only have a maximum width of 2mm. If the sample had a width larger than 2mm,
then current crowding would effect the measurement. With the second TLM
mask (Figure 3), many different sample widths could be accommodated. The
spacing of the contacts was reduced by a millimeter to allow for shorter samples.

More contacts could be deposited per length allowing for better measurements.

 

Figure 5: Example of deposited contacts to LAST material.

With this new mask, deposition of contacts was sped up, and more data was able
to be collected.

Lastly another mask was developed to hold as many samples as possible
in order to deposit contact materials. The 32 sample holder was developed to do

just that.

23

 

 

01111111111111110

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

O-u- +2- #0
OTTTTTTTTTTFTTO
0.11.11.111111111110
001T liTiTF‘TT’UoO

 

 

 

 

 

 

Figure 6: 32 Sample Holder

Figure 5 shows the top view of the 32-sample holder where each sample can be
up to 5mm by 5mm in cross section, and as tall as 7mm. Each sample is held in
place by a spring-loaded setscrew. The mask is made to allow the user to
deposit contacts on one side of the material and the just ﬂip the entire holder
over to deposit contacts to the other side. This can be done without removing any
of the samples. This makes contact deposition fast and efﬁcient for module

fabrication.

3.6 Deposition of Contacts

The ﬁrst contacts investigated were tungsten which was sputtered
deposited onto the LAST materials. Tungsten was chosen due to its good
contacts reported to lead telluride [11]. As expected, the sputtering rate for

tungsten was found to be very low; however an added difﬁculty of maintaining the

24

proper pressure (at least 5mTorr of argon) led us to investigate alternative

deposition techniques for tungsten.

3.7 Electron Beam Deposition of Tungsten

Although tungsten has a melting point of 3410°C, it can be deposited by
e-beam evaporation. This is possible by focusing the electron beam such that
the local temperature exceeds the melting point of tungsten. Because of the high
melting point, the current needed (500mA) it higher than typically used (200mA).
The deposition rate is relatively low (~ 1.5Als), therefore a thin layer of tungsten

(20nm) was followed with a thick layer of gold (200nm).

3.8 Initial Contact Measurements

After achieving a good deposition of contacts into the desired pattern, the
contacts were tested to verify they were ohmic. To verify a linear current versus
voltage behavior of the metal to semiconductor junction, a LabView program was
written to control the current through the junction and collect the corresponding
voltage data. A diode behavior for the junction would show as a non-linear
current versus voltage curve as the current passes through zero. To avoid
voltage drops along the measurement cables, a four-probe technique in a Janis
cryostat probe station. In the probe station, the environment was pumped down
to ~300mTorr. The gold-tungsten contacts were measured with a current

between —200mA to 200mA.

25

 

 

 

 

 

0.3 - I I' I ‘I r I I I 1 I I I I I l T f I 1 l I I I I ‘
- ~—- y = 0.010272 + 2.0309x R= 0.99971 '1
0.2 '- u
'. J
0.1 _ .1
if“, I I
t o '- I-
p 1 I- »: '
. T7 -
I f I
: 2;” '
.. .1 " ‘
-0.2 l- 1
I 1
-O.3 F I l I l l I l l L l l l l l l I l I l l l I. I l ‘
-0.15 -0.1 -0.05 0 0.05 0.1

thage (V)

Figure 7: Tungsten-Gold Contact IV

Figure 7 shows the collected data are a good ﬁt to the linear approximation
indicating an ohmic behavior for this junction. This came as no surprise as the
semiconductors used in this study were doped to high degeneracy.

In the next step resistance measurements between the adjacent contact
pads were taken to obtain the contact resistance by the TLM technique. The ﬁrst
contact measurements on an n-type LAST sample (ETN1) came out inconclusive
and was believed to be due to inhomogeneities and/or cracks in the sample.

A p-type LASTT sample (ETP2) was then investigated. The conductivity
of the sample was measured to be 1,33OS/cm and is very uniform. For these

contacts a deposition was done of 10nm of tungsten followed by 300nm of gold

26

and a contact resistance of 119p£2cm2 was found. This is a good start, but we
would like to see contact resistance of close to 5 pacm2 to have good overall
module efﬁciency. To help verify this result, a contact scan just in between each
contact pad was also done, and then the difference of the two resistances and
divided by two. The contact scan is done with a four probe measurement
method. This measurement is also done in the Janis probe station. The
measurement is done under vacuum to ensure accurate results. This also gave
a contact resistance of around 120p.Qcm2 in good agreement with the results
from the TLM technique.

A relatively thick layer of tungsten alone on n-type LAST was investigated
next. A thickness of 100nm was deposited at a rate of 1.5A/s, and a contact
resistance of 65p£2cm2 was measured by the TLM technique. The effects of

annealing these contacts were studied next.

3.9 Annealing of Contacts

Annealing of the contacts is needed to help drive the atoms of the contact
metal deeper into the bulk material to hope for a lower contact resistance. A
good example is that most contacts to silicon are made with aluminum. If this
contact is deposited without an annealing step, a Schottky junction at each
interface is obtained, however with a high temperature anneal the aluminum to
silicon junction becomes ohmic. The tungsten contact was annealed at 600°C for

1 hour.

27

0.003

 

 

 

 

 

 

 

 

 

 

voltageMAnnealed

0.002
3 0.001
(I
d)
C
C
<
9 0
I
D)
g
0
> -0.001

-0.002

-0003 ”—TT’—T_TTT'—TTI I I T I I I I I I I I ii I I I T I I I I I I I

-0.15 -0.1 -0.05 0 0.05 0.1 0.15

current (A)

Figure 8: Annealed IV scan vs Preannealed

Figure 8 shows the voltage versus current scan between two contact pads of
tungsten on an n-type LAST sample. The curve after a 600°C annealing step
shows a signiﬁcantly smaller slope, indicating a reduced contact resistance. A
voltage scan along the sample alone indicated no change in the sample
resistance after annealing, thus the improvement shown in Figure 7 is due to the
contact resistance alone.

Annealing the tungsten/gold contact at 400°C for 1 hour inside a vacuum
sealed quartz tube showed signiﬁcant degradation of the contacts as shown in

Figure 9.

28

 

Figure 9: Gold Tungsten Anneal 400°C 1 hour

More investigation is needed to discover what chemically happened
at the contact. Only now assumptions can be made. One being that the gold
just absorbed into the sample. Another could be that the gold evaporated
completely off of the sample. Lastly that the tungsten and the gold formed some
sort of alloy. Once again these are all speculations and need to be veriﬁed

scientiﬁcally.

29

Chapter 4 Investigating Different Metals for Electrical Contacts

4.1 Antimony Contacts

It was then suggested that antimony might be a good contact material. To
investigate this, a 1 pm ﬁlm of antimony was e-beam deposited onto a p-type
LASTT sample (ETP2). The TLM measured contact resistance was found to be
56.83chm2. Next, a bond of LASTT to antimony was attempted at the end of
the sample (in the conﬁguration that would be used for fabricating a device). A
sample of LASTT was placed in a quartz tube with a coin of antimony on top of it
and three stainless steel weights were placed on top of the antimony. The stack
of materials was placed into a quartz tube and vacuum-sealed as shown in
Figure 10. The temperature was raised to 600°C and it remained at high

temperature for only a few minutes.

 

Figure 10: Antimony to LASTT bond. Three stainless steel weights are shown on the right.

As shown in Figure 9, the antimony melted around the sample. The antimony did
bond, however the stainless steel weight bonded as well. Next this sample was

taken and a cross section was cut to see the interface of the two materials.

30

 

Figure 11: Antimony to LASTT interface cross section

There is a clear-cut interface between the two materials. Since the antimony
melted during the bonding process, there is good coverage of the LASTT
surface. After cutting the sample, the sample size was too small for further

contact resistance measurements.

4.2 Cerromatrix TLM Measurements

Diffusion bonding of the hot side contacts is necessary since it will be
exposed to temperatures as high as 600°C during operation. The cold side of the
module is not expect to exceed 200 °C, however also requires low resistance
contacts. This allows for the possibility of solder contacts on the cold side. To
that end, a Cerromatrix solder was investigated. This solder is similar to Woods
metal, and solders very well to the LAST materials. The Cerromatrix was
soldered to a piece of LAST into the TLM pattern for contact resistance

measurements.

31

 

Figure 12: TLM solder contact pad

The contact resistance of the solder to the sample was found to be 33p.Qcm2

which is quite good for device fabrication.

4.3 Other TLM material Testing

Although antimony was found to be a promising contact, many other
materials were also tested. The following table is good to use a reference for
later bonds. All contacts were deposited by electron beam, and measured in the

probe station.

32

Table 2: Contact resistance of different metals to LAST and LASTT samples

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NIP Material Contact Thickness Contact Resistance (pncmz)
P Sb 3.1m 80
N Sb 1 pm 20
P Ag 500nm 500
N AL 500nm 1 800
N Cr 500nm 275
P Cr 500nm 120
N Ti 250nm 1 30
P Ti 250nm 145
P W 100nm 65
P W/Ag 20nm/200nm 140

 

Contact metals were chosen along the lines of what was attempted in the

literature [14]

33

 

Chapter 5 Bonding of Contacts to Thermoelectric Material
5.1 Diffusion Bonding

Diffusion bonding is a system of bonding two similar or dissimilar metals,
alloys, or nonmetals. This is achieved by pressing the two materials that you
would like to be bonded, preferably under vacuum or in the presence of an inert
gas to reduce oxidization. There are three elements that control how well the
bond is formed; ﬁrst is the temperature you are doing the bonding at, second is
the amount of weight is placed onto the samples as pressure keeping the two
interfaces together, and lastly is the amount of time you stay at a certain bonding
temperature. The bonding temperature should be anywhere to 50-70% of the
melting point of the most fusible metal. The temperature aids in the inter-
diffusion of the atoms across the interface. The bonding pressure ﬁlls voids and
pushes out any oxides that may reside on either of the two surfaces. The time of
bonding needs to be minimized. If the bonding time is too long, voids will be
created at the interfaces, and chemical changes may occur.

In the case with LAST we are looking to reduce the sizes of the grains in
our system in order to create higher density samples [12]. Diffusion along grain
boundary paths becomes the dominant diffusion channels. So in our attempts to
decrease the grain size of our materials we will greatly improving the diffusion
bonding of our interfaces, yielding a reduced contact resistance and better
overall efﬁciency of our module. The increase in grain boundaries is a case in
which the ease that the atoms can migrate along the free surfaces of the grain

boundanes.

34

Diffusion Diredion

 

 

 

 

 

 

 

 

 

 

Figure 13: Diffusion bond B will have better interface resistance due to the reduction in the size
of the grain boundaries.

Diffusion bonding attempts were done at many different temperatures,

pressures, and times. Along with good bonds, we also wanted to make sure that

we did not degrade our samples at the same time (through desorption of the

sample) Such desorption would degrade the properties of the sample and result

in a decrease the performance of the device. To minimize this outgassing of

material, the diffusion bonding was done in an overpressure environment of

argon.

35

5.2 SnTe Diffusion Bond

Tin telluride was the ﬁrst alloy investigated for bonding to LAST. The alloy
was prepared in the Chemistry Department and then vacuum-sealed in a quartz
tube, with small quartz weights placed on top. The bonding temperature was
raised at a rate of 50 degrees a minute to 550°C and then slowed to a rate of 20
degrees per minute to 700°C. It was maintained at the maximum temperature for
a few minutes, and then the furnace cooled over night.

When ﬁrst investigated the bond appeard to have bonded well with no

signs of outgassing.

.J‘ ‘

. 1. ' u. w ' a
{‘3 ‘39:?! * 2.9:"
” “43$ 53‘, U. '

 

Figure 14: SnTe interface with the LAST

After opening the quartz tube, however we found that there really was no
bonding of the alloy to the LAST. From this experiment, it was hypothesized that
one or more of the following conditions were needed for better SnTe-LAST
bonds:

0 Increase bonding temperature

36

0 Increase pressure between materials during bonding

. Increase the time of the bonding

5.3 Nickel Antimony Bonds

Based on the earlier results that showed good contact between antimony
and LAST, but too low of a melting point for antimony, we investigated a small
amount of antimony between the sample and a metal interconnect. To choose
the metal interconnect materials with good resilience to oxidization, low electrical
resistivity, and a coefﬁcient of thermal expansion well matched to LAST were
sought. Nickel has such characteristics, along with a relatively high thermal
conductivity for good transfer of heat into the module, and it is readily available to
purchase. To test the bonding characteristics of antimony to nickel, pieces of
antimony and a nickel pellet were vacuum sealed in a quartz tube and annealed
at 700°C for a few minutes. The quartz tube was then removed from the crucible
furnace and allowed to naturally cool. The antimony was found to bond well to
the nickel with good coverage and wetting characteristics similar to a solder.

Next a bond between the antimony coated nickel and a sample of LAST
material was investigated. The two pieces were again vacuum sealed into a
quartz tube and annealed at 700°C for a few minutes. It was found that the
antimony coated nickel bonded well to the LAST material, however there was
excess antimony in the structure that would hinder device fabrication. Therefore,

thin ﬁlm depositions of antimony were investigated. .

37

 

Figure 15: Nickel antimony alloy bonded to LAST

5.4 Antimony Thin Film Diffusion Bonding

To begin an n-type sample of LAST (ETN16) was loaded into a mask, and
placed in the physical vapor deposition system located in the Engineering
Research Complex Cleanroom, and the chamber was pumped down to ~106
Torr.. Argon was then allowed to ﬂow into the chamber to a pressure of
100mTorr, and a plasma etch of the sample was initiated to remove oxides and
hydrocarbons from the sample surface. Then the chamber was evacuated
again, and 500nm of antimony was e—beam deposited onto the sample. A nickel
electrode was then polished to clean mirror ﬁnish with 2,500 grit sandpaper. The
antimony coated LAST sample was then bonded by loading them into a vacuum
sealed quartz tube and raising the temperature to approximately 700°C. The

bond attempt was successful as shown in Figure 16.

38

 

Figure 16: Antimony thin ﬁlm to nickel diffusion bond. In this photo, the sample was mounted in

cutting wax.

To study the temperature dependence of the contact resistance during

annealing, the sample was mounted with electrical current and voltage leads in a

standard four probe conﬁguration as shown in Figure 16. The sample was

mounted in the Ultra High Temperature (UHT) measurement system and the

junction conductance was monitored during annealing under a vacuum.

700.

6(1)‘

5(1)'

Conductivity

4(1)

3(1)

200 ——x

200

Nickel to LAST Bond

--I'— T - T _I ——-1—#-r-AT V 9— ‘—ﬁ—_?—_ AIA W -'—I . I _,

 

Contact Improves

......

300 400 500 600 700
Furnace Temp (K)

Figure 17: Annealing data

39

As seen from the data (Figure 17), near 500K the conductivity curve changes
slope. This is showing an improvement of the contact at this temperature. The
conductivity then continues on this trend to 700K, and overlaps the 500K data
upon cooling. Thus a high diffusion of antimony atoms near 500K appears to
signiﬁcantly improve the contact during annealing. It should be noted that no

pressure was applied to this contact during annealing.

5.4.1 TLM Measurement of Nickel Antimony Contacts
Using the TLM mask 500nm of antimony was deposited onto an n-type

sample of LAST followed by 150nm of nickel.

 

Figure 18: TLM contacts of Ni-Sb on LAST

The contact resistance measured is 140chm2. This is still too high to build an
efﬁcient device. The sample was then vacuum—sealed into a quartz tube and
annealed at 700K for 39 hours. The contact resistance improved after the anneal
to 16¢)sz which was very encouraging. With a contact resistance this low, it

can be considered negligible when fabricating a device.

40

 

Figure 19: TLM contacts after anneal

The contacts darkened some during the anneal, but there was no apparent
change in the LAST and contact metal. The fact that the tube is a small
environment, during the annealing process, anything that does outgas would
create an overpressure. Thus it could be possible that an equilibrium is
established that would keep the material from outgassing further. With this low of
a contact resistance, it was time to begin building devices. At this time we had
no system that was large enough to build modules. All diffusion-bonding
attempts up to this point were done in tiny vacuum-sealed quartz tubes. An
entire module will not be able to be sealed individually and then baked, so the
fabrication of a diffusion furnace was needed which could accommodate the size

of the modules planned for the ﬁnal devices.

5.5 Other Bond Experiments

Two other experiments assocated with the bonding are also worth noting.

First was to do EDS analysis on the interface between the nickel and antimony,

41

by ﬁrst removing the nickel electrode. This interface is shown in Figure 19 after

removal of the nickel electrode.

 

Figure 20: Antimony-LAST interface after bond

The EDS showed the interface was tellurium rich. After subsequent EDS
analysis on other samples, this was common to ﬁnd tellurium at the contact and
on other surfaces of the electrodes.

An attempt at a nickel antimony diffusion bond was done in open air. The
result was that both samples turned black from oxidization and no bond occurred.
The conclusion is that the bond must be done in an oxygen free environment to
ensure a good bond.

Next a bond of tin to LAST was investigated. Since tin has a low enough
melting point, a soldering iron was used to apply a layer of tin to the LAST
material. From this simple connection technique, a contact resistance of
79p.Qcm2 was found. Since tin was thought to be a p-type dopant for these
materials, there was concern of using tin on the LAST materials, and future

experiments focused on tin to LASTT p-type samples.

42

Chapter 6 Diffusion Furnace

6.1 Designing a Diffusion Furnace System

A three zone furnace was utilized in the design of this system to
help obtain a uniform temperature region in the center of the furnace. The size of
the furnace was chosen to accommodate several large thermoelectric generator ‘
modules and/or the individual unicouples for such modules for simultaneous
diffusion bonding of the contacts. Based on this a center zone of 12 inches long
and the two outer zones 6 inches long was chosen. The bore size of the furnace
was chosen to accommodate two or three modules side by side requiring a bore
size of at least 5.125 inches in diameter.

To control the environment around the modules during bonding, a quartz
tube was included in the design, and the length of the tube was chosen such that
o-ring seals at the ends of the tube could be made without the need for water
cooling. To accomplish this, the quartz tube was selected to extend an extra two

feet out each end of the furnace (8’ quartz tube in total length).

43

 

Figure 21: Diffusion furnace with quartz tube
The endcaps for the quartz tube were designed to be easily removable,
provide a vacuum seal of lower than 10mTorr, to be relatively light, and to have
gas inlets and vacuum ports to control the environment in the chamber. In
addition, there was interest in investigating diffusion bonding at atmospheric
pressure of argon. To maintain the position of the endcaps during backﬁlling of
the argon, a structure was built with angle brackets to keep the end caps in place

as shown in Figure 22.

 

Figure 22: Gas inlet end cap

44

The end cap shown in Figure 21 is the gas inlet end cap. Welded to the 6
inch aluminum iso-ﬂange is a QF-25 vacuum connecter which is in turn attached
by a ‘A gas line to a cylinder of argon. The iso-ﬂange is easily removed with
eight bolts. The end cap was designed to seal to the outside of the quartz tube

by compression of a set of o-rings as shown in Figure 23.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I I7 O-rings
I : / I e7 S
I ' ”TIIT'TOIH OI
_- I LOUNtzTube
I l I
II I; #0 I, IGNIH..-
I II _
rel \/
Boltsbetweenthese
partscorrpresstheo—rings.

L ISO Flange

Figure 23: Endcap design exploded view.

45

 

Figure 24: Vacuum connect end cap and pressure gauge attachment

The vacuum port end cap mirrors that of the inlet, but connected to this
end is the vacuum pump, TC pressure gauge, and a valve. The vacuum pump is
a tri-scroll dry roughing pump with a base pressure of around 10mTorr. The
valve is used to seal the system and allow the fumaoe to run independent of a
vacuum pump. This is used when the tube is ﬁlled with a gas and an

overpressure is desired.

46

 

Figure 25: TC gauge for diffusion furnace

The thermocouple gauge is used to monitor the pressure inside the quartz
tube. After backﬁlling the quartz tube with argon, the thermocouple vacuum
gauge becomes inaccurate. This is due to the change in the thermal conductivity
of the gas in the furnace. The proper trends are still read by the gauge, an
example being the pressure rising and falling during heating and cooling.
Occasionally the bolts that squeeze the o~rings need to be retightening to keep a
good seal. Also added to the end of the quartz tube is an o-ring used as a
cushion to keep the aluminum iso-ﬂange from chipping the very end of the quartz
tube during vacuum (a soft bump for the quartz tube). To test the vacuum seals
to the quartz tube, a base pressure of 3mTorr was reached, and the system was
then valved off and found to be able to sustain vacuum over night.

The furnace temperature controller consists of three PID controllers (one
for each zone) which regulate the power to each zone through a three phase
solid state relay. The controllers receive the input by three type K thermocouples

placed in the center of each of the three zones.

47

 

Figure 26: Temperature Controllers

The controllers can be set to slowly increase temperature at a user selectable
ramp rate, and a desired soak time. This makes the furnace very easy to
operate, and does not need to be watched during a run. It is good to make note
of the pressure periodically, and to keep the end caps cool during the diffusion
process by a fan. Desorbed material during diffusion collects on the colder
regions of the quartz tube just outside the furnace, however the center zone
remains relatively clean due to the low sticking coefﬁcients at the higher

temperatures of the center zone.

48

 

Figure 27: Overall diffusion bonding system showing the quartz tube, the inlet endcap, and the
thermocouples for each zone on top of the furnace.

For added safety, a stainless steel top for the table on which this diffusion
furnace system sits was fabricated. Also, within the quartz tube, circular disks of
stainless steel were inserted to further insulate the center zone, and help to
maintain a more isothermal region in the center. This was found to be

particularly helpful with such a large diameter quartz tube chamber.

6.2 Unicouple Boat

To place the thermoelectric materials inside the diffusion fumaoe and
bond them to the electrodes, a quartz “boat” was designed. While the system is
capable of bonding an entire module in a single run, if any of the legs forms a
bad contact or cracks during the diffusion process, the entire module will be
inoperable. By designing a quartz boat individual unicouples could be fabricated,

and tested before assembling into the ﬁnal module.

49

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5rrm 5rrm
i ‘ 5V “ "‘
Closeupend

Figure 28: Preliminary drawings for unicouple boat

The boat was designed to hold up to 140 unicouples at once to rapidly fabricate

many unicouple in a single diffusion run.

50

 

Figure 29: Unicouple Boat

The boat also makes it easy to run many different experiments at the same time.
It is large enough to separate different bond attempts so that one bond will not
contaminate the other. The only negative aspect of this boat is that it is
fabricated to only hold one size unicouple, 5x12mm. If the sizes were to change
a whole other boat would have to be built. This unicouple boat still does serve its

purpose well.

6.3 Furnace Run 1
For the ﬁrst run six n-type and six p-type samples were ﬁrst dipped into a
ZnCl solution and then placed into an antimony powder. The ZnCl was used as

a ﬂux to aid in the bonding process.

51

 

Figure 30: Design diagram

The samples were placed onto nickel strips and then placed into the
furnace for bonding. The furnace was pumped down to a base pressure of
around 10mTorr, and then argon was allowed to ﬂow into the chamber at a
constant rate of 5 liters per minute. The furnace was raised to 650°C. The whole
run had a continuous ﬂow of gas, and the overall pressure of the system never
got above 3Torr. After the furnace sat at 650°C for about 10 minutes, the
temperature was set to 22 °C and allowed to cool over night.

No bonds resulted. The materials also out gassed and changed colors.

 

Figure 31: Run 1 results

52

The nickel also changed to a pinkish color. The conclusion was that the run was
done at too low of pressure and also some oxygen may be leaking into the

system.

6.4 Experiment 2

The second bonding experiment did not use an antimony powder, but a

bar of antimony about 0.5mm thick.

4— antimony

4_ nickel

 

Figure 32: N and P material to antimony bar all to a nickel stn'p

This experiment resembles the ﬁrst nickel antimony bonds, however here it was
done at a lower temperature and a higher pressure to minimize outgassing from
the samples. The chamber was ﬁrst pumped down and then the furnace was set
the temperature to 250°C. Once the furnace reached 250°C a ﬂow of argon gas
into the chamber was initiated to raise the pressure to 300Torr and then sealed
off the tube. The temperature was then raised to 575°C and maintained there for

approximately 10 minutes before cooling to room temperature over night.

53

 

 

Figure 33: Bonded unicouple with antimony bar to nickel

As seen in Figure 33, a bond between the samples and the electrode was
formed. The contacts did, however show some signs of roughening, and ﬂakes
of metal could be easily removed from the electrode. The higher pressures were
found to be helpful for preventing outgassing from the LAST and LASTT

materials.

6.5 Furnace Operation

After a few runs of the furnace, it became apparent that many of the same
routines were being repeated, with only small changes in pressure, maximum
bonding temperature, and bonding time. The following is a list of general
instructions to operate the diffusion furnace.

1. Load samples for bonding into boat and then place into the center of the
hot zone in the furnace.
2. Bolt the iso-ﬂange in the order speciﬁed on the front of the ﬂange.

3. Attach the gas line from the cylinder to the front of the iso-ﬂange.

54

10.

11.

Close the vacuum valve on the outlet port of the furnace.
Turn the rough pump on.
Slowly open the valve to make sure the air is not initially evacuated too
fast.
After the base pressure begins to fall, open the valve all the way to ensure
maximum vacuum ﬂow.
Do a quick inspection to make sure there are no leaks in the system.
Let the furnace pump down for about an hour, or until a pressure of near
20mTorr is reached.
Turn on the gas ﬂow and open the cylinder and allow gas to ﬂow through
the furnace. This process will help sweep away oxygen, water, and
hydrocarbons that remains in the tube.
While the gas is ﬂowing, set the temperature controllers as follows

a. Hit the scroll key once to set point 1

b. Hit enter

c. Set the desired maximum bonding temperature

d. Hit enter, and the controlled will display, “stored”

e. Scroll to “conﬁg” and hit enter

f. Scroll to “ramp” and enable it

9. Enable soak

h. Then enter the desired ramp time. The controllers are programmed

to input the total time to reach the maximum temperature in hours

and minutes

55

12.

13.

14.

15.

16.
17.
18.

19.

i. Enter the soak time in hours and minutes

j. The controllers are now set, and when the run is completed it will
automatically shut off your furnace.

Allow the gas to ﬂow for about 3 minutes, then shut the vacuum valve
completely and turn off the scroll pump.

Let the inert gas ﬁll the chamber till it has reached the desired starting
pressure. It has been found that above 300Torr samples show little
outgassing during the run.

Turn off gas ﬂow and close cylinder valve.

a. It takes at least 8 to 10 hours for the furnace to completely cool, so
it is good to put runs in during the afternoon so the furnace is cool
the following morning.

When the fumaoe has cooled to a reasonable temperature (below 70°C)
turn the vacuum back on and open the outlet valve.

a. This procedure is done to pump out residuals that may reside
inside the fumaoe that are not safe to breath.

Close vacuum valve

Shut pump off and remove vacuum hose

Open outlet valve slowly to allow the furnace to vent

Once back to atmosphere, carefully remove the inlet iso-ﬂange and

remove the sample boat.

This procedure has also be utilized for sintering and thermal etching of samples.

56

6.6 Diffusion Bonds and Results

The following is a list of bond attempts and results of those different

bonds. Photos were taken of each bond, and are available in the appendix.

Table 3: List of diffusion) bonds and results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

BondlmestigatedtoLASl' I Temp(°C)/ Pressue I Visudlnspection

I Time (min) Arz .
ZnClzpre-etclHprowder ‘ 650l10 300mTon NoBond(out-gassing)
Sb 622/10 29Torr Good Bonds (no out-gassing)

I
Bi/Sb(55l45)onN . 500/10 2mTorr NoBonds

i
Bileonly I 50010 300Torr GoodBonds
BiletoN(NoLAST) 50010 :5me Bi/SbblackensonN

I
ﬁ§_m" “ 604/10le 20on Goodbonds
SbonN(NoLAST) ; 604/10 200Torr NotBad-alittlebladtening
222-- - 2 _-. I 2 2 . _ 22 -2 _
Bile(55/45)on‘li I 500110 215Torr Goodbondbetweenmetals
(NOLASD i

i .
Bile(55/45)onw ' 500/10 215l'orr I Nobondbetweenmetals
(NoLAST) I ,
* h(0.5mmthick)onN—‘ _ I 640/10 mm Consumed Samples “M
lnonhi(NoLAST) I 50010 215Torr GoodBondbetweenMetals
_h_onN(pm_—loyed)— E esEio ‘ Sofa; " ' Good“ 3de W H” '—
ln1pmttickonM -2 wéb/TCJ‘_—N450Tm GoodBond 22 “—1
Stainless steel 755/240 250Torr Well Bonded

 

 

57

Indium was found to bond well to LAST with a low contact resistance, therefore
more investigation was done to fabricate unicouples using a thin ﬁlms of indium

between the samples and a nickel electrode.

58

Chapter 7 Module Fabrication

7.1 Indium Thin Film Modules

One of the better mechanical and electrical contacts that has been found
is the thin ﬁlm indium contact. A 1 pm of indium was e—beam deposited onto

nickel electrodes as shown in Figure 34.

 

Figure 34: Indium thin ﬁlm on nickel

Indium has a low melting point of 157°C and can be easily deposited as thick
ﬁlms. The solid stripe down the center of the nickel electrodes in Figure 34 is
where the mask was holding down each individual nickel strip. Bonds formed
with indium to LAST were found to withstand temperatures >600°C suggesting
that an indium compound is chemically forming at the contact during bonding.
The bonding temperature for fabricating the unicouple was set to 630°C
and pressure in the furnace was set to approximately 300Torr. When the run
was completed, the nickel strips were found to be well bonded to the LAST and

LASTT legs. To complete the electrical connections, nickel electrodes were

59

soldered to each leg at the cold side junctions the Cerromatrix solder supplied

from McMaster Carr.

 

Figure 35: Indium-Nickel unicouple

The module resistance was measured to be 10mg, During this measurement,
the electrical current was rapidly ﬂipped from positive to negative current values
to avoid any thermal offset voltages. With an estimated 3m!) per leg, the contact
resistance is approximately 33% of the total resistance of the module. The
contact resistance is such a high part of the module because the total resistance

is only 10 m9.

7.2 Initial Module Fabrication

Utilizing the indium bonding results, a larger module was investigated
consisting of 18 n-type legs and 18 p-type legs in a 4" x 4" conﬁguration. The hot
side and cold side electrodes were made of nickel strips as shown for the cold

side in Figure 36. Each nickel strip used in the module went through a two stage

60

polishing process; a 1000 grit sandpaper followed by 2500 grit sand paper. The

hot side nickel strips had 1 pm indium deposited to them in the PVD.

 

Figure 36: Cold plate and interconnects

The nickel strips were bonded to an alumina plate using silver paste which was

cured on a hot plate to form stronger bonds.

 

Figure 37: Unicouples in boat before diffusion bonding

Twenty unicouples were assembled into the boat and separated by quartz slides.
As done previously with this bond, no weight was added to aid in the diffusion
process. For this bonding run the furnace was initially raised to 150°C under
vacuum to help desorb moisture and residual hydrocarbons. At 150°C the

vacuum valve is closed and argon was allowed to ﬂow into the chamber to raise

61

the pressure to 200Torr. The furnace was then set to 628°C, however the
temperature overshot to approximately 650°C. Then the furnace was set back to
room temperature.

The following morning when the unicouples were taken out of the fumaoe,

a discoloration of the p-type legs (yellow tinge) was noticed as seen in Figure 38.

i -.
3 I »
.MIMIIMQEMIESBQMUHJI

'ﬂmumﬂmmmmmummlsrmwwi

 

Figure 38: Unicouples after removal from the furnace

The sample yellowing was believed to be caused by excess tellurium that
outgassed from the LASTT during bonding. Following this initial investigation the
unicouples were removed from the boat to ﬁnd that all 20 of the unicouples

formed bonds.

62

 

Figure 39: 20 bonded unicouples for module fabrication

To complete the module 18 unicouples were soldered to the nickel strips
previously silver pasted to the alumina plate. This cold side bond was formed

using the Cerromatrix solder and a hot plate.

 

Figure 40: Module fabrication on the hotplate

To help maintain the position of the legs, quartz spacers between unicouple were
used. Once this fabrication was completed, the hotplate was turned off to allow
the solderjunctions to cool. Using an alumina epoxy (Resbond) which cures at
room temperature, the top plate (hot side) of the module was bonded to the

unicouples as shown in Figure 41.

63

 

Figure 41: Completed Module

Initial open circuit testing with a temperature gradient of 176K, the voltage the
module supplied was 407mV. With the 36 legs we got an estimated
therrnopower is 98.4pV/K per leg. This number is very close to the expected
room temperature thennopower. Additional testing revealed that the bond
between the silver and the alumina plate was weak, and after the cold side
alumina plate came off, then several of the hot side junctions also mechanically

failed.

7.3 Indium Foil Modules

To reduce the fabrication time, an investigation of using a thin indium foil
between the sample legs and the nickel electrodes was initiated. This was to be
used in place of the e-beam deposition step. A 50pm thick indium foil. Using the
foil instead of doing a complete deposition would save entire day on module

fabrication. Following the same procedure as for the previous run the results

64

showed signiﬁcant reaction with the sample as shown in Figure 42. Using
smaller pieces of indium foil, showed more promising results as shown in Figure

43.

 

Figure 43: Indium foil bond small pieces

The two unicouples from Figure 43 were put together into a four leg module as
shown in Figure 44. The resistance of the module was found to be

16.82mQ. This would give each unicouple an estimated resistance of

8.4mQ, which is lower per unicouple found for the e-beam deposited thin ﬁlm

indium modules.

65

 

Figure 44: Module formed from the unicouples shown in Figure 43.

Based on these results, a scaled up attempt to fabricate larger modules was

made.

7.4 Indium Foil Module Run
The samples and nickel strips were prepared the same way as for the thin
ﬁlm indium module. For this run 30 unicouples were put into the furnace to be

bonded.

 

Figure 45: Indium foil module pre-bake.

66

The same temperatures and pressures were used as before, however only 12
out of the 30 unicouples bonded this time. Between runs, one adjustment to the
composition was made to reduce the amount of tellurium outgassing. With the

12 unicouples left, six 4 leg modules were made as shown in Figure 46.

 

Figure 46: Modules made with left over unicouples

Tests were done on the modules resistances with the results shown in Table 4.

Table 4: 4 Leg unicouple resistances and calculated contact resistances.

Module # Total Resistance m9 Calculated Contact Resistance

 

67

Further tests are being initiated to investigate these contacts while the unicouple
is under mechanical pressure during bonding.

To better investigate the individual bonds, a scanning probe system was
developed to incrementally measure the voltage drop along the sample and

across the junction.

68

Chapter 8 Diffusion Bonded Contacts lnvesﬁgations and Alloy Bonds

8.1 Improved Contact Resistance Measurement System

As the amount of bonds attempted grew in numbers, the contact
resistance needed to be test as well. The TLM method would not work for these
materials since they are not bonding in the proper conﬁguration for the TLM
measurement. To measure the contact resistance for these bonded junctions, a
voltage scan across the contact was needed. Already in use in the lab was a
room temperature scanning probe that was used to measure electrical
conductivity of materials. It was designed to measure voltage vs. position and

with the proper dimensions electrical conductivity is calculated.

 

Figure 47: Room temperature scanning probe

This sytem was further tested, and a signiﬁcant improvement in the
measurements was found by repositioning the low voltage pin of the

nanovoltmeter used. Positional resolutions as ﬁne as 10 microns could be made,

69

however to scan across the sample this required a signiﬁcant number of

measurements. Sufﬁciently good results were found for scan steps of 100um as

shown for a solder to LAST junction in Figure 48.

Voltage (mV)

 

0.6 IIT1 ITWUV'TTT‘I‘rII'U'I

0.5
0.4
LAST
0 3
Interface
0.2
Nickel "'

 

' Sample
Discontinuity

0.1

   

o I l I I I I l I kl! l I I I I I I

0 1 2 3 4 5
Position (mm)

Figure 48: Room temperature probe scan at 100nm resolution across a solder to LAST material

junction.

A discontinuity in the surface of the sample at the 2.8mm position can be clearly

seen in Figure 48, however the voltage proﬁle at the junction is smooth with a

negligible discontinuity at the 1.5mm position. This method has been utilized to

measure all contact resistance values for diffusion bonded contacts. In addition

to the indium studies described above, several other contacts were investigated.

8.2 Indium Antimony Alloy

7O

A 50-50 weight percent alloy of indium - antimony was fabricated by by a
fellow colleague (Jarrod Short). The alloy was cut to approximately
5mm x 12mm x 1.5mm thick, and bonded at a temperature of 500°C and a
pressure of 338Torr. The resulting bond is shown in Figure 49 where it can be
seen that the alloy melted around the LAST and LASTT legs, and discolored with

a black coating.

 

Figure 49: 1“ alloy diffusion bond

8.3 Silver Diffusion Bonding

Early measurements by the TLM technique showed contact resistances
between silver and n-type LAST of approximately 500p.Qcm2 for un-annealed
contacts. This experiment was to investigate the effects of bonding at a higher
temperature with silver. The silver was bonded at a temperature of 650°C, and

above atmospheric pressure of argon.

71

 

Figure 50: N and P bonds to silver

The results shown in Figure 50 show that the LAST and particularly the LASTT
were partially dissolved by the silver. The n-type held up to the bonding process
a little better, but also shows signs of deterioration near the silver. Scanning
probe measurements across the silver to LAST junction are shown to have very

high contact resistivity values as show in Figure 51.

72

This is a bond of Silver bar to N-Type LAST by way of an indium foil.
The calculated contact resistance is 4.04m Qcm2

 

 

 

0.003 r I T I I 1 I I I I T ﬂ f j I
I r n l 1 I"
r ’ -l
L ..
r LAST Scan .
00025 7 \ o -
0.002 - /' -
(D I . q
S . 1
§ 0.0015 .- 1
:5 : Contact Resistance Calculated from 1
. this Voltage discontinuity ,,
0.001 b .
00005 '_' Silver Swn j
: I . :
O I I I 5 I I I I I I I l I I I I I I

 

K)
O
N
A
O)
on

Position LAST

Figure 51: Silver to N-type LAST voltage proﬁle scan across contact

This contact scan shows a resistance of around 4chm2. This is much too high

for making an efﬁcient device.

8.4 Antimony to LASTT Contact Scans

A bond was done with antimony directly to LAST in order to see what the

contact voltage proﬁle would look like.

73

P-type to Sb scan 50micron resolution

r ;

Voltage (mV)

 

 

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

Figure 52: Antimony to LAST contact Scan

The data shows a large voltage discontinuity at the interface between the LAST
and the antimony and measures a contact resistivity of 525mcm2. This is still
much too high for good module fabrication. The contact resistances previous
results because the temperature and time used during the fabrication processes
are different. A lower temperature was used during this run to help sample
integrity and reduce outgassing.

Figure 53 shows the results of contact resistance measurements for a LAST-Sb-
LAST structure bonded at 590°C at 300Torr for 10 minutes. Previous antimony

bonds were done under in a sealed quartz tube at temperatures around 700°C.

74

Scan of Sb to N-type Room Temperature Scan

 

 

 

 

12

m . I I I I I I I I I I I I I I I r I I I I I I I l I I I ‘
I d
; l=100mA ﬂ/ 1
700 _- 1
60° .- il\ 5
E Antimony E
500 :- 1
E - LAST 1
8. 400 [- 1
(U _ an
E E x E
h / d
300 _ / .
zoo:- If; -:
' This shows a good electrical contact 3
100 ‘ for antimony bonded to LAST "1
a h I I I I I I L LI I I I I I I I I I I I I I I ‘

-2 0 2 4 6 8 10

Position (mm)

Figure 53: LAST-Antimony-LAST Contact Scan

The voltage drop at each Sb—LAST junction is less than 200pV for contact

resistance less than ZmQ. This sample had a cross section of 3.51mm x

5.00mm giving a contact resistivity of less than 461 pQ-cmz. This value is similar

to the previous Sb-LAST contact, and is too high for device fabrication. The

fabrication process used early must be used to achieve low enough contact

resistances to build an efﬁcient device.

8.5 BiSb Alloy Contact Scan

An alloy of bismuth — antimony has been found to make a good

mechanical contacts to LAST. The phase diagram for Bi-Sb is shown in Figure

75

53, where it can be seen that for a solder melting point of less than 500°C, the

weight percent concentration of bismuth must be greater than 60%.

Bi-Sb Bismuth-Antimony

Atomic Percentage Antimony

C IO 20 3O 4O 50 60 7O 80 90

700
IZOOF ~ 630.74'

600 , II

.000; . L / /

500 _ // /

 

 

\

 

BOOF ~

400 ,/

700?»

300' /

50°F ' 271.44?

200 ““1“
Bi IO 20 30 4O 50 60 7O 80 90 Sb

D.T. H. Weight Percentage Antimony

 

\
\

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 54: Phase diagram for bismuth - antimony alloy [13].

The bond was fabricated at a temperature of 500°C for 10minutes at 300Torr.

After the run was completed, there were no signs of outgassing of the LAST.

76

Scan of BiSb bondtoN-Type
m.UIIIIIIIIIIIVIIYtiiljiiiiiiiI'T‘YU'itYr.

 

Contact displays a large contact resistance //

/-’7
l

IIIIJIIIIIIIJIII Ill-

 

 

 

 

E
" LAST
0
a
g
o
>
Contact discontinuity _-
1
I
4 5 6 7

Position (mm)
Figure 55: BiSb alloy to LAST contact scan

The contact resistance calculated for the BiSb bond is 1.152chm2. Although

the bond strength is strong mechanically, this material has to be abandoned due

to the very high contact resistance.

8.6 Gold Silver Alloy Bond

A 50-50 weight percent alloy of gold - silver was made and bonded to
LAST and LASTT materials as shown in Figure 56. The bond was formed at

650°C for 10 minutes with no pressure applied to the junction during bonding.

77

 

Figure 56: Gold silver alloy bonds

The alloy did bond to the LAST and LASTT materials, however the alloy also
catalyzed the melting of the samples. Low melting point materials such as
indium have been shown to form low resistance contacts to PbTe by forming a
new compound during the bonding process (lnTe in this example) which is then
stable to much higher temperatures [14]. Large quantities of the low melting
point material would consume too much of the sample in forming the new
compound, therefore such materials are best investigated as thin ﬁlms or thin
foils at the interface between the sample and a high temperature metal

interconnect.

78

Chapter 9 New Diffusion Bonding Techniques and Stainless Steel Bonds

9.1 Higher Pressure During Bonding

To investigate the pressure dependence during bonding, a stainless steel
clamping structure shown in Figure 56 was constructed. A 500nm thick tungsten
foil was placed on an n-type LAST sample and clamped between the stainless

steel plates as shown in Figure 57.

 

Figure 57: Stainless steel bonding block

The bond to the tungsten was attempted at 755°C for eight hours. The argon gas
pressure in the chamber remained near atmospheric pressure during the 8

hours. It was found that the tungsten did not bond to the LAST, however the
LAST material did strongly bond to the stainless steel plate as shown in Figure

59.

79

 

Figure 58: LAST bonded directly to stainless steel

Although the stainless steel plate was considerably heavier than the LAST
sample, it could be lifted by the LAST sample without breaking it off the stainless
steel. Based on this result further investigations of bonding stainless steel to the
LAST and LASTT materials was initiated. To apply some pressure on the

contacts during bonding, a new sample holder was designed.

9.2 Diffusion Boat Improvements

Prior to beginning more investigations into bonding stainless steel, a way
of adding weight was needed to aid in the bonding process. Glass shop cut
quartz pieces that could individually sit on top of each unicouple to add weight to

the bonding process.

80

 

Figure 59: Added quartz pieces for weights during diffusion bonding

81

9.3 Initial Stainless Steel Diffusion Attempts

 

Figure 60: First diffusion bon of LAST and LASTT to stainless steel
725°C 330Torr for the run

 

Figure 61: N-type bond at 725°C

The p-type turned black and showed signs of pitting.

82

 

Figure 62: P-type bond at 725°C

The conclusion was to keep attempts at lowering the bonding temperature
and increasing pressure. Initial contact measurements gave a contact resistivity
of 140chm2. Later bonds done at lower temperatures (near 700°C) showed

very low contact resistivities to p-type legs as shown in Figure 63.

83

 

0.0007 I I I I I I i i I l I I I I I T I I I I I I I I

0.0006 °
0.0005

0.0004 ‘

Voltage (V)
0

0.0003

a
0.0002 /
0

0.0001 0

O
O
llJlllIlllIIIIIIlIIIlllIllIIIIIII

 

o I l l I l I I I l l l l l I I I l l I I I I I

0 1 2 3 4 5

Position (mm)

Figure 63: P-type to stainless steel contact scan

The bond shown in Figure 63 was formed using additional weight during the

bonding process.

84

 

 

 

 

O.m025 I I I I l I I I I l I I I I I I I I I l I I I I I I I
- 1
- -l
0.0002 b / . a
- 'l
: - '
A 0.00015 - -
2, l .
:6 ' 0.
0.0001 " . "‘
P . u
l' l
510“5 - -
. e r
.4” j
0 I LI IHI l I I I I I I I I I
—1 0 1 2 3 4 5

Position (mm)

Figure 64: N-type to stainless steel contact scan

Bonding stainless steel to the n-type (LAST) samples also showed a
relatively low contact resistivity (Figure 65). To further investigate the inﬂuence

of pressure during the bonding process, a new unicouple boat was designed.

9.4 New Unicouple Boat

A new unicouple boat was design to give each individual unicouple an
amount of pressure through adjustments of a set screw in order to make a good

interface bond. A side proﬁle of this new boat is shown in Figure 65.

85

 

Figure 65: New unicouple boat

The ﬁrst bonds made from the boat showed lower contact resistances of 50-

100chm2 and increased mechanical strength.

9.5 New Reducing Gas

To help remove oxides at the interface, a (5% H2) / (95% Arz) gas mixture

was used in place of the pure argon gas.

9.6 Stainless Steel Bond Investigations

The following is a table listing the bonding temperature and times used to
bond LAST and LASTT to stainless steel. Chromium was also investigated for
bonding to p-type LASTT for comparison. All contact resistance values were

measured using the room temperature scanning probe also developed as part of

86

this work. Photos of some of the bonds investigated can be -found in Appendix A,
and voltage scans for several of the experiments listed in Table 4 can be found in

Appendix B.

Table 5: Diffusion bonds and corresponding contact resistances

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Bridge Bonding Bonding Pressure Contact
Material Temp °C Time Hrs. Torr Reisistance pﬂcm2
PISS 690 1 92 274
PISS 690 1 92 1 95
N/SS 690 1 92 131
N/SS 690 1 92 75

N-SS 690 2 80.3 250
P-SS 690 2 80.3 129
P-CHR 690 0.5 138 112
N-CHR 690 0.5 138 Cracked
N-SS Hand 690 3 168 ~5
Tighten

N-SS new 690 0.75 300 Cracked
boat (NB)

P-SS NB 690 0.75 300 70
P-SS NB 1 297 ~5
P-SS NB 1 297 90

 

 

 

 

 

87

 

Chapter 10 Summary and Future Work
10.1 Summary

This thesis focuses on the investigation of electrical contacts to new
thermoelectric materials for the development of power generation devices. As
part of this effort systems and methods for measuring contact resistances have
been established, including a transmission line technique and a scanning probe
technique. The scanning probe technique is useful for conﬁgurations which are
amiable to device fabrication, and the transmission line technique gives
increased accuracy on the measurement. Many different alloys have been
deposited and measured for contact resistance (Table 2). A diffusion bonding
system has been designed and assembled for investigating high temperature
contacts through diffusion bonding. This system has been designed to
accommodate several modules during the bonding process and/or individual
unicouples. Utilizing this system various metals and alloys have been
investigated for forming these high temperature bonds and tested with the
measurement systems developed (Table 3). Several promising contacts have

been identiﬁed with contact resistivities less than 60pQ-cm2.

10.2 Future Work

Diffusion bonding studies have indicated the formation of tellurides during
bonding, therefore future work will include depositing excess tellurium at the
contact interface to assist in the bonding process and help prevent tellurium

diffusion out of the semiconductor samples. The investigation of lnTe and SnTe

88

contacts will also be further investigated. Pressurization of the diffusion bonding
system during bonding has signiﬁcantly helped to reduce outgassing from the
samples, and high temperature sealant coatings will be investigated for

encapsulation of the samples to further prevent them from outgassing.

89

APPENDIX A

Diffusion Bonds and Results

  
    

‘ 6... » , . .--_ as. M?”
Figure A 1: ZnCIz pro-etch + Sb powder, no bond (out-gassing)

FigureA 2: Sb, good bonds(no ut-gassing)

90

 

Figure A 3: Bile (55l45) on Ni, no bonds

 

    

ﬁys

- . - ‘
0 LAST), iISb blackens on Ni

 

FiguFeWst BiISb to Ni (N

91

 

Figure A 7: Sb on NI (No LAST), Not Bad- A little blackening

Figure A 8: Bile (55l45) on Ti (no LAST), Good bond between metals

92

 
 
 

Figure A879: 3qu (55145) on w (no LAST), no bond between metals

Figure A 10: In (0.5mm thlck) onNI, Consumed samples

Figure A 11: In on'NI (NolL’AS-T) Good bond between metals

 

Figure A 12: In on Nickel to LAST (prealloyed), Good Bond

93

 

Figure A 13: In 1mm thick on Ni, Good Bond

 

Figure A 14: Stainless steel to LAST, good Bond

94

Voltage (V)

APPENDIX B

Contact Resistance Scans

 

0.0005 I I I I l—I 1 I I T I I I I l I I I I I I I I I l I I I I

0.0004

 

 

 

)— .-
0.0003 l- '-
. g 5
0.0002 F g -
P I
- 'l
0.0001 - I 4
I 0’" 2
I o I
0 ‘M I I I4 I I l I I I I l I I I I I I L I_I
-1 0 1 2 3 4 5

Position (mm)
Figure B 1: N-type to Stainless Steel, 690°C 1 Hour, 130mm2

95

Voltage (V)

0.0008

0.0007

0.0006

0.0005

0.0004

0.0003

0.0002

0.0001

 

IIII'IIIIIIIII'IIrI'IIII'IIII'II

 

I III'IIIrjIIII'IIrI' IITI'IIII

I I I I I I I I I I I I I III j I I I I l I I I I I LIJ

IIIlIIIIIJIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

 

I
—L

0

1

2

Position (mm)

3

4

Figure B 2: P-type to Stainless Steel, 690°C 1 Hour, 195pﬂcm’

96

U1

Voltage (V)

0.0005

0.0004

0.0003

0.0002

0.0001

 

I I I I l I I I I l I I I I l I I I I I

 

I I I r' I I I I l I I I I l I I I I l I I I I l I I I I

I I I I I I I I I I I I I I I I I I I I I I I I I I4 I I

I I I I I I I I I I I I J. I I I I I I I l I I I

 

 

I
—L

0 1 2 3 4

Position (mm)

Figure B 3: P-type to Stainless Steel, 630°C 4 Hours, ~5uoem2

97

0'1

Voltage (V)

 

   

 

 

 

0.m025 I I I I l I I I I I I I III I I I I l I I I I I I r ‘
0.0002 - / O -
: - 3
0.00015 - -l
- e l
I o 1
0

0.0001 -
b O. .1
510'5 - -
. e .
. / .1
o I I I I I N I I I I I I IJ I I L I I I I I I I I I I d
-1 0 1 2 3 4 5

Position (mm)
Figure B 4: N-type to Stainless Steel, 700°C 4 Hours, ~5pﬂcm2

98

BIBLIOGRAPHY
[1] H.J. Goldsmid, “Thermoelectric Refrigeration.” Plenum, New York, NY, 1964.
[2] H.J. Goldsmid, “CRC Handbook of Thermoelectrics,” ed. D.M. Rowe (1995)

[3] Sim Loo, “Transport Characterization of New High Temperature Quaternary
Thermoelectric Materials And Devices”, Ph.D. thesis, Michigan State University,
2003.

[4] M. G. Kanatzidis, “The Role of Solid-State Chemistry in the Discovery of New
Thermoelectric Materials,” in Solid State Physics, ed. H. Ehrenreich, F. Spaepen,
vol. 69, pp 51-100, Academic Press, New York, 2001.

[5] Kuei Fang Hsu et al., “Cubic Angme Tam, Bulk Thermoelectric Materials
with High Figure of Merit,” Vol. 303, Science, www.sciencemag.org

[6] A. F. loffe, Semiconductor Thermoelements and Thermoelectric Cooling,
lnfosearch Limited, London, 1957.

[7] L. D. Hicks, M. S. Dresselhaus, "Effect of Quantum-Well Structures on the
Thermoelectric Figure of Merit," Physical Review B, Vol. 47, No. 19, pp.
12 727-12 731, 1993.

[8] T. C. Harman, M. P. Walsh, B. E. LaForge, G. W. Turner, “Nanostructured
Thermoelectric Materials,” Journal of Electronic Materials, 34 (5), pp. L19-L22,
2005.

[9] M. H. Cobble, “Calculations of Generator Performance,” CRC Handbook of
Thermoelectrics, CRC Press, 1995.

[10] H.H. Berger, “Models for Contacts to Planar Devices,” Solid State
Electronics, 1972, Vol. 15, pp. 145-158.

[11] H. E. Bates, F. Wald and Martin Weinstein, “The Bonding of Lead Telluride
with Non-Magnetic Electrodes,” IEEE/AIAA Thermoelectric Specialists
Conference, May 17-19, 1966.

[12] DA. Porter and K.E. Easterling, “Phase Transformations in Metals and
Alloys,” 1993, Chapman and Hall.

[13] Massalski, T.B., “Binary alloy phase diagrams,” Materials Park, Ohio : ASM
lntemational, c1990.

99

[14] Hittman.Associates, Inc. “Thermoelectric Bonding Study,” Report
Prepared for National Aeronautics and Space Administration, Goddard Space
Flight Center, under Contract NASS-3973, 1964.

100

   

llllljlllll‘lllllilllilljfl