Pressureless Silver Nanopowder Sintered Bond for Liquid

Cooled IGBT Power Module for EVs and HEVs

by

Namjee Kim

A thesis submitted in conformity with the requirements

for the degree of Master of Applied Science

The Edward S. Rogers Sr. Department of Electrical and Computer Engineering

University of Toronto

© Copyright by Namjee Kim 2018

ii

Pressureless Silver Nanopowder Sintered Bond for Liquid Cooled IGBT

Power Module for EVs and HEVs

Namjee Kim

Master of Applied Science

The Edward S. Rogers Sr. Department of Electrical and Computer Engineering

University of Toronto

2018

Abstract

Pressureless silver nanopowder sintering of the bonding layer between the insulated gate bipolar

transistor (IGBT) and the direct bonded copper (DBC) for the liquid cooled power modules in

electric vehicles and hybrid electric vehicles is studied. The pressureless silver nanopowder

sintering is analysed using the Differential Scanning Calorimeter / Thermogravimetry /

Simultaneous Thermal Analysis and the Scanning Electron Microscopy. Based on the analysis

results, the sintering is optimized at 200 ˚C for 60 minutes for reliable die attachments. The

bonding of the sintered silver to the IGBT is confirmed by intermetallic layers composition

analysis. The nanopowders are classified by the average sizes into the micro and nano-meter scaled

groups. The porosity of 30% is computed within the sintered silver layer. Overall, the experimental

results verify the feasibility of the pressureless silver nanopowder sintering of the bonding layer

between the IGBT and the DBC.

iii

Acknowledgments

I would like to express my deepest gratitude to my research supervisor, Professor Wai Tung Ng,

for his constructive guidance, continuing support and the patience throughout this project. It is a

great pleasure to have a chance working in the Smart Power Integration & Semiconductor Device

Group under his strong leadership. His knowledge and vision in power electronics and packaging

encourage me to build the insight of the study.

I would like to express my sincere gratitude to Professor Francis P. Dawson, Professor Thomas

W. Coyle and Professor Glenn D. Hibbard for their valuable advices and comments on my project.

Without their supports the project would have not been successfully driven.

I would like to express my appreciation to everyone involved in the project at Dana Ltd., Henkel

and Rogers Co. for their sponsorship.

Special thanks to the current and previous graduate students and researchers of the Smart Power

Integration & Semiconductor Device Group for their supports and companionship over the years.

I would like to thank graduate students of the Cellular Hybrid Materials Research Group and staff

of the TNFC for their technical supports as well.

Lastly but not least, I wish to thank my family and friends for their supports and encouragements.

I would like to acknowledge the financial supports from the Ontario Centre of Excellence (OCE)

and the Natural Sciences and Engineering Research Council of Canada (NSERC).

iv

Table of Contents

Acknowledgments.......................................................................................................................... iii

Table of Contents ........................................................................................................................... iv

List of Tables ................................................................................................................................. vi

List of Figures ............................................................................................................................... vii

List of Acronyms ........................................................................................................................... xi

List of Symbols ............................................................................................................................ xiv

Chapter 1 Introduction ..............................................................................................................1

1.1 Motivations ..........................................................................................................................1

1.2 Thermal Management ..........................................................................................................4

1.3 Research Objective and the Thesis Organization ..............................................................10

Chapter 2 Literature Review...................................................................................................12

2.1 Background ........................................................................................................................12

2.2 Sintering Theory ................................................................................................................13

2.2.1 Polymer Burnout ....................................................................................................13

2.2.2 Theoretical Analysis of Sintering ..........................................................................14

2.3 Sintering Process Management ..........................................................................................21

2.3.1 Sintering Temperature ...........................................................................................21

2.3.2 Sintering Pressure ..................................................................................................22

2.4 Nanopowder Sintering Challenges ....................................................................................23

v

2.5 Direct Bonded Copper (DBC) ...........................................................................................26

2.6 Chapter Summary ..............................................................................................................29

Chapter 3 Experimental Results and Discussion ....................................................................31

3.1 Experiment Design.............................................................................................................31

3.1.1 Structure Design.....................................................................................................31

3.1.2 Sample Preparation ................................................................................................34

3.1.2.1 Thermal Budget .......................................................................................37

3.1.2.2 Nickel/Gold Plating .................................................................................37

3.1.2.3 Silver Bonding Layer Process .................................................................41

3.1.2.4 Silver Nanopowder Pasting .....................................................................41

3.2 Silver Nanopowder Paste Analysis ....................................................................................43

3.2.1 DSC/TGA/STA ......................................................................................................44

3.2.2 SEM .......................................................................................................................47

3.3 Chapter Summary ..............................................................................................................56

Chapter 4 Conclusions and Future Work Plan .......................................................................58

4.1 Conclusions ........................................................................................................................58

4.2 Future Work Plan ...............................................................................................................61

References ......................................................................................................................................63

vi

List of Tables

Table 1.1 Thermophysical Properties of Liquid Coolants at Room Temperature [10] ................ 10

Table 2.1 Sintering Mechanisms in Solids [29]–[31] ................................................................... 18

Table 3.1 DBC Sample Data at 20 °C [51] ................................................................................... 36

Table 3.2 Summary of the DSC/TGA/STA Results ..................................................................... 47

Table 3.3 Weight Percentage at EDS Spots within the Intermetallic Layers ............................... 49

Table 3.4 Atomic Percentage at EDS Spots within the Intermetallic Layers ............................... 49

vii

List of Figures

Figure 1.1 The global electric car sales volume, 2010-2016 .......................................................... 1

Figure 1.2 DC bus voltage for the electric vehicle using boost converter ...................................... 2

Figure 1.3 The cross sectional view of an n-channel thin wafer IGBT cell structure .................... 3

Figure 1.4 Temperature dependence of failure rate of a power semiconductor device .................. 4

Figure 1.5 Illustration of a heat sink ............................................................................................... 7

Figure 1.6 Ranges of heat transfer coefficients for liquid coolants and cooling methods ............. 9

Figure 2.1 Evolution of the relative density during sintering process .......................................... 14

Figure 2.2 Schematics of the sintering process, (a) initial, (b) intermediate, and (c) final ........... 14

Figure 2.3 Geometrical evolution and instantaneous surface vacancy concentration level for the

two initially spherical crystalline particles of equal size sinter by combined surface, volume and

grain boundary diffusion (a) t=0, (b) t=0.1, (b) t=1, (d) t=5 (e) t=20 and (f) t=100 .................... 16

Figure 2.4 Definitions of radius of curvature and the center of curvature ................................... 17

Figure 2.5 Schematic diagram of two particles sintering model (a) diffusion paths and (b)

Dihedral angles at the initial state (120°), equilibrium state (150°) and final state (180°) ........... 19

Figure 2.6 Equisided tetrakaidecaheraon geometry ...................................................................... 20

Figure 2.7 Driving forces for nanopowder consolidation as a function of grain size .................. 23

viii

Figure 2.8 (a) Agglomeration and (b) Aggregation of the nanopowders .................................... 25

Figure 2.9 Sintering results from densification and non-densification diffusion ......................... 25

Figure 2.10 Changes in densification rate in temperature ............................................................ 26

Figure 2.11 Schematics for the eutectic bonding copper process and the Cu-O phase diagram .. 27

Figure 2.12 Cu/Al2O3 void formed at the interface ...................................................................... 28

Figure 3.1 Cross sectional view of the proposed IGBT attachment on a DBC substrate using

silver nanopowder sintered. The DBC is bonded to a liquid cooled aluminum heat sink ............ 33

Figure 3.2 Perspective view of the proposed IGBT attachment on a DBC substrate using silver

nanopowder sintered ..................................................................................................................... 33

Figure 3.3 The DBC design and the actual patterned DBC for a single IGBT die ....................... 34

Figure 3.4 The DBC design for a Half-bridge IGBT (with free wheeling diodes) module

assembly ........................................................................................................................................ 34

Figure 3.5 CTE differences in two layers produces interlayer stresses, leading to potential failure

....................................................................................................................................................... 36

Figure 3.6 Ni and Au electroplating wet bench set up. ................................................................. 38

Figure 3.7(a) Electroplating circuit for Ni and (b) Anode ............................................................ 39

Figure 3.8 DBC substrate samples in conditions of (a) bare (Cu on top), (b) Ni on top, and (c) Au

on top with the IGBT die mounted on pressureless sintered silver nanopowder with the Au wire

connections. .................................................................................................................................. 40

ix

Figure 3.9 Process flow for the IGBT die attachment process using silver nanopowder sintering

....................................................................................................................................................... 41

Figure 3.10 Illustration of the silver nanopowder paste stencil printing technique ...................... 42

Figure 3.11 Temperature profile for drying and sintering processes ............................................ 43

Figure 3.12 DSC/TGA/STA results for the silver nanopowder sintering experiment .................. 45

Figure 3.13 Isothermal DSC/TGA/STA results for silver nanopowder sintering at 200 ˚C ......... 46

Figure 3.14 Cross section image of the silver nanopowder sample sintered at 200 ˚C for 60

minutes, captured using Quanta FEG250 ESEM (magnification 500 X) ..................................... 48

Figure 3.15 (a) Cross-sectional view of sintered silver (magnification 10k X), (b) IGBT die and

sintered silver intermetallic layers EDS analysis on selected areas .............................................. 48

Figure 3.16 . (a) SEM cross-sectional image (magnification 5k X) for the porosity analysis and (b)

the porosity calculation result ....................................................................................................... 50

Figure 3.17 Sample preparations for the top view SEM image analysis along the temperature

profile ............................................................................................................................................ 51

Figure 3.18 Top view SEM images of the silver nanopowder paste samples with heating-up

process stopped at different temperatures before sintering .......................................................... 52

Figure 3.19 Top view SEM images of the silver nanopowder paste samples sintered at 200 ˚C (a)

for 1 min, (b) for 20 min, (c) for 40 min and (d) for 60 min (magnification 10k X) .................... 54

x

Figure 3.20 Images of grain sizes observed on the micro-sized powder after sintering at 200 °C

for 60 minutes, (a) magnification 80k X and (b) magnification 200k X. ..................................... 55

Figure 3.21 Top view SEM images of (a) the dried as-received silver nanopowder paste after the

heating stopped and maintained at 120 °C for 1 minute (magnification 5k X), (b) the silver

nanopowder paste sintered at 200°C for 60 minutes (magnification 5k X), (c) silver nanopowder

paste sintered at 200°C for 10 minutes (magnification 20k X), and (d) silver nanopowder paste

sintered at 200°C for 60 minutes (magnification 20k X) .............................................................. 56

xi

List of Acronyms

AF Acceleration Factor

Ag Silver

Al Aluminum

Al2O3 Alumina or Aluminum Oxide

AlN Aluminum Nitride

Au Gold

C Carbon

Cl- Chloride Ion

CTE Coefficient of Thermal Expansion

Cu Copper

Cu2O Copper Oxide

CuAlO2 Copper Aluminate or Copper Aluminum Oxide

DBC Directed Bonded Copper

DC Direct Current

DIW De-Ionized Water

DSC Differential Scanning Calorimeter

DTG Derivative Thermogravimetry

EDS Energy Dispersive Spectroscopy

EG Ethylene Glycol

EHD Electrohydrodynamics

ESEM Environment SEM

xii

EV Electric Vehicle

FAST Field Activated Sintering Technique

H+ Hydrogen Ion

HCl Hydrochloric Acid

HEV Hybrid Electric Vehicle

HRSEM High Resolution SEM

IGBT Insulated Gate Bipolar Transistor

KFO Potassium Formate

MOSFET Metal-Oxide-Semiconductor Field-Effect-Transistor

MTF Mean Time to Failure in Hours

Ni Nickel

O Oxygen

OH- Hydroxide Ion

PC Pulse Current

PCB Printed Circuit Board

PG Propylene Glycol

rpm Revolutions per Minute

SEM Scanning Electron Microscopy

Si Silicon

Si3N4 Silicon Nitride

SiC Silicon Carbide

SO42- Sulfate Ion

STA Simultaneous Thermal Analysis

TG Thermogravimetry

xiii

TGA Thermogravimetry Analysis

Ti Titanium

TIM Thermal Interface Material

TiO2 Titanium Oxide

ZrO2 Zirconium

xiv

List of Symbols

Rds,on Drain to source on-resistance

R1 & R2 Principal radius of curvature

B Coefficient of the mechanism characteristic

C Vacancy concentration under a curved surface

C0 Equilibrium vacancy concentration

D Particle diameter

ΔTt Test temperature difference between highest and lowest

ΔTo Operating temperature difference between on and off state

EA,v Activation energy of the vacancy formation

EA Activation energy (eV)

FT Failure factor

J Current density (A/cm2)

KB Boltzmann constant

L Length between two particle centers

P Vapor pressure

Pex External pressure

pH Potential of hydrogen

R Gas constant (8.314 J/K·mol)

r1 & r2 Particle radius

t Time

Tm,b Melting temperature of the bulk material (K)

xv

Tm,p Melting temperature of the powder (K)

Tmax,o Maximum operation temperature (K)

Tmax,t Maximum test temperature (K)

Ts Sintering temperature (K)

X Neck diameter

α Coefficient of the geometric and environmental factors

γ Surface energy

δ Material dependent parameter (1.8 - 2.65 nm)

ρ Density

σ Stress

Ω Atomic volume

1

Chapter 1

Introduction

1.1 Motivations

As the awareness on climate change and other environmental issues increases dramatically, the

reduction of greenhouse gas emission is one of the most pressing global interests. Electric vehicles

(EVs) and hybrid electric vehicles (HEVs) are regarded as the key products to lower carbon

emission within the automotive industry [1]. The production volume of EVs and HEVs is on an

exploding trend in the twenty-first century as shown in Figure 1.1 [2]. These worldwide activities

accelerate the development of high performance EVs and HEVs with better efficiency and longer

range. The automotive industry’s effort in reducing greenhouse gas emissions and limiting the

environmental pollution has also created exciting opportunities for the power semiconductor

market.

Figure 1.1 The global electric car sales volume, 2010-2016 [2].

2

The insulated gate bipolar transistor (IGBT) is a special type of power semiconductor device for

delivering electrical power to the drive trains, and it is the device of choice in most EV and HEV

power inverters and converters over the conventional power metal-oxide-semiconductor field-

effect-transistor (MOSFET) due to its better current conduction capability [3]. Common

applications of the IGBTs in EVs and HEVs include inverters for propulsion motors or generators

and DC/DC voltage boost converters for batteries or fuel cells.

IGBTs are used in the inverter to control the flow of the electrical power from the battery to the

propulsion motor in the EVs and HEVs. The commercially available IGBTs are designed to

optimize the power transfer efficiency from the DC bus to the electric motor with a low on state

voltage drop, fast switching speed and a high avalanche breakdown capability. Also, IGBT boost

converters are used to enhance the DC bus allowing more power available to the motor. As power

is calculate by multiplication of voltage and current, if the boost converter raises the DC link

voltage by 2.5 times, the torque and the power output of the motor is also raised by 2.5 times

without increasing the motor current, as the increased DC bus voltage is supplied to IGBT based

inverters to drive the generator and the motor [4]. The topology of the boost converter is illustrated

in Figure 1.2 below

Figure 1.2 DC bus voltage for the electric vehicle using boost converter [4].

3

The IGBT is a power semiconductor device with three terminals; gate, cathode and anode,

alternately known as the gate, emitter and collector, respectively. It is essentially a merged MOS-

bipolar device where the current carrying pnp bipolar transistor is controlled by a MOSFET. The

cell structure of IGBT is shown in Figure 1.3.

One of the major limitations of the power MOSFET is that the value of drain to source on resistance

(Rds,on) increases as the device voltage rating. This is because the doping in the drift region is

inversely proportional to the required breakdown voltage. Thus, the doping decreases as the

voltage rating increases, and the conductivity decreases. The IGBT overcomes this limitation by

using a forward biased pn-junction to inject minority carriers into the reduced doped region to

increase the conductivity. Hence, the structure of the IGBT is similar to the structure of the

MOSFET except that a heavily doped p-layer at the collector terminal is used to inject the minority

carriers [4], [5].

Figure 1.3 The cross sectional view of an n-channel thin wafer IGBT cell structure [4].

4

1.2 Thermal Management

One of the performance indicator for the power converter is its conversion efficiency. The output

power efficiency of the converter is defined as the percentage ratio of the electrical output power

against the total input power. The difference between the input and output power is dissipated as

wasted heat which causes increase in the operating temperature of the power device and the power

module. The typical maximum operation temperature of the silicon (Si) based IGBTs is typically

175 ˚C, and the maximum operation temperature can reach up to 250 ˚C for silicon carbide (SiC)

IGBTs [6]–[8]. To achieve a better power efficiency it is critical to increase the heat dissipation,

and the heat management system is designed to improve the reliability and the electrical

performance [9]. If the rate of the heat removal is not equivalent to or greater than the rate of the

heat generation, the temperature of the device and components increases constantly, therefore

reducing the device reliability and performance, and possibly causing a device failure. As

illustrated in Figure 1.4, the failure factor of a device is derived from the relative failure rate at any

temperature divided by the failure factor at 75 °C, and it increases exponentially with the

temperature increases within the device.

Figure 1.4 Temperature dependence of failure rate of a power semiconductor device [10].

5

The mean time to failure in hours (MTF) due to an increase in the operating temperature can be

estimated using Black’s correlation in forms of:

𝑀𝑇𝐹 =1

𝐴𝐽2 exp (𝐸𝐴

𝐾𝐵(

1

𝑇𝑟−

1

𝑇𝑡)) (1.1)

where A is a constant, J is the current density (A/cm2), EA is the active energy (eV), KB is the

Boltzmann constant, 𝑇𝑟 is the reference junction temperature (K) and 𝑇𝑡 is the junction

temperature during test (K). The Black’s correlation demonstrates that even a slight increase in the

operating temperature can exponentially decrease the MTF and increase the failure rate. The

exponential term in the Black’s correlation, known as the Arrhenius equation, relates how

increased temperature accelerates the aging of the product as compared to the aging under the

normal operating conditions. This term is also called acceleration factor (AF) [10], [11]. The

Black’s correlation is the theoretical failure model for the steady state test temperature. For the

non-steady state test temperature such as the thermal cycle test, the temperature difference between

the highest and the lowest test temperatures and the cycle frequency are addressed in the failure

model.

𝑀𝑇𝐹 =1

𝐴𝐽2 𝐴𝐹 (1.2)

𝐴𝐹 = (Δ𝑇𝑡

Δ𝑇𝑜)

1.9

(𝐹𝑜

𝐹𝑡)

13⁄

exp (𝐸𝐴

𝐾𝐵(

1

𝑇𝑚𝑎𝑥,𝑜−

1

𝑇𝑚𝑎𝑥,𝑡)) (1.3)

Δ𝑇𝑡 is the temperature difference between the highest and the lowest test temperatures, Δ𝑇𝑜 is the

operating temperature difference between the on- and off- states of the testing device, 𝐹𝑡 is the test

cycle frequency (number of cycles per 24 hours), 𝐹𝑜 is the cycle frequency in the device operating

6

condition (number of cycles per 24 hours), 𝑇𝑚𝑎𝑥,𝑜 is the maximum operating temperature, and

𝑇𝑚𝑎𝑥,𝑡 is the maximum test temperature [11], [12].

Most of the thermal management system designs are constrained by two requirements, the high

power dissipation and the size of the power module. The high power dissipation is necessary for

high performance while the high power efficiency and the smaller module size are desirable for a

higher packing density and miniaturization. The local thermal profile and the heat removal

technologies employed are key factors in determining the required heat sinking technique, and

therefore influencing the size and the cost of the module as well [13].

The heat transfer mechanisms for heat removal include conduction, natural convection, forced

convection and other effects such as thermoelectric cooling [13].

The solid state cooling uses solid state bulk materials to remove the heat from the source via heat

dissipation mechanisms. The most common method uses heat sinks, a conduction plate and thermal

interface materials (TIMs), based on the conduction heat transfer mechanism as illustrated in

Figure 1.5. Aluminum heat sinks and a conduction plate are often chosen due to their cost

efficiency, light weight and high thermal conductivity (200 W/m·K). Also, they are less reactive

to the conventional coolants and are non-toxic upon corrosion. A heat spreader maximizes the

contact surface area between the semiconductor and the cooling medium in order to improve the

heat transfer via natural air cooling. The TIMs are used to fill the gaps and to increase the thermal

transfer efficiency by removing the space between the heat sink and the packaging [13].

7

Figure 1.5 Illustration of a heat sink [13].

Magnetic cooling is another method of solid state cooling where changes in magnetic field on a

magnetocaloric material lead to temperature changes. However, the application of the magnetic

cooling is often constrained by the types of materials, volume and weight [13].

The air or gas cooling technologies have the advantages of increases in heat transfer and heat

exchange rates without implementation of the complex structure as often required in the liquid

cooling. The natural air convection is the one of the simplest methods. Assuming the same cooling

module structure as in Figure 1.5 with the heat source sitting on top of the heat sink, the

temperature difference causes the air/gas density to change due to the difference in energy of the

gaseous molecules. This then leads to the decrease in the surface temperature while the air

temperature increases. Only the thermal resistance between the solid surface of the heatsink and

the ambient acts as the controlling factor of the heat dissipation. Despite its advantage in the

simplicity, the heat flux for the natural air convection is as low as less than a few W/cm2. In order

to achieve lower thermal resistances between the heat sink and the ambient, forced convection can

be applied using a fan. The air/gas flow rate increases as the fan rotation speed in revolutions per

minute (rpm) increases. As a consequence, the heat flux is improved. The typical range of the heat

flux for forced air convection heat transfers is about tens of W/cm2 having significant improvement

8

over the heat flux rate of the natural convection. Other air or gas cooling methods employing a fan

using piezoelectric effect, an electrohydrodynamic (EHD) cooler using kinetic energy conversion

based on the corona effect and a thermoacoustic cooler are also available [13].

Liquid cooling is generally preferred over the solid or gas cooling for the cases of the thermal

management in higher power densities. Higher thermal conductivity and the higher heat density

are the main advantages of the liquid coolants. However, liquid cooling has a critical disadvantage

due to the possible leakages of coolants. A mechanical failure of the cooling system may cause

damages to the electronics. Furthermore, flammable or toxic liquid coolants may lead to the

corrosion of the device or module. There are three classifications in liquid coolants according to

their conductivity: where the direct immersion of the hot device is possible, where the direct

immersion is not possible, but a leakage does not damage the electronics, and where the direct

immersion is not possible, and leakage damages the electronics. Therefore, in the liquid cooling

system designs, it is imperative to select the coolant which is non-flammable, nontoxic and with

excellent thermophysical properties including high thermal conductivity, high specific heat, high

heat transfer coefficient and low viscosity.

The mechanisms for heat transfer within the liquid cooling system include free convection, forced

convection, boiling and condensation. The range of heat transfer coefficients of commonly used

coolants are compared in Figure 1.6. Water possesses the highest heat transfer coefficient; thus, it

is the most widely used. However due to its high freezing point at 0 °C it is not ideal for uses in

lower temperature conditions. Moreover, since its volume expands upon freezing water holds a

critical disadvantage to the mechanical design of the cooling system [10], [13].

9

Figure 1.6 Ranges of heat transfer coefficients for liquid coolants and cooling methods [10].

The liquid coolants are classified by their electrical conductivities into dielectric and non-dielectric

groups. Dielectric liquids are electrically insulating materials and non-dielectric liquids are

electrically conducting materials. The dielectric group includes liquids of aromatic base, aliphatic

base, silicone base and fluorocarbons base. The FC72 and the FC77 are two most commonly used

fluorocarbons based liquid coolants for the electronic cooling as they are non-flammable and inert.

Non-dielectric liquid coolants are normally aqueous solutions with high thermal conductivity and

high heat capacity. Water and ethylene glycol (EG) are two most widely used coolants. Propylene

glycol (PG), potassium formate (KFO) and liquid metals such as Ga-In-Sn are also commonly

used coolants. Thermophysical properties of these coolants at the room temperature are listed in

Table 1.1.

10

Table 1.1 Thermophysical Properties of Liquid Coolants at Room Temperature [10]

In addition to the conventional coolants mentioned above, nanofluids and ionic liquid based

nanofluids are newly classified liquid coolants. The nanofluids are made of nanoparticles dispersed

in the conventional liquid coolants to improve the thermal properties for the uses with reduced

sized electronics. Copper (Cu) in water, alumina (Al2O3) in water and titanium oxide (TiO2) in

water are examples of the nanofluids with enhanced heat transfer capabilities up to 18 % in

comparison to the heat transfer capability of water [10].

1.3 Research Objective and the Thesis Organization

In this study, an improved bonding technology using silver nanopowder sintering is proposed to

enhance the heat management and the power efficiency within the IGBT power module for the

HEVs and EVs. The direct IGBT mounting using pressureless silver nanopowder sintering is

studied as a bonding technology to reduce distance of the thermal path on the aluminum liquid

cooling module. The scope of this study is focused on the development and investigation of the

silver sintering bonds on the directed bonded copper (DBC) substrate, while the research on and

development of the liquid cooling module is coordinated by Dana Ltd.

11

This thesis consists of 4 chapters. Chapter 1 introduces the overview of the research background

on the power semiconductor module and the heat management using cooling systems. Chapter 2

contains extensive backgrounds on physics behind the sintering, material selection and failure

mechanisms. In Chapter 3 the sample fabrication processes, experimental and analysis results and

discussions are discussed. Chapter 4 presents the conclusions and future work suggestions.

12

Chapter 2

Literature Review

In this chapter, the background of this study and literature reviews are present and discussed

covering two main subjects of the silver sintering process and the direct bonded copper (DBC).

2.1 Background

The bonding materials between the insulated gate bipolar transistor (IGBT) dies and the direct

bonded copper (DBC) substrate must have a melting temperature above the maximum operating

temperatures of the IGBT dies to prevent delamination, and a high electrical conductivity to allow

electrical conduction between the collector of the IGBT and the DBC. Silver nanopowder is an

excellent candidate meeting these requirements, with the melting temperature at 960 ˚C, electrical

conductivity of 6.3×107 S/m, thermal conductivity of 406 W/m·K for the bulk silver and

240W/m·K for the nanopowder silver at 20 ˚C, and an excellent adhesive strength [6], [14]–[19].

However, the processing temperature of the bonding materials must be lower than the thermal

budget of the IGBT. The thermal budget is the maximum temperature that the IGBT is functional

without a failure during the bonding process. To reduce the thermal stress applied to the IGBT die

during the bonding process, a lower processing temperature is desirable. Sintering is an effective

method in lowering the temperature of the bonding process.

13

2.2 Sintering Theory

The sintering of powder is not a newly developed technique, but has been used for over thousands

of years in ceramic tool manufacturing. However, advanced sintering techniques are employed in

various manufacturing fields including automobile engines, rocket nozzles, dental implants and

semiconductor packaging substrates, etc. To understand the physics behind the sintering,

considerable researches have been carried out and the controllable variables established [20], [21].

The sintering is classified into solid state sintering and liquid state sintering. Solid state sintering

involves solid state diffusion of atoms. Solid state sintering materials include polycrystalline

materials and amorphous materials based on their lattice structures. These two materials exhibit

different diffusion mechanisms. Liquid state sintering technique involves the uses of liquid phase

materials. Liquid phase materials at an optimized condition can provide rapid mass and heat

transport paths therefore promoting fast sintering [22], [23]. In this study, the application of solid

state sintering is reviewed and further examined.

2.2.1 Polymer Burnout

Before the sintering takes place, polymers used as binders and lubricants are removed from the

sintering powder mixture. The polymer burnout is triggered when the powder mixture experiences

a temperature raise and at a certain temperature the polymer molecules become unstable. The

molecular bonds of the polymers are disconnected and the molecules are decomposed into smaller

molecules such as carbon, oxygen, water and other by-products of the combustion [24].

14

2.2.2 Theoretical Analysis of Sintering

The sintering is carried out to fabricate a bulk solid from powders at a low processing temperature.

Because a complete phase change is not required during the sintering, the processing temperature

is significantly lower and the link between the powders are determined by the diffusion

mechanisms. The density of the sintered product increases with the process time. The relative

density during the sintering process is illustrated in Figure 2.1. Starting from the porous structure

at the beginning, the pore sizes decreases as the density of the structure increases.

Figure 2.1 Evolution of the relative density during sintering process [25].

The sintering is classified into three stages; : (a) Initial-point contact, (b) intermediate-neck growth,

and (c) final reduction of pore sizes [24]. Figure 2.2 illustrates the schematics of the sintering

process among four particles in the same plane.

Figure 2.2 Schematics of the sintering process, (a) initial, (b) intermediate, and (c) final [14].

15

At the beginning, a point contact between two powder particles is required to initialize the atomic

diffusion. An empty volume surrounded by the solid particles and filled with gas is created. After

the initial point contact, the atomic diffusion is driven by the force to reduce the excess energy

associated with surfaces. The sintering can initiate by two mechanisms, (1) the reduction of the

total surface area by increasing the powder size, and (2) the elimination of the solid/vapor interface

and generation of grain boundaries.

The vapor pressure, the vacancy concentration and the stress of the sintering microstructure are

the factors considered at the initial stage. The vapor pressure is lower at the neck region where the

contact of two particles is made, and is above the equilibrium at the bulk region of the powder.

𝑃 = 𝑃𝑒𝑥 + 𝛾 (1

𝑅1+

1

𝑅2) (2.1)

where 𝑃 is the vapor pressure, 𝑃𝑒𝑥 is the external pressure applied, 𝛾 is the surface energy and 𝑅1

and 𝑅2 are the principal radius of curvature at the contact point [26]. Likewise, the vacancy

concentration is also far from the equilibrium at the curved surface.

C = 𝐶0 [1 −𝛾Ω

𝑘𝐵𝑇(

1

𝑅1+

1

𝑅2)] (2.2)

where C is the vacancy concentration under a curved surface, 𝐶0 is the equilibrium vacancy

concentration, Ω is the atomic volume, 𝑘𝐵 is the Boltzmann’s constant and T is the absolute

temperature. 𝐶0 has the Arrhenius temperature dependency as presented in Equiation 2.3 below;

𝐶0 ∝ exp (−𝐸𝐴,𝑣

𝑅𝑇) (2.3)

where 𝐸𝐴,𝑣 is the activation energy of the vacancy formation which is proportional to the melting

temperature of the solid, R is the gas constant of 8.314 J/K·mol. The geometrical evolution and

16

surface vacancy concentration levels are shown in Figure 2.3 at the different stages [27]. As the

simulation result shown, at the initial stage of (b), the vacancy concentration is high near the neck

region driving the atomic diffusion to the neck growth. The vacancy concentration finds its

equilibrium throughout the sintering process and two particles merge into one large particle.

Figure 2.3 Geometrical evolution and instantaneous surface vacancy concentration level for the

two initially spherical crystalline particles of equal size sinter by combined surface, volume and

grain boundary diffusion (a) t=0, (b) t=0.1, (b) t=1, (d) t=5 (e) t=20 and (f) t=100 [27].

The atoms are moved along the powder particle surface to fill the valleys in between the curved

surface of the particles. Equation 2.4 describes the driving force to initiate the sintering;

𝜎 = 𝛾

1

𝑅1+

1

𝑅2

(2.4)

17

where 𝜎 is the stress associated with a curved surface. The radius of the curvature is measured by

creating a fitted circle at the point of interest of the curve. The radius from the center of the fitted

circle is equal to the radius of the curvature. The definition of the radius of the curvature and the

center of the curvature is schematically described in Figure 2.4. As the radius of the curvature

decreases, the size of powder also decreases, thus the powder possesses the higher stress at the

surface. Consequently, the smaller powder tends to merge with another powder with a larger radius

of the curvature which in turn has lower surface energy.

Curve

Circle of Curvature

Radius of Curvature

Centre of Curvature

C

P

R

Figure 2.4 Definitions of radius of curvature and the center of curvature [28].

Throughout the sintering process, the surface area of powders decreases, and the pore volume

decreases at the same time. A neck is formed in between two contacted particles. For the solid

state crystalline powders, the neck formation rate depends on the mechanisms of sintering . The

major mechanisms inducing mass transformations are the grain boundary diffusion, the surface

diffusion, the volume (lattice) diffusion, and the viscous flow. Table 2.1 below summarizes the

properties of the solid state sintering mechanisms. Within the table, under the “Source of matter”

column the sources where the atoms are coming from are listed, and under the “Sink of matter”

18

column destinations where the atoms are relocated and diffuse into are listed. The diffusion

sourced from the grain boundary and the dislocation contributes to the densification at the later

intermediate and final stages. Other mechanisms sourced from the surface contribute more to the

surface reduction and the neck formation at the initiation and intermediate stages.

Table 2.1 Sintering Mechanisms in Solids [29]–[31]

Type of solid Mechanism Source of matter Sink of matter Densification

Polycrystalline Surface diffusion Surface Neck No

Lattice diffusion Surface Neck No

Grain boundary Neck Yes

Dislocation Neck Yes

Vapour transport Surface Neck No

Grain boundary diffusion Grain boundary Neck Yes

Amorphous Viscous flow Unspecified Unspecified Yes

Figure 2.5 (a) illustrates the paths of the diffusion for the different mechanisms. The dihedral angle

illustrated in (b) is the angle between the two spherical particles at the grain boundary. As the

dihedral angle increases, more greater degree of shrinkage and the neck growth is found indicating

the progress and the completion of the sintering [24], [27], [32].

19

X/2R

Lattice diffusion

Vacancy diffusion

Vacancy surface diffusion Dihedral angle

Triple junction

Grain boundary

Particle surface

Crystalline particle

(a)

(b)

Figure 2.5 Schematic diagram of two particles sintering model (a) diffusion paths and (b) Dihedral

angles at the initial state (120°), equilibrium state (150°) and final state (180°) [27], [33].

The relationship between the neck growth and the diffusion mechanism is defined by Equation

2.5.

(𝑋

𝐷)𝑛 =

𝐵𝑡

𝐷𝑚 (2.5)

where X is the neck diameter, D is the particle diameter, t is the isothermal sintering time, and n

and m are constants [24], [30], [34]. B is the coefficient of the mechanism characteristic, including

the temperature term;

𝐵 = 𝐵0 exp (−𝐸𝐴

𝑅𝑇) (2.6)

𝐵0 is the coefficient of mechanism, and 𝐸𝐴 is the activation energy. The values for n, m and B are

related to the mass transport. The values of n and m are different based on the mechanisms of the

diffusion. The B value is also different, depending on the diffusion mechanisms [24].

20

During the intermediate stage, pore rounding, grain growth and densification occur. The surface

transport still plays a role in pore rounding and pore migration at this stage. However, the volume

and grain boundary diffusion contribute more for densification. At the intermediate and final stages,

the grain size increases as the pore size increases and the porosity decreases. The geometric model

proposed by Coble is commonly used [34]. The lattice and grain boundary diffusion equation is

derived. The fourteen-sided tetrakaidecahedron in Figure 2.6 is assumed to be the final geometry

of the grains with the complete densification with the cylindrical pores along the edges [35]. Once

the pores are completely closed, the final stage of the sintering occurs. The pores at the final stage

has spherical shapes.. If the gas is trapped in the pore, the pore elimination process would be

extremely slowly or even prevented. For full densification, different sintering process conditions

and techniques may be accounted such as vacuum sintering [36].

Figure 2.6 Equisided tetrakaidecahedron geometry [36].

21

2.3 Sintering Process Management

The literatures and backgrounds on the process parameters for sintering are reviewed in this

section. The key factors reviewed include temperature (𝑇), pressure (𝑃) and time (𝑡).

2.3.1 Sintering Temperature

The sintering temperature is highly related to the sintering powder size, and the processing

temperature can give an advantage in terms of a significant drop in the sintering temperature by

using a nano-sized powder. The sintering temperature (𝑇𝑠) is expressed as in Equation 2.7.

𝑇𝑠 = 𝛼𝑇𝑚,𝑝 (2.7)

where 𝑇𝑚,𝑝 is the melting temperature of the powder, α is the coefficient of the geometric and

environmental factors, and is usually in the range from 0.5 to 0.8 for large sized powders. When

the particle size falls to the nano-meter scale, 𝑇𝑚 becomes lower than the melting temperature of

the bulk material (𝑇𝑚,𝑏). The following Equation 2.8 defines the relationship;

𝑇𝑚,𝑝 = 𝑇𝑚,𝑏(1 −𝛿

𝐷) (2.8)

The term 𝛿 is a material dependent parameter and its value depends on the atomic volume and

bonding energy of the crystalline powder. The reported values of 𝛿 range from 1.8 to 2.65 nm.

The value for α can be reduced to a range between 0.1 and 0.3. Therefore, the nano-meter scale

powder sintering is achieved at temperatures ranging from 0.1 to 0.3 𝑇𝑚,𝑏. For the sintering using

silver with the melting temperature of 960 °C, the sintering temperature can be as low as 110 °C

[26].

22

2.3.2 Sintering Pressure

Either the absence or the presence of the applied pressure during the sintering process classifies

the sintering as the pressureless or the pressure-assisted sintering. As mentioned earlier in Equation

2.1, the external pressure (Pex) leads to the higher flux of the atomic diffusion on the surface during

the sintering. With the pressure applied to the sample, the powder is physically deformed and the

contact surface area increases. Thus, the grain boundary is readily available even without

overcoming the activation energy of the diffusion. The advantage of the pressure-assisted sintering

is that the coarse powder with relatively larger sizes can be sintered at lower temperatures. One

drawback is the higher chance of mechanical failure of the IGBT die. Because the pressure applied

to the die can be as high as 40 MPa for the silver nanopowder, the corresponding force of 4000 N

is applied to a typical silicon die with an area of 100 mm2, which must be handled with care to

prevent the breakage [25].

Pressureless sintering is another option to overcome the drawback from the pressure-assisted

sintering. The Herring law in Equation 2.9 shows the relationship between the sintering times

(𝑡1, 𝑡2) and the particle radiuses(𝑟1, 𝑟2) of two particles. The integer constant 𝑚 ranges from 2 to

4 [24], [37].

(𝑟1

𝑟2)

𝑚

=𝑡2

𝑡1 (2.9)

From this relationship, it is clearly shown that the smaller particle has faster sintering time which

allows a reduction in sintering temperature and pressure. Overall, the pressure applied and the

curvature induced from the size of the particle affect the total driving force of the sintering process.

The pressure applied and the curvature contribution to the driving force is plotted in Figure 2.7.

As presented in this plot, with the smaller grain sizes less than 20-30 nm, the curvature brings the

23

most contribution to the total driving force. The pressure applied maintains the same factor of

contribution within the same range of the grain size. However, for the larger sizes of grains the

pressure applied contributes more to the driving force of the sintering.

Figure 2.7 Driving forces for nanopowder consolidation as a function of grain size [38], [39].

2.4 Nanopowder Sintering Challenges

The term “nanopowder” often refers to the powders with diameters ranging from 1 to 100 nm.

When compared to the conventional materials, a nanopowder has significantly larger surface area

to volume ratio, also meaning that the surface energy is larger. Consequently, the sintering

temperature can be lowered as explained in Section 2.3.1. However, this physical characteristic of

the nanopowders may cause problems. In this section two challenges are addressed – one from the

pre-sintering state and another after the sintering.

24

The agglomeration and the aggregation of the nanopowder brings one of the challenges in the pre-

sintering state of nanopowder. The agglomeration and aggregation are the phenomenon where the

powders are gathered to form a colony of powders without external forces applied and behave as

one large polycrystalline powder. Due to the fine powder sizes and the large surface to volume

ratio, the agglomeration and the aggregation commonly occur in the nanopowder paste. The major

difference between the agglomeration and the aggregation is the bonding strength. The

agglomeration is the state where the powders are weakly bonded by the Van der Waals force or

the electrostatic force. The aggregation is formed by the strong bond such as the metallic bonding

or covalent bonding. Therefore, to disperse the aggregated powder, an application of an external

energy is required. Both the agglomeration and the aggregation result in the inhomogeneous

powder distribution. Even before the sintering, the green density is lowered due to the

agglomerated and aggregated powders. The effective radius is used to characterize the

agglomerated and aggregated powders. The agglomerated and aggregated powder colonies behave

as the large sized powder with the effective radius. When the effective radius exceeds the nano-

scale range, the advantages of the nanopowder is no longer applicable. Figure 2.8 illustrates

descriptions of (a) the agglomerated powders and (b) the aggregated powders with the effective

radius [40]–[42].

25

Figure 2.8 (a) Agglomeration and (b) Aggregation of the nanopowders [22], [42].

The non-densifying diffusion at the low temperature is another challenge that the nanopowder

sintering possesses. The volume (lattice) diffusion from the grain boundary or the dislocations in

the neck region can lead to the densification. Figure 2.9 illustrates the results from the densification

diffusion and the non-densification diffusion. The length between two particle centers (L) differs

with the densification diffusion, but with the non-densification diffusion the length is not affected.

Figure 2.9 Sintering results from densification and non-densification diffusion.

26

The non-densification diffusion and densification diffusion are controlled mainly by the

temperature as shown in Figure 2.10. Within the relatively lower temperature range, the surface

diffusion is dominant thus not triggering the densification. Hence, using the nanopowder with the

higher surface area to volume ratio and processing at a lower temperature allow the sintering

process susceptible to the problem of non-densification. To reduce the effect of the non-

densification diffusion, the heating can be accelerated to bypass the low temperature region so that

the surface diffusion is not sufficiently performed during the short time interval. Techniques such

as the microwave sintering, the plasma activated sintering, the laser sintering and the field activated

sintering technique (FAST) are used for the rapid heating [43]–[47].

Figure 2.10 Changes in densification rate in temperature.

2.5 Direct Bonded Copper (DBC)

The direct bonded copper (DBC) substrate has the sandwiched structure of copper layers at the

bottom and on top and a ceramic layer in between for the electrical insulation. Alumina (Al2O3)

and aluminum nitride (AlN) are the most commonly used materials as the insulating ceramic layer.

27

To withstand a higher operation temperature of electronics and the heat cycling in a harsh

environment such as a large temperature difference between the maximum and minimum operating

temperatures, the module needs a higher thermal dissipation rate. Thus, characteristics such as a

higher thermal conductivity, a low coefficient of thermal expansion (CTE) on each layer and small

differences between CTEs of layers are desired for the module’s performances with better

reliabilities. The DBC is an excellent candidate satisfying these characteristics. ..

As the name DBC refers to, the copper layers are directly bonded to the surface of the ceramic

layer. The ceramic layer is generated first and the copper layers are bonded onto the ceramic layer.

During the bonding process, transition layers are created between the copper and the ceramic to

bond the layers via oxide bridge, the eutectic liquid from the Cu-O system [9], [48]. Figure 2.11

shows the partial binary phase diagram of the Cu-O system. The eutectic temperature is 1065 °C,

and the eutectic liquid forms at the temperature slightly above the eutectic temperature but below

1083 °C which is the melting temperature of copper, to allow the ceramic to bond. This technique

is not only used with Al2O3/Cu but also with AlN/Cu and Si3N4/Cu laminates [49].

Figure 2.11 Schematics for the eutectic bonding copper process and the Cu-O phase diagram [49].

28

The thermal resistance at the interface is considered important since the reaction phase Cu2O or

CuAlO2 is formed as isolated particles. The thermal conductivity of Cu2O and CuAlO2 is 6 W/mK

and 11 W/mK respectively. Compared to the thermal conductivity of Cu and Al2O3, 400 W/mK

and 24 W/mK, the reaction phases behave more likely as resistors and may interfere the current

path along the copper layers. Two critical issues are imposed in the DBC using the eutectic bonding.

First, the large void is formed at the interface. The typical size of the void is larger than 100 µm,

and sometimes void sizes of larger than 500 µm are also found. An image of a void is presented in

Figure 2.12. The void formation is due to the gas generated at the temperature range between 1065

°C and 1075 °C. Upon cooling below 1065 °C, the formed gases are pushed to the ceramic interface

and they are trapped at the interface as the solidification process ends. The voids are possibly the

weak points of failure, so they also impact the long term reliability of the DBC. The differences

between the thermal expansion coefficients of the ceramic and the copper is closely looked into as

the presence of stresses threatens the long term reliability [48], [49].

Figure 2.12 Cu/Al2O3 void formed at the interface [49].

29

2.6 Chapter Summary

In this chapter, the theoretical background and literatures of the sintering mechanisms and process

variables are reviewed discussed extensively. Also, the DBC manufacturing techniques and

possible failure factors are described.

The sintering uses powders to generate a bulk solid without melting. The solid state diffusion is

the driving force of the sintering. The key advantage of the sintering process is the lower process

temperature than the melting temperature of the solid. The sintering temperature, pressure, time,

the heating rate and the powder diameter are the controllable factors affecting the sintering process.

The sintering is further classified into types depending on the phase of powder and the solvent; the

polycrystalline solid state sintering, the amorphous solid state sintering and the liquid state

sintering.

The sintering process has three stages, initial, intermediate and final. At the initial stage, the

powders form a contact and a void/pore volume is created as the surface diffusion is dominant. At

the intermediate stage, the pore rounding and the volume diffusion is presented as the neck grows

among particles, which results in the pore size reduction. The final stage is where the densification

and the shrinkage are observed. The pores are closed and isolated with the gas trapped in. A

vacuum condition is needed to remove the gases completely and to reduce the porosity further.

The diffusion mechanisms are discussed and classified into the surface diffusion, the volume

(lattice) diffusion, the grain boundary diffusion, the dislocation diffusion and the viscous diffusion.

The surface diffusion is the main driving force at the beginning of the sintering and the volume

diffusion, the grain boundary diffusion and the dislocation diffusion play more roles during the

30

later stages. The densification is only governed by the grain boundary diffusion and the dislocation

diffusion.

Challenges of the sintering are also still encountered especially for the nano-scale powders. The

agglomeration and aggregation of the powder compact before the sintering affect the performance

of the sintering and the final porosity. Because the surface energy of the nanopowder is relatively

large, powders tend to form a weak bonds (agglomeration) or strong bonds (aggregation) such as

metallic bonds. By generating colonies of powder, the effective radius of the colonies tends to

behave as those of the large sized particles, and the nanopowder properties are lost. Another

challenge is the non-densification of the powder. Especially at the lower sintering temperature, the

surface diffusion is dominant rather than the grain boundary diffusion or the dislocation diffusion,

which results in the densification. As it has no chance to densify, the pores remain within the final

solidified structure. To resolve this problem and promote the densification at lower temperatures,

new sintering techniques such as the plasma activated sintering, the laser sintering and other rapid

heating sintering techniques are adopted.

The DBC manufacturing techniques and possible failure factors are described. The eutectic

bonding is used to fabricate the DBC using direct bonding between copper layer and the ceramic

layer. Typical ceramics such as Al2O3 and AlN which forms the oxygen bridge to the copper easily

are used. The thermal conductivity is the key factor considered during the selection of the ceramic

material. A drawback of the DBC is the void formation at the interface of Cu and ceramic layers,

as these voids affect the long term reliability of the DBC.

31

Chapter 3

Experimental Results and Discussion

In this chapter, the experimental results are presented and discussed. The contents are organized

as follow: Section 3.1 outlines the experiment design, Section 3.2 provides the analysis on the

silver nanopowder paste, Section 3.3 discusses the optimization for the sintering process, Section

3.4 examines the reliability of the die attachment, and finally Section 3.5 summarizes the chapter.

3.1 Experiment Design

The purpose of this experiment is to develop a solid bond between the silicon IGBT die and the

DBC substrate via pressureless silver nanopowder sintering.

3.1.1 Structure Design

The construction of the IGBT power module consists of two main parts, the IGBT die mounted

DBC substrate and the liquid cooled aluminum heat sink. Alumina (Al2O3) or aluminum nitride

(AlN) ceramic interlayered DBC is chosen due to its electrical isolation property and high thermal

resistivity. The overall structure is described in Figure 3.1. Liquid cooled aluminum plate with the

customized structure is brazed under the interface layers. The interface layer is indicated as the

added layer in Figure 3.1, which is the buffer layer to promote the better bonding between

aluminum cooling plate to copper. This layer is added because of the two reasons; first, to avoid

the brazing, and second, to try other bonding techniques. The ceramic layer in the DBC is too

brittle to experience brazing under high pressure and high temperature to build a direct bonding to

32

the aluminum cooling plate, so brazing is not preferred. Also, other bonding techniques than

brazing and soldering are tests by adopting the additional layers. The developments of the interface

layers and the liquid cooled aluminum plates are the proprietary technique of Dana Ltd. Further

researches on the development of the cooling module is conducted under the scope of Dana Ltd.

as well. A patterned DBC is bonded on the interface layers. The IGBT die is mounted on the DBC

via silver nanopowder sintering without applied pressure. The bondwires are used to connect the

IGBT die to the copper pads on the custom DBC design for the power module. Figure 3.1 illustrates

the perspective view of the IGBT power module. The IGBT is mounted on the patterned DBC with

a Ni/Au electroplated top layer via pressureless sintered silver nanopowder. The DBC in Figure

3.2 shows the single die test pattern that is the same as the one shown in Figure 3.3, also giving

the dimensions of the patterned DBC. The dimension of IGBT die used in the experiments is 5mm

5mm. The silver nanopowder paste is applied with a 2mm oversize to ensure proper coverage of

the IGBT die. The emitter, gate and collector pads are separated by a minimum distance of 2mm.

The IGBT and the pad is connected by using gold bondwires. In this study, only the IGBT die

bonded on DBC using pressureless silver nanopowder sintering is analyzed.

33

Figure 3.1 Cross sectional view of the proposed IGBT attachment on a DBC substrate using silver

nanopowder sintered. The DBC is bonded to a liquid cooled aluminum heat sink.

Figure 3.2 Perspective view of the proposed IGBT attachment on a DBC substrate using silver

nanopowder sintered.

34

Figure 3.3 The DBC design and the actual patterned DBC for a single IGBT die.

Half-bridge and full-bridge IGBT modules are also designed as shown in Figure 3.4. These DBCs

are used for the electrical and thermal analysis of the IGBT module and for the assembled

prototype. Experiment results in this study are limited to the single die DBC substrate.

Figure 3.4 The DBC design for a Half-bridge IGBT (with free wheeling diodes) module assembly.

3.1.2 Sample Preparation

The starting materials from Rogers Co. are 5.5 inches × 7.5 inches blank DBC. They are cut and

patterned according to the design illustrated in Figures 3.3 and 3.4. A variety of thicknesses and

35

ceramic materials are considered in this study. Thermal conductivity is the one of important

properties to be considered in order to prevent the heat build-up and to facilitate efficient heat

transfer. The coefficient of linear thermal expansion (CTE) of the ceramic must be as close as

possible to the CTE of the copper to minimize the mechanical stress applied to the

copper/ceramic/copper sandwiched DBC structure. The CTEs of the pure wrought copper and the

cast copper are in the range from 16 to 18 ppm/K [50]. As shown in Figure 3.5 (a), if the difference

between the CTE of Cu and the CTE of the ceramic is within the acceptable range where the

ceramic A can hold the stress within its elastic deformation range, thus no failure would occur.

However, near the interface the ceramic experiences tensile stress where the CTE of the metal is

larger than the CTE of the ceramic. At the opposite side away from the interface, the ceramic

experiences compression stress. Ceramics have excellent resistances to compression stress but are

less robust to the tensile stress in general. During the thermal expansion, the sandwiched structure

experiences both tensile and compression stresses, but the elastic deformation limit of the ceramic

is easily exceeded at where the tensile stress is applied. It causes the plastic deformation and the

breakage of the ceramic. Thus, it is important that the ceramic undergoes less tensile stress during

the thermal expansion, and that the two layers of metals on the top and at the bottom of the ceramic

are bonded to the ceramic. By bonding the top and bottom metal layers, the ceramic no longer

endures tensile stress on one side, but the stresses applied from both interfaces annihilate the

extreme tensile stress within the ceramic layer. This allows the DBC structure with better long-

term reliability. Figure 3.5 (b) illustrates the failure of DBC due to the breakage of the ceramic

layer. Failure of sandwiched structure increases the thermal resistivity rapidly, and decreases the

ability to dissipate heat.

36

(a) CTE Cu ≥ CTE Ceramic A

(b) CTE Cu ≫ CTE Ceramic B

Figure 3.5 CTE differences in two layers produces interlayer stresses, leading to potential failure.

Sample data including composition, thermal conductivity, CTE and thicknesses for the selected

DBC substrates are listed in Table 3.1.

Table 3.1 DBC Sample Data at 20 °C [51]

Ceramic

Substrate

Composition Thermal

conductivity

CTE Ceramic

thickness

Cu

thickness

Al2O3 Alumina 24 W/mK 6.8 ppm/K 0.32 mm 0.3 mm

0.64 mm 0.3 mm

HPS Alumina + 9%

ZrO2 doped

26 W/mK 7.1 ppm/K 0.32 mm 0.3 mm

AlN Aluminum Nitride 170 W/mK 4.7 ppm/K 0.64 mm 0.3 mm

The DBC substrates have two layers of copper; one at the top and the other at the bottom. The top

copper layer is customized in design and patterned, while the bottom copper layer is not processed

further during the preparation.

37

3.1.2.1 Thermal Budget

The silver nanopowder paste is capable of maximum sintering temperature up to 260 ˚C. The

recommended typical sintering temperatures are 250 ˚C on a silver coated DBC surface and 200

˚C on a gold coated DBC surface. Therefore, considering the maximum operation temperature of

175 ˚C for silicon (Si) based IGBTs, the gold coated DBC is selected in order to maintain low

thermal stress applied to the IGBT throughout the sintering process [6], [8], [52].

3.1.2.2 Nickel/Gold Plating

Once patterning is completed on the top copper layer, Ni and Au layers are then plated. Au is

chosen due to its good wettability, high electrical and thermal conductivity, and immunity to

oxidation. However, since Cu and Au interdiffuse easily, Ni is embedded to form a barrier layer

in between [53]. It is common to have electroless nickel and immersion gold (ENIG) for the

commercial printed circuit board (PCB) and DBC substrates. However due to the availability of

the laboratory equipment, electrolyzed Ni and Au coating is performed.

Watts Ni electroplating is applied and followed by the Au electroplating [54]. Ni and Au plating

set up is shown in Figure 3.6.

38

Figure 3.6 Ni and Au electroplating wet bench set up.

In preparation for the electroplating, the sample is cleaned to eliminate sources of contaminations,

including organic contaminants and native oxides. SP Cleaner manufactured by Caswell Canada

is used as a degreaser removing organic contaminants. 60 g/L of the SP Cleaner powder is

dissolved into 1 L of deionized water (DIW). The degreaser solution is then heated up to the

temperature ranging from 70 °C to 90 °C. The sample is fully dipped in the cleaner solution for a

time period of between 10 and 15 minutes depending on its condition. Once the organic

contaminants are all removed, the degreaser solution is rinsed off using DIW. In order to remove

native oxides, 10 vol% hydrochloric acid (HCl) solution at room temperature is used to etch the

copper surface for 2 minutes. DIW is used to rinse off the remaining HCl solution before the

sample surface is blow-dried using pressurized air. Upon completion of the cleaning processes, the

sample is ready for the electroplating.

The electroplating circuit is set up as illustrated in Figure 3.7 (a). The DBC is connected at the

cathode, and the Watts Ni in the titanium bath is connected at the anode. Watts Ni solution contains

39

300 g/L nickel sulfate, 45 g/L nickel chloride, 45 g/L boric acid, 0.5 g/L sodium saccharin, 0.2 g/L

sodium dodecyl sulfate in 1 L of deionized water (DIW) [54]. The temperature and pH of the bath

solution is maintained at 55 °C and 4.0 respectively. The pH is controlled by adjusting the amount

of 50 vol% sulfuric acid solution added to the bath prior to the Ni plating.

(a) Electroplating circuit (b) Anode

Figure 3.7(a) Electroplating circuit for Ni [55] and (b) Anode.

Pulse Current (PC) is supplied to the Ni bath to yield finer grain depositions. High PC density

results in a higher nucleation rate on the substrate with low porosity [54], [56]. In order to achieve

the minimum 3.5 µm of Ni thickness on the surface of the top copper layer, a peak current of 0.15

A is applied for 4 minutes, whereas the peak current is the amplitude of the current pulse. The

plating speed is maintained at around 1 µm/min·cm2 on both sides of the sample. Due to the nature

of PC, the current is supplied periodically. The time when the current flows is called ON time and

when the current does not flow is called OFF time. The average current is the level of energy

equivalent to the direct current (DC) level, and it is determined by the duty cycle and peak current

as in Equation 3.1. The duty cycle of the power supply is calculated as the ratio of the time during

40

which the power supply is ON against the total time as in Equation 3.2 [57]. The peak current of

0.75 A is applied at the duty cycle of 20 % with 20 ms of ON time and 100 ms of OFF time. The

maximum voltage is set to 40 V.

Average current = peak current × duty cycle (3.1)

Duty Cyle (%) = 𝑂𝑁 𝑡𝑖𝑚𝑒

𝑂𝑁 𝑡𝑖𝑚𝑒+𝑂𝐹𝐹 𝑡𝑖𝑚𝑒× 100 (3.2)

Upon completion of the Ni electroplating, the remaining bath solution is rinsed off using DIW.

The Au electroplating is prepared using a stainless-steel tip as an anode connected to the DC power

supply. The stainless-steel tip is wrapped-around by a cotton ball, and then soaked in the Au

solution in order to ensure continuous contact between the tip and the solution. The Ni plated DBC

is connected to the DC power supply as a cathode. 3 V DC is supplied.

The three images in Figure 3.8 show the surface finishes after each step of electroplating. Figure

3.8 (a) is the bare DBC after cleaning and native oxide etching. Figure 3.8 (b) shows the surface

of the DBC after the Ni electroplating. Shown in Figure 3.8 (c) is the Au plated DBC with the

sintered silver bonding layer and IGBT mounted on. Also, the Au wire is connected from the gate

and emitter terminals of IGBT to the gate and emitter pad, respectively.

(a) Bare DBC (Cu) (b) Ni (c) Au

Figure 3.8 DBC substrate samples in conditions of (a) bare (Cu on top), (b) Ni on top, and (c) Au

on top with the IGBT die mounted on pressureless sintered silver nanopowder with the Au wire

connections.

Ag

Au

IGBT Au wire

41

3.1.2.3 Silver Bonding Layer Process

The IGBT modules is bonded on the surface of the DBC via the silver bonding layer process using

silver nanopowder paste in order to allow electrical conductivity. The overall silver bonding layer

process flow is illustrated in Figure 3.9. The IGBT die is placed and mounted on the surface of the

DBC where the silver nanopowder paste is evenly printed. The sample with the IGBT die mounted

is then placed in the oven for drying and sintering.

Figure 3.9 Process flow for the IGBT die attachment process using silver nanopowder sintering.

3.1.2.4 Silver Nanopowder Pasting

The silver nanopowder paste, Loctite Ablestik SSP 2020-EN from Henkel is spread evenly on the

DBC substrate using the stencil printing technique as illustrated in Figure 3.10. First, the laser cut

stainless steel stencil of the thickness of 80 µm is placed on the surface of the Ni and Au

electroplated DBC, the silver nanopowder paste is applied on the stencil, and a stainless-steel

42

squeegee is used to press the silver paste for stencil printing as shown in Figure 3.10 (a) and (b).

Once the silver nanopowder paste is evenly spread, the stainless-steel stencil is removed, and the

IGBT die is mounted on the silver paste layer as illustrated in Figure 3.10 (c) and (d). Upon placing

the IGBT die on the sample surface, no pressure is applied, while the die and the silver nanopowder

paste layer are handled with care.

(a) Step 1: Stainless steel stencil placement on DBC (b) Step 2: Screen printing

(c) Step 3: paste spreading and removal of stencil (d) Step 4: die placement

Figure 3.10 Illustration of the silver nanopowder paste stencil printing technique.

3.1.2.4.1 Silver Nanopowder Drying and Sintering

After the IGBT die is placed and mounted on the sample surface, the DBC sample is placed in the

oven for drying and sintering processes. The sample is dried in the oven at 120 ˚C for 10 minutes

to ensure the complete removal of polymers such as solvent and additives.

43

After the drying process, the oven temperature is increased up to 200 ˚C for the sintering process.

The temperature is maintained at 200 ˚C for 60 minutes. Then, the sample is allowed to cool down

to the room temperature inside the oven. The temperature profile is depicted in Figure 3.11.

Ramping rates of 5 °C/min and 1 °C/min are used for heating up and cooling down, respectively.

The forced convection constant temperature drying oven DKN402 from Yamato is used. The

temperature distribution accuracy is ±2.5°C at 210°C with the oven in fully closed condition [58].

With no inert gas used, drying and sintering processes are executed with no pressure or vacuum

applied.

Figure 3.11 Temperature profile for drying and sintering processes.

3.2 Silver Nanopowder Paste Analysis

To analyze thermal responses during sintering processes, the silver nanopowder paste is examined

before and after sintering process, using analysis techniques of Differential Scanning Calorimeter

44

(DSC) / Thermogravimetry Analysis (TGA) / Simultaneous Thermal Analysis (STA) and

Scanning Electron Microscopy(SEM).

3.2.1 DSC/TGA/STA

The silver nanopowder paste is analyzed using Differential Scanning Calorimeter (DSC) /

Thermogravimetry Analysis (TGA) / Simultaneous Thermal Analysis (STA) to examine its

characteristic changes along the temperature profile during the sintering process. The analysis is

performed in Walter Curlook Materials Characterization & Processing Laboratory at the

University of Toronto.

The DSC measures the difference in heat required to raise the certain amount of the temperature

of the sample. By capturing the heat generation in cases of exothermic behavior or consumption

in cases of endothermic behaviour, the phase change of the sample is detected. An aluminum

reference sample is prepared in a N2 environment. The result data is extracted from the comparison

between the data measured using the silver nanopowder paste sample and the reference data.

The TGA records the mass changes of the sample along the temperature profile. The ramp rate

applied during the analysis is 5 ˚C/min from 30 ˚C to 300 ˚C. The silver nanopowder paste sample

of total mass of 55.56 mg is used for the analysis.

The results from the DSC/TGA/STA analysis are plotted in Figure 3.12 and Figure 3.13.

In Figure 3.12, it is observed that the endothermic reaction starts at 114 ˚C, indicating that the

liquids additives and solvent evaporate into gaseous forms. At 133 ˚C, the silver nanopowder paste

sample reaches its maximum in mass change, and the phase transformation is completed. These

results show that the evaporation of the polymer in liquids phases, such as additives and solvent is

45

accomplished at the temperature above 133 ˚C. The results also indicate that the additives and

solvent compose 10.9 % of the total sample mass. The plot shows an exothermic reaction starting

at 200 ˚C, therefore indicating that the silver nanopowder begins to diffuse and sinter at this

temperature. The exothermic reaction reaches its peak at 260 ˚C. The exothermic peak indicates

the sample experiencing the phase change from liquid to gas.

Figure 3.12 DSC/TGA/STA results for the silver nanopowder sintering experiment.

For further analysis of the sintering at 200˚C, an isothermal DSC/TGA/STA is performed for 60

minutes as shown in Figure 3.13. During the isothermal analysis, another exothermic reaction is

identified approximately 90 minutes from the beginning of the analysis and approximately 40

minutes after the beginning of the isothermal sintering process. A decrease in the sample mass by

0.2% is also detected.

46

Figure 3.13 Isothermal DSC/TGA/STA results for silver nanopowder sintering at 200 ˚C.

The overall DSC/TGA/STA results are summarized in Table 3.2. The onset temperature is

determined at the point where the mass change starts. While TGA measures the mass change of

the sample, Derivative Thermogravimetry (DTG) is computed as the percentage mass change. The

peak temperature is identified at the point where DTG is at minimum. The TGA result indicates

that the silver nanopowder paste sample is composed of 89 % silver nanopowder and 11 %

additives and solvent. Also based on the DSC and STA results the drying temperature above the

onset temperature of 114.4 ˚C and the sintering temperature above 200 ˚C are determined as

optimum temperatures for the processes. In addition, the results indicate that the sintering time

longer than 40 minutes is necessary for achieving stable sintered bonds among silver nanopowders.

47

Table 3.2 Summary of the DSC/TGA/STA Results

Onset T (˚C) Peak T (˚C) Δm (mg) Δm (%)

1. Solvent Evaporation 114.4 132.6 6.06 10.91

2. Heat up - 266.9 0.26 0.47

3. Isothermal - - 0.11 0.20

Total mass loss (300 ˚C) 6.43 11.58

Total mass loss (200 ˚C, 60 min) 6.17 11.11

3.2.2 SEM

Scanning Electron Microscope (SEM) images are analyzed to verify the results of the silver

nanopowder paste sintering. The instruments used are FEI Quanta FEG250 Environment SEM

(ESEM) and Hitachi S-5200 high resolution SEM (HRSEM) at the Centre for Nanostructure

Imaging in the Department of Chemistry, University of Toronto [59].

A cross sectional sample is prepared in order to analyze the bonding of layers with the sintered

silver nanopowder. Figure 3.14 is captured under the FEI Quanta FEG250 ESEM and shows the

cross section of the sample layers across the DBC, sintered silver nanopowder and IGBT. The

bonding of the sintered silver nanopowder layer with the Ni and Au coated Cu layer of the DBC

at the bottom, and with the IGBT at the top are visually inspected, and the cross sectional image

shows that the sintered silver nanopowder layer is firmly bonded to both the IGBT layer at the top

and the Ni and Au plated Cu layer of the DBC at the bottom.

48

Figure 3.14 Cross section image of the silver nanopowder sample sintered at 200 ˚C for 60

minutes, captured using Quanta FEG250 ESEM (magnification 500 X).

The chemical composition of the intermetallic layers between the sintered silver and the IGBT

bottom layer is examined using an Energy Dispersive Spectroscopy (EDS) analysis. The bottom

layers of IGBT die are identified as Ti and Ni. Figure 3. (a) is the image of the intermetallic layers

between the sintered silver and the IGBT bottom. Figure 3. (b) illustrates the four spots where the

EDS analysis is examined; EDS spot 1 at the deeper area of the sintered silver, EDS spot 2 near

the surface of the sintered silver layer, EDS spot 3 at the first intermetallic layer of the bottom of

the IGBT, and EDS spot 4 at the second intermetallic layer of the bottom of IGBT.

(a) (b)

Figure 3.15 (a) Cross-sectional view of sintered silver (magnification 10k X), (b) IGBT die and

sintered silver intermetallic layers EDS analysis on selected areas.

49

The weight percentage and atomic percentage of each EDS spot are analyzed and summarized in

Table 3.3 and Table 3.4 respectively. Carbon (C), oxygen (O) and aluminum (Al) are detected as

contaminants. Since the sample is processed and analyzed in the atmospheric surrounding, the C

and O contaminations are unavoidable. The Al contamination is introduced during the sample

preparation. During the sample preparation several grinding and polishing steps are required. Since

the alumina is used as an abrasive during the polishing step, excessive alumina abrasive remaining

on the sample surface is detected as a contaminant during the EDS analysis even after several

cleanings.

Table 3.3 Weight Percentage at EDS Spots within the Intermetallic Layers

EDS Spot 1 EDS Spot 2 EDS Spot 3 EDS Spot 4

C 6.02 6.28 8.04 7.58

O 2.71 9.22 5.54 8.15

Ni - - 71.33 7.53

Al 0.50 1.75 1.92 3.32

Si 0.26 2.39 2.39 18.19

Ag 90.52 80.35 10.78 1.14

Ti 0.50 - - 54.09

Ag Ag Ni+Ag Ti+Si

Table 3.4 Atomic Percentage at EDS Spots within the Intermetallic Layers

EDS Spot 1 EDS Spot 2 EDS Spot 3 EDS Spot 4

C 32.59 26.21 26.93 19.84

O 11.02 28.91 13.91 16.03

Ni - - 48.85 4.03

Al 1.20 3.25 2.86 3.87

Si 0.60 4.27 3.42 20.37

Ag 54.60 37.35 4.02 0.33

Ti - - - 35.51

Ag Ag Ni Ti+Si

50

The weight percentage and atomic percentage analyses indicates the EDS spots 1 and 2 within the

sintered silver layer, the EDS spot 3 consisting of Ni with some Ag diffused in, and EDS spot 4

within the Ti layer with some Si and Ni presence as well. The Si is detected at the spot 4 since the

Si layer is on top of the Ti layer. Consequently, the result verifies the formation of an intermetallic

layer by Ni and Ag therefore constructing a bond between the sintered silver layer and the IGBT.

The porosity of the sintered silver layer is examined on the image obtained from the cross-sectional

view of the sample. Figure 3. (a) is the SEM image for the porosity analysis, and the result of the

analysis as well. The porosity is measured by contrast comparison between the pores and the grains

on the image. The areas of dark regions and bright regions are considered as pores and sintered

silver grains respectively. Using the ImageJ software, the number of pixels within each dark and

bright regions is counted, and the result is plotted as shown in Figure 3. (b) [60]. The final porosity

of 29.6% is computed for the sample from the pressureless silver nanopowder sintering process.

Figure 3.16 (a) SEM cross-sectional image (magnification 5k X) for the porosity analysis and (b)

the porosity calculation result.

51

The SEM images of top views of the sample are examined under the Hitachi S-5200 HRSEM. The

shapes and sizes of the nanopowders are identified.

In order to determine the physical characteristics of the nanopowder during the heating-up process

before the sintering, samples are prepared by stopping the heating-up process at temperatures 120

°C, 130 °C, 150 °C and 170 °C and maintaining the temperature for 1 minute before proceeding to

the cooling-down step immediately. Figure 3.17 shows the points 1 to 5 indicating the different

temperatures at which the samples are taken during the heating-up process and after 1-minute

interval proceeded to the cooling-down. The temperature ranges starting from 120 °C is chosen as

the onset temperature of 114.4 °C is the minimum temperature at which the solvent and other

polymer mixtures are evaporated. Presence of remaining polymers from one sample analysis may

cause the cross contamination to another sample analysis, leading to the inaccurate analysis results

and an equipment failure due to debris causing a vacuum pump malfunctioning. Furthermore, since

the Hitachi S-5200 is a high-resolution SEM, the chamber is in high vacuum and thus no polymer

can be inserted into the chamber.

25

120

200

0

50

100

150

200

250

0 50 100 150 200 250

Tem

pera

ture

(ºC

)

Time (min)

120°C, 1min

RT

Te

mpe

ratu

re (

°C)

Time (min)

130°C, 1min

150°C, 1min

170°C, 1min

190°C, 1min

Figure 3.17 Sample preparations for the top view SEM image analysis along the temperature

profile.

52

(a) Process stopped at 120°C for 1 min

(b) Process stopped at 130°C for 1 min

(c) Process stopped at 150 °C for 1 min (d) Process stopped at 170 °C for 1 min

Figure 3.18 Top view SEM images of the silver nanopowder paste samples with heating-up

process stopped at different temperatures before sintering. All images are in

magnification of 8k X.

Top view SEM images (a)-(d) in Figure 3. are obtained at different temperatures. These SEM

images show that two different average powder sizes are present mixed in the paste. The samples

cooled down from 120 °C for 1 minute and from 130 °C for 1 minute are considered to be of the

same physical conditions of the original as-received silver nanopowder paste prior to the heating-

up, except that the solvent polymers are evaporated after heating-up. Thus, the analysis result

defines that the as-received silver nanopowder paste consists of powders of two different average

53

sizes, micro-meter scale powders and nano-meter scale powders. The mixture of micro-meter scale

and nano-meter scale powder has advantages of forming less agglomerated powders colony. As

the sample is not pressured during the sintering process, it is important to keep the nano-scale

powders uniformly mixed into the paste without making agglomerates or aggregates, since no

external energy is applied to break the bonds. However, agglomerated or aggregated powders are

found in all samples prepared below the sintering temperature which are indicated with the red

circles with in the images in Figure 3.. The physical characteristics of the silver nanopowder pastes

during the sintering is also analyzed using top view SEM images. The same method used above

for the sample analyses during the heating-up process is used. The sintering of the silver

nanopowder paste samples at 200 °C is stopped after different time intervals and started cooling-

down.

In Figure 3. the top view SEM images (a)-(d) are captured under the Hitachi S-5200 HRSEM. No

visible difference in porosities is observed among the samples of different sintering times. Also,

the size distribution of the micro-meter sized and nano-meter sized powders is maintained even

with the sintering process completed at 200 °C for 60 minutes.

54

(a) Process stopped at 200 °C for 1 min (b) Process stopped at 200 °C for 20 min

(c) Process stopped at 200 °C for 40 min (d) Process stopped at 200 °C for 60 min

Figure 3.19 Top view SEM images of the silver nanopowder paste samples sintered at 200 ˚C (a)

for 1 min, (b) for 20 min, (c) for 40 min and (d) for 60 min (magnification 10k X).

The overall average size of the silver nanopowder paste is measured at 3.2 µm, with a maximum

powder size at 22 µm and a minimum powder size at 58 nm. The grain size of the micro-meter

sized powder is measured using the images in Figure 3.. The average grain size is measured to be

142 nm (31 counts) with the minimum grain size of 78 nm and the maximum grain size of 602 nm.

The standard deviation is computed as 125.2 nm.

55

(a)

(b)

Figure 3.20 Images of grain sizes observed on the micro-sized powder after sintering at 200 °C

for 60 minutes, (a) magnification 80k X and (b) magnification 200k X.

In Figure 3. below, comparison is made between the top view SEM image (a) of the as-received

silver nanopowder paste and the image (b) of the sintered silver nanopowder paste under the same

magnification. Also, the images (c) of the nanopowder sintered for 10 minutes and (d) of the

nanopowder sintered for 60 minutes are compared to each other. From these comparisons, as a

result, no visible changes in porosity is observed during the sintering process, while the micro-

sized powders and nano-sized powders are observed maintaining their average sizes. The image in

Figure 3. (c), the circled powders show that the neck growth between the two spherical-shaped

powders occurs. In (d), the isolated pores are circled. Also, the non-densification is defined on the

sintered sample after completion of the process at 200 °C for 60 minutes. It is resulted due to the

low sintering temperature and the slow temperature raise rate. Thus, more surface diffusion is

promoted during sintering and not high enough temperature is applied to the powder sample to

lead the grain boundary and dislocation diffusion for the densification and reduction in porosity.

To achieve the more densification of the sample, faster temperature raising rate can be used by

trying other sintering techniques such as plasma sintering, laser sintering or field activated

sintering technique as mentioned in the previous chapter.

56

(a) Process stopped at 120 °C for 1 min,

(b) Process stopped at 200 °C for 60 min,

(c) Process stopped at 200°C for 10 min

(d) Process stopped at 200°C for 60 min

Figure 3.21 Top view SEM images of (a) the dried as-received silver nanopowder paste after the

heating stopped and maintained at 120 °C for 1 minute (magnification 5k X), (b) the

silver nanopowder paste sintered at 200°C for 60 minutes (magnification 5k X), (c)

silver nanopowder paste sintered at 200°C for 10 minutes (magnification 20k X), and

(d) silver nanopowder paste sintered at 200°C for 60 minutes (magnification 20k X).

3.3 Chapter Summary

In chapter 3, experimental results and following discussions are presented.

A DBC substrate is prepared to fabricate the sample. Before mounting the IGBT and examine the

silver paste sintering, Ni and Au electroplated layers is coated on bare Cu layer of DBC. The Au

57

is chosen to reduce the sintering temperature of silver nanopowder paste. Also, Au has good

electrical conductivity and immunity to the oxidation. Ni is used as the buffer layer between Cu

and Au to block the interdiffusions of Cu and Au atoms. When Cu diffuses into the Au layer and

is exposed to air, it oxides easily. Thus, the advantage of the Au layer being passive to the oxidation

is diminished. The sintering process is classified into four different steps: silver nanopowder paste

printing, IGBT die pick and mounting, silver nanopowder drying, and sintering. Stencil printing is

used to evenly spread the paste on DBC substrate. The drying and the sintering temperature and

time is optimized based on the results of DSC/TGA/STA, which is the analysis techniques to find

the phase changes and the mass changes of the sample along the temperature profile. The drying

of the sample is performed at the temperature of 120 °C for 10 minutes since the polymer solvent

evaporation occurs in the temperature range from 114 °C to 130 °C. From the result of isothermal

DSC/TGA/STA, sintering temperature of 200°C and sintering time of 60 minutes are defined as

optimized sintering condition. The sintering results are confirmed and inspected by capturing

images from the SEM analysis. SEM image analysis of cross sectioned sample is performed to

ensure the bonding via sintered silver nanopowder layers in between DBC and IGBT. EDS

analysis confirms the chemical compositions of the sintered silver nanopowder layer. The samples

to inspect the top surfaces is prepared for HRSEM analysis. The powder sizes are measured based

on the SEM images. Sizes of the powders are defined; the average size of 3.2 µm, the maximum

size of 22µm and the minimum size of 58nm. Also, the mixture of micro-scale and nano-scale

powder is defined. 29% porosity of the sintered silver nanopowder is counted using software. In

the as-received sample and pre-sintering samples, agglomerated or aggregated nano-meter scale

powders are found. After the sintering process, the sample has the non-densification problem.

58

Chapter 4

Conclusions and Future Work Plan

In this chapter, the conclusion from the literature review and the experimental result data are

discussed, and a future work plan is suggested for further improvements to the research and

development of the pressureless silver nanopowder sintering for the liquid cooled insulated gate

bipolar transistor (IGBT) power module.

4.1 Conclusions

Pressureless silver nanopowder sintering is studied and proposed for the liquid cooled IGBT power

module development for electric vehicles (EVs) and hybrid electric vehicles (HEVs). The

literatures on the sintering mechanisms, sintering variables, challenges present within the

nanopowder sintering, and the properties of the direct bonded copper (DBC) are reviewed. From

the literature reviews, the viability of the application of the pressureless nanopowder sintering at

the low temperature is reviewed. The processing parameters and the sample structure design for

the sintering analysis are determined accordingly. To determine the optimum sintering conditions,

the silver nanopowder paste is analyzed using the Differential Scanning Calorimeter (DSC) /

Thermogravimetry Analysis (TGA) / Simultaneous Thermal Analysis (STA), and the

characteristics of the sintered silver layer is examined under the Scanning Electron Microscope

(SEM). The result data obtained from the DSC / TGS / STA verify the thermal behaviours of the

silver nanopowder paste, and are used to optimize the sintering temperature profile. The SEM

image analysis verifies the formation of the bonding between the IGBT and the DBC resulting

59

from the pressureless silver nanopowder sintering at the low temperature. The SEM image analysis

results are also examined to identify possible improvements of the pressureless silver nanopowder

sintering due to the observed porosity and the nanopowder agglomeration within the sintered silver

layer. Based on the literature reviews and the experimental result data, the feasibility of the

pressureless silver nanopowder sintering application for the bonding between the IGBT and the

DBC within the liquid cooled power module, and possible improvements on the pressureless silver

nanopowder sintering are suggested.

The literatures on the sintering and the DBC are reviewed and used to determine the process

parameters and the feasibility of the pressureless silver nanopowder sintering. Since the minimum

particle size of the silver nanopowder paste and falls within the range where the effect of the

pressure on the sintering driving force is not significant, the feasibility of the pressureless sintering

is verified. The required sintering temperature is determined above the minimum silver sintering

temperature of 110 °C based on the particle size. The literatures on the sintering mechanisms are

reviewed. With the nano-meter scaled particles with the high surface area to volume ratio the

surface diffusion is determined as the dominant mechanism of the sintering at the low temperature,

therefore the high porosity remains throughout the sintering as no densification occurs. Due to the

high fabrication temperature above 1065 °C, the DBC is verified as a reliable substrate for the

power module. The alumina (Al2O3) and the aluminum nitride (AlN) are chosen as the ceramic

materials within the DBC based on the coefficient of thermal expansion (CTE) values from the

literature review.

The silver nanopowder paste is analyzed using the DSC/TGA/STA to examine the thermal

response and mass changes of the paste along the temperature profile. The evaporation of the

polymer solvent was identified at the temperature range from 114 °C to 130 °C. Therefore, the

60

drying temperature of 120 °C is selected. During the continuous temperature increase, a phase

change and a change in mass are detected at 260 °C. For the isothermal analysis, a phase change

and a change in mass are detected at 200 °C, 40 minutes after the isothermal process begins. Total

mass change of 11% is defined throughout the analysis, so 89% of the silver powder mass % is

verified. Consequently, the sintering temperature of 200 °C for the 60 minutes is selected and

applied to the pressureless silver nanopowder sintering.

From the SEM image analysis, the bonding between the IGBT and the DBC by the sintered silver

layer is inspected. The chemical composition of the layers between the sintered silver and the

bottom of the IGBT is analysed using the Energy Dispersive Spectroscopy (EDS). The

interdiffusions of the intermetallic layers indicates that the bonding is established between the

sintered silver layer and the IGBT. The porosity of 30 % within the final product is determined on

the cross-sectional image. On the top view images, the sizes of the nanopowder is measured with

the average size of 3.2 µm, the maximum size of 22 µm, and the minimum size of 58 nm. Two

different ranges of powder sizes are defined for the micro-meter sized and the nano-meter sized

powders. Also, the agglomerated or aggregated nanopowder colonies are detected within the

temperature range from 120 °C to 190 °C, and remain gathered as colonies throughout the sintering

process. These agglomerated or aggregated nanopowder colonies cause the non-densification of

the sintered silver nanopowder. Overall, the SEM image analysis validates the bonding capability

of the sintered silver nanopowder while also identifying aspects for possible improvements of the

pressureless silver nanopowder sintering.

In conclusion, the pressureless silver nanopowder sintering is capable to create the bonding layer

between the IGBT and DBC at the low sintering temperature. The literatures on the sintering and

the DBC are reviewed to verify the sintering parameters and controllable factors, and to identify

61

and avoid possible failures during the process. The DSC/TGA/STA identifies thermal behaviour

of the silver nanopowder paste and mass changes along the temperature profile. It allows the

optimization of the sintering temperature and time. The SEM analysis allows the visual inspection

of the nanopowder. The porosity of the sample and the sizes of the nanopowders are also

determined. The chemical composition analysis using the EDS confirms that the bonding is built

with the sintered silver nanopowder. Therefore, the experimental result data of the study verifies

that the pressureless silver nanopowder sintering at 200°C is a feasible technique for the bonding

within the liquid cooled IGBT power module. However, improvements can be made to minimize

possible impacts from the non-densification and the agglomerated / aggregated powder colonies.

The mechanical and electrical properties of the pressureless silver nanopowder sintering can also

be enhanced by rapid temperature raise sintering techniques.

4.2 Future Work Plan

Further study is planned for the development of the IGBT power module assembly with the

aluminum liquid cooling plate. The research on the application of the DBC bonding to the

aluminum liquid cooled plate will be executed under the coordination of Dana Ltd.

The thermal dissipation and power efficiency of the IGBT power module will be further analysed

to verify the performance of the module assembled with the aluminum liquid cooling plate. In

parallel, the heat dissipation from the IGBT to the liquid cooling plate through the sintered silver

layer and the DBC will be modeled using a software simulator in order to examine the thermal

paths within the power module and the liquid cooling plate.

The pressureless silver nanopowder sintering can be improved in aspects of the bonding quality

62

by minimizing the impact of the non-densification of the sintered product and the aggregation or

agglomeration of the nanopowder prior to the sintering.

63

References

[1] International Energy Agency, “Technology Roadmap: Electric and Plug-in Hybrid

Electric Vehicles (EV/PHEV),” 2011. [Online]. Available:

http://www.iea.org/publications/freepublications/publication/technology-roadmap-

electric-and-plug-in-hybrid-electric-vehicles-evphev.html. [Accessed: 30-Dec-2017].

[2] International Energy Agency, “Global EV Outlook 2017: Two million and counting,” IEA

Publ., pp. 1–71, 2017.

[3] Z. J. Shen and I. Omura, “Power semiconductor devices for hybrid, electric, and fuel cell

vehicles,” Proc. IEEE, vol. 95, no. 4, pp. 778–789, 2007.

[4] B. J. Baliga, “IGBT Applications:Transportation,” in The IGBT Devices, 2015, pp. 223–

275.

[5] “Power Semiconductor Devices: IGBT - An Introduction,” Power Electronics A to Z,

2017. [Online]. Available: http://www.completepowerelectronics.com/igbt-basics/.

[Accessed: 21-Dec-2017].

[6] Y. Liu, “Trends of power semiconductor wafer level packaging,” Microelectron. Reliab.,

vol. 50, no. 4, pp. 514–521, 2010.

[7] “PrimePACKTM IGBT Modules - Infineon Technologies.” [Online]. Available:

https://www.infineon.com/cms/en/product/promopages/primepack/. [Accessed: 14-Nov-

2017].

64

[8] A. Cissé, G. Massiot, C. Munier, P.-E. Vidal, F. Carrillo, and M. Iturriz, “Development of

a High Temperature Power Module Technology with SiC Devices for High Density Power

Electronics,” 2011.

[9] C. Neeb, L. Boettcher, M. Conrad, and R. W. De Doncker, “Innovative and reliable power

modules: A future trend and evolution of technologies,” IEEE Ind. Electron. Mag., vol. 8,

no. 3, pp. 6–16, 2014.

[10] S. M. Sohel Murshed and C. A. Nieto de Castro, “A critical review of traditional and

emerging techniques and fluids for electronics cooling,” Renew. Sustain. Energy Rev., vol.

78, no. April, pp. 821–833, 2017.

[11] W. Vigrass, “Calculation of semiconductor failure rates,” 2010.

[12] S. Speaks, “Reliability and MTBF Overview,” 2002.

[13] E. Laloya, O. Lucía, H. Sarnago, and J. M. Burdío, “Heat management in power

converters: from state-of-the-art to future ultra high efficiency systems ,” IEEE Trans.

Power Electron., vol. 31, no. 11, pp. 7896–7908, 2016.

[14] S. Fu, Y. Mei, X. I. N. Li, P. Ning, and G. Lu, “Parametric Study on Pressureless

Sintering of Nanosilver Paste to Bond Large-Area (≥100 mm^2) Power Chips at Low

Temperatures for Electronic Packaging,” J. Electron. Mater., vol. 44, no. 10, pp. 3973–

3984, 2015.

[15] D. J. Griffiths, Introduction to electrodynamics. Prentice Hall, 1999.

65

[16] R. A. Serway, Principles of physics. Saunders College Pub, 1998.

[17] D. C. Giancoli, Physics : principles with applications. Prentice Hall, 1995.

[18] “Thermal Conductivity.” [Online]. Available: http://hyperphysics.phy-

astr.gsu.edu/hbase/Tables/thrcn.html. [Accessed: 22-Dec-2017].

[19] V. R. Manikam and K. Y. Cheong, “Die Attach Materials for High Temperature

Applications : A Review,” vol. 1, no. 4, pp. 457–478, 2011.

[20] S. J. L. Kang, Sintering-Densification,Grain Growth,and Microstructure, no. ISBN 978-0-

7506-6385-4. 2005.

[21] W. D. Kingery, H. K. Bowen, and D. R. Uhlmann, Introduction to ceramics. Wiley, 1976.

[22] G. Bai, “Low-Temperature Sintering of Nanoscale Silver Paste for Semiconductor Device

Interconnection,” Mater. Sci., no. October, pp. 45–48, 2005.

[23] R. M. German, P. Suri, and S. J. Park, “Review: Liquid phase sintering,” J. Mater. Sci.,

vol. 44, no. 1, pp. 1–39, 2009.

[24] R. M. German, Powder Metallurgy and Particulate Materials Processing, 1st ed. New

Jersey, 2005.

[25] C. Buttay et al., “Die Attach of Power Devices Using Silver Sintering - Bonding Process

Optimization and Characterization e Morel To cite this version : Die Attach of Power

Devices Using Silver Sintering – Bonding Process Optimisation and Characterization

66

Introduction,” 2012.

[26] P. Peng, A. Hu, A. P. Gerlich, G. Zou, L. Liu, and Y. N. Zhou, “Joining of Silver

Nanomaterials at Low Temperatures : Processes , Properties , and Applications,” Appl.

Mater Interfaces, vol. 7, pp. 12597–12618, 2015.

[27] H. Djohari and J. J. Derby, “Transport mechanisms and densification during sintering: II.

Grain boundaries,” Chem. Eng. Sci., vol. 64, no. 17, pp. 3810–3816, 2009.

[28] “THE CENTER OF CURVATURE AND THE EVOLUTE.” [Online]. Available:

http://www3.ul.ie/~rynnet/swconics/E-COC.htm. [Accessed: 31-Dec-2017].

[29] D. W. Richerson, Modern ceramic engineering : properties, processing, and use in design.

M. Dekker, 1992.

[30] W. E. Lee, “Ceramic processing and sintering,” Int. Mater. Rev., vol. 41, no. 1, pp. 36–37,

1996.

[31] H. Miyoshi, K. Endoh, and S. Kurita, “Application of silver nano particle to pressureless

bonding onto a copper surface-consideration of substitute material for lead solder,” in

International conference on integrated power electronics system, 2014, pp. 25–27.

[32] H. Schwarzbauer and R. Kuhnert, “Novel large area joining technique for improved power

device performance,” IEEE Trans. Ind. Appl., vol. 27, no. 1, pp. 93–95, 1991.

[33] H. Djohari and J. J. Derby, “Transport mechanisms and densification during sintering: I.

Viscous flow versus vacancy diffusion,” Chem. Eng. Sci., vol. 64, no. 17, pp. 3799–3809,

67

2009.

[34] D. L. JOHNSON and I. B. CUTLER, “Diffusion Sintering: I, Initial Stage Sintering

Models and Their Application to Shrinkage of Powder Compacts,” J. Am. Ceram. Soc.,

vol. 46, no. 11, pp. 541–545, 1963.

[35] W. Beere, “A unifying theory of the stability of penetrating liquid phases and sintering

pores,” Acta Metall., vol. 23, no. 1, pp. 131–138, 1975.

[36] P. Thiyagasundaram, B. V. Sankar, and N. K. Arakere, “Elastic Properties of Open-Cell

Foams with Tetrakaidecahedral Cells Using Finite Element Analysis,” AIAA J., vol. 48,

no. 4, pp. 818–828, 2010.

[37] A. Masson et al., “Die attach using silver sintering . Practical implementation and analysis

To cite this version : Die attach using silver sintering practical implementation and

analysis,” 2013.

[38] K. Skandan, “Processing of Nanostructured Zirconia Ceramics,” Nanostructured Mater.,

vol. 5, no. 2, pp. 111–126, 1995.

[39] K. S. Siow, “Mechanical properties of nano-silver joints as die attach materials,” J. Alloys

Compd., vol. 514, pp. 6–19, 2012.

[40] A. Large-area, T. G. Lei, J. N. Calata, G. Lu, X. Chen, and S. Luo, “Low-Temperature

Sintering of Nanoscale Silver,” vol. 33, no. 1, pp. 98–104, 2010.

[41] P. Bowen and C. Carry, “From powders to sintered pieces: Forming, transformations and

68

sintering of nanostructured ceramic oxides,” Powder Technol., vol. 128, no. 2–3, pp. 248–

255, 2002.

[42] M. J. Mayo, “Processing of nanocrystalline ceramics from ultrafine particles,” Int. Mater.

Rev., vol. 41, no. 3, pp. 85–115, 1996.

[43] P. Peng, A. Hu, and Y. Zhou, “Laser sintering of silver nanoparticle thin films :

microstructure and optical properties,” pp. 685–691, 2012.

[44] R. F. K. Gunnewiek and R. H. G. A. Kiminami, “Two-step microwave sintering of

nanostructured ZnO-based varistors,” Ceram. Int., vol. 43, no. 1, pp. 847–853, 2017.

[45] R. Nicula, V. D. Cojocaru, M. Stir, J. Hennicke, and E. Burkel, “High-energy ball-milling

synthesis and densification of Fe-Co alloy nanopowders by field-activated sintering

(FAST),” J. Alloys Compd., vol. 434–435, no. SPEC. ISS., pp. 362–366, 2007.

[46] S. W. Wang, L. D. Chen, and T. Hirai, “Densification of Al 2 O 3 powder using spark

plasma sintering,” J. Mater. Res., vol. 15, no. 4, pp. 982–987, 2000.

[47] S. Ma, V. Bromberg, L. Liu, F. D. Egitto, P. R. Chiarot, and T. J. Singler, “Low

temperature plasma sintering of silver nanoparticles,” Appl. Surf. Sci., vol. 293, pp. 207–

215, 2014.

[48] H. He, R. Fu, D. Wang, X. Song, and M. Jing, “A new method for preparation of direct

bonding copper substrate on Al2O3,” Mater. Lett., vol. 61, no. 19–20, pp. 4131–4133,

2007.

69

[49] W. H. Tuan and S. K. Lee, “Eutectic bonding of copper to ceramics for thermal

dissipation applications - A review,” J. Eur. Ceram. Soc., vol. 34, no. 16, pp. 4117–4130,

2014.

[50] Agilent Technologies, “Chapter 17 Material Expansion Coefficients,” in Laser and Optics

User’s Manual, U.S.A, 2002, pp. 17-1-12.

[51] Rogers Corporation, “curamik ® Ceramic Substrates Technical Data Sheet,” 2016.

[52] T. D. Sheet, “Loctite Ablestik Ssp 2020,” pp. 2–3, 2012.

[53] B. K. Kim et al., “Origin of surface defects in PCB final finishes by the electroless nickel

immersion gold process,” J. Electron. Mater., vol. 37, no. 4, pp. 527–534, 2008.

[54] G. A. Di Bari, “Electrodeposition of Nickel,” in Modern Electroplating, 5th ed., New

Jersey: John Wiley & Sons Inc., 2011, pp. 79–114.

[55] N. Institute, “Nickel Plating Handbook,” Nickel Institute, 2014. [Online]. Available:

https://www.nickelinstitute.org/~/media/Files/TechnicalLiterature/NPH_141015.ashx.

[Accessed: 02-Jun-2017].

[56] P. A. Kohl, “Electrodeposition of Gold,” in Modern Electroplating, 5th ed., New Jersey:

John Wiley & Sons Inc., 2011, pp. 115–130.

[57] Dynatronics Inc., “Operating Manual for DP/DPR/DPD Series Power Supplies with

Mach2 Host Controller Models:DP(R)(D)20-30-100,” 2009.

70

[58] Yamato Scientific Co.LTD., “Instruction Manual for Forced Convection Constant

Temperature Drying Oven Model DKN302/402/602/612/812/912,” Tokyo, Japan, 2004.

[59] U. of T. Cerntre for Nanostructure Imaging, Department of Chemistry, “Centre for

nanostructure imaging-Instruments.” [Online]. Available:

http://www.chem.utoronto.ca/facilities/CNI/instruments.html. [Accessed: 11-Nov-2017].

[60] National Institutes of Health, “ImageJ.” [Online]. Available:

https://imagej.nih.gov/ij/index.html. [Accessed: 21-Dec-2016].