Controlling Interfacial Reaction in Aluminium to Steel Dissimilar Metal Welding

A thesis submitted to the University of Manchester for the degree of

Doctor of Philosophy

in the faculty of Engineering and Physical Sciences

2015

Lei Xu

School of Material

Content

CONTENTS List of Figures ...... 6

List of Tables ...... 21

List of Publications ...... 22

Abstract ...... 23

Declaration ...... 24

Acknowledgements ...... 26

Chapter 1 Introduction ...... 27

1.1 State of the Art and Motivation ...... 27

1.1.1 Greenhouse Gas Emissions ...... 27

1.1.2 Light Weighting ...... 29

1.1.3 Welding Methods ...... 31

1.2 Objectives of the Study ...... 34

1.3 Thesis Overview ...... 35

Chapter 2 Literature Review ...... 37

2.1 Introduction ...... 37

2.2 The Principles of Welding...... 37

2.3 Welding Techniques ...... 39

2.3.1 Fusion Welding ...... 40

2.3.2 Solid State Welding ...... 47

2.4 Base Material ...... 68

2.4.1 Aluminium Alloy ...... 68

2.4.2 Low Carbon Automotive Steel ...... 70

2.5 Intermetallic Compounds ...... 71

2.5.1 Crystallographic Structures of η and θ Phases ...... 72

Content

2.5.2 The Evolution of the IMC Layer ...... 75

2.5.3 Microstructure of the IMC Layers ...... 79

2.5.4 Growth Kinetics of the IMC Layer ...... 82

2.6 Summary ...... 90

Chapter 3 Experimental Methods ...... 93

3.1 Parent Materials ...... 93

3.2 Ultrasonic Spot Welding ...... 94

3.3 Friction Stir Spot Welding ...... 97

3.4 Temperature Measurement ...... 99

3.5 Metallographic Sample Preparation ...... 100

3.6 Heat Treatment ...... 101

3.7 Lap Shear Testing ...... 103

3.8 Hardness Testing ...... 103

3.9 Surface Profile ...... 106

3.10 Scanning Electron Microscope (SEM) ...... 106

3.10.1 Scanning Electron Microscope (SEM) ...... 106

3.10.2 Energy Dispersive X-ray (EDX) ...... 108

3.10.3 Electron Backscatter Diffraction (EBSD) ...... 108

3.11 Focused Ion Beam (FIB) Milling ...... 109

3.12 Transmission Electron Microscope (TEM) ...... 111

3.13 Image Analysis ...... 114

Chapter 4 Weld formation and interface layer in Fe-Al dissimilar joints ...... 116

4.1 Introduction ...... 116

4.2 Parent Sheet Surface Condition ...... 117

4.3 The Microstructure And Mechanical Properties Of Fe-Al Dissimilar Joints Made By USW ...... 118

Content

4.3.1 Temperature Histories Of The USW Joints ...... 118

4.3.2 Weld Characterization ...... 119

4.3.3 Analysis of the Joint Interfaces ...... 120

4.3.4 Lap Shear Testing ...... 125

4.4 The Development Of The IMC Layer In Fe-Al Dissimilar Joints Made By FSSW ...... 129

4.4.1 Temperature Histories In Pin-Less FSSW ...... 129

4.4.2 Microstructure Of The IMC Layer In The Pin-Less FSSW Joints ...... 130

4.4.3 The Interface In The Joints Made By ABC-FSSW ...... 134

4.5 Discussion ...... 135

4.5.1 The Effect Of The Alloy Elements On The Phase Composition In Dissimilar Fe-Al Joints ...... 135

4.5.2 Development of the IMC layer at the interface in joints made by USW and FSSW ...... 136

4.5.3 The Different Growth Rates Of The IMC Layer During The Two Welding Processes ...... 143

4.5.4 Relationship Between Interfacial Microstructure And Mechanical Properties Of The USW Joints ...... 145

4.6 Summary And Conclusions ...... 148

Chapter 5 The static growth behaviour of intermetallic compounds in steel and aluminium dissimilar welds ...... 150

5.1 Introduction ...... 150

5.2 The Influence Of The Initial State On The IMC Layer During Heat Treatment ...... 151

5.2.1 The Microstructure Development Of The IMC Layer In The ABC-FSSW Joints During Annealing ...... 151

5.2.2 The Microstructure Of The IMC Layer In The USW Joints After Annealing ...... 157

5.2.3 The Development Of The Grain Structure In The IMC Layer ...... 161

5.2.4 The Growth Kinetics Of The IMC Layer ...... 164

Content

5.3 The Influence Of Zinc On The IMC Layers During Heat Treatment...... 168

5.4 Discussion ...... 170

5.4.1 The Growth Direction Of The IMC Layer In Dissimilar Fe/Al Couples After Heat Treatment...... 170

5.4.2 Discussion Of The Development Of The IMC Layer During Heat Treatment ....171

5.4.3 Discussion Of The Development Of The Grain Structure In The IMC Layer .....176

5.4.4 The Growth Kinetics Of The IMC Layer During The Heat Treatment ...... 177

5.5 Summary and Conclusions ...... 181

Chapter 6 Modelling Of The IMC Layer Growth In An Fe-Al Dissimilar System ...... 184

6.1 Introduction ...... 184

6.2 The Double-IMC Phase Diffusion Model ...... 185

6.3 The Grain Growth Model ...... 187

6.4 Modelling Results ...... 188

6.4.1 The Kinetic Parameters Of The Two Diffusion Mechanisms ...... 188

6.4.2 Validation Of The Modelling ...... 190

6.4.3 The Application Of The Model To Heat Treatments ...... 190

6.4.4 The Application Of The Model In USW ...... 196

6.5 Discussion ...... 200

6.5.1 The Effect Of Grain Growth Of Each Phase In The IMC Layer ...... 200

6.5.2 The Application Of The Model During Heat Treatment ...... 200

6.5.3 The Application Of The Model In USW ...... 204

6.6 Summary And Conclusions ...... 206

Chapter 7 Conclusions and future work ...... 208

7.1 Conclusions ...... 208

7.2 Future Work ...... 211

Reference ...... 213

Content

LIST OF FIGURES

Fig. 1.1 (a) GDP and (b) predicted CO2 emissions by Venezuela for the period 1980–2025. Dots correspond to the official dataset (Robalino-López 2015). P 28

Fig. 1.2 (a) Sectorial shares of global energy consumption; (b) The global potential of different sectors for CO2 mitigation by 2030 (P. Nejat 2015). P 29

Fig. 1.3 Materials for the key components of (a) a weight-optimised vehicle concept and (b) a cost-optimised concept (Lesemann 2008). P 30

Fig. 1.4 Typical stress-strain curves for (a) η (Fe2Al5) phase and (b) aluminium alloy and mild steel (Hirose 2003, Alumatter). P 32

Fig. 1.5 Phase diagram for Fe-Al system (Massalski 1986). P 33

Fig. 2.1 The principles of welding in achieving metallic continuity by bringing atoms together using (a) cold deformation and lattice strain, (b) hot deformation and dynamic recrystallization; (c) solid-phase diffusion across the original interface, and (d) a liquid provided by melting the parent materials (or substrates), without or with additional filler, and (e) establishing a bond upon epitaxial solidification of the this liquid (Messler 2004). (The original interfaces are shown by the dotted lines). P 38

Fig. 2.2 Illustration of the Resistance spot welding (RSW) process (Mallick 2010). P 41

Fig. 2.3 Reaction layer in a RSW joint made by Qiu et al. (2009) observed by (a) SEM and (b) TEM. P 42

Fig. 2.5 Cross sections of steel/aluminium joints made by (a) Sierra et al. (2008) and (b) Torkamany et al. (2010). The interface layers located in the dashed boxes for each joint are shown at a higher magnification in (c) and (d), respectively. P 44

List of Figures

Fig. 2.6 Examples of a (a) Cross section and (b) interface of steel-Al joints made by GTAW (Sierra 2008). The fusion zones (FZ) and heat affected zones (HAZ) are also indicated. P 46

Fig. 2.7 Schematic diagram of the friction stir welding process (Mishra 2005). P 47

Fig. 2.8 A typical macrograph showing the various microstructure zones found in aluminium welds produced by FSSW (Mishra 2005). P 48

Fig. 2.9 Schematic illustration of the friction stir welding process applied to a dissimilar metal joints between stainless steel and an Al alloy (a) left view, and (b) top view (Lee 2006). P 49

Fig. 2.10 Macroscopic overview of the cross-section of a FSW dissimilar joint between an AA6013 alloy and an X5CrNi18-10 stainless steel (Uzun 2005). P 50

Fig. 2.11 Interface microstructures seen in dissimilar joints in an AA7075 – mild steel by FSW with different rotation speed (a) 600 rpm, (b) 700 rpm, (c) 800 rpm (Tanaka 2009). P 51

Fig. 2.12 (a) STEM BF micrograph of the interface region of a commercially pure aluminium Al99.5 – DC01 FSW after welding, (b) high resolution image with corresponding Fast-Fourier-Transform of the region highlighted in (a); (c) SEM micrograph and (b) STEM HADDF micrograph at the interface region in the joint after annealing at 500oC for 9 min (Springer 2011 (c)). P 52

Fig. 2.13 Relationships between joint tensile strength and IMC thickness in dissimilar joints between steel and aluminium alloy made by (a) Springer (2011 (c)) and (b) Tanaka (2009). In (a) empty symbols indicate failure in the Al weld region, whereas solid filled symbols indicate interfacial failure, semi-filled symbols show both interfacial failure and failure in the Al weld region. P 53

Fig. 2.14 (a) Schematic of the friction melt welding process developed to assemble metallic sheets with different melting temperatures by Rest et al. (2014). (b) Phases identification by EBSD in dissimilar joint between Fe and Al (Rest 2014). P 54

Fig. 2.15 Illustration of friction stir spot welding process (Buffa 2008). P 55

List of Figures

Fig. 2.16 Cross section of dissimilar joint between AA6016 and IF-steel made by FSSW (Bozzia 2010). P 55

Fig. 2.17 EPMA area analysis at the interface between Al alloy 6016 and IF free steel at the hook of FSSW dissimilar joint. Probe diameter: 50 nm, accelerating voltage: 15 kV, irradiation current: 10 nA (Bozzi 2010). P 56

Fig. 2.18 Cross section of a typical dissimilar joint between AA5052 and ductile steel made by FSSW without a pin (Miyagawa 2009). Weld area is highlighted inside the white dashed box. P 57 Fig. 2.19 (a) Sketch of the assigned tool axis path used by Buffa (2008) to incease he weld area by FSSW; (b) Top view and (c) etched AA section of a FSSW joint made by a modified process with a tool path compared to (c) the etched section of a joint made by the traditional process (Buffa 2008). P 58

Fig. 2.20 (a) Schematic diagram of the orbital translation path used in abrasion circle friction spot welding between DC04 and AA6111 by Chen et al. (2012), with (b) an insert showing the tool and (C) the weld surface. (d) high magnification view of the interface of dissimilar joint, and (e) HAADH image of wear debris present in Al alloy with an EDS line scan across it (white dash line) (Chen 2012). P 58

Fig. 2.21 Illustration of the four main stages used in the FSSW-Refill process (sleeve plunge variant) (Tier 2013). P 59

Fig. 2.22 (a) schematic diagram of a dual-reed USW machine, (b) the sonotrode tips used for the aluminium and steel sheets and (c) a schematic diagram of the weld sample (Prangnell 2011). P 61

Fig. 2.23 Weld cross-sections for welds produced between dissimilar alloys (6111 and 6082) showing: (a) microbonded areas at the centre of the weld for a low energy (200 J), and (b) and (c) interface folds and the macroscopic wave-like displacement of the join line for higher welding energies (450 and 700 J). In (d) and (e) examples are given in samples produced with similar 6111 alloy sheets of (d) a microbonded region near the edge of a low energy weld (300 J) and (e) interface folds in a 450 J weld (Bakavos 2010). P 62

List of Figures

Fig. 2.24 Schematic illustrations of aluminium displacement and formation of defects at the weld interface of (a) aluminium to aluminium (Bakavos 2010) and (b) aluminium to steel joint in short welding time (Haddadi 2012). P 64

Fig. 2.25 EBSD orientation maps (Euler contrast) from AA6111 similar weld produced under optimum conditions (750 J, 40 MPa) show: (a) a typical slice through the weld centre, (b) an interface 'swirl' and the grain structure it contains at a higher magnification, (c) and (d) shear bands close to the weld-line (Bakavos 2010). EBSD orientation maps (Euler contrast) from DC04-AA6111 dissimilar weld produced under a 1.4 kN clamping pressure show (e) a typical slice through the weld centre in joint with welding time 2.0 s, and weld interface with welding time of (f) 0.4 s, (g) 1.0 s and (h) 2.0 s with higher magnification images (Haddadi 2012). P 65

Fig. 2.26 Back scatter electron SEM image of the interfacial reaction between aluminium and DC04 bare steel for different welding time of (a) 0.25, (b) 1.5, (c) 3.0 seconds and (d) higher magnification SEM image of a thin sample produced by Focussed Ion Beam (FIB) for a 0.4 sec weld (Haddadi 2012). P 66

Fig. 2.27 Penetration of Aluminium in European Cars; (a) aluminium alloy weight in cars vs. year; (b) growth of aluminium in car from 2000 to 2005 (Hirsch 2011). P 67

Fig. 2.28 The Fe-Al equilibrium phase diagram (Massalski 1986). P 71

Fig. 2.29 The three - dimensional framework of Fe and Al atoms in the Fe2Al5 structure (Burkhardt 1994). P 73

Fig. 2.30 (a) Crystal structure and (b) crystal symmetry of the Fe2Al5 phase (Hirose 2003). P 73

Fig. 2.31 (a) Cross section micrographs of the tongue-like morphology of intermetallic

Fe2Al5 layers at 700 °C. (b) Mild steel coated by hot-dipping for 180 seconds, after etching, revealing the Fe–Al/steel substrate interface morphology of the Fe2Al5 phase (Cheng 2009). P 74

List of Figures

Fig. 2.32 (a) 3-D interactive structure (Griger 1986) and (b) crystal symmetry for the unit cell of FeAl3 phase (Black 1955 (a)). In (a), orange balls correspond to iron atoms, and grey balls aluminium atoms. P 75

Fig. 2.33 Microstructure of IMC layer in dissimilar couples between steel and pure Al after solid/solid interdiffusion at 600oC for (a) (b) 1 h, (c) 8h, and (d) 16 h. In (b) EBSD mapping is used to show the phases in the interface area with different colours (Springer 2011 (b)). P 76

Fig. 2.34 Evolution of the IMC layer in an aluminium coated steel (a) at as-coated state, and at 750oC for (b) 8min, (c) 60 min, (d) 10 hours, (e) 48 hours and (f) 72 hours (Cheng 2008). P 77

Fig. 2.35 Evolution of the IMC layer in an aluminium coated steel by Windmann et al. (2013). P 78

Fig. 2.36 Interface structure in the cross section of dissimilar couples between aluminium and iron (a) in the solid state at 873 K for 3 h (Naoi 2007), and (b) in the liquid state at 1013 K for 1 h (Tang 2012). P 80

Fig. 2.37 EBSD inverse pole figure maps for the reaction layer between steel and pure Al. The η phase region is marked by double white arrows. (a) 600oC 1h, (b) 600oC 16 h, and (c) 675oC 30s (Springer 2011 (b)). P 81

Fig. 2.38 (a) Unit cells of Fe and Fe2Al5 illustrating the orientation relationship between them, and (b) the corresponding [100] Fe stereograph (Wang 2010). P 81

Fig 2.39 The mean thickness l of the IMC layer vs. the square root of the annealing time t by (a) Naoi and Kajihara (2007) and (b) Shibata et al. (1966). Straight lines indicate the best fit from Eq. (2-2). The parabolic coefficient K of the IMC layer vs. the reciprocal of the annealing temperature T by (c) Naoi and Kajihara (2007) and Shibata et al. (1966) with different slopes, and (d) Springer et al. (2011 b) and Eggeler et al. (1985) together with data in (c). Straight lines show the calculations from Eq. (2-3). P 84

Fig. 2.40 (a) The parabolic coefficient K of the η layer vs. the reciprocal of the annealing temperature T shown as open squares. The evaluations for the interdiffusion coefficient D

List of Figures of η phase are indicated as open circles. Straight lines show the calculations from equations (2-3) and (2-4) for k and D, respectively. (b) The interdiffusion coefficient D vs. the reciprocal of the annealing temperature T shown as various straight lines for the α, β,

γ and δ phases, which stand for Fe, η (Fe2Al5), Al and other Fe-Al intermetallic compounds invisible here, respectively. P 86

Fig. 3.1 Microstructure of the parent materials before welding (a) DC04, (b) AA6111-T4, and (c) AA7055-solution in BSE mode. P 95

Fig. 3.2 The ultrasonic welding machine setup used in this work showing; (a) a scheme illustration of the ultrasonic welding components, (b) the tip geometries, (c) the welding power supply and control systems (Haddadi 2012). P 96

Fig. 3.3 The geometry of the test coupon joints made by USW. P 97

Fig. 3.4 Schematic diagram of the translation path, tools and weld surface on the Al sheet for (a) conventional pinless FSSW and (b) the abrasion circle friction spot welding process (YingChun, 2011). P 98

Fig. 3.5 Schematic illustration of the thermocouple measurement positions used in USW at (a) top view and (b) lateral view, and (c) shows the joint with thermocouple after welding. (d) Lateral view of the thermocouple measurement positions used in FSSW. The black solid line in (a) indicates the channel position, and the red lines and circles in (b) and (d) indicate the thermocouples position. P 99

Fig. 3.6 Configuration of the weld sample section studied and cutting position in each joint. (a) the whole joint, (b) the lap section, with the cutting position shown as the dashed-straight line, and (c) a schematic diagram of a cross section of a dissimilar USW joint. The observation area for SEM and TEM is marked in the black box in (c). P 101

Fig. 3.7 Schematic diagramof the lap shear weld coupon test specimen, (a) top view and (b) side view. P 103

Fig. 3.8 Schematic diagram of the microhardness technique (Panteli 2012). P 104

List of Figures

Fig. 3.9 Schematic diagrams for the positions used for microhardness measurement. The positions are shown as the dashed lines. It should be note that the distance between the microhardness measurement positions and interface was at least 100 μm. P 105

Fig. 3.10 (a) The position of three indentation positions in the cross section of welds. (b) and (c) illustrations how marked indentations were used to correlate the rate of IMC growth into each substrate, with and without IMC layer, respectively. The original interface is shown as the dashed line in (c). P 105

Fig. 3.11 Illustrations for the principle of FIB for (a) milling and (b) deposition (Reyntjens 2001). P 110

Fig. 3.12 Schematic diagram for the procedure of TEM sample preparation by FIB.

(a) Pt deposition, (b) Bulk-out, (c) U-cut, (d) Lift-out, (e) Mounting, (f) Thinning and cleaning (FEI 2007). P 111

Fig. 3.13 Signals generated when a high-energy beam of electrons interacts with a thin specimen (Williams 2009). P 112

Fig. 3.14 Illustration of the TEM optical system in (a) imaging mode and (b) diffraction mode (Rodenburg 2004). P 113

Fig. 3.15 Measurement procedure with ImageJ. (a) The example of an original SEM image. (b) The scale setup. (c) Measuring the area by choosing the outline of the target area with 'Freehand selections' function. (d) Measuring the length by drawing a straight line with 'Straight' function. The yellow lines were draw to choose the objects.

Fig. 4.1 Surface roughness (Ra) profiles of the parent sheets after preparation by grinding with #120 grit SiC paper and cleaning in acetone. P 116

Fig. 4.2 Results from thermal measurements made at the weld interface: (a) example temperature histories from AA7055- DC04 welds, and (b) the peak temperatures reached when welding both material combinations, as a function of welding time. P 117

Fig. 4.3 Examples of weld cross sections for the two material combinations for a weld time of 1.5 seconds (SEM montage); (a) AA6111 aluminium – DC04 steel, (b) AA7055

List of Figures aluminium – DC04 steel. The box in each image designates the area used for measuring the IMC reaction layer thickness. P 118

Fig. 4.4 (a) The thickness distribution in the tow combinations with a welding time of 1.5 seconds. (b) Average thickness of the IMC layer found at the weld interface, for both material combinations, as a function of welding time. P 120

Fig. 4.5 SEM (left) and TEM (right) images of the IMC reaction layers seen in the AA6111- DC04 steel ultrasonic welds (a) and (b) after a short 0.3 s, and (c) and (d) medium 1.5 s welding time. In (e) and (f) a comparison is provided to the equivalent AA7055- DC04 steel welds after a welding time of 1.5 s. In the TEM images the dashed lines indicate the interfaces between the different phases. The solid lines indicate the position used for the composition line-scans shown in Fig. 4.6. P 121

Fig. 4.6 (a) EBSD phase map of the dual phase IMC reaction layer seen in the AA6111- DC04 sample after a welding time of 1.5 seconds. Note; the columnar grain structures of the IMC phases and small FeAl3 particles (red) embedded in the aluminium matrix away from the interface. In (b) and (c) pole figures are shown from the η and θ phases, respectively, where ND is normal to the interface plane. P 123

Fig. 4.7 Results from EDX composition line-scans obtained by TEM across IMC interface layers for; (a) the AA6111-DC04 steel and (b) the AA7055-DC04 steel joints after a welding time of 1.5 s. (c) Highlights the concentration of additional alloy elements (without Fe and Al) in the AA7055-DC04 joints using an expanded scale. The vertical dashed lines are the interface positions between different phases shown in the images. P 124

Fig. 4.8 Interface region microstructures observed at longer ultrasonic welding times for both material combinations; (a) AA6111-DC04 joints with a welding time of 2 seconds and the AA7055-DC04 material combination with welding times of (b) 1.5 and (c) 2 seconds. Examples from EDX point analysis results are provided at the positions indicated (refer to key in each image). P 125

Fig. 4.9 (a) Example load-extension curves for the two joints with a welding time of 1.5 seconds. Average (b) Failure loads and (c) fracture energies obtained from lap shear tests

List of Figures conducted on welds produced with both material combinations, as a function of welding time. P 126

Fig. 4.10 Fracture surfaces of the peak lap shear strength samples (1.5 second welding time) showing; macro-images (a), (d) and (g), (j) of the steel and aluminium surfaces for the DC04-AA6111 and DC04-AA7055 weld combinations, respectively, with accompanying higher magnification SEM images from the regions highlighted in the macro-images of the steel side, (b) (c), (h) and (i), and the aluminium side of each joint in (e), (f), (k) and (l). Examples of the local composition of specific regions determined by EDX point analysis are indicated (refer to key in each image). P 127

Fig. 4.11 (a) The temperature histories in pin-less FSSW at different welding times. (b) The peak temperatures in the joints made by ABC-FSSW at different welding times. P 129

Fig. 4.12 Typical cross section of a dissimilar joint made by pin-less FSSW. The steel sheet was placed at the bottom, and the aluminium sheet was placed at the top. P 131

Fig. 4.13 Interface microstructure in the joints produced by pin-less FSSW with welding times of (a) 1 s, (b) 2 s, (c) 5 s and (d) 10 s in SEM. (e) and (f) Thickness distribution of the IMC layer in the joints with different welding times. P 132

Fig. 4.14 Interface microstructure in the joints produced by pin-less FSSW with welding times of (a) and (b) 2 s, (c) 5 s in TEM. (d) EDX line scan profile for the line in (c). The white dashed lines are the interfaces between different materials. P 133

Fig. 4.15 TEM images of the IMC layers in the dissimilar DC04-AA6111 combination made pin-less FSSW with a welding time of 10 seconds; (a) overview, (b) EDX line scan result of the white line in (a), (c) and (d) images with higher magnification in white boxes in (a).

The area between the dashed lines in (c) is the η (Fe2A5) phase and is the θ (FeAl3) phase in (d), respectively. P 134

Fig. 4.16 (a) Cross section for the DC04-AA6111 joints made by ABC-FSSW, with the parameters as following: travel speed 1000 mm/min, rotation rate of 800 rpm, plunge depth of 0.1 mm and a welding time of 1 s. (b) and (c) are the 'clean' interface seen in the SEM and TEM, respectively. P 135

List of Figures

Fig. 4.17 CALPHAD predicted equilibrium phase fractions as a function of Fe content at 500oC for; (a) pure iron and aluminium, (b) AA7055-DC04 and (c) AA6111-DC04 steel. P 137

Fig. 4.18 IMC layer thickness against the distance from the weld centre in the (a) USW joints and (b) pin-less FSSW joints with different welding times. P 140

Fig. 4.19 FE temperature field modelling in DC04-AA6111 joints made by (a) USW with a welding time of 1 second (Jedrasiak 2015) and (b) FSSW with pin less tool with a welding time of 1 second (Jedrasiak 2012). It should be noted that the thickness of the same metal is the same, which is 0.9 mm for AA6111 on the top and 0.97 mm for DC04 on the bottom. P 14

Fig. 4.20 Comparison of thicknesses of the IMC layers been in dissimilar DC04-AA6111 combinations (a) with increasing welding time and (b) with net weld energy made by the two different welding methods. P 144

Fig. 4.21 Comparison of the interface temperatures in USW and FSSW (a) Typical temperature history curves in FSSW and USW. (b) Peak temperatures at the interface of the dissimilar DC04-AA6111 joints during FSSW and USW as function of welding time. The welding conditions were for USW nominal applied power of 2.5 kW, pressure of 29 MPa and welding time of 1.5 seconds, for pin-less FSSW plunge and retraction rates of 1.7 mm/s and 0.8 mm/s, rotation speed of 1600 rpm, plunged depth of 0.6 mm and welding time of 10 seconds. P 146

Fig. 4.22 (a) The peak load and (b) fracture energy against the IMC layer thickness in DC04-AA6111 and DC04-AA7055 joints. P 147

Fig. 5.1 Microstructure development of the IMC layers in an ABC-FSSW joint at 400oC for (a) 0, (b) 8, (c) 16, (d) 64 and (e) 128 hours. (f) EDX line scan result from the arrow in (e). The TEM sample positions are shown in the black box. The Fe-Al IMC particles embedded in the aluminium substrate are highlighted with the white arrows. P 152

Fig. 5.2 Microstructure for the IMC layer in the ABC-FSSW welded joint after annealing at 400oC for 64 hours. (a) STEM image of the interface with EDX analysis along the line

List of Figures indicate. (b) Grain structure of the η phase adjacent to the steel substrate, in b position in (a), with the SADP taken from the dashed circle. (c) STEM image of the thin IMC layer at the interface, in position c in (a), with a dashed line as the interface between the DC04 and IMC layer, with the SADP taken from the dashed circle. (d) EDX map from the box in (a). (e) EDX line scan result from the black line in (a). P 153

Fig. 5.3 TEM images for the IMC layer in the ABC-FSSW joint following annealing at 550oC for 40 minutes with different magnifications. (a) Overview, (b), (c), (d) from the positions marked in (a). (b) Is adjacent to DC04, while (d) is adjacent to AA6111. P 154

Fig. 5.4 (a) IMC layer in the ABC-FSSW joint after annealing at 570oC for 2 hours. (b) TEM overview from the black box marked in (a); (c) and (d) IMC layers in different positions near the aluminium substrate with higher magnification from solid box marked in (b), (e) EDX line scan profile from the line shown in (d), (f) Grain structure in the IMC layer in the dashed box in (b). P 155

Fig. 5.5 (a) Grain structure and (b) colour coded EBSD phase mapping of the ABC-FSSW joint after heat treatment at 500oC for 30 minutes. Yellow- ferrite, Red- η, Green- aluminium. P 156

Fig. 5.6 Microstructure development during annealing of the IMC layers in the USW joints after heat treatment at 500oC for (a) 0, (b) 10 and (c) 30 minutes. (d) the EDX line scan results along the arrow in (c). The dashed lines are the interfaces between the four phases shown in the EDX result. P 157

Fig. 5.7 (a) The IMC layer at the interface of the USW joint in the as-welded state; (b) and (c) the IMC layer at the interface after annealing at 500oC for 10 minutes, with different magnifications. (d) High resolution images for the θ phase in the white dashed box in (c). (e) and (f) are diffraction patterns for the η and θ phases, respectively. P 158

Fig. 5.8 (a), (c) and (d) TEM images for the USW joint after annealing at 500oC for 20 min with different magnifications and places. (b) EDX measurement along the white line in (a). The dashed lines are the interfaces for the different phases shown in the EDX analysis. P 160

List of Figures

Fig. 5.9 (a), (c) and (d) TEM images of the USW joint interface after annealing at 500oC for 60 min with different magnifications. (b) EDX measurement along the white line in (a). The dashed lines are the interfaces for different phases shown in the EDX analysis. P 160

Fig. 5.10 (a) grain structure and (b) colour coded EBSD phase map of the interface in the USW joint after annealing at 500oC for 2 hours. Yellow- ferrite, Red- η, Blue- θ, Green- aluminium. P 161

Fig. 5.11 (a), (c) and (e) EBSD band contrast map, and (b), (d) and (f) grain size distribution in different joints after annealing. In particular, (a) and (b) are from the ABC-FSSW joints after annealing at 500oC for 30 minutes, (c) and (d) from the ABC-FSSW joints after annealing at 550oC for 1 hour, (e) and (f) from the USW joints after annealing at 500oC for 2 hours. The width of the columnar grains was measured as the average grain size. P 162

Fig. 5.12 (a) average η phase grain size in the two joints after heat treatment for different temperatures and times, and (b) average grain size of the η and θ phase in theUSW joints after heat treatment for different temperatures and times. P 163

Fig. 5.13 The distribution of the IMC layer thickness in (a) pin-less FSSW joints and (b) USW joints after annealing at 500oC for different times. P 165

Fig. 5.14 (a) Average thicknesses of the IMC layers in the pre welded joints after annealing vs. square root of annealing time. The joints pre welded by USW had a 1.2 μm continuous IMC layer at the interface with a welding time of 1.5 seconds, and the joints pre welded by ABC-FSSW had no continuous IMC layer at the interface with a welding time of 1 second (Chen 2012). (b) Average thicknesses of the two phases in the USW joints vs. square root of annealing time. The points in all the cases stand for the experimental data, while the straight lines are fitted results using equation 2-2. P 165

Fig. 5.15 The growth kinetics of the IMC layer in the ABC-FSSW joints measured in present study compared with data from previous studies; (a) with data from diffusion couples in the research of Shibata et al. (1966) and Springer’s (2011 a), and (b) data from pre welded couples by FSSW in the research of Springer’s (2011 a). The points in all cases stand for the experimental data, while the straight lines are fitted using equation 2-2. P 167

List of Figures

Fig. 5.16 The natural logarithm of the growth rate k, of each phase in the IMC layer vs. the reciprocal of the annealing temperature, T, for (a) the ABC-FSSW joints and (b) the USW joints. The dashed lines show the fit from Eq. 2-3 for the whole temperature ranges, and

o o the straight-dashed lines with Q1 and Q2 show the fit between 400 C - 500 C and 500 C - 570oC in Fig. 5.3 a. P 168

Fig. 5.17 Colour coded EBSD phase maps (left) and grain structure of the interface (right) in the DC04-AA7055 USW joints after annealing at 450oC for (a) and (b) 2 hours, (c) and (d) 4 hours. Yellow- ferrite, Red- η, Blue- θ, Green- aluminium. P 169

Fig. 5.18 (a) Thickness and (b) grain size of each phase in the two Fe/Al dissimilar couples pre welded by USW after annealing at 450oC for 2 and 4 hours. The growth trend is shown as the solid lines and dashed lines for each phase in DC04-AA7055 and DC04-AA6111 welds after annealing, respectively. P 169

Fig. 5.19 The IMC layer thickness in the same position from the same sample as a function of time: (a) and (b) in the as-welded state made by ABC-FSSW and (c) and (d) after annealing for 30 min at 500oC. It should be noted that (b) and (d) are higher magnification images that correspond to the dashed boxes in (a) and (c), respectively. The white dashed lines in (b) and (c) are the original interface position. P 170

Fig. 5.20 Schematic diagrams of the growth behaviour of the IMC layer at the interface in the three stages of annealing; (a) island shape η phase, (b) a continuous layer of η phase, and (c) the appearance of the θ phase. The black area indicates the η phase, while the grey area indicates the θ phase. P 171

Fig. 6.1 Schematic illustration of applying the model with a loop that divides the weld thermal profile into a series of short time steps and iterates through the full time t, temperature T profile to predict the IMC layer thickness. P 181

Fig. 6.2 (a) Schematic concentration profile of aluminum across the θ and η phases along the diffusion direction in Fe-Al diffusion couple, after modified in Wang et al. (2015). (b) EDX line scan profile in the line in Fig. 5.9. P 184

List of Figures

Fig. 6.3 The grain growth law (equation 6.6) fitted to the experimental results of average grain size of the (a) η phase and (b) θ phase as a function of annealing time from Fig. 5.5 and 5.10. P 187

Fig. 6.4 Effective diffusion coefficients of each phase as functions of annealing time during heat treatment. P 187

Fig. 6.5 Prediction and experimental results of thickness of the (a) η phase and (b) θ phase as a function of annealing time. The experimental data at 600oC was obtained from the results of Springer et al. (2011 b). P 189

Fig. 6.6 The contribution of grain boundary diffusion coefficient (gDgb) and lattice diffusion coefficient ((1-g)Dl) to the effective diffusion coefficient (Deff) as a function of grain size for (a) the η phase at 450oC, (b) the η phase at 500oC, (c) the θ phase at 450oC,

o o (d) the θ phase at 500 C. It should be noted that the gDgb are almost equal to Deff at 500 C when the grain size is over 2 μm. P 190

Fig. 6.7 The contribution of grain boundary diffusion coefficient (gDgb) and lattice diffusion coefficient ((1-g)Dl) to the effective diffusion coefficient, as a function of annealing time for (a) the η phase at 450oC, (b) the η phase at 500oC, (c) the θ phase at 450oC, (d) the θ phase at 500oC. P 191

Fig. 6.8 The effect of grain boundary diffusion on the prediction of the IMC layer thickness, as a function of annealing time for (a) the η phase and (b) the θ phase. In the figure, 'with

Dgb' indicates that the grain boundary diffusion and grain growth was included in the modeling, and the results are shown as the solid lines, while 'without Dgb' indicates that only lattice diffusion was considered in the modeling, and the results are shown as the dashed lines. P 192

Fig. 6.9 The effect of grain size on the effective diffusion coefficient, as a function of annealing time, for (a) the η phase and (b) the θ phase. In the figure, d is the fixed grain size, and the modelling results are shown as the dashed lines and dashed-dotted lines with the fixed grain sizes of d=0.05 μm and d=0.9 μm, respectively. 'Normal' indicates the modelling results are calculated with the grain growth model at each temperature, and the modelling results are shown as the solid lines, while 'without Dgb' indicates that only

List of Figures lattice diffusion was considered in the modeling, and the results are shown as the round dashed lines. P 193

Fig. 6.10 The effect of grain size on the prediction of the IMC layer thickness as, a function of annealing time, for (a) the η phase and (b) the θ phase. In the figure, d is the average fixed grain size, and the modelling results are shown as the dashed lines and dashed- dotted lines with fixed grain sizes of d=0.1 μm and d=0.7 μm, respectively. 'Normal' indicates the modelling results are calculated with the grain growth model at each temperature, shown as the solid lines. P 194

Fig. 6.11 Prediction of the thickness of each phase in the IMC layer, as a function of process time with different grain sizes. In the figure, model results with d=0.02 μm are shown as the solid lines, while model results with d=0.2 μm are shown as the dashed lines. The results for the same phase are shown in the same colour. P 196

Fig. 6.12 Prediction and experimental results of the thickness of each phase and total IMC layer width as a function of process time. In the model, d=0.02 μm was used as the average grain size. The results for the same phase are shown in the same colour. P 198

Fig. 6.13 Predicted thickness of (a) the η phase and (b) the θ phase, as a function of process time with an error of 5oC in the temperature history. The standard curves are the results with measured temperature history, while the curves marked with +5oC and -5oC indicate the results with modified temperature histories. P 199

Fig. 6.14 Predicted thickness of (a) η phase and (b) θ phase as a function of process time with different welding times. The welding times are 1.5 seconds and 4.5 seconds for the 50% shorter and 50% longer welding times, respectively. P 199

Fig. 6.15 Illustrations for the different morphologies of the IMC layer. (a) Island shape particles and (b) uniform layer. The arrows indicate the diffusion direction. P 206

List of Tables

LIST OF TABLES

Table 1.1 Performances comparison of different vehicle body designs with equivalent bending stiffness (Cui 2001). P 31

Table 1.2 Activation energies of the η phase from previous studies. P 34

Table 2.1 Thickness of the IMC layer reaction and corresponding lap shear tensile strength of laser welded joints obtained by different researchers. It should be noticed that different units were used by different researchers. * PIC is the “percentage are” of intermetallic compounds seen in the weld zone. P 45

Table 2.2 Optimised welding conditions for dissimilar welding between aluminium and un coated steel (Haddadi 2012). P 66

Table 2.3 The Fe-Al intermetallic phases in the Fe-Al binary system (Springer 2011 (b), Shahverdi 2002, Lee 2003). P 72

Table 2.4 Activation energies for η phase from previous studies between aluminium and iron by diffusion bonding. P 85

Table 3.1 Chemical compositions (wt. %) of the base materials used in the present work. P 93

Table 3.2 Parameters of the materials used in the experimental. P 94

Table 3.3 Annealing parameters for different couples. P 102

Table 5.1 Activation energies of the η phase from previous studies. P 175

Table 6.1 Calculated diffusion kinetics parameters of the two IMC phases. P 184

Table 6.2 Effect of grain size on the change to the predicted diffusion coefficient and thickness for each phase, compared with the standard values calculated by the grain growth model. P 199

List of Publications

LIST OF PUBLICATIONS

L. Xu, L. Wang, Y. Chen, J. Robson, P. Prangnell. Effect of Interfacial Reaction on the Mechanical Performance of Steel to Aluminum Dissimilar Ultrasonic Spot Welds. Metallurgical and Materials Transactions: A, DOI: 10.1007/s11661-015-3179-7.

L. Wang, Y. Wang, L. Xu, C. Zhang, J. Robson, P. Prangnell. Effect of grain boundary diffusion on growth kinetics of intermetallic compounds between Al alloy and other alloys. Materials Science & Technology 2014, 2014, 1851-1858.

L. Wang, Y. Wang, C. Zhang, L. Xu, J. Robson, P. Prangnell. Controlling Interfacial Reaction During Dissimilar Metal Welding of Aluminium Alloys. Materials Science Forum, 794-796 (2014), 416-421.

Abstract

ABSTRACT

Controlling Interfacial Reaction in Aluminium to Steel Dissimilar Metal Welding

Lei Xu

The University of Manchester for the degree of Doctor of Philosophy in the faculty of Engineering and Physical Sciences

2015

Two different aluminium alloys, AA6111 (Al-Mg-Si) and AA7055 (Al-Mg-Zn), were chosen as the aluminium alloys to be welded with DC04, and two welding methods (USW and FSSW) were selected to prepare the welds. Selected pre-welded joints were then annealed at T=400 - 570oC for different times. Kinetics growth data was collected from the microstructure results, and the growth behaviour of the IMC layer was found to fit the parabolic growth law. A grain growth model was built to predict the grain size as a function of annealing time. A double-IMC phase diffusion model was applied, together with grain growth model, to predict the thickness of each phase as a function of annealing time in the diffusion process during heat treatment. In both material combinations and with both welding processes a similar sequence of IMC phase formation was observed during the solid state welding. η-Fe2Al5 was found to be the first IMC phase to nucleate. The IMC islands then spread to form a continuous layer in both material combinations. With longer welding times a second IMC phase, θ-FeAl3, was seen to develop on the aluminium side of the joints. Higher fracture energy was received in the DC04-AA6111 joints than in the DC04-AA7055 joints. Two reasons were claimed according to the microstructure in the two joints. The thicker IMC layers were observed in the DC04-AA7055 joints either before or after heat treatment, due to the faster growth rate of the θ phase. In addition, pores were left in the aluminium side near the interface as a result of the low melting point of AA7055. The modelling results for both the diffusion model and grain growth model fitted very well with the data from the static heat treatment. Grain growth occurred in both phases in the heat treatment significantly, and was found to affect the calculated activation energy by the grain boundary diffusion. At lower temperatures in the phases with a smaller grain size, the grain boundary diffusion had a more significant influence on the growth rate of the IMC phases. The activation energies for the grain boundary diffusion and lattice diffusion were calculated as 240 kJ/mol and 120 kJ/mol for the η phase, and 220 kJ/mol and 110 kJ/mol for the θ phase, respectively. The model was invalid for the growth of the discontinuous IMC layers in USW process. The diffusion model only worked for 1-Dimensional growth of a continuous layer, which was the growth behaviour of the IMC layer during heat treatment. However, due to the highly transient conditions in USW process, the IMC phases were not continuous and uniform even after a welding time of 2 seconds. Therefore, the growth of the island shaped IMC particles in USW was difficult to be predicted, unless the nucleation stage was taken into consideration.

Declaration

DECLARATION

No portion of the work referred to in this thesis has been submitted in support of an application for another degree or qualification of this or any other university or institute of learning.

The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made.

The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual-property.pdf), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s policy on presentation of Theses.

Acknowledgements

ACKNOWLEDGEMENTS

Once completed, it is necessary to look back over the journey and remember all of the people who have supported and helped me along this tough but fulfilling road.

First and foremost, I am indebted to my supervisor Professor Phil Prangnell whose help and guidance allowed me to get started, and whose patience and occasional kicks up the backside enabled me to finish. During the writing up of my thesis, he provided me with encouragement and loads of good ideas. Thanks are also due to Dr. Joseph Robson for his support throughout my PhD.

Special thanks go to Yingchun Chen, Dave Strong, Li Wang, Yin Wang, and Chaoqun Zhang, all of whom have contributed immensely to the contents of this thesis.

I am also hugely grateful to all of my friends in Manchester who accompany with me through the wonderful time. Ruizhi, Yuanyuan, Xuekun, Tianzhu, Hao Zhao, Hao Wu, to name but a few.

I also wish to acknowledge my parents, Lin Xu and Huiling Ge, for their unwavering support in my life.

Last but not least, I would like to thank my wife, Ting Xiao, for her appearance lighting my whole life.

Chapter 1 Introduction

CHAPTER 1 INTRODUCTION

The work described in this thesis forms part of the LATEST2 (light alloy towards environment sustainable transport) portfolio of work and was funded by the EPSRC and CSC (China scholarship council). In this chapter, the background and motivation for this work is introduced, the objectives of the study are defined, and an overview of the thesis is given.

1.1 State of the Art and Motivation

1.1.1 Greenhouse Gas Emissions

Gross Domestic Product (GDP) is increasing rapidly across the world due to the development of science and technology, especially in developing countries, like the BRIC nations (Brazil, Russia, India and China), resulting in an increase in the quality of life for their citizens. However, some problems are also being caused by this development, of which greenhouse gas emissions (e.g. CO2, N2O, O3, etc) are one of the most important issues, as they have potential to greatly impact on global temperatures. Therefore, studies are being carried out to investigate resource use and the influence of greenhouse gas, as well as to find ways to control emissions.

For example, A. Robalino-López built a model to predict the GDP growth and the emission of CO2 in the medium term in Venezuela (Robalino-López 2015). In their model, they proposed four scenarios: the growth of GDP, the evolution of the energy matrix, the productive sectoral structure, and the improvement of energy efficiency for the period from 2011–2025. Four scenarios were investigated in the model, and the results for project GDP and the associated emission of CO2 are shown in Fig. 1.1:

1 Baseline scenario (BS): GDP, the energy matrix and productive sectoral structure can be extrapolated to 2011–2025 by the geometric growth rate method, evolving from a smooth trend for the period 1980–2010.

2 SC-2 scenario: GDP in 2025 will be more than double that of 2010.

3 SC-3 scenario: SC-2 scenario, with a share of renewable energies employed to reduce the use of fossil energy.

Chapter 1 Introduction

4 SC-4 scenario: SC-3 scenario, with an improvement in efficiency of energy usage implemented.

By 2025, GDP for the three scenarios could be 200 billion USD more than that in the baseline scenario, but SC-4 scenario is the only one which predicts a similar emission of

CO2 to the baseline scenario. This work thus shows that an improvement in efficiency of energy usage is more effective than the share of renewable energies in reducing the emission of CO2. As a conclusion, improvement in efficiency of energy usage could balance the additional emission of CO2 which is generated due to economic development when combined with an increase in the proportion of renewable energies. Therefore, the area of improvement in energy efficiency is an attractive topic for controlling the future emission of CO2 during economic growth.

Fig. 1.1 (a) GDP and (b) predicted CO2 emissions by Venezuela for the period 1980–2025. Dots correspond to the official dataset (Robalino-López 2015).

At present, the top three energy consumers in the world are transportation (33%), industry (29%), and residential (27%), and this composition is similar to the proportions of

CO2 emission, as shown in Fig. 1.2 (P. Nejat 2015). Consequently, controlling energy consumption is an effective way of reducing the emission of CO2, and the control of energy consumption in transportation becomes an interesting area as it takes up one third of the whole of global energy consumption.

Chapter 1 Introduction

Fig. 1.2 (a) Sectorial shares of global energy consumption; (b) The global potential of different sectors for CO2 mitigation by 2030 (P. Nejat 2015).

1.1.2 Light Weighting

In the automotive sector, there are several technologies being developed to increase the efficiency of energy usage, such as improving fuel efficiency, developing better engine management systems, and introducing hybrid power trains (ACEA 2014). However, by far, the most important factor is to reduce vehicle weight. For instance, a 10% reduction in mass can result in a 6 to 8% reduction in fuel consumption (WorldAutoSteel 2015). In addition, light weighting is essential to extend the range of electric vehicles to a level more compatible with customer expectations.

Significant weight savings have already been achieved in the engine and drive train by replacing steel with aluminium alloys and magnesium alloys in some components (Aghion 2001). Car manufactures are now targeting the car body as the main structure for weight reduction, as the body constitutes about 27% of the vehicle weight of a medium sized car (Jambor 1997). Early attempts have been to optimise an all-steel body, which has result in a weight saving of 7% (Jambor 1997), but this reduction needs to be substantially increased to control the emissions more effectively. However, other factors, such as performance, durability and especially cost, are also critical for the customer experience. As a result, the concept of multi-material body structures in vehicle design is becoming an important trend in the automotive industry, where the most efficient material is considered for each component to obtain the best compromise between performance and cost.

Chapter 1 Introduction

The advantage of multi-materials is that the optimal material can be employed for each component by considering all the factors that control performance and cost. There could be several different concepts for multi-material design. Two contrasting examples are given by Lesemann et al. (2008): one is a cost optimised concept with steel-intensive design, close to a traditional body design, while the other is a weight optimised concept with an aluminium-intensive design. Examples for both concepts are shown in Fig. 1.3, where the goal of the lightweight concepts was a weight reduction of at least 30% while maintaining economic feasibility.

Fig. 1.3 Materials for the key components of (a) a weight-optimised vehicle concept and (b) a cost-optimised concept (Lesemann 2008).

As another example, Cui et al. have demonstrated that proper selection procedures can result in a multi-material combination with lightweight and little cost penalty (Cui 2001). They optimized the car body by replacing some steel parts with aluminium alloys, and the crash results show that there was little difference in the overall deformed shapes of the two designs, and most importantly, the optimal multi-material body achieved a weight reduction of 13%, with the material cost increasing by 7%, as shown in Table 1.1. Therefore, it is clearly highly advantageous to apply multi-materials design in the automotive sector.

However, some disadvantages still exist in the multi-material approach, and the most notably one is the challenge of joining dissimilar materials. Due to their different thermo- physical properties (Springer 2011 a), using traditional welding methods between different materials causes significant problems, such as high residual stresses and chemical incompatibility. Most importantly, interface reactions between dissimilar metals

Chapter 1 Introduction are harmful for the joining of multi materials. Therefore, dissimilar welding becomes an interesting area, and a large amount of research work is being carried out on welding different combinations, such as aluminium to steel, aluminium to titanium, aluminium to copper, aluminium to magnesium. Of these, joining aluminium to steel is particularly vital in the automotive sector, as both metals rank among the most important engineering materials, and form the majority of vehicle designs. Consequently, studying joining aluminium to steel has becomes an important topic.

Table 1.1 Performances comparison of different vehicle body designs with equivalent bending stiffness (Cui 2001).

Body structures Bending stiffness a (mm) Weight (kg) Cost ($) V value ($) Original design 1.5744 236.3 191.2 545.6 Optimal design 1.5735 205.4 205.5 513.6

a The bending stiffness was evaluated as the deflection at the load application point under the fixed load.

1.1.3 Welding Methods

Several fusion welding methods have been applied to welding steel and aluminium alloys. For example, Sierra has made successful dissimilar joints by key-hole laser welding (Sierra 2007). However, the failure in tensile lap shear test was always located at the weld- aluminium interface, where thick intermetallic compounds (IMC) layers were formed.

Typical strain-stress curves for Fe2Al5 phase and aluminium alloy and steel are shown in

Fig. 1.4. It is obvious that the Fe2Al5 phase is more brittle than either aluminium alloy or steel. Resistance spot welding (RSW) has also been applied to dissimilar Al-steel joints by Qiu et al.. They found that the effect of the IMC layer on the tensile shear strength was related to the thickness and composition of IMC layer in the joints (Qiu 2009). Consequently, due to the high growth rate of the IMC layer of the joint interface in the liquid phase, fusion welding techniques are not the best choice for dissimilar joints between steel and aluminium.

In contrast, solid state welding methods are a better option due to the lower energy input, and less IMC formation in the welding process. Friction stir welding (FSW) is one possible

Chapter 1 Introduction solid state method, and has been studied by many researchers (Springer 2011 (c), Taban 2010, Tanaka 2009, Lee 2006), and it was found that with the increasing of IMC layer, joint tensile strength of dissimilar joints was decreased dramatically. However, problems still exist in FSW, such as the limitation of the size of the sheet, slow welding efficiency, and disruption of the oxide surfaces. Therefore, some improvements have been applied to FSW, such as the abrasion circle friction stir spot welding by Chen (Chen 2012), which can produce a high quality joint in a short welding time. Another effective technique is ultrasonic spot welding (USW), with a shorter weld cycle and less energy consumption than FSW (Bakavos 2010, Watanabe 2009). Although USW has advantages in welding dissimilar metals, like high efficiency, low temperatures and breaking of the oxide between the two surfaces, only a few studies have been carried out in this area (Prangnell 2011). Panteli and Haddadi have done a lot of work on optimization of welding aluminium to magnesium and aluminium to steel, respectively (Panteli 2012, Haddadi 2012). The relationship between welding parameters and mechanical property was reported, and the microstructure of base metals was observed. However, further study is needed on the growth behaviour and microstructure of IMC layer in the aluminium-steel system to fully understand how the welding process affects the reaction kinetics.

Fig. 1.4 Typical stress-strain curves for (a) η (Fe2Al5) phase and (b) aluminium alloy and mild steel (Hirose 2003, Alumatter 2015).

Chapter 1 Introduction

As the interface IMC layer plays a key role in dissimilar welding, it is important to have a good understanding of it. According to the phase diagram in Fig. 1.5 (Massalski 1986), there are many compounds in Fe-Al system; among these phases, η (Fe2Al5) is considered as the stable phase during the solid state welding process (Chen 2012), while θ (FeAl3) has also been observed in fusion welding processes (Qiu 2009) and diffusion bonding when the temperature ranged from 700 to 900oC (Bouche 1998). The reason less phases are formed is because of the lower kinetic coefficients and higher activation energies of other phases, like FeAl and FeAl2, than η and θ phases. However, the activation energies of η and θ phases are sensitive to many factors, such as temperature, presence of a solid or liquid metal phase, and other alloy elements in the system. As a result, the activation energies reported have a wide range of different values, as shown in Table 1.2. Thus it is currently difficult to build a model to predict the growth of IMC layer in a particular welding process.

Fig. 1.5 Phase diagram for Fe-Al system (Massalski 1986).

The literature review in Chapter 2 will demonstrate that compared to traditional liquid state welding techniques, solid state joining methods, such as ultrasonic spot welding (USW) and friction stir spot welding (FSSW), are more suitable for the welding of aluminium alloys and steel, due to the lower temperature reached in the welding process. However, the growth behaviour of the IMC layer in joints made by solid state welding methods is still not clear, so more detail is required of for the interface and reaction

Chapter 1 Introduction behaviour at a higher magnification and resolution. In addition, the reaction kinetics and the factors are that affect the growth rate have only been widely studied in static situations, and the IMC reaction kinetics have not been assessed during dynamic welding techniques, especially in short welding process (< 1 s). Finally, little systematic data exists for the appearance of different phases in interface reactions in joints made by solid state welding processes.

Table 1.2 Activation energies of the η phase from previous studies.

Activation energy (kJ/mol) Temperature (oC)

Bouayad 2003 74.1 800 Eggeler 1986 134 786 Denner 1977 155 771 Tang 2012 123 680-770 Naoi 2007 281 600-650 Shibata 1966 226 600 Springer 2010 190 600

1.2 Objectives of the Study

The primary aim of this study is to improve current understanding of the interaction between aluminium alloys and steel during solid state friction welding processes, from the initial stage of welding to the formation of a sound joint. Therefore, the objectives of the present work are: contribute to the understanding of the fundamental reactions governing the phase formation between steel and aluminium, by observing IMC layer with high resolution techniques, like scanning electronic microscopy (SEM) and transmission electronic microscopy (TEM), in Fe/Al joints made by USW and FSSW with different welding parameters, including composition, thickness and grain structure of IMC layers in different stages; investigate the factors on build-up and growth of reaction layers, including observe IMC layers in joints made by USW among different combinations of aluminium alloys and steel to find the effect of alloy elements, and the appearance of different phases in different stages with different welding parameters; investigate the growth kinetics of IMC layer in joints with different initial states by preparing the joints

Chapter 1 Introduction with different welding parameters, to find the influence of welding parameters on the composition of IMC layer and calculate the active energies of η and θ phases. An initial attempt will be made to predict the growth behaviour of IMC layer by simulation with data collected from this study. Finally, the methods about how to control the growth of the IMC layer in dissimilar Fe/Al joints during welding are summarized with the results.

1.3 Thesis Overview

Chapter 2: Literature Review

In this chapter, the literature is reviewed to summarise the understanding in relevant areas, including the welding methods, materials used in present work, and intermetallic compounds (IMC).

Chapter 3: Experimental Methods

This chapter gives a summary of the materials and techniques used throughout the investigation.

Chapter 4: Weld formation and interface layer in Fe-Al dissimilar joints

In this chapter, the influence of different aluminium alloys and welding methods were investigated on the formation and growth stages of the intermetallic compounds (IMCs) formed at the interface in joints produced between a DC04 steel and two different aluminium alloys (AA6111 and AA7055) by high power ultrasonic spot welding (USW) and friction stir spot welding with a pin-less tool (pin-less FSSW).

Chapter 5: The static growth behaviour of intermetallic compounds in steel and aluminium dissimilar welds

In this chapter, two welding methods were applied to produce the joints with different initial interface condition. The joints with a continuous IMC layer at the interface were welded by USW with a welding time of 1.5 seconds, and the joints without a continuous IMC layer were welded by ABC-FSSW with a welding time of 1 second. Both kinds of joints were then annealed at T=400 - 570oC for different times. The grain structure of each

Chapter 1 Introduction phase in the IMC layer was observed, and the growth rate and activation energy for each phase in the IMC layer were calculated with the measured thickness data.

Chapter 6: Modelling of the IMC layer growth in Fe-Al dissimilar system

In this chapter, a double-IMC phase diffusion model was applied, together with grain growth model, to predict the thickness of each phase in the diffusion process. The effect of grain boundary diffusion for the diffusion was analyzed for each phase.

Chapter 2 Literature Review

CHAPTER 2 LITERATURE REVIEW

2.1 Introduction

This chapter contains a review of the literature relevant to the present study. The principles of welding are first summarised, and this is then followed by a review of the specific joining techniques currently under consideration by industry, for dissimilar aluminium to steel welding in manufacturing car bodies. These joining techniques can be divided into fusion (laser welding, arc welding and resistance spot welding) and solid- state welding techniques (friction stir welding and ultrasonic spot welding). Both the advantages and disadvantages of these processes will be discussed and compared, with a particular emphasis on the solid state welding methods, and especially ultrasonic spot welding that was used in this research. In the next section, the materials used in the present study are introduced, and the intermetallic reaction between aluminium and iron is then discussed with reference to the phases found in the Fe-Al system. This will include a review of the effect of compositions and the growth conditions on the intermetallic compounds formed and their microstructures. Finally, previous studies conducted on the kinetics of intermetallic phase reactions in steel to aluminium welds are reviewed, to illustrate the wide range of different activation energies reported by different researchers.

2.2 The Principles of Welding

Among the different definitions of welding, the broadest one is that it is a process in which materials of the same fundamental class are brought together and caused to become one through the formation of primary (occasionally secondary) chemical bonds under the combined action of heat and pressure (Messler, 1993). The central aim of welding is that multiple entities are made into one component by establishing continuity in the atomic structure, by the formation of chemical bonds between the surfaces of the weld members, such as metallic bonds between similar or dissimilar metals. At the atomic scale, there is an equilibrium separation distance where the attractive and repulsive forces of the atoms are balanced, resulting in a stable bond configuration and this configuration, together with an atomically clean interface, is necessary to make a sound joint by welding.

Chapter 2 Literature Review

Consequently, an ideal joint can be described as one with no weld line remnant at the interface, and with properties in the weld region equivalent to those in the bulk materials (i.e. a 100% weld efficiency). Such a joint can only be produced between two perfectly clean materials with perfectly flat surfaces, by bringing them to their atomic equilibrium spacing. In reality, due to the rough surface, the materials will initially only touch at discrete and irregular points on their surfaces when they are brought together, with a very low probability (one in every billion atoms between two well polished surfaces (Messler 2004)) of being in intimate contact. In addition, the oxide and absorbed layers on the surfaces also create a barrier to welding materials. Therefore, heat and pressure must be generally employed to improve welding performance, as both processes improve the contact of atoms in the surfaces. A schematic diagram of the metallic continuity achieved by different welding methods is shown in Fig. 2.1.

Two extreme conditions will exist depending on the pressure and heat used in a welding process. If only pressure is used, the metallic crystal lattices of each piece involved in the joining process are deformed, or strained, to achieve intimate contact, and left in a

Chapter 2 Literature Review strained state (Figure 2.1a). In this case, pressure plays three important roles in joining materials. Firstly, it increases the contact area by plastic deformation, which also results in lattice strain in surfaces near the interface in both materials (as shown in Fig. 2.1a). Secondly, it enhances the diffusion of atoms across the interface by increasing the temperature locally due to the plastic work. Thirdly, deformation breaks open the oxide on the surfaces of the weld members. In addition, the pressure holds the joint parts together while bonding occurs (Messler 2004). Solid state welding methods, like friction stir welding (FSW) and ultrasonic spot welding (USW), are mainly based on these three mechanisms. In contrast, if only heat is used, the mechanism for obtaining metallic continuity involves gross mass and atom transport via melting and liquid flow, and microscopic transport via diffusion during solidification of the base metals or filler alloys, as a result of growth of the solidifying crystals at the interface, as shown in Fig. 2.1 d and e. All fusion welding processes, like laser welding and arc welding, which involve significant melting are based on this mechanism.

2.3 Welding Techniques

Welding currently plays an important role in the manufacture of steel car bodies. Around 3000 individual joints are required in a conventional body shell. For steel, resistance spot welding (RSW) has been used for several decades (Barnes (a) 2000), together with other cost effective technologies like laser welding. However, it is difficult to weld aluminium alloys by fusion welding techniques, due to the formation of a surface oxide with a higher melting point than the bulk material and electrode damage (Gourd 1995). As a result, one key point in producing a sound dissimilar joint between a steel and aluminium alloy is breaking the oxide film, which is necessary before metallic bonding can occur. This could be done by pre-treatment of the metals. However, there is another problem in dissimilar joints between steel and aluminium alloys, which is the rapid nucleation and growth rate of intermetallic compounds (IMC) in the bonding interface, due to chemical reaction and inter-diffusion between the dissimilar metals (Qiu 2009 (a), Agudo 2007). The formation of such brittle intermetallic compounds (IMC) is difficult to avoid and is controlled by the heat input and temperature distribution at the interface, and results in poor joint mechanical properties (Borrisutthekul 2007, Xue 2010). Therefore, two solid state spot welding methods, friction stir spot welding (FSSW) and ultrasonic spot welding (USW),

Chapter 2 Literature Review have been developed for welding steel and aluminium alloys, as the IMC layer in joints made by solid state welding is much thinner than by liquid state welding. For example, under optimised parameters, the average thickness of IMC layer is 1 μm in joints produced by USW, compared with 6 μm in joint by RSW (Qiu 2009 (a), Haddadi 2012). As a result, several welding techniques are introduced in this section to discuss both their advantages and disadvantages on application to dissimilar welding between steel and aluminium.

2.3.1 Fusion Welding

Fusion welding describes joining processes where the materials in the weld area are melted by a heat source, and the materials re-solidify as a single joined unit. Different fusion techniques use different heat sources, which can affect the weld quality through several factors, such as the heating rate, penetration, and peak temperatures reached. For dissimilar welding, the welding temperature is usually determined by the material with the lower melt point, which is aluminium in Fe-Al welds. However in general, it is not a good choice for joining aluminium to steel, as the high temperatures and appearance of a liquid phase greatly increase the extent of interfacial reaction (Fan 2011, Chen 2011, Torkamany 2010, Qiu 2009 (a)).

Resistance spot welding

Resistance spot welding (RSW) is used widely in automotive manufacture. In RSW heating is caused by the Joule effect from an electric current passed between two electrodes, by transport across the interface between two sheets. As the highest resistance is designed to occur at the weld interface, localized melting occurs in the welding area, and joint is formed in the following cooling stage (Barnes 2000 (a)). A schematic description of RSW is shown in Fig. 2.2. Due to the advantages of low cost, fast operation, ease of robot manipulating, and a non-visible nugget, thousands of resistance spot welds are applied in a typical steel body-in-white (Mallick 2010). Welding parameters, like welding time and current, affect the joint strength by controlling the nugget size, and are changed according to the composition of the steel. Several previous studies have been carried out on welding steel and aluminium alloy by RSW. Oikawa et al. (1999) and Sun et al. (2004) have used aluminium clad steel sheets as an insert metal to suppress interfacial reaction,

Chapter 2 Literature Review and found that the joints with inserts had better mechanical properties. However, the IMC layer thickness in joints with the inserts was even higher than without, and as the fracture modes for the two type's joints were different, the relationship between their properties and IMC layer was not clear.

Fig. 2.2 Illustration of the Resistance spot welding (RSW) process (Mallick 2010).

Satonaka et al. (2006) and Qiu et al. (2009) added a cover plate between aluminium alloy and electrode to make RSW joints between aluminium and steel. In their research, both the thickness of IMC layer and shear strength were found to be sensitive to the composition of the steel. Furthermore, after measuring the thickness of the IMC layer along the interface, it was found that the centre of welding area had the thickest IMC layer, due to the non-uniform temperature distribution. The microstructure of IMC layer seen in this work is shown in Fig. 2.3. Two phases were found in the IMC layer: a thin layer of θ (FeAl3) next to the Al side, and tongue shape η (Fe2Al5) adjacent to the steel side of the weld. This Al-θ-η-Fe sequence has been also found at the interface between solid steel and liquid aluminium by Bouche et al. (1998).

Chapter 2 Literature Review

In this work the discontinuous reaction layer length fraction along the weld interface could be increased from 0.1 to 0.6 by controlling welding parameters, and the strength also increased from 10 to 30 MPa. However, both the aluminium clad steel sheet and cover plate increased the cost and weight of the joints. As a result, although the IMC layer in joints made by this approach could be suppressed to some extent, other problems limited its industrial application.

Fig. 2.3 Reaction layer in a RSW joint made by Qiu et al. (2009) observed by (a) SEM and (b) TEM.

Laser welding

Laser welding is an excellent industrial welding technique due to the benefits of high welding speed, high precision, reliability, and high productivity (Chen 2011). The heat source in laser welding is a high intensity beam of photons, and the concentrated beam heats a small area rapidly, resulting in a controllable interaction time between dissimilar metals. Additionally, other parameters in laser welding, such as peak power and the overlap factor, together with pulse duration in pulsed welding systems can be manipulated in laser processes (Torkamany 2010). It is these controllable parameters that make it possible to employ laser welding for joining dissimilar metals, such as steel and aluminium alloys.

An illustration showing us examples of set ups that have been used for steel to aluminium dissimilar laser welding is provided in Fig. 2.4. After fixing the work pieces a laser beam is

Chapter 2 Literature Review focused on the welding position and welding is normally conducted using a shielding gas. For lap welding (Fig. 2.4 a), it has been found that there are fewer cracks in the weld area when the steel sheet is placed on the top (Sierra 2007). For seam welding (Fig. 2.4 b), by tilting laser beam to get different wetting angles, the joint performance could be controlled (Sierra 2008). It has also been reported that filler materials, like AlSi3Mn, are a good choice to make sound dissimilar joints because the shear tensile strength can be higher than the tensile strength of aluminium sheet by increasing the wetting length of the filler material on the steel surface, which results in increasing the welding area between steel substrate and filler material (Windmann 2015). By changing the parameters, like the power density, welding speed, and penetration depth of the aluminium alloy, both the thickness of IMC layer and mechanical properties of the dissimilar joints can be controlled. Some selected results from different researches are shown in Fig 2.5 and Table 2.1.

Fig. 2.4 Illustration of the two main welding configurations reported for dissimilar joining of steel and aluminum. (a) Conduction spot welding vertical (Sierra 2008) and (b) Filler seam welding (Windmann 2015), with a wetting angle α. Although the interface shapes are different due to the different welding methods, in both cases a thick IMC layer with cracking can be seen at the interface. In this work the η

(Fe2Al5) phase was identified as the main component of the IMC layer by EDX analysis, and contained other elements like Si and Zn (Sierra 2008), while θ (FeAl3) phase was also observed between the η phase and the aluminum alloy. Cracks were formed in the IMC layer due to its high brittleness and the thermal expansion difference between the different materials. Pores were observed in both the IMC layer and interface between the IMC and the aluminum alloy, as a result of the Kirkendall effect (Torkamany 2010).

Chapter 2 Literature Review

Lap shear strengths and IMC reaction layer thicknesses reported in the literature for aluminum to steel laser welds, are summarized in Table 2.1. It should be noticed that different units were used by different researchers, and due to different weld areas and configurations, it is not easy to compare the joints' shear strengths between each other directly. Despite the different base metals, a common conclusion that can be made from Table 2.1 is that the lap shear strength of the dissimilar joints is related to the thickness of the IMC layer. Generally, the shear strength decreased with growth of the IMC layer in all research, and fracture was always found to form first in the IMC layer, which is more brittle than base metals. As a consequence, generally the joints' shear strength was less than that of the aluminum's parent sheet. For example, Windmann et al. (2015) pointed out that the interface shear stress was much less than for aluminum, but the interface shear strength could be improved by increasing the interface area.

Chapter 2 Literature Review

Material Thickness of IMC Shear tensile strength or Reference Aluminium Steel (um) stress Al 99.5 DC01 8-16 - Fan 2011 AA6016-T4 DC04 5-40 100-150 N/mm Sierra 2008

AA6016-T4 DC04 50 50-250 N/mm Sierra 2007

AA6111-T4 JSC270CC 10 71-128 Mpa Shi 2010

AA6016-T4 DC04 2-25 100-200 N/mm Peyre 2007 Torkamany AA5754 ST14 PIC 5-20% * 50-300 MPa 2010 AA5082- XF350 5-20 20-30 kN Meco 2015 H22 Windmann AlSi3Mn 22MnB5 30 74±21 MPa 2015

Some methods have been applied to try to reduce the growth rate of the IMC layer, such as adding a coating on the steel surface (Chen 2011), using filler materials (Windmann 2015), and optimizing the welding parameters (Shi 2010, Sierra 2008). However, this can cause other problems, including increasing the cost and the weight of joints. Therefore, overall it can be concluded it is difficult to weld aluminum alloy to steel by laser welding.

Arc welding

Another fusion welding method is arc welding, which exploits an electric arc as the heat source. In an arc welding, the heat is converted from the kinetic energy generated from the impact between positive ions or thermally emitted electrons and the work-pieces (Messler 2004). Arc welding is cheaper and more flexible than laser welding, due to the less expensive heat source (Sierra 2008, Dong 2012, Su 2014).Two types of electrodes can be used: a consumable rod or wire electrode that not only conducts current, but also melts and supplies filler material to the joint, as in metal inert gas welding (MIG); or a non-consumable electrode that simply conducts current to the weld area, as in tungsten

Chapter 2 Literature Review inert gas welding (TIG) (Mallick 2010) or gas tungsten arc welding (GTAW) (Sierra 2008). Both techniques, together with various derivations of them, are used widely in industry to weld similar metals like steel and aluminium.

However, dissimilar welding between steel and aluminium alloys by arc welding is difficult, and interfacial reaction again is the main problem. One example of a joint made by GTAW is shown in Fig. 2.6. In this work the TIG torch was located on the steel side of the joint, and aluminium melting was induced by heat diffusion through the interface from the steel side. For this weld an IMC layer with thickness of 13 μm was found at the interface, which is fairly typical for arc welding process. Besides changing the welding parameters to reduce the heat input, the main method for reducing interface reaction is to add a filler material between the steel and aluminium alloy, as some alloys, like Si, can reduce the reaction, and the Zn coating on galvanized steel is also supposed to reduce the formation of the IMC layer (Su 2014, Dong 2012). However, overall arc welding has not been found to be very successful for making dissimilar joints between steel and aluminium owing to the high thickness of the IMC layer produced.

Fig. 2.6 Examples of a (a) Cross section and (b) interface of steel-Al joints made by GTAW (Sierra 2008). The fusion zones (FZ) and heat affected zones (HAZ) are also indicated.

Chapter 2 Literature Review

Consequently, due to the high growth rate of the IMC reaction layer or other problems induced by controlling the growth of the IMC layer, fusion welding is not a suitable approach for making sound dissimilar joints between steel and aluminium alloys.

2.3.2 Solid State Welding

As there are so many drawbacks making dissimilar welds between steel and aluminium by fusion welding, solid state welding, which uses temperatures lower than the melting point of the weld members, has been developed to overcome these disadvantages (Barnes 2000 (a), Mallick 2010, Tanaka 2009). Among the different types of solid state welding techniques, friction stir welding (FSW), friction stir spot welding (FSSW) and ultrasonic spot welding (USW) will be discussed as they were investigated in the present study, and these techniques have the most potential for application in automotive industry (Lee 2006, Tanaka 2009, Prangnell 2011, Chen 2012), while diffusion bonding has been applied as a support method for studying the growth kinetics of the IMC layer.

Friction Stir Welding

Friction stir welding (FSW) is a solid-state welding technique that was invented at The Welding Institute (TWI) of UK in 1991. It has initially been applied mainly to aluminium alloys. A schematic diagram of the FSW process is shown in Fig. 2.7 (Mishra 2005). In FSW, a non-consumable rotating tool is inserted into the abutting edges of two work-pieces to be jointed and traversed along the join line. The tool is comprised of a pin and shoulder and designed according to the specific application. Heat generated from the plastic deformation induced in the work pieces by the tool softens the material around the pin. A combination of tool rotation and translation results in severe plastic deformation of the work-pieces, as the material is moved from the front to the back of the pin and refills the pin cavity forming a joint (Mishra 2005). As heating is generated primarily by plastic work, the temperature in the welding process is self-regulating and is typically lower than 0.8 the melting point of the work-pieces. For aluminium, peak temperatures usually range from 450 to 480oC (Threadgill 2009).

Chapter 2 Literature Review

Fig. 2.7 Schematic diagram of the friction stir welding process (Mishra 2005).

In general, the microstructures of welds produced by FSW can be divided into four zones: the unaffected parent material, a heat affected zone (HAZ), and the thermo-mechanically affected zone (TMAZ). The weld Nugget lies within the TMAZ and originates from the large strain induced by the action of the pin during the welding process. A typical macrograph of the cross section of a weld produced in an aluminium 7075-T651 alloy by FSW is shown in Fig. 2.8. The microstructures of the weld zones are affected by the different levels of strain. The nugget is completely recrystallized with fine, equiaxed grains, as a result of intense strain; the TMAZ is partly recrystallized, due to lower strain; while the HAZ has the similar grain structure to the parent material, as there is almost no strain (Uzun 2005).

Fig. 2.8 A typical macrograph showing the various microstructure zones found in aluminium welds produced by FSSW (Mishra 2005).

Chapter 2 Literature Review

The quality of welds produced by FSW is critically depended on the processing parameters, including rotation speed, translation speed, plunge depth, tool angle and tool design. For example, if the plunge depth is too shallow, the shoulder of tool does not contact the work-piece surface, resulting in generation of welds with inner channels due to less movement of the stirred material from the front to the back of the pin. However, if the plunge depth is too deep, the shoulder of tool plunges into the work-piece, and a concave weld is produced with local thinning of the work-pieces (Mishra 2005). Therefore, suitable parameters should be chosen to produce a sound joint by FSW.

By controlling these parameters, defect-free welds have been achieved in previous studies by researchers, such as Leonard et al. (2003), Thomas et al. (2003) and Reynolds et al. (2005). The key benefits of friction stir welding have been summarised by Mishra and Ma (2005), and include their metallurgical, environmental and energetic advantages which indicate FSW is an excellent welding method for metal welding, and especially for aluminium alloys.

Unlike when welding similar materials, when welding dissimilar welds the tool is shifted towards the work-piece with lower hardness, which is fixed on the retreating side. In steel to aluminium systems this is the aluminium side of the joint (Tanaka 2009). A schematic illustration for welding steel to aluminium alloy is shown in Fig. 2.9. In this case, the stirring action of the pin takes place mainly in the aluminium, to prevent the overheating of aluminium alloy and reduce pin wear (Uzun 2005).

Fig. 2.9 Schematic illustration of the friction stir welding process applied to a dissimilar metal joints between stainless steel and an Al alloy (a) left view, and (b) top view (Lee 2006).

Chapter 2 Literature Review

As a result of asymmetry of the work-pieces, seven distinct regions are exhibited in dissimilar joints, as both materials have their own parent material, HAZ and TMAZ, as shown in Fig. 2.10 (a) - (g). Although there is only one Nugget in a dissimilar joint, it is also different from that seen in a similar joint; in that a mixture of aluminium alloy and steel particles is found in the nugget, resulting in a composite structure of steel particles in an aluminium alloy matrix (Uzun 2005, Tanaka 2009). Furthermore, due to the high level of deformation, micro-cracking can be generated by the irregular and inhomogeneous distribution of the steel particles. In the aluminium side, elongated Al grains with a rotation of up to 90o are found in TMAZ, and a similar grain structure is found in the HAZ and base aluminium alloy, while coarse precipitates are only found in the HAZ. On the steel side, a serrated appearance is observed at the interface between nugget and TMAZ, due to high deformation caused by stirring action of the threaded tool pin, which also results in elongated Fe grains. The steel HAZ had similar coarse austenitic grains but different precipitates to the base steel.

Fig. 2.10 Macroscopic overview of the cross-section of a FSW dissimilar joint between an AA6013 alloy and an X5CrNi18-10 stainless steel (Uzun 2005).

In FSW the intermetallic compound layer is affected by the welding parameters, and is not easily observed in some cases in the as-welded state due to the low welding temperature. For example, Tanaka et al. (2009) pointed out that with the increasing rotation speed the IMC layer became thicker, from 0.11 μm at 600 rpm to 0.34 μm at 800 rpm because of the higher energy input into the weld. The interface areas seen in this work are shown in Fig. 2.11. In their research, the welds were defective when the rotation

Chapter 2 Literature Review speed was either too slow (400 rpm) or too fast (>1000 rpm). With too slow a rotation speed, there was insufficient heat generated to form a joint, and with too fast a rotation speed, cracks formed due to thermal contraction after welding, resulting in joint failure. Therefore, this work proves that suitable parameters should be chosen carefully to make a sound joint.

Fig. 2.11 Interface microstructures seen in dissimilar joints in an AA7075 – mild steel by FSW with different rotation speed (a) 600 rpm, (b) 700 rpm, (c) 800 rpm (Tanaka 2009).

In another study by Springer et al. (2011 (c)) on commercially pure aluminium of 99.5 wt.% purity (Al99.5) to unalloyed low carbon steel (DC01), as little IMC reaction was observed at the interface in the as-welded state by TEM analysis (Fig. 2.12 (a) and (b)), they subsequently applied an annealing treatment to the dissimilar joints to produce different thicknesses of IMC layer. After annealing at 500oC for 9 min, a continuous IMC layer of about 8 μm thick but with a variable thickness was observed at the interface between the aluminium alloy and steel. Two intermetallic phases could be identified: a thin layer of θ

(FeAl3) next to the aluminium alloy, and a thick layer of η (Fe2Al5) next to the steel (Fig. 2.12 (c), (d)). This Al-θ-η-steel sequence was observed in all dissimilar joints after annealing, and has also been found in joints made by liquid state welding (Qiu 2009, Bouche 1998). Therefore, both η and θ phases are considered to be the most common intermetallic phases observed in Fe-Al welds produced by both solid and liquid state welding techniques.

Chapter 2 Literature Review

The mechanical properties of dissimilar metal joints have been linked to the thickness of the IMC layer produced by the welding process by many researchers (e.g. Tanaka et al. (2009) and Springer et al. (2011 (c))), and their results are shown in Fig. 2.13 with the cross weld tensile strength plotted against the IMC thickness. The influence of IMC layer on tensile property has two aspects. The first one is due to Kirkendall pores left in the aluminium alloy, resulting in the failure being located between the IMC layer and the aluminium alloy. In this case, research suggest that before the thickness of IMC layer reaches about 7 μm, the tensile strength was controlled by softening of the aluminium alloy as not enough Kirkendall pores had formed. However, when a higher number of Kirkendall pores formed between the aluminium alloy and the IMC layer, as a result of the

Chapter 2 Literature Review faster diffusion rate of aluminium compared to iron, this started to govern the tensile strength of the joints.

The second effect is the brittleness of the IMC layer, which results in the failure being located inside the η phase as this is the most brittle component (Ryabov 1985). This situation mainly happens in some cases where Kirkendall effect produced a fewer number of pores in the aluminium alloy, resulting in less influence on the tensile properties. As a result, the IMC layer became the unique factor controlling the tensile strength, and the strength of the joints dropped and finally remained at about 30 MPa with growth of IMC layer, which was close to 25 MPa reported as the tensile strength of η phase (Ryabov 1985). Consequently, despite the different mechanisms, a common issue is that a thick IMC layer is detrimental to the performance of dissimilar Fe-Al joints, even when the thickness is lower than 10 μm, the critical value given by Achar et al. (1980).

If the stir process is mainly applied to the steel weld member in FSW, overheating is produced and transmitted to the aluminium alloy. If this is sufficient, localization liquation can be formed on the aluminium side at the interface. Rest et al (2014) developed this friction melt welding method with a pin-less tool for making over lap joints between an ultralow-carbon steel and an aluminium AA2024 T3 alloy. A schematic of the welding

Chapter 2 Literature Review process together with the typical interface microstructure is shown in Fig. 2.14. Again, both θ and η phases were observed at the interface with an uneven thickness, which might be due to the short reaction time and low temperature. In this work the rotation speed was claimed to control the microstructure of the interface and strength of the joints, but the influence of the IMC layer on joints' strength was still not clear. Although benefits of this methods, including absence of a keyhole and higher efficiency than traditional FSW, were claimed by Rest et al., it cannot be proposed as a successful method for making dissimilar welds before the link between microstructure and mechanical properties is fully understand.

Consequently, due to the low welding temperature in principle, IMC free dissimilar joints between aluminium alloys and steel can be produced with good mechanical property by friction stir welding.

Friction stir spot welding

Friction stir spot welding is a point joining technique performed in a similar manor to friction stir welding, but without traversing the tool. The whole process involves plunging a rotating tool into the workpieces, holding it there for a certain dwell time and then

Chapter 2 Literature Review retracting. An illustration of FSSW is shown in Fig. 2.15. The rotating tool with a probe/pin penetrates the top sheet (A), and the friction generates both deformation and heat in the metals adjacent to the tool (B), resulting in metal flow near the interface. After the tool is removed, the metals have been joined with a metallurgical bond in the weld region (C) (Buffa 2008, Badarinarayan 2007). A typical cross section for a dissimilar joint between aluminium and steel made by FSSW is shown in Fig. 2.16.

Fig. 2.15 Illustration of friction stir spot welding process (Buffa 2008).

Fig. 2.16 Cross section of dissimilar joint between AA6016 and IF-steel made by FSSW (Bozzia 2010).

In FSSW, several parameters, such as the plunge depth, plunge rate, rotational speed, dwell time and geometry, will affect the weld quality. Many researchers have studied their effect on weld quality, for example, Mitlin et al. (2006) have reported the effect of plunge depth on the remain material beneath the weld area, and Su et al. (2005) found the influence of rotational speed on the toughness and tensile properties of Al-steel joints. Similar to other welding techniques, suitable parameters should be chosen to make a sound dissimilar joint between aluminium alloy and steel.

Chapter 2 Literature Review

For example, welding conditions like rotation speed and plunge depth have been found to affect the weld strength by controlling the IMC layer thickness by Bozzi et al. (2010). The EPMA analysis at the interface between AA6016 and IF free steel is shown in Fig. 2.17. It is obviously that the IMC layer grows with the increasing rotational speed and the penetration depth. By combining the IMC thickness and weld strength, they claimed that an IMC layer was necessary to improve the dissimilar weld strength, but cracks initiated easily in the brittle IMC layer when the IMC layer was too thick. In their research, an optimal IMC layer thickness of 8 μm was measured for a rotational speed of 3000 rpm and a tool penetration depth of 2.9 mm.

However, when using a conventional tool in FSSW, the small welding area in dissimilar joints limits the application of FSSW, as shown in Fig. 2.16. Therefore, three new methods have been developed to enlarge weld area.

The first one is to use a pin-less tool to produce dissimilar joints between aluminium and steel with a larger weld area. A typical cross section of a dissimilar joint between AA5052 and ductile steel using this method is shown in Fig. 2.18 (Miyagawa 2009). Apparently,

Chapter 2 Literature Review compared with the cross section in Fig. 2.16, weld area highlighted inside the white dashed box is much larger. However, it takes a long weld time to form a joint as there is no frictional sliding between the two work-pieces in pin-less FSSW. Therefore, it is not suitable for the industry application and also leads to a thick IMC layer developing (Chen 2011).

Fig. 2.18 Cross section of a typical dissimilar joint between AA5052 and ductile steel made by FSSW without a pin (Miyagawa 2009). Weld area is highlighted inside the white dashed box.

The second method is to provide a tool path after the completion of the sinking phase (Fig. 2.19 and 2.20); i.e. the tool is moved along a trajectory instead of left in its position. Two examples of this approach are shown in Fig. 2.19 (a) and Fig. 2.20 (a). Once it has completed the assigned trajectory, the tool leaves the sheets to be welded. Buffa et al. (2008) have tried this technique to make high quality friction spot welds. In Fig. 2.19 (b) and Fig. 2.20 (c), the tool paths left on the joint's surface are obvious; compared with joints made by traditional process, the stirred zone is much larger, resulting a much better mechanical properties (Fig. 2.19 (c) and (d)). Chen et al. (2012) have also applied this method on dissimilar welding between steel and aluminium alloys using a circular tool trajectory to produce IMC free interface joints with a strong metallurgical bond between the base metals. Therefore, this is a good choice for dissimilar welding.

Chapter 2 Literature Review

Fig. 2.19 (a) Sketch of the assigned tool axis path used by Buffa (2008) to incease he weld area by FSSW; (b) Top view and (c) etched AA section of a FSSW joint made by a modified process with a tool path compared to (c) the etched section of a joint made by the traditional process (Buffa 2008).

Chapter 2 Literature Review

The third approach is refill friction stir spot welding, developed by GKSS for performing FSSWs without leaving a hole when the tool is extracted. According to Rosendo et al. (2011), this technique can be divided into two variants depending on which part of the tool is the plunging element, the pin or sleeve. An illustration of the main stages for sleeve plunging is shown in Fig. 2.21, and the procedure is explained as follows. In stage one, the sheets are clamped together and both the pin and sleeve start to rotate producing frictional heat on the upper sheet surface (Fig. 2.21 a). In stage two, the sleeve is plunged into the sheets while the pin moves upwards to create a cylindrical cavity to accommodate the softened metal displaced by the sleeve (Fig. 2.21 b). After a predetermined plunge depth is reached, the process is reversed, and the sleeve retracts back to the surface of the upper sheet. At the same time the pin pushes the softened metal back into the hole left by the sleeve, refilling it completely (stage three, Fig. 2.21 c). Finally, the welding head is removed from the work-pieces, resulting in a flat surface joint with minimum material loss (Rosendo 2011). The procedure for a pin plunge is similar, with the pin penetrating the metals, while the sleeve retracts. Compared with pin plunge variation, the sleeve plunge variant makes a larger welding area due to the larger cross section of the sleeve than the pin, with a higher plunge force.

Fig. 2.21 Illustration of the four main stages used in the FSSW-Refill process (sleeve plunge variant) (Tier 2013).

Chapter 2 Literature Review

As refill FSSW is a newly developed process, most research is focused on aluminium alloy welding (Tier 2013, Rosendo 2011). However, Fukada et al. (2013) has used this method to make dissimilar joints between AA6064 and hot dip zinc-coated steel. High quality weld could be made with little IMC reaction layer seen at the interface and better mechanical properties than found with RSW.

Consequently, with several improved new variants, FSSW has become an excellent method for making sound dissimilar joints between steel and aluminium alloys with an IMC free interface and high performance mechanical properties.

Ultrasonic spot welding

Ultrasonic spot welding is another effective solid state welding technique for dissimilar welding between steel and aluminium alloy (Prangnell 2011, Chen 2012 (b), Balasundaram 2014). Although ultrasonic metal welding was first introduced in early 1950s for wire and thin foils bonding with high power (1.5 - 2.5 kW), it has only recently become an effective technique for joining thicker gauges sheets (Hetrick 2009, Chen 2012 (b)). A schematic diagram of the ultrasonic welder used in this study is shown in Fig. 2.22, which used the special tips for the aluminium and steel. Images for the interface development are shown in Fig. 2.23.

Three main sub systems are included in a high power ultrasonic spot welding machine (schematic diagram in Fig. 2.22). The first is the power supply and control system, including an electric signal generator and an amplifier which work together to provide electrical energy, a time measurement circuit controlling the welding time and energy input, and an automatic power adjustment to maintain constant power in the welding process. The second is an energy transformation system, including an electromechanical transducer that converts the electrical power into mechanical vibration and a waveguide to amplify the vibration. The last component is the mechanical system, including a frame where the other components are mounted, an air cylinder to provide the clamping force on the work-pieces, and the two welding sonotrode tips used to fix and conduct the vibration to the work-pieces.

Chapter 2 Literature Review

Fig. 2.22 (a) schematic diagram of a dual-reed USW machine, (b) the sonotrode tips used for the aluminium and steel sheets and (c) a schematic diagram of the weld sample (Prangnell 2011).

Mechanism of USW has been discussed by previous researchers, and includes metallurgical adhesion produced by plastic deformation, diffusion at the weld interface, interfacial chemical reaction, the formation of micro bonds and mechanical interlocking (Jahn 2007, Bakavos 2010, Shakil 2014). The evolution of the interface in similar welding process is summarised by Bakavos and Prangnell (2011), and is shown in Fig. 2.23. After welding starting, the work pieces are fixed by a normal load from a constant clamping pressure. Driven by the vibration generated by the signal generator, the two sonotrode tips oscillate out of phase, introducing a small but high frequency linear displacement across the weld interface. Micro welds are formed firstly at the asperities on the work pieces surfaces which are the early contact points, as shown in Fig. 2.23 (a) (Jahn 2007, Prangnell 2011). At the same time, heat is generated by the shear deformation initially by friction sliding firstly, but rapidly followed by the plastic deformation associated with local galling and micro weld formation and these then spread across the interface (Murdeshwar 1997). Welding temperature rise rapidly because more energy is produced by the increasing levels of plastic deformation, and the micro weld formation spreads in both density and size across the weld interface with longer welding times (Prangnell

Chapter 2 Literature Review

2011). Finally, the base metals are joined together when a continuous weld region is formed as a result of the micro bond spreading and merging across the entire interface. However, welding is rarely fully complete and local segment of un-bonded interface can be observed (Prangnell 2011). During the process, the weld parameters will affect the joint performance by controlling the microstructure and weld area, such as interface sliding which is caused by clamping force, and swirls or folds which are caused by severe plastic deformation in long welding times.

Chapter 2 Literature Review

In USW the energy is predominantly generated at the weld-line instead of at the top surface, as is the case in FSSW. Therefore the process is more efficient and a shorter welding time for a sound joint is possible (Bakavos 2010, Elangovan 2009). Furthermore, another notable advantage of USW for welding aluminium alloy is that the high frequency shear vibrations (typically 20 kHz) can break and disperse the oxide layer (Matsuoka 2009, Annoni 2011, Shakil 2014). Therefore, it is worth developing USW for welding aluminium and steel in dissimilar joining.

USW has been used to make joints between dissimilar metals like magnesium and steel (Patel 2013), aluminium and copper (Balasundaram 2014), magnesium and aluminium (Panteli 2012), aluminium and titanium (Zhang 2014), and steel and aluminium (Haddadi 2012, Shakil 2014). In most of these studies the microstructure was observed, and the mechanical properties were measured; however, further study is still necessary to understand the IMC layer development of high resolution.

The formation process of a sound joint in dissimilar metals welding also involves the formation and spread of micro welds at the interface, but there are two different features in a dissimilar welding process compared with similar welding. The first one is the deformation of the base metals. A schematic illustration of aluminium deformation in both similar and dissimilar USW joints is shown in Fig. 2.24 (a) and (b), respectively. Due to the much higher yield stress than aluminium, little deformation occurs in the steel sheet near the interface, and the oxide on the steel sheet surface is dispersed purely by abrasion from the aluminium sheet instead of by the more complex processes seen in Al- Al similar welding (Prangnell 2011). As a result, longer welding times are necessary for aluminium-steel dissimilar welding and a more flat interface is formed. The grain structure of the aluminium is also quite different from that in similar welding, with the EBSD orientation maps (Euler contrast) for the welding cross section for both joints shown in Fig. 2.25.

Chapter 2 Literature Review

Three distinct areas have been observed in Al-Al similar welds, a forge zone beneath the sonotrode tips, a severely deformed region with fine grains at the interface line, and an intermediate region with shear bands connecting from the interface to the teeth impressions (Bakavos 2010). In contrast in Fe-Al dissimilar welds, although a forge zone is also found under the tips in the aluminium side at short welding times, recrystallisation occurs at long welding time, resulting in the more uniform and equiaxed grain structure in the aluminium side. Furthermore, a thin band of ultrafine aluminium grains was also found near the interface with short welding times, and the grain size increased rapidly with welding time increasing, which is totally different from the interface grain structure seen in similar Al-Al joints. At the steel side, much less deformation is induced in the steel grains than the aluminium, except at the surface under the tips and very near the interface, especially with long welding times (Haddadi 2012).

Chapter 2 Literature Review

The second different feature found between similar and dissimilar welding processes is the formation of intermetallic compounds at the interface (Haddadi 2012). The development of an IMC layer with increasing welding time is shown in Fig. 2.26. In the early stage, discontinuously IMC islands are seen distributed at asperities where micro bands form firstly along the interface. With increased welding time, both the size and density of contact points are increased, resulting in bigger contact interface area, and therefore a more continuous and thicker IMC layer forms at the interface. Under optimised welding conditions, a continuous and uniform IMC layer with approximate 1 μm is formed at the interface. Consequently, the IMC layer forms firstly at the micro- bands position, and grows in both size and density together with the spread of micro- bands until a continuous layer is formed. After the layer becomes continuous, thickening is controlled by volume diffusion in the following welding time (Haddadi 2012).

Chapter 2 Literature Review

Haddadi (2012) also researched the influence of the factors that affected the dissimilar joint properties, and the optimum welding conditions required to make sound dissimilar joints between aluminium and un-coated steel, including the effect of welding time, impedance, welding direction, single/double reeds, and clamping pressure. The results are shown in Table 2.2.

Table 2.2 Optimised welding conditions for dissimilar welding between aluminium and un coated steel (Haddadi 2012).

welding single/double clamping impedance welding direction time reeds pressure the vibration is 1.0 s 2 Ω perpendicular to the lap double reed 1.4 kN edge

Chapter 2 Literature Review

In particular, a shorter welding time, higher impedance value, different welding direction, single reed and higher clamping pressure resulted in the bonding area to be insufficient for producing a sound joint, while longer welding times induced too thick an IMC layer, which was brittle and harmful for the joint properties.

Besides the welding parameters, the effect of two types of zinc coating on the welding area and joint properties were investigated by Haddadi (2012). Zinc coatings were found to affect the joint's performance, such as reducing weld temperatures. Furthermore, with hot-dipped zinc coated steel, expulsion of liquid zinc at the interface was observed, resulting in a larger bonding area than that in joint with un-coated steel. However, in the case of a galv-annealed zinc coating steel, no liquid zinc was formed, but cracking was found in the coating during the welding process, resulting in poor joint properties. Therefore, a suitable coating type should be chosen to make a sound dissimilar joint.

With increasing welding time thickening IMC layer was found to change the failure mode from pull out to interface failure in aluminium to un-coated steel joint. In the case of long welding times (3 s in the study), the fracture path was seen to fully revert back to the weld interface, indicating poor bonding between aluminium and IMC layer (Haddadi 2012). Therefore, the IMC growth behaviour was also studied by Haddadi (2012). Due to having the lowest activation energy, θ (FeAl3) phase was found to form first, while with long weld times η (Fe2Al5) phase became the largest fraction of the IMC layer with a higher kinetic growth rate. This Al-θ-η-Fe sequence is the same as that seen in joints made by liquid sate welding (Qiu 2009) and solid state diffusion (Springer 2009 (a)). However, further study is necessary to obtain more details about phase identification and microstructure of the IMC layer formed during welding. Furthermore, predictions for the thickness of IMC layer in the welding process mad using static annealing data showed a lower value than actual measurements, which is the similar trend with Panteli's result (2012), indicating the welding process accelerated diffusion possible owing to the high vacancy concentration, dynamic generation of dislocations, fine grains and substructures developed under the intense dynamic plastic deformation that occurs in ultrasonic spot welding.

Chapter 2 Literature Review

Overall ultrasonic spot welding has been found to be a viable welding technique for making a sound dissimilar joint between aluminium and steel. However, on the basis of the previous studies mentioned above, more detail is required to further understand the IMC growth behaviour during the welding process. For example, as the IMC layer is too thin to be observed in the as-welded state, the microstructure and composition of the IMC layer are still not clear, as well as its growth behaviour during the welding process (Haddadi 2012, Chen 2012, Springer 2011 (b)). Therefore, it is necessary to obtain higher resolution information of the IMC layer growth behaviour in solid state welding, which was one of the main purposes of this study.

2.4 Base Material

The microstructure and properties of the materials used in this study, including the aluminium and steel alloys, are introduced in this section.

2.4.1 Aluminium Alloy

Due to their light weight and good specific properties, the application of aluminium alloys is increasing in the automotive industry. The total amount of aluminium used in a European car rose from 43 kg in 1980 to 157 kg in 2010. It should be noticed that the strongest growth of aluminium is in car body sheet applications, as shown in Fig. 2.27 (Hirsch 2011). By using aluminium alloys in cars, dramatic weight saving can be obtained without losing too much performance. It has been shown that about a 35% of total weight saving for a body-in-white (BIW) is possible with a multi-material superlight-car (SLC) concept (Hirsch 2011). Among the different aluminium alloy series, medium strength 6xxx and high strength 7xxx alloys are being considered as suitable alloys to be used in complex and functional integration in car design, as they are heat-treatable and can be tailored to be formable and sufficiently strong for specific applications.

Chapter 2 Literature Review

(a) (b)

Fig. 2.27 Penetration of Aluminium in European Cars; (a) aluminium alloy weight in cars vs. year; (b) growth of aluminium in car from 2000 to 2005 (Hirsch 2011).

In 6xxx series alloys, magnesium (Mg) and silicon (Si) are the main alloy elements. Mg2Si is considered to be the main equilibrium precipitate in these alloys, which can be formed by optimising the annealing conditions. Artificial aging is always applied to obtain the optimum microstructure, resulting in good mechanical properties (Esmaeili 2003). In general, microstructure evolution in 6xxx series alloys during heat treatment process follows the sequence (Wang 2003):

α (sss) →Mg + Si clusters → Mg-Si co-clusters → GP zones → β’’ + Q’→ β + Q

Where α (sss) is the supersaturated solid solution formed after solution treatment, and GP zones indicates Guinier–Preston zones. Q and β are the equilibrium quaternary

(Al4Cu2Mg8Si7) and binary (Mg2Si) phases, respectively. The β’’ precipitate has a monoclinic crystal structure with a Mg: Si ratio of close to one, with a fine needle-shape aligning along the soft matrix <100> Al directions. Q′ is a semi-coherent form of the equilibrium Q phase, with the same hexagonal crystal structure and an identical composition (Chen 2012 (b)). However, β’ phase is not formed in the aluminium 6111 alloy as a result of the high copper content (Chakrabarti 2004). The microstructure evolution is controlled by the annealing treatment parameters, including the dislocation density and grain structure. Therefore, the mechanical properties can be optimised according to the microstructure.

Chapter 2 Literature Review

Compared with other 6xxx series alloys, it has been reported that AA6111 can obtain better mechanical properties; in particular, the yield stress and strength can reach approximately 350 MPa and 400 MPa, respectively, which are 40% above those of AA6016 (Esmaeili 2003). Therefore, AA6111 is widely used in the automotive industry, and was adopted in the in present study.

As another well-known precipitation hardening aluminium alloy, 7xxx series alloys are also important structured materials due to their high specific strength, toughness, and fatigue resistance, and have been used widely in aviation and aerospace industries (Feng 2014). The precipitation sequences in microstructure evolution in 7xxx series which are most likely to occur during aging are shown below (Liu 2015):

1: α (sss) → GPI zones → Metastable η' phase→ Stable η (MgZn2) phase

2: α (sss) → VRCs→ GPII zones → Metastable η' phase → Stable η (MgZn2) phase

It should be noticed that two types GP zones may be formed in the process. GPI zones are linked with solute-rich clusters and fully coherent zones with internal order described by AuCu(I)-type structure (Berg 2001). As a result, they can be formed over a wide temperature range. GPII zones are likely related to vacancy-rich clusters (VRCs), and thus highly dependent on the vacancy concentration (Sha 2004). Although there may be a difference in the aging process, both GP zones will change to the meta-stable η' phase, which plays a vital role on the maximum hardening effect (Liu 2015).

It should be noted that all the aluminium alloys are heat treatable, indicating the mechanical properties and microstructure will be affected by the welding process and natural aging (Chen 2012 (b)). Therefore, it is important to consider the natural aging time when studying weld sample.

2.4.2 Low Carbon Automotive Steel

Steel is used widely in all aspects of our life, due to its many advantages, such as the low price and easy formability (Soyarslan 2011). Among thousands of types of steel, low carbon strip steels are widely applied in the automotive industry due to their excellent

Chapter 2 Literature Review mechanical properties and easy formability. Of these, a DC04 un-coated formable automotive steel has been selected in present study.

As rolling and heat treatment have been applied to the DC04 steel, the microstructure is affected by the temperature and deformation history. Firstly, hot-rolling is used to produce steel sheets about 2 mm thickness, with a rolling temperature usually above 900oC. Then further cold-rolling is applied to make the find gauge sheet as thin as 0.2 mm. In the cooling process, transformation from γ (fcc) to α (bcc) occurs leading to carbon enrichment of the untransformed austenite. Finally, after the temperature drops to the eutectoid point (~723oC), the remaining austenite transforms to ferrite and cementite

(Fe3C) (Llewellyn 1998).

2.5 Intermetallic Compounds

The Fe-Al equilibrium phase diagram is shown in Fig. 2.28 (Massalski 1986). It is clearly seen that the solubility of aluminium in iron is very limited, resulting in a large number of intermetallic phases in the Fe-Al binary system, which are listed in Table 2.3, together with their crystal structures. It should be noted that chemical composition for each phase is very limited and the preformed phase exists in a certain range. For example, with the η phase, the content of aluminium is in the range of only 70-73 at. %, and 74.5-76.5 at. % for θ phase (Agudo 2007).

Fig. 2.28 The Fe-Al equilibrium phase diagram (Massalski 1986).

Chapter 2 Literature Review

Although there are so many intermetallic phases could form in Fe-Al binary system, η phase has been reported as the predominant phase in most welding techniques, and a thin layer of θ phase between the η phase and the aluminium may form when aluminium is in liquid state (Qiu 2009, Nishihara 2010, Springer 2011). This is explained by the low active energies for these two phases among the whole set (Naoi 2007). Furthermore, although θ phase was reported to have an even lower active energy than η phase, η phase has a higher kinetic growth rate than θ phase in Fe-Al binary system, with the values of 3.41x10-11 cm2/s and 0.01x10-11 cm2/s at 600oC, respectively (Bindumadhavan 2000, Shahverdi 2002, Naoi 2007, Springer 2011). It is believed that the crystallographic structures have dramatic influence on the growth rate of different phases (Hirose 2003). Besides, as different active energies were received by previous researchers, it is necessary to understand the mechanism for the growth of intermetallic phases, which are mainly η and θ, during the diffusion process. Therefore, previous studies about η and θ phases are reviewed in the following sections, including the crystallographic structure, microstructure, and growth kinetics.

Table 2.3 The Fe-Al intermetallic phases in the Fe-Al binary system (Springer 2011 (b), Shahverdi 2002, Lee 2003).

Phase Composition Crystallographic Structure

β'' Fe3Al Cubic β' FeAl Cubic ε Fe5Al8 Cubic ζ FeAl2 Triclinic η Fe2Al5 Orthorhombic θ * FeAl3 (Fe4Al13) Monoclinic κ Fe3Al Cubic * θ phase has different compositions due to the large range of variation in atomic coordination.

2.5.1 Crystallographic Structures of η and θ Phases

With the orthorhombic crystal system, η (Fe2Al5) phase belongs to Cmcm space group, and the parameters are a= 7.6486 Å, b= 6.4131 Å, c= 4.2165 Å (Ellner 1992). Burkhardt et al. (1994) built a refine structure of Fe2Al5 phase, which is shown in Fig. 2.29. After that, the crystal structure and crystal symmetry of Fe2Al5 phase were built up successfully in Fig. 2.30 in a later research (Hirose 2003).

Chapter 2 Literature Review

Fig. 2.29 The three - dimensional framework of Fe and Al atoms in the Fe2Al5 structure (Burkhardt 1994).

(a) (b)

Fig. 2.30 (a) Crystal structure and (b) crystal symmetry of the Fe2Al5 phase (Hirose 2003).

The partially disordered distribution of atoms is a special feature of the Fe2Al5 crystal structure. Three types of aluminium atom’s positions, Al1, Al2 and Al3, and one type of iron atom position have been classified in the framework proposed for its crystal structure. Burkhardt et al. (1994) found that the occupation factors for the Al2 and Al3 positions were 0.36 and 0.23, respectively. Therefore, thermal motion of aluminium atoms between Al2 and Al3 sites explains the different distribution of aluminium atoms in

Chapter 2 Literature Review

Fe2Al5 structure observed by different researches (Burkhardt 1994, Hirose 2003). Furthermore, it is reasonable that the partial occupation of the aluminium positions Al1 and Al3 in the Fe2Al5 structure also results in the Fe2Al5 phase having a small range of Fe/Al ratio.

It should be noted that the vacancy density (or unoccupied sites) of aluminium atoms along the [001] axis is higher than in other directions, which results in the unique morphology and fast growth rate of the Fe2Al5 phase seen in diffusion process (Naoi, 2007, Bouayad 2003, Bouche 1998). The 3-D framework is built by interactions of iron and Al1 atoms, with the interatomic distances from 2.5 Å to 2.66 Å, which are shorter than the sum of the two atomic radium values (rFe + rAl = 2.69 Å, 2rAl =2.86 Å). Some channels exist inside the framework along the [001] axis, and Al2 and Al3 atoms are placed on the axes of these channels whose positions are very close to each other. Furthermore, the distance between iron atoms and these points are 2.36 Å, even shorter than the average interatomic distances in the framework. Therefore, aluminium atoms in these positions have stronger interactions with the iron atoms than those in other sites. Consequently, both positions are occupied by Al2 and Al3 atoms, but not at the same time, resulting in the incomplete occupation of these sites (Burkhardt 1994). It is this incomplete occupation that causes the preferred growth direction of Fe2Al5 phase along the [001] axis, including the fast growth rate and columnar morphology shown in Fig. 2.31.

Fig. 2.31 (a) Cross section micrographs of the tongue-like morphology of intermetallic

Fe2Al5 layers at 700 °C. (b) Mild steel coated by hot-dipping for 180 seconds, after etching, revealing the Fe–Al/steel substrate interface morphology of the Fe2Al5 phase (Cheng 2009).

Chapter 2 Literature Review

Belonging to the space group C2/m, with a monoclinic structure, θ (FeAl3) phase has a more complicated crystal structure than η phase, with 100 atoms per unit cell. Both the 3-

D interactive structure and crystal symmetry for the unit cell of the FeAl3 phase are shown in Fig. 2.32. There are 5 and 4 types of co-ordinations for iron and aluminium atoms, respectively, with around 10 neighbour atoms for each. According to a qualitative measurement of the inter-atomic distances by Black (1955 (a)), strong and specific interactions between iron and aluminium atoms were assumed to play a vital role in the formation of the unique and complex structure of θ phase, which results in a much greater range of variation in interatomic distances and in atomic co-ordination than in the η phase.

2.5.2 The Evolution of the IMC Layer

In order to observe the growth behaviour of IMC layers between aluminium and iron, static annealing has been commonly used at different temperatures with different times (Springer 2011 (b), Naoi 2007, Bouche 1998). In most case, due to the limitations of the annealing temperature and time, only the η (Fe2Al5) and θ (FeAl3) phases have been observed at the interface. Therefore, the evolution of the IMC layer thickness mainly involves the growth of η phase, and the growth kinetics of the IMC layer calculated with this data will be discussed in next section.

For interface layers in dissimilar couples between aluminium and steel, the η phase is always the major part of the IMC layer, while the θ phase is only a minor part (Shahverdi 75

Chapter 2 Literature Review

2002, Luo 2012) or is even not detected (Naoi 2007, Springer 2011 (b)). As a typical example, the development of η phase is shown in Fig. 2.33 from the work of Springer et al. (2011 (b)). It is clear that mainly the η phase is identified in the IMC layer; although κ and β' phases are also identified, but have similar crystal structures to aluminium and steel, respectively, they are not common and only form as the very thin layers. With annealing time increasing, the uneven thin η layer grows into a thick layer with columnar grains, due to the high vacant site occupancy along the c axis of the η phase (Burkhardt 1994). It should be noticed that IMC layer is still irregular and not uniform even after 16 h.

(b)

(a)

It is important to find the reason why other Al-Fe intermetallic phases do not easily form in Al-steel couples; therefore, it is vital to observe the evolution of the IMC layer with other Fe-Al intermetallic phases. As the IMC layer is also observed at the interface between the coating and matrix in aluminium coated steels and is detrimental to the coating performance, researchers have studied IMC layer evolution in this case. Of these

Chapter 2 Literature Review studies, Cheng et al. (2008, 2009) and Windmann et al. (2013) investigated the evolution of the IMC layer between aluminium and steel, by annealing different aluminium coated steels at temperatures from 750oC to 1000oC with different times, and their results are shown in Fig. 2.34 and Fig. 2.35, respectively.

Fig. 2.34 Evolution of the IMC layer in an aluminium coated steel (a) at as-coated state, and at 750oC for (b) 8min, (c) 60 min, (d) 10 hours, (e) 48 hours and (f) 72 hours (Cheng 2008).

With pure aluminium (Cheng 2008), sequence of IMC formation of Al-θ (FeAl3)-η (Fe2Al5)- Fe was found at the interface in the initial as-coated steel, and the two IMCs had similar thicknesses (Fig. 2.34 (a)). However, with increasing annealing time at 750oC, the θ phase remained the similar thickness, while the η phase grew into aluminium side rapidly, until it made up the entire coating layer (Fig. 2.34 (b)). Crystal orientation maps revealed that the high growth rate of η phase with columnar grains was due to the high vacancy site density of aluminium atoms along the [001] direction, mentioned in the last section. At this point, θ phase disappeared, and a new ζ (FeAl2) phase appeared, both inside η phase and at the interface between the η phase and the Fe substrate, as a result of iron diffusion (Fig. 2.34 (c)). Then the ζ phase grew into the η phase layer, and a new β' (FeAl) phase appeared between ζ phase and Fe substrate. Pores in the β' phase were induced by the Kirkendall effect, where iron atoms diffused to the ζ phase causing a net flux of

Chapter 2 Literature Review vacancies (Fig. 2.34 (d), (e)). Finally, the η phase disappeared, and the ζ phase became the major part of the coating, with a layer of the β' phase between ζ phase and Fe substrate.

Both Cheng et al. (2012) and Windmann et al. (2013) have conducted research on the evolution of the IMC layer in Al-Si coating steel (Fig. 2.35). In Cheng's study, the general evolution of the IMC layer was similar with that in pure aluminium coated steel, but with two different features. The first was the appearance of τ5(H) (Fe2Al7Si) phase at the interface between the coating and Fe substrate in the as-coated state, instead of θ and η phase, which still formed in the following annealing process. The second difference was after the η phase consumed the whole coating, only the β' (FeAl) phase appeared at the edge of the coating, and ζ (FeAl2) phase did not appear in the coating until the end of the evolution process, which was 216 hours in their study. Chromium in the iron substrate and silicon in the coating were found to form complicated phases in the initial stage, but then diffuse into the Fe-Al intermetallic phases that formed later, mainly in the ζ phase and the β' phase, due to the high solubility of Cr and Si in them. Similar results were obtained by Windmann et al. (2013), but the ζ phase did not form, and the coating layer was consumed by the β' phase, with η (Fe2Al5) phase left at the edge.

Fig. 2.35 Evolution of the IMC layer in an aluminium coated steel by Windmann et al. (2013).

Although four different phases can form during annealing in aluminium coated steels, there are still only up to two phases which have been identified in dissimilar joints in

Chapter 2 Literature Review either the as-welded state (Qiu 2009 (a), Springer 2011 (b)) or following annealing state (Naoi 2007, Bouche 1998). One possible reason is that in coating system, the content of aluminium is limited and easily consumed by iron atoms, resulting in the formation of a wider range of Fe-rich intermetallic phases, while in the dissimilar joints made by welding, the content of aluminium is not restricted, which is a more similar situation to the initial stage in annealing aluminium coating steel before η phase consumes the entire coating. Consequently, it is reasonable that only η and θ phases are observed in dissimilar joints between aluminium and steel.

2.5.3 Microstructure of the IMC Layers

Although only η (Fe2Al5) and θ (FeAl3) phases were observed in dissimilar joints, the morphology of IMC layer could change with the welding and annealing conditions (Bouche 1998, Naoi 2007, Qiu 2009 (a), Springer 2011 (b), Luo 2012). The typical interface morphologies seen for dissimilar static annealed couples produced with solid/solid and solid/liquid states are shown in Fig. 2.36 (a) and (b), respectively. In the solid/solid state, a gray wavy IMC layer develops, was identified as η phase at the Fe-Al interface. The IMC- Fe interface was more flat than the IMC-Al interface, and crack easily formed at the IMC- Al interface by mechanical polishing of the cross-section (Naoi 2007). In solid/liquid state, where aluminium was liquid during diffusion, the IMC-Al interface was much more flat than the IMC-Fe interface, due to the tongue-like morphology of the η phase inside the iron matrix (Bouche 1998, Tanaka 2009, Cheng 2011). In this case, the θ phase was identified next to the aluminium, with a slightly wavy interface between them. In comparison, to solid/solid couples, the interfacial reaction between iron and liquid aluminium was accelerated, resulting in a much faster growth rate of the IMC layer. Furthermore, growth of η phase into the iron has a preferred orientation, resulting in the formation of a non-uniform, tongue-like shaped interface between the η phase and iron substrate.

Chapter 2 Literature Review

In order to investigate the grain structure in the IMC layer, high resolution techniques like TEM and EBSD have been used to observe the microstructure (Qiu 2009 (a) (b), Cheng 2012, Sun 2013, Springer 2011 (b)). Example images taken by EBSD with inverse pole figure maps are shown in Fig. 2.37. It should be noted that Fig. 2.37 (a) and (b) show the IMC layer at 600oC for 1 and 16 h, respectively, while (c) shows the IMC layer at 675oC for 30s, which is above the melting point of aluminium. The θ phase again only appeared in the solid/liquid state. According to the inverse pole figure maps, although the η phase had a more random level of orientation at lower temperatures and shorter annealing time, it still has a preferred orientation along the [001] direction which is indicated by the green and blue colours, and this had been confirmed by other researchers (Shahverdi 2002, Naoi 2007, Cheng 2009, Tang 2012). The reason why the θ phase with a lower active energy formed later and needed higher temperature than the η phase was explained by the kinetics in the Fe-Al binary system (Shahverdi 2002, Naoi 2007) and the high growth rate of the η phase (Burkhardt 1994).

Furthermore, previous researchers have investigated the orientation relationship between the η phase and iron substrate. The unit cells of Fe and Fe2Al5 illustrating the proposed orientation relationship between them and the corresponding [100] Fe stereograph’s projection are shown in Fig. 2.38 (a) and (b), respectively (Wang 2010).

Wang et al. (2010) found the orientation relationship by TEM to be [110]η// [111]Fe,

(001)η//(01̄1)Fe, (11̄0)η//(21̄1̄)Fe. In particular, (001)η//(01̄1)Fe was found as the most

Chapter 2 Literature Review possible interface due to the low energy and good match of lattice sites. However,

Marder (2000) has illustrated the orientation relationship might also be (311)η // (110)Fe or (221)η // (110)Fe. This well-defined orientation relationship could explain the rapid nuclear of the η phase at interface of aluminium and steel.

(a) (b) (c) θ

Fig. 2.38 (a) Unit cells of Fe and Fe2Al5 illustrating the orientation relationship between them, and (b) the corresponding [100] Fe stereograph (Wang 2010).

Chapter 2 Literature Review

2.5.4 Growth Kinetics of the IMC Layer

According to the microstructure characterization, the growth of IMC layer in dissimilar combinations between aluminium and iron is mainly controlled by the growth of the η

(Fe2Al5) phase, with θ (FeAl3) phase growing very slowly between η and aluminium in liquid/solid state. Therefore, the growth kinetics of the η phase has been studied by many researchers over the past 80 years. This has confirmed the growth of the η phase is volume diffusion (i.e. lattice diffusion) controlled. A parabolic law has therefore been generally used to model the growth behaviour of the η phase, and compare the influence of alloy elements on the growth kinetics. The kinetic data from previous studies will be summarized in this section.

It is important to confirm the growth mechanism of the η phase first. Based on characterization of the interfacial reaction between aluminium and iron and principles governing diffusion controlled reactions in solid state, Richards et al. (1994) pointed out two important deductions for the growth kinetics of the η phase in the Fe-Al system:

1 As both aluminium and iron have homogeneous compact phase structures, the growth of each phase (mainly η) in IMC layer, which is actually the interfacial reaction resulting in the formation of IMC layer, is volume diffusion controlled; therefore, a parabolic law can be used to represent the reaction kinetics as follows:

x2=kt (2-1) or

x=k't 0.5 (2-2) where x is the thickness of IMC layer (μm) after reaction time t (s), while k and k' are rate constant or parabolic coefficient in different expressions, with dimensions of μm2∙s-1 and μm∙s-0.5, respectively.

2 During the reaction process, if the mechanism of the growth behaviour of the IMC layer does not change, the reaction rate should show a linear, Arrhenius type temperature dependence:

Chapter 2 Literature Review

Q k = k exp⁡(− ) (2-3) 0 RT

2 -1 where k is the reaction rate constant (μm ∙s ), k0 is the frequency factor or pre- exponential factor, Q is the reaction active energy or enthalpy (kJ/mol), R is the gas constant (J∙K-1/mol), and T is the reaction temperature (K).

With the equations (2-1) and (2-3), the growth behaviour of the η phase in Fe-Al system can be quantitatively analyzed, and the active energy for η phase can be calculated in one specific system. It should be noted that rate constant k' in equation (2-2) is different from k in equation (2-1), and can't be used directly in equation (2-3) until it is changed to k.

Some typical results from the equations received by previous researchers are shown in Fig. 2.39. In Fig. 2.39 (a) and (b), thicknesses of IMC layer are plotted against the square root of annealing time with experimental data received by Naoi and Kajihara (2007) and Shibata et al. (1966), respectively, and the solid lines show the calculations from equation (2-2). Apparently, most experimental data fits the equation very well, indicating parabolic law is valid for the growth of the η phase in Fe-Al system over these temperature ranges, and this result confirms again that the growth of the η phase is controlled by volume diffusion.

Furthermore, after the parabolic coefficient k is calculated at each temperature, the logarithm of k can be plotted against the reciprocal of temperature (T) to obtain activation energy. The data acquired by previous researchers is shown in Fig. 2.39 (c) and (d), and the straight lines are calculated from equation (2-3). Although Springer et al. (2011 b) calculated an activation energy with the value of 190 kJ/mol by collecting the data from three studies in Fig. 2.39 (d), different activation energies have been calculated by these researchers, and two values are shown in Fig. 2.39 (c) obtained by Naoi and Kajihara (2007) and Shibata et al. (1966). It is interesting that the activation energy of the η phase distributes across a large range, according to the different experimental conditions used by previous researchers, and more values are summarized in Table 2.4. It should also be noted that Springer used an incorrect forms of equation 2-3, resulting in the wrong value of activation energy.

Chapter 2 Literature Review

Besides the activation energy, the interdiffusion coefficient for the IMC layer (main η phase) was also investigated by Kajihara (2006), as the diffusion coefficient and the solubility range of the IMC phase could control the growth rate in a binary system. For the Fe-Al system, this means the growth rate of the η phase (Kajihara 2004). The diffusion coefficient D for each phase had a similar expression as equation (2-3):

Q D = D exp⁡(− ) (2-4) 0 RT

Chapter 2 Literature Review

2 -1 2 -1 Here D is the diffusion coefficient (μm ∙s ), D0 is the pre-exponential factor(μm ∙s ), Q is the reaction active energy or enthalpy (kJ/mol), R is the gas constant (J∙K-1/mol), and T is the reaction temperature (K).

The relationship between parabolic coefficient k and interdiffusion coefficient D of the η phase can be expressed, as the following equation (Kajihara 2006):

k≡4D∙(kβα-kβγ)2 (2-5)

Here kβα and kβγ are dimensionless coefficients, while α, β, γ indicate Fe, η and Al phases, respectively.

Table 2.4 Activation energies for η phase from previous studies between aluminium and iron by diffusion bonding.

Activation energy Temperature Materials (kJ/mol) (oC) Bouayad et al. (2003) 74.1 800 Pure Al- pure Fe Denner et al. (1977) 170-195 771 Pure Al- Mild steel Pure Al - Low carbon Eggeler et al. (1986) 34-155 786 steel Naoi et al. (2007) 281 600-650 Pure Al- pure Fe Pure aluminum- low- Springer et al. (2011 b) 190 600 carbon steel Tang et al. (2012) 123 680-770 Pure Al- pure Fe

With the suitable initial and boundary conditions, D was calculated from equations built by Kajihara (2006), and the values are plotted as open circles in Fig. 2.40 (a), together with the parabolic coefficient k, or K as open squares at the same experimental temperatures. D0 and Q were calculated from the open circles by the least-squares method, shown in Fig. 2.40 (a), and the temperature dependence of D and K can be calculated as a function of temperature, shown as the solid line and the dashed line, respectively. Due to the similar activation energies between k and D, the two lines are almost parallel to each other. Furthermore, the diffusion coefficients for volume diffusion in iron (α), η (β) and aluminium (γ) phases are calculated as a function of temperature, and the results are shown in Fig. 2.40 (b) with different type lines. Apparently, the diffusion coefficient for the η phase in the substrate is a little smaller than that of aluminium in the η phase, but much larger than that of iron in η phase. The upper limit of

Chapter 2 Literature Review the diffusion coefficient for the invisible Fe-Al intermetallic phases was also calculated by Kajihara (2006), with the expression of Dδ. It is clear that the diffusion coefficients for these phases, including FeAl2 and FeAl3, are much smaller than that for the η phase.

Fig. 2.40 (a) The parabolic coefficient K of the η layer vs. the reciprocal of the annealing temperature T shown as open squares. The evaluations for the interdiffusion coefficient D of η phase are indicated as open circles. Straight lines show the calculations from equations (2-3) and (2-4) for k and D, respectively. (b) The interdiffusion coefficient D vs. the reciprocal of the annealing temperature T shown as various straight lines for the α, β,

γ and δ phases, which stand for Fe, η (Fe2Al5), Al and other Fe-Al intermetallic compounds invisible here, respectively.

If the growth of the IMC layer is considered to be controlled by both the lattice diffusion and grain boundary diffusion, the diffusion coefficients in both diffusion procedures can be expressed by the Arrhenius relationship in equations 2-6 (Fisher 1951, Mehrer 2007):

푄 퐷 = 퐷 exp⁡(− 푙 ) (2-6 a) 푙 푙0 푅푇

푄 퐷 = 퐷 exp⁡(− 푔푏) (2-6 a) 푔푏 푔푏0 푅푇

Here 퐷푙⁡and 퐷푔푏 are the diffusion coefficients for lattice diffusion and the grain boundary diffusion, 퐷푙0⁡and 퐷푔푏0are pre-exponent factors for lattice diffusion and grain boundary diffusion, 푄푙⁡and 푄푔푏⁡are the activation energies for lattice diffusion and grain boundary diffusion, respectively. It should be noted that the pre-exponent factors for lattice

Chapter 2 Literature Review diffusion and grain boundary diffusion is same, but the activation energy for grain boundary diffusion is about half of that for lattice diffusion (Dietrich 2011).

The effective diffusion coefficient of the IMC layer can then be calculated by equation 2-7 (Hart 1957, Mehrer 2007):

퐷푒푓푓 = 푔퐷푔푏 + (1 − 푔)퐷푙 (2-7)

Here 퐷푒푓푓⁡is the effective diffusion coefficient, g is the effective volume fraction of the grain boundary diffusion in the whole diffusion, and can be described as equation 2-8 (Mehrer 2007):

푞훿 푔 = (2-8) 퐿

Here q is a numerical factor relating to the grain shape, 훿⁡is grain boundary width, L is the grain size. 훿⁡is usually assumed as 3 times of the atomic diameter. For columnar grains in the present study, q=1, L is the average grain width.

Table 2.4 shows activation energies calculated by previous researchers, where in all cases the η (Fe2Al5) phase comprised the majority of the IMC layer, as a thin θ (FeAl3) layer is only detected when aluminium is liquid during annealing and the thickness remains at a constant value. It can be seen that the activation energy value changes in a wide range, from 34 to 281 kJ/mol. This wide range results from many factors, including dissolution and spalling effects during immersion (Richards 1994), different thickness measurements like maximum or mean thickness (Richards 1994), too long annealing time (Eggeler 1959), annealing temperature (Bouayad 2003, Tang 2012, Naoi 2007, Springer 2011 (b)), content of iron in aluminium melting (Eggeler 1959), carbon content in the substrate steel (Denner 1977), the presence of different alloy elements in aluminium and steel (Akdeniz 1998, Springer 2011 (b)), surface condition of base metal especially aluminium (Naoi 2007). Of these, the annealing temperature and alloy elements are the main factors that can affect the growth kinetics of the IMC layer with solid state diffusion couples. Other factors either are refined by the technique (e.g. the thickness measurement method) or have little influence in solid state diffusion process (too long annealing time, content of iron in aluminium melting). Therefore, the influence of both annealing temperature and alloy elements will be summarised in this section.

Chapter 2 Literature Review

According to Naoi and Kajihara (2007)'s analysis of the binary Fe-Al system, the effect of annealing temperature on the growth of IMC layer is mainly a result of grain boundary diffusion. After investigating and comparing the growth kinetics in Au-Sn, Cu-Al and Fe-Al binary system, it was found that IMC layer in all cases can be described as a power function of annealing time, but the exponent of the power function changes with annealing temperature. They argued that the grain boundary diffusion in IMC layer contributes to the rate-controlling process for the growth of IMC layer as well as volume diffusion, and grain growth occurs at certain rates in the IMC layer. In particular, as the grain growth slows down at low annealing temperature, the effect of grain boundary diffusion is more remarkable than at high annealing temperature. Therefore, the calculated activation energy varies with the annealing temperature in Table 2.4.

Akdeniz and Mekhrabov (1997) investigated the effect of alloy elements on the evolution of IMC layer in a dissimilar couple between aluminium and iron. According to the experimental results, a higher activation coefficient of aluminium atoms in Fe-Al system resulted in a faster growth rate of IMC layer. Therefore, alloy elements could be divided into two types: I-group which decreased the activation coefficient of aluminium atoms, including Si, Ti, Ge, Sb, Mg, Cu, Ca, Ag, Cd, and Cr, and II-group which increased the activation coefficient of aluminium atoms, including Co, Zn, Mn, Pb and Bi. Among these elements, Si and Zn received the most attention, as both elements were used widely in aluminium alloys 6xxx (Al-Si) and 7xxx (Al-Zn) series, respectively (Richards 1994, Springer 2011 (b), Luo 2012).

Besides decreasing the activation coefficient of aluminium atoms, two possible reasons were claimed concerning the influence of Si on the growth rate of the IMC layer in the Fe- Al system. Yin et al. (2013) believed that Si reduced the growth rate of the IMC layer by occupying the large number of aluminium vacancies on the c-axes ([001] direction) of the orthorhombic cells of η phase, while Marder (2000) and Springer et al. (2011 b) agreed that the thermodynamically preformatted formation of Fe-Al-Si ternary phases, with a slower nucleation and growth rate than η phase, reduced the growth rate of IMC layer. It is still not clear which explanation is more accurate, although many researchers have studied the effect of Si.

Chapter 2 Literature Review

The effect of Si on the growth behaviour of IMC layer in dissimilar combinations between iron and aluminium has been widely reported. Stroup and Purdy (1950) pointed out that the suppressing effect of Si on IMC layer growth would gradually decline with more Si being added. Furthermore, Coburn (1964) suggested an optimum Si addition of 8.5-9.5 wt. % to reduce the IMC layer thickness. New results presented by Springer et al. (2011 b) show that Si was found to increase the growth rate of the IMC layer between low carbon steel and Al-5 wt.% Si at 600oC, although Si still depressed the growth rate at 675oC. In

o this case, unlike the composition of IMC layer at 675 C, where τ5 (Fe2Al8Si) and the θ phases constituted a significant fraction, η phase again became the major fraction at 600oC. However, the Si content was similar in the IMC layer in both cases, indicating the occupation of Si in aluminium vacancy sites in the in η phase might have little influence on the growth of IMC layer. Furthermore, growth of the η phase was still well described by the parabolic equation, indicating the interfacial reactions also had little effect on this unusual growth behaviour (Springer 2011 b). Together with data by Eggeler et al. (1986), an activation energy of 17 kJ/mol was calculated by Springer et al. (2011 b) for IMC layer

(including τ5, θ and η phases) between steel and Al-5 wt. % Si which seems unreasonable low. By comparing the growth kinetics of IMC layers in both Fe/pure Al and Fe/Al-Si couples, a critical temperature of around 668oC was found. At higher temperatures, Si suppressed the growth rate of the IMC layer, while at lower temperature, Si increased it. However, the mechanism of the influence of Si is still not clear.

Little studies have been carried out on the effect of Zn on the growth behaviour of IMC layers in the Fe-Al system, but many researchers have focused on the effect of Al on Zn- coated steel. Marder (2000) summarized the metallurgy of zinc-coated steels, including the IMC reactions in the Fe-Zn-Al ternary system with different aluminium contents.

Fe2Al5Znx was found to be formed at the interface between coating and steel substrate, as a result of high solubility of zinc in η (Fe2Al5) phase, and this layer was found to inhibit the Fe-Zn reaction diffusion. Akdeniz and Mekhrabov (1997) suggest that Zn can accelerate the growth rate of the IMC layer by increasing the activity coefficient of aluminium. Luo et

o al. (2012) observed the interface layer between Al96.2Zn3.8/Fe at 300 and 550 C, and the

IMC phase sequence across the interface was identified to be Al - θ (FeAl3) - η (Fe2Al5) - Fe. Apparently, no new phase was formed as zinc diffused into the IMC layer, but it should be

Chapter 2 Literature Review noticed that θ phase was detected when aluminium was in the solid state. There is no research data which can be directly compared to their result on the growth kinetics of the IMC layer. However, the parabolic coefficient k in equation (2-1) at 550oC was calculated by Naoi and Kajihara (2007) with the value of 1.60 × 10-16 m2/s for pure aluminium and iron; therefore, without zinc the thickness of IMC layer should be 15 μm at 550oC for 16 days in theory, which is less than quarter of the experimental thickness of approximately 70 μm observed by Luo et al. (2012). Huge gap between the calculated and experimental thicknesses, indicates the possibility that zinc can accelerate the interface reaction rate in the Fe-Al system.

2.6 Summary

According to the multi-materials design concept, more and more aluminium alloys will be used in the automotive industry, together with steel sheets as the main body structure. Due to their high formability, reasonable strength, and heat treatable properties, both 6xxx and 7xxx aluminium alloys will make up a large part of the application of aluminium alloys in future automotive construction. The low carbon steel DC04 is also selected in present study as a result of its excellent combination of strength, weldability and price. Consequently, if suitable metals are chosen based on the concept of optimum multi- materials design, the problem remains of how to join dissimilar metals.

In general, liquid state welding techniques like laser welding and resistant spot welding are not suitable for making dissimilar joints between steel and aluminium alloys, as a result of the fast growth rate of the intermetallic compounds (IMC) layer. Therefore, solid state welding, including friction stir (spot) welding and ultrasonic spot welding together with variants on these processes, have been developed for steel and aluminium alloy welding. Sound dissimilar joints with little IMC layer can be received by these solid state welding techniques. Therefore, solid state welding techniques are excellent for dissimilar metal welding.

Compared with friction stir welding, some advantages for ultrasonic spot welding have been reviewed, including disturbing the oxide layer into the base metals, the energy being generated at the interface directly, less surface damage, and faster welding efficiency. As a result, more attention has been paid to ultrasonic spot welding than before, and

Chapter 2 Literature Review optimum welding parameters have been worked out for dissimilar welding between steel and aluminium alloys, which limits the thickness of the IMC layer. However, as the formation of an IMC layer is inevitable, it is important to understand the factors that control its formation and growth behaviour.

In Fe-Al couples, in either the as welded or annealed states, η (Fe2Al5) phase is always found as the predominant phase in the IMC layer as a result of interfacial reaction, with θ

(FeAl3) phase only appearing when aluminium melting occurs. Other phases are not likely to appear unless aluminium is consumed entirely by the IMC layer (i.e. in aluminium coating steel case). This is related to the low activation energy and high diffusion rate of the η (Fe2Al5) phase into the substrate, as a result of its unique crystal structure which involves a low aluminium site vacancy along [001] direction. The orientation relationship between the η phase and the iron substrate also favours nuclear, and these two factors together cause the directional growth of columnar grains and a tongue like shaped interface between the η phase and the steel substrate.

The activation energy of the η phase has been reported to have a variety of values in different studies due to many factors affecting the growth behaviour of the IMC layer. Of these factors, the annealing temperature and alloy compositions are thought to be the main influences, as the temperature affects the state of aluminium and grain growth of the η phase, while alloy elements can either increase (like Zn) or suppress (like Si) the growth rate of the η phase for different reasons. It is found that below the melting point of aluminium, which is usually the welding temperature in solid state welding, grain boundary diffusion also affects the growth of IMC layer, but grain growth slows down and the grain boundary density at the interface remains at a certain value; therefore, parabolic law is suitable to calculate the parameters of the η phase used in growth kinetics, which will be used in predication of IMC layer growth in welding process.

Si has two opposite effects on the IMC layer growth behaviour divided by a critical temperature. Si accelerates the growth rate below and suppresses it above this temperature. Three mechanisms have been pointed out to explain the effect of Si, but none of them are satisfactory in all cases.

Chapter 2 Literature Review

Adding Zn into the pure aluminium and iron system could cause a thin layer of the θ phase to be detected between the η phase and aluminium below the melting point of aluminium. When compared with the calculated thickness of the IMC layer in pure aluminium and iron system at the same temperature, it was shown that it is possible that Zn could accelerate the growth rate of IMC layer. However, there is no reliable published data on the effect of zinc on the IMC reaction kinetics in steel-aluminium welding.

Chapter 3 Experimental Methods

CHAPTER 3 EXPERIMENTAL METHODS

In order to investigate the formation and growth of IMC interface reaction layers in aluminium to steel dissimilar solid state welding, two commercial aluminium alloys, 6111- T4 and 7055-T4, were selected and joined to a low carbon steel, DC04, by using ultrasonic spot welding (USW) and friction stir spot welding (FSSW). After welding, the mechanical properties of the joints were analysed by lap shear testing. In order to understand the mechanism of formation and growth behaviour of the IMC layers produced in the joints, their interfaces were analysed using high resolution microscopy techniques, including Scanning Electron Microscopy (SEM), Energy Dispersive X-Ray (EDX) analysis, Electro Backscatter Diffraction (EBSD) analysis, as well as by Transmission Electron Microscopy (TEM), with the samples prepared by Focus Ion Beam (FIB) milling. Heat treatments were also performed on weld samples to collect kinetic data for the growth of IMC layers during long term interdiffusion, which was used to develop a model to predict the interface layer growth during welding. The materials and each of the methods used are explained in this chapter.

3.1 Parent Materials

The materials used in this investigation were the aluminium alloys 6111-T4, 7055-T4, and the low carbon steel DC04. Their nominal compositions are given in Table 3.1.

Table 3.1 Chemical compositions (wt. %) of the base materials used in the present work.

Alloy Al Cr Cu Fe Mg Mn Si Ti Zn C 6111 balance - 0.59 0.26 0.58 0.20 0.82 - - - 7055 balance 0.04 2.0 0.1 1.9 0.05 0.1 0.06 8.4 - DC04 0.025 - - balance - 0.40 - - - 0.08

The thicknesses of the as-received materials are shown in Table 3.2. Both the AA6111-T4 and DC04 sheets were used directly in the as-received state with original thickness, while rolling was used on the 3 mm thick AA7055-T4 sheet to reach a similar final thickness to the other materials. Solution heat-treatment was used before and after rolling to ensure a uniform distribution of alloying elements. The parameters used in these processes with the AA7055 alloy are shown in Table 3.2. It should be noted that the artificial aging treatment used for AA7055 after welding was 24 h at 120oC, resulting in a T6 state for

Chapter 3 Experimental Methods

AA7055, which is not shown in Table 3.2. The hardness of the base materials used in the welding experiments (AA6111-T4, AA7055-solution, DC04) was measured using a microhardness indenter (Section 3.8), taking the average of 10 measurements performed on 3 different sheets of each material, and the results are shown in Table 3.2. The AA7055 alloy was consistently welded 2 hours after solution treatment. Before welding, the welding area in all of the materials was ground with # 180grit SiC paper, and then washed in methanol to produce an oxide free clean surface.

The parent microstructure of all the materials used was characterized by SEM in backscatter electron (BSE) mode before welding, as shown in Fig. 3.1. It should be noted that all the materials shown in Fig. 3.1 were in the state used directly before welding, which means AA6111 was in a T4 state, and AA7055 was in a solution treated state. The average grain size was measured by the linear intercept method, in which the number of grain boundary were counted, and then the mean grain size was calculated by dividing the line length by the number of intercepts. As a result, the average grain sizes determined for the DC04, AA6111-T4 and AA7055 materials prior to welding were 20, 21 and 17 μm, respectively.

Table 3.2 Parameters of the materials used in the experimental.

Original Rolling conditions Solution treatment Base Hardness thickness Temperature Time Final thickness Temperature material Time (minute) (Hv) (mm) (oC) (minute) (mm) (oC) 6111-T4 0.92±0.01 - 530 60 80±3 7055- 3.05±0.01 430 30 1.06±0.05 455 90 100±2 solution DC04 0.97±0.01 - 93±3

3.2 Ultrasonic Spot Welding

All of the ultrasonic spot welds were produced using a dual head MH2016TM machine, manufactured by Sonobond Ultrasonic. A scheme illustration of this machine is shown in Fig. 3.2 (a). The welding tips used for the steel and aluminium welds are shown in Fig. 3.2 (b), and a picture of the machine with its control box is shown in Fig. 3.2 (c). This machine uses the principle in which a series of DC pulses convert the frequency of the electrical power supply from a typical initial frequency of 60 Hz to a high frequency of 20.5 kHz. The electrical energy is then converted into mechanical movement of the same high

Chapter 3 Experimental Methods frequency by a piezoelectric material within a transducer, and this movement is finally amplified and transmitted to the work-pieces via the weld tips, which are placed at the end of the reeds. During the welding process, a constant clamping pressure is applied through the reeds and tips to the work-pieces by a piston driven by compressed air. Lateral vibration is applied by the machine to both the top and bottom workpieces through two separate wedge/reed/tip combinations that vibrate out of phase with one another, and the vibration occurs linearly in one direction only.

Fig. 3.1 Microstructure of the parent materials before welding (a) DC04, (b) AA6111-T4, and (c) AA7055-solution in BSE mode.

The welding tips and welding parameters were optimized for the welding of steel to aluminium prior to this work by Haddadi (2012). The tip used for the Al side was a standard flat serrated sonotrode with dimensions of 8×6 mm∙mm, whereas the tip used for the steel side had a radiuses surface and a surface curvature of 0.2 mm -1, as shown in

Chapter 3 Experimental Methods

Fig. 3.2 (b). There were four welding parameters that can be set by the control box, including the welding time, input power, clamping pressure, and impendence (current/voltage phase matching). In this project, the welding time was changed from 0.1 to 3.0 seconds, to obtain joints with different IMC interface layers thicknesses, while all of the other parameters were fixed. The input power was set to 2500 W, the clamping pressure 60 psi (which means the clamping force was 1.4 kN), and the impendence at 2 Ω for all welds.

The geometry of the work-pieces used in the USW is shown in Fig. 3.3. Both the steel and Al alloys were cut into the strips with the dimension of 100×25 mm∙mm. The overlap between the dissimilar metals was 25×25 mm∙mm, and the welding area was located at the centre of the overlap. The vibration direction was parallel to the rolling direction of the metals, and both directions were perpendicular to the teeth serrations, as shown in Fig. 3.2 (b).

Following welding, artificial ageing was carried out to obtain T6 tempers in both aluminium alloys. For the AA6111-DC04 combination this involved ageing for 24 hours at

Chapter 3 Experimental Methods

180oC (Marceau 2013), while for AA7055- DC04 joints, ageing was performed at 100oC for 8 hours, followed by 24 hours at 120oC (Schreiber 2014).

Fig. 3.3 The geometry of the test coupon joints made by USW.

3.3 Friction Stir Spot Welding

Two FSSW variants were applied to make dissimilar joints for different purposes, as shown in Fig. 3.4. The first method used was the traditional pin-less FSSW method, which was used to compare the interface layer microstructure with that from dissimilar joints made by USW. The second method used was the Abrasion Circle FSSW (or ABC-FSSW) process developed by Chen (2012). ABC-FSSW was used to make dissimilar joints without an interface IMC layer so that an initially clean interface could be used with heat treatments to generate isothermal kinetic data for the interface reaction layer. The sample configuration used in FSSW was similar to that used in USW (Fig. 3.3). The two FSSW weld tools and surface appearance are shown in Fig. 3.4. Both kinds of joints were welded on a CS Power stir Friction Stir Welding Machine, but the tools and travel paths were different (Fig. 3.4).

An illustration for the conventional FSW process is shown in Fig. 3.4 (a). With this process pin-less tool (Φ10 mm) was plunged into the Al top sheet to a certain depth and for a

Chapter 3 Experimental Methods given dwell time and then lifted out, leaving the aluminium highly deformed and creating a bond between steel and aluminium. With a pin-less tool the joint area can be maximized by eliminating the probe exit hole, but it requires a long welding time (>5 s) to obtain a strong bond between two dissimilar materials, because with a harder steel bottom sheet there is little residual material flow across the interface (Chen 2011). The parameters are listed as following: plunge and retraction rates of 1.7 mm/s and 0.8 mm/s, rotation speed of 1600 rpm, plunge depth of 0.5 mm, welding time of 1-10 s.

Fig. 3.4 Schematic diagram of the translation path, tools and weld surface on the Al sheet for (a) conventional pin-less FSSW and (b) the abrasion circle friction spot welding process (Chen 2011).

In the ABC-FSSW process, a tool designed with a probe (Fig. 3.4 (b)) was plunged into the softer Al sheet and then travels along a circular path, with a radius equivalent to the probe diameter, to obtain a continuous joining area. Finally, the tool is moved to the centre of the circle and extracted to make an axisymmetric weld. The interface in the joints made by this method are very clean, and no IMC layer was detected by TEM (Chen 2012), so joints made using this method were produced for the isothermal heat treatment experiments at different temperatures and times, to collect data for the growth kinetics of the IMC layer under controlled conditions. The parameters are listed as

Chapter 3 Experimental Methods following: welding speed 1000 mm/min, rotation rate of 800 rpm, plunge depth of 0.1 mm, welding time of 1 s.

3.4 Temperature Measurement

0.5 mm k-type thermocouples were used to measure the welding temperature at the joint interface at the weld centre between the steel and aluminium sheets. The configuration used in this study is shown in Fig. 3.5.

The thermocouples were inserted through a channel at the edge of the aluminium sheet to the centre of welding areas prior to welding. One typical example for a dissimilar joint containing a thermocouple is shown after welding in Fig. 3.5 (c). For this sample the aluminium surface was broken under the thermocouple channel, but this had little influence on the welding temperature history (Panteli 2012). It should be noted that the thermocouples were incorporated into the welds during welding. A National Instruments datalogger was used to record the temperature at 0.01 s intervals. Temperature measurements were repeated three times for each parameter to check for consistency within the data. Similar methods were used in FSSW, but the thermocouples were

Chapter 3 Experimental Methods embedded in the channel located at the edge of the bottom steel sheet to the centre of welding (Fig. 3.5 (d)).

3.5 Metallographic Sample Preparation

The metallographic samples from the dissimilar joints that were used for SEM characterization were prepared by the following steps, and an illustration of the orientation and position of the sample studied is shown in Fig. 3.6. The welds were all sectioned down their centre line parallel to the welding direction. Firstly, the lap containing the welding area was cut from the joint, as shown in the box in Fig. 3.6 (a). Then a Struers Minitom cut off machine was employed to cut through the centre of weld area along the rolling direction, using a MetPrep HNF type blade at a rotation speed of 20 RPM, and with a diameter of 125 mm and thickness of 0.45 mm. The cutting position is shown in Fig. 3.6 (b) as the dashed-straight line, which also defines the section surface prepared for metallographic observation. A schematic diagram of the sample cross section is shown in Fig. 3.6 (c). Then the samples were mounted in a brass holder, specifically designed for grinding and polishing.

Finally, the joint’s cross section was ground and polished with conventional metallographic method using the following stages. In the grinding stage, a StruerPol-31 automatic polishing machine was used with a 10 N applied force and water as a lubricant. #600 grit SiC paper was used firstly for 4 min, followed by 800 grit and 1200 grit for 1 min each, and 2000 grit paper was then used in the last stage for 3 min. In the polishing stage, 6 μm and 3 μm diamond paste was used for 5 min on each polishing cloth, with oil based polishing solutions and lubricants. 0.2 μm water free colloidal silica OPS solution was then used for the final polish, for 10 min for normal SEM observation and 1 hour for EBSD characterization. Following OPS polishing water was used to clean the surface, followed by washing in ethanol. Warm air was employed to dry the sample surfaces between each stage.

100

Chapter 3 Experimental Methods

3.6 Heat Treatment

As mentioned in 3.1, the AA7055 alloy was solution treated using the parameters given in Table 3.2, both before and after rolling down to the standard 1 mm thick sheet used for welding. This alloy was welded within two hours of solution treatment.

Artificial aging was also applied to the DC04-AA6111 and DC04-AA7055 weld combinations after welding with different welding times, in the range of 0.3 - 2.0 s. The purpose of artificial aging was to maximize the strength of the aluminium sheets for all the joints. For the DC04-AA6111 combinations, aging was performed at 180oC for 24 hours (Marceau 2013), while for DC04-AA7055 joints, aging was performed at 100oC for 8 hours, and then at 120oC for 24 hours (Schreiber 2014). These ageing conditions correspond to the standard practice that is used to produce the peak strength condition in both alloys after solution treatment (T6 temperature). The samples were all stored in a

101

Chapter 3 Experimental Methods freezer directly after welding in both the as-welded and aged states, to inhibit natural ageing.

Isothermal heat treatments were applied to the different pre-welded couples for several purposes. The annealing parameters used for five groups of samples are shown in Table 3.3. Groups 1, 2 and 3 were used to study the IMC layer growth behaviour, including the influence of different welding methods on the microstructure and for collecting kinetic data of the IMC layer’s growth rate. Groups 1 and 2 were used to clarify the effect of the initial welding stages on the growth behaviour of the IMC layer, including for phase identification and to study the morphology of the IMC phases. Groups 2 and 3 were used with the same annealing times to comparing the growth behaviour of IMC layer were produced in the two couples produced with the two different aluminium alloys, in order to find the influence of alloying elements on the IMC layer growth rate. It should be noted that all of the combinations in groups 2 and 3 were pre-welded for 1.5 s by USW. All of the joint configurations used for the welding process were the same as those shown in Fig. 3.3. To conduct the heat treatment the lap sections were cut out by a Minitom cut-off machine before placing the welds immediately into a preheated furnace at the target temperature for certain annealing time. A k-type thermocouple was attached to each sample to monitor the annealing temperature, which was maintained within an accuracy of ± 3oC. After annealing was completed, samples were immediately air cooled to room temperature. The samples were then metallographically prepared using the same stages as described in section 3.5 before analysis.

Table 3.3 Annealing parameters for different couples.

Group Welding Welding Welding Annealing Annealing time number couples method time (s) temperature (oC) (min) DC04- 400, 450, 500, 550, 1 ABC-FSSW 5 10 - 7680 AA6111 570 DC04- 400, 450, 500, 550, 2 USW 1.5 10 - 7680 AA6111 570 DC04- 3 USW 1.5 400, 450, 500 120, 240, 480 AA7055

102

Chapter 3 Experimental Methods

3.7 Lap Shear Testing

Tensile lap shear testing was performed on an MTS Alliance RT/100 mechanical testing machine using a 50 kN load cell. All tests were performed with a crosshead travel speed of 0.5 mm/min, and the data were collected at 50 Hz. Specimens used in the lap shear testing had the same joint configuration as shown in Fig. 3.3, and were tested along their long axis, as shown in the schematic diagram in Fig. 3.7. Tabs of the same thickness and width of each sample were applied in the position shown in Fig. 3.7 to ensure that the tensile force was coaxial with the welding interface. Three tests were repeated for each joint with each parameter set to check for consistency within the data.

After testing, the load vs. displacement curves were processed to obtain the fracture energy by calculating the area under the curves, and the peak load at fracture that was the highest load value recorded in each test. The area under the curves was calculated using Origin 8.5 software with the integration method; therefore, the error associated with this method was sensitive to the incremental values in the data. A high sampling rate of 50 Hz and low travel speed of 0.5 mm/min was therefore applied to minimise the error giving an increment of 0.16 μm between readings.

Fig. 3.7 Schematic diagram of the lap shear weld coupon test specimen, (a) top view and (b) side view.

3.8 Hardness Testing

Micro hardness testing was performed on the parent materials and welded joints in both the as-welded and aged states, to establish the local variation in yield stress and hardness across the weld regions. After being prepared by the standard metallographic techniques

103

Chapter 3 Experimental Methods mentioned in section 3.5, sample cross-sections were placed in a Vickers CSM micro indentation machine, and an integrated microscope with a measuring facility was used to measure the diamond shaped indentation, following the schematic diagram shown in Fig. 3.8. For reliable measurement the samples were first prepared with a mirror polished surface, and parallel top and bottom surfaces. The Vickers hardness value (Hv) was calculated by equation 3.1:

Hv=1.854(P/d2) (3.1) where P is the applied load, and d is the average of d1 and d2.

Fig. 3.8 Schematic diagram of the microhardness technique (Panteli 2012).

In order to ensure that individual micro-hardness indentations had no influence on each other, a distance of 5 indentation diameters was left between measurements. The measurement position was chosen to be a the line at least 100 μm away from the interface, to obtain the hardness values in the weld heat affected zones (HAZ), as shown in Fig, 3.9 (a). In order to obtain sufficient points along the sample with sufficient spatial resolution to detect the weld HAZ, a suitable load was selected for each combination; therefore, 0.2 and 0.5 N were found to be satisfactory for the DC04-AA6111 and DC04- AA7055 combinations, respectively.

104

Chapter 3 Experimental Methods

Micro hardness was also used to mark the original interface of the welds after the generation of IMC layer after annealing, so that the growth direction could be confirmed. An illustration for the measurement method used is shown in Fig. 3.10. Three locations were chosen along the welding interface on the cross section, and four indentations in each place were used to mark the original interface’s position, as shown in Fig. 3.10 (a) and (b). After annealing, the IMC layer grew into a certain thickness, and the original interface was positioned inside the IMC layer by the marked indentations, which is shown as the dashed line in Fig. 3.10 (c). Therefore, the growth direction was confirmed by measuring the thickness of the IMC layer from the original interface into the substrates.

105

Chapter 3 Experimental Methods

3.9 Surface Profile

In order to understand the influence of base metal surface roughness on welding performance, especially for the different type aluminium alloys, the roughness for each metal sheet after being ground prior to welding was determined using a 3D NanoFoucus µscan SC200 profilometer based on a laser source. The samples were placed on a flat table, with the laser focused on the surface. The laser was then scanned in several lines across the area that would be used in the weld region. The surface profile measured was then analysed using μsurf software version 6.1, with the data collected from the reflected laser beam, with respect to a reference beam. The line scan had a step size of 1 μm, and the roughness value was calculated by the software. In particular, the arithmetical mean roughness, Ra, was used to quantify the roughness. Ra is the arithmetic average of the absolute values of the roughness profile ordinates, also known as Arithmetic Average (AA) or Centre Line Average (CLA). It can be calculated by equation 3-2 (DeGarmo 2003):

1 푅 = ∑푛 |푦 | (3-2) 푎 푛 푖=1 푖

Where Ra is the arithmetical mean roughness, and |yi| is the absolute values of the roughness profile ordinates.

3.10 Scanning Electron Microscope (SEM)

Scanning Electron Microscopy (SEM) was used to observe the microstructure across the interface of the dissimilar joints, including for determining the IMC layer growth morphology, grain structure, texture, element distribution, and for phase identification. The samples used in this section were prepared with the method introduced in section 3.5. In this section, the different SEM techniques used in this study will be introduced.

3.10.1 Scanning Electron Microscope (SEM)

In Scanning Electron Microscopy (SEM) a focused electron beam is used to collect information from the sample surface from the interaction between the beam and the sample. SEM images have higher resolution than those that can be obtained by optical microscopy, due to the shorter wavelength of electrons than light. In this study, three types of microscope were used for SEM analysis; a Philips XL30, FEI Quanta 650, and an

106

Chapter 3 Experimental Methods

FEI Magellan. All of these microscopes were equipped with a Field Emission Gun, allowing high resolution images (FEG-SEM) to be obtained.

In the FEG-SEM, electrons are emitted from a sharp tip and accelerated to the sample surface placed in a vacuum chamber, by the high potential difference between the gun and sample surface, which act as the anode and cathode, respectively (Watt 1985). A condenser lenses and objective lens are used to control the diameter and focus the beam. Increasing the accelerating voltage increases the interaction volume, which is the volume of sample material that interacts with the electrons below the surface. In general, two types of electrons are generated from the interaction between the beam and sample surface, secondary electrons (SE) and backscattered electrons (BSE), which provide different information from the sample (Watt 1985).

In this study, the secondary electron (SE) imaging mode was used to characterize the topography of fracture surfaces after lap shear testing and the interface area, including the pores left by Kirkendall's effect and cracking in the interface layer. In this mode, secondary electrons from a low depth below the sample surface are emitted due to the excitation of electrons in higher energy shells of atoms near the surface. Inelastic scattering occurs after energy is transmitted to these electrons, followed by the ionizing electrons. As the surface condition has dramatic influence on the intensity of the ionized electrons emitted from each point, contrast from secondary electrons is very sensitive to surface topography. Therefore, high resolution morphology images can be obtained from the SE mode (Watt 1985).

In the present study, the backscattered electrons (BSE) imaging mode was used to identify different areas with different chemical compositions, like the IMC layer and base metals. It can also reveal the grain structure in each phase, as contrast in this mode reflects the difference in the atomic numbers and grain orientation. In this mode, elastic scattering occurs due to the interaction between the incident beam and inner shell electrons of atoms in the sample material, resulting in brag diffraction of incident electrons (Watt 1985). More back-scattered electrons can be generated from higher atomic number materials and certain grain orientations, giving rise to a brighter parts in the image. Therefore, the contrast in images viewed in BSE mode indicates differences in

107

Chapter 3 Experimental Methods atomic number and grain structure. It should be noted that as the interaction volume is lager and penetration of incident electrons is deeper in BSE mode than that in SE mode, BSE images have lower resolution than SE images (Watt 1985).

Both types of imaging modes were used in this study, with accelerating voltages in the range of 8 to 20 kV. The samples were mounted on specimen holders, with Silver DAG paint joining them, to make a conductive path between the sample surface and microscope.

3.10.2 Energy Dispersive X-ray (EDX)

Energy Dispersive X-ray (EDX) in SEM and TEM is used to characterise the element distribution and content for a certain area. In the EDX technique, the emission of a characteristic X-ray is excited by the interaction between the high energy electron beam and electrons in the sample surface. After the beam excites them, the inner shell electrons are ejected and higher shell electrons drop down to fill the gap. An X-ray is formed from the emission of the energy difference between the higher shell electron and the ejected inner electron. Each element has its characteristic energy difference; therefore, the EDX technique can quantitatively analyse the composition and element distribution in a sample (Watt 1985).

3.10.3 Electron Backscatter Diffraction (EBSD)

Electron back-scatter diffraction (EBSD) in the SEM is used to characterize crystal orientations, grain orientations, and local texture etc. by analysing back scatter diffraction patterns emitted from the sample surface. EBSD was performed on weld samples using an FEI Magellan microscope with an accelerating voltage of 20 kV. Channel5 EBSD acquisition software supplied by HKL Technology ApS., Denmark was used to analyse the results.

The EBSD technique is based on the automated analysis diffraction patterns emitted from the sample surface. The sample is tilted by 70 o related to the axis of the electron beam, and an incident electron beam is focused on the surface. Diffraction for backscattered electrons happens under Bragg's law condition, as a result of the interaction between crystal lattice planes and the inelastically scattered electrons. Pairs of lines are captured

108

Chapter 3 Experimental Methods on a phosphor screen, resulting in the formation of Kikuchi lines, which can be used for phase identification and orientation observation from the material in the interaction volume (Baba-Kishi 2002).

In this study, EBSD was used mainly for phase identification and grain structure observation within the IMC reaction layer, together with orientation characterisation. It should be noted that to obtain sufficient levels of indexing, after normal SEM samples preparation, focused ion beam (FIB) milling was used to obtain a strain free surface for observation. This will be introduced in the next section.

3.11 Focused Ion Beam (FIB) Milling

A Focused ion beam (FIB) milling system resembles an SEM, but uses a finely focused beam of ions (usually gallium (Ga+)) which can be applied for imaging with low beam currents, or sputtering, or milling specific sites with high beam currents. FIB can also be used to deposit material with a precursor gas, like Pt or W. The milling and deposition modes were mainly used in the present study to prepare both TEM and EBSD samples, using an FEI Quanta 3D dual beam FIB.

An illustration of the principle for the milling mode is shown in Fig. 3.11 (a) (Reyntjens 2001). In the milling mode, the gallium (Ga+) ion beam scans over and hits the substrate surface, and sputters away certain material from the substrate (Fig. 3.11 (a)). If the beam current is low, the sputtered material is limited, and a image of the substrate surface can be formed by collecting the signal from the sputtered ions or secondary electrons, which are produced from the interaction between the beam and substrate. However, when the beam current is high enough, the substrate is etched by sputtering the surface material. When used for sample preparation for EBSD, the strained surface material was removed by milling. In the sample preparation for TEM, the whole sample was thinned to less than 300 nm by milling the surface.

An illustration of the principle for the deposition mode is shown in Fig. 3.11 (b) (Reyntjens 2001). In the deposition mode, chemical vapour deposition (CVD) occurs when the precursor molecules (or gases) are injected on the surface by a gas nozzle. After the precursor gas is sprayed on the sample surface, the ion beam decomposes the precursor

109

Chapter 3 Experimental Methods gas into volatile and non-volatile compounds. The non-volatile compounds, like Pt and W, remain as a thin film on the surface. This deposited film is used to protect the underlying sample from the following destructive sputtering of the beam. Therefore, the deposition mode is usually prior to the milling mode to protect the observation area.

Fig. 3.11 Illustrations for the principle of FIB for (a) milling and (b) deposition (Reyntjens 2001).

For TEM sample preparation the following steps were used, which are also shown in Fig. 3.12 (FEI 2007). Firstly, Pt deposition was applied in the FIB with the purpose of protecting the area of interest (Fig. 3.12a). This was followed by cutting trenches either side of the area to be studied (Fig. 3.12b). Then a u-cut was applied to the selected area with the shape shown in Fig. 3.12c. In the sample lift-out stage (Fig. 3.12d), a probe was attached to the sample using Pt deposition. Then the sample was attached to a specific holder, again using Pt deposition (Fig. 3.12 e). Finally, the probe was cut off from the sample, and thinning stages were employed until the sample was thin enough for TEM observation, with a cleaning stage in the last step to remove impurities induced in the process (Fig. 3.12 f). It should be noted that the parameters used, like the thinning and cleaning current, must be optimised to the sample type, and the sample is rotated to an optimum angle in each milling stage (FEI 2007).

110

Chapter 3 Experimental Methods

Fig. 3.12 Schematic diagram for the procedure of TEM sample preparation by FIB.

(a) Pt deposition, (b) Bulk-out, (c) U-cut, (d) Lift-out, (e) Mounting, (f) Thinning and cleaning (FEI 2007).

3.12 Transmission Electron Microscope (TEM)

Transmission electron microscopy (TEM) is a microscopy technique in which a beam of electrons is transmitted through an ultra-thin specimen, which interacts with the sample as they pass through it. TEMs are capable of imaging at a much higher resolution than light microscopes, owing to the small de Broglie wavelength of electrons (Rodenburg 2004). TEM analysis in this study was mainly performed using an FEI Tecnai TF30 microscope operating at 300 kV.

As an electron beam can be considered as both a particle and wave, the interaction between the beam and the sample can also be thought to have two types; for example, the scattering of electrons is considered as a particle behaviour, and the reflection and diffraction is considered as a wave behaviour. In fact, as the electrons can be considered as either particles or waves. The elastically scattered electrons and the coherent electrons are related to each other; e.g. the elastically scattered electrons are usually coherent, if the specimen is thin and crystalline.

Many types of electrons can be generated in the interaction between the electron beam and thin sample. An illustration of the interaction is shown in Fig. 3.13 (Williams 2009). By

111

Chapter 3 Experimental Methods collecting different types of electrons, TEM can be used to analyze different properties of the sample. In particular, the elastically scattered electrons, with small scattering angles, are the major source of contrast in TEM images, and also create much of the intensity in diffraction patterns, which were the two main TEM techniques used in this study (Williams 2009).

Fig. 3.13 Signals generated when a high-energy beam of electrons interacts with a thin specimen (Williams 2009).

An illustration for an entire TEM system is shown in Fig. 3.14 (a). After passing through the condenser system, the beam from the filament is focused on the thin specimen, and then goes through the objective lens to form the first image for either the microstructure image or diffraction pattern. The first image is amplified by the following project system, which contains a group of lenses. The objective lens has a dramatic influence on the resolution of the TEM, because a defect in the first image will be magnified by the projector system. The final image on the phosphor screen has the same resolution, but higher magnification than the first image (Rodenburg 2004).

112

Chapter 3 Experimental Methods

Fig. 3.14 Illustration of the TEM optical system in (a) imaging mode and (b) diffraction mode (Rodenburg 2004).

By changing the beam path through the strength of the lenses, two modes, imaging and diffraction, can be chosen to obtain different information from the specimen (Fig. 3.14 (a) and (b)). If the object plane of the objective lens is the specimen itself, then an image of the specimens microstructure is formed in the first image plane, which is the imaging mode. If the beam is focused on the back focal plane of the objective lens, a diffraction pattern can be formed in the first image plane if the specimen is a crystal, and an illustration for this case is shown in Fig. 3.14 (b). An objective aperture is placed in the back-focal plane, which is used to improve the image quality and contrast, and to form dark field images (Rodenburg 2004). In this study, the conventional imaging mode was used to observe the microstructure at the interface of dissimilar joints, and the diffraction mode was used to identify the phases present.

Besides the forward scattered electrons, back scattered electrons, like X-rays and backscattered electrons can also be collected by TEM, which is similar with the 113

Chapter 3 Experimental Methods interaction between the beam and sample in SEM. These electrons are collected by scanning transmission electron microscopy (STEM) mode (Williams 2009). The STEM mode is used for quantitative analysis techniques, such as energy dispersive X-ray (EDX) spectroscopy, electron energy loss spectroscopy (EELS) and annular dark-field imaging (ADF). In this study, EDX in the STEM was used to perform elemental analysis across the joints interfaces (Williams 2009).

3.13 Image Analysis

ImageJ software (1.45S) was used for quantitative image analysis to obtain microstructure information, such as the average thickness of the intermetallic layer and the area fraction of different features on the fracture surface. For example, the average thickness of the IMC layer could be calculated from SEM images by the software, with the procedure shown in Fig. 3.15. After copying the images to ImageJ, the scale was setup firstly using the scale bar in the images, with the 'Set scale' button from the 'Analyze' tab in the tool bar (Fig. 3.15 (b)). Then the measurement was done with different tools according to the purposes. For the area measurement, the 'Freehand selections' function was used to draw the outline of the target area (Fig. 3.15 (c)). For the length measurement, the ' Straight' function was used to draw a straight line of the object (Fig. 3.15 (d)). Then the area or the length was measured automatically with the 'Measure' button from the 'Analyze' tab. The intermetallic region was selected manually using the 'Freehand Selection' tool in ImageJ.

For a certain interface length after setting an accurate scale, the average layer thickness (x) was calculated by dividing the area (S) by the interface length (l). Furthermore, as the IMC layer was not uniform in the dissimilar welds, the thickness distribution was also studied with ImageJ. In one image, the IMC layer was divided uniformly into 50 units, with a length of l/50 for each unit (Fig. 3.15 (c), (d)). The average thickness of each unit was measured, and then the distribution of the thickness was calculated with statistics method in Origin. In each case, two SEM images were used to study the distribution of the thickness.

114

Chapter 3 Experimental Methods

115

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

CHAPTER 4 WELD FORMATION AND INTERFACE LAYER IN FE-AL DISSIMILAR JOINTS

4.1 Introduction

In order to investigate the effect of IMC reaction at joint interface on the production of sound dissimilar Fe-Al welds, it is critical to choose suitable materials and appropriate welding methods. According to the results of Haddadi (2012), uncoated low carbon steel DC04 is a good choice to be welded to an aluminium alloy to investigate weld formation due to its low content of alloying elements. However, the alloy elements in aluminium alloys are also known to affect the growth and formation of the IMC layer in dissimilar steel-aluminium joints. For example, Si has been found to inhibit the growth of an IMC layer by forming Fe-Al-Si ternary phases (Springer 2011 b), and Zn has also been found to have an influence on the IMC layer (Marder 2000, Luo 2012). In addition, solid state welding methods have been found to have an advantage in producing Fe-Al dissimilar joints compared to fusion processes as a result of the lower welding temperatures (Springer 2011 b, Naoi 2007, Bouche 2003). However, although both ultrasonic spot welding (USW) and friction stir spot welding (FSSW) are solid state techniques, little research has been conducted to compare the different mechanisms of the IMC formation between these two welding processes. Therefore, in this chapter, two different aluminium alloys, AA6111 (Al-Mg-Si) and AA7055 (Al-Mg-Zn), were chosen as the aluminium alloys to be welded with DC04, and two welding methods (USW and FSSW) were selected to prepare the welds.

Both the USW and FSSW are very rapid processes where kinetic considerations can potentially play a critical role in determining the IMC phases that form at the weld interface. Calculations were therefore first used to verify if the phases identified in the experimental results were consistent with those expected from their relative thermodynamic stability within the material couples studied. The interface reaction behaviour in the two welding processes will then be compared and discussed, as well as the effect of the different aluminium alloys used and the influence of the IMC layer on the welded joints mechanical properties.

In this chapter the effect of the material and weld process variables has been investigated and compared between these two processes on the formation of the IMC reaction layer. 116

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

The relationship between interfacial microstructure and mechanical properties of the joints has also been assessed for the USW joints. As the mechanical property of the joints made by FSSW has been analyzed by previous researchers (Chen 2010, Prangnell 2010, Chen 2012), the present study focused on the effect of the IMC layer on the properties of in FSSW joints. The microstructure of the interface reaction layer, including the phases formed, their grain structure, thickness and morphology, has been observed and compared in detail as a function of welding time for USW and FSSW, by high resolution techniques, including scanning electric microscopy (SEM), electron backscatter diffraction (EBSD) and transmission electric microscopy (TEM).

4.2 Parent Sheet Surface Condition

Before welding, the welding areas on the base metal sheets were ground with #120 grit SiC paper and then cleaned in acetone to remove the oxide layer. In order to check the consistence of surface condition for the different materials, surface roughness measurements were performed on the prepared sheets, and the results are shown in Fig.

4.1. As can be seen from the Ra values, the welding area in the three parent metals all had a similar surface roughness of Ra = 0.8 μm. Therefore, in the present study, the initial sheet surface condition would be considered to have little influence on the relative welding behaviour of the different materials.

Fig. 4.1 Surface roughness (Ra) profiles of the parent sheets after preparation by grinding with #120 grit SiC paper and cleaning in acetone.

117

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

4.3 The Microstructure And Mechanical Properties Of Fe-Al Dissimilar Joints Made By USW

4.3.1 Temperature Histories Of The USW Joints

Fig. 4.2 (a) shows examples of typical temperature histories measured during USW with an input power of 2.5 kW at the centre of the weld area at each joint interface. The peak temperatures reached with increasing welding times are summarized in Fig. 4.2 (b). The thermocouples were inserted using a small channel in the Al sheet and positioned at the joint interface as described in 3.4.The thermal data was only used if, in post-weld sample sectioning, the position of each thermocouple after welding was found to be within 1 mm of the weld centre where the temperature is relatively uniform (Robson 2012). The consistency observed in the heating stage in Fig. 4.2 (a), where the curves fully overlap, gives confidence in the accuracy of the measurements.

As soon as welding was initiated, the temperature rose rapidly at a rate approaching 1000oC s-1 for the first 0.4 seconds. For longer welding times, the temperature increased more slowly through the rest of the weld cycle as a steady state was approached. It can be seen that for welds with a duration time longer than 1 second the interface temperature had risen to exceed 500oC. After the power was turned off, the temperature initially dropped very rapidly down to 300oC, before decaying more gradually. The two

118

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints aluminium alloys used in the present study did not produce significantly different temperature histories, although there was a minor systematic difference in the peak weld temperatures. Welds produced under the same conditions with the 6111 alloy combination were consistently around 5oC hotter than for the AA7055-DC04 Steel samples (Fig. 4.2 (b)). This is an extremely small difference (1%) and would not be expected to greatly affect the relative IMC layer thickness seen with welding time.

4.3.2 Weld Characterization

Following welding, the interface between the aluminium and steel weld members was analysed by electron microscopy at increasing resolution (Figs. 4.3, 4.6, 4.7). Fig. 4.3 shows macro views of typical cross sections of the joints made by USW, produced with the two material combinations for a weld time of 1.5 seconds.

Fig. 4.3 Examples of weld cross sections for the two material combinations for a weld time of 1.5 seconds (SEM montage); (a) AA6111 aluminium – DC04 steel, (b) AA7055 aluminium – DC04 steel. The box in each image designates the area used for measuring the IMC reaction layer thickness.

In the weld area, both parent metals were deformed by the sontrode tip teeth, and this macroscopic deformation was transmitted directly through the samples, resulting in a 119

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints slightly wavy interface and local variation in the contact pressure (Haddadi 2012). Therefore, In order to take into account the effect of varying contact conditions along the join line, in all samples the IMC layer thickness was averaged over the central region shown by the black box in Fig. 4.3. It can be noted that for the AA7055 alloy the sheet appeared less compressed after welding, due to its higher hardness compared to AA6111, but the cross sections of both joints still have a similar overall profile.

4.3.3 Analysis of the Joint Interfaces

The development of the IMC interface reaction layer in each sample set was investigated by electron microscopy, as a function of welding time from 0.2 to 2.0 seconds. The average thicknesses of the IMC layers are plotted in Fig. 4.4 and typical images of the joint interfaces are shown in Figs. 4.5 - 4.6.

According to the thickness distribution for the two combinations with a welding time of 1.5 seconds, it is clear that the thickness of the IMC layer varied significantly (Fig. 4.4 (a)). For DC04-AA6111 combinations, 80% of the thickness located in the range of 0.5 to 1.0 μm, with the average thickness of 0.8 μm. For DC04-AA7055 combinations, 64% of the thickness ranged between 2.0 and 3.0 μm, with the average thickness of 2.4 μm. Furthermore, it should be noted that the IMC layer was thicker and more uniform in DC04-AA7055 combinations than the other, which might be the results of higher growth rate in the former.

120

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

The interface layer developed similarly in both material combinations, although the kinetics were significantly more rapid in the case of the welds produced with the higher strength zinc containing AA7055 alloy (Fig. 4.4 ). In the initial stage of welding (<0.7 s),

121

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints individual IMC islands formed at the weld interface, with a thickness of less than 200 nm. These small, discontinuous, IMC regions could only be reliably identified using TEM, as shown in Fig. 4.5 (a) and (b). The IMC islands were ~ 0.5 -5 m wide and a single grain thick, with the grains themselves being elongated along the interface. As the welding time increased above 0.7 seconds, the IMC islands rapidly merged into an irregular continuous layer. At the maximum welding time used of 2.4 seconds, the IMC layer had still only reached 2.0 and 2.5 μm thick in the AA6111-DC04 and AA7055-D04 welds, respectively.

In the first stage of IMC development the intermetallic islands were identified in the TEM by selected area diffraction (SADP) to consist entirely of η (Fe2Al5) phase (Fig. 4.5 (b)). In the second stage, the thicker continuous IMC layer that formed became identifiable in SEM images by its different back scattered contrast to the parent materials (Fig. 4.5 (c) and (e)). At a higher resolution, for welding times greater than 1 second, a dual layer structure was identified in both material combinations that consisted of continuous layers of θ (FeAl3) and η (Fe2Al5) phase. These two distinct layers were indexed both by TEM SADP (Fig. 4.5 (c) and (e)) and high resolution EBSD (Fig. 4.5 (a)). As expected (Springer

2011 (b)), the phase with the highest iron content (η - Fe2Al5) was found to be located on the steel side and that with the highest aluminium content (θ - FeAl3) on the aluminium side of the joint interface. In both sets of welds, as the reaction layer grew thicker, the two IMC phases developed a columnar grain structure (Fig. 4.6(a)) and, although there was strong local variation, the fraction of each phase present was approximately similar.

Furthermore, the pole figures shown in Fig. 4.6 (b) reveal that η phase developed a fibre texture and grew with their [001] direction preferentially orientated in the direction of columnar growth (i.e. normal to the interface plane). The θ - FeAl3 phase also appeared to have a weak fibre texture, with its [010] direction aligned preferentially approximately normal to the interface (Fig. 4.6(c)), although it was difficult to obtain statistically valid data to confirm the texture strength.

In Fig. 4.7 results from EDX composition line-scans, obtained by TEM across the IMC interface layers, are shown after a weld time of 1.5 second. It can be seen that the measured Fe/Al ratios are consistent with the presence of two sub-layers of composition

FeAl3, and Fe2Al5, as identified by SADP. However, more importantly, the EDX profiles

122

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints show that in the AA7055- DC04 welds zinc was measured to be dissolved in both IMC phases at a concentration of approximately 1 - 2 at. %. In contrast, no other alloying element was detected with a significant concentration within the intermetallic reaction layers in any of the welds; indicating that of the all the alloying elements present, only zinc had a significant solubility in the Al-Fe IMC phases.

Fig. 4.6 (a) EBSD phase map of the dual phase IMC reaction layer seen in the AA6111- DC04 sample after a welding time of 1.5 seconds. Note; the columnar grain structures of

123

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

the IMC phases and small FeAl3 particles (red) embedded in the aluminium matrix away from the interface. In (b) and (c) pole figures are shown from the η and θ phases, respectively, where ND is normal to the interface plane.

When the welding time was increased above 1.5 seconds further differences were observed with respect to the two aluminium alloys used to produce the two sets of dissimilar welds (Fig. 4.8). In the AA6111-DC04 joints a large number of IMC particles were found embedded in the aluminium alloy matrix some distance from the weld interface (7 μm). EDX point analysis revealed that the compositions of these iron rich particles were consistent with the presence of the θ (FeAl3) phase. This was also confirmed by EBSD phase identification mapping; as can be seen in Fig. 4.5(a).

In comparison in the AA7055 - DC04 joints, two additional features were found near the join line: namely voids that appeared in the aluminium matrix close to the IMC layer (Fig. 124

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

4.8 (b)) and local regions with eutectic rich, dendritic microstructures, which are indicative of partial melting of the aluminium alloy taking place near the weld interface (Fig. 4.8 (b)). These two types of feature were found only in the central region of each weld, where the highest peak temperatures were reached (Haddadi 2012). However, far fewer embedded FeAl3 particles were found in the aluminium matrix in the AA7055-DC04 welds and no evidence could be found for incipient melting occurring in the welds produced with the AA6111-DC04 material combination.

4.3.4 Lap Shear Testing

In order to obtain the link between IMC reaction layer thickness and weld strength with the USW joints, lap shear tensile testing was used to measure the weld’s failure loads and failure energies, as a function of welding time, for the two material combinations studied. 125

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

Prior to testing, the welds were aged to T6 tempers to obtain maximum strength from the aluminium parent materials, so that the results would represent the worst case scenario in terms of the stress experienced by the weld interface.

The example load-extension curves for the two joints with a welding time of 1.5 seconds are shown in Fig. 4.9 (a). Before reaching the peak load, the DC04-AA7055 joints had a higher elastic modulus than the DC04-AA6111 joints. After reaching the peak load, the interfaces of the DC04-AA7055 joints were broken directly, while connection remained at the interface of the DC04-AA6111 joints, resulting in the stepped fall of the load on the specimen. The different fracture behaviour also resulted in the different fracture surface for the two kinds of joints, which will be discussed later.

Fig. 4.9 (a) Example load-extension curves for the two joints with a welding time of 1.5 seconds. Average (b) Failure loads and (c) fracture energies obtained from lap shear tests conducted on welds produced with both material combinations, as a function of welding time.

126

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

The failure load and fracture energies for the joints after ageing are shown in Fig. 4.9 (b, c). As the welding time increased from 0.3 s to 1.0 s, the failure load for the AA6111-DC04 joints rose slowly from 2.4 kN to 3.6 kN, then to a peak value of 3.2 kN with a welding time of 1.5 seconds. Longer welding times resulted in a progressive reduction in failure load to just below 3 kN for the longest time studied, of 2.5 seconds. The fracture energies for the AA6111-DC04 joints showed a similar pattern, peaking after a welding time of 1.5 seconds, before decreasing again at longer times. In comparison, the AA7055-DC04 joints showed a similar response, in terms of the optimum welding time, but the peak failure load was always approximately 5% lower than for the AA6111-DC04 welds and the fracture energy was less than half that for the welds produced with the lower strength aluminium alloy. In addition, the trend of the failure load and the fracture energy is similar with that of the IMC layer thickness with the welding times, indicating there might be a relationship between the peak load and fracture energy and the IMC layer thickness.

Examples of fracture surfaces from the peak strength lap shear tests, produced with a welding time of 1.5 seconds, are shown in Fig. 4.10. The macroscopic images (Figs. 4.10 (a) – (d) and (g) – (i)) confirm that interface, or near interface, failure occurred in all the welds produced with both material combinations. More detail of the fracture surfaces can be seen in the higher magnification SEM images from both the steel and aluminium sides of each joint shown in the accompanying figures. Example results of EDX composition point analysis of different areas of the fracture surfaces are also shown in Fig. 4.11. With the AA6111-DCO4 joint fracture surfaces (Figs 4.10 (b), (c), (e), (f)) the different contrast and point analysis revealed the presence of two distinct regions on each surface. The majority of the steel and aluminium sheet surfaces had an iron/aluminium composition ratio of 75/25, which suggests that the fracture occurred predominantly in the θ (FeAl3) intermetallic layer. However, a significant surface area of local patches of aluminium was also found on the steel fracture surface, (Fig. 4.10(b)), suggesting that thin pieces of aluminium had been pulled out of the aluminium sheet surface close to the interface. These regions clearly showed evidence of ductile failure and this was probably responsible for the higher fracture energy of this weld combination (Fig. 4.10(c)).

127

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

In contrast, the fracture surfaces of the AA7055-DC04 joints revealed an entirely brittle behaviour (Fig. 4.10(g -l)). With these welds no aluminium was found adhered to the steel fracture surfaces, which appeared to be entirely covered with an IMC layer of

128

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints composition consistent with the θ phase, apart from a low fraction of lighter contrast areas that EDX analysis revealed to be exposed steel, suggesting that no bonding had taken place in such regions, or the IMC layer had been pulled off the steel surface. In comparison, in these welds few IMC areas could be detected on the fracture surface of the aluminium side of the joints, suggesting that failure occurred at the interface between

θ (FeAl3) phase and stronger aluminium alloy (Fig. 4.10(k), (l)). Some areas were also detected with high zinc content, which may correspond to the local melted eutectic regions noted above.

4.4 The Development Of The IMC Layer In Fe-Al Dissimilar Joints Made By FSSW

In present study, two FSSW variants were used for different purposes. Joints made by pin- less traditional FSSW were chosen to study the formation and growth of the IMC layer and compare it to that was seen in USW; therefore, the IMC reaction behaviour could be analyzed for a raise of friction welding process. In addition, the interface in joints made by abrasion circle FSSW (ABC-FSSW) was also observed, as previously conditions have been found where no interfacial reaction occurs at the interface with this process due to the abrasion of the pin during welding (Chen 2012). Therefore, the interface for welds produced with this technique was only checked here to see if they were 'clean' so that the ABC-FSSW joints could be used for the static heat treatment studies described in chapter 5.

From precious research by Chen's (2010) study, the optimum strength for DC04-AA6111 joints made by pin-less FSSW was 2.8 kN, with a rotation speed of 1600 rpm, insert depth of 0.6 mm and welding time of 1 second. The welding time in Chen's (2010) study was fixed, and discontinuous IMC islands already formed at the interface. Therefore, in order to investigate the growth behaviour of the IMC layer, the welding time was extended to 2, 5, 8 and 10 seconds, while other parameters remained at the optimum conditions.

4.4.1 Temperature Histories In Pin-Less FSSW

The typical temperature histories, as function of welding time, measured during pin less FSSW from each joint interface are shown in Fig. 4.11 (a), and the peak temperatures reached with increasing welding times are plotted in Fig. 4.11 (b). Again, the thermal data

129

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints was only used if the position of each the thermocouple after welding was found to be in the correct position which in this case was close at 1/2 of the radius of the tool, as this is where modelling has shown the interface temperature is close to its maximum, as described in 3.4 (Jedrasiak 2012). The accuracy of the measurements is confirmed by the consistency seen in the heating stage in Fig. 4.11 (a) with the fully overlap curves with different welding times.

Fig. 4.11 (a) The temperature histories in pin-less FSSW at different welding times. (b) The peak temperatures in the joints made by ABC-FSSW at different welding times.

In FSSW the welding process started when there was contact between the pin less tool and the aluminium top sheet. As the plunge rate was 1.7 mm/s, it only took 0.35 seconds to reach the target depth (0.6 mm). Heat was generated from the friction between the tool and aluminium sheet, and the heating rate reached 450oC/s for the first 0.8 seconds, which was slower than that in USW. With increasing welding time, the temperature then increased more slowly in the following stage of welding as steady state was approached and soften of the aluminium sheet reduced the rate of plastic work. After a welding time of about 5 seconds, the interface temperature increased even more slowly with the value exceeded 480oC. After the pin-less tool was retracted with a rate of 0.8 mm/s, the temperature dropped quickly to 150oC, followed with a slower cooling rate stage.

4.4.2 Microstructure Of The IMC Layer In The Pin-Less FSSW Joints

A typical cross-section of a dissimilar joint made by pin-less FSSW with a welding time of 10 seconds is shown in Fig. 4.12. In the images in BSE mode, the bright bottom sheet part is DC04 steel (the Fe side), and the darker sheet is AA6111 (the Al side). The area inside

130

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints the black box was always chosen as a representative region for observation which was the same position as temperature measurement. These areas were selected as they were half way between the centre and edge of the interface area.

Fig. 4.12 Typical cross section of a dissimilar joint made by pin-less FSSW. The steel sheet was placed at the bottom, and the aluminium sheet was placed at the top.

The microstructure at the interface with different welding times is shown in Fig. 4.13 and 4.14. Discontinuous islands initially formed at the interface in the pin less FSSW joints with a welding time of 1 second (Fig. 4.13 a). The discontinuous islands then grew to a thin but still discontinuous layer with an average thickness of around 0.1 μm after welding for 2 seconds (Fig. 4.13 b). This thin layer was confirmed as the η phase by SADP in TEM (Fig. 4.14 b). When the welding time increased to 5 seconds, the IMC layer became thicker with an average thickness of 1 μm (Fig. 4.13 c). By TEM, the θ phase was then found to form between the η phase and the aluminium substrate by SADP and EDX line scans (Fig. 4.14 (c) and (d)). The two phase structure IMC layer then grew into the substrates in the followed welding stages (Fig. 4.13 d), and the microstructure was observed in TEM with more detail and is shown in Fig. 4.15. After a welding time of 10 seconds, the IMC layer grew to a continuous layer with a thickness of 1.5 μm. With the

EDX and SADP results in Fig. 4.15, the phase sequence Al – θ (FeAl3) – η (Fe2A5) – Fe was confirmed across the IMC layer, same to that in USW, and no other Fe-Al intermetallic compounds were found. Both phases show columnar grains in the IMC layer.

According to the thickness distribution of the IMC layer in the joints with different welding times, the IMC layer in pin-less FSSW joints was also found to be uneven in the as-welded state (Fig. 4.13 (e), (f)). With welding time increasing from 2, 5 to 10 seconds, 80% of the thickness of the IMC layer mainly located in the range of 0.1 - 0.3, 0.3 - 0.6, and 1.6 - 2.2 μm, with the average thicknesses of 0.2, 0.4 and 1.8 μm, respectively.

131

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

132

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

133

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

The area between the dashed lines in (c) is the η (Fe2A5) phase and is the θ (FeAl3) phase in (d), respectively.

4.4.3 The Interface In The Joints Made By ABC-FSSW

The cross section of the joints made by ABC-FSSW is shown in Fig. 4.16 (a). The interface was found to have no IMC layer in both SEM and TEM, which agreed with previous study with the same optimum welding conditions (Chen 2012). This clean interface without any IMC layer results from two factors. The first one is that abrasion by the pin stirs the steel surface near the interface into the aluminium sheet, resulting in particles with a ferrite core and an intermetallic skin being distributed in the aluminium substrate (Chen 2012). The second one is the low welding temperature in this process which had a peak temperature of only 400oC (Chen 2012). Consequently, this process will be only used in chapter 5 for static heat treatment studies because it can be used to produce samples with a clean initial interface without an IMC layer and strong bonding between the two sheets.

134

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

4.5 Discussion

4.5.1 The Effect Of The Alloy Elements On The Phase Composition In Dissimilar Fe-Al Joints

Results from predictions made by the CALPHAD method, using CompuTherm LCC’s

TM Pandat software with the database PanAl2013, are provided in Fig. 4.17. In Fig. 4.17 example isopleths are shown, as a function of iron content, calculated at close to the peak welding temperature of 500°C for the binary Al-Fe binary system and compositions equivalent to when DC04 steel is welded to each of the AA6111 or AA7055 aluminium alloys investigated. Several important points emerge from these calculations. Firstly, the stable composition ranges of the Al-Fe IMC phases are not greatly influenced by the additional elements present in the alloys investigated, relative to those in the binary system. Secondly, although both θ and η are predicted to be the thermodynamically stable in all cases (Figs.4.17 (a), (b) and (c)), FeAl2 and FeAl are also expected as equilibrium phases around the composition ratios of 66:34 and 52:48 (Al: Fe at.), respectively. Thirdly, the zinc present in the AA7055 alloy is predicted to have a solubility larger than 10 wt. % in the η and θ phases according to the Fe-Al-Zn ternary phase diagram (Marder 2000) and when added to the binary Al-Fe system at the levels found in the AA7055 alloy (~8 wt. %), no ternary Al-Fe-Zn phases are expected. Finally, in the calculation for the AA7055-DCO4 combination in the 500oC section a liquid phase can be seen to appear and the incipient melting point in the AA7055 alloy is predicted to be as low as around 487oC, due to the influence of zinc on the lowest melting point eutectic reaction in this system.

135

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

Fig. 4.17 CALPHAD predicted equilibrium phase fractions as a function of Fe content at 500oC for; (a) pure iron and aluminium, (b) AA7055-DC04 and (c) AA6111-DC04 steel.

4.5.2 Development of the IMC layer at the interface in joints made by USW and FSSW

In USW welding high-strain local deformation first initiates from friction between the two weld members at contacting asperities on their microscopically rough sheet faying surfaces, under the action of the clamping force and high frequency vibration, which creates a cyclic shear displacement across the weld interface (Prangnell 2011, Chen 2012 (b), Balasundaram 2014). Energy is initially dissipated by sliding friction, but galling and 'micro-welding' rapidly occurs due to local breakdown of the oxide coating at abrading surface asperities, which allows metallic bonds to be formed between the two sheets. Energy is then dissipated at a greater rate by plastic deformation (Hetrick 2009, Chen 2012 (b)), leading to a rapid temperature rise when a high applied power is used. Contact then spreads across the interface as the material softens with increasing temperature and deforms more easily under the applied pressure, so that the micro-welds increase in density and expand across the interface as welding progresses (Jahn 2007, Bakavos 2010, Shakil 2014). However, because USW involves the progressive development and

136

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints coalescence of micro-welds, a small fraction of un-welded interface is usually left at the end of the process (Bakavos 2010). For example, in this work EDX analysis revealed small exposed patches of steel on the fracture surfaces of some of the welds, suggesting that some regions remained where no bonding had taken place. In conventional, similar, USW the weld strength therefore increases with welding time as the density of microwelds increases, but also goes through an optimum because the joint strength is limited ultimately by nugget tear out and with long welding times thinning of the weld area occurs as the sontrode tips progressively sink into the sheet surfaces as the temperature rises (Jahn 2007).

When ultrasonic welding dissimilar low and high melting point materials like aluminium and steel, the aluminium alloy predominantly deforms in the welding process (Patel 2013, Balasundaram 2014, Shakil 2014), as it softens much more dramatically with increasing temperature. In the first stage of welding micro-welds are thus first formed when the oxide film breaks down locally by abrasion of asperities on the steel surface. Inter- diffusion of aluminium and iron will then occur, which can lead to the formation of intermetallic reaction products at the joint interface. The reaction layer thus develops simultaneously with the welding process and, despite being relatively thin compared to that seen in fusion welding processes, the brittle nature of the IMC layer that forms still dominated the failure behaviour seen in the lap shear tests.

With both material combinations studied the formation and subsequent growth of the intermetallic phases, produced by the reaction between ferrite and aluminium at the joint interface, behaved similarly. The IMC reaction itself can be divided into three stages. In the initial stage of welding (<0.5s), isolated IMC islands were formed on the steel side of the interface (Fig. 4.5). This first stage took place extremely rapidly when the temperature was still increasing (Fig. 4.2). In the next stage, lateral growth of the IMC islands took place until they merged to form a continuous layer with a thickness of around 0.4 μm. This occurred in the following 0.2-0.5 seconds of welding, where the temperatures approached a plateau value of around 500oC. In the final stage, the now continuous IMC layer continued to grow into both metals, but more rapidly into the aluminium, and a second intermetallic phase formed as a continuous sub-layer on the aluminium side of the joint, leading to the dual layer structure seen in the welds produced with the longest

137

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints welding times. In addition, in the welds produced for longer times with the AA6111 material the thicker reaction layers became damaged by the USW process (Fig. 4.8(a)), resulting in dispersal of IMC particles in the softer aluminium side of the joint, and in the welds made with the lower melting point AA7055 alloy the high interface temperatures reached (>500oC Fig. 4.2) caused some incipient melting at the joint interface.

In the early stage of joining discontinuous IMC islands were observed to form heterogeneously in local areas along the joint interface, and thus developed within the micro-weld regions that first formed when welding was initiated. The first intermetallic compound to nucleate within these reaction centres was identified to be η (Fe2Al5) phase (Fig. 4.5 (b)). As welding continued, the increasing density of micro-welds expands the area of metallurgical contact and this would contribute to a more uniform development of reaction islands over the whole weld region. When combined with lateral spreading of the reaction centres, this would quickly lead to the development of a continuous IMC layer at the weld interface (Fig. 4.5(c)). Lateral growth of the IMC islands along the joint interface would also be facilitated by interfacial diffusion and the high strain rate deformation experienced by the aluminium alloy under USW conditions that has been shown to increase its vacancy concentration (Haddadi 2012).

In pin less FSSW, inter-diffusion of aluminium and iron would also be expected to occur after a strong bond was built at the interface between the work-pieces during welding. This again would require break down of the surface oxide layer. However, the growth behaviour of the IMC layer in the pin-less FSSW joints was similar to that seen in the joints made by USW, but with a different growth rate. Isolated IMC islands also formed firstly at the interface with a short welding time (1 second, Fig. 4.13). With welding time increasing, the bonding area spread across the whole interface, resulting in an increased density of the IMC islands (1-2 seconds, Fig. 4.13). In the next step, the IMC islands grew mainly along the interface and merged with each other to form a continuous layer with a thickness of about 0.5 μm (2-5 seconds). Finally, the continuous layer grew into both metals in the following welding process.

The formation behaviour of the IMC layer was related to the welding process. In FSSW with a pin less tool, deformation of the work-pieces firstly occurs under the edge of the

138

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints tool due to the fastest local line speed, together with the most rapid heating rate in the process (Prangnell 2010, Chen 2010). Strong bonding then forms at the interface between the work-pieces under the edge of the tool as a result of the friction between the two weld members. With welding time increasing, more heating is introduced by the friction between the tool and top sheet with a slower rate, and the top sheet becomes soft at high temperature. The deformation zone is then transferred to the centre of the welding area in zones together with bonding area, due to the rotation of the soften material coupled with the tool under the sticking conditions. A sound joint is welded when the strong bonding forms in the whole interface area (Prangnell 2010, Chen 2010).

Although the growth behaviour of the IMC layer in the joints made by the two techniques was similar with each other, the thickness distribution of the IMC layer along the interface was not the same in the joints made by the two welding methods. In Fig. 4.18, IMC layer thickness in DC04-AA6111 joints made by both USW and pin-less FSSW was plotted against the distance from the weld centre. The thickest part was always found in the weld centre in USW joints, while it was at the position with a distance of 2.0 mm away from the welding centre in the pin-less FSSW joints. This different distribution of the IMC thickness was related to the temperature field in the joints during welding. This different thickness distribution should be related to the temperature field in the joints, and the FE temperature field modelling results during the two welding processes are shown in Fig. 4.19 (Jedrasiak 2012, Jedrasiak 2015).

With the similar the parameters for the geometry and materials, the FE temperature field models in the two welding methods were calculated by Jedrasiak et al. (2012, 2015). In their work, the experimental data was collected from Haddadi (2012) and Panteli (2012) for the model with USW; for the model with pin-less FSSW, the temperature data was collected at the interfaces of work pieces, with the similar temperature measurement method in the present study (Jedrasiak 2012). The welding conditions in both welding methods were similar with those in the present study, and the dimensions of the work pieces were same for the same type of metal in the present study. During the calculation in their work, the net input power was adjusted empirically with a piece-wise linear variation, with a time step size of 0.05-0.1 seconds. In addition, all the surfaces in contact with the air were treated as insulated, as a result of the low heat transfer coefficient to

139

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints air and short welding times. Besides, heat generation was simulated as a uniform surface heat flux at the interface of the welding metals, whose intensity was change as a function of welding time. Finally, the modelling was built to predict the temperature field in each joint with different welding times. The validation of the two FE models was checked to fit the experimental results in each welding process quite well (Jedrasiak 2012, Jedrasiak 2015).

Fig. 4.18 IMC layer thickness against the distance from the weld centre in the (a) USW joints and (b) pin-less FSSW joints with different welding times.

In the temperature field for USW, most heat was generated in the centre of the weld, and the highest temperature is mainly distributed in the aluminium substrate. This temperature field is related to the deformation of work-pieces during welding. The deformation was mainly happened in aluminium alloy as the softer side, and most deformation zone was found to be very close to the interface as a result of the high

140

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints frequency vibration at the interface between the two sheets (Bakavos 2010, Patel 2013, Balasundaram 2014, Shakil 2014). Therefore, the thickest IMC layer was found in the centre of the interface.

However, the temperature field was more complicate in the pin-less FSSW. As the heat was mainly generated from the interface between the aluminium sheet and pin-less tool, the highest temperature is found in the aluminium sheet near this interface. Furthermore, it is clear that the formation of the welds was from the edge to the centre due to the different line rotation speeds. Therefore, the temperature field was sensitive to the radial distribution of the power input, with a peak at radius of 1.5 mm in the case of a fluted tool with a radius of 5.0 mm was used (Reilly 2012, Jedrasiak 2012). As a result, the thickest IMC layer was found at the interface with a distance of 2 mm from the weld centre.

Once the IMC islands merged, and a continuous reaction layer had formed, TEM and EBSD investigation revealed the development of a second IMC layer consisting of θ (FeAl3) phase on the aluminium side of the joint (Fig. 4.5(d, f), Fig 4.6(a)). For both material combinations, these two phases were the only intermetallic compounds detected, which is consistent with previous studies of Al-Fe weld couples (Springer 2011 b, Naoi 2007). CLPHAD calculations (Fig. 4.15) showed that these compounds are the most stable phases predicted within the composition ratios (at.) of Al to Fe of 71:29 to 74:26 and 75:25 to 78:22, for η and θ respectively, but they are not the only stable IMCs in the Al-Fe system.

At the weld temperature the FeAl2 and FeAl phases were also predicted to be stable around the composition ratios (at.) of Al to Fe of 66:34 and 52:48, respectively. However, these phases have only been previously observed in dissimilar Al-steel diffusion couples after long term heat treatments (Cheng 2011), so it can be concluded that they did not form under the rapid USW and FSSW conditions used in the present work for kinetic reasons.

The observation that η is normally the first phase seen to nucleate in Al-steel dissimilar friction welds is probably partly because aluminium has a higher diffusion rate in Fe than Fe in Al, leading to ferrite at the contact surface becoming more rapidly supersaturated in aluminium, than the aluminium in iron, which would favour a phase first forming that has

141

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints a higher Fe to Al ratio. However, it has also been suggested that this maybe because η has a crystallographic habit orientation with ferrite which would favour a lower activation energy for nucleation (Naoi 2007). It has also been proposed that more rapid growth of η can occur because it has a high density of vacant sites within its crystal structure along [001] direction (Springer 2011 b), which would become particularly important when it forms a continuous layer across the interface as this will then effectively act as a diffusion barrier. The present work also shows that the η phase grew with its [001] direction preferentially aligned normal to the interface, as has been previously reported (Bouayad 2003, Wang 2010), which is parallel to the diffusion flux across the interface layer (Fig. 4.6 (b, c)). The fact that both the η and θ phases formed a fine columnar grain structure (Fig. 4.6 (a)) will also further increase the flux of atoms through the IMC layer by grain boundary diffusion.

As a continuous IMC phase thickens, the concentration gradients driving diffusion through the layer will reduce and at some critical point it will become more favourable for a phase with a different stoichiometry to nucleate (Marder 2000, Hirose 2003). The next phase observed to develop was θ (FeAl3), which formed on the aluminium side of the joint in both material combinations, owing to its higher aluminium to iron ratio. This phase has also been observed in other studies on dissimilar welding of aluminium and steel (Springer 2011 b).

In the AA7055-DC04 combination the base aluminium alloy had a relatively high zinc content of 8 wt.%. Consistent the observations, thermodynamic modelling has shown that no ternary Al-Fe-Zn compounds are expected at the zinc level, although zinc was detected by EDX in the TEM to dissolve into the η and θ phases at a similar level of 1-2 at. % (Fig. 4.8(c)). A range of ternary phases can form at higher zinc ratios, but these are predicted to be well outside of the range of the zinc level present in a conventional aluminium alloy (Marder 2000). The average IMC layer thickness was, however, observed to be thicker in the welds produced with the zinc-containing AA7055 aluminium alloy (Fig. 4.4). The difference in peak temperatures reached between these two sets of welds was very small (< 1%). However, due to uncertainty in the temperature measurements and damage to the IMC layer disproportionally affecting the layer thickness in the case of the AA6111 and DC04 welds, long duration isothermal treatments at 450°C were also

142

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints performed on samples pre-welded for 1.5 seconds to study the effect of zinc in more details, and the results of this work will be shown in Chapter 5.

When the AA6111 and DC04 sheets were welded in USW for longer than 1.5 seconds a large number of small fragments of θ phase were found distributed in the aluminium alloy near the interface. The high cooling rates in the welding cycle suggests that it this is unlikely that these particles could have formed by precipitation. This behaviour can therefore be attributed to the high frequency cyclic shear that is imposed across the weld interface in USW, which damaged the outermost IMC layer as it grew into the aluminium material near the weld interface (Haddadi 2012). In high power ultrasonic spot welding of dissimilar metals, material flow and extensive deformation is known to occur in the softer weld member within a highly deformed layer near the weld line (Bakavos 2010). This intense deformation will fracture IMC grains that protrude ahead of the transformation front and incorporate the debris in to the aluminium sheet. As few IMC particles were found in the AA7055-DC04 joints, this suggests that less damage occurred to the reaction layer in this weld couple.

The difference in damage to the IMC layers seen in the two sets of materials welded by USW is related to the observation of local melting in the AA7055-DC04 USW welds, which will limit shear transfer at the join line and thus reduce fracture of IMC grains growing in the reaction layer. The melting point of the AA7055 alloy was predicted to be about 487°C. Thermocouple measurements have shown that at the weld centre this temperature was exceeded within a welding time of 1 second and reached nearly 530°C after a welding time of 2 seconds. The presence of a liquid phase could also have significantly affected the IMC growth rate in the AA7055-DC04 welds at long weld durations.

4.5.3 The Different Growth Rates Of The IMC Layer During The Two Welding Processes

The growth rates of the IMC layer was found to be different in the dissimilar DC04- AA6111 combinations made by the two solid-state welding methods studied (Fig. 4.20 a). In general, the growth rate was faster in the joints made by USW than by pin-less FSSW as can be seen in Fig. 4.20. According to the lap shear test results, the optimum welding time for a sound joint made by USW was 1.5 seconds, when the thickness of the IMC layer is 1.2 µm. However, it took longer time to reach the similar IMC layer thickness in joint

143

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints made by pin-less FSSW (1.5 µm after a welding time of 10 seconds). From the temperature histories in the welding area in joints made by the two welding process (Fig. 4.21 a), it can be seen that both the heating rate and the peak welding temperature are higher than in USW in pin-less FSSW, which are 1000oC/s and 550oC in USW and 450oC/s and 480oC in pin-less FSSW, respectively (Fig. 4.21). Therefore, it is reasonable to assume that the main reason the IMC layer has a faster growth rate in the joints made by USW than by pin-less FSSW is due to the more rapid heating rate that occurs in USW.

Furthermore, less welding energy was needed to form the same thickness IMC layer in USW than in pin-less FSSW. The welding energy was obtained from the welding machine after a specific welding time. For example, it required 2300 J to form a 0.8 µm thickness IMC layer in USW, but 4100 J to form a 0.9 µm thickness IMC layer in pin-less FSSW (Fig. 4.20 b). The reason for this is the different way energy is dissipated in the two welding processes.

The temperature fields for both USW and FSSW were FE modelled in DC04-AA6111 welds made by the two welding methods by Jedrasiak (2012, 2015) in Fig 4.19. It can be seen from these example simulations that the heating area is smaller but more focused at the interface during USW than that in pin less FSSW. Therefore, less welding energy is wasted in the USW process. Consequently, it requires more welding energy to form the same thickness IMC layer in pin-less FSSW than in USW.

144

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

4.5.4 Relationship Between Interfacial Microstructure And Mechanical Properties Of The USW Joints

Without achieving a high energy nugget pull-out fracture behaviour a spot weld will not be fit for purpose in a safety critical application. From the above discussion, it is apparent that when ultrasonic welding dissimilar metals, that can form IMCs by inter-diffusion and metallurgical reaction, there is competition between the formation and spread of microwelds across the joint interface, which increases the bond strength, and the coalescence and thickening of the IMC reaction layer, which will encourage a brittle interface failure (Haddadi 2012).

In the welds produced in this study failure always occurred close to or through the weld interface. Previously it has been found that when ultrasonic spot welding medium strength aluminium alloys, like AA6111 to steel sheet, with the same welding set-up to that used here, it was possible to obtain pull-out failures under optimum welding conditions (Haddadi 2012). However, in this work the welds were shear tested with the aluminium sheet in a weaker T4 temper condition and lower weld strengths (2.8 kN),

145

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints were achieved than have been obtained in the present study. In this prior investigation it was also found that the failure process reverted to interface failure at longer welding times when the IMC layer became thicker (Haddadi 2012). Surprisingly, the weld failure energies for the optimum (pull-out) condition in this earlier work were only 25% greater (~ 4 kN∙mm (Haddadi 2012)) than those measured for the AA6111-DC04 welds in the current investigation, and this reflects the lower strength condition of the aluminium sheet used previously, as well as the fact that (as will be discussed below) the failure process seen here in this material combination was not purely through the brittle IMC layer.

When measured after ageing the aluminium alloys to stronger T6 tempers, in the present study the welds showed failure loads of over 3 kN at an optimum time of 1.5 seconds (Fig. 4.10). In comparison, the highest fracture energy obtained for the AA6111-DC04 combination was 3.2 kN∙mm, whereas for the AA7055-DC04 welds it was only 1.7 kN∙mm. For the AA6111 alloy this is about 40% of that measured for the equivalent similar aluminum-aluminum welds, which exhibit nugget pull-out by the more energy intensive process of ductile tearing (Haddadi 2012).

The lower failure energies seen in the present study, and in particular with the AA7055- DC04 welds, are due to crack propagation predominantly through a brittle IMC weld interface. It should be noted that in the prior study, which tested the AA6111 alloy in a weaker T4 condition, fracture also occurred along the interface in the IMC layer when it grew thicker at longer welding times (Haddadi 2012). The higher strength of the aluminium materials tested here has therefore had the effect of tipping the balance back in favour of interface failure, owing to the higher loads that are transmitted across the joints before yielding occurs. In the present work fractography (Fig. 4.10) suggests that fracture occurred mainly through the IMC layer or at the interface between θ phase and aluminium matrix.

The influence of the IMC layer thickness on the peak load and fracture energy is shown in Fig. 4.22. An optimum IMC layer thickness of 1.3 μm can be found for both the peak load and fracture energy in the two joints made by USW. In the both joints, the highest peak load is found to be around 3.2 kN with the optimum IMC layer thickness, and reduced

146

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints slowly with increased IMC layer thickness, as the thickest IMC layer was still less than 5 μm in the present study, which was reported as the critical thickness (Sprinter 2011 b, Tanaka 2009). However, the highest fracture energy in DC04-AA6111 joint is about twice of that in DC04-AA7055 joint with the optimum IMC layer thickness. This result indicates that the peak load in the dissimilar joints is mainly controlled by the thickness of the IMC layer, while the fracture energy is also controlled by the base metals besides the thickness of the IMC layer.

Fig. 4.22 (a) The peak load and (b) fracture energy against the IMC layer thickness in DC04-AA6111 and DC04-AA7055 joints.

In the fractured AA6111-DC04 samples a significant total area of small patches of aluminium was, however, found adhered to the steel sheet, which in places covered up to 45% of the fracture surface. At these locations the fracture surfaces showed evidence of the ductile tearing out of thin pieces of aluminium from near the join line. This behaviour was probably encouraged by the high density of brittle θ phase particles embedded in the AA6111 side of the joint, due to the partial break-up of the IMC layer in these samples. Indeed, the height of the aluminium patches (~ 7 μm) was measured to be similar to the depth the brittle θ phase particles were seen embedded within the aluminium sheet when the IMC layer partially broke up during the welding process (Fig. 4.8(a). This mixed fracture behaviour is thus responsible for the surprisingly high failure energies seen for the AA6111-DC04 welds, despite the fracture path following quite close to the weld interface.

In contrast, despite being produced with a much stronger parent material, the AA7055- DC04 welds showed a slightly lower strength in the lap shear tests and a much lower

147

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints fracture energy of less than half that of the AA6111-DC04 samples. This difference is clearly associated with the more brittle interface fracture seen in the AA7055-DC04 weld samples, when lap shear tested (Fig. 4.10) which may be related to the thicker reaction layer found in this system. However, the much stronger AA7055 aluminium sheet would also transmit a higher load to the joint interface in these welds. In addition, it is possible in the AA6111-DC04 samples the outer layer of θ IMC grains, that were more prone to fracture/debonding in the AA7055-DC04 welds, were already selectively dispersed into the aluminium matrix by the greater damage induced to the IMC layer by the USW welding process. For the AA7055-DC04 welds, it is also likely that the local melting behaviour and associated porosity seen at the join line reduced the interface strength.

4.6 Summary And Conclusions

The present study has investigated the influence of different aluminium alloys and welding methods on the formation and growth stages of the intermetallic compounds (IMCs) formed at the interface in joints produced between a DC04 steel and two different aluminium alloys (AA6111 and AA7055) by high power ultrasonic spot welding (USW) and friction stir spot welding with a pin-less tool (pin-less FSSW).

By optimizing the welding time, attractive joint strengths could be achieved by USW. However, no suitable process window was found that could produce joints with a nugget pullout failure mode, using the current welding procedure. In all cases the fracture path was still through the weld interface region. A purely brittle fracture, at the interface between the aluminium alloy and the IMC reaction layer, was seen in the AA7055-DC04 welds, whereas a higher energy mixed mode failure was found in the AA6111-DC04 combination.

In both material combinations and with both welding processes a similar sequence of

Intermetallic compound formation was observed during the solid state welding. η-Fe2Al5 was found to be the first IMC phase to nucleate, within a welding time of only 0.3 seconds (in USW), as isolated islands heterogeneously distributed at the weld interface. These early reaction centres probably formed within the micro-weld regions that first develop at initial asperity contact points across the joint interface. The IMC islands then spread to form a continuous layer in both material combinations, by a welding time of 0.7 seconds

148

Chapter 4 Weld Formation and Interface Layer in Fe-Al Dissimilar Joints

(in USW). With longer welding times a second IMC phase, θ-FeAl3, was seen to develop on the aluminium side of the joints.

In AA6111-DC04 joints made by pin-less FSSW, the IMC layer was found to have similar growth behaviour to that seen in the joints made by USW. However, compared with USW, a slower heating rate and lower peak temperature were found in the welding area during the pin-less FSSW process. As a result, the IMC layer at the interface in USW joints had a higher growth rate and cost less weld energy than those in pin-less FSSW joints. In addition, the reaction layer grew more uniformly in USW, compared to that in FSSW, where it started near the edge of the shoulder and spread inwards, due to the different temperature distribution in each process.

The reaction layer in the AA7055-DC04 welds was found to grow thicker than in the AA6111-DC04 samples and this may be related to the dissolution of zinc into the IMC layer. However, thermodynamic modelling showed that no zinc containing IMC compounds are expected in the welds produced with the AA7055 alloy, and the effect of zinc will be discussed further in Chapter 5.

149

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

CHAPTER 5 THE STATIC GROWTH BEHAVIOUR OF INTERMETALLIC COMPOUNDS IN STEEL AND

ALUMINIUM DISSIMILAR WELDS

5.1 Introduction

In this chapter, the formation and growth behavior of the interface reaction, or intermetallic compound (IMC) layer, was investigated during static heat treatment in three groups of dissimilar Fe-Al diffusion couples. A low-carbon steel (DC04) and two commercial aluminium alloys (AA6111 and AA7055) were welded together by abrasion circle friction stir spot welding (ABC-FSSW) and ultrasonic spot welding (USW) to create Fe/Al diffusion couples; USW joints under the optimum welding conditions and ABC-FSSW joints with an IMC free clean interface. The conditions used, in each case, were rotation rate of 800 rpm, plunge depth of 0.1 mm and welding time of 1 s for ABC-FSSW, and a nominal applied power of 2.5 kW, pressure of 29 MPa and a welding time of 1.5 s for USW (see example 4.3). In chapter 4, it has been demonstrated that a continuous IMC layer formed at the interface, with an average thickness of 1.2 µm in the USW joints, and the phase sequence across the interface was confirmed to be Fe – η (Fe2Al5) – θ (FeAl3) – Al, while no IMC layer was found at the interface in the ABC-FSSW joints. DC04-AA6111 joints made by the two welding methods were used to find the effect of the initial state on the growth behaviour of the IMC layer, while DC04-AA7055 combinations were only welded by USW to study the effect of alloy elements. These experiments were also carried out so that the controlled conditions could be used to measure the reaction rates and fit kinetic growth laws to the data. This enables the activation energies for the reactions to be compared to those published in the literature.

The dissimilar couples were annealed for various times of up to 128 hours between T=400 and 570oC, with the parameters shown in Table 3.3, to allow a systematic determination of the growth kinetics and data to be generated for modelling the IMC layer growth behaviour. At the same time, the microstructure of the IMC layer, including grain structure, thickness, and morphology, was observed by high resolution techniques, like scanning electric microscopy (SEM) and transmission electric microscopy (TEM). The phases in the IMC layer were identified by select area diffraction pattern (SADP) and energy dispersive X-ray spectroscope (EDX) in TEM and SEM.

150

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

5.2 The Influence Of The Initial State On The IMC Layer During Heat Treatment

5.2.1 The Microstructure Development Of The IMC Layer In The ABC-FSSW Joints During Annealing

In the annealing process for the ABC-FSSW joints, several stages of the growth of IMC layer were seen in Fig. 5.1. In the as-welded state, no visible IMC layer was detected at the interface (Fig. 5.1 a), which agreed with the results of Chen et al.'s (2012). After annealing at 400oC for 8 hours, some separated IMC particles formed at the interface (Fig 5.1 b). With increasing annealing time, the IMC particles grew both along and normal to the interface, together with the development of an increase in density of the particles at the interface (Fig 5.1 c). Finally, the IMC particles merged into a continuous layer, with the irregularly fringed interface between both sides (Fig. 5.1 d). The continuous layer then grew into both substrates with increasing thickness (Fig. 5.1 e). The EDX analysis from the line across the IMC layer show that only Fe, Al and Si were detected in the IMC layer, and Si only appeared in the IMC layer adjacent to the aluminium side of the joint (Fig. 5.1 f).

The IMC layer was identified as the η (Fe2Al5) phase, from the content of the Fe and Al.

Some Fe-Al IMC particles were also found in the aluminium matrix, which are highlighted with the white arrows in Fig. 5.1 (a) - (e). The EDX point scan results show that the content of Fe in these particles was in the range from 20 at. % to 28 at. %, suggesting they contained both the η (28.5 at. %) and θ (25 at. %) phases. These particles were the results of the welding process, as the steel surface was abraded by the rotating pin, and some steel particles were embedded into the aluminium sheet during welding (Chen 2012). During the heat treatment, diffusion also occurred between the steel particles and aluminium sheet, resulting in the Fe-Al IMC particles.

TEM samples were prepared by FIB from the ABC-FSSW joints after annealing to obtain more detail of the IMC layer. The results are shown in Fig. 5.2, 5.3 and 5.4. The crack and hole were produced in the TEM sample preparation process.

151

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

Three interesting features were observed at the interface in the sample after annealing at 400oC for 64 hours (Fig. 5.2 a). Island shape IMC particles with an average thickness of 3 μm were still found at the interface. The IMC layer was confirmed to be the η phase

(Fe2Al5) by SADP and EDX (Fig. 5.2 (a) and (b)), and no other Fe-Al compound was

152

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints detected. Columnar grains were found near the steel substrate, with an aspect ratio of 2:1 and average width of 0.2 μm (Fig. 5.2 b). Thin remnants of steel were also found inside the IMC layer, which are highlighted by the solid and dotted-dashed arrows in Fig. 5.3 a. The second area was a thin discontinuous IMC layer with an average thickness of 0.4 μm (Fig. 5.2 (a) and (c)). This thin IMC layer was also confirmed to be the η phase by SADP in Fig. 5.2 e and EDX result in Fig. 5.2 c, with a Fe content of 28 at. %. Some IMC free interface points were also found at the interface, such as the point highlighted by the dashed arrows. The last feature was the small particles distributed in the aluminium substrate near the interface. The EDX result shows that only Fe and Al could be detected in these particles with an atomic ratio of 20:80 (Fig. 5.2 (d)).

Fig. 5.2 Microstructure for the IMC layer in the ABC-FSSW welded joint after annealing at 400oC for 64 hours. (a) STEM image of the interface with EDX analysis along the line

153

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints indicate. (b) Grain structure of the η phase adjacent to the steel substrate, in b position in (a), with the SADP taken from the dashed circle. (c) STEM image of the thin IMC layer at the interface, in position c in (a), with a dashed line as the interface between the DC04 and IMC layer, with the SADP taken from the dashed circle. (d) EDX map from the box in (a). (e) EDX line scan result from the black line in (a).

TEM images for the IMC layer in the ABC-FSSW joint after annealing at 550oC for 40 minutes are shown in Fig. 5.3. At this higher temperature a continuous layer with an average thickness of 5 μm had formed at the interface. From the EDX results and SADP shown in Fig. 5.3 (b) - (d), again only the η phase was identified in the IMC layer. The grain structure in the IMC layer was found to be related with their position. Columnar grains with an aspect ratio of 4:1 were found adjacent to the steel substrate, with an average width of 0.4 μm. The aspect ratio of the grains reduced to 2:1 in the middle of the IMC layer and was 1:1 adjacent to the aluminum alloy substrate, and the average widths were 0.3 μm and 0.2 μm, respectively. A crack was seen to across several columnar grains (Fig. 5.3 (c)), indicating that this crack formed during the sample cooling and handling stage.

154

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

After annealing at a higher temperature still 570oC for 2 hours, the IMC layer grew to a continuous layer with a thickness of 20 μm, and the microstructure is shown in Fig. 5.4.

155

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

The layer was found to contain some cracks within it (Fig. 5.4 (a)). The crack shown in Fig. 5.4 (a), between the aluminium sheet and the IMC layer, was produced in the cooling stage probably due to the different thermal expansion coefficients between the materials. A TEM sample for the IMC layer near the aluminium substrate was prepared to observe the microstructure (Fig. 5.4 (b), (c) (d) and (f)). From the result of EDX and SADP, the η phase was still the only phase found in the IMC layer. Columnar grains constituted the major part of the IMC layer, while some equiaxed grains were distributed among them.

In the EBSD phase map of the ABC-FSSW joint, after annealing at 500oC for 30 minutes, quite a complicated grain structure was seen in the IMC layer and the substrate (Fig. 5.5).

Fig. 5.5 (a) Grain structure and (b) colour coded EBSD phase mapping of the ABC-FSSW joint after heat treatment at 500oC for 30 minutes. Yellow- ferrite, Red- η, Green- aluminium.

Two types of grain could be found in the IMC layer, which were columnar grains with an aspect ratio of 2:1 and average width of 0.4 μm, and fine equiaxed grains inside the IMC layer with an average size of 0.3 μm. The areas containing these two kinds of grains were distributed in the IMC layer at the interface. In the steel substrate, fine compressed

156

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints ferrite grains were also found next to the IMC layer, with an average grain size of 0.5 μm. These deformed grains indicate that the severe plastic deformation occurred during the welding process.

5.2.2 The Microstructure Of The IMC Layer In The USW Joints After Annealing

In the USW joints, a continuous IMC layer with both the η and θ phases had already formed in the welding process (Chapter 4). Therefore, the IMC layer grew into the substrates in the following annealing treatment. One example is shown in Fig. 5.6. After annealing at 500oC, the IMC layer became thicker when the annealing time increased from 10 to 30 minutes. However, the IMC layer was still not uniform after annealing for 30 minutes (Fig. 5.6 c) in some places. Such an uneven IMC layer was also observed in previous studies with short annealing times or low annealing temperatures (Naoi 2007, Springer 2011 b).

Fig. 5.6 Microstructure development during annealing of the IMC layers in the USW joints after heat treatment at 500oC for (a) 0, (b) 10 and (c) 30 minutes. (d) the EDX line scan 157

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints results along the arrow in (c). The dashed lines are the interfaces between the four phases shown in the EDX result.

After annealing, both the η and θ phases layer became thicker in the IMC layer with increased time and could be analysed by EDX in SEM (Fig. 5.6 d), together with SADP in the TEM and EBSD phase mapping in the following results. Therefore, it was easy to collect separate thickness data for the two phases to enable analysis of their individual growth kinetics.

Fig. 5.7 (a) The IMC layer at the interface of the USW joint in the as-welded state; (b) and (c) the IMC layer at the interface after annealing at 500oC for 10 minutes, with different

158

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints magnifications. (d) High resolution images for the θ phase in the white dashed box in (c). (e) and (f) are diffraction patterns for the η and θ phases, respectively.

The microstructure of the IMC layer observed by TEM from the USW joints with 1 μm thick continuous reaction layer, prior to annealing is shown in Fig. 5.7 (a), and TEM images and EBSD mapping for the joints after annealing are shown in Fig. 5.7- 5.10 at 500oC for the different annealing times. In the as-welded state, two sub-layers had already formed at the interface, and the IMC phases were identified as the η and θ phases, respectively (Fig. 5.7 (a)). Both the phases show columnar grains, with a growth direction normal to the interface. The aspect ratios of the η and θ phases were 2:1 and 3:1, and the average widths were 0.2 and 0.15 μm, respectively. The average thickness of the IMC layer was 0.9 μm.

After annealing at 500oC for 10 minutes, the average thickness of the IMC layer grew to 2 μm. The grains of the two IMC phases were found to have similar columnar grain structure, with an aspect ratio of 3:1 and average width of 0.2 μm. Fine grains of the IMC phases could be found at both the Fe- η and θ-Al interfaces, with different nucleation sites (Fig. 5.7 (c)). At the Fe- η interface, new η grains formed inside the η layer. At the θ- Al interface, however, θ phase was found to nucleate firstly at the grain boundaries of the aluminium alloy (black arrow in Fig. 5.7 (c)). High resolution TEM images were taken of the θ phase on an [10-1] zone axis, and two crystal planes were identified by the interplanar distances: (101) with 8.08 Å, and (010) with 8.12 Å.

When the annealing time increased to 20 minutes, significant growth of the IMC layer occurred, and the average thickness in this sample reached at 3.5 μm (Fig. 5.8). For the θ phase, the average grain width grew to 0.18 μm, with an aspect ratio in the range of 2:1 to 4:1. At the θ –Al interface, fine θ phase grains with an average size of 0.1 μm appeared at the aluminium side. For the η phase, columnar grains appeared more in the middle near the θ phase, with an average width of 0.3 μm, and fine grains with an average size of 0.2 μm were seen adjacent to the steel substrate. Furthermore, these fine grains were found to grow into the steel sheet and follow the growth direction of the η phase. A similar growth behaviour was found in the USW joint after annealing for 1 h (Fig. 5.9). The

159

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

η phase showed a faster growth rate than the θ phase; their average thicknesses increased from 2.5 μm and 0.7 μm to 3.5 and 1.1 for the η and θ phases, respectively.

160

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

When the annealing time increased to 2 hours, the η phase became much thicker than the θ phase, while grain growth occurred in both phases (Fig. 5.10). Cracks formed easily in the thick IMC layer, especially in the η phase as it comprised the major part of the layer (Fig. 5.10). It should be noted that the crack between the θ phase and aluminium substrate occurred in the sample preparation. However, the crack seen in the η phase was probably generated due to the different thermal expansion coefficients between the IMC layer (mainly the η phase) and the substrates (Masahashi 2005, Wang 2014) in the air cooling stage following annealing.

Fig. 5.10 (a) grain structure and (b) colour coded EBSD phase map of the interface in the USW joint after annealing at 500oC for 2 h. Yellow- ferrite, Red- η, Blue- θ, Green- aluminium.

5.2.3 The Development Of The Grain Structure In The IMC Layer

The grain size in each phase in the IMC layer was measured by TEM and EBSD in the two types of weld after annealing. Some typical EBSD band contrast maps and grain size data are shown in Fig. 5.11 for the different joints after annealing. In particular, in Fig. 5.11 (a)

161

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints and (b) are from the ABC-FSSW joints after annealing at 500oC for 30 minutes, (c) and (d) from the ABC-FSSW joints after annealing at 550oC for 1 hour, (e) and (f) from the USW joints after annealing at 500oC for 2 hours. For the columnar grains, the width of the grains was measured as the average grain size, because this dimension is most relevant to the effect of the grain size on diffusion through the layer. It should be noted that the θ phase only appeared in the USW joints after annealing at 500oC for 2 hours (Fig. 5.11 e and f), while all the other areas were in the η phase.

162

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints annealing at 550oC for 1 hour, (e) and (f) from the USW joints after annealing at 500oC for 2 hours. The width of the columnar grains was measured as the average grain size.

Apparently, the grain size distribution in all cases belonged to the bimodal distribution, with two peaks located at 0.1 μm and 0.3 μm. This distribution indicates that fine grains with grain size of 0.1 μm, which are the equiaxed grains in band contrast maps, accounted for the largest number, followed by the columnar grains with grain size of 0.3 μm.

When the IMC layer was thin (3 μm IMC layer in Fig. 5.11 (a) and (b)) the grain structure was quite complicate. The grain size and morphology were changed along the interface. For example, in Fig. 5.11 (a), both columnar grains with large aspect ratio and equiaxed grains were found in the IMC layer. In addition, in the middle of the interface, columnar grains were found near the steel side, equiaxed grains were observed near the aluminium side.

When the IMC layer became thicker, the grain structure became more uniform along the interface and changed across the interface. In the case of 10 μm IMC layer, three areas were found across the interface (Fig. 5.11 (c) and (d)). Equiaxed grains were found next to both substrates, and columnar grains existed in the middle of the IMC layer. In the case of the 17 μm thick IMC layer, produced by annealing aft 500oC for 2 hours (Fig. 5.11 (e) and (f)), columnar grains were found next to the steel and in the middle of the IMC layer, while equiaxed grains were found next to the aluminium and near a crack.

The grain size in each phase after annealing was plotted in Fig. 5.12. It can be seen that significant grain growth occurred during annealing and the grain growth of the η phase was similar in the two joints at the same annealing temperature, while the grains grew faster for the θ phase than the η phase with increasing annealing temperature. The grain growth model will be built in Chapter 6 with the measured data.

163

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

5.2.4 The Growth Kinetics Of The IMC Layer

The thickness distribution of the IMC layer in the joints after annealing was also calculated (Fig. 5.13). In general, the thicker IMC layer became more uniform with the increasing annealing time; for example, for the pin-less FSSW joints after annealing at 500oC for 4 hours, over 70% of the IMC layer thickness located in the range of 10 - 11 μm, with the average thickness of 11 μm. However, the pre welding conditions were found to affect the uniform of the IMC layer. With a continuous IMC layer at the interface before annealing, the USW joints were found to have a more uniform IMC layer than the pin-less FSSW joints. For example, after annealing at 500oC for 1 hour, the average IMC layer thicknesses in both joints were around 8 μm, but only 38% of the thickness located in the range of 7 - 9 μm in the pin-less FSSW joints, while the percent was 65% in the USW joints. Furthermore, 20% of the thickness was over 10 μm in the pin-less FSSW joints, compared to 0% in the USW joints. Apparently, this was the result of the formation of the island shape particles at the interface in pin-less FSSW joints (Fig. 5.1-5.2).

The average thicknesses of the IMC layers in the dissimilar DC04-AA6111 joints made by the two welding methods after annealing are shown in Fig. 5.14 (a). As discussed in the literature review 2.5, a parabolic law is suitable to describe the growth of the IMC layer in dissimilar Fe-Al combinations, and equation 2-2 was used to fit the thicknesses of the IMC layer at each temperature. The experimental data is shown as the points and the fitted results, according to equation 2-2, are shown as the straight lines. In particular, the fitted results are shown as dashed lines for the USW joints and solid lines for ABC-FSSW joints in

164

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

Fig. 5.14 (a). It should be noted that when the annealing temperature was above 500oC, the USW joints after annealing were broken in the air cooling stages, even after a very short annealing time like 10 minutes. Therefore, the growth data at 550oC and 570oC were only collected in the ABC-FSSW joints in the following growth kinetics calculation.

Fig. 5.13 The distribution of the IMC layer thickness in (a) pin-less FSSW joints and (b) USW joints after annealing at 500oC for different times.

165

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

In general, the parabolic law fitted well to the results in both cases. For the two joint types the overall layer growth rates were close to each other at each temperature, with values of 4.0 × 10-15 m2/s at 450oC and 1.7 × 10-14 m2/s at 500oC in ABC-FSSW joints, but the IMC layers in the USW joints started thicker than those in the ABC-FSSW joints. As a result, the dashed lines and solid lines are parallel to each other at each temperature. The different thicknesses but similar growth rates are related to the interface condition in the as-welded state, as a continuous IMC layer had already formed at the interface in the USW joints, resulting in the different start points for the growth of the IMC layers in the two joints (Fig. 5.14 a). However, the above is surprising, given that the θ phase was not detected in the ABC-FSSW samples, but was seen in the USW specimens.

In the USW joints, as it contained two phases, the thicknesses of both the η and θ phases were measured, and the results are shown in Fig. 5.14 (b). This indicated that the η phase has a much faster growth rate than the θ phase; for example, the growth rates of the η and θ phases were 1.5 × 10-14 m2/s and 5.8 × 10-16 m2/s at 500oC, respectively. Therefore, the η phase formed the major part of the overall IMC layer, and the growth rate of the IMC layer in USW joints was mainly controlled by the η phase as well in the ABC-FSSW joints.

The growth rates of the IMC layer were calculated and compared from the experimental data in present study and previous studies at the same annealing temperature (Shibata 1966, Springer 2011 a), and the results are shown in Fig. 5.15 (a) and (b). The data in the present study used here is from the results of ABC-FSSW joints, as there was no IMC layer formed in the chosen studies before heat treatment. Pure aluminium and pure iron were selected in the research of Shibata (1966), and pure aluminium and the same low carbon steel (DC04) were used in the research of Springer (2011 a). In Fig. 5.15 (a), traditional Fe/Al diffusion bonding couples were used to collect the growth kinetics data, while the DC04-pure Al joints were pre welded with FSSW by Springer et al. (2011 a). The thicknesses of the IMC layers are compared in Fig. 5.15 (b).

166

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

The growth rates measured in the present study were higher than the growth rates in diffusion bonding couples at a similar temperature, but comparison was more complicated with the results of pre-welded joints. At low temperatures (500oC), the growth rate in the present study was similar with Springer’s result, but at high temperatures, the growth rate in the present study at 570oC was even slightly higher than that in Springer’s (2011 a) study at 600oC.

After calculating the growth rate (k) at each annealing temperature, the activation energy for each phase in the IMC layer was calculated, according to equation 2-3. The calculated results are shown in Fig. 5.16. In the ABC-FSSW joints, the average activation energy of the IMC layer (η phase) was 160 kJ/mol across the whole annealing temperature arrange (400oC - 570oC), and this value agreed well with the activation energy of the η phase in the USW joints after heat treatment (169 kJ/mol). However, there was an inflection at 500oC, and the activation energy increased from 116 kJ/mol between 400oC - 500oC to 248 kJ/mol between 500oC - 570oC. Moreover, the activation energy for the θ phase was calculated as 199 kJ/mol, which is larger than the 169 kJ/mol found for the η phase. This is reasonable given the slower growth rate of the θ phase.

167

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

o o the straight-dashed lines with Q1 and Q2 show the fit between 400 C - 500 C and 500 C - 570oC in Fig. 5.3 a.

5.3 The Influence Of Zinc On The IMC Layers During Heat Treatment

In order to investigate the influence of zinc on the growth behaviour of the IMC layer, the DC04-AA7055 joints were annealed together with the DC04-AA6111 joints to keep the joints under the same heat treatment conditions. Both joints were pre welded by USW with a welding time of 1.5 seconds, and the microstructure was observed in Chapter 4. As the incipient melting point in the AA7055 alloy was predicted to be around 487oC, the heat treatment temperature was set as 450oC. After heat treatment for 2 and 4 hours, the interface microstructure was observed by SEM and EBSD. As the interface in the DC04- AA6111 joints was observed in 5.2, only the EBSD maps for the DC04-AA7055 joints were shown in Fig. 5.17. However, the thickness and grain size for the η and θ phases in both joints are plotted against the annealing time and shown in Fig. 5.18. It should be note that before heat treatment, the θ phase in both joints was thicker than the η phase.

168

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

From the colour coded EBSD phase maps (Fig. 5.17 (a) and (c)), the θ phase is found to be thicker than the η phase. The growth rate of the θ phase is similar with that of the η phase in the DC04-AA7055 joints, and both of them are even higher than the growth rate of the η phase in the DC04-AA6111 joints (Fig. 5.17 (a)). From the grain structure of the IMC layers (Fig. 5.17 (b) and (d)) and the grain growth behaviour (Fig. 5.18 (b)), smaller grain sizes were found in both phases in the DC04-AA7055 joints.

Fig. 5.18 (a) Thickness and (b) grain size of each phase in the two Fe/Al dissimilar couples pre welded by USW after annealing at 450oC for 2 and 4 hours. The growth trend is shown

169

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints as the solid lines and dashed lines for each phase in DC04-AA7055 and DC04-AA6111 welds after annealing, respectively.

5.4 Discussion

5.4.1 The Growth Direction Of The IMC Layer In Dissimilar Fe/Al Couples After Heat Treatment

The growth direction of the IMC layer was first confirmed by measuring the average distance of the layer from the original interface position into the substrates, using the method described in section 3.8. The dissimilar Fe/Al couples used here was the DC04- AA6111 joints pre welded by ABC-FSSW, with the parameters of a rotation rate of 800 rpm, plunge depth of 0.1 mm and welding time of 1 s. The Fe/Al couples were then heat treated at 500oC for 10 – 60 minutes. One example of these results is shown in Fig. 5.19.

The annealing time for this sample was 30 minutes in Fig. 5.19. Fig. 5.19 (a) and (c) show the microstructures at the same sample in the as-welded condition and after annealing, respectively, while (b) and (d) show higher magnification images from the dashed boxes in (a) and (c). The original interface position was identified using four micro hardness indentations and is shown by the white dashed lines in Fig. 5.19 (b), (d). The average growth distance of the IMC layer was calculated to be 2.1 and 7.8 μm from the original interface to the steel side and aluminium sides of the joint, respectively. Therefore, the growth rate ratio of the IMC layer into the steel side relative to the aluminium side of the weld was approximately 1:4.

The reason for the different growth rate into the two substrates is the much higher diffusion coefficient of the aluminium atoms in iron and through the IMC layer than that of the iron in aluminium (Naoi 2007, Kajihara 2006), resulting in a higher growth rate of the IMC layer into the aluminium alloy. This growth direction agreed with the results from the Springer's (2011 a) research, with both the DC04- Al-5wt.% Si couples and DC04-pure Al couples during annealing at 600oC for different times.

170

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

5.4.2 Discussion Of The Development Of The IMC Layer During Heat Treatment

Two welding methods were applied to produce the joints with a different initial interface condition. The joints with a continuous IMC layer at the interface were welded by USW with a welding time of 1.5 seconds, and the joints without any prior interface reaction were welded by ABC-FSSW with a welding time of 1 second. Both kinds of joints were then annealed at T=400 - 570oC for different times. The microstructure development of the IMC layer in each type of weld was observed by SEM, TEM and EBSD, and the phase composition was identified by EDX, SADP and EBSD.

The growth behaviour of the IMC layer was observed by TEM in both ABC-FSSW and USW joints annealed under several combinations of temperature and time. In the ABC-FSSW joints, the η phase formed first at the interface at the nanometres scale and grew rapidly

171

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints into columnar grains normal to the interface, as a result of the higher aluminium vacancy sites along the [001] direction in the η phase (Burkhardt 1994). However, nucleation only occurred locally and scattered η phase islands first formed at the interface, with IMC free interface between each island (Fig. 5.1). These η phase islands became a barrier for the diffusion of the both aluminium and iron atoms, resulting in the formation of new nucleation sites at the edge of the large η phase islands at the interface (Fig. 5.2). New η phase islands formed on these new sites in the followed annealing process and grew faster than the pre-existing islands due to the larger local concentration gradient. Finally, with increased density along the interface, the islands merged with each other, resulting in the formation of a continuous IMC layer at the interface. With a further increased annealing time, the IMC layer grew into both the steel and aluminium substrates, with the growth ratio of 1:4. A similar growth behaviour has been found in the Cu6Sn5 layer between Sn-3.5Ag solder alloy and Cu substrate, from the island shape to a layer type (Choi 2000).

The appearance of the coarsely spaced η phase islands in the ABC-FSSW joints with short annealing time or low annealing temperature is related to the different nucleation behaviour of the η phase along the interface. Two possible reasons could be used to explain locally different growth rates. Firstly, the η phase had a preferred growth direction along c axis, and the growth rate of the η phase was, therefore, sensitive to the grain orientation (Burkhardt 1994). As a result, η phase grains with the c axis normal to the interface had a faster growth rate than other grains. This was also the reason for the appearance of columnar grains in the η phase islands. In addition, this preferred growth direction affected the morphology of the η phase layer in the following annealing, which was not uniform even with a thickness of 35 μm after annealing at 600oC for 8 hours (Springer 2011 b). Furthermore, during heat treatment, this preferred growth direction resulted in the wave-like interface between the η phase and the substrates in solid dissimilar Fe/Al diffusion couples (Naoi 2007, Springer 2011 b), and the 'finger-like' shape seen growing into an iron substrate after heat treatment above the melting point of aluminium by Bouche (1998) and Tang (2012).

The second reason for the sensitivity of the IMC formation to position could be the uneven interface condition in the ABC-FSSW joints. In the FSSW process, welding

172

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints temperature was also not uniform at the interface and sensitive to the distance away from the pin (Prangnell 2010, Chen 2010). The further reason could be that the joint formation occurred by heterogeneous across the interface. Micro-welds would form firstly at the position with highest local pressure and then spread across the whole interface (Prangnell 2010, Chen 2012). Therefore, the contact condition of the dissimilar Fe/Al joints made by ABC-FSSW was not uniform along the interface, resulting in the uneven nucleation site density for the η phase.

In the growth process of the IMC layer, the growth behaviour was different at the interfaces with the two substrates. At the IMC-Al interface, new IMC grains were found to form at the grain boundaries of the aluminium alloy (Fig. 5.7 e), and fine IMC grains were found near the interface in the aluminium alloy in the USW joints after the IMC layer grew to a thickness of 3 μm. Consequently, the growth of the IMC layer into the aluminium substrate involved the formation and growth of the new IMC grains at the aluminium side, of either the η phase in the ABC-FSSW joints or the θ phase in the USW joints. At the IMC- Fe interface, columnar η phase grains were always found next to the Fe substrate, and some grains were found to grow in a finger like fashion into the Fe grains (Fig. 5.8). Therefore, it was reasonable to conclude that the growth of IMC layer into the Fe substrate was a result of the growth of the η phase into the ferrite grains.

The IMC phases were identified in the two joints after annealing with different combinations of temperature and time by EDX, SADP and EBSD phase mapping. From the results, the phase compositions in the two kinds of joints after annealing were different with each other. In ABC-FSSW joints, only the η phase (Fe2Al5) was detected in the IMC layer in the present study, which agreed with the previous studies on dissimilar Fe/Al couples with no IMC layer formed before heat treatment (Naoi 2007 with pure Fe - pure Al diffusion couples, Springer 2011 (a) with DC04- pure Al diffusion couples). However, both the η phase and θ phase (FeAl3) were already present in the USW joints before annealing, and this result agreed with the study on the dissimilar Fe/Al joints at high temperature (Springer 2011 a, with DC04-pure Al couples pre welded by FSSW at 600oC - 675oC).

173

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

In the solid state annealing temperature range, the appearance of the θ phase was related to the initial state of the dissimilar Fe/Al couples before heat treatment. Schematic diagrams of the growth behaviour of the IMC layer at the interface in the three stages of annealing are shown in Fig. 5.20. The island shape η phase particles formed first at the interface (Fig. 5.20 a), and the particles merged with each other to form a continuous layer (Fig. 5.20 b). In this stage, the continuous layer played a role as the diffusion barrier for aluminium and iron atoms, and the concentration gradient was reduced across the barrier layer. Therefore, a second phase with higher aluminium content, θ, formed between the η phase and aluminium substrate.

However, another condition was also necessary for the formation and growth of the θ phase. According to the activation energy and growth rate, it is more difficult for the growth of the θ phase than the η phase at 500oC. As a result, enough input energy was required for the θ phase. In USW process, however, both the η and θ phases formed at the interface at the solid state temperature. This might because high deformation was also induced in the welding process, which provided energy for the formation of θ phase.

In all cases, whether in the present study or in the previous studies listed above, the η phase was always the major part in the IMC layer. According to the growth kinetic calculations in the present study, the θ phase was found to have a higher activation energy than the η phase, with values of 199 kJ/mol and 169 kJ/mol, respectively. In addition, the high density of the aluminium vacant sites along the c axis in the η phase increased the diffusion rate of aluminium atoms across the η phase to be react with iron (Burkhardt 1994). Therefore, the η phase had a faster growth rate than the θ phase, with

174

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints values of 9.6 × 10-15 m2/s and 5.8× 10-16 m2/s at 500oC, respectively. Consequently, it was reasonable that the θ phase only made up a minor part of the IMC layer, or even did not form at the interface.

The reason why only the η and θ phases formed at the interface was related to the growth kinetics of the Fe-Al intermetallic compounds. According to the Fe-Al phase diagram (Fig. 2.28), β' (FeAl) and ζ (FeAl2) are also stable intermetallic compounds in the Fe-Al system as well as the η and θ phases at T= 400 - 600oC (Desai 1986). However, it requires specific conditions for the formation of the β' and ζ phases, such as high temperature and long annealing times, and the consumption of aluminium substrate (Cheng 2008, Windmann 2013). For example, in the work of Cheng et al. (2008), both phases formed only after an aluminium coating was entirely consumed by the η phase, and ζ phase formed after annealing at 750oC for 1 hour, and the β' phase formed at 750oC for 10 hours (Cheng 2008). Apparently, the β' and ζ phases could not form in the conditions seen in the present study. In addition, Naoi and Kaijhara (2007) concluded that the interdiffusion coefficient is more than two orders of magnitude smaller for the β' and ζ phases than that for the η phase at T = 823–913 K.

The alloy elements in the present alloys have been found to affect the phase composition only when their content was high enough (Springer 2011 a, Luo 2012). For example, several Fe-Al-Si ternary phases were seen to form in DC04-Al 5wt.% Si couples by Spinger el al. (2011 b)after annealing at 600oC for different times, but the content of Si was much higher than that detected in the IMC layer in the present study, such as 20 at. % in the τ6

(FeAl4.5Si) phase or 10 at. % in the τ5 (Fe2Al8Si) phase (Springer 2011 a). Therefore, although Si was detected in the IMC layer, the content was not high enough to form Fe- Al-Si ternary phases.

As Zn was the only element detected to have a content higher than 1 at.% in the IMC layers in the DC04-AA7055 welds made by USW (Fig. 4.6, 4.8), the influence of Zn on the growth of the IMC layers was investigate in the DC04-AA7055 welds after heat treatment. From the results (Fig. 5.16, 5.17), the θ phase was found to be thicker than the η phase in the DC04-AA7055 welds after heat treatment at 450oC for 2 and 4 hours. Compared with the IMC layers in the DC04-AA6111 welds, the growth rate of the θ phase was speed up

175

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints significantly, while the growth rate of the η phase was almost close to each in both welds. As the grain sizes measured for each phase in both welds were not quite different from each other, the effect of the grain size was not the main reason for the thicker IMC layers observed in the DC04-AA7055 welds than those in the DC04-AA6111 welds.

Two kinds of cracks were found in the as-annealed combinations. The cracks at the interface between the θ phase and aluminium alloy were produced in sample preparation. Kirkendall pores (Fig. 5.1) were seen produced at this interface due to the faster diffusion rate of the aluminium into iron relative to that of the iron atoms into aluminium (Kajihara 2006), resulting in a weak bond at this interface, which caused it to break in sample preparation. The second type of cracks were seen in the η phase, and were caused by the thermal expansion coefficient mismatch between the IMC layer (mainly the η phase) and the substrates. The thermal expansion coefficients are 12.2 × 10-6 K-1 for Fe, 23.5 × 10-6 K-1 for Al, 18.9× 10-6 K-1 for the η phase, and 19.7 × 10-6 K-1 for the θ phase (Masahashi 2005, Wang 2014). The average thickness of the IMC layer in the samples seen with cracks was found to be similar to that in the samples seen without cracks and both fitted the parabolic growth law. Therefore, it can be concluded that the cracks were produced in cooling the sample after annealing and had little influence on the growth of the IMC layer.

5.4.3 Discussion Of The Development Of The Grain Structure In The IMC Layer

The grain morphology of the IMC layer changed in different positions with different thickness (Fig. 5.11). In the thin IMC layer (3 μm), the grain structure was not uniform along the interface, and both the columnar grains and equiaxed grains appeared with a similar fraction. In the thicker IMC layer (10 μm), a more uniform grain structure was seen in the IMC layer along the interface, with columnar grains in the middle of the IMC layer and equiaxed grains near the substrates. When the η layer became even thicker (17 μm), columnar grains were also found in the IMC layer near the steel side, with a reduced aspect ratio in the position closer to the aluminium, and the equiaxed grains only appeared near the aluminium side.

The columnar grains have also found in the η layer in the results from previous studies, with the aspect ratio reduced at the position closer to the aluminium (Springer 2011 b, Qiu 2009, Bouayad 2003, Bouche 1998). Furthermore, the aspect ratio of the columnar

176

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints grains became larger in thicker IMC layers, which were annealed for longer times or using higher annealing temperatures, and a typical finger-like shape or tongue-like shape morphology has been seen formed in the IMC layer near the Fe side when reacted with liquid aluminium (Bouayad 2003).

The different grain morphologies of equiaxed and columnar might be related to the different growth stages of the grains. After observation of the η phase grain structure in the DC04-pure Al couples after annealing at 600oC for different times (Springer 2011 a), proposed grains with small aspect ratio, or even equiaxed, grains formed first (average aspect ratio was 2:1 for 1 hour), followed by the formation of columnar grains with large aspect ratio near the Fe side (average aspect ratio was 4:1 for 8 hours). The columnar grains develop as the result of the preferred growth direction along the c axis which results in the texture shown in Chapter 4 (Fig. 4.6) (Burkhardt 1994). Therefore, in the mixed structures seen in the thin layers the grains with a large aspect ratio probably formed earlier than the grains with small aspect ratio or the equiaxed grains.

As a result, the reason for the different grain structure in the IMC layers with different thicknesses is clear. The areas with large columnar grains seen in thin continuous η layer (Fig. 5.11 a) developed from the island shape IMC particles in the earlier stage (Fig. 5.2), and the areas with small equiaxed grains (Fig. 5.11 a) formed later than the previous areas, and therefore had a less growth time. After heat treatment with enough time, the grain structure of the η phase became coarse and more uniform in the thick IMC layers along the interface. However, as the IMC layer grew into both substrates, the areas near the substrates formed later than the middle of the IMC layer, resulting in the smaller grain size and aspect ratio been on either side of the layer. With annealing time further increasing, the growth rate slowed down due to the barrier effect of the thick IMC layer on the diffusion of iron and aluminium atoms, resulting in less nucleation near the substrates. Therefore, on average larger grains were seen in the thicker IMC layer.

5.4.4 The Growth Kinetics Of The IMC Layer During The Heat Treatment

After the observation of the microstructure of the IMC layer in the joints after heat treatment, growth kinetic data was collected; such as the thickness and grain size of each phase in the IMC layer. As the growth behaviour of each phase in all the cases fitted the

177

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints parabolic law very well, the growth rate was calculated with equation 2-1 or 2-2, and the activation energy of each phase was then pointed out with equation 2-3.

The growth rate of the IMC layer was mainly controlled by the η phase in both joints (Fig. 5.13). In the USW joints, although the θ phase also appeared and grew during heat treatment, the growth rate was much slower than that of the η phase, with the value of 5.8× 10-16 m2/s and 1.5 × 10-14 m2/s at 500oC, respectively. The faster growth rate of the η phase was related to the high density of aluminium vacancies along the c axis (Burkhardt 1994). Therefore, the growth of the θ phase had little influence in the thickening process of the IMC layer during heat treatment. As a result, although the IMC layer was always thicker in the USW joints than ABC-FSSW joints due to a continuous IMC layer formed before heat treatment, the growth rate of the IMC layer was close to each other at each temperature, with values of 4.0 × 10-15 m2/s at 450oC and 1.7 × 10-14 m2/s in the USW and ABC-FSSW joints at 500oC.

As Si was the only alloy element detected in the IMC layer in the DC04-AA6111 joints, besides Fe and Al, it was worth considering the influence of Si on the growth of the IMC layer. Although Si has been found to depress the growth ratio of the IMC layer from the work of Springer et al. (2011 b), the growth rate of the IMC layer in a DC04-pure Al couples was close to that in the DC04-AA6111 (Al-Mg-Si) joints (Fig. 5.14 b) (Springer 2011 a). According to the previous studies, Si could reduce the growth rate of the IMC layer by either occupying the large number of aluminium vacancies on the c-axis ([001] direction) of the η phase (Yin 2013), or the formation of Fe-Al-Si ternary phases with slower nucleation and growth rate than η phase (Marder 2000, Springer 2011 b). In both cases, the content of the Si should be significant in the η phase or in the ternary phases. For example, in Springer’s (2011 b) study, the content of Si was 20 at. % in the τ6 (FeAl4.5Si) phase, and Coburn (1964) suggested an optimum Si addition of 8.5-9.5 wt. % to reduce the IMC layer thickness. However, only a small content of Si (<4 at. %) was detected in the IMC layer near the aluminium substrate, and no ternary phase was characterized in the IMC layer. The reason for this was the low Si content in the alloy investigated (0.82 wt. % in AA6111) and Si was mainly distributed as precipitates in the AA6111. As a result, the effect of Si on the growth kinetics of the IMC layer in DC04-AA6111 joints was limited.

178

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

Pre welding was found to affect the growth rate of the IMC layer by comparison to the present study with previous studies, in Fig. 5.14 (Shibata 1966, Springer 2011 a). In general, the growth rate in the pre welded joints was faster than that in clamped diffusion couples that were not welded prior to heat treatment. One reason for this is that the pre welded joints already had full interface contact and a metallurgical bond compared to only local contract in a traditional diffusion couples. The pre welding conditions used in both joints were selected from the optimum parameters, while the diffusion couples were just polished and then forced together by pressure and creep during heat treatment. Another reason was the deformation in the substrates after pre welding, resulting in the deformed fine grain structure in the substrates after welding. The fine grains, especially in steel side, induced higher grain boundary density, and therefore improved the diffusion rate of aluminium and iron atoms (Haddadi 2012, Chen 2012, Springer 2011 b).

Using the growth rate constant k measured at each temperature, the activation energy of each phase in the IMC layer was calculated. In the ABC-FSSW joints, the average activation energy of the IMC layer (η phase) was 160 kJ/mol across the whole annealing temperature arrange (400oC - 570oC), which agreed with the activation energy of the η phase in USW joints (169 kJ/mol). However, there was an inflection at 500oC, and the activation energy increased from 116 kJ/mol between 400oC - 500oC to 248 kJ/mol between 500oC - 570oC. This change of the activation energy might be related to the different contribution of grain boundary diffusion and lattice diffusion of the η phase at different temperatures.

The calculated activation energies of the η phase are compared to those from different studies in Table 5.1 (Fig. 5.15). It is quite strange that the activation energies of the η phase are located in the range between 30 and 30 kJ/mol in the different studies. As pure Al-pure Fe couples were used in several studies with different activation energies, the alloy elements were not the main reason. As only lattice diffusion was considered in the previous studies, it is reasonable to take the effect of the grain boundary diffusion in the calculation. The relationship between the two diffusion mechanisms was introduced with equations 2-6 2-8. Qualitative analysis will be discussed in the following discussion, while quantitative analysis will be induced in a diffusion model in Chapter 6.

179

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

Table 5.1 Activation energies of the η phase from previous studies

Activation energy Temperature Materials (kJ/mol) (oC) Bouayad et al. 74.1 800 Pure Al- pure Fe (2003) Denner et al. 170-195 771 Pure Al- Mild steel (1977) Eggeler et al. 34-155 786 Pure Al-Low carbon steel (1986) Naoi et al. 281 600-650 Pure Al- pure Fe (2007) Springer et al. Pure aluminum- DC04 190 600 (2011 b) (Diffusion) Tang et al. 123 680-770 Pure Al- pure Fe (2012) Present study 160-169 400-570 DC04-AA6111

The growth of the IMC layer in dissimilar Fe/Al couples was controlled by both the grain boundary diffusion and lattice diffusion, and the calculated activation energy was a combination of Qgb and Ql. The activation energy of grain boundary diffusion (Qgb) is lower than that of lattice diffusion (Ql), resulting in the different diffusion coefficients of the IMC phases in the substrates for the two diffusion mechanisms (Wang 2015).

However, the contribution of grain boundary diffusion coefficient (Dgb) to the effective diffusion coefficient (Deff) is related to the grain size (Belova 2004, Bokstein 2004, Wang

2015). According to equations 2-6 – 2-8, the contribution of Dgb in Deff, (1-g) Dgb, is reduced with grain growth, while the contribution of the lattice diffusion coefficient (Dl) in Deff, gDl, is almost a constant for the IMC phases with the changed grain size. Therefore, the (1-g) Dgb takes a larger fraction in Deff in the case of IMC phases with smaller grain. In addition, grain growth of the IMC phases always occurred during heat treatment; for example, the average grain size of the η phase was 0.3 µm at 500oC for 2 hours and 0.4 µm at 500oC for 8 hours in the ABC-FSSW joints. Therefore, there is a competition between (1-g) Dgb and gDl during heat treatment, as the grain size is controlled by both the annealing time and temperature (Belova 2004, Heitjans 2005, Wang 2015).

180

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints

In the present study, the average grain size of the η phase was generally smaller at 400- 500oC than that at 500-570oC in the ABC-FSSW joints (Fig. 5.12), resulting in the larger fraction of (1-g)Dgb in Deff at low temperatures. Therefore, the influence of the grain boundary diffusion was more significant at 400-500oC than that at 500-570oC. As a result, the calculated activation energy was closer to Qgb than Ql. A more accurate calculation of the effective activation energy and grain growth will be included in chapter 6.

The competition between (1-g)∙Dgb and gDl during heat treatment can also explain the different activation energies in previous studies, which were listed in Table 5.1. Besides other factors like alloy elements, the annealing temperature and time were quite different in each study, resulting in the possibility of different grain size in the IMC layer. However, due to the limitation of research condition, the grain size could only be measured in a few studies. For example, in Springer’s (2011 b) study, grain growth occurred significantly with increased time and temperature, from 1 μm at 600oC for 1

o hour to 5 μm at 675 C for 30 seconds. As a result, the fraction of (1-g) Dgb and gDl in the effective diffusion coefficient were related to the annealing conditions, and the calculated activation energies of the IMC layer in Table 4.2 were the combination of Qgb and Ql, with different fractions in each study.

The USW joint growth kinetic data was collected at 400-500oC, and no inflection point found. The activation energies of the η and θ phases were calculated as 169 and 199 kJ/mol, respectively. Again, both calculated activation energies include a combination of

Qgb and Ql for each phase. The calculated activation energy of the η phase was close to the value (160 kJ/mol) in ABC-FSSW joints, resulting in similar growth rate of the IMC layer in both joints. The difference between the activation energies of the two phases was only 30 kJ/mol, but the growth rate of the η phase was much larger than that of the θ phase, indicating that the growth rate was not only determined by the activation energy, but also related to other factors, such as the crystal structure.

5.5 Summary and Conclusions

Two welding methods were applied to produce the joints with different initial interface conditions. The joints with a continuous IMC layer at the interface were welded by USW with a welding time of 1.5 seconds, and the joints without a continuous IMC layer were

181

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints welded by ABC-FSSW with a welding time of 1 second. Both kinds of joints were then annealed at T=400 - 570oC for different times. The microstructure of the IMC layer was observed by SEM, TEM and EBSD, and the phase composition was identified by EDX, SADP and EBSD. Kinetics growth data was collected from the microstructure results, and the growth behaviour of the IMC layer was found to fit the parabolic growth law. Consequently, the growth rate and activation energy for each phase in the IMC layer were calculated with equations 2-1. 2-2 and 2-3.

In general, the IMC layer grew into both the substrates, but more quickly into the aluminium alloy than steel, with a growth rate ratio of approximately 4:1. The growth of the IMC layer was found to fit the parabolic law very well.

The phase composition was different in the two joints. The η (Fe2A5) was found to take the major part of the IMC layer in both joints, while the θ (FeAl3) only appeared in USW joints with a continuous IMC layer produced in the pre welding process. The appearance of the θ phase had two prerequisites; a continuous η phase formed at the interface and enough input energy.

The effect of alloying elements on the growth of IMC layer was different. The effect of Si was not significant until the content in aluminium alloy was higher than a specific value. Zn was found to accelerate the growth of the θ phase and therefore improve the growth of the IMC layers in heat treatment.

Grain growth occurred in both phases in the heat treatment significantly, and was found to affect the calculated activation energy by the grain boundary diffusion. A grain growth model will be built with the measured data in Chapter 6.

The growth rates of the IMC layer were controlled by the growth of η phase and close to each other in the two joints, with the activation energies of 160 kJ/mol in ABC-FSSW joints and 169 kJ/mol in USW joints. The growth rate of the θ phase was slower than that of the η phase at each temperature, with a higher activation energy of 199 kJ/mol.

The contribution of the lattice diffusion and grain boundary diffusion was found to be related to the grain size, which was a function of annealing time and temperature. The

182

Chapter 5 The Static Growth Behaviour of IMCs in Fe-Al Dissimilar Joints grain boundary diffusion took a larger fraction with smaller grain size at lower temperature and shorter time. Quantitative analysis will be introduced in Chapter 6.

183

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

CHAPTER 6 MODELLING OF THE IMC LAYER GROWTH IN AN FE-AL DISSIMILAR SYSTEM

6.1 Introduction

In this chapter, a double-IMC phase diffusion model was used to attempt to predict the growth behaviour of the IMC layer in Fe-Al dissimilar couples, during both heat treatment and in the welding process. A schematic illustration for applying the model to predict the thickness of the IMC layer is shown in Fig. 6.1. A grain growth model was firstly built to predict the grain growth of each phase. As both the grain boundary diffusion and lattice diffusion occurred in the growth of the IMC layer, the activation energy (Q) and pre- exponent factor (D0) for the two diffusion mechanisms in the two phases, η (Fe2Al5) and θ

(FeAl3), were calculated from experimental data, to predict the inter-diffusion coefficients for the two diffusion mechanism in each phase, Dgb and Dl. Therefore, the effective diffusion coefficients for each phase could be calculated. Finally, the double-IMC phase diffusion model was used to predict the thickness of the IMC layer with the given temperature and time in the heat treatment or the temperature history in ultrasonic spot welding (USW). The modelling results were compared with the experimental data to check the validation of the model.

The model was not only used to predict the thickness, but also to investigate the factors that could affect the growth behaviour of each phase. In particular, the effect of grain boundary diffusion and grain growth was examined by calculating the contribution of the two diffusion coefficients, gDgb and (1-g)Dl, to the effective diffusion coefficient, Deff. The

184

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System effect of the measured temperature and welding time on the growth of each phase was also used to see how accurately IMC growth could be predicted in the welding process.

6.2 The Double-IMC Phase Diffusion Model

Based on the single IMC phase growth model developed by Kajihara (2004, 2006), a double-IMC phase growth model was developed by Wang et al. (2015). As both the grain boundary diffusion and lattice diffusion occur, the effective diffusion coefficients of the η and θ phases were used in the model to predict the thickness of each phase in the IMC layer in present study. Diffusion data in Fe-Al system, like the diffusion coefficients of Fe in η and θ phase, were collated from Richards (1994) and Kajihara (2006).

In the model, the initial boundary conditions are shown in Fig. 6.2, and the specific values were obtained from the EDX line scan results shown in Fig. 5.9.

The initial boundary conditions are expressed as the equations 6-1:

cAl (x, t=0)= cAl0 (6-1 a)

cFe (x, t=0)= cFe0 (6-1 b)

The boundary conditions of each phase after heat treatment can be described as: 185

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

cAl (x= zθAl, t>0) = cAlθ (6-1 c)

cθ (x= zθAl, t>0) = cθAl (6-1 d)

cθ (x= zηθ, t>0) = cθη (6-1 e)

cη (x= zηθ, t>0) = cηθ (6-1 f)

cη (x= zFeη, t>0) = cηFe (6-1 g)

cFe (x= zFeη, t>0) = cFeη (6-1 h)

Here, the values of c are the concentrations of each phase in mol pre unit volume.

The positions of the three interfaces can be described as functions of the annealing time t as equations 6-2 (Wang 2015):

푧Feη = 퐾Feη√4퐷퐹푒푡 = 퐾ηFe√4퐷η푡 (6-2 a)

⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡푧ηθ = 퐾ηθ√4퐷η푡 = 퐾θη√4퐷θ푡 (6-2 b)

⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡푧θAl = 퐾θAl√4퐷θ푡 = 퐾Alθ√4퐷Al푡 (6-2 c)

Here, t is the annealing time,퐷퐹푒 , 퐷η , 퐷θ and 퐷Al are the effective inter-diffusion coefficients of which include grain boundary diffusion and lattice diffusion in the ferrite, η,

θ and Al phases, respectively, and 퐾Feη, 퐾ηFe,⁡퐾ηθ,⁡퐾θη,퐾θAl and 퐾Alθ are dimensionless coefficients. All of the K values can be calculated from the equations 6-3 (Wang 2015):

퐶휂퐹푒 − 퐶퐹푒휂 =

퐶퐹푒0−퐶퐹푒휂 2 퐶휂퐴푙−퐶휂퐹푒 2 푒푥푝 [−(퐾퐹푒휂) ]⁡+ 푒푥푝[−(퐾휂퐹푒) ] (6-3 a) 퐾퐹푒휂√휋[1−erf(퐾퐹푒휂)] 퐾휂퐹푒√휋[erf⁡(퐾휂퐹푒)−erf⁡(퐾휂퐴푙)]

퐶θ휂 − 퐶휂θ =

퐶θ퐴푙−퐶θ휂 2 퐶휂퐹푒−퐶휂θ 2 푒푥푝 [−(퐾θ휂) ] + 푒푥푝[−(퐾휂θ) ] (6-3 b) 퐾θ휂√휋[erf⁡(퐾θ휂)−erf⁡(퐾θ퐴푙)] 퐾휂θ√휋[erf⁡(퐾휂퐹푒)−erf⁡(퐾휂θ)]

퐶퐴푙θ − 퐶θ퐴푙 =

퐶θ휂−퐶θ퐴푙 2 퐶퐴푙0−퐶퐴푙θ 2 푒푥푝 [−(퐾휂퐴푙) ] + 푒푥푝[−(퐾퐴푙θ) ] (6-3 c) 퐾θ퐴푙√휋[erf⁡(퐾θ휂)−erf⁡(퐾θ퐴푙)] 퐾퐴푙θ√휋[1+erf⁡(퐾퐴푙θ)]

186

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

The K values were calculated from the equations 6-1 – 6-3, with the values of aluminum concentration, effective diffusion coefficients of each phase and annealing times. The positions of the three interfaces were then determined by equations 6-2, and the thicknesses of both the η and θ phases calculated using equations 6-4:

푙η = 푧Feη − 푧ηθ = (퐾ηFe − 퐾ηθ)√4퐷η푡 (6-4 a)

푙θ = 푧ηθ − 푧θAl = (퐾θη − 퐾θAl)√4퐷θ푡 (6-4 b)

Furthermore, the growth rate k of each phase could then be calculated from equations 6- 5:

2 푘η = 4퐷η(퐾ηFe − 퐾ηθ) (6-5 a)

2 푘θ = 4퐷θ(퐾θη − 퐾θAl) (6-5 b)

6.3 The Grain Growth Model

In general, the grain growth kinetics for the equaxied grains in bulk metals in 3D follows equation 6-6 (Thompson 1990):

푛 푛 퐿2 − 퐿1 = 푘(푡2 − 푡1) (6-6)

Here, L1 is the average grain size at the time t1, L2 is the average grain size at the time t2, n is the exponent, generally expected to be 2, and k is the grain growth rate and depended on the temperature.

For the columnar grain structure seen in the IMC phases in the present study (Fig. 5.5 and Fig. 5.10), equation 6.6 can also be applied when the average grain width is used as the average grain size, and the grain size is smaller than the thickness of the IMC layer (Thompson 1990). In this case, the exponent factor n should be greater than 2 (Gilmer 1976, Atwater 1988). Therefore, the values of k and n were derived by fitting the experimental measurements of grain size for each phase at different annealing temperatures. The modeling results are shown as the solid lines in Fig. 6.3, with the experimental data shown as the points. For example, n=2.5 and k =8E-6 were calculated for the η phase, while n=2.6 and k= 3E-6 were calculated for the θ phase at 500oC. This

187

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System grain growth model fitted the experimental data very well, and could be used in the following calculation.

Fig. 6.3 The grain growth law (equation 6.6) fitted to the experimental results of average grain size of the (a) η phase and (b) θ phase as a function of annealing time from Fig. 5.5 and 5.10.

6.4 Modelling Results

The application of the model to the prediction of the thickness of each phase is shown in the schematic illustration in Fig. 6.1, and the modelling procedure used is described as following. The diffusion kinetic parameters were calculated with the model, and then the diffusion coefficients, Dgb and Dl, were calculated at each temperature for each phase.

With the grain growth model, the effective diffusion coefficients, Deff, were then calculated with the equations 2-6 – 2-8. The thickness for each phase was then predicted with Deff, the K values in the diffusion model, and the temperature and time profile.

The modelling results are presented in this section, including calculation of the kinetic parameters for the two diffusion mechanisms (lattice and grain boundary), the validation of the modelling, and the application of the model to both heat treatments and the USW process.

6.4.1 The Kinetic Parameters Of The Two Diffusion Mechanisms

The growth rates of each phase were calculated from the experimental isothermal heat treatment data in chapter 5, and the effective diffusion coefficients that gave these

188

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

growth rates were then calculated with the diffusion model. For example,⁡푘η is equal to

o 3.7 and 푘θ is equal to 3.8 at 500 C. The results are shown as the points in Fig. 6.4. With the grain growth model, the effective volume fraction of the grain boundary diffusion in the whole diffusion, g, was calculated using equation 2-8 at each temperature for each annealing time. The diffusion coefficients for lattice and grain boundary diffusion of each phase were then calculated using equation 2-7. Consequently, with the calculated diffusion coefficients of each diffusion mechanism in each phase, the pre-exponent factors and the activation energies were calculated using equation 2-6 and the results are shown in Table 6.1, using the grain size data in Fig. 6.3 and equation 6-6.

Fig. 6.4 Effective diffusion coefficients of each phase as functions of annealing time during heat treatment.

Table 6.1 Calculated diffusion kinetics parameters of the two IMC phases

2 2 Phase Dl0 (m /s) Dgb0 (m /s) Ql (kJ/mol) Qgb (kJ/mol) -4 η (Fe2Al5) 500 5×10 240 120 -5 θ (FeAl3) 3 3×10 220 110

189

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

6.4.2 Validation Of The Modelling

In order to check the validation of the modelling, the thickness of each phase was predicted with the double-IMC phase diffusion model and grain growth model as a function of annealing time during static heat treatment under isothermal conditions. The results at 450oC and 500oC are shown as the solid lines in Fig. 6.5. The experimental data at each temperature is also shown in Fig. 6.5 as the points. From the results, the model is able to give a good prediction of the thickness for both phases in present study. This model was also used to predict the thickness of the η phase at 600oC, with the grain size data collected from Springer et al. (2011 b). Encouragingly, the modelling results agree well with the thickness data from Springer et al. (2011 b) as can be seen in Fig. 6.5.

6.4.3 The Application Of The Model To Heat Treatments

In order to find the influence of the grain size and annealing time on the inter-diffusion coefficient for each phase, the contribution of the two diffusion coefficients, gDgb and (1- g)Dl, was plotted against the grain size and the annealing time in Figs. 6.6 and 6.7, together with the effective diffusion coefficient, Deff. As g is usually quite small, (1-g) is approximately equal to 1, resulting in the constant value of (1-g)Dl which is almost equal to Dl in all the cases.

In Fig. 6.6, the diffusion coefficients are plotted against the grain size. It can be seen that

o in general, gDgb takes a much larger fraction of Deff at low temperatures (450 C) than at

190

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System high temperatures (500oC) for both phases, resulting in a similar behaviour with

o temperature for both diffusion coefficients. At 450 C, the values of gDgb and (1-g)Dl are of the same order of magnitude when the grain size is small at 1.0 μm for both phases, and then gDgb becomes much smaller than (1-g)Dl when the grain size grows to 1.5 μm.

Therefore, Deff reduces dramatically when the grain size grows above 1.0 μm, and becomes close to Dl after the grain size is over 1.5 μm.

The effect of grain size on the diffusion coefficient is different for the two phases at 500oC.

For the η phase, gDgb is smaller than (1-g)Dl even when the grain size 0.1 μm. As a result,

o the effect of grain size on Deff for the η phase at 500 C is more limited when the grain size

o is smaller than 1.0 μm. For the θ phase even at 500 C, gDgb still contributes a significant fraction to Deff before the grain size becomes larger than 1.5 μm.

o o (d) the θ phase at 500 C. It should be noted that the gDgb are almost equal to Deff at 500 C when the grain size is over 2 μm.

191

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

By using the grain growth model, the diffusion coefficients for each phase can be plotted against the annealing time during heat treatment, and the results are shown in Fig. 6.7. The predicted value agrees well with the experimental data in Fig. 6.4. It is clear that the contribution from gDgb reduces quickly at the beginning of annealing and then tends to a constant value in all cases, as a result of the grain growth occurring in the heat treatment, while (1-g)Dl almost remains constant value in the present study conditions. Consequently, the overall effective diffusion coefficient is affected by the annealing conditions (temperature), the initial IMC phase grain size, and the grain growth kinetics for each phase.

For the η phase, because of the higher activation energies, the grain boundary diffusion has a more significant influence during the growth behaviour at 450oC than 500oC. At

o 450 C, gDgb is larger than (1-g)Dl until the annealing time is over 16 hours, resulting in intersection of the two curves in Fig. 6.7 (a). Therefore, the sum of gDgb and (1-g)Dl, Deff

o has a similar trend with gDgb during heat treatment. At 500 C, however, gDgb is about one order of magnitude less than (1-g)Dl after 8 hours, and Deff is almost equal to (1-g)Dl, resulting in the curves almost coinciding after long annealing times in Fig. 6.7 (b).

Fig. 6.7 The contribution of grain boundary diffusion coefficient (gDgb) and lattice diffusion coefficient ((1-g)Dl) to the effective diffusion coefficient, as a function of

192

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System annealing time for (a) the η phase at 450oC, (b) the η phase at 500oC, (c) the θ phase at 450oC, (d) the θ phase at 500oC.

For the θ phase, however, grain boundary diffusion always plays an important role in the

o thicken process. At 450 C, gDgb is always larger than (1-g)Dl, by a factor of small grain size

o in the present study conditions. Therefore, Deff is always dominated by gDgb. At 500 C, however, gDgb becomes smaller than (1-g)Dl after annealing for 20 minutes, but still contributes a significant fraction to Deff. As a result, Deff is always higher than (1-g)Dl, but the two curves tend to parallel with each other, as the reduction in gDgb slows down owing to grain growth on annealing for long times.

As the effect of the grain boundary diffusion on the effective diffusion coefficient is related to the annealing conditions, the specific phase and its grain size is important to find the influence of these factors on the predicted thickness of the IMC layer. Therefore, the thickness of each phase was predicted with and without considering the effect of grain boundary diffusion and grain growth, and the results are shown in Fig. 6.8. In this case, Deff=Dl was used in the model.

Fig. 6.8 The effect of grain boundary diffusion on the prediction of the IMC layer thickness, as a function of annealing time for (a) the η phase and (b) the θ phase. In the figure, 'with

193

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

In Fig. 6.8, the predicted thickness for each phase is plotted against the annealing time. The solid lines are the results with grain boundary diffusion included, while the dashed lines are the results without considering grain boundary diffusion. In general, the influence of the grain boundary diffusion on the predicted thickness is dramatic at the low temperature (450oC) in both phases, but the influence becomes less important at the high temperature (500oC). Furthermore, the influence is more significant in the θ phase than in the η phase, which also agrees with the effective diffusion coefficient results. In particular, the predicted thickness of the η phase is not changed significantly after taking

o the grain boundary diffusion into account in the model at 500 C (Fig. 6.8 (a)), as gDgb is too small at this temperature (Fig. 6.7 b).

In order to find the influence of grain growth on the predicted thickness for each phase, the thickness was predicted in two ways; in the first case, using the diffusion model with the grain growth model, while in the using the diffusion model with a fixed grain size. In the present study, the grain size of both phases was mainly in the range between 0.1 μm and 0.7 μm, so the fixed grain sizes d=0.1 μm and d=0.7 μm were chosen as upper and lower limits. The predicted effective diffusion coefficient and thickness for each phase are plotted against the annealing time in Fig. 6.9 and 6.10, respectively. The results are plotted as the solid lines, dashed lines and dashed-dotted lines, for modelling with grain growth and fixed grain sizes of d=0.1 μm and d=0.7 μm, respectively. In addition, the lattice diffusion coefficients are shown as the round dashed lines and found to be lower than all of the effective diffusion coefficients.The effective diffusion coefficient in the grain growth model is called as the 'normal coefficient' in the following discussion.

194

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

Fig. 6.9 The effect of grain size on the effective diffusion coefficient, as a function of annealing time, for (a) the η phase and (b) the θ phase. In the figure, d is the fixed grain size, and the modelling results are shown as the dashed lines and dashed-dotted lines with the fixed grain sizes of d=0.1 μm and d=0.7 μm, respectively. 'Normal' indicates the modelling results are calculated with the grain growth model at each temperature, and the modelling results are shown as the solid lines, while 'without Dgb' indicates that only lattice diffusion was considered in the modeling, and the results are shown as the round dashed lines.

For the η phase at 450oC, the initial grain size was 0.1 μm. Compared with the normal coefficient, the effective diffusion coefficient with d=0.1 μm thus shows a much larger value, while the value with d=0.7 μm is smaller. When the temperature rises to 500oC, the effective diffusion coefficient with d=0.1 μm is still larger than the normal coefficient, but with a smaller gap between the solid and dashed lines than that at 450oC, and the value with d=0.7 μm is closed to the normal coefficient. These results confirm that grain growth affects the thicken process more significantly at low temperatures (450oC) than that at high temperature (500oC) in the η phase. A similar behaviour could be found in the θ phase which had a starting grain size of 0.15 μm. In this case, the diffusion coefficient at 450oC with a grain size of d=0.1 μm was found to be close to the normal coefficient at 500oC. It should be noted that when annealing at 500oC after 16 hours, the grain size became larger than 0.7 μm, resulting in a slower diffusion coefficient in the modelling with grain growth model, especially in the θ phase, indicating the huge influence of grain growth on the effective diffusion coefficient in both phases.

As expected, the effect of the grain growth on the predicted IMC layer thickness was similar to that of the effective diffusion coefficients for each phase (Fig. 6.10). For example, in the η phase, a fixed fine grain size (0.1 μm) resulted in a much thicker layer than predicted with grain growth model at 450oC, while the coarse grains (0.7 μm) resulted in the thinner layer. At 500oC, there was a cross over between the prediction thicknesses with d= 0.7 μm and grain growth model, as the grain size became larger than 0.7 μm at 16 hours. A similar behaviour can be seen for the θ phase.

195

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

6.4.4 The Application Of The Model In USW

If the temperature history is known, the model can also be used to predict the thickness for each phase in USW. For this study, the temperature history at the interface in the joints with a welding time of 3 seconds was chosen, as it was the longest welding time used in the present work. From the microstructure observation in DC04-AA6111 joints in Chapter 4, the grain size for each phase was found to be in the range between 0.02 μm and 0.2 μm with welding time of 0.5 seconds to 3 seconds. However, it was difficult to build a good grain growth model for the welding process, as the experimental data was not enough and the deformation in welding could also change the grain growth rate. Therefore, fixed grain sizes of d=0.02 μm and d=0.2 μm were used in the diffusion model as the upper and lower limits. The predicted thickness, using the measured interface temperature history shown in Fig. 4.21, is plotted against the process time with these fixed grain sizes in Fig. 6.11. The model results with grain sizes of d=0.02 μm and d=0.2 μm are shown as solid lines and the dashed lines, respectively. It should be noted that the lower limit of the grain size in the welding process (0.02 μm) is smaller than that in the heat treatment studies (0.1 μm).

196

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

The grain size can be seen to greatly affect the IMC layer thickness prediction in both phases. The predicted thickness changes nearly 37% and 50% for the η and θ phases with a change of grain size for 90%, respectively. As the two grain sizes are the upper and lower limits of both phases measured in the welding process, the measured grain growth behaviour should be between the two limits.

The predicted thickness of each phase with d=0.02 μm was compared with the experimental data, and the results are shown in Fig. 6.12. It can be seen that when compared, the predicted thickness is much lower than the experimental data, for the both phases, even when the lower limits of grain size are used. Therefore, d=0.02 μm will be used as the average grain size in the following modelling result. The contrast seen with the accurate prediction for the static annealing study indicates that grain growth is not the main reason for the under-prediction of the thickness of each phase. The reasons for this will be explored with the microstructure of the interface in the discussion below.

197

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

Although for reason that will be discussed further below, the model cannot be used to predict the accurate thickness for each phase in USW, it can still be used to see if temperature measurement errors are affecting the results and test the sensitivity of the growth behaviour of each phase to the welding parameters. The welding temperature and time were chosen as the two factors, as these two factors were found to affect the growth behaviour dramatically in the controllable parameters in the model (Robson 2012). The results are shown in Fig. 6.13 and 6.14.

In Fig. 6.13, the accuracy of the welding temperature measured in the welding process is found to have a significant influence on the predicted thickness in both phases. The measured temperature history result in the curves marked with 'standard', while the modified temperature histories with an error of 5oC result in the curves with +5oC and - 5oC. In general, a change in measured temperature of 1% results in the final predicted thickness change of 10% in both phases. Therefore, a lower welding time is very beneficial in the control of the IMC layer thickness. However, more important accurate modelling is very depended on highly accurate temperature measurements which are extremely difficult to obtain.

198

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

In Fig. 6.14, the predicted thicknesses for both phases in the joints with different welding times are plotted against the process time. It can be seen that longer welding time results in a much thicker layer. The gap between the standard curves and 50% shorter curves is larger than that between the standard curves and 50% longer curves, as the temperature rise slows down after welding of 3 seconds. Consequently, the welding time should be as short as possible when producing the dissimilar joints with thin IMC layer, as it results in both a lower temperature and exposure time.

199

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

6.5 Discussion

6.5.1 The Effect Of Grain Growth Of Each Phase In The IMC Layer

With the measured grain sizes in Chapter 5, a grain growth model was developed with equation 6-6 (Thompson 1990, Martin 1997). Two parameters, the exponent factor n and grain growth rate k, were calculated using the measured grain sizes. The modelling results were found to fit the measured grain sizes in each phase very well as can be seen in Fig. 6.3 with an exponent larger than 2.

As the equation 6-6 is for normal grain growth in bulk, three-dimensional systems, care is needed in using this model in the thin IMC layer in the present study (Thompson 1990). In the as-welded state, the average grain size in each phase was comparable to the thickness, resulting in quite slow grain growth rate. However, the thickness for both phases increased quickly after heat treatment even for a quite short time, like growing to over 2 μm at 500oC after 20 minutes for the η phase, which was 10 times of the average grain size. In the case when the grain size is smaller than the thickness of the layer, equation 6-6 can be used to describe the grain growth behaviour. Therefore, the grain growth model is available for both phases in the present study.

In the case of the columnar grain structures seen in the thin IMC layers, deviations were expected from equation 6-5 (Thompson 1990, Martin 1997, Heitjans 2005). In particular, the exponent factor n has been found to be greater than 2 (Atwater 1988). The reason for this was the smaller driving force for grain growth in a thin layer with columnar grains than that in a bulk system (Thompson 1990). As a results, it is reasonable that n>2 was used in the grain growth model; for example, n=2.5 was used in the model for the η phase and 2.6 for the θ phase at 500oC.

6.5.2 The Application Of The Model During Heat Treatment

The double-IMC phase diffusion model developed by Wang et al. (2015) gives the relationship between the thickness growth rate and the effective diffusion coefficient in each phase. Therefore, with the growth rates calculated in Chapter 5, the effective diffusion coefficients were calculated and are shown in Fig. 6.4. Then the diffusion coefficients for the two diffusion mechanisms in both phases were calculated using

200

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System equations 2-6 - 2-8 (Fisher 1951, Buehler 2005). Finally, the diffusion kinetic parameters for the two IMC phases were worked out and listed in Table 6.1.

As expected, the calculated activation energy of the grain boundary diffusion (Qgb) was half of that of the lattice diffusion (Ql) in each phase (Belova 2004, Bokstein 2004, Wang 2015). Therefore, the change in the temperature had a more influence on the lattice diffusion than that on the grain boundary diffusion. For example, when the temperature

o o increased from 450 C to 500 C, the value of (1-g)Dl increased by about one order of magnitude, while the increase in gDgb was less than one order of magnitude (Fig. 6.6).

Furthermore, the lower activation energy resulted in the higher diffusion coefficient (Dgb) for the grain boundary diffusion (Dl) in each phase. In general, Dgb was usually several orders of magnitude higher than Dl depending on the temperature (Wang 2015).

However, the effect of grain boundary diffusion on the diffusion process was also dependent highly on the ratio of grain boundary thickness and grain size, which is given by g in equation 2-8 (Belova 2004, Heitjans 2005). As the grain boundary thickness was much smaller than the grain size, g was quite small and generally controlled by the grain size (Belova 2004, Bokstein 2004, Heitjans 2005). Therefore, the contribution of grain boundary diffusion in the effective diffusion coefficient (Deff), gDgb, was highly depended on the grain size, while the contribution of lattice diffusion, (1-g)Dl, was almost constant and close to Dl.

The activation energies of the two diffusion mechanisms can be used to explain the large range of different activation energies of the η phase listed in Table 2.4 and Table 5.1 from previous studies in the literature. As the grain boundary diffusion was ignored, the calculated activation energy was actually the combination of the Qgb and Ql, resulting in the value between the two activation energies. Thus, the calculated activation energy is highly dependent on the temperature range used by the researchers. Furthermore, the inflection that appears for in the ln(k) vs. 1/T plot in Fig. 5.15 can now be seen to also be the result of the different contributions of grain boundary diffusion in the growth process at different annealing temperatures, which controlled the grain growth behaviour of each phase.

201

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

When the diffusion kinetic parameters were used for each phase and the two models (double-IMC phase diffusion model and grain growth model) were used, the thickness of each phase was accurately predicted as a function of static annealing time (Fig. 6.5). In addition to the good fit seen with the measured data in the present study, the thickness of the η phase was also accurately predicted, using the grain sizes from Springer's (2011 b) research. The modeling results fitted the measured thickness in all the cases very well, indicating that the model can be used to reliably predict the thickness of each phase in long duration static heat treatment.

As the contribution of grain boundary diffusion was controlled by the grain size, the diffusion coefficients of both phases were calculated as a function of grain size (Fig. 6.6).

There was no surprise that gDgb reduced with the grain growth, while (1-g)Dl was little affected by the changed grain size. Therefore, when the grain size was large enough (2

o o μm at 450 C and 1 μm at 500 C, Deff was close to (1-g)Dl, indicating that grain boundary diffusion could be ignored.

With the grain growth model, the influence of the annealing time on the diffusion coefficients was analyzed in Fig. 6.7. At 500oC, as grain growth occurred quickly at the beginning of annealing, gDgb reduced dramatically before the annealing time was over 8 hours, resulting in the same trend of Deff. Subsequently for longer times gDgb reached a much smaller value than (1-g)Dl and changed little with annealing time. Therefore, the value of Deff reduced and became close to (1-g)Dl.

As the grain size was still small enough at 450oC, and grain growth behaviour was not the same in the two phases, the diffusion coefficients had different trends as a function of

o annealing time at 450 C (Fig. 6.7 (a) and (c)). In the η phase, the values of gDgb and (1-g)Dl were of the same order of magnitude at 450oC, and an intersection was found between gDgb and (1-g)Dl after 16 hours. Therefore, Deff consisted of significant contributions from both gDgb and (1-g)Dl. In the θ phase, however, the values of gDgb was even higher than

(1-g)Dl. As a result, Deff was closer to gDgb than (1-g)Dl.

In order to find the influence of the grain boundary diffusion on the growth behaviour of the IMC phases, the thickness of both phases was predicted with, or without, gDgb in Deff (Fig. 6.8). A similar conclusion was found to the diffusion coefficient's result, which was

202

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System that grain boundary diffusion had more influence at low temperatures. In particular, when the temperature increased from 450oC to 500oC, the change in thickness reduced from 29% to 1% in the η phase, and 42% to 8% in the θ phase after annealing for 64 hours.

From calculation of Deff at different temperatures in each phase, and the effect of the grain boundary diffusion on the prediction of the IMC layer thickness, it is clear that grain boundary diffusion has more influence on the thicken process of the IMC phase at low temperature (450oC) than that at high temperature (500oC), and it also affects different phases differently, depending on their grain size and activation energies. Therefore, it is reasonable that two IMC layer growth activation energies, Q400-500 and Q500-570, were calculated of the η phase at 400oC to 500oC and 500oC to 570oC, which was 116 kJ/mol and 248 kJ/mol, respectively (Fig. 5.15). It can be seen that Q400-500 is closer to Qgb (120 kJ/mol), while Q500-570 is closer to Ql (240 kJ/mol), indicating that the grain boundary diffusion makes the major contribution at low temperature, while lattice diffusion takes the major part at high temperatures.

As the grain size affects gQgb more significantly with fine grains, it was vital to find the effect of grain growth on the growth kinetics for each phase. With the upper and lower grain size limits of d=0.1 μm and d=0.7 μm, the predicted effective diffusion coefficients and thicknesses were compared to the normal coefficients and thickness (Fig. 6.9, 6.10). A summary of the change of predicted diffusion coefficient and thickness for each phase with different grain growth behaviour is listed in Table 6.2. Besides the influence of temperature and phase, the different grain growth behaviour in each phase was also found to affect the prediction results.

One remarkable phenomenon demonstrated by these results is that a reduction in grain size can even offset a large temperature difference. In the θ phase, the effective diffusion coefficient at 450oC with d=0.1 μm was calculated to be quite close to, and even larger than that at 500oC, with d=0.7 μm. The same trend can be found in the predicted thickness. Therefore, grain growth dramatically can affect the prediction of the thickness, especially with fine initial grain size.

203

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

Table 6.2 Effect of grain size on the change to the predicted diffusion coefficient and thickness for each phase, compared with the standard values calculated by the grain growth model (with an annealing time of 8 hours).

The change of The change of The change of The change of o o o thickness at Deff at 450 C Deff at 500 C thickness at 450 C 500oC η(Fe2Al5), 220% 140% 100% 60% d=0.1 μm θ(FeAl3) 230% 310% 100% 100% d=0.1 μm η(Fe2Al5), 40% 5% 20% 1% d=0.7 μm θ(FeAl3) 50% 20% 30% 6% d=0.7 μm

As the grain size in the as-welded joints was similar, or even smaller than 0.1 μm, the abnormal thickness fraction of the θ phase in the IMC layer could be explained by the effect of the grain boundary diffusion. After welding, the θ phase was found to be even thicker than the η phase, which was contrary to the growth kinetics (Fig. 4.6). This fast growth rate of the θ phase might be the result of the grain boundary diffusion, as the fine grains resulted in a greater change of the effective diffusion coefficient and thickness in the θ phase than that in the η phase, when the temperature was above 500oC, which was the main temperature range in USW (Table 6.2).

6.5.3 The Application Of The Model In USW

Using temperature histories from the welding process, the thickness of each phase was predicted with the diffusion model. As the grain size had a significant influence on the predicted thickness, care should be used in choosing the grain growth model. However, as the IMC phases grew from island shape particles to a thin layer, normal grain growth could not be applied, as the grain size was equal to the layer thickness (Atwater 1988, Thompson 1990). Therefore, the fixed grain size d=0.01 μm and d=0.2 μm were used in the diffusion model as the upper and lower limits. The predicted thickness of each phase was plotted against the process time in Fig. 6.11, and the 'accurate thickness' with grain growth should be between the solid and dashed lines.

204

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System

However, the predicted thickness of each phase was still much lower than measured data (Fig. 6.12). The error between the modelling results and measured data was as large as 36% and 700% for the η and θ phases with a welding time of 1.5 seconds, respectively. This error was too large to be explained by just the grain size.

Varying the temperature in the predictions also showed that this large difference could not be explained by an error in the interface temperature measurement. The temperature change of 1% resulted in the predicted thickness change of about 10% for both phases (Fig. 6.13), indicating the important to reduce the welding temperature. The temperature measurement in the welding process was not accurate, as it damaged the aluminium sheet. However, it is difficult to test the accurate temperature profile in the welding. Therefore, the predicted thickness could not fit the experimental data.

Combined with the microstructure in the joints made by USW in Chapter 4, the reason for the model failed in the USW process is because the model can only be used for the 1- Dimensional diffusion, which is the growth of a continuous layer with the same diffusion coefficient along the interface. The IMC phases started from island shaped particles, and the layer was always not uniform at the interface (Fig. 4.5, 4.6). Illustrations for the different morphologies of the IMC layer are shown in Fig. 6.15.

The growth behaviour of the island shaped IMC particles is quite different from the continuous layer. At start, IMC nucleates at the micro-band and grows to particles (Fig. 6.15 (a)). Then the IMC particles grow both sideways and by thickening. A continuous layer forms when the island shaped IMC particles begin to impinge (Fig. 6.15 (b)). In Fig. 6.15 (a), the diffusion at the locations without IMC particles is controlled by the interface diffusion instead of the diffusion through the IMC layer, as there is no barrier layer for the diffusion of aluminium and iron atoms, resulting in a much faster diffusion coefficient than at the locations with IMC particles. This is the nucleation process, and new model is required to predict the particles' nucleation behaviour in this stage. In Fig. 6.15 (b), the growth rate of the IMC layer is uniform along the interface and controlled by the lattice and grain boundary diffusion. The model can only be used in the situation of a continuous layer at the interface (Fig. 6.15 (b)). However, due to the short welding time in USW, the IMC layer was not continuous at the whole interface even after a welding time of 2

205

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System seconds. Therefore, it is reasonable that the model is invalid in the growth of the IMC layer during welding in USW.

Fig. 6.15 Illustrations for the different morphologies of the IMC layer. (a) Island shape particles and (b) uniform layer. The arrows indicate the diffusion direction.

Besides the temperature, welding time was also found to affect the thickness (Fig. 6.14). For example, with welding time increased from 1.5 seconds to 3.0 seconds, the thickness of the η phase increased 56%, while the increased thickness reduced to 44% when the welding time further increased to 4.5 seconds. The change of growth rate of the IMC layer was related to the heating rate in the welding process, which was fast in the first 2 seconds, and then slows down until the end of the welding. Consequently, it is important to control the welding time to control the IMC layer thickness.

6.6 Summary And Conclusions

In this chapter, a double-IMC phase diffusion model was applied, together with grain growth model, to predict the thickness of each phase in the diffusion process. The modelling results fitted very well with the data from the static heat treatment, but did not fit well with the data from the ultrasonic spot welding process.

Both the grain boundary diffusion and lattice diffusion occurred in the growth of the IMC layer. The activation energies for the grain boundary diffusion and lattice diffusion were calculated as 240 kJ/mol and 120 kJ/mol in the η phase, and 220 kJ/mol and 110 kJ/mol in the θ phase.

The contribution of the grain boundary diffusion in the growth of the IMC phases was related to the temperature and grain size. At lower temperatures in the phase with a smaller grain size, the grain boundary diffusion had a more significant influence on the

206

Chapter 6 Modelling of The IMC Layer Growth in an Fe-Al Dissimilar System growth rate of the IMC phase. In addition, the effect of the grain boundary diffusion was more significant on the growth of the θ phase than that on the growth of the η phase because of its fine grain structure. As a result, it was reasonable that θ phase was thicker than the η phase in the as-welded state.

The model is invalid for the growth of the IMC layers in USW process, as a result of the short welding time and uneven IMC layer. The growth of islands shaped particles is controlled by the interface diffusion instead of the lattice and grain boundary diffusion. As a result, the diffusion coefficients for the IMC phases into the substrates are changed along the interface, depending on whether there is an IMC particle formed at the position. Therefore, the diffusion model could not be used directly in the welding process, and a nucleation model is required.

207

Chapter 7 Conclusions and Future Work

CHAPTER 7 CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

In the present study, the nucleation and growth behaviour of the intermetallic compounds (IMCs) were investigated in dissimilar steel and aluminium alloy combinations, in both welding condition and during static heat treatment. The factors that affected the growth behaviour of the IMC layer were studied, including the pre treatment, alloy elements present and grain structure of the IMC layer. A double-IMC phase diffusion model was used to predict the thickness of each phase in the IMC layer, to obtain a better understanding of the growth kinetics of the IMC layer. In particular, the aim of this work was to address the large scatter seen in claimed kinetic parameters for Fe-Al IMC growth in the literature. The main conclusions from the study are listed as below.

With high power ultrasonic spot welding (USW), dissimilar joints were welded between a DC04 steel and two different aluminium alloys (AA6111 and AA7055). By optimizing the welding time, attractive joint strengths could be achieved by USW. However, no suitable process window was found that could produce joints with a nugget pullout failure mode, using the current welding procedure. In all cases the fracture path was still through the weld interface region. A purely brittle fracture, at the interface between the aluminium alloy and the IMC reaction layer, was seen in the AA7055-DC04 welds, whereas a higher energy mixed mode failure was found in the AA6111-DC04 combination with the similar IMC layer thickness.

Two reasons were claimed for the higher fracture energy for the DC04-AA6111 joints. Firstly, the IMC layers grew faster in the DC04-AA7055 joints than in the DC04-AA6111 joints in USW, especially for the θ phase. After heat treatment, the θ phase was found to have a similar growth rate with the η phase in the DC04-AA7055 joints, which was much faster than that of the θ phase in the DC04-AA6111 joints. Secondly, liquefaction occurred at the interface in the DC04-AA7055 joints due to the effect of Zn, left some pores in the aluminium side near the interface. Consequently, AA7055 is not a suitable aluminium alloy to be welded with steel.

208

Chapter 7 Conclusions and Future Work

A similar sequence of Intermetallic compound formation was observed in both material combinations and with both welding processes. In USW, η-Fe2Al5 was found to be the first IMC phase to nucleate, within a welding time of only 0.3 seconds, as isolated islands heterogeneously distributed at the weld interface. These early reaction centres probably formed within the micro-weld regions that first develop at initial asperity contact points across the joint interface. The IMC islands then spread to form a continuous layer in both material combinations, by a welding time of 0.7 seconds. With longer welding times a second IMC phase, θ-FeAl3, was seen to develop on the aluminium side of the joints.

With friction stir spot welding (FSSW) with a pin-less tool, the IMC layer was found to have a similar growth behaviour to that seen in the joints made by USW in AA6111-DC04 joints. However, compared with USW, a slower heating rate and lower peak temperature were found in the welding area during the pin-less FSSW process, due to the different temperature field generated from the welding process. As a result, the IMC layer at the interface in USW joints had a higher growth rate as a function of welding time and required less weld energy than in pin-less FSSW joints.

In the joints made by the solid state welding processes, the θ phase was found to be even thicker than the η phase, which was contrary to the reported growth kinetics. After the diffusion model was used to predict the thickness of each phase, the effect of the grain boundary diffusion on the growth behaviour was found to be more significant in the θ phase than that in the η phase, especially with the fine grain structures found in both phases in the as-welded state, due to the low welding temperature and short welding time. As a result, it was reasonable to conclude that in solid state welding conditions the θ phase had a faster growth rate than the η phase.

The appearance of the θ phase had two prerequisites; a continuous η phase layer formed first at the interface and there was enough input energy. After a continuous η phase layer formed, the concentration gradient across the interface was reduced, resulting in the possibility of formation of the θ phase with higher aluminium ratio. However, the growth rate of the θ phase was much slower than that of the η phase during static heat treatment. As a result, higher temperatures or longer annealing times than the conditions in the present study are required to for the θ phase.

209

Chapter 7 Conclusions and Future Work

The pre treatment was found to affect the growth behaviour of the IMC layer dramatically. In general, the IMC layer grew faster in the pre-welded dissimilar joints than in the diffusion couples. After pre welding, the dissimilar metals had a stronger bond at the interface than in diffusion couples, and the deformation occurred in the substrates also improved the inter-diffusion of the iron and aluminium atoms.

Grain growth occurred in both phases during heat treatment. In the present study, the grain size was smaller than the thickness of each phase, and the grain growth followed standard rate control laws, so that a grain growth model could be built with the measured data. However, the exponent factor was found to be larger than 2, as columnar grain formed in the IMC layer.

The grain structure of each phase in the IMC layer was found to greatly affect the predicted thickness due to the grain boundary diffusion. The grain boundary diffusion was found to have more influence at low temperatures (400oC-500oC) than at high temperatures (500oC-550oC) due to the different grain growth rates, resulting in the two calculated activation energies for the η phase., which was close to the activation energy of grain boundary diffusion (240 kJ/mol) and lattice diffusion (120 kJ/mol), respectively. Therefore, the grain growth model was important to the predicted thickness, especially for the fine grain size.

The diffusion model worked well to predict the thickness of each phase in the IMC layer in static heat treatment, but not in the welding process. The model could only be used in 1- dimension diffusion process, which was the growth behaviour of the continuous IMC layer during the static heat treatment. In the welding process, however, nucleation stage had a more significant influence than the growth stage due to the highly transient conditions in USW. As a result, the IMC phase layers were not uniform and continuous even after a welding time of 2 seconds. Therefore, the growth of the IMC phases was difficult to be predicted without the nucleation stage.

According to the diffusion model, a 1% change of the reaction temperature could lead to a 14% change of the IMC thickness, and the 50% increasing reaction time could lead to the 55% thicker IMC layer. Therefore, it is important to measure the interface

210

Chapter 7 Conclusions and Future Work temperature extremely accurately to predict the growth of the IMC layer and control the welding temperature and time for dissimilar steel to aluminium alloy welding.

7.2 Future Work

On the basis of this work, a number of fields are highlighted for a further study in the area of controlling the growth behaviour of the IMC layer in Fe-Al system. The specific recommendations for the further work are summarised below.

1 The orientation relationship between the IMC phases and substrates

TEM and EBSD could be applied to investigate the orientation relationship between the IMC phases and substrates, in both the as-welded and annealed states, to obtain the growth mechanism of the IMC phases into the substrates.

2 The effect of the alloy elements and coating

Although Zn was found to accelerate the growth of the IMC layers in both the welding conditions and during static heat treatment, the mechanism is still not clear. Further investigation could be focused on the distribution of Zn in the IMC layer and the grain structure of the IMC layer as a function of annealing time. It would be ideally if pure aluminium with certain content of zinc and pure iron could be used in the future. In addition, more work could be induced on the welding between aluminium alloy and steel with Zn coating.

Aluminium alloys with higher content of Si than AA6111, or with an Al-Si coating, could be welded with steel, and static treatment could be applied on the pre-welded joints, to check the effect of Si on the growth of the IMC layer.

The effect of other elements in aluminium alloy or steel, like Mn or Cu, on the growth of the IMC layer could also be studied, and the mechanism should be investigated with detailed microstructure observation.

3 Accurate temperature measurements in welding process

211

Chapter 7 Conclusions and Future Work

As the measured temperature has a significant influence on the growth of the IMC layer, a new temperature measurement method should be worked out in the welding process, without damaging the work-pieces.

4 The residual stress in as-welded joints

The residual stress in the dissimilar joints could be measured by XRD or FIB after welding, and a FE model could be built based on the experiment measurement. The effect of the residual stress on the growth of the IMC layer during static treatment could be studied further, to find the mechanism for the faster growth rate of the IMC layer in the pre- welded combinations.

5 Nucleation model of the IMC phase

A modelling approach could be taken to predict the nucleation stage of the IMC phase during static heat treatment. After the validation of the model is checked, it could then be used in the welding conditions. Therefore, the growth of the IMC phases could be predicted in the welding process.

212

Reference

REFERENCE www.acea.be/news/news_detail/reducing_co2_emissions. Accessed on December 10th, 2014.

D.R.G. Achar, J. Ruge, S. Sundaresan. Metallurgical and mechanical investigations of aluminium steel fusion welds (II). Aluminium, 56 (1980), 465.

E. Aghion, B. Bronfin, D. Eliezer. The role of the magnesium industry in protecting the environment. Journal of Materials Processing Technology, 117 (2001), 381-385.

L. Agudo, D. Eyidi, C. H. Schmaranzer et al.. Intermetallic Fex-Aly-phases in a steel/Al-alloy fusion weld. Journal of Materials Science, 42 (2007), 4205-4214.

M. V. Akdeniz, A. O. Mekhrabov. The effect of substitutional impurities on the evolution of Fe-Al diffusion layer. Acta Materialiaerialia, 46 (1998), 1185-1192. http://aluminium.matter.org.uk/content/html/eng/default.asp?catid=217&. Accessed on December 10th, 2014.

M. Annoni, M. Carboni. Ultrasonic metal welding of AA6022-T4 lap joints: Part I – Technological characterization and mechanical behavior. Science and Technology of Welding and Joining, 16 (2011), 107-115.

H. A. Atwater, C. V. Thompson and H. I. Smith. Ion‐ bombardment‐ enhanced grain growth in germanium, silicon, and gold thin films. Journal of Applied Physics, 64 (1988), 2337.

K. Z. Baba-Kishi. Review Electron backscatter Kikuchi diffraction in the scanning electron microscope for crystallographic analysis. Journal of Materials Science, 37 (2002), 1715- 1746.

H. Badarinarayaran, F. Hunt, K. Okamoto. Friction stir welding and processing, Ch12: friction stir spot welding. Hitachi America Ltd., 2012.

D. Bakavos, P.B. Prangnell. Mechanisms of joint and microstructure formation in high power ultrasonic spot welding 6111 aluminium automotive sheet. Materials Science and Engineering: A , 527 (2010) 6320-6334.

213

Reference

R. Balasundaram, V. K. Patel, S. D. Bhole et al.. Effect of zinc interlayer on ultrasonic spot welded aluminum-to-copper joints. Materials Science and Engineering: A, 607 (2014). 277-286.

T. A. Barnes, I. R. Pashby. Joining techniques for aluminium space frames used in automobiles: Part I - solid and liquid phase welding. Journal of Materials Processing Technology, 99 (2000), 62-71.

T. A. Barnes, I. R. Pashby. Joining techniques for aluminium spaceframes used in automobiles: Part II - adhesive bonding and mechanical fasteners. Journal of Materials Processing Technology, 99 (2000), 72-79.

I. V. Belova, G. E. Murch. Analysis of the effective diffusivity in nanocrystalline materials. Journal of Metastable and Nanocrystalline Materials,19 (2004), 25- 34.

L.K Berg, J. Gjonnes, V. Hansen et al.. GP-zones in Al–Zn–Mg alloys and their role in artificial aging. Acta Materialia, 49 (2001), 3443-3451.

P.N. Bindumadhavana, S. Makeshb, N. Gowrishankarc et al.. Aluminizing and subsequent nitriding of plain carbon low alloy steels for piston ring applications. Surface and Coatings Technology, 127 (2000), 251-258.

P. J. Black. The structure of FeAl3 I. Acta Crystallographica, 8 (1955), 43-48.

P. J. Black. The structure of FeAl3 II. Acta Crystallographica, 8 (1955), 175-182.

B. S. Bokstein, M. A. Ivanov, Y. R. Kolobov et al.. Grain boundary diffusion in consolidated nanomaterials: Stress effect on grain boundary diffusion. Journal of Metastable and Nanocrystalline Materials, 19 (2004), 69- 106.

R. Borrisutthekul, T. Yachi, Y. Miyashita et al.. Suppression of intermetallic reaction layer formation by controlling heat flow in dissimilar joining of steel and aluminum alloy. Materials Science and Engineering: A, 467 (2007), 108-113.

A. Bouayad, C, Geromett, A. Belkebir et al.. Kinetic interactions between solid iron and molten aluminium. Materials Science and Engineering: A, 363 (2003), 53– 61.

214

Reference

K. Bouche, F. Barbier, A. Coulet. Intermetallic compound layer growth between solid iron and molten aluminium. Materials Science and Engineering: A, 249 (1998), 167-175.

S. Bozzia, A. L. Helbert-Ettera, T. Baudinb et al.. Intermetallic compounds in Al 6016/IF- steel friction stir spot welds. Materials Science and Engineering: A, 527 (2010), 4505-4509.

M. J. Buehler, T. J. Balk, E. Arzt, H. Gao. Handbook of Theoretical and Computational Nanotechnology. American Scientific Publishers, Los Angeles, 2006, 1-35.

G. Buffa, l. Fratini, M. Piacentini. On the influence of tool path in friction stir spot welding of aluminum alloys. Journal of Materials Processing Technology, 208 (2008), 309-317.

U. Burkhardt, Y. Grin, M. Ellner et al.. Structure Refinement of the Iron-Aluminium Phase with the Approximate Composition Fe2Al5. Acta Crystallographica, 50 (1994), 313-316.

D. J. Chakrabarti, D. E. Laughlin. Phase relations and precipitation in Al–Mg–Si alloys with Cu additions. Progress in Materials Science, 49 (2004), 389-410.

Y. Chen, H. Farid, P. B. Prangnell. Feasibility study of short cycle time friction stir spot welding thin sheet Al to un-galvanised and galvanized steel. 8th International Friction Stir Welding Symposium, Timmendorfer Strand, Germany, 18–20 May 2010.

H. Chen, A. J. Pinkerton, L. Li et al.. Gap-free fibre laser welding of Zn-coated steel on Al alloy for light-weight automotive applications. Materials & Design, 32 (2011), 495-504.

Y. Chen, D. Bakavos, A. Ghilinia et al.. HAZ development and accelerated post-weld ageing in ultrasonic spot welding aluminum 6111-T4 automotive sheet. Acta Materialia, 60 (2012), 2816-2828.

Y. Chen, A. Ghilinia, P. B. Prangnell. Interface structure and bonding in abrasion circle friction stir spotwelding: A novel approach for rapid welding aluminium alloy to steel automotive sheet. Materials Chemistry and Physics, 134 (2012), 459-463.

W. Cheng, Y. Chang, C. Wang. Observation of high-temperature phase transformation in the aluminide Cr– Mo steel using EBSD. Surface and Coatings Technology, 203 (2008), 401-406.

215

Reference

W. Cheng, C. Wang. Growth of intermetallic layer in the aluminide mild steel during hot- dipping. Surface and Coatings Technology, 204 (2009), 824-828.

W. Cheng, C. Wang. EBSD characterization of high-temperature phase transformations in an Al–Si coating on Cr – Mo steel. Materials Characterization, 64 (2012), 15-20.

W. K. Choi, H. M. Lee. Effect of Soldering and Aging Time on Interfacial Microstructure and Growth of intermetallic Compounds between Sn-3.5Ag Solder Alloy and Cu Substrate. Journal of Electronic Materials, 29 (2000), 1207 – 1213.

X. Cui, H. Zhang, S. Wang et al.. Design of lightweight multi-material automotive bodies using new material performance indices of thin-walled beams for the material selection with crashworthiness consideration. Materials and Design, 32 (2001), 815-821.

E. P. DeGarmo, J. T. Black, and R. A. Kohser. Materials and Processes in Manufacturing (9th ed.). John Wiley & Sons, Hoboken, USA, 2011.

S. G. Denner, R. D. Jones. Kinetic interactions between aluminium (liquid) and iron/steel (solid) for conditions applicable to hot-dip aluminizing. Metals Technology, 3 (1977), 167.

P. D. Desai. Thermodynamic Properties of Selected Binary Aluminium Alloy Systems. Physics Chemistry, 16 (1986), 109-124.

D. Dietrich, D. Nickel, M. Krause et al.. Formation of intermetallic phases in diffusion- welded joints of aluminium and magnesium alloys. Journal of Materials Science, 46 (2011), 357-364.

H. Dong, W. Hu, Y. Duan et al.. Dissimilar metal joining of aluminum alloy to galvanized steel with Al–Si, Al–Cu, Al–Si–Cu and Zn–Al filler wires. Journal of Materials Processing Technology, 212 (2012), 458-464.

G. Effenberg, S. Ilyenko. Light Metal System. Part 3. Springer, Berlin, 2005.

G. Eggeler, W. Auer, H. Kaesche. On the influence of silicon on the growth of the alloy layer during hot dip aluminizing. Journal of Materials Science, 21 (1986), 3348.

216

Reference

S. Elangovan, S. Semeer, K. Prakasan. Temperature and stress distribution in ultrasonic metal welding—An FEA-based study. Journal of Materials Processing Technology, 209 (2009), 1143-1150.

M. Ellner, J. Mayer. X-ray and electron diffraction investigations on the liquid-quenched Fe2Al5. Scripta Metallurgica et Materialia, 26 (1992), 501-504.

S. Esmaeili, X. Wang, D. J. Lloyd et al.. On the Precipitation-Hardening Behavior of the Al- Mg-Si-Cu Alloy AA6111. Metallurgical and Materials Transactions: A, 34 (2003), 751–763.

J. Fan, C. Thomy, F. Vollertsen. Effect of Thermal Cycle on the Formation of Intermetallic Compounds in Laser Welding of Aluminum-Steel Overlap Joints. Physics Procedia, 12 (2011), 134-141.

FEI Company. Quanta 3 D FEG User Operation Manual. 2007.

D. Feng, X. Zhang, S. Liu et al. Non-isothermal ‘‘retrogression and re-ageing’’ treatment schedule for AA7055 thick plate. Materials & Design, 60 (2014), 208-217.

J. C. Fisher. Calculation of Diffusion Penetration Curves for Surface and Grain Boundary Diffusion. Journal of Applied Physics, 2 (1951), 74-77.

S. Fukada, R. Ohashi, M. Fujimoto et al.. Refill friction stir spot welding of dissimilar materials consisting of A6061 and hot dip zinc-coated steel sheets. Proceedings of the 1st International Joint Symposium on Joining and Welding. Osaka, Japan, 6–8 November 2013.

G. H. Gilmer. Grain-boundary diffusion in thin films: II. Multiple grain boundaries and surface diffusion. Journal of Applied Physics, 47 (1976), 4373- 4380.

A. Griger, V. Stefaniay, T. Turmezey. Crystallographic Data and Chemical Compositions of Aluminium-Rich Al-Fe Intermetallic Phases. Zeitschrift für Metallkunde, 77 (1986), 30-35.

F. Haddadi. Joint Performance and Interface Reactions in Dissimilar Aluminium to Steel Ultrasonic Spot Welding. United Kingdom, 2012.

H. N. Han, J. P. Clark. Lifetime costing of the body-in-white: steel vs. aluminium. The Journal of The Minerals, Metals & Materials Society, 47 (1995), 22–28.

217

Reference

E. W. Hart. On the role of dislocations in bulk diffusion. Acta Metallurgica, 5 (1957), 597.

J. L. Harthoon. Ultrasonic metal welding. Netherlands, 1978.

P. Heitjans, J. Karger. Diffusion in condensed matter. Springer, Berlin, 2005.

E. Hetrick, J. Baer, W. Zhu et al.. Ultrasonic metal welding process robustness in aluminum automotive body construction applications. Welding Journal, 7 (2009), 149-158.

S. Hirose, T. Itoh, M. Makita et al.. Defect structure of deformed Fe2Al5 intermetallic compound. Intermetallics, 11 (2003), 633–642.

J. Hirsch. Aluminium in Innovative Light-Weight Car Design. Materials Transactions, 52 (2011), 818-824.

R. Jahn, R. Cooper, D. Wilkosz. The Effect of Anvil Geometry and Welding Energy on Microstructures in Ultrasonic Spot Welds of AA6111-T4. Metallurgical and Materials Transactions: A, 38 (2007) 570-583.

A. Jambor, M. Beyer. New cars - new materials. Materials & Design, 18 (1997), 203-209.

P. Jedrasiak, A. Reilly, H. R. Shercliff et al.. Thermal modelling of Al-Al and Al-steel Friction Stir Spot Welding, 10th International Seminar Numerical Analysis of Weldability, Graz, Austria, 24-26 September 2012.

P. Jedrasiak, H. R. Shercliff, Y. C. Chen et al.. Modelling of the Thermal Field in Dissimilar Alloy Ultrasonic Welding. Journal of Materials Engineering and Performance, 24 (2015), 799-807.

M. Kajihara. Analysis of kinetics of reactive diffusion in a hypothetical binary system. Acta Materialiaerialia, 52 (2004), 1193–1200.

M. Kajihara. Quantitative Evaluation of Interdiffusion in Fe2Al5 during Reactive Diffusion in the Binary Fe–Al System. Materials Transactions, 47 (2006), 1480-1484.

S. A. Kalogirou. Solar Energy Engineering. Elsevier, Oxford, 2014.

S. Katayama, H. Nagayama, M. Mizutani et al.. Fiber laser welding of aluminum alloy. Welding International, 23 (2009), 744–752

218

Reference

W. E. King, H. D. Barth, V. M. Castillo et al.. Observation of keyhole-mode laser melting in laser powder-bed fusion additive manufacturing. Journal of Materials Processing Technology, 214 (2014), 2915- 2925.

A. Kubošová, M. Karlík, P. Haušild et al.. Fracture Behaviour of Fe3Al and FeAl Type Iron Aluminides, Materials Science Forum, 567-568 (2007), 349-352.

T. Kuo, Y. Lin. Effects of different shielding gases and power waveforms on penetration characteristics and porosity formation in laser welding of Inconel 690 alloy. Materials Transactions, 48 (2007,) 219–26.

J. M. Lee, S. B. Kang, T. Sato et al.. Evolution of iron aluminide in Al/Fe in situ composites fabricated by plasma synthesis method. Materials Science and Engineering: A, 362 (2003), 257-263.

W. Lee, M. Schmuecker, U. Alfaro Mercardo et al.. Interfacial reaction in steel–aluminium joints made by friction stir welding. Scripta Material, 55 (2006), 355-358.

A. J. Leonard, A. A. Lockyer. Flaws in Friction Stir Welds. 4th International Symposium on Friction Stir Welding, Utah, USA, 14-16 May 2003.

M. Lesemann, M. Brockerhoff, P. Urban. The prospects of multi-material design for a compact-class front section. ATZautotechnology, 8 (2008), 16-20.

S. Liu, C. Li, S. Han et al.. Effect of natural aging on quench-induced inhomogeneity of microstructure and hardness in high strength 7055 aluminum alloy. Journal of Alloys and Compounds, 625 (2015), 34-43.

T. Liyanage, J. Kilbourne, A. P. Gerlich et al.. Joint formation in dissimilar Al alloy/steel and Mg alloy/steel friction stir spot welds. Science and Technology of Welding and Joining, 14 (2009), 500-508.

D. T. Llewellyn, R. C. Hudd. Steels: Metallurgy and Applications. Butterworth-Heinemann, Oxford, UK, 1998.

219

Reference

Q. Luo, J. Cheng, Y. Li et al.. Experimental study and thermodynamic assessment of the Al–Fe rich side of the Al–Zn–Fe system at 300 and 550°C. CALPHAD: Computer Coupling of Phase Diagrams and Thermochemistry, 37 (2012) 116–125.

P.K. Mallick. Materials, design and manufacturing for lightweight vehicles. Woodhead Publishing Series, Cambridge, 2010.

R. K. W. Marceau, A. de Vaucorbeil, S. P. Ringer et al.. Analysis of strengthening in AA6111 during the early stages of aging: Atom probe tomography and yield stress modelling. Acta Materialiaerialia, 61 (2013), 7085-7303.

J. W. Martin, R. D. Doherty, B. Cantor. Stability of microstructure in metallic systems. Cambridge University Press, Cambridge, 1997.

N. Masahashi, S. Watanabe, N. Nomura et al.. Laminates based on an iron aluminide intermetallic alloy and a CrMo steel. Intermetallics, 13 (2005), 717-726.

T. B. Massalski. Binary Alloy Phase Diagrams. American Society for Metals, Metals Park, OH, 1986.

S. I. Matsuoka, H. Imai. Direct welding of different metals using ultrasonic vibration. Journal of Materials Processing Technology, 209 (2009), 954-960.

S. Meco, G. Pardal, S. Ganguly et al.. Application of laser in seam welding of dissimilar steel to aluminium joints for thick structural components. Optics and Lasers in Engineering, 67 (2015), 22 - 30.

H. Mehrer. Diffusion in Solids: Fundamentals, Methods, Materials, Diffusion-Controlled Processes. Springer Science & Business Media, 2007.

R. S. Mishra, Z. Ma. Review Article: Friction stir welding and processing. Materials Science and Engineering: R, 50 (2005), 1-78.

D. Mitlin, V. Radmilovic, T. Pan et al.. Structure–properties relations in spot friction welded (also known as friction stir spot welded) 6111 aluminum. Materials Science and Engineering: A, 441 (2006), 79-96.

220

Reference

K. Miyagawa, M. Tsubaki, T. Yasui et al.. Spot welding between aluminium alloy and Zn- coated steel by friction stirring. Welding International. 23 (2009), 648-653.

N. Murdeshwar, J. Krzanowski. A microstructural study of dislocation substructures formed in metal foil substrates during ultrasonic wire bonding. Metallurgical and Materials Transactions: A, 28 (1997) 2663-2671.

B. Myers. TEM Sample Preparation with the FIB/SEM. Northwestern University –Evanston, IL, 2009.

D. Naoi, M. Kajihara. Growth behavior of Fe2Al5 during reactive diffusion between Fe and Al at solid-state temperatures. Materials Science and Engineering: A, 459 (2007), 375-382.

P. Nejat, F. Jomehzadeh, M. M. Taheri et al.. A global review of energy consumption, CO2 emissions and policy in the residential sector (with an overview of the top ten CO2 emitting countries). Renewable & Sustainable Energy Reviews , 43 (2015), 843-862.

K. Nishihara, T. Watanabe, T. Sasaki. Effect of Weld Tip Geometry on Ultrasonic Welding between Steel and Aluminum Alloy. Advanced Materials Research, 89-91 (2010), 419-424.

H. Oikawa, S. Ohmiya, T. Yoshimura et al.. Resistance spot welding of steel and aluminum sheet using insert metal sheet. Science and Technology of Welding and Joining, 4 (1999), 80–88.

A. Panteli. Friction joining of aluminium-to-magnesium for lightweight automotive applications, United Kingdom, 2012.

V. K. Patel, S. D. Bhole, D. L. Chen. Formation of zinc interlayer texture during dissimilar ultrasonic spot welding of magnesium and high strength low alloy steel. Materials & Design, 45 (2013), 236-240.

P. Peyre, G. Sierra, F. Deschaux-Beaume et al.. Generation of aluminium–steel joints with laser-induced reactive wetting. Metallurgical and Materials Transactions: A, 444 (2007), 327-338.

P. B. Prangnell, D. Bakavos. Novel Approaches to Friction Spot Welding Thin Aluminium Automotive Sheet. Materials Science Forum, 638-642 (2010), 1237-1242.

221

Reference

P. B. Prangnell, F. Haddadi, Y. Chen. Ultrasonic spot welding of aluminium to steel for automotive applications – microstructure and optimisation. Materials Science and Technology, 27 (2011), 617-624.

R. Qiu, C. Iwamoto, S. Satonaka. Interfacial microstructure and strength of steel/aluminum alloy joints welded by resistance spot welding with cover plate. Journal of Materials Processing Technology, 209 (2009), 4186-4193.

R. Qiu, C. Iwamoto, S. Satonaka. The influence of reaction layer on the strength of aluminum/steel joint welded by resistance spot welding. Materials Characterization, 60 (2009), 156-159.

R. Rai, J. W. Elmer, T. A. Palmer et al.. Heat transfer and fluid flow during keyhole mode laser welding of tantalum, Ti-6Al-4V, 304L stainless steel and vanadium. Journal of Physics D: Applied Physics, 40 (2007), 5753–5766.

W. Rasband. ImageJ (Version 1.45S) [Computer program]. Available at http://imagej.nih.gov/ij/download.html, accessed on May 9th, 2012.

A. Reilly, H. R. Shercliff, G. J. McShane et al. Novel approaches to modelling metal flow in friction stir spot welding. 10th International Seminar on Numerical Analysis of Weldability, Graz, Austria, 24-26 September 2012.

C. Rest, P. J. Jacques, A. Simar. On the joining of steel and aluminium by means of a new friction melt bonding process. Scripta Materialia, 77 (2014), 25–28.

A. P. Reynolds, W. Tang, Z. Khandkar et al.. Relationships between weld parameters, hardness distribution and temperature history in alloy 7050 friction stir weld. Science and Technology of Welding and Joining, 10 (2005), 190-199.

S. Reyntjens, R. Puers. A review of focused ion beam applications in microsystem technology. Journal of Micromechanics and Microengineering, 11 (2001), 287–300.

R. W. Richards, R. D. Jones, P. D. Clements et al.. Metallurgy of continuous hot dip aluminising. International Materials Reviews, 39 (1994), 191-212.

222

Reference

A. Robalino-López, Á. Mena-Nieto, J. García-Ramoset et al.. Studying the relationship between economic growth, CO2 emissions, and the environmental Kuznets curve in Venezuela (1980–2025). Renewable & Sustainable Energy Reviews, 41 (2015), 602-614.

J. D. Robson, A. Panteli, P. B. Prangnell. Modelling intermetallic phase formation in dissimilar metal ultrasonic welding of aluminium and magnesium alloys. Science and Technology of Welding and Joining, 17, (2012), 447-453.

J. Rodenburg. Tutorials in Transmission Electron Microscopy. http://www.rodenburg.org/. Accessed on May 2nd, 2013.

T. S. Rosendo, B. Parra, M.D. Tier et al.. Mechanical and microstructural investigation of friction spot welded AA6181-T4 aluminium alloy. Materials & Design, 32 (2011), 1094- 1100.

V. R. Ryabov. Aluminizing of Steel. Oxonian Press, New Delhi, 1985.

S. Satonaka, C. Iwamoto, R. Qiu et al.. Trends and new application of spot welding for aluminum alloy sheets. Journal of Japan light metal welding construction, 44 (2006), 41– 44.

J. M. Schreiber, Z. R. Omcikus, T. J. Eden et al.. Combined effect of hot extrusion and heat treatment on the mechanical behavior of 7055 AA processed via spray metal forming. Journal of Alloys and Compounds, 617 (2014), 135-139.

G. Sha, A. Cerezo. Early-stage precipitation in Al–Zn–Mg–Cu alloy (7050). Acta Materialia, 52 (2004), 4503-4516.

H. R. Shahverdi, M. R. Ghomashchi, S. Shabestari et al.. Microstructural analysis of interfacial reaction between molten aluminium and solid iron. Journal of Materials Processing Technology, 124 (2002), 345-352.

M. Shakil, N. H. Tariq, M. Ahmad et al.. Effect of ultrasonic welding parameters on microstructure and mechanical properties of dissimilar joints. Materials & Design, 55 (2014), 263-273.

223

Reference

Y. Shi, H. Zhang, T. Watanabe. CW/PW dual-beam YAG laser welding of steel/aluminum alloy sheets. Optics and Laser in engineering, 48 (2010), 732-736.

K. Shibata, S. Morozumi, S. Koda. Formation of the Alloy Layer in the Fe-Al Diffusion Couple. Journal of the Japan Institute of Metals and Materials, 30 (1966), 382-388.

G. Sierra , P. Peyre , F. Deschaux-Beaume. Steel to aluminum key-hole laser welding. Materials Science and Engineering: A, 447 (2007), 197-208.

G. Sierra , P. Peyre , F. Deschaux-Beaume et al.. Galvanised steel to aluminium joining by laser and GTAW processes. Materials Characterization, 59 (2008), 1705-1715. http://www.superlightcar.com/. Accessed on December 10th, 2014.

C. Soyarslan, D. P. F. Fassmann, B. Plugge et al.. An Experimental and Numerical Assessment of Sheet-Bulk Formability of Mild Steel DC04. Journal of Manufacturing Science and Engineering, 133 (2011), 061008-9.

H. Springer. Fundamental research into the role of intermetallic phases in joining of Aluminium alloys to steel. Germany, 2011.

H. Springer, A. Kostka, E. J. Payton et al.. On the formation and growth of intermetallic phases during interdiffusion between low-carbon steel and aluminum alloys. Acta Materialiaerialia, 59 (2011), 1586–1600.

H. Springer, A. Kostka, E. J. Payton et al.. Influence of intermetallic phases and Kirkendall- porosity on the mechanical properties of joints between steel and aluminium alloys. Materials Science and Engineering: A, 528 (2011), 4630-4642.

P. T. Stroup, G. A. Purdy. Effect of coatings on the fatigue behavior of 0.22% carbon steel. Metal Progress, 5 (1950), 57-59.

P. Su, A. Gerlich, T. H. North et al.. Material flow during friction stir spot welding. Science and Technology of Welding and Joining, 11 (2006), 61-71.

Y. Su, X. Hua, Y. Wu. Quantitative characterization of porosity in Fe–Al dissimilar materials lap joint made by gas metal arc welding with different current modes. Journal of Materials Processing Technology, 214 (2014), 81-86. 224

Reference

X. Sun, E. V. Stephens, M. A. Khaleel et al.. Resistance spot welding of aluminum alloy to steel with transition material – From process to performance – Part 1: experimental study. Welding Journal, 83 (2004), 188–195.

Y. Sun, H. Fuji, N. Takaki et al.. Microstructure and mechanical properties of dissimilar Al alloy/steel joints prepared by a flat spot friction stir welding technique. Materials & Design, 47 (2013), 350-357.

E. Taban, J. E. Gould, J. C. Lippold. Dissimilar friction welding of 6061-T6 aluminum and AISI 1018 steel: Properties and microstructural characterization Materials & Design, 31 (2010), 2305-2311.

T. Tanaka, T. Morishige, T. Hirata. Comprehensive analysis of joint strength for dissimilar friction stir welds of mild steel to aluminium alloys. Scripta Material, 61 (2009), 756-759.

N. Tang, Y. Li, S. Kurosu. Interfacial reactions of solid Co and solid Fe with liquid Al. Corrosion Science, 60 (2012), 32-37.

G. Tani, A. Ascari, G. Campana et al.. A study on shielding gas contamination in laser welding of non-ferrous alloys. Applied Surface Science, 254 (2007), 904–907.

W. M. Thomas. Friction stir welding - recent developments. Materials Science Forum, 426-432 (2003), 229-236.

C. V. Thompson. Grain growth in thin films. Annual Review of Materials Science, 20 (1990), 245-268.

P. L. Threadgill, A. J. Leonard, H. R. Shercliff et al.. Friction stir welding of aluminium alloys. International Materials Reviews, 54 (2009), 49-93.

M. D. Tier, T. S. Rosendo, J. F. dos Santos et al.. The influence of refill FSSW parameters on the microstructure and shear strength of 5042 aluminium welds. Journal of Materials Processing Technology, 213 (2013), 997-1005.

M. J. Torkamany, S. Tahamtan, J. Sabbaghzadeh. Dissimilar welding of carbon steel to 5754 aluminum alloy by Nd:YAG pulsed laser. Materials & Design, 31 (2010), 458-465.

225

Reference

J. Tsujino, T. Ueoka. Ultrasonic multi-spot continuous welding of metal plate specimens using a two-vibration-system welding equipment. Ultrasonics, 34 (1996), 229-233. http://www.twi-global.com/technologies/welding-surface-engineering-and-material- processing/joining-processes/mechanical-fastening/. Accessed on December 10th, 2014.

H. Uzun, C. D. Donne, A. Argagnotto et al.. Friction stir welding of dissimilar Al 6013-T4 To X5CrNi18 -10 stainless steel. Materials & Design, 26 (2005), 41-46.

X. Wang, J. Embury, W. Poole et al.. Precipitation strengthening of the aluminum alloy AA6111. Metallurgical and Materials Transactions: A, 34 (2003), 2913-2924.

K. Wang, L. Chang, D. Gan. Heteroepitaxial growth of Fe2Al5 inhibition layer in hot-dip galvanizing of an interstitial-free steel. Thin Solid Films, 518 (2010), 1935 – 1942.

Q. Wang, X. Leng, T. Yang, et al.. Effects of Fe—Al intermetallic compounds on interfacial bonding of clad materials. Transactions of Nonferrous Metals Society of China, 24 (2012), 279-284.

L. Wang, Y. Wang, P. Prangnell et al.. Modelling of intermetallic compounds growth between dissimilar metals. Metallurgical and Materials Transactions: A, 46 (2015), 4106- 4114.

T. Watanabe, H. Sakuyama, A. Yanagisawa. Ultrasonic welding between mild steel sheet and Al–Mg alloy sheet. Journal of Materials Processing Technology, 209 (2009), 5475- 5480.

I. M. Watt. The principles and Practice of Electron Microscopy. Cambridge University Press, Cambridge, 1997.

D. B. Williams, C. B. Carter. Transmission Electron Microscopy (2nd edition). Springer, US, 2009.

M. Windmann, A. Röttger, W. Theisen. Phase formation at the interface between a boron alloyed steel substrate and an Al-rich coating. Surface & Coatings Technology, 226 (2013), 130–139.

226

Reference

M. Windmann, A. Röttger, H. Kügler et al.. Laser beam welding of aluminum to Al-base coated high-strength steel 22MnB5. Journal of Materials Processing Technology, 217 (2015), 88-95. http://www.worldautosteel.org/why-steel/fuel-efficiency/. Accessed on December 10th, 2014.

Y. Wu, Y. Lo. Surface protection for AA8090 aluminium alloy by diffusion bonding. Theoretical and Applied Fracture Mechanics, 38 (2002), 71-79.

P. Xue, B. Xiao, D. Ni. Enhanced mechanical properties of friction stir welded dissimilar Al– Cu joint by intermetallic compounds. Materials Science and Engineering: A, 527 (2010), 5723-5727.

T. Yamada, K. Miura, M. Kajihara et al.. Kinetics of reactive diffusion between Au and Sn during annealing at solid-state temperatures Materials Science and Engineering: A, 309 (2005), 118-126.

B. Yan, S. H. Lalam, H. Zhu. Performance evaluation of GMAW welds for four advanced high strength steels. 2005 SAE World Congress Soc. of Automotive Engineers, Warrendale, PA, 19-21 April 2005.

M. Yilmaz, M. Col, M. Acet. Interface properties of aluminum/steel friction-welded components. Materials Characterization, 49 (2003), 421-429.

F. Yin, M. Zhao, Y. Liu et al.. Effect of Si on growth kinetics of intermetallic compounds during reaction between solid iron and molten aluminum. Transactions of Nonferrous Metals Society of China, 23 (2013), 556-561.

C. Zhang, J. D. Robson, O. Ciuca et al.. Microstructural characterization and mechanical properties of high power ultrasonic spot welded aluminum alloy AA6111–TiAl6V4 dissimilar joints. Materials Characterization, 97 (2014), 83-91.

227