Interface Design of /Alumina Composites with Interpenetrating Structure (Preform-MMCs)

by Oliver Patrick Lott

Thesis submitted to fulfil the requirements of

Doctor of Philosophy

School of and Engineering, University of New South Wales, Sydney, Australia

2012

Certificate of Originality I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.

______

Oliver Lott, January 2012

i

Acknowledgements This thesis was accomplished in cooperation with UNSW School of Materials Science and Engineering, Sydney, Australia, and Aalen University, Materials Research Institute, Aalen, Germany. I express my deepest gratitude to my supervisors at UNSW, Professor Mark Hoffman and Professor Charles Sorrell, for their excellent supervision, advice and guidance; and thank them for fruitful discussions and being always welcomed by them in Sydney. My work would not have been possible without their help and support.

I am grateful to Professor Gerhard Schneider of Aalen University for enabling my work in Aalen and for his constant support. My sincere thanks go to Dr Alwin Nagel for supervision and guidance there, and for his trust in me and my project, and for all our discussions.

I owe my sincere thanks to the research teams at UNSW, especially to Professor Paul Munroe and Dr Charlie Kong from the EM Unit, and to the Materials Research Institute at Aalen University. Pars pro toto I thank Dirk Staudenecker and Günter Tilk for their constant help: I would not have been able to carry out all the experimental work without them. For their ‘open ears’ I would like to thank Timo Bernthaler, Wilfried Salzwedel, and Thomas Weidler at Aalen University, and thank especially Andreas Häger for his work during his internship at UNSW.

My gratitude goes to the staff at UNSW for their assistance and help during my stay here – to Lana Strizhevsky and Jane Gao in particular.

A great debt of thanks is owed to Robert Bosch GmbH for financial support during my work: I want to thank Dr Matthias Leonhardt and Jan Göhler especially for their assistance during mechanical testing there.

I would like to express my appreciation to Professor Michael Hoffmann and Thomas Waschkies from Karlsruhe Institute of Technology for providing freeze cast preforms.

Most of all I would like to thank my family for their attention, support and encouragement during the past years; and especially my girlfriend Jasmin, with the utmost gratitude for her patience throughout.

ii

Abstract In composite science, interface characteristics have a significant influence on overall composite properties, especially with regard to co-continuous structures. Interface design, therefore, plays a significant role in the manufacture of reliable composites with enhanced mechanical properties; copper , if added in small amounts to alumina preforms in the preparation process, enhances the overall composite properties considerably.

The aim of this study is to understand the underlying phenomena and to explain the enhancement of the mechanical properties for reliable fabrication of copper/alumina composites with an interpenetrating phase structure in a squeeze casting infiltration process. The copper used for infiltration has low content (OF-quality), melted in a reducing atmosphere and, hence, has poor wetting properties. The addition of copper oxide enhances the infiltration process which is not governed by the infiltration initiation, but results in reactions during leading to the formation of compounds in the microstructure. The aluminate phase is determined as brittle in nature by nanoindentation. The mechanical properties of the composite, however, are enhanced by a factor of two, exhibiting bending strength of up to 800 MPa for additions of as little as 2 wt% copper oxide with good reliability (Weibull modulus of 25.6). Fracture toughness and Young’s modulus are also enhanced to 9.6 MPa.m-0.5 and 220 GPa, respectively. These observed improvements are determined to be the result of reactions during infiltration, as the copper partly dissolves the aluminate phase leading to enhanced infiltration and interface bonding behaviour due to small levels of oxygen enrichment. Residual porosity is reduced and, consequently, mechanical properties enhanced. The dissolution of aluminate to fine-structured alumina phases remaining in the microstructure.

iii

Table of Contents

Certificate of Originality ...... i Acknowledgements ...... ii Abstract ...... iii Table of Contents ...... iv Symbols ...... vi 1 Introduction ...... 1 2 Literature Review ...... 4 2.1 Static Wetting ...... 4 2.1.1 Wetting in Metal/ Systems ...... 7 2.1.2 Manipulating the Wetting System ...... 9 2.1.3 Reactive Wetting ...... 11 2.2 Dynamic Wetting and Infiltration of Porous Media ...... 17 2.3 Preform Fabrication ...... 28 2.3.1 Freeze Cast Preforms ...... 29 2.3.2 Particle Preforms ...... 30 2.4 Pressure-assisted Infiltration Techniques ...... 32 2.4.1 Squeeze Cast Infiltration ...... 33 2.4.2 Gas Pressure Infiltration ...... 38 2.5 Interfaces in the Copper/Alumina system ...... 40 2.5.1 Different Phases and their Formation ...... 41 2.5.2 Characteristics of Phases ...... 49 2.6 Influence of Interfacial Chemistry on Mechanical Properties ...... 49 2.7 Mechanical Properties of Interpenetrating Phase Networks ...... 54 2.8 Summary ...... 56 3 Prelimary work, Hypotheses and Approach ...... 59 4 Experimental Procedure...... 63 4.1 Characterisation of Raw Materials ...... 63 4.2 Preform Processing ...... 68 4.2.1 Fabrication of Sintered Porous Particle Compacts ...... 68 4.2.2 Processing of Preforms for Model Investigation ...... 72 4.3 Composite Fabrication ...... 74 4.3.1 Direct Squeeze Cast Infiltration ...... 74 4.3.2 Gas Pressure Infiltration ...... 76 4.4 Characterisation Methods ...... 77 4.4.1 Permeability of Preforms ...... 78 4.4.2 Surface Preparation, Microscopy and Image Analysis ...... 79 4.4.3 Mechanical Properties ...... 82 4.4.4 Nanoindentation testing ...... 84 4.4.5 ...... 85 5 Results and Analysis ...... 87 5.1 Ceramic Preform ...... 87 5.1.1 Preparation of Ceramic Compound...... 87 5.1.2 Preform Fabrication ...... 89 5.1.3 Preform Characteristics ...... 96 5.1.4 Characterisation of Preforms for Simplified Investigation ...... 103

iv

5.2 Infiltration Behaviour ...... 106 5.2.1 Characteristics of Preform Infiltration ...... 106 5.2.2 Performing the Infiltration Process ...... 110 5.2.3 Instrumented Squeeze Cast Infiltration ...... 112 5.2.4 Infiltration at Constant Pressures ...... 119 5.3 Microscopic Investigation ...... 128 5.3.1 Structure of ...... 128 5.3.2 Interface Appearance ...... 141 5.4 Investigation of Interface Strength ...... 147 5.5 Characterisation of Phase Properties by Nanoindentation Testing...... 149 5.6 Physical Properties of Metal Matrix Composites ...... 156 5.6.1 Strength ...... 156 5.6.2 Elastic Modulus ...... 160 5.6.3 Fracture Toughness and Fracture Behaviour ...... 161 5.6.4 Thermal Conductivity ...... 170 5.7 Summary ...... 173 6 Discussion ...... 175 6.1 Preform Preparation and Phase Formation during Sintering ...... 175 6.2 Infiltration Behaviour ...... 181 6.2.1 Saturation of Porous Media ...... 181 6.2.2 Influence of Aluminate Phase ...... 184 6.3 Microstructure and Interface Appearance ...... 189 6.4 Aluminate in the Composite Structure ...... 194 6.5 Hypotheses revisited ...... 198 6.6 Summary ...... 200 7 Conclusions ...... 201 8 References ...... 203

v

Symbols Greek Symbols

ߙ Geometrical factor ߠ௠௜௡ Minimum contact angle ߛ Surface tension ߣ Thermal conductivity

ߛு௚ Surface energy of ߥ଴ Superficial velocity

ߛ௟௩ Interface energy of the liquid – ߩ௙ Fibre vapour interface Interface energy of the – Fluid density ߛ௦௟ ߩ௟ liquid interface

ߛ௦௩ Interface energy of the solid – ߩ௠ Density of melt vapour interface

ߛ௥௟ Interface energy of the liquid – ߩ௣ Particle density solid reaction product interface

ߛ௥௩ Interface energy of the reaction ߩ௥ Raw density product – atmosphere interface Amount of interfacial tension due Characteristic strength ߂ߛ௥ ߪ଴ to reaction

ߟ Viscosity ߪ஻ Splitting tensile strength

ߠ Contact angle ߪ௖ Strength of failure

ߠ଴ Initial contact angle ߪସ௉஻ 4 point bending strength

ߠௗ௬௡ Dynamic contact angle ߔ௖ Closed cell porosity

ߠ௘௤௨௜௟ Contact angle at equilibrium ߔ௢௣ Open porosity conditions

ߠு௚ Contact angle of mercury on ߔ௧௢௧ Total porosity ceramic

Latin Symbols

ܣ Area of cross section ܭ௥ Relative permeability

ܽ Crack length ܭ௦ Specific intrinsic permeability

ܽ௖ Critical crack length ݈ Length of fibre Infiltration depth ܿ௣௙ ܮ preform

ܦ Average diameter ݉ Weibull modulus

ܧ Elastic modulus ݉௅ Slope

ܧ௥ Reduced elastic modulus ݉௫ Mass (x=1,2,…) ܨ Load ܲ Pressure

vi

ܨ௖ Critical load ߂݌ Pressure drop Maximum applied load Threshold pressure ܨ௠௔௫ ܲ଴ Heat of reaction Applied pressure ߂ܩ௥ ܲ௔௣௣௟Ǥ

ܪ ܲ௖ Capillary pressure

݄ Height ܲ௜௡௙Ǥ Infiltration pressure

ܪ௠ Heat of fusion of melt ݌ሺܱଶሻ Oxygen partial pressure ܭ Permeability ܳ Heat quantity Fracture toughness Radius ܭூ௖ ݎ Interface toughness Surface roughness ܭ௖ ܴ௔

ܴ௘ Reynolds number ܸሶ Volume flow Equivalent radius of largest Fibre volume fraction ݎ௘௤Ǥ௠௔௫ ܸ௙ interspace

ܵ Saturation ܸ௜௡௙ Infiltration velocity

ܵ௜ Surface area per unit volume ܸ௠ Volume fraction of melt Surface area per unit mass of Particle volume fraction ܵ௦ ܸ௣ the particles

ܶ Temperature ܹ௔ௗ Work of adhesion

ݐ Time ܹ௜ Work of immersion

௠ Liquidus temperature ݖ Directionܶ

ܶ௣௖௧ Critical minimum preheating temperature

Abbreviations

CTE Coefficient of thermal expansion PFA Pore-forming agent EDS Energy dispersive X-ray SEM Scanning electron microscopy spectroscopy FIB Focused beam SPM Scanning probe microscopy MMC Metal matrix composite STEM Scanning transmission electron microscopy OM Optical microscopy TEM Transmission electron microscopy

vii

1 Introduction Composite materials in general possess the desirable properties of the combined materials; they offer a unique balance of physical and mechanical properties. Metal matrix composite (MMC) materials, for example, combine the ductility of a metal – due to their metallic matrix – with the high stiffness of the ceramic used as reinforcement. In MMC materials, therefore, the metallic matrix contributes some plasticity and thermal and electrical conductivity to the microstructure. As well as the reinforcement the composite material usually has good impact and erosion resistance, enhanced fatigue strength, a high degree of hardness and good resistance to aggressive environments. In addition to this suite of characteristics, MMCs provide higher strength and stiffness than the matrix material, excellent wear resistance and a lower coefficient of thermal expansion (CTE). As the metal matrix and the ceramic reinforcement have different physical, electrical, thermal and mechanical properties, the resulting composite properties can vary over a broad range, which offers a degree of design ability unusual in engineering materials. Thermal and electrical properties, for example, can vary from metal-like to ceramic-like behaviour by appropriate adjustment of reinforcement volume fraction, morphology and distribution.

A special group of composites are those with an interpenetrating network structure. Here, both phases can be defined as the matrix dependant up which phase is most abundant. In this work the description of metal matrix composite (MMC), rather than (CMC) has been used for the materials fabricated, regardless of ceramic volume content.

In interpenetrated network structures, the ceramic phase can be responsible for stiffness, whereas the metal phase retains some plasticity and, especially in the case of copper, offers high thermal and electrical conductivity. The interface is the frontier zone between these two phases and an essential part of MMCs, but potentially the weakest point in most metal/ceramic composite systems due to the chemical diversity that results in a relatively poor interface adhesion. The manufacture of such composites can be characterised by infiltration of the liquid metal into a solid ceramic preform with open cell

1

porosity; the infiltration process is directly related to the wetting properties of the system. Whereas systems with wetting behaviour will be infiltrated spontaneously, systems with non-wetting behaviour need to be force-infiltrated. Furthermore, systems with a wetting character show a better adhesion of the liquid to the solid than non-wetting systems. Thus the wetting is important for both the infiltration process and the interfacial properties of the resulting composite.

The overall mechanical properties of the resulting composite are often also limited by the interface adhesion. Bonding at the interface develops from physical or chemical interactions and interfacial frictional stresses and thermal stresses can occur from a mismatch of the thermal expansion coefficients of the reinforcement and matrix phases. Thus the main way to improve MMC properties is to understand and control the underlying interfacial phenomena; the transmission of thermal and electrical as well as mechanical properties throughout the whole composite can be of enormous importance when creating metal matrix composites by interface design.

The interface is the main factor in composite materials. By understanding the underlying phenomena of bonding and adhesion of the system with its specific interfacial phases and their influence, the composite, or closer the interfacial bonding character, can be characterised and designed. In this work, the Cu-Al2O3 system is characterised in terms of understanding the underlying phenomena in bonding of copper and alumina. The copper-alumina system of is of special interest due to its importance in industrial applications (electronic devices). Here the reliability of the copper-alumina joints is the challenge. Therefore, the significance of understanding the interfacial phenomena with designing thereof is obvious. This study aims to understand the interfacial behaviour in copper-alumina joints for reliable composite fabrication.

The work concentrates on preforms manufactured in a wet powder preparation process followed by drying, green body pressing and sintering. These preforms – ceramic structures with open cell porosity – consist mainly of alumina. A squeeze casting process is used for their infiltration with pure copper. Microstructure, infiltration progress and interface design are mainly

2

maintained (other than by the infiltration ) by designing the ceramic structure. To evaluate and understand the underlying phenomena of the latter composite material and its characteristics, the interface, infiltration progress and material properties are analysed.

Following this introductory section the thesis is organised into 6 chapters. In chapter 2 a literature review is presented. This includes a description of wetting in static and dynamic states and the infiltration of porous media. Furthermore a review of preform preparation is given as well as infiltration methods used for composite fabrication. This is followed by the interfacial aspects, most important in composite systems, with its influence on mechanical characteristics. Based on this review the Hypothesis and Approaches are outlined in chapter 3. The Experimental procedure is given in chapter 4, containing the preform manufacturing, composite fabrication by pressure- assisted infiltration and the characterisation methods used. In chapter 5 the results are given. This includes the preparation of ceramic preform as well as the infiltration processing. Furthermore the microstructure of the resulting composites is shown and analysed in microscopic investigations with specific regard to the interface region. The appearance of an aluminate phase has been additionally characterised separately. The mechanical properties of the composite materials are given. In chapter 6 the results are discussed in relation to the literature. This includes beside the preform fabrication, the infiltration and the microstructure of composite the role of aluminate during infiltration and in the composite structure. Finally, in chapter 7 the conclusions are given.

This work was undertaken as a part of Bosch’s industrial research. Hence, it was not possible to submit results from the work for immediate publication prior to assessment and protection of relevant IP. Some results have now been submitted for journal publication and are under review, as well as recently presented1.

3

2 Literature Review The literature review concentrates first on the wetting theory. This is followed by the possibilities of manipulating the wetting of liquid metals on solid . For infiltration the dynamic wetting, rather than the static wetting is of importance, hence, this is described in the following section with the infiltration of porous media. Preform fabrication and infiltration processes are described to outline the fabrication process of composite materials. The interface and its influence on mechanical characteristics are described before the literature review is summarised.

2.1 Static Wetting Wetting is a phenomenon that describes contact between a solid surface and a liquid. In explaining wetting the entire system has always to be considered, as solid, liquid and vapour phases present affect wetting properties. Wetting is generally expressed by a contact angle θ related to the interface energies of the system’s materials. If θ < 90° a system is described as ‘wetting’, if θ > 90° then it is ‘non-wetting’. Describing the wetting behaviour, therefore, can be done by measuring the contact angle of a liquid on a solid surface. Measuring wetting behaviour and contact angle can be done in sessile drop experiments, wherein a liquid drop is placed on a solid surface in a defined atmosphere, and the contact angle can be measured at the triple line metal/ceramic/ atmosphere in accordance with time and temperature2, 3. The two different modes, non-wetting and wetting, of a liquid drop in contact with a solid material are shown schematically in Figure 1.

(a) (b)

Liquid Liquid droplet droplet

Solid material Solid material

Figure 1: Schematic illustration of a system with (a) non-wetting and (b) wetting behaviour, where a liquid drop is in contact with a flat solid surface.

4

Basic wetting theory equations of wetting and work of adhesion are based on thermodynamic relations. Young4 was the first to describe a relationship between the contact angleߠ and the interface energiesߛ. His equation (also called the ‘Young and Dupré’ equation) balances the horizontal forces of surface tension on a liquid drop with a solid. It is expressed by:

 JJ slsv cosT0 Equation 1 J lv

Thus the characteristic contact angleߠ depends on the interface energies of the liquid-vapourሺߛ௟௩ሻ, solid-vapourሺߛ௦௩ሻ and solid-liquidሺߛ௦௟ሻ interfaces.

Figure 2 shows the relationship expressed by Young’s equation of the interface energies and the wetting angle:

Figure 2: Relationship between interface energies and wetting angle (from Diemer 5).

Young4 considered only the horizontal forces in this equation. Looking at the interface energies as conventional means the triple point is not in a steady state condition, because the single vertical force component inߛ௟௩ cannot be balanced byߛ௦௩ orߛ௦௟ – it would result in a vertical shift of the triple line. However, in metal/ceramic systems especially, it is proposed that the vertical forces can be overlooked because of the stiffness and strength of the ceramic substrate in comparison with the liquid melt3. Nevertheless, there are suggestions explaining an observed phenomenon of ridge formation6-8. It should be mentioned that Young’s equation is not without controversy9, but it is generally adopted.

5

The description of the relationship between interface energies and contact angle expressed by Young’s equation indicates that the contact angle is controlled by the variation of interface energies. This becomes necessary for wetting experiments and for applications where wetting plays a role. Whereas a reduction ofߛ௦௟ always results in reduction of the contact angle, the influence ofߛ௟௩ is contact angle-dependent. With contact angles of below 90° a reduction ofߛ௟௩ results in a lower contact angle; while the contact angle increases at initial angles of above 90°. In contrast, a decrease ofߛ௦௩ always results in an increase of the contact angle. Figure 3 describes a schematic view of this influence.

Figure 3: Schematic overview of how changes in interface energy result in changes to contact angle.

The work of adhesion Wad is another parameter to describe the wetting. It is in direct relation to the interface energies and defined as:

ܹ௔ௗ ൌߛ௦௩ ൅ߛ௟௩ ൅ߛ௦௟ Equation 2 and it characterises the work necessary to form a new surface area when separating a liquid/solid interface; whereas the interfaces solid/vapour and liquid/vapour have to be formed10.

6

The combination of equation 1 and 2 results in:

ܹ௔ௗ ൌߛ௟௩ሺͳ ൅ ‘• ߠሻ Equation 3

The work of adhesion is now directly related to the surface tension of the liquid

(ߛ௟௩) and the contact angle θ.

A special case in wetting is called ‘spreading’, or ‘complete wetting’. This

ݒ. Theݏare equal or lower thanߛ ݈ݏoccurs if the sum of the terms ofߛ݈ݒ andߛ contact angle becomes zero and the liquid spreads the solid surface.

2.1.1 Wetting in Metal/Ceramic Systems The work of adhesion in non-reactive metal melt/ceramic systems is subdivided into two different categories.

The first is related to the physical van der Waals forces and the second relies on chemical relations between the surface atoms of the phases being in contact2, 11. Non-reactive systems are characterised by positive reaction enthalpy of possible phase formation, whereas reactive systems show negative reaction enthalpies. In general, most metal/oxide ceramic systems are specified as non-wetting systems and hence exhibit a contact angle larger than 90°, as reported by Gallois12. Table 1 gives an overview of contact angles and work of adhesion of selected metal/oxide ceramic systems.

Table 1: Selected metal/ceramic systems with their specific contact angle θ and work of 2 adhesion Wad .

System Temperature Contact angle Wad Metal Ceramic [°C] [ °] [mJ/m2]

Au Al2O3 1150 135 308

Cu Al2O3 1150 129 461

Cu MgO 1150 133 422

Cu SiO2 1150 128 474

Ni Al2O3 1500 110 955

Sn Al2O3 1150 127 185

7

In describing wetting properties, many authors state that the concentration of oxygen plays a major influencing role (as mentioned above) so it is important to control – or at least record – the oxygen partial pressure13-15.

Pure metal/oxide ceramic interfacial energies are characterized by weak van der Waals forces and electronic interactions; for better understanding of this, and a description of the wetting of oxide ceramic substrates by liquid-metal melts, various models of observed phenomena2, 11, 13, 16, 17 are created. Naidich2 acts on the assumption that the basic interaction of a liquid metal and ceramic is between the metal and the oxide of the ceramic compound – this has been described in more detail by other authors. Batyrev et al.13 observe in their studies of interactions within metal/ceramic systems in-plane relaxation at the interfaces; for the Cu and Al/Al2O3 interface especially they find that the in-plane relaxation results in the rotation of the O triangle and expansion of the O–O bonds at the alumina surface. Johnson et al.16 find a chemical bond established between the metal d-orbital electrons and the non-bonding 2p-orbital electron of the oxygen anion of the Al2O3 surface in the Fe, Ni, Cu, Ag/alumina systems. Furthermore, Li17, 18 describes the adhesion between a metal and oxide ceramic with an electron transfer from the metal into the oxide valence band, which is not completely filled at high temperatures. The alumina surface can be treated either as Al- or O-terminated. Experimental work19 and ab initio studies20 investigating the bonding behaviour of metal on alumina dependent on the first layer of the alumina surface are available – seemingly in accordance with the authors quoted, the oxygen-metal bond tends to be stronger (to be further described in section 2.6).

Saiz et al.21 review existing data on wetting and work of adhesion of metal/oxide interfaces and deduce the existence of a plateau regime of contact angle and work of adhesion at intermediate oxygen partial pressures. For several metal/alumina joints they find an intrinsic wetting angle of 110°–130°, dropping to and falling below 90° at the high-p(O2) and low-p(O2) limits respectively, and state them to be potentially representative of other metal/oxide systems. However, it is generally acknowledged that oxygen content is a fundamental variable in metal/oxide systems; so the joints are of ternary

8

character, with oxygen activity being relevant2, 13-15. The influence of oxygen on the wetting angle and other aspects influencing the wetting will be focused on in the following sections.

2.1.2 Manipulating the Wetting System In general, metal/ceramic systems, especially if oxide ceramics are present, are of a non-wetting nature with contact angles above 90°12, 22. According to Young’s equation, various factors can influence and modify the wetting properties. The interface energies, dependent on system parameters such as materials (solid, liquid) and atmosphere (vapour), determine the contact angle. By influencing at least one of the three phases the contact angle will alter, as shown schematically in Figure 3.

Changing the liquid component to improve wetting is investigated by many authors2, 14, 22-26: they agree that with certain specific treatments it is possible to improve wetting, or even change non-wetting systems into systems with wetting characteristics. This is shown by Diemer et al.14 in the

Cu-Cu2O/Al2O3 system, where oxygen dissolved in the metal is adsorbed at the metal/vapour and metal/ceramic interfaces, respectively. This leads to improved wetting with contact angles as low as 70°. Other authors24-26 investigated the influence of CuO additions to and copper, and observe the same wetting improvement effects. Chaklader et al.27 observe a decrease in wetting angle down to 21° by adding up to 72 wt% CuO to copper on . Additionally, they report complete wetting of copper on single crystal alumina in air due to the complete oxidising of the copper drop within their sessile drop experiments. Another possibility of manipulating the wetting properties – besides alloying with copper oxide in both structures – is doping the metal melt with reactive elements leading to chemical reactions at the interface. Mortimer and Nicholas28 investigate the copper/graphite system and report a change of its non-wetting behaviour to a wetting system by adding Cr and V to the copper melt. The metal/alumina system is also investigated by other authors, who propose an improvement to wetting through to the addition of metallic to the liquid: Ti29-33 especially is seen as improving the wetting, but also the effects of B, Cr,

9

Mn, V and Zr2 are investigated. Table 2 gives an overview of the influence of Ti doping of different metal melts to the wetting on alumina:

Table 2: Effect of doping on the contact angle in selected metal/ceramic systems2.

Contact angle [°] System Pure melt Titanium addition of 2 at% 4 at%

Au / Al2O3 135 90 76

Cu / Al2O3 148 58 32

Ni / Al2O3 110 99 92

Sn / Al2Oe 127 49 42

Reactive wetting has been investigated predominantly by the addition of Ti, therefore the improvement of wetting through the addition of Ti will be described exemplified; it is largely found that the addition of such a reactive element acts in two ways when adding it to copper melts on alumina substrates. Its first beneficial effect on wetting comes from the liquid side of the interface, where Ti forms a layer with high concentrations through adsorption at the interface. Its second effect involves the modification of the solid side of the interface by local 2, replacement of Al2O3 with TiO, leading to the formation of an interfacial layer 30, which contributes to reactive wetting (further described in the next section). The influence of the release of free energy by the reaction

ௗ௜௦௦௢௟௩௘ௗ Equation 4݈ܣʹ ଶܱଷ ֜ ͵ܱܶ݅ ൅݈ܣௗ௜௦௦௢௟௩௘ௗ ൅݅ܶ͵ due to the dissolution of alumina in the alloy, contributes only about 30% to the total reduction of contact angle, as investigated by Kritsalis et al.30. The authors directly compare both wetting experiments by investigating the Cu-Ti/Al2O3 system, taking into account the investigation by Naidich2 of the Cu-Ti/TiO system; and they state that the major contribution to wetting comes from the formation of a continuous layer of TiO. Further investigations of the effect on wettability show the influence of Ti being stronger than that of Zr in Ag- Cu alloys on alumina (for example)34.

10

Mortensen and Jin35 divide treatment of the ceramic substrate to improve wetting into two possibilities:

(i) Surface coating with a metal that may destroy the oxide layer of the liquid or react with the liquid, leading to reactive wetting; and

(ii) heat treatment of the ceramic component in a defined atmosphere leading to desorption of vapour components at the ceramic surface before the liquid is applied.

There are some studies covering the effects of surface anisotropy of alumina on the influence of wetting by molten metal liquids. Nevertheless, Shen et al.36 investigate the influence of the surface orientation of single crystal alumina on wetting through molten and copper. Measuring the contact angles and work of adhesion in a slightly reducing atmosphere, the crystal plane orientation shows no significant influence on the Cu/Al2O3 system; whereas for the Al/Al2O3 system, a noticeable dependence is seen. This behaviour of 37 38 Cu/Al2O3 systems has also been reported by Vikner and by Ownby and Liu .

As mentioned, the oxygen activity – and thus the oxygen content – significantly influences the wetting behaviour of alumina using liquid copper. Many authors investigate the wetting improvement by alloying the copper melt with oxygen – i.e., copper oxide additions. Some14, 15, 39 investigate the wettability of alumina through copper and its dependence on the oxygen partial pressure. In general, the results are comparable: starting at very low values and increasing the oxygen partial pressure, the metal melt improves its wetting properties on alumina due to the absorption of oxygen of the atmosphere. Backhaus-Ricoult40 evolves a model, based on Gibbs’ absorption theory, to predict the contact angle dependence on oxygen activity in the surrounding atmosphere that fits well with wetting experiments in the literature14, 38, 39.

2.1.3 Reactive Wetting Although most metal/oxide ceramic systems are non-reactive, reactive systems should also be considered. Some of the aforementioned effects are in combination with the formation of a reactive compound or layer between the melt and the ceramic substrate: this reactive layer, formed before, after or in

11

conjunction with the sliding of the triple line, affects the wetting of the system. This means that there are chemical changes at the interface and, therefore, resultant changes to the specific surface tensions; and the apparent free reaction enthalpy has to be considered. Aksay et al.41 and Naidich2 were the first to propose that the enthalpy released in the closer area around the triple line is responsible for the spreading of the liquid on the solid material. Naidich2 divides the work of adhesion into two terms, thus:

ݑ݈݅ሻ Equation 5ݍݑ݈݅ሻ ൅ܹ௔ௗሺ݊݋݊Ǧ݁ݍ௔ௗ ൌܹ௔ௗሺܹ݁ calculating, for the interfacial region:

ݑ݈݅ሻ Equation 6ݍݑ݈݅ሻ ൅ܹ௔ௗሺ݊݋݊Ǧ݁ݍߛ௦௟ ൌߛ௟௩൅ߛ௦௩ െܹ௔ௗሺ݁

ݑ݈݅ሻ is calculated by the integration of free enthalpy ofݍThe term ܹ௔ௗሺ݊݋݊Ǧ݁ the formation of a reaction compound between the starting conditions and the final equilibrium condition along the interface. A rapid reaction and the formation of a monolayer between the substrate and the liquid melt are assumed. In regard to the accordance between the calculated and experimental obtained values, Naidich2 proposes a strong chemical metal/ceramic reaction as necessary for improving the wettability.

According to Aksay et al.41, the major contribution of enthalpy of formation to wetting improvement takes place at the beginning of spreading.

Starting from the initial wetting angle ߠ଴, it decreases towards an intermediate minimum wetting angle ߠ௠௜௡, to increase again until the equilibrium conditions are reached with the wetting angle ߠ௘௤௨௜௟. This behaviour is displayed schematically in Figure 4. Therefore the time-dependence of the spreading power is subdivided into different stages. As proposed by Laurent et al.42, the initial contact angle ߠ଴, minimum contact angle ߠ௠௜௡ and the equilibrium contact angle ߠ௘௤௨௜௟ can be calculated using:

'J r 'Gr cos min cosTT 0   Equation 7 lv JJ lv

12

whereby ‘• ߠ௠௜௡ is the minimum wetting angle, ‘• ߠ଴ the wetting angle at equilibrium conditions without any reaction taking place, ȟߛ௥ the contributions of reduced interfacial tension due to reaction, ȟܩ௥ the heat of reaction due to formation of an interfacial phase, and ߛ௟௩ the surface tension of the liquid metal melt.

Figure 4: Schematic display of the different conditions in reactive wetting of a substrate (S) by a liquid metal melt (M). A reaction product (P) is formed along the interface22.

Landry et al.43 investigate reactive metal/ceramic systems: in their experiments on the wetting of copper on vitreous , the initial contact angle is measured at 137°, decreasing with the addition of a reactive alloying element such as Si or Cr. Neither a minimum contact angle nor an inflection point is observed. This relates to the formation of a reaction compound and, hence, to the interface energies of the new system. They compare the contact angles of the related non-reactive systems – the wetting of copper on the reacted compounds SiC and CrC, respectively – and state that the final contact angle of a reactive metal/ceramic system is nearly equal to the Young contact angle of the metal on the reaction compound. The slight differences are attributed to surface roughness of the reactant. Comparisons with the results of Espié et al.44 establish good agreement.

13

In contrast with Aksay et al.41 and Naidich2, some authors44, 45 propose the influence of interfacial energy changes to be predominant for systems with limited or moderate reactivity; so the free energy release of the reaction is negligible. They suggest, therefore, that the changes in interfacial structure due to adsorption processes and phase formation at the interface are the decisive factor in reactive wetting. Espié et al.44 investigated the wetting of , quartz and alumina by CuPd-Ti alloys, varying the reaction energy term ȟܩ௥ of equation 7 while keeping the interface energy term ȟߛ௥ constant. Kritsalis et al.45 investigated the contact angle of NiPd-Ti alloys on alumina changing the interface energy term ȟߛ௥ of equation 7 at constant reaction energy term ȟܩ௥; this is possible because of the free energies of the formation of these different phases (TiO and TiO2) being roughly the same. They conclude that the reduction of the wetting angle is due to the formation of different reaction products rather than the negative free energies. Kritsalis et al.45 propose, in agreement with Espié et al.44, that in systems with low or moderate reactivity an interfacial reaction should be regarded mainly as a way to modify in situ the metal/oxide interface.

In reactive wetting on ceramic substrates by liquid metals, Zhou and de Hosson46 propose that the volume change of the ceramic induced by the reaction phase is the dominating factor of wetting behaviour – that by increasing the volume of the ceramic during reaction taking place the wetting is improved, whereas it declines when decreasing the ceramic volume. They explain this behaviour through cavities that may be formed during reaction, due to shrinkage of the ceramic substrate. These surface cavities may hinder a further spreading of the liquid drop, as shown schematically in Figure 5. In contrast, with a volume increase after reaction and formation of a dense thin layer at the interface, the wetting can be improved due to reaction taking place at the triple line, leading to a dynamic balance between the three interface tensions: thus the reaction can improve the wettability. For example they refer to the Al/SiO2 system, where a volume decrease due to the formation of alumina and solution of in the melt can be observed. In contrast, the Ti/Al2O3 system represents an example

14

of volume increase from the formation of TiO2 and titanium such as

TiO and Ti2O3.

(a) (b)

Figure 5: Surface cavities are formed in front of the wetting triple line due to the shrinkage of the ceramic substrate after reaction (a), which may hinder the spreading of liquid. A positive volume change (b) may promote the wetting behaviour, until a dynamic balance of the three interface tensions is reached at the triple line46.

The simple criterion of volume change proposed by these authors is challenged 47 by Shen et al. , who investigate intensively the Al/SiO2 system: they observe an improved wetting enforced by chemical reaction. They state that wetting is at first a surface phenomenon and does not seem to have an essential relationship with the volume change of the ceramic substrate at all; further, that the wettability improvement attributed to the chemical reaction, and that contributed by the Gibbs free energy involved in the chemical reaction, are two different concepts. In addition, they propose that the volume decrease of the ceramic substrate can enhance the wetting, as the molten aluminium can penetrate along the cracks and small channels produced by the volume shrinkage of the ceramic – as schematically shown in Figure 5.

If the change in interface energy due to the reaction ȟߛ௥ is the dominating factor in the reactive wetting, there are two possible configurations at the triple line of the metal/ceramic interface, as shown in Figure 6:

15

(a) (b)

Reaction layer θ Reaction layer θ Substrate Substrate

Figure 6: Different configurations in front of the triple line in reactive wetting after equilibrium is reached, according to Espié et al.44, with the growth rate of reaction product being slow (a) or fast (b) relative to the flow rate of the liquid drop.

Considering case (a), the diffusion rate of the reacting components – and hence the growth rate of the reaction product – is slow relative to the flow rate of the liquid drop. The reaction product will reach, but not extend, the triple line; hence the liquid in the periphery of the drop will remain in contact with the unreacted solid. In this case, the equilibrium contact angle ‘• ߠ௘௤௨௜௟ is expressed by

 JJ rlsv cosTequil Equation 8 J lv where ߛ௥௟ corresponds to the interface energy of the liquid – solid reaction product interface.

In the configuration displayed in Figure 6(b), the diffusion rate of the reacting component in the solid is fast, relative to the flow rate of the liquid drop: after equilibrium is reached the reaction layer exceeds the triple line of the droplet. In this case, the equilibrium contact angle is given by

 JJ rlrv cosTequil Equation 9 J lv with ߛ௥௩ expressing the interface energy of the reaction product/vapour interface.

The copper alumina system with Ti additions is investigated by Naidich2 and Kritsalis et al.30, who show the contribution of wetting improvement as being twofold. The first contribution to improve the wetting involves the adsorption of

Ti at the Cu/Al2O3 interface; and the second involves the reaction with the

16

substrate. By comparing their results with the results of Naidich2, Kritsalis et al.30 propose the former being responsible for about 30% of the improvement; whereas the major contribution to wetting comes from the modification of the solid side of the interface by the formation of a continuous layer of TiO, in this specific investigation. In general, it can be stated that the wetting is enhanced as the metallic character of the interfacial phase increases. An increase of wetting accompanies the interfacial phase changes from Al2O3 to Ti2O3, TiO2,

TiO or even sub-stoichiometric TiO0,86 from wetting angles of 129° to 72° at 1150°C in vacuum2.

There is no definite model generally describing the influence of reactivity on wetting behaviour in metal melt/ceramic systems. The enthalpy of formation and the changes in the interface are the two factors leading to changes in surface tensions: which is the dominating factor is controversial, and depends on the particular system. In dynamic infiltration systems, where liquid metals are forced under pressure into a porous medium, non-reactive and reactive wetting are even more difficult to model. Nevertheless, evaluation of system variables and parameters of interest of the specific systems can be managed, helping understanding of the underlying phenomena.

2.2 Dynamic Wetting and Infiltration of Porous Media The aim of processing metal matrix composite with interpenetrating networks is to force the metal melt into a porous preform, especially if the system is of a non-wetting nature. This infiltration process consists of two steps:

(i) the initiation, where a certain capillary or threshold pressure has to be applied to overcome the capillary backpressure, and (ii) the molten metal flow into the capillaries of the preform.

The critical parameter in dynamic wetting that governs the wettability of a solid

35 by liquid is the work of immersion ܹ௜ which is given by

ܹ௜ ൌߛ௦௟ െߛ௦௩ Equation 10

If ߛ௦௟ > ߛ௦௩, work is required to generate the solid/liquid interface; hence the system is of a non-wetting nature. Thus, infiltration by liquid of a porous body in

17

the non-wetting case will require a minimum external pressure. This threshold

35 or capillary pressure ܲ଴ can be written in terms of the work of immersion as

௜ Equation 11ܹכ ଴ ൌܵ௜ܲ where ܵ௜ is the surface area of interface per unit volume of the surface to be covered with the liquid. ܲ଴ can be related to the contact angle ߠ at the liquid/solid interface by using a combination of equations 1, 10 and 11; the threshold or capillary pressure is expressed by:

ߠ Equation 12•‘ כ ߛ௟௩כ ଴ ൌܵ௜ܲ

Assuming a specific morphology of the porous medium, the surface area ܵ௜ can be calculated using relevant parameters. For instance, if the medium is comprised of packed spherical particles with an average diameter ܦ, ܵ௜ takes the form 48

D **6 Vp Si Equation 13 VD p )1( where ܸ௣ is the particle volume fraction and ߙ is a geometrical factor introduced to describe deviations from sphericity, surface roughness and actual particle size distribution. This factor may also affect the capillary pressure ܲ଴. As an alternative, it may be possible to measure the specific surface area ܵ௦. So ܵ௜ can be written as48

Vp SS U psi ** Equation 14 1Vp where ܵ௦ is the specific surface area of the particulates (surface area per unit mass of the particulates) and ߩ௣ the density of the particulate. Then, by combining equations 12 and 14, the capillary pressure takes the form

Vp 0 SP ps lv *** cosTJU dyn * Equation 15 1Vp

18

This equation allows the dynamic contact angle to be determined, once the surface area ܵ௦ and the threshold pressure ܲ଴ are measured, by knowing the system’s wetting behaviour – expressed by the wetting angle ߠ and surface tension ߛ௟௩.

Two different methods for determination of the surface area are used in the case of packed ceramic particulates, as reported by Garcia-Cordovilla et al.48. The most common is the Brunauer, Emmett and Teller (BET) technique, based on absorption. The diffraction method applied to the particulates before compaction shows inaccurate results for applications where wettability is evaluated49. Molina et al.50 investigate the validity of the above equation to determine the threshold pressures of Al into SiC compacts with bimodal grain size distribution. They find that using the linear rule of mixture and BET measurements of the surface area of particles offers a reasonable interpretation of the experimental results investigated.

Regarding the different infiltration abilities of gaseous media and liquids, especially with respect to surface tensions, mercury intrusion porosimetry is an alternative method which may be used for determination of the surface area of porous media. Most of the mercury porosimetry measurements in MMC fabrication have been performed solely to determine the pore size distribution of particulate preforms with ceramic volume contents of more than 50%, as reported by Travitzky and Shlayen51 and Bahraini et al.52, 53. In some cases, preforms with much lower ceramic volume contents are measured, as investigated by Cardinal et al.54, with ceramic fractions of 20% performed on hybrid preforms build-up of Al2O3 platelets and 35 wt% Al2O3 short fibres. The volume of intruded mercury at a specific pressure is the result of penetration into cavities larger than a particular size. In mercury intrusion measurements, Washburn’s equation is applied:

4 Hg cosTJ Hg D  Equation 16 Pappl.

19

where ߛு௚ is the surface tension of mercury, ߠு௚ the wetting angle of mercury on ceramics and ܲ௔௣௣௟Ǥ the applied pressure. This describes the dynamic equilibrium between the external forces pressuring a liquid into a capillary of diameter ܦ and the internal forces repelling entry into that capillary.

The infiltration of a liquid into a porous medium in non-wetting systems will require a minimum external pressure. Once exceeding the capillary back pressure, the infiltration starts; and the threshold pressure can be determined by the capillary pressure as:

2 cosTJ P lv c r Equation 17 where ߛ௟௩ is the liquid/vapour surface tension, ߠ the contact angle of the liquid on the solid and r an average capillary radius. Washburn’s equation relies upon the more general capillary law given in equation 17.

The commonly-used sessile drop experiments to determine the wetting behaviour of a liquid on a solid replicates only poorly the wetting conditions encountered in MMC fabrication55. In infiltration, the three-phase contact line generally moves at a velocity of 10-1 cm*s-1 inside a porous medium. The oxide layer covering the metal surface and the influence of chemical interaction at the interface are found to be very different from those observed in sessile drop experiments by Mortensen55; so experimental procedures using ceramic particulates are put in place to determine the wettability in infiltration48. The molten metal is forced by means of a gas pressure into a packed bed of powder held at the same temperature as the metal. The threshold pressure ܲ଴, necessary to initiate the movement of melt, is recorded as a measure of wettability and expressed by equation 15. In the experiments of Asthana and Rohatgi56 the volume fraction of the ceramic powder bed ranges between 50% and 60%. They investigate the infiltration of Al-alloy into porous SiC compacts dependent on the chemical alteration of the SiC particulates; and observe better infiltration ability in terms of infiltration length for Cu-coated particles than for plain ones accompanied by a lower measured threshold pressure and hence

20

lower estimated contact angle. To determine the threshold pressure of infiltration, the square of infiltrated height is plotted against the applied pressure

57 ܲ௔௣௣௟Ǥ. as exemplarily shown in Figure 7 for the investigations of Molina et al. on infiltration of SiC preforms with Al and Al-Si alloy. The data can be fitted to a straight line in agreement with Darcy’s law48 and the pressure axis gives the initial threshold pressure ܲ଴.

Figure 7: Plots of the square of the infiltrated height h2 vs. applied pressure P for pressure infiltration in Al(Si)/SiC systems from Molina et al.57 Lines are fittings to the data by means of Darcy’s law.

Bahraini et al.52 describe a method of determining the threshold pressure of liquid metal in situ, analogous to the mercury intrusion porosimetry but at elevated temperatures – up to 1500 K. They find with their “high temperature molten metal porosimetry” agreement with prior data using different methods, such as drainage curves58, 59 or infiltrations at different pressures48 for evaluation of capillary data.

Apart from packed ceramic powder beds, investigations using porous sintered preforms are rather rare. Mortensen and Michaud60 report infiltration of Saffil fibre preforms with ceramic fraction ranging from 10% to 25% using constant gas pressure. Infiltration with pure aluminium results in contact angles of 102° to 111°. The results are much closer to the sessile drop experiments conducted in ultra-high vacuum conditions than to those measured in air or lower vacuum.

21

Hilden and Trumble61 investigate numerically the capillarity in packed spheres. They analyse hexagonally packed spheres, either close-packed or separated by 1/10 of their radius. The capillary pressure ܲ௖ at any point on the meniscus is directly proportional to the curvature of the meniscus, as given by the Laplace-Young equation:

§ 11 · ¨ ¸ Pc J lv ¨  ¸ Equation 18 © rr 21 ¹ where ߛ௟௩ is the liquid vapour surface tension and ݎଵ and ݎଶ two orthogonal radii of curvature at a point on the meniscus – the curvature of meniscus on a pore depends on the shape and size of the pore, as well as on the contact angle ߠ. Analytical solutions describing such surfaces of constant curvature have been obtained only for menisci in simple-shaped capillaries, like parallel plates, cylinders, parallel fibres or toroid pores. For more complex pore shapes, a solution for the meniscus shape and curvature typically requires numerical methods61. Regarding this, most attempts at predicting the capillary pressure and corresponding meniscus shape and position in packed particulates are based on simplifications of pore geometry. Common examples are the hydraulic radius approximation62, 63, the toroid pore of Purcell64, the parallel cylinders of Mayer and Stowe65, the extrapolation of parallel cylinders to account for diverging/converging pore structures by Mason and Morrow66, Marmur’s axially- symmetric tube with sinusoidal radius variation67 and ‘Pore-Cor’ software68, 69. A comparison made by Hilden and Trumble61 of these models with the numerical analyses shows most being in significant error: the authors state that for ߠ < 90° the toroid pore model of Purcell gives close approximation, whereas for ߠ > 90° the model of Mason and Morrow gives a reasonable approximation. (Note: these numerical analyses of capillarity apply strictly to liquid penetrating a single layer of spheres, and do not apply generally to 3D arrays of spheres.)

Hilpert evolves a (semi)-analytical model to describe the dynamic contact angle in horizontal70 and non-horizontal71 capillary tubes. Whereas Washburn72 assumes the contact angle to remain constant during the flow, obtaining an

22

analytical solution for the travel time of the gas/liquid interface as a function of interface position, Hilpert71 slightly modifies this solution: he considers two functions – the power law model and the power series model, deriving new (semi)-analytical solutions for liquid infiltration into inclined tubes under the action of gravity. These solutions account for the dynamic contact angle depending on the capillary number, either in the form of the power law or the power series. Considering the power law (݂ሺܥܽሻ ൌߙܥܽఉ, where ܥܽ is the capillary number, ߙ and ߚ model parameters) with ߚ = 1, Hilpert71 obtains an analytical solution for travel time as a function of interface position, as Washburn did for constant contact angle. The (semi)-analytical solutions typically allow a more efficient calculation of interface position and velocity than other approaches, as stated by Hilpert71. Additionally, significant differences may occur between Washburn’s solution and his model prediction, accounting for dynamic contact angle. Further models of the dynamic contact angle are described by Joos et al73, who write of the kinetics of wetting in a vertical capillary of Newtonian fluids, where the fluid wets spontaneously the interior part of the capillary. Cox74 describes the phenomenon of the dynamic in the spreading of fluid on fundamentally; and De Coninck75 reviews the molecular dynamic simulation to explain the dynamics of wetting of liquid on solid surfaces, which does not directly reflect the phenomena in infiltration.

Fluid flow in a porous medium and, hence, in preform infiltration depends on the medium penetrated and the fluid properties. Henry Darcy establishes empirically that the flow rate of water through a porous and permeable medium is proportional to the penetrated height, the prerequisite being a constant cross- sectional area: this is the basis of a simple model for infiltration of porous

media. The superficial velocity ݒ଴ in the flow direction ݖ is calculated with respect to the fluid viscosity ߟ and the pressure gradient ݀ܲΤ ݀ݖ at the infiltration front:

K dP v  s 0 K dz Equation 19

23

The coefficient of proportionality in Darcy’s law is called “the specific intrinsic

76 permeability”, ܭ௦ . The superficial velocity can also be related to the actual

velocity ݀ݖΤ ݀ݐ by use of the ceramic particle volume fraction ܸ௣. Appropriate combination and integration leads to the expression of the square of infiltration depth ܮଶ 48:

2 2 stK L 'P Equation 20 K Vp )1(* with ȟܲ being the pressure drop in the infiltrating liquid, which can be calculated by

ȟܲ ൌ ܲ௔௣௣௟Ǥ െܲୡ Equation 21

Even though Darcy’s law is empirically-based, it represents a simplification of the general equation of viscous fluid flow governed by the Navier-Stokes equation. The simplification is achieved by assuming incompressible fluids, laminar flow and unidirectional saturated flow. It should be noted that the idealizing of the macroscopic assumption does not reflect the microscopic sense of flow of a pressurised fluid in a porous media48. Particulate preforms used to produce composite structures generally have highly complex internal void geometries: this complexity and capillarity during infiltration is too complicated to enable realistic prediction of the metal flow path at the microscopic level of the individual particles making up the preform. Darcy’s equation is limited to saturated flow; so complete saturation of pores is assumed before further penetration takes part. Fluid flow through porous medium can be characterised by the Reynolds number, which is given by

U ** rv Re m 0 K Equation 22 with ߩ௠ being the density of liquid metal and ݎ the effective pore radius. Darcy’s law is valid for laminar flow – hence as long as the Reynolds number does not exceed a value between 1 and 10 76, which is for normal preform infiltration, the flow remains within the laminar regime48, 52, 56.

24

The process of infiltration is often challenging because the phenomena governing the infiltration process are complex and difficult to access. Physical, mechanical and chemical phenomena may interact, including a multiphase flow of liquid and air in a porous medium, heat and mass transfer due to solidification processes, equilibrium of mechanical forces, and chemical interfacial reactions between the matrix and the reinforcement. Analytical and numerical solutions for infiltration of porous bodies are provided and compared with experimental data for

(i) unidirectional infiltration under constant applied pressure, including non-isothermal infiltration60, 77,

(ii) taking preform deformation into account78, 79 and

(iii) isothermal infiltration with respect to capillary phenomena59.

Modelling with numerical tools such as finite element and finite difference methods is reported for non-isothermal saturated flow, including solidification aspects80, 81. All these models, apart from Michaud et al.59, consider only the case of saturated flow, either by ignoring any capillary pressure drop or by using the slug flow assumption that implies the infiltration front to be flat. It is shown for low applied pressures; and if preform structure exhibits a broad pore size distribution, this assumption breaks down and penetration takes part in a gradual manner, with large pores are filled first59, 82.

For most relevant metal/reinforcement systems, isothermal metal infiltration is similar to drainage in soil mechanics58. During drainage, wetting water is displaced by non-wetting air in a porous soil. In MMC infiltration, air generally constitutes the wetting phase and metal the non-wetting. Non- saturated flow through porous media and drainage phenomena is dealt with in the soil mechanics literature76, 83. Based on soil mechanics, Dopler et al.58 develop a model for isothermal infiltration of ceramic fibres based on capillary phenomena. The relationship between local pressure and non-wetting fluid velocity is classically described by Darcy’s law (equation 19). When neglecting gravity, laminar flow in a porous medium is valid in the following form:

׏ܲ Equation 23כܭݒ଴ ൌെ

25

where ݒ଴ is the superficial velocity of the non-wetting phase, defined as the volumetric flow rate of metal per unit area, ܭ the permeability tensor and ܲ the pressure. The permeability can be expressed as a function of three independent terms:

* KK K rs K Equation 24 where ܭ௦ is the specific preform permeability tensor, ܭ௥ the relative permeability, varying with the preforms saturation between 0 and 1, and ߟ the dynamic fluid viscosity. The saturation ܵ indicates the ratio of filled void space to initial void space, and is defined in similarity to soil mechanics by

Vm S Equation 25 1V f with ܸ௠ and ܸ௙ being the volume fractions of metal melt and ceramic fibres, respectively. Saturation in soil mechanics is in general expressed by a function of pressure. The functional relationship is measured by establishing a drainage curve for the system under consideration as a function of the applied pressure

ܲ௔௣௣௟. The curves obtained can be fitted to phenomenological equations, introducing a threshold pressure (ܲ଴) that must be overcome to initiate infiltration, and a shape factor ߙ58, 84.

1 S 1 2 2 Equation 26 D appl  PP 0 )(*1

This equation is valid for ܲ௔௣௣௟ ൐ܲ଴. The curve shape described by the shape factor value ߙ varies with the pore size distribution, the size and type of reinforcement and the wetting behaviour. Dopler et al.58 propose an equation for modelling the infiltration behaviour of Saffil fibre preforms:

26

dS wP * KK V ’ (**)1( rs ’P 0)* f dP wt K Equation 27

In most metal melt/ceramic preform systems only the volume fraction ܸ௙ and the dynamic viscosity ߟ are known. Thus

(i) the drainage curve ܵሺܲሻ using the parameters ܲ଴ andߙ,

(ii) the specific permeability tensor ܭ௦ and

(iii) the relative permeability ܭ௥ need to be determined experimentally58. Once these system parameters are determined, the function ܲሺݔǡ ݐሻ can be found by solving the nonlinear partial differential equation 27 with given initial and boundary conditions. The authors report agreement between the numerical analyses and experiments in terms of both infiltration kinetics and porosity distribution.

Molina et al.57 investigate the influence of contact angle measurement for infiltration of liquid metals into porous ceramic preforms. Contact angles used in preform infiltration can be measured by sessile drop experiments or derived from drainage curves or capillary law. It has been shown by Trumble85 that in the case of close-packed spheres the contact angle has to be smaller than 50.7° for spontaneous infiltration to occur. Therefore the threshold for wetting – on a flat surface 90° – can be much lower in an infiltration experiment than in sessile drop experiment, as the complex geometry of the preform plays an extraordinary role. Thus, much of the information relevant to the infiltration process is missing, and data obtained from the sessile drop experiments for wetting systems must be treated with caution – as stated by Molina et al.57. The validity of the simplest form of Darcy’s law, in equation 20, relies upon the slug- flow assumption – known to fail in many cases, contributing a non-uniform, gradual filling. However, many experimental data obey Darcy’s law, suggesting that the use of an effective constant permeability is still meaningful57; but the interpretation of drainage curves by means of Brooks-Corey’s law tends to show similar results. Additionally, these contact angles are close to those from sessile drop experiments shortly after initiation, corresponding with a metal surface

27

covered by a discontinuous layer of oxide as investigated for aluminium alloys57. And Molina et al. say the Brooks and Corey model is justified only when the pore or particle size distribution obeys a power law. The processing of pressure- assisted infiltration techniques, which further describes the infiltration phenomena, is reviewed in section 2.4.

2.3 Preform Fabrication The reinforcement of metal matrix composites produced by casting methods can vary in shape and structure86: composite materials reinforced by dispersion particles, platelets, non-continuous (short) fibres, and continuous fibres as well as hybrid reinforcement composed of particles and fibres are produced87, 88. The former elements are also used to produce composites by stir casting89, 90: in this process, the reinforcing phases are generally distributed and suspended in the molten metal or alloy via high-energy mixing or another appropriate process91. Another production method is to infiltrate a porous preform with molten metals or alloys. Preforms are made of the other reinforcing elements, like short or long fibres and particles, as well as combining both fibres and particles – “hybrid preforms”. Ceramic preforms make the basis for metal/ceramic composites with interpenetrating or co-continuous network structure92. They can either enhance the component locally at the position where higher modulus or better tribological behaviour, e.g, is essential; or fully, with the ceramic preform’s being infiltrated, showing the entire component. Because of the infiltration process the porosity of the preforms must remain open and the volume fraction is typically required to be between 20% and 70% to provide sufficient mechanical stability to withstand infiltration forces while maintaining sufficient permeability91. Closed porosity cannot be infiltrated; hence it remains as un-infiltrated pore or un- reinforced area in the composite structure. For non-wetting metal melts a minimum pressure, the threshold pressure, has to be overcome to realise the infiltration process as described in detail in section 2.2.

Alumina possesses good mechanical strength, thermodynamic stability and exhibits a non-wetting nature in contact with most metals; most of the published work on oxide ceramic preforms used for MMC fabrication is based on alumina. MMC design starts with the preform design: the reinforcement can

28

be made of whiskers, short or chopped fibres, long or continuous fibres, or particles93. Preforms made of fibres, such as Saffil® (and others) were mainly infiltrated with aluminium melts. Fibre preforms differ in long and short fibre preforms94. Due to the multiple processing steps of fibre fabrication, the preforms made of chopped fibre materials are by factor of 30 to 50 more expensive than particulate preforms. Continuous fibre preforms result in very anisotropic behaviour; whereas isotropic behaviour becomes more evident with long and short fibres, whiskers and particulate preforms, if distributed homogenously.

2.3.1 Freeze Cast Preforms Three-dimensional preforms have been manufactured by the freeze-casting process for a long time95 – high-concentration aqueous ceramic or metallic suspensions being used mostly. By freezing the slurry in a suitable form, the green body is frozen into the desired shape. Before sintering the green body has to be dried; and depending on its constitution, a freeze-drying process is applied to avoid changes in shape and homogeneity. This process was originally developed as an alternative to injection-moulding, to avoid high injection pressures and problems of the polymer or wax matrices normally used at the debinding step. One aim was to get very dense preforms; another, porous preforms with homogeneous dispersed isolated pores for the building industry (e.g.). Big ice crystals leading to inhomogeneous macro pores must be avoided during the freeze process to achieve homogeneous preforms structures96.

Preforms with controlled inhomogeneous pore structure can be manufactured by unidirectional freezing97-100. The ceramic volume fraction can be controlled easily in the range of 15%–85%. The principle of freeze casting of aqueous solutions is shown in Figure 8, aided by a phase diagram of water. By unidirectional freezing of ceramic slurries, the ice forms a lamellar structure in solidification direction, pushing the ceramic particles into the interlamellar spaces. After freeze-drying a porous ceramic structure remains, with pores in the size and shape of the prior ice lamellae. After sintering a porous ceramic preform remains101.

29

Solid Liquid

freezing

sublimating Pressure (freeze drying) sintering

Gaseous

Temperature Figure 8: Schematic view of the phase diagram for water, with the steps of the freeze- casting process used for preform fabrication101.

The potential of these structures is seen in filters, porters for catalysts, sensors and bioreactors. Mattern102 is the first to investigate the potential of metal infiltration of preforms with directional pore structure: he infiltrates alumina freeze-cast preforms with the AlSi12. Investigations of the interface show weak or no bonding between both phases, and he proposes infiltrateable porosity, within the ceramic lamellae, as suitable to increase material strength; or a reactive coating for increasing interface and hence materials strength. The interface aspect with pre-treatment of ceramic structure due to coating and subsequent heat treatment is investigated with the elastic properties of such composites103.

2.3.2 Particle Preforms One of the most common ways to fabricate MMCs is to press ceramic particles into a cavity and infiltrate the packed bed with a molten metal: because of its simplicity, many research groups concentrate on this method48, 53. If the particles are assumed to have a spherical shape with the average diameter D, the specific surface area ܵ௜ of the packed particles is calculated according to equation 13. As no binder is added to the particles, a removal of the pressed

30

compact out of the pressing cavity can result in compact breaking. Thus no consistent preform shape is fabricated using this method.

To achieve consistent ceramic preforms that exhibit sufficient strength to withstand the subsequent infiltration step, the particles have to be (partially) sintered. This technique involves influencing the total porosity (to some extent) by changing the powder compaction and sintering temperature, though it is difficult to achieve porosities above 50%. It is possible to slightly alter the metal ligament diameter by changing the size of the particles104-106: the particle size itself sets an upper limit. It is not possible to influence the pore structure independent of the ceramic microstructure.

A very flexible method to achieve porous ceramic parts is to use a sacrificial pore-forming agent as a placeholder for pores. In this method pyrolysable organic agents are added to the ceramic powder during processing. In a burn-off step before sintering the particles are pyrolysed, leaving pores of corresponding shape and size in the ceramic particulate, as shown schematically in Figure 9:

Figure 9: Schematic of pore formation using a pyrolysable pore-forming agent (PFA). The PFA is added to the ceramic slurry. After pressing and heat treatment the organic agent evaporates leaving pores in the ceramic structure.

A large variety of pore-forming agents (PFAs) is reported by others, including carbon fibres107 or flakes108, 109, starch110, 111, polyethylene110 and cellulose102. This method is very flexible, as the total porosity can be controlled in the range of 0 to approximately 80% by the amount of added PFA, depending on the densification behaviour of the initial particulates.

31

2.4 Pressure-assisted Infiltration Techniques Pressure-assisted infiltration techniques are necessary when systems are of a non-wetting nature. Other than special systems – such as the Lanxide process112 and metallization of ceramic particles with copper or nickel113, 114, or mixing high amounts of metal with the ceramic115, 116 – the non-wetting nature of metals such as copper on oxide ceramics prevents spontaneous infiltration of metal melt into the preform. Thus a differential pressure between the inside of the capillaries and the metal melt has to be applied to force the liquid into the preform pore cavities. In general, infiltration methods are divided into two different ways of liquid transport: constant pressure, realised in gas pressure infiltration48, 105, 109; and the constant , realised in squeeze casting107, 117, 118 or high pressure die casting119, 120. The former is widely used in gas pressure infiltration of particulate compacts and is the single method published to determine dynamic wetting angles, as reviewed by Garcia-Cordivilla et al.48. In constant flux infiltration mode, experimentally done using direct squeeze casting and high-pressure die casting set-ups, the molten metal is forced into the ceramic network in a relatively short time. Bahraini et al.52 recently proposed an apparatus for the direct measurement of capillary forces in the infiltration process of MMCs: relevant systems can be directly investigated through gas pressure infiltration as analogous with mercury intrusion porosimetry, but with elevated temperatures of up to 1500 K and infiltration pressures up to 20 MPa.

In all published infiltration experiments, the preforms are preheated to increased temperatures – depending on the matrix metal, up to 1000°C. Nagata and Matsuda121 suggest in their work on metal/particle composites a critical minimum preheating temperature ܶ௖௣௧, which is independent of the size and morphology of the reinforcement, melt temperature and pressure – as seen in equation 28:

.0 233 U ** H mm cpt TT m Equation 28 U ff ** cV pf

The influencing factors are those of the melt, like liquidus temperature

ܶ௠, melt density ߩ௠ and heat of fusion of melt ܪ௠; as well as those of the fibre

32

preform, like volume fraction ܸ௙, fibre density ߩ௙ and its specific heat capacity

ܿ௣௙. The critical preheating temperature is important in constant flux infiltration, where the temperature is usually below melt liquidus. In gas pressure infiltration, the preform is heated synchronously with the matrix metal; thus the temperature of the porous compact is normally the same as the liquid melt.

2.4.1 Squeeze Cast Infiltration The process of squeeze casting involves three essential steps, as Ghomashschi and Vikhrov122 show in their review. First a specified amount of molten metal is poured into a die cavity of a permanent mould; then the die is closed and the liquid metal is pressurized, as fast as possible so as to prevent premature solidification. Afterwards pressure is held on the metal until complete solidification; and this application of pressure during solidification has an influence on solidification behaviour and the resulting microstructure. In accordance with the Clausius-Clapeyron equation, the high metallostatic pressure, which can rise up to 200 MPa, leads to an increase in the of the metal, as reported by Catterjee and Daas123.

(a) (b)

Figure 10: Production of cast composite materials by (a) direct squeeze casting method and (b) indirect squeeze casting method: 1, pouring of the molten metal into the pressure casting device, followed by 2, the squeeze casting process124.

33

Depending on the gating system, squeeze casting is divided into direct and indirect squeeze casting: in Figure 10 both systems are schematically shown. In the direct mode (Figure 10a), there is no gating system at all: the metal melt is poured directly into the die cavity, which is subsequently closed and pressurized122, 125, 126. In this mode, simple geometries without undercuts are realizable. In contrast, in indirect squeeze casting the die is filled and pressurized through a gating system (Figure 10b): the filling velocities in direct and indirect mode are below the threshold of turbulent flow, preventing entrapment of oxide films in film-forming liquid metals126.

The metal is cooled and solidified under pressure in squeeze casting. The pressure forces close contact of the melt with the die material, resulting in a large heat transfer to the colder die. The heat transfer coefficients as reported by Lee et al.127 enhance from 4.5 kW/m2K for gravity die casting to 125 kW/m2K for squeeze casting, with 100 MPa applied pressure. In fact, the overall cooling rate is governed mainly by the interfacial heat transfer between the melt and the die. The solidification time decreases asymptotically with an increasing heat transfer coefficient, where coefficients of more than 20-40 kW/m2K have little influence on reducing the solidification time further128. With the enhancement of the cooling rate of the casting, solidification is reflected in small secondary dendrite arm spacing and fine-grained structures. Furthermore, the pressure application reduces gas porosity size and compensates feeding defects, thus reducing voids in the microstructure126-128.

The first commercial application of the squeeze cast for infiltration was the fabrication of aluminium alloy diesel pistons containing alumina short fibres, done by the Toyota Motor Corporation129. The porous preform, made of alumina short fibres, was inserted into a preheated die and infiltrated with an aluminium alloy melt by a direct squeeze casting process where the pressure was applied in a hydraulic press. The composite pistons possess better performance attributes than the non-reinforced.

The work of Cappleman et al.130, who infiltrated Saffil ߜ-alumina fibre preforms with aluminium alloys by squeeze casting, is concerned with the fibre/matrix interface. In no case do they observe an intermediate phase

34

between the fibre and the matrix; and their conclusion is that there is no chemical reaction in the composites due to the rapid cooling of the melt when using the squeeze casting infiltration process. This is in contrast to the 131 publication of Levi et al. , where formation of spinel MgAl2O4 on alumina fibres is observed when alumina fibres are immersed in containing aluminium alloy. An interesting point is the suppression of formation at the melt surface during this fast infiltration process. As molten aluminium and its alloys show an extraordinary high affinity for oxygen, a thin oxide layer would be expected to continually reform at the infiltration front as displayed in Figure 11. When assuming an outlet behind the infiltration front connected to the atmosphere, the calculations by Cappelman et al.130 show that a monolayer oxide formation can keep pace with rapid infiltration up to about 10-2 to 10-1 m/sec.

Oxide Melt Fibre layer

Air

v

Figure 11: Schematic of oxide layer formation at the advancing infiltration front in fibre (f) preform infiltration130.

Depending on the specific surface area of the preform, the surface layer cannot be formed on all fibre surfaces when the outlet of the preform is closed or the preform is enveloped within the melt prior to infiltration. The calculation shows that in preforms composed of more than about 50% of fibre by volume, a monoxide layer cannot be formed on all fibre surfaces.

35

Long et al.119, 120, 132 evolve a model to describe pressurized metal melt infiltration in fibre preforms using an indirect squeeze casting process; because of geometrical assumptions, it can only conditionally be transformed quantitatively to other preforms but qualitatively. They propose that in liquid metal infiltration two hydrodynamic parameters are associated with the infiltration quality: infiltration speed ܸ௜௡௙ and infiltration pressure ܲ௜௡௙. The effect of both parameters during unidirectional preform infiltration can be correlated with the geometrical conditions of the preform and the physicochemical properties of the melt-fibre contact system – like surface tension ߛ௟௩, viscosity ߟ and contact angle ߠ of the melt flow to the fibres – by:

lv *cosTJ 40K *Vinf Pinf 2   PL 0 Equation 29 req.max D Vf )1(*3 where ݎ௘௤Ǥ௠௔௫ is the equivalent radius of the largest voids in the ceramic phase,

ܮ the infiltration depth, ߙ a geometrical factor and ܸ௙ the fibre volume fraction. According to their model, theoretically described in Long et al.120, 132, the infiltration starts when the external pressure exceeds the minimum capillary resistance of the preform to the melt penetration. This is followed by a stable infiltration stage, characterised by a linear pressure/time relationship, until the melt penetrates through the preform. The last infiltration step to the pre-selected maximum pressure is dominated by air compression and the filling of microporosity. Figure 12 shows the pressure/time relationship of this model, including its three different stages. For evaluation of the threshold pressure the aforementioned plotting of the square of infiltration height against the applied pressure is applicable with the data measured.

36

Figure 12: Schematic infiltration curve (pressure/time relationship) in unidirectional infiltration of a ceramic preform at constant speed120.

The effect of the infiltration speed determines not only the time required to achieve full infiltration, but also the infiltration gradient, the penetration-through pressure and the saturation degree of the infiltrated preform; so it affects the infiltration quality via preform deformation if the pressure drop exceeds the compression strength of the ceramic preform120 and the infiltration defects through determining the amount of entrapped air in the infiltrated preform and the associated non-infiltration defects.

Etter et al.133 investigate the strength and fracture toughness of graphite aluminium composites with interpenetrating phases. The graphite preforms, with only 14.5% porosity, have been infiltrated by an indirect squeeze casting process with adjusted parameters, leading to a carbide-free interface. The preheating of the preform is moderate, and due to fast cooling rates and short cycle times of about 60 s, including placement of the preform in the chamber, infiltrating of the preform and ejection, Al4C3 formation can be avoided. Mechanical properties are increased by a factor of two for the flexural strength and fracture toughness, respectively.

There is limited literature describing the squeeze cast infiltration process of a porous ceramic preform in copper/ceramic systems. Prakasan et al.134

37

investigate the coefficient of thermal expansion of copper-infiltrated alumina- silicate and graphite fibre preforms prepared by squeeze cast infiltration. They observe a lower CTE of the composite materials than for pure copper, with the carbon fibre-reinforced MMC exhibiting lower values of CTE than alumina- silicate fibre-reinforced MMC at volume fractions of 30% of reinforcement. Xing et al.135 investigate the interfacial reactions appearing during the squeeze cast infiltration process of SiC preforms with copper.

2.4.2 Gas Pressure Infiltration Gas pressure infiltration is the predominant method of producing MMCs of continuous fibre preforms: the main reason is the possibility of realising a low pressurization rate, necessary to prevent fibre breakage and destruction of the preforms of compacts48. To prevent gas inclusions in the preform (or in the resulting MMC) it must be evacuated before infiltration; so the entire process has to be performed in an autoclave. The preform and the melt are heated simultaneously in the chamber; after the melting of the metal, the chamber is filled with pressurised inert gas such as argon or helium – the maximum pressure that can be applied depends on the chamber used. Once the capillary pressure is reached, the melt starts to infiltrate the preform.

Garcia-Cordovilla et al.48 use unidirectional infiltration to infiltrate particulate compacts enclosed in tubes. The experiments are performed using long contact times – in the range of minutes: in reactive systems this period exceeds the incubation time generally necessary to initiate reaction. The reaction leads in some cases to a partial blocking of the preform porosity and hence prevention of further infiltration, as shown by Molina et al.136. To maintain the saturation grade in the filled cavities when liquid to solid transition of the matrix takes place, the solidification shrinkage has to be directed toward the non-reinforced region; so infiltration experiments are carried out using a chill in the furnace to enforce directional solidification137, 138. Knechtel et al.104 infiltrated alumina with pure copper using a gas pressure infiltration process: they use an extraordinarily high processing temperature of 1350°C and gas pressure of 15 MPa, but no infiltration of preforms with a pore diameter below 0.15 μm was observed. They find pore size-dependent debonding of copper to alumina in the

38

resulting fracture surfaces of infiltrated samples. Coarse grain matrix shows a correspondence of good debonding with low fracture strength, whereas at lower grain sizes limited debonding is observed but no interfacial reaction resulting in oxygen coordination of the alumina being shown by energy loss spectra. Nevertheless they propose that there is a possibility of oxidation of copper leading to enhanced bonding.

Bahraini et al.53 investigate the effect of volume fraction, particle geometry and capillary parameters on drainage curves, and compare the results with the expression proposed by Brooks and Corey83, using a gas pressure infiltration process. Drainage curves are obtained by mercury intrusion at room temperature and Cu infiltration at 1200°C by plotting the volume fraction of the non-wetting Cu-melt in the porous preform versus the pressure difference between the melt and the atmosphere in the pores. They observe good agreement of the semi-empirical model with drainage curves relevant to the infiltration of packed alumina powders by molten copper or mercury respectively.

With the exception of some authors’ work 58, 59, 139 the most common modelling of preform infiltration deals with abrupt infiltration fronts48, 77, 81, 139, 140. Michaud and Mortensen139 summarise model structures of preform infiltration in polymer, metal and ceramic matrix composites, slug flow assumption and the model of gradual infiltration. In gradual infiltration the molten metal gradually fills the smaller pores as local pressure increases. In the last few years drainage curves, adapted from soil science, have come more into consideration53, 136 in relation to the semi-empirical model of Brooks and Corey83.

In general, fluid flow in a porous medium and, hence, in preform infiltration depends on the medium penetrated and the fluid properties. Darcy establishes empirically that the flow rate of water through a porous and permeable medium is proportional to the penetrated height at a constant cross- sectional area. This is the basis of a simple model of infiltration of porous media, though simplified by assuming saturated flow. However, it does not reflect the infiltration progress at the microscopic scale, and the prediction of infiltration, especially for higher saturations, is rather poor. Garcia-Cordovilla et

39

al.48 nevertheless propose it to be applicable for their infiltration problem. Even though the simplification provides rather poor accuracy for infiltration modelling with high saturation rates, the infiltration initiation appear clearly, as it is not affected by the infiltration progress. In accordance with Darcy’s law, the infiltration depth ܮ is a square root function of the applied infiltration pressure and can be described as shown in equation 2048.

2.5 Interfaces in the Copper/Alumina system Most materials used in technical applications are polycrystalline: they consist of small crystallites (grains) that meet at internal interfaces. These two- dimensional defects strongly influence many technically-relevant properties; therefore, macroscopic properties depend on the microstructure of the materials19. In composite materials, where different material phases meet at a macroscopic interface – in the case of laminates, for example – the bonding properties are as important as the internal interfaces themselves: a lack of interfacial bonding results in delamination of the different phases. In metal matrix composite materials, especially where there is an interpenetrating network structure of particulate preforms, the interface – the surface area of the reinforcement – is large, compared with its volume; thus the interface region plays a specific role in the properties of resultant MMCs.

The copper/alumina system has been extensively investigated for more than two decades, experimentally141-147 and theoretically13, 16, 20, 146-148. The practical reason for investigating the Cu/Al2O3 interface is that it is used in microelectronic devices, where interfacial delamination is a severe problem for reliability149, 150. Additionally, this system acts as a model for fundamental understanding of interfacial adhesion between dissimilar materials. Strong adhesion is needed, which is greatly affected by the presence of an interfacial layer (or reaction phase) and impurities; thus fundamental knowledge of these factors is a prerequisite for designing the properties of the Cu/Al2O3 system. As reported by Yoshino and Ohtsu149, there is widespread speculation that a reaction layer forms at the copper/alumina interface, which appears to have promoted in the work of Chaklader et al.15, 27; whereas other research groups already have proven and detected the interface layer141. Yoshino151 discusses

40

the lack of visualisation of the interface layer at this stage by pointing to difficulties in the preparation of thin films good enough for microscopy, due to the extreme difference in hardness of these phases; but acknowledges later149 that the reaction phase, if it exists, is either cuprous aluminate (CuAlO2) or cupric aluminate (CuAl2O4) or both, in the range of tens of nanometres. Today’s researchers can more easily investigate thin films with appropriate preparation methods. As reported by Scheu et al.152, a useful method of determining the interfacial structure between different materials is transmission electron microscopy (TEM). Investigations of the atomic structure of interfaces and possible reaction phases using TEM or Z-contrast imaging, chemical composition down to the sub-monolayer level using energy dispersive X-ray (EDS), and electron energy-loss spectroscopy (EELS) or the electronic structure and bonding characteristics by analysing electron energy-loss near- edge structures (ELNES) in the EELS spectra, are all examples of available methodologies153. In all these investigations materials and sample preparation remain supremely important.

2.5.1 Different Phases and their Formation Wetting experiments were made by Chaklader et al.27 as mentioned (section 2.1.2). They observe a strongly decreasing wetting angle between copper and sapphire when increasing the oxygen content of copper by alloying it with CuO. They attribute the decrease in contact angle to the presence of an interfacial layer of CuAlO2. In general, they believe the wetting to be dependent on the oxygen content within the system.

The ternary Cu/Al/O system accommodates different phases. In addition to the binary oxide phases, aluminium oxide and copper , CuO and

Cu2O, two stable ternary copper/aluminium oxides – CuAlO2 and CuAl2O4 – are 154-156 possible . The aluminate phase CuAlO2 has a rhombohedral structure in contrast with the spinel structure of CuAl2O4. To establish interface phases for property characterisation, diffusion bonding141, 157-160, eutectic bonding141, 149-151, 161, 162 and brazing163 processes are applied. In the first, both components are pressed together while heated to 50–98% of the melting temperature of the lower melting material, unless no liquid phase occurs. The pressure is

41

maintained until bonding163, 164. A drawback – interface porosity, which weakens the interface strength – can arise when employing this process164, 165. In eutectic bonding a reduction in the melting point of copper, if in contact with oxygen, is used. The phase diagram of the system Cu/O for oxygen concentrations up to

25 wt% is shown in Figure 13. Figure 14 shows the system CuO/Al2O3 and 166 Cu2O/Al2O3 investigated experimentally by Misra and Chaklader .

Figure 13: The phase diagram of the system Cu-O with oxygen concentrations of up to 25 wt%167.

42

166 Figure 14: Binary phase diagram of the system CuO/Al2O3 and Cu2O/Al2O3 .

Eutectic bonding occurs when oxidised copper is present at the interface, whether formed in situ by controlling the atmosphere or by directly applying a paste of copper oxide. The oxidised copper melts locally at 1066°C, joining the alumina. In this process the oxygen concentration at the interface is already increased while joining the materials, leading to Cu2O segregations and . By regulating the oxygen partial pressure or a downstream heat treatment, the interfacial chemistry of diffusion-bonded materials can be varied5, 168. The influence of oxygen content in the atmosphere on the formation of phases can be seen in Figure 15, where the Cu/Al/O system is shown with dependence on the oxygen partial pressure. Brazing processes always work with brazing alloys, which may have a copper basis, and contain active elements such as titanium or chrome due to the poor wetting behaviour of copper on alumina.

43

Figure 15: Equilibrium stability diagram of the system Cu/Al/O in dependence of the oxygen partial pressure, calculated by Rogers et al.168.

Two ternary phases, copper/aluminium oxide in aluminate and spinel structure, are stable in the Cu/Al/O system, as mentioned. Trumble156, who investigates the aluminate formation theoretically in the case of diffusion bonding, describes a minimum threshold activity of oxygen in the metal required for the aluminate formation at the interface. This minimum amount of oxygen is less than the limit of of oxygen in the metal and, furthermore, the required amount is temperature-dependent. Figure 16 shows the minimum concentration of oxygen needed in solid Cu for {Cu, CuAlO2, ߙ-Al2O3} equilibrium as a function of temperature.

44

Figure 16: Stability diagram of oxygen in the Cu solid solution for the {Cu, CuAlO2, ࢻ-Al2O3} equilibrium and the {Cu, Cu2O} equilibrium as a function of temperature: concentrations calculated by Trumble156.

In equilibrium conditions, solid Cu and ߙ-Al2O3 can only coexist with the ternary copper/aluminium oxide in the aluminate structure CuAlO2, not with the oxide in 156 the spinel structure CuAl2O4 . In the copper/alumina system there have been several reports of aluminate formation5, 14, 27, 141, 149, 156, 161, 168. A ternary phase diagram of the system Cu/Al/O at 1273 K is given in Figure 17. Jacob and Alcock155 electrochemically measured the standard free energy of the formation of CuAlO2 and CuAl2O4: they show that at sufficiently low oxygen partial pressures, solid metallic copper equilibrates with aluminate and alumina. The results have been taken into consideration for the calculations of Trumble156, shown above. Yoshino and Shibata150 show, in their study of copper/alumina bonding, that there are two reactions thermodynamically favourable to forming an interfacial layer of either aluminate or spinel: aluminate will be formed of cuprous oxide, whereas the spinel will more likely be formed of cupric oxide, as shown in equations 30 and 31:

ଶ Equation 30ܱ݈ܣݑܥ ଶܱଷ ՜݈ܣݑଶܱ൅ܥ

ଶܱସ Equation 31݈ܣݑܥଶܱଷ ՜݈ܣݑܱ ൅ܥ 45

These predictions fit well with the ternary phase diagram shown in Figure 17 where the aluminate structure appears within the Cu2O/Al2O3 quasi-binary phase diagram and the spinel structure with higher oxygen concentrations within the CuO/Al2O3 quasi-binary phase diagram, respectively.

Figure 17: Calculated ternary phase diagram of the Cu/Al/O system at 1273 K – phase relations of high oxygen and high aluminium concentration omitted156.

Beraud et al.141 report on the bonding nature in the copper/alumina system for different bonding processes – diffusion and eutectic. They observe an increase in bond strength with increase in bonding temperature, time and pressure of diffusion bonded joints measured by shear tests. In addition, they find surface roughness influencing the overall bond strength. With surface roughness of the alumina of up to about ܴ௔ ൌ 0.5 μm, the bond strength increases, with the fracture being located at the metal/ceramic interface. This behaviour alters for increasing surface roughness, with the fracture occurring in the bulk alumina, resulting in a drastic decrease of bond strength. An optimal parameter set is given, resulting in maximum bond shear strength of 50 MPa. Liquid phase bonding results in a maximum strength of 140 MPa, measured by tensile

46

testing. This is attributed to the interfacial presence of CuAlO2 which acts in two ways:

(i) enhancement of wetting15, 27 resulting in fewer voids or pores in the interface region and

(ii) bonding nature improvement at the interface.

As mentioned, the reaction between copper and alumina is strongly dependent on the reactive atmosphere, where oxygen plays an important role – in agreement with the phase diagram in Figure 15 and the stability diagram in

Figure 16. In their investigations of the formation of the CuAlO2 phase, Beraud 141 et al. observe no crystallographic relationship between Cu2O, Al2O3 and

CuAlO2; and the aluminate formation takes place only when Cu2O is present, with the reaction described as fast – in the order of about 0.1 μm within a few minutes. This leads to their conclusion that Cu2O acts as a promoter for aluminate formation.

Kim and Kim161 investigate the effect of an interfacial reaction product on strength in eutectic bonded copper/alumina. Bonding is performed in a tube furnace at 1075°C, in flowing nitrogen gas between oxygen-free copper and alumina sintered to 98% theoretical density. Bonding time is prolonged up to 24 h and the interfacial region examined via microstructural investigations and bonding strength, using a 4-point bending test setup. They observe linear time- dependent aluminate formation expressed by the thickness of the aluminate layer. The oxygen content dissolved in the melt is about 5 at% – higher than the equilibrium concentration of oxygen of 1.6 at% at 1075°C, resulting in undissolved Cu2O remaining in the Cu-Cu2O eutectic melt. A schematic model for the CuAlO2 formation has been proposed, wherein Cu2O precipitations will form CuAlO2 by a solid state reaction at the interface when in contact with alumina. This is in accordance with Beraud et al.141, who propose that the aluminate formation needs the formation of an initial Cu2O layer at the interface. Because of further observations of the formation of a continuous interface layer of aluminate, Kim and Kim161 state that this could not be the only reaction for

CuAlO2 formation. They propose a theory where alumina has some solubility in oxygen containing copper (Cu[O]): then the reaction of aluminate formation will

47

take place at the Cu[O]/Al2O3 interface where CuAlO2 precipitates once the Cu[O] is saturated with Al and O. This solution of alumina is thought to be very slow, which controls the slow growth rate of CuAlO2. This observation is 141 contrary to the observations of Beraud et al. , who observe a very fast CuAlO2 growth rate in eutectic bonding. Nevertheless Kim and Kim161 observe an increasing strength with bonding time up to a value of 250 MPa at 20 h bonding time: with longer bonding time, the strength was observed to decrease slightly.

Several other authors describe the aluminate formation, when copper is in contact with alumina, with different formation conditions. Fujimura and Tanaka169 use bonding conditions of 1025°C–1194°C in air or nitrogen atmosphere for their polycrystalline ߙ-alumina samples deposited with copper thin films. They investigate the formation of different phases near the interface by in situ x-ray observation, and state the aluminate formation to be taking place from 1138°C to 1194°C in air whilst obeying a parabolic rate law. Yi et al.170 use pre-oxidised copper discs with high purity recrystallized alumina for eutectic bonding in a vertical tube furnace at 1066°C–1078°C in argon atmosphere. Their thermodynamic analysis of phase equillibria in the Cu-Al-O system, which is in good agreement with their experimental results, shows the existence of a -7 five-phase equilibrium at the invariant state T = 1075°C, p(O2) = 5.5 x 10 bar with the phases solid Cu[O], liquid Cu[O], solid CuAlO2, solid Al2O3 and oxygen gas. Scheu et al.171 investigate the interface between copper and alumina single crystals. Bonding is performed by solid state diffusion bonding at 1000°C for 4 h in ultra-high vacuum with a subsequent annealing step at 1000°C for 96 h in controlled atmosphere with oxygen partial pressures of p(O2) = 0.02 and 0.32 Pa. They found for the higher oxygen partial pressure, besides the aluminate layer being slightly thicker, aluminate needles of several millimetres length and up to 1 μm thickness, which are highly twinned. Hasegawa et al.172 use cold-rolled high purity copper foil, with thickness varied between 10 and 100 μm, sandwiched between two single crystal alumina plates for their diffusion bonding process at 1048°C for 24 h under vacuum condition and an applied pressure of 3 MPa. They found an increase in fracture resistance with copper thickness and state that decohesion occurs by a ductile rupture

48

mechanism involving plastic void growth accompanied by interface rupture between the voids.

2.5.2 Characteristics of Phases In the copper/aluminium/oxygen system several phase compositions are possible, as stated – the main phases being alumina and copper. Their characteristics and mechanical properties, as well as physical and thermal properties, are well known, and will not be discussed here. The more interesting phases are the copper oxide phases and the copper/aluminium oxides. Although there has been much work done in characterising properties of aluminate and copper oxide, there are few results of mechanical characterisation. As reported earlier, authors characterising the aluminate phase as interfacial describe its behaviour and failure in a brittle manner. The structural configuration of the copper oxides is cubic for the red/brown- 173 appearing Cu2O and monoclinic for the black CuO phases, respectively .

Whereas copper/aluminium oxide by means of the aluminate phase CuAlO2 is 154, 174, 175 light grey and of rhombohedral structure , the CuAl2O4 phase is of spinel structure. Of more interest is the preparation of CuAlO2 thin films with delafossite structure, as they exhibit optical transparency and p-type conductivity. More investigations have been done regarding the electrical properties of aluminate (which will not be extensively discussed). Transparent conductive oxides (TCOs) are normally n-type semiconductors, whereas aluminate is known to show p-type semiconductivity – a special feature176. Several authors175, 177-180 investigate this material’s thermo-electrical properties.

Copper oxide (Cu2O) exhibits p-type conductivity, too – albeit its appearance is strongly colourful, being red-brown173, 177.

2.6 Influence of Interfacial Chemistry on Mechanical Properties The presence of an interfacial phase – or at least interfacial bonding – is necessary to exhibit good mechanical properties in the resulting composite. This is in agreement with the investigations of several authors141, 156, 181, 182, who establish that metal/Al2O3 joints, together with the metals Fe, Ni, Cu, Ag prepared in an oxygen-rich atmosphere, often exhibit higher fracture strength and toughness than samples produced in a reducing atmosphere. This is

49

explained by the formation of reaction products at the interface and high oxygen partial pressure conditions. Johnson and Pepper16 show that a direct chemical bond can be established between metal and alumina – specifically between metal and the oxygen on the Al2O3 surface. The bond is primarily covalent, and its strength decreases in the aforementioned series Fe, Ni, Cu, Ag due to the increasing occupation of antibonding orbitals established by the metal/oxygen interaction. A more detailed study of the atomic structure of the bonding nature is done by Zhang et al.20. In general, the bonding properties (specifically the strength) of the interface depend directly on the atomic conditions there, as investigated experimentally and theoretically by Rühle et al.19.

Different phases at the interface exhibit different interfacial properties. Several authors investigate the copper/alumina system’s bonding nature (see also section 2.5) and interfacial properties. Yoshino151 investigates the role of oxygen in bonding copper to alumina. He observes no interfacial phase formation, but an increase in bonding strength, expressed by strength, with increasing oxygen content dissolved in the copper. As good bonding strength is often related to an interfacial reaction phase, he attributes the lack of interfacial phase to his analytical methods. In accord with Chaklader et al.27, who report the formation of CuAlO2 at the interface of oxygen-rich copper to alumina with oxygen content above 0.18 wt%, he states the interfacial phase to be possible, as in these experiments the oxygen concentration certainly reaches the eutectic concentration of 0.4 wt%, even if not well analysed. The bond strength is affected twofold by the appearance of cuprous oxide particles and voids induced from the oxide formation, resulting in a maximum of the overall bond strength due to the described mechanisms. While the increase of oxygen leads to an increase in bonding strength, it simultaneously promotes the formation of cuprous oxide, resulting in more voids at the interface; this also causes a reduction of bonding area and hence reduces overall bonding strength.

In further research, Yoshino and Ohtsu149 prepare eutectically-bonded joints with CuAlO2 as an interfacial phase to investigate the influence of an interface on bond strength in the Cu/Al2O3 system. Samples are made by indirect bonding of copper to alumina with a prepared aluminate surface, which

50

has been formed by printing 5 μm CuO paste on the alumina surface and calcinating it at 1077°C. They observe, with the resulting aluminate interfaces of about 10 μm thickness, a somewhat lower bond strength compared with the directly-bonded materials containing no CuAlO2 at the interface, as mentioned151, but a better resistance to embrittlement and thermal . The authors relate this to the brittleness of the aluminate phase and the porous formation of this layer, resulting in less bonding contact area but providing channels for the water vapour and absorbing thermally induced stresses.

Contrary to this, Reimanis et al.159 report in their investigations into the influence of interfacial Cu2O and CuAlO2 on crack behaviour an increase of bond strength if aluminate is present at the interface. They form and investigate interfaces of only one discontinuous phase – either cuprous oxide or aluminate.

In the case of Cu2O, which forms a needle-like structure at the interface, they observe crack growth promoted at the Cu2O/Al2O3 interface, showing weaker bond strength between Cu2O and Al2O3 than between Cu and Al2O3. This appears to contradict the earlier studies of Sun and Driscoll162 and Yoshino151, who find the Cu/Al2O3 interface to be weaker than the Cu2O/Al2O3 interface in eutectic bonding. Reimanis et al.159 relate their observation to poor lattice- matching between Cu2O and Al2O3, as well as to the thermally-induced stresses resulting from different coefficients of thermal expansion (CTE) and the brittle behaviour of the Cu2O needles. Further results show that interfaces with aluminate formation exhibit a strong interface, as also reported by others141, 149, 161 . In contrast with Cu2O, which forms only at the copper side of interface, the

CuAlO2 needles grow into both copper and alumina phases, resulting in a rough interface and hence pinning both phases together. The bond strength of CuAlO2 and Al2O3 is relatively strong compared with that between copper and alumina, resulting in crack-pinning at the interface. Additionally, the authors suggest that an extra processing step to remove the Cu2O – always present in commercial processes of Cu/Al2O3 bonding – via reduction could increase the bond strength, rating the observed weakness of Cu2O/Al2O3 bonding.

51

The observed and reported contradiction of bond strength behaviour regarding discrete phases is explained by Reimanis et al.159 as differences in testing methods leading to different crack modes and phase angles. This is in agreement with the observation of Wang183, who investigates the interfacial fracture toughness in the copper/alumina system. He observes an increase of interfacial fracture toughness as the loading phase angle increases, especially if the load mixture is high; thus the fracture toughness of the interface is higher as mixed mode loads are applied, as normally happens in real applications. So the investigations of other authors of the Cu/Al2O3 system regarding the interface bond strength might not be direct comparable, due to differences in testing procedures leading to different crack modes and phase angles, as indicated by Reimanis et al.159. Another reason why direct comparison of the results of Yoshino151 and Sun and Driscoll162 is difficult is the difference in bonding processes. Whereas Reimanis et al.159 used solid state bonding, the other studies examined liquid state-bonded samples. Furthermore, neither study satisfactorily clarifies whether no aluminate is apparent at the interface between cuprous oxide and alumina. Typically, the interface region in eutectic bonding rarely contains only one interfacial phase. Beraud et al.141, for example, report the presence of CuAlO2 at the Cu2O/Al2O3 interface. As has been reported by 161 Kim and Kim , Cu2O is a prerequisite for aluminate formation at the interface. With the formation being a diffusion-controlled process resulting in a slow growth rate of aluminate, the thickness or size of phase in the study of Sun and Driscoll162 might be below their detection limit for the given bonding time. In fact, they report the possibility of an intermediate reaction product for the Cu/BeO bonding, which seems to be of spinel structure. Yoshino151 contributes to the sensitivity of analytical methods used and states that the possibility of a thin oxide film cannot be excluded.

5 Diemer investigates the Cu/Al2O3 system extensively. As well as the aforementioned wetting investigations, he also investigates the fracture toughness of the interface, depending on the phases occurring. Sandwich specimens are prepared with specific interfaces, and toughness measured by compact tension. He shows the fracture toughness of the interface as strongly affected by the oxygen content, with a maximum of around 5–9 at% oxygen at

52

the interface. Pure Cu/Al2O3 and Cu2O/Al2O3 (NO = 33 at%) interfaces, respectively, fail both at a lower applied load. The formation of aluminate is excluded for NO = 5 at% but stated to be possible for NO = 9 at%. Nevertheless no aluminate can be found by optical investigation at the interface, hence he states aluminate is not responsible for the increase in fracture toughness in this regime. The appearance of interfaces up to 9 at% oxygen shows Cu2O inclusions in the metallic copper matrix: during compact tension experiments he observes microcracks arising within the Cu2O particles in front of the actual crack tip. Further investigations show plastic deformation at the copper side of the fractured surface with cracks in the remaining Cu2O particles and delaminations at the Cu/Cu2O interface. At the alumina side, numerous Cu2O particles are present but no copper residue can be found; concluding that the 162 bonding of Cu and Cu2O to alumina shows, in accord with Sun and Driscoll , that the bonding of the oxide to the ceramic is stronger than the metal to the ceramic. This also contradicts the observation of Reimanis et al.159, without considering the conclusions regarding differences in fracture mode and bonding experiments. Aluminate strengthens the interface bonding in copper/alumina joints, as has been reported often in literature. Diemer5 investigates the fracture toughness of interfaces and shows a strong increase of fracture toughness compared with Cu/Al2O3 interfaces if an aluminate layer, created by heat treatment of Cu/Al2O3 interfaces, is present. For interfaces existing of pure

Cu2O (NO = 33 at%) it seems obvious that the crack path is within the copper oxide: actually, the crack path never reaches the interface at either side and

CuAlO2 has no influence in this case on the fracture toughness.

Several authors investigate the copper/alumina interface in an atomistic manner19, 20, 144, 184, 185. Theoretically it is proposed that the atom species neighbouring each other at the interface influence the bonding strength dramatically. Depending on the substrate cleaning method, the ߙ-Al2O3 surface is either Al- or O-terminated as reported by Scheu185: Ar+ ion sputtering leads the alumina surface to be Al-terminated, whereas a chemical treatment results in an O-terminated surface. The ߙ-Al2O3 interface is prepared to be either

Oxygen- (Al2O3(O)) or Aluminium- (Al2O3(Al)) terminated, in the investigations of

53

Rühle19: the copper coating at the corresponding sapphire surface realised by molecular beam epitaxy is several 100 nm thick. He reports a bonding between the metallic copper and aluminium at the Al2O3(Al)/Cu interface, whereas at the oxygen-terminated interface, bonding occurs between oxygen and copper with higher adhesion strength. This is in accordance with the investigations of Zhang et al.20, who describe the connection between ab initio calculations and interface measurements. Although their preparation of interfaces using diffusion bonding is different, they observe high adhesion, expressed by the work of separation, for the proposed oxygen-terminated alumina/copper interface and alumina/ interface, respectively. Investigations of the proposed aluminium- terminated alumina/nickel interface reveal the work of adhesion to be several 184 times smaller than for either bulk Ni or Al2O3. In contrast, Hashibon et al. propose in their ab initio study that, despite the different atomistic structures of the aluminium-terminated and the oxygen-terminated alumina/copper interfaces, their stabilities, in terms of work of separation, are similar. Rühle19 observes in his high resolution TEM investigations of these interfaces a coherent Al2O3(Al)/Cu interface, whereas the oxygen-terminated alumina/copper interface exhibits some dislocations, indicating a semi-coherent

Al2O3(O)/Cu interface. In addition, he mentions that the interface adhesion can be increased by the formation of an oxide layer, as previously mentioned and reported by several other authors5, 141, 149, 159, 161.

2.7 Mechanical Properties of Interpenetrating Phase Networks Interpenetrating phase composites are characterised by at least two three- dimensionally interconnected phases, as reported earlier; each phase retains its own specific properties in the macroscopic composite material. Whereas one phase may provide high strength, the other may improve other physical properties, like electrical and thermal conductivity186. Such an attractive combination of properties is not possible, if one of the phases is isolated, i.e. if the composite is not of interpenetrating phase structure187. Considering monolithic ceramics, the metal phase contributes to higher conductivity and ductility, whereas for pure metals or alloys the ceramic reinforcement improves the stiffness and wear resistance of the resulting composite material. The main

54

toughening mechanisms in such composite materials rely on the crack bridging due to the ductile metal ligaments105, 188, 189. Here, the stress intensity factor at the crack tip is reduced due to the bridging in the crack wake by plastically deforming metal ligaments leading to closure stresses189. This effect can be more effective with residual compressive stresses in the ceramic phase; this is apparent due to the fabrication process resulting from the thermal expansion coefficient mismatch of the metal and the ceramic phases106, 190, 191. Due to the special microstructure in interpenetrating phase composites, the potential of strengthening mechanisms is enhanced, as the crack has to pass both phases133. Stresses in geometrically constrained metal ligaments can exceed the yield strength of the resulting bulk material several times190, 192; furthermore, small ligaments with good interface bonding result in a high degree of geometric constraint. With increasing the metal ligaments for constant volume fraction the toughening increase is more effective192. Residual stresses in the metal phase can rise up to 620 MPa for Cu/Al2O3 composites as measured by Agrawal et al.190 which is higher than the yield strength of 330 MPa of the copper alloy used for infiltration; they state this to be possible as the stress tensor is predominantly hydrostatic.

Alumina based composites, manufactured with aluminium and copper by gas pressure infiltration exhibit bending strength of 710 MPa and 551 MPa with fracture toughness of 5.4 MPa√m and 4.3 MPa√m for ceramic volume contents of 75% and ligament diameters of about 0.15μm104, 105. Maximum strength for the Al/Al2O3 composite up to about 800 MPa is found at a metal fraction of 40% with a median pore diameter of 0.15 μm; both studies report increasing the pore diameter results in decreasing strength but increasing toughness. Menon and Fahrenholtz193 investigate composites fabricated by pressure-less infiltration of alumina preform with Cu-O alloys; the bending strength is measured to be within 400-500 MPa with fracture toughness of 9-11 MPa√m. Whereas 194 Travitzki reports bending strength of 355 MPa for his Cu/Al2O3 composites manufactured by gas pressure infiltration and 275 MPa for the material fabricated by pressure-less infiltration with Cu-O alloy with fracture toughness of 6.7 MPa√m and 6.5 MPa√m, respectively. The investigations of Agrawal and 195 Sun show the fracture in the Cu/Al2O3 composite material occurring

55

preferably at the metal/ceramic interface and within the metal phase, whereas in the Al/Al2O3 system the crack grows at the metal/ceramic interface and in the ceramic phase. He states this to be evident due to the higher tensile stresses in the metal phase for Cu/Al2O3 compared to Al/Al2O3 composites resulting from fabrication at higher processing temperatures.

2.8 Summary

The Cu/Al2O3 system is of a non-wetting and non-reactive nature, as stated by many authors. Because of its direct relevance to industrial applications many research groups focus on investigating the bonding behaviour of copper to alumina. Directly-bonded copper (DBC) stacks, for example, are commonly used in heat sink applications. The reliability of bonding emphasises the need for a thorough understanding of the system and encourages research groups.

The wetting of the naturally non-wetting system Cu/Al2O3 can be improved by the addition of dopants to the melt. Dopants such as Ti, Zr, Cr, V, are all seen to improve wetting2. Beneficial effects include the release of free energy for reactive wetting, modification of the solid side in the interfacial contact area and the adsorption at the interface due to the formation of a layer with high concentrations of dopant on the liquid side. Overall, the most important copper alloying element for changing the wetting condition is seen to be oxygen. The behaviour of oxygen addition is investigated in more detail by several authors because no other element is added in the ternary Cu-Al-O system. In changing the oxygen content of the melt, either by alloying the copper with copper oxide or by controlling the oxygen partial pressure in the atmosphere, the wetting behaviour of copper on alumina can be influenced. Furthermore, it is found that interfaces where oxygen is present are much stronger.

Composite fabrication is governed more by dynamic wetting, and can be described as the infiltration of porous media. Systems of wetting behaviour measured in static wetting experiments do not necessarily show spontaneous infiltration; for spontaneous – i.e. pressure-less – infiltration wetting angles much lower than 90° are necessary. But information from static wetting experiments can be useful. The infiltration of liquid in porous media has been 56

thoroughly described in soil science by the drainage of water in sand. Several models describing the infiltration process are provided in literature, the most common being the empirically-established model of Darcy, which represents a simplification of the general equation of viscous fluid flow governed by the Navier-Stokes equation. In the non-wetting copper/alumina system, infiltration can be realised only when the melt is force-driven by external pressure. Reactive wetting to improve wetting behaviour is applicable but, depending on the improvement, some reaction products may hinder further infiltration by blocking finer porosity. Still, improvement of wetting is beneficial to composite fabrication as it results in easier and more reliable infiltration processing and better adhesion of phases.

The squeeze casting process of sintered preforms is normally applied for infiltration with aluminium alloys. Infiltration with higher melting-point metals – copper, e.g. – has been investigated mostly through gas pressure infiltration of packed powder beds. In gas pressure infiltration, the atmosphere can be easily controlled and the reinforcement evacuated before infiltration to reduce porosity in the resulting composite. The infiltration progress is relatively slow, and as both the reinforcement and matrix phases are heated up in the same chamber, the infiltration is isothermic. Most modelling of infiltration is based upon this. Squeeze cast infiltration is very fast and therefore favourable for industrial applications. In squeeze casting, undesirable reactions can be suppressed through kinetic circumstances. In contrast to gas pressure infiltration, preform, melt and infiltration tool are heated up separately, and infiltration can be seen as a duel between pore filling and solidification of the melt. With the preform having a critical preheating temperature and the melt being superheated, tool material is preheated to elevated temperatures as well, and has to be able to withstand such temperatures at given pressure.

The system Cu/Al/O has been investigated widely. In model experiments like sandwich specimens it has been shown that aluminate is the most favourable phase at the interphase to enhance bonding. But even oxygen, if present at the Cu/Al2O3 interface, enhances the bonding strength. Additionally, the wetting improves with the addition of oxygen to the copper melt, which leads

57

to better and more reliable infiltration processing. Hence, oxygen plays a major role in the system and appears important for composite fabrication.

58

3 Prelimary work, Hypotheses and Approach The copper/alumina system has been investigated widely; the establishment of strong bonds between copper and alumina on flat surfaces is well known. The wetting behavior of copper on alumina is enhanced by the addition of oxygen to the melt: the oxygen addition also promotes stronger bonding at the copper/alumina interface. Manufacturing of Cu-based composites is described in the literature with regard only to gas pressure or spontaneous infiltration and/or powder beds, rather than to sintered particle preforms. Composites with a co-continuous network structure are of special interest, as the ceramic structure provides a high degree of stiffness, and the metal phase holds some plasticity and, especially in the case of copper, high thermal and electrical conductivity. During a preliminary study196 it was found that by adding copper oxide in the preform preparation process for the fabrication of co-continuous copper/alumina composites the preforms show better infiltration ability and improved composite bending strength compared to those manufactured from pure alumina preforms.

The aim of this thesis is to understand the underlying phenomena of the improvement of infiltration ability and consequent bending strength due to the addition of copper oxide and further to develop a composite material with a co- continuous network structure with excellent properties, with copper being the reinforcing material and alumina the reinforcement. The nature of this system is non-wetting. The manufacturing process of the composite material should be fast and reliable, for possible transfer to industrial applications; so fabrication will be carried out through the industrially applicable squeeze casting process. To date infiltration of copper/alumina composites has been by gas pressure infiltration; no process is known for reliable infiltration of alumina preforms by squeeze casting, suitable for industrial application. Oxygen improves the wetting character: to provide it at the interface without having a Cu/O alloy, which is difficult to handle because the oxygen partial pressure has to be controlled, it will be provided in the preform structure. By adding copper oxide in the preform preparation process, oxygen can be brought easily into the reinforcement; and by the possible reaction of Cu2O with Al2O3, an interface

59

enhancing phase – aluminate – can be formed. Aluminate at the interface results in strong copper-to-alumina joints as described many times in the literature. No description has been found, which shows the behavior of copper oxide or aluminate phase during infiltration of a preform with copper melt.

The hypotheses, which form the basis of this work to fabricate copper/alumina co-continuous composites and for better comprehension of the underlying phenomena are as follows:

1. Infiltration of alumina preforms with copper by squeeze casting becomes reliable, when providing better wetting properties and high thermal energy in the infiltration system to prevent premature solidification. 2. Oxygen improves the wetting properties of the system copper/alumina. Instead of using a Cu-O alloy, which is more difficult to handle, it is possible to provide the oxygen in the preform structure. If distributed homogeneously on the internal surfaces of the preform structure, it further ensures oxygen being present where it is most useful. 3. Copper oxide added in the powder preparation process reacts with the alumina powder to form aluminate. This phase will form on the inner surfaces of the preform, providing an internal interface of the composite structure. Aluminate at the interface leads to higher bond strength which results in improved mechanical properties of the entire composite material and due to the better adhesion of phases enhances the thermal conductivity.

Infiltration of alumina preforms with and without copper oxide added will be performed in an Ni- alloy infiltration tool, which allows tool temperatures up to 900°C. The melt will be superheated and the preform preheated above the minimum critical preheating temperature. Insulation material will protect the melt, once poured onto the preform, from rapid cooling before a pressure is applied. Hence, the thermal energy of the entire infiltrating system will be increased. It is well known that oxygen improves the wetting properties of copper on alumina. This work will deal with the approach of providing the oxygen within the ceramic structure, the preform, by adding copper oxide

(Cu2O) during the powder preparation process. Appropriate preparation will 60

distribute the copper oxide homogeneously in the preform structure. A reaction of copper oxide with alumina will form an aluminate phase, which is known to enhance interface bonding. Heat treatment conditions to ensure the appropriate phases being present will be investigated and analysed; the use of aluminate instead of copper oxide will be considered as well. Investigations of preform structure, the interface of the composite material and fractured surfaces will show up the distribution of the copper oxide and/or aluminate phases. With the copper oxide thought to be located at the inner surfaces, the formation of aluminate leads to an interfacial layer for better bonding strength of the phases in the composite material; this will result in an overall improvement of mechanical properties. With the adhesion of the phases being improved, the thermal conductivity will be improved due to fewer voids in the microstructure and better heat transition between phases. Microscopic and analytical investigations as well as mechanical characterisation will focus addressing the hypotheses given.

Concluding the approaches: aspects of the Cu/Al/O system will be used in combination with wetting aspects and infiltration phenomena to:

x fabricate co-continuous copper/alumina composites by infiltration of a sintered particle preform; x provide infiltrate-able preforms; x improve and establish the infiltration progress; x apply the relatively fast squeeze cast infiltration process; x reliably fabricate composite materials, which may be used to transfer to industrial application.

Bonding of Cu/Al2O3 joints has been described in literature for temperatures below the melting point of pure copper and mainly in a one-step process, where copper and alumina is in contact with or without added interfacial copper oxide.

Aluminate will form at the interface if Cu2O is present. This layer enhances the nature of the bonding, but has been investigated only on Cu/Al2O3 joints where copper was solid. Here, the aluminate phase is formed before liquid copper is added to the composite and, hence, is present at the interface (two-step

61

process above copper melting temperature). This behaviour will be investigated and discussed relative to the literature. As well, a characterisation of the microstructure of the composites will be provided, with discussion of the related mechanisms.

62

4 Experimental Procedure In this chapter the procedure of composite fabrication is experimentally described; the raw materials used, the preform preparation and the infiltration of the preforms with molten metal is described. Furthermore, the characterisation procedures and techniques used to describe the preforms and composite materials are given with the procedure for materialographic imaging, and hence, microstructural investigations. Nano-indentation testing for single phase characterisation and measurement of thermal conductivity are also specified.

4.1 Characterisation of Raw Materials The materials investigated in this study are described below in detail. The infiltration material or the matrix material of the MMC is copper, distributed by Norddeutsche Affinerie (Hamburg, Germany) and is specified in Table 3; while the reinforcing ceramic structure was alumina. The copper used was of OF- quality, which means high quality, non-deoxidized, oxygen-free copper suitable for electronic devices because of its high purity and conductivity.

Table 3: Typical analysis spectra of the copper used for the fabrication of metal matrix composites in this study Pb Bi As Sb Sn Zn Mn Cr Co Cd Fe Ni Ag S Se Te O weigth-ppm < 1 < 0,5 1 1 < 0,5 < 1 < 0,5 < 1 < 1 < 1 ≤ 6 < 2 10 4 < 0,5 < 0,5 ≤ 5

Alumina powders from different manufacturers were used for preform preparation: all were of technical purity. Compaction ability due to different grain size distribution, resulting in different ceramic volume fractions of the preform, is powder-dependent. The powders used are summarised and specified in Table 4. Some selected properties of dense alumina and pure copper of the type and quality used in this study are listed in Table 5. The alumina powders CL2500 and CT3000SG were from Almatis (distributed by Bassermann GmbH, Mannheim, Germany); while the alumina powders NO315 and NO316-30 were from Nabaltec (Schwandorf, Germany).

63

Table 4: Specifications of alumina powders used for preform preparation197, 198.

Alumina powder

CL 2500 NO 615-30 NO 315 CT 3000 SG

Al2O3 % 99.8 99.6 99.6 99.8

Na2O % 0.06 0.3 0.3 0.08

Fe2O3 % 0.02 0.03 0.03 0.02

SiO2 % 0.01 0.05 0.03 0.03

Specific surface area / m2/g 0.95 6 1.5 7.5 BET

Particle size μm 80%>63 2 3 n/a

Primary crystal size / D50 μm 1.9 1.5 2 0.8 cilas

Green density g/cm3 2.22 2.7 n/a 2.25

The alumina powder CT 3000 SG was used explicitly for the fabrication of the freeze cast preforms described in section 4.2.2. In addition, copper oxide (from Sigma Aldrich Chemie GmbH, Taufkirchen, Germany) has been used in this study, the particle size given was 97% < 5 μm.

Table 5: Selected properties of raw materials used for composite preparation199, 200.

Alumina Copper

Raw density g/cm3 3.96 8.94

Strength MPa 470 (bending) 230-390 (tensile)

Young’s modulus GPa 380 127

½ Fracture resistance KIC = 4,3 MPa*m A10 = 50 - 3%

CTE α (20-300°C) 10-6 K-1 7. 17.7

Thermal conductivity W/mK 30 >394

Hardness 1500 HV0.5 50-11 HB

The mixed oxide of alumina and copper oxide, copper aluminate with the sum formula CuAlO2, was not commercially obtainable. Therefore, a small amount of

64

copper aluminate was prepared for use according to the reaction already shown in equation 30:

ଶܱ݈ܣݑܥʹ ଶܱଷ ՜݈ܣݑଶܱ൅ܥ

The preparation process is described detailed in section 5.1.1. The data in the tables listed above form the basis for the calculation of ceramic volume content and other specific properties of the preforms. The particle shape and appearance of the different raw powders are shown in Figure 18.

(a) (b)

1 μm 1 μm

(c) (d)

1 μm 1 μm

Figure 18: Shape and appearance of commercially-available raw oxide ceramic materials used in this study: (a) shows the alumina powder CL 2500 of Almatis; (b) the bimodal powder NO 316-30 and (c) the alumina powder NO 315, both from Nabaltec; (d) shows the copper oxide powder Cu2O from Sigma Aldrich (all pictures are taken before powder preparation).

In the pictures above, taken before powder preparation, the alumina powders CL 2500 and NO 615-30 appear agglomerated; the size of the agglomerates for the CL 2500 is 120–160 μm, the powder of NO 615-30 appears finer, with agglomerate sizes of 40–60 μm. The NO 315 powder from Nabaltec, with a grain size of 3 μm, is found largely to be without agglomerates. The idiomorphic form of Cu2O from Sigma Aldrich can be seen in Figure 18d. Typical of the

65

appearance of CL 2500 are some platelets within the normal, roundish-shaped particles of alumina. In Figure 18b the bimodality of NO 615-30 with a submicrometer scaled powder and a powder with grain size of around 2 μm is observable.

Pore-forming agents (PFA) in ceramic green bodies allow a wide variety in design of preform characteristics, such as pore structure and ceramic volume content. With partial sintering of ceramic compacts alone the pore fraction cannot be varied in a wide range. In this study the organic agent cellulose, specified in Table 6, was used as the pore-forming agent to form relatively large pores.

Table 6: Specified properties of the organic pore-forming agent (PFA) used for preform fabrication201.

Cellulose

Main chemical constituents C6H12O6

Supplier JRS, Holzmühle (Germany)

Type Arbocell P290

Average particle diameter d50 μm 150

Theoretical density kg/m³ 1500

The particle size distribution of the cellulose was measured using a laser diffraction particle analyser of the type Sympatec Helos. In Figure 19 a typical cellulose particle is shown. The behaviour of the pore-forming agent, which should be cracked and oxidized in the sintering process, was investigated using thermo gravimetric analysis (STA 502 - simultaneous thermal analysis - from Baehr, Germany). The decomposition behaviour between 20 and 800°C in inert and oxidising atmosphere was evaluated.

66

15 μm

Figure 19: SEM micrograph of a compactly-shaped cellulose particle used as pore- forming agent for preform fabrication.

As well as the ceramic compounds, several additives – the pore-forming additive, e.g. – were used for preform fabrication. These also included binder or dispersing additives, which are obligatory for ceramic preparation; those used for preform preparation and the concentrations used are summarised in Table 7.

Table 7: Additives, besides cellulose, used for ceramic preparation.

Binder Deflocculant

Type Mowiol 18-88 Dolapix CE64

Zschimmer&Schwarz, Supplier Clariant Lahnstein (Germany)

Active substance Polyvinyl [CH CHOH] Polycarboxylic (Structure) 2 n

Density [g/cm³] 0.4 – 0.6 1,10

Fraction of addition [%] 1 0 – 0.5

The binder, polyvinyl alcohol, was manufactured from polyvinyl acetate by alcoholysis202. The dispersing agent, Dolapix CE64, was an organic, alkali-free

67

deflocculant that does not foam. As it is liquid and thus completely dissociated, the deflocculating effect commences directly after addition to the slip203.

4.2 Preform Processing The fabrication of the composite is subdivided into two separate steps: first, a ceramic preform has to be manufactured; then a liquid metal can be forced into the pores by applying an external pressure. For investigation of interfacial effects, a coating process was applied on specific preforms before infiltration.

4.2.1 Fabrication of Sintered Porous Particle Compacts The preparation process for preforms – ceramic bodies with open cell porosity – is divided into three steps: the liquid powder preparation, followed by a drying and shaping procedure and, finally, the preform sintering process. For model investigation of the interfacial behaviour, preforms of a specific, laminar structure (see section 4.2.2) were used.

Powder preparation was performed in small and big batches. For testing of different powder compositions the preparation process was performed in a laboratory, using small batches. The powders were milled with water using a Fritsch Pulverisette ball mill for 5 min to give full dispersion and for de- agglomeration. In case the slurry contained fine particles (d50 < 1 μm) the dispersant Dolapix CE64 was added. The balls made of -stabilised zirconia were 5 mm in diameter and the container made of polyethylene. Subsequently 1 wt% of the binder Mowiol 18-88 was added to enable pressing of the final powders and handling of the green bodies. Finally, the pore-forming agent, cellulose, was added where appropriate and the slurry stirred for at least another 5 min to give enough time for dispersion. The final slurry was then poured into pre-cooled aluminium alloy plates, placed into a freezer set to -24°C and stored for at least 24 h. To dry the slurry, the plates were placed in a Christ freeze-drying unit of the type Alpha 1-4 and drying was performed at 256 Pa for a minimum of 12 h. The dried powders were then passed through a 250 μm mesh sieve to produce a granulated powder. For activation of the binder 3 wt% of water was dispersed in the powder, which was conditioned for at least 24 h to ensure uniformly-distributed binder activation.

68

As the second step, favourable powders were prepared in big batches, each of at least 100 kg. The preparation process was the same except the drying process, which was done in an industrial manner by a spray drying unit (in which process slips can be dried continuously): it was a 1C/FM/125/UP from ICF INDUSTRIE CIBEC S.p.a. (Maranello, Italy) with a vaporisation capability of 180 l/h, and the drying was done at Alpha Ceramics in Aachen, Germany.

Shape forming of the prepared powders was carried out with uniaxial pressing at ambient temperatures. The shaping die, especially the lower punch, was fixed during the pressing process. Samples for pre-testing various sintering temperatures and programs were pressed using a cylindrical cavity with adjustable dumping height and a diameter of approx. 30 mm on a handpress from Nelke. Die and punches were made of hardened tool steel. Samples prepared for mechanical testing and material analysis were pressed in a rectangular die with inner dimensions of 65 x 46 mm² and adjustable dumping height. The height of the resulting green bodies was set to 8–9 mm with compaction ratios from 2–4.5, depending on powder composition. Pressing was performed on a hydraulic press (Fontijne TP400 Spezial, The Netherlands) with a forming pressure of 100 MPa. The density of the green body was determined geometrically before sintering. A schematic of pressing process is shown in Figure 20.

Figure 20: Schematic of the uniaxial pressing process for green body preparation of the prepared powders.

Consolidation of the preforms was done in different furnaces. Copper oxide containing preforms were sintered at temperatures of 1200°C: a muffle furnace of the type Nabertherm N21 was available for this low temperature. For

69

preforms without copper oxide the sintering temperature was higher, up to 1610°C. Sintering with the higher temperatures was performed using a tube furnace (Gero HT-RH 70-300/16, alumina tube with an inner diameter of 60 mm) and a high temperature batch furnace (LinnHT1800). The maximum temperature of the tube furnace was 1600°C, whereas the batch furnace temperatures could reach up to 1750°C. For comparison, all types of preforms, where possible, were sintered at least twice in the tube furnace to ensure no furnace-dependent behaviour. All three types of furnaces were adapted with an electronic control to program sintering cycles with different heating rates and consolidation times. Depending on the preform type, the sintering temperature differed. A typical sintering cycle for copper oxide containing preforms is demonstrated in Figure 21:

Figure 21: Typical sintering cycle for copper oxide containing preforms.

Whereas all copper oxide containing preform types were sintered at 1200°C for 2 h, copper oxide-free specimens were sintered at different temperatures to account for the desired ceramic volume content. Table 8 gives an overview.

70

Table 8: Sintering temperatures an corresponding ceramic volume contents of copper oxide-free preform types

Preform type [ ] AO A10 A20 N3/5 N3/10 N6/5 N6/10 N6/20

Sintering temperature [°C] 1200 1590 1500 1550 1600 1400 1450 1600 Ceramic content [vol%] 53 53 35 59 50 65 60 60 Sintering temperature [°C] 1500 1620 1600 1450 Ceramic content [vol%] 60 60 45 50

For all type of preforms the sintering cycle starts with a slow heating rate to smoothly pyrolyse the organics in the green body; 20 K/h up to 475°C were used at an air flow of 20 l/h in the tube furnace. Preform types containing copper oxide were able to be sintered in a muffle furnace with a high ratio of furnace volume to preform volume. Apart from that, all preforms were sintered for 2 hours at their respective temperatures. After sintering the specimens were measured and weighed. The ceramic volume content was calculated using the theoretical as described in section 4.4.

As many different powders were used, a unique notation for each variation, respective to preform type, was designed. The nomenclature to distinguish the types prepared is illustrated in Figure 22: AC20/3S_xxx

Specimen number Type of drying – freeze drying S – Spray drying Content of pore forming agent in wt% C – copper oxide Content of A – aluminate addition in wt‰ A – Almatis Type of powder N3 – Nabaltec NO315 N6 – Nabaltec NO615-30

Figure 22: Nomenclature of the preforms prepared.

71

4.2.2 Processing of Preforms for Model Investigation For investigation of the interface adhesion in regards to interface appearance, highly permeable preforms with nearly lamellar pore structure were used: these were made by colleagues at the Karlsruhe Institute of Technology (Karlsruhe, Germany) in a freeze casting process. Water-based suspensions of the alumina powder CT 3000 SG were prepared with distilled water, dispersant Dolapix CE64 and 10 wt% binder Optapix PAF60 (both Zschimmer & Schwarz, Lahn- stein, Germany). The solid content was 22% by mass. Before freezing, the dispersion was pre-cooled to 5°C and degassed. Freezing and freeze drying took part in a Sublimator 400K freeze drier (Zirbus technology, Bad Grund, Germany). Rectangular PTFE-dies were used for shaping the preforms. The dispersion, filled into the dies, was stored on a pre-cooled (T = -10°C) copper plate in order to achieve directional freezing of the water.

After freezing a freeze drying process develops the preform body. A mechanical grinding process on both sides to lay the structure bare completes the preparation process of the preform after sintering. In Figure 8 a phase diagram is shown with a schematic of the freezing process for preform shaping, freeze drying and preform completion. The process is described in detail by Mattern102 and Waschkies101. The nomenclature of this type of preforms is different and given with FC22m10.

To investigate the interfacial behaviour the preforms were treated at the interface before infiltration. As proposed by Mortensen and Jin35, coating of the internal surfaces of the ceramic substrate is an effective treatment to enhance the wetting by changing the ceramic phase; so the interface modification was performed using an electroless coating process of the ceramic preform. Coating of non-conducting surfaces can be subdivided into two steps – activation of the surface and coating by a metal. In the activation step an activator is brought onto the surface. Water-based solutions, which were used here, commonly work with as an activator for metallization. The activated surface in contact with the coating bath results in a metal deposition, which relies on redox reactions of the constituents of the coating bath. The palladium acts only as

72

activator and electron transmitter. Figure 23 shows a schematic of the coating process including the redox reactions:

Redox reaction:

z+ (z+x)+ - Activation R R + xe cluster Me* + xe- Me Me* + Rz+ Me + R(z+x)+

non-conducting surface (activated)

Figure 23: Schematic of the electroless coating process on an activated non-conducting surface with the corresponding redox reaction, where R is the reducing agent and Me* a complex bound metal ion.

Coatings are only adhesive if the surface is clean before coating occurs. The samples were cleaned in an ultrasonic bath and, additionally, heat treated at 1000°C for 2 h to ensure an organic-free surface. For electroless plating of alumina substrates a chemical copper bath Enplate MS Cu-9070® (Enthone, Germany) was used: it contains the soluble metal salt, stabilizer and reducing agents. The surface of the substrates was first activated by Udique 879 W® (Enthone, Germany) containing Pd/Sn clusters in a hydrochloric acid solution. Coating and activation were performed using a coating unit with beaker and specimen-holder made of Teflon. Figure 24 shows a schematic of the coating unit. The preforms were mounted in the specimen-holder – consisting of two disks with rectangular cutouts – and the contact area was sealed with butyl rubber. For activation the mounted preform was dipped into the preheated activation solution and repeatedly moved up and down for liquid flow through the pore channels. The same procedure was used for rinsing thoroughly with distilled water and the actual coating steps of the pre-activated surface.

73

Figure 24: Schematic of the coating unit used for activation, rinsing and coating of the preforms for interface modification.

4.3 Composite Fabrication Spontaneous infiltration of the manufactured preforms did not occur; therefore, pressure had to be applied in order to infiltrate the samples. For this, two kinds of infiltration processes were used: direct squeeze casting and gas pressure infiltration.

4.3.1 Direct Squeeze Cast Infiltration In direct squeeze casting pressure is applied by means of a plunger in order to infiltrate the preform. The infiltration process was done using a set up of three parts – a matrix, a lower and an upper punch, as shown in Figure 25 and Figure 26(a). All parts were made of hot working steel (mat. no. 1.2343) for infiltration at ambient temperatures and of Haynes® 230 for infiltration at high temperatures. The matrix and the lower punch of the steel tool formed a rectangular cavity of 65 x 46 x 35 mm3. Prior to infiltration the cavity and the upper punch were coated with a thin layer of carbon (Graphit 33, Kontaktchemie, Germany) then preheated to 420°C in a laboratory platen press of the type TP400 Spezial, from Fontijne LTD, The Netherlands. For infiltration at high temperatures the cavity made of Haynes® 230, with inner dimensions of 76 x 57 x 48 mm3, was coated with a thin layer of nitride (HeBoCoat 401E, Henze, Germany) then preheated to 800°C – in two different external

74

furnaces for handling reasons. The cavity was preheated in a furnace with a wide opening (Cu4 of Kilns & Furnaces, England), placed on a steel sheet for transportation, whereas the upper punch was preheated in a smaller furnace (MR170, Heraeus, Germany) without any handling. The cavities were lined with 3 mm of highly porous insulation material (Superwool 607 HT Paper, Contherm, Germany). For infiltration at ambient temperatures two layers of insulation were used for the lower punch. The preforms were pre-heated up to 1170°C in a muffle furnace (N41, Nabertherm, Germany), at a heating rate below 200 K/h to prevent thermally-induced cracking. To achieve infiltration the following steps occurred sequentially: the cavity and upper punch were transferred from the external furnaces to the 450°C hot plate of the laboratory press. The preform was set into the cavity. The metal in a graphite , molten at 1250°C in an argon atmosphere in a LKs 15/8/25 furnace from Degussa, Germany, was poured directly onto the preform. One layer of insulation material and the upper punch were set onto the die and the cavity was closed as the setup was pressed together between the platens – the press closes at an average velocity of 0.017 m/s to a load-controlled pressure of 100 MPa. The time between setting the cavity onto the plates and the infiltration process was measured to be in the range of 15 to 20 sec. Figure 25 shows the direct squeeze cast infiltration schematically. After solidification under pressure the samples were ejected.

Figure 25: Schematic of the steps for infiltration using the direct squeeze casting process.

A data acquisition system was used to record the pressure increase over time and the displacement of the punch during the squeeze cast infiltration with a sample rate of 500 Hz. The compliance of the press was taken into account.

75

True pressure was calculated as the ratio of the force to the cross-sectional area of the die cavity, and the infiltration ratio as that of volume reduction in the die due to plunger displacement to the nominal preform porosity volume.

4.3.2 Gas Pressure Infiltration Infiltration via gas pressure consists of forcing the liquid metal by means of pressurized gas. The tool preparation was similar to the squeeze casting process. The upper punch was replaced by a punch geometrically similar to the lower – a schematic view is shown in Figure 26(b) – then fitted with a copper sealing ring to allow pressurization.

(a) (b)

Ar Infiltration tool Copper Copper melt sealing Preform

Figure 26: Schematic view of infiltration tools for -a- squeeze casting infiltration and -b- gas pressure infiltration set-up.

The preheated preform is placed into the cavity and the metal melt poured on top of it. In contrast with the squeeze casting process, no insulation between the metal melt and upper punch – preheated to 420°C – was used, as there is no contact with the molten metal. The upper punch was set onto the cavity, preheated to 800°C, and the die was closed for the set-up to be pressed between the platens; then the gas pressure was applied within 1 sec to the chosen preset value: the gas in the preform replaced by the melt volume could then flow through the air gap between the lower punch and the matrix. After solidification the pressure was reduced and the sample ejected.

To ensure exactly the same infiltration parameters on different preform types, up to 4 preform parts were joined together with a ceramic adhesive (Keratin K, Rath AG, Germany) and infiltrated in one shot.

76

Vacuum / argon pressure Graphite heater Graphite crucible

Preform

Copper

Figure 27: Schematic view of infiltration set-up for isothermal gas pressure infiltration.

Additional to the infiltrations performed in the squeeze cast set-up, specimens were infiltrated isothermally by gas pressure infiltration. The infiltration was performed with a commercial gas pressure sintering furnace. The preform, with its density of about 4 g/cm³, would float on the copper melt, leading to a gas bypass preventing copper infiltration when applying gas pressure. So the infiltration set-up, as shown in Figure 27, was described thus: a crucible was fixed upside-down over the preform, lying on top of the copper, and this was placed into another crucible. During heating the copper melted and the preform floated on the liquid, but located within the upside down crucible. By applying gas pressure, the preform rose with the copper melt in the crucible until it reached the top. From this, the copper liquid penetrated the preform porosity. During heating and melting the chamber was evacuated and infiltration took part at 1200°C with an applied argon gas pressure of 10 MPa.

4.4 Characterisation Methods Different methods for the characterisation of preforms were carried out. In order to validate sintering processes and preform variations, the porosity and the ceramic volume content were used as a rough/fast method. The total porosity

Ȱ௧௢௧ and the volume fractions of the ceramic were calculated using their theoretical densities. Other preferred preform types were measured using the Archimedes principle to determine the fraction of open-cell porosity, in accordance with DIN 51918204. To do this, the sample was first dried in an oven at a temperature of 100°C for 1 h and then weighed (݉଴). After evacuating down to 50 mbar and immersing in distilled water, atmospheric pressure was

77

applied for 30 min to ensure that the water entered the open pores of the preform fully. It was re-weighed immersed in water (݉ଶ). Finally, the sample was taken out, excessive water drops were removed from its surface and it was again weighed (݉ଵ). The raw density ߩ௥ and the fraction of open porosity Ȱ௢௣ in the preform were calculated from the following equations:

m0 Ur * Ul Equation 32  mm 21

 mm 01 ) op Equation 33  mm 21 where ߩ௟ is the density of the fluid at a given test temperature. The values were 173 taken from tables such as these provided by Lide . Closed cell porosity Ȱ௖ was calculated as the difference between total porosity Ȱ௧௢௧ and open cell porosity Ȱ௢௣.

4.4.1 Permeability of Preforms Permeability was measured along the uniaxial pressing direction of the preforms. Although expecting anisotropic permeability of preforms containing PFA, according to the work of uniaxial formed fibre preforms presented by Mortensen et al.77, permeability was not measured perpendicular to the pressing direction, as infiltration is in the parallel direction. For measurements along the pressing direction, disc-shaped samples with a diameter of 30 mm and a height of about 7 mm were used. The samples were bonded into aluminium rings with a conical inner ring (Figure 28b) using a high viscosity adhesive to prevent infiltration. Prior to testing, the samples were evacuated and immersed in distilled water for 12 h to completely fill the open porosity. For measuring, the ring was mounted at the end of a vertical tube filled with the fluid, and controlled gas pressure was applied. A schematic of the assembly is seen in Figure 28c. The mass of water was measured continuously using a Mettler-Toledo PG-5002 SA balance with a resolution of 10 mg. In measuring the fluid temperature the flow rate was calculated using values of water density. The data of density and viscosity were taken from tables173.

78

(a) Green body (c) pressing direction

water seal (b) aluminium ring

glue joint

40 mm

Figure 28: Schematic of (a), samples used for permeability measurements dependent on pressing direction; (b), geometry of ring used as mount for the testing of disk-shaped samples; and (c), apparatus used for preform permeability measurements.

The calculation of the permeability for lamellar flow follows Darcy’s law. The permeability ܭ is defined by

x K ** hV K *'PA Equation 34 with the Volume flow ܸሶ , the viscosity ߟ, the height of preform ݄, the measured cross section ܣ and the pressure drop ȟܲ.

4.4.2 Surface Preparation, Microscopy and Image Analysis All microscopic investigation regarding composite microstructural characterisation was done on polished samples. For the preparation, samples were mounted using a cold (EpoFix of Struers, Germany) or a hot (DuroFast of Struers, Germany) mounting resin. The materialographic preparation was carried out using a semi-automatic grinding and system (Struers RotoPol-31, RotoForce-4) – the procedure being described in Table 9. The preparation process was carried out as carefully as possible, to avoid microstructural damage.

79

Table 9: Preparation procedure for microscopic investigation of composite samples

GRINDING Step 1 Step 2 Surface MD-Piano 2201 MD-Largo2 [μm] 6 Lubricant water blue3 Speed [RPM] 300 150 Force [N] 50 50 Time [min] Planar 10

POLISHING Step 1 Step 2 Step 3 Surface MD-Dac4 MD-Chem5 MD-Chem5 Abrasive [μm] 3 OP-S6 Lubricant blue3 distilled water Speed [RPM] 150 150 150 Force [N] 60 10 10 Time [min] 6 0.5 1

1 disc with resin bonded for 4 disc with textile cloth (fine acetate) for grinding (particle size of 63 μm) polishing with high removal rates 2 composite disc for fine grinding 5 disc with neoprene foam cloth 3 cooling and lubrication liquid on 6 colloidal silica suspension with a grain alcohol basis size of about 0,04 μm

Optical microstructure investigations were done using a Leica Aristomet microscope with reflected light configuration and an adapted 1.3 megapixel camera (Leica DC500).

Apart of fracture analysis and preform characterisation, investigation of the MMC samples by SEM was conducted on polished surfaces – prepared as described in Table 9. In addition, samples of interest were ion milled to investigate surfaces free of mechanically-induced preparation artefacts. In the ion milling process, conducted by the Forschungsinstitut Edelmetalle und Metallchemie, Schwäbisch Gmünd, Germany, a mechanically-polished sample is bombarded by argon , resulting in an artefact-free and very smooth surface. (The process of ion milling and its suitability for composites used in this study is described by various authors205-207.) For comparison purposes vibration polishing was used as a second process for smooth final polishing of surfaces, with minimal mechanically-induced artefacts.

80

Microstructure investigation was done using a SEM (Zeiss Leo Gemini 1520) with field emission gun. To avoid electrostatic charging, preform samples were coated with a thin layer of AuPd, using a sputter coater (Bal-Tec MED020) where necessary.

Quantitative microstructure analysis was performed on images using the MosaiX algorithm, where a large area is displayed in high magnification by imaging of single images (MosaiX-tiles). These were adjusted with overlap within the tiles to acquire one MosaiX-image with integrated stitching algorithms to align the tiles correctly within the overlap zone. Those images were processed with the image analysis software Zeiss AxioVision (version 4.7.1).

Transmission electron microscopy investigation was done using a Phillips CM200 TEM on samples prepared by focused ion beam (FIB) milling (FEI Company, XT Nova NanoLab 200 DualBeam) at the UNSW node of the Australian Microscopy & Microanalysis Research Facility (AMMRF). In focused ion beam technology, ions at high beam currents rapidly sputter away the specimen surface, allowing subsurface cross-sections to be prepared. FIB can be used for nanomilling very thin sections suitable for TEM examinations. Therefore, a thin layer of 20 nm Au was sputter coated (Emitech K550, 2006) on the surface of copper/alumina samples to achieve good conductivity to eliminate the charging effects during SEM imaging and ion beam milling. To prevent the ion beam damage, a strip of Pt about 15 x 2 x 1 μm in size was deposited on to the region of interest in the FIB. The thin section about 100 nm thick was then prepared by removing the materials on both sides with the high energy ion beam at 30 kV and 5 nA, followed with polishing at lower beam currents (3 nA, 1 nA, 0.3 nA and finally 0.1 nA step by step). The samples prepared by FIB used for TEM investigations were placed on a carbon/formvar film supported copper grid using a needle with Kleindiek micromanipulator (Kleindiek, Germany) and stereo microscope.

Further microstructure analysis was carried out using 3-D computed tomography. Preforms and MMC samples were investigated using a 3-D computertomograph V|tome|x s (Phoenix|x-ray systems, Germany) with a Paxscan 2520 V detector of 1920 x 1536 pixels and a lateral dimension of

81

127 μm/pixel. Depending on material and sample geometry as well as investigation aim, the tube of the 3-D computertomograph could be changed from a microfocus tube with a maximum acceleration voltage of 240 kV to a nanofocus tube with an acceleration voltage of maximum 180 kV. Investigations of preforms were realised with the nanofocus tube; while for investigation of MMC samples the microfocus tube with high acceleration voltages was necessary in regard to the metal density. Preform testing was carried out to investigate the distribution of copper oxide or of aluminate before and after sintering. The acceleration voltage was set to 55 kV with a current of 260 μA and an imaging time of 750 ms. The resolution was 1 μm/voxel with a focal spot size of 0.7 μm. In MMC investigation, the parameters were adjusted depending on the thickness and copper content of the specimen.

Phase analysis has been performed with a Bruker AXS D8 micro- diffractometer with Bragg-Brentano Optics. For investigation of phases, preforms and composite materials have been analysed. Additionally, as no aluminate powder was commercial obtainable, the verification of the composed phase of aluminate powder fabrication has been performed by XRD measurements of the calcinated powders. Measurement was performed with

40 kV / 50 mA in ߠǦʹߠ geometry and Cu KD radiation.

4.4.3 Mechanical Properties Preforms and MMC were characterised differently. Whereas for MMCs a normal 4-point bending test was applicable, the preforms were mechanically characterised using the Brazilian disc test. In both methods the evaluation of characteristic strength ߪ଴ and Weibull modulus ݉ for each series (with n- samples) was performed using the maximum likelihood method according to equations 35 and 36 with ߪ௖ being the strength of failure:

1 ln VV m )* 1 ¦ cici  lnV m ¦ ci Equation 35 m ¦V ci n

1 m VV m 0 n ¦ ci Equation 36

82

For characterising preform strength, the Brazilian disc test was used on a universal testing machine Zwick Z100, according to the ASTM standard D 3967- 05208. Cylindrical specimens with a diameter of 18 mm and thickness of about 13 mm, hence a thickness-to-diameter ratio of about 0.75, were used for testing, which was performed with a cross-head speed of 2 mm/min (a schematic of the set-up for testing is given in Figure 29). For handling reasons the steel ball was flatted at one side and attached to the Si3N4 plate. The remaining point contact allowed enough degree of freedom to prevent non- orthogonal loading.

100Cr6 ball

Si3N4 Preform

Figure 29: Schematic of the set-up for the Brazilian disc test used for mechanical preform characterisation.

The load to the preform disc resulted in tensile stress along the median axis between the contact points of loading, meaning a measurement of tensile strength (splitting tensile strength) in the given test set-up. The calculation of the splitting tensile strength ߪ஻ of the specimen was calculated using following equation208:

2F V B ShD Equation 37 where ܨ is the maximum applied load in [N], ݄ the thickness and ܦ the diameter of the specimen in [mm].

83

The elastic modulus, bending strength ߪସ௉஻ and fracture toughness ܭூ௖ of the composite material were determined using bending test bars of ≥45 x 4 x 3 mm³. The testing was done according to standard DIN EN 843-1209 at ambient temperatures on a MTS mechanical testing machine with a 40 mm/20 mm configuration. Testing was performed using a crosshead speed of 2 mm/min. In order to characterise the deflection of the specimen, an inductive linear transducer with MGC Plus measuring amplifier (from HBM, Germany) was used. Determination of elastic modulus was performed on the same specimens before 4-point bending took part with an ultrasonic USLT 2000 system from Krautkrämer, Germany. Fracture toughness was evaluated using the single-edge V-notched-beam (SEVNB) method, according to draft standard DIN 14425-5210. The mechanical tests were conducted at Robert Bosch GmbH in Stuttgart, Germany.

4.4.4 Nanoindentation testing Häger211 determined the different phase characteristics in situ in the MMC structure by using nanoindentation. Nanoindentation was performed with a Hysitron TriboIndenter® to measure the mechanical properties of each phase in the MMC structure. The nanoindenter is accurate for very low forces in the range of 10 nN to 10 mN with a resolution of less than 1 nN and 0.2 nm in depth212. The patented transducer technology uses a capacitive displacement measurement technique combined with electrostatic force generation. Its piezo- electric scanner allows very precise indenter tip positioning; and the machine can be used for SPM (scanning probe microscopy) imaging by scanning the indenter tip over the specimen surface. In Figure 30 a typical load displacement curve is shown schematically with an in situ SPM image of MMC after indentation of the metal phase.

84

F F : load h : depth Ft Ft : maximum load ht : maximum depth

Figure 30: Schematic load displacement diagram of nanoindentation with in situ SPM (scanning probe microscopy) image of specimen surface after nanoindentation. 2 indents are marked with a green banner for reference.

Due to the fine microstructure and high differences in hardness of the phases of the material investigated, different loads had to be applied for metal and ceramic phases. The TriboIndenter® is load-controlled, and as only indentation depths of more than 40 nm are of validity for characterisation212, 400 μN were used for investigation of the copper phase and 2 mN for characterisation of the ceramic phases, alumina and aluminate. In all tests, a Berkovich indenter was used and the load function was set to 10 s for loading and unloading. The maximum force was held for 30 s to avoid any creep behaviour effect in the unloading curve. Properties were calculated from load displacement data according to ISO 14577.

4.4.5 Thermal Conductivity Thermal conductivity was measured at room temperature using a steady state test set-up. The principle is based on the theory of Fourier, the heat quantity Q evaluated by dQ dT   O A** dt dx Equation 38

the cross section, ݔ the length, ݐ the time ܣ ,where ߣ is the thermal conductivity of measurement and ݀ܶ the temperature difference. The experimental set-up is shown schematically in Figure 31:

85

Vacuum Heating rod

Reference specimen

Measured Thermocouple specimen

Cooling coil

Figure 31: Schematic of the test set-up for thermal conductivity determination.

Samples were of 6 mm in diameter and 40 mm in length. The length between the thermocouples was 35 mm. Thermocouples of type T with a diameter of 0.5 mm were used. The entire set-up was placed in a vacuum chamber to limit external influences of radiation and convection. Since the reference specimen, with a known thermal conductivity ߣோ௘௙, and the measured specimen had same dimensions and the heat flow was equal, the conductivity can be evaluated by

' *tT O AQQ ** Re f Spec l Equation 39

ORe f *'TRe f OSpec Equation 40 'TSpec

For the reference specimen a sample of an aluminium alloy (AlMgSi1) with a thermal conductivity of 188 W/mK was used. During testing the specimen temperatures were between about 30°C for the hot side and about 10°C for the cold side, depending on the conductivity of materials measured.

86

5 Results and Analysis This section describes the results and some analysis of the experimental investigations undertaken for fabrication of composite structures and the characterisation thereof, including powder preparation for preform synthesis, infiltration and microstructural investigations of the materials are made. Furthermore, especially the infiltration process is investigated more precisely with the effects occurring. Individual phase characterisation via nano- indentation and mechanical characterisation of the composite material is also given.

5.1 Ceramic Preform 5.1.1 Preparation of Ceramic Compound

The ceramic phase aluminate with the formula CuAlO2 was prepared in the manner shown in Figure 32 in 200 g batches. For a smoothly closed alumina crucible was used in a muffle furnace (Nabertherm N41). The atmosphere in the furnace was determined by flowing Argon gas.

41.61wt% Al2O3 110 ml dist. Water + + Calcination: 1050°C, 8 h, Ar 58.39wt% Cu2O 1400 g ZrO2 balls

Ball milling: 2,5h (ZrO2 balls) Attritor milling: 4h

Drying: Drying: Freeze drying Freeze drying

Specimen Sieve: 250 μm

Calcination: Calcination: 1000°C, 4 h, Ar 1050°C, 4 h, Ar XRD-Analysis (Bruker AXS D8)

Particle size distribution XRD-Analysis (Bruker AXS D8) (Mastersizer 2000)

Figure 32: Flowchart of manufacturing steps for preparation of aluminate used in preform fabrication.

87

Evaluation of calcination temperature and time was done using two different calcination temperatures – 1000°C and 1050°C. As the material had to be in powder form these temperatures were chosen to prevent liquid phase sintering. A subsequent XRD analysis showed aluminate formation at both temperatures: at the lower temperature reaction was incomplete in the remaining Cu2O, and at the calcination temperature of 1050°C the compound reaction was almost complete. With an increase in time at calcination temperature of 1050°C the compound reaction could be completed. In XRD analysis of the compound, processed at 1050°C for 8 h in Argon atmosphere, only aluminate with a little amount of dioxide was detected, as shown in Figure 33:

18840 18840 Lin (counts) Lin (counts) 0 0 20 40

18840 18840 Lin (counts) Lin (counts) 0 0 60 80 2-Theta - Scale

Figure 33: Diffraction pattern of the manufactured compound calcinated at 1050°C for 8 h in argon atmosphere, showing aluminate in two modifications – rhombohedral (red) and hexagonal () – with some abrasion product from balls (green).

Aluminate was detected in two different configurations – hexagonal (blue peaks) and rhombohedral (red peaks in Figure 33) – and the rhombohedral configuration clearly prevailed. ZrO2 was the abrasive of the milling balls used for milling and distribution. After heat treatment the reacted powder was milled for 2.5 h. In Figure 34 the particle size distribution of the manufactured powder is shown:

88

Volume (%)

Figure 34: Particle size distribution diagram of the calcinated aluminate powder with a specific surface area of 8.28 m²/g and d50 of 0.915 μm.

The particle size was characterised with d50 = 0.915 μm. An additional subsequent milling process to further refine the aluminate particles before powder preparation was not necessary. The particle size allowed good distribution of the aluminate particles in the preform structure during the powder preparation process.

5.1.2 Preform Fabrication In the powder preparation a wide variety of powder compositions was prepared. To investigate the influence of different additives – e.g., pore-forming additive or copper oxide – each composition necessary for characterisation was produced. All powder compositions result in specific preforms with their own characteristic behaviour. In Figure 35 the variation in preparation of different powder compositions for investigation of the parameters copper oxide and pore-forming additive is shown schematically. Resulting from this, in a further step the influence of sintering parameters’ time, temperature and atmosphere on the preform phase composition was investigated.

89

Composition Sintering Time

Atmosphere Copper oxide

Pore forming additive

Figure 35: Schematically parameter analysis of preform composition and subsequent sintering parameter analysis of one specific preform type.

The results of XRD-analysis of the preform type AC50 are shown in Figure 36 for an overview. With regard to the nomenclature shown in Figure 22, this preform type contains 5 wt% of copper oxide and no pore-forming additive. It is made of the alumina powder CL2500 with monomodal grain size distribution from Almatis. This type of preform was chosen as a compromise between easy phase detection and mechanically-relevant composition.

Time

24h CuAlO2 CuAlO2

CuAl2O4 + 2h CuAlO2 CuAlO2 CuAlO2

1200°C air vacuum argon Atmosphere CuAl O + 1400°C 2 4 CuO

Temperature

Figure 36: Results of the XRD-analysis of different sintering conditions on AC50 preform (main phase – alumina – excluded): the analysed ceramic phases were aluminate (CuAlO2), with Cu-spinel (CuAl2O4) and copper oxide (CuO).

Preforms without copper oxide addition were investigated for reference in the XRD and no copper/aluminium oxide phase was detected. The investigated preforms of type AC50 (5 wt% Cu2O addition) showed different phase appearance. Whereas preforms sintered at 1200°C and 2 h soaking time showed copper/aluminium oxide phases in the aluminate and spinel structure, increasing the soaking time to 24 h resulted in aluminate being present in the 90

preform and a spinel structure was no longer detected. An increase of sintering temperature to 1400°C resulted in the spinel structure being detected in the preform. When sintering in argon atmosphere or vacuum the aluminate phase was apparent: this was consistent with the isothermal section of the ternary phase diagram shown in Figure 37 (the initial preform system is marked with a red line). Alumina and cuprous oxide reacted to form the aluminate structure as stated in equation 30: only with abundant oxygen the spinel structure would form. In addition to alumina and the above-mentioned oxides, some of the preforms showed little monoclinic zirconia, which is the abrasion product of the milling balls.

Figure 37: Isothermal section of the ternary phase diagram of Cu-Al-O at 1200°C213. The initial preform system Al2O3-Cu2O is marked with a red line.

For MMC fabrication the ceramic volume content of the preform was evaluated and varied: with pore-forming additive the ceramic volume content decreased. Apart from copper oxide containing preforms, all of which were sintered at 1200°C, the ceramic volume content could be adjusted by varying the sintering temperature; meaning that for copper oxide containing preforms, the adjustment of ceramic volume content had to be made by varying the compressibility of the resulting powders by using different raw alumina materials or the addition of pore-forming additives. Further, for large additions the content of copper oxide 91

influenced the ceramic volume content. In Figure 38 an overview of some preforms is given, showing the influence of PFA and copper oxide on ceramic volume content. A wide variety of different preform types was prepared in the process described.

55

50

45

40

35

30 Copper oxide Pore forming agent Ceramic volume content [%] content Ceramic volume 25 0 5 10 15 20 25 30 35

Cu2O / PFA content [wt%] Figure 38: The influence of the addition of copper oxide to pure alumina in preform preparation on the resulting ceramic volume content (blue square): the influence of the addition of the pore-forming agent is shown for preforms containing 2 wt% Cu2O (grey triangles).

The influence of the addition of copper oxide on the ceramic volume content of the preform is only shown where it was in large amounts – these were probably not of technical relevance. The ceramic volume content for additions of up to 20 wt% of copper oxide was in a stable level at around 50–53%. In contrast, the pore-forming additive cellulose had an immense influence on the ceramic volume content, as expected. This influence was shown for preforms containing 2 wt% of copper oxide, all sintered at the same temperature. The ceramic volume content decreased from 52% to 30% by adding 30 wt% of cellulose. Adding just 5 wt% of cellulose resulted in a preform of around 45% ceramic volume content. Although not shown in Figure 38, there was again no influence on ceramic volume content for the pore-forming additive containing preforms when adding copper oxide in moderate amounts. The influence of the volume decrease was obvious, as the cellulose acted as a placeholder and would be pyrolised during sintering, whereas the copper oxide reacted with the alumina powder and remained in the preform structure. The resulting pore structure of cellulose particles was much coarser than the fine porosity between the 92

particles of the alumina powder, providing a bimodal pore structure for preforms containing a pore-forming agent. The structure resembled areas of preforms without pore-forming agent between the large pores resulting from the cellulose particles, as can be seen later in Figure 74.

Figure 39 shows an overview of preform compositions. For preform types with added copper oxide only, types prepared with CL2500 are shown; whereas for those without copper oxide addition the powders from Nabaltec were also taken into account (the preform types prepared in an industrial manner are replicates of some of these and therefore not shown):

35

AC20/30 AC200/30 30

25 N6/20 N6/20A20 AC20/20 AC100/20 20 PFA 15 N6/10 N3/10 A10 AC10/10 AC20/10 10 N6/5 N3/5 AC20/5

5 ≈ AC300

Pore Forming Additive Additive [wt%]Pore Forming AO AC1 AC5 AC10 AC20 AC50 AC100 AC200 AC400

0 ≈≈≈≈≈≈≈ AA17AA17 AA34AA34 AA85 0,01 0,1 1 10 100 Cu O [wt%] Cu2O2 Figure 39: Variety of preform types prepared in laboratory powder preparation. For types with copper oxide addition only the compositions with the alumina powder CL2500 of Almatis is shown, whereas for compositions without copper oxide the alumina powders of Nabaltec are also taken into account.

The preform types shown were all from laboratory preparation with copper oxide addition of 0 to 40 wt%, except for the types labelled ‘AA’: these were also prepared in the laboratory but with an according amount of aluminate, assuming a complete reaction of copper oxide with alumina during sintering for the preforms prepared with copper oxide (yellow circles). Therefore no phase formation of the components was expected, rather a finer distribution of the aluminate phase with regard to the grain size and the lack of reaction.

Sintering processes were determined regarding the data obtained by simultaneous thermal analysis (STA), in which several preform compositions 93

and raw materials were analysed to obtain information about pyrolysis behaviour and reaction temperatures. Analysis was performed using pressed powders to achieve preform-like behaviour. In Figure 40 a typical diagram of thermal analysis is shown for the preform type A10 with 10 wt% pore-forming agent in oxidising atmosphere.

0 30

-4 22 V] μ

-8 14 Weight [mg] Weight DTA-Signal [ DTA-Signal -12 exothermic 6

-16 -2 0 200 400 600 800 Temperature [°C]

Figure 40: Diagram of simultaneous thermal analysis of a preform of type A10 with 10 wt% of cellulose.

The green body analysed still exhibited a certain amount of water, resulting from binder activation for pressing: this could be seen at the loss of weight at around 100°C. The concluding reaction with weight loss could be attributed to the organic content of the green body from around 200°C up to 470°C. According to the material data sheet202, the polyvinyl alcohol Mowiol 18-88 decomposes above 180°C. Cellulose burns from temperatures above 200°C201. It is therefore not appropriate to differentiate the reactions of both organic ingredients. The first endothermic peak seen in the diagram above with major weight loss can be attributed to both. The second endothermic peak also appears when measuring only cellulose particles (FIG), which contributes to the cellulose with its specific particle distribution. As already seen in Figure 19 (and to be shown in microscopic images of composite samples), the cellulose investigated exhibited fibrous parts as well as particles. The sample investigated here was organic-free when the temperature reached 470°C – in 94

regard to which the heating ramp for preform sintering in the range of organic burning had to be slow. As a consequence a slope of 20 K/h was chosen, in the range 100°C–475°C. A typical sintering cycle for preforms containing copper oxide is shown in Figure 21. Preforms containing more than 10 wt% of cellulose had to be pyrolised in an argon atmosphere before a sintering cycle in air could be applied, or the high organic content would to spontaneous ignition, resulting in undefined and unusable preforms.

0 25

-5 20

-10 15 V] μ

-15 10 exothermic

Weight [mg] Weight -20 5 DTA-Signal [ DTA-Signal

-25 0

-30 -5 0 200 400 600 800 Temperature [°C]

Figure 41: Diagram of simultaneous thermal analysis of cellulose particles.

The appearance of the freeze cast preforms is shown in Figure 42.

Figure 42: Surface appearance of the freeze cast preforms.

95

The preforms exhibited a specific structure of lamellar ceramic ligaments. The ceramic volume content was about 40%. As the ligaments were of dense sintered alumina, the permeability was highly anisotropic and interfaces could easily be investigated providing an excellent model system.

5.1.3 Preform Characteristics 5.1.3.1 Computed Tomography Investigation Preforms were investigated using computer-assisted tomography x-ray analysis. As the resolution of tomography relies (beside the x-ray parameter) on the greatest geometrical magnification possible, small samples of preforms were prepared. The volume investigated for each analysis was about 3 mm³. Figure 43 shows an image of the computer-compiled scans, from a frontal direction, of the microstructure of a preform with 5 wt% copper oxide added.

Figure 43: Image of the computer-compiled scans from frontal direction of a sample with 5 wt% Cu2O addition: big agglomerates (bright phases) are observable.

The microstructure showed a relatively homogeneous alumina distribution with agglomerates (bright): these were well distributed, and the fine microstructure of alumina particles can be seen. There were no irregular large pores or other defects like cracks detected. In Figure 44 the microstructure of a specimen prepared with pre-reacted aluminate powder, investigated in the same manner, is illustrated. The amount of aluminate powder equalled the amount of 5 wt% copper oxide addition – the prerequisite being a full reaction of copper oxide with alumina to aluminate. Therefore the amount of the copper/aluminium oxide was the same for both preforms, but prepared in different ways. As the pre- reacted aluminate was of fine particle size distribution and the reaction complete, no coarsening of the particles due to reactions or agglomeration during powder processing was expected.

96

Figure 44: Image of the computer-compiled scans from frontal direction of a sample with pre-reacted aluminate. Amount of aluminate equals the amount of 5 wt% Cu2O (prerequisite being a full reaction). A few little agglomerates (bright phases) are observable.

The microstructure was nearly free of agglomerates and the particles of aluminate seemed much more finely distributed. The microstructure was homogeneous in particle and pore distribution. By adding the pore-forming additive the pore size distribution was expected to differ. In Figure 45 a sample with a high amount of copper oxide and pore-forming additive is shown: 10 wt% of copper oxide and 20 wt% of pore-forming additive was added to the alumina powder. The figure displays an image of the computer-compiled scans from axial direction.

Pore -resulting from PFA-

Agglomerate

-resulting from Cu2O-

50 μm

Figure 45: Image of the computer-compiled scans from axial direction of a sample with 10 wt% Cu2O and 20 wt% pore-forming additive addition (AC100/20).

The agglomerates were larger and of different shape, compared with the sample seen in Figure 43, but again well distributed. The bimodal pore structure can easily be seen, with the shape of cellulose particles formed in the micro- structure remaining as big pores, illustrated by black areas. The surrounding structure of alumina particles was comparable with the samples without pore-

97

forming additive, seen above. In Figure 46 a 3-D image of this sample (AC100/20) compared with the sample containing the pre-reacted aluminate powder (AA85) is shown.

AA85 AC100/20

100 μm 100 μm

Figure 46: A 3-D view of two different preform samples (AA85 – with pre-reacted aluminate powder, equalling the amount of 5 wt% Cu2O and AC100/20 – containing 10 wt% Cu2O and 20 wt% pore-forming additive) with 600 μm edge length.

The use of pre-reacted aluminate resulted in an homogeneous microstructure almost free of agglomerates. The agglomerates already formed from copper oxide in the preform preparation process, as measurements of green bodies highlighted the same microstructural appearance. In green bodies the agglomerates were of copper oxide with a subsequent reaction to copper/aluminium oxide. The use of pre-reacted aluminate powder therefore prevented the reaction during sintering and no agglomerate formation occurred in the powder preparation process. Due to the fine particle size, the distribution of the pre-reacted aluminate was fine and homogeneous in the preform microstructure.

5.1.3.2 Permeability The permeability of preforms was investigated with regard to the ceramic volume content and the content of copper oxide. In Figure 47 the measured permeability dependence on the ceramic volume content of preforms is shown. The permeability is defined by the volume flow of a medium through a porous body at a defined pressure drop, as given in equation 34. The law of Hagen- Poiseuille, which can be derived from the Navier-Stokes equations, describes exponential behaviour of the volume flow in dependence of the cross-section of

98

flow-through medium. Taking this into account, an exponential behaviour in dependence of ceramic volume content of preforms is expected.

1E-12

1E-13 ] 2

1E-14

Permeability [m Permeability 1E-15

1E-16 25% 35% 45% 55% 65% Ceramic volume content

Figure 47: Permeability depending on ceramic volume content of different preform types.

The strong influence of ceramic volume content specifically the addition of pore- forming additive can easily be seen. The diagram shows permeability measurement values of different preform types with their specific ceramic volume content. Regardless of the copper oxide content, the values follow the function ௏೛ with a coefficient of determination of 94.7%. TheכൌͳͲିଵଵ݁ିଵହǤଶଵܭ permeability was dependent on the ceramic fraction. This seems obvious, as both the alumina and the copper oxide or aluminate showed non-wetting behaviour to water and did not react. So the measurement was performed in an inert condition and no reaction occurred during contact. Preforms exhibiting pore-forming additive are expected to show slightly higher values perpendicular to the pressing direction, because of the fabrication of green bodies. As compaction was performed in uniaxial pressing, the big irregularly-shaped cellulose particles aligned perpendicular to the pressing direction, leading to a somewhat texturized microstructure, as can easily be seen after infiltration (see

99

Figure 72). However, this has not been measured as the infiltration direction is parallel to the pressing direction.

5.1.3.3 Appearance of Sintered Compacts Preform structure was observed using SEM. For investigation, the preforms were fractured and the resulting surface used for imaging. Figure 48 shows the surface of a preform type with 2 wt% copper oxide added after sintering at 1200°C. The agglomerates resulting from the copper oxide addition are indicated by circles. Imaging was performed in backscattered mode, hence the aluminate agglomerates appear more bright.

5 μm

Figure 48: SEM image of a broken surface of a preform containing 2 wt% copper oxide: resulting agglomerates are indicated by circles.

Alumina particles were partly sintered, resulting in fine porosity preforms. The platelet-shaped alumina particles within the microstructure were typical of the alumina powder used. Beside the agglomerates of aluminate finely distributed bright dots can be observed: as the imaging was done in backscattered mode for phase contrast, this might also represent aluminate, but finely distributed. This will be microscopically investigated further in the composite structure.

100

5 μm

Figure 49: SEM image of the broken surface of a preform sample containing 10 wt% pore- forming additive: bimodal pore structure can easily be seen.

Figure 49 shows the surface of a preform of type A10. The bimodal pore structure resulting from the addition of 10 wt% cellulose can be seen by the big pore that is the result of a cellulose particle. The surrounding area exhibits fine porosity and the alumina particles are partly sintered. As the sintering temperature was 1600°C, the consolidation of the particulates was progressed further than the specimen in Figure 48, as can be observed from the alumina particulates.

5.1.3.4 Preform Strength Preform strength was evaluated using the Brazilian disc test, in which the tensile splitting strength is characterised. Determination of characteristic strength ߪ஻ and Weibull modulus ݉ was carried out using the maximum likelihood method, as described in section 4.4.3. In Figure 50 the Weibull distribution of various preform types is shown.

101

1,5

1,0

0,5

0,0

-0,5

-1,0 ln(ln(1/(1-F))) -1,5 VB munb. AC20/30AC20/30 0.31 6.5 -2,0 AC20/3SAC20/3S 1.0 4.0 AC20SAC20S 2.7 5.7 AOAO -53% 1.3 5.5 -2,5 AOAO high-59% 12.5 4.1 A1SA1S 17.6 8.6 -3,0 -20246 ln(V)

Figure 50: Weibull distribution for preform types containing different amounts of Cu2O and pore-forming additive.

The unbiased Weibull modulus of preform type AC20S (2 wt% Cu2O added) was ݉௨௡௕ = 5.75 and the characteristic strength ߪ஻ = 2.7 MPa. The fracture stress varied from 1.7 MPa, to 3.0 MPa representing the highest fracture stress. The average strength was ߪത = 2.5 MPa with a standard deviation of 0.5 MPa. The strength of the preform was highly dependent on the fraction of ceramic it contained. A reduction of ceramic fraction at identical sintering parameters was obtained by the addition of the pore-forming additive; the preform type AC20/30, prepared with 2 wt% Cu2O and 30 wt% cellulose exhibited a ceramic volume content of 31% compared with the 51% of the above. The strength was

ߪ஻ = 0.34 MPa and the Weibull modulus ݉௨௡௕ = 6.47. Taking into account the preform type AC20/3S with a ceramic volume fraction of 47% and a characteristic strength of ߪ஻ = 1.0 MPa showed the increase of strength with ceramic volume content to notice. Table 10 gives a short overview of some preform types with their characteristic properties measured.

102

Table 10: Overview of some preform types with their characteristic properties, measured by the Brazilian disc test.

Preform type Ceramic fraction Tensile splitting strength Weibull modulus [ ] [vol%] V B [MPa] munb [ ]

AC20/30 31 0.34 6.5

AC20/3S 47 1.0 4.0

AC20S 50 2.7 5.7

AO 53 1.3 5.5

AOhigh 59 12.5 4.1

A1S 60 17.6 8.6

Preform types without added copper oxide were sintered at different conditions.

Whereas the type AOhigh (no addition of pore-forming additive) was sintered at 1500°C, the type A1S (10 wt% cellulose added) was sintered at a temperature of 1600°C. The tensile strength for the higher sintered preform type was

ߪ஻ = 17.6 MPa with ݉௨௡௕ = 8.6, whereas the lower sintered type showed a strength of ߪ஻ = 12.5 MPa with ݉௨௡௕ = 4.1 exhibiting almost the same ceramic volume content. The reference sample without the addition of copper oxide and pore-forming agent (AO) exhibits a tensile splitting strength of ߪ஻ = 1.3 MPa with ݉௨௡௕ = 5.5 and a ceramic volume content of 53%. These samples were sintered at the same temperature as the copper oxide-containing samples to allow comparison of microstructure. Densification during sintering did not occur, showing poor sinter ability at such low temperatures for the pure alumina sample. Nevertheless, preform strength was high enough to ensure handling for further processing. In fact, the strength was comparable with the AC20/3S sample, overlooking the difference in ceramic fraction.

5.1.4 Characterisation of Preforms for Simplified Investigation Plating was performed on dense ceramic substrates in order to justify the parameters for copper plating, like temperature, time and solution agitation simply. In Figure 51 some alumina substrates used for prelimary bath experiments are shown.

103

Ground surface Polished surface

50 μm

Figure 51: Deposition of copper on dense alumina substrates with different surface quality: surface appears reflective on a polished sample, whereas a ground sample results in a matt surface.

Copper plating was effective for the given parameter set. Plating without surface activation was not possible – as can be easily seen in the figure above, at the front left corner of the samples. The appearance of the coating was strongly dependent on the substrate surface. With polished surfaces the coating acts like a mirror in contrast to a relatively rough substrate: this reveals that the bath is working correctly. As chemical coatings do not deposit selectively, even microscopically, the coated layer is uniform in thickness and replicates the original surface. This applies if the bath flow allows uniformly ion concentrations close to every point of surface. Figure 52 shows a freeze cast preform after coating with copper from the top side and a cross section.

2 mm 10 mm

Figure 52: Top view and cross section of a freeze-cast preform coated with copper.

The layer of copper was uniform over the height of the preform examined macroscopically. For the investigation of the layer thickness and deposition behaviour, a fragment of coated preform was characterised in SEM. As an example, an image of a copper-coated lamella is shown in Figure 53: in addition, an edge of coating used for thickness analysis and deposition behaviour is shown in higher magnification. 104

(a) (b)

5 μm 1.5 μm

Figure 53: SEM images of copper coating on freeze cast lamella: the coating covers the surface almost completely, representing the alumina grain structure. In (b) the edge of the deposition film can be seen.

Even microscopically the layer deposition was uniform over the height of the preform: the layer reproduced the grain surface of the alumina lamella. In the SEM investigation some individual sites of the preform’s lamellae appeared without surface coating: this could be the result of a preform structure with some single walls covered. The layer thickness depends on coating parameters, especially the plating time.

RT 600°C 1150°C

Cu CuO CuAlO2

10 mm

Figure 54: Above – copper-coated dense ceramic substrates treated at different temperatures with their main phases of layer: below – a Cu-coated freeze cast preform, heat treated at 1150°C.

105

In Figure 54 the influence of thermal treatment on the copper layer is shown. Before infiltration the preforms had to be preheated to high temperatures, leading to a reaction of the copper with the ceramic substrate. This was analysed and investigated by coating the dense ceramic substrates with copper and a subsequent thermal treatment. Phase characteristics were evaluated using XRD analysis. The main phases are noted on Figure 54: as can be seen, the copper formed copper oxide for the sample treated at 600°C and reacted at 1150°C with the ceramic substrate, resulting in a surface layer mainly of aluminate. As a consequence of preheating before infiltration, the interface of the latter composite was of aluminate phase. The appearance of such a preform, coated and subsequent heat treated, is shown at the bottom of Figure 54.

5.2 Infiltration Behaviour 5.2.1 Characteristics of Preform Infiltration The infiltration behaviour was dependent on the preform characteristics such as ceramic volume fraction, preform chemistry and composition – and, of course, pore structure. This was observed macroscopically during infiltration experiments. Infiltration depth and quality in the macroscopic sense could easily be seen by investigation of cross-sections in the infiltration direction. In Figure 55, three coarse-ground cross sections of MMCs with different preform compositions are shown: (a) pure alumina, (b) alumina with 2 wt% Cu2O and (c) alumina with 2 wt% Cu2O and 5 wt% pore-forming additive. The excess copper remaining from the infiltration progress, was milled on the upper and lower sides for better observation, while the copper casting around the MMC was still present, as can easily be seen in the images. The milling depth was undertaken to lay the MMC bare without changing its height.

106

(a)

Cu Composite L = 6.2 mm

(b)

Cu Composite

(c) Cu Composite 10 mm

Figure 55: Cross-sections of copper infiltrated preforms: the composites are of different composition for infiltration investigation conducted with the same infiltration parameters. (a) pure alumina preform; (b) alumina preform with 2 wt% Cu2O addition; and (c) preform with 2 wt% Cu2O addition and 5 wt% pore-forming agent. Ceramic volume content is comparable at around 50% for all three.

Infiltration was performed with a tool made of hot working steel at a temperature of 420°C. The infiltration conditions for all preforms were the same (Squeeze casting with 1250°C melting temperature, 1170°C preform preheating and 100 MPa infiltration pressure). All three preforms exhibit a comparable ceramic volume content of around 50%. Figure 55(a) shows the infiltration behaviour of a preform of pure alumina (AO): complete infiltration did not occur and the infiltration front reached only about 6.2 mm in depth. Further infiltration was not possible for this preform type with given infiltration conditions. In Figure 55(b), infiltration of the entire height of 8.15 mm of preform can be seen, where a preform with 2 wt% Cu2O was infiltrated. There are still macroscopic infiltration defects observable in microstructure. Infiltration of the entire preform height could also be performed for preforms with thickness of 9 mm and more. With adding copper oxide (2 wt%) and pore-forming additive (5 wt%) to the preform, the infiltration was complete and defect-free over the entire preform height – as can be seen in Figure 55(c).

Figure 56 shows three cross-sections of composite samples – one of a pure alumina preform (Figure 56(a)), one with addition of 10 wt% cellulose to pure alumina (Figure 56(b)) and one with 5 wt% copper oxide added at no pore- 107

forming additive (Figure 56(c)). Infiltration was performed with the same infiltration conditions as the samples shown in Figure 55; this can be monitored at the sample without copper oxide addition (Figure 55(a) and Figure 56 (a)), as it provides the same appearance of infiltration progress. In contrast to the images in Figure 55 the excess copper was not milled, exhibiting the entire height of the samples with the amount of copper used for infiltration. Considering the improvement of infiltration by adding pore-forming additive; the change in pore structure improved the infiltration behaviour of both copper oxide-free and copper oxide-containing preforms. Pure alumina preforms became able to be fully infiltrated with the addition of 10 wt% cellulose, as shown in Figure 56 (b).

(a) Copper

Composite

Uninfiltrated preform (b) Copper

Composite

(c) 5 mm Copper

Composite

Figure 56: Cross-sections following copper infiltration of preforms of (a) pure alumina showing partial infiltration, (b) alumina with 10 wt% cellulose added and (c) alumina containing 5 wt% copper oxide both showing full infiltration.

The improvement of infiltration with the addition of the pore-forming agent was expected, because of the changes in pore structure. Assuming cylindrical pore channels the infiltration was strongly dependent on the radius of the capillary, with regard to the Laplace-Young equation, in which the pressure is related to

108

the surface tension and curvature, as already shown in equation 18, but re- presented:

§ 11 · ¨ ¸  'P J lv ¨  ¸ © rr 21 ¹ where ȟܲ is the difference in pressure between the droplet and the surrounding medium, ߛ௟௩ the surface tension and ݎଵ and ݎଶ the radii of curvature. This means that the capillary back-pressure, which has to be overcome for infiltration, is directly related to the pore diameter. With the addition of the pore- forming agent (cellulose), large pore channels are introduced into the preform which promote infiltration because of the lower capillary back-pressure and, hence, the lower infiltration pressure needed (this will be shown again later in this section, with infiltration images).

Comparing the samples containing copper oxide, the infiltration improves with copper oxide content. Whereas the sample with 2 wt% Cu2O (Figure 55(b)) still shows some infiltration defects, the sample shown in Figure 56(c) exhibiting

5 wt% Cu2O appears fully infiltrated. Apart from the fluid flow aspect of the enhancement of the infiltration ability by designing big pore channels for promoting the infiltration front, there are also chemical aspects. With the addition of copper oxide the pore structure of the preform remained essentially unaffected. Referring again to the Laplace-Young equation, for infiltration a decrease in surface tension leads, as well as the increase of pore radius, to a lower capillary back-pressure and, hence, better infiltration, due to the lower infiltration pressures needed. Chaklader et al.27 describe in their work the improving of wetting due to the reaction of copper with alumina to aluminate via the interim product copper oxide. They attribute the enhanced wetting to the lowering of the interfacial energy caused by the chemical reaction at the interface. Aksay et al.41 investigate the wetting in equilibrium and non- equilibrium conditions theoretically, and conclude for non-equilibrium that mass transfer at the interface initially results in a decrease of interfacial free energy and interfacial tension and hence promotes wetting due to the chemical reaction. Accordingly, the improving of infiltration due to the small copper oxide

109

addition was not expected, as the reaction of copper oxide to aluminate has already been stated in the sintering process, as was described in section 5.1.2. Diemer5 extensively investigated the wetting properties of copper on alumina. He reports the wetting angle of copper on aluminate to be similar to the wetting angle of copper on alumina in the oxygen partial pressures investigated but

ିସ states, also that, the wetting angle at ݌ሺܱଶሻ൐ͳͲ bar is time-dependent. Macroscopically shown in Figure 55, the infiltration ability of preforms containing copper oxide improved significantly. As well as a reaction at the interface, an alloying of the melt with oxygen or oxygen active metals enhances the infiltration by decreasing the contact angle. These alloying elements absorb at the interface or surfaces, reducing their tension2, 30. Referring to the above, the infiltration of preforms containing copper oxide might not be of inert behaviour, even if the reaction of copper oxide to aluminate had already happened during the sintering process. The infiltration behaviour of preforms with added copper oxide during powder preparation was enhanced, compared with pure alumina preforms. As the macroscopic characterisation of infiltration depth and appearance was not suitable for evaluation, the infiltration behaviour was measured using an instrumented infiltration set-up – described in section 5.2.3.

5.2.2 Performing the Infiltration Process The infiltration was dependent of the preform composition, as shown above. For investigation of the infiltration initiation pressure, infiltration conditions had to be found where complete infiltration was possible for all preforms of interest. As the infiltration was a duel between solidification and pore-filling, the process was highly temperature- and time-dependent. While the time was being set by the experimental conditions, the temperature was variable. Preform preheating, melt superheating and tool temperature were possible methods of increasing the system thermal energy: the tool being the coldest part, an increase in tool temperature promised the best results. Furthermore an increase in preform preheating higher than sintering temperature was not appropriate. Cross- sectional infiltration samples at different tool temperatures are shown in Figure 57. Infiltration was performed with samples containing 2 wt% of copper oxide and no pore-forming additive.

110

tool temperature of 600°C tool temperature of 600°C

tool temperature of 700°C tool temperature of 700°C

tool temperature of 800°C tool temperature of 800°C

tool temperature of 900°C tool temperature of 900°C

Figure 57: Cross-sections of infiltrated alumina preforms containing 2 wt% Cu2O. The infiltration behaviour was investigated macroscopically by varying the tool temperature from 600°C to 900°C.

The quality of the infiltrated preform, the MMC, and hence the infiltration increased with tool temperature. At 600°C the infiltration was still incomplete and many cracks occurred. A further increase in tool temperature led to better infiltration and fewer cracks: with temperatures of 900°C complete and crack- free infiltration became possible. Crack-free infiltration of preform types with pore-forming additive, even if no copper oxide was added, could be performed at a tool temperature of 800°C, as can be seen in Figure 58. As an example, a preform without copper oxide addition but 10 wt% pore-forming additive at two different temperatures is shown (Figure 58).

tool temperature of 600°C tool temperature of 800°C

Figure 58: Cross-sections of infiltrated, pore-forming additive containing preforms (10 wt%) at different tool temperatures for infiltration.

The preforms containing pore-forming additive were able to be infiltrated at 800°C tool temperature. Due to the cold tool, compared with the melting 111

temperature of copper, the infiltration was not complete on the side where the preform was in contact with the tool. As a consequence the preforms had to be 1-2 mm higher than the required infiltration height. This is acceptable, as it leads to one-sided infiltration, with no air entrapment in the microstructure. Double-sided infiltration, as shown in Figure 58 for a tool temperature of 600°C, leads to a centre line: this region in the middle of the infiltrated specimen is not infiltrated because of air entrapment. The air in the preform cannot evaporate as the infiltration fronts move towards each other driving the air before them. One- sided infiltration, therefore, is a favourable process for fabrication of composite samples and can largely be managed by appropriately lining the infiltration die with insulation material. For characterisation of the composite material only samples which were well-infiltrated from one side were used.

5.2.3 Instrumented Squeeze Cast Infiltration The measurement of infiltration initiation pressure was performed at 900°C tool temperature. Two different models of infiltration are common to describe infiltration in the literature:

(i) slug flow assumption handles the infiltration with an abrupt infiltration front and is based on Darcy’s law in its simplest form. Once the capillary back-pressure has been overcome, the infiltration pressure increases linearly with the distance of the infiltration front. This model is most common in modelling infiltration. (ii) The second model, adapted from soil science, is infiltration in a gradual manner. This means large pores are filled first and permeability decreases with saturation of porosity. Thus the infiltration takes place gradually and the infiltration front is somewhat diffuse.

A schematic infiltration curve for fibre preforms with its termination, as investigated by Long et al.120, 214, was shown in Figure 12. In Figure 59 the infiltration behaviour of a preform with and without pore-forming additive for comparison are shown, with the infiltration parameters (ܶ௧௢௢௟: 900°C,

ܶ஼௨: 1250°C, ܶ௣௥௘௙Ǥ: 1170°C, ܲ௔௣௣௟Ǥ: 92 MPa) kept constant. The diagram shows only the section of interest of the entire measurement with the applied

112

hydrostatic pressure as a function of relative plunger displacement, where the appearance of the model mentioned by Long et al.120 can be recognised.

10

8

6 0% PFA [MPa] 4 applied

P 3% PFA 2

0 0246 relative displacement [mm]

Figure 59: Graph of melt penetration in preform infiltration showing the applied pressure as a function of plunger displacement of a preform with (3 wt%) and without added pore- forming agent. Characteristic infiltration initiation pressures cannot be evaluated from the drawing.

The preform with pore-forming agent (3 wt% cellulose) exhibited a ceramic volume content of 47.5%, about 5% lower than that of the preform without cellulose addition. Additionally the preform height was different, resulting in a higher total porosity of the preform with the pore-forming additive. This can be seen by the difference in plunger displacement for infiltration in the penetration curves shown above. Other than this, the pressure needed for infiltration was very similar – at about 6 MPa. This observation might be obvious, as infiltration pressure can be related to the pore diameter, according to capillary law (equation 17). This reveals the pore diameter, so the microstructure of preforms are similar, related to the fine porosity not governed by pore-forming additives – to be shown in microscopic images in section 5.3. The bimodality of pore structure of this preform type results, due to the big pores of cellulose addition, in a lower pressure gradient in the beginning of infiltration, as can also be seen in Figure 59. The initiation of infiltration, expressed by the threshold pressure

113

ܲ଴, could not be evaluated directly from these infiltration curves: in order to determine this parameter the infiltration initiation stage was investigated in more detail.

The infiltration depth is a square root function of the applied metal 48 pressure ܲ௔௣௣௟Ǥ, according to Darcy’s law ; so linear behaviour was expected ଶ for plotting ܮ against ܲ௜௡௙Ǥ. The first 2–5 mm of infiltration depth were plotted assuming a one-sided infiltration. The data were fitted to a straight line by linear regression with slope mL and intercept on the pressure axis, with coefficients of determination of 97.2–99.2%. The intercept equalled zero infiltration, corresponding to the starting point of melt intrusion expressed by the infiltration initiation pressure ܲ଴. As shown in Figure 60 the behaviour was close to linear for both preform types, with and without added copper oxide.

5,0

4,5

4,0

3,5 [MPa] inf. P 3,0

Al2O3, Cu infiltrated 2,5 Al2O3, Cu-O infiltrated Al2O3/Cu2O, Cu infiltrated Al2O3/Cu2O, Cu-O infiltrated 2,0 0 5 10 15 20 25 Square of infiltration depth L² [mm²]

Figure 60: Characteristic behaviour of infiltration initiation of pure alumina preforms (circles) and preforms containing 2 wt% of copper oxide (triangles). Infiltration with oxygen-free copper is plotted in closed symbols, while the open symbols indicate the infiltration with copper-oxygen (0.9 at%) alloy. Data-fitting and evaluation of infiltration initiation (L=0) by linear regression.

Infiltration was performed with samples containing 2 wt% of copper oxide and no pore-forming additive. The alloying of copper melt was done by stirring

114

2 wt% of copper oxide into the melt before infiltration. The preform with added copper oxide showed an infiltration initiation pressure of 2.6 MPa while the pure alumina preform exhibited 2.7 MPa (dashed lines) for infiltration with oxygen- free copper. The microstructure of a copper oxide-containing and a copper oxide-free sample is provided in Figure 80, showing comparable alumina structure. Since the permeability and pore structure of both preform types are similar, no big difference of ܲ଴ was expected. The ceramic volume content for both particular preforms was 50.7% for the specimen with added copper oxide and 51.3% for the pure alumina sample. By adding oxygen to the melt, the initiation pressures were reduced to 2.3 MPa for the copper oxide-containing preform and 2.35 MPa for the pure alumina preform. The ceramic volume content was 51.2% for the preform with added copper oxide and 52.5% for the pure alumina sample, which would not explain the difference in ܲ଴. In fact, the slightly higher ceramic volume content compared with the infiltration with oxygen-free copper should have led to a higher initiation pressure. Hence the reduction of ܲ଴, can be attributed to the oxygen content in the infiltration melt. 48 According to Garcia-Cordovilla et al. ܲ଴ is calculated, and an infiltration initiation pressure of 3.3 MPa for the infiltration with pure copper and a straight reduction to about 0.5 MPa for the alloying with oxygen (0.9 at%) expected due to the enhanced wetting. The discrepancy between the measured and the calculated infiltration initiation pressures with Cu-O alloy might be a consequence of the experimental set-up. As the alloying of copper was done five minutes before infiltration, and the melting conditions were of a reducing nature (carbon crucible, argon atmosphere), a lowering of oxygen content in the melt was the necessary consequence. The contribution of atmosphere and dissolved oxygen in the melt is described in the investigations of Diemer5; and as seen in the literature48, 51, a reduction of infiltration initiation pressures due to oxygen influence is to be expected.

The increase in pressure, indicating the progress of infiltration, can be expressed by the slope ݉௅ of the regression line. It is interesting to note that the complete infiltration curve for the copper oxide-containing sample was shifted in parallel with smaller pressures under oxygen alloying of the copper

115

4 melt. The slope ݉௅ on the infiltration curve was 0.0855 N/mm for copper infiltration and 0.0824 N/mm4 for copper-oxygen infiltration. For pure alumina reference sample this behaviour could not be seen physically, as a steeper 4 increase for oxygen-alloyed copper (݉௅ = 0.1174 N/mm ) – but starting at lower initiation pressure – compared with infiltration with pure copper 4 (݉௅ = 0.1093 N/mm ) was measured. Comparison of the preform compositions showed the pure alumina samples to have a steeper increase. This indicated better infiltration behaviour for copper oxide-containing preforms, as shown in Figure 55, macroscopically. Since the permeability and the pore structure of both preform types were similar, no big difference of ܲ଴ and ݉௅ was expected. In contrast, the characteristic of the plots varied with preform composition in spite of the copper oxide content as shown in Figure 61. Infiltration was performed with oxygen-free copper.

5,0

4,5

4,0

3,5 [MPa]

inf. 3,0 P

AO_390 wt% Cu2O AC1_20.1 wt% Cu O 2,5 2 AC5S_80.5 wt% Cu2O

AC20S_4272 wt% Cu2O 2,0 0 5 10 15 20 25 Square of infiltration depth L² [mm²]

Figure 61: Characteristic infiltration behaviour of the initiation stage of preforms with copper oxide contents of 0, 0.1, 0.5 and 2 wt%.

The rate of increase in the applied pressure with infiltration of pure alumina preforms was steeper than for those containing copper oxide. The slope ݉௅ of the regression line of preform infiltration became progressively reduced with the addition of copper oxide to the preform (preforms with contents of 0, 0.1, 0.5

116

and 2 wt% Cu2O were used for this investigation). An overview is given in Figure 62.

0,12 5

slope mL threshold 4,5 0,10 pressure P0 [MPa] 0 ] 4 4 0,08

[N/mm 3,5 L 0,06 3 slope m 0,04 2,5 Threshold pressure P Threshold pressure

0,02 2 0 0,5 1 1,5 2 2,5 copper oxide content [wt%]

Figure 62: Slopes and threshold pressures of the infiltration behaviour of preforms with 0, 0.1, 0.5 and 2 wt% copper oxide content.

The slopes ݉௅ were evaluated by linear regression with coefficients of determination between 98.2 and 98.7% and could be expressed with slopes ଴ 4 ଴Ǥଵ 4 ଴Ǥହ 4 of ݉௅ = 0.1093 N/mm , ݉௅ = 0.0866 N/mm , ݉௅ = 0.0677 N/mm and ଶ 4 ݉௅ = 0.0565 N/mm , where the superscripts 0, 0.1, 0.5 and 2 represented the matching copper oxide content of the powder composition. Evaluated initiation pressures, in contrast, remained fairly unaffected, showing no real difference. The threshold pressures were between 2.6 MPa for the sample with 0.5 wt% added copper oxide and 2.7 MPa for the sample of pure alumina, showing no specific effect from the amount of copper oxide added to the preform on threshold pressure. This was in accordance with the theory of capillary law of infiltration, where the infiltration initiation pressure is dependent on pore or capillary size: the pore structure is not affected by copper oxide addition, but alters with the addition of pore-forming additive. A difference in threshold pressure was therefore expected for preforms with bimodal pore structure: the bimodality resulted in a lowering of threshold pressure, as illustrated in Figure 63, creating the bigger pores resulting from the pore-forming additive. 117

3 bimodal pore structure 2,5 [MPa] 0 2

1,5

1

0,5 Threshold pressure P Threshold pressure

0

Figure 63: Evaluated threshold pressures of different preform types: infiltration was performed with oxygen-free copper.

Threshold pressures with an average of 2.7 MPa for pure alumina preforms to 2.6 MPa for preforms with 2 wt% copper oxide addition were evaluated, showing no remarkable difference. With the addition of the pore-forming additive the threshold pressure reduced to 1.7 MPa for 3 wt% addition and 1.6 MPa for 7 wt% addition. This is in accordance with the theory of the capillary law of infiltration, in which infiltration initiation pressure is dependent on pore or capillary size. Since the pore structure of preforms with or without copper oxide added is the same, and no chemical issues change the wetting properties, no difference in infiltration initiation can be expected. Aluminate, also present at the preform surface and, hence, in initial contact with the melt, does not affect the wetting properties as reported by Diemer5, who shows the contact angle of copper on alumina as equal to that of copper on aluminate. The addition of pore-forming additive, in contrast, changes the pore size of the preform and thus improves the initial infiltration in terms of a reduction of threshold pressure. The strong improvement in infiltration behaviour due to the addition of copper oxide could therefore not be related to the early infiltration stage, governed by infiltration initiation or expressed by the threshold pressure.

118

5.2.4 Infiltration at Constant Pressures A set-up for constant pressure infiltration was used to verify the initiation pressures measured. Pressures of 1–5 MPa were applied in 1 MPa steps in order to review the pressures measured in the constant flux method. Investigation was done by evaluation of cross-sections after infiltration. For infiltration and direct comparison a design of specific samples was necessary. To gain comparable results the preforms were quartered: four of those quarters of different preform types were stuck together with high temperature-resistant mortar adhesive to represent one single infiltration specimen for constant pressure infiltration. For those four different preform types the infiltration conditions are exactly the same so they can be compared directly. Figure 64 shows a specimen: the dotted lines indicate the resulting MMC cut for investigation of infiltration progress in cross sectional view.

AC20/3S AC20S N6C20/3S N6C20/7S

Figure 64: Example of specimen for constant pressure infiltration containing 4 different preform types; the dotted line indicates the MMC being cut for investigation of infiltration progress in cross-sectional view.

Figure 65 shows a cross-sectional view of different sample types infiltrated at pressures of 1 MPa and 3 MPa. Infiltration was performed with 4 samples in one step for each pressure with different combination of preform types.

119

(a) Gas pressure of 1 MPa:

10 mm 10 mm

AO A10 AC100/20 A20

(b) Gas pressure of 3 MPa:

10 mm 10 mm

AO AC50 AC20S AA34

Figure 65: Gas pressure infiltration in one step of 4 samples: at pressure of (a) 1 MPa infiltration is seen for highly porous preforms (A10, A20 and AC100/20) and (b) 3 MPa the progress of infiltration can be observed showing different behaviour on preform types exhibiting no pore-forming additive.

No infiltration is obvious for the preform type without copper oxide and pore- forming additive added (AO) at 1 MPa gas pressure. Preform types with 10 wt% (A10) and 20 wt% (A20, AC100/20) pore-forming agent (PFA) added exhibit a certain progress of infiltration. Improvement of infiltration due to the addition of pore-forming additive has been expected – considering the expression of Laplace-Young, where the capillary back pressure, hence the infiltration pressure, is directly related to the pore diameter of the preform. No difference in infiltration progress is obvious between the addition of 10 wt% and 20 wt% PFA. By adding 10 wt% copper oxide to the preform containing 20 wt% PFA (AC100/20) the infiltration progress improves further showing almost complete infiltration. The influence of copper oxide to infiltration progress is shown at 3 MPa gas pressure for preform types without pore-forming additive. Whereas for the reference specimen (AO – no copper oxide) the infiltration has just started, samples containing 2 wt% (AC20S) and 5 wt% (AC50) copper oxide are obviously half infiltrated. The specimen labelled AA34 contained the corresponding amount of pre-reacted aluminate to 2 wt% copper oxide assuming full reaction. Nevertheless, no difference in infiltration progress is observed.

120

(a) Gas pressure of 1 MPa:

N6C20/3S N6C20/7S AC20/3S AC20S

(b) Gas pressure of 2 MPa:

N6C20/3S N6C20/7S AC20/3S AC20S

(c) Gas pressure of 5 MPa:

N6C20/3S N6C20/7S AC20/3S AC20S

Figure 66: Gas pressure infiltration in one step of 4 samples at different pressures (a) 1 MPa no infiltration is obvious, (b) 2 MPa the progress of infiltration dependent on preform type can be observed and (c) 5 MPa 2 preforms show double-sided infiltration (N6C20/3S and AC20/3S) and AC20S has not been infiltrated due to a gas bypass.

A different set of preform types in a cross-sectional view at pressures of 1 MPa, 2 MPa and 5 MPa is shown in Figure 66. This type of specimen contained the specific preform types AC20S (2wt Cu2O), AC20/3S (2 wt% Cu2O, 3 wt% cellulose), N6C20/3S (2 wt% Cu2O, 3 wt% cellulose) and N6C20/7S (2 wt%

Cu2O, 7 wt% cellulose) with ceramic volume contents of 51%, 48%, 60% and 53%, respectively. The 4 samples, infiltrated in one shot, showed different infiltration behaviour due to varied pore-forming additives or ceramic volume content and are displayed for comparison. No infiltration can be seen at pressures of 1 MPa, independent of preform types investigated in this set of preform types. This was expected from the results of the measurement of infiltration initiation pressure (Figure 60 and Figure 61), where the infiltration initiation pressures are all above 1 MPa. In contrast, infiltration at 2 MPa is obvious, at least for pore-forming additive containing preforms, consistent with initiation pressures measured (Figure 63). The infiltration saturation was dependent on the preform structure and an influence of pore-forming additive can easily be seen, as also shown in Figure 65 for copper oxide-free samples.

121

With high amounts of PFA in preform preparation the infiltration became easier. Although the ceramic volume content of the preform N6C20/3S with 60% was the highest, the infiltration of AC20S is worse. The content of pore-forming agent was the predominant factor for infiltration progress between these samples. This can be seen by comparing the samples containing 3 wt% pore- forming additive: whereas the sample AC20/3S exhibits a ceramic volume content of 48%, the sample N6C20/3S holds 60%, with the infiltration progress very slightly smaller. The sample N6C20/7S, with a ceramic volume content of 53%, shows an infiltration front progressed nearly twice as far in comparison with the AC20/3S, (with an even lower ceramic volume content) in consequence of the higher content of pore-forming agent (7 wt% cellulose) added in preform preparation. Considering the infiltration fronts of the samples N6C20/3S and N6C20/7S, the model of slug flow infiltration might be applicable. The infiltration fronts seemed to be flat, as the model predicted. The smooth crossover between those two samples was a result of a better infiltration ability of one preform type due to higher porosity, leading to an additional horizontal infiltration of the neighbouring preform. Taking this into account, both infiltration fronts were still straight. Consideration of the preform types AC20/3S and AC20S showed a somewhat more diffuse behaviour at the infiltration front. In the latter preform the infiltration progress had just started, showing no distinct infiltration front but single areas to be infiltrated. Taking the appearance of the load-displacement curves of infiltration into account led to the observation that the preforms being infiltrated followed the model of infiltration in a gradual manner. For evaluation of this observation microscopic investigations were necessary (as will be shown). Figure 66 shows also the cross-section of samples after an infiltration with 5 MPa: in two preforms a double-sided infiltration took place, whereas in another the effect of a gas bypass during infiltration is shown. The macroscopic appearance of the sample with a high amount of pore-forming additive looks like fully-infiltrated. The sample AC20S could not be infiltrated due to a gas bypass during infiltration. A gas bypass in the constant pressure method led to balancing the pressure drop forcing infiltration, resulting in the same pressure within the preform as applied to the molten metal. The pressure drop between the melt and preform is a prerequisite

122

for infiltration, so no infiltration could be expected. In contrast with the experiments at 2 MPa the infiltration was double-sided, as can easily be seen in the preforms N6C20/3S and AC20/3S; so a direct comparison of infiltration progress with the specimen seen above is not relevant. But, again, the infiltration progress in the samples containing 3 wt% pore-forming additive, despite the ceramic volume content of the preform, was roughly the same. In double-sided infiltration the air contained in the preform is entrapped. Because of the applied pressure, infiltration progresses as long as the pressure of the entrapped and compressing air equals the pressure drop at the infiltration front – if no solidification occurs. Therefore the pressure drop and, hence, the infiltration pressure at the infiltration front decreases with further infiltration. This leads to a separation of infiltrateable pore diameters, again referring to the Laplace-Young equation, as can also be seen at the more diffuse infiltration fronts.

Figure 67: Optical microscopy of N6C20/7S specimen (bimodal pore size distribution) infiltrated at 1 MPa: the big pores resulting from the pore-forming agent started to saturate at the infiltration front.

Further investigation of the samples with 1 MPa applied is shown in Figure 67. The sample N6C20/7S, which was found to be that best infiltrated at 2 MPa, pointed to some initial infiltration at 1 MPa when investigated microscopically.

123

The large pores, replicating the pore-forming agent, were filled in the closest contact region to the melt. This behaviour could only be observed with this specimen presenting the highest amount of pore-forming agent – 7 wt% for the four samples. As only the large pores were filled, the assumption of infiltration in a gradual manner became even more significant. The big pores, but not the surrounding finer porosity, were infiltrated, as can easily be seen in the microscopic image. The infiltration progress of specimens is shown by investigation of the same region of each sample. Figure 68 shows the microscopic image of the specimen AC20S for infiltration at 2 MPa and Figure 69 for infiltration at 4 MPa. The area investigated is on the top side of the specimen near the middle, and marked in Figure 67. This area has been chosen as the hottest spot is in the middle of the cavity with no edge influence affecting infiltration. Images were obtained with the MosaiX-algorithm as outlined previously.

100 μm

Figure 68: MosaiX-image of an AC20S specimen (alumina preform containing 2 wt% Cu2O) infiltrated at 2 MPa, at 200x magnification in optical microscopy: the bright phase is copper; the dark phase represents the non-infiltrated area (preform and pores).

124

100 μm

Figure 69: MosaiX-image of an AC20S specimen (alumina preform containing 2 wt% Cu2O) infiltrated at 4 MPa at 200x magnification in optical microscopy: the bright phase is copper; the dark phase represents the non-infiltrated area (preform and pores).

At 2 MPa a small amount of infiltration filled some pores in the sample containing 2 wt% Cu2O and no pore-forming additive (AC20S, Figure 68). Some pores on the upper side of specimen were filled, and single pore channels had started to be infiltrated. At 4 MPa the infiltration progressed much further (Figure 69): whereas the main part of the investigated area still had some infiltration defects, the upper part of the sample was completely filled. The same behaviour – better infiltration with higher pressure – is seen for the other specimen types in Figure 70.

125

2MPa gas pressure 4 MPa gas pressure

O / 3/ wt% PFA 2 50% CVC 50%

100 μm 100 μm

2 wt% Cu 2 wt%

O / 3/ wt% PFA 2 60% CVC 60%

100 μm 100 μm

2 wt% Cu 2 wt%

wt% PFA / wt% PFA O / 7 2 50% CVC 50%

100 μm 100 μm

Cu 2 wt%

Figure 70: MosaiX-images of specimen AC20/3S (top), N6C20/3S (middle) and N6C20/7S (bottom) infiltrated at 2 MPa (left column) and 4 MPa (right column) at 200x magnification in optical microscopy: the bright phase is copper; the dark represents ceramic and pores.

The influence of pressure increase for this pressure drop was not as impressive as shown on the specimen without the pore-forming agent, due to the further progressed infiltration at low pressure; but it shows the powerful role of the pore-forming agent in preform infiltration. Infiltration improved with higher infiltration pressures, beginning from the top of specimen; and was strongly influenced by the content of pore-forming agent in the preform – the fraction of

126

large pores resulting from pyrolisation of the cellulose, acting as infiltration channels. These infiltration channels were filled first to infiltrate the smaller pores in the resulting microstructure from medium to finer pores, as can easily be seen in the figures above.

The infiltration initiation pressures of preforms containing pore-forming agent were evaluated by squeeze cast infiltration experiments to be 1.74 MPa for 3 wt% addition and 1.65 MPa for 7 wt% additions, respectively. These preforms contained additionally 2 wt% Cu2O. The addition of pore-forming agent in preform preparation leads to a bimodal pore structure in the resulting preforms. This contributes to the investigations done into constant pressure infiltration. Whereas at 1 MPa no infiltration was obvious, infiltration could be observed at 2 MPa, as shown in Figure 66. The initiation of infiltration at 1 MPa investigated in microscopic images of the specimen with 7 wt% cellulose added could be attributed to a capillary pressure of 0.94 MPa. This was evaluated by averaging the pore size measured in Hg-porosimetry in capillary law (equation 17). The initiation pressures evaluated of constant pressure infiltration were between 2 MPa and 3 MPa for preforms without the pore-forming agent; and between 1 MPa and 2 MPa for a preform containing PFA, which corresponds with the threshold pressures measured in the constant flux method shown in Figure 63.

In conclusion, the behaviour of preform saturation shows the process of gradual infiltration is appropriate for describing the infiltration, consistent with the investigations of other authors59, 82. A bimodal pore structure promotes the infiltration and the amount of pore-forming additive heavily influences the infiltration behaviour - more than the ceramic volume content. Copper oxide additions improve the saturation progress, but rarely affect infiltration initiation. The influence of copper oxide on preform infiltration microscopically is shown in the following section.

127

5.3 Microscopic Investigation 5.3.1 Structure of Metal Matrix Composite The microstructure and microscopic saturation of infiltration of metal matrix composites was characterised by microscopic investigation. Using optical microscopy the overall phase appearance of MMC is shown and described in

Figure 71: a composite of a preform containing 2 wt% Cu2O with a monomodal pore size distribution (no added pore-forming agent) was chosen to introduce the phase appearance.

Figure 71: Optical microscopy image of a Cu-infiltrated preform containing 2 wt% Cu2O and no pore-forming agent, having a ceramic volume content of 52%: the brown particles are alumina, the big grey agglomerates are of aluminate and the yellow/orange is copper.

The polished sample shows a typical example of a composite manufactured through infiltration of a copper oxide containing preform with copper. The copper oxide addition occurred after sintering and infiltration in relatively big agglomerates of aluminate: these agglomerates appeared grey in the microstructure. The small brownish particles were of alumina embedded in the copper matrix. The sample shown above, exhibits a ceramic volume content of

128

52%. The fine microstructure of the composite is obvious, showing clearly the big size difference when comparing the alumina particles and the aluminate agglomerates: the agglomerates were much bigger than the raw particle size of copper oxide (97% < 5μm and shown in Figure 18). Preforms prepared without copper oxide did not show these agglomerates, as was expected.

A composite manufactured by infiltration of a preform prepared with pre- reacted aluminate powder is shown in Figure 72(a). The amount of aluminate powder added equals the same amount of aluminate in the specimen shown in Figure 71, the prerequisite of a full reaction of the embedded copper oxide. In Figure 72(b) a composite structure of a preform with the addition of 5 wt% pore- forming agent is shown: the preform exhibits the expected bimodal pore structure due to the cellulose addition.

(a) (b) Pressing direction

20 μm 100 μm

Figure 72: Optical microscopic images of composite samples of two different specimen types: (a) is of a preform containing 3.4 wt% pre-reacted aluminate; (b) shows a specimen with bimodal pore size due to the addition of pore-forming agent (5 wt%).

The sample with pre-reacted aluminate (Figure 72(a)) did not show agglomerates. Taking into account the small particle size of pre-reacted aluminate with d50 = 0.9 μm, which is even smaller than the ceramic particles (d50 = 1.9μm), the homogeneous microstructure without agglomerate was expected. Comparing both types of preforms (Figure 71 and Figure 72(a)), the microstructures looked similar, but the agglomerate appearances differed. The ceramic volume content for both is 52%. The microstructure of the specimen prepared from the bimodal Nabaltec powder looks somewhat different: a typical appearance is shown in Figure 73, of a specimen containing 2 wt% Cu2O and 7 wt% pore-forming additive. The ceramic volume content of this type is 54%.

129

Pressing direction

20 μm

Figure 73: Microstructure of composite of a preform type from the bimodal Nabaltec powder with 2 wt% Cu2O and 7 wt% PFA (N6C20/7S) at 500x magnification in optical microscopy.

The microstructure is dominated by large pores resulting from the pore-forming additive. The pores of the preform in the surrounding structure were smaller to the samples prepared with the Almatis powder. The bright phase represents the copper, the dark brown particles the alumina and the grey agglomerates the aluminate phase. The microstructure of samples with bimodal pore structure looked quite textured – a result of the big cellulose particles orientating orthogonally to the pressing direction during green body pressing, due to their non-spherical shape. Therefore, the specimen above was pressed along the vertical direction. By adding more copper oxide to the preform, more aluminate agglomerates could be found in the microstructure. The same effect with pore- forming agent: the more cellulose the more big pores filled with copper, and the clearer the texture of microstructure, as a result of axial green body pressing.

As the imaging resolution of optical microscopy is limited and alumina’s transparency complicates investigation at higher magnifications, SEM was used for further investigation. The image acquisition was done in backscattered mode with copper and alumina providing good contrast. The typical appearance of a pore-forming additive-containing sample is given in Figure 74. Due to the axial

130

pressing process, the big cellulose particles aligned perpendicularly to the pressing direction, as mentioned, showing a preferred orientation horizontally in this image. Pressing direction

20 μm

Figure 74: Appearance of a copper-infiltrated sample containing the pore-forming agent (20 wt%): SEM image of a not fully infiltrated region, visible at the infiltration defects within the fine microstructure.

The images obtained in backscattered mode in the SEM show the phase contrasts due to different phase densities; the copper, with its high density of 8.92 g/cm³, is displayed by the bright phase, and alumina (ߩ = 3.98 g/cm³) as the dark grey particles. If the aluminate phase (ߩ = 4.89 g/cm³), appears within the sample, it is displayed by the grey values between. In the above image the textured microstructure of the pore-forming agent-containing samples can be observed. The infiltration of this specimen was not complete, as can easily be seen by the infiltration defects, expressed by the black colour, within the finer microstructure. In Figure 75 a sample is shown where the copper phase was removed with nitric acid after the mechanical grinding and polishing process.

131

(a) (b)

100 μm 2.5 μm

Figure 75: SEM images of polished MMC samples containing 2 wt% Cu2O in the preform with the copper phase removed by use of nitric acid. (a) shows the distribution of aluminate particles in a sample without PFA addition; (b) shows a detail of a sample with 5 wt% cellulose added.

Alumina and aluminate remained in the microstructure. The distribution of the aluminate particles can be seen more clearly without the copper phase. There was good adhesion – a direct crossover between aluminate and alumina particles which can easily be seen, as there is no gap between those two phases. Other than the agglomerates of aluminate, no obvious layer around the alumina particles can be observed. But beside the big agglomerates of aluminate small amounts were found, well distributed within the microstructure, as can be seen in Figure 76. This effect can also be observed in preform structures before infiltration, as has been seen in Figure 48, where a preform with 2 wt% copper oxide is shown. Small bright dots indicate the aluminate phase, as the image was taken in backscattered mode.

1 μm 1 μm

Figure 76: Small amounts of aluminate well distributed in the microstructure, other than the big agglomerates shown in the figure above. The copper on the polished composite samples containing 2 wt% Cu2O was removed by nitric acid. SEM imaging was performed in backscattered mode.

132

So aluminate appeared in large agglomerates and was well distributed in small amounts in the microstructure. For higher resolution microscopy and investigation of interface appearance, TEM samples were prepared, as shown in section 5.3.2. Further investigation of microstructure with the focus on copper/alumina bonding was performed in the SEM. In addition to the mechanical grinding and polishing process an ion milling step was applied to reduce artefacts, such as breakage of the ceramic particle edges or smearing of the copper phase. Vibration polishing was also applied, to reduce preparation artefacts from the mechanical procedure. Through horizontal vibrations and low loads, the finish of polishing is gentle but effective, and shows results similar to the ion milling process. In Figure 77, microstructural images are shown for comparison of both finishing processes: both can be forced to polish selectively by higher abrasive rates for copper than for alumina.

Vibration polished Ion beam milled

1 μm 1 μm

Figure 77: Comparison of finishing processes for vibration polishing and ion milling, by samples containing 5 wt% Cu2O: both surfaces reveal the fine microstructure of composites. SEM imaging was performed in backscattered mode.

The effect of ion polishing on the microstructural appearance is shown in Figure 78 in comparison with the normal grinding and polishing process.

133

(a) (b)

1 μm 500 nm

(c)

1 μm

Figure 78: SEM images in backscattered mode of polished samples: (a) and (c) show MMC of pure alumina preform and (b) of alumina preform with 2 wt% Cu2O added. (c) is exactly the same specimen as (a) but with additional surface treatment by ion milling.

The MMC sample in Figure 78(a) is of a pure alumina preform mounted in and appears blurry after the mechanical grinding and polishing process; edges cannot be clearly detected and some breakage of the ceramic phase can be seen. A somewhat clearer image of a copper oxide (2 wt%) containing- preform MMC, again mounted inepoxy, at higher magnification is displayed in Figure 78(b). In both images artefacts from the polishing process, such as breakage of the ceramic phase and copper smearing, can be observed. In Figure 78(c) the same specimen as shown in Figure 78(a) is displayed, after an additional ion milling process, which is seen to be a suitable method for final polishing of samples containing phases of different hardness – e.g.,

134

composites205-207. The ion milled surface highlights some infiltration defects: fine gaps and small edges were not filled with copper and some intragranular porosity in alumina particles becomes visible. In many edges where alumina particles are scarcely sintered, and between narrow platelets of alumina, no copper was found.

1 μm

Figure 79: SEM image in backscattered mode of mechanically polished and additionally ion milled sample of MMC of a preform with 5 wt% copper oxide added.

In Figure 79, a sample containing 5 wt% copper oxide in the preform prepared by the same ion milling process is shown. Microstructural defects concerning microscopically incomplete infiltration are obviously lower. The intragranular porosity is of the raw material and thus not influenced by copper oxide addition or any other infiltration-improving effect; but less infiltration defects around the corners of sintered particles and platelets are observed. No infiltration defect is detected next to aluminate particles; although very fine needles can be observed. In Figure 80, images of MMC samples of pure alumina preform (Figure 80(a)) and a preform with 5 wt% copper oxide added (Figure 80(b)) at lower magnification are shown.

135

(a)

5 μm

(b)

5 μm

Figure 80: SEM images obtained in backscattered mode of polished and additionally ion- milled samples of MMC containing pure alumina preform (a) and 5 wt% Cu2O addition (b) at 5,000x magnification – apparent infiltration defects marked.

136

Quantitative microstructural analysis of the microsporosity of both samples showed an influence of copper oxide addition: whereas the specimen without the copper oxide addition revealed a microporosity of about 0.09 area%, the specimen with 5 wt% copper oxide addition showed 0.02 area%. For both measurements about 2900 μm2 of polished composite surface were taken into account. In direct comparison, the influence of copper oxide addition to the preform in terms of fewer infiltration defects can be seen in the micrographs above (Figure 78 and Figure 79): even at low magnification the infiltration defects in the sample without copper oxide addition are visible (Figure 80). The frequency of defects found is high, while the sample with added copper oxide shows fewer infiltration defects. As the intragranular porosity of particles can still be observed, infiltration defects can barely be identified.

The copper oxide addition, therefore, improves the infiltration behaviour not only macroscopically, as shown before in section 5.2, but also microscopically. As stated, the improvement of infiltration relies only on the addition of copper oxide in the preform preparation process, the preform structure being the same. The preforms were sintered before infiltration; and copper oxide does not react to aluminate during infiltration, leading to a lowering of interfacial energy or mass transfer at the interface, as reported by Chaklader et al.27 and Aksay et al.41 – resulting in a lowering of the wetting angle to improve infiltration. So an improvement in infiltration ability cannot be explained by the presence of the aluminate phase, even though Diemer5 shows in his study no obvious difference between the wetting angles of copper on aluminate to copper on alumina. Nevertheless he states the wetting angle on aluminate to be time-dependent due to the instability of aluminate at low oxygen partial pressures and the solubility of atmospheric oxygen in the copper melt at higher partial pressures, which promotes the wetting.

137

(a) (b)

500 nm 500 nm

(c)

500 nm

Figure 81: Aluminate agglomerates at the polished surface of an ion-milled sample: the specimen is a composite of a preform containing 5 wt Cu2O. SEM images are obtained in backscattered mode at 50.000x magnification.

In Figure 81 detailed images of aluminate agglomerates in a specimen containing initially 5 wt% copper oxide are shown: the final preparation of the sample surface was performed by ion milling. The aluminate agglomerates had a different appearance. The aluminate in Figure 81(a) looked as per the mechanically-polished samples: the surface was even and there was a single phase appearance. As well, single parts had built finely-structured lamellae: in Figure 81(b) this behaviour becomes more and more dominant, until in Figure 81(c) none of the original aluminate surface could be seen. As well as normal alumina particles in this image there were lamellae of what was detected at lower magnification as aluminate. The lamellae resembled reacted alumina particles with specific grain boundaries remaining; the structure phases of the

138

eutectic reaction. Within the fine lamellae, copper could be found. Looking more closely at Figure 81(b) showed a darker phase at the edges of the lamellae and around the agglomerate of the forming aluminate. The appearance of this phase looked similar to the alumina particles. Beraud et al.141 investigate the formation of aluminate at the interface in their study of copper/alumina bonding. They report the dissolution of the aluminate phase and formation of alumina without any crystallographic relationship with the bulk material where there is insufficient oxygen content in the surrounding copper.

(a) (b)

175 μm 1 μm

Figure 82: Microstructure of the specimen prepared with the bimodal powder from Nabaltec with 3 wt% of cellulose and 2 wt% of Cu2O: surface finishing by ion milling before investigation. SEM imaging was performed in backscattered mode.

Figure 82 shows the microstructure of a composite sample of type N6C20/3S

(Nabaltec powder, 2 wt% Cu2O, 3 wt% cellulose), with its specific structure, prepared in an ion milling process. The microstructure of the preform type is shown at low magnification in Figure 82(a): the bimodal pore structure is obvious and the microstructure appears to be in layers around the large pores – this appearance is typical of preforms prepared with the bimodal alumina powder from Nabaltec. As well as these large pores resulting from the pore- forming agent, the microstructure exhibits fine porosity. The ceramic volume content of this specimen is at 60%. In Figure 82(b), an aluminate agglomerate is shown. While smaller particles of the aluminate appear in the lamellae structure or in fine particulates embedded in the copper matrix, large particles are only partially evolved in this process beginning at the edge of particle.

139

10 μm

Figure 83: Typical microstructure of a composite sample, infiltrated by copper in isothermal gas pressure infiltration at 1200°C, of a preform with 2 wt% copper oxide added. No aluminate agglomerate can be seen in the microstructure. Imaging performed in SEM in backscattered mode: alumina appears dark, copper bright.

Figure 83 shows the microstructure of a composite sample from a preform with 2 wt% copper oxide added, manufactured by isothermal gas pressure infiltration. The infiltration was performed with a commercial gas pressure sintering furnace at 1200°C and 10 MPa applied argon gas pressure. After infiltration, no aluminate agglomerate is apparent in the resulting microstructure; the aluminate agglomerates appear to be dissolved during infiltration. Other isothermal infiltrations are performed and described by Reymann215, all showing the same dissolution of aluminate agglomerates. This is in agreement with the observations of Beraud et al.141 who report the dissolution of the aluminate phase if there is insufficient oxygen in the copper matrix. Here, melting of oxygen-free copper was performed in vacuum/argon atmosphere in a furnace with graphite environment; hence this state can be assumed.

140

5.3.2 Interface Appearance Model interface characterisation was established with the use of freeze cast preforms: some of these specimens were coated with copper and subsequently heat-treated before infiltration. The infiltrated samples were characterised on cross-sectional polished surfaces. The appearance is illustrated in Figure 84 – a specimen infiltrated with an aluminium alloy (AlSi12) for better detection and contrasting of the copper layer. As described, the copper layer reacted during heat treatment to aluminate.

Al2O3 AlSi12

CuAlO2-layer

40 μm

Figure 84: Polished cross-section of a copper-coated freeze-cast preform infiltrated by an aluminium-silicon alloy: image obtained in backscattered mode in SEM. Alumina appears dark.

The copper-containing layer can be seen, as the image was taken in backscattered mode in SEM. The bright thin phase between the alumina (the dark grey phase) and the aluminium alloy matrix is that containing copper. As the preform had to be preheated, the Cu layer was reacted before infiltration. In Figure 85 a copper-coated freeze-cast sample infiltrated with copper is displayed: ascertaining layer detection is much more demanding due to the low thickness of the layer and lower density contrast. These samples were used for investigating a model of interface strength with manipulated interfacial phase: for comparison, samples without coating were used. 141

Al2O3 Cu

CuAlO2

2,5 μm

Figure 85: Cross-section of copper-coated freeze-cast preform infiltrated with copper: enlarged detail shows the layer being coated and thermally treated before infiltration. Imaging was performed in SEM in backscattered mode.

The interfacial appearance of different selected samples was analysed by TEM investigation: the phases look different in TEM when compared with those from SEM.

Al2O3 Copper CuAlO2

15 μm

Figure 86: A sample used for TEM investigation of a composite containing 20 wt% Cu2O in preform with the phases for investigation indicated. Imaging was performed in TEM.

142

In Figure 86 a sample used for TEM investigation is shown in low magnification with the phases indicated. In contrast with the images obtained in SEM, the copper phase appears dark and the alumina bright. For investigation of the interface appearance, MMC of preforms with a high amount of copper oxide were used, for easy phase detection. The sample shown and investigated contains 20 wt% of copper oxide in the preform (AC200). Figure 87 provides a closer look at one interface:

aluminate (a) copper (b) copper

alumina

alumina 50 nm

Figure 87: Detailed TEM image of a triple point of the three phases – copper, alumina and aluminate (a). In (b) the changeover of alumina to copper is shown.

No obvious phase between the alumina particle and the copper was visible: the interface was smooth, and no gap could be observed. The crossover from the alumina particle to the copper matrix seemed to have a good bonding nature at the direct interface, without visible gaps or defects. The interface between the alumina particle and copper close to a triple point (Figure 87(b)) appeared different at higher magnification: possibly a different phase, extending from the aluminate, was established between the copper and the alumina, with smooth bonding between both. The gap between the aluminate and the alumina particle shown in this figure does not reveal the overall behaviour of this sample: at no other interface could such a gap be observed between those two phases.

143

200 nm 41.0 kx

Figure 88: Interface appearance of a sample without copper oxide addition, investigated by TEM.

In Figure 88 a sample without addition of copper oxide is shown. In contrast to the specimen with copper oxide addition, the interface of alumina and copper showed defects. The nature of the adhesion of copper to the alumina particle was not good, as can be seen at higher magnifications. The gap was quite narrow and revealed poorer bonding behaviour than the specimen with copper oxide addition (Figure 87). As the content of Cu2O was very high and this behaviour was found next to the aluminate particle, another investigation was performed on a sample with copper oxide addition in a technically-relevant amount, and not in direct relation to an aluminate particle. This is shown in Figure 89.

144

(a) (b)

200 nm 200 nm 57.0 kx 73.0 kx

Figure 89: A composite sample with 5 wt% copper oxide addition to the preform, obtained by TEM.

Again the bonding seemed of a better nature. No delamination, as found in the specimen of pure alumina could be detected: even narrow gaps between alumina particles were fully infiltrated. A different bonding nature was found as illustrated in the enlarged detail in Figure 89(b): direct bonding between alumina and copper in combination with small bonding defects is found. The structure of the bonding appeared similar to the area close to the triple point seen in Figure 87 of the specimen AC200, next to the aluminate agglomerate, but with defects.

Further work was done to investigate the lamellar appearance of the aluminate phase. Not only large agglomerates of the former aluminate phase show that structure but also the aforementioned small amounts of aluminate found in the microstructure, as illustrated in Figure 76. This could be seen in high resolution SEM images of a composite sample containing 2 wt% Cu2O prepared by ion milling, shown in Figure 90, where the reaction is incomplete.

145

Al2O3 Cu

CuAlO2

250 nm 250 nm

Figure 90: SEM image showing alumina with small amounts of aluminate in close contact with a composite sample (2 wt% Cu2O) prepared by ion milling. The aluminate phase decomposes during infiltration, remaining a lamellar structure.

The copper oxide reacted with the alumina particles, where in contact, to result in aluminate. Within this reaction the aluminate structure grew into the alumina particles, as can be seen in the images. During infiltration the aluminate phase seemed to further react. To ensure the shown lamellar structure was not preparation-induced, cross-sections were cut using FIB milling. In Figure 91 an image obtained in an STEM is shown (performed at Zeiss, Oberkochen, Germany, with an AURIGA® System).

Al2O3

40 nm

Cu

200 nm

Figure 91: STEM image of the cross-section of decomposed aluminate, appearing in the lamellar structure found in the composite.

The cross-section of decomposed aluminate confirmed the structure. Combining both images, the structure was lamellar in 3-D expansion. Close 146

contact between both phases could be observed. The laminate structures appearing to be copper and alumina accords well with the observations made by Beraud et al.141, who find that aluminate can decompose into copper and alumina if oxygen is absent.

5.4 Investigation of Interface Strength Urena et al.216 propose a technique for characterising interfacial strength by nano-indentation using a Berkovich indenter: one corner of the indentation mark made by the indenter must be in contact with the interface. They investigate the bond strength between coated carbon fibres and the aluminium alloy matrix. Characterisation was performed by single indents in the interface proximity to evaluate the fracture tendency. Here, coated and uncoated freeze-cast preforms infiltrated with copper were investigated in the same manner. Figure 92 shows a row of Berkovich indents of varying proximity from the interface on a freeze cast preform in optical microscopy.

Figure 92: Lamellae of a freeze-cast preform in Cu-MMC with a row of Berkovich indents directed from one corner towards the interface at decreasing distances from it (OM).

Even at the maximum load, regardless of the distance between indent and inter- face, no debonding load or pop-in effect in the nanoindentation load-depth curves could be observed – no cracks could be found in composite structure after testing: so the load was increased and testing performed in a microhardness indentor with a Vickers indenter. For this, indentation was done with the diagonal of the indent along the interface of metal/ceramic structure in order to wedge the two phases for crack initiation – this is described as the interfacial indentation test. Several authors report on this test and evolve models for determining the critical load, bonding strength or interfacial toughness to characterise plasma-sprayed coatings217-219. Marot et al.220

147

compare different tests for determining the interfacial adhesion and conclude that the interfacial indentation test is the most universal. The principle of this test is to make several indentation indents at an interface to create a crack. In Figure 93, two different indents at the interface, using a load of 2.9 N, are shown in optical and electron microscopy, each for a composite of a copper- coated and uncoated freeze-cast preform.

(a) (b)

20 μm 20 μm

(c) (d)

20 μm 20 μm

Figure 93: Interfacial Vickers indentation (load of 2.9 N) in Cu-infiltrated freeze-cast preforms: in (a) and (b) there is no coating whereas in (c) and (d) there is coating with copper before infiltration. Imaging performed in OM – (a) and (c) and SEM – (b) and (d).

In Figure 93(a) and (b) a crack can be observed: the preform for this MMC was not coated and no heat treatment was done after infiltration. Figure 93(c) and (d) show no crack after indentation was performed: the preform was coated with copper before infiltration took place. For evaluation of its presence, the interface layer is still observable in the SEM image (Figure 93(d)). Several indents show the same behaviour as that in Figure 93: obviously, the interface is stronger as an interfacial layer is present. This accords with the work of many researchers5, 27, 141, 159, who investigate the interfacial behaviour of copper and alumina where aluminate is present at the interface.

148

5.5 Characterisation of Phase Properties by Nanoindentation Testing The properties of each phase were evaluated using the nanoindentation technique, requiring the use of different specimens: first, one for reference without copper oxide; second, one with a 2 wt% Cu2O addition infiltrated with copper in a squeeze casting process. Additionally, a copper reference sample exhibiting the same thermal and casting treatments as the infiltrated samples was investigated. Each phase was characterised by at least 14 valid indents and a resulting average is given. Validity of indents was evaluated after testing by investigation of the surrounding microstructure: indents in contact with another phase or on rounded edges of particles for example were excluded. In Figure 94, load-depth curves of some copper indents are shown:

400

300 μ N] 200 Load [ 100

0 0 1020304050 Depth [nm]

Figure 94: Load displacement curves of indents in the copper matrix of a specimen containing 2 wt% Cu2O.

Since copper is ductile and the nanoindentation technique evaluates the reduced elastic modulus Er from the unloading curve, creep effects can influence the evaluation. Therefore a holding time of 30 s at maximum load before unloading was applied: this resulted in an increase of depth at constant load due to creep in the copper. Overlaid on the load-depth curves is the thermal drift of the set-up: the combination of creep and thermal drift can be seen in the load-depth curve in the diagram. For calculation the thermal drift has been automatically corrected for evaluation of hardness and reduced elastic modulus. In Figure 95, a 3-D image created from scanning probe microscopy

149

(SPM) of the area with indents in the copper phase is shown. Each indent is marked with a small banner.

Figure 95: 3-D image obtained from the topographic data during in situ imaging of a selected area of MMC with indents of a Berkovich indenter in the copper matrix, used for characterisation of the discrete phases.

The relief of the surface can be seen. Due to the strong difference in hardness between copper and alumina the nanoscaled microstructure is rough. The alumina particles are quite flat on the surface but rounded at the edges, and somewhat higher than the copper matrix. The copper phase is preferentially removed during the mechanical preparation process. The indents in the copper phase replicate the three-sided/pyramidal indenter tip well: the indentation marks show no recognisable pile-up or sink-in effect in the copper matrix. The hardness and reduced elastic modulus Er of the copper phase is shown in Table 11:

Table 11: Result of the evaluation of the reduced elastic modulus Er and hardness of the copper phase in MMC of 2 wt% copper oxide containing preform: calculation based on 18 valid indents.

Average Er [GPa] Std. dev. Average HV Std. dev.

221 26 308 54

150

The reduced elastic modulus of 221 GPa was much higher than expected. In investigation of the copper phase detached from the composite structure more extensively it was found, that with higher applied loads Er reduces to about 129 GPa at 2 mN load – which is close to the specifications given in Table 5. This behaviour could be generally attributed to the indentation size effect of testing, where apparent hardness is direct related to the force used for testing. With higher applied load, and hence indentation size, the apparent hardness reduces221. Nevertheless, investigation into hardness of copper in composite samples was performed with 400 μN due to the fine microstructure, as hardness measurements are affected by phases lying underneath. Since the indents have to be deeper than 40 nm212 the tests could not be performed with the same load for all three: alumina is much harder than copper, so an indent with 400 μN would not affect an indent with a valid depth. The tests in the alumina and aluminate phases were therefore performed with 2 mN. In Figure 96 some load- depth curves of alumina and aluminate are shown:

Alumina Aluminate 2000

1500 μ N] 1000 Load [

500

0 0 102030405060708090 Depth [nm]

Figure 96: Examples of load-depth curves of selected indents in alumina and aluminate of a specimen with 2 wt% Cu2O – pop-in effects marked.

Some of the load-depth curves for alumina show pop-in effects: in this diagram one curve with and one without pop-in is seen. Apart from this, all curves of the aluminate indents show pop-in effects. Some indents of alumina and aluminate are shown in Figure 97 and Figure 98.

151

Figure 97: 3-D image of an area of MMC with Berkovich indents in the alumina phase, used for characterisation of the discrete phases: each indent is marked with a banner.

Indentation in the alumina phase is affected by the rounding of the alumina particles. The uneven alumina surface and microstructure is a result of the mechanically grinding and polishing process. Due to the appearance of the micro-structured surface not every indent can be used for evaluation of the characteristic properties. Indents on a rounded edge, for example, are not considered in the calculation. The hardness and reduced elastic modulus Er of the alumina phase is shown in Table 12.

Table 12: Results of the evaluation of the reduced elastic modulus Er and hardness of the alumina phase in MMC of 2 wt% copper oxide containing preform: calculation based on 16 valid indents.

Average Er [GPa] Std. dev. Average HV Std. dev.

273 30 2342 345

The reduced elastic modulus is in the range of alumina ceramics of technical purity, which is 200 to 350 GPa222 and matches quite well the specification of 275 GPa (see Table 5) for the alumina used in this work. The values of hardness are relatively high. Very pure alumina results, with up to 2300 HV10; whereas alumina with technical purity is at 1200 to 2000 HV10 and the alumina

152

used for preform preparation 1500 HV0.5. Note: we must consider the indentation size effect of the hardness measurements, where the hardness increases with decreasing indentation loads. The wide range of standard deviation can be explained by micropores and microcracks within this phase, and the rounding of the alumina ligaments. Furthermore, the values vary strongly within the area of the MMC chosen for characterisation.

Figure 98: 3-D image of an area of MMC with Berkovich indents in the aluminate phase used for characterisation of the discrete phases: each indent is marked with a banner.

The microstructure of the aluminate phase in the 3-D image is difficult to separate from the copper phase. Alumina particles can easily be seen and identified. For clear identification of the aluminate phase a gradient scanning probe microscopy (SPM) image, as shown in Figure 99(a), is used.

153

(a) (b)

5 μm 5 μm

Figure 99: Gradient scanning probe microscopy (SPM) images of two areas used for discrete phase characterisation with (a) indents in the aluminate phase and (b) indents in the alumina phase: red-marked indents excluded from evaluation.

In the gradient SPM images the aluminate phase can be identified because of its similar appearance in other microscopy images. Indents marked green were valid and used for evaluation; red-marked indents were invalid due to edge effects of particles, uncertainty regarding subsurface material, e.g., so unusable and therefore excluded from calculation. The hardness and reduced elastic modulus Er of the aluminate phase is shown in Table 13:

Table 13: Result of the evaluation of the reduced elastic modulus Er and hardness of the aluminate phase in MMC of 2 wt% copper oxide containing preform. Calculation based on 14 valid indents.

Average Er [GPa] Std. dev. Average HV Std. dev.

192 19 1207 117

The aluminate phase seemed brittle as all loading curves showed pop-in effects; and by use of higher magnifications cracks around the indentation marks became visible. The first breakage for most indentations occurred at a load of around 1000 μN as can be seen in Figure 96. Due to the insufficient resolution of the SPM images at high magnification, no measurement of radial crack length was possible for fracture toughness evaluation. Furthermore, the indentation marks showed a pile-up effect, more typical for ductile materials such as metals. 154

Figure 100 is an overview of hardness and reduced elastic modulus of indents of the specimen with 2 wt% Cu2O added.

350 3500 Reduced elastic modulus 300 3000 Hardness 250 2500

200 2000

150 1500 Hardness 100 1000

50 500 reduced elastic modulus [GPa] 0 0 0,5 1,5 2,5 3,5 Copper Alumina Aluminate Figure 100: Reduced elastic modulus (blue rings) and hardness (grey triangles) of the discrete phases copper, alumina and aluminate in composite structure: all values are obtained from a specimen containing 2 wt% Cu2O.

Alumina presented the highest reduced elastic modulus (blue) of the three phases, as expected. Aluminate had a somewhat smaller elastic modulus than copper Alumina exhibited the highest values for hardness, copper exhibited a softer behaviour. The hardness of aluminate was roughly between the main phases. In considering the different specimens investigated, the characteristic of the alumina phase varied marginally, while the copper phase was strongly influenced by other factors: the copper phase in MMC without copper oxide added in preform preparation exhibited lower hardness and elastic modulus. Values in the copper reference sample presenting the same thermal treatment were even lower. This behaviour for non-reinforced material was expected, as metals react sensitively to mechanisms such as residual stresses and circumstances within the microstructure. For comparison, therefore, the values given are all for MMC made of copper oxide-containing preforms.

Measurements of the copper phase showed different behaviour depending on the preform chemistry. The increase in hardness of the copper phase of a pure copper sample compared with a composite sample could easily be attributed to the different microstructures. Adding copper oxide to the preform increased the hardness and reduced elastic modulus of the copper

155

phase in comparison with the copper oxide-free preform. As stated, no difference in pore structure or ceramic volume content of these preforms were present, which could also have affected the measurements. Generally, increasing the hardness of a metal by alloying it with different elements is a common method. As has already been stated, aluminate can dissolve into alumina and copper oxide at low oxygen partial pressures141. Taking into account the enhanced infiltration of copper oxide-containing preform compositions, alloying of the melt with oxygen became a possible effect: the dissolved oxygen could have led to greater hardness of the copper phase223. The residual alumina of the dissolved aluminate agglomerates was shown in section 5.3.1.

5.6 Physical Properties of Metal Matrix Composites 5.6.1 Strength Figure 101 shows the bending strength of specimens manufactured from preforms of different copper oxide contents but the same raw alumina powder. The ceramic volume content for all type of preforms is 52% ± 2%. The infiltration of pure alumina preforms without cellulose addition is difficult. Nevertheless, these samples were tested, indicating a median strength of 363 MPa. However, samples with apparent infiltration defects were not considered for median calculation. The evaluation of this preform type according to the maximum likelihood method was not appropriate, as samples without infiltration defects were rare – in fact, only 4 specimens were taken for evaluation. However, samples containing copper oxide showed improved infiltration behaviour, and at least 12 specimens of each sample type were taken for evaluation of strength – except for 0.5 wt% Cu2O addition, where evaluation was based on 9 specimens. Due to the limited number of specimens evaluated, especially the Weibull modulus is afflicted with uncertainty regarding to Khalili and Kromp224 who propose the evaluation of at least 30 specimens to be required for good characterisation. This is only applicable to the sample with 2 wt% copper oxide added.

156

1000 50 Bending strength Weibull modulus m 800 40

600 30

400 20 Weibull modulus Weibull

200 10 Bending strength [N/mm²]

0 0 0123456

Content of Cu2O in preform [wt%]

Figure 101: Bending strength (blue diamonds) and Weibull modulus (grey triangles) of composite samples as a function of the addition of copper oxide in preform preparation process (error bars showing the 90% confidence interval). The strength of the reference sample is given approximately (average value), as an evaluation of strength, and Weibull modulus is not appropriate because of apparent infiltration defects.

The bending strength of the composite material was enhanced significantly with the addition of copper oxide to the preform. The strength of the composite of pure alumina preform was given as about 360 MPa; with additions of 0.5 wt% up to 2 wt% copper oxide the bending strength doubled to roughly 800 MPa, with a Weibull modulus of 17 for 0.5 wt% Cu2O and 25 for 1 wt% Cu2O. Further addition of copper oxide resulted in a decrease of bending strength to 760 MPa at 3 wt%, with a Weibull modulus of about 19 and 660 MPa at 5 wt% with a Weibull modulus of 7. Not shown in this figure, but evaluated, was the continuing decrease of bending strength with higher copper oxide content: specimens with 30 wt% addition showed a bending strength as low as 497 MPa, with a Weibull modulus of 4.8. The influence of pore-forming additive to bending strength and Weibull modulus was investigated on samples containing 2 wt% copper oxide; reference samples without copper oxide additions were also included. Figure 102 shows the bending strength and Weibull modulus of samples with regard to their content of pore-forming

157

additive, divided into two parts – copper oxide-containing (2 wt%) and copper oxide-free samples.

1000 50 Bending strength Weibull modulus m 800 40

600 30

2wt% Cu2O 400 20

0wt% Cu2O Weibull modulus Weibull

Bending strength [N/mm²] 200 10

0 0 0246810 Content of pore forming agent [wt%]

Figure 102: Bending strength (blue diamonds) and Weibull modulus (grey triangles) of composite samples depending on the pore-forming addition in preform preparation: two sets of samples– copper oxide-containing (2 wt%) and copper oxide-free – are shown.

Pore-forming agent slightly affected bending strength. Looking first at the samples with 2 wt% copper oxide added, adding 3 wt% of cellulose to the preform (AC20/3S) slightly reduced the strength to 727 MPa with a Weibull modulus of 23. A further increase of cellulose to 7 wt% (AC20/7S) increased the Weibull modulus up to 29.9, while the bending strength stayed at 725 MPa; an addition of pore-forming agent, therefore, slightly reduces bending strength. Looking at the samples without copper oxide added, reduction of bending strength due to pore-forming addition was still exhibited: the addition of 10 wt% cellulose resulted in a reduction of bending strength from 360 MPa to 260 MPa. In terms of the infiltration properties as described in section 5.2, the addition of pore-forming agent led to the fabrication of reliable composites but weakened the material’s strength slightly; so the addition of pore-forming agent results in more reliable composite materials, as can be seen by the increase in Weibull modulus. The increase in bending strength with a small amount of copper oxide added was immense compared with the pure alumina sample.

158

Strength, reliability and stiffness were slightly reduced for preforms containing pre-reacted aluminate (AA34), as shown in Table 14. The amount of aluminate equalled the amount of the 2 wt% copper oxide added, showing a full reaction during sintering. In the preparation process of preform powders, the drying process was different: while the sample with pre-reacted aluminate has been dried using a freeze-drying process, the sample with copper oxide addition has been spray-dried. For comparison a sample with 2 wt% copper oxide prepared by freeze-drying exhibited a bending strength reduction by about 5%. Taking this into account, both sample types exhibited comparable strength, with

ߪସ௉஻ = 767 MPa for the sample with Cu2O added and ߪସ௉஻ = 761 MPa for the sample with pre-reacted aluminate addition.

In Table 14 some properties of material types investigated are given for overview.

Table 14: Mechanical properties of selected composites of various preform types.

Sample type Ceramic Bending Weibull Specimens Young’s Fracture (Cu2O / PFA) content strength modulus evaluated modulus toughness KIC [ ] ([wt%]) [vol%] [MPa] [ ] [ ] [GPa] [MPa*m½]

AO (0/0) 54 363 n/a 4 177 n/a

A10 (0/10) 52 257 9.6 15 203 5.3

A10 (0/10) 59 238 6.0 15 235 5.4

AC20S (2/0) 52 808 25.6 37 219 9.8

AC20/5S (2/5) 47 727 23 10 209 10.4

AC50 (5/0) 52 657 7.0 16 210 9.2

AA34 (2/0) 53 761 18.6 11 237 n/a

159

5.6.2 Elastic Modulus Young’s modulus of the composite material is affected by the ceramic backbone and increases with the ceramic fraction. Considering the sample type A10 sintered at two different temperatures exhibiting different ceramic fractions at comparable microstructure, shows the aforementioned and expected behaviour. The different sintering of the alumina particles led to ceramic volume contents of 52% and 59% with 203 GPa and 236 GPa, respectively. The bending strength and Weibull modulus, instead, reduced with the increase of ceramic fraction. Nevertheless some influence of the copper oxide addition to Young’s modulus was visible in the sample types AO and AC20S, exhibiting 177 GPa and 219 GPa. The increase in modulus might not have been the result of aluminate being present in the microstructure, but caused by the improved infiltration behaviour due to fewer infiltration defects. Porosity was measured according to Archimedes’ principle, showing 5.1% and 3.8% residual porosity for the samples without and with 2 wt% Cu2O added, respectively. Adding more than 2 wt% copper oxide, Young’s modulus decreased slightly to 210 GPa for 5 wt% addition with residual porosity of 1.9%.

270

250

230

210

190 meassured

Young's modulus [GPa] 170 HS-upper HS-lower 150 AC5S AC20S AC30S AC50 AO A10 AC20/5S AC20/10S (0/0.5) (0/2) (0/3) (0/5) (0/0) (0/10) (2/5) (2/10)

Sample type (addition of Cu2O / PFA [wt%])

Figure 103: Young’s modulus of selected sample types with predictions based on the upper and lower boundaries of the model by Hashin and Shtrikman225.

160

Hashin and Shtrikman225 present a model to determine the Young’s modulus of a composite material based on the modulus and volume fractions of their constituents. They propose the calculation of upper and lower boundaries by neglecting the morphology of the reinforcing phase. Samples with copper oxide added in preform preparation showed good agreement with this model, as indicated by the experimental values lying between the predicted upper and lower boundaries shown in Figure 103. For pure alumina preforms the model overestimates the Young’s modulus: this discrepancy was reduced by adding pore-forming additive in preform preparation, hence enhancing the infiltration. So Young’s modulus contributes to the infiltration behaviour rather than to the presence of copper oxide or resulting aluminate in the microstructure, as was expected.

5.6.3 Fracture Toughness and Fracture Behaviour An increase in fracture toughness with the addition of copper oxide to the preform was observed, in addition to the reported increase in bending strength and reliability, as shown in Table 14. Fracture toughness improved with the addition of pore-forming agent. The bimodal pore structure of the preforms resulted in a composite microstructure exhibiting relatively large copper ligaments which could deform in a ductile manner, contributing to better fracture resistance through crack wake bridging. This behaviour was observed in similar composites102, 226. Adding more than 2 wt% copper oxide the fracture toughness ½ decreased slightly to 9.2 MPa*m for 5 wt% Cu2O addition.

Tensile experiments were conducted on samples prepared for microstructural analysis before testing. The surfaces were mechanically ground and polished for microscopic investigation. After testing, the specimens were investigated in SEM to analyse the area next to the fracture. In Figure 104 a region flanking the fracture plane of a specimen without copper oxide addition is shown. The images were obtained in backscattered mode. For clarification, the tensile experiment on the area investigated is shown schematically:

161

delamination

F

F

5 μm

Figure 104: Cracks near fractured surface of a tensile specimen without copper oxide added in the preform: one side crack is shown in detail, others marked. Image was performed in SEM in backscattered mode.

Cracks evolved during testing flanking the fracture plane could be seen as both sides. The crack, as can be seen in the detailed view, evolved along the interface of copper and alumina. The crack sides were flat, without any deformation visible at the surface and particles undamaged. This behaviour could be observed in the other side, as well. In the material close to fracture plane some isolated delamination or disjointed alumina particles were found. There are also some platelets, broken perpendicular to the longer side.

Figure 105 shows at a higher magnification the edge of fracture in the same specimen:

162

ductile tearing of ligament

fractured surface

polished surface 2,5 μm

Figure 105: Microstructure near the fracture of a sample without copper oxide addition after tensile testing. SEM image obtained in backscattered mode.

The appearance close to the fractured surface equalled that described and seen in the side cracks of the image above. Delamination of copper and barely- detached or broken alumina particles can be recognised. Looking down at the fractured surface, round alumina particles can be observed and the copper phase ligaments have been plastically deformed partly during fracture.

Figure 106 shows the behaviour of an analogous sample containing 2 wt% of copper oxide in the preform. The image shows the crack’s edge and a side crack parallel to the actual crack.

163

5 μm

Figure 106: Edge of fracture and side crack parallel with the actual crack of a sample containing 2 wt% Cu2O in the preform. SEM image obtained in backscattered mode.

The microstructure of this sample showed many more defects when compared with the sample without copper oxide addition: as well as delamination, breakage of alumina particles could be observed. The side crack itself showed less debonding than actual breakage, either in the alumina or in the copper phase. The interface showed a stronger behaviour evidenced by less debonding of phases. Here the copper phase adheres to some alumina particles along the crack and no crack growth along the interface can be observed.

Figure 107(a) is a detailed view of the tip of a crack along a copper/alumina interface, whereas in Figure 107(b) a fractured surface is shown.

164

(a) (b)

250 nm 500 μm

Figure 107: (a) 100,000x magnification of a tip of a crack at a copper/alumina interface, and (b) the fractured surface at 50,000x magnification. SEM images obtained in backscattered mode.

The images show that there is fairly good bonding between alumina and copper, as was expected. This is illustrated by the copper phase sticking to the alumina particle found at the crack flank (Figure 107(a)) and on the fractured surface (Figure 107(b)).

The role of aluminate agglomerates in fracture is shown in Figure 108, where agglomerates are displayed near the crack flank.

(a) (b)

20 μm 5 μm

Figure 108: Aluminate agglomerates near the fractured surface at different magnifications. Agglomerates and a side crack are marked in (a), with a detailed view in (b). SEM images obtained in backscattered mode.

The agglomerates of aluminate were damaged, although 50-100 μm from the fracture surface. Two agglomerates are shown and marked in Figure 108(a), with the one closer to the crack flank displayed in detail in Figure 108(b). Also marked in the image is another crack evolving in the microstructure. The aluminate agglomerates seemed to be brittle as the crack began to close at the edge of the agglomerate.

165

In Figure 109 another aluminate agglomerate is shown, 20 μm from a crack flank.

2.5 μm

Figure 109: Aluminate agglomerate with severe damage, located about 20 μm from a crack flank. SEM image was performed in SEM in backscattered mode.

The agglomerate was severely damaged. Several cracks were encountered and the bonding to both the alumina and copper seemed to be stronger than the phase itself. No debonding occurred. The ductile copper phase seems to bridge the crack between the two particles of aluminate and alumina.

In Figure 110 the microstructure near the fractured surface is displayed for comparison with the image of a specimen without the copper oxide addition shown in Figure 105.

166

2.5 μm

Figure 110: SEM image of microstructure near the fracture of a sample with 2 wt% copper oxide addition to the preform after tensile testing.

Several alumina particles close to the edge are damaged and debonding not to be found. Small cracks are seen with alumina on both sides, stopping in the copper phase. Particles without any copper adhered to it were rarely found at the edge of the crack. At the fractured surface the copper seemed to be somewhat more deformed compared with the surface in Figure 105, and replica particles equally rarely found. At several alumina particles an adherent copper phase was observed, deformed plastically but in contact with the alumina.

Fracture surfaces of specimens tested in 4-point bending were also investigated. Figure 111 shows the fracture surface of a composite specimen without copper oxide addition. The strength of this particular specimen was evaluated as 315 MPa, the average strength of this composite type is

ߪସ௉஻ = 360 MPa. For comparison, in Figure 112 the surface after fracture of a composite specimen containing 5 wt% copper oxide with a mean strength of

ߪସ௉஻ = 620 MPa is shown; the specific strength of the sample is measured at 777 MPa.

167

Al2O3

Cu

1.5 μm

Figure 111: Fracture surface of a 4PB specimen: a pure alumina preform has been used for manufacturing the composite sample. Replicates of the alumina particles in the copper phase can be observed (indicated by arrows), indicating poor adhesion.

The appearances of the copper phase on both fracture surfaces were different. In the sample without copper oxide addition the copper phase appeared to replicate the alumina phase, indicating poor bonding behaviour. Deformed copper ligaments were rarely found. The alumina particles showed no residual copper at the surface, in contrast with the specimen with copper oxide addition shown in Figure 112: here the copper phase adhered to the alumina surface. Ligaments of copper were plastically deformed, indicating ductile behaviour of the metal phase. Replicates of alumina particles in the copper phase were rarely found, in contrast with the specimen with no copper oxide addition (which leads to the formation of aluminate agglomerates in the composite structure, as described earlier). These agglomerates showed brittle failure, but good adhesion to copper in accordance with the investigations described. Yoshino and Ohtsu149 also describe the behaviour of aluminate as of a brittle nature. Fracture toughness increased with the addition of copper oxide to the preform structure, as shown in Table 14. The fractured surfaces shows the reasons for this; ductile deformation of copper ligaments, indicating crack bridging, can be seen for the copper oxide containing preform rather than for the pure alumina

168

preform. Large agglomerates of copper in the composite microstructure, resulting from the artificial pore-forming additive, improve the fracture toughness. Considering Figure 111 and Figure 112 it is seen that the toughening effect of a certain amount of large pores cannot balance the effects shown for small copper oxide addition; a preform with 10 wt% cellulose addition exhibiting fracture toughness of 5.3 MPa*m½, whereas the addition of 5 wt% ½ Cu2O results in fracture toughness of 9.2 MPa*m with both sample types showing 52% of ceramic volume content.

CuAlO2

Al2O3

Cu

1.5 μm

Figure 112: Fracture surface of a 4PB specimen from infiltration of a preform containing 5 wt% copper oxide: good adhesion between copper and alumina, as well as deformed copper ligaments (indicated by arrows) are observed.

The different fracture behaviour of preforms with and without copper oxide addition was shown: this is consistent with the investigation into bending strength. The addition revealed higher bending strength, as is reflected on the fracture surface. Aluminate agglomerates showed cracks, even several microns from the fracture plane. The bonding between the copper phase and the alumina particles was of a better quality and the microporosity decreased because of the copper oxide addition, as shown in the microscopic investigation. The early breakage of the aluminate agglomerates led to energy dissipation – microporosity weakens the creep behaviour of a material and 169

promotes fracture growth, as investigated by Requena and Degischer227 for aluminium piston alloys and composites. This effect of microporosity could be transferable to investigation into bending strength: its strong influence on the higher bending strength is probably related to the better bonding of copper to alumina. The origin of this might be an effect of oxygen solution in the melt and, hence, promoted wetting. The decrease in microporosity could be attributed to enhanced wetting behaviour.

5.6.4 Thermal Conductivity The thermal conductivity of different samples was measured at room temperature. The diversity of samples ranged from copper oxide-containing preforms with 0.5 – 5 wt% Cu2O and pore-forming agent of 0 – 5 wt%, to copper oxide-free preforms with 0 - 20 wt% pore-forming addition. Due to the different infiltration behaviours and ceramic volume content, the residual porosity of infiltrated samples was investigated. In Figure 113 the thermal conductivity of different samples depending on the copper fraction is shown: residual porosity, measured using Archimedes’ principle, has been taken into account by determining the copper fraction with regard to porosity and ceramic volume content.

180 Containing copper oxide 170 Copper oxide free

160

150

140

130

120

Thermal conductivity [W/mK] conductivity Thermal 110

100 0,3 0,4 0,5 0,6 Copper volume fraction

Figure 113: Thermal conductivity of composite samples of Cu2O-containing preforms (blue diamonds) and samples of copper oxide-free preforms (grey triangles) depending on copper content.

170

Thermal conductivity of composite samples increased with copper fraction. This was expected, considering the thermal conductivity of the main phases copper and alumina, are >394 W/mK and 30 W/mK, respectively, as in Table 5. After squeeze casting, the rest of raw copper was measured, having a thermal conductivity of 385 W/mK. The addition of copper oxide was believed to enhance the overall thermal conductivity of the composite material, due to the improving of infiltration and bonding of phases. Evaluation therefore could only be done by considering the copper fraction. All samples measured exhibited somewhat different ceramic volume content and residual porosity – hence, fraction of copper phase. A ratio was defined by the relation of thermal conductivity ߣ and to the copper fraction by:

O ratio %Copper Equation 41

The ratio gives the specific thermal conductivity as a fraction of copper content.

Figure 114 shows the average of the ratio of measured composite samples (Equation 41) depending on copper oxide content.

4 arithmetic average 3,8 min. / max. value 3,6

3,4

3,2

Cu% 3 / λ 2,8

Ratio 2,6

2,4

2,2

2 012345

Cu2O addition [wt%] Figure 114: Arithmetic average of the ratio defined as the thermal conductivity per 1% of copper, with the deviation showing the minimum and maximum value of samples evaluated.

171

Composition types with and without added pore-forming additive were taken into consideration, as no influence on conductivity resulting from the bimodal pore structure could be observed. In contrast with the behaviour expected, thermal conductivity related to the copper phase reduced with the addition of copper oxide in preform preparation. The average of the ratio of copper oxide- free compositions could be given as 3.13 W/mK decreasing to 2.8 W/mK for

5 wt% copper oxide addition. The average of samples containing 2 wt% Cu2O without pore-forming additive with 3.04 W/mK and showed no noticeable difference from samples with pore-forming additive – 3.08 W/mK. Despite the reduction of the specific thermal conductivity, infiltration behaviour of samples improved, residual porosity reduced and the bonding behaviour of phases was enhanced with the addition of copper oxide. Considering the investigations of some authors228-230 who state that the enhanced bonding of phases can improve the overall conductivity of a composite sample by the enhanced transition of conductivity of one phase to the other, there seems to be a contradiction: in their models of particulate-reinforced metals they describe this behaviour as a factor of interfacial thermal barrier resistance. But these models are not directly comparable with the material investigated in this work; as in them the single particulates considered are solely in contact with the matrix. However, the effect of interfacial thermal barrier resistance can be taken into consideration. On the other hand, they investigate particulates that are chemically stable in the matrix: this might not apply for the material investigated here, as has been stated: the enhanced infiltration suggested a solution of oxygen in the melt through dissolution of aluminate agglomerates, which could interfere with the effect of enhanced conductivity due to better bonding and less residual porosity, resulting in enhanced infiltration through the alloying of copper. Thermal conductivity in general reduces with the metal alloying, but in the case of copper the influence of oxygen is only marginal231; so the decrease in thermal conductivity might not have been related to alloying with oxygen. According to the phase diagram shown in Figure 39, copper can also dissolve aluminium: this could have contributed to the decrease in thermal conductivity, as copper can dissolve up to 9.4 wt% Al232: alloying elements with the formation of solid solutions is known to affect strongly the conductivity of copper231, 233.

172

An influence of pore-forming addition on thermal conductivity could not be shown. The specimens containing pore-forming agent were measured perpendicular to the pressing direction: direct comparison with specimens taken parallel with the pressing direction was not possible due to the infiltration set-up and specimen geometry used for measuring the thermal conductivity. Still, the thermal conductivity of specimens with cellulose addition was measured along the aligned copper ligaments resulting from the cellulose particles, but not showing remarkably higher conductivity, as described. This led to the conclusion the copper fraction and internal defects were most responsible for the overall thermal conductivity, rather than the microstructural appearance. Oxygen dissolution in the copper melt acted in two ways: the infiltration behaviour was enhanced, leading to less microporosity and better bonding; while dissolved alloying elements like oxygen and aluminium probably caused a decrease of thermal conductivity.

5.7 Summary The manufacturing of ceramic preforms with open cell structure containing small amounts of copper oxide has been described. Copper oxide, which appears homogeneously distributed, primarily in large agglomerates with the remainder being small particles, reacted during sintering with the alumina phase being in contact to aluminate. The infiltration behaviour of such preform types is improved, compared to those without the addition. Whereas the improvement is not governed by the infiltration initiation, it can easily be seen in cross-sectional views of infiltrated specimen. The microstructure of copper showed the main phases copper and alumina in addition to the brittle aluminate phase, which was analysed in nanoindentation testing. However, bending strength Young’s modulus and fracture toughness increase with the addition of small to moderate amounts of copper oxide to the preform structure.

Aluminate is seen to enhance bonding of copper/alumina joints when located at the interface. However, in the composite structure no continuous interfacial aluminate phase was observed. A different process – dissolution of aluminate during infiltration with the in-situ alloying of the infiltrating copper melt – is under investigation. The behaviour of aluminate during infiltration, therefore,

173

will be discussed shortly with the strengthening effects described in composite manufacturing.

174

6 Discussion This chapter presents a discussion of the results reflecting on the existing literature. It is structured as follows: the preform preparation process with phase formations during sintering followed by the infiltration of preforms, the microstructure of the resulting composite and the influence of aluminate in the structure. Finally the hypotheses given in chapter 3 are revisited.

6.1 Preform Preparation and Phase Formation during Sintering Preform preparation was done using an established wet powder processing method234, 235 without much adaption for this investigation. The main adaption could be described as the use of a second ceramic phase; one phase being the main ceramic (alumina), and the other the copper oxide addition. The active component Cu2O, added in small amounts to the powder system in preform preparation, led to a different behaviour for composite manufacturing, as will shortly be discussed.

Most powder compositions were made without the pore-forming agent cellulose. The targeted porosity range of preforms for further mechanical investigation was 50 ± 5%: preform compositions with small–medium copper oxide additions were within this targeted porosity range. For compositions without copper oxide addition the sintering temperature and, hence, densification during sintering, were higher. Therefore, for those preforms, especially, the reduction of ceramic volume content was done by the addition of the pore-forming additive cellulose. Comparison and other investigative reasons meant subsequent powder compositions included copper oxide and cellulose additions.

Mattern102 reports a swelling of cellulose particles during an aqueous preform forming, leading to preform cracking during drying: in this work, rather than the aqueous method, a dry-pressing process was applied in which preform cracking was prevented. It is interesting to note that preforms with very high amounts (more than about 30 wt%) of the pore-forming additive (PFA) cellulose could not be used for further MMC processing, as they were cracked, and often too weak to allow handling. This behaviour was attributed to the expansion of

175

preforms during sintering as a result of the thermally-induced relaxation of cellulose particles, accompanied by weakening of the polyvinyl alcohol binder during heating. Preforms with more than 65–70% total porosity were therefore not possible to fabricate.

The fundamentals of pore formation with cellulose in copper oxide-free preforms are investigated by Staudenecker235. Porosity was achieved by using incomplete densification of the particles to give fine intergranular pores in combination with larger pores, which were formed by the pyrolysis of the PFA. At the relatively low sintering temperatures and times used, the volume diffusion was kept low and sintering led to small contact points, as shown in the preform microstructure (e.g. Figure 49). Despite this ability to control the microstructure, the targeted porosity range for copper oxide-free compositions was only reached with PFA; and the infiltration ability of pure alumina samples was found to depend strongly on the existence of pores resulting from the pore-forming agent.

Preforms of powder compositions with copper oxide content showed a somewhat different sintering behaviour. No sintering shrinkage of the green bodies to the resulting preforms could be observed, whereas pure alumina preforms exhibited approximately 15% volume shrinkage, depending of the targeted porosity range and composition. Considering the copper oxide-alumina phase stability diagram shown in Figure 14 and in Figure 115(b), a liquid phase would form locally during sintering. The sintering temperature for preforms with copper oxide addition is held strictly to 1200°C, above the Cu2O-CuAlO2 eutectic shown in the phase diagram. Liquid phase sintering normally indicates either densification or swelling of the compact, depending on the contact angle of the liquid; whereas low contact angles result in compact densification and high contact angles, indicating poor wetting and leading to compact swelling with the liquid retreating from the solid as reviewed by German et al.236. Neither phenomenon was observed after sintering, even if the copper oxide addition relative to the alumina particles was high. The liquid phase sintering process is normally applied to derive bodies with high theoretical densities; so a general

176

assumption of the process might not be applicable to preforms with about 50% residual porosity.

The sintering temperature of 1200°C resulted in a relatively low sintering activity of the alumina particles. In a different system, Staudenecker235 suggests that for alumina preforms sintered at 1000°C, with a glassy frit binder that normally softens at 570°C, the preform strength is only attributable to the glassy binder – so pure alumina preforms sintered at 1000°C are too weak to handle and no sintering can be observed. In this work the sintering temperature for the compositions containing copper oxide was about 200°C higher. Reference samples of pure alumina without PFA were sintered at 1200°C as well, and showed enough strength (cf. Table 10) for further MMC fabrication. The ceramic volume content for both preform types (with and without copper oxide addition) sintered at 1200°C was roughly the same, showing no influence of densification due to copper oxide addition. The ceramic volume content for copper oxide- containing compositions is preset by the compaction of green compacts; gaining preforms with higher porosities, therefore, can only be managed by the addition of PFA. On the other hand, higher ceramic fractions were achieved using alumina powders with higher compaction ratios – hence, higher green densities (Nabaltec powders). Additionally, these alumina powders indicated higher sintering activity, and the preforms containing copper oxide showed a sintering shrinkage of about 9% by volume at 1200°C.

In powder preparation, the alumina and copper oxide phases were prepared by wet powder processing, resulting in degglomeration and homogeneous distribution: subsequent drying and moistening steps did not alter or decompose the powder composition, so green compacts contained copper oxide well distributed in the main constituent, the alumina powder. In the sintering process, the subsequent heat treatment, influenced the phase composition in the preform, as per the phase stability diagram shown in Figure 14. As described, a liquid will form above a temperature of 1165°C, which is the Cu2O-Al2O3 eutectic temperature: this happens on the Cu2O-rich side of the phase diagram. Even if the powder composition is a long way towards the alumina-rich side of the phase diagram, this may be applicable

177

locally, where copper oxide agglomerates are in contact with some alumina particles. It can also be assumed that where copper oxide is locally abundant, the Cu2O-Al2O3 eutectic melt will form: further reaction being rapid in the liquid state and the equilibrium balanced towards the alumina side of the phase diagram, resulting in solid aluminate and alumina. Aluminate melts at temperatures above approximately 1260°C, which is well above the sintering temperature. He et al.237 investigate a two-step bonding process of copper to alumina: they first apply a Cu2O paste on the alumina surface followed by firing at 1150°C for 60 min – this is close to the sintering process in this work, wherein at the interface the cuprous oxide layer reacted with alumina by the reaction shown in equation 30, and restated here as the compound aluminate

ଶܱ݈ܣݑܥ ଶܱଷ ՜݈ܣݑଶܱ൅ܥ

He et al.237 observe the appearance of an aluminate phase by XRD- examination and find a diffusion layer of 50 μm between copper and alumina investigated by EDS analysis. Nevertheless, a discrete aluminate layer is not seen in the microscopic image the authors provide. In regard to the phase stability diagram, the reaction of aluminate formation appears in solid state, as the firing temperature is below the Cu2O-Al2O3 eutectic temperature.

The reaction time in diffusion bonding is slow. Beraud et al.141 report a thickness of interfacial aluminate layer of about 0.2-0.4 μm after a diffusion bonding process of 2 h at 1000°C in Ar atmosphere where Cu2O was present at the interface. Kim and Kim161 find a linear growth behaviour of the aluminate layer in eutectically-bonded copper alumina joints. They claim the formation is reaction-controlled and propose a thickness growth rate of about 8x10-5 μm/s. But in their study the aluminate thickness of about 8 μm after 20 h bonding time is shown, which is about 40% thicker than expected by calculating the given growth rate. Nevertheless, they state that the formation of aluminate is a very slow reaction. They also investigate, as mentioned, the liquid phase bonding; stating the reaction of Cu2O with alumina to be very rapid, and CuAlO2 formed by the liquid phase reaction in accordance with equation 30. The aluminate thickness reached 0.1 μm in only a few minutes.

178

The formation of aluminate had to be fast, given the Cu2O agglomerates observed in the green bodies in this work and the resulting aluminate agglomerates with diameters of up to about 40 μm, as can be seen in the computed tomography investigation shown in Figure 43, or microscopy images shown in Figure 71. Sintering time was held to 2 h and in the preform structure no remaining copper oxide could be found by XRD examination. Hence the reaction of cuprous oxide with alumina to aluminate was fast, which is not likely to happen in solid state or diffusion bonding. Therefore, liquid phase reaction can be assumed for the formation of aluminate taking place, as proposed earlier according to the phase diagram of Jacob and Alcock155 shown in Figure 115(b).

Phase formation during sintering was examined by XRD investigation of preform bodies after sintering, as shown in Figure 36. Whereas for prolonged sintering times and sintering at reduced oxygen partial pressure, only aluminate and alumina was apparent in the preform structure, the spinel copper/aluminium oxide CuAl2O4 was also detected in preforms with 2 h sintering in air. He et 237 al. examine the Cu2O-Al2O3 bonding behaviour in their study: they paste a thin layer of Cu2O powder mixed with 100% alcohol on the alumina surface and during subsequent heat treatment for 60 min at 1150°C, the copper oxide near the alumina reacts to aluminate according to the reaction shown in equation 30, as mentioned. However, the surface (side towards atmosphere) of pasted cuprous oxide reacts with oxygen to form CuO. According to Jacob and Alcock155 the spinel will form from the reaction of cupric oxide with alumina, as follows and previously shown in equation 31:

ଶܱସ݈ܣݑܥଶܱଷ ՜݈ܣݑܱ ൅ܥ

The spinel structure in the preforms investigated in this work, therefore, may be an attribute of the predominance of oxygen leading to the oxidation of cuprous oxide to cupric oxide and, hence, the formation of copper aluminum oxide by the reaction shown in equation 31. Misra and Chaklader166 investigate the phases resulting from heat treatment from 600 to 1600°C of mixtures of CuO and ߙ-Al2O3 in equilibrium conditions by x-ray diffraction, and propose the phase diagram shown in Figure 14. Accordingly, above 800°C the spinel phase

179

CuAl2O4 will form and above 1000°C the aluminate phase CuAlO2 will 155 equilibrate with Cu2O and Al2O3. Jacob and Alcock electrochemically measure the standard free energy of the formation of CuAlO2 and CuAl2O4: the proposed phase diagrams of both studies are shown in Figure 115. The latter evaluate the spinel structure to be stable from 612°C up to 1171°C at the alumina-rich side. Nevertheless, for the equilibrium aluminate instead of spinel was present for the preforms sintered in this work, the sintering temperature was above the 3-phase equilibrium CuAlO2-CuAl2O4-Al2O3 of both authors. And this is in accordance with the x-ray phase investigations shown in Figure 36.

(a) (b)

Figure 115: Stability phase diagram of (a) Jacob and Alcock155 and (b) Misra and Chaklader166 for the copper oxide/alumina system.

Copper oxide-containing preforms processed in this work were sintered at 1200°C, as mentioned. In preform preparation copper oxide formed agglomerates, but appeared homogeneously distributed in the powder composition. During sintering the copper oxide formed, in contact with alumina particles, the copper/aluminium oxide with the sum formula CuAlO2 as per the reaction shown in equation 30. Though some of the cuprous oxide reacted with oxygen of the atmosphere, resulting in spinel phase formation according to the reaction shown in equation 31, for longer sintering times only aluminate was

180

present, in good accord with the calculations of Trumble156, who states aluminate as being the more stable phase.

The reaction of Cu2O with Al2O3 to form CuAlO2 causes an increase in volume and size of the agglomerates: to prevent this, preforms with pre-reacted aluminate particles were prepared. Finely-distributed aluminate without phase reaction was expected during sintering, as conditions were equivalent to those for copper oxide-containing preform compositions – in accord with phase diagrams shown. In Figure 44 the fine distribution of pre-reacted aluminate in the preform microstructure was shown: the influence of the use of pre-reacted aluminate to the properties of the resulting composite will be discussed further. The surface area of both preform types – with and without added copper oxide – is comparable. Measured by BET technique and Hg-porosimetry, the surface area was 1.10 m²/g for the pure alumina preform with a ceramic volume content of 52.8% and 1.17 m²/g for the preform with 2 wt% copper oxide addition, exhibiting a ceramic volume content of 49.7%.

6.2 Infiltration Behaviour In this work, the targeted preform porosity range for further mechanical characterisation was 50 ±5%. Overlooking compression, the size of the porosity measured in the ceramic preforms represented the size of the metal phase after infiltration. So the pores in the preform was replaced by the metal during infiltration, leading to the desired composite material: this replacement happened in the infiltration process and will be discussed further.

6.2.1 Saturation of Porous Media Infiltration is governed by the replacement of porosity in the ceramic structure with metal: residual porosity is detrimental to the composite properties, so it is important to strive for full infiltration. The prerequisite is an open cell structure – hence, permeable preforms. The preforms investigated in this work exhibited closed cell porosity below 1%; and the permeability of both preform types, with and without copper oxide addition, was almost the same – this also applied to moderate copper oxide additions. Moderate additions will be discussed; and if not further specified, preform compositions with 2 wt% Cu2O addition are taken

181

as exemplar copper oxide containing preforms. Pure alumina preforms exhibited a permeability of K = 4.6x10-15 m2 with a standard deviation of 1.4x10- 15 at a ceramic volume content of about 52.5%. This was almost the same as for copper oxide-containing samples, at K = 5.6x10-15 m2 with a standard deviation of 1.4x10-15 at a ceramic content of about 53%. The strong improvement in composite fabrication due to the copper oxide addition in powder preparation, therefore, could not be attributed to the permeability of resulting preforms. The infiltration could be characterised by infiltration initiation – the first penetration of liquid into the open cell porosity – stable infiltration and penetration trough, according to Long et al.119.

Infiltration initiation pressures were evaluated as shown in Figure 60, e.g. The evaluated threshold pressure for both pure copper and copper-oxygen alloy infiltration differed from that calculated using equation 15, according to Garcia- Cordovilla et al.48. The calculated infiltration initiation pressure was 3.3 MPa for the infiltration with pure copper and reduced to 0.5 MPa for the oxygen alloyed copper. The evaluated threshold pressures for pure copper infiltration were 2.6 MPa for the copper oxide-containing sample and 2.7 MPa for the alumina sample. For copper-oxygen alloy infiltration a reduction to 2.3 MPa for both preform types could be observed. As stated, the ceramic volume content, especially for those specific preforms, was almost the same and within 50.7% and 52.5%, and therefore not correlated with the reduction of threshold pressure. It is generally acknowledged (and described earlier) that oxygen promotes the wetting of copper on alumina2, 5, 14, 15, 21, 23, 25-27, 51. The discrepancy of calculated and measured infiltration-initiation pressures of approximately 0.6 MPa for pure copper infiltration may be a consequence of the assumptions made for calculation: whereas the surface area of preforms was measured, the contact angle and surface tension were taken from the literature5, 238. A contact angle of 125° was taken, measured in high vacuum conditions5 as the copper was melted in a reducing atmosphere. The infiltration process in this work was performed in an ambient laboratory atmosphere, which could have affected the melt’s wetting properties. A reduction of about 5% in wetting angle to 118°, which can be related to oxygen partial pressures of -9.3 -3 p(O2) = 10 or oxygen content in the melt of 7x10 at% in equilibrium

182

condition at 1300°C as reported by Diemer5, would reflect the measured infiltration initiation pressures in this work. With infiltration times below 20 s, and thus contact of melt with normal atmosphere, equilibrium conditions could not be achieved. But the investigations of Diemer5 show the sensitivity of contact angle on oxygen addition and may explain the discrepancy between measured and calculated infiltration initiation pressures for copper infiltration.

The addition of copper oxide to the melt and, hence, alloying with oxygen, reduced infiltration initiation pressures. The discrepancy between the measured (2.3 MPa) and calculated (0.5 MPa) infiltration initiation pressures with Cu-O alloy may have been attributable to a lower than expected oxygen content in the melt, as the alloying of copper was undertaken five minutes before infiltration and the melting conditions were of a reducing nature (carbon crucible, argon atmosphere). The contributions of atmosphere and dissolved oxygen to the melt are described in the investigations of Diemer5: the oxygen content of the melt and, hence, the wetting angle, exhibit a significant influence upon variables investigated. Accordingly, the influence on the measured infiltration initiation pressures could be related to a reduction of wetting angle to 113°. The infiltration initiation was influenced by the pore size of preform according to capillary law. The average pore size of a preform containing 2 wt%

Cu2O was measured, in mercury intrusion porosimetry, to be 0.81 μm. Therefore, the contact angle of the melt could be calculated with the measured threshold pressure for infiltration taken as the capillary pressure, according to equation 17. This was evaluated for infiltration with pure copper as 116°. Taking the threshold pressure for the infiltration with Cu-O alloy, the contact angle evaluated was as low as 112°, which correlated well with the derived values. A calculation of dynamic contact angle for infiltration initiation, according to equation 15, showed contact angles of 117° ± 2° for infiltration of copper oxide- containing preforms with pure copper alloy, in good agreement. The calculation was based on 9 single infiltrations, showing good reliability for the experimental set-up. The infiltration of pure alumina preforms showed wetting angles of 118° for pure copper infiltration and 113° for copper-oxygen alloy infiltration. Copper oxide-containing samples infiltrated with Cu-O alloy exhibited wetting angles calculated at 112°. This did not lead to a significant change in either the wetting

183

angle or infiltration initiation pressure due to the addition of oxygen to the melt in the described way: the atmosphere, however, played an important role in composite fabrication because of its influence on the infiltration alloy. As a consequence, more reliable investigations of infiltration initiation would need to be undertaken in a controlled atmosphere, where the entire squeeze cast infiltration set-up with preform and melt was in an autoclave.

6.2.2 Influence of Aluminate Phase The preform structure in equilibrium condition exhibited aluminate, as well as the main ceramic alumina, if copper oxide was added in powder preparation (as shown previously) – aluminate does not affect infiltration initiation pressure. This was in agreement with the investigations of Diemer5, who show that the wetting angles of copper in contact with alumina and aluminate are the same. The oxygen content in the melt was a sensitive parameter influencing the wettability in the copper/alumina system2, 5, 14, 21, 23, 25-27, hence affecting the initial infiltration pressures. As the microstructure was mainly unaffected by the addition of copper oxide, and the wetting properties did not differ by the formation of aluminate, no difference of threshold pressure could be expected, in accordance with capillary law, where the average pore diameter r and the wetting properties ߛ௟௩ and ߠ are the influencing factors for infiltration pressure (equation 17). Referring to Figure 61, where the threshold pressure of preforms with different copper oxide additions were evaluated, no real tendency of copper oxide content to threshold pressure can be seen. All threshold pressures shown are within 0.1 MPa, which is within the standard deviation of 0.11 MPa for Cu infiltration of preforms with 2 wt% copper oxide addition, based on 9 samples.

The slope mL of the regression line indicates the progress of infiltration. As shown in (e.g.) macroscopic images – Figure 55 – the presence of aluminate in the preform structure improves the infiltration progress. This seems to be in contradiction to the observations made by Diemer5 in terms of wetting improvement due to the appearance of the aluminate phase; but a reduction in the slope mL depending on copper oxide addition can be seen in Figure 61, where the first infiltration stage, to about 5 mm of infiltration depth, is shown. Aluminate dissolves peritectically to alumina and Cu-liquid at about 1260°C, as

184

shown in the phase stability diagrams of Misra and Chaklader166 and Jacob and Alcock155 in Figure 115. By the dissolution of aluminate, the resulting Cu-liquid will be alloyed with oxygen, leading to enhanced wetting properties. This could explain the improvement of infiltration progress. Infiltration was intended to occur with superheated copper of 1250°C. According to the temperature control of the furnace used for melting, the regulation of temperature is estimated as ± 20°C. As the melt temperature was seen to be better slightly above rather than below 1250°C, the temperature control was set to 1260°C, thus ranging between 1240°C and 1280°C. The temperature of the melt could be above the aluminate formation temperature, leading to its dissolution during contact: considering the infiltration set-up, this was unlikely to happen. The copper melt, molten in an external furnace, had to be transferred to the infiltration die. While the entire infiltration, including all transfer steps, was completed within 20 s, as has been stated, the transfer of melt until first contact with the preform occurred within 2–3 s. Assuming a melt temperature of 1280°C – unlikely to be reached – and a transfer time of just 2 s, the melt temperature was calculated at 1260°C in terms only of heat radiation. Considering further heat loss due to convection the melt would be below the aluminate dissolution temperature by first contact.

Another approach to contribute to enhanced infiltration could be elucidated through considering the XRD examination more closely. As stated, in equilibrium conditions only aluminate is apparent; but normal sintering for further composite processing was performed with 2 h soaking time. These preforms presented a certain amount of the copper/aluminium oxide in the spinel structure, CuAl2O4. No further attention was paid, as in the composite structure only the aluminate phase was detected. Two different processes could be accountable:

(i) the circumstances of infiltration process, or

(ii) the contact with the melt during the actual infiltration taking place.

According to Nagata and Matsuda121 a minimum preheating temperature of the preform is necessary for infiltration, calculated at 926°C. According to this, the preforms in this work were preheated to almost their sintering temperature of 1200°C, to ensure full infiltration of the entire preform. Preheating was done in

185

an external furnace in air at 1170°C, performed at moderate heating rates (50 - 100 K/h) to prevent thermal cracking. Many preforms were preheated simultaneously for infiltration, leading to different preheating times. During preheating a further equilibration of phases could be assumed, leading to aluminate instead of spinel phase in the preforms. This was in accordance with 237 the investigations of He et al. , who anneal Cu2O-Al2O3 joins at 1150°C, leading to aluminate at the interface; and the observations of Misra and 166 Chaklader who investigate fired powders of Cu-Al2O3 mixtures by XRD- analysis and state CuAl2O4 is not stable above 1000°C; but was in contradiction with the electrochemical measurements provided by Jacob and Alcock155 indicating the spinel structure being stable up to 1171°C.

The second process possibly accounting for the lack of spinel detection after infiltration can be described as follows: liquid copper poured onto the preform saturated the pores of the ceramic structure when applying the external pressure. During infiltration the contact of liquid copper with spinel led to the formation of aluminate according to the phase diagram provided by Gonzales et al.213 shown in Figure 37. And as Trumble156 states, only aluminate can be in equilibrium with copper and alumina. The formation of aluminate out of the spinel structure involved the dissolution of oxygen and aluminium in the liquid copper. The oxygen alloying affected the wetting behaviour, as stated, leading to the enhanced infiltration progress. Dissolution of spinel during infiltration and further equilibration during preform preheating would lead to different infiltration behaviour depending on preheating time; hence the amount of residual spinel structure in the preform. This had not been observed; and preforms with preheating times as low as 1 h, for example, exhibited infiltration ability comparable with preforms preheated for 10 h or more. Even preform compositions prepared with pre-reacted aluminate instead of copper oxide or preforms sintered in vacuum – hence providing only aluminate in the alumina structure – showed comparable infiltration ability and bending strength of the resulting composites, as will be discussed.

Beraud et al.141 observe the formation of aluminate at the interface in their copper/alumina solid state bonding experiments at 1000°C in argon

186

atmosphere. They use superficial oxidised copper – copper with a Cu2O-layer at the surface – in contact with alumina leading to a Cu2O-containing interface. The increase in bonding time from 2 h to 4 h firstly increases the thickness of interfacial aluminate layer; with increase of the bonding time to 6 h, they observe the dissolution of aluminate in conjunction with the formation of an alumina layer without any crystallographic relationship to the bulk alumina at the interface. The authors report the dissolution of the aluminate phase as due to the decrease of oxygen content in the surrounding copper by diffusion, giving rise to the destabilization of CuAlO2 for the benefit of Al2O3. This is in accordance with the thermodynamical analysis of Trumble156, who states that a threshold activity of oxygen is necessary for aluminate formation. The minimum concentration of oxygen in solid copper for the {Cu(ss), CuAlO2, Al2O3} equilibrium as a function of temperature is shown in Figure 16. Aluminate formation requires a threshold activity of oxygen in the metal that is less than the solubility limit. The threshold for CuAlO2 formation increases in the temperature range of 700°C to 1000°C from approximately 1/6 (about 1x10-4 at%) to 1/3 (about 6x10-3 at%) of the solubility limit of oxygen in Cu in equilibrium with Cu2O. Hence, if oxygen is abundant aluminate will form, but if oxygen is absent aluminate will dissolve.

In composite fabrication in this work, the copper used was oxygen-free, melted in a graphite crucible in argon atmosphere; hence, no oxygen enrichment during melting was expected. During infiltration the liquid copper saturated the preform by a one-directional infiltration. Considering the liquid state, where copper molten at 1250°C was poured onto the preform, preheated to 1170°C, leading to infiltration with temperatures up to 1250°C, liquid copper can dissolve much more oxygen than solidified copper, so the dissolution of oxygen was fast. While the infiltration front progressed throughout the preform, aluminate agglomerates were, dependent on their location, in contact with unsaturated copper melt. This could have led to enrichment with oxygen due to the dissolution of aluminate to alumina in regard to the earlier observations. The phase stability diagram of Misra and Chaklader166 displayed in Figure 14 shows the Cu2O-CuAlO2 eutectic temperature at 1165°C. Consider the isothermal section of the ternary phase diagram Cu-Al-O, already shown in Figure 37,

187

displayed again in Figure 116: the preform composition of copper oxide containing preforms is schematically shown, marked by a red rectangle. The preform consists of alumina and aluminate, predominately alumina. During infiltration the balance will move towards the copper-rich side, which is indicated by a red arrow in the ternary phase diagram.

1200°C

Figure 116: Isothermal section at 1200°C of the ternary phase diagram of Cu-Al-O213: preform composition is schematically marked by a rectangle. Infiltration with copper moves the balance (indicated by an arrow) towards the copper-rich side of the diagram.

The composite manufacturing started with the sintered preform being preheated at 1170°C, in equilibrium condition waiting for infiltration. Therefore, the preform consisted of alumina (Al2O3) with agglomerates of aluminate (CuAlO2) distributed in the microstructure, represented in the quasi-binary system Cu2O-

Al2O3 (Figure 14) – the composition on the far end side of alumina towards aluminate. The superheated oxygen-free copper melt could have led to local dissolution of the aluminate phase, where the temperature was above the eutectic temperature of 1165°C. Copper dissolves oxygen and aluminium: the oxygen dissolved in the copper led to promoted wetting, hence enhanced infiltration. Dissolved aluminium can precipitate with oxygen to alumina, as also

188

reported by Kim and Kim161: precipitates of alumina were observed in the microstructure, as will be discussed in the next section.

The dissolution of aluminate described during squeeze cast infiltration is in agreement with the observations made in this work by isothermal gas pressure infiltration. A sample with 2 wt% copper oxide added was taken; infiltration was performed at 1200°C. Therefore, the ternary phase diagram of Figure 116 is relevant, and during infiltration the preform equilibrated with the copper melt. Extrapolating the arrow in Figure 116 to about 50% copper in equilibrium condition brought up the balance within the 2-phase section {Cu(L1),

Al2O3} instead of the 3-phase section {Cu(L1), CuAlO2, Al2O3}. The microstructure showed no aluminate, as has been shown in Figure 83. The dissolution of aluminate during infiltration, therefore, is possible for improving infiltration ability resulting in the in situ alloying of the melt.

6.3 Microstructure and Interface Appearance Aluminate at the interface enhances the overall bonding strength, as stated by different authors5, 141, 149, 159, 161: this has been approved for this system by investigations with model-like, almost lamellar preforms – to which the different fabrication of joints is attributed. Whereas in the literature joining is performed with solid copper, hence temperatures below 1086°C, in this work the copper melt temperature was about 1250°C. Infiltration took part by squeeze casting; hence, at relatively fast solidification and cooling times. The contact times at temperatures above the diffusion or liquid state bonding processes were below about 1 min, which is less than in the literature. Apart from this, in literature the joining is 1-step processes, except in the investigation of Yoshino and Ohtsu149 and of He et al.237, who describe the formation of aluminate on alumina at higher temperatures before bonding to copper at lower temperatures. This applied to this work as well, as the aluminate formed before copper was in contact with the preform. Nevertheless, bonding was enhanced if aluminate was present at the interface, compared with preforms without an interfacial layer, as described in section 5.4 and shown in Figure 93 by interfacial Vickers indentation testing.

189

The dependence of bonding between copper and alumina on aluminate in the preform was shown in TEM images as well. It was found that for pure alumina samples the contact of copper to alumina showed by gaps at the interface, whereas the interface for copper oxide containing preform compositions was smooth and without phase delamination, even if no aluminate agglomerate was in the immediate vicinity. The TEM investigations in this work are not satisfactorily statistically verified as the volume investigated is very small, but show a tendency of different bonding characteristic for both preform types, as expected. An interfacial phase could not be detected in the investigations, regardless of imaging method. The bond strength is enhanced also if oxygen is present at the interface, even if no aluminate phase has formed, as reported by Yoshino151 and Diemer5. Therefore, the enhancement of bonding was not necessarily an effect of an interfacial phase but possibly related to the in situ alloying of copper with oxygen, which could not be visualised by the techniques used here for microscopic investigations. A prerequisite for microstructural investigations is the appropriate preparation of the specimen. The general mechanical grinding and polishing procedure for materialographic analysis is not appropriate for thorough analysis of the composite structures investigated in this work in terms of near-interfacial aspects. This is due to the single phase characteristics. Whereas copper shows softness and ductility, alumina is stiff and exhibits good abrasive resistance: this leads to a selective abrasion of the surface in combination with the smearing of copper. As reported earlier205-207, ion polishing is a suitable process for preparation of two- (or more) phase composite structures where one is very stiff and the other(s) ductile or soft. The resulting images are clear and show detailed interfacial sections in SEM investigations, as shown in section 5.3.1. Specimens prepared with the finishing step of vibration polishing instead of ion milling show similar results in spite of clear detail development in microscopic images – shown in Figure 77. Figure 81(c) shows a fine-structured section of a composite sample containing 5 wt% copper oxide: these lamellar-appearing structures have been found where aluminate agglomerates were expected in the microstructure. The appearance is eutectic and can be compared with

190

pearlite in steel, which is the eutectoid of and cementite, representing fine- structured iron and cementite.

In this work it would be alumina and copper, which is indeed a possible reaction. Aluminate dissolves into copper (L1) and alumina according to the ternary phase diagram. These lamellar structures were only found near the upper surface of a preform, where infiltration had started. This can be explained by the temperatures used for infiltration: the melt was superheated to 1250°C which is well above the temperature needed for dissolution. With contact with the colder preform and the tooling material, the melt cooled rapidly below 1170°C needed for dissolution. As can be seen in the phase diagram of

Figure 116, the copper melt Cu(L1) can dissolve oxygen and aluminium. Copper can dissolve more aluminium than oxygen at room temperature, resulting in a stronger influence of Al than O on hardness due to the formation of a solid solution rather than precipitates. The hardness of copper, therefore, remains mainly unaffected by the addition of oxygen, but increases noticeably with the alloying of aluminium231, 233. This was investigated by nanoindentation, where copper in a pure alumina composite sample, copper in a sample with added copper oxide and pure copper after squeeze casting were all analysed. The hardness of copper decreased correspondingly, showing the highest values for the specimen where aluminate was present; hence the dissolution may lead to the alloying of copper with oxygen and aluminium. Oxygen promotes the wetting, hence improves the infiltration progress, and aluminium affects the hardness of copper in the composite structure. Further microstructural analysis appears to verify these observations, even if it was not of statistical relevance. A half-quantitative EDS investigation of the Al-content in the microstructure showed a possible decrease of aluminium content depending on the distance to the dissolved agglomerate (Figure 117).

191

2

3 1 40 nm

4 200 nm

1 2 3 4

Figure 117: EDS spectra of the copper phase in a composite sample with different distances to the decomposed aluminate agglomerate in STEM lamellae.

A STEM investigation was performed on the lamellae to ensure no influence from beneath the surface. Hence no alumina particle complicated the Al- content, as in the analysis of normal polished surfaces, where the area below the surface is always included in the sampling volume. The aluminium content in the surrounding melt seemed to reduce with distance from the decomposed aluminate. The Al-content is not influenced by a hidden alumina particle as the investigation has been made with STEM lamellae, as described. However, as some of the microscope’s components are made of aluminium, the EDS analysis might be influenced; but it shows the expected behaviour of aluminium reduction in the copper melt, even if not reliable in terms of giving the Al-content of the copper phase.

The infiltration progress in squeeze casting, including the solidification of the copper melt, is fast – a description of the composite in equilibrium condition, therefore, is not appropriate. Not all aluminate agglomerates are dissolved, as was shown in the microscopic characterisation. In contrast with that and with regard to the phase diagram in Figure 37, the aluminate phase disappeared 192

when adding copper to the system in equilibrium. Composite samples fabricated in isothermal infiltration by gas pressure showed no residual aluminate in the microstructure, as seen in Figure 83. The infiltration temperature of 1200°C was above the dissolution temperature of the 1170°C mentioned. The agglomerates seemed to dissolve entirely in the copper matrix, as no residual aluminate or lamellae-shaped copper/alumina resulting from the dissolution and precipitation could be found in the microstructure. The copper liquid can dissolve up to 2.1 at% oxygen at 1200°C as measured by Jacob and Alcock155 and shown in Figure 116. Preforms with 2 wt% copper oxide added in preform preparation contained about 3.4 wt% aluminate, the prerequisite being a full reaction of copper oxide and alumina. Assuming about 50% ceramic volume content and considering only the copper within the porosity of the preform and, hence, of the composite structure, the oxygen content was calculated at about 1.55 at% and the aluminium content at about 0.78 at% – well below the oxygen solubility limit of 2.1 at% at 1200°C infiltration temperature.

Taking the aforementioned analysis into account, the in situ alloying of copper during infiltration is an effect possibly contributing to enhanced infiltration. The dissolved oxygen may adsorb at the copper/alumina interface, leading to lowering of surface tension and strengthening of bonding. Aluminium dissolved in the copper contributes to higher hardness in the copper of composite samples containing copper oxide. Furthermore, it was shown in section 5.6.4 that the specific thermal conductivity of the copper phase in composite samples reduces with the content of copper oxide added in preform preparation. Considering the enhanced infiltration and bonding behaviour due to the copper oxide addition, the overall thermal conductivity is believed to be enhanced, in accordance with the work of several authors228-230. The observed decrease in specific conductivity of the copper phase can now be attributed to alloying with oxygen and, especially, aluminium. As oxygen influences the thermal conductivity of copper marginally, it is better to attribute the observed decrease in conductivity to alloying with aluminium, with its higher solubility of up to 9.4 wt% in copper232 and strong influence on conductivity, as reported by Nitzsche231.

193

6.4 Aluminate in the Composite Structure Aluminate appears in the majority of larger agglomerates within the microstructure. As discussed, the dissolution appears only in the area where infiltration starts; hence, only a few agglomerates will dissolve. Investigation by nanoindentation showed the aluminate phase to be brittle. This corresponds with the observations made by Yoshino and Ohtsu149, who also report brittle behaviour of the aluminate phase. As reported by Clyne and Withers86, brittle interfacial phases are detrimental to elastic and plastic properties of the composite. Both observations seem to be in contradiction with many authors5, 141, 149, 159, 161, who state that aluminate at the interface enhances bonding. This has also been investigated in this work in the manner of a model, as discussed, and shown in Figure 93. But Kim and Kim161 observe the bond strength of copper/alumina joints, with aluminate as the interfacial layer, to be thickness- dependent. The bond strength increases with the formation of aluminate, but decreases slightly as the thickness of aluminate increases further. They find the highest bond strength with bonding times of 20 h, resulting in aluminate layer thickness of 8 μm. However, the agglomerates investigated in this work are up to about 40 μm in diameter, well above the thickness responsible for the decrease in bond strength in the aforementioned study. The brittle behaviour of aluminate agglomerates was also observed in tensile investigations, as described in section 5.6.3 and shown in Figure 108 and Figure 109. As shown in Figure 101, the bending strength of the composite reduces slightly when adding more than about 1–2 wt% copper oxide in preform preparation. This effect was visualised in Figure 118, with two areas highlighted – the first showing a strong increase with the addition of very small amounts of copper oxide to the preform, followed by the second showing a decrease in overall bending strength for the addition of more than about 1-2 wt% Cu2O. Young’s modulus and residual porosity were also given for reference. Apart from the sample type with 5 wt% addition, all the copper oxide-containing powder compositions tested were spray-dried in the preparation process. Comparison of the drying processes showed little difference in pourability, expressed by the compaction ratio of powders during pressing, but exhibited the same green density and, hence, ceramic volume content after sintering. The bending

194

strength of the resulting composites was investigated for the sample type with

2 wt% Cu2O addition; exhibit strength of 808 MPa, about 5% higher strength for the spray-dried composition than for the freeze-dried composition with 767 MPa.

1 2 1000 250

800 200 ] 000 / 0 600 150

400 100

200 Bending strength 50 Residual porosity [ Residual porosity Young's modulus [GPa] modulus Young's

Bending strength [N/mm²] residual porosity Young's modulus 0 0 0123456

Content of Cu2O in preform [wt%] Figure 118: Mechanical strength, Young’s modulus and residual porosity of composites depending on the copper oxide content in the preform composition. Two areas are marked, increasing and decreasing strength, with Cu2O addition.

Residual porosity was evaluated by use of Archimedes’ principle. The compression of about 3-5% of preforms during infiltration was taken into account for calculating the residual porosity, where applicable. While the specimen type without copper oxide addition exhibited about 5.1% residual porosity, it reduced with the content of copper oxide, almost linearly, to 1.9% for

5 wt% Cu2O addition, showing again the improvement of infiltration ability. This might have been of relevance but could not be considered responsible for the entire improvement of bending strength, considering that with further improvement of infiltration and, hence, decrease of residual porosity in the composite structure with more than 2 wt% Cu2O addition, the bending strength decreased again. However, Young’s modulus increased with the addition of copper oxide in the preform preparation process, hence, with the reduction of

195

residual porosity. The increase of up to 2 wt% Cu2O addition was steeper than the increase following more copper oxide addition. This might have contributed to the increased amount of a third phase, aluminate, in the microstructure, with lower elastic modulus leading to a slight decrease in overall stiffness overlayed by the enhancement of better infiltration. Elastic modulus calculated by the simple rule of mixture showed a decrease of about 3.7% for the sample type with 5 wt% Cu2O added, if taking into consideration the reduced elastic modulus of the aluminate phase as evaluated in nanoindentation. Whereas the increase of bending strength might be explained by infiltration and improved bonding behaviour due to oxygen presence, the subsequent decrease of the bending strength could not be attributed to the further improvement of infiltration and, hence, decrease of residual porosity. A different effect resulting in weakening the composite structure for higher additions might have been present. Investigation into the surfaces of broken samples showed plastically-deformed copper ligaments and a brittle behaviour of alumina, as expected (Figure 112). The aluminate phase also showed brittle behaviour, which was investigated by nanoindentation and shown in tensile loading (Figure 109). This led to the proposition that the decrease in bending strength could be an effect of the agglomerates in the microstructure. With addition of more copper oxide, the average distance of the agglomerates reduced, resulting in a weaker overall composite structure. The early breakage of aluminate agglomerates if a load is applied has been stated. Sample types with pre-reacted aluminate promised better characteristics, as the aluminate particles were in the range of 1 μm, as analysed (Figure 34). A bending strength of a sample with about 3.4 wt% aluminate added, the equivalent of 2 wt% Cu2O (prerequisite a full reaction), measured at 761 MPa with a Weibull modulus of 18.6. The pre-reacted aluminate was prepared as described in section 5.1.1, as it is not commercially obtainable; so the preparation of powder composition was done in small batches, using the freeze-drying process. For comparison, the sample type with 2 wt% copper oxide prepared using the freeze-drying process rather than the spray-drying process had to be considered: strength was measured at 767 MPa with a Weibull modulus of 11.7. The strength, therefore, is comparable: but the reliability expressed by the Weibull modulus is better for the sample with pre-

196

reacted aluminate. This could contribute to the aforementioned aluminate agglomerate size and its distribution in the microstructure.

The large increase in bending strength for small additions of copper oxide appeared to correlate with the enhanced infiltration behaviour. The interface is the key issue in designing composite materials: with weak bonding between the phases the strength of the overall composite structure will be low. Several authors5, 141, 149, 159, 161 show the bonding between copper and alumina to be enhanced if oxygen is present. This has been explained by the formation of reaction products at the interface. Aluminate is one such reaction product. Rühle and co-workers19, 152, 171 extensively investigate the behaviour of copper/alumina interfaces by TEM observation and show, for example, small layers of aluminate to be responsible for enhanced adhesion in copper/alumina joints. Others5, 19, 151 report enhanced bending strength in copper/alumina joints when oxygen is present, but without the presence of an interfacial phase. Fracture surfaces investigated in this work, as shown in Figure 111 and Figure 112, showed better bonding of copper to alumina if aluminate was present in the microstructure. The aluminate phase was not necessarily present at the interface, as described. These observations were consistent with mechanical properties measured.

Considering these observations the effect of aluminate in the microstructure seems to be twofold. For small particles or thin layers the bonding will enhance, but large agglomerates of aluminate will weaken the overall strength by their brittle behaviour. Apart from this, oxygen at the interface enhances the bonding strength of copper to alumina; therefore, the effect of aluminate apparent in the preform structure shows an enhancement of the infiltration behaviour and improves the composites’ mechanical properties by stronger interface bonding.

Although in this work a clear interfacial phase was not observed in the composite samples, the bonding of copper to alumina was enhanced if copper oxide was added in preform preparation.

197

6.5 Hypotheses revisited The hypotheses given in chapter 3 are now revisited and addressed.

Reliable squeeze cast infiltration of alumina preforms with copper is possible due to providing better wetting properties and preventing premature solidification by raising the thermal energy of the infiltrating system.

The infiltration process is characterised by infiltration initiation and the progressing infiltration front. Oxygen dissolved in the copper melt leads to enhanced wetting properties; the infiltration initiation, expressed by reduction of threshold pressure is achieved, if oxygen is added to the infiltration alloy as has been shown in section 5.2. Furthermore, by the addition of copper oxide to the preform instead of alloying the infiltration melt, the overall infiltration progress is enhanced as also shown in that section. The reliable manufacturing of composite material has become possible by raising the thermal energy of the system due to the increase in tool temperature. Premature solidification can be prevented through raising the temperatures of the infiltration tool to 900°C, which has been possible by use of Ni-base materials.. Reliability of the infiltration process is expressed by an increase in Weibull modulus of the resulting composite as shown in section 5.6.1.

Oxygen improves the wetting properties of the system copper/alumina. Instead of using a Cu-O alloy, which is more difficult to handle, it is possible to provide the oxygen in the preform structure. If distributed homogeneously at the inner surfaces of the preform structure, it further ensures the oxygen being present where it is most useful: at the latter interface of the entire composite.

To provide oxygen in the preform structure copper oxide can be used. Adding copper oxide in the powder preparation process is simple and the oxygen therefore is implemented early in the manufacturing process into the ceramic structure. Sintering of the green bodies is limited in temperature due to reactions taking place (aluminate is formed); varying the ceramic volume

198

content, therefore, can only be achieved by use of artificial pore-forming agents and raw alumina powders with higher compressibility. The addition of copper oxide in fact enhances the overall infiltration progress. As has been shown in section 5.3.2 the copper oxide content of the preform affects the infiltration process (Figure 61). The distribution of copper oxide has been shown to be homogeneous within the alumina structure, and hence, at the latter interface of the composite material. Apart from small well distributed particles, the main fraction of copper oxide appears in relatively big agglomerates. Using pre- reacted aluminate instead of copper oxide, the distribution is finer but does not further improve the mechanical properties of the resulting composite material. With the aluminate phase in the preform structure the oxygen content is elevated and improves the infiltration behaviour, even if the wetting behaviour of copper on aluminate does not explain this effect as described by Diemer 5.

The interface of the composite structure is of aluminate, which is formed by the reaction of alumina and copper oxide, added in the powder preparation process. The aluminate phase leads to better adhesion of phases, enhancing the thermal conductivity and due to higher bond strength it results in improved mechanical properties of the entire composite material.

The overall interfacial bond strength of the composite materials is improved, when adding copper oxide in the powder preparation process. A continuous film of aluminate at the interface was not found. Investigations of the fracture behaviour (section 5.6.3) show better bond strength of the composites containing aluminate. As discussed previously (section 6.2.2) aluminate dissolves during infiltration and alloys in-situ; whereas the dissolved oxygen leads to better bond strength, the aluminium is responsible for higher hardness and reduction of thermal conductivity of the resulting composite (section 6.3). The majority of aluminate is present in large agglomerates with the remainder being well distributed in the microstructure. As has been measured in nano- indentation experiments the aluminate phase is brittle. Fractographic images

199

(Figure 109) show the interface of both, copper/aluminate and alumina/aluminate being stronger than the aluminate phase itself. The improving effect of aluminate, therefore, is dependent on the thickness which is in agreement with the observations made by Kim and Kim161. Nevertheless the bonding of copper to alumina is enhanced in composite materials containing aluminate with the overall mechanical properties being improved as well.

6.6 Summary The powder preparation with copper oxide addition for composite manufacturing resulted in preforms with aluminate agglomerates. The infiltration was enhanced with the presence of aluminate in the preform structure but not governed by infiltration initiation, as the microstructure remained mainly unaffected. However, the penetration of melt in those preform structures was enhanced.

This could be attributed to the dissolution of aluminate into copper liquid Cu(L1) and the precipitation of alumina, according to the ternary phase diagram in Figure 116. Copper can dissolve oxygen and aluminium, giving rise to different effects. While the aluminium content mainly affected copper hardness and thermal conductivity, oxygen promoted wetting behaviour, leading to enhanced infiltration and improves the bonding strength of copper to alumina. As the dissolution of aluminate was temperature-dependent, as in the phase diagrams in Figure 115, few agglomerates were dissolved, leaving the eutectically- structured alumina precipitates as shown in Figure 81 (e.g.). These regions in the microstructure were, therefore, only observed near the composite surface where infiltration had started. The possibility of an entire solution of aluminate agglomerates was shown by isothermal infiltration of a copper oxide-containing sample. Hence, the in situ alloying of copper melt due to the dissolution of aluminate agglomerates present in the microstructure seemed to be responsible for the improvement of infiltration behaviour and mechanical properties. As described in the literature5, 141, 149, 151, 159, 161, copper content and aluminate contribute to enhancing the bonding nature of copper to alumina, which agrees well with the observations in this work. Due to the brittle nature of aluminate and solubility of oxygen in copper, there are limits to the content of aluminate in the microstructure.

200

7 Conclusions This Chapter summarises the major findings of this thesis on the interface design of copper/alumina composites with an interpenetrating phase structure (preform-MMCs).

The major findings of this work can be summarised as follows:

1. Copper/alumina composites with interpenetrating phase structure can be reliably fabricated by squeeze cast infiltration. Copper oxide is added during preform fabrication at the powder preparation process. With this, typical the processing steps are slightly altered, but lead to significantly enhanced properties for small to moderate copper oxide additions.

2. Preforms with added copper oxide exhibit enhanced infiltration behaviour. This improvement is not governed by the infiltration initiation stage expressed by the threshold pressure, but leads to reliable infiltration processing of composite structure.

3. During sintering, the copper oxide (Cu2O) reacts with alumina (Al2O3)

and forms aluminate (CuAlO2). As sintering temperature is above the

Cu2O-Al2O3 eutectic temperature, the reaction is fast and no copper oxide is present after infiltration. The aluminate phase appears mainly in large agglomerates which are well distributed with the remainder being well dispersed in the microstructure.

4. Infiltration takes place with superheated copper melt leading to the dissolution of the aluminate agglomerates in the vicinity of preform

surface, where the temperature is still above the Cu2O-Al2O3 eutectic at 1170°C. Copper alloys in situ with oxygen and aluminium; alumina precipitates can be seen in the microstructure. The oxygen content in the melt promotes infiltration by the reduction of surface tension and enhances the bonding behaviour of copper to alumina. Aluminium dissolved in the copper liquid contributes to somewhat lower specific thermal conductivity and higher hardness in the copper phase.

201

5. The aluminate phase is characterised as being of brittle nature, which limits the amount used for addition, due to a weakening effect with high content in the microstructure.

6. Best results for reliable composite structure in terms of bending strength are found with the addition of between 1 wt% and 2 wt% copper oxide in powder preparation, or the equivalent amount of pre-reacted aluminate. The addition of 2 wt% copper oxide resulted in composite materials with mean strength of 808 MPa – more than twice the bending strength of the sample type without addition – with a Weibull modulus of 25.6.

A major increase in mechanical strength is shown for small to moderate additions of copper oxide. Fracture toughness can be slightly enhanced by the addition of pore-forming additive, providing large copper ligaments for crack bridging.

202

8 References 1 O. Lott, A. Nagel, J. Göhler, and M. Hoffman, "Interpenetrierende Kupfer- Aluminiumoxid Verbundwerkstoffe mit ausgezeichneter Festigkeit". in Verbundwerkstoffe und Werkstoffverbunde, Vol. 41. Edited by B. Wielage, Chemnitz, 2011. 2 J. Naidich, "Wettability of solids by liquid metals", pp. 354-484. in Progress in Surface and Membrane Science, Vol. 14. 1981. 3 C. Englisch, "Maßgeschneiderte Interfaces für Metall-Keramik Komposite", Dissertation, EPF Lausanne, 1999. 4 T. Young, "Phil. Transactions of the Royal Society London", pp. 3 - 98. in, Vol. 65. 1805. 5 M. Diemer, "Benetzungsverhalten und mechanische Eigenschaften der Cu/Al2O3-Grenzfläche", Dissertation, Technische Universität Darmstadt, Darmstadt 1998. 6 E. Saiz, A. P. Tomsia, and R. M. Cannon, "Ridging effects on wetting and spreading of liquids on solids", Acta Materialia, Vol. 46 [7] 2349-61 (1998). 7 E. Saiz, A. P. Tomsia, and R. M. Cannon, "Triple line ridging and attachment in high-temperature wetting", Scripta Materialia, Vol. 44 [1] 159-64 (2001). 8 M. E. R. Shanahan and A. Carre, "Viscoelastic Dissipation in Wetting and Adhesion Phenomena", Langmuir, Vol. 11 [4] 1396-402 (1995). 9 G. Liu and J. K. Shang, "Fracture Behavior of Interface-modified Metal / Ceramic Interfaces", Vol. (1996). 10 J. G. Li, "Wetting of Ceramic Materials by Liquid Silicon, Aluminium and Metallic Melts Containing Titanium and Other Reactive Elements: A Review", Ceramics International, Vol. 20 [6] 391-412 (1994). 11 D. Chatain, I. Rivollet, and N. Eustathopoulos, "Estimation of the thermodynamic adhesion and the contact-angle in the nonreactive metal- ionocovalent oxide systems", Journal De Chimie Physique Et De Physico-Chimie Biologique, Vol. 84 [2] 201-03 (1987). 12 B. M. Gallois, "Wetting in nonreactive liquid metal-oxide systems", Jom- Journal of the Minerals Metals & Materials Society, Vol. 49 [6] 48-51 (1997). 13 I. G. Batyrev and L. Kleinman, "In-plane relaxation of Cu(111) and Al(111)/alpha-Al2O3 (0001) interfaces", Physical Review B, Vol. 64 [3] art. no.-033410 (2001). 14 M. Diemer, A. Neubrand, K. P. Trumble, and J. Rodel, "Influence of oxygen partial pressure and oxygen content on the wettability in the copper-oxygen-alumina system", Journal of the American Ceramic Society, Vol. 82 [10] 2825-32 (1999).

203

15 T. E. Obrien and A. C. D. Chaklader, "Effect of Oxygen on Reaction between Copper and Sapphire", Journal of the American Ceramic Society, Vol. 57 [8] 329-32 (1974). 16 K. H. Johnson and S. V. Pepper, "Molecular-orbital model for metal- sapphire interfacial strength", Journal of Applied Physics, Vol. 53 [10] 6634-37 (1982). 17 J. G. Li, "Wetting and Interfacial Bonding of Metals with lonocovalent Oxides", Journal of the American Ceramic Society, Vol. 75 [11] 3118-26 (1992). 18 J. G. Li and H. Hausner, "Wetting and adhesion in liquid silicon/ceramic systems", Materials Letters, Vol. 14 [5-6] 329-32 (1992). 19 M. Rühle, "Struktur und chemische Zusammensetzung von inneren Grenzflächen in verschiedenen Materialien", Vol. 429 - 37 (2004). 20 W. Zhang, J. R. Smith, and A. G. Evans, "The connection between ab initio calculations and interface adhesion measurements on metal/oxide systems: Ni/Al2O3 and Cu/Al2O3", Acta Materialia, Vol. 50 [15] 3803-16 (2002). 21 E. Saiz, R. M. Cannon, and A. P. Tomsia, "High-temperature wetting and the work of adhesion in metal/oxide systems", Annual Review of Materials Research, Vol. 38 197-226 (2008). 22 N. Eustathopoulos and B. Drevet, "Interfacial bonding, wettability and reactivity on metal-oxide systems", Journal De Physique III, Vol. 4 [10] 1865-81 (1994). 23 M. D. Baldwin, P. R. Chidambaram, and G. R. Edwards, "Spreading and interlayer formation at the copper-cooper oxide polycrystalline alumina interface", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 25 [11] 2497-506 (1994). 24 J. Y. Kim, J. S. Hardy, and K. S. Weil, "Effects of CuO content on the wetting behavior and mechanical properties of a Ag-CuO braze for ceramic joining", Journal of the American Ceramic Society, Vol. 88 [9] 2521-27 (2005). 25 A. Meier, M. D. Baldwin, P. R. Chidambaram, and G. R. Edwards, "The effect og large oxygen additions on the wettability and work of adhesion of copper-oxygen alloys on polycrystalline Alumina", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 196 [1-2] 111-17 (1995). 26 A. M. Meier, P. R. Chidambaram, and G. R. Edwards, "A comparison of the wettability of copper-copper oxide and silver-copper oxide on polycrystalline Alumina", Journal of Materials Science, Vol. 30 [19] 4781- 86 (1995).

204

27 A. C. D. Chaklader, A. M. Armstrong, and S. K. Misra, "Interface reactions between metals and ceramics .4. wetting of sapphire by liquid copper-oxygen alloys", Journal of the American Ceramic Society, Vol. 51 [11] 630-& (1968). 28 D. A. Mortimer and M. Nicholas, "The wetting of carbon by copper and copper alloys", Journal of Materials Science, Vol. 5 [2] 149 (1970). 29 N. Eustathopoulos, D. Chatain, and L. Coudurier, "Wetting and interfacial chemistry in liquid metal-ceramic systems", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 135 83-88 (1991). 30 P. Kritsalis, L. Coudurier, and N. Eustathopoulos, "Contribution to the study of reactive wetting in the CuTi/AI203 system", Journal of Materials Science, Vol. 26 [12] 3400-08 (1991). 31 J. G. Li, "Kinetics of wetting and spreading of Cu-Ti alloys on alumina and glassy carbon substrates", Journal of Materials Science Letters, Vol. 11 [23] 1551-54 (1992). 32 A. Meier, P. R. Chidambaram, and G. R. Edwards, "Modelling of the spreading kinetics of reactive brazing alloys on ceramic substrates: Copper-titanium alloys on polycrystalline alumina", Acta Materialia, Vol. 46 [12] 4453-67 (1998). 33 R. Standing and M. Nicholas, "The wetting of alumina and vitreous carbon by copper--titanium alloys", Journal of Materials Science, Vol. 13 [7] 1509-14 (1978). 34 R. E. Loehman and A. P. Tomsia, "Reactions of Ti and Zr with AlN and Al203", Acta Metallurgica Et Materialia, Vol. 40 S75-S83 (1992). 35 A. Mortensen and I. Jin, "Solidification processing of metal matrix composites", International Materials Reviews, Vol. 37 [3] 101-28 (1992). 36 P. Shen, H. Fujii, T. Matsumoto, and K. Nogi, "Influence of substrate crystallographic orientation on the wettability and adhesion of alpha- Al(2)O(3)single crystals by liquid Al and Cu", Journal of Materials Science, Vol. 40 [9-10] 2329-33 (2005). 37 P. Vikner, "DEA Report ", Materials Science, Vol. 40 2 (1993). 38 P. D. Ownby and J. Liu, "Surface-energy of liquid copper and single- crystal sapphire and the wetting behavior of copper on sapphire", Journal of Adhesion Science and Technology, Vol. 2 [4] 255-69 (1988). 39 V. Ghetta, J. Fouletier, and D. Chatain, "Oxygen adsorption isotherms at the surfaces of liquid Cu and Au-Cu alloys and their interfaces with Al2O3 detected by wetting experiments", Acta Materialia, Vol. 44 [5] 1927-36 (1996). 40 M. Backhaus-Ricoult, "Gibbs' adsorption at alpha alumina-copper interfaces", Journal of the European Ceramic Society, Vol. 23 [15] 2747- 59 (2003).

205

41 I. A. Aksay, C. E. Hoge, and J. A. Pask, "Wetting under chemical- equilibrium and nonequilibrium conditions", Journal of Physical Chemistry, Vol. 78 [12] 1178-83 (1974). 42 V. Laurent, D. Chatain, and N. Eustathopoulos, "Wettability of SiO2 and oxidized SiC by aluminium", Materials Science and Engineering a- Structural Materials Properties Microstructure and Processing, Vol. 135 89-94 (1991). 43 K. Landry, C. Rado, R. Voitovich, and N. Eustathopoulos, "Mechanisms of reactive wetting: The question of triple line configuration", Acta Materialia, Vol. 45 [7] 3079-85 (1997). 44 L. Espie, B. Drevet, and N. Eustathopoulos, "Experimental Study of the Influence of Interfacial Energies and Reactivity on Wetting in Metal/Oxide Systems", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 25 [3] 599-605 (1994). 45 P. Kritsalis, B. Drevet, N. Valignat, and N. Eustathopoulos, "Wetting transitions in reactive Metal/Oxide systems", Scripta Metallurgica Et Materialia, Vol. 30 [9] 1127-32 (1994). 46 X. B. Zhou and J. T. M. DeHosson, "Reactive wetting of liquid metals on ceramic substrates", Acta Materialia, Vol. 44 [2] 421-26 (1996). 47 P. Shen, H. Fujii, T. Matsumoto, and K. Nogi, "Reactive wetting of SiO2 substrates by molten Al", Metallurgical and Materials Transactions a- Physical Metallurgy and Materials Science, Vol. 35A [2] 583-88 (2004). 48 C. Garcia-Cordovilla, E. Louis, and J. Narciso, "Pressure infiltration of packed ceramic particulates by liquid metals", Acta Materialia, Vol. 47 [18] 4461-79 (1999). 49 J. M. Molina, R. Prieto, M. Duarte, J. Narciso, and E. Louis, "On the estimation of threshold pressures in infiltration of liquid metals into particle preforms", Scripta Materialia, Vol. 59 [2] 243-46 (2008). 50 J. M. Molina, R. Arpon, R. A. Saravanan, C. Garcia-Cordovilla, E. Louis, and J. Narciso, "Threshold pressure for infiltration and particle specific surface area of particle compacts with bimodal size distributions", Scripta Materialia, Vol. 51 [6] 623-27 (2004). 51 N. A. Travitzky and A. Shlayen, "Microstructure and mechanical properties of Al2O3/Cu-O composites fabricated by pressureless infiltration technique", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 244 [2] 154-60 (1998). 52 M. Bahraini, J. M. Molina, M. Kida, L. Weber, J. Narciso, and A. Mortensen, "Measuring and tailoring capillary forces during liquid metal infiltration", Current Opinion in Solid State & Materials Science, Vol. 9 [4- 5] 196-201 (2005).

206

53 M. Bahraini, L. Weber, J. Narciso, and A. Mortensen, "Wetting in infiltration of alumina particle preforms with molten copper", Journal of Materials Science, Vol. 40 [9-10] 2487-91 (2005). 54 S. Cardinal, M. R'Mili, and P. Merle, "Improvement of high pressure infiltration behaviour of alumina platelet preforms: manufacture and characterization of hybrid preforms", Composites Part a-Applied Science and Manufacturing, Vol. 29 [11] 1433-41 (1998). 55 A. Mortensen, "Interfacial phenomena in the solidification processing of metal matrix composites", Materials Science and Engineering a- Structural Materials Properties Microstructure and Processing, Vol. 135 1-11 (1991). 56 R. Asthana and P. K. Rohatgi, "Melt Infiltration of Compacts", Zeitschrift für Metallkunde, Vol. 83 [12] 887-92 (1992). 57 J. M. Molina, J. Narciso, and E. Louis, "On the triple line in infiltration of liquid metals into porous preforms", Scripta Materialia, Vol. 62 [12] 961- 65 (2010). 58 T. Dopler, A. Modaressi, and V. Michaud, "Simulation of metal-matrix composite isothermal infiltration processing", Metallurgical and Materials Transactions B-Process Metallurgy and Materials Processing Science, Vol. 31 [2] 225-34 (2000). 59 V. J. Michaud, L. M. Compton, and A. Mortensen, "Capillarity in Isothermal Infiltration of Alumina Fiber Preforms with Aluminum", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 25 [10] 2145-52 (1994). 60 A. Mortensen and V. Michaud, "Infiltration of Fibre Preforms by a Binary Alloy: Part I Theory", Metallurgical Transactions a-Physical Metallurgy and Materials Science, Vol. 21 [7] 2059-72 (1990). 61 J. L. Hilden and K. P. Trumble, "Numerical analysis of capillarity in packed spheres: Planar hexagonal-packed spheres", Journal of Colloid and Interface Science, Vol. 267 [2] 463-74 (2003). 62 P. C. Carman, "Capillary rise and capillary movement of moisture in fine sands", Soil Sci., Vol. 52 [1] 1-14 (1941). 63 L. R. White, "Capillary Rise in Powders", Journal of Colloid and Interface Science, Vol. 90 [2] 536-38 (1982). 64 W. R. Purcell, "Interpretation of capillary pressure data", Petroleum Transactions of AIME, Vol. 189 369-71 (1950). 65 R. P. Mayer and R. A. Stowe, "Mercury Porosimetry-Breakthrough pressure for penetration between packed spheres", Journal of Colloid Science, Vol. 20 [8] 893-& (1965). 66 G. Mason and N. R. Morrow, "Effect of Contact Angle on Capillary Displacement Curvatures in Pore Throats Formed by Spheres", Journal of Colloid and Interface Science, Vol. 168 [1] 130-41 (1994).

207

67 A. Marmur, "Capillary Rise and Hysteresis in Periodic Porous Media", Journal of Colloid and Interface Science, Vol. 129 [1] 278-85 (1989). 68 T. Mathews, G. P. Matthews, and S. Huggett, "Estimating particle size distributions from a network model of porous media", Powder Technology, Vol. 104 [2] 169-79 (1999). 69 M. Spearing and G. P. Matthews, "Modelling Characteristic Properties of Sandstones", Transport in Porous Media, Vol. 6 [1] 71-90 (1991). 70 M. Hilpert, "Effects of dynamic contact angle on liquid infiltration into horizontal capillary tubes: (Semi)-analytical solutions", Journal of Colloid and Interface Science, Vol. 337 [1] 131-37 (2009). 71 M. Hilpert, "Effects of dynamic contact angle on liquid infiltration into inclined capillary tubes: (Semi)-analytical solutions", Journal of Colloid and Interface Science, Vol. 337 [1] 138-44 (2009). 72 E. W. Washburn, "The dynamics of capillary flow", Phys. Rev., Vol. 17 [3] 273-83 (1921). 73 P. Joos, P. Vanremoortere, and M. Bracke, "The Kinetics of Wetting in a Capillary", Journal of Colloid and Interface Science, Vol. 136 [1] 189-97 (1990). 74 R. G. Cox, "The dynamics of the spreading of liquids on a solid surface. Part 1. Viscous flow", Journal of Fluid Mechanics, Vol. 168 169-94 (1986). 75 J. De Coninck and T. D. Blake, "Wetting and molecular dynamics simulations of simple liquids", Annual Review of Materials Research, Vol. 38 1-22 (2008). 76 J. Bear, "Dynamics of Fluids in Porous Media". Dover Publications, (1988). 77 A. Mortensen, L. J. Masur, J. A. Cornie, and M. C. Flemings, "Infiltration of Fibrous Preforms by a Pure Metal: Part 1. Theory", Metallurgical Transactions a-Physical Metallurgy and Materials Science, Vol. 20 [11] 2535-47 (1989). 78 V. J. Michaud, J. L. Sommer, and A. Mortensen, "Infiltration of fibrous preforms by a pure metal: Part V. Influence of preform compressibility", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 30 [2] 471-82 (1999). 79 J. L. Sommer and A. Mortensen, "Forced unidirectional infiltration of deformable porous media", Journal of Fluid Mechanics, Vol. 311 193- 217 (1996). 80 E. Lacoste, M. Aboulfatah, M. Danis, and F. Girot, "Numerical-Simulation of the Infiltration of Fibrous Performs by a Pure Metal", Metallurgical Transactions a-Physical Metallurgy and Materials Science, Vol. 24 [12] 2667-78 (1993).

208

81 X. Tong and J. A. Khan, "Infiltration and solidification remelting of a pure metal in a two-dimensional porous preform", Journal of Heat Transfer- Transactions of the Asme, Vol. 118 [1] 173-80 (1996). 82 H. Kaufmann and A. Mortensen, "Wetting of SAFFIL Alumina Fiber Preforms by Aluminum at 973 K", Metallurgical Transactions a-Physical Metallurgy and Materials Science, Vol. 23 [7] 2071-73 (1992). 83 R. H. Brooks and A. T. Corey, "Hydraulic properties of porous media", Hydrology papers, Vol. 3 (1964). 84 M. T. Vangenuchten, "A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils", Soil Science Society of America Journal, Vol. 44 [5] 892-98 (1980). 85 K. P. Trumble, "Spontaneous infiltration of non-cylindrical porosity: Close-packed spheres", Acta Materialia, Vol. 46 [7] 2363-67 (1998). 86 T. W. Clyne and P. J. Withers, "An introduction to metal matrix composites". in Cambridge solid state science series. Edited by E. A. Davis and I. M. Ward. Cambridge University Press, Cambridge, 1993. 87 T. W. Clyne, "Metal Matrix Composites: Matrices and Processing". in Encyclopedia of Materials: Science and Technologie. Edited by A. Mortensen. Elsevier, 2001. 88 P. K. Rohatgi, "Cast Metal Matrix Composites: Past, Present and Future", American Foundry Society, Silver Anniversary Paper, Vol. 2 1- 25 (2001). 89 W. Ames and A. T. Alpas, "Wear mechanisms in hybrid composites of graphite-20 PCT SiC in A356 aluminium-alloy (Al-7 PCT Si-0.3 PCT Mg)", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 26 [1] 85-98 (1995). 90 P. N. Bindumadhavan, H. K. Wah, and O. Prabhakar, "Dual particle size (DPS) composites: effect on wear and mechanical properties of particulate metal matrix composites", Wear, Vol. 248 [1-2] 112-20 (2001). 91 D. B. Miracle, "Metal matrix composites - From science to technological significance", Composites Science and Technology, Vol. 65 [15-16] 2526-40 (2005). 92 A. Mortensen and J. Llorca, "Metal Matrix Composites", Annual Review of Materials Research, Vol. 40 243-70 (2010). 93 W. C. Harrigan, "Commercial processing of metal matrix composites", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 244 [1] 75-79 (1998). 94 H. Hegeler, R. Buschmann, and I. Elstner, "Herstellung von faserverstärkten Leichtmetallen unter Benutzung von faserkeramischen Formkörpern (Preforms) ", Vol. 101-16 (1994). 95 US Patent 2,765,512. "Formation of Ceramic, etc., Articles" R. A. Nesbit (1956).

209

96 DE Patent 32 11 083 C2. "Schlicker-Gießverfahren" G. Q. Weaver and B. G. Nelson (1983). 97 K. Araki and J. W. Halloran, "Porous ceramic bodies with interconnected pore channels by a novel freeze casting technique", Journal of the American Ceramic Society, Vol. 88 [5] 1108-14 (2005). 98 Y. Chino and D. C. Dunand, "Directionally freeze-cast titanium foam with aligned, elongated pores", Acta Materialia, Vol. 56 [1] 105-13 (2008). 99 S. Deville, E. Saiz, R. K. Nalla, and A. P. Tomsia, "Freezing as a path to build complex composites", Science, Vol. 311 [5760] 515-18 (2006). 100 T. Moritz and H. J. Richter, "Ceramic bodies with complex geometries and ceramic shells by freeze casting using ice as mold material", Journal of the American Ceramic Society, Vol. 89 [8] 2394-98 (2006). 101 T. Waschkies, "Untersuchung der Einflussfaktoren auf die Mikrostrukturausbildung beim Gefriergießen (Freeze-Casting)", Dissertation, Universitätsverlag Karlsruhe, Karlsruhe 2009. 102 A. Mattern, "Interpenetrierende Metall-Keramik-Verbundwerkstoffe mit isotropen und anisotropen Al2O3-Verstärkungen", Dissertation, University of Karlsruhe, Karlsruhe 2004. 103 S. Roy, "Metal/ceramic composites from freeze-cast preforms domain structure and mechanical properties", Dissertation, Universität Karlsruhe, Karlsruhe 2009. 104 M. Knechtel, H. Prielipp, H. Mullejans, N. Claussen, and J. Rodel, "Mechanical-properties of Al/al2O3 and Cu/Al2O3 composites with interpenetrating networks", Scripta Metallurgica Et Materialia, Vol. 31 [8] 1085-90 (1994). 105 H. Prielipp, M. Knechtel, N. Claussen, S. K. Streiffer, H. Mullejans, M. Ruhle, and J. Rodel, "Strength and fracture-toughness of Aluminum Alumina composites with interpenetrating networks", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 197 [1] 19-30 (1995). 106 S. Skirl, M. Hoffman, K. Bowman, S. Wiederhorn, and J. Rodel, "Thermal expansion behavior and macrostrain of Al2O3/Al composites with interpenetrating networks", Acta Materialia, Vol. 46 [7] 2493-99 (1998). 107 F. F. Lange, B. V. Velamakanni, and A. G. Evans, "Method for Processing Metal-Reinforced Ceramic Composites", Journal of the American Ceramic Society, Vol. 73 [2] 388-93 (1990). 108 S. F. Corbin, J. Lee, and X. Qiao, "Influence of green formulation and pyrolyzable particulates on the porous microstructure and sintering characteristics of tape cast ceramics", Journal of the American Ceramic Society, Vol. 84 [1] 41-47 (2001). 109 Y. S. Xu and D. D. L. Chung, "Low-volume-fraction particulate preforms for making metal-matrix composites by liquid metal infiltration", Journal of Materials Science, Vol. 33 [19] 4707-09 (1998).

210

110 S. F. Corbin and P. S. Apte, "Engineered porosity via tape casting, lamination and the percolation of pyrolyzable particulates", Journal of the American Ceramic Society, Vol. 82 [7] 1693-701 (1999). 111 J. B. Davis, A. Kristoffersson, E. Carlstrom, and W. J. Clegg, "Fabrication and crack deflection in ceramic laminates with porous interlayers", Journal of the American Ceramic Society, Vol. 83 [10] 2369-74 (2000). 112 A. W. Urquhart, "Novel reinforced ceramics and metals - A review of lanxide composite technologies", Materials Science and Engineering a- Structural Materials Properties Microstructure and Processing, Vol. 144 75-82 (1991). 113 J. Michalski, I. Wejrzanowski, S. Gierlotka, J. Bielinski, K. Konopka, T. Kosmac, and K. J. Kurzydlowski, "The preparation and structural characterization of Al2O3/Ni-P composites with an interpenetrating network", Journal of the European Ceramic Society, Vol. 27 [2-3] 831-36 (2007). 114 R. Zhang, L. Gao, and J. Guo, "Effect of Cu2O on the fabrication of SiCp/Cu nanocomposites using coated particles and conventional sintering", Composites Part a-Applied Science and Manufacturing, Vol. 1301-05 (2004). 115 Y. J. Lin and T. H. Shen, "Pressureless infiltration of porous Al2O3/Ni compacts with 6061 Al alloy", pp. 37-44. in Composite Materials III, Vol. 249. Key Engineering Materials. Edited by L. Zhang, J. K. Guo, andW. H. Tuan, 2003. 116 D. J. Lloyd, "Particle reinforced aluminium and magnesium matrix composites", International Materials Reviews, Vol. 39 [11] 1-23 (1994). 117 D. Coupard, J. Goni, and J. F. Sylvain, "Fabrication and squeeze casting infiltration of graphite/alumina preforms", Journal of Materials Science, Vol. 34 [21] 5307-13 (1999). 118 A. Mattern, B. Huchler, D. Staudenecker, R. Oberacker, A. Nagel, and M. J. Hoffmann, "Preparation of interpenetrating ceramic-metal composites", Journal of the European Ceramic Society, Vol. 24 [12] 3399-408 (2004). 119 S. Long, O. Beffort, G. Moret Uzwil, and P. Thevoz, "Processing of Al- based MMCs by Indirect Squeeze Infiltration of Ceramic Preforms on a Shot-Control High Pressure Die Casting Machine", Aluminium Vol. 76 82-89 (2000). 120 S. Long, Z. Zhang, and H. M. Flower, "Characterization of liquid metal infiltration of a chopped fibre preform aided by external pressure - II. Modelling of liquid metal infiltration process", Acta Metallurgica Et Materialia, Vol. 43 [9] 3489-98 (1995). 121 S. Nagata and K. Matsuda, "Effects of some Factors on the Critical Preheating Temperature of Particles in Producing Metal-Particle Composites by Pressure Casting", Journal of the Japan Foundrymen Society, Vol. 53 [12] 35 - 43 (1981).

211

122 M. R. Ghomashchi and A. Vikhrov, "Squeeze casting: an overview", Journal of Materials Processing Technology, Vol. 101 [1-3] 1-9 (2000). 123 S. Catterjee and A. A. Daas, "Effects of Pressure on Solidification of some comercial Aluminium-Base Casting Alloys", The British Foundryman, Vol. (1972). 124 J. W. Kaczmar, K. Pietrzak, and W. Wlosinski, "The production and application of metal matrix composite materials", Journal of Materials Processing Technology, Vol. 106 [1-3] 58-67 (2000). 125 J. Campbell, "The concept of net shape for castings", Materials & Design, Vol. 21 [4] 373-80 (2000). 126 J. Campbell, "Castings - Second edition". Elsevier Butterworth- Heinemann: Oxford, (2003). 127 J. H. Lee, H. S. Kim, S. I. Hong, C. W. Won, S. S. Cho, and B. S. Chun, "Effect of die geometry on the microstructure of indirect squeeze cast and gravity die cast 5083 wrought Al alloy and numerical analysis of the cooling behavior", Journal of Materials Processing Technology, Vol. 96 [1-3] 188-97 (1999). 128 H. Chattopadhyay, "Simulation of transport processes in squeeze casting", Materials Processing Technology, Vol. 186 174- 78 (2007). 129 T. Donomoto, K. Funatani, N. Miura, and N. Miyake, "Ceramic Fiber Reinforced Piston for High Performance Diesel Engines" SAE Technical Papers, Paper No. 830252 (1983). 130 G. R. Cappleman, J. F. Watts, and T. W. Clyne, "The interface region in squeeze-infiltrated composites containing delta-alumina fiber in an aluminium matrix", Journal of Materials Science, Vol. 20 [6] 2159-68 (1985). 131 C. G. Levi, G. J. Abbaschian, and R. Mehrabian, "Interface Interactions During Fabrication of Aluminum Alloy-Alumina Fiber Composites", Metallurgical Transactions a-Physical Metallurgy and Materials Science, Vol. 9 [5] 697-711 (1978). 132 S. Long, Z. Zhang, and H. M. Flower, "Characterization of Liquid Metal Infiltration of a Chopped Fibre Preform Aided by External Pressure- III. Model validation and quantification", Acta Materialia, Vol. 44 [10] 4233- 40 (1996). 133 T. Etter, J. Kuebler, T. Frey, P. Schulz, J. F. Loffler, and P. J. Uggowitzer, "Strength and fracture toughness of interpenetrating graphite/aluminium composites produced by the indirect squeeze casting process", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 386 [1-2] 61-67 (2004). 134 K. Prakasan, S. Palaniappan, and S. Seshan, "Thermal expansion characteristics of cast Cu based metal matrix composites", Composites Part a-Applied Science and Manufacturing, Vol. 28 1019 - 22 (1997).

212

135 H. W. Xing, X. M. Cao, W. P. Hu, L. Z. Zhao, and J. S. Zhang, "Interfacial reactions in 3D-SiC network reinforced Cu-matrix composites prepared by squeeze casting", Materials Letters, Vol. 59 [12] 1563-66 (2005). 136 J. M. Molina, M. Bahraini, L. Weber, and A. Mortensen, "Direct measurement of drainage curves in infiltration of SiC particle preforms: influence of interfacial reactivity", Journal of Materials Science, Vol. 43 [15] 5061-67 (2008). 137 E. Carreno-Morelli, T. Cutard, R. Schaller, and C. Bonjour, "Processing and characterization of aluminium-based MMCs produced by gas pressure infiltration", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 251 [1-2] 48-57 (1998). 138 A. J. Cook and P. S. Werner, "Pressure infiltration casting of metal matrix composites", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 144 189-206 (1991). 139 V. Michaud and A. Mortensen, "Infiltration processing of fibre reinforced composites: governing phenomena", Composites Part a-Applied Science and Manufacturing, Vol. 32 [8] 981-96 (2001). 140 T. R. Jonas, J. A. Cornie, and K. C. Russell, "Iinfiltration and wetting of Alumina particulate preforms by Aluminum and Aluminum-Magnesium alloys", Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, Vol. 26 [6] 1491-97 (1995). 141 C. Beraud, M. Courbiere, C. Esnouf, D. Juve, and D. Treheux, "Study of copper alumina bonding", Journal of Materials Science, Vol. 24 [12] 4545-54 (1989). 142 G. Dehm, C. Scheu, G. Mobus, R. Brydson, and M. Ruhle, "Synthesis of analytical and high-resolution transmission electron microscopy to determine the interface structure of Cu/Al2O3", Ultramicroscopy, Vol. 67 [1-4] 207-17 (1997). 143 G. Dehm, C. Scheu, R. Raj, and M. Ruhle, "Growth, structure and interfaces of Cu and Cu/Ti thin films on (0001)alpha-Al2O3", pp. 217-20. in Intergranular and Interphase Boundaries in Materials, Pt 1, Vol. 207-. Materials Science Forum. Edited by A. C. Ferro, J. P. Conde, andM. A. Fortes, 1996. 144 G. Dehm, C. Scheu, M. Ruhle, and R. Raj, "Growth and structure of internal Cu/Al2O3 and Cu/Ti/Al2O3 interfaces", Acta Materialia, Vol. 46 [3] 759-72 (1998). 145 Q. L. Guo and P. J. Moller, "On the thermal stability of copper deposits on a (0001) sapphire surface", Surface Science, Vol. 244 [3] 228-36 (1991). 146 J. A. Kelber, C. Y. Niu, K. Shepherd, D. R. Jennison, and A. Bogicevic, "Copper wetting of alpha-Al2O3(0001): theory and experiment", Surface Science, Vol. 446 [1-2] 76-88 (2000).

213

147 C. Niu, K. Shepherd, D. Martini, J. Tong, J. A. Kelber, D. R. Jennison, and A. Bogicevic, "Cu interactions with alpha-Al2O3(0001): effects of surface hydroxyl groups versus dehydroxylation by Ar-ion sputtering", Surface Science, Vol. 465 [1-2] 163-76 (2000). 148 G. L. Zhao, J. R. Smith, J. Raynolds, and D. J. Srolovitz, "First-principles study of the alpha-Al2O3(0001)/Cu(111) interface", Interface Science, Vol. 3 [4] 289-302 (1996). 149 Y. Yoshino and H. Ohtsu, "Interface Structure and Bond Strength of Copper-Bonded Alumina Substrates", Journal of the American Ceramic Society, Vol. 74 [9] 2184-88 (1991). 150 Y. Yoshino and T. Shibata, "Structure and Bond Strength of a Copper- Alumina interface", Journal of the American Ceramic Society, Vol. 75 [10] 2756-60 (1992). 151 Y. Yoshino, "Role of Oxygen in Bonding Copper to Alumina", Journal of the American Ceramic Society, Vol. 72 [8] 1322-27 (1989). 152 C. Scheu, M. Gao, S. H. Oh, G. Dehm, S. Klein, A. P. Tomsia, and M. Ruhle, "Bonding at copper-alumina interfaces established by different surface treatments: a critical review", Journal of Materials Science, Vol. 41 [16] 5161-68 (2006). 153 K. van Benthem, S. Kramer, W. Sigle, and M. Ruhle, "Structural and chemical analysis of materials with high spatial resolution", Mikrochimica Acta, Vol. 138 [3-4] 181-93 (2002). 154 H. Hahn and C. Delorent, "Untersuchungen über ternäre Chalkogenide. 7. Versuche zur Darstellung ternärer Oxyde des Aluminiums, Galliums und Indiums mit einwertigem Kupfer und Silber", Zeitschrift Fur Anorganische Und Allgemeine Chemie, Vol. 279 [5-6] 281-88 (1955). 155 K. T. Jacob and C. B. Alcock, "Thermodynamics of CuAlO2 and CuAl2O4 and phase-equilibria in system Cu2O-CuO-Al2O3", Journal of the American Ceramic Society, Vol. 58 [5-6] 192-95 (1975). 156 K. P. Trumble, "Thermodynamic analysis of aluminate formation at Fe/Al2O3 and Cu/Al2O3 interfaces", Acta Metallurgica Et Materialia, Vol. 40 S105-S10 (1992). 157 B. J. Dalgleish, E. Saiz, A. P. Tomsia, R. M. Cannon, and R. O. Ritchie, "Interface formation and strength in ceramic-metal systems", Scripta Metallurgica Et Materialia, Vol. 31 [8] 1109-14 (1994). 158 L. Esposito, A. Bellosi, S. Guicciardi, and G. de Portu, "Solid state bonding of Al2O3 with Cu, Ni and Fe: characteristics and properties", Journal of Materials Science, Vol. 33 [7] 1827-36 (1998). 159 I. E. Reimanis, K. P. Trumble, K. A. Rogers, and B. J. Dalgleish, "Influence of Cu2O and CuAlO2 interphases on crack propagation at Cu/alpha-Al2O3 interfaces", Journal of the American Ceramic Society, Vol. 80 [2] 424-32 (1997).

214

160 D. Treheux, P. Lourdin, B. Mbongo, and D. Juve, "Metal-Ceramic solid- state bonding - mechanisms and mechanics", Scripta Metallurgica Et Materialia, Vol. 31 [8] 1055-60 (1994). 161 S. T. Kim and C. H. Kim, "Interfacial reaction product and its effect on the strength of copper to alumina eutectic bonding", Journal of Materials Science, Vol. 27 [8] 2061-66 (1992). 162 Y. S. Sun and J. C. Driscoll, "A New Hybrid Power Technique Utilizing a Direct Copper to Ceramic Bond", Ieee Transactions on Electron Devices, Vol. 23 [8] 961-67 (1976). 163 M. G. Nicholas, "Joining of Ceramics". Chapman & Hall: London, (1990). 164 M. G. Nicholas and D. A. Mortimer, "Ceramic/metal joining for structural applications", Mater. Sci. Technol., Vol. 1 [9] 657-65 (1985). 165 B. Gibbesch and G. Elssner, "Ultra high-vacuum diffusion bonded Nb- Al2O3 and Cu-Al2O3 joints - The role of welding temperature and sputter cleaning", Acta Metallurgica Et Materialia, Vol. 40 S59-S66 (1992). 166 S. K. Misra and A. C. D. Chaklader, "The System Copper Oxide- Alumina", Journal of the American Ceramic Society, Vol. 46 [10] 509-09 (1963). 167 J. P. Neumann, T. Zhong, and Y. A. Chang, "Copper binary alloy phase diagrams, Cu-O (Copper - Oxygen)". in ASM Handbooks Online, http://www.asmmaterials.info, Vol. 3. Edited by H. Baker. ASM International, 1992. 168 K. A. Rogers, K. P. Trumble, B. J. Dalgleish, and I. E. Reimanis, "Role of Oxygen in Microstructure Development at SoIid-State Diffusion-Bonded Cu/a- AI203 Interfaces", Journal of the American Ceramic Society, Vol. 77 [8] 2036-42 (1994). 169 T. Fujimura and S. I. Tanaka, "In-situ high temperature X-ray diffraction study of Cu/Al2O3 interface reactions", Acta Materialia, Vol. 46 [9] 3057- 61 (1998). 170 S. Yi, K. P. Trumble, and D. R. Gaskell, "Thermodynamic analysis of aluminate stability in the eutectic bonding of copper with alumina", Acta Materialia, Vol. 47 [11] 3221-26 (1999). 171 C. Scheu, S. Klein, A. P. Tomsia, and M. Ruhle, "Chemical reactions and morphological stability at the Cu/Al2O3 interface", Journal of Microscopy- Oxford, Vol. 208 11-17 (2002). 172 M. Hasegawa, S. J. Zhu, Y. Kagawa, and A. G. Evans, "Effect of metal layer thickness on the decohesion of high purity copper-sapphire interfaces", Acta Materialia, Vol. 51 [17] 5113-21 (2003). 173 D. R. Lide, "CRC Handbook of Chemistry and Physics", Vol. 86. CRC Press: Boca Raton, FL, (2005).

215

174 A. S. Reddy, P. S. Reddy, S. Uthanna, and G. M. Rao, "Characterization of CuAlO2 films prepared by dc reactive magnetron sputtering", Journal of Materials Science-Materials in Electronics, Vol. 17 [8] 615-20 (2006). 175 J. H. Shy and B. H. Tseng, "Characterization of CuAlO2 thin film prepared by rapid thermal annealing of an Al2O3/Cu2O/sapphire structure", Journal of Physics and Chemistry of Solids, Vol. 66 [11] 2123- 26 (2005). 176 H. Kawazoe, M. Yasukawa, H. Hyodo, M. h. Kurita, H. Yanagi, and H. Hosono, "P-type electrical conduction in transparent thin films of CuAlO2", Nature, Vol. 389 [6654] 939-42 (1997). 177 A. Buljan, M. Llunell, E. Ruiz, and P. Alemany, "Color and conductivity in Cu2O and CuAlO2: A theoretical analysis of d(10)center dot center dot center dot d(10) interactions in solid-state compounds", Chemistry of Materials, Vol. 13 [2] 338-44 (2001). 178 K. Koumoto, H. Koduka, and W. S. Seo, "Thermoelectric properties of single crystal CuAlO2 with a layered structure", Journal of Materials Chemistry, Vol. 11 [2] 251-52 (2001). 179 K. Park, K. Y. Ko, H. C. Kwon, and S. Nahm, "Improvement in thermoelectric properties of CuAlO2 by adding Fe2O3", Journal of Alloys and Compounds, Vol. 437 [1-2] 1-6 (2007). 180 K. Park, K. Y. Ko, and W. S. Seo, "Thermoelectric properties of CuAlO2", Journal of the European Ceramic Society, Vol. 25 [12] 2219-22 (2005). 181 S. V. Pepper, "Shear strength of metal-sapphire contacts", Journal of Applied Physics, Vol. 47 [3] 801-08 (1976). 182 K. P. Trumble and M. Rühle, "The Thermodynamics of Spinel Interphase Formation at Diffusion-Bonded Ni/Al203 Interfaces", Acta Metallurgica Et Materialia, Vol. 39 [8] 1915- 24 (1991). 183 J. S. Wang, "Interfacial fracture-toughness of a copper Alumina system and the effect of the loading phase-angle", Mechanics of Materials, Vol. 20 [3] 251-59 (1995). 184 A. Hashibon, C. Elsasser, and M. Ruhle, "Structure at abrupt copper- alumina interfaces: An ab initio study", Acta Materialia, Vol. 53 [20] 5323- 32 (2005). 185 C. Scheu, "Manipulating bonding at a Cu/(0001)Al2O3 interface by different substrate cleaning processes", Interface Science, Vol. 12 [1] 127-34 (2004). 186 D. R. Clarke, "Interpenetrating phase composites", Journal of the American Ceramic Society, Vol. 75 [4] 739-59 (1992). 187 L. D. Wegner and L. J. Gibson, "The fracture toughness behaviour of interpenetrating phase composites", International Journal of Mechanical Sciences, Vol. 43 [8] 1771-91 (2001).

216

188 J. Rodel, "Interaction between crack deflection and crack bridging", Journal of the European Ceramic Society, Vol. 10 [3] 143-50 (1992). 189 L. S. Sigl, P. A. Mataga, B. J. Dalgleish, R. M. McMeeking, and A. G. Evans, "On the toughness on brittle materials reinforced with a ductile phase", Acta Metallurgica, Vol. 36 [4] 945-53 (1988). 190 P. Agrawal, K. Conlon, K. J. Bowman, C. T. Sun, F. R. Cichocki, and K. P. Trumble, "Thermal residual stresses in co-continuous composites", Acta Materialia, Vol. 51 [4] 1143-56 (2003). 191 M. Hoffman, J. Roedel, S. Skirl, A. Zimmermann, E. Fuller, and H. Mullejans, "Tailoring of an Interpenetrating Network Ceramic/Metal Microstructure to Improve Strength: Al2O3/Ni3Al and Al2O3/Al", Key Engineering Materials, Vol. 159-160 311-18 (1999). 192 M. F. Ashby, F. J. Blunt, and M. Bannister, "Flow characteristics of highly constrained metal wires", Acta Metallurgica, Vol. 37 [7] 1847-57 (1989). 193 S. Menon and W. G. Fahrenholtz, "Microstructure and mechanical properties of alumina-copper composites", pp. 285-95. in Advances in Ceramic Matrix Composites Ix, Vol. 153. Ceramic Transactions. Edited by N. P. Bansal, J. P. Singh, W. M. Kriven, andH. Schneider, 2003. 194 N. A. Travitzky, "Microstructure and mechanical properties of alumina/copper composites fabricated by different infiltration techniques", Materials Letters, Vol. 36 [1-4] 114-17 (1998). 195 P. Agrawal and C. T. Sun, "Fracture in metal-ceramic composites", Composites Science and Technology, Vol. 64 [9] 1167-78 (2004). 196 O. Lott, "Komponentenspezifische Materialanpassung an Metall-Matrix Verbundwerkstoffen auf Kupferbasis", Diplomarbeit, Hochschule Aalen, Aalen 2005. 197 Almatis, "Reactive and Calcined Aluminas for the Ceramic Industry." 2005 198 Nabaltec, "Nabalox Aluminiumoxide." 2006 199 K. D. Linsmeier, "Technische Keramik", Verlag Moderne Industrie, Vol. 208 (2000). 200 Wieland-Werke, "Wieland-K30 Kupfer Cu-OF." Ulm, 2005 201 Rettenmaier & Söhne, "Sicherheitsdatenblatt Arbocel P290." Holzmühle, 1999 202 Clariant, "Mowiol - Polyvinyl Alcohol." 1999 203 Zschimmer & Schwarz, "Produktdatenblatt Dolapix CE 64." 2008 204 Deutsches Institut für Normung, "Bestimmung der Rohdichte nach der Auftriebsmethode und der offenen Porosität durch Imprägnieren mit Wasser" DIN 51 918 (1986).

217

205 I. Gräf, "lonenätzen - Stand und Perspektiven für die Gefügekontrastierung keramischer und metallischer Werkstoffe. Teil Ill: lonenätzen in der Metallographie", Praktische Metallographie, Vol. Band 35 [7] 359 - 83 (1998). 206 S. Höhn, P. Obenaus, J. Hohlfeld, and C. Lies, "Ionenstrahlpräparation - ein treffliches Werkzeug zur Charakterisierung der Zwischenprodukte des schmelz-pulvermetallurgischen Aluminiumschäumverfahrens", Praktische Metallographie, Vol. 45 [3] 122- 34 (2008). 207 M. Koch, "Materialographische Gefügeentwicklung durch Anwendung des lonenstrahlätzens", Praktische Metallographie, Vol. 36 [5] 233 - 49 (1999). 208 American Society for Testing and Materials, "Standard Test Method for Splitting Tensile Strength of Intact Rock Core Specimens" D 3967-05 (2005). 209 Deutsches Institut für Normung, "Hochleistungskeramik – Mechanische Eigenschaften monolithischer Keramik bei Raumtemperatur – Teil 1: Bestimmung der Biegefestigkeit; Deutsche Fassung EN 843-1:2006" (2008). 210 Deutsches Institut für Normung, "Vornorm: Advanced technical ceramics - Test methods for determination of fracture toughness of monolitic ceramics -" Part 5: Singleedge veenotch beam (SEVNB) method; German version CEN/TS 144255:2004, DIN 14425-5 (2004). 211 A. Häger, "Interface Characterisation And Micromechanical Properties of Copper Alumina MMC", Report, University of New South Wales, Sydney 2008. 212 Hysitron Incorporated, "TriboIndenter - User manual." Minneapolis, 2007 213 E. J. Gonzalez and K. P. Trumble, "Spontaneous infiltration of alumina by copper-oxygen alloys", Journal of the American Ceramic Society, Vol. 79 [1] 114-20 (1996). 214 S. Long, O. Beffort, C. Cayron, and C. Bonjour, "Microstructure and mechanical properties of a high volume fraction SiC particle reinforced AlCu4MgAg squeeze casting", Materials Science and Engineering a- Structural Materials Properties Microstructure and Processing, Vol. 269 [1-2] 175-85 (1999). 215 S. Reymann, "Processing of High-temperature Resistant Metal Matrix Composites", Master thesis, Université de Metz, 2006. 216 A. Urena, J. Rams, M. D. Escalera, and M. Sanchez, "Characterization of interfacial mechanical properties in carbon fiber/aluminium matrix composites by the nanoindentation technique", Composites Science and Technology, Vol. 65 [13] 2025-38 (2005). 217 P. Demarecaux, D. Chicot, and J. Lesage, "Interface indentation test for the determination of adhesive properties of thermal sprayed coatings", Journal of Materials Science Letters, Vol. 15 [16] 1377-80 (1996).

218

218 S. S. Kim, Y. H. Chae, and S. Y. Choi, "Characteristics evaluation of plasma sprayed ceramic coatings by nano/micro-indentation test", Tribology Letters, Vol. 17 [3] 663-68 (2004). 219 Z. Mohammadi, A. A. Ziaei-Moayyed, and A. S. M. Mesgar, "Adhesive and cohesive properties by indentation method of plasma-sprayed hydroxyapatite coatings", Applied Surface Science, Vol. 253 [11] 4960- 65 (2007). 220 G. Marot, P. Demarecaux, J. Lesage, A. Hadad, S. Siegmann, and M. H. Staia, "The interfacial indentation test to determine adhesion and residual stresses in NiCrVPS coatings", Surface & Coatings Technology, Vol. 202 [18] 4411-16 (2008). 221 W. D. Nix and H. J. Gao, "Indentation size effects in crystalline materials: A law for strain gradient plasticity", Journal of the Mechanics and Physics of Solids, Vol. 46 [3] 411-25 (1998). 222 Verband der keramischen Industrie, "Brevier technische Keramik", pp. 1- 183, (2003). 223 E. Roos and K. Maile, "Werkstoffkunde für Ingenieure. Grundlagen, Anwendung, Prüfung", 2nd ed. Springer Verlag: Heidelberg, (2005). 224 A. Khalili and K. Kromp, "STATISTICAL PROPERTIES OF WEIBULL ESTIMATORS", Journal of Materials Science, Vol. 26 [24] 6741-52 (1991). 225 Z. Hashin and S. Shtrikman, "A variational approach to the theory of the elastic behaviour of multiphase materials", Journal of the Mechanics and Physics of Solids, Vol. 11 [2] 127-40 (1963). 226 G. Schneider, "Interpenetrierte Metall-Keramik-Verbundwerkstoffe (Preform-MMC) für thermisch und tribologisch hochbeanspruchte Leichtbauteile", Abschlussbericht (2006). Financed by Ministerium für Wissenschaft, Forschung und Kunst Baden-Württemberg, Landesstiftung Baden-Württemberg. 227 G. Requena and H. P. Degischer, "Creep behaviour of unreinforced and short fibre reinforced AlSi12CuMgNi piston alloy", Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, Vol. 420 [1-2] 265-75 (2006). 228 Y. Benveniste and T. Miloh, "The effective conductivity of composites with imperfect thermal contact at constituent interfaces", International Journal of Engineering Science, Vol. 24 [9] 1537-52 (1986). 229 D. P. H. Hasselman and L. F. Johnson, "Effective thermal-conductivity of composites with interfacial thermal barrier resistance", Journal of Composite Materials, Vol. 21 [6] 508-15 (1987). 230 H. Hatta and M. Taya, "Thermal conductivity of coated filler composites", Journal of Applied Physics, Vol. 59 [6] 1851-60 (1986). 231 K. Nitzsche and B. J., "Funktionswerkstoffe der Elektrotechnik und Elektronik", (1993).

219

232 J. L. Murray, "Aluminium binary alloy phase diagrams, Al-Cu (Aluminium - Copper)". in ASM Handbooks Online, http://www.asmmaterials.info, Vol. 3. Alloy Phase Diagrams Edited by H. Baker. ASM International, 1992. 233 W. Bergmann, "Werkstofftechnik - Grundlagen und Anwendung Teil 2", Vol. 3. Carl Hanser Verlag: München, (2002). 234 B. Huchler, "Herstellung und Charakterisierung von Metall-Matrix- Verbundwerkstoffen für Friktionsanwendungen", Diplomarbeit, HTW Aalen, Aalen 2002. 235 D. Staudenecker, "Entwicklung von porösen Keramik-Preforms und Herstellung von Metall-Matrix Verbundwerkstoffen", Diplomarbeit, Hochschule Aalen, Aalen 2001. 236 R. M. German, P. Suri, and S. J. Park, "Review: liquid phase sintering", Journal of Materials Science, Vol. 44 1-39 (2009). 237 H. He, R. L. Fu, D. L. Wang, X. F. Song, and M. Jing, "A new method for preparation of direct bonding copper substrate on Al2O3", Materials Letters, Vol. 61 [19-20] 4131-33 (2007). 238 T. Matsumoto, H. Fujii, T. Ueda, M. Kamai, and K. Nogi, "Measurement of surface tension of molten copper using the free-fall oscillating drop method", Measurement Science & Technology, Vol. 16 [2] 432-37 (2005).

220