advances in metal matrix composites : proceedings of an international meeting

Advances in Metal Matrix Composites

Edited by Lorella Ceschini

Roberto Montanari

Advances in Metal Matrix Composites

Special topic volume with invited peer reviewed papers only.

Edited by:

Lorella Ceschini and Roberto Montanari

Copyright 2011 Trans Tech Publications Ltd, Switzerland

All rights reserved. No part of the contents of this publication may be reproduced or transmitted in any form or by any means without the written permission of the publisher.

Trans Tech Publications Ltd Laubisrutistr. 24 CH-8712 Stafa-Zurich Switzerland http://www.ttp.net

Volume 678 of Materials Science Forum ISSN 0255-5476

Full text available online at http://www.scientific.net

Distributed worldwide by and in the Americas by

Trans Tech Publications Ltd Trans Tech Publications Inc. Laubisrutistr. 24 PO Box 699, May Street CH-8712 Stafa-Zurich Enfield, NH 03748 Switzerland USA

Phone: +1 (603) 632-7377 Fax: +41 (44) 922 10 33 Fax: +1 (603) 632-5611 e-mail: [email protected] e-mail: [email protected]

Preface Metal matrix composites (MMCs) have been the subject of scientific investigation and applied research for about three decades, but only in the past few years these materials became realistic candidates in engineering components, such as electronic heat sinks, automotive drive shafts, ground vehicle brake rotors, fighter aircraft fins and jet engine components. Compared to conventional materials the advantage of MMCs is that their structure can be tailored to get various combinations of physical and mechanical properties and meet the highest requirements. For example, they offer superior specific modulus, strength, thermal stability and wear resistance. Therefore, these innovative materials open up new possibilities for modern material science and technological development. This special issue presents 12 selected peer reviewed papers on different aspects of MMCs and aims to highlight recent findings in the field. The papers are mainly focused on light metal matrix composites, based on Al, Mg, Ti alloys, reinforced by particles, nano-dispersoids and long fibres. The results come from both experimental investigations and computer simulations and the contributions deal with different key issues: production processes, microstructural characteristics, mechanical behaviour, welding by advanced techniques, workability and tribology. We would like to express our gratitude to all the authors who contribute to the special issue; a special thank is due to Dr Riccardo Donnini for his help and dedication in collecting and organizing the selected papers. Lorella Ceschini and Roberto Montanari

Table of Contents

Manufacture of Aluminum Nanocomposites: A Critical ReviewC. Borgonovo and D. Apelian 1

Micro-Chemistry and Mechanical Behaviour of Ti6Al4V-SiCf Composite Produced by HIPfor Aeronautical ApplicationsP. Deodati, R. Donnini, S. Kaciulis, M. Kazemian-Abyaneh, A. Mezzi, R. Montanari, C. Testaniand N. Ucciardello 23

Simulation of the Mechanical Behaviour of Metal Matrix CompositesS. Schmauder, U. Weber, A. Reuschel and M. Willert 49

Dry Sliding Behaviour of Peo (Plasma Electrolytic Oxidation) Treated AA 2618/20% Al2O3pCompositeL. Ceschini, C. Martini, G. Sambogna and F. Tarterini 61

Strengthening Evaluation in a Composite Mg-RE Alloy Using TEMM. Cabibbo 75

Friction Welding of Particle Reinforced Aluminium Based CompositesL. Ceschini, A. Morri and F. Rotundo 85

Hot Drilling of Aluminium Matrix CompositeR. Donnini, L. Santo and V. Tagliaferri 95

Effect of Mechanical Mould Vibration on Solidification Behaviour and Microstructure ofA360-SiCp Metal-Matrix CompositesG. Timelli, E. Della Corte and F. Bonollo 105

Processing of Lightweight Metal Matrix Composites via In Situ Gas/Liquid ReactionC. Borgonovo and D. Apelian 115

Effects of Reinforcement Parameters on Fatigue Strength of Aluminium-Based Particulate-Reinforced CompositesM. Vedani 125

Production and Characterization of Aluminum Iron Powder Composites withFerromagnetic PropertiesS. Amadori, E. Bonetti, E.G. Campari and L. Pasquini 135

Comparison between Roll Diffusion Bonding and Hot Isostatic Pressing ProductionProcesses of Ti6Al4V-SiCf Metal Matrix CompositesC. Testani, F. Ferraro, P. Deodati, R. Donnini, R. Montanari, S. Kaciulis and A. Mezzi 145

Manufacture of Aluminum Nanocomposites: A Critical Review

Cecilia Borgonovoa and Diran Apelianb

Metal Processing Institute, Worcester Polytechnic Institute

Worcester, MA 01609 USA [email protected], [email protected]

Keywords: lightweight, nanocomposites, agglomeration, manufacturing routes, ex-situ processing, in-situ processing, gas-liquid reactions.

Abstract. In the last two decades, metal matrix nanocomposites have witnessed tremendous growth. Particulate-reinforced nanocomposites have been extensively employed in the automotive industry for their capability to withstand high temperature and pressure conditions. Several manufacturing approaches have been used to fabricate them. Non-homogeneous particle dispersion and poor interface bonding are the main drawbacks of conventional manufacturing techniques. A critical review of nanocomposite manufacturing processes is presented; the distinction between ex-situ and in-situ processes is discussed in some detail. Moreover, in-situ gas/liquid processes are elaborated and their advantages are discussed. The thermodynamics and kinetics of the reaction between the precursor gas and the liquid metal have been analyzed and their role on particle formation studied. This critical review will provide the reader with an overview of nanocomposite manufacturing methods along with a clear understanding of advantages and disadvantages.

Metal-matrix Composites in Context

Metal-matrix composites are a hybrid material in which rigid ceramic reinforcements are embedded in a ductile metal alloy matrix. They tailor the best properties of two different materials, such as ductility and toughness of the metallic matrix and the high modulus and strength of ceramic reinforcements. Their first application can be traced back to the late 1960s, with the development of a steel-wire reinforced copper alloy [1]. The aerospace industry led the application and use of composite materials in spacecrafts components. High-performance and high-integrity materials are required for extreme environments and critical applications such as for space missions. It is interesting to note that during its lifetime, the International Space Station will undergo 175,000 thermal cycles from +125 C° to -125 C° as it moves in and out of the Earth’s shadow. During the last 4 decades, aluminum matrix composites were specifically developed to meet both aerospace and defense needs. Continuous boron fiber reinforced aluminum was used in the Space Shuttle Orbiter as the frame and rib truss members in the mid-fuselage section; there are other applications such as landing gear drag link yielding 45% weight savings. A Gr/Al composite is the constituent of a high-gain antenna boom for the Hubble Space Telescope. This boom (3.6 m long) offers the stiffness required to maintain the position of the antenna during space maneuvers. In the 1980's and early 1990's, metal matrix composite development programs were in vogue and there was much activity at all major aluminum producers. Alcan, through its Duralcan subsidiary, established a 25 million pound per year production capability for particulate-reinforced aluminum composites. The Aluminum Association convened the Aluminum Metal Matrix Composites Working Group, a product of which was the ANSI H35.5 standard that established a nomenclature system for aluminum composites [2]. As expected, metal matrix composites found applications in a variety of other markets such as automotive, electronic packaging, industrial product and recreational products [3]; though not a conclusive list, the components given below illustrate applications that utilize Al based composites:

© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.1

• Chevrolet Corvette and GM S/T pick-up truck drive shafts

• Plymouth Prowler brake rotors and GM EV-1 brake drums

• Toyota diesel engine pistons

• Pratt & Whitney 4000 series engine fan exit guide vanes

• Motorola’s Iridium Satellites and GM EV-1 electronic packaging applications

• F-16 fighter aircraft ventral fins and fuel access covers

• Bicycle components and golf clubs

Fig.1. Global outlook of metal-matrix composites by application segment (2004-2013). Source: BCC Research.

An almost 70% increase in metal matrix composites is estimated to occur in the use of Al in vehicles from 2004 to 2013, see Fig.1. The choice of aluminum alloys as matrix is dictated by the compelling need to have vehicles with low fuel consumption and reduced emissions for a sustainable future. Because of their high strength-to-weight ratio, aluminum alloys are considered to be an alternative to conventional steels and to the more expensive superalloys. The amount of aluminum per automobile produced in USA has increased from 251 lb. of 1999 to 280 lb. forecast for 2014 [4,5]. In Europe it went from 220 lb. of 1999 to 462 lb. forecast for 2014 [6], see Fig.2. Aluminum-based composites have contributed to such growth by improving strength and hardness of the aluminum matrix, broadening the application field to more highly-rated regimes.

2 Advances in Metal Matrix Composites

Fig.2. Amount of aluminum per automobile in USA and Europe (1999-2014) [4-6].

When compared to ferrous sand casting, high-production of metal matrix composite components through die casting, squeeze casting and semi-solid molding can compete effectively in terms of cost. In the commercial aircraft industry, weight savings has been estimated to be around $450/kg; and in spacecraft, it can reach $40,000/kg. For what concerns the automotive industry, Ford Motor Co. has placed the value of weight reduction at between $0.35-3.50/kg depending on vehicle platform. In freight transport, the weight savings of a component translates to additional freight that can be hauled. For heavy-duty trucks, such savings has been valued from $2-16/kg depending on the equipment's operational cycle [7]. Aluminum metal matrix composite also win out on iron components in terms of marketability and maintainability. Though metal matrix composites offer many advantages, they do have shortcomings such as low fracture toughness, low strength and hardness at high temperatures and poor machinability. The main concern of machining particulate metal matrix composites is the extremely high tool wear due to the abrasive action of the ceramic reinforcing particles. Tool selection is limited to a small group of extremely hard and expensive materials. The cutting tool must be able to withstand intermittent cutting of hard (reinforcement) and soft (matrix) materials. Polycrystalline diamond tools are often recommended for machining this particular class of materials and the high cost of such tools together with the need of frequent tooling changes increases the cost of the machining process [8]. Conventional machining methods have applied on composites with poor results. Non-traditional processes like waterjet, abrasive waterjet cutting, electrical discharge machining, ultrasonic machining and laser cutting provide precision finish but are characterized by very high costs and slow machining rates [9]. Therefore, machining still remains an issue to address since it will continue to be a necessary step to produce the required close dimensional tolerances and surface finish. There is a compelling need for an aluminum-based material whose strength at high temperatures is retained and whose manufacturing process can be adapted to existing industrial infrastructures. Nanocomposite aluminum matrix materials have emerged as a viable alternative to overcome the limitations of aluminum (micro-) composites. Tensile strength, hardness and fracture toughness are enhanced as well as dimensional stability at high temperatures, see Fig.3 [12]. They currently represent a market segment of $ 250 million, 80% of which is covered by automotive applications. Nanoparticles in castings are considered to be the most promising segment in casting material development [10]. However nanocomposites are challenging to produce as structural components due to difficulties in attaining a homogeneous distribution of the nano-phased particles. Clusters of secondary phases are detrimental for the final component performances and also affect post-processing techniques and the ability to machine the part. Representative metal nanocomposite systems and associated attributes are given in Table 1 [11].

0

100

200

300

400

500

1999 2010 2014

Lb.USA

Europe

Lorella Ceschini and Roberto Montanari 3

Fig.3. The variation of change in length versus temperature for aluminum and its composites at constant SiC content (7.5 Vol% SiC) [12].

Matrix/Nano-sized

Reinforcement Properties

Al/SiC Mg/SiC Al/Al2O3 Mg/Al2O3

Improved ultimate

strength, hardness

and elastic modulus

Al/AlN Higher compression

resistance and low

strain rate

Ni/PSZ (partially-stabilized zirconia) and Ni/YSZ (yttria-fully stabilized zirconia)

Improved hardness

and strength

Cu/Al2O3 Improved

microhardness

Table 1. Metal Nanocomposite Systems of Interest and Associated Attributes [11].

Nano-particle reinforced composites. Nano-particles have progressively replaced other discontinuous reinforcement structures such as nano-fibers, nano-wires or nano-platelets. SiC, TiC, WC, TaC, TiB2, AlN, and Al2O3 are some of the most common types of nano-particles that have been utilized. The characteristics of nano-particle reinforced composites can be summarized as follows:

- drastic change of fracture mode from inter-granular fracture in monolithic metal to trans-granular fracture in nano-composites;

- moderate to significant improvement in strength;

- moderate improvement of fracture toughness;

- significant improvement of creep resistance, thermal shock resistance, and wear resistance;

- enhancement of dimensional stability at high temperatures.


Zebarjad et al. [13] compared the effect of 25 µm, 5 µm, and 70 nm SiC particles on dimensional stability in an aluminum alloy. The temperature sensitivity of aluminum decreases in the presence of both micro and nano-sized silicon carbide, though the effect of nano-sized silicon carbide on dimensional stability is much higher than that of micro-sized ones. Ren and Chan [13] added SiC nano-particles (50 nm) to 7075 aluminum alloy. They pointed out that this resulted in increased wear resistance and high temperature creep resistance compared to the same alloy reinforced with larger sized 13 µm SiC particles. Furthermore, the volume percentage of nano-particles needed to achieve this result was considerably smaller than in the case of the 13µm SiC particles. Similarly, the tensile strength of an aluminum alloy reinforced with 1 % volume of Si3N4 (10 nm) has been found to be comparable to that of the same alloy reinforced with 15 % volume of SiC particle in the micro-sized range (3.5 µm); the yield strength of the nano-metric composite being significantly higher than that of the micro-metric one [14]. The existence of a threshold size (“critical size”) below which the addition of particles improves properties has been reported – see Table 2) [11]. It must be noted that the mechanism responsible for property improvements remains a matter of debate among researchers.

Table 2.Critical Size for Properties Improvement in Nanocomposites [11].

Strengthening theory based on a continuum approach is not useful; since it ignores the influence of particles on micromechanics of deformation - i.e., location of particles, grain size, and dislocation density. Several discontinuous approaches have been formulated to include particle effects. The modified shear lag theory [16] of Nardone and Prewo, the Eshelby- based particle-compounded model and the EMA (effective medium approximation) model by Stroud are the most popular ones [16]. They take into account one or more of the following strengthening mechanisms:

- Orowan mechanism: the stress that must be applied to force dislocations to by-pass an obstacle (such as a particle) is the principle of the Orowan strengthening mechanism. The stress arises due to the resistance of closely spaced hard particles as dislocations pass through. If the particles are coarse (in the micro-size range) and the inter-particle spacing is large, the Orowan effect is not significant [16]. When highly dispersed nano-sized particles are present, Orowan strengthening becomes more favorable. Creep resistance and thermal stability are consistently enhanced. TEM (transmission electron microscopy) observations

Properties Critical Reinforcement

size (nm)

Catalytic activity <5

Softening of hard magnetic materials

<20

Change of refractive index

<50

Producing electromagnetic phenomena such as super paramagnetism

<100

Strengthening and toughening

<100

Modifying hardness and plasticity

<100


reveal strong dislocation bowing and tangling around the particles, confirming the operating mechanism [15, 16].

- Thermal mismatch: matrix and reinforcement have different coefficients of thermal expansion. During cooling of the component, plastic deformation is produced in the matrix at the various interfaces. These deformations increase the density of dislocation [16].

- Load-bearing: the strong bond due to the cohesion between particles and the matrix contributes to load-bearing capacity [16].

When all these factors are taken into account, the increase in mechanical properties with the decrease in size can be estimated. Critical Issues in Processing of Nanocomposites

The main challenge for nanocomposites is how to make them – the processing routes to manufacture them. Dispersing the second phase particles in the matrix and achieving a strong interfacial bond are the two main processing challenges. Most fabrication processes fall short of fulfilling these tasks. Clusters of particles and weak matrix-reinforcement interfaces compromise the ability of the composite material to function under extreme conditions, such as high temperature and pressure typical of automotive applications (especially Diesel engines).

Uneven dispersion and agglomeration. Agglomeration is a common phenomenon that occurs when a solid particle comes into contact with a non-wetting medium [17, 18]. The clustered particles significantly reduce the failure strain of the composite; degradation is attributed to preferential nucleation of cracks in clustered regions. Final fracture is produced by the crack propagating to other clusters. Clustering occurs due to combined effects of agglomeration, sedimentation (particle settling rate) and particles pushing by the advancing solidus-liquidus interface. Particle clustering occurs since the system tends to minimize its free energy. A solid inclusion is never perfectly smooth: its surface is covered with cavities filled with gas, which contribute to increasing the system’s Gibbs energy. This is can be seen by analyzing the equation describing the Gibbs energy of a gas-liquid-solid system [17]:

lg lg( ( , ) ( , ))µ µ γ γ γ∆ = − + ∆ + ∆ + ∆g l sg sg sl sl

G T P T P S S S (1)

where T is the temperature, P the pressure in the liquid, µg and µl

the chemical potentials of gas

and the liquid, ∆S is the change in interfacial areas and γ surface energies. When the particle size is brought down to the nano-scale range, surface energy is enhanced by three orders of magnitude (Table 3), introducing strong instability in the system and hindering particle wetting by the molten metal.

Table 3. Variation of Surface Energy with Particle Size (1 g of sodium chloride) [20].

Particle size [cm] Total surface area [cm²] Surface energy [J/g]

0.1 28 5.6 410−×

0.01 280 5.6 310−×

0.001 2.8 310× 5.6 210−×

410− 2.8 410× 0.56

710− 2.8 710× 560


The natural tendency towards equilibrium is the “spring” that allows the system to assume a physical configuration for which the Gibbs energy is lowered to a minimum value. With this perspective, agglomeration acts like a “stability configuration”: several nano-particles cluster in one micro-agglomerate (Fig.4), providing a less extended total interfacial area. The dynamics of the relative motion of two nano-sized particles has been extensively studied [18, 20]. Due to the complexity of the problem, the analysis is usually limited to two main mechanisms: Brownian diffusion/motion (or perikinetic aggregation), and inter-particle forces (electrostatic and Van der Waals). External forces are not considered and particle inertia is neglected.

Fig. 4. Clusters of SiC nano-particles [19].

Brownian motion. It has been demonstrated [18] that a suspended particle is randomly bombarded from all sides by thermally-excited molecules coming from the liquid. Brownian diffusion ensures continuous collision between particles [19]. It can be defined as the incessant random motion exhibited by microscopic particles immersed in a fluid. Einstein noticed that if one solid inclusion is small enough to behave like a gas molecule, it is continuously collided by liquid molecules and displaced as a consequence. The magnitude of the displacement follows a Gaussian statistic distribution according to the relation:

2

6

kTtd

rηπ= (2)

where η is the viscosity of the medium, t the time, r the particle radius, T the temperature and k the Boltzmann’s constant. The displacement increases with decreasing particle radius, thus enhancing the probability of a collision to occur. It has been confirmed [18] that for particles smaller than 3.5 µm, Brownian motion totally dominates the agglomeration dynamics. The aggregation rate for 20 nm particles has been evaluated to be four orders of magnitude higher when compared to particles in the range of 1 µm [20]. This behavior can be explained by the fact that as the particle size increases the potential energy of repulsion increases, thus making aggregation less likely.

Inter-particle forces: Van der Waals attraction and electrostatic repulsion. According to Van der Waals, the non-ideality of gases can be attributed to the existence of molecular or atomic interactions [21]. Such dynamic interactions are established between the instantaneous dipoles formed in an atom’s orbiting electrons. Thus, the resulting force is weak and becomes significant only at a short particle distance. Hamaker [21] found such interactions to exist between particles and modified Van der Waals’ formulation through the so called “additivity concept” (single atoms or molecules make up the particle). When the cavities located on a solid inclusion are filled with


gas, negative Van der Waals forces come into play, causing particle agglomeration. Attraction is favorable because it reduces the value of the Gibbs free energy by θ:

212

Ar

Hθ

−= (3)

where A is the Hamaker constant, which depends on the polarization properties of the molecules on the particle surface, r is the reduced particle radius and H the inter-particle distance [18]. When the dimension of the particle is smaller than 1 µm, Van der Waals forces dominate. Coulomb force of repulsion competes with Van der Waals attraction. It can be noted from Fig.5 that the electrostatic repulsion is overcome by the Van der Waals attraction force for a inter-particle distance down to 1 nm. For smaller values, the Born repulsion of adjacent electron clouds dominates.

Fig. 5. Forces acting between two particles [20].

Interface debonding. Interface bonding between particles and the matrix is critical as it affects load transfer from the matrix to the particle and for delaying the onset of particle–matrix de-cohesion. Voids nucleation and growth have also been observed to be correlated with the loss of coherency at particle/matrix interface. All these aspects have a profound effect on the strength and stiffness of the composite. Interface debonding caused by large thermal mismatch between metal and ceramic has been noticed to be the main mechanism responsible for fracture of the material [22]. Matsunaga et al. [23] measured the effect on strength and fracture toughness of surface oxidation of SiC particles, according to the reaction:

2 22 3 2 2 ( )SiC O SiO CO gas+ → + (4)

They detected enhanced strength only for thick oxide layers (1.4 µm), while fracture toughness consistently decreased after the oxidation process for all temperatures and exposure times. Therefore, crack initiation on particle surface is more likely to occur, affecting life duration of the component. It’s difficult to determine whether cracking of the oxide layer is responsible for the frailure mechanism of the composite materials. Exposure of clusters of bare particles on the fractured surface (Fig.6) could be an indication of such phenomenon. EDS analysis confirms the presence of silicon dioxides on particles surface (Fig.7). Other studies [24,25,26] found that the wettability of the reinforcement by liquid aluminum is improved when an oxide coating is applied. However, the very thin film character of silicon dioxide makes it brittle, fragile and easy to break-down during particle incorporation and vigorous stirring. In addition to this, when a high percentage


of coating material is used in the oxidation process the interfacial bonding between particle and matrix is degraded and a typical bondless morphology underlines the non-wetting characteristic between both surfaces. Therefore, wettability is enhanced only for specific coating thickness and for layers that are continuous, which is a feature connected to the nature of the heat treatment. Oxidation in air has shown not to improve the contact angle between particle and matrix [27], whereas it is improved in oxygen supported atmosphere. Large thermal mismatches between particle and matrix can also cause interface debonding and fracture upon cooling to room temperature [28].

Fig.6. SiC nano-particles on an A356 aluminum alloy fractured surface.

Fig.7. EDS spectrum of a SiC nano-particle on the fractured surface.


Manufacturing Routes

Classification of processing routes. Metal matrix composite manufacturing processing can be divided into two general categories: ex-situ and in-situ. Ex-situ is when the reinforcement is externally added to the matrix. In-situ synthesis involves the production of reinforcements within the matrix during the processing stage [33, 34]. The same classification applies for nanocomposite manufacturing as well. Ex-situ manufacturing techniques can be further classified into two main processing schemes [33,36]: solid-state and liquid-state. In some instances when the processing is in the semi-solid range (such as in droplet consolidation or similar techniques) then the classification could be further expanded to solid-state, liquid-state and semi-solid state. For the purposes of this review we will limit ourselves to the first two processing routes. Among solid-state techniques, powder metallurgy and mechanical attrition are the most popular ones. The nano-scale can be easily reached, although the cost of the powder is significantly high. Interfacial and surface wetting issues are considerably diminished. This is because both phases remain in the solid state, where diffusivity is much lower [29, 30]. The final products are generally affected by a high amount of porosity, which strongly decreases the fatigue resistance and requires further metalworking. When the process involves attrition at high temperatures chemical modification of the initial constituents is likely to occur [31, 32]. Liquid-state routes can be sorted into four major categories: infiltration, agitation, spraying and ultrasonic cavitation based solidification. Semi-solid casting of nanocomposite materials is still an open field; a novel method of melting, compacting and solidifying semi-solid billets has been tested in [35]. Liquid metal is generally less expensive and easier to handle than powders, and the shape flexibility constitutes a significant advantage. Liquid-state processes are generally fast and easy to scale-up. Despite this, they are affected by the lack of wettability of the reinforcement and by interfacial reactivity. Moreover, they are often limited to low melting point metals [29, 30]. In-situ metal matrix composites are not affected by the shortcomings typical of ex-situ composites, although control of process variables still remains an issue. In-situ fabrication methods can be divided into two major categories according to the physics of the process itself: “reactive” routes, where the reinforcement is synthesized within the metal matrix through a gas-liquid, liquid-liquid, or solid-liquid reaction, or “morphological” routes,

where a favorable composite architecture evolves as a consequence of processing. Other methods, which cannot be used for mass production of near net shape parts can be traced in the literature [31,36]. The most important are laser deposition, spray deposition, sol gel synthesis, nano-sintering and electroplating. They are costly, time and energy consuming processes. Therefore, their application is unlikely to be extended to the industrial scale. Such techniques are generally used for coating and thin films deposition. In this review, only mass production methods see table, which could be adapted to existing industrial infrastructure and can meet the need to large production volumes will be taken into account.


Process System (matrix/reinforcement) Reinforce

ment size Main features

Ex-situ: solid-state

(Section 3.2.1)

+ Near net shape; +Industrially scalable; -Non homogeneous particle size distribution; -Costly.

- Powder metallurgy Al/ 2 3Al O , Al/ 3 2Si N

15-100 nm

- Mechanical attrition and alloying

Al-Fe/ 5 2Al Fe , Al/ 4 3Al C ,

Al/SiC

9-27 nm

Ex-situ: liquid state

(Section 3.2.2)

- Stir casting Al/SiC 40 nm

+Industrially compatible +Industrially scalable; +Inexpensive; -Particle clustering and debonding.

- Ultrasonic cavitation based solidification

Al-Si/SiC, Al/ 2 3Al O

< 100 nm, 10 nm

+Good particle dispersion; +Inexpensive; -Industrially non-scalable.

- Infiltration Al-Cu-Mg/ 2 3Al O

50 nm +Good mechanical properties; -Expensive equipment (preform); -Un-easy to scale up.

In situ: reactive routes

(Sections 3.3.1,3.3.2,3.3.3)

- Combustion synthesis

Al/ 2TiB 30-100 nm

+Good particle dispersion and particle/matrix bonding; +Inexpensive; +Industrially scalable; -Difficult process control.

- Exothermic dispersion

Al/ 2TiB < 0.7 µm

- Substitutional chemical reaction

Al/ 3Al Zr + 2 3Al O

Cu-Ti/ 2TiB

80 nm 50 nm

- MixAlloy by Sutek Cu/ 2TiB 50 nm

- Gas-liquid process Al alloys/AlN, SiC, TiC 100-500 nm

In-situ: morphological

(Section 3.3.4)

- Rapid solidification Al-Fe/ 100 x xAl Fe− , Al/TiC 20-150 nm, 40-80 nm

Table 4. Manufacturing Methods for Metal Matrix Nanocomposites (Mass Production).

Ex-situ methods

Solid state

Powder metallurgy. Prior work in synthesizing nanocomposites involves the use of powder metallurgy techniques, which are usually not cost-effective. Blending of matrix and reinforcement


powders followed by hot or cold pressing and sintering is a standard fabrication sequence; a schematic of a typical powder metallurgy (P/M) processing scheme is shown in Fig.8. In P/M processing, agglomeration can be minimized only if the size of the matrix powder is in the size range of the reinforcement phase. In addition, further working of the product via P/M may cause the reinforcement phase to break up and deform the surrounding matrix, leading to stress concentration and cracking [34]. The advantages of the process are flexibility and the ability to produce near-net shaped components. The size range of metal powders available on the market is quite wide which it is an advantage. P/M has been used [14] to add 50 nm alumina particles to aluminum powder. The process consists in wet mixing (aluminum powder mixed with varying volume fraction of Al₂O₃ powder in a pure ethanol slurry), followed by drying at 150ºC and cold isostatic pressing to compact the powder. The compacted powder is then vacuum sintered at 620ºC (approximately 60ºC below the melting temperature of aluminum). Massive clustering has been observed, and its occurrence increases with decreasing particle size. Ma et al. [37] fabricated via P/M processing nanometric silicon-nitride reinforced aluminum composites. They reported the presence of several agglomerates in the aluminum matrix. Peng et al. [38] created a novel and simplified process for producing aluminum matrix nanocomposites reinforced with oxide particles. The novelty lays in the use of Al₂O₃ surface layers existing on matrix aluminum particles as the ceramic reinforcement. A good distribution has been achieved, although the process does not allow satisfactory control of the process. Moreover, the effectiveness and the scalability of the method remain to be proven.

Fig.8. Processing routes for particulate Fig. 9. Grain size and strain vs. milling for reinforced composites [34]. WC particles [39].

Mechanical attrition and alloying. Mechanical alloying was invented in 1980 to manufacture particle strengthened metal alloys. In the last ten years, the method of high-energy milling gained much attention as a non-equilibrium process able to produce nano-scale microstructures. A variety of ball mills have been developed for different purposes including tumbler mills, attrition mills, shaker mills, vibratory mills, and planetary mills [32]. In the high-energy ball milling process, alloying occurs as a result of repeated breaking up and welding of matrix and reinforcement particles. Both powders are subjected to severe plastic deformation due to collision with the milling tool. Deformation occurs at high strain rates; thus, after extended milling (Fig.9), the average powder grain size can be reduced to few nanometers [32,39]. It should be noted that aluminum nanocomposites with the trade-name DISPAL, reinforced with Al₄C₃ particles, are manufactured via mechanical alloying [14]. Flexibility and scalability are key advantages of the process; contamination by the milling tool and the atmosphere are the main disadvantages of the process. Milling of refractory metals (tungsten) in a high-frequency shaker for extended times can result in iron contamination of more than 10 at% [38]. To prevent contamination, the process should be carried out in an inert atmosphere and the mills ought to be coated. Another major issue is the


occurrence of chemical reactions as a consequence of converting mechanical energy into thermal energy [32]. Zhang et al. [40] proved that there exists a particle size below which further size reduction cannot be performed, since the stress necessary to break the particles is above the process capabilities. The stress required for processing can be expressed as:

cf

c

K

aσ

π= (5)

Where f

σ is the fracture stress, c

K the fracture toughness and c

a size of material defects. When the

particles are reduced to the nano-range, the likelihood of having internal defects and surface notches

are considerably reduced. In this case, f

σ will approach the theoretical strength of the ceramic

material. The impact stress of silicon-carbides is over 15 GPa, which is the value needed to fracture a “perfect” (with no defects) ceramic. Such stress is not achievable with conventional high-energy mechanical mills. Furthermore, nano-particles produced by attrition do not possess uniform size distribution and the process is limited to materials with very poor thermal conductivity [41].

Liquid state

Stir casting. Stir mixing techniques, widely utilized to mix micron size particles in metallic melts [34, 41] have recently been modified for dispersing small volume percentages of nanosize reinforcement particles in metallic matrices [41]. The restraints correlated with mixing nanosize particles in metallic melts are:

- Particle introduction into the melt; - Particle clustering; - Weak bond between matrix and reinforcement because of surface contamination of the externally added reinforcement.

Because of increased surface area together with the reduction in particle size, inserting the particles in the melt and homogeneously dispersing them is a challenge. The increase of interfacial energy raises the free energy of the system, causing agglomerates to form. Xiaodan et al. [42] managed to avoid agglomeration of 40 nm SiC particles in aluminum by designing an experimental setup consisting in fusion, vacuum, and stir parts. In fact, simple stirring by means of a lance or rod does not overcome particle clustering. Alternative stirring tools have also been developed to improve the dispersion. Ultrasonic based solidification has been the most successful one. Ultrasonic cavitation based solidification. High-intensity ultrasonic waves (above 25 W/cm²) can generate strong non-linear effects in the liquid such as transient cavitation and acoustic streaming [43]. These waves produce a dispersive effect and tend to homogenize the microstructure of the melt [44]. An ultrasonic probe is immersed into the melt to create the acoustic field (Fig.10) and nano-sized particles are added during the process. The acoustic bubbles burst, creating hot micro-spots that locally raise the temperature of the melt. This enhances particle wettability and favors good dispersion. It has been measured [43] that with a 3.5 kW ultrasonic power, the ultimate strength and yield strength were improved more than 60% and 100% (Fig.11). In addition, 2.0 vol% SiC nano-particles improved hardness by 20% [45].


Fig.10. Schematic of ultrasonic solidification Fig. 11. Strength vs. percentage nano-particles processing [43,44]. added percentage [45].

One drawback of this technique is the dissolution of the oscillating probe in contact with the molten metal. To overcome such difficulty a non-contact method, where the ultrasonic probe is not in direct

contact with the liquid metal, was attempted to disperse 10 nm 2 3Al O particulates in aluminum

matrix [46]. In this method the mold was subjected to ultrasonic vibration. The reinforcement was found to be uniformly distributed. The amount of material processed with ultrasonic cavitation based solidification generally does not exceed 200 g. The ultrasonic power necessary to achieve good particle dispersion is proportional to the amount of material processed. Therefore, industrial scale quantities would require enormous and costly power supplies. Infiltration. The process consists of infiltrating a porous preform. Capillary forces and viscous drag through preform interstices hinder wetting of the preform by the melt. Evans et al. [30] observed from an “energetic” standpoint that metals generally do not bond to non-metals, and that the chemistry of the system must be modified, or external pressure must be applied. Chemical modification includes coating, adding special elements to the melt, or using special atmospheres [47,30]. Mechanical force reduces porosity and improves interfacial bond. Pressures of around ten atmospheres are needed to infiltrate the melt into 1 µm wide pores [30]. As a result, preform fragmentation, deformation and unevenly reinforced castings [47] may result. Kaptay [48] noted that that when the partially infiltrated liquid metal reaches the “equilibrium depth” (the depth at which interfacial forces are zero), further infiltration will occur by additional pressure. The threshold pressure is given by:

(1.63 )

3threshold lv

P WR

πσ= − (6)

Where R is the particle radius, W the adhesion energy and lv

σ the interfacial energy between the

liquid and vapor phases. The lower the particle radius, the higher is the threshold pressure. When pressures of some GPa are applied, nano-materials can be manufactured. Gierlotka et al. [49] used a toroid cell at pressures up to 7.7 GPa and temperatures up to 2000 °C for the infiltration of an alumina preform with a grain size of 10 nm. Schultz et al. [50] succeeded in the infiltration of an alumina preform with particle size of 50 with Al alloy A206. The composite showed an increase in hardness by 19% compared to the base alloy. The downside of the infiltration technique is the high cost of nano-sized ceramic preform. The latter is a significant disadvantage.


In brief, Ex-situ processes as described above have their distinct advantages and disadvantages. In general however, Ex-situ processes suffer from:

- Thermodynamic incompatibility: interfacial reactions between the reinforcements and the

matrix may occur. Detrimental phases such as 4 3Al C and 5 3Ti Si have been detected in

composite materials manufactured through mechanical stir casting; - Contamination: oxide layers around the particles increase the surface energy, decreasing

wettability of the system [51]; - Inhomogeneous microstructures: particle agglomeration and clustering occur.

In-situ methods

When nano-composite materials are synthesized via In-situ processes, fabrication issues associated with ex-situ methods are mitigated or completely alleviated. The benefits that in-situ manufacturing methods provide are [52]:

- Thermodynamic stability at high temperatures; - Clean interface between particle and matrix, resulting in strong interfacial bonding.

Detrimental phases are eliminated and the creation of the nascent interface can be guided by process control. Wear resistance is enhanced as a result;

- Range of particle size in the nanocomposite are lower than via Ex-situ processes; - Improved distribution yields to superior mechanical properties; - Composites with a broad variety of matrix materials (aluminum, titanium, copper, nickel and

iron) and reinforcing particles (borides, carbides, nitrides, oxides and their mixtures) can be produced;

- Process is scalable and cost effective.

Commercial applications are still limited by the complexity of the reactions and the lack of knowledge concerning these techniques. The two classes of processes –reactive and morphological are described and discussed below.

Reactive processes: solid-liquid state

Elements or compounds react in the presence of a third liquid metallic phase that acts as a solvent medium. The reinforcement is generated via diffusion of components in the metal matrix [52]. Combustion synthesis, XD process, mixed salt reaction, direct metal oxidation and reactive synthesis are examples of solid-liquid processes. There are detailed below.

Combustion synthesis. Combustion synthesis (see Fig.12) -or self-propagating high-temperature synthesis (SHS)- was invented by Merzhanov et al. [53]. A mixture of powdered elements is initially prepared and pressed into cylindrical pellets. Electrically heated coils or a laser act as the heat source that initiates a chemical reaction between the various elements. The solvent can be molten Al, Mg, or Ti where other non metallic elements, such as C and B, are present. The ceramic compounds are burnt via ignition waves at a temperature higher than the melting point of the metal matrix. A typical reaction is:

Al + Ti + 2B → Al + TiB₂ + HEAT = Al/TiB₂ (7)

The highly exothermic nature of the process allows it to be self-sustaining and is energy efficient. The heat released during the reaction keeps the propagation front stable by heating up the un-reacted portion of the sample. The equipment is simple, processing times are short due to very high combustion rates (0.15 m/s) and metastable phases can be synthesized. In addition, volatile impurities are evaporated due to high temperature of the process. Although a variety of shapes and


geometries can be attained, porosity (up to 10%) in the final component still remains an issue. Further processing such as high-pressure consolidation is a necessary step. At present, a major program is underway between WPI and Colorado School of Mines to explore using combustion synthesis to die cast Al and Mg engine components that contain 20-40% second phases.

Fig.12. Combustion synthesis process [54].

Exothermic dispersion (XD process). The XD process was developed by Martin Marietta Corporation and has been extensively applied to the manufacturing of light-weight materials [52]. Jet engine turbine blades with weight savings of 30% to 50% have been fabricated with this process. It is a sustained high-temperature synthesis whose driving force is the difference of melting temperatures of the components. Ceramic phases and a third metallic phase are emplaced together and heated above the melting point of the metallic phase. The ceramic phases release heat and interact, forming very fine (nano-sized) particulates [52, 55], Fig.13. Particle size and distribution are system-dependent. It depends on the thermal conductivity of the environment and the amount of heat developed during the reaction. Tailoring the composition of the initial species can regulate the volume percentage of reinforcement. The exothermic reaction eliminates oxides and provides clean interfaces [52]. Hot isostatic pressing of the final component is necessary in order to reduce porosity.

Fig.13. Schematic diagram of XD process [52].

Substitutional chemical reaction. An in situ copper matrix composite with 3.5 wt.% 2TiB was

prepared by thermic reactions of 2 3B O , carbon as reduction agent and titanium in copper–titanium


melt [56]. The in situ-formed 2TiB particles with a size of about 50 nm exhibited a homogenous

dispersion in the copper matrix. Due to their reinforcement, the tensile strength and hardness of the

in situ Cu– 2TiB composite significantly improved. The in-situ composite also had a high electrical

conductivity. Zhao et al. [57] synthesized nano-sized 2 3Al O and 3Al Zr particles in aluminum in the

system 3 2Al Zr(CO )− according to the reaction:

3 2 2 3 2 3l3Zr(CO ) + 13Al = 6CO + 3Al Zr + 2Al O (8)

A magnetic field is also applied in order to enhance the chemical reaction. The mean particle size is about 80 nm, and the nano-sized particles are well distributed in the Al matrix. The ultimate tensile strength and yield strength of the nanocomposites are enhanced with increasing of particulate volume fraction, and are higher than that of the Al nanocomposites synthesized under zero magnetic field.

Reactive processes: liquid-liquid state

The MixAlloy Process patented by Sutek Corporation [58] has been applied to manufacture nanocomposite materials. Two streams of metal melts containing ceramic inclusions interact with each other in a reaction chamber to form refractory particles. The mixture is then rapidly cast or atomized. Titanium boride particles in a copper matrix have been manufactured with this method. It has been reported [52] that particle sizes around 50 nm have been achieved. In the first process disclosure by Nam.P.Suh [58], the impingement between the metal streams is direct, while in a subsequent patent [59] the impingement is indirect. In this manner, instability in the metal streams are mitigated. The impingement may not provide adequate energy to mix the metal streams; in addition, un-reacted elements have been detected, even though the stoichiometry is locally maintained [59].

Reactive Processes: gas-liquid state

The gas-liquid process belongs to the category of in-situ techniques. A gas is injected into the aluminum melt composed by one or more elements. Such gas reacts chemically with the melt and form the reinforcement phase (Fig.14). Refractory elements can also be added to the melt to tailor the precipitates. Table 5 shows gases, matrices and secondary phases that can be synthesized, together with the chemical reactions involved [60-66] (Fig. 15). Tyagi et al. [67] manufactured aluminum nitrides with a diameter smaller than 1 µm, by bubbling ammonia gas in an Mg-Al melt. The temperature was kept at 900 C° and the gas was purged for 70 minutes with a constant flow rate. Shyu et al. [65] bubbled methane gas in Al-Ti melt to form TiC particles. The yield strength increased up to 18 % and the hardness by 20%. The size of the particles was smaller than 0.1 µm.


Table 5. Gas-Liquid Process Gases, Matrices, Products and Reactions.

The process is characterized by:

- Negligible costs. Gas is relatively inexpensive [60]. The particles are found in-situ alleviating the cost of expensive second phase nano-particles;

- Surface contamination is eliminated thus enhancing interfacial bonding;

- The thermodynamics of the process can be controlled to suppress the formation of unfavorable phases [60,61].

- Homogeneous microstructures are obtained. The particles are naturally dispersed in the metal matrix, Fig. 15 [60].

Some limitations of the process are [65]:

- The temperatures necessary for the reaction to occur are high (1300-1600 K depending on the gas and the matrix); - High apparent viscosity hinders the production of high percentages of reinforcement;

- Process times may be lengthy as the kinetics are challenging; - The method is not applicable to materials with high melting temperatures.

Fig.14. Schematic of gas-liquid process [61]. Fig. 15. AlN particles in Al matrix via gas-

liquid process [60].


Morphological processes: rapid solidification

Nayak et al. [68] have melted under argon atmosphere Al-Fe alloys. Rapid solidification processing of the molten alloys was carried out by a single roller melt spinner with a copper wheel at different linear wheel speeds with cooling rates estimated to be in the range of 104–105 K/s.

Ultra-fine 100 x xAl Fe− precipitates embedded in the α-Al matrix were found in the melt spun Al–2.5

% Fe alloy as shown in Fig.16. Most of the precipitates here are less than 20 nm in size that structurally resemble some nanoquasicrystalline (NQ) phase. Increasing iron content up to 5% gives a cellular microstructure of around 150 nm in size. TiC have also been fabricated by melting a mixture of Al, Ti, and graphite powder under argon atmosphere [36]. Chill block melt spinning was used to prepare rapidly solidified samples in ribbon form. The TiC particles were found to be 40-80 nm in size and some clusters detected at the grain boundaries.

Fig.16. 100 x xAl Fe− precipitates embedded in the α-Al matrix [68].

Concluding Remarks

The various pathways to manufacture metal matrix nanocomposites have been presented and discussed in this critical review. It is quite clear that the challenges we face in manufacturing nanocomposites for structural applications are daunting. Scalability is a critical issue; there are many reported methods for producing small quantities in a laboratory setting. However, commercial production on a large scale is another matter. To be able to manufacture nanocomposites with a homogeneous distribution of the second phase nano-sized particles is also another critical issue. As presented and discussed in this review, this requirement remains to be the most difficult one especially for ex-situ processing methods. Homogeneous distribution of the nano-sized particles is more readily attainable via in-situ processing methods. Ex-situ methods are characterized by the difficulty to introduce the reinforcement in the melt and effectively disperse it (liquid state), as well as porosity and distortion in the final component (solid state). Lastly, cost is a major factor, as the processing method selected needs to be cost-effective. Composite materials (both micro- and nano-scale) are difficult to machine because of the wear action of reinforcement particles on the cutting tool. Therefore, there is the impellent need to select manufacturing methods which can provide near-net shape, so that the machining step could be eliminated. The knowledge of properties of the composite material, such as tribological properties, is fundamental for the design stage. Such data greatly differ from the matrix properties and have a consistent impact on the behavior of the final component. For instance, friction coefficients influence coupling and therefore lubrication between parts of an automotive assembly, as well as coefficients of thermal expansion have to be taken account when the cooling system of a component subjected to high temperatures is designed. The optimal method to determine such properties for nanocomposite materials needs to be established.


References

[1] S.Rawal: J.Mater. Vol. 53 (2001), pp.14-17.

[2] Information on

http://www.secat.net/answers_view_article.php?article=Aluminum_Matrix_Composites.html

[3] Information on http://www.almmc.com/

[4] Information on http://www.thefreelibrary.com/Increased aluminum in automobiles saving

weight and fuel consumption-a0159786251

[5] A.Wielgat: Automot. Ind. (2002).

[6] Information on http://news.alibaba.com/article/detail/metalworking/100187048-1-metals-

knowledge%253A-automotive-trends-aluminum%252C.html

[7] Information on http://www.thefreelibrary.com/Justifying Aluminum Metal Matrix Composites

in an Era of Cost Redution-a060015013

[8] M. El-Gallaba and M. Skladb: J. Mater. Process. Tech. Vol. 83 (1998), pp. 277–285.

[9] S. Jahanmir, M. Ramulu and P. Koshy in: Machining of Ceramics and Composites, edited by

CRC Press, (1999).

[10] Mod. Cast. (September 2010), p.3.

[11] P.E.C. Camargo, K. G. Satyanarayama and F. Wypych: Mater. Res. Vol. 12 (2009), pp. 1-39.

[12] S. M. Zebarjad, S. A. Sajjadi and E. Z. Vahid Karimi: Res. Lett. Mater. Sci. (2008), pp. 1-4.

[13] Information on http://www.materials.unsw.edu.au/NanoWeb

[14] Information on scales.colorado.edu/classes/MCEN5208_F2004/.../lit_Fischer.doc

[15] Information on http://etd.lib.nsysu.edu.tw/ETD-db/ETD-search-c/view_etd?URN=etd-0717106-184541

[16] Q. Zhang and D. L. Chen: Scripta Mater. Vol. 54 (2006), pp. 1321–1326.

[17] M. Cournil, F.Gruy, P. Cugniet, P. Gardina and H. Saint-Raymond in Model of Aggregation of

Solid Particles in Nonwetting Liquid Medium, Centre SPIN, URA CNRS 2021, Ecole

Nationale Supérieure des Mines de Saint-Etienne.

[18] Information on http://www2.ocean.washington.edu/oc540/lec02-26/

[19] G Cao, J. Kobliska, H. Konishi, and X. li: Metall. Mater. Trans. Vol. 39A (2008), pp. 880-886.

[20] S. Melis and A. Sorti: AIChE Journal Vol. 45 No. 7 (2009).

[21] Rhonda Lee-Desautels: Educ. Reso. for Part. Techn. 051Q-Lee.

[22] D. Bozic, M. Vilotijevic, V. Rajkovic and Z. Gnjidic: Mater. Sci. Forum Vol. 492 (2005),

pp.487-492.

[23] N. Matsunagaa, K. Nakahamaa, Y. Hiratab and S. Sameshimab: J. Ceram. Process. Res. Vol.

10 No. 3 (2009), pp. 319-324.

[24] S. Shamsudin, M. M. Wahid and S. Jamaludin: 1st International Conference on Sustainable

Materials (ICoSM2007) (2007).


[25] W. Zhou and Z. M. Xu: J. Mater. Process. Tech. Vol. 63 (1999), pp. 358-363.

[26] N.J. Fei, L. Katgerman and W. Kool: J. Mater. Sci. Vol. 29 No. 24 (1994), pp.6439-6444.

[27] C. Tekmen, F. Saday, U. Cocen and L.Y. Ljungberg: J. Compos. Mater. Vol. 42 No. 16

(2008), pp. 1671-1771.

[28] R. Kolhe, C.Y. Hui, E. Ustundag and S.L. Sass: Acta Mater. Vol. 44 No. 1 (1999), pp. 279-

287.

[29] S.Suresh, A. Mortensen and A. Needleman in Fundamentals of Metal Matrix Composites,

edited by Buttleworth-Heinemann, (1993).

[30] A. Evans, C. San Marchi and A. Mortensen in Metal Matrix Composites in Industry: An

Introduction and a Survey, edted by Springer, (2003).

[31] P. M. Ajayan, L. S. Schadler and P. V. Braun in Nanocomposite Science and Technology,

edited by Wiley-VCH, (2003).

[32] C. C. Koch in Nanostructured Materials: Processing, Properties, and Applications, edited by

William Andrew, (2006).

[33] A. Evans, C. San Marchi, A. Mortensen, “Metal Matrix Composites in Industry: an

Introduction and a Survey, Volume 1”, Springer, 2003.

[34] N. Chawla and K. Chawla in Metal Matrix Composites, edited by Birkhäuser, (2006).

[35] N.K.Tolochko, A.A.Andrushevic and Yu A. Shienok: Adv. Mater. Res. Vols. 79-82 (2009),

pp.425-428.

[36] G.Cao in Nanostructures & Nanomaterials: Synthesis, Properties & Applications, edited by

Imperial College Press, (2004).

[37] Z.Y. Ma, Y.L. Lia, Y. Liang, L F. Zheng , J. BP and S.C. Tjong: Mat. Sci. Eng. Vol. A219

(1996), pp. 229-231.

[38] J. Peng, U.S. Patent 7,297,310 (2007).

[39] C. L. De Castro and B. S. Mitchell in Synthesis, Functionalization and Surface Treatment of

Nanoparticles, edted by American Scientific Publishers (2002).

[40] D.L. Zhang, J. Liang and J. Wu: Mat. Sci. Eng. Vols. A375–377 (2004), pp. 911–916.

[41] Information on http://www.sigmaaldrich.com/etc/medialib/docs/Aldrich/Brochure/al_chemfile_v5_n3.Par.0001.File.tmp/al_chemfile_v5_n3.pdf

[42] L. Xiaodan: TMS 2010 139th Annual Meeting and Exhibition, Supplemental Proceedings,

Vol. 2, Materials Characterization, Computation, Modelling and Energy.

[43] Y.Yang and X.Li: J. Manuf. Sci. Eng. Vol.129 (2004), pp.497-501.

[44] Y.Yang, J.Lan and X.Li: Mat. Sci. Eng. Vol. A380 (2004), pp. 378-383.

[45] Y.Yang, X.Li and X.Cheng: J. Mater. Sci. Vol. 39 (2004), pp.3211-3212.

[46] P.Padhi, S.C.Panigrahi and S.Ghosh: AIF Conf. Proc. 1063 (2008), pp.371-375.


[47] R. Ashtana in: Solidification Processing of Reinforced Metals, Trans. Tech Publications,(1997).

[48] G.Kaptay in Proc. of the 2nd Int. Conf. High Temperature Capillarity, edited by

N.Eustathopoulos and N.Sobczak, published by Foundry Research Institute, Cracow, Poland

(1998), pp. 388-393 .

[49] S. Gierlotka: Solid State Phenom. Vols. 101-102 (2005), pp. 157-164.

[50] B. F. Schultz: Ph.D, The University of Wisconsin – Milwakee (2009), 208 pages, 3363450.

[51] Q.Zheng, B.Wu and R.G. Reddy: Adv. Eng. Mater. Vol. 5 No. 3 (2003), pp. 167-173.

[52] S.C. Tjong and Z.Y. Ma: Mater. Sci. Eng. Vol. 29 (2000), pp. 49-113.

[53] A.G.Merzhanov, A.S.Rogachev and A.E.Sychev: Dokl. Phys. Chem. Vol.362 No. 1-3 (1998),

pp.217-221.

[54] Information on www.spaceflight.esa.int/.../combustion.jpg

[55] C.M.Ward-Close, R. Minor and P.J. Doorbar: Intermetallics Vol. 4 (1996), pp.217-229.

[56] J.Tu, N.Wang, Y.Yang and W.Qu: Mater. Lett. Vol. 52 No. 6 (2002), pp. 448-452.

[57] Y.Zhao, S.Zhang, G.Chen, X.Cheng and Q.Wang: Compos. Sci. and Techn. Vol. 68 (2008),

pp. 1453-1470.

[58] Nam P. Suh Sutek Corporation, U.S. Patent 4278622 and 4279843 (1981).

[59] Nam P. Suh Sutek Corporation, U.S. Patent 4,706,730 and 4,890,662 (1987).

[60] Q.Zheng and R.G.Reddy: Metall. Mater. Trans. Vol. 34B (2003), pp. 793-805.

[61] Q. Zheng, B.Wu and R.G. Reddy: Adv. Eng. Mater. Vol. 5 No. 3 (2003), pp. 167-173.

[62] Q. Zheng and R.G. Reddy: J. Mater. Sci. Vol. 39 (2004), pp. 141-149.

[63] R. G. Reddy, US Patent 6,343,640 (2002).

[64] R.F. Shyu, F.T. Weng and C.T. Ho: J. Mater. Process. Technol. Vol. 122 (2002), pp. 301–304.

[65] R.F. Shyu and C.T. Ho: J. Mater. Process. Technol. Vol. 171 (2006), pp. 411–416.

[66] M.Dyzia and J.Sleziona: Arch. Mater. Sci. Eng. Vol. 31 (2008), pp.17-20.

[67] S.Tyagi, Q.Zheng and R. Reddy: Aluminum 2004, edited by S. K. Das, TMS, Warrandale, (2004), pp. 63-72. [68] S. S. Nayak, B.S.Murty and S.K.Pabi: Bull. Mater. Sci. Vol. 31 No. 8 (2008), pp. 249-254.


Micro-chemistry and mechanical behaviour of Ti6Al4V-SiCf composite produced by HIP for aeronautical applications

P. Deodati1,a, R. Donnini1,b, S. Kaciulis2,c, M. Kazemian-Abyaneh3,d, A. Mezzi2,e, R. Montanari1,f, C. Testani4,g and N.Ucciardello1,h,

1 Department of Mechanical Engineering, University of Rome “Tor Vergata”, Via del Politecnico 1,

00133 Rome, Italy

2 Institute for the Study of Nanostructured Materials, ISMN-CNR, P.O. Box 10, 00016 Monterotondo Stazione, Rome, Italy

3 Sincrotrone Trieste SCpA, SS14-Km163.5 in Area Science Park, 34149 Trieste, Italy

4 Centro Sviluppo Materiali (CSM), Via di Castel Romano 100, 00128 Rome, Italy

[email protected], [email protected], [email protected], [email protected], [email protected], [email protected],

[email protected] , [email protected],

Keywords: Ti6Al4V-SiCf composite, matrix-fibre interface, microstructural stability, mechanical properties, anelastic behaviour

Abstract. The paper reports the results of an extensive characterization of the Ti6Al4V-SiCf composite produced by hot isostatic pressing (HIP) to assess its capability to withstand the in-service conditions of turbine blades operating at middle temperatures in aeronautical engines. The microstructure of composite, in as-fabricated condition and after long-term heat treatments (up to 1,000 hours) in the temperature range 673-873 K, has been investigated by means of different techniques. Particular attention was paid to the micro-chemical evolution of fibre-matrix interface which is scarcely affected also by the most severe heat treatments examined here. This leads to stable mechanical properties as evidenced by hardness, tensile and FIMEC instrumented indentation tests. Therefore, the composite can operate at the maximum temperature (873 K) foreseen for its aeronautical applications without remarkable modifications of its microstructure and degradation of mechanical properties. The mechanical characterization has been completed by internal friction and dynamic modulus measurements carried out both at constant and increasing temperature, from 80 to 1173 K.

Introduction

In the last years great efforts have been devoted to the development of composites with titanium alloy matrix reinforced by unidirectional long fibres [1-6] because they exhibit an excellent strength/weight ratio. The Ti6Al4V-SiCf composite is a promising material for mechanical components operating at middle temperatures, especially turbine blades and structural high stressed parts of aeronautical engines. The performances mainly depend on the fibre-matrix interface and the chemical reactions occurring during fabrication process and in-service life, when temperatures up to 873 K are reached for long exposure time. Fig.1 shows the stratified structure of the SiC fibres (SCS-6 type, Φ = 0.14 mm) used to produce the material. A carbon layer (thickness = 3 µm) separates the SiC fibre from the Ti6Al4V matrix, that consists of α phase (hcp) plus few percent of β phase (bcc). Under thermo-mechanical stresses diffusion phenomena and chemical reactions may occur at the fibre-matrix interface causing its degradation and leading to mechanical instability of the composite [7,8]. Therefore, the target of this work has been to assess the real suitability of the material for the


foreseen applications, by means of an exhaustive microstructural and mechanical characterization before and after long-term heat treatments in the 673-873 K range.

Fig. 1. Section of the composite: internal structure of a SCS-6 fibre embedded in the Ti6Al4V matrix.

Composite production

The composite has been produced by Hot Isostatic Pressing (H.I.P.) at C.S.M. laboratories. As schematically shown in Fig.2, it was prepared by hot isostatic compaction into autoclave of metallic sheets alternated by SCS-6 fibre layers. At first preforms have been produced by overlap of 4 layers of unidirectional fibres alternated with 5 plates of Ti6Al4V, being the thickness of the two external plates of 0.6 mm and the three internal ones of 0.1 mm. The approximate thickness of the fiber layers is 0.14 mm. The sheet dimensions were 450 mm x 200 mm. The preforms have been then placed inside a AISI 304 steel die.

Fig. 2. H.I.P. process for the fabrication of the composite.

FIBRE

MATRIX

SiC

C coating

C core


The internal sheet surface and the die cap have been covered by boron nitride to prevent diffusion welding thus to safely remove the composite from the die after HIP process. Before processing the dies have been welded by a TIG process and subjected to a vacuum-sealing cycle (3 x 10-6 mbar). The HIP process has been carried out by an hot isostatic press type ASEA-QH21, that permits an independent control of temperature and pressure parameters. Process parameters (Tmax = 1163 K, Pmax = 1200 bar) have been chosen on the basis of literature data [6-10]. The cooling to room temperature has been performed after the extraction of composite panels from the die. The samples (size 10 mm x 20 mm) for the experimental tests have been cut from the sheets by spark erosion.

Heat treatments

Composite samples have been subjected to the heat treatments listed in Table 1, which were carried out in vacuum (P = 2 x 10-5 mbar).

Table 1. Heat treatments carried out on the composite samples.

Microstructural characterization

The microstructure of Ti6Al4V-SiCf composite, before and after the heat treatments, has been examined by X-ray diffraction (XRD), transmission electron microscopy (TEM), energy dispersion spectroscopy (EDS), X-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES) and scanning photoemission microscopy (SPEM) [11]. XRD analysis. XRD measurements have been carried out at increasing temperatures by means of an Anton Paar HT-16 camera mounted on a Philips diffractometer. Spectra have been collected up to 873 K in an atmosphere of argon of commercial purity. Before each measurement the samples were kept 1.8 x 103 s at the temperature test to guarantee thermal homogeneity. Spectra were collected by using the Mo-Kα radiation (λ = 0.71 Å) in step-scanning mode with 2Θ steps of 0.05° and counting time of 2 s per step in the angular range 10°-50°. High precision peak profiles of the most intense reflections of α phase were recorded with 2Θ steps of 0.005° and counting time of 20 s per step. From the peak positions of the {100}, {002}, {101}, {102}, {110} and {103} reflections the interplanar spacings dhkl have been determined at different temperatures; for comparison the same tests have been carried out on the monolithic Ti6Al4V alloy. For the composite the strain εhkl on {hkl} planes has been calculated by:

(1)

where d0 is the interplanar spacing of the alloy at the same temperature. Fig.3a shows how the strain strongly depends on the specific {hkl} set of planes. The cell parameters, a and c, of alloy and composite have been also determined from dhkl and are displayed in Fig.3 (b,c).

S1 S2 S3 S4 S5 S6 S7

Temperature [K] as - fabricated 673 673 673 873 873 873

Time [x 105 s] - 3.6 18 36 3.6 18 36

0

0

d

ddhklhkl

−=ε


Fig. 3. Strain εhkl measured by XRD for different {hkl} planes of composite at increasing temperature (a), lattice parameters a and c of alloy and composite at temperatures up to 873 K

(b,c)

Experimental data of Fig.3 (b,c) can be fitted by the following relationships:

a = a0 (1+α ∆T) (2) c = c0 (1+β ∆T) where a0 and c0 are the cell parameters at room temperature (300 K), α and β the expansion coefficients, ∆T the temperature increase. The values obtained by the best fitting are reported in Table 2.

Table 2. Ti6Al4V alloy and composite: cell parameters (a0, c0) at room temperature and thermal expansion coefficients (α, β).

a0 [nm] c0 [nm] αααα [°C-1

] ββββ [°C-1

]

Ti6Al4V 0.2934 0.4681 2.32 x 10-5 2.45 x 10-5 Composite 0.2933 0.4679 1.36 x 10-5 2.14 x 10-5

b) c)

a)


At room temperature, the a0 and c0 values of composite are very close to those of the alloy. When temperature rises the cell parameters linearly increase but with different rates, namely the h.c.p. unit cell expands modifying its shape. Moreover, the expansion of composite is lower than that of monolithic alloy because the fibres, whose thermal expansion coefficient is 4.1 x 10-6 °C-1, represent a constraint for the free expansion of the surrounding matrix. Results of analogous experiments performed on the monolithic Ti6Al4V alloy are reported in literature [12-15]. Expansion coefficients determined from tests in vacuum (pressure of 2.5 x 10-3 mbar) [13], α = 1.043 x 10-5 °C-1 and β = 1.448 x 10-5 °C-1, are lower than those determined in the present experiments because argon of commercial purity used by us contains residual amounts of oxygen and nitrogen, which are absorbed by the metal during the test in temperature. Cell expansion at high temperature has two components: one is thermal, the other one is due to gas absorption. The change of a and c in titanium as a function of the amount η of interstitial atoms in the lattice, da/dη and dc/dη respectively, are 9.0 x 10-5 and 4.0 x 10-4 for oxygen, 2.0 x 10-4 and 6.7 x 10-4 for nitrogen [16], being the values of da/dη and dc/dη expressed in nm (at %)-1. More details about the effects of gas absorption in Ti6Al4V alloy can be found in [17]. Unlike the monolithic alloy, the composite is not free to expand at high temperature due to fibre constraint. Gas absorption involves a major expansion perpendicularly to {002} planes because dc/dη > da/dη thus the grains with [002] direction parallel to the fibre axis are more affected by constraining than those with [002] direction perpendicular. So, from the different values of α and β of composite and alloy, it is possible to conclude that gas is preferentially absorbed by those grains with a favourable orientation with respect the fibres ( [002] perpendicular to major fiber axis). XRD analyses permitted to determine the dislocation density of composite and alloy. Peak profiles of the composite are always narrower than those of the alloy; for example, Fig.4 displays the {100} reflections whose intensities have been normalized for making the comparison easier. The half-height line widths β have been determined and corrected from instrumental broadening. For each XRD reflection the total line broadening βT is basically due to two contributions, the size of coherently diffracting domains ( βD ) and the micro-strains ( βε ). βT can be written as:

(3)

where D is the domain size, ε the average micro-strain, ϑ the Bragg angle, λ the wavelength and K a constant = 0.89 . In the case of Ti alloys the coherently diffracting domains are the grains, which from metallographic observations result to be of similar size (D ≈ 30 µm) in alloy and composite. Therefore, the major peak broadening of alloy can be ascribed to a higher density of dislocations. Being the average grain size very large, the βD term in Eq.(3) can be neglected. Introducing the measured βT values of the main XRD reflections into the simplified Eq.(3), the micro-strain ε has been determined. The dislocation density ρ was calculated by means of the Williamson-Smallman relationship:

ρ = Ξ ε2 / k0 b2 (4) where Ξ =16 is a constant, b is the Burgers vector modulus and k0 ≅ 1 is a factor depending on dislocation interaction. The calculation gave the values of 6.1 x 109 cm-2 for the composite and 5.5 x 1010 cm-2 for the monolithic alloy.

ϑεϑ

λβββ ε tan2

cos+=+=

D

KDT


40,5 41,0 41,5 42,00

500

1000

1500

2000

2500

3000

X-r

ay in

ten

sity (

a.u

.)

2Θ (deg)

Alloy

Composite

100

Fig. 4. Precision XRD peak profiles of alloy and composite.

TEM observations. The specimens have been prepared by mechanical polishing up to ~180 µm, a thickness just a little larger than the fibre diameter, then etched using a Tenupol apparatus with a reagent 100 ml HClO4 and 900 ml CH3OH at -20 °C, V= 20 mV. TEM confirms the complex structure of the fibre-matrix interface in the as-prepared composite. Fig.5a shows titanium carbide (TiC) particles with mean size of about 200 nm forming a layer of irregular thickness around the fibres. The particles grow during HIP process due to carbon diffusion at high temperature from the fibre graphite coating toward the Ti6Al4V matrix and subsequent chemical reaction with titanium. Fig. 5b displays the matrix structure near the TiC layer surrounding the fibres. The distribution of defects is not homogeneous and some bands with very low dislocation density are observed. Dislocations induced by plastic flow around the fibres by HIP are in large part recovered by the high temperature soaking and subsequent slow cooling with a resulting final density lower than that of the monolithic alloy, as evidenced by XRD measurements.

a) b)

Fig. 5. TiC particles form a layer of irregular thickness around the fibres (a). Bands with low dislocation density (b).


EDS, XPS, AES and SPEM analyses. Mechanical performances of composite strongly depend on the fibre-matrix interaction in fabrication process and in-service life where temperatures up to ~ 873 K are reached. Direct contact of Ti6Al4V matrix with SiC gives rise to brittle compounds like Ti5Si3, which deteriorate the mechanical behaviour of composite [18-19], therefore the fibres are coated with a thin (∼ 3 µm) carbon layer. Carbon coating hinders chemical reactions, maintains the integrity of fibres, reduces the interfacial debond strength and deflects the propagation of microcracks along the fibre. However, when the composite is operating for a long time at middle-high temperatures (up to 873 K), carbon may diffuse into the matrix promoting the formation of TiC, according to ref. [1] and TEM observations. Therefore, the interface stability has been examined by EDS, XPS, AES and SPEM, before and after the heat treatments reported in Table 1. The analyses have been performed on different areas of the samples: fibres, fibre-carbon interface, carbon, carbon-matrix interface and matrix. To extend the observation zones, some samples have been prepared by mechanical polishing with surfaces forming a small angle (≈ 2°) with the major fibre axis (Fig.6).

Fig. 6. Image of a surface forming a angle of ≈ 2° with the major fibre axis (a) and corresponding sketch of the SCS-6 fibre structure (b).

The thickness of the reaction zone (see Fig.7), i.e. the zone between carbon coating and matrix affected by chemical reactions, has been measured by SEM. Since local irregularities are present at fibre boundary a statistical approach (at least 10 measurement in 8 different fibre-matrix positions rotated of 45°) has been used. The results, reported in Table 3, demonstrate that the growth of the reaction zone is quite slow. For example, its thickness, which is 0.85 µm in the as-fabricated material, becomes 0.98 µm after 1,000 hours at 873 K. Moreover, it has been observed that long-term heat treatments do not substantially affect grain size that remains of about 30 µm.

b) a)


Fig. 7. Measurements along different directions (a) permitted to determine the mean thickness of the reaction zone shown in (b).

Table 3. Thickness of the reaction zone after the heat treatments listed in Table 1. Fig.8a shows the SEM image of the as-fabricated sample, mechanically polished to get a surface forming an angle of ≈ 2° with the major fibre axis. EDS measurements have been carried out along the direction indicated by the arrow from P1 to P10, with a distance of 10 µm between each test position. The concentrations of carbon, titanium and silicon are displayed in Fig.8b.

Fig. 8. As-fabricated composite: SEM micrograph of a fibre-matrix zone (a) and elemental concentration profiles measured by EDS microanalysis along the arrow direction (b).

Heat treatment S1 S2 S3 S4 S5 S6 S7

Mean thickness [µµµµm] 0.85 0.86 0.95 0.98 0.94 0.97 0.98

a) b)

P1 P2 P3 P4 P5 P6 P7 P8 P9 P10

0

10

20

30

40

50

60

70

80

90

100

% A

t.

C

Si

Ti

b) a)


The diameter of the interaction volume was 1.5 µm in the matrix and 4 µm in the carbon coating, a resolution sufficient to avoid experimental artefacts. In spite of some surface irregularities produced by mechanical polishing in a material with phases of different mechanical characteristics, some important aspects can be observed: there is a transition region (P8-P10) where progressively the signal of carbon decreases while that of titanium increases, silicon has been found in graphite (P3-P9) and silicon signal disappears at the graphite-matrix boundary. In order to get a better lateral resolution, XPS and AES measurements have been carried out. The main features of the XPS and AES techniques have been described elsewhere [20-22]. Photoelectron images were registered by using an Axis Ultra spectrometer (Kratos Analytical), equipped with a monochromatized Al Ka X-ray source (1486.6 eV). Charge neutralization system was used to maintain the binding energy (BE) scale at 285.0 eV for the hydrocarbon C 1s peak. X-ray photoelectron and Auger spectra were collected by using an Escalab Mk II spectrometer (VG Scientific) equipped with 5-channeltron detection system. Photoelectrons were excited by using a standard Al Kα excitation source, while Auger electrons were excited by using an electron gun LEG 200, operated at 10 keV and 1 – 10 nA current. XPS spectra were registered at constant pass energy of 20 eV, while AES spectra were registered in a constant ratio retard (1:2) analyzer mode. All the experiments were performed at a base pressure below to 1 x 10-10 mbar. The depth profiling was carried out by an EX-05 Ar+ ion gun set at 1.0 keV energy. A resistive heater placed in the preparation chamber (1 x 10-9 mbar) was used for the thermal treatments of the samples up to the temperature of 873 K. The BE scale was calibrated positioning the C 1s peak from graphite at 284.6 eV. XPS data were processed by the CasaXPS v.2.2.84 software, using a peak-fitting routine with symmetrical Gaussian–Lorentzian functions. The background was subtracted from the photoelectron spectra by using Shirley method. XPS chemical images of a single fibre (in cross-section) are displayed in Fig.9 (a, b). The image of graphitic carbon (a) has been acquired by separating the chemical species identified as C 1s peak at BE = 284.6 eV [23]. The Si image (b) is related to Si 2p peak at BE = 99.9 eV, that is typical of SiC [21]. The outer and inner borders of SiC ring confine with graphite. The presence of lower amount of graphite in the zone of SiC ring can be explained as carbon contamination.

Fig. 9. XPS chemical images (235 x 235 µm2) of: (a) C 1s at BE = 284.6 eV,

(b) Si 2p at BE = 99.9 eV. A higher lateral resolution (∼ 200 nm) has been achieved by the multipoint AES examinations performed for all the samples. The AES spectra collected across the interface showed that the heat treatments induced the interdiffusion of titanium in carbon and viceversa [24]. Unfortunately, the resolution of XPS and AES did not allow to separate the contributions of SiC and TiC, therefore it was difficult to distinguish the possible formation of TiC at the outer border of carbon layer.


Seeking to improve further the resolution, spatially resolved XPS images were acquired at the ESCA microscopy beamline of the ELETTRA synchrotron light laboratory in Trieste by using a SPEM, equipped with zone-plate focusing optics which provides an X-ray nanoprobe with a diameter of 150 nm. Photoemission spectra of selected regions and chemical maps were acquired with 0.2 eV energy resolution by using 650 eV photon energy. More details on this microscope have been reported elsewhere [11, 25-26]. Fig.10 shows the chemical maps of the C 1s peak at BE = 284.6 eV (a) and Ti 2p peak at BE = 455.0 eV (b), acquired across the fibre-matrix interface. The spectra of Al 2p, C 1s, Si 2p and Ti 2p were collected in the points A, B and C, marked in the Fig.9. The metal matrix (point A) is characterized by the presence of C, Ti and Al. The spectra of vanadium are not reported here, because the main V 2p peak was very low and was overlapping with the high O 1s peak, present as a consequence of the atmospheric oxidation. After a cycle of sputtering with Ar+ ions, the concentration of oxide was drastically reduced, but still it was not low enough in order to study the V 2p peak. The deconvolution of C 1s spectra revealed two main components (see Fig.11a). The first one, positioned at BE = 284.6 eV, is attributed to graphitic C–C bond, while the second one at BE = 282.6 eV is assigned to Ti-C bond at BE = 281.6 – 282 eV and/or Al-C bond at BE ≈ 282.4 eV [21]. Ti-C and Al-C bonds can be due either to single pairs of metal-carbon atoms or to carbides. The comparison of the C 1s spectra, acquired in the points A, B and C (graphite layer), revealed a change of the intensity of the component at lower BE. In particular, the data evidenced a higher concentration of metal-carbon bonds in the matrix near the carbon coating. The same trend was also verified by the investigation of Ti 2p signal (Fig.11b). The most intense Ti 2p signal was determined in the point B. The peak-fitting revealed a Ti 2p3/2 component centered at BE = 455.0 eV, that is characteristic for the Ti – C bond. However, the assignment of this component is not certain due to the possible presence of TiO created by ion sputtering. Finally, Fig.12 shows the Al 2p spectrum acquired in the points A of the Fig.9. The signal was deconvoluted by using the Shirley background and the spin–orbit doublets of 2p3/2 and 2p1/2 for the two species: Al–Al at BE = 72.2 eV and Al2O3 at BE = 74.4 eV.

Fig. 10. SPEM chemical images (23 µm × 2.1 µm) of C 1s (a) and Ti 2p (b), corrected for the sample topography by using the formula (Peak-Background)/Background.

a)

b)


Fig. 11. Photoemission spectra of C1s (a) and Ti2p (b) in the points A, B and C, which are marked

in the SPEM image (Fig. 10a).

Fig. 12. Peak fitting of Al 2p spectrum collected in the point A of the Fig. 10a: peak 1 - BE = 74.6 eV; peak 2 - BE = 72.4 eV.


0

10

20

30

40

50

60

70

80

90

100

0,0E+00 2,5E-03 5,0E-03 7,5E-03 1,0E-02 1,3E-02 1,5E-02

x [µµµµm]

C (

%)

b)

D0 = 4,1x10-8 m2s-1 Q = 207 kJmole-1

0

10

20

30

40

50

60

70

80

90

100

0 2 4 6 8 10

x [µµµµm]

C (

%)

D0 = 5,1x10-4 m2s-1 Q = 182 kJmole-1

a)

To better understand the carbon-titanium interdiffusion occurring at fibre-matrix interface some physical simulations have been carried out. Ti6Al4V and Ti 99.99 foils have been coated by thin graphite films (thickness of about 30 nm), heated in vacuum at 773 K for 8 hours and then examined by XPS profiling. XPS depth profiles of C-covered Ti 99.99 samples before and after the heating in vacuum have been recorded and compared. These depth profiles are plotted in Fig.13 as the peak areas of Ti 2p and components of C 1s: titanium-carbon bond (carb.) at BE = 282.0 eV and graphite at BE = 284.6 eV.

Fig. 13. XPS depth profiles of the carbon covered Ti99.99+ sample at room temperature (R.T.) and after heating for 8 hours at 773 K.

After sample heating, the signal of titanium-carbon bond in the C-matrix interface was increased, moreover, the thickness of graphite layer was reduced. Considering that the ion sputtering rate was about 0.2 nm/min, the width of reactive zone due to thermal process can be estimated as about 10 nm. Similar results have been obtained from the study of the depth profile of the C-covered Ti6Al4V sample.

Fig. 14. Concentration profiles calculated for the diffusion of carbon in Ti-α (a) and in TiC (b) after 8 hours at 773 K.


0,0 0,1 0,2 0,3 0,4 0,50

1000

2000

3000

4000

5000

6000

7000

Pre

ssu

re (

MP

a)

Depth (mm)

pY

= 3528 MPa

Experimental depth profiles can be compared with carbon concentration profiles calculated by the 2nd Fick’s law in the case of a semi-infinite solid. The calculated profile for the diffusion of carbon in Ti-α after 2.88 x 104 s at 773 K is shown in Fig.14a. The values of the pre-exponential factor D0 and of the activation energy Q used in the calculation have been taken from ref. [27] and are displayed in Fig. 8a. Carbon diffuses over a distance of about 10 µm, which is about 3 orders of magnitude higher than the experimental one (10 nm) thus the carbon concentration in the interface can not be explained by the process of carbon diffusion in the Ti-α matrix. The same calculation has been performed for the diffusion of carbon in TiC (Fig. 14b). The values of D0 and Q have been taken from ref. [28]. In this case carbon diffuses over a distance of about 12 nm, a value in a good agreement with the experimental one. The comparison between experimental and calculated data suggests that carbon diffusion in Ti-α occurs only at the beginning of the process, then carbon reacts with titanium, forming a layer of TiC, which separates graphite from metallic matrix. The interface growth is governed by the diffusion of carbon in TiC that is much slower than diffusion of carbon in Ti-α.

Mechanical characterization

The mechanical properties of the composite in as-fabricated condition and their evolution after heat treatments have been investigated by means of instrumented penetration, tensile and hardness tests. Moreover, the mechanical characterization has been completed by internal friction (IF) and dynamic modulus measurements carried out from 80 to 1173 K to describe the anelastic behaviour of the material.

FIMEC test. FIMEC (Flat-top Cylinder Indenter for Mechanical Characterization) is an instrumented indentation test employing a cylindrical punch (diameter = 1 mm and axial length = 1.5 mm). During the test applied load and indentation depth are recorded. When load is divided by contact area pressure-penetration curves are obtained. The characteristics of the method have been described in detail in several papers, e.g. [29-32]. Fig.15 shows the FIMEC curve of composite tested at room temperature. For p > pY the curve is characterised by a sudden slope decrease and the material starts to protrude around the imprint.

Fig. 15. FIMEC curve of the as-fabricated composite recorded at room temperature. When indentation test is carried out with a penetration rate of 1.7 x 10-3 mm s-1 or lower, it is possible to compare directly indentation data with those of tensile tests made with a strain rate of


10-3 s-1; in these conditions the yield stress σy ≅ pY/3. The relationship has been verified for a lot of pure metals, alloys and composite materials; the relative difference, ∆ = (σY − pY/3) / σY , between pY/3 values coming from indentation curves and σY values from tensile tests does not exceed 7% , i.e. it is similar to data scattering observed in different tensile tests on the same material [33]. In Fig.15 the slope change occurs for a pressure pY = 3528 MPa thus σY = pY / 3 = 1176 MPa. The corresponding value obtained by tensile tests is 1154 MPa thus ∆ = (σY − pY/3) / σY is ∼ 2%. FIMEC tests have been carried on composite and, for comparison, on the monolithic Ti6Al4V alloy at increasing temperature up to 773 K. The curves are reported in Fig.16 (a,b) while Table 4 reports the yield stress values (σY = pY / 3) obtained from the tests. The results show that the yield stress of the composite is always higher than that of the matrix alloy but the difference becomes progressively smaller as temperature increases.

Fig. 16. FIMEC curves of as-fabricated composite (a) and monolithic Ti6Al4V alloy (b).

b)

a)


Table 4. Yield stress σY, expressed in [MPa], of as-fabricated composite and monolithic Ti6Al4V alloy from FIMEC tests at increasing temperature

Tensile and hardness tests. The probes for tensile tests (ASTM E21 standard) have been cut from the composite sheets. Fig.17 shows the results of tests carried out at room temperature and 873 K on probes in as-fabricated condition and after heat treatments in vacuum (P = 5 x 10-6 mBar) at 873 K with exposure times of 3.6 x105, 18 x105 and 36x105 s (the treatments listed in Table 1). The yield stress and ultimate tensile strength are not affected by heat treatments also in the most severe conditions.

Fig. 17. Yield stress (Y0.02%) and ultimate tensile strength (UTS) at room temperature and 873 K of the composite in as-fabricated condition and after the heat treatments.

Fig.18 shows the fracture surface of a probe heated 36 x 105 s at 873 K. The fracture surface is not planar with several pull-out of fibres and the external mantle of the fibres shows a clear reaction with the matrix indicating a correct load transfer from the matrix to the fibres. The main reason of such mechanical performances is the stability of fibre-matrix interface after long-term heat treatments. In fact, as discussed before, a thin TiC layer forms all around the carbon coating during the fabrication process and hinders further carbon diffusion towards the matrix retarding interface degradation. Another factor which contributes to preserve the mechanical properties is the grain size stability after long-term heat treatments.

R.T. 373 K 473 K 573 K 673 K 773 K

Composite 1176 849 807 722 595 550

Ti6Al4V alloy 904 807 747 671 569 548


Fig. 18. Fracture surface of a probe exposed at 873 K for 36 x 105 s.

Elastic and anelastic behaviour

The elastic and anelastic characteristics of the material have been investigated in an extended range of temperature from 80 to 1173 K by dynamic modulus and internal friction (IF) measurements carried out on bar-shaped samples mounted in free-clamped mode using the method of frequency modulation. The VRA 1604 apparatus used in the experiments has been described in detail in [34]. The tests have been performed with resonance frequencies f in the range 600-3500 Hz while strain amplitude was kept lower than 1 x 10-5.

A set of experiments have been made at increasing temperature with heating rate of 1.7 x 10-2 Ks-1, another one isothermally for 8.64 x 104 s at different temperatures up to 873 K. To simplify the discussion, the results will be presented in three points: a) tests at increasing temperature above room temperature, b) tests at increasing temperature below room temperature, c) tests at constant temperature. Tests at increasing temperature above room temperature. The dynamic modulus E is proportional to f 2:

(5)

where m is a constant (m=1.875), L the length of vibrating reed, h its thickness and ρ the material density. The trends of composite and monolithic alloy vs. temperature are displayed in Fig.19. The modulus of composite is always higher (∼ 20%) than that of the matrix alloy up to 873 K.

ρπ

E

L

hmf

2

2

122=


Fig. 19. Dynamic modulus of Ti6Al4V alloy and composite at increasing temperature up to 873 K.

Fig.20 shows the Q-1 spectrum of monolithic alloy and composite above room temperature. That of composite displays a peak superimposed to an exponentially increasing background.

Fig. 20. Comparison of Q-1 and (f/f0)2 vs. T trends of composite (f0 = 898 Hz) and monolithic alloy.

The composite spectrum is the superposition of a Debye peak and an exponential background. The IF spectrum vs. temperature T can be fitted by the sum of two contributions: an exponential curve for the background, 1−

BQ (T), and a single Debye peak, 1−PQ (T):

(6)

−

∆+=+= −−−−

P

BPBTTR

HhTQTQTQTQ

11sec

2)()()()( 1111


being ∆/2 the peak maximum, H the activation energy of the physical process giving rise to the IF peak, R the gas constant and TP the temperature of peak centre. Since the peak position TP depends on the resonance frequency f , it is a relaxation peak. For a relaxation peak TP and f obey to the following relationship:

(7)

Therefore, the activation energy H and the pre-exponential factor τ0 can be determined by tests at different resonance frequencies. From the Arrhenius plot in Fig.21, the values of H = 186 kJ mol-1 and τ0 = 2.3 x 10-15 s have been obtained.

Fig. 21. Arrhenius plot for determining H and τ0.

The IF curve of the alloy shows only an exponential background which is higher than that of the composite. The result can be explained by considering a different contribution to background from dislocation damping in the two materials: the grain size is near the same in both of them while they have a different dislocation density, which is about one order of magnitude higher in the alloy. The IF spectrum of the composite exhibits a Debye peak, not observed in the alloy, so its origin is connected to the presence of the fibres and their effects on the surrounding matrix. To identify the physical origin of this peak, some hypotheses have been considered. Several phenomena, giving rise to energy loss, may occur at the fibre-matrix interface, when the composite is subjected to thermal and/or mechanical stresses. At high stresses, plastic flow and interface de-bonding may occur, while at low stresses, the matrix and the fibres undergo only elastic distortions. The effects of plastic flow and fibre-matrix de-bonding on damping have been discussed by Schaller [35]. Due to the different thermal expansion of matrix and fibres, internal stresses arise during the production process, when the samples are cooled from HIP temperature of 1163 K to room temperature. On the other hand, a simple calculation shows that these stresses are not sufficiently high to induce the formation of dislocations. The mean stress αM in the matrix caused by cooling from HIP temperature (THIP) to room temperature (TR) can be expressed by:

(8)

02 1P

H

RTf eωτ π τ= =

))(())1(( RHIPFM

MF

MFM TT

EE

EE−−

−+= ααφ

φφσ


where EF , αF and EM , αM are the Young’s modulus and the coefficient of thermal expansion of fibre and matrix, respectively; φ is the volume fraction of fibres. Introducing in Eq.(8) the values of EF = 400 GPa, EM = 114 GPa, αM = 9.6 x 10-6 K-1

, αF = 4.1 X 10-6 K-1 and φ = 0.3, was obtained

σM = 350 MPa. This value is much lower than the matrix yield stress (904 MPa). Therefore, plastic flow and de-bonding at the fibre-matrix interface can be ruled out as possible causes of the IF peak. The attention has been focused then on elastic strains at the interface. This condition has been analysed by He and Lim [36] on the basis of the interfacial diffusion mechanism [37]. When a shear stress is applied to the interface, which is not perfectly planar on a microscopic scale, it is under tension at some locations and under compression at some others, as schematically shown in Fig.22.

Fig. 22. Schematic view of the fibre-matrix interface, which is not perfectly planar on a microscopic scale. When a shear stress is applied some locations are in tension, others in compression.

The stress gradient induces the atom diffusion along the interface, causing anelastic behaviour. However, the activation energy H = 186 kJ mol-1, determined from present experiments, is quite different from those controlling the atomic diffusion of Ti, Al and V in both α and β phases, present in the matrix, thus the IF peak can not be ascribed to such a mechanism. Also the diffusion processes inside the fibres, which have a stratified axial-symmetric structure, are not compatible with the peak activation energy because H = 318 kJ mol-1 for C in SiC [38] and H = 911 kJ mol-1 for Si in SiC [39]. As shown in Fig.1, the fibres are coated by a carbon layer (thickness ≈ 3 µm) which separates SiC from the matrix. During the process of composite fabrication, carried out at high temperature (1163 K), carbon reacts with titanium, forming a thin layer (few nanometers thick) of titanium carbide (TiC) [6, 7, 9]. TiC thickness is increasing when the material is heated, but its kinetics is very slow. The activation energy for TiC growth, independently determined by Naka et al. [40], is of 194 kJ mol-1, i.e. it is very close to that of IF peak. Therefore, the peak seems connected to the growth of the TiC layer between the carbon coating and the matrix. Furthermore, it is necessary to consider that TiC activation energy is very close to that of carbon diffusion in the α phase of Ti (H* = 182 kJ mol-1). The distribution of elemental composition determined by XPS

ττττ


depth profiling microchemical profiles testifies that in the matrix around the fibre exists an extended zone (at least 100 nm wide), where the content of carbon is relatively high. Of course, this process can not occur in monolithic alloy, where a very low concentration carbon is homogeneously distributed. On these grounds, it is possible to suppose that the IF peak is caused by the stress-induced reorientation of i-s pairs (C-Al and C-V) in the hcp α phase of the matrix near the fibres. Fig.23 illustrates the presence of i-s pair in the α phase. This mechanism for hcp metals has been discussed by Gupta & Weining [41] and Povolo & Bisogni [42].

Fig. 23. Interstitial substitutional (i-s) pair in the h.c.p. lattice of the composite α phase.

Tests at increasing temperature below room temperature. Anelastic behaviour below room temperature has been described and discussed in detail in a previous paper [43]. Fig.24a shows Q-1 and (f/f0)

2 (f0 is the value at the lowest temperature when the measurement starts) vs. temperature for the monolithic alloy and the as-prepared composite. A peak is observed in both materials at approximately 120 K (f = 1 kHz) with an activation energy of 20.3 kJ mol-1; the relaxation strength is significantly higher in the composite in comparison with the monolithic alloy. A second very broad peak (activation energy of 48.2 kJ mol-1) localized at 250 K in the alloy and 270 K in the composite spectra has been observed, too. The background damping in the whole temperature range investigated appears higher in the composite. After thermal ageing up to 900 K in vacuum (10–4 Pa) a remarkable reduction of peak relaxation strength (up to 50%) in both the monolithic alloy and composite is experienced (Fig. 24b). In the composite the background damping with respect to the as-prepared material increases slightly with temperature, moreover a small reduction of the modulus (frequency) occurs with reference to the values experienced in as-fabricated condition. The first peak with a relaxation strength similar to that observed on the composite and significantly higher than that of the monolithic matrix alloy is observed. On the basis of the IF spectra presented above, the major peak can be attributed to hydrogen. The activation energy obtained from the peak shift is slightly lower than that reported for hydrogen diffusion in the β phase of vanadium free titanium [44], but is slightly higher than those obtained by NMR in titanium with vanadium additions [45]. A value comparable with our results, within the uncertainty of measurements, was obtained by IF measurements on the Ti6Al4V alloy [46]. The prominent peak can be therefore attributed to a Snoek-type relaxation caused by interstitials hydrogen atoms in the β phase.


a) b)

Fig. 24. Internal friction and normalized resonance frequency (arrows indicate the peaks) of the monolithic alloy and the composite in as-prepared condition (a) and after ageing at 900 K in

vacuum (b) [43].

A significant difference of relaxation strength is yet observed between the monolithic alloy and the composite. The overall hydrogen content, measured was respectively: 116 ppm in the Ti6Al4V alloy and 187 ppm in the composite. These values are within the low concentration solubility limit where a linear dependence of the (200) β phase XRD line breadth in Ti6Al4V on total hydrogen content was reported [47], and a proportionality between the relaxation strength and the amount of interstitial hydrogen responsible for Snoek relaxation , in the β phase is observed [48]. Therefore an apparent inconsistency for the differences in the relaxation strength of the alloy and the composite (see Fig.24), could be explained by assuming different redistribution of hydrogen in different phases or segregation at dislocations. In this regard it must be considered that dislocation density evaluated by XRD analysis (Fig.4) is one order of magnitude higher in the alloy with respect to the composite. The relaxation strength reduction after ageing, which is of the same order (50%) in the alloy and in the composite for similar ageing treatments, is anyway consistent with a similar mechanism of hydrogen occupancy reduction of the interstitial sites in the β phase, responsible for the Snoek-type relaxation . The major peak is much broader than the corresponding Debye peak and the broadening effect is larger in the composite. This is likely connected to micro-strain and lattice distortion. The average activation energy for the peak observed around 250 K in the matrix alloy corresponds well within the measurement uncertainty, to that reported for hydrogen diffusion in the α phase [42]. A peak with similar activation energy was observed in Ti6Al4V [46] and ascribed to hydrogen in α phase. The mechanism proposed is the same for the Debye peaks present at high temperature, i.e. the stress induces re-orientation of i-s pairs in the hcp α phase [41]. Tests at constant temperature. Isothermal tests have been carried out on composite and monolithic alloy at different temperatures from 300 K to 873 K. Typical results are shown in Fig.25a displaying the evolution of Q-1 and E of composite during experiments at 673 K and 873 K for 8.64 x 104 s (24 hours). Q-1 and E exhibit opposite trends for increasing treatment time: damping decreases whereas dynamic modulus increases. Such variations become larger as temperature increases. Analogous trends were observed for the monolithic alloy with changes a little larger with respect those of composite. For example, Fig.25b compares the modulus trends of composite and Ti6Al4V alloy in tests at 873 K.


It is believed that such behaviour can be due to dislocation pinning by impurities, especially carbon, whose migration velocity increases with temperature. According to the Granato-Lücke dislocation string model [49], the changes of dynamic modulus and Q-1 can be described in terms of dislocation density ρ and mean distance between pinning points l:

(9)

(10)

where G and G0 are the values of shear modulus in the material with and without dislocations, β a constant and ω = 2πf. Although Eq.(9) refers to the shear modulus, the same effect also occurs for the Young’s modulus, since longitudinal deformation may be analyzed into pure shear plus hydrostatic deformation.

a) b)

Fig. 25. Q-1 and E vs. time in isothermal tests for 8.64 x 104 s (24 hours) at 873 K (a) and comparison between the dynamic modulus evolution at 873 K of composite and monolithic alloy.

The corresponding dislocation densities are indicated (b).

Conclusions

Exhaustive microstructural and mechanical investigations have been carried out on the Ti6Al4V-SiCf composite produced by HIP at C.S.M. laboratories. From the results of micro-chemical analysis, the following issues can be emphasized:

1- At increasing temperatures, the lattice expansion of composite is lower than that of monolithic alloy because fibres act as a constraint on the surrounding matrix. Different elastic strains have been measured on the examined {hkl} planes.

2- The fibre-matrix interface is substantially stable also after long-term heat treatments (1,000 hours at 873 K) because a thin TiC layer, which forms all around the carbon coating during the fabrication process, hinders further carbon diffusion towards the matrix and retards interface degradation.

Therefore, the composite can operate at the maximum temperature (873 K) foreseen for its aeronautical applications without remarkable modifications of its microstructure, in particular of the fibre-matrix interface. This is confirmed by mechanical tests (FIMEC indentation, tensile and hardness) which evidence that mechanical properties remain stable also after the most severe long-term heat treatments considered here.

2l

G

G

o

⋅⋅−=∆

ρβ

ωρ ⋅⋅∝− 41lQ


The anelastic behaviour of the composite has been investigated by IF and dynamic modulus measurements from 80 to 1173 K. The IF spectrum of the composite exhibited:

1- a relaxation peak at high temperature (∼880 K) with H = 186 kJ mol-1 and τ0 = 2.3 x 10-15 s due to the re-orientation of C-Al and C-V pairs in the α phase of Ti6Al4V matrix around the fibres. This peak is not present in the monolithic alloy.

2- two peaks at low temperature. The first one appears at approximately 120 K (f = 1 kHz) with an activation energy of 20.3 kJ mol-1 and has been attributed to a Snoek-type relaxation caused by interstitials hydrogen atoms in the bcc β phase. The second very broad peak localized at 270 K was ascribed to hydrogen diffusion in the α phase because its activation energy (48.2 kJ mol-1) is the same.

In isothermal tests the values of Q-1 and E of composite change with an asymptotic trend. These variations depend on diffusion phenomena of interstitial atoms to dislocations that reduces the mean distance between pinning points.

Acknowledgments

S. Kaciulis and A. Mezzi are grateful for the financial support provided by the ELETTRA synchrotron for the project n. 20095130. R. Montanari thanks Mr. Benedetto Iacovone and Mr. Piero Plini for their assistance in FIMEC indentation tests.

References

[1] C. Arvieu, J.P. Manaud and J.M. Quenisset: J. Alloys Comp. Vol. 368 (2004), p. 116.

[2] Y.C. Fu, N.L. Shi, D.Z. Zhang and R. Yang: Mater. Sci. Eng. A Vol. 426 (2006), p. 278.

[3] G. Das: Metall. Trans. A Vol.21A (1990), p. 1571.

[4] P. Martineau, R. Pailler, M. Lahaye and R. Naslain: J. Mater. Sci. Vol. 19 (1984), p. 2749.

[5] T.W. Clyne, in: “3.7.12. Metal Matrix Composites: Matrices and Processing , Encyclopaedia of Materials: Science and Technology”, §3.7“Composites: MMC, CMC, PMC”, ed. A Mortensen, Elsevier, (2001), p. 12.

[6] M.A. Mittnick, in: Metal and Ceramic Matrix Composites 605, Proc. TMS annual meeting in Anaheim, CA, U.S.A., (1990) p. 413.

[7] S.V. Nair, J.K. Tien and R.C. Bates: Inter. Metals Rev. Vol. 30 (1985), p. 275.

[8] P.R. Smith and F.H. Froes: J. Met., Vol. 36 (1984), p. 19.

[9] P.D.Nicolau, S.L.Semiatin et al.: Scri. Met. et Mat, Vol. 32 (1995), p. 57.

[10] H.D. Hanes, D.A. Seifert and C.R. Watts, in: Hot Isostatic Pressing, Battelle Press, Columbus, OH, (1979), p. 16.

[11] S. Gunther, B. Kaulich, L. Gregoratti and M. Kiskinova: Prog. Surf. Sci. Vol. 70 (2002), p. 187.

[12] S. Malinov, W. Sha and Z. Guo, SRS Annual Reports (2000- 2001).

[13] S. Malinov, W. Sha , Z. Guo, C.C. Tang and A.E. Long: Mater.Charact. Vol. 48 (2002), p. 279.

[14] H. Choo, P. Rangaswamy, M.A.M.Bourke and J.M.Larsen.: Mater. Sci. Eng. A Vol.325 (2002), p. 236.

[15] R. Pederson, O. Babushkin, F. Skystedt and R. Warren: Mater. Sci. Tech. Vol. 19 (2003), p. 1533.


[16] E. Fromm and G. Horz: Inter. Metals Rev. Vol. 25 (1980), p. 269.

[17] R. Montanari , G. Costanza , M.E. Tata and C. Testani: Mater. Charact. Vol. 59 (2008), p. 334

[18] J.F. Silvan, J.C. Bihr and Y. Lepetitcorps: Composites Vol. 27A (1996), p. 691.

[19] Y.C. Fu, N.L. Shi, D.Z. Zhang and R. Yang: Mater. Sci. Eng. A Vol.426 (2006), p. 278.

[20] S. Kaciulis: “Spettroscopia elettronica delle superfici”, Tecniche sperimentali per la

caratterizzazione dei materiali, ed. R.Montanari, Associazione Italiana Metallurgia (2005), p.103.

[21] J.F. Moulder, W.F. Stickl, P.E.Sobol and K.D. Bomben, in: Handbook of X-ray Photoelectron

Spectroscopy, Eden Prairie, Physical Electronics Inc.,(1995) U.S.A.

[22] K.D. Childs, B.A. Carlson , L.A. La Vanire, J.F. Moulder, D.F. Paul, W.F. Stickle, D.G. Watson, in: Handbook of Auger Electron Spectroscopy, Eden Prairie, Physical Electronics Inc.(1995)U.S.A.

[23] R. Donnini, S. Kaciulis, A. Mezzi, R. Montanari and C. Testani: Surf. Interface Anal. Vol. 40 (2008), p. 277.

[24] S. Kaciulis, A. Mezzi, R. Donnini, P. Deodati, R. Montanari, N. Ucciardello, M. Amati, M. . Kazemian-Abyaneh and C. Testani, Surf. Interface Anal., Vol. 42 (2010) p. 707.

[25] M. Kiskinova, K. Marsi, E. Di Fabrizio, M. Gentili: Surf. Rev. Lett. Vol. 6(2) (1999), p. 265.

[26] M. Marsi, L. Casalis, L. Gregoratti, S. Gunther, A. Kolmakov, J. Kovac, D. Lonza, M. Kiskinova: J. Electron. Spectrosc. Vol. 84 (1-3) (1997), p. 73.

[27] L.H. VanVlack, in: Elements of Materials Science and Engineering, Publ. Addison-Wesley (1985).

[28] H.C. Swart, A.J. Jonker, C.H. Claassens, R. Chen, L.A. Venter, P. Ramoshebe, E. Wurth , J.J. Terblans and W.D. Roos: Appl. Surf. Sci. Vol.205 (2003), p. 231.

[29] P. Gondi, R.Montanari and A.Sili: J. of Nucl. Mater. Vol. 212-215 (1994), p. 1688.

[30] P. Gondi, A. Donato, R. Montanari and A. Sili: J. of Nucl. Mater. Vol. 233-237 (1996), p.1557

[31] A. Donato, P. Gondi, R. Montanari, F. Moreschi, A. Sili and S. Storai: J. Nucl. Mater. Vol. 258-263 (1998), p. 446.

[32] B. Riccardi and R. Montanari: Mater. Sci. Eng. A 381, Vol.1-2 (2004), p. 281.

[33] B.W. Christ, in “Effect of Specimen Preparation, Setup and Test Procedures on Test Results”, Metals Handbook, 9th ed., Vol. 8 (1985) p. 32.

[34] S. Amadori, E.G. Campari, A.L. Fiorini, R. Montanari, L. Pasquini, L. Savini and E. Bonetti: Mater. Sci. Eng. A Vol. 442 (2006), p. 543.

[35] R. Schaller: Mater. Sci. Eng. A Vol. 442 (2006), p. 423.

[36] L.H. He and C.W. Lim: Compos. Sci. Technol. Vol. 61 (2001), p. 579.

[37] R. Raj and M. Ashby: Metall. Trans. Vol. 2 (1971), p. 1113.

[38] M.H. Hou and R.F. Davies: J. Mater. Sci. Vol. 14-10 (1979), p. 2411.

[39] M.H. Hou, R.F. Davies and D.E. Newbury: J. Mater. Sci. Vol. 15-8 (1980), p. 2073.

[40] M. Naka, J.C. Feng and J.C. Schuster: Metall. Mater. Trans. Vol. 28 A (1997), p. 1385.

[41] D. Gupta and S. Weining: Acta Metall. Vol. 10 (1962), p. 292.

[42] F. Povolo and E.A. Bisogni: Acta Metall. Vol. 14 (1966), p. 711.

[43] S. Amadori, E. Bonetti, P. Deodati, R. Donnini, R. Montanari, L. Pasquini and C. Testani:


Mater. Sci. Eng. A Vol. 521-522 (2009), p. 340.

[44] R.J. Wasilewsky and G.L. Kehl: Metallurgia Vol. 50 (1954), p. 225.

[45] C. Korn and D. Teitel: Phys. Status Solidi A Vol. 44 (1977), p. 755.

[46] Du Jia Ju : J. Phys., C5 Vol.42 (1981), p.775.

[47] X.S.Guan, H. Numakura, M. Koiwa, K. Hasegawa and C. Ouchi: Mater. Sci. Eng. A Vol. 272 (1999), p. 230.

[48] G.F. Pittinato and W.D. Hanna: Met. Trans. Vol. 3 (1972), p. 2905.

[49] A.S. Nowick and B.S. Berry, in: Anelastic Relaxation in Crystalline Solids (Academic, New York, 1972), p.356.


Simulation of the Mechanical Behaviour of Metal Matrix Composites

Siegfried Schmaudera, Ulrich Weberb, Andreas Reuschelc and Markus Willertd

Institute for Materials Testing, Materials Science and Strength of Materials (IMWF), University of Stuttgart, Pfaffenwaldring 32, 70569 Stuttgart, Germany

[email protected], [email protected],

[email protected], [email protected]

Keywords: Metal/Matrix-Composite, Self Consistent Unit Cell Model, Mechanical Behaviour, Homogenization

Abstract. A model based on the geometry of the phases is introduced in order to investigate the

mechanical properties of interpenetrating microstructures. In order to characterize the elastic and

elastic-plastic properties of the composite a self consistent unit cell model is applied on a wide

range of volume fractions for an Al/TiO2 composite. Besides the volume fraction a microstructural

based parameter is used, the matricity, to describe the mutual circumvention of both phases.

Computations are carried out for different temperatures and void volume fractions. In addition a

conservative fracture criterion based on critical normal stresses is applied to derive realistic stress

strain curves.

Introduction

The prediction of the overall elastic-plastic behaviour of metal matrix composites with

interpenetrating phases is a research topic since long time. Different numerical approaches exist to

achieve this goal. The method which costs the most effort is using real microstructures which have

to be meshed and calculated in 3D. This approach requires high computation times. Therefore, there

had always been intentions to reduce the computational effort especially for parametric studies. A

advanced method for time efficient simulations of the mechanical behaviour of MMC-materials is

based on unit cell models. They have been developed by simplifying the microstructures to

spherical inclusions embedded in a matrix. In the present work a parametric study is carried out to

calculate the material response of a wide range of composite materials with interpenetrating

microstructures. The aim is to identify the most favourable material properties which could be

obtained in TiO2-AlSi9Cu3 system automotive applications. The composite manufacturing route

selected is the infiltration of a ceramic preform by squeeze casting. The low wettability of the

ceramic material was met by the application of pressure during the infiltration process. Therefore,

preheating of the melt as well as of the ceramic is of influence on the infiltration and the

solidification of the metal. Furthermore the infiltration path of the melt has to be taken into account

to reduce the porosity in the MMC-material. Manufacturing of the calculated composites is not part

of the present pater.

Nomenclature

matricity a microstructure dependent parameter which describes the

matrix character of a phase

skeleton line a line that will stay if from a binary image of a microstructure

step by step the outer pixels are removed

representative volume element a cut out of a real microstructure which possesses the same

overall distribution of the inclusion as the bulk material


unit cell a simplified simulation cell representing the material behaviour

of the bulk material

Unit Cell Models

The embedding of the inclusions in the matrix is described by self consistent unit cells [1, 2]. Even

though these models are based on spherical inclusions in a matrix, they have been proven to predict

the properties of statistically distributed inclusions. These self consistent unit cell models can be

extended to take into account the geometrical mutual circumvention of both phases which will

occur to different degrees, for different volume fractions of ceramic and materials (see next

section).

The reduction of computing time for the prediction of stress – strain curves is a reason why unit cell

models are often used. In these models a representative volume element (RVE) is transferred into a

geometry as simple as possible which usually is a sphere of the inclusion material, surrounded by a

cylinder of the matrix material. From such unit cells the behaviour of the composite can be studied.

By using special rotational symmetric boundary conditions and appropriate elements it is possible

to mesh these unit cell models with a 2D FE-mesh. Further reductions of computing time can be

achieved by using of geometric symmetries. Self consistent unit cells (Fig. 1) extend this approach

by enclosing the inclusion and the matrix, which form the inner cell, with an outer surrounding

homogenized material. The mechanical properties of the embedding composite are computed

through an iterative approach to the behaviour of the inner cell. Therefore, the elastic-plastic

behaviour of the homogenised material can be calculated. Unmodified unit cell models can be

applied for all sizes and shapes of reinforcement phases above a size of two microns. Below this

size the “Mechanism-based Strain Gradient (MSG) Plasticity Theory” [3, 4] has to be considered to

take the geometric necessary dislocations into account. Recent work implemented this approach in a

unit cell model to describe the properties of a dual phase steel which contains martensitic particles

in a ferritic matrix [5].

Fig.1. Schematic model of a self consistent embedding cell [6].

Force- or displacement Controlled loading

Embedding composite

Matrix

Inclusion or fiber

inner cell

Planes of symmetry


Matricity Model

The matricity model is a combination of two self consistent unit cells, which consider the enclosure

of the phases of a two-phase structure [6 - 8]. Thereby, in every cell, depending on the volume

fraction, every phase is once considered as the matrix and as inclusion. Thus, there are two stress-

strain characteristics for the composite material. With the aid of a weighting function, which

depends on volume fraction and the topology of the micro structure (matricity), the stress-strain

curve is calculated. For the derivation of the microstructure parameter, matricity, which was

introduced in [9], the length of the skeleton lines (Fig. 2) of the two phases has to be compared

(Eq.1). The matricities of both phases are complementary to each other and sum up to one. If a

phase has a matricity of zero, it is in a globular shape and is totally enclosed by the other phase

(matricity one).

βα

αα

SS

SM

+= (1)

( )( ) ( )111

11

31

32

32

+++−

+−=

ββ

β

βfWfW

fWM (2)

Fig.2. Two-phase microstructure with skeleton lines for each phase [6].

Fig. 3 shows the two self consistent unit cells. The volume fractions of both phases α and β are the

same, but the arrangement of the phases α and β is interchanged, so that every phase once appears

as “inclusion” and once as “matrix”. Since the volume fractions fα and fβ of the phases are the same

in both unit cells, the matricities Mα and Mβ can be represented in terms of W1 and W2 (Eq. 2).

In Fig. 4 an iterative approach to the composite behaviour of an interpenetrating microstructure with

a phase composition of 50 vol.% metal and accordingly ceramic can be seen. The diagram

intentionally shows results for strain levels which the composite is not likely to reach because of

failure initiation. However, by displaying results for larger strain levels, the changes appearing in

higher iteration steps can be seen. The default for the iteration step of a volume fraction of 50%

ceramic was close to the stress-strain behaviour of the metal phase. Therefore, the stress-strain

curve after the first iteration shows a stiffer material behaviour than before. The slope of the elastic

part of the curve is after one iteration similar to the one after the last iteration. In this case the

plastic part of the phase composition can be sufficiently exact approximated after about seven

iteration steps.

Sα

Sβ


Fig.3. Schematic of the two embedding cells fort the iterative calculation of the homogenized

material properties with the matricity model [6].

Fig. 4. Iterative computation of the stress strain curve (fTiO2=50 vol.%).

In Fig. 4 the stress stain curves for the 7th

up to the 15th

iteration are marked with “higher

iterations”. At these iterations the curves only differ in stress- and strain-levels that are no more

relevant for the composites on hand. Hence, the underlying convergence criterion, which finishes

the calculation if successive iteration curves differ by less than 3%, can be used without any loss in

quality.

The model was applied to calculate the mechanical properties of different interpenetrating

microstructures. In [10 - 12] the model was applied to W/Cu and Ag/Fe composites, where the

comparison with experimental results showed good agreement. Furthermore the model was used

successfully to calculate the material response for composite materials with randomly distributed

particle reinforcements [13].

TiO2 TiO2


Elastic Modulus and Thermal Expansion

With the aid of an analytical approach for linear material characteristics, boundaries for the

composite modulus can be defined. The calculated elastic modulus must be within the area between

the boundaries. Otherwise the chosen preconditions for the simulation are not appropriate. The

analytical calculation of an upper and lower value boundary for the elastic modulus is derived by

using rules of mixture. As shown in Fig. 5, the upper boundary is obtained by the rule of mixture

for a model of both materials, the lower boundary is obtained from a serial arrangement of the

materials. These two assemblies represent the extreme combinations of two material phases in a

composite C. Hence, it is possible to determine the upper and lower boundary for the elastic

modulus through their extreme arrangement relative to the direction of loading. The distance

between the two limiting curves gets larger, the more the quotient EA/EB differs from 1.

Fig. 5. Young´s modulus of interpenetrating microstructures versus the ceramic content.

When using the simple forms of the mixture rules it is obvious that they only allow a very rough

estimation of the composite behaviour since in the present case of an interpenetrating

microstructure a rather inhomogeneous arrangement of the phases prevails. From the shape of the

Young´s modulus-curve of the composite the area by area classification of the composite material

can also be derived. Up to a ceramic volume fraction of 30% the material behaviour is mainly based

on the logic of the series connection of the phases in the stress-strain behaviour. As soon as the rise

of ceramic material in the composite has limited the plastic deformability of the metallic parts, the

increase of the Young´s-moduli is orientated to the conditions of a parallel assembly of the phases.

For more than 40% of ceramic the Young´s modulus increases stronger with increasing ceramic

phase fraction as compared to a ceramic volume fraction of less than 40%. The simulations show

that the change of the Young´s modulus is by far not directly proportional to the volume fraction of

the ceramic phase. For this reason composite models based on the rules of mixture can only be used

as a rough approximation, while self-consistent unit cell models are capable to represent the actual

material behaviour of the composites with interpenetrating microstructures.


Fig. 6. Young´s modulus of Al/ceramic interpenetrating microstructures at room temperature and

200°C versus the ceramic volume fraction.

Fig. 6 shows the dependence of the Young´s modulus and the volume fraction of ceramic at a

temperature of 200°C. To compare the dimensions, the dependence of the Young´s modulus at

room temperature is also shown. This comparison demonstrates that the development of material

properties, depending on the composition of the material, follows the same tendency even at

different temperatures.

Fig. 7. Coefficient of thermal expansion drawn against the ceramic content for two temperatures

(room temperature and 200°C).

As can be seen from Fig. 7, there are differences in the coefficient of thermal expansion (CTE)

depending on the temperature. From above simulations, however, it is known that the thermal

expansion coefficients only slightly changes between 400°C and room temperature. This constancy

also exists in the simulation, which only considers cooling form 400°C to 200 °C.

at room temperature

at 200°C

CTE

ceramic content in vol. %

ceramic content in vol. %


Stress-Strain Curves

In Fig. 8 simulation results for different ceramic volume fractions are shown. For each simulation

the convergence criterion was met and the iteration process stopped. Thereby the phase

compositions in the simulations have been changed in 10% steps to allow a continuous evaluation

of the change in mechanical behaviour depending on the phase composition. The shown stress-

strain curves are simulation results in which no fracture criterion has been used.

Fig. 8. Stress strain curves for different phase compositions for the Al/ceramic composite.

The material behaviour in this combination group can be divided into three areas. Microstructures

with a ceramic phase fraction up to 30 vol.% show distinctive plastic behaviour. In this combination

area the properties of the metallic phase are dominant.

In the second area the change in deformation behaviour from metallic dominated to ceramic

dominated takes place. In this area, microstructures with about 40% ceramic can be assigned. From

50% ceramic phase on the plastic part in the stress-strain behaviour gets noticeably smaller and the

deformation behaviour of the composite approaches the one of ceramics. In principle according to

Fig. 8 there will be no proportional transition of material behaviour from the pure metal phase to the

pure ceramic phase with increasing ceramic volume fraction.

Conservative Fracture Criteria

Fig. 9 shows the application of the fracture criterion. For this conservative criterion the maximal

shear stress criterion is applied to the metal phase of the model and the maximum stress theory is

applied to the ceramic phase of the model. Due to the incremental analysis of the stresses and

strains within each model phase it is possible to determine which criterion is reached first. If one of

the criterions is met the composite is considered at its failure strength.

The straight lines in Fig. 9 connect the failure criterion of the Aluminium and the Failure criterion

of the ceramic for each phase composition. From the slope of the lines the external stress and strain

distribution on each phase of the microstructure can be derived. At low ceramic volume fractions

the strains within the metallic phase will be much higher than in the ceramic phase at similar stress

levels. This can be explained by the inclusion of the ceramic within a metal matrix: In this

TiO2


constellation it is possible for the metallic phase to absorb the appearing load through plastic

deformation, whereas the brittle ceramic parts of the composite interfere in this yielding process at

first only slightly and with increasing volume fraction more and more. Up to compositions of 30%

the material reacts to the applied loads in this way. A ceramic volume fraction of over 30% leads to

the formation of a ceramic network within the composite material. The deformation of the metallic

phase is therefore restrained and the same strain level in both phases is assumed.

Fig. 9. Realisation of the fracture criterion.

These principles can be applied for the present composite materials in the way that at high volume

fractions of ceramic the strain in the metallic parts of the microstructure is restrained. In the extreme

case of a metallic inclusion in a ceramic surrounding the metal is separated from the applied loads

since only strains according to the ceramic surrounding are conveyed to the metal, which lead to

small stresses in the metal. The influence of the ceramic phase on the stress-strain ratio in the

metallic phase is very noticeable at more than 50 vol.% ceramic. The change between the model

representation of the series connection to the parallel connection takes place at a ceramic volume

fraction of 30% to 50%.

After the evaluation of the fracture criterion for each micro structural composition the stress-strain-

ratio is shown in Fig. 10. For its determination the intersection points of the stress-strain curves

with the lines, that represent the stress distribution between the two phases and were shown in

Fig.9, must be defined. The intersection points also stand for the failure of the phase compositions.

It becomes obvious that interpenetrated microstructures with a ceramic volume fraction of up to

30% can take large plastic deformations, which are much higher than expected for a composite with

a TiO2 structure. The striking influence of the ceramic part to the behaviour of the whole

microstructure with a ceramic volume fraction of more than 50% is a result of the restrained plastic

deformation in the metal phase. Because of the pure elastic characteristic of the ceramic the

restrained deformation of the aluminium through the ceramic phase is approved.

TiO2


Fig. 10. Stress-strain curves after the application of the presented fracture criterion.

Fig. 11. Stress-strain curves after the application of the presented fracture criterion for an

Al/ceramic interpenetration microstructure at 200°C.

After the application of the fracture criterion, which applies the maximal shear stress criterion to the

metal phase and the maximum stress theory to the ceramic phase, stress-strain curves for 200°C

shown in Fig. 11. In comparison with the stress-strain curves at room temperature in Fig. 11 it can

be seen, that the overall strain is higher for the temperature of 200 °C.

TiO2

TiO2


This change in the behaviour of the composite is together with the lower fracture stresses a result of

the decreasing strength of the metal at elevated temperatures. The material behaviour of the ceramic

phase is not changed by the higher temperature which was applied in this study.

Fig. 12. Stress-strain curves after the application of the presented fracture criterion for porous

Al/ceramic interpenetration microstructures at room temperature.

The stress-strain curves in Fig. 12 show that the influence of porosities only applies to metal

dominated microstructures. From the figure it can be seen that the development of the stress-strain

behaviour according to the applied loads and depending on the composition of the microstructure in

principle stays the same as without pores. Porosities show an influence on the macroscopic material

behaviour for microstructures with less than 50% of ceramic phase. In the material system at hand,

the interpenetrating microstructures react very sensitive to changes of the volume fractions of the

phases. Microstructures with a ceramic volume fraction up to 30% show metal dominated

behaviour, which are displaced at lower stress levels in comparison to the calculations of the

aluminium/ceramic interpenetrating microstructure with ceramic volume fractions between 50%

and 70%.

The increased plasticity of the microstructure induced by pores may not distract from the

problematic impact of the porosities. Because of their properties porosities will not take part in the

transmission of loads. Thus, they reduce the load carrying area of the Material. Hereby the stress

levels are increased regarding to the macroscopic loads. Furthermore, porosities induce strong notch

effects which additionally increase the stress levels.

The effects of virtual lowering the elastic modulus of the ceramic to 60% of its value are shown in

Fig. 13. It displays the stress-strain curves for the phase compositions from 30% to 60% ceramic for

both the base value for the Young´s modulus for the ceramic (250 GPa) and the reduced Young´s

modulus for the ceramic (150 GPa). Despite the more elastic composition of the ceramic phase the

plastic part of the stress-strain curves has been reduced compared to the original calculation with

250 GPa. For all phase compositions the material behaviour is almost purely elastic. This is the

result of a reduced stress loading of the metal phase during deformation. The fracture criterion in

the ceramic phase is reached as the strain level increases and due to the restricted deformation of the

metal phase the load is not distributed to the metal but rather has to be carried by the ceramic.

TiO2


Fig. 13. Stress-strain curves of interpenetrating microstructures with variation of young´s modulus

of the ceramic phase.

Conclusion

The described self-consistent approach shows a good possibility to predict a wide range of

composite properties including the Young´s modulus, plastic behaviour and the coefficient of

thermal expansion. Moreover this model applies to a wide range of temperatures and different phase

distributions, which is achieved by introducing the matricity microstructure parameter, and volume

fractions. By applying the conservative fracture criteria, which represents the maximal shear stress

criterion the metal phase and the maximum stress theory phase, is a way to approximate the

material failure.

References

[1] Z. Li, S. Schmauder, A. Wanner and M. Dong: Scripta Metall. Mater. Vol.33 (1995) p. 1289-

1294,.

[2] S. Hönle and S. Schmauder: Comp. Mater. Sci Vol.13 (1998) p. 56-60

[3] H. Gao, Y. Huang, W.D. Nix and J.W. Hutchinson: J. . Mech. Phys. Solids Vol.41 (1999)

p.1239-1263

[4] Y. Huang, H. Gao, W.D. Nix, J.W. Hutchinson: J. Mech. Phys. Solids Vol.48 (2000) p.99-128

[5] I. Sahni: Micromechanical Simulation of the Effect of Particle Size and Volume Fraction in

dual Phase Steels: A study by the theory of Mechanism-based Strain Gradient Plasticity,

Student Work, IMWF, 2009

[6] P. Leßle, M. Dong and S. Schmauder: Comp. Mater. Sci. Vol. 15 (1999) p 455-465

[7] P. Leßle, M. Dong, E. Soppa and S. Schmauder: Scripta Mater. Vol. 38 (1998) p.1327-1332

[8] M. Dong, P. Leßle, U. Weber and S. Schmauder: Mater. Sci. Forum Vol. 308-311 (1999)

p.1000-1005

[9] M.H. Poech and D. Ruhr: Prakt. Met. Sonderband Vol.24 (1993) p. 385-391

TiO2 TiO2

TiO2 TiO2


[10] S. Schmauder, U. Weber, I. Hofinger and A. Neubrand: Tech. Mechanik Vol.19 (1999) p.313-

320

[11] S. Schmauder, U. Weber, in: Modelling the Deformation Behaviour of W/Cu Composites by a

Self-Consistent Matricity Model, ECM'99, Progress in Experimental and Computational

Mechanics in Engineering and Material Behaviour, edited by D. Zhu, M. Kikuchi, Y. Shen,

M. Geni, Northwestern Polytechnical University Press, Xi'an, China, pp. 54-60, 1999.

[12] P. Leßle, M. Dong, E. Soppa, S. Schmauder, in: Simulation of Interpenetrating

Microstructures by Self Consistent Matricity Models, Materials Mechanics, Fracture

Mechanics, Micro Mechanics, An Anniversary Volume in Honour of B. Michels 50th

Birthday, edited by T.Winkler, A.Schubert, Fraunhofer IZM Berlin, Chemnitzer

Werkstoffmechanik GmbH, Chemnitz, pp. 456-461. 1999.

[13] U. Weber: Modellierung von Verformung und Schädigung in Werkstoffgefügen mit

unterschiedlich großen Teilchen und unter Wasserstoffeinfluss, Dissertation, Uni Stuttgart,

2006


DRY SLIDING BEHAVIOUR OF PEO (PLASMA ELECTROLYTIC OXIDATION) TREATED AA 2618/20% Al2O3P COMPOSITE

Lorella Ceschinia, Carla Martinib, Giuliano Sambogna and Fabrizio Tarterini

Department of Metals Science, Electrochemistry and Chemical Techniques (SMETEC) Università di Bologna – V.le Risorgimento 4, Bologna, 40136 Italia

[email protected], [email protected]

Keywords: Metal matrix composites (MMC), AA2618/20%Al2O3p, Plasma Electrolytic Oxidation (PEO), tribology, wear, sliding

Abstract. The present study focuses on the influence of the PEO (Plasma Electrolytic Oxidation)

treatment on the tribological behaviour of the AA2618/20 % vol. Al2O3p composite, dry sliding

against induction hardened UNI C55 steel. Particle-reinforced Al based composites offer a higher

wear resistance by comparison with the corresponding unreinforced alloys, however, the presence

of critical loads and/or velocities which lead to transition towards severe wear regime, was often

observed. In such conditions, the composite can show higher wear rates than those of unreinforced

alloys. For this reason, surface modifications, such as PEO, might contribute to improve wear

resistance. In this paper, topography, microstructure, phase constitution and surface hardness of the

PEO-treated composite were investigated and its tribological behaviour was studied by dry sliding

tests using a block-on-ring tribometer. The results were compared with those from the uncoated

composite, demonstrating a very positive effect of the PEO treatment, which moved transitions

from mild to severe wear towards more severe test conditions, in terms of both load and velocity.

Introduction

Aluminium alloy matrix composites reinforced with ceramic particles are characterised by a

particularly good combination of specific strength and stiffness, thermal stability and wear

resistance. This means that they can be applied to a wider range of fields than unreinforced alloys,

with the cost of such application being reasonable [1-2]. For these reasons, they have been

considered as potential candidates for various applications, mainly in the transport sector, such as

pistons, cylinder liners and brake discs. With reference to such components, the study of

tribological behaviour is of particular interest and, in fact, several studies have been carried out on

their wear resistance in conditions of both sliding and abrasive wear [1-5]. It has become clear how

Al based composites reinforced with ceramic particles are characterised by a substantial

improvement in wear resistance compared to unreinforced alloys. This is due to their capacity to

support applied loads, which limits damage to the matrix. However, the presence of critical loads

and/or velocity values have also been highlighted. In correspondence with these critical values,

transitions towards a severe wear regime, as a consequence of fragmentation and avulsion of the

ceramic reinforcement, which lead to the formation of a particularly abrasive “third body” [6], were

observed. In such conditions, the composite can show higher wear rates than those of unreinforced

alloys.

Recently, various studies have been carried out on plasma electrolytic oxidation treatment (PEO)

applied to aluminium [7-13], titanium [14-16] and magnesium alloys [17, 18]. PEO, also known as

Micro-Arc Oxidation (MAO), is a conversion treatment based on anodic oxidation, which is carried

out at a low temperature (T <60 °C) in an electrolyte consisting of a dilute alkaline aqueous solution

[6]. This process is different from conventional anodic oxidation as it involves the modification of

the growing oxide film, by applying a low-frequency alternating electric field (approx. 50 Hz) [7,

8]. In these conditions, a succession of arc micro-discharges is created on the treated component


(duration 0.25÷3.5 ms; involved volume<0.03 mm3 according to estimations in [19] in the case of

Al): the temperatures reached locally give place to the densification of the layer that is being formed

and the rapid cooling that follows modifies the layer produced, which will be made up of both

amorphous and nano-crystalline phases. The structure of the conversion layers, produced by PEO,

consists of a succession of layers [10], which have different properties: (i) external or so-called

“technological” layer (equal to 20÷30% of total thickness), characterised by high roughness and

porosity, to such an extent that it can be effective for lubricant retention; (ii) functional layer, which

has the highest levels of hardness, due to its dense and compact structure.

PEO can produce oxide layers on Al alloys with a thickness ranging from 50 to 150 µm [10]. PEO

is also advantageous as it produces hard and relatively uniform conversion layers, even on complex

geometries, which is not the case for other conversion treatments such as hard anodising [20].

Moreover, the costs of PEO treatment are comparable to those of hard anodising, with the

additional advantages of low environmental impact [7].

Various studies show that the application of this type of treatment to aluminium alloys improves

their tribological performance and corrosion resistance [21-23]. It is therefore important to also

evaluate the effects of PEO treatment on aluminium alloy matrix composites reinforced with

ceramic particles, with the aim of widening the field of tribological application of these materials, in

terms of applied loads and sliding velocity. There are currently few studies on the effects of plasma

oxidation treatments on the tribological behaviour of Al-matrix composites [24, 25]. The aim of this

research was, for this reason, to study the effect of PEO treatment on the microstructure and wear

resistance of a composite, based on the aluminium alloy 2618, reinforced with Al2O3 particles, in

sliding conditions against induction hardened UNI C55 steel.

Experimental Methods

Material. The experiment described in this work was carried out on the

AA2618/20%volAl2O3p composite, treated with PEO and compared with untreated material

(hereafter indicated as MMC (Metal Matrix Composite) and MMC+PEO, respectively). The

composite was produced by Duralcan through Compocasting, then was T6 heat treated

(solubilisation at 530°C for 2 h, water quenching and ageing at 195°C for 29 h), to a Brinell

hardness of 135±11.

The PEO treatment was carried out by Keronite Ltd (Cambridge, UK) under the following

conditions:

• Alternating current, frequency = 50 Hz

• Constant current density, equal to 20 A dm-2

• Treatment time: approx. 50 minutes

• Electrolytic solution of Na2SiO4, SiAl2 and other additives, maintained at 45°C.

Microstructural and tribological characterisation. The characterisation of the material involved

metallographic analysis through optical microscopy (OM) and image analysis, scanning electron

microscopy (SEM) with an energy dispersive spectrometer (EDS). The phase constitution of the

coatings was determined by x-ray diffraction (XRD), performing θ-2θ scans from 20° to 90° with a

0.02° step size and a 3 s dwell time. A CuKα radiation source was used, with a 40 kV accelerating

voltage and a 30 mA filament current. XRD traces were collected from coating free surfaces as well

as from the inner layer at about 10 µm of depth from the free surface. Topographic measurements

were carried out on coating free surfaces by optical and stylus profilometry, in order to measure the

surface roughness and characterise surface morphology. The thickness of the PEO treatment on the

composite and the characteristics of the coating/substrate interface were evaluated in both cross

section, on samples mounted in resin (preceded by the deposition of a protective layer of approx. 20


µm of cyanacrylic resin) and polished, and in fracture section (obtained by Charpy impact test at

room temperature). The effect of PEO treatment on the tribological behaviour of the untreated and

PEO treated composite was evaluated by dry sliding tests, using a “slider-on-cylinder” test [26],

which corresponds to the block-on-ring contact geometry and allows continuous acquisition, as a

function of the sliding distance, of the friction force, by means of a bending load cell, and system

wear (fixed slider + rotating cylinder), by linear vertical displacement transducer. At the end of the

test, the wear scar depth was measured by stylus profilometer (curvature radius: 5 µm). The tests

were carried out in laboratory atmosphere (18÷24 °C, relative humidity 40÷60%), with applied

loads ranging between 10 and 50 N, sliding velocity of 0.6 and 1.8 m/s and for a sliding distance of

10 km. The stationary sliders (5x5x70 mm) were produced with untreated MMC and were also

submitted to PEO. For the rotating cylinder (diameter: 40 mm) induction hardened UNI C55

steel was used (hardening thickness of approx. 400 µm, hardness 650 HV1, roughness Ra=0.15 µm)

as a countermaterial. Both the cylinder and the sliders were characterised, before the tribological

tests, by stylus profilometer and sclerometric measures (Brinell hardness of the untreated MMC;

Vickers microhardness HV0.1 of the coated system MMC+PEO). For a more complete evaluation of

the mechanical properties of the PEO layer, nano-indentation measures were also performed on the

MMC+PEO, after removing the outermost porous layer (Berkovich indenter, peak load 500 mN

reached in 100 s, acquisition of 20 load-unload curves with indentations 50 µm apart). The study of

wear mechanisms was carried out by morphological and compositional analysis of the wear scars

and debris, which was performed by stereomicroscopy, SEM and EDS microanalysis.

Results and Discussion

Material. Since the layer that forms during plasma electrolytic oxidation (PEO) derives from the

substrate material and, consequently, its microstructure and composition can influence its properties

and morphology, the microstructural characteristics of the MMC were evaluated before the PEO

treatment was carried out. Fig. 1 shows OM (a) and SEM (b) micrographs, characteristic of the

untreated composite, which highlight a fairly even distribution of reinforcement (Fig. 1-a),

with some particle clusters, which is typical of as-cast MMC.

(a) (b)

Fig. 1. MMC AA2618/20 % Al2O3p microstructure: reflected polarised light optical micrograph

(a) and back-scattered electron (SEM-BSE) images that show the precipitates at grain

boundaries (light areas) (b).

100 µm


The reinforcement content was controlled by image analysis, with Image Pro-Plus software;

the volume content of the Al2O3 particles ranged from 19 to 21% and was therefore in

agreement with data from the producer (20% in vol.). Adopting the main axis of the

equivalent ellipse associated with the particles as a characteristic dimension of the

reinforcement, the following results were obtained: the size of the greater amount of particles

ranged from 10 to 20 µm, even though particles of up to 30-40 µm in size were observed. The

back-scattered electron image in Fig. 1-b shows the presence of prevalently Al-Cu-Mg based

precipitates and some Fe based precipitates, which are mostly found at the grain boundaries of

the matrix (with an average grain size of approx. 115 µm) and at the particle/matrix interface.

XRD analysis showed that PEO treatment on MMC produced a layer mainly consisting of α-Al2O3

and γ-Al2O3, as well as amorphous phases, which is typical in PEO treatments [12, 13], and also of

phases due to the interaction with the electrolyte. Underneath the outermost layer, these interaction

phases disappear and the coating mainly consists of α-Al2O3 and γ-Al2O3 (diffraction lines from the

underlying substrate were also observed due to the relatively low thickness of the internal layer), as

shown in Fig. 2.

Fig. 2. XRD pattern measured on the inner layer of the PEO coating on the treated MMC, after

removing by mechanical polishing the outermost layer (about 10 µm).

Position [°2Theta]

20 30 40 50 60 70 80 90

Counts

0

500

1000

1500

CMMKer-10mic; Measured: 28/11/2007 08:53:17

Al

α-Al2O3

γ-Al2O3


A representative SEM image of the PEO-treated free surface is shown in Fig. 3-a, which displays

the typical morphology of craters and bubbles, that are generated by the way the oxide layer grows

in conditions of localised and repeated arc discharges [12]. Also the optical topography of Fig. 3-b

shows the morphology of the treated surface: the PEO treatment causes a considerable increase in

roughness (Ra goes from 1.5 to 2.1 µm) of the MMC, which is in agreement with the observed

morphologies. EDS analysis of the surface (Fig. 3-c) confirmed the presence, in the outermost layer

of the treated surface, of compounds containing K, P, Si and Ca, which form due to the interaction

with the electrolyte.

(a) (b)

(c)

Fig. 3. SEM image (a) and optical topography (b) of the PEO layer on the composite, which

highlight the typical morphology of the material with craters and bubbles that are ejected during

solidification of the PEO layer. The EDS spectrum in (c) shows the presence, in the outermost layer

of the treated surface, of compounds containing K, P, Si and Ca, which form due to the interaction

with the electrolyte.


The SEM image of the fracture section of the MMC+PEO in Fig. 4-a shows the uniformity of the

thickness of the PEO layer (evaluated by image analysis of 30±5 µm) and also makes it possible to

evaluate the excellent adhesion of the conversion layer to the substrate, given no detachment on the

interface. The fracture section shows, moreover, increased compactness that is manifested in the

innermost zone of the coating, where repeated discharge events favour densification of the layer [6].

From the optical micrograph in Fig. 4-b, it can be noted how the reinforcement particles are

incorporated into the layer that is in the process of being formed, as PEO treatment consists of the

chemical conversion of the substrate: in particular, reinforcement particles (Al2O3) can be seen

crossing the substrate/coating interface and helping improve the adhesion of the conversion layer.

PEO treatment gives rise to a remarkable increase in the surface hardness (which goes from 141 HB

to 1100 HV0.1) of the MMC. The nano-indentation measures (elaborated according to the Oliver and

Pharr model [27] using a Poisson’s module value of ν=0.22) carried out after removal of the

“technological layer”, also show that the PEO layer has intrinsic hardness of 8.1±1.5 GPa and

elastic modulus of 170±15 GPa. The average indentation depth was 1946 nm, therefore it amounted

to less than 1/10th

of the coating thickness, thus excluding any significant substrate contribution to

the mechanical response [28].

(a) (b)

Fig. 4. SEM image of the fracture section of MMC+PEO (a) and optical micrograph in cross section

of the same (b): reinforcement particles can be noted, shown with a white arrow, which cross the

substrate/coating interface and help improve the anchorage of the conversion layer. Reinforcement

particles, incorporated into the PEO layer during oxidation of the Al matrix, are also visible in (b).

Tribological behaviour. Carrying out tribological tests in slider-on-cylinder configuration

made it possible to evaluate the positive effect of PEO treatment on the sliding wear resistance of

the composite under investigation. The results are outlined in the histograms in Fig. 5, which

show the maximum wear scar depth, revealed by stylus profilometer, both on the sliders in

MMC+PEO and on the untreated MMC, as a function of the test conditions. It is clear how the

treatment lead to a significant increase in the wear resistance of the composite, particularly

when increasing the applied load and sliding velocity. On the one hand, this result can be

attributed to the increase in surface hardness induced by the conversion treatment, on the other hand

by the good adhesion of the PEO layer to the substrate, which clearly gives it greater load bearing

capacity.

PEO Layer

Mounting resin


Fig. 5. Maximum wear scar depth measured on stationary sliders (untreated composite (MMC) and

composite treated by plasma electrolytic oxidation (MMC+PEO)) at the end of the tests (10 km), as

a function of the applied load and sliding velocity.

The untreated composite only presented mild tribo-oxidative wear regime at the lowest load (10 N)

and slowest sliding velocity (0.6 m/s) considered in the experiment, while an increase in both the

load and velocity increasingly lead to transition towards a severe wear regime by delamination, as

the test conditions became more severe.

In the mild wear regime, the reinforcement particles of the composite firstly support the applied

loads, limiting the wear of the aluminium matrix and, secondly, exert a microabrasion action

towards the steel countermaterial. This leads to the formation of iron based debris that, during

sliding contact, remaining interposed between the contact surfaces, oxidise and continually transfer

onto themselves, producing a compact layer of Fe oxides onto the wear scars. This is shown in the

SEM images and x-ray maps of the wear scars in Fig. 6 (a and b-d) and is typical of the tribo-

oxidative wear regime. Typical fine and powdery wear debris, mainly consisting of Fe oxides, were

collected in these conditions, as shown in Fig. 6 (e and f).

1

10

100

1000

10 20 30 40 50

MMC 0.6 m/s

MMC+PEO 0.6 m/s

MMC 1.8 m/s

MMC+PEO 1.8 m/s

Wea

r d

ep

th,

µµ µµm

Load, N


(a) (b) (c) (d)

(e)

(f)

Fig. 6. SEM micrographs (a) and corresponding x-ray maps (b, c, d) of untreated MMC, in mild

tribo-oxidative wear regime. X-ray maps show the presence of a Fe-O transfer layer on the wear

scar. The morphology of the corresponding wear debris is shown in (e), whilst the composition of

debris (mainly consisting of Fe oxides) is shown in the EDS spectrum (f).

Al kα1 Fe kα1 O kα1 Fe kα1 O kα1


The graph in Fig. 7-a shows, for this regime, the characteristic results of the friction coefficient and

system wear (slider + cylinder), as a function of the sliding distance (test conditions: 10 N, 0.6 m/s).

It can be noted how the friction coefficient reaches a stationary value of approx. 0.55, which is

consistent with the presence of the compact layer of interposed Fe oxides [29]. The system wear,

however, has a negative value and is consistent with the fact that the interposition of the oxide layer

leads to a positive shift of the displacement transducer.

As the applied load is increased, the reinforcement particles partially fracture and are pulled out of

the metal matrix, therefore they loose their ability to limit the wear of the MMC and begin to form a

particularly abrasive “third body” [6]. This leads to a considerable increase in the wear damage of

both the MMC and the countermaterial. In such conditions, this helps transition towards a severe

wear regime by delamination, with an additional abrasion component exerted from the wear debris,

which is clear from the morphology of the scars (Fig. 8-a) and wear debris (Fig.8-b). In particular,

flake-like wear debris (as those shown in fig. 8-b) detach from the mechanically mixed layer

(MML), which forms on the worn surface as a consequence of plastic deformation, material

transfer, interactions with the environment and mechanical mixing [30]. The MML therefore

consists of material from both the MCC and the steel counterface, as shown by the EDS spectrum in

Fig. 8-c.

(a) (b)

Fig. 7. Coefficient of friction and system wear as a function of sliding distance for the untreated

composite, in mild wear (load 10 N, velocity 0.6 m/s) (a) and severe wear (load 20 N, velocity 1.8

m/s) (b) regime.

The tests carried out by increasing the sliding velocity to 1.8 m/s, led to a severe wear regime

already at the lowest load in the experiment (10 N), as highlighted by the data in Fig. 5. An increase

in sliding velocity leads to, in fact, an increase in the heat dissipated due to friction [31, 32], with a

consequent reduction of the resistance to the plastic flow of the material, of a growing nature to the

increase of the applied loads. This involves, on the one hand, higher wear rates, on the other, a

reduction of the coefficient of friction, up to values of approx. 0.4, as the graph in Fig. 7-b shows.

0

0,2

0,4

0,6

0,8

1

0

200

400

600

800

1000

0 2000 4000 6000 8000 10000

MMC, 20 N - 1,8 m/s

coefficient of friction

system wear

Co

eff

icie

nt

of

fric

tio

n Sy

ste

m w

ear, µµ µµ

m

Sliding distance, m

0

0,2

0,4

0,6

0,8

1

-60

-40

-20

0

20

40

60

80

100

0 2000 4000 6000 8000 10000

MMC 10 N - 0.6 m/s

coefficient of friction

system wear

Co

eff

icie

nt

of

fric

tio

n Sy

ste

m w

ear, µµ µµ

m

Sliding distance, m


(a) (b)

(c)

Fig. 8. SEM micrographs of wear scar (a) and a large fragment from the mechanically-mixed layer

(b) of untreated MMC, in severe delamination wear regime. The typical composition of wear debris

from the MML, consisting of elements from both the MMC (Al) and from the countermaterial (Fe),

is shown by the EDS spectrum in (c).

In the case of the PEO-treated composite, in order to highlight the possible presence of critical load

and/or velocity values for the transition from mild to severe wear, the applied loads were increased

up to 50 N. Fig. 5 shows how damage for wear, is a substantial minor result and is typical of a tribo-

oxidative regime for loads up to 50 N, at 0.6 m/s, and up to 30 N, at 1.8 m/s. The transition from

tribo-oxidative mild wear to severe wear by delamination, with complete removal of the PEO layer,

is clear from the coefficient of friction and system wear plots, shown in the graph in Fig. 9.

unworn MMC

wear scar


Fig. 9. Coefficient of friction and system wear, as a function of sliding distance for the PEO-treated

composite, corresponding to the critical load for transition from the mild wear to the severe wear

regime.

Fig. 10 shows a portion of the wear scar on MMC+PEO, produced in conditions of mild tribo-

oxidative wear and with compact Fe oxide layers, as a consequence of wear of the countermaterial.

(a) (b) (c) (d)

Fig. 10. SEM micrographs (a) and corresponding x-ray maps (b, c, d) of the wear scar on the PEO-

treated MMC, in mild tribo-oxidative wear regime. X-ray maps show the presence of a compact Fe-

O transfer layer on the wear scar, formed as a consequence of wear of the countermaterial.

Therefore, the wear behaviour of the untreated composite can be summarized as shown in Fig. 11:

under applied load and sliding speed lower than critical values, mild wear by tribo-oxidation takes

place (Fig. 11-a), while severe wear by delamination (with detachment of flake-like debris from the

mechanically mixed layer (MML) on the surface of the MMC) is the main wear mechanism (Fig.

11-b) above critical load and speed values.

In particular, in the mild wear regime, tribo-oxidative wear of the steel counterface, mainly due to

the abrasive action of hard reinforcement particles, leads to the formation of a protective Fe-O layer

on the surface of the MMC. Instead, in severe wear, the avulsion of reinforcement particles starts to

occur due to the plasticization of the Al matrix. These phenomena enhance the severity of wear and

do not allow the formation of a stable and protective Fe-O layer.

Conversely, the PEO-treatment shifts the transition from mild to severe wear towards higher values

of applied load and sliding speed (Fig. 5), because the alumina PEO layer protects the matrix from

both plastic deformation and frictional heating and holds the reinforcement particles in place.

0

0,2

0,4

0,6

0,8

1

1,2

1,4

0

200

400

600

800

1000

0 2000 4000 6000 8000 10000

MMC+PEO, 40 N - 1,8 m/s

coefficient of friction system wear

Co

eff

icie

nt

of

fric

tio

n Sy

ste

m w

ea

r, µµ µµm

Sliding distance, m

Al kα1 Fe kα1 O kα1


(a) (b)

Fig. 11. Schematic of wear mechanisms for the MMC: mild tribo-oxidative wear (a)

and severe delamination wear (b) regimes.

Conclusions

The dry sliding tests, carried out on untreated and PEO treated AA2618/20%volAl2O3p composite

against steel, under various load and velocity conditions, have made it possible to highlight the

following behaviours.

The untreated composite only shows mild tribo-oxidative wear at low loads and low sliding speeds

(10 N - 0.6 m/s), while when the severity of the test conditions is increased, transition towards

severe wear by delamination is observed. This was ascribed to two main factors: (i) the achievement

of a critical temperature of the matrix, which led to a reduction in the plastic flow resistance of the

matrix, as well as to (ii) phenomena of fragmentation and avulsion of the reinforcing particles. The

application of PEO, thanks to the increase in surface hardness, the excellent adhesion to the

substrate and the thermal barrier effect, moved transitions from mild to severe wear towards more

severe test conditions, in terms of both load and velocity, compared to the untreated material.

WEAR

CYLINDER

SLIDER

CYLINDER

SLIDER

CYLINDER

SLIDER

CYLINDER

SLIDER

CYLINDER

SLIDER

CYLINDER

SLIDER

WEAR

WEARWEAR

WEAR

WEAR

Al2O3

Al2O3

FeOx

Al2O3

FeOx

Al2O3

Al2O3

Al2O3

Debris

MML


Acknowledgements

We would like to thank Dr. James A. Curran of Keronite International Ltd (Cambridge, UK) for

carrying out the treatment and for the information he provided.

References

[1] F. Bonollo, L. Ceschini, G. L. Garagnani, G. Palombarini, A. Zambon: Mater. Sci. and Techn.

Vol. 10, Issue 6 (1994), p.481.

[2] L. Ceschini, C. Bosi, A. Casagrande. G. L. Garagnani: Wear Vol. 251 (2001), p.1377.

[3] L.Ceschini, A.Morri, G.Sambogna, M.C.Breslin, M.Fuller: Int. J. of Mat. & Prod. Techn.,

Vol. 17, Issue 3/4 (2002), p.165.

[4] R. L. Deuis, C. Subramanian, J. M. Yellup: Wear Vol. 201 (1996), p.132.

[5] R. L. Deuis, C. Subramanian, J. M. Yellup: Comp. Sci. and Techn., Vol. 57 (1997), p. 415.

[6] M. Godet: Wear Vol. 136 (1990), p. 29.

[7] A. L. Yerokhin, X. Nie, A. Leyland, A. Matthews, S. J. Dowey: Surf. and Coat. Techn. Vol.

122 (1999), p.73.

[8] A. L. Yerokhin, A. Shatrov, V. Samsonov, P. Shashkov, A. Pilkington, A. Leyland, A.

Matthews: Surf. and Coat. Techn. Vol. 199, Issue 2-3 (2005), p.150.

[9] L.O. Snizhko, A.L. Yerokhin, A. Pilkington, N.L. Gurevina, D.O. Misnyankin, A. Leyland,

A. Matthews: Electrochimica Acta Vol. 49 (2004), p. 2085.

[10] E. Matykina, R. Arrabal, P. Skeldon, G. E. Thompson: Surf. Interface Anal. Vol. (2010), p.

221.

[11] R.O. Hussein, X. Nie, D.O. Northwood, A. Yerokhin, A. Matthews: J. Phys. D: Appl.

Phys. Vol. 43 (2010), p.1.

[12] J. A. Curran, T. W. Clyne: Surf. and Coat. Techn. Vol. 199 (2005), p.168.

[13] J.A. Curran, T.W. Clyne: Surf. and Coat. Techn. Vol. 199 (2005), p.177.

[14] A.L. Yerokhin, X. Nie, A. Leyland, A. Matthews: Surf. and Coat. Techn. Vol. 130 (2000) p.

195.

[15] A.L. Yerokhin, A. Leyland, A. Matthews: Appl. Surf. Sci. Vol. 200 (2002), p.172.

[16] L. Ceschini, E. Lanzoni, C. Martini, D. Prandstraller, G. Sambogna: Wear Vol. 264 (2008) p.

86.

[17] R. Arrabal, E. Matykina, T. Hashimoto, P. Skeldon, G.E. Thompson: Surf. and Coat.

Techn. Vol. 203 (2009), p. 2207.

[18] S.V. Gnedenkov, O.A. Khrisanfova, A.G. Zavidnaya, S.L. Sinebryukhov, V.S. Egorkin, M.V.

Nistratova, A. Yerokhin, A. Matthews: Surf. and Coat. Techn. Vol. 204 (2010), p. 2316.

[19] A. L. Yerokhin, L. O. Snizhko, N. L. Gurevina, A. Leyland, A. Pilkington, A. Matthews: J.

Phys. D: Appl. Phys. Vol. 36 (2003), p. 2110.

[20] M. Cartier, Handbook of Surface Treatments and Coatings (HEF Group, Professional

Engineering Publishing, New York 2003).

[21] X. Nie, E.I. Meletis, J.C. Jiang, A. Leyland, A.L. Yerokhin: Surf. and Coat. Techn. Vol. 149

(2002), p. 245.

[22] R.C. Barik, J.A.Wharton, R.J.K. Wood, K.R. Stokes, R.L. Jones: Surf. and Coat. Techn. Vol.

199 (2005), p.158.


[23] Y. Guan, Y. Xia, G. Li: Surf. and Coat. Techn. Vol.202 (2008), p. 4602.

[24] S. Cui, J. Han, Y. Du, W. Li: Surf. and Coat. Techn. Vol. 201 (2007), p. 5306.

[25] J. Lee, S. Kang, J. Han: Wear Vol. 264 (2008), p. 75.

[26] L. Ceschini, G. S. Dahen, G. L. Garagnani, C. Martini: Wear Vol. 216 (1998), p. 229.

[27] W. C. Oliver, G. M. Pharr: J. Mat. Research Vol. 7 (1992), p.1564.

[28] L.E. Samuels, T.O. Mulhearn: J. of the Mech. and Phys. of Solids 5 (1957) 125-134.

[29] J.L. Sullivan, S.G. Hodgson: Wear Vol. 121 (1988), p. 95.

[30] B. Venkataraman, G. Sundararajan: Wear Vol. 245 (2000), p. 22.

[31] G. Straffelini: Attrito e usura. Metodologie di progettazione e controllo (Tecniche Nuove,

Milano 2005).

[32] I. M. Hutchings: Tribology, Friction and Wear of Engineering Materials (Edward Arnold,

UK 1992).


Strengthening Evaluation in a composite Mg-RE Alloy using TEM

Marcello Cabibbo

Department of Mechanical Engineering, Università Politecnica delle Marche, 60131-Ancona, Italy

email: [email protected]

Keywords: Mg-RE alloy, composite, strengthening, TEM, compression

Abstract. Magnesium alloys containing rare earth elements are known to have high specific

strength and corrosion resistance. The addition of SiC ceramic particles makes the metal matrix

composite stronger with better wear and creep resistance and a still good machinability. The role of

the reinforcement particles to the enhanced strength can be quantitatively evaluated using

transmission electron microscopy (TEM). This paper presents a quantitative strengthening

evaluation in a SiC Mg-RE composite alloy. The different contributions were determined by TEM

inspections. The microstructure strengthening mechanism was studied after room temperature

compression specimens. The way of combining the different contributions and the comparison to

the measured yield stress, is also discussed and justified.

Introduction

The increasing demand on lightweight materials in aerospace, automotive, electronics and other

technological applications, makes the magnesium alloys an attractive metallic material [1, 2]. In the

last two decades, magnesium alloys have being used in a broad variety of structural applications due

to their high strength to density ratio. Magnesium alloys are versatile and can be easily shaped,

using nearly all methods including casting and metal working, as extrusion and rolling. Yet, the

major drawbacks in using magnesium alloys are the limited mechanical properties (strength,

hardness, corrosion resistance) especially at high temperatures [2]. Precipitation hardening usually

improves the magnesium alloy mechanical properties resulting in a severe reduction of the

corrosion resistance [2,3].

Rare earth (RE) elements are known to improve the mechanical properties and corrosion resistance

of magnesium alloys [4-18]. These aspects justify the high cost of production, especially when used

in aircraft or space applications. Rare earth elements, which are usually added to the magnesium

alloys, are: yttrium [6-13], gadolinium [7-10, 14-18], neodymium [6,12,14,17], dysprosium [10,17],

samarium [11]. The three most frequently added rare earth elements are Y, Gd and Nd. The

maximum solubility of yttrium in solid magnesium is about 11 wt.% at 840 K. However, the

response to ageing is promoted primarily by the neodymium [4].

The high performance of these alloys are generally obtained through the nano-scaled precipitates

that are able to effectively hinder the dislocation sliding on basal plane during deformation

[6,19,20]. The precipitation sequence has been reported to be as: supersaturated solid solution (sss)

→ β”→β’→β(Mg24Y5) stable bcc precipitate, in the case of Y [21,22]. The sequence is the same in

the case of Gd with the formation of a stable fcc phase β Mg5Gd [23]. In the case of magnesium-RE

alloys containing two different rare earth elements (alloys containing Gd and Y, Gd and Nd, or Dy

and Nd) the sequence of precipitation of the secondary phase particles changes very little giving a

stable β-phase of the Y-rich and Gd-rich types. The stoichiometry of the stable phase is of the

Mg5X type, where X is the sum to 1 of the existing rare-earth elements [9,20,22]. Other studies,

Antion et al. [12] to cite one, reported different and more complex stoichiometry for the β-stable Y-


Nd-containing precipitates. For instance, in the case of WE54 alloy, the stable β-precipitate was

reported to have a Mg12NdY or Mg24Y2Nd3 stoichiometry.

A further effective strengthening method of the Mg-RE alloy is by addition of ceramic particles

[24,25]. The technological interest on the discontinuous SiC alloys composites is due to the

economic production of the Si-carbides [25-27]. One of the key advantage is also on the possibility

of shaping the SiC particles by standard metallurgical processes such as forging, rolling, extrusion

[26,27]. Composite metals usually widen the range of applications to technological fields where

high specific stiffness is needed and to the creep resistant temperature regimes.

A formulation of the strengthening contributions of the discontinuous composites in a metallic

matrix has been proposed by Chou and Kelly in [28]. This formulation has several assumptions,

which, other than the alignment of the fiber particles, are that the matrix microstructure must not be

affected by voids and there must be a perfect bonding between the SiC particles and the matrix.

The strength contributions coming from the presence of the SiC particles is strongly related to the

local stress that the presence of the particles introduces within the metal matrix. Particle volume

fraction and distribution affect the stress between particles and the surrounding matrix under plastic

flow [29-32]. Some results showed that plastic flow initiates in microstructure regions where the

reinforcement concentration is low to propagate toward regions with clustered reinforcement

particles [33].

Strengthening contributions are of two types: matrix strengthening and SiC particle strengthening.

Matrix contributions to strengthening is due to the grain and cell size (Hall-Petch), to the presence

of secondary phase particles (Orowan strengthening), and to the reinforcement generated by twining

[34,35]. On the other hand, the dislocation density increment in the surrounding regions of the SiC

particles can be quantitatively described taking into account all the significant contributions to

strengthening, generated by the presence of discontinuous SiC particles under plastic deformation.

These contributions are: the load transfer from the matrix to the SiC particles, the thermal mismatch

and a geometrical mismatch between particles and surrounding matrix, a small contribution coming

from the presence itself of the coarse SiC particles in the matrix (SiC Orowan), and the cell size

reduction induced during plastic deformation by the presence of the SiC particles [27,29,36-42].

In the present work, matrix and SiC particle strengthening contributions of a Mg-RE SiC composite

have been quantitatively determined through TEM inspections. These results were directly

compared to the yield strength obtained by room temperature compression test. The way all the

strengthening contribution were added up is also discussed.

Experimental details

An Mg-RE alloy reinforced by 13pct. SiC was used for this study. The chemical composition of the

alloy is (wt.%): 5.25Y, 1.75Nd, 1.0Gd, 0.75Dy, 0.5Zr, Mg bal. The alloy was produced by powder

metallurgy. Mixing of the matrix powders with SiC particles was carried out in an asymmetrically

moving mixer and then in a ball milling. The sintered alloy was extruded at 673 K with a 4MN

horizontal extrusion press.

Microstructure was characterized by optical microscopy (OM) and by electron microscopy (TEM).

Specimen surfaces for LM were prepared by mechanical polishing and chemical etching with a

solution of HNO3 (40 pct.) and C2H5OH (60 pct.). TEM discs were polished, dimpled and prepared

mechanically using a precision ion polishing system Gatan-691 PIPSTM

working with an incident

angle of 5° and a beam energy of 4.5 eV. A TEM Philips®

CM200 equipped with double-tilt

specimen holder was used. All the TEM inspections were carried out at 200 kV.

Mean grain and cell size were determined with line intercept methods. Particle size and mean

particle distance were measured using an image analysis, Image Pro® plus 4.5, software.

Compression tests were carried out at room temperature using a servo-hydraulic INSTRON testing

machine. Cylindrical specimens of 8 mm in diameter and 12 mm long were deformed at an initial

strain rate of 2.8×10−4

s−1

.


Experimental Results

Fig. 1 shows the microstructure of the untested Mg-RE composite alloy; Fig. 1(a) is a low-

magnification OM, Fig. 1(b) is a TEM image. SiC particles are not uniformly distributed within the

matrix, they rather cluster leaving zones of the matrix with few small isolated reinforcing particles.

The mean SiC particle size is of 12.2 ± 0.2 µm. Traces of the extrusion process are clearly visible as

microstructure streaks generated in the matrix by the presence of the hard undeformed SiC

reinforcing particles. Grain sizes are quite small, they are not well visible in the LM image, while,

in the TEM image of Fig. 1(b) a detailed matrix zone, constituted by few grains, is shown.

Fig. 1. OM (a) and TEM (b) images of the undeformed SiC Mg-RE composite alloy.

Grains are equiaxed with an average size of 3.2 ± 0.1 µm. Grains are generally decorated by

twining and cuboid-shaped secondary phase particles. Twins are quite narrow, parallel each other

and distant apart within the grains less than one micron (typically 0.3 to 1.0 µm). In some cases,

twins extend from one to the adjacent grain, causing grain boundary deflections. Their length ranges

~ 1 to 3.5 µm.

Fig. 2 shows the compression true stress-strain curve and the microstructure after deformation.

Stress-strain curve is clearly affected by work hardening (Fig. 2(a)). The maximum stress was 372

MPa, the yield stress 258 MPa.

SiC particles tend to cluster and their mean size was not affected by the deformation (Fig. 2(b)).

Dislocations pile-up was observed in the vicinity of the secondary phase particles. Twins were still

present in the microstructure and their length and spacing seemed not to be affected by the

compression deformation. Grain size did not change significantly, its mean size being 3.05 ± 0.05

µm, with a mean cell size of 1.20 ± 0.05 µm. Cuboid-shaped particles are quite diffuse within the

grains and along grain boundaries (Fig. 2(c) and 2(d)). Distribution and size of these particles seem

not to be affected by the plastic deformation. These particles appear to be randomly distributed.

Their size ranges from 20 nm to 150 nm.

Other studies reported cuboid-shaped and rectangular particles in Mg-RE (WE54, WE43 and WN42

alloys) with similar size and distribution [6,7,9,10,12-14,18,23,43]. In most of these studies the

secondary phase was identified as β Mg5RE where RE stands for all the rare earth elements

contained in the alloy. The effective stoichiometry of the rare earth elements has been proposed in

different studies as RE = AxBy, with A and B two generic rare-earth elements. The stoichiometry

factors x and y can be respectively 0.66 and 0.33, as reported by Apps et al. in [10] for a Y- Nd-

containing magnesium alloy, using quantitative dispersive energy spectroscopy. These coefficients

can alternatively be more complex as reported by Nie and Muddle in [6] for Y- La- and Y- Ce-

containing alloys, but also more simple as in [7] in a Mg-Y-Ga alloy. In the present case the studied

alloy has four rare-earth elements and this can surely make the stoichiometry more complex. The

EDS analyses performed during TEM inspections revealed the presence, in the secondary phase

particles, of all the four rare earth elements. Thus, on the basis of the above mentioned

considerations and the EDS results, the identification of the β phase has generically been left as

Mg5RE, with RE being of the form RE = Yx1Ndx2Gdx3Dyx4, with x1+ x2+ x3+ x4 =1.

25 µm a) b) 1 µm b)


Fig. 2. (a) Room temperature compression true stress-strain curve of the Mg-RE SiC composite

alloy. Specimens were deformed at a strain rate of 2.8×10−4

s−1

. In order to better determine the

Yield strength, the reported σ-ε curve is a datapoints average over three different room temperature

compression tests. TEM microstructure after deformation; (b) low-magnification BF-TEM, (c) and

(d) are two higher magnification BF-TEM taken in two different areas of the TEM disc, where

dimension and distribution of the secondary phase particles are documented.

Discussion

The strengthening mechanism in a metal matrix composite is related to the microstructure feature of

the matrix, in one hand, and to the presence of the reinforcing particles within the matrix, on the

other. Thus, the meaningful metal matrix contributions needed to be taken into account are: Hall-

Petch strengthening due to grain and cell size [7,26,35,36,39,44,46-48], twining [11,34],

dislocation-secondary phase particle interaction by the Orowan process [7,26,36,37,39,42,49]. The

contributions due to the presence of the reinforcing particles are: load transfer from matrix to

particles [29,40,45], dislocation generation due to the different thermal expansion existing between

the soft matrix and the hard reinforcements [29,37,38,41,50-52], dislocation generation required

geometrically during deformation [53], Orowan due to the coarse reinforcing particles [42,47].

In the following all these strengthening contributions for the Mg-Y-Nd-Gd-Dy-Zr SiC composite

alloy will be addressed and discussed.

Matrix strengthening.

Hall Petch strengthening. The Hall-Petch relationship (Eq. 1) includes the cell contribution and

thus the relationship used in this study is a modified Hall-Petch equation, according to McQueen

[35,48], Miller et al. [46], Hausselt and Nix [42]:

2/12/1

'

cg

HPd

K

d

K+=∆σ (1)

0

100

200

300

400

0 0,008 0,04 0,08 0,12 0,16 0,2εεεε

σ,

σ,

σ,

σ,

MP

aa)

1 µm b) 200 nm c) 50 nm d)


where K and K’ are the Hall-Petch constants for the grains and cells (K = 0.28 MNm-3/2

for

magnesium [36], and K’ = 0.05 MNm-3/2

[46]), dg and dc are the mean grain and cell size,

respectively.

Putting in the Eq. 1 the calculated mean grain and cell size values, the resulting Hall-Petch

contribution to strengthening (∆σHP) is 205 ± 5 MPa.

Twining. Twins can contribute to the matrix strengthening as the reorientation of grains during

plastic deformation can determine textural modification which in turns reflect a slight change in the

lateral spacing and orientation of the existing twins. Twins are reported to be as { }2110 type [34].

During plastic deformation, they slightly developed a sideways growth, and in some cases twins

were observed to entirely consume the grains where they had formed. This can explain the quasi-

equiaxed grain structure under compression, where work-hardening still acts, and a limited grain

and cell size changes occur. The grain and cell size and morphology strongly resemble the

microstructure of the as-extruded undeformed alloy. Twining strengthening contribution was found

to be 6 ± 1 MPa [34,53,54].

Orowan strengthening due to the secondary phase particles. Secondary phase precipitation

hardening plays an important role in the alloy matrix strengthening. The moving dislocations bow

out as they encounter the nanometric particles, they bypass them leaving back looping dislocations.

The Orowan strengthening contribution for cuboid-shaped particles can be written as [55]:

+

Λ

+=∆ B

r

AEbMOrowan

0

ln)1(4 λνπ

σ (2)

where M = 6.5 is the Taylor factor for Mg, A = 1/(1-ν) = 1.37, B = 0.6, for screw dislocations, A =

1, B = 0.7, for edge dislocations, ν = 0.27 is the Poisson ratio, E is the Young modulus (44.4 GPa)

determined by the shear modulus of Mg, G = E/2(1+ν), λ is the average interparticle spacing, Λ is

the value of the harmonic mean between λ and r

rrs

2

22

π= , where r is the particle radius, r0 the

inner cut-off particle radius. Geometric mean was used for the A and B coefficients. The mean

spacing λ was determined using Eq. (3):

rf

r23

225.1

2/1

ππλ −

= (3)

r and f being the particles mean size and volume fraction, respectively.

The Orowan relationship of Eq. (2) is equivalent to the ones reported by other authors

[36,45,48,50]. The r, λ values were 44 ± 2, and 155 ± 5 nm, respectively. The particle volume

fraction, f, was 0.12 ± 0.01. Using Eq. (2) and Eq. (3), introducing the values of r and λ, the

corresponding Orowan strengthening contribution after compression was: 44 ± 1 MPa.

Composite particle strengthening.

Load transfer from matrix to particles. According to the shear-lag model proposed by

Nardone and Prewo in [29], the composite particles do contribute to alloy reinforcement carrying a

fraction of the load from the matrix. This strengthening strongly depends on the shape and

morphology of the particles; it specifically depends on the particle aspect ratio [45]. Thus, the

proposed relationship is (Eq. (4a)):

)1(4

)(1 ff

L

AtLymymLT −+

++= σσσ (4a)


where σym is the unreinforced matrix yield stress, f the particle volume fraction, L the particle size

facing the load direction, t the mean particle thickness, A = L/t the particle aspect ratio. For

equiaxed particles, as in the present case, the Eq. (4a) reduces to the Eq. (4b):

fymymLTLT 5.0σσσσ =−=∆ (4b)

Thermal expansion between the matrix and the reinforcement. An increase of dislocation

density due to the different thermal expansion coefficients (CTE) between the SiC particles and the

magnesium matrix gives an additional strengthening contribution. On cooling from extrusion,

dislocations are generated around the SiC particles resulting in high local dislocation densities.

Therefore, this matrix strengthening contribution is a local phenomenon. The amount of the thermal

stress induced by the presence of the reinforcement depends upon the volume fraction, morphology,

size of the particles and the effective temperature change. The relatively large thermal expansion

difference between the matrix and the SiC particles creates a substantial misfit strain at the SiC-Mg

interface. Thermal stress can be partially released by the dislocation generation and accumulation in

the surroundings of the reinforcement surfaces. Arsenault et al. [37,41] assumed that no plastic

relaxation of the thermal stresses occurs for temperatures up to 473 K.

Thus, according to [36], [43], [50] and [51], the induced dislocation density can be calculated as Eq.

(5a),

'

1

)1('

1

)1( tfb

TBf

tf

f

b

BT

−

∆∆=

−=

αερ (5a)

where B is a constant which is equal to 12 for equiaxed particles, ε = ∆α⋅∆T is the misfit strain, ∆T

is the temperature variation, ∆α = 21⋅10-6

K-1

is the thermal expansion difference between matrix

and SiC, f the particle volume fraction, b = 3.21⋅10-10

m the magnesium Burgers vector, and t’ the

minimum size of the SiC particles. The increased dislocation density yields a strengthening factor

of (Eq. (5b)):

TT Gb ρµασ 1=∆ (5b)

where α1 = 0.35, and G = 17480 MPa is the shear modulus for Mg.

Since the average residual stress generated by the thermal expansion is of tension nature, it is

actually a negative contribution to the strengthening of the magnesium composite alloy [37, 38].

Reinforcement particle geometrically mismatch. The different nature of the ceramic SiC

particles respect to the metallic magnesium matrix causes the generation of geometrically necessary

dislocations resulting in a strengthening of the alloy. The resulting matrix-to-particle misfit depends

on the reinforcement size and morphology [53]. The density of the geometrical necessary

dislocations is given by Eq. (6a) [27]:

'/8 btf pGEO ερ = (6a)

where εp = 0.28 is the plastic strain. The corresponding strengthening contribution is thus (Eq.

(6b)):

( )GEOGEOM Gb ρµασ 1=∆ (6b)


Orowan due to the coarse reinforcing particles. The presence of coarse SiC particles can also

generate a strengthening contribution of the type of fine secondary phase particles. That is, an

Orowan strengthening contribution, even if very low, can be also generated by the SiC particles.

The Orowan relationship of Eq. (2) has thus been used for calculating this strengthening

contribution. Mean SiC particle size, r, and mean interparticle spacing, λ, were, respectively: 12.2 ±

0.2 and 13.4 ± 0.2 µm. Mean interparticle spacing was measured using line intercept method.

Calculated alloy strengthening vs. mechanical compression yield stress. Tab. 1 reports all the

strengthening contributions due to the matrix and to the SiC reinforcement.

The different strengthening contributions coming from the magnesium matrix and from the SiC

reinforcements are expected to be fully consistent to the mechanical yield stress value obtained by

the compression test. A straightforward linear sum of all the calculated strengthening contributions

(∆σHP+∆σTwin+∆σOr+∆σLT+∆σT+∆σGEO+∆σOr-SiC) is not the proper solution, as discussed by Liholt [56],

and Clyne and Whithers [57]. Another solution could be the combination through square root sum

of squares (Eq. (7)):

((∆σHP)2+(∆σTwin)

2+(∆σOr)

2+(∆σLT)

2+(∆σT)

2+(∆σGEO)

2+(∆σOr-SiC)

2)

1/2 (7)

but this would mean that all the strengthening contributions are unevenly distributed throughout the

metal matrix, which is not the case.

∆∆∆∆σσσσHP ∆∆∆∆σσσσTwin ∆∆∆∆σσσσOr ∆∆∆∆σσσσLT ∆∆∆∆σσσσT ∆∆∆∆σσσσGEO ∆∆∆∆σσσσOr-SiC

205 ± 5 6 ± 1 44 ± 1 18.0 ± 0.5 28 ± 1 11.0 ± 0.5 5.8 ± 0.2

Table 1. All the calculated strengthening contributions are listed in this table. Matrix: Hall-Petch

(∆σHP), twining (∆σTwin), secondary phase Orowan (∆σOr); SiC reinforcing particles: load transfer

from matrix to SiC particles (∆σLT), different thermal expansion between matrix and SiC (∆σT),

strength due to geometrically induced dislocations (∆σGEO), Orowan due to the coarse SiC particles

(∆σOr-SiC).

In the present case it is possible to assume that all the matrix contributions are uniformly distributed

throughout the microstructure of the alloy. Secondary phase particles are distributed in the grains

interior with no evidence of clustering or preferential paths. After compression, twins were

observed; grain and cell boundaries are by definition all over the microstructure in a polycrystalline

metallic material.

On the other hand, the reinforcing SiC particles are unevenly distributed throughout the

microstructure, as they are preferentially located at the grain boundaries.

Thereafter, a proper way of combining all the different contributions can be, a linear sum of only

those strengthening terms coming from the matrix, while the ones coming from the reinforcing

particles can be added quadratically, as square root of sum of squares (Eq. (7)):

∆σHP+∆σTwin+∆σOr+((∆σLT)2+(∆σT)

2+(∆σGEO)

2+(∆σOr-SiC)

2)

1/2.


room temperature

compression test

yield stress, [MPa] 258

linear, [MPa] 262 (+0.02)

quadratic, [MPa] 209 (-0.23)

linear + quadratic, [MPa] 254 (-0.01)

Table 2. Three different methods of combining all the calculated strengthening terms. A linear sum

(linear), a square root sum of squares (quadratic), linear sum of strengthening terms evenly diffused

throughout the matrix and square root sum of squares of the unevenly contributions coming from

the SiC particles (linear + quadratic). The yield stress as determined by compression test, is also

reported. In parenthesis, beneath the corresponding calculated strength, are reported the

discrepancies between the calculated and the mechanically obtained alloy yield strength.

Tab. 2 compares the three different methods of combining the strengthening terms to meet the yield

stress obtained by compression test. It appears that, the linear sum of strengthening contributions

slightly overestimate the mechanical yield stress, while the quadratic sum gives a significant

underestimation of the measured yield stress (23 pct.).

On the other hand, the linear sum of the evenly distributed matrix contributions, combined with a

quadratic sum of the unevenly SiC contributions, stays quite close to the measured yield stress.

Thus, the linear + quadratic sum appeared to be a good solution and actually the best among the

three possible combination methods.

Summary

Strengthening contributions of a SiC Mg-RE (RE =Y, Nd, Gd, Dy) composite alloy have been

determined by TEM inspections. The calculated yield strength has been compared to the yield stress

obtained by room temperature compression test.

It was found that the majority of the strengthening contribution comes from the Hall-Petch strength

due to grains and cells, which accounts for approximately 0.8 of the yield stress.

The secondary particles contribution is only one-fifth of the Hall-Petch strength, while the twin

contribution is negligible. Among the contributions coming from the presence of the SiC

reinforcement particles, the most important appeared to be the strength due to the dislocation

density increment generated by the different thermal expansion between the hard particles and the

softer matrix.

A mixed combination of the different strengthening terms was used to meet the yield stress obtained

mechanically. That is, a linear sum of the evenly distributed strengthening terms in the matrix, and a

quadratic sum of the unevenly distributed contributions (the ones coming from the presence of the

SiC particles): ∆σHP+∆σTwin+∆σOr+((∆σLT)2+(∆σT)

2+(∆σGEO)

2+(∆σOr-SiC)

2)

1/2.

This approach was compared to a simple linear and a quadratic sum of the strengthening terms,

showing a much better and satisfactory fitting to the measured yield stress.

Acknowledgment

The author is grateful to Mr. D. Ciccarelli for the mechanical compression tests.


References

[1] D.J. Lloyd: Int. Mater. Rev. Vol. 39 (1994), p.1.

[2] S. Logan in: Magnesium Technology 2007, edited by R.S. Beals, A.A. Luo, N.R.

Neelameggham and M.O. Pekguleryuz TMS publications, Warrendale, Pennsylvania (2007), p.

41.

[3] H. Ferkel and B.L. Mordike: Mater. Sci. Eng. Vol. A298 (2001), p. 193.

[4] I.J. Polmear: Metall. Trans. JIM Vol. 37 (1996), p.12.

[5] B. Smola, I. Stulikova, J. Pelcova and B.L. Mordike: J. All. Comp. Vol. 378 (2004), p. 196.

[6] J.F. Nie and B.C. Muddle: Acta Mater. Vol. 48 (2000), p. 1691.

[7] S.M. He, X.Q. Zeng, L.M. Peng, X. Gao, J.F. Nie and W.J. Ding: J. All. Comp. Vol. 427 (2007),

p. 316.

[8] D.J. Li, X.Q. Zeng, J. Dong, c.q. Zhai and W.J. Ding: J. All. Comp. Vol. 468 (2009), p. 164.

[9] S.M. He, X.Q. Zeng, L.M. Peng, X. Gao, J.F. Nie and W.J. Ding: J. All. Comp. Vol. 421 (2006),

p. 309.

[10] P.J. Apps, H. Karimzadeh, J.F. King and G.W. Lorimer: Scripta Mater. Vol. 48 (2003), p. 475.

[11] M. Suzuki, T. Kimura, J. Koike and K. Maruyama: Scripta Mater. Vol. 48 (2003), p. 997.

[12] C. Antion, P. Donnadieu, F. Perrard, A. Deschamps, C. Tassin and A. Pisch : Acta Mater. Vol.

51 (2003), p. 5335.

[13] C. Sanchez, G. Nussbaum, P. Azavant and H. Octor: Mater. Sci. Eng. Vol. A221 (1996), p. 48.

[14] K.Y. Zheng, J. Dong, X.Q. Zeng and W.J. Ding: Mater. Sci. Eng. Vol. A454-455 (2007), p.

314.

[15] Z. Zhang, L. Peng, X. Zeng, P. Fu and W. Ding: Mater. Charact. Vol. 60 (2009), p. 555.

[16] D. Li, Q. Wang, W. Ding: Mater. Sci. Eng. Vol. A428 (2006), p. 295.

[17] D. Li, J. Dong, X. Zeng, C. Lu and W. Ding: J. All. Comp. Vol. 439 (2007), p. 254.

[18] J.F. Nie, X. Gao and S.M. Zhu: Scripta Mater. Vol. 53 (2005), p. 1049.

[19] G. Song and D. StJohn: J. Light Met. Vol. 2 (2002), p. 1.

[20] L.L. Rokhin: Magnesium alloys cotaining rare earth metals (Taylor & Francis, London 2003).

[21] I. Anthony, S. Kamado and Y. Kojima: Mater. Trans. Vol. 42 (2001), p. 1206.

[22] G.W. Lorimer in: Porceedings magnesium technology, edited by G.W. Lorimer Institute of

Metals, London, UK (1986), p. 47.

[23] B. Smola, I. Stuhikovai, F. von Buch and B.L. Mordike: Mater. Sci. Eng. Vol. A324 (2002), p.

113.

[24] I.B. Cutler: Am. Ceram. Soc. Bull. Vol. 52 (1973), p. 425.

[25] J. Nasser, W. Riehemann and H. Ferkel: Mater. Sci. Eng. Vol. A234 (1997), p. 467.

[26] R.J. Arsenault: Mater. Sci. Eng. Vol. 64 (1984), p. 171.

[27] J.W. Luster, M. Thumann and R. Baumann: Mater. Sci. Techn. Vol. 9 (1993), p. 853.

[28] T.W. Chou and A. Kelly: Ann. Rev. Mater. Sci. Vol. 10 (1980), p. 299.

[29] V.C. Nardone and K.M. Prewo: Scripta Metall. Vol. 18 (1986), p. 25.


[30] I. Dutta, J.D. Sims and D.M. Seigenthaller: Acta Metall. Mater. Vol. 41 (1993), p. 885.

[31] T. Christman, A. Needleman and S. Suresh: Acta Metall. Vol. 37 (1989), p. 3029.

[32] P.B. Prangnell, S.J. Barnes, S.M. Roberts and P.J. Withers: Mater. Sci. Eng. Vol. A220

(1996), p. 41.

[33] S.F. Corbin and D.S. Wilkinson: Acta Metall. Mater. Vol. 42 (1994), p. 1311.

[34] S.E. Ion, F.J. Humphreys and S.H. White: Acta Metall. Vol. 30 (1982), p. 1909.

[35] H.J. McQueen: Metall. Trans. Vol. 8A (1977), p. 807.

[36] M. Mabuchi and K. Higashi: Acta Mater. Vol. 11 (1996), p. 4611.

[37] R.J. Arsenault, L. Wang and C.R. Feng: Acta Metall. Mater. Vol. 39 (1991), p. 47.

[38] G. Garmong: Metall. Trans. Vol. 5 (1974), p. 2183.

[39] W.S. Miller and F.J. Humphreys: Scripta Metall. Mater. Vol. 25 (1991), p. 33.

[40] G. Bao, J.W. Hutchinson and R.M. McMeeking: Acta Metall. Mater., Vol. 39 (1991), p. 1871.

[41] R.J. Arsenault and M. Taya: Acta Metall. Vol. 35 (1987), p. 651.

[42] J.H. Hausselt and W.D. Nix: Acta Metall. Vol. 25 (1977), p. 595.

[43] B. Smola and I. Stulikova: J. All. Comp. Vol. 381 (2004), L1-L2.

[44] P. Andersson, C.H. Caceres, and J. Koike: Mater. Sci. Forum Vol. 419-422 (2003), p.123.

[45] R.M. Aikin and L. Christodoulou: Scripta Metall. Mater. Vol. 25 (1991), p. 9.

[46] W.S. Miller, F.J. Humphreys and S.T. Derham, in: Deformation processing and modelling of

MMCs edited by R. Bahgrat, TMS (1990).

[47] J.K. Lee, Y.Y. Earmme, H.I. Aaronson and K.C. Russell: Metall. Trans. Vol. 11A (1980), p.

1837.

[48] H.J. McQueen and J.E. Hackett: Metal. Trans. Vol. 1 (1970), p. 2997.

[49] S.V. Kamat, J.P. Hirth and R. Mehrabian: Acta Metall. Vol. 37 (1989), p. 2395.

[50] R.J. Arsenault and N. Shi: Mater. Sci. Eng. Vol. 81 (1986), p. 175.

[51] Y. Flom and R.J. Arsenault: Mater. Sci. Eng. Vol. 75 (1985), p. 151.

[52] M. Vogelsang, R.J. Arsenault and R.M. Fisher: Metall. Trans. Vol. 17A (1986), p. 379.

[53] J.E. Harris and B.C. Masters: Proc. Roy. Soc. Vol. 292A (1966), p. 240.

[54] P.G. Partrige: Metall. Rev. Vol. 118 (1967), p. 169.

[55] R.O. Scattergood and D. Bacon: Phil. Mag. Vol. 31A (1975), p. 179.

[56] N. Liholt, in: Deformation of Multi-phase and Particle Containing Materials, edited by. N.

Hansen, A. Horsewell, T. Leffers and H. Liohot. Risǿ Nat. Lab., Roskilde, Denmark (1983),

p. 381.

[57] T. Clyne and P.J. Whithers: An Introduction to Metal Matrix Composites (Cambridge Press,

Cambridge, 1993).


Friction welding of particle reinforced aluminium based composites

Lorella Ceschini1,a, Alessandro Morri1,b and Fabio Rotundo2,c

1 Department of Metals Science, Electrochemistry and Chemical Techniques (SMETEC), University of Bologna, V. le Risorgimento 4, I-40136 Bologna, Italy

2DIEM, University of Bologna, V.le Risorgimento 2, I-40136 Bologna, Italy

[email protected], [email protected], [email protected]

Keywords: Metal matrix composites (MMCs), friction stir welding, linear friction welding, aluminium, particles, AA6061, AA2124, alumina, silicon carbide

Abstract. The widespread use of metal matrix composites (MMC) is often limited due to the

difficulties related to their joining by means of traditional fusion welding processes. The aim of this

work was to evaluate the effect on microstructure and mechanical properties (hardness and tensile

strength) of two different friction welding techniques used for joining two Al-based metal matrix

composites. In particular, Friction Stir Welding was applied to a 6061 (Al-Mg-Si) alloy matrix,

reinforced with 20vol.% of Al2O3 particles (W6A20A), while Linear Friction Welding was applied

to a 2124 (Al-Cu-Mg) alloy matrix reinforced with 25vol.% of SiC particles (AMC225xe). Both the

welding processes permitted to obtain substantially defect-free joints, whose microstructures was

found to be dependent on both the initial microstructure of the composites and the welding

processes. Hardness decrease was in the order of 40% for the FSW joint and of 10% for the LFW

joint, mainly due to overaging of the matrix induced by the frictional heating, while the joint

efficiency in respect to the ultimate tensile strength was 72% and 82%, respectively. Elongation to

failure increased in the FSW joint due to coarsening of precipitates, whereas it decreased in the

LFW joints due to the fibrosity in the thermomechanically altered zone. Fracture surface analysis

showed good matrix/reinforcement interface for both composites.

Introduction

The synergic interaction between the ductile metal matrix and the hard reinforcement bestow metal

matrix composites (MMCs) with enhanced specific stiffness and strength, better wear resistance and

greater thermal stability with respect to the corresponding monolithic alloys [1-3]. Among the

MMCs, Al matrix composites reinforced with ceramic particles (such as Al2O3 or SiC) offer many

advantages, such as: relatively simple production routes, suitability to undergo conventional

secondary processes (forging, extrusion, welding etc.), as well as isotropic behaviour [3–6]. Over

the past few years, significant efforts have been devoted to the development of these MMCs, which

have hitherto found applications in the aerospace, automotive and motor sport fields. One of the

main limitation to the widespread use of this class of material is related with their joining, as

traditional fusion welding techniques, such as TIG, MIG or laser, generally lead to microstructural

defects, thus badly affecting their mechanical properties. Besides the typical defects of Al alloy

fusion welds, such as solidification shrinkages, oxide inclusions and gas pores [7], the addition of

the ceramic reinforcement causes higher viscosity in MMCs melts, particle segregation, evolution

of the occluded gas and undesired matrix-reinforcement reactions [7,8]. Most of these problems can

be overcome by means of solid state welding processes that, avoiding melting and re-solidification

processes, significantly limit the defects formation in the joints. In friction welding the heat

necessary to activate the joining process is generated by friction, due to the rubbing motion of

abutting surfaces; the subsequent softening of the metal components favours the formation of the

joint. In Friction Stir Welding (FSW) a rotating tool, cylindrical in shape with a pin of smaller

diameter extending from the tool shoulder, is roto-translated along the joint line, generating

frictional heating and severe plastic deformation of the material, due to the stirring effect of the pin

[9]. Several studies showed that this process, initially developed for welding aluminium alloys,


could be also successfully applied to particles reinforced Al based composites [10-14]. One of the

main limitation of this technique, when applied to particle reinforced MMCs, can be the sever wear

of the tool, due to the abrasive action of the ceramic reinforcement [10]. This phenomenon could be

limited by means of recently developed highly-wear resistant FSW tools [15], or avoided by the use

of other friction welding techniques, such as Linear Friction Welding (LFW). In this process,

bonding of two flat-edged components is achieved through frictional heating, caused by their

relative reciprocating motion under an axial compressive force, without using any consumable tool

[15]. LFW was successfully applied to Ti alloys [16-18], steels [19] and Ni-based superalloys [20],

and only recently its application to Al-based MMCs was evaluated [21-22].

The aim of the present study is to discuss the effect of the FSW and LFW processes on two Al-

based particle reinforced MMCs, in terms of microstructural modification and mechanical

properties.

Experimental Methods

Materials. The experiment described in this work was carried out on two different particle

reinforced aluminium based composites. The first was produced by Duralcan (USA) using a

stir casting process and consists of a 6061 (Al-Mg-Si) alloy matrix, reinforced with 20vol.% of

Al2O3 particles (W6A20A). The cast was extruded at 480 °C to a rectangular plate (cross section of

100x7 mm2) and then heat treated to the T6 condition (solubilization at 540 °C for 1 h, water

quenching and aging at 145 °C for 16 h). The other composite was produced at Aerospace Metal

Composite (UK) by a powder metallurgy processing route and consists of a 2124 (Al-Cu-Mg)

alloy matrix reinforced with 25vol.% of SiC particles (AMC225xe). The billets were forged into 15

mm thick plates, then heat treated at the T4 condition (solubilization at 505 °C, water quenching

and aging at room temperature). Chemical compositions of the two matrix Al alloys are reported in

Table I.

Si Fe Cu Mn Mg Zn Ti Cr Al

AA6061 0.65 0.15 0.18 0.10 0.97 0.009 0.02 0.19 Bal. AA2124 ≤0.20 ≤0.300 3.80-4.90 0.30-0.90 1.20-1.80 ≤0.25 ≤0.15 ≤0.100 Bal.

Table 1. Chemical composition of the matrix Al alloys.

Friction welding. The W6A20A composite was Friction Stir Welded at the GKSS Research

Institute (Geesthacht, Germany), using a Neos Triceps 805, CN-controlled, five-axis robot. The

FSW tool, with a 18 mm diameter shoulder and a pin with a diameter of 8 mm, a length of 6.8 mm

and a left hand screw with a 1.25 pitch, was made of high wear-resistant material, heat treated to 63

HRC. The joint parameters were: vertical force 12 kN, rotation speed 600 rpm, welding speed 300

mm/min.

The AMC225xe composite was Linear Friction Welded at TWI (The Welding Institute, Cambridge,

UK), using welding parameters optimized after preliminary tests: friction/forge force 100 kN,

pressure 185 MPa, frequency 50 Hz, amplitude ±2 mm, burn off (axial shortening at which the

oscillating motion is stopped) 2 mm.

No post-weld heat treatment was carried out on both FSW and LFW MMCs joints. Schematics of

the FSW and LFW processes are reported in Fig.1.

Fig. 1. Schematic of (a) Friction Stir Welding and (b) Linear Friction Welding.

a b


Microstructural and mechanical characterisation. Samples for microstructural investigations

were cut in a transverse direction with respect to the welding line. The microstructural

characterisation, involving metallographic analyses through optical microscopy (OM) and scanning

electron microscopy (SEM) with an energy dispersive spectrometer (EDS), was aimed at

investigating the effect of welding processes both on the Al alloy matrix and the reinforcement

particles. Image analyses both on OM and SEM micrographs were performed using the software

Image Pro-Plus. Vickers microhardness measurements, with a 20 g load (HV0.02), were made across

the welded joints of the W6A20A FSW composite, to evaluate hardening or softening effect

induced by the FSW on the matrix alloy. In the case of the AMC225xe, Vickers hardness

measurements were taken across the joints, with a 30 kg load (HV30), on a central line parallel to the

y axis (Fig. 1-b), in the transverse cross-sections used for the microstructural analyses. No

interparticle microhardness were taken on these joints due to the very small size and uniform

distribution of the reinforcing particles that didn’t allow for this measurement.

The tensile tests on the W6A20A composite were carried out using flat specimens with a gauge

length of 25 mm, a gauge width of 6 mm and a thickness of 3 mm for the base material. For the

FSW joints, tensile specimens with a gauge length of 50 mm, a gauge width of 12 mm and a

thickness of 7 mm were machined perpendicularly to the FSW line. In the case of the LFW joints,

flat specimens were machined with the tensile axis perpendicular to the welded plane x-z (Fig.1-b),

having a gauge length of 20 mm, a gauge width of 14 mm and a thickness of 4 mm. The tensile data

for the base material was taken from the literature [23].

Three specimens for each joint were tested at a strain rate of 10-4

s-1

on a servo-hydraulic machine.

SEM analyses of the fracture surfaces were carried out, in order to investigate the influence of the

welding process on the mechanisms of failure.


Microstructural characterization. A low magnification optical micrograph of the cross section of

the FSW W6A20A joint is reported in Fig. 2, with presence of semicircular features, the so-called

onion rings, similar to those induced by a conventional milling process. FSW induced a surface

roughness increase from Ra=0.7 µm to Ra=3.5 µm on the side in contact with the shoulder, while

the other side was substantially unaffected.

Fig. 2. Optical micrograph of the W6A20A FSW joint cross section [14].

No evidence of typical fusion welding defects, such as gas porosity or particle segregation, were

found. Micrographs at higher magnification (Fig. 3) showed how the stirring action of the abrasive

tool modified both particle size and distribution in the weld nugget and Thermo-Mechanically

Affected Zone (TMAZ).

Weld nugget TMAZ

HAZ


Fig. 3. Optical micrographs of the W6A20A FSW joint: (a) base material; (b) transition from

TMAZ to weld nugget.

The results of the image analyses performed on the optical micrographs (Fig.4) showed a significant

reduction both of the particle average area (about 50%) and particle shape factor (from 2.1 to 1.9),

due to the abrasive action of the FSW tool. Moreover, the stirring effect induced by the pin,

favoured by the frictional heating, led to a better particle distribution mainly in the central part of

the weld (nugget), with respect to the base material (Figs. 3-4). Al matrix grain size was also

affected by the FSW process: the concomitant effect of the frictional heating and sever plastic

deformation reduced the grain size from about 30 µm in the base material to 20 µm in the nugget

(Fig. 4). This effect could be attributed to the dynamic recrystallization induced by the process,

favoured by the presence of the reinforcing particles, which act as preferred nucleation sites.

Fig. 4. Reinforcement particle and grain size of the W6A20A composite: base material (BM) and

FSW joint.

The low magnification optical microscope, under polarized light, of the cross section of the Linear

Friction Welded joint (Fig. 5), evidenced a relevant plastic flow induced by the process and a

significant fibrosity induced in the material, which was partly expelled as flash during welding. A

complete weld penetration was observed and, as already observed for the FSW joints, the

microstructural analysis didn’t identify any defects, such as particle segregation, gas pores or

undesired matrix-particle interfacial interaction.

Weld nugget Base material

TMAZ

100 µm 100 µm


Fig. 5. Optical micrograph of the AMC225xe FSW joint cross section.

Three characteristic zones were detected in the LFW joints (Fig.5): Weld Centre, with a uniform

particle distribution and, possibly, grain refinement (Fig. 6-a), due to the concurrent effect of

frictional heating and severe plastic deformation, as in the nugget of FSW joints [11]; Thermo-

Mechanically Affected Zone (TMAZ), characterized by a strong fibrosity of the matrix (Fig.5) and

small particle-free zones elongated in the flash extrusion direction (Fig. 6-b); Heat-Affected Zone

(HAZ), without any plastic deformation, but possibly affected by overaging. Both in the base

material and the HAZ the presence of large particle-free bands were found, as a consequence of the

forging process prior to welding (Fig. 6-c).

Fig. 6. Particle distribution in: (a) weld centre, (b) TMAZ and (c) HAZ.

As highlighted in the SEM micrographs in Fig. 7 and in the statistical distribution of the particle

size for the base and LFW composite (Fig. 8), no evidence of particle refinement or cracking was

found in the weld centre. Particle shape factor remained almost constant, being 1.91 for the base

material and 1.97 in the weld centre. This result can be related both to the absence of the abrasive

stirring tool, in contrast to FSW, and the smaller particle size characteristic of the base AMC225Xe

composite (about 3 µm), with respect to that of the W6A20A (about 60 µm).

Fig. 7. SEM micrograph of the AMC225xe LFW joint: (a) base material, (b) weld centre.

HAZ

TMAZ

Weld centre

HAZ

TMAZ

2.5 mm

a b c

25 µm 25 µm 25 µm

a b


Fig. 8. Particle size distribution for the base and welded zones of the AMC225xe LFW joints [21].

Hardness. In order to investigate the effects induced by the FSW on the Al alloys matrix,

microhardness profiles (HV0.02) were taken on cross-sections of the W6A20A welded plates. The

hardness (Fig. 9) decreased from the base material (80 HV0.02 ) to the centreline of the FSW zone

(50 HV0.02), with a reduction of about 37%, despite the matrix grain refinement experienced in this

zone. This softening of the Al alloy matrix was already reported in the literature [5] and was

associated to dissolution and growth of the strengthening precipitates induced by the frictional

heating during the FSW process [24].

Fig. 9. Microhardness profiles on the cross sections of the W6A20A FSW composite.

For the LFW joints, the hardness decrease in the welded zone was approximately 10% in respect to

the base material (Fig. 10); the complex microstructural modifications induced some fluctuations in

the hardness values, as for the FSW joints, due to the concurrent effect of severe plastic deformation

and frictional heating that favour recrystallization of the Al alloy matrix, as well as coarsening of

the intermetallic phases (Cu–Mg co-clusters) [10]. The lower hardness decrease of the AMC225xe

LFW joint in respect to the W6A20A FSW one, can be related to the different initial heat treatment

condition of the composites, respectively T4 and T6.

MMC FSW MMC Base


Fig. 10. Hardness profile on the cross sections of the AMC225xe LFW composite.

Tensile tests. The results of the tensile tests on the base and friction welded composites are reported

in Table 2. While the FSW specimens were tested in the as-welded condition, then with the typical

roughness induced by the process, the LFW specimens were surface finished to remove the flash.

The FSW W6A20A composite showed a joint efficiency (JE), defined as the ratio between the joint

and the nominal base material properties, of about 70% with respect to the tensile strength

(UTSFSW/UTSBM) and about 60% with respect to the proof strength. On the contrary, the elongation

to failure increased by about 60%. This behaviour is probably related to the previously mentioned

overaging effect induced by the frictional heating on the T6 heat treated matrix alloy, which cause

the transformation of the reinforcing precipitates from coherent or semi-coherent, to incoherent. As

a consequence, a change from the Ashby to the Orowan mechanisms of plastic deformation occurs

[25], leading to a reduction of the proof strength and an increase of the elongation to failure.

The tensile tests for the LFW joints showed a higher JE. In particular, the JE with respect to proof

strength (Rp02) was equal to 78% while, with respect to the ultimate tensile strength (UTS), JE was

equal to 83%. On the contrary, the elongation to failure significantly decreased in the LFW

specimens (about 40%).

E Rp02 UTS Elongation

[GPa] [MPa] [MPa] %

W6A20A-T6 96 340 364 1.7

FSW-W6A20A 90 193 262 2.8 Joint efficiency % 57 72 168

AMC225xe – T4 [23] 115 464 659 4

LFW- AMC225xe 105 364 542 2.4

Joint efficiency % - 78 82 60

Table 2. Tensile test results for W6A20A FSW, AMC225xe LFW and respective base materials.

Fracture surfaces. The fracture of particle reinforced aluminium-based composites occurs by three

main failure mechanisms: (i) nucleation, growth and coalescence of voids in the matrix, (ii)

interfacial decohesion at the particle–matrix interface, (iii) cracking of large reinforcing particles

[10–13].

In the W6A20A FSW tensile specimens, the fracture occurred at the interface between the FSW

zone and the base material. Both for the welded and base material, the fracture surfaces showed a

bimodal distribution of large voids, due to the presence of the particle reinforcement, and small

dimples, caused by ductile failure of the Al matrix. In the base material, decohesion at matrix-

particle interfaces and fracture of the larger particles were often observed (Fig.11-a). In the FSW

joints (Fig. 11-b) particle cracking was less evident, due to the particle refinement induced by the

process, while a higher volume fraction of minute dimples, due to the Al matrix grain refinement,

was observed.


Fig. 11. Fracture surface of (a) W6A20A base material and (b) FSW tensile specimen.

In the LFW joints fracture usually occurred in the TMAZ (Fig. 12), propagating in the same

direction as the fibrosity in this zone, where the hardness also reaches its minimum value. The

presence of this fibrosity could be therefore the cause of the relevant decrease in ductility of the

LFW joints.

SEM analyses of the fracture surfaces evidenced a good adhesion between the Al matrix and the

ceramic reinforcement, with failure occurring along the TMAZ (Fig. 12-a). The deformed matrix

showed the presence of tear ridges and small dimple, whose fine dimension could be related to the

large volume fraction of the reinforcement phase. Despite this large amount of ceramic particles,

only a small amount of cracked particles was observed and no particles clusters were found, due to

the small particle size. The mechanism of failure demonstrated the excellent matrix/reinforcement

interface, with no decohesion induced by undesired interfacial reaction products such as Al4C3 (Fig.

12-b). In fact, the pull-out of the reinforcement usually initiated in the matrix, adjacent to the SiC

particles, and propagated in the surrounding matrix [21].

Fig. 12. (a) Secondary cracks in the tensile fracture cross section; (b) Fracture surface of the

AMC225xe LFW joint.

Conclusions

Friction Stir Welding and Linear Friction Welding techniques were used to joint two different Al-

based MMCs, respectively a T6 heat treated 6061Al/20vol.%Al2O3P and a T4 heat treated

2124Al/25vol.%SiCP composite.

Complete weld penetration was found in both joint types and none of the typical fusion welding

defects were identified by means of the microstructural characterization. Excellent particle

distribution was found in the weld centre for both composites. Particle size distribution was not

affected by the LFW process, due both to the small size of the SiC particles and the absence of the

stirring tool. On the contrary, FSW led to a significant refinement of both reinforcement particles

and matrix grain size, due to the stirring effect exerted by the tool.

The harness decrease, in respect to the base materials, was 40% for the FSW joint and only 10% for

the LFW joint. This difference was attributed to the overaging of the Al alloy matrix, induced by

a b

a b


the frictional heating, which was of higher entity for the T6 heat treated composite. Overaging also

induced a decrease of the UTS, of 28% and 18%, respectively for the FSW and LFW joints, with

respect to the base composites. The superior reduction in the elongation to failure of the LFW joints

was related to the strong fibrosity, perpendicular to the applied load, induced by the process.

Fracture surface analyses showed good matrix/reinforcement adhesion for both welded composites.

References

[1] T.W. Clyne, P.J. Withers: An introduction to metal matrix composites (Cambridge University

Press, UK, 1993).

[2] M. Taya, R.J. Arsenault: Metal matrix composites – Thermomechanical Behaviour (Pergamon

Press, NY 1989).

[3] D.J. Lloyd: Int. Mater. Rev. Vol. 39(1) (1994), p. 1

[4] J.M. Torralba, C.E. Da Costa, F. Velasco: J. Mater. Proc. Tech. Vol. 133(1-2) (2003), p. 203

[5] K.N. Subramanian, T.R. Bieler, J.P. Lucas: Key Eng. Mater. Vol. 104-107 (1995), p. 175

[6] C.H.J. Davies: Key Eng. Mater. Vol. 104-107 (1995), p. 447

[7] M.B.D. Ellis: Int. Mater. Review Vol. 41(2) (1996), p. 41

[8] A. Urena, M.D. Escalera, L. Gil: Compos. Sci. Technol. Vol. 60 (2000), p. 613

[9] R.S. Mishra, Z.Y. Ma: Mater. Sci. Eng. R Vol. 50 (2005), p. 1

[10] G.J. Fernandez , L.E. Murr: Mater. Charact., Vol. 52 (2004), p. 65

[11] I. Boromei, L. Ceschini, A. Morri, G. Garagnani: Metall. Sci. Technol. Vol. 24 (2006), p. 12

[12] G. Minak, L. Ceschini, I. Boromei, M. Ponte: Int. J. Fatigue. Vol. 32 (2010), p. 218

[13] L. Ceschini, I. Boromei, G. Minak, A. Morri, F. Tarterini: Comp. Sci. Technol. Vol. 67

(2007), p. 605

[14] L. Ceschini, I. Boromei, G. Minak, A. Morri, F. Tarterini: Composites: Part A, Vol. 38

(2007), p. 1200

[15] H.J. Liu, J.C. Feng, H. Fujii, K. Nogi: Int. J. of Machine Tools & Manuf. Vol. 45 (2005), p.

1635

[16] A. Vairis, M. Frost: Wear, Vol. 217(1) (1998), p. 117

[17] A. Vairis, M. Frost: Mater. Sci. Eng. Vol. A271 (1999), p. 477

[18] A. Vairis, M. Frost: Mater. Sci. Eng. Vol. A292 (2000), p. 8

[19] W.Y. Li, T.J. Ma, S.Q. Yang, Q.Z. Xu, Y. Zhang, J.L. Li, H. L. Liao: Mater. Lett. Vol. 62(2)

(2008), p. 293

[20] M. Karadge, M. Preuss, P.J. Withers, S. Bray: Mater. Sci. Eng. A, 491(1-2) (2008), p. 446

[21] F. Rotundo, L. Ceschini, A. Morri, T.-S. Jun, A.M. Korsunsky: Composites: Part A Vol. 41

(2010), p. 1028

[22] T. S. Jun, F. Rotundo, X. Song, L. Ceschini, A.M. Korsunsky: Mater. Design Vol. 31 (2009),

p. S117

[23] Information on http://www.amc-mmc.co.uk.

[24] L.M. Marzoli, A.v. Strombeck, J.F. Dos Santos, C. Gambaro, L.M. Volpone: Comp. Sci. and

Technol. Vol. 66 (2006), p. 363

[25] G.E. Dieter, Mechanical metallurgy (Mc Graw-Hill, UK, 1988).


Hot drilling of aluminium matrix composite

Riccardo Donninia, Loredana Santob, Vincenzo Tagliaferric

Department of Mechanical Engineering, University of Rome “Tor Vergata”, Via del Politecnico 1, 00133 Rome, Italy


Keywords: Metal Matrix Composites (MMC), Al2009 and Al6061 Aluminium Alloy, SiC and Al2O3 reinforcement, Hot drilling.

Abstract. The aim of this paper is to investigate the behaviour in terms of drilling forces and

roughness of Metal Matrix Composites (MMC) in hot drilling machining. In particular,

Al2009/(SiC)w, Al6061/(SiC)w, and Al6061(Al2O3)p metal matrix composites were used, and the

adopted temperature were in the range 20°C-160°C. A comparison with drilling at room

temperature has been discussed.

The results have shown the sensible influence of the working temperature on drilling forces and on

surface material properties. In the case of Al2009/(SiC)w a minimum in the drilling forces has been

found, making possible the dry machining and improving the cutting conditions. Instead, for

Al6061/(SiC)W and Al6061(Al2O3)p in the used temperature range no minimum appears.

Introduction

Drilling is one of the most common industrial machining processes [1-2] and some researches have

been carried out to try the best work conditions to reduce the cutting forces, so to enlarge life tool.

For this issue the use of refrigerant lubrication fluid permits to limit the drill heating, however

posing the problem about the environmental impact that regards its use [3]. Therefore, the trend is

to reduce more and more this use and to work in dry condition [4-5], but this situation requires

cutting forces lower and lower in order to reduce tool wear, or more generally excessive stress that

can damage it. A new idea, proposed in Santo et al. [6] regards the metal heating during drilling

operation by means of an infrared lamp. By this way a reduction in thrust force and torque is

expected, having temporarily changed the material properties. In fact in these conditions, during

drilling operation, the local heat disposal is more difficult, and in the hole point it is possible to

reach more easily the conditions of material “softening”. The reduction of lubricants is also

expected, making the issue very important for ecological and economic reasons.

In [6] the study has been mainly developed on the non-ferrous metals such as aluminium alloys.

The results, obtained in the case of Al6082 aluminium alloys, not only confirmed the force

proportionality with the feed az [mm/rev] = vfeed/n, but also a minimum temperature has been

identified. In order to predict the cutting force in hot drilling machining and to evaluate the effects

of the process parameters on thrust force a neural network has been also implemented obtaining a

good agreement between experimental and numerical result [7].

In this work the aforesaid experimental procedure has been extended to aluminium matrix

composites. They are advanced materials which combine a tough metallic matrix with a hard

reinforcement to obtain composite materials with properties superior to conventional metallic alloy.

Due to such attractive properties MMCs are very interesting for application in several aerospace and

automotive applications. However, drilling of these materials is complex because of the abrasive

characteristics of the reinforced particulates [8-10].

In the paper experimental tests were carried out in order to record the drilling forces in different

process conditions, about different typologies both of matrix and of reinforcement for the

composite. Moreover, roughness measurements were performed to evaluate change in surface

material properties.


Material and experimental methods

The materials used for the experimentation are MMC, aluminium metal matrix and reinforcement in

the form of SiC whiskers (SiC)w or alumina particles reinforcement (Al2O3)p . In particular, about

the whiskers reinforcement, the aluminium alloys Al2009 and Al6061 were used (volume fraction

25%). About the Al2O3 reinforcement (volume fraction 10%) only the Al6061 alloy was used, as

reported in table 1.

MMC Matrix Reinforcement Volume fraction

Al2009/(SiC)W Al2009 SiC whiskers 25%

Al6061/(SiC)W Al6061 SiC whiskers 25%

Al6061/(Al2O3)P Al6061 Al2O3 particles 10%

Table 1. Materials

In the case of Al/ SiCw composite, the samples have been cut by extruded bars, having a length of

about 500 mm. These bars were cut along the perpendicular direction to the extrusion axis, in order

to obtain samples, 40 mm x 75 mm in size. In the case of Al6061/Al2O3 composite, the samples

were obtained by cutting a circular extrusion bar, having a radius of 40 mm. For all the samples the

thickness was of 10 mm.

Experimental tests were performed in dry drilling condition, using a conventional milling machine

and HSS-Co 8% (DIN338) twist drills of 5 mm in diameter, and at least five valid tests were carried

out for each experimental condition. The adopted experimental system is shown in Fig.1.

The feed values as combination of the cutting parameters indicated in Table 2 were used, varying

the temperature in the range 25-140°C step 20°C, by means of a 2 kW power IR lamp for material

heating (Fig.1). The temperature was measured on irradiated surface by an optical pyrometer (Irtec

P500). Each hole was obtained immediately after heating, turning off the lamp. No significant

gradient of temperature was measured along the thickness.

Fig.1. Assembled system on the milling machine for the drilling operations


Table 2. Cutting parameters

During the drilling tests, the torque and the thrust force were recorded by a Kistler 9273

dynamometer coupled with a charge amplifier (PCB 443B102). The process parameters were

chosen considering the results found in [6-7].

The signals were sent to a digital data acquisition system for storage and subsequently, a computer

was used for data handling and analysis.

The surfaces of the holes were analysed by a 3D Surface Profilometer (Taylor Hobson, Talysurf

CLI 2000). In particular, Ra (the arithmetic average of the deviation for the roughness profile)

values have been calculated.

Finally, to characterize material properties, Flat-top cylinder Indenter for Mechanical

Characterization (FIMEC) tests were carried out varying the temperature. Such test provides the

local value of yield tensile stress by means of the curve load-penetration depth recorded during a

compression test, utilising a flat punch [11].The indenter was 1.5 mm in diameter, the speed 0.01

mm/s.

Results and discussion

Al2009/SiCw. At first to evaluate the influence of process parameters on the thrust force and torque,

drilling tests at room temperature have been considered. In particular, Fig.2 shows the drilling

forces as a function of feed az (where z axis coincides with the drilling rotation axis) where az [mm /

rev] = vfeed/n, which is a fundamental parameter for a careful analysis based on the process

productivity.

a) b)

Fig. 2. Al2009-SiCw: Fz and Mz as function of az for drilling at room temperature

It can be seen that, in the case of Mz, theories relating to conventional materials are confirmed about

the proportionality in az. In this case, working at high n values, minor stresses are involved for the

az [mm/rev] vz [mm/min] n [rev/min]

0.014 77.5 15000

0.0155 77.5 5000

0.042 210 5000


cutting edges and thus higher quality and less wear are expected. In addition, Fig.2a shows Fz as a

function of az, and no clear trend has been found, probably due to the high scattering of the results.

One of the aims of the present work regarded the drilling operations at “high speed” as rapid

process implementation, so the focus has been about the influence of the workpiece heating and

about the two highest values for az (0.015 and 0.042 mm/rev). The results are shown in Fig.3 (a, b).

a) b)

Fig. 3. Al2009-SiCw: Fz and Mz vs. T, for 5000 rev/min

Table 3. Force thrust and torque reduction % in hot drilling of Al2009-SiCw

A minimum temperature for all the process conditions, both for Fz and Mz (Fig.3-a,b), is evident.

Particularly, Tab.3 shows the maximum percentage reductions of the forces for hot drilling,

calculated making a comparison between the thrust force and the torque values at room temperature

and in correspondence of the minimum force values (temperature 80° C).

The reduction relative to Mz, with a minimum of 80°C, appears especially significant. This situation

is positive in particular about the study of tool wear as Mz acts more than anything else directly on

the cutting edges. Regarding Fz, the reductions (with a Tmin about 100 ° C) is smaller, reflecting

however the real effectiveness of the hot drilling concepts. In order to correlate hot drilling with the hole surface morphology, Fig. 4 and 5 show two roughness

profiles and 3D surface maps for drilling at room temperature and at 80 °C respectively. By

analyzing the results, a value of Ra=0.53 µm has been found at room temperature, while at 80°C

Ra=0.35 µm. A more uniform surface (except for a central not uniformity probably due to

accidental sliding of chip) than that at room temperature has been found (Fig.5 a, b).

In the obtained results, reduction in thrust force and torque was found, having temporarily changed

the material properties by heating. This last occurrence is evident in Fig. 6 where the results of the

FIMEC test are shown. In particular, a decrease in the yield stress has been found by increasing the

temperature.

In drilling operation there is a minimum value for drilling forces, while by analysing the FIMEC

results a decrease of the yield stress up to 140°C can be seen. Probably, in hot drilling this

rid.% Fz rid.% Mz

210 mm/min 38.3% 52.8%

77.5 mm/min 25.3% 45.8%


phenomenon is due to the built-up edge generation, evident at the high temperature. Over 80°C the

mechanism of chip formation changes, affecting significantly the drilling forces.

Fig. 4. Al2009-SiCw: roughness profiles for drilling at room temperature (a) and at 80°C (b),

az=0.04 mm/rev

a) b)

Fig. 5. Al2009-SiCw: roughness 3D surface maps for drilling at room temperature (a) and at 80°C

(b), az=0.04 mm/rev

Fig. 6. FIMEC test: Yield Stress vs. temperature for the Al2009-SiCw

a)

b)


Al6061/SiCw. In the case of Al6061/SiCw, at room temperature, it is confirmed the proportionality

of torque compared to feed; instead the trend of Fz seems to be the opposite of that of Al2009/SiCw.

The reason is related to the properties material, which involves, in addition to minor drilling forces,

different cutting conditions.

a) b)

Fig.7. Al6061-SiCw: Fz and Mz as function of az for drilling at room temperature

The phenomenon of chip removal becomes more important than the usefulness obtained by the

temperature increasing. Moreover, this increase can lead to an easier achievement of the conditions

for built-up edge formation, as above mentioned, and in this case to a highest scattering for the

values at higher temperatures. This was evident by the observation of the drill surfaces and tools.

Fig.8. Al6061-SiCw: Fz and Mz vs.T, for 5000 rev/min

By analyzing the error bars on each obtained values (Fig.8), if the result is good for the Tmin and

Tamb values, more uncertainty is found especially at “contour” temperatures to that of minimum,

because evidently there is a certain instability in the cutting process in those conditions. In addition,

there is a greater uncertainty for Mz, where the acquired vibration signals are usually higher than in

Fz. Tab. 4 reports the maximum force thrust and torque reduction percentage in hot drilling of

Al6061/SiCw.


Table 4 . Drilling forces reduction % from r.t. to the obtained minimum at the specific temperature

80°C, for Al6061-SiCw

a) b)

Fig. 9. Al6061-SiCw: roughness profiles for drilling (a) roughness 3D surface map (b) for drilling

at room temperature, az=0.04 mm/rev

Fig. 10. 3D surface maps for the drilling considering to a) 60°C (Ra = 1.15 µm),

b) 80°C (Ra = 0. 95 µm),

c) 100°C (Ra = 0.8 µm ) and d) 120°C (Ra = 0.65 µm), at az = 0.04 mm/rev

Fig.9 and Fig.10 show roughness profile, 3D surface maps and the Ra values for different

temperatures. Therefore, in the hot drilling of Al6061/SiCw, the heating causes no noticeable effect

on the drilling forces. The surface quality of the holes is worse, especially considering the results

obtained for the room temperature drilling (Ra =0.61µm).

rid.% Fz rid.% M+z

210 mm/min - 5.34%

77.5 mm/min - 23.92%

d) c)

a) b)


Fig. 11. FIMEC test: Yield Stress vs. Temperature for the Al6061-SiCw

Fig.11 shows the FIMEC tests results for Al6061/SiCw. A decrease with the increasing of the

temperature was found, also in this case.

Differently from the case of Al2009/SiCw, no evident minimum was found in drilling forces for

each temperature. Probably, the effect of heating strongly influences the mechanisms of chip

formation, inhibiting the effect of Fig.11.

This can be concluded observing the roughness results in Fig.9 e Fig.10: the worse surface quality

at high temperature mainly depends on the joined chip.

Al6061/Al2O3. Finally, to verify also the influence of other reinforcements on the drilling

operations of composite, experimental tests have been performed on Al6061/Al203. About drilling

forces, the obtained results have been compared with the Al6061/SiCw results. The comparison is

shown in Fig.12. In this case the drilling operation were carried out at vfeed=77. 5 mm/min and n

=500 rev/min.

Fig.12. Fz and Mz vs. Temeperature for Al6061-SiCw and Al6061-Al2O3p, az=0.0155 mm/rev

It is necessary to recall that the reinforcement percentage is lower in the actual composite (10%)

and the nature is different. From Fig.12 a different behaviour in terms of thrust force was found: the

higher values for the Al6061/SiCw depends on the different reinforcement, as above mentioned.


In the case of the torque the trend is the same for low temperature, while starting from 100°C the

decreasing as function of temperature is more evident for Al6061/Al203.

Fig. 13. FIMEC test: Yield Stress vs.Temperature for the Al6061-Al2O3

Fig.13 shows the results of FIMEC test for Al6061/Al203, the yield stress is practically constant for

each temperature, showing that the material results unchanged in the adopted range of temperature.

This is in agreement with the drilling force trend as a function of temperature, while is different

from the case of Al6061/SiCw, where, even if the material properties change by increasing the

temperature, no significant change was observed in the drilling forces values. The reinforcement so

plays a very important role.

Conclusions

Hot drilling of MMC Al2009/SiCw, Al6061/ SiCw and Al6061/ Al203 has been carried out and the

main conclusions are as follows.

The drilling forces strongly depend on process condition and the effect of temperature on workpiece

can be very significant. In particular, in the case of Al2009/SiCw a minimum in the drilling forces

has been found. The results of the FIMEC tests of this material show that also a change in the

material properties is found by increasing the temperature. This change influences the mechanisms

of chip formation and so the morphology of the hole surface.

A minimum was not observed in Al6061/SiCw, probably caused from the complexity of the cutting

mechanism for each considered temperature. Analysing the aforesaid material but with different

reinforcement (Al203), both FIMEC test and drilling operations provided constant values in terms of

yield stress and drilling forces versus temperature, respectively. These results confirm the

importance of the FIMEC test to give indication on material properties related to the possible

working conditions.

Moreover, the role of reinforcement in MMC drilling is decisive, in fact it strongly influences the

drilling forces based on their nature and volume fraction amount.

Further studies are necessary to deepen this topic, for instance the analysis of chip formation, tool

wear and a micro-structural analysis of the material around the hole.


References

[1] S. Kalpakjian and S.R.Schmid, in: Manufacturing Engineering and Technology, 2001, Prentice

Hall, Inc.

[2] M.C. Shaw, in: Metal Cutting Principles, Clarendon Press, Oxford, 1984.

[3] J.F. Kelly and M.G. Cotterell: J. Mater. Process. Tech., Vol.120 (2002), p.327

[4] M. Nouari, G. List, F. Girot and D. Coupard: Wear, Vol. 255 (2003), p.1359

[5]� M. Nouari, G. List and D. Géhin: Int. J. Mach. Tool Manu., Vol. 45, (2005) p. 1436

[6] L. Santo, F. Trovalusci and V. Tagliaferri, Proceedings of ASME ESDA 2006 Conf., Turin,

Italy, 4–7 July 2006, ref. 95486.

[7] R. Donnini, R. Montanari, L. Santo,V. Tagliaferri and N. Ucciardello: Int. J. Comput. Mater.

Sci. Surf. Eng., Vol. 3, (2/3) (2010), p.175.

[8] M. Ramulu, P.N. Rao and H.Kao: J. Mater. Process. Tech., Vol.124 (2002), p. 244

[9] G. Tosun and M. Muratoglu: Compos. Sci. Technol., Vol.64 (2004), p. 299

[10] S. Barnes, J.R. Pashby and B. Hashim: Appl. Compos. Mater., Vol.6 (1999), p. 121

[11] Β. Riccardi, P. Gondi, R. Montanari, L. Moreschi, A. Sili and S. Storai: 2001, Fusion Eng.

Des., Vol.58-59, p. 755.


Effect of mechanical mould vibration on solidification behaviour and microstructure of A360-SiCp metal-matrix composites

Giulio Timelli1,a, Emilia Della Corte2,b and Franco Bonollo1,c 1 University of Padova, Department of Management and Engineering, Stradella S. Nicola, 3 I-

36100 Vicenza, Italy 2 Enginsoft Spa, via Giambellino, 7 I-35129 Padova, Italy


Keywords: Metal matrix composites, solidification, vibration, numerical simulation, heat transfer coefficient

Abstract. In this work the microstructural evolution of an A360 alloy reinforced with 10vol.% SiC

particulate is described. During the material solidification, mechanical vibration, in the range of 0-

41 times the gravity acceleration, g, has been applied to a steel die. It has been observed that

vibrations can promote a quite homogeneous SiC dispersion on macroscopic scale. On the other

hand, by using too high vibrations’ intensity, segregation phenomena have been pointed out in the

castings. Furthermore, it has been evidenced that the reinforcement distribution is influenced by

mechanical entrapment of the particles at grain boundaries and in the interdendritic channel. The

metallographic analysis has emphasized a finer microstructure with increasing vibrations’ intensity.

By comparing simulated and experimental temperature curves of the mould in the different cases,

different HTC made the best fit. By increasing the vibrations’ intensity, the HTC increases in the

temperature range of solidification of the composite.

Introduction

The possibility of combining different properties in one material, extending the field of use, has

been achieved, and it is steadily improving, with metal matrix composites (MMC) [1]. The MMC

based on light alloys, particularly aluminium alloy matrix, are characterized by a high strength-

weight ratio, good wear resistance and thermal expansion if compared to standard alloys [2].

Important factors affecting the final cost of the material are the type of reinforcement and the

process technique. Actually, a consolidated approach is to use particles such as silicon carbide (SiC)

or alumina (Al2O3) for MMC applications targeted to automotive industry, where the cost factor is

crucial, and to use foundry processes that are able to produce near-net shape castings, reducing the

machining operations [3,4]. Therefore, the capability to know and to control the solidification

phenomena of MMC is essential for the optimization of the microstructure and mechanical

behaviour [5,6]. During the casting solidification, the ceramic particles interact with the advancing

solidification front. The balance between the repulsive forces, arising from the surface tension, and

the attractive forces, caused by the viscosity of liquid metal, determines the wettability of the

matrix-particulate system [1,7-10]. When the repulsive forces prevail over attractive ones, the

ceramic particulate is driven by the solidification front (pushing phenomenon). On the contrary,

when SiC or Al2O3 particles are homogeneously absorbed by the solidification front (engulfment

phenomenon), a continuous interface particle-matrix is generated with good adhesion and dispersion

of the particulate [3]. Further, when the ceramic particles are physically unable to be pushed

because of the converging solidification fronts and the wettability is poor, a mechanical entrapment

of the particulate takes place [3, 11]. Since the wettability of ceramic particles with liquid Al alloys

is generally poor, it is necessary to supply the necessary energy to the system for the formation of a

stable interface. Various procedures have been recommended to improve the wetting of ceramic

particles by molten metal, and include increasing metal liquid temperature [12], pre-treatment of

particles [9], coating or oxidizing the ceramic particles [13], and by adding some surface-active

elements such as Mg and Li into the matrix [14]. By considering the manufacturing processes of the


MMC, this energy can actually be provided by applying a “mechanical pressure” on the alloy

constituting the matrix [7, 8].

An alternative approach may be the application of vibration energy, already used in several

metallurgical and engineering processes [15]. The variations of the amplitude and frequency of

vibration have a strong influence on the microstructural characteristics of casting, showing however

several limits: high values of amplitude lead to an increase of internal defects [16], while high

frequency of vibration is limited by the equipment used. The application of controlled mechanical

vibration on Al-Si foundry alloys leads to a refinement of grain size [17-19], a variation in the

morphology and the size of eutectic silicon particles [18-20] and an increase in the final mechanical

characteristics of the casting [17, 20].

In this study, the effects of mechanical mould vibrations on the microstructure of A360-10%SiCp

metal-matrix composites and on the heat transfer from the casting to the die were investigated.

Experiments

Material and experimental procedure. The material used in the present work was an A360

composite foundry alloy reinforced with 10 vol.% SiC particulate. The MMC was supplied by

DURALCAN Co. in the form of 12 kg notched ingots, with commercial designation F3N.10S.,

which were produced by Compocasting technology within a low vacuum of approximately 1-5 torr

[21]. The chemical composition of the matrix is shown in Table 1. The SiC particles showed a mean

size of 7.7 ± 2.9 µm and an aspect ratio of 1.8 ± 0.7.

Si Fe Cu Mn Mg Zn Ni Ti Al

9.65 0.8 0.180 0.5 0.53 0.003 0.005 0.1 bal.

Table 1. Chemical composition of the A360 matrix alloy [wt.%]

The material was remelted in an electric furnace set up at 650 ± 5°C. Periodically, the molten metal

was stirred with a coated paddle. Further, the metal was allowed to sit, without stirring, for 1 h and

then manually skimmed. After skimming, the oven was set up at 730 ± 5°C and the bath

temperature increased up to 720 ± 5°C. The temperature is carefully controlled to avoid unwanted

reactions between the liquid metal and SiC particles [22]. Before pouring, the metal was carefully

stirred to prevent settling of ceramic particles at the bottom of the crucible. The molten alloy was

poured into a steel mould, with dimensions shown in Fig. 1a, preheated at 440 ± 15°C. The mould

is attached to an electromagnetic shaker which was connected to a power amplifier CREST AUDIO

10001 and to a signal generator HP33120A. A detailed description of the casting procedure and the

process parameters is given elsewhere [23]. Briefly, 3 to 5 castings were scrapped after the start up,

to reach a quasi-steady-state temperature in the die. The vibration parameters, such as the frequency

and the amplitude, were controlled by means of a high-speed oscilloscope and a triaxial

accelerometer. The experimental set up is shown in Fig. 2. Six K-type (Chromel-Alumel)

thermocouples with a diameter of 1 mm were inserted into the mould and the die cavity to monitor

the temperature evolution of the mould and the metal during the solidification. The thermocouples

were connected to a data acquisition system (National Instruments SCXI-1000) to record the

temperature at a sampling frequency of 20 Hz. The numbers and locations of the thermocouples are

also shown in Fig. 1. Throughout the present work, the thermocouples Th1, 2 and 5 were only

considered, since representative of the macro- and microstructural results. For each test, the shaker

was vibrated according to user-specified amplitude and frequency, inducing an horizontal vibration

on the mould, i.e. parallel to the plate of the shaker. In the present work, the intensity of mechanical

mould vibration is given in units of gravity. Eq. 1 relates acceleration to vibration amplitude and

vibration frequency [19]:

24.024g D f= × × (1)


where g is the acceleration in units of gravity (1g = 9.81 m/s2), D is the displacement or double

amplitude (m), and f is the frequency of vibrations (Hz). The microstructure and the solidification

behaviour of the MMC castings were studied in the range of 0-41g with suitable variations of

vibration amplitude and frequency. Three or more specimens were tested for each condition.

Fig. 1. Cross sectional view of the mould

showing the location of the thermocouples

used.

Fig. 2. Mould assembly mounted on the shaker and

controlled by and accelerometer.

Microstructural characterization. The microstructure examination was made 85 mm from the

bottom edge of the casting, i.e. close to the locations of Th1, 2 and 5. The samples cut from the

cross section of the casting were mechanically prepared to a 3-µm finish with diamond paste and,

finally, polished with a commercial fine silica slurry for metallographic investigations.

Microstructural analysis was carried out using an optical microscope and a scanning electron

microscope (SEM) equipped with an energy-dispersive spectrometer (EDS), and quantitatively

analyzed using an image analyzer. Various microstructural parameters were investigated and

measured, such as the secondary dendrite arm spacing (SDAS) of α-Al phase, and the volume

fraction and distribution of SiC particles. The local variations of SiC volume fraction were studied

over the cross section of castings. Each data point represents the volume fraction in an area 250 ×

188 µm2 and it was obtained from micrographs taken from casting edge to casting centre. At least

three profiles were measured and subsequently averaged. The secondary phases, such as the Mg2Si

and Fe-rich particles, were excluded from the measurements.

While some samples were chemically etched with 15vol.% HNO3, 10vol.% HCl, 5vol.% HF and

70vol.% H2O for macrostructure, other specimens were anodised in a solution of HBF4. The macro

and microstructural characteristics were correlated with the vibration parameters used.

Casting simulation. The MAGMASOFT®

v4.4 (2009) commercial software, with its modules for

gravity die casting and for optimization process (MAGMAfrontier®

), was used to calculate the heat

transfer coefficient, HTC. The numerical code employs the finite volume approach to convert

differential equations into algebraic ones and solve them on a rectangular grid. The CAD model of

the casting was imported in the simulation software where a controlled volume mesh of 6204 cast

cells was automatically generated by the software as shown in Fig. 3. Due to symmetry, only one

half of the model was simulated in the vertical direction. The initial conditions for numerical

simulation were defined to reproduce the real casting conditions. The pouring temperature was set

at 720°C, while, for the die, the temperature was set up according to the experimental

measurements. The materials used in the simulation for the die and the alloy were chosen among

those present in the software database. To define the whole set of boundary conditions in the model,

the process parameters and the cycle time, collected from the casting process, were imported in the

software, increasing the reliability of HTC calculation. Virtual thermocouples were inserted in three


different zones of the die in order to control the temperature profiles and to compare these values

with the real ones (Fig. 3). The simulation results were then correlated with the microstructures

observed at this location of the casting. As a first approach to determine the HTC value, a previous

reported experiment [24] was used as a first approximation. Simulated and experimental

temperatures were then compared and the HTC value was determined based on best fit as a function

of the temperature.

Fig. 3. Mesh generation for analysis with total control volumes 50544 and metal cells 6204. Three

virtual thermocouples were used for simulation and HTC calculation.


Heat transfer coefficient at the casting-die interface. Temperature profiles of thermocouples 1-6

are generated for the different cases of casting with and without vibration. A close examination of

the temperature profiles of the mould and a comparison with the central temperature readings inside

the die cavity show that the relative close proximity to the die cavity made also the reading of

thermocouple 1 indicative of the solidification evolution of the A360-10%SiC composite.

Analysing together Th1 and Th5, it is possible to observe how the first temperature rise for

thermocouple 1 is due to the latent heat of the α-Al dendritic network, while the second rise is

produced by the latent heat of the Al-Si eutectic structure as it transforms from liquid to solid. The

different peak height is explained by the different amount of latent heat released during

solidification phenomena, which is proportional to the fraction solid formed [25]. Under steady-

state conditions for A360 alloys, the total fraction of Al-Si eutectic is ~0.68 [26]. The rest of the

thermocouple readings represent the traditional transient decay of the temperature inside the steel

mould.

(a)

(b)

(c)

Fig. 4. Comparison between the experimental and simulated temperature curves in the A360-

10%SiC: (a) without vibrations, and with vibrations at (b) 7g and (c) 29g.


The thermocouple readings in the die show a relative increase in temperature as a result of applying

vibration, while the central thermocouples in the die cavity evidence that the cooling rate of A360-

10%SiC composite, estimated from the straight line portion of the cooling curve just before the start

of solidification, slightly increases with increasing the vibrations’ intensity. A slight increase in

cooling rate with the intensity of mould vibrations may be attributed to an increase in the forced

convection in the melt brought about the increased vibration levels. These findings are in agreement

with the results reported in References 19-21.

The temperature readings inside the mould cavity show evidence that vibration reduces the

longitudinal temperature gradient (∆T) in the mould compared with the un-vibrated case. While ∆T

is around 10°C in the castings solidified without vibrations, it progressively decreases at 7g (~8°C),

reaching a steady value (~3°C) for vibrations higher than 20g. This result evidences that the

temperature longitudinal gradient inside the casting is also more uniform under the influence of

vibration. This temperature uniformity will result in a more homogenous microstructure and

properties.

Fig. 5. Variations of HTC as a function of

temperature for specimens solidified without

vibrations, and with vibrations at 7 and 29g.

Fig. 6. SEM micrograph with EDS spectra where

A indicates to secondary α-Al15(Fe,Mn)3Si2

compounds and B to SiC particles.

By comparing simulated and experimental temperature curves of the mould in the different cases, as

shown in Fig. 4, different HTC made the best fit. In the present work, the casting simulation does

not consider filling of the die, but only the solidification stage. Fig. 5 shows the HTC values,

determined from an inverse modelling approach, of specimens solidified under different vibrations’

conditions. By increasing the vibrations’ intensity, it is observed how the HTC increases in the

temperature range of solidification of the A360-10%SiCp composites, i.e. 600-575°C [22]. The

simulated HTC that fits the experimental results in the un-vibrated conditions is included between

2700 and 1000 W/m2K as a function of temperature. At vibrations of 7g, the simulation results in a

higher HTC ranging between 8600 and 1000 W/m2K. With still higher vibrations, as at 29g, the

HTC increases and it is in the range of 10000 to 1000 W/m2K.

Microstructure and distribution of SiC particles. The matrix (A360) of the composite consists of

a primary phase, α-Al solid solution, and an eutectic mixture of aluminium and silicon. α-Al

precipitates from the liquid as the primary phase in the form of dendrites. Typical secondary

intermetallic compounds, such as β-Mg2Si, π-Al8Mg3FeSi6 e θ-Al2Cu, are identified as the main

intermetallics constituting the matrix alloy. Due to the high Fe and Mn content, secondary coarse α-

Al15(Fe,Mn)3Si2 particles with polyhedral morphology, as revealed by the EDS analysis, are

observed (Figure 6). It is worth mentioning that the presence of needle-like β-Al5FeSi phase is less

evident. These intermetallic compounds are located predominantly in the interdendritic region of

the microstructure.


The scale of microstructure was characterized by means of SDAS measurements and then

correlated with mechanical vibrations. The SDAS measurements revel a very fine microstructure,

with values in the range of 10-16 µm, with higher values in the un-vibrated specimens. A general

refining of microstructure occurs by increasing the vibrations’ intensity due to a slight increase in

cooling rates with the intensity of mould vibrations. The SDAS values are comparable with the size

of SiC particles. Therefore, the ceramic particles are unable to move due to converging

solidification fronts and the reinforcement distribution throughout the casting is controlled by a

mechanical entrapment mechanism [3, 11].

In un-vibrated samples, the microstructure is coarser in the central regions of the specimens, if

compared to the outer zones, with coarse secondary intermetallic phases. By increasing the

vibrations’ intensity, the intermetallic compounds show similar dimensions proceeding from areas

near the die walls towards inner regions of the castings. Fig. 7 compares the α-Al15(Fe,Mn)3Si2

particles in the central region of the specimens solidified under different vibration intensities.

(a)

(b)

Fig. 7. Microstructures of the A360-10%SiC composite solidified under (a) no-vibrations and (b)

vibrations at 41g. Arrows indicate the α-Al15(Fe,Mn)3Si2 intermetallic particles.

In un-vibrated specimens, the examination of the eutectic microstructure reveals the presence of

both coarse and acicular, and fine and fibrous eutectic Si particles. This can be due to a non-

homogeneous solidification rate throughout the casting. It is well established that rapid

solidification changes the eutectic Si shape so that it is similar to chemically modified eutectic Si

(quench modification) [27], but if solidification conditions are not similar in the whole casting, a

non-homogeneous microstructure can develop [28]. On the other hand, vibrated castings exhibit

shortening and a reduction in size of their eutectic Si particles, which also assume a flake-like

morphology, even if a few islands of unmodified eutectic cells are still present. The observed

shortening of the eutectic Si particles with increased vibration intensity can be attributed to their

fragmentation during the early stages of solidification. In contrast with the work of Kocatepe et. al.

[18], Deshpande et al. [19] and Abu-Dheir et al. [20], no thickening of eutectic Si flakes are

observed over certain vibration intensity. It is however observed that the changes of size and

morphology of eutectic Si depends strongly on the vibration parameters, i.e. amplitude and

frequency, and therefore can not be reduced to a single parameter like the gravity acceleration [18-

20, 29]. Further studies are therefore required. The nucleation of intermetallic phases and eutectic Si

particles on the SiC ceramics are observed in Fig. 6 and 7.

The dispersion of the ceramic particles is not homogeneous on a macroscopic scale in un-vibrated

specimens. The heat transfer trough the die wall and the poor wettability of the SiC particles are the

preferred conditions to push the reinforcement segregating in the center of the casting, the last zone

to solidify (Fig. 8a). By applying a mechanical mould vibration, the increase of convective forces

and movements in the melt reduces the temperature gradient within the casting. This produces a

more homogenous dispersion of SiC particles on a macroscopic scale (Fig. 8b). The improvement

of the reinforcement distribution is observed up to mould vibrations at 20g. At higher vibrations’


intensity the distribution of ceramic particles is again non-homogeneous and it appears as a series of

almost concentric macrosegregation bands (Fig. 8c).

(a)

(b)

(c)

Fig. 8. Volume fraction of SiC particles from the casting surface to the centre in the A360-10%SiC:

(a) without vibrations, with vibrations at (b) 7g and (c) 41g. Linear fits are also shown.

The quantitative analysis of the distribution of SiC particles over the cross section of the castings

confirms these observations. Fig. 8a shows how the volume fraction of SiC particles increases from

the edge to the center of the samples solidified without vibrations. The mean volume fraction is

~9.9%vol., confirming correct melting and pouring procedures of the MMC. Under vibrations at 7g,

the distribution of the ceramic is more uniform, as evidenced by the linear fit, with an average

volume fraction of ~9.7%vol. (Fig. 8b). An example of specimens solidified under high vibrations’

intensity (41g) is shown in Fig. 9c. The mechanical mould vibration apparently induces a SiC

distribution that is on average constant along the cross of the specimen (~10.5%vol.), as it is

evidenced by the trend in Fig. 9c. The volume fraction profile evidences however the presence of

bands with high SiC concentration (~14.5%vol.) alternated with depleted bands (~7.5%vol.). The

distance between consecutive bands is ~1.9 ± 0.1 mm, which is comparable with the vibration

amplitude value (1.8 mm) used for vibrations’ intensity higher than 20g. This segregation

phenomenon can be explained considering the density difference between the SiC particles (ρ =

3200 kg/m3 [30]) and the molten matrix (ρ = 2400 kg/m

3 [30]), and therefore different force

intensity acting on the solid ceramic particles than on the molten metal. The different density values

produce a different response to inertial forces, in terms of spatial movements, of the ceramic

particles within the molten metal. Other features to be considered are the lower thermal diffusivity

and conductivity values of SiC than the molten matrix alloy. This involves a high inertia cooling of

the ceramic particles than the surrounding melt. Therefore, SiC particles are unable to cool down as

fast as the melt. As a result, the temperature of the particles is somewhat higher than the liquid

alloy. The hotter particles may heat up the liquid in their immediate surroundings, and thus delay

solidification of surrounding molten metal [31].

The macrostructure analysis evidences a grain refinement in the specimens solidified under

different vibrations’ intensity. The structure of the specimen in the as cast condition exhibits coarse

equiaxed grains in the center and finer grains in the chill zone close to the die wall. The effect of

vibrations on the grain size is presented in Fig. 9. By progressively increasing vibrations, the

structure remains equiaxed but finer throughout the casting. In general, the main effect of vibration

on the structure of solidifying metals and alloys is the suppression of columnar growth and the

formation of small equiaxed grains. The observed grain refinement is explained in terms of

fragmentation of α-Al dendrites [17-19, 32]. The vibration induced turbulent movement of the

liquid between the solid dendrites subjects the growing dendrites to bending stresses. These

dendrites have very little strength and ductility because the temperature is so close to the melting

temperature. Fragmentation occurs due to impact of the liquid with the dendrites, and the small

crystals generated by fracture of the dendrites will act as nuclei [18]. Campbell [33] suggested that,


under vibrations’ conditions, the Reynolds number, Re, indicating the transition from laminar to

turbulent flow regime, can be expressed as:

2 ed f aRe

π ρ

η

= (2)

where de is the diameter of dendrite arm, ρ and η are the density and viscosity of the liquid, f and a

are the vibration frequency and amplitude. Campbell [33] considered that flow is laminar below a

Reynold’s number of 10, and turbulent above 103, while between these values it is mixed. The

Reynold’s number used in the present experiment lies between 66 and 641, for vibrations in the

range 3-29g, and thus the flow of liquid should be mixed. Re increases up to 1159 for specimens

solidified under vibrations at 41g, indicating turbulence conditions.

(a)

(b)

(c)

(d)

Fig. 9. Typical etched cross section of A360-10%SiC specimens solidified (a) without vibrations,

and with vibrations at (b) 7g , (c) 20g and (c) 41g.

Fragmented dendrites created by turbulence of the liquid, temperature fluctuations, bending stresses

and flow of liquid around dendrite arms, are carried to the other parts of liquid, in particular to the

central region of the mould, where they act as nuclei. As this solid is replaced by liquid, the

temperature of liquid and the temperature gradients decrease even faster. Therefore, the temperature

in the mould drops more rapidly and heat will be extracted faster from the mould [19]. Campbell

[32] reported as the limit for dendrites fragmentation can be given by the product of vibration

frequency (f) and amplitude (a):

-10.10 msf a⋅ = . (2)

The product of frequency and amplitude has to exceed 0.01 ms-1

for 10% refinement, 0.02 ms-1

for

50% refinement, and 0.1 ms-1

for 90% refinement [32]. Table 2 summarizes the quantitative results

of the product of frequency and amplitude used in the present work, evidencing how the grain

refinement is actually completed under vibrations at 29g.

Vibrations’ intensity f·a [ms-1

]

3g 0.01

7g 0.03

20g 0.07

29g 0.11

41g 0.19

Table 2. Product of vibration frequency and amplitude used in the present work, as limit for

dendrites fragmentation


Fig. 10 shows the effect of the grain refinement on the SiC distribution. It is noticed that the

reinforcement particles are preferably entrapped in the interdendritic regions in the castings

solidified without vibrations and only few SiC particles segregate at grain boundaries (Fig. 10a). By

increasing vibrations’ intensity up to 20g and reducing the grain size, a higher number of ceramic

particles is detected at grain boundaries rather than in the interdendritic channels (Fig. 10b). SiC is

pushed by the growing α-Al dendrites and gathered together at grain boundaries. At the highest

vibrations’ conditions, with a grain size about 100 µm, the reinforcement exclusively segregates at

grain boundaries (Fig. 10c).

(a)

(b)

(c)

Fig. 10. Polarised light images of specimens solidified (a) without vibrations, and with vibrations at

(b) 7g and (c) 41g. The arrows indicate entrapped SiC particles in the interdendritic regions, while

the dashed arrows show segregations of SiC at grain boundaries.

Conclusions

The effect of mechanical mould vibrations on the microstructure of A360-10%SiCp metal-matrix

composites and on the heat transfer from the casting to the die has been investigated. Based on the

results obtained in the present study, the following conclusions can be drawn.

• Increasing vibrations, the grain size and the microstructure of the castings become finer.

• The dispersion of the ceramic particles is not homogeneous on a macroscopic scale in un-

vibrated specimens with segregation phenomena in the center of the casting.

• By applying a mechanical mould vibration, the increase of convective forces and movements in

the melt produces a more homogenous dispersion of SiC particles on a macroscopic scale. The

improvement of the reinforcement distribution is observed up to mould vibrations at 20g. At

higher vibrations’ intensity the distribution of ceramic particles is again non-homogeneous and

it appears as a series of almost concentric segregation bands.

• The reinforcement distribution throughout the castings is controlled by a mechanical

entrapment mechanism. While the SiC particles are preferably entrapped in the interdendritic

regions in the castings solidified without vibrations, by increasing vibrations’ intensity and

reducing the grain size, a higher number of ceramic particles is detected at grain boundaries

rather than in the interdendritic channels.

• By comparing simulated and experimental temperature curves in the different vibrations’

conditions, different HTC made the best fit. By increasing the vibrations’ intensity, it is

observed how the HTC increases in the temperature range of solidification of the composites.

References

[1] K.U. Kainer, in: Basics of Metal Matrix Composites, edited by K.U. Kainer, Metal Matrix

Composites, chapter 1, Wiley-VCH Verlag GmbH & Co., Weinheim (2006).

[2] B. Verlinden and L. Froyen, in: Aluminium Matrix Composite Materials, TALAT CD-ROM,

Lecture 1402, , European Aluminium Association, Bruxelles (1999).

[3] M.K. Surappa: J. Mater. Process. Technol. Vol. 63 (1997), p. 325.


[4] F. Pinna, in: Proc. Conf. Tecnologie Innovative nella Fonderia Leghe Leggere, Montichiari,

Italy (1998).

[5] P. Bassani, B. Previtali, A. Tuissi, M. Vedani, G. Vimercati and S. Arnaboldi: Metal. Sci.

Tech. Vol. 23 (2005), p. 3.

[6] S. Arnaboldi, P. Bassani, M. Pini, A. Tuissi, M. Vedani and G. Vimercati: Metall. Ital. Vol.

98 (2007) p. 19.

[7] J. Hashim, L. Looney and M.S.J. Hashmi: J. Mater. Process. Technol. Vol. 92-93 (1999), p. 1.

[8] J. Hashim, L. Looney and M.S.J. Hashmi: J. Mater. Process. Technol. Vol. 123 (2002), p.

252.

[9] P.K. Rohatgi, R. Asthana and S. Das: Int. Met. Rev. Vol. 31 (1986), p. 115.

[10] A. Mortensen, V. Michaud and M.C. Flemings: J. Met. 45 (1993), p. 36.

[11] P.A. Karnezis, G. Durrant and B. Cantor : Mater. Charact. 40 (1998), p. 97.

[12] C.G. Levi, G.J. Abbaschian and R. Mehrabian: Metall. Trans. A Vol. 9 (1978), p. 697.

[13] J.P. Rocher, J.M. Quenisset and R. Naslain: J. Mater. Sci. Lett. Vol. 4 (1985), p. 1841.

[14] F. Dellaney, L. Rozen and A. Deryttere: J. Mater. Sci. Vol. 22 (1987) p. 1.

[15] T.P. Fisher: Br. Foundryman Vol. 66 (1973), p. 71.

[16] K. Kocatepe : Mater. Des. Vol. 28 (2007), p. 1767.

[17] J. Bast, J. Hubler and C. Dommaschk: Adv. Eng. Mater. Vol. 6 (2004), p. 550.

[18] K. Kocatepe and C.F. Burdett: J. Mater. Sci. Vol. 35 (2000), p. 3327.

[19] J. Deshpande and M.M. Makhlouf: AFS Trans., paper 08-104, May, 2008, Atlanta, GA.

[20] N. Abu-Dheir, M. Khraisheh, K. Saito and A. Male: Mater. Sci. Eng. A Vol. 393 (2005), p.

109.

[21] D.R. Herling, G.J. Grant and W. Hunt Jr.: Adv. Mater. Processes 159 (2001), p.37.

[22] DURALCAN, in: Composites for High-Pressure die casings: Mechanical and Physical

Property Data. Duralcan, USA, San Diego, CA, (1995).

[23] G. Timelli, P. Ferro and F. Bonollo: Metall. Ital. Vol. 102 (2010), p. 11.

[24] F. Grosselle, G. Timelli, F. Bonollo, A. Tiziani and E. Della Corte: Metall. Ital. Vol. 101

(2009), p. 25.

[25] F. Piasentini, F. Bonollo and A. Tiziani: Metal. Sci. Tech. Vol. 23 (2005), p. 11.

[26] L. Bäckerud, G. Chai and J. Tamminen: Solidification Characteristics of Aluminum Alloys,

Volume 2: Foundry Alloys (American Foundrymen’s Society Inc., Des Plaines, IL, 1990).

[27] M.M. Makhlouf and H.V. Guthy: J. of Light Metals 1 (2001), p.199.

[28] A. Manente and G. Timelli: Metall. Ital. 100 (2008), p.37.

[29] X. Jian, T.T. Meek and Q. Han: Scripta Mater. 54 (2006), p. 893.

[30] J.C. Baez, C. Gonzalez, M.R. Chavez, M. Castro and J. Juarez: J. Mater. Process. Technol.

153-154 (2004), p. 531.

[31] W. Zhou and Z.M. Xu: J. Mater. Process. Technol. 63 (1997), p. 358.

[32] J. Campbell: Casting (Elsevier Butterworth-Heinemann , Oxford 2003).

[33] J. Campbell: Int. Met. Reviews 2 (1981), p. 71.


Processing of Lightweight Metal Matrix Composites via In-Situ Gas/Liquid Reaction

Cecilia Borgonovoa and Diran Apelianb

Metal Processing Institute, Worcester Polytechnic Institute

Worcester, MA 01609 USA [email protected], [email protected]

Keywords: lightweight, composites, aluminum, magnesium, gas/liquid reaction, nitridation.

Abstract. Aluminum nitride (AlN) possesses superior thermal and electrical properties and is an

ideal candidate for high-temperature, as well as for packaging and optoelectronic applications.

Aluminum based composites reinforced with AlN have been manufactured via an in situ gas-

assisted process, where a nitrogen-bearing gas is injected in the molten aluminum at 1273-1323 K.

The process is carried out in an inert atmosphere in order to avoid oxygen contamination. Addition

of Mg lowered the oxygen content in the melt by forming MgO and thus favoring the nitridation

reaction. Particle size formed in the matrix varied from 1- 3 µm to sub-micron scale depending on

the gas injection time. Longer bubbling times give rise to improved reinforcement dispersion.

Addition of Si is detrimental for the synthesis of AlN; 2Mg Si phase precipitates, replacing the

formation of MgO and hindering the nitridation reaction. The challenges of controlling the kinetics

are discussed.

Introduction

Aluminum based nanocomposites have increasingly gained attention as weight-saving functional

materials with improved mechanical properties. Carbide or nitride particles are added to the

aluminum matrix to attain superior hardness, wear resistance and dimensional stability at high

temperatures. Several manufacturing methods such as mechanical stirring, infiltration and powder

metallurgy have been employed to date and the subject has been covered in some detail in a critical

review by the authors [1]. None of the ex-situ conventional processes meet the three key challenges:

scalability, homogeneous distribution, and cost-effectiveness. In contrast, in-situ synthesis routes

offer pathways that address key challenges in the manufacture of nanocomposites for structural

applications [2,4,5]. The secondary phase is created in-situ through a chemical reaction; the in-situ

creation of the reinforcement phase ensures clean and thermodynamic stable interfaces and good

particle dispersion. Moreover, it is possible to produce composites with a broad variety of matrix

materials (aluminum, titanium, copper, nickel and iron) and reinforcing particles (borides, carbides,

nitrides, oxides and their mixtures).

Among the wide range of in-situ techniques, the synthesis of nitride particles by means of a gas-

assisted reaction has shown promise. Hou et al. [3] have been able to manufacture aluminum matrix

composites reinforced with AlN with a diameter smaller than 0.1 µm. Zheng et al. [4] have

converted 14% weight of Mg-Al alloy into aluminum nitrides. The process involves the

introduction of a nitrogen-bearing gas in the melt so that nitridation of aluminum takes place.

Control of process variables (processing temperature and time amongst others) can tailor the

amount and size of the reinforcement in the matrix [4,5]. Aluminum nitride is a refractory

compound characterized by attractive properties such as high thermal conductivity, high electrical

resistance, low dielectric constant, and a thermal expansion coefficient similar to silicon [6]. It is

suitable for producing substrates and packaging materials in high-power integrated circuits, as well

as coatings, insulators and optoelectronic devices. Although liquid nitridation has been widely


investigated over the years, the mechanism (and kinetics of formation) of AlN is not well

understood. Two different formation mechanisms have been identified: direct nitridation according

to the reaction 22Al N 2AlN+ → and indirect nitridation assisted by a catalyst such as magnesium.

The latter involves the formation of an intermediate phase ( 3 2Mg N ) through the reaction

2 3 23Mg N Mg N+ → followed by the substitution reaction 3 2Mg N 2Al 2AlN 3Mg+ → + . Several

publications refer to this mechanism as more likely to be responsible for nitride formation than

direct nitridation [8,9,11]. Shtapitanonda and Margrave [10] observed the tendency of magnesium

nitrides to form in the gaseous phase after the volatilization of magnesium when melted in a

controlled atmosphere. The substitution reaction takes place once the 3 2Mg N phase is in contact

with the melt. Pech-Canul et al. [11] pointed out how the formation of magnesium nitrides is

kinetically more favored than the formation of AlN. Moreover, they confirmed the occurrence of

the substitution reaction to form AlN, which is a more thermally stable compound at around the

process temperature range (1273-1373 K). Despite this, there is no agreement about the formation

mechanism (direct or indirect). Dopants (Mg, Si) and oxygen content in the reactive gas have been

found to be important parameters. Scholz and Greil [12] stated that for higher Mg/Si ratio and for

decreasing oxygen content in the gas, the conversion from Al to AlN is more favorable. Jinxiang et

al. [13] investigated the influence of Mg and Si on the rate of nitride formation, underlining the

predominant role of magnesium over silicon. Zheng and Reddy [7] found that amount of nitrides

formed are increased when ammonia is used as a reactive gas. Ammonia’s oxygen-getter action,

due to the dissociation of nitrogen and hydrogen at around 1273 K, results in lower oxygen content

and thus a lower oxygen partial pressure in the melt. The detrimental effect of oxygen is clear when

the thermodynamics of the system are analyzed. The Ellingham diagram of the reaction through

which aluminum oxides are formed 2 2 34Al 3O 2Al O+ → shows a lower Gibbs free energy

compared to the Gibbs energy of the nitridation reaction (Fig.1).

-1200

-1000

-800

-600

-400

-200

0

0 500 1000 1500 2000

Temperature (K)

Gibbs Free Energy (kJ)

Al2O3

AlN

Fig.1. Ellingham diagram for the nitridation and oxidation of aluminum.

Studies of the initial nitridation period of aluminum at 673 K and higher show that the rate of

nitridation is much slower than the rate of oxidation at a given temperature [13]. The aim of this

work is to establish the feasibility of aluminum nitridation via in-situ gas/liquid reaction. The effect

of the addition of catalysts such as Mg and Si has been investigated and a mechanism of nitride

formation purposed.

Experimental Procedure

Experiments were carried out in a sealed stainless steel resistance furnace with a temperature range

of 1223-1273 K; the setup is shown in Fig.2. The temperature is measured by two K-type


thermocouples placed in the furnace walls and inside the crucible. Alloy compositions selected for

the experiments are 100wt% Al, Al-15wt% Mg, Al-15wt%Mg-8wt%Si (see Table 1). Both

commercial 3NH -99.998% pure- and High-Purity 2N -99.9999% pure- (purchased from AIMTEK)

were employed as reactive gases. Pure Al ingots (purchased from ALCOA) were sectioned and

ultrasonically cleaned in acetone for 20 minutes. For each experiment, 150 g of metal was melted;

alloying elements were added to the pure Al in the reaction crucible and placed in the furnace. A

uniform temperature distribution in the crucible was ensured by properly placing it in the furnace. A

fiberscope camera was inserted laterally in the furnace walls to ensure alignment between the

crucible and the nitrogen-bearing injection tube. Prior to every run, the chamber was cleaned in

order to avoid contamination of the melt by impurities such as dust and coating material. Once the

gas-delivery tubes and the thermocouples were fixed in place, the furnace was sealed. The chamber

was subsequently evacuated and purged with High Purity Argon Grade 5 gas four times in order to

minimize oxygen presence inside the furnace. During the heating process, inert atmosphere is

maintained by constantly injecting Argon at a flow rate of 0.2 l/min. The reaction temperature is

1273 K and it is held constant by an adjustable power controller. When the reaction temperature

was reached, an alumina tube of 1.5 mm diameter is submerged in the melt and nitrogen-bearing

gas is bubbled through the tube at a flow rate of 0.1 l/min and a gas pressure of 0.1 MPa. Two high

capacity oxygen-and moisture-removal traps were used in series at the gas inlet. Each trap can

lower the oxygen content to less than 1 ppb and moisture levels to less than 10 ppb.

Fig.2. Schematic of in-situ gas/liquid process.

The gas was bubbled through the melt for a designated time and the furnace power was turned off.

The metal is left to cool down in the inert atmosphere in order to avoid oxygen contamination.

Samples were taken from bottom, middle and top part of the crucible to characterize reinforcement

distribution at different lengths. The samples were mounted in green phenolic powder and polished

according to standard procedures; the sample was then cleaned ultrasonically for 20 minutes to

remove residuals of alumina and colloidal silica. The samples were sputter-coated with carbon so

that the AlN particles are conductive for Scanning Electron Microscopy (SEM). X-Ray Diffraction

(XRD) analysis was performed in order to detect the presence of nitrides and secondary phases.

Field Emission Gun SEM has been employed for microstructure observation, Energy Dispersive X-

ray (EDS) microanalysis and X-ray mapping.


Experiment Gases Al alloying elements

Mg (wt%) Si (wt%)

Process time (hr.)

1

Al + 2N

Al + 3NH

0 0

0 0

2, 6, 8

2

Al + 2N

Al + 3NH

15 0

15 0

2, 6, 8

3 Al + 2N

Al + 3NH

15 8

15 8

2, 6, 8

Table 1. Details of Nitridation Experiments.


Pure Al was bubbled with both nitrogen and ammonia gas for 30 minutes, 1 hour and 2 hours,

respectively, under evacuated and inert atmosphere (Experiment 1- Table 1). XRD analysis of the

top, middle and bottom part of the crucible revealed that no nitrides were formed. This result

suggests that a catalyst (such as Mg) needs to be added to the metal in order for the nitridation

reaction to occur - as suggested by previous work [7, 8, 11]. In order to investigate the role of

magnesium on the nitridation reaction, 15wt% was added to pure Al (Experiment 2). The results

differ depending on the bubbling time. When the de-oxidized nitrogen is injected in the melt for 30

minutes no aluminum nitrides are detected. Whereas, when Nitrogen gas was injected in the Al-Mg

melt for 1 h, a consistent amount of nitrides was observed. XRD analysis confirms strong peaks of

AlN in the upper part of the crucible along with MgO (Fig.3). SEM analysis shows the presence of

AlN with two different morphologies: embedded in the microstructure (Fig.4a) or as AlN powder

(Fig.4b). In the powder phase, MgO is observed on the AlN particles and tightly connected with

them. The size of the aluminum nitrides ranges from 1 to 3 µm while submicron MgO is also

detected. The bubbling time was further increased to 2 hours and AlN was observed throughout the

whole casting. AlN is present with two different morphologies - embedded in the microstructure

(Fig.5a), and pockets of powder (Fig.5b). XRD analysis reveals AlN and MgO peaks also in the

middle/bottom part of the crucible. The peaks in middle/bottom part of the crucible are less intense

than at the top of the casting and have XRD patterns similar to that in Fig.3. In sum, the amount of

AlN in the middle of the casting is less than the top of the casting. The average size of AlN is

smaller for shorter injection times compared to when gas injection for longer times – i.e., 2 hrs. The

size of the AlN formed is around 1 µm for the particles embedded in the microstructure and ~ 0.5

µm in the powder phase. Size control still remains an issue and kinetics and control of particle size

work is continuing at the Metal Processing Institute (WPI). No difference was noticed in AlN

formation between the use of ammonia and nitrogen gas. Ammonia quickly dissociates into

nitrogen and hydrogen when in the injection tube. In addition, the use of ammonia is undesirable

because of the high amount of porosity that is observed at the center/bottom of the casting

(Experiment 3, when injected for 2 hours). This is explained by the fact that the fraction of AlN

formed is initially limited to the upper portion of the melt, which increases the viscosity of the melt

in this region. As a result, the melt traps more hydrogen causing porosity.


Fig.3. XRD pattern of the upper part of the crucible for 1 h injection time (Experiment 2).

a) b)

Fig.4. a) SEM image of AlN imbedded in the matrix in the upper part of the crucible; b) Pockets of

AlN and MgO powder in the upper part of the crucible Experiment 2) - 1 hour injection time.

AlN MgO


a) b)

Fig.5. a) SEM image of AlN imbedded in the matrix in the middle part of the crucible; b) Pockets of

AlN and MgO powder in the middle part of the crucible (Experiment 2, 2 hours gas injection time).

Silicon is an important element for fluidity and its influence on the nitridation of aluminum was

investigated (Experiment 3). The literature contains contradictory information about Si’s effect

when it is added to the melt together with Mg [10,11,12]. In these experiments, when 8wt% Si was

added to the Al-Mg melt, AlN was not formed whether nitrogen or ammonia gas was used. XRD

pattern reveals strong peaks of silicide phase - 2Mg Si (Fig.6). EDS (Fig.7b) and X-ray mapping

further confirm the presence of the silicide phase. Oxygen is undesirable since it favors the

formation of aluminum oxides versus aluminum nitrides. It can be noticed (Fig.8) that the

permissible oxygen partial pressure for nitridation is e-16 MPa at 1273 K, value that can be hardly

achieved with commercial oxygen traps. The presence of MgO together with AlN suggests that the

former could act as an oxygen-getter to reduce the oxygen partial pressure in the melt. At this point,

aluminum nitrides form by direct nitridation according to the reaction 22Al N 2AlN+ → .

Therefore, nitridation does not occur indirectly by substitution with Mg but through a Mg assisted

direct reaction. The Ellingham diagram of MgO, AlN and the Al-Mg substitution reaction is given

in Figure 9. It can be noted that MgO is thermodynamically stable over a wide range of

temperatures and that indirect nitridation 3 2Mg N 2Al 2AlN 3Mg+ → + is less favorable than the

sequence 22Mg O 2MgO+ → and 22Al N 2AlN+ → . When silicon is added to the melt, AlN and

MgO are not formed, while Mg2Si phase formed as evidenced by the microstructural analysis. It is

hypothesized that the synthesis of the silicide phase is favorable compared to synthesis of MgO, and

that Mg in the melt was depleted by the precipitation of Mg2Si. An important note concerning the

temperature of formation of MgO, and therefore of AlN is worth making. The formation of Mg2Si

starts at 923 K and is completed at 823 K (Fig.10). Therefore, MgO formation must occur at

temperatures equal or smaller than 953 K. This leads us to note that direct nitridation of aluminum

takes place during the cooling process and not at temperatures ~ 1273 K.


Fig.6. XRD pattern of Al-Mg-Si microstructure (Experiment 3).

a) b)

Fig.7. a) SEM magnification of the magnesium silicide phase; b) EDS analysis.


Fig.8. Maximum oxygen partial pressure Fig.9 Ellingham diagram for indirect nitridation

for nitridation vs. temperature. and Mg assisted direct nitridation.

Fig.10. Fraction solid of magnesium silicide formed vs. temperature (Pandat Software).

Conclusions

Gas-assisted nitridation of aluminum is feasible. The in-situ route to manufacture nanocomposites

has the potential to be a commercial process where scalability, homogeneous distribution and cost-

effectiveness are important criteria.

AlN particles, whose thermal and electrical properties are exceptional, have been successfully

synthesized. Specifically:

- Particle sizes in the sub-micron range were achieved when the gas was injected in the melt

for 2 hours.

- Distribution is improved for longer injection times. For shorter bubbling times (1 hour) AlN

were observed only on the upper section of the casting, while for longer times (2 hours) AlN

was found throughout and particularly the middle and bottom sections of the casting.


- Ammonia does not improve the rate of nitride formation and causes an increase in porosity

especially for long injection times. This can be attributed to the entrapment of hydrogen in

the upper part of the crucible where viscosity is higher due to AlN and MgO that synthesize

at an early stage.

- Addition of Mg in the casting is fundamental for the mechanism of formation of AlN. When

pure aluminum was used as matrix, no reinforcement was formed. XRD analysis and SEM

observation showed the presence of MgO along with AlN. This suggests an alternative

hypothesis about the mechanism of formation of nitrides. Oxygen content is lowered and

AlN forms through direct nitridation or through direct Mg-assisted nitridation.

- Silicon totally hinders the nitridation reaction. No MgO has been detected but magnesium

silicide is present in the microstructure. This suggests that 2Mg Si suppresses the formation

of MgO. Since the latter precipitates during cooling, aluminum nitridation may take place at

lower temperatures during cooling.

The control of AlN particle size and the kinetics of the nitridation process need further study. This

work has shown that the process is feasible and that this in-situ approach has merit and has

commercial potential. The aim of producing nano sized (30-40 nm particles) was not achieved in

these experiments; however, the pathway to do so was clearly laid out. This work is continuing to

establish the mechanisms to address the kinetics of the reaction in order to enable us to attain

particles that are not submicron but rather in the nano range.

Acknowledgements

The authors gratefully acknowledge the member companies of the Advanced Casting Research

Center (ACRC) of the Metal Processing Institute for their support of this work, and for their

continued support of research focused on the science and technology of metal casting at Worcester

Polytechnic Institute.

References

[1] C.Borgonovo and D.Apelian: submitted to Mat. Sci. Forum (2010).

[2] S.C. Tjong, Z.Y. Ma: Mat. Sci. Eng. Vol. 29 (2000), pp. 49-113.

[3] Q. Hou, R. Mutharasan and M. Koczak: Mat. Sci. Eng. Vol. A195 (1995), pp. 121-129.

[4] S.Tyagi, Q.Zheng and R. Reddy: Aluminum 2004, edited by S. K. Das, TMS, Warrandale,

(2004), pp. 63-72.

[5] H.Z. Ye, X.Y. Liu and B. Luan: J. Mater. Process. Tech. Vol. 166 (2005), pp. 79–85.

[6] J. Haibo, K. Chen, Z. Heping, S. Agathopoulos, O. Fabrichnaya and J.M.F. Ferreira: J. Cryst.

Growth Vol. 281 (2005), pp. 639–645.

[7] Q. Zheng and R. Reddy: Metall. Mater. Trans. Vol. 34B (2003), pp. 793-805.

[8] Q. Zheng and R. Reddy: Adv. Eng. Mater. Vol. 5 No. 3 (2003), pp. 167-173.

[9] Q. Zheng and R. Reddy: J. Mater. Sc. Vol. 39 (2004), pp. 141-149.

[10] P. Shtapitanonda and J. Magrave: Symposium at University of Wisconsin-Madison (1956).

[11] M. I. Pech-Canul, R.N. Katz and M. M. Makhlouf: Metall. Mater. Trans. Vol. 31A (2000), pp.

565-573.

[12] H. Scholz and P. Greil: J. Mater. Sc. Vol. 26 (1991), pp. 669-677.

[13] L. Jinxiang , G. Xiuying , C. Jianfeng , W. Qun, S. Yuhui and G. Qin: Thermochim. Acta Vol.

253 (1995), pp. 265-273.


Effects of reinforcement parameters on fatigue strength of aluminium-based particulate-reinforced composites

Maurizio Vedani

Politecnico di Milano, Dipartimento di Meccanica

Via G. La Masa 1, 20156 Milano Italy

[email protected]

Keywords: aluminium-matrix composites, particulate, fatigue, damage

Abstract. A study on low-cycle and high-cycle fatigue behaviour of 6061-Al2O3 composites

reinforced with nominal volume fractions of 10% and 20% of Al2O3 particulates is presented. The

effects of reinforcement geometrical features (volume fraction and size) and of the loading mode

experienced during the different kind of fatigue tests (strain controlled and stress controlled tests)

were evaluated. A relation with crack growth mechanisms was drawn by analyses on fracture

surfaces and on longitudinal sections of specimens subjected to the fatigue tests. The

micromechanisms of cyclic deformation and of microstructural damage acting in the materials are

discussed and compared to data and observations available from the wide published literature.

Introduction

Aluminium-matrix composites have been of interest as structural engineering materials owing to

their high specific stiffness and strength as well as interesting wear properties compared to the

unreinforced aluminium alloys. Within the class of metal matrix composites (MMCs), particulate or

more generally, discontinuously reinforced composites (DRCs), are of special interest due to their

low cost and ease of processing by conventional means such as forging, casting, rolling, extrusion

and machining.

DRCs are thus excellent candidates for structural components in the aerospace and automotive

industries, where they are often subjected to cyclic loads. As a result, the fatigue behaviour of

MMCs has received considerable attention during last decades. An extensive review on this subject

was published by Llorca in 2002 [1] while further focus on specific aspects of fatigue performance

were given by several authors in more recent times. The effects of both SiC particulate and Al2O3

short fibres volume fraction on high-cycle fatigue (HCF) response were investigated in [2] and [3],

respectively, whereas recent data on a wide number of materials differing in their reinforcement

volume fraction, shape and size are available from [4-6]. The influence of matrix aging condition

was also considered in a study on a 6061-SiC composite [7]. As a rule, the published data show that

the overall HCF performance and the fatigue limit of DRCs exceed those of the unreinforced alloys

and improved with reinforcement volume fraction. The reinforcement size has an inverse effect on

fatigue limit of DRCs, especially when considering the actual maximum size of the particles found

in the composite rather than the average size of the reinforcement [1]. It is to note that these

improvements in fatigue limit and fatigue strength were achieved only when the composite

materials were correctly processed and their structure was free of flaws. In fact, porosity, particle

clusters, reinforcement debonding and other microstructural defects related to ceramic

reinforcements distribution and processing may reduce the composite cyclic performance well

below that of the unreinforced matrix [8].

In the low-cycle fatigure (LCF) regime, the composite life revealed to be generally inferior to that

of the unreinforced matrix when the comparisons are made in terms of cyclic strain amplitude. This

behaviour was usually accounted for by the premature failure of the ceramic particles under the


relatively high tensile peak stresses. However, it must be considered that comparisons among these

different material families (DRCs vs. unreinforced alloys) in terms of total strain amplitude is

highly favorable to the unreinforced alloys which carry significantly lower stresses and plastic

strains due to their lower stiffness and strain hardening capacity [1, 9].

In the present paper experimental data on LCF and HCF of particulate reinforced 6061-Al2O3

MMCs reinforced with a nominal volume fraction of 10% and 20% are presented. The composite

performances are discussed in light of the reinforcement geometrical features (volume fraction and

size) and of their actual loading mode experienced during the different kind of fatigue tests.

Metallographic and fractographic analyses allowed identifying the micromechanisms of cyclic

deformation and microstructural damage acting in the MMCs investigated. These are discussed and

compared to data and observations available from the wide published literature.

Materials and experimental procedures

The materials investigated are two 6061-Al2O3 composites nominally reinforced with 10 vol.% and

20 vol.% particulates. The composites were produced by a proprietary molten metal process and

supplied in the form of extruded bars having diameters of 20 mm. Specimens were cut from the

bars, heat treated to T6 temper (solution treatment for 1,5 hours at 530°C and aging for 18 hours at

160°C for both the composites) and machined to final shape.

Microstructural characterization of the materials was carried out on the supplied materials by

standard metallographic grinding, polishing and etching with a 2% HF aqueous solution. Optical

(OM) and scanning electron microscope (SEM) analyses were carried out to elucidate actual

particle size and volume fraction as well as matrix grain size and distribution of intermetallics.

Tensile tests at room temperature were performed on specimens having a diameter of 6 mm and a

gauge length of 30 mm, strained at an engineering strain rate of 3·10-4

s-1

. Several different fatigue

testing conditions were then selected according to the fatigue regime to be investigated. LCF tests

(from 10 to 104 cycles to failure) were performed on both composites under uniaxial strain

controlled testing conditions, at a frequency of 0,5 Hz, on specimens having a diameter of 12 mm

and a gauge length of 36 mm. HCF performance was evaluated on the 20 vol.% reinforced 6061-

Al2O3 composite. Data were collected by rotating bending fatigue tests (from 104 to 10

7 cycles to

failure), on hourglass shaped specimens with a minimum diameter of 8 mm, cyclically loaded with

a frequency of 30 Hz.

Finally, fractographic analyses were performed on selected fatigue-fractured specimens. Both

fracture surfaces and polished and etched longitudinal sections of broken specimens were used to

elucidate the damage and fracture mechanisms as a function of testing condition and material.

Results

Microstructure. Representative optical micrographs of the materials investigated are depicted in

Fig. 1. It can be observed that the alumina particle distribution in both composites is rather uniform

without any evidence of large clusters. Quantitative analyses on particle features, whose results are

summarized in Table 1, showed that the average reinforcement size and the actual volume fraction

were of about 22,5 µm and 21,0% for the highly reinforced composite and of 12 µm and 14,0% for

the composite with the lowest reinforcement amount. It can also be inferred from the data given in

Table 1 that differentiation between the two composites was obtained by the producer primarily by

changing the particle size, resulting in a mean free distance among particles (interparticle spacing)

of the same order of magnitude for both materials.


Fig. 1. Representative SEM micrographs of the 6061-10%Al2O3

(a) and of the 6061-20%Al2O3 (b) composites

Material Vf [%] d [µµµµm] λλλλ [µµµµm]

6061-10%Al2O3 14,0 11,5 34,8

6061-20%Al2O3 21,0 22,5 47,0

Table 1. Geometrical features of the Al2O3 particles found in the composites investigated. Vf:

volume fraction; d: average particle size; λ: mean interparticle spacing

Tensile properties. The tensile properties of the materials investigated recorded by uniaxial static

tests at room temperature are given in Table 2. Of particular interest for the following investigation

on fatigue behaviour is the marked increase in elastic modulus recorded for the 6061-20%Al2O3

composite and a slightly higher value of the yield strength for the 6061-10%Al2O3 material owing

to the different shapes of the stress vs. strain curves (see Fig. 2).

Material E [MPa] YS [MPa] UTS [MPa] εεεεu [%] εεεεf [%]

6061-10%Al2O3 77200 320 354 5,5 9,9

6061-20%Al2O3 95700 304 380 2,2 5,2

Table 2. Tensile properties of the materials investigated. E: elastic modulus; YS: 0,2% offset yield

strength; UTS: ultimate tensile strength; εu: uniform elongation; εf: fracture elongation

Low-cycle fatigue behaviour. In Fig. 2 a comparison between the monotonic tensile curves and the

data points corresponding to peak stresses of LCF cycles is presented. It can be observed that in

both composites, a significant strain hardening occurs during fatigue, especially for the 6061-

20%Al2O3 composite.

Fig. 3 supplies further details on evolution of peak stress during fully reversed strain-controlled

fatigue tests. It can be stated that after an initial strain hardening phase, both materials feature a

substantially stable behaviour for the largest fraction of their fatigue life. It is also worth noting that

only at the highest stress values, a cyclic softening in the last stages is recorded, especially for the

6061-20%Al2O3 composite. It is also worth noting in Fig. 3 the systematically higher values of the

peak stresses recorded in compression with respect to those found in the tension part of the cycles.

(a) (b)


The fatigue life of the two materials investigated as a function of both total and plastic strain

amplitude is depicted in Fig. 4. The data clearly show that the 6061-10%Al2O3 composite

systematically has the best fatigue behaviour at any plastic strain amplitude level. The same trend

becomes even more evident when considering the fatigue life of the materials as a function of total

strain amplitude (see Fig. 4a).

0

100

200

300

400

500

0 1 2 3 4 5 6

Str

ess (M

Pa

)

Strain (%)

Monotonic tension

Cyclic deformation / tension

Cyclic deformation / compression

0

100

200

300

400

500

0 1 2 3 4 5 6 S

tre

ss (M

Pa

)Strain (%)

Monotonic tension

Cyclic deformation / tension

Cyclic deformation / compression

Fig. 2. Comparison of monotonic tensile curves and cyclic peak stress data of the composites

investigated. (a) 6061-10%Al2O3; (b) and 6061-20%Al2O3

200

250

300

350

400

450

0,0001 0,001 0,01 0,1 1

Peak s

tress (

MP

a)

N/Nf

Tension

Compression

200

250

300

350

400

450

0,0001 0,001 0,01 0,1 1

Peak s

tress (

MP

a)

N/Nf

Tension

Compression

Fig. 3. Evolution of peak stress in tension and compression during low-cycle fatigue tests of the

6061-10%Al2O3 (a) and 6061-20%Al2O3 (b) composites

0

0,2

0,4

0,6

0,8

1

1,2

10 100 1000 10000

Tota

l str

ain

, ∆

ε/2 (

%)

Reversals to failure, 2Nf

6061-10%

6061-20%

0

0,2

0,4

0,6

0,8

1

1,2

10 100 1000 10000

Pla

stic s

train

, ∆

ε pl/2

(%

)

Reversals to failure, 2Nf

6061-10%

6061-20%

Fig. 4. Total strain amplitude (a) and plastic strain amplitude (b) vs. cycles to failure during LCF

tests of the composites investigated

(a) (b)

(a) (b)

(a) (b)


High-cycle fatigue behaviour. The 6061-20%Al2O3 composite was further investigated in the HCF

regime in order to draw information on the material response at lower-amplitude fatigue cycles and

hence higher fatigue lives. Thus, the HCF behaviour was evaluated by a series of rotating bending

tests aimed at defining the fatigue limit and at drawing the Wöhler curve. The estimate of the

fatigue limit at 107 cycles by the staircase method gave a value of 160 MPa with a standard

deviation of 10,6 MPa. The curve was completed by further tests at five increasing stress levels,

using four samples for each level. Fig. 5 depicts the results obtained by these tests.

100

150

200

250

300

350

400

1E+4 1E+5 1E+6 1E+7 1E+8

Str

ess a

mp

litu

de

(M

Pa

)

Cycles to failure

Fig. 5. HCF strength of the 6061-20%Al2O3 composite

Fractography. The analyses on the fracture mechanisms observed in the specimens broken under

the different fatigue conditions were carried out by SEM either on the fracture surfaces and by

sectioning (by diamond-saw cutting) and polishing a longitudinal plane crossing the specimen axis

and the expected regions of crack nucleation and growth.

Fig. 6. Typical microstructural flaws leading to nucleation of fatigue cracks in the composites. (a)

Coarse Al2O3 particle close to specimen surface in a 6061-10%Al2O3 composite; (b) cluster of small

particles not wetted by the Al matrix during manufacturing in a 6061-20%Al2O3 (detected in a

sectioned specimen)

Usually, nucleation of fatigue cracks took origin from defects associated to the presence of coarse

Al2O3 particles, clusters of particles or other oxide inclusions located close to the external surfaces

of the specimens, as shown in Fig. 6.

(a) (b)


The stage of crack growth revealed to be strongly dependent on the amount and size of

reinforcement as well as on the stress intensity factor (∆K) felt by the materials at the different

stages of crack propagation. These differences were easily detected in HCF tested specimens where

the ∆K experienced a much wider range at initial and final stages of crack development. Indeed, the

fracture surface corresponding to initial fatigue crack growth mainly developed through the matrix

and intersected only a relatively low number of alumina particles (Fig. 7a). Moving on toward the

region corresponding to final growth stage, the fracture surface underwent a progressive

modification with an increasing number of fractured ceramic particles exposed on the crack path, as

depicted in Fig. 7b. In these latter zones, the fracture mechanism was of ductile type, governed by

void nucleation at cracked alumina particles and growth through the matrix ligaments.

Fig. 7. Fracture appearance of a 6061-20%Al2O3 composite tested under HCF regime with a stress

amplitude of 150 MPa and failure at 6·106 cycles. (a) Region of early crack propagation; (b) region

close to final specimen fracture

The analyses on the sectioned specimens confirmed the above results, showing the presence of an

increasing number of cracked particles close to the fracture line when approaching the regions of

high ∆K. This feature clearly demonstrates that the process of particle cracking and linkage through

the matrix ligaments was the leading mechanism for failure at high ∆K values, especially in the

6061-20%Al2O3 composite featuring larger particles (see Fig. 8). On the contrary, in the early

stages of crack propagation, at much lower ∆K values, the reinforcement remained intact and the

cracks were forced to bend through the matrix, developing a more tortuous path, as shown in Fig.

8a.

Discussion

It is known from literature that the incorporation of ceramic particles in Al alloys does not usually

lead to significant improvement in LCF resistance, especially when relatively coarse reinforcement

particles are considered [1, 10]. This trend is also confirmed for the 6061-20%Al2O3 composites

here investigated. Even though a direct comparison with experimental data of the unreinforced alloy

is missing, the comparison between the 10 vol.% and the 20 vol.% Al2O3 particle reinforced

composites distinctively suggests that a higher fraction of coarser particles reduces the low cycle

fatigue life, especially at high plastic strain amplitudes (see Fig. 4b). The peculiar effect of particle

size on fatigue performance of composite has been clarified since several years [9, 11-16]. Already

in early nineties Kumai and co-authors were able to state that in powder metallurgy processed 6061-

SiC composites the interaction mechanisms of cracks and reinforcement particles depended on

crack length and/or on the applied stress range [15]. Fatigue cracks developed by avoiding the

reinforcement at low ∆K ranges (thus leading to slower growth rates over the unreinforced alloy)

whereas at high ∆K ranges, the cracks appeared to proceed by linking the fractured SiC particles

(a) (b)


that had failed due to overload within the process zone ahead of the crack front. Li and Ellyin

studied in great detail the stage of particle cracking and crack growth in 10 and 20 vol.% Al2O3

reinforced Al 6061 alloys, apparently similar to the materials here investigated [14]. Their results

highlighted that ceramic particles in MMCs are much stronger barriers for short cracks than many

other microstructural defects such as grain boundaries and that the threshold for the short crack

growth (i.e. the fatigue limit) could be directly related to the ability of the particles to trap the crack-

tip cyclic plasticity.

Fig. 8. SEM images of longitudinally sectioned specimens of the 6061-20%Al2O3 composite broken

under LCF regime at imposed plastic strain amplitude of 0,30 %. (a) Early stages of crack growth

with limited evidence of particle cracking; (b) stage of fast crack growth with extensive damage of

reinforcement beneath the fracture line. (c) stage of fast crack growth, showing cracks developing

through the matrix by linking already fractured particles

In more recent times, Uematsu, Tokaji and Kawamura focused on the effect of particle size in

2024/SiC reinforced composites, fatigue tested at room and high temperatures (up to 250°C)

[10,17,18]. They confirmed that under HCF regime and at room temperature, the composites

reinforced with small particles (5 and 20 µm) featured an improved resistance over the unreinforced

matrix while the composite with relatively large particles (60 µm) had an opposite trend. At

increasing temperature, the particle size dependence became less evident and eventually, at 250°C,

the fatigue strength of all the composites was nearly the same, irrespective of particle presence and

size. These results were accounted for by an increasing dominant role played by the matrix at high

temperature and by changes in the damage mechanisms associated to reinforcements.

(c)

(a) (b)


As for the damage mechanisms, it is well accepted that weak interfaces in MMCs, generated due to

insufficient wetting and absence of a strong chemical bonding between the two phases, prevent a

proper stress transmission from the matrix to the reinforcement to be achieved [19]. Chemical

reaction at the interfaces can in fact improve the interface strength. However, too coarse and brittle

reaction products as well as excessive interface roughening can also lead to premature

reinforcement failure due to interface-induced brittle fracture and notch effects acting on the micro-

scale. A representative case of this latter behaviour is given by the composite here investigated,

where a strong interfacial cohesion was caused by the reaction during material manufacturing

between the Mg, present as alloying element in the matrix, and the alumina particles to form the

MgAl2O4 spinel [20]. The tiny spinodal particles observed on the Al2O3 surfaces (see for instance

the irregular surfaces of the particles in Fig. 8c) in turn, may affect particle strength by creating

notch effects on the brittle ceramics, leading to possible failure of the reinforcement, depending on

extension of crack-tip process zone and particle size, as depicted in figure 9.

Fig. 9. Cracking of a reinforcement particle presumably stimulated by the notch effects related to

spinodal MgAl2O4 crystal (arrowed) formed on the surface of the Al2O3 (6061-20%Al2O3

composite specimen broken in HCF condition)

An interesting research work was also published by Srivatsan [21] focusing on the fracture

resistance of Al 2014-Al2O3 particulate reinforced composites subjected to LCF at room and high

temperature. Also in this study it was stated that a progressive deterioration of the composite

microstructure and the concurrent development of microscopic flaws can strongly affect the long

term integrity of the material. The authors demonstrated that, under total strain-amplitude controlled

conditions, the elastic strains were much lower in the composite than they would be in the

unreinforced alloy due to the higher elastic modulus of the composites, thus generating additional

plastic flow in the composite matrices and inducing reinforcement damage. A similar mechanism is

believed to hold also for the materials investigated in the present study, where the significant

increase in stiffness of the 6061-20%Al2O3 composite (see Table 1) significantly contributed to

worsen the overall low-cycle fatigue behaviour with respect to the 6061-10%Al2O3 composite.

When considering the systematic differences found in compression and tension peak stress

evolution during LCF lives of the composites (see Fig. 3), reference can be made again to the data

published by Srivatsan [21]. The author published a set of LCF curves showing marked differences

between the tension and compression peak stresses achieved in fully reversed strain-amplitude-

controlled cycles. This trend was discussed considering both the cyclic behaviour of the matrix (the

2014 alloy featured a progressive softening) and the progressive damage of the reinforcement,

ascribing to it the rapid decrease in tension stress in the last cycles of the composite fatigue lives.

For the 6061 alloy-based composites here investigated, the gap between tension and compression

phases during LCF was actually not as large as that found in [21] and no particular evidence of

depletion in the load carrying ability in tension of the materials was detected prior to fracture. It is


therefore believed that damage of the reinforcement, although being clearly present and strongly

affecting the crack growth mechanisms, is not extended to the degree that it can reduce significantly

the overall material strength during the last stages of fatigue life of the 6061-Al2O3 composites.

Finally, it must be reminded that the typical fatigue crack initiation sites found either in LCF and

HCF tests are represented by defects or inclusions located close to specimen surfaces, very often

related to lack of process control during composite manufacturing (e.g. clusters of non-wetted

particles) or on raw materials selection (e.g. coarse Al2O3 particles). It is well known that the phase

of nucleation plays a major role especially in HCF behaviour, representing a significant part of the

whole fatigue life. Hence a careful control of manufacturing processes for composites can preserve

intrinsic material homogeneity and reinforcement distribution, thus improving the overall fatigue

performance of the MMCs investigated.

Conclusions

The investigation carried out on fatigue behaviour of two Al 6061-Al2O3 particulate reinforced

composites allowed to draw the following conclusions.

• Low-cycle fatigue tests performed on the 6061-10%Al2O3 and 6061-20%Al2O3 composites

showed that the reinforcement volume fraction and particle size strongly affect the fatigue life in

strain-controlled regime.

• The different values of elastic modulus and work-hardening behaviour of the two composites

also have to be considered when drawing comparisons since, for the same total strain amplitude,

the 6061-20%Al2O3 composite experiences a significantly higher amount of plastic strain and

peak stress.

• Damage of Al2O3 particles plays a fundamental role on fatigue life by affecting the crack growth

mechanisms, especially in the stage of high stress intensity factor.

• Under these conditions, coarser alumina particles are prone to cracking under load and lead to

reduced fatigue properties of the highly reinforced composite.

• On the contrary, in the high-cycle fatigue regime, during early stages of crack propagation, the

crack path mainly develops through the matrix and intersects only a relatively limited number of

alumina particles. The reinforcement thus plays a positive role by promoting more tortuous

fatigue crack paths at low ∆K values.

• The importance of a careful control of composite microstructure and reinforcement distribution

was highlighted by stating that most of the nucleation sites for fatigue failure of the samples

occurred from defects or inclusions located close to specimen surfaces such as clusters of

particles or coarse Al2O3 particles.

References

[1] J. Llorca: Prog. in Mater. Sci. Vol. 47 (2002), p. 283

[2] C. Kaynak and S. Boylu: Mater. and Design Vol. 27 (2006), P. 776

[3] Y. Ochi, K. Masaki, T. Matsamura and M. Wadasako: Mater. Sci. Eng. A Vol. 468-470 (2007),

p. 230

[4] B.G. Park, A.G. Crosky and A.K. Hellier: Comp. part B: Vol. 39 (2008), p. 1257

[5] L. Ceschini, G. Minak and A. Morri: Comp. Sci. and Techn. Vol. 66 (2006), p. 333

[6] S.C. Tjong, G.S. Wang and Y.-W. Mai: Comp. Sci. and Techn. Vol.65 (2005), p. 1537

[7] K. Mahadevan, K. Raghukandn, B.C. Pai and U.T.S. Pillai: J. Mater. Proc. Techn. Vol. 198

(2008), p. 241


[8] C. Bosi, G.L. Garagnani, R. Tovo and M. Vedani: Int. J. Mater. Prod. Techn. Vol. 17 (2002), p.

228

[9] J. Llorca and P. Poza: Acta Metall. Mater. Vol. 43 (1995), p. 3959

[10] K. Tokaji, H. Shiota and K. Kobayashi: Fat. and Fract. Eng. Mater. Struct. Vol. 22 (1999),

p. 281

[11] I. Sinclair and P.J. Gregson: Mater. Sci. Techn. Vol. 13 (1997), p. 709

[12] T. Wilkins and Y.-L. Shen: Comp. Mater. Sci. Vol. 22 (2001). p. 291

[13] B.R. Crawford and J.R. Griffiths: Fat. and Fract. Eng. Mater. Struct. Vol. 22 (1999), p. 811

[14] C. –S. Li and F. Ellyin: Fat. and Fract. Eng. Mater. Struct. Vol.18 (1995), p. 1299

[15] S. Kumai, K. Yoshida, Y. Higo and S. Nunomura: Int. J. Fat. Vol. 14 (1992), p. 105

[16] S. Qu, T. Siegmund, T. Huang, P.D. Wu, F. Zhang and K.C. Hwang: Comp. Sci. Techn.

Vol. 65 (2005), p. 1244

[17] Y. Uematsu, K. Tokaji and M. Kawamura: Comp. Sci Techn. Vol. 68 (2008), p. 2785

[18] K. Tokaji: Fat. Fract. Eng. Mater. Struct. Vol. 28 (2005), p. 539

[19] J.M. Howe: Int. Mater. Rev. Vol. 38 (1993), P. 257

[20] J.C. Lee, G.H. Kim and H.I. Lee: Mater. Sci. Techn. Vol. 13 (1997), p. 182

[21] T.S. Srivatsan: Int. J. Fat. Vol. 17 (1995), p. 183


Production and characterization of aluminum iron powder composites with ferromagnetic properties

Stefano Amadoria, Ennio Bonettib, Enrico G. Camparic and Luca Pasquinid

Dipartimento di Fisica Università di Bologna and CNISM, v.le Berti Pichat 6/2, 40127 Bologna,

Italy

[email protected], [email protected], [email protected],

[email protected]

Keywords: aluminum, iron, compacted powders, hardness, anelasticity, induction heating

Abstract. Composites made from compacted powders blends of Al with different Fe contents were

produced and characterised with respect to mechanical and induction heating properties.

Mechanical spectroscopy and hardness measurements were employed to follow the evolution of

Young modulus and internal friction after ageing. It was found that above a critical iron content

(>30% of the volume) a percolation network of Fe grains is obtained inside the specimen and the

induction heating characteristics become comparable with those of ferritic steel samples.

Introduction

Powder compaction [1] is a well established method for the production of composites that presents

many advantages with respect to other processing techniques like casting, forming and machining.

Complex net or near net shaped components combining metals with different physical and

mechanical properties, can be produced, in a way that would be impossible using casting.

Aluminum in particular, thanks to its low density, high thermal and electrical conductivity,

excellent machinability and competitive cost, is widely used in blends of elemental or pre-alloyed

powders, to produce high performance composite materials. A commonly employed procedure

consists first in pressing the powders to obtain partially compacted material, which is then sintered,

hot pressed or hot extruded to obtain a full density material with the desired properties. Yet this

procedure is time consuming, expensive and problematic because of the high temperatures (750 to

900 K for Al and its alloys), and the controlled atmosphere [2].

Purpose of this work is to investigate the possibility to prepare by a simple and cost effective

procedure, composite by mixing Al and Fe powders. The composites synergically combine different

properties: the good compressibility, electrical and thermal conductivity of Al with the magnetic

properties of Fe. This mixing of properties is particulary suited to meet the needs of the induction

cooking industry for the realization of pans to be employed with induction hobs. An induction hob

generate an alternate magnetic field that directly heats a ferromagnetic pan by joule heating due to

eddy currents [3]. Up to now pans used in induction heating systems are produced in two ways:

with multiple layers of different stainless and ferritic steels, or by sticking a ferritic steel layer at the

bottom of an Aluminiun pan. Our composite materials aim to combine the best features of both

materials: the Fe powders should made the composite heatable by any common induction hob while

the use of Al should give the material lightness and the good thermal conductivity necessary for

good cooking properties.

In the following the results of hardness and anelastic parameters (internal friction and Young

modulus) measurements on composites with different Al/Fe ratio, and their evolution after different

heat treatments will be described. Some measurements have been performed also on partially

compacted samples (0,75<ρ<1). To verify the effective use of these materials with electromagnetic

induction devices, a commercial induction heating apparatus was used to investigate their heating


rates, and to compare their performances with that of a reference sample obtained from a

commercial pan for induction cooking.

Sample production and experimental conditions

Fig. 1. Rectangular shaped samples and disc shaped sample of Al-Fe powder composites obtained

by consolidation.

The samples were produced from commercial Aluminium and iron powders, 99% purity with grain

size < 60 µm. Optical microscopy observations showed that powder particles were roughly

spherical with a gaussian size distribution centered at a diameter of 17 µm. Powders were mixed

with the desired Al/Fe weight ratio in the range 8:2-1:9 and consolidated in form of bars (20 mm x

4 mm x 0.5 mm) or discs (Φ=100 mm, h=6 mm) (Fig. 1) using uniaxial compaction, under a

pressure of ~1 GPa. The Fe volume fraction for fully consolidated samples span the range 0.08-

0.76. Two typical optical images of fully consolidated samples with different Fe/Al weight ratio are

reported in Fig. 2.

The compaction curves (density vs. applied pressure) of the elemental components (Al and Fe),

have been compared with those obtained for different composition blends by the rule of mixtures

[4]. The samples density was measured by the Archimedean method for the fully compacted

samples and by the geometrical method for both fully and partially compacted ones. The measured

values have been compared with theoretical ones as predicted from the weight (wt%) of the

powders in the composite and using Eq. 1, 2 :

VAl=[(wtAl/100)ρFe]/[ ρAl-wtAl(ρAl-ρFe)/100] (1)

ρc=ρAlVAl+ρFeVFe (2)

where VAl is the volume fraction of Al, ρAl , ρFe and ρc are respectively the densities of Al (2.7 x

103 Kg/m

3),Fe (7.87 x 10

3 Kg/m

3) and composite, VFe=1-VAl. The calculated values are in good

agreement within 1% with the measured ones.

Samples were submitted to different thermal treatments, one set was heated to 473 K at 0.17 K/s

and aged at this temperature for 7.2 x 103 s (2 hours), a second one was submitted to a heating run

up to 723 K at a rate of 3.33x 10-2

K/s. Both treatments were made in argon atmosphere or under a

pressure <10-4

Pa and are used to improve the conductivity [5] and stabilize the structure of the

samples. The Vickers hardness of fully compacted samples was measured with a Shimadzu micro

hardness tester. For each Al/Fe composition, and for the reference samples, the measurements have


been made on several samples subjected to the same processing route and thermal treatments at 0.3

Kg load and along all the surface of the sample.

Fig. 2. Optical images of fully compacted (ρ=1) samples with different Fe content: a) Fe volume

fractions: 0.08, b) Fe volume fraction: 0.34

The inverse mechanical quality factor (Q-1

) and dynamic Young modulus of partially and fully

compacted samples, have been measured by a completely automated vibrating reed analyzer VRA-

1604 (CANTIL srl), in the 1.5-2.5 kHz frequency range at a strain amplitude ε <3x10-5

and at a

pressure of 10-4

Pa. A few measurements at low frequency (0.1-10 Hz) were performed by a

Dynamic Mechanical Analyzer (TA instruments). A commercial E.G.O. induction heating

apparatus has been used to test the heating rate of the compacted disks with different compositions

aged 7.2 x 103 s (2 hours) at 473 K and to compare it to that of a reference sample made of ferritic

steel. The samples surface temperature was monitored with thermocouples and all heating tests

were realized with the same constant power output and experimental conditions, so that any

difference in the heating rates of the disks was linked only to their composition.


Fig. 3 shows the Vickers hardness values, as a function of the Fe volume fraction, for the fully

compacted samples in the as pressed state and after the two heat treatments. Each experimental


value reported is the mean of about 20 measurements performed on the whole sample surface. It is

worth noting that hardness greatly decreases after thermal treatment up to 723 K, with the only

exception of the 100% Fe samples. The as-pressed samples are harder, owing to cold work

hardening, therefore the dislocation density and the internal stresses of the material increases. The

heating of the samples induces a relaxation of the internal stresses, lowers the dislocation density

and so the hardness decreases.

0.0 0.2 0.4 0.6 0.8 1.00.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Vic

kers

Hard

ness (

GP

a)

Fe volume fraction

Fig. 3. Vickers hardness as a function of Fe content for: as pressed samples (■), after isothermal

treatment for 2 hours at 473 K (○) and after a temperature ramp up to 723 K at 3.33 x 10-2

K/s (▲).

These results were compared with the Voigt and Reuss models derived from the rules of mixtures

(ROM), a simple method for the estimation of effective mechanical properties of a composite based

on its constitutents [6]. The Voigt and Reuss models represent the two phases in iso-strain and iso-

stress condition respectively. The actual stress/strain state in the composite lies between these two

conditions, however the models described by Eq. 3, 4 can be used as a good estimate of the upper

(Voigt) and lower (Reuss) bound of the effective hardness value.

HVoigt=fAlHAl+fFeHFe . (3)

HReuss=(fAl/HAl+fFe/HFe)-1

. (4)

In these equations H is the effective hardness of the sample, HAl and HFe respectively those of Al

(the soft phase) and Fe (the hard phase), fAl and fFe the respective volume fractions. The measured

hardness values (Table 1), for as pressed and thermally treated samples with a Fe volume fraction >

0.3, are well within these upper and lower bounds whereas at lower iron content they fall under the

lower limit set by the Reuss model and remain close to those of the un-reinforced aluminum matrix.

This behavior is more evident for samples submitted to a temperature ramp. This can be explained

considering that the aluminum hardness is only 1/3 of that of iron (respectively 0.54 and 1.37 GPa

for the as-pressed samples, 0.52 and 0.35 GPa versus 1.4 and 1.42 GPa for the heat treated ones),

therefore, at low Fe volume fractions strain is accounted for mainly by the aluminium matrix. At

higher iron volume fraction (≥ 34%) the Fe particles are closer to each other and under compression

the load is transferred by the softer matrix to adjacent harder particles which shares part of the

deformation.


Sample

(Fe volume

fraction)

Hardness (±4%)

[GPa]

Dynamic Young modulus (±4%)

[GPa]

Compacted Isothermal Ramp Compacted Isothermal Ramp

Al (0) 0.54 0.52 0.35 76.6 76.9

Al8-Fe2 (0.08) 0.55 0.54 0.37 77.2 72.9 77.4

Al6-Fe4 (0.19) 0.6 0.6 0.38 80 82.1

Al4-Fe6 (0.34) 0.68 0.65 0.52 76.8 74.7 78.2

Al2-Fe8 (0.58) 0.87 0.78 0.71 82.7 76.7 79.1

Al1-Fe9 (0.76) 1.06 1.13 0.98 81.9 84.4

Fe (1) 1.37 1.4 1.42 83.8 81.9

Table 1. Vickers hardness and dynamic Young modulus of ρ=1 samples with different composition

ad after isothermal ageing (2 hours at 473 K) and a 3.33 x 10-2

K/s ramp to 723 K

As a matter of fact this behaviour is more pronounced in samples submitted to a temperature ramp

up to at 723 K, which show the higher hardness difference between the Al matrix and the

reinforcing particles (0.35 and 1.42 GPa). A similar behavior has been observed by Kim et al. [6]

and is in agreement with the literature data [7,8]. It is worth noting that the cross over between the

different defomation behavior correspond to a Fe volume fraction ≈ 0.3, very close to the

percolation thresholds for a mixture of spheres of the same size of the particles, wich occurs at ~

0.29 [9]: this result enforce the model of an Al matrix bearing almost all the deformation when

containing isolated iron particles. Further information regarding microstructure and Young modulus

of compacted powder samples have been obtained from mechanical spectroscopy results.

0 2000 4000 6000 80002

3

4

5

6

7

Q-1

Time (s)

Q-1 (

x 1

0-3)

0.84

0.88

0.92

0.96

1.00

E/E0

E/E

0

Fig. 4. Internal friction and normalized Young modulus for fully (ρ=1) (■) and partially (ρ=0.75)

(○) compacted samples with 0.58 Fe volume fraction during a 0.17 K/s heating ramp up to 473 K

followed by a 2 hours isothermal treatment. Resonance frequency ≈2 KHz. The vertical line

separates the heating ramp from the isothermal stage of the treatment.

Figure 4 reports the internal friction (Q-1

) and normalized Young modulus (E/E0) during a 0.17 K/s

ramp to 473 K, followed by a 2 hours isotherm, for samples with a Fe volume fraction of 0.58, in

the fully (ρ=1) and partially (ρ=0.75) compacted state. Q-1

increases during the ramp reaching a

maximum and then slowly decreases during the isothermal treatment. Q-1

of the ρ=1 sample rises to

a higher Q-1

final value (6.3 x 10-3

), which could be consequence of the higher amount of cold work


and deformation exercised on the sample during compaction. The normalized Young modulus for

ρ=1 samples decreases during the ramp and then remains constant, while for ρ=0.75 samples it

rapidly drops and then gradually rises reaching a constant value. The ρ=1 sample curves (Fig. 4) do

not reveal significant structural changes, whereas at ρ=0.75 the modulus behavior can be linked

with a strengthening of the bonds between particles due to the relaxation of the heavily deformed

contact zones [9]. It is known [2,10,11] that Al powders are covered by a thin oxide layer, a main

factor in hindering the sintering process (also a diffusion controlled mechanism); in partially

compacted powders this layer is only broken in the deformed contact areas between particles.

Heating the sample allows the diffusion of material at the contact points and strengthen the bonding

between powder grains. Measures on Al and 0.34 Fe volume fraction samples, both with ρ=0.8,

shows the same E/E0 behavior but a with a smaller rise; whose presence in Al suggests that the

effect is linked to Al powders, and its intensity decreases as the relative density increase.

300 400 500 600 700 8000

5

10

15

20

25

2nd

2nd

1st

1st

Temperature (K)

Q-1 (

x1

0-3)

0.7

0.8

0.9

1.0

1.1

1.2

ρ=1E

/E0

300 400 500 600 700 8000

5

10

15

20

25

2nd

2nd

1st

1st

Temperature (K)

Q-1 (

x1

0-3)

0.7

0.8

0.9

1.0

1.1

1.2

ρ=0.75

E/E

0

Fig. 5. Internal friction (■) and normalized Young modulus (■) during two successive (1

st, 2

nd )

heating ramps at 3.33 x 10-2

K/s up to 723 K on a fully (ρ=1) and partially (ρ=0.75) compacted

samples with 0.58 Fe volume fraction. Resonance frequency ≈2 KHz

The Q-1

and normalized Young modulus measured during two consecutive ramp at 3.33 x 10-2

K/s

up to 723 K on samples with 0.58 Fe volume fraction and ρ=1 and ρ=0.75, are shown in Fig. 5. The

damping curve of the fully compacted sample show an anelastic relaxation peak at 600 K. The

second heating run measurements show a strong reduction of the background damping and of the

peak relaxation strenght. Structural recovery with concomitant Al grain growth are a possible

explanation of this behavior. Measurements performed at lower frequency (1-10 Hz) (Fig. 6) clearly

disentangle the structural and relaxational processes which are partly superimposed in the curves of

Fig. 5. The relaxation peak (P1) is shifted at lower temperature whereas during the first heating

ramp a structural recovery (P2) is evinced from the strong damping reduction at temperatures higher

than 500 K, with a concomitant inversion of the decreasing modulus trend. During the second

heating ramp remains only the anelastic relaxation peak superimposed on an exponential

background. The relaxation peak P1 can be tentatively associated with that observed on

polycristalline Al [12,13,14]: this is confirmed by the value of the activation energy and pre

exponential factor (HP1=1.66±0.12 eV , τ0=9x10-(17±1)

s) calculated for fully compacted samples

with a 0.58 Fe volume fraction from the shift of the P1 peak temperature with the resonance

frequency (Fig. 7), employing the Arrhenius expression [12].

τ = τ0e(-H/kT)

. (5)

The internal friction curves of different samples show that the peak relaxation strength depends on

composition, showing a linear growth with the Al volume fraction .


300 400 500 600 700 8000

10

20

30

40

50

60

1st ramp

Temperature (K)

Q-1 (

x10

-3)

0.8

1.0

1.2

P2

P1

E/E

0

300 400 500 600 700 8000

10

20

30

40

50

60

Temperature (K)

Q-1 (

x 1

0-3)

0.8

1.0

1.2

2nd

ramp

E/E

0

Fig. 6. Low frequency damping curves of a fully compacted (ρ=1) sample with 0.58 Fe volume

fraction during two successive heating ramps at 3.33 x 10-2

K/s.

Frequency 1 Hz (■) and 10 Hz (○)

Fig. 7. Arrhenius plot for the evaluation of the activation energy (HP1) and the pre exponential

factor (τ0) of the anelastic peak (P1) for samples with 0.58 Fe volume fraction [12]

In order to compare the heating rates, revealing how samples absorb energy from an

electromagnetic induction system, Al-Fe mixtures heating tests were conducted on disk shaped

samples. Figure 8 shows the heating curves of disks with different composition and ρ=0.75,

together with that of a reference steel disk. As previously specified, the experimental set up ensures

that differences in the heating rates only depends on samples composition. The compacted mixtures

with a Fe content below the percolation threshold show a poor heating behaviour and they are

unable to reach temperatures above 350 K. As far as they have such a low energy absorption these

specimens are not suitable to be employed with most commercial elettromagnetic induction

systems. Moreover in these specimens the induced electric currents are too low to match the safety

requirements of said systems. The heating rate of the samples greatly improves for Fe volume

fraction above 0.3 . In the heating tests they reach far higher temperatures ( > 450 K) and show a

behaviour similar to the ferritic steel reference whose curve is almost superimposed with that of the

disk with a Fe volume content of 0.58. The explanation for this glaring difference stems from the

presence above the percolation threshold of Fe particles extended substructures that substains

macroscopic induced currents. These currents from one side satisfy the safety requirements of

1,6 1,8 2,0

1

2

3

4

log

(2ππ ππf)

1000/T (1/K)


commercial induction system, on the other result in a strong energy absorption by the specimen and

therefore a fast heating.

Fig. 8. Heating curves of samples with different Fe volume fraction and ρ=0.75±0.02, together with

a reference ferritic steel sample. All samples were previously submitted to an isothermal treatment

at 473 K for ~104 s.

Conclusion

In conclusion, these tests show that cold compaction of Al and Fe powders followed by a relatively

low temperature ageing can be used as a cost effective and practical way of producing composites

mixing the characteristics of Al with the ferromagnetic properties of Fe. The composites are good

candidates in induction heating applications being lighter and less expensive substitutes of steel.

Mechanical spectroscopy measures show structural changes at working temperatures above 600 K.

Akowledgements:

The authors wish to thank ISTA s.r.l. Italy for its financial support and thechnical assistance.

References

[1] G.O’Donnel and L.Looney: Mater. Sci. Eng. A Vol.303 (2001), p. 292

[2] F.Tang, I. E. Anderson, and S. B. Biner: J. Light Met. Vol.2/4 (2002), p. 201

[3] V. Rudnev, D. Loveless, R. Cook and M. Black, in: Handbook of Induction Heating, edited by

Marcel Dekker, New York (2003)

[4] K.Yamaguchi, N. Takakura and S. Imatani: J. Mater. Process. Tech. Vol.63 (1997), p. 364

[5] L. P. Lefebvre, S. Pelletier and C. Gelinas: J. Magn. Magn. Mater. Vol.176 (1997), p. L93

[6] H.S.Kim: Mater. Sci. Eng. A Vol.289 (2000), p. 30

[7] A.L. Greer, Mater. Sci. Eng. A Vol.304-306 (2001), p. 68

[8] Z.C.Zhong, X.Y. Jiang and A.L. Greer: Phil. Mag. Vol.76 (1997), p. 505

[9] R.Consiglio, D.R.Baker, G,Paul and H.E.Stanley: Physica A Vol.319 (2003), p.49

0 200 400 600 800

300

350

400

450

500

550

Te

mp

era

ture

(K

)

Time (s)

Reference

Fe(0.58)

Fe(0.40)

Fe(0,26)

Fe(0.13)


[10] R.S.Timsit, in: IEEE Transaction on components, Hybrids and Manufacturing Technology, Vo.

CHMT-3, No.1,march 1980

[11] S.Gedevanishvili and S.C. Deevi: Mater. Sci. Eng. A Vol.325 (2002), p. 163

[12] A.S. Nowik and B.S. Berry, in: Anelastic Relaxation in Crystalline Solids, Academic Press,

New York (1972).

[13] X.S.Guan, H.Numakura, and M.Koiwa: J. Physique 3, Vol.6 (1996), p. 219

[14] H.Zuhang and M.Gu: J. Alloys Compd. Vol. 426 (2006), p. 247


Comparison between Roll Diffusion Bonding and Hot Isostatic Pressing production processes of Ti6Al4V-SiC f metal matrix composites

C. Testani1,a, F. Ferraro1,b, P. Deodati2,c, R. Donnini2,d, R. Montanari2,e,

S. Kaciulis3,f and A. Mezzi3,g

1Centro Sviluppo Materiali (CSM), Via di Castel Romano 100, 00128 Rome, Italy

2Department of Mechanical Engineering, University of Rome “Tor Vergata”, Via del Politecnico 1, 00133 Rome, Italy

3Institute for the Study of Nanostructured Materials, ISMN-CNR, P.O. Box 10, 00016 Monterotondo Stazione, Rome, Italy

[email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]

Keywords: Roll Diffusion Bonding, Hot Isostatic Pressing, Ti composites, Microstructure, Mechanical Properties

Abstract. Titanium-metal-matrix composites (Ti-MMC) are materials with very large specific resistance and potential operative temperature up to 800° C. At present these composites are produced by Hot Isostatic Pressing (HIP), a reliable but expensive manufacturing method. To cut production costs, Centro Sviluppo Materiali SpA (CSM) has developed and patented an experimental plant for co-rolling at high temperature sheets of titanium alloy and silicon carbide monofilaments fabrics. The experimental Roll Diffusion Bonding (RDB) pilot plant permits a reduction of process costs of about 40% with respect to the HIP process. This work reports the results of microstructural and mechanical examinations carried out on composites realized by RDB and HIP. The comparison shows that the fibre-matrix interface is stable in both the composites while the mechanical properties of RDB composite are better due to its smaller grain size and high dislocation density.

Introduction

Roll-Diffusion Bonding (RDB) is a process for preparing Ti-MMC reinforced with mono-directional SiC fibres [1-2]. It represents a promising alternative route to Hot Isostatic Pressing (HIP), a well known and reliable manufacturing process [3-6] that, however, requires complex and expensive equipments, sealed steel dies and batch manufacturing approach. A RDB innovative laboratory-equipment has been developed and patented (Patent n° 0001371787 March 2010- Application n° IT2006A000261 - May 2006) at CSM laboratories. A preliminary study showed that it is possible to cut costs up to the 40% with respect the HIP process [1] because RDB involves shorter production-time and does not employ steel dies. This paper reports the results of microstructural and mechanical examinations carried out on composites produced by both the processes to assess the differences of features and properties and to verify whether the quality of RDB composites is comparable or better than that of HIP ones.

RDB Process

RDB method consists in manufacturing Ti-MMC reinforced with mono-directional SiC fibres by means of a semi-continuous approach. The materials used in the process are sheets of Ti6Al4V alloy and SiC long fibres. As shown in Fig. 1, the precursor material is assembled as a lay-up of two metallic sheets with a SiC fabric in the middle.


Fig.1. Precursor material of the Ti-MMC. During the Roll-Diffusion-Bonding process the precursor is put under backward and forward strip-tension and “co-rolled” under controlled pressure; the operation is carried out in an atmosphere of inert gas. The special-design rolling-tools, the inert-gas chamber and the sealing gates of the RDB experimental arrangement are sketched in Fig. 2. Rolling speed, rolling force and mill-stand temperature are controlled by a suitable software to keep the bonding reaction within the process window. Under the combined effects of temperature and pressure the metallic strips undergo a plastic flow through the fibre interspaces with the result of a complete metal-metal and ceramic-metal bonding. The process stages can be summarised as follows:

1. heating-up of the precursor in the rolling chamber; 2. plastic deformation of the metal matrix that flows between the fibres; 3. chemical reactions and atomic inter-diffusion at the fibre-matrix interfaces; 4. contact and joining of the two metal flows in the spaces between the fibres; 5. matrix recrystallization in the joining surfaces.

Further details about the RDB process and experimental equipment are reported in a previous paper [1-2].

Fig. 2. Schematic view of the RDB apparatus.

Ti-foil

SiC-fibres

Rolling

Ar-Chamber

Sealing gate Precursor

Force-roll Work-roll

Sealing gate

Thermal insulator


Pilot equipment design and realisation

As schematically shown in Fig. 2, each rolling-tool consists of an inner force-roll and an external work-roll made of a high-temperature resistant alloy. In order to reduce heat loss, the space between inner and external rolls is insulated with ceramic shields (shaded area in fig. 2). The solution presents the advantage that only the external rings reach high temperatures thus inner work-rolls can be manufactured by cheap tool-steel. The process parameters needed for the design of the pilot-equipment (rolling speed; strain-rate; temperature; pressure etc.) have been defined by means of literature data and experimental hot-compression tests including some pack-rolling tests. One of the main problems has been the large difference of thermal capacity between the precursor and the rolling tools that requested a fine control of the isothermal state in the arc-contact during RDB. A main effort has been devoted to the optimisation and FEM process simulation approach has been very useful to design a set of possible solutions. The FEM model permitted to simulate the possible experimental arrangements, compare the results and select the best solution for the superplastic flow and bonding in the rolling contact arc. The MSC MARC™ code has been used for the thermal simulation of the rolling tools (work-roll and force-roll) during rotation (Fig. 3). The rolling-tool rotation starts when the surface of work-roll reaches a temperature that is sufficient to guarantee the correct process condition in the contact-arc. The core of the plant is the RDB chamber (Fig.4), the operating temperature could be raised up to 1150°C in inert gas (Ar) to avoid any contamination of the Ti-sheets during the RDB process. An oxygen spectrometer is continuously controlling the quality of the atmosphere inside the chamber.

Fig. 3. Simulation of the thermal trend in °C during the counter-clockwise rotation

Fig. 4. The RDB plant core


Experimental

RDB and HIP sheets have been produced with the following process parameters: 1- HIP: T = 920°C, applied pressure 1200 bar for 30 minutes; 2- RDB: T = 920°C, roll-pressure and arc-contact time have been adjusted with an DB-

preliminary testing phase. Samples for observations and analyses were obtained by spark erosion from composite sheets. Optical and electron microscopy observations have been performed on cross-sections after mechanical polishing and etching in Kroll reagent. Auger Electron Spectroscopy (AES) analyses have been done by using an Escalab Mk II spectrometer (VG Scientific, UK) equipped with 5-channeltron detection system [7]. Photoelectrons were excited by using a standard Al Kα excitation source, while Auger electrons were induced by electron gun LEG 200, operated at 10 keV and 1 – 10 nA current. XPS spectra were collected at constant analyzer pass energy of 20 eV, while AES spectra were registered in constant retard ratio (1:2) mode. XRD measurements have been carried out using the Co-Kα radiation (λ = 1.79 Å). Spectra were collected in step-scanning mode with 2Θ steps of 0.05° and counting time of 2 s per step in the angular range 10° - 100°. High precision peak profiles of the most intense reflections of α phase were recorded with 2Θ steps of 0.005° and counting time of 20 s per step. Dislocation density has been determined by half-height line widths, corrected by instrumental broadening. Micro-hardness tests have been carried out on cross-sections of the both composites. Tensile tests have performed from room temperature up to 800 °C in accordance with ASTM E 21 standard. Finally, the anelastic behaviour of the materials has been investigated by mechanical spectroscopy experiments, i.e. internal friction and dynamic modulus measurements. The tests have been carried out on bar-shaped samples using the method of frequency modulation. The VRA 1604 apparatus used in the experiments has been described in detail in [8]. The resonance frequencies were in the kHz range. The samples have been heated from room temperature to 850 °C with a heating rate of 1.7 x 10-2 °C s-1.


The microstructures of RDB and HIP composites are shown in Fig. 5. The grain size of the RDB composite (∼ 10 µm) is smaller than that of the HIP one (∼ 30 µm). In RDB composite defects due to an incomplete metal flow in the interstices between the fibres have been observed near the rim of the sheet (Fig. 5 c). To analyze chemical diffusion around the fibres AES measurements have been carried out in different positions. An example is reported in Fig. 6a. The AES spectra measured in the five points across the fibre/matrix interface are shown in Fig. 6b. Peak intensities demonstrate that C penetrates into the matrix during RDB process. Similar results has been found by examining HIP composite [7] and it is known that C and Ti chemically react to form a layer of TiC which retards further C diffusion toward the matrix preventing interface degradation.


Fig. 5. Structure of RDB a) and HIP b) composites. Close to the rim of the sheet RDB composite exhibits an incomplete metal flow in the interstices between the fibres c).

Fig. 6 . a) SEM image (80 x 80 µm2) of the fibre-matrix interface; the markers (1-5) indicate the points of Auger spectra acquisition. b) Auger spectra measured in the points 1-5.

Fig.7 shows the XRD spectra of the two composites. From the comparison the peaks of RDB composite are much broader and shifted to lower angles.

a) b)

a) b)

c)


Fig.7. XRD spectra of HIP and RDB composites.

The broader peak profile of RDB composite is evident in Fig.8, where the intensities and positions of the {100} reflections have been normalized to make easier the comparison.

Fig. 8. Precision {100} peak profiles of HIP and RDB composites.

For each XRD reflection the total line broadening βT , corrected from instrumental broadening, is basically due to two contributions, the size of coherently diffracting domains (βD ) and the micro-strains ( βε ). βT can be written as:


(1)

where D is the domain size, ε the average micro-strain, ϑ the Bragg angle, λ the X-ray wavelength and K a constant (= 0.89). In the case of Ti and Ti alloys the coherently diffracting domains are the grains. Introducing in Eq.(2) the D values determined for the two composites by metallographic observations the micro-strain ε has been determined. Finally, the dislocation density ρ was calculated by means of the Williamson-Smallman relationship [9]:

ρ = Ξ ε2 / k0 b

2 (2)

where Ξ =16 is a constant, b is the modulus of Burgers vector and k0 ≅ 1 is a factor depending on dislocation interaction. From this calculation the values of ρ = 1.1 x 1012 cm-2 for the RDB composite and ρ = 6.0 x 109 cm-2 for the HIP one have been obtained. These data are in good agreement with the results of mechanical tests which evidence better mechanical properties of RDB composite. In fact, hardness is 500 HV and 323 HV for RDB and HIP composites, respectively. Tensile tests carried out at increasing temperatures up to 800°C (Fig.9) show that RDB composite has higher values of yield and ultimate strength.

a) b)

Fig. 9. Yield strength (a) and ultimate tensile strength (b) of RDB and HIP composites obtained from tensile tests carried out at increasing temperatures.

Dynamic measurements (Fig.10) up to 550°C show that both the composites exhibit higher modulus than the corresponding monolithic alloy; the E value of RDB composite is ∼ 8% higher than that of HIP one. Mechanical spectroscopy experiments provided not only the modulus trends vs. temperature but also information on the anelastic behaviour of the materials. Fig. 11 a) shows Q-1 and (f/f0)

2 vs. T trends of the RDB composite. Dynamic modulus E is proportional to the resonance frequency f :

(3)

where m is a constant (m=1.875), L the length of vibrating reed, h its thickness and ρ the material density. Therefore, (f/f0)

2 represents the variation of E with respect its value at room temperature.

ϑεϑ

λβββ ε tan2

cos+=+=

D

KDT

ρπ

E

L

hmf

2

2

122=


Fig. 10. Dynamic modulus trends of Ti6Al4VAlloy, RDB and HIP composites vs. temperature. The Q

-1 curve shows a peak at about 627 °C superimposed to an exponentially increasing background; in correspondence of the peak the elastic modulus exhibits a change. In tests with different frequencies the peak position changes indicating that the IF peak is a relaxation peak. The corresponding activation energy H = 189 kJ/mol and the relaxation time τ0 = 2 x 10-15 s are very close, inside the experimental error, to those determined for the HIP composite (H = 186 kJ mol-1, τ0 = 2.3 x 10-15 s) [10] thus it is the same IF peak in the two materials. The physical phenomena giving rise to the peak have been extensively discussed in [10]: it was ascribed to stress induced reorientation of interstitial-substitutional (i-s) pairs (C-Al and C-V) in the hcp α phase of the matrix near the fibres according to the mechanism discussed by Gupta & Weining [11] and Povolo & Bisogni [12] for hcp metals. Moreover, comparing the IF curves of the two materials in Fig.11 b) the RDB composite exhibits a higher background. The background is strongly structure-sensitive [13] thus the result can be easily explained by considering the smaller grain size and the higher dislocation density of the RDB composite which provide a greater contribution to the damping.

Fig. 11. a) IF and (f/f0)2 trends of RDB composite. b) Comparison between IF curves of RDB and

HIP composites.

a) b)


The similar anelastic behaviour of the composites, characterized by a Q-1 peak due to the re-

orientation of C-Al and C-V pairs present in the matrix around the fibres, confirms that the structure of the fibre-matrix interface is substantially the same. This is very important because the evolution of mechanical properties mainly depends on the degradation of fibre-matrix interface during in-service life and HIP composite does not modify its properties after long-term exposure to the temperatures foreseen for aeronautical engine applications [14]. Therefore, it is reasonable to expect that RDB composite exhibits behaves in the same way.

Conclusions

The feasibility of RDB process has been demonstrated by realizing a laboratory-pilot-plant at CSM and a preliminary evaluation showed that the costs can be reduced of about 40% with respect to HIP. AES analyses showed that the structure of the fibre-matrix interface is similar in both the materials: a TiC layer has been observed to form around the fibres also in the RDB composite. TiC retards C diffusion toward the matrix preventing interface degradation. The similar structure of the interface is confirmed by the anelastic behaviour which is characterized in both the materials by a Q-1 peak arising from the re-orientation of C-Al and C-V pairs present in the matrix around the fibres. Since the worsening of mechanical properties mainly depends on the degradation of fibre-matrix interface, it is expected that the properties of RDB composite do not change after long-term exposure to the temperatures foreseen for aeronautical engine applications, as previously verified for the composites produced via HIP. With respect to HIP, RDB process gives rise to grains of smaller size and to higher dislocation density in the matrix leading to better mechanical properties. An incomplete metal flow in the space between the fibres has been observed in some RDB sheets close to the rim; to remedy to this drawback the ending part ( 1-2 cm wide) must be removed. Since the mechanical stability is similar, the mechanical properties better and the costs lower, it is concluded that RDB is a valid alternative to HIP. References

[1] C. Testani and F. Ferraro: J. Mater. Eng. Perform., Vol. 19(4) (2010), p.521.

[2] C. Testani and F. Ferraro: Mater. Sci. Forum, Vol. 638-642 (2010), p. 991.

[3] M.E. Tata, R. Montanari, C. Testani and G. Valdrè: La Met. Ital. Vol. 7-8 (2005), p. 43.

[4] P.D. Nicolau, S.L. Semiatin and H.R. Piehler: Scr. Metall. Mat., Vol.32 (1995), p.57.

[5] P. R. Smith and F. H. Froes: J. of Met., Vol.27 (1984), p.19.

[6] H. D. Hanes, D. A. Seifert and C.R. Watts, ‘‘Hot Isostatic Pressing’’, Battelle Press, Columbus, OH, (1979).

[7] R. Donnini, S. Kaciulis, A. Mezzi, R. Montanari and C. Testani: Surf. Interface Anal. Vol. 40

(2008), p. 277.

[8] S. Amadori, E.G. Campari, A.L. Fiorini, R. Montanari, L. Pasquini, L. Savini and E. Bonetti:

Mater. Sci. Eng. A Vol. 442 (2006), p. 543.

[9] G.K.Williamson and R.A.Smallman: Phil. Mag. Vol.1 (1956), p. 34.


[10] P. Deodati, R. Donnini, R. Montanari and C. Testani: Mater. Sci. Eng. A, Vol.521–522

(2009), p.318.

[11] D. Gupta and S. Weining: Acta Metall. Vol.10 (1962), p. 292.

[12] F. Povolo and E.A. Bisogni: Acta Metall. Vol.14 (1966), p. 711.

[13] A.S. Nowick and B.S. Berry, in: Anelastic relaxation in crystalline materials, Academic Press,

New York and London (1972), p. 454.

[14] P. Deodati, R. Donnini, R. Montanari, C. Testani and T. Valente: Mater. Sci. Forum, 604-605

(2009), p.341.


Keywords Index

A

AA2124 85

AA2618/20%Al2O3p 61

AA6061 85

Agglomeration 1

Al2O3 Reinforcement 95

Al2009 Aluminium Alloy 95

Al6061 Aluminium Alloy 95

Alumina 85

Aluminium Matrix Composite 125

Aluminum 85, 115, 135

Anelastic Behavior 23

Anelasticity 135

C

Compacted Powder 135

Composite 75, 115

Compression 75

D

Damage 125

E

Ex Situ Processing 1

F

Fatigue 125

Friction Stir Welding (FSW) 85

G

Gas-Liquid Reaction 1, 115

H

Hardness 135

Heat Transfer Coefficient (HTC) 105

Homogenization 49

Hot Drilling 95

Hot Isostatic Pressing (HIP) 145

I

In Situ Processing 1

Induction Heating 135

Iron 135

L

Light-Weight 1, 115

Linear Friction Welding 85

M

Magnesium 115

Manufacturing Route 1

Matrix Composite 49

Matrix-Fibre Interface 23

Mechanical Behaviour 49

Mechanical Property 145

Metal Composites 49

Metal Matrix Composite (MMC) 61, 85, 95,105

Mg-RE Alloy 75

Microstructural Stability 23

Microstructure 145

N

Nano-Composite 1

Nitridation 115

Numerical Simulation 105

P

Particle 85

Particulate 125

Plasma Electrolytic Oxidation(PEO)

61

R

Roll Diffusion Bonding 145

S

Self Consistent Unit Cell Model 49

SiC Reinforcements 95

Silicon Carbide (SiC) 85

Sliding 61

Solidification 105

Strengthening 75


T

TEM 75

Ti Composites 145

Ti6Al4V-SiCf Composite 23

Tribology 61

V

Vibration 105

W

Wear 61

Authors Index

A

Amadori, S. 135

Apelian, D. 1, 115

B

Bonetti, E. 135

Bonollo, F. 105

Borgonovo, C. 1, 115

C

Cabibbo, M. 75

Campari, E.G. 135

Ceschini, L. 61, 85

D

Della Corte, E. 105

Deodati, P. 23, 145

Donnini, R. 23, 95, 145

F

Ferraro, F. 145

K

Kaciulis, S. 23, 145

Kazemian-Abyaneh, M. 23

M

Martini, C. 61

Mezzi, A. 23, 145

Montanari, R. 23, 145

Morri, A. 85

P

Pasquini, L. 135

R

Reuschel, A. 49

Rotundo, F. 85

S

Sambogna, G. 61

Santo, L. 95

Schmauder, S. 49

T

Tagliaferri, V. 95

Tarterini, F. 61

Testani, C. 23, 145

Timelli, G. 105

U

Ucciardello, N. 23

V

Vedani, M. 125

W

Weber, U. 49

Willert, M. 49

advances in metal matrix composites : proceedings of an international meeting

Documents