advances in metal matrix composites : proceedings of an international meeting
TRANSCRIPT
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.
4 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 5
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
6 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 7
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
8 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 9
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.
10 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 11
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
12 Advances in Metal Matrix Composites
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].
Lorella Ceschini and Roberto Montanari 13
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.
14 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 15
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
16 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 17
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].
18 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 19
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).
20 Advances in Metal Matrix Composites
[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.
Lorella Ceschini and Roberto Montanari 21
[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.
22 Advances in Metal Matrix Composites
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.23
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
24 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 25
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)
26 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 27
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).
28 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 29
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)
30 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 31
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)
32 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 33
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.
34 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 35
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)
36 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 37
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=
38 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 39
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
40 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 41
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.
42 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 43
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
44 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 45
[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:
46 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 47
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.49
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
50 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 51
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
52 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 53
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. %
54 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 55
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
56 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 57
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
58 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 59
[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
60 Advances in Metal Matrix Composites
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.61
(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
62 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 63
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
64 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 65
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
66 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 67
(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
68 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 69
(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
70 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 71
(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
72 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 73
[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).
74 Advances in Metal Matrix Composites
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-
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.75
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
.
76 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 77
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)
78 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 79
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)
80 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 81
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.
82 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 83
[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).
84 Advances in Metal Matrix Composites
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,
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.85
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
86 Advances in Metal Matrix Composites
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.
Results and Discussion
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
Lorella Ceschini and Roberto Montanari 87
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
88 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 89
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
90 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 91
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
92 Advances in Metal Matrix Composites
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).
Lorella Ceschini and Roberto Montanari 93
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
[email protected], [email protected], [email protected]
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.
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.95
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
96 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 97
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%
98 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 99
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.
100 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 101
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.
102 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 103
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.
104 Advances in Metal Matrix Composites
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
[email protected], [email protected], [email protected]
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.105
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)
106 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 107
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.
Results and discussion
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.
108 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 109
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’
110 Advances in Metal Matrix Composites
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,
Lorella Ceschini and Roberto Montanari 111
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
112 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 113
[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.
114 Advances in Metal Matrix Composites
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.115
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
116 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 117
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.
Results and Discussion
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.
118 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 119
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.
120 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 121
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.
122 Advances in Metal Matrix Composites
- 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.
Lorella Ceschini and Roberto Montanari 123
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
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.125
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.
126 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 127
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)
128 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 129
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)
130 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 131
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
132 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 133
[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
134 Advances in Metal Matrix Composites
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],
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
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.135
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
136 Advances in Metal Matrix Composites
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.
Results and discussion
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
Lorella Ceschini and Roberto Montanari 137
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.
138 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 139
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 .
140 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 141
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)
142 Advances in Metal Matrix Composites
[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
Lorella Ceschini and Roberto Montanari 143
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.
© (2011) Trans Tech Publications, Switzerlanddoi:10.4028/www.scientific.net/MSF.678.145
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
146 Advances in Metal Matrix Composites
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
Lorella Ceschini and Roberto Montanari 147
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.
Results and discussion
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.
148 Advances in Metal Matrix Composites
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)
Lorella Ceschini and Roberto Montanari 149
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:
150 Advances in Metal Matrix Composites
(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=
Lorella Ceschini and Roberto Montanari 151
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)
152 Advances in Metal Matrix Composites
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.
Lorella Ceschini and Roberto Montanari 153
[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.
154 Advances in Metal Matrix Composites
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
156 Advances in Metal Matrix Composites
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