This article was downloaded by: [Eindhoven Technical University]On: 22 November 2014, At: 05:09Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Materials and Manufacturing ProcessesPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/lmmp20

Solid-State Joining of Metal Matrix Composites: ASurvey of Challenges and Potential SolutionsTracie Prater aa Vanderbilt University , Nashville, Tennessee, USAPublished online: 22 Apr 2011.

To cite this article: Tracie Prater (2011) Solid-State Joining of Metal Matrix Composites: A Survey of Challenges and PotentialSolutions, Materials and Manufacturing Processes, 26:4, 636-648, DOI: 10.1080/10426914.2010.492055

To link to this article: http://dx.doi.org/10.1080/10426914.2010.492055

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Materials and Manufacturing Processes, 26: 636–648, 2011Copyright © Taylor & Francis Group, LLCISSN: 1042-6914 print/1532-2475 onlineDOI: 10.1080/10426914.2010.492055

Solid-State Joining of Metal Matrix Composites:A Survey of Challenges and Potential Solutions

Tracie Prater

Vanderbilt University, Nashville, Tennessee, USA

This article examines the challenges associated with joining particulate-reinforced metal matrix composites (MMCs). A survey of publishedresearch reveals the problems inherent in fusion welding of MMCs, namely, porosity and the formation of a deleterious theta phase precipitatedby the reaction of molten Aluminum with the reinforcement. The theta phase is absent in joints produced using friction stir welding (FSW), asolid-state process. FSW, however, is not a panacea, as FSW welds of MMCs are characterized by rapid and severe wear of the welding tool.Research efforts to characterize and mitigate wear incurred during this process are also documented.

Keywords Joints/Joining; Metal-matrix composites (MMCs); Wear.

Introduction

An emerging area of friction stir welding (FSW) researchis the joining of metal matrix composites (MMCs), aclass of composite materials valued by the aerospaceindustry for their high strength-to-weight ratio. In 2010,the market share for MMCs grew to 4.9 million kilograms,increasing 6.3 percent over the past five years [1]. Theuse of MMCs in industrial applications is limited by thedifficulties associated with joining MMCs to themselves orother materials. This review details the problems associatedwith fusion welding of MMCs: foremost among theseis the formation of theta phase, a precipitant formedfrom the reaction of the reinforcement material withmolten Aluminum in the surrounding matrix. A solid-state process, which precludes formation of the undesirabletheta phase, has the potential to produce strong jointswith strengths comparable to those of the parent material.Previous literature reviews in this area have focused almostexclusively on fusion welding [2, 3]. This article provides anassessment of past and current research in the field of joiningMMCs using FSW, a relatively new solid-state joiningprocess. Characteristics of MMC joints are presented, andthe problem of tool wear, caused by contact between thetool and the much harder reinforcing particles, is discussedin detail. The review synthesizes research on this topicpublished in the last decade and targets issues that shouldbe addressed in future work.

The FSW process

Patented in 1991 by researchers at The Welding Instituteof Cambridge, England, FSW is a novel solid-state joiningprocess used in applications worldwide. The process, which

Received March 31, 2010; Accepted May 3, 2010Address correspondence to Tracie Prater, Vanderbilt University,

VU Station B, Box 1592, Nashville, TN 37235-1592, USA; E-mail:tracie.j.prater@vanderbilt.edu

occurs below the melting temperature of the joint material,represents a departure from traditional fusion weldingmethods. The FSW process is illustrated in Fig. 1. Inconventional FSW, a rotating tool is plunged into the surfaceof adjoining metal plates. The rotation of the tool generatesheat at the interface, resulting in local plasticization of thematerial due to shear stress. As the tool traverses along thejoint line, the material behind the tool consolidates, forminga welded region with a width roughly corresponding tothe diameter of the tool in contact with the surface [4].Advancing and retreating sides of the weld are definedrelative to the direction of tool rotation. The advancingside is the side of the weld in which the tool rotation isin the same direction as the traverse; the retreating side isthe region where tool rotation and welding direction arein opposition. During a weld, material is swept from theadvancing side and deposited on the retreating side [5].As a relatively nascent technology, much research has

devoted to the effect of process parameters on the quality ofthe finished joint. There are four major process parameterswhich can be varied in FSW: rotation speed, traverse speed,tilt angle, and plunge depth. Rotation speed ��� is in unitsof rotations per minute (rpm) and designates the rate atwhich the tool rotates. Traverse speed ��� indicates thespeed at which the tool traverses the material; � is usuallyspecified in units of inches per minute (ipm). Dependingon the system configuration, the tool may remain stationarywhile the material is advanced at traverse rate � (or viceversa). The selection of � and �, often expressed in terms ofthe weld pitch ratio �

�, largely governs heat input. Excessive

heat input can contribute to the formation of voids inthe joint, while insufficient heat input can result in toolfracture. The establishment of an operating window—a setof parameters which produce acceptable welds—is essentialfor many applications.Friction Stir Welding tools are usually fabricated from

tool steels; a cursory review of the literature suggests thatAISI H13 is the most common tool material. H13 is a strongmaterial (elastic modulus 30,500ksi) with a melting point of

636

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 637

Figure 1.—Schematic of the FSW process [4].

2600F, approximately three times as great as temperatureslikely to be experienced by the tool during the FSW process[6, 7]. While FSW tools made from conventional materialshave a nearly infinite life when used to join Aluminumalloys, they exhibit wear in the welding of harder materialssuch as metal composites and steel.

MMCs

The solid-state nature of the FSW process makes it idealfor joining Aluminum alloys or other materials, such asMMCs, which are difficult to weld using fusion methods.MMCs consist of two separate phases. The continuousphase, termed the matrix, is present in larger quantities.Embedded within the matrix is the reinforcement phase,which consists of abrasive material in the form of fibers orparticles. MMCs are classified according to the materialswhich comprise the matrix and reinforcement phases, theshape of the reinforcement, and the concentration of thereinforcement Aluminum MMCs (metal composites inwhich the matrix is an Aluminum alloy) are categorizedusing a scheme developed by the Aluminum Association[8]. For instance, the classification Al 359/SiC/20p indicatesthat the material is Aluminum alloy 359 reinforced with 20percent Silicon Carbide particulate by volume.A microscopic view of Al 359/SiC/20p appears in Fig. 2.

The Silicon Carbide particles in this material displayan irregular, angular geometry, although some particulatereinforcements may have a spherical or cubic structure.The particles in Fig. 2 are classified as F500 on the FEPAscale; this means that 90% of the particles have a diameterbetween 11.8 and 13.8 In most cases, the particles arerandomly oriented and dispersed throughout the matrix,a condition which gives rise to the term discontinuouslyreinforced Aluminum (DRA). MMCs can also have fibrousreinforcements consisting of long fibers of a harder material.If the aspect ratio, defined as the ratio of the length tothe cross-sectional diameter, is large the reinforcement isclassified as continuous; discontinuous fiber reinforcementsare characterized by lengths and diameters on the sameorder [9]. While both fiber and particulate reinforced MMCsfind applications in industry, particulate reinforcement ispreferred due to its lower cost and ease of manufacture.Additionally, particulate reinforced MMCs are isotropic.

Figure 2.—Microscope image of Al 359/SiC/20p at 100× magnification.

Although the anisotropy of fiber reinforced MMCs canbe exploited to enhance the strength of a structure in aparticular direction, the directional dependence of materialproperties makes them more difficult to characterize andmodel.Though the constituent materials of MMCs are relatively

inexpensive, the price of composites is driven upward bycosts associated with their manufacture. The most commonforming methods for MMCs are liquid metallurgy andpowder metallurgy. In a stir casting method describedby Sahin, reinforcement particles or fibers are introducedinto initially molten aluminum through a pipette [10].Alternately, particulate reinforcement can be inserted intothe solid matrix alloy through drill holes and mixedthroughout once the metal has melted. The mixture ispoured into a cast iron mold and allowed to solidify;hydraulic pressure is applied before billet removal to reducethe incidence of porosity [10]. A more uniform particledistribution can be attained through the powder metallurgytechnique of hot isostatic pressing, which applies heatand pressure concurrently to the elemental matrix andreinforcement to induce consolidation [11], both solid andliquid metallurgy techniques for MMC fabrication requirethe development of specialized equipment. The primaryadvantage of solid metallurgy is that it can produce near-netshape parts. While liquid metallurgy yields a more uniformreinforcement distribution, significant machining time isrequired to transform the part into the desired shape.The degree of enhancement in mechanical properties

attained with use of a metal composite (such asincreased strength, temperature resistance, and hardness)is determined by the percentage reinforcement. Since thematerial property data available for MMCs is limiteddue to both their diversity and categorization as specialtymaterials, mechanical properties are estimated using theLaw of Mixtures. Commonly used in calculations involvingcomposite materials, the Law of Mixtures expresses a givencomposite material property as a sum of weighted averages,where the weighting factors are the fraction of matrix andreinforcement present in the material:

Xc = xmvm + xrvr � (1)

Xc corresponds to the mechanical property of the composite(such as density or elastic modulus); xm and xr and denote

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

638 T. PRATER

Table 1.—Comparison of mechanical properties for Al-MMCs andcorresponding unreinforced Aluminum alloys [1].

Material E (MPa) �y (MPa) Elongation at failure (%)

Al 6061 68,000 330 18Al 6061/Al2O3/20p 97,000 360 4Al 7005 72,000 325 12Al 7005/Al2O3/10p 84,000 345 7

the value of this property for the matrix and reinforcementmaterials, respectively; vm and vr and represent the volumefractions for the matrix and reinforcement (for MMCs whichcontain only two phases, vm + vr = 1). The expression in(1) is also valid for weight fractions (volume fractions canbe converted to weight fractions using density) [9]. Table 1,adapted from Pirondi et al. [1], compares the mechanicalproperties of two composites (Al 6061/Al2O3/20p and Al7005/Al2O3/10p) with their unreinforced counterparts. Theelastic modulus and yield stress are increased by additionof particulate reinforcement (as Eq. (1) predicts, the degreeof this enhancement is proportional to the volume ofreinforcement). The enhancements in E and �y observedin the composites are attributed to the inclusion of thereinforcing particles, which function as a load-sharingmechanism [1]. The decrease in the elongation at failure isrelated to the reduced ductility and higher wear resistanceof the composite.It is apparent from Eq. (1) that MMCs are in many

respects “customizable”: that is, we can design a compositewith a desired property value by carefully selecting thematrix and reinforcement materials and controlling theproportions in which they are present. In the case ofparticulate reinforcement, which is usually added to theAluminum alloy as a powder, the elastic modulus ofthe alloy can significantly be augmented with only anominal increase in weight. Kunze and Bampton reportthat the elastic modulus of conventional Aluminum can beincreased by 300% through the addition of SiC particulatereinforcement present in a volume fraction of 70% [11].However, while the increase in particle concentrationsubstantially enhances the strength of the alloy, it makesmachining more difficult, as cutting through the particulatesresults in rapid and severe wear of the cutting tool. Asdemonstrated by the work of Gugger, diamond tools haveshown great potential in overcoming the problem of toolwear in machining metal composites [12].As a result of the challenges and costs incurred

in machining MMCs, their use is currently relegatedto applications where the added strength is critical toperformance (and thus justifies the additional investment).Some “weight-saving” structures in which MMCs havebeen successfully implemented include the Space ShuttleOrbiter’s structural tubing, the Hubble Space Telescope’santenna mast, control surfaces and propulsion systems foraircraft, tank armors, and braking systems for roller coasters[11, 13]. The potential applications of MMCs are equallyvaried, ranging from nuclear fuel containers to auxiliarycomponents for satellites [11].

Problems with fusion welds of MMCs

An additional barrier to the widespread use of MMCsis the difficulty encountered when joining them to otherMMCs or unreinforced materials in a larger structure.When MMCs are welded together using fusion methods,the quality of the joint is degraded by solidificationand chemical reactions, effects documented by Stojohannet al. and Ellis in [14]. During fusion welding, thedisparity in densities between the matrix and reinforcementmaterials results in particle segregation; this “clumping”of the reinforcement creates nonuniformity in the joint.Additionally, the higher viscosity of the composite materialinhibits material flow, potentially impacting the weld stressdistribution and contributing to a reduction in strength [15].The most problematic aspect of fusion welding

Aluminum MMCs is the formation of a deleterious thetaphase induced by the reaction of molten Aluminum withreinforcement particles. In the case of Silicon Carbidereinforcement, liquid Aluminum reacts with SiC to produceAluminum Carbide (Al4C3�. Stojohann et al. indicate thatthis reaction, initiated when the weld temperature exceedsthe melting point of Aluminum, is highly dependent ontemperature. The amount of Al4C3 present in the completedjoint (visible in the microscope image of Fig. 3 as needle-like formations) is directly proportional to the differencebetween the weld temperature �Tw� and the melting pointof the matrix �Tm�.Stojohann et al. assessed the propensity for Al4C3

formation and particle degradation in three different fusionwelding processes: gas tungsten arc (GTA), electron beam(EB), and laser beam (LB). Two composite materials wereconsidered, one with Al2O3 particulate reinforcement andanother with SiC fiber reinforcement. Optical evaluationof the microstructures in the welded Al2O3 reinforcedcomposites indicated degradation of the reinforcing particlesfor each of the processes considered, a result of thedecomposition of Al2O3 reinforcing particles to gas inthe presence of molten Aluminum [14]. The clumpingof particulates, observed for GTA and EB welds, createsporosities in the joint. For the SiC reinforced composites, theSiC whiskers remain intact, but the Al4C3 phase is presentin all fusion welds. Though the amount of Al4C3 producedcan be somewhat mitigated through careful control of heatinput (since the reaction which governs the formation ofAl4C3 is accelerated by an increase in temperature above

Figure 3.—Aluminum carbide in the fusion zone of a laser-beam weld of SiCreinforced Aluminum [14].

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 639

Figure 4.—Reorientation of SiC whiskers in FSW of Al–SiC composite [14].The left SEM image shows the whiskers as they appear in the base metal,while the right image captures the whiskers in their reoriented state in theweld nugget.

Tm), Al4C3 is still prevalent in the EB welds, the fusionprocess with the lowest heat input per unit length. ThusStorjohann concludes that the formation of the Al4C3 phaseis unavoidable in fusion welding of Al–SiC composites.Due to the undesirable chemical reactions which take

place in the composites when the matrix alloy is melted, asolid-state welding method may yield better results. Sincethe weld temperature in FSW is less than the melting pointof the matrix alloy �Tw < Tm�, the undesirable chemicalreactions which lead to the dissolution of reinforcementin Al–Al2O3 and the formation of Al3C4 in Al–SiC areunable to proceed. Storjohann used FSW to produce defect-free welds in both the Al–Al2O3 and Al–SiC composites.Although intermittent particle clustering was observed inthe Al–Al2O3, the particle distribution was more uniformthan that found in fusion welded joints. While the numberof particles present in the parent material and the joint is onthe same order (as the Al2O3 will not dissolve in the absenceof liquid Aluminum), the particle distribution in the joint isskewed toward smaller diameter particles. This outcome isa sign of particle breakage, an effect Storjohann attributesto the brittleness of the reinforcement and its inability to“flow” with the surrounding Aluminum alloy.In the samples of Al–SiC composite joined by FSW, none

of the deleterious Al3C4 phase was detected. Additionally,no breakage of the SiC whiskers was observed, owing totheir small size and ability to reorient themselves in the flowfield of plasticized Aluminum. In the parent material, thewhiskers are aligned along the x direction; in the joint theirorientation is primarily in the z direction, the axis coincidentwith the direction of welding (Fig. 4). Storjohann postulatesthat this change in orientation may prove favorable to themechanical properties of the joint, though further researchis required to substantiate this claim.

Features of MMC joints produced using FSW

As documented by Storjohann et al., FSW is the preferredmethod for joining metal composites primarily because itprecludes formation of deleterious phases within the joint.There have been a number of published studies that assessthe characteristics of MMC joints produced using FSWprocesses. These distinguishing features are enumeratedbriefly in the following list:

1. Reduction of reinforcement particle size in weld zone.When compared with the shape of the particles in theparent material, particles in the weld region have smallerdiameters and are round, rather than angular, in shape[16]. This size reduction, which may be as large as

Figure 5.—Comparison of SiC particles in base metal Al 6061/Al2O3/20p(left) and FSW joint (right) [1].

20%, suggests that the brittle particulate reinforcementfractures in the wake of the stresses that accompanyplastic flow of the metal matrix [17]. These changes insize are particularly prominent in the weld nugget, thezone coincident with the largest amount of heating andplastic deformation [18]. The size reduction and dispersalof particles is evident in Fig. 5, which compares SEMimages of the particles in the parent material and jointfor Al 6061/Al2O3/20p [1]. The particles in the parentmetal are broken up during FSW, creating a weld regionwith smaller particles that are more densely packed [19].

2. Changes in particle distribution. In addition to roundingthe reinforcement particles, stirring of the metalcomposite in FSW induces their rearrangement. Inthe base metal of a discontinuously reinforced metalcomposite, there exist aggregates of reinforcement whichare significantly larger than the median-sized particle;thus the particle size distribution (plotted as a histogramof frequency versus size) will be skewed toward largerparticle sizes. These large particle conglomerates aretransformed into a homogeneous distribution through theparticulate rounding and scattering effects of FSW [20].A comparison of the pre- and post-weld particle sizedistributions for Al2O3 particulate reinforcement revealsthat the Al2O3 particles are broken up and redistributedin the weld by the stirring action of the tool. The abrasiveparticles in the nugget are of similar size and shape(aspect ratio), an attribute that is not observed in the basemetal. The redistribution of particles induced by FSWis evident in specimens of Al 6061/Al2O3/10p analyzedusing electron backscatter diffraction (EBSD) as partof the work of Root, Field, and Nelson [19]. As thereinforcement is broken up by FSW, the smaller particlesmigrate toward the center of the weld on the advancingside. Imaging software reveals that welding reduces thevolume fraction on the retreating side from 10% to9.22%; meanwhile the volume fraction in the center ofthe weld has increased to 11.18% [19]. Particles in theweld center are rounded, smaller, and closer together, adistribution that characterizes as optimal for a weldedcomposite, since it enhances grain refinement. Fusionwelding techniques do not alter the particle distributionin this manner.

3. Grain refinement. The phenomenon of grain refinement,well documented in FSW of Aluminum alloys, isparticularly pronounced in FSW of Al-MMCs. While theprecise mechanism by which grain refinement occurs isa topic of scholarly debate, it is generally agreed that the

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

640 T. PRATER

effect is closely related to the dynamic recrystallizationof the plasticized material which consolidates behindthe tool to form a weld. In FSW of metal composites,grain refinement is enhanced by the reinforcing material,as the reinforcement particles (which are absent inconventional metal alloys) provide additional nucleationsites to facilitate grain growth. The inverse relationshipbetween number of nucleation sites and grain size canbe exploited in MMCs to reduce the size of the grainsin the finished joint [17]. Humphrey has developed amodel which predicts the recrystallized grain size, D,from the volume fraction of the composite �F�� andthe initial average diameter of the reinforcing particles�d� [21]. According to Humphrey’s empirically derived

expression d = F−2s

� , the grain size in the weld nuggetdecreases with increasing reinforcement percentage, aprediction that is consistent with the nucleation sitehypothesis proposed by Ceschini et al. [17]. Based on thisrelationship, The presence of reinforcing particles standsto improve the efficiency with which recrystallizationoccurs.According to Root, Field, and Nelson, the enhancedgrain refinement in composites is directly related to thechanges in the particle distribution brought about byFSW, an effect they call “particle stimulated nucleation”[19] Since the particles in the weld nugget are smaller andmore abundant in the as-FSW MMC (an increase createdby the break-up of large particles and the stirring of theresulting smaller fragments into the center of the weld),there are more particles available to serve as nucleationsites. This gives rise to a grain structure that is morerefined than either that observed in the parent materialor the as-FSW Aluminum alloy.

4. Hardness. There is some disagreement in the publishedliterature regarding the hardness of the weld zone inMMC joints produced using FSW. Marzoli et al. observea hardness profile similar to that found in Al 6XXXalloys (Fig. 3), with a minimum hardness value in theheat-affected zone [22]. The hardening of the HAZ whichoccurs in fusion welding, a consequence of reinforcementparticle migration toward the center of the joint line, isnot detected in FSW joints by Marzoli. Similar findingsare reported by Ceschini et al., who record an 11 to25% drop in hardness in the weld zone as comparedwith the base material [18]. However, other researchershave documented a hardening of the HAZ by as muchas 30%, an increase evident in the hardness profiles ofFig. 6 [23]. Though this hardening appears to be ananomalous result, Shindo et al. attribute the discrepancyto their use of an as-cast Al–SiC composite, which hasdifferent characteristics than the Al–Al2O3 compositesused in the other studies. Shindo claims that the rise inhardness of the weld zone is more pronounced for slowertraverse rates. Additional variations in hardness coincidewith wear of the tool .

5. Joint Efficiency and Weld Envelope. Joint efficiency,defined as the ratio of the yield strength of the weldspecimen to that of the parent material, is highlydependent on process parameters. Most studies on FSWof MMCs report joint strengths in the range of 70 to

Figure 6.—Comparison of hardness profiles for FSW of Al 359/SiC/20p fora range of travel speeds [23].

80% [1, 17, 18]. Feng, Xiao, and Ma were able to obtainjoint efficiencies exceeding 100% in transverse tensionand longitudinal tension [16]. Interestingly, they foundthat heat treating the specimens to a T4 condition, atechnique frequently used in FSW joints of unreinforcedalloys to improve their strength, has the opposite effectin as-FSW MMCs. As documented by Root, Field, andNelson, FSW refines the grains in the MMC beyondthat observed in the parent material [19]. The highlyevolved grain structure in the finished MMC jointmeans there is little to be gained by heat treatment.In contrast, the unreinforced as-FSW joint, where thegrain structure is not nearly as refined, can be improvedupon significantly by heat treatment. Feng, Xiao, and Maspeculate that heat treatment may facilitate formation oftheta phase at the particle-matrix interface, which in turndegrades the strength of the joint [16]. Comparison ofjoint efficiencies for welds produced at various processparameters is used to define a set of parameters whichproduce acceptable joints, sometimes referred to as theweld envelope. The process envelope for the FSW of Al6061/Al2O3/20p is shown in Fig. 7 [22]. Compared withthe weld envelope for unreinforced Aluminum alloys,the range of operating parameters available for FSW ofmetal composites is much narrower [16]. The compositeis in some ways less forgiving than its unreinforcedcounterpart, as the presence of the brittle reinforcingparticles inhibits plasticized flow of the matrix alloy [16].This decreased metal flow contributes to the formationof voids that can negatively impact the tensile strength.It may be possible to counteract the tendency for voidformation in MMCs by implementing more complex toolgeometries that enhance the vertical flow of material.The rounding of the particles and the more uniformparticle distribution created by the FSW processincreases the ductility of the weld specimen, achange which translates into an improvement in tensileproperties. This claim is supported by the computersimulations conducted by Marzoli et al., which foundthat distributions with smaller, similarly-sized particlesdemonstrate increased ductility [22]. Joints with these

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 641

Figure 7.—Weld envelope for FSW of Al 6061/Al2O3/20p. The trianglesdelineate a process limit, below which no acceptable welds were produced.The circles denote welds with satisfactory joint efficiencies. Squares indicatemechanical limits of the welding apparatus [22].

distributions are able to withstand higher tensile stresses.The high joint efficiencies, which comprise the weldenvelope for MMCs, thus correspond to parameterswherein the material is stirred sufficiently to create afavorable particle distribution.

6. Temperature. MMC materials possess a high specificheat and resistance to temperature, properties whichmake the plasticization of the material required for FSWmore difficult. To achieve plastic flow in MMCs, the toolmust impart additional heat to the material at the startof the weld. The higher heat input necessitated by thetemperature resistance of MMCs is evident in Fig. 8, aplot of thermal cycles for six locations along the weldjoint in FSW of Al 6061/Al2O3/20p [22]. Although thisneed for a “dwell” period in which the tool is stationarybut rotates to generate heat for plasticization is presentin conventional materials, it is of longer duration inMMCs, since more heat is required to induce flow of thematrix alloy. Based on the operating envelope developedby Marzoli (Fig. 7), it appears that larger weld pitches(consistent with higher heat input) produce joints withimproved properties.

Figure 8.—Temperature profiles recorded in FSW of Al 6061/Al2O3/20p.Each curve corresponds to a thermocouple location (shown in the inset) [22].

7. Surface characteristics. The surface appearance of anAluminum joint made with FSW is characterized by apolished, mirror-like finish. MMC joints have a coarsersurface structure, presumed to be a consequence ofdecohesion between the matrix alloy and reinforcementat the surface of the joint [15]. Tools made of hardermaterials with a lower lubricity, such as Boron Carbide�B4C�, produce a smoother joint surface due to slippageat the interface between the workpiece and the tool [15].

8. Fatigue. Although data on fatigue life of as-FSW MMCjoints is sparse, in general the fatigue life of the weldedMMC specimen is lower than that observed in the parentmaterial [17, 18, 24]. This reduction in fatigue strengthis most prevalent in samples with void defects: whensubjected to high strain-amplitudes, these specimensfail prematurely, a consequence of the high stressconcentration present in the vicinity of the void [17].The reduced fatigue life in specimens that do not containvoids can be attributed to the disturbances to the basematerial caused by the FSW process [18].Cavaliere, Rossi, Sante, and Moretti used a novelevaluation technique known as thermo-elastic stressanalysis (TSA) to assess the fatigue behavior in as-FSWMMCs [24]. TSA relates the temperature variationson the surface of the specimens during loading to thestresses, effectively exploiting the relationship betweenthermal energy and mechanical deformation. Cavaliereet al. are thus able to create 3-D stress maps ofeach specimen at various stages in the load cycling.Examination of the samples post-fracture reveals thatthe site of initial failure (and also the origin of thepropagating crack) is most often located at an interfacebetween a reinforcing particle and the matrix alloy.These interfaces function as stress risers and are readilyapparent in the plots of the stress contours. Both theparticle/matrix interfaces and the presence of voidscontribute to a reduction in the fatigue life of the as-FSWMMC. Overall the as-FSWMMCs exhibit classicalfatigue behavior, with cycles to failure decreasing as theapplied load is increased.

9. Ductility and failure mechanisms. As discussed inthe section on joint efficiency, the ductility of metalcomposites is reduced by the presence of brittlereinforcement particles. This decreased ductility isevident in the mechanical response of the specimento tensile stress; elongation in the weld is slightlyless than that of the base material [25]. However,FSW welds of composites exhibit increased ductilitycompared to fusion welded specimens of the samematerial, owing to the rounding of the reinforcementparticles induced by the FSW process and the enhancedductility of the particle distribution in the FSW joint.Tensile failure in FSW samples is consistent with ductilefracture originating in the weld’s thermo-mechanicallyaffected zone (TMAZ). This ductile fracture mechanismis indicative of the good plasticity properties of the jointand the refined, recrystallized grain structure createdby FSW [25].

According to Ceschini et al., failure of an FSW metalcomposite can generally be attributed to one of three causes:

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

642 T. PRATER

1. Cracks, originating in large reinforcement particles thathave not been broken up or rounded by the stirring actionof the FSW tool, propagate and increase the specimen’ssusceptibility to fracture.

2. Decohesion of the matrix and the reinforcement particles,often apparent in the weld’s surface structure (coarsesurfaces are characteristic of decohesion).

3. Presence or growth of voids. As mentioned previously,voids introduce stress concentrations which cancompromise the mechanical properties of the specimen(most notably fatigue life). Voids which are initiallypresent can increase in size as a result of decohesion,eventually leading to fracture [18].

The cause of failure, when discernable, can inform selectionof process parameters which produce joints with morefavorable properties. For instance, an increase in weld pitch���� results in increased stirring of the material, an effect

which may reduce the formation of the voids that are oftenresponsible for specimen failure.

Wear of FSW tools in the joining of MMCs

A final characteristic of MMC joints made using FSWis the presence of debris products in the weld specimendeposited along the joint line as the tool wears. Wearof the tool in the FSW process is a phenomenon uniqueto high-strength alloys, having been documented in theliterature only for FSW of steels and metal composites[18]. Prado et al., who used FSW of Al 6061 as a baselinefor comparison with tool consumption in MMCs, reportno detectable wear for FSW of conventional Aluminumalloys [26]. The wear of tools in composite materials israpid and severe enough to be discerned with the nakedeye. Figure 9 shows the shape changes observed in a steelFSW probe for successive welds of Al 6061/Al2O3/20p.After only two welds (1.82m) the initially threaded toolresembles a smooth cylindrical probe with rounded tip.Tool material eroded as a result of contact with hard

reinforcing particles is deposited along the joint line.This abraded material can be detected and quantified byperforming an energy dispersive X-ray (EDX) analysison a transverse cross-section of the weld [16]. EDX candetect the presence of particles originating from the toolas well as reveal the regions in which these erosionproducts are concentrated. The results of Feng, Xiao, andMa show an increased iron concentration in the weld nugget,

Figure 9.—Images from Prado et al. showing shape changes in steel FSWtool for two successive welds of Al 6061/Al2O3/20p. Parameters are 500RPMand 1mm/s, and each weld was 0.76m in length [27].

an indication of a rise in wear in regions with greaterconcentrations of reinforcement. Thus this observationcorroborates the prevailing theory that contact between thetool and the reinforcement particles is the primary cause ofwear in FSW of metal composites.Although EDX provides useful metrology information,

destructive methods (i.e., sectioning of the joint) arerequired to obtain a sample suitable for analysis. Thereare a number of alternative methods that analyze wearbased on post-weld examination of the tool rather thanthe welded specimen. The easiest to implement of thesemethods compares the weight of the tool before and afterwelding; the weight lost by the tool is equal to the amountof material eroded. Unfortunately, sensing tool wear viamass comparison is complicated by the Aluminum thataccumulates on the tool during welding. Although thisAluminum residue can be etched away using a base solution,weighing is regarded as a crude method for this applicationdue to the concurrent acquisition and deposition of materialinherent in FSW [8]. Mechanical gauging methods, such asthe use of calipers to record changes in probe length anddiameter, have similar limitations [8]. Gauging techniquesare largely insensitive to the nuanced shape changes thatoccur as the tool wears (Fig. 9). In contrast, measurementsystems which rely on optical comparisons of the tool aregenerally regarded as more reliable indicators of wear [8].Lienert et al. use a shadowgraph technique to quantify thepercentage of the tool probe consumed during welding. In anovel hybrid of weighing and optical comparator methods,Fernandez and Murr measure wear by capturing close-upphotos of the tool probe, cutting out the probe in theseimages, and comparing the weights of the cutouts. Theassumption that the weight of the two-dimensional imagecutout is indicative of the tool’s material loss is substantiatedby a series of parallel experiments which calculate percentwear by comparing masses of the etched tool after eachweld [28].Many studies of wear in FSW of composite materials

are exclusively concerned with wear of the probe. Thesusceptibility of the shoulder to wear is a subject of somedebate. Liu et al. measured the shoulder diameter betweensubsequent welds but found the differences to be negligible[29]. A conclusive indicator of shoulder wear would be anincrease in the probe length after welding due to erosionof tool material at the probe/shoulder interface, an effectthat is observed by Lee et al. [15]. However, since wearof the shoulder is not observed in any other publishedresearch, it may be the case that the outcome documentedin [15] is an anomalous result, possibly attributable to anexcessive plunge depth used to increase heat input andensure plasticization of the temperature-resistant workpiecematerial.

Variation of wear with process parameters

Once a robust method of wear measurement has beenestablished, the variation of wear with process parameterscan be investigated. The wear experiments detailed inreferences [26–31] relate the volumetric wear of the probe tothe process parameters rotation speed ���, travel speed ���,and distance welded �l�. Though the studies utilize different

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 643

metal composite materials and tool geometries, there aresome trends that seem to hold in general for FSW of MMCs.An obvious direct proportionality between wear and linearweld distance is reported in all studies. In particular, Liuet al. notes the dramatic reduction in the diameter of theprobe with increasing weld length [29]. Lee et al. determinethe average reduction in probe diameter to be 0.010 inchesper foot of weldment, degradation substantial enough tonecessitate replacement of the probe after approximately 5feet of weld [15]. It is clear from the plots of percent toolwear versus weld distance displayed in Figs. 18 and 19 thatthe variation of wear with distance is nonlinear. In fact,there is usually some critical distance, xcr , beyond which noadditional wear is observed. The significance of this plateauin the wear curve observed for values of l greater than xcris discussed in Section 7.The dependence of wear on rotation speed and traverse

rate is less palpable. To isolate the effect of rotation speedon wear, Fernandez and Murr performed welds at a fixedtraverse rate while varying the spindle speed from 500RPMto 1000RPM [28]. As evidenced by the wear curves plottedfrom their data (Fig. 10), percent tool wear increases withincreasing rotation speed. This trend is also observed byPrado et al., who find that wear increases with rotation speedup to 2000RPM [27]. To characterize the dependence ofwear on traverse rate, Fernandez and Murr expanded theirexperimental matrix to include variations in traverse speeds.The wear curves for the seven combinations of traverseand rotation speeds considered are plotted in Fig. 11. Bycomparing wear data for parameter sets with the samerotation speed but different traverse rates, it is apparentthat wear decreases with increasing traverse speed. Thoughthe inverse relationship between wear and traverse rate isnonintuitive, it provides experimental evidence that toolwear in the FSW process is a shear, rather than drag,phenomenon.While the data cited above is by no means a complete

characterization of the variation of wear with processparameters, it indicates that material loss in FSW of metal

Figure 10.—Plot of wear versus weld distance for 500, 750, and 1000RPM[27].

Figure 11.—Plot of wear versus weld distance for a range of rotation andtraverse rates [28].

composites is inversely proportional to traverse rate anddirectly proportion to the rate of rotation [30]:

W ∝ �

�� (2)

The symbol W denotes percent wear, the complement ofthe ratio of the tool volume after welding to the originaltool volume expressed as a percent (for the Fernandez/Murrdata, differences in the weights of the photo cutouts areregarded as indicators of volumetric wear). Based on thefunctional dependence of wear on � and � indicated in (2),slow rotation speeds and high traverse rates are paramountto the minimization of wear. It is thus not surprisingthat of the parameters selected for the Fernandez/Murrstudy, the least amount of tool wear was observed inthe 500RPM/6mm/s and 500RPM/9mm/s cases, as thevalues of the parameters in these cases correspond to themaximum traverse speed (11mm/s) and minimum rotationspeed (500RPM) considered. Although Fernandez and Murrdo not interpret their results in terms of the weld pitchratio �

�of Eq. (1), it is apparent from Fig. 19 that lower

weld-pitch conditions result in significantly reduced wear.Unfortunately, minimization of weld pitch is not a viablesolution to tool wear when weld quality is a concern:Crawford et al. argue that the formation of wormholes ismore likely to occur at low weld pitches because of thelack of adequate flow [32]. The competing effects of defectformation and rapid wear discussed suggest that controllingwear through the selection of process parameters may be adifficult prospect.The process parameters �, �, and l are not the

only factors that influence wear. A study conductedat NASA Marshall Space Flight Center by Diwanfound that the amount of wear incurred in weldinga metal composite is inversely proportional to thecomposite’s percentage reinforcement. Categorizing thematerials based on reinforcement concentration, Diwan

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

644 T. PRATER

ascertains that category IV and V composites (which haveSiC reinforcement percentages of 40 and 55, respectively)are “unweldable” using conventional tool materials [33].Diwan’s findings are similar to those of Lee et al., whodetermine that the production of satisfactory welds usingsteel tools is limited to materials with a reinforcementpercentage no greater than 25% [15].

Rate of wear and evolution of the self-optimized

shape in FSW of metal composites

It is evident from the wear curves of Figs. 10 and 11 thatthe rate of wear (recorded in units of percent effective wearper centimeter) varies with process parameters � and � aswell as the distance welded. In general, the initial wear rateof the tool decreases with increasing traverse rate, a trendshown in the plot of Fig. 12 [23]. In the Fernandez/Murrdata (Fig. 11), the wear curve for the maximum traversespeed considered (11mm/s) has a slope of zero (indicatingthat no wear of the tool occurs) until the weld distanceexceeds 150cm [28]. For parameters with greater valuesof �

�, the wear rate is most rapid at the outset of welding,

an effect that is particularly pronounced in Fig. 11 for the1000RPM/1mm/s and 1000RPM/3mm/s cases. Prado et al.postulate that this increased initial wear rate is related to thetemperature of the workpiece material: because the materialis colder at the start of the weld, higher forces are requiredto move the tool through the workpiece [26]. Similar resultsappear in reference [28], wherein the maximum wear rateis always recorded during the early phases of welding.Stellwag and Lienert document an analogous increase ininitial wear rates for FSW of steel (approximately 8% of thetotal wear occurs during the initial plunging and dwellingstages), a rise they also attribute to greater material flowstress at reduced temperatures [8].There are some parameters for which the wear curve

plateaus with increasing weld distance. The termination ofwear with some critical distance xcr is clearly visible in the1000RPM/6mm/s curve of Fig. 11, where wear effectively

Figure 12.—Plot of initial wear rate versus traverse rate (rotation speed isconstant at 1000RPM). The plot combines data from Prado et al. [27] andShindo et al. [23].

Figure 13.—Evolution of self-optimized shape for 1000RPM, 9mm/s case.The leftmost image (0m welded) is the original probe geometry; the middleimage shows the probe after a weld distance of 1.52m; the right imagecorresponds to the self-optimized shape (attained after 3.96m of weld) [23].

ceases for distances greater than 150cm. This criticaldistance is not observed for wear curves correspondingto parameters with comparatively large values of �

�(such

as 1000RPM/1mm/s and 1000RPM/3mm/s in Fig. 11).Shindo et al. contend that the high initial wear ratesconsistent with these parameter sets preclude the existenceof a critical weld distance, as material loss during the initialweld phase is substantial enough to make the steady-stateplateau associated with zero wear unattainable [23].The shape of the tool that is present at the cessation of

wear is referred to as the self-optimized shape [23, 28]. Thesequence of images in Fig. 13 shows the evolution of a probefrom its original state to self-optimization [23]. During theinitial stage of the weld, Fernandez and Murr claim that theworkpiece material “fills” the threads, inhibiting the wearmechanism and delaying the onset of wear. However, thethreads are eventually eroded by the abrasive reinforcementparticles, producing the smooth frustum-like shape seen inthe rightmost image of Fig. 13. The figure also demonstratesthat radial wear along the length of the probe is nonuniform.For the 1000RPM/9mm/s, maximum radial wear occurs atthe root of the probe, while the diameter near the shoulderof the probe remains virtually unchanged (indicating thatlittle to no wear occurs in this region). Liu et al. also studiedthe variation of radial wear with distance, ultimately findingthat different regions of the probe experience different ratesof wear [29]. The amount of wear recorded at a particularlocation additionally depends on the process variables, anobservation that is apparent in Fig. 14, which compares self-optimized shapes for two of the parameter sets consideredby Shindo et al. [23]. These images, which correspond tothe zero wear rate regimes in the wear curves of Fig. 11,clearly illustrate that the optimized shape varies significantlywith welding speed. The left image (1000RPM/6mm/s)exhibits the pseudo-hourglass shape characterized by Liuet al., wherein the maximum wear occurs at a location 1/3of the distance between the root and the end of the probe.Using the results of Liu et al., and defining x as the lineardistance from the shoulder and l as the length of the pin, theprobe can be divided lengthwise into three distinct sectionscorresponding to low, moderate, and high wear, as follows:

i) 0 < x < l/3. Little to no radial wear of the probeoccurs in the vicinity of the shoulder �x ≈ 0�, but wearincreases with increasing distance from the shoulder x.

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 645

Figure 14.—Comparison of images of self-optimized probe shapes taken fromShindo et al. [23]. 1000RPM/6mm/s (distance welded 3.1m) is on the left;1000RPM/9mm/s (distance welded 3.96m) is on the right.

ii) x ≈ l/3. This is the region of maximum wear—for theoptimized shape, the probe diameter is smallest in thisregion.

iii) l/3 < x < l/3. The amount of wear decreases fromx ≈ l/3 to x = l. While the diameter of the probe in thisregion is greater than that in (ii), it is slightly reducedfrom the diameter in (i). Thus the wear regime for thisrange of x values is classified as moderate.

This hourglass shape described by Liu et al. is morepronounced for parameters that experience comparativelygreater wear (such as the 1000RPM/6mm/s in Fig. 14).Not surprisingly, parameters resulting in less wear producemore subtle changes in shape. There is no existing literaturethat explains the observed shape variations in terms of aphysical wear mechanism. Additionally, the current researchon optimization of tools in FSW of metal composites offerslittle explanation as to why the self-optimized shape isattainable for some parameters but not others.A research question which naturally arises from the

studies presented in references [23, 26–31] concerns whateffect, if any, the variations in tool shape accompanyingwear impose on the quality of the weld. The resultsof hardness tests of welds in the wearless regime (theplateau regions of select wear curves in Fig. 11) indicate asignificant increase in the hardness of the HAZ [26]. Thisrise in strength associated with the use of a self-optimizedprobe is consistent with a homogeneous weld zone formedin the absence of a wake effect. The gradual disappearanceof threads during the course of welding reduces the verticalflow of material around the tool; once the tool has reacheda self-optimized shape, the threads have been completelyremoved, and there is little to no circulating flow of material,a result manifested in the hardness measurements of Pradoet al. [27]. Additionally, when the probe has reached theoptimized condition, the onion structure characteristic of thesurface of an FSW joint is less apparent [28]. These changesin the weld attributes can be ascribed to differences in theflow fields for the initially threaded geometry and the self-optimized shape, illustrated in Fig. 15 [28]. Although thevortex flow is absent in the self-optimized shape, frictionat the tool/workpiece interface generates sufficient heat toproduce acceptable welds [26]. Shindo et al. contend that theability of the self-optimized tool to yield satisfactory weldspecimens renders the use of featured tools in FSW of metal

Figure 15.—Comparison of flow regimes for threaded probe (left) and probewith self-optimized shape (right) [27].

unnecessary. Fernanzdez and Murr agree that the inclusionof features is inefficient, as they will eventually be erodedby the abrasive action of the particles in the workpiece [28].However, since some vortex flow is desirable to reducethe likelihood of void formation, Prado et al. suggest thatthe ideal tool design for this application is a smooth probedesigned with a minimum amount of features to mitigateporosity [26]. Prado et al. also advise the use of toolsmachined to closely resemble the self-optimized shape.Stellwag and Lienert contend that the wear of the

tool seems to have no adverse effect on the mechanicalproperties of the weld [8]. This claim is called intoquestion by Feng et al., who conclude based on tensile teststhat the presence of erosion products in the joint resultsin deterioration of mechanical properties [16]. It is alsopossible that the reduced vertical flow associated with theself-optimized shape facilitates the growth of voids that cannegatively impact joint strength. When wear of the toolbecomes severe enough to compromise weld quality, thetool must be replaced. Replacement may prove a costlyventure when the tool geometry is complex and/or there isa large amount of material to be welded. For these reasons,considerable research has focused on reduction of wear inFSW of MMCs.

Combatting wear in FSW of metal composites

Preservation of the original tool geometry eliminates costsassociated with the manufacture of replacement tools andrenders concerns about reduction in tensile strength relatedto porosity obsolete. Assuming that the mechanism whichunderlies the wear in this process is abrasive, the mostobvious approach to combatting it is to select a tool materialharder than the reinforcement particles that cause abrasion(this plan of attack is founded upon the principle of abrasivewear, as defined by Rabinowicz [34]). PolycrystallineBoron Nitride (PCBN), B4C, and diamond are suitable toolmaterials for this application, having hardness values greaterthan the ceramics (Al2O3 and SiC) frequently used as thereinforcing phase in metal composites. Figure 16 presents arough visual comparison of the hardness values for severalof these materials [29].The selection of a wear-resistant tool material is fraught

with challenges. While materials such as PCBN, WC, anddiamond are harder than the reinforcing particles (and thusshould guard against wear), they are both expensive anddifficult to machine. Many of these materials are brittle andwould thus fracture under the high forces associated withFSW [13]. The problem of fracture in more exotic tool

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

646 T. PRATER

Figure 16.—Chart used to convert between Vickers and Rockwell C hardnessscales. The plot can be used to compare the hardness values for tool materialstypically used in FSW (denoted by the shaded square corresponding to steelsand metals) with the hardness of the reinforcement phase in MMCs (usuallyeither Al2O3 or SiC). It is apparent that CBN and diamond, which havehardness values greater than the reinforcement particles, should demonstrateimproved wear resistance in FSW of metal composites [35].

materials, such as Molybdenum, for FSW applications isdocumented by Prater [36]. In an effort to improve ductility,a brittle material can be alloyed with a less brittle material,resulting in a new material with mechanical properties thatare more suitable for FSW applications. Two commonalloys of this type are WC-Co and Mo-Rh; the addition ofthe comparatively more ductile Cobalt and Rhenium limitsthe propensity for fracture and improves machinabilitywithout significantly reducing material hardness.Because it is expensive to fabricate the entire tool from

harder, wear-resistant materials, researchers often rely oncoatings of substrates made of materials that are cheaper andeasier to manufacture. Lee et al. used an h13 steel tool with aB4C coating to join Al 6092/SiC/17.5p [15]. Unfortunately,the tool did not demonstrate increased wear resistance, asthe B4C coating delaminates from the h13 substrate afteronly a short weld distance. Once the coating has wornaway, the coarse joint surface characteristic of decohesionbetween matrix and reinforcing particles reappears. Weldsmade using the coated tool show a marginal improvement inmechanical performance; for the same process parameters,the ultimate tensile strength (UTS) associated with thecoated tool is 61.9ksi, a value approximately 13% greaterthan the UTS measured for the uncoated tool [15]. Thisslight improvement in UTS can be explained in termsof Feng’s hypothesis, which argues that the presence oferosion products negatively impacts joint strength [16]. Inthis scenario, the hard B4C coating functions as a barrierbetween the tool and abrasive particles, affording a delay inthe onset of wear. This in turn prevents the accumulation oferoded tool material, which can reduce the UTS. Lee et al.intimate that, although the B4C coating marginally improvesjoint strength, its tendency to delaminate under FSW

Figure 17.—Functionally Gradient Material (FGM) used by Lee et al. [15].

conditions makes it ineffective from a cost perspective.While coatings are commonly employed in other machiningprocesses to mitigate wear, their use in FSW of metalcomposites is largely unexplored. Lee et al. note that theuse of diamond as a tool base material or coating may yieldmore promising results and suggest that further investigationon this subject is needed [15].Another option for reducing wear involves alteration of

the workpiece material. Lee et al. investigated replacingconventional metal composites, which have a uniformdistribution of reinforcement throughout, with FunctionallyGradient Materials (FGM), a type of composite that has ahigh percentage of reinforcement everywhere except nearthe joint line [15]. This reduced presence of reinforcementsurrounding the weld joint limits the tool wear, whilethe large concentration of reinforcement present at allother locations in the material preserve the enhancedstrength associated with MMCs (Fig. 17). The gradientin the reinforcement is created via infiltration casting: aceramic perform material (Saffil) comprises the 0.25” edgesections of the plate. Lee et al. were able to successfullyjoin Aluminum 2195 alloy to FGM composites with SiCreinforcement percentages of 5, 18.5, and 27 near the edgeof the plates and 50% elsewhere.Wert [37] proposes the joining of metal composites to

monolithic alloys as a possible means of wear mitigation.This method is advantageous because it does not requirealteration of the composite (i.e., the reinforcement gradientwhich is created in FGM) and limits the use of themetal composite to regions where increased strength ismost critical. There is thus the potential for a threefoldreduction in cost: there are lower costs associated withthe production of the MMC (as no special modificationto the material is required), the amount of material thatmust be purchased (since the MMC is used in less of thestructure), and replacement of tooling (if the joining ofMMCs to monolithic alloys is shown to reduce wear). Whilethe method posited by Wert is attractive for economicalreasons, it proves exceedingly difficult to implement owingto the dissimilarity in material properties between the metalcomposite and the unreinforced alloy. In order to softenthe MMC, which has a substantially higher flow stress, theheat input to the weld must be increased. As a result, MMCto monolithic alloy joints are characterized by overheatingand the formation of undesirable eutectic phases containingMn, Si, Cu, and Al [37]. Additionally, the placement ofthe MMC on the advancing or retreating side of the weldhas a great effect on the surface morphology. Since stirringharder material into softer material is easier than stirringsofter material into harder material, it is recommended that

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

SOLID-STATE JOINING OF METAL MATRIX COMPOSITES 647

Figure 18.—Comparison of transverse cross-sections for welds of dissimilarmaterials (MMC and Al 2024). In (A), the metal composite is on the advancingside; in (B), the composite is on the retreating side. The morphology in (B)indicates little to no mixing of the materials [37].

the comparatively harder MMC be placed on the advancingside. Positioning the composite on the retreating side withrespect to the tool rotation produces joints with cross-sections characteristic of insufficient mixing (Fig. 18). Theeffect of joining MMCs to conventional Aluminum alloyson tool wear has yet to be investigated; at the time ofthis writing, there are no studies demonstrating that FSWof MMCs to Aluminum reduces the rapid consumption oftool material documented in welds of similar compositematerials.

Summary and conclusions

This review provides an assessment of the problemsinherent in the joining MMC. As documented by Storjohannet al., fusion welding of these materials leads to formationof a deleterious theta phase, precipitated by the reactionof melted aluminum with the reinforcement phase of thecomposite. While the accumulation of theta phase in thejoint can be somewhat mitigated through careful control ofheat input, a solid-state joining process such as FSW is amore viable option. Since FSW occurs below the meltingpoint of the workpiece, the reaction producing the thetaphase in fusion-welded MMC joints is unable to proceed.The absence of theta phase produces MMC joints thatare stronger than their fusion welded counterparts. Thestirring action of the FSW tool results in a redistribution ofreinforcement particles, producing what Pirondi et al. [1]describe as the “optimal microstructure,” a characterizationthat reflects a refined grain structure exceeding that presentin the parent material. FSW of MMCs is complicated byrapid and severe wear of the tool, a result of contact betweenthe tool and the much harder reinforcement particles thatgive the material its enhanced strength. The amount ofwear incurred by the tool in FSW of MMCs is thoughtto depend on several factors: the hardness of the toolrelative to the reinforcement material in the composite,the composite’s percent reinforcement, and the processingparameters. Most studies on wear have focused on the latter.In general, the total amount of material removed from thetool is in direct proportion to the rotations speed of the

tool and the length of the joint but inversely proportionalto traverse rate. There is a need for additional statisticalprocess modeling to characterize the dependence of thewear on both process parameters and material properties,similar to work done by Davim and others to characterizewear in cutting and turning of particulate-reinforced MMCs[38, 39]. The results of such studies could point theway toward a fundamental, physics-based model of thewear process in FSW of MMCs. An understanding ofthe wear mechanism is essential for determining effectivemethods to combat wear. The development of wear-resistanttools is necessary to make solid-state joining of MMCscost effective by eliminating the expenses associated withconsumable tooling. The ultimate goal of research intoFSW of MMCs is twofold: to produce repeatable, robustwelds while simultaneously minimizing the frequency oftool replacement. The ability to weld these materials usingFSW could expand their application to a wide range ofstructures where weight reduction is of critical importance.

Acknowledgment

The author would like to thank Dr. Art Nunes of NASAMarshall Space Flight Center, as well as Drs. George Cook,Alvin Strauss, and Jim Davidson of Vanderbilt University.

References

1. Pirondi, A.; Collini, L.; Fersini, D. Fracture and fatigue crackgrowth behavior of PMMC friction stir welded butt joints.Engineering Fracture Mechanics 2008, 75, 4333–4342.

2. Ellis, M.B.D.; Gittos, M.F.; Threadgill, P.L. Joining Aluminum-based metal matrix composites. Materials World 1994, 8,415–417.

3. Ellis, M.B.D. Joining of Al-based metal matrix composites: Areview. Materials and Manufacturing Processes 1996, 11, 45–66.

4. Mishra, R.S.; Ma, Z.Y. Friction stir welding and processing.Materials Science and Engineering: Research Reports 2005, 50,1–78.

5. Threadgill, P.L. Terminology in friction stir welding. Science andTechnology of Welding and Joining 2007, 12, 357–360.

6. MatWeb Material Property Data. 2009. MatWeb. Available athttp://www.matweb.com. Last accessed October, 2009.

7. Threadgill, P.L.; Leonard, A.J.; Shercliff, H.R.; Withers, P.J.Friction stir welding of aluminum alloys. International MaterialsReviews 2009, 54, 49–93.

8. Stellwag, W.L.; Lienert, T.J. EWI. Friction Stir Welding ofAluminum Metal Matrix Composites Progress Report; Columbus,2001.

9. Matthews, F.L.; Davies, G.A.; Hitchings, D.; Soutis, C. FiniteElement Modelling of Composite Materials and Structures;Woodhead Publishing Ltd.: Cambridge, 2003.

10. Sahin, M. Joining of stainless-steel and aluminium materialsby friction welding. The International Journal of AdvancedManufacturing Technology 2009, 41, 487–497.

11. Kunze, J.M.; Bamptom, C.C. Challenges to developing andproducing MMCs for space applications. Journal of the Minerals,Metals, and Materials Society 2001, 53, 22–25.

12. Gugger, M.D. Machining of DRA-MMCs: Yesterday, today, andtomorrow. Journal of Advanced Materials 2001, 33, 7–12.

13. DWA Aluminum Composites. 2009. DWA-DRA. Available athttp://www.dwa-dra.com. Last accessed October, 2009.

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14

648 T. PRATER

14. Storjohann, D.; Barabash, O.M.; Babu, S.S.; David, S.A.;Sklad, P.S.; Bloom, E.E. Fusion and friction stir welding ofaluminum metal matrix composites. Metallurgical and MaterialsTransactions: A: Physical Metallurgy and Materials Science2005, 36A, 3237–3247.

15. Lee, J.A.; Carter, R.W.; Ding, J. National Aeronautics and SpaceAdministration. Friction Stir Welding of Aluminum Metal MatrixComposites (MSFC Center Director’s Discretionary Fund FinalReport, Project No. 98-09); GPO: Washington, 1998.

16. Feng, A.H.; Xiao, B.I.; Ma, Z.Y. Effect of microstructuralevolution on mechanical properties of friction stir weldedAA2009/SiCp composite. Composites Science and Technology2008, 68, 2141–2148.

17. Ceschini, L.; Boromei, L.; Minak, G.; Morri, A.; Tarterini, F.Effect of friction stir welding on microstructure, tensile, andfatigue properties of the AA7005/10vol.%Al2O3p composite.Composites Science and Technology 2007, 67, 605–615.

18. Ceschini, L.; Boromei, I.; Minak, G.; Morri, A.; Tarterini,F. Microstructure, tensile and fatigue properties ofAA6061/20vol.%Al2O3p frictions stir welded joints. CompositesPart A: Applied Science andManufacturing 2007, 38, 1200–1210.

19. Root, J.M.; Field, D.P.; Nelson, T.W. Crystallographic texture inthe friction-stir-welded metal matrix composite Al 6061 with 10volume percent Al2O3. Metallurgical and Materials TransactionsA 2009, 40A, 2109–2114.

20. Lee, W.-B.; Lee, C.Y.; Kim, M.K.; Yoon, J.I.; Kim, Y.J.; Yoen,Y.M., Jung, S.B. Microstructures and wear property of friction stirwelded AZ91 Mg/SiC particle reinforced composite. CompositesScience and Technology 2006, 66, 1513–1520.

21. Humphreys, F.J.; Hatherly, M. Recrystallisation and RelatedAnnealing Phenomena; Elsevier Ltd.: Oxford, 2004.

22. Marzoli, L.M.; Strombeck, A.V.; Dos Santos, J.F.; Gambaro,C.; Volpone, L.M. Friction stir welding of an AA6061/Al2O3p

reinforced alloy. Composites Science and Technology 2006, 66,363–371.

23. Shindo, D.J.; Rivera, A.R.; Murr, L.E. Shape optimization fortool wear in the friction stir welding of cast Al 359-20% SiCreinforcement. Journal of Materials Science 2002, 37, 4999–5005.

24. Cavaliere, P.; Rossi, G.L.; Di Sante, R.; Moretti, M.Thermoelasticity for the evaluation of fatigue behavior of7005/Al2O3/10p metal matrix composite sheets joined by FSW.International Journal of Fatigue 2008, 30, 198–206.

25. Cavaliere, P.; Cerri, E.; Marzoli, L.; Dos Santos, J. Frictionstir welding of ceramic particle reinforced Aluminium basedmetal matrix composites. Applied Composite Materials 2004, 11,247–258.

26. Prado, R.A.; Murr, L.E.; Soto, K.F.; McClure, J.C. Self-optimization in tool wear for friction-stir welding of Al 6061+

20%Al2O3 MMC. Materials Science and Engineering 2003, 349,156–165.

27. Prado, R.A.; Murr, L.E.; Shindo, D.J.; Soto, K.F. Tool wearin the friction-stir welding of aluminum alloy 6061 + Al2O3:a preliminary study. Scripta Materialia 2001, 45, 75–80.

28. Fernandez, G.J.; Murr, L.E. Characterization of tool wearand weld optimization in the friction-stir welding of castaluminum 359 + 20% SiC metal matrix composite. MaterialsCharacterization 2004, 52, 65–75.

29. Liu, F.J.; Feng, J.C.; Fujii, H.; Nogi, K. Wear characteristicsof a WC-Co tool in friction stir welding of AC4A + 30%volSiCp composite. International Journal of Machine Tools andManufacture 2005, 45, 1635–1639.

30. Prater, T.; Strauss, A.M.; Cook, G.E.; Machemehl, C.; Sutton, P.;Cox, C. Statistical modeling and prediction of wear in friction stirwelding of a metal matrix composite (Al 359/SiC/20p). Journalof Manufacturing Technology Research 2010, 2, 1–13.

31. Prater, T.; Cook, G.E.; Strauss, A.M.; Davidson, J.Parameterization of Friction Stir Welding of Al 6061/SiC/17.5p.International Conference on Trends in Welding Research, PineMountain, GA, June 2–8, 2008.

32. Crawford, R.; Cook, G.E.; Strauss, A.M.; Hartman, D.A.;Stremler, M.A. Experimental defect analysis and force predictionsimulation of high weld pitch friction stir welding. Science andTechnology of Welding and Joining 2006, 11, 657–665.

33. Diwan, R. National Aeronautics and Space Administration.Investigation of Friction Stir Welding of Al Metal MatrixComposite Materials; Huntsville, AL: 2002.

34. Rabinowicz, E. Friction and Wear of Materials; New York: JohnWiley & Sons, 1957.

35. Estimated Hardness Equivalents between Vickers andRockwell C. Gordon England Surface Engineering Forum. 2008.Gordon England Company. Oct. 2009 http://www.nasa.gov/centers/marshall/pdf/104835main_friction.pdf

36. Prater, T. Friction Stir Welding of Metal Matrix Composites:The Joining of Al 6061/SiC/17.5p Using Diamond-Coated Tools;VDM-Verlag: Germany, 2009.

37. Wert, J.A. Microstructures of friction stir weld joints betweenan aluminium-base metal matrix composite and a monolithicaluminium alloy. Scripta Materialia 2003, 49, 607–612.

38. Muthukrishnan, N.; Paolo Davim, J. Optimization of machiningparameters of Al/SiC-MMC with ANOVA and ANN analysis.Journal of Materials Processing Technology 2009, 209,225–232.

39. Basavavarajappa, S.; Chandramohan, G.; Paolo Davim, J. Somestudies on drilling of hybrid metal matrix composites based onTaguchi techniques. Journal of Materials Processing Technology2008, 196, 332–338.

Dow

nloa

ded

by [

Ein

dhov

en T

echn

ical

Uni

vers

ity]

at 0

5:09

22

Nov

embe

r 20

14