Effects of CuO Content on the Wetting Behavior and MechanicalProperties of a Ag–CuO Braze for Ceramic Joining

Jin Yong Kim,* John S. Hardy,* and K. Scott Weil*,w

Pacific Northwest National Laboratory, Richland, Washington 99352

A silver-based joining technique referred to as reactive airbrazing has been recently developed for joining high-tempera-ture structural ceramic components of the type used in gasturbines, combustion engines, heat exchangers, and burners. Itwas found that additions of copper oxide to silver exhibit atremendous effect on both the wettability and joint strengthcharacteristics of the subsequent braze relative to polycrystallinealumina substrates. The effect is particularly significant at lowcopper oxide content, with substantial improvements in wettingobserved in the 1–8 mol% range. The corresponding strength ofthe brazed polycrystalline alumina joints appears to be max-imized at a copper oxide content of 8 mol%, with a maximumroom temperature flexural strength approaching that of mono-lithic alumina. While further increases in oxide content lead toimproved wetting on polycrystalline alumina, the effect on jointstrength is deleterious. It appears that the formation of acontinuous brittle copper-based oxide layer along the interfacebetween the braze and alumina faying surface is responsible forthe poor mechanical behavior observed in joints fabricated withhigher copper oxide content brazes.

I. Introduction

AS the operating temperatures of advanced power generationequipment, such as gas turbines, combustion engines, high-

temperature heat exchangers, and burners, continue to bepushed upward by thermal efficiency considerations, there isan ever increasing need to develop materials suitable for theseapplications, particularly for use under oxidizing conditions.Ceramics are attractive because of their excellent high-tempera-ture mechanical properties and their high level of wear andcorrosion resistance. What limits their usefulness, however, isthe current inability to economically manufacture large orcomplex-shaped ceramic components that exhibit reliable per-formance. One alternative is to fabricate small, simple-shapedparts that can be assembled and joined to form a larger, morecomplex structure. Although a considerable research effort hasbeen directed toward developing various methods of ceramicjoining, intrinsic to each technique is some form of trade-off interms of joint properties, ease of processing, and/or cost.

Glass joining, for example, is a cost-effective and relativelysimple method of bonding ceramics. However, the maximumtemperature to which a glass joint may be exposed is limited bythe softening point of the glass. Additional complications arise ifglass devitrification occurs during service, as its thermomecha-

nical properties will begin to deviate from the original carefullyengineered state.1 An alternative joining technique, diffusionbonding, is conducted at high temperatures and under highpressures. Because of the pressure requirement, however, com-ponents fabricated by diffusion bonding are typically limited tosimple shapes. Reaction bonding, also a high-temperature join-ing process, often yields joints that contain residual porosity,unconverted reactants, and undesired secondary product phases,any of which can reduce joint strength by acting as sites forcrack initiation.2 Joints formed by converting a polymericprecursor to the final ceramic bonding phase often experiencecracking during processing because of the large volumetricshrinkage that accompanies pyrolysis. The use of a ceramicfiller material can partially mitigate this problem, but the jointoften retains a significant amount of porosity, which reducesits strength and reliability.3 Active metal brazing4–6 requires astringent firing atmosphere, either high vacuum or reducing-gasconditions, to prevent the active species, typically titanium, frompre-oxidizing. This represents a high capital expense and higheroperating costs relative to air-fired processes. In addition, recentstudies on the oxidation behavior of active metal brazes haveshown that they are unreliable at temperatures beyond 5001C, atwhich point they eventually oxidize completely, conferring littleor no strength to the joint.7,8

With the exception of diffusion bonding, all of these joiningtechniques rely on an intermediary material to bond the twofaying surfaces. In order to deliver a high-integrity joint, theintermediary and joining surfaces must achieve a chemicalequilibrium, that is, form a chemical bond across the intermedi-ary-to-ceramic interface. Otherwise, only weak Van der Waalsbonding is possible. Thus, one of the conditions necessary forretaining joint integrity under operating conditions is that thereaction products formed at these interfaces must remain stable.In an effort to satisfy these conditions and offer a joiningtechnique, which, like glass joining, can readily be carried outin air, an alternative reactive brazing approach was developed.The objective of this technique, referred to as reactive airbrazing (RAB), is to reactively modify one or both oxide fayingsurfaces with an oxide compound that has been at least partiallydissolved in a noble metal solvent, e.g. silver, gold, or platinum,such that the newly formed surface is readily wetted by theremaining molten filler material. Conceptually, the technique issimilar to the metallization of ceramic surfaces by Cu–Cu2Omelts,9 except that in the present method, the joining operationis conducted in air without the use of an inert cover gas andtherefore the final joint is expected to be resistant to oxidation atmoderate-to-high temperatures.

One system that appears to be readily suited for RAB is thepseudobinary CuxO–Ag system, shown in Fig. 1. Equilibrium-phase studies conducted by Shao et al.10 indicate that there aretwo invariant points in the CuO–Ag phase diagram from whichpossible braze compositions could be developed: (1) a mono-tectic reaction at 9641731C, where CuO and a liquid L2 coexistwith a second liquid-phase L1 at a composition of 30.65 mol%Ag, and (2) a eutectic reaction at 9321731C, where CuO and Agcoexist with the liquid-phase L2 at a composition of 98.6 mol%Ag. Extending between these three-phase reaction points is a

2521

Journal

J. Am. Ceram. Soc., 88 [9] 2521–2527 (2005)

DOI: 10.1111/j.1551-2916.2005.00492.x

r 2005 The American Ceramic Society

B. Derby—contributing editor

Presented at the 105th Annual Meeting of the American Ceramic Society, Nashville,TN, April 28, 2003 (Basic Science Division, Paper No. AM-S14-4-2003).

This work was supported by the U.S. Department of Energy, Office of Fossil Energy,Advanced Research and Technology Development Program. The Pacific NorthwestNational Laboratory is operated by Battelle Memorial Institute for the United StatesDepartment of Energy (U.S. DOE) under Contract DE-AC06-76RLO 1830.

*Member, American Ceramic Society.wAuthor to whom correspondence should be addressed. e-mail: scott.weil@pnl.gov

Manuscript No. 20266. Received September 15, 2003; approved March 07, 2005.

two-phase liquid miscibility gap. In addition, in a series of sessiledrop experiments conducted in inert atmosphere, Meier et al.11

demonstrated that the contact angle between silver and aluminais greatly reduced by small additions of copper oxide. Theimprovement in wetting apparently results because of: (1) anincrease in the oxygen activity of the melt and (2) the formationof an interfacial compound, CuAlO2.

12,13 Based on these stu-dies, Schuler et al.14 recognized that the copper oxide–Ag systemcould be exploited to bond ceramics in air, reporting theirfindings on a 1 mol% copper oxide–Ag braze compositionused to join alumina substrates for power semiconductor packa-ging. In the study described below, CuxO–Ag compositions nearthe monotectic end of the phase diagram were used as a startingpoint for developing a reactive air braze for alumina. Threetypes of experiments were performed in this investigation: (1)sessile drop measurements to determine the wetting behavior ofthe different copper oxide–Ag braze compositions on polycrys-talline alumina, (2) analysis of the braze/alumina interfacialregions within as-formed and exposure-tested joints to deter-mine the nature of wetting and to identify changes in the brazebecause of high-temperature oxidation, and (3) four-point bendtesting of as-brazed joints to measure joint strength at roomtemperature.

II. Experimental Procedure

(1) Materials

The polycrystalline alumina (Al-23, Alfa Aesar Inc., Ward Hill,MA) used in this study was approximately 98% dense and99.7% pure, containing a small amount of silicate material.Alumina discs measuring 50 mm in diameter and 6 mm highwere used in the wetting and initial joining experiments. Thediscs were polished on one face to a 10 mm finish usingsuccessively finer grit diamond paste, cleaned with acetone,and rinsed with propanol, air-dried, and finally heated in staticair to 6001C for 4 h to burn off any residual organic contam-ination. The alumina used for four-point bend testing wereplates measuring 100 mm� 25 mm� 4 mm. They were joinedalong the long edge using one of the experimental brazes to forma 100 mm� 50 mm� 4 mm plate from which the bend speci-mens were cut.

As listed in Table I, eight different braze compositions wereselected based on the pseudobinary CuxO–Ag phase diagramshown in Fig. 1.9 These compositions were formulated bydry mixing the appropriate amounts of copper powder (99%,2.5 mm average particle size; Alfa Aesar Inc.) and silver powder

(99.9%, 0.75 mm average particle diameter; Alfa Aesar Inc.)using a mortar and pestle. The copper oxidizes in situ, formingcopper oxide as the braze is heated. For the wetting studies, themixtures were cold pressed into pellets measuring approximately7 mm in diameter and 10 mm in height. The pellet densitiesaveraged B65% of the theoretical density based on a rule ofmixtures calculation for the dry starting materials. To preparebraze pastes for the joining experiments, a liquid polyvinylbutyral binder (BX-18, Ferro Corp.15) was added to the drypowder mixture in a 1:5 weight ratio.

(2) Testing and Characterization

Wetting experiments were performed in a static air box furnace,furnished with a large quartz window on the front door throughwhich the heated specimen could be observed. A high-speedvideo camera equipped with a zoom lens was used to record thewetting specimen during an entire heating cycle. Each brazepellet was placed on the polished side of an alumina disc andheated at 301C/min to 9001C, at which point the heating ratewas reduced to 101C/min for the subsequent heat treatment. Thefurnace temperature was raised to 10001C, where the tempera-ture was held for 15 min, and then increased again to 11001Cand held for 15 min. In this way, the contact angle between thebraze and alumina was allowed to stabilize for measurement ateach temperature. Using VideoStudio6t (Ulead Systems Inc.,Torrance, CA) software, selected frames from the videotapewere converted to computer images, from which the wettingangles between the braze and alumina substrate were measuredand correlated with the temperature log for the heating run.

Joining samples were prepared by spreading a thin amount ofthe braze paste on the faying surface of each alumina part. Abead of excess paste was allowed to remain along the side of thejoint to flow back and fill in as the binder burned out and thepowders melted. Disc samples were fixtured within a steel springclip, which kept the joint under compression partway throughthe heating cycle, up to B5001–6001C. The following heat-treatment schedule was used in fabricating all but one of thejoining specimens: heat in static air at 51C/min to 10001C, holdat 10001C for 1/2 h, and cool to room temperature at 51C/min.CA80 was heated to 11001C, as it was determined that this brazedoes not melt completely at the lower temperature. Once joined,the discs were cut in half using a diamond saw and cleaned withacetone, followed by a propanol rinse and air-drying. One halfof each joining specimen was set aside for metallographicpreparation and analysis, while the other half was subjected toexposure testing. Testing was conducted in a static air mufflefurnace held at 8001C for 100 h to simulate prototypical serviceconditions. After exposure under these conditions, the sampleswere mounted and polished for comparative metallographicanalysis.

Plate-shaped samples for fracture specimen fabrication wereheld together by the weight of one of the alumina plates restingon the other. Spring steel side clips and appropriately positionedrefractory brick kept the arrangement from slipping or topplingduring heating. The braze paste was applied to the fayingsurface of each piece and the parts were brought together andjoined using the same procedure described above for the disc

Fig. 1. Pseudobinary CuxO–Ag phase diagram.6

Table I. Braze Compositions Used in This Study

Braze ID CuO content (mol%) Ag content (mol%)

CA80 80 20CA69M 69.35 (monotectic composition) 30.65CA60 60 40CA08 8 92CA04 4 96CA02 2 98CA01E 1.4 (eutectic composition) 98.6CA01 1 99

2522 Journal of the American Ceramic Society—Kim et al. Vol. 88, No. 9

specimens. Once joined, each sample was machined into rectan-gular bend bars measuring 4 mm� 3 mm� 50 mmwith the jointmidway along their lengths. The edges of the side to be placedunder tension during bending (top on bending test) werechamfered to remove machining flaws that could intiate pre-mature failure. Strength testing was conducted by four-pointbending. The spans between the inner and outer contact pointswere 20 and 40 mm, respectively, and testing was performed at adisplacement rate of 0.5 mm/min. Flexure strengths were calcu-lated from the load at failure using standard relationshipsderived for monolithic elastic materials:16

sF ¼3PL

4bd2(1)

where P is the applied load, L is the length of the outer span, andb and d are the respective width and height of the specimen. Fivespecimens, each cut from the same plate, were used to determinethe average room temperature joint strength for a given brazecomposition.

Microstructural analysis of the joints was performed onpolished cross-sectioned samples using a JEOL JSM-5900LVscanning electron microscope (SEM, Tokyo, Japan). The SEMis equipped with an Oxford energy-dispersive X-ray (EDX,Ametek, Mahwah, NJ) analysis system, which uses a window-less detector for quantitative detection of both light and heavyelements. To prevent electrical charging of the samples in theSEM, they were carbon coated and grounded. Elemental pro-files were recorded across joint interfaces in the line-scan mode.

III. Results and Discussion

(1) Sessile Drop Experiments

As plotted in Fig. 2(a), all eight braze compositions displayedmeasurable wetting on the polished polycrystalline aluminasubstrates. The effect of the copper oxide is particularly evidentat low oxide content. As seen in Fig. 2(a), the addition of copperoxide causes a rapid decrease in contact angle (inverse ofwettability) at low copper oxide concentration and a moremoderate decrease at higher concentrations. Within a 101–201C difference, the melting of the binary braze compositionsintitiated at the temperature predicted by the pseudobinaryCuO–Ag phase diagram, indicating that the copper oxide andAg are not impeded from reaching equilibrium by diffusion orsolid-state reaction kinetics. The 15-min hold time used fortaking the sessile drop measurements appeared to be longenough for interfacial equilibrium to be established; in all cases,the contact angle reached its stable value within 5 min. With theexception of CA80, each braze was completely molten at 10001Cor lower. The CA80 braze was not expected to melt fully untilB10271C.9 In Fig. 2(b), a plot of contact angle as a function ofbraze composition for the two hold temperature is shown.Again, between the eutectic and monotectic compositions, thecontact angle displays a monotonic decrease with increasingcopper oxide content, but only a marginal change with increas-ing temperature (Fig. 2(b)). This suggests that the wettingphenomena that take place on the braze/alumina interface arerapid and essentially complete by 10001C, being unaffected byan increase in temperature, but may be hindered from reachingtheir maximum effect by the lack of reactant, i.e., too littlecopper oxide. As will be seen in the metallographic results, alikely reason for this is that a copper oxide-rich liquid phaseinteracts with and prewets the alumina surface. Thus, maximumsurface coverage will be achieved when the braze containsgreater than a critical concentration of copper oxide at thealumina faying surface.

(2) Microstructural Analysis of As-Joined and As-OxidizedRAB Specimens

Back-scattered electron images of the as-joined RAB specimensare shown in Fig. 3. As seen in Fig. 3(a), the sample that was

joined using the CA80 braze, which is hypermonotectic withrespect to copper oxide composition, exhibits an extensive andcontinuous copper alumunum oxide (CuAlO2) reaction zoneadjacent to the braze/alumina interface. The composition of thisreaction product is consistent with that predicted by the copperoxide (Cu2O and CuO)–alumina-phase diagrams.17 Extendingfrom the reaction zones toward the center of the braze is aregion that consists of two blocky phases that compositionallyare nearly pure CuxO and nearly pure Ag. The pseudobinaryCuxO–Ag phase diagram indicates that at 11001C, the braze is asingle liquid phase that is rich in copper oxide, possibly Cu2O aspure CuO will spontaneously reduce above 10201C. The moltenliquid reacts with the alumina at the joining surfaces to form asolid CuAlO2 interfacial phase at 11001C. A similar phenom-enon has been observed previously at this temperature in thecopper oxide–alumina system.18 Given that the CuAlO2 zones inFig. 3(a) are fairly extensive in comparison with the thickness ofthe braze region, the molten phase will become depleted incopper oxide. As this continues, the composition of the liquidcould easily reach the miscibility boundary in the pseudobinaryCuxO–Ag phase diagram and thus form two immiscible liquidsthat would have a tendency to segregate. The minor phase liquidwould be silver rich, containing on the order of 90 mol% Ag,whereas the major liquid phase would be silver poor, with acomposition of approximately 35 mol% Ag. Liquid-phaseseparation would account for the significant amount of silversegregation observed in the solidified braze of Fig. 3(a); virtuallyno silver is found in the copper oxide region of the braze. Uponfurther cooling, as the two liquids reach the monotectic tem-perature, the silver-poor liquid would become further depletedin silver and eventually copper oxide would precipitate out ofsolution. After the monotectic reaction, the remaining small

0

10

20

30

40

50

60

70

80

0 20 40 60 50 100Ag Content (mol%)

Co

nta

ct A

ng

le (

Deg

ree)

1000°C1100°C Eutectic

Monotectic

(a)

0

10

20

30

40

50

60

70

80

950 1000 1050 1100 1150Temperature (°C)

Co

nta

ct A

ng

le (

Deg

ree)

CA80

CA69M

CA60

CA08

CA04

CA02

CA01E

CA01

(b)

120

Fig. 2. Contact angle of Ag–CuO brazes on Al2O3 in ambient air as afunction of (a) composition and (b) temperature. The estimate error foreach data point is 110%. The hold time for each soak temperature was15 min.

September 2005 Effects of CuO Content on the Wetting Behavior 2523

amount of liquid would contain very little copper oxide,B2 mol%. When it solidifies, it will form predominantly silverand a small amount of copper oxide.

As seen in Fig. 3(b), the microstructure of the monotecticbraze CA69M, which was heated to 10001C, contains a randommixture of two phases, one that is nearly pure Ag and the othernearly pure copper oxide. This microstructure is typical for amonotectic reaction. At the monotectic temperature, solid cop-per oxide and a silver-rich liquid nucleate simultaneously fromthe silver-poor monotectic liquid. Proportionally, copper oxideis the major product. As it forms and grows during the invariantreaction, the oxide precipitates will eventually impinge with eachother trapping the still molten silver-rich phase within. Underfurther slow cooling, the liquid eliminates additional copperoxide, presumably at the interface with the proeutectic copperoxide. Eventually, the eutectic liquid solidifies forming predo-minantly silver and a small amount of copper oxide. Unlike theCA80 brazed joint, this sample does not contain an extensivereaction zone and no CuAlO2 was found. However, the EDXanalysis shown in Fig. 4 suggests that a thin, o1 mm thickCuxO–Al2O3 diffusion zone (marked in the figure by the twoasterisks) exists along the braze/alumina interfaces. This result isconsistent with the pseudobinary CuxO–Al2O3 phase diagram,which displays complete solubility between copper oxide andAl2O3.

The CA60 brazed joint, shown in Fig. 3(c), also displays twophases, copper oxide and Ag, but the morphologies of each arequite different from that of CA69M.Most of the copper oxide inCA60 is found in a nearly continuous layer along each interfacewith the alumina. When this braze is heated to 10001C, two

liquid phases form, one that is rich in silver (the minor phase)and the other silver poor (the major phase). Because the phasesare immiscible, it is expected that they will segregate, with thesilver-poor liquid preferentially migrating to and wetting thealumina surfaces because of its higher copper oxide content and

Fig. 4. Energy-dispersive X-ray analysis (line profiles) of Cu, Al, andAg obtained on the CA69M braze: the regions indicated by ‘‘�’’represent the presence of Cu in the alumina plates near the interface.

Fig. 3. Cross-sectional scanning electron microscope micrographs (back-scattered images) of braze/alumina interfaces: (a) CA80, (b) CA69M, (c)CA60, and (d) CA01E. Each specimen was heated in air at a final soak temperature of 10001C for 30 min, except for the joint containing the CA80 braze,which was heated to 11001C.

2524 Journal of the American Ceramic Society—Kim et al. Vol. 88, No. 9

therefore lower expected interfacial energy with alumina. Uponcooling to the monotectic temperature, copper oxide beginsto precipitate from this liquid, nucleating along the aluminaboundary. As it does, the silver-rich liquid becomes furtherenriched with silver. At the eutectic temperature, solid copperoxide and Ag will simultaneously nucleate from the remainingliquid, presumably heterogeneously on the surface of the pre-viously formed copper oxide layers that coat the alumina fayingsurfaces. Because the eutectic liquid is predominantly silver,the central region of the braze is nearly free of copper oxideprecipitates. Those that do exist within this region assume anequiaxed morphology.

The joint formed using the eutectic braze, with a compositionof 1.4 mol% copper oxide and 98.6 mol% Ag, displays discretecopper oxide particles both within a nearly pure Ag centralregion and along the interface with alumina (Fig. 3(d)). At10001C, the CA01E braze will form a single silver-rich liquidphase between the two alumina faying surfaces. Upon cooling tothe eutectic temperature, solid Ag and copper oxide formsimultaneously from the eutectic liquid. However, because thecopper oxide concentration of the molten phase is very low, thecopper oxide nucleates as fine, discrete particulates that decoratethe interface with alumina and the interior of the braze. Inaddition, an occasional pore can be found along the braze/alumina interface, indicating that the wettability of this brazeis not as good as the others, even at the micro-scale level. Asmentioned previously, the contact angles observed in Fig. 2(a)can be directly related to the phenomena occuring at the braze/alumina interface. It is apparent that for the silver-rich eutecticliquid to appropriately wet the alumina, a nearly continuous

CuxO–Al2O3 diffusion zone and/or copper oxide interfaciallayer is required. Obviously, the discrete copper oxide particlesthat populate the interfaces in the CA01E brazed joint arespaced closely enough to provide sufficient wetting with thealumina, but do not offer the type of wettability exhibited by theCA69M and CA60 brazes.

Microstructural changes within the brazed joints that under-went air oxidation at 8001C for 100 h are shown in the series ofback-scattered electron images in Fig. 5. It is immediatelyobvious from this set of micrographs that neither the bulkmatrix nor the braze/alumina interfaces change significantlyduring exposure to high-temperature air. As found in the as-joined samples, the exposure-tested CA80 brazed joint stillexhibits the CuAlO2 reaction layer along each interface withalumina. This phase does not appear to thicken as a result of the100 h thermal treatment. The other three joints also appear toretain their as-joined microstructural condition after exposuretesting. There was no indication, for example, that the copperaluminate phase observed in CA80 formed in any these joiningspecimens, suggesting that the RAB brazes possess good oxida-tion resistance.

(3) Fracture Strength and Fractography of As-JoinedRAB Specimens

The four-point bend strengths of each joint, as measured atroom temperature, are plotted along with contact angle as afunction of braze composition in Fig. 6. The results reveal thatbending strength improves with increasing copper oxide contentup to 8 mol%. The specimen joined using the braze with 1 mol%

Fig. 5. Cross-sectional scanning electron microscope micrographs (back-scattered images) of the same four alumina/braze/alumina joints shown in Fig.2, after testing for 100 h at 8001C in air: (a) CA80, (b) CA69M, (c) CA60, and (d) CA01E.

September 2005 Effects of CuO Content on the Wetting Behavior 2525

of copper oxide (CA01) displays a bending strength of 117MPa,while the specimen brazed with 8 mol% of copper oxide (CA08)exhibits almost a twofold improvement with an average strengthof 221 MPa. However, compositions containing higher copperoxide content display rather low bending strengths (101MPa forCA60 and 73 MPa for CA80) despite the significant improve-ment in wettability.

The fracture surfaces of selected samples are shown in Fig. 7.The sample brazed with the 80 mol% copper oxide composition,CA80 in Fig. 7(a), shows signs of crack propagation througha brittle mixed oxide phase region consisting of CuAlO2

(dark phase) and copper oxide (bright phase). The CA60sample, which was joined with the 60 mol% copper oxidebraze, also displays a typical brittle fracture mode, in this casealmost entirely through copper oxide (Fig. 7(b)). On theother hand, as seen in Figs. 7(c) and (d), samples joined withthe lower copper oxide content brazes (CA08 and CA01) do notexhibit brittle fracture surfaces. The CA08 sample, characterizedby the maximum average bending strength in the entire set,displays ductile fracture through the continuous silver phaseadjacent to the braze/alumina interface. In the CA01 sample,however, there is evidence of debonding at the braze/aluminainterface, in addition to ductile fracture of the few islands ofsilver that remain well adhered to the alumina faying surface.This mode of failure is believed to be the direct result of poorwetting between the 1 mol% copper oxide braze and thealumina substrate.

In general, brazes with high copper oxide content containcontinuous brittle phases such as CuAlO2 (CA80) and copperoxide (CA60) along the interface with the alumina substrate.These phases can act as sites for crack initiation and propaga-tion, particularly if they contain microcracks or exist undertensile residual stresses because of the mismatch in thermalexpansion with the substrate, and thus give rise to the lowfracture strengths observed despite improved wettability.Although wetting is greatly ameliorated by the addition ofcopper oxide to silver, it appears that the presence of a con-

40

60

80

100

120

140

160

180

200

220

240

260

0 20 40 60 80 100 120

Ag Content (mol%)

Ben

d S

tren

gth

(M

pa)

0

10

20

30

40

50

60

70

80

Wet

tin

g A

ng

le (

Deg

ree)

Bend StrengthWetting angle

Fig. 6. Room-temperature four-point bend strength and contact angleas a function of Ag content. Note that all of the bend bar specimens werejoined under the same conditions: heat in static air at 51C/min to 10001C(or 11001C for CA80), hold at the soak temperature for 30 min, and coolto room temperature at 51C/min.

Fig. 7. Scanning electron microscope micrographs of the fracture surfaces for the following fractured alumina/braze/alumina joint specimens: (a)CA80, (b) CA60, (c) CA08, and (d) CA01.

2526 Journal of the American Ceramic Society—Kim et al. Vol. 88, No. 9

tinuous phase of copper oxide either within the interior of thebraze, as in the cases of CA80 and CA69M, or along the braze/alumina interface (CA60) is deleterious to the strength of thejoint. On the other hand, the specimens joined with extremelylow copper oxide-containing brazes such as CA01 exhibit ameasurable amount of porosity at the faying surfaces becauseof poor wettability, which also leads to poor joint strengths as aresult of interfacial debonding.

The specimen joined using the braze with intermediate copperoxide content (CA08) displays sound joint interfaces and im-proved wettability because of the formation of discrete islands ofcopper oxide along each interface that essentially anchor the restof the braze, i.e., the silver matrix. The higher degree of wettingprevents the interfacial debonding seen in the CA01 specimen,forcing failure to occur solely by the ductile fracture within thesilver matrix. In this way, joint strength is optimized to a valuethat is comparable with that of monolithic alumina. Sincethe discrete interfacial copper oxide particles may still act ascrack initiation sites, further improvements in strength maybe possible by refining the size of the particles through modica-tions in heat treatment and braze composition. That is, webelieve that both the wetting behavior of the braze and theresulting joint strength can be enhanced simultaneously bycarefully controlling the nucleation and growth events of theinterfacial copper oxide.

IV. Conclusions

A new joining technique, reaction air brazing, was investigatedas a potential method of developing oxidation-resistant ceramicjoints. In this study, it was found that the addition of copperoxide to silver significantly improves the wetting behavior of thenoble metal on alumina. In general, the observed decrease incontact angle is related to the formation of a nearly continuousCuxO–Al2O3 diffusion zone and/or copper oxide layer along thebraze/alumina interface. The effect of temperature on the wet-ting characteristics of these brazes was found to be marginal.Despite the significant improvement in wetting because of theaddition of copper oxide, the resulting joint strengths aremaximized when the braze contains 8 mol% of copper oxideand decrease dramatically at higher copper oxide content.Microstructural results suggest that the presence of continuouscopper oxide-related phases along the interface with aluminadegrades joint strength in brazes containing high contentof copper oxide. Only when the copper oxide is present as adiscrete phase, decorating the braze/alumina interface, does

the flexural strength of the RAB joint approach that of mono-lithic alumina.

Acknowledgments

The authors would like to thank Nat Saenz, Shelly Carlson, and Jim Colemanfor their assistance in polishing the wetting samples and conducting the metallo-graphic and SEM analysis work.

References

1K. Eichler, G. Solow, P. Otschik, and W. Schaffrath, ‘‘Degradation Effects atSealing Glasses for the SOFC;’’ pp. 899–906 in Proceedings of the Fourth EuropeanSolid Oxide Fuel Cell Forum, Oberrohrdorf, Switzerland, Edited by A. J. McEvoy.European Fuel Cell Forum, Oberrohrdorf, Switzerland, 2000.

2C. A. Lewinsohn, M. Singh, T. Shibayama, T. Hinoki, M. Ando, Y. Katoh,and A. Kohyama, ‘‘Joining of Silicon Carbide Composites for Fusion EnergyApplications,’’ J. Nucl. Mater., 283–287B, 1258–61 (2000).

3E. Pippel, J. Woltersdorf, P. Colombo, and A. Donato, ‘‘Structure andComposition of Interlayers in Joints Between SiC Bodies,’’ J. Eur. Ceram. Soc.,17 [10] 1259–65 (1997).

4M. G. Nicholas and D. Mortimer, ‘‘Ceramic/Metal Joining for StructuralApplications,’’ Mater. Sci. Technol., 1 [9] 657–65 (1985).

5Y. S. Chung and T. Iseki, ‘‘Interfacial Phenomena in Joining of Ceramics byActive Metal Brazing Alloy,’’ Eng. Fract. Mech., 40 [4–5] 941–9 (1991).

6W. B. Hanson, K. I. Ironside, and J. A. Fernie, ‘‘Active Metal Brazing ofZirconia,’’ Acta Mater., 48 [18–19] 4673–6 (2000).

7J.-H. Kim and Y.-C. Yoo, ‘‘Bonding of Alumina to Metals with Ag–Cu–ZrBrazing Alloy,’’ J. Mater. Sci. Lett., 16 [14] 1212–5 (1997).

8K. S. Weil and J. P. Rice, ‘‘Substrate Effects on the High-TemperatureOxidation Behavior of a Gold-based Braze Filler Metal;’’ Scripta Materialia, 52[11] 1081–5 (2005).

9J. F. Burgess and C. A. Neugebauer, ‘‘Direct Bonding of Metal with a Metal-Gas Eutectic;’’ U.S. Patent No. 3,744,120, July 10, 1973.

10Z. B. Shao, K. R. Liu, L. Q. Liu, H. K. Liu, and S. Dou, ‘‘Equilibrium PhaseDiagrams in the Systems PbO–Ag and CuO–Ag,’’ J. Am. Ceram. Soc., 76 [10]2663–4 (1993).

11A. M.Meier, P. R. Chidambaram, and G. R. Edwards, ‘‘A Comparison of theWettability of Copper–Copper Oxide and Silver–Copper Oxide on PolycrystallineAlumina,’’ J. Mater. Sci., 30 [19] 4781–6 (1995).

12Y. Yoshino, ‘‘Role of Oxygen in Bonding Copper to Alumina,’’ J. Am. Ceram.Soc., 72 [8] 1322–7 (1989).

13C. Beraud, M. Courbiere, C. Esnouf, D. Juve, and D. Treheux, ‘‘Study ofCopper–Alumina Bonding,’’ J. Mater. Sci., 24 [12] 4545–54 (1989).

14C. C. Shuler, A. Stuck, N. Beck, H. Keser, and U. Tack, ‘‘Direct SilverBonding—An Alternative for Substrates in Power Semiconductor Packaging,’’J. Mater. Sci.: Mater. Electron., 11 [3] 389–96 (2000).

15Ferro Corporation, Screen Printable Binder, Ferro Electronic Materials Pro-duct Literature Publication, pp. 1–2. Ferro Corporation, Cleveland, 2003.

16M. R. Locatelli, A. P. Tomsia, K. Nakashima, B. J. Dalgleish, and A. M.Glaeser, ‘‘New Strategies for Joining Ceramics for High-Temperature Applica-tions,’’ Key Eng. Mater., 111–112, 157–90 (1995).

17S. K. Misra and A. C. D. Chaklader, J. Am. Ceram. Soc., 46 [10] 609 (1963).18D. W. Sosnitzky and C. B. Carter, ‘‘The Formation of Copper Aluminate by

Solid-State Reaction,’’ J. Mater. Res., 6 [9] 1958–63 (1991). &

September 2005 Effects of CuO Content on the Wetting Behavior 2527