WELDING RESEARCH

-s41WELDING JOURNAL

ABSTRACT. Alumina-niobium inter-faces were fabricated at 1400°C via solid-state diffusion brazing of a 127-µm-thickniobium foil between alumina blocks.Prior to brazing, some of the alumina mat-ing surfaces, both polished and unpol-ished, were evaporation-coated with cop-per films ª1.4, ª3.0, and ª5.5 µm thick toinduce liquid-film-assisted joining at thebrazing temperature. The effects of cop-per film thickness and surface roughnesson fracture characteristics and ceramic-metal interfacial microstructure were in-vestigated by room-temperature four-point bend tests, optical microscopy,profilometry, and atomic force mi-croscopy. The average strength of bondsbetween niobium and polished aluminasubstrates increased with the introductionof copper film interlayers, and the scatterin strength tended to decrease, with an op-timum combination of strength andWeibull modulus arising for a copper filmthickness of 3.0 µm. The strength charac-teristics of niobium bonded to unpolishedalumina substrates were also improved byliquid-film-assisted joining, but were unaf-fected by the thickness of the copper interlayers.

Introduction

Bonded ceramic-metal interfaces play avital role in modern materials applications.Precise control of interfacial microstruc-ture through processing is therefore essen-tial, and the development of processing-microstructure-properties correlations isof sound fundamental value. Among themore widely studied ceramic-metal sys-tems is alumina-niobium, which, due toclosely matched thermal expansion coeffi-cients, results in bonded interfaces that arevirtually free of thermal stresses. Consid-erable research has previously been con-ducted on the mechanical properties andinterfacial characterization of this system

(Refs. 1–9). Niobium and alumina are alsochemically compatible, resulting in inter-faces with no chemical reaction layer whenbonded in vacuum (Refs. 2, 7, 10).

In the present study, alumina wasjoined using copper/niobium/copper in-terlayers via liquid-film-assisted joining(LFAJ). The LFAJ approach to joining ce-ramics employs a multilayer metallic in-terlayer composed of two thin claddinglayers of a low-melting-point metal (cop-per) and a thick core of a high-melting-point or refractory metal (niobium) be-tween sections of the ceramic to be joined(alumina). Copper was chosen as the liq-uid former because of its low meltingpoint, ease of deposition, previous re-search on the joining of copper and alu-mina via diffusion brazing and partialtransient liquid phase (PTLP) bonding,previous fracture studies of alumina-copper interfaces, and the past success ofPTLP bonding with copper (see Ref. 7 andreferences therein). The joining tempera-ture is above the melting point of copperbut below that of niobium; consequently,during the initial stages of bonding, a thin,copper-rich liquid film develops betweenthe alumina and niobium, resulting in het-erophase liquid-phase sintering. Redistri-bution of this liquid layer fills voids at theinterface and provides a path for the rapiddiffusion of niobium, which, in turn, ac-celerates contact formation between thealumina and niobium. Fractography of theinterfaces indicates that the liquid copperfilm results in more extensive alumina-niobium contact compared to solid-statediffusion brazing, and concomitantly im-proved strength (Refs. 11, 12). The copperfilm becomes discontinuous, and upon

cooling from the bonding temperature,discrete particles of copper remain at theinterface due to the limited solubility andslow diffusion of copper in niobium. Plas-tic deformation of the ductile copper par-ticles increases the toughness of the inter-face (Ref. 13), and tearing of this ductilemetal during fracture has been observed(Refs. 11, 12). The goal of this researchwas to explore the effects of selected pro-cessing conditions (copper film thickness,alumina surface finish) on the interfacialmicrostructure and mechanical propertiesof joined assemblies.

Background

Extensive discussions of the alumina-niobium and sapphire-niobium systemscan be found in the literature (Refs. 2, 4,5, 8, 14–30) and prior publications (Refs.7, 9, 11, 12). Copper-niobium is an attrac-tive brazing system for joining alumina be-cause there are no brittle intermediatephases, and at temperatures above themelting point of copper, the compositionof the equilibrium liquid contains a fewat.-% of niobium (Ref. 31), which hasbeen shown to enhance the wetting of cop-per on alumina (Refs. 32–34).

Prior studies by Shalz et al. (Ref. 7) andMarks et al. (Ref. 11) have establishedthat pressure and temperature can havean important influence on joint character-istics. At a fixed bonding temperature of1150°C, increasing the applied load duringvacuum bonding from 2.2 to 5.1 MPa in-creased the average strength in four-pointbend tests from 78 ± 22 to 181 ± 45 MPa.1In both joints, failures occurred primarilyalong the alumina-niobium interface. At afixed bonding pressure of 2.2 MPa, in-creasing the bonding temperature from1150° to 1400°C increased the four-pointbend strength from 78 ± 22 to 241 ± 18

Processing of Alumina-Niobium Interfacesvia Liquid-Film-Assisted Joining

The average strength of bonds between niobium and polished alumina substratesincreased with the introduction of copper film interlayers

BY J. T. MCKEOWN, J. D. SUGAR, R. GRONSKY, AND A. M. GLAESER

J. T. MCKEOWN, J. D. SUGAR, R. GRONSKY,and A. M. GLAESER are with the Department ofMaterials Science & Engineering, University ofCalifornia, and Division of Materials Science,Lawrence Berkeley National Laboratory, Berke-ley, Calif.

KEYWORDS

AluminaNiobiumDiffusion BrazingLiquid-Film-Assisted JoiningCopper FilmFracture PathArea Fracture of Contact

1. The average strength represents the mean of themeasured fracture strengths for a given set of join-ing conditions, not the median strength in aWeibull distribution. The error ranges represent±1 standard deviation of the measured strengthsfor a given set of joining conditions. These defini-tions are used throughout the paper.

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 41

WELDING RESEARCH

MARCH 2005-s42

MPa, and ª75% of the samples failed inthe ceramic. For samples processed at1400°C, ceramic failures were also ob-served in high-temperature fracture testsup to 1100°C, suggesting the formation ofa strong interface with potential for use atelevated temperature (Ref. 11).

In prior studies with copper/niobium/copper interlayers (Ref. 11), local varia-tions in the copper film thickness were ob-served to affect strength characteristics. Intwo samples bonded under the same con-ditions (1150°C, 2.2 MPa), the weakersample had a copper thickness of 3.8 µm,while the stronger had a copper thicknessof 3.0 µm. The (interfacial) fracture sur-faces of the weaker samples were domi-nated by copper, while the stronger sam-ples displayed more alumina-niobiumcontact. Local variations in copper thick-ness can cause spatial variations in thetime required to initiate widespread alu-mina-niobium contact and the extent ofensuing alumina-niobium contact growth(Ref. 11). The kinetics of copper filmbreakup depends on the film thickness,and excessively thick films are expected todegrade joint properties.

Liu et al. (Ref. 35) examined the effectsof a thin-film niobium interlayer on thefracture strength of sapphire-copper dif-fusion brazes. Niobium interlayers weredeposited by e-beam evaporation onto thejoining surfaces of the sapphire prior todiffusion brazing. The introduction of thinniobium films greatly improved the sap-

phire-copper bond strength. The fractureenergies of sapphire-niobium-copperjoints were significantly higher than sap-phire-copper joints processed without aniobium interlayer. This was attributed tostrong adhesion of niobium to aluminarelative to that of copper to alumina com-bined with plastic deformation in the cop-per during fracture. A theoretical valuefor the work of adhesion of a pure sap-phire-niobium interface is 0.8 J/m2 (Ref.36), while an experimental value of 1.63J/m2 was reported by Jilavi (Ref. 37) (ascited in Ref. 35). Experimental values forthe work of adhesion for a sapphire-copper interface range from 0.49 to 0.54J/m2 (Refs. 38, 39). The fracture energy isrelated to the work of adhesion by a powerlaw (Ref. 27). Plastic deformation in sap-phire-copper interfaces without a niobiuminterlayer is still possible, but the weakerbonding of copper to sapphire leads tobrittle debonding along the interface andlimited plastic deformation (Ref. 35).

Sugar et al. (Ref. 12) have also re-ported strength degradation in assembliesin which the copper film thickness was de-creased sufficiently. This is attributed toan increase in the density of interfacialflaws when an inadequate amount of cop-per is available to fill irregularities andgaps between the two mating surfaces.Careful polishing and surface preparationshould in principle reduce the severity ofsurface irregularities and thus reduce theamount of copper required to fill interfa-

cial gaps. An interplay between film thick-ness and surface preparation can be ex-pected, with the potential emergence of anoptimum film thickness and surfacepreparation combination.

Surface roughness will not only influ-ence the size and spatial distribution of thegaps between mating surfaces, but canalso have multiple additional and oppos-ing effects on the strength of a ceramic-metal interface. A roughened surface canprevent ceramic-metal contact at the in-terface with large deviations from pla-narity leading to sharp, crack-like interfa-cial flaws. Deep scratches on a ceramicsurface will not be completely filled by anonwetting liquid metal, introducingnear-interfacial flaws. Surface roughnesswill have an effect on the wettability of aliquid metal on a ceramic substrate. In-creased roughness may also introducenear-surface damage and flaws, particu-larly in brittle materials. However, a roughinterface can lead to an anchoring effectthat promotes joining by mechanical in-terlocking at the interface (Refs. 40, 41).Together, these effects influence the me-chanical properties of a joint. Relativelyfew studies have explored and quantifiedthe effects of surface roughness on ulti-mate ceramic-metal joint properties.

Suganuma et al. (Ref. 40) investigatedthe effects of surface damage introducedby grinding on the properties of silicon ni-tride joints prepared with a pure alu-minum braze. Brazing involved a 10 min

A

B

Fig. 1 — Failure probability vs. fracture strength for joints processed withpolished Al2O3 substrates. Filled symbols indicate failure along the Al2O3-Nb interface; open symbols indicate failure in the ceramic.

Fig. 2 — AFM images showing regions on the surfaces of (A) as-receivedNb and (B) flattened Nb. Note the depression on the surface of the flattenedNb. The average roughness of the as-received and flattened Nb is, respec-tively, 107 and 97 nm.

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 42

WELDING RESEARCH

-s43WELDING JOURNAL

hold at 800°C under an applied load of0.05 MPa for three grades of roughness.The average roughness, Ra, was calculatedas

where L is the length of the measured dis-tance and f(x) is the distance to the surfacemeasured from the centerline. Smootherjoining surfaces were found to yieldstronger joints in three-point bending withless scatter in the data. The averagestrength and the Weibull modulus de-creased with increasing roughness, whilethe scatter in the data, as characterized bythe standard deviation, increased with in-creasing roughness.

Suganuma and coworkers (Ref. 42)also explored the effects of surface rough-ness on the properties of alumina-niobiumjoints prepared by diffusion brazing. Poly-crystalline alumina blocks were bonded toan intervening 1-mm-thick niobium diskusing a pressure of 20 MPa for 1 h at1500°C. The roughness of the polycrys-talline alumina and niobium surfaces werevaried, with Ra values of 0.97, 1.62, and 3.1µm, and 0.41, 2.22, and 3.49 µm, charac-terizing the alumina and niobium sur-faces, respectively. The results of room-temperature four-point bend tests andfractographic analyses were correlatedwith changes in surface roughness. In-creasing alumina roughness resulted in amodest (£10%) decrease in averagestrength, increased scatter in strength, anda decrease in the area fraction of alumina-niobium contact (98% Æ ª80%). In con-trast, increasing niobium roughness re-sulted in a modest (£10%) increase inaverage strength, and an increase in thearea fraction of alumina-niobium contact(ª80% Æ ≥ 99%).

Experimental Procedures

Joint Processing and Mechanical Testing

The materials and the majority of theexperimental procedures used in this workare identical to those of prior studies(Refs. 7, 9, 11, 12). Joints were fabricatedusing a 99.5% pure, ≥ 98% dense alumina(AD995, Coors Technical Ceramic Co.,Oak Ridge, Tenn.) in the form of 19.5 ¥19.5 ¥ 22.5 mm blocks. The joining sur-faces of the alumina blocks were groundflat using a diamond wheel (400 grit) on asurface grinder (K. O. Lee Co., Aberdeen,S.Dak.). Joints processed with unpolishedalumina substrates were then cleanedwhile those processed with polished alu-mina substrates were polished with pro-gressively finer diamond suspensions(South Bay Technologies, San Clemente,Calif.) before cleaning. After polishingwith a 1-µm diamond suspension, a finalchemical-mechanical polish was per-formed using colloidal silica (Struers,Westlake, Ohio).

A flattened and cleaned, 99.99% pure,127-µm-thick niobium foil (GoodfellowCorp., Malvern, Pa.) and a commercial-grade copper (Consolidated CompaniesWire and Associated, Chicago, Ill.) servedas the interlayer materials. For joints pro-duced via liquid-film-assisted joining, cop-per films 1.4, 3.0, or 5.5 µm thick were de-posited directly onto the polished orunpolished alumina joining surfaces byevaporation of the copper wire source in ahigh-vacuum chamber. Film thickness wasdetermined using profilometry (TencorInstruments, Inc., San Jose, Calif.) andweight-gain measurements (Ref. 7).

Joints were processed under high vac-uum (pressure maintained below 7.6 ¥10–5 torr, equivalent to 10–7 atm) in agraphite element vacuum hot press. Thebrazing process coupled a constant ap-plied load of ª2.2 MPa with heating at4°C/min, soaking at the brazing tempera-ture of 1400°C for 6 h, and cooling at2°C/min. After brazing, the assemblieswere machined into beams ª3 mm ¥ ª3mm in cross section and ª4 cm in length,

R L f x dxaL

= ( ) ( )Ú1 (7)0

/

Table 1 — Surface Roughness*

Material Ra (µm) la (µm) r aa (deg)

Unpolished alumina 0.154 (AFM) 8.9 (AFM) 1.014 (AFM) 8.7 (AFM)0.299, 2.72 (Prof.) 30 (Prof.) 1.005, 1.411 (Prof.) 4.9, 45.3 (Prof.)

Polished alumina 0.027 (AFM) 32.5 (AFM) 1.00003 (AFM) 0.42 (AFM)0.051, 0.057 (Prof.) 30 (Prof.) 1.00014, 1.00018 0.85, 0.95 (Prof.)

(Prof.)As-received niobium 0.107 11.3 1.0045 4.9Flattened niobium 0.097 (AFM) 6.8 (AFM) 1.01 (AFM) 7.3 (AFM)

0.213 (Prof.) 20 (Prof.) 1.006 (Prof.) 5.3 (Prof.)

* Ra is the average deviation in surface height, la is the average distance between peaks of surface features, r isthe roughness parameter, and aa is the average slope of surface features. Calculations of r and aa use the meth-ods described by Hitchcock et al. (Ref. 43). All values of r and aa were calculated using average values of 50 and9, respectively, for the arithmetic factors K1 and K2. Where two values are provided, the first is parallel to the scandirection, and the second is perpendicular to the scan direction.

Fig. 3 — AFM images showing regions on the surfaces of (A) polished Al2O3 and (B) unpolished Al2O3. The average roughnesses of the polished and unpol-ished surfaces are, respectively, 28 and 154 nm.

A B

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 43

WELDING RESEARCH

MARCH 2005-S44

with the metal interlayer at the center ofthe beam. The tensile surfaces of thebeams were polished to a 1-µm finish andthe edges of the beams were beveled to re-move machining flaws that could initiatefailure. This allowed for a more meaning-ful measurement of the fracture strengthof the joined assembly, and the observedfracture path provided insight on the rela-tive strengths of the ceramic-metal inter-face and the bulk ceramic.

Beams were tested at room tempera-ture using four-point bending. The innerspan of the test jig was 9 mm; the outerspan was 25 mm. Testing was performedwith a displacement rate of 0.05 mm/min.Strengths were calculated from the load atfailure using standard relationships de-rived for monolithic elastic materials.

Surface and Interface Characterization

The surface roughness of the as-ground and polished alumina blocks, aswell as the roughness of the niobium foilbefore and after flattening, were deter-

mined by atomic force microscopy (AFM)to allow comparison between the surfaceroughness of the substrates and interlayerand the copper film thickness. The AFMscans typically covered a 50 ¥ 50 µm area,and provided accurate measurements ofthe local fine-scale roughness. To assesslonger wavelength (and larger amplitude)variations in the surface topography, pro-filometer scans spanning a length of ª1 cmwere also conducted on polished and as-ground alumina and flattened niobiumfoils. For the ground alumina, scans wereconducted both parallel and perpendicu-lar to the grinding direction. The resultsare summarized in Table 1. The roughnessdefinitions of Hitchcock et al. (Ref. 43)(see Appendix) were used as the basis forthe calculations of the cited roughness values.

For selected samples in which fractureoccurred at or near the alumina-interlayerinterface, fracture surfaces near the ten-sile edge were examined using optical mi-croscopy. Fracture surfaces were mountedadjacent to one another so that equivalent

fractographic locations of the metal andceramic were in mirror symmetry posi-tions. The microstructure at matching lo-cations, the pore structure, and the frac-ture path could thus be identified.

During bonding, in regions where theceramic and metal make contact, the ce-ramic and metal grain boundaries areetched, and the surfaces mutually con-form. Thus, when failure occurs along theceramic-metal interface, an imprint of theceramic microstructure appears on themetal surface in regions where contact wasachieved. The area fractions of contactand interfacial failure (for fracture pathstatistics) were determined by a point-counting method employing a referencegrid superimposed onto a micrograph ofthe metal fracture surface. The grid wasrotated with respect to each micrograph,to yield several distinct grid orientations.Approximately 800 intersection pointswere evaluated per micrograph. Areafractions were calculated by averaging re-sults for each micrograph, and are re-ported in Tables 3 and 4.

Fig. 5 — Optical micrographs of a polished sample joined with a 5.5-µmCu film, showing matching regions of (A) metal and (B) ceramic sides ofthe fracture surface. The regions marked “a” indicate unbonded area. Thesample failed at 278 MPa.

Fig. 4 — Failure probability vs. fracture strength for joints processed withunpolished Al2O3 substrates. Filled symbols indicate failure along theAl2O3-Nb interface; open symbols indicate failure in the ceramic.

A

B

a

a

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 44

WELDING RESEARCH

-s45WELDING JOURNAL

Results and Discussion

Effects of Copper Thickness andSubstrate Roughness on Room-Temperature Mechanical Properties

Polished Alumina Substrates

Figure 1 is a plot of failure probability vs.beam fracture strength for bonded assem-blies processed with varying copper thick-nesses. The alumina “reference” was un-bonded and unannealed before testing, andFig. 1 clearly shows the beneficial effects ofthin liquid copper films on the strengthcharacteristics of joined assemblies.

As indicated in Table 1, the three filmthicknesses exceed Ra and are also com-parable to or greater than the maximumasperity height (0.32 µm) and maximumcavity depth (1.38 µm) in profilometryscans of polished alumina. While “flatten-ing” of the niobium foil appears to havelittle effect on the average roughness, thesurface of the flattened niobium containslarge divots and depressions. Flatteningalso increased the roughness parameter, r,and the average slope of surface features,aa, on the niobium foil (see Appendix andTable 1). Figure 2 shows regions on the as-received and flattened niobium foil. Dur-ing LFAJ, some niobium dissolution oc-curs. As a result, although large divots anddepressions are likely to persist at thebonding temperature, the finer details ofthe surface topography change. Whetherthese changes are beneficial or detrimen-tal is unknown. Smoothing and surfacefaceting are possible depending upon thesurface orientation and stability. It is rea-sonable to expect that in samples preparedusing polished alumina and copper films,sufficient liquid should be available to fillmost interfacial gaps at the bonding temperature.

Solid-state diffusion brazing at 1400°Cand a pressure of 2.2 MPa led to low aver-age strength, which, in the combined dataset taken from two bonded assemblies,was 130 ± 20 MPa; the Weibull moduluswas 5.7. The data do show a kink (dashedlines in Fig. 1) indicating that two failuremodes might be operative; however, frac-tographic analysis showed that all samplesfailed along the alumina-niobium inter-face, with considerable tearing of the nio-bium. Thus, there is no mechanistic basisfor such treatment of the data. Interfacialfailures at low applied stress were due tolarge unbonded regions and a high areafraction of interfacial porosity, a conse-quence of the low applied load and the lowrate of solid-state diffusion.

The application of a 1.4-µm copper filmto the bonding surfaces of the alumina sub-strates increased the average fracturestrength to 197 ± 37 MPa. The weakestbeam failed at 136 MPa, comparable to thestrength of the strongest diffusion-brazedsample (155 MPa). The Weibull moduluswas 5.6, essentially the same as that for dif-fusion brazing. All samples continued to failalong the alumina-niobium interface, butwith more limited tearing of the niobium.Here, LFAJ allowed for the filling of inter-facial voids by liquid copper and provided ahigh-diffusivity path for the transport ofniobium. Copper dissolves ª3 at.-% nio-bium at 1400°C (Ref. 31), and the diffusioncoefficient for niobium in liquid copper isorders of magnitude higher than the self-diffusion coefficient for niobium. Conse-quently, the rate of niobium redistributionat the interface during LFAJ can be ex-pected to be much greater than that in con-ventional solid-state diffusion brazing.

Increasing the copper film thickness to3 µm yielded a further improvement in thestrength to 241 ± 18 MPa, a decrease instandard deviation by slightly more than a

factor of two (from 37 to 18 MPa), and in71% of samples tested, there is evidenceof crack initiation and propagation com-pletely within the ceramic. As Fig. 1 shows,interfacial failures do not necessarilyoccur at lower stresses. The Weibull mod-ulus increased to 14.9. This is comparableto that of the unbonded reference alumina(13.8).

Prior work had suggested that a furtherincrease in the copper film thicknesswould degrade the strength characteristics(Ref. 11) due to a larger area fraction ofcopper at the interface, though no studieshad previously been pursued. When thecopper film thickness was increased toª5.5 µm, the average strength increasedslightly, from 241 to 246 MPa. However,the standard deviation increased byroughly a factor of two, from 18 to 37 MPa.Only about 57% of the samples failed inthe ceramic, but the majority of interfacialfailures occurred at lower applied stresses.Samples with the highest ceramic and in-terfacial fracture strengths failed atstresses of, respectively, 289 and 278 MPa;the lowest ceramic and interfacial failuresoccurred at stresses of, respectively, 206and 136 MPa. This spread in the data is fargreater than that of samples processedwith 3.0-µm films, as evidenced by theWeibull parameter of 6.0, approximatelythe value of diffusion-brazed and 1.4-µm-film assemblies. There is also a large dis-crepancy between the average fracturestrengths and standard deviations for ce-ramic (265 ± 20 MPa) and interfacial (223± 41 MPa) failures. This degradation instrength is attributed to large regions ofcopper at the interface. The work of ad-hesion of an alumina-copper interface isless than that of an alumina-niobium in-terface (Refs. 35–39). Thicker films aremore difficult to break up, and large plate-like copper regions persist at the interface

Fig. 6 — Optical micrographs of the fracture surface from an unpolished diffusion bond that failed at 48 MPa, showing matching regions of (A) metal and (B)ceramic sides. The region labeled 1 is deformed metal. The regions marked “b” indicate bonded regions; regions marked “c” indicate the presence of silicides.

A B

b

c

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 45

WELDING RESEARCH

MARCH 2005-s46

after bonding, yielding higher area frac-tions of alumina-copper interface.

Unpolished Alumina Substrates

Atomic force microscopy (AFM)(Table 1) suggests that it might not be nec-essary to polish the alumina, since the av-erage roughness of unpolished as-groundsurfaces over 50 ¥ 50 µm areas (0.154 µm)remains much less than the minimum liq-uid film thickness. Similar conclusions aresupported by profilometry data over 50-µm length scales. However, comparisonsof profilometry scans over larger dis-tances, up to 1 cm, show significant differ-ences between the polished and as-groundalumina. The Ra value increased fromª0.050–0.060 µm for polished alumina to0.299–2.72 µm in the as-ground alumina,depending upon the scan direction. Maxi-mum asperity heights increased fromª0.18–0.32 µm in polished alumina to1.7–14.1 µm in as-ground alumina de-pending upon the scan direction. Maxi-mum surface cavity depths increased from1.38 µm in polished substrates to ª11 µmin unpolished substrates. In contrast to

polished alumina, the maximum asperityheight and cavity depth far exceed eventhe maximum film thickness. Conse-quently, unpolished alumina substrateswould be more likely to have more severeflaws and larger unbonded regions at thealumina-interlayer interface, and thecharacteristics of the most severe defectswould be relatively insensitive to the filmthickness.

Figure 3 provides a comparison ofAFM images, and differences are appar-ent even over limited spatial scales. Theflaws on polished alumina surfaces areresidual scratches and gouges from grind-ing, while the flaws on unpolished aluminaare more severe. Although the depositedcopper film is generally thick enough to filland cover scratches of this severity at alocal scale, point-by-point assessment of50,000-point profilometry scans of 1 cmtotal length indicate that >3% of thescanned length on the as-ground surfacehas multiple independent asperities >10µm in height. When the ceramic-metalcontact initiates at such extreme asperi-ties, large and relatively wide interfacialgaps are anticipated.

Solid-state diffusion brazing was at-tempted with unpolished alumina sub-strates, but only 4 of 25 beams survivedmachining into test samples, and all 4 ofthese surviving beams failed interfaciallyat low stresses (48–57 MPa). Large un-bonded areas, high porosity, and devia-tions from planarity that introduce localstress concentrations at the interface, allcontribute to crack initiation and failure.

Figure 4 is a plot of failure probabilityvs. beam fracture strength for bonded as-semblies processed with varying copperthicknesses. The average strength of jointsprocessed with copper films of 1.4, 3, and5.5 µm are, respectively, 224 ± 27 MPa,218 ± 23 MPa, and 230 ± 24 MPa, demon-strating that the thickness of the copperfilm has little effect on the strength char-acteristics of assemblies processed withunpolished alumina substrates. The aver-age roughness of unpolished alumina sub-strates is 0.299 and 2.7 µm in orthogonaldirections. Filling a sawtooth surface pro-file with average surface asperity height of2.7 µm and average surface cavity depth of2.7 µm would require at least 2.7 µm ofcopper. However, if during bonding thepeaks on the alumina surface became em-bedded in the niobium and liquid copperwere displaced to adjoining regions, thin-ner films could be sufficient to fill interfa-cial gaps. Unfortunately, the most severeheight deviations in the as-ground alu-mina surface create significantly larger in-terfacial gaps. Displacement of liquid cop-per into these most severe interfacial gapswould reduce their size, but could not fillthem completely. Consequently, the sta-tistical distribution of such flaws at the in-terface may explain the observed strengthcharacteristics.

Fracture characteristics appear to im-prove slightly with increasing film thick-ness (Table 2). The Weibull modulus in-

Table 2 — Strength Characteristics

Substrate Finish Film Thickness Average Strength Weibull Ceramic(µm) (MPa) Modulus Failures

(%)

Polished 0 103 (± 20) 5.7 0Unpolished(a) 0 52 (± 4) — 0Polished 1.4 197 (± 37) 5.6 0Unpolished 1.4 224 (± 27) 8.0 40Polished 3.0 241 (± 18) 14.9 71Unpolished 3.0 218 (± 23) 10.0 29Polished 5.5 246 (± 37) 6.0 57Unpolished 5.5 230 (± 24) 10.2 46

(a) Only four out of a possible 25 beams survived the machining process.

Fig. 7 — Optical micrographs from an unpolished sample joined with a 5.5-µm Cu film, showing matching regions of (A) metal and (B) ceramic sides. The re-gion labeled 2 is a ceramic grain that has pulled out and adhered to the metal surface. The sample failed at 246 MPa.

A B

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 46

WELDING RESEARCH

-s47WELDING JOURNAL

creases from 8.0 for the 1.4-µm film to 10.0for the 3.0-µm film to 10.2 for the 5.5-µmfilm. In bonds prepared with polished sub-strates, the maximum Weibull modulus(14.9) coincided with the highest percent-age of ceramic failures (71% for 3.0-µmfilms). In the present case, the percent-ages of failures that occur in the ceramicfor 1.4-, 3.0-, and 5.5-µm films are, re-spectively, 40%, 29%, and 46%. Thus, theWeibull modulus does not parallel thepercentage of ceramic failures, but seemsto increase with film thickness.

Under optimum conditions, jointsprocessed with polished substrates exhibitslightly better strength characteristics thanthose processed with unpolished sub-strates, and have higher values of theWeibull modulus. However, for many ap-plications unpolished substrates may be ad-equate and even desirable, especially sinceelimination of the polishing process wouldsave considerable time and reduce cost.Further research and more data would behelpful to augment the statistics presentedhere in support of such decisions.

Fractography

Fractography was performed on diffu-sion-brazed and LFAJ samples in bothpolished and unpolished conditions. Forall images, the tensile surface lies near theadjacent edges of the micrographs show-ing the metal (A) and ceramic (B) sides ofthe fracture surface.

By comparing the extent to which theceramic microstructure was imprinted onthe metal surface (Refs. 7, 9, 11, 12, 14, 44,and 45), regions in which contact was notachieved during bonding were identified.In diffusion-brazed samples, approxi-mately one-half of the interface was ob-scured. The ceramic side of the interfacewas obscured by adherent islands of metalor silicide (Refs. 13, 46, 47); the metal sideof the interface was obscured by deformedmetal (labeled ! in Fig. 6A) or adherentalumina grains (labeled " in Fig. 7). Theobservable regions of contact are there-fore only within the fraction of the totalfracture surface associated with interfacialfailure. An upper limit on the area fractionof contact is obtained by assuming that re-gions in which the crack deviates from theinterface are associated with completealumina-niobium contact. However, if theinterfacial porosity is uniformly distrib-uted, the average area fraction of contactis given by the area fraction of contactwithin regions of interfacial failure. Thislatter interpretation seems to better ratio-nalize the significant strength differencesbetween diffusion-brazed and LFAJ sam-ples (Ref. 12). Both values are provided inTables 3 and 4 for polished and unpolishedsamples, respectively. For unpolished

samples processed with copper films,there does not appear to be any tearing ofthe niobium foil, and hence there is no ob-scured interface. Although more ceramicpullout occurs than in polished samples,the failure is predominantly interfacial,and thus, only a single value is reported.

Polished Alumina Substrates

Increasing the copper thickness af-fected both the strength and the fracturepath. Comparing the fracture surfaces ofsamples processed with 0-, 1.4-, 3.0-, and5.5-µm copper films, the average areafraction of interfacial failure wasª48(±5)%, ª67(±6)% (Ref. 12),ª94(±5)%, and ª71(±10)%, respectively.However, it should be noted that the per-

centage of samples that fail along the in-terface varies with copper film thickness(Table 2). In samples with no copper and1.4-µm films, significant tearing of the nio-bium interlayer occurred, along with re-gions where there was no contact betweenthe alumina and niobium interlayer. Nio-bium tearing and unbonded regions aremuch more limited in samples processedwith 3.0- and 5.5-µm films. In these sam-ples, better contact was achieved, and ce-ramic grains pull out and adhere to theniobium fracture surface. However, the5.5-µm copper films resulted in an in-crease in the number of interfacial failuresand unbonded regions as compared to 3.0-µm samples. Figure 5 shows the metal andceramic fracture surfaces of a sample thatfailed at 278 MPa. The copper particles at

Table 3 — Fracture Path and Contact Area Statistics for Polished Assemblies

Fracture Path Statistics Area Fraction Bonded

Area Area FractionFraction Interlayer or

Interfacial CeramicFailure Failure

Diffusion bond79 MPa 0.495 0.505 0.812 0.621102 MPa 0.473 0.527 0.809 0.572119 MPa 0.54 0.46 0.813 0.595

1.4 µm Cu136 MPa 0.695 0.305 0.84 0.77197 MPa 0.665 0.335 0.877 0.808260 MPa 0.68 0.32 0.938 0.908

3.0 µm Cu206 MPa 0.965 0.035 0.843 0.837243 MPa 0.897 0.103 0.788 0.764263 MPa 0.95 0.05 0.97 0.968

5.5 µm Cu192 MPa 0.721 0.279 0.765 0.674213 MPa 0.744 0.256 0.903 0.87261 MPa 0.686 0.314 0.929 0.897

1Area

Area1

Area

Area unbonded

fracture surface

unbonded

interfacial failure

- -

Table 4 — Fracture Path and Contact Area Statistics for Unpolished Assemblies

Fracture Path Statistics Area Fraction Bonded

Area Area FractionFraction Interlayer or

Interfacial CeramicFailure Failure

Diffusion bond48 MPa 0.451 0.549 0.679 0.28857 MPa 0.489 0.511 0.679 0.344

1.4 µm Cu213 MPa — — 0.687 —246 MPa — — 0.686 —

3.0 µm Cu216 MPa — — 0.647 —242 MPa — — 0.729 —

5.5 µm Cu218 MPa — — 0.699 —246 MPa — — 0.809 —

1Area

Area1

Area

Area unbonded

fracture surface

unbonded

interfacial failure

- -

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 47

WELDING RESEARCH

MARCH 2005-s48

the interface tear during fracture, andthere is a slight increase in niobium tear-ing relative to samples processed with a3.0-µm film (see Table 3, Area Fraction In-terlayer or Ceramic Failure). There arestill large unbonded regions between thealumina and niobium (indicated as “a” inFig. 5A), and the area fraction of these re-gions decreased as the fracture strengthincreased. The increase in interfacial fail-ures correlates with a slight increase in theamount of copper in the braze, increasingthe area fraction of alumina-copper inter-face.

Unpolished Alumina Substrates

Fractography of a diffusion-brazed sam-ple that failed at 48 MPa is shown in Fig. 6.From inspection of the metal side (Fig.6A), it is evident that the majority of the in-terface remained unbonded. Small regionsof bonding between the alumina and nio-bium are marked “b” in Fig. 6A. Niobiumgrain boundaries are seen in the large un-bonded regions. Energy-dispersive spec-troscopy in both SEM (Ref. 13) and TEM(Refs. 46, 47), and electron diffractionanalysis (Refs. 46, 47) indicate that silicidesform in regions of alumina with smallgrains. These silicide particles appear asbright regions on the ceramic side of the in-terface, and are labeled “c” in Fig. 6B. Thelow strength of these samples is a result ofpoor contact between the alumina and nio-bium due to the roughness of the aluminasubstrates.

Figure 7 shows metal and ceramic sidesof a sample processed with a 5.5-µm cop-per film that failed at 246 MPa. The frac-ture surfaces for the three film thicknessesdo not appear to show any major differ-ences, which is to be expected based on thesimilar strength characteristics. Thebonded area fraction is approximately thesame for all samples (Table 4). Comparingmating fracture surfaces, the unbondedregions on the metal side correspond withregions on the alumina side that aredarker in contrast. Such unbonded regionsare attributed to large localized depres-sions on the alumina surface, as shown in Fig. 7. The amount of copper on the frac-ture surfaces appears to increase slightlywith increasing film thickness, but doesnot appear to have had any significant ef-fect on the strength characteristics. Thereis also more ceramic pullout in these sam-ples relative to polished samples, possiblyindicative of grinding-induced near-surface damage. These regions that pulledout and adhered to the metal surfaces areelevated regions on the alumina joiningsurfaces. Better contact relative to surfacedepressions can be made with the niobiumduring the bonding cycle, since these ele-vated regions will penetrate and deform

the metal layer due to the applied pressureand temperature. The majority of samplesthat failed interfacially also exhibitedsmall regions of ceramic along the tensileedge that adhered to the niobium foil, in-dicating a more tortuous crack path.

Polished vs. Unpolished Substrates

For samples that fail primarily alongthe alumina-interlayer interface, the dif-ferences between the fracture surfaces ofpolished and unpolished substrates are ev-ident upon inspection. The individual un-bonded regions are significantly larger inthe unpolished samples than in the pol-ished samples (see Figs. 5 and 7, bothprocessed with 5.5-µm copper films), andthe bonded area fraction is lower in un-polished samples. The copper distributionis also different. Again comparing sampleswith 5.5-µm copper films, the average areafraction of copper coverage in unpolishedsamples (ª13%) is roughly twice that inpolished samples (ª7%), and the averageprojected area of residual copper patchesis three to four times larger in unpolishedsamples, ª11–17 µm2 vs. ª4 µm2. As a re-sult, copper is more uniformly dispersedand isolated patches are smaller in the pol-ished specimens.

These general microstructural differ-ences exist regardless of the copper filmthickness. At constant copper film thick-ness, the only variable is the alumina sur-face roughness. Therefore, it is useful toassess and rationalize how a change inroughness could cause or contribute to theobserved microstructural trends.

Models of solid-state metal-metal dif-fusion brazing (Refs. 48, 49) depict theearliest stages of bonding as asperity con-tact, and for small contact area, the ap-plied stress is large. In the present ce-ramic-metal case, asperities on thealumina surface will likely be pushed intothe softer niobium, increasing the contactarea until the stresses decrease to the nio-bium yield stress. Increasing the appliedload promotes more extensive contact for-mation during this initial stage, and re-duces the size of interfacial defects. Inter-facial gaps flank large asperities and deepdepressions in the alumina. Copper, oncemolten, can withdraw from or redistributewithin these cavities leaving voids at thealumina-niobium interface, which, if theypersist, provide potential sites forcrack/failure initiation. As the contactarea increases, regions of the interface be-come isolated by continuous perimeters ofceramic-metal contact. Pockets com-pletely filled with liquid increase the ef-fective area supporting the applied load.Since access to additional copper by liquidflow is precluded, if these isolated regionsare not completely filled with copper at

the bonding temperature, void closureand the development of full contact willrequire either additional deformation/creep of the foil or longer-range diffusion.In view of the significant differences in as-perity heights, significant differences be-tween interfacial microstructures andfracture characteristics in polished andunpolished samples can be expected. Evenwhen the cavities are filled at the bondingtemperature, voids within the copper areexpected at room temperature. The dif-ferential thermal expansion of copper rel-ative to niobium and alumina, and the vol-ume change due to copper solidification,are expected to generate ª12 vol-% poros-ity within the copper-rich phase.

Bonding experiments using highly pol-ished (transparent) single-crystal sapphiresubstrates have allowed nondestructivestudies of the morphological evolution ofthe copper film and the growth of sap-phire-niobium contact during postbond-ing annealing (Ref. 50). The results con-firm that sapphire-niobium contactinitiates at asperities on the adjoining sur-faces. Some of these asperities are initiallypresent on the bonding surfaces, whileothers form and grow during bonding andsubsequent annealing, such as those dueto liquid copper etching niobium grainboundaries and forming grain boundarygrooves. Pairs of protruding ridges thatflank each grain boundary groove developon the niobium surface and grow to locallybridge the liquid copper film. Subsequentcapillary instabilities at the edges of cop-per patches lead to film breakup and theisolation of small copper particles alongthe sapphire-niobium interface.

In the present work, the initial points ofcontact and the progressive growth of alu-mina-niobium contact are highly influ-enced by the alumina surface roughness.Since the roughness of the polished sam-ples approaches that of the sapphire sub-strates, it is reasonable to assume that theearly stages of contact formation will againbe ridges due to preexistent asperities andgrain boundary grooves, and the spacingand location of these points of contact willbe a function of the surface topographyand the niobium grain size. Grooving ofthe alumina grain boundaries and facetingof alumina grains may also contribute tothe breakup and evolution of the copperfilm. Recent work by Saiz et al. (Ref. 51)has demonstrated enhanced rates of alu-mina grain boundary grooving at liquid-metal–alumina interfaces.

Prior work by Suganuma et al. (Ref. 42)and findings in the present study (see Ta-bles 3 and 4) provide evidence of a de-crease in the area fraction of contactachieved during solid-state bonding as theroughness of the alumina increases. Asubstantially larger height difference be-

McKeown Supplement 3/05corr 4/1/05 4:40 PM Page 48

WELDING RESEARCH

-s49WELDING JOURNAL

tween initial points of contact and valleysin the alumina surface make it more diffi-cult to achieve complete ceramic-metalcontact in both diffusion brazing andLFAJ of unpolished samples. Larger in-terfacial gaps relative to polished samplesmay also reduce the impact of grainboundary groove ridging on film breakup.

Optical microscopy of cross sectionsnormal to the alumina-niobium interfaceprepared from bend beams that failed inthe ceramic reveals that the niobium con-forms to the alumina topography. Nio-bium is pushed into depressions (e.g.,pores) in the alumina surface and deformsto accommodate elevations on the alu-mina surface. In unpolished samples, localelevations in the alumina surface of up toseveral microns are evident along the in-terfaces. These may define the initialpoints and areas of alumina-niobium con-tact and define gaps between the aluminaand niobium surfaces during the initialstages of bonding that are larger thanwould be inferred from either the AFM oraveraged profilometry data.

In unpolished samples that fail alongthe interface, fracture surfaces suggest thegaps are too large for complete removal.Isolated, rough mottled regions on themetal side of the fracture surface are be-lieved to reflect the imprint of rough ele-vated regions on the alumina surface asthey are forced into the niobium. The flowand removal of copper liquid from theseregions of contact and the low joining tem-perature inhibit decay of the resulting in-terfacial roughness. The resulting rough-ness at the interface will provide a moretortuous interfacial crack path relative tothat of the polished samples.

The difference in the roughness of thealumina can also affect the redistributionof copper along the interface during pro-cessing and its ultimate morphology. Animportant issue is whether gaps largerthan the local thickness of the copper filmcan be filled by liquid flow. If the contactangles of liquid copper on niobium andalumina are denoted q1 and q2, respec-tively, and the niobium and alumina sur-faces are parallel, then liquid copper flowsinto voids along the interface providedthat q1+q2 < 180 deg. If instead, local de-pressions on the opposing niobium andalumina surfaces cause angular deviationsof a1 and a2, respectively, from this paral-lel surface geometry, then flow of liquidinto voids will only occur if a more strin-gent condition, (q1+a1)+(q2+a2)

WELDING RESEARCH

MARCH 2005-s50

per film thickness in samples producedwith polished substrates but not in sam-ples produced with unpolished substrates,as summarized in Tables 3 and 4. For pol-ished substrates, there is progressively lesstearing of the niobium interlayer with in-creasing copper film thickness. There is atransition from mainly interfacial failureto ceramic failure at a thickness of 3.0 µm.The area fraction of contact increases withincreasing film thickness to 3.0 µm, thendecreases when the film thickness is in-creased further to 5.5 µm. For unpolishedsubstrates, the fracture path and area frac-tion of contact are unaffected by the cop-per film thickness. The interfacial mor-phologies of samples produced withpolished surfaces and unpolished surfacesare significantly different.

Acknowledgments

This research was supported by the Di-rector, Office of Science, Office of BasicEnergy Sciences, Division of MaterialsScience and Engineering, of the U.S. Department of Energy under ContractNo. DE-AC03-76SF00098. The efforts ofDaniel Carvajal in performing the area-fraction measurements are gratefully acknowledged.

References

1. Nicholas, M. G., Forgan, R. R. G., andPoole, D. M. 1968. The adhesion of metal/alu-mina interfaces. J. Mater. Sci. 3: 9–14.

2. Florjancic, M., Mader, W., Rühle, M., andTurwitt, M. 1985. HREM and diffraction stud-ies of an Al2O3/Nb interface. J. Phys. (Paris)46(C4): 129–33.

3. Burger, K., Mader, W., and Rühle, M.1987. Structure, chemistry and diffusion bond-ing of metal/ceramic interfaces. Ultrami-croscopy 22(1–4): 1–14.

4. Mader, W., and Rühle, M. 1989. Electronmicroscopy studies of defects at diffusion-bonded Nb/Al2O3 interfaces. Acta Metall.37(3): 853–66.

5. Mayer, J., Flynn, C. P., and Rühle, M.1990. High-resolution electron microscopystudies of Nb/Al2O3 interfaces. Ultramicroscopy33(1): 51–61.

6. O’Dowd, N. P., Stout, M. G., and Shih, C.F. 1992. Fracture toughness of alumina-niobium interfaces: Experiments and analyses.Philos. Mag. A, Phys. Condens. Matter DefectsMech. Prop. (UK) 66(6): 1037–64.

7. Shalz, M. L., Dalgleish, B. J., Tomsia, A.P., Cannon, R. M., and Glaeser, A. M. 1994. Ce-ramic joining III bonding of alumina viaCu/Nb/Cu interlayers. J. Mater. Sci. 29(14):3678–90.

8. Rühle, M. 1995. Structure and composi-tion of metal/ceramic interfaces. J. Eur. Ceram.Soc. 16(3): 353–65.

9. Marks, R. A., Chapman, D. R., Daniel-son, D. T., and Glaeser, A. M. 2000. Joining ofalumina via copper/niobium/copper interlay-ers. Acta Mater. (USA) 48(18-19): 4425–38.

10. Knauss, D., and Mader, W. 1991. Chem-ical composition and lattice relaxations at dif-fusion-bonded Nb/Al2O3 interfaces. Ultrami-croscopy 37(1–4): 247–262.

11. Marks, R. A., Sugar, J. D., and Glaeser,A. M. 2001. Ceramic joining IV. Effects of pro-cessing conditions on the properties of aluminajoined via Cu/Nb/Cu interlayers. J. Mater. Sci.36(23): 5609–24.

12. Sugar, J. D., McKeown, J. T., Marks, R.A., and Glaeser, A. M. 2002. Liquid-film as-sisted formation of alumina-niobium inter-faces. J. Am. Ceram. Soc. 85(10): 2523–30.

13. Kruzic, J. J., Marks, R. A., Yoshiya, M.,Glaeser, A. M., Cannon, R. M., and Ritchie, R.O. 2002. Fracture and fatigue behavior at am-bient and elevated temperatures of aluminabonded with copper/niobium/copper interlay-ers. J. Am. Ceram. Soc. 85(10): 2531–41.

14. Elssner, G., Reidel, S., and Pabst, R.1975. Fractography and fracture paths in ce-ramic-metal composites. Prakt. Metallogr. 12(5):234–43.

15. Morozumi, S., Kikuchi, M., and Nishino,T. 1981. Bonding mechanism between aluminaand niobium. J. Mater. Sci. 16(8): 2137–44.

16. Turwitt, M., Elssner, G., and Petzow, G.1985. Manufacturing and mechanical proper-ties of interfaces between sapphire and nio-bium. J. Phys. (Paris) 46(C4): 123–27.

17. Rühle, M., Burger, K., and Mader, W.1986. Structure and chemistry of grain bound-aries in ceramics and of metal/ceramic inter-faces. J. Microsc. Spectrosc. Electron 11(3):163–77.

18. Rühle, M., Backhaus-Ricoult, M.,Burger, K., and Mader, W. 1987. Diffusionbonding of metal/ceramic interfaces — Amodel study at the Nb/Al2O3 interfaces. Ed. J.A. Pask and A. G. Evans. Proceedings of Ce-ramic Microstructures ’86. New York, N.Y.:Plenum Press. pp. 295–305.

19. Fischmeister, H. F., Mader, W.,Gibbesch, B., and Elssner, G. 1988. Prepara-tion, properties, and structure of metal/oxideinterfaces. Mater. Res. Soc. Symp. Proc. Ed. W.A. T. Clark, C. L. Briant, and M. H. Yoo. Mate-rials Research Society, pp. 529–40.

20. Rühle, M., and Evans, A. G. 1989. Struc-ture and chemistry of metal/ceramic interfaces.Mater. Sci. Eng. A107: 187–87.

21. Burger, K., and Rühle, M. 1989. Mater-ial transport mechanisms during the diffusionbonding of niobium to Al2O3. Ultramicroscopy29(1–4): 88–97.

22. Mader, W. 1989. Structure and chemistryof interfaces between oxide precipitates and ametal matrix. Z. Metallkde. 80(3): 139–51.

23. Kuwabara, M., Spence, J. C. H., andRühle, M. 1989. On the atomic structure of theNb/Al2O3 interface and the growth of Al2O3particles. J. Mater. Res. Soc. 4(4): 972–77.

24. Ohuchi, F. S. 1989. Surface science stud-ies of Nb-(0001)Al2O3 interfacial reactions. J.Mater. Sci. Lett. 8(12): 1427–9.

25. Mayer, J., Dura, J., Flynn, C. P., andRühle, M. 1990. Electron microscopy studies ofNb-Al2O3 interfaces formed by molecular beamepitaxy. Surface and Coatings Technol. 43/44:199–212.

26. Reimanis, I. E. 1992. Pore removal dur-ing diffusion bonding of Nb-Al2O3 interfaces.Acta Metall. Mater., Suppl. 40: S67–74.

27. Elssner, G., Korn, D., and Rühle, M.1994. The influence of interface impurities on

fracture energy of UHV diffusion bondedmetal-ceramic bicrystals. Scr. Metall. Mater.31(8): 1037–42.

28. Gupta, V., Wu, J., and Pronin, A. N.1997. Effect of substrate orientation, rough-ness, and film deposition mode on the tensilestrength and toughness of niobium-sapphire in-terfaces. J. Am. Ceram. Soc. 80(12): 3172–80.

29. Soyez, G., Elssner, G., Rühle, M., andRaj, R. 1998. Constrained yielding in niobiumsingle crystals bonded to sapphire. Acta Mater.(USA) 46(10): 3571–81.

30. Zhang, W., and Smith, J. R. 2000. Stoi-chiometry and adhesion of Nb/Al2O3. Phys. Rev.B 61(24): 883–89.

31. Massalski, T. B. 1990. Binary Alloy PhaseDiagrams. Metals Park, Ohio: ASM Interna-tional.

32. Hodkin, E. N., Nicholas, M. G., andPoole, D. M. 1970. The surface energies of solidmolybdenum, niobium, tantalum and tungsten.J. Less-Common Met. 20(2): 93–103.

33. De Lima, O. F., Krehl, M., and Schulze,K. 1985. Wetting characteristics of copper onniobium. J. Mater. Sci. 20(7): 2464–70.

34. Locatelli, M. R., Tomsia, A. P.,Nakashima, K., Dalgleish, B. J., and Glaeser, A.M. 1995. New strategies for joining ceramics forhigh-temperature applications. Key Eng. Mater.111-12: 157–90.

35. Liu, W., Elssner, G., and Rühle, M. 2001.Effects of thin film Nb interlayer in Cu/sapphirebonds. Mater. Sci. Eng. A317: 153–62.

36. Girifalco, L. A., and Good, R. J. 1957. Atheory for the estimation of surface and inter-facial energies I. Derivation and application tointerfacial tension. J. Phys. Chem. 61(7): 904–8.

37. Jilavi, M. 1994. Ph.D. dissertation.Stuttgart, Germany: University of Stuttgart.

38. Eustathopoulos, N., Chatain, D., andCoudurier, L. 1991. Wetting and interfacialchemistry in liquid metal-ceramic systems.Mater. Sci. Eng. A135: 83–88.

39. Li, J. G. 1992. Wetting and interfacialbonding of metals with ionocovalent oxides. J.Am. Ceram. Soc. 75(11): 3118–26.

40. Suganuma, K., Okamoto, T., Koizumi,M., and Shimada, M. 1986. Effects of surfacedamage on strength of silicon nitride bondedwith aluminum. Advanced Ceram. Mater. 1(4):356–60.

41. Suganuma, K., Miyamoto, Y., andKoizumi, M. 1988. Joining of ceramics and met-als. Ann. Rev. Mater. Sci. 18: 47–73.

42. Suganuma, K., Okamoto, T., andKoizumi, M. 1989. Effect of surface grindingconditions on strength of alumina/niobiumjoint. Ceram. Eng. Sci. Proc. 10(11-12):1919–33.

43. Hitchcock, S. J., Carroll, N. T., andNicholas, M. G. 1981. Some effects of substrateroughness on wettability. J. Mater. Sci. 16(3):714–32.

44. Shalz, M. L., Dalgleish, B. J., Tomsia, A.P., and Glaeser, A. M. 1993. Ceramic joining. I.Partial transient liquid-phase bonding of alu-mina via Cu/Pt interlayers. J. Mater. Sci. 28(6):1673–84.

45. Shalz, M. L., Dalgleish, B. J., Tomsia, A.P., and Glaeser, A. M. 1994. Ceramic joining IIPartial transient liquid-phase bonding of alu-mina via Cu/Ni/Cu multilayer interlayers. J.Mater. Sci. 29(12): 3200–8.

46. McKeown, J. T. 2003. Processing andcharacterization of alumina/niobium inter-

McKeown Supplement 3/05corr 4/1/05 4:41 PM Page 50

WELDING RESEARCH

-s51WELDING JOURNAL

faces produced via liquid-film-assisted joining.M.S. thesis. Berkeley, Calif.: University of California.

47. McKeown, J. T., Sugar, J. D., Gronsky,R., and Glaeser, A. M. 2004. Niobium silicideformation at alumina-niobium interfacesprocessed via liquid-film-assisted joining. Un-published research.

48. Derby, B., and Wallach, E. R. 1982. The-oretical model for diffusion bonding. Met. Sci.16(1): 49–56.

49. Derby, B., and Wallach, E. R. 1984. Dif-fusion bonding: Development of theoreticalmodel. Met. Sci. 18(9): 427–31.

50. Sugar, J. D. 2003. Mechanisms of metal-lic-interlayer/ceramic interface microstructuredevelopment during liquid-film-assisted bond-ing. M.S. thesis. Berkeley, Calif.: University ofCalifornia.

51. Saiz, E., Tomsia, A. P., and Cannon, R.M. 1999. Energetics and atomic transport at liq-uid metal/Al2O3 interfaces. Acta Mater. (USA)47(15): 4209–20.

52. Dauskardt, R. H., Lane, M., Ma, Q., andKrishna, N. 1998. Adhesion and debonding ofmultilayer thin film structures. Eng. Fract.Mech. 61(1): 141–62.

53. Wenzel, R. N. 1936. Resistance of solidsurfaces to wetting by water. Ind. Eng. Chem.28(8): 988–94.

54. Shuttleworth, R., and Bailey, G. L. J.1948. The spreading of a liquid over a roughsolid. Disc. Farad. Soc. 3: 16–22.

55. Nicholas, M. G., Crispin, R. M., andFord, D. A. 1984. Some effects of surface tex-ture on melt-mold interactions during invest-ment casting. Proc. Brit. Ceram. Soc. 34: 163–72.

Appendix

Surface roughness can impact the wet-ting characteristics of the surface and liq-uid redistribution. A number of variableshave been defined to characterize the sur-face roughness including Ra the average

deviation in surface height, la the averagedistance between peaks of surface fea-tures, r the roughness parameter, and aathe average slope of surface features. Val-ues of these parameters are given in Table1, and some are used in the discussion. Abrief summary of these parameters andtheir significance is provided here.

Wenzel (Ref. 53) proposed that an in-crease in surface area of a roughenedplane causes a change in contact anglegiven by

cosqR/cosq0 = r (1)

where qR and q0 are the contact angles ofsessile drops on, respectively, rough andsmooth horizontal surfaces and r is theroughness factor, equal to the ratio of thetrue area to the apparent (projected) area.Asperities on a rough surface can act asbarriers to liquid flow, which can signifi-cantly affect the contact angle predictedby the Wenzel relation. Shuttleworth andBailey (Ref. 54) developed the followingrelation:

qR = q0+am (2)

where the angle am represents the maxi-mum inclination that surface featuresmake with the average plane of the sur-face. While the Wenzel model does notaccount for hysteresis (a difference in con-tact angles between advancing and reced-ing liquid fronts), Shuttleworth and Bai-ley’s analysis showed that am could bepositive or negative (and therefore qRgreater or less than q0) depending onwhether the liquid front was advancing orreceding. Minimization of the surface en-ergy of the drop led to the conclusion that

an advancing front comes to rest on a de-scending slope (am positive) and a reced-ing front comes to rest on an ascendingslope (am negative).

Hitchcock et al. (Ref. 43) and Nicholaset al. (Ref. 55) investigated the wetting be-havior of various liquid metals on ceramicsurfaces. The irregularity of the ceramicsurfaces made it difficult to derive valuesof r and am using profilometry. Using thestatistical parameters Ra, the average de-viation in height of random points on theceramic surface from a line drawn suchthat the cross-sectional areas of asperitiesabove and grooves below are equal, andla, the average distance between peaks ofsurface features, values of the roughnessfactor, r, and the average slope of surfacefeatures, aa, were obtained from the fol-lowing expressions:

r = 1+K1(Ra/la)2 (3)

aa = tan–1K2(Ra/la) (4)

where K1 and K2 are arithmetic factors de-pendent on surface topography equal toabout 50 and 9, respectively. For values of(Ra/la) up to about 0.06, tan aa is a nearlylinear function of aa and Equation 4 canbe simplified to

aa @ 500(Ra/la) (5)

when aa is expressed in degrees. A linearrelationship was observed between qR andaa:

qR = q0+aa (6)

This relationship is in agreement with theanalysis of Shuttleworth and Bailey.

CAN WE TALK?The Welding Journal staff encourages an exchange of ideas with you, our readers. If you’d like to ask a question, share an idea or

voice an opinion, you can call, write, e-mail or fax. Staff e-mail addresses are listed below, along with a guide to help you interact withthe right person.

Publisher/Editor Andrew Cullison cullison@aws.org, Extension 249Article Submissions

Senior EditorMary Ruth Johnsenmjohnsen@aws.org, Extension 238Feature Articles

Associate Editor Howard Woodwardwoodward@aws.org, Extension 244Society News, Personnel

Assistant Editor Kristin Campbellkcampbell@aws.org, Extension 257New Products, New Literature

Production Editor Zaida Chavezzaida@aws.org, Extension 265Design and Production

Advertising Sales Director Rob Saltzstein salty@aws.org, Extension 243Advertising Sales

Advertising Production Coordinator Frank Wilsonfwilson@aws.org, Extension 465Advertising Production

Advertising Sales & Promotion Coordinator

Lea Garrigangarrigan@aws.org, Extension 220Production and Promotion

Peer Review Coordinator Doreen Kubishdoreen@aws.org, Extension 275Peer Review of Research Papers

Welding Journal Dept. 550 N.W. LeJeune Rd. Miami, FL 33126 (800) 443-9353FAX (305) 443-7404

McKeown Supplement 3/05corr 4/1/05 4:41 PM Page 51