Electronic Structure of Metal-Ceramic Interfaces

ISIJ International, Vol. 30 (1990), No. 12, pp. 1059-l065

Fumio S. OHUCHIand Qian ZHONG1)

Central Research and DevelopmentDepartment, E. l. DuPontde Nemoursand Company,Experimental Sation, Wilmington, DE19880- 0356, U.S A. 1)Department of Materials Science and Engineering, University of Pennsylvania, Piladelphia, PA19104-6272, U.S.A. (Received on March 1, 1990, accepted in the final form on May18. 1990)

Increasing technological applications of metal-ceramic systems have demandeda fundamental understanding of the properties of interfaces In this paper, wedescribe our approach to the study of electronib structure of metal-ceramic intertaces. Electron spectroscopies have been used as primary techniques to investigate the interaction betweenmetal overlayers and ceramic substrates under various experimental conditions. Thesedata are further elucidated by theoretical calculations, from which the electronic structures of the interface have beendeduced Atemperature dependenceof the band structures of Al203 is first discussed, then the evolution of the electronic structure and bonding of Cuand Ni to Al203 is studied. The relationship between electronic structures and interfacial properties are also addressed.

KEYWORDS:metal-ceramic interface; alumina; copper; nickel; electron spectroscopy; electronic structure; UPS; XPS; LEELS.

Introduction have the capability of resolving the distinct nature of 1. evolving chemical and electronic states of both metal Increasing technological applications of metal-ce- and ceramic components. Lastly, the experiments ramic systems, such as structural and electronic mate- need to be free from artifact, particularly contamina- rials, have generated great interest in a variety of' sci- tions, thus an ultra high yacuum(UHV) condition is entific issues concerning interfacial properties. Me- required so as to control the environmcnt. chanical and structural aspects of the interfaces, and With these features highlighted, wedescribe a tech- their relationship have been most frequently studied nique representing a comprehensive methodology for in the past, because these macroscopic properties are the study of' electronic structures of rnetal-ceramic directly relevant to the mechanical strength of the interfaces. As the uses of Al203 expand into new interfaces. While electronic structure of the inter- and more demanding_applications, we first study the faces can be the basis for the macroscopic properties, surf~ce electronic structure of A1203, then discuss its this has not been experimentally investigated until interface interactions with Cu and Ni, in particu]ar, recently. This paper deals with the experirncntal focusing on the electronic structures at early stages of methodologyused to study the electronic structures of the interface formation. The kno\4'1edge thus obtained will assist in overall metal-ceramic interfaces. - understanding the prop- The study of metallic monolayers on well charac- erties, and more importantly, provide a basis for terized material surfaces pro_vides a fruitful approach controlling the structure and compo_sition of the inter- to the clarification of the physics and chemistry of face. adsorbate-adsorbate and adsorbate-substrate inter- characteristics. action Major advances have been 2. Experimental Approach achieved through applications of modernsurlace science techniques, which provide information about the Our experimental approach to metal-ceramic in- electronic and structural properties of such systems. terface studies involves the preparation of the inter- Wehave adopted similar concepts to the investigation faces via formation of metal layers onto well charac- ol' the intrinsic bonding and clectronic properties at terized ceramic surfaces in UHVunder controlled the metal-cerarnic interfaces. To demonstrate these, conditions. This process is schematically illustrated however, several criteria must be met in designing in Fig. l. Wehave designed the experimental ap- and performing_ the experiments. First, the interface paratus, as seen in Fig. 2, such that we are able to must be fabricated so that variation of the chemical control thc init.ial condition of ceramic surfaces, the or electronic state of each componentcan be probed rate of interface formation, and the incorporation of during formatio_ n of' the interi~Lce. Secondly, the con- desired dopants. Details of the experimental method figuration of the experimental apparatus is such that have been discussed elsewhere.1) Electron spectro- is the measurement compatible with the interface scopies have been used as primary tools for the ad- formation in terms of their geometry. Thirdly, the vantage of their sensitivity to different electronic experirnental techniques that have been chosen must states and a range of probing depths.

C1990 ISI J l059 ISIJ International. Vol. 30 (1990). No. 12 metal (TOp View] (Side View] vapor ceramic surface ~~ e ' e . LEED e e' Metal Sources MS Molecular Doser ~¥ X-ray source metal-ceramic interface Sample / ~~~ lon Gun / ~ F ' _+~ CMA \ ) f Optica[ Microscope ~ Sample I Heater "'~ " Al-Ka Fig. 1. Concept diagram for the Energy' Metal source Analyze\_r -- Monochromatized with metal-ceramic interface x_ray Diff. Pump Energy Ana[yzer lon b'lolecu[ar Doser study. Gun Metal Source Fig. 2. Schematic diagram for thc apparatus for the metal-ceramic interface studies.

Ultraviolet photoelectron spectroscopy (UPS) is a mid 10-11 torr range, while the pressure maybe in- technique that is used to characterize the surface creased to the 10-ro torr during the deposition. The electronic structure or ~ensity of states (DOS)of' clean metal source is held at a fixed temperature, thus or adsorbate covered solid surfaces.2) In this tech- yields a constant vapor pressure, from which the vapor nique a collimated beamof monoenergetic ultraviolet flux at the specimensurface maybe calculated. The photons is directed at the surface of interest. The metal deposition rate is typically maintained in the ~nergy ~istribution curve (EDC) of electrons that are range of O.1-1.0A/min. This means the time re- photoemitted is then determined and the occupied quired to form one equivalent monolayer is on the electron DOScan be inferred from the EDC. Be- order of 3-30 min. The time dependent changes of cause the escape depth for photoelectrons excited by the signals are monitored during metal deposition. typical photons is short, this technique samples only If necessary, the local atmosphere can be chan~"ed by the surface energy levels. Thus, the experimental delivering gas from a molecular doser located near DOScurve can be comparedwith theoretical surface the specimen. DOScalculated by appropriate methods. Also, in- In order to ensure the stoichiometry and the crys- formation about surface bonding and surface reac- tallographic structure of the ceramic surfaces, the tivity can be obtained. substrate routinely underwent the following proce- ~~-ray photoelectron ~Pectroscopy (XPS) is a simi- dures before any run of experiment: in case of Al203, lar technique but uses photons with muchhigher ex- an Ar ion-sputtering (5 x l0-5 torr, 1-3 kV, 25 mA) for followed in citation energies, and as a result, both EDCofvalence about 10 min, by annealing UHVat 10-5 electrons and core electrons are measured. There- l OOO'Cfor 5-lO min, and in oxygen (Po,=5X fore, XPSis used to determined the electronic struc- torr) at I OOO'Cfor 5min. ture oi' solid surfaces as well as to chemically identify surface components. Low-energyelectron energy loss 3. High Temperature Electronic Structure of spectroscopy (LEEI.S) gives information about transi- Al203 l tions between initial and final states, in particular, excitations from valence or relatively shallow core Al203 rs a mixed ionic and electronic conductor at states into both bulk and surface-related final states. high temperatures. The high temperature electrical Combining LEELSand photoemission spectroscopies conductivity behavior has been studied extensively in (UPSand XPS)during the course of' metal-ceramic the past to elucidate the defect chemistry with a goal interface formation, information can be obtained of' understanding the mass transport associated with .joints, l about the jo_ int density oi' both filled and emptystate; sintering process.3) In metal-ceramic high from this an energy level schemefor the interface of temperatures are usually involved in the process, interest can be deduced. While the primary empha- therefore, a clear understanding of the electronic sis for the techniques described here is on the study structure of Al203 at high temperatures is important. of the electronic structure, additional information Measurementsof absorption edges and reflectivity about interface chemistry and structure, as well as, peaks of Al203 with increasing temperature using the mechanismsof' interface formation can be ob- yacuumI~ltra-yiolet (VUV) spectroscopy have been tained. madeonly recently by French.4) This is a first di- The interface is fabricated at the surface by de- rect measurementof this kind. The conclusion made positing metal atoms from a differentially pumped frorn this study was that the temperature depen- is electronic leads decrease metal source. The base pressure maintained in the dence of the structure to a l060 ISIJ International, Vol. 30 (1990), No. 12 in the optical band gap energy from 8.8eV, at teracted by a broadening of the valence band caused 23'C, to 7.2eV, at 1490'C with a linear rate of by electron-phonon interaction. The net effect of meV/'C. Theseobservations further eluci- both terms is to produce virtually change in the - I . I were no dated by the theoretical work on the temperature valence band width with increasing temperature. dependenceof' the electronic structure, where calcu- The band-gap temperature dependence therefore lations were madeby incorporating thermal lattice arises from the broadening of the conduction bands, shift expansion as well as various distortions of the lattice duc to the electron-phonon interaction, and the to qualitatively model the effect of phonons on the of conduction bands, due to the thermal expansion. electronic band gap. Details are cited elsewhere.5) It was found that effects of both thermal lattice ex- 4. Electronic Structure of Cu-Al203 Interfaces pansion and electron-phonon interaction contribute linearly to band gap temperature dependence. Esti- Diverse and valuable technological application of mates madeon the basis of the co.mbined effects metal-Al203 systems as electronic and structur.al ma- it f'or research, reveal that one third of the temperature dependence terials made a favorable subject basic is attributed to the thermal lattice expansion with the yet only a few fundamental theoretical studies have remainder arising from the electron-phonon inter- beenmadeso far. Johnsonand Pepper7) have studied action. Temperature dependent valence band mea- the bonding of metals to the cluster (AI06)-9 which surements give further insight into the origin of the represents the Al203 surface. The general conclu- observed band gap temperature dependence. Shown sion is that a chemical bond is formed between the in Fig. 3are the occupied DOSofA12O3as a function d-orbital clectrons and the nonbonding p-electrons of in- of temperature measured by XPS. A co_mparison the O anions. They further suggested that an with the computed total and A1 part,ial density of creasing numberof occupied metal Al203 antibond- states at roomtemperature is also shown. The O(2p) ing orbitals explained qualitatively the observed de- valence band is observed to consist of two main peaks crease in contact shear strength through the series Fe, with a total band width of 7eV. The peak near the Ni, Cu and Ag,8) thus, the decrease oi' the energy upper edge of the valence band corresponds to the gained in the bondformation. Theseconclusions were O(2p) nonbonding orbital, and the other peak con- generally confirmed by recent calculations performed sists ol' the bonding O(2p) orbital, with someAl(3s) by Nath and Anderson9) and Kohyamaet al.10) and (3p) admixture. With increasing temperature, Previously wehave attempted a fully self-consistent, of in- there was no observable difference in the band width total energy computat.ion, of the properties an finite adsorbate.11) as well as the valence band edge position. Thus, the Al203 film with Cu The main valence bands appear to be insensitive to the com- conclusion was that Cu atoms bond to surface O bined effects of the lattice expansion and the electron- atoms instead of surface A1 atoms. This assignment phonon interaction. Wehave calculated that the was based on the result of a lower total energy for lattice expansion leads to a small decrease in the Cu-Osurface bonds than that for Cu-AI bonds. In valence-band width.6) This valence band width de- Fig. 4, the energy band for a half monolayer of Cu crease is due to the lattice expansion and is coun- coverage is shown. Details for this calculation can be found elsewhere.11) This banddiagram indicates that the d-bands overlap the top of the filled O(p) states (1012) Al203 Valence Band XPS and the Fermi encrgy occurs at a hybridization be- (Al-Ka source] tween the O(p) band and the Cu(s) orbitals. As most of the O(_p) band are below the narrow Cu(d) bands, it would appear that the Cu(d) bands are in 1000'C the band gap of the Al208 bulk band, as a first ap- proximation. = ~ A very careful UPSwork was conducted by de- c9 682 'c positing Cuonto Al203 surface successively from sub- :5 monolayer covcrage up to several monolayers at room t c' temperature. To search for any overlapping_ of or- ~ 418 'C bitals occurring between Cu and Al203, the surface gradually heated dri_ve off the e= was then to Cu over- o layers and performed at each = UPSmeasurementwas 25 'c step of temperature rise. Shownin Fig_s. 5(a) and 5(b) are a series of spectral changes for the evolution Total DOS of the Cu overlayer on the Al203, at room temperature, and with increasing temperatures, respectively. In Fig. 5(a), for a clean Al203 surface is similar Al DOS EDC to what was discussed above. The only difference is o -15 -10 -5 Ef +5 +1o +1 that the present spectrum was taken by UPS,so that Electron Binding Energy (eV) the sampling depth and the photoionization cross 3. sections different. initial of the Fig. Temperature dependence of valence band of a- are For coverage Al203 comparedto the calculated DOS, Cu deposition, the emission from Cu(3d) orbitals

l061 ISIJ International, Vol 30 (lggO), No. 12

(a)

2 CulAi203 7

O 6 = 1 = 5 18 > ,g .2 2 n 4 \ cu (d) Is 17 3 ~ :¥~, co 3 a,c 5 :~ 13 = 6 o, 2 7 ~ -6 11 8 6 1 .8 14 12 10 8 6 4 2 O Binding Energy (eV) Ef -1 2 O (a) Cucoveragc dependence: 1 1=Al203 substrate 2=0, I monolayer Cucoverage 3=O.2 monolayer X r Y 4=0,33 monolayer :~l Fig. 4. Band structure of the Cu-Al203 interface. This 5=0,5 monolayer calculation for wasmade 0.5 monolayerCucoverage 6= I.3 monolayer ol' l.83: A. with Cu--O interatomic distance The 7=3monolayer Cucoverage at ro(~m temperature Al203 substrate has 18 filled pstates. (b) slightly higher binding than those appears at energy CuIAI203 from metallic Cu, but the shift is no more than 0.5 eV. Theseresults indicate that copper interacts very 1 weakly with forms clusters Al203 and as evidenced by = = the observed binding energy shift (i.e., Cu-Cuadatom > 2 interactions). After approximately 3 monolayers of ,g Cu deposition over Al203, no contribution from the ~ is ca l substrate seen and Cu(3d) and Cu(4s) spectra re- 4 present the metallic Cuoverlayer. Whenthe surfac.e :b co 3 is gradually heated, a successive change in is c EDC o observed (Fig. 5(b)). At 210'C, the Cu overlayer i its appears to change morphologyas evidenced by the 2 decrease in Cu(3d) peak height and the appearance of Al203 substrate feature in the spectrum. With 1 3 fhrther heating to higher temperatures, the main Cu(3d) peak decreases its intensity substantially, and 4 splits into two components. While one component is simply the attenuation of the metallic Cu(3d) peak 14 12 10 8 6 4 2 O maintaining> the binding of' metallic Cu, same energy Ef a second peak appears 1.5 eVhigher than the metallic Bindlng Energy (eV) Cu binding energy. This is clearly seen in EDC (b) Heating effects: after heating to 310'C. By 400'C, most of the rne- l =Aftcr 3monolayers of Cuat 25'C tallic Cu overlayer is now thermally desorbed and 2=After heating to 2lO'C only the second componentof C,u(3d) derived peak 3=Heating to 310'C remains. This pe.ak appears at the higher binding 4=Heating to 400'C energy side, nearly overlapping the substrate DOS, Fig. 5. UPSspectra for Cu-Al203 interfaces. which is attributed to the orbitals of'the tightly bound Cu atoms to the Al203 surface. Note that this ap- picture is very consistent to what was predicted by parent shift is not due to a re-clustering of' Cuatoms, calculation, where the charge transfer from Cu-d but must be due to the electronic interaction. This states to the non-bonding O(2p) orbitals of Al203 l062 ISIJ International, Vol. 30 (1990), No. 12

5. Electronic Structures of Ni--AI208 Interfaces CuIAI203 Bonding of nickel often involves for- 1 to alumina a f' mation o_ a nickel-aluminate spinel at the intcrlace. While the formation of NiAl204 rs believed to pro- ceed through counterdiffusion of Ni+2 and Al+3 ions across the interface,12,13) the role of oxygen in bonding has not been well understood. This motivatcd us to ~~~~ study the Ni-A]203 interl~cial chemistry in detail using our surface science approach. Details of the results elsewhere.14) Brief- e experimental can be found = ly summarizing, identified of' in- > we have tow types c9 teractions at the Ni-Al203 mterface, in which the is inter- JQ l* absence or presence of oxygen crucial to the ,:, 2 facial structure. In the absence of oxygen, a Ni-Al :b lr intermetal]ic alloy is formed, whereas a NiAl204 ,n / = ~*"/ spinel formation is identified in the of o presence oxy- = gen only if the oxygen partia] pressure is high enough i to maintain the stoichiometry of nickel oxidation prod- jl, uct to be NiO at the interface. Little work has been *~'/ reported for the electronic structure of NiAl204 either j' 3 experimentally or theoretically. The early stage of the spinel phase at the interface maybe different from \ that ol' the bulk fbrm. It is, therefbre, oi'our interest to study the electronic structure at the Ni-Al203 m- terface. XPSand LEELShave been used to obtain a preliminary picture of the e]ectronic structure of the interface. 14 12 10 8 6 4 2 O Weshowfirst an examination by LEELStechnique ':,jl Binding Energy (eV) Ef of a single crystal A12O3to mapout the band structure. Shownin Fig. 7(a) is a LEELSspectrum for Fig. 6. I=UPSspectra for Cu-Al203 Interfaces fabricated A12O3at the incident energ_~y of 179 eV. In the. Iower in the presence of oxygen at 25'C loss region ( 30 eV), energ>y loss peaks at 3.3, 2 Subsequentheating to 450'C energy = 9.4, 12.6, 15.7 21 .4 observed. higher 3=Subsequentheating to 580"C and eVare The 4=Al203 substrate UPSspectrum for reference energy loss peaks are shown, at 34,7, 77.1 and 79.6 eV. The significance of I*EELS is that these transitions can be assigned whencoupled with XPS, and :il occurs. the local surface electronic structure can be dcduced While the overlayer the alumina from the data. Here of the procedures. Cu can stay on weshowsome ~ll" substrate, the numberof Cu atoms actively interact- Based on a free electron model,15) we have assigned ing with surface oxygen is rather limited. This ex- peaks observed at 21.2 and 12.9 eV as bulk and sur- plains qualitatively whythe Cu film does not adhere face plasmons, respectively. Loss peaks representing well to the alumina surface. However this situation interband transitions from the core levels to the con- maybe changedwhenCuoverlayer is formed in the duction band are assigned first. The loss peak at presence ofoxygen. Shownin Fig. 6is a comparison 79.6eV corresponds to a transition from Al(2p) at of UPSspectra for Al203 surfaces, after (1) Cu de- 75.6 eVbelow the Fermi level to the conduction band, position in I x l0-4 torr oxygen at room temperature, yieldi_ng a conduction band minimum (CBM) at (2) subsequent heating to 450'C, and (3) to 580'C. about 4.0 eVabove the Fermi level. A rnuch broad* is AnUPSspectrum for a clean Al203 Is also shownfor ened peak, centered at about 86.2 eV also con- reference. As expected, the Cu(3d) orbital from the sidered to be transitions from Al(2p) to the conduc- oxygenated Cu overlayer was observed at slightly tion band. The differences in the amplitude and the higher binding energy as comparedto those observed broadening of~ the peaks may be attributed to the for the mctallic Cu. With heating to 450'C, this transition probability associated with the initial and peak splits into two components, exactly identical to final states. Ciraci and Bartral6) have shown, in those observed in Fig. 5(b). In the present case, their electronic structure calculations, the existence of however, the peak heights from the Cu(3d) derived two surl~ce states, S1' and Sa, Iocated near conduction peaks (higher binding energy side) are significantly and valence bands in the band gap, respectively. higher than those observed in Fig. 5(b) at identical These states are produced by the dang_Iing bonds of temperatures. This indicates that more Cu atoms the surface A1 with a considerably small O orbital are involved in the bonding at the interface of the contribution from the bulk. The loss peak at 77.l oxygenated overlayer. In this case, the adhesion eV, therefore, can be assigned to the transition from would be greatly enhanced. the A1(2p) to the surface state, Sp, Iocated near the

roe3 ISIJ International. Vol. 30 (1990), No, 12

oc-Al203 79.6 Interface Splnel (NIAl20,] 7e.4 8e 8

71.

B6.O 70.0 77.1 88.2

21.2 21.4 9.4 ~fuJU p,

Z~UJ 12.e 1:,~ 7.4 ~~U , 3.3 15. l } 34.7 12.9 35.0

(a) (b)

90 O 90 o Energy Los5 (eV] Energy Loss (eV, Fig. 7. LEELSspectra for (a) Al203 taken at 179 eV and (b) " interface " NiAl~04 taken at 178 eV.

CBM. The loss peak at 34.7 eV represents a transi- (a) Al2p3 (b) interface NiAl204 tion from the lower valence ~and (LVB), O(2s), to previously, the 15~!n~~-' Al(3p,3s)3s) the conduction band. As described 9'3 93 yalence band (UVB) consists of O(2p) l~pper an 4'o CBM Ni(3d,4s) 3.0 bonding state and non-bonding state with their peak sp l ~~ EF CBM 0.5 maximalocated at 12.7 and 8.0 eV, respectively. The l~E~_, F ~~~-~~~ Sd -~ loss peaks at 9.4 and 15.7 eV, therefore, correspond VBMTI -3 O Ni(3tNi(3d) -6.0 to transitions from the to the surface state, Sp. 15'o VBM UVB -8'o o(2p) non-bondingn-bonding -8.0 Howevcr, the 3.3 eV peak is consistent with a transi- 12 o(2p) bonding -1 2.7 tion from UVBto another surface state, Sd, which is 6 about 3eV above the yalence band maxima(VBM). An energy band diagram constructed based on these analyses is shownin Fig. 8(a). An estimated band -25 4 O(2s) LVB -25.4 gap of about 9eV is also consistent with the band ~~9 gap value measuredby other techniques. Wenowapply this method to elucidate the electronic structure of " interface " spinel phase. In the Ni(3p) -69 5 LEELSspectrum shownin Fig. 7(b), valance electron E~~~~ -75.6 energy loss peaks positioned at 7,4, 12.9, 21.2 and -75.6 Al(2p) ~~~I 35.0 eV, and the core electron energy loss peaks at Fig. 8. E~~Energy band diagrams for (a) Al203 and (b) 66.5, 70.0, 77.1 and 79,4eV are identified. Again " interface " NiAl204' the 12.9 and 21.2 eV peaks are energy losses frorn the surface and bulk plasmonpeaks, respectively. Analo- the surface states associated with surface dangling gous to those observed in Al203, the 79.4 eVIoss peak bonds of A1203, as a result, the surface state density comesfrom the transition from Al(2p) to an unfilled maybe signiflcantly reduced. This is, in fact, evi- state in the conduction band, mainly consisting of denced by the decrease in the peak intensity at 77.1 .7 Al(3p, 3s). Further, the interaction between Ni~O) eV, as well as, the disappearance of the 3.3 and 15 and Al203 Ieads to a Ni contribution to the c_onduc- eV Ioss peaks in NiAl204. However, the 66.5eV tion band, which maybe seen through the interband peak still predicted to be a transition from the Ni(3p) transition from the Ni(3p) Ievel to C~.BM. NiAl204 to the surface state, S,1' below the Fermi leve]_. The being a double oxide of NiO and Al203, the Ni(3d, 4s) UVBof NiAl204 differs from that of' Al203 in that conduction band overlaps that oi' Al203 by occupying there is a contribution from the Ni(3d), Iocating at l064 ISIJ International, Vol. 30 (1990), No. 12 about 6.0eV below Fermi energy (Ef)' The in- vania. terband transition from the Ni(3d) to the conduction REFERENCES band is seen through the loss peak at 7.4 eV. The in Fig. 1) F. S. Ohuchi: Bonding, Structure and Mcchanical Prop- "il energy levels are thus constructed and shown erties of Metal-Ceramic Interfaces, ed, by M. Rilhle and 8(b). This provides a preliminary picture of the A. G. Evans, PergamonPress, Inc., Elrnsford, NY, (1990), electronic structure of NiAl204 spinel. Thebandgap in press. is estimated to be about 3.5eV. Obviously these 2) G. Ertl and J. Kuppers: LowEnergy Electrons and Sur- preliminary results necd to be further clarified by face Chemistry, VCHVerlagsgesellshc.aft mbH,VVeiheun optical measurements,which are in progress. (1985). 3) F. A. Kroger: " Structure and Properties of MgOand ", in Ceramics, 10, ed. 6. Conclusions Al203 Ceramics Advances Vol. by W.D. Kingery, Am. Ceram. Soc., Inc., Columbus, OH, The atomistic and electronic nature of metal- (1984), 1. 4) R. French: J. Am.Ceram. Soc., 73 (1990), 477. ceramic interfaces provides us a newinsight into how H. dissimilar materials interact each other. The studies 5) R. Car and M. Parrincllo: Phys. Rev. Lett., 55 (1985), 2471. 6) R. H. French, R.L. Coble, R.V. Kasowski and F. S. to datc serve as first view of' the electronic structure a Ohuchi Physica, B150 1988), 47. of metal-ceramic interface. Morework and in depth : ( 7) K. H. Johnsonand S. V. Pepper: J. Appl. Phys., 53 (1982), analysis will further elucidate the fundamental role oi' 6634. the electronic structure of metal-ceramic systems. 8) S. V. Pepper: J. Appl. Phys., 47 (1976), 801. interfacial the "Addition of oxygen reduces lhe energy at 9) K. Nath andA. B. Anderson: Phys. Rez'., B39(1989), 1013. metal-ceram,ic interface, and hence promotes spreading of 10) M. Kohyama,Y. Ebata, S. Kose, M. Kinoshlta and R. metal onto substrate ".17) While this statement is useful Yamamoto: Proc. of MRSInt. Meeting on Advanced in describing the phenomena,the relevance to the Materials, Vol. 8, Metal-Ceramic Joints, Mater. Res. Soc., electronic structure is yet to be demonstrated. The Pittsburgh, PA, (1989), 183. 11) R. V. Kasowski, F. S. Ohuchi and R. French: Physica, present study is, in part, aimed at filling this gap by H. B150(1988), 44. showing the influences of external variables, such as 12) F. S. Pettit, E. H. Randklev and F..J. Felten: J. Am. incorporation of the electronic structure oxygen, on Ceram. Soc., 49 (1966), 109. and ultimately the properties of' interl~ces. 13) G. de Roos, J.H.W. de Wit, J. M. Fluit, J. W. Geusand Acknowledgments R. P. VclLhuizen: Surf. Inter. Anal., 5 (1983), 119. 14) (~L¥ Zhong and F. S. Ohuchi: Mat. Res. Soc. Symp. Proc., 1~he authors like to acknowledge would R.H. 53 (1989), 71. for R. V. Kasowski French VUVmeasurementsand 15) D. Bohmand D. Pines: Phys. Rev., 85 (1952), 338; 92 for band structure calculations. Q. Zhongwishes to (1953), 609. acknowledge the support of the National Science 16) S. Ciraci and I. P. Bartra: Phys. Rev., B28 (1983), 982. oi' Foundation through the Laboratory Research on 17) B. Gallois and C.H.P. Lupis : Metall. Trans. B., 12B(1981), the Structure oi' Matter at the University of Pennsyl- 549.

l065