Materials Science and Engineering A261 (1999) 16–23

Boron-doped molybdenum silicides for structural applications

Mufit Akinc *, Mitchell K. Meyer 1, Matthew J. Kramer, Andrew J. Thom,Jesse J. Huebsch 2, Bruce Cook

Ames Laboratory and Department of Materials Science and Engineering, Iowa State Uni6ersity, Ames, IA 50011, USA

Abstract

The addition of as little as 1 wt.% (=3 at.%) boron improved the oxidation resistance of Mo5Si3 by as much as five orders ofmagnitude over a temperature range of 800–1500°C. The mechanism of oxidation protection is the formation of a borosilicateglass scale that flows to form a passivating layer over the base intermetallic. The compositional homogeneity range for T1(Mo5Si3Bx) phase was determined to be much smaller than that reported previously by Nowotny. Compressive creep measure-ments show that materials based on the phase assemblage of T1-T2 (Mo5SiB2)–Mo3Si have high creep strengths similar to singlephase Mo5Si3. Electrical resistivity of selected compositions was also measured and varied from :0.06 mV-cm at roomtemperature to 0.14 mV-cm at 1500°C. Temperature coefficient of resistivity (TCR) was estimated to be on the order of 1×10−4

C−1 for most compositions. © 1999 Elsevier Science S.A. All rights reserved.

Keywords: Molybdenum silicides; Oxidation protection; Creep strength

1. Introduction

In order to meet an ever-increasing demand on theoperating temperature capability of structural compo-nents, a considerable amount of research has beenfocused on developing new materials. In particular,ceramic matrix composites and intermetallics have re-ceived considerable attention.

Recently, a large body of research has centered onnickel aluminide and titanium aluminide. These materi-als exhibit significant room temperature ductility com-pared to other intermetallics. Unfortunately, theiroperating temperature is typically limited to less than1000°C. The end result is that these materials providelittle material property advantage over existing nickel-based superalloys.

MoSi2 has also been studied extensively and has beenfound to have several significant applications. As anumber of other articles in this volume allude to, MoSi2possesses a number of unique properties, which makes

it an attractive high temperature material. In particular,excellent oxidation resistance up to 1700°C and rela-tively easy processibility makes this material an attrac-tive candidate for structural applications. On the otherhand, low creep strength at T\1000°C, pest oxidationat moderate temperatures, and poor low temperaturefracture toughness are severe limitations and are thesubject of significant current research activity.

Development of single-phase ceramic or intermetallicsystems capable of meeting the demanding require-ments for ultra-high temperature applications doesn’tseem likely in the near future. Additionally, composite-based materials typically pose a large number of pro-cessing challenges which again hinder their near futuredevelopment. Multiphase (in-situ composite) inter-metallics based on Mo–Si–B compositions around theT1 phase present some unique solutions to these prob-lems. The present paper reviews the work conducted bythe authors in this system. Most of the material dis-cussed in the present paper is based on articles previ-ously written by the authors. Experimental details havemostly been omitted, and readers should consult thesearticles for specific details. Where available, new datahas been incorporated in the present paper to bring upto date the state of understanding of the Mo–Si–Bresearch effort at Ames Laboratory.

* Corresponding author.1 Present address: Argonne National Laboratory West, Idaho Falls,

ID 83408, USA.2 Present address: Seagate Technology, Bloomington, MN 55435,

USA.

0921-5093/99/$ - see front matter © 1999 Elsevier Science S.A. All rights reserved.

PII: S09 21 - 5093 (98 )010 45 -4

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–23 17

2. Mo5Si3-a structural intermetallic material

Mo5Si3 is the highest melting compound in theMo–Si binary phase diagram with a melting point of2180°C. In addition, it exhibits a small compositionalhomogeneity range. It crystallizes in tetragonal W5Si3-type structure with no phase transformation fromroom temperature to its melting point, and it meltscongruently. Compared to molybdenum disilicide,Mo5Si3 has a more complex structure with 4 formulaunits per unit cell, hence it is expected to have bettercreep resistance. In fact, compressive creep rate forthis material was measured to be an order of magni-tude lower than that of MoSi2 [1].

Mo5Si3 has been previously considered for hightemperature structural applications. It was quicklyabandoned for mainly two reasons, low oxidative sta-bility at moderate to high temperatures, and microc-racking presumably due to thermal expansionanisotropy. Bartlett, et al. [2], modeled the oxidationof MoSi2 and Mo5Si3 using a kinetic model based onthe rate of dissociation of respective silicides to gener-ate silicon and its consumption by the available oxy-gen at the reaction interface. Though the modelignores the microstructural aspects of the SiO2 scaleformation, it predicts that Mo5Si3 will oxidize activelyin air atmosphere unless the temperature is above1650°C. Thus, Mo5Si3 was considered intrinsically un-stable during oxidation in air at temperatures below1650°C. These predictions were corroborated by ex-perimental data at various oxygen partial pressuresand temperatures by several studies including onescarried out in our laboratory. Cyclic oxidation byAnton and Shah [3], at 1149°C also showedcatastrophic failure in 20 1 h cycles, indicating thatan adherent, passivating layer does not form at thesetemperatures.

3. Mo–Si–B intermetallics as oxidation resistantstructural materials

In order to improve the oxidation behavior ofMo5Si3 by scale modification, the Mo–Si–B systemwas chosen for investigation [4,5]. Compositionsaround the Mo5Si3 compound requiring small boronadditions were selected. These materials were exposedto a flowing air atmosphere at elevated temperatures,and Fig. 1 shows the oxidation behavior of a typicalboron-doped Mo5Si3 composition (16.1 Si, 1.24 B bywt.%). Mass change was measured as a function oftime at temperatures ranging from 800 to 1500°C.For comparison, the oxidation behavior of undopedMo5Si3 over a temperature range of 800–1200°C isalso illustrated.

Common features of oxidation plots for both B-doped and pure Mo5Si3 samples include a small massgain followed by an abrupt mass loss during initialheating to the test temperature. Catastrophic massloss occurs for undoped samples whereas boron-doped samples passivate and show little to no masschange over several hundred hours.

Undoped samples exhibit active oxidation due tothe formation of an initial porous oxide scale and itsinability to subsequently form a protective scale.Small additions of boron to Mo5Si3 promote the for-mation of a non-porous, protective scale as shown inFig. 2. The micrographs indicate that the oxide scaleis several hundred micrometers thick for undopedMo5Si3 (Fig. 2A), whereas the boron-doped Mo5Si3(Fig. 2B) forms a continuous, non-porous scale thatis B10 mm thick.

Obviously, the addition of boron modifies the scalemicrostructure dramatically. Oxidation resistance pro-vided by the boron addition of :1.2 wt.% as previ-ously mentioned is not limited to a narrowcomposition range. Several other Mo–Si–B composi-tions around Mo5Si3 were also tested for oxidationresistance. These materials also exhibit excellent oxi-dation resistance as shown in Fig. 3. The magnitudeof the initial weight loss is dependent on the molyb-denum content of the sample and the B/Si ratio. Ini-tial mass loss ranged from about 2 mg cm−2 forB/Si=0.02 to about 11 mg cm−2 for B/Si=0.24.Again, all compositions show the similar behavior ofa rapid mass loss followed by a region of near zeromass change. The isothermal portions of the oxida-tion curves at long times (several hundred hours)show some clear differences in oxidation rates. Com-positions with low B/Si ratio or low (B+Si) contentshow a negative linear oxidation rate (kl= −3.6×

Fig. 1. Isothermal oxidation of undoped Mo5Si3 and boron-dopedMo5Si3 (Mo 16.1 Si–1.2 B) from 800 to 1500°C in flowing air.Dashed curves denote undoped Mo5Si3 while solid line curves denoteboron-doped Mo5Si3.

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–2318

Fig. 2. Micrographs of scale cross-sections of samples oxidized in air.(A) Undoped Mo5Si3 oxidized at 1100°C for 8 h. Note only smallcenter fraction of the original silicide remains unoxidized. (B) B–Mo5Si3 oxidized at 1300°C for 120 h. External scale is B10 mm. Forboth micrographs, U, underlying base alloy, S, scale.

4. Oxidation protection mechanism

The mass gain followed by large mass loss duringthe initial heating to the test temperature can be at-tributed to oxidation of silicon and molybdenum, fol-lowed by sublimation of MoO3 A clearerunderstanding of transient state oxidation and the ox-idation protection mechanism provided by boron ad-dition was determined by characterizing the initialstages of oxidation for several boron-doped materialsusing SEM (scanning electron microscopy) and XRD(X-ray diffraction) [6].

Fig. 4 shows SEM surface micrographs of the de-velopment of the oxide scale on a specimen with ahigh B/Si ratio. The material is designated as V, andthe composition is given in Fig. 3. The starting mate-rial consists of an Mo3Si matrix with both T1 and T2as distributed phases. At 600°C, the scale is com-posed of small pockets of borosilicate glass, molybde-num oxide grains, and a matrix consisting of mixedmolybdenum and silicon oxides. There are no crys-talline phases in the scale since the XRD pattern at600°C in Fig. 5 only reveals the intermetallic basematerial. At 633°C, MoO3 crystals begin to grow overthe scale surface, and growth continues up to 725°C,as shown in Fig. 4. The vapor pressure of MoO3

increases significantly at 750°C, and MoO3 begins tosublime, evident in Fig. 4(B). By 775°C, a significantfraction of the MoO3 crystals has evaporated, leavingbehind a surface containing borosilicate glass and asizable fraction of porosity (Fig. 4C). At 1000°C, Mopresence is suggested by the appearance of the Mo(100) diffraction maxima. Viscous flow and closure ofsubmicron scale porosity occurs after holding isother-mally for 20 min at 1000°C. The scale looks like aglass surface after 40 min at 1000°C (Fig. 4D), andthere are no longer any signs of MoO3 formation.The enrichment of Mo at the oxidation interface isindicated by the Mo (100) peak that grows in inten-sity with time at 1000°C relative to Mo5Si3, T2, andMo3Si intermetallic peaks near 41° 2U (Fig. 5). Theseobservations suggest the selective oxidation of siliconover molybdenum during passivating scale growth.

For all compositions tested, oxidation occurs intwo stages, an initial transient period followed bysteady state oxidation. For some compositions, a pro-tective scale forms rapidly so that transient mass lossis relatively small and essentially completes by 850°C.On the other hand, transient mass loss continues wellinto the isothermal temperature regime and a protec-tive scale forms more slowly where Si+B content isrelatively low or B/Si ratio is low.

There is a transition to a slower, steady state oxi-dation regime, regardless of the magnitude of the ini-tial mass loss. This transition is the crucial step in

10−3 mg cm−2 h). Compositions with a higher B/Siratio or higher (B+Si) content exhibit a near zero topositive linear oxidation rate (kl= +7.3×10−4 mgcm−2 h).

Fig. 3. Isothermal oxidation of Mo–Si–B compositions I–V at1000°C. Silicon and boron content for each composition is indicatedby its respective curve.

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–23 19

Fig. 4. SEM micrographs of the surface of composition V to demonstrate formation of the initial oxide scale. (A) 725°C, (B) 750°C, (C) 775°C,and (D) 1000°C for 40 min. Samples in A–C were immediately quenched to room temperature after exposure to the test temperature.

formation of a protective surface layer. If the substratesurface is sealed by a continuous scale, oxygen trans-port through the scale to the reaction interface occursby atomic scale diffusion. On the other hand, a microp-orous scale will force oxygen and oxidation products todiffuse through a torturous path to the substrate sur-face. The resulting passivating scale will be rate-limitedby diffusion of oxygen and oxidation products throughthe porous scale.

If oxidation is limited by atomic oxygen diffusionand the silicon activity at the oxidation interface issufficient, oxygen partial pressure will be fixed by theSi/SiO2 equilibrium [7]. Silicon dioxide has a muchlower free energy of formation than any of the molyb-denum oxide species and hence molybdenum will notoxidize. The dominant oxidation reaction may be ex-pressed as:

Mo5Si3+3O2�5Mo+3SiO2 (1)

A silicon depleted and molybdenum rich interlayer

will thus form. It has been shown that SiO2 can exist inequilibrium with Mo and Mo5Si3 so that formation of aMo3Si layer according to Mo–Si phase diagram is notthermodynamically necessary [8]. Eq. (1) predicts a netmass gain on oxidation. Fig. 4 describes a compositionthat exhibited an isothermal mass gain and formationof a molybdenum interlayer after heating to 1000°C.For this composition, oxidation at 1000°C is controlledby oxygen diffusion through a completely passivatingscale.

If oxidation is limited by diffusion of oxygen througha microporous layer, the partial pressure of oxygen atthe interface is high enough to oxidize molybdenum.Oxidation will proceed according to:

2Mo5Si3+21O2�10MoO3 (volatile)+6SiO2 (2)

A mass loss is predicted by Eq. (2) since molybde-num oxide is volatile above :750°C. Also noteworthyis that a metallic molybdenum interlayer will not form.A slow steady state mass loss was observed for compo-

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–2320

sitions with low Si+B content or low B/Si ratio. Nomolybdenum interlayer was detected by XRD, andtherefore Eq. (2) is the dominant oxidation reactionfor these compositions at 1000°C. A transition frommass loss to near zero mass change was observed forintermediate compositions after about 30 h of oxida-tion. In this case, Eq. (1) and Eq. (2) must be com-peting to yield an essentially zero net mass change.Another possibility is that micropores close off withtime so that the extent of Eq. (2) decreases with timeto become negligible.

For the sake of simplicity, the preceding discussionneglects the presence of boron and/or other phasesusually present such as T2 and Mo3Si. This assump-tion however, does not change the qualitative assess-ment of the oxidation mechanism proposed.

It is important to consider the scale porosity andscale viscosity when describing the transition tosteady state oxidation. If viscous flow does not occurto close pores on a reasonable time scale, initial scaleporosity and pore size will affect the steady state oxi-dation rate. The steady state oxidation rate will becontrolled by the diffusion rate of oxygen through thescale if viscous flow of the scale to form a coherentpassivating layer does occur. Fig. 6 shows the varia-tion of isothermal oxidation rate at 1000°C with theestimated viscosity of the scale. The steady state massloss rate does increase with increasing scale viscosity.The scale viscosity was calculated assuming that theB/Si ratio in the scale is identical to the initial B/Siratio of the substrate [9].

Fig. 6. Variation of isothermal mass loss rate at 1000°C with thecalculated viscosity of the scale. The atomic B/Si ratio for eachcomposition is given.

5. Mo–Si–B ternary

All of our oxidation tests were conducted on threephase samples. T1 was usually the matrix phase withT2, Mo3Si, MoSi2, or MoB as the distributed phases,depending on the overall composition point. There-fore, it was not possible to accurately assess the rolethat individual phases play in the oxidation behavior.Realizing the need to assess the oxidation behavior ofsingle-phase T1, we began a concerted effort to syn-thesize phase pure T1 [10]. A number of compositionswere prepared using the 1600°C isothermal cut of theternary phase diagram according to Nowotny shownin Fig. 7. Numerous preparations with careful consid-eration of the starting materials failed to produce sin-gle phase T1. The resulting multi-phase assemblagematerials were analyzed by XRD and EMPA (elec-tron micro probe analysis). These data suggested thateither Nowotny’s phase diagram was not accurate orthat equilibrium was not achieved during annealing ofthe chosen points.

However, careful heat treatment periods at 1800°Cand chemical analysis indicated that the single-phasehomogeneity region for T1 is much smaller than thatreported by Nowotny. Fig. 8 shows the narrow ho-mogeneity range measured for T1 phase. During thesestudies, arc-cast samples were originally heat treatedfor 2 h at 1800°C in an attempt to reach equilibrium.XRD patterns indicated that even this high tempera-ture heat treatment was not sufficient for thermody-namic equilibrium. In order to reach the equilibrium

Fig. 5. XRD patterns from 600 to 1000°C for composition V shownin Fig. 4. Dashed lines indicate MoO3. Arrows at top of plot give the100% intensity peak for both molybdenum and cristobalite.

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–23 21

Fig. 7. Isothermal cut at 1600°C of the Mo–Si–B ternary phasediagram [14].

Fig. 9. Arrhenius plot of steady state creep rate with varying com-pressive loads for undoped and boron-doped Mo5Si3.

Preliminary oxidation testing of single phase T1 sam-ples suggests that the material also has good oxidationresistance [11]. This is consistent with the previousoxidation data for which T1 containing materials ex-hibit excellent oxidation resistance. Obviously, the oxi-dation protection afforded to Mo5Si3 by boron dopingis independent of the resulting phase assemblage inwhich it resides. Very small quantities of boron (B3at.%) are sufficient to render the Mo5Si3 oxidativelystable.

6. Creep behavior

As mentioned earlier, Mo5Si3 and boron-doped com-positions around the T1 phase field are expected toexhibit higher creep strength than other silicides due toits complex unit cell structure. For creep testing, bothMo5Si3 and a typical Mo–Si–B sample were investi-gated [12]. Material was prepared by powder com-paction and sintering. The undoped Mo5Si3 material isdesignated A, and the boron-doped sample of composi-tion 13.0 Si–1.3 B (wt.%) is designated B. Sample Bpossessed a three-phase microstructure with T1 as thematrix phase and T2 and cubic Mo3Si as the distributedphases. The sintered specimens were about 97% oftheoretical density as determined by Archimedesmethod. The average grain size for all phases was about4 mm. Compressive creep rates were determined from1220 to 1320°C and stress levels of 140–180 MPa. Attemperatures below 1200°C, and loads less than 100MPa, no measurable deformation was detected.

Fig. 9 shows a reciprocal temperature plot of creeprate for each applied load for both compositions A andB. From the data in Fig. 9, apparent activation energieswere obtained from a least squares fit. The calculatedactivation energy for sample A was 399 kJ mol−1, andthe activation energies for sample B ranged from 377 to

phase distribution, arc-melted samples were groundinto powders, compacted, and sintered in an argonatmosphere for 48 h at 1800°C. Although this heattreatment promoted the equilibrium condition, the va-por pressure of silicon at this temperature is significantand shifts the initial composition to the molybdenumrich side. Nevertheless, the homogeneity range for theT1 phase was determined to be 61.5–62.9% Mo (at.%)and 1 to 2% B (at.%). The maximum solubility for Boccurs at about 62.0% molybdenum content.

Fig. 8. Experimentally determined Mo–Si–B ternary phase diagramat 1800°C in the region of T1 phase.

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–2322

412 kJ mol−1. Since the values for sample B andbetween samples A and B were similar, the same orsimilarly activated creep processes dominate the creepbehavior of both A and B. Fig. 10 shows calculatedcreep stress exponents for B, along with the standarddeviation associated with the calculation of each stressexponent. The stress exponent appears to decrease withincreasing temperature from a value of 5.0 at 1242°C toa value of 3.7 at 1302°C. The errors associated with thecalculation of the stress exponent are in the order of0.5–0.7. Since the differences between stress exponentsare in the order of this error, no conclusions about thechange in creep mechanism with temperature can bedrawn from this data. The average stress exponent forall conditions tested is 4.3. Stress exponents in thisrange often indicate creep by dislocation relatedprocesses.

The creep rate of the undoped and boron-dopedMo5Si3 samples was about the same. A very low dislo-cation density was observed in the T1 phase duringTEM examination of the crept material. Only a fewdislocations were noted in the T2 phase, while a highdislocation density was observed for Mo3Si. At 5%strain levels, Mo3Si actively undergoes polygonizationand shows dislocation climb. Cracks were seen in theT1 phase at higher strain levels (:13%), although thespecimen retained its integrity. The density of crackswas much higher in the T1 phase than in Mo3Si or T2.It is believed that cracking and sliding of the T1 phaseis the mechanism by which deformation is accommo-dated at higher temperatures and strain rates. Sometoughness enhancement for the composite microstruc-ture may be afforded by Mo3Si because cracks ap-peared to be arrested at the Mo3Si phase.

Fig. 11. Temperature dependence of electrical resistivity for boron-doped Mo5Si3 with commercial Kanthal® Super MoSi2 heating ele-ment material for comparison.

7. Electrical resistivity

One potential application for these materials is elec-trical furnace heating elements. This application istargeted for providing improved performance over ex-isting MoSi2 heating elements in use today.

In order to assess the utility of the material aspotential heating elements, we measured the electricalresistivity of select compositions using the four pointmodified Van der Pauw method [13]. Electrical resistiv-ity was measured in air atmosphere up to 1100°C.Measurements up to 1500°C, were conducted in aninert atmosphere. Thin oxide scales that formed in airabove 1100°C contributed to high resistance electrodecontacts and erroneous measurements. Data taken inan inert atmosphere and in air atmosphere comparevery well up to 1100°C. Therefore, the inert atmospheredata up to 1500°C are considered to be reliable andaccurate. Fig. 11 shows the variation of electrical resis-tivity with temperature for select compositions. Su-perkanthal® heating elements were also measured andare compared to the manufacturer’s published data(solid line). The data obtained in our laboratory forMoSi2 agrees well with the published data, and thislends strong support to the validity of the electricalresistivity measurements. The materials developed inour lab have a room temperature electrical resistivity of0.06–0.08 mV-cm, depending on the specific composi-tion. The electrical resistivity increases slightly up to1500°C. Even after several heating/cooling cycles, nohysteresis was observed. No drift in resistivity wasobserved as evidenced by heating a sample up to1500°C and holding at the temperature for 6 h.Mo5Si3Bx compositions have a relatively low tempera-ture coefficient of resistivity (TCR), indicating thatthese materials are excellent candidates for electric fur-

Fig. 10. Creep rate versus stress plot for boron-doped Mo5Si3 show-ing the temperature dependence of the stress exponent. Uncertaintyvalue for each estimate is given in parentheses.

M. Akinc et al. / Materials Science and Engineering A261 (1999) 16–23 23

nace heating elements. Rigorous current control tomaintain certain power levels upon heating may not berequired.

8. Conclusions

The oxidation behavior of Mo–Si–B compositionsnear the T1 (Mo5Si3Bx) phase field was studied from800 to 1500°C. Steady state oxidation in a flowing airatmosphere indicates that the addition of smallamounts of boron dramatically increases the oxidationresistance of these materials. Steady state linear oxida-tion rates on the order of kl= +10−4 mg cm−2 h weremeasured.

Scale formation proceeds with an initial oxidation ofthe surface to form borosilicate glass and crystallineMoO3. Sublimation of MoO3 and reflow of the porousborosilicate glass promotes formation of a passivatingscale. Further oxidation is limited to diffusion of oxy-gen through this scale to the reaction interface.

Boron doping of Mo5Si3 does not appear to decreasethe creep strength of Mo5Si3. Compressive creep data atseveral temperature and stress levels yields a creepstress exponent in the range of n=3.8–5.0. The aver-age activation energy was calculated to be Q=400 kJmol−1. No dislocation motion was evident in the T1phase. Some dislocation motion was observed in T2,and a much higher density of dislocations was observedin Mo3Si.

Electrical resistivity measurements of several compo-sitions indicated that these materials show electricalconductivity much like metals. Electrical resistivities onthe order of 0.1 mV-cm were measured up to 1500°Cwith a TCR in the order of 10−4C−1.

It appears that boron-doped Mo5Si3 compositionsoffer promise for use in high temperature oxidativeenvironments. Oxidative stability, compressive creepstrength, and electrical resistivity characteristics suggestthat these materials have significant potential for struc-tural applications. Several other engineering propertieshave not yet been studied including cyclic oxidation

and mechanical properties such as flexural strength andfracture toughness. Different phase assemblages avail-able around the T1 phase offer the possibility of design-ing microstructures to provide a combination ofchemical, physical, and mechanical properties for agiven application.

Acknowledgements

Ames Laboratory is operated for the US Departmentof Energy by Iowa State University under contractnumber W-7405-ENG-82. This research was supportedby the Office of Basic Energy Science, Materials ScienceDivision. This work was also supported by the Office ofEnergy Research, Office of Computational and Tech-nology Research, Advanced Energy Projects Division.

References

[1] D.L. Anton, D.M. Shah, Mater. Res. Soc. Symp. Proc. 213(1991) 733.

[2] R.W. Bartlett, J.W. McCamont, P.R. Gage, J. Am. Ceram. Soc.4811 (2) (1965) 551.

[3] D.L. Anton, D.M. Shah, Mater. Res. Soc. Symp. Proc. 213(1991) 733.

[4] M.K. Meyer, M. Akinc, J. Am. Ceram. Soc. 794 (4) (1996) 938.[5] M.K. Meyer, M. Akinc, J. Am. Ceram. Soc. 7910 (10) (1996)

2763.[6] M.K. Meyer, A.J. Thom, M. Akinc, Intermetallics, in press.[7] E. Fitzer, in: R.B. Tressler, M. McNallan (Eds.), Ceramic Trans-

actions, Corrosive and Erosive Degradation of Ceramics, vol. 10,American Ceramic Society, Westerville, OH, 1989, p. 19.

[8] R. Beyers, J. Appl. Phys. 56 (1) (1984) 147.[9] M.F. Yan, J.B. MacChesney, S.R. Nagel, W.W. Rhodes, J.

Mater. Sci. 15 (1980) 1371.[10] J.J. Huebsch, M.J. Kramer, M. Akinc, Intermetallics (1998),

submitted for publication.[11] J.J. Huebsch, M. Akinc, Intermetallics (1998), submitted for

publication.[12] M.K. Meyer, M.J. Kramer, M. Akinc, Intermetallics 4 (1996)

273.[13] B. Cook, S. Zelle, M. Akinc, Ceram. Eng. Sci. Proc., in press.[14] H. Nowotny, E. Dimakipoulu, H. Kudiekla, Monatsh. Chem. 88

(1957) 180.

.