J O U R N A L OF M A T E R I A L S SCIENCE LETTERS 14 (1995) 1668-1671

Transmission electron microscopy characterization of a fluorine-doped Si3N4 H . - J . KLEEBE

University of Bayreuth, Institute of Materials Research, Ludwig-Thoma-Str. 36B, D-95447Bayreuth, Germany G. PEZZOTTI

Toyohashi University of Technology, Department of Materials Science, Hibarigaoka, Tempaku-cho 1-1, Toyohashi 441, Japan T. NISHIDA

Department of Materials, Kyoto Institute of Technology, Matsugasaki, Kyoto 606, Japan

Densification of Si3N4-based ceramics via liquidphase assisted sintering is a widely used processing technique to overcome the low self-diffusion coefficients of Si3N4 [1]. The most common sintering additives utilized to achieve complete densification are MgO, Y203 and Y203 + A1203 [2, 3]. Numerous research studies have been focused on the investigation of the influence of type and amount of sintering aid on the densification behaviour, microstructural development and resulting mechanical characteristics [3, 4]. Since Si3N4 ceramics are of interest regarding their application as structural materials at elevated temperatures, the influence of densification aids on the high-temperature mechanical performance has been studied extensively. Based on these investigations, it is well established that (i) the high-temperature behaviour of Si3N4 depends primarily on the volume fraction, composition, distribution and crystallinity of the residual glass [5, 6] and (ii) amorphous grain-boundary phases limit t h e high-temperature properties of Si3N4-based ceramics, in particular, above the glass softening temperature of about 9001100 °C. Hence, since Si3N4 ceramics consist of highly refractory matrix grains surrounded by a less refractory vitreous phase, the modification or even elimination of such secondary phases is required, in order to overcome the aforementioned obstacles hampering the application under severe service conditions. One approach to improve the hightemperature performance of Si3N4 materials is to form a more refractory intergranular phase. For example, the substitution of lanthanide oxides for the commonly used metal oxides as sintering additives has been reported as a method of improving the refractoriness of the intergranular phase [3-7]. Moreover, post-densification heat treatment is also a widely accepted method for minimizing the residual fraction of glass by converting it into refractory crystalline phases [8-11]. However, it should be emphasized that complete crystallization of the glass residue cannot be achieved. Amorphous intergranular phases are still present along grain and phase boundaries upon heat treatment [12]. Moreover, crystallization of secondary phases surely reduces the volume fraction of residual glass, but can also 1668

simultaneously lead to an enrichment of impurities within the remains of the liquid, formed at elevated sintering temperatures, when these impurities are insoluble in the crystal lattice of the newly formed secondary phase. This leads to the opposite effect, an even more pronounced deterioration of the hightemperature properties [13]. Thus, in addition to volume fraction and refractoriness of the intergranular phase, the structure and Chemistry of the amorphous films, located at grain and phase boundaries, is important with respect to high-temperature properties [14]. It has been shown both theoretically and experimentally [ 15-17] that the film width depends on the composition of the adjacent grains as well as on the composition of the residual amorphous phase. However, the way in which the intergranular film thickness itself influences the resulting thermo-mechanical properties (in contrast to the glass chemistry/viscosity) is still an open question. Apart from crystallizing the residual glass pockets upon post-sintering heat treatment, a number of attempts have been reported to fabricate Si3N4 materials without external addition of sintering aids. It has been shown that complete densification can be achieved during hot isostatic pressing (HIPing) of pure undoped o~-Si3N4 powders at temperatures I> 1950 °C [18, 19]. However, high resolution electron microscope (HREM) studies performed on these materials unequivocally showed the presence of residual glass, observed both at triple-grain junctions and at grain boundaries [20]. Since o~-Si3N4 starting powders are in general oxidized during exposure to the atmosphere, an SiO2-rich surface layer covers the particles. At elevated HIPing temperatures (~>1730°C) this layer forms a silica rich eutectic liquid, which promotes densification and results in the observed amorphous residue upon cooling. Although these materials contained small fractions of residual glass, they revealed superior hightemperature properties, compared to other liquidphase sintered materials [21-23]. This improved material performance is related to the glass chemistry, i.e. the inherent refractoriness of the amorphous phase. However, in contrast to the improvements 0261-8028 © 1995 Chapman & Hall

reported, the materials still lack appreciable ductility, e.g. they are highly brittle [21-24]. It is worth noting that an Si3N 4 material densified without sintering aids revealed a very low intrinsic fracture energy of 10.7 _+ 0.8 Jm -2, as reported by Pezzotti et al. [24], as compared to the markedly higher values commonly measured in Si3N 4 alloys. As a consequence of the aforementioned results, further detailed information about the structural and chemical characteristics of such ceramic residual glasses, as well as their correlation with micromechanical processes, dominating the macroscopic material performance are required. In this letter we report on the microstructure of a HIPed and fluorine-doped Si3N 4 ceramic with no further sintering aid addition. Emphasis is placed on the structural and chemical characterization of the vitreous secondary phase performed by transmission and analytical electron microscopy. A high purity oz-Si3N4 starting powder (E-10, Ube Industries Ltd, Ube, Japan) was doped with fluorine by adding Teflon powder (Teflon, E.I. du Pont de Nemours, Wilmington, Delaware) via mechanical mixing. The powder blend was pre-heated in vacuum to 1200°C in order to depolymerize the Teflon structure to tetrafluorethylene C2F4, which was thought to be incorporated into t h e SiO2 glass structure at elevated temperatures (1>1000 °C) under the formation of gaseous CO. The powder compact was densified by HIPing at 1900°C for 2 h at 180 MPa Ar gas pressure. Specimens were encapsulated in an evacuated B-silicate glass tube to enable complete densification and to avoid fluorine evaporation. Full density (>99.5%) was achieved under the applied processing conditions. The grain-boundary structure and interface chemistry of the Si3N 4 material, doped with fluorine as the only sintering aid, were studied by transmission electron microscopy (TEM). TEM foils were prepared by the standard techniques of grinding, dimpling and ion beam thinning, followed by coating with a thin layer of carbon to minimize charging under the electron beam. TEM and analytical electron microscopy (AEM), were performed using a Philips CM20FEG (field emission gun) operated at 200 kV with a point resolution of 0.24 nm. This instrument was equipped with an energy-dispersive X-ray spectrometer (EDS; Tracor Voyager 2100) with an ultra-thin window Gedetector. Moreover, an electron energy loss spectrometer (EELS; Gatan 666) was attached to the microscope. To determine accurately the thickness of the intergranular films at two-grain junctions, HREM studies were carried out utilizing a Jeol 4000EX (top entry) microscope operated at 400 kV with a point resolution of 0.18 nm. The overall microstrucmre of the material investigated cohsisted of /3-Si3N4 grains, which were surrounded by a small fraction of amorphous secondary phase. No crystalline phases other than /3-Si3N4 were observed. The formation of the siliconoxinitride phase, Si2N20 , could be excluded, since its decomposition temperature of about 1830 °C [25] was exceeded during HIPing. The ~3-Si3N4

grains were about 0.2-1/xm in diameter, with only a small amount of elongated ]3-grains observed by TEM, which exhibited an apparent aspect ratio of 34. The size of the triple-grain pockets containing the amorphous residue was on the order of 10-200 nm in diameter. A close TEM inspection of the triple pockets in conjunction with specimen tilting and examining for contrast changes revealed no crystalline phases within these pockets. It was concluded that no devitrification of the secondary phase had occurred upon cooling, which was consistent with experimental results obtained by electron microdiffraction and diffuse dark field (DDF) imaging. Hence, the relatively fine grained material consisted of equiaxed /3-Si3N4 matrix grains in addition to small triple-grain junctions, as shown in Fig. 1. The glass pockets which contained most of the amorphous residue were interconnected by a threedimensional network of thin grain-boundary films. Detailed HREM studies confirmed the presence of an amorphous interlayer along all/3-Si3N4 grain boundaries investigated. In order to determine precisely the width of such intergranular films, the HREM imaging technique was utilized [26]. This technique was shown earlier to be most accurate (+0.1 nm), as compared to DDF imaging and the defocus Fresnel fringe imaging technique [27, 28], both of which can also be used for the detection and thickness evaluation of such interlayers. Fig. 2 depicts two adjacent/3-Si3N4 grains separated by an intergranular film, which emanated from a small triple-grain pocket. At the interface, the amorphous intergranular phase seemingly acquires a characteristic film thickness. This observation is consistent with previous studies indicating that, for a given interface composition, the grain-boundary film thickness is constant [29]. Hence, the major parameters that influence the intergrannlar film thickness are the chemistry of the adjacent grains and the chemical composition of the amorphous phase itself. Moreover, recent HREM observations led to the conclusion that intergranular

Figure 1 Low-magnification TEM bright-field image showing the overall microstructure of the fluorine-doped Si3N4 with a grain diameter of typically 0.2-1 /xm. Note the rather globular morphology of the matrix grains and the small extention of the triple junctions, e.g. 10-200 nm in diameter. No crystalline phases were observed in any of these glass pockets.

1669

0,~

8

E_ / Energy loss (eV)

Figure 2 HREM micrograph of the intergranular film observed along an Si3Ng-Si3N4 interface. The width o f the amorphous grainboundary interlayer was quantitatively evaluated to 1.1 nm. Note that fluorine could be detected at the interface by EELS (see also Fig. 3).

film thickness is independent of grain misorientation, with the exception of low-energy boundaries or special orientations [29, 30]. Measurements of the grain-boundary films in this HIPed and fluorinedoped material indicated an equilibrium film thickness of 1.1 nm (see also Fig. 2). The thickness of the glass interlayer was quantitatively evaluated from eight boundaries (HREM images) and three measurements at each boundary. In comparison to Si3N4 ceramics fabricated with pure S i Q as sintering aid, a small increase in grain-boundary film thickness of 0.1nm was observed. For the undoped material, a film width of 1.0 nm was reported, independent of the silica glass volume fraction [17, 20]. It should be noted that the different results on intergranular film thicknesses reported in the literature thus far are mainly based on the variation in the cation concentration, i.e. observed changes due to different sintering aids used. The grain-boundary film thickness of 1.1 nm in this F-doped material is less than that evaluated in either Y203 + A1203 or Sc203-doped Si3N4, which both revealed a film thickness of 1.5 nm [31], because the cations are present at these interfaces. However, the small increase in film thickness of 0.1 nm monitored here, when compared to the undoped material with a film width of 1.0 nm, is thought to be due to the analysed change in anion concentration along the grain boundaries, i.e. the segregation of fluorine at the interface. Electron energy-loss spectroscopy revealed the presence of fluorine both at the triple-grain junctions and the grain-boundary films. On determining the F:O ratio for the spectra obtained from triple pockets and interface regions, no major compositional changes were detected. An average F:O ratio of about 0.30 was evaluated, as shown in Fig. 3, which apparently suggests a homogeneous secondary phase composition within the material. Assuming a constant grain-boundary chemistry, a constant film thickness was to be expected. Furthermore, with respect to the incorporation of fluorine into the SiO2-glass structure, a small increase in film thickness was also expected. 1670

Figure 3 EELS measurements of (B) a glass pocket (triple junction) and ( I ) a grain-boundary film (interface). The F:O ratio of the intergranular film and the triple pocket are about identical with an average value of 0.30, hence a homogeneous composition of the vitreous phase throughout the sample is assumed.

This is due to the fact that the replacement of oxygen by fluorine, which have almost identical ionic radii with 0.132nm and 0.133nm for O e- and F-, respectively, requires two fluorine ions to allow for charge balance. Hence, two F - ions are incorporated into the glass structure replacing one 0 2 . ion, therefore leading to the observed slight increase in film thickness, when compared to the pure SiOe glass structure. It can be concluded that the HREM observations are consistent with the obtained EELS results and, in addition, that the grain-boundary structure is also influenced by the anion concentration, i.e. the fluorine content at the interface. Although a detailed characterization of the mechanical properties of such F-doped Si3N4 materials is beyond the scope of this letter, it should be mentioned that first preliminary results showed an enhanced intergranular fracture (improved toughness) as well as a lower creep resistance [22, 32]. This latter data is consistent with the lower HIPing temperature needed in the F-doped system to achieve full densification as compared to the undoped system (i.e. 1900 °C rather than 1950 °C). The variation in mechanical response of the F-doped samples compared to the undoped materials can also be attributed to changes in glass structure and chemistry, owing to the incorporation of fluorine.

Acknowledgements Professor M. Rfihle is greatly acknowledged for his support of the HREM study, which was performed at the Max-Planck Institute in Stuttgart.

References 1. K. KIJIMA and S. S H I R A S A K I , J. Chem. Phys. 65 (1976) 2668. 2. I. C. HUSEBY and G. PETZOW, Powder Metall. Int. 6 (1974) 17. 3. K . S . M A Z D I Y A S N I and C. M. COOKE, J. Amer. Ceram. Soc. 57 (1974) 536.

4. 5. 6. 7. 8. 9. 10. 11. 12.

13. 14. 15. 16. 17.

18. I9.

W. A. S A N D E R S and D. M. M I E S K O W S K I , ibid. 64 (1985) 304. F. F. L A N G E , Amer. Ceram. Soc. Bull. 62 (1983) 1369. G. Z I E G L E R , J. H E I N R I C H and G. W O T T I N G , J. Mater. Sci. 22 (1987) 3041. N. H I R O S A K I , A. O K A D A and K. M A T O B A , J. Amer. Ceram. Soc. 71 (1988) C-144. A. T S U G E , K. N I S H I D A and M. K O M A T S U , ibid. 58 (1975) 323. L . K . L . F A L K and G. D U N L O P , ,L Mater. Sci. 22 (1987) 4369. D. A. B O N N E L L , T. Y. TIEN and M. R1JHLE, J. Amer. Ceram. Soc. 70 (1987) 460. M. K. C I N I B U L K , G. T H O M A S and S. M. J O H N S O N , ibid. 75 (1992) 2037. J. S. V E T R A N O , H.-J. K L E E B E , E. H A M P P , M. S. H O F F M A N N and M. R U H L E , J. Mater. Sci. Lett. 11 (1992) 1249. H.-J. K L E E B E , J. B R U L E Y and M. R O H L E , J. Eur. Ceram. Soc. 14 (1994) 1. D. R. C L A R K E , J. Amer. Ceram. Soc. 72 (1989) 1604. Idem, ibid. 70 (1987) 15. D. R. C L A R K E , T. M. S H A W , A. P. P H I L I P S E and R. G. H O R N , ibid. 76 (1993) 1201. H.-J. K L E E B E , M. K. C I N I B U L K , I. T A N A K A , J. B R U L E Y , R. M. C A N N O N , D. R. C L A R K E and M. R O H L E , Mater. Res. Soc. Symp. Proc. 287 (1993) 65. K. H O M M A , H. O K A D A , T. F U J I K A W A and Y. T A T U N O , Yogyo Kyokaishi, 95 (1987) 229. a . P E Z Z O T T I , J. Amer. Ceram. Soc. 76 (1993) 1313.

20. 21. 22.

23. 24. 25, 26. 27. 28. 29. 30. 31. 32.

I. T A N A K A , K, I G A S H I R A , H.-J. K L E E B E and M. R U H L E ibid. 77 (1994) 275. I. T A N A K A , a . P E Z Z O T T I , T. O K A M O T O , Y. M I Y A M O T O and M. K O I Z U M I ibid. 72 (1989) 1656. G. P E Z Z O T T I , K. M A T S U S H I T A , H.-J. K L E E B E , Y. O K A M O T O and T. N I S H I D A , Acta Metall. Mater. submitted. I. T A N A K A and G. P E Z Z O T T I , a~ Amer. Ceram. Soc. 75 (1992) 772. G. P E Z Z O T T [ , I. T A N A K A and T. N I S H I D A , Phil. Mag. Lett. 67 (1993) 95. Z. K. H U A N G , P. GREIL and G. P E T Z O W , Ceram. Int. 10 (1984) 14. M. K. C I N I B U L K , H.-J. K L E E B E and M. R l J H L E , J. Amer. Ceram. Soc. 76 (1993) 426. D. R. C L A R K E , Ultramicroscopy 4 (1979) 33. J . N . NESS, W. M. S T O B B S and T. F, P A G E , Phil. Mag. A, 54 (1986) 679. H.-J. K L E E B E , M. K. C I N I B U L K and M. R U H L E , J. Amer. Ceram. Soc. 76 (1993) 1969. H. S C H M I D and M. R O H L E , J. Mater. Sci. 19 (1984) 615. H.-J. K L E E B E , M. J. H O F F M A N N and M. RfJHLE, Z Metallkd. 83 (1992) 610. G. P E Z Z O T T I , H.-J. K L E E B E , K. M A T S U S H I T A , T. N I S H I D A and M. S A K A I , Phil. Mag. A submitted.

Received 6 March and accepted 24 May 1995

1671