J. MATER. CHEM., 1994, 4( lo), 1595-1602 Role of Additives in the Sintering of Silicon Nitride: A 29Si,27AI,25Mg and *'Y MAS NMR and X-Ray Diffraction Study K. J. D. MacKenzie* and I?.H. Meinhold New Zealand Institute for Industrial Research and Development, PO. Box 37-370, Lower Hutt, New Zealand Multinuclear MAS NMR in conjunction with X-ray diffraction (XRD) has been used to study the role of A1203, Y,p03and MgO, both singly and in combination, in the sintering of silicon nitride at 1500-1 800 "C. Under the present experimental conditions, Al,03 enters the silicon nitride to form a low-z fi'-SiAION, whereas MgO reacts both with the cixidised surface SiO, layer to form forsterite (Mg,SiO,), and with the Si3N4 to form an X-ray amorphous Mg-Si-0-N phase characterised by a broad 25Mg signal at about -50 ppm and a fast ,'Si relaxation time.Y,O, forms an yttrium-rich Y-Si-0-N phase at 1500 "C which progressively becomes silicon-rich at higher temperatures. The "Y spectra of these phases are broad and could be detected only in samples containing added Yb203 to shorten the relaxation timtl?. When used in combination, the A1,03/Mg0 and Y,O,/MgO pairs behave similarly to the separate components, in terms of intergranular phase formation, but A1,03/Y,03 forms Y,oAI,Si,0,BN4, for which the ,'Si and MAS NMR spectra are reported. Silicon nitride is a versatile high-technology ceramic which has found use as refractories, cutting tools, engine wear parts and biomedical implants. In practice, it is prepared as a powder which is then formed to the required shape by pressing or casting and sintered at high temperatures, with or without the application of pressure, to a hard, dense material.Because sintering is difficult in pure silicon nitride, a variety of additives have been used, which react to produce a phase which is liquid at the sintering temperature. Three of the most common additives used to promote sintering in Si3N4 are MgO, A120, and Y203, introduced either singly or in combination. The addition of up to 5 wt.% of MgO has been used to produce fully dense Si3N4 bodies by hot pressing.' The Mg-Si-A1-0-N phase diagram2 predicts the formation of forsterite (MgSiO,) in the presence of free Si02 which is normally present as a surface impurity.The silicate is molten above 1550 "C, and promotes liquid-phase sintering. The phase diagram also predicts the further reaction of the molten silicate with Si3N, to produce Mg-Si-0-N liquid,2 which when recrystallised by re-heating at 1350"C is reported to give a mixture of enstatite (MgSiO,) and silicon oxynitride (Si2N20).3 Y203 has been used, both on its own and in combination with Al,03 to produce grain-boundary phases which are more refractory than magnesium silicates, and therefore more appropriate to higher temperature applications. The Y-Si-A1-0-N phase diagrams4 indicate the formation of several stable phases within this system, including an yttrium- nitrogen melilite ( Y2Si303N4), a nitrogen apatite Y5Si,012N (designated 'H-phase'), and a substituted yttrium aluminate Y4Si207N2 (designated 'J-pha~e').~ Other phases which com- monly occur within this system are Y2Si207 and yttrium aluminium garnet (Y3A15012).The presence of Al,O, in the system lowers the temperature of the Y,03-Si02 eutectic and decreases the viscosity of the liquid Y-Si phases. Combinations of A120, and Y203 have been reported to produce a liquid phase at temperatures as low as 1400"C;6at higher tempera- tures, some of the alumina reacts with the Si3N, to form SiAlON and increases the yttrium concentration of the liquid.6 The crystalline intergranular phases which eventually appear are reported to be similar to those formed with Y203 alone, and include N-melilite, H-pha~e,~,' J-phase7 and K-phase (YSiO,N).' The use of MgO in combination with Y203 has a similar effect to that of A1203 in lowering the viscosity of the liquid phase.The crystalline compounds reported in this system are Y2Si3O3N, (N-melilite), Y,Si20,N2 (J-phase), YSi02N (K-phase), Y2Si207 and Mg5Y,Si502,.9 One of the problems in studying the reactions of the intergranular phases formed by the sintering additives is that they are often non-crystalline; to study these phases by X-ray diffraction methods, they may be re-crystallized hy heat treatment at lower temperatures, which may also ai€ect the physical properties of the sintered body. Solid-statc NMR with magic-angle spinning (MAS NMR) is a technique which is useful for studying the atomic environments of atoms in non-crystalline and poorly crystalline phases.The 1eported MAS NMR data for a number of the phases which may be encountered in Si3N4, sintered with various additives, are summarised in Table 1. Despite the existence of this body of information about the pure phases implicated in sintering processes, very little in situ work has been published on the NMR spectra of these phases in actual sintered bodies, probably because of the small quantities of interganular material involved. To solve this sensitivity problem, C'arduner et all2 spun cylinders of Si,N4 sintered with Y203 and A1203 in their NMR rotors, and recorded their 29Si and 2741 MAS NMR spectra. The aim of the present work is to extend the studies of the in situ phases formed in Si3N4 sintered with additions of MgO, A1203 and Y203 and mixtures thereof, using ;L combi-nation of X-ray powder diffraction to monitor the ciystalline phases and solid-state MAS NMR to study both the CI ystalline and non-crystalline components. Recent improvertients in 25Mg2' and "Y MAS NMR23 have re-kindled an interest in the glassy grain boundary phases in samples containing MgO and Y20,, and these are of particular interest in the present study.Experimental The silicon nitride used in this work (H.C. Starck, grade LC 10) was blended with the additive (A1203, MgO or Y203) by ball-milling under ethyl alcohol for 4 h in polypropj lene jars using Si,N, milling media. Several experiments using longer milling times (up to 16 h) did not give significantly different results, suggesting that the additives are adequately dispersed by milling for 4 h.Initial experiments used 10 wt.% of each additive, corresponding to 12.5, 28 and 6.4 mol% of Al,03, MgO and Y203 respectively, but other experiments were also J. MATER. CHE_M.,1994, VOL. 4 Table 1 MAS NMR parameters for systems containing Si,N,, Al,O,, MgO or Y,O, phase 6 ref. 29Si NMR (wrt TMS) b-Si,N, -48.5 10-12 r-Si,N, -47.1, -49.5 10-12 Y,Si,O,,N (H-phase) -73.7 to -74.9, -67.4 to -67.5 12,13Y,Si,O,N, (J-phase) -73.7 to -74.4 12-14 YSi02N (K-phase) -64.7 to -65.3 12,13Y,Si,03N4 (N-melilite) -56.7 to -57.1 12,13 Y10A12Si3018N4 -76.1 this work Y Si,N, -42.3 to -42.8, -45.5 12,13Y,Si05 -79.8 to -80.0 12.13 y-Y2Si,0, -92.8 13 Si,N,O -63.0 13 F-SiAlON -47.6 to -48.8 15 SiAlON glass (ca.Si,4AI,0,0N) -113 (broad) 16 Mg,SiO, -62 17 MgSiO, -82 18 MgSiN, -44.4 this work MgAlSiN, -42.5 (broad) this work NMR [wrt A1(H20),3+] P-SiAlON 66 to 69.1, 106 to 112.5, -6.6 to 10 15,19 SiAlON glass (ca. Si24A160s0N) 35, 50, 0 16 Y10A12Si30 1SN4 30, 54.7, 105 this work Y,A1,09 9.4, 114, 0.8 19 Y3A15012 74, 0.8 19 YAlO, 9.4, 0.8 19 Al,,O,?N5 (approx.) 114, 65, 12 19 MgAlSiN, 101.8, 112.3, 10.1 this work "Mg NMR (wrt MgSO,) MgO 26 20,21 MgSiO, -18.0 21 MgSiN, 79, 45, -57 (broad) this work MgAlSiN, 45, -59 (broad) this work "Y NMR (wrt YCl,) y203 314, 272.5 22 Y3A15012 222 22 YAlO, 214.5 22 Y,SiO, 237, 148 22 a-Y,Si20, 114 22 j?-YzSi207 208 22 ;t-Y,Si20, 198 22 6-Y2Si,0, 122 22 Y,Si,O,N, (N-melilite) 185 (broad) 23 Y4Si20,N, (J-phase) 202 (broad) 23 carried out using different additive concentrations, e.g. up to positive pressure during the firing.The heating rate was 15 mol% Y203. Additions of 3 wt.% Yb203 (ca. 0.03 mol per selected to take 120 min to reach the sintering temperature mole of mixture) were made to the Y-containing samples to (1500-1800°C) at which the sample was held for a further decrease the 89Y relaxation time;23 89Y spectra could not be 120 min before being cooled rapidly by turning the furnace obtained for samples without this Yb203 addition. Other off. Temperature measurement was by boron graphite-graph- samples were also prepared containing 10mol% each of Y203 ite coaxial thermocouple, with periodic checks also made by and A120,, Y203 and MgO, and A1203 and MgO.After the optical pyrometry. sample had been milled, the solvent was removed in a vacuum After measurement of their radial firing shrinkage by vernier rotary evaporator and the powder brushed through a 600 pm callipers, the pellets were broken, their densities and porosities sieve before being pressed at 100 kPa into 2.0 g pellets of measured by a water penetration method, and a portion 200 mm diameter. The green densities of all the pellets, both ground to pass a 100 mesh sieve for examination by X-ray doped and undoped, were very similar (1.62-1.68 g ~m-~). powder diffraction using a Philips PW 1700 computer-Since the primary focus of this study was the chemistry and controlled goniometer with a graphite monochromator and composition of the intergranular phases, the addition of other Co-Kol radiation.The room-temperature MAS NMR compounds was avoided, to minimise possible complications. measurements were carried out at 11.7 T on a Varian Unity Thus, the low pressing pressure and consequent low green 500 spectrometer using a 5 mm Doty probe spun at typically density resulted from the decision not to use an organic 10 kHz as follows: "Si, spectrometer frequency 99.3 MHz, 90" binder. Pressureless sintering was carried out in a boron pulse of 6 ps, recycle times of up to 3000 s, referenced to nitride powder bed in an alumina pot with lid, using a graphite tetramethylsilane (TMS); 27Al, spectrometer frequency resistance furnace (Thermal Technology Inc).The furnace was 130.3MHz, 15" pulse of 1 ps, recycle time 0.1 or 1 s, referenced evacuated to 100 kPa or better for at least 0.5 h before filling to Al( H20)63+ ;25Mg, spectrometer frequency 30.6 MHz, 60" with oxygen-free nitrogen which was kept flowing at a slight (solids) pulse of 3 ps, recycle time 0.1 s, up to 500 000 transients J. MATER. CHEM., 1994. VOL. 4 acquired, referenced to saturated MgS04 solution; "Y, spec-trometer frequency 24.5 MHz, 90" pulse of 18 ps, recycle time 100 s, referenced to aqueous YC1, solution. Results and Discussion The change in radial shrinkage with temperature for the series of samples containing 10 wt.% of the single additives is shown in Fig.1. By comparison with the undoped control samples, all three additives exert a positive influence on the sintering behaviour, increasing with temperature, and in approximate proportion to the molar concentration of the additive. A similar trend is seen in the bulk density, with an inverse trend in the apparent porosities. The absolute values of density and porositj indicate that the most highly sintered of these samples has achieved ca. 80% of theoretical density, a consequence of the low unsintered densities resulting from the decision not to use a binding agent. However, the changes in density with sintering conditions are of more interest than the absolute densities which could be achieved, since the primary purpose of this work was to study the intergranular phases rather than to produce highly densified materials.Typical X-ray powder diffraction traces of the various sintered bodies are shown in Fig. 2. XRD indicates that the starting material contains predominantly a-Si3N4; with increasing temperature, in the absence of additives, the pro- portion of p-Si3N4 increases only slightly. With the addition of 10 wt.% Y,03, conversion to p-Si,N, is much enhanced at 1600-C and is complete by 1700°C. The X-ray patterns also indicate the presence of small amounts of other crystalline phases: Y,Si,012N (H-phase) at 1500"C, giving place to Y2Si3N40, (N-melilite) at 1600-1750 "C [Fig. 2(f),(g)]. Samples prepared specifically for "Y NMR (see later) contain- ing higher concentrations (15 mol%) of Y203 behave similarly to the 10wt.% samples, except that they form Y4Si20,N2 (J-phase) in preference to H-phase at 1500°C [Fig.2(e)].At 20.0 h 15.0 -(II.-10.0 0.0 I I I I1 I I 1500 1600 1700 1800 71°C Fig. 1 Radial shrinkage of Si,N, pellets sintered in nitrogen with 10wt.Yo additive at 150O-175O0C, and 120 min dwell time at sintering temperature. Additives: (M) MgO, (+) A1,0,, (A)Y,03, (0)blank. 18OOcC, the Y-containing samples were found by XRD to contain elemental Si, and a degree of melting had occurred. Similar results were found at 1800 "C for samples containing A1203 [Fig. 2(c)] and MgO. The formation of elemt~~tal Si has previously been reported by Deeley et a/.' in Si,N4 sintered in graphite at 1850cC,and by D~tta~~ in commercial pressureless sintered (Kyocera) Si3N4.The reasons for its formation in the present samples are not clear, but may be due to some peculiarity of the reducing atmosphert in the graphite furnace, acting in combination with the porous nature of the pellets, which are of lower green density than would normally be encountered in sintering experiments. The addition of 10 wt.% MgO similarly promoted the formation of P-Si3N4, this conversion being complete by 1700 "C . At the lower temperatures, there was also X-ray evidence of a small amount of poorly crystalline forsterite [Fig. 2(d)] which progressively decreased with increasing temperature. At 1800"C, the sample still contained predominantly Si.N4, but decomposition to Si had progressed appreciably. In samples containing 10 wt.% A1203, conversion to p-Si3N4 was also well advanced by 1600°C and complete by 1700°C. At 1500"C, a small amount of a-A1203 was also detected [Fig. 2(a)], but by 1600"C this decreased signifi- cantly. No other discrete Al-bearing phases were dtbtectable by X-ray diffraction, but slight shifts in the positions of the P-Si3N4 peaks suggest the incorporation of some A1 into this phase to form b'-SiAION (IFig. 2(b)].Careful measurements of the X-ray peak positions, using silicon as the angular calibrant, were used to deduce the z values of these p-SiAlONs from the relationship given by Ekstrom et These values indicate very little incorporation of A1 at 1500"C, consistent with the presence of unreacted A1203in thi:, sample.At increasing temperatures, the SiAlON z values progressively increase, up to a maximum of 0.64 at 1700T. The XRD results, which are generally consisttmt with predictions from the phase diagrams, are summarised in Table 2. 29SiNMR Fig. 3 shows a selection of typical 29Si MAS NMR spectra. The major features of all these spectra are the 4 and P-Si,N4 resonances at -47 and -49 ppm (a) and -48 ppni (p), the relative areas of which were used to provide an indication of the progress of the r to p transformation with temperature (Fig. 4). Fig. 4 indicates that all three additives facilitate the a-p transformation above 1500"C, the greatest efft:ct being obtained with Y203 and MgO.The effectiveness of Y203 as a transformation additive is in contrast to its performance as a sintering aid (Fig. l), suggesting that the mechanisms of these two processes are influenced as much by the chemistry of the system as by the formation of phases which ;ire liquid at the reaction temperature. The 29Si spectra of the samples containing A120, [Fig. 3(a),(b)]show no additional features attributable to aluminosilicate or oxynitride phases, but the spectra are not inconsistent with the presence of D' SiAION, in which the Si resonance occurs at -47.6 to -48.4 ppm,15 i.e. for practical purposes indistinguishable from P-Si,N,. The 29Si spectra of samples containing MgO [Fig. 3(c)] contain, in addition to the two polymorphs of Si,N4, an additional resonance at about -62 ppm, corresponding to that of f0r~terite.l~ The relative intensity of this I esonance decreases as the heating temperature increases, possibly due to the vapour-phase removal of MgO in the flowing nitrogen atmosphere at higher temperatures.These MAS NMR obser- vations are consistent with the XRD results. Samples containing 6.4 mol% Y203 showed no 29Si reson- ances other than those of Si,N4, despite clear XRD evidence J. MATER. CHEM., 1994, VOL. 4 IN N tf) I lB lBlB 10 20 30 40 50 60 70 80 28(Co-Ka)/degrees Fig. 2 Typical XRD traces for Si2N4 sintered for 2.0 h with various additives: (a) A1203, 1500-1600 "C; (b)A1,03, 1700-1750 T; (c) A1203, 1800°C; (d) MgO, 1600-1650°C; (e) Y203, 1500°C; (f)Y,03, 1600°C; (g)Y203, 1700-1750°C; (h)A1203+Y203, 1700'C.Additive contents: (u)-(d) 10 wt.%; (e)-(g) 15 mol%; (h) 10 mol% each component. Key to phases: A, r-Si3N4 (PDF no. 9-250), B. P-Si3N4 (PDF no. 33-1160); B', P'-SiAlON (PDF no. 25-1492); C, corundum (PDF no. 10-173); Si, silicon (PDF no. 27-1402); F, forsterite (PDF no. 34-189); J, J-phase (PDF no. 32-1450); N, N-melilite (PDF no. 28-1457); Y, Y,,A12Si3018N4 (PDF no. 32-1426). Table 2 Phase formation in Si3N4 sintered with additives, by XRD and MAS NMR additive" T/"C XRD phases NMR phases 1500 A, B, C 1600 B', AlN(tr) 1700 B', AIN 1750 B', AIN 1800 B', AlN, Si 1500 A, B, F, Mg-Si-0-N 1600 B, A, F, Mg-Si-0-N 1700 B, F(tr), Mg-Si-0-N 1750 B, Mg-Si-0-N 1800 B, Si, Mg-SI-0-N 1500 A, B, J 1600 B, J, N 1700 B, N 1750 B, N 1800 B, Si 1700 B', Mg-Si-(1-N, MgO 1700 B, N, Mg-Si-0-N, MgO 1700 B, y Key: A =r-Si3N4, B =P-Si3N4, B' =P'-SiAlON, C =corundum, (r-A103), F =forsterite (Mg2Si04), J =J-phase (Y4Si,0,N2), N =N-melilite (Y,Si303N4),Y =Y,,,A12Si30,,N4, A1N =aluminium nitride, Si =silicon, Mg-Si-0-N =glassy phase containing these elements.Tr =trace. a All additive concentrations 10 wt.%, except Y203 (15 mol%) and binary combinations (10 molyo each component). J. MATER. CHEM., 1994, VOL. 4 -48.9-4i 4 -4 -47 A (f) I, I1 I ,0#l.l.ltl~ -40 -60 -80 ' -20 -60 -100 -140 29~i6 wrt TMS Fig.3 Typical room-temperature 29Si MAS NMR spectra of Si,N, sintered with various additives at various temperatures: (a) A120,, 1500°C; (b) Al,O,, 1750°C; (c) MgO, 1500°C; (d) Y203, 1500°C; (e) Y,03, 1700 "C, (f)MgO +Al,O,, 1700"C.Additive content: (a)-@) 10 wt.%; (d)-(e) 15 mol%; (f) 10 mol% each additive. Recycle time for spectra (u)-(e),3000 s, recycle time for spectrum (f) 10 s. 100 80 h. $! 2 60 m.-? 4c 2c I I I I I IC 1 30 1600 1700 1800 TI% Fig. 4 Transformation of r-to P-Si,N, estimated by ,'Si MAS NMR, in the presence of 10wt.% of various additives, as a function of temperature, and 120 min dwell time at sintering temperature. Additives: (M)MgO, (+) Al,03, (A)Y203,(@) blank. of H-phase at 1500 "C and N-melilite at > 1600 "C By con- trast, the spectrum of the sample containing higher concen- trations of Y203 (15mol%) heated at 1500°C show, in addition to the double peaks of a-Si3N4, a weak resonance at -74.4 ppm [Fig.3(d)], corresponding to the reported position for J-pha~e.'~,'~ When the sample is heated at higher tempera- tures, this resonance is replaced by a broad feature at -57 ppm [Fig. 3(e)], corresponding to the reported position of N-melilite.'23'3 Thus, the 29Si NMR spectra of samples containing appreciable Y203 are consistent with 1 he XRD results, but at lower Y concentrations, the low sensitivity and broadness of these resonances makes MAS NMR 1t:ss useful than XRD as a diagnostic technique for the intergranular yttrium silicate phases. 27A1NMR The 27Al NMR spectra of A120, samples sintered for 2 h at various temperatures are shown in Fig.5. At 1500"C, the A1 is predominantly six-coordinate ( 14 ppm), corresponding to unreacted corundum, a-A1203 [Fig. 5(u)], but the smidl, broad resonance at 60 ppm indicative of tetrahedral A1 indicates that some degree of reaction has taken place. By 1600°C,this reaction is well advanced [Fig. 5(b)],the broad tetrahedral and octahedral resonances at 65 and 7 ppm, respectively, corresponding well with those of P'-SiAlON (66 to 69.1 and -6.6 to 10 ppm, re~pectively'~,~~). This spectrum also contains a hint of residual corundum at 12 ppm and a shoulder which may mark the beginning of the P'-SiAlON peak at 104 ppm; this resonance, which corresponds to A1-N bonding is much better developed at 1700°C [Fig.5(c)]. By 1750°C [Fig. 5(d)] the degree of A1-N bonding has become more significant, as reflected in the more highly positive shifts in all the resonances, and the dominance of the resonance at 113 ppm sug,gests the appearance of A1N as a discrete phase.lg By 18OO"C, A1N has become the only A1 phase detectable by NMR [Fig. 5(e)], the silica having been removed by the vapour-phase formation of SiO and the progressive reduction to elemental Si, confirmed by XRD and 29Si NMR. None of these samples >,how the 102.5 I I I I 1 r A" l l I I 1 I 1 L- 200 0 -200 200 0 -200 27AI 6 wrt AI(H20)6& Fig. 5 Typical room-temperature 27Al MAS NMR spectra of Si,N, sintered with 10 wt.% A1,0, for 120 min at (a) 1500'C, (0)1600"C, (c) 1700"C, (d) 1750"C, (e)1800"C.(f) Sample sintered with 10 mol% each of A1,0, and Y203for 120 min at 1700"C, (Y,oA12Si.,018N,). spectrum reported for a highly siliceous SiAlON glass,16 in which a resonance ascribed to five-coordinate A1 was found at 35 ppm, in addition to four- and six-coordinate resonances at 50 and 0 ppm, respectively. Since the principal reaction product found here (p-SiAlON) is itself refractory and difficult to sinter, the addition of A1203would not by itself be expected to assist sintering by the formation of any phase which is liquid at the sintering temperature. 2sMgNMR A selection of 25Mg MAS NMR spectra of samples containing MgO and sintered at various temperatures is shown in Fig.6. The low concentrations of Mg-bearing phases in these samples and the low sensitivity of "Mg make these natural-abundance spectra noisy, but at all temperatures, a sharp resonance is seen at ca. 33-36ppm, most probably resulting from the presence of unreacted Mg0.20,21 The other major feature of these spectra is the broad resonance at ca. -69 to -85 ppm, the origin of which was the subject of several further experi- ments. The XRD and 29Si NMR results suggest that the crystalline Mg-bearing phase in these samples is forsterite, Mg,SiO,. A similar broad feature at about this position has been observed in chrysolite heated to 850 "Cwhich contained a significant amount of crystalline forsterite;21 it is unlikely, however, that the 25Mg resonance in that sample was due to forsterite, and probably arises from some Mg-bearing impurity. To examine this point further, mixtures of MgO and Si02 in molar proportions corresponding to both forster- ite and enstatite (MgSiO,) were melted in air at 1600 "C and quenched in water.XRD and 29Si NMR of these materials showed them to be very crystalline forsterite, with unreacted MgO also present, but the 25Mg spectra showed only a single sharp resonance due to MgO. A similar 25Mg result was found for a sample in which part of the SiO, was replaced by Si,N4, which also contained principally forsterite, according to XRD. These results suggest that forsterite is not responsible for the "Mg spectra of these samples.A similar broad feature at ca. -50 ppm was found in equimolar mixtures of Si3N4 I I I , I , I , 1-A I , I , I , I , I 400 0 -400 '400 0 -400 25Mg6 wrt MgS04 solution Fig. 6 Typical room-temperature 25MgMAS NMR spectra of Si,N, sintered with 10wt.% MgO for 120 min at (a) 1500"C, (b) 1600"C, (c) 1700'C. (d) Equimolar mixture of Si,N, and MgO reacted in air for 120 min at 1650°C. (e) Synthetic crystalline MgSiN, and ( f)MgAISiN,. J. MATER. CHEM., 1994, VOL. 4 and MgO fired at 1650 for 2 h both in nitrogen and air [Fig. 6(d)] which according to XRD contained only Si3N, and a small trace of forsterite. The broad ,'Mg spectrum therefore appears to be associated with the X-ray amorphous or glassy phase predicted from the phase diagra~i~.~ This phase may contain four-coordinate Mg; the broad resonance at 62 ppm [Fig.6(d)] is in the region reported for tetrahedral Mg in spinel, MgA12022 and in akermanite, C'a,MgSi,O, .27 The "Mg resonance arising from six-coordinate Mg in the sintered silicon nitride samples falls in a similar range as that of MgSiN, [Fig. 6(e)] and MgA1SiN3 [Fig. 6(j')], suggesting that the Mg-containing glassy phase is stabilised by nitrogen. The results of heating equimolar Si,N,-MgO mixtures suggest that the amount of glassy phase formed is not restricted by the amount of Si available, as would be expected if the sole Si source was the oxidised surface layer on the nitride particles, and that further reaction with Si,N4 was viir the molten magnesium silicates thus formed.2' Rather, the glassy phase appears to be the product of a direct reaction of MgO with the Si3N4 itself, with the small amount of forsterite also produced resulting from a different reaction with the surface SO2, since, by contrast with the glassy phase, its concentration is essentially independent of the concentration of available MgO.Thus, to summarise, the mechanism by which MgO assists sintering appears from 29Si NMR and XRD (but not "Mg NMR) to involve the formation of some forsterite, probably by reaction between the MgO and the oxidised surface layer of the nitride. However, the 25Mg NMR spectrum provides evidence of further reaction between the MgO and Si,N, to form a glassy Mg-Si-0-N phase which is not readily recrys- tallised, and at higher MgO concentrations contains appreci- able proportions of four-coordinate Mg."Y NMR Fig. 7 shows typical *'Y MAS NMR spectra of silicon nitride samples containing 15 mol% Y203, with 3 wt.O/,l Yb,O, added before firing to facilitate the acquisition of the spectra by A , I I I I I L 800 400 0 -400 89~ solution6 wrt YCI~ Fig. 7 Typical room-temperature *'Y MAS NMR spectra of Si,N, sintered with 15 mol% Y,O, and 3 wt.% Yb,O, for 120 min at (u) 1500"C, (b)1600"C, (c) 1750°C. J. MATER. CHEM., 1994, VOL. 4 shortening the "Y relaxation time. The spectra are all very broad. and similar to those observed in equimolar mixtures of Y203 and Si3N4 heated together.23 The position of the 89Y resonance in the sample heated at 1500 "C [Fig.7(a)] is close to that reported for the J-phase (202 ~pm~~), consistent with the 29Si NMR and XRD results (Table 2). As the heating progresses to higher temperatures, the position of the centre- of-gravity of the broad "Y resonance tends downwards, able. This is confirmed by the similarity of the 29Si spectrum to that of the Y203-containing sample [Fig. 3(e)],although if forsterite is present its resonance at -62 ppm could be obscured by the broad N-melilite feature centred at -57 ppm. The 25Mg spectrum is similar to Fig. 6(c), showing the reson- ances of unreacted MgO and the broad resonance centred at ca. -70 ppm; possible confirmation that this represents an N-stabilised Mg glass was found in the 29Si spectrum acquired towards the position reported for pure N-melilite (185 ~pm~~).with a delay of 10 s [similar to Fig. 2(f)], which shows a Concomitantly, minor spectral features such as that at 526ppm and a possible shoulder upfield of the major peak become more easily distinguishable [Fig. 7(b),(c)]; these fea- tures are associated with the spectrum of N-melilite.23 Thus, although the usefulness of the 89Y MAS NMR spectra is limited by their broadness, their general features support the conclusions of 29Si NMR and XRD, that the intergranular phases formed during sintering with Y203 at lower tempera- tures are Y-rich, but become progressively Si-rich as the temperature is raised. This sequence is at variance with that proposed by Hirosaki et ~l.,~who reasoned that if the reaction is between the surface Si02 layer and the Y,03, the liquid initially formed should be Si-rich, and become progressively Y-rich with increasing temperature.It is, however, clear the sole source of Si is not the surface oxide, and significant reactions occur between Y203 and Si3N4, especially during the comparatively long soaking times of the present experi- ments. The reaction products also appear to be dependent on the reaction atmosphere, being different and less reproducible when the heatings are carried out in a carbon pot. The phases detected by MAS NMR of the various nuclei are summarised in Table 2. Effect of Additives in Combination A brief survey was made in which the techniques developed above were applied to more complex Si3N4 systems containing 10 mol% each of the additive pairs A1203/Y203, MgO/Y203 and A1203/Mg0, heated at 1700°C for 2 h.The samples containing Y203were prepared both with and without 3 wt.% added Yb203, the former samples being used for ''Y MAS NMR. The sample containing A1203/Mg0 shrunk radially by 19.5% on firing, corresponding to a final bulk density of 2.70. The only crystalline phase detectable by XRD was P-Si3N4, with no crystalline forsterite, spinel, A1203 or MgO found. The 29Si NMR spectrum acquired using a long delay (3000 s) shows only the expected silicon nitride spectrum, but using a shorter delay (10 s) and acquiring more transients, an additional broad component appears, centred at ca.-100 ppm [Fig. 3(f)]. The 27Al NMR spectrum is very similar to that of a corresponding sample containing A1203 alone [Fig. 5(c)], except for an upfield displacement of all the resonances by about 10 ppm. There was no indication of the reported resonance of Mg5A1Si30,,N at 67 pprn." The 25Mg NMR spectrum is similar to those of samples containing MgO alone [Fig. 6(c)], containing a sharp resonance at ca. 30 ppm corresponding to unreacted MgO, and a broad feature centred at ca. -50 tentatively ascribed to an N-containing Mg-Si glass. Thus, the sintering action of this pair of additives appears to consist of a combination of their individual reac- tions (i.e. SiAlON formation by the A1203 and formation of an N-stabilised Mg glass by the MgO), with no evidence of the interaction of the additives to form new phases.The combination of MgO with Y203 resulted in a radial shrinkage and final bulk density of 15.8% and 2.5 gcmV3, respectively. XRD detected only the crystalline phases present in samples containing Y203alone (P-Si3N4 and N-melilite), with no forsterite or any other Mg-containing phase detect- broad feature centred at ca. -100 ppm and possibly associated with this glassy phase. If, as has been suggested," the composi- tion of the amorphous phase formed in this system includes Y, in addition to Mg, Si, 0 and N, its presence does not significantly change the 25Mg and 29Si spectra. Thus. as is the case with MgO/A1203, the additives appear to be acting independently of each other, at least with respect to the formation of the crystalline intergranular phases.By contrast, the addition of A1203 and Y203, wliich gives a radial shrinkage and bulk density of 20.3% and 2.74 g cmP3, respectively, results in the formation of the phase Y,OA12Si30,8N4 (PDF no. 32-1426). The 29Si spectrum shows, in addition to an intense P-Si3N4 resonance at -48.8 ppm, another peak at -76.1 ppm, which apparently coi responds to the previously unreported spectrum of Y 10A12Si3018N4. The 27Al spectrum [Fig. 5(f)] is significantly different from that of the corresponding sample containing A1203 alone [Fig. 5(c)]; both spectra contain a broad resonance at ca. 60 ppm and a sharp peak at ca. 104 ppm, but the A1203/Y203 spectrum also contains a significant resonance at .?0.7 ppm, and no resolved octahedral resonance at ca.15 ppm. Resonances at 20-30 ppm are often ascribed to five-cc lordinate Al, but the only previous report of a resonance in this position in N-containing systems is in a highly siliceous SiAlON glass of approximate composition Si24A16050N.'6 Since the only crystalline phase other than P-Si3N4 identified in this sample is Y,oA12Si3018N4, Fig. 5(f) probably represents the pre- viously unreported 27Al spectrum of this compound. The combined evidence suggests that this additive pair reacts to form a new phase, which may behave similarly to the related yttrium phases formed with Y,03 alone, while possessing some of the structural features of the SiAlON glass in which the similar A1 resonance has been reported.This result is different from that reported by Carduner et who detected Y,Si3OI2N (H-phase) and YSi02N (K-phase) by hoth 29Si NMR and XRD in solid Si3N4 samples sintered with A1203 and Y203. In other experiments with A1203/Y203 ;idditions sintered in a carbon pot, we detected the formation of Y2Si303N4 (but not H-phase or K-phase); the expt.rimenta1 sintering conditions thus clearly play a significanI role in determining the nature of the intergranular phases fix-med. Conclusions 1. Multinuclear MAS NMR in conjunction with XRD is capable of providing useful information about both the degree of a-P transformation and the nature of the intergranular phases formed when Si3N4 is sintered in the presence of additives.At ca. 1800 "C the system becomes unstable, decom- posing to elemental Si, especially when the reaction 1s carried out in a carbon pot. 2. When used alone, A120, enters the Si,N, structure at >1600"C, forming a low-z P'-SiAION which decomposes to A1N with the loss of SiO at ca. 1800°C. 3. Below 1750"C, MgO reacts with the residual SO2 in the system to form forsterite which is detectable by XRD and 29Si NMR but not by "Mg NMR; the latter suggests, however, that MgO also reacts with the Si3N4 to form m X-ray amorphous phase, suggested to be an Mg-Si-0-N gldss, from 1602 J. MATER. CHEM., 1994, VOL. 4 the similarity of its ,'Mg spectrum with that of MgSiN,.The 29SiNMR spectra suggest that this glassy phase has a short 29Si relaxation time. 4. The use of Y,O, alone produces intergranular Y-Si-O-N phases which change in composition with tempera- 5 6 7 Ceramics, ed. R. E. Tressler, G. L. Messing, C. G. Pantano and R. E. Newnham, Plenum Press, New York, 1986, p. 79. K. H. Jack, J. Muter. Sci., 1976, 11, 1135. N. Hirosaki, A. Okada and M. Mitomo, J. Muter. Sci., 1990, 25, 1872. 0.Abe, J. Muter. Sci., 1990,25,4018. ture. At 1500"C the yttrium-rich J-phase was found in samples containing 15mol% Y203, progressively converting to the more siliceous N-melilite at 1600"C. The "Y spectra, which could be detected only in samples containing added Yb203 to decrease the relaxation time, are all very broad, but in 8 9 10 G.M. Crosbie, J. M. Nicholson and E. D. Stiles, Bull. Am. Ceram. SOC.,1989,68, 1202. A. Gianchello, P. C. Martinengo, G. Tommasini and P. Popper, Bull. Am. Ceram. Soc., 1980,59, 1212. K. R. Carduner, R. 0. Carter 111, M. E. Milberg and G. M. Crosbie, Anal. Chem., 1987,59,2794. general confirm the XRD and ,'Si NMR results. 5: In terms of intergranular phase formation, the additive pair Al,O,/MgO behaves as separate entities, the A1,0, forming /j'-SiAlON and the MgO forming a glassy phase, with no evidence of spinel or other A1-Mg phase formation. The same is true of MgO/Y,03, in which the same Y-Si-O-N 11 12 13 14 G. H. Hatfield and K. R. Carduner, J. Muter. Sci.. 1989,24,4209. K. R. Carduner, R. 0. Carter 111, M. J. Rokosz, C. Peters, G. M. Crosbie and E. D. Stiles, Chem.Muter., 1989, 1, 302. R. Dupree, M. H. Lewis and M. E. Smith, J. Am. Chem. Soc., 1988, 110,1083. D. S. B. Hauck, R. K. Harris, D. C. Apperley and D. P. Thompson, J. Muter. Chem., 1993,3, 1005. phases are formed as with Y203 alone, together with the Mg-containing glass. By contrast, in samples containing Al,03/Y,03, an Al-Y-Si-O-N phase is found, with a 27Al NMR spectrum similar to that of a SiAlON glass, containing a resonance ascribed to five-coordinate Al. 15 16 17 18 R. Dupree, M. H. Harris, G. Leng-Ward and D. S. Williams, J. Muter. Sci. Lett., 1985,4, 393. R. K. Sato, J. Bolvin and P. F. McMillan, J. Am. Cerum. Soc., 1990,73,2494. J. S. Hartman and R. L. Millard, Phys. Chem. Mineral., 1990, 17, 1. M. Magi, E. Lippmaa, A. Samosan, G. Engelhardt and A-R. 6. The present NMR and XRD results are consistent with expectations from previously published phase diagrams for these systems. 19 20 Grimmer, J. Phys. Chem., 1984,88,1518. R. Dupree, M. H. Lewis and M. E. Smith, J. Appl. Crystallogr., 1988,21, 109. R. Dupree and M. E. Smith, J. Chem. Soc., Chem. Commun., 1988, 1483. We are indebted to Dr. W. A. Groen, Philips Research Laboratories, Eindhoven, for the samples of MgSiN, and MgAlSiN,. 21 22 23 K. J. D. MacKenzie and R. H. Meinhold, Am. Mineral., 1994, 79, 250. R. Dupree and M. E. Smith, Chem. Phys. Lett., 1988,148.41. R. H. Meinhold and K. J. D. MacKenzie, to be published. 24 S. Dutta, J. Am. Ceram. SOC.,1982,65, C2. References 25 T. Ekstrom, P. 0.Ka11, M. Nygren and P. 0. Olsson, J. Muter. Sci., 1989,24, 1853. 1 G. G. Deeley, J. M. Herbert and N. C. Moore, Powder Met., 1961, 26 J. Sjoberg, R. K. Harris and D. C. Apperley, J. Muter. Chem., 1992,2,433. 2 8, 145. K. H. Jack, in Nitrogen Ceramics, ed. F. L. Riley, Martynes Noordhoff, Leyden, 1977,p. 109. 27 28 P. S. Fiske and J. F. Stebbins, Am. Mineral., to be published. D. R. Messier, F. L. Riley and R. J. Brook, J. Muter. Sci., 1978, 13, 1199. 3 4 D. P. Thompson, in Tailoring Multiphase and Composite F. L. Harding and R. J. Ryder, Glass Tech., 1970,11,54. Paper 4/02639J; Received 4th May, 1994