"0NMR investigation of hafnia and ternary hafnium oxides Timothy J. Bastow,"Mark E. Smith*".band Harold J. Whitfield' "CSIRODivision of Materials Science and Technology, Private Bag 33, Rosebank MDC Clayton, Victoria 31 69, Australia bDepartment of Physics, University of Kent, Canterbury, Kent, UK CT2 7NR 'Department of Applied Physics, Royal Melbourne Institute of Technology, Box 2476W Melbourne, Victoria 3001, Australia Structural evolution in alkoxide-derived amorphous HfO, gels is followed by I7Omagic angle spinning (MAS) NMR in I7O-enriched samples, with crystallisation being identified readily at 500 "C. A series of I70-enriched alkali-metal and alkaline-earth- metal hafnates with sodium chloride- and perovskite-related structures are also examined by I7OMAS NMR spectroscopy.The number of oxygen resonances is related to the crystal structure. Comparison of I7ONMR data is made to isostructural zirconium- containing compounds with the isotropic chemical shifts (Hf/Zr) in the ratio 0.86 k0.02. I7ONMR data are also presented from Hf0,-SiO, and Hf0,-GeO, gel mixtures. Oxygen-17 is a very significant probe nucleus for solid-state NMR of oxides since it exhibits a large isotropic chemical shift range and a relatively small nuclear electric quadrupole moment. Initial studies demonstrated its large chemical shift range and showed that in ionic materials the static linewidths were narrow enough for magic angle spinning (MAS) of 3 kHz to be effective.'q2 The major drawback to the widespread use of 170NMR is its low natural abundance (0.037%) which usually necessitates isotopic enrichment for ready observation.I7O has been used mostly in specific applications on enriched samples such as examination of ceramic materials3 and the different layers in high-temperature superc~nductors.~~~ One of the most important attributes of NMR as a structural probe is its short-range nature, which means that investigation of amorphous materials is possible where conventional charac- terisation techniques are often limited in the information they can provide. Sol-gel produced materials are of great tech- nological interest, allowing the production of materials at relatively low temperatures (that would otherwise require application of very high temperatures, e.g.glassy Zr02-Si02 mixtures) and atomically mixed compounds. Structural evol- ution in a gel occurs uia a series of distinct amorphous states that differ in their atomic arrangement and it is these states that NMR can probe effectively. The properties of materials can differ widely depending on the local bonding arrangement that occurs and the behaviour can be determined at an early stage of their formation. Recently a series of sol-gel produced simple oxides including ZTO~,~Ti026 and La203,7 have been studied by I7O NMR spectroscopy. Hydrolysis of alkoxide precursors with I7O-enriched water is an extremely effective route for incorporation of the 170label into oxide gels. The detailed route to oxide formation varies widely with amorphous mixtures of tetragonal and monoclinic precursor oxides forming in ZrO,, mixtures of OTi3 and OTi, coordinations in TiO, gels and La203 forming from LaO(0H).The structural chemistry of each oxide is thus very different and needs to be studied separately in detail. In binary mixtures the variety of behaviour increases. I7O effec-tively monitors both metals since it is bonded to all such nuclei. In metal silicate gels, I7O NMR spectroscopy has been shown to be a much less ambiguous structural probe than 29Si NMR, with phase separation and atomic-scale mixing readily detectable in Ti02-SiO, Direct crystallisation of TiO, gels into rutile is made possible by the incorporation of small amounts of tin in the original gel and ''0NMR showed that the room-temperature gel had a different structure to pure Ti02 gels as the oxygen was already rutile-like in the tin- doped case, as indicated by the I7O chemical shift." Similarly the behaviour of ZrO, dissolved in an SiO, matrix is very different to TiO, with NMR showing that almost complete solid solution of zirconium in SiO, can occur." Hafnium is a very similar element in its chemical and physical properties to zirconium, with hafnia (HfO,) and zirconia (ZrO,) showing complete solid solution, consistent with their similar atomic radii.Both oxides are stable in the monoclinic form at room temperature, but transform on heating successively to tetragonal and cubic forms. In the present work 170-enriched preparations of HfOz gels, ternary hafnates and Hf02-SiO, are examined by I7O MAS NMR and a comparison is made to the previous work on zirconium analogues.The work is extended to Hf0,-GeO, gels. The sensitivity of I7O NMR as an atomic scale probe in such materials is examined. In crystalline structures of ionic com- pounds, such as perovskite- and NaC1-structured hafnates, I7O MAS NMR spectra can show high resolution. This observation is a consequence of the small quadrupole interaction (C,= e2qQ/h, where eQ is the nuclear electric quadrupole moment and eq is the maximum component of the electric field gradient tensor) in these ionic materials. The resonances become broader in the less ionic silicon- and germanium-containing compounds. Experimental Hafnia gel was prepared by dissolving hafnium isopropoxide, 2 mmol Hf(OPr'), *PrOH, in a mixed solvent of 20 cm3 toluene and 8 cm3 isopropyl alcohol, and then adding dropwise 0.5 ml of 10 atom% 170-enriched water while stirring continuously.This produced a sol which gelled rapidly and the gel was aged at room temperature for 24 h. The excess solvent was slowly evaporated at room temperature under a flow of dry nitrogen. The dry gel was then heated successively at 100, 300, 500 and 800°C for 2 h periods. The hafnia gel heated in nitrogen at 500°C and then at 800°C gave an X-ray powder diffraction (XRPD) pattern in agreement with monoclinic baddelyite-type HfO, (ASTM-JCPDS no. 34-104). '70-enriched Li,HfO, and Na2Hf03 were prepared by solid- state reaction of lithium and sodium carbonate, respectively, with I70-enriched hafnia gel by heating at 800 and 950°C J.Muter. Chem., 1996, 6(12), 1951-1955 1951 respectively for 6 h under a dry nitrogen atmosphere The correct stoichiometry of the final product was ensured by using amounts of Li2C03 or Na,CO, that correspond to the amount of hafnium initially added as Hf(OPr'),.PrOH The XRPD pattern of the Li,HfO, product was in excellent agreement with it being the correct single-phase product (ASTM-JCPDS no 23-1183) The XRPD peaks of Na,HfO, had a similar intensity distribution to those of Na,ZrO, (ASTM-JCPDS no 25-770) but with slightly shifted positions in accord with a slightly larger unit cell for the hafnate The alkaline-earth-metal hafnates were prepared by the metathetical reaction of Li,HfO, with an excess of a melt of a eutectic mixture of NaCl and the appropriate Group 2 chloride at 800 "C for 4 h in a nitrogen atmosphere The eutectic mixture was used because the minimum temperature that causes reac- tion is desirable as it minimises exchange of 170with H,O, COz or 0, even in a nominally inert atmosphere, and with the reaction vessel Thus 2 mmol dm-3 Li,HfO, was heated with a melt of 9 2 mmol dmP3 SrCl, and 9 4 mmol dm-, NaCl The melt was cooled and the unreacted halide extracted by dissolution in water The dried product (94% of theoretical yield) gave an XRPD pattern with very sharp peaks in accord with crystalline SrHfO, (ASTM-JCPDS no 44-991) CaHfO, and BaHfO, prepared similarly gave powder patterns in accord with ASTM-JCPDS no 36-1473 and 24-102 respectively A gel with stoichiometry HfSiO, was prepared by dissolving 2 29 mmol dm-, Hf(OPr'), *PrOH in 20 ml toluene and then adding an equimolar quantity of Si(OEt), dissolved in 8 ml isopropyl alcohol 170-enriched water (1cm3) was then added with continuous stirring A gel formed which was allowed to dry at room temperature under a stream of nitrogen and then heated for successive 2 h periods at 500 and 850°C The mass of the final product was 0 551 g, which is 93% of the theoretical yield of HfS104 XRPD studies of the product showed no evidence of crystallisation of the product Selected area electron diffraction (SAED) gave a nng pattern consistent with short- range order but no crystallisation To prepare hafnium germanate, HfGeO,, 8 mmol dmP3 Hf (OPr'), -PrOH was dissolved in toluene and an equimolar amount of Ge(OEt), in isopropyl alcohol was added 170-enriched H20 (1 cm3) was then added stirring continuously which produced a gel which was allowed to dry at room temperature under a stream of dry nitrogen The dried gel was then heated stepwise to 100, 300, 500 and 950°C XRPD studies of the product formed by heating at 950°C were in excellent agreement with crystalline HfGeO, (ASTM-JCPDS no 34-408) The X-ray peaks were not as sharp as those from the alkaline-earth-metal hafnates The XRPD of the product formed at 500 "C showed no evidence of long-range ordering X-Ray powder diffraction of these samples was performed on a Siemens Type F diffractometer using Ni-filtered Cu radiation Selected area electron diffraction patterns were obtained in a JEOL 2010 electron microscope 170MAS NMR spectroscopy was performed at 54 24 MHz on a Bruker MSL 400 spectrometer equipped with a 94T magnet The spectra from HfO, gel and binary crystalline hafnates were accumulated using a 7 mm double-bearing (DB) MAS NMR probe spinning at 4-5 kHz using 15 ps (ca 30" tip angle) pulses and a 5-10 s recycle delay For HfO, mixed with SiO, and GeO, a 4mm DB-MAS NMR probe spinning at >8 kHz was employed with the other acquisition parameters the same as for the 7 mm probe For the 7 mm probe a Si3N4 rotor was used but for the 4 rnm probe only ZrO, rotors were available and these give a 170background signal at 6 ca 385 which can be seen in Fig 3(b) and 4 (later) where the 4mm probe was used The rotor peak can be identified readily and does not interfere with the spectrum so that spectral subtraction was unnecessary There is little known about 170 relaxation times in such solids but conditions are chosen that practically allow an adequate signal-to-noise to be collected in reasonable time, which is often a trade-off between allowing complete relaxation and adding sufficient scans For BaHfOJ the I7O was determined as 32s which means some spectra observed here are partially saturated 29S1 MAS NMR spectra were accumulated by spinning at 4-5 kHz in a 7 mm DB-MAS probe using a 1 5 ps (ca 30") pulse and a 15 s recycle time at an operating frequency of 79 4 MHz Spectra were referenced externally to H20 (6 0, 170)and SiMe, (6 0, 29S1) Results and Discussion Hafnia The 170MAS NMR results from the enriched HfO, gels are shown in Fig 1 and summarised in Table 1 The gel at room temperature is characterised by two peaks at 6 245 and 336 with the latter resonance the stronger The spectrum remains unchanged after 2 h at lOO"C, and although it superficially appears similar at 300°C the peaks have moved to more positive shifts by ca 10 ppm (Table 1) After heating at 500 "C some narrow peaks appeared, indicating crystallisation, shifted to 6 267 2 and 336 8 and of changed relative intensity These narrow peaks sit on top of the residual peaks of the former broader resonances On further heating to 800°C the broad components have been completely removed and the two sharp Fig.1 170MAS NMR spectra of hafnia gel (a) dned at 25 "C and then heated to (b) 100"C, (c) 300 "C, (d) 500 "C, (e) 800 "C and (f)of unenriched bulk hafnia Table 1 Heat treatments and peak positions of HfOz gels sample heat treatment ii(170) H1 gel dried at 25 "C 245, 336f2 5 H2 H1+2h, 100°C 245, 335 k2 5 H3 H2+2 h, 300°C 258, 346 f2 5 H4 H3 +2 h, 500 "C 2672, 3368k04 H5 H4+2 h, 800 "C 267 2, 336 Ok 0 4 natural abundance HfO, 2669, 3362f04 1952 J Muter Chem, 1996, 6(12), 1951-1955 resonances and some spinning sidebands remain. A comparison with calcined monoclinic Hf0, containing I7O at natural abundance is given in Fig.l(f) with peaks at 6 266.9 and 336.2. The peaks formed when the gel crystallised and for normal bulk HfO, are consistent and agree with the previous natural- abundance study.' Of the peaks in the room-temperature gel at 6 245 and 336, the latter is very close to that from the final crystalline product, although the lines are much broader. In previous studies of the field dependence of the linewidth in similar ionic gels6v7 it has been demonstrated that the line broadening is caused by chemical shift dispersion in such ionic oxides, as a result of the disorder present in the samples in the gel state. However, it would be misleading to regard the gel spectrum as simply a broadened version of that from the crystal since the intensity ratio (ca.1 :2) is not the same in the crystalline material (1: 1). Heating to 100"C leaves the spec- trum and hence the structure unchanged from the room-temperature gel. However, at 300 "C the peaks have shifted significantly, which must be an indication that structural changes are taking place as a precursor to crystallisation. At 500 "C the much narrower resonances indicate that crystallis- ation has largely taken place, although some significant amorphous domains remain as broad underlying resonances are observable. The shifts indicate structural change at 300 "C, showing opposite trends for the two peaks. The lower shift peak has continued to become more positive whilst that at 6 ca.346 has shifted back to 6 337 at 500°C (Table 1). This observation indicates that 170NMR is a subtle probe of structural change in amorphous oxides and crystallisation. At 800°C all of the amorphous component has been removed, leaving the two resonances of crystalline monoclinic HfO,. X-Ray powder diffraction (XRD) shows no evidence of crystallisation at 500 "C, which is a consequence of the typically nanocrystalline sample dimension of such gel-produced oxides that result in particle-size broadening of the XRD reflections. NMR, being short-range in nature, shows crystallisation read- ily as all the local environments become the same causing the significant line narrowing. Assignment of the two resonances as 6 267 from OHf, and 6 336 from OHf, is in accordance with decreasing chemical shift corresponding to increasing coordination number.It is also interesting to compare the structural behaviour observed here with that of other pure oxides. For TiO, the initial gel was a mixture of OTi, and OTi, and increasing heat treatment gradually eliminated the OTi, environment, with the shift of the OTi, site becoming more anatase-like.6 The chemically very similar ZrO, again showed peaks in the gel that largely line up with the monoclinic lines. However, throughout a third peak from the tetragonal form could be observed for ZrO,. On crystallisation there were clearly tetragonal crystallites present. This is not the case for HfO,, despite a tetragonal form existing at higher tempera- tures.The difference in the structure of the gels between ZrO, and HfO, must be a reflection of the relative stability of the monoclinic and tetragonal forms in the two systems and the influence of surface stabilisation. These subtle differences could have a profound influence on the sintering behaviour of these oxides. Alkali-metal and alkaline-earth-metal hafnates The 170MAS NMR spectra from some alkali-metal and alkaline-earth-metal hafnates are shown in Fig. 2 and the isotropic chemical shifts are summarised in Table 2. The spectra are characterised by some very narrow resonances and accompanying sidebands. Such high-resolution solid-state 170 NMR spectra are possible in highly ionic systems as C, is small as the charge density around the oxygen ion becomes closely spherically symmetric.The (1/2, -1/2) transition from such systems is then dominantly broadened by chemical shift anisotropy and the sidebands observed in Fig. 2 do resemble h h. *c-r 1 I 400 I300 I200 I 100 I0 I 6 Fig.2 170 MAS NMR spectra of (a) Li,HfO,, (c) CaHfO,, (d) SrHfO, and (e) BaHfO, (b) Na2Hf03, such powder patterns, although accurate analysis would require both chemical shift anisotropy and quadrupole effects to be taken into account. Most spectra show two isotropic resonances that lie in the range 6 240-332. For these ternary hafnates their crystal structures have two inequivalent oxygen sites in the ratio 2 : 1 and, for sodiumt, calcium and strontium samples the spectra integrate to accurately reproduce this ratio.Given the long T' measured for BaHfO, the accurate 2: 1 ratios determined here indicate that either the T,s are very much shorter or more likely that the sites have very similar Tls and hence similar degrees of saturation. The separation of the two isotropic resonances in these compounds decreases, 18.6, 8.6 and 2 ppm for sodium, calcium and strontium respect- ively, which is related to decreasing distortion of the crystal structures (Table 2) away from cubic symmetry. Li,HfO, reson-ances show a large deviation from 2: 1 with the spectrum giving a ratio of ca. 4: 1. Although T' relaxation effects cannot definitely be ruled out, a large range of relaxation delays has been tried with no observed change in the spectrum, and a similar discrepancy in intensity was seen for the isostructural Li2Zr0313so that relaxation effects are unlikely to be causing this intensity variation.These layer structures may be suscep- tible to some systematic stacking faulting that could lead to this change of intensity, but for confirmation would need to be subjected to a detailed electron microscopy study. BaHfO, shows a single resonance as expected from a cubic system where all the oxygen sites have become equivalent. An intriguing observation is that the ratio of the isotropic 170 chemical shifts from isostructural zirconates and hafnates fall in a very limited range of 0.84-0.88 (Hf/Zr, Table 2). An extensive study of such shifts in ternary titanates and zirconates shows that the isotropic 170chemical shifts also give nearly constant ratios for the isotropic shifts of isostructural corn-pounds13 so that the titanium, zirconium and here hafnium 7 A detailed discussion of the structural chemistry and 23Na MAS NMR spectra of NaHfO, is given in ref.12. J. Muter. Chem., 1996, 6(12), 1951-1955 1953 Table 2 Crystal data on HfO, and ternary hafnates and 170 isotropic chemical shifts compound crystal system space group Hf02 monoclinic P21lC Li,HfO, monoclinic cc Na,HfO, CaHfO, monoclinic orthorhombic C2/m Pbnm SrHfO, orthorhombic Pbnm BaHfO, cubic Pm3m are having the strongest effect in determining the 170chemical shift with the alkali metal having a secondary effect Hf02-SO2 An Hf0,-Si02 gel consolidated at 500 "C and then 850 "C for 2 h penods gave a 29S1 NMR resonance at 6 -96 [Fig 3(a)] This indicates that silicon is probably in a Q3 environment (ze three bridging and one non-bridging oxygen bond) The 170NMR spectrum using MAS at 9 4 kHz shows four isotropic peaks at 6 343, 252, 139 and -6 4 [Fig 3(b), the shoulder at 6 cu 380 is from the ZrO, rotor] All peaks, even those associated with ionic hafnium oxide-like environments are quite broad, indicating that the material is still highly amorph- ous Fast spinning is advantageous as it largely removes spinning sidebands and causes efficient narrowing of even Si-0-Si resonances which have quite large C, values and these sites produce the structured resonance at 6 -64 The four resonances gwen above can be attributed to OHf,, OHf,, HfOSi and OSi2 respectively 170and 29S1 give a consistent picture of the structure which is amorphous and has silicon- and hafnium-rich parts, but there is some hafnium present in the silica-nch part as indicated by the presence of HfOSi bonding and the Q3 nature of the silicon environment It is also clear from this data that via this gel route single-phase HfSiO, has not formed, as all the oxygen would be present as Hf-0-Si linkages and the silicon resonance would be much more positive corresponding to a Qo unit Comparison of the 170peak positions from ZrOSi" and HfOSi gves a ratio of 090, close to the range for the crystalline hafnates, again showing that the metals are having some influence (although equating the peak position with the isotropic chemical shift is not as certain for such a resonance compared to the more ionic linkages) The effect of the silicon is to move the shift Fig.3 (a) 29S1 and (b) 170MAS NMR spectra of an equimolar Hf0,-SiO, gel heated to 850°C (the shoulder at 6 ca 385 is from the ZrO, rotor) 1954 J Muter Chern, 1996,6(12), 1951-1955 ~~~ structure type 6,so (l70) ratio 6,,, ("0) Hf/Zr baddelyite-t ype 267, 336 0 84 Li2Zr0,-type 237 8,249 2 0 85 Na,ZrO,-t ype GdFe0,-type 239 8, 258 4 292 7, 284 1 0 84 0 87 GdFe0,- type perovskite 296, 298 331 9 0 87 0 88 well outside the range associated with the crystalline alkali- metal and alkaline-earth-metal hafnates and zirconates Hf02-GeO, For comparison, gels in the system HfO2-GeO2 were formed, heating the gel successively from room temperature to 100, 300, 500 and 950°C, and subsequently at room temperature recording the 170MAS NMR spectrum (Fig 4) Again several broad 170 resonances are observed which are summarised in Table 3 The peaks at 6 ca 330 and 240 approximately agree with OHf3 and OHf, from pure Hf02 gels The identity of the resonance around 6 115 is less certain Given the stoichiometry A Fig.4 170 MAS NMR spectra of an equimolar Hf0,-GeO, gel (a)dned at room temperature and then heated to (b) 100 "C, (c) 300 "C, (d)500 "C and (e)950 "C (the peak at 6 ca 385 is from the ZrO, rotor) Table 3 Heat treatments and 170 NMR peak positions of Hf02 GeO, gel sample heat treatment W70)(k5 ppm) G1 dried at 25 "C 327,239, 112 G2 G3 G4 G5 G1+ 1 h, 100 "C G2 + 1 h, 300 "C G3 + 1h, 500 "C G4 +6 h, 950 "C 332, 243, 118 340,248, 118 340, 245, 115 357,295, 115 The peak at 6 ca 385 is from the ZrO, rotor of the gel an immediate assignment could be to GeOGe linkages, indicating that the system is completely phase separ- ated into GeO, and Hf02 domains.However, GeO, gels produce broader I7Oresonances with well defined quadrupolar structure and different shift.', With the magnetic field applied here and the MAS employed it is likely that such linkages would not be narrowed and hence are mostly lost from the MAS NMR spectrum. The XRPD pattern of the product heated to 950°C shows it contains a crystalline component of HfGeO,.This has the scheelite structure with GeO, units linked by hafniums and a possible assignment of the 6 115 peak is to HfOGe linkages. The shift is quite close to the position of HfOSi at 6 139. The gradual increase of the relative intensity of the 6 115 peak reflects an increase in the fraction of GeO, units linked by hafniums, which may be an indication that the sample is becoming more single phase. The HfOGe resonance also begins to show some structure and can be fitted to a quadrupole lineshape with C, =5.2 MHz, an asymmetry parameter of 0.65 and an isotropic chemical shift of 6 185. Such behaviour contrasts with the HfSiO, sample, which at 850°C still showed marked phase separation.This paper indicates that 170NMR is a very subtle atomic-scale probe of the structures of a wide range of hafnate-based materials. Conclusions Hf02 gels show two resonances that can be attributed to OHf, and OHf, that shift with heat treatment, reflecting structural development in the amorphous gel and show marked nar- rowing on crystallisation into the monoclinic form, but unlike ZrO, there was never any evidence of the tetragonal form being present. Crystalline alkali-metal and alkaline-earth-metal hafnates show very high-resolution spectra, with the number of resonances and their splitting reflecting the distortion of the crystal structure. Comparison of the isotropic 170chemical shifts of the isostructural hafnates and zirconates shows an almost constant ratio of 0.86f0.02.Binary gels of HfO, with SiO, and GeO, show there is a higher tendency to form a solid solution with GeO,. M.E.S. thanks the EPSRC for funding research into the development of 170characterisation of oxide materials through GRlJ2393 8. References 1 T. J. Bastow and S. N. Stuart, Chem. Phys., 1990,143,439. 2 E. Oldfield, C. Coretsopoulos, S. Yang, L. Reven, H. C. Lee, J. Shore, 0.H. Han, E. Ramli and D. Hinks, Phys. Rev. B, 1989, 40, 6832. 3 R. K. Harris, M. J. Leach and D. P. Thompson, Chem. Muter., 1992,4,260. 4 R. Dupree, Z. P. Han, A. P. Howes, D. MckPaul, M. E. Smith and S. Male, Physicu C, 1991, 175,269. 5 T. J. Bastow, M. E. Smith and H. J. Whitfield, J. Muter. Chem., 1992,2, 989. 6 T. J. Bastow, A. F. Moodie, M. E. Smith and H. J. Whitfield, J. Muter. Chem., 1993,3, 697. 7 F. Ali, M. E. Smith, S. Steuernagel and M. E. Smith, J. Muter. Chem., 1996,6,261. 8 M. E. Smith and H. J. Whitfield, J. Chem. Soc., Chem. Commun., 1994,723. 9 P. J. Dirken, M. E. Smith and H. J. Whitfield, J. Phys. Chem., 1995, 99,395. 10 T. J. Bastow, L. Murgaski, M. E. Smith and H. J. Whitfield, Muter. Lett., 1995,23, 117. 11 P. J. Dirken, R. Dupree and M. E. Smith, J. Muter. Chem., 1995, 5, 1261. 12 J. V. Hanna, M. E. Smith and H. J. Whitfield, J. Am. Chem. Soc., 1996,118,5772. 13 T. J. Bastow, P. J. Dirken, M. E. Smith and H. J. Whitfield, unpub- lished results. 14 T. J. Bastow, R. Dupree, M. E. Smith and H. J. Whitfield, unpub- lished results. Paper 6/04406I; Received 25th June, 1996 J. Matev. Chem., 1996, 6(12), 1951-1955 1955