Journal of Nanoparticle Research 3: 433–441, 2001.© 2001 Kluwer Academic Publishers. Printed in the Netherlands.

Homogeneity of lanthanide doped ceria nanocrystal dispersions usinghigh-resolution transmission electron microscopy

A. Hartridge1, A.K. Bhattacharya∗,1, R.E. Dunin-Borkowski2 and J.L. Hutchison2

1Warwick Process Technology Group, Department of Engineering, University of Warwick,Coventry CV4 7AL, UK; 2Department of Materials, University of Oxford, Oxford OX1 3PH, UK;∗Author for correspondence

Received 1 May 2001; accepted in revised form 29 August 2001

Key words: nanocrystal, sol–gel, TEM, oxide

Abstract

Aqueous sols of crystalline solid solutions of the general formula Ce1−xLnxO2−x/2�x/2 [Ln = entire lanthanide range,x = 0–0.50 and � = anion vacancy] were synthesised using inorganic materials. The nanocrystals were dispersedon an amorphous silica support and individual crystals examined for composition and structure by high-resolutiontransmission electron microscopy (HREM) and energy dispersive analysis of X-rays (EDX). High-resolution picturesshowed the nanocrystals to be mainly between 5 and 10 nm in size, rarely elongated and with atomically cleanfaceted surfaces.

Diffractograms were used to measure the lattice parameters of the face-centre-cubic (fcc) fluorite structures givingvalues, depending on dopant species and concentration of dopant, in the range 5.5–5.8 Å. EDX analysis of individualcrystals with the same nominal composition compared closely to each other and to that of an average over a largearea indicating good homogeneity from the preparation method.

Introduction

The synthesis and study of nanocrystalline materialshas only recently, in the last decade or so, been exploredto examine those solids whose properties show a grad-ual transition from molecular structures to solid-statematter as the particle size increases. Such materials canshow greatly different optical, electronic and catalyticproperties from their macro-crystalline counterparts,as atomic orbital overlap gradually increases from dis-crete molecular orbital energy levels to splitting of thequantised energy levels and finally to the continuum ofenergy bands in a macro-crystalline lattice. Band gapsbetween valance and conduction bands can be seen toincrease by several electron volts in CdS, for example,as the size of the nanocrystal is reduced (Weller, 1993).As a result of the decreasing size and increasing surfaceto volume ratio of such particles, magnetic ordering can

collapse and crystal lattices become distorted (Gleiter,1991).

So far, the study of nanocrystalline ceramic oxideshas been limited to the powder form where inter-particle boundary effects can greatly influence theproperties of the individual nanocrystals (Gleiter,1991). Analysis of individual, multi-componentceramic oxide nanocrystals would require synthesistechniques that allow the crystalline materials to bedispersible in a liquid medium, without any organicadditives, and therefore enabling dispersion on a non-interactive medium. It is important however, to ensurethat in such a preparation method, the composition,size and crystallographic structure of all dispersednanocrystals are homogeneous.

This paper therefore, describes the preparationand subsequent use of high-resolution transmissionelectron microscopy (HREM) to characterise such

434

nanocrystals produced for the lanthanide doped ceriasolid solution series.

Lanthanide doped CeO2 of the general formulaCe1−xLnxO2−x/2�x/2 [Ln = all lanthanide elements,x = 0–0.50 and � = anion vacancy] possessesthe face-centre-cubic (fcc) fluorite structure, spacegroup Fm3m, over a wide range of dopant concen-trations. Partial replacement of CeIV ion with LnIII

ions in the lattice creates a corresponding number ofanion vacancies. The resulting solid solutions havehigh ionic conductivity and as such have been widelystudied in the last two decades for their applicationsin fuel cells (Inaba, 1996; Takahishi, 1995; Eguchi,1992; Ushida, 1998; Mishima, 1998), as inorganic pig-ments (Sulkova, 1998; 1999), catalysts (Cho, 1991;Kundakovich, 1998) and oxygen sensors (Henault,1983; Gnorich, 1991). A vast majority of the tech-niques used to prepare these solid solutions employsolid-state methods, firing at high temperatures inexcess of 1400◦C to produce single-phase dense pow-ders, which in turn must be pressed and sintered intodense components suitable for devices. More recently,techniques such as oxalate co-precipitation (VanHerle,1996; 1998) and hydrothermal syntheses (Yamashita,1995; Huang, 1997; Shuk, 1999) have been used tolower the temperature of synthesis to below 1000◦Cand the crystallite sizes to around 50 nm, but still pro-duce materials in powder form which need to be pressedand sintered. With the obvious benefits in miniaturis-ing devices such as fuel cells and sensors by coatingthin layers of electrolytes and ionic conductors, moreadvanced preparation techniques are needed to producefilms of these compounds on metal or glass substrates.Some researchers have tackled this problem with acomplicated chemical vapour deposition approach toproducing thin films of ceria and gadolinium dopedceria (McAleese, 1998), but with limited success,whilst others have used expensive organic sol–gel pre-cursors to produce electrolyte bi-layers (Mehta, 1998).By the use of an inorganic sol–gel synthesis route how-ever it was shown for the first time some years ago(Bhattacharya, 1996), that single-phase materials ofthis type could be produced at very low temperatures asnanocrystals but more importantly, they are dispersiblein an aqueous medium. These aqueous sols are stableover a wide concentration range up to 700 g l−1 and con-tain particles well below 25 nm with crystallite sizes aslow as 3 nm. Using these sols it was possible to producehigh quality dense thin films with minimal effort withpotential use as counter electrodes in opto-electronicdevices. With the fabrication of these thin transparent

films came for the first time detailed structural and opti-cal data of the lanthanide doped ceria solid solutionnanocrystals (Hartridge, 1998; 1999; Krishna, 1998).Transmission electron microscopy (TEM) is the idealtechnique to explore the structure and compositionof nanocrystalline materials. One research group haslooked at yttrium doped ceria nanocrystals in the bulkagglomerated form (Bai, 1999) and another has lookedat bulk ceria nanocrystalline structure with temperature(Guillou, 1997). However, until now the structure andcomposition of individual unaggregated nanocrystalsof lanthanide doped ceria produced by this method orany other has not been explored.

It was found that dispersions of these nanocrystalson a high surface area amorphous silica would allowcharacterisation of individual nanocrystals in a high-resolution electron microscope without interferencefrom the support. The advances in nanoprobe electronmicroscopy linked with improved energy dispersiveanalysis of X-rays (EDX) technology allow the crys-tallographic and compositional analysis of individualcrystals to be explored and thereby be compared witheach other and to those same properties in the bulkpowders and films.

Experimental

A series of solid solutions with the general for-mula Ce1−xLnxO2−x/2�x/2 [Ln = lanthanide (III/IV),x = 0–0.5 and � = anion vacancy] was preparedusing techniques described elsewhere [22]. Briefly,Cerium (III) Nitrate was simultaneously precipitatedand chemically oxidised to produce a crystalline ceriahydrate with a crystallite size of around 5 nm. Repeatedwashing and conditioning produced a bright yellowcoloured moist ceria powder, which was impreg-nated with the requisite quantity of lanthanide nitrateand dried to a granular solid. This was heated in amuffle furnace under strict temperature–time param-eters, which depended on the amount of lanthanidedoping. The resulting powder was dispersed in dis-tilled water to yield a transparent sol, which was inturn centrifuged to remove any undispersed product(<5%). The transparent sols produced were filteredto <200 nm to remove any dust, and their concentra-tion adjusted to 25 g l−1. Separately, synthetic amor-phous silica (Crossfield C200) was calcined to 600◦Cto remove any adsorbed water/CO2 etc, cooled to roomtemperature and divided into gram quantities. To eachgram, a different sol was added dropwise to just fill thepores without over-wetting the support. The silica was

435

dried at 105◦C and briefly calcined to 450◦C to removeany anions and water. This process was repeated onceto give a resulting supported nanocrystallite concentra-tion of ∼15% by weight.

To prepare the TEM samples, individual powdersamples were dispersed in methanol and placed in anultrasonic bath for 5 min to break up the silica aggre-gates. The suspension was allowed to settle for 30 sand a holey carbon copper grid was then immersed atthe surface, pulled out immediately and dried underan infrared lamp for 1 h. The grid was then mountedin a double-tilt low-background holder and placed inan oxygen plasma to decarburise the surface, ready foranalysis.

TEM was performed on a JEOL JEM-3000F instru-ment with a field-emission-gun operating at 300 kV andin the ultra-high resolution configuration, with a lowspherical aberration coefficient of 0.60 mm, a point res-olution of below 0.17 nm and an information limit ofbelow 0.10 nm.

The microscope was equipped with an OxfordInstruments ISIS 300 EDX system with an ultra-thinwindow Si(Li) detector. Its features are described indetail in the paper ‘The development and assessmentof a high performance field-emission-gun analyticalHREM for materials science applications’ (Hutchison,1999).

High-resolution images were recorded directly ontoa 1024 × 1024 pixel Gatan model 794 CCD cameraat a nominal microscope magnification of 600 k×corresponding to a 29.1 nm field of view. Latticefringe spacings were measured from Fourier transformdiffractograms of images of single crystals. The magni-fication had been previously calibrated using a Si [110]lattice image with the sample at the same height in themicroscope.

EDX spectra were obtained from individualnanocrystals using a probe below 5 nm in diameterwithout tilting the sample. The acquisition time wastypically 60–100 s. The spectra were only interpretedqualitatively mainly due to the strong overlap of theceria signal with those of the dopant cations. At leastthree images and spectra were obtained for each sam-ple and the lattice spacing from several crystals withineach image were analysed.

Results and discussion

All sols produced by this preparation route were trans-parent and stable over a wide concentration range and

with time (over a period of several months). Ideal sols,that is, those with narrow, low size distributions andhigh stability, had a pH of around 4.0 and a conduc-tivity at 25 g l−1 of 1–2 mS. Sol colours ranged frompale yellow to white for La3+, Gd3+, Y3+, Dy3+, Yb3+,with dopants increasing in concentration; pale brown toblack for Pr3+,4+ as dopant, pale green to green–blue forNd3+, yellow for Sm3+ and pale peach to pale pink forEr3+. Photon correlation spectroscopy (PCS) measure-ments showed the particle size for all sols was between5 and 30 nm, the average size generally increasing withincreasing dopant concentration and increasing withthe larger ionic radii, Ce0.50La0.50O1.75 having the largestaverage particle size whilst Ce0.95Yb0.05O1.975 and pureCeO2 have the smallest. Details of the sol particle sizeand structural information on the gels, powders and thinfilms of these materials have been published elsewhere(Bhattacharya, 1996; Hartridge, 1998; 1999; Krishna,1998).

The theoretical basis of electron diffraction in theelectron microscope is discussed in detail elsewhere(Eddington, 1975). Briefly, when a monochromaticbeam of electrons passes through a single crystal, thelattice planes act as a diffraction grating scattering theelectrons into a characteristic spot pattern. The charac-teristics of this pattern depend on the basic crystallo-graphic structure of the crystal and its zone axis that isthe axis normal to the incident beam. The interpreta-tion of the created electron diffraction pattern is aidedby the principles of the reciprocal lattice (Eddington,1975) which assigns each spot in the pattern with a cor-responding (hkl) plane. For ease, basic spot patternsfor each zone axis of a particular crystal system andits corresponding assigned spot pattern are available inthe text (Eddington, 1975). This allows crystal systemsand zone axes to be assigned for each crystal throughthe ratio of lengths to each spot in the pattern and theirangle to each other.

In such a spot pattern, the radius from the centralspot R to the spot produced by a specific hkl plane isalso related to the interplanar spacing d of the zone axisby the expression

R = λL

D

where λ is the electron wavelength and L is the cameralength. For a fcc structure, the plane separation d isrelated to the lattice parameter a by the equation

1

d2= 1

a2{h2 + k2 + l2}

436

Figure 1. High-resolution TEM picture of a Ce0.85Nd0.15O2−x

nanocrystal with corresponding spot diffraction pattern.

where h, k and l are the Miller indices of theplane. Hence, by accurately measuring the distancesto various spots from the central beam, the latticefringe spacings for the single crystals can be mea-sured and the lattice parameter for the crystal systemcalculated.

An example of a high-resolution image of ananocrystal along with its spot diffraction pattern isshown in Figure 1 specifically for a Ce0.85Nd0.15O1.925

crystal. The following is a detailed description of theanalysis procedure used for all the nanocrystals anal-ysed. First, the size of the nanocrystal is noted usingthe size bar as a reference. This shows the size of thecrystal in Figure 1 to be about 8 nm in its longest direc-tion with a very slightly oval shape. This is consistentwith the crystallite size obtained from X-ray diffrac-tion data and particle size data from PCS. The atomicplanes within the nanocrystal can be clearly seen as aseries of white dots separated by a black background.The square outline surrounding the crystal indicates thearea used for obtaining electron diffraction and EDXanalysis data.

The next step is to assign a zone axis to the spotdiffraction pattern inset to the top right of the picture,which shows a hexagonal array of dots surrounding thecentral beam. To do this we note the angle and distancesbetween the highlighted spots of the diffractogram inFigure 1. The ratio of a/b in this case is 1.16 sepa-rated by an angle of 51.4◦. Now, referring to the stan-dard reference spot patterns mentioned before, it can beseen that this spot pattern is derived from the fcc struc-ture and closest represents the [011] zone axis plane.The slight variation in angles and ratios between mea-surements from the same crystal arises, mainly from

the small tilts of individual crystals away from exactzone axis orientations, and is discussed in detail in thepaper ‘Lattice spacing in high-resolution micrographsof metal nanoparticles’ (Malm, 1997). From this stan-dard spot pattern we can now identify the three labelledspots in the Figure 1 as having been derived from the(111), (200) and (111) planes respectively. Using theequations above, we now can convert these interplanardistances (d) into lattice parameter (a) by the simplemultiplication of

√3, 2 and

√3 for the three planes

respectively. These give lattice parameter values for thecrystal of 5.65, 5.62 and 5.99 Å. The first point to noteis the size of the lattice parameter values compared tothe 5.434 Å for bulk ceria doped with neodymium ata level of 15% reported earlier. This is again an indi-cation of the extreme lattice distortions involved inthese nanocrystals when the surface to volume ratiois very large, an ∼5% increase in the observed lat-tice parameters. The large value compared to the othertwo closer values is again due to deviation of the crys-tal orientation away from the exact zone axis (Malm,1997).

The EDX spectrum of the particular crystal shownin Figure 1 is shown in Figure 2. Peaks in this spectrumshow the silica substrate and copper grids along withpeaks generated by the elements cerium, neodymiumand oxygen. No other peaks are evident. Qualitatively,we can state that the crystal contains only cerium andneodymium in the approximate ratio of 5 : 1 judgingroughly by the peak heights. This is in approximate linewith the nominal value of 15% Nd in the bulk powder.The procedure was repeated for many other crystalswithin each sample and a lattice parameter average isgenerated.

Figure 3 shows examples of micrographs and corre-sponding EDX spectra used to calculate lattice param-eters and compositions for other lanthanide dopants.The crystals analysed were almost all between 3 and15 nm in size and more commonly between 5 and10 nm. They were often strongly faceted with appar-ently atomically clean surfaces and most had fairly evendimensions in all directions that is they were rarelyvery elongated. During acquisition of EDX spectrawith an intense electron beam concentrated on a sin-gle crystal, an increase in faceting was often observed.EDX spectra in all cases confirmed the presence ofthe intended dopant species within the lattice of eachcrystal examined, and the dopant peak heights relativeto those of ceria varied in the correct sense accord-ing to the concentration of dopant expected by thenominal composition. Within each composition, the

437

Figure 2. EDX spectrum of the Ce0.85Nd0.15O2−x nanocrystal pictured in Figure 1.

dopant peak heights for each crystal were reasonablyconsistent but on occasion varied by upto a factorof 2. However, this variability was rare and may becaused by occasional differences in sample geome-try, shadowing, or perhaps by damage to certain crys-tals by the intense electron beam used to record thespectra.

Table 1 shows some of the average lattice param-eter averages versus lanthanide dopant for the indi-vidual nanocrystals analysed with a 15% doping. Thefirst point to notice is that no obvious trend existsbetween the ionic radius and the lattice parameter inthese isolated nanocrystals as it does in the unsupportedbulk powder samples where the crystals are aggre-gated. The change in lattice parameter (a), althoughsmall in the powders, is systematic, increasing withincreasing ionic radius from ytterbium to neodymiumin samples of the same composition. With thesenanocrystals however, the size of the crystals has moreeffect on the lattice parameter than the ionic radiusof the dopant, those compositions having the small-est average crystallite sizes consequently showing thelargest lattice parameters, apparantly due to the crys-tal strain. This phenomenon has been noted in EXAFSstudies on other crystal systems before (Decker,1998).

It is also clear from this table that the average lat-tice parameters for individual crystals are much largerin every case than those of the bulk powder values,5.56–5.67 Å for the nanocrystals and 5.41–5.44 Å forthe bulk powders. This relatively large difference may

be ascribed to the large surface to volume ratio of thenanocrystals, as previously mentioned, exerting exces-sive strain on the cubic lattice. The high proportionof surface atoms in the nanocrystal with respect tothe larger crystals encountered in bulk powders, opensthe structure resulting in larger lattice parameters.This phenomenon has been reported before with othersystems for example (Gleiter, 1991).

Having now analysed nanocrystals with a fixed levelof doping but varying in ionic size, the effect of dopantconcentration for a fixed lanthanide element was nowstudied. For this, two series were studied, Ce1−xYxO2−x

nanocrystals and Ce1−xPrxO2−x nanocrystals. Since wehave seen already that the effect of nanocrystal sizehas more bearing on the lattice parameter than that ofany doping, for each of these series, a large dopantconcentration range (5–50%) was used to show anytrends to the best effect. Figure 4 shows a couple ofexample analysis areas for yttrium doped ceria com-positions. Since yttrium’s EDX analysis peak at 15 eVis solitary, the relative peak heights corresponding tothe amount of yttrium in each crystal can now be seenmuch more clearly, the peak at 15 eV is roughly dou-ble in height in Figure 4(b) than it is in Figure 4(a)complying with the average bulk powder nominalcomposition.

Lattice parameters from numerous crystals were cal-culated for the yttrium and praseodymium doped seriesand the results are shown in Table 2. Lattice param-eter calculations again show values above 5.5 Å, con-siderably larger than the expected values of around

438

5.42 Å for the bulk powders. There is again thepresence of a large difference in lattice parameterwith nanocrystallite size, for crystals of the samecomposition. We can now see, despite the changes incrystallite size that a trend of increasing lattice param-eter average with increasing Pr and Y concentration

Figure 3. (Continued)

is evident. This effect, which is systematic in the bulkpowders as analysed previously, is now apparent dueto the large compositional changes between the twosets of three samples, and is again a good indicationthat the nanocrystals contain the same amount of dop-ing according to composition as the bulk averaged

439

Figure 3. Examples of high-resolution micrographs and corresponding EDX spectra (a) Ce0.85Sm0.15O2−x , (b) Ce0.85Gd0.15O2−x ,(c) Ce0.85Dy0.15O2−x , (d) Ce0.85Er0.15O2−x and (e) Ce0.85Yb0.15O2−x nanocrystals.

Table 1. Lattice parameter versus dopant ionic radius for a seriesof Ce0.85Ln0.15O2−x nanocrystals

Ln Ce0.85Ln0.15O2−x nanocrystals

Nd Sm Gd Dy Er Yb

Av. lattice 5.60 5.57 5.67 5.62 5.62 5.56parameter/Å

nominal composition. We can also see that the over-allvalues for Y doping are smaller than those for Prdoping of the equivalent dopant concentration. Thisis consistent with the Y3+ ion (90 pm) being smallerthan the Pr3+ ion (99 pm) confirming the correct com-position and homogeneity of the nanocrystals for eachnominal composition.

A series of zirconia doped ceria solid solutionnanocrystals was also studied during this investiga-tion showing identical behaviour to the yttria doped

ceria series but with the fluorite lattice contracting withincrease in zirconia concentration.

Conclusions

By using advanced TEM techniques linked withEDX data, it has been possible to show that dis-persible lanthanide doped ceria nanocrystals producedby an inorganic sol–gel route over the entire range oflanthanides are homogeneous crystalline entities withthe cubic fluorite structure and that each individualcrystal has a very similar composition for a given nom-inal composition. This is an important factor when con-sidering using materials of this type for fuel cell, opticaland sensor applications since a completely homoge-neous material will have a more even response. Sincethe individual crystallites are very small and thereforepossess a high surface free energy, relatively low tem-perature treatment will result in dense thin films or

440

Figure 4. Sample analysis areas and EDX spectra of (a) Ce0.75Y0.25O1.875 and (b) Ce0.50Y0.50O1.75 dispersed nanocrystals.

Table 2. Average lattice parameter values versuscomposition for the series’ Ce1−xPrxO2−x/2 andCe1−xYxO2−x/2 nanocrystals

Nanocrystal Average latticecomposition parameter/Å

Ce0.95Y0.05O2−x 5.56Ce0.75Y0.25O2−x 5.61Ce0.50Y0.50O2−x 5.62Ce0.90Pr0.10O2−x 5.60Ce0.75Pr0.25O2−x 5.65Ce0.50Pr0.50O2−x 5.66

devices and avoid problems that arise from high sin-tering temperatures such as surface segregation, metalevaporation and excessive oxygen loss.

References

Bai W., K.L. Choy, N.H.J. Stelzer & J. Schoonman, 1999. SolidState Ionics 116, 225–228.

Bhattacharya A.K., A. Hartridge & K.K. Mallick, 1996. J. Mater.Sci. 31, 5005.

Cho B.K., 1991. J. Catalysis 131, 74–87.Eddington J.W., 1975. “Practical Electron Microscopy in

Materials Science”, Monograph 2, MacMillan (Philips Tech-nical Library).

Eguchi K., T. Setoguchi, T. Inoue & H. Arai, 1992. Solid StateIonics 52, 165–172.

Eguchi K., T. Setoguchi, T. Inoue & H. Arai, 1992. J. Electrochem.Soc. 139, 2875–2880.

Gleiter H., 1991. J. Appl. Cryst. 24, 79–90.Gnorich A. & H.G. Beie, 1991. Sens. Actuators B Chem. 4,

393–399.Guillou N., L.C. Nistor, H. Fuess & H. Hahn, 1997. Nanostruct.

Mater. 8, 545–557.Hartridge A., M. Ghanashyam Krishna & A.K. Bhattacharya,

1998. J. Phys. Chem. Solids. 59, 859–866.Hartridge A., M. Ghanashyam Krishna & A.K. Bhattacharya,

1998. Int. J. Mod. Phys. B 12(14), 1573–1583.Hartridge A., M. Ghanashyam Krishna & A.K. Bhattacharya,

1998. J. Mater. Sci. Lett. 17, 1329–1331.Hartridge A., M. Ghanashyam Krishna & A.K. Bhattacharya,

1999. Mater. Sci. Eng. B B57, 173–178.Henault M. & J. Fouletier, 1983. Solid State Ionics 9–10,

1277–1281.

441

Huang W., P. Shuk & M. Greenblatt, 1997. Chem. Mater. 9,2240–2245.

Hutchison J.L., R.C. Doole, R.E. Dunin-Borkowski, J. Sloane &M.L.H. Green, 1999. JEOL News 34E, 10–15.

Inaba H. & H. Tagawa, 1996. Solid State Ionics 83, 1.Ghanashyam Krishna M., A. Hartridge & A.K. Bhattacharya,

1998. Mater. Sci. Eng. B B55, 14–20.Kundakovic L. & M. Flytzani-Stephanapoulos, 1998. App.

Catalysis A General 171, 13–29.Malm J.O. & M.A. O’Keefe, 1997. Ultramicroscopy 68, 13–23.McAleese J. & B.C.H. Steele, 1998. Corros. Sci. 40, 113–123.Mehta K., R. Xu & A.V. Virkar, 1998. J. Sol–Gel Sci. Technol.

11, 203–207.Mishima Y., H. Mitsuyasu, M. Ohtaki & K. Eguchi, 1998.

J. Electrochem. Soc. 145, 1004–1007.

Shuk P. & M. Greenblatt, 1999. Solid State Ionics 116,217–223.

Sulkova P., M. Trojan & Z. Solc, 1998. Dyes and Pigments 37,65–70.

Sulkova P. & M. Trojan, 1999. Dyes and Pigments 40, 87–91.Takahashi T., 1995. Bull. Electrochem. 11, 1.Ushida H., H. Suzuki & M. Watanabe, 1998. J. Electrochem. Soc.

145, 615–620.VanHerle J., T. Horita, T. Kawada, N. Sakai, H. Yokokawa &

M. Dokiya, 1996. Solid State Ionics 86–88, 1255–1258.VanHerle J., T. Horita, T. Kawada, N. Sakai, H. Yokokawa &

M. Dokiya 1998. Ceram. Int. 24, 229–241.Weller H., 1993. Angew. Chem. Int. Ed. Engl. 32, 41–53.Yamashita K., K.V. Ramanujachary & M. Greenblatt, 1995. Solid

State Ionics 81, 53–60.