J O U R N A L O F C H E M I S T R Y Materials Extended defect structures in zinc oxide doped with iron and indium Tom Ho� rlin,*a Gunnar Svenssonb and Eva Olssonc† aDepartment of Inorganic Chemistry, Stockholm University, SE-106 91 Stockholm, Sweden. E-mail: tom@inorg.su.se bDepartment of Structural Chemistry, Stockholm University, SE-106 91 Stockholm, Sweden cDepartment of Physics, Chalmers University of Technology and Gothenburg University, SE-41296 Gothenburg, Sweden Received 8th July 1998, Accepted 26th August 1998 The eVects of iron- and indium oxide doping on the structure and magnetic susceptibility of ZnO have been studied.The nominal compositions were InxFe2-xO3(ZnO)n with 0x1 and n=23, 48 and 98. Magnetic measurements showed the iron-doped samples to be paramagnetic, with a behaviour indicating antiferromagnetic coupling between the iron ions.HREM studies showed that indium and iron are incorporated as layer defects of two kinds. One type forms cubic close packed (ccp) planes perpendicular to the c axis, and the other appears as corrugated layers inserted between the former. The folds in the corrugated layers consist of alternating (114) and (1149) planes.Analytical transmission electron microscopy studies revealed that indium prefers the ccp layers, whereas iron can be found in both types of defects. Structural models based upon the experimental results are presented. The In2O3–ZnO system was first studied by Kasper,5 who Introduction found an homologous series with the general formula Zinc oxide is one of the most important semiconductors used In2ZnnO3+n.Cannard and Tilley6 later reinvestigated the in large quantities. Although ZnO devices are quite diVerent system using X-ray powder diVraction and transmission elecfrom components made of e.g. silicon and gallium arsenide, tron microscopy. They suggested an intergrowth structure there are similarities that can be utilised.Examples are the consisting of slabs of wurtzite-type ZnO separated by bixbyiteenhancement of conductivity by doping and the creation of type layers of In2O3. The structures suggested by Cannard non-linear junctions. The project described in this article was and Tilley have recently been confirmed by McCoy et al.7 initiated by a study of the possibility of using ZnO as electrode The phase compsitions in the In2O3–Fe2O3–ZnO system material in the smelting of aluminium.have been investigated by Nakamura et al.,9,15 Kimizuka It is easy to promote the electric conductivity of zinc oxide et al.8,10,11 and Siratori and Kimuzuka.13 They found a number by doping it with normally trivalent metals. Some of these of phases which can be described by the formula ions may enter as divalent, with one rather loosely bound InxFe2-xO3(ZnO)n were n varies between 1 and 13 and x as electron that may dissociate and participate in n-type conduc- 0.72x1.00 for n=1 and as 0x2.0 for n=13.They tion. Only n-conduction has been confirmed in zinc oxide; suggest that the compounds with n=1–6 and x=1 are isostrucp- conduction has not yet been observed. tural with their lutetium analogues LuFeO3(ZnO)n, which It is known1–3 that the addition of gallium or indium ions have been determined by X-ray single-crystal diVraction techenhances the electric conductivity of ZnO, which soon becomes niques.23 The structures consist of wurtzite-type ZnO slabs metallic.For gallium a maximum of the conductivity has been intergrown with single layers of indium atoms in octahedral observed around 0.4 atom% of the dopant.3 This result is in cavities formed by cubic close packed oxygen. The iron atoms agreement with what we have found for both gallium and form a dilute solution in the ZnO slabs.For structural reasons indium.4 Iron-doped ZnO stays semiconducting, on the other the ZnO tetrahedra have to be connected with their apices to hand, with a maximum iron content of 0.3–0.5 atom% at the ccp layers.The ZnO slabs have diVerent polarity on either 1000 °C. The reason for this is that Fe2+ has a higher ionisation side of the indium-containing layers. Consequently, an inverenergy than In2+. Simultaneous doping with 1 atom% indium sion of the polarity must occur within the ZnO slabs between and iron does not markedly increase the conductivity at low these layers, which can be achieved by a gradual displacement temperatures, since at this doping level there is enough trivalent of the metal atoms from one tetrahedron to the neighbouring iron to trap all loosely bound electrons according to the one via the shared triangular face.reaction:4 HREM studies on InFeO3(ZnO)n, however, revealed a more complex picture.19–22 At least for n6, wave-like defects were In2++Fe3+�In3++Fe2+ (1) found between the indium oxide layers.EDS spot analysis If doped ZnO is prepared in air, only a fraction of the trivalent showed an increased signal from iron in these defects. dopant ions will be reduced to divalent. Divalent dopants may In this article we will describe how low concentrations of substitute for zinc without structural consequences, but the iron and indium dopant, lower than those reported in the trivalent ions entering ZnO must be accompanied by structural literature, are structurally accommodated in ZnO.We will defects such as metal vacancies or other alien structure elements discuss what structure elements are formed and how are they in order to conserve charge neutrality.Sooner or later these related to the phases found at higher dopant concentrations, defects may start to order in the host structure, eventually and how this doping influences the magnetic properties of the leading to new phases. compounds. There have been numerous studies of binary and ternary systems with ZnO and one or two trivalent metal oxides.2–22 Experimental Powders of ZnO (Fotofax), Fe2O3 (Merck p.a.) and In2O3 †Present address: Analytical Materials Physics, The A° ngstro�m Laboratory, Uppsala University, Box 534, SE 75121 Uppsala, Sweden.(Aldrich 4N) were milled with SIALON balls in high-density J. Mater. Chem., 1998, 8(11), 2465–2473 2465polyethylene bottles. The powder mixtures were pressed to pellets, which were heated at 1400 °C for 3 days, embedded in loosely packed ZnO powder.After the heating period the furnace was switched oV and the door was opened. Earlier work9 indicates that very long heating periods (weeks) are required to obtain ordered phases. Our dopant (In+Fe) compositions—2, 4 and 8% of the total metal content—were more dilute, so we did not expect to reach complete threedimensional ordering. The formula can be written as InxFe2-xO3(ZnO)n where n=98, 48, 23, and x=0.0, 0.5, 1.0.The inner parts of the pellets were used for characterisation. Solid pieces were cut for the magnetic susceptibility measurements and for the electric measurements. Fine powders were ground for X-ray and HREM investigations. X-Ray powder diVraction patterns were recorded with a focusing camera of Guiner–Ha�gg type, using Cu-Ka1 radiation and with silicon added as internal standard (a=5.431 A° ).The Fig. 1 The XRD pattern for the n=23, x=1 sample. The additional films were evaluated with a scanner system.24 lines not indexed with a ZnO type unit cell are marked with arrows. For high-resolution electron microscopy studies (HREM) we used a JEOL JEM 200CX operated at 200 kV, capable of 2.4 A° point resolution, and a JEOL JEM 3010 operated at 300 kV, capable of 1.7 A° point resolution; and for the analytical TEM studies a field-emission gun Philips CM200 supertwin microscope operated at 200 kV, equipped with an Oxford LINK ISIS EDX system and a Gatan imaging filter for electron energy loss spectroscopy (EELS) and electron spectroscopy imaging (ESI).For high-resolution and analytical transmission electron microscopy studies, dispersions in nbutanol of finely crushed samples were put on perforated carbon films supported by a copper grid. A model 7000 AC susceptometer from Lake Shore, equipped with a ‘Cryodyne’ closed-cycle helium refrigerator from CTI Cryogenics, was used for the measurements of magnetic susceptibility.We have also aomputer-controlled diVerential transformer to the system, because the excellent balance of the pick-up coils at room temperature and moderate frequencies declines at low temperatures and high frequencies, and it is therefore necessary to have a device that can restore the balance when the temperature and frequency vary. The system is controlled by an in-house developed computer program.The samples were pieces around 0.7 g in weight. The magnetising field (H) ranged from 177 A m-1 up to 707 A m-1; the stronger fields were used for samples with less iron. The frequency was 2000 Hz. The temperature was varied from ca. 11.3 K up to 320 K. The susceptibilities of the samples were found to be independent of frequency and magnetising field.Fig. 2 (a) Unit cell volume and (b) c/a ratio versus 1/n and x, for InxFe2-xO3(ZnO)n, n=23, 48, 98, 2 and x=0, 0.5, 1.0. Results 1.0 samples, as shown in Fig. 2(b). In all samples c/a was less X-Ray powder diVraction than the ideal value for hcp packing: c/a=2(2/3)1/2#1.630. All the X-ray powder patterns showed strong lines from a HRTEM studies hexagonal unit cell closely similar to that of ZnO.The XRD pattern for the n=23 and x=1.0 sample is shown in Fig. 1. We first looked for crystal fragments oriented along a [100] (i) The XRD patterns of the n=98 and x=0.0 samples direction of ZnO, since the two shortest projected distances revealed no additional lines, and only one faint extra line was between zinc atoms (2.82 and 2.77 A° ) in the corresponding seen for the x=0.5 and 1.0 samples.(2910)-projection are well within the resolution of our HRTEM. (ii) The XRD patterns of all the n=48 samples contained [The diVerence between these distances is due to a slight the same extra line mentioned above, and for the x=0.5 and distortion of the tetrahedra (c/a<1.630).] We then looked 1.0 samples of n=48, five faint additional extra lines were parallel to possible planar defects perpendicular to the c-axis, observed.such as those reported by Uchida et al.21 Fig. 3 shows the (iii) The X-ray powder pattern of the n=23, x=0.0 sample HREM image and the corresponding electron diVraction showed three weak extra lines, and those of the x=0.5 and pattern taken along [100] of a crystal found in a sample of 1.0 samples seven and five weak extra lines, respectively.the nominal composition InFeO3(ZnO)23. A magnification of The unit cell volume for the ZnO phase in the n=98 samples the edge is shown in Fig. 3(c). In the images we see two types was the same as for undoped ZnO, while it increased for n= of irregularities in the ZnO sructure, similar to those reported 48 and 23 samples as shown in Fig. 2(a).The increase was by Uchida et al.21 One type is a sharp uninterrupted defect, largest for samples containing both indium and iron. The c/a unwrinkled and perpendicular to the c-axis. Most probably these defects are extended in depth, forming layers parallel to ratio increased with decreasing n value, but less so for the x= 2466 J. Mater. Chem., 1998, 8(11), 2465–2473Fig. 3 (a) An electron diVraction pattern and (b) corresponding lattice image taken along [100] of a crystal found in a sample with nominal composition InFeO3(ZnO)23. The width of the bouncing defect at the arrows corresponds to a crystal thickness of 60 A° , see text. (c) Magnification of a part of the edge in (b). The shift of 0.9 A° along [100] when crossing the (001) defect (marked with an arrow) is emphasised with a line.The shift when crossing the bouncing defects (marked with arrows) is emphasised with two lines. the (001) planes of the ZnO network. In the higher magnifi- cation one can see that the dark spots ascribed to zinc atoms at this defocus (ca. -400 A° ) are shifted 0.9 A° along [120] when crossing this defect. This distance corresponds to a third of the height of a tetrahedral face.The second type of defect was not always seen in the crystallites viewed along [100] or its equivalents. These defects have a more diVuse contrast, bouncing back and forth between the (001) defect layers forming zigzag patterns. In this projection (2910), they have the shape of ribbons with more or less pronounced rims, increasing in width with the distance from the edge of the crystallites.(They appear as inclined layers when viewed along the [100] axis.) A shift of the contrast along the c-axis is observed in Fig. 3 when crossing a defect of the second type, which might be interpreted as a small shift of the zinc atoms along the c-axis. Weak streaking along the c-axis is seen in the corresponding ED patterns in Fig. 3(a), caused by the variation of the distance between the defects parallel to the (001) planes of ZnO. It has to be mentioned that anomalies of the same type were found in samples doped with less indium: x=0 and 0.5. In order to determine the orientation of the defects in three dimensions, some crystallites were tilted 30° around the c-axis to the [1190] or the equivalent [210] directions.After this operation the bouncing defects were still seen in the HREM images in some cases, while in others they had disappeared. In both cases the first type of defect parallel to (001) remained clearly visible. An HREM image with bouncing defects visible in the new direction, and the corresponding electron diVraction pattern, are shown in Fig. 4. In accordance with reports on highly doped compounds in this system,19–21 we suggest that one half of the impurity atoms (indium or iron) are located in the plane defect layers.The number m of ‘ZnO’-planes between the defect planes would Fig. 4 (a) HREM image of a crystallite viewed along [1190] found in then be m=n+1 [n from the formula InxFe2-xO3(ZnO)n]. As a sample with nominal composition InFeO3(ZnO)48.(b) The correexpected due to the short firing period, this number m varies sponding electron diVraction pattern, and magnification of a diVraction spot exhibiting a star of intensity along [114] and [1149]. markedly in the crystallites. However, the average of m for J. Mater. Chem., 1998, 8(11), 2465–2473 2467Table 1 Average distance between defect layers in the HREM images electron beam will thus be either approximately 60 or 0°, and of crystallites studied in InxFe2-xO3(ZnO)n samples, m=n+1 the bouncing defects will be visible in one case out of three.Clearly this defect structure breaks the hexagonal wurtzite m x Observed ma symmetry. We can therefore introduce an orthorhombic cell with reduced symmetry, related to the hexagonal wurtzite cell 24 0.0 24 0.5 30 by: 1.0 24 49 0.0 55 1.0 51 base(orth)=A1 19 0 1 1 0 0 0 1Bbase(hex) (2) 99 1.0 90 aApproximate value.This is of course a subcell of the structures that can occur in this system. However, for simplicity of description we will the crystallites studied corresponds well with the amount of retain the hexagonal cell of ZnO. doping, as shown in Table 1.The planar defects seem to be The bounds of the ribbons formed by the bouncing defects, very sturdy and without faults. This indicates that ordering seen in the (2910) projection, are the intersections between the towards a smaller spread in m is very sluggish; otherwise we bouncing defect layers and the surfaces (top and bottom) of would see that some layers were ‘on the move’ by showing the crystal fragment.The width of the ribbons b is related to steps. A fast structural rearrangement is not to be expected, the thickness t of the crystal as b=(t sin a)/Ó3 (a=38.7°, the since it must involve interchange of atoms. angle between (114) and (001)), assuming the crystallites to When viewed along [1190] or equivalent axes in ZnO, the be perpendicular to the beam.It is thus possible to determine shortest distances between zinc atoms in the projection are the thickness of the crystals by measuring the width of the c/2=2.60 A° and a/2=1.625 A° . We can therefore only expect ribbons: for example the crystallite shown in Fig. 3(b) would them to be resolved in the c direction. In the crystallite shown then have a thickness t#60 A° at the position marked with the in Fig. 4(a), both types of extended defects are very clearly two arrows. distinguished, the bouncing defects being more narrow and Our conclusions on the orientation of the defect planes are distinct in this projection than when viewed along [100]. This shown in a drawing. In Fig. 5(a) we look straight down on gives the impression that the defect planes are now oriented the (114) and (1149) defects along [1190] in ZnO.The resem- parallel to the beam. The angle between the two types of blance to corrugated cardboard is striking. In Fig. 5(b) the defect planes is approximately 39°. The estimated value of this model is tilted 30° with respect to in ZnO. The model is wedge- angle does not vary much within diVerent parts of the crystal shaped from the edge (at the bottom) to the position marked or between diVerent crystals.If we define the direction in ZnO with arrows, and then has constant thickness. The ribbons where the bouncing defects are sharply visible to be [1190], then the bouncing defects will be layers alternately parallel to (114) and (1149). The angle between (114) or (1149) and (001) planes in ZnO should be arctan(c/2a)#arctan(1.60/2)#38.7°, which is well in accord with the observed value.In the ED-pattern taken along [1190] there is streaking along the c axis as in the [100] patterns. The (114) or (1149) boundary defects are also seen in Fig. 4(b) as a faint streaking in the shape of a weak star around the diVraction spots in the ED pattern. An enlargement of a diVraction spot is inserted in Fig. 4(b). The directions of the star points are close to the [114] and [1149] directions in reciprocal space, in agreement with the interpretation of the HREM images. The distinct 39° angle found between the planar defects and the bouncing defects when viewing along [1190] decreases a few degrees, to ca. 35°, when viewing along [100], with a larger spread in observations.The preferred cleavage surface for ZnO is reported to be (2910).25,26 It is therefore not unreasonable that the crystallites studied here are flakes extended in the equivalent (2910)-planes—perpendicular to the [100]-axis. The angle between the defect types projected on a (2910)-plane should be arctan[(c/2a)cos(30°)]#34.72°. The spread of observations is caused by the fact that the crystallites are wedge-shaped at the edges and increase in thickness with the distance from the edge, and they may not necessarily be exactly extended in the (2910)-plane.The structure of the bouncing defects together with the planar (001) defects very much resembles corrugated cardboard. The corrugated defect layers are not always seen, since in both viewing directions, [100] and [1190], there are three equivalent cases in the hexagonal setting of the sublattice.In the first view along [100] the planes of corrugated defects make a 30° angle with the incident beam, but they may also be perpendicular to the incident beam and thus not seen. In the other projection, [1190] and [210], the corrugated planes are either aligned with or 60° oV the beam.In the latter case Fig. 5 Reconstruction of the (114), (1149) and (001) defect planes. the angle may be too large to allow the observation of the (a) Viewed along [1190] and (b) along [100] in ZnO. The model is wedge-shaped from the edge to the position marked with arrows. defect layer. The angle between the corrugated layer and the 2468 J. Mater. Chem., 1998, 8(11), 2465–2473slabs remains to be answered, but it seems probable that some iron is incorporated there.The sample just described did not contain indium, so a crystallite found in a sample with nominal composition InFeO3(ZnO)48 was investigated to locate this element and the zinc, indium and iron elemental maps are shown in Fig. 8. This crystallite is viewed along an equivalent to the [1190], zone axis, in a direction where the bouncing defects are not viewed edge on, e.g.[210]. The indium map reveals a preference for the planar defects, whereas the iron and zinc signals indicate deficits in these defects. A uniform iron signal is found between the planar defects, as expected for a crystallite of finite thickness. The conclusions from this elemental mapping are (i) indium prefers the planar defect layer.Whether it is possible for indium to enter the corrugated Fig. 6 HREM image of a crystallite found in a sample with nominal layers is not clear. (ii) Iron may enter both types of defects. composition In0.5Fe1.5O3(ZnO)23, showing a well developed net of corrugated defects. Magnetic susceptibility The magnetic susceptibility measurements revealed all doped found in the HREM images are clearly seen in this orientation.samples to be paramagnetic. The diamagnetic contributions Their width increases with the thickness of the ‘crystal,’ exactly from zinc and oxygen have been removed in presented data. as observed in the HREM images. For this we used the measured susceptibility of ZnO, which An impact of a corrugated layer on one side of a plane was in excellent agreement with Pascal’s constants tabulated layer is frequently matched with an impact on the other side in ref. 26. For the various species we expect the mB2 to be 35 at almost the same place (exceptions occur). This means that for Fe3+, 24 for Fe2+ and 3 for electrons or In2+, assuming the corrugated layers are mirrored through the planar layers, spin-only contribution and high-spin configuration.(In oxides i.e. the waves of adjacent corrugated layers are 180° out of the ligand field is usually not strong enough to induce a lowphase. The traces of the corrugated layers form a diamond spin state.) One may expect divalent indium and free electrons net. This net is of course only well developed when there is to be trapped by iron according to a prior study:4 some order in the c direction.Fig. 6 (of In0.5Fe1.5O3(ZnO)23 In2++Fe3+�In3++Fe2+ e-+Fe3+�Fe2+ (3) [100]) illustrates a rather well developed net with one planar and two corrugated layers are joined at each node. This is not There are two kinds of trivalent iron ions and a small fraction always the case, as shown by the HREM image in Fig. 4(a), of divalent in the compounds studied.The squared eVective where both crossing and in-phase patterns are seen. There are Bohr magneton number per iron atom is shown in Fig. 9–11 also cases where corrugated defects turn before reaching the for n=98, 48 and 23 respectively. The curves marked A and planar defect. The general appearance of our images is slightly B are from the samples with x=0.0 and 1.0, respectively.The diVerent to that reported by Uchida et al. and by Bando and results of the analytical TEM show that iron is located in both coworkers.19–21 In the (2910) projection they observed sinus- types of defects when no indium is present (x=0.0). For x= oidal contrast waves, in phase with each other and thus not 1.0 the iron has a preference for the corrugated layers.A forming a net. Uchida et al. also report the period lengths of simple estimation of meff2 for iron in the planar defects can be the waves for two compositions [InFeO3(ZnO)6 and calculated from InFeO3(ZnO)13] in the (2910) and the (1190) projection. They meff2 (C)=2meff2 (x=0)-meff2 (x=1) (4) found the length shorter in the latter, which is consistent with our observations that the projected angle between the two which is shown in the curves marked C.At low temperatures types of defects is larger in the latter projection. The ratio the calculated value of the iron meff2 value in the planar defects between the lengths in the two projections is, however, not in becomes negative for two of the samples (n=23 and 48). This accordance with our findings, and is not the same for the two indicates of course that the model represented by eqn.(4) is compositions. rather crude, but it can be used for a qualitative discussion. The low meff2 values, and the appearance of the curves in Analytical TEM Fig. 9–11, clearly indicate an antiferromagnetic interaction between Fe3+ ions in both types of defect layers. This antiferro- Spot analysis using EDX was performed on a crystallite magnetic interaction is more pronounced in the planar defects.oriented along [100], found in the InFeO3(ZnO)48 sample. Wea hexagonal layer of edge-sharing FeO6 octahedra there are made point analyses of the planar defects, the bouncing defects six nearest neighbours. Only four of these yield favourable and the intervening triangular ZnO parts.The beam size used antiferromagnetic interactions in an ordered structure, and the was 10 A° . Although the low signal-to-noise ratio was unsatisother two contacts will be unfavourable. In a hexagonal lattice factory, the analysis gave a clear indication: indium prefers there is no preference for the orientation of these interactions, the planar defects while iron seems to be enriched in the but if the hexagonal symmetry is distorted this degeneracy is bouncing defects in this compound.The result is in agreement broken. The corrugated cardboard structure with orthorhom- with the report of Bando and coworkers.19 To obtain a better bic symmetry may at least to some extent stabilise and orient signal-to-noise ratio we used EELS to perform elemental an antiferromagnetic structure by giving a stronger interaction.mapping. Two samples were investigated: (i) Fe2 O3(ZnO)48 The sharp bend of the C-curves at low temperatures may be and (ii ) InFeO3(ZnO)48. The results confirmed the impression an indication of a transition to a state where the magnetic from the EDS analysis. A crystallite in the Fe2O3(ZnO)48 structure of the planar defect is lined up with the corrugated sample was oriented along [1190], of which an image using the layer structure.zero beam is shown in Fig. 7(a). Two images using electrons that have lost energy due the K-absorption edges of iron and zinc are shown in Fig. 7(b) and (c). The iron map clearly Discussion shows that iron is enriched in the defects, and a corresponding dark contrast due to depletion of zinc in the planar defects is The structure of ZnO, wurtzite, can be described as a hexagonal close packing (hcp) of oxygen atoms, containing two diVerent seen in the zinc map.To what extent there is iron in the ZnO J. Mater. Chem., 1998, 8(11), 2465–2473 2469Fig. 7 Elemental distribution map, using EELS, of a crystallite in the Fe2O3(ZnO)48 sample oriented along [1190] to give the best contrast of the defects.(a) An image using the zero beam. The resolution is rather low due to the limitations of the EELS system, but the defects are clearly resolved. (b), (c) Images using electrons that have lost energy due to the (b) Fe-K absorption edges and (c) Zn-K absorption edges. The light contrast in the iron map clearly shows that iron is enriched in the defects, while a corresponding dark contrast due to deficiency is seen in the zinc map. types of tetrahedral voids and one octahedral.The zinc atoms atoms. The insertion of the extra close-packed layer is such that a portion of the oxygen layers are cubic close packed, occupy only one type of the tetrahedral voids. The empty and the occupied tetrahedra share a basal plane, so as to form a and the voids with octahedral coordination are filled with trivalent ions.The formation of such a ccp slab shifts the zinc trigonal bipyramid (in contrast to ccp). All the filled tetrahedra point their apices in the same direction, resulting in a polar atoms in the ZnO layers on opposite sides through a/2Ó3 A° (ca. 0.9 A° ) relative to each other.The indium and/or iron structure. By shifting the zinc atoms through the equatorial plane of the bipyramids, the structure is mirrored and the atoms are situated in octahedral cavities in these ccp layers. This model is in agreement with results of single-crystal studies polarity is reversed. The (001)-plane defects in the doped structures are caused of LuFeO3(ZnO)n (n=1, 4, 5 and 6) reported by Isobe et al.23 The directions of the shift of the zinc atoms on opposite sides by insertion of an extra close-packed oxygen layer plus a reversal of the polarity of ZnO to one side.The result is that of the ccp layers will depend on the number of ‘ZnO’ planes, m=n+1, between the defect planes, as observed in the HREM the apices of the ZnO4 tetrahedra from each side of the (001) defects point towards each other without sharing oxygen images.In the ordered structures an even number results in a Fig. 8 Elemental distribution mapping, using EELS, of a crystallite oriented along [1190] to give the best contrast of the defects found in a sample with nominal composition InFeO3(ZnO)48. (a) Zero-beam image. Images using electrons that have lost energy due to the (b) In-K absorption edges and (c) Fe-K absorption edges and (d) the Zn-K absorption edges.The orientation of the crystallite is such that the bouncing defects are inclined to the beam. The image shows that indium prefers the planar defects, which are avoided by the zinc and iron atoms. 2470 J. Mater. Chem., 1998, 8(11), 2465–2473Fig. 9 The squared eVective Bohr magneton number per iron ion for Fig. 11 The squared eVective Bohr magneton number per iron ion for n=98. A, The value for the samples x=0.0 with iron in both types of n=23. A, The value for the samples x=0.0 with iron in both types of layer defects. B, The contribution from samples x=1.0 with iron layer defects. B, The contribution from samples x=1.0 with iron atoms only in the bouncing defect layers, since there is indium in the atoms only in the bouncing defect layers, since there is indium in the planar defects.C, The magnetic moment from iron in the planar planar defects. C, The magnetic moment from iron in the planar defects is calculated from eqn. (4) in the text. defects is calculated from eqn. (4) in the text. of these compounds revealed no preference of iron for the layer of trigonal prisms halfway between the ccp layers.(One would expect the zinc atoms to avoid the trigonal bipyramidal positions, as this co-ordination is very unusual for zinc, while there are some compounds known with iron in trigonal bipyramids, see e.g. ref. 28. The lutetium atoms were found in the ccp layers, as expected. As already mentioned, Uchida et al.21 observed the wave structures discussed above when investigating InFeO3(ZnO)n (n6), but they did not present any structural explanation for the phenomenon.If similar waves also occur in the structure of the lutetium compound with n=6, which was determined by single crystal diVraction, it could explain why no preference for the trigonal bipyramids was found for iron.The single-crystal studies give the average structure, and this partly disordered modulation therefore cannot be seen. As a consequence, there are not necessarily any perfect (001) layers Fig. 10 The squared eVective Bohr magneton number per iron ion for of trigonal bipyramids half-way between the (001) layers n=48. A, The value for the samples x=0.0 with iron in both types of when n>0.layer defects. B, The contribution from samples x=1.0 with iron In dilute structures, n>6, most of the material is ZnO which atoms only in the bouncing defect layers, since there is indium in the planar defects. C, The magnetic moment from iron in the planar naturally relaxes to the wurtzite structure. Therefore the defects is calculated from eqn. (4) in the text.reversal of polarity must be confined to rather narrow ranges, less than the width of the slabs between the planar defect layers. The corrugated layers may contain such a transition. hexagonal space group, P6/3mcm, while odd numbers yield the rhombohedral symmetry R3m, see Appendix. We have It is diYcult, however, to explain why such a domain boundary should be locked to certain crystallographic planes (114) or simulated HREM images of the ccp layers, using the coordinates from the ordered structures with n=5 and 6.These show (1149) if the transition is gradual. Moreover, a gradual transition does not fit in with the fact that the electric charges are good agreement with our observed HREM images and also with published calculated and observed images.21 The diVer- local.The charge required for polarity reversal is provided by proper localisation of trivalent iron. This idea is supported by ence between the simulated images having indium or iron in the ccp layers is rather small. the observation that the iron is enriched in the bouncing defects. It is very reasonable to anticipate that the these defects As mentioned above, the ZnO4 tetrahedra are always oriented with their apices towards the ccp layers.If the polarity cause the shift in polarity in ZnO necessary to fit the ccp layers. The small shift in contrast in the HREM images when were to be reversed by letting three oxygen atoms of the ZnO4 tetrahedra be shared by the octahedra, then the metal ions crossing these bouncing defect layers corroborates this idea, see Fig. 3. Since these structures are disordered, detailed would come to close to each other. The polar axis of the ZnO slabs thus must change direction somewhere between the ccp information cannot be obtained by using X-ray diVraction techniques. Still, there is some information that can be used layers. The single-crystal structure studies23 of the ordered lutetium-containing compounds [LuFeO3(ZnO)n] with small n to construct a structural model for the defects: (i) There is a diVerence in polarity between the ZnO slabs values, n6, show that this change of polarity is accommodated by a gradual shift of the metal atoms from one tetra- on opposite sides of the ccp layers. Consequently there has to be a reversal of polarity somewhere between these layers. hedron to the neighbouring empty one.As a result the metal atoms have trigonal bipyramidal coordination halfway (ii) There must be reasonable charge balances, interatomic distances and coordination spheres around the atoms. Fe3+ between the ccp layers. The detailed crystal structure analysis J. Mater. Chem., 1998, 8(11), 2465–2473 2471second, with octahedra, for higher n-values. The most important argument for the octahedral model is that there are rather few oxides known with iron in trigonal bipyramidal coordination except the lutetium compounds.23 One such compound is InFeO3.28 To finally settle which model is correct, one of our suggestions or some other, additional information is needed from, for example EXAFS and Mo�ssbauer studies.Appendix The common chemical formula for these types of compounds is: MM¾O3(ZnO)n where M is the metal residing in the plane defect layers (marked below with dots in the layer sequences).In the ordered compounds there are then n+3 close packed oxygen layers within each subperiod (distance between the plane defect layers). If n is an integer there are two cases: Fig. 12 A structural model showing the bouncing defects as formed (i) n is odd and the sequence of oxygen layers is: by trigonal bipyramids viewed along [1190].| subperiod | •(A B)k • (C A)k • (B C)k • (A B)k • (C A)k • can occupy tetrahedra, trigonal bipyramids and octahedra, whereas Zn2+ prefers tetrahedra. | period | (iii) EDS and EELS investigations clearly show an Here: k=(n+3)/2 and period=3×subperiod; enrichment of indium or iron in the ccp layers and of iron in the bouncing defect layers ((114) and (1149) planes).the symmetry is trigonal; (iv) Magnetic data show a coupling between the iron atoms. (ii) n is even and the sequence of oxygen layers is: They are located close to each other in both types of defects (if not diluted by indium in the (001) defect). | subperiod | (v) The defects are narrow and distinct in the crystallites •(A B)k A• (C B)k C• (A B)k A• (C B)k C• viewed along [1190].In accordance with the list above, two structural models | period | have been constructed to describe the bouncing defect layers. Here: k=(n+2)/2 and period=2×subperiod; In the first model we suggest that the change in polarity occurs via double chains of trigonal bipyramids, with trivalent ions the symmetry is hexagonal.in the equatorial plane. These chains run along [1190] and are linked to form planes parallel to (114) and (1149) in ZnO. This This work originates as a spin-oV from a multi-national structural model viewed along [1190] is shown in Fig. 12. The industrial project concerning high-temperature electrode mate- second model comprises a movement of the cations from the rials.In the Northern countries it was co-ordinated by Michael equatorial positions in the bipyramids to neighbouring octa- Hatcher at ‘Permascand’. Our work was later supported by hedral cavities. The latter model is shown in Fig. 13. The NUTEK (Swedish Board for Industrial and Technical octahedra share edges as in a-PbO2, forming chains along Development) and NFR (Swedish Natural Science Research [1190] in (114) and (1149) planes in ZnO.None of the models Council ). We are in debt to the participants in the early stages includes large movements of the oxygen atoms. The hcp net of this work: Mats Nygren, Jekabs Grins, and the late Thomasz of oxygen atoms is in principle preserved, and only the cations Niklewski.are moved. These two models conform to the list of observations and requirements above and are simple enough to be tangible, but it is not obvious which is the correct one, if any. References One may also consider the possibility that the first, with 1 G.Heiland, E.Mollwo and F.Sto�ckmann, Solid State Phys., 1959, bipyramids, is appropriate for lower n-values and that the 8, 193. 2 R.Wang, A. W. Sleight and D. Cleary, Chem.Mater., 1996, 8, 433. 3 R. Wang, A. W. Sleight, R. Platzer and J. A. Gardner, J. Solid State Chem., 1996, 122, 166. 4 T.Ho�rlin and J. Grins, unpublished results. 5 H. Kasper, Z. Anorg. Allg. Chem. 1967, 349, 113. 6 P. J. Cannard and R. J. D.Tilley, J. Solid State Chem., 1988, 73, 418. 7 M. A. McCoy, R. W. Grimes and W. E. Lee, Philos. Mag. A, 1997, 76, 1187. 8 N. Kimizuka, T. Mohri, Y. Matsui and K. Siratori, J. Solid State Chem., 1988, 74, 98. 9 M. Nakamura, N. Kimizuka and T. Mohri, J. Solid State Chem., 1990, 86, 16. 10 N. Kimizuka, M. Isobe, M. Nakamura and T. Mohri, J. Solid State Chem., 1993, 103, 394. 11 N. Kimizuka and T. Mohri, J. Solid State Chem., 1989, 78, 98. 12 M. Nakamura, N. Kimizuka and T. Mohri, J. Solid State Chem., 1991, 93, 298. 13 K. Siratori and N. Kimizuka, J. Solid State Chem., 1992, 99, 243. 14 M. Nakamura, N. Kimizuka, T. Mohri and M. Isobe, J. Solid State Chem., 1993, 105, 535. Fig. 13 A structural model showing the bouncing defects as formed 15 M. Nakamura, N. Kimizuka, T. Mohri and M. Isobe, J. Alloys Comp., 1993, 192, 105. by a-PbO2 type chains of octahedra viewed along [1190]. 2472 J. Mater. Chem., 1998, 8(11), 2465–247316 T. Tsubota, M. Ohtaki, K. Eguchi and H. Arai, J. Mater. Chem., 23 M. Isobe, N. Kimizuka, M. Nakamura and T. Mohri, Acta Crystallogr., Sect. C, 1994, 50, 332. 1997, 7, 85. 17 H. Ohta, W-S. Seo and K. Koumoto, J. Am. Ceram. Soc., 1996, 24 K. E. Johansson, T. Palm and P.-E. Werner, J.Phys. E, 1980, 13, 1289. 79, 2193. 18 N. Kimizuka, M. Isobe and M. Nakamura, J. Solid State Chem., 25 V. D. Frechette and C. F. Cline, Am. Mineral., 1963, 48, 1381. 26 R. A. Powell, W. E. Spicer and J. C. McMenamin, Phys. Rev. B, 1995, 116, 170. 19 N. Uchida, Y. Bando and N. Kimizuka, 13th Int. Conf. Electron. 1972, 6, 3065. 27 F. E. Mabbs and D. J. Machin, Magnetism and Transition Metal Microsc., Paris, 1994, vol. 2, p. 891. 20 Y. Bando, 13th Int. Conf. Electron. Microsc. Paris, 1994, vol. 1, Complexes, Chapman and Hall, London, 1973. 28 D. M. Giaquinta, W. M. Davis and H.-C. Zur Loye, Acta p. 591. 21 N. Uchida, Y. Bando, M. Nakamura and N. Kimizuka, Crystallogr., Sect. C, 1994, 50, 5. J. Electron Microsc., 1994, 43, 146. 22 E. Olsson, G. Svensson and T. Ho� rlin, Eur. Meet. Electron. Micr. Paper 8/05291C Dublin, 1996. J. Mater. Chem., 1998, 8(11), 2465–2