Mossbauer Studies in the Colloid System P-FeOOH-P-Fe,O, : Structures and Dehydration Mechanism BY ARTHUR T. HowE*-~ Inorganic Chemistry Laboratory, South Parks Road, Oxford OX1 3QR KEVIN J. GALLAGHER Chemistry Department, University College, Swansea SA2 8PP AND Received 6th February, 1974 The Mossbauer spectrum of the recently reported P-iron(ur) oxide has been measured and that of j3-iron(uI) oxide hydroxide re-investigated. The NCel temperature of P-iron(m) oxide lies between 300 and 380 K and the average magnetic field at 4.2 K is 49.5 tesla. Magnetic relaxation effects are observed. The spectrum above the NCel temperature is a broad asymmetric doublet, consisting of peaks having a continuous distribution of velocities, with the most probable quadrupole splitting occurring at 1.0 mm s-l, and the most probable isomer shift, extrapolated to 295 K, occurring at 0.31 mm s-'.While the distribution is within the known range of parameters of 6 or 5 co-ordinated Fe3+ in oxides, a proportion of tetrahedral Fe3+, with a lower isomer shift, could also exist within the broad spectral envelope, and a possible concentration range of from 0 to 40% was estimated. The spectrum of /3-iron(m) oxide hydroxide is consistent with octahedral Fe3+. The high pro- portion of ions on the internal surfaces of the tubular structure results in a non-stoichiometric surface anion excess, and the conventional formula 13-Fe00H has been reformulated as #?-FeOx(OH)3-,,, where xw 0.9. Four models of the anion vacancy distribution in /.?-iron(rrr) oxide are considered within the frame- work of the tubular structure retained during dehydration, and a defect structure is proposed which is consistent with the Mossbauer, X-ray, magnetic and B.E.T.data: all the internal Fe3+ ions are 5 co-ordinated whilst the surface Fe3+ ions are tetrahedrally co-ordinated. It has been reported that a new structural form of iron(r1r) oxide is obtained by vacuum dehydration of B-iron(Irr) oxide hydroxide. The new form, designated as the /3-form, retains all of the structural features of the parent compound. Colloidal particles of p-iron(rrr) oxide hydroxide are composed of tubular subcrystalline straws packed into cigar-shaped bundle^.^-^ The micro-structure is based on the hollandite unit cell (fig. l), which has a small central tunnel. The internal tube walls of the colloid are approximately 2 unit cells thick and define the large tunnels which are approximately 3 unit cells wide.2 A supercell, 5 x 5 unit cells, which describes such a structure is shown in fig.2. (The assignment of ions to the surface sites will be justi- fied later in the paper.) When the colloid is initially formed chloride ions occupy the small tunnels, as represented approximately by the formula FeO -%(OH) +&1, with water in the large tunnels. The structure shown in fig. 2 is obtained by thorough washing and subsequent drying, and is the material usually referred to as washed and dried /3-FeOOH. Upon dehydration of this compound to P-iron(rr1) oxide, of nominal formula P-Fe203, neither the morphology, as shown by electron microscopy, nor the total j- present address : Department of Inorganic and Structural Chemistry, The University of Leeds, Leeds LS2 9JT.22A . T. HOWE AND K . J . GALLAGHER 23 a = 1.048 nm 0 O/O H at z = 0 0 O/O H at z = &+ O F e a t z = O O F e a t z = &+ FIG. 1.-Projection along (001) of' the unit cell of the hollandite skeletal structure, upon which the structure of p-iron(m) oxide hydroxide is based. The small tunnels are shown unoccupied. The A site anions (uncrossed) are co-ordinated to 3 Fe3+ ions in a trigonal pyramid. The B site, bridging anions (crossed), are co-ordinated to 3 Fe3+ ions in a plane. This arrangement, together with the fact that the A sites are probably occupied by OH- ions while the B sites are occupied by 02- ions, inevitably leads to a distorted octahedral arrangement around each Fe3+ ion.FIG. 2.-One unit cell of the idealised super-structure of /3-iron(m) oxide hydroxide, FeI28O1 6(0H)15a, in which the numerous smaller hollandite cells can be recognised. The inset shows how replication of the supercell produces the regularly repeating large tunnels within the structure. B site anions are crossed, and the surface sites S1 and $2 are shaded.24 M o SSB A u E R ST u D I E s o F /3-FeOOH-p-Fe203 surface area, as obtained from B.E.T. measurements, alters significantly. Similarly no change can be observed in selected area electron diffraction, while the X-ray diffraction pattern reveals only a 2 % contraction along the c-axis of the tetragonal, hollandite-like cell, together with minor intensity changes, indicative of a topotactical reaction.It is surprising that removal of 25 % of the anions from such an open lattice does not result in more noticeable structural changes, and we have therefore investigated the Mossbauer spectrum of the new p-iron(m) oxide, together with that of the parent (washed and dried) oxide hydroxide, since the particle dimensions, important in determining the magnetic relaxation effects, were not known in a previous investiga- tion of the parent c~mpound.~ Mossbauer studies have also been reported *-lo for the unwashed oxide hydroxide (misleadingly also referred to as p-FeOOH), which, we point out, has some significantly different Mossbauer parameters from the washed and dried compound. EXPERIMENTAL The washed and dried p-iron(m) oxide hydroxide was prepared by the same method as for the previous B.E.T.and X-ray studies.ll The material thus obtained from slow hydrolysis of a hot ferric chloride solution was washed 12 times with distilled water and vacuum dried at room temperature to remove all water from the inner and outer surfaces. Dehydration in high vacuum at 443 K produced p-iron(n1) oxide which was stable in air at room temperature. This product, however, still retains 0.2 moles of H20 per mole of Fe203, as determined by the weight loss upon transformation at 873 K to the stable form, a-Fe203, and subsequent high- t emperature firing. Thin Mossbauer absorbers containing 3.7 mg cm-2 Fe were used to obtain the high quality spectra in the paramagnetic region so that saturation corrections would not be necessary. These samples were mixed with a tenfold excess of inert boron nitride powder to obviate the possibility of non-random particle orientation.The other spectra were ob- tained from undiluted samples having 15 mg cm-2 Fe. Mossbauer equipment of the type described by Cranshaw l2 was used with a 30 mCi J7Co/Pd source. Isomer shifts are quoted with respect to Fe metal at 295 K as zero. Several samples were heated in vacuum in the BN sample holder of a Ricor Mossbauer furnace regulated to within 1 K by a Eurotherm temperature controller. Allowance for a parabolic baseline caused by the geometric effect was included in the Harwell computer fitting programs. RESULTS AND DISCUSSION MOSSBAUER SPECTRA AND MAGNETIC PROPERTIES The spectrum of the oxide hydroxide is shown in fig.3. At 295 K only a broad paramagnetic doublet is observed. The parameters are given in table 1 and lie within the error range of those previously rep~rted.~ DCszi identified 295 K as the NCel ternperat~re,~ and found that, below this temperature, a magnetically-split 6 peak pattern developed while the paramagnetic doublet, although diminished in intensity, simultaneously persisted down to about 200 K. Our very broad spectrum at 77 K shows the residual effects of such magnetic relaxation behaviour. The most pro- nounced magnetic hyperfine field of 46.0 tesla (1 tesla = 10 kG), when extrapolated to 0 K, gives a limiting value of 47.0 tesla, as compared to 47.5 tesla found by DCszi.’ These values are significantly different from the value of 48.5 tesla at 80 K obtained from unwashed samples of the oxide hydroxide.1° The difference, which probably arises from the change in Fe3+ co-ordination due to the additional protons needed to balance the charge of the C1- in the small tunnels of the unwashed samples, has not been previously noticed, and is to be compared with the similarities in the otherA .T. HOWE AND K. J . GALLAGHER 25 Mossbauer parameters and in the magnetic behaviour. We found that the presence of water in the tunnels, as in washed but undried material, did not alter the spectrum compared to the washed and dried samples. FIG. -10 -5 0 5 10 velocity/mm s-' 3.-Mossbauer spectra of B-iron(nI) oxide hydroxide (more exactly ~-FeO~OH)3-2x where x = 0.91) at 77 and 295 K.The baseline is the same for both spectra. TABLE 1 .-MOSSBAUER PARAMETERS most Drobable isomer shift/ most probable mm r1 at 295 K w.r.t. Fe at 295 I< quadrupole sqlittingl most probable H / mm s- T at 4.2 K p-iron(m) oxide hydroxide 0.37 0.7 47.0 p-iron(m) oxide 0.31 1 .o 49.5 Fig. 4 shows the Mossbauer spectrum of p-iron(III) oxide over a range of tempera- tures from 4.2 to 383 K. The parameters are given in table 1. The most noticeable change from the spectrum of the oxide hydroxide is the increase in the Ndel temp- erature, which now lies above 295 K but less than 383 K. Because of slight variations in the shape of the spectrum from sample to sample a more accurate determination of the NCel point was not warranted. The superimposition of magnetically-split and paramagnetic components in the spectrum at 295 K resembles the behaviour of the oxide hydroxide, although the peaks in the spectrum at 77 K are now much sharper (cf.fig. 3 and 4). At 4.2 K relaxation effects should be absent, and the large half- widths of the peaks, which are up to 4 times the expected single peak value, providing a range of magnetic fields from 48 to 51 tesla about the mean of 49.5 tesla, indicate a range of Fe3+ environments in the lattice. The application of a 0.5 tesla magnetic field did not alter the shape of the spectrum at 295 K, indicating an antiferromagnetic ordering, in agreement with the absence of particle movement in a magnetic field. Our samples, of mean dimensions 60 x 60 x 350 m, were at no stage ground, and the two methods of preparing the Mossbauer samples gave the same Mossbauer spectra.26 MOSSBAUER STUDIES OF P-FeOOH-P-Fe,O, .. h -10 - 5 0 5 I0 velocity/mm s-l FIG. 4.-Mossbauer spectra of ,!3-iron(rn) oxide (more exactly fl-Fe0,(OH)3-2x where x = 1.45) over the temperature range 4.2 to 383 K Since the basic double tunnel structure and antiferromagnetic ordering are com- mon to both the unwashed and washed oxide hydroxide and to the oxide, it is likely that the origins of the magnetic relaxation effects found for the oxide are lsimilar to those proposed for crystals of the other two compounds of comparable size. DCszi ’ attributed the effects in the washed and dried oxide hydroxide to domain magnetisa- tion reversal in the small particles,l in a way resembling superparamagnetic behaviour in ferromagnetic mafe~ia1s.l~ However, Syzdalev l has proposed that similar data for the antiferromagnetic a-FeOOH and ferrimagnetic a-Fe,O, are best interpreted in terms of low-temperature paramagnetic behaviour in particles having less than a critical volume.Voznyuk and Dubinin l6 proposed, along the same lines, the pres- ence of paramagnetic surface layers on the otherwise antiferromagnetic particles of unwashed p-iron(II1) oxide hydroxide. These effects in small antiferromagnetic particles are, however, not yet well understood. DEHYDRATION OF P-IRON(III) OXIDE HYDROXIDE TO P-IRON(III) OXIDE A N D THE PARAMAGNETIC MOSSBAUER SPECTRA The asymmetry just evident in the paramagnetic spectrum of the oxide at 383 K (fig. 4) was further investigated using a low-velocity scan on a sample prepared in situA .T . HOWE AND K . J . GALLAGHER 27 in the Mossbauer furnace from a thin absorber of the oxide hydroxide. Fig. 5(a) shows the spectrum of the oxide hydroxide taken in vacuum at 295 K. The tempera- ture was then slowly raised to 443 K and after the initial vacuum of approximately Torr had been restored the spectrum, now of the oxide, shown in fig. 5(b), was collected. The spectrum was unaltered by continued heating at 443 K. A compar- ison of the two paramagnetic spectra could not be made at the same temperature since the oxide is below its Ntel temperature in the stability range of the oxide hydrox- ide. The oxide sample was heated further to 523 K. Between 383 and 523 K there was no observable change in either the value of the quadrupole splitting or the asymmetry of the peaks, providing evidence that the asymmetry did not originate from magnetic relaxation effects., I I I I I l I l I - 1 0 1 2 velocity/mm s-* FIG. 5.-The change in the paramagnetic Mossbauer spectrum in going from p-iron(rr1) oxide hyd- roxide (a), recorded at 295 K, to &iron(m) oxide (b), brought about by heating in vacuo at 443 K and recorded at this temperature. The horizontal lines show the previously observed ranges of the isomer shifts of tetrahedral (lower velocity) and octahedral (higher velocity) Fe3+ in oxides (a) at 295 K and (b) extrapolated to 443 K. One of the many possible computer fits for the oxide is shown, and this particular one is consistent with the structural model proposed, and has 15% tetrahedral Fe3+ (dashed lines), compared to 17 % in the model.The remaining doublets have isomer shifts at the low end of the octahedral range and are therefore consistent with the 83 % 5 co-ordinate Fe3+ proposed. The asymmetry of the oxide spectrum could not have arisen from preferred particle orientation since the material was diluted with boron nitride. Dilution did not noticeably alter any of the spectra, but was done as a precautionary measure in view of the cigar shape of the crystals. However, in such tubular structures the surface ions could possess an anisotropic recoil-free fraction, or f factor, leading to unequal areas of the quadrupole components. ' 9 * The effect, however, is not evident in the spectrum of the oxide hydroxide, and computer fits to both spectra, to be discussed later, were consistent with components of equal area. Neither the oxide hydroxide nor the oxide showed any evidence of having unusually low overallffactors due to the tunnel or defect structures.Only a small reduction (ca. 10 %) in the total area response at room temperature accompanied the conversion of the oxide hydroxide to the oxide, showing that theffactor of the oxide was only marginally lower than that of the oxide hydroxide. Available evidence indicates that the f factors of the surface ions would still be appreciable. Iron ions located in the large tunnels of zeolite A have anffactor at room temperature of 0.57, corresponding28 to a Debye temperature of 265 K. The conclusion is also supported by calculations for the surface layer of Fe The asymmetry of the oxide spectrum would appear to arise from a range of isomer shifts of the many unresolved components arising from the tubular structure, and in view of the high proportion of anion vacancies the isomer shifts may indicate the presence of 5 or 4 co-ordinate Fe3+.In oxides, the isomer shift decreases with increas- ing proportion of s character in the bonds, the more covalently bonded tetrahedral configuration producing a lower isomer shift than octahedral coordination. Isomer shifts can be validly compared between different oxides provided that the actual temperature of measurement was not appreciably lower than the Debye temperature of the solid. At low temperatures departures from the classical limit of the second order Doppler shift occur and render isomer shift comparisons less reliable.A limit of the second order Doppler shift of -7 x mm s-l K-l is indicated by studies on Fe3+ oxides,21 and applies to typical oxides above approximately 200 K,22 for which the Debye temperatures are 300-400 K.23 The possible low Debye temperature of our tubular materials indicates that our measurements have been made well within the applicability of the high temperature limit and can be validly compared to values from other oxides. For comparison, the velocities of components in the oxide spectrum measured at 443 K were calculated, by use of the above figure, to be 0.10 mm s-1 lower than they would have been at 295 IS. A survey of the numerous oxides studied 24p 2 5 shows that, at 295 K, the isomer shifts of Fe3+ in a state of 6 co-ordination range from 0.31 to 0.41 mm s-l, and those in a state of 4 coordination from 2 5 * 26 0.13 to 0.27 mm s-l.The few systems studied in which Fe3+ is 5 coordinated 27-31 cover a range from 0.22 to 0.33 mm s-l at 295 K. The established octahedral and tetrahedral ranges are drawn as horizontal lines for the oxide hydroxide (at 295 K) and the oxide (at 443 K) in fig. 5. It can be seen that both spectral doublets have a centroid within the range for octahedral Fe3+. That of the oxide is towards the lower end of the octahedral range and is therefore also consistent with 5 fold coordination. Furthermore, the asym- metry of the oxide envelope may indicate the presence of a small proportion of tetra- hedral Fe3+.It is clear, though, that the spectrum could not represent a predomin- ance of tetrahedral Fe3+ in the structure. Working on the basis of only the hollandite unit cell (fig. l), where each anion is coordinated to 3 Fe3+ ions, one might expect the removal of 25 % of the anions in going from FeOOH to Fe203 to result in 75 % tetrahedral Fe3+, with 25 % being 6 coordinated. Such a discrepancy with the observed spectrum may suggest the presence of 5 coordinated Fe3+ or that the features introduced by the supercell (fig. 2), primarily the internal surface, play a crucial role in determining the coordination of Fe3+ in the oxide. Since the coordination of the Fe3+ on the internal surfaces of the parent oxide hydroxide has not been considered previously, we shall, after the next section, establish this for use as a basis for considering the dehydration to the oxide.We shall first, however, estimate more precisely the maximum proportion of tetra- hedral Fe3+ in the oxide consistent with the observed Mossbauer envelope and the known isomer shift ranges for 4 and 6 coordinated Fe3+. M OSSB A u ER s T UD I ES OF P-FeOOH-jI-Fe,O, SPECTRAL ANALYSIS The surface induced distortions of the structure would result in a large number of Fe3+ environments and hence a large number of considerably superimposed Moss- bauer peaks, and a complete resolution of the envelope into these components would not be possible. The maximum proportion of the oxide envelope consistent withA . T. HOWE AND K. J . GALLAGHER 29 tetrahedral Fe3+ was therefore estimated by the following process of trial and error and successive approximation.Initially doublets (e.g. AA’), where the intensities and halfwidths of the lorentzian peaks A and A were constrained to be equal, were fitted to the envelope. A minimum of 4 doublets was required, and by choosing starting parameters covering a wide range of possibilities, several solutions were found with satisfactory l2 values in the range 190 to 230 for 205 degrees of freedom. The simplest solution was of the symmetrical form ABCDD’C’B’A‘. The 4 isomer shifts were all higher than the tetrahedral range and were consistent with 6 or 5 coordinate Fe3+. Tetrahedral components could only be introduced by choosing the doublet pairs in an unsymmetrical fashion. A satisfactory fit with the grouping ABCDC‘D’B’A was found with the doublet CC’ having an isomer shift at 295 K of 0.27 mm s-l, at the top of the tetrahedral range and accounting for 8 % of the total area, while the isomer shift of the doublet DD’ (0.40 mm s-l at 295 K) was at the top of the octahedral range.A satisfactory fit could also be obtained with the grouping ABCDD’C‘A’B’ provided that an additional small doublet with an isomer shift in the 6 or 5 coordinate range was included at the velocities of peaks B and A’. In this case the doublet AA’ had an isomer shift in the tetrahedral range and accounted for 33 % of the area, while the doublet BB’ was still consistent with 6 or 5 coordinate Fe3+. Satisfactory groupings of the type ABCDD’B’C’A’ could not be found.As some of the peaks in the above fits had halfwidths up to 0.50 mm s-l, which is considerably Jarger than the value expected from a single absorption (0.23 mm s-l in the present case), the possibility exists of further subdivisions of these peaks to yield component peaks with a new set of isomer shifts. This was achieved by further constraining halfwidths of groups of doublets to be equal, a procedure which maxi- mised the choice of final doublet assignment since any one peak could be proportioned between two peaks at the same velocity and reallocated to two new doublets. Succes- sive introduction of components into the fits led to a statistically satisfactory 14 peak fit with component halfwidths of 0.23, 0.38 or 0.45 mm s-l. The symmetrical doublet grouping gave 7 doublets with isomer shifts, 6, between 0.30 and 0.36 mm s-l when extrapolated to 295 K, in the 6 or 5 coordinate range, with quadrupole splittings, A, between 1.71 and 0.40 mm s-l.By reassignment into 8 doublets, one new doublet consistent with tetrahedral Fe3+ (6 = 0.26 mm s-l at 295 K, A = 1.16 mm s-l) and another doublet having 6 = 0.35 mm s-1 at 295 K and A = 1.15 mm s-l could be introduced. The area of the tetrahedral doublet could be varied from 0 to 15 %, and the 15 % grouping, which is consistent with the model proposed later (17 % tetrahedral and 83 % 5 coordinate Fe3+) is shown in fig. 5(b). The remaining isomer shifts of between 0.30 and 0.36 mm s-l at 295 K are at the lower end of the octahedral range where 5 coordinate Fe3+ would be expected.Additional tetrahedral components could not be introduced without having isomer shifts higher than the known limit for both tetrahedral (0.27 mm s-l) and octahedral Fe3+ (0.41 mm s-I). In order to assess the effect of a rather unlikely uncertainty in these limits, all possible doublets with isomer shifts of up to 0.30 mm s-1 were assigned to tetrahedral Fe3+, which raised the total tetrahedral proportions to 38 %, with the highest isomer shift in the envelope being 0.42 mm s-l at 295 K. We have therefore taken 40 % as the upper limit for the tetrahedral Fe3+ proportion. Further refine- ments, such as an allowance for possible non-lorentzian behaviour, saturation effects and cosine effects would not be expected to alter the basic conclusion of the analysis. Fits to the spectrum of the oxide hydroxide were consistent with octahedral Fe3+, in accord with the known structure. The range of possible fits to both spectra prevents a unique assignment of the isoiner shifts and quadrupole splittings, and for30 MOSSBAUER STUDIES OF P-FeOOH-P-Fe,O, this reason the parameters derived from the individual fits have not been quoted.The data in table 1, indicating the parameters of the general envelope, should be interpreted in this light. The limiting magnetic hyperfine fields for the two compounds are consistent with the above possible assignments. This parameter is not generally diagnostic of coordination number. THE SURFACE STRUCTURE OF P-IRON(III) OXIDE HYDROXIDE: SURFACE CONTROLLED NON-STOICHIOMETRY I N P-Fe0,(OH)3-2x WHERE XEO.9 In the basic hollandite unit cell (fig.1) all the cations sites are identical, and there is an equal proportion of A and B anion sites. Four double strings containing edge- sharing [FeX,] octahedra run the length of the crystal in the c direction, and contain the A site (non-bridging) anions. The double strings are linked together through corner sharing via the B site (bridging) anions to form the small central tunnel. A recent neutron diffraction study 32 of the unwashed material, containing C1- in the small tunnels, showed that the octahedra were distorted in a manner which retained the unique Fe site, but which allowed for the preferential proton positioning closer to the A site oxygens within the double strings, but on the side of these oxygens facing the two adjacent bridging (B site) oxygens.A similar site preference of the protons for the A site oxygens may exist in the washed and dried p-iron(Irr) oxide hydroxide. Because the crystal diameter is only of the order of 10 times the 5a x 5a supercell dimensions, diffraction techniques would not detect the supercell, and the neutron diffraction study was therefore unable to determine the internal surface structure, as evidenced by the high R factor of 11 %. Inspection of the supercell (fig. 2) shows that the internal surface of the large tunnels, as well as containing A and B anion sites, also has 12 sites coordinated to one Fe3+ (S1 sites) and 12 coordinated to two Fe3+ (S2 sites). These are both shaded in the figure. The absence of anions from the S1 sites would create twelve 5 coordinate Fe3+ ions per supercell, while the absence of site S2 anions would create twelve 4 coordinate Fe3+ ions.Such a significant breakdown of the stable [FeX,] octahedra upon which the structure is based, resulting in a change of coordination of 18 % of the Fe3+, would be unlikely to occur under the mild conditions of room temperature vacuum drying, suggesting that these sites are in fact occupied. However, if the surface anion sites were occupied, there would be an excess of anions in the crystal over and above that represented by the formula FeOOH. Each hollandite cell exposed to the internal surface will be of composition M8X1,, compared to the normal cell of M8X16. The total composition will be M128X268. The form- ula which satisfies the charge balance is Fe12s0116(OH)152, or FeOo.906(OH)1.188. provide definite evidence for the anion excess structure.After a 5 % weight loss during prolonged vacuum drying at room temperature, during which water is removed from the internal and external surfaces, the oxide hydroxide lost a further 12 % water, as a percentage of the initial weight, upon dehydration and final conversion at 873 K to a-Fe203. This value agrees well with the value of 11 -2 % water loss calculated for the OH- excess structure above, but is considerably higher than the value of 7.6 % calculated for the OH- deficient structure corresponding to unoccupied S1 and S2 sites. The general formula which expressed the variability of the 02- to OH- ratio depending on the extent of surface controlled non-stoichiometry is Fe0,(OH)3 - 2x.This formula is preferable to the more conceptual formula for the anion excess oxide hydroxide of Fe01-x(OH)1+2, since it covers the whole range from Fe(OH), to Fe203. Crystalline products with the composition Fe(OH)3 are unknown, so the The results of the previous thermogravimetric studiesA. T . HOWE AND K. J . GALLAGHER 31 oxide hydroxide, in which x = 0.906, is probably as OH- rich as can be obtained in a crystalline phase. Since all the investigations of p-iron(II1) oxide hydroxide indicate the presence of both large as well as small tunnels, it would appear that the stoichio- metric formula of FeOOH does not apply to the normally prepared material, which should be represented as FeO,(OH),-,, with x M 0.9, the exact value depending on the exact dimensions of the large tunnels.Inclusion of the effects of the external surface of the crystals in this study only altered x from 0.906 to 0.898. The anion excess structure, in which all the Fe3+ ions are 6 coordinated, is consist- ent with the small range of isomer shifts found in the Mossbauer spectrum of the oxide hydroxide, and the surface induced distortions would alone be sufficient to account for the range of quadrupole splittings. Possible proton disorder may contribute further to the spread of quadrupole splittings. [The structure would allow an ordered proton arrangement, with protons occupying all the A (non-bridging) sites, as suggested by Gallagher,, together with all the surface sites to give Fe12 8(OH)t2 80: 1 (j(OH);:(OH);;l. THE DEFECT STRUCTURE OF P-IRON(III) OXIDE : P-FeO,(OH),-,, WHERE XM 1.45 Fe00.9(OH)1.2-+FeOo.g+,(OH)l.2-2,+yH,0.Anion vacancies (0-) are created in the retained supercell according to the equation 20H-+H20+02-+ 0-. If complete dehydration to P-Fe203 (supercell formula FelzzOl 92) owurred, 76 anion vacancies must be distributed amongst the 128 A sites, the 116 B sites and the 24 surface anion sites in Fe12801,,(OH)l,2. This figure may be reduced to about 62 anion vacancies due to the small quantity of water equivalents (2 % of the initial weight) still present in the oxide after prolonged evacuation. The reported i.r. peak indicates that these residual protons, like those in the oxide hydroxide, are not hydrogen bonded, but the shift of the stretching mode to even higher frequencies compared to the latter indicates that the protons are more tightly bound to oxygen, which could be produced by OH- bound to Fe3+ in a state of low coordination in the highly defective structure of the oxide. If OH- was present in the structure it would imply that the dehydration process had ceased before the eventual expected product of P-Fe203 had been reached, and we have therefore assumed, in the following, the minimum extent of dehydration corresponding to the residual presence of OH-.The maximum value of x in the formula Fe0,(OH)3-2, is thus calculated to be 1.45, slightly less than the value of 1.5 for Fe203. Four model anion distributions have been considered for the removal of 62 anions. If the surface anions are retained, the average Fe3+ coordination drops from 6 to 4.55 (model I).If the surface anions are removed together with anions from the interior, the average Fe3+ coordination drops to 4.83 (model 11). For each of these two basic possibilities, we have calculated the proportions of Fe3+ in each type of coordination assuming (a) that only 6 and 4 coordination exists and (b) that only 5 and 4 coordina- tion exists. The values are given in table 2. The models show such a wide range in the proportions of the three coordinations that we can confidently discriminate between them on the basis of (1) the Mossbauer results, showing less than 40 % tetrahedral Fe3+, (2) the constancy, to within 2 %, of the unit cell dimensions upon dehydration, and (3) the increased magnetic coupling in the oxide. Removal of water from the oxide hydroxide can be expressed by32 MOSS B AUER s T UDI E s OF /?-FeOOH-P-Fe,O, The Mossbauer results are highly inconsistent with the presence of 73 % tetrahedral Fe3+ [model I(a)] and are also inconsistent with model II(a) with 59 % tetrahedral Fe3+.The results would, however, be in very good agreement with model II(b), with only 17 % tetrahedral Fe3+, the remaining Fe3+ being 5 coordinated. TABLE 2.-POSSIBLE PERCENTAGE DISTRIBUTIONS OF DIFFERENT Fe3+ COORDINATION NUMBERS IN B-IRON(III) OXIDE six five four model I 73 (b)* - 55 45 - no surface anions removed (a) 27 model I1 all surface anions removed (a) 41 - 59 (6)" - 83 17 * shows models consistent with Mossbauer results. The a unit cell dimension depends primarily on the axial distances of the coordina- tion polyhedra, while the c dimension depends primarily on the equatorial distances.Since Fe3+ in a square planar coordination is unknown in oxides, 4 coordinate Fe3+ would distort to the tetrahedral position with a consequent substantial reduction in both the axial and equatorial polyhedral dimensions, and this would result in decreased unit cell dimensions. On the other hand, the removal of an A site anion from an octahedron would, in the absence of site relaxation, leave a square pyramid without altering either the axial or equatorial distances. Relaxation of the former equatorial anions towards the more symmetrical trigonal bipyramidal configuration would only be expected to affect the equatorial distances slightly, predicting only small changes in the unit cell dimensions, as found.The X-ray evidence thus strongly favours model II(b), with a predominance (83 %) of 5 coordinated Fe3+, and only 17 % tetrahedral Fe3+. The magnetic evidence also favours this model since a major alteration of the Fe-0-Fe angles is likely to significantly alter the superexchange couplings. The weight of evidence therefore suggests model II(b) as the most likely structure. Consideration of the supercell shows that the site distribution is quite a stable one for such a structure. Of the 128 Fe3+ ions in the supercell, 24 are at surface sites and 104 at internal sites. Now least disruption to the structure and magnetic interactions will occur if the tetrahedral ions are those at the surface, and 17 % of 128 is 21.8, which corresponds closely to the 24 surface Fe3+.The internal sites are therefore all 5 coordinated (106.2 calculated, as compared to 104 in the supercell). Creation of the internal anion vacancies exlusively on the A sites would leave all the B (bridging) sites occupied to interlock the double strings of ions. Five coordination can then be achieved by the staggered removal of every third A site anion down each c-axis string. Further dehydration would necessitate the energetically unfavourable creation of internal tetrahedral Fe3+ ions, suggesting a reason why the dehydration ceased at this stage when OH- still remained in the lattice. In fig. 5(b) a particular doublet grouping consistent with 17 % tetrahedral Fe3+ is shown for a 16 peak fit.The isomer shift at 295 K is 0.26 mm s-l. The other doublets have isomer shifts at 295 K ranging from 0.30 to 0.36 mm s-l and are consistent with 5 fold coordination. Five coordinate Fe3+, in the form of trigonal bipyramidal coordination, while not common, is well established in a number of oxides.A . T . HOWE AND K. J . GALLAGHER 33 For instance, in FeV04 there are 2 octahedral Fe3+ sites and one trigonal bipyramidal Fe3+ site 29 (having an isomer shift 6 = 0.325 mm s-l and quadrupole splitting A = 1.11 mm s-l at room temperature). In YMn03 the Mn site is trigonal bipyra- midal, and Fe3+ doped into this site retains the configuration 30 [6 = 0.30 mm s-l, A = 2.13 mm s-l, H(0) = 45.5 TI. In the oxygen deficient phase Sr,-,La,Fe0,.5+x/2 a trigonal bipyramidal site has been proposed 28 (6 = 0.33 mm s-', A = 0.33 mm s-', H(0) = 48.8 T), together with tetrahedral and octahedral sites in the structure.It would therefore not seem unreasonable for most of the Fe3+ [as in model II(b)] to have this coordination. The magnitudes of the quadruyole splittings in the oxide, which are larger than those in the oxide hydroxide, arise from the electric field gradients created around the 4 and 5 coordinate Fe3+ by the unsymmetrical anion environment. The range of quadrupole splittings would result from both the surface-induced distortions and the distribution of tetrahedral Fe3+ and residual OH-, which are probably randomly distributed over the A sites. The presence of the initially occupied surface sites and the comparative freedom of movement of the internal anions due to the tubular structure has allowed the structure to sustain such a large degree of ionic removal and rearrangement without collapse.The analysis has provided a valuable insight into the structure, which, because of its tubular form and range of surface-induced distortions would not be readily amenable to an accurate positional analysis by either X-ray or neutron diffrac- tion. Financial support from the S. R.C. through the Physico-Chemical Measurements One of us (A. T. H.) is grateful to the Royal Unit, Harwell, is acknowledged. Commission for the Exhibition of 1851 for a Fellowship. H. Braun and K. J. Gallagher, Nature Phys. Sci., 1972, 240, 13. K. J. Gallagher, Nature, 1970, 226, 1225. W. Feitnecht in 4th Int.Symp. Reactivity oj*Solids, ed. J. H. de Boer (Elsevier, Amsterdam, 1960), p. 583. A. L. MacKay, Miner. Mag., 1960, 32, 545. A. L. MacKay, Miner. Mag., 1962, 33, 270. J. H. L. Watson, R. R. Cardell and W. Heller, J. Phys. Chem., 1962, 66, 1757. 22, 617. M. J. Rossiter and E. A .M. Hodgson, J. Inorg. Nuclear Chem., 1965, 27, 63. 9T. Takada, M. Kiyama, Y . Bando, T. Nakamura, M. Shiga, T. Shinjo, N. Yamamoto, Y. Endoh and H. Takaki, J. Phys. SOC. Japan, 1964, 19, 1744; N. Yamamoto, T. Shinjo, M. Kiyama, Y . Bando and T. Takada, J. Phys. SOC. Japan, 1968,25,1267. l o P. 0. Voznyuk and V. N. Dubinin, Fiz. Tverd. Tela, 1973, 15, 1897 (Sou. Phys. Solid State, 1973, 15, 1265). l 1 K. J. Gallagher and D. N. Phillips, Chimia, 1969, 23, 465. T. E. Cranshaw, Nuclear Znstr. Mehthods, 1964, 30, 101. l 3 A. A. van der Giessen, Philips Res. Rep. Suppl., 1968, 12,40. l4 D. W. Collins, J. T. Dehn and L. N. Mulay, in Mussbauer Methodology, ed. I. J. Gruverman l 5 I. P. Syzdalev, Fiz. Tverd. Tela, 1970, 12, 988 (Sou. Phys. Solid State, 1970, 12, 775). l6 P. 0. Voznyuk and V. N. Dubinin, Fiz. Tuerd. Tela, 1973, 15, 1897 (Sou. Phys. Solid State, l7 P. A. Flinn, S . L. Ruby and W. L. Kehl, Science, 1964, 143, 1434. l8 A. A. Maradudin in Solid State Physics, ed. F. Seitz and D. Turnbull (Academic Press, New l9 L. Kumer, H. Posch and J. Kaltseis, Phys. Letters A , 1972, 40, 59. 2o L. Dobrzynski and P. Masri, J. Phys. Chem. Solids, 1972, 33, 1603. 21 J. J. van Loef, Physica, 1966, 32, 2102. ' I. Dkzsi, L. Keszthelyi, D. Kulgawczuk, B. Molnar and N. A. Eissa, Phys. Stat. Solid., 1967, (Plenum, New York, 1967), p. 103. 1973, 15, 1265). York, 1966), vol. 19, p. 1. 1-234 M o SSB A UER STUD IE s OF ~-FeOOH-/3-FFe,O3 22 F. van der Woude, Phys. Stat. Solid., 1966, 17, 417. 23 G. A. Sawatzky, F. van der Woude and A. H. Morrish, Phys. Rev., 1969, 183, 383. 24 N. N. Greenwood and T. C. Gibb, Miissbauev Spectroscopy (Chapman and Hall, London, 2 5 J. Danon in Chemical Applications of Mossbauer Spectroscopy, ed. V. I. Goldanskii and R. H. 26 C. Le Corre, A. Malve, C. Gleitzer and J. Foct, Compt. rend. C, 1972, 274, 466. 27 J. G. Rensen and J. S. van Wieringen, Solid State Comm., 1969, 7, 1139. 28 H. Yamamura and R. Kiriyama, Bull. Chem. Soc. Japan, 1972,45,2702. 29 B. Robertson and E. Kostiner, J. Solid State Chern., 1972, 4, 29. 30 J. Chappert, J. Physique, 1967, 28, 81. 31 F. P. Glasser, F. W. D. Woodhams, R. E. Meads and W. G. Parker, J. Solid State Chem., 32 A. Szytula, M. Balanda and Z . Dimitrijevic, Phys. Stat. Solid. (a), 1970, 3, 1033. 1971), Ch. 5. Herber (Academic Press, New York, 1968). 1972, 5, 255.