American Mineralogist, Volume 69, pages 937-942, 1984

X-ray powder diffraction profile refinementof synthetic hercynite

Rooerucr J. Hrt-rCSIRO Division of Mineral Chemistry P.O. Box 124. Port Melbourne Victoria 3207 Australia
Abstract A full profile X-ray powder diffraction structure refinement has been completed on a sampleof synthetichercynite,FeAl2Oa, CuKa step-scan usinggraphitemonochromatized data (20^u* : 140") and a flexible profile shape of the pseudo-Voigt type. The most satisfactory was achievedat R*o = 0.0781 usingfully ionized convergence and Rs : 0.0125 atomic scattering factors and a peak shape parameter of the form e = 0.233(17) + 0.0055(3)20: Lorentzian the individual peak shapesvary from 33.7Vo Lorentzian to 99.8Vo t characteras 20 increases. The derivedstructuralparameters at20"C areia = 8.15579(6)4, (inversion) = 0.163(5),r(0) : 0.2633(2),B : 0.56(3), 0.44(3) and 1.07(5)42for the tetrahedral(Fe), octahedral(Al) and oxygen atoms, respectively.Calculatedvaluesfor a andx basedon the ionic modeland i : 0.163are in good agreement with the refinedvalues.

Introduction

The crystalchemistryof the spinelhercynite,FeAl2Oa, is of considerableinterest to mineralogistsand materials scientists becauseof its importance in metamorphic petrology (Osborneet al., l98l; Rumble, 1976; Turnock and Eugster, 1962; Palache et al., 1944, p.687), ceramics (Masonand Bowen, l98l; Meyerset al., 1980; Baldwin, 1955) (Zhukovskayaet al., 1980; and device applications Dehe et al.,1975; Hoffmann and Fischer, 1956). However, the detailed crystal structure of this phase(in particular its cation distribution) remains uncertain, primarily becausepure hercynite is rarely, if ever, encounteredin nature (Turnock and Eugster, 1962;Palacheet aI., 1944, p. 692) and because synthetic hercynites in the FeOFe2O3-Al2O3 system often contain significant levels of non-stoichiometry arising from valence state variations, interstitial atoms and oxygen atom vacancies(Mason and Bowen, l98l; Halloran and Bowen, 1980). Indeed, many of the natural "hercynites" documentedin the literature are more properly classifiedas membersof the MgAl2O2FeAlzO+(pleonaste)or Fe3Oa-FeAl2Oa solid solution series. A compound AB2X4 with the spinel structure crystallizes in spacegroup Fd3m with the A cation in site 8d at (l/8 l/E l/8), the B cation in site l6d aI (ll2 ll2 ll2), and the anion in site 32e at (xxx), when the origin is at l6d (Bragg,1915; Hafner, 1960;Blass, 1964; Hill et al., 1979). Experimental The structure then requires only three parameters (excluding thermal vibration coefficients) to describe its The sampleof FeAl2Oaused in the present study was atomic arrangementcompletely: the unit cell size, a, the synthesized for 4 daysusing in an evacuated silicatubeat 1050"C and yAl2O3followedby mixtureof Fe, Fe2O3 anion positional coordinate, x, and the cation ordering, or a stoichiometric 0003-004x/84/09r0-0937$02.00 %7

inversionparameter,i. For i : 0 and l, respectively,the spinel has the so-callednormal, A[Bz]&, and inverse, B[AB]X4, cation distributions,where the ions indicated in parenthesisare in octahedral coordination by X, and the remainingions (including X) are in tetrahedral coordination. Intermediate cation distributions may be representedas (Ar-i B)[Ai Bz-i]Xn. as FeAlzOr which havebeendescribed For thosephases in the literature the unit cell dimensions have been to lie in the range8.08-8.164 (Segnit, 1984). determined This rather large range of values suggeststhat not all of the samplesare stoichiometric and/or they contain some Fe3+ or other cation. The oxygen coordinatehas been determined twice, yielding values of "about" 0.265 (Barth and Posnjak, 1932)and 0.2635(l) (Roth, 1964). Estimates of the distribution of Fe2* over the octahedral and tetrahedral sites obtained from diffraction and Mtissbauer measurementsvary between i : 0.0 and 0.25 (Osborneet al., 1981;Dickson and Smith; 1976,Yagnik Roth, 1964). and Mathur, 1968;Barth and Posnjak,1932; The present study of synthetic hercynite was initiated (1) in order to accurately characterize the crystal structure of this important phase, and (2) as part of a broader investigation into the application of full-profile Rietveldtype powder diffraction structure refinementmethods and detailed peak-shapemodelling in the case of two-wavelength X-ray data.

938

HILL: REFINEMENT OF SYNTHETIC HERCYNITE

Table l. Crystal structurerefinementresultsfor hercynite

Pare[etet X-ray proflle (predent study) Neutron ltrtegrated (Roth,1954)

9 (A) !(tet,) (A2) g(oct') {42) !(o) (A2)

e.rsszllo;*
0.s6(3) 0 . 4 4 (3 ) r . 0 7( 5 ) 0,2633(2)
I

E,1t2

0 . 9 9 (s )

r(o)
Inverslon,

0.2535(1) 0.1t4(r0)

0 . r 5 3 (5 ) 0.00r195(6) 0.0r54(1r) -0.0r4r(2t) 0.0296(8) - 0 . 0 8 5 75 () 0.96(5) 0.233(r7) 0.0055(3) t207

SceIe

ut
v IT b zeto

Aay-ettyrr -tt

No. of ob6ervatlono

14

R -p

0.0568 0.0781 0,0125

ln parentheels to the lft. are the ead.s ln teru

R {P

:}
P

0.010
of the 1a6t Eltniflcent

N-b"a" flgure

t Coefflclent ln the expregton: et al, 195E). **Ae deftned by E.rvcld (1969).

nml2

- Uton2o + vtene + t{ (Cegltort

ttcqefflclent

ln the exptesslonr

p.eudo-vol8t

e '

eo + I.2O. E! t. rouShty fim tntegratdd

t Agreenent lndlces deflned by yostrg.nd prtnce (19E2)i equlvelent to trlce th colvqntlonel I vllre obtained peak structure reflneoentE.

annealing at E50"C for 5 days (pers.comm., V.J. Wall, Dept. of Earth Sciences, Monash University).The maximum amountof Fe3+is estimated from Mossbauer measurements tobe about4Vo of the total Pe content (personalcommunicationfrom W.A. Dollase,Dept. of Earth Sciences, ucLA to V.J. Wall). The hercynite powder was back-pressedinto a standard aluminium holder and mounted in a Philips PW1050 diffractometer equippedwith a PWl7l0 automaticstep-scanning systemand a diEracted-beam curved graphitemonochromator. Intensitymeasurements weremadeat 20'C at intervalsof0.04.20 over the range 17-140'20using CuKa radiation and a step countingtime of 5 s. The X-ray tube was operatedat 45 kV and 26 mA, with l'divergence and receivingslits. Theseconditions allowed the collection of profile data for a total of 72 Bragg reflections with background counts in the range 150-1200 and a maximumpeakintensityof 45,000 counts.Dead-time corrections were appliedautomatically during data collection.

Structure refinement The least-squares structurerefinements were undertaken with the full-profile, Rietveld-type,program DBW3.2 (Rietveld, 1969;Wiles and Young, l98l) locally modified to include the applicationof a Z&variable profile shape function of the form:

where s i. ,n" r.""ii]':::":t^n character or the peaksin the pseudo-Voigtfunction (Hindelehand Johnson, 1978). This programcan acceptX-ray data obtained from a conventional diffractometer since it allows the (i.e., ar and simultaneous refinementof two wavelengths az) if their intensity ratio is known (2:l in the present case).The weight assignedto the intensity observedat each step i in the pattern is 4 = llYioand the function procedureis I/,(Iio- Yi")2, minimizedin the least-squares where Yloand I'i" are the observedand calculated intensities, respectively. The refinedquantitieswere a scalefactor, the pseudoVoigt profile shape parameters 0 (and er in the final refinement cycles),a20 zeroparameter,a peak full width at half maximum(F'WHM) function (of the form: FWHM2 : IJ tan20+ V tan? + W; Caglioti et al. 1958) calculated for 3.5 half-widthson either side of the peak position, a peak asymmetry parameter(Rietveld, 1969)for reflections with d > 2.24, the unit cell constant,a, the x coordinateof the oxygen atom, x(O), isotropic thermal parameters, B, for all three atoms, and a site occupancy parameter, i, arrangedto modelcooperative cation disorder over both the octahedraland tetrahedralpositions. The backgroundwas evaluated by linear interpolation between 18 points selectedat roughly 7o intervals in 20 over the entire pattern.Note that when the background is estimated by interpolation, rather than by refinement, only thosedata points within 3.5Vo FWHM on either side of the peak positions are included in the profile refinement. The number of "observations" is, therefore, reducedfrom 3076to 1207. factors (includingthe Scattering anomalous dispersion coefficients) for Fe, Fe2*, Al, Al3*, O and O'- were obtainedfrom Ibers and Hamilton (1974) and that of 02- was obtainedfrom Hovestreydt (1983). An experimentally determinedvalue of 0.91was usedfor polarizationcorrection. the monochromator The refinements were initiated using a : 8.159A,x = 0 . 2 6 5e , s : 0 . 5 ,e r : 0 . 0 ,t : 0 , A ( t e t . )= B ( o c t . )= 0 . 5 4 2 , B(O) : 1.0A2,and a scalefactor and half-width parameters estimatedby inspectionof the observeddiffraction pattern. There was no evidence for the presence of preferredorientation although several very small peaks phasewere observed.Refinement due to a contaminating was continueduntil the parametershifts in the last cycle were all less than one quarter of their associated esd's. Combinations of scattering factors corresponding to neutral, half-ionizedand fully-ionizedatoms were carried to convergence separately.The pseudo-Voigtes parameter was only releasedfor the last few cycles of full-matrix refinement.The complete set of results for the best-fit model are given in Table I where they may be compared to the integrated intensity neutron powder diffraction resultsof Roth (1964). The raw digital intensitydata have been deposited,rbut a plot of the observed,calculated and diference profilesfor the final refinementis given in Figure l. A list of observed and calculatedintegrated

HILL:

REFINEMENT OF SYNTHETIC HERCYNITE

939

SnT|T|CrrcntE
aa.5 rt0!

- FELU4 - FILLYlO{Eo nlt6

o
N T

''

([c) no DTn

Fig. 1. Observed,calculatedand diference X-ray powder diffraction profilesfor synthetic hercynite. The observeddata are indicatedby plus signsand the calculated profile by the continuousline overlyingthem. The short vertical lines below the pattern represent the positionsof all possibleBraggreflections and the lower curve showsthe value of (yi., - yi.) at each step.

peak intensities,together with their d-spacings,for all peakswith intensities greaterthan lVoof the maximumis given in Table 2. With the profile and structural parametersset equal to the values obtainedfrom the best-fit model the 12 most intense Bragg peaks were considered individually in refinementsof the scale factor, 20 zero, half-width W term, and pseudo-Voigts0parameter. The resultantpeak

widths and e0parameters are plotted as a function of 20 in Figure 2. Discussion Structural parameters

Refinement of the model with fully-ionizedatom scattering factors produced a slightly better fit between the by Ro and observedand calculatedprofiles,as measured I To receivea copy of this data, order Document AM-84-248 R.o, than the neutral and half-ionized models.However, from the BusinessOffice, MineralogicalSociety of America, includingthe relatively sensitive none of the parameters, 2000 Florida Avenue, N.W., Washington,D.C. 20009.Please .8, and inversion paramethermal vibration parameters, remit $5.00in advancefor the microfiche. ter, i, were affectedby more than 1.5 combinedesdls by the differentform factor models. Table2. Observed and calculated peak intensities integrated and The mean isotropic thermal vibration parameter of d-spacings for hercynitereflections with intensities greaterthan 0.8242 is in reasonableagreement(by X-ray powder lVo of the maximumcalculatedvalue with the singlevalue 0.99(5)deterdiffraction standards) q (a ) mined for hercynite by Roth (1964)(Table l), with the values 0.56(4) and 0.3(1)42 determined for synthetic ll I 4.709
I f

I.1

I.I

o2
I1 00 22 l1 04 o2 33 22 44 26 13 35 00 06 22
E I 7 5 6

2.4E4 2.459 2.O39 1.871 I.665

57.4 102.6 17.2 t6.E 41.4 49.1 10.0

5 7. 9 I00.0 L 7. 2

17.0 41.5 49.9

o.5 ; t

:t 4

1.570 1.442 1,290 1 .2 4 1 l. 230 1 .I 7 7 1.090 1.062 1.019 o.96L2 0.9418 0.9355

g 08

E
o
o

.I ''li/ l/l
-t-tl-r-l-{l/r

o.4 g

l-zt
0.6

9.6

o.3 E 02 :
a

f o.o
8.0
7.8 14,9

oz
60 80 ?9(.1 loo 120 140

o.t t c oo
o 20 40

6.2 4.t r 1 ,3 1.3 2.1 I.6 12.6 22,2

6.I

))
26 04 46 I3 44

l )

10.9

E o 9 I

0 . 9I t 9 0. E694 0.E550 o,8324

1.7

22.O

Fig. 2. Variation of the individual peak pseudo-Voigt (solid circles)and peak full widths at half maximum parameters (crosses)with 20 for the 12 most intense peaks in the X-ray diffraction pattern of hercynite. The peak width and shape of the entire patternare relationships determinedby refinement lines. shownas continuous

940

HILL: REFINEMENT Table 3. Interatomic distancesand anglesfor hercynite

AO4tetrahedron: A - Fe0.84A1O.15 A-o o-A-o o-0 I'41 tx6l t'61
1.9s4(2) A 109.47(10)" 3.190(3) A

OF SYNTHETIC HERCYNITE

806 octahedron: 8 - A l t , g 4 F e O , t 6 B-0 o-o o'-0' o-E-o o'-B-o' Ix51 [x5l t "51 t'61 tr5l
1.937(2) A 2.577(!) 2.892(I) 83,41(7)" e6.59(7)" A.hared ed8e

than the unshared ones, and the valence angles are distorted 6.6ofrom their ideal values. The tetrahedron retails its 43m symmetrywith a bond lengthexpanded by 0. l9A over the value corresponding to idealclosepacking of the anions (x : 0.25), while the octahedronhas 3m symmetry with bond lengths decreasedby 0.104 from their ideal values. The bond lengthsare consistentwith values calculated using the observed site populations (Table l) and the efective ionic radii of O2-, Fe2* and Al3* (O'Neill and Navrotsky, 1983). Parameter esd's The subject of the accuracy of parameter esd's determined in profile structure refinements has been extensively debatedin the recent literature (Scott, 1983;Cooper,1982:'Cooper et al., l98l; Hewat and Sabine, l98l; Prince, l98l; Bacon and Lisher, l9E0; Prince, 1980; Sakataand Cooper, 1979). The generalconclusionis that profile refinement esd's may be underestimatedby a factor of 2 or so compared to the results obtained from integratedintensity refinements when the model, including both the structure and the profile parts, is not correct. However, in cases, like the present study, where the difraction peak profile shapeis itself a refineableparameter, it may be expectedthat the esd's of the structural parameters more closely reflect the true distribution about their mean values. Indeed, while the major contributions to the least-squares residual do lie under the peaks in Figure I, the satisfactory value for R*o indicates that the peak shapemodel is reasonable, by X-ray standards,in the presentcase. Comparison of the profile refinement results obtained in the present X-ray study with those obtained in the integrated intensity neutron data refinement of Roth (1964) (Table l) demonstratesthat the values of the parameters themselves are unaffected by the different refinementmethods.Moreover, the esd's determinedfor all parameterscommon to both studiesare of very similar magnitude,lending support to the estimatesobtainedin the profile refinement. Peak shape and half-width The power of the pseudo-Voigt profilefunction is that it allows a continuous variation in shapefrom pure Lorentzian or Cauchy(e = 1.0)throughto a pure Gaussian (e = 0.0).Indeed,Young and Wiles (1982) haveconcluded that the pseudo-Voigtand PearsonVII (Hall et al., 1977)peak shape functions are consistently the best options for Rietveld refinements of both X-ray and neutron diflraction data. In the presentstudy the best fit betweenthe observed and calculatedpeak shapeswas achievedby varying the shapelinearly as a function of 20 (Fig. 2). At the low 20 end of the diffraction pattern the peaks have about 34Vo Lorentzian character, whereas at the high 20 end they approximate to a pure Lorentzian form. This indicates that for a given FWHM, a greater proportion of the integratedintensity lies in the peak tails as the diffraction

A undhared edSe

oAB3 polyhedroo o-A o-8 A-B B-B A-0-B B-o-B t'Il t'31 t'31 t'31 tr3l I'31 1 . 9 5 4 ( 2 )A 1.937(2)A 3 . 3 8 1 ( 0 )A 2.884(0) A scro36 ehered edge r2o.72(8)' e6.23(7r"

Shortest A-A - 3.532 A

MgAl2Oa and ZnAl2Oa, respectively, from neutronpowder diffraction data (Fischer, 1967), and with the value 1.0(2)42determined for synthetic Fe3Oafrom neutron singlecrystal data (Hamilton, 1958). The inversionparametervalue of0.l63(5), corresponding to the cation distribution (FessaAls. 16)lv[Fe0 r6All ro]utOo, lies in the middleof the rangei = 0.0-0.25determined for hercynite in the past, and is very close to the value 0.154(10) obtained by Roth (1964)(Table l). The large range exhibited by the previous X-ray and Mcissbauer determinationsis probably a reflection of the difficulty in estimating the areas of weak shoulder peaks in the Mcissbauer spectraand/orproblemsassociated with nonstoichiometryand phaseimpurity in the earlier samples. The refinedpositionalparameter for oxygen,0.2633(2), is very closeto the value obtainedby Roth (1964), and the unit cell size, 8.15579(6)A, falls ar the upper end of the rangeofliterature valuesclusteringaround8.16A(Segnit, 1984). The a andx valuesdetermined in the presentstudy may be compared with values of 0.2640 and 8.136A calculated with the relationships of Hill et al. (1979) using the observed inversion parameter and the spinel-optimizedcation radii of O'Neill and Navrotsky (1983). Note that the lattice parameter determined from the profile refinement, even allowing for the cubic symmetry and well crystallizednature of the sampleused, is unusually precise becauseof the requirementthat the calculated and observedprofilesmatch all alongtheir steepsidesas well as at the peak center. The bondingdimensions resultingfrom the determined a and x values are summarizedin Table 3. As expected for a spinelwith x > 0.25,the three edgessharedbetween adjacent octahedraare shorter (in this case by 0.324)

HILL: REFINEMENT OF SYNTHETIC HERCYNITE

941

angle increases.A similar increasein Lorentzian character has beenobservedin other X-ray and neutrondifraction studies undertaken in our laboratories (Hill and Howard, in preparation) and in the caseof FelOaneutron diffraction data by Santoro and Prince (1980),who suggested that the variation in shape could be related to structural distortions and/or the presenceof particular distributionsof crystallite sizes within the sample analyzed. For hercynitethe neglectofthe peakshapedependence poorer fit between on 20 not only resultsin a significantly the observedand calculatedpeak profiles (Rwp = 9.999 compared to 0.07E), but also results in a systematic overestimation of the atomic thermal vibration parameters. However, in spite of this, none of the structural parameters are affected by more than two combined standarddeviations. When individual strong peaks are isolated from the pattern as a whole and refinementsare made of the individual e0 and W parameters,the resultant FWHM valueslie closeto the relationship betweenhalf-widthand 20 determined from the overall U, V and W parameters (Fig. 2). The similarity betweenthe form of the individual peak width variation with 20 and the variation produced by the Cagliotiet al. (1958) formula not only indicatesthat the Cagliotiet al. relationshipbetweenhalf-width and 20 is satisfactoryfor X-ray data, but also shows that the crystallites are relatively isotropic in shape and unstrained. The systematic variation in the individual e valuescan not, therefore, be related to the presenceof eitherof theseproperties in the sample.The only parameter with which e showseven a modestcorrelationcoefficient (0.42) is the peak width parameter 1,7and this relationshipis to be expected. The source of the profile shape change therefore remainsundetermined. Possible explanations could involve the effect of thermal diffuse scattering,changesin the diffractometer resolutionfunction, or variablecrystallite size contributionsto differentparts ofthe pattern.In any event, if the presenceof peak shape variations within powderpatternsprovesto be widespread it is clear that a 2&dependence for the profile shapefunction should be routinely implementedin the profile refinementprocedure. For hercynite the function = ao I e1 .20 adequatelyaccountsfor the shapevariation (Fig. 2), but this may not be the casefor other samples.

Baldwin, B. G. (1955) The formation and decompositionof hercynite(FeO . AlzOr). Journal of the Iron and Steel Institute. 179. 142-146. with and Barth, F. W. and Posnjak,E. (1932)Spinelstructures: without variate atom equipoints. Zeitschrift f0r Kristallographie,82, 325-341. properties Blass,G. (1964) Crystalchemistryand somemagnetic of mixed metal oxides with spinelstructure.PhilipsResearch Reports, Suppl.no. 3, l-139. Bragg,W. H. (1915) The structure of the spinelgroupof crystals. Philosophical Magazine,30, 305-315. Caglioti, G., Paoletti, A. and Ricci, F. P. (195E)Choice of collimators for a crystal spectrometerfor neutron difiraction. Nucfear Instruments,3, 223-228. Cooper, M. J. (1982)The analysisof powder diffraction data. Acta Crystallographica, A38, 264-269. Cooper, M. J., Rouse, K. D., and Sakata, M. (1981)An alternative to the Rietveldprofilerefinement method.Zeitschrift fiir Kristallographie,157, 101-117. Dehe, G., Seidel, B., Melzer, K., and Michalk, C. (1975) Determinationof a cation distribution model of the spinel . systemFe3-*Al*Oa. PhysicaStatusSolidi (a), 3 l, 439-447 Dickson, B. L. and Smith, G. (1976)Low-temperature optical absorptionand Mossbauerspectra of stauroliteand spinel. Canadian Mineralogist,14, 206-215. Fischer, P. (1967)Neutronenbeugungsuntersuchung der Strukturen von MgAlzOa-und ZnAl2oa-Spinellen, in Abhiingigkeit von der Vorgeschichte. Zeitschrift fiir Kristallographie,124, 275-302. Hafner, S. (1960)Metalloxyde mit Spinellstruktur.Schweizerische Mineralogische und Petrographische Mitteilungen,40, 207-243. Hall, M. M., Veeraraghavan, V. G., Rubin,H., andWinchell, P. G. (1977)The approximationof symmetric X-ray peaks by Pearson type VII distributions. JournalofApplied Crystallography, 10,66-68. Halloran,J. W. and Bowen, H. K. (1980)Iron diffusionin ironaluminatespinels.Journalof the AmericanCeramicSociety, 63,58-65. Hamilton, W. C. (1958)Neutron diffractioninvestigation of the ll9oK transition in magnetite.Physical Review, ll0, 1050r057. Hewat, A. W. and Sabine,T. M. (1981)Profile refinementof singlecrystaland powderdata:the accuracyof crystallographic parameters. AustralianJournalof Physics,34,707-712. Systematics of Hill, R. J., Craig,J. R., and Gibbs,G. V. (1979) the spinelstructuretype. Physicsand Chemistryof Minerals, 4,317-339. Crystallinity and Hindeleh,A. M. and Johnson,D. J. (1978). crystallite size measurement in polyamide and polyesterfibres. Polymer. 19, 27-32. Hoffman, A. and Fischer, W. A. (1956)Bildung des Spinells Acknowledgments und Al2O3. FeO . Al2O3und seinerMischkristallemit Fe3O4 I am grateful to my colleagues I. C. Madsen for assistance Zeitschriftftir Physikalische Chemie,Neve Folge, 7, 80-90. during datacollection andDr E. R. Segnit for discussions about Hovestreydt,E. (1983)On the atomic scattering factor for 02-. natural hercynite crystal chemistry, andto Dr H. G. Scottof the Acta Crystallographica, A39, 268-269. CSIRO Division of Materials Science for his constructive critiIbers, J. A. and Hamilton, W. C. (Eds.) (1974)International cismof an earlyversion of the manuscript. Vol. IV, Revisedand SupTablesfor X-ray Crystallography, plementaryTables to Volumes II and III. Kynoch, BirmingReferences ham. Bacon,G. E. and Lisher, E. J. (19E0) Cationdistribution and A neutronpowder Mason,T. O. and Bowen,H. K. (1981) difraction study of deuterated defect chemistry of iron-aluminatespinels. Journal of the a- and B-resorcinol. Acta Crystallographica, 836, l90E-l9l AmericanCeramicSocietv.64. 86-90. 6.

942

HILL: REFINEMENT OF SYNTHETIC HERCYNITE fitting method. NBS Technical Note I142, US Department of CommerceA.{ational B ureau of Standards,8- I 9. Scott, H. G. (1983) The estimation of standard deviations in powder diffraction Rietveld refinements. Journal of Applied . Crystallography , 16, 159-163 Segnit, E. R. (1984) Ferroan spinel from the Harts Ranges, Northern Territory, Australia. Australian Mineralogist, 45, 2s9-2ffi. Turnock A. C. and Eugster, H. P. (1962)Fe-Al oxides: phase relationshipsbelow 1000'C.Journal of Petrology, 3, 533-565. Wiles, D. B. and Young, R. A. (1981) A new computerprogram for Rietveld analysis of X-ray powder diffraction patterns. Journal of Applied Crystallography, 14, 149-151. Yagnik, C. M. and Mathur, M. (1968) Journal of Physics C (ProceedingsPhysics Society) l, 469472. Young, R. A. Prince, E., and Sparks, R. A. (1982)Suggested guidelinesfor the publication ofRietveld analysesand pattern decomposition studies. Journal of Applied Crystallography, 15,357-359. Young, R. A. and Wiles, D. B. (1982) Profile shapefunctionsin Rietveld refinements.Journal of Applied Crystallography, 15, 430-43E. Zhukovskaya,T. N., Camokhvalova, O. L., Bushueva,E. V., and Chichagov, A. V. (1980) Hydrothermal synthesis and study of somephysical propertiesof spinelsin the composition Mg1-*Fe*Al2Oa.Zapiski Vsesoyuznogo Mineralogicheskogo Obshchestra.109. 367-369. 14, 1983; Manuscriptreceived,September acceptedforpublication,May I, 19E4.

Meyers,C. E., Mason,T. O., Petuskey, W.T., Halloran, J. W., and Bowen, H. K. (19E0) Phaseequilibriain the systemFeAl-O. Journal of the American Ceramic Society, 63,659-663. O'Neill, H. St. C. and Navrotsky, A. (1983)Simple spinels: crystallographicparameters,cation radii, lattice energies,and cation distribution. American Mineralogist, 68, I 8l-194. Osborne,M. D., Fleet, M. E., and Bancroft,G. M. (1981) Fe'?+Fe3* ordering in chromite and Cr-bearing spinels. Contributions to Mineralogy and Petrology, 77,251-255. Palache, C., Berman,H., and Frondel,C. (1933) Dana'sSystem of Mineralogy, Seventh ed. Wiley, New York. Prince, E. (1980) Estimates of standard deviations in Rietveld powder refinement. NBS Technical Note 1142, US Department of Commerce/NationalBureau of Standards,20-23. Prince,E. (1981)Comparisonofprofile and integrated-intensity methodsin powder refinement. Journal of Applied Crystallography, 14, 157-159. Rietveld, H. M. (1969)A profile refinement method for nuclear and magnetic structures. Journal of Applied Crystallography, 2,65-11. Roth, W. L. (1964)Magnetic properties of normal spinels with only A-A interactions. Journal de Physique (Paris), 25, 50751 5 . Rumble,D. III (Ed.) (1976) OxideMinerals.Short CourseNotes, Vol. 3, MineralogicalSocietyof America, Washington, D.C. Sakata,M. and Cooper,M. J. (1979) An analysisof the Rietveld profile refinement method. Journal of Applied Crystallography, 12, 554-563. Santoro, A. and Prince, E. (1980) Use of the Pearsontype VII distribution in crystal structure refinements by the profile