THE SAFFLORITE-LOELLINGITE SOLID SOLUTION SERIES Dbnnts Reoclrnnn, Department of Geology, Uniuersity of Georgio, a Thens, Georgi

THE AMERICAN MINERALOGIST, VOL.53, NOVEMBER DDCEMBER, 1968

THE SAFFLORITE-LOELLINGITE SOLID SOLUTION SERIES

DBNNTsReoclrnnn, Department of Geology,Uniuersity of Georgio,A thens,Georgia AND , L. G. Bonnv, Departmentof GeologicalSciences, Queen's . UniaersitY,Kingston, Ontar'io.

ABSTRACT

Electron microprobe analyses of transition metal diarsenide minerals show that there is complete solid solution between coAs2-!'eAs2, limited solution of co in NiAsz and no solid solution between FeAs2-NiAs2. The only ternary Co-Fe-Ni diarsenide compositions are in the more cobalt-rich range (saffiorite). The variations ol a, b and c of the marcasitelike lattice with composilion follow smooth curves in synthetic (Co, Fe)Asz and fall close to these curves for natural (Co, Fe, Ni)Asr minerals. X-ray difiraction data are given for synthetic and natural Co-Fe diarsenides and the effect of Ni is described. The X-ray data for Berry and Thompson's (1962) saffiorite I, II, III, IV, V, and loellingite, are correlated with chemical composition' Synthetic CoAsz and two natural specimens of high-cobalt saffiorite have a measurable monoclinic lattice. The name saffiorite is retained for natural (co, Fe, Ni)Asz compositions frorn 3-100 mole percent CoAsr, with nickel from zero in low Co minerals to about 30 percent in low Fe minerals. Safflorite is dimensionally orthorhombic in the range from 97 to about 20 mole percent Fe and often measurably monoclinic in compositions lower in iron' The exact iomposition at which saffiorite becomes monoclinic has not been established. Studies of the stoichiometry of saffiorite-loellingite show that the minerals are characterized by vacant arsenic lattice positions which in most cases are fully occupied by metals in excessof a 1:2 metal-arsenic ratio.

INrnopucrroN Various workers have attempted to determine the exact chemical characteristicsof the samorite-loellingitesolid solution series (CoAsz- FeAsr).Noteworthy are Roseboom(1963) for his studiesin the synthetic system and Holmes (1947) for his work on natural phases. Holmes pointed out the heterogeneityof natural phasesand their closeassoci- ation with NiAsz and CoAsa.Thus many early analysesare probably incorrect. Until recentl-u-it was not possible to analyze the exact range of compositionsexisting in nature. Electron microprobe studies have been conducted (Radcliffe, 1966) on 80 natural samplesobtained from most known localities,through the courtesyof Queen'sUniversity, Royal Ontario Museum, University of Toronto, and the United States Na- tional Museum. Berry and Thompson (1962)published six difierent X-ray' diffraction patterns for saffiorite-loellingiteand named these types samorite I, II' 1856 SA F FLO RI T E-LOELLI N GI T]J 1857

III, IV, V and loellingite. These determinationswere made on minute single crystalsand small scratchsamples and it was not possiblethen to correlate these types with chemical composition. By use of electron microprobe, X-ray diffraction, and density measurementson small fragments (20 mg on average), the chemical, stoichiometric, and physical parametersof this solid solution serieshave been correlated.

AN,\ryrrc,c,rTncuwrquos

Eighty polished sections of transition metal arsenides were measured at various points for Co, Ni, Fe, As, and S. Of these, 39 were selected for more detailed study involving complimentary X-ray difiraction analysis and density measurements. Studies were made in the following order: rnonophase areas of about 2-3 mm in diameter were delineated by drilling a circle on the polished surface of the sample. The encircled areas were then re- checked for homogeneity in the microprobe. Some of the samples were seen to be multi- phase or heterogeneous and thus unsuitable for density measurements. Following microprobe analysis, the delineated area was extracted by means of a percussion drill. The density of the solid fragment u'as measured, prior to crushing and analysis by X-ray difiraction. As an aid in X-ray diffraction analysis and in the establishing of conversion parameters for microprobe output into chemical composition, six standards were synthesized for the CoAsr-FeAsz series.

Synthesis oJ cobaltiron diorsenides. The compounds were prepared by a subsolidus dif- fusion technique at 800'C and 10-a mm Hg in the anhydrous system with a low vapor volume. The exact weights of reduced metal sponges were mixed in the stoichiometric ratio of 1:2. Pure CoAsz (DR-1) and FeAsz (DR-6) were prepared together with 4 intermediate compositions of (Co,Fe)As2 at intervals of 20 mole percent. Four separate annealing processes,each of one week duration, lyere necessary to produce a homogeneous diarsenide reaclion product.

Electron microprobe onal,ysis. An Applied Research laboratories AMX microprobe was employed in this study An electron beam of approximately 1 micron diameter, accelerated to 20 KV, impinging on the sample surface, generates a sample current of about 0.020 microamps and a Ka X-ray counting rate below 5000 cps which is within the linearity range of the detection system. Correction was made to the integrated 40-second count for instrumental dead time and for the small differences of emission wavelengths of sample and standards (Radc1ifie 1966). Pure metal standards were mounted directly into the polished sections and ari intensity ratio obtained. This ratio was converted into weight percentage using Duncumb and Shield's backscatter correction, Nelm's ionization efficiency correction, Philibert's absorption correction and Castaing's fluorescence correctionl. The method is satisfactory for alloyed phases but not for phases with ionic and covalent bonding unless the absorption correction is rnodified to allow for the difierences of interatomic spacings of pure metal standards, and chemically combined phases. A modification of Philibert's absorption correction was made and basic mass-absorption data were obtained from Kelly (1966) for Ka ra,liation as follows.

I Unpublished manual for electron probe microanalyzer, M.I.T., 1965. 1858 DENNIS RADCLIFFD AND L G. BERRY

Endtter CoAbsorber Fe Absorber As Absorber Co 58 61 110 Fe 68 76 135 As 160 l/J 38

The mean mass absorption for the alloyed phase was computed equivalent to coAsz and FeAsz. This was multiplied by the mean density of each composition which was extrapolated from the pure metal end members after their densities had been measured on a Berman Balance. The resulting linear absorption coefficient was subsequently divided by the mean density of the minerals to give the new mass absorption coefficient' A series of natural samples were analyzed independently by X-ray fluorescence but due to the difficulty of measuring Co, Fe, Ni, and As, in high concentrations, and to sample heterogeneity, calibration curves utilizing these data have a poor correlation. However their conversion parameters of X-ray intensity ratio (I /Io) into weight percent are close to the calculated parameters. The mean of the measured conversion parameters, a constant derived by a least squares straight line on the same data, and the calculated theoretical proportionaiity constants Q e. I IIoXK) are given below:

Parameter K(Co) K(Fe) K(,4s) means of measured 0.945 0.924 1.035 least squares o.997 o.921 1.035 calculated 0.940 0.926 r.037

The mean analytical total of 100 analyses after correction was 100.2 percent with a standard deviation of * 1.4 percent.

Density measurem,ents.The density of samples averaging 20 mg was determined by a weight loss procedure on a Berman Balance. Carbon tetrachloride was used as a liquid medium as it has a low surface tension (26.95 dynes/cm at 20"c) and a higher density (1.595) than toluene. The precision error, obtained with Herkimer quartz and pure galena is *0.39 percent. The density of transition metal arsenides approach 7.0 and the acceptable measuring precision on three successive measurements rvas thus *0.03. x-ray ilifraction analysis. The mineral fragments were ground to a fine powder and mired with Herkimer quartz as an internal standard. Smear mounts were scanned at ]' per minute on a Norelco diffractometer using cu Kar radiation (r:1.54051 A) and a'r,iF monochromater. No beta filter is necessary but the increase in intensity of the X-ray beam is balanced by the inefficiency of the LiF diffraction monochromater. The method gives sharp peaks, and permits operation at very high sensitivities with a low even background. Theonlywell-definedpeakswhichcanbereadilyindexed are2l0, lol, l20, and 111. consequentiy all samples were scanned 10 times through the interval 33"-40"20. The unit-cell dimensions were then determined simultaneously by least-squares calculations on all four diffraction peaks. The values obtained for the cell constants (Tables 4 and 5) compare favorably to +0.01 A with earlier results on synthetic and natural material'

CnBurcar- CollposrrroN I{omenclature.The terminology of the saffioriteJoellingite compositions is not well defined. Saffiorite and loellingite have been considered as SAFFLORITL-LOELLINGIT]' 1859 distinct mineral species.Peacock (1944) reported that saffioriteis mono- clinicl with rectangularaxes for a Nordmarken specimen(Sweden) and loellingite is orthorhombic with a composition close to FeAsz. Holmes (1947) noted a concentration of analysesalong the mid-region of the CoAsz-FeAszjoin, and closeto FeAs2,and proposeda cut-ofi between saffiorite and loellingite at 70 mole percent FeAs:. Roseboom (1963) cited the previously reported coexistenceof saffiorite and loellingite. He also describedthe symmetry of synthetic CoAszas monoclinic.For thesereasons saffiorite and loellingitehave beenconsidered to be distinct mineral species,loellingite being nearly pure orthorhombic FeAs: and saffiorilemonoclinic (Co, Fe)As, with a beta angle closeto 90o. Further study of Nordmarken saffiorite (saffiorite I) and of saffiorite III by Weissenbergmethods reveals no evidenceof monoclinicsymmetry. The composition reported for Nordmarken safforite by Peacock (1944),based on previous analyseson material from the same locality, is suspect.Microprobe analysesof Berry's sample (M4035) and Pea- cock's sample (M4035 pk) gave similar results which are significantly different from that quoted by Peacock(Palache, et al. 1944,p. 308):

Mole 7o Peacoch M4035 pk M1035 CoAsz 44.3 1o.., t/.J FeAsz 55.0 84.7 82.5 NiAsz o.7

The present investigatorshave found a complete range of compositions from CoAszto FeAsz(Fig. 3) and, with only two exceptions,the lattice of these compositions is dimensionally orthorhombic with a marcasite-likestructure. Of the natural samplesexamined, 25 percent lie in the field of loellingite with less than 2 mole percent metals other than iron. The remaining 75 percent are distributed fairly evenly along the CoAsz-FeAs2join. It is now recommendedthat loellingitebe usedfor FeAsz with less than 3 mole percent CoAsz2in solid solution and safflorite for the CoAsz-FeAs2series. This terminology is in keeping with that used by Berry and Thompson (1962),and with previous usage.

Rangeof natural compositions.Roseboom (1963) studied the synthetic Co-Fe-Ni diarsenide system and showed complete miscibility of Co- Fe-Ni in a diarsenidestructure at 800"C exceptfor a small region close to Ni on the Ni-Fe join (Fig. 1). Further work on the extent of this region at higher and Iower temperaturesled him to concludethat the immiscibility gap probably constitutes a solvus dome. The complete

1 Based on asymmetric intensities at hrgh 20 values on Weissenberg films. 2 n-ickel was not found in iron diarsenide. 1860 DENNIS RADCLIFFE AND L G. BERRY

-- solvus qt 8OO"C

ffisotvus tor noturol phos6s bca€d on Holmog A Jourovsky

Frc. 1. Solvus for synthetic (800"C) and natural ((800'C) phascs, according to Roseboom (1963). solubility along the Co-Fejoin in the synthetic systemwas substantiated by Radcliffe (1966) at 800oC. Roseboomalso compiled analysesfrom Holmes (1947) and Jouravsky (1959) and sketched the extent of the solvus for natural phases at temperatures much lower than 800"C (Fig. 1). Holmes (1947) compiled all available analysesof natural diarsenides (about 50) published in the period 1810 1945 and obtained the extent of the compositionlimits of natural phases(Fig. 2). Chemicalanalyses obtained in this study (Radcliffe 1966)are plotted on Figure 3. This showsthe extent of natural compositionsdetermined by microprobe techniques. Some representative samples are listed in Table 1. This limit representsthe approximate extent oI the solvus at low temperatures(200-300'C). It is pointed out that coexistingnickel- rich (generally niccolite) and nickel-poor phases were frequently ob- servedduring this study (Table 2). Of the 88 analysesplotted on Figure 3, only three Iie outside the limits indicated. These phasesare quite abnormal, of very limited extent in the sample,and are only detectable with a microprobe e.g. one of the phasesoccurred only at one point in sample 1862 in a 5 micron zone enveloping a core of native silver. Further away from the silver a more normal compositionwas measured with lessthan 5 mole percent NiAsz. No cobalt-freeFe-Ni diarsenideswere found; this is in accord with SA F FLORI T]],-LO ELLI N GI T ]i 1861

6xt0nl nolurol

Frc. 2. Extent of natural phases,according to Holmes (1947).

Mikheev's observations (1951). In addition, no ternary compounds exist in Fe-Ni diarsenidesunless the cobalt content is in excessof about 25 mole percent.These observationsare almost completely contrary to the relationships given by Holmes (1947) and reproduced here in Figure 2. However examination of Holmes' paper shows that the existenceof compositionsin the nickeliferousloellingite field is based on only three analyses,numbers 80, 81, and 129.Analysis 80 is by Kobell (1868) who notes the presenceof arsenopvrite.Analysis 81 was per- formed by Rammelsberg(1873) who describedthe sample as loellingite surroundedby breithauptite. Sample 129was analyzedby Duerr (1907) who notes admixed pyrite and rammelsbergite.Thus in all casesthe analyseswere probably done on impure material, the impurity probably having a nickel content. The validity of these analysesmust be seri- ously questionedand the existenceof a nickeliferousloellingite is un- Iikely. The existenceof ferriferous rammelsbergiteis also suspectfor similar reasons. The complete absenceof any natural compositionsalong the Fe-Ni join is in sharp contrast with the completerange of compositionsalong the Co-Fe join. There is an indication of a weak diffuse maximum near the mid-region,which is a little more Ni-rich than previousl,vdescribed' This is due to the type of sample analyzedas thesemore Ni-rich members (5-10 mole percent NiAsr) commonlv occur as thin veneers(aver- 1862 DE.NNIS RADCLIFFD AND L G, BIIRRY

/ erlcnl for mosl I \ nolurol compo3ition8

,'oi'^.

Frc. 3. Extent of natural phases analyzed in this study. aging 1 mm) surroundingnative silver. Such samplesare easilya\alyzed with a microprobe,but not by other, more conventionaltechniques. The existence of nearly pure cobalt end-member compositions in mineralshas not beenpreviously reported. These samples are intimately intergrown with skutterudite (Radcliffe, 1966)on a fine scale(0.05 mm lenses)and thus are only amenableto analysisby microprobe. The range of compositionsalong the Co-Ni join is somewhatpeculiar

T,s.rr-r1. Serrtonrtn: Srr-rcrBnMrcnopnoen ANr,vsns nnon Reocr,rnrr(1966)

1B75 lB52 1B63

Co 2.7 10.8 16.8 162 20.9 22.2 21 2 Ire 25.9 207 9.7 5.6 6.9 5.0 Ni 2.1 6.9 0.3 ZJ As 712 677 71.7 7r.2 69.3 68.4 704 S 0.4 1.1 0.5 0.5 2.6 19 1..)

Tota] 100.2 100.3 100 8 100.4 100.0 99.8 100 7

0/6 Mole 90 326 45.5 Jt .l 55.6 734 745 72.4 fe 91 0 67.4 47.6 350 204 )\6 27.6 Ni 6.8 /.J 23.9 f.i 7.8 S1 F I.-LO RI TE- LO ]'LLI N GI T E 1863

'lrntl. 2. Onsonvrl ConxrsrrNcPn.tsrs

Ita Ni

Sa SK I,O Ita Ni Ce

Ag:native silver Ni: niccolite Sa: saffiorite Qs: gersdorffite Sk: skutterudite Ar: arsenopyrite Lo: loellingite P6: pyrrhotite Ra: rammelsbergite coexisting discretephases; - : coexisting phasesno I observed. as there is apparently a limited solubility of Co in NiAsz but no sollr- bility of Ni in CoAsz.Synthetic studies (Roseboom,1963) along this join show that some Ni could be expectedin CoAsz.Possibly the reason for this discrepancylies in the composition of the geochemicalliquids from which these phases crystallized. The processesof geochemical differentiation of Co-Fe-Ni towards an arsenic-richliquid are such that iron-free Co-Ni liquids do not separate,nor do cobalt-poor (lessthan 25 mole percent) Fe-Ni liquids. This explanation is supported to some extent by the compositionsof observed coexisting phases (Table 2). For example loellingite (FeAsr) was not found in coexistencewith a nickel-rich phase,whereas saffiorite (Co, Fe)As'rand niccolite are often intimately associated.

Zoning in natural phases. Saffiorite associated with a polyminerallic assemblageis invariably zoned or shows definite heterogeneity with regard to the distribution of the Co-Fe-Ni content. Table 3 gives some data for samplesassociated with native silver and those Co-rich mem- members associatedwith skutterudite. The width of the zonesis quite variable but is often 5-10 p wide in contact with the other phasesbut may be up to 50-100 p. A diagnostic feature of samplesfrom the Cobalt-Gowgandaregion, Ontario, is the occurrenceof coresof native silver which forms a recti- linear pseudodendriticpattern. Saffioriteforms a thin veneer (average 1864 I)].NNIS RADCLIFtr.E AND L. G. BERRy.

Tanr.a 3. CouposrrroNs(Mor-r /) ol ZoNro AnsnNrors (Frc. 4)

Metallic Cores & CaL Co # Calcite Margins

lB34 Ag core 512 488 masslve 51.9 20.3 27.9 'rt 1835 Ag core 68.5 7 6.9 masslve 54.5 31.9 13.6

tB52 Ag core 90 1 10.0 massrve 746 1.0

1B60 Ag core 400 49.7 9.3 lnASSIVC 47.1 Jl.o t.J margrn 52.3 47.0 0.7

tB62 Ag core 27.8 55.2 17.o InASSIVC 31.0 64.2 4.7

rR76 Niccolite core 95.8 J.I MASSlVE 845 8.5

t c01 Ag core +z.J masslve 36.6 634 margln 286 71.+

Two Phase Zoning

1875 Skutterudite Safflorite

1B83 Skutterudite 5.8 Safflorite 9.1

1 mm) around the silver, and both phasesapprently replace calcite. Microprobe analyses of these samples show strong asymmetry with regard to the nickel distribution. Nickel is either concentratednext to silver, or alternatively towards the marginal phase in contact with calcite. Figure 4 shows that saffioritewith a Ni-poor core tends to be more cobalt-rich, and that with a Ni-rich core tends to be more iron- rich. This implies that the nickel content is dependent on the Co/Fe ratio (Fig. 5). It is pointed out that the direction of migration of the composition of the saffloritefrom the core phase outwardsis towards a field in which there is a greater frequency of sampleswith similar compositions(c/. SA F F LORI T N. LOELLI N GI T E 1865

distonce.5o,1) f"H""

o silv6r coro

eniccolile coro . skutlcrudil6

Frc. 4. Compositions of zoned saflorites. Lines with arrows represent the first seven samples of Table 3, lines without arrows the lower two samples in which safforite and skutterudite coexist.

+ I 6 t .EI

o o o c' o o c s

7o decreose of Co/Fq rotio ---.;

Frc. 5. Nickel content of zonedsafflorites. 1866 DENNIS RADCLIFFE AND L. G. BERRY

Fig. 3). In most casesthe saffioritephases in immediate contact with native silver have abnormal compositions. Figures 6 and 7 illustrate the types of zoning found in saffiorite- loellingite.Figure 6 is of a loellingite with an average1.4 mole percent CoAszin solid solution. Halos of saffiorite which vary from 50 to 100

As Fs

Frc. 6. Top left: Electron backscatter image of loellingite with discrete grains of native silver. Top right: X-ray scanning image showing a concentration of Co in halos surround- ing native silver. Bottom left: X-ray image shorving homogeneousdistribution of As in loellingite. Bottom right: X-ray image showing a more heterogeneousdistribution of Fe in loellingite. Grid scale:0.03 mm. S A F F LO RI T E- IN E L LI N GI T I':. 1867

!r$ 4r {tqr4}

€e i*

. :. ...:.:.-:-. : :i:.. rir:j,:rr ir:i:::ir:::tjri:t:!:::r:i:

Frc. 7. Top left: Electron image of saffiorite and native silver which replaces calcite (black). Top right: X-ray irnage showing the distribution of Ni in saffiorite Nickel is concentrated at the margins of saflorite in contact with both silver and calcite. Note the presenceof some Ni in native silver. A spot check gaveO.30/6Ni in this silver. Bottom: X-ray images of Co and Fe. The Co/l-e ratio decreases from left to right across both photographs and the zoning is thus difierent from that of Ni. Grid scale 0 03 mm. microns in width occur immediately in contact with native silver. Two types of zoning are shown in Figure 7, in which two areasof native silver, separatedby calcite, are all in contact with saffiorite. Nickel is concentrated at the margins of the samorite in contact with silver and 1868 I)L,NNISRADCLIFFE AND L. G. BERRT/ calcite, whereasthe Co/Fe ratio of the saffioritevaries acrossthe sec- tion without apparent regard for the phaseswith which it is in contact. A puzzling feature is the small content of Ni in native silver, a spot check gave 0.3 percent nickel.

X-Rev Cnvsrerr-ocRAPHY The six X-ray patterns of Berrl'and Thompson (1962) which char- acterize the saffiorite-loellingite series, index on the following orthorhombic cell dimensions:1 i a^ bA cA

Loellingite 5.29 5.98 2.88 Saffiorite I 526 5.98 294 Saffiorite II 5.11 591 3.11 Safflorite III 5 175 5.950 3.015 Saffiorite IV 5.16 5.93 299 Saffiorite V 5.06 5.89 3.11 This variation of the elementarv lattice constants gives rise to a complicated seriesof diffraction peaks which overlap and interchange position. These reversalsare most apparent in the following pairs of peaks; 120-l}l, 2l}-lll, 220-121,2lt-130, and these give the patterns their characteristicappearance. In addition a group of four peaks at higher 20 positions(310,031, 221, l3L) show more complicatedinter- changes. The relative positions of these peaks for Berry and Thompson's six types are plotted on Figure 8 as bar graphs,in the order of increasing cobalt content. The saffioriteI powder pattern differs only slightly from loellingite with a decreasein A20in the iine pairs 120-101,210-lll, 220-121,and |3O-2II. In saffioriteIII and IV, which are quite similar, the lines in each pair noted above have interchangedposition. This arrangement covers the range from about 30 to 70 atom percent cobalt with very Iow nickel. In saffioriteII and V the order of peaks remains the same as in III and IV but Ad within each pair increases,apparently with increasein Co/Fe or (Cof Ni) f Fe ratio, approachingpure CoAszin V. 'Ihe changebetween saffiorite I and III and between IV and II seems abrupt but resultsfrom a continuousincrease in c and decreasein o and D, as expressedin the shift of 011, 101, 111, 121 and 2ll to larger d values, accompaniedbv a shift of l2O, 210, 220 and 130 to smaller d vaiues.In the presentstud1, saffiorite frorn IB20 and IB85 (both inter-

I Appelations Salilorite I V refer to X-ray patterns orly, and are not strictly mineral or varietal names, although they are characterized bi'distinct chemical compositions. SA FF LORI TE. LO]iLLI NGI T N 1869

o ri'$AI

o N O1

Co cYK., t39 3,2 q4 l2o i*l's4oslA. 46 4.a I

Frc 8. 29 versus intensity (ordinate) for natural safflorites. 1870 DLNNIS RADCLIFFE AND L. G. BERRY grown with skutterudite) gave powder patterns in which hol and hkl peaks are split or broadened, these require monoclinic indexing. All other natural specimensgave patterns which could be indexedas orthorhombic. Weissenbergstudy of several natural saffiorite fragments (including one of saffiorite V) has revealedno divergence{rom orthorhombic svmmetry. Peacock (1944) recorded monoclinic symmetry with 0:90o from one set of Weissenbergfilms, but this has not been observedagain. A similar bar graph (Fig. 9) has beenprepared for s-vntheticsaffiorite. The cell constantsare given in Table 4. DR-1 (CoAsz)shows a splitting oI h\l and hkl peaks but no change in hk} and 0&1.This is distinct evidencefor monoclinicsymmetry. DR-2 is dimensionallyorthorhombic although the hhl and h\l peaks are somewhat broader than similar peaks for less cobalt-rich compositions.Apparently 80 mole percent CoAsr is closeto a measurabletransition from orthorhombic to monoclinic symmetry. For easeof comparison,the mean hol and hJl, and fikl and hkl, were obtained and ceil constantscalculated as though for orthorhombic symmetry. This is a reasonableprocedure as B of monoclinic CoAsris very closeto 90o. The X-ray difiraction patterns for s1'nthetic saffiorite show very regular trends which are analogousto those found in natural material. DR-6 (FeAs2)is identical to ioellingite, DR-5((Feo sCoo:)Asr)is very similar to saffiorite I and DR-1 (CoAsz)is very similar to saffioriteV except for the splitting of hkl and h}l reflecLionsin the former. Comparison of saffioriteI to V with the synthetic data and microprobe analysesof samplespreviously classifiedb1' Berrv and Thompson (1962) into a saffioritetype gives a distribution of types shown chemically in Figure 10; boundariesare arbitrary. Onll-saffioritelI and V contain any appreciablenickel in the structure and this has a strong effect on the X-ray patterns. The effect is to 'normalize' the pattern to that of pure CoAsz.This may be explained in terms of the available valency electronsof Ni(4), Co(3), and Fe(2). The effect of Ni and Fe in a saffioritestructure is to produce a mean of 3 valency'electrons,i.e. the sameas cobalt, and the effectivecovalent radii and electronic structure of Co, and Ni plus Fe, are thus very similar. As a consequence,the lattice dimensionsof CoAs.zand saffiorite with some Ni and Fe, are quite similar. Clear evidence of the monoclinic character of synthet.ic CoAsz to CooaFeo.zAs2, seen in the doubling of several hol and hkl peaks, is available in this study, in the powder work of severalearlier authors,and in the single crystal study by Darmon and Wintenberger (1966). Rose- boom (1963) also recorded a monoclinic cell for a synthetic sample S A F F LO R I 7' E- LO E I. LI N GI T I1:,

= o o$ !ol ot -l N N A o N- T9 A A

9 l-: o 6t 6l INN ENot

oze CuK4l I=1.54O51 32 3.4 46 4^e I

Frc.9. 20 versus intensity (ordinate) for synthetic saffiorites. 1872 DDNNIS RADCLIFFI], AND L, G. B]],RRY

o ,' o,r'o------/ / O ^t' o t?it od--_ t,/ -. e vno i'o .oap.o o / "-a"^ o .. ,t,64 ,--6.---6T?9-:-=--

Itc. 10. Comnositions of saffiorite I to V. with Co:Ni:7:1. All other compositionsyield patternswhich can be indexed on orthogonal axes.In addition a few weak lines require the doubling of o and c, as noted below.

Cnvsrar S:rnuc'runr Buerger (1932)established that loellingitehas the marcasitestructure with space group Pnnm. Peacock (1941) confirmed this space group; Peacock and Dadson (1940) reported the same space group for rammelsbergite(NiAsr) and Kaiman (1947)confirmed that rammelsbergite also has the marcasitestructure. Roseboom(1963) showed that complete solid solution occursin synthetic phasesat 800oCin NiAsz-CoAsz and in FeAsr-CoAsz.Studies of natural specimensshow a complete solid solution in FeAsr-CoAs2,but synthetic CoAsz and some of the cobalt-rich natural phases show a splitting andfor broadening of iefrl and h\l reflectionsindicating a changeto monoclinicsymmetry. Quesneland Heyding (1962), Roseboom (1963) and Swansonet al. (1962, 1966)have all indexed powder diffractometerpatterns of CoAs.z using a monoclinic lattice with a B centered cell in which 0 = 90o30', a:10.1 and c:6.24 h are doublethe valuesfound for the loellingite- like ceil.The latticecan be describedby a P cellwith a:5.916,b:5.872. SAI,']''LORITI':.-I,OF:,LLINGITTi 1873

,:5.960 A, P:1I6"27' (Swansonet al., 1966). Darmon and Winten- berger (1966) have now confirmed this Iattice geometrlr with single crystal measurementson synthetic CoAs2,and have deducedthe crystal structure of monoclinic CoAsz. The variations of cell dimensionsof CoAsr-FeAszsolid solution series are shownin Figuresll,12, and 13.With increasingcobalt content there is an increasein c and cell volume, and a decreasein o and 6. Hulliger and Mooser (1965)have explainedthese distortions in terms of "Jahn- Teller" stabilities with respect to octahedral symmetry of the metal- arsenic atoms. They show that hybridization of the electronic con- figuration of the metal atoms, which favors the stability of compressed arsenic octahedra, can occur. Hulliger and Mooser think that the existenceof phases with the marcasite structure is most readily explained in terms of "Jahn-Tellert' distortions. However they do point out that phasescan exist (including marcasiteand rammelsbergite)in ttJahn-Teller" which no distortions can occur. These are called "anom- alous" marcasites.They also point out that the properties of phases within each group are variable and the only classifyingparameters are the ratios of the cell edges.Thus:

c/a c/o

ttanomalous" marcasites : o.73-O.74 0. 61-0. 63 (marcasite, rammelsbergite) CoAsz : 0.62 0.53 "Jahn-Teller" marcasites : 0.54-0.56 0.480.49 (loellingite)

CoAsz(Fig. 13) falls midway between the two main groups, and the ratios cf a and c/b vary along smooth curves with increasingiron content to those of loellingite. The marcasite structure is characterizedby chains of metal-sulfur octahedrawhose shared edgesare perpendicular to c. In marcasiteand rammelsbergitethe metal-metal distancealong c is 3.39 and 3.54. In loellingite this distanceis reduced to 2.88, which increaseswith cobalt substitution to 3.13 (for pseudo-orthorhombic subcell) in saffiorite. Darmon and Wintenberger (1966)in their structural study of CoAs2 find Co-Co distancesof 2.77 and 3.47 approximately parallel to c. The sum of these two is equal to the c length of the B centred monoclinic Iattice of CoAs:u.These two distancesalternate along the c axis. The structure of CoAszis similar to arsenopyritein which Fe-Fe distances arc 2.89and 3.53 alternatingalong the c axis of the B centredmonoclinic lattice. The interatomic distancesof 2.9, or less,indicate the existence r874 DENNIS RADCLIFFE AND L. G, BERRY o---'- --o-.------u-.___u_. o ._l:

,--"t"'-o-'

O--\_

Frc. 11. Synthetic saffiorites. Composition versus lattice parameters.

94.O

l3 9t.o

6.O DI

5.8 5.4

5.1 curves ore tor synthelic somDles. 3.1 ------:.-.-.-..=-. cA 29

Frc. 12. Natural saffiorites. Comoosition versus lattice parameters. SA F F LO RI T E-LO LLLI N GI T IJ

o.50

o.48 CoAsZ FeAe2

Frc. 13. Natural and synthetic safflorite. Co/I.e ratio versus c/s and c/b (loellingite-type cell). of polycation bonds (Hulliger and Mooser 1965). Nickel (1968) has recently offered an explanation of this phenomenain terms of ligand field theory and the number of nonbonded d-electrons.He points out that substanceswith 6 nonbonding d-electronsare stable as marcasite and rammelsbergite or as cobaltite, while those rvith 5 form arsenopyrite or saffiorite (CoAsz) and one with 4 occurs as Ioellingite (FeAsz).In marcasite iron atoms repel each other across the shared octahedral edge,in loellingite they attract, and in saffiorite they alternately attract and repel. The saffiorites,in general, are intermediate between loellingite and CoAszin composition,must develop an increasingnumber of longer intermetallic bonds with increasing Co content until an ordered arrangement of alternate long and short bonds is possiblein high Co compositions.The increasing c from loellingite through in- creasingCo-bearing saffiorites is a measureof the averagemetal-metal distance acrossthe shared octahedraledge. The number of nonbonded d-electronsalso increases from 4 in loellingiteto 5 in CoAsz.The presence of nickel in high Co saffioritestends to increasethe number of nonbonded d-electronstowards the 5 found in CoAsz,nickel is not found in iron-free saffiorite. 1876 I)LNNIS RADCLIFFD AND L. G, BN,RRY

Cnvs:rar Cnnursrnv Propertiesof sffiorite. l'he cell dimensions of CoAsz-F'eAszare given in Tables 4 and 5. The variation of cell dimensionsalong the CoAsz-FeAsz solid solution serieshas been determined for the synthetic system by Roseboom (1963) and by the present authors. Both sets of data are plotted on Figure 11 and good agreemenlwas obtained. The value of c decreaseswith increasing iron content, while o and b increase.The onll'Iinear relation of compositionis with molar volume and the equation of this line is: mole percent FeAsr : 5629.280- (60.547X molar volume At) The caiculateddensities of all the synthetic compositionswas 7.47+ 0.01. Thus the decreasein the cell weight due to increasingiron is balanced bv a concomitant decreasein molar volunte, probably due to smaller effectivecovalent radius of iron. In the natural system (Fig. 12) there is reasonablygood agreement of similar plots for nickel-free compositions (i.e. compositionswith greater than 50 mole percent FeAsi). For compositions CoAsz- (Cos.5Fe6.5)Aszthere is a wide scatter of points and this is probably related to the effect of nickel on the structure. The plot showing least scatter involved the chemicalratio of (Cof Ni) /Fe (Fig. 12). Since c is the Ieast variable parameter the data can be somewhat normalized by comparing the cf a and cf b ratio with the Co/Fe ratio. Figure 13 showsthat the valuesfor the natural material are consistently higher than those of the synthetic samples. The reasonfor the scatter of points in the more cobalt-rich samples is imperfectly understoodbut is most probably related to variable nickel and sulphur contents,variable temperatureof formation, and variable stoichiometry. The stoichiometry of six saffioritesof various compositionsand six nearly pure loellingitesare given in Table 6. This was determined by consideringonl-v the simplest models (a.g.Frenkel or Schottky defects) and matching the calculated density of these simple models with the measureddensity. The compositionof the six saffioritesvaried from (Co6z;Feo.z;)Asr to (Coo.rFeos)Asz. The stoichiometry is quite variable and no regular trends have been established.However a tendency appearsin natural saffiorite for part of the anion lattice positions to be occupied by the metals in excessof a 1: 2 metal-arsenicratio. The metal-arsenicsolu- bility limits for synthetic CoAs2were establishedby Roseboom(1963) as CoAsr.ss-CoAs1.ee.Natural samplesapparently show much wider solubility limits, i.e., (Co,Fe,Ni)As1.3x-(Co,Fe,Ni)As2. SA FF I,ORI T E-LOELLI N GI T E t877

T.crlr 4. Crrr, CoNsraxrs e,no CouposrrroNsrN rrrn Sor,rnSor,urroN Snnrns CoAszl-eAs:

Specimen Number Co Fe Ni As S

DR 1" 5.072 5.878 3.130 100 0 0 100 0 z 5.1& 5.913 3.037 80 20 0 100 0 3 5.207 5.946 2.979 60 40 0 100 0 4 s.262 5.979 2.940 40 60 0 100 0 5.282 5.975 2.91r 20 80 0 100 0 6 5.291 5.987 2.880 0 100 0 100 0

1B20b 5 048 5.886 3.rM lJ. t lJ. / 10.6 91.5 8.5 1852 5.179 5.820 3.122 73.4 25.6 1.1 9l.9 8.1 tB32 5.005 5.962 3.r42 58.7 IJ. / 25.6 94.8 5.2 1BU-2 5.173 5.946 3 002 590 37.5 J-.) 92.2 7.8 1850 J.JZI 5.815 2.955 45.5 47.6 6.8 99.8 0.2 18 16 5.237 5.926 2 958 32.6 67.4 0 100.0 0 LB77 5.259 s.970 2.919 16.3 847 96.5 3.5 1C03 5.259 6.008 2.903 9.0 91.0 98.7 l..J 1815 J. J l.) 5.949 2.906 7.4 98.6 984 1.6 rB05 5.281 6.000 2.888 100 100

Monoclinic, " B not determined, Swanson et at. (1966) give 90"2g1t. b Monoclinic with B:96"16,

Taer.n 5. Crr,r, DrunxsroNs on puno CoAszano tr-eAsz

Author bA cA Material

CoAsz A M. Rosenqvist (in T. R. Rosenqvist, 1953) 5.08 5.89 310 synthetic Heyding & Calvert (1960) s 0054 587 3 114 synthetic & Heyding (1962) Quesnel 4 8934 5 .885 3.167e 90o311/2' synethtic Swanson et al (1960) 5.0474 5 872 3 1274 90'51', synthetic Swansonet al (1966) 5 0484 5.872 3.1274 90o28r/2' slmthetic Roseboom (1963) 5.0514 5 .873 3 1274 90"27', slmthetic Darmon & lVintenberger (1966) 5 064 s86 3 122 90036', synthetic Radclifie (DR-l) 5 056 5.878 3. 130 synthetic Berry & Thompson (1962) 506 s89 3 11 natural (safforite V) 5 048 5.886 3.144 s0.10' natural IreAsr Buerger (1932) 526 5.93 285 natural Peacock(1941) 529 s .98 288 natural Heyding & Clavert (1960) s 300 5 982 2.882 synthetic Swanrcn et al (1960) 5 300 5.983 2.882 synthetic Berry & Thompson (1962) 5 .3023 5 9818 2 8802 natural Roseboom(1963) 5 301 5.979 2 882 synthetic Radclifie (DR-6) 5.291 5 981 2 880 synthetlc Radclifie (extrapolated) s 285 s.998 2 887 natural a Pseudo-cellvalues-true cell lengths of B lattice are double. (; 1878 .D,/'lTNlSRIDCLIFFI| AND L. BIiRRY

TeBr,s 6. Stotcnrourrtv ol Nlruner- ARSENTDES

A/B ratio Sample I Composition Stoichiometry

Saffiorite r.c03 high iron Ar(A olBr m)z 7.34 t:1.89 J AA tB77 Ar ozBroo l:1.95 1816 A1(A 68r sz)r Iil.92 1B44-l Ar(A uBr m)z 6.99 l: | .97 1B63 A1(A mBr ss)z 722 1:1.93 1.'.) 1B20 high cobalt AqBt 7.t2

Loellingite S" 1ts05 A1(A mBr u)z 7.40 1:1 .83 1C16 2.3 Ar(A orBrm)z ,/.JJ l:1.92 tB7 | 26 Ar(A oeBrsg)r sz 7.21 1:1.84 1C14 4.1 Ar(A orBrse)r m 7.29 1:1.93 1865 5.3 (A eeBm)B: 7.28 1:2.08 1815 t.6 (A wB o')Bz 6.97 1:2.04

Ni; B:As' S * compositionsin mole /, Dm:measured density; A:Co, Fe, Propertiesof toellingite. The cell dimensions, compositions, and stoichi- 6' ometr,vof natural and syntheticloellingite are given in Tables4, 5' and The stoichiometry of loellingite has been investigated and shows defect structures similar to that of saffiorite.In Fe(AsS)u,the sulphur content apparently controls the stoichiometry, i.e. with increasing posi- sulphur, the amount of iron occupying the vacant arseniclattice a tions tends to decrease(Table 8). This trend is also accompaniedby generalincrease of the metal-arsenicratio. These solubility limits vary Lom FeAsr.r, to FeAsr.osand may be compared with svnthetic Fe(Feo.o+Asr.go),2which has a metal-arsenicratio of 1:1'93 (Heyding and Calvert. 1960). The variation of sulphur with other physical and chemicalproperties is shown in Figure 14. Samples1871 and 1C14 both fall below the correlation lines of a and.b.These are from Edenville, New York, and are distinctive in that their stoichiometriesshow overall anion deficiency

discussedabove and loellingite from Franklin, New Jersey' In this Iatter case,Buerger (1932) and Peacock(1944) independently reported Fe(Feo.osAsr.ru):which compareswith the presentdetermination (1B05) of Fe(FeoouAsrun)r. SA F F LORI T E- I.OE LLI N GI T E r879

7.45

7.40 Donsiiy

7.30

7.20

6.O5 6.OO bE

5.90

5.85

5.35

5.30 ol 5.25 5.20

2.95

2.9O c I za5 ro (o o b @F .D dl od) dt 2.80 Mol€ 7" Sulphur

Frc. 14. Loellingite. Sulphur versus lattice parameters and density. 1880 DENNIS RADCLIFFF AND L. G. BERRY

'foronto, The united States National Museum, Royal Ontario Museum, University of and Dr. E. H. Roseboom, kindly supplied samples for this research work.

RBnnnnNcBs

Bnnnv, L. G. eNo R. M. Tnoupson (1962) X-ray powder data Ior ore minerals Geol.soc. Amer. Mem.,85. 165- Burncnn, M. J. (1932) The crystal structure of loellingite, Fe\sz. Z' Kristallogr ,32, r87. Danuor, R. & M. WlNrnNsnncnn, (1966) Structure cristalline de CoAsz.BulL Soc. Frane. M ineral Cristal'l'ogr., 89, 213'215. Dunn, L. (1907) Die Mineralien der Markircher Erzgange. Mitt. Geol. LoniJesonst. Flsass- Lothringen. 6, 183-248. HovorNc, R. D. ,c.r,toL. D. CAr,vrnr (1960) Arsenides of the transition metals, part III' Can. J. Chetn.,38'313-316. Amer' Holws, R. J. (1947) Higher mineral arsenidesof cobalt, nickel, and iron' Geol.Soc' BuIl'.58,299-392 Hulllcrn, F., aNo E. Moosrn (1965) Bond description of semiconductors: polycom- pounds. Prog. Solid State Chem.,2, 33V337. de Jounevsrv, G. (1959) Composition chemique et nomenclature des bi- et triar-seniures colbalt nickel et fet. Maroc Min. Econ. Nat. Direct. Mines GeoL Sent. Geol'. Notes Mem. 18, 161-178. Kerua.m S. (1947) The crystal structure of rammelsbergite, NiA$. (Jni,t. Toronl,oStud" Geol'. Ser., 51, 49-58. Krr-r-v, T. K. (1966) Mass absorption coefficients and their relevance in electron-probe microanalysis. Inst. Mi.neral'.Met. Trans., Sec' 8.75r 69-73' Kornlr,, F. (1963) Uber das Auffinden das Nickels und Kobalts in Erzen und uber einen Chathenit von Andreasburg am Harz. Akail. Munich. Sitzurgsber.,l' 39G403' Mrrnnrv, V. I. (1951) X-ray investigations of naturally occuring tri- and diarsenides. Trudy Feilorott- Nauch. Sessi,p' 75-112, (Transl. Nat. Res' Counc. Can' 944, 196l) Nrcrnr,, E. H. (1963) Structural stability of minerals with the pylite, marcasite, arsenopyrite and Iollingite structures. Can. Mineral'., 9, 3ll-321. Par-rcnr, C., H. Bnnuen eNn C, Fnowost- (l9M) System oJ Mineralogy . ' ',7th ed. L, John Wiley and Sons, New York. Pnrcocr, M. A. (1941) On the identification of minerals by means of x-rays. Trans. Ro1-. Soc.Can. 3rtl.Ser. Sec.1V.,35, 105-113. -- (1944) On loellingite and saffiorite (abstr). Roy' Soc.Con' Proc.,38,ll5. A. S. D.cDsoN (1940) On rammelsbergite and parammelsbergite' Amer' Minerol.,25, 561-577. PutsoN, W. B. (1955) Compounds with the marcasite structure. Z' Kristallogr', l2l I M9- 462. on the QunsNor-, J. C. aNn R. D. Hnvnrl{c (1962) Transition metal arsenidesV. A note rhodium-arsenic system and the monoclinic diarsenides of the cobalt Iamily. can. J. Chem., 4O,814-818. R.trcrrrrr, D. (1966) Mineralogical stuilies on sffiorite-loellingite.Ph.D. Thesis, Queen's University, Kingston, Ontario, Canada. Raulrnr-ssnnG, C. F. (1373) Untersuchung einiger naturlichen Arsen-und-Schwefel-ver- bindungen. Z. D eutsch. Geol. Ges., 25, 282-285. RosneooM, E. H. (1963) Co-Fe-Ni diarsenides: compositions and cell dimensions. ,4zrar. M iner al . . 48 . 27 l-299 . SA F F LORI T E-IN ELLI NGI T E 1881

RosnNqvrsr,T. R. (1953)Magnetic ond crystall,ographicstudies on thehighu antimonidesoJ i,ron,cobdt antl nickel. NorgesTek. Univ. NTII TRYKK, Trondheint. SwaNsoN,et al. (1960)Standard X-ray patterns.Nat. Bur. Stand.(U.5.) Ci,rc.539,LO,2l- 1i --t (1966) Standard X-ray difiraction powder patterns. I[al. Bur. Stand,. (U.5.) Mono., 25, (Sect.4), 10-11.

Manuscript re6ei1)eil,December 8, 1967; accepted,Jzr publication, fune 12, 1968.