the colors of sillimanite - mineralogical society of america

American Mineralogist, Volume 67, pages 749-761 ' 1982

The colors of sillimanite

GeoncB R. RossunN

Division of Geological and Planetary Sciences

California Institute of TechnologyPasade na, C alifurnia 9 I 1251

Eownnn S. Gnew nuo W. A. Dot-use

Department of Earth and Space SciencesU niv e rsity of C alifurnia

Los Angeles, California 90024

Abstract

Sillimanite occurs in three colored varieties: yellow, brown, and blue. The yellow color

is characteristic of unaltered single crystals of sillimanite from high-grade metamorphic

rocks and associated pegmatites. Such sillimanites contain up to 1.8 wt.VoFe2O3 and up to

0.3/o Cr2O3. The yellow color is due to Fe3+, or in a few cases' to Cr3*. The ion dominant

in the optical spectrum of sillimanite containing negligible Cr3* is Fe3* in tetrahedral

coordination. However, Mdssbauer data suggest that about 80% of the iron is in octahedral

coordination and only 2O% in tetrahedral coordination. The salient features in the optical

spectrum from Fe3* are absorption bands at 462, 44O, and 412 nm in a, 616, 474, and 438

nm in 7, and 361 nm in c and 7. No evidence for Fe2+ was found. Absorption from Cr3*

occurs near 620 and 423 nn. The brown variety is also formed in high-grade metamorphic

rocks and associated pegmatites; some brown sillimanite is chatoyant from abundant

acicular inclusions oriented parallel to c. Brown sillimanite generally contains I or more

wt.Vo Fe2O3. Optical absorption spectra of brown sillimanite include some of the iron

features characteristic of yellow sillimanite and prominent bands at -452 nm and 542 nm.

These last two bands may be due to incipient exsolution of an iron-rich phase' which

appears to constitute the inclusions. Blue sillimanite has been documented from crustal

xenoliths in basalt at two localities and from alluvial deposits at two other localities.'Blue

sillimanite contains no more than I wt.%o FezOr. The prominent absorption features,

restricted to 7, consist of two bands at 595 and -836 nm. The spectra are similar to those of

blue kyanite. As with kyanite, it is likely that the coloration is the result of intervalence

charge transfer although the intensity ofthe absorption bands is not correlated with either

Fe or Ti content.

Introduction

Sillimanite, AlzSiOs, occurs in three principalcolored varieties-yellow, brown, and blue. Theyellow and brown varieties are widespread whereasthe blue variety is a mineralogical curiosity that hasbeen well-documented from only four localitiesworldwide. Iron, chromium and titanium are themost abundant minor elements in sillimanite and themost likely to be responsible for the color. Grewand Rossman (1976a, b) and Grew (1976) had previ-

'Contribution No. 3655

0003-004)v82l070E-0749$02.00

ously communicated the results of a study of theoptical absorption of a blue and a yellow sillimaniteand observed that the absorption spectra of thesevarieties are fundamentally different. They suggest-ed that the yellow color is largely due to Fe3* intetrahedral coordination' On the other hand, Hilen-ius (1979) reported an absorption spectrum of ayellow sillimanite and concluded that iron in the six-coordinate site was responsible for the color.

In a study of the electron paramagnetic resonanceof Oconee County, South Carolina, sillimanite, LeMarshall et al. (197t) observed signals from twocrystallographically inequivalent Fe3* ions and

749

750

based the interpretation of their data on the assump-tion that Fe3+ occurred in both the six-coordinateand four-coordinate sites. The present study, anextension of our earlier work, attempts to provide asystematic examination of a number of sillimaniteswhich span the range of color variation. It is thepurpose of this study to examine the origin of thesecolors, and to relate the colors to the petrologicenvironment in which sillimanite forms.

Methods

Samples of sillimanite for this study were select-ed from (1) granulite-facies rocks collected in Ant-arctica, (2) crustal xenoliths collected at KilbourneHole, New Mexico, U.S.A. and Bournac, Haute-Loire, France, and from (3) specimens obtainedfrom museum and private collections (Tables l-3).

Compositions of sillimanite were determined oncarbon coated thin sections and isolated grains atthe electron microprobe facilities at UCLA andCaltech. The average, maximum, and minimumvalues for Fe2O3 for each sample (Tables I and 3)are based on analyses at four to six points on threeto six grains in each thin section of the sillimanite-bearing rock (sample 15, Table 3, consists of twograins broken from one piece). The TiO2 contents ofsamples 14 and 19 are based on an analysis at onepoint per sample. The tabulated analyses are aver-ages ofboth referenced literature analyses and newdata. Optical spectra were obtained using methodsdescribed by Rossman (1975). e-Values are in unitsof liters per mole per cm. The intensities presentedin Figure 2 were obtained by constructing a linetangent to local minima on the sides of the absorp-tion band. There are consequently uncertaintiesassociated with the values obtained for bands whichwere shoulders on steeply rising baselines, such asthe bands at 390 nm and 361 nm in a and at 361 nmin 7. Mossbauer spectra were obtained on about 210mg of acid-washed hand-picked grains. The sampledensities were about 0.012 mg 57Fe per cm2. Isomershifts are reported relative to iron in pd.

Yellow sillimanite

Mode of occurrence and composition

We have analyzed samples of yellow sillimanitefrom six localities (Table l). Samples l, 2, and, 4 arefrom granulite-facies rocks in Antarctica and theAdirondack Mountains, New York (Grew, 1980,1981a), The sillimanite in samples I and 2 is assoc!ated with ilmenite, and that in 4, with magnetite.

ROSSMAN ET AL.: COLORS OF SILLIMANITE

Sdople

A1203

Table L Chemical composition and localities of yellow-greensillimanite

I{ax 1.33Fe203 Ave 1.28

Mln 1 .22

Tt02 0 .01

CrZ03 0.003**

0 . 0 2 0 . 0 3 0 , 0 2

0.21* 0 .08 0 .00

6

6 2 . 1 7

0.02 0 .02

0 . 1 s t 0 . 0 0

1 2 3 4 5

- 62 .21 6L .46 6r .74 60 .E4

5102 - 36 .19

0 . 8 50 , 8 s0 . 8 4

- 36 .77 36 .52 37 .19

1 , 8 7 L , 9 7 , 0 . 5 9r . 7 8 t , 7 9 0 . 6 2 t . 4 l1 . 6 3 7 . 7 0 0 . 5 6

Nunber ofpotnts (pEobea n a l y s e e ) 1 0 4 8 9 3

Retnbolt gi l16, Antarct lca (70'2E'S, 72"27,Er. Crew #556 and NltNH1,37011. Froo peg@ttte ln granul l te-facles rocks.Molodezhdaya Stat lon, Antarct ica (67'40rS, 45'50'E), Gtew l l2748,Fron garnet-blot l te quertzofeldepathlc gnelss. Analysls: crew( 1 9 8 0 ) .Kl lbourne Hole, New t4extco, U.S.A. Gtew #76-5-I . FroE crustelxenol l th ln core of beselElc boob. AneLysls: Grew (19E0).Benson Mines, New York, U.S.A. NMN}{ 115586, Froi l nagnet i te-quattz-sLII lMnlte l ron-ore. Anely616: Grew (1980).F o r e f l n g e r P o l n t , A n t a r c t i c e ( 6 7 ' 3 7 ' S , 4 8 ' 0 4 ' E ) . c r e w # 2 1 1 3 , F r o ncoerae - gtained s1l l lEnl. te-orthopyroxene-cordlert te-blot i te 1sapphir lne rock.Brevlg, Nomay. Analyalst Halentus (1979).

Table entr iea ln welght percenL.Analyaeg ete electron microprobe, excepE:* Energy dlsperslve analysls on scannlng electron dcroscope (ED{)** Edlsslon specrrotraphtc adalysls (ESA). Also by ESA on nuuber l :

0 . 0 0 6 2 B e a n d 0 . 0 0 5 2 V .T X-ray f luorescence anelysls: 0.012 V; l t r absent,Abbreviet lons: NMNH: Nat lonal Museun of Netural l l istory,

Soith6onlan Ine! i tut ion.

Sample 1 is part of a cm-sized single crystal from asubconcordant garnetiferous pegmatite lens in gar-net biotite quartzofeldspathic gneiss. Fe2O3 con-tents of the pegmatitic sillimanite from this localityvary from crystal to crystal (0.76 to 1.30 wt.%Fe2O3, Grew, 1980) and the Fe2O3 contents in TableI were obtained on the piece used in the opticalabsorption woqk.

Sample 5 is from granulite-facies rocks of theArchean Napier complex at Forefinger Point in thewestern part of Casey Bay, Enderby Land, Antarc-tica (Grew and Manton, 1979; c/. sample 12 ofbrown sillimanite, below). It was collected from alayer 1.5 m thick of aluminous granulite containingorthopyroxene, sillimanite, cordierite, biotite, andsapphirine; other rock types present at ForefingerPoint are pyroxene-hornblende granulite, garnet-quartz-feldspar gneiss, and charnockitic gneiss. Sil-limanite in sample 6 from Brevig, Norway, is asso-ciated with quartz, K-feldspar, biotite, hematite andmagnetite (Hilenius, 1979). Samples 1,2, 4,5,and6are representative of prismatic, Fe2O3-bearing, silli-manite that is found in many high-grade metamor-phic terrains.

Sample 3 is from a sillimanite-quartz-K-feld-


spar-ilmenite-magnetite gneiss xenolith containingabundant glass, which was found among the ejectaat Kilbourne Hole maar, New Mexico (Grew, 1979,1980). Xenoliths of quartzofeldspathic gneiss suchas sample 3 and the samples containing blue silli-manite (see below) may be derived from a granulite-facies complex in the lower third of the earth's crustunder Kilbourne Hole (Padovani and Carter,1977a). During their incorporation in the basalt hostand transport to the earth's surface, the xenolithshave been partially fused either by heating or de-compression, and subsequently quenched (Pado-vani and Carter, 1977b; Grew, 1979).

Optical spectra

Millimeter thick plates of sample I have thepleochroic formula,2g : yellow, I: green-yellow,and Z: colorless. Its optical absorption spectrum(Fig. 1) is representative of yellow sillimanitesexcept sample 5. The spectrum is rich in structure;salient features are absorption bands at 462, 440,and 412 nm in c and 616, 474, and 438 nm in 7, andan additional band at 361 nm, prominent in a and 7,but off scale in Figure 1. Bands in the 700 to 900 nmregion are weak or absent. An absorption feature ispresent at 1027 nm in the a and y spectra of all thicksamples examined. Its width is much narrower thaniron features which occur in this region in thespectra of other minerals. Because its intensity isproportional to the thickness of the sample but notproportional to iron content, it is tentatively as-cribed to a vibrational overtone of sillimanite. The

S i l I i m o n i f eR e i n b o l t H i l l s

, . ^ , l / \ :r , i - \

i tIII

\\ c\ . .\ -"... --

p - - - - -

400 600 800 roooWovelengf h (nm)

Fig. 1. Optical absorption spectrum of yellow sillimanite fromReinbolt Hills (#1). Thickness: 5.0 mm, Ella = a = dots; Ellb = F= dash; Ellc =

" : solid line.

depth of color and the intensities of all other absorp-tion bands, except the 616 nm 7 band (the dominantcause ofabsorption in the 620 nm region), generallyincrease with iron content (Fig. 2). Consequently,we attribute all the absorption bands except 616 nmto iron. The absorption of sample 3 is anomalouslylow at most wavelengths. We doubt if this is due tocompositional inhomogeneity because it was ob-served in multiple optical and chemical analyses offragments of sample 3.

The intensity of the -616 nm 7 band variesamong the different samples. The band is lessintense relative to the other features in the spectrumof sample a Gig. 3). Although its intensity is notcorrelated with iron content, its intensity behaviorin Fig. 2 suggests that it still could be related to iron'In sample 3, where the iron absorption bands,notably the 361 nm (y) band, have intensity belowthe general trend, the 616 nm band is prominent' Apossible interpretation of this behavior is that the616 nm band represents iron in a different environ-ment in sillimanite. The origin of this band has yet

to be established. This band imparts a greenish tintto the otherwise yellow crystals. Chromium alsocauses absorption in this region, which we willdiscuss below.

Stoichiometry (Grew, 1980), Mossbauer spectra(see below), absorption band positions and intensi-ties indicate that the oxidation state of iron is *3.The intensities of the prominent absorptions ex-pressed as absorption coefficients (e-values), calcu-lated for sample #l under the assumption that all of

# 5

SILLIMANITE(Ye l low)

o.4 0.8 t . ? 1 .6 2 0

wt % FerO.

Fig. 2. Correlation of the intensities of the optical absorption

bands of yellow sillimanites with average Fe2O3 concentration.Bands at 361, 390, 410, and 440 nm in a and at 390' 410' and 440

nm in 7 follow trends which generally parallel these illustrated

for 361 nm (y) and 460 (nm (a). Absorption in the 620 nm region

includes both the 616 nm and and the Ct'* bands.

@

c

!

E

a zoo

co

!

752 ROSSMAN ET AL,: COLORS OF SILLIMANITE

U(-) o.zz

c0uoam a l

l^IAVENUI'1BER, cm- 130aao 2AAAO 1500a

SILL IMAN]TE

B e n s o n M r n e s

300 400 s00 600 700 8a8t^ /AVELENGTH, nm

904 taoo

Fig. 3. Optical absorption spectrum of yellow sillimanite fromBenson Mines (#4), showing the near absence of the 620 nmband. Thickness; 400 pm; a : broken line: y = solid line.

the iron contributes to each peak, are: 4l2nm(a) e= 3.9;449 nm (a) € : 3.0;361 nm (y) e - 3.8. Theseintensities, low for most metal ions in crystals areconsistent for spin forbidden Fe3* transitions. If theFe3* were present in more than one site, these e-values would be lower limits for one of the sites.

Absorption features of Fe2+ are expected in the800-1200 nm region with an e greater than 3. Fromthe intensity of the absorption of sample #l in the850 nm region in the B spectrum, an upper limit forthe Fez+ content can be calculated assuming € : 3.At most, the Fe2+ content could be 0.012%, FeOcorespondingto 0.9Vo of the total iron. The absorp-tion features at 826 and 1205 nm reported byHilenius (1979) and associated with Fe2+ were notobserved in any ofour spectra, including that ofhissample (#6).

Assignment of the Fe3+ to a particular site in thecrystal structure of sillimanite is more difficult.Both a tetrahedral and an octahedral Al site (Burn-ham, 1963a) are available for Fe3+ substitution. Thecomplexity of the spectrum suggests Fe3* substitu-tion on both sites. Moreover, the spectrum differsfrom the patterns for minerals containing Fe3* inonly octahedral coordination, such as green kya-nite, andradite, and chrysoberyl. Kyanite is an idealcomparison standard because aluminum occupiesonly sites of six coordination with average Al-Obond distances comparable to those of sillimanite(Burnham, 1963b). Stoichiometry, electron spinresonance data (Hutton and Troup, 1964), andoptical spectra indicate that trivalent cations, in-cluding Fe3+, substitute in the aluminum sites. The

absorption pattern of Fe3+ in kyanite consists ofbroad bands at 1075 (0.35) and 610 nm (0.6) andpairs of sharp bands at 444 and 437 nm (1.6) and 377and 370 nm (-3.5) (White and White, 1967; Fayeand Nickel, 1969). The e values, which we deter-mined for a Brazllian kyanite, follow the wave-lengths in parentheses. The 1075 nm band occurs atan energy unusually low for Fe3*, a direct conse-quence of the substitution of Fe3* into an Al3* siteof smaller average metal-oxygen bond distance thanis normally found for Fe3*. The absence of such aband at lower energy in the sillimanite spectrumindicates the absence of a major spectroscopiccontribution from Fe3* in the six-coordinate site.However, the low intensity of Fe3* in the 1075 nmregion of kyanite means that in the spectrum ofsillimanite #1, nearly 0.7 wt.Vo octahedral Fe3+would have to be present to be reliably detected bya band in the 1000 nm region. If this much octahe-dral Fe3+ were present, the kyanite spectral modelpredicts that the contribution from octahedral Fe3*would only be as intense as the B spectrum inFigure 1. Because the sillimanite a and 7 spectra aremuch more intense than the B spectrum, the moreintense features in the sillimanite spectrum are notdue to octahedral Fe3+.

Experimental spectroscopic standards for tetra-hedral Fe3+ are less well established than those foroctahedral Fe3*. Orthoclase: Fe3* (Faye, 1969) andAIPOa: Fe3* (Lehmann, 1970) are relevanr sran-dards because they contain low concentrations ofFe3+ in a tetrahedral site and have spectra whichhave several similarities to the sillimanite spectra.Because the positions of the stronger absorptionbands in the 350-500 nm region are in the samerange as those of octahedral Fe3*, site occupancycannot be readily determined by a wavelength crite-rion. The e values of tetrahedral Fe3* bands in the300-500 nm region can be as much as an order ofmagnitude higher than those of octahedral Fe3*.The intensity of the sillimanite pattern can beexplained if a portion of the iron is in the tetrahedralsite.

Consequently, we propose that: (1) the moreintense features in a and 7 are due to Fe3* in thetetrahedral site; (2) the weaker features could bedue to Fe3* in either the octahedral site or thetetrahedral site; (3) tetrahedral Fe3* is the spectro-scopically dominant ion and primary cause of thecolor in the crystal; and (4) even if most of the ironin sillimanite were in octahedral coordination, itscontribution to the color would be minor.


Miissbauer spectral analysis

Mcissbauer spectra are consistent with the pres-ence of Fe3* in two sites. Spectra were obtained onthe two samples richest in iron, numbers 3 and 4.Although a long counting time was employed, withbackground counts of over 6 x 106 counts perchannel, the spectrum (Fig. a) of sample #4 is weakdue to its low iron content. It consists of two mainlines; a more intense lower velocity peak and a lessintense high velocity peak. When a doublet con-strained to have equal area lines but independentlinewidths is matched to this spectrum, a statistical-ly satisfactory fit is obtained (rms chi value 1.09).The asymmetric doublet would have an isomer shift(LS.) of 0.18 mm/sec and a quadrupole splitting(Q.S.) of 1.04 mm/sec. The higher velocity, broaderpeak has an apparent width of 0.81 mm/sec and thelower velocity peak is narrower with an apparentwidth of 0.58 mm/sec. The isomer shift is somewhatlow but not inconsistent with octahedrally coordi-nated ferric iron. The quadrupole splitting is a littlehigher than usually observed in such compounds.The apparent line widths and asymmetry would beanomalous for a high-iron content mineral, butcould be rationalized as due to relaxation broaden-ing (Wignall,1966) brought about by the long Fe-Fedistances in such a low-iron content phase.

There are several sources of dissatisfaction withthis fit. The two-line fit does not account for ananomalous absorption area between the two fittedpeaks and the correspondence in shape between theobserved and fitted peaks is poor. The observedpeaks both seem to be more intense and narrowerthan their best-fit approximations. Because the ap-pearance of the two-peak fit suggests that a second,weaker doublet was present, a two doublet fit wasattempted. With two-doublets, the fit is improved(rms chi : 1.02), but this improvement is notsignificant compared to the large statistical uncer-tainty present in this weak spectrum. The two-doublet fit is reported because it is both consistentwith the observed spectrum and is thought to leadto an interpretation which is more realistic in viewof the known crystal chemistry of sillimanite.

In order to minimize the number of fitted varia-bles and ensure convergence ofthe fit, the doubletswere constrained to be symmetric, i.e., each halfhaving the same area, and all four peaks having thesame width. In addition, as the low velocity peaksof each doublet are essentially completely over-lapped, the location of the weaker low-velocity

oooozo

I ooolEof; oooz

F ooo3z

fr oooa

Eooo5

753

V E L O C I T Y ( m m p e r s e c o n d )- 1 0 1

Fig. 4. Mdssbauer spectrum of yellow sillimanite from Benson

Mines (#4) fitted with two constrained doublets. Velocities aregiven in mm/sec relative to Fe in Pd' Absorbance is given as the

fraction of the background count.

peak was held fixed at the location of the envelopemaximum. As a result of applying these constraints,the two-doublet fit has only one more free variablethan the one-doublet fit.

The constrained two-doublet fit (Fig. 4) con-verged yielding the following parameters:

outer doublet: 79(!5)% of area,I .S. : 0.19(r.02) mm/sec,

Q.S. = l . l l (1 .03) mm/sec .

inner doublet: 2l(t5)% of area,I .S. : _0.031-rg.5y mm/sec.

Q.S. : 0 .53( i - l ) mm/sec .

The major doublet in either the one or two-doublet fit has an isomer shift appropriate for Fe3*in octahedral coordination. In analogous minerals,ferric ions in octahedral sites give I.S. values ofabout 0.201.03 mm/sec. The weaker inner doubletof the two-doublet fit has a smaller isomer shift ofca. -0.03 mm./sec, which is in the range expectedfor ferric iron in tetrahedral coordination. Analo-gous silicates and oxides with tetrahedrally coordi-nated ferric ions give I.S. values of ca. 0'01-f .05mm/sec.

The quadrupole splitting of ferric doublets gener-ally increases with increasing distortion of the Fesite. The Q.S. of the major, or octahedral, doublet,at ca. l.l mm/sec is moderately large suggesting amoderately distorted site. For cornparison, octahe-dral Fe3+ in pyroxeneso garnets, amphiboles and

micas give Q.S. values of about 0.4 to 0'6 mm/sec,and in such very distorted sites as the M(3) site inepidote, the Q.S. is 2.0 mm/sec or higher. The Q.S.of the weaker, or tetrahedral, doublet is smaller.However, it is difficult to assess the significance ofthis value, as the Q.S. values reported for the fewtetrahedrally coordinated cases studied range from

754 ROSSMAN ET AL.: COLORS OF SILLIMANITE

near zero for some spinels to about 1.5 mm/sec forsome pyroxenes.

The fitted linewidth in the two-doublet case,0.441-+.921 mm/sec, is significantly narrower andprobably closer to reality than those given by theone-doublet fit. However, the true linewidths areprobably all unequal considering the low iron con-tent of this mineral. Meaningful refinement of sepa-rate linewidths is precluded by the quality of thespectrum.

The Mrissbauer spectrum of sample 3 is too weakand ill-defined to be of quantitative value. Thisspectrum (two mirror image halves) was recordedover 1024 channels with a velocity increment of0.02 mm/sec-channel. A background count of 2.45x 106 counts per channel was accumulated. Themaximum dip was only 0.003 of background and theRuby "S" parameter which is a measure of theinformation content of the spectrum, is only about1/3rd the value of the spectrum of sample 4. A one-doublet fit results in Mossbauer parameters thatagree with those of sample 4 within the largeexperimental error. There is no indication of ferrousiron. The low quality of this spectrum precludesfitting separate octahedral and tetrahedral doublets.

It is not apparent why sample 3 shows distinctlylower resonant absorption. The optical absorptionfeatures we attribute to iron, except the 616 nm 7band, are also anomalously low in this sample (seeabove). This may possibly reflect a more homoge-neous iron atom distribution resulting in more iso-lated iron atoms and therefore more severe relaxa-tional broadening of the M<issbauer spectrum. Amore random distribution of iron in this sillimanitefrom Kilbourne Hole is also suggested by the highlydisordered Al-Si distribution in K-feldspar of theKilbourne Hole xenoliths (Grew, 1979).

During the course of this study Hilenius (1979)published Mrissbauer spectra of a sillimanite fromBenson Mines, New York, the same locality as thatfor sample #4, and of a Norwegian sillimanite (#6)whose optical spectra we have also studied. Hisspectra not only show greater dips and smallerstatistical noise than ours, which is consistent withhis use of a stronger source and thicker samples,but also his conclusions differ from ours. Hflleniusfitted his sillimanite spectra with three symmetricaldoublets with the strongest feature being a Fe3*doublet at I .S. : 0.19(t) , Q.S. : 1.07(1) andlinewidth : 0.63(1) mm/sec. These values are es-sentially identical with the octahedral Fe3+ doubletwe observed except for the broadened linewidth.

The remaining doublets have been assigned byHilenius to an Fe2+ component in sillimanite and ahematite impurity phase present in his sample. TheBrevig, Norway, sillimanite in particular had aprominent Fe2* component in its Mcissbauer spec-trum (Hilenius, 1979). However, we observed noprominent Fe2* features in the optical spectra of hisBrevig, Norway, sample, and Fe2* or hematitefeatures could not be detected in our Mcissbauerspectra. The Fe2+ features in Hilenius' spectra maybe due to another contaminant phase. These con-taminant phases are not observable in our silliman-ite samples, which were carefully picked free ofvisible impurities by hand.

The conclusions from our combined Mossbauerand optical spectra are: (1) iron in yellow sillimaniteoccurs in the ferric oxidation state: (2) about 4/5thsof this ferric iron is in octahedral sites; (3) theremaining 1/5th is in tetrahedral coordination-nodoubt substituting for the trivalent aluminum pres-ent on both these sites; (4) the yellow color isprimarily caused by the tetrahedral Fe3*; (5) the 616nm band, when present, imparts a green componentto the yellow color; (6) the prominent absorptionbands in the spectrum of yellow sillimanite arepresent in the spectra of nearly all sillimanitesincluding these for brown and blue sillimanite dis-cussed below.

Heating experiments

To determine whether the proposed dual octahe-dral-tetrahedral Fe3* occupancy is temperature de-pendent, samples 1, 3 and 4 were heated undervarious conditions at temperatures from 500 to1430" C for up to 130 hours under air, charcoal and

fOz = 10-r0. Essentially no changes in the color orspectra occurred up to 1050" C. In the 1430'Cexperiment, sample I lost the trace OH- evident inthe infrared spectrum, became translucent, butshowed essentially no change in the Fe3* spectrum.

Other cations

Chromium is an important minor element in sam-ples 2 and 5. In the spectrum of sample 5, absorp-tion bands not due to Fe3+ occur at -62A,523 and423 nm (Fig. 5). The bands at620 and 423 nm are atwavelengths similar to those from the spectrum ofCr3* in emerald which is in a six-coordinated site.Their approximate e values in 7 based on Cr3* are:620 nm e : 5; and 423 nm e : 15. The cause of the523 nm band has not been established.

ROSSMAN ET AL.: COLORS OF SILLIMANITE 7s5

Brown sillimanite


We examined samples of brown sillimanite fromseven localities (Table 2). Six of the sillimanites arefrom quartz veins, and the seventh, from a pegma-tite. The prisms of brown sillimanite we examinedrange from less than 1 mm to over 3 cm in width.

The sillimanite-bearing quartz most likely occursas veins in mica schist or gneiss. The sillimanite isassociated with ilmenite, a hematite-ilmenite inter-growth, or rutile, as well as biotite, muscovite, orminor secondary sericite. Sample 10 from OconeeCounty, South Carolina, is a loose crystal. In F. H.Pough's samples from this locality, brown silliman-ite occurs with ilmenite, biotite and muscovite inqvartz. This paragenesis is the same as that ofsillimanite from the type locality in Saybrook, Con-necticut (Bowen, 1824). Although our samples ap-pear to be found only in quartz, sillimanite atNorwich (presumably also brown) also occurs inmica schist (Schairer, 1931, p. ll2).

The Connecticut localities are in the sillimaniteand sillimanite-K-feldspar zones of regional meta-morphism, roughly equivalent to the upper amphib-olite facies and possibly the hornblende granulitefacies (Thompson and Norton, 1968). The OconeeCounty sillimanite is most likely derived from am-phibolite facies metamorphic rocks, which underliethis part of South Carolina (Overstreet and Bell,1965). As we have no information on the exactlocality for the Sardinian sample, we do not knowthe grade of metamorphism of the source terrain forthis sample.

The pegmatitic sillimanite (#12) is from "Christ-

I" IFVENUMBEB < cu-t >30000 20000 15000

S I L L I M R N I I EFOREFINC'ER POINT

q00 500 600 ?00 800NFVELENGTH <Hll>

Fig. 5. Optical spectrum of yellow sillimanite containingchromium from Forefinger Point (#5). Thickness : 3.0 mm; a =

broken line, y = solid line.

Table 2. Fe2O3 contents and localities for brown sillimanite

S a o p l e 7 8 9 1 0 1 1

F e 2 0 3 l . , 0 2 1 . 2 0 I ' 1 2 1 . 3 3 0 . 9 8

L2 13

1 . 3 5 1 . 3 0

mas Point" (informal name), a part of an unnamedisland near Mclntrye Island in the eastern part ofCasey Bay, Enderby Land, Antarctica. This peg-matite cuts granulite-facies rocks of the ArcheanNapier complex, the same complex from which theForefinger Point sillimanite (#51 originated.Coarse-grained portions of the pegmatite consistlargely of quartz, microcline (in part red), biotite,apatite and sillimanite; and medium-grained por-tions of surinamite, cordierite, garnet, apatite, py-rite, orthopyroxene, wagnerite, ilmeno-hematite'and sillimanite (Grew, 1981b). Sillimanite, some ofwhich is dark brown and chatoyant, forms singlecrystals up to 10 cm long and 3 cm across. Thispegmatite and the enclosing granulite-facies rocksconstitute a block of relatively unaltered rock in atectonic zone (possibly of early Paleozoic age) ofintense deformation, retrogression of the granulite-facies rocks under amphibolite facies conditions,and of emplacement of pegmatite veins' In general,chatoyancy and brown color are characteristic ofsome sillimanites in relatively unaltered masses ofrock exposed in Casey Bay and intersected by thesetectonic zones. Away from these zones' EnderbyLand sillimanite is gray, white, or pale yellow (e.9.,

#5, Forefinger Point).In addition to the above examples, patches pleo-

chroic in brown (maximum absorption llc) havebeen reported in thin sections of sillimanite ingranulite from Wilson Lake, Labrador, Canada(Leong and Moore, 1972, p.787 Gtew,1980). Thissillimanite is associated with magnetite and titanif-erous hematite, plagioclase, biotite, sapphirine, andorthopyroxene. Metamorphism was under condi-tions of the granulite-facies (Leong and Moore,1972; Morse and Talley, l97l)' Yrdna (1979, p' 26)reports "patches of light brown coloration . . .

r0.I I .

Tdble entr lee tn wetght percent, by energy dispersLve analyses.wl l l iMnt ic, Connect icut, U.S.A. NMNH C3019-2. With he@tl!e-l lnenlte and raae blot l te ln quattz.vernon, Connect icut, U.S.A. NMNH 814835. l , l l th blot i te, rut l le,and heno-i1@nite in quartz.Gul l ford(?), Contrect icut, U.S.A. l tarvard Unlv ' ColI . E5919.In quartz.oconee county, South carol lna, U.S.A. NMNH R11616'Sardlnta(?). N}!NH R3741. In quartz."ChrlstMs Polnt", on an unmned 131and 5 kD WSW of MclntyreIetand (67" 22t5, 49' 05' t , l ) , Antarct lca, ctee 112292D. Inp e g @ t l t e .Nomich, Connect lcut, U.S.A. Harvard Univ ' col1. 85918. withblot i te, i loenlte, and oinor ouscovlte ln quartz.

3 o . ezGCDEO

3 o . uG

0 . 0

ROSSMAN ET AL,: COLORS OF SILLIMANITE

passing in their centers into areas of opaque dustyparticles" in sillimanite in a borosilicate rock con-taining magnetite and hemo-ilmenite from Zambia.

Brown sillimanite we have examined typicallyhas numerous fine acicular inclusions, most ofwhich are oriented parallel to [001]. Viewed down[001] brown sillimanite appears pale blue because ofscattering from these inclusions; these inclusionscause the chatoyancy noted in some of our samples.The brown color is not evenly distributed in thecrystal, and is typically most intense in patcheswhere the inclusions are most numerous. The inclu-sions are either opaque or are translucent andbrown; some have a hexagonal outline in a viewperpendicular to [001]. The brown inclusions maybe hematite; the opaque inclusions may be ilmenite,magnetite, or some other iron-rich oxide mineral.

Brown sillimanite analyzed in this study does notcontain significantly less than I weight percentFe2O3 $able 2);the Wilson Lake material contains1.6 to 1.8 wt.Vo (Grew, 1980). Vrdna (1979) reportsl.2wt.Vo Fe2O3 in the colorless portions of crystalshaving brown patches. Thus, the brown color ap-pears only to develop in sillimanite containing atleast 1 wt.Vo Fe2O3.

Optical spectra \The pleochroic formula'in our samples is : X and

Y: colorless to pale yellow in thin section (30 p.m)and orange-brown in thick sections (0.1 mm) and Z: violet brown (see also Dana, 1920, p.498; Win-chell and Winchell, 1951, p. 520).

The spectrum of sample 11 (Fig. 6) is repre-

NRVENUMBEB <cr- t r20000 10000 7000

S I L L I M R N I T ESRRDINIR

q00 ,800 1200r^IRVELENGTH <rulr>

Fig. 6. Optical absorption spectrum of brown sillimanite fromSardinia(?) (#ll) obtained with (100) and (0t0) slabs, 200 trcmthick; a = dots, F = dash; 7 = solid line. Data obtained in aregion of the crystal chosen to minimize the contribution fromscattering.

sentative of brown sillimanite. The broad region ofabsorption in the 7 spectrum extending throughoutthe visible portion of the spectrum displays twomaxima corresponding to absorption bands at -452and 542 nm. The 361 nm (7) sharp Fe3+ absorptionband characteristic of the pattern of yellow silliman-ites and weaker Fe3+ features in a and B, are alsopresent. Other absorption bands, in y, at about 894nm and about 730 nm, appear to be present. Theeffects of scattering can be best observed in the B-spectrum obtained on a (001) slab where the appar-ent absorption rises in a broad band centered near600 nm and decreases near 400 nm to preferentiallytransmit blue (Fig. 7). The inclusions responsiblefor this effect presumably have a dimension compa-rable to the wavelength of visible light. A back-ground absorption, steadily rising toward shorterwave-lengths as a result of scattering from presum-ably larger inclusions is in many cases superim-posed upon the absorption pattern (Fig. 8).

The dominant minor element is iron Oable 2).Based on total iron content. the e value for the 452nm band of sample 11 is 300. This value is too highto be caused solely by Fe3*. An intervalence chargetransfer transition is a likely cause of this absorp-tion, and has been proposed to cause an intenseabsorption in the spectrum of andalusite. The opti-cal spectrum of andalusite shows weak absorptionfrom Fe3* and a strong, dominant absorption bandpolarized parallel to [001] at 481 nm (Faye andHarris, 1969). The 481 nm band is reminiscent of the452 nm and 542 nm absorption bands of sillimanite.Both the andalusite and sillimanite bands are in-

q0000NFVENUMBER <cr t>

10000 7000

S i L L I M R N I T ESFBOINIF

500 1000 I 500I IFVELENGTH rw>

Fig. 7. Optical absorption spectrum of a (001) slab of brownsillimanite from Sardinia(?) (#ll) obtained in a portion of thecrystal which maximizes the spectral dependence of thescattering from the numerous inclusions oriented along [00t],Sample 237 pm thick; p polarization; the a polarization is nearlyidentical,

t n

N A

0 . 8

trJ(Jz . n Ac r - ' -

f

O n r rc D v . a

G

0 . 2

LU

zCE

mG 0 , 2

0 . 00 . 0

ROSSMAN ET AL.: COLORS OF SILLIMANITE 757

I . I f lVENUMBEB <cu-t >20000 10000 ?000

t . 1

0 . 8

0 . ' t

0 . 0

S I L L I M R N I T EC,tJ I LFOBD

q00 800 1200| ' IRVELENGTH ruur

Fig. 8. Optical absorption spectrum of brown sillimanite fromGuilford, Conn. (#9) showing the effects of backgroundscattering. Thickness = 600 g.m; a = broken line; 7 = solid line.

tensely polarized parallel to the direction of thechain of edge-shared octahedra and have intensitiesmuch greater than can be accounted for by single-ion electronic transitions. Faye and Harris haveproposed for andalusite a mechanism involvingintervalence charge transfer betwen Ti3* and Tia*.We do not consider an explanation involving Ti3* tobe likely for sillimanite in view of the relatively highFe2O3 contents of brown sillimanite. Interactionsinvolving Fe2* such as Fe2*/Ti4* and Fe2*/Fe3*are more likely to be the cause. The absence ofFe2* absorption indicates that if such mechanismsare operating, the amount of Fe2+ present is small.The association of the color with regions withinclusions suggests that the color originates fromincipient exsolution. We propose that submicro-scopic regions of an iron-rich phase in the processof exsolving are the source of the color. Thissuggestion is consistent with analytical electronmicroscopic observations that sillimanite #7 con-tains numerous small particles of titaniferous hema-tite ranging from 0.005 to 0.2 pm in dimension(David Veblen, personal communication, 1982).

Blue sillimanite


Samples of blue sillimanite available for study(Table 3) are gem quality single crystals up to 2 cmlong from alluvial deposits at Mogok, Burma and inSri Lanka (Spencer, 1920; Adams and Graham,1926) and as prisms 0.1-5 mm long in xenolithicquartzofeldspathic gneisses at Kilbourne Hole,New Mexico. and Bournac. Haute-Loire, France

(Cailldre and Pobeguin,1972; Padovani and Carter,1973, 1974, 1977a; Grew, 1976, 1977a). Blue silli-manite has also been reported from Kenya (Bank,1974) andfrom Davis Creek, New Zealand (Hattori,1967). No details are available on the Kenya occur-rence. Grew (1977b) found blue kyanite, but no bluesillimanite in the sample studied by Hattori (1967).The blue sillimanite reported in the New Zealandsample is possibly kyanite misidentified as silliman-ite.

To our knowledge, there are no definitive reportsof blue sillimanite in situ in Burma or Sri Lanka,and consequently, no data are available on itsparagenesis. The Bournac paragenesis is similar tothat at Kilbourne Hole (see above). Sillimanite-bearing xenoliths at Bournac may originate from thelower crust (Leyreloup et al., 1977). By analogywith Kilbourne Hole, the Bournac xenoliths haveprobably also been subjected to decompression orheating by the basalt host during the xenolith'stransport to the earth's surface. In contrast to theKilbourne Hole xenoliths, however, glass and or-thopyroxene-spinel intergrowths (replacing garnet)are much less abundant.

Sillimanite in eight xenoliths from KilbourneHole (including the 3 samples in Tables 1 and 3) andin 7 xenoliths from Bournac (including sample 19/20) were analyzed for iron (Grew, 1977a)' The ironcontent, reported as FezOr, of blue sillimanite at

14 15 16 17 lE 19 20 2L 22

AlZ0: 62.20 63.L5 - 62.12 - 62'63 6I '74 62'14 6L'16

S i O r 3 6 . 7 4 3 7 . 4 I - 3 7 ' 0 3 3 E ' 0 0 3 7 ' 3 6 3 7 ' I E

I . D

LTJ(JzCEcoEo<n@CE

hxFe203 Ave

E n

Tt02

C r 2 0 3 0 . I 3 *

Number ofpointe 5 Z

0 . 3 7 0 , 2 50 , 3 5 0 . 2 5 - 0 . 3 20 , 3 5 0 . 2 4

0 . 0 4 0 . 0 1 - 0 . 0 2

0 . 2 6 - 0 . 5 2 0 , 3 50 . 2 r 0 . r 7 0 . 4 6 0 ' 3 50 , 1 8 - 0 . 4 1 0 . 3 5

0,02 0 .09 0 ,08 0 .05

- o , to - 0 .00* 0 .00* 0 .02 0 .05

- 5 - 6 e 1 b 2 2

14. Fron xenol l lh of quert tofeldsPathlc gnelss' Kl lbourne Hole'

Nee Mexlco, U.S.A. Crew #76-5 (col lected by E. R' Pedovani) '

Anri lye16r Grew (1980) '

15. l togol, BurE. l ' tu6' Net. drHlstolre Nature1le M1n6ralo8le(Parts) I21.U4. Cryotal f roo al luvlal dePoslt '

16. t torok. Bur@' Brl ! . I tu6. BM I920,I7. Cryotel f rotr al luvlal

aeiosis at e ruby olne (sPencer, 1920).L7. l togok, BurM. Acr ' ldrs ' Net. HlsEoty 34923. crystal f roE

el luvlal deposlt '18, srt Isnka Et l t . l tus' EM 1945,39. crydtel t roE el luv{el d€Poolt '

19/20. Bournsc, Haute-Llrer France. Mua' l lst . dr l l lstolre Neturel le

l , t lnerelogle (Perts) 129.7. Ihou aenol l th of Sarnet qualtzofelda-

oathlc cnetee ln b68alt lc tuf f (Cst l lere snd Pobeguln, 1972) '

i tnetyet i : Grew (1980); 0.24/ Fe2o3 on e crydtel by energv

d{€perolve analYdlB.ZL/ZZ.

'*o. aenol l th of quertzofeldsPEthlc 8oele6, Kl lbourne Hole,

Nev Meslco' U. s ' A' cres 1176-5-24.

Table entr le6 tn selght Percent. Atrslyae6 by electron olcroProbe_

6ff i Iy81e, excep!:*Eneriy dtsperalve gpectroscopy under Ecennlng electron oictogcope'

ROSSMAN ET AL,: COLORS OF SILLIMANITE

Kilbourne Hole ranges from 0.35 to 0.94 wt.Vo (6samples), and at Bournac, 0.17 to 0.23 wt.Vo (5samples). On the other hand, colorless and yellowsillimanite at Kilbourne Hole contain 0.99 and 1.82wt.Vo, and colorless sillimanite at Bournac, 0.70 and0.87 wt.Vo. Padovani and Carter (1977a) also reportthat the Fe2O3 content of blue sillimanite is lessthan that of colorless sillimanite at Kilbourne Hole.Moreover, at both localities, graphite is found onlyin those xenoliths in which the sillimanite containsless than 0.4 wt.% Fe2O3. Padovani and Carter(1977a) report the close association ofgraphite, bluesillimanite, and purple to blue rutile in the Kil-bourne Hole xenoliths. Ilmenite is present in somexenoliths with blue sillimanite. Recalculated micro-probe analysis (from stoichiometry) of two suchilmenites from Kilbourne Hole indicate Fe2O3 con-tents of 1 moleVo or less (samples 76-5 and 76-5-9,Grew, 1977a), which is less than in ilmenite associ-ated with yellow sillimanite (e.g., Table l, andGrew, 1980). These relations suggest that develop-ment of the blue color is associated with low Fe2O3contents and may be promoted by a reducing envi-ronment.

The xenolithic sillimanites are typically chatoy-ant from fine acicular inclusions oriented parallel to[001], a feature noted in some crystals from SriLanka (Spencer, 1920). Such chatoyancy is notlimited to these blue sillimanites. Some colorlesssillimanite we have examined from the exposedgranulite-facies terrain in Enderby Land, Antarcti-ca, is also chatoyant and pale blue as a result ofscattering by these fine inclusions. However, thissillimanite lacks the absorption spectra characteris-tic of the blue sillimanite descibed below.

The relative abundance of blue sillimanite in thexenoliths and its apparent absence in exposed meta-morphic terrains suggests that interaction betweenthe xenoliths and basalt host during their incorpo-ration in the host or during transport to the earth'ssurface, or rapid quenching following transport mayplay a role in developing the blue color. Theseprocesses have evidently led to the nearby com-plete Al-Si disorder in associated K-feldspar (Grew,1979). The alluvial blue sillimanite most probably isderived from nearby exposed granulite-facies ter-rains; the origin of this material remains an enigma.

Optical spectra

The pleochroic scheme for blue sillimanite is Xand I : colorless to pale yellow and Z: blue (seealso Spencer, 1920; Winchell and Winchell, l95l).

The absorption spectra of all blue samples aresimilar to one another, and are distinct from thespectra obtained of yellow and brown sillimanite.The pattern of a sillimanite from Sri Lanka (Ceylon)is representative of blue sillimanite (Fig. 9).

The prominent absorption features are two bandspolarized llc (7) centered at 595 and -836 nm. Thereis little absorption in the a and B directions. Weakabsorption features in the 350-500 nm region thatcorrespond to the Fe3+ features in the spectra ofyellow sillimanite are present in a, B, and 7. Theintensities of these weak features do not correlatewith the intensity of the two strong features. Theintensities of the 595 and 836 nm features arecorrelated (Fig. 10) and consequentially are as-sumed to arise from the same source.

Chemical analyses of the blue sillimanites (Table3) indicate that Fe and Ti are the major traceelements. The intensity of the 595 nm feature is notcorrelated with either the iron or the titaniumcontent.

Relationship to blue lcyanite

Several similarities between the spectra of bluekyanite and blue sillimanite are evident: (l) absorp-tion bands occur at nearly the same wavelengths,(2) the lower energy band is about % as intense asthe high energy band, and (3) both absorption bandsare polarized along the octahedral chain axis. Thesesimilarities suggest that the blue color in bothminerals has the same origin.

The origin of the blue color in kyanite has beenaddressed several times (White and White, 1967;

NFVENUMBER <or-t >20000 10000 7000

S I L L I M F N ] T ECEYLON

q00 800 1200NRVELENGTH <mr>

Fig. 9. Optical absorption spectrum of blue sillimanite (#18).Thickness = 6.0 mm; a = dots, y = dash. The p-spectrum (notillustrated) is nearly identical to the a-spectrum.

0 . 8

u . l f ,L!(Jz.CE@ n r lr " ' =ocoC E ^ ^

u . z

0 . 0


I N T E N S f T Y A T 5 9 5 n m ( A b s , / c m )

Fig. 10. Correlation of the 595 nm and 836 nm absorptionbands of blue sillimanite. The numbers designating dots aresample numbers from Table 3.

Faye and Nickel, 1969; Faye, 1971; Smith andStrens, 1976; Parkin, et al., 1977). All authorsrecognize the likelihood that intervalence chargetransfer absorption is the process responsible forthe color. Two broad, overlapping absorptions areprominent in the absorption spectrum of blue kya-nite" one at -600 nm and weaker one at -850 nm.They are generally believed to arise from eitherFe2*-Ti4* or Fe2+-Fe3* intervalence charge trans-fer.

Iron and titanium are present in blue sillimanite,but both are also present in yellow sillimanite. Thecolor-producing mechanisms proposed for kyaniteinvolving trace amounts of Fez* are also appropri-ate for blue sillimanite in view of its associationwith reducing environments. The amount of Fe2* insillimanite must be small because the characteristicpair of Fe2* absorption bands has not been ob-served in the sillimanite spectrum. The details ofthe mechanism by which Fe2* would produce thecolor are not established: it is not known whetherthe color-causing ions are in true solid solution orare associated the incipient exsolution of inclusionssuch as rutile.

Conclusions

Yellow and brown appear to be the characteristiccolors of sillimanite in high-grade metamorphicrocks. Sillimanite contains a small amount of Fe2O3

(up to 1.8 wt.%) and Fe3* is the primary cause ofthe color, as absorption increases with Fe2O3 con-tent. A variety of experiments suggest that the Fe3*occurs in both octahedral and tetrahedral sites. Afew sillimanites contain a small amount of Cr3* (upto 0.3 wt.Vo CrzOt in Antarctic sillimanite, Grew, inpreparation) which contributes a green hue to theyellow in such samples.

Brown color in sillimanite appears to be related toincipient exsolution of iron-titanium oxides fromsillimanite. It is restricted to samples relatively richin Fe3* (>l wl% FezO:) which generally originatefrom relatively oxidized rocks. Most brown silli-manite contains acicular inclusions of opaque ordark brown material, presumably an Fe or Fe-Tioxide. In contrast to yellow sillimanite, the intensi-ty of color does not increase proportionally to theFe content.

Sillimanite from xenoliths of granulite and gneissincorporated in basaltic rocks differs from the silli-manite of exposed high-grade terrains. In the onlyyellow xenolithic sample we had, the optical ab-sorption intensity and Mossbauer spectrum aremuch weaker than those of a non-xenolithic silli-manite containing the equivalent amount of iron.Possibly the iron in the xenolithic sillimanite isdistributed more evenly throughout the crystal thanin the non-xenolithic sillimanite. This more evendistribution may be related to the thermal history ofthe xenolithic samples. The xenoliths were heatedto magmatic temperatures and rapidly quenched tothe ambient temperature at the Earth's surface. K-feldspar in a xenolith at the same locality (Kil-bourne Hole) has an unusually high degree of Al-Sidisorder, one of the highest reported in the litera-ture (Grew, 1979). This rapid quenching may havepreserved the highly diffused Fe3* distribution in-duced at high temperatures. In the sillimanite fromexposed metamorphic terrains, less rapid coolingallowed the Fe3* to migrate and cluster to a limitedextent.

Xenolithic sillimanite containing less than I wt.%Fe2O3 is blue in some cases. Rapid quenching fromhigh temperatures under reducing conditions (i.e.,low FezOr) may be instrumental in developing theblue color in sillimanite. However, this mechanismdoes not explain the blue color of alluvial gemquality sillimanite from Sri Lanka and Burma. Nordoes it account for the related blue color in kyanite,the characteristic color of kyanite from exposedregionally metamorphosed terrains, which musthave cooled slowly.

2 4 . 4E

U(n_0v 3 . O

€L

(oa 2 .O

F

- t a- t - aHazulF

7 a.o1 0a

I ]r8 r6

aaa

7ffi ROSSMAN ET AL.: COLORS OF SILLIMANITE

AcknowledgmentsGrew thanks members of the lSth and lgth Soviet Antarctic

Expeditions (197 3 J 4) and of the 197 7 -:l 8 and I 979-80 Au stralianNational Antarctica Research Expeditions for providing logisticsupport for field work in Antarctica. Grew also thanks p. C.Grew for assistance in collecting xenoliths at Kilbourne Hole andBournac.

We acknowledge the following people and institutions for theirgenerosity in providing us with samples for this study: C.Frondel-Harvard University; J. S. White, Jr.-The U. S. Na-tional Museum (Smithsonian Institution): J. P. Fuller-The Brit-ish Museum; J. Fabrids-The Mus6um d'Histoire Naturelle(Min6ralogie); G. E. Harlow-the American Museum of NaturalHistory; U. H&lenius, E. R. Padovani and F. H. pough. DavidVeblen provided electron microscopy results; K. Langer andR. G. Bums provided helpful discussion of our results.

This work was funded in part by National Science Foundationgrants DPP 75-17390 and DPP 76-80957 to UCLA, and EAR 79-M80l to Caltech; electron microscopy studies were funded bygrant EAR 8l-15790 to Arizona State University.

ReferencesAdams, F. D. and Graham, R. P. D. (1926) On some minerals

from the ruby mining district of Mogok, Upper Burma. Trans-actions of the Royal Society Canada 20, ll3-136.

Bank, H. (1974) Durchsichtiger farbloser bis bliiulicher Sillimanitaus Kenya. Zeitschrift deutschen Gemmologischen Gesell-schaft, 23, 281-282.

Bowen, G. T. (1824) Description and analysis of the sillimanite,a new mineral. American Journal of Science. 8. ll3-118.

Burnham, C. W. (1963a) Refinement of the crystal structure ofsillimanite. Zeitschrift fiir Kristallographie, I | 8, 127 -148.

Burnham, C. W. (1963b) Refinement of the crystal structure ofkyanite. Zeitschrift fiir Kristallographie, I 18, 337-360.

Cailldre, S., and Pobeguin,T. (1972) Sur la pr€sence de silliman-ite i Bournac (Haute-Loire). Bulletin de la Soci€t6 franEaisede Min6ralogie et de Cristallographie, 95, 4ll.

Dana, E. S. (1920) Descriptive Mineralogy, John Wiley & Sons,New York.

Faye, G. H. (1969) The optical absorption spectrum oftetrahe-drally bonded Fe3* in orthoclase. Canadian Mineralogist, 10,tt2-117.

Faye, G. H. (1971) On the optical spectra of di- and trivalent ironin corundum: a discussion. Canadian Mineralogist, 56, 344-348.

Faye, G. H. and Harris, D. C. (1969) On the origin of color andpleochroism in andalusite from Brazil. Canadian Mineralogist,10, 47-56.

Faye, G. H. and Nickel, E. H. (t969) On the origin of color andpleochroism of kyanite, American Mineralogist, 10, 35-46.

Grew, E. S. (1976) Blue sillimanite (abstr.). Transactions of theAmerican Geophysical Union, 57, 1019.

Grew, E. S. (1977a) Blue sillimanite and the origin of crustalxenoliths at Kilbourne Hole, New Mexico, and Bournac,Haute-Loire, France (abstr.). Extended Abstracts, contribut-ed to the Second International Kimberlite Conference, SantaFe, New Mexico, Oct. 3-:7, 1977.

Grew, E. S. (1977b) First report of kyanite-sillimanite associa-tion in New Zealand (Note). New Zealand Journal of Geologyand Geophysic s, 20, 797 -802.

Grew, E. S. (1979) Al-Si disorder of K-feldspar in crustal

xenoliths at Kilbourne Hole, New Mexico. American Mineral-ogist, 64, 912-916.

Grew, E. S. (1980) Sillimanite and ilmenite from high-grademetamorphic rocks of Antarctica and other areas. Journal ofPetrology, 21, 39-68.

Grew, E. S. (1981a) Granulite-facies metamorphism at Molo-dezhnaya Station, East Antarctica. Journal of Petrology, 22,297-336.

Grew, E. S. (l98lb) Surinamite, taaffeite, and beryllian sapphi-rine from pegmatites in granulite-facies rocks of Casey Bay,Enderby Land, Antarctica. American Mineralogist, 66, 1022-1033.

Grew, E. S. and Manton, W. I. (1979) Archean rocks in Antarc-tica: 2.5-billion-year uranium-lead ages of pegmatites in En-derby Land. Science, 206,443-445.

Grew. E. S. and Rossman. G. R. (1976a) Iron in some Antarcticsillimanites (abstr.). Transactions of the American Geophysi-cal Union, 57, 137.

Grew, E. S. and Rossman, G. R. (1976b) Color and iron insillimanite (abstr.). Abstracts, 25th International GeologicalCongress, 2, 5G-565.

HAlenius, U. (1979) State and location of iron in sillimanite.Neues Jahrbuch fiir Mineralogie Monatshefte, 165-174.

Hattori, H. (1967) Occurrence of sillimanite-garnet-biotitegneisses and their significance in metamorphic zoning in theSouth Island, New Zealand. New Zealand Journal ofGeologyand Geophysics, 10, 269-299.

Hutton, D. R. and Troup, G. J. (1964) Paramagnetic resonanceof Fe3* in kyanite. British Journal of Applied Physics, 15,1493-1499.

Lehman, G. (1970) Ligand field and charge transfer spectra ofFe(III)-O complexes. Zeitschrift fiir Physikalische ChemieNeue Folge, 72, 279-297.

LeMarshall, J., Huton, D. R., Troup, G. J., and Thyer, J. R. W.(1971) Paramagnetic resonance study of Cr3* and Fe3+ insilfimanite. Physics Status Solidi A., 7 69-:773.

Leong, K. M. and Moore, J. M., Jr. (1972) Sapphirine-bearingrocks from Wilson Lake, Labrador. Canadian Mineralogist,r l ,777J90.

Leyreloup, A., Dupuy, C., and Andriambololona, R. (1977)Catazonal xenoliths in French Neogene rocks: cpnstitution ofthe lower crust. 2. Chemical composition and consequences ofthe evolution of the French Massif Central Precambrian crust.Contributions to Mineralogy and Petrology 62,238-1W.

Morse, S. A. and Talley, J. H. (1971) Sapphirine reactions indeep-seated granulites near Wilson Lake, Central Labrador,Canada. Earth and Planetary Science Letters, 10,325-128.

Overstreet, W. C. and Bell, H., III (1965) The crystalline rocksof South Carolina. U. S. Geological Survey Bulletin llE3.

Padovani, E. R. and Carter, J. L. (1973) Mineralogy and mineralchemistry of a suite of anhydrous, quartzo-feldspathic, gamet-bearing granulites from Kilbourne Hole maar, New Mexico(abstr.). Geological Society of America Abstracts with Pro-grams, 5, 761-:162.

Padovani, E. R. and Carter, J. L. (1974) Blue sillimanite ingarnet granulite xenoliths from Kilbourne Hole, New Mexico(abstr.). Transactions of the American Geophysical Union, 55,482.

Padvoani, E. R. and Carter, J. L. (1977a) Aspects of the deepcrustal evolution beneath south central New Mexico. In J. G.Heacock, Ed., The Earth's Crust, p. 19-55. American Geo-physical Union Geophysical Monograph 20.


Padovani, E. R. and Carter, J. L. (1977b) Non-equilibrium par-tial fusion due to decompression and thermal effects in crustalxenoliths. In H. B. J. Dick, Ed., Magma Genesis, p. 43-57.Oregon Department of Geology and Mineral Industries, Bulle-tin 96.

Parkin, K. M., Loeffier' B. M. and Burns, R. G. (1977) Mdss-bauer spectra of kyanite, aquamarine, and cordierite showingintervalence charge transfer. Physics and Chemistry of Miner-a l s . 1 . 3 0 1 - 3 1 1 .

Rossman, G. R. (1975) Spectroscopic and magnetic studies offerric iron hydroxy-sulfates: intensification of color in ferriciron clusters bridged by a single hydroxide ion. AmericanMineralogist, 60, 698-704.

Schairer, J. F. (1931) The minerals ofConnecticut. ConnecticutState Geological and Natural History Survey, Bulletin No. 51.

Smith, G. and Strens, R. G. J. (1976) Intervalence-transfer ab-sorption in some silicate, oxide and phosphate minerals. InR. G. J. Strens, Ed., The Physics and Chemistry of Mineralsand Rocks, p. 583-612. John Wiley & Sons, New York.

Spencer, L. J. (1920) Fibrolite (:sillimanite) as a gemstone fromBurma and Ceylon. Mineralogical Magazine, 19, 107-112.

Thompson, J. 8., Jr. and Norton, S. A. (1968) Paleozoic region-al metamorphism in New England and adjacent areas. In

E. Zen, W. S. White. J. B. Hadley and J. B. Thompson, Jr ' ,Eds. Studies of Appalachian Geology Northern and Maritime,p. 319-327 , John Wiley and Sons, New York.

Vr6na, S. (19'19) A polymetamorphic assemblage of grandidier-ite, kornerupine, Ti-rich dumortierite, tourmaline, sillimaniteand garnet. Neues Jahrbuch fi.ir Mineralogie Monatshefte1979,22-33.

White, E. W. and White, W. B. (1967) Electron microprobe andoptical absorption study of colored kyanites. Science, 158,915-917.

Wignall, J. W. G. (1966) Mctssbauer line broadening in trivalentiron compounds. Journal of Chemical Physics, 44,2462J467.

Wilkins, R. W. T. and Sabine, W. (1973) Water content of somenominally anhydrous silicates. American Mineralogist, 5E,50E-5 16.

Winchell, A. N. and Winchell, H. (1951) Elements of OpticalMineralogy. An introduction to Microscopic Petrography. PartII. Descriptions of Minerals. Fourth Edition, John Wiley andSons. New York.

Manuscript received, January 6, 1982;acceptedfor publication, March 18, 1982.

the colors of sillimanite - mineralogical society of america

Documents