Origin of Optical" Variations in Grossular-Andradite Garnet

American Mineralogist, Volume 69, pages 328-338, 1984

Origin of optical" variations in grossular-andradite garnet

Mrzuntro Axrzuxr Institute of Mineralogy,Petrology, and EconomicGeology Faculty of Science,Tohoku University,Sendai 980,Japan

Abstract

Optical textures in birefringent grossular-andraditegarnets were studied from the standpoint of crystal growth. Grossular from Eden Hill, vermont, shows one-to-one correlation between growth surface features consisting of well-defined pyramidal hillocks and the internal texture. Grossular from Fujikaramari, Japan, consists of growth sectors with fine lamellaewhich apparpntly correspondto grofih steps.Andradite from Kamihogi, Japan, is composed of many triangular sectors which have the character of so-called lineage structure. The three specimens show sector twins whose planes are parallel, normal, or inclined to the [001] aiis in a thin section parallel to the (110)growth plane. The texture can be interpreted by a general mechanismof formation of order-disorder growth sectors suggestedby Akizuki (1981a).That is, various degreesof Fe3+/Al ordering in the octahedral cation sites were produced on the side faces of surface steps during growth.

Introduction origin of sectors and twinning, although Tak6uchi et al. (1982) Optically non-cubic garnets, showing birefringence, suggestedthat the twinning ofgarnets they usedfor have been reported by many mineralogists and petrolo- X-ray refinement was produced during a transformation gists since the last century. Brauns (lE9l) summarized from disorderedto ordered structure after crystal growth. optical studies of birefringent garnets with sectors, and Through studies of a number of minerals showing so- Meagher (1980)reviewed recent studies. Birefringence is called optical anomalies, such as adularia (Akizuki and especially common in grossular-andraditegarnets (gran- Sunagawa, 1978),topaz (Akizuki et al.,1979), analcime (Akizuki, dite garnet), though garnets near both the end-member 1981a),and chabazite(Akizuki, l98lb), it was compositions are isotropic. Some grandite garnets show found that the internal sectors observable between alternating isotropic and birefringent lamellae parallel to crossed nicols correspond to growth planes and further- the {ll0}t growth planes(e.g., Murad, 1976).Also, gran- more that fine textures correspond to the various vicinal dite garnets show optical growth sectors which corre- faces on the growth planes, on which various degreesof cation or anion ordering may occur during growth. Also, spond to the {110} faces with or without {2ll} faces. Birefringent grandite garnet was transformed irreversibly a general mechanism of formation of order-disorder growth into optically isotropic garnet by heating at 860.C by sectors was suggested by the present writer (Akizuki, Stoseand Glass (1938),though the birefringencewas still l98la). problems observablein some garnets after quenching from a tem- Two of optically "abnormal" garnets are perature of 1060"C(Ingerson and Barksdale, 1943'5. considered here: (l) why is the optical symmetry lower The origin ofbirefringence in garnetshas been various- than that usually shown by X-ray analysesand morpholo- gy, (2) ly attributed to residual strain in the structu.re(Chase and and what is the origin of the complicated textures Lefever, 1960;Kitamura and Komatsu, l97E) and order- that the "abnormal" garnets show without exception? ing of octahedral cations, Fe3+ and Al, thereby reducing The former question was explained in terms of Fe3*/Al (Tak€uchi the symmetry (Tak€uchi et al., 1982). Takduchi et al. ordering et al., 1982), while the latter has (1982) refined the structures of sorne grandite garnets remained an open question since the last century. with X-ray data and found cation or{ering. In grandite The objective of the present paper is to describe the growth garnetfrom Munam, North Korea, the cont€nt of Fe3+in sectors and optical properties of some grandite garnets the eight non-equivalentoctahedral positions varied from showing distinct birefringence and to discuss the l5(l)% to 49(l)%, and the spacegroup was 1T. origin of the optical sectors and twinping using the general Winchell and Winchell (1951)summarized the sectoral mechanism suggestedby Akizuki (19E1a). twinning in birefringent garnets, though fine twinning in each sector was not described. Littte is known about the Observations Although birefringent grandite garnets from several I Throughout this paper indices refer to the ideal iscimetricaxes. localities were studied, only the garnetsfrom three local- m03-fix)v84l03M-032E$02.00 32E AKIZUKI: OPTICAL VARIATION IN GROSSULAR.ANDRADITE 329 ities are described in this paper. Most observationswere feature varies during different stages of crystal growth, madeon thin sections(30 g.mthick) parallel and normal to and therefore the correlation between optical properties the growth surface(110). and surface topography is not always good in some minerals(see Akizuki, 1981a).However, from the present is clear that the internal textures were Grossular from Eden Hill, Vermont observations, it producedduring growth. The garnethas cinnamon color and transparentdodeca- The optical orientation and 2V value in the light- hedralform {ll0}, which is the so-calledhessonite type, colored sectors of Figure lb are shown in Figure 2. The crystals are as large as 5 mm in diameter, and chemical optical vibration direction X is normal to the growth plane composition is Grs57 And1a.3.The surface microtopo- (110), and the other two directions, I and Z, rotate tp to graphy of crystal faces was studied with reflection-type 4ofrom [001] and [Tl0] around the X-axis. The 2V' value interferencecontrast microscopy. The (110)growth surface is characterizedby growth hillocks elongatedparallel to the [001] direction (Fig. la). Each hillock consists of four vicinal faces in a pyramidal form. The faces intersect in edges parallel to [001] and at about 50' to [001] (in projection perpendicular to (110)). Some indistinct boundaries, across which optical contrast differs, shown by white lines at the lowerleft on Figure la; Summits of growth hillocks are observed at the i tions of the boundaries. There are six summits bet two arrows marked A and A'. Also. the summit of a hillock with some growth layers in the central area irili Figure la is shown by arrow B. ,l A thin section parallel to the growth surface prepared from the specimen, whose as-grown face cementedon a glass slide, and was observed on thd'. ! polarizing microscope with a universal stage in order to :; 1; '.: measurethe optical orientation. Figure lb shows a photo-i' ; . "

...^^--... "'. summits.Six centersof fourlingsectors between A and' -/ A' of Figure 1b correspond to the six summits of growt hillocks shown in Figure la. Likewise, the center shown by arrow B in Figure lb correspondsto the summit of the growth hillock B in Figure la. Although the fourling vicinal faces cannot be observed clearly on the photograph (Fig. 1a),because of the small diference of inclina- tion between two of the vicinal faces, the fourling sectorsi are clearly observedbetween crossedpolars (Fig. lb).; The indistinct boundaries shown by the white lines at the lower-left in Figure la correspond to sector boundaries inclined about 50'to [001] in Figure lb. The angleof these boundaries varies from 90oto about 50'with respect to il ttre [001] axis from specimen to specimen. The growth hillock with summit B developedat a late stageof crystal i: growth, and one of the vicinal faces covered part of the growth hillock with summit indicated by a letter C (Fig. la). The sector boundary under the vicinal face is shown by a broken line. The growth hillock at C is much thicker cross polarized transmission (b) photomicrographs of a (ll0) as compared with the vicinal face covering the hillock C, section of garnet from Eden Hill. [001] is vertical, and the and therefore the sector boundary corresponding to the polarizing plane is rotated 3'clockwise from vertical. Seetext for broken line is distinct in Figure lb. Frequently, a growth details. 330 AKIZUKI : OPTICAL VARIATION IN GROSSULAR-ANDRADITE

lar sectors, shown by arrows, correspond to one of the fourling sectors produced on the four vicinal faces of Eden Hill garnet (Fig. lb). The sectorsshown by stars have rims parallel to the outline of the (ll0) plane. A magnifiedschematic sketch of the texture in the rectangle of Figure 4 is shown in Figure 5. The extinction directions shown by arrows are inclined to [001] in the (110)thin section, though they are nearly parallel to [001] in some areas. Lamellae with different extinction angles repeat alternatelyparallel to one ofthe outlinesof{llOi facesin each sector. Thus, a clear correlation exists between orientationsof the lamellaeand extinction (Fig. 5). One of the growth sectors shown in Figure 5 has a rim with two extinction anglesinclined 15'to the (ll0) outline. This suggeststhat a thick hillock with large faces existed on the surface during growth. The fourling sectoral texture as found in Figure lb was not observed in this specimen. Figures 6a and 6b show the sectoral textures of another Fig. 2. Stereogaphicprojection of the optical properties measuredon a (ll0) thin sectionof garnetfrom Eden Hill. Principalvibration direction (X) is normalto the (ll0) growth surfaceand optic axesare indicated by OA. is (+)80'. The four sectors corresponding to the four vicinal faces on the growth hillocks are reflection twins to each other. Other transparentgrandite garnetsfrom Tori- no, Italy, showed a texture similar to that of Eden Hill gamet. However, detailed study was impossible, because of the small size of the crystals.

G r o ss ul ar fr o m F uj ika r ama r i, F uku shi ma P r efec - ture, Japan, The specimens, which are red-brownish and translu- cent, have well-developed{ll0} faces with small {2ll} faces and are about l0 mm in diameter. The {110} faces have someirregular growth stepsand the {21l} faces have fine striations. Fine cracks, some of which are filled by calcite, are abundantin the crystal. Growth bandsparallel to the (ll0) face are due to chemical variations, which were observedwith an electron microprobe. The average chemical composition is Gr7e.7And2s.3. Figure 3a shows fine growth stepson the (110)surface, whose growth center seemsto exist near the crystal edge. In general, the crystal faces do not show regular growth patterns such as those of Eden Hill garnet. Frequently, some straight lines, which continue through several crystal faces, are observed to modify the growth steps slightly. The straight lines may be produced during crystal growth by slip. Such straight lines were not observed in garnets from the other two localities. A cross-polarized microphotographshows fine lamellae correlated with the growth steps on the surface (Fig. 3b). The internal Fig. 3. Correspondingreflection interference-contrast(a) and textures of the fine lamellae and their optical properties cross polarized transmission (b) photomicrographs of a (ll0) are describedbelow. section of garnet from Fujikaramari. [001] is vertical, and the Figure 4 shows another thin section with growth sec- polarizing plane is rotated 4'clockwise from vertical. Roughly tors corresponding to the outline of the (110) face. The horizontal lines may be produced by crystal slip. Roughly many fine black lines are due to irregular cracks. Triangu- vertical lines are due to cracks. See text for details. AKIZUKI : OPTICAL VARIATION IN GROSSULAR.ANDRADITE 331

Fig. 4. Crosspolarized photomicrographofa thin sectionpaxallel to the growth surface(110) ofa garnetfrom Fujikaramari. Black arrow is [001], and the polarizing plane is rotated l2o counter-clockwisefrom vertical. The three edgesofthe crystal with black bands are natural, while the fourth is cut. See text for details. specimenobserved between crossed polars. The crystal Thickness of the lamellae varies parallel and normal to is divided into sectors by vertical boundaries,and extinc- the crystal surfaces, and frequently the lamellae cross tion anglesare roughly symmetrical with respect to [001] each other in thin sections parallel to the growth plane. on the (110) face. Two kinds of lamellae alternate in the Also, the lamellaeare inclined to the growth surface(110) sectors (Fig. 6b). A magnified photograph of the rectan- and cross growth bands in thin sections normal to the gular area in Figure 6a is shown in Figure 7. Both surface. A small difference of AVFe3* ratio was detected elongationdirection and optical extinction of the lamellae between the two kinds of lamellae by electron micro- are roughly symmetrical with respect to the sector bound- probe. Although an attempt was made to determine the aries in some specimens.However, the lamellae are not optical orientation and 2V value, they were not clearly always parallel to the outlines of (l l0) faces, but may be defined, because the optical extinction was not clear. inclined, curved or kinked (Fig. 7). Extinction angles of However. the structure is assumed to be triclinic, be- the lamellae vary along the curves. The sectoral bound- cause the crystal is optically biaxial and the indicatrix aries are sectoral twin planes roughly parallel to [001]. rotates variously with respect to the crystal axes. Internal texture similar to that of the Fujikaramari specimen commonly is observed in garnetsfrom other localities in Japan.

Andradite from Kamiho gi, Yamag uchi P refe c ture, Japan The specimens,which are greenishbrown and translu- cent, consistofwell-developed {110} faces and are about 10 mm in diameter. The averagechemical composition is AndTe.sGr2q2. Very fine non-isotropic chemical zoning, which causesa strong iridescence, was observed in thin sections normal to the growth surface (Akizuki' Nakai' and Suzuki, in prep.). Figure 8 shows growth features on a (ll0) surface consisting of many blocks whose forms are the same as the outline of a rhombic (1l0) growth plane. The blocks, which have fine striations parallel to the [001] direction, are sometimesslightly misoriented, that is, this garnethas the so-called lineage structure as found in galena Fig. 5. Schematicdiagram of growth sectors in the rectangle (Buerger,1932). shown in Fig. 4. Growth lamellae, optical extinction directions between the surface fea- (arrows), and extinction anglesare shown. The sectorsconsist of One-to-one correspondence observable in the alternate lamellae showing larger extinction angle (left number) tures and the internal texture is not just growth and smaller angle (right). [001] is vertical. specimen,because the area below the surface 332 AKIZUKI: OPTICAL VARIATION IN GROSST]LAR-ANDRADITE

* !:ii:::!iii.n

,,,

Fig. 6. Cross polarized photomrcrograph(a) and schematicdiagram (b) of a thin section parallel to the growth surface (ll0) of garnetfrom Fujikaramari. [001] is vertical and the polarizing plane is rotated about l2o counter-clockwisefrom vertical. Seetext and caption of Fig. 5 for details. shows a fine granular texture whose development is Figure l0 shows growth sectors which correspondto a presumably due to strain. However, many triangular hillock with four vicinal faces. This arrangement of sectors correlating to the outline of the (110) face are growth sectors is extremely rare. The different optical observedin the area deeper than about 0.5 mm below the contrasts shown by the four starts in Figure 10 are the surface(Figs. 9a, 9b and l0), and schematicfigures ofthe result of chemical zoning. Figure 12 shows a cross- sectors are shown in Figure I l. The optical orientations polarized photomicrographof a thin section normal to the of sectorswith solid and dashedlines (Fig. l l) are roughly (ll0) growth surface. The texture crossing the horizontal symmetrical with respect to the [001] direction. The growth lamellae is a profile of growth sectors like those sectorsare more or less optically homogeneousand have shown in Figure 9. The textures are optically modified no lamellaecorresponding to the striations on the surface where they intersect. One example of the triclinic optical (Fig. 8). Figures 9a and 9b show some triangular sectors orientation is shown in Figure 13. Although the rotation indicated by letters B and E, corresponding to those of of the optic axial plane is 33" from the [Tl0] axis in the Figure 11. Since the triangular sectors are in contact or (110)plane in this example, the rotation anglesvary from overlap with each other, their shapesare modified from about 25" to 35'in the specimen,and frequently the angle the ideal shapesof Figure ll. and the 2V value are not clearly defined. This type of AKIZU KI : OPTICAL V ARIATI ON I N GROS SULAR.AN DRADITE JJJ

Fig.7. High-magnificationcross polarized photomicrograph of finelamellae in therectangle shown in Fig.6a. [001] is vertical, andthe polarizingplane is rotated3' clockwisefrom vertical.

Fig. 9. Cross polarized photomicrographshowing sectorsln a texture was not observedin garnetfrom other localities in thin section parallel to the (ll0) growth surface of garnet from Japan.Iridescent grandite garnet from Nevada, however, Kamihogi.The letters B and E correspondto those in Fig' 11' 20'counter- shows a texture similar to that of the present garnet [001] is vertical and the polarizing plane is rotated (Ingersonand Barksdale,1943,their Fig. 8). clockwise from vertical.

Fig. 10. Cross polarized photomicrographof sectoralfourling Fig. 8. Reflection interference contrast photomicrograph of twins in thin sectionparallel to the (1l0) growth surfaceofgarnet plane is the growth pattern on a (110) surface of garnet from Kamihogi. from Kamihogi. t0011 is vertical and the polarizing vertical' [001] is vertical. rotated 2E' counter-clockwisefrom 334 AKIZU KI : OPTICAL VARIATI ON IN GROSSU LAR.AN DRADITE

Fig. 13. Stereographicprojection of optical properties measuredin a (ll0) thin sectionof garnetfrom Kamihogi. Principalvibration directions (X,Y,Q and optic axialplane are shown.X is notnormal to thegrowth surface (ll0).

1981a),and chabazite(Akizuki, l98lb). Therefore, the origin of optical variations in the present specimenswill be interpreted by the generalmechanism of formation for order-disorder growth sectors suggested by Akizuki (1981a). Fig. ll. Schematic figures of sectors observed in thin sectiors parallel to the (ll0) growth surface. oi gurn"t f.o. Crystal growth mechanism Kamihogi. Extinction directions of the sectors wi6 solid and dashedlines are pyramidal symmetrical with respec, t;i00ii:-shown by The shape of the growth hillocks on garnet arrows in Fig. A. t00ll is vertical. from Eden Hill suggestsspiral growth. The garnets from both Fujikaramari and Kamihogi, however, consist of Discussion many sectors without a growth center. In Fujikaramari garnet, fine lamellae may correspondto growth stepsthat Correlations growth between features on the crystal nucleate at the sector boundary or the crystal edge, surfaceand the internal texture as found in Eden Hill and suggestingthat Fujikaramari garnet grew by the two- Fujikaramari garnets have been observed in adularia dimensional nucleation mechanism. Kamihogi garnet, (Akizuki and Sunagawa, 1978), analcime (Akizuki, with lineagestructure, was presumablyproduced by two- dimensionalcrystal growth as well, becausesectors with a growth center are extremely rare. The clear correlation between surface features and internal textures strongly suggeststhat the internal textures of garnet from Eden Hill and Fujikaramari were produced during growth, not by phase transition after growth. In minerals whose growth conditions changedabruptly at the latest stageof crystal growth, surfacefeatures are not always correlated to internal textures (Akizuki, l98la). The internal textures in Kamihogi garnets(Figs. 9a, 9b and l0) are similar to the growth sectors of garnets from the other two localities (Figs. la, lb and 4); therefore they were presumablyproduced during growth as well, though the one- to-one correlation between the surfacefeatures and internal texture is not observed in the crystal. Thin sections normal to the (110)growth surfacesof somegarnets show various textures crossing zonal growth lamellae parallel Fig. 12. Cross polarized photomicrograph of a thin section to the (110) surface (as in Fig. l2), and therefore it has normal to the (ll0) growth surface of garnet from Kamihogi. been suggestedthat the textures were produced after AKIZUKI: OPTICAL VARIATION IN GROSSULAR.ANDRADITE 335 growth by a phasetransition (Tak6uchi et al., 1982)or by strain. Such a texture, however, is frequently produced during growth. The textures crossingthe growth lamellae shown in Figure 12 correspond to growth sectors which were observed in thin sections cut parallel to the growth surface (110) (Figs. 9a and 9b).

Fet*lAl ordering Tak6uchi et al. (1982) represented the structure of garnet using an array of octahedra on a pseudo-trigonal axis, which is similar to the rod-packing model of the garnet structure suggestedby O'Keefe and Andersson (1977). The cation ordering can be explained^by differ- encesin the immediate surroundingsof the Fe'*/Al sites of the octahedral array exposed on the growth steps. Fig. 15. Relationbetween formation of fourlingsectoral twins the relation between the pseudo- Figure 14 indicates andgrowth directions (heavy arrows) of vicinalfaces on the(1 l0) of trigonal axes shown by arrows and the crystal form surfaceof garnet.The letters"a" correspondto the "a" of Fig' garnet. Four ofthe eight octahedralarrays in the unit cell 14. lie on the four axes. Assuming growth of the crystal is in the direction shown by the arrow on the (110) plane, the and (001), the same situation will occur on each vicinal immediate surroundingsof octahedrain these four arrays face, and fourling twins will be formed during growth differ from eachother with respect to the growth direction (Figs. l0 and 15).Ifthe growth steps kink or curve during on the surface. Thus, the four octahedral sites are not growth, the Fe3+/Al partitioning at each site may differ equivalent during growth. If this non-equivalenceresults from growth step to growth step, and complicated optical in different relative preferencesfor iron and aluminum on textures will be produced in the crystal during growth the sites, then ordering will occur and the symmetry of (Figs. 6a, 6b, and 7). Ordered structures with various the crystal will be reduced. Fe3*/Al ratios can be produced in the garnets except the Although the formal chargesof the two cations are the end-members. Birefringence is observed even near the same,partitioning migtrt occur on the growth surface due end-members,because optics are very sensitive to order- to the different ionic radii (0.57A for Al3*, 0.67A for ing in the structure. Fe3*) and electronegativities. Akizuki et al. (1979\ pre- The three specimensare optically biaxial, and of them, dicted from optical observations that ordering of F and the two specimensfrom Kamihogi and Fujikaramari may OH could occur on growth surfacesoftopaz, becauseof be triclinic, judging from the relations between optical different ionic radii. This hypothesis was confirmed by and crystallographic orientations. However, the symme- the neutron diffraction studies of Parise (1980). try of Eden Hill garnet may be monoclinic. Growth steps If a crystal grows in the directions of the arrows in on the growth hillocks of garnet from Eden Hill are Figure 15, which are symmetrical with respect to (Tl0) inclined to the [001] axis on the (110) surface' and therefore the optic vibration direction does not coincide with the [001] axis, because one-dimensionally ordered arrangementsrepeat parallel to the growth step on the surface, resulting in a two-dimensionally ordered structure. If the two-dimensionally ordered structures stack regularly in the direction normal to the growth plane (110) during slow gowth, a diad axis may occur in the direction parallel to the [110] axis, which coincides with one of the principal optical vibration axes. If so, the garnet from Eden Hill will be monoclinic (Fig. 2). An orthorhombic structure cannot be produced on the vicinal surfacewith growth stepsinclined to the [001]axis on the (110) face, but may be formed on the vicinal surface with growth steps parallel or nonnal to the [001] axis. Whether such a structure exists on the (110) sur- of some garnets is not known. According to X-ray Fig. 14. Relation between the crystal form of garnet and faces garnets from Kamaishi, Japan, by directions of the octahedral arrays shown by arrows labelled diffraction analysis of c for the a,b,c, and d. Growth steps discussed in text move in the Tak6uchi et al. (1982),the cell parametersa and direction shown by the thick arrow on the (ll0) face. cubic cell are equal but the angle B is not 90'; the cubic a 336 AKIZUKI: OPTICAL VARIATION IN GR.OSSULAR.ANDRADITE

hedral sites, positive and negative octahedral, by ditrer- ences in the ratio of Fe3*/Al. For instance, in Munam garnet (Gr66And3)analyzed by Tak6uchi et al. (1982),the Fe content varies frorn 4l atom percent to 49 atom percent in.the fou.r sets of positive octahedraand from 15 atom percent to 29 atorn percent in negative octahedra. Figure 17 shows schemat.icallythe twinned structure on the symmetrical vicinal faces of the (110) growth plane. This figure was modified from the twinned structure shown by Tak6uchi et al. (1982)to interpret the growth I twinning observed in the present garnets. In grandite garnet, the twelve {110} growth planes produce twelve I growth sectors,and each surfacehas various vicinal faces which make many fine sectors. These complexities make it essentially irqpossible to ascertain the type of twin .J using a random section of garnet. Fine larnclloe Fig. 16. Monoclinicunit cell of garnet. Pseudo-cubiccel (Figs. dimensionsa andc areequal, as are a and7, whichare not equal The optical orientations of the fine lamellae 5 and to 90'. Solidand dashedlines show two-dimensional unit cells 6b) are rough$ symmetrical with respect to the sectot parallelto eachother. Diad axis is normalto the (ll0) growth twin boundaries. The lamellae never cross a sector twin surface.The unit cell data are thoseof garnetfrom Kamaishi, boundary, though they cross growth bands parallel to the Japan,which was analyzed by Takduchier al. (1982). (ll0) growth surface. The internal lamellae in garnets from Fujikararnari and Eden Hill correlate with the growth steps on the surface (Figs. I and 3). Garnet with and Tare equal but are not 90'. The garnet is monoclinic, fine striations o{r a small (2ll) face from Fujikaramari with the diad axis parallel to [110] of the cubic cell as shows fine larnellae normal to the erowth surface in the shown in Figure 16,though Takduchi et al. (1982)reduced their data in the orthorhombic system, becausethe lattice angles a and 7 are very close to 90'. Monoclinic symmetry is consistent with the optical observations for Eden Hill garnet. ; c Twinning o ,tt According to crystallographic definition, optical prop- b \" erties such as2V andoptical orientation must be symmet- ^ a I etO oo rical with respect to a twin plane. Becausethe sector twin { components of the garnets discussed above, especially 3'o o {l those from Kamihogi and Fujikaramari, are not optically homogeneous,the Fe3+/Al ordered structure is not always rigorously symmetrical with respect to the "twin" o t t o o planes,although the optical orientations of the sectorsare D roughly symmetrical with respect to the [001] axis. There- o c o o \-./ o fore, these sector twins are not twins in a strict crystallographic sense, and the term sector twin here is used to n a) o denote a relation between two sectors whose growth t t o directions and optical orientations are roughly symmetrical with each other. vii The traces of sector twin planes of garnetsfrom Kami- hogi and Fujikaramari are parallel and normal to [001] on produced the (110)face, while in garnet from Eden Hill, one sector Fig. 17. Schemaficdiagram of twinned structure on vicinal facesofthe (l l0) growth surfaceofgarnet. The figure was twin plane is parallel to the whereas the other [fi)l], modified from tbe twinned structure shown by Tak€uchi et al. varies from 90'to about 50ofrom the [001] axis. These (1982).Solid and opeo circles show octahedral cation sites, Mr differences are a result of diferences in the growth and M2, at which Fc3*/Al occupancyis different. The twin plane features. is normal to tlp (l 10)growth surface.The crystal axes a and b of Takduchi et al. (1982)distinguished two kinds of octa- the pseudo-cubiccell are shown. AKIZUKI : OPTICAL VARIATION IN GROSSULAR-ANDRADITE t37

structure is produced by the two-dimensional atomic arrangementexposed on the side faces of growth steps and is frozen during and after growth (Akizuki, 1981c). Hariya and Kimura (1978)synthesized non-cubic grandite garnet at hydrothermal conditions and determined the "stability" field of pressure and temperaturefor the formation on non-cubic garnet. They compared the "stability" field with the phase transition temperature of natural non-cubic garnet as determined by heating. The non-cubic garnet formed at temperatures lower than Fig. lE. Directionsof motionof fine stepson thick growth about 700'C at low pressure,while the transition tempera- steps.Degree of orderingand/or ratio of Fe3+/Alare different in ture of the natural garnet was slightly higher than 800'C. the two kindsof sectors.A andB. In the present interpretation, the metastable Fe3+/Al ordered structure of synthetic garnet is produced on the free surface during growth, whereas the order-disorder l} sector between crossedpolars. From these observa- {21 transition occurs in the three-dimensionalstructure. tion it is thought that the lamellae were produced during Some garnets show complicated internal textures growth. which may be attributed to residual strain after crystal Figure diagram showing a possible 18 is a schematic growth, and these textures may overlap with growth mechanism for formation of lamellae. If a growth step sectors in some thin sections. It is dfficult to distinguish consistsof monatomic layer, a step will move in only one between the two textures in thin sections normal to the if the growth step is direction along the surface, whereas growth plane in some cases. The optical properties pro- thick, thin growth layers may move in two different duced during growth may be modified by strain after directions (Fig. l8). If so, differencesof chemical compo- growth in some specimens. It appears that Eden Hill Fe3+/Al ordering might occur on sition and/or degree of garnet maintains the original structure and texture and the two kinds of surface (A and B in Fig. 18), and that the modification of the garnets from Fujikaramari therefore lamellae may be observed in the crystal. Stria- and Kamihogi is not large. tions on the (2ll) surfaceconsist of alternating(110) and (l0l) faces, and polysynthetic twinning develops in the General conclusions {2ll} sector. The distance between steps varies during growth, and Optical textures of natural grossular-andraditegarnet the thicknessof stepscan changeby bunching or dispers- can be interpreted in terms of Fe3+/Al ordering which is ing ofsteps. This explains the variation ofthickness ofthe produced on the surface during growth. Growth twins, internal lamellae.If a surfacefeature changesby intemrp- whose mirror planes are in orientations parallel, normal, tion of growth, it mieht result in two kinds of lamellae or inclined to t00ll in (110)thin sections,correspond to crossing each other in a thin section parallel to (ll0). sector boundaries at which growth steps are roughly symmetrical. Various types of twins and optical proper- Stability ties of garnet which were summarized by Winchell and It is likely that only the cubic structure is stable at the Winchell (1951)may be interpreted by the sameprinciples growth temperaturein granditegarnet, becausenon-cubic describedin the present paper. grandite garnetsare not homogeneousbut have a hetero- Even if a growth face is normal to the mirror planes of geneoustexture characteristicof metastableminerals. No the ideal crystal structure and a growth hillock possesses unequivocal evidence of phase transitions from the high the mirror planes, vicinal faces on the face are not always form to the low form have beenobserved in thesegarnets. normal to the mirror plane. If the crystal has a growth The crystal symmetry of the meta-stable garnet is not hillock on the surface as shown in Figure l9a, the vicinal merely the result of the chemical composition, but results faces will be normal to the mirror planes (m), and from Fe3+/Al ordering in the octahedral sites. The degree therefore the structure produced on the hillock will have of ordering in minerals can be affected by growth tem- one mirror plane at least and will be disordered with perature and growth rate (Akizuki, 1981a,b,c). If the respect to that plane. However, since the vicinal faces growth temperatureof garnet is very high, a cubic disor- shown in Figure l9b are not normal to the mirror plane, dered structure may be produced during growth. If the the crystal structure does not possessthe mirror plane growth rate is very high or extremely low, a cubic and may be ordered, thereby reducing the symmetry. The disordered garnet may occur as well, becausethe disor- crystal symmetry is not controlled by the macroscopic dered structures are favored by high growth rates, or are form of the crystal, but by the microscopic form of obtained by equilibrium growth at low growth rates. At growth hillocks, that is, the order-disorder structure is intermediate growth rates and temperatures typical of determinedby the directions of growth steps. In the case hydrothermal or skarn deposits, the metastable ordered of Figure 19b, even if the crystal form has the mirror 338 AKIZUKI: OPTICAL VARIATION IN GROSSULAR-ANDRADITE

Akizuki, M. (l9Elc) Crystal growth and phase transition of intermediate microcline. Neues Jahrbuch fiir Mineralogie, Monatshefte,18l-189. Akizuki, M. and Sunagawa, I. (1978) Study of the sector structure in adulariaby meansof optical microscopy, infra-red absorption, and electron microscopy. Mineralogical Maga- zine, 42, 453-462. Akizuki, M., Hampar, M. S. and Zussman,J. (1979)An explana- tion properties Fig. 19. Growth hillocks (summitsin center) with mirror of anomalousoptical of topaz. Mineralogical planes(m). Growthsteps on hillock(a) are normal to the mirror Magazine, 43,237-241. (1891) planes,while the growthsteps on hillock(b) areinclined to the Brauns, R. Die OpJischenAnomalien der Krystalle, Bei S. mirrors.Hillock (c) consistsof a combinationof two kinds of Hirzel, Leipzig. (1932) vicinalfaces (a and b). Seetext for details. Buerger, M. J. The significanceof "block structure" in crystals. American Mineralogist, 17, 177-191. Chase,A. B. and Lefever, R. A. (1960)Birefringence of synthet- planes, the disordered structure with the mirror planes is ic garnets. American Mineralogist, 45, l 126-1129. not necessarily produced during growth. An example is Hariya, Y. and Kimura, M. (1978)Optical anomalygarnet and its chabazitecrystals with rhombohedral form consisting of stability field at high pressuresand temperatures.Journal of triclinic ordered structure (Akizuki, 1981b).The vicinal Faculty of Science, Hokkaido University, SeriesIV, lE, 6l l- faces ofgrowth hillocks on dodecahedralgarnets are not 624. normal to a mirror plane. Therefore, the ordered struc- Ingerson, E. and Barksdale, J. D. (1943)Iridescent garnet from ture produced on the hillock is not orthorhombic. but the Adelaide Mining District, Nevada. American Mineralogist, 28,303-312. monoclinic or triclinic. If a growth hillock is composedof Kitamura, K. and Komatsu, H. (1978)Optical anisotropy associ- a combination of the two (a cases and b of Fig. l9), as ated with growth striation of yttrium garnet, Yr(Al,Fe)5O12. shown in Figure 19c, the sectoral texture may become Kristall und Technik. 13.811-816. more complicated. Surface features must be observed in Meagher, E. P. (19E0)Silicate garnets. In P. H. Ribbe, Ed., order to understand minerals with anomalous optical Reviews in Mineralogy, 5, Orthosilicates, p. 25-66. Mineral- properties. X-ray analyses should be performed on indi- ogical Society of America, Washington, D.C. vidual sectors after determining the optical properties as Murad, E. (1976)Zoned, birefringent garnetsfrom Thera Island, well as the chemical compositions. Optically homoge- Santorini Group (Aegean Sea). Mineralogical Magazine, 40, neous crystal flakes for X-ray analysis may be obtained 715-119. (1977) packing from thin sections parallel to the growth surface. O'Keeffe, M. and Andersson, S. Rod and crystal chemistry. Acta Crystallographica, A3l, 914-923. Acknowledgments Parise, J. B. (1980) Order-disorder phenomena in Minerals: Structural studies.Ph.D. Dissertation, JamesCook University I thankDr. D. R. Veblen,Johns Hopkins University,and Dr. of North Queensland. E. Dowty, AmericanMuseum of NaturalHistory, for valuable Stose,G. W. and Glass,J. J. (1938)Garnet crystals in cavitiesin suggestionsand Englishcorrection of the manuscript.Also, I metamorphosedTriassic conglomeratein York County, Penn- thank Dr. R. Aines, CaliforniaInstitute of Technology,for sylvania. American Mineralogist, 23, 430435, critical readingof the manuscript.I am gratefulto Drs. K. Tak6uchi, Y., Haga, N., Umizu, S. and Sato, G. (1982)The Yamaokaand H. Konno,Tohoku University for providingthe derivative structure of silicate garnets in grandite. Zeitschrift samples. ftir Kristallographie, 158, 53-9. Winchell, A. N. and Winchell, H. (1951)Elements of Optical References Mineralogy. Part II, 4th edition, John Wiley and Sons, New Akizuki, M. (l9E1a)Origin of optical variationin analcime. York. AmericanMineralogist, 66, 403-409. Akizuki, M. (l98lb) Origin of optical variationin chabazire, Manuscript received,December 27, 1982; Lithos,14, 17-21. acceptedforpublication, September 14, 1983.