Alexandrite FromLake Manyara, Tanzania*

By EDWARD GUBELlN, Ph.D., C.G., F.G.A.Honorary Professor of the University of Stellenbosch

Lucerne, SWitzerland

IntroductionThe far-flung emerald deposit at

Lake Manyara was mentioned byTHURM (1972) and described moreprecisely by BANK (1974), by thepresent author (1973 & 1974), andmore recently by MWAKlSUNGA(1976). BANK (1974) reported a smallpiece of mother rock, on which, as-sociated with feldspar and mica,alexandrite, emerald and ruby oc-curred, i.e. three extremely rare gem-stones all of which owe their color toan admixture of chromium. From themineral paragenesis he also surmisedthat the Manyara deposit was analo-gous to the famous occurrence ofberyllium minerals at Tokowaya in theUral mountains (described by FERS-MANN, 1929), where the same gemminerals are found in close association.Towards the end of his paper "TheEmerald Deposit at Lake Manyara,Tanzania" (Lap. Journ., May, 1974)the writer refers to several other gemssuch as apatite, garnet, spinel andchrysoberyl, of which the detection of

*Manuscript received July, 1976.

FALL 1976

alexandrite is the most intriguing(Figure 1).

Since then further investigationcould be carried out on the alexandritefrom Lake Manyara in its quality as acrystal and precious stone as well as onits mode of occurrence, justifying aspecial publication on this fascinatinggem.

The mother rock of the Manyaraalexandrite has meanwhile been care-fully analyzed. Each individual mineralcomponent was X-rayed. The redgrains which can easily be seen in therock all proved to be chondrodite. Inthin sections they are yellow. Contraryto the previous information (GUBE-LIN, 1974), garnet seems to be ratherscarce or then very sparsely dispersed;at least it has not been encountered ina more recent and more thoroughanalysis. The mineral componentswhich were definitely identified are insequence of frequency: actinolite, en-statite, fayalitic olivine, chondrodite,chrysotile and pleonaste. Conse-quently the name actinolite-schist isproposed. In contrast to the formerconjecture it is not a peridotite, be-

203

cause it does not contain enougholivine. The association of actinoliteand alexandrite indicate a metamor-phic rock, hence: it is a schist.

According to BANK (1976) andOKRUSCH (1971) the possibilities offormation are more limited for thealexandrite than for ordinary chryso-beryl. Alexandrite owes its great rarityto the fact that the geochemical condi-tions are highly accidental becauseberyllium and chromium do notnormally occur within the same rocksuites. Beryllium is concentrated in thepegmatitic and pneumatolytic phaseswhich themselves are devoid of chro-mium. The latter, on the other hand, is

more frequently present in maficrocks. Therefore the formation ofalexandrite may only be expected inplaces where the ingredients of theserocks can meet, that is to say, wherethe pegmatitic and pneumatolytic re-sidual phases carry the beryllium intocontact with the chromium in suchultramafic rocks as actinolite, perido-tite, pyroxenite, and their metamor-phic de rivates (particularly serpen-tinites). These were the favorable con-ditions found in the deposit at LakeManyara manifesting close similarityto the deposit of Tokowaya as well asof the Girdlestone Farm at Novello,Rhodesia (BANK, 1964).

38·rr------------,2

t1!~AQUATERNARV

1;:::-:-:1 VOLCANITES(TERTIARY)

!t:;>\F] 6ASEr,lENT COMPLEX

@ ALEXANDRITE DEPOSIT30 MIS.

Figure 1. This simplified geological map of northern Tanzania shows the situation of thealexandrite deposit at Lake Manyara, where alexandrite occurs together with some otherehromium-beerinq gem minerals.

204 GEMS & GEMOLOGY

External AppearanceConsidering this geochemical simi-

larity of occurrence, the author wasnot astonished to observe that all theManyara alexandrites which he hasseen so far looked puzzlingly similar tothose from Tokowaya and from No-vello. In daylight they all displayed thesame exquisite bluish-green hue with aremarkable change of color to pro-nounced raspberry red in incandescentlight.

With a few exceptions, most of thespecimens seen or investigated dis-played those characteristic formationsof three crystals intersecting eachother, which yield the well knowntrillings. The habit of the individualcrystal was mainly determined by theprism. Some were just broken-offcrystals and others merely fragments.Some were absolutely clear and ofexceptional gem quality and when cutwould have resulted in high gradeprecious gems up to 5 carats. However,many crystals and fragments were in-tensely etched and marred by toomany mineral and liquid inclusions aswell as cracks.

Chemical CompositionBy means of the electron micro-

probe, an analysis was carried out ontwo crystals, the results of which arecommunicated in Table I. The calcula-tion program was started off in assum-ing B~O to amount to 20% (theoreticalvalue), because due to the low atomicnumber (Z=4) of Be the microprobedoes not register this element. In thealexandrite from Novello BeO amountsto 19% and thus very closely ap-proaches the theoretical value.

FALL 1976

TABLE I

Alexandrite

No.1 No.2

%2077

0.250.7

%2078

0.200.7

0.120.13

0.080.4

The percentage of the trace ele-ments present reveals that chromiumand iron are responsible for the color.Vanadium seems to be entirely absent.However, as will be explained furtherdown, the striking alexandrite effect iscaused solely by chromium, while ironis of secondary importance and mainlyacting as an inhibitor (see later).

Physical PropertiesThe physical properties do not

manifest any peculiarities - the oneexception of course being thechameleon color change - and maytherefore be mentioned hereaftermerely for statistic purposes.

Density:The frequency medium of numer-

ous measurements taken results ind = 3.7lg!cm3 at 4°C

R. I. and birefringence:Again the medium data of numer-

ous readings are:nexn(3n.y

whereas the birefringence varied from0.008 to 0.010. The optical characteris of course negative.

1.7451.7501.754

205

Trichroism:This property is very distinct and

depending upon the three main opticaldirections produces the colors shownin Table II.

TABLE IIOptical Incandescent

direction Daylight light

fl", reddish carmine red

nl3 yellow green orange red

n'Y blue green greenish

For this accentuated pleochroismthe transition-metal chromium alone isresponsible.

Absorption:The absorption spectrum is a typi-

cal chromium spectrum whose spectralimage is characterized by two promi-nent bands and marked by several indi-vidual absorption lines in the redregion (Figure 2). The two bands aredesignated A and B. They range astridethe critical wavelengths of 580 nm and415 nm. If they trespass these valuesthe alexandrite is green, otherwise it isred. The A band extends from approx-imately 550 nm to roughly 610 nmand culminates at 571.4 nm. It isaccompanied on its long-wave side bythe telltale chromium lines at 680,

A

4!J4nm

nm 350 400 450 500 550 600 650 700

Figure. 2. Absorption spectrum of thealexandrite from Lake Manvara in unpolar-ized light ranging from 325 nm to 700 nmat room temperature.

206

678, 665, 655, 649, and 645 nmwhich result from spin-forbiddentransitions. The B band is somewhatnarrower ranging from 395 nm to 443nm with its peak at 415 nm. It ischaperoned by four satellite lines at470 nm and 465 nm on thelong-waveside and at 385 nm and 375 nm on theshort-wave side. The latter two linesare caused by trivalent iron. Thenarrow lines in the red and in theshort-wave section become more con-spicuous when a polaroid is used. Theabsorption minimum is situated at 494nm and very closely corresponds tothe intensity maximum of the solarspectrum in the green (at 500 nm)imparting the alexandrite a green colorin daylight. Incandescent light excelsin a predominance of longer wave-lengths which are transmitted by thealexandrite's second absorption mini-mum at 725 nm thus conferring thegem a red appearance.

The breadth of band A suggeststhat it must be provoked by twooverlapping absorption spectra of twodiffering color centers. This becomesevident when polarized light is ap-plied. In unpolarized light the breadthof the A band complex is produced bythe superposition of two bands withmaxima slightly displaced against eachother. The polarized absorptionparallel to the z-direction (A liz} culmi-nating in longer wavelengths at 571.4nm is responsible for the green color,while the polarized absorption per-pendicular to the z-direction (Alz)sweeps to a maximum among shorterwavelengths at 551 nm and inducesred coloration (Figure 3). This changeof color is called "Alexandrite Effect."

GEMS & GEMOLOGY

Figure 3. Polarized absorption spectra ofband A parallel and perpendicular to thez-direction (Hassanet et., 1974).

The alexandrite-effect:This phenomenon is an excep-

tionally well-pronounced specialty ofthe Manyara alexandrite, for the colorchange from green to red and viceversa is very definite and so complete,that the beholder cannot help wonder-ing about the nature of its cause. Theattempt to explain essence and causeof this unusual feature was undertakenin numerous publications. For a verylong time it was simply connectedwith the mineral's strong trichroismand it was merely regarded as an effectprovoked solely by the differen t

spectral composition of daylight andincandescent light. None of these olderand rather one-sided explanations ofthe very complex process were con-vincing. Nowadays the contexts arebetter understood, and hence theinterpretation of the causal conditionsacquire a new significance. Today'sapprehension is based upon a muchmore profound comprehension of theintrinsic structure of the alexandrite.While POOL (1964) and partly alsoWHITE et al. (1974) attributed thecolor change to a rather psychologicaleffect of the human eye and brainthan to the inherent structure andproperties of the alexandrite, FAR-REL et al. (1963 & 1965), NEWN

FALL 1976

HAM et al. (1964), WHITE et al.(1967), and HASSAN et al. (1974)investigated the structural backgroundof the question. Their results providean instructive insight into the mecha-nisms which take place, when lightfalls into an alexandrite. The authorfeels that his present account of theManyara alexandrite might be a wel-come opportunity of making someintegrating reflections on these morerecent studies.

Alexandrite is isostructural witholivine and sinhalite, i.e. it has anidentical structure as these two min-erals (FARRELL, NEWNHAM, 1965and STRUNZ 1968). The structure ofthe alexandrite is determined by adense packing of the relatively largeoxygen ions in a hexagonal arrange-ment. Within this lattice the tetra-hedral interstices are occupied by Bewhile the Al partly sits inside thecoordination octahedra. This results ina combination of tetrahedral Be04groups and octahedral A104 groups.Numerous chemical analyses as well asrecent refinements of the structure ofchrysoberyl have lead to the convic-tion that the transition elementchromium (Cr}') (NASSAU, 1975) isthe chromophorous ion, yet not vana-dium. The aforementioned chemicalanalyses did not furnish any alibi forthe existence of vanadium.

It is not the mere presence of Cr3+

ions within the structure of the alex-andrite which is responsible for thecolor change but rather their array inthe crystal lattice. Within the structureof the chrysoberyl there are variouslattice sites with octahedral symmetry,which are held by Cr3+ instead of AI3+ions. However, normally only two of

207

them are being claimed by Cr3'-. Therefinement of the structure of chryso-beryl by FARRELL et al. (1963)disclosed that these two octahedrallattice sites preferred by the Cr3'- ionsdiffer with regard to size - not verymuch but significantly enough to pro-duce the color change (Figure 4). Halfof the AI3+- ions are perched in coordi-nation polyhedra with inversionsymmetry = A13+ (1) and the other onmirror plane positions = A13+ (2). Theaverage distances to the nearest oxy-gen ions measure 1.890A for theA 13+ (1) sites and 1.934A for theA13+ (2) sites (the mean value betweenthe two (= 1.914A) happens to bealmost identical to the distance Al -o in alpha-corundum). If trivalentcations replace others they favor oneplace to another depending uponsymmetry and size.. The Cr3+ ion (¢ = 0.615A) isslightly larger than the AIYion (¢ =

0.530A) and hence prefers the morespacious A13+ (2) sites. Yet, if intro-duced under pressure and hightemperature some Cr3+ ions mayaccommodate themselves on narrowerA 13+ (1) sites. The dispersion of theCrY ions onto transition-metal sitesdiffering in size and symmetry withinthe crystal field cannot happen with-out drastic influence on the colorproperties of the alexandrite, espe-cially so because the absorption spec-trum of the alexandrite consist of twoindividual spectra of the CrY (1) andthe Cr3+ (2) color centers overlappingone another in the A band.

The intrinsic cause of the cha-meleon effect of the CrY ions on theirtwo different lattice sites is their dif-

208

Figure 4. The refined structure of chryso-beryl projected on (001). Heights of theatoms are expressedin cell fractions (Farrellet al., 1963).

ferent distance to the oxygen ions. Onthe smaller octahedral sites (1) of thealexandrite lattice the Cr - 0 distanceis shorter, the Cr3'" ion finds itself in acompressed position and thus pro-motes the transposition of the absorp-tion maximum from 571.4 nm(assigned to the more numerous Cr3'"(2)centers) to 551 nm (ascribed to theless populated Cr3'" (1) centers). Thedenser the population of the Cr3'" (1)centers the greater the spectral disloca-tion and hence the more intense thered coloration. This condition concursexactly with observations made ingemstones of the ruby type (ruby,spinel) where the red color also resultsfrom compressed space at the Cr3+

lattice sites.The investigation of the two dif-

ferent coordination sites elucidatesthat the Cr3'" (1)/Cr3'" (2) populationratio decides the color itself as well as

GEMS & GEMOLOGY

the intensity of the color change. Thegreater the Cr3'" (1)jCr3'" (2) ratio themore pronounced is the color of thealexandrite in daylight as well as inincandescent light. Due to the denserpopulation of the Cr 3'" (2) centers,but also because of the polarizinginfluence of the adjacent Be ion actingas a central ion, the conditions for theCr3'" ions are very similar as in the Cr3'"doped silicates of the emerald type(emerald, demantoid, hiddenite etc.)as long as the alexandrite is impelled bythe intensity as well as by the spectralcharacter .of the light in which the gemis being viewed. Daylight or any otherwhite illumination with an identicaldistribution of wavelengths interactswith both Cr3+ centers imparting thegreen as well as the red hue in de-pendence on the Cr3'" (1 )jCr3'" (2)population ratio. In incandescent lighton the other hand, which is relativelypoor in short (blue, high energy) wave-lengths but rich in long (red, lowenergy) wavelengths the alexandriteappears red, because the strong band Bwith its culmination at 415 nm ab-sorbs the majority of the scanty shortwavelengths up to about 443 nm thusblotting out almost all of the blueregion. The residual light is beingcontrolled by the broad A band com-plex at 571 A nm, i.e. it absorbsmainly in the green. On account of itscomplex nature the band A also blacksout parts of the bluish green and of .the yellowish green between 625.5-516.6 nm. This extensive absorptionof almost all the shorter wave-lengths results in the complementarypurplish red color: the characteristicraspberry red of the alexandrite inincandescent light. From this it may

FALL 1976

be concluded that the smaller the Cr3'"(1 )jCr3+ (2) population ratio, i.e. themore concentrated the Cr3+ centers,the more more important the absorp-tion effect of band A.

After all these reflections it must,however, be considered that the Aband complex is neither the sole northe principal factor for the quality ofthe color change but rather the corre-lation of both bands A and B. Aboveall the absorbence of band B, that is tosay, the sum of the absorbing centersin band B must apparently reach acertain critical value to completelyabsorb the shorter blue and greenwavelengths which are in any wayquite scanty in incandescent light.

In summarizing it may be recapitu-lated that the alexandriteeffect ma-terializes in that chromoforouschromium atoms in the state as Cr3+ions occupy octahedral coordination.sites in the crystal which are otherwiseheld by AI3'" ions. These sites differ insize; some of them are of normallargeness, others are constricted. Thecolor change is provoked when incomparison to the Cr3'" (2) ions onregular spacious sites sufficient Cr3'"ions can slip onto the compressed sitesCr3+ (1), that is when a certain criticalvalue of the Cr3'" (1 )jCr3'" (2) ratio isreached. The distribution of the Cr3'"ions onto the different transition-metal sites certainly depends upon thegrowth conditions of the alexandrite;in order to force enough Cr3'" ionsonto the compressed lattice sites eithervery high temperatures and lowpressures or low temperatures withhigh pressures are necessary. Such ex-treme conditions happened to occurvery rarely in the birth chambers of

208A

the alexandrite (not to mention theextreme rarity of the chromium) sothat natural specimens with a goodcolor change are purely accidental!

The foregone explanation may bepreliminary and perhaps it is an over-~iniplificati0n but it bad to be adaptedto different grades of seientific educa-tion of the readers. To understandthe miracle does not deny its ex-istence!

In the above considerations the ironwas not mentioned because it exertsno influence on thealexandrite effect.If present in sufficient quantity as toaffect the color by imposing its ab-so~tion ~pon. that of ~e chromium,Fe may impair the punty of the huesin day and incandescent light. Thusiron is often responsible for thebrowrilsh tint of some alexandritesfrom Sri Lanka. In addition, iron actsas a powerful inhibitor in that it

prevents the alexandrite from emittingluminescence.

Behavior under shott wave radiation:Table 1 reveals that iron partici-

pates with 0.7% of the total amount ofchemical elements in the compositionof the Manyaraalexandrite.This isapproximately three times that ofchromium. Yet, it appears to be insuf-ficient for completely suppressing theluminescence, forcontrary to alexan-drites from other sources which arenormally inert to shortwave illumina-tion, the Manyaril alexandrite glowsfaintly in UV light but does notrespond to X-rays. This·distinct Iurni-nescenceservesas a welcome virtue ofdistinction from (a) genuine alexan-drites fromptber deposits and (b)both types of synthetic alexandritepresently known. Table Ill offers aclear survey of the behaviour under

208B

Figure 5. A "peres!" of dark blackish brown biotite flakes (32x).

GEMS & GEMOLOGY

Figure 6. A section of a''fi"ptprliW' int;Jusion, l.e. a I1Sttijlly h,tNflild fracturemarked by a system of inteTf:OmmunicBting channels of mid_'Iquid (40x).

FitJP'ie:1.~exagonal arrangement of color zones alternatively dark and-;!MIII',- (32x1.

FALLl~76 208(:

short wave radiation of the variousgenuine and synthetic alexandrites.

The immediate deduction to beassumed from the positive response ofthe Manyara alexandrite to UV radia-tion would be that it contains less ironthan the alexandrites from the otherlocalities. This observation is also con-firmed by the high degree of trans-mission of the Manyara alexandrite inlong wave UV light. Under short waveUV radiation all natural alexandrites aswell as flux grown synthetic alexan-dri tes are opa que, while theCzochralski pulled synthetic alexan-drites are transparent.

InclusionsThe internal paragenesis of the

Manyara alexandrite is typicallymoulded by its mode of formation andthe component minerals of the mother

rock. Therefore it is not surprising toencounter some minerals of the ex-ternal mineral association such asactinolite and biotite. The actinoliteoccurs in the form of fibres or stalkseither discretely distributed or thenconcentrated in irregular masses or asslightly divergent sheaves. Close anddense arrangement of fine actinolitefibres may cause chatoyancy in somespecimens. The biotite also eitherassembles in the well known form ofso-called "books" or hovers as indi-vidual flakes in the body of its hostgem (Figure 5). In one small alexan-drite an individual, well formed crystalof apatite was observed.

The majority of those Manyaraalexandrites excelling in inclusionscontain mainly fanciful liquid inclu-sions which traverse the whole crystal

TABLE IIILUMINESCENCE Transparency Opacity inALEXAN-

long wave short wave in long wave short waveDRITE UV light UV light X-Rays UV light UV lightFROM:365 nm 254nm 365 nm 254 nm

Lake Manyara medium weak inert good opaqueTanzania dull red yellowish

Novello inert inert inert medium opaqueRhodesia

Ratnapura inert inert inert medium opaqueSri Lanka

TokowayaUral Mts. inert. inert inert medium opaqueSiberia

SyntheticAlexandrite

flux grown distinct very weak inert very good opaque(Patterson) orange red reddish

Czochralskipulled strong red strong red inert excellent transparent

(Dr. Morris)

208D GEMS & GEMOLOGY

at random and display the character-istic appearance of secondary inclu-sions or partly healed fractures. Theirpattern is decided by liquid filledtubes or channels which are moreloosely disseminated or then by denseand irregular arrangement of ramifyingveins and capillaries (Figure 6). Nosmall number of Manyara alexandritesdisplay periodic growth either by alayered distribution of the color or apseudohexagonal zonal accord of lightand darker shades (Figure 7).

Unfortunately the supply of thismagnificent precious stone from LakeManyara will presumably always re-main rather uncertain or at least veryirregular and in any case too scarce asto satisfy a receptive market, and theexhorbitant prices which were askedfrom the beginning did not help topromote the gem's popularity.

Acknowledgements:The author wishes to extend his

gratitude to Prof. Dr. M. Weibel forthe chern. analyses of the native rocksand the alexandrite crystals as well asto G. Brunner for running the absorp-tion curves (both at the Institute ofCrystallography and Petrology inZurich) and to C.A. Schiffmann (Gem.Lab. Gubelin, Lucerne) for his assis-tance with the gemological experi-ments,

REFERENCESBank, H. (1964), Alexandritvorkommen inStklrhodesicn. Z. Dt. Gemmol. Ges. 47,11-15.Bank H. (1974), Smaragd, Alexandrit undRubin als Komponenten einer Paragenesevom Lake Manyara in Tansania. Z. Dt.Gemmol. Ges. 23, 1.

Bank, H. & GUbelin E.J. (1976), Smaragd-Alexandritvorkommen von Lake Manyara,Tansania. Z. Dt. Gem mol. Ges.

FALL 1976

Carstens, H. (1973), The Red-Green Changein Chromium-Bearing Garnets. Contr. Min.Petr; 41, 273-276.

Farrell, E. F., Fang, J. H., Newnham, R. E.(1963), Refinement of the ChrysoberylStructure.Am. Min. 48, 804-810.

_______ (1965), Crystal fieldspectra of chrysoberyl, alexandrite, peridot,and sinhalite. Am. Min. 50, 1972-1981.

Fersmann, A.E. (1929), Geochemische Mi-gration der Elemente Ill. Smaragdgruben imUralgebirge. Abh. prakt. Geologie und Berg-wirtschaftslehre. 18.Grum-Grzhimailo, S.V. (1946), On theColor of Alexandrite Crystal, Zap. Vser.Min. Obsh, 75, 253-255. Translated by G.Tunnell. Gems & Gemology, Spring 1949,143-145.

Giibelin, E. J. (1974), The Emerald Depositat Lake Manyara, Tanzania. The LapidaryJournal, No.5.

Hassan, F. and El-Rakhawy (1974),Chromium III Centers in Synthetic Alexan-drite.Am. Min. 59,159-165.

Mwakisunga, M.A. (1976), Gemstone occur-rences in Tanzania with special reference toMayoka Emerald Deposit, Lake Manyara,Tanzania. (Thesis to obtain a Ph.D. title)

Nassau, K. (1974/5), The Origins of Color inGems and Minerals. Gems & Gemology, Vol.XIV, 12, 354-361.

Newnham, R., Santoro, R., Pearson, J."andJansen, C. (1964), Ordering of Fe and Cr inchrysoberyl. Am. Min. 49,427-430.

Okrusch, M. (1971), Zur Genese vonChrysoberyll - und Alexandritlagerstatten,Eine Literaturubersicht, Z. Dt. Gemmol.Ges. 20, 1971, Heft 3,114-124.

Pool, C.P. (1964), The optical spectra andcolor of chromium containing solids. J.Phys. Chem. Solids, 25, 1169-1182.

Strunz, H. (1968), Atomare Struktur, Eigen-schaften und Klassifikation der Edelsteine.Aufschluss 18,69-102

Thurm, R. E., Smaragde vom Lake Manyarain Tansania Z. Dt. Gem mol, Ges. 21, Heft 1.

Troup, G.J. (1969), The Alexandrite Effect,Austr. Gemmologist Febr. 1969,9.12.

White, W. B., Roy, R., and Crichton J.M.(1967), The "Alexandrite-Effect": an opti-calstudy.Am. Min. 52,867-871.

209