A merican M ineralog is t, Volwne 6 3, pages 92 4 -9 29, I 97 8

Implicationsof a titanium-richglass clod at OceanusProcellarum

Unsule B. MenvtN H aruard- S miths onian C ente r fo r A strophysi cs Cambridge,M assachusetts 02I 38

lNp Devto Wnlrrn Departmentof Geology,Haruard Uniuersity Camb ridge, M assachuset ts 02 I 38

Abstract

A friable clod consistingof spherulesand conchoidalfragments of red-blackglass with an exceptionallytitanium-rich composition (-16 percentTiOr) was discoveredin the 2-4 mm fractionof Apollo l2 sample12033. Despite a diligentsearch of the remainingApollo l2 soils, no similarfragment has beenfound in the samplesfrom that or any other lunar mission.By analogywith the glassspherule clods collectedat the Apollo 15 and Apollo 17 sites,this uniqueclod of glassmay representa lunar magmaof which no crystallineequivalent has yet beenfound. The probabilityis smallthat magmaswith the compositionof the red-blackglass could be parentalto high-K titaniferousApollo I I basaltby low-pressurefractional crystallization. Red-blackglass composition cannot be parentalto any other lunar basalttypes. If red-black glassis a magmatype, and not an accidentallyremelted cumulate, its existencerequires that gross heterogeneityof TiO, characterizelunar magmasource regions and/or that simple modelsof magmagenesis cannot produce this type.

Occurrence suggestan affinity with mare basaltsrather than no- rites. A friable spheruleaggregate, which looked like a When the clod was discoveredin the summerof dollopof blackcaviar (Fig. I ), wasrecognized as an 197l,it wasimmediately mounted in epoxyand cut exceptionalparticle of lunar soil during a macro- into a thin section.As no more of the clod remains, scopicsurvey of Apollo 12 sample 12033,66.The the opportunitywas thus lost to treat this 2-mm samplerepresented a layerof light gray soil that was particleas a consortiumrock (an approachthat was exposedat a depthof a few centimetersin the wallsof developedlater) in order to determineits age and a trench the astronautsdug through the rim of Head trace elementcontent and to analysethe spherule Crater.The light color of the soil signaledan abun- surfacesfor micromoundsof condensedvolatiles, as danceof small ropy fragmentsof KnErp-richglass wasdone subsequently on theApollo l5 greensphe- coatedwith gray-whitedust. The ropy glasseshave rulesand the Apollo 17orange and black spherules beeninterpreted as eitherray materialfrom the crater (for reviewsee Butler and Meyer, 1976). Copernicus(Meyer et al., l97l) or as shock-melted A singlesoil particle,regardless of how exotic its ejecta from a more local impact crater that pene- composition,hardly seemsadequate for formulating trated thin mare flows and excavatednoritic materi- a petrogenetichypothesis. In order to find additional als (Quaide et al., l97l). Although the dark clod, examplesto be usedfor surficialanalyses and other which differedstrikingly in textureand composition studies,an arrangementwas madeby U. B. Marvin from all other particlesin the Apollo 12 soils,oc- in 1977to examineand catalogue the Apollo l2 soils curredin a sampleconsisting of 40 percentropy glass in storagein the CuratorialFacility at Houston(cata- fragments,no geneticassociation between these par- logue in preparation).A diligent searchof all the ticle types need be assumed.On the contrary, high fractions coarserthan 0.55 mm failed to locate a valuesof titanium and chromiumin the dark glasses single spheruleor spheruleaggregate of red-black 0003-004x/7 8/09 I 0-0924$02.00 924 MARVIN AND WALKER' TITANIUM-RICH GLASS CLOD 92s glass.Inasmuch as no similarclod has been found in -'r'i'ti anyother lunar soil, the original clod remains unique. Texture In thin sectionthe clod is nearlyopaque in trans- mittedlight, but the extraillumination provided by a substagecondenser shows that the main constituent is a very dark red-blackglass. A single50-pm frag- ment of clinopyroxeneis locatednear the centerof the section,but most if not all of the many other small transparentphases are crystallitesof olivine and pyroxene.Reflected light (Fig. 2a) revealsthat the clodis a porousaggregate of spherules(5- I 30pm dia.), conchoidalspherule fragments, and irregular chunks of glass.Some of theseglass chunks and partially devitrifiedfragments of spheruleshave been compressedtogether into new aggregates,but clasts incorporatinglithologies other than red-blackglass and its partial devitrificationproducts were not ob- served.About 35 percentof the glassyparticles are brownishin colorand contain well-developed crystal- lites of ilmenite,olivine, and/or pyroxenewith a maximumsize of 2.5x 1.4pm (Figs.2b,2c,2d). The remaining65 percentof the spherulesand fragments are dark red and cloudy,with opaquesubmicron Fig 2. Views in reflected light of the polished thin section ofthe crystallitesdistributed uniformly or alignedin feath- glassy clod. (a) View of the entire particle showing the distribution erypatterns in theglass. The presence of theseminute of spherules in the porous matrix of conchoidal glass fragments. pre-existing portion crystallitesgives the dark glassesan exceptionally The dark diagonal streaks are fractures. (b) A of the section illustrating the contrast between the highly reflective bright reflectivity which makes the mafic silicates spherulesand fragments containing minute, uniformly distributed look dull grayin comparison.This is theopposite of opaque crystallites and the fragments of darker glass with the usual reflectivity contrast betweenglasses and prominent crystallites (c) Two particles with crystallites in maficsilicates. different stages of growth. The particle at the upper left has a delicate dendritic pattern of ilmenite. The one at lower left includes Composition ilmenite crystallites plus one elongated olivine crystallite which occupies the band between the lobe at the left and the larger Electron microprobe analysesof the homoge- portion at the right. The remaining matrix material is brownish neous-appearingdark spherulesand fragmentsshow glass that partially covers the olivine, making it indistinguishable in this photograph. (d) A "homogeneous" spherule and an irregular mass of glass with the more uniform variety grading into the more coarsely crystalline type.

thenarrow compositional range listed in TableI and plotted in Figures3 and 4. The red-blackglass is olivine-normativeand containsabout 3l weightper- cent of normativeilmenite. The growth of olivine, ilmenite,and a few clinopyroxenecrystallites in the red-blackglass produces a lighter-colored,quartz- normative, pyroxene-plagioclase-richresidual glass (Table l). Most attemptsto analyzethe crystallites failed to yield mineralcompositions unmixed with glass.However, two good crystalliteanalyses are of olivine(Forr) containing 0.3 percent Cr. Thisvalue of Fig. l The 2-mm clod of spherulesand fragments of red-black chromiumin a lunarrock suggestsa genetic relation- glass from Apollo 12 sample 12033,66. shio with mare basaltsrather than with noritesor 926 MARVIN AND WALKER: TITANIUM.RICH GLASS CLOD

Table l. Electronmicroprobe analyses of the glassclod Glassspherules and clods A. The averageand range of valuesin 8 particlesof red-black Impactand volcanism are both glass-forming proc- glass "homogeneous" esses,and both aregenerally believed to havecontrib- B. Bulk compositionof the clod,derived from theaverageof8 100 pm broad-beamanalyses recalculated to 100percent uted glassyparticles to the lunar regolith.To distin- C, The compositionof I particle of brown quartz-normative, guish impact from volcanicglass is sometimesa residualglass containing crystallites of olivineand ilmenite relatively straightforward exercise:impact glasses may havethe compositionof non-igneoustarget ma- terials,whereas volcanic glasses should duplicate or

Avg. Range showa closegenetic relationship to the compositions of volcaniclavas. It becomesmore dimcult to distin- - ?? nt S102 33.44 31.83 35.14 42.8r guish betweenimpact and volcanic origins when T102 ro.J/ ro.uz - ro./o r). oo 10.54 glassyparticles are the product of impacts into ig- 41203 4.60 3.90 - 5.41 6.2r 5.15 neousrocks or remeltedbreccias with largepropor- Ct20g 0.84 0.71 - 0.98 1.00 0,22 tions of igneousclasts. Particles of eitherimpact or Fe0 23.87 23.42- 24.43 zJ.04 20.88 glass in association ltn0 0.30 0.24 - 0.36 0.30 0.30 volcanic would logicallyoccur Mgo 13.00 12.04 - 13.87 1a <1 8.15 with unmeltedfragments of equivalentcomposition. Ca0 6.27 5.14 - 6.90 6.69 LO.74 Thus many lunar glassesare not easilyclassified. Na20 0.05 0.00 - 0.13 0.15 0.30 Spherulesof clearglass in a wide spectrumof col- Kzo o,r2 0.03 - 0,42 0.40 o.7r ors are ubiquitousin the lunar soils,and thoseindi- N10 0.02 0.00 - 0.06 o.o2 0.03 viduals that mimic the compositionof individual Pzos n.d. 0.20 0.17 mineralsor of bulk soilsare readily ascribedto im- qn^ - v-5 0.08 0.01 0,16 0.r2 0.09 pact. If impact was the processresponsible for pro- spheruledeposits, the highland AOEAI 98.96 100.00 100,09 ducing aggregated soils should include discretelayers or lensesrich in

anorthositicgabbros (Marvin et al., l97l). Oneanal- ysis yielded a normative compositionof 94 percent -A PSEUDOTERNARY I II HIGH-K clinopyroxene(EnrrFsnoWols). The singlepyroxene 2'A il,t7LoW-K LIQUIDUS S.ORANGESOIL fragment not occurring as a crystalliteis more Ca- PROJECTION 4'At2 ILM rich (EnarFsroWor,). FROM 5=Lt6 6=At2 PtC-PtG In an attemptto determinethe bulk compositionof CnAL2St2Os 7-A t5 the clod, a numberof 100-pmbroad-beam analyses S.GREENGLASS, L24,VLT were made at random on the section.The clod is highly porous,and the results(which wereunexpect- edlyuniform) summed to about84 percent. The aver- age values,recalculated to 100percent, are given in Table l. We believethat the highercontent of alu- minain the bulk analysis(column B us.A) represents smallamounts of AlzOspolishing compound trapped < PYROXENE> in pore spaces. The analyticalresults show that the red-blackglass l-< Lrcrxpr;sflTASES > | St0z is richer in TiO, by about 6 percentthan the orange glassof theApollo I I andApollo l7 soils,and by 4 to Fig. 3. Liquidus surfacefor a portion of the FmO-FmTiOe- 5 percent than some red-black glassesfrom the (Fm, * Ca.Fm)SLOeplane projected from CaALSLOs.Liquidus phasevolumes for armalcolite,ilmenite, olivine, pyroxene,and Apollo l l site analyzedby Keil et al. (1970).No mare olivine * pyroxene are shown from Walker et al. (1975). basaltssampled have equivalent compositions. Thus Equilibrium crystallizationof red-blackglass produces residual the.darkspherule clod representsan exceptionallu- liquidswhich follow the heavyarrow to A, B, C, thenD (seetext). .nar composition.The clodis especiallyout of placeat Numberson compositionfields are keyedin box to Apollo and glass possible the Apollo 12 site, where the basaltsof Oceanus Luna basalticrock and types.The only suiteof rocks that can be related to red-black glass by crystal-liquid Procellarumcontain a maximumof only 5.5percent fractionationat low pressureand low pOz(to which theseliquid Tior. relationsapply) is thetitaniferous basalt suite ofApollo 1I and 17. MARVIN AND WALKER: TITANIUM-RICH GLASS CLOD 927 milky glassbeads, produced by the crateringevents St02 that have sculptured the crustal anorthositesand anorthositicnorites during the past 4 billion years. PSEUDOTERNARY Yet no such concentrationswere observedat the LIQUIDUS Apollo 14or Apollo l6 sites,where the spherules that PROJECTION are present(including feldspar-rich glasses) are rela- FROM FurTrO3 < PYROXENE> tively dispersedcomponents of the regolith. At the two siteswhere large volumes of spherulesiof uniform composition have been discovered,the glassesare ILH ARML highly mafic and not of highland-relatedcomposi- tions. In a negativeway the absenceof'aggregates < PLAGIOCLASE> rich in feldspathicspherules in the higlhlandsoils, whichhave been generated by impact,lends support /.or,r, r, to the hypothesisof volcanic origin for the aggre- gatedconcentrations of greenspherules nLear the rim of MareImbrium at theApollo 15site ancl the orange and black spherulesat the margin of the Littrow Ftr,r2SrOa CnAL2Sr2Os Valleyat the Apollo 17site. Fig. 4. Liquidussurface for SiOz-FmzSiOr-CaAlrSirO,projected The hypothesisthat thesespherules originated by from FmTiOr. Liquidus phasevolumes for plagioclase,olivine, the eruptions of fire fountainsalong fissiuresat the pyroxene,and olivine + pyroxeneare shownfrom Walker el a/. (1975). rims of the mare-filled primarily The traceof the armalcolite/ilmenitereaction is shownfor basinsis popular low pO, and low total pressure.Crystallization of red-blackglass becauseof the largevolumes and unifonn composi- producesresidual liquid which followsthe heavyarrow from A to tions of the glasses,the presencein a few ripherulesof B to D (sameletters as in Fig. 3).This sequenceintersects the high- large euhedralolivine phenocryststhat must have K titaniferous basalt suite from Apollo I l. Numbers on existedin a largervolume of liquid prior to the dis- compositionalfields are keyedin box to titaniferousbasalt types. The displacementof the phase persal of glass olivine*pyroxene boundary is droplets,and the occurrenceon the shown for total pressureof 5, 10,and 20 kbar as well as at low spherulesurfaces of condensedvolatile elements not pressure.Red-black glass would intersect this equilibriumat 15-20 found in comparableconcentrations elseu'here on the kbar or about 300-400km depthon the moon. moon.The hypothesisis popularin spite,of the fact that marebasalts at theApollo l5 andApollo l7 sites any of the other types?(3) What is the origin of this neitherduplicate the compositionsnor showobvious chemicaltype, irrespective of the glass-formingproc- geneticrelationships to the glassesof the spherule essresponsible for its distribution around the lunar deposits,as might be expectedif the spheruleswere surface? alsoof igneousorigin. Both the Apollo l5 andApollo To addressthese questions, red-black glass compo- l7 depositsinclude a few friableclods o,f spherules sitions are plotted in Figures 3 and 4, which are admixedwith glass,mineral, and lithicfragments that liquidusprojections appropriate to titaniferousbasalt aggregatedafter the spheruleswere deposited in the petrogenesistaken from Walker et al. (1975).Figure regolith.We suggestby analogythat the clodof red- 3 projects from CaAlrSirO' onto the plane FmO- black glassin sample12033,66 was projected to the FmTiOr-(Fm, * Ca.Fm)SLO' (FmO is combined Apollo 12 site from a depositrich in high-titanium FeO and MgO), which includesolivine, pyroxene, glassspherules. In the followingsection we explore armalcolite,and ilmenite.The plane is subdivided the inferencethat thesespherules were volcanic in into liquidus crystallizationfields for thesephases; origin. arrows indicate the direction of decreasingtemper- ature on multiple saturation curvesseparating liq- Petrogenesis of the red-blackglass uidus phasespaces. Red-black glass lies in the olivine The existenceof red-blacklunar glassol'bizarre Ti- primary phasevolume and crystallizesolivine until rich composition posesthe following quLestionsfor the titaniferousoxide saturationsurface is reachedat lunar petrogenesis.(l) If substantialquantities of A, in this casethe one for armalcolite.Equilibrium magmaof this compositionexisted, is it possiblethat crystallizationof red-blackglass commences with oli- this material could have been parental to other vine, which is subsequentlyjoined by armalcolite. known magmatypes by fractionalcrystallization? (2) Continuedcrystallization results in the reactionrela- lf such is the case,is the red-blackglass parental to tion betweenarmalcolite and liquid, producingilme- 928 MARVIN AND WALKER: TITANIUM-RICH GLASS CLOD nite at the expenseof armalcoliteat B. Further crys- diminishesthe credibilityof the hypothesis.Further- tallizationdrives the residualliquid further down on more, the singularoccurrence of red-blackglass in the diagram until pyroxene saturation is reached the Apollo l2 soil,where the bedrockis not high-Ti from C to D. basalt,and the lack of reportsof this materialat Figure 4 showsliquidus relations in the part of Apollo I I and 17, wheretitaniferous basalt is com- SiOr-CaAlrSirOr-FmzSiOrspace appropriate to ti- mon, doesnot encouragethe view that thereis any taniferous basalts,projected from FmTiOr. Again intimategenetic connection. one can follow the equilibriumcrystallization se- It may be useful to inquire whethercrystal-liquid quenceolivine + plroxenein this diagram.Note the fractionationat high pressureis capableof producing trace of the armalcolite- ilmenitereaction on this the low-K parentalmagma from eitherthe orangeor liquidus.This projectionshows that the last major red-blackglass. In this inquiry, then, one must as- phaseto crystallizeunder equilibrium conditions will sumethat the answerto the third questionposed is: be plagioclase.This saturationcurve is reachedas "the red-blackglass originates by direct partial melt- crystallizationadvances the residualliquid alongthe ing at depthwithin the moon." This couldbe accom- olivine* pyroxeneintersection. Olivine remains in plishedby partialmelting of an olivine-pyroxeneas- the final recrystallizationproduct of ilmenite,py- semblageat 300-400km. This can be deducedfrom roxene,and plagioclase(even though it is in reaction Figure 4, which showsthat the olivine-pyroxene relationto the final liquids) by virtue of the high cosaturationcurve intersects the red-blackglass com- normativeolivine content of red-blackglass. positionsat 15-20kbar pressure.Slow ascentto the An important feature of this crystallizationse- lunar surfaceof sucha magma,if it maintainedequi- quenceis that the residualliquid from crystallization librium with its olivine-pyroxenitesource region, of red-blackglass intersects the compositionfield of would resultin modificationof the melt composition some of the Apollo 11 titaniferousbasalt samples, in responseto the shift of the olivine/pyroxenesatu- namelythe high-K suite.It doesnot intersecteither ration boundaryto lessolivine-normative composi- the Apollo ll or the Apollo l7 low-K titaniferous tions with decreasingload pressure.In this situation suites(see Fig. 4). Subtractionof -50 percentolivine pyroxenefractionation in the 5-10 kbar pressure and - l0 percentarmalcolite would be necessaryto range(100-200 km) couldyield magmaswith major bring the residualliquid to the appropriatecomposi- elementchemistry similar to the parentof the low-K tions whereolivine and armalcolitereact with the suite.The residuumin this casewould be ilmenite- liquidto givepyroxene and ilmenite. Minor pyroxene saturated.On the otherhand, if the ascentof magma fractionationmay be required,but plagioclasewould segregatedfrom the residueat l5-20 kbar wererapid play no role in this differentiation.The red-black and essentiallyadiabatic, eruption of the titaniferous glasscompositions appear to be suitableparents to red-blackglass could result. This couldthen give rise the high-K Apollo 11suite with regardto KrO con- to the high-K suite by fractional crystallizationas tent as well as with respectto major elementchemis- describedabove. try. The KzO contentof red-blackglass is variable This complexscheme involving partial melting of but may be ashigh as0.42percent.In this respect the olivine pyroxenitewith subsequentfractionation of red-black glass makes a more suitable parent for pyroxene,olivine, ilmenite, and armalcoliteat differ- high-K materialthan the Apollo 17 orangeglass, ent depthsis capablein principle of generatingthe which appearsto be uniformly low in KzO. From major elementchemistry of the titaniferousbasalts as major element considerationsalone, Apollo 17 daughterproducts of the red-blackglass composi- orangeglass could also be parentalto the high-K tion. A family of suchschemes is possiblewith differ- Apollo l1 group by simpleolivine fractionationat ing degreesof melting, depths of segregation,and low pressure.However, neither orange nor red-black subsequentcrystal fractionation.Unfortunately this glass compositionscan be parental to the closely family of modelsdoes not appearto be consistent associatedlow-K seriesfrom Apollo ll and 17.The with the incompatibleelement abundances. A feature failureof eitherglass type to be ableto producethe commonto all of the modelsis that the low-K suite associatedlow-K seriesby crystal fractionation at must be the resultof eitherthe largerdegree of frac- low pressurecasts doubt on the propositionthat the tional crystallizationor the smallerdegree of partial red-blackglass is, in fact, parentalto the high-K melting when estimatedon major elementcriteria. suite.The lack of intermediatedifferentiates between This conclusionresults because the low-K suiteis red-blackglass and its supposeddifferentiates also more aluminousthan the high-K suite.The fractiona- MARVIN AND WALKER: TITANILJM-RICH GLASS CLOD 929 tions discussedinvolve crystallinephases less alu- Acknowledgments minousthan the participatingliquid. Henceliquids We aregrateful to ProfessorsClifford Frondeland CorneliusS. resultingfrom a processleading to a smallerpropor- Hurlbut for providing us with instruction and our interestin tion of liquid should be more aluminous.For in- mineralogy,both terrestrialand extraterrestrial.We thank Patrick Butler stance,the product greater Jr. and an anonymousreviewer for their comments.This of a amount of fractional work wassupported by NASA Grants09-015-150 to J. A. Wood crystallizationshould be morealuminous. However, and NGL 22-007-247to J. F. Haysand the Committeeon Experi- the greater enrichmentof the alkali and REE ele- mentalGeology and Geophysicsof Harvard University. ments,which suggests a smaller proportion of liquid, in the high-K group,is contradictedby the lessalu- Referencesr minouscharacter of this group which suggestsa proc- Butler,P., Jr. and C. Meyer,Jr. (1976)Sulfur prevailsin coatings essleading to a higherproportion of liquid. on glassdroplets: Apollo l5 greenand brownglasses and Apollo Alternate proposalswhich derive red-blackglass l7 orangeand black(devitrified) glasses. PLC, 7, l56l-1581. Green,D. H., A. E. Ringwood,W. O. Hibbersonand N. G. Ware by advanceddegrees of melting of shallowsource (1975)Experimental petrology of Apollo 17 marebasalts. PZC, regionssimilar to thosethat givetitaniferous basalts 6, 87t-893. by lesserdegrees of meltingmay be considered.How- Keil, K., T. E. Bunchand M. Prinz(1970) Mineralogy and compo- ever,these proposals are also unable to givein detail sitionof Apollo ll lunarsamples. PLC, I,561-598. a coherentaccount of the relation betweenthe red- Marvin, U. 8., J. A. Wood, G. J. Taylor,J. B. Reid,Jr., B. N. propor- black glass Powell,J. S.Dickey, Jr. and J. F. Bower(1971) Relative and the other two titaniferousmagma tions and probablesources of rock fragmentsin the Apollo l2 types,for the samereasons. Evidently there is no such soil samples.PLC, 2,6'19-699. simplerelation of themagma types, or elsethe source Meyer,C., Jr., R. Brett,N. J. Hubbard,D. A. Morrison,D. S. regionsproducing various related batches of magma McKay, F. K. Aitken, H. Takedaand E. Schonfeld(1971) are heterogeneouswith respect,at least,to their in- Mineralogy,chemistry, and origin of the Knrrp componentin soil samplesfrom the Oceanof Storms.PLC, 2, 393-411. compatibleelement content. Just this sortof dilemma Quaide,W., V. Oberbeckand T. E. Bunch(1971) Investigations of has led Greenet al. (1975)to postulatethat mare the natural historyof the regolithat the Apollo 12 site.PLC, 2, basaltsource regions are heterogeneouswith respect 701-718. to TiO, content. Ringwood,A. E. and S.Kesson (1976) A dynamicmodel for mare -1722. In summary,there exist several possibilities of re- basaltpetrogenesis. PLC, 7, 1697 Walker, lating other titaniferous D., J. Longhi,E. Stolper,T. Groveand J. F. Hays(1975) basalt types to a red-black Origin of titaniferouslunar basalts.Geochim. Cosmochim. Acta, glassparental liquid generatedwithin the lunarinte- 39,r2t9-t236. rior by partial melting.However, in detailmost of the Wood, J. A., U. B. Marvin, J. B. Reid,Jr., G. J. Taylor,J. F. proposals are unsatisfying,suggesting that simple Bower,B. N. Powelland J. S. Dickey,Jr. (1971)Two unique modelsare not viable.It is possiblethat the red-black microbrecciasfrom the Apollo 11 and Apollo 12 soil samples: petrology glasscould result process mineralogyand of theApollo 12lunar sample.Smith- from some as complexas sonianAstrophys. Obs. Spec. Rep. 333,165-171. the Ringwood-Kessondynamic assimilation model (1976), or, in view of the very limited occurrence, I PLC, 1,2, etc. : ProceedingsLunar ScienceConference, lst, other serendipitousprocesses such as impact-remelt- Znd,etc. ing of titaniferouscumulate rocks may havegener- Manuscript receiued,April 24, 1978;accepted ated the red-blackslass. for publication,May 22, 1978.