Table Of ContentReports 673
I_IN tLO. (1980) Rare _ in the past and pmumt solar wind. InProf. W1EL._.R., BAURH.ANDSIGNERP.(1986")Noble gases from solar energetic
Conf. Ancient Sun (e_:ks1.LO. P_in, L .-%E.ddy andtLB. Morrill), pp. particles r_vealed byclnsedsystem step,bee etching of lunarsoil mineral_
411-421. P_gamon Press,New York. Geochim Cosmoch_m.Acta 50, 1997-2017.
A new titanium-bearing calcium aluminosilicate phase: I.
Meteoritic occurrences and formation in synthetic systems
JULm M. PAQL_I, Jom¢R. BEC_TT2, DAVIDJ. BARBVa3* ANDEDWARDM. STOr2ER2
tSETI Institute, NASA-Ames Research Center, MS 244-11, Moffett Field, California 94035-1000, USA
2Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California 91125, USA
3Physics Department, University of Essex, Colchester, Essex, CO4 3SQ, U. K.
*Current address: Physics Department, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong
(Received 1993 September 28; accepted in revised form 1994 May. 3)
Abstract-A new titanium-bearing calcium aluminosilicate mineral hasbeen identified in coarse-grained calcium-aluminum-rich inelus'enz
(CA/s) from carbonaceous chondri_ Timformula forthis phase, whichwehave temporarily terrnocl_NK," isCa3Ti(ALTih(S_,A1)3014,
and it ispresent inatleast 8of the 20roams-grained CAIs from the Allende CV3 chondrite examined as partof this project. The phase
occurs inTypes AandB1inclusions assmalltabular crystals oriented along twomutually peqxmdicular planesinmelilite.
UNK crystallizes from melts in dynamic crystallization e_ents conducted inair from fourbulk composifiom modeled after Types
A, B1, B2 and C inclusions. Cooling rates resulting in crystallization of LINKranged from 0.5 to 200 0C,%from maximum (initial)
temperatures of 1375to 1580 *(2.Only below1190 °(2does UNK itsalfbegln tocrystallize. Tofirstorder, theprmenc_ orabsence clUNK
from individual experiments can beunderstood interms ofthe compositions ofresidual melts andnucleationprobabilities.
Compositions ofsyrghetic andmeteoriticUNK arcverysimilar interms of major oxides, differingonly inthe small amounm oftrivalent
Ti(%13% oftoud Ti) inm_ri'dc samples. UNK crystallized fromtheType Aanalog issimilar texturally tothat found inCA/s, although
glass, which istypically associated with synthe,ic UNK, is not observed in meteoritic occurrences. Alow Ti end-mwaber of UNK CSi-
UNK") with acomposition near tl_ of Ca3AI2Si4OI4 was produced inafew samples from the Type BI analog. This phasehas not been
foundinthe meteoritic inclusions.
INTRODUCTION ANALYTICAL PROCEDURES
Calcium-aluminum-rich inclusions (CAIs) from carbonaceous Phases were identified and textures observed by backscattared and
secondaryelectron imaging on a J"EOL733 microprobe at the Smithsonian
chondrites have been studied intensely to determine physical
Astrophysical Observatory. Chemical analyses were also performed on this
conditions during the early history of our Solar System microprobe. Accelerating voltage was 15kV and sample current 15 to30 nA
(MacPherson et aL, 1988; and references therein). Coarse-grained asmeasured ona Faradaycupplaced inthe path of the beam. Counting times
CAIs, those that can be studied by standard petrographic for each element rangedfrom 15 to 30 s for the majority of analyses, and
recalculation of therawdatafollowed the method ofAlbee andRay (1970). A
techniques, are host to minerals formed at different stages in the
glasssimilarincompositiontoaTi-fnssaiwtaesused asan internal standard.
evolution of the inclusions so that each mineral constrains specific
aspects of the history of CAIs. This is the first in a series of three EXPERIMENTAL PROCEDURES
papers dealing with a new mineral, which we term "UNK" Starting Materials
following Paque et al. (1986). This work describes the
CAIs are clarified according to textures and relative amounts of major
characteristics of meteoritic UNK from coarse-gained inclusions in phases (e.g., MacPherson er al., 1988). Type A inclusions are ,jurnposed
the Allende meteorite, the conditions under which it can be mainly ofmelilite, with minor amounts of spinel, hibonite, perovskite, and Ti-
fassaite. They can be subdivi&d intocumpaet Type A inclusions (CTA)
produced in dynamic crystallization ex'periments, and the relevance
characterized by rounded shapes and axiolitie intergrowths of melilite, and
of melting to the origin of meteoritic UNIC The second paper deals fluffy TypeAinclusions (I:TA)withirregular shapes. Type B2 inclusions have
with the crystallographyand crystalchen_stryof syntheticUNK sub-equal amounts of melilite, ano,.hite, pyroxene, and spinel, while Type BI
crystallized during a cooling rate ex'periment on a CAI bulk inclusions are charaetarizedbyamelilite-rich mantle surrounding acornthat is
pctrographically similar to Type B2s. Type C inclusions arc similar to Type
composition and hypotheses relating to its formation and
B2s but are unusually rich inanorthite and poor inmelilke. Compact Type As
significance to the early history of the Solar System (Barber et al., andTypes B and C indnsions are generally thought to have crystallized from
1994). In the third paper, the space group of UNK is constrained partially molten droplets. The role,if any, of melts in the evolution of fluffy
using crystals from amixer furnace slag (Barber and AgreU, 1994). Type As isstill controversial (e.g., MacPherson et aL, 1988). We synthesized
four bulkeompositiom representingthe differenttypes of Allenda CA/s aspart
UNK, with the chemical formula Ca3Ti(AI,Tih(Si,AI)3014 ofa larger study onthe crystallization properties ofCA/s.The compositions
(see Barber er aL, 1994), was first noted in synthetic slag samples represent melilite-rich Type A("98": SiO.,_25.8; TIC2, 1.51; A1203, 32.6;
by Agrell (1945) and in a CA/ from the meteorite Essebi by El MgO, 6.65; CaO, 33.7), pyroxene-rich Type B1 ('CA/": SIC2, 31.4; TIC-2,
1.13; A1203, :28.3;MgO,9.9; CaO, 29.1) and Type B2("B2C": SiOT_35.7;
Goresy et al. (1984). The phase has also been noted in synthetic
TIC2, 1.39; AI203, 27.1; MgO, 12.9; CaO, :22.4) and anorthite-rich Type C
samples produced in several experimental studies of CAIs (Paque ('TCAN": SiC_ 38.8; TiO,o 1.31; AI203, 29.0; MgO, 5.77; CaO, 24.9)
and Stolper, 1984; Paque et al., 1986; Beckett and Stolper, 1994; inclusions.
this study). More recently, Floss at at (1992) described an Starting materials were prepared by mixing high purity oxides and
carbonates under ethanol in an automated asam mortar for 5 h or more,
unusually Ti-rich, Al-poor phase probably related to UNK from an followed bydecarbonation at1000*C,melting in airforone dayat eithar1525
Allende Type A inclusion. °C or 1550 °C, and quenching indeionlzed water after removing the oruca'ble
674 Reports
throughthe top of the furnace. Furtherdetails on preparation of the starting
rnatcriaisandthe experimental techniques canbe found inStolpcr andPaque (3 IO/Jm
(1996).
CrystallizationExperiments
A summary, of isothermalcrystalIm_tiosnequencesforthe bulk
compositions used inthisstudy canbefound inPaqueandStolpar(1984) along MEL
with preliminary, results onthecontrollodcooling experiments. C.rystallization
behavior oft,he CA/bulk composition hasalso been studiedbyStolper (1982),
MaePh_rson etaL (1984), Stolperand Paque(1986) and Bcckett etaL(1990).
UNK was P.._thesized for this study."in dynamic .crystallization experiments
performed inairina verticat[ubeDeltcchVT-31furnace.The sample
ternperamrweasmonitoredwithaPt/PIIoP(.ThypeS)thermocouplcealibrated
agaim_themeltingpoirasofAu (1064*C)andPd(I554°C)andplacedintbe UNK
hot spot adjacent tothe samples. Ineach experiment,several powdered sarnple_ A
ofthesynthetic materialwer_ suspended inthehotspot usingPtloops. Sampl_
were heldfor 3h atthemaximum temparamrt of theexperiment (Tm_), tlmn
cooled at a controlled, approximately linear rate. Individual samples were
quenchedindeiomzed water atvarious temIxr'amr_salongthecoolingpathin
ordertodeterminethetemperaturoefappem-anceforindividupahlasesandthe
liquildineofdescent.
RESULTS
MeteoriticOccurrence ofLINK
A survey of 20 coarse-grainedAllende CA]s (fourcompact
Type As, two fluffyType As, eleven Bls, and threeB2s) was
conductedtodeterminetheextenttowhichUNK occursinnatural
materialsand the variabilityin itscompositionand nature of
occurrence.The searchwas carriedoutusingaJEOL 733 electron
microprobe.Initiallyt,hesample was scannedusingbackscattcred
electron(BSE) imaging tolocatepoteatialgrainsofUNIt, which
hasalower BSE albedothanperm'skitc,buthigherthanmelilitc,
anorthite,spinel,and allbutthemost Ti-richfassailes(>15 wt%
TiCh)due toitshigh mean atomicnumber (Z). Ener_ dispersive
spectroscopy(EDS) v_asthenused toconfirmtheidentificatioonf
UN_. Final analyses were made by wavelength dispersive
spectroscopic(WDS) analysis.UNK isa minor constituentof FlO.I.Backscattereoldectro(nBSE)imagesofUNK fromAllen&TypeBI
approximately half of the Type A and Type Bl inclusions inclusion(sa.)NMNH 3682.TabularcrystalosfUNK areorientewdithlong
directiopnarallteolthec-axisinthemelilit(*MEL). (b)NMNH 3529-41.
examined. Itwas positivelyidentifiedinone CTA (4691),one
ThisisthelargesmteteoritgircainofUNK foundtodate.
FTA (A-WPI), and sixType B1s (3529-30, 3529-33, 3529-41,
3655A, 3658, 3682). No likely candidates for lINK grains were slag•Extinctionisparalleltothe lengthofthecrystals,and the
observed in CTAs 3643 or 3898, FTA 3529-46, Bls 3529Y, 3529Z, reflectidtiysslightlhyigherthanthatoffassaite.
3529-21 or 3529-31, orB2s A11-201, RB-83-1 or 3529-32. For two Chemistry of MeteoriticlINK and Coexisting Melilite-:
inclusions, CTA 3529-45 and the Type BI 3732-I, the grain size Table I shows representativeanalysesofUNK from meleodtic
(<I gin) of phases tentatively identified asUNK was so srnalI that samples. The phase containssubequal amounts ofCaO (29-32
itwas not possible to demonstrate conclusively that the grains were wt%), Si(>,.(24-28%), and TiO2 (20--27°,6)with lessA1203
UNK. In the Type A CAIs, UNK can be found in melilile (16--20%) and very low IvlgO (<1%). Calcium consistently
throughouttheinclusiort.InType Bls, UNK occursonly inthe representsone-thirdof the cationspresent,and deviationsin
melilitcmantleand ismost commonly foundinregionsclosetothe analysesfromCa + Mg = 3.0cationsinaformulaunitbasedon9.0
cationscan be attributedto contamination of the analysesby
rim (i.e., the outer quarter of the melilile-rich mantle) of the
inclusion. UNK was not observed in any of the Type B2 inclusions mounding melilite.Figure2a shows cationsofSivs.cationsof
Ti forUNK analysesfrom Type B1 inclusionswith 3.00± 0.05
surveyed, although the sample set included only three inclusions.
cationsofCa ina formulaunitbased on 14 oxygens. Allofthe
Type C inclusions were not studied as part of this project.
Petrography of Meteoritic UN_-AI1 of the meteoritic LINK analysesplotcloseto,but slightlyabove,theSi + Ti = 4.0 line,
crystalsareenclosedinmelilitc.The crystalsaretypicallytabular probablyaconsequenceofassuming thatallTi istetravalent.Ifa
in thin section(Fig.la),although occasionallythey occur as stoichiometryof 9 cationsand 14 oxygens isassumed and the
anhedralcrystalsembedded inmelilitc(Fig.Ib). CrystalsofUNK valencestaleofTi allowed tofloat,then analysesare consistent
areoftenfound attheintersectioonftwo perpendicularplanesof with the presence of small amounts, 7-13%, oftrivalent TL Due to
cleavage({001} and {110};cf.Deer etaL, 1992) in melilite. the limited spatial resolution of the electron microprobe, zoning can
There are usually several to many grains within different meIilile be characterized only in the largest crystals in Type B1 inclusions.
crystals in a given CAI, all less than 10 _m in maximum These are consistent with slightly higher Ti in the core relative to
dimensioR. They are colorless in plane polarized fight, and the the rim (Fig. 3a).
larger crystals display third order interference colors under crossed In Fig. 4, compositions of meleoritic UNK are cast in terms of
nicols consistent with indices of refraction measured by Barber and the three components:
AgreU (1994) on synthetic crystals of UNK from a mixer furnace (Ca,Mg)3Ti_* Ahgi,O_, (Ca,Mg))AhSi4Ol( and
Reports 675
TABU_I. Sde_zd analyses of UNKfromAllende CAI_
Inclusion A-W'PI 4691 3655A 3529-33 3529-41 3529-30 3682 3658 Essebi*
Type FTA CTA BI BI BI BI B1 B1
SiO2 25.7 24.0 24.8 27.3 27.0 27.6 24.0 25.4 24.6
TiO2*" 20.4 24.1 27/ 24.1 23.9 23.6 26.9 26.9 27.1
Al203 19.5 18.0 17.9 17.0 17.6 17.2 16.6 16.2 16.1
MgO 0.77 0.34 0.29 0.49 0.14 0.12 0-39 0.35 0-36
CaO 312 31.7 31,9 31.2 30.5 31.0 30.6 30.6 29-3
Total 98.6 98.1 102.1 100.I 99.1 99.5 98.5 99.5 97.5
Cationsbasedontotal- 9
Si 2.37 2.20 2.21 2.46 2.45 2.51 2.23 2.33 2.30
Ti 1.41 1.67 1.82 1.64 1.64 1.61 l.gg 1.85 1.91
Al 2.11 1.95 1.gg 1.81 1.88 1.84 1.82 1.75 1.78
Mg 0.11 0.05 0.04 0.06 0.02 0.02 0.05 0.05 0.05
Ca 3.17 3.12 3.05 3.02 2.97 3.01 3.04 3.02 2.94
Si+ Ti 3.7g 3.87 4.03 4.10 4.09 4.12 4.11 4.1g 4.21
,El Goresy ctaL (1984); _¢ pha.sealsocontains 1.1wt% oftl_ rare earth elements and0.2%eachof7_.rO-2andHtO2.
All Ti iscalculated as TiO2.
(Ca,Mg)3Ti*'Ti['Si301,. Each open s'ymbol represents the average by a halo of anomalously birefringent melilite relative to nearby
of all UNK analyses obtained from an individual CA[, corrected for portions of the host melilite crystal. This is borne out by measured
melilite compositions at melilite-UNK coataets, which are highly
melilite conm_minafion and variable Ti_+/Ti 'v, by reqaidng that the
variable (Ak12 to Ak_i) but generally (Fig. 5b), though not always
formula unit have 3.00 Ca + Mg cations, 9 total cations and 14 (Fig. 5a), more magnesian than expected based on regional zoning
oxygens. The meteoritic UNKs are composed mostly of of the melilite. The zoning profile shown in Fig. 5b is for melilite
Ca,Til" AhSizO,, with lesser amounts of (Ca,Mg)3AlzSi40,, and in the vicinity of the UN'K crystal depicted in Fig. lb.
Ca3Ti" TiZSi30 t4. The calculated mole fraction of
SyntheticUgCK
Ca3Ti4*Tiz_'Si3Ot4 is sensitive to the assumed composition of co-
existing melilite and the MgO wt% in the analysis, leading to Conditions of Formation and Chemistry-UNK is an
occasional constituent of run products from cooling rate
uncertainties on the order of ±5 tool%. There are no obvious
experiments from all four bulk compositions under conditions
compositional differences among UNKs from different classes of summarized in Fig. 6, but absent from all isothermal experiments.
CAIs, but alI are consistent with the presence of small amounts of Typically, several samples were run under the same conditions of
trivalent Ti and, therefore, growth under reducing conditions. initial temperature (Tin=) and cooling rate (CR), then quenched at
Melilite in CAIs is essentially a binary solid solution between various temperatures. LINKwas noted as being present in Fig. 6 if
the end-member components geh.lenite (C_re: Ca2A12SiO7) and any of the samples produced under the stated conditions of Tm_
Lkermanite (Ak: Ca2MgSi2OT). When viewed in transmitted and CR contained the phase. To eliminate from consideration run
cross-polarized light, U'NK crystals oRen appear to be surrounded sequences that did not include samples cooled to temperatures low
2\
2IN ' j ' I ' I ' 1 ' I ' I ' I
x_LM_ete°ritic_' I Syntheticl _
1.5 1.5 -
.i
imi d
cam
O
_- 1 1 -
.-q
• TCAN!
L) L)
• B2C
0.5 0.5 -
• 98
• CAI
, I , ,I , I , '
0 , 1 , I 1 i , O
2 2.5 3 3.5 4 2 2.5 3 3.5 4
Cations of Si Cations of Si
I_O.2. Relationship between TiandSi inUNK, basedonatotal of9 cations, forallanalyses withCa= 3.00 ±0.05cations. The solid lineindicates Si +Ti = 4.0.
(a) Meteoritic TypeB1and(b)Synthetic.
676 Reports
2.8
a • 98
•3* .4+ .
2.4 fCa,Mgk_T! z T= S=30=4 B2C
2.2
3529-41
c-
o 2 Type B1 Type B!
L)
Essebi
1.8
Type A
1.6
1.4 ' ' I I I
2 4 6 8 10
Distance (_m)
(Ca.Mg,_T:.'kl,Si O,,
(Ca,Mg)aAlzSi4034 (Ca,Mg)_Ti:'Al_$i _O:4
2.6
b
FI_. 4. Compositions in tool% of _'r_¢ (dosed symbols) and met¢_'itie
2.4
(open s3_nbolsU)NK intermsof thethree components (Ca.Mg)_Ti_'A12$i:O:4,
2.2 (Ca,Mg)j_h_Si(01( , and (C..=3vig)_Ti:_Ti4-*SAin3a0l:y(s.esof UNK were
¢3 correctefdorsmallamounts(<I0wt%) ofmeliIitc¢ontaminatiosnuchthat
r- 2 98-39 cations of big + Ca =3.00 in aformula unit based on a sum total of 9.00
m Tmax=1500°C catiom and 14.00 oxygens. Among analys_s of synthetic UNIL grainsfrom
CR=20'_C/hr
1.8 intergrowths are notplotIed becau._ the bulk composition ofthe contamination
wasuncertain. Also, Si-UNKs were assumed tohave Ti_+;ri'_, =0becauseTi
1.6 contents aretoo low(<I wt% asTiO2_)to allow quantitative calculations.
1.4 J _ ! ... 1 analyses all fail near the Si + Ti = 4.0 line on the graph. Those
0 5 10 15 20 analyses plotting significantly above the line reflect contamination
Distance ktm) by mtergrowths of melilite, glass, anorthite, pyroxene and a poorly
charactm'ized hi# Z phase. Compositions of meteoritic and
FIG.3.ZoningpattcrinnUNK from(a)NMNH 3529-41,aTypeB1CAIfrom synthetic UNKs are very similar with overlapping ranges in major
Allcnd¢a,ndCo)98-39(Tmax=1500°C,CR =20=C/h).Cationsarchascdo_
oxides (Tables 1-2; Fig. 2). Only in Ti_+ contents are there
a sum total of 9. Traverses are instraight linesfrom onecrystaledge through
thecentertotheopposite¢dg,'. significant differences (Fig. 4). The Si-UNKs crystallized in our
experiments are compositionaliy distinct from both the Ti-rich
enough to enter the range in which LINK was likely to crystallize, meteoritic and synthetic UNK. This phase is characterized by 4
only those series of experiments extending to quench temperatures cations of Si in a formula unit based on a total of 9 cations and
below 1100 °C are plotted. In the CAI (Fig. 6a) and B2C (Fig. 6b) contains very litre Mg or Ti (Table 2; Fig. 2b). It is possible that
bulk compositions, UNK ¢rystalliz_ over a restricted range in T_= Si-UNK and UNK analyses shown in Fig. 4 lie on opposite limbs of
and CIL Only those CAI experiments cooled from Trrm = 1375 and amiscibility gap.
1420 °C with CR _<50 =C/h, and B2C experiments cooled from
Textural Relationships-In an individual experknental charge,
1400-1500 °C with CR < 20 °C_ produced UNIC LINK failed to UNK occurs in one and only one of the four distinct i_trographic
appear in experiments at hi_er _!ower T,,_ or at higher cooling
associations iltuslmted in Fig. 7 and summarized by run conditions
rates. In contrast, UNK crystallization occurred over viru_ly all in Fig. 6. (I) UNK may occur as small (several/zm) tabular
conditions examined for the 9g bulk composition ('Fig. 6c). The crystals oriented along planes in melilite or fiIIing the acute ends of
data base for TCAN (Fig. 6d) is insufficient to characterize the glass pockets (Fig. 7a). This t)pe of ocourrence has been observed
crystallization behavior of UNK.
only in experiments on the 98 bulk composition with Tr_x = 1500
The synthetic UNK occurs in two distinct compositional °C and CR = 2-20 °C_. Meteoritic occurrences of LINK are
varieties: a Ti-rich variety (UNK) similar to that found in texturally similar, although _ass has not been observed. (2)
meteoritic samples that can be described in terms of the two Synthetic LINK from the 98 bulk composition can also form thin
components C,sA_,S_O, and C_Ti_'AI2Si_OI( and a Ti-poor, Si-rich (5-10 _m) rims on melilite laths along with small prismatic
variety (Si-UNK) on the same composition line, essentially crystals protruding from the ends of the melilite laths (Fig. To). (3)
ca,AI_Si40:(. Like its meteoritic counterpart, synthetic LINK is In CAI experiments, UNK is typically part of an intergrowth
characteristically hi_ in TiO: and low in MgO (Table 2). Zoning containing anorthite, melilite, pvroxene and _ (Fig. 7c). One
in the largest crystals is similar to that found in meteoritic CAIs experiment on B2C also produced this texture. (4) UNK often
with cores higher in Ti and lower in Si than rims (Fig. 3b). crystallizes from the residual liquid between major phases in the
Melilite surrounding larger UNK crystals is -Akso (Figs. Yo, c), sample (Fig. 7d). There is no obvious correlation between
similar to that found surrounding UNK crystals in some Type BI composition and textural type among the Ti-rich UNK.
CAIs. In Fig. 2b, cations of Si are plotted vs. cations of Ti for The textural habit of Si-UNK is quite different from the Ti-
synthetic UNK and Si-UNK analyses with Ca = 3.0 ± 0.05. The bearing variety, appearing as a fibrous mass when viewed in
R_r_ 677
16 6O
fl c
14
¢D ¢,50 UNK t 98.2o
-_12
_4o '; Trn_. -ISO0"C
3
,_30
N 6
4691
a 2o
CTA
IO
2
0 :.1.,, , ,,,,,.I ==.! .... r.... i. .,
10 20 30 40 50 60 0 5 10 15 20 25 30 35 40
distance (Ixm) distance (IJ.m)
b 60 d 6O
50
3 3o 3o
= 20 3529-41 _ 2o
°,oJ \
TypeB1
-'10
0 ' 0 i l,. _,, 1 ,- _ • 1,-- I_,, I
0 20 40 60 80 100 0 10 20 30 _0 50 60 70
distance (lim) distance (ixm)
FIO.5. Zoninginmelilit¢surroundingoradjacenttoUNK. (a)Melilit¢surroundingUNKgraininNM'NH4691,aCompactType AfromAllend¢. Co)M¢lilite
surroundingL.'NKgraininNMNH3529-41,aTypeBl inclusionsfromAllende(seeFig.Ib). (c)M¢liliteadjacenttoLINKloaned asarimonmelilitelaths(98-
20;Tmax= 1500°C,CR=200°G_h).(d)M¢lilit¢compositionsurroundingUNKthatcrystallizedfrompatchesofliquidwithinthemelilite(98-56;Tmax"1500
°c,CR =2_).
transmitted light.Figure 7e shows a backscattered electron image 4M'gAI20,R,+p)7Cap,12SL1Ola(Si.t_,'_=tc)
of one example. A very bri_t, hence, relatively high mean atomic
number phase co-crystallizes with the Si-UNK. The phase contains 4Ca._MgSi:OT(Ak)+9Ca_2Si_.Os(=)+ 2Ca2DJ2SiO'/(o=).
significant concentrations of sulfur but has little, if any, Ti (i.e., it Equ.(l)
might be oldhamite, but it is not perovskite or UNK). Due to the
However, some ofthe UNK-bearing glasses are not in equilibrium
small gram size, the composition of this phase could not be
with spinel, based onthe reaction
quantified.
Factors Controlling the Crystallization of UNK in Kxper- MgO(li_+AllOz(liq)=MgAl20_gsp)
Iments--Based on our results, UNK crystallizes from residual melts
and Berman's (1983) activity model for oxide components in the
derived from a variety of CAI bulk compositions analogous tothose
melt This is probably due to overgrowths of pyroxene and/or
of meteoritic inclusions, but the factors controlling the appearance
melilite onthe pre-existing spinel grains. In addition, nearly all of
of UNK are subtle because several sets of samples from the CAI
the glasses coexisting withUNK are supersaturated with respect to
bulk composition with identical Tmx, cooling rate, and quench
anonhite based onthe reaction
temperature, differ only in the presence or absence of UNK. This
suggests that the statistical nature of nucleation plays a role (e.g., CaO(lia3+AI2OxliI)+ 2SiO2(liq)=CaA12Si20s(an)"
Gibb, 1974) or that the appearance of UNK is Controlled ind{recfly
through changes in interstitial melt composition or boundary layer Equation (1) is, therefore, an inappropriate basis for interpreting
our ex-'periments. We can nevertheless crudely constrain the
formation caused by crystallization of one or more of the phases
thermodynamic properties of the phase based on exchange
melilite, spinel, anorthite and fassaite. Analyses of glass from
equilibria involving components in the melt (liq). For the
multiple samples run under identical conditions are given in Table
3 and plotted in Fig. 8. Glass in samples containing UNK is Ca3A12Si4Ol4component (of.Fig. 4), the reaction
significantly higher in MgO and lower in CaO than glass fi-om
3CaO(lic0+A1203(li¢0+4SiOxliq) = Ca3A12Si4014(UNK) Equ. (2)
samples run under identical conditions that didnot produce UN-K.
The phases spinel (sp), pyroxene, melilite (~Akso; reel) and is applicable. An equilibrium constant for reaction (2) can be
anorthite (an) may also occur in UNg-bearing experimental run written..
products and can potentially be used to constrain the
_UNK o
thermod)namic properties ofUNK. Forexample, ifmel]lite, spinel
inK.(2)=in[ asi.unR ] -aO...a)
and anorthite were simultaneously in equilibrium with UNK, then
(a_o3'(a_?,o,)(7a;0:3' RT ' Equ.(3)
the activity of the Si-UNK component would be fixed via the
reaction
678 Reports
1800 1800
_''1''''t .... 1''''1''*' ' ' ' ' i ' ' ' ' I .... I .... l ....
®
16O0 1600
O
o o Z=8 o.: • v
G 1400 14O0
• @ • • %
t •
1200 j 121)(I
UNK alv_m
r@l UNK pl¢_nL inler_J;il to_ l_,e s
Si-UNK proem
UNK I_"_¢n[. wilhm _idn_ I
t i UNK _h,.eat I i
O UNK pr_nl. _ of ifllergtwwlh } _' UNK _'=_,¢nLbort_r aloeg_lilllc gmm_ I
8O0
800 ' , I , , , , t .... I .... I ....
-1 0 1 2 3 4 -! 0 ] 2 3 4
log [Cooling Rate (°C/hr)] 10o[Cooling Rate (°C/hr)]
1800 1800
.... t .... t .... I ' ' ' ' I '@' ' '
-Nrc-1 ®
1600 1600
• • I
[TCAN I
- O •
]4oo
-_ 1400
-=I200 b_ 1200
UNK ilb_ent 1 r • U._-Kabf_.m I
I000 nl UNK FeM:m. inler,,uti.d tootherpha_, [ 1O00
O UNK p_:_nt, pan ormiercmwlh J
800 .... I , , _ , I _ , _ , I .... I .... 8O0 I I I I ] ' , T _ ] i t ; .t. I , , I , , ,
-I 0 I 2 3 4 -1 0 I 2 3 4
log [Cooling Rate (°C/hr)] log [Cooling Rate (°Clhr)]
FIG.6. Experimental conditionsunder whichUNK isproducedandthe resulting textures.Eachpointrepresents aseriesofexperimenm inwhich samples were held
atamaximum t_rapevamrefor3It,cooledataconstantrate,anddrop quenched atvariousmmperatures. Open symbolsindicaz_that one ormoreexperimmts inthe
series produced UNK. Onlyseries including samples quenched from below 1100°Careplotu_ (a)CAl. (b)B2C. (c) 98. (d)TCAN.
or upon rearranging, In Equations (3) and (4), at refers to the activity, of component i in
phase j, ao '.(:_ to the standard state free energy of reaction for
In K _) = 31n(a tiqc,o)+ in(a_a:o"3) + 4in ( s_._)
equilibrium (2), R to the gas constant and T to the temperature in
o o
= AG _'G) +In(_UNK degrees IC The acti_fity-composition relationships for UNK solid
_T "",- Si-UNK, • Equ. (4)
solutions are not known so activi_ and standard state terms on the
TABLE2. Selected analysesofUNKandSi-UNK fromsynthetic samples. fight hand of Equ. (4) cannot be separated. K _ is that part of the
Sample CAI- CAI- 98-21 98-55 98-69 B2C- B2C- TCAN- equilibrium constam arising from activities of components in the
226 252 68 29 32 melt. Given that the range in composition of the synthetic UHK is
Phase UNK Si-UNKUNK LINK UNK UNK UNK lINK small (Fig. 4) and aL"&_ may therefore be approximately constant,
Tmax(°C) 1420 1420 1500 1500 1500 1420 1400 1390 K_, may in fact be approximately constant for agiven temperature.
CR(_'CJh) 2 5 200 2 20 20 2 5
Using Be-man's (1983) activity model for silicate melts (assuming
T_uen_(°(2) 1101 1001 1050 1071 1067 1025 1045 1042
Oxide w{% that all Margules parameters involving TiO2 are zero), In K6 was
SiO2 24.8 45.8 28.5 28.6 28.2 26.4 25.0 26.9 evaluated for synthetic glasses and plotted against inverse temper-
TiO2 25.6 0.83 20.5 19.6 19.6 23.2 24.8 23.7 ature for glass compositions from UNK-bearing and UNK-absent
A1203 17.2 19.4 18.6 19.2 19.1 17.6 18.1 17.4 run products in Fig. 94. A linear regression line,
MgO 0.68 0.62 0.62 0.12 0.31 0.39 0.48 0.69
CalO 30.1 32.7 30.9 31.1 31.8 31.2 31.4 30.7
, -- liq liq liq -- ? 1.84x10 4
Total 98.4 99.4 99.1 98.6 99.0 98._ 99.8 99.4 In K(_- 31n (a cao) +ln (a _a2o_) +4 In (a sio2) -'12"5" T
Cations basedontotal - 9
for the UNK-bearing experiments is also shown. We emphasize
Si 2.29 3.92 2.57 2.60 2.54 2.41 2.35 2.44
thatthis expression incorporates both standardstate (free energy of
Ti 1.77 0.05 1.40 1.34 1.33 1.59 1.57 1.62
formation from the solid oxides) and activity terms for the
AI 1.87 1.96 1.98 2.05 2.03 1.89 2.02 1.86
C83/_2Si4014 component in UNK (eft Equ. 4). This treatment also
Mg 0.09 0.08 0.08 0.01 0.04 0.05 0.04 0.09
C_ 2.97 3.00 3.00 3.03 3.06 3,05 3.08 2.99 implicitlyassumes that the composition of glass measured at some
$i+ Ti 4.06 3.97 3.97 3.94 3.87 4.00 3.92 4.06 distance from crystals is the relevant measure of LTNK stability
* Cooling rale.
Reports 679
50]Jm
FIO. 7, The BSE images illustrating mod_ of
occurrence ofUNK insynthetic samples. GL:glass; SP:
spinel; TPX: Ti-Hchpyroxene; AN: anorthite, Other
abbceviatiorm as inFig. 1. (a) Euh*dralcrystalsoEUNK
oriented along faces inmelilite (9_-39, Tmax= 1500 oC,
CR = 20 °C,q_). (b) LINKas rimson melilite crystals.
Note alsotheprismatic UNK crystalsextendingfi'omthe
ends of melilite laths. (98-21; Tmax= 1500 :C, CR =
200 "Ca'h). (c) UNK as part of an imergrowth w/th
melilite, fassaite, and anorthite (CAI-2g6; Trr_.x= 1420
°C,OR= 50 oct). (d)UNK interstitial to other phases
(CAI-226; Tmax = t420 °C, CR "=2 °C'h). The
variable bdghmess of the pyroxene refl_-..s sector
zoning. (el The Si end m-rebec of UNK (C._I-252;
Tra_x = 1420 =C,CR = 5°CJh). Someof thebright
regionaslong cracksaredueto charging,but most of*he
brightvertically trending arcuate arrays aracomposed of
apoorly characterized, relatively high Zphase..
(i.e., botmdary layers are ignored). Nevertheless, liquid composi- saturation. For example, UNK saturation may not be possible in a
tions from _c UNK-bearing experiments are consistent with a specific instance because the Ti-coment of the melt may be
single expression for/a K (/,, wh/ch can be taken as approximating insu_cient to stabilize Ti-bearing L_'K with the appro-priate
the conditions required for stabilizing LINK in CA.I-like melts. The a,,.,._=_. The present analysis, therefore, should be viewed as
linearity suggests that either values of a,,._.., for rINK in our providing necessary but insufficient conditions for the stability of
ex'periments are all similar or that they change systematically with UNK. Keeping these caveats in mind, CAI-like melts with In Ka)
temperature. It is significant that glasses for all of the isothermal plotting beneath a particular isoacti_StT line will to first order be
and most of the cooling rate experiments plot beneath the line in undersaturated with respect to UNK of a given activity, and hence
Fig. 9a (at lower values of I/T) and, therefore, are predicted to be of some composition, provided that a,%% is physically achievable.
undersaturated with respect to LINK, This can explain simply _'ny Melts with ha K_ plotting above an isoactivity line will be super-
UNK is not observed in these experiments - it was never stable. saturated with respect to that UNK. Any melt with in K_h plotting
Figure 9b shows schematic curves of constant activity, of the Sl- above the a,.u_,_ = I line will be supersaturated with respect to Si-
UNK component for coexisting lINK and melt together with four rich UNK regardless of the UNK composition.
schematic cooling paths in In K@-104fr for a CAI bulk compo-
For illustrative purposes, we consider a "zone", represented by
sition. These curves can be used to illustrate factors influencing the UNK-saturated points in Fig. 9a and by the roughly parallel
the appearance or non-appearance of UNK in a particular cooling lines of constant a=_o,_._in'_Fig. 9b, where saturation with respect to
sequence. The basis for our treatment is that as a particular bulk UNK is possible. Upon cooling of a droplet of CAI bulk
composition crystallizes, the residual melt moves to the right in the composition from above the liquidns, t'n-st spinel and then melilite
figure. If it reaches or crosses over the diagonal line in Fig. 9a + spinel crystallize with decreasing temperature. The melt changes
defined by the UNK-saturated ex'periments, it becomes saturated or composition, but In K,h increases only slightly based on liquid
supersaturated with respect to UNK. Therefore, provided that lines of descent determined experimentally. In path A, UNK satu-
nucleation can occur, we can expect to observe rINK in the run ration is reached before any other additional phases crystallize.
product. Tkis treatment is necessarily schematic because precisely UNK crystallizes, and In K,h changes along an UNK-saturated
what a .%_.,=should be at satv..ation for a specific liquid is tmknown path. /zx detail, a,,.%.= changes as the composition of UNK and
(i.e., we are unable to separate the contributions of a,,_., _d
other phases change, but the precise path is n°t quantified by our
ao "...) to ha K,i)) and because we are neglecting the activities of
experiments. In path B, p.vroxene crystnllizes before UNK, and In
other components in the melt and UNK, In particular, the Ti-
K&, decreases relative to a p_oxene-absent liquid line of descenL
content of the UNK and melt are important in determining This has the effect of delaying the appearance of UNK of a given
680 Reports
TASL._3. Comparison of glass analyses from CAIbulk composition samples compositions pass through an appropriate region of In K_-I/T for
with and without LINK.allotherconditionsequal. UNK crystallization, there is insufficient time for nucleation of
Run _ 263* 264* 265* 288 287" 290 289* UNK (or hadeed of pyroxene or anorthite in many eases). At higher
Tmax(=C) 1420 1420 1420 1420 1420 1420 1420 Trrax, crystallization of all of the silicates including UNK are
CR (*C,la)-+. 2 2 2 50 50 50 50 .
greatly suppressed relative to their equilibrium appearance
Taueneh_C) 1048 1048 1048 1117 1117 1059 1059
temtxa'atures (Stolper and Paque, 1986)due to desmmtion of
UNK?'" yes no no yes no yes no
Oxide wt% nuclei. Therefore, the probabHit-y that rINK will crystallize is
SiO2 42.7 44.8 44.9 42.3 43.3 42.3 45.9 ®
TiO'2 1.37 0.5g 0.55 3.53 1.53 3.23 1.75
A1203 18.2 20.3 21.9 20.9 19.2 21.9 14.0
-2O
MgO 10.g 1.05 0.71 5.8 2.93 4.79 1.30
CaO 27.0 31.3 29.7 25.7 30.7 24.5 34.3
,j
Total 100.1 9g.0 97.8 98.2 97.7 96.7 97.3 -22
.-_-
* Average oftwo analyses.
"_Cooling rate. _ r,i -24
** All samples containing LINKhave the phaseassemblage glass - spinel +
melilite _-Tifassaite + anorthite + UNK. Sampleswithout LINKhave the z-, -26
phase assemblage glass + spinel + melilite + Ti-fassaite. CAt-289 also
contaJr_ anorthit¢.
-28
a_ to lower temperatures. If pwoxene begins to crystallize -30
5.5 6 6.5 7 7.5 8 8.5
early enough, UNK saturation may be delayed to such low
Idn'(K)
temperatures that LINK never crystallizes (path C). It is possible
that anorthite has an effect similar to that of pyroxene, althou_ our ®
data are insufficient to COnfL___this. In path D, no additional
phases join melilite and spinel and eventually the melt becomes
i I I
supersaturated with respect to UNK (as well as for other phases
!stability of UNK ]
such as anorthite and pyroxene).
¢,J
The four paths illustrated schematically in Fig. 9b can also help
to rationalize the association of UNK with particular ranges in UNK supersaturated
_J
experimental conditions (Fig. 6) because they are crudely
correlated with Tin= and cooling rate. For CAI experiments cooled
at intermediate to rapid cooling rates from Trr_ - 1420 *C, near
melilite saturation for the CAI bulk composition (Stolper, 1982),
S
the liquid line of descent intersects the temperature-hquid
composition regime within which UNK can crystallize because the
appearances of anorthite and fassaite are delayed so that UNK is
stabilized at relatively high temperatures (e.g., path A in Fig. 9b) 104/T(K)
and, therefore, can crystallize provided appropriate nucleation sites
are available. UNK crystallization for Trmx - 1420 *C is unlikely _o.9.Stability ofUNK. (a)Experimental results. Oxide activities inthemeR
at the highest cooling rates because, although the liquid were obtained using Berman's (1993) model and are relative to solid lime,
corundum and cristobalit=. Closed squares representUNK-bearing controlled
40
coolingrateexperimentsiawhichthe phasesspinel,meIilitep,yroxene,
mmrthlte, UNK and glassarcallpraseat Isothermal ¢xp=_ems (open circles)
i 'i______, , , , j , l '
andcoolinrgateexperimeats(opentrianglefso)rwhichUNK wasnotobserved,
35 ix o UNKp_,,_ntl represenvtarioupshaseassemblagesC.o)Fourschematilciquidlineosfdescent
foran initialmloyltenliquidofbulk CAI compositioninremusof ha
Km-IO4/T. Curve A: spinel _ spinel +melilite _ spinel + melilhe + LINK
30
(pyroxene failsto nucleate before UNK; in_-mediate to high T_ low to
intermediate cooling rates). Curve B: spinel --"spinel + melilita ---spinel +
melilite + pyroxm,_ -- spinel + mdilke + pyroxene + UNK (.pyroxene crystal-
25
lizes before LINKanddelays but do_ not prevent later LINK erystailizatlon;
intermediate Tmaxandcooling rate). CurveC: spinel -* spinel + melilite ""
spinel+ melilite +pyroxe_ (pyroxene appears at ahigh enough temperature Io
20 , I A LI , I , I z I r
completely su_ LINKeryslailization; low Trmx and cooling rate). Curve
0 2 4 6 8 10 12
D: spir_l --_spinel + melilite (pymxene andUNK fail to nucleate; IfighTmax
wt% MgO
and cooling rate). Thermal histories inwhich oneorboth ofmelilit- and spinel
FIo. g. Wt% CaO vs. wt% MgO inglass from CAI exper_ents. Lines join
samples with and without UNKthat were tan under identical conditions aIthe failed to nucleate (I.e.,high Trnax with very fast cooling rates) would lead to
same time. pathssimilar toDthough withsomewhat differantslopes.
I.[
Reports 6g1
reduced at all cooling rates (path D inFig. 9b). At lower Tr=x and Regardless of its origin the invariably low Mg and relatively
cooling rates, pyroxene begins to crystallize atahi_er temperature high AI contents ofUNK make it potentially useful for 26AIstudies
(paths B mad C in Fig. 9b) so that the appearance of LINK is analogous to those on corundum and anorthite ('Podosek et aL,
suppressed as discussed above. UNK may still crystallize (curve 1991; Vizag et al., 1991). The grains may yield useful information
B) if cooling rates are not so slow that the early appearance of onthe timing of whatever event produced them. Finally, it is worth
pyroxene leads toentirely suppressing LINK stability under igneous noting that if a significant proportion of the Ti in meteoritic UNK
conditions (curve C). The probability of encountering rINK there- coe,xi_ang with perovskite istrivalent (e.g., Floss et aL, I992), then
fore decreases for both higher and lower Tr,_ and for very high qua.ndtative calibration ofoxygea baromete_ such as
cooling rates. The data base is less complete for the B2C and 9g
bulk compositions but the results areconsistent with the same basic 3Ca3Ti 4+AI2 Si3014 - 3CaTiO3 =
behavior. 3Ca2AI2SiO7 +2Ca3Ti4+ Ti_+Si3O14 +02
could provide useful insight into redox conditions.
Origin of LINK
CONCLUSIONS
There are several possible modes of origin for UNK in
meteorites, including crystallization from a late-stage residual UNK is found as small, tabular inclusions in melilite in
liquid, exsolution from the enclosing melilite, or alteration. It approximately half of the T}pe A and BI CAIs examined. It
should be emphasized that different occurrences of meteoritic LrNK contains small amounts of Ti3÷ (7-13% of total Ti) sugges'.Jng
growth under reducing conditions. 15NK of similar composition is
may have different modes of origin. Here, we address the possibil-
produced in dynamic crystallization experiments over a well-
itythat meteoritic LINKcrystallized from amelt. A general discus-
defined range of maximum temperatures and cooling rates. The
sion of other possibilities is given in Barber eral.(1994)
crystallization of UNK from Ca,I-like bulk compositions requires
Based onFig. 9a, UNK will not be stabilized asanear-liquidus
residual melt compositions unlike those found under equilibrium
phase for any ofthe Types A, B, or C inclusions described by Wark
conditions, thereby explaining its restriction to dynamic
(1981), Wark and Lovering (1982), Beckett (1986), or Wark (1987) crystallization expe_ents, which generate a wider range of liquid
because In K,_) is too low. The dynamic crystallization experi- compositions, some ofwhich reach saturation with respect to LINK_
ments demonstrate that LINK with composition very. similar to
Acknowledgmems-Review5 of A. E1 Gore_'y and G. J. MacPhersort led to
meteoritic examples can crystallize from a range of bulk composi-
substantive improvements in the manuscript This work was supported by
tions representing these CAIs but only from very late stage liquids NASA grants NAG 9-105 and NAGW- 3533 to EMS, NAG 9-28 to John
after extensive crystallization of melilite and spinel * anorthJte4- Wood, and SETI grant NCC 2-758 to Lauranee Doyle. Caltcch Division of
Geological and Planetary Sciences, Division Contribution No. 5215.
fassaite, which drive the residual liquid near to UNK saturation. If
meteoritic lINK crystallized from a melt, it, therefore, must have Editorial handling: K. Keil
formed late in the crystallization sequence, but this poses potential REFERENCES
problems for meteoritic occurrences of UNK described in this
AORELL $. O. (1945) Mineralogical observations on some basic open-hearth
work. UNK is included in texturally early melihte at least in Type slags. ,.7.Iron Steel In:t. Land. 152, 19P-55P.
BI inclusions (i.e. it is preferentially in the outer portions of the ALBEE A. L AND RAY L (1970) Correction factor_ for electron probe
microanalysis of silicates, oxides, carbonates, phosphate, and sulfate*.
melilite mantle), although it is possible that meteoritic UNK in me _ Anal Chera.47,1408-1414.
melilite monde is actually a late-crystallizing phase analogous to BAP.BEP, D. J. A/,q2 AO1ZXLL$. O. (1994) A new titanium.bearing calcium
aluminosilicate phase: III.Crystals from amixer furnace -_lag. Mereoritfcs
fassaite inclusions in mantle melihtes (MacPherson et aL, 1984;
29, 691--695.
Simon et aL, 1991) and the increase in XAknear_ inclusions BARBEP, D. J., BECKETT .,I.R., PAQU_ J. M. AND STOLPE_ F..(1994) A new
(Fig. 5b) is consistent with this. Ifmeteoritic UNK formed by such titanium-bearing calcium aluminosilicate phase: II. Crystallography and
crystal chemistry of grains ton'ned in slowly cooled melts with bulk
a mechanism, then extreme rare earth element (R.EE) enrichments
compositions of calcium-, aluminurn.rieh inclusions. Meteorities 29,
would be expected whereas its REE content would be exPected to 682-690.
be low if it had exsolved from the surrounding metilite. The UNK - BECX:m'rI ]t. (1986) The origin of-_aleium-,aluminum-richinelusiortsfrom
carbonaceous chondrites: Anexperimental taudy. Ph.D. thesis, University
described by E1 Goresy et aL (1984) is REE enriched and, there-
of Chicago. 373 pp.
fore, consistent with crystallization from a melt, but the host BECKEI"r J. P,- AND STOLPER ]_.(1994) The stability of hibonite, melilite and
other aluminous phases in silicate melts: Implications for the origin of
inclusion is so unusual that it is not possible based on this occur-
hibonite-bearing inclusions from carbonaceous chondrites. Meteorff_cs 29,
rence to make general inferences about occurrences of UNK in the 41--65.
Allende CAIs described here. There are currently noREE analyses BE_ J. R., SP[VACK A. J., HUTCH_ON L D., WAS_P, BtmO G. J.
STOL?ER E.M. (1990) Crystal chemical effects onthe partitioning oftraeg
of UNK from these CALs. Another problem with crystallization of
elt_aents between mineral and melt: An experimental study of melilite
meteoritic UNK from a melt is that glass is often associated with with applicstiom to refractory inclusions from carbonaceous chondrites.
UNK included in melilite in experimental samples, but this feature GeocMm. Cosrnochlm. Acta ._, 1755-1774.
BERMArt R. G. (1983) Alhermod_amie model for multieomponent melts with
is not observed in CAIs. Floss et al. (1992) described a phase
application to tl_ system CaO-MgO-AI203-SiO2. Ph.D. thesis, University
similar to UNK occurring as symplectic intergrowths inaCompact ofBritish Col_-nbia. 17_ pp.
Type A inclusion apparently replacing perovakite. The phase is DE£R W. A., HOWIER..au AND ZL'S_rtAN J. (1992) An Introduction to the
Rock-forming Minerals. Longman Scientific and Technical Press, Essex,
enriched in REE and although subsolidus reactions may be U.K. 696 pp.
responsible, crystallization from near-solidus melts is a more likely EL GOP,.ESY A., P_._ H., YABUKI H., NAGEL K., HE_RW_mTR I.._a, rD
RAMI_FIR P. (1984) A calcium-aMmmum-6ch inclusion from the Essebi
origin for this type of texture.
(CM2) chondrite: Evidence for captured spinel.hibonite _herules and for
682 Reports
an ultra-refractory rimming sequence. Geoehtm. Cosraochim. Acta 48, suite of refractory inclusions from the Allande meteorite. Geochim.
2283-2298. Cosmochim. Acta 55, 1083-1110.
FLOSS C., EL GORESY A.. PALME H., SPETI-c.L B. ANDZINNER E. (1992) An SIMON S. B., GROSSMAN L AND DAVIS A. Iv[. (1991) Fassaite composition
unusual Ca-Ti-AI silicate in aTHee A Allende inclusion (abstract). trends daring crystallization of Allende Type B refi'actory inclusion mdt.%
MeteorTtics 27, 220. Geochim. Cosmoc htm. Acre 55, 2635-2655.
GIBBF. G. I:. (1974) Supercooling and the mTsxallization of plagiocla._t from a STOLPER E. (1982) Crys'tallization se.queaees of Ca-Al-rieh inclusions from
basaltic magma. Mineral Meg. 39, 641-653. Allen,: An expe6menml study. Geoeh_m. Cosmochim. Acta 46,
MACPHERsoN G. J., PAQUEJ. M., STOLP_-RE. ANDGROSSMAN L (1984) The 2159-2180.
origin and signkfieanee of revere zoning in melilite from Allende Type B STOLPER E. AND PAOUE J.M. (1986) Cryaallization SeXlUences of Ca-Al-fich
inclusions. J. Geol. 92, 289-305. inclusions from A]lmde: The effects of cooling r_ and maximum
MACON G. J., WARK D. A. AND ARMSTRONG J. T. (1988) Primitive temperature. Geochim. Cosmochtm. Acta 50, 1785-1806.
_al surviving in chonddtes: Refractory inclusions. In Meteorites and Vm.AO A., _ E., AMAI_ $. AND A,_IDE1LSE. (1991) An ioa microprobe
the Early Solar System (¢ds. L F. Kerfidge aad M. S. Matthews), pp. study, of corundum inthe Murehi_n meteorite: Implications for 26.4.1and
746---807. Univ. Arizona Press, Tucson. Arizona. 160 in the early solar system. Geoehim. Cosmochim. Acta 55,
PAQUEJ.M. AND STOLPERE. (1984) Cr>mallization experiments onarange of 2045-2062.
Ca-AI-rieh inclusion compositions (abstract). Lunar Planet. Sci. 15, WAP,.K D. A. (1981) The pre-alteration compositions of Allende Ca-Abrich
631--632. condensates (abstract). Lunar Planet. SeL 12, 1148-I 150.
PAQUE J. M., BECXETI" J. R-AND STOLP/---RE. (t986) A new Ca-AI-Ti silicate WAgX D. A. (1997) Plagmclase-rich inclusions in carbonaceous _on_e
in eoar_-grained Ca-Al-rieh inclusions (absaraet). Lunar Planet. Set. 17, meteorites: Liquid condensates? Geochim. Cosmochsm. Acta 51,
646-647. 221-242.
PODO_ F. A., ZINNER E. K., MACRON G, J., LUNDBEg.0 L L, WARK D. A. ANDLOVERINO J. F. (1982) The nature and origin of type B1 and
BRANNON .l.C. AND FAHEY ,4. J. (1991) Correlated study of initial B2 Ca-A.l-rich inclusions in the ADende meteorite Geach_m. Cosmoehtm.
ZTSr/g6Sr and AI-Mg isotopic sys'temati¢_ and petrologic properties in a Acta 46. 2581-2594.
A new titanium-bearing calcium aluminosilicate phase: II.
Crystallography and crystal chemistry of grains formed in slowly cooled melts
with b_lk compositions of calcium- aluminum-rich inclusions
DAVID J. BARBER]*, JoIffN R. BECKE'I'r2, JULm M. PAOUE3 AND'EDWARD STOLPEP,2
lPhvsies Deparanent, University. of E.ssex, Colchester, Essex CO4 3SQ, U. K.
2Division of Geological and Planetary Sciences, California Institute of Tedmolo_, Pasadena, California 91125 USA
3SETI Institute, NASA-Ames Research Center, MS 244-11, Moffett Field, California 94305 USA
*Current address: Physics Department, The Hone Kong University of Science and Technology, Clear Water Bay, Kowloon, Hone Kong
(Received 1993 September 28; accepted in revised form 1994 May. 3)
Abstract-The crystallography and er_ daernistry of a new ealeiurn-titanium-aluminosiiieate mineral CIJNK) observed in synthetic
analogs to calcium-aluminum-rich inclusions (CAIs) from carbo_eeeus chondrites was studied by electron diffraction techniques. The unit
cell isprimitive hexagona/or trigonal, with a= 0.790 ±0.002 nm and e =0.492 ±0.002 ran, similar to the lattice parameters of melilite and
eoraistent with cell dimensions for crystals inamixer furnace slag described by Barberand Agrell (1994). The phase frequently displays an
epitactie relationship in which melilite acts as the host, with (0001)UNK _(001)n_l and <10"r0>L_ K]]<100>mc l. If one oft he two spa_
groups determined by Barber and Agrell (1994) for their _mple of UNK is applicable (P3ml or P31m), then the straeture is probably
characterized bypuckered sheets ofoctahedra and tetrahcdra perpendicular to the e-axis with successive sheets coordinated by planar arrays
of Ca. In this likely structure, each unit celt eontaim three Ca sites located in mirror plan_, one oetahedrally eoordinatexi cation located
along athree-fold axis and five tetrahedrally coordinated eationg three inmirrors and two along triad_ The octahedron contains Ti but,
because there are 1.3-1.9 cations of Ti/formula unit, some of the Ti mug also be in tetrahedral coordiaation, an unusual but not
unprecedented situation for asilicaS. Tatahedral sit_ in mirror planes would contain mostly $i, with lesser amounts oral while those along
the triads correspondingly contain mostly AIwith subordinate Ti. The structural fommln, therefore, canbe expressed as
Ca_vm(TLAI)w (Al,Ti.$i) ['v(SiAl)70.
with Si + Ti = 4. Compositions of _c and synthaic Ti-bearing samples of lhe phase can be desen_l in tmns of a binary solid
solution between the end-members Ca3TiAl2Si3Ol,, and Ca3Ti(AITiXMSi2)O]4. A Ti-fi'¢¢ aradog with a formula of Ca3AI2Si40]4
synthesized byPaque et aL (1994) isthought to be related su-ucturally but with the octahedral s_ being occupied by A1,that is,
c,_ A1_ (A_si)7(si) _' o..
INTRODUCTION silicate described by E1 Goresy et el. (1984) and Paque et al.
Coarse-grained calcium-aluminum-rich inclusions (CAIs) in (t 986). This phase could also contain genetic information about
carbonaceous chondrites have chemical and isotopic si_atures CA.Is, but the crystal structure, phase relations, thermodynamic and
indicative of processes dating back to the origin and earhest kinetic properties that might allow such information to be extracted
evolution of the Solar System (see Grossman, 1980 and have not been characterized. In this study, we consider the crystal
MacPherson et aI., 1988 for reviexx_). The mineralogy is generally structm-e of the phase. The mineral remains unnamed because
dominated by a combination of one or more of ctinopyroxene, there are no x-ray diffraction (XRD) data currently available on the
mel.ilite, spinel, anorthite, hibonite and perovskite. For each phase, natural material, but for the purposes of this paper, we refer to the
the compositions and textures reflect the history of the host phase as UNK.
inclusion and can be used to impose constraints On physical UNK is a ubiquitous, albeit minor, constituent of c,oarse-
conditions early in the Solar System. There are, in addition, grained meLilite-rich (Type A: Grosmaan, 1975) and fassaite-rich
various minor and trace minerals in CAIs including a new Ca-Ti-AI (Type B) CAIs. In occurrences of UNK reported thus far (El
