Origin of Th/U Variations in Chondritic Meteorites Citation Goreva, Julia S. (2001) Origin of Th/U Variations in Chondritic Meteorites. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/hfsz-ym11. https://resolver.caltech.edu/CaltechTHESIS:04252014-150216789 Abstract Isotope dilution thorium and uranium analyses of the Harleton chondrite show a larger scatter than previously observed in equilibrated ordinary chondrites (EOC). The linear correlation of Th/U with 1/U in Harleton (and all EOC data) is produced by variation in the chlorapatite to merrillite mixing ratio. Apatite variations control the U concentrations. Phosphorus variations are compensated by inverse variations in U to preserve the Th/U vs. 1/U correlation. Because the Th/U variations reflect phosphate ampling, a weighted Th/U average should converge to an improved solar system Th/U. We obtain Th/U=3.53 (1 -mean =0.10), significantly lower and more precise than previous estimates. To test whether apatite also produces Th/U variation in CI and CM chondrites, we performed P analyses on the solutions from leaching experiments of Orgueil and Murchison meteorites. A linear Th/U vs. 1/U correlation in CI can be explained by redistribution of hexavalent U by aqueous fluids into carbonates and sulfates. Unlike CI and EOC, whole rock Th/U variations in CMs are mostly due to Th variations. A Th/U vs. 1/U linear correlation suggested by previous data for CMs is not real. We distinguish 4 components responsible for the whole rock Th/U variations: (1) P and actinide-depleted matrix containing small amounts of U-rich carbonate/sulfate phases (similar to CIs); (2) CAIs and (3) chondrules are major reservoirs for actinides, (4) an easily leachable phase of high Th/U. likely carbonate produced by CAI alteration. Phosphates play a minor role as actinide and P carrier phases in CM chondrites. Using our Th/U and minimum galactic ages from halo globular clusters, we calculate relative supernovae production rates for 232 Th/ 238 U and 235 U/ 238 U for different models of r-process nucleosynthesis. For uniform galactic production, the beginning of the r-process nucleosynthesis must be less than 13 Gyr. Exponentially decreasing production is also consistent with a 13 Gyr age, but very slow decay times are required (less than 35 Gyr), approaching the uniform production. The 15 Gyr Galaxy requires either a fast initial production growth (infall time constant less than 0.5 Gyr) followed by very low decrease (decay time constant greater than 100 Gyr), or the fastest possible decrease (≈8 Gyr) preceded by slow in fall (≈7.5 Gyr). Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: Geochemistry Degree Grantor: California Institute of Technology Division: Geological and Planetary Sciences Major Option: Geochemistry Thesis Availability: Public (worldwide access) Research Advisor(s): Burnett, Donald S. (advisor) Rossman, George Robert (advisor) Thesis Committee: Rossman, George Robert (chair) Burnett, Donald S. Farley, Kenneth A. Wasserburg, Gerald J. Stolper, Edward M. Defense Date: 20 December 2000 Record Number: CaltechTHESIS:04252014-150216789 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:04252014-150216789 DOI: 10.7907/hfsz-ym11 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 8202 Collection: CaltechTHESIS Deposited By: Benjamin Perez Deposited On: 28 Apr 2014 23:01 Last Modified: 13 May 2024 21:39 Thesis Files Preview PDF - Final Version See Usage Policy. 30MB Repository Staff Only: item control page

Origin of Th/U variations in chondri tic meteorites Thesis by Julia S. Goreva In Partial Fulfi llment of the Requirements For the Degree of Doctor of Philosophy Californi a Institute of Technology Pasadena, California 200 1 (Submitted December 20, 2000) II Acknowledgments I will not be able to thank everybody who filled the void around me during my tenure at Caltech. But I would like to point out several people who contributed the most to what I have become. Fir t of all, I wou ld like to thank my advi e r professor Don S. Burnett. I have enjoyed workin g with him during last few years immensely; I would not have become what I am now without his guidance and support. In particular. he formu lated the problem this thesis is dealing with, and strategically maneuvered me through all my ups and downs with patience. technical expertise and always provided moral support. I thank him for exposing me to his unlimited enthusiasm and multi-faceted interests. I would like to thank my other adviser. George R. Rossman. who introduced me to the mineral spectroscopy. and under whose supervision I have done work included at the end of this thesis as Attachments . I have benefited from his talent of being a navigator rather than a driver. I would like to thank George for his curiosity. fo r the spark in his eyes and his open door. I am thankfu l to many people who have helped me out around various Caltech labs: Mihai Ducea and Jim Chen (radiogenic isotope clean lab), Ken Farley and Yaniv Dibowski (ICPMS). Paul Carpenter (electron microscopy), Liz Aredondo (mineral pectroscopy). and others. Thanks to NASA grant NAGS-4319 for the financial support of the meteorite work, and to the White Rose Foundation for the support of the rose quartz work. I have made a number of wonderful friends at Caltech, who brightened my days (years) here. But two of them became especially close to me and deserve special Ill appreciation. I thank Ronit Kessel for hour of serious talks, meaningless chats and silly laughs. And for hundreds of gallons of coffee we had together. And I express my endle s gratitude to Mihai Ducea, who has been for me a source of inspiration, support and encouragement. I thank Mihai for his friendship and love, and for becoming part of my life. I thank my mother Natalia Goreva and my brother Georgiy Gorev for their optimism, constant support. and for letting me move to the other hemisphere in pursuit of my goals. Finally, I would like to dedicate this thesis to my grandfather, Korolev Valeriy G .. whose boundless devotion to science established three generations of geologists in our family. IV Abstract Isotope dilution thorium and uranium analyses of the Harleton chondrite show a larger scatter than previously observed in equilibrated ordinary chondrites (EOC). The linear correlation of Th/U with 1/U in Harleton (and all EOC data) is produced by variation in the chlorapatite to merrillite mixing ratio. Apatite variations control the U concentrations. Phosphorus variations are compensated by inverse variations in U to preserve the Th/U vs. 1/U correlation. Because the Th/U variations reflect phosphate ampling, a weighted Th/U average hould converge to an improved solar system Th/U. We obtain Th!U=3.53 0-mean=O.l 0), significantly lower and more precise than previous estimates. To test whether apatite a lso produces Th/U variation in CI and CM chondrites. we performed P analyses on the solutions from leaching experiments of Orgueil and Murchison meteorites. A linear Th/U v . 1/U correlation in CI can be explained by redistribution of hexavalent U by aqueous fluids into carbonates and sulfates. Unlike CI and EOC. whole rock Th/U variations in CMs are mostly due to Th variations. A Th/U vs. 1/U linear correlation suggested by previous data for CMs is not real. We distinguish 4 components responsible for the whole rock Th/U variations: (I) P and actinide-depleted matrix containing small amounts of U-rich carbonate/sulfate phases ( imilar to Cis): (2) CAis and (3) chondrules are major reservoirs for actinides. (4) an easily leachable phase of high Th/U . likely carbonate produced by CAl alteration. Phosphates play a minor role as actinide and P carrier phases in CM chondrites. v Using our Th/U and mmtmum galactic ages from halo globular clusters, we calculate relative supernovae production rates for m Thf3 8U and 235U/238U for different model s of r-process nucleosy nthesis. For uniform galactic production, the beginning of the r-process nucleosynthesis must be < 13 Gyr. Exponentially decreasing production is also consistent with a 13 Gyr age. but very slow decay times are required (<35 Gyr), approaching the uniform production. The 15 Gyr Galaxy requires either a fast initial production growth (infall time constant <0.5 Gyr) followed by very low decrease (decay time constant> I 00 Gyr), or the fastest possible decrease (==8 Gyr) preceded by slow in fall (==7.5 Gyr). VI Table of Contents Acknowledgments._ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ II Abstract _ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Iv Table of Contents_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ VI List of Tables _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _v11 List of Figures VIII Chapter 1: Introduction - - - - - - - - - - - - - - - - - - - - - - 1.1 Background and motivation. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ 1.2 Thesis outline _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ !! Chapter 2: Analytical techniques 13 2.1 Sample preparation 13 2.2 Inductively coupled mass-spectrometry 16 2.3 Instrumentatio n used in this work 23 Chapter 3: Ordinary chondrites 37 3. 1 Sample selection and preparation 37 3.2 U. Th, P ana lyti cal data 43 3.3 The phosphate control o n the actinide abundances 51 3.4 Implications for ordinary chondrite petrogenesis 67 3.5 An improved solar system Th/U 71 Chapter 4: Carbonaceous chondrites 76 4.1 Th/U variatio ns in carbonaceous chondrites 76 4 .2 CI chondrites 78 4.3 CM2 chondrites 82 4.4 Interpretation 99 Chapter 5: Implications to cosmochronology I 06 Summary 120 References 125 Appendix I 13 1 Appendix II 156 VII List of Tables Table 1.1 U and Th mixed spike concentrations_ _ _ _ _ _ _ _ _ _ _ _ _ 14 Table 3.1.1 Sources of meteorite samples_ _ __ _ __ _ __ _ _ _ _ _ _39 Table 3.1.2 Chemical composition of Harleton chondrite compared to L chondrites_41 Table 3.1.3 Normative composition (vol. %) of Harleton and L chondrites_ _ _ _42 Table 3.2.1 U, Th, P analytical data _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _46 Table 3.2.2 St. Severin phosphates (data from Crozaz, 1979) _ _ _ _ _ _ _ _ _47 Table 3.3.1 Th and U ion probe data of phosphates _ _ _ _ _ _ _ _ _ _ _ _54 Table 3.5.1 Calculated average Th/U for Ordinary Chondrites _ _ _ _ _ _ _ _ _75 Table 4.2.1 U, Th and P analyses of Orgueil (CI) _ _ _ _ _ _ _ _ _ _ _ _ _ 81 Table 4.3.1 Murchison meteorite. Whole rock and leaching experiments_ _ _ _ 95 Table 5.1 Production ratios from r-proces calculation - - -- - - - -- 113 Table 5.2 Minimum decay time constant for 13 Gyr galactic age as a function of t 2 and Th!U______________________________________________ II9 Table 5.3 Dependence of the minimal infal l time constant on Th!U and the galactic age _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ __ __ _ __ _ _ _ _ __ 119 VIII List of Figures Figure 1.1 Th/U vs. U diagram for CI, CM2 and CV3 chondrites _ _ _ _ _ _ _ _6 Figure 1.2 Refractory-element ratio . represented by element/AI ratios in chondrites _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 8 Figure 1.3 Whole rock ID Th and U literature data for ordinary chondrites, showing bi-modal distribution of Th!U ratio 9 Figure 2.2.1 Schematic of an inductivel y coupled plasma-mass spectrometer 19 Figure 2.2.2 Schematic of a quadrupole mass filter 21 Figure 2.3.1 Schematic diagram of Element ICPMS 28 Figure 2.3.2 CET AC microconcentric nebulizer (MCN-1 00) 29 Figure 2.3.3 Electrostatic scan of the 23 U peak. illustrating flatnes of the peak top 32 Figure 2.3.4 Medium resolution scan of the P peak 34 Figure 2.3.5 Standard Addition method for P without internal standardization 35 Figure 2.3.6 Standard Addition method with normalization of the P counting rates against Cs and 25 M 0 o - - - - - - - - - - - - - - - - -- - ---------------36 IX Figure 3.1.1 Backscattered secondary electron image of the Harleton meteorite thin section. showing big, IOOX200 11m apatite grain _ _ _ _ _ _ _ _ _ _ _ _40 Figure 3.2.1 Th and U isotope dilution data for Harleton _ _ _ _ _ _ _ _ _ _48 Figure 3.2.2 Harleton data and literature data on ordinary chondrites _ _ _ _ _ _49 Figure 3.2.3 Model I. Apatite-merrillite mixing _ _ _ _ _ _ _ _ _ _ _ _ _ 50 Figure 3.3.1 Th/U vs. P for Harleton._ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _56 Figure 3.3.2 Plot of measured U concentration vs. P content in Harleton _ _ _ _ _57 Figure 3.3.3 Plot of measured Th concentrations vs. P content in Harleton _ _ _ _ 58 Figure 3.3.4 Plot of measured whole rock U vs. fraction of apatite (Model II) _ _64 Figure 3.3.5 Same plot as Figure 3.3.4. for different [Th!UJ.patite-- - - - - - - 6 5 Figure 3.3.6 Model II. Plot of calculated concentrations of U in apatite vs. measured phosphorus concentrations 66 Figure 4.1.1 Th/U vs. 1/U plot of the whole rock ID Th/U analyses for C Chondrite _77 Figure 4.3.1 Th/U v . 1/U plot for CM2 chondrites _ _ _ _ _ _ _ _ _ _ _ _ 84 Figure 4.3.2 Schematic diagram of the sample preparation and leaching steps for the Murchison _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ __ __ _93 X Figure 4.3.3 Illustration of the 2 ml centrifuge vial used for separation of the leaching solution from the residue 94 Figure 4.3.4 Ni-Fe-silfide grain from Murchison _ _ _ _ _ _ _ _ _ _ _ _ _97 Figure 4.3.5 CAl inclusion from Murchison meteorite _ _ _ _ __ _ _ _ _ __ 98 Figure 4.4.1 Combined literature and this study data for CM2 chondrites meteorite_ I 03 Figure 4.4.2 Whole rock. CAL chondrules and matrix analyses in Allende ____ 104 Figure 4.4.3 Illustration of mixing of nebular material and CAl in CV3 chondrites_ I05 Figure 5.1 Examples of different models of nucleosynthetic production historie _114 Figure 5.2 Relative actinide production rates for the uniform galactic production_ll5 Figure 5.3 Relative production rates for the exponential galactic production I 16 Figure 5.4 Relative production rates for the exponential galactic production with varying infallrates 11 7 Figure 5.5 Calculated infall and decay time constants for different 235 U/238U supernovae production rates I 18 Chapter 1: Introduction 1.1 Background and motivation The actinide elements (Th. U; Pu for meteoritic materials) are most famous because of the chronological applications as ociated with their radioactivity. This has the consequence that their unique and interesting chemical propertie are often overlooked. In this work we re-emphasize actinide chemistry. We performed an analytical study to understand mechanisms of actinide element fractionation in meteoritic material . In turn this enabled us to obtain a solar system Th/U with significantly improved accuracy for cosmochronological applications. The relative abundance of Th and U is close to 3.7 (by atom) in a wide variety of planetary materials. resulting in this value a being regarded as a solar system constant. Th and U are highly incompatible elements that are strongly depleted in the terrestrial upper mantle (e.g., Collerson and Kamber, 1999). In addition to overall dep letion, both the lithospheric mantle and asthenospheric mantle can maintain domains with highly variable Th/U ratio, partial melting of which contributes to the variability of Th/U in the crust (Asmerom. 1999). Uranium has 2 common oxidation states (+4 and +6). whereas thorium has only one (+4). Therefore. oxygen fugacity as well as fluid content and magma composition affects the Th/U ratio (e.g .. Keppler and Wyllie. 1990). As a result of these. and potentially other mechanisms. reported ratios for crustal igneous rocks vary greatly from at least as low as I in oceanic tholeiites to as hi g h as 7 in a labradorite-andesine alkali basalt from Guadeloupe Island (e.g. Tatsumoto. 1966). Such a 2 big spread in Th/U ratios makes it difficult to infer Th/U ratio of the bulk sil icate Earth from crustal rocks. Beyond merely assuming that the relative terrestrial abundances of refractory lithophile elements are chondritic. the principal evidence for the applicability of Th!U=3.7 for the bulk Earth come from 208 PbP06Pb data beginning with Patterson (1955). From the 20 PbP06Pb ratio in a variety of ancient terrestrial rock , the calculated Th/U ratio of the bulk Earth is 3.9±0.1 (Manhes et al .. 1979). The Th/U deduced from the lead isotope compositions of conformable lead ore deposits is 3.85 (e.g., Stacey and Kramers, 1975). Based on the Pb-isotope study of komatiites and ophiolites, Allegre et al. ( 1986) infer the Th/U ratio of the bulk silicate Earth to be 4.2. The Th/U ratios in lunar rocks and regolith cluster strongly around 3.7 (Korotev, 1998). Because the Moon is an extremely dry and moderately reducing environment, the fO~ effects on Th/U are minimal. Assuming only tetravalent U. the whole planet Th/U ratio is seen in crustal rocks. The 9 samples of 5 different SNC meteorites show a range in Th/U from 3.75 to 4.4 (Chen and Wasserburg, 1985 and Chen et al. 1993). Given the apparently oxidizing conditions of Martian materials, it may be difficult to infer a Th/U for Mars from SNC meteorites. In April. 2001 the Mars Odyssey spacecraft will be launched with a gammaray spectrometer aboard. Because of the poor quality of the uranium data, the gamma-ray spectrometer will only obtain an average U and Th data for the Mars surface. The gamma-ray escaping into the space come from depth of at most several ten of g/cm 2• which i -I 0 to 20 em in a regolith (Boynton et al., 1992). Therefore. the Th/U average of the Martian surface will come mostly from the surface dust. Since the dust appears to 3 be oxidized (i.e., red). we expect the Th!U ratio from Odyssey data to be lower than the actual mar Th!U. The discussed evidence for the constancy of Th/U in a wide variety of planetary materials suggests a limited range of fractionation mechanisms, making it conceivable that an accurate average olar system ratio can be obtained. Whole rock Th/U variations in carbonaceous chondrites The CI carbonaceous chondrites are normally regarded as providing the best source of average solar system elemental abundances. Thus. Anders and Ebihara ( 1982) (later adopted by Anders and Grevesse,1989) based their estimation of Th and U solar abundances on CI chondrites. to a large extent on Orgueil, the most extensively studied CI meteorite. U and Th analyses used by Anders and Ebihara ( 1982) are mostly done by the neutron activation (NAA) technique. which has relatively poor accuracy for U and Th. Moreover, a significant number of data points were rejected. The mean chondritic Th/U ratio inferred from Ander and Ebihara·s solar abundances is 3.7±0.4 (Ia). Recent. high precision, analyses show surprisingly large, up to factor of 4. intersample variation in Th!U within Cl type chondrite (Rocholl and Jochum; 1993). These variations are unique to Th!U. and not seen in other refractory lithophile elements. Given the e variations, use of CI compo. itions for precise U and Th solar abundance becomes questionable . Rocholl and Jochum ( 1993) also noticed large variations In other groups of carbonaceous chondrites. These authors observed very interesting systematic decreases in Th/U with U concentrations for CI, CM2 and CV3 chondrites (Figure 1.1 ). Individual 4 carbonaceous chondrite groups display similar trends in Th/U vs. U. In the sequence CI. CM. CV, there is a shift of the correlation curve towards higher U concentrations. Compared to CI chondrites. which display a concentration range of a factor of 6, CV chondrites show a much lower range of U of about a factor of 2 corresponding to smaller Th/U variations. Figure 1. 1 hows high precision isotope dilution TIMS and SIMS analyses (filled symbols) as well as radiochemical neutron activation analy i (RNAA. open symbols). Th/U variations in ordinary chondrites Ordinary chondrites appear to have average solar abundances of refractory lithophile elements. Figure 1.2 (from Larimer and Wasson, 1988) is a plot of refractory lithophile element ratios versus the ratio of AI (refractory) to Si (more volatile). In this plot carbonaceous. ordinary and en tatite chondrites can be distinguished by AI/Si ratio. because the abundance of more volatile Si varies. But the refractory-element ratios plotted along they-axis are constant within each group of chondrites to within± 15% total spread. The analyses presented in Figure 1.2 are collected by the same analyst. G.W. Kallemeyn. using INAA. and much of the scatter of the data is expected in bulk analyse obtained by this method. Refractory element ratios are remarkably constant within, as well as between. variou chondrite group . The average agrees with CI. which is accepted a the average solar value. The largest scatter in refractory-element ratios, observed in enstatite chondrite . probably reflects inhomogeneous samples. rather than refractory lithophile element fractionation. For ordinary chondrites. Srn/AI:::2.74* 106 (±8o/c total spread), Sc/AI:::4.3 * 10~(±12o/c ). Ca/Al:::74(± 10 o/c). Th and U are both refractory lithophile 5 elements. so ba ed on the arguments above. Th!U ratios of the bulk ordinary and CI chondrites should be the same and Th!Uoc=Th!Uso1ar· 5 6 Cl <> 4 t 3 0 <> 0 ~ 0 • ~ !!:, 0 <> 2 • • 0 5 CM2 0 t•.. 4 ~ r: 3 • 0 2 0 7 CV 3 6 0 5 •\ -4 " • • .1: ~-' 3 2 • ;J A 0 0 5 10 15 20 25 U, ppb 30 35 40 Figure 1.1 Th/U vs. U diagram for CI, CM2 and CV3 types carbonaceous chondrites. IDisotope dilution TIMS. SIMS data from Rocholl and Jochum ( 1993), Tatsumoto et al. ( 1973.1976). Chen et al. ( 1993), Chen and Tilton ( 1976). All groups display negati ve correlation Th/U vs. U. In the sequence CI - CM2 - CV3 . there i a shift of a correlation c urve towards higher U concentratio ns. 7 Restricting the discussion to high accuracy measurements of Th/U in ordinary chondrites by isotopic dilution, prior to this work the literature data show a strong clu tering of analyses around 3.6. However. 4 independent analyses of two L6 ordinary chondrites, Harleton and Glatton, show ratios around 6 (Unruh. 1982; Chen et al., 1993, and private communication) suggesting a bimodal distribution (Figure 1.3). The bimodal Th/U distribution is surprising for primitive materials such as chondrites. especially in light of Figure 1.2. The objectives of the ordinary chondrite study are: (1) check the apparent bimodal distribution with additional high accuracy measurements of Th/U in ordinary chondrites. (2) under tand the fractionation mechanisms causing the variations, (3) interpret the variation in terms of ordinary chondrite petrogenesis. Unexplained intersample variations provide a quantitative measure of the uncertainty m estimates of solar abundances. Abundance variations for which the mechanisms are understood can be interpreted in terms of an original unfrac ti onated set of solar abundances. 8 2. ' I • ... A - • ' • - '' .- . • .. ••sr- .. . .. ..•.. or• < .. ~ , )( " .. l • -~ -~.,.., 4'~Y C• ' ' l I k • *~ • ~"' "· p .. (. 0 • ~ ...:.- • . ...... • c ....¢ r- loC <. .,.. ~- • .. ....•.. • II '..) ·t • ... , 1l ... -: . i " n ...• I . I 1 • < .. . ;. !'" :... • .. .. . .. ..,. _,.,.... • .p - J0 • • • j .~ a D- ..• - • t. ,. f,A -- _...-.. • ' I .. I .. :: ~ .; --· • l I ! •t-1 • Cl ~ :: L •<'"' A I!H• C O , • £l • CVi ... *- '<i ~ 0 . I -~ ~ .... ... ·- 2 ..- I - •• , - .. .. . ,. •• • • ·• ... •' • I ltf1llll70 Figure 1.2 Refractory-element ratios. represented by element/AI ratios in chondrites (from Larimer and Wasson. 1988), based on the data by Kallemeyn and W as on 1981, 1982. 1986. 9 14 12 10 # 8 6 4 2 0 2 3 5 4 6 7 Th/U Figure 1.3 Whole rock ID Th and U literature data for ordin ary chondrites, showing bimoda l di tribution of Th/U rati o. Arrow indicates Anders and Grevesse ( 1989) value. 10 Nucleochosmochronlogy models Why do we need to establish more precise relative Th and U abundances in the solar system? In the classic paper on nucleosynthesis by Burbidge, Burbidge, Fowler, and Hoyle ("B2FH.'' Burbidge et al. 1957), it was recognized that a model age for the galaxy could be derived from the solar system Th/U and 235 U/~38U. The rephrasing of the problem as "cosmochronology" recognize that there both the age of the galaxy and the time dependence of the rate of r-proces nucleosynthesis are involved. Assuming the time interval between the termination of galactic nucleosynthesis inputs into the solar system reg ion and the formation of meteorites is negligible. for every nuclear radioactive species with a decay constant A.,. the abundance , at time T (time before the formation of the Solar system) is defined as a solution of the growth equation: N = ( P (t )e-A..IT-n dt . I Jo I There are two unknowns in the equation above: net production .. history;· or change in the production rate with time (representing supernova input to the interstellar medium and loss of mass in star formation), and T. Considering two ratios of two species. such as ?' ' ? '8 ?' ? ? '8 -~'u;-~ U and -~-Th/-~ U. one wou ld have two equations with two unknowns. setting constraints on T and P,(t). although P,(t) can have severa l parameters. Schramm and Wasserburg ( 1970) (summarized in Podosek and Swindle. 1988) showed that long-lived isotopes define a "mean age of the elements," regardless of the time-dependent model used, i.e .. the average time of production from the beginning of the galaxy to 4.55 Gyr ago. II This is an important result, but cosmochronology modeling has continued because of the interest in the age of the galaxy and the r-process production rates separately. All models are quite sensitive to the assumed solar system Th/U abundance ratio. The reason for the sensitivity is that the r-process Th!U production rate ratio is P 23/P238 "" 1.9 (Podo ek and Swindle. 1988), whereas 4.6 Gyr ago the solar system Th/U abundance ratio was 2.3 (assuming 3.8 today). All of galactic evolution i tied up in the difference between 1.9 and 2.3. The precision of the calculated relative production rates (order of 30% for the uranium ratio and 15% for the Th/U ratio) makes it difficult to distinguish between different model . However. in light of a new. improved Th!U solar system ratio we are able to test the compatibility of different models. specifically with different estimations of relative production rates. 1.2. Thesis outline Thi thesis ha been organized in six chapters, including this introduction; two other chapters were written as separate manuscripts and included as Appendix I and 2. Chapter 2 describe sample preparations and analytical techniques used in thi study. Chapter 3 presents results for the study of uranium and thorium distributions in equilibrated ordinary chondrites, proposing a new. improved solar system Th/U ratio. Most of this chapter is a manuscript in press. to be published in the January 2001 issue of Meteoritics and Planetary Science. Chapter 4 is a study of actjnide distributions in carbonaceous chondrites. Its main focus is leaching experiments. aimed to distinguishing the earner phases for U, Th and P. Chapter 5 discusses nucleocosmochronological models in light of an improved Th!U ratio. different 12 Appendices I and II contain the results of a project unrelated to the main subject of thi thesis. a study. that started as a part of the qualifying exam requirements in the Division of Geological and Planetary Sciences at Caltech. Thi work was done under the supervision of Dr. George G. Rossman. Appendix I presents Part I of the study of nanofibrous inclusions in the massive rose quartz, resolving a long-standing problem of the source of rose quartz coloration. This Appendix is a manuscript in press, to be published in the American Mineralogist. April 200 I. Appendix II addresses questions. raised in the Appendix I. by performing detailed single-crystal HR-TEM and AEM study of the structure of this nanofibers. determining a new phase. the polymorph of dumortierite. This manuscript is Part II of the rose quartz study, yet to be submitted to the American Mineralogist. 13 Chapter 2: Analytical techniques 2.1 Sample preparation In this work whole rock analyse as well as leaching experiments of meteorites were performed. Following the method of Shinotsuka and Ebihara 1997, we performed inductively-coupled mass spectrometry (ICPMS) analyse on dissolved samples without chemical separation of U and Th even though our samples were up to IOOx larger than these authors. In addition. we used the isotope dilution method to increase the precision of our results. A detailed procedure for the sample dissolution is described below. Procedures for the leaching experiments are described in the Chapter 4 for specific applications. Chips of chondrites, usually weighing about I gram. are surface cleaned with a dental dril l. ultrason ically cleaned in spectroscopic grade methanol. and crushed to submillimeter grain size with stainless steel tools. Sample handling is done in a laminar flow c lean bench. Chemical procedures are carried out in the geochemical clean lab of J. Saleeby. Spike amounts are estimated to give 23~ThP3n-rh and 238Uf3 5U ratios around I. Procedural blanks are run wi th each set of samples. For di solution. samples are mixed with a ~ 3 o-rh+ 235 U spike solution and treated for 10 hours in a 30 cc open Teflon beaker with a mixture of concentrated acids (10 ml HF. 2 ml HN0 3 and a few drops of HCIOJ at room temperature. The beaker is heated on a hot plate at Isooc with a screw-cup on, periodicall y hand shaken until the solution is clear. This step usually take 3 to 4 days. The solution is evaporated to a thick paste and taken up with I0 ml lOCK HN0 3 . To assure 14 dissolution, the sealed beaker is ultrasonically agitated for I 0 min and then heated on a hot plate (800C) for I hour. The solution is centrifuged. After HF+HN03+HCIO~ treatment, there is always a visible solid residue. fine metallic black specks of. presumably. chromites. The residue is treated with 0.5 ml cone. HCIO~ in an open Teflon beaker at room temperature for I0 hours. The walls of the beaker are washed with H 20. evaporated almost to dryness. and the residue dissolved in a known amount of I 0% nitric acid. All final solutions were combined and diluted to 20m I 10% HN03. For ordinary chondrites. this gives solutions of about 0.5 ppb U and 2 ppb Th. Amounts of spike and standard solutions are measured both gravimetrically and volumetrically. Amounts of reagents are measured volumetrically with digital pipettors. so that the final 20 ml volume is accurately (on the order on 1-2%) known. The mixed Th-U spike was calibrated against two different commercial standard solutions ( 235U-depleted) as well as natural U solution (courtesy of J.Chen). Results from the different standards agreed to within 1 %. Table 1.1 presents U+Th mixed spike concentrations. Table 1.1 U and Th mixed spike concentrations. Isotope Concentration. ng/g 17.03 2:12Th mu 1.509 3.265 0.0052 Uncertanty is< I% I a for masses 230. 232 and 235 ; -3% for mass 238. 15 A standard solution (Specpure® Alfa JESAR) is always analyzed along with the spiked samples. giving independent measurements of U and Th concentrations. Agreement of the Th analyses is usuall y better than a few percent. This suggests that coprecipitation losses of other unspiked lithophile elements are not important. Over the period of 3 years. changes in concentrations of standard solutions where not observed based on the periodic analyses freshly mixed standard and spike olutions. Procedural blanks (contamination from acids used for digestion plus sample handling). were measured using the arne mixtures of concentrated acids as for samples. These were evaporated and re-dissolved in 7o/c nitric acid for analyses. The procedural blank are at the level of the background 7% HN0 3 solution (::::I Oppq, corresponding to -I o-u g, negligible). Reagents Spectroscopic grade methanol was used for cleaning of the samples and as a cutting lubricant. The cleanest commercially available concentrated nitric (HNOJ). hydrofluoric (HF), hydrochloric (HCI). perchloric (HCI04 ) and acetic (HAc. CH,C0 2H) acids (Seastar Chemicals Inc., Th. U<O.OO I ppb) were used at every step of the sample preparation. For diluton, 4x distilled. 2x deionized HP (Bi-18. from D.A. Papanasstassiou. Caltech) was used. Ammoniacal EDTA (ethylenediaminetetraacetate. (0 2 C H ~hNCH 2 CH~N(CH ~C0 2 - h). used in the leaching experiments. was prepared by adding EDTA acid powder, to the O.IM NH3 solution (concentrated Sea tar HN 3 diluted with Bi-18 H 20 ) until pH= 9. 16 2.2 Inductively coupled mass-spectrometry Atomic emission spectrometry, using the inductively coupled plasma source (ICPAES) wa extensively used for multi-element analysis at trace levels at the end of the 1960s. However. soon it was recognized that AES analysis of trace elements in rock samples presents a number of problems because of insufficient sensitivity. For example, REE are measured using a time-consuming cation separation and preconcentration procedure. These problems stimulated a search for alternative methods of trace element analysis. while preserving the ease of sample introduction and speed of analysis of ICPAES. Spark ource mass spectrometry (SSMS), based on ana lyzi ng ions from the application of a discharge to a solid sample. was a lready extensively u ed for trace element analysis, but relati vely poor precision. and low sample turnaround rates made it clear that a new ICP technique with higher ion pressures wa required for adequate analysis. It appeared that an inert gas plasma might prove to be a good ion source (Gray. 1989 for review). In ICPMS. the sample is introduced to a reg1on where the gas temperature is approaching I 0.000 K. increasing the degree of ionization of even e lements with hi gh ionisation potentials. Positive ions, generated in the plasma. are extracted via a differentiall y pumped gas-vacuum interface, into a rna s analyzer designed to scan a wide mass spectrum very rapidl y. Since the introduction of the fir t commercial inductively coupled mass spectrometry (ICP-MS) instruments in 1983, the technique has gained rapid and wide acceptance for the analysis of geological. environmental. medical. biological and 17 industrial samples. Measurements can be made rapidly and with minimal sample pretreatment. making the technique both cost effective and versatile. An ICP mass spectrometer for aqueous solution analysis is shown schematically in Figure 2.2.1. The sample solution is continuously pumped into the inlet system. The flow of the solution into and out of the inlet system is usually controlled by peristaltic pumps. The solution passes into a nebulizer that disperses the sample using a stream of argon gas. A sample solution mist forms and passes through a spray chamber, where the larger sample droplets are removed by collision with chamber walls. A drain pump con tantly removes excess sample solution. The sample aerosol passes from the spray chamber into the center tube of a horizontally mounted ICP torch consisting of three concentric quartz tubes. The end of the torch is situated inside the 4-turn load coil, through which a radio frequency current is passed. The outermost concentric tube of the torch carries the plasma gas (argon). Conduction in the argon gas is initiated with a highvoltage spark. allowing energy to be transferred from the coil to the torch. The RF magnetic field produced by the coil causes oscillatory motion of free electrons within the argon gas. Further collisions of electrons and Ar atoms within the intense RF field produces additional ionization generating an approximately steady-state argon plasma. In the center of the plasma. the temperature ranges from 8,000 to I 0.000 K. The carrier gas flows through the innermost tube of the torch and delivers the sample to the plasma. This gas upply i mass flow controlled. which contributes to good signal stability. The sample aero ol is instantaneously ionized. Argon plasmas are good source of positive ions (mostly singly charged) and over 60% of the elements have an ionization efficiency of greater than 97%. The differentially pumped interface region. which transmits ions from 18 the atmospheric pressure plasma into the mass spectrometer. con ists of two co-axial conical apertures (made of nickel or platinum), the sampling cone and the sldmmer cone. This set of cones allows ions to pass into mass spectrometer. leaving behind most of the uncharged particles. The region between the cones is evacuated by a rotary pump, whereas turbo pumps maintain high vacuum in the focusing region behind the skimmer. 19 _, ·.::" ~ ~ ~ - '-' ~ i~ ~ ~ Plume Jon beam Plruma tord1 Quadrupole Electmn multiplier [[~~~~~~~~3=5l~unple aerosol moss filter "'•-++-+-+-+-1magner secwr " Sil!,lllll Out HiRh \"CICIIIIIIJ 1-1&" -10'' mbarJ Electroswtic " le11s assemb/,· I lmemredwte Ll)W \'ll('lllllll \'l/( 11111/J t-2mlmr) f-lo-' mbar) pump Stmtf>h' .\olutum Figure 2.2.1 Schematic of an inductively coupled plasma-mass spectrometer. 20 Quadrupole ICPMS The rapid scan rates of ICPMS with quadrupole mass analyzer allows nearsimultaneous determination of most elements in the periodic table at level down to I 0 pg/ml. A diagram of a typical quadrupole mass filter is shown in Figure 2.2.2. In a quadrupole analyzer. mass separation is achieved solely by electric fields. The analyzer has a hyperbolic cross section and consists of four metal rods parallel and equidistant from the axis. Opposite pairs are connected together. and to each pair a potential U+U 1cos(2rrut) with opposite polarities. is applied (see Fig. 2.2.2a). The direct current voltage (DC) is positive for one pair and negative to another. The radio frequency (RF) voltages on each pair have the same amplitude but are opposite in sign (i.e .. they are 180° out of phase). By varying these voltages. the rods act as a mass filter. allowi ng only ion of a specific-to-charge ratio to pass through the center of a quadrupole at any given combination of applied voltages (Fig. 2.2.2b). All other masses are unstable and collide with the rods. These voltages a re ramped very rapidl y so the quadrupole can scan the whole mass range (2-260 m/z) in I 00 milliseconds. 21 Tons +U DC supply RF generato -U cos 21tYt I Jk:e~ns --.... I -~~Analyzed + Positive rods Negative rods I ion; ·~ Analyzed ion; Heavier ions Figure 2.2.2 a) Schematic of a quadrupole mass filter. U is a fixed direct current (DC) potential (up to 70 volts) and U 1 is the peak amplitude for a radiofrequency (RF) potential (maximum of about 400 volts) of freque ncy u (typica ll y 2 MHz); b) Illustration of the ion paths in each of the two rod planes. In the positive rod plane quadrupole acts like a hi ghpass mass filter. in the positive pla ne- li ke a low-pass mass fi lter. Since high-pass a nd low-pass filtering are superimposed, the two form a system that transmits io ns only at the m/z of interest to the detector. M odified from Jarvis ( 1997). 22 Quadrupoles are limited effectively to unit mass resolution so they can't resolve polyatomic or isobaric inteferences. However. of all of the elements detectable by ICPMS. only indium does not have an isotope that is free from overlap by another element. After passing through the quadrupole. ion currents of a specific rn/z ratio are measured by an electron multiplier detector. Depending on the concentration of the ana lytes being measured, the detector operates in two mode : the pul e count mode for concentrations ranging from detection limit to -I ppm in solution. and the analog mode for concentrations between I and I 00 ppm. Single collector magnetic sector ICPMS Since the introduction of the first commercial system for ICP-MS employing a quadrupole filter as mass analyzer, the problem of spectral interference (molecular and doubly charged ions) has always been a significant impediment, because the unit mass re olution of quadrupoles is not sufficient to separate the ions from interfering species. The most straightforward approach to overcome limitations from spectral interferences is the application of higher mass resolution by focusing both energy and angle of the ion beam in the magnetic sector field mass-spectrometers. 23 2.3 Instrumentation used in this work HP 4500 ChemStation Quadrupole mass spectrometer The HP 4500 is one of the most stable. accurate. and sensitive of the current commercial single-quadrupole ICP-MS instruments. It can detect as little as lppt (by mass) of an element in solution and as much as I ppm of most metals without significant deadtime cou nting losse and can do this in one minute. For routine quantjfication 1-1000 ppb is the usual working range. The HP 4500 features the Babington-type nebulizer, in wh ich the analyte solution run down a narrow V -shaped groove. The aerosol is produced when solution passes over a small hole in the bottom of the grove. through which argon gas flows at high pressure. A Peltier cooler cools the spray chamber. enabling the temperature inside the chamber to be precisely controlled, which give the HP 4500 a very stable ion signal. HP4500 Tuning for U and Th analyses To achieve high signal sensitivity. the ion beam extracted from the plasma must be focused before entering the quadrupole mass analyzer. The quadrupole and detector are mounted off-axis to the ion beam entering the system from the interface. preventing photons and neutral species from reaching the detector thu giving the HP 4500 a very low random background re ulting in high detection limits. The ions are focused into the analyzer. where they are separated by their mass-to-charge ratio (m/z). Although it is not necessary to tune HP 4500 daily for many applications, in this work optimization was performed routinely to achieve optimum sensitivity. 24 Optimization of instrument performance consists of tuning for: (I) sensitivity, (2) reducing oxide ions, (3) reducing doubly charged ions. and (4) adequate mass resolution. Within the tuning windows of the HP 4500 ChemStation software, one can adjust parameters for the sample introduction rate. torch box, analyzer and detector systems. Sensitivity is strongly dependent on the sample uptake rate. carrier gas flow, temperature of the spray chamber (2°C for aqueous samples), radio frequency power. and sample "depth" (distance between the edge of the load coil and the tip of sampling cone). Tuning for the highest sensitivity almost inevitably increases the level of oxide/hydroxide or doubly charged ions. To reduce the level of oxides, key factors are plasma temperature. lower sample introduction rate, lower carrier gas flow and longer residence time for ions in the plasma (sample depth ) produced by increasing distance between end of the torch and the sample cone. The level of doubly charged ions is anticorrelated with the amount of oxides, so a compromise must be reached for higher sensitivity within acceptable limits of polyatomic compounds and doubly-charged ions. For the mass 230240 region. there is no evidence for molecular interferences. Using a mixed Th and Ba solution. the instrument was tuned to have counting rates of oxide /hydroxides and doubly charged ions in the plasma below I and 2 percent respectively. For most applications the HP 4500 is optimized using a tuning solution with a few elements covering the whole range of masses. assuring good sensitivity at all masses simultaneously. However, for this study the HP4500 was optimized using a I ppb uranium standard solution to obtain maximum count rates. usually on the order of 1600 cps for both U and Th. There is a tradeoff between sensitivity, mass resolution, and shape of the peak. The width of every peak is adjusted to give good resolution for adjacent 25 masses. For sufficient resolution. the peak width at I0% of a peak height for all masses should be 0.65-0.8 AMU and a peak center should be within 0.1 of the selected mass. Depending on the mass analyzed and tuning parameters. the top of the peak can be made flat to within a few percent for four voltage steps. In this work, for quantitation of U and Th, counts are taken for 3 voltage points on the peak top to provide a single mass peak intensity reading. Exploiting the rapid scan capabilities of the quadrupole to minimize the effects of drift. each mass peak is counted for 0.3 sec (0. 1 sec. per voltage point). A data block of 5 such measurements is averaged, and the counting rates and ratios of average counting rates (232/230 and 238/235) tabulated. For such blocks. drift in the intensity is not important. For our work 40 such data blocks have been taken for a given sample, which consumes about 3 cc of solution or about 15% of the sample. leaving the remainder for analysis of other elements in the same solution. The di stribution of ratios from the blocks appears normal , with a standard deviation only slightly above that expected from counting statistics. justifying the calculation of an error of the mean for all the data. Counting rates for pure HN0 3 for masses 230. 232. 235. 238, as well a for masses 23 I. 233. 234. 236 and 237 are roughly 4 cps. Background corrections are made from counts at masses 230. 232. 235. and 238 from the 7% HN03 carrier solution. Counting rates for heavy (e.g .. > REE) elements are about the same as for U and Th. but for lig ht elements (masses <56) re-tuning of the instrument is needed for optimum performance. Instrument drift is initially about 3% per hour, but after first 1.5-2 hr. of work, drift becomes undetectable. Sample and standards can be switched in about 7 minutes in 26 programmable sequences. enabling good precision for analyses comparing counting rates between sample and standard. Of the 7 minute intersample time, 5 minutes are spent for washing the machine with pure 7% HN03. This time is shown to be sufficient to bring the counting rates to the background level after analysis of I ppb ample olution. Typical observed standard deviations for ~3~ThP<>rh and 238 U/235U are <1.5%. Instrumental mass fractionation correction was made on the basis of measurements of an isotopically normal U solution on three separate days. The correction of 0.5% per amu was applied to both 23<>rh!232Th and 235 UP 8U. ELEMENT Finnigan MAT HR-ICP-MS A diagram of this single collector ICPMS is shown in the Figure 2.3.1. The mass spectrometer consists of po t-extraction focusing lens system, the magnet sector. the electrostatic analyzer and the electron multiplier. Sample is introduced into the spray chamber through the nebulizer. In thi work both standard pneumatic (Meinhardt) and self-aspirating rnicroconcentric nebulizers (CETAC MCN-1 00; Figure 2-3-2) were used. The Meinhardt is essentially two concentric tubes. Through the inner one the sample solution is introduced with a help of peristaltic pump. Through the outer one Ar gas is supplied. To achieve maximum sensitivity. the flow rate through the Meinhardt should be approximately 1 ml/min. The MCN- 100 consists of a PVDF body with a Teflon nebulizer capillary and a sapphire nebulizer insert. The MCN-1 00 i a self-aspirating nebulizer. Argon gas flow produces the vacuum which drives the sample solution into the nebulizer. The MCN-1 00 offers benefits of low sample uptake with high fraction of the solution injected into the plasma. 27 With flow rates of only 30 ~1/min. the MCN- I 00 can provide the detection limits of a pneumatic nebulizer. Using the MCN-1 00. analyses can be performed on small volume (less than I ml) samples. The decrease in the amount of sample analyzed combined with high sensitivity of ICPMS translates into small sample consumption, allowing for the analyses of wide range of element in the same sample aliquot. The beam of plasma with ionized sample passes through the cones, extraction lens. and a set of focusing and shaping (to a line) lenses before entering the flight tube through the entrance slit. The beam then becomes the source for the analyzing system which, in tum. focuses the beam on the exit slit (see Figure 2.3.1 for diagram). The ELEMENT system has an inverse Nier-Johnson geometry. i.e., the magnet is located in front of the electrostatic analyzer (ESA). The ELEMENT has been designed for multi element analysi in low. medium. and high resolution mass modes. The electro tatic analyzer (ESA) focu e the ion beam both in angle and in energy. Ma re olution of the analyzer is accomplished by means of two mechanical slits, one at the entrance (source- or entrance-slit) and one at the exit (detector- or exit-slit) of the ion optical system (see Figure 2.3.1 ). The smaller entrance slits are used for higher resolution modes. Analyses are possible with the use of different resolution modes for different mas es in the same sample. Time delay for the switch of the resolution slits is approximately 0.3 sec, so for muli-resolution analyses it i more convenient to scan all the masses at one resolution. then switch slits. and measure all the other masses. The peak shapes produced by the ELEMENT are either triangular (medium or high mass resolution) of tlat-topped (low mass resolution). The mass range extends from 2 to 260 amu . 28 ·-- ·- , CJ I Focusand rotation high resolution quadrupoles • 1 Exit slit -~ EM Focus and rotation quadrupoles ~-~-~ co .--..-- ----..., Entrance slit ShaJ:ie oenJction Foc4s Lens system Skimmer and Sampling cone Ions from torch Figure 2.3.1 Schematic diagram of Element ICPMS. ESA- Electrostatic Analyzer; CDConversion Dynode; EM -Electron Multiplie r. Modified from Element Operating Manu aL Revision A. 29 Nebulizer gas port (2) Sample port Sapphire :::C=>-t-4=~==:;i==~==:;}- nebulizer PVDF nebulizer body orifice Polyimide nebulizer capillary ( I) Used to attach the polyimide nebulizer capillary to the sample tubing (2) Used to connect the nebulizer gas supply from the ICP instrument to the MCN Figure 2.3.2 CETAC microconcentric nebulizer (MCN-1 00). 30 Finnigan ELEMENT HR-ICPMS U and Th analysis The instrument was optimized using a multi-element I ppb solution for occasional mass calibration and I ppb Indium solution for everyday tuning for sensitivity. Routine tuning includes sample gas flow. lens parameters and plasma power tuning for maximum sensitivity. and plasma parameters for minimum multiple ionisation and minimum molecule-generation. The sensitivity is the highest for high rna se (i.e .. for U). This is because light masses tend to be deflected from a straight trajectory upon passing through the cones. U and Th analyses were obtained using low resolution mode of the ELEMENT. The resolution R of ELEMENT is defined by the peak distance A between 56 (Ar0) and 56 Fe : R- 2 5 00 B *A ' where B is the width of the ArO peak at S'k height. For low resolution. R=300, which is much hi gher than the rna s resolution of the quadrupole ICPMS. The counting rate for the I ppb In solution is ideall y I 0 6 counts per second (cps) for both nebulizers used in this tudy. The sensitivi ty over the whole mass range spans between I 0.000 cps for I ppb Li and 2* I 0 6 for U. Dark noise is usually 0.1 cps. The only molecular interferences expected for the rna ses of interest in this study are platinum argides. The most abundant is 195 PrmAr. interferring with mu. After tuning the machine for minimum oxide/hydroxides (on the order of Ba0/Ba+::::Q.003). there is no evidence of ignificant input of PtAr in the peak of mass 235. Mas es 230. 232. 235 and 238 are measured usi ng e lec tro tatic scans with the magnet sitting at the lowest mass. At low re e lution. peaks have a flat top. Figure 2.3.3 illustrate a U peak at mass 238. An average of 50 scans with I00 point steps was taken over the whole width of the !Jsu peak. For a nalyses. I 0 point teps were taken over only 31 the central 10% of the peak width for each of the 4 masses. Each point step was scanned for 0.0 I second. or 0.1 sec per peak. One scan consist of 4 measured masses; 49 such scans were performed, with a total analysi time of T=4*0.1 sec*49+peak shift time < I min. Typical standard deviation for measured 232/230 and 238/235 ratios is a few permit. Corrections for instrumental mass fractionation were based on the measurement of the 238/235 ratio in the normal U solution. Typical correction i 0 .5% per amu. The background level from the HN03 solution is usually < 10 ppq for all masses. Isotopic ratio analy es of the same samples with the HP quadrupole ICP and ELEMENT sector ICP agree within the uncertainty. 32 400000 350000 300000 rll Q. 250000 u .....;. 200000 c ....c 150000 ·r;; Q,j 100000 50000 0 237.7 237.8 237.9 238.0 238.1 238.2 238.3 238.4 238.5 238.6 Mass, amu Figure 2.3.3 Low resolution electrostatic scan ofthe 238 U peak. illustrating flatness ofthe peak top. 50 scans were col lected with a I 00 point steps in every scan. 33 Finnigan ELEMENT HR-ICPMS P analysis Phosphorus analyses were done with the Finnigan magnetic sector single collector ICPMS. Medium resolution (4000) was used in order to resolve the 3 1P peak from 15N 160 and 14 N 160 1H (Figure 2.3.4). At medium resolution, the shape of P peak is triangular. Counts were taken stepping over a range of 0.0 I amu. centered at the P peak, wich covers >I 00% of the P peak width at the base. 100 point counts per scan were taken. and I 00 scans were performed. ELEMENTs software is programmed to find top of the peak for each scan. The 100 scans were averaged and 30% of the peak top was integrated. It was found necessary to use the method of standard addition for phosphorus ana lyses to avoid matrix effects. Three aliquots of the sample were analyzed. one with no P added. one with a known added amount of P approximately equal to the sample solution and one twice as much. For quantitative analysis. the sample solution and the two P-spiked sample solutions must produce a linear calibration trend. Preliminary results showed that the common standard add ition method does not produce satisfactory results. Figure 2.3.5 shows that although the three sample solutions follow a general positive trend, the linear fit is not good. This is probably due to ion beam current instability. To avoid this effect as well as other processes affecting counting rates. C was added as an internal standard to aJI analyzed solutions. Since the sample olutions contain high concentrations of Mg. ~5 Mg as well as Cs was used for normalization of P counting rates. One hundred P/Cs and P/Mg ratios were collected for every sample. Results for P concentrations are consistent between Cs and Mg normalized counting rates (Figure 2.3.6). Typical sigma standard deviation was <3%. Reproducibility of analyses performed on different days is on the order of few percent. 34 !\=0.0103 L1=0.02l9 Figure 2.3.4 Medium resolution scan of the mass interval 30.975-31.020 amu, showing resolution of the phosphorus peak and 15N 16 0 a nd 14N 16 0H. arising from the HN0 3 carrier solution. Peak corresponds to 7 ppb P concentration. 35 60000 R2 = 0.9728 50000 • 40000 • 30000 fll Q. u ~"' 20000 10000 • 0 -500 0 500 1000 1500 P, ng added Figure 2.3.5 The Standard Addition method for P without internal standardi zatio n. Slope of the line (cps/ppm P) used to calculated P concentratio n in the unspiked sample sol utio n. Graphicall y, extrapolation of the cps P vs. ng of Padded line to P(cps)=O gives the concentratio n of Pin the unspiked sol uti on. 36 0 .08 R2 = 0.9991 • 0.07 • P/Cs • P/Mg 0.06 0 +.J 0.05 ~ I... +.J c 0.04 E Q) 0 .03 P concentration in the sample solution: P=275 ppb • • R2 = 0.9997 Q) Q) ......... a. 0 .02 0.01 0 -500 • • • 0 500 1000 1500 P, ng added Figure 2.3.6 Demonstration of the Standard Addition technique with normalization of the P counti ng rates against Cs and 25 Mg. 37 Chapter 3: Ordinary chondrites 3.1 Sample selection and preparation As discussed in Chapter 1, the bimodal distribution of Th!U in ordinary chondrites is surprising. Based on the previous literature data and issues raised by Figure 1.3, we focused on the anomalous (with respect to Th/U) chondrites, Harleton and Glatton. Rather than assume that all specimens of these have anomalously high Th!U, we e tablished experimental procedures to be able to perform high accuracy isotopic dilution Th and U analyses. With this approach. we could guarantee that the samples we were characterizing did in fact have anomalous Th/U. This conservative approach turned out to be very important and led to solution of the problem. Analyzed samples are listed in the Table 3.1.1. Two fragments of Glatton and 5 individual multigram fragments of Harleton (I. II. U. 4.5.6) were subdivided into approximately gram- ized ample . Only 2 samples (H6-3 and H6-5) are substantially smaller, about 30 mg. These two samples are round. presumably fragments of relict chondrules. Both are gray and visuall y homogeneous, H6-3 being darker than H6-5. The last column in Table 3.1 reflects which samples came from the same rock fragment. USNM 2576 and USNM 6625 are separate stones in the Harleton fa ll. All samples were prepared as described in the Chapter 2. For reference. samples of Leedey (L6) and Allende (CV3) meteorites were analyzed. Petrography of Harleton chondrite 38 The Harleton meteorite fell four miles east of Harleton, Harrison County, Texas, on 30 May I96 I (Clarke at a!.. 1974-75 and references therein). A single stone. weighing 8.36 kg was recovered within ten minutes of the time of the fall. Macroscopically. the interior of the meteorite appeared remarkably homogenous and compact, with a few, poorly defined. chondrules. Except for a slightly higher FeS content, there is no significant difference in the chemical composition of the Harleton meteorite with an average L(4-6) chondrite composition (Table 3.1.2). Thin sections under the electron microscope (SEM) show granular agregate of olivine, orthopyroxene and opaque minerals (troilite, metal and chromite). Uniformity of the reflectance of the olivine and orthopyroxene grains in SEM BSE image mode implies homogeneously distributed Fe and Mg contents (characteristic of the equilibrated chondrites). Relict chondrules are rare and poorly defined. Phosphate grains are both chlorapatite and merrillite. Apatite grains as big as 200 ~m were obseved (Figure 3.1.1 ). Modal analyses of one I Ox4 mm section of Harleto n are reported in the Table 3.1.3. These are compared with normative composition. calculated from the chemical analy e from Table 3.1.2. From the chemical and mineralogical compo ition in Tables 3.1.2 and 3.1.3. it is evident that Harleton is typical equilibrated L6 chondrite. The sizes of the pho phate grains seem to be the only characteristic feature of thi s meteorite. It should be noted that total amount of phosphates (as well as total phosphorus in the rock) is within the ran ge of L-chondrites (see Tables 3. 1.2. 3. 1.3). 39 Table 3.1.1 Sources of meteorite samples. #,year received and source Leedey (L6) #489 .15 Center for Meteorite Studies. Arizona State University Allende (CV3) USNM3529 split 9. position 2, 1998, National Museum of Natural Hi story, Smithsonian Institution (NMNH) Glatton (L6) Glatton I BM, 199l,M3 British Museum. UK Glatton II BM, 1991.M3 British Museum. UK Glatton W G .J. W asserburg Harleton (L6) Harle ton I USNM2576, 1992, NMNH Harleton II USNM2576, 1992, NMNH Harleton U D . Unruh H4-2 USNM2576, 1998. NMNH H4-4 USNM2576, 1998, NMNH H4-6 USNM2576, 1998. NMNH H5-2b USNM6625, 1998, NMNH H6-1 USNM2576, 1998,NMNH H6-3* USNM2576, 1998, NMNH H6-4 USNM2576, 1998, NMNH H6-5* USNM2576, 1998, NMNH H6-6 USNM2576, 1998. NMNH · Small (- 30mg) relict chondrules. 40 200p.m Figure 3.1.1 Backscattered secondary electron Image of the Harleton meteorite thin section. outlining big. 1OOX200 ~m apatite grain. 41 Table 3.1.2 Chemical composition of Harleton chondrite compared to average L chondrites. Constituent 1 Harleton. L61 L(4-6) chondrites3 Si02 39.85 39.69 MgO 24.77 24.72 Al203 2.14 2.30 Cr203 0.52 0.53 MnO 0.33 0.34 Ti02 0.10 0. 12 CaO 1.81 1.82 Na20 0.92 0.94 K20 0.1 0.11 P205 0.23 0.23 FeO 14.3 14.3 1 FeS 6.58 5.78 Ni 1.14 1.24 Co 0.06 0.07 Fe-metal 7.28 7.37 Fe-total 22.6 22. 17 1 Composition in weight percent:--·------1 3 Clarke et al. 1974-75. Jarosewich and Dodd 1985. 42 Table 3.1.3 Nonnative composition (volume percent) of Harleton compared to average L chondrites. - - - - - - - - - - - - - - - - - - -- Constituent Harleton ' Harleton" L(4-6) 3 Olivine 41.3 43 .7 43.6 Orthopyroxene 31.6 27.4 27.5 Clinopyroxene 4.3 5.3 6.0 Feldspar 11.1 13.4 13.0 Apatite~ 0.4 0.6 0.6 Chromite 0.6 0.6 0.7 Troilite 4.2 5.0 4.4 Nickel-Iron 6.4 4.0 4.2 1 R. Radtke (personal communication), made by electron microprobe on section from USNM2576. " Clarke et al. 1974-75; norm based on chemical analysis in Table 3.1.2. -' Jarosewich and Dodd 1985; nonn. ~ "Apatite" represents total amount of phosphates. 43 3.2 U, Th, P analytical data Th and U results, given in Table 3.2.1 , are based on the quadrupole ICP analyses. They are corrected for mass fractionation. Mass fract ionation correction was made on the basis of three measurements of an isotopically nom1al U solution. The correction of 0.5% per amu was applied to both "-1<Thf3 2Th and 235 UP 8 U. Procedural blanks (contamination from acids used for dige tion plus sample handling during dissolution) were measured using the same mixtures of concentrated ac id as for samples. These were evaporated and re-dissolved in 7 % nitric acid for ana lyses. The procedural blanks are at ""10· 13 g for U and Th (negligible. see analytical techniques in Chapter 2). Table 3.2.1 gives Th, U. Th/U, and P for the analyzed samples. We also tabulate data for previous Th and U analy es of Harleton and Glatton. Samples of Harleton and Glatton were expected to be anomalous. so a sample of Leedey (L6) was analyzed a representative of a " normal'· L chondrite. Our Leedey sample was an aliquot of an o lder homogenized chondrite si licate sample prepared by crushing a multigram sample. Larger metal fragments were removed during the crushing: conseq uently. our U and Th concentrations are somewhat higher than typical values. However. as given in the table, our Th/U for this meteorite is in good agreement with previous literature data. Table 3.2.1 shows that all of our Glatton ample have typical U and Th concentrations for L c hondrites. The Glatton P concentrations show a total range of about a factor of 1.5 and the average is somewhat higher th an the average L chondrite literature 44 value. None of our Glatton samples show the anomalously high Th/U found by Chen et a/., 1993. By analyzing a large number ( 12) of Harleton samples. we obtained a wide range of Th/U, including low as well as normal va lues of Th/U, in addition to only high values reported by Unruh, 1982. The bimodal distribution in the previous Literature Th/U values (Figu re 1.3) for ordinary chondrites is only apparent. Specimens of Harleton show a range in Th/U from 2.5 to 6, larger than the total range observed in previous isotopi c dilution analyses. The variations are mostly due to variations in U content. U varies by a factor of 5 ; Th by a factor of 2. We will focus on the Harleton data in the subsequent discussion. Figure 3.2.1 shows that the Th/U variations tn Harle ton display a linear correlation when plotted aga inst 1/U. Taken at face value, the linear trend implies that the Th/U variations are the result of different amounts of mi xing of hig h U, low Th/U and low U. high Th/U end member . There is no obvious grouping among the samples from the same large fragments (H4, H6, Hl,ll). In fact, the largest Th/U differences among gmsized samples is between H6-4 and H6-6. Sample H6-5 (smaller sample, presumably relict chondrule, see Table 3.2. 1) showed the highest Th/U ratio. Figure 3.2.2 shows that. in hindsight, all high precision data for ordinary (types 4-6) chondrites also conform reasonably well to the Harleton trend. An exception is the ample of Glatton (Chen et al .. 1993). whic h is characterized by a high Th and Th/U (Table 3.2. 1) and lies distinctly off the Harl eton tre nd. No hi gh precision data are ava ilable for type 3 unequilibrated ordinary chondrites. No differences are observed among H. L and LL chondrites. 45 For equilibrated ordinary chondrites. there are three components which affect Th/ U and 1/ U. Two phosphates (merrillite and chlorapatite) are well characterized as phases in which actinides are concentrated. Based on Crozaz ( 1974. 1979), the Th/U for merrillite is high (9-11 ), but much lower in apatite ( 1-3) (Table 3.2.2). So. 2 actinide-rich end members with sufficiently different Th/U are known to be present. The third component is the proportion of admixed inert (i.e. actinide-free) phases to the phosphates. The mi xing line between the two phosphates is shown in the left-hand side of Figure 4 .2.3. Random mixtures of the two phosphates and inert material might be expected to show a lot of scatter on a Th/U vs. I!U plot (addition of the inert material will dilute U content, but will not change Th/U ratio), but a good correlation is observed, especially clear in the Harleton data. Thus, a different mechanism is required to explain Figures 3.2.1 and 3.2.2 than just random mixing of phosphate end members and inert material. 46 Table 3.2.1 U. Th. P analytical data. Sample ID Weight (g) A&G* 238 U. ppb* 232 Th, ppb· Th/U+ p (ppm) 8.1 29.4 3.63 1180 Allende 0.992 12.54 43.28 3.43 1235 Leedey A 0.999 17.48 60.6 3.44 1040 Glatton5 I 12.5 74 5.87 na Glatton I 0.979 9.34 36.1 3.83 920 Glatton II 0.891 11.7 43.8 3.70 1260 Glatton W 0.853 12.2 46.3 3.77 14 10 Harleton- I# 0.096 6.17 37.5 6.03 na Har1eton-2# 0.075 5.47 32.7 5.93 na Harleton-3# 0.073 5.07 32.8 6.42 na Harleton I 0.937 12.38 39.6 3.17 1910 Harleton II 0.957 12.78 43. 1 3.35 1460 Harleton U 0.836 11.27 39.8 3.50 1790 H4-2 0.756 10.14 35.8 3.5 1 936 H4-4 1.154 12.47 36.7 2.93 11 40 H4-6 0.992 11.49 39.3 3.39 1110 H5-2b 0.466 8.26 36.5 4.39 9 16 H6- l 1.118 12.94 43.6 3.34 985 H6-3 0.029 14.1 4 1.4 2.91 2240 H6-4 1.088 7.03 3 1.0 4.38 1550 H6-5 0.028 8.03 48.8 6.04 424 H6-6 0.688 25.05 63.5 2.52 630 • ID-ICPMS analyses. corrected forMS isotope fractionation. "' Th/U is a weight ratio. accuracy< 1.5% ( Icr). ! Cl chindrites. Anders and Grevesse, 1989. 5 Chen eta! .. 1993 (and personal communications). 4 Unruh. 1982. 47 Table 3.2.2 St. Severin phosphate (data from Crozaz, 1979). Pha e (#of grains) U, ppm Th. ppm Th/U (wt. ) Chlorapatite ( I) Merrillite (7) 7.5±0.8 0.275±0.025 11.2±3.8 3.21 ±0.25 1.5 11 .67 48 7 0 • 6 0 0 5 ;:) • • 4 ..c E-< • I • • •• 3 • 2 1 0 .00 0.05 0.10 0.15 0.20 0.25 1/U, ppb- 1 Figure 3.2.1 Th and U isotope dilution data for Harleton. Fil led circles - this work, open circles from Unruh. 1982. The linear trend o n the plot Th/U v . 1/U implies a two component mixing between hi gh U. low Th/U. and low U and high Th!U components. 49 7 • 6 • • • 5 ~ ..c: E- • • 4 /}. ll 3 ll 6'- ~ • AI! 2 1 0.00 0.05 0.10 0.15 0.20 0 .2 5 1/U, ppb-1 Figure 3.2.2 Combined Harleton data (circles) and literature data (triangles) on ordinary chondrites (from Tatsumoto et at., 1973. Unruh. 1982, Hagee et at., 1990, Chen et at., 1993. Shinotsuka et at., 1995 and this study) data. Ordinary chondrites are mostly types L4-6 plus 5 data points of St. Severin (LL6) and 3 point of of H5. This plot shows that ordinary chondrite data follow the same linear trend as Harleton. 50 11 r merrillite 9 7 ..c E- • • • • ~ 5 3 1 2 0 apatite 0.004 • 0.05 .,.. • • 0.1 0.15 0.2 0.25 1/U, ppb Figure 3.2.3. Model I. Phosphates data from (Crozaz. 1979). Line on the left-hand side represents mixing between 2 phosphates. Dilution of point on this line by random factor will produce scatter in the right-hand side of the p lot (note change in scale). Fil led circles are Harleton data. L inear trend through these data is calculated in the Model I apatitemerrillite mjxing. using dilution factor of 100. 51 3.3 Phosphate control on the actinide abundances Model I A very simple model can quantitati vely fit the trend observed in Figure 3.2.1. The left-hand side of Figure 3.2.3 shows the location of the merrillite and apatite points from St. Severin. LL6 (Crozaz, 1979). Relati ve to Figures 3.2.1 and 3.2.2. an expanded 1/U scale is used here to illu trate the much higher U concentration in apatite compared to merrillite. In Model I we use the St. Severin phosphate Th and U concentrations from Crozaz ( 1979). Numerical values are summarized in Table 3.2.2. The dashed curve on the right-hand side of Figure 3.2.3 is calculated by simply taking the St. Severin phosphate data and diluting the U concentrations of both phosphates by a factor of I 00. simulating the effect of mixing I part phosphates with 99 parts of inert material. The slope of thi curve fits the data for equilibrated ordinary chondrites. The very fact that the slope of the data trend in Figure 3.2.3 matches the phosphate mixing slope says that phosphate variations are indeed the reason for the whole rock Th/U correlation. In this model, analyses that are known to involve three components (two phosphates and inert material) described by a two component mixing line because the phosphate mixture is uniformly diluted by inert material. The variation along the line reflect the different proportions of phosphates. with apatite (the high U end member) being especially important. The observed data points correspond to a range of apatite/merrillite from 0.03 (hi ghest Th/U) to 0.3 (lowest Th/U). 52 Petrographic observation on polished sections of Harleton revealed the special nature of this meteorite. It appears to have a larger proportion of apatite relative to merrillite with large (up to 200 micron. as shown in Figure 3.1.1) apatite grains. In the context of Model I, the proportions of apatite are not unprecedented; however, the unique aspect appears to be the large grain size. With more-or-less normal amounts of U-rich apatite concentrated in large grains, fluctuations in the number of these grains drives large variations in Th/U in gram-sized Harleton samples. In particular. samples which by chance have few or no large apatite grains. but typical amounts of merrillite, will show high Th/U ratios, explaining the Unruh. 1982 data which were the motivation for this study . lt is quite ironic that control by apatite (a phase with low Th/U) was first revealed by samples that showed high Th/U ratios. Furthermore, given this non-intuitive feature in the data, it is the ability of mixing models to quanTitaTively describe the Th/U vs. 1/U trend that strongly argues that we have the correct interpretation. The success of Model I must be respected; however. it is over implified and not correct in detail, as we now proceed to discuss. Before launching into complications, it is important to note the conclusion, which we believe to be generally valid. that the chlorapatite control of the U and Th/U distributions very likely reflects heterogeneities in the Cl distribution in the original parent bodies. These Cl heterogeneities then are the ultimate control on the U and Th/U distributions. Model I make testable predictions. The required constant dilution factor of I 00 implies a total phosphate abundance of I %. or equivalently. a whole rock P content of around 2000 ppm. but these are a factor of 2 higher than typically observed in ordinary chondrites. Thus. Model I predicts: (I) similar U and Th contents in Harleton and St. 53 Severin phosphates; (2) a uniform P content in the Harleton samples (variations of up to about ± 20% would be permitted by the scatter in Figure 3.2. J ); (3) P concentrations for Harleton that are roughly twice that typically found in ordinary chondrites. Preliminary ion probe data for Harleton phosphates Our preliminary ion probe data for Harleton phosphates show that in Harleton, as wel l as in St. Severin, phosphates display the same characteristic high Th/ U (merrillite) and low Th/U (apatite) values. Ion probe analyses were performed with UCLA IMS-1270. Data for ThO and UO were collected relative to 40Ca. Standard error for ThOJUO is between 1.5 and 3.7%. Standard error is based on 20 cycles that were taken in the measurement. Durango apatite wa used a a standard (Ian Hutcheon. Lawrence Livermore National Laboratory). Concentrations of U and Th in Harleton phosphates. St. Severine apatite from this work as well as Crozaz. 1979 are presented in Table 4.2.3. Analyses of the Durango apatite standard are also listed. Although these results show that acti nides in Harleton phosphates follow the same rules as phosphates from St. Severin. actinide concentrations are lower. especially for apatite. However. our data a lone are not sufficient to make definitive conclusions. since only the two largest phosphate grain were analyzed. 54 Table 3.3.1 Th and U ion probe data of phosphates. U.ppm Th, ppm Th/U (wt.) Chlorapatite ( I ) 1 7.5±0.8 11 .2±3.8 1.5 Merrillite (7) 1 0.275±0.025 3.21±0.25 11.7 Merrillite (If 0.325 3.25 lO Chlorapatite ( I)2 3.6 4.8 1.3 Merrillite ( 1)~ 0.2 1.9 9.5 16.3±1.1 313±9 19.2 Phase (#of grains) St. Severin Harleton Durango apatite 3 1 Crozaz, 1979. ~ Thi s work. see text for uncertanties. 3 I. Hutcheon. personal communication. 55 Phosphorus analyses. Major advantages of our ICPMS techniques are that only a small fraction of the sample is consumed in the Th/U determination and that it has excellent sensitivity for the determination of other elements. Thus we were able to test the predictions of Model I by measuring P on exactly the same samples for which U and Th are measured. Figure 3.3.1 summarizes the results of P analyses: (I) In spite of the prediction of Model I, P concentrations are not constant in Harleton. A total range over a factor of 5 is observed. The most deviant P concentrations are observed on our smallest (30 mg) samples; however, even when these are omitted, a range over a factor of 3 is observed. (2) The observed P contents are not uniformly high compared to the average 1000 ppm for L chondrites (Moore, 1971 ). Based on large ( I 020 g) samples. the P from the chemical analysis of Clarke et al., 1974-75 (Table 3.1.2) is I 004 ppm. Our average Harleton P concentration is 1240 ppm. Only 2-3 samples have near the 2000 ppm predicted by Model I. (3) There is no clear correlation of Th/U with P. Similarly, there is no correlation of P with U or Th separately (see Figures 3.3.2 and 3.3.3). 56 7 • 6 5 4 - 3 ;::J • ..c ~ •• • • • •• • • • 2 1 0 0 500 1000 1500 2000 2500 P, ppm Figure 3.3.1 Th/U vs. P for Harleton. P concentrations show much bigger scatter than predicted by Model I. 57 30 • 25 20 ~ c. c. 15 • •• • • • • •• 500 1000 1500 "' ::J 10 5 •• 0 0 2000 2500 P, ppm Figure 3.3.2 Plot of measured U concentrations vs. P content in the same sample aliquates. 58 70 • 60 so • • 40 .Q Q. Q. ... -= •• •• •• • • • 30 ~ 20 10 0 0 500 1000 1500 2000 2500 P, ppm Figure 3.3.3 Plot of measured Th concentrations vs. P content in the same sample aliquots. 59 Thi s is a failure of the Model I. However, this is probably not a failure of the concept of apatite control, but instead, it indicates the need for a more complex model. The most interesting feature of the Harleton data is the variability of P concentrations. By itself, the failure of the Harleton samples to show the 2000 ppm predicted by Model I could have been explained simply by abandoning the St. Severin phosphate U concentrations and assuming instead that the Harleton phosphate U concentrations are roughly twice as large. In the mi xi ng model there is a tradeoff between dilution factor and phosphate U concentrations. Our ion probe data show that U concentrations in apatite are roughl y the same as for other ordinary chondrite apatites ( 13.5 ppm, Crozaz et al., 1989). The St. Severin apatite U concentration is already the highest. Thus, there is no observational support for increasing U in apatite to compensate for this failure of Model I. Howe ver. all actinide measurements are biased towards the biggest grains. This will be considered later. The lack of uniform P content as well as lack of correlation between P and Th!U (Figure 3.3. 1) among the Harleton samples should have produced scatter in the Th/U vs. 1/U plots (Figure 3.2. 1 and 3.2.2), but regul arity is observed instead. There is a special process, associated with the petrogenesis of ordinary chondrite. which produced the obser ved correlation. 60 Model II Here we retain the successful core of model I that Th/U variations are a reflection of varying proportions of apatite and merrillite. However, we drop the explicit assumption of the St. Severin phosphate Th and U concentrations, and allow [U] and [Th] to vary within phosphates and among the different Harleton samples. For model IL we assume instead that we should take "equilibrated" literally, a nd assume that in each sample the relative amounts of U (and Th separately) between the two phosphates at least approach an equilibrium partitioning ratio defined as Du = [U],P I [U] merr (and similarly forTh). We assume that the values for Du (27) and DTh(3.5) can be inferred by the Crozaz (1979) data for St. Severin. Thus, for model II we have: (I ) where the subscripts phos, ap, and m refer to total phosphate, apatite and merrillite, f; is the fraction of phosphate which is phase i (i.e., f.P + fm = I). Simlarly. the whole rock [U] concentration is: (2) where F is the dilution factor (gm rock I gm phosphate). but also F = [P]phos I [P] (3) 61 where [P] is the whole rock P concentration and [P]phos IS the concentration of P in phosphates. Accepting the D as known and constant. there are still four unknowns in the model: f,P' [U]ap- [Th]ap, and F. Strictly speaking [P]phos in equation (3) depends on f since the P contents of merrillite and apatite differ slightly. Our detailed calculations allow for this. but the effect is totally negligible and [P] phos can be regarded as a constant. Thus. F is completely determined by the measured P concentration. However, even with this simplification, the model remains underdetermined. Although the model doesn't make explicit predictions, there are no required specific correlations between P and Th/U. [U]. or [Th] as observed. If we make an additional assumption, we can u e the measured concentrations of U. Th. and P to solve for the unspecified model parameters. A reasonable assumption is that the Th/U in apatite is the same in all samples and the value lies in the range of I to 2.5 . With this additional assumption we can calculate the actual concentrations of U and Th in the phosphate phases. The additional assumptions in Model II build in the mixing trend from Figure 3.2.1; however. the fit parameters of Model II provide insight into the origin of the mixing trend. Figure 3.3.4 shows a well-defined correlation between the measured total rock U concentration and the amount of apatite calculated from Model II. It demonstrates that the apatite control over U concentration inferred from Model I is maintained in Model II. The failure to have uniform dilution factors does not affect this major conclusion. Figure 3.3.4 is shown for an assumed apatite ThlU = 2. but the observed trend is independent of what is assumed for apatite Th!U (see Fgure 3.3.5). What does change is the calculated 62 range of values for fw For lower values of apatite Th/U, lower apatite abundances are calculated. Figure 3.3.6 shows that there is a general inverse correlation between the U concentration in apatite calculated from Model II and the measured P concentration. Although an apatite (Th!U) must be assumed to calculate the model [Ulw the trend in Figure 3.3.6 is totally insensitive to the value of apatite (Th!U) chosen; only the [U]ap scale is affected, analogous to Figure 3.3.5. In other words. the assumed Th!U ratio of apatite determines the intercept of the trend in Figure 3.3.6, but does not influence the slope. The data point in Figure 3.3.6. which deviates the most from the trend. is the smallest (relict chondrule) sample H6-3 (see Table 3.2.1).This sample has an anomalously high P content. If a relict chondrule, the high P concentration is surprising. In general we have no explanation as to the anomalous location of this sample on Figure 3.3.6. H6-3 has the lowest Th/U, but plots close to the trend on Figure 3.2. 1. In Figure 3.3.6 the U concentration in apatite is in no way fixed when a P concentration is specified. The correlation in Figure 3.3.6 is not built into the assumptions of Model II. It instead represents a property of ordinary chondrite metamorphism. Figure 3.3.6 helps understand why the lack of uniform P contents and the lack of a correlation between P and Th/U (Figure 3.3.1) did not produce scatter on the Th!U vs. 1/U plots (Figures 3.2.1 and 3.2.2) instead of the observed regularity. The correlation shown in Figure 3.3.6 is probably best viewed as the response of the U and Th concentrations to P inhomogeneities. With an increase in P, the U concentration in apatite decreases. i.e.. a dilution effect. Figure 3.3.6 illustrates a 63 secondary control on the U concentrati ons, with apatite (and chlorine) abundances (Figure 3.3.4) being the primary control. 64 30 • 25 -- 20 - .c Q.. Q.. ~ ~ Q -.c ""' 15 • I '•• • ~ Q ~ ... 10 - ;:J • 5 - •• 0 0 0.1 0.2 0.3 0.4 0.5 0.6 model fraction of apatite Figure 3.3.4 Plot of measured whole rock U vs. fraction of apatite calculated from Model II. Th/U ratio in apatite is assumed equal to 2. 65 30 25 ,--, .D 0.. 0.. 20 IJ 0 ~ '-" ~ <..> 15 - ~6 8c /lll:l cP 0 6 IJ 1-o <1.) - 10 0 ..c ~ ~ 5 Be c 0 8 8 0 0 0 r '-- " ~ IJ 6 0 0 r 0 0 0.2 0.4 0.6 model fract ion of apat ite Figu re 3.3.5 Same plot as Figure 3.3.4, but showing calculated fraction of apatite for different [Th!U]apatire. Circles, squares and triangles correspond to [Th!U]apatite equal 2, 1.5 and I respectively. 66 14 - • 12 10 • 8 • ~ = .. Q, ........ • • ~ ........ • • 6 •• • 4 • 2 0 500 1000 1500 2000 2500 P, ppm Figure 3.3.6 Model II. Plot of calculated concentrations of U in apatite vs. measured phosphorus concentrations. (Th/U),P= 2 gives most reasonable values for Uw The two most deviant points (one in with high P, high [UJ.P and one off plot) are small relict chondrules (see Table 3.2.1 and Figure 3.3. 1). 67 3.4 Implications for ordinary chondrite petrogenesis Our data show that the large range of Th/U seen among ordinary chondrites can be found in one meteorite and that the bimodal distribution (see Introduction) of this ratio among meteorites is only apparent. The nature of the uranium carrier phases in ordinary chondrites is a long standing problem. In a detailed study of the uranium distribution of the St. Severin (LL6) chondrite with the use of fission track radiography, Jones and Burnett (1979) were able to locate reliably apatite grains down to about 1 micron in size. From these data only about 5% of the phosphates were apatite. Apatites greater than I micron accounted for 15% of the total meteorite U, and all phosphates accounted for only 24%. Greater than 60% of the U resided in unknown interstitial submicron phases. Jones and Burnett noted that submicron apatite was a plausible candidate for the missing U. This was supported by a lack of a Cl material balance if only the larger apatite grains were considered. Hagee et al. (1990) measured Th/U ratios of 2.60 and 3.63 for samples of the dark and light phase of St Severin. Equation (I) and the St. Severin data give apatite abundances of 25 and 13% for the dark and light phases, significantly larger than the observed 5% for the large apatite grains by Jones and Burnett. The samples of St. Severin used by Jones and Burnett and by Hagee et al. (1990) were not identical; nevertheless, the above calculation gives strong support to the interpretation that submicron apatite grains control the U distribution in St. Severin. The St. Severin analyses of Hagee et al. ( 1990) are included on Figure 3.2.2 and are consistent with the Harleton trend, also supporting apatite control for St. Severin. 68 Possibly contradictory evidence comes from leaching studies on non-magnetic fractions of equilibrated ordinary chondrites (including St. Severin) by Ebihara and Honda (1984). For St. Severin, 94% of P was extracted by EDTA and 6M HCl leaches; whereas only 57% of the U was leached, arguing that 43% of the U is not associated with phosphates (M. Ebihara, personal communication). However, the missing 6% of P could represent apatite grains separated with the magnetic fraction, or small grains sealed on the boundaries of binary grains and therefore difficult to leach. If all of the U is associated with phosphates, these small apatite grains might have much higher U concentrations than easily leachable larger phosphates, accounting for 43% of the total U. This could also reconcile the higher model apatite U concentrations (average for whole meteorite) in Figure 3.3.6, with the lower Uapa•i•e concentrations measured in large grains (Table 3.3.1 ). In unequilibrated ordinary chondrites, most of U located in chondrule glass, whereas most of P is in metal. In regions with low P and high U (i.e., chondrule glass) isolated from equilibration with big phosphates associated with metal, small U-rich phosphate grains will not grow during metamorphism, explaining the inferred correlation of the grain size and U concentrations. Finally, it is worth noting that, if the Th/U vs. 1/U systematics for equilibrated ordinary chondrites were dictated by four components (two phosphates, inert material and the 43% non-phosphate carrier), then it would be surprising to observe the correlations shown in Figures 3.2.1 and 3.2.2. Direct observations on unequilibrated ordinary chondrites (e.g., Yabuki and El Goresy 1986) and inferences from petrographic associations of metal and phosphates in equilibrated ordinary chondrites (e.g., Jones et al., 1980; Murrell and Burnett, 1983), especially an excess of coarse phosphate grains at the margins of metal in the metal-rich 69 Portales Valley chondrite (Ruzicka et al. 2000), indicate that, prior to metamorphism, P was associated with metal. Association with sulfide cannot be ruled out and P-rich sulfides are reported in CM chondrites (e.g., Bunch et al. 1979, Nazarov et al. 1997). However, the phosphate-sulfide preferential associations are probably best interpreted as PeS formation from a P-rich metal followed by (or simultaneous with) nucleation of a phosphate during metamorphism. The alloying of P in Fe-Ni metal is a predicted consequence of ca1culations of gas solid equilibrium under solar nebular conditions (e.g., Grossman and Olsen, 1974). Parent body conditions appear to have been sufficiently oxidizing that phosphates formed early, even in the most primitive chondrites (Yabukj and El Goresy 1986). Phosphate nucleation probably occurred primarily on metal grain boundaries, preserving the preferential association with large metal grains observed in equilibrated chondrites. Large (I 00 mkron) phosphate grains are rare in unequilibrated ordinary chondrites but common in equilibrated ordinary chondrites. Thus, continued growth on some phosphate nuclei at the expense of others over the duration of re-crystallization has produced relatively large phosphate grains which in turn produced a gm-scale heterogeneity in the P concentrations. We have documented that this is quite large in Harleton, but Harleton is likely anomalous in this respect. For an order of magnitude larger sample size, the data of Von Michaelis et al. (1969) indicate that equilibrated ordinary chondrite P contents are remarkably homogeneous. Both apatite and merrillite are observed in unequilibrated chondrites (Murrell and Burnett. 1983; Yabuki and El Goresy, 1986), but petrographic observations to date cannot dis6nguish which phase formed first. 70 Similarly, the host phases in which Cl entered into ordinary chondrites are poorly constrained. Besides apatite, other Cl-rich phases in ordinary chondrites are rarely observed. uch as lawrencite (FeCI 2 ) and halite (NaCI). Lawrencite was described in only one, unusually high-Cl. L6 chondrite Gomez (Sipiera et al., 1980). Halite has been found in ordinary chondrites of intermediate petrographic grade (Zolensky, 1999), and if it is indeed a hydrothermal mineral. it may have formed very early. If halite was a major source of Cl in ordinary chondrite parent bodies, its apparently gradual destruction during metamorphism may have produced a transfer of the Cl to apatite. The survival of halite in intermediate grade chondrites might indicate that the first formed phosphates were merrillite, followed, possibly relatively quickly, by apatite. However, limited by the chondritic P/Cl abundance ratio, conversion to apatite was only partial. This scenario is supported by the rarity of the observed apatite grains in the low petrographic grade ordinary chondrites (Y abuki and El Goresy, 1986). Regardless of details. everal authors have noted the great variability (up to a factor of 20) of chondritic Cl concentrations (e.g., Reed, 197 ! ) or of the apatite/merrillite ratio (e.g., Pellas and Storzer, 1975). It is plausible to regard this heterogeneity as representing the 'freezing-in'' of a heterogeneous Cl parent body distribution on a more uniform distribution of early nucleated merrillite grains. The coarsening of phosphates during metamorphism has not produced a more homogeneous di stribution of apatite. This i probably because the apatite/merri Ilite heterogeneity is only a property of the larger. more easily observed phosphates and that most apatites are really very fine-grained as discussed above. Based on our study it is unlikely that Unruh's ( 1982) high Th/U measurements (see Table 3.2.1) accidentally happened to sample three different parts of 71 one rock, a11 with "anomaleously" low Cl content. Unruh's samples are fairly small, -100 mg, so it is likely that Cl variations have to be on a much larger scale, with Unruh samples all coming from a low-Cl region. In unequilibrated chondrites, U and Th are concentrated in chondrule mesostasis (Murrell and Burnett, 1983) like other refractory lithophile elements. Phosphates in unequilibrated chondrites have much lower U and Th concentrations compared to those of phosphates in equilibrated chondrites (Crozaz et al., 1989; Murrell and Burnett, 1983). Thus, the migration of U and Th (and other lithophile elements) into phosphates occurs gradually over the epoch of chondritic metamorphic recrystallization. It is plausible to assume that the strong partitioning of U into apatite forces a heterogeneous U distribution, matching the frozen-in initial chlorine distribution. 3.5 An improved solar system Th/U As noted in the Introduction, cosmochronology models require precise knowledge of the average solar system Th!U. Moreover, a significant scatter in Th!U in CI chondrite analyses (Rocholl and Jochum, 1993) makes this traditional source of solar system abundances unrel iable. Relative abundances of other refractory lithophile elements are indistinguishable between CI and ordinary chondrites (Larimer and W asson. 1988). Consequently, ordinary chondrites are expected to provide a valid average solar system Th!U, although additional tests of the equality of refractory lithophile abundances between ordinary and Cl chondrites would be important. 72 Available high precision Th!U for ordinary chondrites prior to this study showed significant variations well outside of analytical errors. It was tempting to average these data in the hope of obtaining an improved solar system Th/lJ. However, in the absence of any understanding of the causes of the variations. one could not rule out that the average solar system Th/U was the minimum or the maximum of the range, as opposed to the average. Our study shows that the observed Th!U variations m equilibrated ordinary chondrites are produced by the redistribution of U and Th among the co-existing phosphate phases, coupled with fluctuations in the sampling of apatite grains in gramsized ordinary chondrite specimens. With sampling established as the cause of the observed variations, a sample-mass-weighted average of high precision ordinary chondrite Th/U should give an improved average solar system Th/U. This conclusion is supported by the data of Von Michaelis et al. (1969), which show that ordinary chondrite P abundances converge to precise values when large samples are analyzed. Thus, and 73 where g; is the mass of i sample and N is a number of the samples in the set with g;>0.5 gr. The calculated weighted Th/U of Harleton and weighted Th/U for other ordinary chondrites are listed in Table 3.5.1. ln this table we present weighted averages for OC data as well as arithmetic Th/U averages for comparison. Calculated crmean is given in the brackets. Calculation of weighted averages separately for Harleton and other ordinary chondrites is a test of our approach since the total mass of analyzed samples of Harleton (9.3 grams) and other meteorites (13 grams) is comparable. The OC weighed average is highly sensitive to one 1-gram sample of Glatton (Chen et al. 1993 and private communication). This sample is anomalous in two respects: first, this sample falls >4cr from the weighted average, and second, the high Th/U in this sample is due to enriched Th. Other (this work and Unruh. 1982) high Th/U whole rock data are mostly due to U depletion. Thus, the Chen et al. ( 1993) Glatton sample can be considered truly anomalous. In Table 3.5.1 we present data for Harleton (line I); other than Harleton OC literature data (line 2), and OC data omitting the anomalous Glatton analysis (lines 3 and 4). The striking result is that weighted averages of Th/U for Harleton and other OC (without Glatton sample) strongly converge to a value of 3.53. If, as it was discussed at the end of the previous section, strong partitioning of U into apatite forces a heterogeneous U distribution. matching the frozen-in initial chlorine distribution, then the whole rock analyses of unequilibrated chondrites should converge more readily to the average chondtritic value because of a more homogeneously (on a I gram scale) distribution of the chondrule mesostasis. Indeed the average of 5 available 74 analyses of lA chondrites (Bjurbole and Tennasilm, included m the Figure 3.2.2) is 3.55±0.10. It is necessary to note that almost all of the literature ID data of OC are for L-type chondrites. However, the relati ve abundances of refractory lithophile elements are the same in all ordinary chondritic groups and CI, thus the L chondrite bias in the Th!U data should not make a difference for solar system abundance calculations. The identification of U as a refractory lithophile element can be questioned. High Th/U ratios in Allende CAis (e.g., Chen and Tilton, 1975; Boynton. 1978) are usuall y interpreted as reflecting a lower condensation temperature for U. However, the general issue may be much more complex as suggested by a systematic study of CAls in Mighei (CM2) by MacPherson and Davis (1994), who found Th/U ratios ranging from I to greater than I 8. Depletions of moderately volatile elements with condensation temperatures significantly below that of U (e.g., Cr, Mn, Li, Na) in ordinary chondrites are not large (~20%; Palme et al., 1988), but the possibility of the ordinary chondrite Th/U being slightly higher (- I 0 %?) than the average solar system value cannot be ruled out. 75 Table 3.5.1 Calculated average Th/U for ordinary chondrites. Weighted average 1 Arithmetic average Harleton 3.53 (0.20) 4.12 (0.42) 0Chondrites 2, w/o Harleton 3.71 (0.24) 3.55 (0. 19) OC, wlo Harleton and Glatton 3 3.53 (0.14) 3.46 (0.13) 3.53 (0.10) 3.67 (0.20) Ordinary Chondrites 4 3.63 (0.37) 1 See equation (4) and explanation in the text for weighted average and 1crmean (in brackets). 2 Literature (Tat umoto et al. 1973; Unruh, 1982; Hagee et al. 1990; Chen et al. 1993; Shinot uka et al. 1995) and thi tudy data for H. Land LL 4-6, omitting Harleton. Multiple mea urements of the arne meteorite are treated as separate amples. Total 31 ample . of which one i H5, 5 are LL6 and the re t are L4-6. 3 Same a above, with I Glatton ample (Chen et al. 1993) omitted ( ee text for explanation). 4 Combined literature (Giatton omitted) and Harleton analyses from Unruh ( 1982) and this tudy. 5 Anders and Grevesse 1989 value for Solar System abundance , based on CI chondrite . 76 Chapter 4: Carbonaceous chondrites 4.1 Th/U variations in carbonaceous chondrites As discussed in the introduction, carbonaceous chondrites show a surprisingly big catter in Th/U ratios, mostly because of U variations ( ee Figure 1.1 ). Restricting our di cu sion only to high-precision i otope dilution (ID) literature data, Figure 4.1.1 plots Th/U versus 1/U for CI (Orgueil, lvuna), CM2 (Mighei, Murrey, Murchison, Nagoya and Yamato 74-662), CV3 (Allende) and C03 (Lance. Omans). The simplest interpretation of the positive linear trends in Th!U vs. 1/U is two-component mixing between high Th/U, low U and low Th/U, high U endmembers, strikingly analogous to what observed in ordinary chondrites (Figure 3.2.2). Samples within every group appear to yield linear trends, although the trend i le s defined for CM chondrites. Trends in the Figure 4.1 .1 eem to converge at the same value, possibly a common high-U endmember. Rocholl and Jochum ( 1989) consider apatite to be a possible endmember, because apatites are reported as trace phases in carbonaceous chondrites and because apatites in ordinary chondrites are enriched in Th and U, and tend to have Th/U ratios around unity. Objectives of the carbonaceou chondrite study are ( I) to test whether phosphates are the cause of the heterogeneou distributions of Th/U shown in Figure 4.1 .I, and (2) to determine other possible proce e for redistribution of actinide within parent body (such as aqueous alteration). 77 5.0 4.5 4.0 I 3.5 • 3.0 ::I • ..... .c 2.5 I- • • .Ao • •' j • ~ 2.0 1.5 1.0 • • CI • CM2 o C03 • CV3 • 0 .5 0.0 0.00 ~ ~ A&G 0.05 0.10 0.15 1 /U, ppb-1 Figure 4.1.1 Th/U vs. 1/U plot of the whole rock ID Th/U analyses for Carbonaceous Chondrites (data from Tatsumoto et al., 1976, Rocholl and Jochum, 1993 and J. Chen, personal communication). Different types of carbonaceous chondrites display linear trends with different slopes. Large variations in Th/U are apparent for CI and CY3 chondrites. Star is Anders and Greves e ( 1989) value. 78 4.2 CI chondrites The bulk compositions of CI chondrites are a very close match to the composition of solar photosphere (Anders and Ebihara, 1982), excluding a few volatile elements and Th/U (Chapters I and 3), hence, the elemental abundances in bulk CI chondrites are used as a reference composition for other types of solar system materials. In general, the other groups of carbonaceous chondrites are enriched in refractory lithophile elements relative to CI, both in absolute concentrations and in Si-normalized ratios, presumably because of the presence of refractory Ca-AI-rich inclusions (CAl). Petrographically, CI chondrites are constituted almost entirely of fine-grained, opaque matrix material, with virtually no chondrules (<<I%, see Brearley and Jones (1998) for review). A remarkable characteristic of CI chondrites is the fact that despite being the "most primitive" composition among solar system materials, CI chondrites are the most severely hydrothermally altered. Their matrices consist mostly of very fine-grained phyllosilicate minerals (60-50%) with imbedded magnetite crystals and a variety of accessory sulfides, carbonates and sulfates. Minor phases include Ca-phosphates (e.g., Endress and Bischoff, 1996). CI leaching experiments The only attempt to distinguish between different actinide carriers in CI-type chondrites was made by Chen et a!. ( 1993). In this work a number of stepwise leaching experiments were performed for the Orgueil sample. The Chen et al. ( 1993) results are included in the Table 4.2.1. These showed that there is a large degree of Th/U 79 fractionation in different leaches. The dilute acids (acetic and 0.1 M HCl) removed about 35% of total U, but only few percent of Th. The presence of these easily leachable sites with relatively high U concentration and low Th/U ratio could be the cause of the variations in Th/U, mostly due to U variations, in whole rock CI chondrites as shown in Figure 4.1.1 and Table 4.2.1. It is evident that a large fraction of U in Orgueil resides in sites which may have been produced by aqueous alteration. Such phases are most likely carbonates or sulfates. Other plausible U-enriched phases would be phosphates, which may or may not be a product of the hydrothermal alteration. In order to test whether high Th-U phosphates are the cause of the heterogeneous distribution of U and Th in Orgueil, we analyzed phosphorus concentrations in exactly the same leaches in which Th and U were previously determined by Chen et al. (1993). For TIMS analyses. Chen et al. (1993) extracted U and Th via anion exchange column, collecting and combining all elutes other than the U, Th fraction. There is no evidence of P being in the same elute as actinides (Kraus and Nelson, 1958). Phosphorus was analyzed as described in Chapter 2. The total phosphorus content for the Chen et al. (1993) sample (sum of all elutes) came to 1035 ppm. In CI chondrites, average phosphorus concentration is I 180 ppm (Anders and Greveesse, 1989). This confirms that there was very little or no P lost in Chen et al. (1993) chemistry. Table 4.2.1 shows that very little P is associated with easily leachable phases. enriched in U. Most of the P (87%) came out in the same stage (6N HCl) that is enriched in both U and Th. but with a high, not low, Th/U. This experiment shows that ThfU variations in CI chondrites are likely dictated by the presence of 2 major actinide carriers: 80 (1) easily leachable phases, rich in U and very little Th, most likely carbonates and sulfates, and (2) 6NHC1- leachable phase enriched in both actinides, with relatively high Th/U (-6), possibly phosphates, although the correlated release of actinides and P could also be coincidental. So, for CI chondrites, variations in whole rock Th/U ratios can be explained solely by the redistribution of actinides (mostly U) by the aqueous fluid into secondary mineral phases. The linear trend seen in Figure 4.4. 1 is produced by mixture of low- and high- Th/U phases. The observation of uranium redistribution during hydrothermal alteration is not surprising, because upon oxidation to higher valence states, this element becomes more mobile, staying in the aqueous fluids and depositing as alteration phases with fractionated actinides. 81 Table 4.2.1 U. Th and P analy es of Orgueil (CI). ·········--·····-·-········-····-·----·-----·--···-········-·-··-··-- ----··--··-····-······----·· ········-···--·------·-··---·--·-·-··---·--··--·---·--· 1 2 2 U.CJI Th % -- u. ppb ~ Th. ppb Th/U P, o/c 1.57 0.23 0.15 1.4 14.6 P. ppm Leaching exp. HAc 21 O. IM HCI 14 5 1.07 1.46 1.36 1.1 I 1.2 6M HCI 55 81 4. 16 23.43 5.63 87.2 903 7M HN03 7 10 0.57 2.95 5. 16 9.6 99 Re idue 3 3 0.21 0.77 3.57 0.8 7.9 Total 100 100 7.59 28.8 3.80 100 1035 Orgueil 3 8.2 28.6 3.45 4 Oroueil-1 0 12.1 30.4 2.5 1 Orgueil-24 8.0 29.2 3.65 4 8.5 30.3 3.56 Orgueil-536: 8.4 31.0 3.69 5 Oroueil-T 0 7.9 27.8 3.5 1 Orgueil-T 5 7.9 30.0 3.80 27.4 39. 1 1.43 36.0 38.0 Whole rock Oroueil-3 0 5 Oro-318/1-1 0 Org-3 18/1-2 1 5 ------- 1.06 ---·---------------- From Chen et al. 1993 ~ Concentrations in leaches (ppb) are in ng of element in the leach per total mass of the initial sample. Tat umoto et al. 1976 .~ J. Chen. personal communication ' Rocholl and Jochum. 1993 3 82 4.3 CM2 chondrites CM2 chondrites are more petrologicall y and chemically heterogeneous than CI. Whi le the C I consist almost enti rel y of fine-grained materiaL the matrix in CM occupies o nl y 70% by volume (Brearley and Jones. 1998) . Other constituents include chondrules (20%), refractory incl usions (5 %), isolated o livine grains and sulfides. Metal is rare in CM2 chondrites (0. 1% by volume), with Fe present as Fe-oxides and sulfides. Matrix material consists mostly of hydrous phases. which are believed to be products of hydrothermal al teration of a fine-grained anhydrous precursor (see Brearley and Jones, 1998 for review). Chondru les and CAis also have been altered to a d iffe rent degree by hydrous flu ids in different CM c hondrites (e.g .. Clayton a nd Mayeda, 1989). An alogous to Cis, aqueous alteration of CM chondri tes could produce redistribution of U by hydrotherma l fluids and fractionate Th and U if conditions are sufficientl y ox idizing. Carbonates and sul fates occur as major phases in CM matrices. as wel l as secondary alterati on phases in CAis (Brearley and Jones, 1998. MacPherson and Davis. 1990). Clfre e apatite was found in the matri x of the Cochabamba CM chondrite (Muller et al., 1979). In add ition to aqueous al teration. the heteroge neous distribution and mineralogical and chemjcal diversity of refrac tory incl usions in CM2 chondrites can affect the who le rock Th/U rati o. Refractory inclusio ns appear to be the fi rst materials formed in the solar nebul a. but have very comple x evolutionary histories. Thermodynamic calcul ati o ns of the condensati on temperatures of U and Th in the solar nebu la show that Th is slightl y more re frac tory than U. D ifferences between 50% condensatio n temperatures (K ) of Th and U at differen t pressures and C/0 ratios are from I 0 to 100°K U (Kornack i and Fegley, 1986, 83 Ladde rs and Fegley. 1993). So refractory incl usions in carbonaceous chondrites might not reflect the solar Th/ ratio. A thorough ion probe study of bulk major a nd trace element composition of the refractory incl usions from Mighei (CM2) cho ndrite was done by MacPherson and Davis ( 1994). Figure 4.3 .1 shows who le rock ID analyses for CM2 chondrites as well as in-situ analy es of the refractory inclusion fro m Mighei. Refractory inclusio ns exhibit a wide range o f Th/U (from I to 20). S urprising ly. in spite of the big error bars on the io n probe data points. refractories also seem to fo llow a positive trend in Th/U vs. 1/U coordinates. The objectives of CM-chondrite study are si milar to C)-chondrites: ( I ) T o test whethe r phosphates are the cause of the heteroge neous di stributio n of Th/U. and (2) to determine other possible processes for redi stributi on of actinides within parent body (such as aqueous alteration). 84 + CAl 14 o Whole rock 12 • •• • 8 • •• •• • 10 .••..... 6 4 0 0 -~ 0 0 cw'P 2 • •••• o* 0 0.02 0.04 0.06 0.08 0.1 0.12 1 / U ppb _, Figure 4.3.1 Th/U vs. 1/U p lot for CM2 chondrites. Whole rock analyses same as in Figure 4.1.1. CAl- ion probe data from MacPherson and Davis ( 1994). 85 CM leaching experiments The Murchi son CM2 meteorite was chosen for thi tudy. We have performed leaching ex peri me nts in order to distinguish between different actinide carriers. To gai n a qua lita ti ve unde rstanding of hig h and low U compone nts. we have tri ed to distinguish U and Th associated wi th carbona tes. sul fa tes. phosphates. matri x material (ph ylosilicates) and with chondrule plus CAl mate rial. Leaching expe rime nt are c halle ng ing to interpret becau e of the te ndency of di solved e leme nts to precipitate from the solution and readsorb on the residues. ForTh thi s te nde ncy is especiall y well known. Without isotope diluti o n, measured Th concentrations might be significantl y lower than the real conce ntratio n in the leached phase. So one should be careful in choosi ng the reagents and conditi ons of the experiment (such as te mperature and du ratio n). Three pieces of the arne fragment of Murchison (Me2683) were surf ace cleaned wi th a d ia mo nd de ntal drill and rin ed in spectroscopic grade methano l. Sample Mu I was spiked and di ssolved fo llowing the technique described in Chapter 2.4. Sampl e Mu2 was: I) subjected to ultrasoni c di gesti o n for 15 minutes in 3 ml 1.6M HN03 ; 2) sample was centrifuged. leachate separated and spiked: 3) 3 miH ~ O wa added. centrifuged. and the leach combined with the one eparated in the step 2. Step 3 wa perfo rmed one more time. The combined solution is Mu2-leac h in the Table 4.3.1. The residue (sample M u2re idue) after leaching was spi ked and dissolved followi ng Chapter 2.4. The results of the Mu2 leach were not definite, so a more e laborate experime nt was designed to distinguish be tween di ffe rent easil y soluble phases. 86 Sample separation for Mu3 is illustrated in the schematic diagram Figure 4.3.2. Mu3 was carefull y (not to break c hondrules and CAls) crushed in a tainless steel pestle. and separated into two fract ions - fine (MuF) and coarse (MuC) . Separation was performed usi ng 270 mesh (53 ).lm) Ny lon screen under methano l. The goal was to have the MuF fraction be a re lativel y pure matrix material. Consequently. the MuC fraction does have matrix lumps. in addition to expected constituents of the coarse fraction. The fine (<53 ).lm) fraction was di vided into two splits MuFI and MuF2. MuFI was spiked and di ssolved. The coarse (>53 ).lm) fracti on was divided into two splits: MuCsp I and MuCsp2. Split MuCsp I was put aside for the future study. Split MuCsp2 was powdered and split in two: MuCsp2a and MuCsp2b. Sample MuCsp2b was spiked and di ssolved. Samples MuF2 and MuCsp2a underwent stepwise leac hing. Paral le l to the amp les. blank solutions were carried through the same step . In the Figure 4.3 .2. th ree leaching steps are indicated: I. HAc. Carbonates and sulfates di ssolve readil y in acetic acid (HAc). The duration of the HAc leaching step was short to minimize time fo r Th precipitation. For preparation of the samples MuF2- I and MuCsp2a- 1. one ml IOo/c HAc was added to the samples. allowed to sit for 5 minutes at room temperature. periodical ly hand shake n. and then di gested ultrasoni call y for 5 minutes. For the separation of the leach from the residue. samples were centrifuged using 2 ml Nalgene acid-cleaned centrifuge via l with a 0.2 mm Nylon filter insert (Figure 4.3.3). All fo llowing residue-leach separations were perfo rmed using these centri fuge vial . So lution was removed from the lower part of the vial. and Th-U spike was added . An additi onal 0.4 ml of HAc was added ri ght o n top of 87 the filter and centrifuged. This solution was combined with the spiked portion . Two additiona l 0.4 and 0.6 ml H, O rinses were made and the rinse solution combined with previous solutions. 2. NH ,-EDTA . Ebihara and Honda ( 1984) have demonstrated that ammoniaca l EDT A (preparation of the reagent described in the Secti on 2.4) efficiently attacks phosphates. Additional experiments on the di ssolution of powdered (-200 ).lm) apatite grains in EDTA were performed to show that the dissolution works and to determine a sufficient duration for the phosphate dissolution. We performed test e xperiments to determine whether Th stays in EDT A solution without adsorbtion on the solid residue. 23 <>rh spike was added to scrap powdered material from Murchison and H~-EDTA for 7 leached with hours at room temperature. After centrifugation. the sample powder was rinsed with fresh EDT A 3 times . The leach and rinse solutions were combined for ana lysis. The procedure above was repeated 2 more times. Finally. the solid residue was dissolved and ana lyzed along wi th the leaching solutions for the presence of 23<>rh. The results of this experime nt showed that Th complexes. formed upon reacti o n with NH~-EDTA. keep Th in solution. Approximately 96lk of 23<Th from the spiked a mount came out with the first EDTA leach, and an additional 4lk was in the followi ng EDTA rinse. The following leaching steps as well as the di ssolved residue did not carry detectable amounts of 23<Th . For preparation of the samples MuF2-2 and MuCsp2a-2. 0.6 ml of ammoni acal EDTA solution was added to the residue in the centrifuge vial. These were left for reaction for 7 hours at room temperature. The leach was centrifuged. separated, and 88 spiked. Additio nal rinses were made with one additional ml of N H.~-EDTA and one ml of Hp. The rinse solutio ns were combined and added to the original leach. 3. NH 1-EDTA (re-extract). Preparation of the samples MuF2-3 and MuCsp2a-3 was as in step 2. 4. Dissolution of the residues MuF2-4 and MuCsp2a-4. Inserts from the centrifuge vials, carrying residua l samples on top the filter, were removed and placed upside down into crew-cap Teflon beakers with 1.5 ml 0.5N HCI. Ultrasonic digestion for 15 minutes was performed to separate sample from the fi lter. The filter was removed and cleaned an additional two times in the ultrasonic bath in HCI. The HCI solutions and sample residues were combined. spiked. and the excess HCI evaporated. Mixture of concentrated HF. H 0 , and HCIO.~ were added for dissolutio n as described in Chapter 2.4. Results Anal ytical results for U . Th and P from Murchison are listed in Table 4.3.1. For Th and U the analytical blank was of the order of Io- J-1 g. which is significantl y lower than most of the samples a nalyzed. A blank space for U. Th or P in Table 4.3. 1 means concentrations below blank levels. The Th/U ratio of the whole rock shows large variati ons between 2.85 and 4.5 . It IS evident. that unlike C I and OC whole rock analyses, Th/U variations in CM2 chondrites are mostly due to Th variations. The whole rock sample Mu I has the hi ghest U. Th and Th/U. This suggests that this sample mi ght contain excess of CAl materi al. 89 Leaching of the whole rock sample Mu2 shows that about half of the U, Th and most of the P are in 1.6 M HN0 3-leachable phases. These incl ude carbonates. sulfates. sulfides. and phosphates. However. some of the cho ndrule mesostasis and si licate material may have been leached as well. making the results of thi s experiment somewhat ambiguous. Th/U ratio in the 1.6 M H 0 3 is somewhat lower than the who le roc k ratio. suggesting the presence of more soluble U-rich phases. Unlike Mu I. total U and Th concentrations in Mu2 are not elevated. probably indicatin g no excess of CAl material in the sample. The amounts of U and Th leached in Mu2 compare well with those from the 6M HCI leach (sample Murc hison#6 in Table 4.3. 1) used by Chen et al.. 1993. 60% of U extracted with 1.6 MHNO , (63 % - with HCI). 53% of Th (40% - with HCI). The 6M HCI will attack the same phases as nitric acid. But Th stays better in nitric solution than in HCI. The second leaching step of Chen eta!. ( 1993) was 7MHNO 3 which extracted a signifi cant part of the U (22% of total) and approximately half (47%) of the Th . The 7M HN03 is concentrated enough to start dissolving more resistant si licate phases. so detailed interpretation of the 7M HN0 3 -leac h data is not possible. The U. Th analyses of the fine (matri x) separate (MuF I) and total for matri x sample MuF2 are consiste nt. The P contents differ by 50%. The Th/U rati o for both is low. 2.57 and 1.42 respectively. A maj or result is that the matri x U. Th and P conte nts are signifi cant ly lower than the who le rock concentrations. Whereas there are big reservoirs for U and Th in CM chondrites (chondrul es a nd CAls). most of the phosphorus might 90 have been expected to be in the matrix because phosphates were observed in CM matrix (Brearley and Jones, 1998). Results of the stepwise leaching of the matri x materi al are also shown in Table 4.3.1 . The HAc leach. targeted to the carbonates and sulfates. removed a significant portion of U (42%) and virtual ly no Th (0.7%) or P, suggesting a mobile U-rich component. But on a total sample basis. th is component accounts only for 10% of total U content. Chen et at. ( 1993) found no U associated with 0.1 N HCI. which should have dissolved carbonates. Of course. we cannot rule out a possible particular contaminatio n of small. low in U samples prior to chemistry. For example, to produce the amount of U "contamination" extracted in our HAc-leac h (0.24ng). the sample needed to pick up 25 f..lg of dust with I 0 ppm U concentration or ten million of I f..lm size du st particles. This seems large . It may be th at our sample and and that of Che et at. ( 1993) are different. The first EDTA leaching step. which preferentially dissolves phosphates, extracted -3 % of Th, 0.8% of U. and 3.6 % of phosphorus. The EDTA re-extract showed that no leachable phosphate is left in the sample. Most of the Th and P. and half of U. were left in the residue. This result provides strong evidence for the interpretation from MuFI analyses: Pis nor concentrated in phosphates. One of the explanations is that Pis ac tua ll y assoc iated with sulfides, which did not break during the separation of the fine and coarse fractions. Figure 4.3.3 shows such a Ni-Fe sulfide grai n. ri ch in P. Cr and Prich sulfide phases are recognized in CM chondrites (Bunch et at.. 1979. Nazarov et at.. 1997). Recent TEM studies suggest th at these phases most likely are mi xture of two phases. one of whi ch is likely to be a phosphide (Devouard and Buseck. 1997). 91 The total P differ by 50% in Mufl and MuF2, indicating that P i concentrated in relati vely large. possibly sulfide, parti cles. thus difficult to sample repre entatively. Assuming our matri x separation was successful. one can predict the concentrations of U and Th in the coarse fraction. In CM2 chondrites, matrix material constitutes 70% by volume (Brearley and Jones, 1998). Thus, from the samples MuF I and MuF2 it fo llows that matrix can account only for 5 % of total U and Th and for 30- SOCK of P. Then. the estimated Th, U and P concentration of the non-matri x material (including chondrules and CAis) are 105 ppb. 25 ppb and 2300 ppm respectively. Surprisingly, the an alytica l data for both the coarse fractions of Murchison (sample MuCsp2b and the total of sample MuCsp2a) re vea led fac tors of 3-5 higher actinide concentrations than was expected! Phosphorus concentrations are within the expected range . These actinide concentrations are way too high to be explained by particulate sample contamination. Very low concentrations in fine fractions show that the overall sample was clean. The o nl y plausible explanation is presence of an anomalously big CAl fraction in the sampled rock fragment. High U and Th concentrations in both coarse fractions MuCsp2a and MuC p2b suggest possibly just o ne big CAL which broke on further crushing after sieving. To produce the observed Th and U co ncentrations -I 0 mm 3 CAl is required to have on the orde r of I ppm U and few ppm Th. MacPherson and Dav is ( 1994) analyzed man y CAl from Mighei (CM2) with U and Th concentrations this high . The split of the coarse fraction (sample MuCsp I ) was in vestigated in order to find pieces of broken CAL The I I biggest chunks ( -0.5 mm diameter) were mounted 92 into the e poxy and polished. X-ray mapping for Ca and AI di d not reveal CAl materi al amo ng tho e fragme nts. However. in a doubl y polished thin ection. made from an adjacent piece to the initial roc k a mple . a big - 7mm d iamete r CAl was observed. Backscattered secondary electron image as we ll as X-ray maps of CAl inclusion fro m our Murchi on sample are presented in Figure 4.3.4. No phosphates we re o bserved within the inclusio n based on a P X-ray map. The petrography of this CA l inclusion was not studied in deta il, since onl y the pre ence of uch bi g CAl inclusions is important fo r this study . In the stepwise leachin g of the coarse fractio n of Murchi on. the HAc step ex tracted phases carrying 4 ppb of · and unusually hi g h (30 ppb) conce ntrati ons of Th . No phosphorus was assoc iated w ith this step. So it is likel y that these HAc-leac habl e phases are e ither in-situ a lterati o n products of high Th/U CA I phases (according to MacPhe rson and Davis. 1993. carbonate-barring CAis have wide range of Th/U rati os). or re idua l alterati on products after removal of more mobi le U by aqueous fl uids. An amount of phosphorus (So/c o f total in the coarse fracti on) comparable to the matri x fracti on was extracted by EDT A with very little actinide content (0.1 4 9c: o f U and 0.8% of Th). supporting the conclusion from the m atri x leac hing o f the overall low phosphate co ntent in Murc hison. 93 Murchison sample l\1u3 Gentle grinding (not to break inclusions) spike. dissolbe. analyze saved MuCsp2b Spi ked. d issolve, analyze leach Spike, ana lyze leach Spike, ana lyze leach Spike. ana ly ze Figure 4.3.2 Schemati c diagram of the sample preparation and leaching steps for the M urchison sample Mu3. Both samples MuF2 and MuCsp2a have undergone leac hing steps I through 4. 94 Snap cap Insert ( I ml ) with 0.2 )l m filter Figure 4.3.3 Illustration of the 2 ml centrifuge vial used for separation of the leaching solutio n from the residue. ll") 0\ Weight, g u, ppb4 Th, pph 4 Th/U P, % P, ppm ...................................................................................................................................................................................................................... ................................ Th, % 4.52 '''''•''''''' 65.03 4.02 ......... ,,, 14.37 51.1 3.40 U,% 12.7 37. 1 3.9 1 Leachate Table 4.3. 1 Murchison meteorite. Who le rock and leaching ex pe riments analytical results. Sample Whole rock 10.9 41.1 3.26 Murchi son# 11 Murc hison#2 1 Leaching exp-ts ! 22 12 I 0.2 i i 0 .1 10.5 39.5 2.85 100 i 0 .2 12. 1 37 0 .2523 Murchison#3 1 1 2.9 13 1 Murchison#4 1 j 0.181 3.67 Mu l Murchi son2 39.3 1925 I0.7 8 1.9 I 0 .2 15 2.84 Murc h(S IMS / 20.65 3.68 7.27 39 53.05 10.6 60.05 425 i l .6MHNO, 18. 1 Murc h(TIMS)' Mu2-leach 3.78 2350 18.28 100 4.83 3.22 548 38.93 100 12. 10 2.57 100 ............................................................................................................................ ' 46.95 ............... 39.95 0.574 6.9g5 0.035 4.95 0.023 ().{)2 3.60 0.06 13 . .......................................................................,....................................................... 1.5 12 0. 146 - 2.7 14 0.67 0.029 0.049 100 4 1.6 2.82 - 100 100 resid ue total 0 .0325 HAc 0.8 0.95 none Mu2-residue MuFI MuF2- I EDTA - . Muf2 -2 EDTA '"" MuF2-3 34H 36 1 96.38 100 2.36 1.42 4.952 5. 182 2.()97 3.639 95.6 100 57.6 100 residue 0 . 158 MuF2-4 total \0 0\ U, ppb 4 T h, ppb~ 7.05 T h/U Table 4.3.1 Cont. Th, % 29.2 P,ppm U, % 4. 14 P, % Leachate 10.88 Weight, g 6.07 Sample HAc 168 MuCsp2a- l 5. 17 2 1.9 0 .1 4 2. 13 0.004 3082 0.09 94.83 0.79 3.68 0. 14 235.6 EDTA 63.97 0.05 100 0 .001 0.008 100 10.33 0.005 0.005 39.46 39.46 0 .0005 0.003 417 .7 2.29 3.80 3.79 0. 1 0.68 3. 13 1.70 87.70 100 0.63 93.79 3.94 MuCsp2a-2 residue 268.6 0.04 100 10.34 15.68 8.04 I I 13.1 -I 2730 3250 68.20 water 99.9 100 6.78 18 .70 3. 12 ........ 100 O. IM HCI 100 40 2.3 1 5.31 100 100 EDTA o.2228 residue 63 47 1.69 none 6MHC I 2 1.5 13.4 10 .1884 j 7M HN03 15.6 ········!····· ................... '"""""""' ... ................... ............... .. I residue ···r······························· ·········································································· ................. ·································· .............................. 1 I MuCsp2a-3 MuCsp2a-4 total MuCsp2b Murc hi son#5 1 tota l Murchison#6 1 i I 100 100 10.78 39.69 3.66 total 1 1 Chen et al. , 1993. TIMS analyses . 2 Rocho ll and Jochum, 1993. 10-S IMS analyses. Prec is ion fo r Th/U is 3'1o. ·' Rocho ll and Jochum, 1993. ID-SJMS and TTMS ana lyses of the same a liquot. Precisio n for Th/U is 3% and 0 .3 % respecti ve ly. ~Concentrations in leaches (ppb) are in ng o f element in the leach per total mass of the initia l sample. 97 :! llfJ. ID Figure 4.3.4 Ni-Fe-silfide grain from Murchi on a) BSE SEM image; X-ray map for b) Ni, c) P, and d) Ca of the same sulfide grain. Ca-enrichment on the edge of the grain is likely oxidation. probably beginning of theCa-phosphate formation. 98 Figure 4.3.5 CAl inclusion from Murchison meteorite. a) SEM BSE image of the inclu ion b), c), and d) - X-ray map for AI , Ca and Ti correspondingly. Width of all images is 7 mm. 99 4.4 Interpretations From our size separation and leaching data it is evident that the distribution of Th/U in CM2 chondrites cannot be described by mixture of only 2 high- and low-U components. Figure 4.4.1 combines literature and whole rock data from this study, including the matrix and coar e fractions. If two-component mixing exists, we would expect data from the matrix separates and the coarse fraction to fall on the same linear trend. They clearly do not. Presumably, whole rock compositions can be reconstructed by mixing matrix with an appropriate amount of coarse fraction to produce data points laying on a pseudo-mixing line. However, when the complete whole rock data set is considered, there is very little correlation observed. Unlike the CM . the Th/U data in ordinary and CI chondrites can be described by pseudo-mixing of only two component . There is a mixing of two phosphates with an inert material in ordinary chondrite , and redistribution of uranium into carbonates and sulfates in CI chondrites. Figure 4.4.1 shows that, in fact, there are more than 2 components involved in producing vari ations in Th/U. Two of the components are chondrules and CAis, which are the biggest reservoir of actinides. Both chondrules and CAis are combined in our coarse fraction, although in fact, they represent two different re ervoirs. As shown in Figure 4.3.1. CAl materials have a wide range of Th/U . Becau e of this variability. CAis cannot be considered as one single component. A linear trend in Th/U vs. 1/U coordinates for CAis seems to hold. This may be a coincidence, but there are reasonable interpretation of that trend. The ion probe study of Murchison perovskites (Ireland et al.. 1990) shows that thi highly actinide enriched phase has Th/U well below 100 unity. Such perovsldte-bering CAis can represent a low-Th!U endmember. Our data shows presence of high-Th!U, HAc-leachable phases (sampleMuC p2a-l , Table 4.3.1 ). likely carbonate. This carbonate material could be produced either by the in-situ aqueous alteration of the high-Th!U material (complementary to perovsldte?) or be a residue after removal of more mobile U by aqueous fluids. Carbonate is a common secondary mineral phase in CM CAis (Brearley and Jones, 1998). Variation in the amount of high-Th/U carbonates in CAl materials will produce the mixing trend een in Figure 4.3.1. If the high Th/U ratio in carbonates i a re ult of preferential leaching of more mobile uranium, then the hydrated matrix material should show a relatively high U content, which it does not. This work shows that the Th/U ratio in matrix is indeed relatively low (-2), and the Th/U ratio of the HAc-leachable matrix phases are below unity. But the matrix material accounts for only a very small fraction of the total rock uranium (-I 0% ). Chondrules in ordinary, CV3 and C03 chondrites show small (1-2x CI composition) enrichments in refractory lithophile elements (Grossman et al.. 1988). Specifically. the data of Murrell and Burnett (1983) and Chen and Tilton (1975) for ordinary and CV3 chondrites respectively, show that U and Th are similarly enriched in chondrules with respect to CI. By analogy, it is very likely that CM chondrules are also ignificant reservoirs of Th and U. Contrary of the suggestion of Rocholl and Jochum (1993), our data shows that phosphates play only a minor role as actinide carrier phases in both CI and CM chondrites. The fine-grained matrix can be regarded as another component. However, there is fractionation of Th!U within matrix material. The Hac-leachable phase has very low 101 Th!U, compared to the bulk matrix analyses. The matrix contains much lower concentrations of actinides compared to CI materials. Matrices in carbonaceous chondrite are often regarded as primitive. but our matrix data in Table 4.3.1 and Figure 4.4.1 clearly shows that the matrix is strongly acti nide and phosphorus depleted. The Th!U linear correlation in Figure 4. 1.1 for whole rock CV3 chondrites is as good as that for CI chondrites. The amount of CAl material in CVs is twice as much as in CMs (B rearley and Jones, 1998). So, the CAl input into the Th!U whole rock variations has to be significant. To produce the CV -trend in Figure 4.1.1, CAl materials have to be mixed with a low-U component (matrix?). The amount of Th and U data for CAis of CV3 chondrites is not overwhelming. Whole rock CV3 and i otope dilution TIMS analyses of CAis from Allende (CV3) meteorite are plotted in Figure 4.4.2. The only available analyses of the CAl and chondrules are ID TIMS from Tatsumoto et al. (1976) and Chen and Tilton (1975). CAl Th!U ratio range between 2.5 and II , whereas chondrule data from Chen and Tilton (1975) are cattered in the area between CAl and whole rock analyses. From the data plotted in Figure 4.4.2 it is not obviou that mixtures of all the components should produce a linear correlation. A possible problem with the Chen and Tilton ( 1975) data is that these authors used HCI leaching of their samples as a cleaning procedure. Potentiall y, such leaching could remove microscopic carbonates and phosphates before analysis. producing disturbed actinide data. The HCI-leached CAis and chondrules are filled symbols in Figure 4.4.2. Omitting those leached samples, mixture of the chondrule Th and U analyses average out closer to the whole rock data. 102 Figure 4.4.3 demonstrates one possible way to produce linear whole rock correlation in CV chondrites. As shown in Figure 4.4.3, mixture of a primitive (nebular) CI-Iike material (" A&G") and 5% of CAl component with variable Th/U will produce a whole rock trend seen in Figure 4.1.1. Uranium data for chondrules and, possibly, matrix are higher than primitive CI composition, which argues that these components were "contaminated" with CAis. Unlike CM chondrites, phosphates are more commonly observed in CV chondrites. Our leaching experiments show that in CMs very little phosphorus is in phosphate phases, but Ebihara and Honda ( 1987) have found that -60% of P in Allende (CV3) is associated with phosphates, so their role in the CV Th/U distribution can not be ruled out. 103 5 _I • 4.5 4 0 3.5 3 ~ ~ 0 • Whole rock o coarse ·; A matrix •• • • 2.5 2 1.5 1 0.5 0 0 0.1 0.2 1/U, ppb"1 0.3 0.4 Figure 4.4.1 Combined literature and this study data for CM2 chondrites meteorite from Table 4.3. 1. Matrix and coarse fractions as well as leaching experiment data are "total" analyse . 104 12 10 8 ~ ~ 6 6 • •• 4 • 6 <> <> • <15'~ 0 X 0 2 0 0.00 X 00 X X 0.02 0.04 0.06 0.08 0.10 0. 12 1/U ppb-1 Figure 4.4.2 Whole rock (circle ), CAl (triangles), chondrules (diamonds) and matrix (cro ses) analyses in Allende (CV3) meteorite. Whole rock analyses same as in 4.1.1. Data from Tatsumoto et al. ( 1976) and Chen and Tilton, ( 1975). Filled labels are samples leached with HCl (see text for detail ). 105 15 13 CAl 11 ::::::1 9 ...... 15% CAl .&: ~ 7 - 5 5% CAl / 6. A&G / / 6. 3 ~ cP8 ° 0 / / / 0 0 0 1 0 0.05 0.1 1/U, ppb -1 Figure 4.4.3 Illustration of the mixing of the primitive nebular material (A&G, Anders and Grevesse, 1989) with CAis of variable Th/U (Tatsumoto et al. 1976) to produce whole rock CV3 variations. Whole rock data as in Figure 4. 1.1. The line through the whole rock CV3 data is a calculated trend with 5% of CAl mixed with A&G component. For comparison, trend for 15% CAl admixture is also shown. 106 Chapter 5: Implications to cosmochronology A it was briefly discussed in the Introduction. U and Th solar system abundances are especiall y valuable for determining the age of r-process nucleosynthesis. Cosmochronological time is usually divided into three periods: J) T, a period of active nucleosynthesis, during which the production of a nuclear species (P) occurs parallel to the free decay; 2) ~ . a free-decay period before the formation of the solar system, during which Pi=O, and 3) 4.6 Gyr of the solar system existence, from T +~ until today, during which free decay occur within solar system (e.g., Podo ek and Swindle, 1988). Constraints on~ come from short-lived isotopes, such as 244Pu. We assume that ~<<-r235 (where -r235 is the lifetime of 235U. the shortest-living of nuclides considered), so that ~=0 is a good approximation for for the models considered. For every nuclear species with a decay constant A,., the abundance N, at time T is defined as a solution of the growth equation: N = fr P.(t)e -1.. ,rr-t )dt I Jo I ( 1) There are two unknown in the equation above: production "history," or change in the production rate with time, and T . Instead of dealing with absolute abundances and absolute production rates, it is only possible to consider two ratios of two species, such as 235 UF-18U and 2.12Th/238U. Thus in general we have two equations with two unknowns, although Pi(t) can have several parameters. There are various simple model of nucleosynthetic production histories (see Figure 5. 1). such as ( I ) con tant production rates until solar ystem matter is isolated 107 from further nucleosynthetic input, (2) production exponentially decreasing with time (e.g., Fowler, 1977), and (3) initial rapid increase in production, followed by the smooth decrease (for example, Clayton 1988). These models have received substantial mathematical treatment (Schramm, 1973, Clayton, 1984, Fowler 1977, 1987, Clayton, 1987), and a number of galactic ages (T +4.6 Gyr) were determined, ranging from 8.7 to 12.8 billion years relative to the present. All the models are limited by how precisely we know the two abundance ratios and two relative production rates of 235U, 238U and 232Th. The solar uranium isotopic composition 235 UP 8U=lll37.88 is precisely defined (e.g., Chen and Wasserburg, 1981). In this work we have improved the knowledge of the solar system Th!U=3.53±0.1 (weight ratio), which translates to 232Th/~38U:::3.64±0.1 atomic ratio. The relative production rates are known from a combination of the decay schemes and nuclear physics theoretical predictions. As was recognized by B 2 FH, P232/P238 and P 235/P238 are unique among nucleosynthetic production rates, being accurately determined simply by counting the alpha decay progenitors ending at masses 232 and 238 based on the well known decay properties of nuclei near the line of ~-stability between A= 230260. The uncertainties on relative production ratios come mostly from the consideration of ~-delayed neutron emission. Although by astrophysical standards these uncertainties are small, for present purposes they are big. The quoted values for U, Th relative production rates vary by -25%, and uncertanties on those values are up to 30%. During the 1980s, it was proposed that, during the long chain of ~-decays connecting the r-process path and the line of ~-stability, spontaneous fission from nuclear excited states following beta decay (" beta-delayed fission" ) would significantly change 108 2 the B FH progenjror production rate model. However, in the last 10 years, estimates of the importance of beta-delayed fission have steadily dropped (Meyer et al., 1989; Hall and Hoffman, 1992) to the point where it is acceptable to neglect this effect (D. Hoffman, private communication). Table 5.1 lists literature production rate ratios of chronometer pairs (modified from Thielemann et al. 1993). llipparcos astrometry satelJjte observations have dramaticaJly improved knowledge of our Galaxy. Interesting results come from the re-evaluating distances, hence ages of globular clusters. The globular clusters of the galactic halo are the oldest datable objects in our galaxy, thus setting a lower limit to the galactic age. Clusters as old as I 5 billion years (e.g., globular cluster M92, Carretta et al., 2000) are observed. Differing from the starting point of other researchers, we will accept the estimates of the minimum galactic age based on the globular clusters. So the age of the galaxy is not a parameter. On the other hand, we will permit maximum production to occur significantly more recent than 15 Gyr, corresponding to the slow infall into the galactic disk. The uncertainties of the estimates of the relative production rates is the major limiting factor for all models. Consequently, our approach is to calculate the production rate ratios (P/Pj), defined by the galactic age, nuclide abundances, and time-dependence of the production rate evolution assumed in the model. Then the calculated production ratios can be compared to the estimates from Table 5.1. Let us consider 3 relatively simple scenarios, illustrated in Figure 5.1 , for the evolution of the galactic production rates. 109 J ). In case of the uniform production (Figure 5. J a), P;(t)=constant, equation (I) becomes (2) Then, for every pair of nuclides i and j , N; _ P; /y * (l-e-1vT) M - Pj Ai ( 1- e-~.;r ) (3) For the galactic ages (T+4.6 Gyr) 13-15 Gyr, and Th/Utoday=3.55-3.75, the calculated range of possible solutions of the equation (3) for relative production rates is illustrated in the Figure 5.2. It is evident that calculated production ratios for the given time interval T are systematically off the production estimates (circles in the Figure 5.2 are from Table 5.1 ). Only for the galactic ages as young as 13 Gyr, the production ratios calculated in this model overlap with the estimated theoretical values of P232/P238 and higher than P 235/P238 • The height of the bar is determined by allowing today's solar system Th/U to vary between 3.55 and 3.75, witch is in the limits of lcr uncertainty on our Th/U. For 2cr, the overlap will be greater, allowing 14 Gyr calculations to also overlap with the estimates. Most previous nucleocosmochronology models indicated galactic ages significantly lower. Indeed, according to Figure 5.2, lower ages are more compatible with estimates of the production rates, would be consistent with the beginning of the r-process production in the disk long after the formation of the halo globular clusters. 2). In case of exponential galactic synthesis (Figure 5.1.b), it is assumed that the rate of the star formation and the rate of nucleosynthesis are exponential functions of the time (Fowler, 1977). In this case for every nuclide i with A; decay constant and initial production P;0 • 110 dNi -ri T 'l . - =Pioe -NNi dt ' (4) where T, is the decay time constant in the rate of r-process nucleosynthesis. From the equation (4) it follows that for every pair of nuclides i and j , Ni =(Pi) /y- 1/T, *e-r n - e-rA., P·J o 'l/1.,,. - ]/'~' 1 r e - rn - e-rlv NJ. Figure 5.3 shows the field of possible solutions for 232Th/238U (5) and 235Uf"3 8U production ratios from the equation (5), given the Th/U ratio today between 3 .55 and 3.75 and age of the Galaxy between 13 and 15 Gyr. There is a full range ofT, from essentially zero to infinity. As follows from this plot, the field of existing solutions barely overlaps with estimated production ratios. It is possible to calculate possible T, for the time interval and production rates, defined by the overlap of the solutions for equation (5) with the field of production rate estimates. This range corresponds to only 13 Gyr galactic age and T, >35 Gyrs, which is approaching a uniform production (case of Figure 5.2). 3). Now let us consider a more general case, which includes the two scenarios above as special cases. We assume that initial exponential growth in the galactic production in the beginning of the galactic history is followed by an exponential decrease (Figure 5.1.c). This scenario allows for the delayed supernovae production in the disc relative to the globular cluster age. Thus we assume r-process production rate to be proportional to (e.'11' 1 -e·"~2), where 't 1 > 't2 , starting with P;(O)=O: dM -P·(e -rt t • - e -r t t ' ) - N 'l .N dt - - I I (6) Figure 5. I .c shows only a general shape of change in supernovae production. The position and amplitude of the peak in Figure 5.4 depends on values 't 1 and 't2• The Ill parameter 't 1 IS the decay constant for exponential decrease in the production rates (analog to T, in the second model). The parameter 't 2 is a time constant defining infall from the halo to the disc. From the equation (6) for the nuclide pair: N; Nj I P; . e-r '' . = -p [ e-~' -1 e-~-1] 11 'ti - II 't I 1I 'ti- 1I 't 2 >.: ~ >.: ~-, ------:-,--~ j e e -1/ TJ-I/t i _) (7) e -l ltr l l o -1 where 't1 and 'ti and are mean Jives of the two nuclides. The parameters 't 1 and 't2 can be calculated assuming 13-15 Gyr Galactic age and solar Th!U ratio 3.55-3.75. Figure 5.4 shows fields of possible solutions for relative production rates for 13, 14 and 15 Gyr galactic age. Although we cannot distinguish between the three fields, this model has a bigger overlap with estimated relative production rates than the previous ones. A semi-log plot of t 2 vs. 't 1 in Figure 5.5 shows contours of P 235/P238 relative production rates calculated for 15 Gyr galactic age. The lower part of the plot (field of 't2 > t 1) is not allowed in the frame of this model. The shaded area, where 1.2<P 235/P238< 1.52, is the field where solutions of equation (7) overlap with the field of estimated ratios. Most solutions correspond to large t 1, i.e., slow decay of the production rate. For a 15 Gyr galactic age, the fastest possible decay (==8 Gyr decay time constant) corresponds to a very slow infall rates (==7.5 Gyr time constant). Faster decays are 112 possible with younger galactic ages. The dependence of minimum decay constants for 13 Gyr on the infall time constant and Th!U ratio is shown in Table 5.2. To achieve a short infall time constant ('t 2 <0.5 Gyr), the decay constant must be very big (>100 Gyr), approaching uniform production. The dependence of the minimum infall time constants on the assumed ranges of Th!U ratios and the galactic ages is shown in Table 5.3. This table shows, that production starting earlier in the galactic hystory, the minimum infaJI times become progressively higher. From the models considered for 15 Gyr galaxy, model 3 is the only one that is consistent with theoretically estimated actinide relative production rates. Thus, taking in account galactic infall, we can limit the range of time constants 't 1 and 't2 for 15 Gyr galaxy. Time-dependence of the galactic production rates can be described as a case between nearly uniform production for extremely fast infall rates and moderately fast decay time with very slow infall (8 Gyr). 113 Table 5.1 Production ratios from r-process calculations. 232Thl238u m u;238u Reference 1.65 1.42 Seeger et al., 1965 1 .74~~ 1.46~~ Schramm and Wasserburg, 1970 1.9 1.89 Seeger and Schramm, 1970 1.8±0.1 1.42±0.19 Fowler 1977 1.9 1.5 Symbalisty and Schramm, 1981 1.5 1.1 Krumlinde et al. , 1981 1.9~~ 1.5~; Schramm, 1982 1.4 1.24 Thielemann et al., 1983 1.71±0.07 1.34±0.19 Fowler, 1987 Same as Fowler, 1977, with 25°Cm-SF corrected 1.6 1.16 Cowan et al., 1987 With ~-delayed fission 1.53 1.26 Thielemann et al. , 1989 Without ~-delayed fission 1.73 1.79 Thielemann et al., 1993 Equal abundances assumed Notes 114 a) • b) Solar System formation c) Time Today Figure 5.1 Examples of different models of nucleosynthetic production histories. a) constant production until the formation of the Solar System, b) exponential decrease in production, c) rising production in the beginning of the galactic history followed by exponential decrease. 115 2.0 1.9 1.8 1.7 1.6 00 I 1.5 ~ 1.4 - ~ 1.3 ~ ~ ~ 1.2 1.1 0 ~ ~ ~ 0 / 0 0 ./ i' 1/.Y' I I • .i O ' ..__.<"-· 0 ... . --~~ ·- _,...-'" -· ·' {3Gyr lSGyr t 1.0 1.0 1.2 1.4 1.6 1.8 2.0 P235/238 Figure 5.2 Solutions for the relative production rates for uniform production (equation 3) for the galactic ages from 13 to 15 Gyr. The height of the bar is determined by allowing today's solar system Th/U to vary between 3.55 and 3.75. Circles are production values from Table 5.1 for comparison (filled circles are three the most recent values). 116 2 /; / 1.8 I fo • 00 ~ ~ ~ !• .I 1.4 o) 0 / 0 0 1.6 ......----... ,.' 0 / i •• r -< / l y .-·/ ~ ~ ~ 1.2 1 0.8 1 .5 2 2.5 3 3.5 P235/P238 Figure 5.3 Calculated possible solutions for today's Th/U between 3.55 and 3.75 and age of the galaxy between 13 and 15 billion years. Circles are theoretical estimations of rprocess production rates as in Figure 5.2. There is no overlap of two fields for an age> 13 Gyr. 118 l.E+ lt '~~ mSJP'23$=l~- 1.E+09 S.OE+OB S.SE+OO 1. 1E+10 1.6E+10 2.1E+10 2.6E+l0 tau2, infall time constant Figure 5.5 Calculated parameters -r, and -r2 from equation (7) for different 235Uf38U supernovae production rates and 15 Gyr galactic age. Shadowed area is the field of overlapping solution with area of theoretical estimations of the production rates (see Figure 5.4). 119 Table 5.2 The minimum decay time constant for 13 Gyr galactic age as a function of t 2 and Th!U. Th!U today -~~:_2~~....___].~~9 ! 5 : 3 ! 1i o.5 1 0.2 ! 1 5 ' 6.5 . 35 i so i X 1 i 3. 62-! _ _]_}_?_ 51 5 5.3 : 4.6 17 i 13 3o : 20 1 X i 31 no overlap with production estimates from Table 5.1. Table 5.3 The dependence of the minimal infall time constant on Th/U and the galactic age. Th!U today 120 Summary Non-volatile elements in CI chondrites are considered to have average solar system abundances. However, literature data show big variations in Th/U. Ordjnary chondrites have refractory element abundances identical to CI chondrites; therefore, we should be able to rely on ordinary chondrites for estimation of a more precise solar Th/U ratio. Whole rock Th. U analyses prior to our work showed a bimodal distribution in ordinary chondrites, with some meteorites showing an anomalously high Th/U of 6. One of those meteorites, Harleton. was the key for the current study. Isotope dilution thorium and uranium analyses by inductively-coupled plasma mass spectrometry of 12 samples of Harleton (L6) show a much larger scatter than was previously observed in equilibrated ordinary chondrites. The bimodal distribution of the older data was only apparent. Th/U linearly correlates with 1/U in Harleton and in the total equilibrated ordinary chondrite data set as well. Such a correlation suggests a two component mixture and this trend can be quantitatively modeled as reflecting variations in the mixing ratjo between two phosphate phases: chlorapatite and merrillite. The major effect is due to apatite variations, which strongly control the whole rock U concentrations. Phosphorous variations will tend to destroy the Th/U vs. 1/U correlation, and measured P concentrations on exactly the same samples as U and Th show a factor of 3 range. It appears that the P variations are compensated by inverse variations in U (a dilution effect) to preserve the Th/U vs. 1/U correlation. In ordinary chondrites, data on actinide, P, and Cl chemistry are consistent with contemporary views on the origin and evolution of these materials. The parent material of 121 ordinary chondrites resulting from nebular processes had Th and U primarily concentrated in chondrule glass with P present in metal (possibly sulfide) phases, the latter a consequence of condensation or at least high temperature nebular phase equilibrium. The chlorine host phases are unknown, but apparently very heterogeneousl y distributed. Very early after the accumulation of parent bodies, more oxidizing conditions resulted in phosphate nucleation, preferentially as binary grains with metal or sulfide (Jones and Burnett, 1979). The relatively large chondritic P/Cl ratio compared to apatite stoichiometry and the heterogeneous Cl distribution resulted in a relatively homogeneous distribution of merrillite and a very heterogeneous distribution of apatite. During the later stages of ordinary chondrite metamorphism, following crystallization and recrystallization of chondrule glass, the actinides redistributed themselves into the phosphate phases with gm-scale heterogeneities impressed on bulk actinide (especially U) concentrations. The unequal distribution of Th and U among the two phosphates is probably a consequence of equilibrium partitioning. Thus, except for an inability of apatite grains to coarsen during metamorphism, equilibrated chondrites possibly can be regarded as truly equilibrated from the point of view of actinide chemistry. Because variations in whole rock Th/U are consequences of phosphate sampling, a weighted average of high accuracy Th/U measurements in equilibrated ordinary chondrites should converge to a significantly improved average solar system Th/U. Based on the combination of our and literature isotope dilution data, we calculate a solar system Th/U value of 3.53 (by weight), with 1crmean=O. I 0, which is significantly lower and more precisely defined than previous estimates. 122 If this number is indeed an average solar system value, it brings up an issue of a significantly higher Th/U ratios of the Earth (-3.9, see Chapter 1.1 ). Such difference can be attributed to U, behaving more as moderately volatile element, rather than refractory lithophjJe. Interestingly, in this case, the true average solar system Th/U must be even lower than our estimate of the ordinary chondrite ratio. An other possibility is that the Earth reservoir was more contaminated with the high Th!U refractory material than ordinary chondrites. Analogous to ordinary chondrites, an apparent two-component correlation of Th!U vs. 1/U in carbonaceous chondrites was previously suggested by Rocholl and Jochum ( 1993). To explore the possibility of apatite producing Th!U variations in CI and CM chondrites, we performed P analyses on the solutions from previous CI leaching experiments from Chen et al. (1993) and our own leaching experiments in CM-type carbonaceous chondrites. Th!U variations in Orgueil (CI) can be explained by the redistribution of actinides (mostly U) by the aqueous fluid into secondary mineral phases, such as carbonates and sulfates. Our data show that the mobile phase is not a phosphate, contrary to the interpretations of Rocholl and Jochum. The redistribution of alteration products, rich in uranium. but not thorium, can produce a real 2-component mixing line for the whole rock CI analyses. To become mobile, uranium has to be partially oxidized to higher valence tates. Unlike CI and OC whole rock analyses, Th!U variations in CM chondrites are mostly due to Th variations. The Th!U vs. 1/U correlation is an artifact of the older data. 123 There are at least 4 components in CM chondrites responsible for the Th!U variations we see in the whole rock analyses. First. fine matrix material in Murchison contain U-rich carbonate/sulfate phases (analogous to Orgueil), but the total amount of matrix U is not significant on a total rock basis. Matrices in carbonaceous chondrites are often regarded as primitive. Our study of the Murchi on (CM2) matrix showed that it is P and actinide-depleted. Second, CAl and chondrules are major reservoirs for actinides, which is consistent with contemporary views on the origin and evolution of these materials. Finally, our data show an easily leachable phase of high Th/U, likely carbonates produced by alteration of CAis as with Cis. Phosphates play only a minor role as actinide carrier phases in CM chondrites. Establishing a more prec1 e solar system Th/U allowed us to re-visit nucleocosmochonology models, but starting from a point different from other re earchers. We accept the estimates of the minimum galactic age based on the halo globular clusters and calculate the required relative supernovae production rates for 232 Th/238U and 235U/238U, assuming different models for the time dependence of r-process nucleosynthesis. The required production rates can be compared with independent nuclear physics estimates. We considered three simple models of the galactic supernovae production: (I) constant production over the galactic history until! solar was formed; (2) production exponentially decreasing with time, and (3) initial rapid increase in production (i nfall), followed by the exponential decrease. The calculated production rates for the first two models are compatible with part of the range of estimated value for the galactic age < 13 Gyr. The older galaxy requires time-dependence of the galactic production rates to 124 be a case between nearly uniform production for extremely fast infall rates and moderately fast decay time with very slow infall (8 Gyr). 125 References Allegre C.J .. Dupre B. and Lewin E. (1986) Thorium/Uranium ratio of the Earth. Chem Geol., 56,219-227. Anders E. and Ebihara M. ( 1982) Solar-system abundances of the elements Geochim. Cosmochim. Acta, 46, 11 , 2362-2380. Anders E. and Grevesse N. (1989) Abundances of elements: meteoritic and solar. Geochim. Cosmochim. Acta, 53, 197-214. Asmerom Y. ( 1999) Th-U fractionation and mantle structure. Earth Planet. Sci. Lett. 166, 163-175. Boynton W.V. (1978) Fractionation in the solar nebula, II. Condensation ofTh, U, Pu and Cm. Earth Planet. Sci. Lett. 40, 63-70. Brearley A.J. and Jones, H.J. (1998) Chondritic meteorites, in Planetary Materials (ed. J.J.Papike), Reviews in Mineralogy Series, MSA, Washington, DC, 3-01 - 3-398. Bunch T.E., Chang S. , Frick U., Neil J. and Moreland G. (1979) Carbonaceous chondrites -I. Caracterization and significance of carbonaceous chondrite (CM) xenoliths in the Jodzie howardite. Geochim. Cosmochim. Acta, 43, 1727-1742. Burbidge E.M .. Burbidge G.R., Fowler W.A. and Hoyle F. ( 1957) Rev. Mod. Phys. 29, 547. Burnett D.S., Stapanian M.I. and Jones J.H. ( 1982) Meteorite actinide chemistry and cosmochronology. In Essays in Nuclear Astrophysics (eds. C.A. Barnes, D.O. Clayton and D.N. Schramm), pp. 141-158. Cambridge University Press. Carretta E., Gratton R.G., Clementini G., et al. (2000) Distances, ages, and epoch of formation of globular clusters. Astrophys J. 533, 215-235. Carretta E., Gratton R.G. , Clementini G., and Pecci F.F. (2000) Distances, ages, and epoch of formation of globular clusters Astroph. Jour., 533, 215-235. Chen J.H. and Tilton, G.R. ( 1975) Isotopic lead investigations on the Allende carbonaceous chondrite. Geochim. Cosmochim. Acta, 40, 635-643. Chen J.H. and Wasserburg G.J. (1981) Isotopic determination of Uranium in picomole and subpicomole quantities. Analit. Chem. 53, 2060-2067. Chen J.H. and Wasserburg G.J. ( 1985) Peculiar U-Th-Pb systematics in Shergotty: A Martian lead? (abstract). Proc. Lunar Plan et. Sci., 16, 7-8. 126 Chen J.H., Wasserburg G.J. and Papanastassiou D.A. (1993) Th and U abundances in chondritic meteorites (abstract). Lunar Planet. Sci., 24, 277-278. Clarke, Roy S., Jr. and Henderson, E.P. (1974-75) The Harleton, Texas, Chondrite. Smithsonian contributions to the Earth Sciences, 19, 61-70. Clayton R.N. and Mayeda T( 1989) Oxygen isorope classification of carbonaceous chondrites (abstr,) Lunar Planet. Sci., 67, 151-161. Collerson K.D. and Kamber B.S. (1999) Evolution of the continents and the athmosphere inferred from Th-U-Nb systematics of the depleted mantle. Science , 283, 1519-1522. Crozaz G. (1974) U, Th and extinct 244Pu in the phosphates of the St. Severin meteorite. Earth PLanet. Sci. Lett., 23, 164-169. Crozaz G. (1979) Uranium and thorium microdistributions in stony meteorites. Geochim. Cosmochim. Acta, 43, 127-136. Crozaz G., Pellas P., Bourot-Denise M., de Chazal S.M., Fieni C., Lundberg L.L. and Zinner E. (J 989) Plutonium. uranium and rare earths in the phosphates of ordinary chondrites; the quest for a chronometer. Earth Planet. Sci. Lett. 93(2), 157-169. Devouard B. and Busek P.R. ( 1997) Phosphorus-rich iron, nikel sulfides in CM2 chondrites: condensation or alteration products (abstr) Met. Planet. Sci. 32, A34. Ebihara M. and Honda M. ( 1984) Distribution of rare earth elements and uranium in various components of ordinary chondrites. Meteoritics, 19(2), 69-77. Ebihara M. and Honda M. ( 1987) Rare earth elements in Ca-phosphates of Allende carbonaceous chondrite. Meteoritics, 22(3), 179-190. Endress M . and Bischoff A. (1996) Carbonates in CI chondrites: clues to parent body evolution Geochim. Cosmochim. Acta, 60, 489-507. Fowler W.A. (1977) Nuclear cosmochronology. In The R.A. Welch Found. Conf on Chern. Research XXI. Cosmochemistry, pp. 61-133, Houston, Texas. Fowler W .A. (1987) The age of the observable uni verse. Q. Jl R. Astr. Soc. 28, 87-I 08. Grossman L. and Olsen E. ( 1974) Origi n of the high temperature fraction of C-2 chondrites. Geochim. Cosmochim. Acta, 38, 1119- 1 140. Grossman J.N .. Rubin A.E., Nagahara H. and King E.A.( 1988) Properties of chondrules. In Meteorites and the early solar system (ed . J. F. e. Kerridge and M.S. e. Matthews), 619-659. The Uni versity of Arizona Press, Tucson. 127 Hagee B., Bernatowicz T.J., Podosek F.A., Johnson M.L., Burnett D.S. and Tatsumoto M. ( 1990) Actinide abundances in ordinary chondrites. Geochim. Cosmochim. Acta, 54, 2847-2858. Ireland T.R., Compston W., Williams I.S. and Wendt I. ( 1990) U-Th-Pb systematics of individual perocskite grains from the Allende and Murchison carbonaceous chondrite . Earth Planet. Sci. Lett. 101 , 379-387. Irion R. (2000) A crushing end for our Galaxy. Science, 287, 62-64. Jarosewich, E. and Dodd, R.T. (1985) Chemical variations among L-chondrites; IV, Analyses, with petrographic notes, of 13 L-group and 3LL-group chondrites. Meteoritics, 20, 1' 23-36. Jarvis, K.E. (1997) Inductively coupled -plasma-mass spectrometry. In Modern Analytical Geochemistry (ed. R. Gill), pp.171-187. Longman, Singapore. Jones J.H. and Burnett D.S. (1979) The distribution of U and Pu in the St. Severin chondrite. Geochim. et Cosmochim Acta, 43, 1895-1905. Jones J.H., Burnett D.S., Criswell P.R. and Merrill R.B . (1980) Laboratory studies of acti nide metal-silicate fractionation. Meteorite and regolith studies (abstract). Lunar Planet. Sci. , 1 I , pp. 995-100 I. Kallemeyn and Wasson ( 1981. 1982, 1986) reference in Larimer J.W. and Wason J.T. (1988). Keppler H. and Wyllie P.J. (1990) Role of fluids in transport of uranium and thorium in magmatic processes. Nature, 348, 531-533. Kornacki A.S. and Fegley B. Jr ( 1986) The abundance and relative volatility of refractory trace elements in Allende Ca.Al-rich inclusions; implications for chemical and physical processes in the solar nebula. Earth Planet. Sci. Lett. vol.79, 217-234. Korotev R.L. (1998) Concentrations of radioactive elements in lunar materials. 1. Geophl. Res. 103 (EJ ), 1691-1701. Kraus K. .A. and Nelson F. ( 1958) Metal separations by anion exchange. Am. Soc. fo r Testing and Materials Spec. Pub!., 195, 27-57. Larimer J.W. and Wasson J.T. (1988) Refractory lithophile elements. In Meteorites and the early solar system (eds. J. F. e. Kerridge and M.S. e. Matthews), pp. 394-415. The University of Arizona Press, Tucson. 128 Ladders K. and Fegley B. Jr. ( 1993) Lantanide and actinide chemistry at high C/0 ratios in the solar nebula. Earth Planet. Sci. Lett. 117, 125-145. Mackinnon l.D.R., Kaser S.A. (1990) Trace phases in CI chondrites Alais and Orgueil. Proceedings of the abstracts of the 53rd meeting of the Meteoritical Society 25(4), 381382. MacPherson G.J. and Davis A.M. ( 1994) Refractory inclusions in the prototypical CM chondrite, Mighei. Geochim. et Cosmochim Acta, 58, 5599-5625. Manhes G .. Allegre C.J., Dupre B. and Hamelin B. (1979) Lead-Lead Systematics, the "Age of The Earth" and the chemical evolution of our planet in a new representation space. Earth Planet. Sci. Lett. 44, 91-104. Moore, C.B. ( 1971 ) Phosphoru . In Handbook of elemental abundances in meteorites. (B. Masoned.) Gordon and Breach Science Publishers. p. 555. Muller W.F .. Kurat G. and Kracher A. ( 1979) Chemical and crystallographic study of cronstedtite in the matrix of the Cochabamba (CM2) carbonaceous chondrite. Tschermaks Min. Petr. Mitt. 26, 293-304. Murrell M.T. and Burnett D.S. ( 1983) The behavior of actinides, phosphorus. and rare earth element during chondrite metamorphism. Geochim. Cosmochim. Acta, 47, 19992014. Nazarov M.A .. Brandstatter F. and Kurat G. ( 1997) Comparative chemi try of P-rich phase in CM chondrites (abstract) LPSCXXVIII, I003-1004. Palme H., Larimer J.W. and Lipschutz M.E. ( 1988) Moderately volatile elements. In Meteorites and The Early Solar System (eds Kerridge F. and Matthews M.S.) pp. 436461 . University of Arizona Press, Tucson. Pellas P. and Storzer D. ( 1975) Urani um and plutonium in chondri tic phosphates. Meteoritics 10,471-473. Podo ek F.A. and Swindle T.D. ( 1988) Nucleocosmochronology. In Meteorites and the Early Solar System (eds. J. F. Kerridge and M. S. Matthew), pp. 1114-1126. University of Arizona Press. Tucson. Reed G.W. ( 1971) Chlorine. In Handbook of elemental abundances in meteorites. (ed. B. Mason), pp. 143-148. Gordon and Breach Science Publishers. Rocholl A. and Jochum K.P. ( 1993) Th, U and other trace elements in carbonaceous chondrites: Implications for the terrestrial and solar-system Th/U ratios. Earth Planet. Sci. Lett. 117, 265-278. 129 Ruzicka A., Killgore M., Boesenberg J. and M. Prinz (2000) Portales Valley: not just another "ordinary" chondrite (abstract). Meteoritics, 35 supplement, A 139-A 140. Schramm D.N. (1982) The r-process and the nucleochosmochronology. In Essays in Nuclear Astrophysics, eds. Barnes C.A., Clayton, D.D. and Schramm D.N., Cambridge Univ. Press, 325-353. Schramm D.N. and Wasserburg G.J. ( 1970) Nucleocosmochronologies and the mean age of the elements. Astrph. J. 162, 57-69. Seeger P.A.. Fowler W.A. and Clayton D.D. (1965) reference in Thielmann et al. (1993) Shinotsuka K. and Ebihara M. ( 1997) Precise determination of rare earth elements, thorium and uranium in chondritic meteorites by inductively coupled plasma mass spectrometry- a comparative study with radiochemical neuron activation analysis. Analytica Chimica Acta, 338, 237-246. Shinotsuka K., Hidaka H. and Ebihara M. (1995) Detailed abundances of rare earth elements, thorium and uranium in chondritic meteorites: An ICP-MS study. Meteoritics 30, 694-699. Sipiera P.P., Tarter J., Moore C.B., Dodd B.D. and Johnson R.A. (1980) Gomez, Terry County, Texas: a new meteorite find. Meteoritics, 15(3). 201-210. Stacey J.S. and Kramers J.D. ( 1975) Approximation of terrestrial lead isotope evolution by two-stage model. Earth Planet. Sci. Lett. 26, 207-221. Tatsumoto M . (1966) Genetic Relations of Oceanic Basalts as Indicated by Lead Isotopes. Science 153, 1094-1101. Tatsumoto M., Knight R.J. and Allegre C.J. ( 1973) Time differences in the formation of meteorites as determined from the ratio of lead-207 to lead-206. Science, 180, 1279-1283. Tatsumoto M., Unruh D.M. and Desborough, G.A. (1976) U-Th-Pb and Rb-Sr systematics of Allende and U-Th-Pb systematics of Orgueil. Geochim. Cosmochim. Acta, 40, 617-634. Thielemann F.-K., Bitouzet J.P., Kratz K.-L., Cowan J.J. and Truran J.W. (1993) Operation of the r-process and cosmochronology. Phys. Rep. 227, 269-281. Unruh D.M. (1982) The U-Th-Pb age of equilibrated L chondrites and a solution to the excess radiogenic Pb problem in chondrites. Earth Planet. Sci. Lett. 58, 75-94. Von Michaelis H., Ahrens L.H. and Willis J.P. ( 1969) The composition of stony meteorites II. The analytical data and an assessment of their quality. Earth Planet. Sci. Lett. 5, 387-394. 130 Yabuki H. and El Goresy A. ( 1986) Phosphate-bearing microspherules in chondrules of unequilibrated ordinary chondrites. Yanai, Keizo, Takeda, Hiroshi and Shimoyama. Akira (editors), Proceedings of the Tenth symposium on Antarctic meteorites, Memoirs of National Institute of Polar Research. Special Issue, 41 , p. 235-242. Zolensky M.E., Bodnar R.J. , Gibson E.K. Jr., Nyquist L.E., Reese Y., Shih, ChiYu, Wiesmann H. (1999) Asteroidal water within fluid inclusion-bearing halite in an H5 chondrite, Monahans. Science, 285, 1377-137. 131 Appendix I. Fibrous nanoinclusions in massive rose quartz. Part 1: The origin of rose coloration. Thjs is a manuscript in press, to be published in the American Mineralogist, April 200 1. 132 Fibrous nanoinclusions in massive rose quartz. Part 1: The origin of rose coloration. Julia S. Goreva, ChiMa and George R. Rossman Division of Geological and Planetary Sciences, California Institute of Technology, MS 100-23, Pasadena, CA, 91125 ABSTRACT Pink nanofibers were extracted from rose quartz from 29 different pegmatitic and massive vein localities throughout the world. Their width varied from 0.1 to 0.5 _m. On the basis of optical absorption spectra of the fibers and the initial rose quartz, we conclude that these nanofibrous inclusions are the cause of coloration of massive rose quartz worldwide. These fibers do not occur in the rare, euhedral variety of pink quartz. Redox and heating experiments showed that the pink color of the fibers is due to Fe-Ti intervalence charge transfer that produces an optical absorption band at 500 nm. Based on the XRD patterns and characteristics of pleochroism, the best match for these inclusions is dumortierite. FTIR and Raman spectra consistently did not exactly match the standard dumortierite patterns, suggesting that this fibrous nano-phase may not be dumortierite itself, but rather a closely related material. INTRODUCTION The rose variety of quartz is known and valued from time immemorial as an item of beauty, a source of rock for decorati ve carvings and as a jewelry stone. Even though it is 133 one of the less common colored varieties of quartz, rose quartz is found at numerous localities around the world. Massive rose quartz occurs most commonly in granitic pegmatites and sometimes in massive hydrothermal veins. It makes up the whole or part of the core and/or joins together the potassium feldspars of the blocky zones of pegmatites (Cassedanne and Roditi 1991 ). Associated minerals are schorl, cassiterite, beryl, and phosphate minerals. Coloration of massive rose quartz varies from bright to pale pink, sometimes with lavender or orange overtones. It is always slightly- to highly turbid. In spite of extensive studies over the last century, the source of the rose coloration in quartz still remains a subject of active debate. A number of possible mechanisms for the color of rose quartz have been proposed (reviewed in Rossman, 1994), but no consensus has developed. The proposed causes of pink color in quartz can be divided into 3 groups: (1) the presence of individual substitutional/interstitial ions in the quartz structure; (2) charge transfer between pairs various transition elements in the quartz structure; (3) the presence of mineral inclusions. Based on chemical analyses, Holden (1924ab) argued that rose color in quartz is due to Mn 3+ pigmentation. On the other hand, number of researchers later attributed the colorto the presence ofTi 3+ (Wright et al. 1963; Lehmann 1969; Cohen and Makar 1984). Other workers have suggested that the color of rose quartz is due to intervalence charge transfer (IVCT) between Fe2+ + Ti4+ ~ Fe 3+ + Ti3+ (Smith et a!. 1978) or between substitutional Ti4+ and interstitial Ti 3+ (Cohen and Makar 1985). Turbidity and scattering of light in massive rose quartz suggests the presence of a second mineral phase. Fine, subrnicrometer diameter oriented needles were observed and identified as rutile by many workers (e.g. Holden 1924a; Gordon 1945; Wright et al. 1963; Lehmann 1969). Vultee (1955) and Vultee and Lietz (1956) argued that scattering of light by rutile is, in fact, the real cause of the rose color. Later work by Dennen and Puckett ( 1971 ) supported this argument and concluded that the color of rose quartz results from 134 Tyndall scattering by tiny oriented rutile needles. Other pink phases such as tourmaline and manganiferous clays have been identified as inclusions in pegmatitic quartz, but their relationship to common, massive rose quartz has not been explored. More recent studies presented different inclusions in rose quartz as an alternative to rutile. By HF etching of metaquartzite, Folk ( 1989) noticed fibrous resistant inclusions in quartz, which he presumed to be aluminosilicates. Applin and Hicks ( 1987) found that pink microfibers included in massive rose quartz from Montana and South Dakota yielded an X-ray diffraction pattern compatible with that of dumortierite, Si 3 B[A~ _ 75 [] 0 _ 25 0 17 _ 25 (0H) 0 75 ] (Moore and Araki 1978). Ignatov et al. (1990) observed macroscopic (2-3 mm) as well as submicron size inclusions in a rose quartz specimen from a vein in a biotite-muscovite schist from the Caucasus Mountains. Fine inclusions were oriented perpendicular to the c-axis of the quartz crystal. Clusters of larger prismatic needles were identified as dumortierite. The authors extrapolate their phase identification to the submicron size inclusions and conclude that these two types of inclusions (oriented and clustered) represent two generations of the same mineral, dumortierite. Optical absorption spectra, taken from a -2 mm cluster, are similar to the optical spectra of the rose quartz itself. This study and the work of Applin and Hicks ( 1987) motivated us to raise the question of a more global role of dumortierite or other mineral inclusions in rose quartz coloration. A further complication comes from reports about the lack of photo- and thermostability of rose quartz. Cassedanne and Roditi ( 1991 ) report, that in some Brazilian deposits, specimens on the surfaces of mine dumps turned milky white over a fe w years. Fading is often a characteristic of color centers from radiation damage. In other deposits, fragments dumped more than 50 years ago do not show any fading. There are reports that massive rose quartz is unstable under heat. Experiments by lgnatov et al. ( 1990) showed, that the color of rose quartz is lost at -500°C. At this temperature, the larger, clustered inclusions in the quartz lost their color but did not change to another phase. 135 Although normally found in massive aggregates, rose quartz also occurs in a few pegrnatites as euhedral crystals. Based on their optical properties and chemical composition, the coloration of these two types of rose quartz appears to be of a different nature. Massive quartz is slightly- to highly turbid, indicating the presence of the second phase, whereas single-crystal rose quartz (sometimes called pink quartz) is nearly devoid of internal scattering when probed by a narrow laser beam. In addition, single-crystal rose quartz is photochemically unstable, and, unlike massive rose quartz, it is extremely rare. EPR spectra of natural euhedral rose quartz showed that this type of rose quartz contains hole centers induced by ionizing radiation. Maschmeyer and Lehmann ( 1983) have attributed rose quartz color to these centers. Their model consists of substitutional Al and P atoms bridged by 0 ions. This is further substantiated by Balitsky et al. (1996) who synthesized single-crystal rose quartz that contains these types of centers. The Al-0-P centers are responsible for the photochemical instability. EPR studies of some synthetic euhedral rose quartz also revealed [SiO/Li:f centers (Bailey and Well, 1991) and [Ti04 r. In this work we concentrate only on common massive rose quartz. By extraction of inclusions from pegmatitic and hydrothermal vein rose quartz from different localities, we explored whether there is a general relation between the presence of nanofibers and the origin of color in rose quartz. We tried to understand whether massive rose quartz commonly contains fibrous inclusions, whether nanofibers are the reason for its coloration, and what mineral phases make up those nanofibers. MATERIALS AND METHODS Sample preparation The massive rose quartz samples selected for analysis (Table 1) exhibit small differences in tones and intensity of coloration. They all scattered light when probed with a narrow laser beam by, presumably, a second phase. For comparison, a sample of singlecrystal pink quartz (GRR-1790a) with no turbidity, and a sample of dumortierite from 136 Dehesa, CA (GRR 122), were also analyzed. The quartz and dumortierite were crushed to a grain size of 1-5 mm, cleaned successively in methanol, near-saturated oxalic acid, 2N HCl and water. After cleaning, samples were left over night in an oven at 9QOC to dry. The visual intensity of the coloration did not change during sample preparation and heating. The quartz samples were then dissolved in hot (100°C) concentrated (48%) HF. All 29 massive quartz samples produced insoluble fibrous residues (between 50 and 150 ppm of initial sample weight), with a color visually corresponding to the initial coloration of the samples. Residues from dissolution were treated with a 3:1 mixture of concentrated HF and HN0 3 and washed several times (until no visible fluorides were left) in 5% HN03 and H 2 0 followed by centrifugation. Residues were dried on a hot (150°C) plate. Upon drying, all samples (with the exception of GRR 1790a) produced residues of visually homogeneous, fine textured, flaky pink-colored mats. Sample GRR-1790a dissolved entirely. The dumortierite sample from Dehesa, CA. did not experience significant weight loss after acid digestion. Ultimately, thi acid-treated dumortierite sample was compared to the resistant residues from quartz. Analytical methods Imaging and qualitative analysis were performed with Caltech' s CAMSCAN Series II scanning electron microscope and associated Link thin-window SiLi detector and EDS analyzer system. Semi-quantitative X-ray fluorescence spectra of all the fiber mats were taken with a Kevex!IXRF 770 EDS system using Ge and Gd secondary targets. For the XRD analyses, fibers from GRR1815 and dumortierite (Dehesa, CA) were finely ground and loaded on the tip of a glass fiber. Conventional powder diffractometry was not used for two reasons: first, the available sample size was too small, and second, even our finest grinding of the sample produced particles with a very high length to width ratio, minimizing X-ray incidence in the elongated direction of the crystals. X-ray diffraction patterns were obtained with a 57.3 mrn diameter Gandolfi camera that 137 introduces 2 axes of rotation to the fibrous sample to ensure a powder diffraction pattern. The films were read with an estimated accuracy of 0.02° in the 2-theta measurements. Infrared reflectance spectra of all fiber samples and dumortierite were obtained with a Nicolet 60SX FfiR spectrometer. Whole mats of fibers were first placed over a pinhole aperture. Spectra were then obtained with a NicPlan IR microscope at 4 cm·1 resolution at near-normal incidence in the reflection mode. In addition, IR transmission spectra from samples GRR1815, GRR1811 and GRR1864 dispersed in KBr pellet were obtained. A Renishaw 1000 micro-Raman spectrometer with a 2 cm· 1 resolution was used to obtain Raman spectra of dumortierite and fiber samples. Five samples were examined. Mat of fibers were placed at the microscope focus (20x) and probed with a polarization-scrambled 514.5 nm laser operating at 50 mw. Optical spectra were obtained using a home-built 1024 element silicon diode-array spectrometer. Spectra of the fiber mats were taken in a reflectance mode by focusing an approximately 2 mm diameter pot of light on the mat and collecting the reflected light at 90° to the incident light. These spectra were compared to massive rose quartz spectra obtained in a diffuse transmission mode. In these experiments light was directed into a rectangular block of rose quartz and light was collected from an adjacent face and directed to the spectrometer. Milky quartz was used as a reference standard in both of these experiments. RESULTS Physical characteristics All massive rose quartz amples produced fibrous residues that, among different samples, varied from only slightly pink to bright pink. Cluster of parallel fibers show strong pleochroism under the optical micro cope. They are bright pink parallel to the direction of elongation and near-colorle s perpendicular to that direction. 138 SEM observations and sample chemistry Fiber samples obtained from dissolution experiments were examined with SEM. Representative back-scattered electron images of the fibers are shown in Figure 1. Figure l a shows a residual mat from the quartz sample GRR 1820 from the Black Hills, South Dakota. The length of the fibers was probably controlled by the grinding of the initial quartz sample. Figure 1b shows a more detailed image of a fiber cluster from the same sample. The width of individual fibers recovered from different samples generally varied between 0.1 to 0.5 J.lm. The SEM-EDS analyses of the wide (relative to the width of individual crystals) areas of the fiber mats show an aluminosilicate framework with traces Ti and Fe in all samples. The relative intensity of the Al and Si peaks is approximately 2:1. Unfortunately, because of the small ize of the individual fibers, quantitative analysis could not be obtained by EDS. No other elements were detected by EDS XRF. In particular, Sb, Ta, Nb and As were absent even though they are found in holtite, a phase related to dumortierite (Hoskins et al. 1989). An attempt to obtain chemical analyses of fibrous residue with ICP-MS (Elan SOOOA) was also made. Residue were dissolved in lON KOH. The ICP-MS analysis confirmed that boron was present in the fibers at levels of at least 2%, but repeat analyses did not yield reproducible results, presumably as a result of boron loss during the analytical procedure. Occasional larger tabular cry tals were observed in the fiber mats extracted from GRR1819 and GRR1 820. A Si-Al matrix and morphology of these crystals suggest an A12 Si05 polymorph, probably illimanite. X-ray diffraction The ten most intense line for fibers extracted from sample GRR 1815 (Madagascar) and dumortierite (Dehesa, CA) are compared (Table 2) with those of the ICDD standard dumortierite (12-0270). The d-spacings from the diffractograms of the 139 powdered fibers and natural dumortierite are close to those of the JCPDS standard dumortierite pattern, although there are differences in intensities. FTIR spectrometry The IR spectra of the rose fibers are broadly compatible with the spectrum of dumortierite (Fig. 2). AJJ samples show the borate band at 1450 cm· 1 that indicates the presence of the BO/ ion. Comparison of the absorption spectra of dumortierite and the fibers suggests that these phases have approximately the same borate to silicate ratios. The spectrum of dumortierite from Dehesa, CA, closely matches the synthetic dumortierite spectrum (e.g., sample B 12) of Werding and Schreyer (1990), but these spectra consistently differ from those produced by the various fiber samples. Relative band intensity differences in the 1000 and 500-600 cm· 1 regions are the salient differences. Raman Spectrometry Figure 3 shows Raman pectra of a dumortierite crystal (Dehesa, CA), acid-treated dumortierite powder, and three sample of fiber mats from different localities. The two dumortierite spectra are similar, with only minor differences in the relative intensities of two bands in the 500-600 cm· 1 region. These differences can be attributed to the anisotropy of the single dumortierite crystal, a the spectrum was taken with the incident light perpendicular to the c-axis. The spectra produced by the fibers are virtually identical to each other, but they are ignificantly different from the spectra generated by dumortierite (Figure 3). Vertical dashed lines in Figure 3 help to illustrate differences in the positions and intensities of the bands. Optical absorption Optical absorption spectra of fiber mats from GRR1815, GRR1818, GRR1820 and GRR 1874 were compared to the spectra of the corresponding initial rose quartz sample . 140 The spectra of fiber mats showed an absorption band centered at approximately 500 nrn. Moreover, the absorption bands of mats and their rose quartz precursors were nearly identical, indicating that their colors share a common origin (Figure 4). For comparison, an optical spectrum of pink dumortierite (Dehesa, CA) is also presented. In addition to the 500 nm band, it also has a longer wavelength shoulder arising from a second component previously described in Platonov et al (2000). Irradiation and thermal stability of color As noted earlier, the heating experiments by lgnatov et al. ( 1990) showed that their rose quartz and its associated clusters of fibrous inclusions in the quartz loose their color at - 500°C. For several minerals, color loss on beating is associated with radiation-induced colors. Since rose quartz occurs in pegmatites, where a number of minerals such as smoky quartz have radiation-induced color, it is plausible that the pink color of rose quartz was developed upon natural irradiation within the pegmatite body. In order to explore this possibility, we subjected quartz sample GRR1815 to three weeks of heating at 750°C in air. The quartz sample lost its rose color, although the overall turbidity of the quartz did not visually increase. The fibrous inclusions extracted from the heated quartz were colorless. The FTIR spectrum of the colorless fibers was virtually identical to the spectrum of the pink fibrous residue from an unheated aliquot of the same rose quartz (Figure 2) and indicates that a phase change did not occur. The colorless fibers then were irradiated in a 137 Cs source. After exposure to 10 Mrad of gamma radiation, the fibers did not tum pink, suggesting that a radiation-induced center is not a cause of the color. After ruling out the irradiation hypothesis, we designed a number of heating experiments to determine if this loss of color was due to the oxidation of transition elements. Experiments were performed on fibers extracted from 13 of the 29 samples studied. All 13 samples ofrose fibers lost their color upon heating in air at temperatures between 430° and 750°C. The higher the temperature, the more quickly the fibers lost their 141 color. At 750°C it takes few minutes, whereas at 430°C it takes several days. The colorless fibers were reheated at the same temperatures under highly reducing conditions with a mixture of 95% N 2 and 5 % H 2 • Eleven out of 13 samples regained their pink color. Samples GRR-1897 and GRR-1904 remained white. Both of these samples contained thicker fibers (-1 _m), so the reaction kinetics might be slower than for the other samples. No phase transitions occurred during the reduction experiments. DISCUSSION AND CONCLUSIONS Our work confirms previous observations of alurninosilicate fibers in rose quartz and extends it to numerous localities around the world. After separation from the host quartz, these insoluble fibers are the same color as the initial rose quartz. Thus, we argue that the color of rose quartz itself arises from these fibrous inclusions. Several experiments indicate that these fibers are related to dumortierite. However, there are significant differences between their spectroscopic signatures and those of dumortierite. Nevertheless, dumortierite remains a valid point of comparison to the properties of these fibers. Dumortierite itself is commonly associated with quartz in hydrothermal and pegmatitic environments. It is common for dumortierite to contain traces of Fe and Ti which are associated with its color (Alexander et al. 1986). An absorption band in the dumortierite spectrum centered at about 500 nm has been attributed to Fe2+ - Ti4+ IVCT by Ignatov et al. (1990) who also associated the color of dumortierite in association with rose quartz to the color of their sample of rose quartz. Our work supports their proposal for the origin of rose color in dumortierite. Because the fibers we recovered from rose quartz always show traces of Ti and Fe, we propose that the same Fe-Ti charge transfer mechanism operates in the dumortieriterelated fibrous phase extracted from rose quartz. Smith et al. (1978) had previously proposed that a Fe-Ti charge transfer mechanism is responsible for the color of rose quartz but thought that it originated with substitutional components within the quartz itself. 142 Heating experiments in open air showed that the fibers could easily be rendered colorless. Because we were able to return color to the fibers by heating them under reducing conditions, we believe that oxidation and reduction of iron (Fe2+ H FeJ+) in Fe-Ti IVCT centers are responsible for the change of the fiber color from pink to colorless and back. Hemingway et al. (1990) have speculated that the color change that occurs in dumortierite may reflect oxidation of iron and loss of charge transfer among Fe 2+, Fe3+ and Ti+r rather than dumortierite decomposition. Our experiments demonstrate this mechanism in the fibers. Our FTIR study of fibers extracted from rose quartz after heating to 750°C showed that the fibers remained structurally unaltered, even though dumortierite reportedly decomposes above- 700°C at atmosphere pressure (Werding and Schreyer, 1996). Ignatov et al. ( 1990) showed that the color of quartz with the macroscopic dumortierite inclusions is stable below 500°C at 1 atm. and does not depend on the duration of the heating. By contrast, our heating experiments revealed, that there is a time dependence to the color stability of the heated fibers. Fibers lo e their color within a few hours when heated at 500°C, whereas decoloration takes several days at 430°C. Fading of rose quartz The stability of the color of the fibers appears inconsistent with the often reported fading of the color of some fraction of rose quartz after prolonged exposure to the sun in mine dump . We have not observed this process in any of our samples and have no direct experimental results to bear on thi problem. We speculate that the photo oxidation of iron in the fibers is a plausible but unlikely process. It is further possible, but unproven, that common massive rose quartz could contain a component of the photosensitive phosphorous center found in crystalline pink quartz. This would not explain the lore that orne faded rose quartz regain its color when soaked in water. Another, more conservative, 143 explanation for the fading of massive rose quartz would be the solar heat-induced expansion and shattering of C02 inclusions causing microfractures and scattering of light near the surface of the quartz. Origin of the fibers The origin of the fibers is, in part, related to the origin of rose quartz. A study of primary fluid inclusions in the rose quartz from Black Hills, SD (the locality of the sample ORR 1820), shows that crystallization of the quartz occurred in the temperature range 400° to 450°C, at 3.5 to 4 kbars pressure (Sirbescu, 1997, pers. comm., 1998). Because these temperatures are in or near the temperatures at which the rose color was lost in our heating experiments, it seems unlikely that the pink fibrous phase was captured during the growth of the quartz. Another scenario is that an initially colorless fibrou phase grew within quartz and rose color was gained after crystallization. The lack of respon e of the colorless fibers to irradiation appears to rule out color generation from a radiation-induced color center. Sirbe cu ' s (pers. cornm., 1998) work shows that the composition of the primary inclusions in this quartz is mainly C02 with minor CH4 . While the latter indicates plausible reducing conditions in the fluid during the quartz crystallization or crystallization of the inclusions that might affect the oxidation state of Fe, this scenario also seems unlikely. Our favored explanation is that the pink fibrous phase exsolved from crystallized quartz at temperatures below 430°C. There are many report of asterism in rose quartz. The most detailed was that of Kalkowsky (1915-20). The asterism in rose quartz implies orientation of fibers perpendicular to c- axis and along three equivalent crystallographic directions, which, presumably. is diagnostic of exsolution. Fine differences between the FTIR and Raman spectra of the fibers and dumortierite could also be due to the presence of an impurity phase. For example, the 144 fibrous mineral inclusions could be a mixture of true dumortierite and a related phase. To address these questions, a more detailed HRTEM and AEM investigation of the chemistry and structure of individual crystals of these fibers is in progress (Chi Ma et al., in prep.). ACKNOWLEDGMENTS We thank the White Rose Foundation for their support of our research on color in minerals. We also thank John Brady, Milka de Brodtkorb, Rock H. Currier, Richard V. Gaines, George Harlow, Brian D. Hicks, Lothar Jung, Knut E. Larsen, Taijing Lu, Ole Peterson, Gunnar Raade, Ed Swoboda, Matt Taylor, and Jer Wednt for their interest in our work and for providing quartz samples. We also thank A.N. Platonov for multiple discussions regarding his work, and Liz Arredondo for assistance with laboratory measurements, George Harlow and an anonymous reviewer for valuable comments and Jeffrey Post for the editorial review of the manuscript. 145 REFERENCES CITED Alexander, V.D., Griffen, D.T., and Martin T.J. (1986) Crystal chemistry of some Fe- and Ti-poor dumortierites. American Mineralogist 71, 786-794. Applin, K.R. and Hicks, B.D. ( 1987) Fibers of dumortierite m quartz. American Mineralogist 72, 170-172. Bailey, Phillip, and Well, J. (1991) EPR study of the [SiO./Lif centre in _-Quartz. J. Chern. Soc. Faraday Trans. 87(19), 3143-3146. Balitsky, V.S., Makhina, LB ., Emelchenko A.G., et al. (1996) Experimental'noe izuchenie usloviy obrazovaniya kristallov rozovogo fosforosoderzhaschego kvarca i prichiny ego redkoy vstrechaemosti v prirode. [Trans. title: Formation of rose phosphorus-bearing quartz and causes of its rarity in nature - an experimental-study]. Geokhimiya, 11 , 1074 - 1081 (in Russian). Cassedanne, J.P. and Roditi, M. (1991) Crystallized and massive rose quartz deposits in Brazil. Journal of Gemmology 22, 5, 273-286. Cohen, A.J. , and Makar, L.N. ( 1984) Differing effects of ionizing radiation in massive and single crystal rose quartz. N eues J ahrbuch fur Mineralogie-Monatshefte 11 , 513-521. Cohen, A.J., and Makar, LN. (1985) Dynamic biaxal absorption spectra of Ti 3+ and Fe2+ in a natural rose quartz crystal. Mineralogical Magazine 49, 709-715. Dennen, W.H. and Puckett, A.M. (1971) On the chemistry and color of rose quartz. Mineralogical Record 3, 226-227 Folk, R.L. ( 1989) Internal structuring of quartz as revealed by boiling concentrated hydrofluoric acid. Journal of Geological Education 37, 250-260. Gordon, S.C. ( 1945) The inspection and grading of quartz. American Mineralogist, 30, 269-295. Hemingway, B.S., Anovitz, L.M., Robie, R.A. and McGee, J.J. (1990) The thermodynamic properties of dumortierite Si 3 B[A~ 75 [] 0 25 0 17 _25 (0H) 0 75 ]. American Mineralogist 75, 1370-1375. 146 Holden, E.F. (1924a) The cause of color in rose quartz. American Mineralogist 9(4), 7589. ------- (1924b) The cause of color in rose quartz (concluded). American Mineralogist 9(5), 101-108. Hoskins, B.F., Mumme, W.G. and Pryce, ( Si 2 _ 25 Sb 0 _ 75 )B[A~A~..t 3 T(\)_27 )[] 0 _30 )0 15 (0,0H)2.2 5 ] ; M .W . crystal (1989) structure and Holtite, crystal chemistry. Mineralogical Magazine 35, 457-463. lgnatov, S.l., Platonov, A.N., Sedenko, V.S., and Taran, M.N. (1990) Pozheve zabravlennya kvarcu, zvyazane z rnikrovklyuchennyarni dumortyeritu. [Trans. title: Pink coloration of quartz related to dumortierite microinclusions.] Dop. AN URSR. Ser. B. him. ta bioi. nauki 7, 23-26 (in Ukranian). Kalkowsky, E. (1915-20) Opalezierender Quarz. [Tran . title: Opalescent quartz] Zeitschrift fur Krystallographie und Mineralogie, 53, 23-50 (in German). Lehmann, G. (1969) Zur Farbe von Rozenquarz. [Trans. title: On the color of rose quartz] Neue Jahrbuch fur Mineralogie-Monatshefte, 222-225 (in German). Maschmeyer, D. and Lehmann, G. (1983) A trapped-hole center causing ro e coloration of natural quartz. Zeitschrift fur Kristallographie 163, 181-196. Moore, P.B. and Araki, T. (1978) Dumortierite, Si 3 B[~_ 75 [] 0 _ 25 0 17 _ ~ 5 (0H ) 0 _ 75 ]: A detailed structure analysis. Neue Jahrbuch fur Mineralogie Abhandlungen, 132,23 1-241. Platonov, A.N., Langer, K. , Chopin, C., Andrut, M. and Taran, M.N. (2000) Fe 2+ -Ti++ charge-transfer in dumortierite, European Journal of Mineralogy, 12, 521-528. Rossman, G .R. ( 1994) Colored varieties of the silica minerals. In P .J. Heaney, C. T. Prewitt, and G.V. Gibbs, Eds., Silica. Physical Behavior, Geochemistry and Material Applications. Reviews in Mineralogy 29, Mineralogical Society of America, Washington, D.C., p. 433-462. Sirbe cu, M-L. C. and Nabelek, PJ. ( 1997) Fluid evolution in a simple pegmatite from the Black Hills pegmatite field, South Dakota, Geological Society of America, 1997 annual 147 meeting, Abstracts with Programs - Geological Society of America, 29, p. 457-458, 1997. Geological Society of America, 1997 annual meeting, Salt Lake City, UT, United States, Oct. 20-23, 1997. Smith, G., Vance, E.R., Hasan, Z., Edgar, A., and Runciman, W.A. (1978) A charge transfer mechanism for the colour of rose quartz. Physica Status Solidi (A) 46, K135-K140. Vultee, J. von (1955) Uber die orientierten Verwachsungen von Rutil in Quartz. [Trans. title: The intergrowth of the oriented rutile inclusons in quartz.] Neues Jahrbuch Fur Mineralogie-Abhandlungen 87, 389-415 (in German). Vultee, J. von and Lietz, J. ( 1956) Uber die Rolle des Titans als Farbungsursache von Blau- und Rosenquarzen. [Trans. title: About the role of Ti in the pegmentation of the blue and rose quartz.] Neues Jahrbuch fUr Mineralogie-Monatshefte, 49-58 (in German). Werding, G. and Schreyer, W. (1990) Synthetic dumortierite, its PTX-dependent compositional variations in the system Al2 0 3 -B 2 0 3 -Si0 2 -H 2 0. Contributions to Mineralogy and Petrology 105, 11-24. --------- ( 1996) Experimental Studies on S orosilicates and Selected Borates. In E.S. Grew and L.M. Anovitz (eds), Boron. Mineralogy, Petrology and Geochemistry. Reviews in Mineralogy 33, Mineralogical Society of America, Washington, D.C., p. 117-159. Wright, PM., Weil, J.A., Buch, T., and Anderson, J.H. (1963) Titanium colour centers in rose quartz. Nature 197, 246-248. 148 TABLE I. Sample locations and descriptions. Sample ID Location, City GRR-1790a· Alto da Pitora. Galileia State. Country Type of deposit (if known) Minas Gerais. Brazil Pegmatite California, USA Weathered pegmatite GRR-1815 Madagascar Pegmatite GRR-1818 Namibia Pegmatite GRR-181 I Anza Valley GRR- 1819 Olivera dos Brejinhos Bahia, Brazil GRR-1820 Scott Mine, Custer South Dakota, USA Massive vein GRR-1821 Tobacco Root Mnts Montana. USA Massive vein GRR-1822 Ramshorn Gulch. Tobacco Root Mnts Montana, USA Massive vein GRR-1845 Miass Ural Mnts. Russia Pegmatite GRR-1855 Custer South Dakota, USA Pegmatite GRR- 1856c Shirley Meadows, Kern County California, USA Massive vein GRR-1864 Joai'ma Minas Gerais, Brazil GRR-1871 Obrazek Quarry. Pisek Czech Republic GRR-1874 Lavra Schupp. Joai 'ma Minas Gerais, Brazil GRR-1877 Bedford New York. USA Pegmatite GRR-1885 Madeleine. La Toma San Lui s, Argentina Massive vein GRR-1896 Gronoy Nordland. Norway Granitic pegmatite GRR-1897 Amdal Aust-Agder, Norway Granitic pegmatite GRR-1899 Sarniresy, Antsirabe Madagascar GRR-1900 Tsaramanga (Tongafeni), Mahaiza (Fivondronana) Betafo, Madagascar GRR-1903 Rossing Berge Namibia GRR-1904 Tammela Finland GRR-1905 Ilvafjord. Julianehab Fjord Greenland GRR-1906 Rabenstein Bayeni, Germany GR-7 Schinder Mine, Riverside County California. USA Granitic Pegmatite GR-8 Fcno Mine, Riverside County California, USA Granitic Pegmatite CJT-2998 Santa Cruz County Arizona. USA CIT-10416 Washoe Count:t Nevada, USA GRR-1861 Bahia, Brazil GRR-1907 Granitic Pegmatite Granitic Pegmatite Granitic Pegmatite China • Thi specimen is the single-crystal pink quartz. 149 TABLE 2. XRD d-spacings. ICDD 12-0270 Dumortierite (Dehesa, CA) GRR-1815 hkl 5.85 ( 100) 5.06 (90) 4.26 (50) 3.84 (50) 3.43 (60) 3.22 (60) 2.91 (60) 2.54 (20) 2.09 (80) 1.33 (50) 5.84 ( 100) 5.07 (85) 4.27 (60) 5.87 ( 100) 5.06 (85) 4.25 (50) 3.83 (45) 3.44 (85) 3. 15 (50) 2.90 (50) 130 140 3.83 (55) 3.43 (85) 3.21 (80) 2.90 (75) 2.54 (55) 2.09 (90) 1.34 (70) 2.08 (90) 1.33 (50) d-spacings are in Angstroms, relative intensity (brackets) in %. I I I 240 22 1 231 260 440 371 I 12 2 150 FIGURE CAPTIONS Figure 1. SEM backscattered-electron images of the purified fibrous residue obtained after dissolution of rose quartz. a) sample GRR-1820 (Custer, SD), low resolution image showing the large length-to-width ratio of the fibers; b) sample GRR1820, higher resolution of a separate cluster of fibers. Individual fibers in this strand are about 100 nm wide. Figure 2. Ff infrared spectra of a) the natural dumortierite (Dehesa, CA), and b) fibrous material from sample GRR-1815 (Madagascar) before and c) after the heating as described text. The intensity differences observed in the 1000 and 500-600 cm·1 regions between the standard and the fibers were observed in the spectra of all fibers examined. Figure 3. Raman spectra of a) dumortierite (Dehesa, CA) crystal, b) dumortierite (Dehesa, CA) acid-treated powder; and fibrous mats from the samples c) GRR-1811, d) GRR-1815, e) GRR-1861. Vertical dashed lines help to illustrate differences in the position of the bands. The spectra of fiber samples are nearly identical, whereas the apparent differences in the 350-600 cm· 1 region between the dumortierite standard and the fibers are observed. Figure 4. The optical absorption spectra of the extracted fibrous mat and corresponding crystal of rose quartz from GRR-1815 (Madagascar). The two spectra, offset vertically for clarity, both have an absorption band at about 500 nm. The spectrum of dumortierite from Dehesa, CA, is shown for comparison. 151 Figure Ia. 152 Figure lb. 153 3 ~--------------------------------------~ a 1 b c 2000 1800 1600 1400 1200 1000 800 600 400 1 Wavenumber (cm- ) Figure 2. 154 120 100 80 60 Raman Shift (cm- 1) 40 20 Figure 3. 155 1 0.8 (].) 0.6 ~ 0.4 0.2 , 0 400 500 I 600 700 800 900 1000 Figure 4 156 Appendix D. Fibrous nanoinclusions in massive rose quartz. Part 11: HRTEM & AEM investigations. This manuscript is to be submitted to the American Mineralogist, 2001. 157 Fibrous nanoinclusions in massive rose quartz. Part II: HRTEM & AEM investigations ChiMa, Julia S. Goreva, and George R. Rossman Division of Geological and Planetary Sciences California Institute ofTechnology, Pasadena. CA 91125, USA ABSTRACT Pink fibrous crystals within massive rose quartz from localities in California, South Dakota, Brazil, Madagascar and Namibia were examined with high-resolution transmission electron microscopy (HRTEM) and analytical electron microscopy (AEM). This study reveals that the nano-fiber in all samples are related to dumortierite. Selectedarea electron diffraction (SAED) patterns and HRTEM images indicate that the dumortierite-related fibers have a superstructure with a doubled repetition period along the .K- and)!- axes of dumortierite, giving cell parameters a= 2adum = 2.36 nm, b = 2bdum = 4.05 nm, c = cdum = 0.47 nm. Computer simulations suggest that periodic arrangements of two different M(l ) site occupation in the octahedral face-sharing chains give rise to the superstructure. One type of M( I) site is occupied mainly by AI, whereas the other type is dominated by Ti and Fe. Simulated structural images based on our proposed model match the experimental structural images. Most of the fibers, elongated along the £:-axis, are free of defects. AEM analysis shows that the chemical composition of the dumortierite-related fibers is similar to well characterized dumortierite. but contains a greater amount of Fe !58 substituting for AI in the M( I) sites. B was detected in all fibers examined by electron energy loss spectroscopy as well as in sillimanite crystals found as a minor component in one rose quartz from Brazil. Key Words: Fibers, rose quartz, borosilicate, dumortierite, superstructure, sillimanite, HRTEM, AEM, SAED, Image simulation. 159 INTROD CfiON Fine fibers in massive rose quartz have been observed and, since the 1930's, were generally presumed to be rutile (reviewed by Rossman, 1994, and Goreva et al, 2000). Applin and Hicks (1987) concluded that the fiber phase in rose quartz from a Montana locality is dumortierite based on their TEM and powder XRD analyses. However, their fibers· XRD peaks are not well matched with dumortierite·s peaks in many details. Our XRD on bulk fibers from rose quartz from a variety of localities also show similar results (Goreva et al., 2000). Significant differences observed in the IR and Raman spectra of these fibers, however, cast doubt on their identification as dumortierite. These differences were the motivation for a more detailed TEM study of the fibers. Most fibers in massive rose quartz are less than I 11m in thickness and ideal for TEM examinations (Fig. I). Previously the only published TEM results on those fibers were those of Applin and Hicks who showed that the c-axis repeat of the fibers from the Montana locality was appropriate for dumortierite. We present here a detailed analysis of the structural nature of indi vidual fibers based on HRTEM and AEM studies. EXPERIMENTAL PROCEDURES Samples Rose quartz samples were selected from Madagascar (G RR 1815), Olivera dos Brejinhos in western Bahia. Brazil (GRRI819). Joai'ma, Minas Gerais, Brazil (GRR 1864), Rossing (not stated in the original sample description as received, but inferred from the sample), Namibia (GRRI818), Scott Mine, Custer, South Dakota (GRRI820), and Schindler Mine. Riverside County. California (GRR 1996). All rose 160 quartz samples are from pegmatite deposits. Under routine examination at 400x with an optical microscope, mineral inclusions were observed in only two of the samples selected for this study. In these two samples numerous (GRRI864) and occasional (GRRI819) needle-like crystals up to I 0 )..tm in thickness are present. Fibers were extracted by dissolution ofthe crushed quartz in HF and then were concentrated and purified. More information on these specimens and fiber purification procedure is given in Goreva et al. (2000). For comparison, detailed, parallel studies were conducted on a well-characterized blue dumortierite standard from Saharina, Madagascar (GRR347), studied by Moore and Araki (1978). Electron diffraction patterns were also obtained from pink dumortierite from Dehesa, San Diego County, California (GRR122), and blue dumortierite from the Cargo Muchacho Mountains. Imperial County, California (CIT14110). Transmission electron microscopy A JEOL Atomic Resolution Microscope (ARM) operated at 800 kV and an In-situ JEOL 200CX TEM in the National Center for Electron Microscopy (NCEM), Lawrence Berkeley Laboratory (LBL), a Philips EM430 STEM operated at 300 kV at Caltech and an lSI TEM operated at 200 kV at the Jet Propulsion Laboratory were used for HRTEM imaging and/or electron diffraction analysis. AEM analysis was performed with a JEOL 200CX at the NCEM. LBL. The JEOL 200CX is equipped with a high angle energy dispersive x-ray spectrometer (EDS) and a Gatan 666 electron energy loss spectrometer with parallel detection (PEELS) capable of routinely achieving I.SeV resolution. The k 161 factors for the fiber quantitative EDS analysis were obtained using dumortierite, sillimanite. titanite and andradite as standards under the same analytical conditions. TEM specimens ofthe fibers were prepared in two ways: (I) crushed fiber particles dispersed on a carbon film supported by a copper grid so the elongation of fibers is almost parallel to the film , (2) sectioning with a diamond microtome to produce thin sections of the fiber normal to the electron beam. The dumortierite standard was prepared by dispersing crushed crystals on a carbon film. early I 000 photos were recorded during this study, allowing confidence that the observations are representative. HRTEM simu lations Computer simulations ofHRTEM images and electron diffraction patterns were preformed using the NCEMSS (version 1.6) software package provided by Dr. Roar Kilaas. The calculations were carried out with a Sun workstation. The input structure parameters for dumortierite (Pmcn, a= 1.1828 nm, b = 2.0243 nm, c = 0.4700 I nm) and sillimanite (Pbnm, a= 0 .7479 nm, b = 0.7670 nm, c = 0.5769 nm) were baed on Moore and Araki (1978) and Peterson and McMullan (1986), respectively. Two-dimensional structural images were simulated with the multi slice approach for a typical range of microscope defocus values, crystal thicknesses, and crystal orientations, using a set of optical constants appropriate for the ARM, Philips EM430 and JEOL 200CX micro copes. The simulated and experimental images are compared visually for interpretation. Figure 2 illustrates computed HRTEM images of the dumortierite structure at different orientation, thickness and defocus conditions. The simulations show that image details are highly sensitive to defocus, orientation and the calculated range of 162 specimen thickness. Hundreds of simulated images were generated in thjs study to assist in interpreting experimental atomjc images. RESULTS Electron diffraction Selected-area electron diffraction (SAED) patterns of the zone-axes were obtained from a number of single fiber crystals. Figure 3 shows representative zone-axis SAED patterns of single fibers from rose quartz, comparing with same zone-axis patterns of the dumortierite standard. The SAED patterns, obtained from different orientations of the fibers, are assigned to [100], [010], [110], [Ill] and [001] zone-axis electron ruffraction patterns based on a dumortierite structural model. The [010], [110] and [111] patterns were observed from the same area of a single fiber through tilting. The [100] zone-axis pattern is almost identical to the dumortierite pattern. However, the [II 0] and [Ill] zoneaxis patterns of the fibers show weak but relatively sharp reflections along the [1-10] * direction at _( 1-1 O)dum *. This indicates that the fiber has a periodicity along [ 1-1 0] * = 2d(l -l O)dum• which implies a superstructure of dumortierite with a doubled repetition period along both the,!- and the ~-axes (i.e., a super = 2adum = 2.38 nm; bsuper = 2bdum =4.05 nm). The [II 0] zone-axis pattern of a fiber can be easily mistaken for a [I 00]-zone pattern (Fig. 3A). However, the intensity distributions of the reflections are different. This superstructure was observed in the electron diffraction pattern of fibers from all localities examined. It was not observed in any of the electron diffraction patterns of the dumortierite samples. 163 Figure 4A contains a series of computed electron diffraction patterns from a welloriented dumortierite crystal with uniform I 0 nm thickness. The almost identical [I 00] zone-axis patterns from a fiber and from the dumortierite standard (Fig. 3) are not well matched with the simulated [1 00]-zone pattern of dumortierite (Fig. 4A) since dynamic diffraction causes the presence of forbidden reflections such as (00 1) and (OkO) where k = 2n+ 1 in the experimental SAED patterns. The [001] zone-axis SAED patterns of the fibers and dumortierite are also similar and are matched closely with the simulated pattern of dumortierite (Fig. 3 & 4A). SAED patterns show that nearly all fibers in rose quartz have a superstructure of the dumortierite structure, and are elongated along the ~-axis. Few [11/]-zone singlecrystal SAED patterns indicate that a very small number of fibers might have the dumortierite structure because there are no weak reflections along the [I I0] * direction at _(II 0)*. Such patterns, however, might be from a thin domain where the superstructure information has weakened and/or has been damaged by the electron beam. The zone-axis SAED patterns of reference dumortierite (GRR347) examined in this study reveal that the dumortierite crystal has the ideal structure proposed by Moore and Araki ( 1978) (e.g., Fig. 3B and 4A). These patterns are consistent with the simulated dumortierite electron diffraction patterns. Most SAED patterns of single fiber crystals are clear without diffusive streaking. This indicates that most fiber crystals have an ordered structure. Only a few electron diffraction patterns revealing the superstructure show weak streaking (e.g., Fig. 3A) along [ 1-1 0] *. The HRTEM images show that the streaking was not derived from stacking faults but more likely was from uneven cation occupations in the M( I). 164 Convergent-beam electron diffraction patterns were also obtained from some dumortierite-related fibers. Figure 5 illustrates a [010]-zone whole pattern of a fiber showing a weak First-Order Laue Zone (FOLZ) and a sharp Second-Order Laue Zone (SOLZ). The FOLZ reveals that the repeat along [010] is 4.05 nm (i.e., the periodicity of the fiber along the ~-axi s = 2bdum). HRTEM imaging Bright-field images at low-magnification reveal the fibers' morphology (Fig. I). In general, the fibers vary in thickness mainly from I 00 nm to 800 nm but can be up to 1200 nm. Two-dimensional structural images from a number of single fibers were obtained along with their zone-axis SAED patterns. HRTEM images indicate that the fibers are almost free of stacking faults, deformation and dislocations. HRTEM images (Fig. 6) of single fibers corresponding to the orientation of the [II 0] and [Ill ]-zone SAED patterns (Fig. 3A) show that the true periodicity along the [II 0] direction is 2.04 nm and the subperiodicity of 1.02 nm appears in the thin area. Figure 7 illustrates the [I 00) zone-axis images showing sub-periodicity of 2.02 nm along [0 I 0). Cross-sections of the fibers (Fig. 8) reveal (0 10), (I 00), (II 0) and ( 140) surface planes indexed according to the dumortierite superstructure. A single ( I I0) twin was observed in only one fiber. The irregular-shape of the surface layer (< 2 nm thick) of the fibers in Figure 68, is likely caused by electron beam damage. In general electron beam damage on the fiber phase is not serious although the knock-on damage plays a ro le in degrading its structural ordering. 165 Chemistry AEM analysis on individual fibers shows that in all samples the fibers from rose quartz are aluminum-borosilicates, having a chemical composition similar to dumortierite. Mean structural formulae of the fibers are given in Table I . EDS analysis reveals that the fibers contain a small amount ofTi and Fe substituting for AI. The Fe content varies slightly. The contents ofTi and Fe are slightly higher than in blue dumortierite (Fig. 9), but overlap the range of these elements in pink dumortierite (Alexander et at. 1986). In the fibers from rose quartz, Ti content is generally greater than Fe content. Ti and Fe are believed to occupy the M( I) sites in dumortierite (Piatonov et at. 2000). The composition of the fibers is plotted together with the compositions of dumortierites in Figure 9. FTIR analyses on bulk, extracted fibers, including those used in this study, reveal that borate is present in about the same B to Si ratio as in dumortierite (Goreva et at. 2000). PEELS analysis reveals that B is present in each single fiber from rose quartz (Fig. I OA). Although attempts to obtain reproducible quantitative PEELS analyses were, in part, frustrated by the irregular shape and thickness of the individual sam ples, these analyses also show that the B content ofth~ fibers is comparable to that of dumortierite (Fig. I OB). Other fibers and phases In several recent books and articles, rutile fibers (Ti0 2) are still considered to be the cause of color in rose quartz. A control experiment with light ye llow rutilated quartz (CIT 5585, Minas Gerais, Brazil) demonstrated that fine rutile needles ( 1.5 jlm wide- 166 identity verified by Raman spectroscopy) survive our acid dissolution process intact. Thus, if rutile fibers were present in rose quartz in significant numbers, we should detect them. However, rutile fibers were not found in any of our rose quartz samples. Along with the dumortierite-related fibers, a significant proportion of acicular sillimanite nanocrystals (- I 0%) were admixed with the fibers extracted from one sample of Brazilian rose quartz (GRR 1864). Figure II illustrates a HRTEM image of a sillimanite crystal with its [11 0] zone-axis SAED pattern, which is closely matched to the calculated image and electron diffraction pattern. The composition ofthe sillimanite crystals is almost identical to AbSi0 5 (Table 1). PEELS analysis shows that boron occurs in the sillimanite (Fig. I OC). The B content appears not to be caused by contamination of dumortierite-related fibers. or inclusions in sillimanite of borosilicates with 803 groups. This is consistent with a prior observation of boron in macroscopic sillimanite (Grew and Rossman 1985). Similar, but larger (up to 15 J.lm wide) acicular AbSi0 5 crystals were observed as a minor component in the SEM images of fibers extracted from samples GRR 1819 and GRR 1820 of Goreva et at. (2000) and were presumed to also be sillimanite. The intensity of the boron Kedge in the PEELS spectrum ofthe sillimanite nanocrystals is visually as evident as it is for dumortierite (Fig. I 0). However, this does not necessarily indicate that B content of the sillimanite needles is as high as it is in dumortierite because the PEELS spectra were obtained under different conditions (e.g., crystal thickness, orientation). DISCUSSION 167 Experimental TEM results in this study indicate most fibers from rose quartz are a superstructure of dumortierite with a doubled repetition period along both the !- and the .1::-axes. The superstructure can be modeled based on a periodic variation of the M(l) position in the dumortierite structure. The ideal dumortierite structure consists of three chain types (each built of AI06 octahedra) which run parallel to the ~-axis: x1 [M(1 )0 3] 1 face-sharing chain, x1 [M(4) 4012) double chains, and x [M(2)2M(3)20 12] chains (Moore and Araki 1978). The M(2), M(3) and M(4) sites are occupied totally by AI. The M(l) site is occupied mainly by AI and cation vacancies in a 3: I ratio in normal dumortierite (Moore and Araki 1978), and by Mg in magnesiodumortierite (Chopin et al. 1995). The AEM analyses of individual fibers show the presence ofTi and Fe (Table I). These two cations are assigned to theM( I) site in dumortierite (Alexander et al. 1986). We suggest that there are two types of cation occupations in the M( I) site; they are Type 1: AI, Type II : Ti + Fe. On this crystal-chemical basis, a model for the superstructure is constructed, as illustrated in Figure 12. In this model, viewed down the [00 I] direction, the Type II site is surrounded by Type I sites along the [I 00), [ 11 0) and [0 I 0] directions of the dumortierite structure. The superstructure derives from the regular distribution of these two types of cation occupations in the M(l) site. In the superstructure unit, three quarters of the M( I) sites are Type I (AI) and one quarter are Type II (Ti and Fe). The stoichiometry of this model, [AI 6(AI 0 75) 1(Ti,Fe) 11 025 )Si3B0 1s, comes close to predicting the experimentally determined composition of the fibers in Table I. The model does not exclude minor 168 amounts ofTi and Fe in the Type I AI position nor does it exclude cation vacancies in either the Type I or Type II cation sites. The simulated electron diffraction patterns of this model model (Pmcn. a = 2.3656 nm, b = 4.0486 nm, c = 0.4700 I nm) (Fig. 4B) match well with the experimental SAED patterns of the dumortierite-related fibers (Fig. 3A). The simulated HRTEM images also closely match the experimental images of the fibers (Fig. 13). These simulations indicate that the superstructure occurring in the fibers indeed has such a cation configuration. The streaking along the [II 0] direction in some SAED patterns of single fibers might correspond to some cation disordering in theM( I) sites. The chemistry of the fibers (Table 1) shows an average A I o.66 Ti o 18 Fe 0. 13 per formula in the M(l) site, which suggests that about 20% of the Ti and Fe are located in the Type I position with AI. It is likely that uneven distribution of those Ti and Fe in the Type I position give rise to the streaking along [1-10]* on some ofthe fibers· SAED patterns. It is apparent that Ti and Fe in the Type II position of the M( I) site are disordered along the ~-axis , whereas AI in the Type I position is ordered in the M(l) site or partly disordered when minor Ti, Fe or vacancies are present in the Type I position along with AI. This model predicts that in the fibers c = Cdum = 0.47 nm without a doubled or tripled repetition, as is observed in the e xperimental electron diffraction results. If AI. Ti and Fe are randomly distributed in the M( l ) site along z*. the dumortierite structure occurs. This might be the case for a few fibers whose [II/] SAED patterns are almost identical to these of ideal dumortierite crystals. Our model indicates that the pink color of the fibers results from intervalence 4 charge transfer between Fe2+ and Ti - in the M(J) site. A similar mechanism has 169 previously been proposed as the cause of color of dumortierite (Alexander et al., 1986; Platonov et al., 2000). Although the fibrous borosilicate phases observed by HRTEM may seem volumetrically insignificant in rose quartz (Goreva et al. 2000), they are crucial to the very definition of rose quartz. It is possible, that the fine acicular sillimanite crystals extracted from GRR 1864 are cogenetic with the comparably sized dumortierite-like fibers. This would indicate that both phases were formed by exsolution from the quartz. However, formation of larger sillimanite crystals microscopically observed in GRRI864. and in samples GRR1819 and GRR 1820 of Goreva et al. (2000) were most likely incorporated through crystal capture during the growth of the quartz. Ironically, these microscopically visible needles that captured the attention of many researchers appear to have no relationship to the origin of color of rose quartz. These observations raise the question of whether the ordering we observe in the fibers is a result of growth at a lower temperature than usual for dumortierite. It is reasonable to expect that exsolved phases might display higher ordering than their directly crystallized counterparts because lower temperature formation generally leads to higher order. In our case, exsolution presumably occurred at a temperature below 400°C. In this respect, it would be interesting to find a locality with multiple generations of dumortierite formed at different temperatures and examine the state of ordering in the mineral. Co CLUSJONS 170 Borosilicate fibers exist in massive rose quartz from vein deposits and pegmatites worldwide and they are the cause of the pink color of rose quartz (Go rev a et al. 2000). These sub-micrometer wide, pink fibers are related to dumortierite. Nearly all of them have a previously uncharacterized superstructure of dumortierite with a doubled repetition period along both the ;r-and the ~-axes. Computer simulations show that the superstructure is due to an ordered arrangement of two types of cation occupations on the dumortierite M(l) site. We propose that the color ofthe fibers is due to Fe2+ - Ti 4"'" intervalence charge transfer in one of the M( 1) sites. ACKNOWLEDGMENTS We thank Ulrich Dahmen and Christian Kisielowski for allowing us to access the ARM and the JEOL 200CX microscopes at the National Center for Electron Microscopy. Lawrence Berkeley National Laboratory, Roar Kilaas for helping with the NCEMSS image simulation program, Chengyu Song for assisting with the ARM operation, and Thomas George for allowing us to use the lSI 200 kV HRTEM at the Jet Propulsion Laboratory. Lothar Jung (Quartz Technology, Inc.), Rock H. Currier (Jewel Tunnel Imports) and Edward Swobota (Beverly Hills. CA) provided quartz samples used for this study. Funding for this project was provided by the White Rose Foundation. 171 REFERENCES CITED Alexander. V.D., Griffen, D.T., and Martin, T.J. (1986) Crystal chemistry of some Feand Ti-poor dumortierites. American Mineralogist, 71 , 786-794. Applin, K.R., and Hicks, B.D. ( 1987) Fibers of dumortierite in quartz. American Mineralogist, 72, 170-172. Chopin, C .. Ferraris, G .. lvaldi, G., Schertl, H.-P., Schreyer, W., Compagnoni, R., Davison, C .. and Davis. A.M. (1995) Magnesiodumortierite. a new mineral from very-high-pressure rocks (western Alps). II. Crystal chemistry and petrological significance. European Journal of Mineralogy, 7, 525-535. Goreva, J., Ma, C., and Rossman, G.R. (2000) Fibrous nanoinclusions in massive rose quartz. Part I: The source of rose coloration. (Submitted to the American Mineralogist). Grew, E.S .. and Rossman. G.R. ( 1985) Co-o rdination of boron in sillimanite. Mineralogical Magazine, 49, 132-135. Moore. P.B.. and Araki, T. (1978) Dumortierite, SbB[AI6 1s[]o250 172s(OH)o.7s]: A detailed structural analysis. Neues Jahrbuch fur Mineralogie Abhandlungen, 132, 231-241. Peterson. R.C., and McMullan. R.K. ( 1986) Neutron diffraction studies of sillimanite. American Mineralogist, 71. 742-745. Platonov, A. ., Langer, K., Chopin. C., Andrut. M .. Taran, M. . (2000) Fe2+-Ti 4+ charge transfer in dumortierite. European Journal of Mineralogy. 12, 521-528. 172 Rossman, G.R. (1994) Colored varieties ofthe silica minerals. In: Silica Physical Behaviour, Geochemistry and Materials applications. Heaney. P.J., Prewitt, C.T. and Gibbs, G.V. (editors). Reviews in Mineralogy, 29, 433-467. 173 Table I. Mean chemical formula of the fibers in rose quartz and reference dumortierite by AEM (EDS) analysis. Sample No. Locality Dumortierite-related fibers • ------------------ ----·------------------------- ---------------------------------------------- GRR1815 Madagascar (Al 6.71 Ti o.15 Fe o.l3) Si 3.00 B 01 8 GRRI818 Namibia (Al 6.64 Ti 0.11 Fe o 15) Si 3.02 B 01 8 GRR1819 Bahia, Brazil (AI 668 Ti 0.11 Fe o.l3) Sh o1 B 01 s GRRI864 Minas Gerais, Brazil (AI 6.58 Ti o.25 Fe oo9) Si 302 B 01 8 GRR1996 Schindler Mine, CA (Al 6.68 Ti o.11 Fe o 13) Si 301 B 01 s Reference Dumortierite • GRR347 Saharina, Madagascar (A1 6.86 Ti o06 Fe o06) Si 302 B 01 8 Sillimanite microcrystals t GRRI864 Minas Gerais, Brazil • Formulas are normalized to B0 18. t ormalized to 0 5. AI 1.98 Fe o.o2 Si 1.00 0 5 174 Table 2. Representative cell parameters Cell parameters dumortierite • fibers (GRR 1815) ao 1.1828 nm 2.36 nm 2.3656 nm bo 2.0243 nm 4.05 nm 4.0486 nm ~ OA700l nm OA7 nm OA7001 nm * From Moore and Araki ( 1978) superstructure model 175 Figure Captions Figure 1. Low-magnification bright field image showing the dispersed crushed fibers on a carbon film. The fibers are from a rose quartz from Madagascar (GRR1815). Figure 2. Simulated HRTEM images of dumortierite with different orientations and at different thickness and defocus conditions using optical constants for the Philips EM430. (A) [ 100]-zone, (B) [ 11 0]-zone. Figure 3. (A) Representative zone-axis SAED patterns of single fibers obtained from different orientations: [I 00], [0 I 0], [II 0], [Ill] and [00 I). (B) Zone-axis SAED patterns ofthe dumortierite standard from [100], [110], [Ill) and [001). A ll indices are based on the dumortierite structure determined by Moore and Araki ( 1978). Figure 4. (A) Simulated electron diffraction patterns of dumortierite in perfect orientations with uniform I 0 nm thickness from [1 00]. [0 I 0], [II 0], [Ill] and [00 I). (B) Simulated electron diffraction patterns of the superstructure model (see Discussion) with uniform I 0 run thickness from [ 100], [0 I 0], [ 110], [ 112] and [00 I). Figure 5. The [0 10]-zone whole pattern of a fiber from GRR 1819 showing the FirstOrder Laue Zone (FOLZ) and the Second-Order Laue Zone (SOLZ). Figure 6. HRTEM images of single fibers showing 2.04 nm real-periodicity along the [II 0]* direction and 1.02 nm subperiodicity at a thinner area. (A) The [II 0]-zone. (B) the [I 12]-zone. Figure 7. HRTEM images of single fibers down the [100] zone-axis showing only subperiodicity along the [0 I 0] direction. That is d(020)super = b dum · 176 Figure 8. [00 1] zone-axis HRTEM images of a fiber from GRR 1815 showing its cross section. Figure 9. Octahedral cation contents in the fibers from rose quartz (this study), classic dumortierite (Alexander et al. 1986) and magnesiodumortierite; Chopin et al. 1995). Figure 10. PEELS spectrum of (A) a dumortierite-related fiber (GRR1819), (B) the dumortierite standard (GRR347), and (C) a sillimanite nanocrystal (GRRI864). The boron K edge is evident in all spectra. Figure 11. (A) A thin sillimanite crystal from GRRI864. (B) [110] zone-axis HRTEM image of the sillimanite crystal from GRR 1864 and its SAED pattern. Figure 12. The superstructure model (a= 2adum, b = 2bdum, c = Cdum)· The superstructure is strongly psuedohexagonal, with ?. the pseudo hexad axis. There are two types of M(l) chain: Type I filled with AI and Type II occupied by Ti and Fe. Figure 13. HRTEM images of single fibers viewed down the (A) [112] and (B) [II 0] directions, matched with the inserted simulated images using the superstructure model. 177 aoonm Figure I 178 IOwn 15nm (A) ·TOO nm · 120 nm -lou run -120 nm Ulllmn Thi~ .Snm I ~ run (Bl Fig ure 2 179 (A) Fiber Figure 3 (B) Dumortierite [I 00] ,. • f' • • • • • • • oo:::(I • • • [01 0] [ I I 0] • • • • • [I II] • • • ' • •' • f".'. . • ••+ • • ! '" 1 • • • ' • . • ., • .. "' • .. i • • • "' • .. [001] • • • • • .. • • • • C.::' ) • • .. • .. 180 Figure 4 (B) Superstructure (A) Dumortierite .... . . " . . . . ....... ...._...... · · · - · · · - -.. . .. . ~ • • .s.ao• . . . .. ... .. ..... ... ~ . ........ ~ t 't • +• ... . - . • .._*••. · · ···· ····-···-··---····· · ·--::..' ••• . ....................... ............ ..···h· ~ ... .. ......... .................. .... ... ............. •·•• .. •·••a:••-•••-•••..••--•••••a•••• • • ... • ···-········ ............. [100] + •· ~ ~ .. + •.. • • • .. .. "'.. • 4 • . • • • • ~ ~ .• " .. • • • • • •+- + • ..+ "« .. •.. • • [010] • •.• · · ···++• ·4· ····~ • " • ~ .,~ •• [II 0] . .. .. . . . . . . . . .. .. .. . . . .... .. .. - .. . . . . . . " . . .. -~ . ~ •• !lr' .. ....... -~·"! ·· ·"t · ·~ . ... ~ [Ill] [ 112] .. [001] t • • ~ • ... - . . .. ....... . "" ~ ., .. • • IF 181 Figure 5 182 Figure 6 183 Figure 7 184 A 30nm 4nm Figure 8 185 '1~-~-~----~-. I 0 (). s . AI (VI) p r. LL Figure 9 186 (A) . 250 1- . I ! l Fiber in rose quartz - OiL : • 100 . 200 300 400 Energy Loss (eV} ·--·-·- 500 600 (B) - 300 1Dumortierite 0 8 zso - - ',_ B (K ) '.. I o ~·--~--~~~~~ -~ ~-- = ~ ~-==~====~~~T-~·-~--=-~-~~-~- 100 200 300 400 Energy Loss {eV) 500 600 Figure IOa,b 187 (C) l 3001- Sillimanite - 0 0 0 250 - 2200 c ;J 0 (.) 150t- - 50 - - ol\ .I , 100 200 Energy Loss {eV) Figure JOe 188 A 300nm ,, -, - - -- - Figure II 189 Figure 12 190 Figure 13