Carbon Dioxide in Synthetic and Natural Silicate Glasses Citation Fine, Gerald Jonathan (1986) Carbon Dioxide in Synthetic and Natural Silicate Glasses. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/xmqc-xv31. https://resolver.caltech.edu/CaltechTHESIS:09032019-144830666 Abstract Infrared spectroscopy has been used to study the speciation of CO 2 in both synthetic silicate glasses quenched from melts held at high temperatures and pressures and in natural basaltic glasses. In glasses near the NaAlO 2 -SiO 2 join, absorption bands resulting from the antisymmetric stretches of both molecular CO 2 (2352 cm -1 ) and CO 2- 3 (1610 cm -1 and 1375 cm -1 ) are observed. The latter are attributed to distorted Na-carbonate ionic-complexes. Molar absorptivities for each of the absorption bands have been determined; these molar absorptivities allow the quantitative determination of species concentrations in sodium aluminosilicate glasses with a precision on the order of several percent of the amount present. The accuracy of the method is estimated to be ±15-20% at present. The ratio of molecular CO 2 to CO 2- 3 in sodium aluminosilicate glasses varies little for each silicate composition over the range of total dissolved CO 2 contents (0-1.5%), pressures (15-33 kbar) and temperatures (1400-1560°C) studied. This ratio is, however, a strong function of silicate composition, increasing both with decreasing Na 2 O content along the NaAlO 2 -SiO 2 join and with decreasing Na 2 O content in peraluminous compositions off the join. The molar absorptivities determined for sodium aluminosilicate glasses have also been used to measure the concentrations of CO 2 in albitic (NaAlSi 3 O 8 ) glasses quenched from melts equilibrated with CO 2 vapor at high pressures (15-30 kbar) and temperatures (1450-1625°C). The results show that total CO 2 solubility increases with increasing pressure at constant temperature. Both molecular CO 2 and CO 2- 3 concentrations increase with pressure. At constant pressure, the solubility of molecular CO 2 decreases with temperature while the concentration of CO 2- 3 increases. The net effect is that total CO 2 solubility is not significantly dependent on temperature, decreasing slightly with increasing temperature at constant pressure. The speciation of CO 2 in both synthetic Ca ± Mg-composition glasses and natural basaltic glasses contrasts with the case of CO 2 -bearing sodium aluminosilicate glasses. CO 2 is inferred to be dissolved in these glasses as distorted Ca- or Mg-carbonate ionic-complexes that result in unique infrared absorption bands at 1515 cm -1 and 1435 cm -1 . The molar absorptivities for each of these absorption bands were also determined. No detectable molecular CO 2 is dissolved in these glasses. Infrared spectroscopic measurements of species concentrations in glasses provide insights into the molecular level processes accompanying CO 2 solution in melts and can be used to test and constrain thermodynamic models of CO 2 -bearing melts. CO 2 speciation in silicate melts can be modelled by equilibria between molecular CO 2 , CO 2- 3 , and oxygen species in the melts. Consideration of the thermodynamics of such equilibria can account for the observed linear relationship between molecular CO 2 and carbonate concentrations in sodium aluminosilicate glasses, the absence of molecular CO 2 in Ca ± Mg silicate glasses, the proposed linear relationship between total dissolved CO 2 content and the activity of CO 2 in melts, and observed variations in CO 2 solubility in melts. Dissolved CO 2 contents of natural basaltic glasses can also be determined from the intensities of the carbonate absorption bands at 1515 cm -1 and 1435 cm -1 . The uncertainty of the method is estimated to be ±15% of the amount present. The infrared technique is a powerful tool for the measurement of dissolved CO 2 contents in natural basaltic glasses since it is nondestructive, can be aimed at regions of glass a few tens of micrometers in size, and discriminates between dissolved carbonate and carbon present as carbonate alteration, contained in fluid inclusions or adsorbed on the glass. A set of submarine basaltic glasses dredged from a variety of locations contain 0-400 ppm dissolved CO 2 , measured using the infrared technique. These concentrations are lower than most previous reports for similar basaltic glasses. No general relationship is observed between dissolved CO 2 content and depth of magmatic eruption. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: Geology Degree Grantor: California Institute of Technology Division: Geological and Planetary Sciences Major Option: Geology Thesis Availability: Public (worldwide access) Research Advisor(s): Stolper, Edward M. Thesis Committee: Rossman, George Robert (chair) Stolper, Edward M. Wyllie, Peter J. Ahrens, Thomas J. Burnett, Donald S. Epstein, Samuel Defense Date: 15 November 1985 Funders: Funding Agency Grant Number NSF EAR-8212765 NSF EAR-8417434 Record Number: CaltechTHESIS:09032019-144830666 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:09032019-144830666 DOI: 10.7907/xmqc-xv31 Related URLs: URL URL Type Description https://doi.org/10.1007/bf00377759 DOI Article adapted for Chapter 1. Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 11789 Collection: CaltechTHESIS Deposited By: Mel Ray Deposited On: 04 Sep 2019 23:31 Last Modified: 16 Apr 2021 22:23 Thesis Files Preview PDF - Final Version See Usage Policy. 42MB Repository Staff Only: item control page

CARBON DIOXIDE IN SYNTHETIC AND NATURAL SILICATE GLASSES Thesis by Gerald Jonathan Fine In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California 1986 (Submitted November 15, 1985) ii Acknowledgments The research presented in this thesis was completed in collaboration with Professor Tom Tombrello, Dr. Marcus Mendenhall, Dr. Rogerio Livi, and Thomas Johnson. Their help is appreciated. I have benefitted from the support and cooperation of many people while at Caltech. I am especially grateful to Roger Aines, Professor Art Boettcher of U.C.L.A., Professor Don Burnett, Phil Ihinger, Stephanie Mattson, Paula Rosener, Martin Ruzek, Henry Shaw, and Lynn Silver. Professor George Rossman has been a continual source of both knowledge and advice. I sincerely thank him for his time and energies. Finally, I thank my thesis advisor, Professor Edward Stolper, who has been a good advisor, a g ood teacher, and a good friend. This research was supported by NSF grants EAR-8212765 and EAR8417434. iii Abstract Inf rared spectroscopy has been used to study the speciation of C0 2 in both synthetic silicate glasses quenched from melts held at high temperatures and pressures and in natural basaltic glasses. In glasses near the NaAl02 -Si02 join, absorption bands resulting from the antisymmetric stretches of both molecular co2 (2352 cm- 1 ) and co~- (1610 cm- 1 and 1375 cm- 1 ) are observed. The latter are attributed to distorted Na-carbonate ionic-complexes. Molar absorptivities for each of the absorption bands have been determined; these molar absorptivities allow the quantitative determination of species concentrations in sodium aluminosilicate glasses with a precision on the order of several percent of the amount present. The accuracy of the method is estimated to be ±15-20% at present. The ratio of molecular C0 2 to CO~- in sodium aluminosilicate glasses varies little for each silicate composition over the range of total dissolved co 2 contents (0-1.5%), pressures (15-33 kbar) and temperatures (1400-1560°C) studied. This ratio is, however, a strong function of silicate composition, increasing both with decreasing Na 2 o content along the NaAl02-Si02 join and with decreasing Na 20 content in peraluminous compositions off the join. The molar absorptivities determined for sodium aluminosilicate glasses have also been used to measure the concentrations of co 2 in albitic (NaA1Si 3 08 ) glasses quenched from melts equilibrated with C0 2 vapor at high pressures (15-30 kbar) and temperatures (1450-1625°C). The results show that total C0 2 solubility increases with increasing pressure at constant temperature. Both molecular C0 2 and CO~- iv concentrations increase with pressure. At constant pressure, the solubility of molecular co2 decreases with temperature while the concentration of CO~- increases. The net effect is that total C0 2 solubility is not significantly dependent on temperature, decreasing slightly with increasing temperature at constant pressure. The speciation of C0 2 in both synthetic Ca ± Mg-composition glasses and natural basaltic glasses contrasts with the case of sodium aluminosilicate glasses. co 2 -bearing co2 is inferred to be dissolved in these glasses as distorted Ca- or Mg-carbonate ionic-complexes that result in unique infrared absorption bands at 1515 cm- 1 and 1435 cm- 1 • The molar absorptivities for each of these absorption bands were also determined. No detectable molecular C02 is dissolved in these glasses. Infrared spectroscopic measurements of species concentrations in glasses provide insights into the molecular level processes accompanying co 2 solution in melts and can be used to test and constrain thermodynamic models of co 2-bearing melts. co 2 speciation in silicate melts can be modelled by equilibria between molecular C0 2 , oxygen species in the melts. co 23 , and Consideration of the thermodynamics of such equilibria can account for the observed linear relationship between molecular co 2 and carbonate concentrations in sodium aluminosilicate glasses, the absence of molecular C0 2 in Ca ± Mg silicate glasses, the proposed linear relationship between total dissolved activity of melts. co2 content and the co2 in melts, and observed variations in co2 solubility in v Dissolved C0 2 contents of natural basaltic glasses can also be determined from the intensities of the carbonate absorption bands at 1515 cm- 1 and 1435 cm- 1 • The uncertainty of the method is estimated to be ±15% of the amount present. The infrared technique is a powerful tool for the measurement of dissolved C0 2 contents in natural basaltic glasses since it is nondestructive, can be aimed at regions of glass a few tens of micrometers in size, and discriminates between dissolved carbonate and carbon present as carbonate alteration, contained in fluid inclusions or adsorbed· on the glass. A set of submarine basaltic glasses dredged from a variety of locations contain 0-400 ppm dissolved technique. co 2 , measured using the infrared These concentrations are lower than most previous reports for similar basaltic glasses. between dissolved No general relationship is observed co 2 content and depth of magmatic eruption. vi Preface This thesis includes condensed versions of four published or soonto-be published papers listed in the references: Fine and Stolper (1985ab) and Fine et al. (1985ab). Some revision or rewriting of each manuscript has occurred and new text has been added. vii Table of Contents Acknowledgments ••••••••••••••••••••••••••••••••••••••••••••••••••••••• i i Abstract ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• i i i Pref ace ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• vi Introd.uction ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 1 Chapter 1. The Speciation of Carbon Dioxide in Sodium Aluminosilicate Glasses ••••••••••••••••••••••••••••••••••• 6 'Experi111ental Techniques ••••••••••••••••••••••••••••••••••••••• 7 Experimental Results ••••••••••••••••••••••••••••••••••••••••• 17 Band Assignments ••••••••••••••••••••••••••••••••••• 17 Calibration of Band Intensities Versus Species Concentrations •••••••••••••••••••••••••• 36 Precision and Accuracy ••••••••••••••••••••••••••••• 42 Speciation of Carbon Dioxide in Sodium Aluminosilicate Glasses ••••••••••••••••••••••••• 50 Discussion ••••••••••••••••••••••••••••••••••••••••••••••••••• 63 Comparison with Previous Spectroscopic Investigations •••••••••••••••••••••••••••••••••• 63 Solubility Mechanisms of Carbon Dioxide in Sodium Aluminosilicate Melts ••••••••••••••••• 69 Conclusions •••••••••••••••••••••••••••••••••••••••••••••••••• 7 2 viii Chapter 2. The Solubility of Carbon Dioxide in Molten Albite ••••••••• 74 Experimental Techniques •••••••••••••••••••••••••.•••••••••••• 76 Sample Synthesis ••••••••••••••••••••••••••••••••••• 76 Infrared Spectroscopy of C02 -Saturated Glasses ••••••••••••••••••••••••••••••••••••••••• 80 Results . ........•..........................•...............•. 86 Comparison with Previous Results ••••••••••••••••••••••••••••• 98 Conclusions ..... ........•.....................•.............• 99 Chapter 3. The Speciation of Carbon Dioxide in Ca ± Mg Silicate Glasses .... .......•...•....•..••.......•.....•.........•. 100 Synthesis of C02 -Bearing Glass Standards •••••••••••••••••••• 101 Results ........... ........•................................. 109 Band Assignments and Intensities •••••••••••••••••• 109 Band Intensity Calibrations ••••••••••••••••••••••• 125 Conclusions . •................................•.............. 13 2 Chapter 4. Thermodynamic Considerations ••••••••••••••••••••••••••••• 134 Speciation of C0 Solubility of C0 2 2 in Silicate Melts ••••••••••••••••••••••••• 135 in Silicate Melts ••••••••••••••••••••••••• 145 Water Versus co2 Speciation•••••••••••••••••••••••••••••••••l49 Conclusions .....................••...•................•..... 151 ix Chapter S. The Concentration of Carbon Dioxide in Natural Basaltic Glasses ••••••••••••••••••.•••••••••••••••.•••••• 153 Results .............•.•••..........•.••.•••.••.....••.....•• 160 Conclusions . ..••......••........•••••••.•....••..•••....•.•. 17 2 Conclusions ...•.........•.....•..•...••.•..••..•....•....•........... 17 3 Appendix 1. Measurement of the Carbon Content of Silicate Glasses Using the 12 C(d,p 0 ) 13 c Nuclear Reaction ••••••••• 176 Experimental Procedure •••••••••••••••••••••••.•••••••••••••• 178 'Results •••..•.•..•••......•..•...•••••••..•••••••••••....•. • 184 Conclusions .............••.•..•••..•........•............... l 9 3 References • •.•.....................................•.......•.•....... 19 4 1 Introduction 2 Stimulated by the important role played by carbon dioxide in the petrogenesis of a wide range of igneous rocks, particularly alkaline basic magmas (e.g., Eggler, 1973; Wyllie, 1979), there has been considerable recent interest in the solubility mechanisms of co silicate melts. 2 in Two contrasting approaches to understanding these solution mechanisms have been pursued. The first is an indirect approach that uses systematic variations in solid-liquid phase equilibria to infer structural changes that occur in melts when they dissolve C0 2 • For example, the liquidus volume of pyroxene, a polymerized phase, enlarges relative to that of olivine, an unpolymerized phase, when C0 2 is added to melts. This has been used to infer that addition of co2 results in increased polymerization of melts (Eggler, 1974, 1978; Kushiro, 1975). Unfortunately, this intuitively satisfying line of reasoning has given, at best, only qualitative insights. The second approach used to understand co involves spectroscopic techniques. 2 solution mechanisms Both infrared (Mysen et al., 1975; Brey and Green, 1975, 1976; Mysen, 1976; Brey, 1976; Mysen et al., 1976; Eggler et al., 1979) and Raman (Verweij et al., 1977; Sharma, 1979; Sharma et al., 1979; Mysen and Virgo, 1980ab; Rai et al., 1983) spectroscopy have been used to study co2-bearing silicate glasses quenched from melts to elucidate aspects of the speciation of co2bearing melts. These spectroscopic techniques can provide direct, quantitative information on how, at a molecular level, carbon dioxide is dissolved in silicate glasses and how it influences the structure of the silicate framework. In particular, infrared and Raman spectroscopy can 3 be used to distinguish between carbon dioxide that is dissolved as carbonate (CO~-) ions and that which is dissolved as molecules of C0 • 2 Dissolved molecular C0 2 and CO~- each have unique infrared signatures, distinct from co 2 contained in vesicles, carbonate occurring in any crystalline form, or other carbonaceous material. Infrared spectroscopy can thus discriminate between dissolved and extraneous forms of C0 2 and allow the accurate identification and measurement of the concentrations of dissolved species in both natural and synthetic glasses. Spectroscopy is also a non-destructive technique that does not require fusion of the glass being analyzed, eliminating the possibility of reactions between C species during heating. An infrared beam can be directed at spots as small as tens of microns, avoiding crystals and vesicles, and can be used to examine sample homogeneity. These qualities make infrared spectroscopy a potentially powerful tool for the investigation of dissolved co2 in natural and synthetic glasses. Most recent spectroscopic investigations have not taken full advantage of the quantitative potential of either infrared or Raman spectroscopy. Most investigators have simply noted the existence of absorption or emission bands and thus of particular species in the glasses; the intensities of the bands have not generally been used to determine the concentrations of the absorbing or emitting species. Exceptions include the work of Verweij et al. (1977) in which the potential for using Raman spectroscopy to determine the concentrations of CO~- groups in iso-Si0 2 -co2 glasses was discussed and the work of Sharma and co-workers in which Raman spectroscopy was used to determine relative concentrations of CO~- ions as a function of pressure and 4 temperature in glasses of sodium melilite, diopside, and akermanite composition (Sharma, 1979; Sharma et al., 1979). In addition, Mysen (1976) used infrared spectroscopy to estimate co~- concentrations in glasses on the NaAl02 -Si0 2 -co2 join. In that study, Mysen concluded that the relative proportions of molecular co2 groups and carbonate groups in co2-bearing glasses are strong functions of silicate composition and of the pressures and temperatures from which melts are quenched to glasses. In this thesis, I report the results of experiments designed to develop more fully the potential of infrared spectroscopy for quantitative determinations of the concentrations of carbon-bearing species in silicate glasses. In Chapter 1, molar absorptivities (or extinction coefficients) are determined for absorption bands due to molecular C02 and carbonate in glasses near the NaAl02 -Si02 -co2 join. These molar absorptivities are used to measure the concentrations of molecular co2 and carbonate in sodium aluminosilicate glasses. Using this information, the dependence of the speciation of co2 on sodium aluminosilicate glass composition, on total dissolved co2 content, and on the pressure and temperature of equilibration of the melt from which the glass was quenched is investigated. In Chapter 2, infrared spectroscopy is used to determine the concentrations of total dissolved co co 2 2 in albitic (NaA1Si 0 ) glasses quenched from melts equilibrated with 3 8 vapor at a variety of pressures and temperatures. In the next chapter, it is demonstrated that infrared spectroscopy is also a useful tool for the identification and measurement of dissolved co 2 species in a variety of synthetic Ca ± Mg silicate glasses, including basaltic 5 compositions. I show that the speciation of C0 2 in these glasses is similar over a wide range of glass compositions, yet different from the speciation of co 2 in NaA1Si 2o6 -Si02-co 2 glasses. Chapter 4 presents a simple thermodynamic treatment to guide thinking about the dependence of co 2 speciation and solubility on melt composition. This treatment provides insights into the solubility mechanisms of co2 in silicate melts, into the thermodynamics of co2- bearing silicate melts, and into the effects of dissolved co2 on the phase equilibria of silicate melts. Finally, in Chapter 5, the applicability of these results to the study of natural basaltic glasses is demonstrated and the concentration of to.tal dissolved co 2 in a variety of ocean floor basalts is measured. 6 Chapter 1. The Speciation of Carbon Dioxide in Sodium Aluminosilicate Glasses 7 The experiments described in this chapter were designed with three immediate goals: first, to develop infrared spectroscopy as a quantitative analytic tool for the measurement of the concentrations of species in NaA1Si 2o6 -Si0 2-co 2 glasses; nature of co 2 second, to understand both the co2 solution in fully or nearly-fully polymerized glasses and the effects of changing silicate composition on the speciation of third, to establish a framework for the thermodynamic model of co 2 ; co2 solution discussed in Chapter 4. Experimental Techniques Powdered silicate starting materials were prepared at three compositions along the NaA10 2-Si0 2 join: NaA1Si 2o6 (_j_~_), NaA1Si 3 0 8 (_~~) and NaA1Si0 4 o10 (a composition close to the 1 atm. albite-silica eutectic designated as~). Glasses of .ii. composition (batches jd-D and jd-E) were synthesized by grinding Na 2co 3 , Al 203 and Si02 (JohnsonMatthey SpecPure reagents) in agate under ethanol for 6 hours followed by melting at 1580°C for 12 hours in air at 1 atm. of Si0 E. 2 Appropriate amounts glass were then added to jd-E to prepare batches ab-E and ~ These decarbonated mixtures were ground under ethanol for 6 hours and dried at 850°C in air at 1 atm. for 2 days to remove any hydrocarbon residue or adsorbed co2. When mixtures not dried in this fashion were used as starting materials for high pressure experiments, co 2 was detected spectroscopically in the run products even when no loaded into the charges. significant Na-loss. co2 was Drying times of more than 1 week resulted in Some Na-depleted compositions were used as 8 starting materials; these are designated jd~X, ab-X, and eu-X. All starting materials were stored over dessicant after drying. Selected glasses of each composition were analyzed for major elements in order to verify glass homogeneity and to determine the extent of Na-loss during the drying process. Analyses were obtained using an automated MAC-5 electron microprobe with a 15 KV operating voltage, a sample current of 5 nA on brass and a 40-45 micron spot size (Table 1). The results of these analyses are shown graphically in Figure 1. The starting materials were loaded along with known amounts of silver oxalate (Ag 2 c2 o4 ) into Pt capsules, which were then sealed by arc-welding. These Pt capsules were then loaded into larger Pt capsules containing a dry hematite-magnetite buffer or sintered hematite. The amounts of silver oxalate and silicate were weighed on a Cahn electrobalance to 0.0005 mg precision out of 0.0750-20.0000 mg total and were chosen to produce vapor-undersaturated melts. The silver oxalate was analyzed by gas manometry in the laboratory of Professor s. Epstein and found to be stoichiometric to within 5% of its nominal total co2 content. It was stored in a light-proof container over dessicant to minimize decomposition and H 0-adsorption. 2 The double capsules were held at constant pressure (15-33 kbar) and temperature (1400-1560°C) in a 0.5 inch solid media piston-cylinder apparatus using a NaCl and pyrex cell. used: A hot piston-in technique was runs were brought cold to a pressure 5 kbar below the desired pressure, the temperature was raised to the final temperature and then the final pressure was applied. Temperature was monitored using W3%Re - 9 Table 1- Analyses of glass compositions used. are described within the text. Microprobe techniques Starting Batch jd ideal jd-D jd- E jd-X ab ideal ab-E ab-X eu ideal eu-E eu-X Na 2o 15. 33 14.52 13.81 12.66 11.82 11. 28 10.80 9.62 9.80 9.25 Al 2 0 3 25.22 25.47 24 .BS 24.80 19 .44 19. 34 19 . 28 15.83 15.07 15.93 Si0 2 TOTAL 59.45 60.10 61.43 62.42 68.74 69.66 69.81 74.55 74 . 80 75.46 100.00 100.09 100.09 99.88 100.00 100. 38 99 .89 100.00 99.67 100.64 12 4 6 7 7 6 3 I of mialyaes ....... 0 11 Figure 1- Graphical representation of the starting compostions used in this chapter. Ideal stoichiometric compositions are marked by X. The fields show the range of microprobe analyses, while the solid circles show the average of all analyses for each sample. 12 ID .o.., O _N <% .Q ~ I ON xI xI xI 0 I I • (/) UJ I ., ;:, '«? "o q,s UJ I .ll ~ o .'11.,() ~ 0>10 ~ :2. 0 ~I 10 ~I -~ 13 W25%Re thermocouples. Pressure was calibrated against the reaction: ab = .i! +~at 29.5 kbar and 1100°C (Holland, 1980) and a -7% correction applied. Upon heating, silver oxalate is presumed to decompose according to the reaction: Agz C2 04 = ~ + 2C02. This results in melts containing co2 • Run times of 1 hour or more were disseminated Ag and dissolved needed to produce co2 homogeneity and uniform species concentrations in the glass run products. Runs were quenched by turning off the power to the furnace; quench rates were 150-200°C/sec. At the end of most runs, the contents of the outer capsule were examined and both hematite and magnetite found to be present. co2 -free glasses of each composition were synthesized at high pressures by running both unwelded and welded silver oxalate-free inner capsules. Infrared spectra of these glasses did not differ from the infrared spectra of glasses of the same compositions prepared at 1 atm. in air. After quenching, the inner capsules were sectioned and doublypolished plates of the glasses were prepared. In most samples disseminated Ag metal gave the glass in the bottom of the capsule an orange color or blueish-orange streaked appearance. The thickness of each glass plate was measured using a micrometer or dial indicator. The plates were then placed over metal apertures 150-1000 microns in diameter and their transmission infrared spectra were obtained. Spectra discussed in this chapter were gathered on a Perkin-Elmer 180 infrared spectrophotometer in the linear absorbance mode (base ten logarithm) over the frequency range 4000 cm- 1 to 1200 cm- 1 • The sample chamber was purged with high purity N2 while each spectrum was recorded. Spectra 14 were recorded on chart paper and digitized with a Tektronix 4052 computer. The spectra of anhydrous, decarbonated glasses of each composition were then computer-subtracted from these spectra. Both absorption intensities and integrated absorption intensities were determined from the background-subtracted digital data. The precision of absorbance data obtained from the PE 180 is estimated to be about 0.01 (lcr) at an absorbance of 0.5 (Stolper, 1982a). Many of the glasses synthesized early in this study were not suitable for calibration of absorption band intensities versus species concentration because the final step in the drying procedure was not performed and the glasses were dark, presumably containing either graphite or a hydrocarbon residue. In these instances, total dissolved carbon contents were presumed not to be accurately known. The infrared spectra of all these glasses were still taken and data obtained from these samples, discussed below, were still useful. After infrared spectra were taken, some sodium aluminosilicate glasses were independently analyzed for total carbon using the 12 c(d,p ) 13c nuclear reaction. 0 This technique, used by Filleux et al. (1977), Oberheuser et al. (1983), and Mathez et al. (1984), is discussed more fully in Appendix 1. It allowed the independent determination of total C contents (within 5% relative certainty) of the same pieces of glass on which spectra were taken. These analyses show that after 1 hour long runs, the amount of C analyzed after the run was generally within 10% of the amount loaded into the capsule (Table 2). Two glasses compositionally off the NaA1Si o -Si0 join were also 2 2 6 used in this study. The first,a 56% Na 2 o, 43% Si0 2 (wt.%) glass was 15 Table 2- Comparison of wt. % C0 2 loaded into Pt capsules containing sodium aluminosilicate glass with total carbon (reported as wt.% co 2 ) analyzed using the 12 C(d,p 0 ) 13 c nuclear reaction. conditions are given in Appendix 1. Run 16 Sample II CO(, Loaded wt. %) C02 Analyzed (wt. %) ABC-S3 0.67 0.69 ABC-S4 0.38 0.32 ABC-SS 0.20 0 .19 ABC-S8 0.40 0.36 ABC-62 0.94 0.88 ABC-63 o.oo 0.07 ABC-64 0.7S o. 77 JDC-111 0 .17 0.20 JDC-112 O.S4 o.ss JDC-116 0. 56 o.ss NC-lS o.ss o.s2 NC-17 0. 2S 0. 28 NC-19 0.43 0.47 17 equilibrated with co 2 vapor at 1 atm., 1100°C for 24 hours. The second, a 30% Na 2 0, 10% Al 2 o3 , 60% Si02 glass was equilibrated with 0.31 wt.% co2 at 1450°C, 25 kbar for 1 hour. A few glasses were prepared using starting materials consisting of C02 -free glass and Na 2 co 3 enriched with 90% l3c as a C02 source rather than Ag 2 c2 o4 • These glasses were enriched in Na and compositionally not on the NaAl0 2 -Si02 join. co2 -bearing ~and .ii_ glasses prepared in the laboratory of Professor A. Boettcher were also used in the study. were prepared using Na 2 co 3 or Ag 2 c2 o4 as They co2 sources and were run at a variety of pressures and temperatures. Densities used in this study were estimated from the data of Kushiro (1978); the presence of disseminated Ag or bubbles (Chapter 2) in the run products precluded direct density measurements of most glasses. It is assumed that the effect of dissolved C0 2 on glass density is not significant. Experimental Results Band Assignments The spectra of a C0 2 -free and a C0 2 -bearing jadeite glass are shown in Figure 2. Four strong absorption bands are present in the C02 bearing glass at frequencies of higher than 1200 cm- 1 that are not present in the co2 -free glass. Although their relative intensities vary with co 2 content and glass composition (Figures 3, 4, and 5 and Table 3) the locations of these four bands are essentially identical in all of the co -bearing NaAl0 -Si0 glasses examined. 2 2 2 At wavenumbers lower than 18 Figure 2- Infrared spectrum (solid line) of a jadeite glass (JDC-84) containing 0.80 wt. % total dissolved spectroscopy. co2 as measured by infrared Band assignments are given. The dashed line is the spectrum of a co 2 -free jadeite glass synthesized at 1 atm. The silicate background has not been subtracted from either spectrum, hence the spectra go offscale at approximately 1200 cm- 1 due to aluminosilicate vibrations. WAVELENGTH,, 3 ,ym 4 5 6 7 1.4 C02 1.2 w u z <( 1.0 m 0.8 (/) 0.6 O'.'.: 0 m <( c 03-2 ...... -OH r-~/ 0.4 0.2 4000 ------3500 3000 --------2500 WAVENUMBER,, cm I // 2000 -1 1500 I "' 20 Figure 3- Background subtracted spectra, from 2500 - 2200 cm- 1 and all normalized to 100 microns sample thickness, of four jadeite composition glasses containing variable amounts of total dissolved C0 2 • Sample numbers and total dissolved co2 contents are shown. The band at 2352 cm- 1 increases in a regular fashion with increasing total dissolved co2 • 3.5 3.0 w u 2.5 z ca 2.0 <{ Ct: 0 (/) ca <{ 1.5 t 1.0 ~ 0. 5 E I /\. \ j ~I/\\ \._JDC-108 JDC-116 J '- JDC-113 JDC-Ill 2500 WAVENUMBER~ 2300 cm -t 0.96% 0.56% 0.39% 0.17% N ...... 22 Figure 4- Background subtracted spectra, from 1800 - 1200 cm- 1 and normalized to sample thicknesses of 100 microns, of the same jadeite glasses shown in Figure 3. contents are shown. Sample numbers and total dissolved C0 2 The bands at both 1610 and 1375 cm- 1 increase with increasing total dissolved co2, yet their intensities are always approximately equal (1.17 absorbance 1610 = absorbance 1375 ). 23 I e 0 CS) CS) LI) Ck:' w m ~ :J z w > <( 3 N . lJ) - - 3~NV8~0S8V . lJ) CS) 24 Figure 5- Background subtracted spectra, from 2600 to 1200 cm- 1 , of C0 2 -bearing silicate glasses of four different compositions, jd-E, ab-E, eu-E, and eu-X. All of the spectra are normalized to constant intensity of the 2352 cm- 1 band. The intensities of the carbonate absorption bands increase with increasing Na content and decreasing silica content. 108.) (Samples shown are NC-10, NC-19, ABC-53, and JDC- WAVELENGTH;' .ym 5 4 7 6 2.5 w u 2.0 z <{ m 1.5 I (!'.: 0 t (/) m 1 . 0 t- <{ I.I /\ Al~ jd-E Ill I~ ab-E I \ 0.sµ/\~ eu-X 2600 2300 2000 WAVENUMBER;' 1700 -1 cm 1400 1 N \JI 26 Table 3- Data obtained for Chapter 1. dur•t1on St•rt1ng S.apl• ' ec.p. p (kber) T (•c) (atn.) NC NC NC NC 2 eu-! eu-1 eu-1 eu-1 2S 2S 25 25 25 25 25 25 25 25 25 25 25 25 H 14SO 14SO 14SO 14SO 14SO 14SO 14SO 14SO 14SO 14SO 14SO 14SO 1450 14SO 14SO 14SO 14SO 14SO 1450 1450 1560 1560 1450 14SO 14SO 14SO 14SO 14SO 14SO 14SO 1400 1400 1400 1400 1400 1400 30 60 60 60 60 67 60 60 60 60 60 60 30 60 60 60 60 60 60 60 MC NC NC NC NC NC NC NC ABC UC ABC UC ABC UC ABC UC AIC AIC AIC AIC AIC AIC AIC AIC ABC ABC JDC JDC JDC JDC JDC JDC ) 4 6 7 10 13 15 17 19 20 21 14 22 21 24 25 27 29 30 n JS ll 54 II 16 17 18 62 64 67 68 69 72 7l 74 ...-. •u-1 eu-1 eu-1 eu-l eu-1 eu-l •u-1 ab-I •b-• ab-I ab-! ab-I ab-! ab-! eb-1 ab-I eb-! ab-! ab-I ab-X ab-X •b-1 ab-I ab-I •b-1 Jd-D Jd-D jd-D jd-0 Jd-D Jd-D )) H 27 IS 15 25 25 25 20 20 20 20 20 20 20 21 21 25 25 21 21 lO 30 60 60 60 60 60 60 60 60 30 30 30 30 30 30 OO(v1:L.oald)•d 1.81 c t. lS c 0.4S c 0 . 90 c 0.60 c l.2S c 0.00 0.55 0.2S 0.4) 0.20 0. 34 c 0.6S c 1. 83 c O.S2 c 1.01 c 0.64 c 1.02 c 0. 78 c 0.6S c 0.44 c 0.6) O. ll 0.20 0.49 0.97 0.40 0.94 • 0.75 0.99 c o.s8 c 1.64 c o. )) c 0.26 c 0.11 c (vHt2.0~) 0.)7 0.41 0.29 0.42 0.26 0. )9 0.14 0.24 0.18 0.21 0.11 0.)9 0.10 0 . 21 0.65 0.22 0.19 0. )9 0 . 86 0 . 49 0 . 21 0 . 21 0.14 0.16 0.16 0.30 0.29 0.11 o. ) j 0.11 0.26 o. )9 0.42 0.)7 o.so 0.60 0.nettyb ThtckneH C1/ll (ca) 2440 2440 2440 2440 2440 2440 2440 2440 2440 2440 2440 2440 2130 2530 2530 2S30 2S30 2510 2440 2440 2510 2530 2510 2510 2510 2SIO 2110 2HO 2110 2110 2600 2600 2600 2600 2600 2600 o.ooss 0.004) 0.0058 0.006) 0.0067 0.0043 0. 0 167 0 . 0074 0.008) 0.0044 0.0073 0.0075 0.0200 0.0060 0.0026 O.OOS6 0.0037 0 . 0046 0.0030 0.0036 0.0046 0.0067 0.0045 0.0011 0.0081 O.OOSI 0.0040 0.0070 0.0036 0.0018 0.001) 0.0073 o.oon o.ooao 0.0062 0.0011 lend Locatlon ..... 2150 2353 2153 235S 2152 2153 2154 21S4 2153 2)55 2154 2153 21SS 2356 21S4 21SS 2ll5 2)5) 2154 2352 2154 2154 2113 2113 2113 2156 2lS4 2112 2113 2ll2 211) 2111 2113 2111 2ll2 21Sl 1. 80 0.11 0.18 0.96 1.06 I. 30 0.19 1.n 1.08 1.04 0.15 1.46 0.0) 1.31 1.29 0.72 1.11 0.41 0.31 0.41 0.11 o.n 1.0 0.11 0.41 1.06 0.97 1.01 I. l7 1 :as 0.76 0.57 0.16 0.44 0.34 0.26 -· ln1:e1r. 67 .96 4.19 24.14 )5.0l l0.92 31.01 4.11 31.11 29.51 27.79 22.21 31.92 0.14 44.70 )7 .11 21.11 34. 34 10.6) 1.12 11.11 21.97 21.17 36 . 60 18.91 11.6) 28.)) 2s.21 21.46 36.49 49.55 19.52 ll.56 20.92 9.61 1.10 l.16 lend Locatlon 16S7 16SO 16S6 16SO 1660 1652 d d d d d latear. Abe. 0.09 0.01 0.04 0.01 o.o5 o.06 ••• 11.11 1.44 4.02 1.26 5.00 ..,, 1379 1424 1351 1341 1407 1))0 d d d ..... ••• la1:e1r. 0.11 0.02 0.04 O.OI 0.01 0.05 6.60 2.41 4.42 7 .55 l.00 5.06 0.16 0.16 0.12 0 . 11 0.05 0.01 0.05 0.09 0.13 0.13 0 . 01 0.10 0.06 0.09 0.09 0.10 0.11 0.21 0.12 0.11 0.24 0.24 0.11 10.90 11.s1 I.SI 11.21 •• • d 1620 1614 1620 1611 1619 1606 1611 1611 1622 1625 1626 1622 1619 166) 1610 1606 1616 1605 1601 1607 1604 1601 1606 lead Locattoa 0.14 0.16 0.09 0 . 15 o.os o.os 0.04 0.06 0.01 0 .01 0.04 0.02 0.04 0.06 0.06 0.08 0.13 0.24 0.29 0.12 0.21 0.21 0.15 ll.91 21.22 11.10 20.U •••• 5. lO 4.15 l.54 1.51 1.21 4.42 2.41 l.11 l.15 a.14 10.SO 11.11 21.n l0.00 lS.31 21.61 21.22 16.lS d d 1313 ll72 1)72 1)11 ll7J llll 1312 1))4 1))6 1))4 1311 1380 1))1 1314 lll4 1315 1379 1)11 llll ll15 1]71 1311 1314 J.13 l.IO 4.ll •.n 11.41 10.36 5.21 1.62 5.21 6.81 6.1) 6.51 11.22 20.05 22.14 24.)9 16.60 11.60 14.36 N --.J S.aple JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC , 76 84 85 86 89 90 91 101 108 110 Ill 112 113 11• 116 llO!T2582 I IO!T2990 I IOIT2995 I Stantnc ec.p. P (kbar) T (°C) Jd-0 jd-0 jd-0 jd-0 jd-0 jd-0 jd-0 jd-K jd-1 jd-1 jd-1 jd- 1 jd-1 Jd-1 jd-1 Jd 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 20 20 20 1400 1450 1450 1450 1450 1450 14SO 1450 14SO 1450 14SO 1450 1450 1450 1450 1450 1450 100 •b •b durat too (•la.) lO 10 lO lO 30 30 lO 60 60 60 60 60 60 60 60 15 360 60 ro(vt'-:'•x•>•• 1!'.<>"1> 0.00 c I. 56 c 0.14 c l. ll c 0.12 c 1.21 c 0.12 c o . 33 0.96 0.00 0.17 0.54 o. 19 0.66 0.56 SlT. I 1.00 h 1.00 0.65 0.35 0.24 0.10 0.12 o.n o. )() 0.28 0.11 0.11 0.12 0.14 0.18 0.16 0.12 0,41 0.49 0.20 0.n.itJb ThickneH (1/l) (c•) 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2600 2510 2510 0.0049 o.oou 0.0089 0.0055 0.0027 0.0052 0.0041 0.0066 0,0041 0.0104 0,0115 0.0139 0.0059 0.0012 0.0109 0.0018 0.0053 0.0012 .... Locetton 2lS2 2352 2352 2353 2l55 2153 2148 2153 2154 d 2)54 2154 2353 2354 2l51 2154 2354 2355 -· ..... la tear. 0.23 0.11 o. 31 0,79 0.31 0 .66 0,05 0.65 0.91 4.66 19.64 10.13 22.01 10.64 16.76 1.80 17.62 24.11 o. 34 1.23 0.51 0.46 1.11 0.63 1.12 1.22 9.16 36.61 12.90 ll.47 31.11 14.05 28.76 35.21 .. Loc•tlon Aha. 1643 1611 1612 1610 1612 1613 1625 1623 160t d 1607 1601 1617 1622 1607 1595 o.05 -· O.lO 33.36 IUI O.OI 1612 lot . . r. 0.11 0.36 0.11 0.21 0,04 O.OI 0.33 0.16 0,60 0.19 0.20 0,54 o.n o.oa 4.13 ll,t5 31.lt 21.0 31.42 5.05 1.57 31.52 19,U H.51 21.71 25,60 61.12 36,05 12.57 10.61 .... ..... t.ocauoo 1397 lllO 1371 1312 1376 1311 1371 1313 1374 d 1378 lllO 1372 llll lllO 137t 1363 136t 0.12 0 .31 0.17 0.31 0.11 0.2' 0.04 0.05 O,lt -· lat. .r. 7.13 u.21 11.54 u.21 13,17 I0.2' 2.10 2.lt 27.H 0.11 10.45 0.64 O.lt 43.91 o.n 0.60 o.u o.ot o.ot 12,t4 11.51 42.42 IO.t5 6.42 6. 11 H 0 contenu deter•ined uelna the .-thod of Stolper ( 1982a) and an Htlnctlon coefficient of 80 Uten/.ole-c:a. . c d 2 DendtiH ctttwtr obtalMd or utrapohted from luehlro ( 1978). CbHH wn dtirk. preeu . . bly contafntaa arepblte MMt other r.duced carbon •P•ci••· land lnunetty too la..r to charectertse • Cl••• conuln• • fe" ... n bubbl••· Sa•plH provided by ProfeHor A. Soettcher . Vapor eeturated. Na,co 5 "8ed .. a 00 aourcc rether then A& 2 C,~. Tout loaded c:o cooteot believed uJnown due to loot run lenat (a .. ten). 2 110 thet total C02 contente ere unknown. N CXl 29 1200 cm- 1 , the vibrations of the aluminosilicate networks comprising these glasses result in absorption bands that go offscale. The broad band at approximately 3550 cm- 1 is due to the stretching of -OH groups (Stolper, 1982a) and was used to quantitatively measure the amount of H2 o dissolved in each sodium aluminosilicate glass using a molar absorptivity of 80 liters/mole-cm. Water contents of the C0 2 bearing glasses quenched from melts at high pressure typically range from 0.2 wt. % to 0.5 wt. % H2 0. These levels of dissolved water are believed unavoidable . in experiments of this type, independent of drying procedure or the use of a double-capsule, buffered assembly. Water concentrations are higher in co2-bearing glasses than in those without C02 • This suggests that adsorbed H2 0 on the silver oxalate and/or hydrogen that diffuses through both the buffer and Pt capsule wall and reacts with co2 to produce H2 0 and reduced carbon species during the experiments are responsible for most of the water in these nominally anhydrous runs. Absorptions due to dissolved hydrocarbons or CO were not observed in any spectra. If reduced carbon is produced by reaction of hydrogen and dissolved co2 in the experiments, it is either not detectable via infrared spectroscopy (e.g., graphite) or diffuses out of the Pt capsules (Watson et al., 1982). Preliminary experiments conducted in our laboratory and in the laboratory of Professor Boettcher have indicated that run lengths of 3 or more hours are necessary to result in significant depletions in spectroscopically measured total dissolved co2 • If substantial carbon is lost or substantial reduced carbon is produced in shorter runs, it could modify some of the conclusions presented here. However, graphite is not 30 observed in the run products and both the hydration of the glasses and obvious depletions of the glasses in spectroscopically observable carbon have been minimized by run lengths of only 1 hour. It is emphasized that despite the most rigorous precautions, water seems to be unavoidable in these piston-cylinder experiments. In fact, water contents seem to be largely independent of drying procedures and even the experiments conducted by Professor Boettcher using special precautions to ensure dry conditions (Boettcher et al., 1981) contain dissolved water. Even C-free runs contain water. The only consolation is that our laboratory is able to precisely measure water concentrations. The sharp band at 2352 cm- 1 is attributed to the v 3 antisymmetric stretch of molecular 12 co 2 (Mysen et al., 1975; Mysen et al., 1976; Mysen, 1976; Brey, 1976). molecular For several reasons it appears that this co2 is homogeneously dissolved at molecular scale, probably as weakly bound units located in holes or cages in the overall melt structure, rather than present as trapped fluid inclusions. The intensity of this band increases in a regular fashion with increasing total co 2 content (Figure 3). Most glasses are bubble free when viewed at high power (500X) with an optical microscope and their total dissolved co2 contents are generally much lower than published co2 solubilities (Mysen, 1976). Finally, the location of the band is slightly shifted from that of gaseous 1978). co 2 at 2349 cm- 1 (Nakamoto, When bubbled glasses are intentionally prepared and run, a double band suggestive of both gaseous co2 in bubbles and dissolved co2 31 is observed. Other arguments suggesting structurally bound co , 2 including SEM observations, are presented by Mysen et al. (1976). The v 3 antisymmetric stretch of a free, undistorted carbonate group (n 3h symmetry) results in a degenerate absorption band at ~1415 cm-1. Distortion of the CO~- group (to Czv on Cs symmetry) results in a loss of the degeneracy and the appearance of two bands (White, 1974; Nakamoto, 1978). Following Brey and Green (1975), the broad bands at 1610 cm- 1 and 1375 cm- 1 are attributed to the v carbonate groups. 3 stretch of distorted Sharma (1979), Sharma et al. (1979), and Mysen and Virgo (1980a) assigned analogous bands in other glass compositions the same way. The constancy of the splitting of the v 3 mode (~235 cm- 1 ) of carbonate ions in the glasses studied suggests that the local environment of the carbonate group does not vary much over the range of Na 2 o, Al 2 o3 , Si02 , and co2 contents of the glasses. The degree of splitting observed is much larger than that generally caused by symmetry lowering in crystalline carbonates (White, 1974), but Nakamoto (1978) notes that either a unidentate (Cs) or bidentate (Czv) carbonate complex (a carbonate ion coordinated to a metal atom) could have approximately the required splitting. The details of the observed splitting presumably provide important information on the local environment of CO~- ions in the glasses, but I am unable to completely decipher these details at this time. Nevertheless, the large splitting is consistent with previous suggestions that carbonate ions are coordinated with sodium atoms in sodium aluminosilicate glasses (Mysen and Virgo, 1980b). That the local carbonate environment is almost certainly 32 dependent upon, if not dominated by, the charge-balancing or network modifying cations of the melt (Holloway et al., 1976) is confirmed by the fact that the v3 bands of co~- groups in a wide range of co2-bearing glasses with dominantly divalent network modifying cations, including basaltic glasses, glasses in the system CaO-Al o -Si0 , and diopside and 2 3 2 aRermanite glasses (Rai et al., 1983; Sharma, 1979; and Sharma et al., 1979; Chapter 3) are split almost identically, but quite differently from the Na-bearing aluminosilicate glasses. Although it is suggested that carbonate is present in the glasses as distorted sodium carbonate ionic-complexes, aluminum probably plays an important role in the local environment of carbonate groups. This is suggested by the fact that the v 3 carbonate bands of the Al-poor Na 2 0Al2o3-Si02 and Na 2 0-Si02 glasses examined are, as in crystalline Na 2 co 3 , not split to the extent observed in this study, if at all. One extremely intense band attributed to dissolved co~- is observed at ~1400 cm- 1 in the spectra of these glasses. The precise nature of the interaction between aluminum and sodium in the distortion of carbonate complexes and in influencing the relative proportions of molecular co2 and carbonate in our glasses cannot be addressed with these results and will require study of a wider range of silicate compositions on the Na 2 0-Al 2 o 3 -sio2 -co 2 join. Other interpretations of the bands at 1610 cm- 1 and 1375 cm- 1 can be found in the literature. Often, the high energy band has either not been recognized or not assigned to any C-bearing species (Mysen et al., 1975; Mysen, 1976; Mysen et al., 1976). Mysen (1976) and Mysen et al. (1976) also suggest that in some instances, these bands may be due to 33 Rai et al. (1983) conclude that the analogous doublet in rn2 bearing diopside glass is due to two distinct carbonate sites in the glass. None of these suggestions can be readily reconciled with the observations that although the intensities of both bands increase with increasing total dissolved C0 2 (Figure 4), their intensities are approximately equal to each other (1.17 absorbance 1610 regardless of total dissolved = absorbance 1375 ) co2 content, silicate composition, or total dissolved H2 0 (Figure 6). It is also noted that the doublets in the diopsidic glasses of Rai et al. (1983), in the akermanite glasses of Sharma et al. (1979), and in basaltic glasses and glasses in the system CaO-Al 2 o3 -Si02 -co2 (Chapter 3) mimic to within a few wavenumbers the splitting observed in the infrared spectrum of distorted c~- groups in scapolite. Scapolite contains only a single carbonate site (Papike and Stephenson, 1966; Chamberlain et al., 1985), rather than two carbonate sites or Hco; groups. Table 3 presents the quantitative results obtained during this study, including the run conditions of all syntheses, the exact locations, intensities and integrated intensities of all of the aforementioned absorption bands, and the thickness and density of each sample studied. Three other bands of low intensity were also observed during the course of this study. The band at 1610 cm- 1 is distorted due to the presence of an unassigned, broad, low intensity absorption band centered at ~1680 cm- 1 • Another band, at approximately 3710 cm- 1 , appears to vary directly with the intensity of the 2352 cm- 1 band. It is tentatively attributed to a combination of the v 1 (Raman active only) + 34 Figure 6- Intensity of the carbonate absorption band at 1375 cm- 1 versus the intensity of the carbonate absorption band at 1610 cm- 1 , both normalized to 100 micron sample thickness, for all of the samples studied. indicated. The silicate composition of each glass is The intensities of both bands are approximately equal (1.17 absorbance 1610 composition, total = absorbance 1375 ) regardless of silicate co 2 concentration, dissolved H2 0 content, or the pressure and temperature from where the glass was quenched. 35 - N x :2.I -x W1 . ..Q I :Q. 0 -- cr w w 0 --- q 000 d E I 0 . I I "'O ..Q ·0 :::3 <1.> <.O 0 N (X) r") en T ~ u U5 U5 U5 co LO r<> <! <! <! 0 0 0 0 0 <l d I'- zzz 0 0 0 ~ w . (/) 0 u z <t ~ CD a:: 0 0 ~o 0 0 OCD DO 0 <t v. 0 0 0 0 ~ d N d 0 d <l q O'> d CO d ~ d ~ ~ V ~ d d d d 1 _w~ 0191 3~N'i8CIOS8'1 N d 0 d 36 v 3 modes of molecular C0 2 • The intensity of a small band at 2287 cm- 1 also varies directly with the intensity of the 2352 cm- 1 band. This band is assigned to the v 3 antisymmetric stretch of molecular 13co 2• Both of these bands are used in Chapter 2 for the determination of molecular C0 2 concentrations in glasses enriched in this component. All of the band assignments attributed to C-bearing species are consistent with the spectra of 13 co -enriched glasses (Figure 7). 2 The band at 2352 cm- 1 is offset to 2287 cm- 1 as expected (Nakamoto, 1978, p. 116), and the bands at 1610 cm- 1 and 1375 cm- 1 are lowered in frequency 50 cm- 1 and 35 cm-1 respectively. The unassigned band at 1680 cm- 1 is also offset ~so cm- 1 , indicating that it too is due to a C-bearing species. Calibration of Band Intensities Versus Species Concentrations The Beer-Lambert law expresses the relationship between absorption band intensity and the concentration of either c where co2 or co~-: 44.01 x absorbance 1 x£ d x p (1) c is the weight fraction of either co2 or co~- (expressed as a weight fraction of C0 2 ), 44.01 is the molecular weight of C02 in g/mole, absorbance (base ten logarithm) is expressed in absorbance units, d is the sample thickness in cm, p is the sample density in g/liter, and £ is the molar absorptivity (or extinction coefficient) in liters/mole-cm. If integrated absorbances are measured instead of absorbances, the integrated molar absorptivity (£*in liters/mole-cm2 ) is substituted for 37 Figure 7- Spectra of 12 c- (top) and 90% 13 c- (bottom) enriched jadeite glasses. The lowered ratios of the intensities of the molecular co bands to the intensities of the carbonate bands in the 13 c enriched glass are due to the additional Na2 o added to the glass. 2 38 E <D 0 0 T I I" r- E ' ~ (.9 z _J I u ' w --- 0 0 0 N Lf) w > <( w en L: ::::) z w > <( 3 0 0 (V) N 3JNV8CiOS8V 3 39 molar absorptivity in equation (1). The total amount of carbon dissolved in a glass as molecular carbon dioxide and carbonate can be readily determined via infrared spectroscopy using equation (1) provided that the molar absorptivities of the band due to molecular co2 and one of the carbonate bands are known. Molar absorptivities (and integrated molar absorptivities) for the 2352 cm- 1 , 1610 cm- 1 , and 1375 cm- 1 bands are listed in Table 4 and were determined by the following procedure: An equation was written giving total dissolved co2 as the sum of dissolved molecular co2, based on the absorbance of the 2352 cm- 1 band, and of co 2 dissolved as carbonate, based on the absorbance of the 1610 cm- 1 band, for each of the 20 sodium aluminosilicate glasses for which total accurately known (see Table 3). co2 contents were believed to be The molar absorptivities of the 2352 cm- 1 and 1610 cm- 1 bands were then determined based on these 20 equations using a least-squares method similar to that given by Albarede and Provost (1977) by which errors could be assigned to all of the parameters given in equation (1) and to the best fit molar absorptivities. Three glasses (NC-15, ABC-57, and JDC-114) were anomalous, probably due to either errors in sample weighing or co 2 loss during synthesis, and were ultimately rejected from the fitting procedure. The best fit value of the ratio of the extinction coefficients of the 1610 cm- 1 and 1375 cm- 1 bands was determined from the data shown in Figure 6 and the molar absorptivity of the 1375 cm- 1 band given in Table 4 is based on this ratio. The ratios of the integrated absorbances to absorbances for each of these bands were also 40 Table 4- Calculated molar absorptivities and integrated molar absorptivities for infrared absorption bands due to C-bearing species. Band (cm- 1 ) Assignment Molar Absorptivitya (liters /mole-cm) Integrated Molar Absorptivitya (liters/mole-cm2 ) 2352 molecular co2 945 ± 45 25200 ± 1200 1610 c~- 200 ± 15 24100 ± 1900 1375 co2- 235 ± 20 16800 ± 1500 3 a Errors in these reported values were calculated assuming uncertainties of ±10% total loaded co , of ±2% estimated density, of ±0.02 2 absorbance units, and of ±3 microns sample thickness. To the extent that reduced carbon species are present in these glasses, these absorptivities have been underestimated. -I'- ...... 42 determined by least squares and the integrated molar absorptivities listed in Table 4 are based on these ratios. The procedure for determining molar absorptivities is based on two critical assumptions. First, molar absorptivities are assumed to be constant over the range of compositions studied, and second, all dissolved carbonate. co2 is assumed to be present as either molecular co2 or The internal consistency of these assumptions can be tested by comparing total co 2 concentrations, obtained by summing spectro- scopically determined molecular co 2 and carbonate concentrations, with the amounts of C02 originally loaded into the capsules. Figure 8 shows that the spectroscopically determined values adequately reproduce the total amount of co 2 originally added to these glasses over the range of silicate compositions studied. Precision and Accuracy From these results, infrared spectroscopy appears to be a useful tool for the measurement of co 2 concentrations in glasses along the NaAl0 2 -Si0 2 join to low levels of total dissolved carbon. Estimated detection limits are 25 ppm molecular co 2 and 75 ppm CO~- for 200 micron thick pieces of glass. Thicker samples would, from equation (1), have lower detection limits, but absorption bands due to aluminosilicate vibrations would tend to overwhelm the bands at both 1610 cm- 1 and 1375 cm- 1 in thicker specimens. The precision of the infrared technique for measuring co 2 concentrations in sodium aluminosilicate glasses is assessed to be on the order of several percent and no worse than ±5-6%. This estimate is 43 Figure 8- Comparison of total co 2 contents based on amount of silver oxalate loaded into the capsule and based on summing spectroscopically determined molecular co2 and co~- concentrations using the bands at 23S2 cm- 1 and 137S cm- 1 • Table 4. reference. A line of loaded C0 2 Molar absorptivities are from = measured C02 is shown for Samples shown are those used to determine molar absorptivities by least squares; the close correspondence of the data to a 4S 0 line reflects the internal consistency of the calibration procedure. (Samples shown are NC-13, 17, 19, 20, 21; ABC-S3, S4, SS, S6, S8, 62, 64; JDC-103, 108, llO, 111, l12, l13, 116, and BOET-229S.) 44 1.2 a. NaAISi 2 0 6 0 NaAISi 3 0 8 [J • NaAISi 3 0 8 (Boettcher) >. 0 [J 1.0 6 (.) (/) - NaAISi 4 0 10 0 ~ ( .) Q) a. (/) 0 0.8 0::::: ......... c: 0 "O Q) (/) 0 0.6 .c - 0~ -~ 0 N 0.4 u ....J <! .,.._ 0 .,.._ 0 .2 0 0 .2 0.4 0.6 LOADED C0 2 (wt. %) 0.8 1.0 45 based on potential uncertainties in absorbance (0.01 for an absorbance of 0.5 with the PE 180, based on the reproducibility of individual measurements from Stolper, 1982a), estimated density (±2%), and sample thickness (±3 µm) measurements and varies from sample to sample. Infrared spectroscopy can thus precisely measure the relative amounts of co2 dissolve in sodium aluminosilicate glasses. On the other hand, the accuracy of the technique for sodium aluminosilicate glasses, as reflected by Figure 8, is estimated to be ±15-20% of the amount present due to uncertainties in both the molar absorptivities that have been determined and in the co2 contents of the standard glasses. It has already been suggested that possible causes of these uncertainties include minor loss of co2 during the pist9n-cylinder experiments and some heterogeneities in co2 within individual glasses. These hypotheses were not originally testable with the PE 180 spectrophotometer, but with the acquisition of a Nicolet Instruments 60SX FTIR, I was able to use smaller beam sizes and discovered that both suspicions are probably correct. undersaturated Figures 9 and 10 demonstrate that a typical co 2-bearing glass is zoned, with higher co 2 and lower H20 contents in the core of the glass than in the rim next to the Pt capsule enclosing the sample. This profile again suggests that H2 enters the capsule and reacts with co 2 to form water and reduced carbon species. These species either then diffuse through the Pt wall, as originally suggested by Watson et al. (1982), or concentrate at the rim of the glass. Again, no evidence for reduced carbon species in the glasses has been found. While the nuclear microprobe apparently verified that total C contents integrated over a substantial area of a glass plate (from 46 Figure 9total Concentration profiles for dissolved CO~-, molecular co 2 , and co2 in sample ABC-58. A map showing the approximate location of each measurement is also shown. section of a piston-cylinder run. The glass used is a horizontal 47 ~ - 0.30 >~ - 0 (/) - 0 . .~ .! 0.20 0 C\J co c;u ~ c 0.30 :::J (.) Q) ~ ~ ~ 0.20 "'C - ~ C\J C5 0 (/) u 0.10 (/) 0 C\J- u ~ -: 0.20 0 ~ 0.15 ~ ~!(') 0.10 -o ~(/) u 0.05 0 t9 0 "'C - A B c D POINT Pt Glass E 48 Figure 10- Concentration profile for total dissolved H o in sample # 2 ABC-58. The location of each measurement is the same as shown in Figure 9. 49 - 0 . 15 0~ ~ .......... 0 N I 0 .10 "O (1) > 0 (/) (/) 0 0.05 o~....i....~~~_._~~~__.'--~~~-'--~~~_._---1 A B C POINT D E so ~0.2S mm 2 to ~a.so mm 2 , generally in the center of the glass plate) did not significantly decrease (Table 2), the scatter to the data was probably introduced by making infrared measurements near the rims of some glass samples rather than at their cores. I note that improved accuracy of the technique might be possible given this information. Speciation of Carbon Dioxide in Sodium Aluminosilicate Glasses The extinction coefficients given in Table 4 have been used to obtain the amounts of molecular co2 and co~- dissolved in each of the glasses synthesized in this study. The results are presented in Table s. Figure 11 shows molecular co2 versus carbonate con~entrations for jd-E, ab-E, and eu-E glasses synthesized at 14S0°C and 2S kbar with up to 1.1 wt.% total dissolved C0 2 • For each silicate composition, the ratio of molecular co2 to co~- dissolved in the glasses varies little over the range of total co2 contents studied. The ratio is, however, a strong function of silicate composition, increasing with increasing silica and decreasing soda and alumina content along the NaAl0 -Si0 2 2 join. co2 dissolves in .i! glass predominately as c~ - , the ratio of molecular C0 ~glasses, 2 to C0 2 dissolved as CO~- is approximately 0.6. · In ab and molecular co2 predominates; the ratios are approximately 2.3 and S, respectively. Glasses prepared from compositions that lost Na during the initial drying process (jd-X, ab-X, eu-X) have higher, but still constant molecular C0 2 to CO~- ratios. glasses and 2.7 in ab-X glasses. The ratio is 3.4 in jd-X Eu-X glasses contain negligible CO~-. 51 Table 5co 2 Concentrations (in wt.%) of molecular C0 , CO~-, and total 2 based on spectroscopic measurements. CO~- conce.ntrations are given as wt. % co2 dissolved as co~- and are based on the intensity of the 1375 cm- 1 band. are also given. The mole fractions of molecular C02 and CO~- Mole fractions are computed on the basis of moles of assumed mixing species (i.e. co2,mol' co32- , and 0). 52 Sample # P (kbar) T (°C) co2 (mol.) cor co2(Tot.) Xco2 Xe~3 NC NC NC NC NC NC NC NC NC NC NC NC ABC ABC ABC ABC ABC ABC ABC ABC ABC . ABC ABC ABC ABC ABC ABC ABC ABC ABC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC JDC 2 3 4 6 7 10 13 15 17 19 20 21 14 22 23 24 25 27 29 30 33 35 53 54 55 56 57 58 62 64 67 68 69 25 25 25 25 25 25 25 25 25 25 25 25 25 25 33 33 33 27 15 15 25 25 25 20 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 1560 1560 1450 1450 1450 1450 1450 1450 1450 1450 1400 1400 1400 1400 1400 1400 1400 1450 1450 1450 1450 1450 1450 1450 1450 1450 1450 0.62 0.05 0.29 0.29 0.30 0.58 0.02 0.34 0.25 0.45 0.22 0.37 0.15 0.04 0.05 0.10 0.08 0.09 0.78 0.08 o. 34 o. 39 0.38 0.67 0.02 0.34 o. 25 0.45 0.22 o. 37 0.0046 0.0004 0.0021 0.0021 0.0022 0.0042 0.0002 0.0025 0.0018 0.0033 0.0016 0.0027 0.0000 0.0032 0.0068 0.0018 0.0043 0.0012 0.0018 0.0016 0.0023 0.0017 0.0044 0.0019 0.0008 0.0029 0.0034 0.0021 0.0053 0.0044 0.0020 0.0011 0.0036 0.0008 0.0008 0 .0005 0.0006 0.0022 0.0005 0.0020 0.0019 0.0017 0.0002 0.0014 0.0033 0.0000 0.0004 0.0011 0.0003 0.0004 0.0007 0.0006 0.0007 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0015 0.0034 0.0012 0.0025 0.0006 0.0006 0.0008 0.0011 o.oo 11 0.0016 0.0007 0.0007 0.0006 0.0012 0.0007 0.0015 0.0016 0.0028 0.0024 0.0059 0.0017 0.0021 0.0015 0.0014 0.0039 0.0011 0 .0038 0.0037 0.0028 0.0005 0.0004 0.0053 0.0000 0.0008 72 73 74 76 84 85 86 89 90 91 103 108 110 111 20 20 20 20 20 20 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 25 o.oo o.oo o.oo o.oo o.oo o.oo o.oo o.oo o.oo 0.42 0.91 o. 24 0.58 0.16 o. 24 o. 22 0.31 0.23 0.59 0.25 0.11 o. 39 0.45 0 .28 0.71 0 .59 0.26 0.14 0.47 0.10 0.10 0.07 0.08 0.29 0.06 o. 26 0.25 0.23 0.02 0.18 0.43 0.20 0.46 0.16 0.34 0.08 0.08 0.11 0 .15 0.14 0.22 0.10 0.09 0.09 0.17 0.10 0.21 o. 22 0.37 0.32 0.77 0.22 0.28 0.19 0.18 0.51 0.14 0.50 0.48 o. 36 0.07 0.05 0.69 0.62 1. 37 0.40 0.92 0.24 0.32 0.32 0.46 0.37 0.81 0.35 o. 20 0.47 0.62 0.38 0.91 0.81 0.62 0.46 1. 23 0.32 0. 38 0.26 o. 26 0.80 0.20 0.76 0.73 0.59 0.09 0.23 1. 12 0.05 0.11 0 .16 o.oo o.oo o.oo 53 TABLE 5 (cont'd.) Sample JDC JDC JDC JDC BOET BOET BOET ' 112 113 114 116 2582 2990 2995 P (kbar) 25 25 25 25 20 20 20 T (°C) 1450 1450 1450 1450 1450 1450 1450 co2 (mol.) 0 .16 0.15 0.11 0.19 0.63 0.39 0.71 c~- 0.33 0.23 0.25 0.40 1. 29 0.13 0.21 co2 (tot.) Xco 0.49 o. 39 o. 37 0.59 1.91 0 . 52 0.92 0.0012 0.0012 0.0009 0.0015 0.0048 0.0029 0.0053 2 XCO"j 0.0026 0.0018 0.0019 0.0031 0.0099 0.0009 0.0016 54 Figure 11- co 2 concentration versus the concentration of co 2 dissolved as CO~- for glasses on the NaA1Si 2 0 -Si0 2 -co2 join quenched from 6 melts held at 1450 °c and 25 kbar. Concentrations are spectroscopically determined. The trends defined by each composition as a function of total reproduced in Figures 12-14. co 2 content are outlined and (Data points are for NC-2, 3, 4, 6, 7, 10, ABC-14, 22, 27, 53, and JDC- 84, 85, 86, 89, 90, 91, 108, 111, 112, 113, 114, 116.) 1.01~--r--r----i---r----i-r-~-,--r-r---r- 1450°C 25 kbar NoAISi30a ~ -~ N \J1 \J1 0 (.) 0.5 a:: :3 0.4 0 ::::> u w _J 0 ~ No Al Si206 (jd-D,jd-E) o NoAI Si30a (ob-E) "' No Al Si4010 (eu-E) o 0.2 0.1 6 a 0 0.1 0.2 0.3 0.4 0.5 C02 0.6 0.7 0.8 0.9 2as C0 3 (wt%) 1.0 I.I 1.2 1.3 56 I have also studied glasses of ab-E, ab-X and jd-E compositions quenched from melts held at pressures and temperatures other than 1450°C and 25 kbar. While there is some variation suggesting a slight increase in the ratio of molecular co2 to carbonate with decreasing temperature and pressure (Figures 12 and 13), the variations are minor for each of the compositions studied. The pressure and temperature dependence that is observed is small when compared to previously reported results (Mysen, 1976 and Figure 14). It is concluded that the speciation of co2 in glasses on the NaAl02 -Si02 join is not strongly dependent on pressure and temperature over the 15-33 kbar, 1400-1560°C ranges. The ratios of molecular C0 2 to CO~ - of the ab and .1i_ C02 -bearing glasses synthesized in the laboratory of Professor A. Boettcher are similar to those of our co2 -bearing ab-E and jd-E glasses (Figure 12). Since these glasses were synthesized using both Na2co3 and Ag 2c2o4 as C0 2 sources and were held at run conditions for a variety of run lengths, this correspondence is taken to be indicative of equilibrium between co2 species in the glasses we have studied. Since none of the runs were anhydrous, it is difficult to assess the role of H 0 in the C0 speciation trends observed. 2 2 Previous workers (Eggler and Rosenhauer, 1978) have stated that H2 0 enhances the formation of CO~-. There is no discernable relationship between the ratio of molecular co studied. 2 to CO~- and the H 0 contents of the samples 2 However, the water contents of the glasses are low; higher water contents may well influence the speciation of co2 • 57 Figure 12- Molecular C0 2 versus C0 2 dissolved as carbo~ate for glasses quenched from melts held at 1450°C and pressures other than 25 kbar. The pressure of each run is marked. Trends are for glasses equilibrated at 1450°C and 25 kbar (Figure 11). Filled data points are from samples synthesized in the laboratory of A. Boettcher. (Data points are for ABC-23, 24, 25, 29, 30, and BOET 2582, 2990, 2995.) 1.0,-r-r-~--r---r-.--~-.----.-~-,------ - 1450°C, P indicated NaAISi30a 0.9 0.8 ~ "i 0.7 - ~0.6 20kbor u a::: 0.5 \J1 00 <l: _J :::> 0.4 u w _J 0 ~ • NaAISi 2 0 6 (Boettcher) o No Al Si 3 0 8 (ob-E) • No Al Si30a (Boettcher) 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 2- 0.8 C02 as C0 3 (wt.%) 0.9 1.0 I.I 1.2 1.3 59 Figure 13- Molecular C0 2 versus C0 2 dissolved as carbonate for glasses quenched from melts held at 25 kbar and temperatures other than 1450°C. The temperature of each run is marked. Trends are for glasses equilibrated at 1450°C and 25 kbar (Figure 11). (Data points are for ABC-33, 35, and JDC-68, 69, 72, 73,74, 76.) 1.01=:-=---r--r--r--r~-r--r--r----.----.,....-- - 25 kbar, T indicated NaAISi30s NoAISi4010 0.9 0.8 ~ -= 0.7 ~ ~0.6 u °' 0 ct: 0.5 o/140ooc <{ _J :::> 0.4 u w 6 0.3 ~ 0.2 o No Al Si 2 0 6 (jd-0, jd-E) o NoAI Si30a (ob-E) 0.1 1400°C O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 2- 0.8 C02 as C0 3 (wt.%) 0.9 1.0 I. I 1.2 1.3 61 Figure 14- Molecular C0 2 concentration versus the concentration of C0 2 dissolved as CO~- for glasses on the NaAl0 2 -Si02 -co2 join comparing the data of Mysen (1976) with the results (shaded regions) presented in this study. Mysen's data points are labelled with the pressure and temperature of the melt prior to quenching. There is essentially no correlation between the two data sets. 62 c 30, 1625 2.0 1.8 1.6 -~ 1.4 C\I 0 u 1.2 •30, 1625 a:: <l: _J ::) 1.0 u 020,1625 w _J c 30, 1450 0 20, 1525ec 10, 1625 ~ 30, 1450 • 020, 1450 0.4 20, 1625• 020, 1525 •10, 1450 0.2 0 •I~ 1625 0.4 010, 1450 0.6 0.8 1.0 1.2 1.4 C02 as CO~- (wt. %) 1.6 1.8 2.0 63 Discussion Comparison with Previous Spectroscopic Investigations The results are comparable with those of Mysen (1976), who used infrared spectroscopy to investigate co2 dissolution in glasses quenched from vapor-saturated melts on the NaAl0 -Si02 -co2 join. 2 Mysen et al. (1976) also reported infrared studies of co -bearing NaA1Si o glass. 3 8 2 In addition, Brey (1976) studied C0 -bearing glasses on the NaA1Si o 2 3 8 CaA12Si2o8 join with infrared spectroscopy, and Raman spectroscopy was used to investigate glasses in this system by Mysen and Virgo (1980b). In all of these studies, as in this study, glasses were synthesized by quenching from melts held at high pressures and temperatures in pistoncylinder apparatuses. The pressures and temperatures from which the melts were quenched were similar in all of these investigations. One difference between these previous studies and this study is that they focussed on vapor-saturated systems whereas most of the glasses examined here were quenched from vapor-undersaturated melts. There is little disagreement about the existence of both molecular co2 and carbonate in co2-bearing sodium aluminosilicate glasses, although Brey (1976) states, without supporting evidence, that the molecular C02 is a quenching phenomenon. This study confirms the conclusion of Mysen (1976) that the speciation of co2 in silicate melts is strongly dependent on silicate composition, with the ratio of molecular C0 2 to carbonate increasing with increasing silica content on the NaA10 2 -Si02 join. There is, however, little quantitative correspondence between the measured molecular co2 and carbonate contents and those reported by Mysen (1976) (Figure 14). There is no evidence of 64 the strong pressure and temperature dependence of Mysen (1976) reported. co2 speciation that Finally, there are essentially constant carbonate to molecular co 2 ratios for each silicate composition as total dissolved C02 content increases; Mysen and Virgo (1980b) reported that the co~- to molecular co2 ratio increases sharply with total dissolved co2 in albite glasses. Why are these results so different from those of Mysen (1976)? The total co 2 contents of the glasses studied by Mysen tend to be at the high end of those studied here, since his experiments were vaporsaturated while most of these were not. The P-T conditions and total co 2 contents of these two studies overlap, however, and the results should be comparable. The explanation of the discrepancies between results probably lies principally in differences in analytic techniques. Mysen (1976) used a mixture of Ba 14 co 3 and Na 2co 3 as a co 2 source and measured the total dissolved carbon content of each of his glasses by 6-track radiography. Then, using the intensity of the 7.2 micron (~1375 cm- 1 in this study) carbonate band, which he calibrated using mixtures of powdered C0 -free silicate glass and Na co 2 2 3 embedded in KBr discs, he determined the concentration of CO~- dissolved in each glass. Molecular co 2 concentrations were determined by the difference between the total carbon content and the co~- content. One limitation of the procedure employed by Mysen (1976) is that uncertainties in bulk co2 content determined by 6-track mapping and in carbonate concentration determined by infrared spectroscopy propagate into estimations of molecular C0 2 content. One potential problem is that all of the carbon detected by 6-tracking may not be dissolved or 65 present as oxidized species. Other possible uncertainties associated with the use of 8-track mapping for bulk carbon analysis have heen discussed by Brey (1976), Rai et al. (1983), and Tingle (1985). There are also uncertainties associated with Mysen's (1976) determination of carbonate concentrations via infrared spectroscopy. Infrared spectra were obtained on discs of KBr in which powdered glasses were embedded. Numerous studies (Duyckaerts, 1959; Tuddenham and Lyon, 1970; Russell, 1974; Wong and Angell, 1976) have indicated that extreme care must be taken with KBr pellets and that quantitative results are difficult to obtain. Experience in our laboratory has confirmed that KBr spectra are often quite difficult to interpret, especially in the case of these co2-bearing glasses, in which relevant band intensities are quite low. Particle size has a large effect on absorption spectra and care must be taken to reproducibly grind samples. Grinding can, in some cases, structurally alter the material of interest. Adsorption of gases on both the KBr and the ground sample may also be significant (Barker and Torkelson, 1975). It is clear that many of the earlier studies that used infrared spectroscopy on KBr pellets to study volatile speciation in glasses suffered from these difficulties. Therefore, many of the conclusions drawn from them are suspect. In particular, the conclusions that molecular co2 is a detectable constituent of nearly all C-bearing glasses, including basic compositions such as diopside and several natural rock compositions (Mysen et al., 1975; Mysen et al., 1976), and that its concentration varies systematically with conditions of melt equilibration were almost certainly artifacts. In the case of diopside, molecular C02 has not been detected by Raman spectroscopy 66 (Sharma, 1979; Mysen and Virgo, 1980a; Rai et al., 1983). I have studied one of the glasses synthesized by Rai et al. (1983) [Di #89] and conclude, using a molar absorptivity for the 2352 cm-1 band of 945 liters/mole-cm, that the concentration of molecular co2 is less than 25 ppm by weight (Chapter 3). I have also failed to detect any molecular co2 in synthetic co2-rich basaltic glasses or in glasses in the system CaO-Al 2 o 3 -Si02 -co2 (Chapter 3). A combination of adsorbed co and 2 incomplete purging of the sample chamber (Brey, 1976) may have been responsible for the spurious molecular co2 observed in the spectra of Mysen et al. (1975, 1976). Examination of the spectra of Mysen et al. (1975), Mysen et al. (1976), and Mysen (1976) suggests that water adsorbed on the KBr was probably also a problem, resulting in major absorptions at ~3500 and ~1600 cm- 1 • Eggler at al. (1979) also suggest that adsorbed acetone can produce significant absorptions in spectra of powdered materials. There may also be problems with the procedure used by Mysen (1976) to calibrate the intensity of the ~1375 cm- 1 carbonate band. The carbonate band is split by ~zoo cm-l in NaAl0 -Si0 glasses and only 2 2 slightly split in Na2 co 3 (White, 1974), suggesting that caution should be used in equating the integrated molar absorptivities for the carbonate g roups in these two systems. Indeed, my measured integrated molar absorptivity for the 1440 cm-l (actual location) band in Na co 2 3 2 (determined using KBr pellets) is ~36500 liters/mol-cm , more than twice the value obtained for the 1375 cm- 1 band. Mysen's (1976) calibration curve for the ~1400 cm-l carbonate band (his Figure 10) is also troublesome since it is not linear and does not pass through zero and 67 that the area he measures to determine concentration is described in his text as "transmittance", whereas absorbance would be the usual measurement used for determining the CO~- concentration. By taking spectra on doubly- polished glass plates, our laboratory has avoided the problems associated with KBr pellets. Most important, however, is the fact that molecular co2 and carbonate band intensities are measured simultaneously in each glass plate. Thus, even if determinations of molar absorptivities are in error, leading to inaccuracies in the measured concentrations of the various species, the ratios of band intensities, and thus variations in the relative proportions of dissolved C-bearing species, can be precisely determined. This contrasts with the approach used by Mysen (1976) in which the concentration of molecular co 2 was determined by difference between independent determinations of total dissolved concentrations. co2 and carbonate Using Mysen's approach, meaningful determinations of the relative proportions of molecular co2 and carbonate depend on both the accuracy and precision of the total trations. co2 and carbonate concen- These factors probably explain the differences between the two data sets in Figure 14. The results presented in this chapter indicate that at the relatively low concentrations of molecular co2 in silicate glasses, the ratio of co2 to carbonate in a glass varies little with total dissolved C02 content. Using Raman spectroscopy, Mysen and Virgo (1980b) found that over a similar range of pressure, temperature and co2 content, the ratio of carbonate to molecular co2 increased sharply with increasing total C0 2 content for albite glasses . It is difficult to evaluate this 68 conclusion since Mysen and Virgo (1980b) give no quantitative details. There is neither a demonstration of a linear variation in band intensity with species concentration, nor a discussion of the precision and accuracy of their determinations of band intensities, nor a consideration of their detection limits for co2 and co~- in glasses. The band at ~1075 cm- 1 that they use to infer c~- concentration is deeply buried among three silicate bands, and its intensity can only be determined by deconvolution procedures. This probably means that the detection limit and precision for co~- concentrations determined by this procedure are poor. Apparently confirming this, they saw no CO~- in an albite glass containing 0.68 weight percent total co2, whereas in equivalent glasses studied here, carbonate is readily detected. appears from Figure 3 of Mysen and Virgo (1980b) that the intensity of the band at ~1378 cm-l they have assigned to molecular roughly in proportion to their total co • 2 CO~- It also co2 increases band between 2.05 wt.% and 3.20 wt.% Their data might be consistent with a roughly constant molecular C0 2 to carbonate ratio. Thus, the apparent discrepancy between the Raman spectroscopic results of Mysen and Virgo (1980b) and the infrared results may simply reflect the fact that infrared spectroscopy is, at this stage, more useful as a tool for measuring the concentrations of C-bearing species in silicate glasses. A similar conclusion was reached by Stolper (1982a) regarding measurement of the concentrations of water and hydroxyl groups in glasses. 69 Solubility Mechanisms of C02 in Sodium Aluminosilicate Melts Silicate melts are dynamic, with the homogenous equilibria between oxygen species continually readjusting as they are perturbed by the addition of co2 or other components. oxygen actually reacts with molecular Even if only a single kind of co2 to form carbonate, the concentrations of other oxygen species will be affected as they act to buffer the single reactive oxygen species. Thus, if the concentrations of the different oxygen species could be determined, it may be extremely difficult to specify which classes of oxygen are directly involved in carbonate-producing reactions and which are passively responding to these reactions. abo~t There has, however, been considerable speculation the structural role played by the oxygens that react with co 2 molecules to produce carbonate groups and about the structural changes that occur in the silicate framework of these melts when carbonation reactions proceed. It is widely believed, based on Raman spectra (Mysen and Virgo, 1980b), X-ray scattering data (Taylor et al., 1979), and thermochemical studies (e.g., Navrotsky et al., 1982), that NaAl0 -Si0 2 2 glasses and melts are nearly fully polymerized in the sense that every oxygen is shared between two silicate and/or aluminate tetrahedra. It is difficult to envision how co2 molecules can react with a melt in which there are only bridging oxygens to produce carbonate. One possibility is for the resulting carbonate ion to participate in the bridge between adjacent tetrahedra by forming silicon- or aluminum-bearing carbonate complexes (Sharma et al., 1979). ~= An example of such a structure might 70 CO 2 + [=(Si,Al) - 0 - (Si,Al) =l =<si,Al) - o ___ c11 ---0 - (Si,Al)= 0 Carbonate concentration decreases as silica is added to melts, suggesting that oxygens bridging two Si tetrahedra do not react with molecules in this way. co2 However, the formation of such structures between Al and Si tetrahedra or between two Al tetrahedra cannot be ruled out. In these cases, sodium atoms that provided charge-balance for the bridging oxygens prior to reaction with co2 molecules would probably be associated with the newly formed structure in some way. This study yields no data indicating if complexes like this are present in the glasses, but note that these co~- groups would undoubtedly be highly asymmetric and that this unusual environment could, in principle, account for the splitting of the v 3 carbonate band that we observe. Another possibility is that bridging oxygens are removed from the silicate-aluminate network when they react with co2 molecules to produce carbonate ions. In such cases, structural and charge balance considerations are usually taken to require local rearrangements in the silicate and aluminate framework. suggested that reaction of For example, Mysen and Virgo (1980b) co2 molecules with bridging oxygens to produce carbonate groups is accompanied by the generation of octahedrally-coordinated aluminum. Other configurations, such as five- fold or three-fold coordination of aluminum (McKeown et al., 1984) or 71 aluminum triclusters (Lacy, 1963) could also be generated by such reactions. None of these reaction products has been positively identified by spectroscopy in co2-free or co2-bearing glasses, although their concentration may be expected to be quite high in very carbonaterich aluminosilicate glasses. Mysen and Virgo (1980b) include weak bands they interpret as reflecting the presence of octahedrallycoordinated aluminum in their deconvolutions of the Raman spectra of co2-rich glasses. It is doubtful that co -free melts on the join NaAl0 -Si0 are 2 2 2 fully polymerized. They undoubtedly contain non-bridging and even free oxygens at some concentration level, probably some non-tetrahedrally coordinated aluminum (Boettcher et al., 1982; 1984), and perhaps aluminum triclusters. It may be that molecules of C02 react with two non-bridging oxygens to produce a bridging oxygen and a carbonate group in much the same way that carbonate groups are thought to be produced in depolymerized melts such as diopside (Eggler and Rosenhauer, 1978). Alternatively, free oxygen could react with C0 2 to form carbonate complexes (Tomlinson, 1953; Pearce, 1964; Wagner, 1975). Only small amounts of non-tetrahedrally coordinated aluminum or of aluminum triclusters would be required to produce enough non-bridging or free oxygens to account for the observed concentrations of CO~- groups - in the glasses I have studied. In addition, as reactive oxygens are consumed and carbonate ions are generated, the concentrations of the reactive oxygens might be regenerated to some extent by the equilibria between these and the major oxygen species; relatively large amounts of carbonate could be produced by reaction of co2 molecules with non- 72 bridging or free oxygens even if the concentrations of these species in melts are small. Most of these and the equivalent statements of previous workers are highly speculative. Reliable and sensitive techniques for quantitatively determining both the concentrations of different types of oxygens and local configurations in glasses are needed before statements about changes in melt structure that accompany carbonate formation can be made with confidence. Conclusions 1. Infrared spectroscopy can be used to quantitatively determine the concentrations of molecular co 2 and carbonate groups in sodium aluminosilicate glasses • . If doubly-polished glass plates are used instead of powders embedded in KBr pellets, species concentrations can be measured with a precision on the order of several percent of the amount present. This technique has the potential for analyzing individual spots with diameters on the order of 10 µm. The accuracy of species concentrations measurements is currently limited to about ±15-20% in glasses studied due to heterogeneous glass standards, but this accuracy can be improved with suitable standards. 2. Glasses in the system Na 2 0-Al 2 o 3 -sio 2 -co 2 near the NaA1Si 2 06 -Si0 2 co2 join that have been quenched from melts at 15-33 kbar, 1400- 15600C contain both carbonate and co 2 molecules. Carbonate groups are in distorted environments that are very different from those in crystalline carbonates. The carbonate groups may be in the form of Na-carbonate ionic-complexes. 73 3. There is an approximately linear relationship between molecular C0 2 and carbonate concentrations for each of the sodium aluminosilicate compositions studied, up to at least 1 wt.% total dissolved co2. The molecular co2 to carbonate ratio is strongly dependent on the silicate composition of the glass, with the proportion of molecular co2 increasing as silica is approached on the NaA1Si2o6 -Si02 join. Compositions that differ only in their soda concentrations have different molecular 4. co2 to carbonate ratios, with those that have lower Na 0 contents having higher ratios. 2 The speciation of C02 in glasses quenched from melts is only weakly dependent on the pressure or temperature at which the melt was equilibrated over the 15-33 kbar, 1400-1560°C range. with previously reported results. This contrasts 74 Chapter 2. The Solubility of Carbon Dioxide in Molten Albite 75 The solubility of C02 in both natural and synthetic silicate melts has been extensively studied and a large volume of data has been accumulated. Unfortunately, little of the data agree and no clear picture of the interactions between co2 vapor and molten silicates has emerged. There have been a number of motivations behind obtaining data of this sort. First, solubility data ultimately provide upper limits to the amounts of carbon dioxide contained in evolving magmas. There is, however, little reason to suspect that magmas are ordinarily saturated with C02 vapor in the range of pressures and temperatures associated with most experimental studies (10-40 kbar, 1250-1750°C), although a saturated state is probably reached at pressures approaching 1 atm. (e.g., Moore, 1979). This minimizes the importance of the absolute values of carbon dioxide solubilities in melts equilibrated at high pressures. However, information on the dependence of carbon dioxide solubility on the pressure and temperature of melt equilibration may provide some insight into the mechanism of carbon dioxide dissolution, into the nature of carbon dioxide evolution and degassing during magmatic ascent, and into the structures and properties of silicate melts (e.g., Mysen, 1976, 1977). Techniques previously used to measure co2 solubilities include S- track radiography (Holloway et al., 1976; Mysen et al., 1975, 1976; Mysen, 1976; Mysen and Virgo, 1980ab), high temperature mass spectrometry (Rai et al., 1983), vacuum fusion (Pearce, 1964; Faile and Roy, 1966), gas chromatography (Strnad, 1971; Brey, 1976; Brey and Green, 1976; Mysen et al., 1976), differential thermal analysis (Eggler 76 and Rosenhauer, 1978), and Raman spectroscopy (Verweij et al., 1977; Sharma, 1979; Sharma et al., 1979). As discussed previously (Intro- duction), infrared spectroscopy offers a major advantage over most of these techniques in that it is species-specific. can be used to distinguish between molecules of dissolved in a glass. and Infrared spectroscopy co2 and co~- ions It can also discriminate between dissolved carbon co2 present in fluid inclusions, co~- ions contained in quench or alteration phases, and contaminant carbon. Since solubility studies by their nature require the study of vesicular glasses, yet are concerned only with the dissolved co2 content of the glasses, this is a very desirable feature. In addition, spot sizes as small as tens of microns in diameter are attainable with the FTIR. This allows the beam to be directed between bubbles or alteration phases, and permits examination of sample heterogeneity. Since fusion of the sample is not required during analysis, the possibility of reaction between species is also eliminated. published It is probable that many of the discrepancies found in co2 solubility data are attributable to the problems that infrared spectroscopy eliminates. Experimental Techniques Sample Synthesis The synthesis of in Chapter 1. co2 -bearing glass samples is discussed in detail For this particular study, powdered silicate starting material of approximate albite composition was synthesized by grinding and mixing Johnson-Matthey Specpure Na2 C03 , Al2 03 , and Si02 in an agate mortar for 6 hours, followed by melting at 1580°C for 12 hours in air at 77 1 atm. This decarbonated glass was ground under ethanol for 6 hours and dried at 850°C in air for 2 days to remove any hydrocarbon residue. A microprobe analysis of the decarbonated glass is given in Table 6; the glass is deficient in Na2 o and enriched in Si02 relative to stoichiometric albite, probably due to a combination of Na o 2 volatilization during decarbonation and drying and Si02 contamination during grinding in the agate mortar. It is emphasized that the drying process was necessary to reduce the adsorbed H2 0 content of the starting material. The result~ presented in Chapter 1 suggest that slight deficiencies in Na2 0 content could affect C02 solubility. It is again noted that less emphasis is placed on the actual values of co2 reported than on the pressure and temperature dependence of these glasses. co2 solubility in It is doubtful that these characteristics of silicate glasses vary significantly with small changes in Na2 0 content. Bubbles from 5 µm to lOOµm in diameter and presumably containing gaseous or liquid C02 are evident in the quench carbonate is visible. co2 -saturated glasses. No The glasses are often yellow or blueish- orange due to the disseminated Ag and the bubbles tend to be present in streaky clusters distributed throughout the glass. The C02 -saturated samples studied in this report are not zoned, probably due to excess co2 present throughout the capsule as a vapor phase at all times during melt equilibration. The quenched glasses were sectioned by diamond saw, ground into plates 50-250 µm in thickness, polished in a slurry of Al 2 o3 and H2 0, and measured for thickness using a digital-dial indicator. They were placed on metal apertures 50-1000 µm in diameter and examined 78 Table 6- Average of 5 microprobe analyses of the albite glass used in this part of the study. Microprobe conditions were: 15 kV accelerating voltage; sample current: 5 nA on brass; 40-50 micron beam size. Also given is an analysis of stoichiometric albite. 79 Analyzed Ideal Na 2 0 10.38 ll .82 Al 2 03 20 .43 19.44 Si02 69.66 68.74 TOTAL 100.47 100.00 80 microscopically to verify that bubbles were not present in the region of glass exposed by the aperture. In some instances, a few small (~5 µm diameter) bubbles were unavoidable, but based on the similarity of spectra of both bubbled and unbubbled regions of other glasses, these have not affected the measurements presented here. Infrared Spectroscopy of C02-Saturated Glasses Infrared spectra were obtained for this study on both the PerkinElmer 180 Infrared Spectrophotometer and the Nicolet Instruments 60SX Fourier Transform Infrared Spectrometer (FTIR). Both machines yield quantitatively similar results, but the FTIR is quicker and has the advantage of the smaller beam sizes (> 10 µm) previously mentioned. Infrared spectra from the FTIR were obtained using a HgCdTe 2 detector, KBr beamsplitter, globar source, a mirror velocity of 1.57 cm/sec., and 1012 to 8096 scans. Typical spectra are shown in Figure 15, which shows the spectra of both a co 2 -bearing and a co2 -free albite glass. The co2 -free glass was synthesized at 1 atm. in air in a vertical quench furnace. The assignments of six absorption bands unique to the co2-bearing glass are reviewed: The first, a weak band at ~3710 cm -1 is tentatively attributed to a combination of the Vl + V3 modes of C02 dissolved in the glass in the form of molecular co • The second, a broad absorption band 2 at ~3550 cm- 1 is attributed to the stretch of hydroxyl groups and is due to water dissolved in the glass (Stolper, 1982a). For reasons discussed previously (Chapter 1), H2 0 appears to be an unavoidable consequence of the experimental technique. The sharp band at 2352 cm- 1 is due to the 81 Figure 15- The upper box shows the spectra of a C0 -bearing (top, 2 sample # ABC-53) and co 2 -free (bottom) glass of albite composition, both scaled to a sample thickness of 45 microns. assignments are noted. Absorption band The spectra go offscale at ~1200 cm-l due to the vibrations of the aluminosilicate glass matrix. band at 2287 cm-l due to molecular 13 co in the bottom box. 2 The absorption is highlighted and enlarged 82 . N -OH (T') 1 00 U1 N (T') 0 . 2 2 02 83 v3 mode of co2 dissolved in the glass in the form of molecular 12co2. Another weak band at 2287 cm- 1 is attributed to the v 3 antisymmetric stretch of molecular 13 C02 • The absorption bands at 1610 cm-l and 1375 cm -1 are attributed to co2 dissolved in sodium aluminosilicate glasses in the form of distorted co~- ionic-complexes. Computer collection of the spectra presented in this chapter and in the following chapters obviated the need for digitizing spectra by hand. To determine the intensity of absorption bands for this portion of the study, the computer collected spectrum of co2 -free glass was scaled to the thickness of each co2 -bearing glass and computer subtracted from each spectrum of co2 -bearing glass. This resulted in a spectrum with a relatively flat background from which the intensity of each absorption band could be simply determined. In the cases of the absorptions at 3710 cm- 1 and 2287 cm-1 , background effects due to the more intense bands at 3550 cm- 1 and 2352 cm- 1 were estimated and subtracted by hand. These estimates were straightforward in the case of the band at 2287 cm- 1 , but less obvious for the band at 3710 cm-1 • For these measurements only band heights, rather than the integrated area under bands have been determined. In silicate glasses saturated with C02 , the intensity of the band at 2352 cm- 1 is usually too intense to adequately measure and characterize. This necessitated calibration of the less intense absorption band at 2287 cm- 1 due to molecular 13 co2 • The intensity of this band has been calibrated against the intensity of the band at 2352 cm-l in Figure 16, using a set of undersaturated co2 -bearing sodium aluminosilicate glasses described in Chapter 1. Details regarding each 84 Figure 16- Intensity of the absorbance band at 2287 cm-l versus the intensity of the absorption band at 2352 cm -1 • A line given by absorbance 2352 = 38.2 absorbance 2287 is shown for reference. 85 2.0r-----r----r---.,...--r----""T""'"P"--1.8 1.6 • 1.4 I E (..) C\.J L{) 1.2 r0 C\.J Q) (..) 1.0 c c -e0 0.8 CJ) _Q <! 0.6 0.4 0.2 0 o 0.01 0.02 0.03 0.04 0.05 0.06 Absorbance 2287cm- 1 86 of the samples used in the calibration are presented in Table 7. Table 7 demonstrates that absorbance 2352 = (38.2 ± 2.1) absorbance 2287 • Using this ratio of absorbances and assuming s 2352 = 945 ± 45 liter/mole-cm, a value of s 2287 = 24.7 ± 1.8 liters/mole-cm is calculated. This value is used for determination of the total dissolved molecular co 2 content of glasses, not simply dissolved 13 co 2 and assumes no significant variable fractionation of 13 c from experiment to experiment. The band at ~3710 cm-l, attributed to molecular C0 2 dissolved on the molecular scale in the glasses, could also be used to determine the molecular C02 content of these glasses. A preliminary calibration of the band indicates that absorbance 2352 = (47 ± 6) absorbance 3710 , yielding a molar absorptivity of 20.1 ± 2.7 liters/mole-cm. The molecular co 2 content of the saturated albite glasses was not determined from the intensity of this band, because the difficulties in background subtraction mentioned previously preclude simple determination of the intensity of this band for most of the samples. Results The result of the infrared measurements are presented in Table 8. Total co 17a. 2 solubility is shown as a function of pressure in Figure Isothermal C02 solubility increases dramatically with pressure over the 15-30 kbar range. There is also a positive correlation between equilibration pressure and the concentration of both molecular co CO~- at constant temperature (Figures 17b and 17c). 2 and At constant pressure, total co2 solubility in albite glasses is not a strong function of equilibration temperature, although the results suggest a 87 Table 7- Samples used in the calibration of the intensity of the absorption band at 2352 cm-l versus the intensity of the absorption band at 2287 cm- 1 • Sample numbers prefixed ABC are of albite (NaA1Si 3o8 ) composition, JDC and 1::!._ are of approximate jadeite (NaA1Si 2o6 ) composition. with co 2 except TJ-18. None of the samples used were saturated Sample Thickness (cm) Absorbance (2352 cm- 1 ) Absorbance (2287 cm- 1 ) abs2352/abs2287 ABC 53 0.0045 1.09 0.028 38.9 ABC 54 0.0057 0.74 0.021 35.2 ABC 58 0.0075 1.05 0.028 37. 5 JDC 103 0.0069 0.87 0.023 37.8 JDC 108 0.0036 0.80 0.021 38 .1 JDC 111 0.0113 0.40 0.011 36.4 JDC 112 0.0141 1.52 0.036 42.2 TJ 18 0 .0105 1.92 0.049 39.2 AVG. a 1 standard deviation. 38. 2 ± 2.1 a 00 00 89 Table 8- Data obtained from co 2-saturated albite glasses. Sample 41 42 37 48 4 5 12 15 36 43 40 39 8 T Duration Density 8 (g/l) (kbar)(°C) (min.) p 15 15 20 20 20 20 25 25 25 30 30 30 1450 1525 1450 1450 1525 1650 1450 1525 1650 1450 1525 1650 65 62 60 60 100 60 60 65 60 60 90 70 2445 2445 2510 25 10 2510 2510 2530 2530 2530 2535 2535 2535 Thickness Band (cm- 1 ) Abs. (cm) 3516 3522 3517 3520 3534 3540 3552 3518 3524 3514 3495 3518 0.81 1.11 0.66 0.06 0.53 0.41 0.11 1.10 0.39 0.55 1.17 0.87 0.0118 0.0115 0.0092 0.0052 0.0116 0.0130 0.0095 0.0098 0.0062 0.0092 0.0116 0.0112 Band Thickness (cm-1 ) Abs. (cm) Band Thickness (cm-1) Abs. (c•) 2287 2287 2287 2288 2285 2288 2286 2287 2288 2287 2288 2287 1597 1598 1600 1600 1600 1605 1610 1598 1609 1615 1631 1600 0.032 0.069 0.011 0.078 0.084 0.078 0.096 0.090 0.054 0.109 o. 120 0.108 0.0052 0.0150 0.0092 0.0052 0.0116 0.0130 0.0095 0.0098 0.0062 0.0092 0.0116 0.0112 0.26 0.38 0.27 0.16 0.37 0.43 0.44 0.46 0.12 0.64 0.93 0.94 0.0118 0.0150 0.0088 0.0052 0.0116 0.0130 0.0095 0.0098 0.0062 0.0092 0.0116 0.0112 Molecular C02 COi-b Total C02 (wt.%) (w .%) (wt.Z) 0.45 0.34 0.55 0.53 0.51 0.43 0.11 0.65 0.61 0.83 0.73 0.68 0.20 0.23 0.27 0.21 0.28 0.29 0.40 0.41 0.45 0.60 o. 70 0.73 0.65 0.56 0.82 0.80 0.79 0.72 1.11 1.06 1.06 1.44 1.42 1.41 co 2 ---r- H Oc cwLx> 2.26 1.47 2.04 1.96 1.82 1.47 0.63 0.94 0.64 0.10 0.41 0.28 0.35 1.00 0.56 0.53 0.90 0.69 C03 1.11 1.58 1.37 1.38 1.04 0.93 Densities estimated from Kushiro (1978). b co~- concentrations are expressed as wt.% co2. c u2o contents obtained using the method of Stolper (1982a) and a molar absorptivity of 80 liters/mole-cm. 0 '° 91 Figure 17- Total C0 2 (a), molecular C02 (b), and CO~- (c) concentration versus pressure. The temperature of melt equilibration is listed for each set of runs and is also indicated by triangles (1450°C), circles (1525°C), and squares (1625°C). #37 is shown at 1450°C, 20 kbar. For clarity, only sample 92 1.6..------.--------.--------.--------.---.. a) 1.4 1.2 "'O Q) - 2= o-e ~...; 1.0 -~ !t o- - C\J -E 8 0.8 I- 1450 1525 1625 0.41-----+-------+--------+-------+-----I 0.6 1.0 ::l (.) b) 0.8 Q) - o-e ~ ~ 0.6 0 "'O Q) C\J 2= 0 ~ u 0.4 1450 (/) 1525 __•_ ---1625 0 ---- Q21----4-------+------+-------t-----1 0.8 c) (/) 0 C\J o~ uo 0.6 ~ !t 0.4 > I r<i - C\J "'O Q) - oo ~ u 0 1625 0.2 1525 1450 OL-~~~15~~~~~2~0~~~~~2~5~~~~--.:3~0:"'"""'"-' Pressure (kbar) 93 slight negative temperature dependence of 16250C temperature range (Figure 18a). co2 solubility over the 1450- Within the precision of the analyses, this temperature effect may be negligible, rather than negative. The concentration of molecular co2 decreases with increasing temperature at constant pressure; this effect is largely counterbalanced by a positive correlation between equilibration temperature and co~­ concentration (Figures 18b and 18c) Molecular co /CO~- ratios are plotted as functions of pressure and 2 temperature in Figures 19a and 19b. The ratio is dependent on both pressure and temperature of melt equilibration and generally decreases both with increasing pressure and increasing temperature. This slight variation is comparable with the range for albite glasses reported in Chapter 1 and is not of the magnitude reported by Mysen (1976). The water contents of all of the run products are disturbingly high. The possible sources of this water in these nominally anhydrous runs are discussed at length in Chapter 1. The belief that it is unavoidable in piston-cylinder experiments due to a combination of adsorbed water on our starting materials and diffusion of hydrogen into the capsules during run conditions is reiterated. previous studies of It is doubtful that co2 solubility using analogous methods of glass preparation have had more success establishing anydrous conditions. While previous workers have suggested that H2 o would increase the solubility of co2 in silicate melts (e.g., Mysen, 1976; Eggler and Rosenhauer, 1978), there is no apparent correlation between H2 o content and co2 solubility. It is noted that two glasses equilibrated at the same pressure and temperature (#37 and #48, Table 3) contain different 94 Figure 18- Total C0 2 (a), molecular C02 (b), and CO~- (c) concentration versus temperature for the glasses studied. The pressure of melt equilibration for each set of runs is also listed. only sample #37 is shown at 1450°C, 20 kbar. For clarity, 95 1.6 a) 30 1.4 1.2 "'O Q,) - 0 0 -~ ~ • >~ en ..: I 0 o- . -E 8 0.8 - - C\J r- 0.6 . • • 25 I 20 • ---------._ -- ---- 0.41--------+----___.,__ _-_15_ 1.0 ..... c :::;::, u _,j b) 0.8 -Q,) -......o 0 O' ~ "i 0.6 "'O Q,) ..::= 0 • C\J ~ u 0.4 (/') 0 ~- ---- • - 20 0.2 .,___ _ ___..,_ _ _ _____.,_-_-_-_15_--1 0.8 c) 30 (/') c uo 0.6 ~ 0.4 C\J o~ _: "'O • Q,) - 2= c!Jr0 oo ~ u 0 0.2 • • • • --------1500 1600 Temperature (°C) 25 20 15 1700 96 Figure 19- Molecular co /CO~- ratios versus temperature (a) and 2 pressure (b). Temperature is again indicated in (b) by triangles (1450°C), circles (1525°C), and squares (1625°C). The 15 kbar, 1525°C data point might be inconsistent with the rest of the data and is shown by a dashed circle in both figures. 97 ~~ 2.6 8 a) CJ) c 22 C\I 0 u .......... 1.8 C\I 0 u "c 1.4 20 25 :::l ~ 1.0 30 0 ~ o~~~~~--'-~~~~__..~~~~_._~~~~...._~~~---' 1400 I C\I~ 2.6 u (/'J c N 2.2 0 u " 1,8 u "- 1.4 1700 b) 1450 1525 ........... ........ 1625....... .... ................ ......... ........ N 0 1500 1600 Temperature (°C) ........ - .... .... 0 c :::l u <1' 1.0 0 ~ 0.6 15 20 25 Pressure (kbar) 30 98 amounts of dissolved water (0.64 wt.% and 0.10 wt.%, respectively), yet co2 and c~- contents, and hence their total their dissolved molecular dissolved co2 contents (0.82 wt.% and 0.80 wt.%, respectively), are essentially the same. Comparison with Previous Results Mysen et al. (1976), Mysen (1976), Mysen and Virgo (1980b) and Brey (1976) studied the solubility of co2 in albite glasses. The first three studies used S-track radiography to find a positive temperature and pressure dependence of C02 solubility in the 10-30 kbar; 1450-1750°C ranges. The latter study, on the basis of gas chromatography, inferred a negligible or slightly negative temperature dependence for co2 solubility over a comparable range of pressures and temperatures. The results agree with the Mysen studies in many ways. The absolute C02 solubilities are similar to those reported by Mysen (1976). Compare, for example, 1.44 wt.% Mysen's (1976) 1.6 wt.% C0 pressure and temperature. pressure dependence of C0 2 co2 at 30 kbar, 1450°C with and Brey's (1976) 2.1 wt. % C02 at the same The results also confirm the strong positive 2 solubility found by Mysen (1976). Note, however, that there is no evidence for a strong positive temperature dependence of co2 solubility found in this study, but there is evidence for a slight negative temperature dependence, in accord with the inference of Brey (1976). Studies of the temperature dependence of C02 solubility in other silicate compositions have also contradicted the data produced by 8-track radiography. Rai et al. (1983) used high temperature mass 99 spectrometry and Sharma (1979) used Raman spectroscopy to find a slight negative temperature dependence of co2 solubility in melts of diopside composition at 20 kbar, 1550-1725°C. Brey (1976) found a slight negative temperature dependence on co2 solubility in olivine melilite melt at 30 kbar, 1450-1650°C. In addition, although there are no comparable 8-track studies, Pearce (1964) used a vacuum fusion technique to find a strong negative temperature dependence of co2 solubility in sodium silicate melts at 1 atm. C02 pressure. Faile and Roy (1966) also report a negligible temperature dependence of co2 solubility in potassium silicate glasses to 10 kbar pressure. While I can make no firm assessment of the reasons for the discrepancies between the results produced by 8-track radiography and the results produced by all other techniques (see Rai et al., 1983 or Tingle, 1985), I must again arrive at the conclusion that aspects of the results produced by 8-track radiography are suspect. Conclusions 1. In C0 2 -saturated albite glasses equilibrated at constant temperature, both molecular co2 and c~- concentrations increase with increasing pressure. The ratio of molecular C0 2 to CO~­ decreases slightly with increasing pressure. 2. In C0 2 -saturated albite glasses equilibrated at constant pressure, the concentration of molecular co2 decreases with increasing equilibration temperature while the co~- concentration increases; the net effect is a slight decrease in total increasing temperature at constant pressure co2 content with 100 Chapter 3. The Speciation of Carbon Dioxide in Ca ± Mg Silicate Glasses 101 The dissolution of C0 2 as both molecular co 2 and CO~- is a characteristic not only of the sodium aluminosilicate glasses discussed in the last two chapters, but of many aluminosilicate glasses dominated by cations with a +l valence state (e.g., orthoclase glass, Figure 20). Preliminary work has also indicated that natural rhyolitic glasses contain dissolved molecular C02 (Figure 21). This could be the consequence of a number of factors discussed below (Chapter 4), but is probably due in large part to similarities in structure between these glasses, all of which are probably nearly fully polymerized (Chapter 1), with monovalent cations providing charge balance for aluminum in the glass framework. Previous work (e.g., Brey, 1976; Mysen and Virgo, 1980ab; Rai et al., 1983) has indicated that the speciation of dissolved co2 in glasses rich in divalent cations (such as Ca2 + or Mg 2 +) is quite different from the speciation of C0 2 in sodium aluminosilicate glasses. This chapter explores these differences in detail. Sinthesis of C02 -Bearing Glass Standards Three starting compositions were used: .ii_, CaMgSi 2 o6 ; anc, a calcium aluminosilicate composition near the 1 atm. anorthite-tridymitewollastonite eutectic with approximately the chemical formula CaA1 0 • 8 si 2 • 9 o8 ; and 519B, natural basaltic glass ALV-519-4-1 FAMOUS region of the mid-Atlantic Ridge. from the The first two compositions were prepared by mixing and grinding appropriate amounts of JohnsonMat they Specpure CaC0 3 , Si02 , and Al 2 o3 or MgO in agate under ethanol; the third was prepared by grinding selected chips of ALV-519-4-1 under 102 Figure 20- Infrared spectrum of a 39 micron thick piece of orthoclase glass (ORC-3) containing 0.38 wt.% capsule. co2 , as measured into the Pt Tentative band assignments are shown. The glass was equilibrated at 1700°C and 20 kbar for 1 hour using the techniques described in Chapter 1. 103 a> CS> "' CS) IJ) I - ~(") (0 0 CS) CS> t..n CS) - Nt E E 0 J I ' v CS) CS) IJ) OCl4 0 I(!) N z w _J w CS) CS> CS) ('t) > < 3 0::' w al l: ::J z w > <( 3 ('t) CS) CS) L/) ('t) J: 0 I v N CS> - - - a> (0 v N (S) CS) (S) CS) 3JNV8~0S8\f CS) CS) CS) v 104 Figure 21- Infrared spectrum of a 33 micron thick piece of rhyolite glass (KS-2) containing 0.43 wt.% capsule. co2 as measured into the Pt Tentative band assignments are shown. The glass was equilibrated at 1550°C and 25 kbar for 1 hour using the techniques disussed in Chapter 1. 105 co CS) "" (S) LI) ~~ .., <O - (S) (S) (S) ..... LI) NI E J E 0 (S) (S) LI) '\ I r- ~ . (\.I 0 N (.) (.!) z w _J w (S) CS) (S) (\") > <( 3 Ct::' w ca L ::J z w > <( 3 ('t) (S) (S) J: LI) (\") 0 I (S) (S) ~ - N CS) CX> <O ~ N (S) (S) (S) (S) 3JNV'8~0S8V' (S) ~ 106 similar conditions. The chips were free of alteration, contained less than 5% phenocrysts by volume, and were ultrasonically cleaned in dilute HCl before grinding. for 12-24 hours: di f0 All three compositions were then held at 800-850°C and~ in air, 519B in a C02 -H 2 mixture giving an approximately at the QFM buffer. This heat treatment resulted in 2 decarbonation, dehydration and crystallization or recrystallization of the starting compositions, which were then stored over dessicant. Major element analyses of one or more glasses of each composition studied were obtained . using the MAC-5 electron microprobe with a 15 kV operating voltage, a sample current of 5 nA on brass and a 40-50 micron beam size (Table 9). Pt capsules used for experiments with the 519B composition had been pre-doped with Fe to reduce Fe loss from the 519B melts during run conditions. To pre-dope the capsules, they were filled with glass of 519B composition and held for 24 hours at 1580°C and the QFM buffer at 1 atm. The silicate glass was removed using HF and these capsules were reloaded with silicate and Ag 2 c2 o4 • This technique effectively eliminated loss of Fe (Table 9). In all cases, the melts quenched to glass. glasses was saturated with noted. None of the synthetic co 2 vapor and no bubbles or inclusions were Some high pressure runs were done without added silver oxalate in order to synthesize co 2 -free glasses. Anhydrous, decarbonated glasses of each composition were also synthesized at 1 atm. on Pt loops: anc and di at 1450°C in air, 519B at 1315°C at the QFM buffer. The infrared spectra of these glasses did not differ from the spectra of glasses synthesized at high pressure without added Ag 2 c2 o4 • 107 Table 9- Microprobe analyses of glass compositions used for piston- cylinder runs. The 519B analysis given is of glass run in an Fe- doped capsule (519B-S). Ideal analyses of the anc and die compositions and an analysis of natural 519B glass are also included. Microprobe techniques are described in the text. ANC Ideal DIC Ideal 519B Naturala --- -- -- -- 2.30 2 .13 -- 17.52 18.62 10. 76 10 .15 Al 2 0 3 14.22 14.40 -- - 15.83 16.44 Si02 61.67 62.35 55.07 55.49 47 .77 49.07 K2 0 -- -- -- -- 0.10 0.07 Cao 23.02 23.25 25.68 25.89 11.40 ll .65 Ti0 2 ----- ---- --- -- 0.73 0.74 - 0.10 0.03 -- -- 0.16 0.16 -- 8.52 - 8.86 98.91 100.00 98.27 100.00 97.67 99.45 2590 2820 Na 2 0 MgO Cr 2 o 3 MnO Fe Ob TOTAL DENSITY (g/l) -- aGlass analysis from Bryan and Moore (1977) bAll Fe reported as FeO. 2790 ...... 0 CXl 109 Two synthetic glasses provided by Dr. S. Sharma were also used in this study: a CaMgSi 2 o6 + co2 glass (Di #89; Sharma et al., 1979) and a Ca 2 MgSi 2 07 + C0 2 glass (Ak #51; Rai et al., 1983). Caco 3 was used as a C02 source for both runs, which were saturated with co2 -rich vapor. Results Rand Assignments and Intensities Figure 22 shows the spectra of both a C0 -bearing and a C0 -free 2 2 glass of ~ composition. Three absorption bands are observed in the C0 2 -bearing anc glass that are not present in the C0 2 -free sample. Both spectra go offscale at approximately 1200 cm- 1 due to vibrations of the aluminosilicate glass matrix. In Figure 23, a portion of the spectrum of a co2-bearing anc glass after subtraction of the co2-free anc glass spectrum is shown. This spectrum is overlain by the spectrum of NaA1Si 2 06 + C02 glass discussed in Chapter 1. The three absorption bands evident in the NaA1Si 2 06 + C0 2 glass spectrum were shown to be characteristic of all NaA1Si 2 o6 -Si02 -co2 glasses. The spectrum of the anc + co 2 glass (Figures 22 and 23) has absorption bands at 1515 cm- 1 and 1435 cm- 1 • These bands are considered to be analogous to the bands in the spectrum of NaA1Si 2 06 + C0 2 glass (Figure 23) at 1610 cm- 1 and 1375 cm- 1 • bands in the anc + co 2 Like the latter bands, the spectrum are always approximatel y equal in intensity regardless of total dissolved co2 content or total dissolved H2 0 (Figure 24). Recall that the bands at 1610 cm- 1 and 1375 cm- 1 in the NaA1Si 2 06 + co 2 spectrum were assigned to the v 3 antisymmetric stretch of distorted 110 Figure 22- Infrared spectra of ANC-10 (solid), containing 0.61% dissolved C0 2 , and a BL, dashed). thickness. co 2 -free anc glass synthesized at 1 atm (ANC- Both spectra have been scaled to 42 microns sample Aluminosilicate vibrations cause the spectra to go offscale at -1200 cm- 1 • WAVELENGTH .. ;-tm 4 3 1 •4 5 6 I 7 8 co~- I 1.2 w u z v1t 1.0 <( m CY 0 (j) m <( 0.8 I 0.6 0.4 0. 2 t= 4000 -OH _, -------3500 3000 -- 2500 WAVENUMBER .. cm _.// 2000 -1 --- / ---./ 1500 I ...... ...... ...... 112 ~igure 23- Background subtracted spectra of ANC-9 glass (dashed) containing 0.12 wt. % dissolved co2 and a jadeite composition glass (solid, JDC-84) containing 0.80 wt.% dissolved infrared spectroscopy. co 2 as measured by 113 \ E ~ (Q I ' I(.!) z 'I (S) (S) ' Ct'.'. w w _J LJ) w -'~ } N <! 3 l I I N 0 I I I u (S) (X) (Q ~ N (S) (S) (S) (S) 3JNV8~0S8V E "" u (S) (S) (S) > I en L: :J z w > (S) (S) (Y) N . (S) (S) (Q N <! 3 114 Figure 24- Intensity of the carbonate absorption band at 1515 cm- 1 versus the intensity of the carbonate absorption band at 1435 cm- 1 for synthetic Ca ± Mg silicate glasses. Open circles are data from glasses of anc composition; filled circles are di composition; filled squares are 519B composition. 115 I. 2 ....-----.-----..---.......---.------.---- 0 1.0 I E (.) L() 8.. - 0 .6 0 .8 0 .8 r0 ~ Q) (.) 0.6 • c 0 ~ ~ 0 (/) 0.4 • ~ <{ 0 .2 0 0 0 .2 0.4 Absorbance I 515 cm- 1 I .0 1.2 116 Na-carbonate ionic complexes. The hands at 1515 cm- 1 and 1435 cm-1 in the spectrum of anc + co2 glass are also assigned to the antisymmetric stretch of distorted CO~- groups. Similar assignments have been made previously in some infrared (Brey and Green, 1975) and Raman (Sharma , 1979; Sharma et al., 1979; Mysen and Virgo, 1980ab) studies of Ca± Mgbearing glasses. However, in some studies of C0 -bearing Ca-rich 2 silicate glasses, the high energy band at 1515 cm- 1 has either not been identified (Mysen et al., 1975, 1976; Eggler et al., 1979) or has been attributed to a second structural site for co~- in the glass that differs from that of the CO~- producing the low-energy band (Rai et al., 1983). It is reiterated that this splitting mimics to within a few wavenumbers the splitting observed in the infrared spectrum of scapolite, a mineral that contains distorted CO~- groups. Note that Ca+ 2 is associated with the CO~- groups in scapolite (Papike and Stephenson, 1966; Chamberlain et al., 1985). Figure 25 shows infrared spectra typical of synthetic C0 2 -bearing glasses of~, di and 519B compositions. All display absorption bands centered at ~1515 cm- 1 and ~1435 cm- 1 despite the compositional differences between glasses. The same splitting is observed in the diopside and akermanite glasses provided by Dr. S. Sharma (Sharma, 1979; Sharma et al., 1979; Rai et al., 1983). Figures 26 and 27 show that the same doublet is observed in natural basaltic glasses, although the intensities are more than an order of magnitude lower. It is likely that the distinctive splitting of the carbonate band observed in all of these glasses is characteristic of metal-carbonate ionic-complexes in which the metal cation is Ca (and/or Mg), based on the striking contrast 117 Figure 25- Infrared spectra, from 2000 cm- 1 to 1200 cm- 1 , of 519B-6, DIC-8, and ANC-10. Sample thicknesses are noted in Table 10. 118 1.4 w u 1.0 z <( co ~ 0 <./) co <( 0.6 5198-6 0.2 2000 1700 WAVENUMBER,, 1400 cm -1 119 Figure 26- Infrared spectrum, from 2000 cm- 1 to 1200 cm- 1 , of a 150 micron thick piece of basaltic glass TT152 29-1 from the Juan de Fuca Ridge. Absorption bands at 1515 cm- 1 and 1435 cm- 1 due to dissolved carbonate are evident. 120 1.4 TTl52 w u 29-1 1.0 z <! co Ct::: 0 (/) co 0.6 <! 0.2 2000 1700 WAVENUMBER,, 1400 cm -1 121 Figure 27- Background subtracted infrared spectra, from 1700 cm- 1 to 1350 cm- 1 of synthetic glass 519B-6 (top, 59 microns sample thickness) and natural glass ALV519-4-l (bottom, scaled to 1470 microns sample thickness). Both the natural and synthetic glass have similarly shaped spectra. 122 I E E ~ (S) (S) I I ' r- ¢ I (.!) z w _J w (j) L.() u LJ) ' ~ w co L: ::::J z w > <( > <( 3 3 v N (S) (X) CO v. 3JNV'8Cj0S8V' N 123 with the splitting of this band in sodium aluminosilicate glasses. This may reflect specific distinctions between Na and Ca (or Mg) cations or more general distinctions between mono- and divalent cations associated with carbonate groups. In either case, it is concluded that the local environments of the carbonate groups dissolved in all of these Ca ± Mg glasses are similar to the extent sensed by infrared spectroscopy, despite the wide range of silicate composition that they span. It is also noted that the doublets in the synthetic and natural basalt spectra shown in Figures 26 and 27 are slightly broader at their base and split slightly more than in the simpl~r compositions, possibly ref;ecting either the participation of minor Na in carbonate complexes or a wider range of local environments for carbonate in more complex compositions. The splitting of the v 3 band in the systems studied here also appears independent of Al-concentration, in contrast to the conclusion for Na-aluminosilicate glasses in Chapter 1, in which the presence of Al appears to be essential for the degree of splitting observed. Another important difference between the infrared spectra of the Ca ± Mg-bearing glasses studied here and the NaA1Si 2o6-Si02-co2 glasses studied in Chapter 1 is also illustrated in Figure 23. The _sharp band at 2352 cm- 1 in the spectrum of the NaA1Si 2 o + co2 glass was attributed 6 in Chapter 1 to the v molecular 3 antisymmetric stretch of molecular C0 2 • This co2 was concluded to be homogeneously dissolved on the molecular level and not present in fluid inclusions. No molecular C0 2 was detected in any of the glasses reported on here, based on the absence of an absorption band at 2352 cm- 1 • If the molar absorptivity 124 for the 2352 cm- 1 band in sodium aluminosilicate glasses is used, none of the synthetic or natural glasses studied here contains more than 25 ppm molecular C0 2 and most must contain less than 10 ppm. This once again contrasts with the conclusions of Mysen et al. (1975, 1976) based on studies of KBr pellets. The likely reasons for the spurious identification of molecular co2 in those early studies is also discussed in Chapter 1. No bands attributed to C-H vibrations such as those expected for dissolved methane were observed in any of the natural or synthetic glasses. Surprisingly, no bands attributed to methane were detected in the diopside glass provided by Dr. S. Sharma despite the great thickness of the sample (0.039 cm), and the relatively high concentration of methane (0.27 wt.%) that was determined by high temperature mass spectrometry (Rai et al., 1983). If the extinction coefficient or molar absorptivity for the C-H stretch band at ~3000 cm- 1 is of the order of 500 liters/mole-cm, similar to the value for solid methane (Chapados and Cabana, 1972) or the value for methane contained in melanophlogite (~200 liters/mole-cm), a clathrate-type framework silicate (spectrum obtained by Professor G. Rossman from a sample referenced in Cooper and Dunning, 1972), an intense absorption band would have been expected for the high reported methane concentration. Graphite was observed in the diopside glass sent by Dr. Sharma, so the methane found by Rai et al. (1983) in the diopside glass might be produced by reaction during heating of the sample. Alternatively, the extinction coefficient for methane in silicate glasses could be anomalously low, rendering it undetectable. 125 Table 10 lists the band locations and intensities for all of the infrared spectra obtained during this study. Three different intensities are listed for the carbonate bands for each sample: the total area under the doublet, the intensity of the 1515 cm- 1 band above background, and the intensity of the 1435 cm- 1 band above background. Ideally, it might have been best to deconvolve the doublet into two bands and measure the deconvolved intensities. However, since all of the bands have essentially identical shapes, this was not regarded as necessary. Band Intensity Calibrations The intensities of the 1515 cm- 1 band, the 1430 cm- 1 band and the integrated area under both bands are plotted versus total dissolved content in Figures 28a, b and c. Total co 2 contents are based on the amounts of silver oxalate loaded into the capsules. some of these total co 2 The accuracy of co 2 contents has been confirmed via the deuteron beam measurements made on some of these glasses (Table 11). intensities increase regularly with total The band co2 for all compositions and indicate that molar absorptivities and integrated molar absorptivities are roughly independent of composition over the range of compositions reported here. wt.% co 2 • The molar absorptivity is essentially constant below 0.8 At high total C0 2 contents these trends begin to curve, suggesting that molar absorptivities decrease with increasing co2 content at high total C0 2 contents. These deviations from constant molar absorptivity may indicate that c~- groups dissolved in the glasses are interacting electrostatically with one another at high 126 Table 10study. Infrared data obtained for synthetic Ca ± Mg glasses in this Run conditions of glass syntheses are also included. Sample ANC-BL ANC-7 ANC-9 ANC-10 ANC-11 DIC-BL DIC-6 DIC-7 OIC-8 OIC-9 5 l 9B-BL 519B-5 519B-6 p ( kbar) T ("C) c 20 20 20 20 c 15 15 15 15 c 20 20 1450 1450 1450 1550 1550 1450 1625 1625 1625 1625 1315 1550 1550 Duration Loaded C02 Thickness (min.) (wt.%) (cm) Band (cm-1) Intensity (abs.) H20a (wt.%) o.oo 3550 3525 3530 3520 3545 3520 3515 3505 3520 3550 d 3505 3550 0.02 0.12 0.09 0.07 0.13 0.02 0.03 0.02 0.03 0.02 d 0.09 0.05 0.02 0.31 0.17 0.19 0.22 0.03 0.09 0.04 0.06 0.05 d 0.29 0.09 480 60 60 60 60 480 60 60 60 60 480 60 15 0.96 0.12 0.61 0.38 0.00 1.29 0.32 0.51 0.63 o.oo 0.87 0.54 0.0145 0.0057 0.0061 0.0042 0.0068 0.0083 0.0037 0.0050 0.0055 0.0051 0.0126 0.0033 0.0059 Band (cm- 1 ) 1525 1510 1510 1510 1515 1520 1515 1515 1525 1525 Intensity (abs.) d 1.07 0.22 0.55 0.55 d 0.90 0.35 0.62 0.75 d 0.59 0.77 Band (cm- 1 ) 1445 1425 1435 1430 1440 1440 1440 1435 1425 1435 Intensity (abs.) Areab (ca- 1 ) d 1.13 0.23 0.58 0.58 d 0.88 0.35 0.61 0.71 d 0.60 0.12 d 197.22 37 .26 102.26 99.30 d 149.87 60.28 104.53 126.33 d ll6.06 145.26 aWater analyses done using the method of Stolper (1982a) and a molar absorptivity of 60 liters/mole-cm for the infrared absorption band at 35~ ca- 1. Densities are given in Table 9. ~otal area under carbonate doublet. cl atm. dBelow detection limit . ...... N -....J 128 Figure 28- Intensities of the 1515 cm-1 (a) and 1435 cm-l(b) bands and the integrated intensity under both bands (c) versus wt.% co2 loaded into the capsules for synthetic glasses from this study. All intensities are normalized to 100 microns sample thickness. Open circles are data from glasses of anc composition, filled circles are _i!_ composition, and filled squares are 519B composition. Best fit lines for £=375 liters/mole-cm and e*=69500 liters/mole-cm 2 are shown, assuming a density of 2735 g/liter, and are constrained by the origin. 129 2.4 a) €=375 • 2.0 0 Cl.>::- g' E 0 ..0 (.) 1.6 ~ L{) .2 L{) <I: -= 1.2 0.8 0.4 2.4 - 2.0 § E 1.6 0 b) €=375 0 • Cl)_ UI ..0 ,_ 0 I.fl • (.) L{) r0 ..0 '¢ 1.2 <I: - 0.8 0.4 -Si c) §· ~ J3 T 0 €=69500 500 j ~400 '¢ "tJ ~ "tJ 300 j' T 200 0c: 0,_ c: E u L{) L{) • 0 • • • 100 0.2 0.4 0.6 0.8 1.0 Loaded C02 (wt.%) 1.2 1.4 130 Table 11- Comparison of loaded C0 2 (wt.%) with co 2 (wt.%) analyzed using the 12c(d,p 0 ) 13 c nuclear reaction. in Appendix 1. Analysis conditions given 131 ON u - -- o~ w .µ• 0 ....:i .._,, 0 N >o :J < ~ 0 N • 0 co U"\ U"\ ('I") c 0 N u0 ~ • - -'° w .µ 0 0 N <'-' 0 0 0 0 :J • 0 co ('I") 0 ....:i -0 I u ~ I u ~ 132 concentrations or that carbon loss through Pt capsules (Watson et al., 1982; Chapter 1) is more severe in C-rich experiments. It is emphasized that the deviation from linearity in Figure 28 occurs at total C0 2 contents much higher than in most natural basaltic glasses and will not affect the quantitative measurement of dissolved C0 2 contents of such samples. Molar absorptivities of 375 ± 20 liters/mole-cm were determined by the method of least squares (Albarade and Provost, 1977) for the absorption bands at both 1515 cm-l and 1435 cm- 1 , excluding sample DIC~. which contains 1.29 wt. % dissolved co2 • An integrated molar absorptivity of 69500 ± 3000 liters/mole-cm 2 was determined for the total area under the doublet. These values are illustrated in Figure 28 and should only be used for Ca ± Mg glasses containing s0.80% C0 2 by weight. It is noted that these extinction coefficients are much higher than those we determined for the carbonate bands in sodium aluminosilicate glasses. The premises on which this calibration is based are again emphasized: the total co2 contents of the synthetic glasses are accurately known and all of the C is dissolved as carbonate that contributes to the observed doublet. At present, there is no reason to doubt these statements; indeed, the linearity of Figures 28a-c at low concentrations lends them strong support. Conclusions 1. The CO~ dissolved in both natural and synthetic Ca± Mg glass ~ compositions is dissolved essentially entirely as carbonate. carbonate is probably in the form of distorted ionic-complexes This 133 associated with Ca2 + and/or Mg 2 +, and is spectroscopically different from carbonate dissolved in sodium aluminosilicate glasses. than 25 ppm (by weight) molecular 2. No more co2 is dissolved in these glasses. Carbon dioxide analysis of both natural and synthetic Ca ± Mg silicate glasses is possible using infrared spectroscopy. 134 Chapter 4. Thermodynamic Considerations 135 There have been two approaches used to understand the thermodynamics of volatile solution in silicate glasses. The first is a macroscopic empirical approach used in the case of co 2 by Spera and Bergman (1980). While this type of treatment can account for many aspects of the behavior of volatiles in glasses, it neither takes into account nor explains molecular level processes occurring during volatile solution. The second approach, used most recently in the case of water by Stolper (1982b), uses the statistical thermodynamics of molecular level equilibria to understand macroscopic processes. The latter type of treatment is more amenable for analysis of the data generated by infrared spectroscopy and is the approach taken in this chapter. Speciation of Carbon Dioxide in Silicate Melts The results presented in Chapter 3 indicate that C02 in basaltic and Ca ± Mg glasses dissolves predominately as carbonate. This has been suggested previously (Brey and Green, 1975; Brey, 1976; Mysen et al., 1975, 1976; Mysen and Virgo, 1980ab; Eggler and Rosenhauer, 1978; Eggler et al., 1979; Sharma, 1979; Sharma et al., 1979; Rai et al., 1983), although early infrared studies concluded that significant molecular co 2 is also dissolved in comparable glass compositions (Mysen et al., 1975, 1976). The results contrast with the results for glasses in the NaA1Si 2 o -Si02 -co2 system reported in Chapter 1, which contain both 6 dissolved carbonate and molecular co 2 • In this section, I speculate on the causes of the apparently strong compositional dependence of carbon dioxide speciation in silicate glasses. 136 It must be emph~sized that although the . spectroscopic measurements are on co2-bearing glasses quenched from melts, the principal interest of this study is the behavior of C0 2 -bearing melts. Throughout this chapter, I will assume that the speciation of co2 that has been measured in glasses accurately reflects that of the melts from which they were quenched. Support for this assumption comes from numerous comparisons between the properties of melts and glasses at 1 atm (e.g., Taylor et al., 1980; Seifert et al., 1981; Okuno and Marumo, 1982; Aines et al., 1983) and from the fact that the speciation of water in glasses quenched from melts does not appear to be affected by quenching provided that quenching is done rapidly (Stolper et al., 1983). Since the type of thermodynamic treatment used here has been used to understand the dissolution of water (Stolper, 1982b), the case of water is briefly reviewed. Water, like carbon-dioxide, dissolves in silicate glasses and melts both in molecular form and as ions (hydroxyl in the case of water and carbonate in the case of carbon dioxide) produced by reaction between the molecular species and oxygen atoms of the silicate melt (Stolper, 1982a). The homogeneous equilibrium between melt species has been described as follows for water-bearing melts: H2 0(molecular) + o 2 - ( 2) with an equilibrium constant: ( a~H- ) 2 K = ~~~~~~~~~~l (a~ 2 -) (~zO,mol) (3) 137 where am refers to the activity of each of the melt species. At very low concentrations of dissolved water, the concentration of unreacted oxygen is essentially the same as in anhydrous melt, because it has not decreased either by dilution with molecular water or by extensive conversion to hydroxyl groups, and a~2- can be treated as constant. The activities of hydroxyl and molecular water can be approximated by their mole fractions (Stolper, 1982b) and therefore: = (a~2-) (K 1) = constant • (4) Therefore, at low total water concentrations the concentration of molecular water in melts is expected to be proportional to the square of the concentration of hydroxyl groups. This is related to the oft-quoted "square-root" relationship between the fugacity of water in mel.ts and its mole fraction at low total water contents (Stolper, 1982b). A similar treatment can be applied to the equilibrium between molecules of C0 2 and carbonate in melts. The homogeneous equilibrium between melt species can be described with the following sort of reaction: co 23 where Or represents a reactive oxygen. constant given by: (5) This reaction has an equilibrium 138 K (a~ r ) (a~02,mol) C0 2 solution is envisioned as the mixing of co silicate melt species. (6) 2 molecules with C-free These react according to reaction (5), producing carbonate, until the equilibrium constant K is satisfied. It is likely that not all oxygens in silicate melts are equally reactive with molecular co 2 to produce carbonate. Thus, for each distinguishable type of oxygen in the melt, a reaction such as (5) can be written, each having its own value of K (equation 6). Many of the oxygen species in silicate melts may be effectively inert with respect to C0 2 • Kin equation (6) would be approximately zero for these oxygen species. For purposes of discussion, it will be assumed that one reaction of the form of equation (5) is dominant and thus that Or in equation (5) refers to a subset of the oxygens in the melt (e.g., to the bridging, non-bridging, or free oxygens in the sense of Toop and Sammis, 1962). Thus a m will be less than 1 for C02 -free melts and will 0 r probably vary with the silicate composition of the melt. The relationships between the activities of molecular C02 , carbonate, various types of oxygen atoms in the melt, and melt composition are not obvious and are not likely to be simple. In this discussion, the melt is envisioned to be a mixture of co2 molecules, carbonate groups , and oxygen atoms, analogous to the model developed by Stolper (1982b) for molecular water, hydroxyl groups, and oxygen atoms in hydrous melts. In the latter case, an ideal mixture of these species was assumed, giving activities equal to mole fractions. In the case of 139 a mixture of co 2 molecules, carbonate groups_, and oxygen atoms, the fact that the each species has a different size complicates the relationship between activity and composition even for ideal (in the sense that all configurations are energetically equivalent) mixtures. In these situations, volume fractions have greater significance than mole fractions (Flory, 1944; 1953, Chapter XII; Guggenheim, 1952, Chapter X). For the simple analysis given here, however, it will be assumed that activity equals mole fraction for each species, keeping in mind that this is only an approximation. Mole fraction in this context refers to the number of moles of a particular mixing species in the mixture over the total number of species. For example, for a melt consisting of 1 mole of co2 molecules, 3 moles of co~- groups, and l mole of NaA1Si 2 o6 (6 moles of O), Xco2 , mol = 0.1 and Xc 0 2= 0.3. 3 If only low total co 2 contents are considered, such that a m 0 , the r activity of the reactive oxygen species, can be treated as approximately constant and the activities of the carbon-bearing species can be approximated by their mole fractions, equation (6) can be approximated by: ( x~o~-) - - - - - " ' (a~) (K) "' constant • ( X~o , mol) r (7) 2 The observation that for each sodium aluminosilicate composition there is an approximately linear relationship between the concentration of molecular C0 2 and carbonate up to ~1% total C0 2 (Figures 11, 12, and 13) is thus precisely what we would expect based on homogeneous 140 equilibrium between melt species. In Ca ± Mg-bearing systems, xmCO ,mo l 2 is effectively zero. 2 Although the C0 /CO~- ratio increases as silica increases along the NaAl0 2 -Si0 2 join, it is not clear whether this should be attributed to decreasing soda and alumina or to increasing silica. The experiments on silicate compositions that lost sodium during extended drying at 800°C (ab-X, jd-X, and eu-X, Chapter 1) point to a major role for the sodium concentration. Even the relatively minor soda losses experienced by these compositions (resulting in slightly peraluminous bulk compositions) result in dramatic increases in the ratio of molecular co 2 to carbonate in the glasses. Indeed, this was the first indication that these compositions were depleted in soda. The observed variations in speciation with melt composition in the sodium aluminosilicate glasses can consequently be readily rationalized in terms of equations (6) and (7). Suppose that the equilibrium constant K in equation (6) does not vary over the range of compositions that we have studied. It is then clear from equation (7) that at low total dissolved co2 contents, the ratio of co~- to molecular co2 would increase with increasing soda content, provided that the basicity of the m melt (i.e., the activity of reactive oxygens, ao ), increases with r increasing soda content. This would be the case, for example, if co2 molecules reacted primarily with non-bridging oxygens or free oxygens (Toop and Samis, 1962; Pearce, 1964; Wagner, 1975), since in both cases the concentrations of these oxygens are expected to be correlated with 141 Is this effect likely to be the entire explanation of why no molecular C02 was detected in any of the Ca ± Mg silicate glasses reported on here? That is, do the Ca± Mg melts studied here simply have much higher concentrations of reactive oxygens (higher a m) than 0r the sodium aluminosilicate melts? true (Binsted et al., 1985). In some instances, this is probably However, another way to alter the ratio of carbonate to molecular C0 2 in melts is to change the value of K as defined in equation (6) with the silicate chemistry of the melt. Consider the possibility that the reaction of co 2 with oxygen atoms associated with sodium cations has a much lower value of K than the reaction of co 2 with structurally analogous oxygen atoms associated with calcium cations. In other words, suppose that calcium carbonate ion- complexes are stabilized relative to unreacted Cao and co 2 to a greater extent than are sodium carbonate complexes relative to unreacted NaOand co 2 • If this were the case, then it might be expected, as is in fact observed (Brey, 1976; Mysen and Virgo, 1980ab), that as the silicate composition varies from albite to anorthite, for example, the ratio of carbonate to molecular co 2 would increase. However, simple thermodynamic considerations regarding the variation in the K for equation (6) from Na- to Ca- and Mg-bearing glasses are ambiguous. (Compare Spera and Bergman (1980) to Holloway (1981).) At this point, I cannot say with confidence whether the absence of molecular C0 2 in the Ca ± Mg glasses is due to high a 0r in these glasses, to variations in K for equation (6) from Na- to Ca ± Mg-rich systems and as functions of other compositional variables, or to some combination of these effects. Nor can it be determined from the data 142 whether the differences are cation-specific or reflect more general differences between systems rich in monovalent cations and systems rich in divalent cations. It will be possible to definitively address these issues only by detailed study of carbonate to molecular co2 ratios as functions of melt composition. As it is written, reaction (5) does not explicitly take into account the changes in the speciation of the C-free aluminosilicate matrix of the melt that accompany formation of carbonate (e.g. changes in degree of polymerization, etc.) between However, a linear relationship Xc0 2and XCo mol for sodium aluminosilicate melts is expected 3 2' to hold at low total co 2 contents for each silicate composition regardless of the details of the reactions describing the molecular CO 2 carbonate-oxygen interactions. The linearity results from the fact that one carbon dioxide molecule is converted into one carbonate group and from the approximation that the activities of the oxygen species are constant. For example, a linear relationship would also be expected to hold at low co2 concentrations if reaction (5) were replaced by: (8) where 0- and 0° refer to non-bridging and bridging oxygens, respectively, in the sense of Toop and Samis (1962). This is the kind of reaction proposed by Eggler and Rosenhauer (1978) for the production of carbonate groups in diopside and other basic melt compositions. At high total dissolved co concentration of molecular co 2 2 contents the relationship between the and carbonate for all melt compositions 143 is expected to deviate from linearity when the approximations that ao r in equation (6) or that a _ and a in the equivalent expression for 0 0° equation (8) are constant break down. Under these circumstances, it would be expected that the molecular C02 /carbonate ratio would begin to increase with increasing total co , as in the case of the solution of 2 water, where the ratio of molecular water to the square of the hydroxyl concentration eventually begins to increase as water content increases (Stolper, 1982b). Just where the expected deviation from linearity would become noticeable is unclear, but it must be at )1% total co the compositions that have been studied (Figures 11 and 28); 2 in If a reaction such as that described in equation (8) describes the molecular level interactions between molecular co2 and carbonate in sodium aluminosilicate melts, the linear relationship could extend to relatively high total co2 contents. In such a case, the activity of the major species 0° might be changed little by the addition of small amounts of carbonate and the activity of 0-, a minor species, might be buffered by interactions with other C-free melt species and change very little even after substantial molecular co2 has been converted to CO~-. This will be true for any carbonate-producing reaction involving relatively minor oxygen species whose concentrations might be essentially buffered by the concentrations of major species not greatly affected by the addition of carbonate or molecular co 2 • The thermodynamic treatment presented here is consistent with some aspects of the bulk thermodynamic treatment of Spera and Bergman (1980) and, in fact, explains why part of their treatment is successful. and Bergman (1980) analyzed available data on the pressure and Spera 144 temperature dependence of co2 solubility in silicate melts using a thermodynamic model of co2-bearing silicate melts in which the activity of co 2 dissolved in the melt was assumed to be proportional to the mole fraction of total dissolved C0 2 • The results presented in Chapter 1 also suggest that the relationship between activity and composition in C0 2 -bearing melts can be approximated as linear. This follows from the fact that the activity of co 2 in the melt can be assumed to be roughly equal or proportional to the mole fraction of dissolved molecular co2, which is in turn approximately proportional to total dissolved co2 because carbonate and molecular C0 2 contents are proportional. The results can thus explain why Spera and Bergman's "ideal" model for the relationship between a and the mole fraction of total dissolved co2 co 2 is so successful at explaining phase equilibrium data. Spera and Bergman (1980) also used available solubility data to determine VCO , the partial molar volume of C0 2 2 in silicate melts, for a number of melts including albite, jadeite, and nepheline. Recall that most of the dissolved co2 in albite is molecular co2, whereas most in nepheline is expected to be carbonate. Vco Spera and Bergman (1980) found to be essentially identical for these melts despite the difference 2 in their speciation. This observation suggests that the volume change accompanying reactions such as (S) and (8) are minor and is consistent with the observation that co2 speciation is only weakly dependent on pressure. 145 Solubility of cn2 in silicate melts In addition to accounting for changes in the speciation of C0 2 with melt composition, the thermodynamic approach used here can rationalize changes in the solubility of co 2 in silicate melts with changing pressure, temperature, and silicate composition. The equilibrium between silicate melt and co2-bearing vapor can be modelled by the following heterogeneous equilibrium: = C02melt (molecular) • (9) The difference between the volume of co2 in the vapor and the partial molar volume of co2 in the melt (e.g., Spera and Bergman, 1980) requires that with increasing pressure at constant temperature, the concentration of molecular co2 dissolved in a co2-saturated melt of constant silicate composition should increase. This is in accord with the observations in Chapter 2 for albitic melt (Figure 17b). Entropy considerations further dictate that with increasing temperature at constant pressure, the lefthand side of reaction (9) should be favored and the concentration of molecular co 2 dissolved in a melt of constant silicate composition should decrease, again in agreement with the results (Figure 18b) for albitic melt. The results shown in Figure 19 might suggest that the value of K increases slightly, or that reaction (5) favors the formation of co~with both increasing temperature and pressure. These effects are small and again suggest that the volume and entropy changes associated with reaction (5) are small. Recall that at constant pressure and 146 temperature, the activity of reactive oxygen (equation 7) has been treated as constant at low total dissolved co 2 concentrations. noted in the last section that at some high total dissolved It is co 2 content, or will decrease, causing the ratio of carbonate to .molecular co2 to am decrease. The small, but systematic changes in the molecular co /co~- 2 ratio observed under vapor-saturated conditions are attributed to changes in the value of the equilibrium constant K (equations 6 and 7) as functions of pressure and temperature coupled with possible changes or as total dissolved co 2 content varies. The total solubility of co2 is the sum of the dissolved molecular in am C0 2 and carbonate. Its dependence on pressure and temperature can only be evaluated by considering the effects of pressure and temperature on both reactions (5) and (9). Thus, the sum of the negative temperature dependence of molecular co2 solubility for albitic melt and the positive temperature dependence of the ratio of co~- to molecular co2 for albitic melt is a slight negative temperature dependence of the total co2 solubility. A similar competition between decreasing solubility of molecular H2 o and increasing OH/H2 0 ratios with increasing temperature (Stolper et al., 1983) may explain the slight negative temperature dependence of total water solubility in melts (Karsten et al., 1982). The positive pressure dependence of both molecular C0 2 solubility and the ratio of c~- to molecular co2 (except perhaps at low pressure and high temperature, Figure 19b), assures a positive correlation between pressure and total co 2 solubility. A likely consequence of these considerations is a correlation between both the pressure and temperature dependence of co2 solubility and melt composition. For 147 instance, in the case of molten Si0 2 , which probably rlissolves co 2 nearly entirely as molecular co2 (see below)~ the decrease in co2 solubility with increasing temperature can be expected to be more dramatic than the decrease observed for molten albite. Likewise, the increase in total co2 solubility in silica melt with increasing pressure can be expected to be less dramatic than for albitic or jadeitic melts. There is no evidence for abrupt changes in C0 2 solubility or speciation with changing pressure and temperature (as suggested by Mysen et al., 1976 and Holloway et al., 1976) over the ranges of pressure and temperature studied here. Since abrupt changes in melt polymerization (i.e., ag ) might be expected to result in abrupt changes in the ratio r of molecular co2 to co~- (equation 7), it is concluded that no evidence for abrupt changes in melt structure exists in the data presented here. Reaction (9) has an equilibrium constant: a m C02, mol a m C02,mol (10) where a~o 2 , mol and aco 2 refer to the activities of molecular co2 in the The fugacity of C02 in the vapor is fco and 2 of C0 in pure C0 2 vapor is fco • The standard state for C0 2 in the 2 2 melt and co2 in the vapor. vapor is pure co2 vapor at pressure and temperature; i.e., a~o is 1 at 2 all pressures and temperatures for pure co 2 vapor and less than one if the vapor contains other constituents. The standard state for molecular co2 in the melt is a fictive form of pure co2 at pressure and 148 temperature such that the activity coefficient of molecular eo2 in the melt approaches 1 as its concentration approaches zero. of Keo The reciprocal can thus be regarded as a Henry's law constant and the 2 approximation given in equation (7) (a~ 02 ,mol strictly true as long as Henry's law holds. If Keo is only weakly dependent on melt composition for melts near 2 the NaAl0 2 -Si02 join, this would mean that the mole fraction of molecular eo 2 (X~ 0 mol) would be similar in all melts equilibrated with 2' pure eo 2 vapor at a given pressure and temperature. The total eo 2 solubility is the sum of eo2 dissolved as molecular eo2 and carbonate; therefore, since the amount of dissolved carbonate at constant X~ 02 ,mol increases from silica to jadeite (Figure 11), it should be anticipated from the results that eo2 solubility should also increase in this order. This is precisely what is observed (Mysen, 1976). This form of thermodynamic treatment also suggests that it should be possible, given the measured solubility of eo2 in one melt composition at a given pressure and temperature, to use our eo 2 speciation data to calculate the solubility of eo 2 in other melt compositions at the same pressure and temperature. The speciation data suggest that the ratio of the solubility of total eo2 in albite melts to that in jadeite melts should be ~0.54. Mysen (1976) reported that this ratio ranged from 0.65 at 1625°e and 20 kbar to 0.90 at 1450°e and 10 kbar. The agreement is considered to be satisfactory given the possible uncertainties in Mysen's measurements and the absolute eo 2 /eo~measurements presented here. This analysis also suggests that pure silica liquid would dissolve eo 2 essentially entirely as molecular eo 2 ; 149 the solubility of co 2 will then be ~70% of that in albite melt, assuming that at saturation Xco2 , mol is the same in all melts. Based on the data from Chapter 2, this suggests a solubility of 0.46 wt.% co 2 in metastable silica melt at 20 kbar, 1625°C and 0.99 wt.% at 30 kbar, 1625°C. The results of Boettcher (1984) on phase equilibria in the system Si02 -co2 -H2 o also suggest a non-trivial solubility of co2 in silica melts. Water versus co2 speciation There are some important differences between the behavtor of water and carbon dioxide in silicate melts that can be evaluated in light of the previous discussion of homogeneous equilibria in melts. In contrast to the case of carbon dioxide, the speciation of water in melts shows very little dependence on silicate composition. For example, variations in the proportions of molecular water and hydroxyl groups between melts on the NaA1Si o -Si02 join or between these and melts on the CaA1Si 2 08 - 2 6 Si02 join appear to be minor (L.A. Silver and E. Stolper, unpublished results), whereas composition range. co 2 speciation varies dramatically over this Likewise, total water solubility is only weakly dependent on silicate composition, whereas carbon dioxide solubility is strongly dependent on melt composition (Mysen et al., 1975, 1976; Mysen, 1976; Mysen and Virgo, 1980ab; Eggler and Rosenhauer, 1978). Mono- and divalent cations probably play roles in the solubility mechanisms of water in the silicate melts in which they are present. However, these roles appear to be minor compared to those that they play in the case of co 2 solubility, where such cations seem to be essential 150 for the formation of carbonate groups and where energetic differences between different metal-carbonate complexes may dominate both the speciation and solubility behavior of CO~- in melts. In other words, all of the oxygens in silicate melts can be regarded as nearly equally available for reaction with water molecules via equation (2). Although the OH groups formed by such a reaction may be associated with the Na, Ca, etc. in melts that contain such cations, these metal hydroxide complexes seem to be energetically similar. The result is that variations in K1 in equation (4), which is an expression of the variation in the free energy differences between these species in the melt, are minor, and K1 values of .1-.3 seem to satisfy most hydrous systems. In contrast, it seems likely that only certain types of oxygens are available for reaction with co 2 molecules in equation (5). Thus, even if K in equation (6) is constant for a series of melts, the carbonate to molecular co 2 ratio can vary from melt to melt due to variations in the activity of the reactive oxygen species (a~ ) in r equation (6). In addition, the energetics of the reaction (i.e., the stability of the carbonate complex relative to the left hand side of equation (5)) may depend dramatically on the cations present in the melt; that is, K for reaction (5) may be strongly dependent on the monoand divalent cations that are present and perhaps on other compositional factors. The strong composition dependence of C02 speciation and solubility in melts may thus be viewed as a combination of the 0 variations in concentrations of "reactive" oxygen species (i.e., a rin equation (6)) and of the equilibrium constants of the homogeneous 151 equilibria governing the interactions between co 2 and CO~- (i.e., K in equation (6)) with melt composition. Conclusions 1. The speciation of C02 in silicate melts is controlled by the homogeneous equilibria between co2 molecules, carbonate groups, and oxygen atoms in the melts. The approximately linear relationship between molecular co 2 and carbonate concentrations in sodium aluminosilicate melts can be readily understood by considering the thermodynamics of such equilibria. molecular The strong dependence of the co 2 to carbonate ratio can also be accounted for if the melts are considered to contain several different oxygen species, only some of which react with molecular co2 to produce carbonate. The observed compositional dependence of the molecular C0 2 to CO~­ ratio in sodium aluminosilicate melts is consistent with the hypothesis that soda concentration is the major factor controlling the availibility of oxygens for reaction with molecular co 2 to form carbonate and that the carbonate is dissolved as sodium carbonate complexes. The lack of dissolved molecular C0 2 in Ca ± Mg melts is consistent with an increased stability of the carbonate ion in melts with a greater concentration of divalent cations and/or a higher concentration of reactive oxygens. 2. Observed variations in total co 2 solubility with both melt composition and the pressure and temperature of melt equilibration can be understood using the measured species concentrations and a simple thermodynamic model. The approximately linear relationship 152 between the activity of co2 in melts and the mole fraction of total dissolved co2 that has been proposed by Spera and Bergman (1980) can also be readily understood using the results. 3. The strong dependence of C0 2 speciation on melt composition contrasts with the speciation of water in glasses. 4. The results demonstrate that infrared spectroscopic measurements of species concentrations in glasses provide a direct approach to developing a quantitative understanding of the molecular level interactions involved in the solubility of co2 in silicate melts and of the thermodynamic properties and phase equilibria of co2-bearing silicate melts. 153 Chapter 5. The Concentration of Dissolved Carbon Dioxide in Natural Basaltic Glasses 154 Although there is considerable uncertainty about the concentrations of dissolved carbon in oceanic basalt glasses (Delaney et al., 1978; Harris, 1981; Harris and Anderson, 1983; Byers et al., 1983, 1984, 1985; Pineau and Javoy, 1983; Des Marais and Moore, 1984; Mattey et al., 1984; Sakai et al., 1984) and much speculation about the nature of the dissolved carbon (Brey and Green, 1975; Mysen et al., 1975; Eggler et al., 1979), there is general agreement that carbon dioxide plays an important role in the evolution of oceanic basalts. The carbon dioxide- rich vesicles and vapors exsolved from these magmas are the critical evidence for this (Moore et al., 1977; Moore, 1979; Gerlach and Graeber, 1985; Greenland et al., 1985). These vapor compositions are probably the .result of the limited solubility of co2 relative to water in basaltic melts at low pressures and have implications for the understanding of eruptive processes (Gerlach and Graeber, 1985; Greenland et al., 1985), the oxidation state of magmas (Mathez, 1984) and the carbon budget of the oceans and atmospheres (Berner et al., 1983). At deeper levels, heterogeneities in the C0 2 contents of the source regions of basalts may correlate with heterogeneities in the concentrations of other incompatible elements and in isotopic ratios (Schilling et al., 1980; Harris, 1981), and may strongly influence the phase equilibria of these source regions and the petrochemistry of their melting products (Eggler, 1973, 1974, 1978; Wyllie, 1979) C0 2 dissolved as both molecular C02 (Mysen et al., 1975) and carbonate (CO~-) (Mysen et al., 1975; Brey and Green, 1975; Eggler, 1978; Eggler et al., 1979) as well as CO (Byers et al., 1983, 1984) and hydrocarbons (Harris and Anderson, 1983; Byers et al., 1983, 1984) have 155 all been suggested to be dissolved in MORB (mid-ocean ridge basalt) or MORB-like glasses. The structural sites of these dissolved species, even if they are present in basaltic glasses, are matters of speculation. Furthermore, since vesicular C02 , carbonaceous material (either as graphite, carbynes, hydrocarbons or amorphous matter located in microcracks and on vesicle walls (Mathez and Delaney, 1984)) and carbonate alteration (Pineau and Javoy, 1983) are widespread in basaltic glasses, measurements of the absolute concentrations of dissolved C (or C0 2 ) have been difficult. Infrared spectra have consequently been obtained for ~ series of natural submarine basaltic glasses (Table 12) in an effort to resolve this problem. Spectroscopic techniques used are the same as those discussed in Chapter 2. listed in Table 13. Major element analyses of the glasses are Also listed is the name of the scientist from whom each sample was received, the location and depth of the ocean floor from which it was collected, and notes on the crystal and vesicle content of each sample. Density measurements were made on each natural glass with a Berman balance using toluene as a reference liquid and are listed in Table 13. For sample sizes greater than 5 mg, a precision of ±0.2% is estimated, based on the reproducibility of measurements of a quartz standard. This contrasts with a precision of ±2% given in Stolper (1982a) and is revised as the consequence of author error. Unknown uncertainties in the densities of natural glasses are contributed by the variable phenocryst and vesicle contents of the samples. These uncertainties are presumed to be minor both because the samples were 156 Table 12- Infrared data obtained for the natural basaltic glasses used in this study. Sample ALV-519-4-1 ALV-906-R3 ALV-918-Rl ALV-923-R6 ALV-981-R23 TT152-21 TTl52 29-1 TT152 37 TT152 43-19 KN54-2-2 RAMA 26-1 TR123 4D-20 HU81-0l 7 6-11 LOIHI BENTH-2d Kl2-40 Kl697 Thickness (cm) Band (cm- 1 ) Intensity (abs.) H oa (wi.%) Band (cm- 1 ) Intensity (abs.) Band ( cm- 1 ) Intensity (abs.) Areab (cm-1 ) 0.0210 0.0160 0.0180 0.0155 0.0196 0.0126 0.0150 0.0159 0.0181 0.0196 0.0151 0.0185 0.0145 0.0146 0.0117 0.0109 3540 3550 3545 3540 3535 3535 3525 3550 3545 3530 3530 3550 3540 3530 3530 3550 0.34 0.17 0.29 0.18 0.26 0.44 0.29 o. 36 0.28 0.66 0.69c o. 21 0.38 0.88 0.15 o. 27 0.17 0.11 0.17 0.13 0.14 o. 37 0.21 0.24 0.16 o. 37 1.31 0.12 0.29 0.64 0.14 0.26 1520 1520 1515 1525 1520 1525 1530 1530 1525 1525 1520 1535 1515 0.085 0.125 0.061 0.039 0.168 0.078 0.081 0.082 0.090 0.149 0.025 0.053 0.037 e 0.061 0.021 1440 1430 1450 1430 1435 1440 1445 1430 1440 1430 1430 1435 1435 0.084 0.126 0.054 0.040 0.160 0.075 0.082 0.081 0.092 0.134 0.026 0.062 0.037 e 0.064 0.028 14. 35 21.60 8. 73 6.75 28.68 12.23 13.76 12.42 14.94 22.12 2.65 9.28 5.26 e 10.06 3.01 1515 1520 1430 1400 aWater analyses done using the method of Stolper (1982a) and a molar absorptivity of 60 liters/mole-cm for the infrared absorption band at 3550 cm- 1 • Densities are given in Table 13. bTotal area Wlder carbonate doublet. cSample thickness 0.0058 cm .f or H 0 measurement only. 2 dTwo separate pieces run, results agree within 5%. eBelow detection limit. ....... \.J1 -...J 158 Table 13- Major element analyses, densities, depths of eruption and dredge sites of natural basaltic glasses studied. Estimates of vesicle and phenocryst contents and the name of the donor of each sample are also included. ALV519 4- 1 ALV906 2.13 10 . 15 16.44 49.07 0.07 11.65 0.74 0.03 0.16 8.86 99.30 2.91 8 . 15 15.56 49.83 0.07 11.49 1.31 0.10 0.14 9.41 98.97 Source of Analyaia: b c Pbenocryata:j • s Na o Hg Al o 3 Si 2 K20 Cao T10 cr 3 Hn FeOa TOTAL 6 6 6 6 R3 ALV918 RI ALV923 R6 ALV981 R23 TT152 TT152 STA21-35 STA29-l TTl52 STA37 TTl52 STA43-19 KN542-2 RAMA TR123 40-20 HU81-017 STA6-l I LOIHI Kl2-40 BENTH-2 Kl697 26-1 2.61 8 . 04 15.13 50.38 0.10 12.02 1.32 2.26 8.29 15.64 50.34 0 . 07 12.61 1. 21 2.63 7. 11 13. 71 50.92 0.16 10.88 2.02 2.58 7.25 14.39 49.91 0.15 11.77 1. 72 2.79 6.32 13.72 50.98 0.23 10.75 1.93 2.35 8.01 15.10 50.46 3.45 7 . 18 17.05 52.14 0.21 10.35 1.57 0.18 11.04 98.99 0.18 11.95 98.85 0.16 9.94 99.99 8.72 101. 27 0.22 11.94 98.61 2.45 6.65 13.74 47.95 0.44 11.99 2.53 0.13 0.33 11.97 98.18 2.23 6.80 13.60 52.60 0 . 41 6.80 2.50 0 . 19 12.56 100.08 2.42 9.25 18.11 46.78 0.03 10.90 0.82 0.07 0.08 9.78 98.24 2.15 8.79 15.51 49.29 0.05 12.03 1.02 0.18 9.53 99.78 3.15 7.73 17.05 50.06 0.33 10.11 1.16 0.07 0.17 7.62 97.45 2.69 6.39 13.40 51.19 0.21 10.48 2.09 0.16 9.47 99.37 3.05 8.52 16.92 49.70 0.05 11.05 1. 25 0.07 0.15 8.70 99.56 0.19 12.49 101.54 10.90 95 . 84 d d c e e e e f c c e g h s s n s s p • • • p - . - - - . m - . O. IO 12.45 1. 42 -- --- - - Veaiclea:j • .. s n s s s p s s n m • • • s Denaity (g/l): 2790 2830 2810 2770 2790 2840 2810 2820 2830 2760 2730 2870 2680 2850 2800 2810 W.B. Bryan D. s. s. Huenov Newman Newman H. Craig J. Delaney J. J. Delaney Delaney J. Delaney W.B. Bryan Nevaan J . G. J. Schilling Delaney H. Craig D. Huenov Harris J uan J uan Juan de Fuca de Fuca Ridge Ridge Juan de Fuca Ridge Cayman Trough Mariana Trench HidJuan Atlantic de Fuca llidge Ridge Loi hi de Fuca Ridge Galapagos Kilauea Spreading Center 24102520 2310 2360 1950 11061130 Donated by: i.ocatiOD: HidEast East East Atlant. Pacific Pacific Pacific Ridge Rise Rise Rise East Pacific Rise Depth (a) : 2710 2600 2614 2635 2540 22352240 2282 4795 s. a All Fe reported aa FeO. b Bryan and Moore (1977). c Thia study. Hicroprobe conditions: 15 kV accelerating voltage; aample current: 5 nA on brass ; 40-50 alcron beam size. d Ne....,n et al. (1983). e J.R. Delaney, pers. co... W.B. Bryan, pers. coma. g Hawkins and Melchior (1984) . ~Byers et al. (1984). Harr ls (1981). j n - none; • • •inor (<IX); s •some (1% - 5%); p - prevalent (>5%). 2600 2650 D. 2960 ....... \Jl '° 160 generally poor in these constituents and no unusually high or low densities were recorded. Results Figures 26 and 27 show that carbon dioxide dissolved as carbonate can be readily detected in submarine basaltic glasses via infrared spectroscopy. By measuring the intensity of the carbonate band, 'the concentration of co2 dissolved as carbonate can be determined using Beer's law and the value of molar absorptivity derived in Chapter 3 (Figure 28). Carbonate concentrations (expressed as ppm co2 by weight) determined in this way for natural basaltic glasses are · listed in Table 14; values based on the total area under the doublet as well as on the individual band intensities of the 1515 cm-l and 1430 cm- 1 bands are given along with the average of these three values • . The intensity of the 1515 cm- 1 band is the easiest to determine and is probably the most reliable of the individual values tabulated in Table 14. No absorption bands attributable to any other C-bearing species (e.g., molecular co , 2 CO, hydrocarbons) have been detected in natural basaltic glases; I conclude that the carbonate concentration corresponds to the total dissolved carbon content. Also tabulated are dissolved water contents based on the intensity of the 3550 cm-l band (Table 12). No correlation is observed between the dissolved H2 0 and C02 contents or between depth of magmatic eruption and total dissolved co2 content (Figures 29 and 30). The infrared technique that has been developed contrasts in important ways from most of the other techniques in current use for the 161 Table 14- Total dissolved C0 2 concentrations (in ppm by weight) in natural basaltic glasses, based on the intensities of the infrared absorption bands at 1515 cm- 1 and 1435 cm-l and the integrated intensity under both bands. The means of these determinations are given and are also given as C concentrations (also in ppm by weight). 162 Sample co co co2 co2 c 1515 ~- 1 (ppm) 1435 ~- 1 (ppm) area (ppm) avg. (ppm) avg. (ppm) ALV-519-4-1 171 168 156 165 45 ALV-906-R3 325 326 304 318 87 ALV-918-Rl 142 126 110 126 34 ALV-923-R6 108 109 100 106 29 ALV-981-R23 363 344 335 347 95 TT152 STA 21 257 246 218 240 65 TT152 STA 29-1 227 228 208 221 60 TT152 STA 37 216 212 177 202 55. TT152 STA 43-19 207 211 186 199 54 KN54-2-2 325 291 261 292 80 RAMA 26-1 72 74 41 62 17 TRI23 4D-20 118 137 111 122 33 HU81-017 STA 6-11 112 111 86 103 28 LOIHI BENTii-2 a a a a a Kl2-40 220 229 196 215 59 Kl697 81 108 63 84 23 3 Below detection limit. 163 Figure 29- Dissolved C0 2 (ppm) versus dissolved H2 0 (wt.%) for all natural basaltic glasses studied. Data are given in Tables 12 and 14. Sample locations are lahelled. 164 Dissolved C (ppm) 0 1.8 25 50 I I 75 100 I I 0 Mid-Atlantic Ridge Juan de Fuca Ridge •o East Pacific Rise - 1.6 I - • Galapagos Spreading Center 6. Cayman Trough • East Rift Kilauea x Loihi Seamount + Mariana Trough 1.4 ..... + 1.2 I 0~ - 1.0 - - 0 - - x 0.6 - 3: {\J I 0.8 -0 (1) > 0 en en - 0 0.4 0.2 - - 00 •• 0 I 100 - • 0 0 0 •• •• I 0 0 I 200 300 Dissolved C02 (ppm) 400 165 Figure 30- Dissolved co 2 (wt.%) versus depth of erupti?n (where available) for all the natural basaltic glasses studied. indicate uncertainties in eruption depth. labelled. Error bars Sample locations are Rama 26-1, KN54-2-2, and Loihi Benth-2 are not included. 400 I o Mid -Atlantic Ridge -1100 I -- 300 E 0. • Juan de Fuca Ridge o East Pacific Rise • Galapagos Spreading Center 1 East Rift, Kilauea D D I• 75 0. ....__ C\J I 0 u ""O Q) • • 200 > .......... • 0 - Cf) Cf) I 0 < • <D 0... 50 () 0 .......... 0 (/') (/') ""O -0 - 0 3 D 0 D 1001- • • --125 O.__~.._~_._~_,_~~.__~.._~_._~_,_~_____,.__~.._~_._~_. 1900 2100 2300 2500 Depth (meters ) 2700 2900 ...... O'\ O'\ 167 measurement of carbon contents of glasses. First, it is species specific; it probes only for carbon dissolved in a particular form, in this case, carbonate. This species specific nature may not be advantageous if there are forms of dissolved carbon that are not detectable via infrared analysis (e.g., C0 4 groups). However, on balance, this is an advantage since it discriminates between dissolved and extraneous forms of carbon (e.g., co 2 -rich vesicles or crystalline carbonate alteration). resolution. Second, this technique has excellent spatial With modern instrumentation, "spot" sizes as small as a few tens of microns or better can be analyzed. This allows sample homogeneity to be examined and allows vesicle- and crystal-free regions of glass to be studied. Third, it minimizes contributions from surface contamination (Barker and Torkelson, 1975), since the samples are highly polished shortly before analysis and because it is likely that adsorbed carbon species would have a distinctive infrared signature. the technique is non-destructive. Finally, Consequently, the analyst is not concerned with total extraction as in most vaporization techniques and there is no danger of reaction among various species during analysis. It is emphasized again that no molecular co 2 was observed in any of the spectra of natural basaltic glasses. In general, analyses were performed on vesicle-free regions, but in some cases a few bubbles were in the light path. Even in these latter cases, no absorptions due to gaseous C0 2 were observed. Presumably the vesicles are C02 -rich and the absence of a molecular co2 band in these cases probably reflects either the very low co 2 content of a single bubble or the fact that vesicles more than several microns in size are not well sampled by the infrared 168 beam (e.g., the light may be diffracted around the vesicle). noted, however, that significant molecular It is co2 is not contained in submicroscopic vesicles or clusters of vesicles. If molecules of co 2 were present in this form, they would have been readily detected by infrared spectroscopy, since co2 molecules absorb at wavelengths of light (4-5 µm) longer than the diameter of such vesicles or clusters. The accuracy of the analyses is dependent on the calibration shown in Figure 28. The molar absorptivities are estimated to be accurate to better than ±15%; the availability of a wider range of standards might improve this. The precision (i.e., the reproducibility) of these analyses is on the order of selected specimens. < ±5-6% based on repeated analyses of However, there are discrepancies (Table 14) of up to ±15% between analyses of total co2 for each sample by the three different intensity measurements listed in Table 12. Below 100 ppm dissolved C0 2 , these discrepancies can be as large as ±30%. This is a real limitation of this technique since the bands are very broad and become difficult to discriminate from the silicate background when they are less intense than ~0.04 absorbance units using the PE 180 spectrophotometer. This precision at low absorbances may be improved with further development of the FTIR. Many analyses of C0 2 in ocean-floor basaltic glasses have been reported in the literature. A comparison between the analyses presented here and literature values for the same samples is shown in Table 15. Although it is not always clear, most previous authors have also tried to determine dissolved co2 contents; the analyses of dissolved co2 given here are always lower than the values in the literature and in some 169 Table 15- Comparison of co 2 contents determined in this study with values determined in the literature. References are noted. "Dissolved" C02 (ppm by weight) Sample ALV-519-4-1 ALV-981-R23 Kl697 Kl2-40 This study 165 347 84 215 Literature Reference 447 Sakai et al. , 1984 230 Des Marais and Moore, 1984 645 Pineau and Javoy, 1983 436 .D.J. Des Marais, pers. comm. 300 Harris, 1981 310 Harris and Anderson, 1983 1100 Byers et al., 1984 ...... --.J 0 171 cases the deviations are substantial. For example, Pineau and Javoy (1983) report 645 ppm co2 as dissolved carbon in sample ALV-981-R23 as compared to 347 ± 52 (±15%) ppm C0 2 by this analysis. However, Pineau and Javoy's analysis includes the contents of vesicles under 10 microns in diameter, some alteration carbonate, and graphite, which may account for the differences between analyses. Likewise; Sakai et al. (1984) report 447 ppm co2 in ALV-519-4-1 as compared to 165 ± 25 ppm. Sakai et al.'s analysis is dependent on their estimation of the contribution of C0 2 filled vesicles to their analysis of C0 • 2 However, there is no simple explanation of the discrepancy between the value presented here and that of Des Marais and Moore (1984) for the same sample since they took great pains to eliminate co2 from surfaces and vesicles; do note, however, that they could not conclusively rule out a contribution from vesicles in their highest temperature combustion. Harris (1981) and Harris and Anderson (1983) use a vacuum fusion technique to arrive at an analysis of 300 ppm co2 for sample Kl697. 25 ppm C0 2 • The analysis here yields 84 ± This difference might again be attributed to vesiculation, as noted by Harris (1981). Although I again have no explanation, the results are also consistently lower than those given by either Byers et al. (1983, 1984) or Mattey et al. (1984) for comparable basaltic glasses (compare 1100 ppm co 2 in Kl2-40 analyzed by Byers et al. (1984) with 215 ± 32 ppm in this study). While I can make no definitive assessment of all of the reasons for these possible discrepancies, it is again emphasized that the technique used here avoids co2 in vesicles and the problems of C contamination that have plagued previous studies. It is also true that for some problems total co2, including that in vesicles 172 and perhaps even on surfaces, may be more relevant (e.g., pre-eruptive C content); in these cases the analyses presented here may be of limited value since they only measure the dissolved carbonate content. Despite the analyses of Kl697 and ALV-519-4-1, .t he results are most consistently similar to those of Harris (1981), Harris and Anderson (1983) and Des Marais and Moore (1984). for the correlation between dissolved However, there is no evidence co 2 content and depth of eruption that they proposed, and it is concluded that no such general relationship exists (Figure 30). In localized geographic regimes such as the Juan de Fuca Ridge (samples TT152-21, 29-1, -37, -43) ·there is a hint of such a relationship, but explanations other than the pressure dependence of solubility are easy to formulate. Given all of the factors contributing to the actual co2 content of submarine basalts [e.g., Moore, 1977; Moore et al., 1979; Des Marais and Moore, 1984), a simple relationship between depth of eruption (pressure of quenching) and co2 content would be surprising. Conclusions 1. Based on the calibration of the infrared technique using synthetic samples containing known amounts of dissolved co2 presented in Chapter 3, submarine basalt glasses contain from 0-400 ppm dissolved co2, all as carbonate. No general correlation is observed between carbonate content and dissolved water content or depth of magmatic eruption. 173 Conclusions 174 The experiments presented in this study have not lead to a complete understanding of the solution mechanisms of co2 in silicate glasses; the individual conclusions summarized at the end of each chapter represent only an initial effort. necessary data. Obviously, no one technique . can yield all the Infrared spectroscopy, as used in this study, is probably limited by an inability to distinguish subtle changes in the aluminosilicate framework of glasses that result from co2 dissolution. Despite this, a number of fundamental conclusions have been reached via this study: Infrared spectroscopy is a useful beam technique for studying the speciation and concentration of dissolved co2 in both synthetic and natural glasses. The infrared technique offers some advantages over other co2 analysis techniques, most importantly its species specific nature. co2 dissolves as both molecular co2 and co~- in silicate glasses. The relative proportions of the two species vary significantly as a function of silicate composition; the most important compositional variables appear to be the concentration and valence state of the dominant cation in the glass. This influence can only be confirmed with more extensive work in simple synthetic systems (e.g., NaA1Si 3 o8 CaA12 Si2o8-co2) and better knowledge of the structure of silicate glasses. In silicate glasses where both species are present, the relative proportions of molecular co2 and co~- also vary, though less significantly, with the pressure and temperature of melt equilibration and probably, at some high total dissolved co 2 content, with total dissolved co2. 175 In addition, the solubility of co 2 in molten albite as a function of pressure and temperature has been determined. The concentration and speciation of co2 in natural basalt glass has also been studied. Carbon dioxide is clearly a major component of both volcanic gasses and many volcanic glasses and is apparently an important constituent of many basic and ultrabasic magma types. extensive study of the effects of co 2 These observations have prompted on phase equilibria in an attempt to understand the effects of C0 on melt behavior. 2 The results of these studies have been used to infer the structural changes that occur in melts as co2 is added and, ultimately, have been used to construct bulk thermodynamic models. This type of empirical treatment has given little insight into processes occurring on the molecular level. Even if incomplete, the data presented in this thesis demonstrate the superior power of both infrared spectroscopy and the statistical thermodynamics approach for understanding solution mechanisms. The beginnings of a thermodynamic model of volatile solution are presented in this study; this model correctly accounts for some aspects of the behavior of co2-melt interaction (i.e.' solubility relations' changes in co2 speciation as a function of silicate composition and the pressure and temperature of melt equilibration). The model takes into account observable processes occurring on the molecular level which the contrasting treatment can only infer. Little has been concluded in this dissertation regarding the reasons why co 2 effects silicate phase equilibria. The data presented here should ultimately be an integral part of future efforts to understand the effects of co 2 on natural systems. 176 Appendix 1. Measurement of the Carbon Content of Silicate Glasses Using the 1 2c(d,p 0 ) 13 c Nuclear Reaction 177 Beam techniques such as infrared spectroscopy have several advantages over bulk techniques for the analysis of C in geological materials. Beam techniques are non-destructive, the beam may be aimed at specific sites in the sample to avoid some of the contaminants that complicate the interpretation of bulk analytic techniques, and sample homogeneity may be checked. However, infrared spectroscopy may also fail to detect forms of dissolved carbon that are of interest. Electron microprobe techniques have also been used for the qualitative analysis of low level C in silicate glasses (Mathez and Delaney, 1984), but quantitative results are difficult to obtain for a variety of reasons, including the rapid attentuation of low energy X-rays and problems with surface contamination (Goldstein et al., 1976). More significantly, sample currents required for the analysis of C result in mobility of sodium in silicate glasses. Sodium is present in many geologically relevant glasses in relatively high concentrations. Since the mechanism of C dissolution in these glasses is apparently intimately associated with Na o (Chapter 1), electron microprobe techniques must be used with 2 caution. Even if sodium mobility under the electron beam can be minimized, care must be taken to avoid mobilization of the carbon itself. In this appendix, I describe the use of the 12 C( d, p ) 13 C nuclear 0 reaction for the absolute measurement of dissolved C in a series of Grich, synthetic silicate glasses. This beam technique avoids most of the problems encountered during analysis of low-level C contents by electron microprobe and by bulk analysis. It has been used previously for the analysis of C in metals (Cookson, 1979), lunar samples (Filleux 178 et al., 1979), minerals (Oberhauser et al., 1983; Mathez et al., 1984), and meteorites (Heymann et al., 1985). Here the usefulness of the nuclear technique is demonstrated by the measurement of the C content of a variety of the synthetic silicate glasses used in this study. Experimental Procedure The C analyses were performed using a collimated, 0.25 um2 , 1.4 MeV deuteron beam from the Caltech EN tandem Van de Graaf accelerator and a 500 mm2 silicon detector placed at an angle of 160° to the beam. Samples were doubly polished, mounted on tantalum holders with 0 superglue, and coated with a ~so A gold or gold-palladium layer. During analysis, the samples were held under vacuum (10-7 torr) using siliconebased diffusion oil and were in close proximity to an LN2 coldfinger. The glasses luminesced under deuteron bombardment, facilitating beam placement. The principles of C analysis by nuclear reaction have been reviewed previously (Cookson, 1979; Filleux et al., 1979; Oberhauser et al., 1983; Mathez et al., 1984). Some features merit review. Deuterons impinging on the surface of the sample lose energy with increasing depth (x) as they penetrate the sample (Figure 31). At an energy ultimately determined by the cross-section of the 12 C(d,p 0 ) 13 c reaction, the deuterons no longer react to produce 13 c. Similarly, protons emitted by the reaction lose energy as they leave the glass; their ultimate energy as seen by the detector is determined both by the energy of the reactive dE deuteron and the characteristic stopping power ((dx)p) of the glass as expressed by: 179 Figure 31- Representation of the energetics of the nuclear reaction as deuterons impinge on the glass surface. (After Filleux et. al., 1977) Ed O > Ed and Ep max > Ep· 180 dBEAM I I . I I I I I I I Edo GLASS I I I I I I I I I E'd I I I I x 1 181 E (11) p where Ep is the maximum proton energy, Ed is the .deuteron energy at max o the surface, Ed is the deuteron energy at depth x, and 20° = 180° - 160°, where 160° is the angle between incoming deuterons and outgoing protons. The highest energy deuterons (Ed ) thus react at the surface 0 of the glass (x=O); these reactions result in the highest energy protons ) at the detector. As defined here, (~;)P is negative. In max principle, use of the relationship permits depth versus concentration (Ep profiling by monitoring the number of protons emitted from a sample as a function of proton energy. 1.4 MeV deuterons were chosen for their large cross-section, which allows us to reduce irradiation times to a mean of 10 minutes. This deuteron energy should also reduce the sensitivity of the method to surface contamination, but the spectra are clearly resolvable into a surface and volume component, shown schematically in Figure 32. Each glass spectrum was fit as a linear combination of the spectrum of a blank glass of the same composition (or Si02 glass) and of calcite (CaC0 3 ) from the Hilton deposit, San Diego, Calif., weighted by (dE) dX p (Anderson and Ziegler, 1977). This allowed calibration of the integrated intensity of the volume component versus C content. A spectrum of calcite was obtained after every 4 to 6 glass irradiations, to minimize errors due to variations in charge integration. The absolute accuracy of this method is estimated to be ±5% total C. 182 Figure 32- Idealized depiction of proton energy versus protons counted for a thin surface layer (dotted), a thick target (dashed), and a thick target with a thin surface contaminant layer (solid). Filleux et. al., 1977) (After 183 t ...-~ ,, ,, ~ ,,, I w I 0 + I I \ c c ••• ••• t-a_ ~ Q) Q) /.,,, I Ol ', ' 13NN'1HJ/SlNnOJ •••• 0 +0 ~ a. _..... _J w z z <( I ,, u ' 184 Results Figure 33 shows the proton spectra for an NaA1Si o glass 3 8 containing 0.19 wt.% dissolved c. The peak due to the 12c(d,p )13c 0 reaction is noted, as are peaks due to resonant 0 reactions and Rutherford backscattering edges from deuteron scattering on the Na, Al, and Si in the glass. The increase in intensity of the C peak with increasing dissolved C content is shown in Figure 34. Note that no surface component is visible in these spectra, which were obtained during 10 minute runs. During 30 minute irradiations, the surface component slightly increases in intensity with time due to deposition of C on the glass. No change is observed in the volume component, implying no significant loss of C from the samples and confirming the ability to distinguish between the surface and volume components. The fit of a spectrum of NaA1Si 2 o8 + co 2 glass as a linear combination of the spectra of CaC0 3 and Si02 glass is shown in Figure 35. Tables 2 and 11 list the C analysis of each glass (expressed as wt. % C0 2 ) accompanied by the amount of C originally loaded into each glass; the results are shown graphically in Figure 36. Considering both the ±10% uncertainty in loaded C content and the ±5% uncertainty in the C analyses, the amount of loaded C is adequately reproduced. Analyses of these glasses and natural basaltic glasses containing <300 ppm C tend to be inadequate; preliminary work with glasses containing C at less than these levels has yielded inconsistent results. that four factors are probably responsible: Experience indicates a) a surface C film on the samples was not adequately removed before analysis; b) these experiments were not performed under ultra-high (10-9 torr) vacuum conditions; 185 Figure 33- Spectrum of ABC-53. Resonant 16 0(d,p) reaction peaks and the deuteron Rutherford backscattering edges due to Na, Al, and Si are noted. Irradiation time is approximately 10 minutes. 186 _J w z z <! I u "t(f) z :::> 0 u ~ 0 _J CHANNEL 187 Figure 34- C reaction peaks for Si0 (Q), silicon (S), and NaA1Si o 2 2 8 glasses containing 0.085 (1), 0.049 (2), and 0.19 (3) wt.% dissolved c. 188 _J w z z <! I u ' Cf) ~ z ~ 0 u CHANNEL 189 Figure 35- NaA1Si 0 + C0 glass spectrum fitted as a linear 2 2 8 combination of Si0 2 and CaC0 3 spectra. 190 _J w z z ?l ~exp. points I • ~fitting <! u "r(J) z " rl • :::> 0 " " u CHANNEL 191 Figure 36- Analyzed C versus loaded C for all glasses analyzed. 192 0 .26 0.22 - ~ 0.18 -u ~ 0.14 0 w 0 <( 0 0.10 _J 0 .06 0.02 0 0.02 0.06 0.10 0.14 0.18 0.22 ANALYZED C (wt. /o) 0 0 .26 193 observable surface contaminant C becomes significant relative to dissolved C at low levels of dissolved C, making deconvolution of the volume component in Figure 34 quite difficult; c) the presence of contaminant C in microcracks in the glass generated during polishing cannot be ruled out; and d) run times (10 minutes) were not long enough. All four of these factors can he eliminated and it is clear that this method is a useful technique for low-level C analyses in natural glasses. Conclusions 1. The 12 C(d,p ) 13 c nuclear reaction is a useful, non-destructive 0 technique for the analysis of C in geologically relevant glasses. The method is unencumbered by many of the problems associated with other beam techniques, although more care in sample handling is necessary for adequate low-level (<300 ppm) C analyses. The use of a collimated 100 micron diameter beam is eventually possible, accompanied by run times of reasonable length. 194 References Aines RD, Silver LA, Rossman GR, Stolper E, Holloway JR (1983) Direct observation of water speciation in rhyolite at temperatures up to 850°C. Geol Soc Am Abstr with Prog: 15, 512 Albarede F, Provost A (1977) Petrological and geochemical mass-balance equations: an algorithm for least-squares fitting and general error analysis. Comp Geosc. 3: 309-326 Anderson HH, Ziegler JF (1977) Hydrogen Stopping Powers and Ranges in All Elements. New York, Pergamon Press Barker C, Torkelson BE (1975) Gas adsorption on crushed quartz and basalt. Geochim Cosmochim Acta 39: 212-218 Berner RA, Lasaga AC, Garrels RM (1983) The carbonate-silicate geochemical cycle and its effect on atmospheric carbon dioxide over the last 100 million years. Am J Sci 283: 641-683 Binsted N, Greaves GN, Henderson CMB (1984) An EXAFS study of glassy and crystalline phases of compositions CaA1 Si o (anorthite) and 2 2 8 CaMgSi 2 o6 (diopside). Contrib Mineral Petrol 89: 103-109 Boettcher AL (1984) The system Si0 2 -H 2 0-C0 2 : melting, solubility mechanisms of carbon, and liquid structure to high pressures. Am Mineral 69: 823-833 Boettcher AL, Windom KE, Bohlen SR, Luth RW (1981) Low-friction, anhydrous, low- to high-temperature furnace sample assembly for piston-cylinder apparatus. Rev Sci Instrum 52: 1903-1904 Boettcher AL, Burnham c. Wayne, Windom KE, Bohlen SR (1982) Liquids, glasses, and the melting of silicates to high pressures. 127-138 J Geol 90: 195 Boettcher AL, Guo Q, Bohlen S, Hanson B (1984) Melting in feldsparbearing systems to high pressures and the structures of aluminosilicate liquids. Brey G (1976) Geol 12: 202-204 co2 solubility and solubility mechanisms in silicate melts at high pressures. Contrib Mineral Petrol 57: 215-221 Brey G, Green DH (1975) The role of C02 in the genesis of olivine melilitite. Contrib Mineral Petrol 49: 93-103 Brey G, Green DH (1976) Solubility of C02 in olivine melilitite at high pressures and the· role of co 2 in the earth's upper mantle. Contrib Mineral Petrol 55: 217-230 Bryan WB, Moore JG (1977) Compositional variations of young basalts in the Mid-Atlantic Ridge rift valley near lat. 36°49'N. Geol Soc America Bull 88: 556-570 Byers CD, Muenow DW, Garcia MO (1983) Volatiles in basalts and andesites from the Galapagos Spreading Center, 85° to 86°W. Geochim Cosmochim Acta 47: 1551-1558 Byers CD, Christie OM, Muenow OW, Sinton JM (1984) Volatile contents and ferric-ferrous ratios of basalt, ferrobasalt, andesite, and rhyodacite glasses from the Galapagos 95.5°W propagating rift. Geochim Cosmochim Acta 48: 2239-2246 Byers CD, Garcia MO, Muenow OW (1985) Volatile in pillow rim glasses from Loihi and Kilauea volcanoes, Hawaii. Geochim Cosmochim Acta 49: 1887-1896 Chamberlain CP, Docka JA, Post JE, Burnham CW (1985) Scapolite: alkali atom configurations, antiphase domains, and compositional variations. Am Mineral 70: 134-140 196 Chapados C, Cabana A (1972) Infrared spectra and structures of solid CH 4 and CD 4 in phases I and II. Can J Chem 50: 3521-3533 Cookson JA (1979) The production and use of a nuclear microprobe of ions at MeV energies. Nuc Inst Meth 165: 477-508 Cooper JF, Dunning GE (1972) Melanophlogite from Mount Hamilton, Santa Clara County, California. Am Mineral 57: 1494-150 Delaney JR, Muenow DW, Graham DG (1978) Abundance and distribution of water, carbon and sulfur in the glassy rims of submarine pillow basalts. Geochim Cosmochim Acta 42: 581-594 Des Marais DJ, Moore JG (1984) Carbon and its isotopes in mid-oceanic basaltic glasses. Earth Planet Sci Lett 69: 43-57 Duyckaerts G (1959) The infrared analysis of solid substances - a review. Analyst 84: 201-214 Eggler DH (1973) Role of co 2 in melting processes in the mantle. Carnegie Inst Washington Yearb 72: 457-467 Eggler DH (1974) Effect of co 2 on the melting of peridotite. Carnegie Inst Washington Yearb 73: 215-224 Eggler DH (1978) The effect of co 2 upon partial melting of peridotite in the system of Na2 0-CaO-Al 2 o3 -MgO-Si02 -co2 to 35 kb, with an analysis of melting in a peridotite-H2 o-co2 system. Am J Sci 278: 305-343 Eggler DH, Rosenhauer M (1978) Carbon dioxide in silicate melts: Solubilities of II. co 2 and H2 0 in CaMgSi2 o6 (diopside) liquids and vapors at pressures to 40 kb. Am J Sci 278: 64-94 Eggler DH, Mysen BO, Roering TC, Holloway JR (1979) The solubility of carbon monoxide in silicate melts at high pressures and its effect on silicate phase relations. Earth Planet Sci Lett 43: 321-330 197 Faile SP, Roy DM (1966) Solubilites of Ar, N2 , pressures to 10 kbars. co 2 and He in glasses at J Am Ceramic Soc 49: 638-643 Filleux C, Tombrello TA, Burnett DS (1977) Direct measurement of surface carbon concentrations. Proc Lunar Sci Conf 8th: . 3755-3772 Fine G, Stolper E (1985a) The speciation of carbon dioxide in sodium aluminosilicate glasses. Contrib Mineral Petrol 91: 105-121 Fine G, Stolper E (1985b) Carbon dioxide in basaltic glasses: Concentrations and speciation. Earth Planet Sci Lett (in press) Fine G, Johnson T, Stolper E (1985a) The solubility of carbon dioxide in molten albite. Am Mineral (submitted) Fine G, Stolper E, Mendenhall MH, Livi RP, Tombrello TA (1985b) Measurement of the carbon content of silicate glasses using the 12 c(d,p ) 1 3c nuclear reaction. 0 In: JT Armstrong (ed) Microbeam Analysis-1985. San Francisco Press: pp. 241-245 Flory PJ (1944) Thermodynamics of heterogeneous polymers and their solutions. Jour Chem Phys 12: 425-438 Flory PJ (1953) Principles of polymer chemistry. Cornell U Press Gerlach TM, Graeber EJ (1985) Volatile budget of Kilauea volcano. Nature 313: 273-277 Goldstein JI (1976) Practical Scanning Electron Microscopy. ~ew York, Plenum Press Greenland LP, Rose WI, Stokes JB (1985) An estimate of gas emissions and magmatic gas content from Kilauea volcano. 49: 125-130 Guggenheim EA (1952) Mixtures. Oxford U Press Geochim Cosmochim Acta 198 Harris DM (1981) The concentration of basalts. co 2 in submarine tholeiitic J Geol 89: 689-701 Harris DM, Anderson Jr. AT (1983) Concentrations, sources, and losses of H2o, co 2 and S in Kilauean basalt. Geochim Cosmochim Acta 47: 1139- 1150 Hawkins J, Melchior J (1983) Petrology of basalts from Loihi Seamount, Hawaii. Earth Planet Sci Lett 66: 356-368. Heymann D, Vis R, van der Stap C (1985) Carbon concentration mapping in a surface of the Allende meteorite. Abst. Lunar Sci. Conf. 16th: 348-349 Holland TB (1980) The reaction albite = jadeite +quartz determined experimentally in the range 600-1200°C. Am Mineral 65: 129-134 Holloway JR (1981) Volatile interactions in magmas. In: RC Newton, A Navrotsky, BJ Wood (eds) Thermodynamics of minerals and melts. Springer-Verlag, New York: 273-293 Holloway JR, Mysen BO, Eggler DH (1976) The solubility of C02 in liquids on the join CaO-MgO-Si0 -co • 2 2 Carnegie Inst Washington Yearb 75: 626-630 Karsten JL, Holloway JR, Delaney JR (1982) Ion microprobe studies of water in silicate melts: Temperature dependent water diffusion in obsidian. Earth Planet Sci Lett 59: 420-428 Kushiro I (1975) On the nature of silicate melt and its significance in magma genesis: regularities in the shift of the liquidus boundaries involving olivine, pyroxene, and silica minerals. 411-431 Am J Sci 275: 199 Kushiro I (1978) Viscosity and structural changes of albite (NaA1Si o8 ) 3 melt at high pressures. Earth Planet Sci Lett 41: 87-90 Lacy ED (1963) Aluminum in glasses and in melts. Phys Chem Glasses 4: 234-238 Livi RP, Mendenhall MH, Tombrello TA, Fine G, Stolper E (1984) High sensitivity carbon content analyses of geological materials using 1.4 MeV deuterons. Proc Int Symposium on Nuclear Accelerator Physics (Laboratori Nazional di Legarno, Italy) Mathez EA (1984) Influence of degassing on oxidation states of basaltic magmas. Nature 310: 371-375 Mathez EA, Delaney JR (1981) The nature and distribution of carbon in submarine basalts and peridotite nodules. Earth Planet Sci Lett 56: 217-232 Mathez EA, Blacic JD, Berry J, Maggiore C, Hollander M (1984) Carbon abundances in mantle minerals determined by nuclear reaction analysis. Geophys Res Lett 11: 947-950 Mattey DP, Carr RH, Wright RP, Pillinger CT (1984) Carbon isotopes in submarine basalts. Earth Planet Sci Lett 70: 196-206 McKeown DA, Waychunas GA, Brown GE (1984) Na and Al environments in some minerals and a series of glasses within the Na 2 0-Al o -Si02 2 3 system. Geol Soc Am Abstr with Prog 16: 589 Moore JG (1979) Vesicularity and co 2 in mid-ocean ridge basalt. Nature 282: 250-253 Moore JG, Batchelder JN, Cunningham CG (1977) mid-ocean basalt. co 2-filled vesicles in J Volcan Geotherm Res 2: 309-328 200 Mysen BO (1976) The role of volatiles in silicate melts: solubility of carbon dioxide and water in feldspar, pyroxene and feldspathoid melts to 30 kb and 1625°C. Am J Sci 276: 969-996 Mysen BO (1977) The solubility of H2o and co 2 under predicted magma genesis conditions and some petrological and geophysical implications. Rev Geophys Space Phys 15: 351-361 Mysen BO, Arculus RJ, Eggler DH (1975) Solubility of C0 2 in natural nephelinite, tholeite and andesite melts to 30 kb pressure. Contrib Mineral Petrol 53: 227-239 Mysen BO, Eggler DH, Seitz MG, silicate melts and crystals: Holloway JR (1976) Carbon dioxide in Part I. Solubility measurements. Am J Sci 276: 455-479 Mysen BO, Virgo D (1980a) Solubility mechanisms of carbon dioxide in silicate melts: A Raman spectroscopic study. Am Mineral 65: 885- 899 Mysen BO, Virgo D (1980b) The solubility behavior of co2 in melts on the join NaA1Si 3 o8 -CaA1 2 Si 2 o -co2 at high pressures and temperatures: 8 Raman spectroscopic study. A Am Mineral 65: 1166-1175 Nakamoto K (1978) Infrared and Raman spectra of inorganic and coordination compounds, third edition. John Wiley and Sons, New York Navrotsky A, Peraudeau G, McMillan P, Coutures JP (1982) A thermochemical study of glasses and crystals along the joins silicacalcium aluminate and silica-sodium aluminate. Acta 46: 2039-2047. Geochim Cosmochim 201 Newman S, Finkel RC, MacDougall JD (1983) 230 Th-23 8u disequilibrium systematics in oceanic tholeiites from 21°N on the East Pacific Rise. Earth Planet Sci Lett 65: 17-33 Oberheuser G, Kathrein H, Demortier G, Gonska H, Freund F (1983) Carbon in olivine single crystals analyzed by the 12 C(d,p)l3c method and by photoelectron spectroscopy. Geochim Cosmochim Acta 47: 1117-1129 Okuno M, Marumo F (1982) The structures of albite and anorthite melts. Mineral Jour Japan 11: 180-196 Papike JJ, Stephenson NC (1966) The crystal structure of mizzonite, a calcium- and carbonate-rich scapolite. Am Mineral 51: 1014-1027 Pearce ML (1964) Solubility of carbon dioxide and variation of oxygen ion activity in soda-silica melts. J Am Ceramic Soc 47:342-347 Pineau F, Javoy M (1983) Carbon isotopes and concentrations in midoceanic ridge basalts. Earth Planet Sci Lett 62: 239-257 Rai CS, Sharma SK, Muenow DW, Matson DW, Byers CD (1983) Temperature dependence of co2 solubility in high pressure quenched glasses of diopside composition. Geochim Cosmochim Acta 47: 953-958 Russell JD (1974) Instrumentation and techniques. The infrared spectra of minerals. In: VC Farmer (ed) Mineral Soc London Mon 4: PP• 11- 25 Sakai H, Des Marais DJ, Ueda A, Moore JG (1984) Concentrations and isotope ratios of carbon, nitrogen and sulfur in ocean-floor basalts. Geochim Cosmochim Acta 48: 2433-2442 Schilling JG, Bergeron MB, Evans R (1980) Halogens in the mantle beneath the North Atlantic. 178 Phil Trans Royal Soc London Series A 297: 147- 202 Seifert FA, Mysen BO, Virgo D (1981) Structural similarity of glasses and melts relevant to petrological processes. Geochim Cosmochim Acta 45: 1879-1884 Sharma SK (1979) Structure and solubility of carbon .dioxide in silicate glasses of diopside and sodium melilite compositions at high pressures from Raman spectroscopic data. Carnegie Inst Washington Yearb 78: 532-537 Sharma SK, Roering TC, Yoder HS (1979) Quenched melts of ackermanite composition with and without C02 -- Characterization by Raman spectroscopy and gas chromotography. Carnegie Inst Washington Yearb 78: 537-542 Spera FJ, Bergman SC (1980) Carbon dioxide in igneous petrogenesis: I. Aspects of the dissolution of C02 in silicate liquids. Contrib Mineral Petrol 74: 55-66 Stolper E (1982a) Water in silicate glasses: study. An infrared spectroscopic Contrib Mineral Petrol 81: 1-17 Stolper E (1982b) The speciation of water in silicate melts. Geochim Cosmochim Acta 46: 2609-2620 Stolper E, Silver LA, Aines RD (1983) The effects of quenching rate and temperature on the speciation of water in silicate glasses. EOS, Trans Am Geophys Union 64: 339 Strnad Z (1971) Determination of the solubility of carbon dioxide at various partial pressures in soda-silica melts using gas chromatography. Phys Chem Glasses 12: 152-155 203 Taylor M, Brown GE, Jr, Fenn PM (1980) Structure of mineral glasses III. NaA1Si 3 o8 supercooled liquid at 805°C and the effects of thermal history. Geochim Cosmochim Acta 44: 109-117 Tingle TN (1985) An evaluation of absolute carbon concentrations measured by the beta track technique. EOS Trans Am Geophys Union 66: 423 Tomlinson JW (1953) Some aspects of the constitution of liquid oxides. In: Physical chemistry of melts. London Inst Min Metall pp. 22-33 Toop GW, Samis CS (1962) Activities of ions in silicate melts. Trans Met Soc AIME 224: 878-887 Tuddenham WM, Lyon RJP (1960) Infrared techniques in the identification and measurement of minerals. Anal Chem 32: 1630-1634 Verweij H, Van den Boom H, Breemer RE (1977) Raman scattering of carbonate ions dissolved in potassium silicate glasses. J Am Ceram Soc 60: 529-534 Wagner C (1975) The concept of basicity of slags. Met Trans 6B: 405-409 Watson EB, Sneeringer MA, Ross A (1982) Diffusion of dissolved carbonate in magmas: Experimental results and applications. Earth Planet Sci Lett 61: 346-358 White WB (1974) The carbonate minerals. infrared spectra of minerals. In: WC Farmer (ed) The Mineral Soc London Mon 4: Wong J, Angell CA (1976) Glass structure by spectroscopy. 227-284 Marcel Dekker, Inc. Wyllie PJ (1979) Magmas and volatile components. 500. Am Mineral 64: 469-