Compressibility of molten “green glass” and crystal-liquid density crossovers in low-Ti lunar magma

Geochimica et Cosmochimica Acta, Vol. 61, No. 10, pp. 2139-2145, 1997 Copyright 0 1997 Elsevier Science Ltd Printed in the USA. All rights reserved 0016-7037/97 $17.00 + .OO

Pergamon

PII SOO16-7037(97)00055-O

Compressibility of molten “green glass” and crystal-liquid density crossovers in low-Ti lunar magma Department

of Earth and Planetary

J. R. SMITH and C. B. AGEE Sciences, Harvard University, Cambridge, Massachusetts 02138, USA

(Received October 9, 1996; accepted in revised form January 28, 1997)

Abstract-Density measurements of molten Apollo 15 “green glass” have been performed in the pressure range 0.5-3.5 GPa using the floating and sinking spheres technique in piston-cylinder and multi-anvil devices. A density crossover with equilibrium orthopyroxene is predicted for green glass at 3.5 GPa, or -800 km depth in the lunar interior. Equilibrium olivine should be neutrally buoyant in molten green glass at a pressure slightly greater (5 GPa) than the lunar core value of -4.7 GPa. At the olivine-orthopyroxene cotectic ( -2.0 GPa), molten green glass is less dense than both crystalline phases. Thus, the results are consistent with models that propose generation and buoyant rise of green glass magma from the depth of the olivine-orthopyroxene cotectic in the lunar interior. Molten green glass has a compression curve slope of 0.093 g/cc/GPa, along the liquidus, in the pressure range investigated. The values of the Birch-Murnaghan isothermal bulk modulus (K) and the pressure derivative of the bulk modulus (K’) at 1645°C are described by the relationship K(GPa) = 19.5 / (1 - (0.25 - 0.063 K' )) . Combining this relationship with a calculated isothermal bulk modulus value of 18 GPa, derived from 1-atm ultrasonic measurements, gives K’ = 5.3 for molten green glass. Copyright 0 1997 Elsevier Science Ltd 1. INTRODUCTION

tional melting over a range of depths from -4-2 GPa (-1000-400 km). Longhi’s mechanism does not address the high-Ti compositions, although Hess ( 199 1) argued that they can be produced by adiabatically rising diapirs originating at the thermal boundary between the cumulates and the primary lunar interior (the ascent rate of the diapir would be rapid enough to compensate for the tendency of the dense melt to sink). In this study, we assess the importance of crystal/liquid density crossovers in the genesis of low-Ti mare basalts by experimentally determining the compressibility of Apollo 15 Green Glass “Cl’, the most Ti-poor of the “pristine” lunar picritic glasses (Delano, 1986). The pristine lunar glasses are logical candidates for study of mare basalt petrogenesis because they seem to represent the most primitive mare basalt compositions available. The Mg#s (Mg# = [ Mg/ (Mg+Fe)]) of the glasses are generally, though with some overlap, higher than those of the mare basalts (Delano, 1986). Since a magma’s Mg# will decrease as a result of the fractional crystallization of olivine and pyroxene, the glasses’ high Mg#s show that they are better reflections of a primary, unfractionated melt composition than the crystalline basalts, and, as such, logical choices for experiments designed for understanding lunar basalt petrogenesis. Our experiments on molten Apollo 15 green glass give additional insight into the effect of melt composition on compressibility. Circone and Agee ( 1996) found that Apollo 14 black glass was the one of the most compressible ultramafic silicate melts ever measured. Further analysis of the black glass data (Agee et al., 1997) suggests that this high compressibility may decrease rapidly with pressure. One explanation for this behavior could be linked to the high TiOz concentration in molten black glass. If so, comparison with

Circone and Agee ( 1996) showed that high-Ti lunar picrite (Apollo 14 “black glass”) is denser than coexisting liquidus phases orthopyroxene and olivine at pressures of -0.6 and -2.2 GPa, respectively, which corresponds to depths of -120 and -470 km in the lunar mantle. These data argue against high-Ti magmas being generated and transported from regions below the level of crystal-liquid neutral buoyancy and support the original prediction of density crossovers by Delano ( 1990). In fact, it is possible that molten black glass is denser than the solid residue at the orthopyroxene-olivine co&tic pressure (Wagner and Grove, 1997) of -1.5 GPa. This runs counter to earlier models of lunar ultramafic genesis (Ringwood and Kesson, 1976) which produce these magmas through static partial melting of chemically heterogeneous cumulates formed at depths of -400 km from lunar magma ocean cumulates. The high-Ti picrites were thought to have been produced by melting late-stage ilmenite-pyroxene-olivine cumulates after these late stage cumulates sank below earlier formed olivine-pyroxene cumulates. In this class of model, the low-Ti basalts are produced from the ilmenite-free, olivine-pyroxene cumulates. The depth of origin of the magmas in this model is taken as that of the olivine-pyroxene cotectic (2.0-2.5 GPa, or 400500 km), as olivine and orthopyroxene are believed to be the primary residual phases in the lunar mantle. Other models (Hubbard and Minear, 1975; Wagner and Grove, 1997) involve a low-Ti magma being produced at depth and traveling through the heterogeneous cumulates, assimilating different proportions of Fe and Ti from the ilmenite-pyroxene cumulates at relatively shallow depths ( lOO- 150 km). An alternative to these two models is that of Longhi (1992), which would produce low-Ti mare basalts through polybaric frac2139

2140

J. R. Smith and C. B. Agee

molten green glass compressibility, which has low TiOz content, may test this possibility. A forthcoming study on the compressibility of intermediate-Ti Apollo 17 “orange glass” should elucidate this issue further.

Table 2. Sphere compositions. Forsterite ideal Analyzed SiOZ -41203

2. EXPERIMENTAL METHOD The density of a molten Apollo 15 Green Glass ‘C’aualogue was measured using the sinking and floating spheres technique (Agee and Walker, 1988, 1993; Agee, 1992; Circone and Agee, 1996), in which mineral spheres are packed into the top and bottom of a sample capsule with the powdered starting material, the sample is pressurized and melted, and the spheres are driven to one end of the capsule by buoyancy forces. The density of the liquid was bracketed by achieving a “sink” (sphere density greater than liquid density) and a “float” (sphere density less than liquid density) within a narrow range of pressures; experimental temperatures were just above the liquidus of the glass. This technique makes use of the relatively high compressibility of silicate liquids when compared with the crystalline spheres; as the density/pressure curve of the spheres has such a shallow slope, a sphere which is just slightly denser than the liquid at some pressure and temperature conditions will, when pressure is increased, quickly become less dense than the rapidly compressing liquid. The transition from the spheres being more dense to less dense is occasionally marked by a neutral buoyancy result, which yields a direct measurement of liquid density at that point (i.e., sphere density equals liquid density). Experiments from 0.5-3.0 GPa were performed in a Boyd and England (1960) piston-cylinder apparatus and at 3.5 GPa, in a Walker-style (Walker et al., 1990) multi-anvil module. The piston-cylinder was calibrated at 0.8, 1.4, 1.8, and 2.7 GPa (cold-piston in) by differential thermal analysis using the melting point of gold as reference ( Akella and Kennedy, 197 1) Uncertainties in piston-cylinder experimental conditions are estimated as 20.1 GPa and 10°C (Circone and Agee, 1996). The multi-anvil pressures were calibrated for Hertel K05 and Hertel KF-1 tungsten carbide cubes (8 mm truncation edge length) by bracketing the following phase transitions: quartzlcoesite (at lOOO”C,3.0 GPa), fayalite crl y (12oo”C, 5.8 GPa), and coesite/stishovite (12OO”C, 9.3 GPa). Estimated uncertainties for the multi-anvil are as follows: 20.2 GPa and 240°C (Agee et al., 1995; Circone and Agee, 1996). The starting material used was an oxide mixture with composition equal to that of Apollo 15 Green Glass “C” (Delano, 1986; Table 1) . This material was conditioned for 24 h at the Fe-Fe0 0 fugacity buffer, 1075°C and 10’ Pa. Mineral spheres were made by grinding fragments of natural single-crystal an&l&e and synthetic for&rite (Form) (Table 2) in an air mill until spheres of diameter 200-550~ Were-formed. The spheres were cleaned ultrasonically in ethanol before being used.

Table 1. Lunar basalt compositions. Apollo 15 Green ‘C’ low-Ti

Apollo 17 Orange medium-Ti

Apollo 14 Black hieh-Ti

48 0.26 7.74 0.57 16.5 0.19 18.2 8.57 nd nd 100.03

39.4

34 16.4 4.6 0.92 24.5 0.31 13.3 6.9 0.23 0.16 101.32

wt% SiOZ TiOr A1203 Cr203

Fe0 MnO MgG CaO NarO kG T&l

8.63 6.21 0.67 22.2 0.28 14.7 7.53 0.41 0.04 100.07

Fe0 MgG CaO Total wt%

42.71 -

57.29 100

42.29 -

0.15 56.16 0.01 98.61

Anorthite ideal Analyzed 43.2 36.7 20.1 100

44.09 34.81 0.46 18.73 98.09

The capsules used in all experiments were of high purity MO; one mineral sphere was placed near the bottom of the capsule and another, near the top, with a small amount of powdered green glass material between the spheres and the actual top and bottom of the capsule. Powder was packed around the spheres to prevent them from adhering to the capsule walls. Molybdenum capsules have many advantages: they maintain a low f0,; they provide a large sample chamber, which lessens the chance of the spheres adhering to the walls of the capsule and minimizes wall drag effects; and they have a high thermal conductivity, which minimizes the thermal gradient along the length of the capsule. Sample set-up for the piston-cylinder is a 0.5“ talc + Pyrex assembly; the sample capsule is placed inside a dense alumina sheath and rests on a crushable alumina rod, inside a graphite heater. The W3R9,/WZJRer5thermocouple sits directly on top of the capsule lid, - 1.6 mm from the sample center. All experiments were performed “cold piston in,“ i.e., the sample was brought to pressure before being heated. For the multi-anvil experiments, the pressure medium used is a cast, finned, 12 mm octahedron (8 mm truncation edge length on comer of 25 mm tungsten catbide cube) composed of Ceramacast 584, which consists primarily of MgO, SiOZ, and A1203. The octahedra are heated at 1000°C for 3 h to drive off the binder. A 3.3 mm bore hole is drilled through the octahedron and lined with a Re foil (0.02 mm thick) heater, coiled twice. The W3R97/W25Rer5thermocouple is placed between the heater and the pressure medium, centered along the heater. A shallow channel dug in the pressure medium around the bore hole to accommodate the thermocouple aids in assembly and helps the thermocouple to remain as close as possible to the hotspot. The sample capsule, surrounded by a dense alumina sheath, is placed in the center of the octahedron, between two crushable alumina spacers (for more graphical details of this method, see Circone and Agee, 1996). Run procedure was essentially the same in both high-pressure devices. The sample was pressurized, then ramped quickly to the desired temperature, held there for 75- 120 s, and quenched in 24 s by cutting off power to the heater. The temperatures desired in the experiments were slightly above the green glass liquidus because sphere motion can be impeded crystal mush. When the desired temperature is overshot by a significant margin (i.e., -2O”C), the mineral spheres tend to dissolve rapidly into the melt. Potential sphere compositions are restricted to those which have well known equations of state and have the appropriate densities relative to the liquid being studied; thus, sphere material cannot be chosen based only on considerations of compositional stability. After decompression (15-20 min for the piston-cylinder, - 12 h for the multi-anvil), samples were cast in epoxy and sectioned. Once the positions of the spheres were established, the experiments were polished with 6 and 1~ diamond slung and analyzed by electron microprobe The standards used were natural and synthetic silicates and oxides, and data collection conditions were 15 kV and 15 nA. A Bence-Albee correction routine converted data to wt% oxide compositions. Liquid composition was measured by taking approximately 15 point analyses per sample (rastering the beam over an 8 p square area to compensate for inhomogeneities due to quenching) on transects from top to bottom and side to side of the capsule in order to achieve an average composition of the liquid despite inhomogeneous quenching. Mineral sphere compositions were also

2141

Compressibility of molten green glass Table 3. Equation of state parameters. Sphere composition (Mg, Fe)&04

CaA12Si208 (Mg, Fe)SiO,

dKldT

KT @Pa)

(GPaP)

K’

128.54” 125.2” lOO.sh

-0.022” -0.022’ -0.021h

5.3b 4.5’ 5.oh

a1

a2

Pwc

65693E-09 -3.892E-Ogg

-5.8733E-01”

3.229 XF. + 4.417 xp: 2.762’ 3.206 X, + 4.066 X,:

a0

3.1131E-05 1.628E-05 4.847OE-05’

’Suinino et al. (1977). b Kumazawa and Anderson (1969). c Suzuki (1975). d Hazen (1977). ’Hariya and Kennedy (1968). f Hackwell and Angel (1995), Angel et al. (1988). g Gnmdy and Brown (1974), Foit and Peacor (1973). ’ Circone and Agee (1996). i Frisillo and Buljan (1972). JKrupka et al. (1985). ~r,~ was calculated from the expression:

s T

Pzo = Pm (1 +

25

4T)dT)

With thermal expansion, a(T), being defined as a(T) = a0 + a,*T + az*TZ

measured.The spheres were examined for reaction rims, and no significant rims were found. Calculation of sphere density (p) is required in the sink-float technique as it is these densities that provide a constraint on the relative liquid density. Calculations are based on the third order Birch-Mumaghan equation of state

where Kr is the isothermal bulk modulus in GPa, defined as KT = &c

+ dK/dT(T

- 25)

(2)

K’ is the pressure derivative of the bulk modulus, P is pressure, pro is the density of the mineral at temperature and 1 bar, and pT,Pis the density of the mineral at temperature and high pressure. BirchMumaghan parameters for anorthite, olivine, and pyroxene are given in Table 3. (Pyroxene data are used in later calculations of equilibrium residual phase densities.)

3. EXPERIMENTAL RESULTS

Results of experiments are summarized in Fig. 1 and Table 4. Synthetic olivine ( Foloo) spheres sank in Apollo 15 Green Glass “C” at 2.5 GPa and 1600°C and floated at 3.5 GPa and 1720°C. The experiment at 3.0 GPa and 1645°C is considered a neutral buoyancy in that no movement of the spheres was observed. This was taken to indicate that the density of the spheres was so similar to the density of the liquid at that pressure that their velocity resulting from the buoyancy force was essentially zero. In addition to the three experiments described above, a float of anorthite was observed at 0.5 GPa and 136O”C, constraining liquid density to be >2.75 g/cc. It is consistent with the 1 bar calculated reference density of Apollo 15

Green “C” (Lange and Carmichael, 1987). Experiments were also run with natural San Carlos olivine ( -Fopg) at 5.0 GPa to further constrain the compression curve shape. Unfortunately, the San Carlos olivines always dissolved in these experiments. It is interesting to note that olivine spheres become highly unstable above 3.5 GPa in this melt. This is in accord with the fact that olivine is not a liquidus phase at these pressures. Gn the other hand, Agee (1992) and Circone and Agee (1996) showed that garnets are ideal marker spheres for pressures of 5 GPa and greater. Because the present study is primarily focused on density relations at lunar pressures, garnet flotation in green glass (-8-12 GPa) was not attempted. The liquid compositions measured by electron microprobe in the sectioned experiments differed somewhat from the actual Apollo 15 Green “C” composition (Tables 1,4). In experiments where forsterite spheres were used, differences included small depletions in SiOZ, A1203, and CaO ( < 1 wt%), a larger depletion in Fe0 (-2 wt%), a significant enrichment in MgO (-2-3 wt%), and small enrichment MO (0.07-0.8 wt%). Iron-molybdenum exchange occurs between the silicate liquid and the MO capsule, which is the source of the MO contamination and some of the Fe0 depletion (see Circone and Agee, 1996, for discussion of Mooxide partial molar volume). Magnesium oxide enrichment is most likely an artifact of crystal dissolution. In the experiment with anorthite, higher A1203 in the liquid also suggests some crystal dissolution; however, no increase in CaO was observed. The 1 bar calculated densities of the post-run liquids were usually within 0.01 g/cc of the calculated 1 bar densities of the ideal green glass liquid composition. 4. DENSITY CROSSOVERS IN THE LUNAR MANTLE Circone and Agee (1996) showed, as Delano ( 1990) predicted, that density crossovers between high-Ti lunar basalt

2142

J. R. Smith and C. B. Agee Table 4. Summary of experimental conditions, analyzed liquids (one standard deviation in parentheses), calculated ideal and actual I bar liquid densities, and sphere densities at experimental conditions.

1360 float

3 1645 neutral

537A8 Fo 3.5 1720 float

46.55 (0.95) 0.23 (0.04) 8.11 (1.37) 0.53 (0.06) 15.14 (0.66) 18.41 (3.80) 8.11 (1.16) 0.07 (0.07) 97.14 (0.64) 2.83 2.82 2.75

47.16 (0.68) 0.20 (0.03) 6.81 (0.39) 0.53 (0.03) 14.44 (0.59) 20.98 (0.99) 7.32 (0.35) 0.75 (0.72) 98.19 (0.41) 2.78 2.78 3.108

47.18 (0.92) 0.20 (0.03) 6.90 (0.39) 0.54 (0.05) 14.13 (1.13) 20.83 (0.94) 7.52 (0.35) 0.28 (0.61) 97.58 (0.80) 2.77 2.76 3.116

47.27 (0.71) 0.22 (0.03) 7.20 (0.60) 0.54 (0.03) 14.95 (1.08) 20.22 (1.54) 7.70 (0.65) 0.72 (0.5 1) 98.82 (0.59) 2.76 2.76 3.12

176PC An

SiOz TiO, A1203 Cr203 Fe0 MgO CaO MOO* Total p ideal (g/cc) p actual (g/cc) p sphere (g/cc)

0.5

and its equilibrium olivines and pyroxenes occur between 2.0 and 2.3 GPa and 0.5 and 0.8 GPa, respectively. Thus, high-Ti basalts are denser than a harzburgite mantle residue over a wide range of pressures. The density crossovers may have important implications for the origin of high-Ti magmas and for various petrogenetic models of the lunar mantle (e.g., Hess, 1991; Longhi, 1992). Do the same processes influence the genesis of low-Ti basalts? Delano ( 1990) predicted that density crossovers for low-Ti basalts occur too deep in the lunar mantle ( - 1000 km) to have much, if any, effect on their petrogenesis, based on the calculated pressures at which density crossovers occur between Apollo 15 green glass and its equilibrium olivines and pyroxenes. In order to test this, equilibrium compositions of olivines and pyroxenes in Apollo 15 Green “C” glass were calcu-

T= T Liquidus

3.3

I

3.26 5

3.1

B

3

E

2.9-

0”

j

2.c

Molten Green Glass

.h 2.7~

*“loo

2.511 0

1

2

3

18OPC

165PC Fo 2.5 1600 sink

Run ID Spheres P @Pa) T (C) Result

4

5

Pressure (GPa)

Fig. 1. Density vs. pressure diagram summarizing the sink/float data for molten black glass. Arrowhead up is a float, arrowhead down is a sink, filled circle is neutral buoyancy. FolOO = forsterite, An100 = anorthite. Compression curve for molten green glass is shown for liquidus temperatures.

FO

lated using the experimentally liquid distribution coefficient: K

=

D

determined Fe-Mg crystal-

(X X)(X ki,) (X &xx k)

K,, is calculated from a TiO,-dependent relationship fit by Delano ( 1990) to data from the suite of lunar glasses: KD = 0.333 - 0.0071 (mol% TiOz in melt)

(4)

The pressure effect on KD can be evaluated using an equation from Jones ( 1988): KD = 0.320 + 0.120 XMg + 0.106 XFe

- 0.863 XTi - 0.00007P

(5)

Equation 5 shows, as Delano (1990) proposed, that K. hardly varies within the pressures relevant to the moon; equilibrium olivine compositions changed by only 0.1 mol% forsterite from O-4.7 GPa. Values of KD for olivine were assumed to lie within the range of 0.33-0.36, accommodating both Eqns. 4 and 5. Following Circone and Agee ( 1996), a slightly larger range of values for the orthopyroxene KD was allowed (0.32-0.37) as Fe-Mg exchange between pyroxene and silicate melt has not been studied in as much detail as olivine and silicate melt. Resulting equilibrium compositions were well represented by average values of FoS5and En*,. Their densities were calculated across the range of lunar pressures at projected liquidus temperatures of Apollo 15 Green “C” glass and compared to the compressibility of the green glass (Fig. 2). The green glass composition is less dense than its equilibrium olivines and has a neutral buoyancy zone with its equilibrium pyroxenes at 3.5 GPa, or - 800 km depth. At its cotectic pressure (2.0 GPa; Wagner and Grove, 1997)) green glass is less dense than both coexisting crystalline phases. However, at pressures greater than 3.5 GPa, there is the potential for Apollo 15 Green “C” to be neutrally buoyant in an orthopyroxene-rich residue. At 3.5 GPa, the residue is

Compressibility of molten green glass

2143

T=T liquidus 3.4

3.2

F”85.6

M

0 0 n

2.8

Molten Green

opx

T

S-

cross over

I--

Pressure (GPa) Fig. 2. Density vs. pressure diagram giving the molten green glass liquidus compression curve, the compression curves for an average liquidus olivine composition (Fo85), and a range of pyroxene compositions (En85). Also shown is the pressure of the olivine/pyroxene cotectic (-2 GPa) and the pressure at the Moon’s center ( -4.7 GPa). Density crossover exists for molten green glass and pyroxene at 3.5 GPa. The density crossovers between molten green glass and olivine occurs at -5 GPa.

required to be solely pyroxene, for the densities of melt and solid residue are equal. As olivine is added to the residue, the pressure of solid-melt neutral buoyancy increases. For example, at 4.0 GPa (- 1000 km), green glass would be neutrally buoyant with a residue of 37% olivine, 63% pyroxene and negatively buoyant in any residue with ~37% olivine. Residues consisting of more than 82% olivine will be less dense than molten Apollo 15 Green Glass “C” at the Moon’s center. Density crossovers, then, would not seem to be as crucial in the genesis of low-Ti ultramafics as they are for their high-Ti counterparts. The crossover for green glass does not occur at pressures near that of the olivine-pyroxene cotectic, as does that for black glass. Thus, the experimental results are consistent with buoyant rise of green glass magma generated at the depth of the olivine-pyroxene cotectic. However, density crossovers could impact models such as that of Lon-

it can be calculated that the initial density/pressure slope of molten green glass, along the liquidus ( 1300°C at 1 bar; 1645°C at 3 GPa), is 0.093 g/cc/GPa. (Fig. 3). The slope, however, is mostly likely not constant and is expected to decrease with pressure. That is clearly the case for molten black glass, which is also presented in Fig. 3. At a compres-

-._ 0.09s

(lsoo~c)

3.7-

3.63.5-

ghi ( 1992), in which low-Ti mare basalts are produced from inefficient fractional fusion over a range of depths from 42 GPa. Given the right proportions of olivine and pyroxene, a low-Ti magma could be negatively buoyant at the highest model pressures, and a mechanism such as adiabatically rising diapirs (Hess, 1991) would be required to explain the eruption of more than just the high-Ti magmas. 2.8

5. MOLTEN GREEN GLASS COMPRESSIBILITY The information gained from the sink-float experiments can also be used to elucidate the effect of composition on silicate melt compressibility. From the experimental results,

0

.

, 1

,

,

2

3

Pressure

.,

,

.,

4

5

6

(GPa)

Fig. 3. Density vs. pressure diagram comparing linear compressions of molten green and black glass at liquidus temperatures.

2144

J. R. Smith and C. B. Agee

sion of 1.5 GPa, molten black glass has a slope of 0.125 g/ cc/GPa, but at 5.5 and 6.0 GPa, it has decreased to 0.099 and 0.092 g/cc/GPa, respectively. A reasonable estimate for a compression of 3 GPa might place the slope of molten black at a value of 0.115 g/cc/GPa. This is -24% greater than that observed for molten green glass and suggests that black glass is more compressible. Because the liquidus temperatures of green and black glass are similar, this comparison is not biased by the effect of temperature on compressibility. We can also present the green glass data in the form of an isothermal bulk modulus for more general comparisons. Because the most precise measure of liquid density is the neutral buoyancy of forsterite at 1645°C and 3 GPa, we adopt that as the reference temperature. Two known densities, at 1 bar and 3 GPa, permit a family of solutions for BirchMumaghan K and K’ to be calculated. Figure 4 illustrates two such possible solutions: K = 19.5 GPa and K’ = 4; K = 13.0 GPa and K’ = 12. Figure 5 displays the relationship between K and K’ in these solutions which is given by K(GPa)

= 19.5/(1 - (0.25 - 0.063 K’))

Molten Green Glass 1645°C 3ox

dtVdP (K’)

(6)

Hence, large values of K are matched with small values of K’ and vice versa. More data at higher pressure are required to narrow the range of K and K’ possibilities. Until such data are available, we refer to the partial molar compressibility of liquid oxides from Lange and Carmichael ( 1987). We calculate, from these quantities, a 1645°C bulk modulus for molten green glass of 18.0 GPa. Figure 5 shows that the Lange and Carmichael bulk modulus corresponds to a K’ value of 5.3 for this melt. For comparison, we note that the Lange and Carmichael bulk modulus for molten black glass at 1645°C is 16.6 GPa, a smaller value, consistent with our conclusion discussed above that molten black glass is more

Molten Green Glass 1645°C

Fig. 5. Bulk modulus vs. dKldP (K’, i.e., first pressure derivative of the bulk modulus) diagram for green glass at 1645°C. The dotted path defines K-K’ solutions to the third order Birch-Mumaghan equation from this study. A unique solution for K’ is possible if K is known. For illustrative purposes, the figure shows that the calculated molten green glass bulk modulus from Lange and Carmichael ( 1987) gives K’ = 5.3.

compressible than molten green glass. What compositional effects can account for the differences in the melt compressibilities? As Table 1 shows, green glass and black glass differ significantly in SiOz, TiOs, FeO, and MgO concentration, but when summed, the Si02 + TiOl and Fe0 + MgO of the two melts are nearly equal. At 164X, Lange and Carmichael give the following values for the liquid oxides: K = 14.4 for SiOz, K = 10.2 GPa for TiOz, and 19.0 GPa for FeO, and 19.2 GPa MgO. Hence, large variations in SiO*/ Ti02 will have a marked effect on compressibility while the FeO/h4gO does not. Agee et al. (1997) give a K’ value of 6 for molten black glass, assuming K = 16.6 GPa at 1645°C. This is similar to the K’ = 5.3 for molten green glass; however, according to Agee et al. ( 1997), the black glass value may increase to 7 or more at higher pressures (>6 GPa) . Whether this is also the case for green glass remains to be determined. 6. CONCLUSIONS Compressions

Pressure (GPa)

Fig. 4. Density vs. pressure diagram showing some of the possible Birch-Murnaghan fits to the molten green glass data at 1645°C. The examples are K = 19.5 GPa paired with K’ = 4 (weaker curvature) and K = 13.0 GPa paired with K’ = 12 (stronger curvature). The curves start at the l-bar calculated density (Lange and Carmichael, 1987) and pass through the high pressure datum (neutral buoyancy experiment, this work).

of molten Apollo 15 Green glass “C” have

been performed up to 3.5 GPa using the sinking and floating spheres technique. From these data, green glass is predicted to be less dense than its equilibrium olivines over the range of pressures applicable to the lunar interior but to have a density crossover with its equilibrium pyroxene at pressures at 3.5 GPa ( -800 km depth). This indicates the potential for green glass to be negatively buoyant in pyroxene-rich mantle residue above 3.5 GPa. Were such a density inversion

Compressibility of molten green glass to exist, the low-Ti mare basalt petrogenesis model of Longhi

( 1992) might require revision as it produces the magmas from about 4 GPa, or 1000 km. The Birch-Mumaghan isothermal bulk modulus of the green glass has been constrained to a range of K-K’ values. Comparison with the bulk modulus of molten Apollo 14 black indicates that Ti-rich lunar basalts are more compressible than Ti-poor compositions. Acknowledgmenrs-We thank Tim Grove and two anonymous reviewers for useful comments. This work was supported by NASA grant NAGW-3996. REFERENCES Agee C. B. ( 1992) Isothermal compression of molten Fe2Si04. Geophys. Res. Len. 19, 1169-1172. Agee C. B. and Walker D. (1988) Static compression and olivine flotation in ultrabasic silicate liquid. J. Geophys. Rex 93, 34373449. Agee C. B. and Walker D. (1993) Olivine flotation in mantle melt. Eurrh Planer. Sci. Len. 114, 315-324. Agee C. B., Li J., Shannon M. C., and Circone S. ( 1995) P-T phase diagram for the Allende meteorite. J. Geophys. Res. 100, 1772517740. Agee C. B., Smith J. R., and Courtial P. (1997) Density crossovers in lunar picrites. Lunar Planet. Sci. Cof XXVIII. Session: Lunar Maria. Akella 3. and Kennedy G. C. ( 1971) Melting of gold, silver, and copper-proposal for a new high-pressure calibration scale. J. Geophys. Res. 76, 4969-4977. Angel R. J., Hazen R. M., McCormick T. C., Prewitt C. T., and Smyth J. R. ( 1988) Comparative compressibility of end-member feldspars. Phys. Chem. Mineral. 15, 313-318. Boyd F. R. and England J. L. (1960) Apparatus for phase equilibrium measurements at pressures to 50 kilobars and temperatures up to 1750°C. J. Geophys. Res. 65, 741-748. Circone S. and Agee C.B. ( 1996) Compressibility of molten highTi mare glass-evidence for crystal-liquid density inversions in the lunar mantle. Geochim. Cosmochim. Acta 60, 2709-2720. Delano J. W. ( 1986) Pristine lunar glasses: Criteria, data, and implications. Proc. Lunar Planet. Sci. Conf 16, D201 -D212. Delano J. W. (1990) Buoyancy-driven melt segregation in the Earth’s moon, I. Numerical results. Proc. Lunar Planer. Sci. ConJ: 20,3-12 Foit F. F. and Peacor D. R. (1973) The anorthite crystal structure at 410 and 830 degrees C. in high temperature crystal chemistry. Amer. Mineral. 58, 665-675.

2145

Frisillo A. L. and Buljan S. T. (1972) Linear thermal expansion coefficients of orthopyroxene to 1000°C. J. Geophys. Res. 77, 7115-7117. Grundy H. D. and Brown W. L. ( 1974) A high-temperature X-ray study of low and high plagioclase feldspars. In The Feldspars; Structure, Bonding and Order-disorder, pp.l62-173. Hackwell T. P. and Angel R. J. (1995) Reversed brackets for the Pl * 11 transition in anorthite at high pressures and temperatures. Amer. Mineral. 80,239-246. Hariya Y. and Kennedy G. C. (1968) Equilibrium study of anorthite under high pressure and high temperature. Amer. J. Sci. 266,193203. Hazen R. M. (1977) Effects of temperature and pressure on the crystal structure of ferromagnesian olivine. Amer. Mineral. 62, 286-295. Hess P. C. (1991) Diapirism and the origin of high TiO? mare glasses. Geophys. Res. Len. 18, 2069-2072. Hubbard N. J. and Minear J. W. (1975) A chemical and physical model for the genesis of lunar rocks: Part II. Mare basal&. Lunar Sci. VI, 405-7(abstr). Jones J. H. ( 1988) Partitioning of Mg and Fe between olivine and liquids of lunar compositions: The role of composition, pressure and Ti speciation. Lunar PIunet. Sci. XIX, 561-562 (abstr). Krupka K. M., Robie R. A., Hemingway B. S., Kerrick D. M., and Ito J ( 1985) Low-temperature heat capacities and derived thermodynamic properties of anthophyllite, diopside, enstatite, bronzite and wollastonite. Amer. Mineral. 70, 249-260. Kumazawa M. and Anderson 0. L. (1969) Elastic moduli, pressure derivatives, and temperature derivatives of single-crystal olivine and single-crystal forsterite. J. Geophys. Res. 74, 5961-5972. Lange R. A. and Carmichael I. S. E. (1987) Densities of Na,OKz0-Ca0-Mg0-Fe0-F~03-Al~O~-TiO~-SiO~ liquids: New measurements and derived partial molar properties. Geochim. Cosmochim. Acta 51,293 l-2946. Longhi J. ( 1992) Origin of picritic green glass magmas by polybaric fractional fusion. Proc. Lunar Planet. Sci. 22, 343-353. Ringwood A. E. and Kesson S. E. (1976) A dynamic model for mare basalt petrogenesis. Proc. Lunar Sci. Con& 7, 1697-1722. Sumino Y., Nishizawa 0.. Goto T., Ohno I., and Ozima M. (1977) Temperature variation of elastic constants of single-crystal forsterite between -190” and 400°C. J. Phys. Earth 25, 377-392. Suzuki I. ( 1975) Thermal expansion of periclase and olivine, and their anharmonic properties. J. Phys. Earth 23, 145-159. Wagner T. P. and Grove T. L. (1997) Experimental constraints on the origin of lunar high-Ti ultramafic glasses. Geochim. Cosmochim. Acia 61, 1315-1327. Walker D., Carpenter M. A., and Hitch C. M. (1990) Some simplifications to multi-anvil devices for high pressure experiments. Amer. Mineral. 75, 1020-1028.