- Email: [email protected]
Solubility and partitioning of Ne, Ar, Kr and Xe in minerals and synthetic basaltic melts
0016-7037/92/$5.00 + .oO
(rr,,ct,,m,cu PI
Solubility and partitioning of Ne, Ar, Kr, and Xe in minerals and synthetic basaltic melts C. L. BROADHURST,I,* M. J. DRAKE, ’ B. E. HAGEE,’ and T. J. BERNATOWICZ’ ‘Lunar and PlanetaryLaboratory,Universityof Arizona, Tucson, AZ 8572 I, USA *McDonnell Centerfor the Space Sciences, Washington University, St. Louis, MO 63 130, USA
(ReceivedApril 16, 199I ; accepted in revised form November 27, I99 1) Abstract-We have measured the solubilities of Ne, Ar, Kr, and Xe in natural samples of anorthite, diopside, forsterite, spinel, and in synthetic basaltic melts. The minerals and melts represent equilibrium pairs in the Fo-An-Di-Si02 system. Samples were suspended in individual crucibles in a one-bar flowing mixed noble gas atmosphere at 1300or 1332°C for times ranging from 7-l 8 days. The solubilities in the minerals increase with increasing noble gas atomic number, and typical solubility values are surprisingly high. Samples of a particular mineral (i.e., anorthite) that came from different localities yield distinctly different and reproducible solubility values. This indicates that properties intrinsic to individual samples can influence solubility. Noble gases are likely to be sited in lattice vacancy defects. In contrast, the solubilities in the melts decrease with increasing atomic number. Our data overlap the low end of the range defined previously for natural basalts; however, the dynamic range of the values from Ne to Xe is not as great as for the natural melts. Solubilities correlate well with melt molar volume but poorly with density. The partition coefficients increase with increasing noble gas atomic number for all mineral/melt pairs. Such trends imply that the terrestrial planet atmospheres were not derived from partial melting of chondritic source material. Significant fractionation of Kr relative to Xe is not observed, ruling out an origin for Earth’s “missing” Xe via magmatic fractionation. Partition coefficient absolute values are frequently near or greater than unity. For example, the ranges for five diopside measurements are the following: DNe,,, 3.2-47. This indicates that magmatic transport 0.013-0.37; D&l,, 0.15-0.84; DKrll, 0.31-2.4; and Dxell, is not an efficient mechanism for degassing the terrestrial planets. Our results favor a catastrophic origin for the atmospheres.
INTRODUCTION
short time period relative to the age of the planet. Very early (i.e., pre-3.8 Ga for Earth) intense volcanic activity is an example of such. In addition, early planetary atmospheres may have consisted of captured solar nebula gas; or several hundred bars of steam (LANGE and AHRENS, 1982; MATSUI and ABE, 1986a,b; ZAHNLE et al., 1988) generated by accretionary impacts. These dense atmospheres were subsequently lost by mechanisms such as massive hydrodynamic escape (WATSONet al., 198 1; ZAHNLE and KASTING, 1986; HUNTEN et al., 1987; ZAHNLE et al., 1990a) or impact erosion ( LANGE and AHRENS, 1982; WALKER, 1986; AHRENS et al., 1989; MELOSH and VICKERY, 1989). Noncatastrophic degassing mechanisms are those which occur over the lifetime of the planet: solid state diffusion, mass transport by convective overturn, and magmatic transport. As discussed previously ( OZIMA and PODOSEK, 1983; BROADHURST, 1989; BROADHURST et al., 1990), diffusion is not effective. Mass transport by convective overturn can be demonstrated to be important only for Earth. Even then, partial melting resulting from adiabatic decompression of the rising mantle plumes is probably the major mechanism of noncatastrophic degassing. Although the production and transport of basaltic magma has long been recognized as an important degassing mechanism, until recently experimental data concerning noble gas partitioning were completely lacking. Noble gases were speculated to behave as uniformly incompatible elements during partial melting events, but quantitative melting models could not be constructed.
THE SOLUBILITY,PARTITIONING,and siting of Ar in minerals
and synthetic
basaltic
melts
were discussed
previously in is a companion to those results, extending them to include Ne, Kr, and Xe. The impetus for the partitioning experiments is the observation that the noble gas abundance patterns and primordial/ radiogenic isotopic component ratios of Venus, Earth, and Mars differ from each other as well as from the chondritic meteorites. The most obvious difference between the terrestrial planet and meteoritic trends is in the relative abundances of Xe. The ‘32Xe/84Kr ratio in the Earth’s atmosphere is over an order of magnitude less than the values typical of carbonaceous chondrites ( MAZOR et al., 1970) and ordinary chondrites ( Z;~HRINGER, 1968). Apparently, this “missing Xe” does not reside in any known surface reservoir, including oceans (OZIMA and POWSEK, 1983), sedimentary rocks ( BERNATOWICZ et al., 1984), polar ice (BERNATOWICZ et al., 1985), and deep-sea amorphous silica ( MATSUDA and BROADHURSTet al. ( 1990). This contribution
MATSUBARA, 1989).
Variations in planetary atmosphere noble gas abundances may be due to primordial heterogeneities or may have been determined by the interplay of the degassing mechanisms operating on each planet. The degassing of planetary mantles can be classified as catastrophic or noncatastrophic. Catastrophic degassing events are those which occur in a very * Present address: Code 693, NASA Goddard SpaceFlightCenter, Greenbelt,MD 2077 I, USA. 709
C. L. Broadhurst et al.
710
HIYAGON and OZIMA ( 1986) performed the first experimental equilibrium noble gas partitioning study. They investigated olivine/ melt and diopside/ melt systems and found that noble gases were only somewhat incompatible, with partition coefficients increasing from He to Xe. In this study, we report an extensive series of partitioning experiments with anorthite, diopside, forsterite, spinel, and synthetic basaltic melts. Our results for all four mineral/melt pairs show the same partitioning trend as found by HIYAGON and OZIMA ( 1986); however, we find the surprising result that noble gases are often compatible in minerals. Since such partitioning behavior greatly complicates efforts in magmatic modelling, we offer only limited interpretations of the data. Instead, we present the results of many experiments, including those in which experimental errors occurred. We wish to demonstrate that noble gases will be fractionated in mineral/melt systems yet they cannot be expected to behave in the relatively predictable fashion that characterizes ionic trace elements. EXPERIMENTAL
PROCEDURES
The following section summarizes the experimental technique; detailed presentation is given in BROADHURST ( 1989) and BROADHURSTet al. ( 1990). Information concerning the starting material sources, compositions, and preparation methods is also given in these references. The additional melt compositions prepared for this study are given in Table 1. Mineral starting materials were eleven natural samples very close to endmember composition (five anorthites, three diopsides, two spinels, and one forsterite). We chose these samples based on availability and purity. Most importantly, tests with candidate mineral samples containing only a few wt% Fe0 gave poor results; strong color changes in the mineral powders indicated that oxidation/reduction was occurring. Since we did not want to introduce the additional variable of controlled oxygen fugacity, only nearly Fe-free material was used. All experiments used cl0 pm powders, except for samples DI 122742 in experiment NGX-VI and FO 12243 in experiment NGXV, which were ~38 pm (400 mesh) powders. The 110 pm powders were produced by grinding the minerals in a sapphire mortar and forcing them through a series of sieves. After several regrindings, a TABLE 1. Bulk compositionsof glassesused for synthetic melt stating materials in oxide weight percar. Microprobe analysesperformed at the University of Atiz~na. All analysesassumeFe0 = Fe total. EQ phace representsthe mineral phasethat the melt compositionis in quilihrium with in the An-Di-Fo-Si02 systemat 1300°C: AN = anotthite, DI = diopside, FO = forerite.
small amount of powder was forced through a 10 pm nylon filter mesh in an acetone slurry. The residue was evaporated to produce the starting material. The ~38 pm material was ground to pass through a 400 mesh sieve but did not pass through the nylon filter mesh. Optical examination of the powders showed them to be roughly spherical and not appreciably elongated or tabular. Silicate liquid starting materials were prepared as melts from oxide powders and quenched to glasses. The compositions are in the system forsterite-diopside-anorthite-silica ( PRESNALLet al., 1978) and were based on the experiments of MALVIN and DRAKE( 1987). The AN, DI, and FO series glasses are in equilibrium at 1300°C with the minerals anorthite, diopside, and forsterite. respectively. Again, these synthetic compositions were chosen in order to avoid FeO/FelOs equilibria problems and to reduce the complication of the experimental system so that we could focus clearly on the noble gas mineral and melt solubilities. The minerals and glasses were loaded into separate open cylindrical Pt crucibles. The crucibles were then suspended in a vertical alumina muffle tube gas-mixing furnace. We used a commercial gas mixture ( Alphagaz Liquid Air) consisting of 5.00% Ne, 93.0% Ar, 1.00% Kr, and 1.OO%Xe by weight. The gas flowed through a metering system at 90 cc/min into the top of the furnace and exited such that the pressure in the muffle tube was constant and less than l-2 torr over ambient. After a specified length of time, the samples were quenched by quickly removing the entire suspension system from the furnace. Experiments NGX-IV and MT were quenched in air and then in water. The suspension bucket was dipped in water, but the samples were not submerged or othenuise wetted. Under no circumstances was there any evidence of reaction of the samples with the crucibles. The first four experiments (NGX-I to IV) were run at 1300 + 2°C for eighteen days. At this temperature, the melt compositions were close to saturation in their respective mineral phases, and some were not analyzed because they grew crystals during the experiments. The last three experiments (NGX-MT, and V and VI) were run at 1332 f 2°C for seven and nine days, respectively. This slight temperature increase ensured that the melts were completely liquid, and it is unlikely to have significantly affected the results as noble gas solubilities have been demonstrated to be rather insensitive to temperature (JAMBONet al. 1986; Lux, 1987; WHITE et al., 1989). Experiment NGX-MT marked the only significant variation of the experimental technique. This experiment was basically a test which confirmed that we could avoid the crystallization problem mentioned above by running at 1332°C. As this was an unanticipated final usage of the first mixed noble gas tank, the gas flow was reduced to 50 cc/ min. After NGX-MT and prior to experiment NGX-V, the furnace was refitted with new heads, muffle tube, o-rings, and gas inlet lines, and a second tank of the noble gas mix was started. Diffusion A major concern of this experimental approach is ensuring that we attain solubility equilibrium in the minerals. Accurate K/Ar dating has long been the major impetus for Ar diffusion studies. FOLAND (1974) gives D = 0.00982 exp[-43.8/RT] cm’/sec for Ar in feldspar (spherical geometry). At 13OO”C,the characteristic diffusion time for 10 pm grains is 62 sec. This stands in contrast to eighteen days, or 1.6 X IO6 sec. In addition, BROADHURSTet al. ( 1990) presented a successful reversal and rate study experiment for Ar. The rate study, in particular, indicated that solubility equilibrium is achieved within two days. Unlike Ar, diffusion rates for Ne, Kr, and Xe in minerals are not available. We can, however, get a feel for the relative diffusion rates by examining data from carbon ( WACKERet al., 1985), diamond (OZIMA, 1989), silicate melts (HAYATSUand WABOSO,1985; JAMBON et al., 1986; Lux, 1987), and glasses (MATSUDAet al., 1989; CARROLLand STOLPER,199 1) . It is clear that Ne will diffuse more rapidly, and Kr and Xe more slowly, than Ar. If we make the conservative estimate that D ( 1300°C) values for Kr and Xe are one to two orders of magnitude less than those for Ar, we have still allowed ample time to achieve diffusional equilibrium. Even if solubility equilibrium has not been achieved for the heavier noble gases, we do not feel that our results are compromised. On the contrary, our solubility values would then be lower limits only, which make the results even more surprising.
711
Solubility of noble gases in minerals and melts
NOBLE GAS ANALYTICAL PROCEDURES Gas cleanup and separation was similar for all the experiments, and the procedures used have been developed from many years of sensitive noble gas measurements at Washington Univ. Full details aregivenin BROADHURST( 1989)and BROADHURST etal.( 1990). However, noble gas measurement techniques in the first two mixed noble gas experiments (NGX-I, II) differed from the remaining five (NGX-III to VI and MT). In the first two experiments, a Dycor model MZOOMquadrupole mass spectrometer (QMS) was used for the analysis of Ar. The absolute Ar ion beam signal was dete~in~ by at least sixteen measurements of the *Ar signal and extrapolation of these data to the time of initial generation of the Ar ion beam. The sensitivity of the QMS depends upon the total ambient gas pressure. and because this pressure is due almost entirely to Ar, we calibrated the sensitivity over a wide range of pressures using known quantities ofatmospheric Ar. Neon, Kr, and Xe were measured in the first two experiments with a Reynolds-type instrument operating in electron multiplier mode. Abundances were determined by peak height comparisons with standard air pipettes. In this procedure, only a fraction of the sample Ne was used (about 10%). and this amount was obtained by trapping the other noble gases on activated charcoal at LN2 temperature and partitioning the Ne between accurately calibrated volumes. The entire sample amount of Kr and Xe was sub~quentiy collected on another charcoal linger and unironed between two volumes, the larger of which was used for Xe m~surement and the smaller for Kr. The gas abundances were determined from the calibrated instrumental sensitivities by extrapolating the measured reference isotope (“Ne, s“Kr, “*Xe) signals to the time of initial generation of the ion beam. The first two experiments also used a two-step extraction, with gases measured at -800°C and - 1550°C. Two steps were also used for experiment AAN in BROADHURST et al.(1990).In all three experiments, very little gas was released at the lower temperature step, so we discontinued this very time-intensive practice for the remainder of the analyses. At this point, it was beginning to be clear to us that properties intrinsic to individual mineral samples could influence noble gas solubility, and we wanted to continue with more precise measurements in order to reduce the overall abundance uncertainty as much as possible. Beginning with ex~~rnent NGX-III, we determined all noble gas abundances using isotope dilution techniques with a mixed noble gas spike. Isotopic ratios were measured with a Reynolds-type instrument operating in static Faraday cup mode. Data acquisition (for either mode) was done by computer-controlled peak-switching, and fifteen scans of each gas spectrum were taken. During Ne runs, the gas was exposed to charcoal at LN2 temperature to minimize 40Ar++ and H,‘sO+ interferences at “Ne and CO?’ at **Ne. The sinelvcharge2 species were also monitored duringNe runs, and all Ne &a were corrected for interferences from these ions. For the isotope dilution procedure, we first collected all the sample noble gases on activated charcoal cooled by a He cryostat to 10°K. We then mixed an aiiquot of spike gas with the sample gas. The noble gas spike contained the following composition: 22Ne= ( 1.394 f 0.06 1) X lo-’ ccSTP with 22Ne/2?Je = 5.612 rt 0.048; %Ar = (2.980 20.043) X 10m7ccSTP with 40Ar/MAr = 0.0987 It 0.0013; *& = 100% and f 1.955 + 0.0169) X 10v9 ccSTP; lJ6Xe = (I.547 t0.0165) X 10e9 ccSTP with ‘36Xe/‘3ZXe = 7.439, ‘29Xe/‘32Xe = 0.9841, ‘3’Xe/‘3*Xe = 0.7885. Other Ne, Ar, and Xe isotopes are present in much lower abundances and are not critical to the analysis. The experimental uncertainties in noble gas abundances determined by isotope dilution are ~2% for Ar, Kr, and Xe, and ~4% for Ne, exclusive of uncertainties due to corrections for instrumental background (“blanks”). Instrumental blank levels were determined between individual sample runs for experiments in which the gas abundance, and consequently instrumental memory, was high. They were determined between groups ofseveral runs for cases where both blanks and sample gas abundances were low. In practically all cases, the blank corrections were less than 10% of the measured Ar abundances, and a 50% uncertainty in the blank corrections was assumed. For abundance dete~inations based upon in~Nrnen~ sensitivity, the blank correction
and uncertainty have been included in the overall uncertainty (25% for the QMS and 15% for Reynolds instrument determinations). For the isotope dilution experiments, the overall abundance uncertainty including blank correction is given. RESULTS For reasons of clarity, results for each mineral are presented separately in the text below, followed by the melt data. In the co~es~nding Figs. la-f, the different samples are identified by their catalog numbers, followed by a Roman numeral indicating the experiment number. Note that the ordinates are scaled differently. Measured isotopic abundances and solubilities for the set of experiments are given in Tables 2 and 3, respectively. Anorthite Noble gas solubilities in anorthites are shown in Fig. la. Samples AN HO 228, AN 1 l-7779, and AN WU 617 were run together in experiment NGX-I in an attempt to reproduce the results from the 99.95% Ar synthesis experiment AAN (BROADHURSTet al., 1990). The behavior was consistent: HO 228 again showed very high soiubilities; 1l-7779, intermediate values; and WU 6 17, the lowest values. The solubility trend for WU 617 was reproduced in experiment NGX-II, except for a slight decrease in the Xe/Kr ratio. All of these samples show a clear trend of increasing solubility with increasing noble gas atomic number. Sample AN 4079 III shows a nearly flat pattern, which reflects some contamination with glass material, probably introduced to the sample as the set of crucibles was removed from the suspension bucket directly after the experimental run. Optical examination of AN 4079 III powder showed no signs of partial fusion during the synthesis experiment. although the powder was somewhat packed together. This sample was also examined by transmi~ion electron microscopy (TEM) and did contain some grains that did not yield a diffraction pattern ( BROADHURST, 1989 ). This topic is discussed further below.
Diopside Diopsides also show an overall trend of increasing solubility with increasing noble gas atomic number (Fig. 1b). DI 122742 yielded consistently higher solubiiities than DI R 18684. The single analysis of DI M2506 cannot be distinguished from the range represented by R 18684, and both of these samples had some of the lowest solubilities of all the minerals studied. The solubilities in the 38 pm DI 122742 material run in experiment VI did not differ significantly from the other samples. The 122742 powders in experiments V and VI were somewhat packed together, similar to AN 4079 above. Other diopsides run in these 1332°C experiments, as well as some test diopsides at 13OO”C,did partially fuse and consequently were not analyzed. Forsterite The highest nobie gas soiubility overall was observed in forsterite FO 12243 (Fig. lc). This sample shows a clear trend of increasing solubiIity and was the most reproducible
C. L. Broadhurst et al.
712
la.
anorthite
Ne
I-3
A7
X.2
AI
Kr
Xe
gases
gases
lo33
1
le. forsterite
Id.
1000
Ne
Kr
AI
Xe
I
I
I
AI
Kr
Xe
Kr
Xe
gases
gases
Ne
Kr
AZ
gases
spine1
Xe
Ne
AI
gases
FIG. I. (a) Noble gas solubility in anorthites. A slight peak is present at Ar in An 4079 VI. Data for AN 4079 III indicates that the sample was probably contaminated by melt. Symbols: WU 6 17 I, - 0 --; WU 6 I7 II, - a --; 40791II,--~--;4079VI,--~--;HO228I,--O--;11-77791,--A--.(b)NoblegassolubiIityindiopsides. The Ar solubility has been interpolated (see text) for DI 122742 VI: the measured value is denoted by *. Symbols: 122742 V,--a--; 122742VI,--0--;R18684I1,-~---;R18684III,-+---;M2506 122742 IV, --•--; IV, - - 0 - -; 122742 VI*, 0. (c) Noble gas solubility in forsterite. A slight peak at Ar is present in the data from V. Symbols: 12243 II, - 0 --; 12243 III, -H -; 12243 IV, -Cl --; 12243 V, - fB -, (d) Noble gas solubility in spinels. The Ar value for SP 1l-4877 VI has been interpolated as described previously. Symbols: 1l-4877 III, - q -; I l-4877 IV, - 0 --; 1i-4877 VI, - m -; B12099 IV, - - 0 - -; 1l-4877 VI*, n . (e) Nobfe gas solubihty in DIseries silicate melts. Argon solubilities for DJM 2DI( ii) have been interpolated. Symbols: 2 DI(i) IV, - - n - -; 2 IX(i) MT, - - 0 - -; 2 Di(ii) V, - 0 -; 2 Dijiif VI, - 0 --; 2DI V*, 0: 2DI VI*. 4. (f) Noble gas solubility in FO- and AN-series meits. Argon values for DJM F05B have been interpolated. Symbols: FO 5B V, - - A - -: FO 53 VI,-A---;7ANMT,-m--;FOV*,&FOVI*,A.
Solubility of noble TABLE 2. MEASURED
NOBLE GAS ISOTOPIC IN ccSTP/gm
gases in minerals
CONCENTRATIONS
TABLE 3. Noble eas absoIute solubilities
in ccSTP/em-atm noble gas syntheses. Experiments conducted at atm&pheric 93% Ar. 1% Kr, and 1% Xe by weight. Ar values which described in text are denoted by &. -
NGX-I AN HO228
0.0956
800 15&l total
2.8(-6) 2.2(-7) 3.0(d)
9.61-8) 96(-S) 9.R.5)
6.5(-10) 1.1(-6) 1.1(-Q
5.3(-IO) 9.q.7) 9.q.7)
AN II-7779
O.l(rx
Boo 1550 total
3.w.7) 1.7.(-7) 3.8(-7)
8.3(-9) l.7(-5) I .U(-5)
3.q.IO) 1.7(-7) 1.7(-7)
3.3(-IO) 1.3(-7) 1.3(-7)
ANWU617
0.1234
SKI 15% total
7.x.9) 6.q.91 ,.4(.8)
8.11-8) 3.4(6) 3.5(-6)
1.6x-10) 3.81-8) 3.8(-S)
1xIO) 2.9(-8) 2.9~.8)
DJM 3AN*
0x690
800 1550 total
1.q-61 l.q.5) 1.1(-S)
7.9(-7) 9.1(-S) 9.2(-5)
5.2(-9) 6.3(-7) 6.3(-7)
3 x-91 2.7(-7) 2.8(-7)
NCX-II AN WU 617
0.0926
ISSD
6.x-8)
7.1(d)
5.S(-8)
1.5(-X)
DI RlX6&1
0.0773
800 1550 lOtal
1.q.8) 1.61-8)
2.8(6) 6.3(-7) 3.4(x$
6.7(-9) ,.4(-E) 2.1(-S)
5.6(-10) 2.0(-E) 2.0(-R)
800 1550 total
28~.8) 6.5(-6) 6.5(-6)
4.8(-7) ,.X4) I 5w
4.9(-9) I .7(-Q I .7(-6)
3.2(-9) I.,(-6) l.l(-6)
FO I2243
0.0792
DJM 3AN’
0.0478
15M
1.3(-S)
l.l(“t)
6.2(-7)
2.0(-7)
DJM IDI’
0.0674
8W 15% tolal
3.2(.6) 2.q-6) 5.2(-a
8.21.7) 8.3(x$ 9.1(d)
5.5(-9) 4.4(-S) 4.91-Q
4.7(-9) 2,3(-S) 2.8~.81
x 10-5 for the mixed pressure in 5% Ne, are slightly high as
experiment
Ne
Ar
Kr
XC
I II Ill VI 1 I
0.016 0.073 0.80 0.096 3.5 0.43
0.39 0.79 0.79 b.6* 10.7 2.0
1.4 2.1 0.80 10.9 41.3 b.3
3.9 1.9 0.82 17.9 I19 17.2
d&p&G R18684 R18684 122742
II III N N
0.018 0.14 0.079 0.80
0.38 g
0.80 1.3 0.9 1 5.4
2.1 4.5 6.8 7.1
122142 M2506
z N
0.41 2.1 0.20
7.0’ 2.6’ 0.41
6.3 15.4 0.74
!% 7.1
forsterite 12243 12243 12243 12243
II III IV V
7.6 6.5 6.2 7.8
16.7 19.0 19.0 30.7*
65.2 32.6
143 35.4
VI N
0.66 2.0 0.56 052 b.0
SZ3IIpk Experiments NGX-I and II mn at 1300@Cand @nospheric pressme using a minti noble gas atmosphere of 5% Ne. 93% AI. I W KI, 1I Xe by weight. Gas concentrstions detcmkxi by peak height comparison with error k 25% for Ar and +_15%for the olhw gases.
713
and melts
onorthite WU617 WIJ617 4079 4079 HO 228 11-7779
32.6
32.9
58.7
96.4
2.0 3.7 5.b*
3.0 4.8 2.4
1.5
4.6 3.9 3.6 1.3
1.0 2.8 1.5 13.0* 11.8* 4.-l* 16.7’ 1.2
1.9 2.4 0.97 2.6 2.6 1.3 3.3 0.61
3.6 2.1 0.33 2.4 2.4 1.1 3.0 0.33
sainel
ii-4877 Il.4877 1l-4877 812099 melts 1DI ZDI(i) 2DI(i) 2DI(ii) ZDl(ii) FOSB FOSB 7AN
II zr ; z tvfr
5.9 4.5 3.9 5.6 2.8 7.7 3.4
i1
Experiments NGX III and IV run at IxKlOC and atmospheric pressure using a mixed noble gas atmosphere. Gas concenmaions dewmined by isotope dilution: emx’s are listed. NGX-III AN 4079
0.0542
1550
6.9t2.1(-8)
7.1+1.1(-6)
2.1?0.02(-8)
&liO.M((-9)
DI RI8684
0.0552
1550
1.2f0.2(-7)
1.3?00.1(-5)
3.4*0.03(-s)
3.4M.03(-8)
0.0124
1550
5.660.3(-6)
1.7M.O4(-4)
8.6+~0.09(-7)
2.7iO.O2(-7)
0.0556
1550
S.?M.3(-7)
1.&0.1(-S)
7.wl.O7(-8)
3.5iO.O3(-8)
NGX-IV DI RI8684
0.0567
1550
6.w2.0(-8)
6.4fl.l(-6)
2.4a.o2(-83
5.1M.05(-8)
DI M2506
0.0515
155O
1.7to2-7)
3.7+1.2(-6)
1.9+0.02(-8)
5.5+0.05(-83
Dl 122742
0.0459
1550
6.9kO.4(-7)
Z.lM.l(-5)
1.4~0.01(-7)
5.2+0.05(-X)
FO 12243
0.0113
1550
5.4to.3(-6)
1.7*.04(-4)
B.&O.OY(-7)
2.5~0.02(-7)
SP II-4877
0.0157
1550
1.7zkO.O8(-7)
3.3~o.08(-5)
1.2k0.01(-7)
2.9+0.02(.X)
SPBlZO99
0.0642
1550
4.4M.24(-7)
l.OaOS(-5)
3.~.04(.8)
9.9M.O8(-9)
DIM 2DI(i)
0.0585
1550
5.OkO.2 (-6)
2.5?0.14(-5)
6.5m.063(-8)
1.6Fo.O1(-8)
FO 12243 SP
i 1.4877
one in the study. The solubilit~es in the 38 pm material in experiment V did not differ significantly from the other samples. Spinet
Expaimenrs NGX-MT, NGX-V, and NGX-VI run at 13324: BLalmospheric pressure using a mixed noble gas almosphere Gas concentrationa derermmed by isotope dilution. Ar values in V and VI arc slkhdy high due ID experimenml diflicuhies as described in text. NCX-MT D&3 7AN
0.0134
1550
2.9+0.1(-6)
1.1*0.2(-s)
I .6+0.02(-R)
?.SiO.4(-9)
DIM 2DI(I)
0.0330
I550
3.9iO.2(-6)
1.3FO.OR(-5)
2.6+0.03(-8)
2.5iO.O5(-9)
NGX-V DI 122742
0.0498
1550
3.6+0.38(-73
2.3&0.16(-S)
4.010.052(-7)
9.3+0.53(-7)
FO 12243
0.0126
ISM
6.7ti.33(-6)
2.8+0.008(-l)
lSkO.O19(-6)
5.8?0.1 If-71
DJM ZDI(d)
O.WrlR
,550
3.4ti.l6(-6)
l.ZM.023(-4)
6.8iO.063~.8)
2.010.014(-X)
DJM FOSB
0.0687
1550
2.4M.I I(-6)
4.210.067(-5)
3.4~.031(-8)
R.O+O.O65(-9)
NtiX-“I AN4079
0.0851
I550
8.2Kl.43(-7)
5.9k0.14(-5)
2.8*0.038(-7)
1.3+0.024(-7)
Dl 122742
0.0575
15%
l.gtO.o88(-6)
6.3+0.17(-S)
1.6aOl7(-7)
7.4kO.O73(-8)
SP I14877
0.0453
1550
4.&0.44(-7)
5.lxO.19(-5)
6.0+0.057(-S)
2.7+0.019(-8)
DIM ZDI(ii)
0.0538
1550
4&O.24<-6)
1.010.017(-2)
6.9+0.0&?+8)
1.&0.013(-8)
DIM FO5B
0.0594
1550
6&0.35(‘5)
1.5*.028(-4)
8.7+O.086(-8)
2.3kO.017(-8)
*melts conramcd crystals, so results are not discussed m text
Spine1 samples (Fig. Id) show a weaker increase in solubility with atomic number than do the silicate minerals. Two analyses show a decrease in the Xe/Kr ratio. Spine1 is more refractory than the other minerals investigated, and we are concerned that it may not have fused completely during gas extraction. The maximum temperature achieved by the RF induction heating was - 1600°C. However, the degassed remains of many previous samples are retained in the crucible, and this dross acts as a flux for refractory materials. Incomplete fusion was not suspected, for example, in the forsterite analyses, but it is a possible explanation for the decrease in the spine1 Xe/Kr ratio since sample degassing becomes diffusion-limited. The overall weaker increase in the solubility trend is also consistent with incomplete fusion. Since many other samples were analyzed both before and after these two spinels, it was not possible to verify this explanation. Very conservatively, we might consider that the gas concentrations measured in the forsterite and spine1 samples are lower limits. This would make the results yet more unusual as we could presumably retrieve even more gas if we melted the samples completely. An anomalous peak is present at Ar in some of the data from experiments NGX-V and VI, which is probably due to a memory effect in the gas extraction manifold during the final set of sample analyses. Further details are given in Appendix 1. Solubility trends in several of the figures have been
C.
714
L. Broadhurst et al.
roughly corrected by allowing the plotting program to calculate an equation for the line and estimate a value. The uncorrected values are denoted by *, as noted in the captions.
taining bubbles were analyzed. We also observed that the samples that were quenched in air and then in water did not contain higher gas concentrations.
Silicate Melts
Solubility Patterns of Melts + Crystals
Figure 1e shows the results for DI series melts, and Fig. 1f shows the 7AN and FOSB series melts. The melts show a clear trend of decreasing solubility with increasing noble gas atomic number. Solubiliti~ are also surprisingly low, often on the order of or lower than solubilities in the minerals. Figure 2 summarizes data from this study, with the range from other authors dashed in. The solubilities in our synthetic melts tend to be lower than those in the natural basal& especially for Ne and Ar. Our results overlap the low end of the natural basalt range and are most similar to molten enstatite from KIRSTEN ( 1968 ) . It is unlikely that our samples suffered gas loss upon or after quenching. Our experimental and analytical techniques were comparable to the other authors ( KIRSTEN, 1968; JAMBON et al., 1986; HXYAGONand OZIMA, 1986; Lux, 1987), all of whom had adapted procedures adequate for retention of He. Equilibmtion of Ne in the melts takes a few hours at minimum, so large amounts of gas are not expected to be lost during our quenching at 600-l 100°C/min. JAMBONet al. ( 1986) investigated noble gas diffusive loss from their glass samples. A one-day heating at 2OO’C resulted in no gas loss. Ten-day heatings at 600°C resulted in total He loss and significant Ne loss, which could be correlated with grain size. No measurable loss of Ar, Kr, or Xe was observed. It is possible, however, that in our study, as well as others, gas was lost directly after removal from the furnace. This gas would be lost not by difftsion but by the formation of bubbles on the melt surface. In general, our glasses were bubble-free and did not have a foamy surface. Specifically, no chips con-
Figure Al in Appendix 1 shows the noble gas solubilities measured in some of the glasses that grew crystals during the course of an experiment or upon quenching. Sample DM842 contained large euhedral olivine crystals that clearly grew suspended in the melt during the experiment. The DJM 3AN melts contained numerous small feathery crystals that may have grown during the experiment, upon quenching, or at both times. DJM ID1 and 2DI each grew only a small proportion of crystals, which sank to the bottom of the crucible, leaving a clear glass cylinder above. The solubility pattern is nearly flat, with a slight V-shape. The sides of the V have shallower slopes than the patterns for either minerals or melts alone. This shape is readily interpreted as a superposition of the solubility trends for the mineral and melt fractions. The exception is sample 2DI (i) , which shows a strong peak at Ar. This sample was run in experiment IV but analyzed along with the final set of V and VI runs. The anomalous pattern seen here was instrumental in determining that an Ar analytical problem occurred in that final set. Some of these melts also have higher overall noble gas solubilities than the melts that did not grow crystals. This may be attributable to gas inclusions in the minerals or trapping of bubbles during quenching. Optical examination of the glasses did not reveal any obvious bubbles; however, in the case of all but DM-842, the numerous small crystals made the images very confusing. It is possible that inclusions could have been overlooked or were too small to resolve; however, there could not have been too many because our gas mix was 93% Ar. Large or numerous bubbles of gas would lead to a huge gas release upon analysis with a composition very close to that of the experimental atmosphere. Assuming that inclusions are not important, then these experimental mistakes provide additional evidence that the solubility of noble gases can often be higher in minerals than in melts. It is to be noted that the noble gas concentrations in the melts are enriched, rather than diluted, by the addition of crystals. These crystal + melt patterns also led us to suspect that AN 4079 III had been contaminated with some glass, and thus follow through with TEM examination. While admittedly anomalous, AN 4079 III shows a strong increase in the Ne abundance and a strong decrease in the Kr abundance as compared to all the other anorthite samples. DISCUSSION
gases FIG.2. Overlapping heavy lines summarize noble gas solubilities in synthetic iron-free melts from this study: (B2) DJM 2DI( i) IV at 1300°C. (B3) DJM 2DI( ii) VI, (B4) DJM F05B V, and (BS)DJM
7AN MT at 1332°C. Range of melt solubility data from HIYAGON and OZIMA(1986), JAMBON et al. (1986), and Lux (1987) is dashed in; (KI ) refers to molten enstatite at 1500°C from KNSTEN( 1986). Symbols: Kl, x; B2, - l --; B3, -Cl--; B4, - 0 -: BS,- A -.
Noble Gas Solubility Trend in Minerals Noble gases show a clear trend of increasing solubility with increasing noble gas atomic number for all the minerals in this study. Some periodic properties of the noble gases are also correlated with atomic number. Of these properties, pola~~bility is most likely to be of importance since noble gases will not be ionized nor cont~bute significantly to co-
Solubility of noble gases in minerals and melts valent bonding in geochemical environments. They do not have permanent dipole moments; therefore, potential solubihty mechanisms must involve induced dipole interactions. which are governed by an atom’s electronic polarizability. Two of these interactions are the weak inverse r”-dependent-dipole/ induced-dipole and induced-dipole/induceddipole interactions. These types of interactions are used to model the formation of noble gas solids and the physical adsorption of noble gases on surfaces. Adsorption of noble gases is reviewed in detail by OZIMA and PODOSEKf 1983) and is further addressed in YANG et al. (1982), YANG and ANDERS (1982a,b), BERNATOWICZand PODOSEK(1986), and BROADHURSTet al. ( 1990). It is known to be a lowtemperature phenomenon: A typical lifetime for physisorption at room temperature is about lo*.* sec. Quantitative physi~~tion of Ne and Ar requires 77% Xe and Kr 192*K. Prior to analysis, we placed the samples in a high-vacuum manifold and heated them at 100°C for -48 h. This is a long-accepted mass spectrometric practice for removal of the bulk of surface adsorbates and/or atmospheric contamination, both in our laboratory and in numerous other noble gas research groups (AMARI and OZIMA, 1988; MATSUDA and NAGAO, 1989; WIELER et al., 1989; SCHELHAASet al., 1990; SWINDLEet al., 1990). Reactive gases are sometimes observed to form chemical bonds with surfaces. Enthalpies of adsorption for such interactions are on the order of -50 to -200 k&/mole, whereas enthalpies for physiso~tion are on the order of > - f 0 kcal/ mole. This energy difference is one of the criteria for hypothesizing a surface chemical interaction. Enthalpies of adsorption for Kr and Xe on a number of silicate substrates are compiled in OZIMA and PODOSEK( 1983), and no value more negative than -6.7 kcal/mole is reported. WACKER ( 1989) investigated the sorption of Ne, Ar, Kr? and Xe on carbon black, acridine carbon, and diamond. These surfaces are considered to be much more active than silicate or oxide minerals; however, enthalpies of adsorption still ranged from -2 to -8 kcal/mol, fully consistent with physisorption. As reported in a series of three papers, YANG et al. ( 1982) and YANG and ANDERS ( 1982a.b) synthesized magnetite, chromite, sulfides, and spine1 in a mixed noble gas atmosphere at temperatures from 167-447”C. One important result of the study was that samples record the noble gas signature of the highest pressure reservoir that they encounter. This, in effect, means that experiments conducted at noble gas partial pressures less than atmospheric yield results that are indistinguishable from air contamination. When partial pressures are higher, distinction between components can be made. Thus, the authors divided gases into groups based on release temperature: ls4OO”C, physisorbed; 18OO”C, chemisorbed; and >8OO”C, trapped. (“Trapped” gases are contained within the mineral lattices; however, this term is used rather that “dissolved” because there was no attempt to attain an equilibrium solubility). It is hypothesized that during mineral synthesis, trapped gases are first surface adsorbed and then become “fixed” in the crystal by some unspecified process, probably involving a combination of recrystallization, chemical reaction, diffusion, and migration to traps. This “fixing” turns ph~isor~d gases into more strongly bound components. However, the majority of gases released
715
from the samples upon analysis are still classified as adsorbed, not trapped. In their “trapped” nobie gas com~nent, YANG et al. ( 1982) and YANG and ANDERS ( 1982a,b) found the same fractionation pattern that we retrieved from our mineral samples. Despite the differences that minerals were grown during the experiments and that temperatures were much lower, this study shares another conclusion with ours: Noble gases introduced into minerals can be persistently retained. The two-step heating procedure used for some of our noble gas analyses shows that the gas atoms are held in very retentive sites. With the exception of Ne, very little gas is released at the 800°C step (see Table 2), and the samples must be fused to extract the majority of the gas incorporated in the minerals at 1300°C or 1332°C. This observation argues strongly against simple physiso~tion and against chemiso~tion as defined by YANG et al. ( 1982). The higher proportions of Ne lost at 800°C were anticipated, and we attribute this to faster diffusion as well as a less energetic interaction with the mineral lattice. Preferential loss of Ne has been observed in shock-implanted noble gases in basalt ( WEINS, 1988; WEINS and PEPIN, 1988), as well as in glasses ( JAMBONet al., 1986; MATSUDAet al., 1989 ). If our results are due to adsorption and/or atmospheric contamination, then they are an unprecedented example of such and would lead to the conclusion that surface-bound gases are so persistent that they cannot be removed from any arbitrary mineral sample. We might also consider the distinction between “adsorption” stemming from many days of exposure to noble gases at 1300°C and “true solution” to be a moot point. Although it is clear that we cannot fully determine where the noble gases are sited in the minerals, we hypothesize that they are situated in lattice vacancy defects. The supposing evidence, more fully detailed in BROADHURST( 1989) and BROADHURSTet al. ( 1990), is as follows:
1) The variable solubilities of noble gases in different natural samples of the same mineral require a solubiiity mechanism that can vary significantly among samples with almost identical major element compositions and, in some cases, geologic occurrence. 2) The two-step heating procedure for some of the noble gas analyses shows that the gas atoms are held in retentive sites. 3) Interstitial sites cannot be ruled out, but it is difficult to explain the solubility variation, as interstitial sites should not vary greatly among nearly identical mineral samples. Purely interstitial siting for noble gases would predict a mineral solubility trend opposite to that observed (i.e., similar to the melt trend). We might also expect to observe much higher ~lubilities in the tectosilicate anorthite, which is far less dense than the three other minerals investigated. 4) Transmission electron microscope examination of some of the samples did not reveal any anomalous microstructures or dislocation densities, or any evidence of healed mi~rofractures. Material was examined both before and after an experiment.
C. L. Broadhurst et al.
716
To first order, the interaction of a lattice-vacancy defect and a noble gas atom may be described by the monopole/ induced-dipole interaction: E = -q?aJ2r4, where q, is the charge on the site, LYZ the noble gas polarizability, and r the distance between the monopole and the gas atom. This interaction is much more energetic than the inverse r6 interactions and is stable at high temperature. The energy .E of a Xe atom 3 A from a +2 site is 32.7 kcallmole. The maximum temperature of stability for this interaction can be estimated by dividing E by the Bolztmann constant: E/k = 1.64 X lo4 “K. The actual interaction energy will be lower because the charge on the site will be only partial. In a crystal lattice, defects often occur in pairs. A vacancy would be accompanied perhaps by the nearby substitution of an M 3+ for an MZi cation, and the hole and the cation have a significant coulombic attraction for one another. More importantly, the lattice will adjust in the vicinity of defects. The polarization P of a crystal with N ions (i) in the formula unit can be written as:
where cut,*= Lliai, Er, is the localelectric field, and polarizability is scalar. Further, 01, = ffe + Lyd. The term (ye is the electronic polarizability, which describes the interaction of ion i with an oscillating electric field. The term ad is the displacement polarizability, which describes the induced-dipole moment due to the relative displacement of an ion from its normal lattice position. This displacement is in response to the static electric field generated by the charged vacancy. The term ad is important in the total polarization vah,te for a crystal, and details for silicates can be found in LASAGAand CYGAN ( 1982), and CYGAN and LASAGA(1986). Short-range repulsive energies may also reduce the interaction energy of the noble gas atom with the defect site. The short range energy arises from interactions between electrons in orbitals of nearest neighbor cations and anions in a crystal lattice ( BURNHAM,1985 ). For nearest neighbors, this energy is repulsive. Analogously, as the noble gas electron cloud becomes polarized, one must consider both the attractive and repulsive forces on the positive and negative portions of the dipole. These types of calculations are complex and beyond the present scope; however, the magnitude of this energy is relatively small. It is possible that noble gas atoms can interact with the electric fields produced by lattice defects; thus, the greater polarizabilities of the heavier noble gases would account for their enhanced solubility. Neon is the least polarizable, hence the least soluble; also, a greater fraction is lost at the lowtemperature heating step. Even if noble gases ultimately reside in interstitial sites or planar defects after quenching, vacancy/ induced-dipole interactions nevertheless may be important for the solution and diihtsion of noble gases during our experiments.
Solubility Trend in Melts As noted earlier, the solubilities of noble gases in synthetic melts reported here and by KIRSTEN ( 1968) are lower, in general, than those reported for natural basalts by JAMBON et al. ( 1986) and LIJX ( 1987 ). The greatest differences occur for Ne and Ar. There is relatively little difference in any of the Kr and Xe values plotted in Fig. 2. The lower overall ~lubilities in synthetic melts relative to natural melts can be att~buted to the greater degree of polymerization and Fe3+ concentration and lesser MgO and CaO concentrations in the natural melts ( LUX, 1987; WHITE et al., 1989; BROADHURSTet al., 1990). BROADHURSTet al. ( 1990) concluded that Ar solubilities in melts of greatly differing compositions were better correlated with melt molar volume than melt density. The new melt data reported here also support this conclusion. Figure 3a plots JSr solubility against melt densities calculated at appropriate experimental temperatures for a number of basalts; Fig. 3b shows Kr solubility vs. melt molar volume. Figure 3a shows two distinct groups, whereas 3b shows a positive correlation of solubility with molar volume for all of the data. The major differences between the two groups in Fig. 3a is the presence of ferric iron in melts 6 through 10. Among the natural melts, our data and the data of KIRSTEN ( 1968 ) are more similar to those of HIYAGON and OZIMA ( 1986). Although these latter authors did not give melt analyses, the starting material consisted of basalt (67 wt%) and harzburgite (33 wt%) powders, and experiments were done at oxygen fugacities of lO-9 to lo-” bars, which is below the fayalite/ quartz/magnetite buffer at 1300°C. These conditions produced a MgO-FeO-rich oiivine basalt. The pattern of decreasing dispersion in solubilities with increasing noble gas atomic number is a feature of all the melts in Fig 2 but is most pronounce for our data. FoILowing UHLIG ( 1937) and BLANDERet al. ( 1958), Lux ( 1987) found that the noble gas solubihties in a series of natural melts could be described by the equation Ki = a exp( -br’), where a and b are constants and r is the gas kinetic radius. The constant a was the same for all liquids studied; however, b increased moderately [ 0.869- 1.14 (kO.048 )] from leucite basanite to ugandite. It was also concluded by LUX ( 1987 ), on the basis of relative solubility (i.e., Ki - Kj/Kj for two melts, i and j) that the larger a noble gas atom, the more its solubility will be athected by the properties of the liquid solvent. While this latter conclusion is intuitively satisfying, it is not supported by our data (Table 4). The major technical difference between our study and the others referenced herein is the composition of the atmosphere. We did not use He, nor did we gas-mix with major atmospheric components (i.e., CO*, CO, 02, and Nz). The other authors used gas mixes containing 10-1000 times less Ne than ours. KIRSTEN( 1968) is an exception: His experimental runs were in pure He, Ne, and Ar at pressures up to 0.55 bars, or a mixture containing 0.18 bars of each. KIRSTEN (1968), JAMBONet al. ( 1986), and Lux ( 1987) found that plots of solubihty vs. noble gas partial pressure were linear, with slopes of the lines decreasing with increasing atomic number. in particular, the slopes for He and Ne are much steeper than the plots for the other gases: Small variations in
717
Solubility of noble gases in minerals and melts
yl s
x
60
3b
3a
1
80
9
.9
.
-
E 2 i
8
n8 6
g VI z
40
-
.
5 .
20
1
.
.
6
.4
.
.3
*I n 2 0
I 2.5
I
I
2.6
2.7
density
2.8
I8
20
V(m)
gm/cm*3
26
24
22
28
cmA3/mole
FIG. 3. Krypton solubility plotted against (a) melt density and (b) molar volume. Molar volume shows a positive correlation for all the data, while density does not. Densities calculated from LANGEand CARMICHAEL( 1987) at 1300°C except where noted. Key to figure: (*1 ) molten enstatite at 1500°C from KIRSTEN( 1968) (solubility is K&en’s extrapolation from light noble gas data); (2) DJM 7AN: (3) average DJM 2DI( i); (4) average DJM F05B: (5) average DJM ZDI(ii); (6) tholeiite at 1400°C from JAMBONet al. ( 1986) (note that composition is given as Fe0 = Fe total); (7) ugandite: (8) alkali-olivine basalt; (9) tholeiite; ( 10) leucite-basanite from LUX ( 1987).
the partial pressure will translate to large variations in the solubility. There is no reason to suspect that our experiments exceeded Henrian pressure regimes. Our results indicate, more likely, that there are yet subtleties to the relationship between noble gas solubility and melt structure. Partition Coefficients
Partition coefficients (Table 5) were calculated from the total concentration data for minerals and melts run in the TABLE 4. Comparison of absolute (AKi_j) vs. relative (AKi:/Kl) solubilities. Example melts are as follows: Leucite basantte and ugattdite brat /cet the range of solubilities in basahs from Lux (1987); tholeiite is intermediate. Hlb is the upper limit for olivine basalt from Hiyagon and Ozima (1986). DJM 7AN and DJM 2DI(i) IV are the respective lower limit and representative value from this study.
i&u-bat.:
I =Ieu-has:
i =Wte
gas
AK
AKIK
AK
AK/K
Ne
20.5
0.49
36.1
0.85
At
1.1
0.63
9.4
0.77
Kt
6.3
0.68
6.9
0.74
Xc
2.5
0.7 1
1.4
0.40
i=leu_bas:i=7AN
t = Hlb:
i =lAN
gas
AK
AK/K
AK
Ne
38.6
0.91
10.8
AK/K 0.76
Ar
11
0.90
2.7
0.69
Kr
8.69
0.93
0.26
0.29
Xc
3.17
0.90
0.83
0.7 I
same experiment. These patterns (Fig. 4) show a clear trend of increasing compatibility with increasing noble gas atomic number-a trend also seen by HIYAGONand OZIMA( 1986 ). However, there is a large variation in the absolute mineral/ melt partition coefficients for a given phase pair. Values are frequently greater than unity, which is a direct result of the low melt solubilities and the surprisingly high mineral solubilities. It seems certain that the partition coefficients for the noble gases cannot be described with the level of specificity considered standard in ionic trace and minor element partitioning studies. This is not the first time that noble gas partition coefficients greater than unity have been reported. BATIZAet al. ( 1979) reported noble gas data for primitive and evolved rocks thought to originate from the same magma. Assuming a closed system, they calculated the amount of crystallization involved in forming the different rock types and, thus, the noble gas partition coefficients. Values were near 10-l for Ne, but greater than unity for the heavier gases. The authors were forced to conclude that their system was not closed. Many of the runs in HIYAGONand OZIMA ( 1986) yielded partition coefficients greater than unity, although the pattitioning trends were the usual (increasing with increasing atomic number). These data were discarded by the authors
TABLE
gas
AK
AK/K
AK
AK/K
Ne
6.8
0.32
15.1
0.7 1
At
0.77
0.17
I.7
0.37
Kr
2.1
0.71
0.6
0.33
Xe
-0.16
-0.16
-1.1
-1.1
exper.
5. Noble gas solid/liquid partition coefficients. Values calculated and melt total gas concenwtions within a given expaiment. mineral
melt Ne
DSil Ar
from mineral
KI
Xe
1.3 0.040
0 38 0.31 2.3 2.0 0.63
3.2 3.3 3.3 1.8 0.61
0. I05 2.7
0.19 67
5.9 45
zl
0.37 0.073
0.63 0.33
2 4 0.73
4.0 I 2
N ;
DI RI8684 DI M2.506 122742
ZDI(i) ”
0.013 0.14 0.033
0.26 0.84 0.15
N N
SP 1 l-4877 SPB12099
” ”
0.034 0.088
V V
DI 122742 FO 12243
2DI(ii) FO 5B
VI VI
DI 122742 SP 1 l-4677
ZDI(ii) FO 5B
C. L. Broadhurst et al.
718
Ar
Kr
Xe
Ne
Kr
AT
gases
Xe
gases
FIG. 4. Noble gas partition coefficient trends for (a) experiment IV and (b) experiments V and VI. All partition coefficients calculated from melt DJM 2DI(i) in (a). In (bl. values for DI 122742 calculated from melt DJM 2DItiil: those for FO 12243 and SP I l-4877 from CjM Fd5B. ?hk patterns generally reflect the total gas concentration data: Partition coefficients increase with increasing noble gas atomic number and are greater than unity for at least Xe. The trends for the VI spinels show a downturn in the Kr/Xe ratio, reflecting the same trend in the total gas concentrations. SP B I2099 in IV is the only sample that yields a set of partition coefficients that are all less than unity. Symbofs: (a) DIR 18684 IV, - q --; DI M2506 IV, - q -: DI 122742 IV, - n --; SP 114877 IV, - - 0 - -; SP BI2099 IV, --~---.(b)DII22742V,-~---;D1122742VI.-~---;FO12243V,--~--;SPi14877VI,--‘~--.
on the grounds that these samples contained gas bubbles or inclusions. Inclusions were indeed observed microscopically; however, the trapped gas component (G) defined from the isotopic mixing discussion had an unusual property: “The composition of component G appears to be fractionated from the original test gas favoring the heavier noble gases: Xe/Ar, Kr/Ar ratios are higher and He/Ar, Ne/Ar ratios are lower than those for the original gas. . . . Possibly, lighter noble gases have diffused out of the bubbles and dissolved in the melt more quickly than heavier noble gases have done” (p. 2053). While we by no means wish to criticize the above interpretations, we find it interesting that our results, as well as those of BATIZA et al. (1979) and HIYAGON and OZIMA ( 1986), can be explained by noble gases acting as relatively compatible, fractionated species in mineral/melt systems. In particular, we note that our data set for minerals + melts shows the same type of behavior-elevated concentrations of noble gases with respect to melts alone and a component fractionated in favor of the heavy noble gases. However, we have no evidence for inclusion control of these phenomena, thus no basis to conclude that the composition of gas bubbles is fractionated. The hypotheses that ( 1) noble gas solubility in minerals is controlled by defect siting, and (2) noble gas solubility in melts is directly proportional to molar volume and polymerization leads to the conclusion that noble gases will become even more compatible with increasing pressure, Highpressure data from HIYAGON and OZIMA ( 1986) do tend to show partition coefficients similar to or greater than 1 bar results; however, these experiments are compromised by the use of 10 or 15 wt% lithium borate in the starting material composition. The partial molar volume of BZ03 is highly
variable and must be calculated for any specific melt composition. However, in silicate melts of the basalt-hanburgite type, boron is likely to be coordinated as B03; hence, the partial molar volume will be somewhere in the range 37-47 cm3/mole (EITEL, 1965; PAUL, 1990). This means that the melt molar volume will be significantly greater than a typical mafic basalt. More studies are required before the effects of pressure on noble gas partitioning can be understood. Average minerals from the mantle are likely to contain many more defects than the minerals from this study. Minerals in the mantle exist at high tem~rat~es and are expected to have higher concentrations of trace elements than the relatively pure natural minerals that we used. Since we have shown ( BROADHURSTet al., 1990) that defects probably can account quantitatively for the one-bar gas concentrations in our minerals, it follows that the much smaller noble gas concentrations in the mantle could be reasonably accommodated in defects. IMPLICATIONS FOR THE TERRESTRIAL PLANET ATMOSPHERES
Contrary to the usual a~umptions, the absolute values of our partition coefficients indicate that magmatic degassing cannot be an efficient process. This is especially true for forsterite sample FO 12243, and for the gases Kr and Xe. Furthermore, regardless of the absolute values, these partitioning trends cannot produce the observed noble gas abundance patterns for the terrestrial planet atmospheres. An example calculation for a C3V chondritic source is shown in Fig. 5. At low degrees of partial melting, the liquid becomes strongly enriched in Ne only, contrary to the trends for the planets. At higher degrees of partial melting, the liquid trend would
Solubility of noble gases in minerals and melts
Ne
ICI
Ar
XC
gases FIG, 5. Equilibrium hatch partial melting of a C3V chondrite source and POLLACK,1983). Noble gas concentrations in the
(DONAHUE
liquid shown at several fractions (F) of partial melting. Partition coefficientsare from sampie SP B12099 in Table 5. We assume that the partition coefficient set for a single mineral is sufficient to represent a variety of mantle com~sitions since noble gases are not sensitive to mineral structure or composition in the same manner that ionic species are. Symbols: C3V, - l -; F = 0.05, - * -; F = 0.10,
--[II--;F=O.lS,--A--;F=OSO,--O--.
follow the starting material composition almost exactly. The salient point is that the distinctive Kr/Xe fractionation in the planetary atmospheres is not reproduced. Krypton and Xe do not partition differently enough that magmatic fractionation can account for the “missing Xe.” Our data imply strongly that the terrestrial planet atmospheres did not accumulate from partial melting of a common chondritic source. As a corollary, if volcanic outsung was mostly responsible for the terrestrial ptanet atmospheres, then the primordial volatile abundances followed the pattern Venus > Earth > Mars. This abundance pattern cannot be produced from a common source with our partition coefficients because noble gases will remain at almost constant concentration levels in the melt throughout a reasonable range of degrees of partial melting (BROADHURST,1989). It must be noted that there is no a priori reason demanding that noble gases other than He act as incompatible elements in mantle melting processes. Among the primordial volatiles, only ‘He has been unambiguously determined to be fluxing from the mantle today. A large number of isotopic m~elling studies find that Earth’s upper mantle and parts of the lower mantle were nearly completely degassed prior to at latest 3.8 Ga (BERNATOW~CZand PODOSEK, 1978; HAMANO and OZIMA, 1978; ALLBGREet al., 1987; TURCO~TE and SCHUBERT, 1988; MARTY, 1989; OZIMA and IGARASHI, 1989; ZHANG and ZINDLER, 1989). These studies are based upon the evolution of the radiogenic isotopes in a magmatic degassing scenario. Most important to the arguments are observations of excess “‘Xe in MORB (STAUDACHERand ALLI?GRE,1982; ALL&RE et al 1983; 1987) and estimates of Earth’s K content. Each is also based on the assumption
719
that nobie gases are uniformly in~mpatible in mineral I melt systems. Despite individual approaches, these models concur that early catastrophic degassing was largely responsible for the Earth’s atmosphere, and only a moderate amount of degassing has occurred since. Our results also favor catastrophic degassing, but for somewhat different reasons. The effect of our partitioning behavior on the isotopic models would be a definite increase (i.e., moves closer to the present so degassing takes longer) in the predicted mean ages for outgassing. There is also potential for change in the relative rates of degassing of the different species; this would have to be tested explicitly for a given model. In addition, since noble gases will be fractionated by partial melting events, the use of mean degassing ages to estimate past rates of tectonic activity ( ALL~GRE et al., 1987) is not strictly correct. In a magmatic degassing scenario, our results would require large degrees of partial melting and vigorous convection in the mantle in order to degass at least the MORB source mantle. Such conditions were possibly present early in the planet’s history. We also note that these high fractions of melting might be inherently self-limiting as they are expected to produce MgO-rich komatiite type magmas. As discussed previously, such melts will have low noMe gas solubilities. AZBELand TOLSTIKHIN( 1990) have introduced fmctionation of noble gases into an isotopic model. They have explained the MORB noble gas fractionation trends relative to air by a superposition of initial mineral/melt partitioning followed by partial degassing of the melt. The theoretically derived partition coefficients of AZBEL and TOLSTIKHIN ( 1990) match the trends found in this study and by HIYAG~N and OZIMA ( 1986). A very important point made by AZBEL and TOLSTIKHIN( 1990) is that single-stage partial melting and degassing cannot yield both a high degree of upper mantle degassing and the observed MORB fractionation pattern; a multistage process is required. As has been pointed out to us, noble gases in MORB and OIB basalts are often concentrated in vesicles, micropores, ‘etc., and can be released quantitatively by simple crushing of the samples. These vesicles are to be expected in basalts from near-surface conditions. The equilibrium noble gas solubilities in this study and others are by definition liquid/ vapor partition coefficients which show that noble gases are highly incompatible elements in this system. The solubilities are insignificant in comparison with those of the reactive volatiles ( H20, Cq, H,S, CH4, etc.), and it seems clear that any fluid phase evolved from a melt is capable of stripping that melt of noble gases. In other words, if a melt can migrate to near the surface, it can outgass. In a multistage process, we are concerned less with the present-day release of volatiles at the MOR and hot spots than we are with the initial, largescale melting of whole or part of the mantle to form the hypothesized degassed mantle reservoir. To reiterate, we wish to call attention to the points that during partial melting ( 1) noble gases can be fractionated from one another, and (2) radiogenic noble gas daughters may or may not be significantly fractionated from their parent elements, depending on the individual situation. An additional concern is the growing body of evidence that the Earth’s interior is enriched in solar-type Ne. 2oNe/
720
C. L. Broadhurst et al.
22Ne enrichments have been reported in diamonds, MORB, and Hawaiian basalts (POREDA and DI BROZOLO, 1984; SARDA et al., 1988; MARTY, 1989; HONDAet al., 1991; OZIMA and ZASHU, 1991). Essentially, the mantle appears to have a Ne endmember component that is entirely solar. MARTY ( 1989) finds that the difference between air and mantle Ne is so great that it precludes a common origin for these reservoirs: Air Ne composition could then be explained by a mixture of planetary (59%) and solar (41%) components, which have remained distinct since accretion. OZIMA and ZASHU ( 1991) conclude from diamond samples that primordial Ne in the ancient mantle is of solar origin, while Ar, Kr, and Xe isotopic ratios are virtually identical with the current atmosphere. Interestingly, OZIMA ( 1989) and OZIMA and ZASHU( 199 1) find that noble gases are of trapped origin, including some of the radiogenic components. There is too much gas in the diamonds for a reasonable in situ production. If the present volcanic regime is not the key to the past, then one must seriously consider that accretionary catastrophic processes are largely responsible for the terrestrial planet atmospheres. Primordial volatile abundances among Venus, Earth, and Mars may or may not have been roughly equal. One can appeal to impact degassing to strip Mars of an early denser atmosphere (WALKER, 1986; AHRENSet al., 1989; MELOSHand VICKERY, 1989 ). Cometary impacts are equally appealing as a means of enhancing the Venusian atmosphere relative to Earth (GRINSPOONand LEWIS, 1988). If the impact flux during the latest stages of accretion were very high, then there could have been nearly continual atmospheric gain and loss, which slowed gradually as accretion slowed. With such an impact flux, effective removal of volatiles from the growing planets is necessary since only a small percentage of volatile-rich material is required to account for the present planetary atmospheres ( ANDERSand OWEN, 1977; DREIBUS and W,&NKE,1987). The final atmospheres could represent a mixture of many different types of impactors, or could reflect the influence of only a few large objects. ZAHNLE et al. ( 1990b) have considered the latter case to explain the Xe problem. They start with a population of porous planetesimals growing in solar nebula. At some point, a planetesimal grows large enough that lithostatic pressure squeezes the pores shut, thus trapping the Xe. These larger planetesimals are then accreted, in turn, to a growing terrestrial planet. Loss of gases during the accretion period may have been by hydrodynamic escape in addition to impact degassing. Hydrodynamic escape can result in isotopic fractionation (HUNTEN et al., 1987; SASAKI and NAKAZAWA, 1988; ZAHNLEet al. 1990a,b), thus it has the potential to reconcile some of the differences in primordial/radiogenic isotopic component ratios among the terrestrial planets and between the planets and meteorites. Escape of a solar composition atmosphere is one explanation for the observed difference between air and mantle Ne. Alternatively, the present atmosphere may also have formed from degassing of a lateaccreting veneer, while the interior is composed of solar composition material. MUSSELWHITEet al. ( 199 1) also provide experimental partitioning data that ultimately supports an accretionary steam atmosphere on Mars. The authors have addressed the
high ‘29Xe/‘32Xe ratio in the Martian atmosphere as compared to SNCs and Earth reservoirs. Partitioning experiments showed that I and Xe are not fractionated from each other during magmatic processes. Instead, it is proposed that both I and Xe were outgassed by 4.45 Ga, or before significant lz91 decay. The highly water-soluble I was retained in the oceans and hydrated crust, while Xe was removed by atmospheric erosion. ‘29llater decayed and was released to the now-present atmosphere. CONCLUSIONS The solubility trend of Ne, Ar, Kr, and Xe in anorthite, diopside, forsterite, and spine1 increases with increasing noble gas atomic number. Different samples of a particular mineral composition have characteristic solubilities, some of which are surprisingly high. Noble gases are likely to be sited in lattice-vacancy defects. In contrast, synthetic silicate melt solubilities decrease with increasing atomic number. Solubility values are on the order of or lower than those in minerals. Our data tend to overlap the low end of the range defined for natural basaltic melts. The difference between Ne and Xe solubility values in our results is not as extreme as that in the natural melts. Noble gas solubilities correlate well with molar volume but poorly with density. A trend of increasing partition coefficient values with increasing atomic number is seen for all mineral/melt pairs. This trend indicates that is not possible for the noble gas abundances in the atmospheres of the terrestrial planets to have been derived from partial melting of a common chondritic source. Furthermore, the partition coefficients determined in this study rule out significant fractionation of Kr relative to Xe in magmatic processes; thus, the “missing Xe” in the Earth’s atmosphere cannot be attributed to partial melting. We call attention to the caveats that during partial melting, ( 1) noble gases can be fractionated from one another, and (2) radiogenic noble gas daughters may or may not be significantly fractionated from their parent elements, depending on the individual situation. These noble gas partition coefficients support a number of other studies, which conclude that early, catastrophic degassing was largely responsible for Earth’s atmosphere, and that degassing is a multistage process. Catastrophic degassing would have to involve either large amounts of partial melting and vigorous mantle convection or accretionary processes, such as impact accumulation/erosion and hydrodynamic escape. Acknowledgments-This research was supported by NSF grant EAR 85-18740 to M. J. Drake and NASA Traineeship NGT-50051 to C. L. Broadhurst. We thank the various reviewers and F. Frey for constructive criticisms. CLB also acknowledges support from NAS/ NRC and NASA GSFC for production of the manuscript. Editorial handling: F. A. Frey
REFERENCES AHRENsT. J., O’KEEFEJ. D., and LANGEM. A. ( 1989) Formation of atmospheres during accretion of the terrestrial planets. In Origin and Evolution of Planetary and Satellite Atmospheres (ed. S. K.
ATREYAet al.) pp. 328-385. Univ. Arizona Press.
Solubility of noble gases in minerals and melts ALL~GREC. J., STAUDACHERT., SARDAP., and KURZ M. ( 1983) Constraints on evolution of Earth’s mantle from rare gas systematics. Nature 303, 762-766. ALLBGREC. J., STAUDACHERT., and SARDAP. ( 1987) Rare gas systematin: Formation of the atmosphere, evolution and structure of the Earth’s mantle. Earth Planet. Sci. Lett. 81, 127-150. AMARIS. and OZIMAM. ( 1988) Extra-terrestrial noble gases in deep sea sediments. Geochim. Cosmochim. Acta 52, 1087-1095. ANDERSE. and OWENT. ( 1977) Mars and Earth: Origin and abundance of volatiles. Science 198, 453-465. AZBEL1. Y. and TOLSTIKHINN. ( 1990) Geodynamics, magmatism, and degassing of the Earth. Geochim. Cosmochim. Acta 54, 139154. BATIZAR., BERNATOWICZ T. J., HOHENBERGC. M., and PODOSEK F. A. ( 1979) Relations of noble gas abundances to petrogenesis and magmatic evolution of some oceanic basalts and related differentiated volcanic rocks. Contrib. Mineral. Petrol. 69, 30 1-3 14. BERNATOWICZ T. J. and PODOSEKF. A. ( 1978) Nuclear components in the atmosphere. In Terrestrial Rare Gases (ed. E. C. ALEXANDER.JR., and M. OZIMA)pp. 99-135. Japan Scientific Society Press. BERNATOWICZ T. J. and PODOSEKF. A. (1986) Adsorption and isotopic fractionation of Xe. Geochim. Cosmochim. Ada SO, 15031507. BERNATOWICZ T. J., PODOSEKF. A., HONDA M., and KRAMER F. E. ( 1984) The atmospheric inventory of Xenon and noble gases in shales: The plastic bag experiment. J. Geophys. Res. 89,45974611. BERNATOWICZ T. J., KENNEDY B. M., and POD~SEKF. A. ( 1985) Xe in glacial ice and the atmospheric inventory of noble gases. Geochim. Cosmochim. Acta 49, 2561-2564. BLANDERM., GRIMES W. R., SMITH N. V., and WATSON G. M. ( 1958) Solubility of the noble gases in molten fluorides. II. In the LiF-NaF-KF eutectic mixture. J. Phys. Chem. 63, 1164-l 167. BROADHURST C. L. ( 1989) Solubility and Partitioning of Noble Gases in Anorthite, Diopside, Forsterite, Spine], and Synthetic Basaltic Melts: Implications for the Origin and Evolution of Terrestrial Planet Atmospheres. Ph.D. thesis, Univ. Arizona. BROADHURST C. L., DRAKEM. J., HAGEEB. E., and BERNATOWICZ T. J. ( 1990) Solubility and partitioning of Ar in anorthite, diopside, forsterite, spine], and synthetic basaltic liquids. Geochim. Cosmochim. Acta 54, 299-309. BURNHAMC. W. ( 1985) Mineral structure energetics and modeling using the ionic approach. In Microscopic to Macroscopic; Reviews in Mineralogy. pp. 347-388. Min. Sot. Amer. CARROLLM. R. and STOLPERE. M. ( 1991) Argon solubility and diffusion in silica glass: Implications for the solution behavior of molecular gases. Geochim. Cosmochim. Acta 55, 211-225. CYGAN R. T. and LASAGAA. C. ( 1986) Dielectric and polarization behavior of forsterite at elevated temperatures. Amer. Mineral. 71, 758-766. DREIBUS,G. and WANKE H. (1987) Volatiles on Earth and Mars: A comparison. Icarus 71, 224-240. DONAHUET. M. and POLLACKJ. B. (1983) Origin and evolution of the atmosphere of Venus. In Venus (ed. D. M. HUNTENet al.) pp. 1003-1036. Univ. Arizona Press. EITELW. ( 1965) Silicate Science. Academic Press. FOLANDK. A. ( 1974) A?‘diffusion in homogeneous orthoclase and an interpretation of Ar diffusion in K-feldspars. Geochim. Cosmochim. Acta 38, 151-166. GRINSFQOND. H. and LEWISJ. S. ( 1988) Cometary water on Venus: Implications of stochastic impacts. Icarus 74, 21-35. HAMANO Y. and OZIMA M. (1978) Earth-atmosuhere evolution model based on Ar isotopic data. in Terrestrial kare Gases (ed. E. C. ALEXANDER,JR ., and M. OZIMA) pp. 155- 17 1. Japan Scientific Society Press. HAYATSUA. and WABOSOC. E. ( I985 ) The solubility of rare gases in silicate melts and implications for K-Ar dating. Chem. Geol. 52, 97-102. HIYAGONH. and OZIMAM. ( 1986) Partition of noble gases between olivine and basalt melt. Geochim. Cosmochim. Acta SO, 20452057.
721
HONDAM., MCDOUGALLI., PATTERSOND. B., DOULGERISA., and CLAGUED. A. ( 1991) Possible solar noble gas component in Hawaiian basalts. Nature 349, 149- I5 1. HUNTEN D. M., PEPIN R. 0.. and WALKERJ. C. G. (1987) Mass fractionation in hydrodynamic escape. Icarus 69, 532-549. JAMBONA., WEBERH., and BRAUN0. ( 1986) Solubility of He, Ne, Ar, Kr, and Xe in a basalt melt in the range 1250-1600°C. Geochemical implications. Geochim. Cosmochim. Acfa SO, 401-408. KIRSTENT. ( 1968) Incorporation of rare gases in solidifying enstatite melts. J. Geophys. Res. 73, 2807-2810. LANGER. A. and AHRENST. J. ( 1982) The evolution of an impactgenerated atmosphere. Icarus S&96-120. LANGER. A. and CARMICHAELI. S. E. ( 1987) Densities of Na,OK~0-Ca0-Mg0-Fe0-Fe203-Al~O~-TiO~-SiO~’ liquids: New measurements and derived partial molar properties. Geochim. Cosmochim. Acta 51,2931-2946. LASAGAA. C. and CYGANR. T. ( 1982) Electronic and ionic nolarizability of silicate minerals. Amer. Mineral. 67, 328-324. . Lux G. ( 1987) The behavior of noble gases in silicate liquids: Solution, diffusion, bubbles, and surface effects, with applications to natural samples. Geochim. Cosmochim. Acta 51, 1549- 1560. MALVIND. J. and DRAKEM. J. ( 1987) Experimental determination of crystal/melt partitioning of Ga and Ge in the system forsteriteanorthite-diopside. Geochim. Cosmochim. Acta 51, 2 117-2 128. MARTYB. ( 1989) Neon and xenon isotopes in MORB: Implications for the earth-atmosphere evolution. Earth Planet. Sci. Lett. 94, 45-56. MATSUDAJ. and MATSUBARAK. (1989) Noble gases in silica and their implication for the terrestrial “missing” Xe. Geophys. Res. Lett. 16, 8 l-84. MA~SUDAJ. and NAGAO K. ( 1989) Noble gas emplacement in shockproduced diamonds. Geochim. Cosmochim. Acta 53, 1I 17- 1 I2 1, MATSUDAJ., MATSUBARAK., YAJIMA H., and YAMAMOTOK. ( 1989) Anomalous Ne enrichment in obsidians and Darwin glass: Diffusion of noble gases in silica-rich glasses. Geochim. Cosmochim. Acta 53,3025-3033. MATSLJIT. and ABE Y. (1986a) Evolution of an impact-induced atmosphere and magma ocean on the accreting Earth. Nature319, 303-305. MATSUIT. and ABE Y. ( 1986b) Impact-induced oceans on Earth and Venus. Nature 322, 526-528. MAZOR E., HEYMANND., and ANDERSE. (1970) Noble gases in carbonaceous chondrites. Geochim. Cosmochim. Acta 34, 78 l824. MELOSHH. J. and VICKERYA. M. ( 1989) Impact erosion of the primordial Martian atmosphere. Nature 338, 487-489. MUSSELWHITED. M., DRAKE M. J., and SWINDLET. D. ( 1991) Early outgassing of Mars supported by differential water solubility of iodine and xenon. Nature 352, 697-699. OZIMAM. ( 1989) Gases in diamonds. Ann. Rev. Earth Planet. Sci. 17,361-384. OZIMA M. and IGARASHIG. (1989) Terrestrial noble gases: Constraints and implications on atmospheric evolution. In Origin and Evolution ofPlanetary and Satellite Atmospheres (ed. S. K. AYEDA et al.) pp. 306-327. Univ. Arizona Press. OZIMA M. and PODOSEKF. A. ( 1983) Noble Gas Geochemistry. Cambridge Univ. Press. OZIMAM. and ZASHUS. ( 1991) Noble gas state of the ancient mantle as deduced from noble gases in coated diamonds. Earth Planet. Sci. Lett. 105, 13-27. PAUL A. ( 1990) Chemistry ofGlasses. Chapman and Hall. POREDAR. and DI BROZOLOF. R. ( 1984) Neon isotope variations in Mid-Atlantic Ridge basalts. Earth Planet. Sci. Lett. 69, 277289. PRESNALLD. C., DIXON S. A., DIXON J. R.. O’DONNELLT. H., BRENNERN. L., SHROCKR. L., and DYCUSD. W. ( 1978) Liquidus phase relations on the join diopside-forsterite-anorthite from I atm to 20 kb: Their bearing on the generation and crystallization of basaltic magma. Contrib. Mineral. Petrol. 66, 203-220. SARDAP., STAUDACHER T., and ALLBGREC. J. ( 1988) Neon isotopes in submarine basalts. Earth Planet. Sci. Left. 91, 73-88.
722
C. L. Broadhurst et al.
SASAK~S. and NAKAZAWA K. ( 1988) Origin of isotopicfractionation of terrestrial Xe: Hydrodynamic fractionation during escape of the primordial HZ-He atmosphere. Earth Planet. Sci. Lett. 89, 323334. ZHELHAAS N., OTT U., and BECEMANNF. ( 1990) Trapped noble gases in unequilbrated ordinary chondrites. Geochim. Cosmochim. Acta 54,2869-2882. STAUDACHERT. and ALLEGREC. J. ( 1982) Terrestrial xenology. Earth Planet. Sci. Lett. 60, 389-406. SWINDLE T. D., GARRISON D. H., GOSWAMIJ. N., HOHENBERG C. M., NICHOLSR. N., and OLINGERC. T. ( 1990) Noble gases in the howardites Bholghati and Kapoeta. Geochim. C~smoc~~rn.ilcfa 54,2183-2194. TURCOTTED. L. and SCHUBERTG. ( 1988) Tectonic implications of radiogenic noble gases in planetary atmospheres. Icarus 74,3646. UHLIG H. H. ( 1937) The solubilities of gases and surface tension. J. Phys. Chem. 41, 1215-1225. WACK~R J. F. ( 1989) Laboratory simulation of meteoritic noble gases. III. Sorption of neon, argon, krypton, and xenon on carbon: Elemental fractionation. Geochim. Cosmochim. Acta 53, 14211434. WACKERJ. F., ZADNICKM. G., and ANDER~E. ( 1985) Laboratory simulation of meteoritic noble gases. I. Sorption of xenon on carbon: Trapping experiments. Geochim. Cosm~him. Acta 49,10351048. WALKERJ. C. G. ( 1986) Impact erosion of planetary atmospheres. Icarus 48, 87-98. WATSONA. J., DONAHUET. M., and WALKERJ. C. G. ( 1981) The dynamics of a rapidly escaping atmosphere: Applications to the evolution of Earth and Venus. Icarus 48, 150- 166. WHITE B., BREARLYM., and MONTANAA. ( 1989) Solubility of Ar in silicate liquids at high pressures. Amer. Mineral. 74, 5 13-529. W~ELERR., GRAF TH., PEDRONIA., SIGNERP., PELLASP., FIENI C., SUTERM., Vffi~ S., CLAYTONR. N., and LAULJ. C. ( 1989) Exposure history of the regolithic chondrite Fayetteville: II. Solargas-free light inclusions. Geochim. Cosmochim. Acta 53, 14491459. WIENS R. C. (1988) On the siting of gases shock-emplaced from internal cavities in basalt. Geochim. Cosmochim. Acra 52, 27752783. WIENSR. C. and PEPINR. 0. f 1988) Laboratory shock empla~ment of noble gases, nitrogen, and carbon dioxide into basalt, and implications for trapped gases in shergottite EETA 7900 1.Geochim. Cosmoehim. Acta 52,295-307. YANGJ. and ANDERSE. ( 1982a) Sorption of noble gases by solids, with reference to meteorites. II. Chromite and carbon. Geochim. Cosmochim. Acta 46, 861-875. YANGJ. and ANDERSE. ( 1982b) Sorption of noble gases by solids, with reference to meteorites. III. Sulfides, spinels, and other substances; on the origin of planetary gases. Geochim. Cosmochim. Acta 46,877-892. YANG J., LEWISR. S., and ANDERSE. ( 1982) Sorption of noble gases by solids, with reference to meteorites. I. Magnetite and carbon. Geochim. Cosmochim. Acfa 46.84 l-860. ZAHNLEK. J. and KASTINGJ. F. ( 1986) Mass fractionation during transonic escape and impIications for loss of water from Mars and Venus. Icarus (i&462-480. ZAHNLEK. J., KASTINGJ. F., and POI.LACKJ. B. (1988) Evolution of a steam atmosphere during Earth’s accretion. fcarzcs 74,62-97. ZAHNLEK., BLASTING J. F., and POLLACKJ. B. ( 199Oa) Mass fractionation of noble gases in diffusion-limited hydrodynamic hydrogen escape. Icarus 84,62-97. ZAHNLE K., POLLACKJ. B., and KASTINGJ. F. (1990b) Xenon fractionation in porous planetesimals. &whim. Cosmochim. Acfa 54,2577-2586. Z~;HRINGER J. ( 1968) Rare gases in stony meteorites. Geochim. Cosmochim. Acta 32,209-237. ZHANG Y. and ZINDLERA. ( 1989) Noble gas constraints on the evolution of the Earth’s atmosphere. J. Geoph~ys.Rex 94, 13.7 1913,737.
APPENDIX
I
Figure A-l shows the patterns for some of the melts that inadvertently grew crystals. These results allowed independent confirmation of the trends observed in the minerals and melts, as well as an understanding of the Ar analytical problem. The particular crucible used for the last set of analyses had been used previously for SiOz samples containing Ar only, Those experiments were unrelated to this work. It appears that the SiOz was incompletety fused during these analyses, so some Ar was retained and degas& into our runs. The Ar did not show up in the blank runs; although the crucible was heated to 15.5O’Cduring blank runs, the dross in the crucible had aheady been degassed at this temperature, so it did not degas further until a sample was dropped. The silicate sample can sometimes act as a flux for the refractory residue material in the crucible and vice versa, as discussed in the text for spinel. It is difficult to attribute the Ar spike to an error in the gas composition for the following reasons:
1) The reported data for DJM 2DI(i) IV is the second of two aliquants. The first run had such a high Ar abundance that the sample could not be analyzed properly and had to be discarded. A leak in the system was suspected at the time, but none was present. 2) The Ar spike is a persistent feature but variable in relative height depending on the overall gas content of the sampies. For example, the peak is hardly noticeable in FO 12243 but very apparent in SP 1l-4877. 3) As shown in Fig. A-l, another sample of DJM ZDI( i) IV containing both melt and crystals was analyzed along with this group so that we could better understand some of the previous-melt results. Since DJM 2DI ( i ), IV samnles bear no relation to the V and VI samples other than the fact that they were analyzed in the same group; this shows that the changes to the furnace and the new tank of gas are not responsible for the Ar spike. 4) If the gas were, for sake of argument, 99% Ar, this would hardly change the Ar solubility values but would lead to astronomical solubilities for the other gases. We would also be obliged to make the same blanket corrections to the minerals run in these experiments, an adjustment for which we have no basis. the solubility behavior of noble gases in melts has already been established and shown to obey Henry’s Law at low pressures. It is unlikely that minerals show the same type of sensitivity to the &ascomposition.
Ne
A,
Iir
I
Xi?
gases FIG. A-l. Solubilities in Melts + Crystals. Symbols: 3AN(a) I, - n --; 3AN(b) 1, - q --; 3AN(c) II, - q -; DM-842 II, -*-O-e-; IDIII, **m****;ZDI(i)IV, *..O.*..
Solubility of noble gases in minerals and melts TABLE
A-L
Analysis of variance for Ne. Kr, and Xe solubiliries in mineral and melt
723 APPENDIX 2
samples for which multiple analyses exist. Units are ccSTP/gm-arm x 10-S.
Reproducibility of Results Total variance
Taking into consideration the data presented here and in BROADet al. ( 1990), it appears that variations of at least a factor of two are inherent in our experimental procedure. If we first consider Ar measurements for the anorthite samples, the Ar solubility in a given sample cannot always be reproduced better than this. AN WU 6 I7 has a total of eight measurements, averaging 1.1 + 0.5 X 10m5 ccSTP/gm-atm. This is not significantly different from the averages ofAN 1l-7779 (1.3 X 10e5) and AN 137041 (0.87 X 10e5) replicates. However, all these values are much lower than the average value for AN HO 228 (9.4 X 10-5). Although our data set is limited with regard to statistical analysis, there are five samples that have three replicates in the mixed noble gas experiments. Table A-l presents a simple analysis of variance for these samples and an additional melt. The purpose of this table is to test the hypotheses that ( 1) the variation between replicates is as great as the variation in samples, and (2) the result that the mineral solubilities are on the order of or greater than the melt solubilities is due to experimental error or uncertainty. While it is acknowledged that the total variance (SST ) is very large in this system, the results show that this is not due to differences among mineral structures, as would be the case for ionic trace elements Furthermore, the calculations show that the variance among replicates is much less than the variance among samples. Table A- 1 identifies DI 122742 as having consistently higher solubilities than the other diopsides and spinels, similar to AN HO 228 and FO 12243. The inclusion of melts in Table A- 1 is not critical to the conclusions. For example, Kr results for the minerals DI 122742, DI R18684, and SP 1 l-4877 gives SST = 393, SS, = 209, and SSw = 184. Adding in a column with two analyses of AN WU 6 17 and one of AN 1l7779 (whose solubilities cannot be distinguished) gives SSr = 184, SSA = 105, and SSw = 79. Therefore, we conclude that hypotheses ( 1) and (2) are likely to be false. HURST
Variance within sample replicates (SS,)
ssw = SST
ss,$
where n = number of samples m = number of replicates N=n*m
Ne
R18684
122742
F05B*
11.4877
12243
ZDI
I :
0.018 0.079 0.14
0.80 0.4 2.1
4.5 2.7 2.8
0.66 0.56 2.0
1.6 7.8 6.5
5.9 5.6 3.9
x
0.079
1.1
5.0
1.07
7.3
5.1
SST = 145 Kr
SSA = 127
SSW = 19
1 2
3
0.80 1.3 0.91
5.4 15.4 673
0.97 1.3 3.3
3.0 4.8 2.4
65.2 32.6 58.7
2.4 2.6 2.6
T7
1.06
9.03
1.9
3.4
52.2
2.5
SST = 6690
Xe
I
1 2 3
SSw = 663
SSA = M)27
2.1 4.5 6.8
7.1 125 9.6
0.33 I.1 3.0
4.6 3.9 3.6
143 35.4 96.4
2.1 2.6 2.4
(rr,,ct,,m,cu PI
Solubility and partitioning of Ne, Ar, Kr, and Xe in minerals and synthetic basaltic melts C. L. BROADHURST,I,* M. J. DRAKE, ’ B. E. HAGEE,’ and T. J. BERNATOWICZ’ ‘Lunar and PlanetaryLaboratory,Universityof Arizona, Tucson, AZ 8572 I, USA *McDonnell Centerfor the Space Sciences, Washington University, St. Louis, MO 63 130, USA
(ReceivedApril 16, 199I ; accepted in revised form November 27, I99 1) Abstract-We have measured the solubilities of Ne, Ar, Kr, and Xe in natural samples of anorthite, diopside, forsterite, spinel, and in synthetic basaltic melts. The minerals and melts represent equilibrium pairs in the Fo-An-Di-Si02 system. Samples were suspended in individual crucibles in a one-bar flowing mixed noble gas atmosphere at 1300or 1332°C for times ranging from 7-l 8 days. The solubilities in the minerals increase with increasing noble gas atomic number, and typical solubility values are surprisingly high. Samples of a particular mineral (i.e., anorthite) that came from different localities yield distinctly different and reproducible solubility values. This indicates that properties intrinsic to individual samples can influence solubility. Noble gases are likely to be sited in lattice vacancy defects. In contrast, the solubilities in the melts decrease with increasing atomic number. Our data overlap the low end of the range defined previously for natural basalts; however, the dynamic range of the values from Ne to Xe is not as great as for the natural melts. Solubilities correlate well with melt molar volume but poorly with density. The partition coefficients increase with increasing noble gas atomic number for all mineral/melt pairs. Such trends imply that the terrestrial planet atmospheres were not derived from partial melting of chondritic source material. Significant fractionation of Kr relative to Xe is not observed, ruling out an origin for Earth’s “missing” Xe via magmatic fractionation. Partition coefficient absolute values are frequently near or greater than unity. For example, the ranges for five diopside measurements are the following: DNe,,, 3.2-47. This indicates that magmatic transport 0.013-0.37; D&l,, 0.15-0.84; DKrll, 0.31-2.4; and Dxell, is not an efficient mechanism for degassing the terrestrial planets. Our results favor a catastrophic origin for the atmospheres.
INTRODUCTION
short time period relative to the age of the planet. Very early (i.e., pre-3.8 Ga for Earth) intense volcanic activity is an example of such. In addition, early planetary atmospheres may have consisted of captured solar nebula gas; or several hundred bars of steam (LANGE and AHRENS, 1982; MATSUI and ABE, 1986a,b; ZAHNLE et al., 1988) generated by accretionary impacts. These dense atmospheres were subsequently lost by mechanisms such as massive hydrodynamic escape (WATSONet al., 198 1; ZAHNLE and KASTING, 1986; HUNTEN et al., 1987; ZAHNLE et al., 1990a) or impact erosion ( LANGE and AHRENS, 1982; WALKER, 1986; AHRENS et al., 1989; MELOSH and VICKERY, 1989). Noncatastrophic degassing mechanisms are those which occur over the lifetime of the planet: solid state diffusion, mass transport by convective overturn, and magmatic transport. As discussed previously ( OZIMA and PODOSEK, 1983; BROADHURST, 1989; BROADHURST et al., 1990), diffusion is not effective. Mass transport by convective overturn can be demonstrated to be important only for Earth. Even then, partial melting resulting from adiabatic decompression of the rising mantle plumes is probably the major mechanism of noncatastrophic degassing. Although the production and transport of basaltic magma has long been recognized as an important degassing mechanism, until recently experimental data concerning noble gas partitioning were completely lacking. Noble gases were speculated to behave as uniformly incompatible elements during partial melting events, but quantitative melting models could not be constructed.
THE SOLUBILITY,PARTITIONING,and siting of Ar in minerals
and synthetic
basaltic
melts
were discussed
previously in is a companion to those results, extending them to include Ne, Kr, and Xe. The impetus for the partitioning experiments is the observation that the noble gas abundance patterns and primordial/ radiogenic isotopic component ratios of Venus, Earth, and Mars differ from each other as well as from the chondritic meteorites. The most obvious difference between the terrestrial planet and meteoritic trends is in the relative abundances of Xe. The ‘32Xe/84Kr ratio in the Earth’s atmosphere is over an order of magnitude less than the values typical of carbonaceous chondrites ( MAZOR et al., 1970) and ordinary chondrites ( Z;~HRINGER, 1968). Apparently, this “missing Xe” does not reside in any known surface reservoir, including oceans (OZIMA and POWSEK, 1983), sedimentary rocks ( BERNATOWICZ et al., 1984), polar ice (BERNATOWICZ et al., 1985), and deep-sea amorphous silica ( MATSUDA and BROADHURSTet al. ( 1990). This contribution
MATSUBARA, 1989).
Variations in planetary atmosphere noble gas abundances may be due to primordial heterogeneities or may have been determined by the interplay of the degassing mechanisms operating on each planet. The degassing of planetary mantles can be classified as catastrophic or noncatastrophic. Catastrophic degassing events are those which occur in a very * Present address: Code 693, NASA Goddard SpaceFlightCenter, Greenbelt,MD 2077 I, USA. 709
C. L. Broadhurst et al.
710
HIYAGON and OZIMA ( 1986) performed the first experimental equilibrium noble gas partitioning study. They investigated olivine/ melt and diopside/ melt systems and found that noble gases were only somewhat incompatible, with partition coefficients increasing from He to Xe. In this study, we report an extensive series of partitioning experiments with anorthite, diopside, forsterite, spinel, and synthetic basaltic melts. Our results for all four mineral/melt pairs show the same partitioning trend as found by HIYAGON and OZIMA ( 1986); however, we find the surprising result that noble gases are often compatible in minerals. Since such partitioning behavior greatly complicates efforts in magmatic modelling, we offer only limited interpretations of the data. Instead, we present the results of many experiments, including those in which experimental errors occurred. We wish to demonstrate that noble gases will be fractionated in mineral/melt systems yet they cannot be expected to behave in the relatively predictable fashion that characterizes ionic trace elements. EXPERIMENTAL
PROCEDURES
The following section summarizes the experimental technique; detailed presentation is given in BROADHURST ( 1989) and BROADHURSTet al. ( 1990). Information concerning the starting material sources, compositions, and preparation methods is also given in these references. The additional melt compositions prepared for this study are given in Table 1. Mineral starting materials were eleven natural samples very close to endmember composition (five anorthites, three diopsides, two spinels, and one forsterite). We chose these samples based on availability and purity. Most importantly, tests with candidate mineral samples containing only a few wt% Fe0 gave poor results; strong color changes in the mineral powders indicated that oxidation/reduction was occurring. Since we did not want to introduce the additional variable of controlled oxygen fugacity, only nearly Fe-free material was used. All experiments used cl0 pm powders, except for samples DI 122742 in experiment NGX-VI and FO 12243 in experiment NGXV, which were ~38 pm (400 mesh) powders. The 110 pm powders were produced by grinding the minerals in a sapphire mortar and forcing them through a series of sieves. After several regrindings, a TABLE 1. Bulk compositionsof glassesused for synthetic melt stating materials in oxide weight percar. Microprobe analysesperformed at the University of Atiz~na. All analysesassumeFe0 = Fe total. EQ phace representsthe mineral phasethat the melt compositionis in quilihrium with in the An-Di-Fo-Si02 systemat 1300°C: AN = anotthite, DI = diopside, FO = forerite.
small amount of powder was forced through a 10 pm nylon filter mesh in an acetone slurry. The residue was evaporated to produce the starting material. The ~38 pm material was ground to pass through a 400 mesh sieve but did not pass through the nylon filter mesh. Optical examination of the powders showed them to be roughly spherical and not appreciably elongated or tabular. Silicate liquid starting materials were prepared as melts from oxide powders and quenched to glasses. The compositions are in the system forsterite-diopside-anorthite-silica ( PRESNALLet al., 1978) and were based on the experiments of MALVIN and DRAKE( 1987). The AN, DI, and FO series glasses are in equilibrium at 1300°C with the minerals anorthite, diopside, and forsterite. respectively. Again, these synthetic compositions were chosen in order to avoid FeO/FelOs equilibria problems and to reduce the complication of the experimental system so that we could focus clearly on the noble gas mineral and melt solubilities. The minerals and glasses were loaded into separate open cylindrical Pt crucibles. The crucibles were then suspended in a vertical alumina muffle tube gas-mixing furnace. We used a commercial gas mixture ( Alphagaz Liquid Air) consisting of 5.00% Ne, 93.0% Ar, 1.00% Kr, and 1.OO%Xe by weight. The gas flowed through a metering system at 90 cc/min into the top of the furnace and exited such that the pressure in the muffle tube was constant and less than l-2 torr over ambient. After a specified length of time, the samples were quenched by quickly removing the entire suspension system from the furnace. Experiments NGX-IV and MT were quenched in air and then in water. The suspension bucket was dipped in water, but the samples were not submerged or othenuise wetted. Under no circumstances was there any evidence of reaction of the samples with the crucibles. The first four experiments (NGX-I to IV) were run at 1300 + 2°C for eighteen days. At this temperature, the melt compositions were close to saturation in their respective mineral phases, and some were not analyzed because they grew crystals during the experiments. The last three experiments (NGX-MT, and V and VI) were run at 1332 f 2°C for seven and nine days, respectively. This slight temperature increase ensured that the melts were completely liquid, and it is unlikely to have significantly affected the results as noble gas solubilities have been demonstrated to be rather insensitive to temperature (JAMBONet al. 1986; Lux, 1987; WHITE et al., 1989). Experiment NGX-MT marked the only significant variation of the experimental technique. This experiment was basically a test which confirmed that we could avoid the crystallization problem mentioned above by running at 1332°C. As this was an unanticipated final usage of the first mixed noble gas tank, the gas flow was reduced to 50 cc/ min. After NGX-MT and prior to experiment NGX-V, the furnace was refitted with new heads, muffle tube, o-rings, and gas inlet lines, and a second tank of the noble gas mix was started. Diffusion A major concern of this experimental approach is ensuring that we attain solubility equilibrium in the minerals. Accurate K/Ar dating has long been the major impetus for Ar diffusion studies. FOLAND (1974) gives D = 0.00982 exp[-43.8/RT] cm’/sec for Ar in feldspar (spherical geometry). At 13OO”C,the characteristic diffusion time for 10 pm grains is 62 sec. This stands in contrast to eighteen days, or 1.6 X IO6 sec. In addition, BROADHURSTet al. ( 1990) presented a successful reversal and rate study experiment for Ar. The rate study, in particular, indicated that solubility equilibrium is achieved within two days. Unlike Ar, diffusion rates for Ne, Kr, and Xe in minerals are not available. We can, however, get a feel for the relative diffusion rates by examining data from carbon ( WACKERet al., 1985), diamond (OZIMA, 1989), silicate melts (HAYATSUand WABOSO,1985; JAMBON et al., 1986; Lux, 1987), and glasses (MATSUDAet al., 1989; CARROLLand STOLPER,199 1) . It is clear that Ne will diffuse more rapidly, and Kr and Xe more slowly, than Ar. If we make the conservative estimate that D ( 1300°C) values for Kr and Xe are one to two orders of magnitude less than those for Ar, we have still allowed ample time to achieve diffusional equilibrium. Even if solubility equilibrium has not been achieved for the heavier noble gases, we do not feel that our results are compromised. On the contrary, our solubility values would then be lower limits only, which make the results even more surprising.
711
Solubility of noble gases in minerals and melts
NOBLE GAS ANALYTICAL PROCEDURES Gas cleanup and separation was similar for all the experiments, and the procedures used have been developed from many years of sensitive noble gas measurements at Washington Univ. Full details aregivenin BROADHURST( 1989)and BROADHURST etal.( 1990). However, noble gas measurement techniques in the first two mixed noble gas experiments (NGX-I, II) differed from the remaining five (NGX-III to VI and MT). In the first two experiments, a Dycor model MZOOMquadrupole mass spectrometer (QMS) was used for the analysis of Ar. The absolute Ar ion beam signal was dete~in~ by at least sixteen measurements of the *Ar signal and extrapolation of these data to the time of initial generation of the Ar ion beam. The sensitivity of the QMS depends upon the total ambient gas pressure. and because this pressure is due almost entirely to Ar, we calibrated the sensitivity over a wide range of pressures using known quantities ofatmospheric Ar. Neon, Kr, and Xe were measured in the first two experiments with a Reynolds-type instrument operating in electron multiplier mode. Abundances were determined by peak height comparisons with standard air pipettes. In this procedure, only a fraction of the sample Ne was used (about 10%). and this amount was obtained by trapping the other noble gases on activated charcoal at LN2 temperature and partitioning the Ne between accurately calibrated volumes. The entire sample amount of Kr and Xe was sub~quentiy collected on another charcoal linger and unironed between two volumes, the larger of which was used for Xe m~surement and the smaller for Kr. The gas abundances were determined from the calibrated instrumental sensitivities by extrapolating the measured reference isotope (“Ne, s“Kr, “*Xe) signals to the time of initial generation of the ion beam. The first two experiments also used a two-step extraction, with gases measured at -800°C and - 1550°C. Two steps were also used for experiment AAN in BROADHURST et al.(1990).In all three experiments, very little gas was released at the lower temperature step, so we discontinued this very time-intensive practice for the remainder of the analyses. At this point, it was beginning to be clear to us that properties intrinsic to individual mineral samples could influence noble gas solubility, and we wanted to continue with more precise measurements in order to reduce the overall abundance uncertainty as much as possible. Beginning with ex~~rnent NGX-III, we determined all noble gas abundances using isotope dilution techniques with a mixed noble gas spike. Isotopic ratios were measured with a Reynolds-type instrument operating in static Faraday cup mode. Data acquisition (for either mode) was done by computer-controlled peak-switching, and fifteen scans of each gas spectrum were taken. During Ne runs, the gas was exposed to charcoal at LN2 temperature to minimize 40Ar++ and H,‘sO+ interferences at “Ne and CO?’ at **Ne. The sinelvcharge2 species were also monitored duringNe runs, and all Ne &a were corrected for interferences from these ions. For the isotope dilution procedure, we first collected all the sample noble gases on activated charcoal cooled by a He cryostat to 10°K. We then mixed an aiiquot of spike gas with the sample gas. The noble gas spike contained the following composition: 22Ne= ( 1.394 f 0.06 1) X lo-’ ccSTP with 22Ne/2?Je = 5.612 rt 0.048; %Ar = (2.980 20.043) X 10m7ccSTP with 40Ar/MAr = 0.0987 It 0.0013; *& = 100% and f 1.955 + 0.0169) X 10v9 ccSTP; lJ6Xe = (I.547 t0.0165) X 10e9 ccSTP with ‘36Xe/‘3ZXe = 7.439, ‘29Xe/‘32Xe = 0.9841, ‘3’Xe/‘3*Xe = 0.7885. Other Ne, Ar, and Xe isotopes are present in much lower abundances and are not critical to the analysis. The experimental uncertainties in noble gas abundances determined by isotope dilution are ~2% for Ar, Kr, and Xe, and ~4% for Ne, exclusive of uncertainties due to corrections for instrumental background (“blanks”). Instrumental blank levels were determined between individual sample runs for experiments in which the gas abundance, and consequently instrumental memory, was high. They were determined between groups ofseveral runs for cases where both blanks and sample gas abundances were low. In practically all cases, the blank corrections were less than 10% of the measured Ar abundances, and a 50% uncertainty in the blank corrections was assumed. For abundance dete~inations based upon in~Nrnen~ sensitivity, the blank correction
and uncertainty have been included in the overall uncertainty (25% for the QMS and 15% for Reynolds instrument determinations). For the isotope dilution experiments, the overall abundance uncertainty including blank correction is given. RESULTS For reasons of clarity, results for each mineral are presented separately in the text below, followed by the melt data. In the co~es~nding Figs. la-f, the different samples are identified by their catalog numbers, followed by a Roman numeral indicating the experiment number. Note that the ordinates are scaled differently. Measured isotopic abundances and solubilities for the set of experiments are given in Tables 2 and 3, respectively. Anorthite Noble gas solubilities in anorthites are shown in Fig. la. Samples AN HO 228, AN 1 l-7779, and AN WU 617 were run together in experiment NGX-I in an attempt to reproduce the results from the 99.95% Ar synthesis experiment AAN (BROADHURSTet al., 1990). The behavior was consistent: HO 228 again showed very high soiubilities; 1l-7779, intermediate values; and WU 6 17, the lowest values. The solubility trend for WU 617 was reproduced in experiment NGX-II, except for a slight decrease in the Xe/Kr ratio. All of these samples show a clear trend of increasing solubility with increasing noble gas atomic number. Sample AN 4079 III shows a nearly flat pattern, which reflects some contamination with glass material, probably introduced to the sample as the set of crucibles was removed from the suspension bucket directly after the experimental run. Optical examination of AN 4079 III powder showed no signs of partial fusion during the synthesis experiment. although the powder was somewhat packed together. This sample was also examined by transmi~ion electron microscopy (TEM) and did contain some grains that did not yield a diffraction pattern ( BROADHURST, 1989 ). This topic is discussed further below.
Diopside Diopsides also show an overall trend of increasing solubility with increasing noble gas atomic number (Fig. 1b). DI 122742 yielded consistently higher solubiiities than DI R 18684. The single analysis of DI M2506 cannot be distinguished from the range represented by R 18684, and both of these samples had some of the lowest solubilities of all the minerals studied. The solubilities in the 38 pm DI 122742 material run in experiment VI did not differ significantly from the other samples. The 122742 powders in experiments V and VI were somewhat packed together, similar to AN 4079 above. Other diopsides run in these 1332°C experiments, as well as some test diopsides at 13OO”C,did partially fuse and consequently were not analyzed. Forsterite The highest nobie gas soiubility overall was observed in forsterite FO 12243 (Fig. lc). This sample shows a clear trend of increasing solubiIity and was the most reproducible
C. L. Broadhurst et al.
712
la.
anorthite
Ne
I-3
A7
X.2
AI
Kr
Xe
gases
gases
lo33
1
le. forsterite
Id.
1000
Ne
Kr
AI
Xe
I
I
I
AI
Kr
Xe
Kr
Xe
gases
gases
Ne
Kr
AZ
gases
spine1
Xe
Ne
AI
gases
FIG. I. (a) Noble gas solubility in anorthites. A slight peak is present at Ar in An 4079 VI. Data for AN 4079 III indicates that the sample was probably contaminated by melt. Symbols: WU 6 17 I, - 0 --; WU 6 I7 II, - a --; 40791II,--~--;4079VI,--~--;HO228I,--O--;11-77791,--A--.(b)NoblegassolubiIityindiopsides. The Ar solubility has been interpolated (see text) for DI 122742 VI: the measured value is denoted by *. Symbols: 122742 V,--a--; 122742VI,--0--;R18684I1,-~---;R18684III,-+---;M2506 122742 IV, --•--; IV, - - 0 - -; 122742 VI*, 0. (c) Noble gas solubility in forsterite. A slight peak at Ar is present in the data from V. Symbols: 12243 II, - 0 --; 12243 III, -H -; 12243 IV, -Cl --; 12243 V, - fB -, (d) Noble gas solubility in spinels. The Ar value for SP 1l-4877 VI has been interpolated as described previously. Symbols: 1l-4877 III, - q -; I l-4877 IV, - 0 --; 1i-4877 VI, - m -; B12099 IV, - - 0 - -; 1l-4877 VI*, n . (e) Nobfe gas solubihty in DIseries silicate melts. Argon solubilities for DJM 2DI( ii) have been interpolated. Symbols: 2 DI(i) IV, - - n - -; 2 IX(i) MT, - - 0 - -; 2 Di(ii) V, - 0 -; 2 Dijiif VI, - 0 --; 2DI V*, 0: 2DI VI*. 4. (f) Noble gas solubility in FO- and AN-series meits. Argon values for DJM F05B have been interpolated. Symbols: FO 5B V, - - A - -: FO 53 VI,-A---;7ANMT,-m--;FOV*,&FOVI*,A.
Solubility of noble TABLE 2. MEASURED
NOBLE GAS ISOTOPIC IN ccSTP/gm
gases in minerals
CONCENTRATIONS
TABLE 3. Noble eas absoIute solubilities
in ccSTP/em-atm noble gas syntheses. Experiments conducted at atm&pheric 93% Ar. 1% Kr, and 1% Xe by weight. Ar values which described in text are denoted by &. -
NGX-I AN HO228
0.0956
800 15&l total
2.8(-6) 2.2(-7) 3.0(d)
9.61-8) 96(-S) 9.R.5)
6.5(-10) 1.1(-6) 1.1(-Q
5.3(-IO) 9.q.7) 9.q.7)
AN II-7779
O.l(rx
Boo 1550 total
3.w.7) 1.7.(-7) 3.8(-7)
8.3(-9) l.7(-5) I .U(-5)
3.q.IO) 1.7(-7) 1.7(-7)
3.3(-IO) 1.3(-7) 1.3(-7)
ANWU617
0.1234
SKI 15% total
7.x.9) 6.q.91 ,.4(.8)
8.11-8) 3.4(6) 3.5(-6)
1.6x-10) 3.81-8) 3.8(-S)
1xIO) 2.9(-8) 2.9~.8)
DJM 3AN*
0x690
800 1550 total
1.q-61 l.q.5) 1.1(-S)
7.9(-7) 9.1(-S) 9.2(-5)
5.2(-9) 6.3(-7) 6.3(-7)
3 x-91 2.7(-7) 2.8(-7)
NCX-II AN WU 617
0.0926
ISSD
6.x-8)
7.1(d)
5.S(-8)
1.5(-X)
DI RlX6&1
0.0773
800 1550 lOtal
1.q.8) 1.61-8)
2.8(6) 6.3(-7) 3.4(x$
6.7(-9) ,.4(-E) 2.1(-S)
5.6(-10) 2.0(-E) 2.0(-R)
800 1550 total
28~.8) 6.5(-6) 6.5(-6)
4.8(-7) ,.X4) I 5w
4.9(-9) I .7(-Q I .7(-6)
3.2(-9) I.,(-6) l.l(-6)
FO I2243
0.0792
DJM 3AN’
0.0478
15M
1.3(-S)
l.l(“t)
6.2(-7)
2.0(-7)
DJM IDI’
0.0674
8W 15% tolal
3.2(.6) 2.q-6) 5.2(-a
8.21.7) 8.3(x$ 9.1(d)
5.5(-9) 4.4(-S) 4.91-Q
4.7(-9) 2,3(-S) 2.8~.81
x 10-5 for the mixed pressure in 5% Ne, are slightly high as
experiment
Ne
Ar
Kr
XC
I II Ill VI 1 I
0.016 0.073 0.80 0.096 3.5 0.43
0.39 0.79 0.79 b.6* 10.7 2.0
1.4 2.1 0.80 10.9 41.3 b.3
3.9 1.9 0.82 17.9 I19 17.2
d&p&G R18684 R18684 122742
II III N N
0.018 0.14 0.079 0.80
0.38 g
0.80 1.3 0.9 1 5.4
2.1 4.5 6.8 7.1
122142 M2506
z N
0.41 2.1 0.20
7.0’ 2.6’ 0.41
6.3 15.4 0.74
!% 7.1
forsterite 12243 12243 12243 12243
II III IV V
7.6 6.5 6.2 7.8
16.7 19.0 19.0 30.7*
65.2 32.6
143 35.4
VI N
0.66 2.0 0.56 052 b.0
SZ3IIpk Experiments NGX-I and II mn at 1300@Cand @nospheric pressme using a minti noble gas atmosphere of 5% Ne. 93% AI. I W KI, 1I Xe by weight. Gas concentrstions detcmkxi by peak height comparison with error k 25% for Ar and +_15%for the olhw gases.
713
and melts
onorthite WU617 WIJ617 4079 4079 HO 228 11-7779
32.6
32.9
58.7
96.4
2.0 3.7 5.b*
3.0 4.8 2.4
1.5
4.6 3.9 3.6 1.3
1.0 2.8 1.5 13.0* 11.8* 4.-l* 16.7’ 1.2
1.9 2.4 0.97 2.6 2.6 1.3 3.3 0.61
3.6 2.1 0.33 2.4 2.4 1.1 3.0 0.33
sainel
ii-4877 Il.4877 1l-4877 812099 melts 1DI ZDI(i) 2DI(i) 2DI(ii) ZDl(ii) FOSB FOSB 7AN
II zr ; z tvfr
5.9 4.5 3.9 5.6 2.8 7.7 3.4
i1
Experiments NGX III and IV run at IxKlOC and atmospheric pressure using a mixed noble gas atmosphere. Gas concenmaions dewmined by isotope dilution: emx’s are listed. NGX-III AN 4079
0.0542
1550
6.9t2.1(-8)
7.1+1.1(-6)
2.1?0.02(-8)
&liO.M((-9)
DI RI8684
0.0552
1550
1.2f0.2(-7)
1.3?00.1(-5)
3.4*0.03(-s)
3.4M.03(-8)
0.0124
1550
5.660.3(-6)
1.7M.O4(-4)
8.6+~0.09(-7)
2.7iO.O2(-7)
0.0556
1550
S.?M.3(-7)
1.&0.1(-S)
7.wl.O7(-8)
3.5iO.O3(-8)
NGX-IV DI RI8684
0.0567
1550
6.w2.0(-8)
6.4fl.l(-6)
2.4a.o2(-83
5.1M.05(-8)
DI M2506
0.0515
155O
1.7to2-7)
3.7+1.2(-6)
1.9+0.02(-8)
5.5+0.05(-83
Dl 122742
0.0459
1550
6.9kO.4(-7)
Z.lM.l(-5)
1.4~0.01(-7)
5.2+0.05(-X)
FO 12243
0.0113
1550
5.4to.3(-6)
1.7*.04(-4)
B.&O.OY(-7)
2.5~0.02(-7)
SP II-4877
0.0157
1550
1.7zkO.O8(-7)
3.3~o.08(-5)
1.2k0.01(-7)
2.9+0.02(.X)
SPBlZO99
0.0642
1550
4.4M.24(-7)
l.OaOS(-5)
3.~.04(.8)
9.9M.O8(-9)
DIM 2DI(i)
0.0585
1550
5.OkO.2 (-6)
2.5?0.14(-5)
6.5m.063(-8)
1.6Fo.O1(-8)
FO 12243 SP
i 1.4877
one in the study. The solubilit~es in the 38 pm material in experiment V did not differ significantly from the other samples. Spinet
Expaimenrs NGX-MT, NGX-V, and NGX-VI run at 13324: BLalmospheric pressure using a mixed noble gas almosphere Gas concentrationa derermmed by isotope dilution. Ar values in V and VI arc slkhdy high due ID experimenml diflicuhies as described in text. NCX-MT D&3 7AN
0.0134
1550
2.9+0.1(-6)
1.1*0.2(-s)
I .6+0.02(-R)
?.SiO.4(-9)
DIM 2DI(I)
0.0330
I550
3.9iO.2(-6)
1.3FO.OR(-5)
2.6+0.03(-8)
2.5iO.O5(-9)
NGX-V DI 122742
0.0498
1550
3.6+0.38(-73
2.3&0.16(-S)
4.010.052(-7)
9.3+0.53(-7)
FO 12243
0.0126
ISM
6.7ti.33(-6)
2.8+0.008(-l)
lSkO.O19(-6)
5.8?0.1 If-71
DJM ZDI(d)
O.WrlR
,550
3.4ti.l6(-6)
l.ZM.023(-4)
6.8iO.063~.8)
2.010.014(-X)
DJM FOSB
0.0687
1550
2.4M.I I(-6)
4.210.067(-5)
3.4~.031(-8)
R.O+O.O65(-9)
NtiX-“I AN4079
0.0851
I550
8.2Kl.43(-7)
5.9k0.14(-5)
2.8*0.038(-7)
1.3+0.024(-7)
Dl 122742
0.0575
15%
l.gtO.o88(-6)
6.3+0.17(-S)
1.6aOl7(-7)
7.4kO.O73(-8)
SP I14877
0.0453
1550
4.&0.44(-7)
5.lxO.19(-5)
6.0+0.057(-S)
2.7+0.019(-8)
DIM ZDI(ii)
0.0538
1550
4&O.24<-6)
1.010.017(-2)
6.9+0.0&?+8)
1.&0.013(-8)
DIM FO5B
0.0594
1550
6&0.35(‘5)
1.5*.028(-4)
8.7+O.086(-8)
2.3kO.017(-8)
*melts conramcd crystals, so results are not discussed m text
Spine1 samples (Fig. Id) show a weaker increase in solubility with atomic number than do the silicate minerals. Two analyses show a decrease in the Xe/Kr ratio. Spine1 is more refractory than the other minerals investigated, and we are concerned that it may not have fused completely during gas extraction. The maximum temperature achieved by the RF induction heating was - 1600°C. However, the degassed remains of many previous samples are retained in the crucible, and this dross acts as a flux for refractory materials. Incomplete fusion was not suspected, for example, in the forsterite analyses, but it is a possible explanation for the decrease in the spine1 Xe/Kr ratio since sample degassing becomes diffusion-limited. The overall weaker increase in the solubility trend is also consistent with incomplete fusion. Since many other samples were analyzed both before and after these two spinels, it was not possible to verify this explanation. Very conservatively, we might consider that the gas concentrations measured in the forsterite and spine1 samples are lower limits. This would make the results yet more unusual as we could presumably retrieve even more gas if we melted the samples completely. An anomalous peak is present at Ar in some of the data from experiments NGX-V and VI, which is probably due to a memory effect in the gas extraction manifold during the final set of sample analyses. Further details are given in Appendix 1. Solubility trends in several of the figures have been
C.
714
L. Broadhurst et al.
roughly corrected by allowing the plotting program to calculate an equation for the line and estimate a value. The uncorrected values are denoted by *, as noted in the captions.
taining bubbles were analyzed. We also observed that the samples that were quenched in air and then in water did not contain higher gas concentrations.
Silicate Melts
Solubility Patterns of Melts + Crystals
Figure 1e shows the results for DI series melts, and Fig. 1f shows the 7AN and FOSB series melts. The melts show a clear trend of decreasing solubility with increasing noble gas atomic number. Solubiliti~ are also surprisingly low, often on the order of or lower than solubilities in the minerals. Figure 2 summarizes data from this study, with the range from other authors dashed in. The solubilities in our synthetic melts tend to be lower than those in the natural basal& especially for Ne and Ar. Our results overlap the low end of the natural basalt range and are most similar to molten enstatite from KIRSTEN ( 1968 ) . It is unlikely that our samples suffered gas loss upon or after quenching. Our experimental and analytical techniques were comparable to the other authors ( KIRSTEN, 1968; JAMBON et al., 1986; HXYAGONand OZIMA, 1986; Lux, 1987), all of whom had adapted procedures adequate for retention of He. Equilibmtion of Ne in the melts takes a few hours at minimum, so large amounts of gas are not expected to be lost during our quenching at 600-l 100°C/min. JAMBONet al. ( 1986) investigated noble gas diffusive loss from their glass samples. A one-day heating at 2OO’C resulted in no gas loss. Ten-day heatings at 600°C resulted in total He loss and significant Ne loss, which could be correlated with grain size. No measurable loss of Ar, Kr, or Xe was observed. It is possible, however, that in our study, as well as others, gas was lost directly after removal from the furnace. This gas would be lost not by difftsion but by the formation of bubbles on the melt surface. In general, our glasses were bubble-free and did not have a foamy surface. Specifically, no chips con-
Figure Al in Appendix 1 shows the noble gas solubilities measured in some of the glasses that grew crystals during the course of an experiment or upon quenching. Sample DM842 contained large euhedral olivine crystals that clearly grew suspended in the melt during the experiment. The DJM 3AN melts contained numerous small feathery crystals that may have grown during the experiment, upon quenching, or at both times. DJM ID1 and 2DI each grew only a small proportion of crystals, which sank to the bottom of the crucible, leaving a clear glass cylinder above. The solubility pattern is nearly flat, with a slight V-shape. The sides of the V have shallower slopes than the patterns for either minerals or melts alone. This shape is readily interpreted as a superposition of the solubility trends for the mineral and melt fractions. The exception is sample 2DI (i) , which shows a strong peak at Ar. This sample was run in experiment IV but analyzed along with the final set of V and VI runs. The anomalous pattern seen here was instrumental in determining that an Ar analytical problem occurred in that final set. Some of these melts also have higher overall noble gas solubilities than the melts that did not grow crystals. This may be attributable to gas inclusions in the minerals or trapping of bubbles during quenching. Optical examination of the glasses did not reveal any obvious bubbles; however, in the case of all but DM-842, the numerous small crystals made the images very confusing. It is possible that inclusions could have been overlooked or were too small to resolve; however, there could not have been too many because our gas mix was 93% Ar. Large or numerous bubbles of gas would lead to a huge gas release upon analysis with a composition very close to that of the experimental atmosphere. Assuming that inclusions are not important, then these experimental mistakes provide additional evidence that the solubility of noble gases can often be higher in minerals than in melts. It is to be noted that the noble gas concentrations in the melts are enriched, rather than diluted, by the addition of crystals. These crystal + melt patterns also led us to suspect that AN 4079 III had been contaminated with some glass, and thus follow through with TEM examination. While admittedly anomalous, AN 4079 III shows a strong increase in the Ne abundance and a strong decrease in the Kr abundance as compared to all the other anorthite samples. DISCUSSION
gases FIG.2. Overlapping heavy lines summarize noble gas solubilities in synthetic iron-free melts from this study: (B2) DJM 2DI( i) IV at 1300°C. (B3) DJM 2DI( ii) VI, (B4) DJM F05B V, and (BS)DJM
7AN MT at 1332°C. Range of melt solubility data from HIYAGON and OZIMA(1986), JAMBON et al. (1986), and Lux (1987) is dashed in; (KI ) refers to molten enstatite at 1500°C from KNSTEN( 1986). Symbols: Kl, x; B2, - l --; B3, -Cl--; B4, - 0 -: BS,- A -.
Noble Gas Solubility Trend in Minerals Noble gases show a clear trend of increasing solubility with increasing noble gas atomic number for all the minerals in this study. Some periodic properties of the noble gases are also correlated with atomic number. Of these properties, pola~~bility is most likely to be of importance since noble gases will not be ionized nor cont~bute significantly to co-
Solubility of noble gases in minerals and melts valent bonding in geochemical environments. They do not have permanent dipole moments; therefore, potential solubihty mechanisms must involve induced dipole interactions. which are governed by an atom’s electronic polarizability. Two of these interactions are the weak inverse r”-dependent-dipole/ induced-dipole and induced-dipole/induceddipole interactions. These types of interactions are used to model the formation of noble gas solids and the physical adsorption of noble gases on surfaces. Adsorption of noble gases is reviewed in detail by OZIMA and PODOSEKf 1983) and is further addressed in YANG et al. (1982), YANG and ANDERS (1982a,b), BERNATOWICZand PODOSEK(1986), and BROADHURSTet al. ( 1990). It is known to be a lowtemperature phenomenon: A typical lifetime for physisorption at room temperature is about lo*.* sec. Quantitative physi~~tion of Ne and Ar requires 77% Xe and Kr 192*K. Prior to analysis, we placed the samples in a high-vacuum manifold and heated them at 100°C for -48 h. This is a long-accepted mass spectrometric practice for removal of the bulk of surface adsorbates and/or atmospheric contamination, both in our laboratory and in numerous other noble gas research groups (AMARI and OZIMA, 1988; MATSUDA and NAGAO, 1989; WIELER et al., 1989; SCHELHAASet al., 1990; SWINDLEet al., 1990). Reactive gases are sometimes observed to form chemical bonds with surfaces. Enthalpies of adsorption for such interactions are on the order of -50 to -200 k&/mole, whereas enthalpies for physiso~tion are on the order of > - f 0 kcal/ mole. This energy difference is one of the criteria for hypothesizing a surface chemical interaction. Enthalpies of adsorption for Kr and Xe on a number of silicate substrates are compiled in OZIMA and PODOSEK( 1983), and no value more negative than -6.7 kcal/mole is reported. WACKER ( 1989) investigated the sorption of Ne, Ar, Kr? and Xe on carbon black, acridine carbon, and diamond. These surfaces are considered to be much more active than silicate or oxide minerals; however, enthalpies of adsorption still ranged from -2 to -8 kcal/mol, fully consistent with physisorption. As reported in a series of three papers, YANG et al. ( 1982) and YANG and ANDERS ( 1982a.b) synthesized magnetite, chromite, sulfides, and spine1 in a mixed noble gas atmosphere at temperatures from 167-447”C. One important result of the study was that samples record the noble gas signature of the highest pressure reservoir that they encounter. This, in effect, means that experiments conducted at noble gas partial pressures less than atmospheric yield results that are indistinguishable from air contamination. When partial pressures are higher, distinction between components can be made. Thus, the authors divided gases into groups based on release temperature: ls4OO”C, physisorbed; 18OO”C, chemisorbed; and >8OO”C, trapped. (“Trapped” gases are contained within the mineral lattices; however, this term is used rather that “dissolved” because there was no attempt to attain an equilibrium solubility). It is hypothesized that during mineral synthesis, trapped gases are first surface adsorbed and then become “fixed” in the crystal by some unspecified process, probably involving a combination of recrystallization, chemical reaction, diffusion, and migration to traps. This “fixing” turns ph~isor~d gases into more strongly bound components. However, the majority of gases released
715
from the samples upon analysis are still classified as adsorbed, not trapped. In their “trapped” nobie gas com~nent, YANG et al. ( 1982) and YANG and ANDERS ( 1982a,b) found the same fractionation pattern that we retrieved from our mineral samples. Despite the differences that minerals were grown during the experiments and that temperatures were much lower, this study shares another conclusion with ours: Noble gases introduced into minerals can be persistently retained. The two-step heating procedure used for some of our noble gas analyses shows that the gas atoms are held in very retentive sites. With the exception of Ne, very little gas is released at the 800°C step (see Table 2), and the samples must be fused to extract the majority of the gas incorporated in the minerals at 1300°C or 1332°C. This observation argues strongly against simple physiso~tion and against chemiso~tion as defined by YANG et al. ( 1982). The higher proportions of Ne lost at 800°C were anticipated, and we attribute this to faster diffusion as well as a less energetic interaction with the mineral lattice. Preferential loss of Ne has been observed in shock-implanted noble gases in basalt ( WEINS, 1988; WEINS and PEPIN, 1988), as well as in glasses ( JAMBONet al., 1986; MATSUDAet al., 1989 ). If our results are due to adsorption and/or atmospheric contamination, then they are an unprecedented example of such and would lead to the conclusion that surface-bound gases are so persistent that they cannot be removed from any arbitrary mineral sample. We might also consider the distinction between “adsorption” stemming from many days of exposure to noble gases at 1300°C and “true solution” to be a moot point. Although it is clear that we cannot fully determine where the noble gases are sited in the minerals, we hypothesize that they are situated in lattice vacancy defects. The supposing evidence, more fully detailed in BROADHURST( 1989) and BROADHURSTet al. ( 1990), is as follows:
1) The variable solubilities of noble gases in different natural samples of the same mineral require a solubiiity mechanism that can vary significantly among samples with almost identical major element compositions and, in some cases, geologic occurrence. 2) The two-step heating procedure for some of the noble gas analyses shows that the gas atoms are held in retentive sites. 3) Interstitial sites cannot be ruled out, but it is difficult to explain the solubility variation, as interstitial sites should not vary greatly among nearly identical mineral samples. Purely interstitial siting for noble gases would predict a mineral solubility trend opposite to that observed (i.e., similar to the melt trend). We might also expect to observe much higher ~lubilities in the tectosilicate anorthite, which is far less dense than the three other minerals investigated. 4) Transmission electron microscope examination of some of the samples did not reveal any anomalous microstructures or dislocation densities, or any evidence of healed mi~rofractures. Material was examined both before and after an experiment.
C. L. Broadhurst et al.
716
To first order, the interaction of a lattice-vacancy defect and a noble gas atom may be described by the monopole/ induced-dipole interaction: E = -q?aJ2r4, where q, is the charge on the site, LYZ the noble gas polarizability, and r the distance between the monopole and the gas atom. This interaction is much more energetic than the inverse r6 interactions and is stable at high temperature. The energy .E of a Xe atom 3 A from a +2 site is 32.7 kcallmole. The maximum temperature of stability for this interaction can be estimated by dividing E by the Bolztmann constant: E/k = 1.64 X lo4 “K. The actual interaction energy will be lower because the charge on the site will be only partial. In a crystal lattice, defects often occur in pairs. A vacancy would be accompanied perhaps by the nearby substitution of an M 3+ for an MZi cation, and the hole and the cation have a significant coulombic attraction for one another. More importantly, the lattice will adjust in the vicinity of defects. The polarization P of a crystal with N ions (i) in the formula unit can be written as:
where cut,*= Lliai, Er, is the localelectric field, and polarizability is scalar. Further, 01, = ffe + Lyd. The term (ye is the electronic polarizability, which describes the interaction of ion i with an oscillating electric field. The term ad is the displacement polarizability, which describes the induced-dipole moment due to the relative displacement of an ion from its normal lattice position. This displacement is in response to the static electric field generated by the charged vacancy. The term ad is important in the total polarization vah,te for a crystal, and details for silicates can be found in LASAGAand CYGAN ( 1982), and CYGAN and LASAGA(1986). Short-range repulsive energies may also reduce the interaction energy of the noble gas atom with the defect site. The short range energy arises from interactions between electrons in orbitals of nearest neighbor cations and anions in a crystal lattice ( BURNHAM,1985 ). For nearest neighbors, this energy is repulsive. Analogously, as the noble gas electron cloud becomes polarized, one must consider both the attractive and repulsive forces on the positive and negative portions of the dipole. These types of calculations are complex and beyond the present scope; however, the magnitude of this energy is relatively small. It is possible that noble gas atoms can interact with the electric fields produced by lattice defects; thus, the greater polarizabilities of the heavier noble gases would account for their enhanced solubility. Neon is the least polarizable, hence the least soluble; also, a greater fraction is lost at the lowtemperature heating step. Even if noble gases ultimately reside in interstitial sites or planar defects after quenching, vacancy/ induced-dipole interactions nevertheless may be important for the solution and diihtsion of noble gases during our experiments.
Solubility Trend in Melts As noted earlier, the solubilities of noble gases in synthetic melts reported here and by KIRSTEN ( 1968) are lower, in general, than those reported for natural basalts by JAMBON et al. ( 1986) and LIJX ( 1987 ). The greatest differences occur for Ne and Ar. There is relatively little difference in any of the Kr and Xe values plotted in Fig. 2. The lower overall ~lubilities in synthetic melts relative to natural melts can be att~buted to the greater degree of polymerization and Fe3+ concentration and lesser MgO and CaO concentrations in the natural melts ( LUX, 1987; WHITE et al., 1989; BROADHURSTet al., 1990). BROADHURSTet al. ( 1990) concluded that Ar solubilities in melts of greatly differing compositions were better correlated with melt molar volume than melt density. The new melt data reported here also support this conclusion. Figure 3a plots JSr solubility against melt densities calculated at appropriate experimental temperatures for a number of basalts; Fig. 3b shows Kr solubility vs. melt molar volume. Figure 3a shows two distinct groups, whereas 3b shows a positive correlation of solubility with molar volume for all of the data. The major differences between the two groups in Fig. 3a is the presence of ferric iron in melts 6 through 10. Among the natural melts, our data and the data of KIRSTEN ( 1968 ) are more similar to those of HIYAGON and OZIMA ( 1986). Although these latter authors did not give melt analyses, the starting material consisted of basalt (67 wt%) and harzburgite (33 wt%) powders, and experiments were done at oxygen fugacities of lO-9 to lo-” bars, which is below the fayalite/ quartz/magnetite buffer at 1300°C. These conditions produced a MgO-FeO-rich oiivine basalt. The pattern of decreasing dispersion in solubilities with increasing noble gas atomic number is a feature of all the melts in Fig 2 but is most pronounce for our data. FoILowing UHLIG ( 1937) and BLANDERet al. ( 1958), Lux ( 1987) found that the noble gas solubihties in a series of natural melts could be described by the equation Ki = a exp( -br’), where a and b are constants and r is the gas kinetic radius. The constant a was the same for all liquids studied; however, b increased moderately [ 0.869- 1.14 (kO.048 )] from leucite basanite to ugandite. It was also concluded by LUX ( 1987 ), on the basis of relative solubility (i.e., Ki - Kj/Kj for two melts, i and j) that the larger a noble gas atom, the more its solubility will be athected by the properties of the liquid solvent. While this latter conclusion is intuitively satisfying, it is not supported by our data (Table 4). The major technical difference between our study and the others referenced herein is the composition of the atmosphere. We did not use He, nor did we gas-mix with major atmospheric components (i.e., CO*, CO, 02, and Nz). The other authors used gas mixes containing 10-1000 times less Ne than ours. KIRSTEN( 1968) is an exception: His experimental runs were in pure He, Ne, and Ar at pressures up to 0.55 bars, or a mixture containing 0.18 bars of each. KIRSTEN (1968), JAMBONet al. ( 1986), and Lux ( 1987) found that plots of solubihty vs. noble gas partial pressure were linear, with slopes of the lines decreasing with increasing atomic number. in particular, the slopes for He and Ne are much steeper than the plots for the other gases: Small variations in
717
Solubility of noble gases in minerals and melts
yl s
x
60
3b
3a
1
80
9
.9
.
-
E 2 i
8
n8 6
g VI z
40
-
.
5 .
20
1
.
.
6
.4
.
.3
*I n 2 0
I 2.5
I
I
2.6
2.7
density
2.8
I8
20
V(m)
gm/cm*3
26
24
22
28
cmA3/mole
FIG. 3. Krypton solubility plotted against (a) melt density and (b) molar volume. Molar volume shows a positive correlation for all the data, while density does not. Densities calculated from LANGEand CARMICHAEL( 1987) at 1300°C except where noted. Key to figure: (*1 ) molten enstatite at 1500°C from KIRSTEN( 1968) (solubility is K&en’s extrapolation from light noble gas data); (2) DJM 7AN: (3) average DJM 2DI( i); (4) average DJM F05B: (5) average DJM ZDI(ii); (6) tholeiite at 1400°C from JAMBONet al. ( 1986) (note that composition is given as Fe0 = Fe total); (7) ugandite: (8) alkali-olivine basalt; (9) tholeiite; ( 10) leucite-basanite from LUX ( 1987).
the partial pressure will translate to large variations in the solubility. There is no reason to suspect that our experiments exceeded Henrian pressure regimes. Our results indicate, more likely, that there are yet subtleties to the relationship between noble gas solubility and melt structure. Partition Coefficients
Partition coefficients (Table 5) were calculated from the total concentration data for minerals and melts run in the TABLE 4. Comparison of absolute (AKi_j) vs. relative (AKi:/Kl) solubilities. Example melts are as follows: Leucite basantte and ugattdite brat /cet the range of solubilities in basahs from Lux (1987); tholeiite is intermediate. Hlb is the upper limit for olivine basalt from Hiyagon and Ozima (1986). DJM 7AN and DJM 2DI(i) IV are the respective lower limit and representative value from this study.
i&u-bat.:
I =Ieu-has:
i =Wte
gas
AK
AKIK
AK
AK/K
Ne
20.5
0.49
36.1
0.85
At
1.1
0.63
9.4
0.77
Kt
6.3
0.68
6.9
0.74
Xc
2.5
0.7 1
1.4
0.40
i=leu_bas:i=7AN
t = Hlb:
i =lAN
gas
AK
AK/K
AK
Ne
38.6
0.91
10.8
AK/K 0.76
Ar
11
0.90
2.7
0.69
Kr
8.69
0.93
0.26
0.29
Xc
3.17
0.90
0.83
0.7 I
same experiment. These patterns (Fig. 4) show a clear trend of increasing compatibility with increasing noble gas atomic number-a trend also seen by HIYAGONand OZIMA( 1986 ). However, there is a large variation in the absolute mineral/ melt partition coefficients for a given phase pair. Values are frequently greater than unity, which is a direct result of the low melt solubilities and the surprisingly high mineral solubilities. It seems certain that the partition coefficients for the noble gases cannot be described with the level of specificity considered standard in ionic trace and minor element partitioning studies. This is not the first time that noble gas partition coefficients greater than unity have been reported. BATIZAet al. ( 1979) reported noble gas data for primitive and evolved rocks thought to originate from the same magma. Assuming a closed system, they calculated the amount of crystallization involved in forming the different rock types and, thus, the noble gas partition coefficients. Values were near 10-l for Ne, but greater than unity for the heavier gases. The authors were forced to conclude that their system was not closed. Many of the runs in HIYAGONand OZIMA ( 1986) yielded partition coefficients greater than unity, although the pattitioning trends were the usual (increasing with increasing atomic number). These data were discarded by the authors
TABLE
gas
AK
AK/K
AK
AK/K
Ne
6.8
0.32
15.1
0.7 1
At
0.77
0.17
I.7
0.37
Kr
2.1
0.71
0.6
0.33
Xe
-0.16
-0.16
-1.1
-1.1
exper.
5. Noble gas solid/liquid partition coefficients. Values calculated and melt total gas concenwtions within a given expaiment. mineral
melt Ne
DSil Ar
from mineral
KI
Xe
1.3 0.040
0 38 0.31 2.3 2.0 0.63
3.2 3.3 3.3 1.8 0.61
0. I05 2.7
0.19 67
5.9 45
zl
0.37 0.073
0.63 0.33
2 4 0.73
4.0 I 2
N ;
DI RI8684 DI M2.506 122742
ZDI(i) ”
0.013 0.14 0.033
0.26 0.84 0.15
N N
SP 1 l-4877 SPB12099
” ”
0.034 0.088
V V
DI 122742 FO 12243
2DI(ii) FO 5B
VI VI
DI 122742 SP 1 l-4677
ZDI(ii) FO 5B
C. L. Broadhurst et al.
718
Ar
Kr
Xe
Ne
Kr
AT
gases
Xe
gases
FIG. 4. Noble gas partition coefficient trends for (a) experiment IV and (b) experiments V and VI. All partition coefficients calculated from melt DJM 2DI(i) in (a). In (bl. values for DI 122742 calculated from melt DJM 2DItiil: those for FO 12243 and SP I l-4877 from CjM Fd5B. ?hk patterns generally reflect the total gas concentration data: Partition coefficients increase with increasing noble gas atomic number and are greater than unity for at least Xe. The trends for the VI spinels show a downturn in the Kr/Xe ratio, reflecting the same trend in the total gas concentrations. SP B I2099 in IV is the only sample that yields a set of partition coefficients that are all less than unity. Symbofs: (a) DIR 18684 IV, - q --; DI M2506 IV, - q -: DI 122742 IV, - n --; SP 114877 IV, - - 0 - -; SP BI2099 IV, --~---.(b)DII22742V,-~---;D1122742VI.-~---;FO12243V,--~--;SPi14877VI,--‘~--.
on the grounds that these samples contained gas bubbles or inclusions. Inclusions were indeed observed microscopically; however, the trapped gas component (G) defined from the isotopic mixing discussion had an unusual property: “The composition of component G appears to be fractionated from the original test gas favoring the heavier noble gases: Xe/Ar, Kr/Ar ratios are higher and He/Ar, Ne/Ar ratios are lower than those for the original gas. . . . Possibly, lighter noble gases have diffused out of the bubbles and dissolved in the melt more quickly than heavier noble gases have done” (p. 2053). While we by no means wish to criticize the above interpretations, we find it interesting that our results, as well as those of BATIZA et al. (1979) and HIYAGON and OZIMA ( 1986), can be explained by noble gases acting as relatively compatible, fractionated species in mineral/melt systems. In particular, we note that our data set for minerals + melts shows the same type of behavior-elevated concentrations of noble gases with respect to melts alone and a component fractionated in favor of the heavy noble gases. However, we have no evidence for inclusion control of these phenomena, thus no basis to conclude that the composition of gas bubbles is fractionated. The hypotheses that ( 1) noble gas solubility in minerals is controlled by defect siting, and (2) noble gas solubility in melts is directly proportional to molar volume and polymerization leads to the conclusion that noble gases will become even more compatible with increasing pressure, Highpressure data from HIYAGON and OZIMA ( 1986) do tend to show partition coefficients similar to or greater than 1 bar results; however, these experiments are compromised by the use of 10 or 15 wt% lithium borate in the starting material composition. The partial molar volume of BZ03 is highly
variable and must be calculated for any specific melt composition. However, in silicate melts of the basalt-hanburgite type, boron is likely to be coordinated as B03; hence, the partial molar volume will be somewhere in the range 37-47 cm3/mole (EITEL, 1965; PAUL, 1990). This means that the melt molar volume will be significantly greater than a typical mafic basalt. More studies are required before the effects of pressure on noble gas partitioning can be understood. Average minerals from the mantle are likely to contain many more defects than the minerals from this study. Minerals in the mantle exist at high tem~rat~es and are expected to have higher concentrations of trace elements than the relatively pure natural minerals that we used. Since we have shown ( BROADHURSTet al., 1990) that defects probably can account quantitatively for the one-bar gas concentrations in our minerals, it follows that the much smaller noble gas concentrations in the mantle could be reasonably accommodated in defects. IMPLICATIONS FOR THE TERRESTRIAL PLANET ATMOSPHERES
Contrary to the usual a~umptions, the absolute values of our partition coefficients indicate that magmatic degassing cannot be an efficient process. This is especially true for forsterite sample FO 12243, and for the gases Kr and Xe. Furthermore, regardless of the absolute values, these partitioning trends cannot produce the observed noble gas abundance patterns for the terrestrial planet atmospheres. An example calculation for a C3V chondritic source is shown in Fig. 5. At low degrees of partial melting, the liquid becomes strongly enriched in Ne only, contrary to the trends for the planets. At higher degrees of partial melting, the liquid trend would
Solubility of noble gases in minerals and melts
Ne
ICI
Ar
XC
gases FIG, 5. Equilibrium hatch partial melting of a C3V chondrite source and POLLACK,1983). Noble gas concentrations in the
(DONAHUE
liquid shown at several fractions (F) of partial melting. Partition coefficientsare from sampie SP B12099 in Table 5. We assume that the partition coefficient set for a single mineral is sufficient to represent a variety of mantle com~sitions since noble gases are not sensitive to mineral structure or composition in the same manner that ionic species are. Symbols: C3V, - l -; F = 0.05, - * -; F = 0.10,
--[II--;F=O.lS,--A--;F=OSO,--O--.
follow the starting material composition almost exactly. The salient point is that the distinctive Kr/Xe fractionation in the planetary atmospheres is not reproduced. Krypton and Xe do not partition differently enough that magmatic fractionation can account for the “missing Xe.” Our data imply strongly that the terrestrial planet atmospheres did not accumulate from partial melting of a common chondritic source. As a corollary, if volcanic outsung was mostly responsible for the terrestrial ptanet atmospheres, then the primordial volatile abundances followed the pattern Venus > Earth > Mars. This abundance pattern cannot be produced from a common source with our partition coefficients because noble gases will remain at almost constant concentration levels in the melt throughout a reasonable range of degrees of partial melting (BROADHURST,1989). It must be noted that there is no a priori reason demanding that noble gases other than He act as incompatible elements in mantle melting processes. Among the primordial volatiles, only ‘He has been unambiguously determined to be fluxing from the mantle today. A large number of isotopic m~elling studies find that Earth’s upper mantle and parts of the lower mantle were nearly completely degassed prior to at latest 3.8 Ga (BERNATOW~CZand PODOSEK, 1978; HAMANO and OZIMA, 1978; ALLBGREet al., 1987; TURCO~TE and SCHUBERT, 1988; MARTY, 1989; OZIMA and IGARASHI, 1989; ZHANG and ZINDLER, 1989). These studies are based upon the evolution of the radiogenic isotopes in a magmatic degassing scenario. Most important to the arguments are observations of excess “‘Xe in MORB (STAUDACHERand ALLI?GRE,1982; ALL&RE et al 1983; 1987) and estimates of Earth’s K content. Each is also based on the assumption
719
that nobie gases are uniformly in~mpatible in mineral I melt systems. Despite individual approaches, these models concur that early catastrophic degassing was largely responsible for the Earth’s atmosphere, and only a moderate amount of degassing has occurred since. Our results also favor catastrophic degassing, but for somewhat different reasons. The effect of our partitioning behavior on the isotopic models would be a definite increase (i.e., moves closer to the present so degassing takes longer) in the predicted mean ages for outgassing. There is also potential for change in the relative rates of degassing of the different species; this would have to be tested explicitly for a given model. In addition, since noble gases will be fractionated by partial melting events, the use of mean degassing ages to estimate past rates of tectonic activity ( ALL~GRE et al., 1987) is not strictly correct. In a magmatic degassing scenario, our results would require large degrees of partial melting and vigorous convection in the mantle in order to degass at least the MORB source mantle. Such conditions were possibly present early in the planet’s history. We also note that these high fractions of melting might be inherently self-limiting as they are expected to produce MgO-rich komatiite type magmas. As discussed previously, such melts will have low noMe gas solubilities. AZBELand TOLSTIKHIN( 1990) have introduced fmctionation of noble gases into an isotopic model. They have explained the MORB noble gas fractionation trends relative to air by a superposition of initial mineral/melt partitioning followed by partial degassing of the melt. The theoretically derived partition coefficients of AZBEL and TOLSTIKHIN ( 1990) match the trends found in this study and by HIYAG~N and OZIMA ( 1986). A very important point made by AZBEL and TOLSTIKHIN( 1990) is that single-stage partial melting and degassing cannot yield both a high degree of upper mantle degassing and the observed MORB fractionation pattern; a multistage process is required. As has been pointed out to us, noble gases in MORB and OIB basalts are often concentrated in vesicles, micropores, ‘etc., and can be released quantitatively by simple crushing of the samples. These vesicles are to be expected in basalts from near-surface conditions. The equilibrium noble gas solubilities in this study and others are by definition liquid/ vapor partition coefficients which show that noble gases are highly incompatible elements in this system. The solubilities are insignificant in comparison with those of the reactive volatiles ( H20, Cq, H,S, CH4, etc.), and it seems clear that any fluid phase evolved from a melt is capable of stripping that melt of noble gases. In other words, if a melt can migrate to near the surface, it can outgass. In a multistage process, we are concerned less with the present-day release of volatiles at the MOR and hot spots than we are with the initial, largescale melting of whole or part of the mantle to form the hypothesized degassed mantle reservoir. To reiterate, we wish to call attention to the points that during partial melting ( 1) noble gases can be fractionated from one another, and (2) radiogenic noble gas daughters may or may not be significantly fractionated from their parent elements, depending on the individual situation. An additional concern is the growing body of evidence that the Earth’s interior is enriched in solar-type Ne. 2oNe/
720
C. L. Broadhurst et al.
22Ne enrichments have been reported in diamonds, MORB, and Hawaiian basalts (POREDA and DI BROZOLO, 1984; SARDA et al., 1988; MARTY, 1989; HONDAet al., 1991; OZIMA and ZASHU, 1991). Essentially, the mantle appears to have a Ne endmember component that is entirely solar. MARTY ( 1989) finds that the difference between air and mantle Ne is so great that it precludes a common origin for these reservoirs: Air Ne composition could then be explained by a mixture of planetary (59%) and solar (41%) components, which have remained distinct since accretion. OZIMA and ZASHU ( 1991) conclude from diamond samples that primordial Ne in the ancient mantle is of solar origin, while Ar, Kr, and Xe isotopic ratios are virtually identical with the current atmosphere. Interestingly, OZIMA ( 1989) and OZIMA and ZASHU( 199 1) find that noble gases are of trapped origin, including some of the radiogenic components. There is too much gas in the diamonds for a reasonable in situ production. If the present volcanic regime is not the key to the past, then one must seriously consider that accretionary catastrophic processes are largely responsible for the terrestrial planet atmospheres. Primordial volatile abundances among Venus, Earth, and Mars may or may not have been roughly equal. One can appeal to impact degassing to strip Mars of an early denser atmosphere (WALKER, 1986; AHRENSet al., 1989; MELOSHand VICKERY, 1989 ). Cometary impacts are equally appealing as a means of enhancing the Venusian atmosphere relative to Earth (GRINSPOONand LEWIS, 1988). If the impact flux during the latest stages of accretion were very high, then there could have been nearly continual atmospheric gain and loss, which slowed gradually as accretion slowed. With such an impact flux, effective removal of volatiles from the growing planets is necessary since only a small percentage of volatile-rich material is required to account for the present planetary atmospheres ( ANDERSand OWEN, 1977; DREIBUS and W,&NKE,1987). The final atmospheres could represent a mixture of many different types of impactors, or could reflect the influence of only a few large objects. ZAHNLE et al. ( 1990b) have considered the latter case to explain the Xe problem. They start with a population of porous planetesimals growing in solar nebula. At some point, a planetesimal grows large enough that lithostatic pressure squeezes the pores shut, thus trapping the Xe. These larger planetesimals are then accreted, in turn, to a growing terrestrial planet. Loss of gases during the accretion period may have been by hydrodynamic escape in addition to impact degassing. Hydrodynamic escape can result in isotopic fractionation (HUNTEN et al., 1987; SASAKI and NAKAZAWA, 1988; ZAHNLEet al. 1990a,b), thus it has the potential to reconcile some of the differences in primordial/radiogenic isotopic component ratios among the terrestrial planets and between the planets and meteorites. Escape of a solar composition atmosphere is one explanation for the observed difference between air and mantle Ne. Alternatively, the present atmosphere may also have formed from degassing of a lateaccreting veneer, while the interior is composed of solar composition material. MUSSELWHITEet al. ( 199 1) also provide experimental partitioning data that ultimately supports an accretionary steam atmosphere on Mars. The authors have addressed the
high ‘29Xe/‘32Xe ratio in the Martian atmosphere as compared to SNCs and Earth reservoirs. Partitioning experiments showed that I and Xe are not fractionated from each other during magmatic processes. Instead, it is proposed that both I and Xe were outgassed by 4.45 Ga, or before significant lz91 decay. The highly water-soluble I was retained in the oceans and hydrated crust, while Xe was removed by atmospheric erosion. ‘29llater decayed and was released to the now-present atmosphere. CONCLUSIONS The solubility trend of Ne, Ar, Kr, and Xe in anorthite, diopside, forsterite, and spine1 increases with increasing noble gas atomic number. Different samples of a particular mineral composition have characteristic solubilities, some of which are surprisingly high. Noble gases are likely to be sited in lattice-vacancy defects. In contrast, synthetic silicate melt solubilities decrease with increasing atomic number. Solubility values are on the order of or lower than those in minerals. Our data tend to overlap the low end of the range defined for natural basaltic melts. The difference between Ne and Xe solubility values in our results is not as extreme as that in the natural melts. Noble gas solubilities correlate well with molar volume but poorly with density. A trend of increasing partition coefficient values with increasing atomic number is seen for all mineral/melt pairs. This trend indicates that is not possible for the noble gas abundances in the atmospheres of the terrestrial planets to have been derived from partial melting of a common chondritic source. Furthermore, the partition coefficients determined in this study rule out significant fractionation of Kr relative to Xe in magmatic processes; thus, the “missing Xe” in the Earth’s atmosphere cannot be attributed to partial melting. We call attention to the caveats that during partial melting, ( 1) noble gases can be fractionated from one another, and (2) radiogenic noble gas daughters may or may not be significantly fractionated from their parent elements, depending on the individual situation. These noble gas partition coefficients support a number of other studies, which conclude that early, catastrophic degassing was largely responsible for Earth’s atmosphere, and that degassing is a multistage process. Catastrophic degassing would have to involve either large amounts of partial melting and vigorous mantle convection or accretionary processes, such as impact accumulation/erosion and hydrodynamic escape. Acknowledgments-This research was supported by NSF grant EAR 85-18740 to M. J. Drake and NASA Traineeship NGT-50051 to C. L. Broadhurst. We thank the various reviewers and F. Frey for constructive criticisms. CLB also acknowledges support from NAS/ NRC and NASA GSFC for production of the manuscript. Editorial handling: F. A. Frey
REFERENCES AHRENsT. J., O’KEEFEJ. D., and LANGEM. A. ( 1989) Formation of atmospheres during accretion of the terrestrial planets. In Origin and Evolution of Planetary and Satellite Atmospheres (ed. S. K.
ATREYAet al.) pp. 328-385. Univ. Arizona Press.
Solubility of noble gases in minerals and melts ALL~GREC. J., STAUDACHERT., SARDAP., and KURZ M. ( 1983) Constraints on evolution of Earth’s mantle from rare gas systematics. Nature 303, 762-766. ALLBGREC. J., STAUDACHERT., and SARDAP. ( 1987) Rare gas systematin: Formation of the atmosphere, evolution and structure of the Earth’s mantle. Earth Planet. Sci. Lett. 81, 127-150. AMARIS. and OZIMAM. ( 1988) Extra-terrestrial noble gases in deep sea sediments. Geochim. Cosmochim. Acta 52, 1087-1095. ANDERSE. and OWENT. ( 1977) Mars and Earth: Origin and abundance of volatiles. Science 198, 453-465. AZBEL1. Y. and TOLSTIKHINN. ( 1990) Geodynamics, magmatism, and degassing of the Earth. Geochim. Cosmochim. Acta 54, 139154. BATIZAR., BERNATOWICZ T. J., HOHENBERGC. M., and PODOSEK F. A. ( 1979) Relations of noble gas abundances to petrogenesis and magmatic evolution of some oceanic basalts and related differentiated volcanic rocks. Contrib. Mineral. Petrol. 69, 30 1-3 14. BERNATOWICZ T. J. and PODOSEKF. A. ( 1978) Nuclear components in the atmosphere. In Terrestrial Rare Gases (ed. E. C. ALEXANDER.JR., and M. OZIMA)pp. 99-135. Japan Scientific Society Press. BERNATOWICZ T. J. and PODOSEKF. A. (1986) Adsorption and isotopic fractionation of Xe. Geochim. Cosmochim. Ada SO, 15031507. BERNATOWICZ T. J., PODOSEKF. A., HONDA M., and KRAMER F. E. ( 1984) The atmospheric inventory of Xenon and noble gases in shales: The plastic bag experiment. J. Geophys. Res. 89,45974611. BERNATOWICZ T. J., KENNEDY B. M., and POD~SEKF. A. ( 1985) Xe in glacial ice and the atmospheric inventory of noble gases. Geochim. Cosmochim. Acta 49, 2561-2564. BLANDERM., GRIMES W. R., SMITH N. V., and WATSON G. M. ( 1958) Solubility of the noble gases in molten fluorides. II. In the LiF-NaF-KF eutectic mixture. J. Phys. Chem. 63, 1164-l 167. BROADHURST C. L. ( 1989) Solubility and Partitioning of Noble Gases in Anorthite, Diopside, Forsterite, Spine], and Synthetic Basaltic Melts: Implications for the Origin and Evolution of Terrestrial Planet Atmospheres. Ph.D. thesis, Univ. Arizona. BROADHURST C. L., DRAKEM. J., HAGEEB. E., and BERNATOWICZ T. J. ( 1990) Solubility and partitioning of Ar in anorthite, diopside, forsterite, spine], and synthetic basaltic liquids. Geochim. Cosmochim. Acta 54, 299-309. BURNHAMC. W. ( 1985) Mineral structure energetics and modeling using the ionic approach. In Microscopic to Macroscopic; Reviews in Mineralogy. pp. 347-388. Min. Sot. Amer. CARROLLM. R. and STOLPERE. M. ( 1991) Argon solubility and diffusion in silica glass: Implications for the solution behavior of molecular gases. Geochim. Cosmochim. Acta 55, 211-225. CYGAN R. T. and LASAGAA. C. ( 1986) Dielectric and polarization behavior of forsterite at elevated temperatures. Amer. Mineral. 71, 758-766. DREIBUS,G. and WANKE H. (1987) Volatiles on Earth and Mars: A comparison. Icarus 71, 224-240. DONAHUET. M. and POLLACKJ. B. (1983) Origin and evolution of the atmosphere of Venus. In Venus (ed. D. M. HUNTENet al.) pp. 1003-1036. Univ. Arizona Press. EITELW. ( 1965) Silicate Science. Academic Press. FOLANDK. A. ( 1974) A?‘diffusion in homogeneous orthoclase and an interpretation of Ar diffusion in K-feldspars. Geochim. Cosmochim. Acta 38, 151-166. GRINSFQOND. H. and LEWISJ. S. ( 1988) Cometary water on Venus: Implications of stochastic impacts. Icarus 74, 21-35. HAMANO Y. and OZIMA M. (1978) Earth-atmosuhere evolution model based on Ar isotopic data. in Terrestrial kare Gases (ed. E. C. ALEXANDER,JR ., and M. OZIMA) pp. 155- 17 1. Japan Scientific Society Press. HAYATSUA. and WABOSOC. E. ( I985 ) The solubility of rare gases in silicate melts and implications for K-Ar dating. Chem. Geol. 52, 97-102. HIYAGONH. and OZIMAM. ( 1986) Partition of noble gases between olivine and basalt melt. Geochim. Cosmochim. Acta SO, 20452057.
721
HONDAM., MCDOUGALLI., PATTERSOND. B., DOULGERISA., and CLAGUED. A. ( 1991) Possible solar noble gas component in Hawaiian basalts. Nature 349, 149- I5 1. HUNTEN D. M., PEPIN R. 0.. and WALKERJ. C. G. (1987) Mass fractionation in hydrodynamic escape. Icarus 69, 532-549. JAMBONA., WEBERH., and BRAUN0. ( 1986) Solubility of He, Ne, Ar, Kr, and Xe in a basalt melt in the range 1250-1600°C. Geochemical implications. Geochim. Cosmochim. Acfa SO, 401-408. KIRSTENT. ( 1968) Incorporation of rare gases in solidifying enstatite melts. J. Geophys. Res. 73, 2807-2810. LANGER. A. and AHRENST. J. ( 1982) The evolution of an impactgenerated atmosphere. Icarus S&96-120. LANGER. A. and CARMICHAELI. S. E. ( 1987) Densities of Na,OK~0-Ca0-Mg0-Fe0-Fe203-Al~O~-TiO~-SiO~’ liquids: New measurements and derived partial molar properties. Geochim. Cosmochim. Acta 51,2931-2946. LASAGAA. C. and CYGANR. T. ( 1982) Electronic and ionic nolarizability of silicate minerals. Amer. Mineral. 67, 328-324. . Lux G. ( 1987) The behavior of noble gases in silicate liquids: Solution, diffusion, bubbles, and surface effects, with applications to natural samples. Geochim. Cosmochim. Acta 51, 1549- 1560. MALVIND. J. and DRAKEM. J. ( 1987) Experimental determination of crystal/melt partitioning of Ga and Ge in the system forsteriteanorthite-diopside. Geochim. Cosmochim. Acta 51, 2 117-2 128. MARTYB. ( 1989) Neon and xenon isotopes in MORB: Implications for the earth-atmosphere evolution. Earth Planet. Sci. Lett. 94, 45-56. MATSUDAJ. and MATSUBARAK. (1989) Noble gases in silica and their implication for the terrestrial “missing” Xe. Geophys. Res. Lett. 16, 8 l-84. MA~SUDAJ. and NAGAO K. ( 1989) Noble gas emplacement in shockproduced diamonds. Geochim. Cosmochim. Acta 53, 1I 17- 1 I2 1, MATSUDAJ., MATSUBARAK., YAJIMA H., and YAMAMOTOK. ( 1989) Anomalous Ne enrichment in obsidians and Darwin glass: Diffusion of noble gases in silica-rich glasses. Geochim. Cosmochim. Acta 53,3025-3033. MATSLJIT. and ABE Y. (1986a) Evolution of an impact-induced atmosphere and magma ocean on the accreting Earth. Nature319, 303-305. MATSUIT. and ABE Y. ( 1986b) Impact-induced oceans on Earth and Venus. Nature 322, 526-528. MAZOR E., HEYMANND., and ANDERSE. (1970) Noble gases in carbonaceous chondrites. Geochim. Cosmochim. Acta 34, 78 l824. MELOSHH. J. and VICKERYA. M. ( 1989) Impact erosion of the primordial Martian atmosphere. Nature 338, 487-489. MUSSELWHITED. M., DRAKE M. J., and SWINDLET. D. ( 1991) Early outgassing of Mars supported by differential water solubility of iodine and xenon. Nature 352, 697-699. OZIMAM. ( 1989) Gases in diamonds. Ann. Rev. Earth Planet. Sci. 17,361-384. OZIMA M. and IGARASHIG. (1989) Terrestrial noble gases: Constraints and implications on atmospheric evolution. In Origin and Evolution ofPlanetary and Satellite Atmospheres (ed. S. K. AYEDA et al.) pp. 306-327. Univ. Arizona Press. OZIMA M. and PODOSEKF. A. ( 1983) Noble Gas Geochemistry. Cambridge Univ. Press. OZIMAM. and ZASHUS. ( 1991) Noble gas state of the ancient mantle as deduced from noble gases in coated diamonds. Earth Planet. Sci. Lett. 105, 13-27. PAUL A. ( 1990) Chemistry ofGlasses. Chapman and Hall. POREDAR. and DI BROZOLOF. R. ( 1984) Neon isotope variations in Mid-Atlantic Ridge basalts. Earth Planet. Sci. Lett. 69, 277289. PRESNALLD. C., DIXON S. A., DIXON J. R.. O’DONNELLT. H., BRENNERN. L., SHROCKR. L., and DYCUSD. W. ( 1978) Liquidus phase relations on the join diopside-forsterite-anorthite from I atm to 20 kb: Their bearing on the generation and crystallization of basaltic magma. Contrib. Mineral. Petrol. 66, 203-220. SARDAP., STAUDACHER T., and ALLBGREC. J. ( 1988) Neon isotopes in submarine basalts. Earth Planet. Sci. Left. 91, 73-88.
722
C. L. Broadhurst et al.
SASAK~S. and NAKAZAWA K. ( 1988) Origin of isotopicfractionation of terrestrial Xe: Hydrodynamic fractionation during escape of the primordial HZ-He atmosphere. Earth Planet. Sci. Lett. 89, 323334. ZHELHAAS N., OTT U., and BECEMANNF. ( 1990) Trapped noble gases in unequilbrated ordinary chondrites. Geochim. Cosmochim. Acta 54,2869-2882. STAUDACHERT. and ALLEGREC. J. ( 1982) Terrestrial xenology. Earth Planet. Sci. Lett. 60, 389-406. SWINDLE T. D., GARRISON D. H., GOSWAMIJ. N., HOHENBERG C. M., NICHOLSR. N., and OLINGERC. T. ( 1990) Noble gases in the howardites Bholghati and Kapoeta. Geochim. C~smoc~~rn.ilcfa 54,2183-2194. TURCOTTED. L. and SCHUBERTG. ( 1988) Tectonic implications of radiogenic noble gases in planetary atmospheres. Icarus 74,3646. UHLIG H. H. ( 1937) The solubilities of gases and surface tension. J. Phys. Chem. 41, 1215-1225. WACK~R J. F. ( 1989) Laboratory simulation of meteoritic noble gases. III. Sorption of neon, argon, krypton, and xenon on carbon: Elemental fractionation. Geochim. Cosmochim. Acta 53, 14211434. WACKERJ. F., ZADNICKM. G., and ANDER~E. ( 1985) Laboratory simulation of meteoritic noble gases. I. Sorption of xenon on carbon: Trapping experiments. Geochim. Cosm~him. Acta 49,10351048. WALKERJ. C. G. ( 1986) Impact erosion of planetary atmospheres. Icarus 48, 87-98. WATSONA. J., DONAHUET. M., and WALKERJ. C. G. ( 1981) The dynamics of a rapidly escaping atmosphere: Applications to the evolution of Earth and Venus. Icarus 48, 150- 166. WHITE B., BREARLYM., and MONTANAA. ( 1989) Solubility of Ar in silicate liquids at high pressures. Amer. Mineral. 74, 5 13-529. W~ELERR., GRAF TH., PEDRONIA., SIGNERP., PELLASP., FIENI C., SUTERM., Vffi~ S., CLAYTONR. N., and LAULJ. C. ( 1989) Exposure history of the regolithic chondrite Fayetteville: II. Solargas-free light inclusions. Geochim. Cosmochim. Acta 53, 14491459. WIENS R. C. (1988) On the siting of gases shock-emplaced from internal cavities in basalt. Geochim. Cosmochim. Acra 52, 27752783. WIENSR. C. and PEPINR. 0. f 1988) Laboratory shock empla~ment of noble gases, nitrogen, and carbon dioxide into basalt, and implications for trapped gases in shergottite EETA 7900 1.Geochim. Cosmoehim. Acta 52,295-307. YANGJ. and ANDERSE. ( 1982a) Sorption of noble gases by solids, with reference to meteorites. II. Chromite and carbon. Geochim. Cosmochim. Acta 46, 861-875. YANGJ. and ANDERSE. ( 1982b) Sorption of noble gases by solids, with reference to meteorites. III. Sulfides, spinels, and other substances; on the origin of planetary gases. Geochim. Cosmochim. Acta 46,877-892. YANG J., LEWISR. S., and ANDERSE. ( 1982) Sorption of noble gases by solids, with reference to meteorites. I. Magnetite and carbon. Geochim. Cosmochim. Acfa 46.84 l-860. ZAHNLEK. J. and KASTINGJ. F. ( 1986) Mass fractionation during transonic escape and impIications for loss of water from Mars and Venus. Icarus (i&462-480. ZAHNLEK. J., KASTINGJ. F., and POI.LACKJ. B. (1988) Evolution of a steam atmosphere during Earth’s accretion. fcarzcs 74,62-97. ZAHNLEK., BLASTING J. F., and POLLACKJ. B. ( 199Oa) Mass fractionation of noble gases in diffusion-limited hydrodynamic hydrogen escape. Icarus 84,62-97. ZAHNLE K., POLLACKJ. B., and KASTINGJ. F. (1990b) Xenon fractionation in porous planetesimals. &whim. Cosmochim. Acfa 54,2577-2586. Z~;HRINGER J. ( 1968) Rare gases in stony meteorites. Geochim. Cosmochim. Acta 32,209-237. ZHANG Y. and ZINDLERA. ( 1989) Noble gas constraints on the evolution of the Earth’s atmosphere. J. Geoph~ys.Rex 94, 13.7 1913,737.
APPENDIX
I
Figure A-l shows the patterns for some of the melts that inadvertently grew crystals. These results allowed independent confirmation of the trends observed in the minerals and melts, as well as an understanding of the Ar analytical problem. The particular crucible used for the last set of analyses had been used previously for SiOz samples containing Ar only, Those experiments were unrelated to this work. It appears that the SiOz was incompletety fused during these analyses, so some Ar was retained and degas& into our runs. The Ar did not show up in the blank runs; although the crucible was heated to 15.5O’Cduring blank runs, the dross in the crucible had aheady been degassed at this temperature, so it did not degas further until a sample was dropped. The silicate sample can sometimes act as a flux for the refractory residue material in the crucible and vice versa, as discussed in the text for spinel. It is difficult to attribute the Ar spike to an error in the gas composition for the following reasons:
1) The reported data for DJM 2DI(i) IV is the second of two aliquants. The first run had such a high Ar abundance that the sample could not be analyzed properly and had to be discarded. A leak in the system was suspected at the time, but none was present. 2) The Ar spike is a persistent feature but variable in relative height depending on the overall gas content of the sampies. For example, the peak is hardly noticeable in FO 12243 but very apparent in SP 1l-4877. 3) As shown in Fig. A-l, another sample of DJM ZDI( i) IV containing both melt and crystals was analyzed along with this group so that we could better understand some of the previous-melt results. Since DJM 2DI ( i ), IV samnles bear no relation to the V and VI samples other than the fact that they were analyzed in the same group; this shows that the changes to the furnace and the new tank of gas are not responsible for the Ar spike. 4) If the gas were, for sake of argument, 99% Ar, this would hardly change the Ar solubility values but would lead to astronomical solubilities for the other gases. We would also be obliged to make the same blanket corrections to the minerals run in these experiments, an adjustment for which we have no basis. the solubility behavior of noble gases in melts has already been established and shown to obey Henry’s Law at low pressures. It is unlikely that minerals show the same type of sensitivity to the &ascomposition.
Ne
A,
Iir
I
Xi?
gases FIG. A-l. Solubilities in Melts + Crystals. Symbols: 3AN(a) I, - n --; 3AN(b) 1, - q --; 3AN(c) II, - q -; DM-842 II, -*-O-e-; IDIII, **m****;ZDI(i)IV, *..O.*..
Solubility of noble gases in minerals and melts TABLE
A-L
Analysis of variance for Ne. Kr, and Xe solubiliries in mineral and melt
723 APPENDIX 2
samples for which multiple analyses exist. Units are ccSTP/gm-arm x 10-S.
Reproducibility of Results Total variance
Taking into consideration the data presented here and in BROADet al. ( 1990), it appears that variations of at least a factor of two are inherent in our experimental procedure. If we first consider Ar measurements for the anorthite samples, the Ar solubility in a given sample cannot always be reproduced better than this. AN WU 6 I7 has a total of eight measurements, averaging 1.1 + 0.5 X 10m5 ccSTP/gm-atm. This is not significantly different from the averages ofAN 1l-7779 (1.3 X 10e5) and AN 137041 (0.87 X 10e5) replicates. However, all these values are much lower than the average value for AN HO 228 (9.4 X 10-5). Although our data set is limited with regard to statistical analysis, there are five samples that have three replicates in the mixed noble gas experiments. Table A-l presents a simple analysis of variance for these samples and an additional melt. The purpose of this table is to test the hypotheses that ( 1) the variation between replicates is as great as the variation in samples, and (2) the result that the mineral solubilities are on the order of or greater than the melt solubilities is due to experimental error or uncertainty. While it is acknowledged that the total variance (SST ) is very large in this system, the results show that this is not due to differences among mineral structures, as would be the case for ionic trace elements Furthermore, the calculations show that the variance among replicates is much less than the variance among samples. Table A- 1 identifies DI 122742 as having consistently higher solubilities than the other diopsides and spinels, similar to AN HO 228 and FO 12243. The inclusion of melts in Table A- 1 is not critical to the conclusions. For example, Kr results for the minerals DI 122742, DI R18684, and SP 1 l-4877 gives SST = 393, SS, = 209, and SSw = 184. Adding in a column with two analyses of AN WU 6 17 and one of AN 1l7779 (whose solubilities cannot be distinguished) gives SSr = 184, SSA = 105, and SSw = 79. Therefore, we conclude that hypotheses ( 1) and (2) are likely to be false. HURST
Variance within sample replicates (SS,)
ssw = SST
ss,$
where n = number of samples m = number of replicates N=n*m
Ne
R18684
122742
F05B*
11.4877
12243
ZDI
I :
0.018 0.079 0.14
0.80 0.4 2.1
4.5 2.7 2.8
0.66 0.56 2.0
1.6 7.8 6.5
5.9 5.6 3.9
x
0.079
1.1
5.0
1.07
7.3
5.1
SST = 145 Kr
SSA = 127
SSW = 19
1 2
3
0.80 1.3 0.91
5.4 15.4 673
0.97 1.3 3.3
3.0 4.8 2.4
65.2 32.6 58.7
2.4 2.6 2.6
T7
1.06
9.03
1.9
3.4
52.2
2.5
SST = 6690
Xe
I
1 2 3
SSw = 663
SSA = M)27
2.1 4.5 6.8
7.1 125 9.6
0.33 I.1 3.0
4.6 3.9 3.6
143 35.4 96.4
2.1 2.6 2.4